Triplett and K. D. McMahon - CAE...
15
Environmental Microbiology (2006) 8(6), 956–970 doi:10.1111/j.1462-2920.2005.00979.x © 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Triplett and K. D. McMahon Received 29 July, 2005; accepted 4 November, 2005. *For correspon- dence. E-mail [email protected]; Tel. (+1) 608 263 3137; Fax (+1) 608 262 5199. † Present address: Department of Natural Resources and Environmental Sciences, University of Illinois at Urbana-Champaign, Urbana, IL 61801, USA. Microbial community dynamics in a humic lake: differential persistence of common freshwater phylotypes Ryan J. Newton, 1 Angela D. Kent, 2,3† Eric W. Triplett 4 and Katherine D. McMahon 1,2 * 1 Microbiology Doctoral Training Program, University of Wisconsin-Madison, Madison, WI 53706, USA. 2 Department of Civil and Environmental Engineering, University of Wisconsin-Madison, Madison, WI 53706, USA. 3 Center for Limnology, University of Wisconsin-Madison, Madison, WI 53706, USA. 4 Department of Microbiology and Cell Science, University of Florida, Gainesville, FL 32611-0700, USA. Summary In an effort to better understand the factors contrib- uting to patterns in freshwater bacterioplankton community composition and diversity, we coupled automated ribosomal intergenic spacer analysis (ARISA) to analysis of 16S ribosomal RNA (rRNA) gene sequences to follow the persistence patterns of 46 individual phylotypes over 3 years in Crystal Bog Lake. Additionally, we sought to identify linkages between the observed phylotype variations and known chemical and biological drivers. Sequencing of 16S rRNA genes obtained from the water column indicated the presence of phylotypes associated with the Actinobacteria , Bacteroidetes , Firmicutes , Proteo- bacteria , TM7 and Verrucomicrobia phyla, as well as phylotypes with unknown affiliation. Employment of the 16S rRNA gene/ARISA method revealed that spe- cific phylotypes varied independently of the entire bacterial community dynamics. Actinobacteria , which were present on greater than 95% of sampling dates, did not share the large temporal variability of the other identified phyla. Examination of phylotype rela- tive abundance patterns (inferred using ARISA frag- ment relative fluorescence) revealed a strong correlation between the dominant phytoplankton suc- cession and the relative abundance patterns of the majority of individual phylotypes. Further analysis revealed covariation among unique phylotypes, which formed several distinct bacterial assemblages corre- lated with particular phytoplankton communities. These data indicate the existence of unique persis- tence patterns for different common freshwater phylotypes, which may be linked to the presence of dominant phytoplankton species. Introduction Bacterioplankton communities are integrally involved in the biogeochemical processes underpinning freshwater ecosystems (Cotner and Biddanda, 2002). In humic lakes, which receive an exceptionally large input of allochtho- nous (terrestrially derived) organic matter, bacterioplank- ton play a critical role in determining the flux of nutrients between the terrestrial and aquatic environment. Humic lakes are further characterized by moderate phytoplank- ton productivity, but high bacterial metabolism (Wetzel, 2001). In addition, these lakes contain a simplified food web due to a lack of planktivorous fish. The combination of continuous nutrient input and a general lack of higher trophic levels result in a system dominated by microbial activity. Significant research efforts have focused on the contribution of bacterial communities as single entities to ecosystem functions in these lakes (Wetzel, 2001); how- ever, the factors influencing bacterial community compo- sition (BCC), and in particular the dynamics of individual community members, remain relatively unknown. Recently, several studies have examined BCC in the epilimnion of freshwater lakes and reservoirs. As a result of these studies, a core group of bacterial phylotypes common to freshwater has emerged (Zwart et al ., 2002). Of the ubiquitous bacterial phylotypes, the acI clade of Actinobacteria and the Beta I and II clades of Betapro- teobacteria are generally reported as numerically domi- nant (Hahn, 2003; Warnecke et al ., 2004; Simek et al ., 2005), with a single clade often comprising 30–50% of the total bacterial cells in the water column. In addition, evidence suggests BCC is significantly impacted by brief intense protistan grazing periods (Kent et al ., 2004; Pernthaler et al ., 2004; Simek et al ., 2005) and may be influenced by the composition and accessibility of
Transcript of Triplett and K. D. McMahon - CAE...
Environmental Microbiology (2006)
8
(6) 956ndash970 doi101111j1462-2920200500979x
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
gy y pp gy g g y y
Triplett and K D McMahon
Received 29 July 2005 accepted 4 November 2005 For correspon-dence E-mail tmcmahonengrwiscedu Tel (+1) 608 263 3137Fax (+1) 608 262 5199
dagger
Present address Department of NaturalResources and Environmental Sciences University of Illinois atUrbana-Champaign Urbana IL 61801 USA
Microbial community dynamics in a humic lake differential persistence of common freshwater phylotypes
Ryan J Newton
1
Angela D Kent
23dagger
Eric W Triplett
4
and Katherine D McMahon
12
1
Microbiology Doctoral Training Program University of Wisconsin-Madison Madison WI 53706 USA
2
Department of Civil and Environmental Engineering University of Wisconsin-Madison Madison WI 53706 USA
3
Center for Limnology University of Wisconsin-Madison Madison WI 53706 USA
4
Department of Microbiology and Cell Science University of Florida Gainesville FL 32611-0700 USA
Summary
In an effort to better understand the factors contrib-uting to patterns in freshwater bacterioplanktoncommunity composition and diversity we coupledautomated ribosomal intergenic spacer analysis(ARISA) to analysis of 16S ribosomal RNA (rRNA)gene sequences to follow the persistence patterns of46 individual phylotypes over 3 years in Crystal BogLake Additionally we sought to identify linkagesbetween the observed phylotype variations andknown chemical and biological drivers Sequencingof 16S rRNA genes obtained from the water columnindicated the presence of phylotypes associated withthe
Actinobacteria
Bacteroidetes
Firmicutes
Proteo-bacteria
TM7
and
Verrucomicrobia
phyla as well asphylotypes with unknown affiliation Employment ofthe 16S rRNA geneARISA method revealed that spe-cific phylotypes varied independently of the entirebacterial community dynamics
Actinobacteria
whichwere present on greater than 95 of sampling datesdid not share the large temporal variability of theother identified phyla Examination of phylotype rela-tive abundance patterns (inferred using ARISA frag-ment relative fluorescence) revealed a strongcorrelation between the dominant phytoplankton suc-cession and the relative abundance patterns of the
majority of individual phylotypes Further analysisrevealed covariation among unique phylotypes whichformed several distinct bacterial assemblages corre-lated with particular phytoplankton communitiesThese data indicate the existence of unique persis-tence patterns for different common freshwaterphylotypes which may be linked to the presence ofdominant phytoplankton species
Introduction
Bacterioplankton communities are integrally involved inthe biogeochemical processes underpinning freshwaterecosystems (Cotner and Biddanda 2002) In humic lakeswhich receive an exceptionally large input of allochtho-nous (terrestrially derived) organic matter bacterioplank-ton play a critical role in determining the flux of nutrientsbetween the terrestrial and aquatic environment Humiclakes are further characterized by moderate phytoplank-ton productivity but high bacterial metabolism (Wetzel2001) In addition these lakes contain a simplified foodweb due to a lack of planktivorous fish The combinationof continuous nutrient input and a general lack of highertrophic levels result in a system dominated by microbialactivity Significant research efforts have focused on thecontribution of bacterial communities as single entities toecosystem functions in these lakes (Wetzel 2001) how-ever the factors influencing bacterial community compo-sition (BCC) and in particular the dynamics of individualcommunity members remain relatively unknown
Recently several studies have examined BCC in theepilimnion of freshwater lakes and reservoirs As a resultof these studies a core group of bacterial phylotypescommon to freshwater has emerged (Zwart
et al
2002)Of the ubiquitous bacterial phylotypes the acI clade of
Actinobacteria
and the
Beta
I and II clades of
Betapro-teobacteria
are generally reported as numerically domi-nant (Hahn 2003 Warnecke
et al
2004 Simek
et al
2005) with a single clade often comprising 30ndash50 ofthe total bacterial cells in the water column In additionevidence suggests BCC is significantly impacted by briefintense protistan grazing periods (Kent
et al
2004Pernthaler
et al
2004 Simek
et al
2005) and maybe influenced by the composition and accessibility of
Freshwater bacterial community dynamics
957
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
autochthonous nutrients made available during phy-toplankton blooms in both freshwater (Eiler and Bertils-son 2004) and marine systems (Pinhassi
et al
2004Rooney-Varga
et al
2005)Although the existence of common epilimnetic freshwa-
ter bacterial phylotypes is apparent we do not yet under-stand the extent of variation and associated ecologicaldrivers of change in these dominant groups or of otherfreshwater bacterial community members over timescales of seasons or years Yannarell and colleagues(2003) showed that lakes experience quite dramaticchanges in BCC within- and between-years regardless oflake trophic status Similarly several other researchershave shown that BCC varies greatly between lake types(Methe and Zehr 1999 Lindstrom 2000 Yannarell andTriplett 2004) However in none of these studies weretemporally persistent and variable organisms identified Inan attempt to reconcile the notion that particularly com-mon freshwater taxa exist with the observation that overallBCC is highly variable we set out to identify and distin-guish between populations that are characteristically vari-able and those that tend to recur over seasonal andannual time scales
A focus on identifying dynamic and persistent bacterialpopulations should lead to a greater understanding offactors influencing bacterial community structure inlakes In addition an examination of factors that may beregulating not only the microbial community dynamicsbut the dynamics of individual members of that commu-nity will lead to an increased understanding of potentialfreshwater microbial-mediated processes linked tospecific organisms Therefore we designed a multiyearstudy of microbial populations in a single lake with asampling frequency designed to capture the pace ofchange in BCC in an effort to provide insight into theecology of both common and transient freshwaterbacteria
An initial study of Crystal Bog Lake revealed annualbut dynamic BCC patterns and a significant correlationbetween bacterial phylotype richness and eukaryoticplankton succession (Kent
et al
2004) However theidentity of individual phylotypes making up the bacteri-oplankton community and their population dynamics werenot described The aim of this study was to investigatepatterns in humic freshwater BCC and diversity at a fine-scale taxonomic level over 3 years to identify the persis-tent and dynamic bacterial community members and toexplore chemical and biological factors influencing individ-ual community member dynamics This was accomplishedby coupling automated ribosomal intergenic spacer anal-ysis (ARISA) fingerprinting to 16S ribosomal RNA (rRNA)gene clone library analysis (Brown
et al
2005) to exam-ine BCC at multiple levels of taxonomic resolution duringa long-term and intense sampling effort
Results
Bacterial community composition
Clone library analysis of Crystal Bog Lake identifiedrepresentatives of six bacterial phyla (
Actinobacteria
Bacteroidetes
Firmicutes
Proteobacteria
TM7
and
Ver-rucomicrobia
) including members of the classes
Alpha-
Beta-
Delta-
and
Gammaproteobacteria
and threeclones with unclassified phylogenetic affiliation (Figs 1ndash4)The
Betaproteobacteria
class contained the most repre-sentatives at each assigned operational taxonomic unit(OTU) (Table 1) while only one representative at eachOTU assignment was detected for the
Deltaproteobacte-ria
class and
Firmicutes
and
TM7
phyla (Table 1)Thirteen freshwater-specific clades (Glockner
et al
2000 Zwart
et al
2002 Warnecke
et al
2004) wereidentified in Crystal Bog Lake which included membersof the
Actinobacteria
Bacteroidetes
Alpha-
and
Betapro-teobacteria
and
Verrucomicrobia
All phylotypes classi-fied in the
Actinobacteria
phylum were members of thepreviously described freshwater-specific acI-B or soil andfreshwater-specific soil IIndashIII clade (Fig 1) Sequencesaffiliated with the soil IIndashIII clade clustered with previouslydefined peat bog clones Many of the Crystal Bog Lake
Betaproteobacteria
16S rRNA gene sequences wereaffiliated with the freshwater clades Beta I II III and IV(Fig 2) The remaining
Betaproteobacteria
clones wereclosely related to
Janthinobacterium
Ralstonia
or
Burkholderia
-type bacteria The majority of
Bacteroidetes
-related clones were members of the previously definedfreshwater clades CF I and CF III (Glockner
et al
2000)All of the
Alphaproteobacteria
clones were identified asbelonging to the freshwater clades Alpha I II III and IVand most of the
Verrucomicrobia
clones belonged to theFukuN18 freshwater clade (Figs 3 and 4) The remaining
Table 1
Number of OTU assignments by phylum
Phylum Clade
a
Species
b
AFL
c
Actinobacteria
2 7 13
Bacteroidetes
5 11 12
Firmicutes
1 1 1
Proteobacteria
18 34 52
Alpha
4 9 8
Beta
7 14 25
Delta
1 1 1
Gamma
6 10 18
TM7
1 1 1
Verrucomicrobia
2 2 6Unknowns 3 3 3Total 32 59 88
a
Clades were determined by the branching patterns obtained follow-ing phylogenetic tree construction and have sequence identity = 90
b
Species are defined as 16S rRNA sequence groups sharing = 97gene identity
c
AFL
=
ARISA fragment length
958
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
identified clones were not affiliated with any previouslydescribed freshwater clades Following alignment andtree inference three singleton clones Crystal Bog 2E1Crystal Bog 2KA12 and Crystal Bog 021B9 did not clusterwith any previously defined phyla (Fig 4) Clone CrystalBog 2E1 was most closely affiliated with 16S rRNA genesequences from the
TM6
and
TM7
phyla (Fig 4) ClonesCrystal Bog 2KA12 and Crystal Bog 021B9 were mostclosely affiliated with each other and are loosely affiliatedwith members of the
Verrucomicrobia
phylum (Fig 4)
Clone library analysis
Four Crystal Bog clone libraries produced 289 16S rRNAgene sequences and their corresponding ARISA fragment
lengths (AFLs measured as the number of nucleotidesamplified with primers 1406F and 23SR) The coverageof the largest clone library (170 sequences from 3 yearpooled DNA) as calculated based on the species OTU(97 16S rRNA gene sequence identity) by Goods CloneCoverage (Good 1953) was 89 and as estimated by theC
ace
statistic was 88 (Kemp and Aller 2004) The prob-ability of drawing a new sequence from this library at thespecies level on the next draw was 17 (Clayton andFrees 1987) The S
chao1
diversity estimate (Chao 1987)of this library predicted 185 unique species sequencesBecause the remaining three libraries were constructedwith a subset of the total dates and one library wasscreened by AFL prior to sequencing they were notincluded in the clone library coverage analyses The final
Fig 1
Unrooted consensus phylogram depicting a subset of common
Actinobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees following 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2D5 [2] 556 (AY792221)Crystal Bog 2KD4 [2] 594 (AY792222)
Crystal Bog 5G8 [16] 556 (AY792223)Crystal Bog 1D11 [5] 545 (AY792224)
Crystal Bog 2KE7 [1] 581 (AY792225)
Lake Fuchskuhle SW10 (AJ575554)Crystal Bog 1D1 [1] 611 (AY792226)
Crystal Bog 1F9 [4] 600 (AY792227)Crystal Bog 022D6 [2] 581 (AY792228)Crystal Bog 022E4 [1] 600 (AY792229)
Lake Fuchskuhle SW9 (AJ575553)Crystal Bog 2F5 [6] 556 (AY792230)
Rimov Reservoir R6 (AJ575502)
Soil Clone Sequences (Group I)01
Marine clone sequences
acIV
Marine clone sequences (Group I)
Crystal Bog 2A7 [11] 636 (AY792231)Crystal Bog 1C7 [6] 633 (AY792232)
Crystal Bog 1C4 [1] 615 (AY792233)
Crystal Bog 1D10 [5] 622 (AY792234)Crystal Bog 022F2 [1] 675 (AY792235)
Peat Bog TM262 (X92710)Peat Bog TM177 (X92701)Lake Fuchskuhle SW3 (AJ575548)Crystal Bog 2A8 [1] 615 (AY792236)
Peat Bog TM232 (X92709)Peat Bog TM210 (X92704)
Acidimicrobium ferrooxidans (U75647)Ferromicrobium acidophilum (AF251436)
So
il II-III
acII
acIIIMicrobacterium arborescens (X77443)Curtobacterium sp VKM Ac-2052 (AB042090)
Lake Wolfgangsee MWH-Wo1 (AJ507464)Lake Constance MWH-Bo1 (AJ507465)
acI-C
100
96
100
75
100
84
100100
75
100
100
100
100
100
93
100
100
100
100100
100
100
100
100
98
100
100
100
100100
10099
82
100
99
acI-B
acI-A
Crystal Bog 2A9 [3] 598 (DQ093399)
100
Crystal Bog 6H11 [1] 660 (DQ093400)100
Freshwater bacterial community dynamics
959
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
library of 96 clones which was screened for unidentifiedAFLs prior to sequencing returned six clones containingan AFL we had not yet identified
Community composition dynamics
ARISA fingerprints were obtained from samples collected
during the ice-off season for 3 years in Crystal Bog LakeThese fingerprints contained 126 different ARISA frag-ments (based on fragment length) and a total of 3041ARISA fragments summed across all 68-sample datesduring the 3 year sampling period Sixty-five (52) of theunique ARISA fragments and 2341 (77) of the totalARISA fragments were assigned a taxonomic identity
Fig 2
Unrooted consensus phylogram depicting a subset of common
Betaproteobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7000 trees following 30 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 022E2 [1] 797 (AY792237)Crystal Bog 2KB10 [2] 865 (AY792238)
bacterium FukuS35 (AJ290013)Crystal Bog 6F1 [6] 797 (AY792239)
Polynucleobacter necessarius (X93019)Clone ACK-C4 (U85124)str LD17 (Z99998)
clone ACK-L6 (U85123)Crystal Bog 022C7 [1] 812 (AY792240)
beta proteobacterium MWH-CaK5 (AJ550655)beta proteobacterium MWH-MoNR1 (AJ550649)
Ralstonia eutropha (M32021)Crystal Bog 2E8 [2] 806 (AY792241)
Ralstonia pickettii (S55004)Crystal Bog 1G9 [2] 806 (AY792242)Crystal Bog 5B11 [2] 755 (AY792243)
Crystal Bog 5C1 [1] 767 (AY792244)Crystal Bog 2G3 [1] 749 (AY792245)Crystal Bog 5F8 [1] 565 (AY792246)
Janthinobacterium lividum (Y08846)Crystal Bog 2KF8 [1] 787 (AY792247)Crystal Bog 1E12 [1] 812 (AY792248)
Crystal Bog 2KC4 [1] 812 (AY792249)Oxalobacter formigenes (U49757)
Crystal Bog 2KE9 [2] 930 (AY792250)
Crystal Bog 571A6 [13] 873 (AY792251)
Crystal Bog 2B1 [8] 873 (AY792252)Crystal Bog 571B4 [1] 619 (AY792253)
Crystal Bog 572G9 [2] 648 (AY792254)Burkholderia spN3P2 (U37344)
Crystal Bog 571H5 [4] 911 (AY792255)Crystal Bog 571B10 [10] 880 (AY792256)
Burkholderia glathei (Y17052)Variovorax paradoxus (AB008000)Rhodoferax fermentans (D16211)
bacterium RB13-C10 (AF407413)bacterium GKS2-122 (AJ290026)
bacterium FukuN55 (AJ289999)Lake Gossenkoellesee GKS16 (AJ224987)
Crystal Bog 021H12 [1] 958 (AY792257)Crystal Bog 2KD10 [1] 1026 (AY792258)
Lake Gossenkoellesse GKS98 (AJ224990)bacterium FukuN65 (AJ290001)
bacterium FukuS93 (AJ290018)Crystal Bog 6C11 [1] 1066 (AY792259)
Crystal Bog 1E9 [1] 937 (AY792260)Crystal Bog 1G5 [1] 930 (AY792261)
Crystal Bog 5E7 [1] 925 (AY792262)Bordetella bronchiseptica (X57026)
clone ACK-C30 (U85120)freshwater bacterium LD28 (Z99999)
Crystal Bog 022E9 [1] 865 (AY792263)Crystal Bog 6D11 [1] 741 (AY792264)
Methylophilus methylotrophus (M29021)Crystal Bog 021G5 [1] 828 (AY792265)
Neisseria gonorrhoeae (X07714)Hydrogenophilus thermoluteolus (AB009828)
01
Beta II
Beta I
Beta III
Beta IV79
79
59
98100
76100
67
65
100
100
82
92
100
100
70
99
100
100
69
100
10059
100
100
100100
100
100
87
70
79
100
100100
100
100
98
100
99
99100
62
82
83
57
8580
100100
100
Gammaproteobacteria
Crystal Bog 5D4 [3] 745 (DQ093405)100
Crystal Bog 2E5 [3] 745 (DQ093406)
99
85
Crystal Bog 6B4 [1] 885 (DQ093407)
Crystal Bog 571C2 [4] 646 (DQ093408)
100
100
960
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
based on matching to 16S rRNA gene sequences repre-senting 83 of the total fluorescence from all ARISAprofiles
The BCC of Crystal Bog Lake was quite dynamicLess than 20 of the AFLs were present on more than50 of the sampling dates but nearly 50 of the AFLswere present in at least one sample in each of the3 years (Fig 5) Although the overall BCC changes quiterapidly during a year some members of the bacterialcommunity did not share the dynamic behaviour of themajority The
Actinobacteria
especially members of theacI-B clade were more consistently present than any
other clade Four AFLs associated with the acI-B cladewere present on more than 90 of the 68 samplingdates (Fig 5) As a clade acI-B contributed more than25 of the total fluorescence units on more than 60 ofthe sampling dates while no other clade contributedgreater than 25 on more than two sampling dates(data not shown) In contrast the majority of phylotypesassociated with the
Beta
- or
Gammaproteobacteria
showed substantial presenceabsence variability Gener-ally the AFLs from these two classes of
Proteobacteria
were present on no more than 50 of the samplingdates (Fig 5)
Crystal Bog 5A11 [2] 492 (AY792266)Crystal Bog 1E2 [1] 622 (AY792267)
Crystal Bog 2D3 [10] 636 (AY792268)Crystal Bog 2E4 [1] 510 (AY792269)Crystal Bog 1E3 [6] 824 (AY792270)Crystal Bog 2C8 [5] 516 (AY792271)
Escherichia coli (Z83205)Crystal Bog 1F6 [1] 664 (AY792272)
Crystal Bog 6E4 [13] 664 (AY792273)Crystal Bog 5H4 [1] 732 (AY792274)
Cystal Bog 5B5 [1] 752 (AY792275)Vibrio vulnificus (X76333)
Aeromonas salmonicida (AJ009859)Crystal Bog 1B1 [4] 915 (AY792276)
Crystal Bog 1B8 [3] 911 (AY792277)Pseudomonas fluorescens (D84013)Crystal Bog 6B2 [1] 689 (AY792278)
Methylobacter BB51 (AF016981)bacterium FukuN13 (AJ290055)
Crystal Bog 1D4 [1] 925 (AY792279)Crystal Bog 6C12 [1] 715 (AY792280)
Methylococcus capsulatus (X72770)Crystal Bog 6E8 [1] 771 (AY792281)
Crystal Bog 021C3 [1] 763 (AY792282)Xanthomonas campestris (X95917)
Crystal Bog 571C8 [18] 806 (AY792283)Crystal Bog 5E2 [1] 1026 (AY792284)Crystal Bog 2KD12 [5] 885 (AY792285)
Beijerinckia indica (M59060)Crystal Bog 021H4 [1] 891 (AY792286)
Crystal Bog 022B5 [3] 950 (AY792287)Bradyrhizobium japonicum (U69638)
str 4-8 (AJ222832)Crystal Bog 5C10 [1] 885 (AY792288)Caulobacter fusiformis (AJ227759)
Crystal Bog 571H1 [1] 905 (AY792289)Sar Sea clone SAR 11 (X52172)
FW bacterium LD12 (Z99997)Caedibacter caryophila (X71837)
Rhodobacter sphaeroides (X53853)Lake Gossenkoellesee (AJ224989)bacterium FukuN22 (AJ289994)
Crystal Bog 2A11 [1] 898 (AY792290)uncultured bacterium FukuS56 (AJ290014)
Crystal Bog 022E8 [1] 911 (AY792291)Sphingomonas paucimobilis (X72722)
bacterium GKS2-124 (AJ2920027)Crystal Bog 5F2 [1] 937 (AY792292)
Sphingomonas sp B18 (AF410927)
Desulfovibrio burkinensis (AF053752)Syntrophus gentianae (X85132)
Bdellovibrio bacteriovorus (M59297)Geobacter metallireducens (L07834)
Nannocystis exedens (M94279)Crystal Bog 021E5 [7] 684 (AY792293)
01
Alp
hap
rote
ob
acte
ria
Alp
ha II
Alp
ha IV
Delta
pro
teo
bacte
ria
Archaea
Gam
map
rote
ob
acte
ria
100
100
100
100
52
100
100
100
9982
100
52100
76
90
100100
10097
10086
10068
71
100
94
94
58
77
100
94
100
100
94
85
100
9494
9493
94
100
gt 90
100
94
Crystal Bog 1H2 [1] 660 (DQ093401)
Alp
ha I
Alp
ha III
Fig 3
Unrooted consensus phylogram depict-ing a subset of
Alpha-
Delta-
and
Gammapro-teobacteria
based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Four common freshwater
Alphaproteobacteria
clades are illustrated Relationships were deter-mined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees fol-lowing 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics
957
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
autochthonous nutrients made available during phy-toplankton blooms in both freshwater (Eiler and Bertils-son 2004) and marine systems (Pinhassi
et al
2004Rooney-Varga
et al
2005)Although the existence of common epilimnetic freshwa-
ter bacterial phylotypes is apparent we do not yet under-stand the extent of variation and associated ecologicaldrivers of change in these dominant groups or of otherfreshwater bacterial community members over timescales of seasons or years Yannarell and colleagues(2003) showed that lakes experience quite dramaticchanges in BCC within- and between-years regardless oflake trophic status Similarly several other researchershave shown that BCC varies greatly between lake types(Methe and Zehr 1999 Lindstrom 2000 Yannarell andTriplett 2004) However in none of these studies weretemporally persistent and variable organisms identified Inan attempt to reconcile the notion that particularly com-mon freshwater taxa exist with the observation that overallBCC is highly variable we set out to identify and distin-guish between populations that are characteristically vari-able and those that tend to recur over seasonal andannual time scales
A focus on identifying dynamic and persistent bacterialpopulations should lead to a greater understanding offactors influencing bacterial community structure inlakes In addition an examination of factors that may beregulating not only the microbial community dynamicsbut the dynamics of individual members of that commu-nity will lead to an increased understanding of potentialfreshwater microbial-mediated processes linked tospecific organisms Therefore we designed a multiyearstudy of microbial populations in a single lake with asampling frequency designed to capture the pace ofchange in BCC in an effort to provide insight into theecology of both common and transient freshwaterbacteria
An initial study of Crystal Bog Lake revealed annualbut dynamic BCC patterns and a significant correlationbetween bacterial phylotype richness and eukaryoticplankton succession (Kent
et al
2004) However theidentity of individual phylotypes making up the bacteri-oplankton community and their population dynamics werenot described The aim of this study was to investigatepatterns in humic freshwater BCC and diversity at a fine-scale taxonomic level over 3 years to identify the persis-tent and dynamic bacterial community members and toexplore chemical and biological factors influencing individ-ual community member dynamics This was accomplishedby coupling automated ribosomal intergenic spacer anal-ysis (ARISA) fingerprinting to 16S ribosomal RNA (rRNA)gene clone library analysis (Brown
et al
2005) to exam-ine BCC at multiple levels of taxonomic resolution duringa long-term and intense sampling effort
Results
Bacterial community composition
Clone library analysis of Crystal Bog Lake identifiedrepresentatives of six bacterial phyla (
Actinobacteria
Bacteroidetes
Firmicutes
Proteobacteria
TM7
and
Ver-rucomicrobia
) including members of the classes
Alpha-
Beta-
Delta-
and
Gammaproteobacteria
and threeclones with unclassified phylogenetic affiliation (Figs 1ndash4)The
Betaproteobacteria
class contained the most repre-sentatives at each assigned operational taxonomic unit(OTU) (Table 1) while only one representative at eachOTU assignment was detected for the
Deltaproteobacte-ria
class and
Firmicutes
and
TM7
phyla (Table 1)Thirteen freshwater-specific clades (Glockner
et al
2000 Zwart
et al
2002 Warnecke
et al
2004) wereidentified in Crystal Bog Lake which included membersof the
Actinobacteria
Bacteroidetes
Alpha-
and
Betapro-teobacteria
and
Verrucomicrobia
All phylotypes classi-fied in the
Actinobacteria
phylum were members of thepreviously described freshwater-specific acI-B or soil andfreshwater-specific soil IIndashIII clade (Fig 1) Sequencesaffiliated with the soil IIndashIII clade clustered with previouslydefined peat bog clones Many of the Crystal Bog Lake
Betaproteobacteria
16S rRNA gene sequences wereaffiliated with the freshwater clades Beta I II III and IV(Fig 2) The remaining
Betaproteobacteria
clones wereclosely related to
Janthinobacterium
Ralstonia
or
Burkholderia
-type bacteria The majority of
Bacteroidetes
-related clones were members of the previously definedfreshwater clades CF I and CF III (Glockner
et al
2000)All of the
Alphaproteobacteria
clones were identified asbelonging to the freshwater clades Alpha I II III and IVand most of the
Verrucomicrobia
clones belonged to theFukuN18 freshwater clade (Figs 3 and 4) The remaining
Table 1
Number of OTU assignments by phylum
Phylum Clade
a
Species
b
AFL
c
Actinobacteria
2 7 13
Bacteroidetes
5 11 12
Firmicutes
1 1 1
Proteobacteria
18 34 52
Alpha
4 9 8
Beta
7 14 25
Delta
1 1 1
Gamma
6 10 18
TM7
1 1 1
Verrucomicrobia
2 2 6Unknowns 3 3 3Total 32 59 88
a
Clades were determined by the branching patterns obtained follow-ing phylogenetic tree construction and have sequence identity = 90
b
Species are defined as 16S rRNA sequence groups sharing = 97gene identity
c
AFL
=
ARISA fragment length
958
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
identified clones were not affiliated with any previouslydescribed freshwater clades Following alignment andtree inference three singleton clones Crystal Bog 2E1Crystal Bog 2KA12 and Crystal Bog 021B9 did not clusterwith any previously defined phyla (Fig 4) Clone CrystalBog 2E1 was most closely affiliated with 16S rRNA genesequences from the
TM6
and
TM7
phyla (Fig 4) ClonesCrystal Bog 2KA12 and Crystal Bog 021B9 were mostclosely affiliated with each other and are loosely affiliatedwith members of the
Verrucomicrobia
phylum (Fig 4)
Clone library analysis
Four Crystal Bog clone libraries produced 289 16S rRNAgene sequences and their corresponding ARISA fragment
lengths (AFLs measured as the number of nucleotidesamplified with primers 1406F and 23SR) The coverageof the largest clone library (170 sequences from 3 yearpooled DNA) as calculated based on the species OTU(97 16S rRNA gene sequence identity) by Goods CloneCoverage (Good 1953) was 89 and as estimated by theC
ace
statistic was 88 (Kemp and Aller 2004) The prob-ability of drawing a new sequence from this library at thespecies level on the next draw was 17 (Clayton andFrees 1987) The S
chao1
diversity estimate (Chao 1987)of this library predicted 185 unique species sequencesBecause the remaining three libraries were constructedwith a subset of the total dates and one library wasscreened by AFL prior to sequencing they were notincluded in the clone library coverage analyses The final
Fig 1
Unrooted consensus phylogram depicting a subset of common
Actinobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees following 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2D5 [2] 556 (AY792221)Crystal Bog 2KD4 [2] 594 (AY792222)
Crystal Bog 5G8 [16] 556 (AY792223)Crystal Bog 1D11 [5] 545 (AY792224)
Crystal Bog 2KE7 [1] 581 (AY792225)
Lake Fuchskuhle SW10 (AJ575554)Crystal Bog 1D1 [1] 611 (AY792226)
Crystal Bog 1F9 [4] 600 (AY792227)Crystal Bog 022D6 [2] 581 (AY792228)Crystal Bog 022E4 [1] 600 (AY792229)
Lake Fuchskuhle SW9 (AJ575553)Crystal Bog 2F5 [6] 556 (AY792230)
Rimov Reservoir R6 (AJ575502)
Soil Clone Sequences (Group I)01
Marine clone sequences
acIV
Marine clone sequences (Group I)
Crystal Bog 2A7 [11] 636 (AY792231)Crystal Bog 1C7 [6] 633 (AY792232)
Crystal Bog 1C4 [1] 615 (AY792233)
Crystal Bog 1D10 [5] 622 (AY792234)Crystal Bog 022F2 [1] 675 (AY792235)
Peat Bog TM262 (X92710)Peat Bog TM177 (X92701)Lake Fuchskuhle SW3 (AJ575548)Crystal Bog 2A8 [1] 615 (AY792236)
Peat Bog TM232 (X92709)Peat Bog TM210 (X92704)
Acidimicrobium ferrooxidans (U75647)Ferromicrobium acidophilum (AF251436)
So
il II-III
acII
acIIIMicrobacterium arborescens (X77443)Curtobacterium sp VKM Ac-2052 (AB042090)
Lake Wolfgangsee MWH-Wo1 (AJ507464)Lake Constance MWH-Bo1 (AJ507465)
acI-C
100
96
100
75
100
84
100100
75
100
100
100
100
100
93
100
100
100
100100
100
100
100
100
98
100
100
100
100100
10099
82
100
99
acI-B
acI-A
Crystal Bog 2A9 [3] 598 (DQ093399)
100
Crystal Bog 6H11 [1] 660 (DQ093400)100
Freshwater bacterial community dynamics
959
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
library of 96 clones which was screened for unidentifiedAFLs prior to sequencing returned six clones containingan AFL we had not yet identified
Community composition dynamics
ARISA fingerprints were obtained from samples collected
during the ice-off season for 3 years in Crystal Bog LakeThese fingerprints contained 126 different ARISA frag-ments (based on fragment length) and a total of 3041ARISA fragments summed across all 68-sample datesduring the 3 year sampling period Sixty-five (52) of theunique ARISA fragments and 2341 (77) of the totalARISA fragments were assigned a taxonomic identity
Fig 2
Unrooted consensus phylogram depicting a subset of common
Betaproteobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7000 trees following 30 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 022E2 [1] 797 (AY792237)Crystal Bog 2KB10 [2] 865 (AY792238)
bacterium FukuS35 (AJ290013)Crystal Bog 6F1 [6] 797 (AY792239)
Polynucleobacter necessarius (X93019)Clone ACK-C4 (U85124)str LD17 (Z99998)
clone ACK-L6 (U85123)Crystal Bog 022C7 [1] 812 (AY792240)
beta proteobacterium MWH-CaK5 (AJ550655)beta proteobacterium MWH-MoNR1 (AJ550649)
Ralstonia eutropha (M32021)Crystal Bog 2E8 [2] 806 (AY792241)
Ralstonia pickettii (S55004)Crystal Bog 1G9 [2] 806 (AY792242)Crystal Bog 5B11 [2] 755 (AY792243)
Crystal Bog 5C1 [1] 767 (AY792244)Crystal Bog 2G3 [1] 749 (AY792245)Crystal Bog 5F8 [1] 565 (AY792246)
Janthinobacterium lividum (Y08846)Crystal Bog 2KF8 [1] 787 (AY792247)Crystal Bog 1E12 [1] 812 (AY792248)
Crystal Bog 2KC4 [1] 812 (AY792249)Oxalobacter formigenes (U49757)
Crystal Bog 2KE9 [2] 930 (AY792250)
Crystal Bog 571A6 [13] 873 (AY792251)
Crystal Bog 2B1 [8] 873 (AY792252)Crystal Bog 571B4 [1] 619 (AY792253)
Crystal Bog 572G9 [2] 648 (AY792254)Burkholderia spN3P2 (U37344)
Crystal Bog 571H5 [4] 911 (AY792255)Crystal Bog 571B10 [10] 880 (AY792256)
Burkholderia glathei (Y17052)Variovorax paradoxus (AB008000)Rhodoferax fermentans (D16211)
bacterium RB13-C10 (AF407413)bacterium GKS2-122 (AJ290026)
bacterium FukuN55 (AJ289999)Lake Gossenkoellesee GKS16 (AJ224987)
Crystal Bog 021H12 [1] 958 (AY792257)Crystal Bog 2KD10 [1] 1026 (AY792258)
Lake Gossenkoellesse GKS98 (AJ224990)bacterium FukuN65 (AJ290001)
bacterium FukuS93 (AJ290018)Crystal Bog 6C11 [1] 1066 (AY792259)
Crystal Bog 1E9 [1] 937 (AY792260)Crystal Bog 1G5 [1] 930 (AY792261)
Crystal Bog 5E7 [1] 925 (AY792262)Bordetella bronchiseptica (X57026)
clone ACK-C30 (U85120)freshwater bacterium LD28 (Z99999)
Crystal Bog 022E9 [1] 865 (AY792263)Crystal Bog 6D11 [1] 741 (AY792264)
Methylophilus methylotrophus (M29021)Crystal Bog 021G5 [1] 828 (AY792265)
Neisseria gonorrhoeae (X07714)Hydrogenophilus thermoluteolus (AB009828)
01
Beta II
Beta I
Beta III
Beta IV79
79
59
98100
76100
67
65
100
100
82
92
100
100
70
99
100
100
69
100
10059
100
100
100100
100
100
87
70
79
100
100100
100
100
98
100
99
99100
62
82
83
57
8580
100100
100
Gammaproteobacteria
Crystal Bog 5D4 [3] 745 (DQ093405)100
Crystal Bog 2E5 [3] 745 (DQ093406)
99
85
Crystal Bog 6B4 [1] 885 (DQ093407)
Crystal Bog 571C2 [4] 646 (DQ093408)
100
100
960
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
based on matching to 16S rRNA gene sequences repre-senting 83 of the total fluorescence from all ARISAprofiles
The BCC of Crystal Bog Lake was quite dynamicLess than 20 of the AFLs were present on more than50 of the sampling dates but nearly 50 of the AFLswere present in at least one sample in each of the3 years (Fig 5) Although the overall BCC changes quiterapidly during a year some members of the bacterialcommunity did not share the dynamic behaviour of themajority The
Actinobacteria
especially members of theacI-B clade were more consistently present than any
other clade Four AFLs associated with the acI-B cladewere present on more than 90 of the 68 samplingdates (Fig 5) As a clade acI-B contributed more than25 of the total fluorescence units on more than 60 ofthe sampling dates while no other clade contributedgreater than 25 on more than two sampling dates(data not shown) In contrast the majority of phylotypesassociated with the
Beta
- or
Gammaproteobacteria
showed substantial presenceabsence variability Gener-ally the AFLs from these two classes of
Proteobacteria
were present on no more than 50 of the samplingdates (Fig 5)
Crystal Bog 5A11 [2] 492 (AY792266)Crystal Bog 1E2 [1] 622 (AY792267)
Crystal Bog 2D3 [10] 636 (AY792268)Crystal Bog 2E4 [1] 510 (AY792269)Crystal Bog 1E3 [6] 824 (AY792270)Crystal Bog 2C8 [5] 516 (AY792271)
Escherichia coli (Z83205)Crystal Bog 1F6 [1] 664 (AY792272)
Crystal Bog 6E4 [13] 664 (AY792273)Crystal Bog 5H4 [1] 732 (AY792274)
Cystal Bog 5B5 [1] 752 (AY792275)Vibrio vulnificus (X76333)
Aeromonas salmonicida (AJ009859)Crystal Bog 1B1 [4] 915 (AY792276)
Crystal Bog 1B8 [3] 911 (AY792277)Pseudomonas fluorescens (D84013)Crystal Bog 6B2 [1] 689 (AY792278)
Methylobacter BB51 (AF016981)bacterium FukuN13 (AJ290055)
Crystal Bog 1D4 [1] 925 (AY792279)Crystal Bog 6C12 [1] 715 (AY792280)
Methylococcus capsulatus (X72770)Crystal Bog 6E8 [1] 771 (AY792281)
Crystal Bog 021C3 [1] 763 (AY792282)Xanthomonas campestris (X95917)
Crystal Bog 571C8 [18] 806 (AY792283)Crystal Bog 5E2 [1] 1026 (AY792284)Crystal Bog 2KD12 [5] 885 (AY792285)
Beijerinckia indica (M59060)Crystal Bog 021H4 [1] 891 (AY792286)
Crystal Bog 022B5 [3] 950 (AY792287)Bradyrhizobium japonicum (U69638)
str 4-8 (AJ222832)Crystal Bog 5C10 [1] 885 (AY792288)Caulobacter fusiformis (AJ227759)
Crystal Bog 571H1 [1] 905 (AY792289)Sar Sea clone SAR 11 (X52172)
FW bacterium LD12 (Z99997)Caedibacter caryophila (X71837)
Rhodobacter sphaeroides (X53853)Lake Gossenkoellesee (AJ224989)bacterium FukuN22 (AJ289994)
Crystal Bog 2A11 [1] 898 (AY792290)uncultured bacterium FukuS56 (AJ290014)
Crystal Bog 022E8 [1] 911 (AY792291)Sphingomonas paucimobilis (X72722)
bacterium GKS2-124 (AJ2920027)Crystal Bog 5F2 [1] 937 (AY792292)
Sphingomonas sp B18 (AF410927)
Desulfovibrio burkinensis (AF053752)Syntrophus gentianae (X85132)
Bdellovibrio bacteriovorus (M59297)Geobacter metallireducens (L07834)
Nannocystis exedens (M94279)Crystal Bog 021E5 [7] 684 (AY792293)
01
Alp
hap
rote
ob
acte
ria
Alp
ha II
Alp
ha IV
Delta
pro
teo
bacte
ria
Archaea
Gam
map
rote
ob
acte
ria
100
100
100
100
52
100
100
100
9982
100
52100
76
90
100100
10097
10086
10068
71
100
94
94
58
77
100
94
100
100
94
85
100
9494
9493
94
100
gt 90
100
94
Crystal Bog 1H2 [1] 660 (DQ093401)
Alp
ha I
Alp
ha III
Fig 3
Unrooted consensus phylogram depict-ing a subset of
Alpha-
Delta-
and
Gammapro-teobacteria
based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Four common freshwater
Alphaproteobacteria
clades are illustrated Relationships were deter-mined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees fol-lowing 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
958
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
identified clones were not affiliated with any previouslydescribed freshwater clades Following alignment andtree inference three singleton clones Crystal Bog 2E1Crystal Bog 2KA12 and Crystal Bog 021B9 did not clusterwith any previously defined phyla (Fig 4) Clone CrystalBog 2E1 was most closely affiliated with 16S rRNA genesequences from the
TM6
and
TM7
phyla (Fig 4) ClonesCrystal Bog 2KA12 and Crystal Bog 021B9 were mostclosely affiliated with each other and are loosely affiliatedwith members of the
Verrucomicrobia
phylum (Fig 4)
Clone library analysis
Four Crystal Bog clone libraries produced 289 16S rRNAgene sequences and their corresponding ARISA fragment
lengths (AFLs measured as the number of nucleotidesamplified with primers 1406F and 23SR) The coverageof the largest clone library (170 sequences from 3 yearpooled DNA) as calculated based on the species OTU(97 16S rRNA gene sequence identity) by Goods CloneCoverage (Good 1953) was 89 and as estimated by theC
ace
statistic was 88 (Kemp and Aller 2004) The prob-ability of drawing a new sequence from this library at thespecies level on the next draw was 17 (Clayton andFrees 1987) The S
chao1
diversity estimate (Chao 1987)of this library predicted 185 unique species sequencesBecause the remaining three libraries were constructedwith a subset of the total dates and one library wasscreened by AFL prior to sequencing they were notincluded in the clone library coverage analyses The final
Fig 1
Unrooted consensus phylogram depicting a subset of common
Actinobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees following 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2D5 [2] 556 (AY792221)Crystal Bog 2KD4 [2] 594 (AY792222)
Crystal Bog 5G8 [16] 556 (AY792223)Crystal Bog 1D11 [5] 545 (AY792224)
Crystal Bog 2KE7 [1] 581 (AY792225)
Lake Fuchskuhle SW10 (AJ575554)Crystal Bog 1D1 [1] 611 (AY792226)
Crystal Bog 1F9 [4] 600 (AY792227)Crystal Bog 022D6 [2] 581 (AY792228)Crystal Bog 022E4 [1] 600 (AY792229)
Lake Fuchskuhle SW9 (AJ575553)Crystal Bog 2F5 [6] 556 (AY792230)
Rimov Reservoir R6 (AJ575502)
Soil Clone Sequences (Group I)01
Marine clone sequences
acIV
Marine clone sequences (Group I)
Crystal Bog 2A7 [11] 636 (AY792231)Crystal Bog 1C7 [6] 633 (AY792232)
Crystal Bog 1C4 [1] 615 (AY792233)
Crystal Bog 1D10 [5] 622 (AY792234)Crystal Bog 022F2 [1] 675 (AY792235)
Peat Bog TM262 (X92710)Peat Bog TM177 (X92701)Lake Fuchskuhle SW3 (AJ575548)Crystal Bog 2A8 [1] 615 (AY792236)
Peat Bog TM232 (X92709)Peat Bog TM210 (X92704)
Acidimicrobium ferrooxidans (U75647)Ferromicrobium acidophilum (AF251436)
So
il II-III
acII
acIIIMicrobacterium arborescens (X77443)Curtobacterium sp VKM Ac-2052 (AB042090)
Lake Wolfgangsee MWH-Wo1 (AJ507464)Lake Constance MWH-Bo1 (AJ507465)
acI-C
100
96
100
75
100
84
100100
75
100
100
100
100
100
93
100
100
100
100100
100
100
100
100
98
100
100
100
100100
10099
82
100
99
acI-B
acI-A
Crystal Bog 2A9 [3] 598 (DQ093399)
100
Crystal Bog 6H11 [1] 660 (DQ093400)100
Freshwater bacterial community dynamics
959
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
library of 96 clones which was screened for unidentifiedAFLs prior to sequencing returned six clones containingan AFL we had not yet identified
Community composition dynamics
ARISA fingerprints were obtained from samples collected
during the ice-off season for 3 years in Crystal Bog LakeThese fingerprints contained 126 different ARISA frag-ments (based on fragment length) and a total of 3041ARISA fragments summed across all 68-sample datesduring the 3 year sampling period Sixty-five (52) of theunique ARISA fragments and 2341 (77) of the totalARISA fragments were assigned a taxonomic identity
Fig 2
Unrooted consensus phylogram depicting a subset of common
Betaproteobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7000 trees following 30 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 022E2 [1] 797 (AY792237)Crystal Bog 2KB10 [2] 865 (AY792238)
bacterium FukuS35 (AJ290013)Crystal Bog 6F1 [6] 797 (AY792239)
Polynucleobacter necessarius (X93019)Clone ACK-C4 (U85124)str LD17 (Z99998)
clone ACK-L6 (U85123)Crystal Bog 022C7 [1] 812 (AY792240)
beta proteobacterium MWH-CaK5 (AJ550655)beta proteobacterium MWH-MoNR1 (AJ550649)
Ralstonia eutropha (M32021)Crystal Bog 2E8 [2] 806 (AY792241)
Ralstonia pickettii (S55004)Crystal Bog 1G9 [2] 806 (AY792242)Crystal Bog 5B11 [2] 755 (AY792243)
Crystal Bog 5C1 [1] 767 (AY792244)Crystal Bog 2G3 [1] 749 (AY792245)Crystal Bog 5F8 [1] 565 (AY792246)
Janthinobacterium lividum (Y08846)Crystal Bog 2KF8 [1] 787 (AY792247)Crystal Bog 1E12 [1] 812 (AY792248)
Crystal Bog 2KC4 [1] 812 (AY792249)Oxalobacter formigenes (U49757)
Crystal Bog 2KE9 [2] 930 (AY792250)
Crystal Bog 571A6 [13] 873 (AY792251)
Crystal Bog 2B1 [8] 873 (AY792252)Crystal Bog 571B4 [1] 619 (AY792253)
Crystal Bog 572G9 [2] 648 (AY792254)Burkholderia spN3P2 (U37344)
Crystal Bog 571H5 [4] 911 (AY792255)Crystal Bog 571B10 [10] 880 (AY792256)
Burkholderia glathei (Y17052)Variovorax paradoxus (AB008000)Rhodoferax fermentans (D16211)
bacterium RB13-C10 (AF407413)bacterium GKS2-122 (AJ290026)
bacterium FukuN55 (AJ289999)Lake Gossenkoellesee GKS16 (AJ224987)
Crystal Bog 021H12 [1] 958 (AY792257)Crystal Bog 2KD10 [1] 1026 (AY792258)
Lake Gossenkoellesse GKS98 (AJ224990)bacterium FukuN65 (AJ290001)
bacterium FukuS93 (AJ290018)Crystal Bog 6C11 [1] 1066 (AY792259)
Crystal Bog 1E9 [1] 937 (AY792260)Crystal Bog 1G5 [1] 930 (AY792261)
Crystal Bog 5E7 [1] 925 (AY792262)Bordetella bronchiseptica (X57026)
clone ACK-C30 (U85120)freshwater bacterium LD28 (Z99999)
Crystal Bog 022E9 [1] 865 (AY792263)Crystal Bog 6D11 [1] 741 (AY792264)
Methylophilus methylotrophus (M29021)Crystal Bog 021G5 [1] 828 (AY792265)
Neisseria gonorrhoeae (X07714)Hydrogenophilus thermoluteolus (AB009828)
01
Beta II
Beta I
Beta III
Beta IV79
79
59
98100
76100
67
65
100
100
82
92
100
100
70
99
100
100
69
100
10059
100
100
100100
100
100
87
70
79
100
100100
100
100
98
100
99
99100
62
82
83
57
8580
100100
100
Gammaproteobacteria
Crystal Bog 5D4 [3] 745 (DQ093405)100
Crystal Bog 2E5 [3] 745 (DQ093406)
99
85
Crystal Bog 6B4 [1] 885 (DQ093407)
Crystal Bog 571C2 [4] 646 (DQ093408)
100
100
960
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
based on matching to 16S rRNA gene sequences repre-senting 83 of the total fluorescence from all ARISAprofiles
The BCC of Crystal Bog Lake was quite dynamicLess than 20 of the AFLs were present on more than50 of the sampling dates but nearly 50 of the AFLswere present in at least one sample in each of the3 years (Fig 5) Although the overall BCC changes quiterapidly during a year some members of the bacterialcommunity did not share the dynamic behaviour of themajority The
Actinobacteria
especially members of theacI-B clade were more consistently present than any
other clade Four AFLs associated with the acI-B cladewere present on more than 90 of the 68 samplingdates (Fig 5) As a clade acI-B contributed more than25 of the total fluorescence units on more than 60 ofthe sampling dates while no other clade contributedgreater than 25 on more than two sampling dates(data not shown) In contrast the majority of phylotypesassociated with the
Beta
- or
Gammaproteobacteria
showed substantial presenceabsence variability Gener-ally the AFLs from these two classes of
Proteobacteria
were present on no more than 50 of the samplingdates (Fig 5)
Crystal Bog 5A11 [2] 492 (AY792266)Crystal Bog 1E2 [1] 622 (AY792267)
Crystal Bog 2D3 [10] 636 (AY792268)Crystal Bog 2E4 [1] 510 (AY792269)Crystal Bog 1E3 [6] 824 (AY792270)Crystal Bog 2C8 [5] 516 (AY792271)
Escherichia coli (Z83205)Crystal Bog 1F6 [1] 664 (AY792272)
Crystal Bog 6E4 [13] 664 (AY792273)Crystal Bog 5H4 [1] 732 (AY792274)
Cystal Bog 5B5 [1] 752 (AY792275)Vibrio vulnificus (X76333)
Aeromonas salmonicida (AJ009859)Crystal Bog 1B1 [4] 915 (AY792276)
Crystal Bog 1B8 [3] 911 (AY792277)Pseudomonas fluorescens (D84013)Crystal Bog 6B2 [1] 689 (AY792278)
Methylobacter BB51 (AF016981)bacterium FukuN13 (AJ290055)
Crystal Bog 1D4 [1] 925 (AY792279)Crystal Bog 6C12 [1] 715 (AY792280)
Methylococcus capsulatus (X72770)Crystal Bog 6E8 [1] 771 (AY792281)
Crystal Bog 021C3 [1] 763 (AY792282)Xanthomonas campestris (X95917)
Crystal Bog 571C8 [18] 806 (AY792283)Crystal Bog 5E2 [1] 1026 (AY792284)Crystal Bog 2KD12 [5] 885 (AY792285)
Beijerinckia indica (M59060)Crystal Bog 021H4 [1] 891 (AY792286)
Crystal Bog 022B5 [3] 950 (AY792287)Bradyrhizobium japonicum (U69638)
str 4-8 (AJ222832)Crystal Bog 5C10 [1] 885 (AY792288)Caulobacter fusiformis (AJ227759)
Crystal Bog 571H1 [1] 905 (AY792289)Sar Sea clone SAR 11 (X52172)
FW bacterium LD12 (Z99997)Caedibacter caryophila (X71837)
Rhodobacter sphaeroides (X53853)Lake Gossenkoellesee (AJ224989)bacterium FukuN22 (AJ289994)
Crystal Bog 2A11 [1] 898 (AY792290)uncultured bacterium FukuS56 (AJ290014)
Crystal Bog 022E8 [1] 911 (AY792291)Sphingomonas paucimobilis (X72722)
bacterium GKS2-124 (AJ2920027)Crystal Bog 5F2 [1] 937 (AY792292)
Sphingomonas sp B18 (AF410927)
Desulfovibrio burkinensis (AF053752)Syntrophus gentianae (X85132)
Bdellovibrio bacteriovorus (M59297)Geobacter metallireducens (L07834)
Nannocystis exedens (M94279)Crystal Bog 021E5 [7] 684 (AY792293)
01
Alp
hap
rote
ob
acte
ria
Alp
ha II
Alp
ha IV
Delta
pro
teo
bacte
ria
Archaea
Gam
map
rote
ob
acte
ria
100
100
100
100
52
100
100
100
9982
100
52100
76
90
100100
10097
10086
10068
71
100
94
94
58
77
100
94
100
100
94
85
100
9494
9493
94
100
gt 90
100
94
Crystal Bog 1H2 [1] 660 (DQ093401)
Alp
ha I
Alp
ha III
Fig 3
Unrooted consensus phylogram depict-ing a subset of
Alpha-
Delta-
and
Gammapro-teobacteria
based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Four common freshwater
Alphaproteobacteria
clades are illustrated Relationships were deter-mined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees fol-lowing 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics
959
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
library of 96 clones which was screened for unidentifiedAFLs prior to sequencing returned six clones containingan AFL we had not yet identified
Community composition dynamics
ARISA fingerprints were obtained from samples collected
during the ice-off season for 3 years in Crystal Bog LakeThese fingerprints contained 126 different ARISA frag-ments (based on fragment length) and a total of 3041ARISA fragments summed across all 68-sample datesduring the 3 year sampling period Sixty-five (52) of theunique ARISA fragments and 2341 (77) of the totalARISA fragments were assigned a taxonomic identity
Fig 2
Unrooted consensus phylogram depicting a subset of common
Betaproteobacteria
freshwater clades based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7000 trees following 30 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 022E2 [1] 797 (AY792237)Crystal Bog 2KB10 [2] 865 (AY792238)
bacterium FukuS35 (AJ290013)Crystal Bog 6F1 [6] 797 (AY792239)
Polynucleobacter necessarius (X93019)Clone ACK-C4 (U85124)str LD17 (Z99998)
clone ACK-L6 (U85123)Crystal Bog 022C7 [1] 812 (AY792240)
beta proteobacterium MWH-CaK5 (AJ550655)beta proteobacterium MWH-MoNR1 (AJ550649)
Ralstonia eutropha (M32021)Crystal Bog 2E8 [2] 806 (AY792241)
Ralstonia pickettii (S55004)Crystal Bog 1G9 [2] 806 (AY792242)Crystal Bog 5B11 [2] 755 (AY792243)
Crystal Bog 5C1 [1] 767 (AY792244)Crystal Bog 2G3 [1] 749 (AY792245)Crystal Bog 5F8 [1] 565 (AY792246)
Janthinobacterium lividum (Y08846)Crystal Bog 2KF8 [1] 787 (AY792247)Crystal Bog 1E12 [1] 812 (AY792248)
Crystal Bog 2KC4 [1] 812 (AY792249)Oxalobacter formigenes (U49757)
Crystal Bog 2KE9 [2] 930 (AY792250)
Crystal Bog 571A6 [13] 873 (AY792251)
Crystal Bog 2B1 [8] 873 (AY792252)Crystal Bog 571B4 [1] 619 (AY792253)
Crystal Bog 572G9 [2] 648 (AY792254)Burkholderia spN3P2 (U37344)
Crystal Bog 571H5 [4] 911 (AY792255)Crystal Bog 571B10 [10] 880 (AY792256)
Burkholderia glathei (Y17052)Variovorax paradoxus (AB008000)Rhodoferax fermentans (D16211)
bacterium RB13-C10 (AF407413)bacterium GKS2-122 (AJ290026)
bacterium FukuN55 (AJ289999)Lake Gossenkoellesee GKS16 (AJ224987)
Crystal Bog 021H12 [1] 958 (AY792257)Crystal Bog 2KD10 [1] 1026 (AY792258)
Lake Gossenkoellesse GKS98 (AJ224990)bacterium FukuN65 (AJ290001)
bacterium FukuS93 (AJ290018)Crystal Bog 6C11 [1] 1066 (AY792259)
Crystal Bog 1E9 [1] 937 (AY792260)Crystal Bog 1G5 [1] 930 (AY792261)
Crystal Bog 5E7 [1] 925 (AY792262)Bordetella bronchiseptica (X57026)
clone ACK-C30 (U85120)freshwater bacterium LD28 (Z99999)
Crystal Bog 022E9 [1] 865 (AY792263)Crystal Bog 6D11 [1] 741 (AY792264)
Methylophilus methylotrophus (M29021)Crystal Bog 021G5 [1] 828 (AY792265)
Neisseria gonorrhoeae (X07714)Hydrogenophilus thermoluteolus (AB009828)
01
Beta II
Beta I
Beta III
Beta IV79
79
59
98100
76100
67
65
100
100
82
92
100
100
70
99
100
100
69
100
10059
100
100
100100
100
100
87
70
79
100
100100
100
100
98
100
99
99100
62
82
83
57
8580
100100
100
Gammaproteobacteria
Crystal Bog 5D4 [3] 745 (DQ093405)100
Crystal Bog 2E5 [3] 745 (DQ093406)
99
85
Crystal Bog 6B4 [1] 885 (DQ093407)
Crystal Bog 571C2 [4] 646 (DQ093408)
100
100
960
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
based on matching to 16S rRNA gene sequences repre-senting 83 of the total fluorescence from all ARISAprofiles
The BCC of Crystal Bog Lake was quite dynamicLess than 20 of the AFLs were present on more than50 of the sampling dates but nearly 50 of the AFLswere present in at least one sample in each of the3 years (Fig 5) Although the overall BCC changes quiterapidly during a year some members of the bacterialcommunity did not share the dynamic behaviour of themajority The
Actinobacteria
especially members of theacI-B clade were more consistently present than any
other clade Four AFLs associated with the acI-B cladewere present on more than 90 of the 68 samplingdates (Fig 5) As a clade acI-B contributed more than25 of the total fluorescence units on more than 60 ofthe sampling dates while no other clade contributedgreater than 25 on more than two sampling dates(data not shown) In contrast the majority of phylotypesassociated with the
Beta
- or
Gammaproteobacteria
showed substantial presenceabsence variability Gener-ally the AFLs from these two classes of
Proteobacteria
were present on no more than 50 of the samplingdates (Fig 5)
Crystal Bog 5A11 [2] 492 (AY792266)Crystal Bog 1E2 [1] 622 (AY792267)
Crystal Bog 2D3 [10] 636 (AY792268)Crystal Bog 2E4 [1] 510 (AY792269)Crystal Bog 1E3 [6] 824 (AY792270)Crystal Bog 2C8 [5] 516 (AY792271)
Escherichia coli (Z83205)Crystal Bog 1F6 [1] 664 (AY792272)
Crystal Bog 6E4 [13] 664 (AY792273)Crystal Bog 5H4 [1] 732 (AY792274)
Cystal Bog 5B5 [1] 752 (AY792275)Vibrio vulnificus (X76333)
Aeromonas salmonicida (AJ009859)Crystal Bog 1B1 [4] 915 (AY792276)
Crystal Bog 1B8 [3] 911 (AY792277)Pseudomonas fluorescens (D84013)Crystal Bog 6B2 [1] 689 (AY792278)
Methylobacter BB51 (AF016981)bacterium FukuN13 (AJ290055)
Crystal Bog 1D4 [1] 925 (AY792279)Crystal Bog 6C12 [1] 715 (AY792280)
Methylococcus capsulatus (X72770)Crystal Bog 6E8 [1] 771 (AY792281)
Crystal Bog 021C3 [1] 763 (AY792282)Xanthomonas campestris (X95917)
Crystal Bog 571C8 [18] 806 (AY792283)Crystal Bog 5E2 [1] 1026 (AY792284)Crystal Bog 2KD12 [5] 885 (AY792285)
Beijerinckia indica (M59060)Crystal Bog 021H4 [1] 891 (AY792286)
Crystal Bog 022B5 [3] 950 (AY792287)Bradyrhizobium japonicum (U69638)
str 4-8 (AJ222832)Crystal Bog 5C10 [1] 885 (AY792288)Caulobacter fusiformis (AJ227759)
Crystal Bog 571H1 [1] 905 (AY792289)Sar Sea clone SAR 11 (X52172)
FW bacterium LD12 (Z99997)Caedibacter caryophila (X71837)
Rhodobacter sphaeroides (X53853)Lake Gossenkoellesee (AJ224989)bacterium FukuN22 (AJ289994)
Crystal Bog 2A11 [1] 898 (AY792290)uncultured bacterium FukuS56 (AJ290014)
Crystal Bog 022E8 [1] 911 (AY792291)Sphingomonas paucimobilis (X72722)
bacterium GKS2-124 (AJ2920027)Crystal Bog 5F2 [1] 937 (AY792292)
Sphingomonas sp B18 (AF410927)
Desulfovibrio burkinensis (AF053752)Syntrophus gentianae (X85132)
Bdellovibrio bacteriovorus (M59297)Geobacter metallireducens (L07834)
Nannocystis exedens (M94279)Crystal Bog 021E5 [7] 684 (AY792293)
01
Alp
hap
rote
ob
acte
ria
Alp
ha II
Alp
ha IV
Delta
pro
teo
bacte
ria
Archaea
Gam
map
rote
ob
acte
ria
100
100
100
100
52
100
100
100
9982
100
52100
76
90
100100
10097
10086
10068
71
100
94
94
58
77
100
94
100
100
94
85
100
9494
9493
94
100
gt 90
100
94
Crystal Bog 1H2 [1] 660 (DQ093401)
Alp
ha I
Alp
ha III
Fig 3
Unrooted consensus phylogram depict-ing a subset of
Alpha-
Delta-
and
Gammapro-teobacteria
based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Four common freshwater
Alphaproteobacteria
clades are illustrated Relationships were deter-mined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees fol-lowing 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
960
R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
based on matching to 16S rRNA gene sequences repre-senting 83 of the total fluorescence from all ARISAprofiles
The BCC of Crystal Bog Lake was quite dynamicLess than 20 of the AFLs were present on more than50 of the sampling dates but nearly 50 of the AFLswere present in at least one sample in each of the3 years (Fig 5) Although the overall BCC changes quiterapidly during a year some members of the bacterialcommunity did not share the dynamic behaviour of themajority The
Actinobacteria
especially members of theacI-B clade were more consistently present than any
other clade Four AFLs associated with the acI-B cladewere present on more than 90 of the 68 samplingdates (Fig 5) As a clade acI-B contributed more than25 of the total fluorescence units on more than 60 ofthe sampling dates while no other clade contributedgreater than 25 on more than two sampling dates(data not shown) In contrast the majority of phylotypesassociated with the
Beta
- or
Gammaproteobacteria
showed substantial presenceabsence variability Gener-ally the AFLs from these two classes of
Proteobacteria
were present on no more than 50 of the samplingdates (Fig 5)
Crystal Bog 5A11 [2] 492 (AY792266)Crystal Bog 1E2 [1] 622 (AY792267)
Crystal Bog 2D3 [10] 636 (AY792268)Crystal Bog 2E4 [1] 510 (AY792269)Crystal Bog 1E3 [6] 824 (AY792270)Crystal Bog 2C8 [5] 516 (AY792271)
Escherichia coli (Z83205)Crystal Bog 1F6 [1] 664 (AY792272)
Crystal Bog 6E4 [13] 664 (AY792273)Crystal Bog 5H4 [1] 732 (AY792274)
Cystal Bog 5B5 [1] 752 (AY792275)Vibrio vulnificus (X76333)
Aeromonas salmonicida (AJ009859)Crystal Bog 1B1 [4] 915 (AY792276)
Crystal Bog 1B8 [3] 911 (AY792277)Pseudomonas fluorescens (D84013)Crystal Bog 6B2 [1] 689 (AY792278)
Methylobacter BB51 (AF016981)bacterium FukuN13 (AJ290055)
Crystal Bog 1D4 [1] 925 (AY792279)Crystal Bog 6C12 [1] 715 (AY792280)
Methylococcus capsulatus (X72770)Crystal Bog 6E8 [1] 771 (AY792281)
Crystal Bog 021C3 [1] 763 (AY792282)Xanthomonas campestris (X95917)
Crystal Bog 571C8 [18] 806 (AY792283)Crystal Bog 5E2 [1] 1026 (AY792284)Crystal Bog 2KD12 [5] 885 (AY792285)
Beijerinckia indica (M59060)Crystal Bog 021H4 [1] 891 (AY792286)
Crystal Bog 022B5 [3] 950 (AY792287)Bradyrhizobium japonicum (U69638)
str 4-8 (AJ222832)Crystal Bog 5C10 [1] 885 (AY792288)Caulobacter fusiformis (AJ227759)
Crystal Bog 571H1 [1] 905 (AY792289)Sar Sea clone SAR 11 (X52172)
FW bacterium LD12 (Z99997)Caedibacter caryophila (X71837)
Rhodobacter sphaeroides (X53853)Lake Gossenkoellesee (AJ224989)bacterium FukuN22 (AJ289994)
Crystal Bog 2A11 [1] 898 (AY792290)uncultured bacterium FukuS56 (AJ290014)
Crystal Bog 022E8 [1] 911 (AY792291)Sphingomonas paucimobilis (X72722)
bacterium GKS2-124 (AJ2920027)Crystal Bog 5F2 [1] 937 (AY792292)
Sphingomonas sp B18 (AF410927)
Desulfovibrio burkinensis (AF053752)Syntrophus gentianae (X85132)
Bdellovibrio bacteriovorus (M59297)Geobacter metallireducens (L07834)
Nannocystis exedens (M94279)Crystal Bog 021E5 [7] 684 (AY792293)
01
Alp
hap
rote
ob
acte
ria
Alp
ha II
Alp
ha IV
Delta
pro
teo
bacte
ria
Archaea
Gam
map
rote
ob
acte
ria
100
100
100
100
52
100
100
100
9982
100
52100
76
90
100100
10097
10086
10068
71
100
94
94
58
77
100
94
100
100
94
85
100
9494
9493
94
100
gt 90
100
94
Crystal Bog 1H2 [1] 660 (DQ093401)
Alp
ha I
Alp
ha III
Fig 3
Unrooted consensus phylogram depict-ing a subset of
Alpha-
Delta-
and
Gammapro-teobacteria
based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Four common freshwater
Alphaproteobacteria
clades are illustrated Relationships were deter-mined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 8000 trees fol-lowing 20 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics
961
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd
Environmental Microbiology
8
956ndash970
Biological chemical and physical drivers of bacterial community composition
The physicalchemical parameters measured in thisstudy by themselves or in combination did not signifi-cantly explain the BCC change observed in 2002 (Kent
et al
2004) Likewise the temporal dynamics of individ-ual bacterioplankton community members (assessed by
AFL relative fluorescence) were not significantly corre-lated to the changes of any single measured chemicaland physical factor or any combination thereof (data notshown)
Phytoplankton community succession and het-erotrophic nanoflagellate (HNF) abundance were closelymonitored during 2002 (Kent
et al 2004) The dynamicsof dominant assemblages (regimes) are described here
Fig 4 Unrooted consensus phylogram depicting a subset of Bacteroidetes TM7 Verrucomicrobia and Firmicutes phyla based on nearly full-length (gt 1300 bp) 16S rRNA gene sequences Two freshwater Bacteroidetes and one freshwater Verrucomicrobia clade is depicted Sequences not belonging to a known phylum are labelled as Unknown Relationships were determined by Bayesian analysis (software MrBayes 30) using a 50 base frequency mask with 100 000 generations yielding 7500 trees following 25 000 generations of burnin Nodes with posterior probability values of gt50 are indicated The unique (lt 97 16S rRNA gene sequence identity or unshared AFL) sequences identified from Crystal Bog Lake are shown in bold with their corresponding AFL in curly brackets The number of clones represented by each depicted sequence is shown in square brackets and the GenBank accession number of all sequences is indicated in parentheses The scale bar indicates 01 changes per site
Crystal Bog 2KG7 [1] 780 (AY792294)Crystal Bog 2KE10 [1] 806 (AY792295)
Crystal Bog 022B7 [1] 817 (AY792296)Crystal Bog 6F6 [1] 749 (AY792297)
bacterium FukuN24 (AJ289995)
bacterium FukuS59 (AJ290042)Crystal Bog 022H6 [3] 905 (AY792298)uncultured bacterium GKS2-106 (AJ290025)
Crystal Bog 022A2 [1] 787 (AY792299)bacterium FukuN23 (AJ290011)
Crystal Bog 2C5 [1] 930 (AY792300)Crystal Bog 5D8 [1] 920 (AY792301)
Flavobacterium aquatile (M62797)
bacterium GKS2-33 (AJ290035)Crystal Bog 1D6 [1] 652 (AY792302)
Flexibacter litoralis (M58784)
Cytophagales bacterium 13 (AF361196)bacterium AH57 (AJ289964)
Taxeobacter gelupurpurascens (Y18836)Crystal Bog 5A2 [2] 495 (AY792303)Crystal Bog 2F6 [1] 626 (AY792304)
Sphingobacterium thalpophilum (M58779)Crystal Bog 5H5 [2] 780 (AY7922305)
clone WCHB1-11 (AF050603)clone WCHB07 (AF050600)
clone WCHB1-58 (AF050610)clone WCHB1-15 (AF050596)
Crystal Bog 2KD8 [4] 1116 (AY792306)
Crystal Bog 2E1 [1] 759 (AY792307)
Crystal Bog 1B6 [1] 660 (AY792308)Crystal Bog 2KH1 [3] 937 (AY792309)
Crystal Bog 1D5 [2] 911 (AY792310)
clone DA101 (Y07576)Verrucomicrobium spinosum (X90515)
Crystal Bog 022E6 [1] 759 (AY792311)
clone WCHB1-25 (AF050559)clone WCHB1-41 (AF050560)
Crystal Bog 021B9 [1] 806 (AY792312)Crystal Bog 2KA12 [1] 749 (AY792313)
Bacillus smithii (Z26935)Staphylococcus aureus (L36472)Crystal Bog 5A7 [2] 586 (AY792314)
Asteroleplasma anaerobium (M22351)
01
Bactero
idetes
TM
7V
erruco
micro
bia
Firm
icutes
Archaea
Fu
kuN
18
Unknown 1
Unknown 2
9689
6689
100
100
100100
72
61100
100
100
90
88
100
70
97
100100
79
61
8282
100
100
100100
100
100
100
94
100
100
100
84
100
Schohsee clone SF11 (AJ697697)
Schohsee clone SF54 (AJ697701)
100100
100
Crystal Bog 6G4 [1] 821 (DQ093402)
Schohsee clone SF21 (AJ697698)100
100
Crystal Bog 021C4 [1] 920 (DQ093403)Crystal Bog 022B10 [1] 905 (DQ093404)
97
79100
82
61
CF
IC
F III
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
962 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
as follows the Chrysophyte Dinobryon and dinoflagellatePeridiniopsis co-dominated (in terms of biovolume) thephytoplankton population during a significant increase inabundance of HNFs during the mid-spring season Cryp-
tomonas a motile unicellular photosynthetic alga domi-nated during late spring the dinoflagellates Gymnodiniumfuscum Peridinium limbatum and Peridinium cinctum co-dominated during early summer and the two Peridinium
Fig 5 Three year plot by sample date showing the presenceabsence of all AFLs associated with clades identified in Crystal Bog Lake The presence of a coloured box indicates that the AFL was present on that sample date The months and years listed across the top row correspond to the first sampled date within that monthyear The phyla clades and AFLs are listed to the left of the respective plot row All AFLs listed below each clade designation belong to that clade The AFL and the total number of sample dates on which the AFL was present are listed to the right of the corresponding plot row AFLs assigned to more than one clade are listed separately at the bottom as mixed assignments Phylogenetic affiliation not listed Verrucomicrobia (Ve) and unknown (Un)
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics 963
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
species alone dominated during mid-summer In additionthe unicellular bristled Chrysophyte Mallomonas showeda significant increase in total biovolume during the end ofearly summer and beginning of mid-summer although itwas not the dominant phytoplankton community memberat any time during that period
Calculation of the Pearson productndashmoment correlationcoefficient revealed significant correlations between indi-vidual bacterial phylotypes (assessed by AFL relative flu-orescence) and individual dominant phytoplanktonregimes over the course of 2002 (Table 2) The majorityof identified AFLs (41 out of 65) exhibited strong cor-relations (P le 0001) to the dynamics of at leastone phytoplanktonHNF regime in 2002 Canonicalcorrespondence analysis (CCA) using individual phy-toplankton species biovolume as explanatory variablesillustrates the relationships between individual AFLs andparticular phytoplankton regimes (Fig 6) Notably AFLsassigned to the Beta and Gammaproteobacteria are asso-ciated with the intense bacterivory period that includedblooms of Peridiniopsis and Dinobryon while the majority
of Actinobacteria AFLs do not appear to be influenced byany of the measured phytoplankton taxa (Fig 6) Alto-gether 69 of the AFLndashphytoplankton relationship isexplained by the first two CCA axes and the relationshipis significant (P = 001)
Several groups of covarying phylotypes related toindividual phytoplanktonHNF regimes became apparentfrom these analyses (Table 2 and Fig 6) An analysis ofsimilarity (ANOSIM) with groups defined by the strongestcorrelation to a phytoplankton regime (listed in boldTable 2) confirmed the significance of these covaryingassemblages (R-value = 08 P-value lt 0001) Althoughthe taxonomic composition of the bacterial communitycomprising the assemblages varied greatly a few trendsemerged The acI-B clade of Actinobacteria relativeabundance was negatively correlated to the presence offlagellate grazers which indicates the acI-B clade wasa less significant part of the community during thisintense bacterivory period On the other hand a largenumber of phylotypes from clades in the Betaproteo-bacteria Bacteroidetes and Gammaproteobacteria
Table 2 Pearson productndashmoment correlation values between bacterial phylotype relative abundance and algal phylotype biovolumea or HNFabundance
Cladeb AFL HNF Per Din Cryp Gym Mal P cin P lim
Beta IV 741 minusminusminusminus051 ndash ndash ndash ndash ndash ndash ndashCF I 817 minusminusminusminus060 ndash ndash ndash ndash ndash ndash ndashSoil IIndashIII 675 051 ndash ndash ndash ndash ndash ndash ndashCB_Ga1 732 063 053 ndash ndash ndash ndash ndash ndashCB_Ga6 763 059 052 ndash ndash ndash ndash ndash ndashacI-B 545 ndash minusminusminusminus064 minus056 ndash ndash ndash ndash ndashacI-B 556 ndash minusminusminusminus053 ndash ndash ndash ndash ndash ndashacI-B 594 minus060 minusminusminusminus068 ndash ndash ndash ndash 054 ndashCF III 652 077 083 079 ndash minus054 ndash minus055 ndashCB_Ga4 715 081 086 083 ndash ndash ndash ndash ndashCB_Be2 755 ndash 064 ndash ndash ndash ndash ndash ndashCB_Ga5 771 051 086 074 ndash ndash ndash ndash ndashBeta IV 828 ndash 063 ndash ndash ndash ndash ndash ndashCB_Be1 880 ndash 074 058 ndash ndash ndash ndash ndashBeta III 1066 070 078 064 ndash ndash ndash ndash ndashAlpha I 950 ndash ndash ndash 061 ndash minus053 ndash ndashDelta 684 ndash ndash ndash ndash 081 070 ndash ndashCB_Ga1 516 ndash ndash ndash ndash 058 084 ndash ndashFirm 586 ndash ndash ndash ndash 056 060 055 ndashacI_B 611 ndash ndash ndash ndash ndash 068 058 054Soil IIndashIII 615 ndash ndash ndash ndash ndash 068 ndash ndashSoil IIndashIII 633 ndash ndash ndash ndash ndash 055 ndash ndashCB_Ga1 664 ndash ndash ndash ndash 063 084 058 ndashCB_Ga1 824 ndash ndash ndash ndash 068 071 ndash ndashCB_Ga1 492 ndash ndash ndash ndash ndash ndash 074 060CB_Be1 619 ndash ndash ndash ndash ndash ndash 068 063Beta II 797 ndash ndash ndash ndash ndash ndash 065 ndashCB_Ba2 821 ndash ndash ndash ndash ndash ndash 082 067Alpha IV 898 ndash ndash ndash ndash ndash ndash 074 ndashTM7 1116 ndash ndash ndash ndash ndash ndash 065 ndash
a In the interest of clarity correlation coefficients are presented only for correlations that were significant at a level of P lt 0001 N = 38 Thestrongest correlations for each clade are in bold text Per Peridiniopsis Din Dinobryon Cryp Cryptomonas Gym Gymnodinium MalMallomonas P cin Peridinium cinctum P lim Peridinium limbatumb Clades were determined by the branching patterns obtained following phylogenetic tree construction and have sequence identity ge 90 Cladegroupings are listed in Fig 5 See trees (Figs 1ndash4) for freshwater clade identification
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
964 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
exhibited significant positive correlations during thissame period Over the course of the summer phy-toplankton succession nearly all of the examined AFLswith a significant correlation to a single phytoplanktonregime showed a significant positive relationship whichindicates that certain bacterioplankton favour specificphytoplankton regimes These assemblages exhibited awide range of taxonomically diverse bacteria but con-tained a large number of Gammaproteobacteria phylo-types (Table 2)
Discussion
The advent of microbial fingerprinting techniques hasallowed microbial ecologists to carry out more efficientanalysis of microbial community composition and dynam-ics during intense andor long-term sampling efforts How-ever the basis of these techniques makes it difficult toobtain truly quantitative data from multiple phylotypeswithout prohibitive effort Yet sampling at appropriate tem-poral scales may be critical for the identification of eco-logical patterns related to BCC (eg Yannarell et al2003) It is also recognized that examining solelypresencendashabsence data may significantly hinder theidentification of ecologically relevant trends in communityanalysis (Yannarell and Triplett 2005) Therefore othershave used the relative abundances of individual AFLs toexamine the change of that AFL across multiple samples(Hewson and Fuhrman 2004 Brown et al 2005Yannarell and Triplett 2005) We also considered theserelative abundance data when examining links betweenchanges in the bacterial community and other chemicalor biological parameter dynamics
Community composition
All phyla identified in Crystal Bog Lake except for the TM7phylum were previously identified in other freshwaterstudies (Zwart et al 2002) Furthermore 13 of the iden-tified clades were formerly recognized as containingmostly freshwater members Included in this freshwatergroup were the acI-B clade of Actinobacteria the CF I andCF III clades of Bacteroidetes several clades of the Beta-and Alphaproteobacteria and the FukuN18 clade of Ver-rucomicrobia (Figs 1ndash4) the majority of which were seenin all three sampling years (see Fig 5) providing addi-tional evidence to the hypothesis that members of theseclades represent a substantial cosmopolitan componentof lake bacterial communities (Zwart et al 2002)Although the majority of 16S rRNA gene sequencesretrieved from Crystal Bog Lake were affiliated with fresh-water-specific clades many were from bacteria notbelonging to recognized freshwater clades As extensivephylogenetic surveys of bacteria in freshwater are rela-tively limited some of these sequences may represent asyet unidentified freshwater-specific bacteria For examplethe soil IIndashIII clade of Actinobacteria contains sequencesobtained from bog lakes and numerous soil environments(Warnecke et al 2004) Our clone libraries contained 16SrRNA gene sequences from the soil IIndashIII clade of Actino-bacteria that formed a distinct monophyletic cluster withsequences from other humic lakes and bogs suggestingthe existence of humic lake-specific populations (Fig 1)Burkert and coworkers hypothesized that the prevalenceof Actinobacteria in humic lakes may be due to an abilityof these organisms to break down humic acid containingcompounds a trait seen in many terrestrial Actinobacte-
Fig 6 CCA biplot showing individual AFLs and their relationship to food web variables during 2002 Phytoplankton (biovolume) and nanoflagellate (abundance) explanatory variables are represented by black arrows (eigenvectors) that indicate the direction of increase for each variable The length of each arrow indicates the degree of correlation with the ordination axes Note that the Cryptomonas eigenvector is not strongly correlated with the first two ordination axes
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics 965
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
ria If the freshwater subset of the soil IIndashIII clade is trulyspecific to humic lakes as the sequence record suggeststhen it is quite possible that these organisms are special-ized to use the allochthonous humic compounds found inbog lakes Alternatively sequences that fell outside ofknown freshwater clades such as the Burkholderia-related and enteric-related sequences may representtransient lake community members more typically associ-ated with environments outside of the lake water column(eg the surrounding sphagnum mat or sediments) thatare periodically transported into the lake by rainfall eventsor wildlife activity The fairly large number of thesesequences obtained in the clone libraries indicates CrystalBog Lake may receive a large and continual flux of alloch-thonous bacteria (see below)
Several 16S rRNA gene sequences retrieved fromCrystal Bog Lake including all those from TM7 andDeltaproteobacteria were not closely (lt 92 identity)related to any other sequences in the NCBI GenBankdatabases (19 May 2005) Additionally three 16S rRNAgene sequences did not affiliate with any known phylaindicating they may represent unrecognized bacterialphyla (GenBank Accession numbers AY792312ndashAY792314) However all three of these unique 16S rRNAgene sequences were obtained only once in the clonelibraries Although extensive chimera detection methodswere employed the possibility that these sequences areartifacts of PCR reactions cannot be dismissed Additionalsequence collection and phylogenetic analyses arerequired to determine if these clones are derived fromnovel phyla
The Betaproteobacteria exhibited the greatest richnessat all OTU definitions (Table 1) The large number ofunique Betaproteobacteria taxa inhabiting freshwater maybe indicative of the diverse metabolic composition of thisgroup (Madigan et al 2002 Burkert et al 2003) whichmight allow phylogenetically similar taxa to occupy sepa-rate niches within the same physical space Alternativelythis high level of observed richness may be due to popu-lations containing multiple compositionally diverse rrnoperons (Klappenbach et al 2000 Acinas et al 2004)In freshwater mesocosm and isolation studies membersof the Betaproteobacteria were observed to respondquickly to nutrient additions (Burkert et al 2003 Hahn2003 Simek et al 2005) and are thought to be highnucleic acid containing bacteria (Simek et al 2005) Bac-teria capable of responding quickly to nutrient concentra-tion fluctuations would be predicted to contain a largernumber of rrn operons and have higher nucleic acid con-centration than slow growers or those that respond lessquickly to nutrient additions (Klappenbach et al 2000)As the internally transcribed spacer (ITS) length is notnecessarily conserved among multiple rrn operons withina single 16S rRNA phylotype the large diversity of Betap-
roteoba cteria phylotypes might actually represent asmaller number of organisms with multiple divergent rrnoperons Further investigations are required to adequatelytest these hypotheses and lie outside the scope of thisproject
BCC variation over time
Previous work illustrated the highly variable nature of lakeBCC within- and between-years (Yannarell et al 2003)The majority of Crystal Bog Lake community members(assessed by AFL) are quite dynamic yet gt70 arepresent at some time during all 3 years (Fig 5) Asdescribed above community members represented dur-ing short continuous intervals illustrated by the CB_Ba2and TM7 phylotypes and numerous AFLs within otherclades (Fig 5) may indicate the presence of transientunsustainable populations that are occasionally trans-ported into the lake (Warnecke et al 2004) Another plau-sible explanation is that these populations are notnumerous and therefore regularly fall below detection lev-els (Yannarell and Triplett 2004) In either case theobserved temporal variation in BCC suggests that manycommunity members are significantly influenced by themultiple ecological drivers known to affect these freshwa-ter communities (Nold and Zwart 1998 Crump et al2003 Kent et al 2004 Yannarell and Triplett 2005) ofwhich food web dynamics are proposed to be the domi-nant factor in this lake (Kent et al 2004)
The acI-B clade of Actinobacteria is a clear exceptionto the otherwise continual variation in BCC over time(Fig 5) Warnecke and colleagues (2004) suggest that theacI Actinobacteria clade does not constitute a transientcomponent of lake communities originating from soil in thecatchment but represents a unique pelagic freshwaterlineage capable of sustaining growth in the lake The datapresented here support their argument because it isunlikely that bacteria washing in from the surroundingcatchment would be found on nearly all 68 sampling datesacross 3 years In addition acI Actinobacteria sequenceshave been obtained in the majority of clone libraries con-structed from freshwater and are often the numericallydominant member of freshwater bacterial communities(Glockner et al 2000 Warnecke et al 2005) Taking intoaccount the prevalence of this group noted in numerousstudies and the persistence seen in this study it appearsthe acI clade of Actinobacteria possesses a significantand sustainable competitive advantage over most bacteriain the freshwater system Pernthaler and colleagues(2001) demonstrated that the small size of freshwaterActinobacteria led to decreased grazing upon this cladeby some bacterivorous protists Recently Warnecke andcoworkers demonstrated a correlation between increasedUV solar radiation and the per cent abundance of clade
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
966 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
acI which suggests another possible mechanism for theprevalence and persistence of these organisms Howeverit seems unlikely that the small cell size and possibleincreased resistance to UV solar radiation are the onlyfactors contributing to the success of this clade in fresh-water In addition to the aforementioned traits the abilityof acI to inhabit a diverse suite of lake types (representedby differences in trophic status hydrology elevationchemistry etc) suggest possible underlying factor(s) thatare unique to freshwater lakes and specifically exploitedby these organisms
Burkert and colleagues (2003) using FISH identified theBeta II clade as the numerically dominant clade in theirhumic lake samples Four 16S rRNA gene sequenceswere attributed to the Beta II clade in Crystal Bog LakeOne of these four phylotypes had an AFL of 797 basepairs for which the ITS length is the same as the culturedPolynucleobacter strains studied by Hahn and coworkers(Hahn et al 2005) The corresponding ITS lengths ofbacteria sampled from two different continents indicatethat this phylotype may have a cosmopolitan distributionin freshwater Furthermore the Beta II clade was detectedon more sampling dates than any other non-Actinobacte-rial clade (Fig 5) This degree of persistence over timesupports the conclusions of previous studies based oncoarser scales of temporal resolution that the Beta IIclade is a common pelagic resident in freshwater systems(Burkert et al 2003 Hahn 2003) and may (like the acIclade) exploit a particular niche that is intrinsic to fresh-water systems
Although Crystal Bog Lake contained many sequencesfrom the Gammaproteobacteria phylum in general thesephylotypes were detected on lt50 of the sampling datesThe extreme temporal variability within this phylum (Fig 5)suggests these bacteria may be transient communitymembers washing in from the surrounding landscape oroften present at levels below detection limits Althoughthere are few Gammaproteobacteria 16S rRNA genesequences recognized as freshwater-specific (Zwartet al 2002) members of this phylum often make up asmaller but still significant portion of the bacterioplanktoncommunity (Pernthaler et al 2004 Simek et al 2005)
Recent studies have shown that members of theBacteroidetes phylum represent a large percentage of thebacterial community in lakes especially during grazingperiods (Pernthaler et al 2004) This trend is seeminglydue to the distinctive filamentous morphology assumed bythese bacteria which significantly increases their resis-tance to grazing by protistan bacterivores The greatestnumber of AFLs that we did not identify occurred duringthe intense 2002 bacterivory period (data not shown)suggesting that our clone libraries may not have ade-quately sampled the diversity of Bacteroidetes communitymembers present on these dates Furthermore filamen-
tous bacteria were enriched during this period (Kent et al2004) If members of the filamentous LD2 Bacteroidetesclade (Pernthaler et al 2004) were highly prevalent dur-ing intense grazing periods then it is quite possible thatthese bacteria were part of the unidentified mid-spring2002 population The lack of detection of these organismsmay be due to the use of universal bacterial primers withmismatches to many members of this phylum (OSullivanet al 2004)
Food web interactions
The data collected in 2000 and 2001 indicated a correla-tion between the change in BCC and the change in dom-inant phytoplankton regime (Kent et al 2004) Howeverlittle evidence existed for a similar relationship betweenthe measured chemicalphysical parameters and BCCdynamics Although the data suggested a relationshipbetween the phytoplankton and bacterioplankton commu-nities it was also apparent that an increased samplingeffort would be needed to perceive this relationship moreaccurately Thus to examine a more relevant temporalscale for the phytoplanktonndashbacterioplankton relationshipthe 2002 samples were taken more frequently than at theprevious biweekly pace
The majority of individual community phylotype persis-tence patterns (assessed by AFL relative fluorescence)were highly correlated to the phytoplankton succession in2002 (Fig 6) Within the overall BCC pattern AFL assem-blages demonstrated unique patterns correlated to indi-vidual phytoplankton regimes (Table 2) The persistencepattern of several AFLs associated with the acI-B cladeof Actinobacteria was negatively correlated with theintense bacterivory period (Table 2) This relative reduc-tion during an intense bacterivory period may indicateeffective grazing on this clade However several studiesincluding controlled mesocosm feeding experiments haveshown that Actinobacteria which are generally very smallare less grazed upon than their freshwater counterparts(Pernthaler et al 2001 Simek et al 2005) On the otherhand these same studies showed a significant decreasein total Actinobacterial cells during increased bacterivoryA significant decrease in total cells combined with theincrease in abundance of filamentous organisms seenduring this period may be the cause of the negative cor-relation between Actinobacteria and flagellate grazersBecause no members of the acI clade of Actinobacteriahave been cultured this clades ecophysiology remainsunknown As described above Burkert and colleagues(2003) conjecture that acI clade members could possessattributes similar to the related soil Actinomycete groupwhich produces peroxidases capable of breaking downrecalcitrant compounds such as humic acids one of themost abundant carbon sources in humic lakes This
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics 967
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
hypothesis suggests that variation in Actinobacteria pop-ulations would be uncoupled to phytoplankton successionand the corresponding unique autochthonous carbonsources made available during each phytoplankton inter-val Our data support this hypothesis as the AFLs asso-ciated with the acI clade of Actinobacteria were generallynot correlated to changes in the non-mixotrophic phy-toplankton regimes (Table 2 and Fig 6)
The majority of AFL assemblages (grouped by signifi-cant correlation patterns Table 2) contained AFLs from awide variety of the represented phyla This observationmay indicate that community assemblages of taxonomi-cally diverse organisms are maintained in this lake andthat these assemblages are selected for during the differ-ent phytoplankton regimes Upon closer inspection theBeta- and Gammaproteobacteria community dynamicsshow the strongest correlation to phytoplankton succes-sion (Fig 6) suggesting that as a whole the Proteobacte-ria phylum is most closely tied to phytoplankton dynamicsMembers of this phylum seem to have the ability to growquickly during shifts in nutrient availability (Burkert et al2003) which may be brought about by phytoplankton suc-cession Thus it appears that the Actinobacteria acI-Bclade and the Betaproteobacteria Beta II clade two of themost abundant and ubiquitous freshwater phylotypesinhabit different ecological niches within the water column
Conclusions
The planktonic bacterial community of Crystal Bog Lakeconsisted of both persistent and transient populationsThese contrasting population dynamics were dividedamong different bacterial phylotypes The acI-B clade ofActinobacteria was by far the most prevalent phylotypeover the 3 year study and showed a significant negativecorrelation to the intense bacterivory period Furthermorethis clades dynamics were seemingly uncoupled to thechanges in phytoplankton regime which may indicate apreference for allochthonous nutrient sources In contrastphylotypes in the Gammaproteobacteria class exhibitedextremely variable presenceabsence patterns suggestinga transient existence in the lake Phytoplankton and grazercommunities provide lsquobottom-uprsquo and lsquotop-downrsquo pres-sures respectively which influence bacterial communitiesIn particular the Proteobacteria phylum contributedheavily to unique bacterial assemblages that wereselected for during the phytoplankton community transi-tions The increase in relative abundance of the majorityof phylotypes associated with the unique phytoplanktonregimes may indicate elevated growth rates linked to theavailability of autochthonous algal-derived nutrients forthese organisms Controlled community manipulationexperiments will be needed to further examine the driversof persistent and transient bacterial community members
as well as the phytoplankton community ndash bacterioplank-ton community relationship in humic lakes
Experimental procedures
Study sites and sample collection
Crystal Bog Lake is a shallow humic lake located in theNorthern Highlands State Forest in Vilas County Wisconsin(89deg36prime W long 46degN lat) It is part of the North TemperateLakes Long-Term Ecological Research program (Magnusonet al 1997) Detailed limnological data for this lake and sam-pling procedures have been described previously (Kent et al2004) The physicalchemical data collected for this studyincluded total chlorophyll dissolved organic carbon ammo-nia nitratenitrite total oxygen pH total nitrogen total phos-phorus total particulate matter and water temperature
Phytoplankton and HNF abundance
Phytoplankton enumeration and identification was carried outto species when possible as previously described (Kent et al2004) Heterotrophic nanoflagellate cells were stained withDAPI and counted on black 02 microm PCTE filters as previouslydescribed (Kent et al 2004)
Bacterioplankton community fingerprints
Bacterial community composition (BCC) and diversity wereassessed using ARISA (Kent et al 2004) Relativeabundance of individual phylotypes was inferred using thefluorescence of each individual peak normalized to totalfluorescence within a profile to account for run-to-run varia-tion during fragment analysis while avoiding the significantdistortion associated with presencendashabsence data transfor-mations as described previously (Yannarell and Triplett2004) and described below
Clone library construction
Clone libraries were constructed from 3 years of combinedCrystal Bog Lake DNA samples combined DNA samplesfrom 2000 and combined DNA samples from 2002 Brieflythe 16S rRNA gene and the 16Sminus23S rRNA ITS region wereamplified from pooled environmental DNA samples usingprimers 8F 5prime-AGAGTTTGATCMTGGCTCAG-3prime (bacteria-specific 16S rRNA gene) and 23SR 5prime-GGGTTBCCCCATTCRG-3prime (bacteria-specific 23S rRNA gene) PCR productswere cloned into the pGEM-T Easy vector following the man-ufacturers instructions (Promega cat A1380)
Sequence analysis
Cloned plasmid inserts were amplified directly from cells asdescribed (Vergin et al 2001) using vector primers The 16SrRNA gene portion of the cloned DNA was initially sequencedusing the ABI Prism BigDye terminator sequencing kit (PEApplied Biosystems) with standard PCR sequencing reactionconditions using the primer 8F Sequences were assigned
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
968 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
preliminary bacterial phylum associations based onthe BLASTN (Altschul et al 1990 httpwwwncbinihgovBLAST) and RDP-II Classifier programs (Cole et al 2003httprdpcmemsueduclassifierclassifierjsp) Followingclassification all sequences were aligned using the ARBsoftware package (Ludwig et al 2004) containing a publiclyavailable 16S rRNA gene ARB database January 2002(Hugenholtz 2002) supplemented with freshwater 16S rRNAgene sequences (described by Glockner et al 2000 Zwartet al 2002 Warnecke et al 2004) Actinobacteria-relatedsequences were also independently aligned Sequencesadded to the existing ARB database were initially automati-cally aligned using the FAST_ALIGNER ARB tool before thealignment was heuristically adjusted using primary and sec-ondary rRNA structure as a guide
Reference sequences were chosen for further sequencingof the 16S rRNA gene and intergenic spacer region A totalof 132 selected clones were additionally sequenced with theprimers 515F 5prime-GTGCCAGCMGCCGCGGTAA-3prime 1100F5prime-CAACGAGCGAGACCCA-3prime 1406F 5prime-TGYACACACCGCCCGT-3prime 1492R 5prime-GGTTACCTTGTTACGACTT-3prime and23SR 5prime-GGGTTBCCCCATTCRG-3prime All partial and full-length 16S rRNA sequences were edited manually andassembled using the software Sequencher 31 (Gene CodesCorporation) Forty-nine clone sequences were identified asputative chimeras by the programs CHIMERA_CHECK (httprdpcmemsuedu) or BELLEROPHON (Huber et al 2004 httpfoomathsuqeduausimhuberbellerophonpl) and were elimi-nated from further analyses Eighty-three nearly full-length(gt 1300 bp) and an additional 206 partial (gt 400 bp) 16SrRNA gene sequences and their corresponding AFLs wereacquired and used during all subsequent analyses
Linking fingerprints to phylogeny
Following amplification with vector primers of the 16S rRNAgene and 16Sminus23S rRNA ITS insert from each clone the ITSof each clone insert was amplified with primers 1406F and23SR The amplified product was then analysed using themethods described previously for community ARISA (Kentet al 2004) This procedure allowed the matching of the AFL(measured as the number of nucleotides amplified with prim-ers 1406F and 23SR) from an individual clone to the fragmentlengths obtained in the 68 ARISA community fingerprintsgenerated over 3 years (procedure recently described byBrown et al 2005) As 16S rRNA gene sequences were alsoobtained for each clone it was then possible to apply multiplehierarchical OTU definitions to each ARISA peak in the fin-gerprint
Phylogenetic reconstruction
Only nearly complete (gt 1300 bp) 16S rRNA genesequences were used for phylogenetic tree reconstruction A50 base frequency filter was calculated on the includedsequences to exclude highly variable positions An alignmentof selected Crystal Bog Lake and other reference sequenceswere exported from ARB into the MrBayes software programv 30 (Ronquist and Huelsenbeck 2003) for phylogeneticreconstruction using Bayesian inference A general time
reversible gamma-distributed rates variation model was spec-ified Three independent Markov Chain Monte Carlo analy-ses each starting with random trees for each of foursimultaneous chains were run for 100 000 generations withsampling every 10 generations to create a posterior proba-bility distribution of 10 000 trees Trees created before chainstabilization were discarded with appropriate burn-in valuesand a 50 majority-rule tree was calculated Partialsequences were added to the alignment and the MrBayesanalysis was rerun Placement of the partial sequences inMrBayes was compared with placement of the same partialsequences by the maximum parsimony tool in ARB whilepreventing changes in tree topology Partial sequences werethen appropriately added to the final trees so as not to affectfinal tree topology
All Crystal Bog Lake 16S rRNA gene sequences weregrouped into defined OTUs (clade and species) based onARB phylogeny and sequence identity determined by theprogram DOTUR (Schloss and Handelsman 2005) Specieswere identified based on a furthest neighbour 97 16S rRNAgene sequence identity threshold Clades were identifiedbased on tree topology and consistently share ge90 16SrRNA gene sequence identity
Community composition data transformations
Presence and absence analysis of ARISA profiles intro-duces a significant arbitrary bias towards rare taxa(Yannarell and Triplett 2005) Furthermore the relative fluo-rescence produced by a single ARISA peak is highly repro-ducible across PCR runs (Yannarell and Triplett 2004) andmay be used to compare samples (Hewson and Fuhrman2004) These authors concluded that significant valuableinformation is lost when the relative contribution of eachindividual peak to the total fluorescence in an ARISA profileis not taken into consideration during data analysis There-fore we used relative fluorescence produced by each indi-vidual phylotype in all analyses to infer relative abundanceof that phylotype in the original sample We do not utilize therelative fluorescence information as a method to comparethe abundance of different phylotypes but instead use itsolely to examine changes in relative contribution of a singlephylotype to the community over time A detailed discussionof sensitivity analyses conducted using the relative fluores-cence data transformation can be found elsewhere(Yannarell and Triplett 2005)
Statistical analysis
The Pearson productndashmoment correlation coefficient (r) wascalculated for each bacterial phylotype (ie unique ARISAfragment) relative fluorescence and the biovolume of thedominant phytoplankton or the abundance of HNFs in CrystalBog Lake across all sample dates in 2002 The Pearsonproductndashmoment correlation coefficient was also calculatedfor each bacterial phylotype and environmental parameters(total chlorophyll dissolved organic carbon total nitrogentotal phosphorus dissolved oxygen lake pH total particulatematter and water temperature) gathered during 2002 A P-value of lt0001 was used to establish a significant correlation
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
Freshwater bacterial community dynamics 969
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
between the variables ARISA fragments associated withmore than one clade were excluded from these analyses Thebiweekly sampling effort during 2000 and 2001 was not at asufficient frequency to accurately follow the pace of changein the bacterial community as related to the above-mentionedparameters these samples were not included in correlationanalyses
The similarity of relative fluorescence patterns from indi-vidual phylotypes was evaluated using the Bray-Curtis Indexof Similarity (Magurran 1988) Phylotype assemblages wereassigned according to their correlation with unique phy-toplankton regimes The largest r-value indicating the stron-gest correlation was used to determine the assemblageassignment for those phylotypes with more than one signifi-cant phytoplankton regime correlation (Table 2) An analysisof similarity (ANOSIM) was carried out using rank dissimilari-ties to test the hypothesis that the relative fluorescence pat-terns within the assigned assemblages were more similarthan the relative fluorescence patterns between assem-blages (Clarke 1993) These multivariate analyses (basedon non-metric multidimensional scaling) were performedusing the statistical package PRIMER 5 for Windows v 527
ARISA fragment length relative fluorescence data weresubjected to CCA using phytoplankton biovolume and HNFabundance as potential explanatory variables Only thoseAFLs exhibiting a significant correlation (P lt 0001) to aphytoplankton regime (listed in Table 2) were included inthe analysis Canonical correspondence analysis repre-sents individual AFLs as occurring in a theoretical environ-mental space or ordination space which is defined by thepotential explanatory variables included in the analysisThe CCA axes represent linear combinations of the protistpopulation data included in the analysis This approachallows us to explore relationships between bacterial popu-lations and food web factors that may influence the dynam-ics of bacterial populations of interest The significance ofthe set of explanatory variables was tested by a MonteCarlo permutation test (999 permutations) The position ofeach AFL in ordination space represents its correlationwith the explanatory variables AFLs plotting close to anarrow representing an explanatory variable are stronglycorrelated with that variable and populations which plotclose together in ordination space have similar responsesto the food web factors included in the analysis Canonicalcorrespondence analysis was carried out using Canoco forWindows Version 451 (Biometris-Plant Research Interna-tional 1997ndash2003) Analysis settings included biplot scal-ing focusing on interspecies distances and permutationtests which accounted for the time-series structure of thedata
Nucleotide sequences and ARISA profiles
All 16S rRNA gene sequences generated in the current studythat were included in the phylogenetic analyses have beensubmitted to GenBank under Accession numbers AY792221to AY792314 and DQ093399 to DQ093408
The ARISA profiles and ancillary environmental data usedto generate Figs 5 and 6 are available in downloadable formin our online database (httpmicrobeslimnologywiscedu)(Jacob et al 2005)
Acknowledgements
The authors wish to thank Anthony Yannarell and StuartJones for helpful discussions A sequence alignment of fresh-water Actinobacteria 16S rRNA genes was kindly providedby Falk Warnecke and Jakob Pernthaler Thanks to MartinHahn for unpublished information about the 16Sminus23S rRNAintergenic spacer region in Polynucleobacter isolates Thisresearch was supported in part by National Science Founda-tion Grant MCB-9977903 and MCB-0401987 to EWT and bythe National Institutes of Health Biotechnology Training Pro-gram Grant 5 T32 G08349 (RJN)
References
Acinas SG Marcelino LA Klepac-Ceraj V and PolzMF (2004) Divergence and redundancy of 16S rRNAsequences in genomes with multiple rrn operons J Bacte-riol 186 2629ndash2635
Altschul SF Gish W Miller W Myers EW and LipmanDJ (1990) Basic local alignment search tool J Mol Biol215 403ndash410
Brown MV Schwalbach MS Hewson I and FuhrmanJA (2005) Coupling 16S-ITS rDNA clone libraries andautomated ribosomal intergenic spacer analysis to showmarine microbial diversity development and application toa time series Environ Microbiol 7 1466ndash1479
Burkert U Warnecke F Babenzien D Zwirnmann Eand Pernthaler J (2003) Members of a readily enrichedbeta-proteobacterial clade are common in surfacewaters of a humic lake Appl Environ Microbiol 696550ndash6559
Chao A (1987) Estimating the population-size for capturerecapture data with unequal catchability Biometrics 43783ndash791
Clarke KR (1993) Nonparametric multivariate analyses ofchanges in community structure Aust J Ecol 18 117ndash143
Clayton MK and Frees EW (1987) Nonparametric-esti-mation of the probability of discovering a New Species JAm Stat Assoc 82 305ndash311
Cole JR Chai B Marsh TL Farris RJ Wang QKulam SA et al (2003) The Ribosomal Database Project(RDP-II) previewing a new autoaligner that allows regularupdates and the new prokaryotic taxonomy Nucleic AcidsRes 31 442ndash443
Cotner JB and Biddanda BA (2002) Small players largerole microbial influence on biogeochemical processes inpelagic aquatic ecosystems Ecosystems 5 105ndash121
Crump BC Kling GW Bahr M and Hobbie JE (2003)Bacterioplankton community shifts in an arctic lake corre-late with seasonal changes in organic matter source ApplEnviron Microbiol 69 2253ndash2268
Eiler A and Bertilsson S (2004) Composition of freshwaterbacterial communities associated with cyanobacterialblooms in four Swedish lakes Environ Microbiol 6 1228ndash1243
Glockner FO Zaichikov E Belkova N Denissova LPernthaler J Pernthaler A and Amann R (2000)Comparative 16S rRNA analysis of lake bacterioplanktonreveals globally distributed phylogenetic clusters including
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
970 R J Newton A D Kent E W Triplett and K D McMahon
copy 2006 Society for Applied Microbiology and Blackwell Publishing Ltd Environmental Microbiology 8 956ndash970
an abundant group of Actinobacteria Appl Environ Micro-biol 66 5053ndash5065
Good IL (1953) The population frequencies of species andthe estimation of population parameters Biometrika 40237ndash264
Hahn MW (2003) Isolation of strains belonging to thecosmopolitan Polynucleobacter necessarius cluster fromfreshwater habitats located in three climatic zones ApplEnviron Microbiol 69 5248ndash5254
Hahn MW Pockl M and Wu QL (2005) Low intraspecificdiversity in a Polynucleobacter subcluster populationnumerically dominating bacterioplankton of a freshwaterpond Appl Environ Microbiol 71 4539ndash4547
Hewson I and Fuhrman JA (2004) Richness and diversityof bacterioplankton species along an estuarine gradient inMoreton Bay Australia Appl Environ Microbiol 70 3425ndash3433
Huber T Faulkner G and Hugenholtz P (2004) Bellero-phon a program to detect chimeric sequences in multiplesequence alignments Bioinformatics 10 1093
Hugenholtz P (2002) Exploring prokaryotic diversity in thegenomic era Genome Biol 3 1ndash8
Jacob C Kent AD Benson BJ Newton RJ TriplettEW and McMahon KD (2005) Biological databases forlinking large microbial and environmental datasets Pro-ceedings of the 9th World Multiconference on SystematicsCybernetics and Informatics Orlando FL USA
Kemp PF and Aller JY (2004) Bacterial diversity inaquatic and other environments what 16S rDNA librariescan tell us FEMS Microbiol Ecol 47 161ndash177
Kent AD Jones SE Yannarell AC Graham JMLauster GH Kratz TK and Triplett EW (2004) Annualpatterns in bacterioplankton community variability in ahumic lake Microb Ecol 48 550ndash560
Klappenbach JA Dunbar JM and Schmidt TM (2000)rRNA operon copy number reflects ecological strategies ofbacteria Appl Environ Microbiol 66 1328ndash1333
Lindstrom ES (2000) Bacterioplankton community compo-sition in five lakes differing in trophic status and humiccontent Microb Ecol 40 104ndash113
Ludwig W Strunk O Westram R Richter L Meier HYadhukumar et al (2004) ARB a software environmentfor sequence data Nucleic Acids Res 32 1363ndash1371
Madigan MT Martinko JM and Parker J (2002) BrockBiology of Microorganisms Upper Saddle River NJ USAPrentice HallPearson Education
Magnuson JJ Kratz TK Allen TF Armstrong DEBenson BJ Bowser CJ et al (1997) Regionalization oflong-term ecological research (LTER) on north temperatelakes Verh Int Verein Limnol 26 522ndash528
Magurran AE (1988) Ecological Diversity and its Measure-ment Princeton NJ USA Princeton University Press
Methe BA and Zehr JP (1999) Diversity of bacterial com-munities in Adirondack lakes do species assemblagesreflect lake water chemistry Hydrobiologia 401 77ndash96
Nold SC and Zwart G (1998) Patterns and governingforces in aquatic microbial communities Aquat Ecol 3217ndash35
OrsquoSullivan LA Fuller KE Thomas EM Turley CMFry CJ and Weightman AJ (2004) Distribution and
culturability of uncultivated lsquoAGG58 clusterrsquo of theBacteroidetes phylum in aquatic environments FEMSMicrobiol Ecol 47 359ndash370
Pernthaler J Posch T Simek K Vrba J Pernthaler AGlockner FO et al (2001) Predator-specific enrichmentof Actinobacteria from a cosmopolitan freshwater clade inmixed continuous culture Appl Environ Microbiol 672145ndash2155
Pernthaler J Zollner E Warnecke F and Jurgens K(2004) Bloom of filamentous bacteria in a mesotrophiclake identity and potential controlling mechanism ApplEnviron Microbiol 70 6272ndash6281
Pinhassi J Sala MM Havskum H Peters F GuadayolO Malits A and Marrase CL (2004) Changes in bac-terioplankton composition under different phytoplanktonregimens Appl Environ Microbiol 70 6753ndash6766
Ronquist F and Huelsenbeck JP (2003) MrBayes 3 baye-sian phylogenetic inference under mixed models Bioinfor-matics 19 1572ndash1574
Rooney-Varga JN Giewat MW Savin MC Sood SLeGresley M and Martin JL (2005) Links between phy-toplankton and bacterial community dynamics in a coastalmarine environment Microb Ecol 49 163ndash175
Schloss PD and Handelsman J (2005) IntroducingDOTUR a computer program for defining operational tax-onomic units and estimating species richness Appl Envi-ron Microbiol 71 1501ndash1506
Simek K Hornak K Jezbera J Masin M Nedoma JGasol J and Schauer M (2005) Influence of top-downand bottom-up manipulations on the R-BT065 subclusterof β-Proteobacteria an abundant group in bacterioplanktonof a freshwater reservior Appl Environ Microbiol 71 2381ndash2390
Vergin KL Rappe MS and Giovannoni SJ (2001)Streamlined method to analyze 16S rRNA gene clonelibraries Biotechniques 30 938ndash944
Warnecke F Amann R and Pernthaler J (2004) Actino-bacterial 16S rRNA genes from freshwater habitats clusterin four distinct lineages Environ Microbiol 6 242ndash253
Warnecke F Sommaruga R Seka R Hofer J and Pern-thaler J (2005) Abundances identity and growth state ofActinobacteria in mountain lakes of different UV transpar-ency Appl Environ Microbiol 71 5551ndash5559
Wetzel RG (2001) Limnology Lake and River EcosystemsSan Diego CA USA Academic Press
Yannarell AC and Triplett EW (2004) Within- andbetween-lake variability in the composition of bacteri-oplankton communities investigations using multiple spa-tial scales Appl Environ Microbiol 70 214ndash223
Yannarell AC and Triplett EW (2005) Geographic andenvironmental sources of variation in lake bacterial com-munity composition Appl Environ Microbiol 71 227ndash239
Yannarell AC Kent AD Lauster GH Kratz TK andTriplett EW (2003) Temporal patterns in bacterial com-munities in three temperate lakes of different trophic statusMicrob Ecol 46 391ndash405
Zwart G Crump BC Agterveld MPKV Hagen F andHan SK (2002) Typical freshwater bacteria an analysisof available 16S rRNA gene sequences from plankton oflakes and rivers Aquat Microb Ecol 28 141ndash155
