AQUATIC MICROBIAL ECOLOGYAquat Microb Ecol

Vol. 82: 225–239, 2019https://doi.org/10.3354/ame01891

Published online January 21

INTRODUCTION

Permanently ice-covered lakes in the McMurdo DryValleys (MDV) of South Victoria Land, Antarctica,have closed basins and extremely stable strata thatvary physically, chemically, and biologically withinand between stratified water columns. The ice-covers(3−4 m in thickness) restrict water column mixing andexternal inputs. Ephemeral glacial melt streams,which flow for only a few weeks in the australsummer, are the primary mechanism of water and nu-

trient input to the lakes (Lyons et al. 2000). This studyfocused on 3 MDV lakes (Lake Bonney, Lake Fryxelland Lake Vanda), which have been sites of a long-term study since 1993 (www. mcm. lter. org). All 3 lakesare light-limited and either oligo trophic (Lakes Bon-ney and Vanda) or mesotrophic (Lake Fryxell) (Priscu1995). Low light and limited nutrients combined withyear-round low temperatures and a lack of sunlight inthe winter greatly limit yearly production. The foodwebs in MDV lakes are microbially dominated, andlack crustacean zooplankton and fish.

© Inter-Research 2019 · www.int-res.com*Corresponding author: wei.li02@montana.edu

Drivers of protistan community autotrophy and heterotrophy in chemically stratified Antarctic lakes

Wei Li1,*, Jenna Dolhi-Binder2, Zev E. Cariani3, Rachael M. Morgan-Kiss3

1Department of Land Resources and Environmental Sciences, Montana State University, 334 Leon Johnson Hall, Bozeman, Montana 59717, USA

2Department of Biology and Toxicology, Ashland University, Ashland, Ohio 44805, USA3Department of Microbiology, Miami University, Oxford, Ohio 45056, USA

ABSTRACT: Single-celled, eukaryotic microorganisms, known as protists, are responsible for 2important, yet opposing, metabolic activities within aquatic food webs. They are major primaryproducers and highly active predators in marine and fresh water systems. While genomics hasaccelerated in recent years for this taxonomically diverse group, our understanding of the meta-bolic capabilities of most protists remains limited. It is also poorly understood how protist trophicmode is affected by biotic and abiotic factors, and therefore it is difficult to predict how eventssuch as global climate change will affect the balance between autotrophic and heterotrophicactivities in protist communities. To address open questions regarding how protist metabolic versatility is influenced by their environment, we characterized the potential for carbon fixationversus organic carbon degradation using enzymatic assays (RubisCO and β-D-glucosaminidase,respectively) within the water columns of ice-covered lakes in McMurdo Dry Valleys (MDV),Antarctica. Steep physical and chemical gradients in the water columns, microorganism domina-tion and minimal allochthonous inputs makes the MDV lakes uniquely suited to investigate envi-ronment−microbe interactions. Spatial trends in RubisCO and β-D-glucosaminidase activitieswere lake-specific and vertically stratified within the water columns. Moreover, bottom-up driverscontrolling the activity of C-fixation vs. organic C-degradation among the MDV protist communi-ties were distinct between the upper photic vs. the deep, aphotic zones. We conclude that differ-ential controls over major C-cycling enzymes have important implications on the influence ofenvironmental change on the carbon and nutrient cycles in the MDV lakes.

KEY WORDS: Aquatic protists · Autotrophy · Heterotrophy · β-D-glucosaminidase · RubisCO ·McMurdo Dry Valleys · Antarctic lakes

Resale or republication not permitted without written consent of the publisher

Aquat Microb Ecol 82: 225–239, 2019

Microbial eukaryotes (or protists) play dual roles inMDV food webs as primary producers and top preda-tors (i.e. consumers of bacteria and small phytoplank-ton) (Priscu et al. 1999, Bielewicz et al. 2011). Recentapplications of 18S rRNA sequencing provided thefirst insight into the diversity and function of the mi-crobial eukaryote populations (Bielewicz et al. 2011,Vick-Majors et al. 2014, Rojas-Jimenez et al. 2017).Spatial and seasonal trends in MDV protist communitystructure are complex and exhibit strong variabilityamong lakes (Bielewicz et al. 2011, Kong et al. 2012b,2014, Vick-Majors et al. 2014, Dolhi et al. 2015). LakeBonney is dominated by chlorophytes Chlamydo mo -nas in the shallow layers, and haptophytes Isochrysisand stramenopiles Nannochloropsis in the chemo-cline. Lake Fryxell is dominated by cryptophytesGeminigera (Bielewicz et al. 2011, Kong et al. 2012b,Dolhi et al. 2015) and prasinophytes Pyramimonas(Vincent 1981, Roberts et al. 2000). The least produc-tive lake, Lake Vanda, harbors nanoplankton relatedto non-flagellated chlorophytes Nannochloris andstramenopiles Heterococcus (Dolhi et al. 2015).

Heterotrophic grazing by protists was measured inseveral MDV lakes (Roberts & Laybourn-Parry 1999,Marshall & Laybourn-Parry 2002, Thurman et al.2012), suggesting that heterotrophic protists mightplay important roles in top-down control of phyto-plankton and bacterioplankton populations. Mixo -trophy is widespread in ice-covered Antarctic lakes,allowing exploitation of alternative nutrient and en-ergy sources (Nygaard & Tobiesen 1993, Laybourn-Parry et al. 2005, Li et al. 2016). Recent studies in LakeBonney and Lake Fryxell supported these observa-tions with phylogenetic evidence of heterotrophicnanoflagellates (e.g. choanoflagellates and heterotro-phic chrysophytes) (Bielewicz et al. 2011, Vick-Majorset al. 2014). Recent reports also found evidence for theimportant role of Cryptomycota and Chytridiomycotafungi (Rojas-Jimenez et al. 2017) and heterotrophicnanoflagellates (Li et al. 2016) as parasites and preda-tors in the food webs in MDV lakes.

Lakes and reservoirs cover the globe, forming anetwork of environmental sensors which can beexploited as sentinels of climate change (Williamsonet al. 2009). Seasonally and perennially ice-covered,permanently low-temperature lakes are highly pre -valent in polar and alpine environments. Unlike tem-perate ecosystems, the MDV lakes are relatively lowtemperature ecosystems which have experiencedlow human impact, representing end members of theglobal lake network (Morgan-Kiss & Dolhi 2012).Relative to lower latitude watersheds, Antarctic, Arc-tic and high alpine aquatic habitats such as the

Tibetan Plateau are more responsive to both climatewarming and increased episodic events; however,physiological responses of polar microbial communi-ties are also influenced by local environmental condi-tions (Morgan-Kiss & Dolhi 2012). Thus, predictingthe impact of climate change on functional diversityof the polar lake protist communities is complex.

Several studies have shown that trophic modes ofprotist communities are sensitive to biotic and abioticfactors. Environmental drivers, including solar radia-tion (Batistas Navarro et al. 2011, Izaguirre et al.2012), temperature (O’Connor et al. 2009, Wohlers etal. 2009), nutrients and other physicochemical condi-tions (Brown et al. 2000, Phillips et al. 2008) can havestrong effects on protist activities and therefore theirroles in the aquatic carbon and nutrient cycles. Theimpact of protist trophic mode is likely amplified instratified, microbial-driven systems such as the MDVlakes; however, an understanding of the relationshipbetween environmental factors and protist trophicactivities in the MDV lakes is currently lacking.Improved quantification of trophic activities for MDVprotists would contribute to a better understanding ofpolar lake food webs, building more reliable and pre-dictive models on plankton community interactions,carbon and nutrient fluxes, and the sensitivity ofthese to environmental change.

Most photosynthetic protists utilize the Calvin-Benson-Bassham (CBB) cycle for autotrophic car-bon fixation. The enzyme Ribulose-1,5-bisphos-phate carboxylase/ oxygenase (RubisCO) catalyzesthe first and rate-limiting reaction. The radiometric14C RubisCO carboxylase activity assay has con-ventionally been used as a measure of maximumpotential of autotrophic carbon fixation and hasbeen adapted for use in natural phytoplanktoncommunity samples (Tortell et al. 2006, Dolhi et al.2012, Kong et al. 2012a).

In phagotrophic and pinocytic protist food uptake,prey cells and other food particles are sequestered inacidic food vacuoles where digestive enzymes suchas lysozymes and proteases degrade complex or ga -nic compounds into simpler substrates for subse-quent heterotrophic metabolism (Vrba et al. 1996,Verni & Gualtieri 1997). While several methods areavailable for estimating protistan grazing rates usingfluorescently labeled bacterial or artificial prey(Sherr & Sherr 1993, Sherr et al. 1997, Smalley et al.1999), quantifying protist heterotrophic activity innatural environments is challenging. Accurate esti-mates can be confounded by selective grazing of thepredators and/or spatiotemporal shifts in prey abun-dance. Low productivity systems such as the MDV

226

Li et al.: Antarctic lake planktonic community trophic activities

lakes are also near the limit of detection. As an alter-native approach, planktonic phagotrophic activitycan be quantified based on specific lysozyme activityassociated with β-D-glucosaminidase (βGAM) enzy-matic activity (Zub kov & Sleigh 1998, Štrojsová &Dyhrman 2008, Sintes & del Giorgio 2010, 2014). Theoptimal pH condition for protist vacuolar βGAM islower (pH <5) compared to the exoenzymes of copi-otrophic bacteria (pH ≥7) (Vrba et al. 1996, Arnosti2011).

In the microbially dominated MDV lakes, protistancarbon fixation and phagotrophic activity are criticalto carbon and nutrient cycling. Several studies havereported on the distribution and activity of mixo -trophic and heterotrophic eukaryotes in these lakesbased on artificial prey feeding experiment or microscopic numeration (Laybourn-Parry et al. 1995,James et al. 1998, Roberts & Laybourn-Parry 1999,Roberts et al. 2004a, Thurman et al. 2012), but directrate measurements of these 2 critical processes in theDry Valley lakes are currently lacking. We hypothe-size that (1) the trophic activities of protists are strat-ified in the MDV lake water columns due to highlystratified environmental conditions; (2) these activi-ties are correlated with specific environmental driv-ers. In this study, we investigated autotroph and pro-tist-specific heterotroph distribution using RubisCO

and acid-βGAM enzyme activities, respectively, inthe physically and chemically distinct MDV lakesand correlated activity with specific biotic and abioticfactors.

MATERIALS AND METHODS

Field sampling and limnological parameters

Locations of study sites within the Taylor (LakesBonney and Fryxell) and Wright (Lake Vanda) Val-leys of the MDV, Antarctica, are shown in Fig. 1.Water column samples were collected during theaustral summer (November− December) of the 2012field season with the exception of samples for LakeVanda RubisCO activity, which were collected inDecember 2011. All sampling depths were measuredfrom the piezometric water level in the ice hole usinga depth-calibrated hand winch. Water samples werecollected with a 5 l Niskin bottle and stored underlow temperature and dark conditions until process-ing. Water samples for RubisCO carboxylase activity(2−5 l) and βGAM activity (0.75−2 l) were vacuum fil-tered (0.3 mbar) onto 47 mm GF/C filters (CAS#:1822-047, Whatman) and hydrophilic nylon mem-brane filters (CAS#: 7404-004, Whatman) with a

227

Fig. 1. Map showing locations of study sites within the McMurdo Dry Valleys, Antarctica (see inset): Lake Bonney east (ELB)and west (WLB) lobes, Lake Fryxell (FRX) and Lake Vanda (VAN). Image was edited from a LIMA based map from the Polar

Geospatial Center (www.agic.umn.edu/maps/Antarctic)

Aquat Microb Ecol 82: 225–239, 2019

0.45 µm pore size, re spectively. Filters were immedi-ately flash frozen in liquid N2 and shipped to our USlaboratory on dry ice. Samples were stored at −80°Cfor less than 2 mo before determining enzyme activ-ity. Prior to the field season, both enzymatic assayswere optimized. Fresh versus frozen mock filterswere also compared to assess the effect of freezingon enzymatic activity.

Photosynthetically active radiation (PAR), temp -erature, chlorophyll a (chl a), primary productivity(PPR), and nutrient (NH4

+, NO3− and soluble reactive

phos phorus [SRP]) concentrations were determinedthrough the water columns of the east and west lobesof Lake Bonney (ELB and WLB, respectively), LakeFryxell, and Lake Vanda. PAR was measured witha LI-193 spherical quantum sensor (LI-COR Bio-sciences). Temperature was measured with a Sea-bird model 25 profiler (Spigel & Priscu 1998). Chl aconcentrations were determined using a profilingspectrofluorometer (FluoroProbe, BBE Mol daenke)(Carraro et al. 2012). The abundance of major phyto -plankton groups including green algae (Chlorophytaand Euglenophyta), Cryptophyta, Haptophyta, Chry -sophyceae and Cyanobacteria were discriminatedusing the FluoroProbe based on their specific lightharvesting pigment fluorescence (Beutler et al.2002). This method has been routinely applied intoother studies in these lakes (Winslow et al. 2014,Dolhi et al. 2015). Light-mediated PPR was deter-mined by measuring NaH14CO3 incorporation induplicate over a 24 h in situ incubation. Nutrientswere measured as part of the NSF-funded McMurdolong-term ecological research (MCM-LTER) pro-gram according to the methodology outlined in theMCM-LTER limnological manual (www.mcm. lter.org). Briefly, inorganic nitrogen species were deter-mined with a Lachat autoanalyzer and SRP wasmeasured manually using the antimony-molybdatemethod (Strickland & Parsons 1972).

RubisCO carboxylase activity assay

GF/C filters were thawed on ice and cut in half. Allsubsequent extraction steps were carried out on ice.Each cut half filter was transferred to a 1.5 ml screwcap tube containing 0.1 mm zirconia/silica beads and650 µl ice cold RubisCO extraction buffer (50 mMbicine, 10 mM MgCl2, 1 mM EDTA, 5 mg ml−1 bovineserum albumin [BSA], and 0.1% triton X-100; addfreshly made: 20 mM NaHCO3, 10 mM DTT, 0.1 mgml−1, pH 7.8). Filters were disrupted with a Minibead-Beater-16 (Biospec Products) for 3 cycles of 30 s at

speed setting 48 with alter nating 1 min on ice incu-bations. The lysate was transferred to a cold 1.5 mlmicrocentrifuge tube and cleared by centrifugationat 4°C for 2 min at 15 000 × g. The soluble lysate wasused in RubisCO carboxylase assays.

RubisCO carboxylase activity was measured in thesoluble cell lysates (200 µl) within 30 min of extrac-tion. Soluble cell lysates were incubated at 25°C for1 min before RubisCO activity was measured using amodified NaH14CO3 assay from Tortell et al. (2006) asdescribed previously (Dolhi et al. 2012, Kong et al.2012a). Briefly, RubisCO was activated with MgCl2(8 mM) and NaH14CO3 (ViTrax, Placentia; specificactivity in final reaction: 0.03 µCi mmol−1) for 5 min at25°C. The carboxylase assay was initiated by addi-tion of 20 µl of the substrate, Ribulose-1,5-bisphos-phate (RuBP; CAS #: 24218-00-6, Sigma-Aldrich;15 mM) and the reaction was allowed to proceed for5 min at 25°C. The reaction was stopped with 100 µlof 100% propionic acid and unincorporated 14C wasexhausted by centrifugation in a fume hood for 1.25 hat 2000 × g. Soluble cell lysate with no RuBP sub-strate added was used in the assay as a negative control. A multipurpose scintillation counter LS6500(Beckman Coulter) was used to quantify acid stableend products. RubisCO carboxylase activity in theLake Bonney psychrophillic alga Chlamydomonassp. UWO241 was previously shown to be stable andoptimal at 25°C (Morgan-Kiss & Dolhi 2012).

Glucosaminidase activity assay

Acidic βGAM activity was measured as particulateenzyme activity retained on filters. Nylon filters werethawed on ice and extraction was carried out on ice.Filters were disrupted by 3 cycles of bead beating in2 ml of ice cold extraction buffer (0.1 M acetate, 0.1%23 lauryl ether [Brig 35], pH 4.6 with glacial aceticacid). βGAM activity was measured in soluble ly -sates (250 µl) using a β-N-Acetyl-glucosaminidaseAssay Kit (CAS#: 9012-33-3, Sigma-Aldrich). Celllysates were incubated with the fluorogenic sub-strate 4-methylumbelliferyl-n-acetyl-α-D-gluco sa mi -nide (4-MUF) in the dark at 20°C for 4 h. Nano-purewater and purified β-N-Acetyl-glucosaminidase fromJack beans (standard with assay kit) were used asnegative and positive controls, respectively. Fluores-cence intensity was quantified in a Lambda-35 spec-trofluorometer (Perkin-Elmer). A standard curve wascreated with known amounts of the product of theenzymatic reaction, 4-MUF (Vrba et al. 1993, Zubkov& Sleigh 1998).

228

Li et al.: Antarctic lake planktonic community trophic activities

βGAM activity was also measured in 3 psychro -philic phytoplankton strains, Chlamydomonas sp.ICE-MDV, Isochrysis sp. MDV and Geminigera cryo -phila to test for temperature sensitivity and speci-ficity of the βGAM activity. Chlamydomonas sp. ICE-MDV and Isochrysis sp. MDV were isolated fromLake Bonney (Li et al. 2016, Raymond & Morgan-Kiss2017), while G. cryo phila (CCMP 2564) was isolatedfrom Ross Sea near McMurdo Sound. All 3 algaewere grown in either autotrophic (F/2 or L1) or het-erotrophic media (F/2 supplemented with 5% cerealgrass). βGAM activity was measured in the solublelysate across several incubation temperatures in ahomemade temperature gradient apparatus.

In order to select filter material with optimal en -zyme recovering capability, various filter types (glassfiber, polyethersulfone, and nylon) were also testedusing water samples from a local lake (Acton Lake,OH). The test results indicated that sample extractionfrom nylon filters produced the highest enzyme yield(data not shown).

Protein concentration determination

Protein concentration was determined in the solu-ble cell lysate extracted for RubisCO and βGAMassays according to the Bradford method using BSA(CAS #: 9048-46-8, Sigma-Aldrich) as a standard(Bradford 1976).

Bacterial enumeration

Quantitation of free-living bacteria in the lakewater columns was carried out using a method modi-fied from Lebaron et al. (1994). Lake water (2 ml) fromeach depth was fixed with paraformaldehyde at afinal concentration of 2% v/v for 30 min. Fixed samples were filtered on 25 mm black polycarbonatemembrane with a pore size of 0.2 µm. Bacteria cellswere stained with a fluorochrome (4’,6-diamidino-2-phenylindole, [DAPI], CAS #: 28718-90-3, Sigma-Aldrich) followed by epifluorescent microscopic enu-meration. Filters were examined using an OlympusAX-70 Multi-mode System with a specific filter set(EX 360/40 nm, EM 460/50 nm) for DAPI staining, anddigital images of at least 15 random views were taken.Images were then calibrated, and bacterial countsfrom each image were recorded using ImageJ soft-ware (V1.47, National Institutes of Health). The bacte-rial concentration of each depth was determined byusing the average bacterial counts of each image.

Statistical analyses

Abiotic parameter data including PAR, tempera-ture, NH4

+, NO3−, SRP−, dissolved inorganic N:P ratio

(N:P), dissolved inorganic carbon (DIC), and conduc-tivity were normalized using z-scores. Principal com-ponent analysis (PCA) was carried out in PAST soft-ware (V3.07, Hammer et al. 2001). All other statisticalanalyses were performed using JMP Pro 12 software(SAS Institute). Data of physical (except tempera-ture), chemical, and biological parameters were log-transformed, and the correlation between theseparameters and enzyme activities were determinedusing Pearson correlation analysis. In order to iden-tify linear models of enzyme activities (respondents)with abiotic and biotic factors (variables), we gener-ated all possible linear models and selected the mod-els with lowest Akaike Information Criterion (AIC)value to obtain the models with the least number ofparameters and the maximum explanatory power(Akaike 1974, Sakamoto et al. 1986).

RESULTS

Optimization of enzyme assays

We have previously reported on optimization of theRubisCO enzyme assay in natural samples from LakeBonney (Dolhi et al. 2012, Kong et al. 2012a) andtested the temperature range of RubisCO activity in aLake Bonney isolate (Chlamydomonas sp. UWO241)(Morgan-Kiss & Dolhi 2012). This latter study showedthat despite the fact that Chlamydomonas sp. UWO241is a psychrophile (Morgan et al. 1998), RubisCO isnot cold-adapted and exhibits maximum activity atassay temperatures ≥20°C (Morgan-Kiss & Dolhi 2012).

Since the βGAM assay has not been previouslyapplied to protist communities of the MDV lakes, wetested βGAM activity in 3 psychrophilic, polar algalisolates: 2 from Lake Bonney (Chlamydomonas sp.ICE-MDV and Isochrysis sp. MDV), and a third froma polar marine environment (Geminigera cryophila).Chlamydomonas sp. ICE-MDV and Isochrysis sp.MDV are dominant phytoplankton of Lake Bonney,while a cryptophyte closely related to the marine G.cryo phila dominates the water column in Lake Fryx-ell (Bielewicz et al. 2011, Kong et al. 2012b, Dolhi etal. 2015). In addition, previous studies have reportedthat Isochrysis sp. MDV and G. cryo phila are mixo-trophic, while Chlamydomonas sp. ICE-MDV is anobligate photoautotroph (Gast et al. 2014, Li et al.2016). We compared βGAM across these 3 psy-

229

Aquat Microb Ecol 82: 225–239, 2019

chrophile isolates grown under autotrophic vs. het-erotrophic conditions. As expected βGAM activitywas below detectable levels in the pure photoauto-troph, Chlamydomonas sp. ICE-MDV (Table S1 in theSupplement at www.int-res.com/articles/suppl/ a082p225 _ supp. pdf). In contrast, βGAM activity was de -tectable in the mixotrophs Isochrysis sp. MDV and G.cryophila, and was maximal when both strains weregrown in the presence of cereal grass (Table S1).

We also tested the effect of assay temperature onβGAM activity in Isochrysis sp. MDV and G. cryo -phila (Fig. S1 in the Supplement). Over a range ofincubation temperatures βGAM activity of bothpsychro philes increased slightly with increasingtempe rature; although, no significant difference wasob served between 10 to 20°C (Fig. S1). Therefore,we concluded that similar to RubisCO, βGAM activ-ity in the psychrophiles does not cold-adapt. Thus,since the temperature in the MDV lakes can widelyvary (from −6 to over 20°C), we chose 20°C as thetemperature for enzyme assays on environmentalsamples.

Lake Bonney

Lake Bonney is separated into 2 lobes, east andwest. The water columns of each lobe remain iso-lated from each other for most of the year, with theexception of a narrow passage which connects theupper photic zones for a few weeks during theaustral summer. Thus, ELB and WLB share somecommon characteristics, including strong thermo-haline stratification (Fig. 2) and phosphorus limita-tion (Fig. 3A,C); however, the deeper layers ofeach lobe are permanently separated. For bothlobes, maximum PAR values were at or below20 µmol photons m−2 s−1 (Fig. 3B,D). Temperaturesin the shallow depths and the chemoclines rangedbetween 2 and 4°C in both lobes, but declined tobelow 0°C in the monimolimnion of WLB (Fig. 3B,D).Both lobes exhibit SRP levels near the limit ofdetection throughout the water columns and earlystudies reported that phytoplankton and bacterialcommunities are limited by phosphorus (Priscu1995, Ward et al. 2003). Nitrogen (NH4

+, NO3−)

was low in the surface waters of both lobes.Spatial trends in several chemical variables aredistinct in the monimolimnia below the chemo-clines, in particular salinity (Fig. 2) and concen -tration of several inorganic nitrogen species(Fig. 3A,C). For example, ELB exhibits high levelsof NO3

− below the chemocline, suggesting a lack

of significant rates of denitrification (Ward et al.2005). Chl a maxima were present in the shallowand deep layers of both lobes of Lake Bonney; thedeep chlorophyll maximum (DCM) was well de -fined in WLB, reaching a maximum chl a concen-tration of 15 µg l−1 (Fig. 4A,B). Spatial trends in bac-terial biomass were uncoupled from phytoplanktonin WLB. Maximum bacterial biomass occurredabove the chemocline in both lobes, but peaked atdifferent depths (5 m for ELB and 10 m for WLB;Fig. 4A,B).

We used RubisCO and βGAM enzyme activitiesto understand how spatial variability in physio -chemical and biological parameters impacts relativecontributions of autotrophic and heterotrophic meta -bolism, respectively, in the food webs of MDV lakes.Enzyme activity was expressed on the basis ofeither total extracted protein (Fig. 5A−D) or volumeof filtered lake water (Fig. 5E−H). Activity for bothenzymes was higher in ELB relative to WLB. On aprotein basis, RubisCO activity peaked at 13 m inboth lobes. Spatial trends in βGAM activity weredistinct from RubisCO activity, indicating that theimportance of the 2 metabolic processes is variableat different layers within the lakes. Both lobes ex -hibited a peak in βGAM activity in shallow depths(5 m); however, WLB also exhibited a dis tinct peakin the chemocline (15 m depth). Spatial trends inβGAM were comparable on a protein or lake waterbasis; however, trends in RubisCO activity were

230

Fig. 2. Salinity profiles for Lake Bonney (east lobe, ELB; westlobe, WLB), Lake Fryxell (FRX) and Lake Vanda (VAN).

Note that Lake Fryxell is plotted on a secondary x-axis

Li et al.: Antarctic lake planktonic community trophic activities

dependent upon which variable was used for nor-malization of the data.

Lake Fryxell

Lake Fryxell is the least saline (maximum conduc-tivity ~7 mS cm−1) and shallowest (20 m maximum

depth) of the 3 lakes in this study(Fig. 2), and exhibited very low PARlevels (maximum PAR <10 µmol pho-tons m−2 s−1) (Fig. 3F). In common withLake Bonney, nutrient concentrationswere low in the upper oxic zone. SRPlevels were significantly higher in thedeeper zones of Lake Fryxell relativeto Lake Bonney and a recent studyshowed that Lake Fryxell phytoplank-ton communities are co-limited for ni-trogen and phosphorus (Teufel et al.2017). Higher nutrient levels supportedthe highest levels of chl a and bacterialbiomass in the water column of LakeFryxell compared with other MDVlakes (Fig. 4C), peaking at 9 m and 7 mfor chl a and bacterial biomass, respec-tively.

On a protein basis, RubisCO activitypeaked above the chemocline (7 m);however, on a lake water volume basisRubisCO activity peaked in the deepanoxic zone (14 m) (Fig. 5C,G). βGAMactivity levels in Lake Fryxell werecomparable with Lake Bonney; how-ever, no spatial trends were noted.

Lake Vanda

Lake Vanda is the deepest lake(maximum depth 80 m) in our studyand exhibited fresh shallow water andhypersaline bottom water which wascomparable to Lake Bonney (Fig. 2).PAR levels under the ice were 5−10 ×higher relative to Lakes Bonney andFryxell, and the maximum tempera-ture was above 20°C at the bottom ofthe water column (Fig. 3F). Nutrientconcentrations for most of the watercolumn (above 60 m depth) were lowor at undetectable levels, and SRP lev-els in Lake Vanda were comparable

with Lake Bonney (i.e. maximum SRP <10 µM)(Fig. 3G). Lake Vanda is considered one of the mostoligotrophic bodies of water in the world (Vincent etal. 1981); both chl a and bacterial biomass levelspeaked within the chemocline (68 m depth) and werethe lowest of the 3 study lakes (Fig. 4D). In contrast,both RubisCO and βGAM activities peaked in thedeeper layer of Lake Vanda (72 m) (Fig. 5D,H).

231

Fig. 3. Depth profiles of physical and chemical characteristics for the 3 lakes inthis study: (A,B) Lake Bonney east lobe (ELB) and (C,D) west lobe Lake Bonney(WLB), (E,F) Lake Fryxell (FRX), and (G,H) Lake Vanda (VAN). Note that photo -synthetically active radiation (PAR) and temperatures for Lake Vanda are plot-

ted on secondary x-axes. SRP: soluble reactive phosphorus

Aquat Microb Ecol 82: 225–239, 2019

Cluster analyses of lake physicochemical parameters

In agreement with past reports (Vincent 1981,Priscu 1995, Roberts et al. 2000), lake physicochem-istry and biology exhibited strong vertical stratifica-tion as well as lake-specific differences (Figs. 2−4).This spatial complexity in lake biogeochemistry wasreflected in spatial trends of both RubisCO andβGAM enzyme activities (Fig. 5). In an effort to de -termine whether specific environmental and biolog-ical factors influence protist autotrophic or hetero-trophic activity, we first performed PCA of majorabiotic factors (i.e. temperature, PAR, conductivityand dissolved inorganic nutrients) across all sam-pling sites. Despite the lake-specific differences inseveral physicochemical parameters, the PCA re -sults indicated that physicochemical parameters col-lected from 93 total samples clustered strongly in -to 2 major groups representing sampling depthslocated in the shallow layers (above chemocline)

versus the deep layers (chemocline anddeeper) (Fig. 6). In the PCA, the first 3principal components explained themajority of the variation between sam-ples (34.6, 23.4 and 20.2% for PC1,PC2, and PC3, respectively) (Fig. 6).The shallow and deeper layers weremajorly separated on the first axis(PC1); the environmental parameterloadings of the first 3 principal compo-nents are shown in Table S2 in theSupplement. The first component lar -gely represented a gradient of nutrientbalance (N:P ratio) and conductivitylevels. It also positively correlated withconcentrations of DIC, NO3

−, NH4+

while PAR levels had a weakly nega-tive effect. In addition, the separationbe tween samples within shallow anddeeper layers was majorly in the sec-ond and third principal components.The second component ma jorly repre-sented a gradient of temperature andconcentrations of NH4

+ and SRP. Thethird component represented a gradi-ent of NO3

−, SRP and DIC, and weaklycorrelated with PAR and temperature.Further, NO3

− PAR and temperaturewere at the positive end of the gradientwhile SRP and DIC were at the nega-tive end.

Correlation of autotrophic and heterotrophicactivities with physicochemical parameters

Given that lake physicochemical parameters strong -ly clustered into 2 groups (i.e. the upper water col-umn and at/below the chemocline), we performedPearson correlation coefficient (PCC) analysis be -tween autotrophic and heterotrophic enzyme activityand a suite of biological and environmental parame-ters in shallow and deep layers separately. Severalenvironmental and biological parameters exhibitedsignificant correlations with RubisCO and βGAM(Table 1). Spatial trends in Log RubisCO activity inthe shallow water column layers weakly correlatedwith abiotic parameters such as conductivity, temper-ature, NO3

−, SRP− and N:P, as well as biotic parame-ters such as PPR and abundance of cryptophytes.However, in deep water columns Log RubisCO activ-ity negatively corresponded with the environmentalparameter conductivity as well as NH4

+, NO3− and

232

Fig. 4. Depth profiles for chl a concentration and bacterial abundance in (A)Lake Bonney east lobe (ELB), (B) Lake Bonney west lobe (WLB), (C) Lake

Fryxell (FRX) and (D) Lake Vanda (VAN)

Li et al.: Antarctic lake planktonic community trophic activities

N:P, and positively corresponded with the biologicalparameters chl a, protein concentration, as well asabundance of cryptophytes and haptophytes.

The relationship between spatial distribution ofheterotrophic activity (i.e. βGAM) and various lakeparameters was distinct from that of RubisCO. Statis-tically significant correlations for βGAM activity inshallow water column layers include those with PAR,temperature, bacteria counts, protein concentration,chl a as well as green algae. In contrast, the enzymeactivity in deep water column layers showed correla-tions with conductivity, temperature, NH4

+ and SRP−,protein concentration (Table 1). Interestingly, in thedeep layers βGAM activity was positively correlatedwith temperature while in the shallow layers, it hadnegative correlation with this parameter.

We selected linear models of RubisCO and βGAMactivities and their significant predictors (i.e. abioticand biotic factors) based on the lowest AIC values(Table 2). RubisCO activity in the shallow layers waspredicted mostly by abiotic variables (i.e. DIC con-centration, conductivity and am mo nia concentration)

(R2 = 0.49, p = 0.02). In the deep lake layers, irradi-ance (PAR), conductivity, major nutrient ratio (N:P)and total protein concen tration were the significantpredictors of RubisCO activity (R2 = 0.80, p < 0.001).On the other hand, physicochemical conditions werethe significant predictors of βGAM activity in bothshallow and deep layer models, with R2 = 0.65, p <0.001 and R2 = 0.80, p < 0.001, respectively. Temper-ature and N:P were present in both models whilechl a and phosphate concentrations were importantto ex plain βGAM activity in shallow layers, and con-ductivity and irradiance were significant predictorsof the activity in deep layers.

DISCUSSION

MDV lakes exhibit strong gradients in the physico-chemistry which drives variations in microbial distri-bution, diversity and abundance within and betweenthe lakes (Laybourn-Parry et al. 1997, Roberts et al.2004b, Kong et al. 2012b). Recent studies combining

233

Fig. 5. Spatial variability in enzyme activity of RubisCO and β-D-glucosaminidase (βGAM) in study sites: (A,E) Lake Bonneyeast lobe, (B,F) Lake Bonney west lobe, (C,G) Lake Fryxell and (D,H) Lake Vanda. Enzyme activity is represented as eithernmol µg protein−1 h−1 (A−D) or nmol h−1 l−1 lake water (E−H). Note that the scales of the x-axes are different across all panels,

and also note the break in the x-axis in (A,C,D) and (G)

Aquat Microb Ecol 82: 225–239, 2019

sequencing and quantitative PCR (qPCR) approacheshave described the diversity and vertical distributionof the protistan communities residing in MDV lakes(Bielewicz et al. 2011, Kong et al. 2012a,b, Dolhi et al.2015). Based on gene and transcript abundance ofform I and II RubisCO genes (rbcL and cbbM, respec-tively), these studies predicted that light is a strongenvironmental driver for the obligate photoauto-trophs (e.g. chlorophytes). On the other hand, spatial

and seasonal variability in mixotro-phic protist populations (e.g. hapto-phytes and cryptophytes) appears tobe independent of PAR availability(Kong et al. 2012b, Dolhi et al. 2015).The present study extends the find-ings from the molecular studies bymeasuring autotrophic and heterotro-phic enzyme activities in the MDVlake communities.

Spatial trends in autotrophic (Rubis -CO) activity peaked at or near thechemocline in both lobes of Lake Bon-ney as well as Lake Vanda. WLB ex-hibited relatively low RubisCO acti vitycompared to ELB despite the presenceof a well-defined chl a maxi mum. Theapparent uncoupling of chl a from car-bon fixation in WLB was also noted byKong et al. (2012b), who observed thattranscriptional activity of 2 types of Ru-bisCO large subunit (rbcL) genes was25 and 500× lower in WLB comparedwith ELB (Kong et al. 2012b). In sup-port of the molecular evidence, PPRwas 4 to 5× lower in WLB comparedwith ELB (Kong et al. 2012b); however,expression of the photosystem II reac-tion center gene psbA was comparablebetween the 2 lobes. These reportscombined with results presented hereindicate that phytoplankton in WLBhave a relatively high light harvestingcapacity combined with a lower capac-ity for fixation of CO2, relative to ELB.This uncoupling between light energycapture and C-fixation may be due tophoto acclimation in obligate photo-autotrophs and/or heterotrophic meta -bolism in mixotrophic phytoplankton(Morgan-Kiss et al. 2016, Obryk et al.2016).

Permanent chemoclines are steepredox transitions associated with high

microbial activity and biomass in meromictic lakes(Van Gemerden & Mas 1995), and represent a richsource of POC for heterotrophic and mixotrophicprotists. While mixotrophy in the chemoclines islikely a strategy for energy acquisition under limitinglight levels, peaks in βGAM activity in the shallowdepths of both lobes of Lake Bonney may reflectdigestion of bacterial prey as an alternative mecha-nism for the acquisition of additional nutrients (nitro-

234

Fig. 6. Principal component analysis (PCA) based on major abiotic factors of 4study lakes. Cond: conductivity; PAR: photosynthetically available radiation; T:temperature; DIC: dissolved inorganic carbon; NO3: NO3

−; NH4: NH4+; PO4:

soluble reactive phosphorus; N:P: dissolved inorganic N:P ratio. Shallow anddeep layers were coded as orange and blue respectively. Circles: 75% similar-ity within each group. Symbols indicated samples from Lake Bonney east lobe(plus), Lake Bonney west lobe (circle), Lake Fryxell (cross) and Lake Vanda

(diamond)

Li et al.: Antarctic lake planktonic community trophic activities

gen and phosphorus) in the highly oli-gotrophic surface layers (Laybourn-Parry et al. 2005). In support of thishypothesis, βGAM activity correlatedwith bacterial abundance (Table 1),and N:P explained βGAM variation(Table 2) in the shallow zones.

Spatial trends in RubisCO andβGAM in both ELB and WLB matchedthat of haptophyte rbcL abundance ina previous study (Kong et al. 2012b).The dominant haptophyte in LakeBonney is an Isochrysis sp. (Bielewiczet al. 2011). While this genus is not typ-ically associated with mixotrophy, theLake Bonney strain was isolated andshown to be capable of mixotrophy(Table S1) (Li et al. 2016). Thus, it islikely that haptophytes contribute sig-nificantly to both RubisCO and βGAMactivity within the chemoclines of ELBand WLB. In addition, a recent reportidentified Isochrysis as a major prey fora heterotrophic nanoflagellate, Pteri -domonas sp. (Li et al. 2016). Thus, Iso -chrysis sp. represents a keystone inLake Bonney, playing a significant rolein the transfer of carbon and nutrientswithin the food web of MDV lakes.

Of the 4 lakes in this study, LakeFryxell exhibited the highest ratio ofRubisCO to βGAM activity (RubisCO:βGAM range of 3−36). Higher PPR inLake Fryxell correlates with its statusas one of the most productive lakes inthe MDV (Priscu 1995). Mixotrophiccryptophytes related to Geminigeracryophila dominate the photic zone(Dolhi et al. 2015) and are likely to bemajor contributors to the carbon cyclein Lake Fryxell. The contribution ofcryptophytes to primary production inall the lakes was evident in the signif-icant correlation between RubisCOactivity and cryptophyte abundance(Table 1). The contribution of cyano-bacteria to RubisCO activity cannotbe ruled out, as Oscillatoria spp. andPhormidium spp. have been dominantmembers of the under-ice and chemo-cline communities, respectively, incer tain years (Spaulding et al. 1994).Conversely, βGAM activity remained

235

Con

dP

PR

PA

RT

emp

NH

4+N

O3−

SR

PN

:PD

ICB

act

Pro

tC

hl

aG

alg

Cya

nH

apt

Cry

pt

Lo

g R

ub

isC

OS

hal

low

(n

= 2

0)−

0.53

*0.

5*0.

27−

0.49

*−

0.35

−0.

48*

−0.

52*

−0.

49*

−0.

040.

3−

0.08

0.46

−0.

29−

0.31

0.26

0.51

*D

eep

(n

= 3

7)−

0.66

***

−0.

210.

26−

0.19

–0.

52**

–0.

42**

−0.

23–

0.69

***

−0.

140.

250.

42**

0.57

***

−0.

36*

0.04

0.5*

*0.

75**

*

βGA

MS

hal

low

(n

= 4

2)0.

1−

0.26

−0.

47**

−0.

39*

0.03

0.04

0.06

0.08

−0.

010.

54**

*0.

39*

0.67

***

0.54

***

0.15

0.17

−0.

15D

eep

(n

= 5

1)−

0.47

**−

0.05

−0.

220.

41**

0.5*

**−

0.18

0.45

***−

0.22

−0.

22−

0.06

0.32

*0.

010.

030.

25−

0.09

−0.

09

Tab

le 1

. Pea

rson

cor

rela

tion

coe

ffic

ien

t va

lues

(r)

for

aver

age

Ru

bis

CO

an

d β

- D-g

luco

sam

inid

ase

(βG

AM

) ac

tivi

ty w

ith

lak

e p

hys

ical

, ch

emic

al, a

nd

bio

log

ical

par

ame-

ters

for

all

sam

ple

s (n

= 5

7−93

). T

he

sig

nif

ican

ce o

f th

e co

rrel

atio

n c

oeff

icie

nt

is in

dic

ated

by

*0.0

1 <

p <

0.0

5; *

*0.0

01 <

p <

0.0

1; *

**p

< 0

.001

. Con

d: c

ond

uct

ivit

y; P

PR

:p

rim

ary

pro

du

ctiv

ity;

PA

R: p

hot

osyn

thet

ical

ly a

vaila

ble

rad

iati

on; T

emp

: tem

per

atu

re; S

RP

: sol

ub

le r

eact

ive

ph

osp

hor

ous;

DIC

: dis

solv

ed i

nor

gan

ic c

arb

on; B

act:

bac

teri

a co

un

ts; P

rot:

pro

tein

; Ch

l a:

ch

loro

ph

yll

a; G

alg

: gre

en a

lgae

; Cya

n: c

yan

obac

teri

a; H

apt:

hap

top

hyt

es; C

ryp

t: c

ryp

top

hyt

es

No.

Dep

end

ent

vari

able

Lin

ear

mod

elA

ICR

2n

1L

og R

ub

isC

Osh

allo

w1.

8***

+ 0

.27*

* D

IC −

1.1

6***

Log

Con

d +

0.1

5 L

og N

H4+

520.

4918

2L

og R

ub

isC

Od

eep

−2.

83 +

0.1

4* P

AR

+ 1

.55*

Log

Con

d −

0.6

5***

Log

SR

P −

0.6

6* L

og N

:P +

0.1

1 C

hl a

+ 0

.07*

** P

rot

780.

8037

3βG

AM

shal

low

1.41

** −

0.2

2***

Tem

p −

0.0

02 C

hl

a+

0.6

8***

Log

SR

P +

0.5

8***

Log

N:P

950.

6539

4βG

AM

dee

p8.

84**

* −

1.1

1***

PA

R +

0.3

7***

Tem

p −

3.8

9***

Log

Con

d +

2.3

2***

Log

N:P

228

0.61

54

Tab

le 2

. Lin

ear

reg

ress

ion

mod

els

exp

lain

ing

th

e en

zym

e ac

tivi

ties

in d

iffe

ren

t p

orti

ons

of w

ater

col

um

ns.

Th

e m

odel

s w

ere

sele

cted

by

low

est

AIC

val

ue

from

all

pos

-si

ble

lin

ear

mod

els

to e

xpla

in l

og-t

ran

sfor

med

en

zym

e ac

tivi

ties

in

sh

allo

w a

nd

dee

p l

ayer

s as

dep

end

ent

vari

able

s an

d l

og-t

ran

sfor

med

ab

ioti

c (e

xcep

t te

mp

erat

ure

)an

d b

ioti

c p

aram

eter

s as

in

dep

end

ent

vari

able

s. β

GA

M: β-

D-g

luco

sam

inid

ase;

Con

d:

con

du

ctiv

ity;

PA

R:

ph

otos

ynth

etic

ally

ava

ilab

le r

adia

tion

; T

emp

: te

mp

erat

ure

;D

IC:

dis

solv

ed i

nor

gan

ic c

arb

on;

Ch

l a:

ch

loro

ph

yll

a; P

rot:

pro

tein

. T

he

sig

nif

ican

ce o

f th

e co

rrel

atio

n c

oeff

icie

nt

is i

nd

icat

ed b

y *0

.01

< p

< 0

.05;

**0

.001

< p

< 0

.01;

**

*p <

0.0

01

Aquat Microb Ecol 82: 225–239, 2019

low in Lake Fryxell, suggesting that the Lake Fryxellmixotrophs were not reliant on heterotrophy duringthe sampling time of this experiment (mid-summer).Thus, while all MDV lakes are limited for light by thepermanent ice-covers, we suggest that nutrient avail-ability is a strong driver of protist community trophicstatus. This hypothesis is also supported by a recentnutrient bioassay study mimicking a climate-relatedsummer flood event which showed that the diversityand abundance of phytoplankton communities inLake Bonney, but less so in Lake Fryxell, was highlyresponsive to nitrogen and phosphorus concentra-tions (Teufel et al. 2017).

The PCC analysis of enzyme activity correlatingwith biotic and abiotic environmental parametersprovided further insights on drivers of auto- and het-ero-trophic groups in shallow and deep lake depths.The correlation between RubisCO activity and bioticparameters (protein concentration, chl a, and hapto-phyte and cryptophyte abundance) in deep layers indicates that haptophytes and cryptophytes arelikely major contributors to inorganic carbon fixationwithin the chemoclines of Lakes Bonney and Fryxell,respectively. On the other hand, βGAM activity cor-related strongly with biotic parameters (bacterialcounts, protein concentration, chl a, and green algaeabundance) in shallow lake layers, whereas βGAMactivity correlated strongly with abiotic parameters(temperature, salinity, NH4

+, and PO43−) in deep lake

layers. This correlation fits well with the assumptionthat heterotrophic protists heavily graze on bacterio-plankton and small phytoplankton (Laybourn-Parryet al. 1995, 1997, Laybourn-Parry 1997). The positivecorrelation between βGAM activity and biotic param-eters in these depths suggests that prey availabilitymight be a limiting factor of heterotrophic activity inshallow layers. In contrast, PCC analysis indicatesthat conductivity was negatively correlated with het-erotrophic activity, which can be explained by thefact that higher salinity depths (e.g. hyper-saline water in deep Lake Bonney) may impact the physiol-ogy of heterotrophic protists (Greenwald & Hurlbert1993). Since many heterotrophic organisms are knownfor their delicate cell structure, it is not surprising thatMDV heterotrophic protists could be adversely af-fected by the extreme salinity levels in the deep zoneof Lake Bonney. A recent paper on the diversity anddistribution of fungal communities in the MDV lakesalso reported salinity as a strong driver of abundanceand diversity (Rojas-Jimenez et al. 2017).

RubisCO activity was measurable at depths belowthe photic zones in all MDV lakes studied, indicatingthat there is the potential for carbon fixation using

alternative energy sources, such as reduced organicor inorganic compounds (i.e. chemolithoautotrophicmetabolism). Evidence for chemolithoautotrophicbacteria in the MDV aquatic ecosystems has beenreported for several MDV lakes. The anoxic, sulfidicbottom waters of Lake Fryxell have yielded culture-dependent (Sattley & Madigan 2006) and independ-ent (Dolhi et al. 2015) evidence for sulfur-oxidizingbacteria (SOB) related to Thiobacillus thioparus. Evi-dence for the existence of a SOB community belowthe photic zone was also found for WLB (Kong et al.2012a, Dolhi et al. 2015). It is predicted that theseorganisms utilize dimethyl sulfoxide (DMSO), abreakdown product of dimethylsulfoniopropionate(DMSP) which is produced by algae including hapto-phytes, chrysophytes, and dinoflagellates as a cryo-or osmo-protectant (Trevena et al. 2000), as an electronacceptor. A second breakdown product, dimethylsul-fide (DMS) may be utilized as an electron donor(Raina et al. 2010). The potential for dark carbon fix-ation by ammonia-oxidizing bacteria also exists inELB and Lake Vanda (Vincent et al. 1981, Voytek etal. 1999).

While the RubisCO enzyme assay measured thepotential for a community to fix inorganic carbon,βGAM activity is a proxy for community respiration,specifically breakdown of complex organic carbonsources by acidic food vacuoles. As the result of localclimate change, it has been reported that water lev-els of the MDV lakes have risen significantly in thepast 30 yr due to greater input of glacial melt andstreams (Fountain et al. 2016). Continuous fresh waterinflow into these closed basin lakes could potentiallyaffect the lake physiochemical conditions. For exam-ple, our linear model predicts that a 10% decrease ofwhole lake conductivity would cause a 12% RubisCOactivity increase in the water column above the chemo-cline or a 16% RubisCO enzyme activity decreaseand a 14% increase of average βGAM activity belowthe chemocline.

Some fraction of the βGAM activity could be attrib-utable to other microorganisms. Sherr & Sherr (1999)observed that both heterotrophic bacteria and phyto-plankton exhibit βGAM activity. However, a morerecent paper tested the contribution of heterotrophicbacteria to particulate βGAM activity and reportedthat particulate activity was magnitudes higher com-pared with that of the filtrate (Sintes & del Giorgio,2010). In addition, βGAM activity is below detectionlimits in a strict photosynthetic alga, Chlamydo mo -nas sp. ICE-MDV (Table S1), a dominant member ofthe Lake Bonney phytoplankton community (Biele -wicz et al. 2011).

236