Host-compound foraging by intestinal microbiotarevealed by single-cell stable isotope probingDavid Berrya, Bärbel Stecherb, Arno Schintlmeistera,c, Jochen Reicherta, Sandrine Brugirouxb, Birgit Wildd,Wolfgang Wanekd, Andreas Richterc,d, Isabella Rauche, Thomas Deckere, Alexander Loya,1, and Michael Wagnera,c

aDepartment of Microbial Ecology, Vienna Ecology Center, cLarge Instrument Facility for Advanced Isotope Research, and dDepartment of TerrestrialEcosystem Research, Vienna Ecology Center, Faculty of Life Sciences, University of Vienna, A-1090 Vienna, Austria; bMax von Pettenkofer Institute of Hygieneand Medical Microbiology, Ludwig-Maximilians-University of Munich, 80336 Munich, Germany; and eDepartment of Microbiology, Immunobiologyand Genetics, Max F. Perutz Laboratories, University of Vienna, A-1030 Vienna, Austria

Edited by Ralph R. Isberg, Howard Hughes Medical Institute, Tufts University School of Medicine, Boston, MA, and approved February 6, 2013 (received forreview November 9, 2012)

The animal and human intestinal mucosa secretes an assortment ofcompounds to establish a physical barrier between the host tissueand intestinal contents, a separation that is vital for health. Somepathogenic microorganisms as well as members of the commensalintestinal microbiota have been shown to be able to break downthese secreted compounds. Our understanding of host-compounddegradation by the commensal microbiota has been limited toknowledge about simplified model systems because of the diffi-culty in studying the complex intestinal ecosystem in vivo. In thisstudy,we introduce an approach that overcomes previous technicallimitations and allows us to observe which microbial cells in theintestine use host-derived compounds. We added stable isotope-labeled threonine i.v. to mice and combined fluorescence in situhybridization with high-resolution secondary ion mass spectrome-try imaging to characterize utilization of host proteins by individualbacterial cells. We show that two bacterial species, Bacteroidesacidifaciens and Akkermansia muciniphila, are important host-protein foragers in vivo. Using gnotobiotic mice we show thatmicrobiota composition determines the magnitude and pattern offoraging by these organisms, demonstrating that a complex micro-biota is necessary in order for this niche to be fully exploited. Theseresults underscore the importance of in vivo studies of intestinalmicrobiota, and the approach presented in this study will bea powerful tool to address many other key questions in animaland human microbiome research.

nanoSIMS | stable isotope probing

A fundamental interaction between the intestinal microbiotaand their host is the foraging of host-derived compounds by

mutualistic and pathogenic microorganisms. Utilization of host-derived substrates is known to confer a selective advantage tosome enteric pathogens during inflammation (1), but little isknown about host-compound foraging in the healthy mammalianintestines. The secreted mucus, a heterogeneous mixture ofcompounds that covers epithelial tissue in two layers (∼150 μmtotal in the mouse colon) (2), is regenerated rapidly, on theorder of less than in a single day (1, 3). The high production andturnover of secreted compounds represents a steady influx ofnutrients and therefore constitutes an ecological niche formicroorganisms capable of, or even specialized at, foraginghost-secreted compounds. Therefore, the secretion not only ofantimicrobial substances (2, 4) but also of specific substrates isa means for the host to shape the structure of its symbiontcommunity (5).Despite the obvious importance of host-compound foraging,

technical difficulties in teasing apart this close interaction havelimited studies to pure culture and simplified in vitro systems orgnotobiotic mouse models (6, 7). Results produced from simpli-fied model systems and metaomics studies are valuable for gen-erating hypotheses, but techniques to directly probe importantintestinal processes such as host-compound foraging are ur-gently needed to test these hypotheses (8). Here, we explored

host-compound foraging using a stable isotope tracer-based ap-proach for in vivo labeling and quantitative imaging of individual gutbacteria with high-resolution secondary ion mass spectrometry(NanoSIMS). We used the amino acid threonine as a marker as,after injection, it is used by the intestines for biosynthesis ofthe threonine-rich mucin glycoproteins (6, 7) but will also be in-corporated into several other secreted protein compounds.

Results and DiscussionTo establish the tracer approach and to define the optimal sam-pling time for the NanoSIMS measurements, a physiologicallyrelevant concentration of stable isotope labeled threonine (both13C and 15N labeled) was given i.v. to mice and the transport of thestable isotopes was tracked in the blood serum, cecum tissue, andgut lumen using elemental analysis isotope ratio mass spectrom-etry (EA–IRMS). Enrichment of 13C and 15N was not observed inthe lumen contents until after 4 h postinjection (h.p.i.) and δ13Cand δ15N values peaked at about 8 h.p.i. (Fig. S1A, SI Results andDiscussion). Concomitantly, acetate, propionate, and butyrate inthe lumen, main products of microbial metabolism of secretedcompounds, were also enriched in 13C (Fig. S1A). Consistent withthe hypothesis that the microbiota plays a role in the transfer oftracer to the gut, the lumen contents of germ-free mice and gno-tobiotic mice colonized with a low complexity intestinal microbiota(Fig. S2) were significantly less enriched than mice with a normalcomplexity (NC) microbiota (Fig. S1C; Student’s t test, P < 0.05).This indicated that themicrobiota richness and/or composition areimportant modifiers of secretion.Using NanoSIMS we were able to visualize the transfer of label

from the mouse cecum tissue to the gut lumen. Semithin sectionsof embedded samples showed distinctive hotspots of 15N in thetissue as well as in the lumen (Fig. 1). The 15N hotspots in thetissue were likely within secretory mucus cells (Fig. S3), whichwould be expected to have a higher demand for threonine tosupport the ongoing biosynthesis of mucins, although verificationwith Muc2 immunohistochemical staining was not possible with-out disturbing the isotope composition of the sample (SI Resultsand Discussion). By combining NanoSIMS imaging and fluores-cence in situ hybridization (FISH) imaging of the same field ofview, we were able to determine the isotope content of individualcells hybridized with specific phylogenetic probes (Table S1).Using gut lumen samples prepared with a FISH-compatible

Author contributions: D.B., A.L., and M.W. designed research; D.B., B.S., A.S., J.R., S.B., B.W.,and I.R. performed research; T.D. contributed new reagents/analytic tools; D.B., A.S., B.W.,W.W., A.R., A.L., and M.W. analyzed data; and D.B., A.L., and M.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.1To whom correspondence should be addressed. E-mail: loy@microbial-ecology.net.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1219247110/-/DCSupplemental.

4720–4725 | PNAS | March 19, 2013 | vol. 110 | no. 12 www.pnas.org/cgi/doi/10.1073/pnas.1219247110

acrylic resin and probes targeting broad or highly abundantgroups, we found widespread 15N enrichment. In addition, 15Nhotspots were associated with cells targeted by the Bacteroidetes-specific (Bac303) and the Lachnospiraceae-specific (Erec482)probes, as well as with cells belonging to a highly abundant

species-level phylotype in the Lachnospiraceae [Lachnospiraceaeoperational taxonomic unit (OTU)_11021, mean relativeabundance in healthy mice of 9.5%] (8) (Table S2). These resultswere consistent with reports that the capacity to degrade mucin invitro is phylogenetically widespread (9).

Fig. 1. Imaging 15N enrichment (δ15N) in the cecum tissue and intestinal lumen 8 h after i.v. injection of 13C,15N threonine. A mosaic of 16 individual high-resolution NanoSIMS images is shown. 12C14N− secondary ion intensity distribution images of the same area are shown to illustrate the structure of the tissueand lumen biomass. 15N hotspots are indicated by white arrows. The patchy distribution of 15N hotspots in the tissue is likely attributable to the heterogeneityin mucus release activity both within and between mucus cells (24, 25).

Fig. 2. Representative NanoSIMS (at% 13C and 15N) and FISH images 8 h after i.v. injection of 13C,15N threonine. (A) B. acidifaciens (blue) and Rumino-coccaceae OTU_5807 (red) cells (bar: 5 μm; prepared without embedding; other Bacteria shown in green), (B) Akkermansia spp. (white) cells (bar: 2 μm;prepared without embedding), and (C) Lactobacillaceae/Enterococcaceae spp. (red) cells (bar: 5 μm; prepared without embedding; other Bacteria shown ingreen). (D) Semithin sections of lumen contents with Lachnospiraceae OTU_11021 (yellow/white, overlap of green and red signals), all other Bacteria (blue),and autofluorescent dietary fibers (green) (bar: 10 μm). Exemplary cells with or without significant enrichment (white or green arrows) are indicated. Noenrichment in carbon is visible in D because of dilution by unlabeled carbon in the resin.

Berry et al. PNAS | March 19, 2013 | vol. 110 | no. 12 | 4721

MICRO

BIOLO

GY

To quantify per-cell C and N isotope composition withoutcarbon contributions from the resin (Fig. 2C, SI Results andDiscussion), we analyzed nonembedded lumen contents. We ap-plied a suite of probes based on abundant bacterial groups ina previous study of dextran sodium sulfate (DSS)–induced acutecolitis in the same mice (8). These probes were specific forAkkermansia spp., Bacteroides acidifaciens, Mucispirillum spp.,Lactobacillaceae/Enterococcaceae spp., and a RuminococcaceaeOTU_5807 species-level phylotype. All targeted microbial groupswere enriched in 13C compared with unlabeled controls (Fig. 3),indicating phylogenetically widespread incorporation of 13C-labeled substrates either via direct utilization of host-derivedcompounds or cross-feeding. Importantly, however, 13C enrich-ment was not uniform between the groups, and significant 13Clabeling was detected in a large subset of the cell population ofB. acidifaciens (42%) as well as Akkermansia spp. (29%), and onlyto a much lesser extent in Ruminococcaceae OTU_5807 (7%),Mucispirillum spp. (12%), and Lactobacillaceae/Enterococcaceaespp. (2%) (Figs. 2 and 3). Past findings lend additional weight tothis differential foraging pattern as Akkermansia muciniphila inpure culture degrades mucin and is an abundant member of themouse and human intestinal microbiota (10). B. acidifaciens hasnot previously been shown to degrade mucin, but another mem-ber of the same genus, the human intestinal bacterium Bacteroidesthetaiotaomicron, uses host compounds in a monoassociatedmouse model (11). This is, however, proof that these organismsforage on host-derived compounds in a complex gut ecosystem,which is important because substrate utilization can be affected byecological complexity (12). The variation in the percentage ofcells enriched in each target population indicates physiologicalheterogeneity within phylogenetic probe-defined cells in the gut.This variation may either be due to the targeting of differentgenotypes or ecotypes with the same probe-stained population, orto heterogeneity in the location or activity of genetically similar oridentical populations. It is, for example, possible that resource

partitioning of mucosal products between related organisms leadsto divergent or complementary ecological niches, a phenomenonthat has been observed for Lactobacillus species in the forest-omach of mice (13).In contrast to 13C, almost all cells analyzed were enriched in

15N, which is likely due to an enrichment of 15N in the luminal-free ammonium pool and subsequent incorporation of the la-beled ammonium via amino acid synthesis into biomass of allgrowing cells. 15N-ammonium can originate from microbial dis-proportionation during fermentation of labeled proteins, butmay also result from secretion of labeled urea into the lumen andsubsequent hydrolysis if threonine in the host was deaminated(14). However, there was a significant correlation betweencellular 13C and 15N content of all labeled cells, indicatinga specific coassimilation of 13C and 15N in addition to the generalenrichment of 15N (Pearson correlation coefficient 0.763, P <0.001) (Fig. S4). The most parsimonious explanation for thisobservation is that bacteria also assimilated organic matter thatwas directly sourced from secreted threonine-derived proteins.We estimated that ∼1.6% of the total cecum microbiota

(targeted by the EUB338 probe mix) was enriched in 13C (TableS2). Given the abundance of B. acidifaciens and Akkermansiaspp. in these communities (relative abundance of 1.0 and 2.1%,respectively; Table S3) and the high proportions of 13C-enrichedcells in these populations (42 and 29%, respectively), these twogroups are the numerically dominant foragers of host-derivedproteins in this mouse cecum community.We used a gnotobiotic mouse model to test the foraging

performance and system-wide impacts of the important host-compound foragers B. acidifaciens and A. muciniphila in a morecontrolled ecosystem. The type strains from these species wereallowed to colonize mice with a four-member gut microbiotaseparately and together for 10 d and reached high abundances(the “4S” community; SI Results and Discussion). Surprisingly,however, colonization with A. muciniphila or B. acidifaciens was

Fig. 3. Single-cell stable isotope labeling of selected bacterial groups in the mouse intestine 8 h after i.v. injection of 13C,15N threonine. At% 13C and 15N wascalculated for each cell. FISH probes targeted Akkermansia spp. (Akk), Mucispirillum spp. (Mcs), B acidifaciens (Bac), Ruminococcaceae OTU_5807 (Rum), andLactobacillaceae/Enterococcaceae spp. (Lab) (probe details in Table S1). Eub338-targeted unlabeled cells (control) from the gut were used as controls. Eachpoint represents a single cell, and box plots summarize the quartiles of the target population. Red points are significantly enriched cells (>95% confidenceintervals).

4722 | www.pnas.org/cgi/doi/10.1073/pnas.1219247110 Berry et al.

unable to increase 13C and 15N in the lumen or 13C enrichment inthe fermentation product pools, and NanoSIMS analysis verifieda low level of enrichment in single cells of the two added strainsas well as the 4S microbiota (Fig. 4). It is notable that A.muciniphila, which is generally considered to be a dedicated mucindegrader, is able to colonize effectively and reach a significantlyhigher relative abundance in 4S than in NC microbiota mice(Table S3) without assimilating much host-derived compounds.Our results suggest that although some organisms benefit dif-

ferentially from components of host-secreted compounds, theactivity of host foragers depends on other community members.Specialized degraders may require the presence of other organ-isms to efficiently harvest host-derived compounds due to thestructural heterogeneity of the glycoprotein substrates and theassociated enzymatic diversity required for complete degradation

(15, 16). Alternatively, host foragers may spread into other nichessuch as degradation of dietary compounds in the absence ofcompetition. The identification of these abundant host-proteinforagers and the ecological dependence of their activity raises in-triguing research questions for future studies to address, includinghow the availability of diet-derived compounds affects foragingbehavior and the mechanisms by which the commensal microbiotais able to stimulate host-protein secretion. As we have demon-strated here, single-cell stable-isotope probing complements pop-ular metaomics approaches (17, 18) by offering an unprecedentedability to explore the eco-physiology of uncultivated gut micro-organisms and directly interrogate their metabolic interactions. Invivo labeling in combinationwithNanoSIMS imaging will thereforebe a powerful approach to address other key questions in animal

Fig. 4. Mice that host a four-member intestinal microbiota (4S microbiota) consisting of Lactobacillus acidophilus [Altered Schaedler Flora (ASF) 360],Lactobacillus murinus (ASF 361), Mucispirillum schaedleri (ASF 457), and a Parabacteroides sp. (ASF 519) were colonized with either A. muciniphila (4S-Akk) orB. acidifaciens (4S-Bac) or both together (4S-Akk/Bac). Isotope ratios (δ) are given in units of permil (‰) and are relative to Vienna PeeDee Belemnite (V-PDB)for δ13C and to atmospheric air (atm. air) for δ15N. Colonization with A. muciniphila or B. acidifaciens or both strains together had no significant effect inenrichment of 13C or 15N in the lumen. (A) EA–IRMS data of carbon and nitrogen in lumen contents 8 h after i.v. injection of 13C15N threonine in mice that hadbeen colonized for 10 d. (B) Amount of total 13C in lumen acetate, propionate, and butyrate, measured with LC–IRMS. Values from NC microbiota mice used in(Fig. 1 B and C) are shown in A and B for comparison. (C) Per-cell at% 13C and 15N of selected bacterial groups. Measurements of ASF 360/361, 457, and 519cells from 4S, 4S-Akk, and 4S-Bac mice were pooled because there were no significant differences between mouse types and Eub338-targeted cells that werenot labeled (control) are shown for comparison. Each point represents the at% of a single cell, and box plots summarize the quartiles of the target pop-ulation. Red points are significantly enriched cells (>95% confidence interval).

Berry et al. PNAS | March 19, 2013 | vol. 110 | no. 12 | 4723

MICRO

BIOLO

GY

and human microbiome research, ranging from host selection ofspecific microbes (5) to the biological basis of enterotypes (17).

MethodsEthics Statement. All animal experiments were approved by the University ofVeterinary Medicine, Vienna, institutional ethics committee and the AustrianMinistry of Science and Research (BMWF) and were conducted in accordancewith protocols approved by the Austrian laws (BMWF-66.006/0002-II/10b/2010) and German authorities (Regierung von Oberbayern).

Animal Experiments. Specific pathogen-free (SPF) C57BL/6 mice that had NCmicrobiotawere used for all experiments not involving directmanipulation ofthemicrobiota andwere housed in a SPF facility according to recommendationsof the Federation of European Laboratory Animal Science Association.Germfree Naval Medical Research Institute (NMRI) mice were purchased fromthe University of Bern and transported to the Max von Pettenkofer Institute(Munich) in germfree shippers. NMRI mice colonized in a low-complexitymicrobiota (Fig. S2) were obtained from the Rodent Center HCI EidgenössicheTechnische Hochschule Zürich. Gnotobiotic female C57BL/6 mice colonizedwith four strains of the altered Schaedler flora (19) (ASF 360, 361, 457, and 519;4S microbiota, 4S) were bred at the Ludwig-Maximilians-University of Munichin flexible film isolators with Hepa-filtered air under germfree conditions andfed with autoclaved chow and water. 4S mice were associated by gavage withcultures of A. muciniphila (ATCC BAA-835) or B. acidifaciens (DSM 15896) thathad been grown overnight at 37 °C in anaerobic brain-heart infusion (BHI)broth [37 g/L BHI (OXOID), 0.25 g/L Cysteine-HCl·H2O, 0.25 g/L Na2S·9H2O].Gavaging of the mice with sterile BHI was used as control. Mice were kept inindividually ventilated cages for the duration of association. Upon harvest,blood, cecum epithelial tissue, scraped mucus, and lumen contents werecollected, and combined tissue and lumen were embedded in either epoxy(low-viscosity epoxy resin, LVR; Agar 100, Agar Scientific) or acrylic (LR White,Ted Pella) resin as described in detail below.

DSS Treatment. For DSS challenge, mice were provided with 2% (mass/vol)DSS (molecular weight, 36–50 kDa; MP Biomedicals) in autoclaved drinkingwater ad libitum for 72 h. No weight loss was observed, and intact gutbarrier was confirmed with a fecal occult blood assay (Hemoccult SENSA,Beckman Coulter).

Collection of Blood Serum. Mice were anesthetized and blood was collectedfrom the retro-orbital sinuses with a microhematocrit blood tube. Blood wasallowed to clot for 20 min at room temperature, then centrifuged in a ta-bletop centrifuge (14,000 × g) for 10 min. Serum was transferred to a newtube, snap-frozen in liquid nitrogen, and stored at –80 °C.

Cecum Contents. Immediately after collection of blood serum, mice werekilled. The intestines were opened, and cecum contents were removed toa microcentrifuge tube. High-purity water (200 μL) was added to the con-tents, the mixture was homogenized by vortexing, and the homogenate wascentrifuged (14,000 × g) for 2 min. The supernatant was transferred to a newmicrocentrifuge tube and snap-frozen in liquid nitrogen for later analysis ofshort chain fatty acids. The solid contents were divided and either flash-frozen for subsequent DNA extraction, dried in a vacuum concentrator(Eppendorf Concentrator 5301) for elemental analysis, or fixed for FISH.Fixation was performed by resuspending the contents in 2% para-formaldehyde and incubating overnight at 4 °C. This fixation was appro-priate for both Gram-positive and –negative cells, and >99% of SYBR green-stained cells also had a positive signal with the EUB338 probe mix. Thecontents were then pelleted and washed with PBS two times and thenresuspended in a solution of 60% ethanol/40% PBS and stored at –20 °C.

Secreted Mucus Collection. Mucus was collected by opening the cecum,carefully removing the lumen contents, and gently washing the tissue withsterile PBS. The insoluble mucus was then collected by lightly scraping theinner wall of the tissue with bent forceps and transferring mucus either toa microcentrifuge tube, where it was flash-frozen in liquid nitrogen forfurther purification, or to a tin cup, where it was dried in a vacuum con-centrator (Eppendorf Concentrator 5301) for elemental analysis.

Cecum Epithelial Tissue Preparation. Cecum tissue was washed in sterile PBSand small sections (∼100 mg) were homogenized in 1 M perchloric acid so-lution to release soluble components in a carbon-free solvent (16,000 rpm,30 s, 4 °C; Polytron PT-DA 07/2EC-B101). The proteins were separated fromacid-soluble amino acids by centrifugation (13,000 × g, 25 min, 4 °C). The

pellet was resuspended in 1 M perchloric acid and centrifuged, and decantedtwo times more to ensure removal of free threonine. Tissue material wasthen resuspended in high-purity water (Milli-Q), added to a tin cup, anddried in a vacuum concentrator for elemental analysis.

Tissue and Lumen Harvesting for Embedding. Cecum tissue and contents wereharvested together by carefully tying the edges of the selected area closedwith thread, excising the area, and fixing in Carnoy’s solution (6:3:1 mixtureof ethanol–acetic acid–chloroform) at 4 °C for 2 h. Fixed samples were thentransferred to a solution of 70% ethanol/30% PBS and stored at –20 °Cuntil embedding.

Resin Embedding and Sectioning. Samples were dehydrated using an in-creasing ethanol series (six steps: 80%, 80%, 90%, 90%, 96%, and 96%) for10 min at each step using molecular-grade ethanol. Samples were thenfurther dehydrated by three successive incubations in 100% ethanol for 20min each. For epoxy resin embedding, which is ideal for preservation of tissuestructure but is incompatible with FISH, the samples were placed in 100%acetone for 20 min three times. LVR (Agar 100, Agar Scientific) was infiltratedin a series of steps: a 1:3 mixture of LVR–acetone for 3 h, a 1:1 mixture for 8–12 h, a 3:1 mixture for 8–12 h, pure LVR for 3 h, and pure LVR overnight.Samples were then placed in a mold and into a 60 °C oven for 24 h to po-lymerize the resin. For embedding in FISH-compatible acrylic resin (LR White,Ted Pella), following the ethanol dehydration, the sample was infiltrated ina series of steps: 3:1 mixture of ethanol–LR White for 3 h, a 1:1 mixture for 8–12 h, a 3:1 mixture for 8–12 h, and pure LR White for 8–16 h three times.Samples in LR White were enclosed in gelatin capsules and incubated at 48 °Cfor 24 h to polymerize the resin. Semithin sections (300–500 nm) were cutwith an ultramicrotome (Leica EM UC6).

Pyrosequencing. DNA was extracted using a standard phenol-chloroformprotocol (8). DNA was subjected to a two-step barcoded PCR targeting the16S rRNA gene as described previously (20). Sequencing was performed ona GS FLX instrument (454/Roche) at the Norwegian High-Throughput Se-quencing Centre and data were analyzed using QIIME (21) as describedpreviously (8).

Isotope-Labeled Compounds and Pulse-Chase Labeling. For all labelingexperiments, stable isotope-labeled L-threonine [98 atom (at) % U-13C4, 98 at%15N, Sigma-Aldrich] or D-glucose (99 at% U-13C6, Cambridge Isotope Labora-tories Inc.) was used. Nonlabeled L-threonine (Sigma-Aldrich) was used asa 12C-control. Compounds were dissolved in sterile saline solution ([0.90%(wt/vol) of NaCl] and delivered to mice via a 50 μL lateral tail vein injection.Each mouse received either threonine (1.8 μmol or 18 μmol) or glucose (12μmol). Mice were then killed at selected time points (pre-injection, 0.5, 1, 2, 3,4, 5, 6, 7, 8, 16, or 24 h.p.i.). (For further details, see SI Results and Discussionand Figs. S1 and S4–S7.)

EA–IRMS. To measure the isotopic composition (at% 15N and at% 13C) ofblood serum, cecum epithelial tissue, and intestinal lumen contents, 0.5–1.2mg (dry weight) of each sample was placed in a tin capsule and dried at 60 °Covernight. Samples were analyzed with an elemental analyzer (EA 1110, CEInstruments) coupled via a ConFlo III device to the IRMS (DeltaPLUS, ThermoFisher). The precision of measurements in standards was better than ±0.00005 at% 15N and better than ±0.00011 at% 13C.

Liquid Chromatography IRMS. For compound-specific carbon isotope analysisof acetate, propionate, and butyrate, an HPLC system (Dionex Corporation)coupled to a Delta V Advantage Mass Spectrometer by a Liquid Chroma-tography (LC) IsoLink Interface (Thermo Fisher Scientific) was used (22).Samples were acidified with 1 M phosphoric acid to remove bicarbonate,and subsequently separated on a Macherey-Nagel Nucleogel Sugar 810Hcolumn at 75 °C with 0.5 mL·min−1 20 mM phosphoric acid as eluent. Peakswere converted to CO2 at 99 °C with 50 μL·min−1 of each 0.5 M sodiumpersulfate and 1.7 M phosphoric acid. Concentrations and isotopic compo-sitions of individual compounds were calibrated against external standardsas described elsewhere (22).

FISH. LR White semithin sections or gut lumen content were mounted oneither boron-doped silicon or indium tin oxide coated glass slides (7 × 7 mm)for use in the NanoSIMS. FISH was performed on these samples with mono-labeled 16S rRNA-targeted oligonucleotide probes using a standard protocol(23). Specificity and coverage of all applied probes with genus or higher cov-erage (Table S1) was confirmed using the Ribosomal Database Project II probe

4724 | www.pnas.org/cgi/doi/10.1073/pnas.1219247110 Berry et al.

match tool with database release 10, Update 31, containing 2,639,157 bacte-rial and archaeal 16S rRNA sequences. The search for each probe was restrictedto sequences of good quality with data in the target region. The specificity ofspecies-level phylotype probes was confirmed in either a previous study (8) bychecking them against 16S rRNA gene pyrosequencing data from the samemice or in this study by recovering near full-length 16S rRNA sequences fromclone libraries of the same samples. The intersection of species-level phylotypeprobes used for a single target had a perfect match only to the target se-quence and/or to similar sequences in the SILVA Small Subunit Non-RedundantReference (SSURef NR) Release 111 database (>97% sequence similarity) (TableS1). For most species-level phylotype targets, more than a single probe wasapplied with fully consistent results between probes, which supported thespecificity of the probes in these samples. Specificity for the probe targetingAkkermansia (Akk1437) was also experimentally tested by simultaneouslyhybridizing Akk1437 and the EUB338 III probe (which targets the phylumVerrucomicrobia), which confirmed that all Akk1437 signals were also EUB338III positive. The optimal hybridization stringency of all genus-level and species-level phylotype probes was determined in either this study or a previous study(8) by hybridizing gut lumen samples enriched in the target group (as de-termined with sequencing data) under a range of formamide concentrationsto produce formamide dissociation profiles for each probe. Hybridized sam-ples were imaged on an epi-fluorescence laser microdissection microscope(LMD, Leica LMD 7000) using a 40× air objective. Markings were made on thesample using the LMD laser to properly orient the sample and to record thesame field of view in the NanoSIMS.

NanoSIMS. NanoSIMS measurements were performed on an NS50L (Cameca).Data were recorded as images by scanning with a finely focused Cs+ primaryion beam (2–4.5 pA) and detection of negative secondary ions and secondaryelectrons. Recorded images had a 512 × 512 pixel resolution and field-of-view ranging from 30 × 30 to 60 × 60 μm2. The mass spectrometer was tunedfor a mass resolving power of ca. 10,000 at mass 26 to separate 12C14N fromthe isobaric species 13C2. Analysis areas were presputtered to establisha steady-state secondary ion formation before multicycle image acquisition.All images were recorded with a dwell time of 5–15 ms/pixel/cycle. Signalintensities were corrected for dead time effects and quasi-simultaneous ar-rival (QSA) effects using QSA sensitivity factors (i.e., “beta” values) of 1.10for C− and 1.05 for CN− ions. (For further details, see SI Results and Discussionand Figs. S3 and S8.)

Image Processing. FISH images were manually aligned to NanoSIMS images inPhotoshop (CS5, Adobe). Cells were identified in aligned FISH images by

automatic thresholding in the Fiji implementation of ImageJ, and resultingregions of interest were manually curated. NanoSIMS images were processedusing the OpenMIMS plugin in Fiji (v2.5, National Resource for Imaging MassSpectrometry). NanoSIMS images were autotracked and atomic percent 13Cand 15N were calculated from 12C− and 13C− and 12C14N− and 12C15N−

channels, respectively. Summary statistics from each region of interest werecalculated for single-cell analysis. Individual cells were considered signifi-cantly enriched in 13C,15N if the mean cellular at% 13C,15N was above the95th percent confidence interval of unlabeled control cells from the gutlumen and if the measurement error (3σ, Poisson) was smaller than thedifference between the at% of the labeled cell and the mean at% of un-labeled control cells.

Isotope Notation. Stable isotopes are given either in atom percent or δ no-tation. For an element X, with heavy isotope H, and light isotope L, atompercent is given by:

at%H X =H

H+ L* 100;  in percent ð%Þ;

and δ is given by:

δH X =�Rsample

Rs tandard− 1

�*1000;  in permil ð‰Þ:

To interconvert isotope values between notations, the following equationcan be used:

at%H X = 100 *δ+ 1000

δ+ 1000+1000

Rs tandard

;

where Rstandard refers to Vienna PeeDee Belemnite for δ13C (0.011180) and toatmospheric air for δ15N (0.0036765).

ACKNOWLEDGMENTS. We thank Margarete Watzka and Julia Ramesmayerfor technical assistance and the Norwegian High Throughput SequencingCentre for pyrosequencing. This work was financially supported by theAustrian Federal Ministry of Science and Research [Austrian Genome ResearchProgram (GEN-AU) III InflammoBiota], the European Research Council[Advanced Grant Nitrification Reloaded (NITRICARE) 294343], and a GermanFederal Ministry of Education and Research Infektionsgenomik grant. Nano-SIMS measurements were supported by large infrastructure grants from theUniversity of Vienna and the City of Vienna.

1. Winter SE, et al. (2010) Gut inflammation provides a respiratory electron acceptor forSalmonella. Nature 467(7314):426–429.

2. Johansson MEV, et al. (2008) The inner of the two Muc2 mucin-dependent mucuslayers in colon is devoid of bacteria. Proc Natl Acad Sci USA 105(39):15064–15069.

3. Faure M, et al. (2002) Development of a rapid and convenient method to purifymucins and determine their in vivo synthesis rate in rats. Anal Biochem 307(2):244–251.

4. Vaishnava S, et al. (2011) The antibacterial lectin RegIIIgamma promotes the spatialsegregation of microbiota and host in the intestine. Science 334(6053):255–258.

5. Koropatkin NM, Cameron EA, Martens EC (2012) How glycan metabolism shapes thehuman gut microbiota. Nat Rev Microbiol 10(5):323–335.

6. van der Schoor SR, Wattimena DL, Huijmans J, Vermes A, van Goudoever JB (2007)The gut takes nearly all: Threonine kinetics in infants. Am J Clin Nutr 86(4):1132–1138.

7. Schaart MW, et al. (2005) Threonine utilization is high in the intestine of piglets.J Nutr 135(4):765–770.

8. Berry D, et al. (2012) Phylotype-level 16S rRNA analysis reveals new bacterial in-dicators of health state in acute murine colitis. ISME J 6(11):2091–2106.

9. McGuckin MA, Lindén SK, Sutton P, Florin TH (2011) Mucin dynamics and entericpathogens. Nat Rev Microbiol 9(4):265–278.

10. Derrien M, Collado MC, Ben-Amor K, Salminen S, de Vos WM (2008) The Mucin de-grader Akkermansia muciniphila is an abundant resident of the human intestinaltract. Appl Environ Microbiol 74(5):1646–1648.

11. Sonnenburg JL, et al. (2005) Glycan foraging in vivo by an intestine-adapted bacterialsymbiont. Science 307(5717):1955–1959.

12. Ze X, Duncan SH, Louis P, Flint HJ (2012) Ruminococcus bromii is a keystone species forthe degradation of resistant starch in the human colon. ISME J 6(8):1535–1543.

13. Tannock GW, et al. (2012) Resource partitioning in relation to cohabitation of Lac-tobacillus species in the mouse forestomach. ISME J 6(5):927–938.

14. Fuller MF, Reeds PJ (1998) Nitrogen cycling in the gut. Annu Rev Nutr 18:385–411.15. Png CW, et al. (2010) Mucolytic bacteria with increased prevalence in IBD mucosa

augment in vitro utilization of mucin by other bacteria. Am J Gastroenterol 105(11):2420–2428.

16. Hoskins LC, Boulding ET (1981) Mucin degradation in human colon ecosystems.Evidence for the existence and role of bacterial subpopulations producing glyco-sidases as extracellular enzymes. J Clin Invest 67(1):163–172.

17. Arumugam M, et al.; MetaHIT Consortium (2011) Enterotypes of the human gutmicrobiome. Nature 473(7346):174–180.

18. Li M, et al. (2008) Symbiotic gut microbes modulate human metabolic phenotypes.Proc Natl Acad Sci USA 105(6):2117–2122.

19. Dewhirst FE, et al. (1999) Phylogeny of the defined murine microbiota: AlteredSchaedler flora. Appl Environ Microbiol 65(8):3287–3292.

20. Berry D, Ben Mahfoudh K, Wagner M, Loy A (2011) Barcoded primers used in mul-tiplex amplicon pyrosequencing bias amplification. Appl Environ Microbiol 77(21):7846–7849.

21. Caporaso JG, et al. (2010) QIIME allows analysis of high-throughput community se-quencing data. Nat Methods 7(5):335–336.

22. Wild B, Wanek W, Postl W, Richter A (2010) Contribution of carbon fixed by Rubiscoand PEPC to phloem export in the Crassulacean acid metabolism plant Kalanchoedaigremontiana. J Exp Bot 61(5):1375–1383.

23. Daims H, Stoecker K, Wagner M (2005) Advanced Methods in Molecular MicrobialEcology, eds Osborn A, Smith C (Bios-Garland, Abingdon, UK), pp 213–239.

24. Halm DR, Halm ST (2000) Secretagogue response of goblet cells and columnar cells inhuman colonic crypts. Am J Physiol Cell Physiol 278(1):C212–C233.

25. Perez-Vilar J, Mabolo R, McVaugh CT, Bertozzi CR, Boucher RC (2006) Mucin granuleintraluminal organization in living mucous/goblet cells. Roles of protein post-trans-lational modifications and secretion. J Biol Chem 281(8):4844–4855.

Berry et al. PNAS | March 19, 2013 | vol. 110 | no. 12 | 4725

MICRO

BIOLO

GY