ORIGINAL ARTICLE

Host genetic and environmental effects on mouseintestinal microbiota

James H Campbell1, Carmen M Foster1, Tatiana Vishnivetskaya1,2, Alisha G Campbell1,3,Zamin K Yang1, Ann Wymore1, Anthony V Palumbo1, Elissa J Chesler3,4 andMircea Podar1,3

1Biosciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, USA; 2Center for EnvironmentalBiotechnology, University of Tennessee, Knoxville, TN, USA; 3Graduate School of Genome Scienceand Technology, University of Tennessee, Knoxville, TN, USA and 4The Jackson Laboratory, Bar Harbor,ME, USA

The mammalian gut harbors complex and variable microbial communities, across both host phylo-genetic space and conspecific individuals. A synergy of host genetic and environmental factors shapethese communities and account for their variability, but their individual contributions and the selectivepressures involved are still not well understood. We employed barcoded pyrosequencing of V1-2and V4 regions of bacterial small subunit ribosomal RNA genes to characterize the effects of hostgenetics and environment on cecum assemblages in 10 genetically distinct, inbred mouse strains.Eight of these strains are the foundation of the Collaborative Cross (CC), a panel of mice derivedfrom a genetically diverse set of inbred founder strains, designed specifically for complextrait analysis. Diversity of gut microbiota was characterized by complementing phylogeneticand distance-based, sequence-clustering approaches. Significant correlations were found betweenthe mouse strains and their gut microbiota, reflected by distinct bacterial communities. Cohabitationand litter had a reduced, although detectable effect, and the microbiota response to these factorsvaried by strain. We identified bacterial phylotypes that appear to be discriminative and strain-specific to each mouse line used. Cohabitation of different strains of mice revealed an interaction ofhost genetic and environmental factors in shaping gut bacterial consortia, in which bacterialcommunities became more similar but retained strain specificity. This study provides a baselineanalysis of intestinal bacterial communities in the eight CC progenitor strains and will be linkedto integrated host genotype, phenotype and microbiota research on the resulting CC panel.The ISME Journal (2012) 6, 2033–2044; doi:10.1038/ismej.2012.54; published online 14 June 2012Subject Category: microbe–microbe and microbe–host interactionsKeywords: Collaborative Cross; intestinal microbial diversity; microbiome; microbiota; pyrosequencing;SSU rRNA gene

Introduction

Throughout their evolutionary history, animalshave been in continuous, direct contact with themicrobial diversity that thrives in all environmentson earth. Specific microbial eco-physiological traitshave led to a wide range of associations betweenmetazoan taxa and members of the bacterial andarchaeal domains. In some cases, extensive geneticcoevolution between the animal host and microbeshas resulted in obligate, highly specific, nutritionalsymbioses involving one or a few vertically trans-mitted microbial species, such as the endosym-bionts of some hydrothermal vent invertebrates and

those of plant sap-feeding insects (Moran, 2007;Dubilier et al., 2008). Even for more complexanimal gut microbial communities, acquired andmaintained dynamically after hatching or birth,there are likely host-microbe specificity determi-nants, as revealed by natural colonization andexperimental microbiota transplantation across hostspecies (Rawls et al., 2004; Rawls et al., 2006;Palmer et al., 2007; Morowitz et al., 2011). Distinctcommunity structure and composition characterizesdifferent vertebrate and invertebrate species in theirnatural environments, global microbiota and inter-species relatedness, reflecting host phylogenyand incorporating elements of developmental andnutritional specialization (Ley et al., 2008a, b;Ochman et al., 2010; Yidirim et al., 2010). Suchcomplex interactions between deterministic (geneticand developmental), environmental and stochasticfactors in the assembly and dynamics of vertebrategut microbiota are being studied intensely, from

Correspondence: M Podar, Biosciences Division, Oak RidgeNational Laboratory, Oak Ridge, TN 37831, USA.E-mail: podarm@ornl.govReceived 8 December 2011; revised 1 May 2012; accepted 1 May2012; published online 14 June 2012

The ISME Journal (2012) 6, 2033–2044& 2012 International Society for Microbial Ecology All rights reserved 1751-7362/12

www.nature.com/ismej

fundamental ecological perspectives to its impact onhost health and disease (Dethlefsen et al., 2006; Leyet al., 2006; Dethlefsen et al., 2007; Palmer et al.,2007; Ley et al., 2008a; Turnbaugh et al., 2009; Reidet al., 2011; Spor et al., 2011).

Significant advances in understanding the indivi-dual roles of host and environmental factors on thecomposition of vertebrate gut microbiota haveresulted from studies on genetically inbred mouselines (reviewed in Spor et al. (2011)). Such studieshave used both conventionally reared and germ-freeanimals inoculated selectively with different bacter-ial isolates or natural microbiota samples. Strongevidence exists that the global host genotypeinfluences specific microbiota composition (betadiversity) (Benson et al., 2010; Kovacs et al., 2011),and mutations and inactivation of specific geneshave been associated with discrete communitychanges, in some cases linked to metabolic diseases(for example, obesity, diabetes and metabolic syn-drome) (reviewed in Spor et al. (2011)). At the sametime, however, studies using embryo transplanta-tion, litter cross-fostering and other variation inmouse rearing and housing have shown that experi-mental manipulations, environmental and stochas-tic factors (for example, founder effects) can exertdominating contributions in microbiota taxonomiccomposition (Friswell et al., 2010). Metagenomicsequencing studies have revealed functionallyequivalent gut communities, with similar genecomposition, that have quite diverse taxonomicstructure (Turnbaugh et al., 2009). Such resultssuggest that physiological interactions, both withthe host and between the microbes, may have adominant role over phylogenetic composition (alphadiversity) of the community. Therefore, linking hostgenetic background with discrete units of themicrobiome (microbial taxa or genes) relies upon acombination of diversity and functional genomic/physiological measurements. With thousands ofsegregating genes and millions of segregating poly-morphisms in mouse populations, comprehensivemapping of potential deterministic associationsbetween host genotype and the hundreds of bacter-ial taxonomic or functional units, as well asdistinguishing environmental and stochastic effects,requires an extensive population genetics andstatistical framework. A recent study using quanti-tative trait locus analysis of advanced intercrosslines identified a subset of microbial lineagesthat cosegregate with host genetic loci (Bensonet al., 2010).

The Collaborative Cross (CC), a large panel ofrecombinant, inbred mouse strains designed by theComplex Trait Consortium, offers a standardizedand reproducible foundation for complex traitanalysis, including microbiota heritability factors(Churchill et al., 2004). The CC will encompass alarge number of inbred strains resulting fromsystematic crossing of eight genetically diversefounder strains that capture B90% of the known

mouse genetic variability (Roberts et al., 2007). TheCC was initiated at several research institutionsincluding Oak Ridge National Laboratory (ORNL)(Chesler et al., 2008; Philip et al., 2011) and recentlyhas been employed in quantitative trait analysis of awide range of phenotypes (Aylor et al., 2011; Philipet al., 2011).

Here we present an analysis of gut microbialcommunity structure associated with the eightfounder strains of the CC and two additional strainsto assess environmental and founder populationeffects using pyrosequencing of two separate regionsof the small subunit ribosomal RNA (SSU rRNA)gene. This study lays the foundation for determiningthe community structure variability in mouse linesresulting from controlled crossing of the founderpopulations at different levels of inbreeding andcorrelating with quantitative host physiologicaland genetic markers.

Materials and methods

MiceMice were bred and housed at the William L andLiane B Russell vivarium at ORNL and at theUniversity of Tennessee (UTK), Knoxville, TN,USA. Mice at ORNL profiled in this study werebred at the facility and weaned at 3–4 weeks afterbirth and distributed in separate cages eitherindividually or with same-gender siblings or non-siblings based on experimental design (Supple-mentary Figure S1) until adult (8–10 weeks of age).The eight parental mouse lines of the CC were used:A/J, C57BL/6J, 129S1/SvImJ, NOD/LtJ, NZO/HILtJ,CAST/EiJ, PWK/PhJ and WSB/EiJ (abbreviated AJ,BL6J, 129S1, NOD, NZO, CAST, PWK and WSB,respectively). Strains were originally obtained fromThe Jackson Laboratory and maintained over nolonger than 10 generations. Because of difficulties inbreeding, mice from the NZO line were the only ageexception, with some 41 year. C3H/Ri and DBA/2JRmice (abbreviated C3HRI and DBAJR, respectively)were also profiled. Replicates of 7–10 mice were usedper strain. Cecum content samples were collected asdescribed in the Supplementary Methods.

For the interstrain cohabitation study, 3-week-oldBL6J and C3HRI mice were purchased from TheJackson Laboratory and were housed in a separatefacility (UTK) until they reached 10 weeks of age, atwhich time they were all euthanized. Thoren cageswith microisolator tops and individual water bottleswere used for this experiment. Separate cagescontained five individuals of only BL6J (cage 1) orC3HRI (cage 4). Cage 2 contained three BL6J andtwo C3HRI mice. Cage 3 contained two BL6J andthree C3HRI mice (Supplementary Figure S1). All themice were fed Harlan Laboratories (Indianapolis,IN, USA) Teklad Rodent Diet 8604, which is similarto Purina Rodent Chow 5053 (high-protein, low-carbohydrate content).

Genetic effects on mouse gut microbiotaJH Campbell et al

2034

The ISME Journal

SSU rRNA gene amplification and pyrosequencingDNA was extracted from cecum contents using aprotocol modified from that of Ley et al. (2008a)(Supplementary Methods). Amplicon libraries of bothV1-2 and V4 regions of 16S SSU rRNA genes wereobtained using barcoded primers and sequenced usinga 454-FLX instrument (Roche, Indianapolis, IN, USA),using 40 samples per plate. Resulting sequenceswere filtered for length, quality and chimeraremoval using the software package mothur(Schloss et al., 2009). High-quality sequences weresubjected to operational taxonomic unit (OTU)-based clustering (Huse et al., 2010) and phylogeny-based analysis using Fast UniFrac (Hamady et al.,2010) to evaluate the effects of host geneticson bacterial community composition. Details ofsequencing and data processing steps are providedin the Supplementary Methods.

Statistical analysesMatrices of OTU-by-sample were imported intoPRIMER-E v6 (Clarke and Gorley, 2006) for down-stream statistical analyses. Raw sequence counts ofeach OTU within each sample were convertedinto percentages, square-root transformed and aBray–Curtis resemblance matrix was calculated.This matrix was used for nonmetric multidimen-sional scaling plots, hierarchical clustering, analysisof similarity and similarity percentage (SIMPER).Permuted (n¼ 9999) multivariate analyses of var-iance were performed on Bray–Curtis matrices usingthe PERMANOVAþ add-on package for PRIMER-E.Permuted calculations of P were used when uniquepermutation values were 4100 and Monte Carlocalculations of P were used when unique permuta-tions were o100 (Clarke and Gorley, 2006). Retro-spective power analyses were performed for eachwithin strain comparison of the sexes. Briefly,critical values (a¼ 0.05) along a t distribution weredetermined for one population of mice in thecomparison and used to determine the overlappingsection of the second population of mice (Sokal andRohlf, 1981). This P-value equals b, and power wascalculated from b (power¼ 1� b).

We used a two-step approach to identify OTUsthat were most influential in differentiating mousestrains in each of the V1-2 and V4 ampliconlibraries. First, SIMPER was used to calculate therelative contribution of each OTU to the overalldissimilarity in each pair of mice. Because of thelarge number of pairwise comparisons, it is difficultto elucidate clear trends. However, we used SIMPERas a data reduction technique to discard OTUs thatdid not contribute at least 0.5% to the dissimilarityof any pair of mice. OTUs in V1-2 and V4 datafound to contribute at least 0.05% to any pairwisedifference in SIMPER comparisons were furtherscreened for differential abundances across strainsusing a discriminant function analysis (DFA) inMatlab (v7.10) using a freely available statistics

toolbox (Strauss, 2010). DFA is a multivariatetechnique used to identify variables (OTUs) thatdistinguish a priori groups (mouse strain). Thus,DFA was used to further reduce V1-2 and V4 OTUmatrices to a suite of OTUs that could be used topredict mouse strain membership. Hierarchicalclustering of strains based upon these predictiveOTUs was performed on Euclidean distances inMatlab.

Sequence depositionNucleotide sequences generated in this study havebeen deposited in the NCBI Sequence Read Archive(Accession no. SRPO12588.1).

Results

Mice representing 10 inbred mouse lines, includingthe 8 progenitors of the CC project, were used todetermine differences in gut microbial diversitylinked to distinct host genetic background.Embedded in this, maternal, sex and cage-sharingeffects were also explored. The mouse lines weremaintained separately but under the same condi-tions at the ORNL facility. Second, to compareeffects of environmental exposure and interstraincontact, we analyzed the gut microbial diversity intwo of the strains raised at a different location andexposed to one another (Supplementary Figure S1).

For the primary study, the cecum microbiota of 94mice were profiled by SSU rRNA gene pyrosequen-cing (Supplementary Table S1). Two regions of SSUrRNA gene (V1-2 and V4) were analyzed to comple-ment differences in taxonomic representation dueto primer bias (Griffen et al., 2012), as well as tocompare and contrast inferred relationshipsbetween the microbiome and the host genetic back-ground. After sequence processing, V1-2 ampliconlibraries contained 293 928 reads (mean of 4982reads/mouse) and V4 libraries contained 605 397reads (mean of 6640 reads/mouse).

Taxonomic analysis of all the sequences usingthe RDP Naıve Bayesian rRNA Classifier (Cole et al.,2009) revealed similar bacterial diversity topreviously observed communities in mouse ceca(Ley et al., 2005), with a dominance of Firmicutes(53–89%) (Supplementary Figure S2). A largedifference was observed for detection of Bacteroi-detes, with the V4 data set containing many fewersequences mapped to that phylum relative to theV1-2 data set (B2% vs 30% median, respectively).Conversely, phyla generally present at low abun-dance in the mouse cecum (o2%), such as Proteo-bacteria, Verrucomicrobia, TM7, Deferribacteria andTenericutes, were detected more efficiently by theV4 than by the V1-2 primer set. Additional taxa weredetected at much lower abundance. For example,the Cyanobacteria-like group (Ley et al., 2005) waspresent in only mice 129S1-352 (39 sequences;

Genetic effects on mouse gut microbiotaJH Campbell et al

2035

The ISME Journal

0.45%) and 129S1-353 (25 sequences; 0.58%) of theV4-amplified microbiota. This same Cyanobacteria-like group was only detected as a single sequencein V1-2 amplicon libraries (mouse 129S1-352).Differences in taxonomic coverage and efficiencyof detection are known to occur between primer sets(Hong et al., 2009; Engelbrektson et al., 2010). Inmany cases, these discrepancies are not predictablebased on sequence complementarity analysis (suchas V4 detection of Bacteroidetes), highlighting theadvantage in targeting more than one SSU rRNAgene region for analyses of taxonomic diversity(Griffen et al., 2012). Analysis of gut microbiotabased on taxonomic classification is limited by thehigh diversity of taxa below the genus level, manywith uncultured relatives, which reduces resolutionof sequence assignment. Therefore, in this study, weprimarily used a taxonomy-independent analysisapproach by classifying the sequences into OTUsbased on sequence similarity (genetic distance).

Amplicon libraries of V1-2 hypervariable regionsof bacterial SSU rRNA gene produced 3821 OTUsacross all samples at 0.03 genetic distance, whereaslibraries of the V4 region produced 1142 OTUsacross all samples at the same genetic distance.Variation observed in the two hypervariable regionsand current analytical methods for such microbialcommunity data led us to adopt a consensusapproach for data analysis. Both OTU-based cluster-ing and phylogenetic (Fast UniFrac) analyses werepursued for all data to ensure that overarchingtrends were not dependent on analytical method.

Strain-wise comparisonsConceptually, analysis of mouse cecum commu-nities by clustering sequences into operational units(OTUs) differs from analysis based on phylogeneticsequence information, but both methods producedsimilar results for both SSU rRNA gene hypervari-able regions, and consistent differences betweenstrains were found. Nonmetric multidimensionalscaling (NMDS) of Bray–Curtis similarity matricesfor OTU-based clustering (Figures 1 and 2) providedsimilar visual separation by strain as principalcoordinates analysis of UniFrac distance matrices(Supplementary Figures S3 and S4). Subsampling toachieve equal sequencing depth for each sampleresulted in slightly lower explained variation foreither hypervariable region, but mice appeared toseparate more clearly by strain with equal sequen-cing depth (Supplementary Figures S3 and S4).Although explained variation was enhanced inUniFrac analysis of V4 sequences, groupwiseseparation of strains was reduced. It is possible thatthis is a result of the larger number of samplessequenced and number of sequences per sample inthe V4 data set. However, it is evident from bothanalyses that BL6J, C3HRI, DBAJR, PWK and WSBstrains harbored distinct microbial assemblages,

whereas individual variation appeared higherwithin 129S1, AJ, CAST, NOD and NZO strains.

Hierarchical clustering was also used to visualizerelationships of individual mice for OTU-based

129S1 AJ BL6J C3HRI CAST

DBAJR NOD NZO PWK WSB

Male Female

2D Stress = 0.19

Figure 1 Nonmetric multidimensional scaling (NMDS) repre-sentation of OTU-based clustering (0.03 genetic distance) of datafrom the V1-2 hypervariable region of SSU rRNA gene. Counts ofeach OTU within each mouse (n¼ 59) were standardized topercentage, square-root transformed and a Bray–Curtis similaritymatrix was calculated.

129S1 AJ BL6J C3HRI CAST

DBAJR NOD NZO PWK WSB

Male Female

2D Stress = 0.21

Figure 2 NMDS representation of OTU-based clustering (0.03genetic distance) of data from the V4 hypervariable region of SSUrRNA gene. Counts of each OTU within each mouse (n¼ 94)were standardized to percentage, square-root transformed and aBray–Curtis similarity matrix was calculated.

Genetic effects on mouse gut microbiotaJH Campbell et al

2036

The ISME Journal

clustering. Branching of V1-2 based on OTUs largelyadhered to strain identification of individuals, withmost strains condensing to discreet nodes of thedendrogram. Strains BL6J and PWK appeared to bemost distinct from other strains with OTUs, butmost strains separated into distinct clades. Someoverlap was seen in one individual of both 129S1and NZO with NOD. NOD mice appeared to be theleast cohesive strain (Supplementary Figure S5).Similarly, the V4 region OTUs (SupplementaryFigure S6) showed clear separation of strains ingood agreement with nonmetric multidimensionalscaling plots, with BL6J appearing most distinct.Again, even though several strains had one indivi-dual outlier, they were quite different from oneanother. Strains C3HRI and DBAJR had the lowestintrastrain variation (Supplementary Figure S6).

Strain-wise separation of UniFrac clusters forV1-2 data (Supplementary Figure S7) was compar-able to OTU-based data. BL6J and NOD mice werebroken into two clusters, and an individual from129S1 and CAST failed to congregate with theirstrains. BL6J appeared to separate by sex. WSBindividuals were more cohesive in this UniFracanalysis than was observed for OTUs (Figure 4).

UniFrac clustering for V4 for individual mice(Supplementary Figure S8) also showed similarresults to OTU data, but OTU-based clusters(Supplementary Figure S6) were separated betterby strain. However, individuals of 129S1, CAST andNZO fragmented into separate clades. Moreover,5 out of 10 strains had at least 1 individual that didnot congregate with their respective strains, but thiscould not be linked to either maternal or cagingfactors and likely reflects stochastic communityassembly.

Due to the difficulty in visualizing three-dimensionalarrangements, box-and-whisker plots of intra- andinterstrain dissimilarities were constructed fromboth V1-2 (Figure 3) and V4 (Figure 4) distancematrices for Bray–Curtis dissimilarities and UniFracdistances (sub-sampled only), and these analysesindicated that mice within all strains were moresimilar to one another than to mice from all otherstrains. Sequences from the V1-2 region displayedgreater variation between mice than V4, but bothanalytical methods produced similar intrastrain andinterstrain relationships. Data from both SSUregions were supportive of one another. In thesesimplified representations, V1-2 and V4 libraries

30

40

50

60

70

80

90

0.4

0.5

0.6

0.7

0.8

129S1 AJBL6J

C3HRICAST

DBAJRNOD

NZOPWK

WSB

Bra

y-C

urtis

Dis

sim

ilarit

yU

niF

rac

Dis

tanc

e

Strain

Figure 3 Box-and-whisker plots of intrastrain (black) andinterstrain (blue) distributional comparisons within V1-2 data.Distributions were formed by parsing strain-wise data from larger(a) Bray–Curtis dissimilarity and (b) UniFrac distance matrices ofmouse-by-mouse comparisons. Outliers are denoted by red pluscharacters (þ ).

129S1 AJBL6J

C3HRICAST

DBAJRNOD

NZOPWK

WSB0.2

0.3

0.4

0.5

0.6

0.7

Strain

Uni

Fra

c D

ista

nce

20

30

40

50

60

70

80

Bra

y-C

urtis

Dis

sim

ilarit

y

Figure 4 Box-and-whisker plots of intrastrain (black) andinterstrain (blue) distributional comparisons within V4 data.Distributions were formed by parsing strain-wise data from larger(a) Bray–Curtis dissimilarity and (b) UniFrac distance matricesof mouse-by-mouse comparisons. Outliers are denoted by redplus characters (þ ).

Genetic effects on mouse gut microbiotaJH Campbell et al

2037

The ISME Journal

both showed C3HRI and DBAJR strains to be themost distinct.

Significance measures calculated around separationof mouse cecum communities (multivariate analysesof variance, MANOVA, and analysis of similarity,ANOSIM) reinforced that mouse strains harboreddistinctly different assemblages. The effect of strain(Fpseudo¼ 5.48; Ppermuted¼ 0.0001) on the V1-2 datawas significant (Table 1), accounting for 38.9% of allvariation. Similarly, strain effects (Fpseudo¼ 8.55;Ppermuted¼ 0.0001) were significant for V4 data(Table 2), accounting for 41.1% of all variation.Pairwise t-tests for individual strains indicated thateach strain differed significantly from other strainsfor both V1-2 and V4 regions (Supplementary TablesS2 and S3). These analyses were supported byanalysis of similarity, in which the V1-2 region(global R¼ 0.818; P¼ 0.001) separated strains withhigher resolution than did the V4 region (globalR¼ 0.795; P¼ 0.001). V1-2 comparisons(Supplementary Table S4) showed C3HRI andPWK to be strongly separated from most otherstrains, and BL6J, DBAJR and NOD also showedlittle overlap with other strains. V4 data(Supplementary Table S5) supported clear distinc-tion of C3HRI and PWK microbiota from otherstrains. However, bacteria detected using this regionof SSU rRNA gene did not strongly separate NZOfrom most other mouse strains.

DFA indicated that relatively few OTUs could beused to reliably predict strain membership. Withinall OTUs detected in V1-2 libraries, SIMPER

analysis found 80 OTUs that explained X0.5% ofthe difference between any two pairwise compar-isons of strains. DFA reduced these OTUs to 44 thatdiffered significantly across mouse strain(Supplementary Table S6). Discriminating OTUswere dominated by uncultured phylotypes amongthe firmicutes (59%) and Bacteroidetes (36%), butone Proteobacteria and one Deferribacteres were alsodifferential across strains. Clustering of only thosediscriminating OTUs (Figure 5) indicated thatsubsets of at least two OTUs could be positivelyassociated with each strain. OTUs showing consis-tently high abundances within a strain usuallyshowed a phylogenetic association, as well. StrainsAJ and BL6J contained differential OTUs found inthe Bacteroidetes. In particular, BL6J contained thehighest abundances of five OTUs most closelyrelated to the genus Barnesiella. Conversely, differ-entially abundant OTUs of strains 129S1, CAST,NOD, NZO and WSB were from the Firmicutes.CAST and NOD mice were enriched for members ofthe Clostridiales. Differential OTUs of C3HRI,DBAJR and PWK lacked phylogenetic relationships.

The same approach was applied to detect differ-entially abundant OTUs for V4 data (SupplementaryTable S7), and again a relatively few OTUs could beused to reliably predict strain membership. Thisregion of the 16S rRNA gene produce 71 differentialOTUs dominated by the phylum Firmicutes (93%),many of which matched most closely to unchar-acterized members of family Lachnospiraceae.A single representative cluster of the phyla Bacter-oidetes, Deferribacteres, Proteobacteria, Tenericutesand TM7 also displayed unequal distributionsacross mouse strains. Again, hierarchical clusteringof just these OTUs (Figure 6) indicated that subsetsof at least two OTUs could be positively associatedwith a mouse strain. With a narrower phylogeneticscope than discriminating OTUs in V1-2 regions, apattern of taxonomic associations by strain is notclear. It is possible that these OTUs are interchange-able with closely related bacteria across the strainssurvey in this study.

Maternal effectsWe used V4 data to investigate the effects ofmaternal lineage on gut microbial communitiesbecause it contains a larger sample size and can,therefore, be considered more comprehensive. Toillustrate the effects of maternal lineage, intrastraindissimilarities (Bray–Curtis) were separated intotwo groups: (1) pairwise distances of siblings and(2) pairwise distances of all non-siblings within astrain. Distributions of non-siblings were plotted(Supplementary Figure S9), and distances ofsiblings were superimposed onto these distribu-tions. Gut communities from siblings were spreadwithout a clear pattern along strain-wise distribu-tions of non-siblings. This suggests that inclusionof siblings with non-siblings had little effect on

Table 1 Permuted multivariate analysis of variance (MANOVA)tests of significance of mouse strain and sex for the V1-2 region

Source Dfa SSb MSc Pseudo-F Uniquepermutations

Pd

Strain 9 59701 6633.5 5.4829 9747 0.0001Sex 1 1858.9 1858.9 1.5364 9862 0.0152Strain�Sex 9 16589 1843.3 1.5235 9605 0.0001Residual 39 4.72Eþ 04 1209.8Total 58 1.26Eþ 05

aDegrees of freedom.bSum of squares.cMean squared.dBolded P-values indicate statistical significance.

Table 2 Permuted multivariate analysis of variance (MANOVA)tests of significance of mouse strain and sex for the V4 region

Source Dfa SSb MSc Pseudo-F Uniquepermutations

Pd

Strain 9 58297 6477.4 8.5513 9766 0.0001Sex 1 1225.7 1225.7 1.6181 9873 0.01Strain�Sex 9 12357 1373 1.8126 9665 0.0001Residual 74 5.61Eþ 04 757.48Total 93 1.30Eþ 05

aDegrees of freedom.bSum of squares.cMean squared.dBolded P-values indicate statistical significance.

Genetic effects on mouse gut microbiotaJH Campbell et al

2038

The ISME Journal

strain-wise bacterial communities. In support of thisconclusion, siblings did not appear to cluster(Supplementary Figure S6) more closely than anyother individuals.

Sex-based comparisonsDistributions of males and females within moststrains overlapped broadly and did not indicatestrongly differential microbial communities in

AJ

C3H

RI

NZ

O

NO

D

129S

1

PW

K

CA

ST

WS

B

DB

AJR

BL6

J

−2.5

−2

−1.5

−1

–0.5

0

0.5

1

1.5

2

2.5

Mouse Strain

OT

U Id

entif

ier

Figure 5 Heatmap of V1-2 OTUs found to vary across strains by discriminant function analysis (DFA). Means of each OTU (n¼44) werecalculated for each strain (n¼ 10). Hierarchial clustering was determined for both dimensions of the heatmap using Euclidean distances.Taxonomic assignments of OTUs can be found in Supplementary Table S6.

AJ

PW

K

CA

ST

NO

D

BL6

J

129S

1

NZ

O

DB

AJR

WS

B

C3H

RI

−2.5

−1.5

−0.5

0.5

1.5

2.5

0

Strain

OT

U Id

entif

ier

Figure 6 Heatmap of V4 OTUs found to vary across strains by DFA. Means of each OTU (n¼ 71) were calculated for each strain (n¼10).Hierarchial clustering was determined for both dimensions of the heatmap using Euclidean distances. Taxonomic assignments of OTUscan be found in Supplementary Table S7.

Genetic effects on mouse gut microbiotaJH Campbell et al

2039

The ISME Journal

males and females. The effect of sex was significantfor both V1-2 (Fpseudo¼ 1.54; Ppermuted¼ 0.015) and V4(Fpseudo¼ 1.62; Ppermuted¼ 0.01), but explained only0.9% and 0.7% of variation in the data, respectively(Tables 1 and 2). Strain-by-sex interactions were alsosignificant for each data set, indicating that malesand females of some mouse strains containeddivergent cecal communities. Retrospective poweranalysis of each comparison (Supplementary TablesS8 and S9) indicated that most t-tests were robust,but some had low sex resolution. Differential cecalcommunities within sex were detected for the BL6Jstrain in V1-2 data (Supplementary Table S8) andfor 129S1, AJ, BL6J, C3HRI and PWK in V4 data(Supplementary Table S9). All individuals of BL6Jwere cocaged with at least one other mouse of thesame sex (Supplementary Table S1), indicating thatseparation of sexes was potentially an artifact.Similarly, one pair of males and females wascocaged within strains 129S1, AJ and C3HRI, andPWK contained a pair of cocaged females. However,strains CAST, NOD, NZO and WSB also had cocagedpairs of the same sex, but did not show significantdifferences in microbial communities. Therefore,sex-based differences could vary with strain, butmore replication is needed for some strains toanswer this definitively.

Cagemate comparisonsSome mice of the same sex and strain were cagedtogether (Supplementary Table S1) and comparedwith mice housed separately to test the effects ofcage environment on the gut microbial community.To analyze this effect, intrastrain dissimilarities(Bray–Curtis) were separated into two groups:(1) pairwise distances of cagemate mice and (2)pairwise distances from all mice kept separately.

Distributions were plotted only for mice thatwere not cocaged, and distances of cagemateswere superimposed onto these distributions(Supplementary Figure S10). Cagemates tended tobe more similar to one another than the majority ofthe isolated mice. This was most evident withinstrains 129S1, AJ and NZO. However, dissimilaritymeasures of most cagemates fell within the ranges ofthose observed for isolated mice. Therefore, overallvariation within strains is of greater magnitude thancocaging effects.

Interstrain cohabitationWe also tested the effects of cohabitation of adultBL6J and C3HRI mice in varied ratios, and againstrain effects appeared to dominate caging effects.Four cages were used for this experiment. Separatecages contained five individuals of only BL6J(cage 1) or C3HRI (cage 4). Cage 2 contained threeBL6J and two C3HRI mice. Cage 3 contained twoBL6J and three C3HRI mice. Mice were purchasedspecifically for this purpose and housed in aseparate facility (UTK) for 8 weeks prior to eutha-nization. Gut communities of mice housed at ORNLdiffered from those at UTK (Figure 7), similar toprevious reports (Friswell et al., 2010). OTU-basedclustering of V4 amplicon libraries found 483 OTUsat a genetic distance of 0.03. An individual mousein BL6J was not closely positioned with any othermice in the experiment. Therefore, it (and allof its unique OTUs) was removed from furtheranalyses. An NMDS plot and hierarchial clus-tering (Figure 7) of these data indicated clearseparation of mice by strain and cage. Hierarchicalclustering also showed clear delineation of miceprimarily by strain and secondarily by cage.Interestingly, cohabitation influenced gut microbial

BL6J - UTK Cage 1

C3HRI - UTK Cage 3

C3HRI - UTK Cage 2

C3HRI - UTK Cage 4

BL6J - UTK Cage 2

BL6J - UTK Cage 3

2D Stress = 0.1

BL6J - ORNL C3HRI - ORNL

Bray-Curtis Similarity100806040

Figure 7 Effects of interstrain cohabitation and housing facility on cecum bacterial communities. BL6J and C3HRI mice were housedseparately (cages 1 and 4) or in cohabitation (cages 2 and 3). Only the V4 hypervariable region was sequenced and OTUs were calculated(0.03 genetic distance) for all mice. Counts of each OTU within each mouse (n¼ 19) were standardized to percentage, square-roottransformed and a Bray–Curtis similarity matrix was calculated and used to produce an (a) NMDS and (b) hierarchical clustering of thegut communities.

Genetic effects on mouse gut microbiotaJH Campbell et al

2040

The ISME Journal

communities, but host genetics appeared to out-weigh this environmental influence. However, micewere cohoused post weaning, possibly renderingtheir microbiota more resistant to change. Therefore,host genetic effects and maternal inoculation couldnot be simultaneously addressed. Further studiesemploying larger populations of mice, temporalsampling and strain cross-fostering would betterdetermine the resilience of established gut commu-nities and the effects of initial colonization.

Discussion

Studies of genetic effects on microbiota areaccumulating in the literature. Some of these studiesaddress fine genetic scales, such as monozygotic,human twins (Turnbaugh et al., 2009) and well-characterized host mutations (Vaahtovuo et al.,2005; Khachatryan et al., 2008). Others haveaddressed the effects of host genetics on the gutmicrobiome on a larger scale with studies of speciesof primates (Ley et al., 2008a; Ochman et al., 2010)and various animals in captivity or the wild (Leyet al., 2008a). In this study, we investigated theeffects of host genetics on cecum microbiota in10 commonly used, inbred strains of laboratorymice, 8 of which are progenitor strains of the CC(Consortium, 2004). Therefore, this study serves as abaseline for determining the nature and extent ofgenetic effects on microbial diversity of these mouselines for future studies of the CC.

Individual variation within strains was observedfor all mouse lines used in this study, but theinfluence of host genetics on bacterial communitiesin the cecum is apparent. This observation wassupported by independent analyses of two regionsof SSU rRNA gene sequence libraries. Individualswithin several strains appear to be more cohesivethan others (for example, C3HRI, DBAJR and WSB),indicating that a gradient of host genetic factorsproduces varied levels of strain-level conformity.Unlike microbial communities of wild primates(Ochman et al., 2010), dendrograms of strain-wiserelationships based on cecum microbiota failed torecapitulate apparent natural histories of the hosts(Petkov et al., 2004; Kirby et al., 2010). Mice of thesame strain purchased from different vendors alsoharbor different microbial communities (Friswellet al., 2010). Therefore, lack of a reflection of thenatural history of the strains in their cecal commu-nities was not surprising.

Other studies have also reported that host geneticsshape gut communities in mice. Two such studies(Benson et al., 2010; Buhnik-Rosenblau et al., 2011)found ties between host genetics and Lactobacillusin mice. Another study (Alexander et al., 2006)in which mice from 23 inbred strains were inocu-lated with and tested for the altered Schaedler’sflora using specific quantitative PCR assays notedsignificant differences for these species. Also, it was

noted that different strains of 129 and BALB micewere similar when supplied through a differentvendor (Alexander et al., 2006). However, Friswellshowed that obtaining the same strain from differentvendors produced varied microbial flora (Friswellet al., 2010). Moreover, Friswell et al. (2010) foundthat C3HRI and BL6J mice harbored distinct micro-bial communities that were more strongly regulatedby host genetics than changes in environment.Interestingly, host genetics could be overcome byimplanting embryos of different strains into asurrogate mother, producing microbial communi-ties within offspring that resembled the surrogatemother (Friswell et al., 2010). Similar to Friswellet al. (2010), we found C3HRI to have low intrastrainvariation and BL6J mice to have high intrastrainvariation. Gut communities in BL10J mice (Lohet al., 2008) have also been shown to vary amongindividuals.

These studies provided a structural basis for us tomore deeply investigate effects of host genetics ongut microbiota. In contrast to employing quantitativePCR (Alexander et al., 2006) or DGGE (Friswellet al., 2010), we used pyrosequencing to compilelibraries of two regions of SSU rRNA gene, allowingus to identify bacterial taxa that were specific tomouse strain. Identification of differentially abun-dant OTUs across mouse strains makes it temptingto speculate as to their potential roles within eachstrain. However, many of the discriminating OTUshave neither cultured relatives nor genomic dataavailable. Moreover, definitive trends in health anddisease cannot be discerned for many taxa closelyrelated to these OTUs. For instance, the genusOscillibacter has been linked to diet in humans(Walker et al., 2011), but this genus appeared to bedifferential in several of our strains, which were fedthe same diet. Also, among the four OscillibacterOTUs found discriminatory in V4 data, no cleartrend across strains was found for this genus, as awhole. Correlations of quantitative trait locusand host gene expression to bacterial diversity datapresented here will likely shed more light onpotential physiological roles of these bacteria inthe mouse cecum. Future isolation and physiologi-cal studies of bacterial taxa that were discriminatoryamong mouse strains will also improve our under-standing of the role of these bacteria. Linkage ofhost genetics, host health/disease and microbialflora will be the ultimate goal of such microbiomestudies, and these mapping studies will enablethe detection of the sources of host molecularvariation and impacts on the intestinal micro-environment.

Our study was not designed to quantify effects ofmaternal lineage on gut microbiota across these10 strains, but we were able to make somecomparisons of siblings to unrelated individuals.Siblings from some strains bore stronger resem-blance to one another than to unrelated mice.However, siblings of other strains were markedly

Genetic effects on mouse gut microbiotaJH Campbell et al

2041

The ISME Journal

dissimilar in microbial flora. Recently, DNA finger-printing techniques revealed no maternal-deriveddifferences in gut bacteria of CC mice (Kovacs et al.,2011). However, Ley et al. (2005) found lineageinfluences to extend to more than one generation. Infact, others have (Hufeldt et al., 2010) observed asufficiently strong effect of maternal lineage thatthey suggested related individuals be used to reducemicrobiome variation for disease studies. Perhapsthe most convincing demonstration of maternaleffects was provided by Friswell et al. (2010).In this study, a BDF1 female was implanted withembryos from BL6J and Agouti strains, resulting inpups with gut bacteria similar to the surrogatemother (Friswell et al., 2010). At this point, factorsdriving maternal differences observed in somestudies and not others remain unclear. In fact, itappeared that maternal influences could be strain-specific, possibly indicating underlying genetic orbehavioral disparities (Alexander et al., 2006).

We did not observe major separation of males andfemales within most mouse strains in this study, butsome minor sex-based differences among indivi-duals in some mouse strains were found. Lack of apredictable response of cecum bacteria to genderdifferences has been substantiated (Spor et al.,2011). Comparatively few studies have addressedsex-associated changes in mouse gut microbiomes.Although more narrow in taxonomic scope than ourstudy, Alexander’s quantitative PCR study(Alexander et al., 2006) detected gender-specificdifferences in only two species of the alteredSchaedler’s flora (species of Firmicutes andClostridium). Moreover, expression of a humancaspase conferred differential susceptibility toListeria monocytogenes infection in male and femaletransgenic mice, apparently due to estrogen interac-tions (Yeretssian et al., 2009). However, Kovacs et al.(2011) found no gender differences in the CC mice oftheir study. As is the case with maternal lineage,undetected genetic or epigenetic factors that werenot adequately controlled could manifest in sex-based differences between strains. Quantitative traitlocus and gene expression analyses of our mice(to be presented elsewhere) have potential forelucidating such mechanisms.

Controlled cohabitation of mice of the same strainoffered the ability to weigh the effects of geneticsagainst environmental pressures. A limited numberof individuals of the same strain were cocaged toevaluate this effect on cecum bacterial communities.Similar to previous reports (Alexander et al., 2006;Loh et al., 2008; Teran-Ventura et al., 2010), somecagemates were more similar to one another thanto isolated mice, but for many no difference wasdetected. This response varied by mouse and wasnot strong for most strains. Teran-Ventura et al.(2010) employed cultivation, fluorescence in situhybridization and terminal restriction fragmentlength polymorphism to detect minor variations inabundances of Enterobacteriaceae, Bacteroides,

Clostridium and Lactobacillus associated withvaried levels of caging isolation (Teran-Venturaet al., 2010). Alexander et al. (2006) also noted thatcage effects appeared to vary by strain, and it wassuggested that behavioral differences in the strains(such as differential coprophagy) could explain thestrain-wise differences.

Results of an interstrain cohabitation experimentwere more informative than observations of intra-strain cohabitation. We did note the same ‘consortialdrift’ (Friswell et al., 2010) between BL6J and C3HRIpopulations used in studies at ORNL and UTK. Whencages containing only one strain were compared withthose containing two strains of mice, genetic strainbest separated mice. This is supported by an experi-ment in Alexander’s study (Alexander et al., 2006), inwhich cages with five inbred strains were monitoredfor members of the altered Schaedler’s flora, demon-strating that host genetics are more influential indetermining host mouse microbial flora than theenvironment. Interestingly, environmental effectswere weaker than underlying host genetics in shapingcecum bacterial communities.

Assessing the causal role of host genetic variationin gut microflora composition and dynamics willenable an understanding of the mechanisms ofcolonization, and in well-characterized mousestrains, the correlation to phenotypes of health anddisease, and will enable comparisons with similarstudies in the human population. Understandingthe mechanisms of community selection and robust-ness of genetic influences on community structurewill have many implications for attempts to altercommunity structures as a therapeutic intervention.Establishing the relationship of microbial commu-nities to the spectrum of variation in physiologicalphenotypes will further our understanding of patho-logical and normal metabolic processes. Emergingmouse resources such as the CC are a powerfulsystem with which to assess these phenomena andwidespread variation in microbial structure.

Acknowledgements

We thank J Mosher, M Shakya, C Brandt, C Schadt (ORNL),M Hauser and J Becker (UTK) for many helpful discus-sions during data collection and analysis. We also thankL Miller (ORNL) for molecular biology assistance. We arealso grateful to P Schloss (mothur; University of Michigan)and R Knight (UniFrac; University of Colorado) forguidance with their respective analytical tools. We alsothank F Bushman (University of Pennsylvania) for criticalevaluation of the manuscript and anonymous reviewersfor helpful suggestions. This research was funded by theUS Department of Energy Office of Science, Biological andEnvironmental Research programs at Oak Ridge NationalLaboratory (ORNL), by the Laboratory Directed Researchand Development Program of ORNL and in part by a grantto MP from the National Human Genome ResearchInstitute (NIH R01-HG004857). ORNL is managed byUT-Battelle, LLC, for the US Department of Energy undercontract DE-AC05-00OR22725.

Genetic effects on mouse gut microbiotaJH Campbell et al

2042

The ISME Journal

References

Alexander AD, Orcutt RP, Henry JC, Baker Jr JBissahoyo A, Threadgill DW. (2006). QuantitativePCR assays for mouse enteric flora reveal strain-dependent differences in composition that are influ-enced by the microenvironment. Mammalian Genome17: 1093–1104.

Aylor DL, Valdar W, Foulds-Mathes W, Buus RJ, VerdugoRA, Baric RS et al. (2011). Genetic analysis of complextraits in the emerging collaborative cross. Genome Res21: 1213–1222.

Benson AK, Kelly SA, Legge R, Ma F, Low SJ, Kim J et al.(2010). Individuality in gut microbiota composition isa complex polygenic trait shaped by multiple envir-onmental and host genetic factors. Proc Natl Acad SciUSA 107: 18933–18938.

Buhnik-Rosenblau K, Danin-Poleg Y, Kashi Y. (2011).Predominant effect of host genetics on levels ofLactobacillus johnsonii bacteria in the mouse gut.Appl Environ Microbiol 77: 6531–6538.

Chesler EJ, Miller DR, Branstetter LR, Galloway LD,Jackson BL, Philip VM et al. (2008). The CollaborativeCross at Oak Ridge National Laboratory: developinga powerful resource for systems genetics. MammGenome 19: 382–389.

Churchill GA, Airey DC, Allayee H, Angel JM, Attie AD,Beatty J et al. (2004). The Collaborative Cross, acommunity resource for the genetic analysis ofcomplex traits. Nat Genet 36: 1133–1137.

Clarke KR, Gorley RN. (2006). PRIMER v6: UserManual/Tutorial. PRIMER-E, Plymouth.

Cole JR, Wang Q, Cardenas E, Fish J, Chai B, Farris RJ et al.(2009). The Ribosomal Database Project: improvedalignments and new tools for rRNA analysis. NucleicAcids Res 37: D141–D145.

Consortium TCT. (2004). The Collaborative Cross, acommunity resource for the genetic analysis ofcomplex traits. Nat Genet 36: 1133–1137.

Dethlefsen L, Eckburg PB, Bik EM, Relman DA. (2006).Assembly of the human intestinal microbiota.Trends Ecol Evol 21: 517–523.

Dethlefsen L, McFall-Ngai M, Relman DA. (2007).An ecological and evolutionary perspective onhuman-microbe mutualism and disease. Nature 449:811–818.

Dubilier N, Bergin C, Lott C. (2008). Symbiotic diversity inmarine animals: the art of harnessing chemosynthesis.Nat Rev Microbiol 6: 725–740.

Engelbrektson A, Kunin V, Wrighton KC, ZvenigorodskyN, Chen F, Ochman H et al. (2010). Experimentalfactors affecting PCR-based estimates ofmicrobial species richness and evenness. ISME J 4:642–647.

Friswell MK, Gika H, Stratford IJ, Theodoridis G, Telfer B,Wilson ID et al. (2010). Site and strain-specificvariation in gut microbiota profiles and metabolismin experimental mice. PLoS One 5: e8584.

Griffen AL, Beall CJ, Campbell JH, Firestone ND,Kumar PS, Yang ZK et al. (2012). Distinct andcomplex bacterial profiles in human periodontitisand health revealed by 16S pyrosequencing. ISME J6: 1176–1185.

Hamady M, Lozupone C, Knight R. (2010). Fast UniFrac:facilitating high-throughput phylogenetic analyses ofmicrobial communities including analysis of pyrose-quencing and PhyloChip data. ISME J 4: 17–27.

Hong S, Bunge J, Leslin C, Jeon S, Epstein SS. (2009).Polymerase chain reaction primers miss half of rRNAmicrobial diversity. ISME J 3: 1365–1373.

Hufeldt MR, Nielsen DS, Vogensen FK, Midtvedt T,Hansen AK. (2010). Family relationship of femalebreeders reduce the systematic inter-individualvariation in the gut microbiota of inbred laboratorymice. Lab Anim 44: 283–289.

Huse SM, Welch DM, Morrison HG, Sogin ML. (2010).Ironing out the wrinkles in the rare biosphere throughimproved OTU clustering. Environ Microbiol 12:1889–1898.

Khachatryan ZA, Ktsoyan ZA, Manukyan GP, Kelly D,Ghazaryan KA, Aminov RI. (2008). Predominant roleof host genetics in controlling the composition of thegut microbiota. PLoS One 3: e3064.

Kirby A, Kang HM, Wade CM, Cotsapas CJ, Kostem E, HanB et al. (2010). Fine mapping in 94 inbred mousestrains using a high-density haplotype resource.Genetics 185: 1081–1095.

Kovacs A, Ben-Jacob N, Tayem H, Halperin E, Iraqi FA,Gophna U. (2011). Genotype is a stronger determinantthan sex of the mouse gut microbiota. Microb Ecol 61:423–428.

Ley RE, Backhed F, Turnbaugh P, Lozupone CA, KnightRD, Gordon JI. (2005). Obesity alters gut microbialecology. Proc Natl Acad Sci USA 102: 11070–11075.

Ley RE, Peterson DA, Gordon JI. (2006). Ecological andevolutionary forces shaping microbial diversity in thehuman intestine. Cell 124: 837–848.

Ley RE, Hamady M, Lozupone C, Turnbaugh PJ, RameyRR, Bircher JS et al. (2008a). Evolution of mammalsand their gut microbes. Science 320: 1647–1651.

Ley RE, Lozupone CA, Hamady M, Knight R, Gordon JI.(2008b). Worlds within worlds: evolution ofthe vertebrate gut microbiota. Nat Rev Microbiol 6:776–788.

Loh G, Brodziak F, Blaut M. (2008). The Toll-like receptorsTLR2 and TLR4 do not affect the intestinal microbiotacomposition in mice. Environ Microbiol 10: 709–715.

Moran NA. (2007). Symbiosis as an adaptive process andsource of phenotypic complexity. Proc Natl Acad SciUSA 104 (Suppl 1): 8627–8633.

Morowitz MJ, Denef VJ, Costello EK, Thomas BC, PoroykoV, Relman DA et al. (2011). Strain-resolved commu-nity genomic analysis of gut microbial colonization ina premature infant. Proc Natl Acad Sci USA 108:1128–1133.

Ochman H, Worobey M, Kuo CH, Ndjango JB, Peeters M,Hahn BH et al. (2010). Evolutionary relationships ofwild hominids recapitulated by gut microbial com-munities. PLoS Biol 8: e1000546.

Palmer C, Bik EM, Digiulio DB, Relman DA, Brown PO.(2007). Development of the human infant intestinalmicrobiota. PLoS Biol 5: e177.

Petkov PM, Ding Y, Cassell MA, Zhang W, Wagner G,Sargent EE et al. (2004). An efficient SNP system formouse genome scanning and elucidating strain rela-tionships. Genome Res 14: 1806–1811.

Philip VM, Sokoloff G, Ackert-Bicknell CL, Striz M,Branstetter L, Beckmann MA et al. (2011). Geneticanalysis in the Collaborative Cross breeding popula-tion. Genome Res 21: 1223–1238.

Rawls JF, Samuel BS, Gordon JI. (2004). Gnotobioticzebrafish reveal evolutionarily conserved responsesto the gut microbiota. Proc Natl Acad Sci USA 101:4596–4601.

Genetic effects on mouse gut microbiotaJH Campbell et al

2043

The ISME Journal

Rawls JF, Mahowald MA, Ley RE, Gordon JI. (2006).Reciprocal gut microbiota transplants from zebrafishand mice to germ-free recipients reveal host habitatselection. Cell 127: 423–433.

Reid G, Younes JA, Van der Mei HC, Gloor GB, Knight R,Busscher HJ. (2011). Microbiota restoration: naturaland supplemented recovery of human microbialcommunities. Nat Rev Microbiol 9: 27–38.

Roberts A, De Villena FP-M, Wang W, McMillian L,Threadgill DW. (2007). The polymorphism architec-ture of mouse genetic resources elucidated usinggenome-wide resequencing data: implications forQTL discovery and systems genetics. Mamm Genome18: 473–481.

Schloss PD, Westcott SL, Ryabin T, Hall JR, Hartmann M,Hollister EB et al. (2009). Introducing mothur: OpenSource, Platform-independent, Community-supportedSoftware for Describing and Comparing MicrobialCommunities. Appl Environ Microbiol 75: 7537–7541.

Sokal RR, Rohlf FJ. (1981). Biometry. W. H. Freeman andCompany: San Francisco.

Spor A, Koren O, Ley R. (2011). Unravelling the effects ofthe environment and host genotype on the gutmicrobiome. Nat Rev Microbiol 9: 279–290.

Strauss RE. (2010). Matlab statistical functions [computersoftware]. Retrieved 1 April 2011, from http://www.faculty.biol.ttu.edu/Strauss/Matlab/matlab.htm.

Teran-Ventura E, Roca M, Martin M, Abarca M, MartinezV, Vergara P. (2010). Characterization of housing-related spontaneous variations of gut microbiota andexpression of toll-like receptors 2 and 4 in Rats.Microb Ecol 60: 691–702.

Turnbaugh PJ, Hamady M, Yatsunenko T, Cantarel BL,Duncan A, Ley RE et al. (2009). A core gut microbiomein obese and lean twins. Nature 457: 480–484.

Vaahtovuo J, Korkeamaki M, Munukka E, Viljanen MK,Toivanen P. (2005). Quantification of bacteriain human feces using 16S rRNA-hybridization, DNA-staining and flow cytometry. J Microbiol Methods 63:276–286.

Walker AW, Ince J, Duncan SH, Webster LM, Holtrop G, ZeX et al. (2011). Dominant and diet-responsive groupsof bacteria within the human colonic microbiota.ISME J 5: 220–230.

Yeretssian G, Doiron K, Shao W, Leavitt BR, Hayden MR,Nicholson DW et al. (2009). Gender differences inexpression of the human caspase-12 long variantdetermines susceptibility to Listeria monocytogenesinfection. Proc Natl Acad Sci USA 106: 9016–9020.

Yidirim S, Yeoman CJ, Sipos M, Torralba M, Wilson BA,Goldberg TL et al. (2010). Characterization of the fecalmicrobiome from non-human wild primates revealsspecies specific microbial communities. PLoS One 5:e13963.

Supplementary Information accompanies the paper on The ISME Journal website (http://www.nature.com/ismej)

Genetic effects on mouse gut microbiotaJH Campbell et al

2044

The ISME Journal