Captivity humanizes the primate microbiomeJonathan B. Claytona,b, Pajau Vangayc, Hu Huangc, Tonya Wardd, Benjamin M. Hillmanne, Gabriel A. Al-Ghalithc,Dominic A. Travisf, Ha Thang Longb,g, Bui Van Tuanb, Vo Van Minhh, Francis Cabanai, Tilo Nadlerj, Barbara Toddesk,Tami Murphyl, Kenneth E. Glanderm, Timothy J. Johnsona, and Dan Knightsd,e,1

aDepartment of Veterinary and Biomedical Sciences, University of Minnesota, Saint Paul, MN 55108; bGreenViet Biodiversity Conservation Center, Danang59000, Vietnam; cBioinformatics and Computational Biology, University of Minnesota, Minneapolis, MN 55455; dBiotechnology Institute, Universityof Minnesota, Saint Paul, MN 55108; eDepartment of Computer Science and Engineering, University of Minnesota, Minneapolis, MN 55455; fDepartment ofVeterinary Population Medicine, University of Minnesota, Saint Paul, MN 55108; gFrankfurt Zoological Society, 60316 Frankfurt, Germany; hDepartment ofEnvironment and Biology, Danang University of Education, Danang 59000, Vietnam; iWildlife Nutrition Centre, Wildlife Reserves Singapore, 729826,Singapore; jEndangered Primate Rescue Center, Cuc Phuong National Park, Nho Quan, Ninh BÌnh, 40000, Vietnam; kPhiladelphia Zoological Garden,Philadelphia, PA 19108; lComo Park Zoo & Conservatory, Saint Paul, MN 55103; and mDepartment of Evolutionary Anthropology, Duke University, Durham,NC 27708

Edited by Lora V. Hooper, The University of Texas Southwestern, Dallas, TX, and approved July 19, 2016 (received for review November 6, 2015)

The primate gastrointestinal tract is home to trillions of bacteria,whose composition is associated with numerous metabolic, auto-immune, and infectious human diseases. Although there isincreasing evidence that modern and Westernized societies areassociated with dramatic loss of natural human gut microbiomediversity, the causes and consequences of such loss are challengingto study. Here we use nonhuman primates (NHPs) as a modelsystem for studying the effects of emigration and lifestyle disrup-tion on the human gut microbiome. Using 16S rRNA gene sequenc-ing in two model NHP species, we show that although differentprimate species have distinctive signature microbiota in the wild, incaptivity they lose their native microbes and become colonized withPrevotella and Bacteroides, the dominant genera in the modernhuman gut microbiome. We confirm that captive individuals fromeight other NHP species in a different zoo show the same pattern ofconvergence, and that semicaptive primates housed in a sanctuaryrepresent an intermediate microbiome state between wild and cap-tive. Using deep shotgun sequencing, chemical dietary analysis,and chloroplast relative abundance, we show that decreasing di-etary fiber and plant content are associated with the captive pri-mate microbiome. Finally, in a meta-analysis including publishedhuman data, we show that captivity has a parallel effect on theNHP gut microbiome to that of Westernization in humans. Theseresults demonstrate that captivity and lifestyle disruption causeprimates to lose native microbiota and converge along an axis to-ward the modern human microbiome.

human microbiome | primate microbiome | dietary fiber | dysbiosis |microbial ecology

Humans with metabolic disorders, autoimmune and inflam-matory conditions affecting the intestines, colorectal cancer,

and infectious diseases often exhibit dysbiosis, a state of micro-bial imbalance (1). Petersen and Round described three categoriesof dysbiosis: loss of beneficial microbial organisms, expansion ofpathobionts or potentially harmful microorganisms, and loss ofoverall microbial diversity, all of which can occur simultaneously(2). In addition to the evident effect of lifestyle on the patho-physiology of many diseases (3), there is mounting evidence that anintimate interplay exists between the gut microbiota and the de-velopment of diseases, including obesity (4–6), Crohn’s diseaseand ulcerative colitis (7), diabetes (8–10), nonalcoholic fattyliver disease (11), Kwashiorkor (12), and many others. Thus,understanding how lifestyle affects the development and main-tenance of the gut microbiome is an important health issue.Specific environmental and genetic factors have been linked to

the development of dysbiosis, including antibiotic use, stress,geography, race, host genetics, and diet (13, 14). Of these factors,diet, gastrointestinal motility, and medication history most stronglyshape the gut microbiome (13, 15–19). Furthermore, previousstudies examining dietary patterns suggest a strong association

between Western lifestyle and dysbiosis (3, 13, 20), especiallyconsidering that in Westernized countries, diet-related chronicdiseases are the largest contributors to morbidity and mortality(3, 21), affecting more than 50% of the population. Dysbiosis inWesternized countries is thought to be mainly a result of diet, asthe Western diet is evolutionarily discordant from the diet of an-cestral humans (3, 21) and tends to be high in fat and animalprotein (e.g., red meat), high in sugar, and low in plant-based fiber(3, 13, 20). The consequences for humans of living in a state ofevolutionary discordance with the native primate diet may includedevelopment of disease, increased morbidity and mortality, andreduction in reproductive success (3, 21).Nonhuman primates (NHPs) are the most biologically relevant

research animal models for humans, as they are our closest livingrelatives. In this study, we sought to explore the relationship be-tween lifestyle and the gut microbiome in NHPs in an effort tobetter understand primate conservation and the etiology ofmodern human dysbiosis. In a unique study design, we examinedthe microbial communities of two NHP species, the red-shankeddouc (Pygathrix nemaeus) and the mantled howling monkey(Alouatta palliata), in different individuals living in captive, semi-captive, and wild conditions. We also examined an additionalnovel captive NHP population comprising eight different species,in addition to previously published human populations living in

Significance

Trillions of bacteria live in the primate gut, contributing tometabolism, immune system development, and pathogen re-sistance. Perturbations to these bacteria are associated withmetabolic and autoimmune human diseases that are prevalentin Westernized societies. Herein, we measured gut microbialcommunities and diet in multiple primate species living in thewild, in a sanctuary, and in full captivity. We found that cap-tivity and loss of dietary fiber in nonhuman primates are as-sociated with loss of native gut microbiota and convergencetoward the modern human microbiome, suggesting that par-allel processes may be driving recent loss of core microbialbiodiversity in humans.

Author contributions: J.B.C., D.A.T., K.E.G., T.J.J., and D.K. designed research; J.B.C., H.T.L.,B.V.T., V.V.M., F.C., T.N., B.T., T.M., K.E.G., T.J.J., and D.K. performed research; J.B.C., P.V.,H.H., T.W., B.M.H., G.A.A.-G., F.C., and D.K. analyzed data; and J.B.C., P.V., T.W., D.A.T., T.J.J.,and D.K. 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.

Data deposition: The sequence reported in this paper has been deposited in the EuropeanBioinformatics Institute database (project no. PRJEB11414).1To whom correspondence should be addressed. Email: dknights@umn.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1521835113/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1521835113 PNAS Early Edition | 1 of 6

ECOLO

GY

the United States (i.e., Western) and in developing nations (i.e.,non-Western). Both red-shanked doucs and mantled howlingmonkeys are folivorous NHPs, consuming a diet that is nutri-tionally poor and difficult to digest compared with diets consumedby nonfolivores. These species are rarely housed in captivity, inpart because of the challenge of replicating their wild diets. Ourability to sample from captive, semicaptive, and wild populationsfrom the same species gave us a unique opportunity to study therelationship between lifestyle and disturbance of the native gutmicrobiota in primates.

ResultsCaptivity Reproducibly Alters the Primate Microbiome. We per-formed amplicon sequencing of the V4 region of the 16S rRNAgene on fecal samples collected from captive and wild red-shanked doucs (n = 93) and captive and wild mantled howlingmonkeys (n = 56). We obtained samples from wild doucs inVietnam and from captive doucs in two zoos in different conti-nents (Southeast Asia, United States). Captive and wild howlersamples were collected in Costa Rica. We first examinedmicrobiome composition differences according to captivity sta-tus. To determine whether significant differences in gut micro-biomes were present between captive and wild populations, wecalculated unweighted UniFrac distances between all samples(22). This distance metric has been effective previously for dis-tinguishing both highly divergent and subtly divergent microbialecosystems (23). Examination of a principal coordinates analysisplot revealed that although the gut microbiomes of wild NHPpopulations (doucs and howlers) are highly divergent, captivitycauses them to converge toward the same composition (Fig. 1and SI Appendix, Fig. S1). This is true despite the highly distinctdiet, gut physiology, and geographical location of the two spe-cies, and is true across three independent zoos in three countries.We use unweighted UniFrac distances throughout our analysis,as they provide much better clustering of our experimental databy population than weighted UniFrac or Bray Curtis distances(SI Appendix, Fig. S2). This indicates that the clustering is likelydriven by presence or absence of key taxa in different pop-ulations, rather than by shifts in the ratios of dominant membersof the microbiota.To confirm this pattern of microbiome convergence, we col-

lected samples from an additional captive population comprising33 individuals from eight different species, housed at a singlezoological institution in the United States (n = 33), differentfrom the US institution housing the sampled doucs. We foundthat the novel captive population had experienced a similar trendof convergence toward the same captive microbiome state(Fig. 1). To further test the hypothesis that this convergencetoward the captive microbiome was a result of the major dis-ruptions to diet and lifestyle associated with captivity, we obtainedmicrobiome samples from 18 individual doucs housed at a primatesanctuary in Vietnam. These “semicaptive” animals are exposed toa less severe form of captivity than the fully captive animals. Theyare provided with diets consisting mostly of local plants obtaineddaily from the jungle by caregivers, but these plants do not rep-resent the full diversity of plants the animals would normally eat inthe wild, the animals are kept in large caged enclosures, and theanimals also have increased exposure to humans relative to wildanimals. However, they are not given antibiotics or other medi-cines and are not fed the manufactured or refined diets typical ofzoos. One of the distinguishing features of our analysis of theseanimals is that they belong to the same species (red-shanked douc)as the wild and captive individuals to which we are comparingthem. Notably, we found these animals to have an intermediatelevel of disruption to the gut microbiota, falling approximately inbetween the wild doucs and the captive doucs in our distance-based analysis (Fig. 1), suggesting that the level of severity of di-etary and lifestyle disruption is associated with the level of

disruption to the native gut microbiota. Analysis-of-similarities(ANOSIM) statistical analysis confirmed that NHP gut micro-bial communities grouped by captivity status (ANOSIM R = 0.69;P = 0.001), and a random-forest classifier obtained 100% cross-validation accuracy at discriminating wild douc, wild howler,semicaptive, and captive groups, with the sole exception of therecently captured howler monkey, which classified with the wildhowlers (Fig. 2 and SI Appendix, Figs. S3 and S4). These analysesconfirmed that the wild primate populations had unique signaturemicrobiota that were lost in captivity. An additional meta-analysisincluding data previously published by Muegge et al. alsoconfirmed that captive primates lose their native microbialbiodiversity and converge toward a similar perturbed state(SI Appendix, Fig. S5) (17).

The Captive Primate Microbiome Is Characterized by Loss of Diversity.To investigate gut microbial diversity, including previously un-known taxa, we performed open-reference operational taxonomicunits (OTUs) picking on the NHP samples. We tested for signifi-cant differences in within-individual biodiversity between primatesin different living conditions. Adding to previously published re-sults describing decreased NHP diversity in captivity (24), we foundthat gut microbial diversity decreased in accordance with theseverity of captivity, with the highest alpha diversity observed inNHPs living under the most natural conditions (i.e., wild), thelowest alpha diversity seen in the NHPs living under the mostunnatural conditions (i.e., captive), and an intermediate numberof OTUs seen in the semicaptive individuals (t test wild vs.

−0.4 −0.3 −0.2 −0.1 0.0 0.1 0.2 0.3

−0.

4−

0.3

−0.

2−

0.1

0.0

0.1

0.2

PC1

PC

2

Douc

Howler

2 days incaptivity

DoucHowlerOtherWildSemi−captive wild−bornSemi−captiveCaptiveNon−western HumanUSA Human

Fig. 1. The captive primate microbiome converges toward the modernhuman microbiome. Principal coordinates plot of unweighted UniFrac dis-tances showing ecological distance between gut microbial communities inwild, semicaptive (from a sanctuary), and captive NHPs, as well as non-Westernized humans, and humans living in the United States (i.e., West-ernized). All samples were obtained with the same protocol for V4 16S rRNAsequencing. Although in wild populations the douc and howler microbiomesare highly distinctive, captivity causes them to converge toward the samecomposition. Semicaptive doucs (green) fall in between wild and captivedoucs along the same axis of convergence. The axis of convergence con-tinues toward non-Westernized human populations (Malawi and Venezuela),and finally to the modern US human microbiome. One howler individual re-sembling wild howlers was sampled after only 2 d in captivity, indicating thatthe transition to the captive microbiome state requires more than 2 d. Semi-captive doucs born in the wild have similar microbiomes to their captivity-borncounterparts, indicating that transition to captivity from the wild is sufficientto produce the captivity-related microbiome.

2 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1521835113 Clayton et al.

semicaptive, P = 2.1 × 10−21; semicaptive vs. captive, P = 1.9 × 10−5)(Fig. 3A). Rarefaction curves showed a similar trend (SI Appendix,Fig. S6).

Factors Influencing Captive Primate Microbiome Convergence. Sev-eral major lifestyle changes known to influence the gut micro-biome are associated both with captivity in NHPs and withmodernization in humans. These changes include dietary shifts,disease, changes in geography, and exposure to modern medicineand hygiene, including antibiotics. We performed several analysesto determine whether these factors, or host genetics, are associ-ated with convergence toward humans and loss of microbiomediversity in captive primates.First, we examined the relative effects of geography and lo-

cation on microbiome variation. Including the additional 33samples from the second US zoo, our samples span four differentcaptive facilities in three different countries, all demonstratingthe same trend of convergence toward the modern humanmicrobiome in captive primates. We found that, although thereis a strong effect of geography and location on the microbiota ofthe captive individuals, with interzoo microbiome distances beingsignificantly greater than intrazoo microbiome distances, there isan even greater difference between captive and wild doucmicrobiomes (Fig. 3B) (permutation test on unweighted UniFracdistances; P < 0.001). We also note that the two captive doucpopulations are in very divergent geographical locations (UnitedStates and Southeast Asia), yet their microbiomes are moresimilar to each other than to the wild doucs. We confirmed ina meta-analysis, including samples from additional species in

additional zoos (17), that although captive animals do clustertogether with animals from the same zoo, all zoo animals tend tobe more similar to each other, even between zoos, than they areto the wild animals (SI Appendix, Fig. S5). In addition, we notethat although the captive and wild howler populations have ap-proximately the same geographical location, we observed sig-nificant microbiome perturbations in the captive howlers. Thus,we find that geography and location alone are not sufficient toexplain the observed changes in captive NHP microbiomes.We also considered the effects of host genetics on the micro-

biome as a potential confounder. However, an important addi-tional component of our study design is that we have controlledfor host genetics at the species level by obtaining samples fromwild and captive individuals within the same species, replicatedacross two different species (howler and douc). Although thisdoes not control for within-species individual genetic variationthat may cause dysbiosis (14), microbiome variation betweencaptive individuals of the same species is smaller than variationbetween captive and wild individuals from the same species (Fig.3B). Interestingly, we do find a correlation between host phy-logeny and host microbiome variation within a given zoologicalinstitution (permutation test, P < 0.0001) (SI Appendix, Fig. S7),as noted previously in wild NHPs (25). However, this associationis potentially confounded by diet and is less strong than the as-sociations of microbiome state with zoological institution andcaptivity in general. Thus, we find that host genetics, althoughlikely affecting the primate microbiome, has a much smaller ef-fect on the microbiome than does captivity.We followed several lines of study to examine the effects of

dietary changes on the captive primate microbiome and foundevidence that dietary perturbation is likely a primary driver ofcaptive primate microbiome composition. Recent studies inhumans and mice have supported the hypothesis that loss ofnatural dietary fiber causes a loss of native microbial diversityand a shift toward the modern Westernized human microbiome(26, 27). To quantify dietary perturbation in a subpopulationof wild, captive, and semicaptive doucs, we collected samplesof typical plants and other dietary components for chemical

Tannerella

Collinsella

Uncl. Coriobacteriaceae 1

Methanosphaera

Uncl. Bacteroidales 2

[Prevotella]

Prevotella

Lactobacillus

Parabacteroides

Phascolarctobacterium

Desulfovibrio

Bacteroides

Methanobrevibacter

Uncl. Christensenellaceae

Uncl. Streptophyta

Uncl. Coriobacteriaceae 2

Clostridium 1

Akkermansia

Dorea

Oscillospira

Uncl. Anaeroplasmataceae

Uncl. ML615J−28

Uncl. Rickettsiales

Dehalobacterium

rc4−4

Uncl. mitochondria

Adlercreutzia

CaptiveSemi-

CaptiveWildDouc

WildHowler

Individuals

Bac

teria

l gen

era

Fig. 2. Discriminative taxa by species and captivity status. Highly predictivetaxa for discriminating the captive, semicaptive, wild douc, and wild howlergroups. Predictive taxa included are those with a 0.001 mean decrease in ac-curacy when removed as determined by the random forests classifier. Theclassifier was able to correctly predict the group label for unseen samples forall but one sample (99.6% accuracy). Semicaptive individuals show an overlapbetween some of the wild douc genera and some of the captive animalgenera. Captive NHPs contain large blocks of bacterial genera that are absentfrom wild NHPs, and vice versa, indicating a massive shift in taxonomic com-position between the populations. Although wild howlers tended to overlapmore with captive NHPs than did the wild doucs, both the doucs and howlerslose most of their native signature gut bacteria in captivity.

Est

imat

ed O

TU

ric

hnes

s(C

hao-

1 es

timat

or)

***

**

Captiv

e

Semi−c

aptiv

eW

ild

020

0040

0060

0080

00

Unw

eigh

ed U

niF

rac

dist

ance

Captive−to−wild(same species)

Between−species(same zoo)

Between−zoo(same species)

Within−zoo(same species)

0.4

0.5

0.6

0.7

0.8 *** ***

*****

A B

Fig. 3. Captivity reduces native primate microbiota. (A) Bar plot of mean andspread of gut microbial biodiversity, as measured by the number of species-likeOTUs in the gut microbiome, of wild, semicaptive, and captive douc pop-ulations of NHPs, using the Chao1 estimator of total OTU richness. This indi-cates a significant loss of biodiversity from wild populations to semicaptivepopulations, and again from semicaptive to captive. Error bars indicate SD, andasterisks denote significance at *P < 0.05, **P < 0.01, and ***P < 0.001.(B) Standard box plot of microbiome variation (unweighted UniFrac distance)explained by different experimental factors, showing that captivity in generalis associated with a greater change in microbiome state than variation in hostspecies, zoological institution, or individual.

Clayton et al. PNAS Early Edition | 3 of 6

ECOLO

GY

analysis. We tracked wild doucs over the course of ∼9 mo andobserved them feeding on 57 different plant species, whereasthe semicaptive doucs were offered 43 different plant speciesover the course of 1 y. In contrast to the high dietary diversity(i.e., number of plant species) consumed by the wild and sem-icaptive doucs, the captive doucs were fed far fewer plantspecies, and thus consumed a much less diverse diet. Specifi-cally, the Southeast Asia zoo doucs were offered ∼15 plantspecies and the US zoo doucs were offered only one plantspecies over the course of 1 year. We measured crude protein,crude fat, soluble sugars, acid detergent fiber, and neutral de-tergent fiber, in addition to several minerals, in the typical dietsof each of these populations (wild, semicaptive, captive in Asia,captive in the United States). For neutral detergent fiber con-tent in the wild and semicaptive douc diets, we used previouslymeasured values from Ulibarri (28) and Otto (29), respectively(SI Appendix, Table S1). Although we do not have specific di-etary chemical analysis for each individual animal in thesepopulations, dietary content tends to be homogeneous withinthese captive and wild primate groups (see SI Appendix fordetailed discussion of captive primate diet homogeneity). Wethen tested the hypothesis that decreased total neutral de-tergent fiber is associated with loss of native microbiota incaptive primates and convergence away from the wild doucmicrobiome (ANOVA, P = 1.4 × 10−74). We found that pop-ulations consuming high-fiber diets had microbiomes moresimilar to those of wild doucs, and populations consuming low-fiber diets had microbiomes more similar to those of modernhumans (Fig. 4A). Functional pathways predicted using Phy-logenetic Investigation of Communities by Reconstruction ofUnobserved States (PICRUSt) (30) indicated that wild NHP(notably douc) microbiomes possessed increased metabolism ofpyruvate, butanoate, glycerophospholipid, and propanoate(random forests feature importance score > 0.01), consistent

with increased plant fiber degradation (SI Appendix, Fig. S8).However, these predicted functional profiles should beregarded as suggestive only because of the lack of accounting forunknown species and functional variability within OTU clusters.We also estimated total raw plant dietary content for all NHPs

and human subjects, using chloroplast sequences observed in the16S amplicon sequencing data. We found that chloroplast con-tent was substantial in wild populations (∼3% of all ampliconsequences, on average), but that chloroplast content decreasedin semicaptive and captive animals and was nearly completelyabsent from the human samples and US-based captive primatesamples (Fig. 4B). Finally, we performed deep shotgun meta-genomic sequencing (28.2 ± 6.6 million sequences per sample)on a subset of 30 captive and wild howlers and doucs and alignedthe sequences to known plant reference genomes at 97% iden-tity. Although the existing plant reference genomes were not thesame species as those being consumed by the wild primates, wenonetheless identified a large proportion of fecal DNA se-quences in the wild primates that were homologous to regions ofknown cultivated plant genomes, ranging up to ∼40% of allobserved sequences. In contrast, we observed almost no plantDNA in the captive samples, with the one exception of an in-dividual howler monkey that had been rescued for electrical burntreatment within the last 48 h who had levels similar to those ofthe wild animals (Mann–Whitney U test, P < 0.0001) (Fig. 4C).This same animal was also an outlier in the principal coordinatesanalysis, clustering with the wild howler monkeys (Fig. 1). PlantDNA in captive NHP microbiomes also represented lower di-etary plant alpha diversity than in wild NHP microbiomes(Mann–Whitney U test of Shannon index, P < 0.0001) (SI Ap-pendix, Fig. S9).We considered whether transition to captivity can produce

major microbiome perturbations in an adult individual. Weobtained birth location records for the semicaptive douc

−0.4 −0.2 0.0 0.2 0.4

−0.

6−

0.4

−0.

20.

00.

2

PC1

PC

2

0

3%

Primary chloroplast variation axis

Chloroplast

ratio

Total dietary fiber (%)

0.2

0.3

0.4

0.5

0.6

12.6 32.0 35.6 53.7

Sim

ilarit

y to

wild

dou

cs(1

−un

w. U

niF

rac

dist

ance

)F

ract

ion

plan

t DN

A0.

00.

10.

20.

30.

4

Captive Wild

Blue−eyed black lemurDe Brazzas monkey DoucEmperor tamarinGeoffroys tamarin

GorillaHowlerHumanOrangutan SakiSpider monkey

A B

C

Fig. 4. Dietary fiber and gut microbiome in NHPs. (A) Mean ecological similarity of a given individual douc’s microbiome to all wild douc microbiomes,plotted against that individual’s estimated dietary fiber content (black); same, but showing mean ecological similarity to humans (blue). Doucs consumingmore fiber more closely resemble wild doucs in their microbiome; doucs consuming less fiber more closely resemble humans. (B) The same samples plotted inFig. 1, colored by host species. Also shown is the primary axis of correlation of chloroplast relative abundance, a proxy measurement for dietary raw plantcontent, with sample positions. A smoothed local regression curve for chloroplast ratio (i.e., relative abundance) along the primary axis of variation showsdecreasing raw plant consumption from wild, to semicaptive, to captive primates, with almost no raw plant consumption in the humans or US captiveprimates. (C) Fraction of whole-genome shotgun data aligning at 97% identity to known plant genomes for 14 captive individuals (9 douc, 5 howler) and 16wild individuals (8 douc, 8 howler). Wild individuals have a high fraction of plant DNA in their stool; captive individuals have almost none, with the exceptionof a single outlier individual who was recently rescued from the wild for treatment of electrical burns (also an outlier in Fig. 1).

4 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1521835113 Clayton et al.

individuals and determined that approximately half of them wereborn in the wild (eight of 18). This provided us with an oppor-tunity to test the hypothesis that transition to captivity has causedthe observed disruption to the natural microbiota in the adultindividuals. If the within-individual transition from wild tosemicaptive caused only part of the observed disruption to thegut microbiota, then we would expect the wild-born sanctuary-housed doucs to resemble the wild doucs more closely than dothose born in the sanctuary. However, we found that wild-borndoucs were no closer to wild doucs than were the captive-borndoucs (unweighted UniFrac permutation test, P = 0.663) (Fig. 1).This supports the hypothesis that transition to captivity can causethe observed microbiome disruption in these animals and indi-cates that a substantial part of the captivity-related microbiomeperturbation can be acquired within an individual’s lifetime. Thismay have implications in the study of migration-related dysbiosisin humans.We examined medical records for differences in antibiotic use

and disease between individuals that could be affecting or af-fected by the microbiome. Sadly, five of the nine captive doucindividuals we sampled died of gastrointestinal-related diseasesincluding wasting and gastroenteritis within the following year.We tested whether individuals who died had microbiomes moredivergent from the wild douc microbiome than their captivecounterparts within the same zoo. We observed a positive trend,but there were only five deceased and four living individuals, andthe trend was not significant (t test, P = 0.35) (SI Appendix, Fig.S10). This suggests there may be an association between thecaptive primate microbiome and disease, but the findings are notconclusive and require larger sample size for proper testing.Finally, we determined that 11 of the 33 animals sampled from

the second US zoo had never been prescribed antibiotics in theirlives, based on their complete medical records. In contrast, theremaining 22 individuals had received an average of 4.6 ± 4.0courses throughout life. We used this information to test whetherthe microbiomes of the 22 individuals that had received antibi-otics more closely resembled modern human microbiomes, butthey did not (t test, P = 0.5507), nor were they lower in diversityor less similar to wild primate microbiomes (SI Appendix, Fig.S11). This finding suggests that antibiotics are not a strong driverof the convergence of the captive primate microbiome towardthe modern human microbiome.

Captivity in Primates Partially Parallels Modernization in Humans.Using standardized methodologies across studies (i.e., sequenc-ing of the V4 region), we were able to conduct a meta-analysiscombining NHP populations with previously published samplesfrom adult humans living in both Westernized (United States,n = 129) and non-Westernized (Malawi and Venezuela, n = 21,34, respectively) countries (23). This analysis demonstrated thatthe axis of convergence seen in Fig. 1 continues from wild tocaptive NHPs, and then toward non-Westernized humans, andfinally to Westernized humans, suggesting that a similar loss ofsignature biodiversity seen in captive NHPs has taken place ashumans adapted to modern society. Interestingly, we observedhigher relative abundances in captive primates of Bacteroides andPrevotella, the two dominant human gut microbiome generacompared with in wild and semicaptive NHPs (Fig. 5). In addi-tion, a higher relative abundance of Bacteroides was seen inWesternized humans compared with non-Westernized humans.The high relative abundance of Bacteroides seen in both captiveNHPs and Westernized humans is suggestive that captivitymoves captive NHPs in the same direction along the Bacteroidesgradient as does Westernization in humans.

DiscussionOur analysis of species-matched wild, semicaptive, and captiveindividuals in two different species (red-shanked douc and

mantled howler monkey) demonstrates that NHPs lose sub-stantial portions of their signature microbiota in captivity andthat they become colonized by human-associated gut bacterialgenera Bacteroides and Prevotella. Species in both Bacteroidesand Prevotella are capable of polysaccharide degradation, sug-gesting their emergence may be associated with a shift in di-versity or types of dietary polysaccharides, rather than withoverall loss of dietary fiber. In an additional captive NHP pop-ulation, represented by 33 individuals across eight species at anindependent zoo, we found that captive rearing caused theseindividuals to converge toward the same perturbed microbiomestate as the captive doucs and howlers. Thus, we found thatmultiple populations of captive primates converged along an axisleading away from the wild primate microbiome state and towardthe modern human microbiome state.It is important to note that we do not know whether the

microbiome perturbations we observe contribute to captive pri-mate disease or are merely a consequence of gastrointestinaldisease caused by other factors. However, perturbation of the gutmicrobiome has known causal and predictive roles in humangastrointestinal diseases, particularly in predisposing individualsto infection (31–33). This suggests there may be a broader rolefor maintaining keystone gut microbial genetic and functionaldiversity in the conservation of NHPs, although further researchis needed to determine which microbiome perturbations repre-sent dysbiosis, and which are merely benign or even beneficialadaptations to changes in lifestyle and diet. Previous studies haveshown that changes in diet are directly associated with shifts ingut microbial community structure (18, 20, 34). By leveraging ourstudy design and zoological medical records, we were able to ruleout geography, host genetics, antibiotics exposure, and birth incaptivity as the primary determinants of the captive primatemicrobiome. In contrast, chemical dietary analysis, deep shotgunsequencing, and tracking of fecal chloroplast ratios pointed toloss of dietary plant fiber as a primary driver of captive primatemicrobiome perturbation.Recent studies have shown that modern humans have lost a

substantial portion of their natural microbial diversity (34–36).Given the massive loss of gut microbiome diversity in captiveprimates in this study, captive NHPs may provide an informativemodel for understanding the effects of modernization and masshuman migration on the development of human diseases linkedto diet and the microbiome, such as obesity and diabetes.Our meta-analysis including Westernized and non-Westernizedhuman microbiomes suggests that loss of dietary fiber may be

Bacteroides0.0 0.1 0.2 0.3 0.4

Wild Douc

Wild Howler

Semi−captive Douc

Captive

Non−western Human

USA Human

Prevotella0.0 0.2 0.4 0.6 0.8

Relative abundance

Fig. 5. Captive primates acquire Bacteroides and Prevotella, the dominantgenera in the modern human gut microbiome. A bee swarm plot of the arc-sine square root relative abundance of bacterial genera Bacteroides and Pre-votella, the two dominant modern human gut microbiome genera, shown inwild, semicaptive, and captive NHPs, as well as in non-Westernized andWesternized humans. As populations shift toward the captive state, theirmicrobiomes become colonized by dominant human gut bacterial genera.

Clayton et al. PNAS Early Edition | 5 of 6

ECOLO

GY

driving loss of core microbial biodiversity in humans and captiveprimates alike.

Materials and MethodsStudy Subjects and Sample Material. Fecal samples (n = 251) were collectedopportunistically immediately after defecation from two wild, one semi-captive, and three captive NHP populations located around the globe be-tween 2009 and 2013. Ten different NHP species were sampled in this study,including the Red-shanked douc (Pygathrix nemaeus), Mantled howlingmonkey (Alouatta palliata), Western lowland gorilla (Gorilla gorilla gorilla),Sumatran orangutan (Pongo abelii), De Brazza’s monkey (Cercopithecusneglectus), Black-handed spider monkey (Ateles geoffroyi), White-faced saki(Pithecia pithecia), Blue-eyed black lemur (Eulemur flavifrons), Emperortamarin (Saguinus imperator), and Geoffroy’s tamarin (Saguinus geoffroyi).A detailed breakdown of population groups is included in the SI Appendix.

The research in this study complied with protocols approved through theUniversity of Minnesota Institutional Animal Care and Use Committee andadhered to the legal requirements of the countries in which it was conducted.

Bacterial 16S rDNA Extraction, Amplification, and Sequencing. The bacterial16S rRNA gene was extracted and amplified using primers 515F and 806R,which flanked the V4 hypervariable region of bacterial 16S rRNAs, followingthe Earth Microbiome Project protocol (37). Amplicons were sequenced onan Illumina MiSeq in 2 × 250 paired-end mode.

Data Analysis. Raw forward-read sequences were analyzed with QIIME 1.8.0pipeline (38) with standard settings, but keeping those sequences with150 bp < length < 1,000 bp and mean quality score > 25. Reverse reads werediscarded because of low average sequencing quality (SI Appendix, Fig. S12).Preprocessed sequences were clustered at 97% nucleotide sequence

similarity level, using closed reference OTU picking against the Greengenes13_8 reference database (39) for the meta-analyses ,including publishedhuman data, and using open-reference OTU picking for the analysis of alphadiversity in NHPs. All the sequences were rarefied to 14,100 reads per samplefor the downstream analysis. Taxonomy assignment, diversity analysis, andprincipal coordinate analysis were obtained through QIIME with defaultsettings and using custom R scripts. Published data used in our meta-analysis(17, 23, 40, 41) were analyzed similarly. Microbiome-based phylogenies forSI Appendix, Fig. S7 were generated using the ape package in R (42).

ACKNOWLEDGMENTS. We thank Dr. Michael Murtaugh and Dr. HerbertCovert for critically reading the manuscript and providing feedback; Dr. LisaCorewyn, Dr. Chris Vinyard, Dr. Susan Williams, Dr. Mark Teaford, andDr. Marta Salas, who were vital in the collection of individualized fecalsamples from the howlers at La Pacifica in Costa Rica; our field assistants, AiNguyen Tam and Nguyen Dat, for their help with sample collection andprocessing in Vietnam; the Endangered Primate Rescue Center, PhiladelphiaZoo, Singapore Zoo, and Como Zoo for providing fecal samples fromsemicaptive and captive nonhuman primates; Tran Van Luong, NguyenVan Bay, and Nguyen Manh Tien for their permission to work in Son TraNature Reserve and for their continued support and help; Kieu Thi Kinh andThai Van Quang for help in obtaining the research permits; the Departmentof Forest Protection, the Danang University, and the Son Tra Nature Reservefor granting the research permits; and Christina Valeri and James Collins atthe University of Minnesota Veterinary Diagnostic Laboratory for theirassistance with acquiring and maintaining shipping permits. This researchwas funded in part by the Morris Animal Foundation through a VeterinaryStudent Scholars grant; the University of Minnesota through a College ofVeterinary Medicine Travel grant; Merial/NIH through a Summers Scholarsgrant; the Margot Marsh Biodiversity Foundation; the Mohamed bin ZayedSpecies Conservation Fund; and the National Institutes of Health through aPharmacoNeuroImmunology Fellowship (NIH/National Institute on Drug AbuseT32 DA007097-32).

1. Yang Y, Jobin C (2014) Microbial imbalance and intestinal pathologies: Connectionsand contributions. Dis Model Mech 7(10):1131–1142.

2. Petersen C, Round JL (2014) Defining dysbiosis and its influence on host immunity anddisease. Cell Microbiol 16(7):1024–1033.

3. Hold GL (2014) Western lifestyle: A ‘master’ manipulator of the intestinal microbiota?Gut 63(1):5–6.

4. Turnbaugh PJ, et al. (2006) An obesity-associated gut microbiome with increasedcapacity for energy harvest. Nature 444(7122):1027–1031.

5. Turnbaugh PJ, et al. (2009) A core gut microbiome in obese and lean twins. Nature457(7228):480–484.

6. Turnbaugh PJ, Bäckhed F, Fulton L, Gordon JI (2008) Diet-induced obesity is linked tomarked but reversible alterations in the mouse distal gut microbiome. Cell HostMicrobe 3(4):213–223.

7. Gevers D, et al. (2014) The treatment-naive microbiome in new-onset Crohn’s disease.Cell Host Microbe 15(3):382–392.

8. Giongo A, et al. (2011) Toward defining the autoimmune microbiome for type 1 di-abetes. ISME J 5(1):82–91.

9. Brown CT, et al. (2011) Gut microbiome metagenomics analysis suggests a functionalmodel for the development of autoimmunity for type 1 diabetes. PLoS One 6(10):e25792.

10. Boerner BP, Sarvetnick NE (2011) Type 1 diabetes: Role of intestinal microbiome inhumans and mice. Ann N Y Acad Sci 1243:103–118.

11. Henao-Mejia J, et al. (2012) Inflammasome-mediated dysbiosis regulates progressionof NAFLD and obesity. Nature 482(7384):179–185.

12. Smith MI, et al. (2013) Gut microbiomes of Malawian twin pairs discordant forkwashiorkor. Science 339(6119):548–554.

13. Hawrelak JA, Myers SP (2004) The causes of intestinal dysbiosis: A review. Altern MedRev 9(2):180–197.

14. Knights D, et al. (2014) Complex host genetics influence the microbiome in in-flammatory bowel disease. Genome Med 6(12):107.

15. Wu GD, et al. (2011) Linking long-term dietary patterns with gut microbial enter-otypes. Science 334(6052):105–108.

16. Hildebrandt MA, et al. (2009) High-fat diet determines the composition of the murinegut microbiome independently of obesity. Gastroenterology 137(5):1716–1724.

17. Muegge BD, et al. (2011) Diet drives convergence in gut microbiome functions acrossmammalian phylogeny and within humans. Science 332(6032):970–974.

18. Falony G, et al. (2016) Population-level analysis of gut microbiome variation. Science352(6285):560–564.

19. Zhernakova A, et al.; LifeLines cohort study (2016) Population-based metagenomicsanalysis reveals markers for gut microbiome composition and diversity. Science352(6285):565–569.

20. Martinez-Medina M, et al. (2014) Western diet induces dysbiosis with increased E coliin CEABAC10 mice, alters host barrier function favouring AIEC colonisation. Gut 63(1):116–124.

21. Cordain L, et al. (2005) Origins and evolution of the Western diet: Health implicationsfor the 21st century. Am J Clin Nutr 81(2):341–354.

22. Lozupone C, Knight R (2005) UniFrac: A new phylogenetic method for comparing

microbial communities. Appl Environ Microbiol 71(12):8228–8235.23. Yatsunenko T, et al. (2012) Human gut microbiome viewed across age and geogra-

phy. Nature 486(7402):222–227.24. Uenishi G, et al. (2007) Molecular analyses of the intestinal microbiota of chimpan-

zees in the wild and in captivity. Am J Primatol 69(4):367–376.25. Ochman H, et al. (2010) Evolutionary relationships of wild hominids recapitulated by

gut microbial communities. PLoS Biol 8(11):e1000546.26. Sonnenburg ED, et al. (2016) Diet-induced extinctions in the gut microbiota com-

pound over generations. Nature 529(7585):212–215.27. O’Keefe SJD, et al. (2015) Fat, fibre and cancer risk in African Americans and rural

Africans. Nat Commun 6:6342.28. Ulibarri LR (2013) The socioecology of red-shanked doucs (Pygathrix nemaeus) in Son

Tra Nature Reserve, Vietnam. PhD dissertation (University of Colorado, Boulder).29. Otto C (2005) Food intake, nutrient intake, and food selection in captive and semi-

free douc langurs. PhD dissertation (University of Cologne, Cologne).30. Langille MGI, et al. (2013) Predictive functional profiling of microbial communities

using 16S rRNA marker gene sequences. Nat Biotechnol 31(9):814–821.31. Weingarden A, et al. (2015) Dynamic changes in short- and long-term bacterial

composition following fecal microbiota transplantation for recurrent Clostridium

difficile infection. Microbiome 3:10.32. Montassier E, et al. (2016) Pretreatment gut microbiome predicts chemotherapy-

related bloodstream infection. Genome Med 8(1):49.33. Fujimura KE, Slusher NA, Cabana MD, Lynch SV (2010) Role of the gut microbiota in

defining human health. Expert Rev Anti Infect Ther 8(4):435–454.34. Martínez I, et al. (2015) The gut microbiota of rural Papua New Guineans: Compo-

sition, diversity patterns, and ecological processes. Cell Reports 11(4):527–538.35. Moeller AH, et al. (2014) Rapid changes in the gut microbiome during human evo-

lution. Proc Natl Acad Sci USA 111(46):16431–16435.36. Clemente JC, et al. (2015) The microbiome of uncontacted Amerindians. Sci Adv 1(3):

e1500183.37. Gilbert JA, et al. (2010) Meeting report: The terabase metagenomics workshop and

the vision of an Earth microbiome project. Stand Genomic Sci 3(3):243–248.38. Caporaso JG, et al. (2010) QIIME allows analysis of high-throughput community se-

quencing data. Nat Methods 7(5):335–336.39. DeSantis TZ, et al. (2006) Greengenes, a chimera-checked 16S rRNA gene database

and workbench compatible with ARB. Appl Environ Microbiol 72(7):5069–5072.40. Rollo F, Ermini L, Luciani S, Marota I, Olivieri C (2006) Studies on the preservation of

the intestinal microbiota’s DNA in human mummies from cold environments. Med

Secoli 18(3):725–740.41. Tito RY, et al. (2012) Insights from characterizing extinct human gut microbiomes.

PLoS One 7(12):e51146.42. Paradis E, Claude J, Strimmer K (2004) APE: Analyses of phylogenetics and evolution in

R language. Bioinformatics 20(2):289–290.

6 of 6 | www.pnas.org/cgi/doi/10.1073/pnas.1521835113 Clayton et al.

Supplemental Text and Methods Random forests classifier and cross-validation estimation of accuracy Using the random forests classifier, we determined the most discriminative genus-level taxa between wild, semi-captive, and captive NHPs. We then assessed the accuracy of the classifier using 10-fold cross-validation. In other words, we trained the classifier on 90% of the samples, and then used the discovered signatures to predict which populations the remaining 10% of samples belonged, and then repeated the process 10 times. This analysis revealed that individual primate populations have such distinct signature microbiomes that they can be identified from their microbiota with an estimated 99.6% accuracy (Figure 2). From this analysis we identified the bacterial genera most important for distinguishing the populations (random forests feature importance score ≥ 0.01). We found that wild NHPs possessed higher relative abundances of a variety of microbes, including Collinsella, Tannerella, Oscillospira, Coprococcus, etc., while captive NHPs possessed higher relative abundances of Bacteroides, Prevotella, Parabacteroides, Treponema, etc. Compared to wild and captive NHPs, our population of semi-captive NHPs possessed higher relative abundances of a variety of microbes, including Akkermansia, Turicibacter, Methylobacterium, and other taxa. Importance of concordant data generation methods for meta-analysis A major challenge in performing the meta-analyses combining microbiome data from multiple sources is the presence of batch effects due to study bias from different processing methods. To extract meaningful patterns from our comparison of NHP and human microbiome data required joint analysis of multiple wild populations of species, together with previously published human data. Previous work has characterized gut microbiome variation in a number of specific groups of primate species, such as chimpanzees, African apes, and baboons (1, 2). However, these published data are generated using varied approaches to sample storage, DNA extraction, amplification, and DNA sequencing, impeding efforts toward large-scale meta-analyses. Quantitation of microbiome data requires application of consistent, standardized methods to avoid batch effects. In this study, all NHP fecal samples were obtained by our group using the same protocols, were processed using comparable methods, and were sequenced at the same sequencing facility using the same method (i.e., EMP method; V4 region) as published human data (3), which resulted in wild, semi-captive, and captive NHP microbiome samples that were amenable to meta-analysis. Detailed discussion of captive primate diet homogeneity Captive NHP diets are dramatically different from those of wild or even semi-wild NHPs. Diets fed to captive NHPs are typically generalized and rarely species-specific. NHP species are regularly categorized as folivores, frugivores, and omnivores based on the dietary niche they occupy. In captive settings, this results in NHPs being given very similar, if not identical diets (4). However, even if wild primates do belong in similar feeding guilds, different feeding ecologies, morphologic and physiologic adaptations, and habitats all contribute to varied nutrient intake in the wild (5), typically rich in plant fibers. In contrast, the recommended diet of most captive NHPs is based primarily on corn and soy. It is high in fat (5%) and protein (23%) while low in fiber (14%) (6) when compared to the diet of wild leaf-eating NHPs that has approximately 0% fat, 10-13% protein and 23-54% fiber (7). Unlike corn and soy, tree leaves contain plant secondary compounds (such as alkaloids, phenolics, and cyanide) in addition to nutrients (8–12). For example the Golden Bamboo lemur consumes four times the human-lethal dose of cyanide every day (13), indicating highly specialized digestive capability across different primates. Thus, loss of dietary fiber content and fiber diversity in captive NHPs is a likely contributor to their concomitant loss of gut health and microbial diversity. Detailed discussion of emergence of Bacteroides One of the marked effects of captivity on the gut microbiome of NHPs, as well as Westernization on the gut microbiome of humans, is an increase in relative abundance of Bacteroides. Using both wild ape and human microbiomes, Moeller et al. (2014) determined that Bacteroides has increased in relative abundance in humans living in the USA greater than fivefold since their divergence from other human populations. The bacterial genus Bacteroides has a known positive association with the consumption of a diet rich in animal fat and protein (1, 14), which are major components of a

Western diet. A Western diet is considered to be a diet high in fat and animal protein (e.g., red meat), high in sugar, and low in plant-based fiber (15–17). Previous studies examining the relationship between dietary patterns and dysbiosis suggest a strong association between Western lifestyle, notably diet, and a dysbiotic gut microbiome (14, 15, 17), as the Western diet is evolutionarily discordant from the diet of ancestral humans (15, 18). Taken together, the relative abundance of Bacteroides in the gut appears to be strongly regulated by dietary intake. Breakdown of Douc and Howler sample population Doucs: Fecal samples (n = 111) were collected from captive (n = 27 samples, 9 individuals), semi-captive (n = 18 samples, 18 individuals), and wild (n = 66 samples from 7 known individuals and 39 unknown individuals) red-shanked doucs (Pygathrix nemaeus) between 2012-2013. One captive population was located at the Philadelphia Zoo in the USA while another was located at the Singapore Zoo in SE Asia. Doucs housed at the Endangered Primate Rescue Center in Ninh Binh, Vietnam served as the semi-captive population. Doucs inhabiting Son Tra Nature Reserve, Da Nang, Vietnam (16°06’—16°09’N, 108°13’—108°21’E) served as the wild population in this comparative study (19). Howlers: Fecal samples (n = 56) were collected from captive (n = 5 samples, 5 individuals) and wild (n = 51 samples, 28 known individuals, 17 unknown individuals) mantled howling monkeys (Alouatta palliata) between July-August 2010. Howlers inhabiting the forests of Hacienda La Pacifica, which is a privately owned cattle/tilapia farm of approximately 2,000 hectares located at the base of the Cordillera de Tilaran in the Province of Guanacaste, Costa Rica (latitude 10°28’N, longitude 85°07’W), served as the wild subjects (20, 21). Howlers housed at Las Pumas Rescue Center, which is located within Hacienda La Pacifica, served as the captive subjects. The remaining eight NHP species sampled consisted of captive individuals housed at the Como Zoo in Saint Paul, MN (Supplemental Table 2).

Supplemental Figures

Figure S1. Primate microbiome clustering by captivity, location, and species does not depend on inclusion of USA zoo #2 samples. Principal coordinates plot of unweighted UniFrac distances between all primate and samples shown in Figure 1, excluding the 33 samples from the second USA zoo. Similar to the plot in Figure 1, this plot shows separation of the wild doucs and howlers, and convergence of the douc and howler microbiomes toward the same state in captivity. Principal coordinates analysis was performed only on the subset of samples to demonstrated that the observed clustering was not driven by the second USA zoo samples.

Figure S2. Captive primate dysbiosis converges toward the modern human microbiome. Principal coordinates plot of (a) weighted UniFrac and (b) Bray-Curtis distances between all samples, showing ecological distance between gut microbial communities in wild, semi-captive (from a sanctuary), and captive nonhuman primates, as well as non-westernized humans, and humans living in the USA (i.e., Westernized). Unweighted UniFrac (Figure 1) provided much stronger clustering of our experimental data by population than weighted UniFrac or Bray Curtis distances, indicating that the clustering is likely driven by presence or absence of key taxa in different populations, rather than by shifts in the ratios of dominant members of the microbiota. These distances based on relative abundance show captive primates overlapping more with the non-westernized modern humans.

Figure S3. Heatmap of most predictive taxa discriminating gut microbiomes of two wild primate species. The transformed relative abundance of the 20 most strongly predictive bacterial genera for discriminating between two species of primates, the red-shanked douc and the mantled howling monkey, as determined by the random forests classifier. This heatmap shows very strong signature species for each of these groups.

Figure S4. Captivity reduces native primate microbiota. Standard box plot of microbiome variation (unweighted UniFrac distance) explained by different experimental factors, showing that captivity in general is associated with a greater change in microbiome state than variation in host species, zoological institution, or individual. (c) Stacked bar plot of relative abundance of the 20 most abundant genera across all wild and captive douc and howler individuals. Bars above zero correspond to genera more prevalent in wild primates; bars below zero correspond to genera more common in captive primates.

Figure S5. Captivity causes nonhuman primate microbiomes to converge toward the same compositional state. Principal coordinates plot of genus-level unweighted UniFrac distances showing ecological distance between gut microbial communities in wild, semi-captive (from a sanctuary), and captive nonhuman primates plotted by population location (left panel) and captivity status (right panel). Novel NHP microbiomes (doucs, howlers, and Como Zoo population) are based on V4 16S sequences (R1 only), and previously published NHP microbiomes (Saint Louis Zoo and Namibia_Wild) by Muegge et al. (2011) are based on V2 16S sequences. Although in wild populations the douc and howler microbiomes are highly distinctive, captivity causes them to converge toward the same composition. Semi-captive doucs (green) fall in between wild and captive doucs along the same axis of convergence. The inclusion of additional captive nonhuman primate populations, represented by 14 distinctive primate species (Como Zoo and Saint Louis Zoo), further highlights the convergence that occurs when primates are kept in captivity. We note that although location is an important driver of microbiome variation (left panel), the effect of zoo location is smaller than the overall effect of captivity. This is also shown in Figure 3b.

Figure S6. Rarefaction curves for different primate groups. (left) Chao-1 estimator as a measure of alpha diversity; (right) Observed number of OTUs as a measure of alpha diversity. Red: wild; blue: semi-captive; orange: captive. We also compared diversity between groups of different captivity status using data at the full rarefaction depth (14100 sequences/sample). Dropping singleton OTUs present in only one sample, the wild NHPs (2544.3 ± 390.9 OTUs) harbored the highest number of OTUs (i.e., greatest diversity), followed by the semi-captive NHPs (2141.4 ± 293.0 OTUs), and captive NHPs (1967.9 ± 538.2 OTUs). By this metric, wild NHPs had significantly higher diversity than captive or semi-captive (t-test p = 2.1×10-21, 1.9×10-5, respectively), and captive had higher diversity than semi-captive (t-test p = 0.040). We repeated this analysis with the Chao1 estimator of the true number of OTUs (i.e., species richness) in our samples. Using the Chao1 estimator differences were significant between all three populations (t-test p = 4.6×10-39, 7.3×10-8, 0.0099 for wild vs. captive, wild vs. semi-captive, and captive vs. semi-captive, respectively) (see Figure 3a).

Figure S7. Association of microbiome and host phylogeny. (a) host phylogeny from Perelman et al. (22) for the 10 unique primate species sampled in this study. (b) microbiome phylogeny of all captive, semi-captive, and wild individuals according to unweighted UniFrac distance, using the Nei-Saitou neighbor-joining method (23). This shows major clustering by zoological institution location, and minor nested clustering by host species. (c) Host phylogeny for subset of species present in USA Zoo #2. (d) Microbiome phylogeny as in (b) for samples from USA Zoo #2. This shows some concordance with (c) but concordance is hard to assess due to multiple individuals sampled per species. (e) Microbiome phylogeny as in (d) but depicting a phylogeny built from the average distances between species, averaging over individuals, for comparison with (c). (f) Direct comparison of 8-species host phylogeny (c) with 8-species microbiome-based phylogeny (e), showing significant concordance between the two phylogenies (p < 0.0001, permutation test of cophenetic distance (23) when permuting species labels).

Figure S8. Heatmap of KEGG level 2 metabolic pathways that discriminate the four major populations. Only pathways with random forests feature importance > 0.01 are shown. Wild NHPs (notably doucs) possessed higher relative abundances of a variety of pathways, including pyruvate metabolism, butanoate metabolism, glycerophospholipid metabolism, and propanoate metabolism (random forests feature importance score > 0.01), consistent with increased plant fiber degradation.

Figure S9. Dietary plant diversity and the non-human primate microbiome. Estimated dietary plant diversity (Shannon index) in captive and wild doucs and howlers based on whole-genome shotgun data aligning at 97% identity to known plant genomes. These data include 14 captive individuals (9 douc, 5 howler) and 16 wild individuals (8 douc, 8 howler). Wild individuals have higher dietary plant diversity based on plant DNA in their stool (Mann-Whitney U test, p < 0.001), despite the fact that whole genome reference databases only contain a small number of cultivated plant genomes.

Figure S10. Severity of captive douc dysbiosis and risk of mortality. Beeswarm plot with median and upper/lower quartiles of average unweighted UniFrac distance from each captive douc to all wild douc individuals, stratified by survival status at 1 year post-sampling. Distances are normalized to the mean within each captive population in order to adjust for microbiome variation associated with sampling location. Individuals who died tended to have microbiomes more divergent from the wild douc microbiome than their captive counterparts within the same zoo, but there were only 5 deceased and 4 living individuals and the trend was not significant (t-test, p = 0.35).

Figure S11. Antibiotic exposure and captive primate microbiome dysbiosis. (a) Beeswarm plot of mean ecological similarity of individual captive primates to all modern humans, stratified by antibiotic exposure (USA zoo #2 samples only). This shows that antibiotic exposure in captive primates was not associated with having a more human-like microbiome. 11 individual animals sampled from the second USA zoo had never taken antibiotics, while the remaining 22 individuals had received an average of 4.6 +/- 4.0 courses throughout life. (b) as in (a) but showing unweighted UniFrac distance of the individuals in the left panel to wild primates. (c) Shannon diversity of 33 captive individuals in the USA Zoo #2, 11 of whom had never had antibiotics. There is no statistical difference between the two groups, indicating that lifetime exposure to antibiotics is not related to captive primate microbiome diversity. These results demonstrate that lifetime exposure to antibiotics is not likely to be causing captive primate microbiome diversity.

Figure S12. Sequencing quality scores in forward and reverse reads. Forward-read quality scores (a) and Reverse-read quality scores (b) plotted against nucleotide position. Forward-read score lower standard deviation drops consistently below q=25 at approximately 185 bases; reverse-read score lower standard deviation drops consistently below q=25 at approximately 100 bases.

Supplemental Tables Supplemental Table 1. Nutrient content from dry matter aggregate dietary material for each population.

Diet component Wild EPRC Southeast Asia Zoo1 USA Zoo

Crude Protein (%) 9.46 16.52 13.37 16.7 Crude Fat (%) * 3.23 3.12 3.71 Soluble Sugars (%) 2.7 2.28 * 7.9 Acid detergent Fiber (%) 46.762 23.23 23.07 8.65 Neutral detergent Fiber (%) 53.672 35.63 31.97 12.64 Calcium (%) 0.49 1.05 0.22 0.72 Potassium (%) 0.96 0.76 0.21 0.29 Sodium (%) 0.01 0.01 0.24 0.27 Zinc (mg/kg) 19.4 10.16 8.25 26.3 Iron (mg/kg) 26.5 33.73 20.74 64.33 1Southeast Asian zoo diet also included a vitamin and mineral supplement which was not included in the analysis. 2,3Values marked with (2) are from Ulibarri (2013) and those marked with (3) are from Otto (2005) as NDF and ADF were not available from the laboratory analyses for these diets. Data from Ulibarri (2013) were weighted according to feeding season to represent the proportion of plants part selected. *Not detected.

Supplemental Table 2. Primate species and associated lifestyles included in this study.

References 1. Moeller AH, et al. (2014) Rapid changes in the gut microbiome during human evolution. Proc Natl Acad Sci U S A

111(46):16431–16435.

2. Tung J, Rudolph J, Altmann J, Alberts SC (2007) Parallel effects of genetic variation in ACE activity in baboons and humans. Am J Phys Anthropol 134(1):1–8.

3. Yatsunenko T, et al. (2012) Human gut microbiome viewed across age and geography. Nature 486(7402):222–227.

4. Kleiman DG, Thompson KV, Baer CK (2010) Wild Mammals in Captivity: Principles and Techniques for Zoo Management, Second Edition (University of Chicago Press).

5. Kaumanns W, Hampe K, Schwitzer C, Stahl D Primate nutrition: towards an integrated approach (Filander Verlag, Furth, The Netherlands).

6. Leaf-Eater Primate Diet - Biscuit # 5M02 - 25 lb Mazuri Shopp Cart. Available at: http://www.mazuri.com/mazurileaf-eaterprimatediet-1-1.aspx [Accessed May 31, 2016].

7. Glander KE (1981) Feeding patterns in mantled howling monkeys. Available at: http://dukespace.lib.duke.edu/dspace/handle/10161/7083 [Accessed May 31, 2016].

8. Ehrlich PR, Raven PH (1964) Butterflies and Plants: A Study in Coevolution. Evolution 18(4):586–608.

9. Feeny PP (1968) Effect of oak leaf tannins on larval growth of the winter moth Operophtera brumata. J Insect Physiol 14(6):805–817.

10. Janzen DH (1969) Seed-Eaters Versus Seed Size, Number, Toxicity and Dispersal. Evolution 23(1):1–27.

11. Rhoades DF, Cates RG (1976) Toward a General Theory of Plant Antiherbivore Chemistry. Biochemical Interaction Between Plants and Insects, Recent Advances in Phytochemistry., eds Wallace JW, Mansell RL (Springer US), pp 168–213.

12. Rosenthal AZ, Matson EG, Eldar A, Leadbetter JR (2011) RNA-seq reveals cooperative metabolic interactions between two termite-gut spirochete species in co-culture. ISME J 5(7):1133–42.

13. Consumption of cyanogenic bamboo by a newly discovered species of bamboo lemur - Glander - 2005 - American Journal of Primatology - Wiley Online Library Available at: http://onlinelibrary.wiley.com/doi/10.1002/ajp.1350190205/abstract [Accessed May 31, 2016].

14. Wu GD, et al. (2011) Linking Long-Term Dietary Patterns with Gut Microbial Enterotypes. Science 105(2011). doi:10.1126/science.1208344.

15. Hold GL (2014) Western lifestyle: a “master” manipulator of the intestinal microbiota? Gut 63(1):5–6.

16. Hawrelak JA, Myers SP (2004) The causes of intestinal dysbiosis: a review. Altern Med Rev J Clin Ther 9(2):180–197.

17. Martinez-Medina M, et al. (2014) Western diet induces dysbiosis with increased E coli in CEABAC10 mice, alters host barrier function favouring AIEC colonisation. Gut 63(1):116–124.

18. Cordain L, et al. (2005) Origins and evolution of the Western diet: health implications for the 21st century. Am J Clin Nutr 81(2):341–354.

19. Lippold LK, Thanh VN (2008) The Time is Now: Survival of the Douc Langurs of Son Tra, Vietnam. Primate Conserv 23(1):75–79.

20. Glander KE Habitat and resource utilization: An ecological view of social organization in mantled howling monkeys.

21. Glander K, Nisbett R (1996) Community structure and species density in tropical dry forest associations at Hacienda La Pacífica in Guanacaste province, Costa Rica.

22. Perelman P, et al. (2011) A molecular phylogeny of living primates. PLoS Genet 7(3):e1001342.

23. Paradis E, Claude J, Strimmer K (2004) APE: Analyses of Phylogenetics and Evolution in R language. Bioinforma Oxf Engl 20(2):289–290.