Abstract
Microbiome studies show that host taxon, diet, and environment influence gut bacteria. However, these factors are rarely studied in animal hybrids and exudivores (which nutritionally exploit indigestible oligosaccharides). To investigate the effects of host taxon, hybridization, and environment on gut microbiota, we conducted 16S V4 ribosomal sequencing of the gut microbiome of marmosets (Callithrix), non-human primate (NHP) specialist exudivores that also hybridize. We sampled 59 wild, translocated, and captive pure and hybrid Callithrix, including endangered C. aurita. Gut microbiome diversity differed significantly between hybrids and non-hybrids, but host environment had the strongest overall effect on the gut microbiome. Captive marmosets showed relatively reduced gut microbiome diversity. Wild Callithrix had the highest relative abundance of Bifidobacterium, which process host-indigestible carbohydrates, while captive marmosets had the highest relative abundance of Enterobacteriaceae, a family containing several pathogenic bacteria. The wild marmoset gut microbiome was enriched predictively for carbohydrate metabolism functions, while that of captive marmosets was enriched for nucleotide and amino acid metabolism function. Our findings show that carbohydrate metabolism is integral to the composition and function of the wild exudivore gut microbiome. Further, captivity perturbs the exudivore gut microbiome, raising implications for captive host health and endangered exudivore conservation.
Introduction
The role of the gut microbiome in host physiology and health has been studied widely in humans [1–4] and increasingly in non-human primates (NHPs) and other animals [5–9]. While these studies show that host taxon, diet, and environment are key influences on animal gut bacteria, hybrid and exudivore hosts are underrepresented in such work. Hybridization may perturb host fitness by disrupting interactions between the host genome and gut microbiome from breakage of parental co-adapted gene complexes in hosts [10–11]. Exudivores subsist mainly on relatively indigestible oligosaccharides of tree gums or, if they gouge, hardened saps, that require fermentation for digestion. Their gut microbiomes may be exclusively dedicated to carbohydrate metabolism. Thus, hybridization and exudivory represent intriguing venues to round out our understanding of animal gut microbiomes.
Hybridization is well documented among NHPs and other animals [12–19]. A few studies of captive, natural, and experimental animals show that hybrid gut microbiomes differ from those of parental species [e.g., 10, 11, 20]. Human activities increasingly facilitate anthropogenic hybridization, which may lead to the rise of animal host populations with compromised viability due to perturbed gut microbiomes. Few data are available on anthropogenic hybrid gut microbiomes, and such data will increase comparative datasets that facilitate our understanding of how hybridization may impact animal health.
In terms of animal dietary strategies, gut microbiomes have been mostly studied in folivores, frugivores, and omnivores. For NHPs exploiting such diets, gut microbiome composition is more sensitive to environmental effects than that of other animals [6, 9, 21]. Further, captive diets have lower fiber and plant content than what NHPs eat in nature [6,9], which may perturb the gut microbiome of captive populations and negatively impact host health. Indeed, regardless of natural diet, captive NHP gut microbiome composition tends to resemble that of humans consuming a Western diet [6]. This diet may cause dysbiosis and be directly associated with chronic diseases that drive morbidity and mortality in industrialized countries [22]. For threatened species, captivity serves a vital conservation role, and a particular concern is whether a “westernized” gut microbiome profile would compromise host digestive or immune function for captive hosts reintroduced into the wild [9].
A number of NHPs are exudivores [23–26], and some exudivore primate species are endangered or vulnerable (e.g., Allocebus trichotis, Microcebus berthae, M.ravelobensis, Nycticebus bengalensis, C. aurita and C. flaviceps) [27] and some are known to hybridize [e.g., Callithrix: 19; Microbecus: 28]. Studies of the exudivore gut microbiome from the wild are rare [e.g., Nycticebus pygmaeus: 29] and even rarer in captivity [e.g., an individual captive Callithrix jacchus: 30]. Further investigation of the gut microbiome of wild exudivore NHPs will provide baseline data, which could improve conservation and management of natural host populations. Additionally, similar data from captive exudivores could facilitate understanding of exudivore welfare in captivity and during reintroduction initiatives.
Our study focused on marmosets (Callithrix), a genus of six species considered to be specialist exudivores due to several morphological adaptations [23, 25, 31]. Two species are critically endangered, C. flaviceps and C. aurita, though conservation plans underway for both species include reintroductions of captive individuals into the wild [32]. Callithrix species are naturally allopatric, but hybridize at species contact points [19]. Nonetheless, anthropogenic hybridization occurs widely among Callithrix species due to illegal pet trade and exotic species introduction outside natural species ranges [19].
Here we applied 16S ribosomal (rRNA, V4 region) gene amplicon sequencing of Callithrix anal microbiotas, and investigated microbiome composition and predictive functional profiles of the Callithrix gut microbiome. Anal swabs were sampled from a total of 59 individuals across four species (C. aurita, C. geoffroyi, C. penicillata, and C. jacchus) and three anthropogenic hybrid types (C. aurita x Callithrix spp., C. jacchus x C. penicillata, C. penicillata x C. geoffroyi). These individuals were sampled in the wild as well as from captive facilities where animals were born in captivity or translocated from the wild. This multifaceted dataset gives us a unique opportunity to study how host taxon, hybridization, and environment shape the gut microbiota in specialist exudivore primates. We hypothesized that: (1) Callithrix hybrids have less diverse gut microbiomes than their parental species; (2) Callithrix hosts found in similar environments possess more similar gut microbiome diversity and predictive function; (3) captive Callithrix have less diverse gut microbiomes than wild Callithrix; and (4) as specialist exudivores, Callithrix hosts show gut microbial composition and predictive function profiles that reflect a relatively strong bias toward carbohydrate metabolism.
Materials and Methods
Sample Collection
We collected anal swabs between 2015 and 2016 from 59 adult individuals (Table 1 and Supplementary Table 1). Figure 1 and Supplementary Table 1 show sampling locations for: (1) wild marmosets in Viçosa, Minas Gerais (MG), Berilo, MG, and Guiricema, MG; (2) captive-born marmosets housed at the Guarulhos Zoo, Guarulhos, São Paulo (SP), CEMAFAUNA Wild Animal Triage Center, Petrolina, Pernambuco (PE), and Rio de Janeiro Primatology Center (CPRJ), Guapimirim, Rio de Janeiro (RJ); (3) a wild group from Natividade, RJ that was caught and housed at CPRJ; (4) and a wild group from Ilha d’Agua Island, Rio de Janeiro, RJ that was caught and kept at SERCAS (Setor de Etologia, Reintrodução e Conservação de Animais Silvestres) at Universidade Estadual do Norte Fluminense (UENF), Campos dos Goytacazes, RJ.
Marmoset sampling was authorized by the Brazilian Environmental Ministry (SISBIO protocol #47964-2), the Arizona State University IACUC (protocol# 15-144R) and a CPRJ internal review. Wild animals were captured with auto-close Tomahawk style traps baited with banana. All sampled animals were immobilized by injection of ketamine (10 mg/kg of body weight) into the inner thigh intramuscular region. Animals were photographed, weighed, measured, and kept under veterinary observation. Copan FLOQSwab tips were inserted and rotated in the anal region and then each swab was submerged in a storage buffer (50 mM Tris pH 8.0, 50 mM EDTA, 50 mM Sucrose, 100 mM NaCl, 1% SDS) for several seconds and then discarded. Under field conditions, the storage buffer was kept at room temperature and then at 4C in the laboratory. After sample collection, animals were returned to cages, given a banana, and allowed to recover. All recovered wild marmosets were released at the original capture site.
Specimens were identified as pure C. aurita, C. geoffroyi, C. jacchus and C. geoffroyi or hybrid (C. aurita x Callithrix spp., C. jacchus x C. penicillata, and C. penicillata x C. geoffroyi) based on phenotypic characters following descriptions [34–37], and whole genome genotyping (unpublished data). Specimens were further classified by their environment as wild (captured as free-range individual), translocated (born in the wild but later put into captivity), or captive (born and raised in captivity based on information from captive facilities).
Sample Processing and Sequencing
Bacterial DNA extraction from Callithrix anal swabs was carried out by adding 0.3g zirconia beads to tubes containing storage buffer and swabbed material following a modified phenol-chloroform protocol [38]. Modifications included beating the samples on a vortex fitted with a horizontal vortex adaptor (#13000-V1-24, Mo Bio, Carlsbad, CA, USA) for 10 minutes at step “2Aiii,” precipitating samples in 100% ethanol in step “2Axvi,” and rehydrating DNA pellets in 25 uL low TE buffer at step “2Axxii.” Extracted DNA was quantified on a Qubit3 (Life Technologies, Carlsbad, CA, USA) with a dsDNA HS Assay Kit (Life Technologies). Then the V4 region of the bacterial 16S rRNA gene was amplified in triplicate using the barcoded primer set 515f/806r [39]. Genetic samples have been registered in the Brazilian CGen SISGEN database (Supplementary Table 2). Amplicon triplicates were combined for each individual and then pooled in equimolar amounts into a multiplexed Illumina sequencing library. The library was purified with a Zymo DNA Concentrator and Cleaner-5 (#D4013, Zymo Research, Irving, CA, USA) and size selected for 375-380 base pairs with Agencourt Ampure XP (#A63880, Beckman Coulter, Indianapolis, IN, USA) magnetic beads. Following international regulations, DNA extraction and library preparation was carried out at the Federal University of Viçosa, Brazil and then sequenced at Arizona State University, USA on an Illumina MiSeq for 2×250 cycles.
Bioinformatics and Statistical Analysis
Sequence data were demultiplexed using default parameters in Qiime2-2018.6 (https://qiime2.org) [40]. The DADA2 Qiime2 plug-in [41] was used to quality-filter and trim sequences and join paired-end reads. Upon trimming, the first 10 and last 30 nucleotides were removed from reverse reads due to low base quality. These steps resulted in feature tables of DNA sequences and their per-sample counts, termed sub-operational taxonomic units (sOTUs) [42]. MAAFT and FastTree Qiime2 plug-ins [43,44] aligned and produced a phylogenetic tree of feature sequences, which was then mid-pointed rooted. Qiime2’s q2-diverity plug-in was used to carry out alpha diversity analyses (Shannon’s diversity index, observed OTUs, and Pielou’s Evenness) and calculate beta diversity weighted UniFrac distance [45]. Based on rarefaction, we set the minimum sampling depth to 45000 reads for alpha and beta diversity measures. Box-plots were generated for each alpha diversity measure according to host taxon, hybrid status, and environment. Kruskal-Wallis group and pairwise tests determined statistical significance for each category. UniFrac principle coordinate analysis (PCoA) plots were also generated according to host taxon, hybrid status, and environment. Statistical significance of UniFrac PCoA plots within each category was determined by PERMANOVA group and pairwise tests [46]. All pairwise comparisons were corrected with the Benjamini–Hochberg procedure [47]. The corrected P-value was considered significant at p<0.05.
Taxonomic composition of samples was determined with the QIIME2 q2-feature-classifier plug-in, which was trained on the Greengenes OTU database [version 13_8, 48] at the 99% level. Reads were limited to 500 nucleotides from the portion of the 16S V4 region bound by the 515F/806R primer pair. Afterward, QIIME2 was used to generate bar plots of bacterial taxonomic abundance based on host taxon, hybrid status, and environment.
The QIIME2 gneiss plug-in [49] was used to carry out differential abundance analysis of bacterial taxa within and among hosts based on host taxon classification, hybrid status, and environment. Briefly, partitions were made for microbes that commonly co-occur together via Ward hierarchical clustering, which then underwent isometric log ratio transformation to control for false positives by testing “balances” (i.e., changes in log ratios between microbial abundances) [42]. Linear regression tested for balances between microorganisms using host taxon, hybrid status, and environment as co-variants by the ordinary least-squares (OLS) method.
PICRUST 1.1.3 [50] was used for predictive functional profiling of bacterial communities via a phylogeny-based approach to correlate sampled bacterial species to the Greengenes 13_5 99% OTU reference dataset [50]. Initially, the Greengenes reference data was used in QIIME2 to make a closed-reference OTU table from our sequence reads and converted to biom format. The OTU table was normalized by 16s rRNA copy number in PICRUST and the extrapolated metagenome results then were input into HUMAnN 0.99 [51]. This program reconstructs microbial pathway abundances from metagenomic data using the KEGG Orthology catalogue [52, 53, 54]. The HUMAnN KEGG pathway modules output was used in LefSe [55] on the Huttenhower lab Galaxy server (http://huttenhower.sph.harvard.edu/galaxy/) to identify how abundance of modules differ between marmosets categorized by taxon, hybrid status, and environment with linear discriminant analysis (LDA) coupled with effect size. These differences were then plotted in GraPhlAn 0.9 [56].
Results
A total of 11,123,817 sequence reads was obtained with an average of 188,539 (129,170 +/− SD) reads per sample. After quality filtering, 9,063,712 reads remained with an average 153,622 (103,648 +/− SD) reads per sample. Afterward, merging of paired-end sequences produced 843,615 reads, with an average of 142,891 (96,076 +/− SD) reads per sample. This information is detailed in Supplementary Table 3.
Gut Microbiome Alpha Diversity based on Host Taxon, Hybridization, and Environment
Rarefaction excluded a total of seven individuals from alpha diversity analysis (Supplementary Table 1). Although taxa differed significantly in Shannon’s diversity index (Figure 2A; Kruskal-Wallis group test, H=16.88, p=0.01), Kruskal-Wallis pairwise tests were not significant (Supplementary Table 4). The number of observed OTUs among taxa ranged from 61 to 571 (Supplementary Table 1), and median values varied significantly among taxa (Figure 2B; Kruskal-Wallis group test, H=28.16, p=0.00). For pairwise comparisons, the median observed OTUs in C. jacchus x C. penicillata hybrids were significantly different from that of C. aurita and C. penicillata x C. geoffroyi hybrids (Supplementary Table 5). Although taxa differed significantly in species evenness (Figure 2C; Kruskal-Wallis group test, H=16.41, p=0.01), subsequent pairwise tests were not significant (Supplementary Table 6).
Shannon diversity index values of hybrid marmosets were significantly higher than those of non-hybrid marmosets (Figure 3A; Kruskal-Wallis pairwise test, H=10.22, p =0.001). Pairwise differences between hybrids and non-hybrids were significant for both observed OTU number (Figure 3B; Kruskal-Wallis pairwise test, H=7.24, p=0.007) and evenness (Figure 3C; Kruskal-Wallis pairwise test, H=8.10, p =0.004) (Figures 3B and 3C). Shannon’s index differed significantly among wild, translocated, and captive marmosets (Figure 4A; Kruskal-Wallis group test, H=9.85, p=0.007), and pairwise differences were significant between captive vs. translocated and captive vs. wild groups (Supplementary Table 7). Differences in median observed OTUs were significant among host environment categories (Figure 4B; Kruskal-Wallis group test, H=15.95, p=0.00), specifically between translocated vs. captive and translocated vs. wild (Supplementary Table 8). Species evenness differences were significant among host environments (Figure 4C; Kruskal-Wallis one-way test, H=10.03, p <0.007), with captive vs. translocated and captive vs. wild comparisons showing significant pairwise differences (Supplementary Table 9).
Gut Microbiome Beta Diversity based on Host Taxon, Hybridization, and Environment
PCoA plots of weighted Unifrac distances suggested that hybridization and environment more strongly affect host differences in gut microbial composition than taxon affiliation (Figures 5A-C). The first PCoA axis explained 33.02% of gut microbial variation and the second axis explained 19.49%. In separating the samples by taxon (Figures 5A), some taxa showed loose clustering, but most of them showed interdispersion in the upper and lower right quadrants. Distances among taxa were significant (PERMANOVA, pseudo-F = 2.67, p-value= 0.002), but there were few significant pairwise comparisons among them (Supplementary Table 10). In Figure 5B, hybrid status showed most non-hybrids clustering in the lower right PCoA quadrant whereas hybrids were more prominent in the lower and upper left quadrants. These differences for hybrid status were significant (PERMANOVA, pseudo-F = 4.43, p-value=0.003). Finally, separation of hosts according to environment resulted in the most prominent clustering patterns between wild, translocated, and captive samples, particularly along the first axis (Figure 5C). Weighted Unifrac distances between samples by environment were significant at both the group level (PERMANOVA, pseudo-F = 8.88, p-value=0.001), as well as for all pairwise comparisons (Supplementary Table 11).
Differential OTU abundance based on Host Taxon, Hybridization, and Environment
The most abundant bacterial classes among hosts (Figure 6, Supplementary Figure S1) were Gammaproteobacteria, Epsilonproteobacteria, Bacteroidia, Actinobacteria, Clostridia, and Alphaproteobacteria; although, many other bacterial classes were recognized. Hosts showed variation in both bacterial class and abundance among individuals (Supplementary Figure S1A), particularly in C. aurita, C. geoffroyi, and C. penicillata- taxa that were sampled across different locations and environments. In considering host hybrid status, Gammaproteobacteria was particularly abundant in pure hosts and Epsilonproteobacteria and Bacteroidia are more prominent in hybrid hosts (Supplementary Figure S1B). Finally, a strong pattern of bacterial taxonomic abundance (limited to class for brevity) emerges upon classifying hosts by environment (Figure 6). At the family level these patterns are dominated by: (1) Enterobacteriaceae (class Gammaproteobacteria) in captive hosts; (2) families Helicobacteraceae and Campylobacteraceae (class Epsilonproteobacteria), Bacteroidaceae (class Bacterioidia), and Sphingomonadaceae (class Alphaproteobacteria) in translocated hosts; and (3) Helicobacteraceae and Campylobacteraceae, Bacterioidia (family undetermined), Bifidobacteriaceae (class Actinobacteria), and Veillonellaceae (class Clostridia) in wild hosts.
Using a multivariate response linear regression on the “balance” of the microbial community of the Callithrix gut model, we evaluated the effect of different types of taxa, hybrid status, and environmental categories on the microbiome. Together, these categories explained 25.15% of the variation of the Callithrix gut microbiota, with environment having the strongest effect by being a covariate that explained 6% of the variation (Supplementary Table 12). A good fit of this “balance” model to our data is indicated by predicted points lying within the same region as the original microbial community and the residuals have roughly the same variance as the predicted points (Supplementary Figure S2).
Predictive Metagenomic and Metabolic Characterization of Marmoset Gut Microbiome
Metagenome prediction by PICRUSt using 16S rRNA gene amplicon data analyzed in QIIME2 resulted in predicted proteins classified as KEGG orthologos (KOs) with a total of 6909 KOs across all hosts. Using these KOs with HUMAnN for metabolic reconstruction detected 113 KEGG modules collectively present in all samples (Supplementary Table 13). Among these modules, 20 were present across all samples at a coverage of >90% and were identified as core modules (Table 2). The core modules were found to be involved with the following Level 3 KEGG BRITE hierarchies: central carbohydrate metabolism (6 modules), purine metabolism (2 modules), pyrimidine metabolism (1 module), cofactor and vitamin biosynthesis (5 modules), ATP synthesis (1 module), photosynthesis (1 module), phosphate and amino acid transport system (2 modules), mineral and organic ion transport system (1 module), and aminoacyl tRNA (1 module). A “secondary” set of 37 modules were present at <90% coverage in no less than one but no more than three sampled marmoset hosts (Supplementary Table 14). Examples of secondary modules present across sampled individuals at Level 3 KEGG BRITE hierarchies include: aromatic amino acid metabolism (2 modules), central carbohydrate metabolism (4 modules), cofactor and vitamin biosynthesis (3 modules), cysteine and methionine metabolism (2 modules), histidine metabolism (2 modules), and mineral and organic ion transport system (2 modules).
Several KEGG modules were differently abundant for host taxonomy, hybrid status, and environment (Figures 7-8, Supplementary Figures S3-S6). A total of 57 modules differed for taxa (Supplementary Figure S3 and S4), with C. jacchus x C. penicillata hybrids having the largest number of differently abundant modules and C. jacchus possessing the second highest number. Hybrids and non-hybrids showed a total of 63 differentially abundance modules (Supplementary Figures S5 and S6), with the former being overwhelmingly different in abundance of modules for metabolism of nucleotides, carbohydrates, and amino acids. Differences among the three groups of host environment were not particularly concentrated around certain categories but rather spread out across clades of modules.
Discussion
Factors that Potentially Influence Callithrix Gut Microbiome Diversity
Our data showed that host hybrid status and environment more strongly shaped Callithrix gut microbiome diversity than host taxon classification. While host taxa differed significantly in microbial diversity measures, few pairwise taxa comparisons were significant. The gut microbiome of hybrid Callithrix was significantly more diverse that of non-hybrids. Captive marmosets had significantly less diverse gut microbiomes than their wild and translocated counterparts. As observed in other NHPs, captive marmosets showed the most reduced gut microbiome diversity in terms of species richness and evenness. However, the strong “humanization” effect of captivity reported for the gut microbiome of other NHPs was not entirely evident in our study.
Regarding hybridization effects on gut bacteria, changes in the hybrid gut microbiome relative to parental species may cause negative epistasis between microbial and chromosomal genes [10]. For example, Nasonia wasps and Mus mouse hybrids possessed altered microbiomes relative to parental species and showed aberrant expression of immune genes [10,11]. Further, Nasonia hybridization was lethal and Mus hybridization resulted in higher histopathology of caecal tissue in hybrids relative to parental species [10,11]. Aberrant immune gene expression in animal hybrids may make them more susceptible to infectious disease than parental species, and alterations in the gut microbiome due to hybridization may negatively affect hybrid fitness.
Anthropogenic Callithrix hybrids are fertile, tend to form hybrids swarms and are frequently present in large Brazilian urban areas and rural landscapes [19, 57; Malukiewicz, pers. obs.], and they possessed more diverse gut microbiomes than non-hybrids. In some cases, hybridization may also lead to favorable physiological functions that can drive adaptation, genetic novelty, or speciation [58]. Our data suggest show that hybridization does not perturb microbial diversity in Callithrix, and increased diversity of the gut microbiome may actually be favorable for hybrid host physiology, fitness, and health. Thus, the connection between the gut microbiome and host functions should be further studied in Callithrix hybrids to determine whether hybridization promotes host health and fitness in specialized exudivores.
Considering host environment, the lack of a strong “humanization” effect of captivity in Callithrix is likely related to differences in their dietary strategy as specialist exudivores to that of other NHPs. Callithrix possesses morphological adaptations for gouging gum from trees [59–61] and can spend more than 50% of their feeding time on gums in the wild (e.g., C. aurita [62]; C. jacchus [63]). However, captive Callithrix husbandry often omits gums from the marmoset diet or uses it as an enrichment ingredient [26]. Regular provisioning of exudates also did not apply to captive and translocated marmosets we sampled across four captive facilities, whose generalized Callithrix diets are shown in Table 3. One concern for a lack of gum in the diet of captive specialist exudivores is the development of health issues as well as a negative impact on breeding and survivability [26, 64]. Thus, our results suggest that provisioning of gum in the diet of captive specialized exudivores may improve host welfare and conservation reintroduction programs by maintaining gut microbiomes closer to that of wild populations.
Bacterial Composition and Abundance of the Callithrix Gut Microbiome Relative to Other NHP and Mammals
We focus here on differences in composition and abundance of bacterial taxa in wild, translocated, and captive Callithrix (Table 4), as host environment showed a stronger effect on these aspects of the gut microbiome than host taxon classification and hybrid status. In a single captive C. jacchus fecal sample, Albert et al. [30] found Proteobacteria (37.1%), Firmicutes (33.0%), Bacteroidetes (28.1%), and Actinobacteria (<0.04%) as the major gut microbiome phyla. Proteobacteria was also the most abundant taxon among our captive and translocated hosts, although with a higher average abundance per individual than that of [30]. For another NHP specialist exudivore, the pygmy loris (Nycticebus pygmaeus), two wild individuals possessed Bacteroidetes (41.19%), Proteobacteria (30.43%), and Actinobacteria (10.98%) as major gut microbiome bacterial phyla [29]. Overall, these results suggest that Actinobacteria may be a gut microbe typically present in wild specialist exudivores, while Proteobacteria and Bacteroidetes are common in specialist exudivores regardless of environment.
The marmoset gut microbiome seems to differ in bacterial taxa abundance from that of non-gummivores. In an early study of mostly captive NHPs and other mammals, bacterial gut microbiome composition consisted of Firmicutes (65.7%), Bacteroidetes (16.3%), Proteobacteria (8.8%), Actinobacteria (4.7%) and some less abundant taxa [65]. In another captive NHP sample, Provotella and Bacteroides were the most abundant gut microbes, in line with the modern human gut microbiome [6]. In a recent large sampling of mammals, including several NHPs, the gut microbiome of captive hosts was largely characterized by Bacteroidetes, class Bacilli (Firmicutes), Gammaproteobacteria (Proteobacteria), and class Verruco-5 (Verrucomicriobia) [9]. On the other hand, Actinobacteria, Prevotella (Bacterioidetes), and class Clostridia (Firmicutes) were most abundant for wild hosts [9].
The relatively higher abundances of gut Actinobacteria in our sampled wild marmosets are notable. Within this phylum, Bifidobacterium was the most common genus found across all host environments, but particularly among wild hosts. This bacterial genus is common in animal gastrointestinal tracts and metabolizes host-indigestible carbohydrates by aiding their importation into cells [30, 66]. The genomes of three isolates of Bifidobacterium callitrichos from a captive C. jacchus fecal sample contained predicted genes associated with galactose and arabinose metabolism, which are major constituents of tree gums eaten by C. jacchus [30]. The relatively higher presence of Bifidobacterium in wild marmosets within our dataset is likely associated with a higher intake of tree gums by wild marmosets than in captivity-underlining the importance of Bifidobacterium in marmoset carbohydrate metabolism and gummivory adaptations.
As several strains of Proteobacteria are pathogenic, the high occurrence of these bacteria across our dataset is of note, particularly differences in Proteobacteria taxa present between captive, translocated, and wild marmosets. Wild marmosets had an average abundance of Helicobacter of 21% and <1% of Enterobacteriaceae. For translocated marmosets, average abundance was 24% Helicobacter and 16% Enterobacteriaceae while captive marmosets had 12% Helicobacter and 41% Enterobacteriaceae. Previous studies have identified potentially pathogenic Helicobacter strains in captive marmosets [67–69]. In humans, Helicobacter is thought to cause hepatic or gastrointestinal disease [70]. However, marmosets infected with Helicobacter do not necessarily show clinical symptoms [68], or may show clinical symptoms of gastric disease but not be infected with Helicobacter [67, 69]. Enterobacteriaceae is frequently associated with intestinal diseases and contains a number of pathogenic bacterial strains of Salmonella, Escherichia, and Shigella [71, 72]. Although Enterobacteriaceae species were not determined for our sample, potentially pathogenic Enterobacteriaceae were previously found in captive and urban Callithrix [73]. Thus, future Callithrix microbiome research should identify which Proteobacteria species are present in the Callithrix gut and how close Callithrix-human contact affects the transmission host dynamics of these bacteria.
Predictive Functional Aspects of the Callithrix Gut Microbiome
The principle core predictive functional KEGG modules expressed in all sampled marmosets were related to central carbohydrate metabolism, purine metabolism, pyrimidine metabolism, cofactor and vitamin biosynthesis, among other modules. Given the differences in the principal bacterial taxa present in the gut of captive, translocated, and wild marmosets, similar functions may be carried out by various bacteria despite differences in host environment. Core predictive modules for carbohydrate metabolism were involved with glycolysis and the pentose phosphate pathway, pathways essential for the production of ATP and NADPH. For predictive cofactor and vitamin biosynthesis modules, biosynthesized products included pantothenate, riboflavin, biotin as well as NAD.
The gut microbiome of wild marmosets was particularly enriched for predictive modules involved in carbohydrate metabolism including that of glycolysis and pentose phosphate biosynthesis. While such modules are part of the core predictive functional modules discussed above, they are over-represented in wild marmosets likely due to the relatively high abundance of Bifidobacterium in their gut microbiome. Carbohydrate metabolism was also the most abundant functional category present in the two wild pygmy lories discussed above [29]. On the other hand, predictive functional KEGG modules enriched in the gut microbiome of captive and translocated marmosets were those involved in nucleotide and amino acid metabolism and environmental information processing. Enrichment of these particular predictive modules in the gut microbiome of captive and translocated marmosets may be related to their diets. Since gums were not a normal part of the marmoset diet in any of the captive facilities we sampled, the marmoset gut microbiome seems to have partially redirected functional and metabolic activities to those that differ from the wild.
Conclusion
The results of this study have given us an initial genus-level look into the gut microbiome of specialized exudivorous NHPs. Our major findings indicate that, just as in other primates, the gut microbiome of Callithrix is sensitive to environmental factors, and these findings likely apply more broadly to other NHP and non-NHP animal exudivores. Marmosets are maintained in captivity for biomedical research [74], in zoological collections, and as part of a recent conservation program in Brazil for C. aurita [32]. Several challenges arise in maintaining captive Callithrix such as marmoset wasting syndrome [26], a rise in obesity in captive populations [75], and lack of a standardized marmoset diet. Additionally, it is concerning that our results indicated a relatively large abundance of Enterobacteriaceae in the gut microbiome of captive, but not wild, marmosets. In the case of biomedical studies, data from unhealthy marmosets may bias research findings, and captive-bred marmosets reintroduced into the wild may not be fit to survive with gut microbiome profiles different from their wild counterparts.
Our findings also suggest that carbohydrate metabolism is a core function of the Callithrix gut microbiome. Yet, although marmosets are specialized exudivores, there is a wide gap between the nutritional intake of marmosets in the wild and captivity, where the marmoset physiology for gummivory is generally not considered [26]. From an evolutionary and health point of view, this disconnect between wild and captive Callithrix diets leaves open several questions concerning maladaptation of the marmoset digestive system, perturbation of gut microbiota, malabsorption of nutrients, and long term health effects. As studies of the exudivore gut microbiome are generally rare, such issues are also applicable to other captive NHPs that have evolved to exploit this nutritional strategy. More studies are needed to understand better the health and reproductive consequences of omitting a major component of the natural diet of specialized exudivores in captivity.
To corroborate the findings of this study, future research should prospectively investigate the specialized exudivore gut microbiome with a larger host sampling across different taxa and environments, as well as through metagenomics and transcriptomics. A systematic study of the effects of different diets on the marmoset gut microbiome, including those with and without gum consumption, would facilitate understanding of how nutrition affects marmoset health and particularly their gastrointestinal health. For Callithrix, such data will add to our understanding of the evolution of marmosets as exudivores, as well as help guide conservation efforts for C. aurita and C. flaviceps. Overall, such information will expand baseline gut microbiome data available for wild and translocated/captive specialized exudivores to allow for the development of new tools to improve the management, welfare, and conservation of these animals.
Conflict of Interests
The authors declare no conflict of interests.
Supplementary Figure S1. (A). Relative abundances of gut bacterial classes by taxon (A). Taxon abbreviations are the same as in Figure 1. (B) Relative abundances of gut bacterial classes by hybrid status. Hybrid status abbreviations are the same as in Figure 2. In both A and B, colors within each bar represent bacterial classes with top colors representing greater bacterial relative abundance than those taxa at the bottom of each bar.
Supplementary Figure S2. Prediction plots of the top two balances for the OLE regression model, with the prediction residuals from the model projected onto these two balances.
Supplementary Figure S3. HUMAnN metabolic reconstruction of marmoset fecal microbiome based on host taxonomy. The cladogram shows 3 KEGG BRITE hierarchical levels with an outer, generalized level, and ending with the inner-level annotated by letters that are detailed further in the accompanying legend. Circles not differentially abundant in any species are clear.
Supplementary Figure S4. Differentially abundant KEGG metabolic modules inferred by LEfSe are colored by their associated taxon classification, which follow the same abbreviations as Figure 1. The plot shows LDA scores resulting from LeFSe analysis for differentially abundant KEGG modules across host taxon classifications.
Supplementary Figure S5. HUMAnN metabolic reconstruction of marmoset fecal microbiome based on host hybrid status. The cladogram shows 3 KEGG BRITE hierarchical levels with an outer, generalized level, and ending with the inner-level annotated by letters that are detailed further in the accompanying legend. Circles not differentially abundant in any species are clear.
Supplementary Figure S6. Differentially abundant KEGG metabolic modules inferred by LEfSe are colored by their associated hybrid classification, which follow the same abbreviations as Figure 3. The plot shows LDA scores resulting from LeFSe analysis for differentially abundant KEGG modules across host hybrid status.
Acknowledgements
This work was supported by a Brazilian CNPq Jovens Talentos Postdoctoral Fellowship, an American Society of Primatologists Conversation Small Grant, and an International Primatological Society Research Grant. We thank Vanner Boere, Ita de Oliveira e Silva, Veronica Souza, Rodrigo S. Carvalho, the Guarulhos Zoo staff, the CEMAFAUNA staff, the CPRJ staff, SERCAS staff, and AMLD staff for assistance with wild and captive populations. We are very grateful to Cecil M. Lewis and Tanvi Honap and the LMAMR staff at University of Oklahoma for donation of PCR primers. We thank Camila Molina for comments on prior versions of this work.