The Gastrointestinal Microbiota of the Common Marmoset (Callithrix jacchus)

Abstract

The microbiota is heavily involved in both health and disease pathogenesis, but defining a normal, healthy microbiota in the common marmoset has been challenging. The aim of this review was to systematically review recent literature involving the gastrointestinal microbiome of common marmosets in health and disease. Twelve sources were included in this review. The gut microbiome composition was reviewed across institutions worldwide, and taxonomic shifts between healthy individuals were described. Unlike the human gut microbiome, which is dominated by Firmicutes and Bacteroidetes, the marmoset gut microbiome shows great plasticity across institutions, with 5 different phyla described as dominant in different healthy cohorts. Genera shared across institutions include Anaerobiospirillum, Bacteroides, Bifidobacterium, Collinsella, Fusobacterium, Megamonas, Megasphaera, Phascolarctobacterium, and Prevotella. Shifts in the abundance of Prevotella or Bifidobacterium or invasion by pathogens like Clostridium perfringens may be associated with disease. Changes in microbial composition have been described in healthy and diseased marmosets, but factors influencing the severe changes in microbial composition have not been established. Multi-institutional, prospective, and longitudinal studies that utilize multiple testing methodologies are required to determine sources of variability in the reporting of marmoset microbiomes. Furthermore, methods of microbial manipulation, whether by diet, enrichment, fecal microbiome transplantation, etc, need to be established to modulate and maintain robust and resilient microbiome communities in marmoset colonies and reduce the incidence of idiopathic gastrointestinal disease.

INTRODUCTION

Callithrix jacchus is a diurnal, arboreal, New World, non-human primate (NHP) native to Brazil. Adults can weigh 300–500 g and have average lifespans ranging from 5 to 12 years of age in research colonies.^1–4 Marmosets live in family troops of up to 15 with 1 dominant breeding pair.⁵ In the wild, their diet is primarily composed of plant exudates and insects, but they will also consume fruits, seeds, flowers, and small animals.^6,7 Due to their size, high fecundity, and similarity to humans, they have become models in biomedical research with applications in aging, vision, behavioral neuroscience, multiple sclerosis, auditory research, neurodegenerative diseases, and toxicology.⁸

Gastrointestinal (GI) diseases are the most common and widespread clinical finding in captive common marmosets.^9–11 Inflammatory Bowel Disease (IBD) prevalence is reported to be as high as 28%–60% in captive adult marmosets and presents with diarrhea, weight loss, enteritis, muscle atrophy, alopecia, and a failure to thrive. IBD is also associated with hypoproteinemia, anemia, and elevated liver enzymes.^9,12 A novel chronic GI disease has been recently reported in young adult marmosets and is characterized by a duodenal stricture or dilation, which accounted for 21% of deaths in a colony.^13,14 As reports have highlighted the effect of captivity, and the associated effects of stress and diet, on the microbiome of NHPs,^15–19 there is an increased interest in understanding the role of the microbiome in both healthy and diseased marmosets.

While the concept of using the 16S ribosomal RNA gene as a phylogenetic marker was pioneered in 1977,²⁰ the advent of next-generation sequencing, along with the reduction of DNA sequencing costs and improved bioinformatics, led to a boom in sequencing-based microbiome studies in the early 2000s.^21–24 The importance of microbiome studies was validated by the creation of the NIH-funded Human Microbiome Project.²⁵ However, studies of the marmoset microbiome are still in their infancy. Comparative studies included other members of the family Callitrichidae, such as Goeldi’s marmoset (Callimico goeldii) as early as 2008,^26,27 but studies focused on the microbiome of the common marmoset (Callithrix jacchus) did not appear until almost a decade later (Table 1).²⁸ This review aims to summarize the current knowledge of the microbiome in marmosets and its association with specific GI diseases. Factors that may affect the microbiome are discussed along with future directions that require further elucidation to better understand the marmoset microbiome and its relationship with itshost.

Table 1.

Summary of 12 Publications Used in the Systematic Review

Author	Research Site	Study Design	Animals (M/F)	Sample Source	Total Samples	Microbiome Analysis	Bioinformatic Pipeline	Key Findings	Reference
Ross et al, 2017	SNPRC & BI	Compared marmosets from SPF barrier at BI (38) with “conventional” animals at SNPRC (16)	54 (NA)	Rectal swabs	54	16S rRNA (V1-V3)	RDP Bayesian classifier and R (vegan, hclust)	Profiles divided into Cluster 1 (pred. SNPRC with high Fusobacterium B/low Bifidobacterium), Cluster 2 (both BI and SNPRC with > 50% Bifidobacterium) and Cluster 3 (pred. BI with high Bifidobacterium/low Fusobacterium B).	28
Kap et al, 2018	BRPC	Evaluated effect of yogurt-based dietary supplement on microbiome and disease incidence in EAE model. Samples collected before diet change, after diet but pre-EAE induction, and post-EAE induction.	16 (8 M, 8 F)	Feces	~59	16S rRNA (V3-V4)	PANDAseq, QIIME and ARB	Diet change alone did not alter baseline microbiome dominated by Actinobacteria (66%), Firmicutes (20%), and Bacteroidetes (11%). EAE induction led to divergent response to perturbation with loss of Bifidobacteriaceae observed in control animals	32
Albert et al, 2018	UMASS	Fecal sample from a single female marmoset was analyzed to isolate Bifidobacteria	1 (1 F)	Feces	1	16S rRNA (V3-V4)	QIIME-1	Microbiome dominated by Proteobacteria, Firmicutes, and Bacteroidetes in a single sample	41
Shigeno et al, 2018	RIKEN	Evaluated laboratory-bred marmosets that were healthy (39) or had chronic diarrhea (19)	58 (27 M, 31 F)	Feces	58	T-RFLP (AluI/MspI)	MiCA3 Virtual Digest analysis	Low resolution but used qPCR to confirm higher abundance of Bifidobacterium in cluster with predominance of healthy animals	33
Reveles et al, 2019	BI	Compared the microbiome of 10 young adult and 10 geriatric marmosets	20 (20 M)	Feces	20	16S rRNA (V4)	USEARCH, Mothur, R	Predominant phyla observed were Bacteroidetes, Actinobacteria, Firmicutes, Proteobacteria, and Fusobacteria. Shannon diversity and Firmicutes decreased, while Succinivibrionaceae increased in geriatric animals.	34
Artim et al, 2019	MIT	Compared the microbiome of paired fecal and rectal swab samples from 20 healthy and 3 marmosets with GI disease to determine feasibility of using rectal swabs for microbiome analysis	23 (12 M, 11 F)	Rectal swabs and Feces	48	16S rRNA (V4-V5)	QIIME-1.9	Microbiome dominated by Bacteroidetes (52%, mainly families Bacteroideceae and Prevotellaceae), Firmicutes (16.6%), and Proteobacteria (14.9%). Rectal swabs represented paired fecal sample unless rectal swab lacked visible fecal matter.	35
Takehara et al, 2019	CLEA	Compared the microbiome in saliva of 8 marmosets with the saliva from 8 humans	8 (4 M, 4 F)	Oral (saliva)	8	16S rRNA (V3-V4)	CD-HIT-OTU, RDP classifier program, QIIME-1.8	Oral microbiome of humans and marmosets were distinct. In marmosets, Proteobacteria (26%), Bacteroides (24%), Fusobacteria (21%), and Firmicutes (19%) most abundant. Compared with humans, Actinobacteria and Firmicutes less abundant in marmosets while Fusobacteria and Spirochaetes more abundant	47
Kobayashi et al, 2020	RIKEN	Comparison of fecal microbiome of 7 monogastric species included marmosets	9 (9 M)	Feces	9	16S rRNA (V3-V4)	QIIME-1.9 and phyloseq	Bacteroidetes (30%), Firmicutes (34%), and Actinobacteria (24.0%) were predominant phyla. Megasphaera and Bifidobacterium proposed as predominant butyrate and lactate producers, respectively.	39
Cooper et al, 2020	JHU	Comparison of JHU colony (18) with imported marmosets from DPG (18) and HHG (15). Imported animals transitioned to JHU diet by day 50 and integrated into colony starting at day 130. Microbiome sampled at days 0, 100, and 390	51 (24 M, 27 F)	Rectal swabs	123	16S rRNA (V4-V5)	QIIME-1.9 & 2, LEfSe	Bacterial communities different at day 0. DPG and HHG converged on day 100 due to diet. Remained distinct from JHU. Diet > housing	38
Malukiewicz et al, 2020	Various	Compared microbiome of 59 wild, translocated and captive pure or hybrid Callithrix in Brazil. Pure C jacchus were captive. Wild animals included 2 C aurita, 1 C geoffroyi, and 5 C jacchus x C penicillata	9 C jacchus (5 M, 4 F)	Rectal swabs	59 (17 evaluated here)	16S rRNA (V4)	QIIME-2	Captive marmosets had decreased alpha diversity compared with wild animals. Wild and captive microbiomes differed in composition. Wild microbiomes predominantly Actinobacteria and Proteobacteria. Captive C jacchus samples predominantly Proteobacteria.	16
Sheh et al, 2020	MIT	Compared the gut microbiome of MIT marmosets including 91 healthy marmosets and marmosets with IBD (n = 59) and strictures (n = 23)	173 (85 M, 88 F)	Rectal swabs and Feces	565	16S rRNA (V4-V5)	QIIME-2 and R	Healthy captive marmosets had Bacteroidetes-dominant, “humanized” microbiome. Source-dependent differences in healthy animals based on Bacteroides and Prevotella. Clostridium perfringens elevated in stricture cases in both feces and duodenum. IBD may be associated with shifts in Bacteroides and Prevotella copri within a source population.	36
Zhu et al, 2020	Nebraska	Evaluated gut microbiome transmission between 8 unrelated male–female marmoset pairs over 2.5 mo encompassing pairing of unrelated animals	16 (8 M, 8 F)	Feces	240 (53 pre-pair and 187 post-pair)	16S rRNA (V4)	QIIME-1.9 and splinectomer	Overall microbiome profile consisted of Firmicutes (39.1%), Bacteroidetes (29.2%), Actinobacteria (26.9%), and Proteobacteria (4.0%). Paired marmosets exhibited higher similarity after pairing. Females transmit more microbes to males.	40

BI = Barshop Institute for Longevity and Aging Studies, USA; BPRC = Biomedical Primate Research Centre, the Netherlands; CLEA = Clea, Japan; DPG = Deutsches Primatenzentrum - Gottingen, Germany; EAE = experimental autoimmune encephalomyelitis; GI = gastrointestinal; HHG = Heinrich-Heine-Universität Düsseldorf, Germany; JHU = Johns Hopkins University, USA; MIT = Massachusetts Institute of Technology, USA; Nebraska = Callitrichid Research Center, University of Nebraska-Omaha, USA; RIKEN = RIKEN Brain Science Institute, Japan; SNPRC = Southwest National Primate Research Center, USA; SPF = specific pathogen-free; UMASS = University of Massachusetts Amherst, USA; rRNA = ribosomal RNA; T-RFLP = terminal restriction fragment length polymorphism analysis; qPCR = quantitative PCR; RDP = Ribosomal Database Project; QIIME = Quantitative Insights Into Microbial Ecology; CD-HIT-OTU = Cluster Database at High Identity with Tolerance Operational Taxonomic Unit; OTU = Operational Taxonomic Unit; LEfSe = Linear discriminant analysis of Effect Size .

METHODS

This systematic review adheres to the relevant criteria from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.²⁹ An electronic search of English language biomedical literature was conducted using PubMed to identify published articles on the microbiome and common marmosets. The final search date was July 2020. This search strategy used the following keywords with their prespecified MeSH headings: “marmoset AND (microbiome OR metagenomics OR 16S).” The keyword “marmoset” was synonymous with “callithrix” and “marmosets.” The keyword “microbiome” also included the terms “microbiome’s,” “microbiomic,” “microbiomics,” “microbiota,” and “microbiomes.” The keyword “metagenomics” was synonymous with “metagenome,” “metagenomes,” “metagenomics,” “metagenomically,” and “metagenomics.” Relevant preprint publications deposited in bioRxiv were searched in August 2020 using the keywords “marmoset microbiome.” Manual searching of abstracts from the American Association for Laboratory Animal Science for the national conferences held from 2017 to 2019 was undertaken to identify studies not identifiable through PubMed. Selected studies were retrieved and assessed for inclusion based on the following predefined inclusion/exclusion criteria. Studies examining the microbiota/microbiome through investigation of oral or intestinal tissue or feces in animals in the species Callithrix jacchus were included. Non-English language studies, only presenting data from select taxa, and data from non-C jacchus callitrichids only were excluded. Directed studies using methods examining an isolate or subset of organisms were excluded. Unless otherwise stated, the marmosets were considered clinically healthy for the purpose of this review.

RESULTS

A total of 33 non-duplicated articles were identified in the PubMed search. After abstract review, 10 fulfilled the inclusion criteria and were included in this review. Three preprint articles were identified. Of these 13 sources, 12 focused on the rectal, anal, or fecal microbiome of callitrichids, and 1 focused on the common marmoset oral microbiome. Of the 12 studies evaluating the lower gut, 12 had C jacchus data with 1 study including data from other callitrichids. The methods used to survey the microbiome included Human Intestinal Tract Chip (n = 1), terminal restriction fragment length polymorphism analysis (T-RFLP) (n = 1), and 16S rRNA microbiome profiling (n = 11). On evaluation of the data, the Human Intestinal Tract Chip study was excluded as the study showed that the hybridization method was not applicable to marmosets (Fig. 1).³⁰ The 12 remaining studies surveyed the microbiome of 334 healthy C jacchus and 104 animals with GI disease. For our purposes, “healthy” was defined by the lack of overt clinical disease based on the assessment of the original publication. Eleven of the 12 studies clearly reported the number of male (n = 202) and female (n = 182) marmosets used for microbiome studies, which resulted in a near even split of 52.6 to 47.4. The 12 studies analyzed for this review are summarized in Table 1.

Figure 1 .

Flow chart of the systematic review process.

The Microbiota in Healthy Marmosets

A survey of marmoset gut microbiota data revealed a wide range of variability between institutions at the highest taxonomical levels. Based on studies presenting phylum-level data of the gut microbiome, the most abundant phylum has been reported alternatively as Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, or Proteobacteria (Fig. 2). This variability in healthy gut microbiomes contrasts data from human studies where the healthy gut microbiome is consistently dominated by Bacteroidetes and Firmicutes.^25,31

Figure 2 .

Relative abundances at the phylum level observed in healthy common marmosets from subpopulations in selected publications (reference no. in parentheses). Horizontal bars denote the most abundant phylum in the population described. Vertical bars represent the relative abundance of phyla within each cohort. The most abundant phylum changed from institution to institution, with Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Proteobacteria being the dominant phylum in at least 1 population. BI = Barshop Institute for Longevity and Aging Studies, USA; BPRC = Biomedical Primate Research Centre, the Netherlands; CLEA = Clea, Japan; DPG = Deutsches Primatenzentrum—Gottingen, Germany; HHG = Heinrich-Heine-Universität Düsseldorf, Germany; JHU = Johns Hopkins University, USA; MIT = Massachusetts Institute of Technology, USA; Nebraska = Callitrichid Research Center, University of Nebraska-Omaha, USA; RIKEN = RIKEN Brain Science Institute, Japan; SNPRC = Southwest National Primate Research Center, USA; UMASS = University of Massachusetts Amherst,USA.

The earliest reported C jacchus metagenomic study compared 2 captive populations at the Southwest National Primate Research Center (SNPRC) and the Barshop Institute (BI).²⁸ The study evaluated the microbiome of marmosets raised in a specific pathogen-free (SPF) environment (BI) compared with animals raised at a primate center (SNPRC). Three clusters were observed in this study based on the levels of Bifidobacterium and Fusobacterium B (Fig. 2). Composed mainly of SNPRC animals, cluster 1 had high levels of Fusobacterium B (27%, median) and low frequencies of Bifidobacterium (4%, median). Cluster 3 associated with the barrier colony and had median Bifidobacterium abundances of 17% and presented with strong colonization with another Actinobacteria (Collinsella), Bacteroides, and Parabacteroides. A small subset composed of both SNPRC and BI marmosets made up cluster 2, in which Bifidobacterium was the most abundant organism in the microbiome (70%, median).

Dominance of the phylum Actinobacteria, of which Bifidobacterium is the most prominent member, was also observed at the Biomedical Primate Research Centre (BPRC) in Rijswijk, the Netherlands.³²  Actinobacteria was the most abundant phylum, at 66%, and was mainly represented by species from the genera Bifidobacteria and Collinsella. In contrast, Bacteroides and Prevotella, which are highly abundant in other colonies, represented <5% of the microbiome each.³² Using terminal restriction fragment length polymorphism (T-RFLP), a low-resolution method of surveying the microbiome, and quantitative polymerase chain reaction (qPCR), Shigeno et al³³ also observed high levels of Bifidobacterium (26.7%) in a colony of healthy marmosets housed at CLEA (Japan).

Multiple studies have shown Bacteroidetes as the most abundant phylum in predominantly biomedical research settings. A second report from the BI comparing young adults and geriatric marmosets found that Bacteroidetes was the most abundant phylum in both groups, representing 35% of the gut microbiome. In young adults, Actinobacteria was the second-most abundant phylum, followed by Firmicutes and Fusobacteria. The 2 most abundant families in young adults were Bifidobacteriaceae and Bacteroidaceae. In geriatric marmosets, the second-most abundant phylum was Proteobacteria, followed by Actinobacteria and Firmicutes. The 3 most abundant families in the geriatric marmosets were Succinivibrionaceae, Prevotellaceae, and Bifidobacteriaceae.³⁴ In geriatric animals, the authors reported a decreased Shannon diversity, lower abundance of Firmicutes, and higher abundance of Succivibrionaceae (both, adjusted P < .05).³⁴

An initial survey of the marmoset colony at Massachusetts Institute of Technology (MIT) showed a predominance of Bacteroidetes with high levels of families Bacteroidaceae and Prevotellaceae.³⁵ Levels of Firmicutes and Proteobacteria were similar to each other, and the most represented families of these 2 phyla were Veillonellaceae and Succinivibrionaceae, respectively. A large follow-up study with 303 samples from 91 healthy animals reported average abundances of 63% Bacteroidetes, 16.7% Firmicutes, 13.8% Proteobacteria, 4% Fusobacteria, and 1.4% Actinobacteria.³⁶ While Bacteroidetes was the most abundant phyla on average, its representation in individual samples ranged from 5% to 85% of the total abundance, which reflects the variability observed in the human microbiome.²⁵ Interestingly, despite being maintained at a single institution (MIT) and receiving the same diet and husbandry care, the microbiome profiles of marmosets from different sources did not fully converge but maintained distinctive characteristics reflective of the original source of importation.³⁶ A similar finding was observed in another study looking at the evolution of the microbiome in marmosets imported from 2 sources in Germany to the Johns Hopkins University (JHU) colony.³⁷ While the JHU colony was dominated by Bacteroidetes at day 0 and 1 year after the study, the imported animals were different at the baseline. The gut microbiome of animals from Düsseldorf were mainly colonized by Proteobacteria, Bacteroidetes, and Firmicutes, but the order of abundance shifted to Bacteroidetes, Firmicutes, Fusobacteria, and Actinobacteria after 1 year of co-housing and diet integration. The microbiome of marmosets sourced from Göttingen was not altered significantly at the phylum level, with a dominance of Bacteroidetes and lower abundances of Proteobacteria (mainly composed of class Epsilonproteobacteria) and Firmicutes. The authors note the convergence at day 100 of the microbiome profiles of the animals from Düsseldorf and Göttingen and suggest that standardization of diet and other environmental parameters played an important role in shifting the microbiome.³⁸ As in the MIT study, the convergence of the microbiome within the institution did not eliminate distinctive bacterial markers from the imported sources, even after 1 year of integration and cohousing.

Several studies also report higher Firmicutes as the most abundant phylum. Healthy females from the RIKEN Center for Brain Science (n = 9) had comparable levels of Firmicutes and Bacteroidetes, with Actinobacteria as the third-most abundant phylum.³⁹ Marmosets housed at the Callitrichid Research Center at the University of Nebraska at Omaha (n = 16) had Firmicutes-dominant gut microbiomes with lower levels of Bacteroidetes and Actinobacteria. The predominant families in the gut microbiome included Bifidobacteriaceae, Veillonellaceae, Bacteroidaceae, Acidaminococcaceae, Prevotellaceae, Lachnospiraceae, Coriobacteriaceae, Enterobacteriaceae, Porphyromonadaceae, and Succinivibrionaceae, accounting for 97% of bacterial abundance in the overall data set.⁴⁰

Lastly, a high representation of Proteobacteria has also been reported in specific cohorts^16,38,41 Samples from zoo-kept C jacchus in Brazil (n = 9) had Proteobacteria ranging from 70% to 80% of the microbial abundance, mostly due to members of Gammaproteobacteria and Epsilonproteobacteria.¹⁶ These 2 classes are predominantly known for the families Enterobacteriaceae, Succinivibrionaceae, Helicobacteriaceae, and Campylobacteriaceae. Similarly, in 8 wild callitrichids, which included 5 C jacchus × Callithrix penicillata hybrids, Proteobacteria (35%) and Actinobacteria (32%) were the most abundant phyla.¹⁶ Another study sampled a single individual at UMASS Amherst, where Proteobacteria (37.1%), Firmicutes (33.0%), and Bacteroidetes (28.1%) were most abundant. Taken together, Actinobacteria, Bacteroidetes, Firmicutes, Fusobacteria, and Proteobacteria (including reads of the novel phylum Epsilonbacteraeota) have been reported as the most abundant phylum in a healthy marmoset (Fig. 2). These reports on clinically healthy marmosets highlight the large plasticity of the marmoset gut microbiome depending on the diet, husbandry, and environment.

While the relative levels of individual bacteria may vary, genera reported across multiple institutions in the marmoset gut include Anaerobiospirillum, Bacteroides, Bifidobacterium, Collinsella, Fusobacterium, Megamonas, Megasphaera, Phascolarctobacterium, and Prevotella. There is a relative paucity of bacteria associated with butyrate production in humans, such as Eubacterium, Roseburia, Faecalibacterium, Anaerostipes, Subdoligranulum, Butyvibrio, Coprococcus, Clostridium, etc.,^42–44 so it has been hypothesized that Megasphaera could be the primary butyrogenic microbe in marmosets.^39,45 In other reports, specific members of Prevotella or Lactobacillus have been reported to act as butyrate producers, given the right environmental conditions,⁴⁶ so further research is necessary to better understand the dynamics of short-chain fatty acid production in marmosets.⁴⁶

To date there has been a single publication looking at the oral microbiome of C jacchus, and those findings indicate a similar distribution of 4 phyla: Proteobacteria, Bacteroides, Fusobacteria, and Firmicutes.⁴⁷ Compared with human volunteers, marmosets had higher levels of Fusobacteria and Spirochaetes but reduced abundance of Actinobacteria and Firmicutes.

The Microbiota in Marmosets With Disease

Of the sources reviewed, only 3 published articles and 2 unpublished studies compared healthy microbiomes with diseased states. Kap et al³² evaluated the effect of yogurt-based diets on a marmoset model of experimental autoimmune encephalomyelitis (EAE), a multiple sclerosis-like disease. The authors observed a reduced incidence of EAE in their model after water-based supplements (WBS) were changed to yogurt-based supplements (YBS). Transition to YBS was accompanied by an increased frequency of supplement availability (4 times/wk to daily) and slight changes to nutritional additives (fresh fruit and play-and-food enrichment). This dietary change caused the reduction of the percentage of marmosets with clinically evident EAE after sensitization against recombinant human myelin oligodendrocyte glycoprotein. The change in diet was also associated with decreased incidence of Callitrichine herpesvirus 3 (CalHV3). In a prospective study, marmoset twins raised on YBS were split into groups receiving WBS and YBS prior to EAE induction. The authors found 100% disease prevalence, with more severe spinal cord demyelination in the WBS group but only 75% prevalence in the YBS group. Prior to EAE induction, microbiome profiling of YBS and WBS marmosets revealed a Bifidobacteria-dominated composition (Actinobacteria, 66%; Firmicutes, 20%; Bacteroidetes, 11%). After 15 weeks on WBS and EAE induction, the percentage of Actinobacteria decreased to 38%, with compensatory increases of 13% and 15% in Firmicutes and Bacteroidetes, respectively. Interestingly, the 75% of YBS-fed marmosets with EAE did not have substantially altered microbiomes. Key bacteria positively correlated with YBS were Bifidobacterium callitrichos, B. stellenboschense, and B. simiae, and bacteria correlated with WBS included Collinsella tanakei, Bacteroides barnesiae, and Blautia stercoris. While dietary change alone did not alter the microbiome profile, the diet, EAE induction, and accompanying immune responses perturbed the composition of the microbiota. The authors hypothesized that reduced CalHV3 levels may play a role and speculated that CalHV3 loss might be linked to short-chain fatty acid production–mediated suppression of viral reactivation, which has been observed in Epstein-Barr virus.⁴⁸

Of the remaining studies reporting microbiome data for marmosets with GI disease, Shigeno et al³³ reported an increased alpha diversity and decreased Bifidobacterium spp. in the cluster associated with diarrhea, but the resolution of T-RFLP did not allow positive identification of other species. Artim et al³⁶ included only 3 animals with GI disease; however, this initial study was followed-up with a much larger study including 303 samples from 91 healthy marmosets, 62 samples from 23 marmosets with duodenal stricture, and 200 samples from 59 marmosets with IBD. Using both Analysis of Composition of Microbiomes analysis and a random forest classifier, a loss of Anaerobiospirillum and an increase of Clostridium sensu stricto 1 was observed in the gut microbiome of marmosets with duodenal strictures. Sequence data matched the 16S rRNA sequence of Clostridium perfringens. Using both microbiological culture and 16S profiling of the duodenal stricture site, an increased level of C perfringens was observed in animals with stricture. C perfringens is a major cause of foodborne death⁴⁹ and has been linked to diarrhea, Clostridial necrotizing enteritis, necrotizing enterocolitis, ulcerative colitis, and enterotoxemia in humans and other mammals.^50–52  C perfringens can encode 4 major toxins (alpha, beta, epsilon, and iota) and other lethal toxins such as perfringolysin O, enterotoxin, and beta 2 toxin.⁵⁰  C perfringens causes gas gangrene and gastric dilatation syndrome in NHP,^53–57 and C perfringens-induced gas gangrene was reported in the same institution as a 21% of mortality associated with duodenal strictures.⁵⁶ Symptoms of C perfringens-induced diseases include necrotizing inflammation of small or large intestine, diarrhea, abdominal pain, ulcers, and, in some cases, intestinal strictures.^50,58–61

Marmoset studies conducted at MIT (n = 59) and JHU (n = 6) observed shifts in Prevotella abundance associated with IBD.^36,62 Marmoset IBD was also associated with decreased alpha diversity in the larger study, as has been observed in humans.^63–66 However, human IBD is often characterized by loss of health-associated genera, including Roseburia, Faecalibacterium, Eubacterium, Ruminococcus, and Subdoligranulum,^{63–65,67–69} but these have not been reported in high abundances in healthy or diseased marmosets.^28,32,34,35 In marmosets, bacteria associated with healthy outcomes in human IBD studies, such as Bifidobacterium, Bacteroides, Collinsella, and Phascolarctobacterium, have been detected. However, Enterobacteriaceae (among these Escherichia-Shigella), Veillonellaceae, and Fusobacteriaceae, which have been associated with IBD,^63,67,68 were also observed. It is important to further determine the roles of these species within the context of marmoset biology. At MIT, a single dysbiotic IBD state was not observed, but local IBD-associated shifts involving increases in Prevotella copri and reductions in Bacteroides from 3 independent healthy baselines were observed. Studies looking at the role of Bacteroides and Prevotella spp. in IBD patients have not been conclusive,^67,70–74 but a large, longitudinal study tracking non-IBD and IBD patients found that expansion and relaxation cycles in P copri abundance were commonly observed in healthy individuals but remained stable in IBD patients.⁷⁵  P copri can thrive in inflammatory environments^76–79 and is capable of inducing IL-17 and Th17 driving cytokines (IL-6 and IL-23), better than other gut commensals such as Bacteroides fragilis, Bifidobacterium bifidum, Lactobacillus acidophilus, and Escherichia coli.⁷⁸  P copri-associated dysbiosis has been observed in the feces of new-onset rheumatoid arthritis (RA) patients, and fecal microbiota transfer from new-onset RA patients to mice induced intestinal Th17 cells and severe arthritis. In a mouse model of Prevotella-induced RA, expansion of regulatory T cells blocked disease progression.⁸⁰ In the gut, Prevotella has been linked to diarrhea, HIV-induced gut dysbiosis, irritable bowel syndrome, and more severe colitis.^81–87 Further research on the importance of increases in P copri from the baseline in healthy marmosets could help determine its role inIBD.

DISCUSSION

Summary of Key Findings

A surprising observation is that in this small number of studies, a large range of variability in the gut microbiome of captive marmosets is observed. In humans, a lot of the variability between geographic regions occurs at the genus level with the competition of Prevotella and Bacteroides. This dynamic was observed within the MIT colony, where 2 sources were strongly identified by high levels of Bacteroides while 2 sources were strongly associated with high levels of Prevotella.³⁶ The microbiome of healthy marmosets at our institution supports the hypothesis that captivity can humanize the primate microbiome,^15,18 as Prevotella 9 or Bacteroides were the predominant genera observed. This pattern mirrors microbiome abundance profiles observed in the human gut microbiome.^25,88 However, across institutions, the gut microbiome of marmosets showed a remarkable amount of variability. Changes were not limited to 2 highly similar genera from the same family, but large-scale phylum level shifts were observed leading to microbiota dominated by Actinobacteria,^28,32  Bacteroidetes,^34–36,38  Firmicutes,^39,40  Fusobacteria,²⁸ or Proteobacteria^16,38,41 depending on the study and institution.

As our understanding of the marmoset microbiome is still in its infancy, our understanding of the role of the microbiome in disease is inconclusive. High levels of Actinobacteria were reported in wild non-C jacchus marmosets, but this was not observed in captive and semi-captive settings.¹⁶ If increased levels of Bacteroidetes signal humanization of the NHP microbiome,¹⁵ high levels of Actinobacteria might better reflect the natural microbiome of the common marmoset. Characterization of Bifidobacteria isolates from common marmosets supports the idea that C jacchus may harbor unique Bifidobacterium species with genetic adaptations that facilitate nutrient uptake, suggestive of the important relationship between these bacteria and their host.^41,89 Indeed, the role of Bifidobacterium spp. in the gut includes the production and liberation of vitamins, antioxidants, polyphenols, and conjugated linoleic acids. In humans, Bifidobacteria are highly abundant in infants and are involved in the maturation of the immune system, preservation of gut barrier, and protection against pathogens.^90–93 As common marmosets are obligate exudativores,⁶ Bifidobacteria isolated from marmosets have been found to possess unique genes, such as ABC transporters, that may contribute to their persistence in the marmoset gut and their role in marmoset nutrition, possibly including the processing of complex, insoluble fibers found in gums.^89,94–97 Supporting this hypothesis, Bifidobacterium-dominated microbiomes reduced the incidence of EAE in an experimental model of multiple sclerosis.³² Shigeno et al³³ also linked decreased abundance of Bifidobacteria with increased reports of diarrhea. Ross et al²⁸ also observed a higher abundance of Bifidobacterium in clusters with good representation of SPF colony animals, which had improved health outcomes compared with conventionally raised marmosets. In humans, a decrease in the relative abundance of Bifidobacterium has been associated with multiple diseases, including diarrhea, irritable bowel syndrome, IBD, obesity, allergies, and regressive autism.^98,99 However, further studies surveying the microbiome of both wild and captive C jacchus and the prevalence of GI diseases are required to support the hypothesis that Bifidobacteria are part of the healthy, normal microbiota of marmosets.

In captive marmosets with Bacteroidetes-dominated microbiomes, shifts in Prevotella have been linked with IBD.^36,62 As the association between duodenal strictures and C perfringens infection was first proposed based on analysis of microbiome data,³⁶ we have highlighted its role in this review. However, it is well established that other enteric bacteria such as Salmonella, enteropathogenic E coli, Campylobacter, Yersinia enterocolitica, Shigella sonnei, and Klebsiella pneumonia can cause GI disease in Callitrichidae⁹ that would be associated with disruption of the microbiome. Future studies across multiple institutions are necessary to confirm these findings and better characterize the effects of GI disease and infections on the microbiome.

Factors That May Modulate Microbiome. The structure and composition of the microbiome can be influenced by many factors, including host genetics, diet, infection, medical interventions (eg, antibiotics), mode of delivery at birth, infant feeding, and geography.^25,26,100 Current reports of the marmoset microbiome suggest the importance of diet, importation, geography (source/environment), age, and co-housing,^{28,34,36,38,40} but systematic studies are lacking that assess the effects of specific dietary components, antibiotics, host genetics, and vertical transmission of bacteria. Reveles et al³⁴ focused on the effects of age on the evolution of the microbiome and found a decrease in alpha diversity and Firmicutes associated with older marmosets, with an accompanying increase in Proteobacteria of the family Succinivibrionaceae. Analysis of microbiome composition of the MIT datasets identified few differentially expressed genera between marmosets younger than 2 years, adults aged from 2 to 8 years, and marmosets older than 8 years. Differences found included changes in the abundances of Helicobacter, Prevotellaceae UCG-001, and unknown genera in the family Prevotellaceae and order Rhodospirillales (A. Sheh, unpublished data, 2020).

Microbiome changes associated with captivity are thought to be influenced by stress, diet, and the environment. While diet and the environment may play an important role in modulating the microbiome, different approaches might converge on similar stable microbial states.^15,17 Two institutions reporting high levels of Actinobacteria had very different approaches to husbandry and diet. At the BI, marmosets were housed within a barrier facility under SPF conditions to limit their exposure to potential pathogens. Marmosets were fed exclusively irradiated purified diet and an irradiated primate enrichment mix consisting of nuts, seeds, and dried fruit.²⁸ Marmosets within the barrier facility had lower levels of known pathogens, more positive reproductive outcomes, increased longevity, and Actinobacteria-dominant microbiomes. In contrast, at the BPRC, marmosets are housed in spacious cages with access to indoor and outdoor spaces enriched with branches, toys, and padded floors.³² Additionally, the diet included pellets, porridge, YBSs, fruits, Arabic gum, etc. Of the studies included in this review, BI (cluster 2) and BPRC marmosets had the highest levels of Actinobacteria.^28,32

Furthermore, it remains difficult to separate the effects of diet and environment. In a study directly comparing animals imported from 2 sources in Germany originally maintained on a distinct diet, the authors noted that baseline microbiomes were different at day 0 but observed convergence of the microbiota of imported animals by day 100, following 50 days of diet integration.³⁸ Unfortunately, due to sampling resolution, it was not possible to isolate effects due to acclimatization to the receiving institution, recovery from importation stress, and the effects of diet itself. However, there is evidence that a standardized environment and diet is not sufficient to cause full convergence of the marmoset microbiome as distinctive, source-specific microbiomes have been identified up years after integration into a colony.^36,38 Further convergence was observed by Zhu et al⁴⁰ when evaluating co-housed marmoset pairs, suggesting that in addition to standardization of diet and environment, physical transmission may be necessary to decrease variability within a colony. However, other challenges may remain because bacteria inhabiting a specific niche may not easily be displaced, as illustrated by the competition between Bacteroides and Prevotella observed within the MIT colony and different human populations.^36,88 This competition for the niche may explain why distinctive features remain after integrating marmosets from different sources. Efforts to normalize the microbiome within an institution will require standardization of the diet, environment, and co-housing as well as other interventions, such as antibiotics, that eradicate the native microbiome and facilitate colonization by other bacteria.

Technical Challenges. Outside of biological sources of variation, it is necessary to consider the effects of sample collection and library preparation as sources of potential variation between the studies. In our previous work, the microbiome of marmoset feces was compared with paired rectal swabs, and rectal swabs were found to be proxies for fecal samples unless visible fecal matter was absent. The microbiome of clear rectal swabs diverged from the fecal microbiome. These rectal swabs can be a source of variation as swabs may be sampling the mucosal microbiome, which is characterized by bacterial species such as Helicobacters that reside close to the epithelial layer in the mucosa.^35,101 In Malukiewicz et al.¹⁶ approximately 25% of anal swabs showed a >50% abundance of the class Epsilonproteobacteria (now reclassified as the phylum Epsilonbacteraeota), but this high level of variability within a study can affect conclusions derived from thedata.

Similarly, variation in DNA extraction techniques, 16S rRNA primers selection, library preparation protocols, and bioinformatics can bias the amplification, detection, or identification of specific taxa. Within the studies reviewed, the V3-V4 region was amplified in 4 studies, the V4 region was used in 3 studies, the V4-V5 region was targeted in 2 studies, and the V1-V3 region was used in 1 study. In a study excluded from this review for its exclusive focus on Bifidobacterium species, the V1-V3 primers detected >25% Bifidobacterium abundances in more than one-half of marmosets surveyed from multiple primate centers,⁸⁹ including the New England Primate Center, which was a founding colony of the MIT cohort. In contrast, our study detected a 1.0% abundance of Bifidobacterium using V4-V5 primers.³⁶ Further investigation is necessary to determine if this is due to changes associated with the change of settings or differences due to techniques. However, unpublished data sampling a subset of the MIT colony using unbiased whole-genome shotgun sequencing shows low abundance of Bifidobacterium, which supports the idea of a microbiome shift that occurred after moving to MIT (A. Sheh, unpublished data,2020).

The authors of Ross et al.²⁸ also allude to another potential source of error in their study, as they highlight that Fusobacterium B is often mischaracterized as Clostridium cluster XIX due to older taxonomic classifications. Similarly, databases used to assign taxa can place the genus Helicobacter either under the phylum Proteobacteria or in the proposed phylum Epsilonbacteraeota. In both of these scenarios, these highly abundant organisms can be classified in different phyla from study to study, so it is important for scientists to be cognizant of taxonomical changes and use the same databases when comparing studies. More rigorous analysis is needed to understand how sampling, methodology, and batch effect may affect comparisons across datasets.¹⁰²

Marmosets as Animal Models. Healthy microbial communities can be considered as a set of dynamic states and not just a single static state (Fig. 3), as supported by the data presented in this review of clinically healthy marmosets. Data from MIT showed the existence of multiple healthy states within a single institution as well as multiple states associated with IBD within the same colony.³⁶ However, this might be more indicative of the IBD diagnosis, an umbrella term for a variety of GI disorders. While no microbial community can shield the host from all disease, we hypothesize that specific microbial communities might be more advantageous due to coevolution (eg, better adaptations to natural diet) or proffer a higher level of resistance and resilience to perturbations that can lead to disease. Is it possible that Actinobacteria-dominant microbiomes are more resilient to transformation into dysbiotic states compared with more humanized microbiomes acquired in captivity and dominated by Bacteroidetes?^15,32 Could easily perturbed healthy states account for the high levels of GI diseases, IBD, and duodenal strictures found in captive common marmosets worldwide?^9–11,13 Do humanized microbiomes improve the modeling of human disease in animal models? It may be possible to leverage the plasticity of the marmoset microbiome to help elucidate the role of the gut microbiome in NHP and human health and disease. As NHP are widely used as models of human disease, these models can be further refined to probe the contribution of the microbiome by using marmoset cohorts with baseline microbiotas that resemble different human populations (eg, Prevotella high vs Bacteroides high). Alternatively, fecal microbiome transplantations from healthy and diseased humans have been commonly used to confer disease phenotypes to rodents and should be further explored in common marmosets.

Figure 3 .

Hypothetical diagram of possible healthy and diseased states in the marmoset gut microbiome. Clinically healthy marmosets exhibit a wide range of microbiome profiles representing multiple healthy microbiome states. Based on advantages conferred in host-microbiome interactions, the range of healthy states confers varying levels of resilience to perturbations such that mild shifts might lead some healthy states to a diseased state, while severe shifts are required to shift other healthy states. Further research is needed to determine which healthy microbiota confer the most robust resistance to perturbations.

CONCLUSIONS AND FUTURE DIRECTIONS

Changes in microbial composition have been described in healthy and diseased marmosets, but factors influencing the severe changes in microbial composition have not been established. Multi-institutional, prospective, and longitudinal studies are required with the inclusion of multiple testing methodologies to determine sources of variability in the reporting of marmoset microbiomes. Currently, the field is dominated by 16S rRNA surveys (Table 1), but incorporating higher resolution metagenomics sequencing data will be of value to provide better resolution of bacterial taxa as well as survey the functions carried out by these bacteria. Furthermore, metagenomics data will provide data regarding the communities of viruses, fungi, and parasites that are an important part of the microbiome. Analysis of the microbiome needs to be paired with other datasets such as transcriptomics, proteomics, metabolomics, etc, to obtain further biological insights. Large sequencing initiatives aimed at analyzing the genetic diversity of marmoset colonies will also be important to determine the effect of host genetics on both GI disease incidence and microbial composition. To fully realize the promise of using marmosets to model different microbiome states observed in humans, methods of microbial manipulation, whether it is diet, enrichment, fecal microbiome transplantation, etc, need to be established to modulate and maintain robust and resilient microbiome communities in marmoset colonies and reduce the incidence of idiopathic GI disease.