The bifidobacterial distribution in the microbiome of captive primates reflects parvorder and feed specialization of the host

Bifidobacteria, which commonly inhabit the primate gut, are beneficial contributors to host wellbeing. Anatomical differences and natural habitat allow an arrangement of primates into two main parvorders; New World monkeys (NWM) and Old World monkeys (OWM). The number of newly described bifidobacterial species is clearly elevated in NWM. This corresponds to our finding that bifidobacteria were the dominant group of cultivated gut anaerobes in NWM, while their numbers halved in OWM and were often replaced by Clostridiaceae with sarcina morphology. We examined an extended MALDI-TOF MS database as a potential identification tool for rapid screening of bifidobacterial distribution in captive primates. Bifidobacterial isolates of NWM were assigned mainly to species of primate origin, while OWM possessed typically multi-host bifidobacteria. Moreover, bifidobacterial counts reflected the feed specialization of captive primates decreasing from frugivore-insectivores, gummivore-insectivores, frugivore-folivores to frugivore-omnivores. Amplicon sequencing analysis supported this trend with regards to the inverse ratio of Actinobacteria and Firmicutes. In addition, a significantly higher diversity of the bacterial population in OWM was found. The evolution specialization of primates seems to be responsible for Bifidobacterium abundance and species occurrence. Balanced microbiota of captive primates could be supported by optimized prebiotic and probiotic stimulation based on the primate host.

Bifidobacterial species detected by MALDI-TOF MS. Bacterial colonies with variable cultivation characteristics from bifidobacterial selective media were isolated for further identifications (Suppl. Tab. 1). From a total of 326 isolates, 210 were F6PPK-positive bifidobacteria and the remaining 116 isolates (isolated mainly from WSP-MUP) were F6PPK-negative gas producing clostridial rods or cells with sarcina morphology. All F6PPK-positive strains were also identified with MALDI-TOF MS using an expanded custom database for bifidobacterial identification. 54% of the strains (n = 112) were assigned to 18 different bifidobacterial species, 36% (n = 76) were assigned only to the Bifidobacterium genus, and 11% (n = 22) were not identified reliably ( Fig. 2A, C).
B. parmae, B. imperatoris/saguini, and B. ramosum were the most frequently identified species in the NWM, whereas B. dentium and B. catenulatum/pseudocatenulatum were most common in the OWM. Interestingly, B. adolescentis was equally represented in both primate parvorders. A more diverse species representation of bifidobacteria was found in the NWM (14 spp.) compared to the OWM (5 spp.). Genus-level assignment and the presence of not reliable identifications (NRI) was mainly detected in the NWM. Related presumed species compliance and the closest match of Bifidobacterium spp. strains was found predominantly with B. parmae and B. stellenboschense in the NWM, and B. angulatum/merycicum in the OWM (Fig. 2B). PR8 Red-handed Tamarin Table 1. List of monkey hosts kept in zoological gardens. General information about primate taxonomy, parvorder and feed classification. Primate general feeders (n = 52) were grouped to 4 individual feed categories based on proportion of dominating feed components -frugivore-omnivore, frugivore-folivore, frugivoreinsectivore, and gummivore-insectivore. Zoo, zoological garden; CZ, Czechia; SK, Slovakia; NWM, New World monkey; OWM, Old World monkey. (B) Cultivation counts of bacteria per feed category: frugivore-folivore (n = 8), frugivore-omnivore (n = 21), frugivore-insectivore (n = 13), gummivore-insectivore (n = 10). Asterisks (*) denote statistically significant differences as determined by t-test and ANOVA (p < 0.05). An agreement between the MALDI-TOF MS species assignment and the sequencing of 16S rRNA gene was confirmed for 38 strains. Only 3 strains were identified differently by the two methods. Namely, strain N127 identified as B. faecale by 16S rRNA gene sequencing was mistaken for B. adolescentis by the MALDI-TOF MS, B. imperatoris for NRI (N40), and PEBJ_s for B. imperatoris/saguini (N50). Interestingly, mentioned strain N50 together with N74, N94, N97, and N115, exhibiting MALDI-TOF MS NRI score (< 1.69), were considered potential novel species of bifidobacteria. In addition, this sample set also contained 5 problematic strains (N16, N70, N81, N119, and N125), whose 16S rRNA gene sequencing failed repeatedly and thus their MALDI-TOF MS identity was not confirmed.
Microbial community shifts were found between the NWM and OWM parvorders. The relative abundance of phylum Actinobacteriota (W = 13) and Campylobacterota (W = 12) was significantly higher in the NWM  www.nature.com/scientificreports/ compared to the OWM as confirmed by the ANCOM statistics. Meanwhile, the phylum Firmicutes showed an opposite trend, which was however not statistically significant (Fig. 4A). The difference in the Actinobacteriota can be attributed specifically to the family Bifidobacteriaceae which was significantly higher in the NWM (16%) compared to the OWM (3%) (W = 139) (Fig. 4B, Supplementary S2). These findings corroborate the cultivation results.

Discussion
Dynamic microbial communities aid the living and surviving of animals in changing environmental conditions, including habitat degradation, captive breeding, and diet. If microbial balance of the host is disturbed and dysbiosis occurs, there is a presumption of disease development 5,53,54 . Among others, commensal microorganisms, such as bifidobacteria, play a crucial role in maintaining the gut homeostasis [55][56][57] . Bifidobacterial diversity and adaptation are connected to their hosts and environments with possession of specific genomic traits 58-60 which includes primates 42 . Two independent approaches, cultivation with subsequent MALDI-TOF MS identification and amplicon sequencing of the V4 region of the 16S rRNA gene, were used to analyse the microbiome composition and the prevalence of bifidobacterial species in primate gut microbiota. NWM are a significant source of cultivable bifidobacteria with average counts of 10 8 CFU g -1 of faeces compared to the OWM with four orders of magnitude lower counts. Interestingly, although no health complications were evident, FS of primate individuals with reduced or undetectable cultivation counts of bifidobacteria contained Clostridiaceae, mainly displaying sarcina morphology. This was mainly observed in individuals belonging to the OWM parvorder (Suppl. Tab. 1). Spore-forming bacteria identified as Sarcina ventriculi (syn. Clostridium ventriculi) were previously isolated also from primates without apparent health problems [61][62][63] . Although they are considered pathogens 64 , this may indicate sarcina as common bacteria of the primate gut microbiota. In the gut of NWM, the abundance of sarcina is probably decreased by the presence of bifidobacteria, which exhibit potential to hamper growth of clostridia [65][66][67] . The inverse ratio and balancing of the bifidobacteria and clostridia are typically described in the gut microbiome of infants [68][69][70] .     71 showed that the screening of bacterial isolates from environmental samples can be performed efficiently, quickly, and inexpensively using MALDI-TOF MS and should be refined by implementation of environmental strains into the database. Within our study, the use of an extended custom database for MALDI-TOF MS allowed reliable species differentiation and identification of wild bifidobacterial isolates. Higher species diversity was observed in NWM. Interestingly, the multi-host species B. adolescentis was present among most screened captive primates. In OWM B. dentium and B. catenulum/pseudocatenulatum, that are common species of the human gut microbiota, as well as B. adolescentis, were found 72 . Lugli et al. 42 detected B. adolescentis and B. dentium in OWM as well, and indicated possible joint development and evolutionary relatedness. In contrast, NWM exhibited the presence of cultivable bifidobacteria mainly with primate origin. Interestingly, Brown et al. 73 pointed out that marmoset bifidobacteria are closely related to those in tamarins. Furthermore, we found that bifidobacterial species variability in NWM significantly exceeds that in OWM. Furthermore, we hereby confirmed that we can re-isolate recently described primate Bifidobacterium spp. also from primate species with various captive locations other than those from which bifidobacteria were originally isolated.
Moreover, MALDI-TOF MS screening allowed us to identify 5 potential novel species of bifidobacteria isolated from tamarins that were confirmed by 16S rRNA gene sequencing. That indicates primate gut as a promising environment for the discovery of novel species of bifidobacteria 42,48,50 . To achieve an accurate identification of potential novel species, a combination with other methods, such as sequencing of phylogenetic markers [74][75][76] , multi-locus sequence typing 77 , and genome sequencing 78 , should be included.
The significantly lower species richness and high relative abundance of bifidobacteria in NWM compared to OWM was confirmed by sequencing of the V4 region of 16S rRNA gene. The relative abundance of Bifidobacteriaceae reached 16% in the NWM and only 3% in the OWM. The same trend was also detected for Prevotellaceae and Veillonellaceae. In particular, marmosets and tamarins exhibited 32% bifidobacterial abundance compared to 0.03% in the OWM 42 . This high relative bifidobacterial proportion in adult marmosets could be a consequence of their housing as family groups and their constant subjection to the gut microbiota of other individuals 73 . Conversely, Lachnospiraceae, Oscillospiraceae, Ruminococcaceae, and Spirochaetaceae showed an opposite trend with high abundances in OWM. Interestingly, we showed that the captive NWM have high relative levels of bifidobacteria, which is similar to what they display in the wild 47,[79][80][81] . It indicates that NWM gut is a rich bifidobacterial environment that is also supported by other studies 42,82,83 . In contrast to our results in captive individuals, some microbiome studies point to a slightly increased bifidobacterial relative proportions in wild OWM as well 84,85 . Although the captivity was previously described as a factor influencing the presence of Actinobacteria in the primate gut microbiome 14,41 , our results suggest that it is probably not as strong as the affiliation to the primate parvorder, which seems to be considerably more significant.
Primate gut microbiome seems to be significantly modified by dietary changes of the host species and geography 14 . Frugivore-insectivores and gummivore-insectivores possessed significantly more abundant Bifidobacteriaceae compared to frugivore-omnivores and frugivore-folivores. Interestingly, if insects constitute an important component of the diet, bifidobacteria are highly abundant. Ecologically beneficial symbionts leading to host evolutionary dependence have been previously described in other animal taxa, such as sap-feeding insects, which generate essential amino acids exclusively for their microbial symbionts 86 . Bifidobacteria are known as a commensal bacterial group of insects with social life 87 , whereas the importance of insects in the diet of primates in relation to bifidobacterial occurrence remains unclear.
Although captive feeding inevitably modifies primate gut microbiome to decreased diversity, the feed optimization could improve the animals health condition 40 . In contrast to Amato et al. 88 , who state that the host phylogeny is stronger driver in shifts of microbial composition than the diet and geographic location, our results suggest that both diet and the host itself affect the microbiome composition, especially the relative abundance of Bifidobacteriaceae. Moreover, it is important to mention, that the diet of captive animals usually includes fruits, vegetables, and leaves that may not completely match the available components present in the wild. In addition, the natural microbiota reflects diet seasonality and location that may affect trophic interactions in the gastrointestinal tract of the host 89,90 .
Clayton et al. 91 confirms that modified diet in captive primates is related to the alteration of microbiome composition and host health. Captive primate individuals susceptible to health disorders may show clinical signs including chronic diarrhoea, weight loss, lethargy, cardiac disease, and poor reproductive success 9,12,92,93 . Therefore, it is necessary to further monitor the relationship between the microbiome, diet, and the health of captive primates 40 . Microbiota modulation is an effective and affordable strategy for host health support of threatened animals 5 . Therefore, applicable mitigation strategies such as optimized dietary 40 and prebiotic interventions 94 could be pursued towards supporting balanced microbiota in captive primates. Moreover, probiotic supplementation with focus on bifidobacteria, that naturally colonize primate guts, can be a further promising approach 42,43,95 . Furthermore, this may provide a potential approach in human probiotic intervention. Due to the ever-decreasing diversity of the human microbiome through diet and antimicrobial intake, the microbiome of originally living evolutionarily close relatives has the potential to design a probiotic that is no longer part of the human microbiota and could have the potential to strengthen health 96 . Probiotic intervention should be optimized according to the gut microbiota composition and should be supported by appropriately selected prebiotic stimulation in synbiotic mixtures for long-term maintenance of balanced microbiome and host health.

Materials and methods
Sampling and cultivation analysis. Faecal samples of primate hosts (n = 52) belonging to two parvorders, NWM (n = 24) and OWM (n = 28), were preliminary screened for quantitative content of cultivable bifidobacteria. The list of primate hosts and classification into parvorders and feed category is shown in www.nature.com/scientificreports/ (all Czechia), Bojnice, and Bratislava (both Slovakia) between 2017-2019. FS were collected in tubes containing dilution buffer (5 g L -1 tryptone, 5 g L -1 nutrient broth No. 2, 2.5 g L -1 yeast extract (all Oxoid, Basingstoke, UK), 0.5 g L -1 L-cysteine, 1 mL L -1 Tween 80 (both Sigma-Aldrich, St. Louis, Missouri, USA), 30% glycerol (VWR, Radnor, Pennsylvania, USA), and glass pearls for homogenization. Media were prepared in an oxygen-free carbon dioxide environment 97 and then sterilized. After sampling, the tubes were stored at -20 °C and within the 14 days transported into the laboratory for analysis. Then, decimal serial dilutions of FS were spread on the following media. Wilkins-Chalgren Anaerobe Agar was supplemented with 5 g L -1 GMO-Free Soya Peptone (both Oxoid), 0.5 g L -1 L-cysteine, and 1 mL L -1 Tween 80 to determine total counts of anaerobic bacteria (WSP medium). Moreover, two selective media were used for bifidobacterial quantification and isolation: WSP-NORF (WSP agar supplemented with 100 mg L -1 of mupirocin, 200 mg L -1 of norfloxacin (both Oxoid), and 1 mL L -1 of acetic acid (Sigma-Aldrich) 52 ) and WSP-MUP (WSP agar supplemented with 100 mg L -1 of mupirocin and 1 mL L -1 of acetic acid 98 ). All plates were incubated anaerobically using GENbag anaer (bioMérieux, Craponne, France) at 37 °C for 2 days.
Isolation and culture identifications. Based on variable cultivation characteristics, the isolation of colonies from selective media and consecutive sub-cultivation was performed in tubes containing WSP broth under anaerobic conditions 97 at 37 °C for 1 day. Whether a culture belonged to Bifidobacterium spp. was verified by fructose-6-phosphate phosphoketolase (F6PPK) test with cetrimonium bromide for cell disruption according to Orban and Patterson (2000) 99 . Subsequently, bifidobacterial isolates were identified to the species level using Matrix-Assisted Laser Desorption/Ionization Mass Spectrometry (MALDI-TOF MS) with ethanol-formic acid extraction procedure with HCCA matrix solution according to the manufacturer's instructions (Bruker Daltonik GmbH, Bremen, Germany). An extended custom database (based on Bruker Biotyper software tools), which included 50 additional bifidobacterial species in addition to the already available entries, was used for identification. An overview about the database entries is provided in Suppl. Tab. 3. Stock cultures of bifidobacteria were stored at -80 °C in 30% glycerol.
Selected isolates (n = 46) were further identified by 16S rRNA gene amplicon sequencing. DNA was isolated from freshly grown bifidobacterial cultures in WSP broth using PrepMan Ultra™ (Applied Biosystems, Waltham, Massachusetts, USA) according to manufacturer's instructions and stored at -20 °C. Primers 285F (5′-GAG GGT TCG ATT CTG GCT CAG-3′) and 261R (5′-AAG GAG GTG ATC CAG CCG CA-3′) were used for PCR amplification of nearly the full 16S rRNA gene according to Kim et al. 100  The DNA concentration of each sample was determined using the Qubit 1X dsDNA HS Assay Kit (Invitrogen, Paisley, UK) and a Qubit fluorometer. Subsequent library preparation and sequencing were performed by Novo-Gene (Cambridge, UK). As amplicon sequencing method supports only shorter fragments, the V4 region of the 16S rRNA gene (300 bp fragments) was amplified using primers 515F (5′-GTG CCA GCMGCC GCG GTAA-3′) and 806R (5′-GGA CTA CHVGGG TWT CTAAT-3′) and a Phusion High-Fidelity PCR Master Mix (New England Biolabs, Ipswich, Massachusetts, USA). The library was prepared using the NEB Next® UltraTM DNA Library Prep Kit for Illumina and paired-end 250 bp sequencing was performed using the NovaSeq machine (Illumina, San Diego, California, USA). The resulting sequences were submitted to the NCBI database with the accession number ERP128111. Amplicon sequence variants (ASV) were obtained using the DADA2 pipeline (bioconductor-dada2 v1.16.0) 103 and Silva non redundant database v138 104 (Supplementary S3) with custom manual species assignment. The depth of sequencing of the resulting data was normalized by rarefaction to the lowest sequencing depth (42 134 sequences/sample) and a relative abundance on several taxonomic levels in different variable groups were explored (Supplementary S4). Total bacterial diversity was expressed as Shannon entropy 105 , the population richness was expressed as simple feature or ASV counts and the evenness was expressed as Pielou's index 106 . Statistical analyses. Counts of bacterial colonies in log CFU g -1 within the parvorders and feed categories are shown as boxplots. The normality of data was evaluated by Shapiro-Wilk W test (α = 0.05). Differences in bacterial counts were assessed using a Mann-Whitney U Test (α = 0.05) within the parvorders, and a oneway ANOVA within the feed categories (α = 0.05) using STATISTICA software (StatSoft, Prague, Czechia) and Microsoft Office Professional Plus 2016.
To detect differentially abundant taxa between the sample categories, the ANCOM statistical test 107 was used from the package skbio v0.5.2 (scikit-bio.org). The one-way F statistics from the scipy package v1.4.1 108 was used to determine that statistical significance with α = 0.05. Several categories of the data were explored on both the Phylum and Family level. Furthermore, the bifidobacterial sub-population was extracted for each sample and the differentially abundant species were calculated. Statistically significant results are presented in form of boxplots (Supplementary S2).