Dynamic age-associated changes and their driver microbes in healthy gut microbiota of captive crab-eating macaques


 Background: Previous population studies have indicated age-associated changes in the gut microbiota. However, the actual age effects on microbiota are inevitably confounded by varying environmental factors such as diets and antibiotic use in the populations. Captive crab-eating macaques reared in a well-controlled environment can provide a useful model to recapitulate dynamic age-associated changes in the healthy primate gut microbiota.Results: We show evidence supporting lifelong age-associated changes in the healthy gut microbiota of captive macaques. The Firmicutes to Bacteroidetes ratio and beta diversity but not alpha diversity changed significantly with age. The most significantly age-associated genera were mainly composed of commensals, such as Faecalibacterium . Unexpectedly, a subset of the age-associated microbes were suspicious pathogens such as Helicobacter and Campylobacter , which were enriched in infant macaques, and possibly associated with gut mucosa development. These age-associated microbes were main contributors to the gut microbiota networks. Importantly, topology analysis showed that connectivity of these networks changed with age, and its rapid decrease in elderly macaques might indicate altered microbial interactions associated with host aging. Prevotella 9 , one of the most abundant age-associated genera, was the driver responsible for the gut microbiota maturation from infants to young adults. In adults, Rikenellaceae RC9 gut group and Megasphaera were two key drivers that continuously played an active role in driving microbial community changes of across different stages of adulthood. We also showed evidence of age-associated changes in gut microbial phenotypes and functions, in particular pathways of immunomodulatory metabolite synthesis, and metabolism of lipids and carbohydrates. The driver microbes were key players involved in these functions.Conclusions: Our current study in captive macaques demonstrate evident age-associated changes during the lifelong process of healthy gut microbiota development. The enrichment of suspicious pathogens in healthy infant macaques might indicate the importance of appropriate exposure to these microbes for the developing immune system. The current study provides new insights into the pivotal role of driver microbes and microbial interactions in gut microbiota, and further underlines the importance of network analysis in microbiome studies. Our findings also provide a baseline for better understanding of disease-related changes in the primate gut microbiota.

0.1% were included in the networks. Surprisingly, although not preferentially selected, the ageassociated genera were found to be the major components of these networks. The gut microbiota network in infants had the lowest connectivity of interactive in infants, as indicated by small Maximal Clique Centrality (MCC) scores (total MCC score = 56) (Fig. 5a and 6a). The network developed into a more mature stage in young adults (total MCC score = 274) (Fig. 5b and 6a), and had the highest connectivity in the middle-aged (total MCC score = 3688) (Fig. 5c and 6a). Unexpectedly, although similar gut microbiota diversities were found between the elderly and middle-aged, the network connectivity dramatically decreased in the elderly (total MCC score = 83) (Fig. 5d and 6a).
We then utilized cytoHubba to analyze hub genera, which were supposed to identified by ranking their centralities MCC and EcCentricity (EPC) scores. Among the hub genera shown in Fig. 6a, Prevotella 9 was the only one shared by all four age groups as well as the network constructed using all samples ( Fig. 6a and 6b). The inter-genera interactions mediated by Prevotella 9 could be of potential importance. The strongest positive interactions in the microbial communities were found in Prevotella 2 and Alloprevotella with Prevotella 9 in infants. In addition to Prevotella 9, Helicobacter and Prevotella 2 were another two important hub genera in infants. The role of such interactions mediated by these genera, in particular Prevotella 9, gradually diminished with age, and were in part replaced by interactions mediated by hub genera negatively associated with age, such as Ruminococcaceae UCG-002 and Rikenellaceae RC9 gut group.
Moreover, we used NetShift analysis to detect rewiring between microbiota networks, and identified key driver microbes responsible for the changes (Fig. 6c and Table. S3). Prevotella 9 was found to be the only driver genus responsible for the microbial changes between infants and young adults.
Novel interactions with Prevotella 9 were established in the gut microbiota of young adults compared to that of infants. As for adults, multiple potential drivers were identified. Among these drivers, Rikenellaceae RC9 gut group and Megasphaera are the two key driver genera that contribute to the long-term development of gut microbiota in adults. Another five genera including Dialister, Christensenellaceae R-7 group, [Eubacterium] coprostanoligenes group, Ruminococcaceae UCG-005 and Ruminococcaceae UCG-002 group are involved in the change of gut microbiota between young adults and the middle-aged. Another five genera including Ruminococcaceae UCG-014, Holdemanella, Succinivibrio, Alloprevotella, Lachnospiraceae UCG-007, and Prevotella 2 are involved in the change of gut microbiota between the middle-aged and the elderly.

Age-associated microbial phenotypes and functions and their correlations with gut microbiota
To understand the potential function impact of age-associated taxonomic changes in gut microbiota, the microbial phenotypes were predicted using BugBase and compared among age groups. Anaerobic and Gram-positive phenotypes was significantly up-regulated, whereas facultative anaerobic and Gram-negative phenotypes were down-regulated in the middle-aged and elderly groups compared to infants (all P < 0.01) (Fig. 7a). In line with these findings, Spearman correlation analysis showed that, the anaerobic and Gram-negative phenotypes significantly decreased (r = -0.37, P=1.2× 10 -4 and r = -0.34, P = 4.3× 10 -4 respectively) with age, whereas the facultative anaerobic and Gram-positive phenotypes significantly increased with age (r = 0.42, P = P = 8.7 × 10 -6 and r = 0.34, P = 4.3 × 10 -4 respectively) (Fig S6).
We also determined age-associated changes in gut microbial function using the software Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt), and identified 152 Kyoto Encyclopedia of Genes and Genomes (KEGG) modules to be significantly associated with age (Table. S2). The principle component analysis (PCA) plot derived from the abundance of KEGG modules revealed remarkable differences in microbial functions among age groups, showing a similar pattern with beta diversity (Fig. 7b). We observed significant correlation between these microbial functions and age. As shown in the heatmap in Fig. 7c, metabolic pathways that were the most positively associated with age were mainly involved in biosynthesis and metabolism of lipids and carbohydrates. And metabolic pathways that were the most negatively associated with age were mainly involved in biosynthesis of immunomodulating metabolites such as lipopolysaccharides, and metabolism of polyunsaturated fatty acids and vitamins such as folate and riboflavin. Noteworthy, strong correlations were found between these age-associated microbial functions and gut microbes, in particular the hub genera and drivers (Fig. S8). Prevotella 9, was the core genus that was involved in these functions.

Discussion
By using the NHP model of captive crab-eating macaques, we revealed remarkable lifelong ageassociated changes in gut microbial composition and functions. Moreover, our study identified hub and driver microbes that holds a potential significance in the age-associated microbial interplay.
Given the similarities between the captive crab-eating and humans, these findings could provide better understanding of age-associated changes in the human gut microbiota.
The gut microbiota of captive macaques in this study showed similarities to that of humans, especially those in developing countries [12,21,23,24]. In line with human and other NHPs, the gut microbiota of our captive crab-eating macaques was dominated by Firmicutes and Bacteroidetes across all ages Age-associated factors, such as diets and life styles, rather than age itself, might actually contribute to the increase of alpha diversity in human populations.
Nevertheless, the remarkable age-associated changes including the F/B ratio and beta diversity as well as network topology emphasized actual effects of age on the gut microbiota in captive macaques ( Fig. 1b). The F/B ratio is considered as an indicator of maturation and development of gut microbiota [28], and has been reported to be involved in health-related conditions or diseases such as obesity [29]. In the current study the F/B ratio increased in adult macaques, and decreased in elderly macaques (Fig. 1b), resembling observation in humans [28,30]. It could be due to increased Firmicutes and decreased Bacteroidetes with age (Fig. 3f). Interestingly, although middle-aged and elderly macaques had similar beta diversity, evident reduction of connectivity in elderly macaques, indicating a decline of microbial interactions. Such findings suggest that, network connectivity could be more sensitive than the F/B ratio and biological diversity to detect age-associated changes in the gut microbiota.
Moreover, the age-associated microbes identified in captive macaques could be involved in the host's development and aging in good health (Figs. 3 and 4). These microbes could play distinct roles dependent of their direction of age-correlation. A large proportion of these age-associated genera decreased with age, including those enriched in infants. The composition and activities in the infant gut microbiota has been engaged in the host's early development and a variety of diseases, such as allergy and autisms [5, 31, 32]. These genera negatively associated with age in fact consisted of at least two distinct groups. First, these genera contained potential commensals, which were active players in the early development of gut microbiota (Fig. 4b, S4, and S6). The interplay between these Faecalibacterium prausnitzii is one of the most abundant anti-inflammatory commensal bacteria in the colon, and was reduced in Crohn disease patients [40]. We also notice that Bifidobacterium, the key probiotics for the metabolism of oligosaccharides in breast milk [43], reduced with age. The abundance of Bifidobacterium was lower in our post-weaning infant macaques than that in lactating macaques reported by Rhoades et al. [9].
Second, these bacteria negative associated with age also contained a number of suspicious pathogens, especially enteropathogens (Fig. 4b, S4, and S6 report that 8-month infant remained asymptomatic for diarrhea were enriched for the species [9]. It should be taken into account that all macaques in the current study were in good health. Therefore, the gradual decrease of these suspicious pathogens with age might associated with the maturation of gut mucosal barrier. In addition, recent studies have reported possible effects of pathogens protecting the host against allergic sensitization [48,49]. In our captive macaques the suspicious pathogens with their abundance under control might allow "good" exposure for the proper training of the host's immune system. In line with such findings, biosynthesis of bacterial toxins was also negatively associated with age, further suggesting a potential role of these age-associated microbes in modulation of the host's immunity. While the roles of the microbes negatively associated with age remained largely unclear, they could be related to the host's healthy aging (Fig. 4b, S4, and S6). A subset of these microbes has been implicated to be involved in diets and energy metabolism, especially lipid metabolism, as well as diseases. Importantly, the genus Lactobacillus, which was highly abundant increased in adult macaques, are widely used probiotics with potential effects on lipid metabolism [50]. Eubacterium coprostanoligenes was identified as a cholesterol-reducing anaerobe [51]. In addition, recently a population-based study linked Genera Christensenellaceae R-7 group, Ruminococcaceae (UCG-002, and UCG-010), and Lachnospiraceae FCS020 group with circulating metabolites related to blood lipids [52]. Candidatus soleaferrea was increased in a randomized controlled trial of hypocaloric diet with Hass avocado [53]. In addition, Treponema 2, Rikenellaceae RC9 gut group, Prevotellaceae UCG-003 were increased in rats with isoproterenol-induced acute myocardial ischemia [54], whereas in a metaanalysis Christensenellaceae R-7 group was found to be reduced in patients affected by intestinal diseases [55]. In line with these findings, changes of microbial functions related to metabolisms of lipids and carbohydrates increased with age (Fig. 7b). Intriguingly, the archaea Methanobrevibacter increased with age in the gut microbiota of our macaques. Although the reported role of these methanogen in host health remain controversial, our results indicated that such increase of Methanobrevibacter abundance with age might not necessarily affect the health of the host.
This study further highlights the substantial role of driver microbes in age-associated changes of the gut microbiota (Figs. 5 and 6). Genus Prevotella 9, with a high abundance in our captive macaques, was identified as the most important hub mediating large proportion of microbial interactions in gut microbiotas across all ages. And it acted as the key driver responsible for the gut microbiota maturation from infants to young adults. The exact biological significance of Prevotella 9 in the context of integrative bacterial community and microbiota development has yet to be further elucidated. A recent reanalysis of existing gut metagenomes from NHPs and humans reported that Prevotella were prevalent in primate gut microbiota of different host species [20]. In line with such finding, the Prevotella 9 genus was highly abundant across all ages with gradual age-associated decrease in our captive macaques. The high abundance of the genus in primates could be strongly associated with plant-based, low-fat diets [22]. In addition, the high abundance of Prevotella in humans and NHPs might also have possible implications for host-microbiota coevolution [56].
Although Prevotella 9 remained abundant in adult macaques, its level decreased with age, and possibly freed up space for other microbes that were necessary for further microbiota development, such as Rikenellaceae RC9 gut group and Megasphaera. Such shift of driver microbes could in turn impact the changes of gut microbiota phenotypes and functions.

Conclusions
In summary, by using captive crab-eating macaques to control confounding factors, the current study demonstrates evident age-associated structural and functional changes in the healthy gut microbiota during the host's development and aging. Our key findings of age-associated microbes, composed of both commensals and suspicious pathogens, highlight the potential importance of appropriate bacterial exposure and community balance for the host. Moreover, the hub genera and drivers identified by network topology analysis probably play a pivotal role as core microbes in the microbial communities, and are responsible for the maturation and development of primate gut microbiota. By characterizing the age-associated healthy gut microbiota, the current study also provides an important baseline for better comparison and understanding of disease-related changes in the primate gut microbiota.

Animals in the study
A total of 104 male crab-eating macaques from Guangdong Xiangguan Biotechnology Co. Ltd.
(Guangzhou, China) were included in the current study. All of the animals were confirmed to be in good health by records and veterinary examination prior to the study. These animals were composed of four different age-groups (N=26 for each group), including infant (1-2 years old), young adult (4-6 years old), middle-aged group (8-10 years old), and an elderly macaques (≥13 years old). Postweaning infant macaques were selected to reduce possible effects of breastfeeding. All animals were kept in a well-controlled environment with moderate room temperature (16-28 °C) and relative humidity of 40%-70%, as well as a 12/12-hour light-dark cycle. The study complied with protocols approved by the Animal Ethics Committees of Guangdong Institute of Applied Biological Resources, and were in compliance with the Guide for the Care and Use of Laboratory Animals [57].

Stool sample collection and DNA extraction
Rectal swab samples were freshly collected from each monkey, and stored at -80 °C immediately until DNA extraction. Microbial DNA was extracted using TIANamp Stool DNA kit (Cat.#DP328, Tiangen, China) according to the manufacturer's instructions, and its concentration and quality were assessed using a Nanodrop One Microvolume UV Spectrophotometer (Thermofisher, U.S.).

Processing of 16S rRNA gene sequencing data
Bioinformatic analysis of the 16S rRNA gene sequencing data was performed using the QIIME2 (version 2018.6.0) analysis pipeline [58]. Briefly, sequencing data were processed by the dada2 program to filter low-quality and chimeric sequences, and generate unique feature tables equivalent to OTU tables at exact match or 100% sequence similarity. Taxonomy was then assigned to these features using the q2-feature-classifier against the full-length SILVA database (release r132) at 99% similarity cutoff [59]. Analysis of microbiota diversities were conducted in QIIME2: alpha diversity metrics including Pielou's evenness, phylogenetic diversity, observed OTUs, Shannon and Simpson's indices, and beta diversity including weighted/unweighted UniFrac distances, and Bray-Curtis dissimilarity. Comparison of beta diversity was performed using the nonparametric method PERMANOVA. Abundance of OUTs were compared among groups by using STAMP [59]. the Linear discriminant analysis (LDA) Effect Size (LEfSe) algorithm was used with a log (LDA) score cutoff of 2 to identify taxa specifically enriched in particular age groups [60]. Phylogenetic cladograms of LEfSe results were visualized using the GraPhlAn tool (https://bitbucket.org/nsegata/graphlan).

Microbial interactive network construction and analysis
The SparCC (https://bitbucket.org/yonatanf/sparcc) algorithm was used to estimate the correlations among gut microbes [61]. 100 bootstrap replicates were used to calculate the pseudo P-values in the SparCC analysis, and correlations with | correlation coefficient (r) | >0.2 and P < 0.01 were considered significant. For each OTU with significant SparCC correlation, a weighted node connectivity score was calculated as an indicator of its weight in the network, by summing up its | r | with all of its   Firmicutes to Bacteroidetes ratio and beta diversity in gut microbiota in different age groups

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.