Seasonal Variation in Gut Microbiota Composition: Comparative Analysis of Siberian Musk Deer (Moschus moschiferus) and Forest Musk Deer (Moschus berezovskii)

Background: Musk deer is famous for its secretion of musk, but due to various factors that the musk deer population has dropped sharply and it is dicult to expand herd sizes. In light of the rapid development of agriculture and technology, as well as the urgent need for wildlife conservation, research on the gut microbiota of musk deer is warranted. The gut microbiota impacts host health is shaped by dietary, environmental, and host factors. Results: We studied the gut microbiota in Siberian musk and Forest musk deer using high-throughput sequencing of the V3–V4 region of the 16S rRNA gene. Fecal samples were collected during two seasons to determine associations between community diversity, composition, and seasonal factors. The results show that the Firmicutes and Bacteroidetes were the predominant phyla in both species, their relative abundance was signicantly different with species and seasonal factors. The Siberian musk deer population sampled in this study had a wider feeding range and no dietary limitations. Affected by the richness and type of food sources, the microbial α-diversity is higher in summer than winter. In the siberian musk deer have only Tenericutes signicant difference with the season change. Conclusions:The gut bacterial composition of these two musk deer is signicantly different, and the abundance of core bacteria is affected by seasonal factors. This study is the rst to analyze the composition of the gut microora in Siberian musk deer, and these results provide a foundation for improving feeding for musk deer populations and further studies investigating the environmental impact of musk deer.

Gut microbial composition and function in wild animals has been widely studied in recent years [7,8]. A number of vital host functions depend on the gut microbiota, which is involved in the synthesis of nutrients and systemic immune regulation [9]. The gut microbiota is a complex ecosystem that is regulated by many factors. The composition of gut microbiota is altered by the host, feed, lifestyle, antibiotics and seasonal changes [10][11][12][13]. Studies in horses [14], mice [15], pandas [16], humans [17], and forest musk deer [18] have shown seasonal shifts in gut microbiome structure. These alterations could be caused by the diet. In summer, the rainfalls increased, and the density of plants will be improved. It means that musk deer have su cient food. However, the food supply will be decreased in winter because of severe weather. Some gut bacteria are opportunistic pathogens, and bacterial overgrowth or an altered community structure are associated with multiple diseases [11,19]. Siberian musk deer and Forest musk deer have a high rate of disease incidence, especially diseases caused by a bacterial infection, and the mortality rate of diseased animals is high. These infections can be mediated by environmental or hostderived bacteria [20,21]. The precise role of the gut microbiota on animal health and disease remains unclear and more foundational work is required. The Siberian musk and forest musk deer are herbivorous ruminants. Although the microbiota can be an important tool for assessing animal health, no reports to date have studied these factors in Siberian musk deer.
We characterized the seasonal variation in the gastrointestinal microbiota of Siberian musk deer and forest musk deer using high-throughput sequencing of the 16S rRNA gene in fecal samples. According to the research results of Hu et al [18] that the in uence of seasonal factors on the gut microbiota of forest musk deer, we selected representative winter and summer seasons for this research.The main objectives of present study are: (1) to compare the composition and diversity gut microbiota of two species; (2) to determine seasonally (i.e., summer and winter) variations in the microbial community composition and disease of Siberian musk deer; (3) to examine the changes in signaling pathways and functional genes, and; (4) to discuss their potential functions in the diagnosing diseases of the digestive system and for improving feeding methods.

Results
Quality ltering and validation of the dataset There was an average of 909,993 reads (average length = 428.38 bps; Table S1) obtained from each sample after quality ltering. There were 1,627 OTUs detected, and each sample had between 410 and 846 OTUs (Table S2). Individually rarefaction curves showed that the sequencing depth was adequate for the study (Fig S1). Good's coverage showed that 28 samples had between 99.28 to 99.62% coverage (mean = 99.44%; Table S2). The results showed that the number of OTUs increased rapidly until it gradually reached a plateau. It shows that the amount of sequencing data is large enough, and the increase of sequencing data can not nd more OTU, which can re ect the information of most microbial species in the sample.
Core bacterial communities and composition difference in the two species There were 18 phyla, 25 classes, 59 orders, 109 families, and 269 genera identi ed in the two musk deer species. There were 658 OTUs that were present in all groups. There were 92, 56, 101, and 12 OTUs that appeared in each sample of the FS, FW, SS, and SW groups, respectively (Fig 1A; FMD in summer and winter: FS and FW; SMD in summer and winter: SS and SW ).
The classi cation of sequences from the samples resulted in six different phyla was identi ed, including Firmicutes, Bacteroidetes, Spirochaetes, Verrucomicrobia, Tenericutes, and Proteobacteria ( Fig 1B). We considered these taxa that were shared by all individuals in a group to be the core bacterial community for that group. Firmicutes (FMD: 52±8.427 to 64.03±11.69, SMD: 84.57±6.305 to 89.86±3.094) and Bacteroidetes (FMD: 38.17±7.489 to 29.09±8.978, SMD:12.78±6.113 to 8.89±2.302) were core and dominant community members for both musk deer species. The relative abundances of Firmicutes were signi cantly higher in SMD than in FMD, and this corresponded to decreases in the relative abundances of Bacteroidetes (Fig 1B Table 2). Spirochaetes and Verrucomicrobia were only found in forest musk deer ( Fig 1B). Also, the difference in gut microbiota composition between the two musk deer can be clearly observed according to the Heatmap . Our results indicated the existence of two core taxa at the class level (Clostidiales and Bacteroidales) (Fig S2). At the order level, core gut bacterial included Ruminococcaceae, Prevotellaceae, Bacteroidaceae, Lachnospiraceae, Rikenellaceae, and Christensenellaceae (Fig S3), in addition to this six core taxa at the family level( Fig S4). These results indicated that the composition of the microbial community may be related to different physiological characteristics in the gut of musk deer.
The alpha diversity of forest musk deer was higher than that of Siberian musk deer based on the Shannon and Chao indices (FMD: Shannon index=4.662±0.42366, Chao=819.29±138.02, p=0.4306; SMD: Shannon=4.5207±0.55078, Chao=761.44±137.5, p=0.4014; Table S3; Fig 2A-2B ), it is concluded that the community abundance and community diversity of the forest musk deer are higher than the siberian musk deer, but there is no signi cant difference. There were differences in the beta diversity of the SMD and FMD communities PCoA analysis (Fig 3). ANOSIM analysis reveals intergroup differences in bacterial communities between two musk deer species (R = 0.6643, p = 0.001). The results of ANOSIM analysis and PCoA analysis mutually veri ed the differences of bacterial communities between the two musk deer species.

Seasonal differences in bacterial communities
The alpha diversity of the bacterial community in the summer was higher than in the winter for the two musk deer species(Shannon index; FS  Table S3). The Chao index have signi cant different in SMD (Chao:p = 0.03038, Q=0.06077; Fig 2D). At the phyla level, there were 8 taxa (Firmicutes, Bacteroidetes, Actinobacteria, Cyanobacteria, Patescibacteria, Liritimatiellaeota, Lentisphaerae, and Elusimicrobia )whose relative abundance was signi cantly different between seasons in FMD; at the same time, and only one taxon (Tenericutes) whose relative abundance was signi cantly different between seasons in SMD ( Table 2; Fig 4A; Fig S5 ). It is worth mentioning that the Firmicutes/Bacteroidetes ratio changed with the season in two musk deer.
We are much interested in musk deer gut microbiota is to determine whether speci c microbiota could be the change with different season factors. To identify speci c microbial candidates present in two specise seasons, LEFse (LDA score ≥ 4) analysis were used to determine speci c biomarker belonged to different taxonomy ( Fig 4B) from phylum to genus. The result revealed that twelve families (Bacteroidia, Bacteroidales, Bacteroidetes, Prevotellaceae, Prevotellaceae_UCG_004, Bacteroides, Bacteroidaceae, Verrucomicrobia, Verrucomicrobiae, Verrucomicrobiales, Akkermansia, and Akkermansiaceae) were signi cantly enriched in the FS group, one family (Coprococcus) was signi cantly enriched in the FW group, two families (Ruminococcaceae_UCG_010 and Ruminococcaceae_UCG_013) were signi cantly enriched in the SS group, and seven families (Firmicutes, Clostridia, Clostridiales, Lachnospiracea, Christensenellaceae_R_7_group, Christensenellaceae, and unclassi ed__Lachnospiraceae) were signi cantly enriched in the SW group ( Fig 4B).

Predicted functions of the microbial communities
There were signi cant differences in the functional capacity of SMD and FMD microbial communities. Speci cally, three KEGG pathways (Environmental Information Processing, Cellular Processes, and None) were enriched in SMD samples, and ve KEGG pathways (Human Diseases, Organismal Systems, Metabolism, Genetic Information Processing, and Unclassi ed) were enriched in FMD samples at Level 1 ( Fig 5A). Also, two groups of samples were signi cant differences in thirty-three KEGG pathways (Level 2; p < 0.005 ). Among them, seven KEGG pathways (Transcription, Membrane Transport, Cell Motility, Signal Transduction, Cellular Processes and Signaling, Environmental Adaptation, and Carbohydrate Metabolism) were signi cantly enriched in SMD samples. Twenty-six KEGG pathways (Translation, Replication and Repair, Nucleotide Metabolism, "Folding, Sorting and Degradation", Enzyme Families, and Metabolic Diseases, et al) were signi cantly enriched in FMD samples (Fig 5B).At level 3, a functional comparison of 143 extremely signi cant differences (p < 0.001) was performed. Among them, the SMD group enriches thirty-ve functions, and the FMD group enriches other functions. The results also re ect differences in transcription and metabolism between the two groups of animals ( Fig S6).
There was a signi cant difference in predicted microbiota function in SMD between summer and winter. The relative abundance of genes (Level 1) involved in Metabolism, Organismal Systems, and Human Diseases was signi cantly different in two seasons ( Fig 5C). The relative abundance of gene families(Level 2) involved in Xenobiotics Biodegradation and Metabolism, Endocrine System, and Metabolism of Cofactors and Vitamins was signi cantly higher in the SS samples than in the SW samples ( Fig 5C). Besides, the In uenza A, Toxoplasmosis, Colorectal cancer, Viral myocarditis, and Smell cell lung cancer functions (Level 3) were signi cantly enriched in the SS group ( Fig 5C).

Discussion
A high-ber diet coupled with long-term evolutionary adaptation has resulted in a large number of symbiotic bacteria in the gastrointestinal tract of herbivores to help the body digest and absorb nutrients.
At the same time, it also gives the ruminant gastrointestinal ora a distinct feature and a similar ora structure. Our data shows the difference in microbial communities between musk deer species as well as the seasonal variations in the microbial communities of each deer species. The α-diversity of the gut microbiota in FMD samples was signi cantly different in two seasons, and the diversity lower in winter than in summer (Shannon index: p<0.05; Chao index: p<0.05); meanwhile, the diversity of the gut microbiota in SMD samples was not signi cantly different between seasons. Gut microbial diversity in mice [22] and humans [23] is signi cantly lower in diseased subjects. It shows that the diversity of gut microbiota is related to animal health. Also, many studies have shown that modern lifestyles have a negative impact on gut microbiota diversity that the simpli cation of diet structure, the use of antibiotics and the reduction of outdoor activities are some of the factors that cause the diversity of gut microbiota to decrease [17,24,25]. In winter, survival strategies are most important for wild populations with low energy intake and high energy consumption. In terms of food types, there are fewer types of food to choose from in winter, and it is more di cult to eat the herbs on the surface, which results in a single diet for Siberian musk deer. It is speculated that the less variety of food is a factor that causes the lower gut microbiota diversity of Siberian musk deer.
The core bacterial phyla in both musk deer species were Firmicutes and Bacteroidetes, which is consistent with previous studies of the microbiota of musk deer [18,26,27] and other ruminants [28][29][30]. The phyla are found in the guts of many mammals, which suggest the importance of these bacteria in the gut ecosystem [31]. Firmicutes are predominant cellulolytic bacteria that able to degrade ber and form short-chain fatty acids (SCFAs), which are utilized by the host [32][33][34]. Ruminants cannot produce the necessary cellulolytic enzymes to degrade ber and must rely on bacteria to produce these enzymes [35]. Certain Bacteroidetes species are specialists that degrade high molecular weight organic material (i.e., carbohydrates and proteins), and the genomes of these species are known to contain large numbers of CAZymes [36]. The seasonal variations in microbiome composition in our study are likely due to seasonal changes in diets. Diet is an important factor that shapes gut microbial communities [37][38][39]. Speci cally, the ratio of ber to carbohydrates in diets can impact microbiome composition in ruminants [40]. For example, the content of animal protein and saturated fatty acids are related to the abundance of Actinomycetes, while the abundance of carbohydrates, monosaccharides (glucose and fructose) and the abundance of Pseudomonas and Prevotella are highly correlated. Bacteroidetes were more abundant in musk deer consuming a high-starch and high-protein diet during the summer; meanwhile, Firmicutes were more abundant in musk deer consuming a high-ber diet during the winter [32,41,42]. Except for various leaves, the food of captive forest musk is mainly carrots and various grains. Compared with the Siberian musk deer, forest musk deer food contains more protein and fat, coupled with a lot of energy to avoid foraging and avoiding natural enemies, it is not di cult to understand that Bacteroides is higher in the forest musk deer.
In Siberian musk deer, the Firmicutes/Bacteroidetes ratio (F/B ratio: 6.6173-10.113) was signi cantly higher than that of forest musk (F/B ratio: 1.3843-2.2011). This may be related to the feeding habits of these animals, and these seasonal changes could increase energy e ciency during the winter [10,43].
Compared to forest musk deer, the Siberian musk deer has a broader diet, a relatively high living altitude, and a colder climate that requires more energy [2,44]. Recent research reports [45,46] have con rmed that the ratio of Firmicutes and Bacteroidetes is related to fat deposition. Although these bacteria can promote fat deposition, when the abundance of Firmicutes increases, it can promote fat. The effect of deposition is more pronounced. Other studies [47] have shown that the genes of intestinal microorganisms with high F/B ratios are rich in encoding lysing polysaccharide start enzymes and eight glycosidic bond hydrolase families involved in sugar metabolism. And the residues of nutrients in mice with a high F/B ratio were signi cantly lower than those in lean mice. These results indicate that intestinal microorganisms with a high F / B ratio have higher fermentation e ciency and obtain more energy from food, thereby promoting fat deposition. Therefore, it is believed that Firmicutes can help the body to absorb more energy from food. In our experiments involving siberian musk deer, life in the wild, winter cold and the shortage of food, a lot of energy on the one hand to nd food, avoid predators, on the one hand, but also consume energy to maintain body temperature. So the Firmicutes in their gut microbiota can help them draw energy from food as much as possible to maintain the body's needs. Spirochaetae were only found in the FMD group. This may be related to feeding conditions and animal habits. In this study, Siberian musk deer lived in a wide range and had a primitive diet, while forest musk deer were raised arti cially in captivity and fed by breeders. Similar phenomena have been found in previous studies on sika deer [30] and human living environment [48] --different living environment makes spirochaetae only appear in intestinal ora of a certain living environment. In addition, studies have shown that animals fed diets rich in fungi, vitamin C and dietary ber have lower levels of spirochaetae in gut microbiota [49], which is in line with our experimental animal Siberian musk deer similar, but need further validation.
The PCoA analyses were also showing the distinct separations and reciprocal relationships among musk deer from different areas, it indicated that the composition and structure of the bacterial community were signi cantly different between forest and siberian musk deer. The season factors would also not be negligible effect factors on musk deer gut microbiota. For forest musk deer, have 8 phyla bacteria were signi cantly different, and this result is similar to that of Hu et al [18,26,27]. However, the SMD group was less affected by seasons, and only one phylum bacteria (Tenericutes) signi cant difference. Phylum Tenericutes, class Mollicutes were rarely found in the gastrointestinal tract before and then were identi ed for the rst time in wild chimpanzees, and Tenericutes also were discovered in herbivorous mammals [50,51]. Consistent with the research results of gut microbiota of wild sika deer [30], which was the relative abundance of Tenericutes higher than Proteobacteria. Hence, Tenericutes should be valued for their special gut microbiota function in the musk deer. As we all know, the gut microbiota can affect the health and metabolism of the host. Among them, the large number of microorganisms and the complexity of the genomes associated with the microbial community limit the study of the metabolic pro le of the bacterial community [23,52,53]. PICRUSt analysis enables robust assessments regarding the impact of environmental factors and biological processes on the composition of microbial communities. Microbial functional analysis showed that the most abundant biological pathways in all samples were translation, amino acid metabolism, carbohydrate metabolism, replication and repair, and membrane transport; this agrees with previous studies [54][55][56]. In the SMD group, some pathways differed between the two seasons. Especially in the disease, the results show a variety of diseases was signi cantly enriched in the summer sample, the results of this follow-up study provides a guideline. PICRUSt analysis does have limitations, as other pathways, such as some disease, were identi ed as signi cant but are not relevant to microbial communities of musk deer. Transcriptomic and metabolomic analyses would be more applicable to determining bacterial functions. This is the rst characterization of the gut microbiota in SMD, and it is also the rst time to compare the structure of the gut microbiota in FMD and SMD. We show that that musk deer gut microbiota is complex and taxonomically plastic, and it varies by species of musk deer. Microbiota composition changes seasonally, likely due to changes in environmental conditions and diet. Limitations of this study include the sample size collected was relatively small, the inability to control for potentially important factors, such as weather and the lack of species-level resolution via 16S rRNA tagged sequencing. Shotgun metagenomic sequencing would provide additional insight into the gut microbial community in musk deer. In light of the rapid development of agriculture and technology, as well as the urgent need for wildlife conservation, further research on the gut microbiota of musk deer is warranted.

Conclusions
We characterized the seasonal variation in the gastrointestinal microbiota of Siberian musk deer and forest musk deer using high-throughput sequencing of the 16S rRNA gene in fecal samples. The results show that the composition and diversity gut microbiota of these two musk deer has signi cant differences. The Firmicutes and Bacteroidetes were the predominant phyla in both species, their relative abundance was signi cantly different with species and seasonal factors. In addition, the microbial αdiversity is higher in summer than winter, and determine seasonal variations in the microbial community composition of Siberian musk deer. This study is the rst to analyze the composition of the gut microbiota in Siberian musk deer, and these results provide a foundation for improving feeding for musk deer populations and further studies investigating the environmental impact of musk deer.

Animals and Sample Collection
Four healthy adult Siberian musk deer were sampled from the musk-deer breeding center in Mongolia (Gachuurt Village, Ulaanbaatar; 47°55'N 107°09'E). The animals used in this experiment were captured from the wild status by the staff of the Mongolian musk-deer breeding center, which is the partner of the project, and kept in pens within a certain range of breeding. The wild habit of the Siberian musk deer is protected to a certain extent, through a wider range of captive breeding (each musk deer is raised in pen of 20 m×40 m) and avoiding the disturbance of the breeders as much as possible (Fig S7). The altitude of the breeding center is about 1783 m, and the annual average precipitation is 200-400 mm. The average summer temperature is 20-30 ℃, which is hot and cool. The average temperature in winter is about -30 ℃, and the lowest temperature is -40 ℃. The siberian musk deer food partly from the leaves of plants grown in the fence, another part of the feed is mainly supplied by keepers such as Cetrarian lichens and mushrooms (Fig S7). They spend 4-5 hours and about 9 hours in summer days and winter days respectively, to feed with a repetition of 5-8 times a day during the dusk and dawn. Stomach lled with food weighs 400-600 gr, on average it is 520 gr. In July, their diet is 80% fresh green plants and 20% moss and lichens. In December, their diet is 90% Cetrarian lichens (e.g., Cetrariahivaeis, Cetrariaislanica, Cetrariacucuetata) and 10% green moss. They also consume pine nuts, barberry leaves, pine, and spruce ( Table 1; Table S4). During the summer Siberian musk deer eats 2 kg of a green plant, 0.6 kg moss lichens or moss, in winter 3.7 kg lichens, 0.4 kg moss per day. According to the research Musk deer drinks 100 ml of water for once, 3-4 times a day.
Ten forest musk deer were sampled from the Inner Mongolia Ulantasb League in China (40°42'N 113°26'E). The average altitude is 1400 meters and the average rainfall is 400 mm. The hottest month is July, the average temperature is 20.4 °C, the highest temperature is 36.5 °C, the average temperature in winter is -13.5 °C, and the lowest temperature is -37.5°C.Forest musk deer diet consisted mainly of leaves (e.g., mulberry leaves (Morus alba), elm leaves (Ulmus pumila), broussonetia papyrifera leaves) and seasonal vegetables (e.g., pumpkin, carrots, and corn). They were fed with fresh leaves in the summer and dried leaves in the winter. Under captive conditions, due to changes in environmental conditions, the type of feeding is greatly restricted, so a proper supply of local vegetables and minerals is essential. Each kilogram of leaves contains 250 mg of carotene, 1200 mg of calcium, 320 mg of iron, and also contains copper, zinc, phosphorus, manganese and other trace elements and vitamins. The content of crude protein in the dry matter of the foliage feed is 17% to 22%. The content was up to 36% in June, and gradually decreased after July, and fell to 12% in November.
Ten fecal samples were collected from forest musk deer in both summer (July, FS1-FS10) and winter (December, FW1-FW10). 4 fecal samples were collected from Siberian musk deer in both summer (July, SS6,SS8-SS10) and winter (December, SW1-SW4). Musk deer usually defecate in the early morning (around 5-6 o 'clock) with xed defecation sites. Fresh fecal samples were collected with a sterile disposable glove to avoid human contamination. Samples were stored in sterile centrifuge tubes, sealed, and placed on liquid nitrogen for transportation to the laboratory. Samples were then stored at -80°C until DNA extraction. And we got a permit of CITES management authority of Mongolia in 2018.

DNA Extraction
Bacterial DNA was extracted using the QIAamp® DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Samples were crushed on dry ice before extraction. The DNA concentration was measured using a NanoDrop 2000 Spectrophotometer and con rmed by 1.0% agarose gel electrophoresis. The concentration and purity of DNA were tested using the Qubit dsDNA HS Assay Kit (Life Technologies, Carlsbad, CA, USA). Extracted DNA was stored at -80 • C until PCR.

Bioinformatics and Statistical Analyses
Sequence assembly and quality ltering were performed using FLASH (version 1.2.11; https://ccb.jhu.edu/software/FLASH/index.shtml) and QIIME (version 1.9.1; http://qiime.org/install/index.html), respectively. The quality control method was: 1) lter bases with a tail mass value < 20 and set a 50 bp window. If the average mass value in the window was lower than 20, the back base was truncated from the window. Reads less than 50 bp were removed; 2) Based on the overlap between the PE reads, the reads were merged with a minimum overlap length of 10 bp; 3) The maximum mismatch ratio allowed by the overlap area of the splicing sequence was 0.2; 4) Sequences were assigned to samples using barcoded primers. Operational taxonomic units (OTUs) were picked using Usearch (version 7.0.1090, http://drive5.com/uparse/).
Taxonomy was assigned to the OTUs using the Ribosomal Database Project (RDP) classi er (version 2.11, http://sourceforge.net/projects/rdp-classi er/) against the Silva 132 (Release132 http://www.arbsilva.de) database with a con dence threshold of 0.97. Alpha diversity indices, including Chao, ACE, Shannon, Simpson, and Good's Coverage, as well as beta diversity analysis, were performed in Mothur (version v.1.30.2, http://www.mothur.org/wiki/Schloss_SOP#Alpha_diversity) and visualized using R (version 3.5.0). Wilcoxon rank-sum tests were used to determine statistically signi cant differences in alpha diversity and taxa relative abundance. Venn diagrams and pie charts were used to visualize shared OTUs, which were de ned as the core microbiome, using the Venn Diagram package in R. Heatmaps were generated using Bray-Curtis dissimilarity with average linkage hierarchical clustering using the vegan package in R. Principal co-ordinates analysis (PCoA) ordination based on Bray-Curtis dissimilarities and an analysis of similarity (ANOSIM) test were used to determine the composition differences and the signi cance of community compositional differences among groups. Perform PCoA Statistical analysis and plotting used the vegan package in R. Differences between groups were determined using Kruskal-Wallis H-tests using the stats package in R and the SciPy package in Python (version 2.6.4). Linear discriminant analysis effect size (LEfse) analysis was conducted to indicate the signi cant factors using the online Galaxy work ow framework (http://huttenhower.sph.harvard.edu/galaxy/) and the cutoff value for the linear discriminant analysis score was set to the default level of 4.0 [58]. PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States, http://picrust.github.io/picrust/index.html) was used to predict the functional gene content of microorganisms (Langille MG et al., 2013) [59]. The predicted function is precalculated for genes in the Kyoto Encyclopedia of Genes and Genomes (KEGG, http://www.genome.jp/kegg/) database were used to predict gene function. STAMP analysis was used to assess functional richness between groups. Welch's t-test was used for STAMP (version v2.1.3) software for differential analysis. Differences were considered as statistically signi cant when P < 0.05.

Consent for publication
Not applicable.

Availability of data and materials
The raw sequences obtained in this study can be found in the NCBI Sequence Read Archive with the accession number: SRP198443.

Competing interests
All authors and funding bodies declare that they have no competing interests.   Figure 1 OTUs and bacterial taxa shared between the four groups. Venn diagrams show the number of OTUs shared by forest musk and Siberian musk deer sampled during two seasons (A). Bar charts showing the core bacterial taxa at the phylum level for each group: (B) . All taxa with relative abundances less than 1% were grouped as "Other taxa".FW: forest musk deer in the winter; FS: forest musk deer in the summer; SS: Siberian musk deer in the summer; SW: Siberian musk deer in the winter.  (0.01 < p ≤0.05 marked as *, 0.001 < p ≤ 0.01 marked as **, p ≤ 0.001 marked as***). LDA histogram (B) showing the differences in taxa relative abundance. The length of the bar column represents the LDA score (LDA>4). A larger LDA score indicates a higher contribution of that species to the overall differences in community structure. FW: forest musk deer in the winter; FS: forest musk deer in the summer; SS: Siberian musk deer in the summer; SW: Siberian musk deer in the winter