Père David’s deer gut microbiome changes across captive and translocated populations: Implications for conservation

Abstract The gut microbial composition and function are shaped by different factors (e.g., host diet and phylogeny). Gut microbes play an important role in host nutrition and development. The gut microbiome may be used to evaluate the host potential environmental adaptation. In this study, we focused on the coevolution of the gut microbiome of captive and translocated Père David's deer populations (Elaphurus davidianus; Chinese: Père David's deer). To address this, we used several different macro‐ and micro‐ecological approaches (landscape ecology, nutritional methods, microscopy, isotopic analysis, and metagenomics). In this long‐term study (2011–2014), we observed some dissimilarities in gut microbiome community and function between the captive and wild/translocated Dafeng Père David's deer populations. These differences might link microbiome composition with deer diet within a given season. The proportion of genes coding for putative enzymes (endoglucanase, beta‐glucosidase, and cellulose 1,4‐beta‐cellobiosidase) involved in cellulose digestion in the gut microbiome of the captive populations was higher than that of the translocated population, possibly because of the high proportion of cellulose, hemicellulose, and lignin in the plants most consumed by the captive populations. However, the two enzymes (natA and natB) involved in sodium transport system were enriched in the gut microbiome in translocated population, possibly because of their high salt diet (e.g., Spartina alterniflora). Thus, our results suggested that Père David's deer gut microorganisms potentially coevolved with host diet, and reflected the local adaptation of translocated population in the new environment (e.g., new dietary plants: Spartina alterniflora). A current problem for Père David's deer conservation is the saturation of captive populations. Given that the putative evolutionary adaptation of Père David's deer gut microbiome and its possible applications in conservation, the large area of wetlands along the Yellow Sea dominated by S. alterniflora might be the major translocation region in the future.

Chinese Père David's deer live in three core areas (Ding, Ren, Wen, Li, & Chang, 2014). Core Areas I (DFI) and II (DFII) are captive environments ( Figure 1b): Deer in these areas are fenced in and eat naturally occurring plants. However, due to overgrazing and the periods when grass is withered (November to April), the naturally occurring plants are insufficient, and the diets of the captive deer are supplemented with human-provided grains: wheat bran, barley, corn, soybean, and soybean straw fibers. The most common food plants in DFI and DFII are Pennisetum alopecuroides (PAL), Imperata cylindrica var. major (ICY), and Phragmites australis (PAU). DFIII is a wild habitat; 53 deer have been translocated into this area since 1998 (Figure 1b; Ding, Zhu, & Ren, 2006). At present, 215 Père David's deer inhabit DFIII (Ding et al., 2014). The most abundant potential food plants in DFIII are high salty Spartina alterniflora (SAL), PAU, Suaeda glauca (SGL), PAL, and ICY ( Figure 1c). The concentration of salty of the Père David's deer dietary plants in DFIII is significantly high than that of dietary plants in DFI and DFII (Zhu, Deng, et al., 2018).
American Bison (Bison bison) adjust their diet continuously over the growing season, possibly dictated by the seasonal availability of high-protein plant species (Bergmann, Craine, Robeson, & II, and Noah Fierer., 2015). The diet of bison during summer and fall has higher caloric and protein (Craine, Towne, Tolleson, & Nippert, 2013).
The seasonal shift in diet is associated with a significant increasing in the abundance of Tenericutes from spring to summer in the bison gut microbiome (Bergmann et al., 2015). Some Tenericutes specifically ferment simple sugars (Manurung, Boye, & Mølbak, 2012).
Differences in gut microbial communities have been identified between the Beijing and Shishou Père David's deer populations using 16S RNA gene sequences; these differences be associated with the differing abundances of the available plant species (Meishan et al., 2018). Père David's deer diets require further detailed investigation to evaluate the possible effects of diet on gut microbe composition (Meishan et al., 2018). Further knowledge of gut microbiome function will increase our understanding of the relationship between diet and gut microbiome.
In this study, we therefore aimed to determine how Père David's deer gut microorganisms are influenced by different food sources in both natural and captive environments at a fine scale. We used different approaches (landscape ecology, nutritional analysis, microscopy, isotopic analysis, and metagenomics) to characterize the regional dynamics of gut microorganism composition and function between translocated and captive populations (two captive, DFI and DFII; one wild, DFIII) with respect to different habitat. We then aimed to assess the relevance of our results for Père David's deer conservation in the future.

| Field observations
Although the feeding behaviors of captive Père David's deer can be observed in close proximity, those of wild/translocated Père David's After 2 min, several epidermal fragments were placed on the slide in a drop of distilled water. The water was then absorbed using filter paper. Small amounts of glycerin were added until the fragment samples were fully covered, and the mixture was stirred, covered with a coverslip, and sealed with neutral gum (Wang & Wang, 2011

| Sample preparation
We used a Delta Plus Advantage Isotope Ratio Mass Spectrometer at the Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences (Nanjing China), for isotopic analyses. Standard reference materials were used: carbon from the Peedee limestone (PDB) and nitrogen gas in the atmosphere (Peterson & Fry, 1987).
Fecal samples were oven-dried at 60°C for 48 hr. After powdering, the material was weighed in a tinfoil capsule using an electronic balance (accurate to 0.000001 g). We weighed 0.1 mg and 2.0 mg of powder to measure the carbon and nitrogen isotopes, respectively.

| Statistical analyses
We expressed the 13 C/ 12 C and 15 N/ 14 N ratios in delta (δ 13 C, δ 15 N) notation in parts per thousand (‰) relative to the PDB standard as follows: We used the Shapiro-Wilk test to determine whether our data were normally distributed (Park, 2008)

| Nutritional analysis of the staple foods of Père
David's deer

| Sample collection
The sample collection method used here was identical to that of the isotope analysis.

| Sample preparation and analysis
Fresh plants were weighed and then oven-dried at 80°C for 72 hr. After recording the dry weight, plant materials were ground over an 80-mesh screen and subsequently placed into a valve bag in an envelope. We were measured with the anthrone colorimetric method; and salinity was measured using conductivity (similar to the test for soil salinity).
Because of the small sample sizes, data normality was determined using the Shapiro-Wilk test (Park, 2008). Homogeneity of variance was measured using Levene's test. Nonparametric statistical methods were used when the raw and converted data were not parametric. Differences in nutritional content between two independent samples were identified using the Mann-Whitney U test. In this way, differences between the staple foods of the deer from different regions were assessed. Data were analyzed using Microsoft Excel 2010 and IBM SPSS Statistics 20.0. We considered p < 0.05 statistically significant.

| Data analysis for the central hypothesis
We treated the two sampling seasons (summer and winter) as separate replicates to evaluate the relationship between gut microbiome composition (alpha and beta diversity) and diet within each sampling season. The alpha diversity (i.e., Shannon index and phylogenetic diversity) for each fecal sample was calculated with QIIME (Caporaso et al., 2010). We identified significant differences in the abundance of bacterial taxa among the three core areas using the linear discriminant analysis effect size method (Lefse; Segata et al., 2011). To identify dissimilarities in community composition, we performed a principal coordinates analysis (PCoA) analysis in QIIME (Caporaso et al., 2010). Supervised learning analyses (random forests) were performed in QIIME to determine whether sampling season or core area

| DFIII habitat analysis
Habitat was evaluated based on the distribution of various plant taxa, as determined from satellite images taken in 2013. These images covered the entire Père David's deer distribution range in DFIII.
Using the maximum-likelihood classification algorithm in supervised classification, forest cover was identified using ERDAS IMAGE 8.7  Figure 1c).

| Stable isotopic differences between the feces of captive and translocated populations within the same sampling season
The stable isotope analysis also indicated that diet changed with season (especially in DFI and DFII) and that diets differed between the captive (DFI and DFII) and wild (DFIII) core areas (Figure 2).
In DFI and DFII, fecal δ 13 C values in the summer were similar to those of C4 plants (Figure 2a,b). This is probably because the pre-   Table S1), but we found the significance over summer and winter using all the samples from DFIII collecting from 2011 to 2014. These reflected the partial or shallow divergence in seasonal diets among the deer from DFIII. We observed the variation of δ 13 C between captive and translocated populations during the same season. In both summer and winter, the δ 13 C values of DFI and DFII were significantly different from those of DFIII, reflecting the differences in diet between the two regions (Supporting information Figure S1).

| Nutritional analysis of plant food components
The

Relative abundance (%)
Relative abundance (%) the random forests classifier was 58 when separating DFI + DFII from DFIII. This ratio was 3.72 for differentiating DFI + DFIII from DFII and 2.22 for differentiating DFII + DFIII from DFI. This finding also indicated that the gut microbial communities of DFI and DFII were more similar than that of DFIII. One-way PERMANOVA showed a significant difference in microbial composition among these groups (Table 2). For example, in the winter season, the pairwise comparisons detected the significant difference between translocated population (DFIII) and captive population (DFI or DFII), and no significant difference existed between DFI and DFII (Table 2). These results further confirmed the previous findings by random forest tests.

| Changes in gut microbial function associated with their diet between captive and translocated populations
Most of the primary functions identified for the gut microbes of deer from the three core areas were similar, including "carbohydrate metabolism" and "amino acid metabolism." Considering the significant difference on the dietary nutrition between captive and translocated populations, we focus the gut microbial genes coding some putative enzymes involved in two primary functions (sodium transportation and cellulose digestion). The genes coding for two enzymes (sodium transport system ATP-binding protein (natA) and sodium transport system permease protein (natB)) involved in the sodium transport system were enriched in the translocated Père David's deer fecal metagenomes (one-way ANOVA, post hoc LSD test at 0.05; Figure 5a,b). Taxonomic assignment of the genes coding for natA indicated that most of them came from Firmicutes (~96%) and Proteobacteria (~4%) (Figure 5c). In genus level, most of these predicted genes came from Roseburia (~47%) and Clostridium (~30%), and others belonged to Bacillus (~4%), Lachnospiraceae_norank (~4%), and Luteimonas (~4%). Taxonomic assignment of the genes coding for natB revealed that most of them came from Firmicutes (~84%), Tenericutes (~10%), and Proteobacteria (~6%) (Figure 5d). In genus level, most of these genes came from Roseburia (~64%), Eubacterium (~7%), Haloplasma (~4%), and Luteimonas (6%). Thus, most of these two enzymes might come from Roseburia. We then investigated the relative abundance of this genus using our 16S dataset and found the relative abundance of this genus in the translocated population feces was highest ( Figure S6; one-way ANOVA test, p < 0.01).
The proportion of putative enzymes (endoglucanase, beta-glucosidase, and cellulose 1,4-beta-cellobiosidase) involved in cellulose digestion in the fecal metagenomes of the captive populations was higher than that in the fecal metagenomes of the translocated population ( Figure 6a). Taxonomic classifications of these genes revealed that most of them possibly come from three phyla (Firmicutes, Bacteroidetes, and Euryarchaeota). For examples, in these main genera having these putative enzymes, Ruminococcus (Firmicutes) and Clostridium (Firmicutes) were the common sources for these three putative enzymes (Figure 6b,d), and the relative abundance (16S) of Ruminococcus was higher in captive population feces than that of translocated population feces (Figure 7a,b). The fecal gut microbiome (16S) of captive populations also has a higher proportion of Methanocorpusculum (Euryarchaeota) compared to that in the feces of the translocated population (Figure 7c). The feces of the translocated population had a high percentage of Bacteroides (Figure 7a,d).
However, many other particular gut microbial genera (Ruminococcus and Methanocorpusculum) had these putative cellulose-digestion enzymes, which might explain the relatively low proportion of these F I G U R E 5 The potential adaptation on high-salt diet by gut microbial from Père David's deer metagenomes. (a, b) The proportion of genes coding for these putative enzymes (natA and natB) related to the potential sodium transport system in Père David's deer gut microbiomes. The taxonomic assignment of the identified genes coding for natA (c) and natB (d)

| D ISCUSS I ON
In this long-term study (2011)(2012)(2013)(2014) of the Dafeng Père David's deer captive and translocated Père David's deer populations, we observed some dissimilarities in Père David's deer gut microbiome community and function that might link microbiome composition to diet. Gut microbial communities are affected by many factors, primarily host phylogeny and diet (Ley et al., 2008). Here, the translocated deer derive from (and are genetically similar to) the captive populations.
These populations have similar gut microbiome.
The proportion of cellulolytic gut microbiome (e.g., Ruminococcus) and their genes coding for putative enzyme (endoglucanase, betaglucosidase, and cellulose 1,4-beta-cellobiosidase) involved in cellulose degradation in the captive populations was relatively higher than that of the translocated population. Indeed, the proportions of cellulose, hemicellulose, and lignin in the plants most consumed by the captive populations were higher than these proportions in the plants most consumed by the translocated population. Moreover, Christensenellaceae abundance was significantly higher in the captive populations than in the translocated population. Christensenellaceae has been isolated from the feces of many mammals, including humans and ruminants (Lima et al., 2015;Morotomi, Nagai, & Watanabe, 2012). Our functional analysis indicated that the Christensenellaceae in the Père David's deer fecal samples were mainly for "carbohydrate metabolism" and "energy metabolism" ( Figure S7).
Interestingly, the abundance of Methanocorpusculaceae in the captive populations (~0.22%) was greater than in the translocated populations (~0.06%), especially in the winter. For example, about 11 percent in these genes coding for putative endoglucanase come from Methanocorpusculum. In the winter, the human-provided forage most commonly straw (high fiber) and bran (Keqing, 2005). The dominant Firmicutes in the microbiota of cattle forestomachs (e.g., Ruminococcaceae, Rikenellaceae, and Christensenellaceae) may play essential roles in the degradation of starch and fiber (Mao, Zhang, Liu, & Zhu, 2015). Thus, the significant higher proportion of cellulosedigestion gut microbiome (e.g., Euryarchaeota_Methanocorpusculum, Firmicutes_Ruminococcus, and Firmicutes_Christensenella) in captive F I G U R E 6 The potential for cellulose degradation by gut microbial from Père David's deer metagenomes. (a) The proportion of genes coding for these putative vital enzymes (endoglucanase, beta-glucosidase, and cellulose 1,4-beta-cellobiosidase) related to the potential cellulose digestion in Père David's deer gut microbiomes. The taxonomic assignment of the identified genes coding for endoglucanase (b), beta-glucosidase (c), and cellulose 1,4-beta-cellobiosidase (d)   (Yamaguchi, Hamamoto, & Uozumi, 2013). In some bacteria, an Na+ circuit is an essential link between exergonic and endergonic membrane reactions (Dimroth, 1990). Here, the fecal metagenomes of the translocated population displayed the enrichment in genes coding for some putative enzymes (natA and natB) involved in sodium transport system, and most of them came from the two genera in Firmicutes, such as Roseburia and Clostridium. The relative abundance of Roseburia in the feces of the translocated population was higher than that in the feces of captive populations. Some Roseburia strains in the human gut can utilize dietary components and produce butyrate short-chain fatty acids (Duncan, Hold, Barcenilla, Stewart, & Flint, 2002), which will affect colonic motility, immunity maintenance, and anti-inflammatory properties (Tamanai-Shacoori et al., 2017).
Thus, we speculated that this finding on translocated Père David's deer gut microbiome might reflect some potential adaptation on host high-salt diet and even had some putative effect on host health. In addition, this might help Père David's deer adapt to the new environment.

| Père David's deer gut microbiome and its application in conservation management
Symbiotic gut microorganisms play an important role in host health and development (Muegge et al., 2011). In natural environments, dietary shifts are common in herbivore mammals and occur in response to food availability and plant nutritional value.
Our results suggested that the difference on the dietary plant nutrition lead to some dissimilar on their gut microbial composition and function. Père David's deer gut microorganisms potentially F I G U R E 7 The relative abundance (16S) of some genera related to potential cellulose and sodium metabolisms in Père David's deer's feces among three populations (two captive populations: DFI and DFII, one translocated population: DFIII). (a) Linear discriminant analysis (LDA) effect size (LEfSe) method identified significant variations in the compositional profile (16S) at genus level among these populations (threshold on the logarithmic LDA score: 2.5). The relative abundance of Ruminococcus (b), Methanocorpusculum (c), and Bacteroides (d)

DFI D FII D FIII
coevolved with host diet, allowing the gut microbiome to adapt to shifts in diet in different habitats. For example, the enrichment in genes coding for some putative enzymes (natA and natB) involved in sodium transport system might help Père David's deer survive in the high-salt diet. Currently, one issue faced by Père David's deer conservation efforts is the saturation of captive populations (Ding et al., 2014). Translocation is an effective strategy with which to address this problem. During initial Père David's deer translocation to DFIII in 1998, there was some concern about Père David's deer diet, as the high-salt plant SAL is widely distributed across this translocation site (Ding, 2009). SAL is an invasive halophyte plant that is widely distributed in the wetlands along the Yellow Sea (Ding, 2009). Père David's deer exhibit a special behavior when feeding on SAL: The deer repeatedly feed on the same plant over a long period, causing the plant to continue to grow fresh leaves (Ding, 2009). In the nearly 20 years since the initial translocation, the Père David's deer population at DFIII has grown well (Ding, 2017). Given that the evolutionary signature gut microbiome of the translocated populations, the large area of SAL-dominant wetlands along the Yellow Sea might be the main region for future Père David's deer translocations. Thus, here, we provide the importance of gut microbiome of the translocation population on the potential adaptation to the new environment (e.g., diet), which can be applied and relevant for the efficiency conservation management.

ACK N OWLED G EM ENTS
This work was supported by grants from the National Natural Science Fund (31222009, 31570489, and 31741112) and the Priority

Academic Program Development of Jiangsu Higher Education
Institutions. We thank LetPub (www.letpub.com) for its linguistic assistance during the preparation of this manuscript.

CO N FLI C T O F I NTE R E S T S
The authors declare that there is no conflict of interest.

AUTH O R CO NTR I B UTI O N S
LZ designed the study. LW, JD, ZY, R.Y, and YD performed the experiments. Q.D carried out the analysis on satellite maps. LZ and HC performed analyses. LZ wrote the manuscript with comments from all co-authors.

DATA ACCE SS I B I LIT Y
DNA sequences have been deposited in figshare (https://doi.