Microbiome differences in wild and captive black rhinoceros 1 Gut microbiome differences between wild and captive black rhinoceros – 1 implications for rhino health 2 3 Keylie M. Gibson1,2, Bryan N. Nguyen1,2, Laura M. Neumann3, Michele Miller4, Peter 4 Buss5, Savel Daniels6, Michelle Ahn1,2, Keith A. Crandall1,7*, & Budhan Pukazhenthi8* 5 6 1Computational Biology Institute, The Milken Institute School of Public Health, George 7 Washington University, Washington, DC, USA 8 2Department of Biological Sciences, George Washington University, Washington, DC, 9 USA 10 3Department of Environmental and Occupational Health, The Milken Institute School of 11 Public Health, The George Washington University, Washington, DC, USA 12 4DST-NRF Centre of Excellence for Biomedical Tuberculosis Research; South African 13 Medical Research Council Centre for Tuberculosis Research; Division of Molecular 14 Biology and Human Genetics, Faculty of Medicine and Health Sciences, Stellenbosch 15 University, Cape Town, South Africa. 16 5South African National Parks, Veterinary Wildlife Services, Kruger National Park, 17 Skukuza, South Africa 18 6Department of Botany and Zoology, University of Stellenbosch, Private Bag X1, 19 Matieland 7602, South Africa 20 7Department of Epidemiology and Biostatistics, The Milken Institute School of Public 21 Health, George Washington University, Washington, DC, USA 22 8Smithsonian’s National Zoo and Conservation Biology Institute, Front Royal, VA, USA 23 *Co-Senior Authors 24 25 Corresponding author: 26 Budhan Pukazhenthi 27 Smithsonian Conservation Biology Institute 28 1500 Remount Road, Front Royal, VA 22630 29 pukazhenthib@si.edu 30 Microbiome differences in wild and captive black rhinoceros 2 Abstract 31 A number of recent studies have shown the importance of the mammalian gut 32 microbiome in host health. In the context of endangered species, a few studies have 33 examined the relationship between the gut microbiome in wild versus captive 34 populations due to digestive and other health issues. Unfortunately, the results seem to 35 vary across taxa in terms of captive animals having higher, lower, or equivalent 36 microbiome diversity relative to their wild counterparts. Here, we focus on the black 37 rhinoceros as captive animals suffer from a number of potentially dietary related health 38 effects. We compared gut microbiomes of wild and captive black rhinos to test for 39 differences in taxonomic diversity (alpha and beta) and in functional diversity of the 40 microbiota. We incorporated a more powerful metagenomic shotgun sequencing 41 approach rather than a targeted amplification of the 16S gene for taxonomic assignment 42 of the microbiota. Our results showed no significant differences in the alpha diversity 43 levels between wild and captive black rhinos, but significant differences in beta diversity. 44 We found that bacterial taxa traditionally associated with ruminant guts of domesticated 45 animals had higher relative abundances in captive rhinos. Our metagenomic 46 sequencing results suggest that unknown gut microbes of wild rhinos are being 47 replaced by those found in conventional human-domesticated livestock. Wild rhinos 48 have significantly different functional bacterial communities compared to their captive 49 counterparts. Functional profiling results showed greater abundance of glycolysis and 50 amino acid synthesis pathways in captive rhino microbiomes, representing an animal 51 receiving sub-optimal nutrition with a readily available source of glucose but possibly an 52 imbalance of necessary macro and micronutrients. Given the differences observed 53 between wild and captive rhino gut microbiomes, we make a number of 54 recommendations for potentially modifying captive gut microbiota to better reflect their 55 wild counterparts and thereby hopefully improve overall rhino health in captivity. 56 57 Keywords: black rhinoceros; Diceros bicornis; ex situ population; critically endangered; 58 metagenomics; microbiome; captivity 59 Microbiome differences in wild and captive black rhinoceros 3 Introduction 60 From more than 100,000 free-ranging African black rhinos in the 1960s, this 61 critically endangered species has declined by more than 90% to approximately 5,000 62 animals today1. On average over 1,000 rhinos are poached annually in range countries 63 that include South Africa, Namibia, Kenya, and Zimbabwe1. Currently, fewer than 100 64 black rhinos (southern and eastern sub-species combined) reside in zoological 65 institutions in Northern America as a reservoir against potential extinction2. However, 66 the ex situ population experiences its own threats to survival, including a myriad of 67 unusual disease syndromes not generally described in the wild3–10, as well as poor 68 reproduction11,12 and fragemented populations3. Across mammals, recent studies have 69 suggested that microbiome differences between wild and captive populations may 70 influence overall health in general and digestive and immune functions in particular4. 71 A number of recent studies have identified differences between captive-wild 72 populations or domesticated-wild populations of mammals. For example, Schmidt et 73 al.10 examined microbiome diversity between wild and captive individuals of deer mice 74 (Peromyscus maniculatus) and found that mice from natural environments had more 75 diverse gut microbiome communities, but that gut microbiomes were more similar by 76 like environments rather than wild versus captive. Likewise, Clayton et al.5 showed that 77 in nonhuman primates, captivity ‘humanizes’ their microbiome showing a convergence 78 to gut microbiota reflective of the human gut via replacement of diverse microbial 79 diversity across species. Wasimuddin et al.13 compared wild versus captive cheetahs 80 and reported differences in gut microbiome between kin and nonkin individuals as well 81 as a higher incidence of pathogenic strains in captive cheetahs. McKenzie et al.4 took a 82 broader taxonomic approach and examined gut microbiome diversity across captive 83 versus wild populations of a variety of mammals. They investigated trends across six 84 mammalian orders and found alpha diversity between wild and captive populations 85 consistent across some mammalian hosts, decreased in captive populations in some 86 hosts, and increased in one host – namely, the rhinoceros4. Interestingly, this 87 conclusion was a combined analysis across both white and black rhinos with limited 88 sampling (especially unbalanced sampling in the black rhino with six captive but only 89 Microbiome differences in wild and captive black rhinoceros 4 one wild individual). Clearly, the jury is still out on the impacts of captivity on gut 90 microbiome diversity, and it may very well be that the impact is species specific. 91 These previous studies testing the associations of gut microbiome diversity in 92 captive versus wild populations have focused on targeted amplicon sequencing of a 93 single gene, 16S rRNA, to characterize the gut microbiome as a metataxonomic 94 approach. While less cost effective, taking a shotgun metagenomic approach to 95 characterizing the gut microbiome provides a number of advantages6. First, the 96 metagenomic approach does not rely on PCR and is therefore not subject to PCR 97 artifacts7. Metagenomics provides greater resolution (down to strain level) compared to 98 metataxonomic approaches. Metagenomics can also identify virus, fungus, and other 99 taxa in addition to bacteria – all in the same sequencing run8. It also provides for greater 100 functional assignments as the data survey across the genome, not a single ribosomal 101 gene9. Given these advantages of metagenomics and the lack of consensus on the 102 impact of captivity on gut microbiome coupled with our focus on black rhino health in 103 captivity for conservation options, we applied metagenomic sequencing to wild and 104 captive black rhino fecal material to characterize microbiome diversity as well as test for 105 differences between wild and captive animals from both taxonomic and functional 106 perspectives. We then make recommendations based on this collective information for 107 adjusting diet to create a normative gut microbiome, which in turn may promote better 108 black rhino health. 109 110 Results 111 Gut Microbiome 112 Read Mapping and Extraction Approaches 113 All samples were sequenced to a depth of at least 6.9 million paired-end reads 114 per sample, with an average of 12,479,613 paired-end reads. Very few reads were 115 discarded during quality trimming (3.16%); the post QC average was 12,085,574 paired-116 end reads per sample. 117 Overall, low mapping rates (<10%; number of mapped reads/number of cleaned 118 reads) were observed across all samples and across all three metagenomic mapping 119 software platforms (PathoScope, Kraken, and Centrifuge; Table 1). Although 120 Microbiome differences in wild and captive black rhinoceros 5 PathoScope mapped fewer reads than Kraken and Centrifuge, PathoScope provided 121 better resolution at the lower taxonomic levels. Prokaryotes made up most of the known 122 mapped reads (average 3% of reads per sample with PathoScope). Rhino and human 123 contamination in the reads were low (~1% average PathoScope mapped reads, 124 respectively). Despite the wild rhino fecal samples having plant material visibly present 125 in the DNA/RNA Shield, very few reads mapped to plant genomes from the Gramene 126 database (average 0.02% or 2,800 reads). On average, 97% of all reads were 127 unmapped, a surprisingly high proportion of reads that suggests our reference 128 databases are not robust to perhaps novel taxa coming from black rhino guts. A slightly 129 higher proportion of unmapped reads were in the wild rhinos compared to captive 130 (average 12,090,472 vs 11,625,025 reads, respectively); one of the wild rhinos (R08) 131 had 36.3% of its reads map to the rhino reference genome, accounting for over 3.5 132 million reads mapped. This was the most reads mapped for any rhino to any database. 133 Wild rhinos had over a 2,000-fold increase over captive rhinos in reads that mapped to 134 the rhino reference genome (average=408,545 reads vs 156 reads, respectively). 135 Notably, the captive samples (R21-R28) had consistently higher proportions of assigned 136 reads for each sample (average 460,939 reads) compared to wild samples (R01-R20; 137 average 287,611 reads). 138 139 Table 1. Average mapping percentage for all metagenomic mapping software platforms. 140 Database Wild Rhinos Captive Rhinos PathoScope Centrifuge Kraken PathoScope Centrifuge Kraken Rhino 0.52% NA NA 0.00% NA NA Human 0.00% NA NA 0.00% NA NA Prokaryotes 2.06% 8.34% 2.97% 3.29% 10.06% 3.55% Eukaryotes NA 4.49% 0.13% NA 3.96% 0.17% Other 0.10% 0.05% 0.00% 0.13% 0.05% 0.00% Unknown 97.21% 87.17% 96.91% 96.55% 85.93% 96.28% 141 The operational taxonomic unit (OTU) richness was inconsistent across samples 142 and was influenced by extraction kit used. The ZymoBIOMICS-extracted samples had 143 more OTU hits on average compared to the MoBio PowerFecal-extracted samples, 144 Microbiome differences in wild and captive black rhinoceros 6 regardless of origin (wild vs captive; p=0.099). PathoScope assigned a larger proportion 145 of reads to the set of wild samples (R2-R11) that were stored in Zymo DNA/RNA shield 146 and extracted using the ZymoBIOMICS DNA Miniprep kit (average 447,354 reads) than 147 the wild samples frozen and extracted with MoBio PowerFecal kit (average 155,729 148 reads). The Zymo wild samples had an average of 350 OTUs at the species level, 149 whereas the MoBio wild samples had 250 OTUs (p=0.036). Thus, collection method did 150 influence the microbiome composition. Observed communities differed markedly 151 between the MoBio-extracted and Zymo-extracted wild samples (PERMANOVA, 152 p<0.0001). However, several bacterial species such as Bacteroides fragilis and 153 Escherichia coli were present in both MoBio- (R01-R08) and Zymo-extracted (R11-R20) 154 samples. 155 156 Taxonomic Composition and Diversity 157 Differences in higher level taxonomic community composition were observed 158 between wild and captive rhinos (Fig. 1A-B). Specifically, Proteobacteria was more 159 abundant in wild compared with captive rhinos while Bacteroidetes was abundant in 160 captive rhinos. The captive rhinos also had elevated levels of Euryarchaeota, while their 161 wild counterparts showed a higher incidence of phylum Actinobacteria and class 162 Acidaminococcales. Additionally, the class Erysipelotrichia was present in five out of the 163 eight captive rhinos, but in only a single wild rhino. 164 At the phylum and class level, abundances of a few bacteria distinguish the rhino 165 microbiomes based on captive versus wild status. Analysis at the genus level revealed 166 greater differences between the wild and captive rhinos (Fig. 2A). Genera Bacteroides 167 and Prevotella were increased in all captive rhinos, while Escherichia, Oscillibacter, 168 Pseudobutyrivibrio, and Treponema were higher in the wild counterparts. However, both 169 wild and captive rhinos expressed the major groups of microbes for digestion 170 (cellulolytic, amylolytic) but were represented by different species (i.e., functionally 171 similar, but taxonomically distinct microbiomes; Fig. 2B). There also were a number of 172 species that were differentially abundant between wild and captive rhinos. For example, 173 the methane producing bacteria Methanocorpusculum bavaricum (p=0.0059) was more 174 abundant in wild rhinos, whereas captive rhinos contained Methanobrevibacter 175 Microbiome differences in wild and captive black rhinoceros 7 ruminantium (p=2.06e-28). Bacteroides fragilis (p=0.0005), Steptococcus suis (p=3.94e-176 15), and Escherichia coli (p<0.001; strains range from p=4.8e-12 through 3.5e-05) also 177 were found to be high in wild rhinos, whereas Ruminoccus albus (p=1.73e-27) and 178 Prevotella ruminicola (p=0.0003) were highly abundant in animals maintained in 179 captivity. However, a quarter of the assigned OTUs were not able to be assigned down 180 to species level, again suggesting the inadequacy of the reference database for 181 microbes. 182 183 Figure 1. Rhino microbiome composition, as determined by PathoScope, broken down 184 by (A) phylum and (B) class, grouped by wild versus captive host. Empty space 185 represents bacterial reads not identified at the corresponding taxonomic rank. Taxa 186 representing less than 1% of reads on average and less than 5% across all samples 187 were filtered out for the sake of visualization. 188 189 Microbiome differences in wild and captive black rhinoceros 8 190 191 192 193 194 195 196 197 198 Microbiome differences in wild and captive black rhinoceros 9 Figure 2. Black rhino bacterial microbiome composition, as determined by PathoScope, 199 broken down by (A) genus and (B) species, grouped by wild versus captive host. Empty 200 space presents bacterial reads not identified at the corresponding taxonomic rank. Taxa 201 representing less than 1% of reads on average and less than 5% across all samples 202 were filtered out for the sake of visualization. 203 204 205 206 207 Microbiome differences in wild and captive black rhinoceros 10 208 Common taxa were found across the wild and captive rhinos, respectively, 209 suggesting a core rhino microbiome based on captivity status (Supplementary Table 210 S1a-b). The phylum Firmicutes dominated the microbiome of both the wild and captive 211 rhinos, which comprised 32.7% and 20.8% of total mapped reads and 51.1% and 48.0% 212 of the core microbiome, respectively. However, Bacteroidetes was the second dominant 213 phylum in the captive rhino microbiome (42.4%), while the next leading dominate phyla 214 in the wild rhinos were Proteobacteria (23.6%) and Bacteroidetes (17.6%). 215 Although the taxonomic composition of the rhinos shows differences between 216 wild and captive gut microbiomes, alpha diversity measures between the two groups 217 were similar. However, the observed species richness (p=0.082) did indicate that the 218 wild rhinos have a higher median observed OTU richness (~335) compared to the 219 captives (~220) (Fig. 3), which is consistent with all the taxonomic results. Both the 220 Shannon (p=0.36) and Simpson (p=0.69) diversity indices indicate that the rhinos show 221 high diversity, independent of their origin (wild vs captive), with wild rhinos showing 222 slightly (but not significantly) higher diversity (Fig. 3). For all three alpha diversity 223 measures, the samples derived from wild rhinos displayed greater variance along with 224 more outliers compared to the captive samples and clustered together based on their 225 origin (wild vs captive; Fig. 4). These patterns were consistent across both the Jaccard 226 and Bray-Curtis indices. Furthermore, these patterns were consistent between the 227 PathoScope (Fig. 4A,B) and PhyloSift results (Fig. 4C,D). The dissimilarity metrics, 228 measured separately with Bray-Curtis, Jaccard, and JSD indices with 10,000 229 permutations, were all highly significant (PERMANOVA, p=9.999e-05), representative of 230 the centroids being different between the two groups. This is indicative of the two 231 groups having distinct and different communities. Together the ordination plots and 232 PERMANOVA indicated that the microbiomes of rhinos were more similar to other 233 rhinos with the same captivity status, with wild rhinos displaying considerably more 234 variation and range than captive rhinos. 235 236 Figure 3. PathoScope sample-level OTU richness and diversity (Shannon and Simpson 237 indices) of the wild and captive rhino populations. 238 Microbiome differences in wild and captive black rhinoceros 11 239 240 Figure 4. Non-metric multidimensional scaling plots of PathoScope data using Jaccard 241 distances (A) and Bray-Curtis distances (B) and of PhyloSift data using Jaccard 242 distances (C) and Bray-Curtis distances (D). 243 244 245 246 Microbiome differences in wild and captive black rhinoceros 12 Functional Analysis 247 A total of 39 gene ontology (GO) terms were found to be differentially abundant 248 between wild and captive rhino microbiomes with a Q-value less than 0.05 249 (Supplementary Table S2). The majority of the GO terms were positiviely associated 250 with captive rhinos’ microbiomes. A total of 127 pathways were differentially abundant, 251 however only two pathways (PWY_5103 L_isoleucine_biosynthesis_III and PWY_6121 252 5_aminoimidazole_ribonucleotide_biosynthesis_I) had Q-values under 0.05, likely due 253 to the small sample size (Supplementary Table S3). Captive rhino microbiomes seem to 254 have higher activity for bacterial replication and amino acid production. Additionally, 255 functional pathways and GO terms show indications of higher starch availability in the 256 captive rhinos. The wild and captive rhinos present different pathways, suggesting that 257 different metabolic activity is occurring between the two groups. 258 259 Discussion 260 The gut microbiome plays a key role in health and the well-being of animals, yet 261 there is no consensus on how the gut microbiome might change between wild and 262 captive animals4. In herbivores, the bacterial population in the gut is involved in the 263 breakdown of fibrous plant material into various metabolites including small chain fatty 264 acids (SCFA) that exert a significant impact on host health. Previous studies on gut 265 microbiome diversity in captive compared to wild animals have limited their inference to 266 a single gene for identifying known bacterial taxa which lacks the genomic breadth and 267 taxonomic depth available through shotgun metagenomics. Here we capitalize on the 268 powerful metagenomics approach to characterized and test for differences in alpha 269 diversity, beta diversity, and functional diversity in the gut (fecal) microbiome of wild and 270 captive black rhinos, a critically endangered species, with the goal of using this 271 information to improve health in captive animals. 272 Because we collected samples both in the US (captive) and South Africa (wild), 273 we used two different kits (ZymoBIOMICs and MoBio [now Qiagen] PowerFecal) for 274 preservation and DNA extraction due to regional availability of these kits. The Zymo-275 extracted samples produced more mapped sequencing read results than MoBio-276 extracted samples, irrelevant of bioinformatic software used (PathoScope, Kraken, 277 Microbiome differences in wild and captive black rhinoceros 13 Centrifuge). There was a difference in OTU richness between the two extraction 278 methods utilized in this study, and the collection/extraction procedures did influence the 279 microbiome composition. Although alpha diversity (Shannon and Simpson diversity 280 indices) were similar between the two extraction kits, there were distinct differences 281 between Zymo and MoBio extracted samples. This difference could be attributed to 282 differences in the kits used. Specifically, the Zymo kit was designed to efficiently isolate 283 bacterial, fungal, protozoan, algae, viral, mitochondrial, and host DNA from mammalian 284 feces, soil, fungal/bacterial cells, biofilms and water. Thus, the Zymo kit is more generic 285 and therefore was better optimized for broader microbiome usage in contrast to the 286 MoBio kit which was optimized for human fecal samples. Furthermore, the samples 287 extracted using the Zymo kit were stored in Zymo DNA/RNA Shield preservation 288 solution, and thus may have preserved more of the microbes between time of collection 289 to extraction. Therefore, care should be exercised during collection, storage and 290 processing of fecal samples from wildlife for metagenomic analyses especially under 291 field conditions. 292 A high proportion of bacteria identified in fecal samples were conserved between 293 wild and captive rhinos but differences that distinguished a significant change in 294 microbial communities due to captivity were also detected, as also seen by several 295 other captivity studies utilizing 16S amplicon sequencing on mammals5,13–17. However, 296 our metagenomic approach resulted in a large number of unmappable sequencing 297 reads (~90%) from the black rhinos, suggesting a lack of relevant and known bacterial 298 genomes in the database. One of the great advantages of the metagenomics approach 299 is that you can discover and quantify the unknown microbes as well as the known and 300 our results indicate that targeted 16S amplicon sequencing is missing much of the 301 microbial diversity given that 90% of the reads could not be mapped to reference 302 genomes. This result was validated across three different software platforms for 303 characterizing microbiome diversity, namely PathoScope, Kraken, and Centrifuge. Due 304 to the significant lack of curated and verified microbial genomes from wildlife in genomic 305 databases (i.e., NCBI’s RefSeq), there is a critical need to investigate these under-306 studied systems (i.e., wildlife and the seasonal dynamics of their diverse microbiomes) 307 to reconstruct new genomes to fully identify the organisms present in their microbiome. 308 Microbiome differences in wild and captive black rhinoceros 14 Identifying the unknown microbes from wildlife microbiomes will provide the entire 309 research and wildlife health communities with the necessary information to accurately 310 characterize and potentially alter the microbiome of captive species to improve health. 311 While we found no significant difference in alpha-diversity of microbial 312 communities between wild and captive populations of black rhino (both had high 313 numbers of microbial species), the beta-diversity was significantly different suggesting 314 there are distinct microbial communities in wild versus captive black rhinos. Our results 315 showed increased assignment of microbial reads in the captive samples to bacterial 316 taxa traditionally associated with ruminant guts (such as Ruminococcus albus, 317 Selenomonas bovis, and Treponema bryantii), suggesting that the unknown (prokaryotic 318 genomes not present in NCBI’s RefSeq) gut microbes of wild rhinos were being 319 replaced by those found in conventional human-domesticated livestock. This 320 replacement could be partially due to the rhinos receiving a similar diet to cows and 321 horses or to the close proximity to cows, horses or humans that captive rhinos are often 322 in contact with. Wild rhinos seem to follow a microbiome profile closer to healthier 323 domestic animals, with greater beta diversity, functional diversity, and variation between 324 individual rhinos compared to captive rhinos18. 325 With our metagenomic data, we were able to establish a core microbiome for 326 both the wild and captive black rhino. The wild black rhinos’ microbiome comprised of 327 Firmicutes (51%), Proteobacteria (23.6%) and Bacteroidetes (17.6%). In contrast, the 328 microbiome of captive black rhinos comprised of Firmicutes (48%) and Bacteroidetes 329 (42.4%). Similarly, an earlier study reported that the white rhinoceros gut microbiome 330 was predominantly comprised of Firmicutes and Bacteroidetes constituting over 90% of 331 total sequences19. Interestingly, Firmicutes and Bacteroidetes represented the most 332 ubiquitous taxa in the vertebrate microbiome and Firmicutes was determined to be the 333 most abundant phyla in the hindgut of healthy humans and other mammals20. Our 334 results show, at the phylum level, that captive black rhino microbiome diversity mirrored 335 the white rhino (captive) with a preponderance of Firmicutes and Bacteroidetes19. A 336 recent study compared microbiome diversity of captive southern white rhino 337 (Ceratotherium simum simum) and captive greater one-horned rhino (Rhinoceros 338 unicornis) using 16S sequencing21. They, too, found predominantly Firmicutes and 339 Microbiome differences in wild and captive black rhinoceros 15 Bacteroidetes, consistent with our captive black rhinos, but in different proportions with 340 the southern white rhino having more Bacteroidetes (55%:30%) whereas the greater 341 one-horned rhino had more Firmicutes (55%:33%)21. Thus, the wild animals seem to 342 have reduced Bacteroidetes and novel Proteobacteria compared to captive rhinos, and 343 captive rhinos of different species seem to have converged on a dominance of 344 Bacteriodetes and Firmicutes to the exclusion of Proteobacteria. 345 With very few studies existing on rhino microbiome, their closest domestic 346 relative, horses, can be utilized as a source of comparison. Previous studies in healthy 347 horses have shown that Firmicutes are seen in a higher ratio compared to 348 Bacteroidetes22, while higher proportions of Bacteriodetes are associated with colitis23. 349 However, there is minimal information on the incidence of colitis in captive rhinos and 350 warrants further investigation. In contrast, Proteobacteria constituted the second most 351 abundant phyla in the wild counterparts. Although Protebacteria is considered a core 352 microbe of herbivores24, this phylum also includes a wide variety of well-known 353 pathogens like Eschericia coli, Salmonella, Vibrio, Helicobactor and others25. These 354 findings may be influenced by the fact that in the wild, rhinos share water sources 355 (water holes) often visited by numerous other species. It is not uncommon that animals 356 defecate in these areas and as a result, contaminate the water with various other 357 microbes that in turn could establish in the gut of animals consuming this water. 358 A comparison of the functional diversity in the black rhino microbiome 359 demonstrated a greater abundance of glycolysis and amino acid synthesis pathways in 360 captive compared wild counterparts suggesting dysbiosis resulting from diet offered in 361 captivity. Captive black rhinos also showed indications of high starch availability. 362 Captive rhino diets consist of ~36% neutral detergent fiber (NDF) and ~25% acid 363 detergent fiber (ADF) in the commercial products, 36-50% NDF and 28-39% ADF in 364 alfalfa hay, and 49-69% NDF and 31-41% ADF in grass hay26. The largest proportion of 365 the diet comes from alfalfa hay and commercial products, which represents a lower fiber 366 content range than what wild rhinos have been observed comsuming with NDF ranging 367 from 30-78% and ADF ranging from 14-59%27. As such, when compared with their 368 captive counterparts, the microbiome of wild black rhinos contained a higher proportion 369 of bacteria involved in breakdown of plant materials. Specifically, we identified higher 370 Microbiome differences in wild and captive black rhinoceros 16 proportions of Escherichia, Oscillibacter, Pseudobutyrivibrio, and Treponema in wild 371 black rhinos. All these taxa are known to be involved in breakdown of fibers. 372 Furthermore, Pseudobutyrivibrio are involved in butyrate production, which has also 373 been reported to be higher in healthy animals by supporting healthier papillae in the 374 gut28. The SCFAs acetate, butyrate and propionate are important in several 375 physiological aspects of the host’s nutrient acquisition, immune function, cell signaling, 376 and pathogen protection29. 377 Our study represents the most extensive analysis of the gut microbiome of free-378 ranging (wild) southern black rhinoceros capitalizing on the more powerful and insightful 379 metagenomic sequencing approach. Similar to earlier studies in other large herbivores, 380 we identified a core microbiome comprising of Firmicutes, Bacteriodetes, and 381 Proteobacteria in the wild black rhinos. These phyla have been reported in most hind 382 gut fermenters and are involved in breakdown of fibrous plant material 383 (polysaccharides). Comparison of gut microbiome between wild and captive rhinos 384 demonstrates a preponderance of bacterial families involved in carbohydrate 385 metabolism. Although this is preliminary, the physiological significance of this new 386 finding cannot be overlooked. Several studies have shown that increased utilization of 387 carbohydrates could lead to dysbiosis in the gut and associated changes in systemic 388 immune function. However, further analysis of a large population of captive managed 389 black rhinos would help confirm these findings and also examine the impact, if any, on 390 metabolic status and immune function. The metagenomic sequencing provides a new 391 minimally invasive and high resolution technique for evaluating nutrition and response to 392 potential interventions. Given the differences discovered between the wild and captive 393 gut microbiomes of the black rhino, there is a clear path to potentially altering the 394 captive gut microbiome to better reflect the wild microbiome diversity and test for 395 improved overall health of captive populations. This could be achieved through a 396 combination of changes in captive black rhino diet, administration of probiotics to better 397 reflect the wild rhino core microbiome, and/or the application of fecal transplant to 398 restore gut microbiome diversity30. Future studies should sample wild and captive rhinos 399 longitudinally to assess temperal and seasonal variation in the gut microbiome to better 400 inform approaches to restore a healthly microbiome in captive populations. 401 Microbiome differences in wild and captive black rhinoceros 17 402 Materials and Methods 403 Collection 404 Permits to collect, process, and transport samples (both within South Africa and 405 to the US) from wild and captive black rhinos were obtained from the South African 406 National Parks, CITES, and US Fish and Wildlife service. A total of 25 fecal samples 407 were collected from 17 wild and 8 captive black rhinos (Table 2). All wild animals were 408 opportunistically sampled during routine translocation efforts in South Africa. Animals 409 were immobilized using a combination of etorphine (9.8 mg/ml, Novartis, Kempton Park 410 1619, South Africa), azaperone (40 mg/ml, Janssen Pharmaceutical Ltd., Halfway 411 House 1685, South Africa) and hyaluronidase (5000 i.u./vial, Kyron Laboratories, 412 Benrose 2011, South Africa) delivered remotely by dart. At the end of the procedure, 413 naltrexone (40 mg/ml, Kyron laboratories) was administered intravenously to reverse 414 the immobilization. All wild rhinos were considered healthy based on physical 415 appearance and behavior and received no supplemental commercial diets. The eight 416 captive samples were collected from black rhinos located on a private ranch in Texas as 417 well as an Association of Zoos and Aquariums accredited institution (also in Texas). 418 Captive animals were fed Alfalfa hay (2 squares), Coastal grass hay (1 square), a grain-419 free hay enhancer (Elephant and White Rhino Supplement, Mazuri, St. Louis, MO; 5 420 lbs), Strategy healthy edge (a high-fat nuggets that delivers a controlled starch and 421 sugar as well as higher fat and fiber; Purina Animal Nutrition, Gary Summit, MO; 5 lbs), 422 a stabilized rice bran supplement (Max-E-, MannaPro, Chesterfiled, MO; 1 lb), and fresh 423 cut huisache twice daily. Animals also received electrolyte powder (Electro Dex; Farnam 424 Companies Inc; Phoenix, AZ; 2 oz per 20 gallons of water daily) as well as apples and 425 sweet potatoes for treats. 426 The eight captive fecal samples were stored frozen (-80˚C) until DNA extraction. 427 Eight of the 17 wild samples were transported to Stellenbosch University in South Africa 428 for DNA extraction and shipped to the United States as purified genomic DNA. For the 429 remaining nine wild fecal samples, between 1-2 grams of feces were stored in 430 DNA/RNA Shield (Zymo Research, USA) and transported into the United States. These 431 samples also were stored frozen (-80˚C) until DNA extraction was attempted. 432 Microbiome differences in wild and captive black rhinoceros 18 433 Table 2. List of all black rhinos sampled with corresponding metadata and captivity 434 status. 435 Sample Number Captivity Status Extraction Kit DNA/RNA Shield Sex Age Sample Type R01 Wild MoBio No M 3.5 yr Feces R02 Wild MoBio No F SAd Feces R03 Wild MoBio No M Juvenile Feces R04 Wild MoBio No F SAd Feces R05 Wild MoBio No F SAd Feces R06 Wild MoBio No M SAd Feces R07 Wild MoBio No F Ad Feces R08 Wild MoBio No F Ad Feces R11 Wild Zymo Yes M Ad Feces R12 Wild Zymo Yes M Ad Feces R14 Wild Zymo Yes M Ad Feces R15 Wild Zymo Yes M Ad Feces R16 Wild Zymo Yes F Sad Feces R17 Wild Zymo Yes M Ad Feces R18 Wild Zymo Yes F Sad Feces R19 Wild Zymo Yes M Calf Feces R20 Wild Zymo Yes M Ad Feces R21 Captive Zymo Yes F Ad Feces R22 Captive Zymo Yes M Ad Feces R23 Captive Zymo Yes F Ad Feces R24 Captive Zymo Yes M Ad Feces R25 Captive Zymo Yes F Ad Feces R26 Captive Zymo Yes F Ad Feces R27 Captive Zymo Yes F Ad Feces R28 Captive Zymo Yes F Ad Feces Abbreviations: adult (Ad), senior adult (Sad), male (M), female (F), MoBio Microbiome differences in wild and captive black rhinoceros 19 PowerFecal kit, which is now QIAamp PowerFecal DNA kit (MoBio), and ZymoBIOMICS DNA Miniprep (Zymo). 436 DNA extraction and metagenomic sequencing 437 Fecal samples R1-R8 (wild rhinos) were processed for DNA extraction in South 438 Africa using the MoBio PowerFecal kit (now QIAamp PowerFecal DNA kit; QIAGEN, 439 USA) per manufacturer’s instructions. Samples (R11-28; consisting of wild and captive 440 rhinos) stored in DNA/RNA Shield were processed (ZymoBIOMICS DNA Miniprep 441 extraction kit; Zymo Research, USA) per manufacturer’s instructions. Different 442 extraction kits were used due to the different locations and availability of kits for the 443 molecular work. In order to minimize biases from extraction method, we placed a 1g 444 scoop of feces from each captive sample into the Zymo Research DNA/RNA Shield 445 resulting in 1 mL of fresh fecal material in solution and were stored frozen until 446 processing (as described above). The only alterations to the Zymo extraction kit 447 manufacturer’s instructions were the following: 1) we began with the sample amount of 448 1 mL of DNA/RNA Shield that contained the fecal sample and 2) we secured the 449 samples in a bead beater and processed at maximum speed for 30 minutes. All DNA 450 samples were processed for sequencing using an Illumina Nextera XT library prep kit 451 (Illumina, Inc., USA) and then sequenced with a single High Output v2 Kit (300 cycles) 452 run on an Illumina NextSeq 500 platform (Illumina, Inc., USA) at the George 453 Washington University Milken Institute School of Public Health Genomics Core Facility. 454 We compared the number of OTUs identified, observed species richness, Simpson and 455 Shannon diversity indices, and dissimilarity metrics between wild samples extracted 456 with the MoBio extraction kit and the Zymo extraction kit to test for biases associated 457 with extraction approaches. 458 459 Bioinformatic analyses 460 Quality of the reads was assessed using FastQC v. 0.11.5 461 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/). Short reads, low quality 462 reads, and reads with adapter contamination and low quality bases were removed from 463 the FASTQ files by trimming using Flexbar v. 3.0.331. Then low complexity reads were 464 Microbiome differences in wild and captive black rhinoceros 20 removed from analysis with PRINSEQ v. 0.20.432. Following data quality check, FastQC 465 was repeated to assess the efficacy of quality trimming and cleaning. The resulting high 466 quality reads were mapped with PathoScope 2.0’s 33 mapping module 34, which utilizes 467 a wrapper for Bowtie235, to the representative and reference prokaryote, viroid, and 468 virus genome databases available from GenBank 469 (https://www.ncbi.nlm.nih.gov/genbank/). Reads that mapped best to the rhino genome 470 (white rhinoceros: Ceratotherium simum, NCBI Assembly ID: 406328), human genome 471 (hg38), plant genomes from Gramene (http://www.gramene.org), the representative and 472 reference genomes databases for fungi and protozoan from GenBank or to the 473 WormBase parasite genome database (http://parasite.wormbase.org/index.html) also 474 were removed. We assessed levels of contamination in the sequencing reads by the 475 number of reads that mapped better to any database but the prokaryotic database. 476 PathoScope 2.0 ID module33 was used to assign taxonomy. We also generated 477 taxonomic assignments using PhyloSift’s v. 1.0.1 core and extended marker sets v. 478 1413946442, which phylogenetically places reads matching conserved marker genes to 479 infer taxonomy36. We utilized additional metagenomic sequence classification software 480 Centrifuge v. 1.0.3-beta37 and Kraken v. v0.10.5-beta38 to assign taxonomy and validate 481 PhyloSift and PathosSope results. Functional analysis was completed using HUMAnN2 482 with the full UniRef90 database (http://huttenhower.sph.harvard.edu/humann2)39. 483 Subsequently, HUMAnN2 results identified pathways and were grouped into GO terms 484 and tested for associations with captivity status using MaAsLin: Multivariate Association 485 with Linear Models40 and filtered with a q-value of 0.05. 486 We analyzed and visualized the PhyloSift and PathoScope with R41 v. 3.5.0 in 487 Rstudio42 v. 1.1.453 using the phyloseq43, vegan44, DESeq245, and ggplot246 packages. 488 For taxonomic composition visualizations, OTUs were transformed into relative 489 abundances and then filtered to include only microbes that had a mean above 1% or a 490 maximum prevalence in any sample greater than or equal to 5%. The core microbiome 491 for each wild and captive rhinos was defined as those that were present in at least 50% 492 of the samples with greater than 0.1% relative abundance13. Differential abundance 493 analysis between all wild and captive rhinos was conducted with the DESeq2 and 494 Phyloseq packages with the PathoScope data. Data (OTU counts) were log-495 Microbiome differences in wild and captive black rhinoceros 21 transformed and variance-stabilized using geometric means to normalize sequencing 496 depth across samples. We determined significant species with an 0.01 alpha level that 497 we then used to filter the adjusted p-value; additional filtering based on relative 498 abundance was not completed. Observed species richness, Shannon diversity index, 499 and Simpson diversity index, which reflect the richness and evenness of microbial 500 representation in a sample, were estimated using the phyloseq and DESeq2 R 501 packages. ANOVA and Kruskal-Wallis tests in R were applied to compare extraction kit 502 differences between the wild microbiome samples and between wild versus captive 503 microbiomes. Alpha diversity metrics were analyzed with lmerTest47 R package. The 504 linear mixed-effects (LME) model analysis was implemented to test for associations 505 between alpha diversity indices and taxa abundances, extraction kit, age and sex. 506 Analysis showed that the only co-variable that showed an impact on the representation 507 of microbial analyses was the extraction kit, and therefore age and sex was not used in 508 final diversity analyses. Additionally, Jaccard (presence-absence) and Bray-Curtis 509 (abundance-weighted) indices were used and implemented in the vegan R package to 510 calculate the similarity of microbial communities between samples using OTU matrices 511 generated from the PathoScope output files. The resulting distance matrices, non-metric 512 multidimensional scaling (NMDS) and calculated dissimilarity metrics with Bray-Curtis, 513 Jaccard, and JSD indices, were compared with permutational multivariate analysis of 514 variance (PERMANOVA)48 with the vegan44 R package and significance was 515 determined with 10,000 permutations. 516 517 References 518 1. Knight, M. African Rhino Specialist Group report. Pachyderm 59, 14–26 (2018). 519 2. Saint Louis Zoo. Black Rhinoceros. Saint Louis Zoo (2019). Available at: 520 https://www.stlzoo.org/animals/abouttheanimals/mammals/hoofedmammals/black521 rhinoceros. (Accessed: 16th March 2019) 522 3. le Roex, N., Paxton, M., Adendorff, J., Ferreira, S. & O’Riain, M. J. Starting small: 523 long-term consequences in a managed large-mammal population. J. Zool. 306, 524 95–100 (2018). 525 4. Mckenzie, V. J. et al. The Effects of Captivity on the Mammalian Gut Microbiome. 526 Microbiome differences in wild and captive black rhinoceros 22 Integr. Comp. Biol. 1–15 (2017). doi:10.1093/icb/icx090 527 5. Clayton, J. B. et al. Captivity humanizes the primate microbiome. Proc. Natl. 528 Acad. Sci. 113, 10376–10381 (2016). 529 6. Hilton, S. K. et al. Metataxonomic and metagenomic approaches vs. culture-530 based techniques for clinical pathology. Front. Microbiol. 7, 1–12 (2016). 531 7. Soergel, D. A. W., Dey, N., Knight, R. & Brenner, S. E. Selection of primers for 532 optimal taxonomic classification of environmental 16S rRNA gene sequences. 533 ISME J. 6, 1440–1444 (2012). 534 8. Jovel, J. et al. Characterization of the gut microbiome using 16S or shotgun 535 metagenomics. Front. Microbiol. 7, 1–17 (2016). 536 9. Silva, G. G. Z., Green, K. T., Dutilh, B. E. & Edwards, R. A. SUPER-FOCUS: A 537 tool for agile functional analysis of shotgun metagenomic data. Bioinformatics 32, 538 354–361 (2015). 539 10. Schmidt, E., Mykytczuk, N. & Schulte-Hostedde, A. I. Effects of the captive and 540 wild environment on diversity of the gut microbiome of deer mice (Peromyscus 541 maniculatus). ISME J. (2019). doi:10.1038/s41396-019-0345-8 542 11. Hermes, R., Göritz, F., Streich, W. & Hildebrandt, T. Assisted reproduction in 543 female rhinoceros and elephants--current status and future perspective. Reprod. 544 Domest. Anim. 42, 33–44 (2007). 545 12. Garnier, J. N. et al. Matings system and reproductive skew in the black 546 rhinoceros. Mol. Ecol. 10, 2031–2041 (2001). 547 13. Wasimuddin et al. Gut microbiomes of free-ranging and captive Namibian 548 cheetahs: diversity, putative functions, and occurrence of potential pathogens. 549 Mol Ecol 26, 5515–5527 (2017). 550 14. Muegge, B. D. et al. Diet drives convergence in gut microbiome functions across 551 mammalian phylogeny and within humans. Science (80-. ). 332, 970–974 (2011). 552 15. Uenishi, G. et al. Molecular Analyses of the Intestinal Microbiota of Chimpanzees 553 in the Wild and in Captivity. Am. J. Primatol. 69, 367–376 (2007). 554 16. Delport, T. C., Power, M. L., Harcourt, R. G., Webster, K. N. & Tetu, S. G. Colony 555 Location and Captivity Influence the Gut Microbial Community Composition of the 556 Australian Sea Lion (Neophoca cinerea). Appl. Environ. Microbiol. 82, 3440 LP-557 Microbiome differences in wild and captive black rhinoceros 23 3449 (2016). 558 17. Delsuc, F. et al. Convergence of gut microbiomes in myrmecophagous mammals. 559 Mol. Ecol. 23, 1301–1317 (2014). 560 18. Henderson, G. et al. Rumen microbial community composition varies with diet and 561 host, but a core microbiome is found across a wide geographical range. Sci. Rep. 562 5, (2015). 563 19. Bian, G., Ma, L., Su, Y. & Zhu, W. The Microbial Community in the Feces of the 564 White Rhinoceros (Ceratotherium simum) as Determined by Barcoded 565 Pyrosequencing Analysis. PLoS One 8, 1–9 (2013). 566 20. Ley, R. Obesity and the human microbiome. Curr. Opin. Gastroenterol 26, 5–11 567 (2010). 568 21. Williams, C. L., Ybarra, A. R., Meredith, A. N., Durrant, B. S. & Tubbs, C. W. Gut 569 microbiota and phytoestrogen-associated infertility in southern white rhinoceros. 570 MBio (2019). doi:http://dx.doi.org/10.1101/451757 571 22. Ericsson, A. C., Johnson, P. J., Lopes, M. A., Perry, S. C. & Lanter, H. R. A 572 microbiological map of the healthy equine gastrointestinal tract. PLoS One 11, 1–573 17 (2016). 574 23. Costa, M. C. et al. Comparison of the fecal microbiota of healthy horses and 575 horses with colitis by high throughput sequencing of the V3-V5 region of the 16s 576 rRNA gene. PLoS One 7, (2012). 577 24. O’ Donnell, M. M., Harris, H. M. B., Ross, R. P. & O’Toole, P. W. Core fecal 578 microbiota of domesticated herbivorous ruminant, hindgut fermenters, and 579 monogastric animals. Microbiologyopen 6, 1–11 (2017). 580 25. Hinchliff, C. E. et al. Synthesis of phylogeny and taxonomy into a comprehensive 581 tree of life. Proc. Natl. Acad. Sci. 112, 12764–12769 (2015). 582 26. National Research Council. Nutrient Requirements of Horse: Sixth Revised 583 Edition. (The National Academies Press, 2007). doi:10.17226/11653 584 27. Dierenfeld, E. S., Toit, R. & Braselton, W. E. Nutrient Composition of Selected 585 Browses Consumed by Black Rhinoceros ( Diceros bicornis ) in the Zambezi 586 Valley, Zimbabwe. J. Zoo Wildl. Med. 26, 220–230 (2011). 587 28. Guilloteau, P. et al. From the gut to the peripheral tissues: The multiple effects of 588 Microbiome differences in wild and captive black rhinoceros 24 butyrate. Nutr. Res. Rev. 23, 366–384 (2010). 589 29. Tremaroli, V. & Bäckhed, F. Functional interactions between the gut microbiota 590 and host metabolism. Nature 489, 242–249 (2012). 591 30. Hourigan, S. et al. Fecal transplant in children with Clostridioides difficile gives 592 sustained reduction in antimicrobial resistance and pathogenic burden. Mol. Biol. 593 Evol. In review, 594 31. Roehr, J. T., Dieterich, C. & Reinert, K. Flexbar 3.0 – SIMD and multicore 595 parallelization. Bioinformatics 33, 2941–2942 (2017). 596 32. Schmieder, R. & Edwards, R. Quality control and preprocessing of metagenomic 597 datasets. Bioinformatics 27, 863–864 (2011). 598 33. Hong, C. et al. PathoScope 2.0: a complete computational framework for strain 599 identification in environmental or clinical sequencing samples. Microbiome 2, 1–600 15 (2014). 601 34. Francis, O. E. et al. Pathoscope: Species identification and strain attribution with 602 unassembled sequencing data. Genome Res. 23, 1721–1729 (2013). 603 35. Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. 604 Methods 9, 357–359 (2013). 605 36. Darling, A. E. et al. PhyloSift: phylogenetic analysis of genomes and 606 metagenomes. PeerJ 2, 1–28 (2014). 607 37. Kim, D., Song, L., Breitwieser, F. P. & Salzberg, S. L. Centrifuge : rapid and 608 sensitive classification of metagenomic sequences. 1–9 (2016). 609 doi:10.1101/gr.210641.116.Freely 610 38. Wood, D. E. & Salzberg, S. L. Kraken: Ultrafast metagenomic sequence 611 classification using exact alignments. Genome Biol. 15, (2014). 612 39. Abubucker, S. et al. Metabolic reconstruction for metagenomic data and its 613 application to the human microbiome. PLoS Comput. Biol. 8, (2012). 614 40. Morgan, X. C. et al. Dysfunction of the intestinal microbiome in inflammatory 615 bowel disease and treatment. Genome Biol. 13, R79 (2012). 616 41. R Core Team. R: A language and environment for statistical computing. (2013). 617 42. RStudio Team. RStudio: Integrated Development for R. (2015). 618 43. McMurdie, P. J. & Holmes, S. Phyloseq: An R Package for Reproducible 619 Microbiome differences in wild and captive black rhinoceros 25 Interactive Analysis and Graphics of Microbiome Census Data. PLoS One 8, 620 (2013). 621 44. Jari Oksanen, F. et al. vegan: Community Ecology Package. (2018). 622 45. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and 623 dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 1–21 (2014). 624 46. Wickham, H. ggplot2: Elegant Graphics for Data Analysis. (Springer-Verlag New 625 York, 2016). 626 47. Kuznetsova, A., Brockhoff, P. B. & Christensen, R. H. B. lmerTest Package: Tests 627 in Linear Mixed Effects Models. J. Stat. Softw. 82, (2017). 628 48. Anderson, M. J. Permutational Multivariate Analysis of Variance (PERMANOVA). 629 Wiley StatsRef Stat. Ref. Online 1–15 (2017). 630 doi:10.1002/9781118445112.stat07841 631 632 Acknowledgements 633 This research was funded by the International Rhino Foundation (R-2015-2). All 634 analyses including those with PathoScope, PhyloSift, Kraken, Centrifuge, and 635 HUMAnN2 were completed on George Washington’s Colonial One High Performance 636 Computing Cluster. 637 638 Author Contributions 639 KAC and BP designed the study and obtained funding. MM, PB, SD and BP collected 640 samples and SD, KMG and BNN extracted DNA from samples. KMG and BNN 641 prepared samples for next-generation sequencing. KMG, BNN, LMN and MA analyzed 642 the data. KMG, KAC, BNN, LMN and BP wrote original draft of the manuscript and all 643 authors reviewed and edited final version of the manuscript. 644 645 Addition Information 646 Data Availablity: The next-generation sequencing data assocatied with this study have 647 been deposited in GenBank under SRA xxx-yyy (provided upon acceptance). 648 649 Supplementary information accompanies this paper at . 651 652 Competing Interests: The authors declare no competing interests. 653 654 Figure and table legends 655 Table 1. Average mapping percentage for all metagenomic mapping software platforms. 656 Table 2. List of all black rhinos sampled with corresponding metadata and captivity 657 status. 658 659 Figure 1. Rhino microbiome composition, as determined by PathoScope, broken down 660 by (A) phylum and (B) class, grouped by population. Empty space represents bacterial 661 reads not identified at the corresponding taxonomic rank. Taxa representing less than 662 1% of reads on average and less than 5% across all samples were filtered out for the 663 sake of visualization. 664 665 Figure 2. Black rhino bacterial microbiome composition, as determined by PathoScope, 666 broken down by (A) genus and (B) species, grouped by population. Empty space 667 presents bacterial reads not identified at the corresponding taxonomic rank. Taxa 668 representing less than 1% of reads on average and less than 5% across all samples 669 were filtered out for the sake of visualization. 670 671 Figure 3. Pathoscope sample-level OTU richness and diversity (Shannon and Simpson 672 indices) of the wild and captive rhino populations. 673 674 Figure 4. Non-metric multidimensional scaling plots of PathoScope data using Jaccard 675 distances (A) and Bray-Curtis distances (B) and of PhyloSift data using Jaccard 676 distances (C) and Bray-Curtis distances (D). 677 678 Supplemental figure and table legends 679 Table S1a. Core rhino microbiome species in wild rhinos. 680 Table S1b. Core rhino microbiome species in captive rhinos. 681 Microbiome differences in wild and captive black rhinoceros 27 Table S2. Differentially expressed gene ontology terms between wild and captive rhino 682 samples. 683 Table S3. Differentially expressed abundant pathways between wild and captive rhino 684 samples. 685 686 0 0.2 0.4 0.6 0.8 1 R28 R27 R26 R25 R24 R23 R22 R21 R20 R19 R18 R17 R16 R15 R14 R12 R11 R08 R07 R06 R05 R04 R03 R02 R01 Actinobacteria Bacteroidetes Euryarchaeota Fibrobacteres Firmicutes Proteobacteria Spirochaetes A) Pathoscope Phylum Composition Relative Abundance S am pl e W ild R hi no s C ap ti ve R hi no s W ild R hi no s C ap ti ve R hi no s B) Pathoscope Class Composition Relative Abundance S am pl e 0 0.2 0.4 0.6 0.8 1 R28 R27 R26 R25 R24 R23 R22 R21 R20 R19 R18 R17 R16 R15 R14 R12 R11 R08 R07 R06 R05 R04 R03 R02 R01 Actinobacteria Bacilli Bacteroidia Clostridia Coriobacteriia Erysipelotrichia Fibrobacteria Gammaproteobacteria Methanobacteria Methanomicrobia Negativicutes Spirochaetia 0 0.2 0.4 0.6 0.8 1 R28 R27 R26 R25 R24 R23 R22 R21 R20 R19 R18 R17 R16 R15 R14 R12 R11 R08 R07 R06 R05 R04 R03 R02 R01 Acetivibrio Bacteroides Blautia Butyrivibrio Clostridium Corynebacterium Dorea Enterococcus Escherichia Eubacterium Fibrobacter Intestinimonas Lachnoclostridium Methanobrevibacter Methanocorpusculum Oscillibacter Phascolarctobacterium Prevotella Pseudobutyrivibrio Roseburia Ruminococcus Selenomonas Senegalimassilia Streptococcus Subdoligranulum Treponema A) Pathoscope Genus Composition Relative Abundance S am pl e W ild R hi no s C ap ti ve R hi no s 0 0.2 0.4 0.6 0.8 1 R28 R27 R26 R25 R24 R23 R22 R21 R20 R19 R18 R17 R16 R15 R14 R12 R11 R08 R07 R06 R05 R04 R03 R02 R01 [Clostridium] citroniae [Eubacterium] rectale Acetivibrio ethanolgignens Bacteroidales bacterium CF Bacteroides fragilis Bacteroides sp. Ga6A1 Clostridium phoceensis Enterococcus casseliflavus Escherichia coli Eubacterium sp. SB2 Fibrobacter succinogenes Lachnospiraceae bacterium AC2031 Methanobrevibacter millerae Methanobrevibacter oralis Methanobrevibacter ruminantium Oscillibacter sp. ER4 Oscillibacter sp. KLE 1745 Phascolarctobacterium succinatutens Prevotella brevis Prevotella copri Prevotella ruminicola Prevotella sp. P4-76 Pseudobutyrivibrio ruminis Ruminococcaceae bacterium AB4001 Ruminococcus albus Ruminococcus flavefaciens Streptococcus gallolyticus Streptococcus ovis Streptococcus suis Subdoligranulum sp. 4_3_54A2FAA Treponema sp. C6A8 B) Pathoscope Species Composition Relative Abundance S am pl e W ild R hi no s C ap ti ve R hi no s 200 400 600 OT U ric hn es s OTU richness 3 4 5 Sh an no n div er sit y Shannon diversity 0.75 0.80 0.85 0.90 0.95 Si m ps on d ive rs ity Captive Wild Simpson diversity −0.4 −0.2 0 0.2 0.4 0.6 −0.6 −0.4 −0.2 0 0.2 0.4 −0.3 −0.2 −0.1 0 0.1 0.2 −0.4 −0.2 0 0.2 0.4 MDS1 MDS1 M D S 2 M D S 2 Stress: 0.1303 A) PathoScope Jaccard NMDS Stress: 0.1395 B) PathoScope Bray-Curtis NMDS Stress: 0.1985 C) PhyloSift Jaccard NMDS Stress: 0.1713 D) PhyloSift Bray-Curtis NMDS Captive rhinos Wild rhinos