DNA was extracted at the Namibia-based CCF Conservation Genetics Laboratory from 0.25 g frozen feces using the QIAamp PowerFecal DNA Kit (Qiagen, MD) following manufacturer’s instructions. For each batch of sample extractions, a negative control was included and carried alongside the other samples throughout the experiment, to identify potential extraction contaminants. Following extractions, DNA concentrations were measured with a nanophotometer (Implen NP80-Touch; Implen, Munich, Germany).

Fecal bacterial DNA was amplified following a previously published two-step polymerase chain reaction (PCR) protocol with dual-index paired-end Illumina sequencing (Keady et al., 2021). The PCR amplified the V4-V5 region of the 16S rRNA gene using universal primers 515F-Y (GTGYCAGCMGCCGCGGTAA) and 939R (CTTGTGCGGGCCCCCGTCAATTC) (Muletz Wolz et al., 2018). PCRs were performed in duplicate for each sample, including the negative extraction and PCR controls. To lower the risk of contamination during transit from Namibia to the USA, the PCR was performed at CCF and PCR products shipped on ice to the Center for Conservation Genomics (CCG), Smithsonian National Zoo and Conservation Biology Institute. Duplicate PCR products were combined before being purified with magnetic beads. After the first bead purification, samples were indexed with i5s and i7s, cleaned again, quantified, and pooled as specified in Keady et al. (2021). An agarose gel was run of the 16S rRNA library and the target band (~578 base pairs) was isolated and removed using a QIAquick Gel Extraction Kit (<ns/>28704, Qiagen, MD) and diluted to 4 nM. All samples were sequenced using an Illumina MiSeq (v3 chemistry: 2 x 300 bp kit) at CCG. For our second objective, to avoid any bias from sequencing, all Day 0 (fresh) sample libraries from USA and NAM collections were pooled and re-sequenced together on an Illumina MiSeq (v3 chemistry: 2 x 300 bp kit) at CCG.

Statistical analyses were performed in RStudio for R. Significance for all analyses was set to p < 0.05 and we adjusted p-values for multiple comparisons using Bonferroni. Analysis pipelines for characterizing microbial structure and composition were based on previous research (Muletz-Wolz et al., 2019a, 2019b; Keady et al., 2021; Bragg et al., 2020). For both, comparisons across sampling days and between two populations, we conducted two metrics of microbial diversity which included alpha diversity (within sample variation), beta diversity (between sample variation), and the changes in relative abundance at the ASV and Phyla levels. For comparisons of fresh samples by population we also predicted functional pathways based on marker gene sequences using PICRUSt2 and linear discriminant analyses effect size using LefSe.

We obtained 5,031,818 high-quality sequences from 32 samples ( Table 1 ). Individual NAM 6 was dropped from the degradation study because of low sequence yields in subsequent samples which led to non-consecutive sampling. The average number of sequences per sample was 157,244 (range 71,066 – 330,423). There were 576 ASVs from eight phyla including Firmicutes, Bacteroidetes, Fusobacteria, Actinobacteria, Proteobacteria and to a lesser extent (≤ 2 ASVs) Candidatus Saccharibacteria, Chloroflexi, and Deinococcus Thermus ( Figure 1A ). The seven fresh subsamples (Day 0) were used as the best representatives of the cheetah gut microbiota in our study, and they consisted of 193 ASVs from five Phyla including Firmicutes (41.2% pooled abundance, range across samples 28.4 - 69%, 114 ASVs), Bacteroidetes (28.1%, 3.6 - 46.9%, 34 ASVs), Fusobacteria (16.4%, 7 – 27.7%, 14 ASVs), Actinobacteria (11.4% 4.8-23%, 18 ASVs), and Proteobacteria (2.9%, 0.65 – 7.5%, 13 ASVs). Only the Phylum Actinobacteria was differentially abundant across days, with lower abundance in Day 4 compared to Day 0 (effect size = -1.64, standard error= 0.43, p.adj = 0.029). Within the Phylum Actinobacteria, no families or abundant ASVs were differentially abundant, suggesting that temporal decay is impacting only the higher-level taxonomic distribution and not one particular family within this phylum. Further, across all bacteria and sampling days, we did not detect any differentially abundant taxa at either the ASV or family level.

Bacterial communities remained similar over time, with little variation between fresh fecal samples at Day 0 and subsequent days. Within samples (alpha diversity), both SR ( Supplementary Figure S2A ) and Faith’s PD ( Figure 1B ) remained similar across the fresh and the subsequent sample days ( Table 1 ). Bacterial SR was 162.43 ± 46.71 (mean ± standard deviation) and PD was 7.10 ± 2.08 at Day 0 with minimal changes thereafter (GLMMs, SR: x² = 0.7639, df = 4, p =0.943; PD: x² = 0.958, df = 4, p =0.916). Similarly, microbial composition did not change between samples over time in presence absence measures (PERMANOVA Sample Day: Jaccard Pseudo-F _4,27 = 0.5638, R² = 0.07709, p = 0.064, UniFrac Pseudo-F _4,27 = 0.4034, R² = 0.05639, p = 0.194; Supplementary Figures S2B, C , respectively). Microbial composition did however vary over time for abundance weighted composition (PERMANOVA Sample Day: Bray-Curtis Pseudo-F _4,27 = 0.9866, R² = 0.12752, p = 0.009; Figure 1C ); although in pairwise post hoc analyses, no days were significantly different from each other including all later sample days (Days 1-4) compared to Day 0 (post hoc Bray-Curtis p adj > 0.156). This indicates no two sample day comparisons show large enough differences to be considered significant and the global effect of the PERMANOVA may be driven by subtle differences undetected in the pairwise comparisons. Additionally, pairwise testing by nature uses smaller sample sizes which may limit power. Within group dispersion of community composition was similar across sample day groups (PERMDISP: Bray-Curtis p = 0. 754, Jaccard p = 0.257, unweighted Unifrac p =0.754).

In the comparison of fresh fecal samples from Namibia and the USA, we obtained a total of 50,800 high-quality sequences from 16 samples after rarefying by lowest sequence sum (3,176). There were 226 ASVs from six phyla including Firmicutes, Bacteroidetes, Fusobacteria, Actinobacteria, Proteobacteria, and Candidatus Saccharibacteria ( Figure 2 ). None of the phyla were differentially abundant between the two populations (Wilcoxon rank test; all phyla p.adj > 0.62).

Analysis of microbial community revealed differences in ASV diversity and composition between the two populations. Within group diversity measures were higher in the Namibian samples in both SR and PD (ANOVA; SR: F_1,14 = 18.96, p = 0.0007 and PD: F_1,14 = 29.79, p = 0.00008; Figure 3 ). SR for NAM was 88.4 ± 12.6 vs USA 64.5 ± 9.06 and PD for NAM was 4.97 ± 0.47 vs USA 4.01 ± 0.17. Between group microbial composition also varied by population (PERMANOVA; Bray-Curtis Pseudo-F _1,14 = 5.1036, R² = 0.267, p = 0.002; Jaccard Pseudo-F _1,14 = 7.8492, R² = 0.359, p = 0.001; Unweighted UniFrac Pseudo-F _1,14 = 8.019, R² = 0.364, p = 0.002; Figure 4 ). Dispersion of the samples within a group (Population) around the centroid were determined to be similar (p > 0.05) across populations (PERMDISP: Bray-Curtis p = 0. 498, Jaccard p = 0.397, unweighted Unifrac p =0.465).

Overall, NAM samples had 104 ASVs not found in the USA samples, corresponding to 46% of the total number of unique 226 ASVs, while the USA had 54 unique ASVs (24%). The two populations shared 68 out of the total 226 ASVs (30%). When ASVs were weighted by abundance (number of sequences per ASV, where more abundant ASVs carry more weight), the proportion of total sequences (50,800 sequences) shared between the two populations increased to 69% (35,052 sequences from 68 ASVs). Whereas 20% of sequences (10,160 sequences from 104 ASVs) were unique to NAM samples alone, only 11% (5,588 sequences from 54 ASVs) were unique to USA samples.

Using LefSe LDA we found 26 specific taxa that were more common in either the NAM (21 taxa) or USA cheetahs (5 taxa; Figure 5A ). Hierarchical clustering of samples based on abundance of these 26 taxa cleanly delineates NAM versus USA samples ( Figure 5B ). Taxa that were enriched in the Namibian samples belonged to three phyla, Actinobacteria (including Coriobacteriaceae Collinsella), Bacteroidetes (including Porphyromonadacea Barnesiella and Prevotellacea Alloprevotella), and Firmicutes (including Erysipelotrichaceae Faecalitalea, Lachnospiracea Ruminococcus2, Peptococcaceae Peptococcus, and Ruminococcacea Clostridium IV). The five taxa enriched in USA cheetahs were also found in the NAM cheetahs, but in lower amounts. By contrast, eight of the taxa present in the NAM cheetahs were missing completely from all of the USA samples ( Figure 5B ). These eight taxa were all members of Bacteroidetes (Prevotellaceae, Alloprevotella, Porphyromonadaceae, Barnesiella, and two unknown species in the Alloprevotella and Barnesiella genera) or Firmicutes (Faecalitalea and an unknown species in the Faecalitalea genus).

When looking at predictive functional analyses, metabolism was the highest predicted Level 1 KEGG Orthology (KO) category function of the microbiome, using PICRUST2 analyses in both populations (NAM 78.4% ± 0.01% and USA 78.6% ± 0.2%). None of the Level 1 KO categories differed between the two populations. For Level 2 KO category comparisons, USA samples had higher proportions of sequences that approached significance (p.adj <0.1) and were associated with lipid metabolism, parasitic infectious disease, and carbohydrate metabolism ( Supplementary Figure S3 ; Wilcoxon test with Benjamini Hochberg corrected p-values; lipid metabolism: effect = 1.29, overlap = 0.06, p.adj = 0.054; Parasitic infectious disease: effect = 1.06, overlap = 0.13, p.adj = 0.076; Carbohydrate metabolism: effect = 1.04, overlap = 0.11, p.adj = 0.09). However, the total proportion of sequences attributed to these categories for both groups was small. Among the most specific categories, Level 3, there were no pathways that differed between the two cheetah populations (see Supplementary Materials for model outputs for all three levels).

In our first aim, we found that after three days of exposure to the dry Namibian environment, cheetah fecals were still similar in microbial diversity, composition, and structure to fresh fecals. However, Actinobacteria relative abundance began to change by Day 4. These patterns of compositional stability are similar to domestic cat samples in tubes at room temperature (Tal et al., 2017) but longer than previously reported in ex situ cheetah fecals sampled during a hot (max daily temperatures 27-33°C, Supplementary Table S2 ) and wet (max cumulative rainfall over a five day study period 2.1 cm, average max daily humidity 98.7%) summer season in Virginia, USA (Maly et al., 2024). In our previous US-based cheetah study, there were shifts in bacterial composition in abundance weighted and presence-absence metrics that occurred by Day 2 post-deposit. There was no precipitation during the sampling period for the current study as opposed to the US based study which experienced heavy rain during temporal sampling. The arid (0 cm of rainfall, average max daily humidity 40.9%) climate during the dry season in Namibia may aid in stabilizing the post-defecation shifts of the fecal microbiota. Similarly, giraffe fecal microbiota composition was stable in the dry Namibian environment until rain occurred before day five sampling in the series (Menke et al., 2015). It is important to note that fecals left in the experiment plot were covered with screens to prevent the fecals from being taken by larger insects (e.g., dung beetles), small animals, etc., which may have offered additional protection against microbial alterations from these or other large insects and animals (Wong et al., 2016). However, by collecting the middle core of the fecal not exposed to the environment directly, we hoped to mitigate these caveats as much as possible. Our sample size was limited due to the nature of working with non-model endangered animals and while our study may benefit from increased sample size to account for potential variation among individuals, our sample size was similar to studies of a similar nature (Menke et al., 2015; Tal et al., 2017; Lafferty et al., 2022; Wong et al., 2016). Together, our data suggest that researchers sampling cheetahs in Namibia during the dry season can collect up to 3-day-old fecal samples, as a proxy for studying gut microbiota. If collecting during a time of higher moisture (e.g., rain), however, based on our previous work, it is likely the collection window is closer to around 24 hours (Maly et al., 2024).

Physical characteristics of the NAM samples over time were similar to those previously reported in ex situ US samples (Maly et al., 2024). In brief, fresh samples were wet and shiny, very dense, pungent, and exhibited frequent insect activity (see Supplementary Figures ). By Day 1, a crust started to form but the outside was still a bit tacky and the inside of the fecal still contained some soft feces and moisture. Day 2 feces had well-formed crusts that appeared dark and were beginning to dry out, even on the inside. Days 3 and 4, most fecals, especially those that were not well-formed logs, were very dark and very dry.

Finally, our study timeline terminated after four days post fresh, but it may be useful to sample for an extended period to identify changes beyond four days to aid in assessing the age of fecals from the field when the deposit event was not witnessed. We were unable to determine fecal degradation day age indicators with the current study because there were so few microbial differences across four days.

We compared fresh samples of the above Namibian project with fresh samples at a US facility from a previous study (Maly et al., 2024). NAM and USA samples consisted of the same main five Phyla previously reported for cheetahs (Wasimuddin et al., 2017; Menke et al., 2014; Maly et al., 2024). All five Phyla were similar in abundance across both populations, exhibiting greater proportions of Firmicutes compared to Bacteroidetes. Though abundance data are not directly comparable between studies, it is noteworthy that the ex situ NAM and USA cheetahs in the current study had numerically lower abundances of Firmicutes [41.2% (time dataset NAM]/40.4% (combined dataset NAM) and 48.0% USA] compared to the wild cheetahs [56.2% (Menke et al., 2014) and 68.5% (Wasimuddin et al., 2017)] which corresponds to a trend seen in a study on bobcats, where those in zoos trended toward lower Firmicute abundance (p < 0.1) compared to wild bobcats (Eshar et al., 2019). Research in domestic cats shows Firmicutes were less abundant in obese compared to healthy weight individuals (Fischer et al., 2017; Ma et al., 2022). The individuals included in our study were not obese, however it is possible that ex situ cheetahs are less lean than their wild counterparts which may explain the lower Firmicute abundance, although other factors may be involved.

We discovered a high degree of variation in the fecal microbial diversity, composition, and structure between the two populations. We can assume the higher diversity of ex situ NAM fresh samples may be due to a variety of factors, including cheetah diet, demographics (age, sex, reproductive status), and environmental factors (climate and location). Since the scope of this study was not to define the effects of these differences, we did not control for them; however, some information appears relevant. Differences in food structure (minced vs whole prey mice) were responsible for significant differences in fermentation profiles in domestic cats, however they did not result in different alpha diversity estimates (D’Hooghe et al., 2024). The diet between the two facilities differed, with primarily donkey meat on the bone for the NAM vs commercial ground beef as the primary diet for USA; however, USA cheetahs were not fed solely minced commercial diet as they were given weekly whole prey items and bones. There is evidence that carcass diets contain higher amounts of non-digestible elements which obligate carnivores such as the cheetah, may have evolved to utilize as a source of dietary fiber (Depauw et al., 2013, 2012), and which may alter the gut microbiomes. Other important aspects that may influence the fecal microbiota include cheetah demographic data such as age (ranges, NAM: 3-14.5 years; USA: 1-6 years (Wasimuddin et al., 2017; Masuoka et al., 2017; Rojas et al., 2023), and sex (Wasimuddin et al., 2017)). Most evident of all, the two populations lived on different continents and the cheetahs and voided fecals were therefore exposed to different environmental conditions including but not limited to weather (Maly et al., 2024; Menke et al., 2015) and habitat (Gani et al., 2024) that likely also influenced the microbiota. Thus, a number of variables could be responsible for the differences we find between populations, and future studies comparing many populations are needed to determine their relative impact on cheetah microbiota.

Bacterial communities differed between the two populations in all measures of diversity and composition examined, with 26 taxa that were enriched in either NAM or USA samples. Interestingly, eight taxa were enriched in NAM samples but absent in the USA samples. A few of these taxa belong to the family Prevotellaceae, some of which are known carbohydrate and protein fermenters in humans (De Filippo et al., 2010; Aguirre et al., 2016). Prevotellaceae are part of the healthy domestic cat microbiome (Ganz et al., 2022). Domestic cats fed whole mice had enriched Prevotellaceae compared to those fed minced mice and were associated with food structure differences between minced and whole prey (D’Hooghe et al., 2024). Alloprevotella, a genus within Prevotellaceae, were enriched in NAM samples and absent in USA samples, and some members of this genus can hydrolyze gelatin from collagen and produce short chain fatty acid acetic acid and succinic acid (which can be metabolized into another SCFA, propionic acid) (Downes et al., 2013; Leaver et al., 1956). Another taxon enriched in the Namibian samples, Barnesiella, has been associated with reduced inflammation in asthma (Zhang et al., 2021) and allergic reactions (Vital et al., 2015). Interestingly, Barnesiella was also found to restrict the growth of an antibiotic-resistant bacteria in mice and humans (Ubeda et al., 2013). The presence of these microbes may play a role in the fewer incidences of symptoms related to GI inflammation (including gastritis) reported in these Namibian cheetahs Mangiaterra et al. (2022), but further research is needed to address their roles in the cheetah GI tract. Interestingly, we found 68 bacterial ASVs (30% of the microbiome) that were shared between the two populations, suggesting that while environment, diet, and demographics may impact the gut microbiota, there is still a conserved cheetah microbiome that likely plays critical functions in cheetah physiology regardless of associated extrinsic and intrinsic factors. Of those shared, ASV1, a Fusobacterium, was the most abundant and is an anaerobic protein fermenter (Mead, 1971) commonly found in domestic cat microbiomes, especially those consuming raw meat diets (Butowski et al., 2019).

PICRUSt2 analyses showed approximately 78% of all sequences in both populations were associated with metabolism. This was expected given the microbes reside in the gut where host and microbial metabolic processes take place to utilize digestive material. Across all levels and categories, only three KO pathways (all Level 2) approached significance between the two populations ( Supplementary Tables 3-5 ). Two of these pathways were types of metabolism. Carbohydrate metabolism is likely higher in the USA samples due to the beet pulp in the commercial ground beef diet. Lipid metabolism was also higher in the USA samples. In domestic cats, multiple types of Level 3 lipid metabolism pathways were higher in cats with acute diarrhea (Bai et al., 2023). Two cheetahs in the USA population were experiencing acute diarrhea at the time of collection, which may at least partially explain these differences. Lastly, the finding that the USA cheetahs had higher proportions of bacterial pathways related to parasite infection was unexpected. The USA cheetahs received monthly parasitic preventatives including Ivermectin for helminths and Frontline (Boehringer Ingelheim, Ingelheim, Germany) for ticks and fleas, so their parasite load should be relatively low. In contrast, NAM cheetahs were regularly monitored for parasites and prescribed treatment only when necessary. It is important to note, that these sequences are based on predicted gene content of the bacteria found in the samples from the two populations and not direct sampling for parasites. These differences may be due to the presence of bacteria that simply carry genes that are known to be involved in pathways of parasitic infectious disease but are not a direct indication of parasite presence. In general, these PICRUSt2 data provide an estimation of the bacterial functional production in the cheetah gut microbiome.

Based on these data, we recommend future studies focus on the microbiome, transcriptome, and metabolome of managed and wild cheetahs to better understand their relationship with host genetics and physiology, diet, and environment. Our main recommendation is to focus on the relationship with diet, particularly in managed cheetahs. We suggest studies to manipulate the cheetah diet in a controlled setting to determine whether the inclusion of non-digestible animal fibers alter the microbiome and metabolome, and at what proportion these dietary fibers produce changes in the microbiota. There are multiple prey species options for carnivores in managed care, but in many cases single-species origin diets are standard. It would be of interest to know if the diversity of animal species in the diet influences the microbiome, metabolome, or susceptibility to GI disease. Additionally, to improve our knowledge on the etiology of GI disease in cheetahs, we recommend an initial direct comparison of cheetahs with and without chronic gastritis. Further, we suggest controlled longitudinal studies on cheetahs starting at cub stage and following them over the course of their life. If animals in the longitudinal studies are fed diets with different levels of non-digestible animal fiber, they should help determine if fiber levels lead to differences in the microbiome and metabolome that may in turn offer protective effects against GI disease.

In conclusion, our data suggest cheetah fecals from the dry season (no rain) in Namibia are acceptable for microbiota collection for up to three days post defecation. This is longer than in our previous study where cheetah fecals exposed to rain and more humid conditions were only stable for 24 hours. These data suggest moisture is an important component to consider for fecal collections when utilized as a proxy for the gut microbiome. We hypothesize these patterns will apply to other members of the Felidae family, and perhaps even other large obligate carnivores under the same climatic conditions, but we recommend additional time series studies to confirm these predictions in new species or environments. These data will provide greater access to cheetah and other large carnivore fecals and reduce costs for sampling efforts, offering an extended period for non-invasive fecal collection. In the second aim we identified differences in fecal microbiota diversity, composition, enriched taxa, and predictive functional relevance between two ex situ cheetah populations. While we were unable to control for or determine which factors were responsible, it does provide evidence that warrants further investigation into managed cheetah gut microbiomes and diet. Many of the differentially abundant taxa between the two populations have known clinical relevance in the gastrointestinal tract, being key taxa to examine in relation to GI health that is negatively influencing captive cheetah husbandry and breeding success. We recommend future studies to assess the functional effects of ex situ diet type on cheetah microbiomes and metabolomes to improve cheetah welfare and breeding program success.