Abstract
Objective
Gut health and its relationship to gut microbiota is an important consideration in the care and well-being of managed endangered species, such as the cheetah (Acinonyx jubatus). Non-invasive fecal sampling as a proxy for gut microbiota is preferred and collecting fresh fecal samples is the current gold standard. Unfortunately, even in managed facilities, collecting fresh samples from difficult to observe or dangerous animals is challenging. Therefore, we conducted a study to determine the terminal collection timepoint for fecal microbial studies in the cheetah.
Methods
We longitudinally sampled eight freshly deposited fecals every 24 h for 5 days and assessed bacterial relative abundance, diversity, and composition changes over time.
Results
Our data indicated that fecal samples up to 24 h post defecation provided accurate representations of the fresh fecal microbiome. After 24 h, major changes in community composition began to occur. By 72 h, individual cheetah fecal microbiota signatures were lost.
Conclusion
Our findings suggest that cheetah fecal samples should be collected within 24 h of defecation in humid environments, especially if precipitation occurs, in order to provide a more biologically accurate representation of the gut microbiome, and we provide visual characteristics that can aid researchers in approximating time since defecation.
Significance statement
Data from this study provides guidelines for researchers investigating cheetah and other large felids and carnivores where the ability to collect fresh fecal deposits is limited.
Introduction
The role of the gut microbiome in animal health has been of increasing interest in wildlife medicine (Sugden et al. 2020, Gillman et al. 2022) and conservation (Redford et al. 2012, Bahrndorff et al. 2016). Understanding the native microbiomes of wildlife is essential to identify crucial taxa necessary for metabolic functions and overall health (Amato 2013). This understanding is particularly relevant for rare and endangered species for which the health of each individual may be critical for population viability. A major challenge for examining gut microbiomes in these animals is the ability to collect non-invasive samples, such as fresh feces, to best represent the gastrointestinal tract. For fecal microbiome studies, the gold standard is to collect fresh fecal samples and immediately freeze them at −80˚C when possible (Fouhy et al. 2015, Choo et al. 2015, Gorzelak et al. 2015). While fresh samples are optimum for fecal sampling, they are often difficult to collect due to challenges posed by defecation timing and frequency, factors that are difficult to predict without continual observation of animals, which is impossible in most free-ranging populations.
Investigations into the temporal stability of the fecal microbiome after defecation are necessary to determine the time window for the most accurate representation of the intestinal microbiome. Time series studies where freshly voided samples are experimentally subsampled over time have become more prevalent in a wide range of species including domestic cats (Tal et al. 2017), minks (Lafferty et al. 2022), cattle (Wong et al. 2016), horses (Beckers et al. 2017), and giraffes and springboks (Menke et al. 2015). With the exception of the domestic cats (Tal et al. 2017), these studies identified shifts in microbial diversity or community composition over sampling time points within 5 days or less (Menke et al. 2015, Wong et al. 2016, Beckers et al. 2017, Lafferty et al. 2022). Feces for captive wildlife studies are often collected during morning routine enclosure cleaning which ensures the samples are collected within 24 h of defecation. However, consistent fresh sampling is not feasible in the wild without intensive effort, and even then, may still be difficult in certain species. Fresh sampling is also difficult for managed animals that are dangerous and/or sensitive to human disturbance in facilities. Identifying a time point beyond which the fecal microbiome shifts and no longer represents the host gut microbiome is key. This will ensure that the samples collected are most appropriate to form valid conclusions regarding the microbial ecology of fecal samples.
Here, we studied the temporal stability of microbiota in cheetah feces over time. Previous research shows that between 40% and 60% of cheetahs under human care suffer from gastrointestinal diseases, such as chronic gastritis, that lead to death or euthanasia (Munson 1993, Munson et al. 1999, Munson et al. 2005, Terio et al. 2018). Due to the prevalence of these diseases, there is a growing interest to better characterize and understand the role of the cheetah gut microbiome in overall individual health. While previous cheetah gut microbiome studies utilize rectal swabs (Becker et al. 2014, Menke et al. 2014, Becker et al. 2015, Wasimuddin et al. 2017), fecal samples are easier to obtain and are non-invasive. However, there are no data on the fecal microbial stability of non-domestic felids, including cheetahs. In domestic cats, fecal microbiota did not change in any community richness or composition measures examined when comparing fresh frozen samples to those stored at room temperature for up to 4 days (Tal et al. 2017). However, as the domestic cat samples were in tubes and not exposed to a natural sampling environment, those data may not be a good predictor of non-domestic felid sampling in the wild.
In this study, we examined the stability of the fecal microbiome in ex situ cheetahs in a US-based facility. Our objective was to determine how many days post defecation the fecal microbiome remained representative of the microbiome of a fresh sample. The goal was to establish timelines to aid the design, implementation, and interpretation of future cheetah, felid, and other large carnivore gut microbiome studies both ex situ (in managed populations) and in situ (from field sampling of wild populations). With ex situ sampling, it is easier to collect a sample within 24 h after deposit; however, it still requires a great deal of time and effort to acquire a fresh (<1 h) sample. For ex situ cheetah research, it is important to know whether the morning fecal sample collection is representative enough of a fresh sample whereas, for in situ sampling, it is even more difficult to find and collect fresh samples. Often, it is more likely to find deposited feces and use some general characteristics to approximate fecal age (color, moisture, texture, mucous layer, insect activity, etc.) (Nardi et al. 2022). Identifying the best time window for fecal collections that can serve as a proxy for gut microbiome structure will strengthen our ability to make inferences about the role of the gut microbiome in cheetah health, both ex situ and in situ.
Materials and methods
Sample collection
Sample names were given to each cheetah fecal as Aju, short for A. jubatus. Eight fresh fecals were collected from two adults (individuals Aju 2 (5.6 years old) and Aju 6 (6 years old) and six juveniles (individuals Aju 1, 3, 4, 5, 7, and 8; all age ~1 year) living at the Smithsonian National Zoo & Conservation Biology Institute (NZCBI) in Front Royal, VA. Smithsonian NZCBI met required animal management and husbandry guidelines put forth by the Cheetah Species Survival Plan (Plan 2009). The cheetahs were housed in 2000 m2 enclosures with free access to both indoor and outdoor areas. Cheetahs were fed a commercial carnivore beef-based diet (Nebraska Premium Canine Diet, North Platte, NE, USA) and were also given one whole rabbit (1× weekly) and horse neck bones (2× weekly). Water was available ad libitum. Fresh fecal samples were collected in June and July between 08:30 h and 10:30 h from cheetah habitats within 30 min of deposit (all defecation events were witnessed by author MM), over the course of 20 days. The fresh fecal samples were placed in an experimental plot outside of but adjacent to the cheetah enclosures with full environmental exposures (in the grass with no cover from sun or weather) to allow for ease of access. A piece (~2.5 cm in length) was removed from one end of the fecal sample and cut in half with a sterile scalpel (Surgical Design Disposable Scalpels #20, Fisher Scientific) (Fig. 1). A subsample was taken from the core of the fecal piece with sterile forceps (Sklar Instruments Econo Sterile Thumb Dressing Forceps, Fisher Scientific) and placed into a sterile DNase/RNase-free 1.5 mL tube (Eppendorf). This core sample was considered day 0. The remaining feces were placed back in the experimental plot. We aimed to mimic realistic sample collection opportunities, whereby fecal is collected in the morning during routine enclosure cleaning. If enclosure cleaning is done daily and fecal is found, it is likely to have been deposited at most, 24 h prior. Therefore, 24 h after the fresh collection at day 0, a new 2.5 cm section was removed from the feces and another subsample was taken from its core and placed into a new sterile 1.5 mL tube. This process continued for each fecal sample for 5 days (days 1–5) or until no feces remained, after the initial fresh collection. The collection date, maximum daily rainfall (cm), temperature (˚C), and humidity (%) were recorded for all subsamples from each of the eight fecals (Supplementary Table 1, see section on supplementary materials given at the end of this article). Maximum temperature and humidity levels did not vary across days, so only rainfall was analyzed. Because feces were exposed until collection, a later subsample (e.g. day 4) would have experienced the total rain amount from all prior days. Therefore, we calculated the cumulative value of the maximum daily rainfall for each subsample. For example, if on day 1 and day 3, we recorded a maximum rainfall of 0.23 cm and 0.46 cm, the maximum daily cumulative rainfall (MDC rainfall) amount for day 4 would be 0.69 cm. Upon collection, subsamples were placed immediately on ice and transported from the experimental plot to a −80°C freezer until processing.
Graphical representation of fecal collection methodology. Samples were collected within 30 min of deposit and transferred to an adjacent experimental plot. After transfer, the day 0 (fresh) sample was collected. A piece of the fecal was removed and cut in half with sterile scalpels. The center of the fecal half was removed with sterile forceps and placed in a labeled tube for storage at −80°C until processing. The internal sample from the second half of the fecal piece can also be stored in case more sample is needed. Maximum daily weather variables were collected after the first 24 h and for every 24 h after. This was repeated for 5 days or until there was no more feces left to sample.
Citation: Microbiota and Host 2, 1; 10.1530/MAH-23-0022
Sample DNA extraction and library prep
DNA was extracted from 0.25 g frozen core feces using the QIAamp PowerFecal DNA Kit (#12530-50, Qiagen) following the manufacturer’s instructions. For each batch of sample extractions, an empty negative control was included to identify extraction contaminants. Following extractions, DNA concentrations were measured via fluorometric quantification (Qubit4 Fluorometer, ThermoFisher Scientific).
Fecal bacterial DNA was amplified following a previously published two-step PCR protocol with dual-index paired-end Illumina sequencing (Keady et al. 2021). The amplicon PCR reaction amplified the V3–V5 region of the 16S rRNA gene using universal primers 515F-Y (5′-GTGYCAGCMGCCGCGGTAA-3′) and 939R (5′-CTTGTGCGGGCCCCCGTCAATTC-3′) (Muletz Wolz et al. 2018). Amplicon PCR reactions for each sample were performed in duplicate, including the negative extraction and PCR controls. The amplicon duplicates were pooled and then purified with magnetic beads, indexed with i5s and i7s, cleaned again, quantified (Qubit4), and pooled as specified in Keady et al. (2021). The target band for the 16S rRNA library (~578 bp) was isolated and removed using a QIAquick Gel Extraction Kit (#28704, Qiagen) and diluted to 4 nM. Samples were loaded onto the flow cell at 8 pM with 25% PhiX and sequenced using an Illumina MiSeq (v3 chemistry: 2 × 300 bp kit) at the Center for Conservation Genomics, Smithsonian National Zoo & Conservation Biology Institute.
Sequence data processing
Demultiplexed Illumina Miseq sequencing reads were imported into R version 4.0.3 (R CoreTeam 2022b) using RStudio (v 2022.12.0+353). We utilized R package ‘dada2’ version 1.16.0 (Callahan et al. 2016, Callahan et al. 2017) to merge paired ends, remove chimeras, and filter out low-quality reads (maxEE > 2). Filtered and merged sequences from two sequencing runs were combined to generate amplicon sequence variants (ASVs) and assign taxonomy using the Ribosomal Database Project (RDP 16S training set (set 16, release 11.5) (Cole et al. 2007, Wang et al. 2007). A phylogenetic tree was built using Quantitative Insights Into Microbial Ecology 2 (vQIIME2-2020.8) (Bolyen et al. 2019) using FastTree (Price et al. 2009). We imported the ASVs, taxonomy table, phylogenetic tree, and metadata into a phyloseq object (McMurdie & Holmes 2013) for processing. Putative contaminant sequences (n = 3) were removed using the combined Fisher method with a threshold of 0.1 in the R package ‘decontam’ (v 1.18.0) (Davis et al. 2018). After contaminants were removed, we filtered out singleton ASVs (ASVs that occur in only one sample), ASVs classified as Cyanobacteria (photosynthetic bacteria found in algae), negative control samples, and low sequence count (<8000 reads) samples. After quality control and filtering, the sequencing depth variation (maximum/minimum) was a 4.89-fold change (maximum = 42,284; minimum = 8948), and therefore, based on current recommendations (Weiss et al. 2017), samples were not rarefied. Seven subsamples were removed due to low sequence coverage (empty boxes in Supplementary Table 1).
Statistical analysis
Statistical analyses were performed in RStudio version 2022.12.0+353 (R CoreTeam 2022a ) for R (v4.2.2) (R CoreTeam 2022b ). Significance for all analyses was set to P < 0.05. Analysis pipelines for characterizing microbial structure and composition were based on previous research in our laboratory (Muletz-Wolz et al. 2017, Muletz-Wolz et al. 2019, Bragg et al. 2020, Keady et al. 2021). For comparisons across sampling days, we calculated two metrics of microbial diversity, which included alpha (within sample variation) and beta diversity (between sample variation), and assessed changes in relative abundance of ASVs.
Differential abundance of phyla and ASVs across sample days
Differential abundance testing was performed using package ‘DAtest’ to identify the most appropriate test given our dataset (Russel et al. 2018). At the ASV level, the phyloseq object was prefiltered to reduce the false discovery rate (min.samples = 5, min.reads = 10), resulting in 160 taxa to be tested. The function testDA() was performed on the prefiltered ASV phyloseq object to identify the most appropriate differential test based on our dataset, where the differential abundance was measured between sample days and paired by fecal ID. The top-ranked three tests were linear model (lmc), log-linear reg.2 (llm2), and log-linear reg (llm). After running the top three tests, ASVs were selected if they appeared in two of the three test results. We calculated post hoc comparisons using lsmeans() between all days, and the output was subset to only include the comparisons between fresh (day 0) samples and all other days (e.g. day 0 vs day 1, day 0 vs day 2, etc.) for further analysis. ASVs that were differentially abundant (after FDR correction) between fresh and all other days were visualized using a heatmap. Heatmaps were created using the plot_heatmap() function in the R package ‘phyloseq’ (McMurdie & Holmes 2013). Differential Phyla were calculated, analyzed, and visualized similarly to the ASV level, except the tax_glom() function was used to merge all taxa of the same Phylum before testing, and there was no prefiltering of the original phyloseq object before running the differential tests.
Influence of sample day on fecal microbiota
Alpha diversity metrics included species richness (SR) and Faith’s phylogenetic diversity (PD) over time (sample day) and by individual fecal (fecal ID). Species richness is the number of unique ASVs in a sample, and PD measures the amount of biodiversity based on the phylogenetic relationships of the taxa and the total tree branch length of ASVs in a sample (Faith 1992). Faith’s PD was calculated for each subsample with the R package ‘picante’ (Kembel et al. 2010). Using the ‘stats’ R package (Version 4.2.1) (R CoreTeam 2022b ), we performed linear models, with SR or PD as the response variable and sample day, MDC rainfall, and fecal ID as fixed effects. Fecal ID was added as a fixed effect because we were interested in how long we could detect individual differences. SR and PD distributions met assumptions of normality (Shapiro–Wilk) and homoscedasticity (Levene). Post hoc tests were assessed using estimated marginal means ± s.d. with ‘emmeans’ package (Lenth 2023) in R with Tukey P adjustments.
To identify shifts in community composition between subsamples across sampling days, we measured Bray–Curtis (abundance-weighted composition), Jaccard (presence–absence composition), and unweighted UniFrac (presence–absence with inclusion of phylogenetic relationships of taxa) distances. We used PERMANOVAs (Anderson 2017) in the package ‘vegan’ (Oksanen et al. 2022) where Bray–Curtis, Jaccard, and unweighted UniFrac distances were the response variables, and sample day, MDC rainfall, and fecal ID were the explanatory variables. Post hoc analyses were performed in the package ‘pairwiseAdonis’ (Martinez Arbizu 2020) using the pairwise.adonis2() function, and P-values were adjusted using the Bonferroni correction. To identify whether dispersion of community composition differed between sample days, we used PERMDISP from the package ‘vegan’, specially thebetadisper() function (Oksanen et al. 2022). The Bray–Curtis and unweighted UniFrac distance values between samples were also used in a sample-wise microbial network to identify connections between samples. We created an igraph microbiome network using the make_network() and plot_network() functions in the ‘phyloseq’ package. The maximum ecological distance between two samples was set to the default value of 0.7, which allowed all samples to appear in the network but minimized over-connectedness.
Results
Relative and differential abundance of phyla and ASVs across sample days
We obtained 984,381 high-quality sequences from 41 samples. The average number of sequences per sample was 24,009 (range 8648–42,284) (Supplementary Table 2). We found 268 ASVs from five phyla including Firmicutes, Proteobacteria, Bacteroidetes, Fusobacteria, and Actinobacteria (Supplementary Fig. 1). Fresh subsamples (day 0), considered the most representative of the cheetah gut microbiome, were dominated by Firmicutes (63% mean relative abundance, 83 ASVs), Bacteroidetes (18%, 14 ASVs), Fusobacteria (13%, 8 ASVs), and with Actinobacteria (3.5%, 6 ASVs), and Proteobacteria (2.5%, 7 ASVs) present in lower amounts.
We identified three phyla (Fusobacteria, Proteobacteria, and Firmicutes) and 12 ASVs that were differentially abundant in at least one of the subsequent days (days 1–5) compared to day 0 (Fig. 2). Fusobacteria were lower in relative abundance on days 3–5 compared to day 0, Firmicutes differed between day 0 and day 3, and Proteobacteria had lower abundance on day 0 compared to days 3 and 4 (Fig. 2). Twelve bacterial ASVs were differentially abundant (Fig. 3, Supplementary Table 3), with two Bacillus ASVs lowest at day 0 then increased with sample day, while all other differential ASVs, four Bacteroides, one Eisenbergiella, one Clostridium XIVa, and four Fusobacterium ASVs were highest on day 0 and decreased with sample day (Fig. 3).
Box plots of differentially abundant phyla across cheetah fecal sample days. Red asterisks indicate which days are different from day 0 relative abundance (P < 0.05 in at least two of the following: linear model, log linear reg.2, and log linear regression).
Citation: Microbiota and Host 2, 1; 10.1530/MAH-23-0022
Heat map of differentially abundant ASVs across cheetah fecal sample day. Differentially abundant ASVs are listed by genus and ASV number in rows and the columns are samples from each fecal ID grouped by sample day. In order, they belong to the following phyla, Fusobacteria (Fusobacterium), Firmicutes (Eisenbergiella, Clostridium XIV, Bacillus), and Bacteroidetes (Bacteroides). Color indicates abundance of the ASV in each sample from lowest (yellow) to highest (red) abundance, where gray indicates absence.
Citation: Microbiota and Host 2, 1; 10.1530/MAH-23-0022
Influence of sample day on fecal microbiota
Over time, bacterial communities in cheetah feces remained similar in richness and diversity but differed in composition. MDC rainfall was a contributing factor to alpha diversity measures but not beta diversity. Among individuals, bacterial communities differed in richness over time, and composition differed until day 3 when individual identities could no longer be distinguished.
For alpha diversity measures, SR and PD were similar among sample days, but they differed by MDC rainfall and among individuals (SR sample day linear model (LM): F5,27 = 1.751, P = 0.156; PD sample day LM: F5,27 = 1.535, P = 0.211; SR MDC rainfall LM: F1,27 = 9.213, P = 0.005; PD MDC rainfall LM: F1,27 = 11.004, P = 0.003; SR fecal ID LM: F7,27 = 3.709, P = 0.006; PD fecal ID LM: F7,27 = 3.78, P = 0.005; Supplementary Fig. 2). At day 0, bacterial species richness was 74.38 ± 14.54 and phylogenetic diversity was 4.25 ± 0.36, and this remained similar over time. However, for every centimeter increase in MDC rainfall, there was a 22 ASV decrease for SR and 0.9 decrease in sum of tree branch lengths in PD. Post hoc analyses of alpha diversity among individuals showed that the two adult cheetahs Aju 2 (estimated mean ± s.d. = SR = 76.6 ± 5.27, PD = 4.37 ± 0.204) and Aju 6 (SR = 81.3 ± 5.12, PD = 4.40 ± 0.198) consistently had greater SR and PD than Aju 3 (47.2 ± 5.61; 3.05 ± 0.217) and greater SR than Aju 7 (42.9 ± 6.83) and greater PD than Aju 5 (3.36 ± 0.214) (Tukey Padj for all ≤ 0.04). For beta diversity measures, microbial composition differed among both sample days and fecal IDs but not by MCD rainfall (PERMANOVA sample day: Bray–Curtis Pseudo: F5,27 = 2.94, R2 = 0.182, P = 0.001; Jaccard Pseudo: F5,27 = 2.37, R2 = 0.170, P = 0.001; unweighted UniFrac Pseudo: F5,27 = 2.22, R2 = 0.175, P = 0.003; Fig. 4) (PERMANOVA fecal ID: Bray–Curtis Pseudo: F7,28 = 5.38, R2 = 0.466, P = 0.001; Jaccard Pseudo: F7,28 = 4.27, R2 = 0.429, P = 0.001; unweighted UniFrac Pseudo: F7,27 = 3.5, R2 = 0.386, P = 0.001; Fig. 4). We focused exclusively on post hoc comparisons of fresh (day 0) vs all later sampling days as our aim was to determine if fecal samples collected later resembled fresh samples. Community composition on day 1 remained similar to fresh samples (pairwiseAdonis: Bray–Curtis, Jaccard, unweighted UniFrac Padj≥ 0.05). By day 2 composition shifted and remained distinct from day 0 fresh samples for the rest of the experiment until day 5 (pairwiseAdonis: Bray–Curtis, Jaccard, unweighted UniFrac Padj ≤ 0.04). Community composition differed among individuals on days 1 and 2 (pairwiseAdonis: Bray–Curtis, Jaccard, unweighted UniFrac Padj ≤ 0.005), but by day 3 individual fecal ID no longer influenced community composition (pairwiseAdonis: Bray–Curtis, Jaccard, unweighted UniFrac Padj ≥ 0.09). By day 1, dispersion in community composition began to increase in all measures (PERMDISP permuted P values: Bray–Curtis P = 0.04, Jaccard P = 0.03, unweighted Unifrac P = 0.03), and the communities remained more dispersed in most measures thereafter (Supplementary Fig. 3).
Principle coordinate analyses (PCoA) of Bray–Curtis distances for all cheetah fecal samples color coded by (A) sample day, where color indicates sample day (80% confidence ellipses by sample day) and (B) fecal ID, where color indicates fecal IDs (80% confidence ellipses by fecal ID, Aju 8 only had three samples, so no ellipse was drawn). Aju 2 and Aju 6 are adults, and the others are juveniles. Aju 7 and Aju 8 were experiencing loose stools and the others had normal fecal consistency.
Citation: Microbiota and Host 2, 1; 10.1530/MAH-23-0022
We built microbial networks based on distances calculated using Bray–Curtis (Fig. 5) measures to better visualize differences over time. Day 0 (fresh) samples are closely connected in the center of the network, except for samples from Aju 7 and Aju 8. Samples from fecal ID Aju 7 and Aju 8 were loose stool samples (compared to normal-shaped fecal samples for all others) and were generally well connected to each other. Fecal ID Aju 2 and Aju 6, the only adults, exhibit multiple close connections between sample days within their own series. However, for samples Aju 1, Aju 3, and Aju 5, the subsequent post-fresh (days 1–5) samples cluster together but are unconnected to their respective fresh samples (day 0s).
Microbiota network showing connections between samples (nodes) using Bray–Curtis dissimilarity. Samples are identified by sample day (color) and fecal ID (shape). Lines connect samples that have a pairwise Bray−Curtis less than 0.7. The length of the connecting line and the distance between two samples indicate the strength of the association, where shorter lines and distances are stronger associations.
Citation: Microbiota and Host 2, 1; 10.1530/MAH-23-0022
Discussion
The gut microbiome of wildlife species is important for their health and survival. Fecal samples are frequently used as a proxy for the gut microbiome as they are non-invasive and easier to collect. Here, we performed time series analyses by subsampling freshly deposited cheetah fecals (day 0) left under natural conditions for up to 5 days post defecation (days 1–5). We investigated the effects of time (sample day) and individual (fecal ID) on microbial relative abundances, diversity, community composition, and dispersion. We also investigated how an environmental factor, rain, influenced the microbiota. We found that overall, fecal microbiota species richness and diversity remained similar over time, but that community composition and relative abundance of particular phyla and ASVs differed as early as day 1 from fresh fecal samples. In general, our data suggest that for fecal microbiome studies, collecting a fecal within 1 day (24 h) of deposit is acceptable when conditions are moist. Within the first 24 h, while we observed small shifts in Firmicutes and increased dispersion in the community suggesting the beginning of changes to the community, the communities overall did not significantly change until 48 h post defecation. Our findings indicate microbial community shifts were likely through changes in endemic bacterial members and colonization of environmental microbes, and that microbial signatures of individual animals can be detected in the earlier days, but not in the later stages of degradation. Other factors should be considered in the future for fecal microbiome stability such as environmental conditions, life stage of the host, and health (e.g. fecal consistency) of the host.
There is generally a consistency in core Phyla across the fecal microbiome in the current study on ex situ cheetahs and previously published data in ex situ and in situ cheetahs. Our most representative samples of the cheetah gut microbiome, day 0 subsamples, had relative abundance similarities with wild cheetah data for Firmicutes (63% this study vs 56.2–68.5% (Menke et al. 2014, Wasimuddin et al. 2017), respectively), Proteobacteria (2.5% vs 4.2–6.3%), and Fusobacteria (13% vs 18.1–18.4%), and similar relative abundances of Actinobacteria with captive cheetahs in Belgium (3.5% this study vs 4.3% (Becker et al. 2014)). Interestingly, we found the relative abundance of Bacteroidetes to be higher than both the wild Namibian cheetahs (18% vs 5.8% and 6.5%) and captive cheetahs in Belgium (18% vs <1%). Bacteroidetes are a dominant Phyla in domestic cats in the USA, Hong Kong, and China (Ritchie et al. 2010, Handl et al. 2011, Tun et al. 2012, Bai et al. 2023) but are rare in other carnivores (Ley et al. 2008, Schwab & Ganzle 2011, Becker et al. 2014). This may be due to the high amount of starch and plant-based fibers found in domestic cat kibble compared to carcass diets of wild carnivores. Bacteroidetes have the ability to breakdown complex carbohydrates into short-chain fatty acids in the large intestine which are then utilized by the host (Thomas et al. 2011, Tremaroli & Backhed 2012). We hypothesize that we detected a higher relative abundance of Bacteroidetes in fecal samples of our ex situ cheetah samples for similar reasons. The primary dietary source of the current study cohort was a commercially available beef diet that contains a mix of insoluble and soluble plant-based fiber and cellulose (de Godoy 2018). Based on the dietary descriptions from the other studies, it seems likely the added source of plant-based carbohydrates may be the reason for the increased relative abundance of Bacteroidetes in the current study.
Relative and differential abundance of phyla and ASVs across sample days
Of the three phyla that were differentially abundant across sample days, Firmicutes abundance was more consistent over time, with only day 3 differing from day 0. Fusobacteria abundance decreased over time, with lower abundance starting at day 3. Fusobacteria abundance was higher in day 0 through day 2 samples, indicating many ASVs were likely cheetah gut-derived and subsequently decreased with exposure to the environment. Unlike Fusobacteria, Proteobacteria abundance increased over time, suggesting an initial low contribution of microbes endemic to the host cheetah followed by an increase due to either the proliferation of endemic microbes or colonization of new ones in the external environment. Generally, fresher samples were characterized by a greater abundance of Fusobacteria, while later-stage samples were more likely to have a greater abundance of Proteobacteria. At the ASV level, both ASVs that were largely absent in day 0 samples belonged to the phyla Firmicutes and the genus Bacillus but could not be identified at the species level. Previous research suggests Bacillus is a common genus in soil (Saxena et al. 2020) as many species within the genus are capable of fixing atmospheric nitrogen (Yousuf et al. 2017). These two ASVs likely arose from environmental sources as they were not present in the majority of fresh samples.
Cheetah fecal bacterial species richness and diversity did not vary over the time course of this study. This outcome is similar to that reported for domestic cat fecal samples stored at ambient temperature for 4 days (Tal et al. 2017) and captive mink fecals sampled for 5 days (Lafferty et al. 2022). We expected species richness to first decrease as the community switched from an anaerobic (inside the host gut) environment to an aerobic (outside the body) environment, followed by an increase as environmental bacteria began to colonize the excreted feces. In general, we observed slight evidence for this pattern (Supplementary Fig. 2), but nothing significant. While species richness and PD did not change over time, they were influenced by individual (fecal ID) and precipitation (MCD rainfall). This indicates that there were strong microbial signatures among cheetahs due to host-related factors, as expected based on previous work (Becker et al. 2015, Wasimuddin et al. 2017). The effect of MCD rainfall indicates that while samples from the same individual were stable over time (sample day), increased rainfall caused greater changes in species richness and PD within individuals compared to sample series without rainfall.
In the first 24 h after defecation, there were minor shifts in phyla relative abundance and some small changes in microbial compositional dispersion. However, large-scale significant changes did not occur until day 2. After day 2, subsamples deviated and were no longer representative of the original fresh sample. By day 3, fecal subsamples had changed so drastically that the individual fecal ID was no longer influential on community differences (Fig. 5, Supplementary Table 4), and individual microbial signatures were lost (Fig. 4). Rainfall was not influential on microbial community structure and composition, indicating the variation in microbial communities between subsamples was not as sensitive to rain, and the effects of fecal ID and sample day accounted for more of the variation. Similarly, community microbial composition was stable over time in studies of other fecal microbiota time series analyses for up to 12 h in bats (Fofanov et al. 2018), multiple days in domestic cats (Tal et al. 2017), and springbok and giraffes (Menke et al. 2015), or a week or more in humans (Lauber et al. 2010, Dominianni et al. 2014). The variation in fecal microbial composition stability across these species may be influenced by host demographics, digestive tract type, diet, and environmental factors. Taken together, these data indicate that fecal microbial communities of cheetahs are robust within at least a 24-h time period and are less susceptible to perturbations than first assumed, even under rainy and humid conditions.
For researchers collecting feces in the field, it is extremely challenging to collect fresh feces, as defecations are difficult to observe and many feces are collected within unknown timeframes. Throughout this collection, we discovered characteristics that may be helpful in determining fecal age. Fresh feces are warm, soft, high in moisture, and show insect activity. By 24 h, the feces are no longer warm and have formed a thin crust around the outside surface. While the feces have lost some moisture and insect activity is reduced, there is still considerable moisture inside the feces, with an almost taffy-like consistency. The inside consistency and overall fecal color were the biggest differences between the 24-h and 48-h feces. At 48 h, the fecal appeared lighter in color, and the inside of the feces was not as moist, resulting in a more crumble-like texture. These are some characteristics that may help researchers identify fresher feces both in captive habitats and in the wild (photos available in Supplemental Materials, see Data Accessibility Statement). Further research is needed to determine whether microbial biomarkers exist (Muletz-Wolz et al. 2021) and when they appear or disappear over time when feces are exposed to the natural habitat of wild cheetahs.
Two additional patterns in fecal microbial diversity and composition were observed that hint at age and gut health as drivers of microbial variation. Samples from the two reproductive-aged adult females in our study (Aju 2 and 6, approximately 6 years old) exhibited greater microbial diversity and less compositional shifts over time relative to samples from healthy juveniles approximately 1 year of age. Little is known about age differences in fecal microbiota of cheetahs, but our data agree with previous studies in domestic cats that showed detectable differences between the fecal microbiota of adults and kittens (Bermingham et al. 2018). While the juveniles in the current study were weaned, they were still pre-pubertal and may have a transitory microbiome between cub and adult. The two juvenile individuals that exhibited GI distress and diarrhea (Aju 7 and 8) during sample collection had species richness on the lower end of the juvenile scale and were also compositionally similar to each other but distinct from all other individuals. A previous 3-year longitudinal study on cheetah fecal microbiomes reported decreased species richness, shifts in microbial membership, and greater variation over time in a cheetah that became ill with vomiting and diarrhea during the study (Becker et al. 2015). These data also agree with previous reports in domestic cats finding fecal microbial differences between healthy cats and those with GI inflammation in domestic cats (Janeczko et al. 2008, Honneffer et al. 2014) and other mammals (Bragg et al. 2020, Keady et al. 2021). Interestingly, the sample series from the adults and the GI-unhealthy individuals exhibited more consistent microbial community composition and structure over time. These factors would be of interest to future studies.
While fresh samples will always be preferred, based on our data we believe collecting cheetah feces up to 24 h post deposit is acceptable for fecal microbial research. We predict that these recommendations will hold for other large carnivores, particularly the large cats of the family Felidae, though empirical studies are needed to confirm this prediction. If samples are collected within the first 24 h, it is unlikely there will be large differences in microbial diversity or community composition and structure, and community signatures distinguishing individual hosts should be retained, especially in adults. These findings will greatly reduce the resources needed and create a safer, more passive sampling strategy for both ex situ and in situ cheetah gut microbial ecology. However, rainfall should be considered, as heavy rain between defecation and sampling may alter microbial species richness. If fecal samples need to be collected past 24 h post deposit, we recommend additional studies on the effects of age, sex, location, and GI health on cheetah fecal microbiota to build a more thorough understanding of microbiome stability over time.
Supplementary materials
This is linked to the online version of the paper at https://doi.org/10.1530/MAH-23-0022.
Declaration of interest
The authors declare that there is no conflict of interest that could be perceived as prejudicing the impartiality of the study reported.
Funding
Funding was provided by grants from Smithsonian Scholarly Studies and Friends of the National Zoo.
Data Accessibility Statement
Raw sequences are deposited in the NCBI SRA repository (BioProject ID PRJNA967019). The metadata, ASV table, taxonomy list, and metadata files and R code used for data processing and analyses are available on Dryad (https://doi.org/10.5061/dryad.2bvq83bvw) and GitHub: github.com/Malytherin/CheetahFecalStability_USA.
Author contribution statement
MAM conceived the idea. MAM, AEC, and CRMW designed the experiment. MAM performed sample collections, field experiments, and lab work. AEC supervised and assisted with sample collections. MAM and MMK performed the data processing and statistical analyses. CRMW supervised the lab work, data processing, and statistical analyses. MAM wrote the manuscript with direct support from CRMW and additional support from AEC, RBR, MB, and MMK. MAM, AEC, CRMW, RBR, and MB contributed to grant writing for acquiring funding for the project. All authors contributed to editing the final version of the manuscript.
Acknowledgements
The authors would like to thank NZCBI cheetah keeping staff, especially Amber Dedrick and Adri Kopp, for their assistance in sample collection. We also thank Nancy McInerney for laboratory and sequencing support.
References
Amato KR 2013 Co-evolution in context: the importance of studying gut microbiomes in wild animals. Microbiome Science and Medicine 1 10–29. (https://doi.org/10.2478/micsm-2013-0002)
Anderson MJ 2017 Permutational multivariate analysis of variance (PERMANOVA). In: Balakrishnan N, Colton T, Everitt B, Piegorsch W, Ruggeri F, Teugels JL, editors. Wiley StatsRef: Statistics Reference Online. John Wiley & Sons.pp. 1–15. (https://doi.org/10.1002/9781118445112.stat07841)
Bahrndorff S, Alemu T & Alemneh T et al.2016 The microbiome of animals: implications for conservation biology. International Journal of Genomics 2016 5304028. (https://doi.org/10.1155/2016/5304028)
Bai H, Liu T, Wang S, et al.2023 Identification of gut microbiome and metabolites associated with acute diarrhea in cats. Microbiology Spectrum 11 e0059023. (https://doi.org/10.1128/spectrum.00590-23)
Becker AA, Hesta M, Hollants J, et al.2014 Phylogenetic analysis of faecal microbiota from captive cheetahs reveals underrepresentation of Bacteroidetes and Bifidobacteriaceae. BMC Microbiology 14 43. (https://doi.org/10.1186/1471-2180-14-43)
Becker AA, Janssens GP, Snauwaert C, et al.2015 Integrated community profiling indicates long-term temporal stability of the predominant faecal microbiota in captive cheetahs. PLoS One 10 e0123933. (https://doi.org/10.1371/journal.pone.0123933)
Beckers KF, Schulz CJ & & Childers GW 2017 Rapid regrowth and detection of microbial contaminants in equine fecal microbiome samples. PLoS One 12 e0187044. (https://doi.org/10.1371/journal.pone.0187044)
Bermingham EN, Young W, Butowski CF, et al.2018 The fecal microbiota in the domestic cat (Felis catus) is influenced by interactions between age and diet; A five year longitudinal study. Frontiers in Microbiology 9 1231. (https://doi.org/10.3389/fmicb.2018.01231)
Bolyen E, Rideout JR, Dillon MR, et al.2019. Reproducible, interactive, scalable and extensible microbiome data science using QIIME 2. Nature Biotechnology 37 852–857. (https://doi.org/10.1038/s41587-019-0209-9)
Bragg M, Freeman EW, Lim HC, et al.2020 Gut microbiomes differ among dietary types and stool consistency in the captive red wolf (Canis rufus). Frontiers in Microbiology 11 5902 12. (https://doi.org/10.3389/fmicb.2020.590212)
Callahan BJ, Mcmurdie PJ, Rosen MJ, et al.2016 DADA2: high-resolution sample inference from Illumina amplicon data. Nature Methods 13 581–583. (https://doi.org/10.1038/nmeth.3869)
Callahan BJ, Mcmurdie PJ & & Holmes SP 2017 Exact sequence variants should replace operational taxonomic units in marker-gene data analysis. ISME Journal 11 2639–2643. (https://doi.org/10.1038/ismej.2017.119)
Choo JM, Leong LEX & & Rogers GB 2015 Sample storage conditions significantly influence faecal microbiome profiles. Scientific Reports 5 16350. (https://doi.org/10.1038/srep16350)
Cole JR, Chai B, Farris RJ, et al.2007 The ribosomal database project (RDP-II): introducing myRDP space and quality controlled public data. Nucleic Acids Research 35 D169–D172. (https://doi.org/10.1093/nar/gkl889)
Davis NM, Proctor DM, Holmes SP, et al.2018 Simple statistical identification and removal of contaminant sequences in marker-gene and metagenomics data. Microbiome 6 226. (https://doi.org/10.1186/s40168-018-0605-2)
De Godoy MRC 2018 Pancosma comparative gut physiology symposium: all about appetite regulation: effects of diet and gonadal steroids on appetite regulation and food intake of companion animals. Journal of Animal Science 96 3526–3536. (https://doi.org/10.1093/jas/sky146)
Dominianni C, Wu J, Hayes RB, et al.2014 Comparison of methods for fecal microbiome biospecimen collection. BMC Microbiology 14 103. (https://doi.org/10.1186/1471-2180-14-103)
Faith DP 1992 Systematics and conservation: on predicting the feature diversity of subsets of taxa. Cladistics 8 361–373. (https://doi.org/10.1111/j.1096-0031.1992.tb00078.x)
Fofanov VY, Furstenau TN, Sanchez D, et al.2018 Guano exposed: impact of aerobic conditions on bat fecal microbiota. Ecology and Evolution 8 5563–5574. (https://doi.org/10.1002/ece3.4084)
Fouhy F, Deane J, Rea MC, et al.2015 The effects of freezing on faecal microbiota as determined using MiSeq sequencing and culture-based investigations. PLoS One 10 e0119355. (https://doi.org/10.1371/journal.pone.0119355)
Gillman SJ, Mckenney EA, Lafferty DJR, et al.2022 Human-provisioned foods reduce gut microbiome diversity in American black bears (Ursus americanus). Journal of Mammalogy 103 339–346. (https://doi.org/10.1093/jmammal/gyab154)
Gorzelak MA, Gill SK, Tasnim N, et al.2015 Methods for improving human gut microbiome data by reducing variability through sample processing and storage of stool. PLoS One 10 e0134802. (https://doi.org/10.1371/journal.pone.0134802)
Handl S, Dowd SE, Garcia-Mazcorro JF, et al.2011 Massive parallel 16S rRNA gene Pyrosequencing reveals highly diverse fecal bacterial and fungal communities in healthy dogs and cats. FEMS Microbiology Ecology 76 301–310. (https://doi.org/10.1111/j.1574-6941.2011.01058.x)
Honneffer JB, Minamoto Y & & Suchodolski JS 2014 Microbiota alterations in acute and chronic gastrointestinal inflammation of cats and dogs. World Journal of Gastroenterology 20 1648 9–16497. (https://doi.org/10.3748/wjg.v20.i44.16489)
Janeczko S, Atwater D, Bogel E, et al.2008 The relationship of mucosal bacteria to duodenal histopathology, cytokine mRNA, and clinical disease activity in cats with inflammatory bowel disease. Veterinary Microbiology 128 178–193. (https://doi.org/10.1016/j.vetmic.2007.10.014)
Keady MM, Prado N, Lim HC, et al.2021 Clinical health issues, reproductive hormones, and metabolic hormones associated with gut microbiome structure in African and Asian elephants. Animal Microbiome 3 85. (https://doi.org/10.1186/s42523-021-00146-9)
Kembel SW, Cowan PD, Helmus MR, et al.2010 Picante: R tools for integrating phylogenies and ecology. Bioinformatics 26 1463–1464. (https://doi.org/10.1093/bioinformatics/btq166)
Lafferty DJR, Gillman SJ, Jeakle LK, et al.2022 Mink (Neovison vison) fecal microbiomes are influenced by sex, temperature, and time postdefecation. Journal of Mammalogy 103 316–327. (https://doi.org/10.1093/jmammal/gyab140)
Lauber CL, Zhou N, Gordon JI, et al.2010 Effect of storage conditions on the assessment of bacterial community structure in soil and human-associated samples. FEMS Microbiology Letters 307 80–86. (https://doi.org/10.1111/j.1574-6968.2010.01965.x)
Lenth R 2023. _ Emmeans: Estimated Marginal Means, Aka Least-Squares Means_. R package, version 1.9.0. (https://cran.r-project.org/web/packages/emmeans)
Ley RE, Lozupone CA, Hamady M, et al.2008 Worlds within worlds: evolution of the vertebrate gut microbiota. Nature Reviews. Microbiology 6 776–788. (https://doi.org/10.1038/nrmicro1978)
Martinez Arbizu P 2020. pairwiseAdonis: Pairwise Multilevel Comparison Using adonis., R package version 0.4.1. (https://github.com/pmartinezarbizu/pairwiseAdonis/tree/master)
Mcmurdie PJ & & Holmes S 2013 phyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One 8 e61217. (https://doi.org/10.1371/journal.pone.0061217).
Menke S, Wasimuddin, Meier M, et al.2014. Oligotyping reveals differences between gut microbiomes of free-ranging sympatric Namibian carnivores (Acinonyx jubatus, Canis mesomelas) on a bacterial species-like level. Front Microbiol 5 526. (https://doi.org/10.3389/fmicb.2014.00526)
Menke S, Meier M, Sommer S, et al.2015 Shifts in the gut microbiome observed in wildlife faecal samples exposed to natural weather conditions: lessons from time-series analyses using next-generation sequencing for application in field studies. Methods in Ecology and Evolution 6 1080–1087. (https://doi.org/10.1111/2041-210X.12394)
Muletz Wolz CR, Yarwood SA, Campbell Grant EH, et al.2018 Effects of host species and environment on the skin microbiome of Plethodontid salamanders. Journal of Animal Ecology 87 341–353. (https://doi.org/10.1111/1365-2656.12726)
Muletz-Wolz CR, Almario JG, Barnett SE, et al.2017 Inhibition of fungal pathogens across genotypes and temperatures by amphibian skin bacteria. Frontiers in Microbiology 8 1551. (https://doi.org/10.3389/fmicb.2017.01551)
Muletz-Wolz CR, Fleischer RC & & Lips KR 2019 Fungal disease and temperature alter skin microbiome structure in an experimental salamander system. Molecular Ecology 28 2917–2931. (https://doi.org/10.1111/mec.15122)
Muletz-Wolz CR, Wilson Rankin E, Mcgrath-Blaser S, et al.2021 Identification of novel bacterial biomarkers to detect bird scavenging by invasive rats. Ecology and Evolution 11 1814–1828. (https://doi.org/10.1002/ece3.7171)
Munson L 1993 Diseases of captive cheetahs (Acinonyx jubatus): results of the cheetah Research Council pathology survey 1989–1992. Zoo Biology 12 105–124. (https://doi.org/10.1002/zoo.1430120110)
Munson L, Nesbit JW, Meltzer DG, et al.1999 Diseases of captive cheetahs (Acinonyx jubatus jubatus) in South Africa: a 20-year retrospective survey. Journal of Zoo and Wildlife Medicine 30 342–347.
Munson L, Terio KA, Worley M, et al.2005 Extrinsic factors significantly affect patterns of disease in free-ranging and captive cheetah (Acinonyx jubatus) populations. Journal of Wildlife Diseases 41 542–548. (https://doi.org/10.7589/0090-3558-41.3.542)
Nardi F, Lazzeri L, Iannotti N, et al.2022 Analysis of scat for gut microbiome identification in wolves from a Mediterranean and an Alpine area. Diversity 15. (https://doi.org/10.3390/d15010037)
Oksanen J, Simpson G, Blanchet F, et al.2022 Vegan: Community Ecology Package. R package version 2.6-4. (http://CRAN.R-project.org/package=vegan)
Plan CSS 2009 Husbandry manual for the cheetah (Acinonyx jubatus). Edited and compiled by K Ziegler-Meeks. White Oak Conservation. (https://www.awjac.org/docs/CheetahManualFinal-1-19-2009.pdf)
Price MN, Dehal PS & & Arkin AP 2009 FastTree: computing large minimum evolution trees with profiles instead of a distance matrix. Molecular Biology Evolution 26 1641–1650. (https://doi.org/10.1093/molbev/msp077)
R CoreTeam 2022a RStudio: Integrated Development Environment for R: Boston, MA: Posit Software, PBC. (http://www.rstudio.com/)
R CoreTeam 2022b R: A Language and Environment for Statistical Computing: Vienna, Austria: R Foundation for Statistical Computing. (https://www.R-project.org)
Redford KH, Segre JA, Salafsky N, et al.2012 Conservation and the microbiome. Conservation Biology 26 195–197. (https://doi.org/10.1111/j.1523-1739.2012.01829.x)
Ritchie LE, Burke KF, Garcia-Mazcorro JF, et al.2010 Characterization of fecal microbiota in cats using universal 16S rRNA gene and group-specific primers for Lactobacillus and Bifidobacterium spp. Veterinary Microbiology 144 140–146. (https://doi.org/10.1016/j.vetmic.2009.12.045)
Russel J, Thorsen J, Brejnrod AD, et al.2018 DAtest: a Framework for Choosing Differential Abundance or Expression Method. biorxiv. (https://doi.org/10.1101/241802)
Saxena AK, Kumar M, Chakdar H, et al.2020 Bacillus species in soil as a natural resource for plant health and nutrition. Journal of Applied Microbiology 128 1583–1594. (https://doi.org/10.1111/jam.14506)
Schwab C & & Ganzle M 2011 Comparative analysis of fecal microbiota and intestinal microbial metabolic activity in captive polar bears. Canadian Journal of Microbiology 57 177–185. (https://doi.org/10.1139/W10-113)
Sugden S, Sanderson D, Ford K, et al.2020 An altered microbiome in urban coyotes mediates relationships between anthropogenic diet and poor health. Scientific Reports 10 2220 7. (https://doi.org/10.1038/s41598-020-78891-1)
Tal M, Verbrugghe A, Gomez DE, et al.2017 The effect of storage at ambient temperature on the feline fecal microbiota. BMC Veterinary Research 13 256. (https://doi.org/10.1186/s12917-017-1188-z)
Terio KA, Mitchell E, Walzer C, et al.2018 Diseases impacting captive and free-ranging cheetahs. Cheetahs: Biology and Conservation 349–364. (https://doi.org/10.1016/B978-0-12-804088-1.00025-3)
Thomas F, Hehemann JH, Rebuffet E, et al.2011 Environmental and gut Bacteroidetes: the food connection. Frontiers in Microbiology 2 93. (https://doi.org/10.3389/fmicb.2011.00093)
Tremaroli V & & Backhed F 2012 Functional interactions between the gut microbiota and host metabolism. Nature 489 242–249. (https://doi.org/10.1038/nature11552)
Tun HM, Brar MS, Khin N, et al.2012 Gene-centric metagenomics analysis of feline intestinal microbiome using 454 junior Pyrosequencing. Journal of Microbiological Methods 88 369–376. (https://doi.org/10.1016/j.mimet.2012.01.001)
Wang Q, Garrity GM, Tiedje JM, et al.2007 Naive Bayesian classifier for rapid assignment of rRNA sequences into the new bacterial taxonomy. Applied and Environmental Microbiology 73 5261–5267. (https://doi.org/10.1128/AEM.00062-07)
Wasimuddin Menke S, J Melzheimer, et al.2017 Gut microbiomes of free-ranging and captive Namibian cheetahs: Diversity, putative functions and occurrence of potential pathogens. Molecular Ecology 26 5515–5527. (https://doi.org/10.1111/mec.14278)
Weiss S, Xu ZZ, Peddada S, et al.2017 Normalization and microbial differential abundance strategies depend upon data characteristics. Microbiome 5 27. (https://doi.org/10.1186/s40168-017-0237-y)
Wong K, Shaw TI, Oladeinde A, et al.2016 Rapid microbiome changes in freshly deposited cow feces under field conditions. Frontiers in Microbiology 7 500. (https://doi.org/10.3389/fmicb.2016.00500)
Yousuf J, Thajudeen J, Rahiman M, et al.2017 Nitrogen fixing potential of various heterotrophic Bacillus strains from a tropical estuary and adjacent coastal regions. Journal of Basic Microbiology 57 922–932. (https://doi.org/10.1002/jobm.201700072)
