Effects of captivity, diet, and relocation on the gut bacterial communities of white‐footed mice

Abstract Microbes can have important impacts on their host's survival. Captive breeding programs for endangered species include periods of captivity that can ultimately have an impact on reintroduction success. No study to date has investigated the impacts of captive diet on the gut microbiota during the relocation process of generalist species. This study simulated a captive breeding program with white‐footed mice (Peromyscus leucopus) to describe the variability in gut microbial community structure and composition during captivity and relocation in their natural habitat, and compared it to wild individuals. Mice born in captivity were fed two different diets, a control with dry standardized pellets and a treatment with nonprocessed components that reflect a version of their wild diet that could be provided in captivity. The mice from the two groups were then relocated to their natural habitat. Relocated mice that had the treatment diet had more phylotypes in common with the wild‐host microbiota than mice under the control diet or mice kept in captivity. These results have broad implications for our understanding of microbial community dynamics and the effects of captivity on reintroduced animals, including the potential impact on the survival of endangered species. This study demonstrates that ex situ conservation actions should consider a more holistic perspective of an animal's biology including its microbes.

These communities of microbes in a host are collectively known as the microbiota. The gut microbiota can play a role in host development, digestion, immunity, and behavior (McKenney, Koelle, Dunn, & Yoder, 2018;Suzuki, 2017) and can therefore influence the survival of relocated animals. Host-associated microbiota are highly dynamic communities, and disrupting their equilibrium can lead to negative direct or indirect effects on their host (Hooks & Malley, 2017;reviewed in West et al., 2019) such as impaired immune function and metabolic disorders Krynak, Burke, Martin, & Dennis, 2017;Rosshart et al., 2017;Wasimuddin et al., 2017).
Housing facilities such as zoos where captive breeding programs are held provide intense veterinary care, sanitized enclosures, a standardized diet, and reduced sexual selection. Captivity has been shown to alter the microbiota of animals compared to wild counterparts (Borbón-García, Reyes, Vives-Flórez, & Caballero, 2017;Clayton et al., 2016;McKenzie et al., 2017;Wasimuddin et al., 2017). The majority of the studies show similar trends: a decrease in bacterial phylotype richness (or α-diversity) among captive individuals compared to their wild conspecifics, as well as differences in community composition (or β-diversity) between the groups. However, some host species show an opposite pattern (Frankel, Mallott, Hopper, Ross, & Amato, 2019;Greene et al., 2019;McKenzie et al., 2017), postulating that the gut microbiota of group taxa respond differently to captivity, mainly through their feeding strategy and gut physiology. Differences observed in gut microbial communities have largely been attributed to altered diets in captivity that can also lead to the extinction of microbial niches and functions in the host's gut over multiple generations in captivity (Sonnenburg et al., 2016). Standardized diets are generally composed of simple fibers in low quantity compared to carbohydrates. A loss of microbial taxa taking part in the digestion of fibers in captivity has been linked to disease as a short-term disadvantage to hosts Krynak et al., 2017;Rosshart et al., 2017) and could also, in the long run, be a disadvantage when hosts are relocated to their natural habitat. In general, there is extensive gut microbiota variation when an animal fed on a more animal-based or plant-based diet for humans and mice, compared to a balanced diet from various food sources (Heiman & Greenway, 2016). The consumption of a diverse diet avoids the loss of crucial microbial function linked to a specific food item in the case of omnivorous host. This has been demonstrated through the evolution of human lifestyles. Human microbial communities have been shaped through changes from hunter-gatherer and nomad societies to farming, sedentary, and urban lifestyles. Especially in Western diets, the lack of fibrous food items and the increased consumption of processed foods have resulted in a reduction of gut bacteria diversity that has been implicated in many diseases linked to impaired immune responses and metabolic disorders (Kolodziejczyk, Zheng, & Elinav, 2019). It is therefore essential to study the effects of changes in diet during captivity on the gut microbiota among different host taxa with variable ecological niches, being dietary generalists or specialists, omnivorous or herbivorous, for example. Previous work has suggested that a change to a more fibrous and less processed diet in captivity changes the gut microbiota compared to a standard diet, but it does not make the gut microbiota of captive animals more similar to their wild counterparts (Allan et al., 2018;Cabana et al., 2019).
However, the impacts of diet change on gut bacteria remain to be investigated during animal relocation.
Few studies have shown how host-associated microbiota vary between captivity and relocation into a natural habitat and mainly focused on the impacts of place of birth and the immediate environmental exposure (Chong et al., 2019;Metcalf et al., 2017;Schmidt et al., 2019;Yao et al., 2019). Even less studies looked at the effect of captive diet on the gut microbiome during animal relocation (Martinez-Mota, Kohl, Orr, & Dearing, 2019). Overall, animals born in captivity have lower α-diversity and more differences in microbial communities compared to animals born in nature reserves or in the wild (Metcalf et al., 2017). Furthermore, deer mice (Peromyscus maniculatus) born in captivity and later released had gut microbial communities closer to their wild counterparts compared to animals that stayed in captivity (Schmidt et al., 2019). Since the literature gap of the effects on gut bacteria of captive diet during relocation remains unaddressed, we focused on the impact of captive diet for the gut bacteria during the relocation process in a generalist species with an omnivorous diet. We hypothesized that diet during captivity will affect the gut microbiota of the host during the relocation process and can maximize the reintroduction success of an animal back into its natural habitat. We predict that a wild-like and nonprocessed diet in captivity would foster the recovery of a wild-like microbiota after the animal is relocated in their natural habitat, compared to a standard captive diet composed of pellets. Therefore, captive diets reflecting a wild diet could have benefits associated with improved degradation of food items by microbes, echoing higher microbial diversity in the gut of mice under a wild-like diet compared to a standard and processed diet in captivity. This study was conducted on the gut bacterial communities of the white-footed mouse (Peromyscus leucopus), an omnivorous rodent native to Ontario (Canada) that feeds primarily on insects, seeds, nuts, and fruits, just like its closely related species, P. maniculatus (Wolff, Dueser, & Berry, 1985). Peromyscus leucopus does not face major threats of extinction, but its large distribution, short generation time, and high capture-recapture rate in general make it an adequate model to study gut microbiota variation across a short period of time to simulate a captive breeding program for reintroduction purposes. 2018-05-02. White-footed mice (P. leucopus) were trapped on the grounds of the Toronto Zoo (ON, Canada) using Longworth traps.

| Sample collection
Trapping occurred five nights a week during the breeding season of June to mid-September 2018. Each mouse was identified with unique numerical tags (National Brand and Tag Co.) and weighed.
Wild juveniles were detected by the color of their fur (gray) and their weight (<15 g) and were excluded from the study. Fecal samples were collected directly from the animal using flamed and 70% ethanol-sterilized tweezers and stored in sterile microcentrifuge tubes in a −20°C freezer until DNA extraction.  Figure 1). The control group received the same diet as their mothers, and the treatment group received a diet composed of sunflower seeds, diced apples, crushed walnuts, mealworms, and crushed corn in equal proportions. Each animal received its respective diet and was kept in these conditions for 30 days until they reached sexual maturity. Fecal samples were collected 8 days and 1 day prior to release for each individual (Table 1). Those samples represent the Captive Control (CC) and

| Experimental design
Captive Treatment (CT) study groups. All offspring were then released at one of three locations on the grounds of the Toronto Zoo.
Fecal samples were collected from all wild adults trapped in this period at least twice, 7 days apart (W for Wild experimental group; Note: (A) Mice and related samples that were born in captivity and were under the two different diets. Once relocated in the wild, some mice were not recaptured and others have been recaptured more than once. All samples collected from the recaptured mice have been included in the study.
(B) Mice and related samples collected from the wild that never experienced captivity. Some had botfly infection, and its effect has been taken into account in the later analysis.
The presence or absence of infection was considered in the sampling and analysis of the data.

| DNA extraction and sequencing
Gene amplicon sequencing was used to study the bacterial communities. DNA extractions from the fecal samples collected were conducted using the Stool DNA Isolation Kit (Norgen Biotek Corp) following the manufacturer's instructions. Two blank extractions were made to control for contamination during the extraction process. After DNA extraction, the targeted gene for taxonomic affiliation (16S rRNA gene) was amplified through polymerase chain reactions (PCRs). The library preparation and sequencing were performed by Metagenombio Inc., as well as the demultiplexing of the sequence reads. Using their designated library protocol, 2 × 300 bp paired-end sequencing was completed using broad bacterial primers of the region V4 of the 16S rRNA gene (515F-806R) using an Illumina MiSeq platform (Illumina Biotechnology Co.).

| Bioinformatics
The quality controls of the already demultiplexed paired-end sequence reads were performed through the software FastQC (Andrews, 2010).

Sequence reads denoising and amplicon sequence variants (ASVs)
picking steps were done with the QIIME2 tool  2019.1), using the DADA2 pipeline (Callahan, Mcmurdie, & Holmes, 2017;Callahan et al., 2016). ASVs-or also referred to as bacterial phylotypes-were then screened to the 97% 16S rRNA gene full-length reference sequences from the Silva v.132 database (Pruesse et al., 2007) for taxonomical association using the VSEARCH classifier implemented in QIIME2 . Sequence alignment and phylogeny building were conducted in QIIME2 for the construction of a generalized UniFrac distance matrix (α = .5; Chen et al., 2012). The cumulative sum scaling (CSS) method was used to normalize the data

| Statistical analysis for α-diversity of gut bacteria between study groups
All statistical analyses were conducted in R (R version 3.5.2, R Core were analyzed using linear mixed models with a restricted maximum likelihood estimation approach with mouse ID and dam ID as random factors, using the lmer function in the lme4 package in R. The study group variable was first considered as two distinct factors: diet and environment. If the two factors and the interaction between the two had a significant effect on the variable, they were combined as "study group. ANOVAs with Satterthwaite's method were run on these models, as well as post hoc Tukey method for p-value adjustments was conducted to investigate differences between groups.
Normality of residuals was validated using the Shapiro-Wilk test.
The significance cutoff was set to p-value < .05 for each test.

| Statistical analysis for β-diversity of gut bacteria between study groups and differential abundance
A generalized UniFrac distance matrix between samples (Chen et al., 2012) was used to investigate differences in gut microbial communities between groups, sex, for maternal effect, and Cuterebra infection. This metric takes into account the differences in phylogenetic removes the arch effect. Finally, a minimum spanning tree was constructed using the phyloseqGraphTest package (Callahan et al., 2016).
The differential abundance analysis was conducted on the ASVs that were present in more than 5% of all the samples and that had a relative abundance of more than 5% among all taxa. It corresponds to the core microbiota of the dataset, represented by 653 phylotypes. The phylotype abundance analysis was made using the  In total, 36 mice were included in the study (n = 18 W; n = 10 CC; n = 8 CT; n = 5 RC; n = 6 RT) and 109 fecal samples were obtained from those mice (n = 36 W; n = 20 CC; n = 16 CT; n = 8 RC; n = 29 RT). The low number of mice in the RC and RT groups is due to the fact that the other mice released were not recaptured after relocation (Table 1).

| α-diversity of gut bacteria between study groups according to the host's birthplace
Some relocated and wild animals were sampled during a Cuterebra infection (n = 26 infected; n = 79 noninfected; Table 1), but they were not treated separately in the statistical analysis as it explained 1% of the community variation (ANOVA: F = 0.0171; p = .896237).
There was a significant difference in terms of phylotype evenness between mice born in captivity and in the wild (Simpson's evenness index: F = 2.785; p = .01877; Figure 2a), so that mice born in captivity carry gut communities less uniform in phylotype abundance.
The interaction of host sex and study group also had a significant impact on the gut bacterial phylotype richness (Fisher's index: F = 6.2087; p = .006176; Figure 2b). Male mice from the CC group had significantly higher gut bacteria phylotype richness compared to females from the same group (Tukey: F = 4.4974; p = .031458) or compared to males from the wild (Tukey: F = 4.4974; p = .03992457; Figure 2b).
From the fecal samples collected, the gut microbial communities of wild mice contained 834 unique phylotypes (5.3% of their gut bacteriome), which is more than captive mice (586, 3%) and relocated animals (525, 1.5%). Relocated and wild mice had 250 common bacterial phylotypes in their gut, which represent a higher proportion (8.8%) than the 238 phylotypes common between relocated and captive (7.3%), and between wild and captive individuals (141, 2.7%).
Similar proportions were found between gut bacteria of wild mice and mice that had the control and treatment diets.

| β-diversity of gut bacteria between study groups
As   Fecal microbiota differed more in composition between captive and wild mice than between relocated and wild mice (Figure 3;

| Differential abundance among groups
The assessment of the differential abundance of bacterial phylotypes using a negative binomial Wald test was conducted on the core microbiota of 653 phylotypes. From those, 62 from four phyla varied significantly among the study groups (Bacteroidetes, Espilonbacteraeota, Firmicutes, and Tenericutes; Figure 6; Table 2). 60% and 22% of the phylotypes with differential enrichment across groups, respectively, belonged to the Lachnospiraceae and Muribaculaceae families. Mice from the CC and RC groups had the greatest loss in abundance in gut phylotypes compared to the other study groups (Table 2; respectively, nine for CC and 14 for RC). Overall, the RT group was the only group that had significant phylotypes enriched and in common with gut communities from W mice compared to the other study groups ( Figure 6; Table 2).

| α-diversity of gut bacteria between study groups
We compared the gut microbiota of mice in captivity under different diets, after relocation, and in the wild. The structures of the gut microbial communities in terms of phylotype richness were similar among study groups, with significantly higher phylotype richness only observed in the gut microbial communities of male CC mice compared to females and to wild males. This result is not common on gut microbiome studies in wildlife and explains no or little variation (Schmidt et al., 2019;Wasimuddin et al., 2017). In that case, captivity could have a sex-specific effect on the gut microbiota of P. leucopus. However, we found no apparent differences in community composition on beta diversity analyses; these results could thus be an artifact from low sample sizes.
The structure of the microbiota in terms of evenness is more uniform in wild-born mice than captive-born mice that have more disparate microbial communities. Similar results were found in studies F I G U R E 6 Heatmap representing the results of the differential abundance analysis. Samples on the x-axis are grouped by mouse identification tag and by study group. The different colors represent the abundance on a log10 scale of each significantly enriched phylotype (median from all samples by mouse). Each phylotype on the y-axis is named by family and genus   691  999  1353  1995  6110  7221  7471  8022  8624  54100  496  1407  6057  7631  7975  8362  8887  9798  691  999  1353  7221  7471  496  6057  7631  7975  8362  8887  1  3  18  51  52  58  59  65  73  82  85  89  92  93  94  124  146  including the place of birth as a factor of variation in gut microbiota for horses and deer mice (Metcalf et al., 2017;Schmidt et al., 2019). Kohl and Dearing (2014) also observed that evenness decreased with time spent in captivity in desert woodrats. It has been hypothesized that this difference could be due to lasting founder effects of colonization of the gut by microbes during the early life of the host.
The natural habitat would be the source of more diverse bacterial phylotypes (interactions with more species, diverse substrates and diets, seasonality, and no antibiotic treatments) compared to captivity. However, the opposite trend was observed in Andean bears and red pandas (Borbón-García et al., 2017;Kong et al., 2014). Host diet, phylogenetics, and position in trophic networks could thus be important factors to consider. Overall, the evenness in bacterial communities can affect the subsequent response to disturbances and is known as the insurance hypothesis (Wittebolle et al., 2009), suggesting that place of birth may have on impact on host survival from the gut microbiome aspect. However, it is worth mentioning that differences in diversity indexes between wild and captive mice might be due to the fact that there is no knowledge about relatedness between animals of the W group, whereas captive-born animals come from a handful of litters that can have an impact on the gut microbiota (Spor et al., 2011).

| β-diversity of gut bacteria between study groups and differential abundance among study groups
From the β-diversity analysis, we observed that RT and RC individuals were the closest to their wild counterparts in terms of microbial structure and composition than CC and CT animals (Figures 4 and 5).
This would imply that the immediate environment has a strong effect on gut microbiota composition. Once the individuals are relocated in their natural habitat, the environment becomes the main source for microbes' horizontal acquisition, in both external exposure but in diet as well (Colston, 2017). Therefore, the captive diet seems to have a smaller impact compared to external exposure but appears to have lasting effects on the gut microbiota, since it influences its composition and structure even 1 month after relocation ( Figure 6).
The reduced influence of the diet compared to the external environment is also reflected in specific phylotype abundances. Similar to Schmidt et al. (2019), the Lachnospiraceae family is differently distributed between the gut microbiota of captive, wild, and relocated animals. Although they are present in all groups, Lachnospiraceae phylotypes are mostly enriched in the W and RT groups rather than the CC, CT, and RC groups. Maurice et al. (2015) examined the variation of Lachnospiraceae in wild Peromyscus species. They hypothesized that seasonal variation in the abundance of this taxon is linked to a diet shift from insects to seeds in mid-summer because these bacterial groups support the degradation of complex plant materials.
These taxa seem to play a role in the degradation of butyrate during fiber degradation that promotes colonocyte health, immune defense, and anti-inflammatory action, reducing the risk of developing metabolic disorders that are a growing concern in captive populations (Meehan & Beiko, 2014;Vijay-Kumar et al., 2010). However, in our study, the abundance of the Lachnospiraceae family is stable between the study groups but, at a lower taxonomical level, genera abundances within this family seem to differ. The Lachnospiraceae NK4A136 group had the highest variation between groups: increasing in RT and W mice and decreasing in the other groups. Not much is known of this genus, but it is associated with the digestive tract of mammals, using carbohydrates and producing short-chain fatty acids (Meehan & Beiko, 2014). Further studies targeting the Lachnospiraceae groups would be needed to investigate to which extent there is variation in these taxa between the study groups and their role in the mouse gut.
However, the Lachnospiraceae taxa seemed to be absent or reduced in abundance in the gut of the RC mice and those hosts may have loss the beneficial microbial function linked to these taxa.  Sonnenburg et al., (2016), mice subjected to standard low-fiber diet recovered less microbiota diversity than mice fed with a high-fiber, less processed diet. However, the diversity was not in terms of total phylotype richness but in terms of common bacterial groups with the wild "original" state of the microbiota.
The generalization that captivity induces an imbalanced microbiota linked to negative effects on the host should be considered with caution, because it can depend on the taxonomy and ecology of the host, as demonstrated by Greene et al. (2019) and Frankel et al. (2019).
It is also worth mentioning that across the studies comparing the gut microbiota of captive and wild animals, some enclosures allow access to open areas, social interaction, and enrichments that favor exposure to the natural habitat of the species (Clayton et al., 2018;Greene et al., 2019). This could shift the gut microbiota of these animals toward a wild-like state; However, this was not the case in our study and microbiota variation between CC, CT, RC, and RT mice could only be due to diet. Our study demonstrates that for the relocation of generalist rodents, it is not only a matter of captivity itself and external exposure, but also about diet manipulation. Even if the treatment diet might not reflect all the aspects and components of a wild diet for P. leucopus and could lead to nutrient deficiency over a long period of time, it is more adequate than standardized pellets for supporting microbiota composition of mice after relocation. Further work on gene expression in the microbiota and on the host's survival should be undertaken to understand the long-term effects of diet and microbiota variation once an animal is relocated.
It is worth mentioning that the recapture rates between the two relocated mice groups were different. The 75% of RT mice released were recaptured 29 times compared to only eight times for the 50% of released RC group (Table 1). This could be a survival rate indicator, but other ecological factors such as dispersal and predation need to be considered. White-footed mice are the prey of many animals such as the eastern screech-owl (Megascops asio) that was seen on site, and the persistent presence of raccoons (Procyon lotor) that disrupted traps and predated on mice (personal observation) might also account for the low recapture rate of RC mice. One explanation could be that RC mice have been more predated than RT mice because of microbiota-induced behavior (Ezenwa, Gerardo, Inouye, Medina, & Xavier, 2012), but further studies and monitoring would need to be undertaken. Finally, no significant results were found in gut microbiota variation due to botfly infection of Cuterebra sp. Even if the high prevalence of this infection in Peromyscus species has been reported, these parasites have been linked to little effect on host population densities or fitness in general (Slansky, 2007). Our results confirm this trend from the microbiota perspective.
This study simulated how captive breeding programs can impact the relocation process of animals under ex situ conservation actions.
We demonstrated that captive diet has an impact on the microbiota of a generalist host, even after relocation to a natural habitat. As the gut microbiota takes part in many aspects of an animal's biology, survival, and reproductive success, one should consider the microbiota aspect as well as the host's nutrition for the development of diets in captive settings. Researchers should continue to study the effect of captivity on the reintroduction process of endangered species at different scale levels: ecosystem, population, individual, and microbiota, and integrate them into management practices.

ACK N OWLED G M ENTS
We are grateful to all staff at the Toronto Zoo for assisting with experimental design, sample collection, and the use of their facilities and also to Jasmine Veitch for manuscript editing. Funding was supported by the NSERC CREATE grant, ReNewZoo.

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