Structural and compositional mismatch between captive and wild Atlantic salmon (Salmo salar) parrs’ gut microbiota highlights the relevance of integrating molecular ecology for management and conservation methods

Abstract Stocking methods are used in the Province of Quebec to restore Salmo salar populations. However, Atlantic salmon stocked juveniles show higher mortality rates than wild ones when introduced into nature. Hatchery environment, which greatly differs from the natural environment, is identified as the main driver of the phenotypic mismatch between captive and wild parrs. The latter is also suspected to impact the gut microbiota composition, which can be associated with essential metabolic functions for their host. We hypothesized that hatchery‐raised parrs potentially recruit gut microbial communities that are different from those recruited in the wild. This study evaluated the impacts of artificial rearing on gut microbiota composition in 0+ parrs meant for stocking in two distinct Canadian rivers: Rimouski and Malbaie (Quebec, Canada). Striking differences between hatchery and wild‐born parrs’ gut microbiota suggest that microbiota could be another factor that could impact their survival in the targeted river, because the microbiome is narrowly related to host physiology. For instance, major commensals belonging to Enterobacteriaceae and Clostridiacea from wild parrs’ gut microbiota were substituted in captive parrs by lactic acid bacteria from the Lactobacillaceae family. Overall, captive parrs host a generalist bacterial community whereas wild parrs’ microbiota is much more specialized. This is the very first study demonstrating extensive impact of captive rearing on intestinal microbiota composition in Atlantic salmon intended for wild population stocking. Our results strongly suggest the need to implement microbial ecology concepts into conservation management of endangered salmon stocks supplemented with hatchery‐reared parrs.

For instance, major commensals belonging to Enterobacteriaceae and Clostridiacea from wild parrs' gut microbiota were substituted in captive parrs by lactic acid bacteria from the Lactobacillaceae family. Overall, captive parrs host a generalist bacterial community whereas wild parrs' microbiota is much more specialized. This is the very first study demonstrating extensive impact of captive rearing on intestinal microbiota composition in Atlantic salmon intended for wild population stocking. Our results strongly suggest the need to implement microbial ecology concepts into conservation management of endangered salmon stocks supplemented with hatchery-reared parrs.

K E Y W O R D S
16S SSU rRNA gene, aquaculture, Atlantic salmon, host-microbiota interactions, metabarcoding sequencing, microbial ecology, stocking
Indeed, Salmo salar gut microbiota is specific to its local environment at the first life cycle stages and changes as soon smolts migrate from fresh to saltwater (Llewellyn et al., 2015), suggesting that intestinal bacterial communities would also differ depending on the rearing environment.
Because host-microbiota interactions are narrowly related to the host physiology (Donaldson et al., 2016;Klaasen et al., 1993;Liu et al., 2012;Scanlan et al., 2008;Wu & Lewis, 2013;Zhang, Lun, & Tsui, 2015), it is suspected that bacterial composition of microbial communities will tightly adapt to artificial rearing conditions (water composition, food, environmental bacterial community), which in turn will affect the ability of hatchery-reared parrs to adapt to natural conditions once released.
To determine the impact of artificial rearing on the gut microbiota composition of parrs meant for stocking, we sampled parrs juveniles from two different populations (Malbaie and Rimouski river) belonging to two different designable units (DU) of Salmo salar. DU are characterized by "an evidence of discreteness, such as in morphology, life history, behavior and/or neutral genetic markers as well as large disjunctions between populations and occupation of different eco-geographic regions" (COSEWIC, 2010). The two populations are subjected to conservation stocking, for which juveniles are reared in hatchery until they reach the stage of 0+ parrs before being released into the wild. Captive 0+ parrs, issued from wild breeders and reared in Tadoussac Hatchery (Quebec, Canada), have been sampled and compared to their wild relatives. By hypothesizing that parrs' gut microbiota composition is influenced more by rearing environment than breeder's genotype, this study aimed to (a) characterize environmental microbiota from hatchery and river waters, (b) characterize the gut microbiota composition from captive and wild-born parrs from the same genetic population (i.e., Rimouski or Malbaie), and (c) identify symbionts that are specific either to hatchery-or wild-born parrs' gut microbiota. Three predictions were made as follows: (a) Environmental microbial communities would significantly differ between the two rivers and the hatchery water, (b) gut microbiota composition of captive and wild-born parrs from the same genetic population would significantly differ, and (c) exclusive taxa would be found in both hatchery-and wild-born parrs' gut microbiota.
Using 16S SSU rRNA gene metabarcoding, we have been able to determine the microbiota composition of 27 parrs gut samples and water samples from each environment. Our results revealed significant differences between environment and gut microbiota from parrs depending on their origin. Furthermore, diet may be the most important factor contributing to the formation of the parrs' gut microbiota composition. Overall, this study suggests that environment may overpass genotype for the commensals recruitment: Parrs from the same genetic population, reared in two distinct environments, host a significantly different gut microbiome in terms of structure, diversity, and taxonomic composition. For instance, captive parrs' microbiota hosts mainly Lactobacillaceae whereas their wild relatives host mainly Enterobacteriaceae. Consequently, stocked parrs' gut microbiota may not confer the same metabolic functions as their wild relatives. Although further investigations are needed to understand how the divergence of the gut microbiota between the reared parrs' microbiota and their wild relatives will affect their survival in the wild, it is now clear that host-microbiome interactions must not be neglected for stocking programs.

| Samples collection and preparation
Captive parrs were sampled from the Tadoussac hatchery (Quebec, Canada), where salmonids from Malbaie and Rimouski rivers are reared for stocking. According to their time of capture, wild breeders from Malbaie and Rimouski rivers were kept in captivity for a period ranging from 2 months to 5 years before spawning. For both groups, artificial spawning occurred between November 28 and December 12, 2012, and hatching between March 29 and April 9, 2013. During incubation, water temperature was set according to the "modified natural thermal regime" to mimic the natural growth rate of wild salmon juveniles. During their growth, reared juveniles were fed with NutraST (Skretting). Captive parrs from both groups were sampled by August 8, 2013, right before stocking.

| Sequence analysis and statistics
The assembly of paired-end sequences was performed using QIIME (v.1.9.1) (Edgar, 2010) and PANDAseq (v.1.0) (Maselle, Bartram, Truszkowski, Brown, & Neufeld, 2012). Only paired sequences between 400 and 500 pb with a minimum overlap of 100 bp were kept for further analysis. Chimeric sequences were then removed using QIIME (Usearch61) to ensure that assembled sequences truly resulted from the same operational taxonomic unit (OTU). OTUs from each sample were assigned to the paired reads using the de_novo method at 97% similarity with the SILVA123 database.
An OTU table was then obtained and processed using R (3.2.3) and Phyloseq package. A total of 241,016 OTUs was obtained. Before data analysis, OTUs were filtered to remove Unidentified taxa. This step reduced the OTUs number to 212,352. All taxa under 0.005% of relative abundance were then removed (Bokulich et al., 2013). The remaining 1,067 OTUs were used for subsequent analysis.
Structure and composition of parrs' gut microbiota were investigated as follows: Alpha-diversity was calculated using the Shannon index; richness and evenness alone were estimated with Chao1 and Pielou's indexes, respectively. A nonparametric variance analysis (Kruskal-Wallis) and Kruskal-Wallis post hoc tests were performed on Shannon index to determine whether alpha-diversity of the gut microbiota from each group was similar. For network analysis, Spearman correlation was calculated between each gut samples. Identification of main OTUs composing the water and gut microbiota was performed by representing the twenty most abundant OTUs of each environment. OTUs were grouped according to family taxonomic rank and visualized with abundance barplots, using ggplot2 package under R. An additional abundance barplot was generated to allow the visualization of the gut and water microbiota composition according to the phylum taxonomic rank. At last, further analysis of gut microbiota composition was carried out by comparing the top 5 OTUs composing each parrs' group microbiota, represented at the genus level.

| RE SULTS
A total of 4,108,663 sequences were obtained after the sequencing and the assembly of paired sequences that were distributed within 23 phyla and assigned to 742 bacterial genera. After filtration (OUT relative abundance threshold of 0.005%), 1,067 OTUs were kept for the analysis of the water and gut microbiota.

| Microbiota structure analysis
Analysis of the alpha-diversity (Shannon index) of the gut microbiota ( Figure 1a) showed that captive-bred parrs housed a much more diversified microbiota than wild parrs. This is especially true This result is mostly explained by one outlier (MWP18), which presented a much higher diversity of symbionts than its relatives.
Alpha-diversity measurement boxplots for Shannon, Chao1, and Pielou's evenness indexes (for water and gut samples), as well as a F I G U R E 1 (a) Alpha-diversity of gut microbiota is lower in wild parrs when compared to their captive relatives. Shannon diversity indexes of gut microbiota are represented in boxplot regarding of parrs location. (b) Environmental conditions are the main driver of gut microbiota. Composition of the 27 microbiota samples, constructed with Cytoscape v.3.2.1, illustrates co-related samples based on Spearman coefficient (r > 0.3, p-value < 0.01). Each node (dot) represents a gut microbiota sample. The node size is proportional to the number of connections a sample makes with other samples, where captive parrs' microbiota shows higher number of connections within its group than wild parrs   Captive-bred parrs are grouped, whereas wild parrs' populations are differentiated from one another and from their captive relatives.
Clustering was even more pronounced with unweighted distance matrices ( Figure 2a). The PERMANOVA based on weighted (Table 1) and unweighted (Table 2) UniFrac metric distances revealed significant to highly significant differences between most groups, the lowest differentiation being detected for captive-bred parrs, which was not significant for weighted UniFrac distances (p-value = 0.116).
The analysis of the multivariate homogeneity of group dispersion (variances) revealed a higher interindividual variation for captive parrs and MWP group when compared to RWP (Figure 3). No significant difference was obtained when comparing MCP interindividual variances to MWP (p-value = 0.4271981). However, it is possible to assess the higher interindividual variation for MWP group by the outlier MWP18, which shows a more diverse gut microbiota composition than its relatives from the same origin as well as a singular bacterial composition.

| Environment microbiota composition
The analysis of the water bacterial community composition at the phylum level revealed the presence of Bacteroidetes, Proteobacteria, and Actinobacteria in every environment (Figure 4), but respective abundance levels varied between sampling sites, particularly for Actinobacteria which was less abundant in hatchery water. The bacterial taxonomic composition, characterized at the family level, however, showed diagnostic taxa for every single microbial niche: environmental water (hatchery and both rivers) and fish gut microbiota composition ( Figure 5). For water samples, the 20 most abun-

| Gut microbiota composition
At the phylum level, the gut microbiota composition revealed the presence of Proteobacteria, Firmicutes and Actinobacteria in each group ( Figure 4). However, a much more heterogeneous composition was highlighted within groups when the bacterial composition was analyzed at the family level ( Figure 5). For each parrs' population,

| D ISCUSS I ON
The purpose of this study was to assess the effect of both host population and rearing environment on gut microbiota taxonomic composition of Salmo salar parrs intended for river stocking.
Hatchery-reared parrs were generated with breeders issued from two wild populations (i.e., Rimouski and Malbaie rivers), whereas wild parrs were naturally born and grown in their respective rivers.
Overall, our results clearly demonstrate that environmental conditions (i.e., hatchery versus river) had the most prevalent effect on gut microbiota taxonomic composition. Substantial impacts on the fish's energetic and behavioral phenotypes resulting from hatcheryrearing conditions have been documented in numerous studies and are suspected of greatly reduce survival rates of stocked fishes (Milot et al., 2013;Stringwell et al., 2014). It has also been demonstrated that reared fish show a different methylation pattern than that of their wild relatives (Le Luyer et al., 2017). As energetic and behavioral phenotypes are controlled by gut microbiota activity (Tremaroli & Bäckhed, 2012) and the latter is suspected to impact epigenetic patterns (Bhat & Kapila, 2017;Cortese et al., 2016;Indrio et al., 2017;Rossi et al., 2011), our study is of prime interest as it is the first to demonstrate the extensive impact of captive rearing on gut microbiota composition in Atlantic salmon parrs meant for stocking. By identifying significant differences in terms of both structure and taxonomic composition between captive parrs and their wild relatives, the present work evidenced that acclimation to artificial rearing is also observable at the host-microbiota level. This result is striking enough as it is now well established that salmonids' microbiota composition is regulated by host genotype (Boutin et al., 2014), population genotype (Dionne et al., 2007), life cycle stage (Llewellyn et al., 2015;Stephens et al., 2016;Yan et al., 2016), and environment (Llewellyn et al., 2015). Importantly, our results suggest that acclimation to artificial rearing overpasses host genotype effect on modeling microbiota composition at both individual and population level.
Environment is an important driver for gut microbiota structuration and composition: Compared to wild parrs, hatchery-reared parrs exhibited a higher bacterial diversity (Figure 1), combined with lower disparity (i.e., most of OTU sharing similar relative abundance), both of which are characteristic of an immature microbial community with low structuration Llewellyn et al., 2015;Sylvain & Derome, 2017 2004). As such, the low structuration of hatchery-reared parrs' microbiota likely results from relaxed selective pressure (Derome, Duchesne, & Bernatchez, 2006;Fisher, 1930), which would translate into a random recruitment of pioneering bacterial symbionts.
The hatchery diet itself would explain such results: Commercial pellets are enriched with nutrients from various origins, including substantial amount of vegetable proteins (e.g., soya) (Feng, Hu, Luo, Zhang, & Chen, 2010;Tanaka, Ootsubo, Sawabe, Ezura, & Tajima, 2004). Enriched food with vegetable proteins and carbohydrates was observed to significantly impact gut microbiota composition in fishes by increasing diversity and richness and more specifically by inducing a significant increase in Firmicutes symbionts, including lactic acid bacteria (LAB), mostly belonging to Lactobacillaceae, Enterococcaceae, and Streptococcaceae (Desai et al., 2012;Gajardo et al., 2016). In our study, commercial diet impact translated into the systematic overdominance of the Lactobacillaceae family in every captive parrs' gut microbiota, whereas this family was absent from the top 20 OTUs composing the wild parrs' gut microbiota.
Consistently, gut bacteria belonging in the Lactobacillaceae family have only been identified in very small amount, if not totally absent, in salmonids coming from the natural environment (Gajardo et al., 2016;Llewellyn et al., 2015). Moreover, plant meal diet (PMD) has been related to the increase in intestinal inflammation and sensibility to various diseases (Krogdahl, Bakke-Mckellep, Roed, & Baeverfjord, 2000) as well as decreasing nutrient digestion and  Water Mucus absorption (Nordrum, Bakke-McKellep, Krogdahl, & Buddington, 2000). Taken together, our results suggest that captive parrs' microbiota composition would therefore be qualified as "generalist" when compared to the highly structured and, thus more specialized, wild parrs' gut microbiota. This more specialized gut microbiota in wild parrs can be attributed to the higher selective pressures occurring in wild rivers, including a more restricted diet, which is mostly composed of insects' larvae, crustacean, and annelids (Bell, Ghioni, & Sargent, 1994). Consequently, the most important environmental factor for the recruitment of intestinal symbionts in teleosts could be associated with diet. This factor could explain most of the heterogeneity of the gut microbiota, the latter being greatly related by its capacity to assimilate nutrients (Tremaroli & Bäckhed, 2012). Therefore, controlling gut microbial symbionts in hatchery could be of prime interest to secure the recruitment of key adaptive microbiota functions of wild parrs.
Regarding both wild-born parrs' populations, gut microbiota composition significantly differed accordingly to their population origin, thus confirming our previous work on wild populations, stating that gut microbiota composition at early life stages is mostly driven by geography (study site) (Llewellyn et al., 2015). Furthermore, net- host-specific genes in salmonids (Boutin et al., 2014), it would be straightforward to investigate further the genetic structure of the Malbaie river Salmo salar population in order to assess whether this population is genetically introgressed with allopatric breeders (i.e., issued from other river populations).
Hatchery-raised parrs issued from both river populations share a similar richness index and are strongly correlated by their composition itself, as shown with both network analysis and PCoA based on weighted and unweighted UniFrac (Figure 2), which clusters both captive parrs populations in a single, isolated, group. Consistently, PERMANOVA on unweighted UniFrac distances showed the weakest, but significant, differentiation between captive parrs from both genetic origins (i.e., Malbaie versus Rimouski), and no differentiation was found when performing the analysis on weighted UniFrac (Table 1). Therefore, even though samples cluster mostly by environment, thus suggesting this factor overpasses genetic origin in controlling gut microbiota composition, it also suggests that genetic origin is still exerting a minimal control on bacteria recruitment. At last, sanitary management in hatcheries, due to high density, impairs microbial environment and gut microbiota (Carlson, Leonard, Hyde, Petrosino, & Primm, 2017;Nakayama et al., 2017), thus amplifying microbiota divergence between hatchery-and wild-born parrs.
In addition to the phenotypic mismatch between captive and wild salmonids from the same genetic population (Araki et al., 2008;Milot et al., 2013;Poole et al., 2003;Stringwell et al., 2014), hatchery-raised parrs are also facing an important microbiota mismatch regarding key microbial symbionts recruited by their wild relatives. Overall, it came out that extensive differences observed between hatchery and river environmental microbiota composition suggest that stocked parrs are exposed to a totally different microbial environment when released into the target river. Knowing that their gut microbiota composition differs from that of wild parrs, exposure to a very different environmental microbial community could lead to an impairment of colonization resistance to wild opportunistic pathogens, thus potentially favoring disease. The sudden transfer from hatchery water to river environment exerts a considerable stress, which is expected to trigger a transient dysbiosis (i.e., altered microbiota activity) giving further opportunities for pathogens to infect fish tissues (Bonga, 1997;Boutin et al., 2013;Seghouani, Garcia-Rangel, Füller, Gauthier, & Derome, 2017). At last, overdominance of Lactobacillaceae in captive parrs is expected to generate a reduced capacity to assimilate nutrients from wild preys. Whether hatchery "imprinting" on microbiome is transient or permanent in stocked fishes after release into the targeted river is yet to be investigated.
Even though two previous studies on rainbow trout observed that the first diet type had no effect on the microbiota composition after a diet shift (Ingerslev et al., 2014;Michl et al., 2017), several studies have also highlighted the long-term effects of the microbiome associated with early life stage cycle diet on the host physiology in human (Indrio et al., 2017;Mischke & Plösch, 2013) and rainbow trout (Ingerslev et al., 2014). Furthermore, Rhossart et al. (2017) demonstrated that host fitness-promoting traits regarding naturally occurring diseases were associated with the "natural microbiome" of wild mice. Indeed, domestic microbiota was associated with greater inflammation and lower resistance to pathogens relatively to wild individual microbiota. Those results are in concordance with other studies reporting that reared fishes fed with a PMD have shown higher level of gut inflammation Nordrum et al., 2000), which is now recognized as an important driver of many diseases (Rajani & Jia, 2018). Because the microbiota composition is proven to be actively involved in several metabolism pathways (Tremaroli & Bäckhed, 2012) and immune responses (Rhossart et al., 2017), investigating the relationship between the host energetic phenotype and the microbiome functional repertoire of captive and wild parrs is crucial to assess to what extent the gut microbiota taxonomic mismatch is actually associated to the loss of microbial functions. Evidence of "metabolic imprinting" (Hanley et al., 2010) related to the microbiota composition at early life stage in hatchery could be suspected to contribute to the mitigated impact of supportive breeding programs, in addition to the potential dysbiosis that may occur during stocking. Therefore, further studies are strongly needed to test whether the taxonomic (and functional) microbial mismatch between hatchery-raised and wild-born salmons could underlie the lower survival of stocked fishes once released into the wild.
Altogether, these results strongly suggest that the extensive discrepancy between hatchery-raised and wild parrs' gut microbiota potentially translates into a phenotypic disadvantage for the former, at least regarding disease resistance and food energetic conversion.
Indeed, the two most important symbionts of the wild parrs' gut microbiota such as Enterobacteriaceae (MWP, RWP) and Clostridiaceae (MWP only) are at very low levels in most captive parrs ( Figure 5) as well as the top 5 OTUs from wild parrs ( Figure 6). Therefore, stocked parrs could have developed different metabolic functions from their wild relatives, regardless of their common genetic origin, which could considerably reduce their fitness in natural conditions. It is interesting that in MWP group, one individual (MWP18) harbored a very distinct microbiota composition. Knowing that Malbaie river is subject to stocking, it becomes even more relevant to identify its origin. However, stocked fish from this river are identified by a clipped caudal fin, but MWP18's fin was not mentioned to harbor such a characteristic. Nevertheless, further investigations are needed to test whether such generalist microbiome from captive parrs provides or not key functions ensuring optimal host physiology in the wild.

| CON CLUS ION
Given that parrs' gut microbiota composition is strongly related to the rearing environment, differences in terms of structure and composition between wild and hatchery-born parrs give valuable information toward improving management of reared fish intended for stocking. Consequently, it becomes even more crucial to investigate the link between environmental and gut microbial communities' taxonomic composition to get new insights on factors driving differences between captive and wild parrs' microbiota and therefore their adaptive ability in a given environment. In light of our results, we support the recommendations stated by Milot et al. (2013) to adopt more natural rearing condition and to release juveniles at a younger stage, before the first feeding occurs. As it may be difficult to exactly mimic the natural conditions in hatchery, the implementation of bacterial ecology in supplementation programs could be one possible avenue to investigate looking forward. Unless the hatchery is connected to the targeted river, one avenue would be to provide beneficial bacteria detected in wild populations to hatchery-reared juveniles through the administration of probiotics. Therefore, further studies are needed to assess how to control the microbiota composition in hatchery and to characterize natural microbiome for each population subjected to supportive breeding programs. To conclude, we strongly believe that implementing host-microbiota evolutionary process and microbial ecology into conservation policies would improve the efficiency of stocking programs for Salmo salar, but also for every teleost species suffering a demographic decline. A P P E N D I X 2 (Continued)