Abstract
Associations between micro- and macro-organisms (forming holobionts) are common and mostly specific in nature. Although microbes might contribute to the holobiont homeostasis, their role in the adaptive capacities of the holobiont remains unclear in most of the living species, and studies testing the links between host fitness and microbiota characteristics are still lacking.
To tackle this issue, we analyzed the microbiota of full-sib progenies from five biparental oyster families in a controlled condition (hatchery) and after transplantation during non-stressful or stressful periods in two geographic regions (Atlantic Ocean and Mediterranean Sea). We used 16S rRNA gene-metabarcoding to study taxonomic compositions of microbial communities. First, comparisons of microbial assemblages showed that environmental conditions highly influenced oyster microbiota. However, different ecological niches existed between oyster tissues and seawater, and even between the five families. Secondly, because two oyster families displayed high mortalities in both stressful periods, we compared their microbiota characteristics to the three other families. All oyster microbiota were highly modified after transplantation and no clear microbiota dysbiosis was observed for the two susceptible families. Moreover, despite similar levels of mortalities in both geographic regions, we identified many differences of microbial assemblages between resistant and susceptible families in the Atlantic Ocean, but few in the Mediterranean Sea. Particularly, susceptible families showed lower diversity of their microbiota than the resistant ones in stressful periods. Secondly, we highlighted that microbiota of susceptible oyster families had higher variability than the resistant ones at the scale of bacterial orders, but also less variability at the scale of OTUs. Lastly, while resistant families had increased relative abundances of cyanobacteria (Subsection III, family I) during the stressful period, susceptible oysters had low stability of Rhodospirillales, Rhodobacterales, Flavobacteriaceae and Erythrobacteraceae.
This study highlighted links between microbiota diversity and low variability of microbial assemblages at higher taxonomic ranks with oyster mortalities. We also identified bacterial taxa related to oyster fitness.
Introduction
Interactions between micro- and macroorganisms are ubiquitous on Earth, from host-pathogen relationships [1] to the widely distributed commensalim or mutualism [2,3]. Previous observations suggested that composition of host-associated microbial communities (hereafter named microbiota) are not random, but might be linked to host taxonomy and physiology. Indeed, specific associations were found for microbiota of phylogenetically distant species of mice, flies, mosquitoes, wasps and hominids [4], but also for microbiota of closely related species of corals [5], hydra [6], or oysters [7]. Moreover, many studies showed that these interactions were not stable when hosts were in stressful conditions or in different physiological conditions. For example, microbiota compositions were found to be very different between healthy and diseased corals [8], or between oysters exposed to different temperatures [9].
Microbial associates might provide benefits to their hosts for survival, but also homeostasis and development [10]. First, a stable microbiota can play a role of physical barriers against pathogens. For example, healthy mice developed Salmonella infection after antibiotic treatments [11]. Other studies highlighted protective effects of (i) secondary endosymbiont of aphids against parasitoid wasps [12], (ii) toxic alkaloids produced by endophytic fungi of grasses against herbivores [13], or (iii) symbiotic bacteria of frogs against pathogenic fungi [14]. Secondly, because microbiota composition can change rapidly, they might represent mechanisms for rapid acclimation of the holobiont to changing environments [15]. In particular, fine-scale modifications of Nematostella vectensis microbiota were observed with environmental variations [16], suggesting that microbiota might participate to the maintenance of holobiont homeostasis. Hence, structure of microbiota might be multilayered with at least two populations of microbes: a stable host-adapted core microbiota and a more variable microbial pool [17]. However, the balance between stability and variability is still poorly studied. In particular, lack of knowledge concerns links between microbiota characteristics and fitness of holobionts facing environmental changes.
To tackle this issue, the farmed oyster Crassostrea gigas is a relevant model, as it filters seawater to feed. As a consequence, all oysters are in contact with the same environmental microbial communities, and microbiota variations are not biased towards different diets [18–20]. In addition, oysters are affected by recurrent mortality syndroms affecting mainly juveniles [21,22]. These mortalities are multifactorial processes due to water temperature and quality, but also development of viral and bacterial pathogens [23–27]. Particularly, different susceptibilities were observed according to oyster physiology and/or genetic backgrounds [28–30], in association with microbiota dysbiosis [9]. Although oyster genetics and microbiota dysbiosis were associated to mortalities, no studies tried to disantangle the links between oyster fitness and microbiota characteristics.
Thus we produced full-sib progenies from five biparental oyster families (see Methods for more details) using genitors from different origins. In particular, males and female genitors were isolated from different geographic regions in France (Atlantic Ocean and Mediterranean Sea), and in different areas (farming areas or sites with lower densities of oysters, and potentially lower concentrations of pathogens). Then we transplanted progenies of each biparental family at two time periods in the Atlantic Ocean or in the Mediterranean Sea. These two time periods were selected according to seawater temperatures in order to study oysters in non-stressful and stressful periods. Particularly, it was previously observed that 16°C was a relevant threshold to define periods of stress, because high mortality rates were observed above this temperature [24,25]. After five days of transplantation, oysters were sampled, seawater was filtered, and we analyzed oyster-associated and seawater bacterial communities using 16S rRNA gene-metabarcoding.
This study aimed at (i) testing environmental effects on oyster microbiota, (ii) comparing oyster microbiota between resistant and susceptible phenotypes for periods of stress, and (iii) testing the balance between stability and variability of microbiota in changing environment.
The identification of such bacterial candidates might help to develop new production techniques for oyster farming in the future.
Methods
Genitors from different geographic origins
Biparental C. gigas oysters families were produced using a recently developed methodology that allowed the production of pathogen-free juveniles [25,27,31]. Four oysters families (F09, F15, F32 and F44) were produced from wild genitor sampled in farming and non-farming area in two geographic regions, the Mediterranean Sea and the Atlantic Ocean (Fig. S1). In addition, family F21 was produced using genitors from a massal breeding program aiming at increasing the resistance of C. gigas oysters against OsHV1-virus. It was performed by breeding disease survivors throughout four generations of selection [32]. Each family (cohort) thus corresponded to the offspring of a biparental reproduction (full-sib progenies). Maturation induction, reproduction and larval breeding were performed as described previously [25,33]. In the larval and post-larval stages, the oysters were fed with the same diet as the genitors at a concentration between 1500-2000 μm3 μl−1 [34]. Before experiments, all oyster families were maintained in controlled condition at the laboratory (Argenton, France) using seawater treated with UV, filtered through 1 μm mesh, and enriched with a bi-specific phytoplankton diet made of Tisochrysis lutea (CCAP 927/14) and Chaetoceros muelleri (CCAP 1010/3) (in equal biomass proportion) at a ratio equivalent to 6% of the oyster dry mass [34]. Finally, all oysters remained free of any abnormal mortality.
Experimental design
About 140 individuals of each oyster family were either kept in the controlled condition or placed in one out of four natural environments for five days. The four natural environments corresponded to two geographic site (Atlantic Ocean and Mediterranean Sea, see Fig. 1), and two time periods. These two time periods coincided to non-stressful and stressful periods according to seawater temperatures (below and above 16°C, respectively) (Table S1), and were confirmed by the observed mortality rates. After five days of transplantation, no mortality occurred and about 15 individuals of each family were flash frozen in liquid nitrogen and stored at -80°C. In addition, 125 oysters per family were also placed in controlled condition to monitor mortality rates. The number of dead oysters was estimated at day 13 (i.e., eight days after the end of transplantation). Moreover, serial filtrations of 5 liters of seawater were performed for the five conditions using porosity of 10 μm (Whatman™ 111115 Nucleopore™ Track-Etch Polycarbonate Membrane filter, 47 mm diameter), 5 μm (Whatman™ 111113 Nucleopore™ Track-Etch Polycarbonate Membrane filter, 47 mm diameter), 0.8 μm (Whatman™ 111109 Nucleopore™ Track-Etch Polycarbonate Membrane filter, 47 mm diameter), and 0.2 μm (Whatman™ 111106 Nucleopore™ Track-Etch Polycarbonate Membrane filter, 47 mm diameter).
DNA extraction, PCR and sequencing
Frozen oysters were ground in liquid nitrogen in 50 ml stainless steel bowls using 20 mm diameter grinding balls (Retsch MM400 mill). The powders were stored at -80°C, and were then used for DNA extractions using the DNA from tissue Macherey-Nagel kit (reference 740952.250) according to the manufacturer’s protocol. In order to improve DNA extractions, we nevertheless added a crushing step, that consisted in an additional 12 minutes mechanical lysis using zirconium beads before the 90 min enzymatic lysis in the presence of proteinase K. The same lysis and DNA extraction protocol was used for the different filters. DNA concentration and quality were checked with Epoch microplate spectrophotometer (BioTek Instruments, Inc.).
Then, the 16S rRNA gene of bacterial communities was amplified and sequenced using the variable V3V4 loops (341F: 5’-CCTACGGGNGGCWGCAG-3’; 805R: 5’-GACTACHVGGGTATCTAATCC-3’) [35]. Paired-end sequencing (250 bp read length) was performed at the McGill University (Génome Québec Innovation Centre, Montréal, Canada) on the MiSeq system (Illumina) using the v2 chemistry according to the manufacturer’s protocol. Raw sequence data are available in the SRA database (BioProject ID PRJNA419907).
Sequence analyses
The FROGS pipeline (Find Rapidly OTU with Galaxy Solution) implemented into a galaxy instance (http://sigenae-workbench.toulouse.inra.fr/galaxy/) was used to define Operational Taxonomic Units (OTU), and computed taxonomic affiliations [36]. Briefly, paired reads were merged using FLASH [37]. After denoising and primer/adapters removal with cutadapt [38], de novo clustering was performed using SWARM that uses a local clustering threshold, with aggregation distance d=3 [39]. Chimera were removed using VSEARCH (de novo chimera detection) [40]. Particularly, this method divided each sequence into four fragments, and then looked for similarity with putative parents in the whole set of OTUs. We filtered the dataset for singletons and we annotated OTUs using Blast+ against the Silva database (release 123, September 2015) to produce an OTU and affiliation table in standard BIOM format. Rarefaction curves of species richness were produced using the {phyloseq} R package, and the rarefy_even_depth and ggrare functions [41]. The alpha diversity metrics (Chao1, Simpson and Shannon) were estimated at the OTU level with the estimate_richness function. Moreover, Pielou’s measure of species evenness was computed using the diversity function in {vegan}. We also used phyloseq to obtain abundances at differents taxonomic ranks (from genus to phylum) (tax_glom function). Multi-affiliation and unknown taxa were removed for the different taxonomic ranks if they corresponded to the addition of multiple OTUs having the same annotation. Because samples had different sequencing depths, we used relative abundances of OTUs and the different taxonomic ranks for subsequent analyses. We computed Bray-Curtis dissimilarities to study beta diversity, i.e., distances between samples for OTU compositions (vegdist function, {vegan}).
Statistical and multivariate analyses
All statistical analyses were done using R v3.3.1 (R: a language and environment for statistical computing, 2008; R Development Core Team, R Foundation for Statistical Computing, Vienna, Austria [http://www.R-project.org]).
Clustering and ordination methods were used to describe and compare composition of microbial communities between samples. Hierarchical clustering (average linkage (hclust {stats})) and principal coordinate analyses (hereafter named PCoA) of microbial communities were computed using Bray-Curtis dissimilarities (vegdist {vegan}). Factors associated to each sample were plotted face to clustering using the heatmap.2 function and the {gplots} package. Variability of oyster microbiota for stressful and non-stressful periods was estimated using PCoA and distances to centroid using betadisper {vegan}.
In order to explain modifications of microbial assemblages, we used redundancy analysis [42] which investigate the variations of the different OTUs under the constraint of differents factors. OTUs were Hellinger-transformed [43] before testing their links with the following variables: family, period (stressful or non-stressful) and site (Atlantic Ocean or Mediterranean Sea) of transplantation. In particular, we estimated the influence of these variables on microbial assemblages using rda and anova.cca functions ({vegan}).
We performed Student’s t-test (t.test {stats}) or non-parametric Wilcoxon test (wilcox.test {stats}) (when normality was rejected with the Shapiro-Wilk test, (shapiro.test {stats})) to compare oyster families for (i) alpha diversity metrics (Chao1, Simpson, evenness and Shannon), and (ii) distances to centroid.
Wilcoxon tests were used to identify candidates (OTUs or orders) related to (i) seawater or oysters, and to (ii) stressful or non-stressful periods.
Finally, we computed enrichment analyses of the significant OTUs to identify over-represented higher taxonomic ranks using Fisher’s exact tests at each taxonomic rank.
For all analyses, the threshold significance level was set at 0.05. P-values were corrected for multiple comparisons using Benjamini and Hochberg’s method [44] (p.adjust, {stats}).
Results
Environmental conditions highly influenced oyster microbiota
Oyster families were produced using wild genitors sampled from sites of high and low density of oyster populations. Full-sib progenies of five biparental families (F09, F15, F21, F32 and F44) were produced and maintained in a controlled condition (laboratory of Argenton, France). In order to study microbiota compositions in changing environments, individuals of each oyster family were then transplanted from the controlled condition into two natural environments at two time periods (non-stressful or stressful) for five days (Fig. 1). For the five conditions of this study, microbial communities were sequenced using the 16S rRNA gene from about 14 individuals of each oyster families (347 oyster-associated microbiota), and from seawater (19 filters). In average, each sample contained ~ 17,550 sequences representing ~1,140 OTUs (Fig. S2-S4, Table S2 and S3).
First, oyster microbiota were highly variable between the different conditions (Fig. 2 and Fig. S5). Particularly, we used redundancy analyses to identify factors (oyster family, period (non-stressful or stressful) and site (Atlantic Ocean and the Mediterranean Sea) of transplantation) that signficantly explained microbiota changes in the Atlantic Ocean and the Mediterranean Sea. We found that the period and site of transplantation were the most important factors (p=0.001) (Fig. 3A and Fig. S5). Hence, the environment highly influenced the composition of oyster microbiota.
Oyster tissues nevertheless exhibited different ecological niches from seawater
Although environmental conditions highly influenced oyster microbiota, we found that oyster tissues constituted different ecological niches from seawater.
First, redundancy analysis also highlighted the effect of oyster family on oyster microbiota (p=0.001) (Fig. 3A). In particular, this variable significantly constrained bacterial assemblages within the five conditions (Fig. 3B). Secondly, analyses of oyster-associated and seawater bacterial communities revealed that both types clustered separately within the five conditions (Fig. 2). Notably, the most abundant OTUs for seawater and oysters belonged to bacterial families Rhodobacteraceae (~16% of total sequences) and Mycoplasmataceae (~11.4% of total sequences), respectively. To identify significant differences between seawater and oysters for OTUs, we computed non-parametric Wilcoxon tests. While we found OTUs with significantly higher relative abundances in seawater compared to oysters (Table S4 and Fig. S6), we did not observed OTUs significantly associated with oysters. This lack of significance might result from our stringent analysis procedure. Indeed, the combination of Wilcoxon tests and p-values corrections for multiple comparisons only kept highly significant OTUs. However, it suggested that most oyster OTUs came from the environment, although oyster tissues constrained their growth. Among them, 48 OTUs from the Proteobacteria, Bacteroidetes, Actinobacteria and Chloroflexi phyla were significantly higher in seawater for the four natural environments. Then we performed enrichment analyses of significant OTUs in each condition to identify taxonomic ranks that were over-represented in seawater. We found in particular that Proteobacteria from Rhodobacteraceae, Litoricolaceae, SAR86 and SAR116 families were enriched in seawater of natural environments.
Overall, these results suggested that oyster tissues constrained bacterial assemblages differently (i) from seawater, and even (ii) between oyster families having differents genetic backgrounds.
Oyster families derived from farming area showed low mortality rates
Then we particularly focused our attention on oyster microbiota and mortality rates in response to period of stress.
While no mortalities were observed for control and non-stressful periods in both the Atlantic Ocean and the Mediterranean Sea, many oysters died during stressful periods (in May and July for the Mediterranean and Atlantic sites, respectively). In particular, two oyster families (F15 and F32) displayed high mortality rates (hereafter named HM) (from 87 to 100%, n=125 oysters per family) in both geographic regions, whereas the three others (F09, F21 and F44) showed lower mortality rates (hereafter named LM) (from 0 to 52%, n=125 oysters per family) (Fig. 4). Notably, these two clusters of phenotypes were not linked to geographic origin of genitors (Atlantic Ocean or Mediterranean Sea), but rather to the fact that they originated or not from farming areas. This suggested that selective pressures by pathogens were higher in farming areas.
Microbiota of LM oysters were more diverse and stable in stressful periods
Although high mortality rates affected HM oyster families, no clear pattern of microbiota dysbiosis was observed during stressful periods (Fig. 2). Oysters were placed for five days in the field, because previous observations showed that this time period was long enough for disease development, but not for mortality [27]. As a consequence, disease possibly just started after five days in our experiment, and we expected to find fine-scale changes of microbial communities rather than complete dysbiosis. We thus studied more precisely microbiota of LM and HM using alpha (Chao1, Simpson, evenness and Shannon) and beta (Bray-Curtis dissimilarities) diversity indices.
First, analyses of alpha diversity (Table S5) showed that LM had higher values of Simpson, evenness, and Shannon indices than HM in control, and in the stressful period for the Atlantic Ocean (Table 1). Moreover, LM had also higher values of evenness in the Mediterranean Sea, but we did not identify significant differences for Simpson and Shannon.
Secondly, we used Bray-Curtis dissimilarities and PCoA to estimate the variability of bacterial assemblages for LM and HM at different taxonomic ranks using distances to centroid as a proxy. We found that oyster-associated microbiota were less variable for LM than HM in controlled condition for many taxonomic ranks (p<0.05, Wilcoxon test; except for family, class and phylum) (Table 2). Furthermore, while no significant differences were observed between LM and HM during the stressful period in the Mediterranean Sea, LM microbiota were also less variable than HM for the stressful period in the Atlantic Ocean considering higher taxonomic ranks (from order to class; p<0.01, Wilcoxon test), but the opposite trend was observed for OTUs in the same region (p=0.02, Wilcoxon test).
Hence, all these results suggested that although no clear pattern of microbiota dysbiosis was observed during stressful periods, microbiota of HM families had lower diversity than LM in the Atlantic Ocean. In addition, HM microbiota assemblages were more variable than LM at higher taxonomic ranks, and less variable for OTUs within this region. In contrast, few significant differences were observed in the Mediterranean Sea for microbiota diversity and assemblage variability. It suggested that disease development did not start before oysters were sampled after five days of transplantation in this region.
Microbiota of LM oysters had significant increase of cyanobacteria
Our results revealed that oyster microbiota of LM showed higher variability and stability than HM in the Atlantic Ocean for OTU and orders, respectively. As a consequence, we used Wilcoxon tests to compare taxa of non-stressful and stressful periods within this region. In particular, we tried to identify parallel changes between oyster families of both LM and HM.
First, LM shared increased abundances of 16 OTUs in stressful period (Fig. 5A and Table S4), and HM had similar decreases for 29 OTUs (Table S4). In particular among them, five cyanobacteria and three Flavobacteriaceae were significant for LM, and 13 Flavobacteriaceae and five Erythrobacteraceae significantly decreased for HM. Then enrichment analyses of the significant OTUs for each oyster family highlighted that variability mostly concerned cyanobacteria (Subsection III, Family I) for all LM. In contrast, decreased OTUs overrepresented Flavobacteriaceae and Erythrobacteraceae families for all HM. Furthermore, we found that although LM had similar increases for 3 bacterial orders in stressful period (Acidobacteria Subgroup 3, Holophagae Subgroup 10, Phycisphaerales), 4 other orders were stable compared to HM (Fig. 5B). Particularly, the relative abundances of Rhodospirillales increased in stressful period for HM, and the fraction of Rhodobacterales, Sphingomonadales, and Cytophagia (Order II) significantly decreased.
Discussion
Oyster microbiota were influenced by the environmental conditions
We found that the environment highly constrained oyster microbiota. This was not surprising because oysters are sessile organisms that filter seawater to feed. Hence, because they face continuous environmental changes, it is expected that microbiota reaches different stable states [45]. Moreover, stable interactions across environments may not be favored in oysters, because they do not have specialized anatomical structures hosting microbiota [17], such as nodules of legume plant [46], or light organ of bobtail squid [47]. As a consequence and according to our observations that oyster microbiota is highly variable, we hypothesize that oyster bacterial assemblages have high diversity, high horizontal transmission of facultative symbionts, high efficiency to purge deleterious mutations, and that they may facilitate rapid acclimation to environmental changes [17].
Mortality rates were linked to the origin of genitors
Oysters were transplanted for five days in the field. According to previous observations [27], this time period was considered sufficient for disease development during stressful periods, and no mortalities were observed before oysters were sampled. As a consequence, we could analyze microbiota modifications linked to perturbations (i.e., environmental changes), and not to moribund individuals.
To date, most oyster diseases were associated to Vibrio strains and herpesvirus (OsHV-1 μvar) [27,48,49], although other studies highlighted the role of Roseovarius crassostreae [50,51], Tenacibaculum soleae [52], and possibly Chlamydiales [53] and Arcobacter sp. [9,52]. Here we did not analyze the causal origin of diseases, but we observed similar mortality rates in both geographic regions. In particular, the five oyster families formed two clusters of low and high mortality rates during stressful periods in both the Atlantic Ocean and the Mediterranean Sea. Moreover, this clustering was highly associated to the origin of genitors from farming areas or sites with lower oyster densities. Thus these results highlighted that resistance to mortalities was (i) as previously observed influenced by oyster genetics [29], (ii) replicable between two French coastal regions, and (iii) higher for oyster families derived from genitors sampled in farming areas. Because aquaculture uses very high oyster densities, and is affected by many mortality syndroms [21,54], it is likely that selective pressures from pathogens are higher in these areas. The low mortalities observed in this study for oyster progenies derived from genitors of farming areas thus suggested that past events might influence their current immune responses. In particular, these different mortality levels might be linked to transgenerational immune priming [55], and/or to natural selection of genetic variants.
Stability for bacterial orders and variability of OTUs were linked to oyster mortalities
No clear patterns of microbiota dysbiosis were observed in both stressful periods for HM oyster families, suggesting that oysters were sampled before the massive community structure disruption that occurred for moribund and dead oysters [9]. In order to identify fine-scale modifications of oyster microbiota during disease development, we thus performed analyses of alpha diversity indices and variability of microbial assemblages using distances to centroid as a proxy.
Higher values of alpha diversity indices (Simpson, evenness, Shannon) was the most striking parallel found for LM compared to HM for the stressful period of the Atlantic Ocean. Particularly, this result highlighted that OTU abundances within LM microbiota were more equally distributed than HM microbiota. For example, evenness may possibly have important effects on oyster fitness, because it was already found to be positively linked to ecosystem productivity [56,57], functional stability [58,59], and invasion resistance [60,61]. Furthermore, for the stressful period of the Atlantic Ocean, LM were more stable than HM for higher taxonomic ranks, and more variable for OTUs. Similar results of stability were also obtained for hydra [16], and in oyster hemolymph [9]. However, we precisely identified the stable taxonomic ranks in this study (i.e., from order to class).
Surprisingly, we did not find significant differences between HM and LM in the Mediterranean Sea for variability of microbial assemblages, and only evenness and Chao1 were significant for alpha diversity indices. Because oysters were all sampled after five days of transplantation, it is very likely that disease development was different between both geographic regions, and even between oyster families. Thus first signs of microbiota dysbiosis were possibly missed in this study if they occurred later in the Mediterranean Sea. For instance, it was observed in a previous work that primate microbiota was altered at the very last stages of a viral infection [62].
Rhodospirillales, Rhodobacterales, Flavobacteriaceae, Erythrobacteraceae and cyanobacteria are putative candidates associated to oyster fitness
Because stability and variability of microbiota are putative beneficial characteristics of the holobiont [11,16], we tried to identify taxa related to LM and HM in the Atlantic Ocean where differences for microbiota diversity and stability were observed. We particularly focused our attention on parallel changes between oyster families, as they could point out factors involved in adaptive processes [63–65]. Furthermore, parallels are likely to be reproducible in future studies.
Similarly to C. virginica oysters from Hackberry Bay [66], Mycoplasmataceae was the dominant taxa of oyster microbiota in this study. Despite this dominance, stability of the Rhodospiralles and Rhodobacterales orders, and Erythrobacteraceae and Flavobacteriaceae families might be the most important factor for oyster fitness. Indeed, HM had high significant changes for both orders between non-stressful and stressful periods. Moreover, enrichment analysis showed that Erythrobacteraceae and Flavobacteriaceae families were over-represented by OTUs that significantly decreased between both time periods. Among them, Erythrobacteraceae were already isolated from marine invertebrates [67], and Rhodospiralles were found in diseased tissues of Platygyra carnosus corals [68]. Conversely, presence and variability of cyanobacteria OTUs might be linked to rapid acclimation of oyster holobiont to environmental changes. Indeed, increased OTUs in stressful period for LM over-represented cyanobacteria (Subsection III, Family I). Although cyanobacteria had a negative effect on other marine invertebrates such as scleractinian corals [69], they were already observed at high relative abundances within oyster microbiota [70]. Particularly, they were abundant in the digestive gland, connective tissue, mantle, and gonad of oysters [71]. Because cyanobacteria persisted in oyster tissues without signs of alterations, a possible endosymbiotic relationship was even proposed [71]. Notably, cyanobacteria were also negatively correlated to culturable Vibrio abundance in free-living microbial communities [72]. Hence, cyanobacteria might play a role of barrier for competitors/pathogens, because vibrios are the main bacterial pathogens of oysters [54],. In particular, photosynthetic activity of cyanobacteria could inhibited pathogen growth through the production of reactive oxygen species, such as observed for Plasmodium infection in Anopheles gambiae [73].
Conclusions
Our study highlighted that oyster microbiota were highly influenced by environmental conditions. However, different ecological niches existed for bacterial communities between seawater and oyster tissues, and even between oyster families having different genetic backgrounds. Although similar mortality rates were observed for each family in the two geographic regions 12 days after transplantation, disease development during stressful periods were possibly more advanced at the time of sampling in the Atlantic Ocean compared to the Mediterranean Sea. Thus, we recommended that future studies control disease development in order to compare microbiota of different individuals at the same stage of infection. For the Atlantic Ocean, we identified many links between microbiota characteristics and oyster mortalities, and particularly that LM had higher diversity, higher variability of bacterial OTUs and higher stability of bacterial orders and classes compared to HM. Furthermore, Rhodospiralles and Rhodobacterales orders, as well as Erythrobacteraceae, Flavobacteriaceae, and cyanobacteria (Subsection III, Family I) families were putative candidates associated with oyster fitness.
However, we did not know whether these differences were causes or consequences of the observed mortalities. Hence, experimental microbiota manipulations might help in the future to understand if such modifications favored oysters facing recurrent mortality syndroms. For example, they might lead to develop farming procedures using bacterial probiotics.
Declarations
Ethics approval and consent to participate
Not applicable.
Consent for publication
Not applicable.
Availability of data and materials
The datasets generated during the current study are available in the Sequence Read Archive repository under BioProject ID PRJNA419907 (to be released upon publication).
Competing interests
The authors declare that they have no competing interests.
Funding
CC benefited of post-doctoral fellowships from CNRS and IFREMER. This work was supported by the French National Research Agency ANR, project ANR-14-CE19-0023 DECIPHER (coordinator G. Mitta) and the DHOF program of the UMR5244/IHPE (http://ihpe.univ-perp.fr/en/ihpe-transversal-holobiont/). This project has received funding from the European Union’s Horizon 2020 Research and innovation programme under grant agreement N° 678589.
Authors’ contributions
CC, JDL, JME, YG, GM and ET were involved in the study concept and design. BP was involved in the generation and maintaining of all animals used in this study. JDL, BP, JME, YG, GM and ET were involved in the collection of samples. CC, JDL, AL and ET were involved in data acquisition and analysis. CC and ET drafted the manuscript and all authors contributed to critical revisions and approved the final manuscript.
Acknowledgements
We thank IHPE members for stimulating discussions. We are grateful to the genotoul bioinformatics platform Toulouse Midi-Pyrenees and Sigenae group for providing help and computing resources thanks to Galaxy instance http://sigenae-workbench.toulouse.inra.fr.