The role of diversity in mediating microbiota structural and functional responses in two sympatric species of abalone under stressed withering syndrome conditions

Background: The withering syndrome (WS) is an infectious disease initially affecting the gastro-intestinal tract (GI) of wild abalone populations of the coasts of Baja California. In spite of its high incidence, structural and functional changes in abalone GI microbiotas under WS-stressed conditions remain poorly investigated. Moreover, it is equally uncertain if interspecic microbiota features, such as the presence or absence of certain bacterial species, their abundances, and their functional capabilities, may prevent or at least lead to different microbiota responses. Healthy Haliotis fulgens and Haliotis corrugata from Baja California Sur (Mexico) harbor species-specic structural and functional microbiota proles; hence, we hypothesize a distinctive microbiota response under WS-stressed conditions. Here, we compared both the structural arrangements and functional capabilities of healthy and dysbiotic microbiotas using 454 pyrosequencing high throughput sequencing technologies and PICRUSt v.2 outputs, respectively. Results: Our ndings suggest that the extent to which WS may involve structural and functional changes in GI microbiotas is contingent on the microbiota diversity itself. Indeed, we report signicant structual alterations in the less complex microbiotas of H. fulgens, which in turn led to a signicant downregulation of several metabolic activities conducted by GI bacteria. Conversely, the effects of WS were marginal in more complex bacterial communities, as in H. corrugata, in which no signicant structural and functional changes were detected. Conclusions: Our results provide new insights concerning the role of microbiome diversity in abalone health and the etiology of WS. Notably, complex bacterial communities appear to be less affected by WS than less complex microbiotas. Moreover, our insights suggest that structural changes observed under WS-stressed conditions may be considered stochastic, as predicted by the Anna Karenina principle, and result in the downregulation of several ecological functions conducted by GI bacteria. Overall, our results support the hypothesis that the occurrence of WS may be associated with shifts in GI microbiotas. Moreover, we propose that susceptibility WS that species


Background
Wild abalone (Haliotis spp.) populations are in decline worldwide due to overexploitation and natural causes, such as epidemic disease outbreaks [1]. Reductions in the populations of these species may result in both strong economic losses and unpredictable ecological impacts. Adult abalone are considered ecosystem engineers, as they graze macro-and microalgae and thus maintain open habitats that may be exploited by other organisms [2][3][4]. Moreover, these gastropods support valuable sheries in many countries worldwide [5][6][7]. The peninsula of Baja California harbors seven exploitable abalone species [8], although the harvest is focused almost entirely on the blue abalone Haliotis fulgens (HF) and the yellow abalone Haliotis corrugata (HC) [8,9]. In recent decades, the main disease contributing to the decline of wild Mexican abalone populations has been withering syndrome (WS), a chronic and fatal disease that has been responsible for moderate and massive mortality events [10,11]. Notably, the degree of susceptibility and occurrence of this disease varies among abalone species and the causes of WS remain poorly understood [1,3,7].
Early studies of WS have linked the occurrence of this syndrome with infection by a GI intracellular Anaplasmataceae bacterium [12] of the order Rickettsiales named Candidatus Xenohaliotis californiensis (CXc) [13]. The primary ndings supporting the connection between WS and CXc have resulted from the identi cation of this novel bacterium using microscopic (in situ hybridization) and molecular (16S rRNA gene) techniques with histological inclusions of infected abalone [13]. Moreover, diseased abalone with clinical signs of WS have been found to regain their original health status after treatment with a Rickettsiales-killing antibiotic [13]. Nevertheless, recent studies have weakened such correlations, and CXs-WS co-occurrence remains a topic of debate. Speci cally, histological analyses have revealed the presence of CXc inclusions in both diseased and healthy abalone with similar prevalence ranges [14,15]. Moreover, the absence of CXc in WS-stressed abalone has also been recorded, suggesting that WS may be trigged by additional pathogenic microbes [16][17][18].
Novel insights regarding animal microbiotas have revealed that microorganism symbiosis is ubiquitous in metazoans, and the understanding of such partnerships is rede ning our knowledge of animal biology [19]. Bacterial microbiotas carry out essential physiological activities that in uence the health, development, disease susceptibility, and the behavior of their hosts [20][21][22]. In this regard, the study of gastro-intestinal (GI) microbiotas has become a growing area of research, and it has become increasingly accepted that GI microbiotas and their associated gene pools play pivotal roles in the overall health of host species [23]. Thus, understanding changes in GI microbiomes and the consequences for their hosts represents one of the main goals in microbial ecology [20].
In this context, the Insurance hypothesis proposed by Yachi and Loreau [24], suggests that the community-level variance would decrease as diversity increases due to both, the possibility that different species may respond differentially to environmental changes and with increased redundancy levels (i.e., different species may carry out the same ecological function) [24][25][26]. The latter appears to be particularly relevant in GI microbiotas, as high diversity has direct consequences on the health status of the host [27]. Moreover, shifts in associated bacterial communities have been linked to epidemiologic outbreaks in different marine invertebrates [28,29]. Furthermore, recent observations suggest that, under normal circumstances, microbiome arrangements tend to favor a small number of bene cial con gurations. Accordingly, stressors and disease will lead to stochastic rather than deterministic changes in the microbiome, which consequently result in a higher number of unpredictable microbiota con gurations. These observations have led to the Anna Karenina principle (AKP), whose name comes from the opening sentence of Anna Karenina by Leon Tolstoy: "All happy families are all alike; each unhappy family is unhappy in its own way." In 2017, Zaneveld et al. [20] rephrased this sentence and created an ecological host-associated microbiome hypothesis: "All healthy microbiomes are similar; each dysbiotic microbiome is dysbiotic in its own way." Although these theories appear common and important for the overall health of the host, little information is available for invertebrate microbiomes, particularly those from marine environments [30][31][32]. In abalone species, only an exiguous number of studies have thus far evaluated structural alterations in the microbiotas of abalone with WS [32]. Consequently, the extent to which WS may be involved in structural and functional alterations in abalone GI bacterial communities remains unclear. In addition, the hypothesis that species-speci c microbiota arrangements among abalone species may lead to different abalone responses to WS remains unexplored. Recently, we compared and reported the structural arrangements and functional capability pro les for the GI microbiotas of healthy HC and HF [33]. The microbiotas of HF appears to be less complex compared to that of HC, as it is mainly dominated by Mycoplasma spp., whereas the microbiotas of HC appears to be composed by a higher number of bacterial operational taxonomic units (OTU) species [33]. Here, we posited that these microbiota characteristics may lead to microbiota-speci c responses in WS-stressed abalone and that dysbiotic microbiotas would be characterized by a high number of new stochastic bacterial con gurations, as predicted by the aforementioned ecological theories.

Methods
Sample collection, DNA extraction, and PCR ampli cation  Table S1). In the eld, landed abalones were morphologically inspected to identify animals bearing morphological signs of WS including low mobility, color alterations, and pedal muscle and mantle retraction, as proposed by Friedman [34]. After the morphological examination, approximately 30 mg of postesophageal tissue was collected from each of the 107 abalone included in the study (n = 46 for HC; n = 61 for HF). Specimens were immediately transferred to sterile 1.5-ml microcentrifuge tubes containing molecular grade ethanol until further analysis.
In the laboratory, DNA was extracted and puri ed from preserved tissues using a DNeasy blood & tissue kit (Qiagen, Valencia, CA, USA) following the protocol of the manufacturer. PCR ampli cation of an internal fragment of the 16S rRNA gene spanning the V1-V3 regions (~500 bp) was conducted using universal eubacterial primers 28F (5' -GAGTTTGATCNTGGCTCAG -3') [35] and 519R 16S rRNA gene library preparation PCR amplicons were sent to the Research and Testing Laboratory (Lubbock, TX) for pre-sequencing preparation and Roche 454 pyrosequencing. Brie y, the amplicons were tagged using Roche 454 adaptors and multiplex identi er (MID) tags for each organism, following the bacterial tag-encoded FLX amplicon pyrosequencing (bTEFAP) approach of Dowd et al. [37]. Roche 454 pyrosequencing was carried out in a GS FLX Titanium platform. The raw 16S rRNA gene libraries were processed as follows: 1) the SFF les were formatted into FASTA sequences and quality les; 2) the reads were demultiplexed from SFF les using the Roche s nfo tool; 3) low quality bases (Phred score < 25), primers, and barcode sequences were removed using a Research and Testing Laboratory internal quality trimming algorithm; and 4) dereplication was carried out using the USEARCH algorithm [38]. The raw reads were deposited in the National Center for Biotechnology Information (NCBI; BioProject: PRJNA494699, accession number: The returned reads were processed and analyzed using Quantitative Insights Into Microbial Ecology v. 2019.4 -2019.7 (QIIME2) [39]. After importing into QIIME2, the reads were de novo clustered into OTUs at 97% identity using the QIIME vsearch cluster-features-de-novo plugin [40]. Chimeras and singletons were detected and removed using UCHIME [41] and the lter-features plug in implement in QIIME2 v.2019-7, respectively. We organized two control steps in order to minimize PCR amplicon noise. Initially, the taxonomic assignments of the OTUs were conducted via a classify-consensus-blast [42] trained on 16S rRNA gene OTUs clustered at 99% similarities within the Silva_132 database [43], and unreliable sequences with no match in the SILVA_132 database (e.g., unassigned) were removed before further analyses. Additionally, we removed reads with frequencies inferior to 0.005% [44]. As a result, OTUs with total read numbers < 4 were removed from both the HF and HC data sets.
Finally, the reads were rare ed by randomly subsampling at 702 and 644 reads for HF and HC, respectively. We selected those sampling depths in order to minimize the loss of both specimens and bacterial diversity. The selected rarefaction values were as close as possible to the asymptotic plateau of the rarefaction curves for both species (Additional le 2: Fig. S1a and Fig. S1b), which also allowed for the retention of most abalone. The number of removed reads and OTUs during each quality control step, as well as the nal numbers of reads and OTUs used as input in downstream analyses are reported in Additional le 1: Table S2.

Ecological analyses
To evaluate how exhaustively the bacterial communities were sampled, rarefaction curves of the detected OTUs were generated using the diversity alpha-rarefaction plugin implemented in QIIME2 v. 2019.7 [39] for each abalone. Also, the number of OTUs obtained was compared against the non-parametric species richness estimator Chao 1 (Additional le 2: Fig. S1c) [45].
Microbiome community structure was evaluated by principal coordinate analysis (PCoA) based on Bray Curtis, Jaccard, and weighted and unweighted phylogenetic Unifrac distance metrics obtained from the abundance of each rare ed OTU. The PCoA results were visualized with EMPeror [46]. Statistical differences were evaluated by a permutational multivariate analysis of variance (PERMANOVA) based on 4999 permutations using the beta-group-signi cance plugin of QIIME2 v.2019.7 [47]. Additionally, the multivariate dispersion based on OTU abundance in morphological categories was estimated by a permutation analysis of multivariate dispersion (PERMDISP) with 4999 permutations implemented in PRIMER+P v.6 [48].
At the OTU taxonomic level and with the aim to include all shared bacterial OTUs and/or exclusive nonrare ed OTUs between morphologically healthy and WS-stressed abalone, the OTUs were visualized using Venn diagrams created with open web-software InteractiVenn [49]. Also, the differential abundance of rare ed OTUs between healthy and WS a icted abalone was determined by linear discriminant analysis (LDA) effect size (LEfSe) [50]. Finally, to determine which OTUs primarily contributed to the dissimilarity between both proposed categories, a similarity percentage (SIMPER) analysis based on rare ed OTU abundance was conducted using PRIMER+P v.6 [48].

Functional prediction
We used PICRUSt v.2.2.0 beta [51] to predict both the potential functional capabilities and the contributions of distinct OTUs to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways. Brie y, the OTUs were normalized by the predicted 16S rRNA gene copy number of each OTU. To implement quality control, we computed the nearest sequenced taxon index (NSTI). Brie y, the NSTI evaluates the prediction accuracy of PICRUSt since it re ects the average genetic distance (measured as the number of substitutions per site) between each OTU against that of the 16S rRNA gene from a reference genome [51][52][53]. Following the suggested guidelines [51], we eliminated OTUs with NSTI values higher than 2.
To visualize the functional dispersion of morphologically healthy and WS-stressed abalone, PCoA plots based on fourth root normalized metabolic KEGG pathway counts were carried out using STAMP v. 2.1.3 [54]. Signi cant differences among enriched pathways (effect size > 2) were evaluated by a Welch test implemented in STAMP v. 2.1.3 [54].

Bioinformatics
Pyrosequencing returned 243,636 and 152,354 demultiplexed 16S rRNA reads for HF and HC, respectively. Among them, 228,058 and 140,958 reads met quality criteria and were assigned to 790 and 596 OTUs for HF and HC, respectively. Rarefaction curves and Chao1 estimates suggest that we were able to sample the major fraction of the bacterial communities in both species given the asymptotic shape of the curves, the similarity between the asymptotic number of taxa for HF and HC, and the Chao1 estimates for both species (Additional File 2: Fig. S1c). All samples passed data quality control tests; however, 4 HF and 3 HC abalone individuals returned a lower number of reads than the selected rarefaction depth threshold and were excluded from subsequent analyses. After read-rarefaction a total number of 590 and 448 bacteria-OTUs were obtained for HF and HF, respectively (Additional le 1: Table S3 and Table S4).
Overall, WS-stressed organisms of both species were characterized by smaller contributions of Mollicutes to the total number of OTUs and read percentages. Nevertheless, both species differed in read-abundance patterns of non-Mollicutes taxa between healthy and WS-stressed specimens. Speci cally, the majority of non-Mollicutes taxa increased congruently with regard to the number of OTUs and the percentage of total reads in WS-stressed HF abalone (Table 1). In contrast, the number of OTUs and read percentages from Mollicutes as well as most bacterial taxa decreased in WS-stressed HC abalone. Furthermore, these alpha-diversity changes were not followed by congruent changes in the number of reads. For example, Alphaproteobacteria showed a slight decrease in diversity (7% of OTUs) but contributed 2-fold the total number of reads in WS-stressed HC compared to that of Spirochaetia, Fusobacteriia and Gammaproteobacteria (Table 1). Notably, signi cant shifts in microbiome compositions were observed between healthy and WS-stressed HF abalone (PERMANOVA; pseudo-f: 3.1358; p < 0.0322), whereas no consistent structural changes were observed in healthy and diseased HC abalone.
Similar abundance patterns were also observed at the OTU taxonomic level. WS-stressed HF abalone possessed a higher fraction of exclusive bacterial OTU species than morphologically healthy animals ( Fig. 1a and b), which resulted in increased microbial diversity (H′ healthy= 2.4 and H′ WS-stressed = 2.7).
The LEfSe analysis revealed that 33 Mollicutes OTUs were either exclusive or predominant in either healthy or WS-stressed HF abalone (Additional le 1: Table S5). Moreover, the results of the SIMPER analysis revealed that total cumulative dissimilarity was mainly explained by the previous Mollicutes OTUs (55%) rather than by the presence of pathogenic bacteria, such as CXc (Additional le 1: Table S5). The PCoA analysis based on Bray-Curtis and weighted Unifrac distance metrics revealed a signi cant structural differentiation of the bacterial communities between healthy and WS-stressed abalone (PERMANOVA; pseudo-f: > 4.88; p < 0.0018; Fig. 2a and Fig. 2b). In addition, WS-stressed HF presented a signi cant increase of multivariate dispersion (51.50 ± 1.5) compared to that of healthy organisms (43.50 ± 2.7; PERMDISP, F: 6.3706, p = 0.0316). On the contrary, a higher number of bacterial OTU species were detected among healthy HC abalone compared to that of WS-stressed HC despite no signi cant structural changes being observed among the microbiotas of HC abalone (Fig. 2c and 2d; Fig.3a and b). Moreover, differences in the presence or absence and/or OTU read abundance were observed in only 18 OTU species among healthy and WS-stressed HC, most of which were assigned to Mycoplasma (Additional le 1: Table S6). Finally, similar multivariate dispersion patterns were observed between healthy (56.30 ± 1.3) and WS-stressed (56.33 ± 2.6) HC abalone.

Functional microbiome capabilities
Functional predictions where estimated for 524 and 529 bacterial OTU species for HF and HC, respectively. The remaining OTUs (n = 266 and n = 67 for HF and HC, respectively) were removed as they presented NSTI values > 2. Overall, a total of 152 and 141 functional pathways were predicted for HF and HC, respectively (Additional le 1: Table S7 and Table S8). According to KEGG hierarchical level 1, a similar abundance of functional pathways was observed among the tested abalone species and morphological conditions (Additional le 1: Table S7 and Table S8). Microbiota functional capability was mainly involved in metabolic pathways (up to 73%) with Mycoplasma being by far the main bacterial genus responsible for those functions in both healthy and WS-stressed abalone. Speci cally, the bacterial genera presenting higher functional contributions (threshold > 1%) were Mycoplasma (92.9%), Vibrio (2.7%), and Mesoplasma (1.4%) in HF abalone and Mycoplasma (63.56%), Spirochaeta (11.87), Mesoplasma (4.51%), and Vibrio (3.54%) in HC.
The PCoA analyses performed using the normalized count numbers of the pathways (or KEGG level 3) revealed a signi cant functional split between healthy and WS-stressed HF abalone ( Fig. 4a; PERMANOVA; pseudo-f: 6.77; p = 0.0018). This separation was explained by 79 of the 152 predicted pathways that presented signi cant differences in the mean proportion of WS-stressed and healthy HF abalone (Fig. 4b). Notably, under-expressed pathways (n = 59; Fig. 3b) in WS-stressed abalone were mainly related with metabolic activities, such as carbohydrate metabolism, amino acid metabolism, glycan biosynthesis, metabolism, and the metabolism of cofactors and vitamins (KEGG level 2; Fig. 3b). Most of the over-expressed genes (n = 20) in WS-stressed HF abalone were related with antibacterial and/or defense mechanisms like cellular apoptosis and antibacterial compound production (Fig. 3b). In contrast, no consistent differences were observed between the predicted pathways of WS-stressed and healthy HC abalone (PERMANOVA; pseudo-f: 2.40; p = 0.089). Moreover, the Welch test results revealed no signi cant differences in the normalized count number of the pathways between healthy and WSstressed HC abalone.

Discussion
Overall, our ndings suggest that WS may involve speci c response patterns with regard to the structural and functional capabilities of the tested microbiotas. Notably, the extent to which WS may be associated with structural changes strongly depends on the con guration of the original unstressed microbiotas, as we observed signi cant alterations in less complex microbiotas (e.g., HF) while marginal changes were detected as bacterial diversity increased (e.g., HC). In this context, the results obtained between healthy and WS-stressed HF strongly support the AKP, as a higher number of both bacterial OTU species and alternative con gurations were observed in dysbiotic microbiotas [20]. Conversely, WS-stressed HC abalone generally presented a reduction in bacterial α-diversity.
These distinctive structural microbiota responses resulted in unequal shifts in functional capabilities. Speci cally, no consistent functional alterations were observed in HC, whereas WS-stressed HF abalone showed a signi cant downregulation of several molecular pathways. The simplest explanation of this downregulation may be found in the Insurance hypotheses that have stated that higher diversity leads to more stable systems, which has been associated with species-speci c responds to perturbation and functional redundancy [24][25][26]. Indeed, in a redundant system, the removal of species may not necessary result in a loss of functionality of the entire system, as other taxonomic groups may carry out the ecological tasks of the species that are removed [27]. This was particular evident in HC, where a higher number of bacterial OTU species were found to enrich the same KEGG pathway [33]; thus, reduced bacterial diversity does not necessarily result in a functional misregulation of the entire microbiota. On the other hand, from a functional prospective, the increase in α-diversity observed in WS-stressed HF should be considered to be stochastic, as predicted by the AKP. Indeed, the acquisition of additional bacterial taxa did not result in either an increase in functional capabilities or in the upregulation of predicted ecological pathways.
These results may be used to elucidate some aspects of WS that remain poorly understood. For instance, the different degrees of susceptibility to WS that have been observed among abalone species [15,34,55] may be related with the complexity level of each abalone species GI bacterial community. In this context, both a lower microbial diversity and a higher degree of susceptibility to WS (up to 100%) have been previously reported for HF; conversely, the higher bacterial α-diversity of HC may explain its lower susceptibility to WS (up to 60%) [33,34]. Recent advances have indicated that more complex microbiotas lead to more e cient systems and improved health conditions [27]. For example, a high microbiota diversity may reduce the susceptibility of individuals to infectious diseases through both direct and indirect mechanisms. Indeed, resident microbiotas may outcompete pathogens for space, metabolites, and nutrients [23,56] or inhibit and even kill pathogens via metabolic byproducts (e.g., bacteriocins, acids, and peptides) [57,58]. This may be particularly important in abalone, as higher interspeci c microbiota diversity is generally explained by an increase in Alphaand Gammaproteobacteria [6, 33,59], both of which represent the main taxonomic groups that produce antimicrobial compounds in abalone species [60,61]. Thus, we posit that the lower susceptibility to WS observed in HC may be explained by both a high prevalence of Alphaand Gammaproteobacteria [33] and the production of associated byproducts. Nevertheless, additional efforts are needed to validate this hypothesis.
Despite the aforementioned structural and functional differences, the taxonomic differences between the healthy and WS-stressed microbiotas from both abalone species were mainly attributed to changes in Mycoplasma abundance rather than the presence of pathogenic bacteria, such as CXc. In this context, the GI microbiotas of all abalone were dominated by Mycoplasma and consequently the differences in the abundance of this bacterial group explained the largest dissimilarity among healthy and WS-stressed specimens. Moreover, although a signi cant increase of CXc was observed in WS-stressed HF abalone, its contribution to the total dissimilarity among morphological categories was generally low, ranging from 0.03-2.87%. In addition, Proteobacteria signi cantly increased in abundance in WS-affected abalone, suggesting that the increase in those bacteria may have resulted from opportunistic responses under unhealthy conditions. Together, these abundance patterns may indicate that the occurrence of WS may be primarily associated with a comprehensive change in the bacterial community or at least with a decrease in the number of certain Mycoplasma species [32], which in turn may result in functional shifts in dysbiotic microbiotas. This may be particularly true considering some ecological and biological characteristics of Mycoplasma. For example, the genomes of these microorganisms present a high number of genes involved in the degradation of glycans, proteins, and complex oligosaccharides [62,63].
Hence, as we previously proposed, the high number of detected Mycoplasma OTUs from both abalone species may indicate that this genus may be highly host-speci c and that different Mycoplasma may bear some degree of speci city to particular metabolic functions and/or to speci c steps along metabolic routes [32,33]. In addition, Mycoplasma may prevent infection from microbial pathogens in host species. Indeed, as Mycoplasma adhere and colonize epithelial GI tracts [64], they may physically prevent infection through the competitive inhibition for surface binding sites on cells [32]. In addition, the presence of some sialic acid lyase genes in the genomes of certain Mycoplasma species suggest that these bacteria may also prevent infection by breaking down the sialic acid residues of outer membrane proteins used by microbial pathogens to avoid the innate immune responses of hosts [62,65].
Finally, the assessment of microbiota structural and functional changes of this study may provide a new point of view to understand WS expression. Additional experimental evidence is required to complement and test the functional predictions of our in silico analyses. Moreover, the main challenge in predicting the ecological functions of the abalone microbiotas is related to their biological novelty. Indeed, up to 30% of detected bacterial OTUs were not included in our ecological prediction analysis because they presented low genetic similarity with database reference genomes (NASTI values > 2), implying their being new to science. Nevertheless, our main ndings concern the differences between morphologically healthy and WS-stressed abalone and do not rely as much on the accuracy of microbiome prediction but rather on the precision required to differentiate among them. In this regard, our data have been shown to be consistent. In addition, no structural or functional changes were observed between HC abalone with and without morphological signs of WS. The simplest explanation may be that a low number of HC abalone were harvested that presented morphological signs of WS. However, as suggested by the same authors that proposed the AKP, several biological patterns may conceal this principle [20]. Hence, additional explanations may be considered. Our results support the presence of seasonally distinct microbiotas in HC. Seasonal variation in microbiotas have been previously reported in other abalone species [6,28]. It is likely that those seasonal changes may have been driven by diet [28], and we posit that they may confer an adaptive value to abalone. Indeed, it is reasonable to assume that abalone diets vary according to seasonal food availability, among other factors, and that this may drive changes in GI microbiotas that are bene cial to hosts. Nevertheless, the analysis of seasonal microbiota changes is outside the scope of this study and was not directly addressed. However, the integration of this information may help to further the understanding of the occurrence of WS and should be considered in further investigations.

Conclusions
Gastro-intestinal microbiota diversity appears to play a pivotal role in both (i) the health of abalone spp. and (ii) the etiology of WS. Notably, signi cant structural and functional alterations were detected in less complex microbiotas. In those microbiotas, structural changes appeared to be stochastic, as they were followed by the down-regulation of essential metabolic pathways conducted by GI bacteria. Conversely, more complex microbiotas appeared marginally affected by WS in terms of both structural and functional capabilities. The latter may be explained by the prediction that highly complex communities are highly stable, as different species may deferentially respond to perturbations and/or carry out the same ecological function (e.g., functional redundancy). In agreement with these observations, we propose that the species-speci c degree of susceptibility to WS may be related with the degree of complexity of the GI bacterial communities harbored by each abalone species; nevertheless, additional studies are needed to test this hypothesis. Competing interests Table   Table 1   Microbiome structural variation. Principal coordinate analysis (PCoA) based on four distance metrics. The PCoAs were obtained using assigned operational taxonomic units (OTUs) assembled at a 97% similarity cut-off for the gut microbiota of Haliotis fulgens (HF) and Haliotis corrugata (HC).