Deciphering effects of symbionts and environmental factors on gene expression in field populations of symbiotic organisms is an important goal in molecular ecology. In the case of the vent mussel B. azoricus, genes regulated and/or involved in its symbiotic relationship with two symbionts were unknown. To address this issue, we used a combination of subtractive libraries and a cDNA microarray approach to characterize putative differentially expressed genes in the hydrothermal vent mussel B. azoricus inhabiting two physically and chemically contrasting vent fields. Then, we determined if the expression of these differentially expressed genes was, in part, influenced by symbiont content and/or symbiont metabolism, and/or environmental factors. Because we used mussels collected in their natural environment, we did not expect to identify genes involved in the establishment of symbiosis, but rather genes involved in an established symbiosis that were differentially regulated with respect to symbiont type, quantity and activity, and environmental factors.
Vent field environmental conditions impact activity and relative abundance of both symbionts as well as host gene expression
The quantification of symbiotic bacteria in mussel gills showed a significant difference of both SOX and MOX content across the two vent fields. The relative abundance of the two symbionts is influenced by the hydrothermal fluid characteristics, especially methane and sulfide concentration. This correlation explains, for example, the higher abundance of MOX in mussels collected at Rb vent field where high concentrations of methane were detected [27, 28], compared to individuals from the MG vent field [14, 15, 23]. The average quantity of each symbiont was significantly different between the two vent fields, but we observed a large inter-individual variation in symbiont abundance within each field. This symbiotic plasticity allows mussels from one vent field to harbor the same amount and/or proportion of symbionts as mussels from another field. Mussels from each vent field were collected in a restricted area of the mussel bed, but given the highly chaotic mixing conditions encountered at hydrothermal vents, we have to consider variations in both sulfide and methane availability, even at the scale of the sampling patch. These variations in gas availability in vent fields impact the distribution and diet of vent fauna at a micro-spatial scale  and may explain these differences in symbiont content.
The symbiotic vent mussel B. azoricus inhabits a variable environment due to the highly chaotic mixing of hydrothermal fluid with seawater within the site [29, 30], and because of the bathymetric position of these hydrothermal fields . Thus, a strong effect for environmental factors was expected due to the very different characteristics of the vent fluid in the two populations sampled (especially gas and heavy metals concentrations, temperature and pressure), and therefore specific signatures of the mRNA expression in mussels were expected for the different vent fields. Our microarray data showed that 120 genes (3.5% of all genes) clearly distinguish both sites, indicating a relatively moderate effect of source vent on transcriptome regulation. However, cluster analysis of all individuals showed a clear separation between samples from MG and Rb, indicating that the physical characteristics of the two vent fields were strong enough to influence transcriptome expression in a way that distinguishes populations.
Among the up-regulated genes in mussels collected at the Rainbow vent field, we identified a 60S acidic ribosomal protein and a selenoprotein. The 60S acidic ribosomal protein, known as a P protein, is mainly associated with the protein elongation step of translation, but potential roles in transcription, DNA repair , in response to pesticide exposure , and in intracellular iron sequestration  have also been described. The selenoprotein has multiple functions such as antioxidant defense, selenium transport and heavy metal chelation . Up-regulation of these genes in mussels from the Rb vent field is consistent with the high metal concentrations at this site, the highest observed in the MAR hydrothermal area [27, 36].
Among the genes significantly up-regulated in mussels collected at the MG vent field, we identified some metabolic genes such as arginine kinase and carbonic anhydrase (CA). Arginine kinase is known to play a key role in cellular energy metabolism in invertebrates  and its regulation in response to temperature has been previously described in B. azoricus . Carbonic anhydrase is known to be involved in the transfer of CO2 from the environment to the cell in many symbiotic animals. This enzyme catalyzes the reversible hydration of CO2, and was found to be regulated at the transcriptome level according to the state of symbiosis, in both plants and animals , but also in B. azoricus in response to temperature variations .
Regarding the limited number of known genes and the lack of experimental studies performed on B. azoricus in response to various stressors, as well as the lack of information about the micro-environmental characteristics around mussels (especially for concentrations of H2S and CH4), it remains difficult to link these significantly regulated genes to environmental parameters. Moreover, our sampling strategy of collecting mussels directly from their environment did not allow us to assess a cause and effect relationship. For example, we cannot establish whether a change in environmental factors drives symbiont metabolism, and in turn host physiology; or conversely, if a change in environment directly effects host physiology leading to a regulation of symbiont population. This point will be discussed more extensively below.
Do symbiotic bacteria drive gene expression in Bathymodiolus azoricus?
Double symbiosis enables B. azoricus to colonize sulfide and/or methane rich environments, in which the primary production of the symbionts ensures a part of the host's nutrition. Additionally, these mussels are able to filter feed and can survive senescent vent conditions. Mixotrophy is a major advantage in highly variable environments. In the particular case of B. azoricus, the mussel has to host two different symbionts that are present in varying abundance in different individuals. The identification of genes showing a similar regulation according to SOX and/or MOX content in two contrasted populations should help to distinguish genes that are mainly regulated by symbiont content from those that are regulated by both symbiont content and environmental parameters. The microarray analysis showed a relatively low number of genes significantly regulated by either SOX (8%) or MOX (14%) content in the MG population compared to the Rb population which had 16% regulated by SOX content and 20% by MOX content. These results suggest that symbiont content is less influential on the transcriptome in mussels from MG. However, because of the lack of studies on this particular dual-symbiont model, we have no information about a potential competition between SOX and MOX symbionts, and in turn, how mussels control each kind of symbiont. One could hypothesize that this difference is partly due to the bathymetric position of the two vent fields. We noticed that very few genes seem to be commonly regulated by both SOX and MOX content at MG, only 1% of all genes, suggesting that each symbiont may affect different pathways in mussels inhabiting this vent field, compared to Rb where 5.6% of all genes are commonly influenced by SOX and MOX content. The MG vent field is located at a depth of 800 m versus 2300 m for the Rainbow vent field. Thus, the mussels at MG benefit from a higher particle flux  which lessens the contribution to carbon nutrition needed from the symbiont, and possibly also the impact of symbionts on the host's transcriptome. In contrast, the mussels at the deeper Rainbow vent field experience lower particle flux and rely more on symbionts to meet their carbon needs. The pattern of gene expression obtained in this study could reflect the relative carbon contribution of symbionts compared to the availability of particles to host nutrition.
While the number of sequences available for bivalves has increased dramatically during the past few years [40–43], very few genes are either fully annotated and/or functionally characterised, often leading to a mean proportion of unknown sequences higher than 50%. In a bivalve such as B. azoricus, it is particularly difficult to find a relationship between the regulation of gene expression and a symbiotic state, even if the gene was described as involved in symbiosis-related functions in other organisms. However, we identified several genes previously described in host/symbiont relationships in other marine models, such as the sea anemone Anemonia viridis , the squid Euprymna scolopes , and the hydrothermal tubeworms Ridgeia piscesae and Riftia pachyptila [4, 6], and one could hypothesize that their roles are potentially quite similar in hydrothermal mussels. It is, for example, well established that participation of sugar residues and lectins is a major process in host-microorganism recognition . In this study, we identified five lectins belonging to different families and showing significant regulation according to symbiont quantity or activity. Lectins are known to bind carbohydrate structures on foreign cells [46, 47]. Previous work conducted on corals  showed that specific lectins may bind to both pathogens and algal symbionts, suggesting that lectins may have been co-opted from an ancient innate immune system into a role of selecting and maintaining the photosynthetic endosymbionts in host tissues. In our analysis, we also identified a lysozyme and observed that this gene is more expressed in mussels with a high SOX content (MG vent field). The ancestral function of this enzyme is in defense against pathogens by degrading bacterial wall [49, 50], but its implication in digestion in ruminants  and mollusks  has also been demonstrated. Lysozyme is strongly involved in the control and maintenance of the bacterial flora in the aphid bacteriocytes  and in the digestion of chemoautotrophic bacteria by their deep-sea bivalve hosts . This change of lysozyme function from anti-biotic defense to digestion may have arisen through convergent evolution via positive selection . An example of such change has been recently identified in the Eastern oyster, Crassostrea virginica in the i-type lysozymes family . The regulation of lysozyme and lectins in B. azoricus agrees with previous observations, an indication that these two gene families are potentially good candidates for proteins that might be involved in the control and maintenance of symbionts. We also noticed the regulation of several genes directly or indirectly implicated in immune defense and inflammatory reaction. Among them are some receptors to melatonin , acetylcholine and laminin , synthaxin , kininogen, cystatin  and prostaglandin E2 synthetase and receptor , and all are significantly regulated by symbiont abundance and/or activity in B. azoricus. However, due to the multi-functionality of these proteins coupled with a lack of knowledge about their roles in hydrothermal mussels, we cannot be conclusive about their respective function(s) in the mussel/symbiont relationship. Complementary analyses of function and biochemical properties should help to determine to what extent these proteins are involved in the breakdown of symbionts and the elimination of microbial intruders in hydrothermal vent mussels.
The influence of bacteria on the cytoskeleton of host cells has been extensively studied in both host-pathogen interactions [61, 62], and host-symbiont relationships [44, 63, 64]. These studies showed that various pathogens and symbionts increase their intimacy with the host tissues by altering the host cytoskeleton. For example, several microfilament and microtubule proteins are strongly regulated at both RNA and protein stages during the establishment of the symbiotic association between the squid E. scolopes and Vibrio fischeri [44, 63]. In B. azoricus, some genes encoding cytoskeleton proteins are differentially regulated according to symbiont content, suggesting a potential effect of symbionts on host cell structure. Among these genes, five (tubulin, dynein, annexin, beta-thymosin and actin-related protein 2/3) present an interesting pattern of up-regulation in mussels hosting a high symbiont content, especially those with a high MOX level.
Disentangling environment and symbiont effects on host gene expression: what is the order of event?
We established that the expression of several genes is correlated with either symbionts (quantity and/or activity) or environmental factors, but we were not able to determine which factor is directly responsible for transcriptome variations in mussels. The analysis of the transcriptome of a symbiotic organism often generates confusion when considering the combined effect of both symbionts and interrelated environmental factors. While they used a robust experimental design, DeSalvo et al.  were not able to determine if a thermal challenge changed coral (Montastraea faveolata) physiology which, in turn, induced a change in symbiont type dominance, or if a thermal stress directly changed symbiont type dominance and, in turn, the physiology of the host. In our case, we could assume that environmental factors directly influenced symbiont abundance in mussels [present study; . But we also observed a large inter-individual variation in symbiont abundance (measured at both sites), indicating that environmental factors alone do not drive symbiont quantity, but probably in association with host and/or symbiont need and/or physiological state.
The expression pattern of several genes was also ambiguous. For example, ferritins are significantly regulated by symbiont content in both populations of vent mussels, and their regulation in host-pathogen as well as in host-symbiont interactions has been previously demonstrated [66, 67]. But ferritins are also known to play a pivotal role in iron homeostasis and the oxidative stress response. In our study, ferritins are more expressed in mussels harboring high SOX and MOX content collected at Rb vent field, compared to MG mussels in which ferritins are more expressed in low MOX content mussels. In this case, it was not possible to link the ferritin expression pattern to either symbiont content or the high level of iron measured at Rb vent field. A similar analysis applies to carbonic anhydrase (CA). We showed that CA is regulated by environmental factors (see discussion above). But, it has previously been demonstrated that CA plays a major role in transport and supply of CO2 to autotrophic symbionts housed in host tissues, such as in the two hydrothermal worms R. pachyptila and R. piscesae [6, 68–70]. In these two species, the metabolism of the thiotrophic symbionts (sulphide oxidation) entails a fast and high production of protons, with which the worms have to cope, partly by induction of CA at both the level of enzymatic activity and mRNA expression. In our study, we observed that CA is highly expressed in the mussels collected at MG in which a higher SOX content was observed compared to Rb.