In this study, we have identified several genomic regions associated with the variability of response to a hyperosmotic challenge in rainbow trout. To our knowledge, this is the first description of QTL affecting plasma indicators of the hydromineral balance after an osmotic challenge in a euryhaline fish. Our results strengthen and extend previous studies which suggested that genetic factors are likely to influence the ability to adapt to high salinity in teleost fish. These studies were based on the comparison of different species and their hybrids (as in sea bass  or in tilapia ) or on the comparison of different strains or populations in teleost fish on striped bass , salmon [13, 14, 16, 17, 46] and artic charr (Salvelinus alpinus) . The discovery of QTL associated with osmoregulatory ability in rainbow trout paves the way for further understanding of the genetic basis of the regulatory mechanisms and physiological pathways involved in the control of hydromineral balance in this species.
We identified several QTL contributing to the variation of plasma ion concentrations after exposure to the salinity challenge. The plasma levels of chloride and sodium, the two main ions responsible for plasma osmotic pressure, are key indicators of the regulation of hydromineral balance and consequently were chosen to monitor the adaptation to 30 ppt salinity in the QTL progenies.
Two significant QTL (on RT10 and RT19) and four suggestive QTL (on RT4, RT5, RT9 and RT23) were found to control the variation of both ions during the two successive challenges. Moreover, in most cases, the QTL affected both concentrations in the same direction. There are no data about the value of the genetic correlation between the two ion concentrations in the plasma and the genetic structure of our sample (5 full-sib families) did not allow a reliable estimate of it. Nonetheless, this result is consistent with our observation that the individual plasma values of Na+ and Cl- are significantly correlated within each challenge. It is also consistent with our present understanding of the hypoosmoregulatory mechanisms in seawater which indicate that the extrusion of Na+ and Cl- is accomplished by distinct but coupled mechanisms. The current model for NaCl extrusion at the gill epithelium proposes a basolateral co-transport of Na+ and Cl- down the electrochemical gradient produced by NKCC co-transporter. This co-transport is coupled with an apical extrusion of Cl- via a low conductance anion channel and a paracellular extrusion of Na+ through Na+-K+-ATPase transporter and diffusion towards the external medium down its electrochemical gradient (see reviews Evans et al. ; Hwang and Lee ). The morpho-functional modification of the gill epithelium characterized by the development of leaky junctions and also accessory cells are fundamental for efficient extrusion of Cl- and Na+ in the gill . Whether these common QTL correspond to genes encoding major regulators of these cellular changes would be an interesting hypothesis to test.
We also found QTL affecting the concentration of one only of the two ions. For sodium, the most significant QTL affected ionic concentration at the second challenge (P < 0.05), and was located on RT14. For chloride, one QTL was consistently found on RT26, and an additional QTL may be located on RT25 (P < 0.05, in unitrait analysis only). Taken together, the levels of significance of those QTL remain moderate and the power of our design may have impeded the detection of some effect on the alternative ion. Nevertheless, this result may also indicate that the regulation of plasma sodium and chloride relies in part on ion-specific mechanisms. Based on the present knowledge of the cellular mechanisms responsible for Na+ and Cl- extrusion in hyperosmotic environment (see review Marshall and Bellamy ), one may hypothesise an association between QTL and one of the ion transporters characterized at the level of the gill epithelium. Additional information may have come from the genes in which the SNP used for the genome scan have been detected although examination of the known biological function of these genes (Gene Ontology) did not suggest any gene relevant for its involvement in ion or water exchange at the level of transporting epithelia.
The QTL families were F2 progenies from two divergent grand-parental lines (HR and LR) previously selected for plasma cortisol response to an acute confinement stress [33, 34]. It is well established that cortisol plays an important role in the success of the adaptive osmoregulation process after transfer to a saline environment . Cortisol represents a major endocrine actor in the regulation of ionic homeostasis, particularly after environmental salinity change, by regulating some seawater specific ion transporters such as Na+-K+-ATPase isoforms, NKCC and CFTR [49–56]. By its actions on gill chloride cells number and activity, cortisol acts in synergy with IGF-1 and GH to increase overall salinity tolerance (see review by McCormick ). Yet, altogether, our QTL results do not support evidence of a unidirectional association between the HR vs LR origin of QTL alleles and their effect on ion concentrations in the plasma. They rather suggest that the role of grand-parental alleles depends on the loci, a result that highlights the functional complexity of underlying mechanisms and is in agreement with the present knowledge that different stressors (as confinement or salinity stressors) have distinct effects on the physiological mechanisms regulated by cortisol . However, by introducing a valuable source of additional variability in comparison to a standard within-population QTL design, the HR and LR fish provide a valuable tool with which to obtain a better understanding of the complex hormonal control of hypo-osmoregulatory mechanisms. An increasing amount of attention is paid to welfare and robustness traits in aquaculture, and many studies on the genetic control of cortisol response to stressors have been carried out, ultimately concerned with reducing stress in farmed fish [28, 57, 59–62]. Thus, a better understanding of the relationship between cortisol responsiveness to acute stress and osmoregulation capacities is a key requirement prior to the adoption of any selection programme designed to manage the cortisol response to stress in aquacultured fish.
Preliminary observations had suggested that LR individuals had relatively larger gill arches than HR individuals. The QTL design was thus appropriate to search for QTL for gill relative development. Four significant and two suggestive QTL were detected for gill weight relative to body weight. This result is in agreement with other studies that have identified the existence of numerous QTL for body traits in fish [62–64]. Four of those QTL are also involved in the regulation of plasma Na+ and Cl- concentration after the osmotic challenge, which raises the question of the possible role of gill size and morphology on osmoregulatory capacities. To our knowledge there are no published data indicating any genetic correlation between gill size and tolerance to high salinity, and the genetic structure of the QTL data set (5 full sib families) did not allow a reliable estimate to be calculated. However, we observed that the QTL alleles associated with a larger gill also affected ionic concentrations in the plasma in different directions (increase as decrease). The total weight of gill arches is not directly representative of the relative development of the gill epithelium, which is the active site for ionic and water exchange and it is therefore unlikely that there is a direct 'mechanical' effect of gill size, through an increase of the branchial ion-exchange surface. This is also in agreement with the lack of correlation between individual values for gill index and any of the plasma ion concentrations in the different challenges. The most likely hypothesis is that QTL harbour multiple syntenic genes regulating the morphological trait (gill size) and physiological traits (ionic concentrations).
Up to now, very few studies have investigated the genetic architecture of osmoregulation capacity or associated traits in salmonids. To our knowledge, only Nichols et al.  have searched for QTL associated with migratory capacities (smoltification) in trout O. mykiss. Using a cross between nonmigratory rainbow trout and migratory steelhead trout line, they discovered 14 genomic regions (on 14 different linkage groups) associated with smoltification-related morphological traits (growth, condition factor, body coloration, morphology). The genome scan in this study was performed using AFLP markers, but at least one microsatellite per linkage group was genotyped and can be used to establish synteny with other rainbow trout maps. Interestingly, several QTL discovered in the two studies co-localize on the same rainbow trout linkage groups. This is the case for the QTL on linkage groups RT9, RT14 and RT23 for which we could establish the synteny with respectively OC9, OC14 and OC23 in Nichols et al. Ambiguity remains for the QTL located respectively on OC3 in Nichols et al. and RT25 in this study. In Nichols et al., OC3 is identified by One8ASC and Ogo2. These two markers belong to a region homologous to RT25, as attested by several shared duplicated loci, including Ogo2 [19, 21]. This suggests that the two aforementioned QTLs could be under control of two homeologous genes. Furthermore, since the duplicated or non-duplicated status of One8ASC (as any marker) is difficult to prove and can vary among populations, the possibility that OC3 and RT25 correspond to the same linkage group in these two particular studies cannot be ruled out. Two additional QTL suggestive in the present study (RT21 and RT31) also locate on linkage groups identified in Nichols et al. For RT21, the synteny can be established from Omy325DU (on OC21 in the study by Nichols et al.) and Omy325UoG (on RT21 in our linkage map ), which correspond in fact to the same marker (R. Danzmann, personal communication).
On the other hand, it is worth emphasising that in our study, we do not find any salinity acclimation QTL on RT20, the linkage group the most strongly associated with smoltification traits (mostly growth and morphological traits) in the study of Nichols et al. As no smoltification process occurred in the fish during our survey, this observation is not surprising. It emphasizes that the smoltification process and the adaptation to salinity after a direct transfer to seawater are two different processes.
Overall, there are at least three linkage groups (and possibly 6) harbouring QTL in both studies. It is very difficult to compare the location of the QTL discovered in the two studies on the linkage groups, as genome scans were performed with a low marker density and different types of markers were used (AFLP vs microsatellites). Moreover, the traits recorded in the two studies are quite different (survey of growth and changes in body colour and morphology over the time course of smoltification vs plasma indicators of the status of the hydromineral balance after an osmotic challenge in a nonmigratory population) and many linkage groups are found to harbour QTL for the traits examined in the two studies (10 and 14 in the present study and the one by Nichols et al. respectively). Although it is thus possible that common linkage groups are observed by chance, these observations may also substantiate the role of those genomic regions in the osmoregulatory process in trout and corroborates the hypothesis that some of the key processes for the control of hydromineral balance during adaptation to salinity (as highlighted by our QTL study) also occur in individuals undergoing smoltification.
During this experiment, a large number of individual fish were submitted twice to the same salinity stressor. The data set provides additional information on basic physiological responses of trout during the early (first 24 hours) adaptation to seawater. The mean plasma ion concentrations were in the same range as those usually recorded in similar tests with fish of similar sizes. However, when the fish were exposed twice to the salinity challenge (24 h at 30 ppt salinity) separated by a 2 week recovery period, plasma Na+ and Cl- levels did not behave similarly: we observed an increase in the mean plasma Na+ levels between the first and second challenges whereas no major changes were observed in plasma Cl- levels (Table 1). It is possible that this difference is linked to the fact that, in rainbow trout, plasma Na+ level changes after transfer to SW were much more sensitive than plasma Cl- level changes (; Brard et al., unpublished data) and this requires further investigation. These results also suggest that, at least for Na+, an acute salinity stressor (24 h exposure to 30 ppt salinity) may lead to permanent modification of the functionality of the Na+ osmoregulatory mechanisms resulting in a change in the capacity to excrete Na+ when the fish is re-exposed to high salinity. To our knowledge, such a mechanism has never been reported in the literature.