Thermal stress can pose a significant challenge to cod at aquaculture cage sites as summer temperatures may approach the upper critical thermal limit for this species and/or change rapidly during the day (due to thermocline inversions, i.e. when bays "turn over"). Heat-stress is known to cause many physiological changes in cod including the release of stress hormones (e.g. cortisol), changes in the expression of immune relevant transcripts, alterations in oxygen consumption and heart rate, and increased mortality [1, 7, 29]. Therefore, a better understanding of the mechanisms mediating the response to thermal stress should provide insights into how to mitigate and/or avoid the deleterious effect of such environmental challenges. We observed a significant elevation in average plasma cortisol in both control transferred (CT) and heat-shocked (HS) cod at CS; plasma cortisol in HS fish was 2.3-fold higher than in CT fish at the CS time point, and 5.5-fold higher in HS when compared to CT at 3ACS. Further, cortisol levels in HS fish remained elevated during the recovery period. In contrast, average plasma cortisol levels in the CT group had returned to basal levels (i.e. not significantly different from the CT group at BHS) by 3ACS. These results confirm that 3 hours of exposure to 18°C was a severe stressor for these juvenile cod.
Some of the terms more frequently represented amongst the GO annotated ESTs derived from the forward libraries were protein folding, signal transduction, immune response, and response to stress (Fig. 2). The genes associated with these terms may be important in the strategies involved with coping with stress, and variability in their sequences (e.g. exonic, intronic, or regulatory region SNPs) and/or timing and magnitude of mRNA expression (i.e. expression profiles) could reveal markers for increased resistance to thermal and other stressors (research ongoing). This is the first study to use high-throughput genomic techniques to investigate the response to heat-shock in cod, and to provide expression profiles of a wide range of transcripts encoding putative chaperone proteins in tissues of fish that have distinct physiological roles. Many of the transcripts validated at the individual level using QPCR in this study had BLASTx hits that were associated with the GO terms mentioned above and exhibited differences in expression profiles between tissues. These findings indicate that the cellular response to heat-shock in cod is complex, involves several genes, and may be controlled by different cues and/or transcription regulation mechanisms in different tissues, as has been observed in human cells . In all three tissues studied we showed an increase in transcript levels of HSP70-1 (a putative orthologue of the human HSPA1A gene as per the nomenclature proposed by Kampinga et al. ). Previous reports on cod and haddock (Melanogrammus aeglefinus) [29, 32, 33] did not detect an increase in HSC71/HSP70-1 protein expression in the gills and liver of cod or haddock in response to thermal stress. It could be that an elevation in HSC71/HSP70-1 protein was not detected in cod or haddock in the aforementioned studies due to the fact that the antibodies [i.e. polyclonal anti-rainbow trout (Onchorynchus mykiss) HSP70 (Agrisera, Sweden) and monoclonal anti-mouse (Mus musculus) HSP70/HSC71 (Sigma Co., St. Louis)] used were not generated against cod or haddock HSP70-1 and may not have recognized the HSP70-1 protein in these species.
Several of the cDNAs identified in this study were represented by more than one contig in a single library. There are several possible reasons for the presence of more than one contig with the same annotation in a given library. Multiple, same-named contigs may represent: a) different paralogues; b) different alleles at a given locus; or c) non-contiguous segments of a given cDNA (the last-mentioned potential cause of multiple same-named contigs may be the most likely since SSH library construction includes a restriction digest with Rsa I, potentially resulting in more than one contig from a given full mRNA). For example, we report two contigs that were annotated as HSP90α in the head kidney forward library (Additional file 1, Table S1A). Further analysis of these contigs using nucleotide alignments against a full-length sequence obtained from Chinook salmon (Onchorynchus tshawytscha) [GenBank: U89945] suggests that these are likely to represent non-contiguous regions of the same cod cDNA (data not shown). However, the HSP family provides important examples of differential expression (i.e. constitutive and induced expression profiles) between distinct paralogues (e.g. HSC71/HSP70-1) , and further studies addressing this question will be needed to determine the roles of different Atlantic cod chaperone paralogues in thermal tolerance.
Molecular chaperones play important roles in cell physiology in both unstressed and stressed situations. These proteins assist with the folding of nascent peptides and the de-novo folding of denatured proteins, the transport of unfolded proteins across membranes, quality control and conformational changes that affect function . Sudden or chronic increases in temperature are known to induce both mRNA and protein expression of several chaperones, such as those belonging to the HSP family .
Among the clients that these proteins bind to are physiologically relevant proteins such as the glucocorticoid  and aryl hydrocarbon  receptors (clients of HSP90), heat-shock factor 1  (client of HSP70), Immunoglobulin (Ig) heavy chain  (client of GRP78) and the Toll-like receptors  (clients of GRP94). Transcripts encoding putative orthologues of all of these chaperones were identified in our libraries, and all of them were confirmed to be heat-shock responsive mRNAs. HSP90α mRNA expression was up-regulated in the liver more than 150-fold at CS relative to the before heat-shock (BHS) time point (Fig 4A). The up-regulation of HSP90s in response to heat-shock has been demonstrated at both mRNA and protein levels in different species of fish [41, 42]. For example, Cara et al.  detected a ~6000% increase in HSP90 proteins and a ~600% increase in HSP70 protein in fasted +10°C heat-shocked rainbow trout larvae. In our study, there was a higher maximum fold-induction of HSP90α transcripts compared to HSP70-1 transcripts in both liver and muscle following heat-shock. We observed a significant increase in HSP70-1 mRNA expression in the liver of CT fish at 12ACS, which may have been a result of fasting. Cara et al.  observed increased HSP70 protein expression in fasted non heat-shocked rainbow trout larvae. Among the many clients of these chaperones are heat-shock factor 1 (client of HSP70) and the glucocorticoid receptor (client of HSP90). Therefore, the increase in the levels of mRNAs encoding these chaperones may indicate that their products are essential in maintaining signal transduction during stress and are likely to be proteins involved in heat-stress tolerance.
Stress also has an impact on the fish's immune system, and temperature stress has been shown to decrease serum IgM content and increase the susceptibility of sea bass (Dicentrachus labrax) to nodavirus . Nodaviruses belong to the family Nodaviridae, and are the causative agents of viral nervous necrosis (VNN). These viral pathogens also infect cod, and can cause high levels of morbidity and mortality . GRP78 is essential for the appropriate folding and secretion of immunoglobulin light and heavy chains from the endoplasmic reticulum (ER) [39, 44]. IgM heavy chain transcripts in liver were significantly up-regulated by handling stress but not by heat-shock in our study (Fig. 6D), and thus, GRP78 may be important for the proper folding of this immune relevant protein following exposure to only some types of stressor. GRP78 mRNA, which encodes the ER-resident member of the HSP70 family, was significantly up-regulated by heat-shock in all tissues studied. GRP94 (synonym: Gp96), the ER-resident member of the HSP90 family, is the major chaperone for the Toll-like receptors (TLRs) . Yang et al.  demonstrated that Gp96 null mice were also macrophage-TLR null and highly susceptible to Listeria infections. In our study, GRP94 transcripts were significantly up-regulated in all tissues after heat-shock, with the head kidney presenting the highest up-regulation (25.1-fold) at 3ACS (Fig. 4B) relative to GRP94 mRNA levels before heat-shock. TLRs may play an important role in the defence against viral infections and have been shown to be up-regulated by the viral mimic pIC in fugu (Takifugu rubripes) .
Therefore, divergent forms of GRP gene sequences or different expression profiles of the mRNAs encoding these proteins between families and/or populations could play an important role in temperature-related immunosuppression. Given that TLR22 mRNA was significantly down-regulated in the head kidney of heat-shocked cod (Fig. 6A) when compared to its levels before heat-shock, and that this receptor in fish recognizes double-stranded RNA and induces genes of the interferon pathway , it is possible that its down-regulation following thermal stress results in reduced protein levels and is linked to decreased resistance to viruses in stressed fish  (a hypothesis we are currently testing). However, stress does not always correlate negatively with disease resistance. Weber et al.  have shown that a single 3 hour crowding event does not affect rainbow trout survival following a challenge with Yersinia ruckeri (the causative agent of enteric redmouth disease). On the other hand, the work of Fevolden et al.  indicates that the impact of stress on immune competence may be pathogen-specific. These authors have shown that rainbow trout strains selected for high cortisol response had lower survival rate when challenged with A. salmonicida (the causative agent of furunculosis), but higher survival rates when challenged with Vibrio anguillarum (the causative agent of vibriosis), when compared to strains selected for low cortisol response. Fast et al.  showed that Atlantic salmon (Salmo salar) subjected to long-term handling stress (i.e. once a day for 4 weeks) had reduced up-regulation of LPS (lipopolysaccharide)-induced macrophage IL-1β mRNA expression compared to control fish. In this study, chronic handling stress appeared to cause reduced immune competence as evidenced by the decreased survival of isolated macrophages from stressed fish (compared with macrophages from control, non-stressed fish) following incubation with A. salmonicida . Clearly, the relationships between stress and immune responses in fish are complex and require further investigation.
Other cod transcripts encoding molecular chaperone-like proteins were identified in this work including several putative members of the T-complex-containing chaperones (CCT), prolyl-peptidyl-isomerase (PPIase, synonym: cyclophilin A), protein disulfide isomerase (PDI) (the two latter being classified as foldases, enzymes that catalyze reactions which accelerate protein folding and are an important part of the ER chaperone machinery) , and an ER-resident chaperone (HSP47) that is essential for the normal synthesis of procollagen and its stabilization during stress . Collagen is an essential and ubiquitous component of the extracellular matrix and a potential target for denaturation and aggregation. We found that maximum up-regulation of HSP47 mRNA by heat-shock was at 3ACS in liver (20.4-fold), and at 12ACS in both skeletal muscle (16.9-fold) and head kidney (1.6-fold) when compared to its levels before heat-shock (Fig. 4E).
In mammals apoptosis induced by the denaturation and aggregation of proteins is one of the causes of death in heat-shocked cells . Over-expression of the HSP70-1 protein (synonym: HSPA1A) plays an important role in protecting cells from apoptosis, presumably by preventing protein aggregation and inactivating the c-jun N-terminal kinase (JNK) pro-apoptotic pathway . In our study, the mRNA encoding the putative cod orthologue of this particular chaperone was one of the most highly induced transcripts in the liver, with a 100.2-fold up-regulation at 12ACS (Fig. 4C) relative to the BHS time point. Although HSP70s have been shown to be anti-apoptotic in sea bream (Sparus auratus) primary macrophage cultures , previous studies have reported that HSP70-1 protein is not responsive to heat-stress in cod [29, 32]. However as previously mentioned, these studies relied on anti-mouse HSC71/HSP70 or anti-rainbow trout HSP70 protein commercial antibodies, which may not efficiently cross-react with the orthologous HSP70 protein in cod. We have demonstrated that, at least at the transcriptional level, there was a significant up-regulation of a HSP70-1-like transcript in response to heat-shock. Moreover, the maximum-fold up-regulation of HSP70-1 mRNA at 12ACS in all tissues (Fig. 4C), is consistent with its reported role in acquired thermal tolerance during the recovery of mildly heat-shocked mammalian cells . GRP78 is also known to protect cells against apoptosis, since it interacts with the key players in the ER stress signalling system (e.g. ATF6 and PERK) in non-stressed cells, preventing pro-apoptotic signalling . Misfolded proteins in the ER interact with GRP78, which causes the activation of the pro-apoptotic ER stress signalling cascades . Thus, up-regulation of the GRP78 transcripts may lead to elevated levels of this protein that would still be able to silence the pro-apoptotic ER stress signalling pathway. The up-regulation of NUPR1 (synonym: p8) mRNA may also lead to increased expression of this protein, and be an indication of increased levels of anti-apoptotic factors in both liver and head kidney. This protein has been correlated with reduced apoptosis in pancreatic cancer cells . Protein aggregation is known to trigger apoptosis , and therefore, cell viability under thermal stress may depend on the ability to elicit a significant anti-apoptotic response through the expression of transcripts such as those encoding HSP70-1 and NUPR1. However, up-regulation of NUPR1 has also been linked to the acute phase response to pancreatitis in mammals . We found that bikunin transcript, which also encodes an acute phase protein, was down-regulated by handling stress in the liver and by heat-shock in the head kidney. Thus, heat-shock may also affect the inflammatory response. In the spleen of Atlantic cod, bikunin transcript levels were not affected by saline control injection (which includes general handling stress), but were significantly suppressed by viral mimic (pIC) injection at 2 and 6 h post-injection, and significantly induced by the viral mimic at 24 h post-injection (these data relative to saline injected controls at these time points) . Finally, it is worth noting that while NUPR1 was identified as a contig of 2 sequences in the reverse liver library (enriched for genes down-regulated by heat-shock), QPCR showed that this transcript was up-regulated by heat-shock in the liver and head kidney of cod. This was not surprising however, as in our hands, the SSH technique sometimes appears to be less effective at enriching for genes that are down-regulated by a stressor (i.e. in reverse subtractions) than at enriching for genes that are up-regulated by a stressor (i.e. in forward subtractions). As evidence of this, three out of four transcripts identified in a reverse spleen SSH library designed to be enriched for cod transcripts that were down-regulated by exposure to a stressor (viral mimic) could not be confirmed by QPCR as significantly down-regulated by the stressor (i.e. no statistically significant differences were detected) . Therefore, the presence of NUPR1 as a contig of 2 sequences in the reverse liver library in the current study could be an artifact of the SSH technique.
The timing of up-regulation of some transcripts encoding putative chaperone proteins suggests that HSP90α may be a first line of defence against heat-stress, while HSP70-1 may be more important during recovery. Moreover, given that the mRNA expression of most of the studied chaperone genes peaked either at CS or 3ACS, we hypothesize that early time points may be crucial in the process of recovery and repair of damaged proteins. Two other transcripts, CCT 1 and CCT 5, putative members of the TCP1 complex, were significantly up-regulated by thermal stress in the liver at 3ACS (Fig. 5A and 5B). Of the 8 known mammalian members of this complex, we identified cDNAs for 6 putative orthologues in cod: CCT 1, 2, 3, 5, 6 and 8 [Tables 2 and 3, and Additional files 1 (Table S1) and 2 (Table S2)]. These chaperonins are known to form heterologous polymers that assist in the folding of actins and tubulins , important components of the cytoskeleton. Structural proteins seem to be among the most heat-labile proteins, and their misfolding and/or denaturation contributes greatly to protein aggregation .
Although we only saw a small, albeit significant, down-regulation (3.0 fold; CS-Fig. 5C) of TCTP mRNA expression in the head kidney in the CT group relative to the BHS time point, this transcript may still represent an important component of the molecular mechanism involved in thermal resistance. The product of the TCTP gene, a ubiquitously expressed protein in most mammalian cells, is known to bind to calcium and tubulin and to be responsive to stressors such as starvation and heat-stress . In addition, Bonnet et al.  have shown that in yeast cells exposed to heat-shock there is a down-regulation of TCTP mRNA, and in rat (Rattus norvegicus) C6.9 glioma cells TCTP mRNA is up-regulated in response to induced programmed cell-death . Down-regulation of TCTP in response to heat-shock in yeast may be one of the mechanisms that prevent heat-induced apoptosis, and it is possible that this down-regulation, which was detected in our experiments (e.g. 1.5 fold down-regulation in HS head kidney at CS), was not significant due to high variance between biological replicates (Fig. 5C). Down-regulation of TCTP may also play a role in preventing apoptosis triggered by other stressors (i.e. handling) since we found it to be significantly down-regulated by 3.0 fold at CS in the head kidney.
We saw little change in the mRNA expression of genes with carbohydrate metabolism related functional annotations (enolase, aldolase, PFK) (Fig. 5D, Table 2) with heat-shock at 18°C. However, this finding does not preclude the possibility that carbohydrate metabolism is increased when Atlantic cod are acutely exposed to elevated temperatures. This is because glycolysis had a relatively high prevalence (3.13%) amongst the biological process GO terms in the muscle forward library (Fig. 2). PFK and glycogen phosphorylase (identified in the head kidney forward library) are the rate limiting enzymes of glycolysis, and allosteric regulation of these enzymes is likely to be the main mechanism through which carbohydrate metabolism is re-organized during acute stress. Finally, the results of Perez-Casanova et al.  suggest that carbohydrate metabolism (based on measurements of plasma glucose) is not up-regulated significantly in cod until temperature reaches at least 20°C during acute thermal stress.
Interestingly, enolase transcript was significantly down-regulated (by 1.9-fold) in the head kidney of control transferred (CT) fish at the CS time point relative to the CT BHS time point, and GRP78 mRNA was significantly down-regulated in the liver in the CT group at all time points relative to the BHS time point. These results indicate that, even though there is a conserved general stress response, some responses at the transcriptome level are stressor specific (i.e. responsive to either heat-shock or handling stress).