De novo assembly of Sockeye salmon kidney transcriptomes reveal a limited early response to piscine reovirus with or without infectious hematopoietic necrosis virus superinfection

Piscine orthoreovirus (PRV) is a non-enveloped, double stranded RNA virus which is a member of the family Reoviridae [1]. Its genome consists of 10 nucleic acid segments L1, L2, L3, M1, M2, M3, S1, S2, S3, and S4 that encode for the λ3, λ2/p11, λ1, μ2, μ1, μNS, σ3/p13, σ2/p8, σNS, and σ1 proteins, respectively [1–3]. PRV was first identified by pyrosequencing in 2010 as part of a Norwegian-based study to identify the causative agent of a recently emerged disease of cultured Atlantic salmon (Salmo salar) known as Heart and Skeletal Muscle Inflammation (HSMI) [1]. In that study, high levels of PRV transcripts were found in 28 of 29 diseased specimens, suggesting an association between PRV and HSMI. In Norway, this association has been strengthened by subsequent challenge studies and viral screenings from HSMI outbreaks which have confirmed that PRV is almost always detected during HSMI [4–8] and host antiviral responses are activated during this disease [8–11]. However, wild and farmed Atlantic salmon have also been shown to carry high loads of PRV in Norway without evidence of HSMI [1, 4, 12].

Shortly after its discovery, PRV was similarly identified by PCR in farmed Atlantic salmon on the west coast of North America (Marty and Bidulka, unpublished data). Subsequent studies have confirmed the presence of PRV in farmed Atlantic salmon in British Columbia [3, 13, 14], and also identified PRV in wild salmonids in both BC and US (Washington and Alaska) waters [3, 14, 15]. Interestingly, although infection with PRV is common in farmed Atlantic salmon in BC, disease manifestations impacting fish performance as observed in Norway has never been reported (reviewed in [16]). Using injection and cohabitation challenge trials, we recently examined the infectivity and pathogenicity of PRV from the west coast of North America to Atlantic and Sockeye salmon (Oncorhynchus nerka) and whether infection would result in HSMI [17]. We demonstrated that PRV was capable of infecting, replicating and persisting within both host species over a 41 week period yet histopathological signs of HSMI were not observed. As with infections associated with HSMI, these non-pathogenic infections showed a similar blood cell tropism and comparable infection dynamics, suggesting that PRV is of low virulence to Atlantic and Sockeye salmon in western North America [17].

The pathogenic versus non-pathogenic outcome associated with PRV infection is puzzling, and suggests multifactorial requirements or PRV variant-specific factors may be required to initiate PRV associated disease. As contributing factors from the host, pathogen(s), or environment are unknown with regard to the development of HSMI, it is important to systematically evaluate PRV pathogenesis prior to or in the absence of the possibly obfuscating physiological changes that may occur as a direct result of this inflammatory pathological state.

In the present study we take advantage of the apparent non-pathogenic relationship which occurs between PRV and Sockeye salmon in western North America to identify host-specific transcriptional responses during early systemic infection and increasing PRV load. We use next-generation RNA-sequencing (RNA-seq) technologies in conjunction with contemporary bioinformatics tools to describe and compare head kidney transcriptomes of PRV infected and non-infected Sockeye salmon. Guided by RNA-seq differential expression analyses, we further explore the significance of selected genes in expanded biological, tissue and temporal datasets with regard to PRV-specific host responses.

A second major objective of this study was to determine whether infection with PRV of no or low pathogenicity would affect how hosts respond to challenge with another RNA virus and if such co-infections could lead to PRV induced pathology; or, conversely, afford the host with a protective advantage against a superinfecting viral pathogen. As PRV has been shown to establish long-term infections in its hosts [17] and is common in natural systems along the western cost of Canada [14], it is inevitable that such co-infections will (or do) occur. One pathogen that is sympatric with PRV is infectious hematopoietic necrosis virus (IHNV); which occurs naturally in the waters of western North America and can cause acute disease (IHN) in nearly all salmonid species given appropriate viral, host, and environmental conditions. In British Columbia, IHNV predominately infects Sockeye salmon where much of the viral replication and associated pathology occurs in the kidney during acute infection [18] – an organ which also manifests high loads of PRV [7, 17] and has a significant proportion of blood associated cells to potentially indicate both systemic and organ-specific responses. We thus explore the consequences of a PRV infection on host cellular functions within the head kidney before and during superinfection with IHNV in hopes of better understanding such dynamics as may occur in natural environs.

Sockeye salmon have previously been shown to develop considerable blood and kidney infections following cohabitation with PRV infected Atlantic salmon [17]. Here we confirm that Sockeye develop similar infections following i.p. injection of PRV prepared from infected Atlantic salmon blood. As with previous work, no weight loss, morbidity or pathology could be attributed to the presence of PRV in Sockeye during this study. These results supports a growing body of evidence that PRV is infectious but non-pathogenic to Pacific salmon in western North America [13, 14, 17]. Nevertheless, even in Norway where PRV has been associated with disease in Atlantic salmon, the widespread prevalence of virus in non-diseased fish, sometimes at high viral load [5, 12] indicates that much is still unknown with regard to PRV pathogenicity; particularly concerning mechanisms involved for inducing pathology.

In this study, we used RNA-seq technologies to develop a reference transcriptome for Sockeye salmon head kidney. The Trinity de novo assembly platform used here has been shown to resolve closely paralogous genes and polymorphic transcripts in a broad range of organisms [24], including an Atlantic salmon leukocyte cell line [42], and is currently considered the forerunning tool for vertebrate transcriptome assembly in the absence of a reference genome [43]. To our knowledge this is the first RNA-seq transcriptome dataset for Sockeye salmon, and provides an important genetic reference for this ecologically and economically important Pacific salmon species. The total number of unigenes identified here (326,410) is comparable to the number of unigenes reported from the de novo assembly of a multi-tissue transcriptome of catfish (370,798) [44]; and although a higher proportion of catfish genes was initially identified as protein coding relative to Sockeye (25 % vs 15 %), this is likely due to differences in categorizing isoforms during assembly [24]. Indeed, the number of putative protein coding genes identified herein (47,429) is consistent with those identified in the notochord transcriptome (55,775) [45] and whole-body transcriptomes (41,735-43,654) [46] of Atlantic salmon using genome guided assembly. Additionally, an all-by-all blastn search of the current Sockeye sequences against the latest salmon repeat database provided by the University of Victoria consortium for Genomics Research on all Salmon Project (cGRASP; http://web.uvic.ca/grasp/) revealed at least 33,456 non-redundant unigenes (Additional file 9). This is higher than the quantity of unique protein encoding genes identified within the catfish transcriptome (25,144) [44] and is consistent with the 4R salmonid-specific whole-genome duplication [47, 48] (Additional file 9).

We further identify the Sockeye kidney transcriptome to be highly responsive to the aquatic single-stranded RNA rhabdovirus IHNV. A large degree of variance in global gene expression could be attributed to IHNV infection at 7 dpc and more than 20,000 unigenes showed putative differential expression following exposure to waterborne virus. This included significant induction of antiviral (Mx) and inflammatory (IL-1β) associated genes as assessed by qPCR, which is consistent with previous observations of IHNV infections of rainbow trout (Oncorhynchus mykiss) [49], and is suggestive that both type-I interferon and NF-κB cellular response pathways are initiated during early IHNV kidney infection. Nevertheless, most of the global changes in Sockeye gene expression caused by IHNV have yet to be explored and we are currently expanding our investigations regarding the IHNV responsiveness of this transcriptome. We intend to provide further details regarding cell signalling pathways and examine the importance of differentially regulated non-protein encoding transcripts in a subsequent manuscript.

In stark contrast to the well-developed responses generated against IHNV, PRV appeared to cause almost no kidney-specific gene expression changes in Sockeye at either 2 or 3 weeks post challenge (wpc). The 17 putative PRV-associated DEGs identified in this study account for only 0.03 % of the unigenes used in the analysis and none of the genes identified as differentially expressed at 14 dpc were shared with 21 dpc. The majority of DEGs identified appear to be non-protein encoding and for those which had blastx annotations (Laminin subunit beta-1, Nebulin, Ependymin and 18 s rRNA) only Ependymin has currently been validated by qPCR.

Ependymin is a glycoprotein that was originally discovered in the cerebrospinal fluid of goldfish [50] and is believed to be involved in learning processes [51] and behaviour [52]. More recently however, a number of Epd paralogues have been identified in a variety of other vertebrate organs, particularly those of fish, although their functions are largely unknown [53]. Here, we identify that at least one Epd gene is constitutively expressed in kidney, heart, and skeletal muscle of Sockeye salmon but not in peripheral blood (Fig. 7). We also identified this gene to be the only qPCR-validated differentially expressed protein coding gene responsive to early PRV infection in head kidney tissues. Interestingly, Epd expression has also been shown to be upregulated in the liver of Atlantic salmon following exposure to the bacterial pathogen Aeromonas salmonicida [54], in the liver of medaka (Oryzias latipes) exposed to dioxin [55], and in the distal intestine of Atlantic salmon fed camelina-derived feeds [56]; suggesting the transcriptional regulation of Epn may be involved in immunological, inflammatory, or cellular repair responses. However, as neither immunological, inflammatory, cellular repair, nor indeed any other transcriptional pathway appeared to correspond to Epn expression in this study its biological role in response to PRV appears limited.

In a previous PRV challenge study with Atlantic salmon, we identified only modest and/or questionable increases (less than fourfold) in gene expression of the classic antiviral response protein Mx during early PRV infection following i.p. injection challenge in both blood and kidney during log-linear viral amplification [17]. Here we confirm a lack of an Mx transcriptional response during early PRV infection in Sockeye kidney and further identify a lack of other antiviral transcriptional responses during this period of significant increasing viral load. Together, these studies indicate that PRV is able to evade or block host responses during primary infection, amplification and hematogenous dissemination within salmonids of western North America during non-pathogenic infection.

Interestingly, an absence of host antiviral signalling during early systemic PRV infections may not be exclusive to non-pathogenic conditions. Dahle et al. [9] examined the transcriptional response of Atlantic salmon erythrocytes to infection with PRV using samples obtained from a cohabitation challenge conducted by Finstad et al. [7] which resulted in a first detection of PRV at 4 wpc and histological signs of HSMI at 7 wpc when blood PRV loads were at their maximum. Microarray analysis conducted on peripheral erythrocyte samples collected at 5 and 7 wpc reported upregulation of a large number of viral responsive genes and down-regulation of non-immune related genes in PRV infected groups at both time points; however, qPCR validation of antiviral genes associated with the interferon response (IFN-1α, IRF-1, RIG-1, Mx and PKR) as well as the immuno-regulatory genes Interleukin-10 and Suppressor of Cytokine Signaling-1 did not identify significant responsiveness of these genes at 5 wpc relative to controls. At 7 wpc, when 4/6 fish sampled had histopathological signs of HSMI and PRV loads were highest, significant induction of these genes (except PKR) was observed although variability between individuals appeared high [9].

In our present study, we identified increased Mx expression in kidney tissues only once peak viral loads were reached, which in conjunction with the work of Dahle et al. [9], suggests that an antiviral response against PRV is initiated late during the mature phase of PRV infection and can occur in both the presence and absence of inflammatory associated disease. However, without disease manifestations, we observed that Mx transcription was not maintained and returned to a comparable control level by 9 weeks post challenge (Fig. 7a). In another study, Johansen et al. [10] examined the transcriptional response of Atlantic salmon cardiac tissues during a mature phase of PRV infection over a period of 5 to 11 wpc which used an injection challenge of PRV inoculum made from hearts of Atlantic salmon experiencing HSMI. This work appeared to encompass a period after peak PRV infections were reached, as decreasing quantities of PRV transcripts were identified in heart tissues from the first sampling at 5 wpc through to the end of the experiment at 11 wpc when only one out of five fish tested positive for PRV. Histopathological signs of HSMI were first evident at 6 wpc (1/6 fish), became prevalent at 7 wpc (6/6 fish), then declined at the end of the 11 week trial (2/6 fish). However, even though PRV loads were in decline and many heart tissues appeared to be in the process of clearing PRV infection, expression of viral responsive DEGs was reported to be consistently elevated in challenged heart tissues relative to time-0 controls throughout this period. Although different tissues make direct comparisons difficult, these results provide counterpoint to the brief antiviral responsiveness of Sockeye kidney to PRV in the absence of disease, and suggest that HSMI may perpetuate antiviral signalling as a direct result of pathology in infected tissues.

Currently, the only study that has observed significant host transcriptional responsiveness to PRV during early PRV amplification was recently reported by Johansen et al. [57] who challenged Atlantic salmon parr and smolts with PRV through cohabitation in freshwater and seawater, respectively, using PRV source material from an outbreak of HSMI. The challenge resulted in heart lesions typical of HSMI in both life history stages at 8 weeks post cohabitation and many antiviral responsive genes were shown to be unregulated in kidney and spleen by microarray analysis during early PRV infection (particularly in parr) for which some genes were confirmed to be significantly elevated by qPCR. However, both microarray and qPCR analysis identified significant induction of antiviral genes at 4 wpc even though PRV transcripts were not detected by qPCR until 6 wpc, which makes the relationship of their expression in reference to PRV very difficult to interpret.

Why Sockeye salmon infected with PRV have such a limited transcriptional response in the presence of increasing viral loads is unclear; however, many reoviruses are known to use blood as a means for disseminating from a port of entry to target tissues and leukocytes, erythrocytes and platelets have all been shown to be susceptible to at least one form of reoviral infection in a wide range of host organisms (reviewed by [58]). Nevertheless, the presence of reovirus in blood is not usually associated with anemia or leukopenia, and to our knowledge reoviruses have not been shown to induce antiviral responses in vivo directly from blood cells due to infection. In our current study we did not observe anemia due to the presence of PRV (Additional file 4) which has also been noted previously in other non-pathogenic PRV infections of Atlantic and Pacific salmon [13, 17] as well as during infections associated with HSMI [7, 11]. In contrast, dissemination of many pathogenic orthoreoviruses into target tissues quickly leads to host immune responses and pathology. For example, induction of host immune responses leading to cellular apoptosis is considered the major cause for mammalian reovirus (MRV) encephalitis and myocarditis, which appears to be initiated shortly after viral cell entry and before the core particles become transcriptionally active in these target cells [59, 60]. Thus, the general lack of cellular responses and apoptosis signalling in blood compared to tissues following mammalian reoviral infection is suggestive of a fundamental difference in the way blood cells and target tissue cells interact with this virus. We hypothesize that PRV also interacts differently with blood cells compared to tissue cells, and that PRV can evade antiviral responses during erythrocyte entry and replication in vivo. Following primary replication in erythrocytes, we suspect the dissemination of virus into resident tissue cells causes host antiviral responses for which Epd may be involved. Still, erythrocytes have been shown to mount antiviral responses against PRV ex vivo [11] and further work will be required to identify a mechanism to explain the variability for PRV to induce host cell responses during natural infection.

Differences in cell and/or tissue type interaction with PRV also does not explain the presence or absence of HSMI following PRV infection, as PRV loads have been observed to reach comparable levels with or without pathology in both heart and skeletal muscle [6, 17]. One potential explanation for this disparity may involve genetic variation between PRV isolates. Siah et al. [3] recently conducted a phylogenetic analysis of all segments of the PRV obtained from western North America as well as selected Norwegian and Chilean PRV genomes. With the exception of segment S4, samples of North American PRV clustered separately from the Norwegian and Chilean samples indicating that there are genetic differences across the genome between Pacific, Norwegian and Chilean forms of PRV. In the case of mammalian reoviruses, small changes in the genome can be reflected in dramatic changes in infectivity and virulence [61, 62]. For example, Doyle et al. [61] identified that a single amino change in the ơ3 protein, which affects outer capsid stability, was sufficient to increase replication, host to host spread, and virulence in mice. Based on the analysis of samples of PRV from Norway, Wessel et al. [11] determined that the ơ3 protein had a dsRNA binding function similar to that reported for the ơ3 protein of mammalian orthoreovirus. The dsRNA-binding activity of the MRV ơ3 protein inhibits the induction of type-1 IFN and activation of protein kinase R (PKR) which through its interaction with translation initiation factor (eIF2α) inhibits cellular mRNA translation, thereby, limiting viral protein production. Siah et al. [3] compared the predicted structure of the North American PRV ơ3 protein to predicted ơ3 proteins based on Norwegian/Chilean sequences. These authors identified an amino acid difference which results in a predicted change in secondary structure from an “alpha-helix” in Norwegian/Chilean PRV source to a “sheet structure” in North American PRV sequences. These authors also identified an amino acid substitution in the μ1 protein between North American and Norwegian/Chilean PRV which results in a predicted change in protein structure. Whether these predicted changes in secondary structure results in changes of function for these proteins and or virulence differences between these PRV sources is unknown.

Epigenetic host factors may also provide a possible explanation for regional or stock inconsistency regarding PRV associated disease. In this study, the majority of unigenes that were expressed in response to PRV infection were non-coding. Additional investigation and validation for the putative PRV responsive long non-coding RNAs (lncRNAs) identified in this study may therefore prove valuable, as lncRNAs have been shown to play critical roles in epigenetic transcription silencing [63, 64] and can have post-translational abilities to alter protein function in mammals [65]. A number of lncRNAs have also been speculated to be important during infectious salmon anemia virus (ISAV) infection of Atlantic salmon [66], although the functions for either PRV-responsive host lncRNAs identified in this study or ISAV-responsive lncRNAs identified previously are so far unknown.

Another potential factor for altering virulence during PRV infection could be the involvement of a coinfecting pathogen. A number of significant human diseases are known to be caused or exacerbated as a result of coinfections involving viral pathogens [67–71], and infection by avian reovirus in chickens has been shown to substantially increase anemia caused by chicken anemia virus during dual infection [72]. Alternatively, a coinfecting agent may provide protection against other viral infection; such as in the increased survival afforded to chum salmon reovirus (CSV) or infectious pancreatic necrosis virus (IPNV) infected rainbow trout during lethal IHNV superinfection [73, 74]. In this study, PRV did not protect Sockeye against superinfection IHNV. This is not entirely surprising, as the induction of interferon antiviral responses is usually the mechanism for protection during viral coinfections and is likely responsible for the previous reports of IHNV protection in rainbow trout afforded by CSV or IPNV mentioned above [75]. As PRV had little to no effect on systemic or kidney specific cellular signalling for the first 3 weeks following exposure which included the first week (peak load) of IHNV superinfection, a similar mechanism for protection seems unlikely to be generated from PRV during this period. Nevertheless, we hypothesize timing may be critical with regard to this type of putative protection, and suspect that exposure to IHNV following peak PRV dissemination which causes host antiviral activation (in this study, a brief period starting at 4 wpc) may indeed contribute to a reduced virulence of IHNV if exposure occurred during this time. Further work in identifying such protection may warrant additional investigation; however any putative protection would likely be short lived, as antiviral responses generated against PRV in this study abated by 9 wpc in the absence of PRV associated disease.

Conversely, one of the more surprising observations of this study was the lack of protection afforded by host antiviral responses against IHNV to inhibit PRV transcription and dissemination. PRV transcriptional loads continued to increase at comparable rates in both blood and kidney with or without the considerable transcriptomic changes induced by IHNV. In a previous Norwegian study, chronic IPNV infections also did not significantly alter PRV transcription, nor was vaccination with bacterial and IPNV antigen 8 weeks before PRV exposure able to reduce PRV loads at 7 wpc [76]. Collectively, these data suggest that although PRV does not appear to induce a significant antiviral state during early dissemination, the activation of host antiviral responses during this time appears inconsequential to PRV transcription in either the blood or kidney.

Here we demonstrate that PRV causes minimal changes to the Sockeye salmon head kidney transcriptome during the first 3 weeks of infection following injection. Viral loads increased substantially during this period in both blood and kidney tissues which indicate that PRV has the ability to evade or block host recognition during hematogenous dissemination and early replication within the Sockeye salmon. In this instance, evasion rather than active blocking of host responses is more likely, as PRV had no measured impact on the substantial host transcriptional responses generated against superinfection with IHNV during this period.

Robust host transcriptome responses that included activation of classic antiviral pathways generated during IHNV superinfection did not affect increases in PRV transcriptional load. The mechanisms for why antiviral responses fail to impact PRV transcription, induce erythrocyte/renal cell apoptosis, or lead to pathology in Sockeye salmon are still unclear; however we suspect variations in the way PRV interacts with different cell types may be important. Here we identified a late increase in Mx expression at 4 wpc during a period of peak viral loads in kidney tissue which was preceded by an increased expression of Epd at 3 wpc in PRV infected kidneys but not blood. Although the role of Epd in PRV infection is currently unknown, its previous association with stress/immune functions in other fish species and its lone transcriptional regulation during early PRV infection in Sockeye suggests an important putative role in host PRV responsiveness and warrants further investigation.

The reason for the very limited transcriptional response in Sockeye salmon to PRV within the first 3–4 weeks of infection remains unclear. Although we hypothesize that disparities in the way PRV interacts with different cell types may explain some aspects regarding host responsiveness to PRV, there are many factors involved in PRV pathogenesis which make comparisons of current studies difficult. For example, some transcriptional differences observed in response to PRV can likely be attributed to differences in host species (Sockeye vs. Atlantic salmon). This study also focused on the head kidney which may have a different transcriptional profile when compared to heart tissues. It is also possible that genotypic differences exist between North American PRV and the types of PRV that are associated with strong transcriptional responses and the presentation of the HSMI disease state in Norway. Lastly, the experiment that we are reporting was conducted in a controlled environment with optimal water conditions. Under these conditions, it appears that PRV is of low virulence to Sockeye salmon and that carriers of PRV are not at a disadvantage when it comes to superinfection with IHNV. Whether these results will hold under less optimal more stressful conditions such as those experienced in the field warrant further consideration.