IPNV isolates have been reported to vary considerably in their virulence and pathogenicity for Atlantic salmon [18, 19]. Previous studies have indicated that a combination of IPNV virulence and host pathogen interactions determines the outcome of IPNV infections. Occurrence of multiple strains indicates rapid changes of the virus but little is known about the rate and character of mutations. It also remains undefined if susceptibility of salmon to IPNV could be associated with low or excessive immune responses. In this study two Norwegian field isolates of IPNV with marked difference in mortality were applied for experimental infection of Atlantic salmon. For the first time the sequence changes and immune responses to low and high virulence strain were compared within a single challenge test.
Previous studies have identified the outer capsid protein VP2 as the main determinant of IPNV virulence as it comprises all the neutralizing epitopes and cell attachment sites that determine host or cell specificity . The amino acid signature associated with virulence of different strains is also identified within the VP2 region [8, 19, 39]. The sequences of the two isolates in this study were determined before and after the challenge. Initially NFH-Ar had Thr217-Ala221-Thr247 while NFH-El had Thr217-Thr221-Thr247, implying high and low virulence of NFH-Ar and NFH-El respectively (Figure 3A). Thus the results from this study are in accordance with previously reports of the virulence motifs within VP2 . Rapid changes were seen during passage in cell culture and also during infection in the fish. After the challenge both isolates acquired Pro217-Ala221-Ala247 motif in VP2 associated with the moderate to low virulence . Sequence analyses revealed a remarkably high rate of non-synonymous substitutions in the HVR containing the virulence motif. Commonly, high Ka/Ks ratios point to the divergent selection meaning that sequence changes increase the rate of reproduction. However in this study we observed accumulation of mutations that correlated with a reduction in virus proliferation, thus being favorable for the fish. End of mortality in NFH-Ar group coincided with the loss of the high virulence motif. To explain this finding, virus modification as the host's defense mechanisms can be hypothesized. Virus editing is a rapidly expanding research area. At present, the best studied actors are adenosine deaminases (ADARs) that target regions of dsRNA, converting adenosine (A) to inosine (I) resulting in an A to guanosine (G) change after second strand synthesis . ADARs target mRNAs, transposable elements and RNA viruses' genomes. In mammals, several families of viruses show A to G mutations thought to be caused by ADARs . The induction of ADAR in the NFH-Ar infected fish observed in the microarray in this study might indicate a possible role for ADAR in editing IPNV. Most of the changes observed in our study, disregarding whether synonymous or non-synonymous, were from A to G or from T to C (or vice versa) associated with deamination (results not shown). However, whether the mutations in IPNV detected in this study are caused by salmon ADARs is an interesting question that needs to be addressed in future studies.
The two virus isolates used in this study showed greater discrepancies in the VP5 region than previously described isolates. The change of VP5 into a shorter protein in the NFH-Ar infected fish might imply a more benign virus since the longer 15.2 kDa VP5 protein was shown to have a potential antagonistic effect on the IFN response by inhibiting IFN-induced expression from the Mx promoter  and might thereby benefit the viral replication. However, the domains responsible for this function have not been mapped and can still be present in the shorter 12.1 kDa version. An early stop-codon located in NFH-El leaves this isolate with a severely truncated form of VP5, only 28 amino acids long, and to our knowledge such a mutation has not been reported in other surveys of IPNV field isolates. It has been suggested that VP5 has an anti-apoptotic function, which is probably not essential for the virulence or persistence of the virus [42, 43]. Functional significance of the observed changes in VP5 remained unclear. The amino acid sequences of VP1 ORF were identical between the two isolates throughout the study and VP4 were subject to very few mutations.
The mortality rates associated with the structural differences described above may be linked to the ability of virus to invade and replicate within the host cells and/or the scale and character of immune responses. Association between mortality and the virus titers was obvious (Figure 1). NFH-El characterized with low replication was avirulent while NFH-Ar infection was fatal at high rate of proliferation at 13 d p.i. To assess the immune responses, we used expression profiling with microarray and qPCR analyses of genes with well-established roles (IFNs, Mx and PRRs) and both approaches produced similar results. The expression levels of VRGs were apparently mirroring the viral titers and expression levels of the IPNV VP2. At 13 d p.i. VRG were induced in salmon infected with NFH-Ar but not with NFH-El consistently with the difference of virus titers. Up-regulation of VRG in NFH-E1 infected fish at 29 d p.i. was in line with the slight increase of virus titer and did not affect mortality. It is likely that the slower replication and concomitant slower spread of the NFH-El strain may allow time for a systemic induction of the host anti-viral system, including adaptive responses.
The IFN system is believed to have a crucial role in the first line of defense against virus infections, and in vitro studies have demonstrated that IPNV replication in cell cultures is efficiently inhibited by salmon IFN-a1 [21, 22]. Additionally, injection of synthetic IFN-inducers like CpG and poly I:C induce protection against IPNV in Atlantic salmon . In this study IFN-a1 was induced by both viral strains and major up-regulation was seen in the IFN-a1 dependent gene Mx. IFN-c, was not induced and even slightly down-regulated at 29 d p.i. for NFH-Ar. IFN-c is suggested to have a separate regulation from a and b and can be produced by a different cell population than IFN-a1 . IFN-b was not detected in this study or showed consistently low expression levels (data not shown). Despite the suggested role of type I IFN in restraining virus production, results in our lab has demonstrated that IFN-a1 does not completely inhibit IPNV growth but causes a delay in viral protein synthesis . Furthermore, our data suggest that IPNV-encoded proteins may be involved in weakening of IFN signaling . As a result high levels of viral proteins may impair the activity of IFN-induced genes, thus the higher replication rate of NFH-Ar compared to NFH-El may cause a more potent IFN-antagonizing effect of the NFH-Ar strain. However when interpreting the results from live pathogen challenges, it is important to keep in mind the complexity of such studies, where it is not straight forward whether an observed response is a strategy employed by the virus for its own benefit or a response by the host to control the virus.
Although innate immunity by itself represents a powerful system to combat viral invaders, many infections can only be cleared in combination with adaptive immunity. In this regard type I IFNs are known to promote the adaptive arm including both T cell mediated cellular responses and antibody production [45, 46]. It is likely that the reduction of virus titers by 29 d p.i. could be associated with the onset of adaptive immune responses. Unlike the genes implicated in the innate immunity, sIgM showed no expression changes at 13 d p.i. and was slightly induced at 29 d p.i.
Results of this study added knowledge to the understanding of the immune responses after IPNV infections. Expression of a panel of PRRs was assessed including several recently identified genes. TLR8, 9 RIG-I and MDA5 showed up-regulation and followed the same trend as other immune genes. Earlier we observed a modest increase of TLR8 and 9 expression during stimulation and infection [27, 28]. TLR22 is reported to recognize dsRNA in pufferfish and when over-expressed it induces type I IFN expression upon IPNV infection, which suggests a possible role for TLR22 in protection against IPNV . However, TLR22 was not induced by IPNV in this study. Unexpectedly, PKR, an IFN-inducible gene, was down-regulated at all time points except 13 d p.i. (Figure 4A). This was in contrast to Mx, another IFN-inducible gene and IFN-a1, which were up-regulated. Functional studies of salmon PKR have to our knowledge not been reported, however flounder PKR was up-regulated both in vitro and in vivo by a negative single stranded RNA virus (SMRV), which also induced Mx expression in vitro . PRRs, PKR and MyD88 were down-regulated at 6 d p.i. in both study groups. This could be explained by migration of leukocytes expressing the PRRs from the head-kidney into the bloodstream at early time-points of the infection.
IFN-γ was together with Mx the most highly induced immune gene in this study, and high levels of IFN-γ has been reported in other IPNV challenge experiments . IFN-γ is regarded as a typical Th1 cytokine which bridges the innate and adaptive immune responses. Fish IFN-γ share several functional properties with mammalian IFN-γ including macrophage activation [48–50] and rainbow trout IFN-γ is shown to signal through STAT1 [50, 51]. Recently, we have observed antiviral activity against IPNV by IFN-γ, although the effect was not as pronounced as described for IFN-a1 . Like in mammals, fish IFN-γ plays diverse roles in different facets of the immune system, and the increased levels of IFN-γ upon IPNV challenge detected here suggest anti-IPNV activity.