In May of 2019, there was an IPN field outbreak in a Western region of Norway. The affected fish were Atlantic salmon of the SalmoBreed strain, delivered as unvaccinated, IPN QTL, all favorable homozygous or heterozygous for the QTL locus (Houston et al., 2008; Moen et al., 2009). Fish were at the post-smolt stage with an approximate average weight of about 700–800 gr at the time of the outbreak. The reported mortality reached about 10%, a high percentage for a field outbreak among QTL carrying fish. The IPNV infection was identified following an inspection by the veterinary authorities. The diagnosis was later validated by detection of the VP2 segment of the viral genome using polymerase chain reaction (qRT-PCR) of the infected head-kidney and through histopathological and immunohistochemical examinations of the liver and pancreatic tissues (Figures 1A,B), all performed at the Fish Vet Group Norway (http://fishvetgroup.no) (Supplementary Table S1). Tissues were fixed in a 4% formalin solution (4% formalin, 0.08 M sodium phosphate, pH 7.0), processed using Thermo Scientific Excelsior® and embedded in Histowax with the aid of Tissue—Tek®, TEC 5 (Sakura) embedding system. The embedded tissues were then sectioned at 1.5–2 µm using a Leica RM 2255 Microtome and stained with Hematoxylin-Eosin (HE). The stained sections were scanned in an Aperio ScanScope AT Turbo slide scanner and read using Aperio Image Scope (Leica Biosystems). The scans were examined and scored in a blind fashion, i.e., without information about the animal’s history. A clip from the adipose fin was also sampled and stored in 95% ethanol for DNA extraction and genotyping of ten fish, five moribund and five dead. We genotyped these fish for the genetic markers used to assign individuals as QTL carriers (Houston et al., 2008; Moen et al., 2009, 2015) and confirmed that all animals carry IPN QTL.

The head-kidney was collected from six infected fish (http://fishvetgroup.no/pcr-2/#1452163253406-d81f6dce-bdf9), all at a moribund state, and stored in RNALater® (Ambion). Total RNA extraction was performed using the RNeasy® Plus mini kit (Qiagen). Nanodrop ND-1000 Spectrophotometer (NanoDrop Technologies) was used to assess the concentration and purity of the extracted genetic materials. The Norwegian Sequencing Centre performed library construction and sequencing of the total RNA on an Illumina HiSeq 4000 platform as paired-end (PE) 150 bp reads.

We first cleaned the sequence data by removing low-quality nucleotides and sequencing adapters using Trimmomatic (v.0.36) (Bolger et al., 2014). The ribosomal RNA (rRNA) was then identified through a similarity search against the SILVA rRNA database (Quast et al., 2013) and excluded from subsequent analysis. The host-specific transcriptome was detected by aligning the remaining reads against the Atlantic salmon reference genome assembly (ICSASG_v2) (Lien et al., 2016) using TopHat (v.2.0.13) (Trapnell et al., 2009, Trapnell et al., 2012). The un-aligned reads were then blasted against the IPNV reference assemblies (ASM397166v1, ASM397170v1, ASM397174v1, ASM397168v1, ASM397172V1 and ViralMultiSegProj15024) by setting “task” as “megablast,” “word size” to 7 and “expectation value threshold” to 1.00 e⁻⁰⁷. All the reads that aligned to the IPNV reference sequences were extracted, pooled into a single sequence datafile per animal, and then fed into Trinity (v.2.11.0) (Grabherr et al., 2011) with the default parameter settings to construct assemblies. We refer to the assembled viral genomes as BG_x_y, where x denotes the animal’s id (1–6) and y refers to the viral genome’s specific segment (e.g., VP1 or polyprotein).

We compared the assembled viral genomes to one another (Supplementary Tables S2, S3), and against various reference isolates, with different degrees of virulence (Dopazo, 2020; Julin et al., 2013; Munang’andu et al., 2016; Mutoloki et al., 2016; Santi et al., 2004; Song et al., 2005) (Table 1) or strains representing different genogroups, as previously suggested (Blake et al., 2001) (Table 2). Further, we assessed the deduced amino acids of these isolates to any IPNV sequence in the NCBI database with a complete polyprotein or VP1 information (Data Sheets S1–S3). The sequences included 90 isolates with the complete coding sequence for the polyprotein and 120 for the VP1. Both the nucleotide and amino acid alignments were performed using ClustalX (Larkin et al., 2007) implemented in Unipro UGENE (v36.0) (Okonechnikov et al., 2012) or Muscle (Edgar, 2004) implemented in Seqotron (v1.0.1) (Fourment and Holmes, 2016). Unipro UGENE and Mega X (v10.1.7) (Kumar et al., 2018) were also used to assess sequence similarity, calculate pair-wise distances for nucleotide and amino acid data, and construct phylogenetic associations using the neighbour-joining (NJ) method. To assess the confidence of node assignments, we performed 1,000 bootstrap replications.

This paper describes the genomic features of IPNV isolates responsible for a field outbreak in an Atlantic salmon population. The post-mortem examination identified IPN infection as the chief cause of mortality among all fish investigated. The histopathological and immunohistochemical examinations (Figures 1A,B; Supplementary Table S1) and the viral genetic material detected through qRT-PCR amplification (Supplementary Table S1) further confirmed earlier diagnosis. We found that the genotypes of all animals examined in this study were consistent with an IPN resistant animal (Houston et al., 2008; Moen et al., 2009; Moen et al., 2015). Further, assessment of the transcriptome sequence data confirmed that all fish carried at least one favorable copy of the causal mutation in the cdh1 gene, conferring resistance against IPNV (Moen et al., 2015). The causal mutation, which is in the form of a single nucleotide polymorphic (SNP) variation, is located in the protein-coding sequence of cdh1 (ssa26:15192533; C/T) and can change Proline (P) to Serine (S), with the former amino acid found to be associated with the IPNV resistance (Moen et al., 2015). It is expected that in a heterozygote fish, the favorable allele exhibit a close to complete dominant effect in response to the virus (Moen et al., 2015). Among the animals investigated in this study, three fish were homozygous for the favorable allele (C/C), and the other three were heterozygous (C/T). The extracted RNA was sequenced to high depth, with an average of 60 million PE reads per animal. Following the rRNA’s exclusion, more than 45 million PE reads per fish remained for the assemblage of the host’s viral genome and profiling gene expression. Together, the two segments of the IPNV genome consist of about 5,800 nucleotides. On average, we obtained 56-fold coverage of the viral genome from each animal in our sequence data.

Comparative genomic analysis between the six assembled genomes revealed at least two different, closely linked variants, caused the field outbreak in the population investigated (Figures 2A,B; Supplementary Table S2A, S2B). The nucleotide sequences of the polyprotein are identical in four of the assemblies (BG_1_polyprot to BG_4_polyprot) and contain nine polymorphic sites compared to the other two assembled sequences (i.e., BG_5_polyprot and BG_6_polyprot). Two of the polymorphisms are nonsynonymous, changing the amino acid residues in positions 47 (aspartate (D) to glutamate (E)) and 717 (lysine (K) to glutamine (Q)).

In a study aiming to trace the IPNV infection signatures from hatcheries to the sea (Kristoffersen et al., 2018), the authors have recovered a partial fragment of the polyprotein (2,772 bp) from an isolate (accession number: MH562009), with very similar nucleotide (∼99.6%) and amino acid (∼99.8–100%) sequences to the ones reported here (Supplementary Figures S1A, S1B). Recently, in controlled infection challenge testing, the mortality induced by this isolate among full-sib families was compared with a standard, virulent form of the virus (AY379740.1). While the infection with the latter variant resulted in approximately 30% mortality, mainly among the non-QTL fish, the former isolate caused more than 75% mortality, treating the QTL and non-QTL fish alike (Ørpetveit et al. in preparation; https://www.vetinst.no/nyheter/endringer-i-ipn-viruset-gjor-fisken-mer-utsatt-for-sykdom, https://www.fishfarmingexpert.com/article/changes-in-ipn-virus-make-salmon-more-susceptible, https://salmobreed.no/articles/benchmark-genetics-intensiverer-forskningen-pa-nytt-ipn-virus). Comparison between the nucleotide sequences of this isolate with our assembled genomes shows divergence ranging from 0.003 to 0.004 (Supplementary Figure S1A). However, this isolate has deduced amino acid sequences identical to the translated sequences from our first four assemblies (Supplementary Figure S1B), indicating that this variant is likely to be the ancestral form and the other isolates (i.e., BG_5_VP2 and BG_6_VP2) are more recently derived. One of the main characteristics of RNA viruses is their high mutation rates, mainly attributed to their lack of an effective proofreading activity. Therefore, finding different sequence variants from the same outbreak might not be surprising, but it indeed is important, as it will allow one to follow the trajectory of pathogen evolution through time and space.

Comparative assessment of the polyprotein nucleotide (data not shown) as well as the amino acid sequence data with multiple reference strains (Figure 2A) shows our assemblies are most closely related to the Sp serotype (genogroup 5). However, they also constitute a distinct group within this serotype compared to Sp isolates assessed in this study (Figure 2A; Supplementary Figure S2A). Several investigations have previously reported variations in some amino acid residues, particularly in the VP2 of the polyprotein, to generally be associated with a higher degree of the IPNV virulence (Bruslind and Reno, 2000; Munang’andu et al., 2016; Mutoloki et al., 2016; Santi et al., 2004; Shivappa et al., 2004; Song et al., 2005). Interestingly, the amino acid motifs found in the six assembled genomes in this study are mainly consistent with the virus’s avirulent form (Table 1) (Santi et al., 2004; Shivappa et al., 2004; Song et al., 2005). For instance, for the most critical positions associated with the virulence, i.e., 217 and 221 of the VP2, the amino acid residues of our assembled genomes are proline (P) and threonine (T) (Table 1; Supplementary File S1), while it is expected that in the virulent form, the corresponding residues to be T and A, respectively (Santi et al., 2004; Song et al., 2005; Mutoloki et al., 2016). Similar patterns have also been reported from the isolates recovered from outbreaks, with the amino acid signature comparable to the non-virulent forms in the key positions (Bain et al., 2008; Maj-Paluch et al., 2020; Ruane et al., 2009, Ruane et al., 2015; Tapia et al., 2015). On the other hand, we identified ten amino acid residues in the assembled genomes’ polyprotein that are either exclusively or with a disproportionate frequency in the reported isolates (Table 2; Supplementary File S1). Nine of these sites are within the VP2, and one is in the VP4. In the VP2, these residues are all close to one another, encompassing the hypervariable region (Blake et al., 2001; Heppell et al., 1995b). Specifically, these residues are in positions 245, 248, 252, 255, 257, 278, 282, 285, 321 (Table 2; Supplementary File S1). This data provides additional support for the previously suggested notion that the amino acid motifs in the VP2 capsid protein hypervariable region might play an essential role in the degree of the virulence displayed between different IPNV isolates (Bruslind and Reno, 2000; Skjesol et al., 2011). It is also likely that this specific constellation of the amino acid residues reported in these isolates had helped the virus to escape the host defence barrier, despite fish being regarded as genetically resistant to IPNV. In the VP4, the deduced amino acid residue, unique to the assembled isolates (Table 2; Supplementary File S1), is located in position 585. The amino acid substitution is a glycine (G) residue, compared to the other sequences investigated in this study, where they carry an aspartate (D) (Table 2).

While this paper was under review, another article also reported similar variations from isolates recovered from outbreaks in Scotland (Benkaroun et al., 2021). Taken together, these results confirm the virulence of this variant and provide some indication of the geographical regions it has spread.

All assemblies contain the VP5, the partial overlapping, non-structural protein, of segment A. Similar to the polyprotein, the phylogenetic analysis of the VP5 grouped the assemblies into two distinct groups and further clustered them with the isolates from the Sp serotype (Figure 2B). The two assembled groups differ in only one amino acid residue at position 90, with the first four assemblies carrying an arginine (R) while the two other assemblies have a glycine (G) (Supplementary File S3). Sequence analysis of the deduced amino acids of VP5 did not reveal any unique residue in the assembled isolates compared to all other sequences. However, the amino acid in position 98 differs between the Sp isolates (Supplementary File S3). At this position, all the investigated Sp isolates carry an R, while in the assemblies, the residue is tryptophan (W). It has been shown that the VP5’s protein is generally produced during the initial stages of replication (Heppell et al., 1995a), and at least in the Ab serotype, it seems that this protein to be involved in anti-apoptotic functions (Hong and Wu, 2002). However, this product’s importance in viral growth or virulence has so far remained ambiguous (e.g., Julin et al., 2013; Dopazo 2020).

The phylogenetic analysis of VP1 clustered the assemblies into three distinct clades (Supplementary Figure S2B) with the amino acid variations in positions 119 (threonine (T) to methionine (M)) and 701 (T to alanine (A) and serine (S)) (Supplementary File S2; Supplementary Table S3A). The nucleotide sequence comparisons show a greater divergence rate in the VP1 of the assemblies than those of the polyprotein, suggesting that the B segment of the genome has a higher rate of nucleotide mutation (Supplementary Tables S1, S2). Comparing the deduced amino acid sequences of the assembled genomes to the VP1 section of 120 IPNV sequences further revealed only a single, almost unique residue in our isolates (Supplementary File S2). The amino acid residue in position 656 of the assembled isolates is glutamate (E), while in the majority of other strains, the amino acid is either aspartate (D) or alanine (A). The only other isolates, carrying an E in this position, are two sequences recovered from Atlantic salmon (KY548520; collection date: 2001) and Brown trout (Salmo trutta L.) (KY548519; collection date: 1989) in Finland following outbreaks (Holopainen et al., 2017). However, it should be noted that so far, the role of VP1 in determining the degree of IPNV virulence has remained ambiguous. For example, in a study conducted by Song et al. 2005, through construction and comparative assessment of chimeric IPNV strains, the authors suggested that VP1 is not involved in this pathogen’s virulence. On the other hand, the data reported by Shivappa et al. 2004 suggests that the amino acids on the B segment of the genome in combination with specific residues in VP2 might help modulate the magnitude of virulence. Work in the infectious bursal disease virus (IBDV) has also shown that variations in the amino acid residues of the VP1 can change the kinetics of viral replication and influence the virus’s virulence (Liu and Vakharia, 2004). VP1 interacts with VP3 to form VP1-VP3 complexes, and as such, can affect various aspects of the virus biology, including its replication efficiency (Tacken et al., 2002; Pedersen et al., 2007).