A potentially novel overlapping gene in the genomes of Israeli acute paralysis virus and its relatives

Colony collapse disorder (CCD) is a syndrome characterized by the mass disappearance of honeybees from hives [1]. CCD imperils a global resource estimated at approximately $200 billion [2]. For example, it has been estimated that up to 35% of hives in the US may have been affected [3]. Many culprits have been suggested as causal factors of CCD, among them fungal, bacterial, and protozoan diseases, external and internal parasites, in-hive chemicals, agricultural insecticides, genetically modified crops, climatic factors, changed cultural practices, and the spread of cellular phones [1]. The Israeli acute paralysis virus (IAPV), a positive-strand RNA virus belonging to the family Dicistroviridae, was found to be strongly correlated with CCD [4]. It was first isolated in Israel [5], but was later found to have a worldwide distribution [4, 6, 7].

The genome of IAPV contains two long open reading frames (ORFs) separated by an intergenic region. The 5' ORF encodes a structural polyprotein; the 3' ORF encodes a non-structural polyprotein [5]. The non-structural polyprotein contains several signature sequences for helicase, protease, and RNA-dependent RNA polymerase [5]. The structural polyprotein, which is located downstream of the non-structural polyprotein, encodes two (and possibly more) capsid proteins.

Overlapping genes are easily missed by annotation programs [8], as evidenced by the fact that several overlapping genes were only detected by using the signatures of purifying selection [9‐13]. Here, we apply a recently developed method for the detection of selection in overlapping reading frames [14] to the genome of IAPV and its relatives.

In the fourteen completely sequenced dicistroviral genomes (Table 1), we identified 43 same-strand overlapping ORFs of lengths equal or greater than 60 codons on the positive strand. Ten overlapping ORFs were found in concordant genomic locations in two or more genomes. The concordant overlapping ORFs were assigned to three orthologous clusters (Table 2). The overlapping ORFs in all three clusters are phase-1 overlaps, i.e., shifted by one nucleotide relative to the reading-frames of the known polyprotein genes. Two of the orthologous clusters overlap the gene encoding the nonstructural polyprotein and one overlaps the reading frame of the structural polyprotein. (In appendix 1, we present the results concerning the overlapping ORFs on the negative strand. We note, however, that dicistroviruses are not known to be ambisense [15].)

We identified a strong signature of purifying selection in cluster A that contains overlapping ORFs from four genomes: IAPV, Acute bee paralysis virus (ABPV), Kashmir bee virus (KBV), and Solenopsis invicta virus 1 (SINV-1) [16‐18]. This ORF overlaps the 5' end of the structural polyprotein gene (Figure 1A). The detection of purifying selection is based on a method for the simultaneous estimation of selection intensities in overlapping genes [14]. To ascertain that each overlapping ORF is indeed subject to selection, we used the likelihood ratio test for two hierarchical models. In model 1, we assume no selection on the overlapping ORF. In model 2, the overlapping ORF is assumed to be under selection. If model 2 fits the data significantly better than model 1 (p < 0.05), then the overlapping ORF is predicted to be under selection and is most probably functional. The signature of selection was identified for the ORFs in the three bee viruses (IAPV, ABPV, and KBV). The protein product of the orthologous ORF in SINV-1 could not be tested for selection because the amino acid sequence identity between the ORF from SINV-1 and the ORFs from the three bee viruses (Table 3) is lower than the range of sequence identities for which the method can be applied (65-95%).

An additional indication for selection on these ORFs was obtained by comparing the degrees of conservation of the hypothetical protein sequences of the overlapping ORFs against the protein sequences of the known genes (structural and nonstructural polyproteins, Table 3). The degree of amino-acid conservation and, hence, sequence identity between orthologous protein-coding genes is influenced ceteris paribus by the intensity of purifying selection. If both overlapping genes are under similar strengths of selection, the amino-acid sequence identity of one pair of homologous genes would be similar to that of the overlapping pair. On the other hand, if a functional gene overlaps a non-functional ORF, the amino-acid identity between the hypothetical protein sequences of the non-functional ORFs would be much lower than that between the two homologous overlapping functional genes. We found that the degree of amino-acid conservation of the overlapping sequence identity between pairs of overlapping ORFs in cluster A is only slightly lower than that of the known gene (maximum of 12% difference between IAPV and SINV-1 in cluster A, Table 3). In contrast, the amino-acid sequence identity between ORF pairs in clusters B and C is much lower than that between the pairs of known genes (maximum of 44% difference between CrPV and DCV in cluster C, Table 3).

The signature of purifying selection on the ORFs in cluster A suggests that they may encode functional proteins. We provisionally term this gene pog (p redicted o verlapping g ene). In Figure 1, we show that pog is found in the genomes of four viruses that constitute a monophyletic clade, but not in any other dicistrovirid genome (Figure 1A). Its phylogenetic distribution suggests that pog originated before the divergence of SINV-1 from the three bee viruses. The phylogenetic distributions of the ORFs in clusters B and C (Figure 1B) are patchy. This patchiness is an additional indication that the overlapping ORFs in clusters B and C are spurious, i.e., non-functional.

An examination of the DNA alignment of pog (Figures 2) reveals a conservation of the first potential start codon (ATG or CTG) in the +1 reading frame in three out of the four viral genomes (IAPV, ABPV, and SINV-1). As seen in Figure 3, this conservation cannot be explained by constraints on the overlapping polyprotein, in which the corresponding site is variable and encodes different amino acids (His, Asn, and Pro, in IAPV, ABPV, and SINV-1, respectively). We note, however, that we did not find a conserved Kozak consensus sequence [19] upstream of the potential initiation site. This situation is similar to that described in [13].

A protein motif search resulted in several matches, all with a weak score. Two patterns were found in all four proteins: (1) a signature of rhodopsin-like GPCRs (G protein-coupled receptors), and (2) a protein kinase C phosphorylation site (Figure 3). Prediction of the secondary structures [20] suggests that the proteins contain two conserved helix domains, separated by 3-5 residues (except for SINV-1, in which one long domain is predicted), at the C-terminus (Figure 3). A search for transmembrane topology [21] indicates that the longer helix may be a transmembranal segment (Figure 3). Although viruses often use GPCRs to exploit the host immune system through molecular mimicry [22‐25], the lengths of the proteins encoded by pog are shorter than the average virus-encoded GPCR. Therefore, these proteins may have a different function.