The use of ARFs in viruses belonging to the Flaviviriade family was first reported for the hepatitis C viruses [26, 27]. Recently, it has been demonstrated that WNV uses a short ARF, termed foo, for the synthesis of NS1’ , a known variant of the canonical NS1 [24, 28]. We earlier reported the presence of other ARFs embedded in the coding frame of the WNV genome . Our bioinformatic analysis detected six ARFs, one of which, designed WARF4, was the longest and restricted exclusively to lineage I of WNV. Since WARF4 is embedded in the NS4B gene, the novel protein has been renamed N-NS4B/WARF4. Our results suggested the production of N-NS4B/WARF4 protein in WNV infected horses because of their ability to mount a humoral immune response to N-NS4B/WARF4. However, there was no direct evidence proving the actual existence of the N-NS4B/WARF4 protein. In order to demonstrate the production of N-NS4B/WARF4 in vitro after WNV cells infection, we produced a monoclonal antibody to the N-NS4B/WARF4 COOH-terminal amino acid sequence (MAb 3A12). MAb 3A12 strongly reacted with VERO WNV infected cells by immunofluorescence and detected a ~ 28-kDa protein by western blotting. The predicted aminoacids of N-NS4B/WARF4 preclude the possibility that MAb 3A12 could react with epitopes shared with the NS4B protein (Figure 3A), however to support this prediction a western blotting analysis was performed. As shown in Figure 3B, MAb 3A12 did not recognize the recombinant COOH terminal portion of the recombinant his-tagged NS4B protein. In addition, the anti-NS4B did not recognize the recombinant His-tagged WARF4 protein. To definitely asses the specificity of MAb 3A12 against the alternative reading frame, four overlapping peptides covering the full N-NS4B/WARF4 COOH- terminal amino acid sequence were synthesized and analyzed for their reactivity to MAb 3A12. As shown in Figure 4 the results confirm the specificity of the monoclonal antibody and allow to recognize its epitope between the amino acids 165–212 of N-NS4B/WARF4 sequence. In addition we demonstrated that N-NS4B/WARF4 protein expression is restricted to WNV lineage I infection and that it is expressed at high level in the late phase of infection (Figures 7, 8). Overall, our results demonstrate that N-NS4B/WARF4 is a novel protein, different from NS4B, and that is expressed in WNV infected cells.
Furthermore, we indirectly demonstrated the “in vivo” production of N-NS4B/WARF4 by showing its immunoreactivity with human sera obtained from WNV infected patients (Figure 9). The heterogeneous reactivity to the recombinant WNV antigens displayed by sera testing positive for WNV reflects the complex humoral response elicited by WNV infection [29, 30]. In addition, it is known that the ARF proteins are expressed with both less and variable efficiency if compared to the canonical proteins .
To date, we have no experimental information on N-NS4B/WARF4 protein translation, but it appears reasonable to assume that a −1 ribosomal frame shifting mechanism produces the novel protein. Indeed, in Flaviviriade the translation process is implemented by a cap-dependent scanning process, which produces a single polyprotein . The sequence encoding N-NS4B/WARF4 COOH-terminal is in −1 frame, moreover it is far from the 5′ terminal end, lacks an AUG codon and no internal ribosomal entry site (IRES) is described for WNV. The translation by ribosomal frame shifting is the only realistic explanation for N-NS4B/WARF4 protein synthesis. Since the proposed model requires the presence of specific RNA structures such as slippery sequences associated with pseudknot [33–35], a bioinformatics analysis was performed to predict these structures. All the complete genomes of WNV available on gene bank were aligned and assigned to the two main lineages (Figure 10, Table 1). The strains belonging to the lineage 1 were first analyzed to confirm the association with WARF4 and then a further analysis was carried out looking for slippery sequences and pseudknot structure within the NS4B coding region. WARF4 was detected in 361 out of 368 genomes belonging to lineage 1. Seven genomes lacked WARF4 because of a single nucleotide substitution that interrupts the alternative frame. In the WARF4 group, two different and mutually exclusive slippery sequences with downstream frameshift-stimulating pseudknot structures were predicted. The first UUUUUUG sequence is the most representative (91%). The pseudknot structure is 80 nucleotides long and includes the initial aminoacids codified by the −1 frame. This slippery sequence is associated with the American viral strains. The second CCCUUUG/T sequence is present in 5% of the genomes, it is positioned 129 nucleotide upstream of WARF4 and has a pseudoknot structure of 40 nucleotides. The second slippery sequence is associated with circulating viral strain in Mediteranean Bacin and Est-Europe. The strain analyzed in this work belongs to this second group [Egyptian strain, an. AF260968]. It should be noticed that both the structures must promote the suppression of a termination codon (UGA) located just before the first codon of −1 frame .
The ribosomal frameshifting model also explains the discrepancy between the predicted and the observed molecular weight of the alternative N-NS4B/WARF4. The predicted alternative reading frame protein consists of 148 AA that account for a molecular weight of 16,7 kDa. However, the protein detected by MAb 3A12 in WNV infected cells migrates with an apparent molecular weight of about 28–30 kDa. Although this discrepancy could be due to a post-translation modification, the proposed model appears the most reasonable explanation for the observed molecular weight of N-NS4B/WARF4 protein. The novel protein would exist as COOH-terminal variant of the NS4B protein, indeed, the ribosomal shift in −1 frame would give rise to a NS4B variant protein where the last 123 AA should be replaced by a longer amino acid tail of 148 A. Thus, the variant protein should exhibit a molecular weight of about 30 kDa, consistent with our results (Figure 6, lane 3). The proposed ribosomal frameshifting model implies that the expression kinetic of N-NS4B/WARF4 should be like that of NS4B protein even if the amount of the novel protein should be less than that of NS4B. Figure 7 shows the expression of N-NS4B/WARF4 and NS4B in time-corse infection, the two proteins exhibit a similar kinetics. N-NS4B/WARF4 is clearly detected in the late phase of infection, such as the NS4B protein. The level of NS4B and N-NS4B/WARF4 proteins expression based on a densitometric analysis (data not shown) indicates a ratio of about 25 to 1 respectively. It should be highlighted that this ratio is estimated for the strain Eg101 [an. AF260968] that our biofinformatic analysis associates with the strain circulating in the Mediterranea area (Figure 10). The predicted pseudoknot associated with the American strain is thermodinamically more stable. It should be intersting to estimate this ratio in the American viral strain.