The amino acid sequence of the MuV SH protein is the most hypervariable in the genome, with differences between some strains exceeding 20% (23), further indicating its nonessential role in the life cycle of the virus. Nonessential proteins typically serve a luxury function, and in the case of the MuV SH protein, this appears to be the blocking of apoptosis in certain cell types (9, 13, 26). Multiple reports have also suggested that the paramyxovirus SH proteins may play an ancillary role in virus-mediated cell fusion (10, 21), aid virus transport through the secretory pathway (19), or act as a viroporin, a group of small, highly hydrophobic proteins that can oligomerize and form ion channels at the cell membrane to modulate membrane permeability (18). In vivo studies of SH gene deletion mutants of related viruses, such as human respiratory syncytial virus (hRSV) and avian metapneumovirus, have been interpreted to indicate a role of the protein in replication, virulence, and immunogenicity (3, 11, 14, 24, 25). However, in these studies, because the SH mutants were produced by deletion of the gene, it was not possible to discern if the observed altered phenotypes are related to the absence of the SH protein itself or if they are a consequence of altering the genome structure (e.g., a decrease in genome length, an altered untranslated region, a change in transcription gradient of downstream genes, or other yet-to-be-identified effects).

To investigate the role of the MuV SH protein in virus neurovirulence, we generated two modified versions of a molecular clone of the highly neurovirulent wild-type MuV clinical isolate 88-1961_WT, one wherein the SH gene was deleted (r88-1961_ΔSH) and one wherein three stop codons were introduced in the SH open reading frame (ORF) (r88-1961_SHstop) to abrogate SH protein expression without altering the structure of the genome.

The plasmid p88-1961_WT (15), which encodes the 15,384-nucleotide positive-strand (antigenomic) RNA sequence of 88-1961_WT, was modified by site-directed mutagenesis (SDM) to introduce the MluI and AscI restriction sites in the intergenic regions flanking the SH gene, resulting in plasmid p88-1961_AscI/MluI (Fig. 1). The restriction sites in p88-1961_AscI/MluI were subsequently used to introduce a linker, comprising only the 5′ and 3′ SH intergenic sequences, into the 88-1961 genome in place of the ORF, resulting in plasmid p88-1961_ΔSH (Fig. 1). Deletion of the SH ORF (174 nucleotides) maintained the genome length at an even multiple of six. The detailed cloning strategy for this construct and for those described below can be obtained from the authors upon request. Plasmid p88-1961_ΔSH was used to produce three independently rescued virus stocks of r88-1961_ΔSH from BHK BSR-T7/5 cells (15), which were amplified once on Vero cells. The genetic identities of the cDNA-derived r88-1961_ΔSH viruses were confirmed by sequencing a 1,366-bp region encompassing the deleted SH ORF. The restriction sites in p88-1961_AscI/MluI were also used to introduce a modified SH ORF containing three translational stop codons at positions 13, 73, and 115 of the SH ORF (genome positions 6279, 6341, and 6381) into the 88-1961 genome, resulting in plasmid p88-1961_SHstop (Fig. 1), which was used to produce two independently rescued virus stocks of r88-1961_SHstop from BHK BSR-T7/5 cells, amplified once on Vero cells. The genetic identities of the cDNA-derived r88-1961_SHstop viruses were confirmed by sequencing a 1,540-bp region encompassing the 173-bp SH ORF containing the stop codons and all sequences in the plasmid that had been produced by PCR. No mutations were identified other than those specifically introduced by SDM.

Our ability to confirm the effectiveness of the three stop codons was hindered by the lack of an antibody against the MuV SH protein. Therefore, the SH_WT and SH_STOP sequences were isolated from p88-1961_WT and p88-1961_SHstop, respectively, and subcloned into the pCAGGS vector (kindly provided by J. C. de la Torre, La Jolla, CA) (17) in-frame with a hemagglutinin (HA) tag at the carboxyl terminus. Following transfection of cells, lysates were evaluated by Western blotting. Expression of HA-tagged SH protein was detected in cells transfected with pSH_WT-HA (Fig. 2, lane 3) but not in cells transfected with pSH_stop-HA, (Fig. 2, lane 1), indicating, albeit indirectly, that the SH protein is not translated in the recombinant virus containing the three stop codons. Of note, insertion of the HA tag into the full-length recombinant viruses themselves was not performed, given the demonstration by others of HA tag interference with SH protein function (26).

To analyze the in vitro growth characteristics and to demonstrate the replication competence of r88-1961_SHstop and r88-1961_ΔSH relative to those of r88-1961_WT, we determined cumulative virus production during the first 7 days postinfection (p.i.) for each virus on Vero cells as described previously (15) (Fig. 3). A one-way analysis of variance (ANOVA) did not reveal statistical differences between the replication kinetics of r88-1961_WT and r88-1961_SHstop (P = 0.95); however, differences between the replication kinetics of r88-1961_ΔSH and 88-1961_WT were significant (P = 0.01). A post hoc Tukey's multiple-comparison test found differences between 88-1961_WT and r88-1961_ΔSH R1 and between 88-1961_WT and r88-1961_ΔSH R2 but not between 88-1961_WT and r88-1961_ΔSH R3. However, the three rescues of r88-1961_ΔSH replicated similarly to each other (P = 0.09; one-way ANOVA), and importantly, all were replication competent. Relative to that for r88-1961_WT, the onset of cytopathic effects (abnormal cell morphology, syncytial formation, and cell lysis) was more rapid in cells infected with r88-1961_ΔSH but not in cells infected with r88-1961_SHstop (data not shown). The faster destruction of the cell monolayer by r88-1961_ΔSH may account for the lower replication kinetics relative to that of r88-1961_WT. However, unpublished observations from our laboratory suggest that there is no correlation between the extent of virus fusogenicity or other in vitro cytopathic effects and virus virulence in our rat model.

Having confirmed that the modified 88-1961 viruses were replication competent, we compared the neurovirulence potentials of 88-1961_ΔSH, r88-1961_SHstop, and r88-1961_WT in vivo in a rat neurovirulence assay (20). No significant differences were apparent in the mean neurovirulence score of r88-1961_WT (17.6 ± 0.9) and that of r88-1961_SHstop (17.2 ± 0.7; P = 0.72; unpaired t test) (Fig. 4). Maintenance of the stop codons in the recombinant viruses from each rescue of r88-1961_SHstop following replication in rat brain was confirmed by extraction of total RNA from brain homogenates at day 3 p.i., the time of maximal virus production (16), followed by reverse transcription (RT)-PCR and sequencing (data not shown). In contrast, the mean neurovirulence score of 88-1961 _ΔSH (9.6 ± 0.6) was significantly lower than that of r88-1961_WT (17.6 ± 0.9; P = < 0.0001; unpaired t test). These data indicate the possibility that the attenuation observed by others may not be due to the absence of the SH protein per se but rather a consequence of disturbing the genome structure as a result of removing the SH transcription unit. The most likely cause of this would be an altered transcription/translation gradient, but this has proved difficult to assess. He et al. (8) and Wilson et al. (26) examined the effect of the removal of the SH gene from parainfluenza virus 5 (PIV-5) and were unable to reproducibly detect differences in levels of protein expression between the two viruses. Fuentes et al. attempted a similar analysis with an SH deletion mutant of hRSV and found variable results ranging from no effect to slight increases in expression levels of proteins both upstream and downstream of the deleted transcription unit (7). Thus, it is not clear from the literature if protein levels are affected by removal of the SH genes and, if they are, if such differences are detectable by Western blotting. While one could examine mRNA levels of the genes of interest, this has proven not to be predictive of protein levels (3). Whether altered levels of mRNA transcripts can affect virus phenotype without affecting protein levels is not known but seems unlikely. In the case of MuV, analyzing differences in gene-specific mRNA levels between two viruses is particularly complicated, given the abundance of polycistronic message species resulting in a laddering of transcripts that are not easily quantifiable (1, 5). Despite our inability to pinpoint the mechanism behind the observation that gene deletion alters virus phenotype, it is clear from our studies that the MuV SH protein per se is not a virulence factor in our animal model, although we cannot rule out that the MuV SH protein might be a virulence factor in a natural setting (i.e., respiratory infection in humans).