NS5A domain I antagonises PKR to facilitate the assembly of infectious hepatitis C virus particles

Hepatitis C virus NS5A is a multifunctional phosphoprotein comprised of three domains (DI, DII and DIII). DI and DII have been shown to function in genome replication, whereas DIII has a role in virus assembly. We previously demonstrated that DI in genotype 2a (JFH1) also plays a role in virus assembly, exemplified by the P145A mutant which blocked infectious virus production. Here we extend this analysis to identify two other conserved and surface exposed residues proximal to P145 (C142 and E191) that exhibited no defect in genome replication but impaired virus production. Further analysis revealed changes in the abundance of dsRNA, the size and distribution of lipid droplets (LD) and the co-localisation between NS5A and LDs in cells infected with these mutants, compared to wildtype. In parallel, to investigate the mechanism(s) underpinning this role of DI, we assessed the involvement of the interferon-induced double-stranded RNA-dependent protein kinase (PKR). In PKR-silenced cells, C142A and E191A exhibited levels of infectious virus production, LD size and co-localisation between NS5A and LD that were indistinguishable from wildtype. Co-immunoprecipitation and in vitro pulldown experiments confirmed that wildtype NS5A domain I (but not C142A or E191A) interacted with PKR. We further showed that the assembly phenotype of C142A and E191A was restored by ablation of interferon regulatory factor-1 (IRF1), a downstream effector of PKR. These data suggest a novel interaction between NS5A DI and PKR that functions to evade an antiviral pathway that blocks virus assembly through IRF1.

functions to block assembly of infectious virus particles.

49
Hepatitis C Virus (HCV) is an enveloped virus in the Flaviviridae family with a positive-50 sense, single-stranded RNA genome (2). HCV infection is a globally prevalent public the same phenotype as V67, ie reduced replication in Huh7 cells, which was restored 145 to wildtype (WT) levels in Huh7.5. Six residues proximal to V67 (S71), or P145 (C142, 146 Q143, P147, C190 and E191), were dispensable for genome replication in either Huh7 147 or Huh7.5 cells. In contrast, four residues proximal to P145 (P102, Y106, W111 and 148 F149) were absolutely required for genome replication in both cell lines and were thus 149 excluded from further analysis in this study. 150 C142 and E191 are required for virus assembly 151 We next investigated whether the eleven residues that were dispensable for genome 152 replication played a role in virus assembly and release. Alanine substitution of all of 153 these residues were generated in the full-length mJFH-1 infectious clone (49). The 154 assembly phenotype of these mutants was evaluated in Huh7.5 cells as in our hands 155 they more efficiently supported both virus genome replication (Fig 2 -compare WT 156 values in A and B), and assembly (data not shown). As shown in Supp. Fig. S1, nine we included C190A as a control in the detailed analysis of the C142A and E191A 166 phenotypes, as described henceforth. 167 We first confirmed virus genome replication for the mutant infectious clones C142A, 168 C190A and E191A both directly by qPT-PCR (Fig 3A), and indirectly by quantifying 169 NS5A positive cells using an IncuCyte S3 cell imager (50) (Fig 3B). Reassuringly, 170 replication of all three mutants was indistinguishable from WT, mirroring the SGR data. 171 Western blot analysis confirmed that NS5A and the structural proteins Core and E2 172 were expressed at equivalent levels to WT (Fig 3C). 173 To assess both virus assembly and release we proceeded to determine intracellular 174 and extracellular virus titres (Fig 3D). This analysis revealed that C142 was absolutely 175 required for virus assembly with levels of both intracellular and extracellular virus 176 indistinguishable from the negative control (NS5B GND). In contrast C190A had no 177 effect on WT levels of infectivity, and E191A exhibited an intermediate phenotype with 178 an approximately 2-log reduction compared to WT. We conclude that C142 and E191 179 play a role in virus assembly, and the fact that C190 is dispensable further suggests 180 that the disulphide bond observed in the 'open' structure of DI is not required for the 181 function of NS5A during genome replication or assembly. 183 To better characterise the role of C142 and E191 on infectious virus production, we 184 used high resolution confocal microscopy (Airyscan) to observe the co-localisation 185 between viral proteins and cellular factors. Key organelles during virus assembly are lipid droplets (LDs), to which both Core (51) and NS5A (52) are recruited. The 187 disruption of LDs either pharmacologically or genetically (53) inhibits virus assembly. 188 We previously showed that in cells infected with the P145A mutant virus, LDs were 189 more abundant and smaller in size compared to WT infected cells (1). As shown in Fig   190   4, this phenotype was recapitulated for both C142A and E191A: cells infected with WT 191 and C190A exhibited an average of 100 LD with a cross-sectional area of 192 approximately 1.0 μm 2 , whereas C142A and E191A displayed >200 LD with a 193 significantly smaller area (0.2 μm 2 ), similar to mock-infected cells (Fig 5A). 194 The differences in both LD size and quantity were consistent with the assembly 195 defective phenotypes of C142A and E191A. To extend this analysis we quantified the 196 colocalisation of Core and NS5A with LDs. As shown in Fig 5B,  with CypA to evade PKR-dependent antiviral responses. The differential sensitivity of 216 HCV to CypA in Huh7 compared to Huh7.5 cells was reminiscent of our initial 217 observation that P145A fails to undergo genome replication in Huh7 cells, but is only 218 modestly impaired in Huh7.5 cells (1) (Fig 2). This led us to assess whether either 219 CypA and/or PKR play a role in virus assembly and whether the NS5A DI mutants 220 studied thus far can shed light on such mechanisms. Consistent with previous studies, 221 we first confirmed that silencing of CypA or PKR in Huh7.5 cells had no effect on 222 genome replication (Supp. Fig. S3A). We did however, note that the production of 223 infectious virus was unaffected by PKR silencing but reduced by ~100-fold in CypA 224 shRNA cells (Supp. Fig. S3B). 225 We proceeded to analyse the genome replication and assembly of the 3 DI mutants 226 in PKR silenced Huh7.5 cells. As expected, genome replication ( Fig 7B) and viral 227 protein production (Fig 7C, D) were unaffected by the lack of PKR. Indeed, overall 228 levels of genome replication as measured by qRT-PCR modestly increased compared to Huh7.5 cells (compare Fig 7B to Fig 3A). A surprising picture emerged when we 230 analysed the assembly and release of the mutants: production of infectious virus by 231 C142A and E191A was restored to the same level as WT and C190A in PKR silenced 232 Huh7.5 cells (Fig 7E). To further confirm that this was due to the lack of PKR, as 233 opposed to an off-target effect of the sgRNA, we treated Huh7.5 cells with the small 234 molecule PKR inhibitor C16 at 24 h post electroporation.

NS5A DI mutants alter the morphology and distribution of lipid droplets
Reassuringly, 235 pharmacological inhibition of PKR function also restored the production of infectious 236 virus by C142A and E191A ( Fig 7F). These data are consistent with a role for PKR in 237 blocking virus assembly and point to a role of NS5A DI in antagonising this previously 238 undefined function of PKR.

239
PKR silencing restores the LD phenotype of the assembly mutants. 240 We next sought to determine whether the restoration of infectious virus production by 241 PKR silencing was associated with concomitant changes in LD morphology, 242 distribution and the association with NS5A and Core. We therefore repeated the and the three mutants with regard to LD number and size were observed (Fig 9A), 246 although it should be noted that overall the size of LDs in infected cells was slightly 247 reduced ( Fig 5A). Co-localisation analysis also revealed that, unlike in Huh7.5 cells, 248 no differences were observed between WT or the three mutants in terms of their co-249 localisation between NS5A and LD (Fig 9B), Core and LD (Fig 9C) or NS5A and Core 250 ( Fig 9D).

251
Lastly, we assessed the distribution of LDs by quantifying their distance from the 252 nucleus. As shown in the representative images in Fig. 10A during this process may be detected by PKR. We therefore assessed the co-dsRNA-specific antibody, J2 (57). As expected, in WT infected cells, we observed co-274 localisation of dsRNA with both NS5A and LDs (Fig 11). This co-localisation was also 275 quantified in cells infected with the three mutants and surprisingly, revealed no 276 significant differences in the co-localisation of NS5A and dsRNA (Fig 12A) or LD and 277 dsRNA ( Fig 12B). However, the number of dsRNA foci in C142A and E191A were 278 reduced compared to WT and C190A (Fig 12C-D NS5A did not. We also investigated whether NS5A was able to bind to activated PKR 300 by performing immunoprecipitations with an antibody to phosphorylated PKR (P-PKR).

301
Although overall levels of P-PKR were low, this analysis clearly showed that (as for 302 the total PKR) only WT and C190A NS5A co-precipitated with P-PKR ( Fig 13B). it is dependent on PKR catalytic activity (59).

318
Western blotting of infected cell lysates with antibodies to either Ser51-phosphorylated 319 eIF2α or total eIF2α revealed no differences between WT and the three mutants 320 (Supp. Fig. S4). We therefore concluded that eIF2α phosphorylation by PKR is not  Huh7.5 cells using CRISPR/Cas9 (Fig 14A), and electroporated these cells with WT 337 or the mutant HCV RNAs. As was observed for PKR silenced cells, genome replication ( Fig 14B) and viral protein production ( Fig 14C, D) was unaffected by the lack of by 339 IRF1 knockout. Reassuringly, when we analysed the assembly and release of the 340 mutants, the production of infectious virus by C142A and E191A were restored to the 341 same levels as WT and C190A in IRF1 knockout Huh7.5 cells ( Fig 14E). These data 342 confirmed that the ability of PKR to inhibit the assembly of HCV is mediated by its 343 activation of the downstream effector IRF1.

345
This study builds on our previously published work (1) and provides further evidence replication. This is consistent with NS5A mediating a switch from virus replication to 354 assembly, perhaps by interacting with a different subset of cellular and/or viral factors.

355
The nature of the switch remains obscure although phosphorylation of the serine 356 cluster in the low complexity sequence linking DI and DII has been proposed. As C142 357 and E191 are close to the C-terminus of DI it is conceivable that they are regulated by In this regard it is interesting to speculate that in the closed dimer C142 is partially 361 occluded within the dimer interface ( Fig 1C)  in the groove between the monomers that is a possible RNA binding motif (Fig 1B).

384
An attractive hypothesis is therefore that NS5A is involved in transporting nascent 385 genomic RNA from sites of replication to sites of assembly (63, 64). In this scenario, 386 the LD is a waystation on the route and at that point NS5A could deliver the RNA to 387 the Core protein, rather like a baton in a relay race. One potential consequence of 388 this is that the RNA would be transiently exposed in the cytosol, permitting detection 389 by innate cytosolic sensors such as PKR. PKR is activated by binding to short (30 bp) 390 dsRNA elements but can activated by imperfect dsRNA or single stranded RNA (65).

391
In this regard, the HCV 5' IRES has been shown to be both a potent activator (66) and 392 inhibitor (67) of PKR. We postulate that DI interferes with the binding of PKR to 393 nascent genomes, possibly by direct binding to PKR (Fig 13)      The funders had no role in study design, data collection and analysis, decision to 580 publish, or preparation of the manuscript.

581
to the mechanism of action for dimeric HCV inhibitors. Protein Science.