Ube2g2-deficient cells are resistant to + RNA viruses
Ube2g2 is a cofactor of Aup1 that we identified as a potential host dependency factor for flavivirus infections 15. Ube2g2 is a E2-conjugating enzyme of the ubiquitylation cascade with several distinct binding domains for RING E3-ligases, ubiquitin and G2BR 16,17 (Fig. 1a). To study the functional impact of Ube2g2 on virus infection, we generated Ube2g2−/− cells using CRISPR/Cas9 gene editing. Several single clones with no detectable Ube2g2 were identified; multiple clones were tested to display the same phenotype and clone I was used in the experiments described in the following sections (Fig. 1b). To further investigate if the catalytic activity of Ube2g2 was essential during Zika virus infection, we generated a cell line stably expressing a catalytically inactive form of Ube2g2 using site-directed mutagenesis. The common catalytic unit of E2 enzymes contains a conserved active-site cysteine (C89), which is essential for Ub/Ubl transfer 16. We reconstituted Ube2g2−/− cells with either the wild-type or catalytically inactive C89K mutants. Both were expressed at comparable levels. While Ube2g2WT was found in its ubiquitin modified form, Ube2g2C89K remained unmodified (Fig. 1b). This was also reflected in total cellular ubiquitylation profiles, which was dramatically reduced in the Ube2g2C89K compared to Ube2g2WT cells (Fig. 1c). Wild-type and Ube2g2−/− cells were infected with flaviviruses (Zika or Dengue), and viral titres in the supernatants were determined using RT-qPCR and plaque assays. Deletion of Ube2g2 resulted in significant reduction of virus production (Fig. 1d). A similar reduction was also observed for coronaviruses (HCoV-229E and OC43) (Fig. 1e). Virus production could be rescued in cells reconstituted with Ube2g2WT but not Ube2g2C89K suggesting that the catalytic activity of Ube2g2 is necessary for its role in the viral lifecycle (Fig. 1f). We and others have previously shown that Ube2g2 interacts with Aup1 via its G2BR domain 8,14,15. To test whether this interaction is retained in virus-infected cells, we immunoprecipitated endogenous Ube2g2 using protein G beads from mock and virus-infected cells. The eluate was resolved by SDS-PAGE and immunoblotted with antibodies for the indicated proteins in total cell lysates and IP elute. As anticipated, Ube2g2 interacted with Aup1 in both mock and virus-infected samples during the course of infection (Fig. 1f).
Ube2g2 Is Necessary For Virus Replication
We next assessed the stage in the viral life cycle that requires Ube2g2. To exclude the possibility that Ube2g2 deletion disrupts viral entry, we took advantage of a fluorescence-based assay to evaluate entry and fusion of the virus with host endosomal membranes. Wild-type and Ube2g2−/− cells were infected with Zika virus pre-labelled with the fluorescent probe octadecyl rhodamine B (R18). R18 is self-quenched at high concentrations, but upon fusion with endosomal membranes, dilution of the R18-labelled virus leads to increased fluorescence intensity that can be quantified using flow cytometry. Viral internalisation can therefore be visualized by confocal imaging of R18-punctae (Fig. 2a). Red punctate structures indicating fusion of internalized virus particles with endosomal membranes were visible in both wild-type and Ube2g2−/− cells. We quantified the percentage of R18 positive cells as a read-out for the proportion of virus particles that gain entry into cells. Wild-type and Ube2g2−/− cells incubated with R18-prelabeled culture medium were used as negative control. ~40% of the population were R18 positive in both WT and Ube2g2−/− cells, with no significant difference detectable between the two, indicating that Ube2g2 has no impact on viral entry (Fig. 2b).
To evaluate the impact of Ube2g2 deletion on virus replication, we developed a Zika replicon system carrying a Renilla luciferase-reporter. Zika replicon is a self-replicative viral RNA which expresses all the viral non-structural genes with the viral structural genes replaced with a luciferase reporter (Fig. 2c). Luciferase intensity is measured as a read out for viral replication. The replicon system therefore provides a useful tool to decouple viral replication from other confounding factors in the viral lifecycle such as entry, assembly and secretion. Zika replicons were transfected into wild-type and Ube2g2−/− cells; at indicated timepoints, the cells were lysed and the luciferase intensity measured (Fig. 2d). The luciferase intensity remained comparable at early time points during translation, but starting from 6h (at the replication phase) increased in the wild-type cells, while that in Ube2g2−/− cells remained significantly lower, indicative of defective replication (Fig. 2d). Accordingly, this difference in replication was even more pronounced at later time points (24 and 48 h post transfection) (Fig. 2e). To confirm these data, we determined the intracellular Zika RNA production in infected wild-type and Ube2g2−/− cells using RT-qPCR. The results presented as 2^-ΔΔCt (WT/ Ube2g2−/−) showing fold difference between intracellular viral RNA in wild-type and Ube2g2−/− cells indicate that viral genome replication was significantly lower in Ube2g2−/− cells compared to wild-type cells (Fig. 2f).
To further confirm the defect in virus replication we visualised Zika virus replication sites using immunofluorescence. Wild-type cells infected with Zika for 24h displayed a typical perinuclear distribution of viral E protein (Fig. 2g). Double-stranded RNA (dsRNA) generated as intermediates during virus RNA replication largely co-localised with Zika E protein in the wild-type cells. In contrast, virus-infected Ube2g2−/− cells displayed a dispersed E protein and dsRNA distribution, indicative of defective replication sites (Fig. 2g, h).
Furthermore, transmission electron microscopy (TEM) analysis of wild-type and Ube2g2−/− cells infected with Zika for 24h revealed viral progenies in wild-type but significantly attenuated in the Ube2g2−/− cells. In addition, appearance of virus replication organelles could be seen in the wild-type cells, but were lacking in the Ube2g2−/− cells (extended Fig. 1a, b 1c). In Zika-infected wild-type cells, ER sheets were substantially dilated; large amount of virus induced membrane invaginations, often referred to as vesicle packets were observed to accumulate. In comparison to WT, the typical condensed convoluted membranes and replication organelles were not observed in Ube2g2−/− cells. Instead, more autophagosome like double membrane vesicles and lysosomes were present in the Ube2g2−/− cells as described later (extended Fig. 4).
Defective Expression Of The Viral Replication Complex In Ube2g2 Cells
The Zika virus non-structural proteins interact with each other to form the replication complex detected as membrane invaginations formed upon ER-remodelling 18,19. Given the defect in replication and replication organelle morphology (Fig. 2), we aimed to determine whether this complex is affected. We first measured the steady state expression levels of all viral proteins in the wild-type and Ube2g2−/− cells at 24h and 48 hours post infection (Fig. 3a). At both time-points, expression of the non-structural proteins was significantly attenuated in the Ube2g2−/− cells. Viral non-structural proteins displayed a more pronounced defect compared to structural proteins prM and E evident at 24 h, supporting a potential defect in formation of the replication complex. Loss in expression of the replication complex proteins e.g. NS5, NS4A-4B at earlier time points in infection indicate that either synthesis or turnover of these proteins is directly affected in Ube2g2−/− cells, resulting in defective replication and therefore a cumulative defect in viral protein expression at later timepoints (Fig. 3a).
To further investigate this phenotype and differentiate between defective synthesis versus increased degradation of the viral proteins, we measured the fate of newly synthesised viral proteins using pulse chase in radiolabelled cells (Fig. 3b-h). Zika-infected wild-type, Ube2g2−/−, Ube2g2WT and Ube2g2C89K cells (12h, 24h and 48 h post infection) were pulse labelled with [35S]cysteine/methionine and chased in cold media for the indicated time intervals. At each time point, viral proteins were immunoprecipitated, resolved by gel-electrophoresis and detected by autoradiography (Fig. 3b). In addition, production of virus particles was also measured from supernatants of these cells (Fig. 3c). As observed with plaque assays, virus production from Ube2g2-deficient cells was substantially lower compared to Ube2g2-proficient cells (Fig. 3c). Interestingly, while synthesis of the structural proteins (E, prM) was only mildly slower in Ube2g2-deficient cells, particularly at 24h post infection (Fig. 3d, g), synthesis of non-structural proteins was followed by their rapid degradation over time in the Ube2g2−/− and Ube2g2C89K cells (Fig. 3e, f, g). Quantitation of protein turnover therefore indicate that the replication proteins do not form a stable complex in the absence of Ube2g2 (Fig. 3g), which very likely results in impaired membrane remodelling necessary for generating replication organelles. Interestingly, although synthesis of the structural proteins was not dramatically reduced in Ube2g2-deficient cells, they did not form the oligomeric complexes that is typical in wild-type cells (Fig. 3h). Collectively, these data indicate that defective membrane rearrangements underpin the block in biogenesis of virus replication organelles.
Ube2g2 Is Required For Virus-triggered Lipophagy For Assembly And Secretion
A consistent feature of membrane remodelling during + RNA virus infection is selective autophagy of lipid droplets 5,7,9,10,20. Our previous study demonstrated that Aup1 is important for virus-triggered lipophagy, for assembly and secretion of viral progenies 8. Since Ube2g2 is a co-factor of Aup1 and was also identified in our initial screen, we aimed to determine whether it was also required for lipophagy and subsequent virus assembly/secretion. To specifically address assembly and secretion of virus particles, we took advantage of the Zika-PrME virus like particles (Zika-VLP) system (Fig. 4a). These cells stably express Zika virus like particles containing only the virus structural proteins (pr, M and E) but not the viral nucleocapsid, thus allowing us to study the virus release in the absence of viral entry and replication 9. We transfected cells constitutively expressing Zika-VLPs with either DsiRNA targeting Ube2g2 or non-targeting control DsiRNA, and the expression of intracellular (cell lysates) and extracellular (supernatant) VLP production was assessed using Western blotting (Fig. 4b).
Compared to Zika-VLP cells transfected with non-targeting siRNA, cells transfected with siRNA targeting Ube2g2 showed no detectable reduction in intracellular VLP production; however, a dramatic reduction of released mature form of VLPs was observed in the supernatants of Ube2g2-depleted cells (Fig. 4b, c). Lipophagy is critical for assembly and secretion of progeny virions 8,9. Contribution of lipid droplet degradation has also been reported for secretion of heavy cargo through the secretory pathway 21. In line with these findings, Ube2g2-deficient cells inhibited maturation and secretion of VLPs as anticipated, in support of its role in lipophagy. This defect in secretion was not specific for VLP secretion, but could also be observed for Vesicular stomatitis virus glycoprotein (VSV-G) (Fig. 4d, e). Wild-type and Ube2g2−/− cells expressing VSV-G were pulsed with [35S]cysteine/methionine for 10min and chased for indicated time intervals. Immunoprecipitated VSV-G were EndoH-treated to measure transport from the ER to the Golgi (Fig. 4d). VSV-G transport was quantitated as amount of EndoH resistant forms as a fraction of total (Fig. 4e). VSV-G transport was found to be impaired in the Ube2g2−/− cells compared to wild-type cells. In addition, the ER-resident pool of VSV-G was found to undergo degradation over time (Fig. 4d, e). Collectively, these data suggest that Ube2g2 might play a dual function in the viral life-cycle – first in the replication process via biogenesis of replication organelles, and second in lipophagy to facilitate secretion of assembled viral progenies.
To determine if lipophagy was affected in Ube2g2-deficient cells, we first stained LDs in cells expressing GFP-tagged Ube2g2 (Fig. 5a-d). Ube2g2 colocalised with lipid droplets, particularly in virus-infected cells, as observed in cells expressing Ube2g2-eGFP either pulsed with a fluorescent fatty acid reporter (Bodipy 558/568 C12) (Fig. 5a, b) or stained with Nile Red to visualize LDs (Fig. 5c, d). This was in concert with Aup1, which displayed a similar distribution in virus infected cells (extended Fig. 2a, b).
To measure lipophagy, we visualised LDs in wild-type and Ube2g2−/− cells at 48 and 72h post infection. While in wild-type cells LDs underwent hydrolysis as anticipated, in Ube2g2−/− cells we observed an accumulation of LDs as measured by Nile red staining of LDs by immunofluorescence and flow cytometry (Fig. 5e, f). These data are in support of Ube2g2 facilitating virus-triggered lipophagy as part of the Aup1-Ube2g2 complex.
To determine how Ube2g2-deficiency impaired lipophagy, we first measured Aup1 levels in wild-type and Ube2g2-deficient cells. Defective lipophagy was not due to loss of Aup1, since its expression levels in Ube2g2-deficient cells remained unaffected (Fig. 6a). We also measured induction of autophagy by immunoblotting for LC3. Interestingly, even in mock infection, basal levels of autophagosomes (LC-II) were significantly high in the Ube2g2−/− cells, which was increased further in virus-infected cells (Fig. 6b). This effect was accompanied by an increase in expression of lysosomes as detected by Lamp2 (Fig. 6c). Confocal imaging of Aup1 and Lamp2 also revealed increased colocalization of the two in Ube2g2−/− cells compared to that of wild-type (extended Fig. 3a, b). To further characterise this process, we visualized autophagic flux in wild-type and Ube2g2−/− cells using cells stably expressing RFP-GFP-LC3. While increased autophagic flux was detected in wild-type cells upon virus infection, in Ube2g2−/− cells, high abundance of autolysosomes was detected even in uninfected cells, which remained at high levels upon infection (Fig. 6d, e), in line with the biochemical data. Quantification of autophagosomes versus autolysosomes by flow cytometry were also in line with confocal imaging analyses, showing increased autolysosomes in Ube2g2−/− cells (Fig. 6f). These data indicate that defective lipophagy is not on account of blocked autophagosome formation, but more likely due to aberrant autophagy in Ube2g2-deficient cells. To specifically rescue the phenotype of impaired lipophagy, we measured viral titres in cells exogenously supplied with free fatty acids 8,22. Interestingly, while virus production in Aup1-deficient cells could be rescued by exogenous free fatty acids, this could not be rescued in Ube2g2-deficient cells (Fig. 6g), suggesting that Ube2g2 performs an additional function besides lipophagy in the viral life-cycle.
Ube2g2-dependent degradation of stress chaperones is a key feature of membrane remodelling during biogenesis of virus replication organelles
Given the increased autophagsomal flux in parallel to impaired selective lipophagy, we hypothesised that the underlying mechanism of defective virus replication and assembly in Ube2g2-deficient cells was due to aberrant autophagy. Replication organelles are formed upon remodelling of the ER 4,23,24. We therefore hypothesised that a combination of (i) defective lipophagy and (ii) increased ER-phagy, resulted in defective membrane remodelling required for biogenesis of replication organelles in flavivirus infection (Fig. 7a). The induction of UPR and ER-expansion is a necessary step for biogenesis of viral replication organelles 25–27. Flaviviruses achieve this via a couple of means: they encode a cysteine protease NS3, which cleaves the ER-phagy receptor FAM134B, to induce ER expansion 28. In addition, infection generates an oxidative environment beneficial to virus replication, which in turn results in stress to induce XBP-1 splicing and induction of downstream effectors essential for viral replication complex formation 29–32.
To verify UPR induction, we measured XBP1 splicing in virus-infected cells. In both wild-type and Ube2g2-deficient cells, infection was accompanied by XBP1 splicing (Fig. 7b, c). Spliced XBP1 was detected by RT qPCR (Fig. 7b, c) as fold change compared to wild-type mock-infected cells. While in both the wild-type and Ube2g2-deficient cells, virus infection triggered increased Xbp1 splicing, the basal levels of spliced Xbp1 were substantially higher in Ube2g2−/− cells compared to that of wild-type (Fig. 7b, c). UPR induction was therefore comparable in wild-type and Ube2g2-deficient cells. One of the viral strategies of inducing ER-expansion is via cleavage of the ER-phagy receptor FAM134B 28,33,34. To evaluate virus triggered ER expansion, we measured FAM134B expression in wild-type and Ube2g2-deficient cells (Fig. 7d). As anticipated, FAM134B cleavage remained unaltered between virus-infected Ube2g2 proficient and deficient cells across all time points. The abundance of FAM134B was found to decrease over the time course of infection in Ube2g2-deficient cells, most likely due to FAM134B-independent ER-phagy (Fig. 7d).
ER expansion is typically followed by stress recovery of the ER to prevent apoptosis and occurs via degradation of specific chaperones. In particular, Herp/Herpud1, Sel1L, Hsp70 are rapidly downregulated to initiate recovery 35–37. To measure rescue and remodelling of the ER we therefore measured specific chaperones associated with ER rescue (Fig. 7e). Infected wild-type and Ube2g2-deficient cells were pulsed with [35S]cysteine/methionine and chased for 45 mins. Interestingly, in infected wild-type cells, Herpud1, Sel1L and calreticulin underwent rapid degradation following infection in wild-type and Ube2g2WT cells, which was blocked in both the Ube2g2−/− and Ube2g2C89K cells with increase in synthesis over time. Synthesis and turover of Hsp70 on the other hand remained unaffected (Fig. 7e). Rtn3 38,39, an ER-phagy receptor remained unaffected. In addition, expression of Sec62, which is also an ER-phagy receptor 40–43, was induced specifically in the Ube2g2-deficient cells. Our data therefore indicate that Ube2g2 deficiency resulted in stabilisation of stress response chaperones along with increased expression of Sec62, which consequently triggers Sec62-dependent ER-phagy during infection.
Sec62-mediated ER-phagy was recently described as Chmp4 dependent 43,44. To test if this pathway was triggered in Ube2g2-deficiency, we depleted either Sec62 or Chmp4 in the wild-type, Ube2g2−/−, Ube2g2WT and Ube2g2C89K cells (Fig. 7f). These cells were challenged with Zika to measure production of infectious virus particles (Fig. 7g, h). Interestingly, in both Sec62 and Chmp4 depleted cells, defective virus production upon Ube2g2-deficiency was rescued to wild-type levels, in support of the hypothesis that Ube2g2 specifically inhibits this pathway during virus infection (Fig. 7g, h). To determine whether the ER-stress chaperones were driven toward rapid degradation by the viral protease, we expressed NS2B-NS3 alone in wild-type, Ube2g2−/−, Ube2g2WT and Ube2g2C89K cells (Fig. 7i). The rapid turnover of the stress response proteins was recapitulated in the wild-type and Ube2g2WT cells, confirming that the viral protease was indeed able to downregulate them. Moreover, this effect was blocked in the Ube2g2-deficient cells, verifying that degradation was due to Ube2g2-mediated ubiquitylation (Fig. 7i). Altered morphology of the ER with increased abundance of lysosomes was also observed by TEM analyses (extended Fig. 4a, b). Collectively, these data highlight that selective autophagy plays a crucial role in the viral life-cycle. While lipophagy is essential for assembly and secretion of viral progenies, inhibition of ER-phagy is equally critical to biogenesis of viral replication organelles.