Roles of ESCRT Proteins ALIX and CHMP4A and Their Interplay with Interferon-Stimulated Gene 15 during Tick-Borne Flavivirus Infection

ABSTRACT Flaviviruses are usually transmitted to humans via mosquito or tick bites. During infection, virus replication and assembly, whose cellular sites are relatively close, are controlled by virus proteins and a diverse range of host proteins. By siRNA-mediated gene silencing, we showed that ALIX and CHMP4A, two members of the host endosomal sorting complex required for transport (ESCRT) protein machinery, are required during flavivirus infection. Using cell lines expressing subgenomic replicons and replicon virus-like particles, we demonstrated specific roles for ALIX and CHMP4A in viral replication and assembly, respectively. Employing biochemical and imaging methodology, we showed that the ESCRT proteins are recruited by a putative specific late (L) domain motif LYXLA within the NS3 protein of tick-borne flaviviruses. Furthermore, to counteract the recruitment of ESCRT proteins, the host cells may elicit defense mechanisms. We found that ectopic expression of the interferon-stimulated gene 15 (ISG15) or the E3 ISG15-protein ligase (HERC5) reduced virus replication by suppressing the positive effects of ALIX and CHMP4A. Collectively, these results have provided new insights into flavivirus-host cell interactions that function as checkpoints, including the NS3 and the ESCRT proteins, the ISG15 and the ESCRT proteins, at essential stages of the virus life cycle. IMPORTANCE Flaviviruses are important zoonotic viruses with high fatality rates worldwide. Here, we report that during infection, the virus employs members of ESCRT proteins for virus replication and assembly. Among the ESCRT proteins, ALIX acts during virus replication, while CHMP4A is required during virus assembly. Another important ESCRT protein, TSG101, is not required for virus production. The ESCRT, complex, ALIX-CHMP4A, is recruited to NS3 through their interactions with the putative L domain motif of NS3, while CHMP4A is recruited to E. In addition, we demonstrate the antiviral mechanism of ISG15 and HERC5, which degrades ALIX and CHIMP4A, indirectly targets virus infection. In summary, we reveal host-dependency factors supporting flavivirus infection, but these factors may also be targeted by antiviral host effector mechanisms.


RESULTS
ESCRT proteins ALIX and CHMP4A are essential for virus production. As ESCRT proteins play specific roles during retrovirus (18) and MBFV infection (23,24), we characterized the effects of the three essential ESCRT proteins-TSG101, ALIX, and CHMP4A-during TBFV infection. We initially silenced expression of these genes by transfecting A549 cells with targeting siRNA. Immunoblotting of cell lysates revealed that gene knockdown reduced protein expression (Fig. 1A). Furthermore, siRNA treatments that targeted these ESCRT expressions did not significantly increase cytotoxicity monitored by the amount of lactate dehydrogenase (LDH) enzyme released into cell culture media (Fig. 1B).
Subsequently, cells were infected with TBFV, including POWV, KFDV, TBEV Sofjin , and LGTV, and were compared to 2 MBFV -WNV Kunjin and ZIKV. The virus titers in cell culture supernatants were then enumerated by plaque assay (Fig. 1C). Interestingly, silencing of ALIX and CHMP4A resulted in reduction of extracellular virus titers between a 0.5 and 1 logarithm (log), compared to the nontargeting (NT) siRNA control (Fig. 1C). Surprisingly, depletion of TSG101 showed modest effects on the viruses investigated, compared to the NT siRNA control (Fig. 1C).
We further examined genome copy numbers of LGTV, WNV Kunjin , and ZIKV in virusinfected cell culture supernatants by qPCR. In accordance with the virus titers, depletion of either ALIX or CHMP4A resulted in significant reduction of genome copy numbers, whereas depletion of TSG101 had significant but very limited effects on the indicated viruses (Fig. 1D). Overall, these data suggest that ALIX and CHMP4A are essential for flavivirus production and may facilitate virus RNA replication.
Reduction of intracellular virus particles and genome copy numbers during ALIX or CHMP4A depletion. As the ESCRT proteins affected virus production, we sought to identify their roles during earlier stages before the virus release. We then determined virus titers and genome copy numbers intracellularly by repeatedly freezing and thawing infected cells. In accordance with virus titers in the supernatants, there was significant reduction of intracellular viruses. Silencing of ALIX or CHMP4A resulted in a 1 log reduction of virus titers, measured by plaque assay (Fig. 2A). Similarly, virus genome copy numbers from cell lysates were reduced when ALIX and CHMP4A were depleted. Particularly, LGTV infected cells showed the highest reduction (1 log) during CHMP4A, compared to the control. This effect was less severe in ZIKV and WNV Kunjin infected cells, with virus production reduced by half of a log (Fig. 2B), which indicated the preferred utility of ALIX and CHMP4A by LGTV. The similarity between results from extracellular and intracellular plaque assays suggests the ESCRT proteins do not act on virus release. The qPCR results from cell lysates suggest ALIX and CHMP4A play a role in either the replication or virus assembly stages.
Differential requirements of ALIX and CHMP4A in the flavivirus life cycle. As we speculated that components of the ESCRT proteins may act during virus replication or assembly, we generated cell lines expressing RNA WNV Kunjin or LGTV replicons to specifically study virus replication. Here, the RNA replicons (Fig. 3A) were transfected into baby kidney hamster 21 (BHK-21) cells. The 3'UTR was engineered with an internal ribosome entry site (IRES) to drive the expression of a neomycin/kanamycin resistance gene IRES-NeoR/KanR, allowing for the selection of cells expressing the subgenomic replicons (28). We also characterized expression of the ESCRT proteins during siRNA treatment of BHK-21 cells. Immunoblotting of cell lysates revealed that gene knockdown resulted in reduction of protein expressions (Fig. 3B).
Depletion of endogenous ALIX resulted in significant reduction in the expression of WNV Kunjin and LGTV replicons, up to 0.5 log, as monitored by the activity of luciferase reporter gene (Fig. 3C) and the genome copy number (Fig. 3D). siRNA knockdown of TSG101 and CHMP4A did not impact the subgenomic RNA replication, a result that reveals a specific role of ALIX in flavivirus replication.
As silencing of CHMP4A did not affect virus replication, we speculated that the protein may have a role in virus assembly. To test this hypothesis, the ESCRT proteins were depleted in cell lines expressing the replicons. This was followed by transfection of the (Continued on next page) C-prM-E constructs into these cells to supply the structural proteins in trans, generating replicon virus-like particles (RVPs). The supernatants containing RVPs were then used to infect naive BHK-21 cells. This was then followed by examining the expression of the luciferase reporter gene. As expected, silencing of CHMP4A resulted in significant reduction in RVPs production (about 0.5 log) (Fig. 3E). There was reduction in luciferase units in RVPs-infected cells during ALIX depletion, which could have resulted from the reduction of replication by ALIX silencing (Fig. 3E). Thus, we conclude that CHMP4A plays a role during virus assembly. These data suggest members of the ESCRT complex perform multiple functions during flavivirus infection.
ALIX is recruited to the virus replication sites. As ALIX has a role during virus replication, we hypothesized that the protein can be recruited to the replication sites. A549 cells infected with LGTV were immuno labeled with antibodies against either dsRNA or E protein together with ALIX. There was a strong overlap between red and green fluorophores during costaining dsRNA and ALIX, suggesting the colocalization The expression levels were normalized using the endogenous proteins, either glyceraldehyde-3-phosphate dehydrogenase (GAPDH) or b-actin. (B) Cytotoxicity assay was performed by measuring the amount of lactate dehydrogenase (LDH) enzyme released into cell culture media during siRNA treatments. When the plasma membrane is damaged, the endogenous LDH is released into media and reacts to substrates from the assay, generating a red formazan product that can be measured spectrophotometrically at 490 nm and normalized to the absorbance at 680 nm.  of the protein at the replication sites ( Fig. 4A and C). There was, however, no colocalization between the ALIX and the E proteins ( Fig. 4B and D). Unfortunately, we could not successfully label other ESCRT proteins during the staining. Thus, these data suggest that ALIX was recruited to the replication sites during LGTV infection. ESCRT proteins close localize and interact with NS3 and E. Since we demonstrated that ALIX is recruited to the replication site, we investigated interactions of ESCRT components with viral proteins. As the depletion of the ESCRT proteins produces a strong inhibitory effect on LGTV and other closely related viruses in the TBFV serocomplex (29-31), we used LGTV and TBEV Toro gene constructs as models for studying protein-protein interactions. We transfected A549 cells with the construct expressing HA epitope tag-TBEV Toro NS3 and using the construct HA-PAR6 as a control. We show that ALIX colocalizes with NS3 but not with the control (Fig. 5A and B).
We observed an L domain sequence motif LYTLA beginning at the 1795th amino acid of the polyprotein within the TBEV Toro NS3 protein, much like in the retrovirus Gag proteins (32,33), which may be the putative interaction site between NS3 and the ESCRT proteins (Fig. 5C). We transfected cells with HA-PAR6, HA-NS5 constructs as controls, HA-NS3, and the double mutants HA-NS3 Y1796L T1797L, followed by immunoprecipitations (IP) with an HA antibody to pull down the endogenous ESCRT proteins. After the IP, both ALIX and CHMP4A could be pulled down from the cell lysates expressing HA-NS3 but not from the other controls ( Fig. 5E), indicating the interaction between the ESCRT proteins and NS3, and the putative L domain motif were essential for interaction between NS3 and the ESCRT complex.
As there is colocalization between E and CHMP4A during JEV infection (23), we investigated whether there are interactions between the proteins in the context of TBEV. We transfected cells with the TBEV Toro anchored prM-E-mCherry or the mCherry construct, followed by IP with the mCherry antibody. Here, the anchored domain of the capsid protein in the anchored prM-E construct allowed the protein to locate at the ER membrane with its correct topology. During expression, the prM is cleaved from the protein by a host peptidase and results in E-mCherry (34). After the IP, we could pull down CHMP4A but not ALIX (Fig. 5F), suggesting that a direct interaction between CHMP4A and E.
Furthermore, we took advantage of the DNA TBEV Toro replicon having firefly luciferase as a reporter gene, as described previously (35). The replicon transfection can be normalized by cotransfecting the replicon with another construct expressing renilla luciferase. Thus, the replicon expression can be enumerated by the ratio of firefly luminescence to renilla luminescence, presented as relative luminescent units (RLUs). We also generated a DNA TBEV Toro replicon construct having Y1796L and T1797L mutations. Compared to the TBEV Toro replicon, there was significant reduction in the RLUs expressed by the Y1796L and T1797L mutation replicon 72 h after transfections (Fig. 5G). A negative-control  replicon with NS5 truncated (35) was used to show that the replicon expressions were not driven by the promoter of the DNA replicon construct. All together, these data suggest that the TBEV NS3 interacts with the ESCRT proteins at the putative late domain that supports virus replication and the E protein interact only with CHMP4A.
Expression of ISG15/HERC5 resulted in reduction in LGTV replication and enhancement of the ALIX-depleting effect. As TSG101 can be ISGylated during IAV infection to reduce the virus budding (20), we investigated the similar interactions between ISG15 and the ESCRT proteins during TBFV infection. A549 cells were treated with ESCRT siRNA and transfected with either GFP as a control or ISG15 construct before LGTV infection. Expression of ISG15 resulted in a 1 log reduction in LGTV titer, compared to the control (Fig. 6A). Remarkably, expression of ISG15 concomitantly with depletion of ALIX synergistically reduced the virus titers (Fig. 6A), compared to that of TSG101. In accordance with virus titers from cell culture supernatants, there was an enhanced 2 log reduction of intracellular virus genome copy number during ALIX depletion and ISG15 expression (Fig. 6B), suggesting ISG15 directly targets ALIX. Furthermore, on the replicon platform, expression of either ISG15 or HERC5 resulted in more than 1 log reduction of luciferase units (Fig. 6C), suggesting their restricting effects on virus replication. These data suggest that ISG15 and HERC5 expression reduce virus replication, and ISG15 may interplay with ALIX during virus infection.
Unconjugated ISG15 and HERC5 elicited ALIX and CHMP4A degradation. As the expression of ISG15 enhances the effects of the ESCRT protein depletion during flavivirus infection and TSG101 can be targeted by ISG15 during IAV infection (20), we investigated the effects of ISG15 on the ESCRT proteins during flavivirus infection. A549 cells were cotransfected with the siRNA targeting TSG101 and constructs expressing either ISG15, HERC5, or the control GFP before LGTV infection (Fig. 7A). As expected, expression of ISG15 or HERC5 reduced generation of virus particles compared to the control, whereas silencing of TSG101 did not alter virus production in both cell lysates and supernatants, indicated by immunoblotting of extracellular E and prM ( Fig. 7A and D). Interestingly, expression of ISG15 or HERC5 resulted in 50% degradation of TSG101 ( Fig. 7A and C) and much stronger degradations of ALIX and CHMP4A: 80% of ALIX was degraded (Fig. 7A and B), and CHMP4A was not detectable during ISG15 and HERC5 expression (Fig. 7A). However, we did not observe alterations in level of the 94-kDa glucose-regulated protein (GRP94) (Fig. 7A and E), which is an ER marker protein (36), suggesting the specific ALIX and CHMP4A degradation by ISG15 or HERC5 expression. Furthermore, we could not detect signs of ISGylation on the blotting membranes as the ISG15 was expressed without the enzymes required for the ISGylation pathway (26,27,(37)(38)(39). Altogether, these results demonstrate that expression of unconjugated ISG15 or HERC5 degraded ESCRT proteins during LGTV infection, which may explain the enhanced virus inhibition during concomitant expression of ISG15 and depletion of ALIX.
Contrary to many other enveloped viruses, where the assembly occurs at the plasma membrane, assembly in flaviviruses is coupled to replication at the ER (6). At the replication site, ALIX-CHMP4A form a complex with NS3 as NS3 can pull down the proteins (Fig. 5). A crystal structural study has revealed that the ALIX protein has two domains: Bro1 and V (40). Studies on retroviruses have shown that ALIX can mutually interact with other ESCRT proteins and the virus proteins; indeed, ALIX interacts with the L domain motif within the retrovirus p6 protein at its V domain (41) and the protein interacts with CHMP4A at its Bro1 domain (42). At the assembly site, CHMP4A is recruited to E, supporting virus assembly as we could pull down the CHMP4A by the E protein. Correlative light electron microscopy has been shown that CHMP4A colocalize at both the replication and virus assembly sites during JEV infection (23). Therefore, we propose a model for the roles of ALIX and CHMP4A at replication and assembly site, respectively, during TBFV infection (Fig. 8).
We identified the LYTLA motif within the helicase domain of the TBEV Toro NS3 protein. The sequence is similar to the L domain motif ФYXФXL recruiting ALIX described in retroviruses (Ф being any nonpolar amino acid) (19,43). The identified motif is highly conserved in the TBFV with the polar YT/S residues flanked by nonpolar residues (Fig. 5C). Based on this observation, we mutated the YT residues, which aborted the interaction between NS3 and the ESCRT complex (Fig. 5E). However, the equivalent ISTRV sequence detected within NS3 of the MBFV is less similar to the putative TBFV L domain motif in this study. Thus, MBFV may have other L motifs within NS3. Indeed, there is another proposed putative L domain motif within YFV NS3 protein supporting ALIX recruitment (Fig. 5D) (24). TSG101 assists post Golgi transporting of IAV HA to the plasma membrane (20). In our study, the effect of TSG101 seems dispensable as their depletions did not highly alter virus production. These findings on TSG101 differ from a previous study showing TSG101 is required for JEV and DENV assembly, although interestingly ALIX appeared to be dispensable (23). However, our findings are consistent with another study on YFV that showed that ALIX is essential for the virus infection (24). These findings suggest that there are variances in the use of either ALIX or TSG101, which are the two early docking proteins of the ESCRT complex to the virus proteins (18,32).
Lastly, we identified that unconjugated ISG15 and HERC5 inhibit LGTV replication, and their expression enhanced the effects of ALIX silencing (Fig. 6) due to the degradation of the ESCRT complex (Fig. 7). We hypothesize that the degradation can be dependent on lysosome as shown in a previous study (44). Furthermore, we do not eliminate the effects of ISG15 on other proteins as ISG15 has a broad spectrum of targeting proteins (45,46). In addition, the ISGylation process can be induced by the interferon response or expressing ISG15 concomitantly with the key enzymes of ISGylation steps (45,46). Although ISGylation has been shown during DENV infection (47), we could not identify any signs of ISGylation during LGTV infection within A549 cells. Thus, to our knowledge, this is the first time it has been shown that the unconjugated ISG15 and HERC5 can inhibit LGTV replication and trigger ESCRT protein degradation. Similarly, it has been shown that ISG15 or HERC5 can inhibit chikungunya (48) and HIV-1 (49), respectively, independently of ISGylation. However, our new findings with ISG15 and HERC5 do not exclude the effects of ISG15 on other stages of the flavivirus life cycle as the recombinant ISG15 has been shown to interfere with ZIKV entry (50) and DENV release (51).
In summary, we identified and characterized checkpoints such as the interaction between ESCRT complex-NS3 and the interplay between ISG15-ESCRT complex, which can function as future targets for the development of antiviral therapies. Future studies should be performed to investigate whether the virus protease NS2B/NS3 can cleave ISG15 to antagonize its antiviral function. Our study has provided insights into virushost cell interactions important for the flavivirus life cycle.
Gene constructs. The WNV Kunjin replicon was constructed based on the WNV Kunjin sequence (accession number AY274504), and the LGTV replicon was constructed based on the LGTV strain TP21 sequence (accession number NC003690). The constructs are driven by a T7 promoter for in vitro transcription to express the replicons as described in (28) (Fig. 3A).
The TBEV NS3 and NS5 (Torö-2003, AH013799) genes were amplified by PCR with primers introducing suitable endonuclease restriction sites. Full-length NS5 and NS3 were cloned into the mammalian cell expression pKH3, which were kindly provided by Ian Macara and Ben Margolis, to have a triple HA tag at the N terminus. NS3 mutations were generated by PCR of the HA-NS3 construct or the replicon construct (35) with 59 phosphorylated primers introducing mutated sites, followed by a ligation. The product was treated with restriction enzyme DpnI (Thermo Scientific) to remove the methylated plasmid PCR template.
The TBEV Toro anchored prM-E (Torö-2003, AH013799) genes were amplified by PCR with primers introducing suitable endonuclease restriction sites. Full-length anchored prM-E was cloned into the mammalian cell expression pmCherry N1 (Clontech) with the mCherry tag at the C terminus.
Renilla luciferase expressing construct (pGL4.74, Promega) was used for normalizing transfections of TBEV Toro DNA replicons. Cytotoxicity assay. Cell lysates and media were assayed using the CyQUANT LDH Cytotoxicity Assay (Invitrogen). The absorbance was read at 490 nm and normalized to the absorbance at 680 nm using the Cytation 3 Multi-Mode Reader (BioTek, Bad Friedrichshall, Germany).
Virus infection and cell harvesting. After siRNA treatments for 24 h, A549 cells were infected by TBEV Sofjin , KFDV, POWV, WNV Kunjin , ZIKV, or LGTV with 0.1 multiplicity of infection. After 24 h, supernatants were harvested. Attached cells were detached by trypsinization, followed by a soybean trypsin inhibitor treatment (Gibco). Cells were then briefly frozen using liquid nitrogen and thawed; this was done three times.
Antibodies. The following antibodies were used in this study: mouse monoclonal anti-TBEV E Plaque assay. Crystal violet-based plaque assay was performed to quantify WNV Kunjin and ZIKV, and immunofocus plaque assay was performed to quantify LGTV, TBEV Sojin , KFDV, and POWV. In brief, a series of virus dilutions in DMEM were used to infect 90% confluent Vero cells for 1 h at 37°C, followed by cell-overlaying with DMEM supplemented with 1.2% Avicel (FMC), 2% HI-FBS (Gibco), 1Â nonessential amino acids (Gibco), and 1% PEST (Gibco). After 3-5 days, the overlays were removed, and cells were fixed by methanol for 20 min before staining. For immunofocus assay, the fixed cells were blocked by 2% BSA (Fitzgerald) for 10 min before labeled with 1:1000 the mouse anti-E antibody, followed by 1:100 the anti-mouse secondary HRP Polymer for 1 h at 37°C. Finally, they were reacted with KPL TrueBlue Peroxidase Substrate (Seracare) for 15 min at room temperature (RT). For crystal violet-based plaque assay, the fixed cells were stained with 2% crystal violet (Sigma), 20% methanol (Fisher), and 0.1% ammonium oxalate (Sigma) solution.
Quantitative real-time PCR. Total RNA was isolated from cell lysates or cell culture supernatants using RNeasy minikit (Qiagen) or QIAamp Viral RNA minikit (Qiagen), respectively. cDNA was synthesized using High-Capacity cDNA Reverse Transcription kit (Applied Biosystems) with specific primers targeting the WNV KUN , ZIKV, or LGTV positive strand (Table 1). qPCR was conducted using a QuantStudio 7 Flex real-time PCR system (Applied Biosystems) with TaqMan Fast Advanced Master Mix (Applied Biosystems) and Custom TaqMan Gene Expression Assays (Applied Biosystems) containing specific primers and probes (Table 1). Virus genome copy numbers were enumerated by extrapolation from the standard curves of Ct values and known genome copy number solutions.
Luciferase assay. Cell lysates were assayed and measured for bioluminescence using Dual-Luciferase Reporter assay kit (Promega) and a Lumi-star Omega machine (BMG Labtech), according to the manufacturer's instructions.
Protein electrophoresis and immunoblotting. Cells were lyzed with 0.5% SDS before boiling in LDS sample buffer (Invitrogen). Proteins were separated on precast 4%-12% polyacrylamide Bis-Tris gels (Invitrogen) in MES running buffer (Invitrogen) for 1 h at 120 V constant before transferred to nitrocellulose membranes using an iBlot 2 Gel Transfer device (Invitrogen). Proteins of interest were detected with the antibodies.
Immunoprecipitation. For pull-down studies, A549 cells were transfected with individual constructs using a Nucleofector 2b device (Lonza) and Cell Line Nucleofector kit T (Lonza). Two days posttransfection, the cells were lyzed by radioimmunoprecipitation assay (RIPA) buffer (Sigma) supplemented with a protease inhibitor cocktail (Sigma) for 20 min at 4°C. Lysates were precleared by centrifuge at 1000g for 10 min, and 300 mg of the precleared lysates were incubated with 4 mg of antibodies and 50 ml of a protein A/G-tagged microbead premix (MACS) at 4°C for 1 h. The mixtures were then loaded on m Columns (MACS) holding a mMACS Separator to retain the microbeads on the columns. The beads were washed four times with 4°C 50% RIPA buffer. Finally, bead-binding proteins were eluted by a preheated 95°C LDS sample buffer.
Immunofluorescence labeling. Cells grown on coverslips were fixed by 4% paraformaldehyde (Scharlau, Spain) for 20 min. They were then permeabilized by 0.1% Triton X-100 (VWR) before being blocked with 2% bovine serum albumin (BSA) (Fitzgerald) and 2% goat serum (Invitrogen) for 1 h. Cells were labeled with the primary antibodies for 2 h at 37°C, followed by the secondary antibodies Alexa Fluor 488 or Alexa Fluor 594. The cells were counterstained with DAPI (Sigma) for 5 min before mounting with ProLong Gold Antifade Mountant (Life Technologies). Images were captured using a confocal laser scanning microscopy SP8 (Leica) and analyzed using ImageJ.
Statistical analysis. Statistical differences between the means were determined using one-way ANOVA followed by the Bonferroni post hoc test, and P , 0.05 was considered statistically significant difference. GraphPad Prism 9 was used for performing all statistical analyses. The values are presented as mean 6 standard error of the mean.
Data availability. The data presented in this study are available in this article. Remaining data supporting reported results are available from the corresponding authors upon reasonable requests.