**3. Results**

#### *3.1. WNV and ZIKV Broadly Inhibited JAK*/*STAT Signaling Following Cytokine Stimulation of Target Cells*

To determine the breadth of flavivirus-mediated inhibition of JAK/STAT signaling, we first assessed acute responses to JAK/STAT-dependent cytokine treatment in cells infected with WNV or ZIKV. We conducted immunoblot analyses of cytokine-induced pY-STAT responses to IFNβ for all six STAT family members in WNV- and ZIKV-infected A549 (human lung epithelial) cells (Figure 1A,B). Mock infected cells responded to IFNβ with pY of all STATs by 30 min post-treatment. In contrast, WNVand ZIKV-infected cells were unable to mediate pY-STATs in response to high dose IFNβ (Figure 1A,B). Immunofluorescence revealed nuclear pY-STAT1 did not accumulate in response to IFNγ in WNV- and ZIKV-infected A549 cells, though pY-STAT1 did accumulate in adjacent non-infected bystander cells (Figure 1C,D). The attenuated response to IFNγ-induced pY-STAT1 was confirmed by immunoblot of WNV-infected cells (Figure S1A). WNV-infected cells also showed inhibited pY-STAT1 and 2 in response to IFNλ3 (Figure 1E). Moreover, WNV and ZIKV infection each led to inhibition of pY-STAT1 and 3 in response to human and mouse IL6 (Figure 1F,G and Figure S1B,C), and of pY-STAT6 following IL4 treatment (Figure 1H,I). Immunofluorescence also showed that WNV-infected THP-1 macrophages failed to accumulate pY-STAT1 and 3 in response to IFNβ and IL10, respectively (Figure S1D,E). Thus, WNV and ZIKV impose a broad blockade to pY-STAT mediated by several unrelated cytokines, implying one or more common features of JAK/STAT signaling could be dysregulated during flavivirus infection. Of note, background levels of pY-STAT1 and total levels of STAT1 and STAT2 were observed as being increased following virus infection in several analyses (Figure 1A,B,E,F and Figure S1A). These observations were fully expected in IFN-competent A549 cells, as IFN signaling is only blocked by flaviviruses after approximately 20–24 hpi [17]. Early antiviral signaling would then be expected induce some pY-STAT1 and ISGs (including total STAT1 and STAT2); importantly, however, neither of these were further increased upon exogenous cytokine stimulation, indicative of the virus-imposed JAK/STAT inhibition by 24hpi.

**Figure 1.** West Nile virus (WNV) and *Zika virus* (ZIKV) infections inhibit pY-STAT and gene expression in response to multiple cytokines. pY-STAT responses were inhibited in WNV- and ZIKV-infected A549 cells at 30 min post-treatment with (**A**,**B**) inferno (IFN)β, (**C**,**D**) 10 ng/mL IFNγ (arrow heads), (**E**) IFNλ3, (**F**,**G**) interleukin (IL)6, and (**H**,**I**) IL4. WNV-infected A549 cells did not respond to 17 h treatment with (**J**) IFNβ (*IFITM1* and *IFIT1* induction) or (**K**) IL6 (γ*Fibrinogen* and *IGFBP1* induction). (**L**) WNV-infected PH5CH8 IFNAR1−/− cells did not respond to 24 h IFNλ3 treatment via expression of the interferon stimulated genes MX1, IFIT1, OAS1, and ISG15. All infections were MOI = 5 for 24 h prior to cytokine treatment unless otherwise stated. All data represent three independent experiments or are a combination of three experiments (mean ± standard error of the mean (SEM)). See also Figure S1. \*\*\*\* = *p* < 0.0001, \*\*\* = *p* < 0.001, ns = *p* > 0.05. Two-way ANOVA with Tukey's posttests.

This blockade to pY-STAT corresponded to inhibition of cytokine-responsive gene induction. Analysis of endogenous interferon-induced protein with tetratricopeptide repeats 1 (*IFIT1*) and

interferon-induced transmembrane protein 1 (*IFITM1*) expression in cells responding to IFNβ (Figure 1J) showed that WNV blocked induction of these ISGs in response to the cytokine (virus-induced ISG mRNA remaining from early antiviral signaling prior to establishment of JAK/STAT inhibition was not further increased). Additionally, WNV blocked induction of γ*Fibrinogen* and insulin-like growth factor-binding protein 1 (*IGFBP1*) in response to IL6 (Figure 1K). To assess the response of WNV-infected cells to IFNλ3, we treated human epithelial cells (hepatocytes) specifically lacking functional IFNα/β receptor expression (PH5CH8 IFNAR1-/- cells) and measured expression of ISG proteins MX dynamin like GTPase 1 (MX1), IFIT1, 2´5´-oligoadenylate synthetase 1 (OAS1), and ISG15. WNV infection abrogated induction of each ISG by IFNλ3 (Figure 1L).

Importantly, NF-κB signaling in response to IL1β treatment in WNV-infected A549 cells was intact, with comparable cytokine-responsive inhibitor of κB alpha (IκBα) degradation and pS-p65 induction (Figure S1F). Upstream signaling components in this non-JAK/STAT pathway were also largely unaffected by infection (Figure S1G).

#### *3.2. Cell-Intrinsic WNV NS Proteins mediated Broad JAK*/*STAT Inhibition*

To define which factors produced during WNV infection contributed to JAK/STAT signaling inhibition, we examined the virus replication cycle using two complementary approaches (Figure 2A–F). Huh7 hepatoma cells stably harboring aWNV replicon [31] (Figure 2A) were used to assess the JAK/STAT pathway in the context of NS protein expression and viral RNA replication, but in the absence of structural proteins. Treatment with IFNλ3 (Figure 2B) showed pY-STAT1 was reduced in WNV replicon cells compared with cells cured of replicon. Similarly, pY-STAT3 abundance was diminished in IL6-treated WNV replicon cells compared to cured controls (Figure 2C).

To assess virion impact on JAK/STAT signaling, we exposed cells to UV-inactivated WNV present in conditioned culture supernatant (Figure 2D). This approach allowed us to assess the JAK/STAT inhibitory role of viral factors such as virions, subviral particles, and NS1 protein released from infected cells [32], as well as host cytokines and metabolites produced in response to infection. Cells were first exposed to UV-inactivated WNV-conditioned media and then treated with either IL6 (Figure 2E) or IL4 (Figure 2F). Immunoblot analysis showed that cytokine treatment induced pY-STAT3 and 6 to levels comparable with control cells treated with conditioned media from mock infections. By comparison, cells exposed to supernatants containing live WNV exhibited a block of cytokine-induced pY-STAT (Figure 2E,F). Thus, extracellular factors secreted from WNV-infected cells are dispensable for the JAK/STAT signaling blockade, whereas one or more viral NS proteins are likely responsible for the broad block to pY-STAT.

To identify viral NS protein(s) responsible for JAK/STAT signaling blockade, we independently expressed each NS protein in cells and assessed their capacity to inhibit pY-STAT3 in response to IL6 (Figure 2G,I) and pY-STAT1 in response to IFNγ (Figure 2H,J) by immunofluorescence assay. As controls, we also assessed pY-STAT in cells expressing WNV capsid protein or vector alone. Expression of viral proteins themselves did not induce pY-STAT in resting cells without cytokine treatment (Figure S2). We found that among WNV NS proteins, expression of NS5 caused significant inhibition of pY-STAT accumulation in cells treated with either cytokine. Expression of NS2A, NS2B, and NS2B/3 showed significant inhibition of pY-STAT1 following treatment with IFNγ only. Expression of any individual NS protein was unable to block JAK/STAT signaling as efficiently as live WNV, suggesting viral NS proteins likely act in a complementary manner to confer broad JAK/STAT antagonism.

**Figure 2.** WNV nonstructural proteins were found to be responsible for Janus kinase (JAK)/STAT signaling inhibition intrinsically within infected cells. (**A**) Huh7 WNV replicon cells showed low pY-STAT responses to (**B**) IFNλ3 and (**C**) IL6 compared to cured controls. (**D**) Methodology to generate infectious and non-infectious WNV-conditioned media. Blockade of pY-STAT in response to (**E**) IL6 and (**F**) IL4 was only observed with infectious conditioned media. Recombinant expression of WNV C and nonstructural (NS) proteins in A549 cells showed differential inhibition of pY-STAT responses to (**G**,**I**) 200 ng/mL IL6 and (**H**,**J**) 10 ng/mL IFNγ (arrow heads), with WNV NS5 significantly responsible

for JAK/STAT inhibition across both cytokine treatments. All cytokine treatments were for 30 min. All data represent three independent experiments or are a combination of three experiments (mean ± SEM) except "Vector" in (**I**), and "Mock" and "Vector" in (**J**) (*n* = 2). See also Figure S2. \*\*\*\* = *p* < 0.0001, \*\*\* = *p* < 0.001, \*\* = *p* < 0.01, \* = *p* < 0.05, ns = *p* > 0.05. One-way ANOVA with Tukey's posttests. ‡ "Counted group" = total cell population in several fields per experiment as measured by 4',6-diamidino-2-phenylindole dihydrochloride (DAPI) for "Mock" and "Vector", all WNV E-positive cells for "WNV", and all hemagglutinin epitope tag (HA)-positive cells for WNV proteins. The red pY-STAT3 channel for images in ( **G**) was enhanced equally post-acquisition across all samples to aid visual clarity.

#### *3.3. Flavivirus Infection Lead to JAK and Cytokine Receptor Degradation by the Proteasome*

Considering the fact that the broad inhibition of pY-STAT responses and gene induction by WNV and ZIKV appeared to be unrelated to the identity of the cytokine used to stimulate cells (Figure 1), we sought to determine whether infection with these viruses a ffected JAK proteins proximal to these events in the JAK/STAT signaling cascade. Immunoblot of WNV-infected A549 cells treated under control conditions (DMSO) demonstrated that at 24 hpi, all JAK family members (JAK1-3 and Tyk2) had reduced protein abundance compared to mock-infected cells. This loss of abundance was partially rescued by treatment with the proteasome inhibitor MG-132 for 6 h, starting at 18 hpi. MG132 led to increased JAK family protein abundance in WNV-infected cells to levels similar to mock, DMSO-treated cells (Figure 3A,B). Time course analyses demonstrated that loss of JAK1 abundance could be observed from approximately 16 hpi (Figure S1H). To assess whether JAK loss was in part due to RNA regulation, we quantified *JAK1*, *JAK2*, *JAK3*, and *Tyk2* transcripts within mock- and WNV-infected cells at 24 hpi (Figure 3C). WNV infection led to induction of JAK mRNA, and thus loss of JAK abundance is likely mediated solely by post-transcriptional processes.

To determine if loss of JAK family protein abundance was a phenotype extending beyond the virulent WNV strain we predominantly used, we evaluated the abundance of JAK1 and Tyk2 in A549 cells infected with a panel of flaviviruses, as well as control viruses including Sendai virus (SeV) and encephalomyocarditis virus (EMCV), at 36 hpi (Figure 3D,E). Compared to mock-infected cells, infection with WNV-TX, WNV-MAD, ZIKV MR766, JEV, and DENV-2 all showed a trend of variably decreased JAK1 and Tyk2 abundance (Figure 3D). Quantification of band intensity across three independent experiments demonstrated that significant decreases in both JAK1 and Tyk2 occurred during WNV-TX and ZIKV MR766 infection (Figure 3E), whereas only decrease in Tyk2 abundance was significant for JEV. Despite the trend towards decreased abundance of these JAK family proteins, quantification from WNV-MAD and DENV-2 infection (both viruses that replicate less robustly in A549 cells) demonstrated that these decreases were not significant. EMCV infection (which shuts down host translation) [33] led to significant loss of JAK1 and Tyk2. In contrast, SeV infection displayed only slight alteration in abundance of these proteins. Therefore, loss of JAK protein abundance is a feature common to several flavivirus infections.

**Figure 3.** Flavivirus infection led to loss of abundance of heat shock protein 90 (HSP90) client kinases including JAK proteins, with interaction between HSP90 and JAK1 disrupted. (**A**,**B**) JAK family proteins are decreased in abundance during WNV infection. Treatment with MG-132 at 18 h post-infection (hpi) partially recovered JAK abundance in WNV-infected cells. (**C**) Transcripts of JAK1, JAK2, JAK3, and Tyk2 increased during WNV infection. (**D**,**E**) Infection of A549 cells with WNV-TX, ZIKV MR766, Japanese encephalitis virus (JEV), and encephalomyocarditis virus (EMCV) led to loss of Tyk2 and (for all except JEV) JAK1 by 36 hpi. Infection with less robustly infecting flaviviruses WNV-MAD and dengue virus (DENV)-2, as well as with Sendai virus (SeV), did not lead to significant protein loss. (**F**) Treatment of A549 cells with HSP90 inhibitor geldanamycin (GA) for 6 and 12 h diminished abundance of JAK1, Tyk2, erythroblastosis oncogene B2 (ErbB2), and AKR mouse strain thymoma-related protein (Akt). (**G**) HSP90 client kinases ErbB2 and Akt displayed decreased abundance in WNV-infected A549 cells in a similar manner to JAK1 and Tyk2 by 24 hpi. (**H**–**J**) HEK-293T cells mock- or WNV-infected (MOI = 5, 10, and 20) were co-transfected with FLAG-epitope tagged JAK1 (FLAG-JAK1) and HSP90-HA. (**H**,**I**) Transfection of JAK1 was sufficient for pY-STAT1 and three responses, and this pY-STAT was significantly blocked with increasing WNV. (**J**) Reciprocal HA- and FLAG-tagged co-immunoprecipitation (co-IP) showed significantly decreased interaction of co-precipitated HSP90 and JAK1 with increasing WNV. All infections were in A549 cells at MOI = 5 for 24 h prior to cytokine treatment unless stated, except SeV which was infected at 40 hemagglutination units per well. All data represent three independent experiments or are a combination of three experiments (mean ± SEM), except (**G**) (*n* = 2). \*\*\*\* = *p* < 0.0001, \*\*\* = *p* < 0.001, \*\* = *p* < 0.01, \* = *p* < 0.05, ns = *p* > 0.05. Paired *t*-tests (**C**). One-way ANOVA with Dunnett's Multiple Comparison post-test (**I**, **J**). Irrelevant lanes were cropped between controls and infected samples in (**H**).

#### *3.4. NS5 Disrupted HSP90-Client Kinase Interaction to Block JAK*/*STAT Signaling*

To identify the mechanism WNV NS5 uses to drive JAK/STAT inhibition, we evaluated cellular processes known to regulate JAK stability and cytokine-responsive pY-STAT. As HSP90 is required to mediate the correct kinase-active conformational folding and stability of JAK proteins (and other specific client kinases) [34,35], we reasoned that viral dysregulation of HSP90 might confer reduced abundance of JAKs and other HSP90 clients within infected cells. Indeed, we found that chemical inhibition of HSP90 activity with geldanamycin (GA) led to loss in abundance of JAK1 and Tyk2, as well as other HSP90 client kinases Akt and ErbB2 at 6 and 12 h post-treatment (Figure 3F). Importantly, we were able to see an analogous loss in abundance in all four of these HSP90 client kinases at 24 hpi with WNV (Figure 3G), directly phenocopying chemical HSP90 inhibition. Thus, viral dysregulation of HSP90 might be responsible for the broad block to pY-STAT during WNV infection.

We directly assessed the effect of WNV infection on the interaction of HSP90 with JAK1 using overexpression of epitope-tagged proteins and co-immunoprecipitation (co-IP) (Figure 3H,J) from virus infected cells. HEK-293T cells were mock- or WNV-infected at MOI = 5, 10, and 20. At 2 hpi, cells were co-transfected with plasmids encoding FLAG-JAK1 and HA-HSP90α (with vector controls). Immunoblot of whole cell lysate (Figure 3H) showed that transfection of FLAG-JAK1 was sufficient to drive ligand- and receptor-independent pY-STAT1 and 3 by 24 hpi. Increased WNV MOI led to significantly decreased levels of pY-STAT (Figure 3I), without loss in abundance of ectopic, overexpressed JAK1. These results demonstrated a fundamental point—degradation of JAK1 is not required per se for pY-STAT inhibition. Degradation presumably occurs subsequent to the mechanism rendering kinase activity of JAKs defective.

Reciprocal co-IP of JAK1 and HSP90 (Figure 3J) revealed the mechanism leading to kinase-inactive JAK proteins. Increasing infection with WNV led to significantly decreased interaction between JAK1 and HSP90. Presumably, this diminished interaction prevented HSP90 from properly chaperoning JAK1 to promote an active conformational fold. Importantly, we found this decline in JAK1-HSP90 interaction occurred concomitant with increasing interaction between HSP90 and viral NS5 (Figure 3J). Thus, the precise mechanism of pY-STAT inhibition involves WNV NS5 targeting of HSP90 to disrupt its interaction with JAKs, preventing the JAK activity that drives pY-STAT (Figure 3H,I), and leading eventually to JAK proteasomal degradation (see Figure 3A,B,D,G).

#### *3.5. HSP90 Antagonism Was Found to be Linked to Interaction with Viral NS5 at Sites of RNA Replication*

To examine the context of virus-induced changes to HSP90 function, we examined localization of this chaperone within flavivirus-infected A549 cells at 24 hpi. Immunofluorescence showed HSP90 co-localized with or is adjacent to double-stranded RNA (dsRNA) within replication complexes of WNV-TX, WNV-MAD, JEV, ZIKV MR766, ZIKV Fortaleza, DENV-2, and DENV-4 (Figure 4A). High-resolution imaging showed both isoforms HSP90α and HSP90β co-localized with WNV NS5 and dsRNA (Figure 4B,C and Figure S3A,B).

We conducted analyses to interrogate the interaction of endogenous HSP90 and NS5 in WNVand ZIKV-infected cells using reversible cross-linked IP (ReCLIP). Endogenous HSP90 bound to virus-produced WNV NS5 in reciprocal co-IP analyses (Figure 4D). Similarly, ZIKV NS5 was found in complex with endogenous HSP90 in infected cells (Figure 4E). The HSP90–NS5 interaction was specifically recovered by anti-NS5 or anti-HSP90, as the NS5-HSP90 complex was not recovered using isotype control IgG2a antibody. We confirmed specific interaction of ZIKV NS5 with HSP90 via further analysis of FLAG-tagged ZIKV NS5 in HEK-293T cells and co-IP of HSP90 (Figure S3C). ZIKV NS5 could co-IP endogenous HSP90, however, neither vector nor ZIKV capsid protein conferred HSP90 interaction.

**Figure 4.** Flavivirus NS5 interacted with HSP90 at sites of viral RNA replication. (**A**) HSP90 co-localized with double-stranded RNA (dsRNA) at 24 hpi with WNV-TX, WNV-MAD, JEV, ZIKV MR766, ZIKV Fortaleza, DENV-2, and DENV-4 (arrow heads). WNV NS5 co-localized with both (**B**) HSP90α and (**C**) HSP90β and dsRNA in infected A549 cells (arrow heads). (**D**) ReCLIP experiments with WNV-infected A549 cells showed NS5 co-precipitation with HSP90, and HSP90 was recovered following NS5 IP. Controls showed NS3 was detected following NS5 IP (replication complexes), and neither glyceraldehyde 3-phosphate dehydrogenase (GAPDH) nor Actin were co-precipitated under any condition. The IgG2a isotype control failed to IP any analyzed protein. (**E**) ReCLIP analysis of ZIKV-infected cells showed an analogous NS5–HSP90 interaction. All infections were at MOI = 5 for 24 h (except DENV-4 where MOI = 3). All data represent three independent experiments, except (E) (*n* = 2).

#### *3.6. HSP90 Activity Was not Required to Support Flavivirus RNA Replication nor to Stabilize Viral Proteins*

We considered the possibility that flaviviruses may usurp HSP90 to promote viral RNA replication and viral protein function, as has been reported for flavivirus interactions with HSP70 [26–28]. To address this notion, we assessed the effect of chemical HSP90 inhibition upon the abundance of viral RNA and proteins, and on dsRNA NS protein localization within Huh7 WNV replicon cells. Compared to DMSO-treated control cells, WNV replicon cells 24 hpt with HSP90 inhibitors GA and EC144 showed stable co-localization of NS3 and NS5 by immunofluorescence analysis (Figure S4A). However, upon HSP90 inhibition in treated cells, NS1 and dsRNA shifted from co-localization with the NS proteins to distribute as bright foci directly adjacent to NS3 and NS5 (Figure S4A,B). Changes in abundance of replicon antigens was undetectable under conditions of HSP90 inhibition in treated cells. In contrast, a decrease in intensity was observed for controls treated with the flavivirus replication inhibitor NITD008. Analysis of replicon genomic RNA (WNrep gRNA; Figure 5A) showed HSP90 inhibition via EC144 treatment significantly increased gRNA abundance by 24 h. In contrast, NITD008 treatment led to decreased gRNA. Analysis of HSP70A mRNA showed significant induction of this chaperone upon HSP90 inhibition as expected [36], but not following NITD008 treatment.

**Figure 5.** Flavivirus RNA replication and protein stability did not require HSP90 activity. (**A**) Huh7 WNV replicon cells were treated with HSP90 inhibitors GA and EC144, or nucleoside analogue NITD008 for 12 and 24 h to assess effects on RNA replication. Analysis of WNV replicon RNA (WNrep genomic RNA (gRNA)) revealed increased abundance following HSP90 inhibitor treatment, in contrast to reduced gRNA following NITD008. This was coincident with significant induction of HSP70A mRNA following HSP90 inhibitor treatment. Vero cells 12 hpi with (**B**,**C**) WNV or (**D**,**E**) ZIKV at MOI = 5 were treated with DMSO or HSP90 inhibitors for a further 12 h. (**B**,**D**) HSP90 inhibitors reduced Akt and ErbB2 abundance, and changes to viral proteins were not detected. Quantities of (**C**) WNV E, NS1, and NS5, as well as (**E**) ZIKV E were moderately ye<sup>t</sup> significantly increased upon HSP90 inhibition. All data represent three independent experiments or are a combination of three experiments (mean ± SEM), with the exception of HSP70A mRNA quantification in (**C**) (*n* = 2). \*\*\*\* = *p* < 0.0001, \*\*\* = *p* < 0.001, \*\* = *p* < 0.01, \* = *p* < 0.05, ns = *p* > 0.05. Two-way repeated measures ANOVA with Holm–Sidak's post-tests (**A**), and one-way ANOVA with Dunnett's multiple comparison post-tests (**C**,**E**).

We also measured the abundance of WNV replicon NS proteins at 12 and 24 hpt by immunoblot analysis (Figure S4C,D). Though reduced abundance of Akt and ErbB2 were found upon GA and EC144 treatment, neither inhibitor caused a decrease in replicon NS proteins. This contrasts with significant 30–50% reduction of NS proteins with NITD008. We also examined the impact of HSP90

inhibition upon protein abundance during live WNV and ZIKV infection. Vero cells 12 hpi were treated with DMSO, GA, or EC144 for a further 12 h. For both viruses, HSP90 inhibitors decreased Akt and ErbB2 abundance, though little change to viral proteins was detected (Figure 5B,D). Quantification of band intensity showed that WNV and ZIKV protein levels did not decrease during HSP90 inhibition (Figure 5C,E). Indeed, WNV E, NS1, NS5, and ZIKV E were significantly increased upon HSP90 inhibitor treatment, as similarly shown for DENV infection [37]. Taken together, these observations indicate that flavivirus replication is not dependent upon HSP90 activity.
