NS5 Sumoylation Directs Nuclear Responses That Permit Zika Virus To Persistently Infect Human Brain Microvascular Endothelial Cells

ZIKV is a unique neurovirulent flavivirus that persistently infects human brain microvascular endothelial cells (hBMECs), the primary barrier that restricts viral access to neuronal compartments. Here, we demonstrate that flavivirus-specific SIM and SUMO sites determine the assembly of NS5 proteins into discrete nuclear bodies (NBs). We found that NS5 SIM sites are required for NS5 nuclear localization and that SUMO sites regulate NS5 NB complex constituents, assembly, and function. We reveal that ZIKV NS5 SUMO sites direct NS5 binding to STAT2, disrupt the formation of antiviral PML-STAT2 NBs, and direct PML degradation. ZIKV NS5 SUMO sites also transcriptionally regulate cell cycle and ISG responses that permit ZIKV to persistently infect hBMECs. Our findings demonstrate the function of SUMO sites in ZIKV NS5 NB formation and their importance in regulating nuclear responses that permit ZIKV to persistently infect hBMECs and thereby gain access to neurons.

ABSTRACT

Zika virus (ZIKV) is cytopathic to neurons and persistently infects brain microvascular endothelial cells (hBMECs), which normally restrict viral access to neurons. Despite replicating in the cytoplasm, ZIKV and Dengue virus (DENV) polymerases, NS5 proteins, are predominantly trafficked to the nucleus. We found that a SUMO interaction motif in ZIKV and DENV NS5 proteins directs nuclear localization. However, ZIKV NS5 formed discrete punctate nuclear bodies (NBs), while DENV NS5 was uniformly dispersed in the nucleoplasm. Yet, mutating one DENV NS5 SUMO site (K546R) localized the NS5 mutant to discrete NBs, and NBs formed by the ZIKV NS5 SUMO mutant (K252R) were restructured into discrete protein complexes. In hBMECs, NBs formed by STAT2 and promyelocytic leukemia (PML) protein are present constitutively and enhance innate immunity. During ZIKV infection or NS5 expression, we found that ZIKV NS5 evicts PML from STAT2 NBs, forming NS5/STAT2 NBs that dramatically reduce PML expression in hBMECs and inhibit the transcription of interferon-stimulated genes (ISG). Expressing the ZIKV NS5 SUMO site mutant (K252R) resulted in NS5/STAT2/PML NBs that failed to degrade PML, reduce STAT2 expression, or inhibit ISG induction. Additionally, the K252 SUMOylation site and NS5 nuclear localization were required for ZIKV NS5 to regulate hBMEC cell cycle transcriptional responses. Our data reveal NS5 SUMO motifs as novel NB coordinating factors that distinguish flavivirus NS5 proteins. These findings establish SUMOylation of ZIKV NS5 as critical in the regulation of antiviral ISG and cell cycle responses that permit ZIKV to persistently infect hBMECs.

IMPORTANCE ZIKV is a unique neurovirulent flavivirus that persistently infects human brain microvascular endothelial cells (hBMECs), the primary barrier that restricts viral access to neuronal compartments. Here, we demonstrate that flavivirus-specific SIM and SUMO sites determine the assembly of NS5 proteins into discrete nuclear bodies (NBs). We found that NS5 SIM sites are required for NS5 nuclear localization and that SUMO sites regulate NS5 NB complex constituents, assembly, and function. We reveal that ZIKV NS5 SUMO sites direct NS5 binding to STAT2, disrupt the formation of antiviral PML-STAT2 NBs, and direct PML degradation. ZIKV NS5 SUMO sites also transcriptionally regulate cell cycle and ISG responses that permit ZIKV to persistently infect hBMECs. Our findings demonstrate the function of SUMO sites in ZIKV NS5 NB formation and their importance in regulating nuclear responses that permit ZIKV to persistently infect hBMECs and thereby gain access to neurons.

INTRODUCTION

Zika virus (ZIKV) is a mosquito-borne flavivirus that has emerged in the Americas as the cause of encephalitis and fetal microcephaly. ZIKV uniquely persists in human bodily fluids for up to 6 months, is sexually transmitted, and traverses placental and blood-brain barriers (BBBs) to damage neurons (1–3). In contrast, other mosquito-borne flaviviruses primarily cause acute febrile infections that are transient and resolve in ∼2 weeks, with ∼80% of flavivirus infections being asymptomatic (4, 5).

The endothelial cell (EC) lining of capillaries is the primary component of the BBB that normally restricts neuronal exposure to blood constituents, immune cells, and viruses (6). We previously observed that ZIKV (PRVABC59) persistently infects and continuously replicates in primary human brain microvascular ECs (hBMECs) without cytopathology and through cellular passage (7). In contrast, ZIKV lytically infects neuronal progenitors, trophoblasts, and other cell types (1), suggesting that ZIKV uniquely regulates cytotoxic responses, permitting hBMECs to serve as a ZIKV reservoir fostering neuronal delivery and systemic spread. ZIKV constitutively induces and evades antiviral interferon (IFN) and IFN-stimulated gene (ISG) responses to continuously replicate in hBMECs, although the mechanisms by which ZIKV balances persistent viral replication and hBMEC survival remains to be determined (7).

Flaviviruses are positive-strand RNA viruses (∼11 kb) that encode a single polyprotein precursor that is cotranslationally cleaved into three structural proteins and seven nonstructural (NS) proteins (8). Although flaviviruses replicate in ER-localized factories in the cytoplasm (9), NS5 proteins from Dengue virus (DENV), yellow fever virus (YFV), and ZIKV are predominantly found in the nucleus of infected cells (10–13). Flavivirus NS5 proteins include ∼900 residues with 2 enzymatic domains, an N-terminal methyltransferase (MT) domain, and a C-terminal RNA-dependent RNA polymerase (RdRp) domain (13). The significance of NS5 localization in the nucleus and functional roles of nuclear NS5 in cellular regulation and viral pathogenesis remain to be elucidated.

Flavivirus NS5 proteins reportedly regulate innate immunity through several mechanisms, including pathway-specific control of IFN induction and the regulation of STAT2 transcriptional responses downstream of IFN-α/β receptors (IFNARs) (13). DENV NS5 reportedly degrades STAT2 (14, 15), and NS5 is stabilized by posttranslational addition of SUMO (small ubiquitin-like modifiers) to site-specific lysine residues (16). SUMOylation is associated with functions and complexes formed by nuclear proteins, yet the NS5 nuclear localization signal (NLS) is suggested to be a structural element only required for RdRp activity and not required for IFN regulation (17). Knockout of the SUMO ligase, UBC9, or mutation of the NS5 NLS causes little change to DENV titers, suggesting that nuclear localization is not required for efficient DENV replication (16, 17). The MTase domain of NS5 also contains a conserved SUMO interacting motif (SIM) that is reportedly required for SUMOylation of DENV NS5 (16), although roles of the SIM and SUMO sites in nuclear localization, antiviral responses, and functional regulation of cellular responses have not been assessed.

Posttranslational modifications dynamically modulate cellular functions and regulate host cell innate immune responses following viral infections. SUMOylation is the posttranslational addition of the ∼11-kDa peptide SUMO to lysine residues. SUMOylated proteins are predominantly found on proteins in the nucleus (18), where they direct protein-protein interactions that regulate cellular gene expression, DNA repair, cell cycle progression, chromatin remodeling, and innate immune responses (19, 20). To replicate successfully, DNA viruses actively regulate host cell SUMOylation, suppressing and interrupting interactions of promyelocytic leukemia (PML) protein, a SUMOylated IFN-enhancing antiviral protein (21, 22). Influenza A virus (IAV) is an RNA virus that reprograms cellular SUMOylation responses; however, like DNA viruses, IAV replicates in the nucleus and is fully dependent on cellular splicing and nuclear export of viral transcripts (23). Regulation of SUMOylation by cytoplasmic RNA viruses remains largely unstudied as a host cell determinant of permissive viral replication. Although ZIKV replicates in the cytoplasm, there is little understanding of how NS5 SUMOylation, nuclear trafficking, or nuclear body (NB) formation regulates cellular responses or affects cell type-specific programs that permit ZIKV to persist in hBMECs.

ZIKV (MR766 and H/PF/2013) NS5s were recently found in nuclear bodies of Huh7 cells and neuronal LN-229 cells that are lytically infected, and one report suggests that NS5 sequesters importin-alpha and modulates neuronal LN-229 cell immune responses (24). In contrast to these cells that are lytically infected by ZIKV, we found that ZIKV (PRV) persistently infects primary hBMECs that form the BBB, a novel cell type that normally restricts viral access to neuronal tissues (7). ZIKV persistence in hBMECs suggests the potential for nuclear NS5 trafficking to uniquely regulate cell survival and IFN responses required for ZIKV to persist in hBMECs.

Here, we examined roles for nuclear NS5 in ZIKV-infected or NS5-expressing hBMECs. We found that in ZIKV (PRV)-infected hBMECs, NS5 proteins are predominantly trafficked to the nucleus, where they form distinct punctate nuclear speckles that increase in size from 18 to 24 h postinfection (hpi). In contrast to ZIKV NS5, expressing DENV3 NS5 in hBMECs resulted in NS5 uniformly dispersed in the nucleoplasm. We unexpectedly found that a conserved SIM site is required for ZIKV and DENV3 NS5 nuclear localization. In hBMECs, ZIKV NS5 colocalizes with SUMO-1 and STAT2 NBs, disrupts SUMO-1 and STAT2 colocalization with PML, and results in PML degradation. Mutating the ZIKV NS5-K252R SUMO motif caused NS5 to be structurally reformed and present in complexes with both STAT2 and PML. In contrast to ZIKV, the dispersed DENV NS5 localization changed to discrete punctate NBs when we mutated the DENV NS5-K546R SUMO site. These findings link site-specific SUMOylation and SIM motifs in DENV and ZIKV NS5 proteins to the formation of novel NBs.

We found that SUMOylation of ZIKV NS5 was SIM site independent, but that ZIKV NS5 SUMO-K252R was required for inhibiting ISG and cell cycle-associated transcript levels in hBMECs. The SUMO-K252R site was required for ZIKV NS5 to efficiently coprecipitate STAT2, dissociate PML-STAT2 nuclear complexes, inhibit ISG induction, and reduce PML expression levels. Failure of the NS5 NLS mutant to coimmunoprecipitate STAT2 further suggests that in hBMECs ZIKV NS5 uniquely binds and regulates STAT2 in the nucleus. Our findings reveal that nuclear ZIKV NS5 and site-specific NS5 SUMOylation control cell cycle-regulating genes and STAT2/PML antiviral innate immune responses that permit ZIKV to uniquely persist and spread in hBMECs.

RESULTS

ZIKV NS5 forms discrete punctate nuclear speckles in primary hBMECs.

Flaviviruses replicate in the cytoplasm of infected cells, yet the viral polymerase, NS5, is trafficked to the nucleus (13, 25). Roles for nuclear NS5 in regulating cellular functions remain enigmatic but potentially distinguish flavivirus pathogenesis. Here, we assess nuclear roles for ZIKV NS5 in regulating responses that permit ZIKV to persistently infect primary human brain endothelial cells (7). In an initial kinetic analysis of ZIKV (PRV)-infected hBMECs, we found that NS5 was highly expressed and nearly completely localized to the nucleus of cells 12 to 24 h postinfection (hpi), despite the envelope (Env) protein being confined to the cytoplasm (Fig. 1). NS5 formed discrete punctate nuclear speckles that increased in size from 18 to 24 hpi, suggesting a fundamental role for NS5 in regulating nuclear hBMEC functions.

FIG 1.

NS5 forms discrete punctate nuclear speckles in ZIKV-infected primary hBMECs. hBMECs grown on microslides were infected with ZIKV-PRV (MOI, 2). Cells were fixed 12, 18, or 24 h postinfection (hpi), immunostained with 4G2 antibody against Envelope protein (Env) or anti-NS5 antibody, DAPI counterstained, and visualized by confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 10 μm.

ZIKV NS5 forms unique nuclear speckles that disrupt PML/SUMO-1 NBs.

Nuclear protein interactions and cellular responses regulated by ZIKV NS5 have yet to be defined. Speckles or nuclear bodies (NBs) are subcompartments within interchromatin regions of the nucleoplasm, e.g., nucleoli, Cajal bodies, SUMO NBs, and PML NBs. SUMO-NBs are enriched in SUMOylated proteins, colocalize with PML (21, 26), and might function as nuclear SUMOylation hot spots (27–29). Since ZIKV NS5 protein contains potential SUMOylation sites, we investigated whether NS5 speckles are associated with SUMO-1 or PML NBs. We found little ZIKV NS5 specifically associated with SUMO-1 NBs at 12 hpi (Fig. 2A). However, by 18 hpi, ∼25% of ZIKV NS5 is colocalized with SUMO-1 NBs, and ∼75% of NS5 was present in discrete SUMO-1-independent speckles (Fig. 2A). The number of colocalized NS5/SUMO-1 NBs increased slightly by 24 hpi, with most SUMO-1-independent NS5 speckles adjacent to and bordering SUMO-1 NBs. These findings indicate that NS5 forms SUMO-1-colocalized and SUMO-1-independent nuclear speckles during a single cycle of ZIKV replication in hBMECs.

FIG 2.

ZIKV NS5 forms unique nuclear speckles that disrupt PML/SUMO-1 NBs. (A) hBMECs grown on microslides were infected with ZIKV-PRV (MOI, 2) and fixed 12, 18, and 24 hpi, immunostained for ZIKV NS5 and SUMO-1 or PML, and visualized by confocal microscopy. (B) Mock-infected or ZIKV-infected hBMECs were immunostained for PML and SUMO-1 24 hpi and visualized by confocal microscopy. (C) ZIKV-infected hBMECs were immunostained for ZIKV NS5 and fibrillarin at 24 hpi and visualized by confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 5 μm.

In uninfected hBMECs, we found that SUMO-1 and PML are constitutively expressed and nearly 100% colocalized in NBs (Fig. 2B). Following ZIKV infection of hBMECs, we observed little NS5 colocalization with PML (Fig. 2B); however, ZIKV infection resulted in a dramatic ∼80% reduction in colocalized SUMO-1 and PML compared to mock-infected hBMECs (Fig. 2B). These findings suggest that in hBMECs, ZIKV infection disrupts normal SUMO-1 association with PML, excludes PML from SUMO-1 NBs, and coordinately reduced PML expression levels. Taken together, these findings demonstrate that ZIKV NS5 plays a fundamental role in the formation of nuclear bodies that govern hBMEC transcriptional responses.

ZIKV does not direct nucleolar remodeling in hBMECs.

Several reports suggest that flaviviruses (DENV, JEV, WNV, and HCV) direct nucleolar remodeling (30–33). To determine if ZIKV alters nucleoli, we analyzed the cellular localization of the nucleolar protein fibrillarin in ZIKV-infected hBMECs. We found that ZIKV NS5 was localized to the nucleoplasm of hBMECs and that the nucleolar localization of fibrillarin was identical in mock- and ZIKV-infected hBMECs (Fig. 2C). These findings reveal no evidence of nucleolar remodeling in ZIKV-infected hBMECs.

SIM sites in ZIKV and DENV NS5 direct nuclear localization.

Flavivirus NS5 proteins contain consensus NLS, SUMOylation sites, and conserved SUMO-interacting motifs (SIM) that are associated with the formation of nuclear bodies. Analyzing ZIKV NS5 using the Joint Advanced SUMOylation Site and SIM Analyzer (JASSA) (34) identified two consensus SUMOylation motifs and one SIM site (Fig. 3A). These motifs are exposed on NS5 surfaces (Fig. 3B) and may direct nuclear interactions of NS5.

FIG 3.

SIM sites in ZIKV and DENV NS5 direct nuclear localization. (A) Multiple-sequence alignment using Clustal Omega comparing conserved SIM site (VIDL or VVDL) domains of flavivirus NS5 proteins. Black highlight indicates fully conserved residues, gray indicates conserved residues with similar properties, and white indicates dissimilar residues. (B) Crystal structures of ZIKV and DENV3 NS5 proteins highlighting the SIM and SUMOylation sites predicted by JASSA. Crystal structure coordinates of ZIKV NS5 (PDB entry 5U0B) and DENV3 NS5 (PDB entry 4V0Q) were used to highlight SIM (green), NLS (blue), and SUMOylation sites (red) using UCSF Chimera software. (C) HEK293T cells were transfected with ZIKV NS5 WT, ZIKV NS5 SIM mutant (SIM_mut), DENV NS5 WT, or DENV NS5 SIM_mut and lysed at 24, 48, or 72 h posttransfection (hpt). NS5 protein levels were determined by Western blotting using anti-FLAG antibody, and GAPDH levels are loading controls. Appended graphs represent NS5 band quantification normalized against GAPDH levels in arbitrary units (AU). Asterisks indicate statistical significance (***, P < 0.001; ns, not significant), as determined by one-way ANOVA. (D and E) Lentivirus-transduced hBMECs expressing ZIKV NS5 WT, SIM_mut, or NLS_mut (D) or DENV NS5 WT, SIM_mut, or NLS_mut (E) were immunostained for NS5 using anti-FLAG antibody and for SUMO-1. Cells were visualized by confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 10 μm.

In general, SIMs interact with SUMOylated proteins to form nuclear speckles (35), but functions of the ZIKV NS5 SIM site have yet to be revealed. We generated SIM site and NLS mutants of ZIKV and DENV3 NS5 proteins and transiently expressed NS5 mutants in HEK293T cells. Suggesting a general effect of SIM sites on NS5 protein stability, both ZIKV and DENV3 NS5 SIM mutant expression levels were reduced compared to that of wild-type (WT) NS5 expression (Fig. 3C).

We found that lentivirus-expressed ZIKV and DENV NS5 proteins localized to the nucleus of hBMECs (Fig. 3D and E). However, the patterns of ZIKV and DENV NS5 localization in the nucleus were completely distinct, with ZIKV NS5 localizing to punctate nuclear speckles (Fig. 3D) and DENV NS5 diffusely spread throughout the nucleoplasm (Fig. 3E). Mutating the conserved bipartite NLS of either ZIKV or DENV3 NS5 resulted in complete cytoplasmic protein localization and functionally substantiates the NLS motif as determinant of nuclear localization (Fig. 3D and E). Despite reduced expression of NS5 SIM mutants, we found that in hBMECs both ZIKV and DENV3 NS5 SIM mutants were also localized exclusively to the cytoplasm, with a distribution similar to that of the NLS mutants (Fig. 3D and E). These findings indicate a novel role for conserved SIM sites in directing ZIKV and DENV3 NS5 proteins to the nucleus and demonstrate that both SIM and bipartite NLS motifs are required for the nuclear localization of NS5 in hBMECs. However, whether SIM interactions function as an NLS by stabilizing nuclear NS5 complexes, or preventing exposure of potential nuclear export signals (NES), remains to be determined.

Mutating ZIKV NS5 K252R SUMO site enhances colocalization with SUMO-1 NBs.

To determine whether SUMOylation directs ZIKV NS5 into punctate nuclear speckles, we mutated lysine residues in consensus SUMO motifs (K252R and K697R) and lentivirus expressed the mutant proteins in hBMECs (Fig. 4A). The NS5-K697R SUMO mutant was too unstable to analyze and prevented further evaluation. Similar to ZIKV infection, expression of WT ZIKV NS5 in hBMECs resulted in the formation of NS5 nuclear speckles that were either SUMO-1 NB colocalized or SUMO-1 independent (Fig. 4B). Most NS5-SUMO-1-colocalized NBs formed unique perispeckle doughnuts (36, 37), with NS5 forming the center surrounded by a discrete concentric ring of SUMO-1 (Fig. 4B). In contrast, expression of the NS5-K252R mutant in hBMECs resulted in enlarged, uniformly colocalized NS5-K252R and SUMO-1 NBs (Fig. 4B). Quantitating the area of SUMO-1 NBs revealed an ∼3-fold increase in the size of NS5 K252R mutant/SUMO-1 nuclear speckles versus WT NS5/SUMO-1 speckles (Fig. 4C). These findings demonstrate that the NS5 K252 SUMO site governs NS5 interactions with SUMO-1 and directs the assembly and potential function of perispeckles. Perispeckle doughnuts are features of nuclear bodies associated with RNA splicing and TREX-directed RNA export complexes that have yet to be studied as regulatory elements during ZIKV infection (36, 37).

FIG 4.

Mutating ZIKV NS5 K252 enhances colocalization with SUMO-1. (A) Schematic of ZIKV NS5 protein SUMO, SIM, and NLS motifs targeted for site-directed mutagenesis. (B) Lentivirus-transduced hBMECs expressing ZIKV NS5 WT or K252R were immunostained for NS5 (anti-FLAG antibody) and for SUMO-1 and visualized by confocal microscopy. Insets show magnified speckles from ZIKV NS5 WT and K252R merged images. Asterisks represent examples of SUMO-1-independent NS5 speckles. Bars represent 5 μm. (C, left) ZIKV NS5 speckle area (μm²) was quantified and plotted against its reciprocal SUMO-1 NB using ImageJ software. (Right) Reciprocal NS5-SUMO-1 speckle area ratio was quantified for WT- or K252R-expressing hBMECs. Experiments were done in triplicate, repeated at least three times, and representative data are presented.

Nuclear localization of Dengue virus NS5 is SUMOylation dependent.

We expressed DENV3 NS5-K279R, -K546R, and -K637R SUMO mutants (Fig. 5A) and assayed their nuclear localization in hBMECs. DENV3 NS5 WT and -K297R and -K637R mutant proteins are dispersed in the nucleus of hBMECs and very distinct from punctate ZIKV NS5 speckles (Fig. 5B). However, like ZIKV NS5, expression of the DENV3 NS5-K546R mutant in hBMECs resulted in the formation of distinct nuclear NS5 speckles and perispeckle doughnuts that colocalize with SUMO-1 (Fig. 5B). Thus, ablating the NS5-K546R SUMO motif caused the DENV3 NS5 to form nuclear speckles and colocalize with SUMO-1 NBs. We observed similar SUMO-2/3 isoform nuclear colocalization with WT and mutant ZIKV and DENV3 NS5 proteins (Fig. 6). Collectively, these findings indicate that discrete SUMOylation sites within DENV and ZIKV NS5 proteins determine their ability to form speckles and colocalize with SUMO-1. These findings suggest flavivirus-specific roles for discrete NS5 protein SUMOylation sites in regulating nuclear interactions and cellular responses.

FIG 5.

Mutating DENV3 NS5 K546 enhances colocalization with SUMO-1. (A) Schematic linear DENV3 NS5 protein SUMO, SIM, and NLS motifs targeted for site-directed mutagenesis. (B) Lentivirus-transduced hBMECs expressing DENV3 NS5 WT, K279R, K546R, K637R, or controls were immunostained for NS5 (anti-FLAG antibody) and for SUMO-1 and visualized by confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 5 μm.

FIG 6.

Mutating ZIKV NS5 K252R and DENV3 NS5 K546R enhances colocalization with SUMO-2/3. Lentivirus-transduced hBMECs expressing ZIKV NS5 WT or K252R mutant (A) or DENV3 NS5 WT, K279R, K546R, or K637R mutants (B) were immunostained for NS5 using anti-FLAG antibody and for SUMO-2/3 and visualized by confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 5 μm.

SUMOylation of ZIKV NS5 is SIM site independent.

We analyzed SUMOylation of WT and mutant NS5 proteins in cells constitutively expressing His-tagged SUMO-1 or SUMO-2 isoforms. We found that WT DENV NS5 was SUMO-1 modified but observed no evidence of posttranslational SUMO-1 addition to the DENV NS5 SIM mutant (Fig. 7, lanes 1 and 2). In contrast, we observed high-molecular-weight SUMOylated forms of ZIKV WT, SIM, and K252R mutant NS5 proteins (Fig. 7, lanes 4, 6, and 7). In cells expressing His₆-SUMO-2, we also found that the ZIKV NS5 K252R mutant was SUMOylated but with a distinct NS5 electrophoretic pattern (Fig. 7, lanes 7 and 8). While both DENV and ZIKV NS5 SIM mutants are localized to the cytoplasm of hBMECs, only SUMOylation of the DENV NS5 protein was SIM dependent. Collectively, these findings demonstrate that SUMOylation of ZIKV NS5 is uniquely SIM and nuclear localization independent.

FIG 7.

SUMOylation of ZIKV NS5 is SIM site independent. HeLa cells were cotransfected with empty vector (−) or His₆-tagged SUMO-1 or SUMO-2 plasmids and the indicated FLAG-tagged DENV3 or ZIKV NS5 plasmids. His₆-SUMO-conjugated proteins were bound to Ni²⁺-NTA agarose beads, and total or SUMO-modified NS5 was detected by Western blotting using anti-FLAG antibody. Experiments were done in triplicate, repeated at least two times, and representative data are presented.

ZIKV NS5 K252R mutant attenuates hBMEC transcriptional responses.

Transcriptional responses that permit ZIKV to persist in hBMECs may be determined by SIM and SUMO sites that mediate nuclear speckle formation. To analyze the effect of NS5 SUMOylation on hBMEC responses, we analyzed basal gene expression of hBMECs constitutively expressing WT or mutant ZIKV NS5 proteins using Affymetrix arrays. The vast majority of changes observed were in transcripts that regulate cell cycle and nucleosome assembly, which were uniquely downregulated 2- to 9-fold in WT ZIKV NS5-expressing hBMECs versus hBMECs expressing NS5 SUMO-K252R or NLS mutants. WT-NS5 downregulated cell cycle-related transcriptional responses of CCNE2, CDKN3, CKS2, RCC1, Notch1, CDKL1, CHEK1, CDK5KAP3, CDCA7L, E2F1, and E2F8. The E2F family plays a crucial role in the control of cell cycle, tumor suppressor protein regulation, and angiogenesis, while others are cyclin-dependent kinases, cell cycle suppressors, or checkpoint proteins or regulate cell proliferation (38).

We also found that hBMECs expressing WT ZIKV NS5 suppressed basal-level ISG transcript levels 2- to 6-fold compared to those of NS5-K252R or NLS mutants. Basal-level ISGs inhibited by WT-ZIKV NS5 include OAS2, IFI6, OAS1, MX1, IFI44L, and IFITM1 (Fig. 8A). To validate array data, we found that expressed WT ZIKV or DENV NS5 proteins suppressed IFIT1, MX1, and OAS2 mRNA transcripts >90% by quantitative reverse transcription-PCR (qRT-PCR) compared to green fluorescent protein (GFP)-expressing hBMECs (Fig. 8B). In contrast, basal ISG transcripts increased 5- to 8-fold in hBMECs expressing ZIKV NS5-K252R or NLS mutant proteins (Fig. 8B). These findings suggest a novel role for NS5 nuclear localization and site-specific SUMOylation in the regulation of basal ISG transcription in hBMECs.

FIG 8.

ZIKV NS5 K252 regulates hBMEC basal ISG transcriptional responses. (A) Affymetrix microarray analysis of basal ISG transcript levels of hBMECs expressing ZIKV NS5 K252R mutant relative to ZIKV NS5 WT. (B) Lentivirus-transduced hBMECs expressing ZIKV NS5 WT, K252R, NLS_mut mutants, DENV3 NS5 WT, or GFP were lysed, total RNA was purified, and changes in basal level ISGs transcripts (OAS2, MX1, and IFIT1) were analyzed by RT-qPCR. Asterisks indicate statistical significance (**, P < 0.01; ***, P < 0.001), as determined by one-way ANOVA with Tukey’s post hoc test. Experiments were performed at least 3 times with similar results.

ZIKV NS5 SUMO K252 is required for regulating IFN-induced ISGs.

As NS5 was previously reported to restrict STAT2-directed ISG responses (13), we evaluated NS5 mutant regulation of IFN-directed ISG induction. Cells expressing WT NS5 from ZIKV or DENV suppressed IFN-α induction of IFIT1, OAS2, MX1, and ISG15 by 60- to 1,250-fold (Fig. 9). In contrast, NS5 NLS and K252R SUMO mutants failed to inhibit ISG induction (Fig. 9). These findings indicate that the NS5-K252 SUMO motif and NLS are required to regulate IFN-directed ISG responses in hBMECs.

FIG 9.

ZIKV NS5 K252 is required for regulating IFN induced ISGs. Lentivirus-transduced hBMECs expressing ZIKV NS5 WT, K252R, NLS_mut, DENV3 NS5 WT, or GFP were induced with IFN-α, total RNA was purified, and changes in ISG induction (OAS2, MX1, IFIT1, and ISG15) were analyzed by qRT-PCR at 24 h. Asterisks indicate statistical significance (*, P < 0.05; ***, P < 0.001), as determined by one-way ANOVA with Tukey’s post hoc test. Experiments were performed at least 3 times with similar results.

Prior studies found that in HEK293T or A549 cells, ZIKV NS5 binds STAT2, and that NS5 directs STAT2 ubiquitination and degradation in response to IFN-α addition (39–41). In hBMECs, STAT2 is constitutively present in the nucleus (Fig. 10A), and we found that STAT2 colocalizes with ZIKV NS5 nuclear speckles (WT or -K252R mutant) in the presence or absence of added IFN-α (Fig. 10A and B). However, in response to IFN-α addition, we also observed a significant reduction in cytoplasmic and nuclear-localized STAT2 in hBMECs expressing ZIKV NS5 or NS5-K252R proteins (Fig. 10).

FIG 10.

ZIKV NS5 speckles colocalized with STAT2 in the nucleus of hBMECs. Lentivirus-transduced hBMECs constitutively expressing ZIKV NS5 WT or K252R mutant proteins (A) or induced with IFN-α (right) and hBMEC controls (B) were immunostained for NS5 using anti-FLAG antibody and for STAT2 and analyzed using confocal microscopy. Experiments were done in triplicate, repeated at least three times, and representative data are presented. Bars represent 10 μm.

Consistent with subcellular localization, expressing ZIKV NS5 in hBMECs resulted in an 80% reduction in endogenous STAT2 protein without affecting IRF9 or glyceraldehyde-3-phosphate dehydrogenase (GAPDH) protein levels by Western blotting (Fig. 11A). In contrast, expressing ZIKV NS5-K252R or NLS mutants reduced hBMEC STAT2 levels ∼50%. In IFN-α-treated hBMECs expressing ZIKV NS5, we observed an ∼80% reduction in STAT2 and pSTAT2 levels compared to those of GFP controls, while IFN-α-treated hBMECs expressing the NS5-K252R mutant reduced STAT2 and pSTAT2 levels by 50% (Fig. 11B). Consistent with this and Fig. 9 transcription data, we found a dramatic reduction in IFIT1 expression in IFN-α-stimulated hBMECs expressing WT ZIKV NS5 that was not observed by expressing NS5-K252R or NLS mutants (Fig. 11B). These results suggest a mechanism by which SUMOylated, nuclear NS5 regulates ISG induction by reducing constitutive and IFN-induced STAT2 levels.

FIG 11.

ZIKV NS5 binds STAT2 in the nucleus and partially requires K252 for binding and degradation. Lentivirus-transduced hBMECs expressing ZIKV NS5 WT, K252R, NLS_mut, or GFP were unstimulated (A) or IFN-α stimulated for 1 h (B). Protein levels of STAT2, pSTAT2, IRF9, and IFIT1 from cell lysates were determined by Western blotting. GAPDH protein level was used as a loading control. Appended graphs represent STAT2 and pSTAT2 band quantification normalized against GAPDH levels. Asterisks indicate statistical significance (*, P < 0.05; ***, P < 0.001) as determined by two-way ANOVA. (C) Cell lysates from hBMECs expressing ZIKV NS5 WT, K252R, NLS_mut, or GFP were immunoprecipitated using anti-FLAG M2 beads for Flag-tagged NS5 and analyzed by Western blotting using anti-FLAG and anti-STAT2 antibodies on immunoprecipitated and total cell lysates.

Analysis of ZIKV NS5 and STAT2 interactions revealed that WT NS5 robustly coprecipitated STAT2 from hBMECs compared to NS5-K252R or NS5 NLS mutants (Fig. 11C). These findings suggest that nuclear NS5 interactions with STAT2 are largely dependent on the presence of the K252 SUMOylation site. Curiously, the overall reduction in STAT2 levels resulting from WT ZIKV NS5 expression (Fig. 11A and B) is not apparent in the nucleus of IFN-treated hBMECs, where STAT2 levels appear similar in the presence of WT or K252R mutant NS5 proteins (Fig. 10). It remains unclear whether cytoplasmic STAT2 or pSTAT2 is selectively degraded in ZIKV NS5-expressing hBMECs or whether high endogenous nuclear STAT2 levels, in the absence of IFN, are targeted by NS5 expression.

In contrast to STAT2, we observed a dramatic reduction in PML NBs in cells expressing WT NS5 versus the NS5-K252R mutant (Fig. 12A and B). Compared to expressing WT NS5, the NS5-K252R mutant is nearly completely colocalized with both nuclear STAT2 and PML in hBMECs (Fig. 2A and 12A) yet failed to reduce PML levels by immunofluorescence assay or Western blot analysis (Fig. 12A and B). In ZIKV-infected hBMECs, PML protein levels decreased concomitantly with increased NS5 expression, 12 to 24 hpi, with total PML expression reduced 70% by 24 hpi (Fig. 12C). Consistent with WT NS5 regulation of PML-directed ISG responses, we found that the NS5-K252R SUMO mutant fails to block basal or IFN-induced ISG responses and only partially reduces STAT2 or pSTAT2 levels. As PML binds STAT2 and promotes ISG transcription (42), our findings suggest that in hBMECs, ZIKV NS5 may reduce ISG responses by directing PML degradation and dissociating PML from STAT2. Collectively, these findings demonstrate that ZIKV NS5 regulation of basal and IFN-stimulated ISG induction is dependent on nuclear NS5 expression, and that K252 SUMOylation is required for STAT2 binding and inhibiting the formation of antiviral PML-STAT2 NBs.

FIG 12.

PML staining is reduced in hBMECs expressing WT ZIKV NS5. (A) hBMECs expressing ZIKV NS5 WT or K252R mutant were immunostained for NS5 (anti-FLAG) and anti-PML antibody and visualized by confocal microscopy. Experiments were repeated at least three times. Bars represent 5 μm. (B) Lentivirus-transduced hBMECs expressing ZIKV NS5 WT, K252R, NLS_mut, or GFP were lysed, and PML protein levels were determined by Western blotting. Bar graph quantifies PML levels relative to cells expressing GFP and are normalized to GAPDH protein. Asterisks indicate statistical significance (**, P < 0.01), as determined by one-way ANOVA with Tukey’s post hoc test. (C) Protein levels of PML and ZIKV NS5 from cell lysates of mock- or ZIKV-infected hBMECs at 12, 18, and 24 hpi were determined by Western blotting with GAPDH protein levels as loading controls. Bar graph quantifies PML levels in ZIKV-infected versus mock-infected controls normalized to GAPDH levels. Asterisks indicate statistical significance (**, P < 0.01), as determined by one-way ANOVA with Tukey’s post hoc test.

DISCUSSION

ZIKV is a mosquito-borne flavivirus that uniquely persists in patients, is transmitted sexually, and crosses blood-brain barriers to cause encephalitis and intrauterine fetal microencephaly (1, 3, 43–45). ZIKV damages placental, corneal, and neuronal tissues; however, ZIKV persistence suggests that viral replication and dissemination occur in discrete cellular reservoirs where ZIKV fosters cell survival and evades innate and adaptive immune responses (1, 44). Persistence provides ZIKV an extended window to enter protected neuronal compartments and spread across the placenta to gain access to fetal neuronal tissues (44). We previously reported that ZIKV persistently infects hBMECs without cytopathology 1 to 9 dpi and after serial passage of infected hBMECs (7). Consistent with systemic and neuronal spread, ZIKV is released from infected hBMECs apically and basolaterally without disrupting endothelial barrier functions (7). These findings distinguish hBMECs from other cell types that are lytically infected by ZIKV (1, 46) and suggest that persistently infected hBMECs serve as reservoirs for ZIKV replication that enables viral access to neuronal compartments. Mechanisms by which ZIKV regulates innate immune responses and fosters cell survival to persistently infect hBMECs remain enigmatic.

ZIKV and DENV infection of endothelial cells results in divergent interferon regulation (7, 47). DENV infection transiently inhibits early IFN-β induction 12 to 24 hpi; however, DENV-induced IFN-β expression 2 to 3 dpi prevents viral spread in endothelial cells (47). In contrast, ZIKV transcriptionally induces IFN-β and IFN-γ 1 to 2 dpi, yet IFNs are not expressed or secreted from ZIKV-infected hBMECs at any time after infection (7). These findings rationalize persistent ZIKV infection, replication, and spread in hBMECs, but the mechanism by which ZIKV establishes persistence in hBMECs remains to be resolved (7). Novel interactions of NS5 in the nucleus of hBMECs suggest mechanisms for ZIKV to restrict innate antiviral responses, foster cell survival, and persist in hBMECs.

Flavivirus NS5 proteins have enzymatic methyltransferase and polymerase functions required for cytoplasmic replication but are highly localized to the nucleus, where their functions remain unresolved. NS5 is also linked to the regulation of IFN, IFNAR, ISG, and STAT2 responses (13, 16, 39–41). Posttranslational SUMOylation of DENV NS5 was suggested as a mechanism that enhances DENV replication, NS5 stability, and STAT2 suppression (16). However, inhibiting SUMOylation has little effect on DENV replication; instead, slight titer changes are consistent with increased IFN-β production that result from blocking UBC9-directed SUMOylation (13, 16, 48). Further complicating this, most studies of NS5 function have been pursued in cells with defective IFN signaling pathways or that are lytically infected (24, 39, 41). ZIKV uniquely establishes a persistent infection in primary hBMECs that differentiates this cell type and provides rationales for ZIKV to control IFN, apoptotic and proliferative responses that are regulated by nuclear NS5 functions.

Studies of NS5 have suggested both cytoplasmic and nuclear roles in the flavivirus life cycle (13). One study suggests that in neuronal LN229 or Huh7 cells, ZIKV NS5 sequesters Importin-α and forms nuclear bodies dependent on a C-terminal monopartite NLS (24). Another suggests that NS5 functions primarily in the cytoplasm and that the NS5 NLS is not required for DENV replication or STAT2 degradation (17). An increase in DENV2 NS5 stability was suggested to be required for suppressing STAT2 (16) and for ZIKV NS5 to suppress IFN signaling (49). However, it is unclear whether these suggested associations are a function of nuclear localization, reduced NS5 expression, dimerization, changes in IFN-β induction, or effects on NS5 SUMOylation.

Our studies reveal roles for NLS, SUMO, and SIM sites in the formation of ZIKV NS5 NBs and the nuclear regulation of STAT/IFN responses. We found that mutating an NLS in the ZIKV NS5 functionally alters both STAT2 interactions and ISG regulation in hBMECs, and this suggests that nuclear localization is a requirement for IFN pathway regulation. In the course of these experiments, we unexpectedly found that a conserved SIM site (VIDL) in ZIKV and DENV NS5 proteins is required for the localization and sequestering of NS5 in the nucleus (24). We found that the DENV NS5 SIM site is required for both NS5 SUMOylation and nuclear localization, while the ZIKV SIM site was required for nuclear localization but dispensable for NS5 SUMOylation. Although roles for SIM sites in NS5 nuclear localization were not previously reported, our findings demonstrate that both the conserved SIM site and bipartite NLS are required for the nuclear localization of ZIKV and DENV NS5 proteins. The mechanism by which conserved SIM sites serve as an NLS remains an enigma, as SIMs could be required for nuclear import, recognition of the NS5 NLS, NS5 dimerization (50), masking of potential nuclear export signals, or formation of complexes with nuclear proteins that restrict nuclear export. Although SIM site mutations reduce NS5 expression, our finding that SIMs confer nuclear localization implies that novel NS5 SIM interactions and functions have yet to be resolved. SIMs direct protein-protein interactions with SUMOylated proteins, function to regulate the assembly of nuclear transcription complexes (20, 35), and could direct or sequester NS5 complexes in the nucleus (34, 51).

Our results demonstrate that in primary hBMECs, ZIKV NS5 forms discrete nuclear speckles that colocalize with SUMO-1 and dissociate SUMO-1 from its nuclear partners, PML and STAT2. Punctate ZIKV NS5 speckles are contrasted by the dispersed nuclear trafficking of DENV3 NS5 in hBMECs. The ZIKV NS5-K252R SUMO mutant restructured SUMO-1 perispeckles to be 100% colocalized and merged with SUMO-1. In contrast, mutating two predicted SUMO sites in DENV3 NS5 had no effect on the diffuse NS5 nuclear localization, while mutating a third SUMO site, NS5-K546R, caused NS5 to form punctate SUMO-1-colocalized nuclear speckles in hBMECs, similar to ZIKV NS5. Thus, SUMOylation sites in both ZIKV and DENV NS5 proteins dictate the formation of NS5 nuclear speckles and their colocalization with SUMO-1 and STAT2. This demonstrates that specific SUMOylation sites contribute to flavivirus nuclear complexes that may regulate cellular functions and cell type-specific responses that permit ZIKV to uniquely persist in hBMECs.

The nucleoplasm of cells is not homogenous, and there are many nuclear bodies, domains, subcompartments, and speckles described that dynamically separate or associate specific factors within restricted areas of the nucleus. NBs are present during G₁/S/G₂ interphase and disassemble during further regulation of cell cycle and cyclin-dependent kinase functions (19). NBs/speckles facilitate the integrated regulation of gene expression, signaling and mRNA synthesis, splicing, maturation, and export (19). SUMO-1 and other posttranslational modifiers regulate protein-protein interactions that form NBs, control their composition structurally and functionally, and are associated with nuclear pore complexes and HDAC1 transcriptional repression (18). Mutating SIM sites in viral proteins has also been shown to inhibit SUMOylation, impair complex formation, and change NB cellular localization (52, 53). SUMOylation also coordinates the repression of tightly controlled basal and IFN-induced ISG responses that direct cellular antiviral programs (48).

Our findings reveal roles for ZIKV NS5 SUMOylation, NLS, and SIM sites in regulating hBMEC transcriptional responses. Comparison of basal-level RNAs in hBMECs constitutively expressing NS5 versus NS5 mutants revealed two discrete clusters of transcriptional changes associated with cell cycle and ISG regulation. Compared to NS5 mutants, expressing ZIKV NS5 in hBMECs resulted in the selective downregulation of 11 cell cycle-related transcripts. The role of NS5-regulated cell cycle genes and their overall effect on cell cycle progression, proliferation, and hBMEC viability have yet to be defined. The ability of ZIKV to cause mitotic dysfunction that inhibits neuronal progenitor proliferation has been reported (54), and the ability to regulate these responses could be fundamental to the unique persistence of ZIKV in hBMECs. There is one report suggesting that in HeLa cells NS5 colocalizes with Cajal bodies, nuclear bodies associated with nucleoli and cell cycle progression (55, 56). However, our findings demonstrate that in hBMECs, ZIKV NS5 forms nuclear speckles that are not associated with nucleoli and fail to alter nucleolar protein localization (55, 56).

Constitutive expression of ZIKV NS5 reduced basal ISG transcripts levels in hBMECs 2- to 6-fold compared to NS5-K252R SUMO or NLS mutants. These findings reveal roles for NS5 SUMOylation and nuclear localization in repressing basal, as well as IFN-induced, antiviral ISG responses. NS5 restriction of basal-level ISG responses is consistent with a recent report showing nuclear STAT2-IRF9 regulation of ISGs without IFN stimulation (57). Analysis of selected ISG transcripts regulated by WT NS5, SUMO, and NLS mutants supports NS5 inhibition of STAT2-directed ISG responses and further reveals that ISG regulation is dependent on SUMOylation and nuclear localization in hBMECs.

Consistent with persistence, ZIKV-infected hBMECs fail to secrete IFNs that would otherwise restrict ZIKV spread (7), and accordingly ZIKV employs novel mechanisms that uniquely regulate antiviral IFN and ISG responses in hBMECs. Prior studies demonstrate that DENV and ZIKV NS5 proteins direct STAT2 degradation, which restricts ISG induction (14, 39, 41), but these findings were obtained in cell lines with altered IFN responses or during lytic infection.

Primary hBMECs are unique at many levels from immune and epithelial cells and perform novel functions in the lining of the vasculature that regulate traffic into restricted tissues (58–62). In hBMECs, high levels of endogenously expressed STAT2 and PML are found in the nucleus colocalized with SUMO-1. Consistent with NS5-STAT2 regulation (39, 40), ZIKV NS5 coprecipitated STAT2 from hBMECs, and both reduced STAT2 and pSTAT2 levels via a mechanism that requires NS5 SUMOylation and nuclear localization. In ZIKV-infected, or NS5-expressing, hBMECs we found that NS5 colocalized with SUMO-1 and STAT2, and that NS5 expression dissociated PML from NS5/SUMO-1/STAT2 NBs, dramatically reducing PML protein levels. In contrast, the ZIKV NS5-K252R SUMO site mutant colocalized with PML but failed to dissociate PML from STAT2 or reduce PML protein expression. These findings reveal that in hBMECs, reductions of STAT2 appear to be linked to changes in the colocalization of SUMO-1 and PML that govern IFN and ISG induction and repression (63). These findings suggest a mechanism by which ZIKV NS5 restricts ISG responses through NS5-SUMO site interactions that disrupt PML-STAT2 NBs in hBMECs.

The reason that hBMECs constitutively express high nuclear levels of PML and STAT2 remains unknown, but nuclear PML/STAT2 expression distinguishes hBMEC responses and the novel role of the vascular endothelium during viral infection. PML positively regulates IFN-directed ISG responses, binds to and promotes STAT2 accumulation, and is associated with restricted cell growth and programmed cell death (42, 64, 65). PML forms complexes containing STAT2 and histone deacetylases (HDAC1/2) that are regulated by protein SUMOylation (42, 66, 67). Viruses have evolved mechanisms to hijack cellular SUMOylation machinery and PML-directed IFN and ISG induction to promote viral replication (64, 68). PML depletion renders STAT2 susceptible to proteasomal degradation (42), and DNA viruses regulate STAT2 by preventing PML interactions with STAT2 or by deSUMOylating and degrading PML (42). ZIKV NS5 binds STAT2 and dissociates STAT2/PML complexes in the nucleus, reducing both PML and STAT2 levels in hBMECs. Given that PML stabilizes STAT2, these findings suggest that reduced STAT2 levels are secondary to NS5-directed PML depletion and dissociation from STAT2. As a result, NS5 SUMO site interactions may determine the assembly of antiviral PML-STAT2 complexes and ISG responses restricted by ZIKV NS5. These findings suggest a fundamental mechanism by which ZIKV disrupts STAT2-PML-NBs in hBMECs.

The ability of ZIKV NS5 to regulate basal hBMEC responses may permit ZIKV to sustain hBMEC proliferation and establish a nonlytic infection. These changes may reflect cell type-specific regulation of ZIKV-infected hBMECs that are contrasted with lytic ZIKV infection of neurons and other cell types. Although the mechanism by which ZIKV maintains the viability of hBMECs during infection remains to be resolved, NS5 SUMOylation and nuclear import, in combination with the unique hBMEC nuclear organization and endothelial cell-specific functions, may in concert enable ZIKV persistence.

MATERIALS AND METHODS

Cells and virus.

HEK293T cells (ATCC) were grown in DMEM (Dulbecco’s modified Eagle’s medium) supplemented with 8% fetal bovine serum (FBS) and penicillin (100 μg/ml), streptomycin sulfate (100 μg/ml), and amphotericin B (50 μg/ml; Mediatech) at 37°C and 5% CO₂. hBMECs (passage 3), derived from elutriation of dispase-dissociated normal human brain cortex tissue, were purchased from Cell Biologics (H-6023) and were grown in endothelial cell basal medium-2 MV (EBM-2 MV; Lonza) supplemented with EGM-2 MV SingleQuots (Lonza) and incubated at 37°C and 5% CO₂. ZIKV (PRVABC59) was obtained from the ATCC, minimally passaged (multiplicity of infection [MOI], 0.1 to 1), and propagated for 5 days in Vero E6 cells in DMEM with 2% FBS.

Plasmids.

pLV-ZIKV-NS5-Flag and pNIC28-DENV3-NS5FL expression plasmids were purchased from Addgene. ZIKV and DENV3 NS5 protein-coding regions were PCR amplified using primers containing BamHI and XbaI restriction sites and a Flag tag and cloned into BamHI- and XbaI-cut pLenti-puro (Addgene). Site-directed mutagenesis was performed using PfuUltra high-fidelity DNA polymerase (Agilent) to generate protein mutants containing amino acid changes for ZIKV NS5 SUMOylation motifs (K252R and K697R), NLS mutant (R393A/V394A), and SIM mutant (V77A/L80A) and for DENV3 NS5 SUMOylation motifs (K279R, K546R, and K637R), NLS mutant (R384A/R386A), and SIM mutant (V72A/L75A) by following the manufacturer’s protocol. Mutants were sequenced, and expression was confirmed by Western blotting. pLenti-GFP was used as a lentivirus transduction control.

Lentiviral vector production.

Lentivirus vectors were generated by transfecting HEK293T cells by using a polyethylenimine (PEI) transfection protocol (69). HEK293T cells (3.8 × 10⁶) were preincubated with 25 μM chloroquine diphosphate for 5 h and transfected with pLenti-Puro or -BLAST expression plasmids in a 3:2:1 ratio with psPAX2 and pLP/VSVG using a DNA/PEI (micrograms) ratio of 1:3. After 18 h, the medium was replaced and the viral supernatants were harvested 72 hpt and filtered through a 0.45-μm polyvinylidene difluoride filter.

Confocal immunofluorescence.

To examine ZIKV NS5 speckles during infection, ZIKV (MOI, 2) was adsorbed to ∼60% confluent hBMEC monolayers for 1 h in Lab-Tek II chambers (Nunc), washed with phosphate-buffered saline (PBS), and grown in supplemented EBM-2 MV with 5% FBS. At the indicated times, cells were washed with PBS, fixed for 10 min with 4% paraformaldehyde-PBS, and permeabilized with 0.1% Triton X-100 in PBS for 10 min. Cells were blocked using 5% bovine serum albumin in PBS for 2 h and incubated with anti-NS5 rabbit polyclonal (GTX133327; GeneTex) and 4G2 mouse monoclonal antibodies diluted 1:2,000 and 1:100, respectively, in blocking solution for 18 h at 4°C.

To examine ZIKV NS5 colocalization with nuclear bodies in hBMECs, mouse monoclonal antibodies anti-fibrillarin (sc-166001), anti-PML (sc-966), and anti-SUMO1 (sc-393144) (1:100; Santa Cruz Biotechnology) were coincubated with anti-NS5 antibody (1:2,000) for 18 h at 4°C. Cells were washed and incubated for 2 h with Alexa 488-conjugated goat anti-mouse IgG antibody (Invitrogen) and with Alexa 546-conjugated goat anti-rabbit IgG antibody (Invitrogen) diluted 1:400 in blocking solution at room temperature. hBMECs were subsequently incubated with 5 μM 4,6-diamidino-2-phenylindole (DAPI; Sigma) for 5 min at room temperature. Slides were mounted using ProLong Antifade (Thermo Fisher) solution and observed using a Zeiss LSM 510 META/NLO confocal microscope. To examine protein localization from hBMECs constitutively expressing ZIKV and DENV3 NS5 WT and mutants, ∼60% confluent monolayers were lentivirus transduced for 6 h at 37°C and 5% CO₂, and staining was performed after 48 h using anti-Flag M2 mouse monoclonal antibody (1:100; Millipore Sigma) coincubated with rabbit polyclonal anti-SUMO-1 (1:1,000; 4930; Cell Signaling), anti-SUMO-2/3 (1:1,1000; PA5-19418; Invitrogen), anti-STAT2 (1:300; 44362G; Thermo Fisher), or anti-PML (1:1,000; A301-167A; Bethyl Laboratories) as described above. Speckles were quantified using ImageJ software.

Affymetrix gene array analysis.

Lentivirus-transduced hBMECs constitutively expressing ZIKV NS5 WT, K252R, or NLS mutants or control hBMECs were lysed, and total RNA was purified using RNeasy (Qiagen). Purified RNA was quantitated, and transcriptional responses were detected with Affymetrix Clariom-S chip arrays at the Stony Brook Genomics Core Facility. Transcriptional responses of ZIKV NS5 WT-expressing cells were compared to control cells and to ZIKV NS5 K252R. Fold changes in RNA transcript levels were analyzed using Affymetrix TAC software.

qRT-PCR analysis.

hBMECs transduced to constitutively express ZIKV NS5 WT, K252R, NLS mutant, DENV3 NS5 WT, and GFP (1 × 10⁶ cells) were stimulated with 1,000 U/ml IFN-α for 30 min, and cells were harvested after 24 h. Alternatively, unstimulated transduced hBMECs were harvested to analyze basal gene expression. RNA purification was performed as described above. cDNA synthesis was performed using a Transcriptor first-strand cDNA synthesis kit (Roche) using random hexamers as primers (25°C for 10 min, 50°C for 60 min, and 90°C for 5 min). qRT-PCR primers for specific genes were designed according to the NCBI gene database with 60°C annealing profiles. Genes were analyzed using PerfeCTa SYBR green SuperMix with ROX (Quanta Biosciences) on an ABI 7300 real-time PCR system (Applied Biosystems). Responses were normalized to internal glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA levels, and the fold induction was calculated using the 2^−ΔΔCT method.

Western blotting.

hBMECs were infected with ZIKV (MOI, 2) or mock infected and harvested at 12 to 24 hpi. Alternatively, transduced hBMECs constitutively expressing ZIKV NS5 WT, K252R, NLS mutant, and GFP (1 × 10⁶ cells) were stimulated with 1,000 U/ml IFN-α for 1 h. Cells were washed with PBS followed by lysis in buffer containing 1% NP-40 (150 mM NaCl, 50 mM Tris-Cl, 10% glycerol, 2 mM EDTA, 10 nM sodium fluoride, 2.5 mM sodium pyrophosphate, 2 mM sodium orthovanadate, 10 mM β-glycerophosphate) with protease inhibitor cocktail (Sigma). Total protein levels were determined in a bicinchoninic acid assay (Thermo Scientific), and 20 μg of protein was resolved by SDS–10% polyacrylamide gel electrophoresis. Proteins were transferred to nitrocellulose, blocked in 5% bovine serum albumin, and incubated with the indicated antibodies. Antibodies used were anti-NS5 (GTX133327; GeneTex), anti-PML (A301-167A; Bethyl Laboratories), anti-FLAG (2368; Cell Signaling), anti-STAT2 (72604; Cell Signaling), anti-pSTAT2 (88410; Cell Signaling), anti-IRF9 (76684; Cell Signaling), anti-IFIT1 (Santa Cruz Biotechnology), and anti-GAPDH (G9545; Sigma-Aldrich). Protein was detected using horseradish peroxidase (HRP)-conjugated anti-rabbit secondary antibody (Amersham) and Luminata Forte Western HRP substrate (Millipore).

In vitro SUMOylation assay.

HeLa cells were cotransfected with FLAG-tagged NS5 and 6×His-tagged SUMO-expressing plasmids using PEI. At 22 to 30 h posttransfection, SUMO conjugates were captured by nickel affinity chromatography (Ni²⁺-NTA agarose; Qiagen) under denaturing conditions and analyzed as described previously (70).

Coimmunoprecipitation.

Transduced hBMECs constitutively expressing ZIKV NS5 WT, K252R, NLS mutant, and GFP for 48 h were washed with PBS, followed by lysis in buffer containing 1% NP-40. Anti-FLAG M2 affinity resin (F2426; Millipore Sigma) was used to immunoprecipitate NS5 constructs. Samples were washed 3 times in lysis buffer, resuspended in SDS sample buffer, separated by 10% SDS-PAGE, and analyzed by Western blotting as described above.

Statistical analysis.

Results shown in each figure were derived from two to three independent experiments with comparable findings; the data presented are means ± standard errors of the means (SEM), with the indicated P values of <0.05, 0.01, and 0.001 considered significant. Two-way or one-way comparisons were performed using analysis of variance (ANOVA). All analyses were performed using GraphPad Prism software version 4.0.

Data availability.

Data obtained from these studies were submitted to the NCBI Gene Expression Omnibus database (GSE147926).