Zika Virus Persistently Infects and Is Basolaterally Released from Primary Human Brain Microvascular Endothelial Cells

1Zika virus (ZIKV) is the cause of encephalitis and fetal microcephaly outbreaks in Brazil and Puerto Rico. ZIKV is transmitted by mosquitoes, and ZIKV is predicted to cause >4 million clinical cases per year in the Americas. Unlike other mosquito-borne flaviviruses (FVs), ZIKV has unique properties that result in persistence (1 to 6 months) and sexual transmission. ZIKV infects fetal and maternal cells, including neural stem cells, astrocytes, neuronal progenitor cells, microglia, Hofbauer cells, and endothelial cells (ECs). ZIKV crosses placental and blood-brain barrier (BBB) ECs, cells which normally prevent mixing of maternal and fetal blood and restrict access to adult and fetal neuronal compartments.

2Mosquito-borne FVs primarily cause acute febrile infections that are transient and resolve in ~2 weeks. Approximately 80% of FV infections are asymptomatic, with neurologic symptoms in <1% of Japanese encephalitis virus (JEV) and West Nile virus (WNV) cases and in an even smaller subset of dengue virus (DENV)-infected individuals. JEV causes more severe disease in children, and WNV encephalitis occurs primarily in elderly or inmunocompromised individuals. DENVs have striking homology to ZIKV, cocirculate in humans and mosquitoes, but cause fever, hemorrhagic fever, and shock syndrome that are not central nervous system (CNS) associated and are instead linked to vascular leakage and EC dysfunction. Preexisting anti-FV immunity was recently found to enhance ZIKV pathogenesis in mice; however, among FVs, only ZIKV causes fetal demise and microcephaly during human infection.

3Hepatitis C virus (HCV) is a divergent blood-borne FV that persistently infects human hepatocytes and is transmitted from person to person. Mechanisms of HCV persistence remain enigmatic but require balanced HCV replication in hepatocytes without cytopathology, as well as evasion of innate and adaptive immune responses that normally limit acute RNA virus infections. In humans, ZIKV is present in patient blood, tears, saliva, semen, and urine for <6 months. Cytotoxicity associated with ZIKV infection of neurons, neuronal progenitor cells, and trophoblasts suggests that discrete cellular reservoirs enable acytopathic ZIKV replication and persistence. Although ZIKV does not establish HCV-like permanence, ZIKV persistence is unique to mosquito-borne FVs and likely to be a key to the ability of ZIKV to be sexually transmitted and enter neuronal compartments.

4Innate and adaptive immune responses normally restrict acute FV replication and spread. Pretreating cells with type I interferon (IFN-α/β) blocks FV infection; however, FVs have developed mechanisms to restrict antiviral IFN-α/β, IFN regulatory factors (IRFs), and IFN-α receptor (INFAR) signaling defenses. In murine settings, ZIKV fails to cause disease unless adapted or grown in immunocompromised (SJL), IRF−/−, or IFNAR-deficient (A129) mice. In these settings, ZIKV replicates to high titers, infects neurons, persists in tears, and causes optic damage, neurologic symptoms, placental damage, and fetal demise. ZIKV infection of immunocompetent macaques results in persistent viremia (for <57 days in pregnant animals) with ZIKV present in cerebrospinal fluid, lymph nodes, saliva, and urine, similar to ZIKV persistence in humans.

5ECs serve a primary barrier function that normally limits the access of blood constituents to the privileged neuronal, retinal, testicular, and placental compartments. Brain capillaries form a BBB composed of unique brain microvascular ECs (BMECs), which restrict the emigration of immune cells and viruses to neuronal compartments. Evidence that the BBB serves as a viral barrier is clear from findings that nonneurovirulent viruses injected into murine brains cause lethal neuronal pathology. ZIKV infects patient ECs and human ECs derived from the aorta, brain, and lymphatic and umbilical vessels, suggesting that brain ECs may serve as potential conduits for ZIKV to bypass BBB restrictions.

6Sites of ZIKV persistence and spread remain to be defined but appear central to the ability of ZIKV to breach barriers that normally restrict viral access to brain, testicular, and fetal tissues. We found that ZIKV (PRVABC59) persistently infects primary human BMECs (hBMECs) for >9 days postinfection (dpi) or following passage in hBMECs, without cytopathic effects. ZIKV was released from basolateral and apical surfaces of polarized hBMECs, but fail to alter hBMEC permeability. ZIKV infection of hBMECs was resistant to IFN-α added at >3 hpi, and IFN-β and IFN-λ were transiently induced but absent from cell supernatants for 1 to 9 dpi. Transcriptome analysis revealed that the chemokine CCL5 was highly induced and secreted by ZIKV-infected hBMECs and that infected hBMECs persistently induced and expressed cellular ISGs, including IFIT1 and MxA, as well as ISG inducers IRF7 and IRF9. ZIKV induced prosurvival factors and apoptosis regulatory genes ATF3, EGR1, IAP-2, and XAF1 and factors associated with HCV persistence, ISG15, HERC5, and USP18. Our findings revealed that ZIKV persists in hBMECs by evading innate antiviral IFN and ISG responses and by uniquely balancing viral replication and cell survival in hBMECs. Persistently infected hBMECs provide a ZIKV reservoir that is uniquely situated to permit systemic ZIKV spread and basolateral emigration into neuronal compartments.

7ECs present in placental, fetal, and brain capillaries are infected in ZIKV patients, and ZIKV transmission to fetal neurons requires the virus to cross placental and brain EC barriers. hBMECs form a protective blood-brain barrier that restricts viral access to neuronal compartments, and mechanisms by which ZIKV bypasses the BBB remain to be defined. We initially assessed the ability of early-passage ZIKV (PRVABC59) to infect primary hBMECs (passages 3 to 10) by immunoperoxidase staining ZIKV-infected cells. We found that hBMECs were 80% infected by 12 to 24 hpi and remained infected for 1 to 3 dpi (Fig. 1A). ZIKV titers, RNA levels, and numbers of infected cells increased from 12 to 48 hpi (Fig. 1B to D), and inoculation of hBMECs with ZIKV at low multiplicities of infection (MOIs) of 0.1 to 1 resulted in viral spread within hBMEC monolayers for 1 to 3 dpi. ZIKV reached titers of 1 × 106 focus-forming units (FFU)/ml in hBMECs (Fig. 1B), similar to ZIKV grown in IFN locus-deficient Vero E6 cells or a telomerase-immortalized human cerebral MEC (hCMEC/D3) cell line (see Fig. S1B and C in the supplemental material).

8The ability of ZIKV to replicate and spread in hBMECs, similar to infection of Vero E6 cells, suggested that ZIKV may potently regulate antiviral IFN responses. We added IFN-α exogenously at various times before or after ZIKV infection of hBMECs and Vero E6 cells (which are IFN-α responsive) and quantitated ZIKV-infected cells 24 hpi. Compared to untreated cells, prior addition of IFN-α inhibited ZIKV infection of both hBMECs and Vero E6 cells (Fig. 1E). Prior IFN-α addition also blocked ZIKV infection of human umbilical vein endothelial cells (HUVECs) and the hCMEC/D3 cell line (Fig. S1D). While simultaneous addition of IFN-α and ZIKV to Vero E6 cells blocked ZIKV infection, effects on ZIKV infection when IFN was added 3, 6, or 12 hpi were sequentially blunted (Fig. 1F; Fig. S1A). In contrast, hBMECs coadministered IFN-α and virus showed reduced ZIKV infection by ~70%, while IFN-α addition 3 to 6 hpi only reduced infected hBMECs by 10 to 20% (Fig. 1G). These findings indicate that ZIKV infection is inhibited by prior addition of IFN-α, but that by 3 hpi ZIKV-infected hBMECs are highly resistant to the effects of exogenously added type I IFN.

9ZIKV pathogenesis is linked to its capacity to damage neurons, and the cytopathic effect (CPE) of ZIKV on Vero cells is well established. However, we observed little, if any, cytopathology in ZIKV-infected hBMECs 1 to 3 dpi (~80% cells were ZIKV infected) (Fig. 1). We compared Vero E6 and hBMEC cell survival following ZIKV infection in propidium iodide (PI)/calcein-AM and CyQuant viability assays. We found that ~50% of ZIKV-infected Vero E6 cells were PI positive (dead) 2 to 3 dpi, with a dramatic reduction in viable cell uptake of calcein-AM (Fig. 2A). In contrast, ZIKV-infected hBMECs were nearly all PI negative and viable based on calcein-AM uptake 2 to 3 dpi (Fig. 2A). Similar viability was observed following ZIKV infection of the hCMEC/D3 cell line (Fig. S2A). CyQuant analysis confirmed that ZIKV-infected Vero E6 cells and hBMECs were, respectively, 55% and >95% viable 2 to 3 dpi (Fig. 2B). These findings indicate that ZIKV fails to cause cytopathic effects in hBMECs.

10Given the survival of ZIKV-infected hBMECs and the continued spread of ZIKV within infected hBMECs, we analyzed the ability of ZIKV to persistently infect hBMECs. We replenished hBMEC culture medium every 3 days for infected or uninfected monolayers and evaluated viral titers and hBMEC viability 2 to 9 dpi. Consistent with ZIKV persistence in hBMECs, we observed ZIKV titers (~1 × 106/ml) and RNA levels 9 dpi that were similar to those observed at 2 to 3 dpi (Fig. 3A and B), as well as expression of ZIKV envelope protein in hBMEC lysates (Fig. 3C). ZIKV-infected and mock-infected hBMECs were similarly PI negative, calcein-AM positive, and ~100% CyQuant viable 9 dpi (Fig. 3D and E). In contrast, ZIKV-infected Vero E6 cells (9 dpi) were nearly all PI positive, calcein-AM negative, and only ~10% viable by CyQuant analysis (Fig. 3D and E). ZIKV also persistently infected HUVECs and hCMEC/D3 cells 9 dpi based on immunoperoxidase staining, with no apparent CPE (Fig. S2B). These findings indicate that ZIKV persistently infects hBMECs without significantly diminishing hBMEC viability.

11ZIKV antigen-positive Vero E6 cells adhere as monolayers regardless of their viability. However, trypsinizing ZIKV-infected Vero E6 cells resulted in <50% of cells being able to reattach to monolayers, and virtually all cells were PI positive after 3 days and unable to be passaged further (Fig. 4A and C). In contrast, trypsinization of ZIKV-infected hBMECs resulted in the propagation of viable ZIKV antigen-positive hBMECs through >3 passages (PI negative, calcein-AM positive) (Fig. 4B and C). Passage of highly infected hBMECs resulted in continued division and the production of ZIKV titers of ~2 to 5 × 106/ml (Fig. 4D) with little or no cytopathology, similar to results in control hBMECs (Fig. 4C). Thus, ZIKV-infected hBMECs are viable following cellular passage and persistently produce ZIKV progeny.

12Persistent ZIKV infection and spread suggests that ZIKV regulates antiviral responses and fosters hBMEC survival. We kinetically analyzed global transcriptional responses of hBMECs to synchronous ZIKV infection (80% infected at 1 dpi) (Fig. 1A). Human gene-level transcriptome arrays (Affymetrix, with >20,000 genes) were used to quantitate ZIKV-induced transcriptional changes in hBMECs compared to responses in mock-infected hBMECs. Selected, transcriptional changes are presented in Table 1, with complete data files available in the NCBI GEO database (GSE98889). ZIKV-infected hBMECs transiently induced IFN-β and IFN-λ at 1 to 2 dpi, and a subset of ISGs was highly induced 1 to 9 dpi (Table 1). Consistent with IFN-λ and ISG induction, IRF1 was transiently induced (1 to 2 dpi) while IRF7 and IRF9 were constitutively induced 1 to 9 dpi. ISG15, ISG15 family ligases (HERC5/6), and ISG15 proteases (USP18/41), which are implicated in HCV persistence, were also induced 1 to 9 dpi (Table 1). Transcription factors ATF3, EGR1, and TRAF1, which are linked to cell survival, were upregulated 39- to 301-fold (1 to 2 dpi), and ATF3 was induced 1 to 9 dpi. ZIKV also induced apoptosis-inhibitory and -regulating factors IAP2 and XAF1 5- to 201-fold (1 to 9 dpi) (Table 1). Two genes associated with cell permeability, Rnd1, a constitutively activated Rho GTPase, and ARHGAP26, a Rho-activating factor, were also induced in ZIKV-infected hBMECs. Chemokine CCL5/RANTES was highly induced (40- to 2,327-fold) 1 to 9 dpi, along with a subset of moderately induced CXCL10/11 and CCL20 chemokines (Table 1) (GEO GSE98889). Selected genes identified as upregulated by array were transcriptionally verified 1 to 9 dpi based on quantitative real-time PCR (qRT-PCR) analysis (Fig. S3). ZIKV-directed hBMEC transcriptional responses suggest that ZIKV evades antiviral IFN and ISG responses while engaging cell survival genes and pathways to establish persistence in hBMECs.

13Analysis via an enzyme-linked immunosorbent assay (ELISA) determined that CCL5 was highly secreted into the supernatants of ZIKV-infected hBMECs at all time points from 1 to 9 dpi (Fig. 5A). Despite high levels of IFN-β secreted by hBMECs in response to poly(I/C) (Fig. 5B), we failed to detect IFN-β or IFN-λ secretion from ZIKV-infected hBMECs 1 to 9 dpi (Fig. 5B and C). In contrast, induced ISGs MxA and IFIT1 were expressed in ZIKV-infected hBMECs 2 to 9 dpi (Fig. 5D) and at levels similar to ISG expression levels induced by IFN-α treatment of hBMECs (Fig. 5E). Distinct from CCL5 secretion, these findings suggest that ZIKV uniquely regulates IFN-β/λ expression or secretion posttranscriptionally and further demonstrate that ZIKV evades antiviral functions of highly induced and expressed ISGs to persist in hBMECs.

14hBMECs form a BBB that prevents paracellular permeability and that in vivo restricts access of blood constituents to neuronal compartments. We evaluated changes in the barrier function of hBMECs following ZIKV infection by assessing the transendothelial electrical resistance (TEER) and fluorescein isothiocyanate (FITC)-dextran permeability of hBMEC monolayers grown on Transwell inserts. We found no significant change in TEER of ZIKV-infected versus mock-infected hBMECs at 1 to 3 dpi (Fig. 6A). After establishing that Transwell monolayers were intact, we disrupted paracellular hBMEC junctions with EDTA and found an ~100-Ω decrease in the TEER of hBMEC monolayers. Consistent with the TEER findings, the permeability of hBMECs to FITC-dextran was not enhanced by ZIKV infection of hBMECs compared to responses of mock-infected hBMEC controls (Fig. 6B). Collectively, these findings indicate that the barrier integrity and permeability of hBMECs is not significantly altered by ZIKV infection.

15hBMECs form polarized monolayers with apical and basolateral surfaces that mimic lumenal and ablumenal capillary surfaces. In order to assess ZIKV entry and release from polarized hBMECs, we grew hBMECs for 5 days on Transwell inserts, infected hBMECs with ZIKV from the apical or basolateral sides, and assessed ZIKV titers in the upper and lower chambers at 1 dpi. ZIKV infection of either the apical or basolateral surfaces resulted in ZIKV release from both apical and basolateral hBMEC surfaces (Fig. 6C), while inoculation controls and TEER revealed no leakage across hBMEC monolayers. These results suggest that ZIKV infects and is released from both the lumenal and ablumenal sides of hBMECs. This observation provides a potential mechanism for ZIKV to cross hBMEC barriers and spread to neuronal compartments (Fig. 6D).

16ZIKV is distinguished from other mosquito-borne FVs by its unique ability to be sexually transmitted, cross placental and blood-brain barriers, cause in utero microencephaly, and persist for up to 6 months in patients. During human infections, ZIKV is found in a variety of body fluids (tears, saliva, semen, cervical mucus, and urine) that facilitate viral replication and dissemination, and ZIKV damages placental, corneal, and neuronal tissues. However, beyond the acute phase of infection, its persistence suggests that ZIKV uniquely replicates in cellular reservoirs where it balances cell survival and viral replication and evades innate and adaptive immune responses for extended periods. Persistence alone is likely to facilitate the ability of ZIKV to be sexually transmitted, spread across the placenta, and gain access to fetal and neuronal tissues. Our findings indicate that ZIKV infects primary hBMECs without the cytopathology that is reported for ZIKV-infected neurons and placental tissues. Cell death, observed following ZIKV infection of Vero E6 cells, was noticeably absent in ZIKV-infected hBMECs even at 9 dpi and after serial passage of infected hBMECs. This suggests the potential for persistently infected hBMECs to serve as cellular reservoirs for ZIKV replication and enable viral spread across BBBs into neuronal compartments.

17Permeabilizing the endothelium is one mechanism for viruses to bypass EC barriers. We found that ZIKV induced Rnd1 and ARHGAP26Rho in infected hBMECs, which direct brain capillary permeability. These findings suggested potential mechanisms for ZIKV to permeabilize and spread across the BBB. However, when we evaluated hBMEC permeability, we found no significant difference between control and ZIKV-infected hBMEC monolayers. Instead, we found that ZIKV exited from apical and basolateral surfaces of hBMECs, suggesting a discrete mechanism for ZIKV to enter neuronal compartments via basolateral release and for systemic spread of ZIKV that is released apically from hBMEC reservoirs. Infection of HUVECs and other EC sites also provides prospective mechanisms for ZIKV to cross testicular, placental, and retinal EC barriers in ZIKV patients.

18Successful viral pathogens have evolved mechanisms to evade antiviral innate immune responses. We found that ZIKV infection of primary hBMECs was inhibited by prior IFN-α addition, but that ZIKV-infected hBMECs were highly resistant to IFN-α by 3 hpi. In addition, ZIKV continued to spread within hBMECs for 1 to 9 dpi without apparent paracrine IFN restriction. In contrast, DENV spread in HUVECs was blocked by IFN-β secreted from DENV-infected cells (1 to 3 dpi) and relieved by an IFN-β antibody blockade. ZIKV’s resistance to IFN and its spread in hBMECs suggests that ZIKV employs discrete mechanisms to regulate innate hBMEC responses.

19ZIKV-directed changes in hBMECs were initially addressed by kinetically analyzing global transcriptional responses to synchronous ZIKV infections. IFN-β and IFN-λ were transiently induced 1 to 2 dpi but, consistent with ZIKV persistence and spread in hBMECs, we failed to detect secreted IFNs in cell supernatants (Fig. 5). The absence of IFN does not appear to be due to altered secretion, as the CCL5/RANTES chemokine was both induced and highly secreted by ZIKV-infected hBMECs, and hBMECs were fully competent in secreting high levels of IFN-β in response to double-stranded RNA (Fig. 5B). In contrast, a subset of antiviral ISGs were highly induced and constitutively expressed in infected hBMECs 1 to 9 dpi (Table 1). In the absence of type I IFN secretion, ISG induction might be explained by the activation of existing IRFs or by ZIKV inducing IRF7 and IRF9 1 to 9 dpi (Table 1), as IRF9 reportedly induces ISGs, including ISG15, without IFN and IFNAR-directed STAT1/2 phosphorylation.

20ZIKV regulation of IFN induction and INFAR signaling pathways has been reported, and both ZIKV and DENV NS5 proteins regulate IFNAR-directed STAT2 signaling responses. STAT2 regulation may contribute to ZIKV resistance to added IFN-α, but this fails to explain ZIKV spread in partially infected monolayers, unless IFN expression is also inhibited. Our findings are supported by a recent study suggesting that ZIKV (PRVABC59) translationally restricts type I IFN in primary human dendritic cells (DCs). However, findings that only ~8% of DCs were infected and that an INFAR antibody blockade increased ZIKV titers by 300% suggest that IFN is expressed and that ZIKV spread is severely restricted in DCs. Potential mechanisms of posttranscriptional IFN regulation during ZIKV infection remain to be defined and factored into an understanding of ZIKV persistence in hBMECs.

21In vitro, ZIKV productively infects and induces ISGs in DCs, skin cells, macrophages, and trophoblasts. Comparison of transcriptional results across studies is complicated by the different ZIKV strains used and in some studies analysis of cells showed that only 2 to 10% ZIKV infected, potentially amplifying paracrine IFN-directed ISG responses of bystander cells. In synchronously infected hBMECs (>80% infected), a large subset of antiviral ISGs were induced 1 to 9 dpi, that are seemingly at odds with ZIKV persistence in hBMECs. IFIT1, which interferes with Alphavirus translation, was highly induced and expressed in ZIKV-infected hBMECs without apparent consequence to ZIKV replication or persistence. In contrast, IFITM1, which is suggested to inhibit early ZIKV infection, was induced at very low levels by ZIKV while IFITM3, which is reported to prevent ZIKV-induced HeLa cell death, was not induced in ZIKV-infected hBMECs. These findings suggest that ZIKV employs novel mechanisms that selectively restrict antiviral ISG responses in order to persistently replicate and spread in hBMECs.

22The lack of cytopathology during persistent ZIKV infection of hBMECs indicates that cell survival responses are engaged and that apoptotic responses are restricted in ZIKV-infected hBMECs. Several prosurvival genes are induced during ZIKV infection of hBMECs, including EGR1, ATF3, and BIRC3, and the constitutively induced chemokine CCL5 (40- to 2,327-fold 1 to 9 dpi) is also associated with cell survival. In contrast, IRF1, which directs apoptosis when expressed constitutively, was only transiently induced, while XAF1, a protein that inactivates inhibitor of apoptosis proteins (IAPs; BIRC2/3), was continuously induced along with its regulatory target, IAP2. Additional prosurvival responses may also be directed by ZIKV-engaging Axl receptors, as Axl performs proliferative and cell survival functions and is linked to IFNAR regulation. These findings suggest that ZIKV establishes persistence by balancing induced apoptotic and cell survival responses in hBMECs.

23HCV replicates at low levels in persistently infected hepatocytes by inducing an IFN-resistant state and promoting cell survival. With HCV, ISG15 induction is actually proviral and contributes to HCV persistence and IFN resistance in hepatocytes. ISG15 knockdown renders HCV-infected cells susceptible to IFN regulation, and human cells devoid of ISG15 have enhanced antiviral protection that is lost by expressing ISG15. Further proviral ISG15 functions stem from human ISG15 interactions, which are crucial for USP18-mediated inhibition of IFN-α/β signaling responses. Consistent with prosurvival roles for ISG15, ZIKV-infected hBMECs constitutively induce ISG15, ISG15 ligases (HERC5/6), and ISG15 proteases (USP11/18/41). ZIKV induction of the ISG15 family further suggests the potential for USP18-directed downregulation of IFN-α/β translation and for HERC5 to enhance IRF stability that may drive ISG induction during ZIKV infection. Whether ZIKV cytopathology is linked to the absence of ISG15 in Vero cells remains to be determined, as do proviral or IFN regulatory roles for ISG15 in ZIKV-infected hBMECs.

24CCL5 is a chemokine linked to antiapoptotic responses, and CCL5 was the only chemokine highly induced and constitutively secreted by ZIKV-infected hBMECs (1 to 9 dpi). CCL5 responses are directed by WNV, DENV, and Tick-borne encephalitis virus (TBEV) infection of brain ECs, and CCL5 induction is associated with increased WNV and TBEV pathogenesis. On the EC surface, CCL5 forms a filamentous complex that is key to the recruitment of immune cells and activated CD4+ T cells to human BMECs. Although the half-life of CCL5 on the surface of ECs is only 30 min, constitutive CCL5 secretion during persistent ZIKV infection of hBMECs may traffic immune cells to the BBB, and both foster ZIKV’s spread to neurons and inflammatory CNS pathology. Further studies are required to determine whether transcytosis is an additional means of ZIKV crossing hBMEC barriers.

25RNA viruses which lack DNA intermediates are rarely associated with extended or persistent infections. Measles and rubella viruses are examples of RNA viruses that persist and cause neuronal disease, but mechanisms of persistence remain perplexing. While ZIKV infection does not establish HCV-like permanence, ZIKV persistence in hBMECs without cytotoxicity suggests that hBMECs are potential ZIKV reservoirs that foster viral spread and pathogenesis. Nevertheless, ZIKV may also persist in additional cell types and compartments. Recent studies suggested that human placental macrophages (Hofbauer cells) and DCs are 90% viable after ZIKV (PRVABC59) infection. However, it remains to be determined whether ZIKV directs survival responses or persistently infects these cells, as only 5% of Hofbauer cells and 2 to 8% of DCs were infected in these studies. In addition to discovering other persistently infected cell types, studies of ZIKV evasion of adaptive immune responses are required in order to understand ZIKV persistence mechanisms in vivo. A recent in vivo study correlated persistent ZIKV infection of rhesus monkey CNS and lymph nodes with transcriptomic responses of peripheral blood mononuclear cells. This study noted the induction of ATF3, EGR1, JUNB, IRF7, ISG15, HERC5/6, additional ISGs, chemokines CCL5 and CXCL10, prosurvival responses, and decreased proapoptotic genes, similar to gene induction levels we found to be induced by ZIKV infection of hBMECs (Table 1) (GEO GSE98889).

26Strategic roles for ECs in ZIKV persistence and spread are suggested by the wide range of restricted tissues targeted by ZIKV. ECs control immune cell recruitment, targeting, and transcytosis into tissues and perform primary fluid barrier functions that prevent capillary leakage. ECs innervate and protect immune-restricted compartments from viral spread, and emigration across ECs is required for ZIKV to bypass blood-retina barriers and blood-testicle barriers to infect retinal ganglion cells, aqueous humor, the cornea, Sertoli cells, seminiferous tubules, and spermatagonia. ECs express tolerizing PD-L1 receptors that prevent immune targeting of the endothelium and which may foster viral immune evasion and persistence. Our findings define hBMECs as ZIKV reservoirs that permit lumenal emigration and spread within blood and potentially provide ZIKV the ability to cross brain EC barriers and enter neuronal compartments. Remarkably, ZIKV persists in hBMECs with little or no cytopathology through ZIKV-directed cell survival responses. These findings suggest mechanisms for ZIKV persistence as well as potential hBMEC targets for restricting ZIKV persistence and spread.

27Vero E6 cells (ATCC CRL 1586) 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). HUVECs were purchased from Cambrex, and the immortalized hCMEC/D3 cell line was purchased from Cedarlane Labs. Inc. hBMECs (passages 4 to 10), HUVECs (passages 3 to 8), and hCMEC/D3 cells were grown in endothelial cell basal medium-2 MV (EBM-2 MV; Lonza) supplemented with EGM-2 MV SingleQuots (Lonza) and incubation at 37°C and 5% CO2.

28ZIKV (PRVABC59) was obtained from the ATCC, minimally passaged (MOI, 0.1 to 1), and propagated for 5 days in Vero E6 cells in DMEM with 2% FBS. ZIKV was allowed to adsorb to ~60% confluent hBMEC monolayers for 2 h. Following adsorption, monolayers were washed with phosphate-buffered saline (PBS) and grown in supplemented EBM-2 MV with 5% FBS. Vero E6, HUVEC, and hCMEC/D3 cells were similarly infected, washed, and supplemented with DMEM with 8% FBS. ZIKV titers were determined by serial dilution and infection of Vero E6 cells, quantifying infected cell foci at 24 hpi by immunoperoxidase staining with anti-DENV4 hyperimmune mouse ascites fluid (HMAF; 1:1,200; ATCC), horseradish peroxidase (HRP)-labeled anti-mouse IgG (1:2,000; KPL-074-1806), and 3-amino-9-ethylcarbazole staining. For IFN-α inhibition studies, medium was supplemented with 1,000 U/ml IFN-α (Sigma-Aldrich) at indicated times and cells incubated at 37°C and 5% CO2. For analysis of apical and basolateral release of ZIKV, hBMECs were grown for 5 days on Costar Transwell inserts (3 μm) and evaluated for confluence by T (EVOM2, STX3; World Precision Instruments, Inc.). hBMECs were ZIKV infected apically or basolaterally (MOI, 5), washed with PBS and the medium, and incubated for 24 h prior to analyzing viral titers in apical and basolateral supernatants.

29ZIKV (MOI, 10) or mock-infected hBMECs or Vero E6 cells were assayed for viability using a CyQuant NF cell proliferation assay kit (Thermo-Fisher Scientific). Medium was removed from mock- or ZIKV-infected monolayers (MOI, 10) at the indicated day postinfection and replaced with CyQuant reaction mixture. Wells were incubated for 15 min before fluorescence was quantified using a BioTek FLx800 fluorimeter (490-nm excitation, 530-nm emission). Fluorescence units of ZIKV-infected versus mock-infected cells were compared to determine the percent fluorescence representing relative cell viability.

30PI (Calbiochem) and calcein-AM (Invitrogen) uptake were used to evaluate hBMEC and Vero E6 cell viability. ZIKV-infected (MOI, 10) or mock-infected hBMECs were seeded into 96-well plates and at the indicated times costained by the membrane-permeable dye calcein-AM (3 μM; green fluorescence in live cells), and 2.5 μM propidium iodide (red fluorescent DNA stain to detect dead cells). Images of calcein-AM-positive versus PI-positive cells were resolved using an Olympus IX51 microscope and Olympus DP71 camera and overlayed using Adobe Photoshop.

31Primary hBMECs (passages 5 to 10; Cell Systems) were synchronously infected with ZIKV (PRVABC59) (MOI, 20) or mock infected. Mock- or ZIKV-infected hBMECs were lysed, and total RNA was purified 12 hpi to 9 dpi by using RNeasy (Qiagen). Purified RNA was quantitated, and transcriptional responses were detected with Affymetrix Clariom-S chip arrays in the Stony Brook Genomics Core Facility. ZIKV-infected cell transcriptional responses were compared to those in mock-infected cells harvested at each time pointm and fold changes in ZIKV versus control hBMEC transcripts were analyzed by using the Affymetrix TAC software. Data obtained from these studies were submitted to the NCBI Gene Expression Omnibus database (GEO GSE98889).

32Quantitative real-time PCR was performed on purified RNAs from mock- or ZIKV-infected hBMECs 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 (provided by Operon). qRT-PCR primers to ZIKV (PRVABC59) RNA were as follows: forward, CCGTGCCCAACACAAG; reverse, CCACTAACGTTCTTTTGCAGACAT. 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 for the difference between mock- and ZIKV-infected hBMEC RNA levels at each time point.

33hBMECs were plated on Costar Transwell inserts (3-μm pore size; Corning) at high density, and 5 days postseeding hBMEC monolayers were analyzed for TEER (EVOM2; STX3; World Precision Instruments, Inc.). Confluent Transwell cultures were infected with ZIKV (MOI, 5) or mock infected, and TEER values for ZIKV-infected versus mock-infected hBMECs were compared at 1 to 3 dpi. To assess Transwell monolayer resistance, 10 mM EDTA was added to apical supernatants for 10 min, and TEER was assessed before and after this addition. EDTA treatment reduced TEER by ~100 Ω, demonstrating the resistance change directed by disrupting interendothelial cell junctions.

34A gold standard Transwell permeability assay was used to assess the permeability of ZIKV-infected hBMECs on Costar Transwell plates in triplicate. Transwells were seeded and infected (MOI, 5) or mock infected as described above. At 3 dpi, FITC-dextran (40 kDa; 0.5 mg/ml; Sigma) was added to the upper chamber and the levels of FITC-dextran observed in the lower chamber were monitored over time by using a BioTek FLx800 fluorimeter (490-nm excitation, 530-nm emission). FITC-dextran fluorescence directed by transit across ZIKV-infected or mock-infected hBMECs was quantitated. The data presented represent results of three independent experiments.

35Levels of IFN-β, IFN-λ, and CCL5/RANTES in the supernatants of mock- and ZIKV-infected hBMECs at 12 hpi to 9 dpi were measured using a DuoSet ELISA (R&D Systems). hBMECs transfected with poly(I/C) (1 µg/ml) and Fugene6 (3:1; Promega) were similarly evaluated for secreted IFN-β at 36 h posttransfection. ELISA plates (Immunolon 2, U-bottom; Dynatech Laboratories) were coated with anti-IFN-β, anti-IFN-λ, or anti-CCL5 ELISA reagents according to the manufacturer. Viral supernatants were incubated on coated plates (2 h) and washed with PBS (0.1% Tween 20), and bound protein was detected with target-specific antibodies conjugated to streptavidin-HRP and developed using tetramethylbenzidine. Protein concentrations were determined based on the optical density, using a BioTek EL312e microplate reader (450 nm), and compared to standard curves for purified IFN-β, IFN-λ, and CCL5 (R&D Systems).

36Western blot assays were performed as previously described. Briefly, MECs were infected with ZIKV (MOI, 10) or mock infected and harvested at 1 to 9 dpi. 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–12% polyacrylamide gel electrophoresis. Proteins were transferred to nitrocellulose, blocked in 5% bovine serum albumin, and incubated with the indicated antibodies. Antibodies used were anti-ZIKV envelope (GTX133314; GeneTex), anti-IFIT1 (Santa Cruz Biotechnology), anti-MxA (D3W71; Cell Signaling), and anti-GAPDH (G9545; Sigma-Aldrich). Protein was detected using HRP-conjugated anti-mouse or anti-rabbit secondary antibodies (Amersham) and Luminata Forte Western HRP substrate (Millipore).

37Results 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 <0.01 and <0.001 considered significant. Two-way comparisons were performed two-tailed analysis of variance and an unpaired Student’s t test. All analyses were performed using GraphPad Prism software version 4.0.