Flavivirus NS1 Triggers Tissue-Specific Disassembly of Intercellular Junctions Leading to Barrier Dysfunction and Vascular Leak in a GSK-3β-Dependent Manner

The flavivirus nonstructural protein 1 (NS1) is secreted from infected cells and contributes to endothelial barrier dysfunction and vascular leak in a tissue-dependent manner. This phenomenon occurs in part via disruption of the endothelial glycocalyx layer (EGL) lining the endothelium. Additionally, we and others have shown that soluble DENV NS1 induces disassembly of intercellular junctions (IJCs), a group of cellular proteins critical for maintaining endothelial homeostasis and regulating vascular permeability; however, the specific mechanisms by which NS1 mediates IJC disruption remain unclear. Here, we investigated the relative contribution of five flavivirus NS1 proteins, from dengue (DENV), Zika (ZIKV), West Nile (WNV), Japanese encephalitis (JEV), and yellow fever (YFV) viruses, to the expression and localization of the intercellular junction proteins β-catenin and VE-cadherin in endothelial cells from human umbilical vein and brain tissues. We found that flavivirus NS1 induced the mislocalization of β-catenin and VE-cadherin in a tissue-dependent manner, reflecting flavivirus disease tropism. Mechanistically, we observed that NS1 treatment of cells triggered internalization of VE-cadherin, likely via clathrin-mediated endocytosis, and phosphorylation of β-catenin, part of a canonical IJC remodeling pathway during breakdown of endothelial barriers that activates glycogen synthase kinase-3β (GSK-3β). Supporting this model, we found that a chemical inhibitor of GSK-3β reduced both NS1-induced permeability of human umbilical vein and brain microvascular endothelial cell monolayers in vitro and vascular leakage in a mouse dorsal intradermal model. These findings provide insight into the molecular mechanisms regulating NS1-mediated endothelial dysfunction and identify GSK-3β as a potential therapeutic target for treatment of vascular leakage during severe dengue disease.


Introduction
The Flavivirus genus is composed of enveloped, positive-sense RNA viruses, including medically important mosquito-transmitted viruses such as dengue (DENV), Zika (ZIKV), West Nile (WNV), Japanese encephalitis (JEV), and yellow fever (YFV) viruses, and viruses transmitted by ticks such as tick-borne encephalitis virus (TBEV) [1]. Human infections with flaviviruses lead to a wide range of outcomes and disease manifestations, from asymptomatic infections to severe life-threatening disease characterized by vascular leakage, hemorrhage, organ failure, neurological disorders, and/or fetal abnormalities [2,3]. DENV NS1 has been shown to modulate the permeability as well as the TJ/AJ proteins (e.g., ZO-1, VE-cadherin) of human ECs from skin (HMEC-1) [21] and umbilical vein (HUVEC) [39,40]; however, the effect of NS1 from other flaviviruses on the integrity of IJCs and the cellular mechanisms leading to their disruption remain obscure. Given the central role of IJCs in preserving the homeostasis of the EC barrier, here we examined the effect of NS1 from five distinct flaviviruses (DENV, ZIKV, WNV, JEV, and YFV) on the fate of VE-cadherin and β-catenin within endothelial cell-to-cell contacts. To do so, we used cultures of human ECs in vitro along with a trans-endothelial electrical resistance (TEER) assay as a proxy for the endothelium that lines the inner face of blood vessels in different tissues and its barrier function, together with a fluorescence-based approach to quantify vascular leak in vivo. Furthermore, we identified a requirement for GSK-3β in NS1-mediated endothelial hyperpermeability and vascular leak, as demonstrated using a GSK-3β-specific inhibitor. Taken together, this study provides new molecular insights into the tissue-specific mechanisms by which flavivirus NS1 proteins trigger the disruption of IJCs and identifies GSK-3β as a host kinase that may be targeted for treatment of panflavivirus NS1-mediated pathology.

Flavivirus NS1 Proteins Alter the Localization of the AJ Proteins VE-Cadherin and β-Catenin in Human Endothelial Cells from the Brain and Umbilical Vein
In previous work, we demonstrated that secreted NS1 proteins from distinct flaviviruses can modulate the permeability of human ECs and cause increased vascular leak in a tissue-specific manner, correlating with disease tropism of the respective virus [19,22]. We have also shown that this phenomenon depends on the activation of cellular enzymes (sialidases, heparanase, and cathepsin L), resulting in the disruption of the EGL and contributing to endothelial hyperpermeability [8]. Here, we tested whether NS1 activates pathways regulating the architecture of the IJC. To do this, we examined the effect of different flavivirus NS1 proteins on the cellular distribution of VE-cadherin and β-catenin in human ECs. We found that NS1 treatment of human umbilical vein (HUVEC) and brain microvascular (HBMEC) ECs resulted in noticeable changes in the localization/distribution of VE-cadherin and β-catenin compared to the untreated controls, in which the distribution of these proteins adopt a "chicken fence" staining pattern outlining the cell-cell junctions ( Figure 1A,B). As expected, the staining for VE-cadherin and β-catenin resulted in a continuous pattern outlining the cell borders in the untreated ECs (HUVEC and HBMEC controls). Six hours post-treatment (hpt) with DENV or ZIKV NS1 proteins (5 µg/mL), altered staining of both proteins in the HUVEC monolayers was observed, as indicated by the arrows, whereas treatment with WNV, JEV, and YFV NS1 did not induce noticeable changes ( Figure 1A,B, top panel). In contrast, the staining pattern for both proteins was altered in HBMEC monolayers exposed to NS1 proteins from DENV, ZIKV, WNV, and JEV, but not YFV ( Figure 1A,B, lower panel). In both human ECs, NS1-specific treatment led to changes in the distribution of VE-cadherin and β-catenin, suggesting their mislocalization from the IJC. These tissue-specific patterns of IJC disruption are consistent with the tissue-specific tropism of flavivirus NS1 proteins observed previously [19]. Although we observed a reorganization of the cellular localization of these IJC proteins, the mean fluorescent intensity (MFI) values, used as a proxy of total IJC protein expression, were not significantly altered ( Figure 1C-F). In agreement, we detected no significant modulation of the VE-cadherin and β-catenin protein expression levels in ECs treated with flavivirus NS1 proteins by Western blot analyses of the cell lysates (Figure 2A-D). Taken together, these data suggest that in vitro NS1 treatment results in mislocalization but not degradation of IJC proteins VE-cadherin and β-catenin in HUVEC and HBMEC monolayers.

NS1 Triggers Endocytosis of VE-Cadherin and Phosphorylation of β-Catenin in Human Endothelial Cells
Next, we investigated the mechanism(s) by which flavivirus NS1 proteins trigger IJC mislocalization in EC monolayers. We first monitored the internalization of VE-cadherin, focusing on DENV and ZIKV NS1 treatment of HUVEC monolayers ( Figure 3A,B). We observed increased colocalization of VE-cadherin with clathrin heavy chain proteins, indicating the potential formation of endocytic clathrin-coated pits, in HUVECs 6 hpt with DENV or ZIKV NS1 proteins, compared to untreated cells or another related flavivirus NS1 protein from WNV ( Figure 3A). Quantification of co-staining of VE-cadherin and clathrin showed increased numbers of colocalized puncta in HUVEC monolayers treated with DENV and ZIKV NS1 compared to untreated cells or WNV NS1-treated cells ( Figure 3B). Additionally, we examined the phosphorylation status of β-catenin 6 hpt with NS1 proteins from DENV, ZIKV, WNV, JEV, and YFV using a commercial cell-based ELISA to determine the level of expression and phosphorylation, as well as the percentage of β-catenin phosphorylated at serine residue 45 (S45) ( Figure 3C,D). We also utilized a pan-phosphorylation polyclonal antibody that can detect β-catenin phosphorylation at serine 33, serine 37, or threonine 41 (S33/S37/T41) by IFA ( Figure 3E). ELISA results showed that DENV and ZIKV NS1, but not WNV, JEV, or YFV NS1, induced a significant increase in phosphorylation of β-catenin at Ser-45 in HUVECs, in comparison to the untreated controls ( Figure 3C,D). Additionally, IFA analysis revealed that treatment of HUVECs with DENV and ZIKV NS1, but not with WNV, JEV, or YFV NS1, led to increased phosphorylation of endogenous levels of β-catenin at any of the three residues S33/S37/T41 ( Figure 3E, green puncta). Of note, phosphorylated β-catenin was largely cytoplasmic, while non-phosphorylated β-catenin (total) was primarily localized in the cell-cell junctions. Together, these results demonstrate that stimulation of human ECs with NS1 triggers the phosphorylation of β-catenin in a tissue-specific manner, consistent with our previous observations [19]. proteins and control cells. VE-cadherin (~137 kDa) and β-catenin (~95 kDa) at 6 hpt. GAPDH (~37 kDa) was used as a protein loading control. The fluorescence intensity of the protein bands, corresponding to the molecular size (kDa) of VE-cadherin and β-catenin in each human EC, was determined by ImageJ Studio analyses and plotted as the ratio of their relative intensity vs. GAPDH from the same experimental condition. Images are representative of three independent experiments.

NS1 Triggers Endocytosis of VE-Cadherin and Phosphorylation of β-Catenin in Human Endothelial Cells
Next, we investigated the mechanism(s) by which flavivirus NS1 proteins trigger IJC mislocalization in EC monolayers. We first monitored the internalization of VE-cadherin, focusing on DENV and ZIKV NS1 treatment of HUVEC monolayers ( Figure 3A,B). We observed increased colocalization of VE-cadherin with clathrin heavy chain proteins, indicating the potential formation of endocytic clathrin-coated pits, in HUVECs 6 hpt with DENV or ZIKV NS1 proteins, compared to untreated cells or another related flavivirus NS1 protein from WNV ( Figure 3A). Quantification of co-staining of VE-cadherin and clathrin showed increased numbers of colocalized puncta in HUVEC monolayers treated with DENV and ZIKV NS1 compared to untreated cells or WNV NS1-treated cells ( Figure  3B). Additionally, we examined the phosphorylation status of β-catenin 6 hpt with NS1 proteins from DENV, ZIKV, WNV, JEV, and YFV using a commercial cell-based ELISA to determine the level of expression and phosphorylation, as well as the percentage of βcatenin phosphorylated at serine residue 45 (S45) ( Figure 3C,D). We also utilized a panphosphorylation polyclonal antibody that can detect β-catenin phosphorylation at serine 33, serine 37, or threonine 41 (S33/S37/T41) by IFA ( Figure 3E). ELISA results showed that DENV and ZIKV NS1, but not WNV, JEV, or YFV NS1, induced a significant increase in phosphorylation of β-catenin at Ser-45 in HUVECs, in comparison to the untreated con- proteins and control cells. VE-cadherin (~137 kDa) and β-catenin (~95 kDa) at 6 hpt. GAPDH (~37 kDa) was used as a protein loading control. The fluorescence intensity of the protein bands, corresponding to the molecular size (kDa) of VE-cadherin and β-catenin in each human EC, was determined by ImageJ Studio analyses and plotted as the ratio of their relative intensity vs. GAPDH from the same experimental condition. Images are representative of three independent experiments.

GSK-3β Is Required for NS1-Mediated Endothelial Hyperpermeability and Vascular Leak
Given the central role of phosphorylation of β-catenin in modulating the barrier function of epithelial cells [33], we assessed the contribution of GSK-3β, which phosphorylates S33/S37/T41, to flavivirus NS1-induced endothelial hyperpermeability in vitro and vascular leak in vivo. We utilized a small peptide competitive inhibitor of GSK-3β (GSK-3β Peptide Inhibitor, GPI) [41]. Using TEER as a measure of the integrity of the endothelial barrier, we demonstrated that the peptide inhibitor alone did not interfere with endothelial permeability, but that GPI was sufficient to inhibit DENV and ZIKV NS1-mediated hyperpermeability in a dose-dependent manner in HBMEC ( Figure 4A,B) and HUVEC ( Figure 4C,D) monolayers. Further, in a murine model of dermal endothelial permeability, in which fluorescently labeled dextran enables quantitation of local vascular permeability in vivo [9,22], GPI was tested in the presence of 15 µg of DENV NS1. GPI significantly reduced the DENV NS1-induced vascular leak in comparison to the NS1 + vehicle treatment without inducing vascular leak alone, in comparison to the PBS control ( Figure 4E,F). These results support the critical contribution of GSK-3β to DENV NS1-induced permeability DENV and ZIKV NS1, but not with WNV, JEV, or YFV NS1, led to increased phosphorylation of endogenous levels of β-catenin at any of the three residues S33/S37/T41 ( Figure  3E, green puncta). Of note, phosphorylated β-catenin was largely cytoplasmic, while nonphosphorylated β-catenin (total) was primarily localized in the cell-cell junctions. Together, these results demonstrate that stimulation of human ECs with NS1 triggers the phosphorylation of β-catenin in a tissue-specific manner, consistent with our previous observations [19].  . Treatment of HUVEC monolayers with DENV and ZIKV NS1 proteins but not WNV NS1 increases the colocalization of VE-cadherin with clathrin and the phosphorylation and cytosolic accumulation of β-catenin. Confluent monolayers of HUVEC were individually treated with five different NS1 proteins from DENV, ZIKV, WNV, JEV and YFV (5 µg/mL), and 6 hpt, cell monolayers were processed by (A,B) IFA for co-staining of VE-cadherin and clathrin heavy chain, (C,D) commercial ELISA for detection of total/phosphorylated β-catenin (S45), or (E) IFA for co-staining of total β-catenin (red) and phosphorylated β-catenin (S33, S37, T41) (green). Images are representative of three independent experiments. White arrows indicate VE-cadherin co-localization with clathrin heavy chain proteins (A) and the cytosolic accumulation of phosphorylated β-catenin (green) (E).

Discussion
In this study, we demonstrated that NS1 proteins from closely related flaviviruses trigger endothelial hyperpermeability in an EC-type-dependent manner through alterations in the normal distribution of VE-cadherin and β-catenin, two important proteins of the IJC that are critical for maintenance of EC barrier function. This phenomenon occurred

Discussion
In this study, we demonstrated that NS1 proteins from closely related flaviviruses trigger endothelial hyperpermeability in an EC-type-dependent manner through alterations in the normal distribution of VE-cadherin and β-catenin, two important proteins of the IJC that are critical for maintenance of EC barrier function. This phenomenon occurred via clathrin-mediated endocytosis of VE-cadherin and phosphorylation of β-catenin, leading to their movement out of the IJC, which resulted in increased endothelial permeability and vascular dysfunction. Further, we demonstrated a critical role for GSK-3β, a cellular kinase that contributes to the phosphorylation of β-catenin, leading to the destabilization of the IJC complex during NS1-mediated endothelial hyperpermeability and vascular leak. A proposed model of this potential mechanism is depicted in Figure 5.  (1). IJCs are major components required for maintaining endothelial barrier integrity, which regulates the passage of fluids and soluble molecules between the bloodstream and underlying tissues. Under homeostatic conditions, the AJ complex forms tight barriers between cells containing VE-cadherin, which is anchored to the cell cytoplasm via association of its cytoplasmic domain with β-catenin, α-catenin, and cytoskeletal components (2). Localization of β-catenin to AJs, and thus IJC barrier integrity, is regulated via several phosphorylation events mediated by kinases, including GSK-3β (3). Our current study shows that tissue-specific flavivirus NS1-mediated endothelial dysfunction is associated with mislocalization of the IJC proteins, VE-cadherin, and β-catenin (B). Our study reveals mechanistic details of AJ disruption, demonstrating that NS1 treatment of endothelial cells is associated with clathrin-mediated endocytosis of VE-cadherin (4) and phosphorylation of β-catenin (5), which is known to destabilize the β-catenin/VE-cadherin complex, resulting in endothelial barrier dysfunction (6). In agreement with these observations, our study demonstrates that a GSK-3β-specific peptide inhibitor (GPI) is sufficient to abrogate NS1-mediated endothelial permeability and vascular leak (7). Thus, our study provides new insights by which TJ/AJ barrier integrity is disrupted by NS1. Abbreviations: EGL (endothelial glycocalyx layer); ECs (endothelial cells); ECM (extracellular matrix); IJC (intercellular junction complex); N (nucleus); β-cat (β-catenin); α-cat (α-catenin); GAGs (glycosaminoglycans); PG (proteoglycan); PM (plasma membrane); VE-cad (VE-cadherin). p-β-cat (phosphorylated β-catenin).
Endothelial cell-to-cell junctions are dynamically regulated structures comprised of TJ/AJ proteins whose expression and localization are critical to cell barrier function, regulating the paracellular transport of polar solutes and macromolecules into the tissues from the blood stream under distinct physiological and pathophysiological conditions [42,43]. The IJC is actively remodeled in response to multiple extracellular stimuli to transiently increase or decrease endothelial permeability [44]. Therefore, breakdown of the IJC results in permeability changes and barrier dysfunction, particularly under pathological conditions [45][46][47][48]. Displacement of IJC proteins such as ZO-1 and VE-cadherin can occur through multiple pathways [49,50]. In epithelial and endothelial cells, clathrin-and caveolae-mediated internalization are critical cellular mechanisms for the transport of IJC proteins from the plasma membrane to the cytoplasm, causing dissociation of the IJC [51]. Additionally, phosphorylation of some of the IJC proteins (e.g., β-catenin) has been shown to reduce the affinity of TJ/AJ protein-protein interactions, which results in destabilization of the TJ/AJ complexes and barrier dysfunction [33,52].
In this study, we identified that NS1 treatment of human ECs changed the distribution of VE-cadherin and β-catenin at cell-to-cell contacts without significantly altering the expression level of these pivotal IJC proteins. This phenomenon coincides with the peak of EC hyperpermeability (decreased TEER) occurring in EC monolayers exposed to soluble NS1 proteins between 6 and 8 h post-treatment, suggesting that altered distribution but not altered expression of IJC proteins may contribute to the transient peak of EC barrier dysfunction occurring during flavivirus infections. One limitation of our study is that IJC protein distribution and expression were only analyzed at a single time point, 6 hpt, when increased endothelial permeability peaks in vitro. Thus, although no changes in IJC protein expression were detected at this time point, it is possible that alterations may be observed at other time points. Further studies analyzing the kinetics of IJC protein distribution and expression after human EC exposure to flavivirus NS1 proteins will provide a more complete picture of the dynamics of NS1-medated barrier dysfunction. Flavivirus NS1 proteins have been shown to selectively bind to and alter the permeability of human ECs and cause tissue-specific vascular leakage in vivo, reflecting the pathophysiology of each flavivirus [9,19,22,23]. For instance, NS1 from neurotropic flaviviruses, such as WNV, JEV, DENV, and ZIKV, can cause endothelial dysfunction in human brain ECs, and those flaviviruses in which transplacental infection has been described, such as DENV and ZIKV [79][80][81][82][83], the corresponding NS1 caused barrier dysfunction in umbilical vein EC used as a proxy for placental tissues. Mechanistically, this phenomenon is driven at the stage of binding and internalization of NS1 into ECs, which in turn triggers flavivirus-conserved steps, such as the activation of EC-intrinsic pathways (e.g., heparanase, sialidase), that result in the disruption of the EGL on the surface of ECs, contributing to vascular leak [8,19]. A complete picture of which amino acids dictate tissue-specific NS1 function is still lacking, but previous structural studies have identified three distinct do-mains of flavivirus NS1, namely, β-roll, wing, and β-ladder [84]. Our current and previous work highlight both conserved residues within the wing and β-ladder domains that all NS1 proteins require to trigger barrier dysfunction, as well as divergent residues within the wing domain that are required for tissue specificity [85]. Further, we generated chimeras and site-specific mutants of DENV, ZIKV, and WNV NS1 proteins that demonstrated the capacity to modulate EC tropism by swapping the wing and β-ladder domains. Specifically, we identified a three-amino acid motif of DENV that confers EC-binding tropism [86]. This is an important area of active investigation, which will improve understanding of the NS1 determinants that trigger EC barrier dysfunction.
While specific host factors required for NS1-medited IJC disruption are still obscure, previous studies have shown that secreted NS1 proteins from DENV and ZIKV disrupt the IJC (e.g., β-catenin and ZO-1) through induction of macrophage migration inhibitory factor (MIF) and matrix metalloproteinase-9 (MMP-9) [10,23,39], or through activation of human ECs via the inflammatory stress-sensing p38 MAPK pathway and/or Angiotensin 1/2 receptor signaling, which were required for NS1-triggered reduction of barrier integrity of human EC monolayers in vitro [40,76]. Here, we observed that treatment of EC monolayers with DENV and ZIKV NS1 proteins resulted in colocalization of VE-cadherin with clathrin heavy chain, suggesting its internalization via clathrin-mediated endocytosis. The phenomenon of EC barrier dysfunction was accompanied by increased staining of clathrin heavy chain proteins in HUVEC monolayers exposed to DENV and ZIKV NS1 proteins but not in the untreated ECs or ECs treated with WNV NS1. NS1 internalization in infected and non-infected cells, particularly via clathrin-mediated endocytosis, has been described as a cellular mechanism required for NS1-induced endothelial hyperpermeability, EGL degradation, and in DENV infection [22,40,76,87].
Additionally, we describe the phosphorylation and disassembly of β-catenin from the cadherin-catenin complex, correlating with the requirement of GSK-3β for NS1-mediated endothelial dysfunction, and its mislocalization in NS1-treated cells. GSK3 is a serine/threonine protein kinase that exists in two isoforms encoded by two different genes, GSK3α and GSK3β, with numerous well-documented roles in distinct cellular processes, including innate immune modulation and glucose metabolism [35,88]. In the past two decades, the recognized contribution of GSK3 kinases in human diseases including diabetes, cancer, and inflammation, as well as many viral infections and pathogenesis, has increased dramatically [36,89]. Of note, GSK3 has been identified as a host factor required for replication of SARS-CoV-1 and -2 [90,91], influenza [92], hepatitis C [93], dengue [94], and other viruses [95,96]. Yet, the role of GSK3 in mediating flavivirus pathogenesis related to the modulation of endothelial barrier function by secreted NS1 proteins has not been addressed. Here, given the pivotal role of GSK3 in regulating the fate of β-catenin in response to distinct external stimuli, we showed in vitro and in vivo that a specific inhibitor of GSK-3β activity significantly reduced NS1-mediated endothelial hyperpermeability and prevented NS1-mediated vascular leak, supporting the contribution of GSK3 kinases to NS1-induced pathogenesis. One caveat to our data is the potential for off-target effects of inhibitors such as the GSK3 peptide inhibitor to alter NS1-mediated IJC disruption. Thus, future genetic investigations into the role of GSK-3B are needed to further probe the role of this kinase in NS1-medated EC dysfunction.
Overall, this work provides new insights into the endothelial, cell-intrinsic mechanisms that contribute to endothelial hyperpermeability triggered by NS1 proteins from clinically important human flaviviruses such as DENV, ZIKV, WNV, and JEV. It confirms the role of the IJC as highly organized cellular structures that maintain the homeostasis of the endothelium and supports previous findings implicating the IJC in the pathogenic processes of DENV and other flavivirus NS1 proteins. Along with previous data, our results together suggest that the mislocalization of AJ proteins from cell-to-cell contacts in human ECs induced by NS1 is a complex multifactorial phenomenon involving several cellular mechanisms that act together to cause endothelial hyperpermeability and vascular leak.
The kinetics and dynamics of this phenomenon are yet to be defined and future studies will define the spatial-temporal relationships between host factors and NS1.

Ethics Statement
All in vivo experiments were performed following the guidelines of the American Veterinary Medical Association and the Guide for the Care and Use of Laboratory Animals from the National Institutes of Health and were pre-approved by the University of California (UC) Berkeley Animal Care and Use Committee (protocol AUP-2014-08-6638) as previously described [9,11,19].

Mice
Six-to-eight-week-old wild-type C57BL/6 (B6) mice were obtained from the Jackson Laboratory. All mice were bred and maintained in specific pathogen-free conditions at the animal facility at UC Berkeley. A mix of male and female six-to-eight-week-old mice were used in all experiments. Mice were housed in a controlled-temperature environment on a 12-h light/dark cycle, with food and water provided ad libitum.

Trans-Endothelial Electrical Resistance (TEER)
The effect of recombinant flavivirus NS1 proteins on endothelial permeability was evaluated by measuring TEER of EC monolayers grown on a 24-well Transwell polycarbonate membrane system (Transwell ® permeable support, 0.4 µm, 6.5 mm insert; Corning Inc., Corning, NY, USA), as previously described [8,19]. Briefly, HUVECs or HBMECs (80-90% confluency) cultured in vented 75 cm 2 flasks (Corning ® ) were detached using a combination of three washing steps using sterile 1× PBS supplemented with EDTA (2 mM) (lifting buffer), followed by two additional washing steps (~30 s total) using a solution of trypsin-EDTA (0.25%) (GIBCO, Thermo Scientific, Waltham, MA, USA). Detached cells were resuspended using fresh culture medium and then counted using a tissue-culture hemocytometer. A total of 1 × 10 5 cells in 300 uL of medium were seeded on the apical side of Transwell inserts (top chamber) placed in a 24-well plate containing 1.5 mL of EC culture medium (basolateral side). Transwells containing ECs were incubated at 37 • C and 5% CO 2 for 3 days, and 50% of culture medium was changed in each well every day post-seeding. Cells were grown until TEER values between 150 and 180 Ohms (Ω) were reached, depending on cell type, indicating 100% cell confluency. Individual flavivirus NS1 proteins (5 µg/mL, 1.5 µg total protein) were then added to the apical side of the Transwell insert (top chamber, 300 µL) containing the cell monolayer. TEER values, expressed in Ohms (Ω), were collected at sequential 2-h time-points over 3-11 h following the addition of test proteins using an Epithelial Volt Ohm Meter (EVOM) with a "chopstick" electrode (World Precision Instruments). Inserts with no cells containing medium alone were used as negative controls to calculate the baseline electrical resistance. Endothelial permeability was expressed as relative TEER, which represents a ratio of resistance values (Ω) as follows: (Ω experimental condition-Ω medium alone)/(Ω non-treated endothelial cells-Ω medium alone).

Fluorescence Microscopy
For imaging experiments, HUVECs and HBMECs were grown on coverslips and imaged on a Zeiss LSM 710 Axio Observer inverted fluorescence microscope equipped with a 34-channel spectral detector. Images acquired using the Zen 2010 software (Zeiss, Jena, Germany) (708.49 × 708.49 µM) were processed and analyzed with ImageJ software [97]. For representative pictures (RGB format), an area of 163.12 × 178.77 µm containing~10-30 cells was used. To assess the effect of flavivirus NS1 on integrity of the endothelial architecture, the distribution of VE-cadherin and β-catenin was examined on confluent EC monolayers treated with distinct flavivirus NS1 proteins, as indicated in each figure legend. EC monolayers were fixed with 2% paraformaldehyde (PFA) and cold methanol (1 mL) at 6 hpt. A permeabilization step was performed using saponin (0.2%) in blocking buffer (3% BSA, 1% FBS in PBS 1×) for 30 min at room temperature. Primary antibodies were added and incubated overnight at 4 • C in PBS (1×), and detection was performed using secondary species-specific anti-IgG antibodies conjugated to Alexa fluorophores (568 and 647). Nuclei were stained using Hoechst. Mean fluorescence intensity (MFI) values for β-catenin staining on human ECs treated with NS1 or controls were obtained from individual RGB-grayscale-transformed images (n = 3). All images were processed, edited, and analyzed using ImageJ software, as previously described [8,9,19,22,23,97].

Western Blot Analyses
For protein expression, confluent EC monolayers (~1 × 10 6 cells/well, 6-well tissue culture-treated plates) were treated with the five flavivirus NS1 proteins (DENV, ZIKV, WNV, JEV, and YFV) separately (5 µg/mL), and at 6 hpt, cell monolayers were scraped on ice using RIPA lysis buffer (50 mM Tris (pH 7.4), 150 mM NaCl, 1% (v/v) Nonidet-P40, 2 mM EDTA, 0.1% (w/v) SDS, 0.5% Na-deoxycholate, and 50 mM NaF) supplemented with a complete protease inhibitor cocktail (Roche, Basel, Switzerland). After total protein quantification using a bicinchoninic acid (BCA)-based colorimetric assay (Pierce BCA Protein Assay Kit, Thermo Scientific, Waltham, MA, USA), 10 µg of total protein per sample was boiled and placed in reducing Laemmeli buffer and separated by 4-20% gradient SDS-PAGE. After immunoblotting using specific primary antibodies for β-catenin and GAPDH (used as housekeeping protein control) and secondary species-specific anti-IgG antibody conjugated to Alexa 680 or Alexa 750, protein detection and quantification was carried out using the Odyssey CLx Infrared Imaging System (LI-COR, Biosciences, Lincoln, NE, USA). Relative densitometry represents a ratio of the values obtained from each experimental protein band over the values obtained from the loading controls (GAPDH) after subtracting the background from both using Image Studio Lite V 5.2 (LI-COR Biosciences, Biosciences, Lincoln, NE, USA) [8,19].

ELISA
Instant one step ELISA for human β-catenin (PhosphoTracer β-Catenin Total ELISA Kit) on adherent cells was performed following the manufacturer's instructions (Thermo Fisher Scientific, Waltham, MA, USA). Briefly, confluent HUVEC monolayers cultured in 48-well plates were treated with distinct flavivirus NS1 proteins (5 µg/mL), and cell protein lysates were collected 6 hpt following the manufacturer's protocol. The amount of total and phosphorylated (S45) β-catenin per well was determined by fluorogenic quantification using a Spectra Max (M3) microplate reader (Molecular Devices, San Jose, CA, USA) equipped with a dual monochromator spectrofluorometer system that provides excitation and emission wavelength selection between 250-850 nm, as previously described [98]. The amount of phosphorylated (S45) β-catenin was expressed as the percentage of the total β-catenin as 100%.

Localized Vascular Leak Murine Model Assay
In vivo NS1-induced endothelial hyperpermeability was measured using a rodent model of localized vascular leak, as previously described [9]. Briefly, the dorsal hair area of 6-weekold female WT C57BL/6 mice (Jackson Labs) was depilated 3-4 days prior to each experiment. On the day of the experiment, mice were anesthetized with isoflurane and injected into the shaved dorsal dermis with 50 µL of four different treatments as follows: 1× PBS plus vehicle (1 µL of DMSO in 50 µL PBS1×), GSK3β inhibitor alone (10 µg in 1 µL DMSO plus 50 µL 1× PBS), DENV2 NS1 (15 µg in 1 µL plus 50 µL 1× PBS) alone, and the mixture of GSK3β inhibitor plus DENV2 NS1 (15 µg NS1 plus 10 µg GSK3β inhibitor in 50 µL 1× PBS). Local vascular leakage at the injection spot was quantified and expressed as the mean fluorescence intensity of 10-kDa dextran conjugated with Alexa Fluor 680 (1 mg/mL; Sigma) delivered via retro-orbital (RO) injection. Two hours post-injection, mice were euthanized using isoflurane, and the dorsal dermis was removed and placed in Petri dishes. Tissues were scanned using a fluorescent detection system (LI-COR Odyssey CLx Imaging System) at a wavelength of 700 nm, and leakage in a 13-mm diameter circle surrounding the sites of injection was quantified using Image Studio software (LI-COR Biosciences, Lincoln, NE, USA).

Data Analyses and Statistics
All experimental conditions were repeated three times, and images are representative of these experiments. Statistical differences between groups were determined by Ordinary one-way ANOVA and considered significant with p-values <0.05. Comparison between MFI, ELISA, and densitometry data was conducted using multiple t-tests. All data analyses, statistics, and graphs were performed and generated using GraphPad Prism v6.07.

Informed Consent Statement: Not applicable.
Data Availability Statement: All data generated or analyzed during this study are included within this published article.