Macrophage migration inhibitory factor is critical for dengue NS1-induced endothelial glycocalyx degradation and hyperpermeability

Vascular leakage is one of the salient characteristics of severe dengue. Nonstructural protein 1 (NS1) of dengue virus (DENV) can stimulate endothelial cells to secrete endothelial hyperpermeability factor, macrophage migration inhibitory factor (MIF), and the glycocalyx degradation factor heparanase 1 (HPA-1). However, it is unclear whether MIF is directly involved in NS1-induced glycocalyx degradation. In this study, we observed that among NS1, MIF and glycocalyx degradation-related molecules, the HPA-1, metalloproteinase 9 (MMP-9) and syndecan 1 (CD138) serum levels were all increased in dengue patients, and only NS1 and MIF showed a positive correlation with the CD138 level in severe patients. To further characterize and clarify the relationship between MIF and CD138, we used recombinant NS1 to stimulate human cells in vitro and challenge mice in vivo. Our tabulated results suggested that NS1 stimulation could induce human endothelial cells to secrete HPA-1 and immune cells to secrete MMP-9, resulting in endothelial glycocalyx degradation and hyperpermeability. Moreover, HPA-1, MMP-9, and CD138 secretion after NS1 stimulation was blocked by MIF inhibitors or antibodies both in vitro and in mice. Taken together, these results suggest that MIF directly engages in dengue NS1-induced glycocalyx degradation and that targeting MIF may represent a possible therapeutic approach for preventing dengue-induced vascular leakage.

Introduction umbilical vein endothelial cells (HUVECs) were stimulated with NS1 for various durations. The results show that CD138 was significantly increased in cell culture medium after 24 h of NS1 treatment (Fig 3A). To confirm that this effect was induced by NS1, anti-NS1 monoclonal antibody (mAb) was used to block the effect of NS1. Anti-NS1 mAb 2E8, which can inhibit NS1-induced vascular leakage, was able to inhibit NS1-induced CD138 shedding (Fig 3B) [19]. In contrast, another anti-NS1 mAb (DN5C6), which was used as a negative control, failed to inhibit NS1-induced CD138 shedding from endothelial cells (Fig 3B) [19]. NS1 stimulation also increased the active HPA-1 level in endothelial cell lysates, which was abolished by mAb 2E8 but not control mouse IgG (S2A Fig). To confirm that HPA-1 is involved in NS1-induced endothelial hyperpermeability and CD138 shedding, recombinant HPA-1 protein and the HPA-1 inhibitor OGT 2115 were used. Inoculating the mice with native but not heat-denatured recombinant HPA-1 directly induced vascular leakage (S2B Fig). Furthermore, cotreatment with OGT 2115 attenuated NS1-induced endothelial hyperpermeability ( Fig 3C) and reduced CD138 release to levels similar to those of the phosphate-buffered saline (PBS) control in vitro (Fig 3D).
In addition to HPA-1, MIF is also capable of inducing endothelial hyperpermeability [19]. As a result, the MIF concentration in the conditioned medium obtained from NS1-stimulated HUVECs was measured. The result shows that 10 μg/ml NS1 was sufficient to induce MIF secretion (S3A Fig). Furthermore, the conditioned medium obtained from NS1-stimulated HUVECs could induce endothelial hyperpermeability and CD138 shedding after incubation with another HUVEC monolayer (S3B and S3C Fig). To clarify which protein mediates NS1-induced endothelial hyperpermeability and CD138 shedding, MIF-blocking antibodies, the HPA-1 inhibitor OGT 2115 and NS1-blocking antibodies were used. The results show that both the anti-MIF antibodies and OGT 2115 attenuated NS1-stimulated conditioned Interestingly, the NS1-blocking antibody 2E8 only partially diminished the conditioned mediuminduced HUVEC hyperpermeability but not the conditioned medium-induced HUVEC CD138 shedding (S3D and S3E Fig). Cotreatment with MIF inhibitors (anti-MIF antibodies, ISO-1, and p425) and NS1 also attenuated the NS1-induced HPA-1 secretion ( Fig 3E) and CD138 shedding of endothelial cells (Fig 3F). In addition, we also visualized the HPA-1 expression, CD138 deposition and sialic acid expression using immunofluorescence with anti-HPA antibodies, anti-CD138 antibodies and wheat germ agglutinin (WGA) lectin which can bind  20 μg/ml NS1, or 20 μg/ml NS1 mixed with 10 μg/ml anti-NS1 antibodies (2E8 or DN5C6). After 24 h of incubation, the culture medium was collected, and the concentration of CD138 was determined by ELISA. (n = 3) (C) HUVECs seeded as monolayers in upper Transwell chambers were treated with PBS, 20 μg/ml NS1 or 20 μg/ml NS1 mixed with DMSO or the indicated concentration of OGT 2115. After 24 h, endothelial permeability was determined by a Transwell permeability assay, as described in the Materials and Methods section. (n = 3) (D) HUVECs were treated with PBS or 20 μg/ml NS1 with or without 5 μM OGT 2115. After 24 h, the cell culture medium was collected, and the CD138 concentration was measured by ELISA. (n = 5) (E) HUVECs were treated with PBS or 20 μg/ml NS1 with or without 100 μM p425, 50 μM ISO-1, or 10 μg/ml anti-MIF antibodies, as indicated. After 24 h, the cell culture medium was collected, and the HPA-1 concentration was measured by ELISA. (n = 3) (F) HUVECs were treated with PBS or 20 μg/ml NS1 with or without p425, ISO-1 or anti-MIF antibodies. After 24 h, the cell culture medium was collected, and the CD138 concentration was measured by ELISA. (n = 5) (G) HUVECs were treated with PBS or NS1 (20 μg/ml) with or without anti-MIF polyclonal antibodies (10 μg/ml) for 24 h. The distribution of HPA-1 (red) and CD138 (green) was assessed by staining with specific antibodies. Sialic acid expression on HUVECs monolayers was assessed by staining with WGA-FITC (green to sialic acids and other sugars such as N-acetylglucosamine. As shown in Fig 3G, NS1-induced HPA-1 expression, CD138 deposition and sialic acid degradation could also be rescued by MIF inhibition. However, the HPA-1 inhibitor OGT 2115 failed to affect MIF secretion, suggested that MIF is the upstream effector of HPA-1 (S3F Fig). To further clarify this hypothesis, recombinant MIF was used. The results show that MIF increased CD138 shedding and the active HPA-1 level in HUVECs (S4A and S4B Fig). The increased HPA-1 expression and CD138 deposition after MIF stimulation could also be observed by immunofluorescence (S4C Fig). These results indicate that NS1 can induce the MIF-mediated secretion of active HPA-1, leading to endothelial glycocalyx degradation.

DENV NS1 induces MMP-9 secretion in THP-1 cells and leukocytes
Since MMPs can degrade the endothelial glycocalyx and several MMPs are upregulated during DENV infection [30,31], we speculated that MMPs are involved in NS1-induced glycocalyx degradation. Because a previous study has indicated that an increase in circulating MMP-9 levels is associated with dengue disease severity [32], we first examined whether NS1 induces MMP-9 secretion. However, we found that NS1 barely induced MMP-9 secretion in HUVECs ( Fig 4A). Since MMPs are primarily secreted by leukocytes (white blood cells, WBCs), including neutrophils and monocytes [33], we tested whether NS1 could induce MMP secretion in HUVECs were treated with PBS or NS1 for the indicated times; then, the supernatants were collected for MMP-9 detection by ELISA. (n = 3) (B) Isolated human leukocytes (WBCs) and (D) PMA-activated THP-1 cells were treated with PBS or NS1, and the culture medium was collected at the indicated times. The concentration of MMP-9 in the culture medium was determined by ELISA. (n = 4) (C) Isolated WBCs were treated with PBS, NS1, or NS1 mixed with anti-NS1 antibodies for 24 h, and the concentration of MMP-9 in the culture medium was determined by ELISA. (n = 4) (E) After PMA activation, THP-1 and primary isolated WBCs were subjected to the desired treatment. After 24 h, the cell culture supernatants were collected. An MMP antibody array that detects various MMPs and TIMPs was used to assess the major subclass of MMPs induced by NS1. THP-1 cells or WBCs were stimulated with PBS or NS1; then, the culture supernatants were collected and analyzed for extracellular matrix proteins. Membranes of the human MMP antibody array were probed with the supernatant collected from bovine serum albumin (BSA)-treated THP-1 cells, BSA-treated WBCs, 20 μg/ml NS1-treated THP-1 cells or NS1-treated WBCs. The quantification of MMPs array membranes was analyzed by ImageJ. PC, positive control; NC, negative control. (F) After treatment with 20 μg/ml NS1 for 24 h, the 5X-concentrated WBCs and 10X-concentrated THP-1 supernatants were analyzed by electrophoresis with a 7.5% acrylamide gel containing gelatin. The gel was stained with Coomassie blue to reveal the white bands corresponding to the proteolysis of gelatin by MMPs. S/N, supernatant; Ã P<0.05, ÃÃ P<0.01, ÃÃÃ P<0.001; unpaired t-test (panel B and D), Kruskal-Wallis ANOVA (panel C).
To obtain the secretion profile of MMPs, we used an MMP antibody array to analyze which MMPs were increased by NS1 in PMA-activated THP-1 cells and leukocytes. The results show that MMP-8, MMP-9, and TIMP-1 were increased in the culture medium of NS1-treated THP-1 cells and leukocytes (Fig 4E). To confirm the activity of MMP-9, cell culture medium from NS1-treated PMA-activated THP-1 cells and leukocytes were analyzed using a gelatin zymography assay, which showed that NS1 induced both THP-1 cells and leukocytes to secrete pro-MMP-9 and activated MMP-9 ( Fig 4F).

DENV NS1-induced MMP-9 secretion from THP-1 cells increases endothelial permeability and glycocalyx degradation
To test whether the NS1-induced MMP-9 secretion of THP-1 cells causes endothelial hyperpermeability, the supernatant from NS1-treated THP-1 cells was incubated with HUVECs, and both permeability and CD138 shedding were examined. The results show that after 3 h of treatment, the supernatant from NS1-treated THP-1 cells increased endothelial permeability (Fig 5A). This phenomenon was attenuated in the presence of the MMP-2/MMP-9 inhibitor SB-3CT and the MMP-9-specific inhibitor MMP-9 inhibitor I (Fig 5B and 5C). The supernatant from untreated or PBS-treated THP-1 cells did not alter endothelial permeability (Fig 5B  and 5C). Similarly, the supernatant from NS1-treated THP-1 cells also induced CD138 shedding from HUVECs (Fig 5D), and this effect was diminished by SB-3CT and MMP-9 inhibitor I (Fig 5E and 5F). Similar results were found for the supernatant obtained from NS1-stimulated leukocytes (S6 Fig). The NS1-blocking antibody 2E8 was used to block NS1 remaining in the supernatant, and it did not alter the endothelial permeability induced by the supernatant, showing that the effect of NS1 remaining in the supernatant is negligible (S6 Fig). These results indicate that NS1 can induce MMP-9 secretion in leukocytes, leading to endothelial barrier dysfunction.

MIF is required for DENV NS1-induced MMP-9 secretion
As MIF is a crucial mediator of NS1-induced vascular leakage and an upstream regulator of MMP-9 [19,26,34,35], we tested whether NS1-induced MMP-9 secretion in leukocytes is also MIF dependent. First, we wanted to confirm whether NS1 induces the secretion of MIF from leukocytes and THP-1 cells. Since a previous study has shown that NS1 increases the expression of IL-6 and IL-8 in PBMCs [16], we also measured the concentrations of IL-6 and IL-8 after NS1 stimulation. The secretion of MIF from NS1-stimulated leukocytes steadily accumu- To clarify whether NS1-induced MMP-9 secretion is mediated by MIF, the MIF inhibitor p425 and MIF short-hairpin RNA (shRNA) were used. The ELISA results show that inhibiting MIF with its inhibitor p425 abolished NS1-induced MMP-9 secretion, while p425 alone did not affect MMP-9 secretion ( Fig 6A). Next, we used shRNA to knockdown MIF expression in THP-1 cells. Western blot analysis showed that the expression of MIF was diminished by shMIF compared to the shLuc scrambled control ( Fig 6B). Furthermore, the knockdown of MIF decreased NS1-induced MMP-9 secretion from THP-1 cells (Fig 6B), and the culture supernatant from shMIF THP-1 cells failed to increase endothelial permeability or CD138 shedding (Fig 6C and 6D). We also knocked down MIF expression in HUVECs and measured the permeability under NS1 stimulation as a comparison. Consistent with our previous study, the knockdown of MIF in HUVECs diminished NS1-induced endothelial hyperpermeability (S8A and S8B Fig). These results suggest that MIF acts on both endothelial cells and leukocytes to mediate NS1-induced endothelial hyperpermeability.

DENV NS1 induces MIF, HPA-1, MMP-9 and CD138 secretion in mice
To further confirm that NS1 can induce MIF, HPA-1, MMP-9 and CD138 secretion in vivo, we injected 50 μg of NS1 into the tail veins of mice, and blood samples were collected every 24 h. The concentrations of NS1, MIF, HPA-1, and MMP-9 were measured by ELISA. The results show that the peak concentration of NS1 in the plasma of mice after injection was approximately 0.75 μg/ml, which falls in the range of NS1 circulating in the bloodstream of DENV-infected patients, estimated as 0.01-50 μg/ml [36]. The concentration of circulating NS1 in mice gradually decreased after the injection and was cleared from the plasma after 96 h ( Fig 7A). The MIF concentration increased 24 h after the injection, peaked at 72 h, and then dropped to basal levels after 96 h ( Fig 7A). The upregulation of HPA-1 occurred later than that of MIF, as it was significantly elevated after 48 h, but it also peaked at 72 h and then dropped to basal levels after 96 h ( Fig 7A). The secretion of MMP-9 did not increase until 72 h, and then it returned to basal levels at 96 h ( Fig 7A), exhibiting an increase over a relatively short period.

MIF inhibition attenuates DENV NS1-induced endothelial glycocalyx degradation in mice
To further investigate whether NS1 causes endothelial glycocalyx degradation in mice, the skin tissues of mice after two sequential subcutaneous injections of NS1 were fixed for immunohistochemical (IHC) staining. Costaining with the endothelial marker α-SMA revealed CD138 only in the samples with two injections of PBS, E or prM (Fig 7B). After two sequential injections of NS1, endothelial cells lost their CD138 staining (Fig 7B). In addition, the intraperitoneal injection of NS1 significantly induced HPA-1, MMP-9, and CD138 secretion, and coinjection of ISO-1 significantly abolished the secretion of MMP-9 and CD138 but not HPA-1 found by the peritoneal lavage (Fig 7C-7E). Furthermore, the inhibition of MIF and MMP-9 also attenuated NS1-induced vascular leakage in mice (S9 Fig). These results suggest that MMP-9 induced by NS1-stimulated leukocytes may play an important role in endothelial glycocalyx degradation.

Discussion
In this study, we first observed that the concentrations of NS1, MIF, HPA-1, MMP-9 and CD138 in the serum of dengue patients were increased. However, only the concentrations of NS1 and MIF showed a positive correlation with CD138 in severe dengue patients. Next, we showed that the DENV NS1 stimulation of endothelial cells and leukocytes could induce HPA-1 and MMP-9 secretion, respectively, causing endothelial glycocalyx degradation and hyperpermeability. Most importantly, both in vitro and in vivo data showed that dengue NS1-induced HPA-1 and MMP-9 secretion was MIF dependent. Therefore, these results suggest that MIF is a central modulator of both direct and indirect dengue NS1-induced endothelial glycocalyx degradation (Fig 8).
Previously, Puerta-Guardo et al. showed that HPA-1 is involved in NS1-induced glycocalyx degradation and hyperpermeability [18]. However, MMPs were not discussed in the mechanism, even though they are the main enzymes that degrade endothelial glycocalyx [11,12]. It is known that DENV infection induces dendritic cells to secrete MMP-9 [31]. In this study, we further demonstrated that the NS1 stimulation of leukocytes but not endothelial cells nor PBMCs could induce MMP-9 secretion. It is known that DENV NS1 can induce neutrophil extracellular traps, which results in the release of tertiary granules containing MMP-9 [37,38]. A previous study has also shown that MIF can mediate the secretion of MMP-9 from neutrophils [39]. Since neutrophils are a major population of leukocytes, taken together, these results suggest that NS1-stimulated neutrophils may represent the main contributors to MMP-9 secretion in the blood. Therefore, even though neutrophils are not the primary target of DENV infection [40,41], the secretion of MMP-9 from neutrophils induced by NS1 may also contribute to vascular leakage during DENV infection.
Interestingly, although it has been shown in previous studies that the concentrations of MMPs are increased in dengue patients [13,32] and MMP-9 upregulation is positively correlated with the disease severity and vascular leakage of dengue [31,32,42,43], we observed a significant increase in the serum level of MMP-9 only in dengue patients with warning signs, not in severe dengue patients. From the in vivo mouse study, we noticed that the secretion of MMP-9 occurred within a smaller time window than that of HPA-1 in mice after NS1 challenge. Therefore, it is possible that the discrepancy in the MMP-9 level in dengue patients between this and previous studies may be due to differences in the timing of sample collection. Because the specific day post-onset of symptoms that samples were collected was not available in the records of our dengue patients, we could not exclude the possibility of variation arising from different sampling times. Further study monitoring the sequential changes in the serum levels of MMP-9 and other glycocalyx-related molecules along with disease development is required to clarify their roles in dengue pathogenesis. From the results of the MMP antibody array, we also found that MMP-8 and tissue inhibitor of metalloproteinases 1 (TIMP-1) were upregulated by NS1-stimulated leukocytes. TIMP-1, which is a potent inhibitor of MMPs, can form a complex with pro-MMP-9 at a 1:1 stoichiometric relationship to inhibit its activation [44,45]. However, neutrophil elastase can inactivate TIMP-1 in the complex to free pro-MMP-9, such that it can be activated by MMP-3 [46]. In addition, myeloperoxidase, which is most abundantly expressed by neutrophils, can also inactivate TIMP-1 via generating hypochlorous acid [47]. These possible mechanisms may explain why MMP-9 activity was not abrogated in the presence of TIMP-1 in NS1-stimulated leukocytes.
A previous study has shown that NS1 can induce PBMCs to secrete IL-6 and IL-8 via Tolllike receptor 4 (TLR4), leading to vascular leakage [16]. However, in this study, we found that the secretion of IL-6 and IL-8 dropped rapidly after 3 h of NS1-stimulation in leukocytes (S7B and S7C Fig). In contrast, MIF steadily accumulated in the supernatant of leukocyte cultures after NS1 treatment, and the concentration of MIF was higher than that of IL-6 and IL-8 (S7 Fig). As NS1 needs at least 24 h to induce endothelial glycocalyx degradation (Fig 3A), we speculated that IL-6 and IL-8 are not very involved in NS1-induced endothelial glycocalyx degradation. This speculation is consistent with a recent study performed by Glasner et al., which found that DENV NS1 does not induce HMEC-1 human endothelial cells to secrete TNF-α, IL-6 or IL-8 and that blocking these cytokines does not affect DENV NS1-induced endothelial hyperpermeability [48]. On the other hand, the same study found that inhibition of HPA-1 prevents DENV NS1-induced endothelial hyperpermeability [48]; however, MIF was not measured. In our previous study and in this study, we demonstrated that NS1 induced HMEC-1 cells or HUMECs to secrete MIF, causing endothelial hyperpermeability [19]. In addition, we further demonstrated that both the secretion of HPA-1 and the shedding of CD138 induced by NS1-stimulation of endothelial cells are mediated by MIF.
Due to MIF regulating the secretion of both MMP-9 and HPA-1 and because CD138 shedding was also directly affected by MIF signaling, MIF may be an upstream regulator of DENV NS1-induced glycocalyx degradation. However, the mechanism of how MIF causes HPA-1 and MMP-9 secretion is still unclear. A previous study has shown that MIF induces MMP-9 expression in macrophages via the MAPK pathway [35]. MIF is also known to activate NF-κB signaling through binding to CD74 [49]. Additionally, HPA-1 mRNA expression is elevated in an NF-κB-dependent manner during hypoxia [50]. Therefore, it is possible that MIF contributes to the secretion of HPA-1 and MMP-9 via the MAPK/NF-κB pathway. However, from our in vivo study, we also noticed that NS1-induced MMP-9 secretion and CD138 shedding were significantly attenuated by MIF inhibition, whereas the attenuation of HPA-1 secretion was not as significant. Serum samples from severe dengue patients also showed no linear relationship between the concentrations of MIF and HPA-1 (S10 Fig). These results may suggest that in addition to MIF, other factors may participate in the regulation of HPA-1 secretion in vivo.
Taken together, our results suggest that NS1 may contribute to vascular leakage through different mechanisms during DENV infection. DENV NS1 may bind to the TLR4 of leukocytes, inducing the secretion of cytokines and MMPs, or it may directly bind to endothelial cells, inducing the secretion of HPA-1, both of which can cause glycocalyx degradation and subsequent vascular leakage. Consequently, NS1 may represent an important viral factor that causes vascular leakage and glycocalyx degradation during DENV infection. Indeed, antibodies against NS1 have been shown to be protective against DENV infection in mice [17,51,52]. Furthermore, MIF may represent the primary host factor that mediates NS1-induced glycocalyx degradation. Studies focusing on the development of neutralizing antibodies or small molecules against MIF may facilitate the development of drugs to prevent or treat severe dengue [53].

Experimental design
The aim of this study was to clarify the mechanism of DENV infection-induced endothelial glycocalyx degradation. From analyzing clinical samples, we correlated glycocalyx degradation to MIF secretion. By applying the results from other studies, we hypothesized that HPA-1 or MMP-9 was involved in MIF-mediated glycocalyx degradation in dengue. This hypothesis was examined via in vitro experiments, which were carried out by recombinant NS1 stimulation, as it was indicated as an important effector in severe dengue. Since the interaction between different cell types is critical under physiological conditions, we assessed the DENV NS1-induced effects on both endothelial cells and leukocytes. To further elucidate the involvement of MMPs in this mechanism, MMP antibody array and gelatin zymography assays were performed. Subsequently, recombinant NS1 was injected into mice systemically or locally to confirm the involvement of MIF, HPA-1 and MMP-9 in NS1-induced endothelial glycocalyx degradation and hyperpermeability in vivo.

Ethics statement
All research involving adult participants has been approved by the Institutional Review Board of NCKUH (IRB #B-ER-104-228). Informed written consent was not obtained from patients because the demographic and clinical information for the patients were delinked prior to analysis.
All animal studies were performed in accordance with the Guide for the Care and Use of Laboratory Animals (The Chinese-Taipei Society of Laboratory Animal Sciences, 2010) and were approved by the Institutional Animal Care and Use Committee (IACUC) of NCKU under the number IACUC 105018.

Patient samples
In this study, serum samples were collected at Clinical Virology Laboratory of NCKUH from dengue patients in the acute stage (days 0-7 after illness onset) of the disease during a DENV outbreak in Tainan, Taiwan, in 2015 [54]. All dengue patient samples were screened via a rapid combo test for NS1 antigen and antibody detection and were assessed by qRT-PCR to quantify the DENV viral load. Patients were categorized as having dengue with warning signs or severe dengue according to the 2009 WHO criteria for dengue severity. The characteristics of these clinical samples are shown in S1 Table. In addition, 26 serum samples from healthy donors were included as the negative control.
Human MIF recombinant proteins were produced as previously described [26]. Briefly, human MIF proteins were cloned, expressed in E. coli, and purified by Sepharose (GE Healthcare). Heparan sulfate and thrombin with protease activity were purchased from Sigma-Aldrich ( In addition, a rabbit anti-MIF polyclonal antibody (10 μg/ml) was used in this study and was purified from recombinant MIF-immunized rabbit serum using a protein G affinity column (GE Healthcare), as previously described [26]. To inhibit HPA-1, OGT 2115 (Tocris Bioscience, Bristol, UK) was used at the indicated concentration. To inhibit MMP-9, SB-3CT (Abcam, Cambridge, UK) and MMP-9 inhibitor I (Santa Cruz, Dallas, TX, USA) were used at the indicated concentrations. Control mouse and rabbit IgGs were purchased from LeadGene Biomedical (Taiwan).

Cells
HUVECs (Bioresource Collection and Research Center, Taiwan) were cultured in EGM-2 (Lonza, Basel, Switzerland), and THP-1 human monocytes (Bioresource Collection and Research Center, Taiwan) were cultured in Roswell Park Memorial Institute 1640 Medium (RPMI 1640; Thermo Fisher Scientific). Medium used to grow both cell types was supplemented with 10% fetal bovine serum (FBS; HyClone Laboratory, Logan, UT, USA), and cells were cultured at 37˚C in a 5% CO 2 atmosphere.
Human leukocytes (WBCs) were isolated from the whole blood of healthy donors. After collecting the blood into EDTA-containing plasma tubes, the whole blood was centrifuged at 1000 g for 5 min. The buffy coat was then collected and treated with red blood cell lysis buffer (Sigma-Aldrich, St. Louis, MO, USA). After one wash with PBS, the cells were cultured in serum-free RPMI 1640 at 37˚C in a 5% CO 2 atmosphere.
Human PBMCs were isolated from the whole blood of healthy donors using Ficoll-Paque (Sigma-Aldrich) according to the manufacturer's instructions. Briefly, blood was collected into EDTA-containing vacutainers (BD, Franklin Lakes, NJ) and transferred to the top layer of Ficoll-Paque. After centrifugation at 2500 g for 30 min, the PBMCs were collected and washed with RPMI 1640 twice, and then cultured in RPMI 1640 containing 10% FBS at 37˚C in a 5% CO 2 atmosphere.

NS1 stimulation of THP-1 cells, human leukocytes and PBMCs
THP-1 cells were suspended in medium containing 5 ng/ml PMA (Sigma-Aldrich). After 16 h, THP-1 cells were resuspended in fresh medium without PMA and incubated for another 8 h. NS1 (20 μg/ml) was used to stimulate THP-1 cells, human leukocytes and PBMCs, and the resultant culture supernatants were collected at the indicated time points.

Transwell permeability assay
A Transwell permeability assay was performed as described in a previous study [58]. HUVECs (2 x 10 5 ) were grown on a Transwell insert (0.4 μm; Corning Life Sciences, Corning, NY, USA) until a monolayer formed. The upper chambers were reconstituted with 20 μg/ml NS1, culture supernatant from NS1-activated THP-1 cells, or the inhibitor-containing medium. After 24 h, the upper chambers were reconstituted with 300 μl of serum-free media containing 4.5 μl of streptavidin-horseradish peroxidase (HRP; R&D Systems, Inc., Minneapolis, MN, USA). Next, 20 μl of medium in the lower chamber was collected 5 min after the addition of streptavidin-HRP and was assayed for HRP activity by the addition of 100 μl of 3,3',5,5'-tetramethylbenzidine (TMB) substrate (R&D Systems). The color development at 450 nm was measured with a VersaMax microplate reader (Molecular Devices, Sunnyvale, CA, USA).

Immunofluorescence staining
HUVECs were seeded as a monolayer onto a microscope cover glass slide and cultured under different conditions. After treatment for indicated time, the cells were fixed in 2% paraformaldehyde and then blocked with Superblock T20 (PBS) blocking buffer (Thermo Fisher Scientific).To measure the integrity of the endothelial glycocalyx and the deposition of CD138, the expression of sialic acid was stained with wheat germ agglutinin (WGA) lectin conjugated to FITC (WGA-FITC, Genetex) and the distribution of HPA-1 and CD138 was detected by antimouse-CD138 mAb (BD, Franklin Lakes, NJ, USA) or rabbit anti-HPA-1 polyclonal antibody (GeneTex). Primary antibodies were incubated with the fixed monolayer overnight at 4˚C, followed by incubation with Alexa 488-conjugated goat anti-mouse IgG secondary antibody, Alexa 594-conjugated goat anti-rabbit IgG secondary antibody (Invitrogen, Carlsbad, CA, USA) (1:500 diluted) and Hoechst 33342 (Invitrogen, Carlsbad, CA, USA) (1:3,000 diluted) for 1 h. Images were captured using a confocal microscope (Olympus FluoView FV1000, Melville, NY, USA).

Human MMP antibody array
The human MMP antibody array (Abcam) was used according to the manufacturer's instructions. Briefly, array membranes were incubated in equal quantities of the culture supernatant from PBS-or NS1-treated THP-1 cells or NS1-treated leukocytes for 24 h overnight at 4˚C. After washing with commercial wash buffer, the membranes were incubated with biotin-conjugated anti-MMP antibodies, followed by HRP-conjugated streptavidin. Bound HRP-conjugated antibodies were detected using the Luminata Crescendo Western HRP substrate (Merck Millipore, Darmstadt, Germany).

Gelatin zymography assay
MMP activity in the culture supernatant was assayed by gelatin zymography using 7.5% acrylamide gel containing gelatin [59]. Briefly, the culture supernatant of NS1-treated THP-1 cells or leukocytes was concentrated. Non-heat-concentrated culture medium samples were mixed with nonreducing sample dye and electrophoresed at 120 V for 90 min. The gels were subsequently renatured and developed before being stained with Coomassie blue to reveal the positions of active gelatinases (clear bands) against the undigested gelatin substrate in the gel.

DENV NS1-induced MIF, HPA-1, MMP-9 and CD138 secretion in mice
Mice were obtained from the animal center of NCKU. Before the injection of PBS or recombinant NS1, blood from 8-to 12-week-old BALB/c mice was collected by orbital sinus sampling with 10% citrate. Next, the mice were intravenously injected with 50 μg of NS1 or 100 μl of PBS. After the intravenous injection, blood from the mice was immediately collected by orbital sinus sampling and every 24 h thereafter until 120 h after the injection. The plasma concentrations of NS1, MIF, HPA-1, and MMP were analyzed by ELISA. For the peritoneal challenge, 500 μl of PBS, 50 μg of NS1, 50 μg of E or 50 μg of prM was injected intraperitoneally. After 24 h, the mice were sacrificed, and the abdominal cavities were washed with 5 ml of PBS. The resultant peritoneal lavage was collected, and the concentrations of MIF, HPA-1 and CD138 were quantified by ELISA.

IHC staining of CD138 in mice
To further confirm that NS1 induced CD138 shedding in endothelial cells in mice, 50 μg of recombinant NS1, E or prM protein or 50 μl of PBS was subcutaneously injected into 8-to 12-week-old BALB/c mice, followed by a second injection of an equal amount of recombinant proteins or PBS 24 h after the first injection at the same site. The mice were sacrificed 24 h after the second injection. The separated skin tissues were fixed in formalin overnight and embedded in paraffin for the preparation of a series of sections. After paraffin removal and antigen retrieval by citrate buffer, the tissue sections were blocked, and immunohistochemistry was performed using the Mouse/Rabbit HRP Detection System with DAB (brown) (BioTnA Biotech, Kaohsiung, Taiwan). Hematoxylin was used as a counterstain. Anti-α-SMA antibody (Arigo, Hsinchu City, Taiwan) was used at 1:200, and anti-CD138 antibody (BD, Franklin Lakes, NJ) was used at 1:100. The resultant images were acquired using phase-contrast microscopy (Olympus, Tokyo, Japan).

Statistical analysis
The patients' sera data were expressed as the median ± interquartile range and tested if the values come from a Gaussian distribution by using D'Agostino and Pearson omnibus normality test. If the data meet Gaussian distribution, the significance of differences between each groups was analyzied using One-way ANOVA with Tukey's method. If the data do not meet the assumptions of normality, they were analyzed with a non-parametric test by Kruskal-Wallis test. The in vitro and in vivo data are expressed as the mean ± standard deviation (SD) from more than three independent experiments. Student's t-test was used to analyze the significance of differences between the test and control groups. One-way ANOVA with Kruskal-Wallis comparison test was used to analyze the significance of differences between multiple groups. All data were analyzed by GraphPad Prism 5 software. P values <0.05 were considered statistically significant.
Supporting information S1 Table. Characteristics of dengue patients. BALB/c mice were intravenously injected with Evans Blue dye, followed by the subcutaneous injection of PBS or different doses of NS1, NS1 with MMP-9 inhibitor I or NS1 with ISO-1 for 6 h. After 5 h, the mice were subcutaneously injected with thrombin as a positive control. After another hour, the mice were sacrificed, and skin samples were collected and processed.