Secreted NS1 of Dengue Virus Attaches to the Surface of Cells via Interactions with Heparan Sulfate and Chondroitin Sulfate E

Dengue virus (DENV) nonstructural protein-1 (NS1) is a secreted glycoprotein that is absent from viral particles but accumulates in the supernatant and on the plasma membrane of cells during infection. Immune recognition of cell surface NS1 on endothelial cells has been hypothesized as a mechanism for the vascular leakage that occurs during severe DENV infection. However, it has remained unclear how NS1 becomes associated with the plasma membrane, as it contains no membrane-spanning sequence motif. Using flow cytometric and ELISA-based binding assays and mutant cell lines lacking selective glycosaminoglycans, we show that soluble NS1 binds back to the surface of uninfected cells primarily via interactions with heparan sulfate and chondroitin sulfate E. DENV NS1 binds directly to the surface of many types of epithelial and mesenchymal cells yet attaches poorly to most peripheral blood cells. Moreover, DENV NS1 preferentially binds to cultured human microvascular compared to aortic or umbilical cord vein endothelial cells. This binding specificity was confirmed in situ as DENV NS1 bound to lung and liver but not intestine or brain endothelium of mouse tissues. Differential binding of soluble NS1 by tissue endothelium and subsequent recognition by anti-NS1 antibodies could contribute to the selective vascular leakage syndrome that occurs during severe secondary DENV infection.


Introduction
Dengue hemorrhagic fever and dengue shock syndrome (DHF/DSS) are severe and potentially fatal complications of infection by dengue virus (DENV), a mosquito-borne RNA virus of the Flaviviridae family. Globally, DENV infects 25 to 100 million people per year, but the life-threatening complications primarily occur in school-age children [1]. Four serotypes of DENV exist, and DHF/DSS is commonly associated with secondary infection with a different virus serotype [2,3]. In the most severe cases, clinical deterioration is characterized by a rapid decline after several days of continuous high fever, thrombocytopenia, and selective vascular leakage at serosal sites [4]. The vascular leakage syndrome results in hemoconcentration, pleural effusions and ascites, and hypotension [4]. An effective strategy for disease prevention or treatment is currently lacking.
The pathogenesis of DHF/DSS reflects a complex interplay of the host immune response and the viral determinants of virulence [5][6][7]. A model of immunopathogenesis has been suggested based on an increased risk of DHF with secondary infection and in children within the first year of life born to DENV-immune mothers [8,9]. From this, the hypothesis of antibody-dependent immune enhancement of infection emerged. In support of it, enhancement of DENV infection in monocytes in vitro with pre-illness serum correlates with increased risk of DHF in vivo [9,10], and peak viremia is higher in patients with severe secondary DENV infection [11][12][13]. Differences in specific genetic determinants among viral isolates [14][15][16] also modulate virulence, as some DENV strains fail to cause severe disease [17,18]. A pathologic cytokine response that occurs after extensive T cell activation also likely contributes to the syndrome [5]. Elevated levels of cytokines including IFN-a, TNF-a, and IL-10 correlate with severe disease [19][20][21][22][23][24], and the expansion of cross-reactive low-affinity DENV-specific T cells produces vasoactive cytokines [25][26][27]. Finally, accumulation of soluble NS1 in serum also correlates with disease severity and is believed to contribute to changes in vascular permeability through antibody-dependent activation of the complement cascade [28].
NS1 is a ;48-kDa glycoprotein that is absent from the infectious viral particle. NS1 is, however, an essential gene within infected cells, as it functions as a cofactor for viral RNA replication, colocalizing with the double-stranded RNA replicative form [29,30]. NS1 is synthesized in an infected cell as a soluble monomer and rapidly dimerizes after posttranslational modification in the lumen of the endoplasmic reticulum, with subsequent transport to the cell surface and release into extracellular milieu [31]. In solution, secreted NS1 behaves as a hexamer [32] and accumulates in serum in high amounts (up to 50 lg/ml) [11,28,33]. The mechanism(s) by which soluble and cell surface-associated NS1 contributes to flavivirus pathogenesis remains uncertain. It has been proposed to facilitate immune complex formation [28,34], elicit auto-antibodies that react with platelet and extracellular matrix proteins [35,36], cause endothelial cell damage via antibody-dependent complement-mediated cytolysis [35,[37][38][39], directly enhance infection [40], and attenuate the alternative pathway of complement activation by binding factor H [41].
The mechanism by which NS1 becomes associated with the plasma membrane is poorly understood. The protein lacks amino-or carboxyl-terminal membrane-spanning or anchoring sequences [42]. Although covalent linkage by a glycosylphosphatidylinositol (GPI) anchor to the plasma membrane of DENV-infected cells has been suggested [43], this mechanism has not been confirmed. Also, it does not explain how uninfected, as well as infected cells, accumulate NS1 on their surfaces. In this study, we demonstrate that soluble DENV NS1 attaches to uninfected cells primarily via an interaction with the glycosaminoglycans (GAG) heparan sulfate (HS) and chondroitin sulfate E (CS-E), and that GAG sulfation is a critical cellular modification necessary for binding. Moreover, soluble DENV NS1 preferentially binds subsets of cells including human microvascular endothelial cells. Our findings suggest that the selective vascular leakage that occurs in severe DENV infection may be related to the relative ability of endothelial cells in different tissues to bind soluble NS1 and be targeted by cross-reactive anti-NS1 antibodies during secondary infection.

Results
Flavivirus NS1 is a secreted nonstructural glycoprotein that lacks a membrane-spanning region yet becomes cell surfaceassociated [31,42]. As NS1 attachment to the cell surface may be important for immune recognition [28,44] and immune evasion [41], we set out to identify the mechanism by which soluble NS1 binds to the surface of cells. Although it has been suggested that DENV NS1 attaches to the plasma membrane via a GPI anchor [43], this is unlikely to explain how DENV NS1 binds to uninfected cells. To assess this issue in greater detail, we developed a flow cytometric-based NS1 cell surface binding assay. We purified soluble DENV NS1 from the supernatants of baby hamster kidney (BHK) cells that stably express a DENV-2 subgenomic replicon (BHK DENV-2 Rep cells [45]). These cells produce high levels of intracellular ( Figure 1A) and soluble NS1 as determined by a capture ELISA (3.6 6 0.2 lg/ml of supernatant) and western blot (unpublished data). Elution profiles of purified NS1 after immunoaffinity or ion exchange chromatography are shown ( Figure 1B and 1C, respectively). The yield of purified NS1 from BHK DENV-2 Rep cells ( Figure 1D) was similar to that previously obtained from DENV-2-infected cell supernatants [28] and retained immunoreactivity as judged by western blot (Figures 1E and S1) or ELISA with a panel of conformationally sensitive anti-NS1 mAbs (unpublished data).

DENV NS1 Selectively Binds to Uninfected Cells of Specific Lineage
Uninfected cells were tested for the ability to bind soluble DENV NS1. Incubation with either purified DENV NS1 or supernatants from BHK DENV-2 Rep cells resulted in rapid binding of NS1 to many cell types ( Figure 2 and Table 1). Binding of NS1 to Chinese hamster ovarian epithelial (CHO)-K1, Vero, and 4/4 RM4 cells was dose-dependent and saturable, and maximum binding was achieved at a concentration of 20 lg/ml, which is in the range reported in vivo during secondary infection [11,28]. DENV NS1 bound to the surface of several types of epithelial and fibroblast transformed cell lines (BHK, CHO-K1, Vero, 293T, HepG2, Hep3B, and L929) including those of human and nonhuman origin. NS1 also bound to primary, untransformed cells including keratinocytes (HaCat, CCD-1102), skin and lung fibroblasts (Detroit-551 and IMR-90), and freshly isolated tonsillar epithelial cells.
As high levels of soluble NS1 are detected in the blood of DHF patients [11,28,33], we extended our analysis to human peripheral blood cells. Notably, DENV NS1 failed to bind to the surface of freshly isolated peripheral blood mononuclear cells or erythrocytes. Similar negative results were obtained with purified CD14 þ monocytes, CD19 þ B lymphocytes, and CD4 þ T lymphocytes. Moreover, DENV NS1 only weakly

Author Summary
Dengue virus (DENV) is a mosquito-transmitted virus that infects humans and has become a global emerging infectious disease threat. Four serotypes of DENV exist, and the most severe cases are associated with secondary infection with a different virus serotype. Clinical deterioration is characterized by bleeding and selective vascular leakage from endothelium in specific tissue sites. An increased understanding of how DENV proteins contribute to this phenotype is vital to developing novel vaccines and identifying individuals at risk for severe disease. DENV nonstructural protein-1 (NS1) is one such protein: during infection, it is secreted and accumulates in the supernatant and on the surface of cells. In this study, we demonstrate that soluble DENV NS1 attaches to subsets of cells, including some but not all endothelial cells, primarily via an interaction with specific glycosaminoglycans (heparan sulfate and chondroitin sulfate E). This was confirmed in tissue binding studies as DENV NS1 bound to lung and liver but not intestine or brain endothelium. Our findings suggest that the selective vascular leakage that occurs in severe DENV infection may be related to the relative ability of endothelial cells in different tissues to bind soluble NS1 and to be targeted by cross-reactive anti-NS1 antibodies during secondary infection.
bound to the surface of monocyte-derived macrophages. In contrast to that observed with primary lymphocytes, DENV NS1 bound strongly to the surface of several malignant T cell lines, including Jurkat, H9, and EL-4 ( Figure 2 and unpublished data).
As vascular leakage is a hallmark of DHF/DSS, and endothelial cells are believed to be targets of immunemediated damage, we analyzed the binding of NS1 to human endothelial cells. Interestingly, DENV NS1 bound strongly to human dermal and lung microvascular endothelial cells (HMEC) and HMEC-lung blood (HEMC-LB), modestly to aortic endothelial cells, but minimally to primary or  (19)(20)(21) were combined and passed over an ion-exchange column, and purified NS1 was eluted (C) with a linear NaCl gradient (diagonal line

Soluble DENV NS1 Binds to GAG
As our data indicated that soluble DENV NS1 bound a subset of uninfected mammalian cells of different lineages, we sought to identify the mechanism of attachment. We hypothesized that NS1 might interact with a highly conserved moiety, such as a GAG. To test this, we compared DENV NS1 binding to wild-type CHO-K1 cells and seven different CHO cell lines ( Table 2) that are either defective in GAG biosynthesis or express GAG with distinct structural specificities. CHO-745 cells, which genetically lack xylosyltransferase, the enzyme required for biosynthesis of both HS and CS [46], showed a 60%-70% reduction (p , 0.0001) in binding of DENV NS1 ( Figure 3A and Table 2). Similarly, CHO-M1 cells, which are defective in HS biosynthesis [47], exhibited a ;50% reduction (p ¼ 0.02) in DENV NS1 binding. However, CHO-H8 cells [48], which are a variant of CHO-M1 HS-deficient cells overexpressing CS 2-O-sulfotransferase (2-OST) (producing a 30-fold increase in double sulfated disaccharide residues on CS), showed levels of DENV NS1 binding that were similar to CHO-K1 wild-type cells. Thus, both HS and CS sustain cell surface binding to DENV NS1, at least on CHO cells.
On CHO cell lines, the specificity of the GAG had a more dominant effect on regulating NS1 binding than the absolute level. Using a previously developed high-pressure liquid chromatography method [51], we quantified the amount of cell surface HS and CS GAG in different CHO cell lines and compared it to NS1 binding ( Table 2). Although CHO-M1 cells expressed twice the amount of total GAG compared to wild-type CHO-K1 cells, CHO-M1 cells bound NS1 poorly. Thus, the total level of GAG did not directly correlate with NS1 binding. Consistent with this, similar amount of GAG were observed on CHO-K1 and other 3-OST-expressing CHO cell lines, yet 3-OST-expressing CHO cells bound NS1 more strongly. Moreover, Vero cells express ;3-fold higher levels of HS compared to CHO-K1 cells (unpublished data) but bound NS1 less well compared to CHO-K1 cells ( Figure 2 and Table 2).
A requirement of GAG for DENV NS1 binding to the cell surface was confirmed independently by enzymatic treatment of BHK and CHO-K1 cells with specific heparin lyases and chondroitinases ( Figure 3B). Treatment of BHK or CHO-K1 cells with heparin lyases I, II, and III, which specifically remove HS, reduced DENV NS1 binding (;50%-60%, p , 0.05) to the levels observed in GAG-negative CHO-745 cells.
In contrast, DENV NS1 attachment was not significantly affected by treatment with chondroitinase ABC, which degrades all types of CS [52]. As expected, treatment with both enzymes had no effect on DENV NS1 binding to CHO-745 cells, which lack GAG. Nonetheless, a small amount of NS1 binding to CHO-745 cells was observed, presumably through a subordinate GAG-independent pathway. Overall, the analysis of CHO cell mutants and enzymatic treatments suggested that HS and highly sulfated forms of CS were the primary cell surface GAG that sustained DENV NS1 binding on CHO and BHK cells.
As an additional confirmation, we performed competitive binding assays with soluble GAG. Pre-incubation with soluble CS-E or heparin (HP) substantially decreased (;50%, p , 0.005), in a dose-dependent manner, DENV NS1 binding to BHK cells ( Figure 4A-4C). Small, albeit significant inhibition (;15%, p , 0.05), was observed even at relatively low concentrations CS-E and HP (0.01 lg/ml or 0.033 nM). In contrast, CS-A and -B inhibited NS1 binding only at higher concentrations (100 lg/ml or 3.3 lM, p , 0.05) whereas CS-C and -D showed no significant effect (p . 0.2). Similarly, preincubation with CS-E and HP reduced DENV NS1 attachment to HMEC cells (;40%, 0.01 lg/ml, p , 0.01) ( Figure 4D-4F). In a more direct in vitro binding ELISA, an interaction between DENV NS1 and HS or CS-E was established, whereas only weak or no appreciable binding was detected with CS-A, -B, -C, and -D ( Figure 5). Taken together, these results suggest that NS1 interacts more strongly with HP, HS, and CS-E and more weakly with CS-A and CS-B. CS-A consists of mainly GlcA-GalNAc4S repeating disaccharides, CS-B contains both IdoA-GalNAc4S and GlcA-GalNAc4S as dominant repeating disaccharides, and CS-C consists of mainly GlcA-GalNAc6S repeating disaccharides. CS-D and CS-E are enriched in disulfated disaccharides, GlcA2S-GalNAc6S and GlcA-Gal-NAc4S6S, respectively. CS-E is unique, as two sulfates are present in the same GalNAc residue compared to other CS, and suggests that clustered sulfates rather than net negative charge on the disaccharide backbone facilitate DENV NS1 binding.

GAG Sulfation Is Critical for DENV NS1 Cellular Attachment
The specificity of GAG-protein interactions often depends on the position and degree of sulfation (reviewed in [53][54][55]). To clarify if GAG sulfation is required for cell binding of DENV NS1, we treated BHK and CHO-K1 cells with the reversible sulfation inhibitor, sodium chlorate [56] ( Figure 6A and 6B). This treatment did not alter cell viability, as analyzed by propidium iodide exclusion (unpublished data). However, binding of DENV NS1 to BHK cells was abolished by the addition of sodium chlorate (.5 mM). In contrast, at comparable doses, DENV NS1 binding to wild-type CHO-K1 was reduced to the level observed with CHO-745 cells lacking GAG. Importantly, the phenotype was reversed by the addition of excess exogenous (10 mM) sodium sulfate ( Figure  6C and 6D). Thus, GAG sulfation is required for optimal binding of soluble DENV NS1 to cells.

Expression of DENV NS1 on the Surface of DENV-Infected Cells Is Largely GAG-Independent
Although our data suggests that soluble DENV NS1 binds to the cell surface of uninfected cells via HS and CS-E, we evaluated if this was the dominant mechanism of attachment of NS1 to DENV-infected cells. Previous studies have suggested that flavivirus NS1 may be displayed on the cell surface of infected cells by a GPI anchor or transmembrane linkage [43], even though the protein lacks anchor attachment sequences or consensus membrane spanning domains. To evaluate this, BHK cells were infected with DENV-2 and cultured in the presence of high concentrations of soluble HP. No reduction of DENV NS1 surface expression on infected cells was observed ( Figure 7A). These results were confirmed by studies with the sulfate inhibitor, sodium chlorate. In contrast to the inhibitory effect of sodium chlorate on soluble NS1 binding ( Figure 6), no reduction of DENV NS1 expression on the surface of infected BHK cells was observed ( Figure 7B). Similar studies with BHK DENV-2 Rep cells also showed no change in surface NS1 expression after sodium chlorate treatment ( Figure 7C). Finally, in direct contrast to that observed with soluble NS1 and uninfected cells ( Figure 3B), treatment of DENV-infected cells with heparin lyases and/or chondroitinase ABC did not decrease HS: 2 6 0.1 CS: CHO cell variants were cultured as described in Materials and Methods. Surface GAG levels were determined by a published high-pressure liquid chromatography method [51]. The data are derived from between one and five independent experiments and the standard deviations were from varying numbers (n) of independent GAG isolations. b DENV NS1 binding was performed as described in Figure 2. surface expression of NS1 ( Figure 7D). Thus, NS1 expressed on the surface of infected cells is not modulated by GAG expression and must attach by an independent mechanism.

Soluble DENV NS1 Binds Selectively to Endothelium In Situ
Soluble NS1 has been hypothesized to contribute to DHF pathogenesis by promoting vascular leakage [28,34], which occurs predominantly into pleural and peritoneal cavities [4].
Our cell culture experiments indicated that soluble DENV NS1 binds selectively to subsets of human endothelial cells. We hypothesized, that in vivo, a preferential interaction of DENV NS1 with specific endothelium could contribute to tissue-specific vascular leakage after immune recognition. This mechanism would not require direct DENV infection of endothelial cells, which has been difficult to establish in vivo by pathological criteria [57,58].
To assess this, sections of uninfected mouse tissues were incubated with soluble DENV NS1 and analyzed by immunofluorescence and confocal microscopy. Histological analysis was performed on parallel hematoxylin and eosin-stained sections (unpublished data). Specific NS1 binding to endothelial cells lining blood vessels of lung and liver was observed, based on costaining with mAbs against DENV NS1 and the endothelial cell-specific marker CD31 ( Figure  8A and 8B). Differences in the pattern of DENV NS1 binding within the same organ were also observed. For example, DENV NS1 bound primarily to endothelial cells lining vessels along the bronchial tree, yet bound poorly to those lining alveolar capillaries. Strong binding of DENV NS1 to the cells lining the outer layer of the adventitia of pulmonary vessels was also apparent ( Figure 8A). In the liver, hepatic arteriolar and sinusoidal endothelial cells bound DENV NS1 weakly, whereas those within central veins showed strong binding ( Figure 8B). As controls, no appreciable staining was observed if sections were incubated with bovine serum albumin (BSA) followed by anti-NS1 mAbs or with purified DENV NS1 and isotype control mAbs. The selectivity of soluble NS1 binding to endothelium in situ was also demonstrated by an absence of NS1 binding to endothelium in the large intestine ( Figure  8C) and brain (unpublished data) of mice. Finally, binding of DENV NS1 to endothelium in situ also was confirmed in human lung tissues ( Figure 9).
Next, soluble GAG and DENV NS1 competitive binding assays were performed on lung tissues and images were analyzed by confocal microscopy and Volocity software. Preincubation of DENV NS1 with soluble CS-E or HP reduced binding to lung endothelium in situ (CS-E: ;40%, p , 0.0001; HP: ;30%, p ¼ 0.0002). In contrast, CS-C marginally decreased DENV-NS1 attachment to lung endothelium (;8%, p ¼ 0.18). Significant binding of DENV NS1 to serosal surfaces of the lung (pleura) and intestine (peritoneum) was also observed ( Figure S2A and S2B). Based on histological analysis, mesothelial cells were also targets for DENV NS1 binding in the lung and intestine. Binding of DENV NS1 to mesothelial cells was specific as negative results were obtained when the sections were incubated with NS1 followed by isotype control mAbs or without NS1 followed by anti-NS1 mAbs. In competitive binding experiments, soluble HP and CS-E, but not CS-C, also decreased DENV NS1 binding to lung mesothelial cells (unpublished data).

Discussion
In this study, we demonstrate that soluble DENV NS1 binds to a subset of uninfected cells via interactions with GAG, primarily HS and CS-E. In cell culture, NS1 bound strongly to epithelial cells and fibroblasts, and weakly, if at all, to freshly isolated human peripheral blood leukocytes. Substantial variability was observed in NS1 binding to cultured endothelial cells and endothelium in situ. Experiments with the sulfation inhibitor sodium chlorate established that highly sulfated forms of GAG are required for optimal binding of soluble DENV NS1. Finally, our experiments suggest that NS1 on the surface of DENV-infected cells is linked primarily by a distinct, GAG-independent mechanism.
Two major types of GAGs, HS and CS, are produced by cells in the form of proteoglycans as linear polymers of repeating disaccharides of uronic acids and glucosamines or galactosamines. For HS, sulfation of glucosamine and uronic acid moieties may occur in a clustered manner to generate highly sulfated domains or in a dispersed pattern to generate less sulfated or non-sulfated domains. Moreover, GlcNS, IdoA, and GlcA can be sulfated at multiple positions. Differential sulfation of GAG results in preferential binding of growth factors, cytokines, chemokines, enzymes, extracellular matrix, and other proteins to these structures [59]. Cell type-specific expression of GAG modifying enzymes and proteoglycan core proteins results in the display of unique GAG structures on different types of cells [60]. Consistent with this, our studies demonstrate that DENV NS1 binding to GAG occurs in a cell type-specific manner.
The specificity of GAG interactions is determined by the pattern of the disaccharide units, the degree of sulfation, and the spacing of basic amino acid residues in GAG-binding domains of ligands (reviewed in [53][54][55]). We observed that highly sulfated GAG, including HS, HP, and CS-E, demonstrated the strongest binding for soluble DENV NS1. The use of mutant CHO lines expressing different levels and forms of GAG established that specific HS and CS structures modified by 2-O and 3-O-sulfotransferases promoted DENV NS1 binding to cell surfaces. These results are analogous, although not identical, to studies with HSV glycoprotein D, which show enhanced cell surface binding when HS is modified by some, but not all 3-OST [50,61]. Sulfation was critical for soluble DENV NS1 binding to cell surfaces, especially for BHK cells,  Figure 2). Data are expressed as the specific MFI of staining as described in Figure 3. The error bars indicate SD corresponding to three independent experiments. Asterisks denote NS1 binding that is statistically different from untreated cells (* p , 0.05, ** p , 0.005). Examples of histogram profiles in which soluble HP and CS-E at 1 lg/ml concentration block DENV NS1 binding to the surface of (B-C) BHK and (E-F) HMEC are depicted. doi:10.1371/journal.ppat.0030183.g004 as treatment with the sulfate inhibitor sodium chlorate abolished DENV NS1 attachment.
Also consistent with our results, CS-E, but not CS-A, -B, or -C, interfered with HSV binding to the target cells [62]. Using radioactive binding or gel mobility shift assays, we observed similarly strong binding between West Nile virus NS1 and HS or CS-E (L. Zhang, K. Chung, and M. Diamond, unpublished data). As binding of soluble DENV NS1 to GAG requires sulfation, it is plausible that the interaction is primarily electrostatic in nature and depends on the relative degree of negative charge on a GAG motif. The major repeating disaccharides in HP are IdoA2S-GlcNS6S (;2.7 sulfates/ disaccharides), whereas the major repeating disaccharides in CS-D and CS-E are GlcA2S-GalNAc6S and GlcA-GalNAc4S6S, respectively (;1.7 sulfates/disaccharide) [62]. Because NS1 preferentially interacts with CS-E and HP compared to CS-D, we hypothesize that NS1 binding may prefer closely spaced sulfates within the same sugar residue. Since HP and CS-E have distinct sugar sequences, the combination of disaccharide unit and position of sulfation on an individual GAG may determine the strength of DENV NS1 binding.
The linear sequences XBBXBX and XBBBXXBX (where B is a basic Arg or Lys amino acid) are common HP binding motifs in proteins [55]. However, GAG binding sites may not be exclusively defined by linear sequences but also can include conformational epitopes that juxtapose basic amino acids from different segments of a protein [55]. Amino acid sequence analysis of DENV NS1 reveals no apparent canonical GAG binding motifs. Although reverse genetic strategies are planned, the identification of the amino acids involved in GAG recognition may await solution of the NS1 structure.
Our data showing that soluble NS1 can bind to the surface of the GAG-deficient line, CHO-745, albeit at significantly lower levels, suggest that additional molecules also serve as ligands for DENV NS1 attachment. This secondary attachment ligand may be expressed only on subsets of cells as addition of sodium chlorate completely abrogated soluble NS1 binding to BHK cells. Consistent with this, soluble HS and CS-E only partially reduced DENV NS1 binding to HMEC. Thus, different cell types may express multiple ligands for attachment of soluble DENV NS1.
The expression of NS1 on the surface of DENV-infected cells was insensitive to treatments that reduce GAG levels. This also suggests the existence of an alternate mechanism of cell surface NS1 attachment. Some have speculated that a transmembrane form of NS1 exists, although the protein sequence lacks a canonical hydrophobic membrane-spanning domain [42]. Others have postulated that at least some fraction of cell surface NS1 is linked via a GPI anchor [43,63]. Based on our data, we hypothesize that DENV NS1 is expressed on cell surfaces by at least two mechanisms: on uninfected cells soluble NS1 binds to the surface via a GAGdependent and GAG-independent mechanism, whereas on infected cells NS1 attaches via a GAG-independent mechanism, possibly via a membrane or GPI anchor. What remains uncertain, and is a direction for future research, is whether the mechanism of cell surface attachment of NS1 has unique functional consequences, especially in terms of immune recognition or evasion.
Our studies demonstrate that soluble DENV NS1 differentially binds to cultured endothelial cells in vitro and endothelium in situ in human and mouse tissues. In contrast, others have shown that intravenous injection of C57BL/6 x SJL mice with high concentrations (250 lg/ml) of soluble DENV-1 NS1 results in binding to and accumulation in hepatocytes but not to other cell types in the liver or other tissues [40]. A possible reason for the disparity in results is that the route of antigen administration may modulate NS1 binding: intravenous injection could prompt rapid first-pass clearance of antigen by the liver. Although further studies are necessary, our data are consistent with specific GAG modification by subsets of endothelial cells in different tissues modulating the level of bound NS1. Indeed, tissuespecific expression of different isoforms of enzymes in GAG biosynthesis has been reported [64,65]. Selective IL-8 binding to endothelial cells has been observed [66] and could be due to subtle differences in the display of GAG on the surface of cells. In favor of this hypothesis, IL-8 selectively binds to subsets of HP and HS [67].
The pathologic mechanism underlying selective vascular leakage at serosal sites during DHF/DSS [4] remains unknown. High levels of intravascular soluble NS1, as observed in DENV-infected patients, could promote binding and surface expression of NS1 on selective endothelium without a requirement for direct viral infection, which has been difficult to establish histopathologically in fatal DHF cases [57,58]. In addition, specific binding to mesothelial cells that line pleura and peritoneum, as observed in our DENV NS1 binding experiments in situ, could contribute to the pleural effusions or ascites that are observed in DHF/DSS patients. Although more experiments are necessary, preferential binding of soluble NS1 to subsets of endothelial and mesothelial cells in vivo could lead to tissue-specific vascular leakage that occurs during severe secondary DENV infection after recognition by anti-NS1 antibodies, immune complex formation, and inflammatory damage [28,68].
Primary human tonsil epithelial cells were isolated according to a published protocol [71] and cultured in keratinocyte serum-free medium supplemented with keratinocyte growth factor (Invitrogen). Primary HMEC-LB and aortic endothelial cells were purchased from Clonetics (Cambrex Bio Science) and maintained according to the manufacturer's protocol. HUVEC were grown in RPMI 1640 containing 10% FBS on 1% gelatin-coated surfaces. Primary human peripheral blood mononuclear cells were isolated from buffy coats obtained from the blood bank or healthy volunteers by Ficoll-Hypaque (Pharmacia) density gradient centrifugation. Human CD14 þ monocytes, CD19 þ B cells, CD4 þ T lymphocytes were purified from primary human peripheral blood mononuclear cells via positive selection using antibody-coated magnetic beads (Miltenyi Biotec). To generate macrophages, monocytes were cultured for 7 d in RPMI 1640 supplemented with 50 ng/ml of GM-CSF, 10% FBS, 2 mM glutamine, 1% nonessential amino acids, 1% sodium pyruvate, 100 units/ml penicillin G, and 100 lg/ml streptomycin sulfate. Human platelets were isolated from whole blood of healthy volunteers. After centrifugation at 750 x g for 20 min at 22 8C, platelet-rich plasma was collected and centrifuged at 1,200 x g for 15 min at 22 8C to obtain platelets. Human erythrocytes were obtained from healthy donors following a published protocol [72]. All primary cells were cultured for five or fewer passages for the NS1 binding experiments.
Purification of DENV NS1. BHK cells that stably propagate a DENV-2 subgenomic replicon (BHK DENV-2 Rep cells [45]) were grown to 80%-90% confluence in DMEM containing 10% FBS and 3 lg/ml puromycin (Sigma). Cell monolayers were washed several times with DMEM and the cells were cultured for another 3 d in serum-free DMEM supplemented with 3 lg/ml puromycin. DENV NS1 secreted in the supernatants of BHK DENV-2 Rep was quantified by NS1 capture ELISA as previously described [28]. Supernatants were collected, centrifuged, pooled, and passed through a 0.2-lm filter prior to immunoaffinity chromatography with anti-NS1 mAb 2G6 [28]. Peak elution fractions were combined, diluted 7-fold with 20 mM Tris (pH 8.0), and loaded onto a 1-ml Mono Q ion exchange column (GE Healthcare) at a rate of 1 ml/min. The protein was eluted with a linear salt gradient (0-1 M NaCl) over 20-column volumes at 1 ml/min. Purity and immunoreactivity of DENV NS1 were confirmed by SDS-PAGE with silver staining and western blot. Concentrations of purified protein were determined by the bicinchoninic acid assay. Purified DENV NS1 from DENV-infected mammalian cells was obtained as previously described [28].
DENV NS1 binding to cells. Adherent cells were removed from tissue culture plates after incubation with an EDTA solution (4 mM EDTA plus 10% FBS in PBS). Cells (5 3 10 5 ) in suspension were incubated on ice for 1 h with 100 ll of purified DENV NS1 at indicated concentrations or 300 ll of serum-free supernatants from BHK DENV-2 Rep cells. After washing once with 3 ml of medium, 50 ll of DENV-2 NS1 specific mAb 2G6 (25 lg/ml) or an isotype-matched negative control mAb was added to the cells and incubated on ice for 45 min. After subsequent washing, bound primary mAbs were detected after 30-min incubation with a 1:500 dilution of Alexa Fluor 647-conjugated anti-mouse IgG (Invitrogen). Of note, experiments were also performed in parallel with serial washing steps and no significant difference in binding or background was observed. Propidium iodide (0.2 mg/ml) was added immediately before flow cytometry to exclude dead cells. Soluble GAG and DENV NS1 competitive binding assays. Cell suspensions were incubated with DENV NS1 in the presence or absence of varying concentrations of soluble GAG for 1 h on ice. Bound DENV NS1 was detected after antibody staining as described above and analyzed by flow cytometry.
Enzymatic digestion of GAGs from cellular surfaces. Cell suspensions were centrifuged at 200 x g for 2 min and were resuspended in PBS containing 0.1% BSA. Heparin lyases I, II, III (0.06 U/ml final concentration) or chondroitinase ABC (0.1 U/ml final concentration) was added to a 50-ll of cell suspension and incubated at 37 8C in a shaking incubator for 1 h. After three washes with cold DMEM, the cells were incubated sequentially with DENV NS1 and DENV NS1 mAbs and analyzed by flow cytometry.
Quantification of surface GAG levels. Different types of CHO cells were cultured as described above. CHO cells were rinsed and detached from cell culture dishes after incubation in PBS/EDTA for 15 min. Cells (10 7 ) were treated with 2 ml trypsin-EDTA (1X solution, Mediatech) at 37 8C for 10 min, which released cell surface proteoglycans. These were collected and used for cell surface GAG isolation and quantification by high-pressure liquid chromatography as reported previously [51].
DENV NS1 and GAG binding ELISA. Maxi-Sorp microtiter plates (Nalge Nunc International) were adsorbed with soluble GAG (1 mg/ml in PBS) at 4 8C overnight. After four washes with PBS (300 ll/well), nonspecific binding sites were blocked with 1% heat-inactivated (65 8C, 15 min) BSA in PBS for 2 h at 37 8C and followed by five washes with PBS. Clarified serum-free supernatants from BHK or BHK DENV-2 Rep cells (100 ll diluted 1:1 in PBS) were added to each well and incubated for 2 h at room temperature. Plates were then washed five times with PBS containing 0.05% Tween-20 followed by a 1-h incubation at room temperature with 100 ll of purified DENV NS1 specific mAb mixtures (1A4, 1B2, 1F11, 2G6, and 2E3; 1 lg/ml of each Figure 8. Binding of DENV NS1 to Mouse Tissues Cryo-sections of mouse (A) lung, (B) liver, and (C) intestine were incubated with serum-free supernatants from BHK DENV-2 Rep or BHK cells for 1 h at room temperature. After extensive washing, bound NS1 was detected by a mixture of NS1 mAbs (1A4, 1F11, 2G6, 1B2) followed by Cy3-conjugated goat anti-mouse IgG. Co-staining with endothelial cell marker was subsequently performed by incubating the sections with rat anti-mouse CD31 (PECAM-1) followed by Alexa Fluor 488-conjugated goat anti-rat IgG. Nuclei were stained with a DNA-specific dye TO-PRO-3. Sections incubated with DENV NS1 followed by an isotype control Ab served as a negative control. Analysis was performed by confocal microscopy. White arrow and yellow arrowhead denote the layer of endothelial cells in the lumen and the outer layer of the adventitia of pulmonary vessel, respectively. doi:10.1371/journal.ppat.0030183.g008 Figure 9. Binding of DENV NS1 to Human Lung Cryo-sections of human lung tissue were incubated with 20 lg/ml of purified DENV-2 NS1 or BSA for 1 h at room temperature. Bound NS1 was detected by staining with a mixture of DENV NS1 mAbs (1A4, 1F11, 2G6, 1B2) followed by Alexa Fluor 488 conjugated with goat anti-mouse IgG. Subsequently, the sections were co-stained with rabbit anti-human CD31 (PECAM-1) followed by Cy3-conjugated donkey anti-rabbit IgG. Nuclei were stained with a DNA-specific dye (Hoechst) and the sections were analyzed by confocal microscopy. Sections incubated with purified DENV NS1 and stained with isotype mAbs served as a negative control. doi:10.1371/journal.ppat.0030183.g009 in PBS containing 0.1% BSA). After washing, biotinylated goat antimouse IgG (1 lg/ml) and horseradish peroxidase-conjugated streptavidin were added sequentially for 1-h incubations at room temperature. After six final washes with PBS, signal was detected by adding 150 ll of TMB substrate (DakoCytomation) and 50 ll of 0.1 N H 2 SO 4 stop solution to each well. Plates were evaluated at 450 nm on a 96well plate reader (Genios Pro; Tecan Instruments).
Sodium chlorate treatment of cell cultures. Sulfation was inhibited by sodium chlorate treatment [56]. BHK and CHO cells were cultured in sulfate-free Joklik Modification Minimum Essential Medium Eagle (Sigma) or Ham's F12 medium (Tissue culture support center, Washington University) supplemented with 10% dialyzed FBS (Sigma) containing different concentrations (1-75 mM) of sodium chlorate. In some experiments, 10 mM sodium sulfate was added to replenish sulfate to the cells. After overnight culture, cells were processed for DENV NS1 binding as described above.
Analysis of NS1 expression on the surface of DENV-infected cells. BHK cells (1.6 3 10 5 /well) were seeded onto 12-well tissue culture plate (Costar). 24 h later, cells were infected with DENV-2 (strain 16681) at a multiplicity of infection (MOI) of 3. After a 2-h incubation at 37 8C, cell monolayers were extensively washed and cultured in DMEM containing 4% FBS. In some experiments, DENV-infected cells were cultured in medium containing 1, 10, or 100 lg/ml of HP or CS-C or 25 mM sodium chlorate. At specific intervals post-infection, cells were harvested from culture plates with an EDTA solution (4 mM EDTA plus 10% FBS in PBS). DENV NS1 on the surface was determined by incubation of 5 3 10 5 cells with anti-DENV-2 NS1 mAb clone 2G6 or mouse polyclonal anti-DENV-2 NS1 (20 lg/ml final concentration) for 1 h at 4 8C. After three washes, cells were incubated with Alexa Fluor 647-conjugated anti-mouse IgG (1:500) and analyzed by flow cytometry. In some experiments, DENVinfected cells were treated with a mix of heparin lyases I, II, III or chondroitinase ABC prior to detection of surface NS1.
DENV NS1 binding to tissues. Human lung tissue: The protocol to obtain human lung tissues from patients was approved by the Institutional Review Board at the Faculty of Medicine Siriraj Hospital, Mahidol University. A human lung, surgically removed with informed consent from a patient with lung cancer, was evaluated macroscopically for cancer-free regions. Selected tissues, cut from the central part of the cancer-free region of the lung, were cryoprotected in 30% sucrose for generation of 6-to 7-lm frozen sections. Cryosections were thawed, fixed in 4% paraformaldehyde (Sigma) in PBS at room temperature for 10 min, and washed extensively with PBS. Tissues were incubated with purified DENV NS1 or BSA at 20 lg/ml for 1 h at room temperature and followed by three washes with 0.1% BSA in PBS 5 min each. Sections were incubated with a mixture of mAbs (1A4, 1F11, 2G6, 1B2; 20 lg/ml final concentration) against DENV NS1. A mixture of IgG1 (MOPC-21, Sigma) and IgG2a (UPC-10, Sigma) at 20 lg/ml was used as isotype control Abs. After three washes, sections were incubated with Alexa Fluor 488 conjugated goat anti-mouse IgG secondary antibody (Invitrogen) and followed by 1-h incubation at room temperature with rabbit anti-human CD31 (PECAM) (Santa Cruz Biotechnology) at the dilution of 1:10. Sections were washed three times and incubated with Cy3 conjugated donkey anti-rabbit IgG (Jackson Immuno Research Laboratories). Nuclear staining was achieved by incubation for 15 min with 1:100 dilution of Hoechst dye (Invitrogen) and visualized using a Zeiss 510 Meta LSM confocal microscope.
Mouse tissues: To obtain tissues from mice, 4-to 6-wk-old uninfected C57BL/6 mice were anesthesized with ketamine and xylazine and perfused with 20 ml of PBS. Tissues were dissected and cryoprotected in 30% sucrose for generation of frozen sections. Serial 6-lm cryosections were air-dried for 1 h at room temperature followed by washing with PBS. Sections were fixed with 50% acetone in PBS for 10 min on ice, incubated with PBS containing 2 M NaCl (pH 7.4) for 10 min at room temperature, and blocked in PBS containing 0.5% BSA for 1 h at room temperature or overnight at 4 8C. Subsequently, sections were incubated with serum-free supernatants from BHK DENV-2 Rep or BHK cells for 1 h at room temperature followed by incubation with a mixture of anti-NS1-mAbs (1A4, 1F11, 2G6, 1B2; 10 lg/ml final concentration) or isotype control mAbs (10 lg/ml) diluted in PBS containing 0.5% BSA. After three washes, sections were incubated with Cy3 conjugated with goat anti-mouse IgG (Zymed) at the dilution of 1:200. Co-staining with endothelial cell marker, CD31 (PECAM), was accomplished by a 1-h incubation with rat anti-mouse PECAM-1 (CD31) (BD Pharmingen) at a dilution of 1:250 followed by incubation with a 1:500 dilution of secondary Alexa Fluor 488 conjugated anti-rat IgG (Invitrogen). Nuclear staining was achieved by incubation for 15 min with 1:2,500 dilution of TO-Pro-3 (Invitrogen). Sections were mounted using fluorescent mounting medium (Vector Laboratories) and analyzed using a Zeiss 510 Meta LSM confocal microscope. For soluble GAGs and DENV NS1 competitive binding experiments, serum-free supernatants from BHK DENV-2 Rep were mixed with soluble CS-C, CS-E, or HP (3.3 lM final concentration) prior to incubation with sections of mouse lung. The sections were further processed for DENV NS1 and CD31 staining as described above. 50-60 images of each condition were captured by confocal microscopy, and fluorescent intensity of each image was analyzed using the Volocity software (Improvision). Regions that stained positively for CD31 (intensity greater than 427 arbitrary units on the intensity scale of 0-4,095) were selected for analysis. Mean intensity of DENV NS1 binding was calculated for all measured regions. For analysis of DENV NS1 binding to pleura, regions were manually selected based on anatomical sites at the surface of the lung.
Statistical analysis. Datasets were compared by a two-tailed, unpaired t test. Multiple comparisons were performed using an ANOVA test. Statistical significance was achieved when p-values were , 0.05. Data analysis was performed using Prism software (Graph-Pad). Figure S1. Immunoreactivity of Purified DENV-2 NS1 with mAbs Purified NS1 from BHK DENV-2 Rep cells (rNS1) or DENV-2 infected cells (iNS1) (250 ng/well) was maintained at room temperature or heated (at 95 8C for 5 min) in SDS nonreducing sample buffer immediately prior to 12% SDS-PAGE. Western blot was performed with the individual NS1-specific mAbs 2E11, 2E3, 1F11, 2G6, 1A4, and 1B2. Monomer and dimer forms of NS1 are indicated. Found at doi:10.1371/journal.ppat.0030183.sg001 (1.5 MB TIF). Figure S2. Binding of DENV NS1 to Mesothelial Cells Lining Pleura and Peritoneum Cryo-sections of mouse (A) lung, (B) intestine were incubated with serum-free supernatants from BHK DENV-2 Rep or BHK cells for 1 h at room temperature. After extensive washing, bound NS1 was detected by a mixture of NS1 mAbs (1A4, 1F11, 2G6, 1B2) followed by Cy3-conjugated goat anti-mouse IgG. Costaining with endothelial cell marker was subsequently performed by incubating the sections with rat anti-mouse CD31 (PECAM) followed by Alexa Fluor 488conjugated goat anti-rat IgG. Nuclei were stained with a DNAspecific dye TO-Pro-3. Sections incubated with DENV NS1 followed by an isotype control Ab served as a negative control. Analysis was performed by confocal microscopy. Arrowheads denote a single layer of mesothelial cells lining pleura (A) or peritoneum (B). Found at doi:10.1371/journal.ppat.0030183.sg002 (38.7 MB TIF).