The Dengue Virus Nonstructural Protein 1 (NS1) Is Secreted from Mosquito Cells in Association with the Intracellular Cholesterol Transporter Chaperone Caveolin Complex

The dengue virus protein NS1 is secreted efficiently from both infected vertebrate and mosquito cells. Previously, our group reported that NS1 secretion in mosquito cells follows an unconventional secretion pathway dependent on caveolin-1. In this work, we demonstrate that in mosquito cells, but not in vertebrate cells, NS1 secretion takes place in association with the chaperone caveolin complex, a complex formed by caveolin-1 and the chaperones FKBP52, CyA, and Cy40, which are in charge of cholesterol transport inside the cell. Results obtained with ZIKV-infected mosquito cells suggest that ZIKV NS1 is released following an unconventional secretory route in association with the chaperone caveolin complex. These results uncover important differences in the virus-cell interactions between the vertebrate host and the mosquito vector, as well as novel functions for the chaperone caveolin complex. Moreover, manipulation of the NS1 secretory route may prove a valuable strategy to combat these two mosquito-borne diseases.

ABSTRACT

Dengue virus (DENV) is a mosquito-borne virus of the family Flaviviridae. The RNA viral genome encodes three structural and seven nonstructural proteins. Nonstructural protein 1 (NS1) is a multifunctional protein actively secreted in vertebrate and mosquito cells during infection. In mosquito cells, NS1 is secreted in a caveolin-1-dependent manner by an unconventional route. The caveolin chaperone complex (CCC) is a cytoplasmic complex formed by caveolin-1 and the chaperones FKBP52, Cy40, and CyA and is responsible for the cholesterol traffic inside the cell. In this work, we demonstrate that in mosquito cells, but not in vertebrate cells, NS1 associates with and relies on the CCC for secretion. Treatment of mosquito cells with classic secretion inhibitors, such as brefeldin A, Golgicide A, and Fli-06, showed no effect on NS1 secretion but significant reductions in recombinant luciferase secretion and virion release. Silencing the expression of CAV-1 or FKBP52 with short interfering RNAs or the inhibition of CyA by cyclosporine resulted in significant decrease in NS1 secretion, again without affecting virion release. Colocalization, coimmunoprecipitation, and proximity ligation assays indicated that NS1 colocalizes and interacts with all proteins of the CCC. In addition, CAV-1 and FKBP52 expression was found augmented in DENV-infected cells. Results obtained with Zika virus-infected cells suggest that in mosquito cells, ZIKV NS1 follows the same secretory pathway as that observed for DENV NS1. These results uncover important differences in the dengue virus-cell interactions between the vertebrate host and the mosquito vector as well as novel functions for the chaperone caveolin complex.

IMPORTANCE The dengue virus protein NS1 is secreted efficiently from both infected vertebrate and mosquito cells. Previously, our group reported that NS1 secretion in mosquito cells follows an unconventional secretion pathway dependent on caveolin-1. In this work, we demonstrate that in mosquito cells, but not in vertebrate cells, NS1 secretion takes place in association with the chaperone caveolin complex, a complex formed by caveolin-1 and the chaperones FKBP52, CyA, and Cy40, which are in charge of cholesterol transport inside the cell. Results obtained with ZIKV-infected mosquito cells suggest that ZIKV NS1 is released following an unconventional secretory route in association with the chaperone caveolin complex. These results uncover important differences in the virus-cell interactions between the vertebrate host and the mosquito vector, as well as novel functions for the chaperone caveolin complex. Moreover, manipulation of the NS1 secretory route may prove a valuable strategy to combat these two mosquito-borne diseases.

INTRODUCTION

Dengue is an arthropod-borne viral disease caused by any of the four dengue virus serotypes (DENV1 to DENV4), which are transmitted by Aedes mosquitoes (1). Dengue infection may develop into a life-threatening illness. There are currently no specific therapeutics, and substantial vector control efforts have not stopped its rapid emergence and global spread. The risk of dengue virus infection and its public health burden are hard to estimate, but dengue currently is endemic to at least 100 countries (2). DENV has a single-stranded RNA genome of positive polarity. The genome of DENV is translated into a single polyprotein and encodes three structural (C, E, and prM/M) and seven nonstructural (NS1, NS2A, NS2B, NS3, NS4A, NS4B, and NS5) proteins. The polyprotein is then cleaved by host and viral proteases to release individual viral proteins (3). The viral genome replication process within the host cell is mainly driven by the NS proteins. The NS1 protein acts as a scaffolding protein that anchors the replication complex to the endoplasmic reticulum (ER) membrane and interacts physically with NS4B (4). The NS1 protein is a 352-amino-acid polypeptide with a molecular weight of 46 to 55 kDa, depending on its glycosylation status. The NS1 protein exists in multiple oligomeric forms and is found in different cellular locations, such as a cell membrane-bound form in association with virus-induced intracellular vesicular compartments, on the cell surface, and as a soluble secreted hexameric lipoparticle (4). The NS1 monomeric form rapidly dimerizes in the endoplasmic reticulum (ER), and then three dimeric forms of NS1 arrange to form a hexamer (5). The hexameric form of NS1 shows an open barrel form filled with lipids and cholesterol, resembling the lipid composition of the HDL particle (6).

Recent studies have shown that the DENV NS1 protein was secreted from vertebrate cells and also efficiently secreted from mosquito cells lines (7, 8). The secretion of NS1 in vertebrate cells follows the classical Golgi pathway (9). However, NS1 secretion in infected mosquito cells is associated with a caveolin-1 (CAV-1)-dependent pathway and was found to be brefeldin A (BFA) insensitive, suggesting a traffic route that bypasses the Golgi complex (10).

Caveolae are made up of interlocking heteropolymers of a family of small proteins (caveolin-1 [CAV-1] to -3) and a second family of accessory structural proteins (flotillins and three families of cavins). The caveolar architecture is connected with unstructured cavin filaments by coiled-coil domains into a polygonal net-like complex. This complex is believed to provide scaffolding for compartmented cellular processes and participates in multiple cellular functions, including endocytosis, transcytosis, membrane homeostasis, inflammation, and signal transduction (11). CAV-1, a 21- to 24-kDa scaffolding protein, is not only a key structural component of the caveolae organelle but also plays an important role in the transport of free cholesterol inside the cell (12, 13). The chaperone caveolin complex (CCC) is a cytosolic complex reported to transport cholesterol synthesized de novo from the ER to cell membranes or other compartments within the cell. CCC has been described as a complex of CAV-1, cyclophilin A (CyA), FK506-binding protein 4 or heat shock protein 56 (FKBP52), and cyclophilin 40 or D (Cy40) (14). CyA, an 18-kDa peptidylprolyl cis-trans isomerase, is a ubiquitous and multifunctional protein. In addition to its role as a host cell receptor for cyclosporine, CyA has diverse functions in inflammatory conditions and diseases (15, 16). The 52-kDa FK506-binding protein (FKBP52), an immunophilin belonging to the FKBP family, is a known cochaperone of heat shock protein 90 (HSP90) and may play a role in the intracellular trafficking of hetero-oligomeric forms of the steroid hormone receptors (17, 18). Cy40, a member of a family of highly homologous peptidylprolyl cis-trans isomerases (PPIases), is known to play a role in mitochondrial permeability transition (MPT), being an integral constituent of the MPT pore (19).

Given the CAV-1-dependent secretion of NS1 protein in mosquito cells and the lipoprotein nature of the released hexameric form of NS1, it was found plausible to study the association of NS1 trafficking to the novo cholesterol transport in DENV-infected mosquito cells. In this work, data are presented indicating that in infected mosquito cells, DENV NS1 enters the unconventional secretory pathway very early after maturation in the ER and usurps the cholesterol transport between the ER and the plasma membrane, mediated by the CCC, to reach the extracellular space. In addition, data are presented suggesting that a similar pathway is used for the secretion of Zika virus NS1 protein in infected mosquito cells.

(This article was submitted to an online preprint archive [20].)

RESULTS

NS1 secretion is not affected by drugs that disrupt early steps of the classical secretion pathway.

Golgicide A (GCA) is a powerful inhibitor of the COPI vehicle transport from ER to Golgi membrane (21). Thus, the cytotoxicity of GCA in the mosquito cell lines (C6/36 and Aag2) and the vertebrate cell line BHK-21, used for comparisons, was measured using the reduction of tetrazolium salts to examine proliferation in cells treated with serial dilutions of GCA. No significant cytotoxicity was observed under 30 μM GCA in any of the three cell types (Fig. 1A). Fli-06 is a novel drug which inhibits the diffusion of ER-synthesized proteins to the ER exit sites (ERES) (22, 23). Fli-06 cytotoxicity was determined in C6/36 and Aag2 cells also using also tetrazolium salt reduction. No significant cytotoxicity was observed under 100 μM Fli-06 in the mosquito cells lines (Fig. 1B). Thus, concentrations of 27 μM and 100 μM were used for GCA and Fli-06, respectively, since these concentrations proved nontoxic yet effective in causing Golgi disruption in the three cell lines (Fig. 1C). In addition, BFA was used at a concentration of 25 μM, as described in Alcalá et al. (10). Figure 1D and E shows that the GCA treatment of DENV-infected C6/36 and Aag2 cells did not cause any change in NS1 secretion. In contrast, NS1 secretion was reduced more than 70% in infected BHK-21 cells treated either with GCA or BFA (Fig. 1F). As expected, virion release was reduced in mosquito and vertebrate cells after GCA or BFA treatment (Fig. 1G to I). To further explore the traffic route of NS1 in DENV-infected mosquito cells, cells were also treated with Fli-06. Treatment of infected C6/36 or Aag2 cells with nontoxic concentrations of Fli-06 showed no effect on NS1 secretion while significantly affecting virion release (Fig. 1G and H). To discard any possible effect of the drug treatments on virus replication, genomic RNA levels were determined by real-time reverse transcription-PCR (RT-PCR). The results indicated that DENV2 and DENV4 RNA replication was not significantly affected by BFA, GCA, and Fli-06 treatments (Fig. 1G to I), although a decrease in DENV4 RNA level of about 1 log₁₀ was observed in the Fli-06-treated C6/36 cells (Fig. 1G). Experiments with Fli-06 were not carried out in BHK-21 due to high drug toxicity. The results with BFA, GCA, and Fli-06 taken together indicate that in mosquito cells NS1 is not secreted by a classical secretory pathway and that newly synthesized DENV NS1 appears not to reach the ER exit sites and leaves the ER compartment very early after synthesis.

FIG 1.

NS1 secretion is not affected in mosquito cells treated with classical secretion pathway inhibitors. (A) Cell toxicity of GCA in C6/36, Aag2, and BHK-21 cells. (B) Cell toxicity of Fli-06 in C6/36 and Aag2 cells. (A and B) Serial dilutions of GCA and Fli-06 were applied for 24 h. Data are means from 3 independent experiments ± standard errors; significant differences compared with controls are denoted by asterisks (P < 0.0001). (C) Golgi matrix integrity after cell treatment with GCA at 27 μM and Fli-06 at 100 μM for 24 h was determined by staining the cis-Golgi matrix. Cells were fixed and stained with anti-GM130 (cis-Golgi, in green) and DAPI (nuclei in blue). (D) NS1 secretion with inhibition of classical secretory pathway in C6/36 cells. (E) NS1 secretion with inhibition of classical secretory pathway in Aag2 cells. (F) NS1 secretion with inhibition of classical secretory pathway in BHK-21 cells. (G) Viral release and DENV RNA copies in drug-treated C6/36 cells. (H) Viral release and DENV RNA copies in drug-treated Aag2 cells. (I) Viral release and DENV RNA copies in drug-treated BHK-21 cells. (D, E, and F) Cells were infected with DENV2 or DENV4 at an MOI of 3 for 1 h. The cells were washed 3 times and treated with DMSO (control), 25 μM BFA, 27 μM GCA, or 100 μM Fli-06, and supernatants were harvested at 24 hpi. DENV NS1 was measured using Platelia NS1 Ag (Bio-Rad) and represented as percentage of NS1 secreted compared to the value for the DMSO control. (G, H, and I) Viral titers were measured in supernatants from treated cells by focus-forming assay in BHK-21 cells by following a standard protocol. Total DENV RNA copies were calculated in 10⁵ cells and values plotted on a secondary y axis. DENV2/DENV4 RNA copies were measured by real-time RT-PCR using TaqMan technology as described in the text. (J) C6/36 cells (pAc5-mCherry-GLuc-neo transfected) and BHK-21 cells (ptk-GLuc transfected) expressing Gaussia luciferase were treated with DMSO and 27 μM GCA for 24 h. Luciferase activity was measured in supernatants and represented as percentage of Luc activity. pAc5-mCherry-GLuc-Neo was constructed by cloning GLuc into the XbaI-HindIII sites of pAc5-STABLE 2-Neo, designed for tricistronic expression driven by the Actin5C promoter. FLAG-tagged mCherry, Gaussia luciferase, and NeoR, each separated by a T2A peptide, are shown. (D, E, F, G, H, I, and J) The experiments were performed in triplicate, and the bars represent means ± standard errors (error bars). Data were evaluated using the 2-way analysis of variance (ANOVA) test, and significant differences compared with results for the DMSO treatment group are denoted by asterisks (P ≤ 0.05). δ, four asterisks. (K) Diagram showing the modes of action of the drugs used and the early ER exit of DENV NS1 in mosquito cells.

Since the release of mature virions, which are multimolecular complexes, may not be equivalent to the release of a single protein, an additional control for protein trafficking through the classical secretory pathway was included. To this aim, a vector, pAc5-mCherry-GLuc-neo, constitutively expressing Gaussia luciferase (GLuc) in C6/36 cells, was constructed (Fig. 1J). To evaluate the effect of GCA treatment on GLuc secretion, cells were treated with GCA for 24 h and the activity of GLuc present in the cell supernatants analyzed using a luminescence assay. A decrease in luciferase activity, in relation to that of the control (dimethyl sulfoxide [DMSO]), was detected upon treatment of C6/36 with GCA and in BHK-21 cells transfected with ptk-GLuc vector, run in parallel as a control (Fig. 1J). These results indicated that the luciferase reporter is secreted in both mosquito and vertebrate cells by following a classical secretory route and reinforce the results indicating that in mosquito cells DENV NS1 is secreted by an unconventional secretory route that bypasses the Golgi membrane (Fig. 1K).

NS1 secretion is dependent on the CCC.

The CCC is formed by the association of CAV-1 with the chaperones CyA, FKBP52, and Cy40 (14). The CCC is responsible for cholesterol transport inside the cell. Previous data from our work (10) indicated that NS1 secretion was dependent on CAV-1. Thus, to fully evaluate the participation of the CCC in DENV NS1 secretion from mosquito cells, CyA was pharmacologically inhibited by treatment with cyclosporine (CsA), and the expression of the chaperones FKBP52 and Cy40 was knocked down using short interfering RNAs (siRNAs). Previously, the noncytotoxic concentrations for CsA in mosquito and vertebrate cells were determined (Fig. 2A). DENV2- or DENV4-infected C6/36 or Aag2 cells were treated with 9 μM CsA for 24 h. NS1 secretion was reduced about 30% and 40% in C6/36 and Aag2 cells, respectively (Fig. 2B and E). However, neither virion release nor viral RNA replication was significantly affected in the treated mosquito cells (Fig. 2C and F). In contrast, the inhibition of CyA in BHK-21 cells did not have any effect on NS1 secretion, virion secretion, or viral RNA replication (Fig. 2H and I). To further corroborate the participation of CyA in cholesterol and lipid traffic, the lipid droplet (LD) count in CsA-treated versus untreated cells was determined. In all cases, C6/36, Aag2, and BHK-21 cells treated with CsA showed a diminished amount of lipid droplets per cell compared to that of untreated cells (Fig. 2D, G and J).

FIG 2.

NS1 secretion is inhibited in mosquito cells treated with CsA. (A) Cell toxicity of cyclosporine (CsA) in C6/36, Aag2, and BHK-21 cells. Serial dilutions of CsA were applied for 24 h. Data are means from 3 independent experiments ± standard errors, and significant differences compared with the control are denoted by asterisks (P < 0.0001). (B) NS1 secretion in CsA-treated C6/36 cells. (C) Viral release and DENV RNA copies in CsA-treated C6/36 cells. (D) Lipid droplet counting in C6/36 cells treated with CsA. (E) NS1 secretion in CsA-treated Aag2 cells. (F) Viral release and DENV RNA copies in CsA-treated Aag2 cells. (G) Lipid droplet counting in Aag2 cells treated with CsA. (H) NS1 secretion in CsA-treated BHK-21 cells. (I) Viral release and DENV RNA copies in CsA-treated BHK-21 cells. (J) Lipid droplet counting in BHK-21 cells treated with CsA. (B, E, and H) Cells were infected with DENV2 or DEN4 at an MOI of 3 for 1 h. The cells were washed 3 times, treated with DMSO (control) or 9 μM CsA, and harvested at 24 hpi. DENV NS1 in supernatants was measured using Platelia NS1 Ag (Bio-Rad) and represented as percentage of NS1 secreted compared to the control level (DMSO). (C, F, and I) Viral titers in supernatants were measured by focus-forming assay in BHK-21 cells according to standard procedures. Total DENV RNA copies in 10⁵ cell lysates were measured by real-time RT-PCR using TaqMan technology as described in the text. (D, G, and J) Lipid droplets were stained with Nile red (red) and nuclei with DAPI (blue). LD were counted in maximum projection images in treated cells from at least 25 cells by group (DMSO or CsA) using the Spot Detector plugin with Icy Image software. LD median count differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05). (B, C, E, F, H, and I) The experiments were performed in triplicate, and the bars represent means ± standard errors (error bars). Data were evaluated using the 2-way ANOVA test, and significant differences are from comparisons to the DMSO treatment group, denoted by asterisks (P ≤ 0.05).

The expression of FKBP52 in C6/36 cells was knocked down by siRNA transfection. Optimal gene knockdown was achieved after 48 h posttransfection (Fig. 3A). The effect of FKPB52-reduced expression in cholesterol and lipid metabolism was evaluated by measuring lipid droplet count per cell, and a reduction of about 50% in the number of lipid droplets was observed after the treatment (Fig. 3B). In cells treated with siRNA FKBP52, a significant reduction in NS1 of about 20% secretion was observed (Fig. 3C). Viral release was not affected by FKBP52 knockdown (Fig. 3D). Given that FKBP52 is a cochaperone of HSP90 (17, 24), we wanted to discard any observed inhibitory effect on NS1 secretion related to a reduction in HSP90 activity. Thus, the effect of the HSP90 inhibitor geldanamycin (GA) in NS1 release from DENV-infected C6/36 cells was tested. As shown in Fig. 3E, NS1 secretion was not affected after GA treatment. The expected reduction in virus yield observed in GA-treated cells (Fig. 3F) showed that effective GA doses were used. These results indicate that the effect in NS1 secretion observed after FKBP52 silencing is not related to HSP90 activity and rather was associated with the function of the CCC. Finally, the expression of Cy40 was significantly knocked down at 48 h after siRNA transfection (Fig. 4A). A reduction in about 60% of lipid droplet counts per cell was observed in transfected cells (Fig. 4B), yet neither NS1 secretion nor virus release was affected by Cy40 silencing (Fig. 4D and E). This result suggests that while Cy40 participates in cholesterol traffic, it is dispensable for NS1 secretion.

FIG 3.

NS1 secretion is inhibited in mosquito cells transfected with FKBP52 siRNA. (A) C6/36 cells were transfected with 50 nM siRNA targeting FKBP52 or with AllStars negative-control siRNA. Protein expression was measured by Western blotting after 48 h (FKBP52 antibody, 1/2,000). After 24 h posttransfection, cells were infected with DENV4 for 1 h at an MOI of 3. (B) LD count in C6/36 cells silenced for FKBP52 expression. LD counting in FKBP52 and negative knockdown were measured by LD staining with Nile red. LD were counted in maximum projection images in at least 20 treated cells using the Spot Detector plugin in Icy Image software. LD median count differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05). (C) Secreted DENV NS1 in siRNA-transfected C6/36-infected cells. NS1 was measured after 48 h posttransfection using Platelia NS1 Ag (Bio-Rad) and represented as percentage of NS1 secreted compared to that of the negative siRNA. (D) Viral titers in supernatants were measured by focus-forming assay in BHK-21 cells by following standard procedures. (E) Independence of the FKBP52 cochaperone activity on HSP90 was assayed by geldanamycin (GA) treatment. HSP90 inhibition is not responsible for reduction in NS1 secretion. C6/36 cells were treated with 10 μM GA, a specific inhibitor of HSP90. Percentage of secreted NS1 in 24 hpi is shown. (F) Viral titers in GA treatment in C6/36-DENV-infected cells. Supernatants were measured by focus-forming assay in BHK-21 cells by following standard procedures. (C, D, E, and F) All data are means from 3 independent experiments ± standard errors (error bars), and significant differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05).

FIG 4.

NS1 secretion is not affected in mosquito cells transfected with Cy40 siRNA. (A) C6/36 cells were transfected with 50 nM siRNA targeting Cy40 or with All Stars negative-control siRNA. Protein expression was measured by Western blotting after 48 h (Cy40 antibody, 1/2,000). After 24 h posttransfection, cells were infected with DENV4 for 1 h at an MOI of 3. (B) LD count in C6/36 cells silenced for Cy40 expression. LD counting in Cy40 and negative-control knockdown were measured by LD staining with Nile red. LD were counted in maximum projection images in at least 20 treated cells using the Spot Detector plugin in Icy Image software. LD median count differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05). (C) Secreted DENV NS1 was measured after 48 h posttransfection using Platelia NS1 Ag (Bio-Rad) and represented as percentage of NS1 secreted compared to that of negative siRNA. (D) Viral titers in supernatants were measured by focus-forming assay in BHK-21 cells by following standard procedures. (C and D) All data are means from 3 independent experiments ± standard errors, and significant differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05).

Since the hexameric NS1 is rich in cholesterol and triglycerides, a possible interaction between DENV NS1 and lipid droplets was tested through colocalization assays. DENV infection results in an increase in LD in mosquito cells (Fig. 5B), yet NS1 did not show significant Pearson correlation coefficient (PCC) values with LD in mosquito or vertebrate cells (Fig. 5C). Finally, the interference with the chaperones that conform to the CCC in noninfected cells resulted in a significant reduction in the count of lipid droplets per cell in all cases. This decrease is similar to that observed in mosquito cells treated with methyl-β-cyclodextrin (MβCD) or subjected to serum starvation (Fig. 5B). These data, generated in mosquito cells, endorse the role proposed for the CCC in lipid traffic and homeostasis in vertebrate cells (12, 14).

FIG 5.

Cholesterol metabolic stress modifies lipid droplets, and DENV NS1 does not colocalize with lipid droplets in mosquito cells. (A) Cell toxicity of MβCD in C6/36 cells. Data are means from 3 independent experiments ± standard errors, and significant differences compared with values for the control are denoted by asterisks (P < 0.0001). (B) Modification in lipid droplet count in mosquito cells during cholesterol stress. DENV4 infection (MOI of 3), MβCD treatment (1 mM), and serum starvation were performed in the absence of bovine fetal serum for 24 h. LD median count differences were compared with mock infection using two-way ANOVA, denoted by asterisks (P ≤ 0.05). (C) Colocalization of NS1 and lipid droplets in DENV2- and DENV4-infected mosquito and vertebrate cells (C6/36, Aag2, and BHK-21). LD were stained with BODIPY (shown in red), DENV NS1 is shown in green, and nuclei are in blue (DAPI). The insets show magnifications of the selected zones. PCC was used to evaluate the degree of colocalization between NS1 and LD. PCC are shown in each square panel. The images were analyzed using an LSM 700 confocal microscope with laser sections (0.38 μm).

Dengue virus infection increases the expression of CCC proteins.

Given the relatively large amounts of NS1 synthesized and the expected large amounts of lipids required to form the hexamer, and given the importance of CCC in NS1 traffic and lipid homeostasis, the effect of dengue virus infection on CCC protein levels was evaluated. Western blot analysis of DENV-infected C6/36 cells showed a 3.5- and 2.0-fold increase in CAV-1 and FKBP52 expression, respectively, at 24 h postinfection (hpi); meanwhile, no changes in the expression of Cy40 and CyA were observed (Fig. 6A). The infection of BHK-21 cells run in parallel showed no changes in the expression of any of the proteins associated with the CCC (Fig. 6B). Of note, a 1.6-fold increase in CAV-1 expression levels was observed in C6/36 cells treated with MβCD for 24 h (Fig. 6C), suggesting that the increase in CAV-1 expression is a response of the mosquito cell to alterations in lipid homeostasis.

FIG 6.

Expression of the chaperone caveolin complex is augmented in DENV-infected mosquito cells. (A) Relative protein expression of CCC in DENV4-C6/36-infected cells at an MOI of 3. (B) Relative protein expression of CCC in DENV4-BHK-21-infected cells at an MOI of 3. (C) CAV-1 expression levels in C6/36 cells treated with 1 mM MβCD. CAV-1 expression was determined by Western blotting and normalized with GAPDH. Values are means ± SEM from 4 independent experiments. (A and B) Cells were mock or DENV4 infected at an MOI of 3 for 24 h. The relative protein expression of caveolin chaperone complex was determined in 20 μg of cell lysates by the ratio of the sample value to an internal standard control (GAPDH). Ratio values are means ± SEM (n = 4 for mock or DENV4). Significant differences were compared using Mann-Whitney U test. Significance is indicated by asterisks (P < 0.05).

In vertebrate cells, NS1 secretion is independent of the CCC.

To assess if the dependence of NS1 secretion on the CCC is exclusive of mosquito cells, we evaluated the effect of silencing of CAV-1, FKBP52 and Cy40 on NS1 secretion in infected BHK-21 cells. As shown in Fig. 7, the use of siRNAs significantly reduced the expression levels of the targeted proteins (Fig. 7A, E, and I). Moreover, a significant reduction in lipid droplet count levels was observed after the knockdown of each of the proteins of the CCC (Fig. 7D, H, and L). However, no changes in NS1 secretion levels (Fig. 7B, F, and J) or virion release (Fig. 7C, G, and K) were observed in the transfected cells. These results indicate that while the CCC plays a role in lipid homeostasis, it plays no role in the secretion of NS1 in vertebrate cells.

FIG 7.

NS1 secretion is not affected in vertebrate cells transfected with CAV-1, FKBP52, or Cy40 siRNAs. (A to D) Knockdown of CAV-1. (A) BHK-21 cells were transfected with 50 nM siRNA targeting CAV-1 or with AllStars negative-control siRNA. (B) Secreted DENV NS1 after CAV-1 knockdown. (C) Viral release after CAV-1 knockdown. (D) LD count in BHK-21 cells silenced for CAV-1 expression. (E to H) Knockdown of FKBP52. (E) BHK-21 cells were transfected with 50 nM siRNA targeting FKBP52 or with AllStars negative-control siRNA. (F) Secreted DENV NS1 after FKBP52 knockdown. (G) Viral release after FKBP52 knockdown. (H) LD count in BHK-21 cells silenced for FKBP52 expression. (I to L) Knockdown of cyclophilin 40. (I) BHK-21 cells were transfected with 50 nM siRNA targeting Cy40 or with AllStars negative-control siRNA. (J) Secreted DENV NS1 after Cy40 knockdown. (K) Viral release after Cy40 knockdown. (L) LD count in BHK-21 cells silenced for Cy40 expression. (A, E, and I) Gene knockdown was assessed using Western blotting. Protein expression was measured and normalized with GAPDH after 48 h. After 24 h posttransfection, cells were infected with DENV4 at an MOI of 3 for 1 h. Data are means from three experiments ± standard errors of the means (SEM). (B, F, and J) Secreted DENV NS1 was measured after 48 h posttransfection of siRNA using Platelia NS1 Ag (Bio-Rad) and represented as percentage of NS1 secreted compared to results for negative-control siRNA. (C, G, and K) Viral titers were measured by focus-forming assay in BHK-21 cells. (B, C, F, G, J, and K) All data are means ± SEM (error bars), n = 3, and significant differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05). (D, H, and L) LD were measured by LD staining with Nile red and counted in maximum projection images in at least 20 cells using the Spot Detector plugin in Icy Image software. LD median count differences were compared using Student's t test, denoted by asterisks (P ≤ 0.05).

DENV NS1 colocalizes with the CCC in mosquito cells.

Based on the dependence of NS1 secretion on the CCC observed, we evaluated the association between NS1 and CAV-1, FKBP52, Cy40, and CyA in mosquito C6/36 (Fig. 8A to D) and Aag2 (Fig. 8E to H) cells infected and then fixed at 18 hpi. BHK-21 cells were run in parallel as controls (Fig. 8I to L). Quantification of the colocalization levels using the PCC indicated significant (PCC of ≥0.5) colocalization between NS1 and all the components of the CCC in both insect cell lines, especially with CAV-1 and CyA. The colocalization between NS1 and CCC was always observed in the cytoplasm, randomly distributed and not forming any specific structure. As expected, no colocalization (PCC of ≤ 0.2) of NS1 with the proteins of the CCC was observed in BHK-21 cells. Finally, colocalization studies carried out in mock-infected cells suggest the association of CAV-1 with FKBP52, Cy40, and CyA, and, thus, the presence of the CCC in both (Fig. 9).

FIG 8.

DENV NS1 colocalizes with proteins of the CCC in mosquito cells. (A, B, C, and D) Colocalization for NS1 and CCC in C6/36 cells. (E, F, G, and H) Colocalization for NS1 and CCC in Aag2 cells. (I, J, K, and L) Colocalization for NS1 and CCC in BHK-21 cells. All cells were infected with DENV4 at an MOI  of  3 and fixed at 18 hpi. The cells were probed against NS1 (in green) and CAV-1, FKBP52, CyA, and Cy40 (in red). Nuclei were stained with DAPI (in blue). In merged fluorescent images of DAPI, red and green channels are shown. Inset panels show magnifications of the detected spots corresponding to the colocalization. The images were analyzed using an LSM 800 confocal microscope. (M) PCC was used to evaluate the degree of colocalization between NS1 and each of the CCC proteins in the 3 cell lines infected with DENV4. (N) PCC was used to evaluate degree of colocalization between CAV-1 and FKBP52, CyA, and Cy40 in noninfected cell lines as a control for CCC. Confocal images of noninfected cells are shown in Fig. 9. (M and N) The images were analyzed using an LSM 800 confocal microscope with laser sections (0.45 μm). The bars represent means ± standard errors from at least 20 independent confocal cell images. Data were evaluated using the 2-way ANOVA test, and significant differences between groups are denoted by asterisks (P ≤ 0.0001). The dotted line indicates the threshold for true colocalization (PCC ≥ 0.5).

FIG 9.

Chaperone caveolin complex colocalization in noninfected vertebrate and mosquito cells. (A, B, and C) Colocalization for NS1 and CCC in C6/36 cells. (D, E, and F) Colocalization for NS1 and CCC in Aag2 cells. (G, H, and I) Colocalization for NS1 and CCC in BHK-21 cells. Mock-infected cells were fixed and probed against CAV-1 (shown in red) and FKBP52, CyA, and Cy40 (shown in green). Merged fluorescent images, with red and green channels, are shown. The images were analyzed using an LSM 800 confocal microscope. PCC were used to analyze the degree of colocalization with laser sections (0.41 μm) (PCC are shown in Fig. 8N).

NS1 interacts with chaperone caveolin complex in mosquito cells.

The colocalization results suggest that NS1 interacts with CAV-1 and the chaperones of the CCC in DENV-infected mosquito cells. To test this possibility, coimmunoprecipitation (co-IP) experiments were carried out using infected C6/36 cell lysates harvested at 18 hpi and mock-infected cell lysates as controls (Fig. 10A). Immunoprecipitation was carried using both anti-NS1 and anti-CAV-1 as primary antibodies. The presence of NS1, FKBP52, CAV-1, Cy40, and CyA in the precipitated immunocomplexes then was revealed by Western blotting. The presence of NS1 was clearly observed in the co-IP with anti-CAV-1, indicating the interaction between NS1 and CAV-1 in infected mosquito cells. This interaction was confirmed with the detection of CAV-1 in the co-IP with anti-NS1. In addition, the presence of FKBP52, Cy40, and CyA was also detected in the co-IP with anti-NS1. No proteins were detected in the co-IP with anti-NS1 when mock-infected cell lysates were used. These results, taken together, indicate that NS1 interacts with the CCC in DENV-infected C6/36 cells. In addition, the detection of FKBP52, CyA, and Cy40 in the co-IP with anti-CAV-1 in both mock-infected and infected cell lysates corroborates the colocalization data, indicating the existence of the CCC in mosquito cells.

FIG 10.

DENV NS1 interacts with proteins of the CCC in mosquito cells. (A) Immunoprecipitation assays with anti-NS1 and anti-CAV-1 antibodies in C6/36 cells. Mock-infected (M) and DENV4-infected (I) cell lysates (18 hpi) were processed for IP using anti-CAV-1 polyclonal antibody (n-20; Santa Cruz) and anti-NS1 polyclonal antibody (GTX124280; Genetex). After extensive washing, eluted protein complexes were analyzed for the presence of CAV-1, NS1, FKBP52, Cy40, and CyA by Western blotting. For the input, mock- and DENV4-infected cell lysates prior to the immunoprecipitation served as controls for protein detection. (B) Proximity ligation assay (PLA-Duolink) between NS1 and the CCC in infected mosquito cells. PLA signals (green) represent dual-recognition PLA against NS1 and CAV-1, FKBP52, Cy40, or CyA. Nuclei were stained with DAPI (blue). (C) Quantification of PLA signals (green spots) per cell on DENV4-infected C6/36 and Aag2 cells (PLA signals greater than 3 pixels) from panel B. PLA signals per cell were counted in each Duolink experiment in at least 20 cells using the Spot Detector plugin in Icy Image software. (B and C) C6/36 and Aag2 mock-infected cells were incubated with both primary antibodies, and infected cells incubated without CAV-1 antibody were included as negative controls. The bars represent mean PLA signals per cell ± standard errors from at least 20 cells. Data were evaluated using the Student's t test, and significant differences are denoted by asterisks (P ≤ 0.05).

To confirm direct interaction between NS1 and proteins of the CCC observed in the co-IP assays, we performed a proximity ligation assay (PLA) analysis in DENV-infected C6/36 and Aag2 cells. The PLA allows the detection of protein-protein interactions while preserving the cell architecture, and the generated signal can be quantified. In agreement with previous results (10), a positive signal (green dots) was detected in DENV-infected mosquito cells stained for both NS1 and CAV-1, confirming that NS1 and CAV-1 directly interact in infected mosquito cells (Fig. 10B). In addition, positive signals were detected between NS1 and each of the chaperone proteins of CCC, Cy40, FKBP52, and CyA (Fig. 10B), indicating that NS1 interacts with the whole CCC in both cell lines. In all cases, the PLA signals were located in the cytoplasm, in agreement with the colocalization results shown in Fig. 8. Quantification of the PLA signal per cell suggests that NS1 interacts equivalently with all components of the CCC, although significantly less signal was observed for FKBP52 and NS1 in C6/36 cells. No fluorescent spots were detected in mock-infected cells incubated with both primary antibodies or in infected cells incubated with anti-NS1 in the absence of anti-CAV-1, signifying the specificity of the signal (Fig. 10B and C).

ZIKV NS1 follows CAV-1-dependent unconventional secretion in mosquito cells.

NS1 is the only viral nonstructural protein secreted from DENV-infected cells. Other mosquito-borne flaviviruses, like ZIKV and yellow fever virus (YFV), also secrete NS1 (25). In order to compare these viruses with DENV, we tested the secretion properties of ZIKV and YFV NS1 in infected C6/36 and Vero E6 cells. Vero E6 cells were used as vertebrate cells in these experiments, since very low levels of infection were detected in BHK-21 cells with ZIKV and YFV (data not shown). Previously, the noncytotoxic concentrations for GCA and CsA were determined in Vero E6 cells (Fig. 11A). The presence of ZIKV NS1 was detected by enzyme-linked immunosorbent assay (ELISA) in the cell supernatants of both cell lines at 48 hpi, whereas YFV NS1 was detected at 72 hpi by Western blot assay (Fig. 11D). The secretion of YFV NS1 was found to be sensitive to BFA treatment in both mosquito and vertebrate cells (Fig. 11D). In contrast, the secretion of ZIKV NS1 was found to be insensitive to BFA and GCA treatment in infected mosquito but not in infected vertebrate cells (Fig. 11B). In addition, the secretion of ZIKV NS1 was reduced in mosquito, but not in vertebrate, cells treated with the CyA inhibitor CsA (Fig. 11B). The secretion of ZIKV and YFV virions was affected by BFA and GCA treatments in both mosquito and vertebrate cells (Fig. 11C and E), yet no effect on the number of ZIKV RNA copies was observed in the drug-treated cells (Fig. 11C). Colocalization experiments carried out in C6/36, Aag2, and Vero E6 cells infected with ZIKV and YFV showed a stronger, and significantly different, colocalization between ZIKV NS1 and CAV-1 than between YFV NS1 and CAV-1 in both mosquito cell lines. No colocalization of ZIKV or YFV NS1 with CAV-1 was observed in the vertebrate cell line (Fig. 11F and G). Of note, the colocalization data support the predictions derived from the presence of a conserved caveolin binding (CBD) domain in ZIKV (and DENV) but not in YFV NS1 (Fig. 11H). All the data taken together suggest that, like DENV NS1, ZIKV NS1 is secreted from mosquito cells by following an unconventional secretory route in association with CAV-1 and the CCC, while a classical secretory route is followed for ZIKV NS1 secretion in vertebrate cells. On the other hand, YFV NS1 seems to be secreted by a classical secretory pathway in both cell types.

FIG 11.

ZIKV and YFV NS1 secretion in infected mosquito cells. (A) Cell toxicity of GCA and CsA in Vero E6 cells. Serial dilutions of GCA and CsA were applied for 24 h. Data are means from 3 independent experiments ± standard errors; significant differences compared with values for the controls are denoted by asterisks (P < 0.0001). (B) NS1 secretion with inhibition of classical secretory pathway and CyA by CsA in ZIKV-infected C6/36 and Vero E6 cells at 48 hpi. (C) Viral release and RNA copies in ZIKV-infected cells treated with inhibitors of the classical secretion pathway and CsA at 48 hpi (n = 2). Total ZIKV RNA copies were calculated in 10⁵ cells and values plotted in a secondary y axis. ZIKV copies were measured by real-time RT-PCR using TaqMan technology as described in the text. (D) NS1 secretion with inhibition of classical secretory pathway in YFV-17D-infected C6/36 and Vero E6 cells at 72 hpi. Secreted YFV NS1 amount was evaluated by Western blotting and estimated by densitometric measuring with ImageJ software (57). Densitometric measuring was changed to percentage of NS1 secreted as described in the text. (E) Viral release in YFV-17D-infected cells treated with inhibitors of the classical secretion pathway at 72 hpi. (F) Colocalization between YFV NS1 or ZIKV NS1 and CAV-1 in mosquito and VeroE6 cells. C6/36, Aag2, and VeroE6 cells were infected at an MOI of 1 for 48 h. Cells were probed for CAV-1 (shown in red), viral NS1 (shown in green), and nuclei stained with DAPI (blue). (G) PCC for CAV1-NS1 were measured in at least 20 confocal independent images with 0.48-μm laser sections. The bars represent means ± standard errors. Data were evaluated using the 2-way ANOVA test, and significant differences are denoted by asterisks (P ≤ 0.05). (H) Protein sequence alignment of caveolin binding domain (CBD) in Flavivirus NS1. Φ, aromatic amino acid (F/Y/W); X, any amino acid. (B, C, D, and E) Cells were infected with YFV or ZIKV at an MOI of 3 for 1 h. The cells were washed 3 times and then treated with DMSO (control), 25 μM BFA, 27 μM GCA, or 9 μM CsA. For ZIKV infection, supernatants were harvested at 48 hpi, and for YFV infection, supernatants were collected 72 hpi. Secreted ZIKV NS1 was measured using in-house ELISA and represented as percentage of secreted NS1 compared to the level for DMSO. Viral titers in supernatants were measured by focus-forming assay in Vero E6 cells according to standard procedures. The experiments were performed in triplicate, and the bars represent means ± standard errors. Data were evaluated using the 2-way ANOVA test, and significant differences compared with levels for the DMSO treatment group are denoted by asterisks (P ≤ 0.05).

Modeling of molecular interactions between DENV NS1 and the chaperone caveolin complex.

To better understand the molecular interactions between the dimeric DENV NS1 protein and the CCC, we constructed a human CAV-1 three-dimensional (3D) model using the RaptorX server (http://raptorx.uchicago.edu) (26). The best template for the 3D CAV-1 model was PBD entry 2A65, with a P value of 2.69e−03. Molecular docking predicts a favorable interaction between the NS1 hydrophobic domain, including the grease fingers, where the CBD is located with a pocket in the CAV-1 scaffolding domain (CSD), with distances of fewer than 8 Å (Fig. 12A and B). The molecular docking results obtained with CAV-1 and NS1 were used as a control to carry on additional docking experiments between DENV NS1 or CAV-1 and the CCC chaperones FKBP52, Cy40, and CyA. Calculations of the energy binding of several docking clusters showed that the DENV NS1 has the lowest and most favorable energy for CAV-1 compared to the energy binding obtained for FKBP52, Cy40, or CyA. In addition, lower energies were obtained for the binding of CAV-1 with the other chaperones of the CCC than for the binding of NS1 with those same chaperones (Fig. 12C). Based on the free-energy data, we predict a model of interaction of NS1 with the CCC proteins in which FKBP52, Cy40, and CyA are directly bound to CAV-1 and NS1 is bound to the CCC through the CAV-1 scaffolding domain (Fig. 12D).

FIG 12.

In silico prediction of the interaction of DENV NS1 with the CCC. (A) Interaction of NS1 and human CAV-1 (isoform alpha; GenBank accession no. NP_001844.2). The CAV-1 3D model (red) was retrieved from the RaptorX server (26). CAV-1 scaffolding domain (CSD) is indicated in orange, and dimeric DENV2 NS1 (PDB entry 4O6B) is in green. (B) Interface distances in docking simulation of DENV NS1 and CAV-1. Dotted lines show measurement distances between amino acids (in Å) as estimated by the PyMOL Molecular Graphics System (version 2.0; Schrödinger, LLC). (C) Intermolecular binding energy for DENV NS1 and the CCC proteins. Energies of twenty clusters of protein dockings were plotted. ΔG (cal/mol) values were estimated by ClusPro server (https://cluspro.bu.edu/home.php). (D) Modeled prediction of DENV NS1 and CCC based on the intermolecular binding energies.

DISCUSSION

Our previous work demonstrated that DENV NS1 secretion in mosquito cell lines follows a BFA-insensitivity secretion pathway that depends on CAV-1; moreover, NS1 was found to interact directly with CAV-1 (10). Therefore, in the present work, we explored the association of NS1 with proteins involved, together with CAV-1, in the intracellular transport of cholesterol and the dependence of NS1 secretion on such association. The results revealed that NS1 is associated with the chaperone caveolin complex (CAV-1, Cy40, FKBP52, and CyA) and that the integrity of this complex is necessary for the secretion of NS1 in infected mosquito cells. Additional data suggest that in virus-infected mosquito cells, ZIKV NS1, but not YFV NS1, also associates with CAV-1 and usurps the cholesterol traffic pathways for secretion. Thus, the lipid content or molecular structure of the DENV NS1 seems to have taken advantage of mosquito cell cholesterol transport to guarantee efficient NS1 secretion in the mosquito vector. Interestingly, even though the CCC is present in vertebrate cells, NS1 is not associated with it and follows a strict classical secretion pathway.

The data obtained with the earlier inhibitors of the classical secretory pathway, GCA and Fli-06, corroborate and expand the previous data obtained with BFA regarding the use by NS1 of nonclassical secretory pathways in mosquito cells. GCA inhibits COPI-II recruitment and ER exit site (ERES) transport to the Golgi membrane (21). GCA insensitivity indicates that the NS1 and CAV-1 interaction initiates in the ER compartment. In agreement, treatment of mosquito cells with Fli-06 showed no effect on NS1 secretion, indicating that the interaction between NS1 and CCC takes place very early after NS1 synthesis, before the viral protein is transported to the ERES (22, 23).

Few works have reported a luciferase secretion system in insect cell lines (27). However, as an additional tool to study classical secretion in mosquito cell lines, an efficient Gaussia luciferase reporter system that enables the detection of luciferase in cell supernatants 6 h after transfection was developed. The secretion of luciferase was significantly affected after treatment with GCA and BFA (data not shown); these data, together with the observed inhibition of viral particle release, indicate the existence of a robust classical secretion pathway in mosquito cells. Although mosquito cells present both classical and unconventional secretory pathways, DENV NS1 is secreted bypassing the Golgi apparatus. A study in mosquito cells demonstrated that insect cells have an active ER retention protein system: the KDEL C-terminal peptide functions as an ER retention signal for some proteins (28). NS1 lacks a KDEL signal, yet it presents a conserved and well-exposed caveolin binding domain (CBD) (10). Thus, NS1 unconventional secretion in mosquito cells is likely associated with this CAV-1 motif. However, the CBD may not be the unique factor to determine unconventional secretion. A role for amino acids at positions 10 and 11 in flavivirus NS1 secretion in mosquito cells has been proposed (29).

The CCC was described in vertebrate cells as a nonvesicular cholesterol transporter within the cell (14). However, to our knowledge there are no previous reports of the presence of the CCC in any insect or mosquito cell line. Thus, given that there are differences in the lipid homeostasis in mosquito cells, including the absence of de novo cholesterol synthesis pathways, it became necessary to demonstrate the existence of the CCC in mosquito cells and to show a role for the complex in lipid homeostasis, as have been shown for vertebrate cells (12, 13). Colocalization and immunoprecipitation experiments indicated that in uninfected mosquito cells, CAV-1 is indeed associated with all 3 chaperones proposed to form the complex, FKPB52, Cy40, and CyA; moreover, knocking down the expression of CAV-1 (data not shown) or any of the associated chaperones resulted in a significant decrease in the lipid droplet count per cell, indicating a role for the CCC in lipid homeostasis in mosquito cells. In agreement with our results, ER chaperones have been reported to be necessary for the secretion of recombinant proteins expressed in insect cells using a baculovirus system (30, 31).

The NS1 and CCC interactions in infected mosquito cells was validated by colocalization, coimmunoprecipitation, and proximity ligation assays. First, the colocalization analysis and coimmunoprecipitation assays carried out in mock-infected cells demonstrated direct interaction of FKBP52, Cy40, and CyA with CAV-1, again supporting the existence of the CCC in mosquito cells. In agreement with our previous report, strong CAV-1 and NS1 interaction were detected by both techniques and used as a reference for the new analysis with the CCC. The capture with anti-NS1 antibodies indicated that FKBP52, Cy40, and CyA interact with NS1 in infected cells. These results were confirmed by proximity ligation assays, a powerful approach which allows us to determine and quantify protein-protein interactions in situ (32). PLA signals between NS1 and each of the CCC proteins in infected mosquito cells were demonstrated to be similar to those observed between CAV-1 and FKBP52, Cy40, and CyA (10). CAV-1-dependent secretion has been reported for nonviral proteins, such as Kallikrein 6 in colon cancer cells and Cyr61 in lung cells (33, 34), as well as nonstructural viral proteins, such as rotavirus NSP4 (35, 36).

The structure of the CCC is unknown, and crystal models exist for all 3 chaperones but not for caveolin-1. We performed docking molecular analysis with the dimeric NS1 and complete CAV-1 (3D modeled in this work), as well as with the CCC proteins. An approximated in silico model for NS1 and CAV-1 interaction showed very high favorable energy binding and suggests interaction through the plasma membrane (10). The molecular docking also allows measuring the energy binding and distances between both crystals. The CBD in NS1 interacts with the scaffolding domain (CSD) with a distance of approximately 5 Å. Further analysis also showed high energy binding through hydrophobic dynamics between the NS1 β-roll domain and FKBP52, CyA, and Cy40, with distances under 6 Å. However, the model predicts all these proteins interact at the same domain in NS1 (data not shown), which seems unlikely due to steric effects. Thus, we favor an alternative molecular docking model, based on interaction energies, in which NS1 is bound directly to CAV-1 and, in turn, FKBP52, CyA, and Cy40 are bound to CAV-1 (Fig. 12D). This model still allowed interaction of NS1 with the CCC through membranes. However, if the NS1-CCC complex will traffic to the plasma membrane as a free cytosolic complex or membrane associated is totally unknown.

Dependence of NS1 secretion on the integrity of the CCC was evaluated using gene knockdown for FKBP52 and Cy40 or drug inhibition for CyA. Again, the observed reduction in the number of lipid droplets per cell after each treatment, similar to the decrease observed after serum starvation and methyl-β-cyclodextrin (MβCD) treatment (Fig. 5B), support a role for the CCC in the lipid homeostasis in mosquito cells. FKBP52 knockdown decreases NS1 secretion while viral release was not affected. HSP90 has been associated with viral replication of chikungunya virus, Arteriviridae viruses, and hepatitis C virus and has a minor effect on DENV replication (37–40). However, our results indicate that HSP90 is not associated with NS1 secretion (Fig. 3E) but indicate the requirement of FKBP52 for NS1 exit in mosquito cells. CyA inhibition by cell treatment with CsA also resulted in diminished NS1 secretion in mosquito cells without affecting viral release. However, CyA has been associated with viral packing and viral release in other viruses, such as coronaviruses, hepatitis C virus, HIV, and others (41–43). Finally, Cy40 knockdown had no effect on either NS1 secretion or viral release. This result suggests a dispensable function of Cy40 in the secretion of NS1 in the mosquito cell; however, this is an unexpected result for which we have no obvious explanation, given the observed association between NS1 and Cy40, which appears to be part of the CCC. Cy40 may also have an unknown regulatory activity in the CCC complex, becoming a dispensable component for NS1 traffic in the infected mosquito cells. However, Cy40 mitochondrial function is essential for viral fitness in human coronaviruses and hepatitis C virus (44, 45). Finally, regardless of the previous data indicating no interaction of the CCC proteins with NS1 in infected vertebrate cells, we wanted to explore the effect of FKBP52 and Cy40 knockdown and CyA inhibition in BHK-21 DENV-infected cells. As expected, alteration in the levels or function of these proteins resulted in a significant decrease in the number of lipid droplets but had no effect on NS1 secretion, indicating that the CCC plays a role in lipid homeostasis in vertebrate cells but has no role in the secretion of NS1 and supporting a classical secretion pathway for NS1 in vertebrate cells.

Profound changes in lipid homeostasis have been observed in DENV-infected mosquito cells (46), which may in part respond to the demand of large amounts of cholesterol required for the hexameric NS1 secretion (6). Interestingly, the levels of CAV-1 and FKBP52 were found to increase during DENV infection in mosquito, but not in vertebrate, cells. Mosquitos are heterotrophs for cholesterol; thus, increased expression of CAV-1 and FKBP52 may result as a response to increments in cholesterol demand upon infection (47, 48), as suggested by the increase in CAV-1 expression observed after MβCD treatment of mock-infected cells.

NS1 lipid cargo indicates high requirements of cholesterol and triglycerides in infected cells (6). In hepatic cells, DENV infection induces an increase in the activity of cholesterol biosynthesis enzymes (49). The results showing lipid droplet decay during CCC inhibition in mosquito cells suggest an association between NS1 and LD during its traffic. To explore this possibility, colocalization parameters between LD and NS1 in infected mosquito cells were measured (Fig. 5C). However, very low values for true colocalization between NS1 and LD were observed in both infected mosquito cell lines. Lipid droplets are a reservoir of triglycerides and cholesterol for viral production. In addition, the C capsid protein has been shown to colocalize with LD in infected cells, and this interaction is necessary for genome encapsidation (50). In general, flaviviruses increase production of free fatty acids and cholesterol to favor assembly of viral particles (51), and NS1 secretion is affected in cells where lipid metabolism is altered, yet LD do not appear to participate in the traffic of this viral protein.

Previous results indicated that the interaction of CAV-1 and NS1 is important for the secretion of NS1 in infected mosquito cells (10). NS1 presents a CBD (ΦXXΦXxXXΦ, where Φ is an aromatic residue and X is any amino acid) that is well conserved among the 4 DENV serotypes and is also conserved in the ZIKV but not in the YFV NS1, where a T-to-F substitution is found in the first aromatic residue. Since changes in the aromatic residues of the CBD are important in the interaction with CAV-1 (52), the possible interactions of the NS1 of ZIKV and YFV with CAV-1 and the role of such interactions in influencing the secretory routes of NS1 from infected mosquito cells were evaluated. The colocalization results obtained for ZIKV- and YFV-infected cells as well as the differential effect of BFA and GCA treatment in the secretion of NS1 of both viruses, taken together, support the notion that in mosquito cells, but not in vertebrate cells, the presence of an intact CBD is necessary for the interaction between NS1 and CAV-1, and that such interaction is necessary to drive NS1 to be secreted by a nonclassical secretory route, in association with the CCC. YFV NS1 showed a favored classical Golgi secretion pathway, while ZIKV NS1 secretion follows an unconventional secretion pathway, as we have described for DENV. Thus, the CBD appears to be a molecular determinant necessary to direct the interaction of NS1 to CAV-1 and facilitate the unconventional secretion using the cholesterol pathway and CCC (52). Remarkably, the CBD in the NS1 sequence by itself is not the only determinant of what type of secretion is going to take the NS1 into the cells, since in vertebrate cells, NS1 from DENV and ZIKV shows no association with the CCC and secretion is Golgi membrane dependent. Several factors acting in concert may be necessary to keep the NS1 from reaching the ERES and entering the classical secretory pathway in mosquito cells: (i) mosquito cells could be defective in a protein necessary to direct NS1 to the ER-Golgi intermediate and then follow classical secretion (28), as is suggested by the insensitivity of NS1 secretion to FLI-06 and GCA treatment, and (ii) the affinity of DENV NS1 for mosquito CAV-1 may be higher for the CAV-1 of vertebrate cells, as is suggested by the predicted interaction energies for both proteins, where DENV NS1 presents stronger interaction for invertebrate CAV-1 than for human CAV-1 (data not shown). Figure 13 presents a schematic model for DENV NS1 secretion in infected mosquito cells.

FIG 13.

Proposed model for unconventional secretion of DENV NS1 through CCC in mosquito cells. (A) DENV replication complexes. Viral RNA is synthesized within RC with the assistance of viral nonstructural proteins. NS1 interacts with the nonstructural proteins to assist membrane bending and envelopment of nucleocapsids. Virions are packed and transported to ERGIC vesicles. NS1 dimer interacts with ER inner membrane in RC. (B) Virions are incorporated into vesicles within the Golgi membrane and transported through the trans-Golgi network. Finally, mature virions are released to extracellular space by following a classical Golgi pathway (BFA or GCA sensitive). (C) NS1 interacts with CAV-1 scaffolding domain through its β-roll domain in the ER-lumen. Proteins FKBP52, Cy40, and CyA interact with CAV-1 and assemble the caveolin chaperone complex (CCC). The NS1-CCC complex is detached from ER exterior membrane by unknown mechanisms, bypassing the ERES. (D) Cytoplasmic NS1-CCC complex follows the intracellular cholesterol pathway until grasping the inner plasmatic cell membrane, and then hexameric NS1 is released to the extracellular space. NS1 does not interact with LD during trafficking. RC, replication complex; ERGIC, ER-Golgi intermediate compartment.

In summary, these results indicate that in mosquito cells, DENV NS1 utilizes the CCC as part of the cholesterol transport machinery to reach the extracellular space, bypassing the Golgi complex. A similar pathway appears to exist for the ZIKV NS1. These results are in congruence with the lipoprotein nature of NS1 and uncover novel functions for the CCC as a protein transporter in mosquito cells. The presence of the CBD in NS1 appears to be important to mediate the interaction between NS1 and CAV-1. However, more research is needed to fully understand the viral and cellular factors that determine the dramatic differences in traffic route followed by DENV NS1 in mosquito and vertebrate cells. NS1 proteins from other mosquito-borne flaviviruses that lack the CBD, such as West Nile virus, YFV, and Japanese encephalitis virus, are secreted apparently at very low levels and by following the classical pathway in mosquito cells. The functions of soluble NS1 in the mosquito are unknown and need to be elucidated but may include facilitation of the spread of viral particles and modulation of innate immunity (31). Finally, manipulation of the lipid and cholesterol system transport in the mosquito cell may become a novel target to reduce NS1 secretion and a novel strategy to block the spread of mosquito-borne flaviviruses such as dengue virus.

MATERIALS AND METHODS

Cells.

C6/36 clone mosquito cells from Aedes albopictus (CRL-1660; ATCC) and Aag-2 cells from Aedes aegypti (kindly provided by Fidel de la Cruz from CINVESTAV-IPN) were grow at 28°C in Eagle’s minimum essential medium (EMEM) (30-2003; ATCC), supplemented with 5% fetal bovine serum (FBS) and 100 U/ml penicillin-streptomycin. Baby hamster kidney cells (BHK-21; ATCC CCL-10) were grown at 37°C and cultured in EMEM supplemented with 5% FBS and 100 U/ml penicillin-streptomycin. Monkey epithelial kidney cells (Vero E6; CRL-1586; ATCC) were grown at 37°C and cultured in EMEM supplemented with 10% FBS and 100 U/ml penicillin-streptomycin.

Viruses.

Dengue virus serotype 2 (DENV2) strain New Guinea, Dengue virus serotype 4 (DENV4) strain Philippines H241, and a Mexican isolate from A. aegypti of Zika virus (ZIKV), Asian genotype, were generously provided by Mauricio Vázquez, (Laboratorio de Arbovirus y Virus Hemorrágicos, Instituto de Diagnóstico y Referencia Epidemiológicos [InDRE], Mexico City, Mexico). The dengue viruses were propagated in suckling mouse brain (ICR; CD-1), provided by the Unit of Production and Experimentation of Laboratory Animals of CINVESTAV-IPN, Mexico (UPEAL-CINVESTAV-IPN), as previously described (53). DENV titers in mouse brain homogenates were determined by focus-forming assay in BHK-21 cells. Serial dilutions of viral stocks or supernatants of experiments were diluted in serum-free medium and added to cell monolayers in 96-well plates. Viral attachment and entry were allowed to proceed for 2 h at 37°C, and then EMEM–10% FBS was added and incubated for an additional 48 h before fixation and quantitation of infected cells by labeling with anti-flavivirus E-glycoprotein antibody using a mouse VECTASTAIN ABC-horseradish peroxidase (HRP) kit (PK-4002; Vector Laboratories) and a DAB peroxidase substrate kit (SK-4100; Vector Laboratories). ZIKV was propagated in C6/36 cells. YFV-17D vaccine strain stock was donated by Juan Salas Benito (Escuela Nacional de Medicina y Homeopatía, IPN) and obtained from passages of the commercial vaccine in C6/36 cells. Titers of infection experiments with YFV and ZIKV and stock viruses were determined in Vero E6 cells using focus-forming assay (as described previously).

Reagents and drug treatment.

BFA (B6542; Sigma-Aldrich), GCA (1139889-93-2; Calbiochem), Fli-06 (SML0985; Sigma-Aldrich), and CsA (C1832; Sigma-Aldrich) were dissolved in dimethyl sulfoxide (DMSO; ATCC). CsA was used to pharmacologically inhibit cyclophilin A. Methyl-β-cyclodextrin (MβCD; C4555; Sigma-Aldrich) was dissolved in water and used to inhibit mobilization of cholesterol and to alter lipids within cells. Unless indicated otherwise, the experimental concentrations of BFA, GCA, Fli-06, CsA, and MβCD were 25 μM, 27 μM, 100 μM, 9 μM, and 1 mM, respectively. Cells were grown in 24-well plates (2 × 10⁵ cells per well) and infected for 1 h and then washed three times with phosphate-buffered saline (PBS). The drugs BFA, GCA, CsA, and Fli-06 were added to the cells in EMEM–5% FBS. Unless indicated otherwise, incubation time was 24 h at 28°C for mosquito cells and 37°C for BHK-21 cells. After this time, cell supernatants were collected to measure secreted NS1 and virus progeny. In other cases, cells were fixed and stained for immunofluorescence.

Quantitative real-time PCR for viral RNA.

Levels of viral genomic RNA were quantified in BFA- and GCA-treated and nontreated cells by real-time RT-PCR. After the supernatants were collected for NS1 and virus titer determinations, the cell monolayers were washed and the total RNA was isolated with TRIzol reagent (Invitrogen) according to the manufacturer’s procedures. A total of 0.3 μg of total RNA was employed to determine the number of genomic DENV2, DENV4, or ZIKV RNA copies. The one-step quantitative RT-PCR was standardized to 25 μl using the QuantiTect probe RT-PCR kit (Qiagen, Valencia, CA) with 500 nM each primer and 50 nM labeled probes (DENV2-Texas red, DENV4-VIC, and ZIKV-FAM; TaqMan). Detection primers and probes for DENV2, DENV4, and ZIKV were as described in Chien et al. (54) and Lanciotti et al. (55). Cycling conditions for ZIKV were 50°C for 30 min, 95°C for 15 min, 45 cycles of 95°C for 15 s, and 60°C for 1 min. Cycling conditions for DENV2/4 were 50°C for 30 min, 95°C for 15 min, 50°C for 30 s, 72°C for 1 min, 45 cycles of 95°C for 15 s, and 48°C for 3 min. Runs were carried out with Applied Biosystems 7500 Fast real-time PCR in accordance with the manufacturer’s instructions (Applied Biosystems, Foster City, CA), and results were analyzed using SDS v1.5.1 software. Each assay was performed in duplicate with two technical replicates, and each assay included no-template negative controls and DENV2, DENV4, and ZIKV positive controls. Results were expressed as the total number of RNA copies determined in 1 × 10⁵ cells.

Cell viability assays.

Cells were seeded in 96-well plates and then treated with serial dilutions of the indicated drugs or siRNAs. Cell viability was measured with the Cell Titer 96 AQueous nonradioactive cell proliferation assay according to the manufacturer’s procedures (MTS assay, G3580; Promega) and expressed as a percentage of the control (DMSO treated). The number of viable cells in each sample (2 × 10⁴ cells per well) was determined in triplicate. Triplicate measurements were then averaged, and the percentage of viable cells was determined relative to cells treated as controls.

Golgi apparatus integrity after drug treatment.

Cells grown to subconfluency on coverslips in 24-well plates were treated with GCA or Fli-06 for 24 h. After fixation with 4% paraformaldehyde, the cells were stained with rabbit monoclonal anti-GM130 antibody 1/300 (Sigma-Aldrich). The samples were stained with the conjugated secondary antibody Alexa Fluor 488, and Golgi apparatus integrity was analyzed with a Nikon epifluorescence microscope (Eclipse Ti-U).

Measurement of secreted NS1 protein.

The Platelia NS1 antigen (Ag) (82830; Bio-Rad, Hercules, CA) commercial kit was used to determine the levels of soluble NS1 in the cell supernatants collected from DENV-infected cells. The assay was carried out by following the procedure indicated by the manufacturer. The presence of ZIKV NS1 was measured using a noncommercial, in-house ELISA. Briefly, ELISA 96-well plates (Nunc-Immuno; Sigma-Aldrich) were coated with 200 ng of purified anti-NS1 monoclonal antibody 7E11 in carbonate buffer (0.05 M, pH 9.6) and incubated overnight at 4°C. Nonspecific binding was blocked by incubating with 100 μl/well of blocking buffer (PBS with 10% fetal bovine serum) for 1 h at 37°C. At that point, supernatant samples were added (50 μl/well) and the plate incubated for 1 h at 37°C. Fifty μl/well of anti-NS1 7E11 conjugated with biotin for 1 h at 37°C was added, followed by 50 μl/well of streptavidin conjugated with HRP diluted 1:10,000 in PBS and incubated for 1 h at 37°C. After each step, wells were washed by rinsing three times with washing buffer (PBS with 0.01% Tween 20). The reaction was developed with the addition of 160 μl/well of TMB (Sigma-Aldrich) for 15 min and stopped with the addition of 50 μl/well of 2 M H₂SO₄. The color development reaction is proportional to the amount of secreted NS1. The mean value of A₄₅₀ obtained under the control conditions (DMSO treatment or irrelevant siRNA) was taken as 100% of NS1 secretion. Absorbance results obtained in the drug- or siRNA-treated cells were expressed as a percentage of the control.

Luciferase reporter construct.

Ac5-STABLE2-neo was a gift from Rosa Barrio and James Sutherland (32426; Addgene) (56). This plasmid was engineered to express the luciferase gene in mosquito cells. The luciferase gene was obtained from plasmid ptk-GLuc vector (N8084; NEB), kindly donated by Susana Lopez (Instituto de Biotecnologia-UNAM). Gaussia luciferase (GLuc) was PCR amplified using the following designed primers with directed cloning sites (underlined): Luc-XbaI-F (5′-GCGCTCTAGAGCCACCATGGGAGTCAAAGTTCTGTTTGCC-3′) and Luc-HindIII-R (5′-GAGTAAGCTTTGGCGGGTCACCACCGGCCCCCTTGATC-3′). Unique sites XbaI and HindIII in pAc5-stable2-neo allowed replacement of green fluorescent protein (GFP). The XbaI-HindIII cassette was cloned into pAc5-STABLE2-Neo, generating the pAc5-mCherry-GLuc-Neo vector. The promoter Actin5C (from Drosophila melanogaster) is efficient in insect cell lines. The T2A peptide sequence derived from Thosea asigna (EGRGSLLTCGDVEENPGP) allowed multicistronic processing, and the neomycin resistance gene (NeoR) confers resistance to G418, allowing stable mosquito cell lines. A schematic representation of pAc5-mCherry-GLuc-neo vector is shown in Fig. 1J.

Vector transfection and luciferase activity assay.

pAc5-mCherry-GLuc-neo vector was transfected into a confluent monolayer of C6/36 cells using Lipofectamine 2000 reagent (Invitrogen). Each 24-well plate was transfected with 1 μg of plasmid DNA and 2 μl of Lipofectamine. After 5 h of transfection, cells were added to EMEM with a final 10% FBS. After 24 h, selective G418 was added to obtain the stable C6/36 cell line. Stable C6/36 cells were treated with BFA or GCA for 24 h. BHK-21 cells were transfected with pkt-GLuc vector, and drug treatment followed the procedure detailed above. Subsequently, the supernatants were analyzed with a BioLux Gaussia luciferase assay kit (E3300; NEB). Firefly luciferase activity in the medium (in relative light units [RLU]) was measured in a Fluoroskan Ascent FL (Thermo Fisher Scientific). RLU means from control, DMSO-treated cells were taken as 100% of luciferase activity, and results are expressed as percentages of the control value.

Gene knockdown with siRNAs.

C6/36 and BHK-21 cells were transfected with a range of concentrations of siRNA targeted against proteins of the chaperon caveolin complex using HiPerFect transfection reagent (Qiagen). FlexiTube siRNAs (4 siRNAs recommended per gene; Qiagen) for FKBP52, CAV-1, and Cy40 were used at the concentrations indicated in the text. AllStars negative-control siRNA (Qiagen) was used as a nonsilencing siRNA control at the same concentration as the siRNAs for the proteins of interest. Gene knockdown was performed using approximately 7 × 10⁵ cells per well and assessed by Western blotting. Protein expression was normalized to the expression of β-actin (from CINVESTAV-IPN), glyceraldehyde-3-phosphate dehydrogenase (GAPDH) (GTX100118; Genetex), or ERK-1 (sc-281291; Santa Cruz Biotechnology, Inc.). Expression levels relative to negative-control cells were estimated by densitometric measuring with ImageJ software (57).

Lipid droplet count.

Lipid droplet count per cell was used as an indicator to determine the effect on lipid homeostasis by drug or siRNA treatment. Cells were fixed and washed three times in PBS. Stock solution of 500 μg/ml Nile red (82485; Sigma-Aldrich) in acetone was prepared and stored protected from light. The dye was then added directly to the preparation at 0.5 μg/ml in PBS, and the preparation was incubated for 30 min. After three washes in PBS, excess dye was removed. From this point forward, none of the suspension media contained serum, albumin, or detergents to avoid draining the dye out of the cells. Coverslips then were mounted in Fluoroshield with 4’,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich). Lipid droplets were numbered in maximum projection images obtained by an LSM 700 confocal microscope. The total amount of lipid droplets was determined in at least 15 cells using the Spot Detector plugin (58) in Icy Image software (Institute Pasteur, France) (59).

Confocal microscopy.

Confluent cell monolayers, grown in 24-well plates containing glass coverslips, were infected with DENV, ZIKV, or YFV-17D at an MOI of 3. After the times indicated in the text, cells were fixed in 4% paraformaldehyde for 10 min. Cells were permeabilized with 0.1% Triton X-100 for 10 min at room temperature and stained for DENV-NS1 using anti-NS1 2B7 (1/300), anti-CAV-1 (1/300) (GTX89541 [Genetex] or sc-894 [Santa Cruz]), anti-FKBP52 (1/300) (ab129098; Abcam), anti-Cy40 (1/300) (GTX104038; Genetex), and anti-CyA (1/300) (GTX104698; Genetex); nuclei were stained with DAPI. Conjugated anti-mouse Alexa-488 or Alexa-598, anti-goat Alexa-568, and anti-rabbit Alexa-648 or Alexa-488 (donkey preadsorbed secondary antibodies; Abcam) were used at 1/800. For lipid droplet colocalization with DENV NS1, cells were incubated with BODIPY 493/503 (Invitrogen) at 2 μM in PBS free of serum or detergents. Coverslips were mounted in Fluoroshield with DAPI (Sigma-Aldrich). The images were analyzed using a Carl Zeiss LSM 700 confocal microscope. To maintain the consistency of the green color for the viral protein, the color of BODIPY was changed to red, like Nile red staining. To evaluate the colocalization between proteins, PCC were obtained from at least 20 independent confocal images (laser sections are indicated in the text) using Icy Image software and the colocalization studio plugin (59).

Western blotting.

Total cells were lysed in lysis buffer (25 mM Tris-HCl, pH 7.6, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 5% glycerol) with protease inhibitor cocktail (P8340; Sigma-Aldrich) added and assayed for total protein concentration using a Pierce bicinchoninic acid protein assay kit (Thermo Scientific). Twenty μg of samples was subsequently prepared for electrophoresis by adding 4× Laemmli gel loading buffer (40% glycerol, 240 mM Tris-HCl, pH 6.8, 8% SDS, 0.04% bromophenol blue, 5% β-mercaptoethanol). Samples were then boiled and resolved in SDS–10% polyacrylamide gels, transferred to nitrocellulose membrane (0.45 μm; Bio-Rad), and incubated with primary antibodies diluted in 5% skin milk powder in PBS–0.1% Tween 20. After washing, membranes were incubated with secondary antibodies conjugated to HRP (anti-mouse-HRP 115-035-003; 1/20,000; Jackson ImmunoResearch; or anti-rabbit-HRP GTX26821; 1/40,000; Genetex) diluted in 5% skin milk powder in PBS–0.1% Tween 20. HRP was detected using SuperSignal West Femto maximum sensitivity substrate (Thermo Scientific). Digital images were acquired with a Fusion FX Spectra (Vilber) and analyzed with ImageJ software (57).

Immunoprecipitation.

Lysates was harvested from DENV-infected cells using ice-cold nondenaturing IP lysis buffer (PBS, pH 7.4, 1% Triton X-100, 150 mM NaCl, 5% glycerol, and protease inhibitor cocktail). Co-IP was done using the Thermo Scientific Pierce co-IP kit (26149) by following the manufacturer's protocol. Briefly, 10 μg DENV NS1 polyclonal antibody (GTX124280; Genetex) and 10 μg anti-CAV-1 polyclonal (n-20; Santa Cruz Biotechnology) were first immobilized for 2 h using AminoLink Plus coupling resin. The resin with the attached antibodies was then washed and incubated with the cell lysates overnight. After incubation, the resin was washed again (PBS, pH 7.4, 150 mM NaCl, 5% glycerol) and proteins eluted using elution buffer (50 mM HEPES, pH 5.0). A mock infection control was used to assess nonspecific binding and received the same treatment as the co-IP samples, including the CAV-1 antibody. In this control, the coupling resin is not amine reactive, preventing covalent immobilization of the primary antibody onto the resin. Samples were analyzed by immunoblotting using anti-CAV-1, anti-FKBP52, anti-Cy40, anti-CyA, and anti-dengue NS1 (antibodies previously described). Membranes were then incubated with HRP-conjugated secondary antibodies (anti-rabbit-HRP or anti-mouse-HRP as described above) and developed using enhanced chemiluminescence (Thermo Scientific).

Proximity ligation assay.

Interactions between the NS1 and the CCC proteins were detected by a Duolink PLA kit (Sigma-Aldrich) in DENV4-infected mosquito cells (MOI of 3). After 18 hpi, cells were fixed and permeabilized in methanol. After preincubation with a blocking agent for 1 h, samples were incubated overnight with the primary antibodies (used at a 1/100 dilution). Duolink PLA probes detecting rabbit or mouse antibodies were diluted in the blocking agent at a dilution of 1:5 and applied to the slides, followed by incubation for 1 h in a preheated humidity chamber at 37°C. The PLA probe anti-rabbit plus binds to the CAV-1, Cy40, CyA, or FKBP52 primary antibody, whereas the PLA probe anti-mouse minus binds to the dengue NS1 antibody. If the distance between both probes is <40 nm, a signal with Duolink PLA is generated and detected in fluorescein isothiocyanate dye emission, indicating an interaction of both proteins. Unbound PLA probes were removed by washing in buffer A (10 mM Tris-HCl, pH 7.4, 150 mM NaCl, 0.05% Tween 20). For hybridization of the two Duolink PLA probes, Duolink hybridization stock (dilution, 1:5) was used. Slides were incubated in a preheated humidity chamber for 15 min at 37°C. The samples were incubated in the ligation solution consisting of Duolink ligation stock (1:5) and Duolink ligase (1:40) for 90 min at 37°C. Detection of the amplified probe was done with the Duolink green detection kit. Duolink detection stock was diluted 1:5 and applied for 1 h at 37°C. Final washing steps were done in buffer B (200 mM Tris-HCl, pH 7.5, and 100 mM NaCl). The slides then were mounted with in situ mounting medium with DAPI (Sigma-Aldrich) and visualized by an LSM 700 confocal microscope. The PLA signals per cell were determined in at least 10 cells in maximum projection images using the Spot Detector plugin in Icy Image software (59). Mock-infected cells incubated with both primary antibodies or DENV-infected cells incubated only with anti-CAV-1 antibodies were included as negative controls.

Protein-protein interaction in silico analysis.

3D crystal structures of dimeric dengue NS1 (PDB entry 4OIG) and proteins of the CCC, FKBP52 (4LAV), CyA (1OCA), and Cy40 (1IHG) were taken from the PDB data bank and analyzed using the ClusPro protein-protein docking experiments server (https://cluspro.bu.edu/home.php). The highest favorable binding energies were obtained by models using scoring schemes in hydrophobic interactions. Sixteen clusters of low-energy docked structures were analyzed with each energy parameter and compared with dimeric dengue NS1. Since no amino acid sequence from CAV-1 of any mosquito species is annotated and no crystal structure for the human CAV-1 is found in the PDB data bank, the CAV-1 putative 3D structure based on the amino acid sequence of human CAV-1 (isoform alpha; GenBank accession no. NP_001844.2) was modeled using a protein structure prediction server (60). CAV-1 putative structure then was used in ClusPro protein-protein docking analysis with NS1 and proteins of CCC (61). Dockings were performed with the balanced free-energy function, and the values were statistically analyzed and compared (26). 3D structures and atom-to-atom distances were represented and measured using PyMOL software (PyMOL Molecular Graphics System, version 2.0; Schrödinger, LLC).

Statistical analysis.

Values of all assays were expressed as means ± standard errors from three independent experiments, each in triplicate or as indicated in the text. Statistical analyses were carried out using GraphPad Prism, version 6.01, software.