Remodeling of the Actin Network Associated with the Non-Structural Protein 1 (NS1) of West Nile Virus and Formation of NS1-Containing Tunneling Nanotubes

The cellular response to the recombinant NS1 protein of West Nile virus (NS1WNV) was studied using three different cell types: Vero E6 simian epithelial cells, SH-SY5Y human neuroblastoma cells, and U-87MG human astrocytoma cells. Cells were exposed to two different forms of NS1WNV: (i) the exogenous secreted form, sNS1WNV, added to the extracellular milieu; and (ii) the endogenous NS1WNV, the intracellular form expressed in plasmid-transfected cells. The cell attachment and uptake of sNS1WNV varied with the cell type and were only detectable in Vero E6 and SH-SY5Y cells. Addition of sNS1WNV to the cell culture medium resulted in significant remodeling of the actin filament network in Vero E6 cells. This effect was not observed in SH-SY5Y and U-87MG cells, implying that the cellular uptake of sNS1WNV and actin network remodeling were dependent on cell type. In the three cell types, NS1WNV-expressing cells formed filamentous projections reminiscent of tunneling nanotubes (TNTs). These TNT-like projections were found to contain actin and NS1WNV proteins. Interestingly, similar actin-rich, TNT-like filaments containing NS1WNV and the viral envelope glycoprotein EWNV were also observed in WNV-infected Vero E6 cells.

In this study, we further investigated the multiple properties of the NS1 protein of the West Nile virus (NS1 WNV ). We tested the capacity of the extracellular, secreted form of recombinant NS1 WNV (sNS1 WNV ) to attach and enter cells, and to possibly affect the cytoskeletal network upon its cellular uptake. The cellular alterations associated with the expression of recombinant NS1 WNV proteins were also analyzed. Three different cell types were used in our study: simian epithelial cells (Vero E6), human neuronal cells (SH-SY5Y), and human glial cells (U-87MG). We observed that sNS1 WNV attachment to cells and cellular uptake varied with the cell type and was more efficient in Vero E6 and SH-SY5Y cells than in U-87MG. Modification of the cellular actin filament network was found to be associated with NS1 WNV in different types of experimental setups: (i) Addition of purified secreted sNS1 WNV protein to the cell culture medium (exogenous sNS1 WNV ) resulted in a significant but transient remodeling of the actin cytoskeleton in Vero E6 cells. (ii) Upon intracellular expression of recombinant NS1 WNV protein (endogenous NS1 WNV ), actin-rich, TNT-like projections containing NS1 WNV were observed in the three cell types. (iii) Similar filamentous projections forming intercellular bridges were observed in WNV-infected Vero E6 cells in the late stages of the virus cycle, and these actin-rich filaments were found to contain NS1 WNV and the viral envelope glycoprotein E WNV .
Interestingly, incubation of cells with sNS1 WNV , added before or concomitantly to WNV, did not significantly change the infectiveness of the virus in any of the studied cell types. Collectively, our results confirmed the multiplicity of functions associated with the flaviviral NS1 proteins, but the common molecular mechanism remains to be determined.
The IS-98-ST1 strain of WNV (provided by Dr. S. Lecollinet; ANSES-ENVA, Maisons-Alfort, France) is a highly virulent strain isolated from a stork with severe neurological symptoms during the 1998 epidemic in Israel [62]. Working stocks of WNV IS-98-ST1 were generated by a single round of amplification in Vero E6 cells. Titers of virus stocks were determined by a standard serial dilution plaque assay in Vero E6 cells, and results expressed as PFU/mL [1,63].

Cloning of NS1
The NS1-coding gene from the WNV strain IS-98-ST1 was cloned in fusion with the coding sequence of a flexible linker (Gly-Ser-Gly) and an oligo-histidine tag (6× His) at the carboxy terminus. A similar tag has been previously added to the NS1 DENV protein, without any detrimental effect on its structural and biological properties [10,64]. The coding sequence of the C-terminal region of the E gene (72 nucleotides) was conserved at the 5' end of the construct, as it contains the ER-targeting sequence and the host protease cleavage site. Bam HI and Xho I restriction sites were added at the 5' and 3' ends, respectively, using polymerase chain reaction (PCR) with the following primers: forward Viruses 2019, 11, 901 4 of 20 5'-GGATCCAATACGGAATTCATGCAGCTGTTGAATTTTGACCTTCTCAAGCTTGCG-3', and reverse 5'-CTCGAGTCACTAGTGGTGATGGTGATGATGGCCGCTTCCAGCATTCACTTGTGACTACG-3'. The fragment thus obtained was digested by Bam HI and Xho I (New England BioLabs, NEB™, Ipswich, MA, USA) and inserted into the Bam HI/Xho I-linearized pcDNA3.1 expression vector, using T4 DNA Ligase (NEB™), according to the manufacturer's instructions. The chemically competent E. coli bacterial strain Top10 was used for plasmid construct propagation. Standard procedures were followed, except for the temperature of transformed bacterial cells propagation (28 • C). The quality of the plasmid construct, abbreviated as pNS1 WNV-IS98 in the present study (Supplementary Figure S1), was verified by DNA sequencing. The empty pcDNA3.1 plasmid vector was used as negative control.

NS1 Production and Purification
One day before transfection, 4 × 10 6 HEK-293T cells were seeded in 100 mm diameter Petri dishes in serum-reduced culture medium (DMEM, 1% FBS). On the following day, cells were transfected with 10 µg plasmid pNS1 WNV-IS98 , using Lipofectamine ® 2000 Transfection Reagent (Invitrogen™, Thermo Fisher Scientific, Carlsbad, CA, USA), following the manufacturer's recommendations. Another batch of cells was transfected with empty pcDNA3.1 vector as negative control. At 6 h post-transfection, the medium was replaced by fresh DMEM supplemented with 1% FBS, and cells were further incubated at 37 • C, 5% CO 2 for 3 days. The culture medium was then collected, filtered through 0.2 µm pores, and stored at 4 • C before NS1 purification. Fresh culture medium was added to the cells for a further 2-day period, then culture medium was again collected, filtered, and pooled with the previous batch stored at 4 • C.
His-tagged native protein sNS1 was purified by affinity chromatography on a nickel resin column. The pooled sNS1-containing culture medium was mixed with an equal volume of equilibration buffer (300 mM NaCl, 10 mM Imidazole, pH 7.4 in PBS). The sNS1-containing mix was then loaded onto a column of Ni-NTA resin (HisPur Ni-NTA resin, Thermo Fisher Scientific™) equilibrated with equilibration buffer. Resin was rinsed with wash buffer (PBS, 25 mM Imidazole, pH 7.4). Elution of the sNS1 protein was carried out with PBS containing 250 mM imidazole, pH 7.4. Eluate was dialyzed overnight against PBS at 4 • C, using SnakeSkin dialysis tubing (7 kDa MWCO, 22 mm; Thermo Fisher Scientific™). In parallel, the same putification protocol was applied to negative control samples (Ctrl pcDNA3.1 ), corresponding to the culture medium of empty pcDNA3.1-transfected HEK-293T cells. The sNS1 concentration was determined in the dialyzed samples, using Bradford protein microassay (protein assay dye reagent, Cat. #5000006, Bio-Rad™, Hercules, CA, USA). Absorbance was measured using the Victor ® microplate reader (Perkin Elmer, Waltham, MA, USA). All the samples were stored at −80 • C until further use.
For kinetic analysis of actin network modifications occurring in sNS1 WNV -treated cells, cell samples were collected at different times of incubation with sNS1 WNV at 37 • C (up to 24 h), fixed, and then stained with rhodamine-phalloidin. Negative controls consisted of mock-treated cells (i.e., cells incubated with samples from the culture medium of HEK-293T cells transfected with empty plasmid pcDNA3.1, subjected to the same chromatographic process as the culture supernatant of pNS1 WNV-IS98 -transfected cells). A total of 140 to 330 individual cells were examined by fluorescence microcopy in at least 5 to 10 separate microscopic fields. Images were acquired with the Leica™ TCS SP5 laser scanning confocal microscope, and cells with absent or disorganized actin microfilament networks were counted using the ImageJ program. The results were expressed as the percentage of total cells examined per field. Statistical analyses were performed using the Mann-Whitney test.

Flow Cytometry Analysis of Cell Attachment and Uptake of NS1 WNV
Cells were seeded with complete media in 24-or 48-well plates to obtain a confluency of 70%-80% the following day. Cells (5 × 10 5 ) were incubated with recombinant sNS1 WNV (10 µg/mL) or control sample (purified supernatant from pcDNA3.1-transfected HEK-293T cells) in PBS containing 10% FBS for 1 h on ice, and rinsed three times with ice-cold PBS. They were detached with PBS containing 0.01% trypsin and 1 mM Na 2 EDTA, then fixed with 4% paraformaldehyde in PBS for 30 min. To determine the level of cell surface-attached sNS1 WNV , cells were incubated with mouse anti-NS1 antibody (1:250 in PBS) for 1 h on ice.
To analyze the cell internalization of sNS1 WNV , cells were first incubated with sNS1 WNV for 1 h on ice to allow cell attachment, rinsed with ice-cold PBS, then transferred to 37 • C for 6 h to allow cellular internalization of sNS1 WNV . Cells were detached with PBS containing 0.01% trypsin and 1 mM Na 2 EDTA, fixed in PBS containing 4% paraformaldehyde for 30 min, and permeabilized with PBS containing 0.1% saponin and 10% FBS for 15 min on ice. NS1 was reacted with mouse anti-NS1 antibody (1:250 in PBS) for 1 h on ice, as above. In both cases, the detection of sNS1 WNV -positive cells was performed using a 1:500 dilution of AlexaFluor 488-conjugated goat anti-mouse IgG antibody for 1 h on ice. Cells were then rinsed 3 times with ice-cold PBS, resuspended in PBS containing 1% paraformaldehyde, and analyzed in a BD FACSCalibur™ flow cytometer.

Transmission Electron Microscopy
Samples of purified sNS1 were applied to a carbon film using the standard mica-carbon flotation technique, and negatively stained with 1% (w/v) uranyl acetate. They were examined under a JEOL (JEM-1200EXII) electron microscope at 100 kV. Images were acquired and analyzed by digital micrograph software (Gatan Inc., Pleasanton, CA, USA).

Transfections
On the day prior to transfection, 5 × 10 4 Vero E6 cells or 8 × 10 4 U-87MG and SH-SY5Y cells were seeded in 8-well plates (Nunc© Lab-Tek, Thermo Fisher Scientific, Carlsbad, CA, USA), and maintained in their respective media. On the next day, cells were transfected with 500 ng of pNS1 WNV-IS98 plasmid or control empty pcDNA3.1, using Lipofectamine ® 2000 Transfection Reagent (Invitrogen™, Thermo Fisher Scientific), according to the manufacturer's recommendations. At 6 h post-transfection, medium was replaced by fresh complete medium. Cells were fixed with 4% paraformaldehyde and processed for immunostaining.

Kinetics of WNV Replication and Virus Titration
WNV titration was performed as previously described [63]. In brief, VeroE6 cells were seeded at 1 × 10 5 cells per well, and SH-SY5Y and U-87MG cells were seeded at 2 × 10 5 cells per well of 48-well plates, in DMEM supplemented with 10% FBS (Gibco). The next day, the cells were incubated with sNS1 WNV (10 µg/mL) or with an aliquot of control sample from empty plasmid pcDNA3.1-transfected cells for 5 h at 37 • C with 5% CO 2 , before or concomitantly to virus addition. WNV-IS98 inoculum at different MOI (in DMEM supplemented with 2% FBS for Vero E6 and U-87MG, and in DMEM/F12 supplemented with 2% FBS for SH-SY5Y) was added to the wells and incubated for 2 h at 37 • C and 5% CO 2 . Cells were rinsed twice with fresh DMEM, then further incubated at 37 • C and 5% CO 2 in DMEM supplemented with 10% FBS for Vero E6 and U-87MG, and in DMEM/F12 supplemented with 2% FBS for SH-SY5Y. Cell supernatants were collected at different times post-infection, and virus progeny titers (expressed as Log 10 TCID 50 /mL) were determined by end point dilution assays in Vero E6 cells, using the Reed and Muench method [1,63].

Expression and Characterization of Intracellular and Secreted Form of NS1 WNV in Mammalian Cells
Recombinant His-tagged NS1 protein was produced in pNS1 WNV-IS98 -transfected HEK-293T cells and isolated from the cell culture medium as the secreted form of NS1 protein (abbreviated sNS1 WNV ). HEK-293T cell lysates and culture supernatants were analyzed by SDS-PAGE and Western blotting, using the standard procedure of protein denaturation by SDS treatment at 95 • C, or, alternatively, heat-denaturation was omitted to preserve the oligomeric structures. In the latter case, samples were only treated with SDS at room temperature, with no chemical cross-linker. The majority of the intracellular NS1 WNV protein occurred as dimers of 110 kDa at 24 h post-transfection (pt; Figure 1a). The secreted sNS1 WNV protein (55 kDa) was detected in low amounts in the culture supernatant as early as 24 h pt, and the maximum extracellular release was observed at 72 h pt ( Figure 1b). No apparent cytotoxicity was observed for the sNS1 WNV -producer cells up to 72 h pt, as shown by MTT assays. After affinity purification, the majority of sNS1 WNV protein in non-heated, SDS-treated samples migrated as dimers, but monomers were also present in significant amounts ( Figure 1c).
After purification by affinity chromatography, the final recovery of sNS1 WNV was about 2 µg protein per 10 6 cells over a 5 day culture period (2.15 ± 0.45 pg/cell; mean ± SD, n = 4). In the absence of an available antibody specific to sNS1 WNV hexamers, their proportion was impossible to assess.
Viruses 2019, 11, x FOR PEER REVIEW 7 of 20 72 h pt ( Figure 1b). No apparent cytotoxicity was observed for the sNS1 WNV -producer cells up to 72 h pt, as shown by MTT assays. After affinity purification, the majority of sNS1 WNV protein in nonheated, SDS-treated samples migrated as dimers, but monomers were also present in significant amounts ( Figure 1c). After purification by affinity chromatography, the final recovery of sNS1 WNV was about 2 μg protein per 10 6 cells over a 5 day culture period (2.15 ± 0.45 pg/cell; mean ± SD, n = 4). In the absence of an available antibody specific to sNS1 WNV hexamers, their proportion was impossible to assess. Electron microscopic (EM) observation of negatively stained sNS1 samples purified by affinity chromatography showed that the secreted recombinant sNS1 WNV protein presented the same structural characteristics as those previously reported for other flaviviral NS1 proteins [10,65]. The top view and side view EM images of our sNS1 WNV protein ( Figure 2) were consistent with a hexameric structure, as previously shown for the sNS1 protein of dengue virus (NS1 DENV ) using 3D reconstruction [4]. The fact that dimers-but no hexamers-were detected by SDS-PAGE analysis of non-heated, SDS-treated samples of sNS1 WNV (Figure 1c), even in gels of low acrylamide concentration, implied that sNS1 WNV dimers interacted via noncovalent links more resistant to SDS dissociation, compared to the other links involved in the whole hexameric structure. on Ni2 + -agarose column, was analyzed by SDS-PAGE and Western blotting after mild denaturation in SDS/β-mercaptoethanol-containing loading buffer at RT. Abbreviations: D = sNS1 WNV dimers; M = sNS1 WNV monomers; Ctrl = negative control samples; mwt = molecular weight (kDa).
Electron microscopic (EM) observation of negatively stained sNS1 samples purified by affinity chromatography showed that the secreted recombinant sNS1 WNV protein presented the same structural characteristics as those previously reported for other flaviviral NS1 proteins [10,65]. The top view and side view EM images of our sNS1 WNV protein ( Figure 2) were consistent with a hexameric structure, as previously shown for the sNS1 protein of dengue virus (NS1 DENV ) using 3D reconstruction [4]. The fact that dimers-but no hexamers-were detected by SDS-PAGE analysis of non-heated, SDS-treated samples of sNS1 WNV (Figure 1c), even in gels of low acrylamide concentration, implied that sNS1 WNV dimers interacted via noncovalent links more resistant to SDS dissociation, compared to the other links involved in the whole hexameric structure.
This was in agreement with a previous study showing that the hexamers of NS1 DENV are unstable in non-ionic detergents, whereas the interfaces between the dimeric subunits are very narrow and contribute to the stabilization of the dimers [4]. Our result confirmed that NS1 dimers constitute the building blocks of the NS1 hexameric scaffold, and that three dimers are held together by weak hydrophobic interaction to form one hexamer [4].   This was in agreement with a previous study showing that the hexamers of NS1 DENV are unstable in non-ionic detergents, whereas the interfaces between the dimeric subunits are very narrow and contribute to the stabilization of the dimers [4]. Our result confirmed that NS1 dimers constitute the building blocks of the NS1 hexameric scaffold, and that three dimers are held together by weak hydrophobic interaction to form one hexamer [4].

Cell Attachment and Uptake of sNS1 WNV
It has been reported that sNS1 DENV can bind to the cell membrane of epithelial and fibroblast cells [15], and that NS1 WNV binds to HEK-293T cells, followed by its internalization within 6 h [16]. Therefore, we compared the cell attachment and cellular uptake of sNS1 WNV in three different cell types: (i) Vero E6, a simian kidney epithelial cell line; (ii) SH-SY5Y, a human neuroblastoma cell line; and (iii) U-87MG, a human astrocytoma cell line. All three cell lines are permissive to WNV infection. The cells were incubated with sNS1 WNV at 10 µg/mL for 1 h and 6 h, respectively, and analyzed by flow cytometry for cell-bound sNS1 WNV (1 h samples, maintained at low temperature) and intracellular NS1 WNV (6 h samples, subjected to temperature increase and cell permeabilization). After a 1 h incubation period, sNS1 WNV proteins were detectable on the surface of non-permeabilized Vero E6 and SH-SY5Y cells, and in significantly lower amounts on the surface of U-87MG cells (Figure 3). After 6 h incubation, sNS1 WNV was efficiently internalized into Vero E6 and SH-SY5Y cells, but not in U87MG cells (Figure 3). These results implied that even though sNS1 WNV could bind to a variety of cells, as previously reported [10,14,15,18], the cellular uptake of sNS1 WNV was cell-type-dependent. previously proposed for NS1 DENV [4], and comparison with enlargements of sNS1 WNV hexamers presented from different incidences-top and side views.

Cell Attachment and Uptake of sNS1 WNV
It has been reported that sNS1 DENV can bind to the cell membrane of epithelial and fibroblast cells [15], and that NS1 WNV binds to HEK-293T cells, followed by its internalization within 6 h [16]. Therefore, we compared the cell attachment and cellular uptake of sNS1 WNV in three different cell types: (i) Vero E6, a simian kidney epithelial cell line; (ii) SH-SY5Y, a human neuroblastoma cell line; and (iii) U-87MG, a human astrocytoma cell line. All three cell lines are permissive to WNV infection. The cells were incubated with sNS1 WNV at 10 μg/mL for 1 h and 6 h, respectively, and analyzed by flow cytometry for cell-bound sNS1 WNV (1 h samples, maintained at low temperature) and intracellular NS1 WNV (6 h samples, subjected to temperature increase and cell permeabilization). After a 1 h incubation period, sNS1 WNV proteins were detectable on the surface of non-permeabilized Vero E6 and SH-SY5Y cells, and in significantly lower amounts on the surface of U-87MG cells (Figure 3). After 6 h incubation, sNS1 WNV was efficiently internalized into Vero E6 and SH-SY5Y cells, but not in U87MG cells (Figure 3). These results implied that even though sNS1 WNV could bind to a variety of cells, as previously reported [10,14,15,18], the cellular uptake of sNS1 WNV was cell-type-dependent.

Cytoskeletal Modifications Associated with sNS1 WNV Internalization
We then analyzed the effect of the internalization of sNS1 WNV on the cytoskeleton of Vero E6, SH-SY5Y, and U-87MG cells. Recombinant sNS1 WNV was added to the culture medium at 10 μg/mL, and the cells were harvested and fixed after 2 h, 6 h, 8 h, and 24 h. The presence of intracellular NS1 WNV and the morphological status of the actin and tubulin networks were analyzed by confocal

Cytoskeletal Modifications Associated with sNS1 WNV Internalization
We then analyzed the effect of the internalization of sNS1 WNV on the cytoskeleton of Vero E6, SH-SY5Y, and U-87MG cells. Recombinant sNS1 WNV was added to the culture medium at 10 µg/mL, and the cells were harvested and fixed after 2 h, 6 h, 8 h, and 24 h. The presence of intracellular NS1 WNV and the morphological status of the actin and tubulin networks were analyzed by confocal fluorescence microscopy. The NS1 WNV signal was detected in sNS1 WNV -treated Vero E6 cells as early as 2 h post-incubation, reaching a maximum at 6 h ( Figure 4A). NS1 WNV was still visible at 8 h, but no longer detectable at 24 h ( Figure 4A).
In parallel, a significant remodeling of the cellular actin network was observed between 6 h and 8 h ( Figure 4B). These alterations were transient and followed the evolution of the NS1 WNV content; after a maximum disorganization at 6-8 h, the actin network was completely restored at 24 h ( Figure  4B and Figure 5). This suggested that the extent of actin depolymerization correlated with the level of intracellular NS1 WNV . A direct negative effect of NS1 WNV on actin biosynthesis was unlikely, as shown by the absence of significant change in actin levels over time in NS1 WNV -expressing HEK-293T cells (refer to Figure 1a).
Our observations, therefore, implied that the viral protein sNS1 WNV alone had a transient negative effect on the actin network of epithelial cells. The effect on actin was apparently specific to epithelial cells, as no remodeling of the actin filament network was visible in sNS1 WNV -treated neuronal cell SH-SY5Y or glial cell U-87MG (not shown). No detectable alteration of the tubulin network was observed in any of the sNS1 WNV -treated cell types, namely, Vero E6, SH-SY5Y, or U-87MG (Supplementary Figure S2).  . Actin microfilaments (red) were visualized using rhodamine-phalloidin, nuclei (blue) were visualized using dihydrochloride (DAPI) reagent, and sNS1 WNV protein (green) was immunolabeled using anti-NS1 WNV rabbit antibody. Note that images in panels (A-C) show the same areas at a given time point.
In parallel, a significant remodeling of the cellular actin network was observed between 6 h and 8 h ( Figure 4B). These alterations were transient and followed the evolution of the NS1 WNV content; after a maximum disorganization at 6-8 h, the actin network was completely restored at 24 h ( Figures 4B and 5). This suggested that the extent of actin depolymerization correlated with the level of intracellular NS1 WNV . A direct negative effect of NS1 WNV on actin biosynthesis was unlikely, as shown by the absence of significant change in actin levels over time in NS1 WNV -expressing HEK-293T cells (refer to Figure 1a).
Viruses 2019, 11, x FOR PEER REVIEW 11 of 20 reagent, and sNS1 WNV protein (green) was immunolabeled using anti-NS1 WNV rabbit antibody. Note that images in panels (A-C) show the same areas at a given time point.

Presence of NS1 WNV Protein in Actin-Containing, TNT-Like Nanofilaments Protruding from NS1 WNV -Expressing and WNV-Infected Cells
Since the cell attachment and cellular uptake of sNS1 WNV differed for the three cell types, the possible effects of endogenous NS1 WNV on the cytoskeletal network of Vero E6, SH-SY5Y, and U-87MG cells were analyzed after transfection with pNS1 WNV-IS98 plasmid. The actin network disorganization was less obvious in NS1 WNV -expressing Vero E6 cells compared to cells treated with Our observations, therefore, implied that the viral protein sNS1 WNV alone had a transient negative effect on the actin network of epithelial cells. The effect on actin was apparently specific to epithelial cells, as no remodeling of the actin filament network was visible in sNS1 WNV -treated neuronal cell SH-SY5Y or glial cell U-87MG (not shown). No detectable alteration of the tubulin network was observed in any of the sNS1 WNV -treated cell types, namely, Vero E6, SH-SY5Y, or U-87MG (Supplementary Figure S2).

Presence of NS1 WNV Protein in Actin-Containing, TNT-Like Nanofilaments Protruding from NS1 WNV -Expressing and WNV-Infected Cells
Since the cell attachment and cellular uptake of sNS1 WNV differed for the three cell types, the possible effects of endogenous NS1 WNV on the cytoskeletal network of Vero E6, SH-SY5Y, and U-87MG cells were analyzed after transfection with pNS1 WNV-IS98 plasmid. The actin network disorganization was less obvious in NS1 WNV -expressing Vero E6 cells compared to cells treated with exogenous sNS1 WNV (Figure 6; compare to Figure 4). Only Vero E6 cells overexpressing NS1 WNV protein showed less distinct stress fibers compared to mock-transfected cells (Figure 6a-d).
Interestingly, NS1 WNV -expressing Vero E6 cells produced thin and long projections joining one cell to another and containing both actin and NS1 WNV proteins (Figure 6b,f-h). These filamentous projections resembled the thin, actin-rich membrane channels referred to as tunneling nanotubes (TNTs), involved in intercellular communication and exchange [40,41,43,44]. In U-87MG cells, TNT-like filaments were observed in control, mock-transfected cells (Figure 7a), but they seemed to occur with a higher frequency in NS1 WNV -expressing cells (Figure 7b). NS1 WNV -expressing SH-SY5Y cells also showed filamentous protrusions containing actin and NS1 WNV , but these structures appeared to be thicker and shorter compared to the TNT-like nanofilaments observed in Vero E6 or U-87MG cells (Supplementary Figure S3).
Interestingly, NS1 WNV -expressing Vero E6 cells produced thin and long projections joining one cell to another and containing both actin and NS1 WNV proteins (Figure 6b,f-h). These filamentous projections resembled the thin, actin-rich membrane channels referred to as tunneling nanotubes (TNTs), involved in intercellular communication and exchange [40,41,43,44]. In U-87MG cells, TNTlike filaments were observed in control, mock-transfected cells (Figure 7a), but they seemed to occur with a higher frequency in NS1 WNV -expressing cells (Figure 7b). NS1 WNV -expressing SH-SY5Y cells also showed filamentous protrusions containing actin and NS1 WNV , but these structures appeared to be thicker and shorter compared to the TNT-like nanofilaments observed in Vero E6 or U-87MG cells (Supplementary Figure S3).  We next determined whether the TNT-like nanofilaments also occurred during WNV infection. Vero E6 cells were infected with WNV at relatively low multiplicity (MOI 0.5), and the cells were fixed at 18 h post-infection for analysis by confocal fluorescence microscopy. Actin-containing filamentous projections emanating from WNV-infected cells were visible (Figure 8), and, interestingly, they contained both the viral envelope glycoprotein E WNV (Figure 8a) and the nonstructural protein NS1 WNV (Figure 8b). In some cases, we observed that TNT-like structures made connections to both WNV-infected and non-infected cells, and certain images suggested a possible transfer of NS1 WNV to non-infected cells (Figure 8c,d). These observations were consistent with the results of an earlier study on WNV-infected Vero cells examined at the late stages of infection, showing the proliferation of filopodia associated with actin filament thickening, and the presence of mature WNV virions in some of these filopodia [66]. We next determined whether the TNT-like nanofilaments also occurred during WNV infection. Vero E6 cells were infected with WNV at relatively low multiplicity (MOI 0.5), and the cells were fixed at 18 h post-infection for analysis by confocal fluorescence microscopy. Actin-containing filamentous projections emanating from WNV-infected cells were visible (Figure 8), and, interestingly, they contained both the viral envelope glycoprotein E WNV (Figure 8a) and the non-structural protein NS1 WNV (Figure 8b). In some cases, we observed that TNT-like structures made connections to both WNV-infected and non-infected cells, and certain images suggested a possible transfer of NS1 WNV to non-infected cells (Figure 8c,d). These observations were consistent with the results of an earlier study on WNV-infected Vero cells examined at the late stages of infection, showing the proliferation of filopodia associated with actin filament thickening, and the presence of mature WNV virions in some of these filopodia [66]. Viruses 2019, 11, x FOR PEER REVIEW 14 of 20

NS1 WNV and Virus Infectiveness
Considering that NS1 WNV was secreted in the extracellular milieu of WNV-infected cells and was also susceptible to spread from cell to cell via actin-containing cellular nanofilaments, we analyzed the possible impact of NS1 WNV on the WNV life cycle. Vero E6, SH-SY5Y, and U-87MG cells were incubated with sNS1 WNV added to the culture medium 5 h prior to, or concomitantly with, the virus inoculum, used at MOI 0.01, 0.1, and 1, respectively. WNV-infected cells were then monitored for late cytopathic effects, and virus replication and progeny yields were determined by end point dilution assays in Vero E6 cells [1,63]. Interestingly, sNS1 WNV did not significantly affect any of these parameters in all three cell types ( Supplementary Figures S4 and S5). This result was somehow expected for U-87MG cells, considering that these cells did not internalize sNS1 WNV in detectable amounts, but not for Vero E6 and SH-SY5Y cells, which showed an efficient sNS1 WNV uptake (refer to Figure 3). Although these negative data were obtained from in vitro experiments, they might have important implications regarding the role of NS1 in WNV pathogenicity.

NS1 WNV and Virus Infectiveness
Considering that NS1 WNV was secreted in the extracellular milieu of WNV-infected cells and was also susceptible to spread from cell to cell via actin-containing cellular nanofilaments, we analyzed the possible impact of NS1 WNV on the WNV life cycle. Vero E6, SH-SY5Y, and U-87MG cells were incubated with sNS1 WNV added to the culture medium 5 h prior to, or concomitantly with, the virus inoculum, used at MOI 0.01, 0.1, and 1, respectively. WNV-infected cells were then monitored for late cytopathic effects, and virus replication and progeny yields were determined by end point dilution assays in Vero E6 cells [1,63]. Interestingly, sNS1 WNV did not significantly affect any of these parameters in all three cell types ( Supplementary Figures S4 and S5). This result was somehow expected for U-87MG cells, considering that these cells did not internalize sNS1 WNV in detectable amounts, but not for Vero E6 and SH-SY5Y cells, which showed an efficient sNS1 WNV uptake (refer to Figure 3). Although these negative data were obtained from in vitro experiments, they might have important implications regarding the role of NS1 in WNV pathogenicity.

Discussion
The NS1 protein used in this study was derived from WNV-IS-98-ST1, a neurovirulent strain of lineage 1 that has been well-characterized and is highly pathogenic in animal models [62,67]. Two forms of recombinant NS1 WNV protein were tested on cells: the exogenous, secreted form (sNS1 WNV ) added to the cell culture medium; and endogenous NS1 WNV , the recombinant protein expressed in pNS1 WNV-IS98 -transfected cells. The possible cytological effects of NS1 WNV were analyzed using three different types of cells: Vero E6, SH-SY5Y, and U-87MG. The rationale for this choice was the following: (i) simian kidney epithelial cells Vero E6 are the most widely used host cell line for WNV propagation and titration; (ii) similar to Vero E6 cells, SH-SY5Y cells are permissive to WNV infection [68]; (iii) human neuroblastoma cells SH-SY5Y and human astrocytoma cells U-87MG represented valuable cellular models to study the sNS1 WNV -mediated neurovirulence.
Cell attachment and uptake have been well-documented for sNS1 DENV compared to the other flaviviral NS1 proteins. It is generally accepted that sNS1 DENV binds to the cell plasma membrane via interactions with heparan sulfate or chondroitin sulfate E, and that TLRs can serve as alternative receptors [10,[15][16][17][18]. After cell attachment, sNS1 DENV follows the endocytic pathway and accumulates in the late endosomal compartment [14]. For the NS1 WNV protein, we observed that the exogenous sNS1 WNV was efficiently internalized into Vero E6 and SH-SY5Y cells, reaching a maximum at 6 h incubation at 37 • C; it was not efficiently internalized into U-87MG. The cellular uptake of sNS1 WNV , therefore, varied with the cell type.
The addition of sNS1 WNV to the cell culture medium was associated with a significant remodeling of the actin filament network, and the pattern suggested the depolymerization of the F-actin. This phenomenon was transient as the actin network progressively reformed in parallel with the decrease of the intracellular NS1 WNV signal. The alteration of the actin network was only observed in Vero E6 cells and not in the other two cell lines, implying that there was no direct correlation between the cell internalization of sNS1 WNV and alteration of the actin network. The absence of actin effect in sNS1 WNV -treated SH-SY5Y cells, although sNS1 WNV was internalized with the same efficiency as in Vero E6 cells (refer to Figure 3), suggested that the sNS1 WNV -associated actin remodeling was cell-type-dependent.
The cell type specificity of actin remodeling could be related to variable expression of surface molecules on the different cell types. Thus, tissue-specific variations in the cell binding of flaviviral NS1 proteins, as well as in the profile of response to inflammatory cytokines in the context of flaviviral infection, have been reported [69,70]. In terms of intracellular mechanism, a significant activation of RhoA by phosphorylation has been found in both DENV-infected and NS1 DENV -treated human dermal microvascular endothelial cells, suggesting that the increased endothelial permeability observed during DENV infection or NS1 DENV treatment would be due, at least in part, to RhoA phosphorylation [71]. A quantitative proteomic analysis of brain tissue in WNV-infected mice has shown that several proteins related to the actin cytoskeleton and Rho GTPase signaling pathway were significantly modified during the course of neuroinvasive WNV infection [72]. The role of molecules of the Rho GTPase signaling pathway in cytoskeleton rearrangement, and the consequences on entry, replication, and release of many viruses, including WNV, have been well-documented [28,29]. It is, therefore, plausible (i) that the binding of sNS1 WNV to a G-protein-coupled receptor present on the surface of certain cells could activate the cascade of the Rho GTPase signaling pathway; or alternatively (ii) that the intracellular localization of cell-internalized sNS1 WNV would vary with the cell type, and, in some particular cases, would allow a direct interaction of sNS1 WNV with molecules of the Rho GTPase pathway. In both scenarios, the regulation of actin polymerization would be significantly affected.
The influence of NS1 WNV on the cytoskeletal network was then analyzed using endogenous NS1 WNV . The pattern shown by NS1 WNV -expressing Vero E6 cells significantly differed from that observed in cells treated with exogenous sNS1 WNV , and only discrete changes in the actin network were observed (compare Figures 4 and 6). This result supported the first hypothesis formulated above, and suggested that the NS1 WNV -mediated modulation of the actin network in epithelial cells would require the attachment of sNS1 WNV to the cell surface and its transit via the endocytic pathway.
Examination of NS1 WNV -expressing Vero E6 and U-87MG cells showed long and thin projections, forming intercellular bridges harboring both actin and NS1 WNV proteins. These nanofilaments presented morphological similarities with TNTs, the actin-rich tunneling nanotubes that form intercellular channels [40,41,43,44]. In contrast, NS1 WNV -expressing SH-SY5Y cells showed thicker and shorter protrusions, but these filamentous structures also contained actin and NS1 WNV . The structural characteristics of the SH-SY5Y protrusions were consistent with the recently elucidated structure of neuronal TNTs, which are composed of bundles of individual TNTs held together by N-cadherin fibers [45]. TNTs have been observed in higher numbers in WNV-infected Vero E6 cells compared to mock-infected cells [66]. Although our study lacked quantitative data, the TNT-like nanofilaments seemed to occur with a higher frequency in NS1 WNV -expressing cells compared to control cells. Specifically designed methodologies would be necessary to quantify these NS1 WNV -containing, TNT-like structures and determine a possible correlation between their density per cell and the level of NS1 WNV expression.
The TNTs that reached out to connect to neighboring cells, as observed in NS1 WNV -expressing and WNV-infected cells, represented a well-designed launching site for the cell-to-cell transfer of NS1 WNV . It was, therefore, tempting to hypothesize that the extracellular form of NS1 WNV (capable of cell entry), the intra-TNT form of NS1 WNV (potentially apt to cell-to-cell trafficking), or both, could prime uninfected cells present in this region, and increase their permissiveness to newly released WNV virions, as described for HIV-1 and Nef-induced filopodia [73,74]. We tested this hypothesis and found that exogenous sNS1 WNV (added to the culture medium of WNV-infected cells prior to, or simultaneously with, the virus) did not significantly change the late cytopathic effects, the replication kinetics, or the virus progeny yields in any of the three cell types. The absence of a direct effect of exogenous sNS1 WNV on the WNV infectiveness in vitro seemed to exclude its role as a virus facilitating factor or enhancer. However, the eventuality of such a role is still conceivable for endogenous NS1 WNV transiting through TNTs.
Another hypothesis might explain the possible advantage for the virus of the frequent occurrence of NS1 WNV -associated TNT-like nanofilaments, and even their proliferation, as previously observed [66]: TNTs could facilitate the passage of WNV virions from one cell to another, so as to avoid their exposure to the extracellular environment, a mechanism of cell-to-cell transmission that would be similar to that of many other viruses [46][47][48][49][50][51][52][53][54][55][56][57][58]. Our observations that NS1 WNV and the envelope glycoprotein E WNV coexisted in the filamentous bridges between WNV-infected and non-infected cells supported this hypothesis. Alternatively, TNTs could be used for the extracellular release of the WNV progeny, a mechanism which has previously been envisaged. Analysis of WNV-infected Vero cells analyses by transmission EM and atomic force microscopy suggested the exit of virus particles via the filopodia [66].
Regardless of the exact functions of the actin-and NS1 WNV -containing TNT-like nanofilaments observed in NS1 WNV -expressing and WNV-infected cells, our data clearly showed that recombinant NS1 WNV protein alone, without any other WNV protein, was capable of modulating-either directly or indirectly via host-cell intermediate factor(s)-the actin network of epithelial cells. These results should be considered in future studies performed in the context of natural infections with WNV, as they might provide further clues to elucidate the molecular mechanisms of the NS1-mediated disruption of endothelial cell monolayers, and the subsequent vascular leakage, as observed with sNS1 DENV [17,24,75].