Human kidney epithelial HEK 293T (ATCC, CRL-11268), human lung epithelial A549 (ATCC, CCL-185), and African green monkey kidney epithelial Vero E6 (ATCC, CRL-1586) cells were maintained in high-glutamine and high-glucose DMEM (Thermo fisher) supplemented with 10% fetal bovine serum (FBS; Thermo fisher), 100U/ml penicillin and 100μg/ml streptomycin (Thermo fisher). Aedes albopictus C6/36 (ATCC, CRL-1660) cells were maintained in Leibovitz’s L-15 medium (Thermo fisher) supplemented with 10% FBS. Transformed human bone marrow endothelial cells TrHBMEC (Schweitzer et al., 1997), human cerebral microvascular endothelial cells (hCMEC/D3; Merck, SCC066), and primary human umbilical vein endothelial cells HUVEC (ATCC, CRL-1730) cells were maintained in complete EndoGRO Basal Medium (Merck) supplemented with 5% FBS, 0.2% EndoGRO-LS supplement, 5ng/ml rhEGF, 10mM l-glutamine, 1μg/ml hydrocortisone hemisuccinate, 0.75U/ml heparan sulphate, and 50μg/ml ascorbic acid (Merck). Alternatively, hCMEC/D3 cells have been maintained in EBM-2 basal medium (Lonza) supplemented with 5% FBS, 1.5μM hydrocortisone (Merck), 5μg/ml ascorbic acid (Merck), 1% chemically defined lipid concentrate (Thermo fisher), 1ng/ml human bFGF (Merck), 10μM HEPES (Thermo fisher), and antibiotics. This culture condition was used when compared to EndoGRO media. All cells were kept at 37°C with 5% CO₂ except for C6/36 which were kept at 28°C with no CO₂.

Two strains of ZIKV were used in this study: the AF/1991/HD78788 strain (GenBank: KF383039) obtained in Senegal in 1991, which belongs to the African lineage, and the Brazil/2016/INMI1 strain (GenBank: KU991811) isolated during the 2016 outbreak in Brazil, which belongs to the Asian lineage. Japanese encephalitis virus JEV/G3/RP-9 (Liang et al., 2009) and yellow-fever vaccine strain YFV/17D (GenBank: MG051217) were used as controls, as they are closely related to ZIKV. Virus stocks were produced in C6/36 cells. All infections were performed in 2% FBS media in BSL-3 facilities.

Gene knockdown was performed using lentiviral vectors carrying shRNA. To produce these lentiviral vectors, HEK 293T cells were co-transfected with psPAX2 (encoding HIV Gag/Pol, 4.68μg; Addgene: 12260), pMD2.G [vesicular stomatitis virus G protein (VSV-G), 2.52μg; Addgene: 12259], and pLKO.1 encoding the shRNA (TRC-Hs1.0 library, 9μg, GE Dharmacon). Transfection was performed using LipoD-293 transfection reagent (Sinagen) according to the manufacturer’s recommendations. After 72h, supernatants were collected, centrifuged, and stored at −80°C.

Cells were exposed for 2h to the lentivirus-containing medium, and gene expression was allowed for 48h then transduced cells were selected with 1μg/ml Puromycin (Merck) for 2weeks. The obtained culture is thus a mix population of transduced cells.

Biotinylated single strand cDNA targets were prepared, starting from 250ng of total RNA, using the Ambion WT Expression Kit (Cat # 4411974) and the Affymetrix GeneChip® WT Terminal Labeling Kit (Cat # 900671) according to Affymetrix recommendations. Following fragmentation and end-labeling, 3μg of cDNAs were hybridized for 16h at 45°C on GeneChip® Human Gene 2.0 ST arrays (Affymetrix) interrogating over 400,000 RefSeq transcripts and ~11,000 LncRNAs. The chips were washed and stained in the GeneChip® Fluidics Station 450 (Affymetrix) and scanned with the GeneChip® Scanner 3000 7G (Affymetrix) at a resolution of 0.7μm. Raw data (CEL Intensity files) were extracted from the scanned images using the Affymetrix GeneChip® Command Console (AGCC) version 4.1.2. CEL files were further processed with Affymetrix Expression Console software version 1.4.1 to calculate probe set signal intensities using Robust Multi-array Average (RMA) algorithms with default settings. Data have been loaded into Array Express, accession number E-MTAB-10813.

Cellular RNA was extracted using the RNeasy Plus Mini Kit (Qiagen), quantified (Nanodrop ND-1000) and 500ng was reverse-transcribed using the SuperScript II Reverse Transcriptase (Thermo fisher) and random hexamer primers (Eurobio) with the following cycle: 2min at 95°C; 90min at 42°C; and 15min at 70°C (GeneAmp PCR System 9700, Applied Biosystems). SYBR Green RT-qPCR was performed on 25ng of complementary DNA with iTaq Universal SYBR Green Supermix (Biorad) using primers listed in Supplementary Table 1 with the following protocol: 15min at 95°C followed by 40cycles consisting of 15s at 95°C, 20s at 60°C, and 30s at 72°C (Mastercycler ep Realplex Thermal Cycler, Eppendorf). The melting curve was also obtained to control the amplified product. Relative expression was analyzed using the −∆∆C_T method.

Functional enrichment analysis of the differentially expressed genes was performed using FunRich 3.1.3 (Pathan et al., 2015).

Cells were lysed using RIPA buffer (Merck) containing protease inhibitor cocktail (Merck) for 15min on ice. Cellular debris were not pelleted since claudins often accumulate in the insoluble fraction. Whole cell lysates were heat-denatured 70°C for 10min in LDS Sample Buffer and 500mM DTT (Novex). Proteins were resolved on NuPAGE 4–12% Bis Tris gel (Novex) in MOPS buffer and were transferred to nitrocellulose membranes (Trans Blot Turbo, Biorad). Blocking followed by immunoblotting was performed on an iBind Flex Western Device (Thermo fisher) using antibodies listed in Supplementary Table 2 according to the manufacturer’s recommendations.

Supernatants were sequentially diluted 10-fold and inoculated onto Vero E6 monolayers in 24-well plates. After 1h incubation, 2% carboxymethyl cellulose (CMC; Merck) overlay containing 2% FBS culture medium was added to each well. For foci-formation assays, plates were then incubated at 37°C for 72h. Following the incubation, CMC overlay was removed and a PBS – paraformaldehyde 4% (Electron Microscopy Sciences) solution was added into each well and incubated at room temperature for 15min. Fixation solution was removed and plates were washed three times with PBS. Cells were then permeabilized with PBS – Triton 0.2% for 3min, washed three times, and incubated in blocking buffer (PBS – Tween 20 0.1% – FBS 1%) for 30min followed by 2h incubation with pan-Flavivirus antibody 4G2 purified from the ATCC hybridoma. Plates were washed three times with PBS followed by an hour-long incubation with a secondary antibody conjugated to horseradish peroxidase (Biorad). Detection was achieved using the VIP Peroxidase Assay (Vector) according to the manufacturer’s recommendations. Titer is expressed as foci-forming units (FFU).

Alternatively, for plaque-formation assays, plates were incubated at 37°C for 5days after adding the CMC overlay. They were then washed and fixed as described earlier. Plates were stained with a 1% crystal violet solution (Merck). Titer is expressed as plaque-forming units (PFU).

Infected cells were washed with PBS, and cellular RNA was extracted using the Nucleospin 96 RNA kit (Macherey Nagel). One-step RT-qPCR was performed using the Power SYBR Green RNA-to-C_T kit (Thermo fisher) with the following protocol: 30min at 48°C, 10min at 95°C followed by 40 cycles consisting of 15s at 95°C, and 1min at 60°C (QuantStudio 6 Flex Real-Time PCR System, Applied Biosystems). The melting curve was also obtained. Genome equivalent (GE) concentration was assessed with the standard curve method using known concentrations of synthetic RNA fragments corresponding to the ZIKV NS5 coding region and was normalized to GAPDH level.

Cells were seeded in 2% FBS medium 24h prior to infection. Pre-chilled cells were incubated with ZIKV (multiplicity of infection – MOI 1 and 10) for 1h at 4°C to block the internalization of the viruses. The cells were then washed three times with cold PBS to remove unbound particles. Cell surface adsorbed ZIKV particles was analyzed by harvesting RNA at this step. ZIKV internalization was induced by adding warm medium and keeping the cells at 37°C for 2h. The cells were then treated with trypsin (Thermo fisher) to remove un-internalized particles, cells were pelleted, and RNA was harvested.

Cells were fixed with 4% paraformaldehyde (PFA; Sigma-Aldrich) for 30min at room temperature, permeabilized with 0.5% Triton X-100 in PBS, and then blocked with PBS containing 0.05% Tween and 5% BSA before incubation with the anti-NS1 17A12 antibody (Schul et al., 2007). Antibody labeling was revealed with Alexa Fluor 488-conjugated antibody. Coverslips were mounted on slides using ProLong Gold Antifade reagent with NucBlue solution (Invitrogen). Images were acquired with a Zeiss LSM 700 inverted confocal microscope.

Specific statistical tests used to analyze experimental datasets are described in the respective figure legends. Graphical representations and statistical analyses were performed using GraphPad Prism version 8.2 for MacOS (GraphPad Software). Nonparametric tests were used, as we could not ascertain that the hypothesis for parametric tests were fulfilled. When comparing cells in the different media (Figure 1 and Supplemental Figures 1, 2), two-way ANOVA (followed by Sidak’s post-hoc tests) or the Mann-Whitney and Kruskal-Wallis tests were used (data are unpaired). When comparing the knockdown, we considered paired tests as a single batch of cells was transduced with the different shRNAs, and a new batch was used for each replicate. Therefore, two-way ANOVA (followed by Sidak’s post-hoc tests) or Friedman tests (followed by Dunn’s post-hoc tests) were used. Values of p indicated in figure legends corresponds to ANOVA result.

The human cerebral microvascular endothelial cell line hCMEC/D3 is a widely used model for in vitro human BBB studies (Weksler et al., 2005). Since ZIKV neuroinvasion relies on a productive infection of BBB endothelial cells (Liu et al., 2016; Mladinich et al., 2017; Papa et al., 2017; Richard et al., 2017; Cle et al., 2020), we examined the susceptibility to ZIKV infection of hCMEC/D3 cells.

We confirmed using RT-qPCR analysis that hCMEC/D3 cells can be productively infected with an African ZIKV strain (Figures 1A,B). The cells released infectious particles (Figures 1C,D), indicating that a complete viral replication cycle. Intriguingly, we found a difference in infection efficiency that was dependent on the cell culture medium, both in terms of viral RNA yield and production of infectious particles (Figure 1, EndoGRO vs. EBM-2). Differences were more marked at a lower MOI (Figures 1A,C vs. Figures 1B,D). Similar differences in viral susceptibility were observed with a Brazilian ZIKV strain (belonging to the Asian lineage), at 72h post-infection (Figures 1E,F).

Our first hypothesis was that factors present in the medium could interfere with viral infectivity. To test this, we suspended the viral stock in one or the other medium (or in DMEM as a control) and the inoculum was let to adsorb for 2h on hCMEC/D3 (previously cultured in EndoGRO). Cells were then washed and cultured in EndoGRO for 48h. RT-qPCR analysis performed at 48h post-infection revealed that the levels of intracellular viral RNA were similar in the three conditions (Supplementary Figure 1). This suggests that the culture media does not alter the cell-virus interaction per se. The differential susceptibility observed in the two culture conditions likely depend on differential gene expression.

We cultured hCMEC/D3 cells in EndoGRO or in EBM-2 prior to infection. Infection and subsequent culture were then performed in DMEM. We found that viral replication efficacy was dependent on the medium in which the cells were initially cultured (Supplementary Figure 2). This suggests that the cells, while being cultured in the different media, have acquired intrinsic differences to ZIKV infection susceptibility.

First, we examined whether hCMEC/D3 cells cultured in the two media differentially expressed AXL and other members of the TAM family (i.e., TYRO3 and MER), which have been found to regulate endothelial susceptibility to ZIKV (Liu et al., 2016; Richard et al., 2017). Western blot analysis revealed that AXL expression remained unchanged in either medium (Figure 2A). Similarly, we did not see any significant difference in the expression of AXL at the mRNA level (Figure 2B). While TYRO3 and MER proteins are not detectable by western blot in hCMEC/D3 cells, we found that the mRNA levels of TYRO3 and MER were comparable in both media (Figure 2B).

We compared the transcriptome of hCMEC/D3 cells cultured in EBM-2 and in EndoGRO to identify cellular genes that could account for differences in susceptibility. Microarray analysis showed that the expression of most genes was left unchanged in the two conditions, at the exception of 101 genes, which were upregulated in cells cultured in EBM-2 as compared to cells cultured in EndoGRO, and of 38 genes that were upregulated in cells cultured in EndoGRO as compared to EBM-2 cultured cells (Figure 2C and Supplementary Tables 3 and 4).

Among the proteins expressed at higher level in EndoGRO medium than in EBM-2, we identified Claudin-7 (CLDN7). This gene caught our attention as claudins have been identified as modulators of cellular sensitivity to viruses from the Flaviviridae family (Colpitts and Baumert, 2017). The endogenous protein expression levels of CLDN7 could not be detected at the protein level in endothelial cells, as previously reported (Vu et al., 2009; Supplementary Figure 3). It was, however, detected in A549 and HEK 293T epithelial cells. Of note, CLDN7 could only be detected when analyzing whole cell lysates (i.e., non-centrifuged, comprising the insoluble fraction). RT-qPCR analysis confirmed that CLDN7 mRNA abundance increased in cells grown in EndoGRO when compared to cells cultured in EBM-2 (Figure 2D), consistent with the microarray analysis. In contrast, the mRNA levels of CLDN5 and CLDN3 – the other predominant claudins expressed in hCMEC/D3 cells (Weksler et al., 2005; Vu et al., 2009; Schrade et al., 2012; Urich et al., 2012) – were unchanged. Of note, at the protein level, CLDN5 expression was comparable in the different conditions (Figure 2A).

We hypothesized that the ability of ZIKV to replicate more efficiently in cells grown in the EndoGRO medium, as compared to cells grown in EBM-2, could be (at least partially) due to differences in CLDN7 expression levels.

To test the impact of CLDN7 knockdown (KD) on hCMEC/D3 susceptibility to African ZIKV infection, hCMEC/D3 cells were transduced with lentiviral vectors expressing shRNA sequences targeting CLDN7 or, as controls, CLDN5 transcripts or non-targeting shRNAs (CTRL). Transduced cells were selected over 2weeks by puromycin treatment. The obtained cultures were a mix of populations of transduced cells.

As CLDN7 cannot be detected by western blot in hCMEC/D3 cells (Supplementary Figure 3), Knockdown was validated by RT-qPCR on cellular mRNA (Supplementary Figure 4). As half-life of claudins is short (6–12h; Van Itallie et al., 2019), one can assume that a significant reduction in claudin mRNA leads to reduced protein levels. Of note, CLDN5 seems consistently overexpressed in the CLDN7-KD cells (Supplementary Figure 4).

Transduction per se had no significant impact on hCMEC/D3 susceptibility (i.e., production of viral RNA and infectious particles) to ZIKV since hCMEC/D3 cells transduced with a lentiviral vector expressing a control shRNA (CTRL) were as susceptible as non-transduced (WT) hCMEC/D3 cells (Figures 3A,B). KD of CLDN5 did not alter viral replication and production (Figures 3A,B). In contrast, viral replication and production were significantly reduced in CLDN7-KD cells, as compared to control cells (Figures 3A,B). CLDN7 expression was also required for optimal replication of the Brazilian strain of ZIKV (Figures 3C,D), as shown by RT-qPCR analysis.

Together these data demonstrate that CLDN7 expression favors ZIKV infection in hCMEC/D3 cells. Importantly, Japanese encephalitis and yellow fever viruses, both closely related to ZIKV, were replicating as efficiently in control hCMEC/D3 cells than in CLDN7-KD cells (Figures 3E,F). This demonstrates that CLDN7 is specifically involved in hCMEC/D3 susceptibility to ZIKV infection.

We evidenced the contribution of CLDN7 in ZIKV replication in hCMEC/D3 cells. We examined whether CLDN7 was also favoring ZIKV infection in two other endothelial cell lines: bone marrow endothelial cells (TrHBMEC) and umbilical vein endothelial cells (HUVEC). We also tested the effect of silencing CLDN7 expression in two epithelial cell lines: alveolar basal cells (A549) and embryonic kidney cells (HEK 293T).

Knockdown efficiency was validated by RT-qPCR in these four cell lines (Supplementary Figures 5A–D). We considered only cells with a −∆∆C_T (compared to WT cells) lower than −2; for HUVEC and HEK 293T, only one shRNA-derived cell line (out of five tested) was considered for further analysis (Supplementary Figures 5A–D). For epithelial cells, downregulation was validated at the protein level by western blot analysis (Supplementary Figures 5E,F).

We found that downregulation of CLDN7 significantly reduced infectious particle production in both endothelial cell types, TrHBMEC (Figure 4A) and HUVEC (Figure 4B). In contrast, no significant difference in viral yield was observed upon CLDN7-KD in A549 (Figure 4C) and HEK 293T epithelial cells (Figure 4D). Likewise, viral RNA production was significantly reduced in CLDN7-KD endothelial cells as compared to control cells (Figures 4E,F) but not in epithelial cells (Figures 4G,H).

Our data demonstrate that CLDN7 favors ZIKV replication in endothelial cells but not in epithelial cells.

We next assessed which step of the viral cycle required the expression of CLDN7. To determine whether CLDN7 could be involved in viral adsorption to the cell membrane, hCMEC/D3 cells were infected with ZIKV for 1h at 4°C, which allows viral binding but prevents virus internalization. Unbound viruses were then removed through extensive washing. ZIKV RNA levels, corresponding to virions that remained attached to cell membranes, were quantified by RT-qPCR analysis. As expected, more viral particles were bound to the cells when incubated at a MOI of 10 than at a MOI of 1 (Figures 5A,B). However, no difference in viral RNA yield was detected between WT and CLDN7-KD cells, suggesting that CLDN7 is dispensable for ZIKV binding to cell membranes.

We next tested whether CLDN7 was involved in virus internalization. Following a 1h incubation at 4°C, cells were cultured at 37°C for 2h. Cells were treated with trypsin to remove non-internalized viral particles. Quantification of intracellular ZIKV RNA present in the cells, representing internalized particles, showed no significant difference between the control cells and CLDN7-KD hCMEC/D3 cells at MOI 1 and 10 (Figures 5C,D), suggesting that CLDN7 does not contribute to ZIKV internalization.

The proportion of infected hCMEC/D3 cells was determined using staining against the viral non-structural protein NS1 at 16h post-infection, and a time that corresponds to a unique replication cycle (Figures 5E, 6). At a MOI of 1, around 11% hCMEC/D3 cells were expressing the NS1 protein. In contrast, in CLDN7-KD cells, only around 4% of cells were positive for NS1, independently of the shRNA that was used. These data confirm our previous RT-qPCR and titration analyses (Figure 3) and thus further validate the role of CLDN7 as a proviral factor in endothelial cells. Of note, the distribution of NS1 was comparable in infected WT and CLDN7-KD cells (Figure 6).

Together these results demonstrate that CLDN7 downregulation does not affect viral attachment or internalization but reduces the number of cells expressing the viral protein NS1, suggesting that CLDN7 facilitates viral replication at a post-entry stage but prior to viral protein translation.