The pan-caspase inhibitor emricasan was purchased from Selleckchem (S7775). The details of the antibody used in this study can be found in the section “Supplementary Material” (Supplementary Tables S1, S2).

Glioblastoma cells, SNB-19, were maintained in RPMI-1640 medium (ATCC) supplemented with 10% fetal bovine serum (Hyclone). Osteosarcoma cells, U2OS, were maintained in McCoy’s 5a Medium (ATCC) supplemented with 10% fetal bovine serum (Hyclone). The wild-type fibroblasts (Coriell Cell Repository, GM05659) were reprogrammed and induced to pluripotent stem cells, followed by neural induction to NSCs, as described previously (Yu et al., 2014). Neural stem cells were cultured in StemPro NSC SFM (Life Technologies) containing knockout Dulbecco’s modified Eagle’s medium-F12, StemPro neural supplement, 20 ng/ml bFGF, 20 ng/ml EGF and 1X GlutaMAX on Matrigel (Corning)-coated flasks. The following viruses were used: ZIKV MR766 strain (Uganda, 1947) and ZIKV PRVABC59 strain (Puerto Rico, 2015).

Cells were fixed with 4% paraformaldehyde (Sigma) for 15 min at room temperature. Samples were permeabilized with 0.25% Triton X-100 (Sigma) for 10 min and were blocked by cell staining buffer (BioLegend) for 1 h. Then, samples were incubated with primary antibody at 4°C overnight, followed by twice PBS washes and incubation with secondary antibody for 1 h at room temperature. Finally, nuclei were stained with 1 μg/ml Hoechst 33342 (Invitrogen) at room temperature for 15 min. After a final wash in PBS, 100 μl of fresh PBS was added to the cells for imaging on the IN Cell 2500 HS (GE Healthcare). Montages were generated using Fiji-ImageJ (NIH).

To determine the ZIKV infection rate, we took images using the IN Cell 2200 imaging system (GE Healthcare) with a 20× or 40× objective lens. Imaging detection was performed using FITC (ENV+/NS1+) and DAPI (nucleus) filter sets. Image analysis was conducted using the IN Cell Analyzer software (Version 3.7.2). With a multi-target analysis protocol, nuclei were segmented using the top-hat segmentation method with a minimum area set at 50 μm² and sensitivity set to 0. ZIKV ENV staining was identified as “cells” by the analysis software and was segmented using the multi-scale top-hat algorithm. Settings for ENV detection were set to 100 μm² and a sensitivity setting of 0. A filter for ENV+/NS1+ cells was set to 800 fluorescence units based on visual inspection of several wells. Infected cells were those with an average intensity greater than 800 fluorescence units. For each field in each well, the ENV+/NS1+ cells were divided by the total cell number as determined by the nuclear staining and multiplied by 100% to calculate the infection percentage per field. The final value is an average of three wells. Montages were generated using Fiji-ImageJ (NIH). To evaluate the NS1 protein expression level in the infected cell, we defined the cell contained 800–2,000 fluorescence unit of NS1 staining as a viral protein low expression cell, and the cell with NS1 intensity over 2,000 fluorescence unit as a viral protein high expression cell.

On the day of the experiment, cells were live-stained with 100 μl/well of 100 nM MitoTracker^TM Deep Red FM (Invitrogen) in the medium at 37°C for 15 min, followed by fixation in 100 μl/well 4% paraformaldehyde (Sigma) for another 15 min and twice washed with PBS. The nuclear staining was performed by the addition of 100 μl/well of 1 μg/ml Hoechst 33342 (Invitrogen) in PBS and incubation at room temperature for 15 min. The cells were then fixed and permeabilized for subsequent immunostaining. The images were taken by the IN Cell 2200 automated fluorescence plate imaging reader (GE Healthcare) with 20× or 40× objective lens, and imaging detection was performed using DAPI (nucleus) and Cy5 (MitoTracker) filter sets. Montages were generated using Fiji-ImageJ (NIH).

Images from the IN Cell 2200 (GE Healthcare) were loaded into FIJI (NIH) for image analysis. The Mitochondrial Network Analysis (MiNA) plugin was used in a blind analysis of randomly selected cells. Regions of interest were created manually, and the MiNA plugin quantitated the parameters described in their manuscript (Valente et al., 2017). Multiple cells per field per well were analyzed in at least three independent experiments.

Cells were fixed in 2% glutaraldehyde, 0.1 M cacodylate buffer, pH 7.2, for 1 h at room temperature and then stored at 4°C until transmission electron microscopy analysis was performed. The cells were washed with 0.1 M cacodylate buffer twice, post-fixed with 1% osmium tetroxide in 0.1 M cacodylate buffer for 1 h and washed again with 0.1 M cacodylate buffer twice and once with 0.1 N sodium acetate buffer before stained en bloc with 0.5% uranyl acetate in 0.1 N sodium acetate buffer for 1 h, pH 4.2. Then, cells were rewashed with 0.1 N sodium acetate buffer twice before gradual dehydration. The cells were then dehydrated in graded ethanol solutions (twice for each step of 35, 50, 70, 95%, and three times for 100%) and infiltrated overnight in pure epoxy resin (Poly/Bed 812, dodecenyl succinic anhydride (DDSA) and Nadic Methyl Anhydride (NMA) and DMP-30, Polysciences). Cells were rinsed with pure fresh resin twice, then cured in pure resin for 48 h at 55°C oven. After removing the polystyrene plates, suitable areas for thin sectioning were selected, cut out with jewelry saw, and glued onto empty resin stubs. About 70-nm thin sections were cut on an ultramicrotome (Leica EM UC6) and mounted on naked 150 mesh copper grids. The thin sections were double-stained (1:1 uranyl acetate and 70% ethanol and 1:1 lead citrate and DDH₂O) and examined with Hitachi H-7600 transmission electron microscope, and images were taken using an AMT CCD camera (Chen et al., 2010).

Cells were lysed in RIPA buffer (Enzo Life Sciences) supplemented with protease inhibitors and phosphatase inhibitor cocktail (Roche). The cell lysates were clarified by centrifugation at 20,000 × g for 15 min and followed by protein quantification with the BCA assay kit (Thermo Fisher Scientific). The cell lysates with similar protein concentrations were subsequently applied to Bis-Tris gels for protein separation, and the proteins were transferred from gels to polyvinylidene difluoride (PVDF) membrane by dry transfer (Thermo Fisher Scientific, iBlot 2 Gel Transfer Device). Immunoblot analysis was performed with the indicated antibodies, and the chemiluminescence signal was visualized with Luminata Forte Western HRP substrate (EMD Millipore) in the BioSpectrum system (UVP, LLC). The chemiluminescence intensities of the bands were quantified in the VisionWorks LS software (UVP, LLC).

An improved variant of GFP, mGFP, was used tagged to the open reading frame of MFN1 (RC207184L4V, Origene), MFN2 (RC202218L4V, Origene) or control (PS100071V, Origene) in a lentiviral particle plasmid. The stable over-expression cells were generated according to the product manual. Briefly, 1 × 10⁵ SNB-19 cells were seeded in each well of a 6-well plate and incubated overnight at 37°C. Then, cells were treated with 5 μg/ml polybrene (Santa Cruz Biotechnology, Inc.) and infected with lentiviral particles at MOI = 5. After overnight incubation with the virus, cells were incubated in fresh medium for another 24 h. The stable over-expression cells were selected by 2 μg/ml puromycin medium treatment for 3 days and kept in the 0.5 μg/ml puromycin (Santa Cruz Biotechnology, Inc.) medium for three passages before the test.

The ATP content assay kit (Promega or PerkinElmer Life Sciences) was applied to monitor the cell viability after ZIKV infection. Cells were seeded in white solid 96-well plates and infected with ZIKV of various MOIs as described in the result section. On the day of the experiment, 70 μl/well assay mixture (prepared according to the manufacturer’s instructions) was added to the assay plates, followed by incubation at room temperature for 10 min. The luminescence signal was determined in the luminescence mode of a Tecan multifunction plate reader.

DePsipher Mitochondrial Potential Kit (6300-100-K, R&D system) was applied to monitor the mitochondrial potential after ZIKV infection. Cells were plated in black 96-well plates with clear bottom and infected with ZIKV at MOI = 0.5, 1, or 2. On the day of the experiment, the DePsipher^TM solution was prepared according to the manufacturer’s instructions. Before DePsipher staining, the nuclear staining was performed by the addition of 100 μl/well of 1 μg/ml Hoechst 33342 (Invitrogen) in medium and incubation at 37°C for 15 min. After that, the medium was removed, and cells were covered with diluted DePsipher^TM solution, followed by incubating at 37°C for 15 min. Then, cells were washed with 100 μl of prewarmed 1× reaction buffer with Stabilizer Solution. The images were taken by the IN Cell 2200 automated fluorescence plate imaging reader (GE Healthcare) with 20× or 40× objective lens, and imaging detection was performed using DAPI (nucleus), FITC (DePsipher monomer) and TRITC (DePsipher aggregates) filter sets. Montages were generated using Fiji-ImageJ (NIH).

The GSE118305 and GSE101878 datasets were employed to predict the biological pathways in ZIKV infected cells and patients; they were downloaded from the GEO database¹. Differentially expressed genes (DEGs) were selected by a cutoff of p-values <0.05 by the T-test. GO & KEGG pathway analysis to identify the enriched functions by DEGs using the DAVID system². The p-value was calculated by hypergeometric distribution, and pathways with p < 0.05 were considered as significant. Rich factor refers to the percentage ratio of the DEGs number in the pathway and the number of all genes annotated in the pathway.

The GSE149775 dataset was download from the GEO database (see text footnote 1). For GSEA analysis of microarray data, the GSEA software was available on the GSEA-Broad Institute website. The statistical significance (nominal p value) of the enrichment score (ES) was estimated by running 1000 gene set permutations. The ES was normalized (NES) to account for the size of the gene set.

All data were presented as mean ± SD with at least three independent experiments unless otherwise stated. All imaging data are presented as mean ± SD and represent data from cells in at least 10 fields from three or more independent experiments. The two-tailed unpaired Student’s test of the mean was used for single comparisons of statistical significance, and the ANOVA test with Tukey’s multiple-comparison was used for the multiple comparisons of statistical significance between experimental groups. The linear regression fit was used to analyze correlation, and R square was used to evaluate the goodness of fit.

To assess the global effects of ZIKV infection on host cells, we analyzed a publicly available genome-wide RNAseq data set (GSE118305) that contains data from ZIKV-FSS13025 infected human blood monocyte-derived macrophages (Carlin et al., 2018). We chose this dataset in order to explore a time-dependent biological reaction during ZIKV infection extensively. Compared to ZIKV-negative data, a total of 6,364; 10,115; and 11,676 DEGs were identified from ZIKV-positive cells 12, 18, and 24 h after infection, respectively (Figures 1A,B). Of these, 3,774 genes began to show differential expression at 18 h, and these changes persisted after 24 h, while 3,114 genes were differentially expressed during all time points (Figure 1B). To further investigate the cellular pathology of ZIKV infection, each set of DEGs was submitted to gene ontology (GO) pathway analysis. In addition to the viral and interferon response pathways that had been reported in the original study (Carlin et al., 2018), our analysis revealed a series of mitochondria-related biological pathways. At the earliest time point, mitochondrial fusion was significantly enriched (p < 0.05) (Figure 1C). At 18 h, more mitochondrial fission-fusion and function-related pathways were enriched, as well as enrichments in mitochondrial respiratory chain assembly, regulation of mitophagy, and mitochondrial-related apoptosis (Figure 1C). After 24 h of infection, there was more enrichment of mitochondrial fission rather than fusion. The analysis also revealed several mitochondria-triggered apoptotic events such as mitochondrial outer membrane permeabilization and cytochrome C release.

To further explore whether these cell-based enrichments were replicated in ZIKV-infected patients, we analyzed another RNAseq data set (GSE101878) using myeloid dendritic cells from ZIKV-infected patients (Sun X. et al., 2017). By comparing the cells of five healthy, non-infected patients (controls) with ZIKV-infected patient cells, we identified a total of 4,398 DEGs that contained 2,276 increased-expression genes and 2,122 reduced-expression genes (Figure 1D). Consistent with the in vitro dataset, GO pathway analysis revealed enrichment of mitochondrial morphogenesis and function pathways (Figure 1E). Additionally, the time-course data from cell-based ZIKV infection experiments in this patient-based data set also showed the enrichment of mitochondrial pathways in dendritic cells (Supplementary Figures S1B,C).

To assess whether mitochondria-related changes also occurred in ZIKV-infected neural cells, analysis was conducted on RNA-seq data set (GSE149775) of ZIKV- (MR766 and PRVABC59) infected SH-SY5Y cells (Bonenfant et al., 2020). After gene set enrichment analysis (GSEA), a significant alteration (p < 0.05) was observed in the GO_MITOCHONDRIAL_PROTEIN_COMPLEX annotation for both MR766- and PRVABC59-infected cells, suggesting a dysregulation and dysfunction of mitochondria in neural cells after ZIKV infection (Supplementary Figures S1H,I). Together, these data support the possibility that mitochondrial morphological and functional changes may play a critical role in ZIKV-induced pathology.

To explore the interaction of ZIKV with mitochondria, we infected NSCs with either ZIKV strain MR766 (Uganda, 1947 isolate, African lineage) or PRVABC59 (Puerto Rico, 2015 clinical isolate, Asian lineage) for 24 h. MitoTracker staining of ZIKV-infected NSCs revealed abnormal mitochondrial morphology. Compared with the intact mitochondrial networks in control cells, ZIKV-infected cells showed more fragmented mitochondrial structures such as short rods and swollen puncta (Figure 2A). The mean mitochondrial branch length, the number of mitochondrial networks, and mitochondrial footprint were significantly decreased, indicating a marked reduction of mitochondrial networks and increased fragmentation (Figures 2B,C and Supplementary Figures S1D–G). We also observed the same morphological alterations in ZIKV-infected SNB-19 cells (Supplementary Figures S2A–G).

Next, we employed electron microscopy to achieve high-resolution images to assess mitochondrial network integrity. As shown in Figure 2D and Supplementary Figure S2H, more fragmented and swollen mitochondria were observed in ZIKV-infected NSCs and SNB-19 cells compared to uninfected cells. Of note, varying degrees of mitochondria structural damage and cristolysis (disappearing cristae) were found in ZIKV-infected cells. Additionally, the hive-shaped ZIKV particles accumulated in the rough endoplasmic reticulum cavity, which was juxtaposed with the fragmented mitochondria (Figure 2D, yellow arrows, and Supplementary Figure S2H). Collectively, these data indicated that ZIKV-infection disrupted healthy mitochondrial structure and caused mitochondrial fragmentation.

Because ZIKV is translated as a single polypeptide that is then post-translationally processed to yield all ten viral proteins in equimolar amounts (Gerold et al., 2017), we next performed immunostaining of ZIKV NS1 protein to monitor viral protein levels in cells. While mitochondria exhibited compact filamentous network structures in uninfected NSCs, fragmented mitochondria were observed in ZIKV infected cells that related to the expression of viral protein NS1 (Figure 2E and Supplementary Figure S2I). Increased mitochondrial fragmentation is positively related to increased viral protein expression (Figure 2F and Supplementary Figure S2J).

Furthermore, fragmented mitochondria were observed surrounding the sites of viral replication in MR766 and PRVABC59-infected cells, supporting the abovementioned electron microscopy observations (Figure 2E). Concurrent with the expression of ZIKV proteins, the number of mitochondrial network structures, the mean length of the mitochondrial branches, and the footprint of total mitochondria in the cells were markedly decreased (Figures 2G–I and Supplementary Figures S2K–M). Taken together, these data indicated that the magnitude of mitochondrial fragmentation was proportional to the amount of ZIKV proteins in host cells (Figure 2J).

Mitochondria are highly dynamic organelles; mitochondrial fusion and fission must be in equilibrium to maintain the healthy structure and function (Cao et al., 2017). To understand the cause of mitochondrial fragmentation during ZIKV infection, we hypothesized that an imbalance in mitochondrial fusion and fission proteins may lead to mitochondrial dysregulation. While only 50% of cells were infected after 24 h of virus exposure (Figure 3A and Supplementary Figure S3A), the protein, but not the mRNA, levels of MFN 2, a mitochondrial fusion-promoting protein (Rojo et al., 2002), was decreased in both ZIKV-infected NSCs and SNB-19 cells (Figures 3B,C). The reduction of MFN 1, which regulates mitochondrial outer membrane fusion (Chen et al., 2003), was only observed in SNB-19 cells but not in NSCs after ZIKV infection (Figure 3B and Supplementary Figure S3B), which might be due to the lower expression of MFN 1 in neuronal cells. As the decrease of MFN 2 in ZIKV-infected cells may be diluted in the homogenate assay due to a 50% infection rate, immunofluorescence staining was employed to detect the expression level of MFN 2 in each ZIKV-infected cell, and the results showed a weaker MFN 2 signal in ZIKV-infected cells than in uninfected cells (Figures 3D,E). Along with the decreased MFN2 protein, fragmented mitochondria were observed in ZIKV ENV-positive cells (Figures 3F–H and Supplementary Figure S3C). The results here indicated a decrease in mitochondrial fusion in ZIKV-infected cells (Figure 3I).

Protein levels of three other mitochondrial fission and fusion-related proteins, OPA1, DNM1L, and FIS1 were not significantly changed in ZIKV-infected cells (Figure 3B and Supplementary Figures S3D,E). Intriguingly, the phosphorylation of DNM1L at Ser616, a post-translational modification that activates mitochondrial fission (Gottlieb and Bernstein, 2016), was not phosphorylated in ZIKV-infected cells (Figure 3B and Supplementary Figure S3E). The dephosphorylation of DNM1L may have been due to a feedback regulation in response to the over-fragmented mitochondria to prevent even more fission. These data suggested that ZIKV infection yielded an imbalance of mitochondrial fusion and fission, leading to mitochondrial network fragmentation.

Typically, healthy mitochondrial morphology is reflected in the bioenergetics and function of the mitochondria (Yu T. et al., 2015). The mitochondrial transmembrane potential (delta-psi(m) or ΔΨm), generated by proton pumps (Complexes I, III, and IV), is a critical, bioenergetic parameter controlling mitochondria respiration, ATP synthesis, and the generation of reactive oxygen species (Nicholls, 2004). A drop in ΔΨm levels may induce dysfunction of mitochondria that results in a decrease in cell viability. Additionally, ΔΨm plays a key role in mitochondrial homeostasis through selective elimination of dysfunctional mitochondria. It is also a driving force for transport of ions (other than H⁺) and proteins which are necessary for healthy mitochondrial functioning (Zorova et al., 2018).

To determine whether the ZIKV-induced mitochondrial fragmentation disrupted mitochondrial bioenergetics, we measured the mitochondrial membrane potential using DePsipher dye, a unique cationic dye (5,5′,6,6′-tetrachloro-1,1′,3, 3′-tetraethylbenzimidazolyl-carbocyanine iodide) to indicate the loss of ΔΨm. The dyes readily enter cells and emit green fluorescence, while the dyes emit red fluorescence in healthy mitochondria as they are aggregated. When the mitochondrial membrane potential collapses, the DePsipher dye cannot accumulate within the mitochondria resulting in a reduction in red fluorescence. Thus, unhealthy mitochondria, showing primarily green fluorescence, are easily differentiated from healthy mitochondria, which show red puncta besides green fluorescence. In ZIKV-infected cells, staining with the dye DePsipher produced fewer red puncta compare with uninfected cells, indicating a reduction in mitochondrial membrane potential and impaired mitochondrial function (Figures 4A,B). In SNB-19 cells, the loss of ΔΨm was proportional to the MOI of ZIKV (started at MOI = 1) (Figure 4C). However, in NSC, infection of ZIKV PRVABC59 at MOI = 0.5 is sufficient to cause a significant decrease in membrane potential, suggesting that NSC is more sensitive to ZIKV infection than SNB-19, especially the PRVABC59 strain (Figure 4D). The cells with reduced mitochondrial membrane potential after ZIKV infection also exhibited increased mitochondrial fragmentation, fewer networks, decreased mean branch length, and a reduced mitochondrial footprint (Supplementary Figures S4A–C). The results revealed impaired mitochondrial function after ZIKV infection combined with mitochondrial fragmentation.

Although mitochondrial fragmentation has been shown to precede cell death, it remains unclear whether it is sufficient to activate the cell death pathway (James and Martinou, 2008). To examine whether mitochondrial fragmentation contributes to ZIKV-induced cell death, we determined the temporal relationship of mitochondrial fragmentation and cell death after ZIKV infection in NSCs and SNB-19 cells using a cell viability assay measuring ATP content in cells. Compared with uninfected controls, intracellular ATP levels began to decrease at 48 h post-infection in both the MR766 and PRVABC59 ZIKV strains (Figures 4E,F). As several studies have shown that ZIKV infection results in apoptosis of infected cells (Ghouzzi et al., 2016; Xu et al., 2016; Zhang et al., 2016; Devhare et al., 2017; Zhu et al., 2017), we also measured the caspase activity in the ZIKV-infected cells. Western blots showed significant activation of caspase-3, -9, -7, and poly (ADP-ribose) polymerase (PARP) (which indicated intrinsic apoptotic pathway activation), but not caspase-8 (which is involved in the extrinsic apoptotic signaling pathways), after 48 h of infection (Figure 4G). We also observed the release of cytochrome C from mitochondria and the aggregation of the apoptosis regulator BCL2-associated X (BAX) on the mitochondrial outer membrane 48 h post-infection (Supplementary Figures S4D,E), indicating that the intrinsic apoptotic pathway played a critical role in ZIKV-induced cell death.

Subsequently, we used the pan-caspase inhibitor emricasan to block cell death caused by ZIKV infection (Hoglen et al., 2004), which validated results from our previous study (Xu et al., 2016). However, mitochondrial fragmentation was still observed in the emricasan treated cells after ZIKV infection (Figures 4H,I and Supplementary Figure S4F). This result suggested that ZIKV-induced mitochondrial fragmentation was prior to or independent of caspase activation and apoptotic cell death in ZIKV-infected cells. Regardless of the presence of emricasan, MFN 2 was significantly decreased in ZIKV-infected cells, and the reduced MFN 2 level correlated with an increased mitochondrial fragmentation (Figures 4J,K and Supplementary Figure S4G). These data suggested that mitochondrial fragmentation could precede or be independent of apoptosis (Figure 4L).

Mitochondrial fission and fusion are balanced to maintain healthy mitochondrial morphology and function in cells. To further explore if blocking mitochondrial fragmentation would help increase cell survival after ZIKV infection, we over-expressed both MFN1 and MFN2 to restore normal fusion and treated cells with Mdivi-1 to inhibit DNM1L activity, reducing mitochondrial fission. Both over-expression of MFNs and treatment with Mdivi-1 produced a hyperconnected mitochondrial network in uninfected cells (Figure 5A). However, a combination of over-expressed MFNs and Mdivi-1 treatment did not further change the mitochondrial structure (Figure 5A). In uninfected control cells, Mdivi-1 treatment significantly increased the mean branch length of mitochondria, but not the numbers of individual mitochondria (Figures 5B,C). After ZIKV infection, Mdivi-1 treatment attenuated mitochondrial fragmentation in control SNB-19 cells. We also observed that in MFN1 or MFN2 over-expressing cells, Mdivi-1 moderately increased mitochondrial branch length and the number of mitochondria in the networks. Although ZIKV infection still shortened mitochondrial branch length in cells overexpressing MFN1 or MFN2, the range of branch lengths in these cells was similar to uninfected controls (Figure 5B). The mitochondrial networks were only spared in MFN2 and not MFN1 overexpressing cells after ZIKV infection (Figure 5C). These data suggested that either inhibition of mitochondrial fission (by inhibiting DNM1L) or induction of mitochondrial fusion (by over-expression of MFN2) could counteract the mitochondrial fragmentation caused by ZIKV infection (Figure 5D).

In addition, we found that Mdivi-1 treatment alone prevented a decrease in cell viability after 48 h of infection without affecting the rate of infection (Figure 5E and Supplementary Figure S5A). Together with our previous data, this result suggested that mitochondrial fragmentation precedes ZIKV-induced cell death. However, we did not find that over-expression of MFN2 or MFN1 protein rescued cell death caused by ZIKV infection (Supplementary Figure S5B). This phenomenon could be due to increased mitochondrial fusion that does not effectively reduce the cytochrome C release from ZIKV-promoted mitochondrial fission, leading to the apoptosis (Suen et al., 2008). Collectively, these data indicated that maintenance of normal mitochondrial dynamics by inhibiting mitochondrial fission corrected ZIKV-induced mitochondrial changes and increased cell survival after ZIKV infection. This finding hints at a potential therapeutic intervention – inhibition of mitochondrial fission in neural cells – as a treatment for ZIKV infection.