**3. Results**

#### *3.1. RIG-I Is the Main Sensor for ZIKV Infection in A549 Cells*

Previously, we showed that total RNA extracted from cells infected with ZIKV contains immunostimulatory RNAs that activate the MAVS pathway when transfected into reporter cells [13]. To study the contribution of RIG-I and MDA5 to sensing of live ZIKV in infected cells, we screened a panel of cell lines for type I IFN induction after ZIKV infection. We found that the lung adenocarcinoma cell line A549 produced robust levels of type I IFNs upon infection (data not shown). Furthermore, this cell line was well established for studies of type I IFN responses and was therefore chosen for this work. We generated A549 cells lacking either RIG-I or MDA5 by CRISPR/Cas9-mediated knock-out (KO) and validated these by western blot analysis and sequencing (Figure 1A,B). We obtained one MDA5 KO clone designated c27. MDA5 protein expression was undetectable in these cells (Figure 1A). Sequencing of the region targeted by the sgRNA suggested a +1 insertion on one allele and a –5 deletion on the other, both of which disrupt the reading frame (Figure 1B). Three RIG-I KO clones were generated, and all had no detectable RIG-I protein (Figure 1A). RIG-I KO clone B05 was used in subsequent experiments; sequencing showed a +1 insertion as well as –1 and –2 deletions at the sgRNA target site, which all disrupt the reading frame (Figure 1B). The presence of three different alleles could be explained by the triploidy of much of the A549 genome [33,34].

In order to compare the amounts of type I IFN produced, wild-type (wt) and KO A549 cells were infected with ZIKV using a multiplicity of infection (MOI) of 0.1 or 1. After 24 h, we collected supernatants and measured IFNβ levels by ELISA. These virus doses and the timepoint were chosen to monitor type I IFN responses to incoming virus early after infection. Similar amounts of IFNβ were present in supernatants from wt and MDA5 KO cells (Figure 1C). In contrast, little or no IFNβ was detectable in samples from RIG-I KO cells. Next, we measured bioactive type I IFN levels in supernatants collected from cells infected (MOI 1) for 48 h by using a bioassay: supernatant samples were transferred onto HEK293 cells with a stably integrated pGF1-ISRE reporter [26]. These cells

harbor an F-Luc gene under control of interferon-stimulated response elements (ISREs) that were bound and activated by STAT1/2 upon engagemen<sup>t</sup> of IFNAR. Cells stimulated with the supernatant of infected wt or MDA5 KO cells induced similar amounts of F-Luc, whereas the supernatant of infected RIG-I KO cells did not lead to significant F-Luc induction (Figure 1D). Furthermore, we tested the activation of IRF3 in infected cells by western blot using an antibody recognizing S396-phosphorylated IRF3 (p-IRF3). This analysis revealed IRF3 phosphorylation upon ZIKV infection in wt and MDA5 KO cells, but not in RIG-I KO cells (Figure 1E). At the selected MOIs and 24-h timepoint analyzed, infection levels were similar in cells of all genotypes as indicated by comparable levels of the viral NS3 protein (Figure 1E). In summary, these data demonstrated that loss of RIG-I abrogated the induction and secretion of type I IFN in A549 cells upon ZIKV infection. To examine the impact of reduced IRF3 activation and type I IFN secretion on ISG induction, A549 cells were infected with ZIKV (MOI 1 or 5) and *IFIT1* and *MX1* mRNA levels were quantified by RT-qPCR. *IFIT1* mRNA was robustly induced in A549 wt and MDA5 KO cells, whereas no induction was detectable in A549 RIG-I KO cells (Figure 1F). Similarly, induction of *MX1* transcripts was not detectable in RIG-I KO cells; however, in contrast to *IFIT1*, *MX1* induction was also reduced in MDA5 KO cells (Figure 1F). This suggested that a subset of ISGs was controlled by both RIG-I and MDA5. To determine the impact of individual RLRs on ZIKV replication in a setting where the infection spreads between cells, we infected cells with a low dose of ZIKV (MOI 0.1) and ZIKV RNA was quantified by RT-qPCR up to 5 days post infection. Virus replication was similar in A549 wt and MDA5 KO cells; however, the virus replicated more potently from day 3 onwards in RIG-I KO cells (Figure 1G). Taken together, these data sugges<sup>t</sup> that RIG-I was the main sensor that detects ZIKV infection in A549 cells leading to the induction of type I IFNs and ISGs. In turn, absence of RIG-I facilitated virus replication.

**Figure 1.** *Cont.*

**Figure 1.** Retinoic acid-inducible gene I (RIG-I) is the main sensor for Zika virus (ZIKV) infection in A549 cells. ( **A**) A549 cells were knocked-out for melanoma di fferentiation-associated gene 5 (MDA5) or RIG-I as described in Materials and Methods. Cells were stimulated with recombinant type I IFN (IFN-A/D) to induce MDA5 and RIG-I to detectable levels and lysates were analyzed by western blot using the indicated antibodies. Actin served as a loading control. The vertical line indicates a cut combining two parts of the same blot. c27, A14, B05 and B20 are individual clones. wt, wild type; (**B**) Genomic DNA was extracted from MDA5 KO clone c27 or RIG-I KO clone B05 cells. A fragment of DNA surrounding the targeted area was amplified by PCR and sequenced (left). Sequences were analyzed using TIDE (right). The number of nucleotides inserted or deleted, and the percentage of sequences a ffected are shown; ( **C**) A549 MDA5 KO or RIG-I KO cells (clone B05) were infected with ZIKV (MOI 0.1 or 1), supernatant was collected 24 h later and IFNβ levels were analyzed by ELISA. The horizontal dashed line indicates the detection limit; n.d., not detectable; ( **D**) A549 cells were infected with ZIKV (MOI 1) and supernatant was collected 48 h later. HEK293 cells stably expressing the pGF1-ISRE reporter were incubated with the supernatant and F-Luc activity was measured after 24 h. Shown is the fold induction relative to supernatant from mock infected cells. (**E**) A549 cells were infected with ZIKV (MOI 0.1 or 1), protein samples were collected 24 h later and analyzed by western blot using the indicated antibodies. Actin served as a loading control; (**F**) A549 cells were infected with ZIKV (MOI 1 or 5) and RNA was isolated 24 h later. Levels of *IFIT1* and *MX1* mRNAs were determined with RT-qPCR and CT values normalized to *GAPDH*; ( **G**) A549 cells were infected with ZIKV (MOI 0.1) and RNA was isolated at the indicated time points. RT-qPCR was performed and ZIKV RNA levels are presented relative to *GAPDH*. Data in **A**, **C** and **E** are representative of two independent experiments. Data in **D**, **F** and **G** are pooled from three ( **D**,**G**) and four (**F**) independent experiments. Each data point is the mean value of two technical replicates. Statistical analysis: One-way ( **D**) and two-way (**F**,**G**) ANOVA with Tukey's multiple comparison (\* *p* < 0.05, \*\*\* *p* < 0.001).

#### *3.2. Transcriptomic Analysis of ZIKV-Infected Cells Indicates that RIG-I Plays a Dominant Role in ISG Induction*

In light of our observation that the induction of *MX1* transcripts after ZIKV infection was not only RIG-I, but also partially MDA5-dependent, we wanted to further investigate how the individual receptors influence transcriptomic changes after virus infection. We therefore performed 3' mRNA sequencing of total RNA that was extracted from ZIKV-infected cells 24 h after infection using an MOI of 5 to robustly induce ISGs. Our analysis included four biologic replicates each for uninfected and infected wt, RIG-I KO or MDA5 KO cells. Across cells of all genotypes, a total of 236 genes were di fferentially regulated upon ZIKV infection (Figure 2A). Most genes were di fferentially expressed in A549 wt and MDA5 KO cells with a substantial overlap between the two. Only 24 genes were di fferentially regulated in RIG-I KO cells, indicating that transcriptomic changes upon ZIKV infection were largely driven by RIG-I. This was also evident from the heatmap in Figure 2B where ZIKV-infected RIG-I KO cells rather clustered with the uninfected instead of the ZIKV-infected cells of other genotypes. Most the 236 genes di fferentially expressed upon ZIKV infection were upregulated while only few were downregulated (Figure 2B). Next, we analyzed ISGs using the gene set defined in [35]. A total of 98 genes di fferentially expressed in ZIKV-infected cells were ISGs. Most ISGs—such as *IFIT1* or *RSAD2* (also known as viperin)—were upregulated in a RIG-I-dependent and MDA5-independent manner (Figure 2C,D). Induction of a small number of ISGs not only required RIG-I but was also partially MDA5-dependent (Figure 2C). As predicted from our RT-qPCR results shown in Figure 1D, this included *MX1* (Figure 2D). Other ISGs were induced to a similar extent upon infection in all three cell lines, including *CCL5* (Figure 2C,D). RIG-I and MDA5 may be redundant for activation of these genes—or their induction could require other signaling pathways. Taken together, these data show that most transcriptional changes in A549 cells upon ZIKV infection occurred in a RIG-I-dependent manner and that RIG-I was particularly important for the induction of ISGs.

**Figure 2.** RIG-I sensing drives most transcriptional changes upon ZIKV infection. (**A**) Venn diagram showing differentially expressed genes in wt, RIG-I KO and MDA5 KO cells 24 h after infection; (**B**–**D**) Heat maps depicting all 236 differentially expressed genes (**B**), 98 differentially expressed ISGs (**C**) and three exemplary ISGs with different expression profiles in A549 wt, RIG-I KO and MDA5 KO cells upon ZIKV infection (**D**). Colors represent *z*-scores that indicate a value's relationship to the mean, measured as standard deviations from the mean. *z*-scores calculated for each row (i.e., each gene) and were plotted instead of the normalized expression values to ensure that expression patterns are not overwhelmed by absolute expression values. Data in (**A**–**D**) pooled from four independent biologic samples.

#### *3.3. RIG-I-Mediated Signaling Protects A549 Cells from Apoptosis*

ZIKV infection causes apoptosis [36–43]. We therefore asked whether reduced innate immune recognition of ZIKV in RIG-I KO cells impacts virus-induced cell death. A549 wt, RIG-I KO and MDA5 KO cells were infected with a low dose of ZIKV (MOI 0.1) to analyze a spreading infection and the confluency of the cells was measured for 6 days using an in-incubator imaging system (Incucyte). Interestingly, we found that after 6 days of infection the confluency of RIG-I KO cells was decreased by about 50%, whereas the confluency of wt and MDA5 KO cells was not a ffected by ZIKV infection (Figure 3A). Furthermore, crystal violet staining revealed virus-induced cell death in two di fferent RIG-I KO clones, but not in wt or MDA5 KO cells six days after infection with two di fferent doses of virus (MOI 0.1 and 0.01) (Figure 3B). To determine whether apoptosis was induced during ZIKV infection in RIG-I KO cells, we performed western blot analysis of PARP cleavage, a molecular signature of apoptosis. Indeed, ZIKV infection resulted in increased levels of cleaved PARP in RIG-I KO cells, but not in wt and MDA5 KO cells (Figure 3C). In addition, we monitored activity of the apoptotic caspases-3 and -7. Four days after ZIKV infection, RIG-I KO cells showed a 4-fold induction of caspase-3/<sup>7</sup> activity, while only a 2-fold induction was observed in wt and MDA5 KO cells (Figure 3D). Furthermore, ZIKV-infected cells displayed shrinkage and membrane blebbing, morphologic changes typical for apoptotic cells (Figure 3E). Taken together, these data showed that a lack of RIG-I signaling in A549 cells led to a loss of protection from ZIKV-induced apoptosis, which may be due to reduced type I IFN production and increased virus replication in RIG-I-deficient cells.

#### *3.4. ZIKV NS5 Inhibits Type I IFN Induction*

ZIKV NS5 inhibits type I IFN signaling by inducing degradation of STAT2 and by blocking phosphorylation of STAT1 [6,13,19,21,44]. Results of overexpression studies suggested that NS5 also blocks the innate immune response by inhibiting the induction of type I IFN [6,13,44,45]. However, it is possible that the latter e ffect is indirect as RLRs and many proteins involved in their downstream signaling are encoded by ISGs [46]. As such, lower levels of type I IFN induction in cells expressing NS5 could be explained by reduced levels of RLRs or other proteins involved in type I IFN induction. To distinguish between such indirect e ffects of NS5 and direct inhibition of type I IFN induction, we generated HEK293 cells lacking IFNAR1 and obtained one clone designated c27. As expected, IFNAR1 KO cells were incapable to phosphorylate STAT1 in response to IFN α2a (Figure 4A). We further validated our IFNAR1 KO cells by sequencing and found −1, −3 and −5 deletions at the sgRNA target site (Figure 4B; HEK293 are largely triploid [47,48]). While the −1 and −5 deletions result in a frameshift, the −3 deletion removes one amino acid keeping the reading frame intact. Our functional data in Figure 4A showing the absence of response to IFN α2a sugges<sup>t</sup> that the protein encoded by the –3 mutant allele is either non-functional or rapidly degraded.

We then transfected wt and IFNAR1 KO HEK293 cells with an expression plasmid for ZIKV NS5. We used empty vector and ZIKV NS3, which we previously found not to block RLR signaling [13] or EMCV L, which blocks IRF3 [49], as negative and positive controls, respectively. Alongside these expression plasmids, cells were co-transfected with an *IFN*β promoter F-Luc reporter construct and R-Luc as a transfection control. Next, we stimulated RIG-I or MDA5 by transfecting 5-triphosphate containing in vitro transcribed RNA (IVT–RNA) or RNA extracted from EMCV-infected Hela cells (Hela–EMCV–RNA), respectively (Figure 4C, [13]). Luciferase activities were measured 24 h after RNA transfection. As expected, both immunostimulatory RNAs induced the *IFN*β promoter in wt cells that had been transfected with empty vector or ZIKV NS3, while the response was strongly reduced by ZIKV NS5 and EMCV L (Figure 4D). Importantly, NS5 inhibited induction of the *IFN*β promoter to a similar extent in IFNAR1 KO cells stimulated with IVT–RNA (Figure 4D). In addition, the response to Hela–EMCV–RNA also appeared to be reduced by NS5 in IFNAR1 KO cells, although this trend did not reach statistical significance. We conclude that ZIKV NS5 blocked RIG-I-mediated IFN induction in the absence of IFNAR signaling. These observations sugges<sup>t</sup> that ZIKV NS5 not only inhibits antiviral

responses downstream of IFNAR signaling, but also has a direct effect on the induction of type I IFNs by RLRs.

**Figure 3.** RIG-I signaling protects A549 cells from ZIKV-induced cell death. (**A**) A549 cells were infected with ZIKV (MOI 0.1) and cell confluency was measured for 6 days in the Incucyte; (**B**) A549 cells were infected with ZIKV, fixed 6 days after infection and stained with crystal violet; (**C**) A549 cells were infected with ZIKV (MOI 0.1) and lysed 4 days after infection. Cell lysates were analyzed by western blot using the indicated antibodies (left). Actin served as a loading control. Signal intensity of full length and cleaved PARP was quantified relative to background and the fold induction of cleaved PARP was calculated (right); (**D**) A549 cells were infected with ZIKV (MOI 0.1) and the activity of caspase 3 and 7 was determined 4 days after infection using the Promega Caspase-3/7 Glo assay; (**E**) A549 cells were infected with ZIKV (MOI 0.1) and images were acquired 4 days after infection. Scale bar: 10 μm. Data in **A**, **C** (right) and **D** are pooled form two (**C**) or three (**A**,**D**) independent experiments. In **A**, mean and SD are plotted; in **C** and **D**, data points correspond to individual experiments and statistical analysis was with one-way ANOVA with Tukey's multiple comparison (\* *p* < 0.05). Data in **B**, **C** (left) and **E** are representative of two (**C**) or three independent experiments (**B**,**E**).

**Figure 4.** ZIKV NS5 inhibits IFN induction. (**A**) wt HEK293 and IFNAR1 KO clone c27 cells were stimulated with 5000 U/mL IFNα2a for 30 min before fixation and intracellular staining with α-pSTAT1 antibody. As controls, some wt cells were not stimulated or the α-pSTAT1 antibody was omitted. pSTAT1 levels were determined by flow cytometry. Mean fluorescence intensity (MFI) of the pSTAT1 signal was calculated and set to 100 in wt cells; (**B**) genomic DNA was extracted from IFNAR1 KO clone c27 cells. A fragment of DNA surrounding the targeted area was amplified by PCR and sequenced (left). Sequences were analyzed using TIDE (right). Number of nucleotides inserted or deleted, and the percentage of sequences affected are shown; (**C**) schematic of the experiment in **D**; (**D**) wt and IFNAR1 KO cells were transfected with the indicated expression plasmids, a plasmid encoding F-Luc under the control of the *IFN*β promoter and a plasmid, which expresses R-Luc. Twenty-four hours later, cells were transfected with 5 ng IVT–RNA or 50 ng Hela–EMCV–RNA per well. F-Luc activity was determined 24 h after RNA transfection and normalized to R-Luc. Data in **A** and **D** are pooled from two and three independent experiments, respectively. Data points correspond to individual experiments and in **D** are mean values of technical triplicates. Statistical analysis: Two-way ANOVA with Tukey's multiple comparison (ns *p* ≥ 0.05, \* *p* < 0.05, \*\* *p* < 0.01, \*\*\* *p* < 0.001).
