**4. Discussion**

Our data demonstrate that RIG-I is the main sensor for ZIKV infection in A549 cells. Genetic ablation of RIG-I in A549 cells led to a loss of type I IFN production and ISG induction as well as to an increase in virus titer. Transcriptomic analysis corroborated that knockdown of RIG-I strongly reduced differential gene expression upon ZIKV infection. Importantly, most ISGs induced after ZIKV infection were RIG-I-dependent. This is likely due to a combination of reduced type I IFN secretion by RIG-I-deficient cells and reduced activation of IRF3, which directly regulates some ISGs [46]. This RIG-I-dependency of ZIKV sensing is in line with a recent publication by Esser-Nobis

and colleagues [16]. RIG-I is thought to recognize the conserved 5-triphosphate group found on nascent RNAs of flaviviruses as shown by Chazal et al. [50]. This is interesting as flaviviruses replicate in complexes formed in invaginations of ER membranes, raising the question as to how RIG-I gains access to viral RNAs [51]. Replication factories are likely to be dynamic structures and viral RNAs can potentially leak into the cytoplasm. Studies on dengue virus, West Nile virus and tick-borne encephalitis virus furthermore revealed 10-nm-wide openings of these invaginations to the cytoplasm using electron tomography [51]. ZIKV RNA is thought to exit to the cytoplasm through these pores to be packaged into virions and for protein translation [52]. RIG-I can be activated by less than 20 RNA molecules per cell [53]. A few nascent positive-stranded RNA molecules that escape replication factories would therefore be su fficient to induce an IFN response by RIG-I.

Our study furthermore revealed that RIG-I prevents ZIKV-induced apoptosis, likely due to RIG-I-induced innate immunity that curtails virus replication. ZIKV-infected A549 RIG-I KO cells succumbed to cell death 4 days post infection with a low MOI and showed increased cleavage of PARP as well as activation of caspases-3 and -7. Several studies suggested that ZIKV-induced cell death in neuronal cells is responsible for neurodevelopmental defects such as microcephaly [36,43,54]. An increase in cell death upon ZIKV infection was modeled in brain-specific organoids derived from human induced pluripotent stem cells (iPSCs) [55,56]. Studies in di fferent cell lines sugges<sup>t</sup> that the induction of apoptosis may be cell-type specific. A549 cells were shown to succumb to apoptosis 48 h after infection when infected with a high MOI [57]. In contrast, human monocyte-derived dendritic cells, Vero cells or mosquito C6/36 cells infected with several di fferent African or Asian ZIKV strains did not induce apoptosis 24 and 48 h after infection [58]. Our work now shows that a functional immune response to ZIKV infection protects cells from apoptosis. The observed cell-type specific di fferences could therefore correlate with how e fficiently a cell senses the virus and how potently the virus is restricted by the initiated type I IFN response. It is thus important to study the levels and functionality of RIG-I in cell types infected by ZIKV, including neuronal cells. Furthermore, ZIKV delays apoptosis by modulating the activities of anti-apoptotic Bcl-2 family proteins [59]. It will be interesting for future studies to determine if and how viral targeting of RIG-I and Bcl-2 proteins is functionally linked.

The importance of RIG-I and type I IFN to ZIKV infection is also evident from the presence of viral antagonists. Here, we confirmed that NS5—one of the most potent viral antagonists—not only blocks type I IFN signaling, but also e fficiently and directly inhibited type I IFN production triggered by RIG-I, as suggested previously [6,13,44,45,60]. It is now important to identify the precise mechanisms by which the RLR signaling cascade is targeted by NS5. Recently, Li et al. described that NS5 directly represses K63-linked polyubiquitination of RIG-I [61]. In addition, an inhibitory e ffect of NS5 on IRF3 activation has been suggested [44,62,63]. Interestingly, Lin and colleagues reported an interaction of NS5 with TBK1 that results in reduced phosphorylation of IRF3 [45]. The latter findings were made by NS5 overexpression in HEK293 cells. It would be very interesting in future studies to confirm these findings at endogenous protein level during infection with live ZIKV. In ZIKV-infected cells, RIG-I-dependent responses are induced despite the presence of NS5 (Figures 1–3). It is therefore likely that NS5s ability to block RIG-I signaling is not absolute. Whether this relates to relative protein levels in infected cells or to cell-to-cell variability remains to be determined. A virus expressing an NS5 mutant that fails to interfere with RIG-I signaling but maintains other functions of NS5 would be useful for such studies. It is noteworthy that a recent report using immortalized human fetal astrocytes and an siRNA approach found that both RIG-I and MDA5 were required for induction of *IFN*β and *ISG* transcripts [17]. This work further described an inhibitory e ffect of ZIKV NS3 on RIG-I and MDA5 signaling, in apparent contradiction to our data shown here in Figure 4D and in [13]. It is possible that these di fferences relate to cell-type specific expression of co-factors of RIG-I and MDA5, an interesting hypothesis for future studies.

Taken together, our study emphasizes the importance of RIG-I-mediated-ZIKV-sensing in controlling virus replication and virus-induced cell death. Targeting viral antagonists to support ZIKV
