Identification of AMOTL2 as an antiviral factor that enhances the human type I interferon response against Zika virus

Significance

Zika virus (ZIKV) can cause severe disease, especially in infants who are infected with the virus in utero. ZIKV incidence is expected to increase in the coming years. Studying how our innate immune system engages with ZIKV can help us better understand how to control the virus. We inactivated different human innate immune genes and observed how this affected ZIKV replication in a cell culture system. We found that the human gene AMOTL2 has an antiviral effect on ZIKV. Specifically, AMOTL2 strengthens the type I interferon (IFN) response, which is responsible for creating antiviral proteins that can directly bind and restrict viruses. This study improves our understanding of ZIKV host–pathogen interactions.

Abstract

Zika virus (ZIKV) has caused multiple human outbreaks, with more recent epidemics associated with severe outcomes in infants. Today, ZIKV is endemic to many countries and presents a persistent threat for future epidemics. The host innate immune proteins that regulate ZIKV replication are incompletely defined. We developed a CRISPR knockout screen to identify host factors that impact ZIKV replication, resulting in the finding of angiomotin-like protein 2 (AMOTL2), a protein that inhibits ZIKV by regulating the host type I interferon (IFN) response. AMOTL2 affects IFN signaling by modulating STAT1 levels and activation in response to type I IFN. Thus, AMOTL2, which has largely been studied for its role in cancer, represents an antiviral protein that interacts with the IFN signaling pathway to promote downstream expression of IFN stimulated genes, resulting in restriction of ZIKV.

Zika virus (ZIKV) is a mosquito-borne flavivirus that was first discovered in a rhesus monkey in Uganda in 1947 (1). While ZIKV caused several human outbreaks in the 20th century, the burden of disease appeared to be limited until ZIKV’s emergence in the 21st century in the Pacific Islands (2007-2015) and South America (2015-2016) (2). In 2016 alone, there were over 500,000 suspected symptomatic cases of ZIKV across the Americas (3). Though ZIKV infection is usually mild or asymptomatic, it was discovered during the 2016 outbreak that ZIKV can cause microcephaly and other serious developmental anomalies in prenatal infections (4). Furthermore, ZIKV can be a trigger for Guillain–Barré syndrome, an autoimmune-mediated paralysis that can cause chronic neurologic impairment (5, 6). Today, ZIKV remains endemic to the Americas, where population-level immunity from the 2015 to 2016 outbreak is thought to be preventing larger outbreaks (7, 8). ZIKV outbreaks occur sporadically in India (9), and there is evidence that ZIKV circulates at low levels in regions of Africa where the host mosquito vector is present (2, 10, 11), potentially providing a reservoir for future outbreaks. Thus, ZIKV remains a serious threat to global public health.

Type I interferons (IFNs) are crucial mediators of the innate immune response against viruses, including ZIKV. Type I IFNs, including IFN-α and IFN-β, are cytokines that are released from infected cells upon detection of virus. They bind to the ubiquitously expressed IFNAR1 and IFNAR2 receptor subunits in an autocrine and paracrine response, resulting in receptor dimerization, autophosphorylation of associated Janus kinase (JAK) tyrosine kinases, and phosphorylation and activation of STAT1 and STAT2 proteins. Activated STAT1 and STAT2 form a complex with IRF9 to make IFN-stimulated gene factor 3 (ISGF3), which translocates to the nucleus to bind IFN-stimulated response elements in gene promoter regions, resulting in the transcription of hundreds of IFN-stimulated genes (ISGs) (12). A subset of these ISGs encodes proteins that act in diverse ways to limit viral replication, with many ISGs having evolved specificity to certain viruses (13). Many ISGs that directly inhibit viral proteins have been described (13). In addition, given the central role of type I IFN in regulating an antiviral state, any host genes that modulate IFN signaling also have the potential to exert a potent effect on viral replication by broadly affecting the expression of some or all antiviral ISGs.

The type I IFN response effectively inhibits ZIKV infection, as evidenced by extensive in vitro data on IFN’s potent effect against ZIKV (14–16) and the fact that mouse models of ZIKV pathogenesis require ablation of IFN signaling for productive infection (17–19). Additionally, polymorphisms in toll-like receptors 3 and 7, which normally trigger IFN production, are associated with severe clinical outcomes in ZIKV-infected newborn humans (20, 21). Several studies have characterized ISGs that block ZIKV (22–35), but only one loss-of-function screen has been performed to identify the larger milieu of host factors that contribute to type I IFN inhibition of ZIKV (36).

In this study, we performed a CRISPR knockout screen to identify genes important to type I IFN-mediated restriction of ZIKV. Our screen results validated a known flavivirus restriction factor, IFI6 (22, 23), as an antiviral ISG with activity against ZIKV. We also identified a gene, AMOTL2, that is constitutively expressed yet possesses antiviral activity against ZIKV only when cells are treated with IFN. AMOTL2 is known to be involved in angiogenesis, tight junction formation, cell polarity, the actin cytoskeleton, and regulation of Hippo signaling (37), but has no described role in innate immunity. We found that AMOTL2 modulates total STAT1 levels, resulting in enhanced nuclear translocation of phosphorylated STAT1 and heightened ISG transcription. AMOTL2 thus represents a host factor contributing to inhibition of ZIKV replication through its effects on the type I IFN signaling pathway.

Results

Development of a CRISPR Screen Based on ZIKV-Induced Cell Death.

We optimized a CRISPR knockout screen to identify genes contributing to IFN-β restriction of ZIKV. A549 epithelial cells were selected because they are highly susceptible to IFN-induced inhibition of ZIKV (14). We repurposed the previously described Packageable ISG Knockout Assembly (PIKA) CRISPR library of single guide RNAs (sgRNAs) targeting 1,905 putative ISGs identified in diverse cell types (38), including microarray data from IFN-stimulated A549 cells (39). To further verify that ISGs in A549 cells are well represented in the library, we performed bulk RNA-Seq on A549 cells stimulated with IFN-β for 24 h. Comparison with genes upregulated at least fourfold upon IFN-β stimulation of A549 cells indicated that the library contains sgRNAs targeting 84% of these genes (n = 174). Thus, the library is enriched for ISGs of interest but also includes many other genes that are not upregulated by IFN-β in A549 cells (Fig. 1A). This afforded the potential to identify both ISGs and non-ISGs that contribute to IFN-β restriction of ZIKV.

Fig. 1.

Development of a CRISPR knockout screen to identify antiviral genes against ZIKV based on ZIKV-induced cell death. (A) Overlap between genes in the PIKA library and genes upregulated at least fourfold by 24 h of IFN-β treatment in A549 cells. (B) Schematic of the CRISPR screen method. (C) CRISPR screen results from two independent replicates. Gene-level MAGeCK enrichment scores were correlated between screen replicates. The dotted lines indicate the most-enriched NTC sgRNA in each screen.

Because ZIKV infection causes apoptosis in A549 cells (15), we leveraged ZIKV-induced apoptotic cell death to detect cells that were permissive to ZIKV infection in the presence of IFN-β. We hypothesized that if a ZIKV antiviral gene was inactivated, cells carrying that inactivated gene would be enriched in a population of IFN-treated, ZIKV-infected dead cells, which would no longer be adherent. To define the optimal window to capture dying cells in the supernatant, untreated and IFN-β-treated A549 KO library cells (A549 cells transduced with the PIKA library) were infected with ZIKV and stained with Annexin-V, which labels apoptotic cells, every 6 h from 24 to 48 h post infection (hpi) (SI Appendix, Fig. S1) (40, 41). We observed that ZIKV infection caused an appreciable increase in Annexin-V+ cells in the supernatant after ~30 h of infection in the absence of IFN treatment (SI Appendix, Fig. S1). By contrast, cells pretreated with IFN-β showed similar levels of dead cells as in noninfected cells, consistent with the protective effect of IFN-β on ZIKV replication (SI Appendix, Fig. S1). From these data, we chose 42 hpi to conduct the CRISPR screen, as the level of ZIKV-induced cell death was clearly above (by ~threefold) the IFN-treated and uninfected conditions. This would allow us to maximize our chances of detecting cells that preferentially succumbed to ZIKV-induced cell death due to inactivation of antiviral genes important in restricting ZIKV.

CRISPR Knockout Screen Reveals Candidate Genes That Contribute to IFN-β Restriction of ZIKV.

With conditions optimized for a ZIKV-induced cell death-based screen readout, we performed CRISPR screens with the A549 KO library cells. Cells were pretreated with IFN-β for 24 h prior to ZIKV infection, followed by replenishment with IFN-β-containing media. In parallel, we included a control condition of A549 KO library cells that were IFN-treated but uninfected to account for cell death from factors other than ZIKV infection. The supernatant from each condition was collected at 42 hpi and pelleted to collect dead/dying cells. Enriched CRISPR-targeted genes in ZIKV-infected, IFN-treated dead/dying cells were detected by sequencing the sgRNAs and comparing to the sgRNAs detected in the uninfected, IFN-treated dead/dying cells (Fig. 1B). sgRNAs targeting members of the type I IFN signaling pathway (IFNAR1, STAT1/2, IRF9), included in the PIKA library as positive controls, were the most enriched in two independent replicates of the screen (Fig. 1C), confirming that this method can successfully identify genes that are critical for IFN-β restriction of ZIKV. Apart from these controls, 11 other genes fell above the most-enriched nontargeting control sgRNA (NTC) (Fig. 1C and Dataset S1). Of these 11 hits, only one is a previously identified ZIKV restriction factor (IFI6) (22, 23).

IFI6 and AMOTL2 Validate as Antiviral Genes in Out-of-Screen Experiments.

A subset of six of the 11 hits identified in the screen were selected for validation experiments in single-gene KO cells based on both their enrichment score and published data that might support their antiviral potential. More specifically, IFI6 was tested to determine whether its inactivation caused reduced IFN inhibition of ZIKV in our system, as suggested by previous studies (22, 23). The top hit, AMOTL2, was chosen for follow-up based on its large effect size in the screen (Fig. 1C). In addition to AMOTL2 and IFI6, the genes HELZ2, GBP3, and BTN3A1 were also selected for follow-up studies based on their known roles in restricting other viruses (42–45). Finally, the gene RMI2 was included because a prior study showed that IFN-α treatment causes an increase in RMI2 mRNA expression, suggesting RMI2 may play a role in the type I IFN response (46). Inactivation of IRF9 was performed as a positive control, and a nontargeting control (NTC) sgRNA was included as a negative control. For each of these six genes and two controls, single-gene KO cell pools were generated by lentiviral transduction. Knockout was assessed by Synthego’s Inference of CRISPR Edits (ICE) tool, which predicted at least 60% functional KO of each gene (see Fig. 2 legend). The single-gene KO cells were pretreated with or without IFN-β prior to infection with ZIKV, which was quantified as ZIKV RNA in the cell supernatant (47). Viral RNA measurements quantified using this assay correlate closely with levels of infectious virus as measured by TCID₅₀ (SI Appendix, Fig. S2).

Fig. 2.

Validation experiments on selected CRISPR screen hits confirm role of AMOTL2 and IFI6 in type I IFN restriction of ZIKV. (A–G) Each indicated gene KO was tested in ZIKV infection experiments in at least three independent replicates; a representative experiment is shown for each KO condition. Cells were pretreated with or without 1,000 U/mL IFN-β for 24 h, inoculated with ZIKV PRVABC59 MOI = 0.5 for 4 h, and replenished with appropriate media (IFN− closed symbols; or IFN+ open symbols). Cell supernatant was collected at multiple time points throughout a 72-h infection and tested by RT-qPCR for ZIKV RNA levels. (H) Aggregate data for three independent experiments using a single cell pool with each indicated gene KO. ZIKV infection is portrayed relative to the values in IFN-treated or IFN-untreated NTC cells. For IFI6, two different cell pools were used. The KO scores (Synthego ICE analysis) are as follows: IRF9, 77; IFI6, 61 and 95; AMOTL2, 100; HELZ2, 96; GBP3, 88; RMI2, 100; BTN3A1, 100.

IFN-treated IRF9 KO cells showed levels of ZIKV replication identical to those in cells that were not treated with IFN, demonstrating complete ablation of IFN restriction of ZIKV (Fig. 2A). Both AMOTL2 KO and IFI6 KO increased ZIKV infection in the presence of IFN-β, and these differences were evident throughout a 72-h time course of infection (Fig. 2 B and C). At 72 h post infection in IFN-β treated cells, ZIKV infection was increased by an average of 49-fold in IRF9 KO cells, 6.7-fold in IFI6 KO cells, and 3.9-fold in AMOTL2 KO cells relative to IFN-treated NTC cells (Fig. 2H). None of these KOs showed any considerable difference in ZIKV replication from NTC cells in the absence of IFN (Fig. 2 A–C and H). We observed a small effect with HELZ2 KO (Fig. 2 D and H), which increased ZIKV infection by 2.2-fold in IFN-treated cells while having no effect in the absence of IFN-β (Fig. 2H). GBP3, RMI2, and BTN3A1 KO had less than twofold average effects on ZIKV RNA levels with or without IFN-β treatment (Fig. 2 E–H).

AMOTL2 Contributes to IFN-β Restriction of ZIKV.

Because AMOTL2 KO had a consistent effect on ZIKV replication approaching the levels observed with IFI6 KO (Fig. 2), and because it has not previously been described to have an antiviral function, we focused on further characterization of this gene. To increase our confidence that the phenotype observed was due to AMOTL2 KO, we generated two additional AMOTL2 KO cell pools by lentiviral transduction using distinct sgRNA sequences and validated the KOs by ICE and/or Western blot, which revealed >90% KO by genotyping and/or visible protein depletion by Western (Fig. 3A). We also generated AMOTL2 KO cells by nucleofection of Cas9 and the original AMOTL2 sgRNAs and likewise observed visible protein depletion in these cells by Western.

Fig. 3.

AMOTL2 contributes to IFN-β restriction of ZIKV. (A) Western blot showing AMOTL2 protein reduction in three AMOTL2 KO cell pools generated using different sgRNAs. The antibody cross-reacts with an unknown protein that is slightly smaller than AMOTL2. (B) Aggregate data from ZIKV infection experiments with and without IFN-β treatment in IRF9 KO and AMOTL2 KO cells. Data are normalized to NTC cells. Data include the three AMOTL2 KO cell pools generated by lentiviral transduction shown in (A), as well as four replicates of AMOTL2 KO cells generated by Cas9 RNP nucleofection (indicated by diamonds), for a total of 14 biological replicates. P values are Kruskal–Wallis test with Dunn’s multiple comparisons comparing NTC cells to each KO, using separate tests for the IFN− and IFN+ conditions. (C) Western blot showing AMOTL2 protein levels in two AMOTL2 KO cell pools generated using different sgRNAs, as well as in the cells complemented with ORF stuffer or AMOTL2 expression vectors. (D) Representative ZIKV infection experiment with and without IFN-β treatment in AMOTL2 KO cells complemented with AMOTL2. Cells were pretreated with or without 1,000 U/mL IFN-β for 24 h, inoculated with ZIKV PRVABC59 MOI = 0.5 for 4 h, and replenished with appropriate media (IFN− or IFN+). Cell supernatant was collected at multiple time points throughout a 72-h infection and tested by RT-qPCR for ZIKV RNA levels. (E) Aggregate data from four replicates of the experiment shown in (D). Data are normalized to AMOTL2 KO cells complemented with ORF stuffer. P values are Mann–Whitney tests comparing the AMOTL2-complemented cells to the ORF stuffer-complemented cells for each condition (IFN− or IFN+). P values in this figure are denoted by asterisks as follows: ns = >0.05, * = <0.05, ** = <0.01, and **** = <0.0001.

Across 14 independent biological replicates that included these various KO cell pools, IFN-treated AMOTL2 KO cells had an average of 5.7-fold higher ZIKV RNA levels at 72 h post infection compared to IFN-treated NTC cells (P = 0.0027, Fig. 3B). Of note, AMOTL2 KO did not significantly change ZIKV RNA levels in the absence of IFN-β treatment (Fig. 3B). In the presence of IFN-β, the effect of AMOTL2 KO did not reach the effect of IRF9 KO (127-fold increase in ZIKV RNA, P < 0.0001, Fig. 3B), suggesting that AMOTL2 contributes a portion of the IFN restrictive effect against ZIKV.

To further validate this phenotype, we complemented the AMOTL2 KO cells with a stably expressed AMOTL2 expression vector and confirmed reconstitution of AMOTL2 protein by Western blot (Fig. 3C). Relative to complementation with a control open reading frame (ORF stuffer), AMOTL2 complementation resulted in an 8.1-fold decrease in ZIKV infection in the presence of IFN-β at 72 h post infection (P = 0.0286, Fig. 3 D and E). We also observed a very small (1.3-fold average) increase in infection in the AMOTL2 KO cells in the absence of IFN-β (P = 0.0286, Fig. 3 D and E). Overall, the direction of the phenotype observed with AMOTL2 complementation in IFN-treated cells is opposite what we observed with AMOTL2 KO, further supporting a role for AMOTL2 as a protein that contributes to the type I IFN response against ZIKV.

AMOTL2 Is a Constitutively Expressed Gene That Enhances ISG Expression Following IFN-β Treatment.

The single-gene KO studies with AMOTL2 demonstrate that it has an IFN-dependent effect on ZIKV replication, which would be consistent with AMOTL2 being an ISG. To determine whether IFN-β regulates AMOTL2 expression in wildtype A549 cells, we examined our RNA-Seq data from cells treated with IFN-β for 24 h, which is the length of time of IFN-β pretreatment in our infection assays. To our surprise, AMOTL2 transcript levels were unaffected by IFN-β treatment in A549 cells, suggesting that AMOTL2 was not upregulated at the time of infection in this cell type (Fig. 4A). By contrast, upregulation was readily detected for the known ISGs IFIT1, OAS1, and IFITM3, with 25-to-400-fold upregulation observed for these genes in the same experiment (Fig. 4A).

Fig. 4.

AMOTL2 is a constitutively expressed gene that enhances ISG expression following IFN-β treatment. (A) Bulk RNA-Seq data showing transcripts per million of each indicated gene in wildtype A549 cells with no IFN treatment or with 24 h of 1,000 U/mL IFN-β treatment. (B) Upregulation of AMOTL2, IFIT1, GBP1, and Mx1 in NTC cells treated with 1,000 U/mL IFN-β for 9 h. Data points represent three independent experiments. (C) Western blot showing AMOTL2 protein levels in NTC cells with and without 24 h of 1,000 U/mL IFN-β treatment. (D) RT-qPCR data assessing IFIT1 and Mx1 upregulation in response to 3 h of 1,000 U/mL IFN-β treatment of AMOTL2 KO cells, relative to NTC cells. Data points represent independent experiments, and shapes represent different AMOTL2 KO cell pools. P values are Mann–Whitney test (* = <0.05).

Since transcription of many ISGs can occur within hours, we also examined AMOTL2 transcript levels at an earlier timepoint. For this purpose, we performed RT-qPCR on A549 cells treated with IFN-β for 9 h and again saw minimal change in AMOTL2 mRNA levels (average 1.3-fold increase), despite high upregulation of the ISGs IFIT1, GBP1, and Mx1 (700-fold, 76-fold, and 1,150-fold, respectively; Fig. 4B). Additionally, AMOTL2 protein is detected in non-IFN-treated A549 cells, and the protein levels do not change with IFN-β treatment (Fig. 4C). These findings presented a paradox: that AMOTL2’s antiviral phenotype is IFN-dependent, but its expression is not regulated by IFN. We therefore hypothesized that AMOTL2 may be interacting with the IFN signaling pathway itself rather than serving as a downstream effector.

To examine whether AMOTL2 impacts the type I IFN signaling pathway, we measured transcript levels of two ISGs highly upregulated by IFN (IFIT1 and Mx1) in AMOTL2 KO cells and NTC cells with and without IFN-β treatment. In AMOTL2 KO cells, both genes were twofold less upregulated by IFN-β treatment compared to their level of upregulation in NTC cells (P = 0.0286 for IFIT1, P = 0.0286 for Mx1, Fig. 4D). To determine whether other ISGs are also affected, we used NanoString nCounter technology to assess the expression of a panel of innate immune genes. Of the 10 genes on the panel with >10-fold upregulation by IFN-β in NTC cells, all were less upregulated by IFN-β in AMOTL2 KO cells compared to NTC cells (ranging from 1.3- to 2-fold reduction) (SI Appendix, Fig. S3). This result indicates that AMOTL2 KO broadly suppresses upregulation of many ISGs, supporting the hypothesis that AMOTL2 may interact with the type I IFN signaling pathway upstream of ISG transcription to increase ISG expression.

AMOTL2 Increases Phosphorylation and Nuclear Translocation of STAT1 and Reduces Total STAT1 Levels.

We next tested whether AMOTL2 affects specific steps of the IFN signaling pathway, beginning with the phosphorylation of STAT1 and STAT2. We detected a lower (1.7-fold) fraction of phosphorylated STAT1 relative to total STAT1 in IFN-stimulated AMOTL2 KO cells compared to NTC cells (P = 0.0238, Fig. 5A, representative blot in SI Appendix, Fig. S4A). In contrast, there was no difference in phosphorylated STAT2 levels (SI Appendix, Fig. S4 A and B). We also measured nuclear translocation of phosphorylated STAT1 and STAT2, which is a key step leading to ISG transcription (Fig. 5B). Across three independent AMOTL2 KO cell pools, we observed higher levels of phosphorylated STAT1 in the cytoplasmic fraction of AMOTL2 KO cells compared to NTC cells, and a corresponding decrease in the nuclear fraction (Fig. 5 B and C). Accordingly, there was a statistically significant reduction in the nuclear to cytoplasmic ratio (1.8-fold average) of phosphorylated STAT1 in AMOTL2 KO cells following IFN treatment compared to NTC cells (P = 0.0091, Fig. 5C). Interestingly, there was no difference in the nuclear translocation of phosphorylated STAT2 (SI Appendix, Fig. S4 C and D). This suggests that AMOTL2 is associated with increased phosphorylation and nuclear translocation of STAT1.

Fig. 5.

AMOTL2 reduces phosphorylation and nuclear translocation of STAT1 and increases total STAT1 levels. (A) Phosphorylated STAT1 levels relative to total STAT1 levels in whole cell lysates from AMOTL2 KO cell pools stimulated with 1,000 U/mL IFN-β for 1 h, normalized to NTC cells. Data show three independent replicates. (B) Representative western blot on nuclear and cytoplasmic fractions of non-IFN treated and IFN-treated NTC and AMOTL2 KO A549 cells. IFN-β treatment was 1,000 U/mL for 2 h. The total STAT1 antibody is known to favor detection of unphosphorylated STAT1. Lamin A/C is a nuclear marker, and MEK 1/2 is a cytoplasmic marker. (C) Nuclear to cytoplasmic ratio of phosphorylated STAT1 for three replicates of the experiment shown in (B), normalized to NTC cells. (D) Total cytoplasmic STAT1 levels for three replicates of the experiment shown in (B), normalized to NTC cells. (E) STAT1 transcripts in non-IFN-stimulated AMOTL2 KO cells, normalized to NTC cells. P values in this figure are denoted by asterisks as follows: * = <0.05, ** = <0.01, and **** = <0.0001.

Using a STAT1 antibody that preferentially detects unphosphorylated STAT1 (U-STAT1; Cell Signaling 9176), we detected 2.4-fold higher levels of total U-STAT1 in the cytoplasm of AMOTL2 KO cells compared to NTC cells, and this difference existed even before IFN treatment (P = 0.0091 both with and without IFN treatment, Fig. 5 B and D). Conversely, there was no difference in total STAT2 levels (SI Appendix, Fig. S4 C and E). To determine whether STAT1 transcripts were also elevated in the AMOTL2 KO cells at baseline, we measured STAT1 mRNA in AMOTL2 KO cells without IFN treatment and found 1.6-fold higher levels compared to NTC cells (P < 0.0001, Fig. 5E). Finally, to determine whether AMOTL2 KO had any effect on STAT1 and STAT2 binding, we performed coimmunoprecipitations of endogenously expressed STAT1 and STAT2 in AMOTL2 KO and NTC cells. We found no difference in STAT1/2 binding in the AMOTL2 KO cells compared to NTC cells (SI Appendix, Fig. S4 F and G). Taken together, these results suggest that AMOTL2 is associated with reduced basal STAT1 expression and increased activation of STAT1 upon IFN-β treatment and thus may affect IFN signaling by modulating STAT1 levels and activation.

Identification of AMOTL2 Interacting Proteins and Critical Domain for AMOTL2 Antiviral Function.

AMOTL2 is known to interact with YAP and TAZ (48, 49), transcriptional regulators that have been reported to antagonize the innate immune response through several mechanisms (50). We hypothesized that AMOTL2 may exert its antiviral function via this interaction with YAP and/or TAZ. To test whether YAP and TAZ affect the type I IFN response against ZIKV in A549 cells, we generated YAP and TAZ KO A549 cells and validated the KOs by Synthego ICE and/or Western blot, which revealed 100% KO by genotyping and/or visible protein depletion by Western. Neither gene KO affected ZIKV levels in the absence of IFN treatment (Fig. 6 A–C). In the presence of IFN, YAP KO had little if any appreciable effect (1.5-fold reduction in ZIKV infection at 72 h post infection, not significant, Fig. 6 A and C), while TAZ KO caused an 8.7-fold average reduction in infection at 72 h (P = 0.0011, Fig. 6 B and C), in addition to causing a consistent reduction at all earlier time points tested (Fig. 6B). We therefore focused our attention on TAZ.

Fig. 6.

Identification of AMOTL2 interacting proteins and critical domain for AMOTL2 antiviral function. (A and B) Representative ZIKV infection experiments with and without IFN-β treatment in YAP KO (A) and TAZ KO (B) cells. Cells were pretreated with or without 1,000 U/mL IFN-β for 24 h, inoculated with ZIKV PRVABC59 MOI = 0.5 for 4 h, and replenished with appropriate media (IFN− or IFN+). Cell supernatant was collected at multiple time points throughout a 72-h infection and tested by RT-qPCR for ZIKV RNA levels. (C) Aggregate data from at least three independent replicates of the experiments shown in (A) and (B). YAP KO cells were generated by lentiviral transduction, and TAZ KO cells were generated by lentiviral transduction or Cas9 RNP nucleofection (the latter indicated by diamonds). Data are normalized to NTC KO cells. P values are Kruskal–Wallis with Dunn’s multiple comparisons comparing NTC KO cells to each KO, with separate tests performed for IFN− and IFN+ conditions. (D) Coimmunoprecipitation assay using A549 cells treated with IFN-β for 0 h, 5 h, 18 h, or 24 h and analyzed by Western blot. An anti-TAZ antibody was used as the capture antibody. (E) Western blot showing AMOTL2 and STAT1 levels in AMOTL2 KO cells complemented with AMOTL2 expression vectors or ORF stuffer. (F) AMOTL2 transcripts in the panel of AMOTL2-complemented AMOTL2 KO cells, relative to AMOTL2 transcript expression in ORF stuffer-complemented AMOTL2 KO cells. (G and H) Representative ZIKV infection experiments with and without IFN-β treatment in AMOTL2 KO cells complemented with AMOTL2 Y213A (G) and AMOTL2 CCdel (H), compared to cells complemented with ORF stuffer and cells complemented with wildtype AMOTL2. Cells were pretreated with or without 1,000 U/mL IFN-β for 24 h, inoculated with ZIKV PRVABC59 MOI = 0.5 for 4 h, and replenished with appropriate media (IFN− or IFN+). Cell supernatant was collected at multiple time points throughout a 72-h infection and tested by RT-qPCR for ZIKV RNA levels. (I) Aggregate data from four independent replicates of the experiments represented in (G and H) for all AMOTL2 expression constructs. Shapes represent different AMOTL2 KO cells that were transduced with the expression vectors. Data are normalized to ORF stuffer-complemented AMOTL2 KO cells. P values are Kruskal–Wallis with uncorrected Dunn’s test comparing ORF stuffer cells to each AMOTL2-complemented cell pool, with separate tests performed for IFN− and IFN+ conditions. P values in this figure are denoted by asterisks as follows: ns = >0.05, * = <0.05, ** = <0.01, and *** = <0.001.

Because the interaction between AMOTL2 and TAZ has not been examined in A549 cells, we tested the interaction between the endogenous proteins by coimmunoprecipitation in cells either untreated or treated with IFN-β for various lengths of time. Similar levels of AMOTL2 were detected in immunoprecipitants using TAZ as bait under all conditions in wildtype cells, suggesting that there is binding between TAZ and AMOTL2 in A549 cells, and it is independent of IFN-β treatment (Fig. 6D).

To determine whether TAZ affects STAT1 levels and activation of pSTAT1, we measured nuclear accumulation of pSTAT1 and total STAT1 levels in TAZ KO cells. We did not observe any detectable effect on the nuclear to cytoplasmic ratio of pSTAT1 or on STAT1 protein levels in the presence or absence of IFN in TAZ KO cells compared to NTC cells (SI Appendix, Fig. S5).

To further verify that TAZ is not involved in AMOTL2’s antiviral phenotype, we tested whether TAZ binding is required for AMOTL2’s effect using an AMOTL2 expression vector containing a mutation in the PPQY motif that mediates TAZ binding (48) (AMOTL2 Y213A). We complemented AMOTL2 KO cells with AMOTL2 Y213A, wildtype AMOTL2, or a control ORF stuffer coding sequence. At the same time, we tested several other AMOTL2 mutation or deletion constructs that disrupt various functional domains of the protein. These included AMOTL2 P105A [disrupting AMOTL2’s LPTY motif, responsible for AMOTL2’s interaction with MAGI-1 and actin (51, 52)], AMOTL2 CCdel [deleting AMOTL2’s coiled-coil domain, responsible for AMOTL2 homo-oligomerization and localization to tight junctions (51)], AMOTL2 PDZdel [deleting AMOTL2’s C-terminal PDZ-binding motif, involved in cell polarity and migration (53)], and S159A [removing a serine that is phosphorylated by LATS1/2 (37)]. AMOTL2 protein levels were markedly increased relative to ORF stuffer-complemented AMOTL2 KO cells in all the expression constructs with the exception of CCdel (a deletion of the amino acids 314-572) (Fig. 6E). This is likely because the AMOTL2 antibody was raised against an AMOTL2 recombinant protein spanning a region encompassed within the coiled-coil domain (amino acids 401 to 480). STAT1 levels appeared reduced relative to ORF stuffer-complemented AMOTL2 KO cells in all the expression constructs with the exception of CCdel (Fig. 6E). To confirm that the AMOTL2 CCdel cells were expressing AMOTL2, we measured AMOTL2 mRNA levels by RT-qPCR using primers that bind all the AMOTL2 expression constructs, including CCdel. All the AMOTL2-complemented cells including CCdel exhibited ~10-fold higher AMOTL2 mRNA levels relative to the ORF stuffer-complemented cells (Fig. 6F), suggesting that the CCdel-complemented cells were indeed expressing the transduced form of AMOTL2.

Compared to ORF stuffer, complementation with either wildtype AMOTL2 or AMOTL2 Y213A reduced ZIKV replication measured at 48 hpi and 72 hpi in the presence, but not absence, of IFN-β (Fig. 6 G and I). This difference was statistically significant in the IFN-treated cells at 72 hpi in four independent replicates (Fig. 6I, 8.1-fold reduction in ZIKV infection for wildtype, 9.1-fold reduction in ZIKV infection for Y213A, P < 0.05 for both). By contrast, complementation with AMOTL2 CCdel had no effect on ZIKV infection compared to ORF stuffer (Fig. 6 H and I; no significant difference across four replicates in the presence or absence of IFN). Complementation with the other AMOTL2 expression variants (AMOTL2 P105A, AMOTL2 PDZdel, and AMOTL2 S159A) decreased ZIKV infection relative to ORF stuffer by 6.8- to 9.4-fold in IFN-treated cells, similar to the effect of complementation with wildtype AMOTL2 (Fig. 6I, P < 0.05 for all). These results suggest that TAZ binding is not required for AMOTL2’s antiviral activity, and that AMOTL2’s antiviral mechanism does not likely act via TAZ. Furthermore, they indicate that AMOTL2’s coiled-coil domain may be essential for its antiviral phenotype.

AMOTL2 Contributes to IFN-β Restriction of SINV.

Given our data suggesting that AMOTL2 interacts with the IFN signaling pathway itself to exert its antiviral effect, we hypothesized that it has broad antiviral activity. To test this, we infected the AMOTL2 KO cells with an IFN-sensitive TE12 strain (54) of the alphavirus Sindbis virus (SINV) and measured infected cells by flow cytometry. Though the magnitude of the effect was variable, AMOTL2 KO consistently resulted in increased SINV infection relative to NTC cells in the IFN-treated condition, with a median effect size of 13.3 (a 16.3-fold median increase for AMOTL2 KO #1 and 10.4-fold median increase for AMOTL2 KO #2; SI Appendix, Fig. S6). Similar to our results with ZIKV, AMOTL2 KO did not affect SINV infection levels in the non-IFN-treated condition (<twofold effect in all experimental replicates).

Discussion

In this study, we present a CRISPR screening platform that uses virus-induced cell death to identify genes contributing to type I IFN restriction of ZIKV. Using the gene inactivation approach allowed us to demonstrate that IFI6 KO reduces type I IFN restriction of ZIKV in A549 cells, reinforcing results of prior overexpression and knockdown studies in other cell types and strengthening the evidence for IFI6 as a major anti-flavivirus ISG. Importantly, we uncovered an antiviral host gene, AMOTL2, that is not an ISG in A549 cells but nonetheless inhibits ZIKV in an IFN-dependent manner. Our studies suggest that AMOTL2 acts by regulating STAT1 activation and downstream ISG expression.

Prior screens for ZIKV restriction factors include a knockdown screen using a limited siRNA library of 386 genes (36) and two other studies that used genome-wide overexpression or CRISPR activation libraries (23, 55). These screens led to the identification of several genes with antiviral activity against ZIKV, including IFI6 and TMEM120A. While these latter studies are useful starting points, relying on gene overexpression or activation can sometimes reveal a phenotype that is not present at a gene’s endogenous expression levels, as has been demonstrated for IFITM3 (14). Our screen results validated IFI6, a known ISG and restriction factor that blocks flavivirus replication by preventing the formation of the viral replication organelles in the endoplasmic reticulum membrane (22, 23). In our validation experiments, we also observed a small effect on IFN restriction with HELZ2 KO, though it was modest in comparison to IFI6. HELZ2 is an ISG that has been reported to restrict two other flaviviruses, dengue virus and hepatitis C virus (42, 43). Given the small impact of HELZ2 inactivation on ZIKV observed here, additional work is required to determine whether HELZ2 is a true ZIKV restriction factor.

In addition to containing sgRNAs targeting most ISGs in A549 cells, our CRISPR library also targeted more than 1,700 other human genes, affording us the opportunity to identify non-ISGs that are important in the type I IFN response. Indeed, the top experimental hit in our screen was AMOTL2, a gene that we show is not regulated by IFN-β in A549 cells yet contributes to IFN-mediated restriction of ZIKV. AMOTL2 is a motin family protein of 780 amino acids with known functions in cancer pathogenesis, angiogenesis, tight junction formation, cell polarity, the actin cytoskeleton, and regulation of Hippo signaling (37, 56). We confirmed the role of AMOTL2 in the IFN response against ZIKV in validation experiments using several distinct AMOTL2 KO cell pools targeting different CRISPR cut sites. We also demonstrated that reconstituting AMOTL2 expression can rescue IFN restriction of ZIKV in the AMOTL2 KO cells. Together, these findings confirm that AMOTL2 has an antiviral effect against ZIKV.

We found that AMOTL2 knockout results in broad blunting of ISG transcription following IFN-β stimulation. In follow-up mechanistic studies, we identified specific defects in the IFN signaling pathway in AMOTL2 knockout cells that could explain this effect, including reduced STAT1 phosphorylation and reduced nuclear translocation of phosphorylated STAT1, with no effect on STAT2. Interestingly, levels of total, unphosphorylated STAT1 (U-STAT1) were increased in AMOTL2 KO cells, even without IFN-β stimulation. U-STAT1 and unphosphorylated STAT2 are known to play an important role in maintaining the antiviral state triggered by type I IFN after levels of phosphorylated STAT1 and STAT2 decline. Specifically, the unphosphorylated proteins can interact with IRF9 and mediate expression of some ISGs, thus prolonging the antiviral state for days after resolution of the acute response mediated by phosphorylated STAT1 and STAT2 (57, 58). However, paradoxically, resting high levels of U-STAT1 have also been reported to reduce the responsiveness of cells to activation by type I IFN treatment (59, 60).

We therefore propose a model in which AMOTL2 suppresses high levels of basal STAT1, relieving the block on IFN-induced STAT1 phosphorylation and nuclear translocation (Fig. 7). When endogenous AMOTL2 is being expressed, creating an AMOTL2-high environment, U-STAT1 levels are attenuated, and STAT1 can be potently activated by type I IFN treatment, resulting in high ISG transcription and an antiviral state that better controls ZIKV replication. In an AMOTL2-low environment, such as the AMOTL2 KO condition, levels of basal U-STAT1 are high, and STAT1 activation by type I IFN treatment is low. ISG transcription is reduced, and the cell is less able to control ZIKV replication. We also found that STAT1 mRNA levels are higher in AMOTL2 KO cells. While this could suggest that AMOTL2 KO increases STAT1 transcription, it is also possible that AMOTL2 KO increases STAT1 protein levels and this feeds forward to drive more STAT1 transcription, as overexpression of exogenous U-STAT1 has been shown to stimulate transcription of STAT1 in other cell types (58).

Fig. 7.

Working model of the impact of AMOTL2 on the type I IFN signaling pathway. (A) With endogenous expression levels of AMOTL2, basal unphosphorylated STAT1 levels are low, and STAT1 activation occurs normally following IFN-β treatment. ISG transcription is high, and ZIKV replication is better controlled. (B) AMOTL2 knockout results in high basal levels of unphosphorylated STAT1. Upon IFN treatment, the cell exhibits reduced STAT1 phosphorylation and nuclear translocation and low ISG transcription, consistent with increased ZIKV replication.

Previous studies hint at several possible mechanisms for how high levels of U-STAT1 may suppress the phosphorylation and nuclear translocation of pSTAT1. It is known that chronic IFN exposure can reduce responsiveness to repeated IFN stimulation (59, 61). In these systems, levels of U-STAT1 tend to be constitutively elevated, similar to the environment in our AMOTL2 KO cells. U-STAT1 may contribute to this IFN-refractory environment by driving the expression of negative regulators of the IFN signaling pathway, such as tyrosine phosphatases targeting pSTAT1 (62). Another study showed that U-STAT1 inhibits the nuclear accumulation of pSTAT1 through interactions with importin-α (63). This literature offers plausible explanations for the association we observed between high U-STAT1 and reduced phosphorylation and nuclear trafficking of pSTAT1.

AMOTL2 is known to interact with TAZ and YAP, transcriptional coactivators that stimulate the upregulation of genes involved in diverse biological processes, including cell proliferation, stemness, and epithelial-to-mesenchymal transition (64). In this context, AMOTL2 can bind and inhibit TAZ and YAP from performing their canonical role (48, 49). Interestingly, TAZ and YAP also antagonize the innate immune response through a variety of mechanisms, including by antagonizing the type I IFN signaling pathway (50, 65). We therefore reasoned that TAZ and/or YAP may be involved in AMOTL2’s phenotype, given their tendency to interrupt innate immune signaling pathways (50), and AMOTL2’s tendency to bind and block TAZ and YAP.

We found that inactivation of TAZ, but not YAP, reduced ZIKV infection in IFN-treated cells. This suggests that TAZ is a proviral gene that antagonizes IFN restriction of ZIKV, a phenotype that may be due to known negative regulation by TAZ on the type I IFN signaling pathway (65). This study presents evidence that TAZ has proviral function for ZIKV. We also demonstrated that TAZ and AMOTL2 interact in A549 cells in an IFN-independent manner. However, TAZ KO had no appreciable effect on nuclear localization of pSTAT1, suggesting TAZ is not involved in AMOTL2’s antiviral phenotype. To definitively test whether TAZ binding is involved in AMOTL2’s antiviral mechanism, we complemented the AMOTL2 KO cells with a panel of AMOTL2 variants with mutations in various functional domains, including in the TAZ-binding PPQY motif. We showed that AMOTL2’s antiviral phenotype was rescued with wildtype AMOTL2 to a similar extent as it was rescued with the PPQY-mutated version, suggesting that TAZ binding is not involved in AMOTL2’s mechanism. On the other hand, an AMOTL2 mutant with a deletion of the coiled-coil domain failed to rescue the phenotype, suggesting that this domain is required for AMOTL2’s antiviral function. Coiled-coil domains frequently mediate protein–protein interactions, and AMOTL2’s coiled-coil domain is known to mediate its homo-oligomerization (51). Therefore, our studies implicate either AMOTL2 homo-oligomerization or interaction with another coiled-coil domain-containing protein in AMOTL2’s antiviral phenotype. This finding provides a direction for future detailed studies of the mechanism and cellular partners contributing to AMOTL2’s ability to modulate STAT1 levels and enhance IFN signaling, which may have larger implications in defining regulation of the type I IFN response.

AMOTL2 is in the motin family of angiostatin binding proteins with angiomotin-like protein (AMOT) and angiomotin-like protein 1 (AMOTL1). Interestingly, AMOT and AMOTL1 have reported proviral roles for some viruses unrelated to modulation of IFN signaling (66–69). While these proteins are considerably divergent, their coiled-coil domains represent the region of highest identity (56%) (70); thus, our findings implicating the coiled-coil domain in AMOTL2’s activity raise the question of whether AMOT and AMOTL1 can also enhance responsiveness to IFN signaling via their coiled-coil domains.

Based on AMOTL2’s impact on IFN-stimulated STAT1 activation, we hypothesized that its antiviral activity is not limited to ZIKV. We explored this idea in experiments with an IFN-sensitive strain of the alphavirus SINV (54) and found that AMOTL2 KO indeed increased SINV infection in IFN-treated cells. We therefore propose that AMOTL2 is likely a broadly antiviral gene—a concept that can be further examined with other viral infections in AMOTL2 KO cells.

In this study, we uncovered a role for AMOTL2, a gene with no previously described function in innate immunity, in IFN-mediated restriction of ZIKV. AMOTL2 mRNA is broadly detectable in primary human tissue, including in the placenta, a site highly relevant to ZIKV infection in vivo (The Human Protein Atlas). Additionally, based on its mechanism and our results with SINV, AMOTL2 appears to be a broadly antiviral gene. Future studies are needed to better define the mechanism and breadth of AMOTL2’s antiviral function across viruses and in other model systems. This work improves our basic understanding of type I IFN-mediated protection against viruses.

Materials and Methods

Cell Culture and Reagents.

Human lung epithelial A549 cells (A. Berger; ATCC CCL-185) were cultured in Gibco RPMI 1640 (ThermoFisher 22400089) supplemented with 10% fetal bovine serum (FBS), 2 mM l-glutamine, and 1× Gibco antibiotic/antimycotic (ThermoFisher 15240062). Vero cells and HEK293T cells were maintained in Gibco DMEM (ThermoFisher 11965092) supplemented with 10% FBS, 2 mM l-glutamine, and 1× antibiotic/antimycotic. The identity of the A549 cells and HEK293T cells was confirmed using STR CODIS fingerprinting prior to beginning experiments.

The A549 cells were regularly tested for Mycoplasma and found to be Mycoplasma-free by the Research Cell Bank shared resource at the Fred Hutchinson Cancer Center.

Vero cells (A. Geballe; ATCC) and HEK293T cells were maintained in DMEM supplemented with 10% FBS, 2 mM l-glutamine, and 1× antibiotic/antimycotic.

Viruses.

ZIKV strain PRVABC59 was kindly provided by BEI Resources. ZIKV PRVABC59 was propagated in Vero cells at an MOI of 0.01, as previously described (71). Viral titers were determined by the TCID₅₀ assay described in SI Appendix, Supplemental Methods.

SINV TE3’2J:mCherry (TE12 with nsP1 T538I mutation) was generated and viral stocks quantified as described previously (72) and in SI Appendix, Supplemental Methods.

Plasmids.

pMD2.G (Addgene #12259, vesicular stomatitis virus glycoprotein [VSV-G] envelope plasmid) and psPAX2 (Addgene #12260, an HIV-based packaging plasmid) were gifts from Didier Trono (Ecole Polytechnique Federale de Lausanne, Switzerland). lentiCRISPR v2 was a gift from Feng Zhang (Addgene #52961) (Broad Institute of MIT and Harvard, Cambridge, MA) (73). The expression plasmids encoding AMOTL2 (NCBI reference sequence NM_016201.4) and ORF stuffer (amino acids 2 to 83 of Escherichia coli beta-galactosidase) were synthesized by VectorBuilder (Vector IDs VB240108-1277gwz and VB230727-1905uxr). Whole plasmid sequencing was performed by Plasmidsaurus using Oxford Nanopore Technology for all plasmid preps used in this study. Plasmids generated in this study are available upon request.

Generation of A549 KO Library Cells.

The PIKA CRISPR library was designed as previously described (38), with details in SI Appendix, Supplemental Methods. Pseudoviruses for lentiviral transduction of the PIKA CRISPR library were made by transfection of 293T cells, as previously described (38), with the following exception: 5 × 10⁵ 293 T cells were plated in 2 mL in six-well plates and the following day were transfected with 667 ng lentiCRISPRv2 plasmid (encoding puromycin resistance), 500 ng psPAX2, and 162.5 ng pMD2.G. The comprehensive lentiviral CRISPR library was titered by colony-forming unit assay on A549 cells and used to transduce A549 cells at an MOI of 0.5. Cells were selected in 1 μg/mL puromycin (Sigma Aldrich P9620) for 14 d. Following selection, aliquots of 1 × 10⁷ A549 KO library cells/mL (constituting ≥500-fold representation of each sgRNA) were stored in liquid nitrogen until CRISPR screening. Cells were maintained at ≥500-fold representation of each sgRNA to avoid bottlenecking the library.

ZIKV Quantification by TCID₅₀.

ZIKV titers were determined by TCID₅₀ assay on Vero cells as previously described in ref. 14 and SI Appendix, Supplemental Methods.

CRISPR Screen.

To perform the CRISPR screen, 2.5 × 10⁷ A549 KO library cells were plated in RPMI complete containing 1,000 U/mL IFN-β (PBL Assay Sciences). Twenty-four hours later, cells were infected with ZIKV (PRVABC59) at MOI = 3 for 4 h. For the uninfected control condition, the method was identical except the cells received serum-free RPMI without virus added. The inoculum was removed and replaced with fresh RPMI complete containing 1,000 U/mL IFNβ. At 42 h post infection, the supernatant was collected. A PBS wash was performed to capture any remaining dead/dying cells not adhered to the flask, and the PBS was collected and added to the cell supernatant. The supernatant was spun to collect cells, and genomic DNA was isolated using a Qiagen QIAamp DNA blood mini kit. Illumina library preparation was performed by two rounds of PCR using Herculase II Fusion DNA Polymerase (Agilent 600679). The primers, thermocycling conditions, product purification, and library preparation are described in SI Appendix, Supplemental Methods. Sequencing was performed on an Illumina HiSeq (Fred Hutch Genomics Shared Resource).

CRISPR Screen Analysis.

Deep sequencing reads were demultiplexed, trimmed, and aligned to the sgRNA library using Bowtie v1.0.0 (74). The aligner’s native –trim5 and –trim3 options were used to trim down the raw reads to 20 nts, which is the expected guide size. In an uninfected, IFN-treated control condition used to assess sgRNA coverage of the A549 KO library cells at baseline and performed in parallel to the experimental conditions of the screen, 97.6% of genes had at least six of their eight sgRNAs sequenced at high representation. This ensured to us that the setup of the screen allowed for high representation of the sgRNAs in the PIKA library. For screen analysis, relative enrichment of sgRNAs (based on all eight sgRNAs targeting each gene) and genes were analyzed and compared between groups of interest using the Model-based Analysis of Genome wide CRISPR/Cas9 Knockout (MAGeCK) screens algorithm, v0.5.9.2 (75, 76). Statistical scores were assigned to identify the most enriched genes, indicating factors important for IFN-mediated restriction of ZIKV infection.

Bulk RNA-Seq in A549 Cells.

Wildtype A549 cells were plated in triplicate in media with or without 1,000 U/mL IFN-β. 24 h later, cells were harvested and total RNA was extracted with the RNeasy Plus Mini Kit (Qiagen 74134). RNA quality was assessed with RNA TapeStation (Agilent), and RNA-Seq was performed with 50 bp paired-end reads on an Illumina machine (Fred Hutch Genomics Shared Resource). RNA-Seq data analysis is described in SI Appendix, Supplemental Methods.

Generation of Single-Gene KOs in A549 Cells.

Single-gene KOs were made by lentiviral transduction for the genes IRF9, IFI6, AMOTL2, GBP3, RMI2, HELZ2, BTN3A1, WWTR1 (TAZ), and YAP. Individual sgRNAs (Dataset S2) were cloned into the pLentiCRISPRv2 plasmid. Nontargeting sgRNA sequences were used to generate NTC cells in parallel. 2 × 10⁶ 293T cells were plated in 10 mL in T-75 flasks and the following day were transfected with 5 μg lentiCRISPRv2 plasmid with the appropriate sgRNA sequence, 5 μg psPAX2, and 2.5 μg pMD2.G. Viral supernatants were collected 2 d later, clarified by centrifugation to remove any cells in solution, and concentrated using 100 kDa MWCO Amicon filter units (MilliPore Sigma UFC910096) until ~250 μL of concentrated virus remained. A549 cells were plated in six-well dishes at 2 × 10⁵ cells per well in 2 mL of RPMI. The following day, cells were transduced with the concentrated pseudotyped lentivirus by spinoculation. In cases where single-gene KOs were generated with more than one sgRNA (Dataset S2), the cells were transduced with the lentiviruses for both sgRNAs. Successfully transduced cells were selected by culturing in RPMI supplemented with 2 μg/mL puromycin for 1 to 2 wk or until flasks with nontransduced control cells were killed by puromycin.

For KOs generated by nucleofection of Cas9/sgRNA ribonucleoproteins (RNPs), the same two sgRNAs per gene as used in dual-guide lentiviral editing experiments were ordered as synthetic sgRNAs (Synthego CRISPRevolution sgRNA EZ Kit, modified with 2′-O-methyl at 3 first and last bases and 3′ phosphorothioate bonds between first 3 and last 2 bases). RNPs were assembled on the day of nucleofection by combining 4 μL of 50 pmol/μL pooled sgRNAs (Synthego), 20 μL of SF Cell Line Complete Nucleofector solution (Lonza V4XC-2032), and 1 μL of 20 μM Cas9-NLS protein (UC Berkeley QB3 MacroLab). Complexes were gently mixed and incubated for at least 10 min at room temperature. 2 × 10⁵ A549 cells were pelleted and washed once with PBS prior to resuspension in 25 μL RNPs and immediate transfer to a 16-well Nucleocuvette Strip in an Amaxa Nucleofector (Lonza). Nucleofection was performed using pulse code CM-130. Cells were immediately supplemented with 80 μL of RPMI complete media and allowed to recover in the cuvette for at least 10 min at 37 °C. Cells were then transferred to six-well dishes and supplemented with an additional 3 mLs of RPMI complete media. Since cells generated by RNP nucleofection do not have a selectable marker, we performed fresh nucleofections for each biological replicate to minimize outgrowth of wildtype cells.

All KOs were verified by Synthego’s ICE genomic editing analysis (described in SI Appendix, Supplemental Methods) and/or Western blotting.

IFN Treatment and ZIKV Infection of A549 Cells.

A549 KO cells were diluted to 8 × 10⁴ cells/mL in RPMI complete media, and 1 mL of cells was plated for each IFN-negative well. IFN-β (PBL Assay sciences 11410) was added to the remaining cells at 1,000 U/mL, and 1 mL of these cells was plated as each IFN-treated well. All conditions were performed in duplicate. Twenty-four hours later, cells were infected with 250 μL PRVABC59 ZIKV at an MOI of 0.5 in serum-free RPMI. Four hours later, the viral inoculum was removed and cells were washed once with PBS, then replenished with fresh IFN-negative or IFN-treated RPMI complete media. For quantification of infectious virus by TCID₅₀, all viral supernatant was collected at the infection endpoint and frozen at −80 °C. For quantification of ZIKV RNA by RT-qPCR, 12 μL of viral supernatant was removed from each well at various time points throughout the infection time course. The viral supernatant was heat-inactivated at 60 °C for 30 min and stored at −80 °C.

ZIKV Quantification by RT-qPCR.

Infected cell supernatant was used as the template in an RT-qPCR assay using the iTaq Universal Probes One-Step kit (BioRad). Forward and reverse primers and probe targeted the NS1 region of the ZIKV PRVABC59 genome (Dataset S2) (adapted from ref. 47). More detailed methods, including thermocycling conditions, are described in SI Appendix, Supplementary Methods and Dataset S2. ZIKV RNA copies per mL were quantified using a standard curve of a positive control DNA fragment (described in SI Appendix, Supplementary Methods).

Human Gene Expression Quantification.

For assessment of gene expression by RT-qPCR, details of each experimental setup are described in SI Appendix, Supplemental Methods. In all cases, cells were lysed in buffer RLT+ (Qiagen) with beta-mercaptoethanol and stored at −80 °C until RNA extraction. RNA was extracted using the RNeasy Plus Mini Kit (Qiagen 74134), cDNA was synthesized with SuperScript III reverse transcriptase (Thermo Fisher 18080093), and qPCR was performed using the iTaq universal SYBR Green mix (Bio-Rad 1725120); further details and slight modifications are described in SI Appendix, Supplemental Methods. Plates were read on an Applied Biosystems QuantStudio 7 Real-Time PCR machine, and relative quantification was calculated using the ΔΔCt method (77).

Assessment of ISG expression by Nanostring nCounter is described in SI Appendix, Supplemental Methods.

AMOTL2 Complementation.

Lentiviral expression plasmids encoding AMOTL2 and ORF stuffer were synthesized by VectorBuilder. To allow for successful complementation of AMOTL2 in the lentiCRISPR AMOTL2 KO cells, the lentiviral expression plasmid encoding wildtype AMOTL2 was edited by site-directed mutagenesis (SDM) to mutate the sgRNA sequence and/or PAM sites for each sgRNA expressed by the AMOTL2 KO cells. An additional mutation was made to mutate the reference sequence’s glutamic acid at position 732 to an aspartic acid, as we noticed the A549 cells express an aspartic acid at this position according to RNA-Seq reads. Then, individual AMOTL2 point (P105A, Y213A, S159A) and deletion (CCdel, PDZdel) mutants were synthesized by SDM as described in SI Appendix, Supplemental Methods.

Complementation of AMOTL2 KO cells with AMOTL2 was performed as described above in “Generation of single gene KOs in A549 cells,” with the following exceptions: Lentiviral expression plasmids were used instead of lentiCRISPRv2 plasmid, and hygromycin B (InvivoGen ant-hg-1) was used as the selectable marker. AMOTL2 protein expression was confirmed by RT-qPCR and Western blot.

Coimmunoprecipitation.

A549 cells with the indicated gene KOs and treatments were harvested at the indicated time points and lysed in Pierce IP buffer (87787) containing protease inhibitor (Thermo Fisher A32955) or protease/phosphatase inhibitor (Cell Signaling 5872). Samples were flash-frozen and stored at −80 °C until further processing. Upon thawing, lysates were sonicated, then spun at maximum speed for 15 min at +4 °C. Protein concentrations were quantified using the DC Protein Assay (Bio-Rad 5000116) against a standard curve of bovine serum albumin. Samples were diluted to equivalent concentrations and volumes. Lysates were precleared by incubating with protein G beads at +4 °C for at least 30 min. Meanwhile, protein G beads were incubated with the indicated IP antibody at +4 °C for at least 30 min. The antibodies used for IP were anti-TAZ (mouse IgG2a monoclonal antibody, Cell Signaling 71192), anti-STAT1 (mouse IgG1 monoclonal antibody, Cell Signaling 9176), anti-phospho-STAT1 Tyr701 (mouse IgG2a monoclonal antibody, Thermo Fisher 33-3400), mouse IgG2a isotype control antibody (Cell Signaling 61656), and mouse IgG1 isotype control antibody (Cell Signaling 5415). Precleared whole cell lysates were then incubated with the antibody-beads overnight at +4 °C. The beads were washed thrice in lysis buffer before eluting off the bound proteins by resuspending beads in LDS sample buffer (Invitrogen NP007) containing 50 mM DTT and heating to 70 °C for 10 min. Immunoblotting was performed as described below.

Nuclear Fractionation.

Confluent T-75 flasks of A549 NTC and AMOTL2 KO cells were treated with 1,000 U/mL IFN-β for 2 h or left untreated. Nuclear fractionation was performed using the NuCLEAR™ Extraction Kit (Sigma-Aldrich) according to the manufacturer’s protocol, with the following changes: Cells were trypsinized instead of scraped, and the nuclear pellet was washed twice in PBS to remove residual cytoplasmic proteins prior to nuclear extraction. Cytoplasmic and nuclear extracts were quantified and processed for immunoblotting as described below.

IFN Treatment and Sindbis Virus Infection of A549 Cells.

Sindbis virus experiments are described in SI Appendix, Supplemental Methods.

Immunoblot Analysis.

Immunoblots were performed and analyzed using standard methods described in SI Appendix, Supplemental Methods. All primary and secondary antibodies are listed in SI Appendix, Supplemental Methods.

Illustrations, Data Display, and Analysis.

Figures were generated using BioRender and eulerr.co. Prism 10 was used for statistical tests and display of quantitative data. Digital images were processed with ImageJ.

Supplementary Material

Appendix 01 (PDF)

Dataset S01 (XLSX)

Dataset S02 (XLSX)

Data, Materials, and Software Availability

Illumina sequencing reads from bulk RNA-Seq and CRISPR screen sequencing can be accessed via the following SRA accession numbers, respectively: PRJNA1238368 (78) and PRJNA1238487 (79). Any additional information required to reanalyze the data reported in this paper is available upon request.