key: cord-0775578-xyyhup28
authors: Soveg, Frank W.; Schwerk, Johannes; Gokhale, Nandan S.; Cerosaletti, Karen; Smith, Julian R.; Pairo-Castineira, Erola; Kell, Alison M.; Forero, Adriana; Zaver, Shivam A.; Esser-Nobis, Katharina; Roby, Justin A.; Hsiang, Tien-Ying; Ozarkar, Snehal; Clingan, Jonathan M.; McAnarney, Eileen T.; Stone, Amy E. L.; Malhotra, Uma; Speake, Cate; Perez, Joseph; Balu, Chiraag; Allenspach, Eric J.; Hyde, Jennifer L.; Menachery, Vineet D.; Sarkar, Saumendra N.; Woodward, Joshua J.; Stetson, Daniel B.; Baillie, J. Kenneth; Buckner, Jane H.; Gale, Michael; Savan, Ram
title: Endomembrane targeting of human OAS1 p46 augments antiviral activity
date: 2021-04-22
journal: bioRxiv
DOI: 10.1101/2021.04.21.440697
sha: ff685d535310044b2c06e9ca9ac71a1cb96f7ed5
doc_id: 775578
cord_uid: xyyhup28

Many host RNA sensors are positioned in the cytosol to detect viral RNA during infection. However, most positive-strand RNA viruses replicate within a modified organelle co-opted from intracellular membranes of the endomembrane system, which shields viral products from host cell innate immune sensors. Targeting innate RNA sensors to the endomembrane system may enhance their ability to sense viral RNA generated by viruses that use these compartments for replication. Here, we reveal that an isoform of oligoadenylate synthetase 1, OAS1 p46, is prenylated and targeted to the endomembrane system. Membrane localization of OAS1 p46 confers enhanced access to viral replication sites and results in increased antiviral activity against a subset of RNA viruses including flavivirus, picornavirus, and SARS-CoV-2. Finally, our human genetic analysis shows that the OAS1 splice-site SNP responsible for production of the OAS1 p46 isoform strongly associates with COVID-19 severity. This study highlights the importance of endomembrane targeting for the antiviral specificity of OAS1 and suggests early control of SARS-CoV-2 replication through OAS1-p46 is an important determinant of COVID-19 severity.

INTRODUCTION rIFNβ for 24h. Cells homozygous for the A allele only produced p42, while cells homozygous for the G allele only produced p46 ( Figure 1C ). OAS1 KO 293T cells generated by 160 CRISPR/Cas9 gene editing ( Figure S1E ) were used to ectopically express OAS1 p42 and p46 as positive controls ( Figure 1C ). Interestingly, as observed in cell-lines, PBMCs had stronger expression of p46 harboring at least one G allele, which may be influenced by splice site preference or mRNA stability. We did not observe expression of the p44, p48, or p52 isoforms consistent with previous reports demonstrating that these isoforms are weakly expressed at the 165 mRNA level, and unstable at the protein level (Carey et al., 2019; .

The OAS1 p46 isoform is geranylgeranylated.

One of the major differences between OAS1 p42 and p46 is the presence of a C-terminal CaaX motif in p46 ( Figure S1C ). Proteins containing CaaX motifs at their C-termini are prenylated and 170

targeted to the cytosolic face of cellular membranes of the endomembrane system, including the endoplasmic reticulum and Golgi apparatus (Wang and Casey, 2016) . Depending on the identity of the "X" amino acid, CaaX-containing proteins are subjected to two types of prenylation, farnesylation or geranylgeranylation. CaaX motifs with a leucine in the "X" position, as observed for p46, ar preferentially geranylgeranylated by geranylgeranyl transferase I (GGTaseI) 175

One of the major differences between OAS1 p42 and p46 is the presence of a C-terminal CaaX motif in p46 ( Figure S1C ). Proteins containing CaaX motifs at their C-termini are prenylated at the ER and targeted to the cytosolic face of cellular membranes of the endomembrane system, including the endoplasmic reticulum and Golgi apparatus (Wang and Casey, 2016) . Depending on the identity of the "X" amino acid, CaaX-containing proteins can be either farnesylated or 180 geranylgeranylated by the respective farnesyl or geranylgeranyl transferases. CaaX motifs with a leucine in the "X" position are preferentially geranylgeranylated by geranylgeranyl transferase I (GGTaseI) (Hartman et al., 2005 ) . In order to test if the CaaX motif in p46 is geranylgeranylated, we performed an in vitro geranylgeranylation detection assay using a click chemistry approach (DeGraw et al., 2010) . OAS1 KO 293T cells were transfected with N-185

terminal FLAG-tagged OAS1 p42, p46, p42CTIL, and p46ATIL expression constructs. The p42CTIL construct represents p42 with the addition of the CaaX (CTIL) motif from p46, which will allow us to test if the CaaX motif from p46 is sufficient to drive geranylgeranylation of p42. The p46ATIL construct was generated by mutating the CaaX cysteine to an alanine (C->A) to disrupt its prenylation, allowing us to test if this motif is necessary for its geranylgeranylation. 190

Immunoblotting of the immunoprecipitated OAS1 proteins revealed CaaX-containing p46 and p42CTIL proteins were geranylgeranylated, while p42 and p46ATIL were not ( Figure 1D ). These data confirm p46, but not p42, is geranylgeranylated and demonstrate the CaaX motif is necessary and sufficient for the geranylgeranylation of OAS1.

OAS1 p46 localizes to the endomembrane system. Based on the differential geranylgeranylation of the p42 and p46 OAS1 isoforms, we hypothesized these proteins would localize to unique subcellular compartments. Since we verified which OAS1 isoforms are produced in primary human fibroblasts and several human cell lines, we could stratify these cells into those producing p42 and or p46 based on if the cells 200

harbored the A or the G allele at rs10774671. Primary human fibroblasts, PH5CH8, and THP-1, harboring both A and G alleles all produced p42 and p46, while primary human fibroblasts and A549 cells that carried only A allele only produced p42 (Figures 1B, 2D and S2C) . We used these cells to evaluate if OAS1 isoforms localize to different compartments under endogenous expression conditions using confocal laser scanning microscopy (cLSM). We observed OAS1 205 localization at a perinuclear compartment in p46-producing cells after rIFNβ treatment, while, in contrast, OAS1 localized to the cytosol and nucleus in cells incapable of producing p46 ( Figure  S1F and S2C). Co-staining with Golgin-97, a marker of the trans-Golgi network, identified this organelle as a site of OAS1 localization in p46-producing cells ( Figures 1E, 1F , S1G and S2C). Cells capable of producing p46 showed a significant increase in OAS1 co-localization with the 210

Golgi over that in A549 cells that only produce p42 ( Figures 1E, 1F , S1G and S2C). Previous reports have suggested mitochondrial localization, which is not observed in our extensive imaging analysis (Kjaer et al., 2014; Kjaer et al., 2009) . These data demonstrate OAS1 localizes to an intracellular organelle membrane in cells capable of producing p46. The endogenous localization patterns of OAS1 suggested the p42 and p46 isoforms 215 localize to distinct subcellular compartments, a hypothesis we tested by performing cLSM on OAS1 KO Huh7 cells ectopically expressing p42 or p46 ( Figures 1G, 1H and S1H). OAS1 localized to the Golgi in cells ectopically expressing p46, while the p42 isoform localized to the cytosol and the nucleus ( Figures 1G and 1H ). Compared with p42, p46 showed significantly stronger localization to the Golgi ( Figures 1G and 1H ). We next tested the contribution of the 220

CaaX motif to the Golgi localization of p46 by expressing the CaaX mutant p46ATIL (C>A) and evaluating its localization by microscopy. p46ATIL instead localized to the cytosol and nucleus, similar to p42, indicating the CaaX motif is necessary for the Golgi localization of p46 (Figures 1G and 1H) . We then tested if adding the CaaX motif to the p42 isoform is sufficient to localize p42CTIL to the Golgi. Consistent with our hypothesis, ectopically expressed p42CTIL localized 225

to the Golgi and showed significantly stronger co-localization with the Golgi over p42, confirming the CaaX motif is both necessary and sufficient to localize OAS1 isoforms to the Golgi ( Figures 1G and 1H ). These findings were further confirmed in OAS1 KO A549 cells ectopically expressing p42, p46, p42CTIL, and p46ATIL (Figures S1I, S1J, and S1K). The p44, p48, and p52 isoforms, which also lack a CaaX motif, localized to the cytosol and nucleus in 230 manner similar to p42 when ectopically expressed in OAS1 KO Huh7 cells (Figures S1L and S1M). These experiments demonstrate the p46 isoform of OAS1 localizes to the endomembrane system, particularly the Golgi, in a prenylation-dependent manner.

OAS1 p42 and p46 are differentially antiviral.

The differential localization of OAS1 isoforms led us to hypothesize these isoforms have differential antiviral activity. Specifically, localization of p46 to the endomembrane system led us to hypothesize this isoform may have enhanced antiviral activity against viruses that use these organelle membranes for replication. Encephalomyocarditis virus (EMCV) is a positive-strand RNA virus sensitive to the OAS/RNase L pathway and EMCV replicates on organelles of the 240 endomembrane system, particularly ER and Golgi membranes (Chebath et al., 1987; Melia et al., 2018) . We used this virus to investigate whether OAS1 isoforms confer different antiviral activity against a virus that uses the endomembrane system for replication. We transfected expression plasmids encoding p42, p46, or an empty vector control in OAS1 KO 293T cells, and infected with EMCV at a multiplicity of infection (MOI) of 0.001 at 24h post-transfection 245 (Figure 2A ). At 24h post-infection, total RNA and culture supernatants were collected for RT-qPCR of EMCV RNA or plaque assay, respectively. We found p42 expression led to a significant five-fold reduction in EMCV RNA over control, while p46 expression led to a significant 50-fold reduction over EV and a significant 10-fold reduction over p42 at 24h postinfection ( Figure 2B ). Quantification of viral titer from supernatants by plaque assay showed p46 250 expression led to a significant 100-fold reduction in EMCV titer over control and a significant 50-fold reduction over p42. In contrast, the p42 isoform reduced EMCV titer by five-fold over control (Figures 2C and S2A) . We also compared the antiviral activity of p44, p48, and p52 against EMCV over a range of doses and found their antiviral activity was inferior to p46 (Figures 2D and 2E) . These data demonstrate that, among OAS1 isoforms, OAS1 p46 confers 255 the strongest antiviral activity against EMCV. Next, we tested the ability of endogenously expressed OAS1 isoforms to restrict EMCV using an siRNA knockdown approach in THP-1 cells (A/G at rs10774671), which express both the p42 and p46 isoforms. We transfected PMA-differentiated THP-1 macrophages with siRNAs against both OAS1 isoforms, or p42 or p46 alone, and then infected with EMCV ( Figure 2F ).

Knockdown of total OAS1 or p46 led to a significant four-fold increase in viral titer compared to a non-targeting control (NC), whereas specific knockdown of p42 had no effect on viral titer compared to NC siRNA ( Figures 2G and S2B ). Finally, we tested if the G allele at rs10774671 correlates with resistance to EMCV in primary human fibroblasts isolated from six donors (3 A/A, 3 A/G). Quantification of EMCV RNA and viral titer at 24h post-infection revealed that 265 cells with at least one copy of the G allele had reduced levels of EMCV burden (Figures 2H and 2I) . These data show major antiviral differences in OAS1 isoforms at endogenous levels of expression and support an important role for p46, but not p42, in restricting a virus that utilizes the endomembrane system for replication.

The antiviral function of OAS1 isoforms requires catalytic activity and RNase L. The differences in the antiviral activity and localization of OAS1 p42 and p46 led us to hypothesize that these isoforms may utilize unique antiviral mechanisms. Both RNase L and 2′-5′A independent antiviral mechanisms have been documented for OAS proteins, but whether human OAS1 p42 and p46 isoforms differentially utilize 2′-5′A or RNase L is unknown (Carey 275 et al., 2019; Elbahesh et al., 2011; Kristiansen et al., 2010; Lin et al., 2009) . In order to ablate catalytic activity and test if both OAS1 isoforms required 2′-5′A synthetase activity to be antiviral, we mutated two key aspartic acid residues required for synthetase activity, D75 and D77, in the catalytic core of p42 and p46 to alanine (D75A; A76A; D77A) which we will refer to as catalytic mutant hereafter (Sarkar et al., 1999) . We expressed these catalytically inactive 280 OAS1 isoforms alongside their corresponding wild type constructs in OAS1 KO 293T cells and tested their ability to restrict EMCV ( Figure 3A ). As before, expression of p42 and p46 reduced EMCV RNA and titer to different degrees. However, the catalytically inactive OAS1 constructs failed to restrict EMCV RNA or titer ( Figures 3B and 3C) . Furthermore, the catalytic mutants failed to reduce EMCV RNA compared to their wild type counterparts over a range of protein 285 expression levels ( Figures S3A and S3B ). These data indicate that the ability to synthesize 2′-5′A is essential for the antiviral activity of OAS1 p42 and p46 against EMCV. We next tested if p42 and p46 require RNase L for their antiviral activity by expressing p42 or p46 in RNase L-deficient 293T cells ( Figures 3D and S3C ). Compared to non-targeted Cas9 control cells, expression of p42 and p46 in RNASEL KO 293T cells had no impact on 290 EMCV vRNA or titers ( Figures 3E and 3F ). Next, we tested if RNase L expression was sufficient to rescue the antiviral activity of OAS1 isoforms in these cells. RNASEL KO 293T cells complemented with RNase L and expressing p46 showed a significant reduction in EMCV RNA and titer over control. Although not significant, complementing p42 expressing cells with RNase L showed a trend in reducing viral RNA and titer ( Figures S3D, S3E , and S3F). These 295 experiments show that RNase L is required for the antiviral activity of OAS1 p42 and p46 against EMCV. Although the RNA binding and catalytic domains are identical between the p42 and p46, we tested if the distinct C-terminal regions of these isoforms could affect their enzymatic activity. To this end, we performed 2¢-5¢A synthetase activity assays of p42, p46, and their 300 respective CaaX mutants. As expected, these isoforms shared a similar capacity to synthesize 2¢-5¢A and the CaaX motif did not alter the catalytic activity ( Figure 3G ). Cleavage of cellular RNA by RNase L has been shown to generate immunostimulatory RNAs that are sensed by RIG-I and MDA5 and increase antiviral protection through the production of interferon (Malathi et al., 2007) . To determine if the differences in antiviral 305 activity of p42 and p46 could be explained by differential production of immunostimulatory RNAs, we tested the antiviral activity of p42 and p46 in wild type or IRF3 KO 293FT cells ( Figure 3H ). In wild type cells, overexpression of p42 and p46 reduced EMCV RNA significantly over the control. Importantly, loss of IRF3 expression did not impede the ability of p46 to restrict EMCV, as EMCV RNA was significantly reduced over control in IRF3 KO cells 310 expressing p46 ( Figure 3I ). Expression of p42 in IRF3 KO cells showed a trend towards reducing EMCV RNA ( Figure 3I ). These data confirm p42 and p46 confer antiviral activity in the absence of additional factors that are induced by IRF3. Collectively, these data demonstrate that p42 and p46 both use the 2′-5′A/RNase L pathway and have a similar capacity to synthesize 2′-5′A to confer antiviral activity. This suggests that some other feature must be important for their 315 differential antiviral activities.

Since the CaaX motif localizes p46 to the endomembrane system, we investigated if this 320 motif played a role in the enhanced antiviral activity of p46 against EMCV. During picornavirus infection, ER-and Golgi-derived membranes form replication organelles that are sites of vRNA replication (Melia et al., 2019; Melia et al., 2018) .synthesis (Melia et al., 2019; Melia et al., 2018) . Therefore, we hypothesized that endomembrane localization of p46 places this viral RNA sensor on membranes utilized by EMCV for replication and would allow p46 enhanced access to 325 EMCV RNA. In contrast, p42 would have limited access to vRNA as it is localized to the cytosol and nucleus. We determined the localization of p42 and p46 during EMCV infection by infecting OAS1 KO Huh7 cells expressing p42 or p46 with EMCV and then stained for OAS1 and doublestranded RNA (dsRNA) followed by cLSM. In mock infected cells expressing p46, this isoform was localized to perinuclear structures ( Figure 4A ). In cells infected with EMCV, p46 was no 330 longer perinuclear and instead redistributed throughout the cell to sites of double-stranded viral RNA replication intermediate accumulation ( Figure 4A ). In contrast, p42 remained in the cytosol and nucleus before and after infection ( Figure 4A ). These data suggest that p46 is in close proximity to sites of EMCV RNA replication where double stranded RNA ligands are present. We next used an RNA immunoprecipitation approach to test whether OAS1 p46 has enhanced 335 access to EMCV RNA and if the CaaX motif is important for its ability to bind EMCV RNA. Since RNase L cleaves RNA and is required for OAS1-mediated restriction of EMCV, RNase L deficient Huh7 cells allowed us to test OAS1 isoform binding to EMCV RNA in samples with equivalent amounts of EMCV RNA ( Figure 4B ). N-terminally FLAG-tagged OAS1 p42, p46, p42CTIL and p46ATIL isoforms were expressed in OAS1 KO Huh7 cells, which are also devoid 340

of RNase L expression ( Figure S4A ). Expression of OAS1 isoforms in OAS1 KO Huh7 cells had no impact on viral RNA levels ( Figure 4B ). However, RT-qPCR for EMCV RNA after FLAG immunoprecipitation revealed that p46 tended to bind more EMCV RNA than p42 ( Figure 4B ). Although adding a CaaX motif to p42 (p42CTIL) did not significantly increase the ability of p42 to pull down EMCV RNA, mutating the CaaX motif on p46 (p46ATIL) caused a significant loss 345 in EMCV RNA binding ( Figure 4B ). These data confirm that endomembrane targeting through the CaaX motif is required for p46 to bind EMCV RNA, as mutating the CaaX motif completely ablated the ability of p46 to access EMCV RNA ( Figure 4B ). To investigate the role of the CaaX motif to the antiviral activity of OAS1, we expressed p42, p46, p42CTIL, or p46ATIL in OAS1 KO 293T cells followed by infection with EMCV 350 ( Figure 4C ). Compared to cells expressing p46, cells expressing CaaX-mutant p46ATIL showed a significant five-fold increase in EMCV RNA as well as an increase in viral titer ( Figures 4D  and 4E ); a decreased antiviral activity similar to that of the p42 isoform. Intriguingly, addition of a CaaX motif to OAS1 p42 did not increase the antiviral activity of p42 ( Figures 4D and 4E ). One possible explanation for this discrepancy could be that the peptide sequence in the unique C-355

terminus of p42 contains a motif that is inhibitory to its antiviral activity. We therefore generated another OAS1 variant which contains a CaaX motif directly downstream of the common OAS1 peptide sequence shared between all OAS1 isoforms, hereinafter referred to as OAS1 common+CTIL ( Figure 4F ). We tested the antiviral activity of OAS1 common+CTIL compared to the p42 and p46 isoforms. At similar protein expression levels, OAS1 common+CTIL did not 360

confer significantly increased antiviral activity over the empty vector control and showed RNA levels similar to those in cells expressing the p42 isoform ( Figures 4G and 4H ) Interestingly, the OAS1 common+CTIL variant showed significantly decreased co-localization with Golgi membranes compared to p46 and appeared cytosolic and nuclear, as observed above for CaaXdeficient p42 ( Figure S4B ). Together, these data indicate membrane targeting of OAS1 p46 365 through a CaaX motif is crucial for its antiviral activity against EMCV. However, a CaaX motif alone is not sufficient to provide optimal antiviral activity against EMCV. Overall, these data suggest other features in the unique C-termini of OAS1 p42 and p46 contribute to their individual antiviral activity. 370

Combined effects of CaaX motif, C-terminus length and oligomerization domain confer differential antiviral activity of OAS1 isoforms.

Since the unique C-terminus of p46 (50 aa) is longer than the unique C-terminus of p42 (18 aa), we investigated if the unique sequence in p46 contains additional motifs important for its antiviral activity. We generated sequential hexameric alanine substitution mutants throughout the 375 C-terminus of OAS1 p46 and compared their antiviral activity against EMCV alongside p46 and CaaX-deficient p46ATIL ( Figure 5A ). As expected from our previous experiments, p46ATIL had reduced antiviral activity when compared with p46 ( Figures 5B, 5C , and S5A). However, compared to p46, none of the alanine substitution mutants affected viral RNA levels significantly ( Figures 5C and S5A) . Notably, mutant 8 showed a trend toward increased EMCV RNA levels, 380

suggesting that the mutated amino acids might contribute to the antiviral activity of OAS1 p46. Although OAS1 p46 mutant 8 localized to the Golgi, compared with p46, this mutant showed a slight but significant decrease in its co-localization with golgin-97 ( Figures S5B and S5C ). In fact, the E 392 -E/N 393 -D/N 394 sequence is conserved in most mammalian OAS1 p46 orthologous isoforms, which suggests a possible function for this motif ( Figure S1B ). Nevertheless, none of 385 the alanine substitution mutants affected EMCV RNA levels as strongly as the disruption of the CaaX motif in p46ATIL.

Previous studies have demonstrated a three amino acid motif (C 331 F 332 K 333 ), present in all OAS1 isoforms, is critical for OAS1 oligomerization and 2′-5′A production (Ghosh et al., 1997) . The CFK motif is in close proximity to the unique C-termini of p42 and p46, however, the 390 requirement of the CFK domain for the antiviral activity of these proteins is unknown. Therefore, we generated OAS1 p42 and p46 CFK mutants (C331A-F332A-K333A) and compared their antiviral activity against EMCV to their wild type counterparts. Relative to wild type p42 and p46, mutation of the CFK domain led to a significant loss in the ability of p42 and p46 to reduce EMCV RNA by three-fold and two-fold, respectively ( Figures 5D and 5E ). Interestingly, 395

disruption of the CFK motif in p42 resulted in complete loss of antiviral activity, demonstrated by similar EMCV RNA copies as cells expressing an empty vector control ( Figures 5D and 5E ). OAS1 p46 with a disrupted CFK motif maintained residual antiviral activity similar to the antiviral activity of wild type p42 ( Figures 5D and 5E ). These data show that the antiviral activity of OAS1 p42 and p46 is partially mediated by CFK motif-dependent oligomerization. 400

This oligomerization facilitates enhanced 2′-5′A synthesis capacity ( Figure S5D ). Furthermore, we show that CFK-mediated oligomerization is required for optimal EMCV RNA binding of OAS1 p46 ( Figure S5E ). Importantly, mutation of the CFK motif did not alter the localization of either isoform ( Figures S5F and S5G ). This suggests that the antiviral activity of OAS1 p42 is completely dependent on the CFK oligomerization motif, whereas p46 only partially depends on 405 the CFK oligomerization motif for its antiviral activity. Since p46 is prenylated and embedded in cellular membranes, the longer C-terminus of p46 may function as a flexible linker to facilitate optimal oligomerization. This may explain why adding a CaaX motif to p42 is not sufficient to enhance its antiviral activity (see Figures 4C-E) .

To test this, we generated several C-terminal truncation mutants of p46 termed p46 D12aa, 410

D22aa, and D32aa, and determined their antiviral activity against EMCV alongside p46 and p42CTIL ( Figure 5F ). Compared to p46, the truncation mutants showed a length-dependent decrease in their ability to reduce EMCV RNA ( Figures 5G and 5H ). The D22aa and D32aa deletion mutants had similar antiviral activities as p42CTIL, which is identical in length to the D32aa mutant ( Figure 5H ). Similar to p46, all the p46 truncation mutants localized to the Golgi 415 ( Figure S5H ). However, there was a slight but significant decrease in the correlation of p46 D32aa with the Golgi ( Figure S5I ). These data suggest that, in addition to the CaaX motif and an oligomerization domain, the C-terminal length of p46 is important for its antiviral activity. Furthermore, we observed the conservation of a longer CaaX-containing p46-specific Cterminal sequence without sequence similarity in various other species. Interestingly, the p46-420 specific C-terminal sequence was truncated in rodents and bats. To test if the length of the p46specific C-terminal sequence affected the antiviral activity against EMCV, we created chimeric proteins by inserting the C-terminal sequences specific to the p46 orthologs of cow (Bos taurus), fox (Vulpes vulpes), flying fox (Pteropus alecto), and alligator (Alligator mississippiensis) into the human OAS1 p46 gene ( Figure S5J ). We selected species with divergent OAS1 p46-specific 425 C-terminal sequences and minimal amino acid sequence identities when compared to human OAS1 p46. We found cow and fox OAS1 chimeras mimicked human OAS1 p46 antiviral activity, whereas the shorter C-terminus of flying fox led to decreased antiviral activity of the chimeric protein compared to human OAS1 p46 ( Figure S5J ). These data are consistent with our hypothesis that the length of the p46-specific C-terminal sequence is important for increased 430 antiviral activity. Intriguingly, the alligator OAS1 p46 chimeric protein, with similar C-terminus length as human p46 but with most divergence in peptide sequence, failed to exhibit similar antiviral activity as human OAS1 p46 ( Figure S5J ). Upon close examination, we found that the alligator (and guinea pig) lack conservation of the E 392 -E/N 393 -D/N 394 motif proximal to the CaaX motif, which we have shown is required for strong antiviral activity (Figures 5A-C and 435 S1B). These data suggest that the length of the p46-specific C-terminal sequence, CFK oligomerization domain, and potentially E-E/N-D/N sequence are required for enhanced antiviral activity of OAS1 p46. OAS1 p46 has broad antiviral activity against viruses that use the endomembrane system for replication. 440 We sought to define the antiviral activity of OAS1 isoforms against other positive-strand RNA viruses that use the endomembrane system for replication. We tested if OAS1 isoforms are differentially antiviral against West Nile virus (WNV) by infecting OAS1 KO 293T cells expressing p42, p46, or control with WNV. WNV, like all flaviviruses, replicates on membranes of the endomembrane system, primarily ER-derived membranes (Westaway et al., 1997) .

Expression of p46 led to a 90% reduction in WNV titer relative to control, while p42 did not significantly impact WNV titer ( Figure 6A ). These data demonstrate that OAS1 p46 has enhanced antiviral activity against WNV. We tested if the enhanced antiviral activity of p46 against WNV could be explained by p46 localizing to sites of flavivirus RNA replication. We performed cLSM to evaluate the 450 localization of p42 and p46 during flavivirus infection. We infected OAS1 KO 293T cells expressing p42 and p46 with WNV, followed by staining for OAS1, dsRNA, and the endoplasmic reticulum protein disulfide-isomerase A3 (PDIA3), since WNV replicates within invaginations of the ER membrane. During WNV infection, we observed p42 in the cytosol and nucleus. Indeed, p42 did not appear to be recruited to sites of WNV RNA production ( Figure 6B , 455

top panel, arrows). However, during WNV infection, p46 localized to PDIA3 and dsRNApositive replication sites ( Figure 6B , bottom panel, arrows). Quantification of OAS1 isoforms relative to PDIA3 revealed a significant increase in the correlation of p46, but not p42, to PDIA3-positive membranes during infection, suggesting that p46 is recruited to sites of WNV replication ( Figures 6B and 6C ). Quantification of OAS1 isoform localization to viral RNA 460 revealed OAS1 p46 has significantly stronger localization to dsRNA during WNV infection than p42 ( Figure 6D ). We evaluated whether p46 was present in the Golgi during WNV infection by staining for OAS1, Golgin-97, and dsRNA in cells infected with WNV. OAS1 p46 was significantly more correlated with the Golgi than p42 in both uninfected and WNV-infected cells (Figures 465 S6A and S6B). However, WNV infection caused a significant decrease in the association of p46 with Golgin-97 ( Figures S6A and S6B ). This suggests that p46 may be recruited from the Golgi to sites of WNV replication and is consistent with previous observations that Golgi membranes and proteins, including RNA binding proteins, are recruited to the replication organelles of flaviviruses (Ward et al., 2016) . Alternatively, p46 might be recruited during prenylation at the 470 ER membrane. We next tested if p46 localizes to sites of Zika virus (ZIKV) replication, another flavivirus that utilizes ER membranes for replication (Cortese et al., 2017) . We observed p46, but not p42, localized to ZIKV replication sites at the ER ( Figure S6C ). Overall, these data suggest that endomembrane localization of OAS1 p46 allows this protein to access sites of flavivirus 475 RNA replication. The human pathogenic picornavirus coxsackievirus B3 (CVB3) replicates within modified Golgi membranes, and thus p46 may also be positioned to readily sense CVB3 RNA (Melia et al., 2019) . We tested the antiviral activity of OAS1 p42 and p46 against CVB3 by infecting OAS1 KO 293T cells expressing p42, p46 or a control. RT-qPCR analysis at 24h post-480

infection revealed a 70% reduction in CVB3 RNA in p46 expressing cells compared to control, while p42 had almost no effect on CVB3 RNA levels ( Figure S6D ). Similarly, CVB3 titers at 48h post-infection were significantly reduced by 50% in p46 expressing cells ( Figure 6E ). These data suggest that OAS1 p46 may have broad antiviral activity against members of the picornavirus family, while p42 is less effective. 485

Coronaviruses, such as SARS-CoV-2, use the endomembrane system for replication using primarily ER-derived membranes (Romero-Brey and Bartenschlager, 2014; Snijder et al., 2020) . We therefore hypothesized that p46 may have enhanced antiviral activity against coronaviruses. We assessed the antiviral activity of p42 and p46 against SARS-CoV-2 in 293T cells expressing the receptor ACE2. Expression of p42 led to a five-fold reduction in SARS-490

CoV-2 titer ( Figure 6F ). However, p46 had a significant five-fold greater antiviral activity over p42 (25-fold over EV) against SARS-CoV-2 ( Figure 6F ). Importantly, the enhanced antiviral activity of p46 against SARS-CoV-2 depended on the CaaX motif, as p46ATIL demonstrated decreased antiviral activity similar to that of p42. This suggests that endomembrane targeting of p46 is critical to its enhanced antiviral activity against SARS-CoV-2 ( Figure 6F ).

We next tested the antiviral activity of p42 and p46 against negative-strand RNA viruses, such as Influenza A virus (IAV) and Indiana vesiculovirus (formerly VSV), which do not replicate on intracellular organelle membranes. IAV replicates within the nucleus where p42 is also present and is sensitive to the OAS/RNase L pathway (Li et al., 2016; Min and Krug, 2006) . To test if OAS1 isoforms are differentially antiviral against IAV, we infected OAS1 KO 293T 500 cells expressing p42, p46, or a control with influenza A/PR/8/34 ( Figure S6E ). Viral RNA analysis revealed neither OAS1 isoform impacted IAV RNA levels significantly at 24h postinfection. Mutations in IAV NS1 disrupting RNA binding render this virus sensitive to the OAS/RNase L pathway, although it is unclear if what antiviral role OAS1 and isoforms playing during IAV infection (Min and Krug, 2006) . We found expression of p42 or p46 had no impact 505 on the replication of this NS1-mutant IAV ( Figure S6F ). Similarly, we observed that GFPexpressing VSV, which replicates in the cytosol, was also insensitive to OAS1 p42 and p46. Compared to an empty vector control, expression of p42 or p46 did not impact the number of GFP+ cells during infection ( Figure S6G ). Collectively these data suggest OAS1 p46 is broadly antiviral against viruses that use the endomembrane system for replication. 510 OAS1 rs10774671 is associated with severe COVID-19 disease.

Since the OAS1 rs10774671 A/G variant generates the OAS1 p42 and p46 isoforms, respectively, that affected the response to SARS-CoV-2 infection, we tested whether this SNP is associated with COVID-19 disease severity. We hypothesized that the G allele, that encodes the 515 p46 isoform would decrease the risk of severe COVID-19 complications due to its superior antiviral activity compared to the A allele that generates p42. To test this, we genotyped the rs10774671 SNP in 34 COVID-19 severe cases (hospitalized, requiring mechanical ventilation) and 99 ancestry matched healthy controls (Table S1 ). Association was tested by logistic regression, adjusting for sex and self-reported ancestry. We detected association of rs10774671 520

in severe COVID-19 cases (p=0.017, Odds Ratio 0.35, 95% CI 0.15-0.83) using a dominant model, indicating that the G allele was protective for severe COVID-19 disease. These data are consistent with results from a recent genome wide association study of 1,676 critically ill COVID-19 patients of European descent and UK Biobank controls (n=8,380) which identified a significant association with the OAS1/OAS2/OAS3 locus (Pairo-Castineira et 525

al., 2020b). The lead SNP in this region, rs10735079 (p=1.65x10 -8 , OR 1.3, 95% CI 1.18-1.42), is in high linkage disequilibrium with the rs10774671 A/G splicing site variant (D'0.91, r 2 =0.79).

To replicate our findings, we tested association of rs10774671 in these cohorts by logistic regression, correcting for age, sex, postal code deprivation decile, and principal components of ancestry. Significant association of the rs10774671 G allele with severe COVID-19 was detected 530 (p=7.38x10 -7 , OR 0.80 95% CI 0.71-0.89). Together these results demonstrate that the G allele at OAS1 rs10774671 encoding the OAS1 p46 isoform contributes protection from severe disease in SARS-CoV-2 infected patients.

In this study, we show endomembrane targeting of OAS1 p46 confers enhanced antiviral activity of this isoform against viruses that use the endomembrane system for their replication. The p46 isoform contains a functional CaaX motif that targets this isoform to the endomembrane system, primarily to Golgi membranes, while other OAS1 isoforms localize to the cytosol and nucleus. Compared with p42, the p46 isoform has enhanced antiviral activity against 540 picornaviruses, flaviviruses, and SARS-CoV-2, all viruses that replicate their RNA within modified organelles of the endomembrane system (Cortese et al., 2017; Melia et al., 2019; Melia et al., 2018; Snijder et al., 2020) . Although replicating in these modified endomembrane compartments is an important immune evasion strategy of positive-strand RNA viruses, out data shows localizing OAS1 p46 to the endomembrane system results in enhanced access to viral 545 RNA and activation of RNase L dependent antiviral activity. In contrast, p42, which is localized to the cytosol and nucleus, has a comparatively weak ability to sense viral RNA and initiate antiviral activity against viruses that replicate within intracellular membranes (Figure 7) .

likely difficult for positive-strand RNA viruses to evade, as these viruses are unlikely to evolve away from this fundamental replication strategy (Romero-Brey and Bartenschlager, 2014). Secondly, many positive-strand RNA viruses depend on lipid products of the mevalonate pathway or prenylation for replication and therefore cannot broadly antagonize this pathway (Mackenzie et al., 2007) . Viral antagonism by specifically targeting OAS1 p46 protein or 555 directly blocking its prenylation or splicing could be possible. Alternatively, viral interference of the OAS-RNase L pathway would negate the antiviral activity OAS1 p46. Such a mechanism is exemplified in the case of some viruses, where the non-structural proteins inhibit the OAS-RNase L pathway (Silverman, 2007; Thornbrough et al., 2016; Zhao et al., 2012) .

We also show addition of the CaaX motif to p42 is not sufficient to enhance its ability to 560 bind viral RNA or increase its antiviral activity. This suggests the CaaX motif is just one of the crucial features in the unique OAS1 p46 C-terminal region that contributes to differential antiviral activity. Tetramerization of OAS1 has been proposed as a requirement for synthesis of the second messenger 2′-5′A, indicating that several OAS1 molecules in close proximity are required to induce 2′-5′A synthesis upon viral RNA binding and to subsequently activate the 565

RNase L pathway (Ghosh et al., 1997) . Aggregation of several OAS1 proteins around viral dsRNA might also increase the number of RNA-binding domains within the OAS1/vRNA complex, thereby allowing access to longer vRNA increasing 2′-5′A synthesis. Mutation of the CFK oligomerization motif in OAS1 p42 led to a complete loss of antiviral activity, while the p46 CFK mutant maintained antiviral activity similar to that of the wildtype p42 isoform. This 570 suggests a model in which OAS1 p46 without a functional CFK oligomerization domain still maintains a residual antiviral activity. These data show p46 proteins might oligomerize which is further aided by CaaX motif-mediated localization at Golgi membranes. Together, we have defined features including the CaaX motif in the unique C-terminus that contribute to the enhanced antiviral activity of p46. The unique alternatively spliced C-terminal region including 575 the oligomerization and prenylation motifs in p46 is maintained across vertebrates presumably because it displays the strongest antiviral activity compared to other OAS1 isoforms (Figure 7) .

The rs10774671 G allele controlling OAS1 p46 expression was originally identified as a West Nile virus resistance allele, although the mechanism accounting for the protection 580 conferred by this SNP was lacking (Lim et al., 2009) . In this study, we identify a functional OAS1 splice site variant affects COVID-19 outcome. We found the A allele contributes to genetic risk for severe COVID-19 disease in patients with respiratory failure compared to healthy control subjects in two cohorts. We propose the G allele association with protection against severe COVID-19 is explained in part by the enhanced antiviral activity of OAS1 p46 585 against SARS-CoV-2, which contributes to early control of viral replication. Additionally, both rs10774671 and the lead GWAS SNP rs10735079 are expression and splicing QTL for OAS1 and OAS3 in our analyses (Pairo-Castineira et al., 2020a). Thus, the association of these SNPs with severe COVID-19 may be two-fold: affecting splicing and expression levels both of those observed in primary human cells. 590 We have shown that the splice-site SNP controls the subcellular targeting of a critical viral RNA sensor, OAS1, which may allow the host to respond to viral evolutionary pressures and replication strategies. Interestingly, the G allele is ancestral and prevalent in African populations, in contrast the A allele is prevalent in rest of the world. Previous studies have documented reintroduction of the G-allele into Homo sapiens population by introgression from 595

Neanderthals (Sams et al., 2016) The broad number of RNA-and even DNA-viruses inhibited by the OAS/RNAse L pathway has raised interesting questions about the determinates of OAS antiviral specificity 610 (Silverman, 2007) . In humans, the OAS family is comprised of three catalytically active OAS proteins: OAS1, OAS2, and OAS3. OAS1 contains an RNA binding domain and a catalytically active domain, while OAS2 contains a catalytically inactive repeat of this unit, and OAS3 contains two catalytically inactive repeats of this unit (Hornung et al., 2014) . The effect of OAS3 SNPs that associate with COVID-19 is still unclear (Pairo-Castineira et al., 2020a). Although the 615 OAS proteins have different RNA binding capabilities and favor synthesis of different lengths of 2′-5′A (Ibsen et al., 2014) , we demonstrate a novel mechanism where subcellular localization determines the antiviral specificity of these proteins. The broad number of RNA-and even DNAviruses inhibited by the OAS/RNAse L pathway has raised interesting questions about the determinates of OAS antiviral specificity (Silverman, 2007) . In humans, the OAS family is 620 comprised of three catalytically active OAS proteins: OAS1, OAS2, and OAS3. OAS1 contains an RNA binding domain and a catalytically active domain, while OAS2 contains a catalytically inactive repeat of this unit, and OAS3 contains two catalytically inactive repeats of this unit (Hornung et al., 2014) . Although the OAS proteins have different RNA binding capabilities and favor synthesis of different lengths of 2′-5′A (Ibsen et al., 2014) , we demonstrate a novel 625 mechanism where subcellular localization determines the antiviral specificity of these proteins (Ibsen et al., 2014) . Based on the reduction of vRNA in our study, we propose a model in which 2′-5′A locally activates RNase L around sites of viral replication. While the function of OAS1 p42 is unclear but might confer resistance to yet unknown pathogen. This is further supported by the genetic association of the A allele with multiple autoimmune disorders, as enhanced immune 630 responses are sometimes made at a tradeoff for overall host fitness Liu et al., 2017; O'Brien et al., 2010) .

Our work highlights intracellular targeting as a crucial determinant for the specificity of OAS1. By positioning viral RNA sensors at different subcellular sites of viral RNA 635 accumulation, the host can potentially respond to diverse replication strategies in the cytosol, nucleus, or on intracellular membranes. Although this work focused on OAS1 and the detection of viral nucleic acids, subcellular targeting is likely also important for function of other OAS proteins. Future studies on the subcellular localization of other OAS proteins will define how this pathway is able to respond to viruses with diverse intracellular replication strategies. 640 Golgin-97 (magenta) antibodies and DAPI (blue). H) Pearson's correlation of OAS1 and Golgin-820 97 in OAS1 KO Huh7 cells expressing p42, p46, p42CTIL, or p46ATIL. Scale on micrographs in E) and G) = 5 μm. F), H) Data were analyzed using one-way ANOVA with Tukey's multiple comparisons test where **P<0.01, ****P<0001. 

confer differential antiviral activity of OAS1 isoforms. A) C-termini of OAS1 p46 alanine substitution mutants 1-9. B) Immunoblot of control EV, p46, p46ATIL, and p46 alanine substitution mutants 1-9 in OAS1 KO 293T cells at 24h post transfection. C) Quantification of EMCV 5´UTR by RT-qPCR from OAS1 KO 293T cells expressing a control EV, p46, p46ATIL, and p46 alanine substitution mutants 1-9 at 24h post EMCV infection (MOI=0.001). D) 890

Immunoblot were analyzed using one-way ANOVA with Tukey's multiple comparisons test where **P<0.01, and ***P<0.001 and ****P<0.001. Golgi apparatus, while OAS1 p42 is cytosolic. During positive-strand RNA virus infection, OAS1 p46 is recruited to virus replication organelles (VROs) of flaviviruses, picornaviruses and coronaviruses. Through this targeting p46 gains enhanced access to viral RNA. OAS1 p42 remains cytosolic and nuclear during infection and has limited access to viral RNA. Both OAS1 isoforms require catalytic activity and RNase L to be antiviral. 925 Supplementary figure 1 (related to figure 1) OAS1 isoforms with last 10 amino acids. D) Immunoblot analysis of OAS1 expression in A549 and PH5CH8 cells following treatment with rIFNβ (1000 U/mL) or Sendai virus (100 HAU/mL) for the indicated times. E) Immunoblot analysis of OAS1 expression in Cas9 and clonal OAS1 KO 293T cells treated with rIFNβ (1000 U/mL) for 24h. F) Representative confocal micrographs of the indicated cell lines treated with rIFNβ (1000 U/mL) for 24h followed by staining with 950

anti-OAS1 antibody (green) and DAPI dye (blue). G) Representative maximum intensity projections from the indicated cell lines treated with rIFNβ for 24h followed by staining with anti-OAS1 antibody (green), anti-Golgin-97 (magenta), and DAPI (blue p46, p48 or p52 stained with anti-OAS1 (green) and Golgin-97 (magenta) antibodies and DAPI (blue). M) Pearson's correlation of OAS1 and Golgin-97 in OAS1 KO Huh7 cells expressing p42, p44 p46, p48 or p52. Scale = 5 μm. K) and M) Data were analyzed using one-way ANOVA with Tukey's multiple comparisons test where *P<0.05 and ****P<0.0001. 965

Supplementary figure 2 (related to figure 2). A) Representative plaque assay plates from the data in shown in Figure 2C . B) Representative plaque assay plates from data shown in Figure  2G . C) Representative confocal micrographs of primary human fibroblasts (donor 3, A/A; donor 6, A/G) treated with 1000 U/mL rIFNβ for 24h followed by staining with anti-OAS1 antibody (green), anti-Golgin-97 (magenta), and DAPI (blue). Scale = 20 μm. 24h (MOI=0.001). E) and F) Data were analyzed using a one-way ANOVA with Dunnett's multiple comparisons test where *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Supplementary figure 4 (related to figure 4) . A) Immunoblot of RNase L expression in untreated cell lines. B) Representative confocal micrographs of OAS1 KO Huh7 cells expressing 985 OAS1 p46 and OAS1 common +CTIL stained with anti-OAS1 (green) and Golgin-97 (magenta) antibodies and DAPI (blue). Co-localization of OAS1 and Golgin-97 is shown as Pearson's correlation. Scale = 5 μm. Data were analyzed by unpaired t test where ****P<0.0001.

Supplementary figure 5 (related to figure 5) and Golgin-97 (magenta) antibodies and DAPI (blue). I) OAS1 and Golgin-97 co-localization from H) expressed as Pearson's correlation. J) Alignment of OAS1 p46 chimeric isoforms with C-termini from different species. OAS1 KO 293T cells expressing chimeric OAS1 p46 isoforms were infected with EMCV (MOI=0.001) and EMCV RNA was quantified at 24h post-infection. Equal expression of chimeric p46 proteins was confirmed by Western blot. A), C), I) and J) Data 1010

were analyzed by one-way ANOVA with Dunnett's multiple comparisons test (vs. p46). E) and G) Data were analyzed by one-way ANOVA with Tukey's multiple comparisons test. *P<0.05, **P<0.01 and ****P<0.0001. n.s. = not significant. Scale = 5 μm.

Supplementary figure 6 (related to figure 6) . A) Representative confocal micrographs from 1015 mock or WNV Texas (MOI=1, 24h) infected OAS1 KO 293T cells expressing p42 or p46 and stained with DAPI (blue) and anti-OAS1 (green), Golgin-97 (red), and dsRNA (magenta) antibodies. B) Pearson's correlation of OAS1 and Golgin-97. C) Representative confocal micrographs from mock or ZIKV MR766 (MOI=5, 24h) infected OAS1 KO A549 cells OAS1 KO 293T cells were seeded in 12 well plates coated with 10 μg/mL poly-L-ornithine 1140 hydrobromide (Sigma) and allowed to adhere overnight. Plasmids were then transfected using TransIT-X2 (Mirus Bio). At 24h post transfection, cells were infected with EMCV, WNV, or CVB at the indicated MOIs for 1h with gentle rocking. After 1h the inoculum was removed, and fresh media was added. For virus titer quantification, culture supernatants were serially diluted in DMEM using 96 well plates. For EMCV and WNV, titration was performed on Vero cells grown 1145

to 90% confluency in 6-well plates. For CVB, titration was performed on HeLa cells grown to 90% confluency in 6-well plates. IAV was titered on MDCK cells grown to 90% confluency in 12-well plates. The inoculum was removed after 1h of gentle rocking and a 0.8% agarose overlay was added containing the following: 0.8% UltraPure Low Melting Point Agarose (Thermo Fisher), 1x DMEM, 0.15% Sodium Bicarbonate, 10% heat-inactivated fetal bovine serum (FBS) 1150

(Atlanta Biologicals) and 1% penicillin-streptomycin-glutamine (Mediatech). For EMCV plates were fixed with 4% paraformaldehyde (Santa Cruz Biotechnology) at 24h post-infection followed by staining with a 5% crystal violet solution prepared by dissolving crystal violet (Sigma Aldrich) in a 50/50 mixture of 100% ethanol and deionized water. For IAV plates were fixed with 4% paraformaldehyde (Santa Cruz Biotechnology) at 24h post-infection followed by 1155 staining with a 5% crystal violet solution. For CVB plates were fixed with 4% paraformaldehyde (Santa Cruz Biotechnology) at 24h post-infection followed by staining with a 5% crystal violet solution. For WNV a neutral red overlay containing 0.01% neutral red (Sigma-Aldrich) 0.8% UltraPure Low Melting Point Agarose (Thermo Fisher), 1x DMEM, 0.15% sodium bicarbonate, 10% heat-inactivated fetal bovine serum (FBS) (Atlanta Biologicals) and 1% penicillin-1160 streptomycin-glutamine (Mediatech) was added at 48h post-infection and plaques were read 24h later. For quantification of Indiana vesiculovirus (VSV-GFP) replication, OAS1 KO 293T cells transfected with OAS1 p42, p46 and EV for 24h prior to infection, were infected with Indiana vesiculovirus (VSV-GFP; MOI=0.5, 6h). Cells were harvested using trypsin, washed 2x with 1165 PBS, and incubated with Zombie NIR fixable viability dye (1:1000) (BioLegend) for 30 min at room temperature in PBS. Cells were washed 2x with PBS, fixed in 4% PFA for 10 minutes at room temperature, and then washed once with FACS buffer (PBS with 0.1%BSA). Flow cytometry was performed on a BD FACS Canto II. % VSV-GFP+ cells were quantified using FlowJo (Tree Star).

SARS-CoV-2 strain USA/WA-1/2020 was propagated and titered on VeroE6 cells (gift of Dr. Ralph Baric). For infections, 293T-ACE2 cells seeded in 24-well plates (100,000 cells/well) were transfected with 250ng of plasmid (Empty vector, p42, p46, and p46 ATIL) using the TransIT X2 kit (Mirus). A duplicate plate was transfected at the same time in order to confirm OAS1 expression by immunoblot. 24 hours post-transfection, cells were infected with 1175 SARS-CoV-2 at an MOI of 0.1 in serum free DMEM for 1 hr, and media was replenished with DMEM containing 4% serum. Supernatants were harvested at 48hpi, and serial dilutions were tittered on Vero E6 cells seeded at 90% confluency in 12-well plates. Inoculum was removed after 1 hr of gentle rocking and replenished with an agarose overlay containing 0.4% Noble Agar (Thermo Fisher) in DMEM containing 10% serum. At 72 hpi, plates were fixed with 10% 1180 formaldehyde and stained with crystal violet solution (0.1% crystal violet and 20% methanol in water).

For all microscopy experiments, cells were seeded on #1.5 12-mm coverslips (Bioscience Tools) 1185 coated with 10 μg/mL poly-L-ornithine hydrobromide (Sigma) and allowed to adhere overnight. For experiments testing endogenous OAS1 localization, cells were treated with rIFNβ for 24h. For experiments testing OAS1 localization in dox-inducible OAS1 A549 cells, cells were treated with 200 ng/mL doxycycline for 24h. For experiments testing localization of OAS1 in Huh7 cells, plasmids were transfected for 24h using Lipofectamine 3000 (Thermo Fisher) according to 1190 the manufacturer's instructions. At 24h post treatment/expression, cells were washed with PBS and then fixed in 4% PFA (Electron Microscopy Sciences) in PBS for 10 min at room temperature, washed with PBS, and then permeabilized with PBS containing 0.1% Triton X-100 for 10 minutes at room temperature. Cells were washed with PBS and then resuspended in a 3% BSA/PBS blocking solution for 1 hour. After blocking, cells were stained with rabbit anti-OAS1 1195

(CST) and mouse IgG1 anti-Golgin 97 (CST , Table S3 ) in PBS containing 1% BSA and 0.3% Triton X-100 for 1 hour in the dark at room temperature (Table S2 ). Cells were washed three times with PBS and then stained with the secondary antibodies goat anti-rabbit IgG Alexa Fluor 488 (Themo Fisher) and goat anti-mouse IgG Alexa Fluor 648 (Thermo Fisher) in PBS containing 1% BSA and 0.3% Triton X-100 for 1 hour in the dark at room temperature (Table  1200 S2). Samples were washed once with PBS, stained with DAPI in PBS for 10 minutes in the dark, followed by washing three times with PBS and then mounted with ProLong Glass antifade mounting media (Thermo Fisher). Samples were cured in the dark at room temperature for 24-48h prior to imaging.

For experiments testing OAS1 localization during viral infections, OAS1 was expressed as described above. At 24h post transfection, cells were infected with the indicated virus for 1h with gentle rocking followed by removal of the inoculum and replacement with fresh media. At 24h post treatment/expression, cells were washed with PBS and then fixed in 4% PFA (Electron Microscopy Sciences) in PBS for 10 min at room temperature, washed with PBS, and then 1210 permeabilized with PBS containing 0.1% Triton X-100 for 10 minutes at room temperature. Cells were washed with PBS and then resuspended in a 3% BSA/PBS blocking solution for 1 hour. After blocking, cells infected with EMCV were stained with rabbit anti-OAS1 (CST) and mouse IgG1 anti-dsRNA 9D5 in PBS containing 1% BSA and 0.3% Triton X-100 for 1 hour in the dark at room temperature. Cells infected with WNV were stained with rabbit anti-OAS1 1215

(CST), mouse IgG1 anti Golgin 97 (CST) or mouse anti IgG1 PDIA3 (Sigma-Aldrich) and mouse IgG2a anti-dsRNA J2 (Scicons) in PBS containing 1% BSA and 0.3% Triton X-100 for 1 hour in the dark at room temperature (Table S2) . EMCV infected cells were washed three times with PBS and then stained with the secondary antibodies goat anti-rabbit IgG Alexa Fluor 488

Interleukin-1beta Induces mtDNA Release to Activate Innate Immune Signaling via 655 cGAS-STING

Regulation of cGAS-and RLR-mediated immunity to nucleic acids

Auto-antibodies to type I IFNs can 660 underlie adverse reactions to yellow fever live attenuated vaccine

Variation in antiviral 2',5'-oligoadenylate synthetase (2'5'AS) enzyme activity is controlled by a single-nucleotide polymorphism at a splice-acceptor site in the OAS1 gene

Recurrent Loss-of-Function Mutations Reveal Costs to OAS1 Antiviral Activity in Primates

Viral evasion of intracellular DNA and RNA sensing

Constitutive expression of (2'-5') oligo A synthetase confers resistance to picornavirus infection Nature 330

Ultrastructural Characterization of Zika Virus Replication Factories

Plasma Exchange to Rescue Patients with Autoantibodies Against Type I Interferons and Life-Threatening COVID-19 Pneumonia

Evaluation of alkyne-modified isoprenoids as chemical reporters of protein prenylation

Single nucleotide polymorphism at exon 7 splice acceptor site of OAS1 gene determines response of hepatitis C virus patients to 685 interferon therapy

The Flvrencoded murine oligoadenylate synthetase 1b (Oas1b) suppresses 2-5A synthesis in intact cells

Enzymatic activity 690 of 2'-5'-oligoadenylate synthetase is impaired by specific mutations that affect oligomerization of the protein

Structure of Human RNase L Reveals the Basis for Regulated RNA Decay in the IFN Response Science 343

Genetic polymorphisms in host antiviral genes: associations with humoral and cellular immunity to measles vaccine

Peptide specificity of protein prenyltransferases is determined mainly by reactivity rather than binding affinity

OAS proteins and cGAS: unifying concepts in sensing and responding to cytosolic nucleic acids

The 2'-5'-oligoadenylate synthetase 3 enzyme potently synthesizes the 2'-5'-oligoadenylates required for RNase L activation

Mitochondrial localization of the OAS1 p46 isoform associated with a common single nucleotide polymorphism

Evolution of the 2'-5'-oligoadenylate synthetase family in eukaryotes and bacteria

Extracellular 2'-5' 715 oligoadenylate synthetase stimulates RNase L-independent antiviral activity: a novel mechanism of virus-induced innate immunity

DNA tumor viruses oncogenes antagonize the cGAS STING pathway

Viral Generated Inter-Organelle Contacts 720 Redirect Lipid Flux for Genome Replication

Identification of a Sjogren's syndrome susceptibility locus at OAS1 that influences isoform switching, protein expression, and responsiveness to type I interferons

Activation of RNase L is dependent on OAS3 expression during infection with diverse human viruses

Genetic variation in OAS1 is a risk factor for initial infection with West Nile virus in man

Distinct antiviral roles for human 2',5'-oligoadenylate synthetase family members against dengue virus infection

A functional variant in the OAS1 gene is associated with Sjogren's syndrome 735 complicated with HBV infection

The Activation Mechanism of 2'-5'-Oligoadenylate Synthetase Gives New Insights Into OAS/cGAS Triggers of Innate Immunity

Cholesterol manipulation by West 740 Nile virus perturbs the cellular immune response

Small self-RNA generated by RNase L amplifies antiviral innate immunity

The Origin, Dynamic Morphology, and PI4P-Independent Formation of Encephalomyocarditis Virus Replication Organelles

The primary function of RNA binding by the influenza A 750 virus NS1 protein in infected cells: Inhibiting the 2'-5' oligo (A) synthetase/RNase L pathway

The Hepatitis C Virus-Induced Membranous Web and Associated Nuclear Transport Machinery Limit Access of Pattern Recognition Receptors to Viral 755 Replication Sites

OAS1 A multiple sclerosis susceptibility gene that influences disease severity Neurology 75

Genetic mechanisms of critical illness in Covid-19

Genetic mechanisms of critical illness in COVID-19

Membranous replication factories induced by plus-strand RNA viruses

Adaptively introgressed Neandertal haplotype at the OAS locus functionally impacts innate immune responses in humans

The Nature of the Catalytic Domain of 2'-5'-Oligoadenylate Synthetases The

RNA-binding protein isoforms ZAP-S and ZAP-L have distinct antiviral and immune resolution functions

Viral encounters with 2',5'-oligoadenylate synthetase and RNase L during the interferon antiviral response

High Anti-Dengue Virus Activity of the OAS Gene Family Is Associated With Increased Severity of

PDIA3 (red), and 1020 dsRNA (magenta) antibodies. D) Relative quantification of CVB3 RNA in EV, p42, or p46 expressing OAS1 KO 293T cells infected with CVB3 (MOI=0.001, 24h). E) Relative quantification of influenza A virus A/PR/8/34 NP RNA in EV, p42, or p46 expressing OAS1 KO 293T cells infected with IAV A/PR/8/34 (MOI=0.01, 24h). F) Viral titers from OAS1 KO 293T cells expressing EV, p42, or p46

Quantification of VSV-GFP + OAS1 KO 293T cells expressing EV, p42

Data were analyzed using one-way ANOVA with Tukey's multiple comparisons test where **P<0.01 and ****P<0.0001 1030 MATERIALS AND METHODS Cells, cell culture conditions and treatments All cells (Table S2) were incubated at 37° C with 5% CO2. HEK293T, A549, Vero, PH5CH8, Huh7, and HeLa cells were grown in DMEM (Sigma) containing 10% heat-inactivated fetal bovine serum (FBS) (Atlanta Biologicals) and 1% penicillin-streptomycin-glutamine 1035 (Mediatech)

PBL Interferon Source) was used at 200-1000 IU/mL. Generation of knockout cell lines using CRISPR/Cas9 gene editing 1045 Cloning of OAS1 targeting guide RNA (gRNA) 5′-GTGCATGCGGAAACACGTGTCTGG-3′ into pRRLU6-empty-gRNA-MND-Cas9-t2A-Puro vector or RNase L targeting gRNA 5′-GTTATCCTCGCAGCGATTGCGGGG-3′ into pRRLU6-empty-gRNA-MND-Cas9-t2A-Blast was achieved using the In-Fusion enzyme mix (Clontech). OAS1 and RNASEL KO 293T were generated using lentiviral transduction as described previously followed by selection in 2 μg/mL 1050 puromycin or blasticidin

Generation of 293T-ACE2 cells Lentiviral expression vector for ACE2 (pLEX-ACE2) was generated by amplifying the ACE2 sequence from cDNA from Huh7 cells (5′-CGACTCACTATAGGGAGACCCAAGCTTGGTACCGAGCTCGATGGACTACAAAGAC-3′ 1100 and 5′-GTCCAGAGATTTGGCTGGGGTATTTCTG AGATCCATCATGCTTGTCATCGTCATCCTTGTAATCGATG-3′ (Table S1).(Schwerk et al., 2019) and Gibson assembly using primers 5′-CGACTCACTATAGGGAGACCCA AGCTTGGTACCGAGCTCGATGGACTACAAAGAC-3′ and 5′--GTCCAGAGATTTGGCTGGGGTATTTCTGAGATCCATCATGCTTGTCATCGT 1105 CATCCTTGTAATCGATG-3′ (Table S1)

FLAG-tagged expression plasmids encoding p46 common+CTIL, p46 alanine mutant 9, p46 D12aa, p46 D22aa, p46 D32aa, p42 CFK mutant, and p46 CFK mutant were generated by site-directed mutagenesis on pCDNA3.1 FLAG-1110 p46. Site-directed mutagenesis was performed using the QuikChange Lightning kit (Agilent) according to the manufacturer's instructions. All primers used for site-directed mutagenesis are listed in Table S1. OAS1 p46 C-terminal alanine mutants 1-8 and OAS1 p46 C-terminus species hybrids were generated by ligation of mutant gBlocks (Table S1). Briefly, pcDNA3.1 FLAG-OAS1 p46 was cut with KflI and ApaI. gBlocks were PCR-amplified using the following primer 1115 pair 5′-TAAGAATTGGGATGGGTCCCCAG-3′ and 5′-GACACTATAGAATAGGGCCCTCTAGA-3′ and then cut with with KflI and ApaI

GGAA) and transfected with 250 ng/mL pCDNA3.1 FLAG-tagged OAS1 expression constructs 1125 3h after addition of GGAA. 24h after transfection, cells were lysed in Co-IP (50 mM Tris-HCl, pH 7.5; 150 mM NaCl; 0.5% NP40; 1 mM EDTA) buffer and OAS1 proteins were immunoprecipitated from the lysate using 20 µg anti-FLAG antibody (Sigma) and Dynabeads Protein G. After 5 washes in Co-IP buffer, the Dynabeads were resuspended in 50 μl 50 mM Tris-HCl, pH 8, and the click chemistry reaction was performed according to the manufacturer's 1130 instruction. Immunoprecipitated protein were immunoblotted and probed for presence of geranylgeranyl azide-biotin labeling using an HRP-conjugated streptavidin antibody. Virus infections and titer quantification Virus and their sources are listed in Table S5. Encephalomyocarditis virus was grown and titered 1135 in Vero cells. West Nile virus Texas from Gale laboratory was grown as previously described

WNV infected cells were washed three times with PBS and then stained with the secondary antibodies goat antirabbit IgG Alexa Fluor 488 (Themo Fisher), goat anti-mouse IgG1 (Thermo Fisher) and goat anti-mouse IgG2a Alexa Fluor 648

Samples were cured in the dark at room temperature for 24-48h prior to imaging. Samples were imaged using a Nikon Eclipse Ti laser scanning confocal microscope using a 60x oil-immersion lens. Images were processed and analyzed using the NIS elements software and Fiji. Quantification of 1230 co-localization was performed using the Fiji Coloc 2 plugin. RNA isolation, reverse transcription, and RT-qPCR Total RNA was isolated using the NucleoSpin RNA kit (Macherey-Nagel) according to the manufacturer's protocol. cDNA was synthesized from 1 μg total RNA using the QuantiTect RT 1235 kit (Qiagen) according to the manufacturer's instructions. RT-qPCR was carried out using the ViiA7 RT-qPCR system

RNA immunoprecipitation

At 12h post-infection, cells were harvested and lysed in RNA-IP lysis buffer (100 mM KCl, 5 mM MgCl2, 10 mM HEPES pH 7.4, 0.5% NP-40, 1 mM DTT, 1× HALT protease inhibitor, 100 U/mL RNasin and 2 mM 1245 ribonucleoside-vanadyl complex). Nuclei and debris were removed from the cytosolic lysate by centrifugation at 8,000 g at 4 °C for 10 min. Next, 400 μg protein from the cytosolic lysate was incubated with 5 μg anti-FLAG mouse

After washing, coprecipitated RNA was 1250 isolated from IgG1-protein complexes by chloroform-isoamyl alcohol extraction, reverse transcribed into cDNA (QuantiTect RT kit, Qiagen) and analyzed by RT-qPCR. OAS1 in vitro activity assay Enrichment of FLAG-tagged OAS1 isoform proteins prior to in vitro activity assay was 1255 performed as described for the RNA-IP above. Briefly, OAS1 KO 293T cells were seeded on a 10 cm dish and transfected with FLAG-tagged OAS1 isoform expression plasmids and harvested 24h post transfection in Co-IP lysis buffer (50 mM Tris-HCl pH 8, 150 mM NaCl, 0.5% Igepal Ca-630, 1 mM EDTA, 1x HALT protease inhibitor)

10 mM MgCl2, 1 mM DTT, 0.1 mg/mL BSA) supplemented with 400 mM ATP, ~80 nM [a-32 P]-labelled ATP (3000 Ci/mmol 10 mCi/mL, 250 µCi; PerkinElmer), and 33.3 µg/mL poly(I:C) (Invivogen). The reactions were 1265 allowed to proceed for 2 hours at 37 °C. The reactions were then analyzed by denaturing gel electrophoresis on a 20 cm tall 20% polyacrylamide 7 M urea gel with 0.5 x TBE running buffer at 12.5 W. The gels were then applied onto Whatman filter paper

labeled 2′-5′ oligoadenylate products were visualized using a Sapphire Biomolecular 1270 Imager (Azure Biosystems)

GGGCCCTCTAGACTCGAGCTAAAAGGAGGTCTGAACATCATCAGTG-3′. This amplicon was cloned into a pLEX lentiviral backbone cut with BamHI and XhoI using the InFusion HD kit (Takara). pLEX-ACE2 was co-transfected with psPAX2 and pMD2.G into 293FT cells for lentiviral packaging. 293T cells were transduced with ACE2-expressing lentivirus and selected with puromycin (2 mg/mL) for 4 days to generate 293T-ACE2 cells. ACE2 expression was 1065 verified by immunoblot (Proteintech).

Cells were lysed in RIPA buffer (+1× HALT protease and phosphatase inhibitor), and 10-30 μg total protein from whole-cell lysates was run on SDS-PAGE and transferred to polyvinylidene 1070 difluoride membranes (Thermo Scientific). The membranes were blocked in 5% milk in PBS-T (pPBS/Tween 20). Primary antibody (Table S3 ) incubation with antibodies against OAS1 (CST), RNase L (CST), or FLAG (Sigma) were performed in 5% milk in PBS-T overnight at 4°C. Membranes were washed for 5 minutes in PBS-T three times. Secondary antibody incubation was performed in 5% milk in PBS-T at room temperature for one hour and after membranes 1075were washed for 5 minutes in PBS-T three times. Membranes were imaged on a Chemidoc XR system.

Dicer-substrate short interfering RNAs against a common region of OAS1 or the unique 3′UTR 1080 of p42 or p46 were custom designed and procured from Integrated DNA Technologies (Table  S1 ). THP-1 cells were differentiated with PMA for 24h. At 24h post treatment, cells were transfected with 20 nM of siRNA using TransIT-X2 (Mirus Bio) according to the manufacturer's instructions. Viral infections were performed at 24h post-transfection. 1085Cloning Expression plasmids encoding OAS1 p42, p42CTIL, p44, p46, p46ATIL, p48 and p52 were generated by Gibson assembly of the common OAS1 sequence with the isoform-specific sequence into the pCDNA3.1 vector (Table S4 ). Empty pcDNA3.1 was cut using BamHI and XbaI. A Gibson assembly compatible fragment for the common sequence of OAS1 was PCR-1090 amplified from an OAS1 expression plasmid (gift from Dan Stetson) using primers 5′-TGGTACCGAGCTCGATGATGGATCTCAGAA-3′ and 5′-CAGCAGAATCCAGG AGCTCACTGG-3′. Gibson assembly compatible fragments for the unique sequences of p42, p42CTIL, and p44 were generated by PCR amplification of annealed sense and antisense oligos (Table S1 ). Gibson assembly compatible sequences for the unique portions for p46, p46ATIL, 1095 and p48, were generated by PCR amplification of gBlocks. N-terminal FLAG-tagged versions of OAS1 p42 and p46 were generated by cutting pcDNA3.1 OAS1 p42, p42CTIL, p46 and p46ATIL with BamHI and cloning of a 3xFLAG fragment by PCR amplification from pEF FLAG-ZAP-L (Schwerk et al., 2019) and Gibson assembly using primers 5′-

Samples from a cohort of 34 severe COVID-19 cases were collected starting in April 2020 at Virginia Mason Medical Center and Benaroya Research Institute. Severity was based on 1275 hospitalization in the critical care unit with mechanical ventilation. A cohort of 99 healthy control subjects matched for ancestry (self-reported) was assembled from participants in the healthy control registry at Benaroya Research Institute. Both studies were approved by the Institutional Review Board at Benaroya Research Institute (IRB20-036 and IRB07109 respectively). A description of the cohorts is presented in Table S3 . DNA samples from these 1280 subjects were genotyped for OAS1 rs10774671 using a Taqman SNP genotyping assay (Thermo Fisher). Genotypes passed Hardy-Weinberg equilibrium analysis. Association testing was performed using gPLINK v2.050 by logistic regression adjusting for sex and ancestry (race/ethnicity). Replication of genetic association was tested using 1,676 critically ill COVID-19 cases collected through the GenOMICC study in the UK and 8,380 population based controls 1285(1:5 cases:controls) from the UK Biobank samples as described (Pairo-Castineira et al., 2020a) . All subjects were of European descent as determined by ancestry informative markers. DNA was genotyped using the Illumina Global Screening Array v3.0+ multi disease bead chips (Illumina) and subjected to standard quality control filters. Tests for association between cases and controls were performed by logistic regression using PLINK, with sex, age (as of April 1, 2020), 1290 deprivation score of residential postal code, and the first 10 principal components as covariates.

Statistical analyses (other than genetic analysis) were performed with Prism 8 and the specific statistical analyses performed are indicated in the figure legends. 1295