Inborn errors of OAS–RNase L in SARS-CoV-2–related multisystem inflammatory syndrome in children

Multisystem inflammatory syndrome in children (MIS-C) is a rare and severe condition that follows benign COVID-19. We report autosomal recessive deficiencies of OAS1, OAS2, or RNASEL in five unrelated children with MIS-C. The cytosolic double-stranded RNA (dsRNA)–sensing OAS1 and OAS2 generate 2′-5′-linked oligoadenylates (2-5A) that activate the single-stranded RNA–degrading ribonuclease L (RNase L). Monocytic cell lines and primary myeloid cells with OAS1, OAS2, or RNase L deficiencies produce excessive amounts of inflammatory cytokines upon dsRNA or severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) stimulation. Exogenous 2-5A suppresses cytokine production in OAS1-deficient but not RNase L–deficient cells. Cytokine production in RNase L–deficient cells is impaired by MDA5 or RIG-I deficiency and abolished by mitochondrial antiviral-signaling protein (MAVS) deficiency. Recessive OAS–RNase L deficiencies in these patients unleash the production of SARS-CoV-2–triggered, MAVS-mediated inflammatory cytokines by mononuclear phagocytes, thereby underlying MIS-C.

INTRODUCTION: Multisystem inflammatory syndrome in children (MIS-C) is a severe, unexplained complication of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection with an estimated prevalence of~1 per 10,000 infected children. It typically occurs 4 weeks after infection, without hypoxemic pneumonia. Affected children present with fever, rash, abdominal pain, myocarditis, and other clinical features reminiscent of Kawasaki disease, including lymphadenopathy, coronary aneurysm, and high levels of biological markers of acute inflammation. Sustained monocyte activation is consistently reported as a key immunological feature of MIS-C. A more specific immunological abnormality is the polyclonal expansion of CD4 + and CD8 + T cells bearing the T cell receptor Vb21. 3. The root cause of MIS-C and its immunological and clinical features remains unknown. RATIONALE: We hypothesized that monogenic inborn errors of immunity to SARS-CoV-2 may underlie MIS-C in some children. We further hypothesized that the identification of these inborn errors would provide insights into the molecular and cellular mechanisms underlying its immunological and clinical phenotypes. Finally, we hypothesized that a genetic and mechanistic understanding of a few patients would provide a proof of principle that would facilitate studies in other patients. We performed whole-exome or whole-genome sequencing on 558 internationally recruited patients with MIS-C (aged 3 months to 19 years). We searched for rare nonsynonymous biallelic variants of protein-coding genes, testing a hypothesis of genetic homogeneity.

RESULTS:
We found autosomal recessive deficiencies of OAS1 (2′-5′-oligoadenylate syn-thetase 1), OAS2, or RNase L (ribonuclease L) in five unrelated children of four different ancestries with MIS-C (~1% of our cohort). There were no similar defects in a cohort of 1288 individuals (aged 6 months to 99 years) with asymptomatic or mild infection (P = 0.001) or 334 young patients (aged 0 to 21 years) with asymptomatic or mild infection or COVID-19 pneumonia (P = 0.046). The estimated cumulative frequency of these defects in the general population was~0.00013. The type I interferon (IFN)-inducible double-stranded RNA (dsRNA)sensing proteins OAS1 and OAS2 generate 2′-5′linked oligoadenylates (2-5A), which activate the antiviral single-stranded RNA (ssRNA)degrading RNase L, particularly in mononuclear phagocytes. Consistent with the absence of pneumonia in these patients, epithelial cells and fibroblasts defective for this pathway restricted SARS-CoV-2 normally. This contrasted with interferon alpha and beta receptor subunit 1 (IFNAR1)-deficient cells from patients prone to hypoxemic pneumonia without MIS-C. Monocytic cell lines with genetic deficiencies of OAS1, OAS2, or RNase L displayed excessive inflammatory cytokine production in response to intracellular dsRNA. Cytokine production by RNase L-deficient cells was impaired by melanoma differentiation-associated protein 5 (MDA5) or retinoic acid-inducible gene I (RIG-I) deficiency and abolished by mitochondrial antiviral-signaling protein (MAVS) deficiency. Exogenous 2-5A suppressed inflammatory responses to these stimuli in control and OAS1deficient cells but not in RNase L-deficient cells. Finally, monocytic cell lines, primary monocytes, and monocyte-derived dendritic cells with genetic deficiencies of OAS1, OAS2, or RNase L displayed exaggerated inflammatory responses to SARS-CoV-2 as well as SARS-CoV-2-infected cells and their RNA.

Results
Identification of homozygous rare predicted loss-of-function variants of OAS1 or RNASEL in two MIS-C patients We performed whole-exome or whole-genome sequencing for 558 patients with MIS-C from the international COVID Human Genetic Effort (CHGE) cohort (https://www.covidhge.com/) (fig. S1). We first searched for homozygous or hemizygous rare predicted loss-of-function (pLOF) variants with a high degree of confidence in human genes with a gene damage index of <13.83 (41). We then restricted the list to genes involved in host response to viruses (Gene Ontology term "response to virus," GO:0009615). We identified two unrelated patients homozygous for stop-gain variants, of OAS1 in one patient (P1, p.R47*) and RNASEL in the other (P5, p.E265*) (Fig. 1A, fig. S2A, and Table 1). OAS1 (2′-5′-oligoadenylate synthetase 1) is one of the four members of the OAS family (OAS1, OAS2, OAS3, and the catalytically inactive OASL). These proteins are type I IFN-inducible cytosolic proteins that produce 2′-5′-linked oligoadenylates (2-5A) upon binding to double-stranded RNA (dsRNA). The 2-5A, in turn, induce the dimerization and activation of the latent endoribonuclease RNase L, which degrades single-stranded RNA (ssRNA) of viral or human origin (42,43). No homozygous variants fulfilling these criteria were identified in any of the 1288 subjects with asymptomatic or mild SARS-CoV-2 infection (SARS-CoV-2infected controls) in the CHGE database (fig. S1). MIS-C patients therefore display significant enrichment (P = 0.013) in homozygous pLOF variants of the OAS1 and RNASEL genes, suggesting that these loci are probably relevant to MIS-C pathogenesis. Moreover, although OAS1, OAS2, OAS3, and RNase L are expressed in various cell types in mice and humans, their levels are particularly high in myeloid cells, including monocytes and macrophages (44)(45)(46). Thus, autosomal recessive (AR) deficiencies of the OAS-RNase L pathway may underlie MIS-C by impairing the restriction of viral replication and/or enhancing the virus-triggered inflammatory response in monocytes, macrophages, dendritic cells, or other cell types.
Identification of biallelic rare experimentally deleterious variants of OAS1, OAS2, or RNASEL in five MIS-C patients OAS1, OAS2, OAS3, and RNASEL have consensus negative selection (CoNeS) scores for negative selection of 2.25, 0.79, 1.46, and 0.66, respectively, consistent with findings for known monogenic IEIs with an AR mode of inheritance (47). We therefore extended our search to all homozygous or potential compound-heterozygous nonsynonymous or essential splicing site variants with a minor allele frequency (MAF) of <0.01 at these four loci in our MIS-C cohort. We identified a total of 12 unrelated patients and 16 different variants of OAS1, OAS2, OAS3, and RNASEL (Table 1). To study the expression and function of these 16 variants in vitro, we first analyzed RNase L-mediated ribosomal RNA (rRNA) degradation after the cotransfection of RNase L-deficient HeLa M cells with the corresponding OAS1, OAS2, OAS3, or RNASEL cDNAs (48-51) (Fig. 1, B to D, and fig. S2, B and C). The p.R47* OAS1 (homozygous in P1) mutant protein was not produced and was LOF ( Fig. 1B and fig. S2D). The three mutant OAS2 proteins detected (p.R535Q, p.Q258L, and p.V290I) were produced in normal amounts, but p.R535Q (homozygous in P2 and P3) had minimal activity, and p.Q258L and p.V290I (both found in P4) had lower levels of activity than the wild-type (WT) protein (Fig. 1, A and C). All the OAS3 variants were produced in normal amounts, and all but one (p.R932Q, found in the heterozygous state in one patient) of these variants had normal levels of activity ( fig. S2C). The RNASEL p.E265* variant (homozygous in P5) was expressed as a truncated protein and was LOF ( Fig. 1D and fig.  S2E), whereas the p.I264V variant was neutral in expression and function (Fig. 1D). We also quantified the function of the OAS1 and OAS2 mutants in a fluorescence resonance energy transfer (FRET) assay, which confirmed that P1's OAS1 variant was LOF and that the OAS2 variants of P2, P3, and P4 were hypomorphic (21 to 43%, 32 to 76%, and 36 to 75% of WT OAS2 activity for p.Q258L, p.V290I, and p.R535Q, respectively) (Fig. 1, E and F). Thus, we identified five unrelated MIS-C patients homozygous or compound heterozygous for rare and deleterious alleles of three of the four genes controlling the OAS-RNase L pathway ( Fig. 1A and fig.  S2A). The patients' genotypes were confirmed by Sanger sequencing and familial segregation. Their clinical and immunological features were consistent with those previously reported for other MIS-C patients (21,22,(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36)52) (Fig. 1, G to I, and    S1). Thus, there was a significant enrichment in such homozygotes among MIS-C patients relative to infected controls (P = 0.001) or controls under 21 years old (P = 0.046), suggesting that AR deficiencies of three genes of the OAS-RNase L pathway (OAS1, OAS2, and RNASEL) specifically underlie MIS-C. We further assessed the probability of AR deficiencies of these three gene products being causal for MIS-C by evaluating the expression and function of all nonsynonymous variants of OAS1, OAS2, and RNASEL for which homozygotes were reported in the Genome Aggregation Database (gnomAD, v2.1.1 and v3.1.1, 28 variants in total) in our RNase L-mediated rRNA degradation assay ( fig. S2, F to H, and table S1). In total, 13 OAS1, OAS2, or RNASEL variants were deleterious and present in the homozygous state in 19 individuals in the gnomAD database ( Fig. 1, J to L). The estimated cumulative frequency of homozygous carriers of LOF variants at the three loci was~0.00013 [95% confidence interval (CI): 7.2 × 10 −5 to 20 × 10 −5 ] in the general population. The rarity of AR OAS-RNase L deficiencies in the general population is therefore consistent with that of MIS-C. Moreover, the enrichment in these deficiencies observed in MIS-C patients relative to the individuals included in gnomAD was highly significant (P = 2 × 10 −6 ). These findings suggest that AR deficiencies of OAS1, OAS2, and RNase L are genetic etiologies of MIS-C.
The expression pattern for the OAS-RNase L pathway implicates mononuclear phagocytes We studied the basal expression of OAS1, OAS2, OAS3, and RNASEL in cells from different tissues. Consistent with data from public databases (44), our in-house human cell RNA sequencing (RNA-seq) and reverse transcriptionquantitative polymerase chain reaction (RT-qPCR) data showed that myeloid blood cells had higher basal mRNA levels for the four genes than did the tissue-resident cells tested (Fig. 2, A and B). In all cell types studied, both type I and type II IFN treatments up-regulated the levels of mRNA for OAS1, OAS2, and OAS3, whereas the levels of RNASEL mRNA were not influenced by these IFNs (fig. S3A). Previous studies reported a relationship between cell type-dependent activation of the OAS-RNase L pathway and basal levels of expression in mice (45,46). MIS-C occurs 3 to 6 weeks after SARS-CoV-2 infection, but the virus and/or viral proteins may still be detectable in nonrespiratory tissues, such as the intestine or heart, at disease onset in some patients (32,34,37). In addition, CD4 + and CD8 + T cells carrying Vb21.3 expand, which implies a superantigen-like viral driver of MIS-C (32,34,(36)(37)(38) and suggests that the virus or its antigens persist. Thus, AR deficiencies of the OAS-RNase L pathway may underlie MIS-C by impairing SARS-CoV-2 restriction and/or enhancing virus-triggered inflammatory responses in monocytes and other mononuclear phagocytes.

OAS-RNase L deficiencies have no impact on SARS-CoV-2 replication in A549 epithelial cells and fibroblasts
Previous studies have shown that the overproduction of exogenous OAS1 can result in the restriction of SARS-CoV-2 replication in A549 lung epithelial cells in the absence of exogenous type I IFN (53,54). However, the five OAS-RNase L-deficient patients had MIS-C without pneumonia. We assessed SARS-CoV-2 replication in A549 cells rendered permissive to SARS-CoV-2 by the stable expression of angiotensin-converting enzyme 2 (ACE2) and transmembrane protease serine 2 (TMPRSS2), which facilitates viral entry. Knockout (KO) of OAS1 or OAS2 did not increase the proportion of SARS-CoV-2-infected cells at 24 or 48 hours relative to that for the parental WT A549 cells, regardless of the presence or absence        ). Only RNase L KO cells resulted in a mild increase in susceptibility to SARS-CoV-2 relative to WT cells in the absence of IFN-a2b, consistent with previous findings (55). We also used patient-specific SV40-transduced human dermal fibroblasts (SV40-fibroblasts) stably expressing ACE2 as a surrogate cell type for studying the impact of OAS-RNase L deficiencies on tissue-resident cell-intrinsic defenses against SARS-CoV-2 (5). Consistent with the lack of pneumonia in these patients, no increase in SARS-CoV-2 susceptibility was observed in any of the fibroblasts with OAS1 (from P1), OAS2 (P3 and P4), or RNASEL (P5) mutations up to 72 hours after infection in the presence or absence of exogenous IFN-a2b, despite the complete loss of expression of OAS1 or RNase L in the cells of P1 and P5, respectively (Fig. 2, E and F). This contrasted with the increased susceptibility reported for fibroblasts from a patient with AR complete IFNAR1 deficiency (56) and critical COVID-19 pneumonia.

OAS-RNase L deficiencies have no impact on SARS-CoV-2 replication in THP-1 cells
Only abortive SARS-CoV-2 infection has been reported in human mononuclear phagocytes, including monocytes and macrophages, which express very little to no ACE2 (57-59). However, basal Oas and Rnasel expression levels have previously been correlated with murine coronavirus or vesicular stomatitis virus (VSV) restriction in mouse macrophages (60). We tested the hypothesis that deficiencies of OAS-RNase L might result in productive SARS-CoV-2 infection in mononuclear phagocytes by assessing the replication of SARS-CoV-2. Unlike WT A549 cells stably transduced with ACE2 and TMPRSS2, in which SARS-CoV-2 can be detected 24 hours after infection, no SARS-CoV-2 was detected in THP-1-derived macrophages (61), whether parental or with a KO of OAS1, OAS2, or RNase L (Fig. 2, G and H, and fig. S3C). Thus, no myeloid Table 2. Demographic and clinical information for MIS-C patients biallelic for deleterious variants of the OAS-RNase L pathway. IEI, inborn error of immunity; SCV2, SARS-CoV-2; IVIG, intravenous immunoglobulins; ND, not determined; CRP, C-reactive protein; sCD25, soluble IL-2Ra.      SARS-CoV-2 replication was detected in the presence or absence of deficiencies of the OAS-RNase L pathway, at least in this cellular model of mononuclear phagocytes (60).

OAS-RNase L deficiencies result in an exaggerated inflammatory response to intracellular dsRNA in THP-1 cells
Sustained monocyte activation has repeatedly been reported to be a key immunological feature of MIS-C (22,(31)(32)(33)(34)(35)(36). We studied the impact of OAS-RNase L deficiencies on cellular responses to intracellular (cytosolic) or extracellular (endosomal) stimulation with dsRNA in THP-1 cells. Consistent with a previous study (62)

The inflammatory response to intracellular dsRNA in THP-1 cells is MAVS dependent
Intracellular dsRNA is known to stimulate the RIG-I/MDA5-MAVS pathway (RIG-I, retinoic acid-inducible gene I; MDA5, melanoma differentiation-associated protein 5; MAVS, mitochondrial antiviral-signaling protein), inducing type I IFNs and other cytokines in various cell types (63), in addition to the OAS-RNase L pathway (42,64). Indeed, unlike WT THP-1 cells, MAVS KO THP-1 cells did not respond to intracellular poly(I:C) stimulation, and RNASEL gene KDn did not result in enhanced activation (Fig. 3C and fig. S5, D and E), confirming that the response to poly(I:C) is dependent on MAVS-mediated signaling in these cells. The enhancement of the intracellular poly(I:C) response after RNASEL KDn was partially attenuated in RIG-I or MDA5 KO THP-1 cells (Fig. 3C and fig. S5, D and E), suggesting that both dsRNA sensors may be involved. Another dsRNA agonist that specifically activates RIG-I, 5′ triphosphate doublestranded RNA (5′ppp-dsRNA), induced enhanced responses in RNase L KO THP-1 cells similar to those seen with poly(I:C) (Fig. 3D). By contrast, the activation of other sensing pathways, including the extracellular ssRNA-sensing toll-like receptor 7 (TLR7) and TLR8 pathways (R848), the TLR4 pathway (LPS), and the intracellular DNA agonist-sensing DAI pathway (ISD), resulted in responses in RNase L KO or KDn THP-1 cells that were similar to those of the parental WT cells (Fig. 3D and  fig. S5F). Thus, the exaggerated inflammatory responses to cytosolic dsRNA observed in THP-1 cells deficient for OAS-RNase L appear to require RIG-I/MDA5 sensing and MAVS activation.
Activation of the OAS-RNase L pathway can suppress inflammatory responses in THP-1 cells Intracellular dsRNA stimulates both the RIG-I/ MDA5-MAVS and OAS-RNase L pathways (42,63,64). We therefore investigated whether the dsRNA-sensing MAVS-dependent signaling pathway was itself hyperactivated as a result of OAS-RNase L deficiency. After intracellular poly(I:C) stimulation, interferon regulatory factor 3 (IRF3) and nuclear factor kB (NF-kB) phosphorylation levels were similar in RNase L KO and WT THP-1 cells (Fig. 3E). Thus, the molecular mechanisms by which OAS-RNase L deficiency results in an exaggerated inflammatory response appears to involve an impairment of RNase L activation resulting in a lack of host RNA transcriptional and/or translational inhibition (65)(66)(67)(68), rather than a hyperactivation of the MAVS-dependent pathways. Consistent with this hypothesis, treatment with exogenous 2-5A, which is normally generated by OASs upon dsRNA sensing and activates RNase L (42,43), rescued the inflammatory phenotype in OAS1 KO THP-1 cells after intracellular poly(I:C) stimulation (Fig. 3F). By contrast, dephosphorylated 2-5A, which is unable to activate RNase L (69,70), had no such effect ( fig.  S5G). Moreover, exogenous 2-5A treatment decreased the response to TLR7/8 activation in WT THP-1 cells (Fig. 3G). Treatment with 2-5A had a much weaker effect or even no suppressive effect in RNase L KDn or KO THP-1 cells (Fig. 3F and fig. S5, G and H). Thus, the exaggerated inflammatory response in OAS-RNase L-deficient mononuclear cells appears to result from the activation of the MAVS-dependent pathway (but not of other nucleic acid-sensing pathways) and an impairment of RNase L activation by OAS1-or OAS2-derived 2-5A after dsRNA sensing. This imbalance creates a phenotype that is probably a consequence of an im-pairment of the posttranscriptional activities of RNase L (65-68).

OAS-RNase L deficiencies result in an exaggerated inflammatory response to SARS-CoV-2 in THP-1 cells
We investigated whether OAS-RNase L deficiencies resulted in exaggerated inflammatory responses to SARS-CoV-2 in mononuclear phagocytes. Bulk RNA-seq on THP-1 cells with KO of OAS1, OAS2, or RNase L stimulated with intracellular poly(I:C) or SARS-CoV-2 revealed transcriptomic profiles different from those of the parental cells (Fig. 4, A and B, and fig.  S6A). Gene set enrichment analysis (GSEA) against Hallmark gene sets (71) revealed an enrichment in genes relating to inflammatory responses and IFN-g signaling in OAS-RNase L-deficient cells, showing that these cells displayed an exacerbated inflammatory response not only to synthetic dsRNA but also to SARS-CoV-2 ( Fig. 4, C and D). Moreover, RNase L KO THP-1 cells had higher levels of IL-6 and CXCL10 secretion than WT cells when cocultured with SARS-CoV-2-infected Vero cells, which support SARS-CoV-2 replication (72, 73) ( Fig. 4E and fig. S6, B and C). Bulk RNA-seq further confirmed this observation at the transcriptome level (Fig. 4F and fig. S6D), revealing an enrichment in the expression of genes relating to inflammatory responses and IFN-a signaling in RNase L KO cells relative to WT cells (Fig. 4G). In addition, transfection with total RNA from SARS-CoV-2-infected Vero cells, but not from uninfected Vero cells, also induced enhanced responses in RNase L KO THP-1 cells relative to parental WT cells, with an enrichment in genes relating to inflammatory responses and IFN-g signaling ( Fig. 4H  and fig. S6E). These findings suggest that OAS-RNase L deficiency results in excessive inflammatory responses in mononuclear phagocytes following both abortive SARS-CoV-2 infection and coculture with SARS-CoV-2-replicating cell types. This is likely due to defective activation of the OAS-RNase L pathway following the engulfment of the virus or infection-related by-products, leading to the release of dsRNA into the cytosol (73).
OAS-RNase L deficiencies result in an enhanced inflammatory response to intracellular dsRNA in primary mononuclear cells P2 (OAS2 deficient), P3 (OAS2 deficient), and P5 (RNase L deficient) secreted larger amounts of the inflammatory cytokines studied than cells from healthy controls ( Fig. 5A and fig.  S7B). This enhanced inflammatory response to intracellular poly(I:C) stimulation was monocyte dependent, as the depletion of monocytes from the PBMCs of healthy controls strongly decreased this response ( fig. S7C). Moreover, the shRNA-mediated KDn of OAS1, OAS2, or RNASEL in monocyte-derived dendritic cells (MDDCs) from healthy controls resulted in an enhanced inflammatory response to intracellular poly(I:C) stimulation, as shown by the higher levels of inflammatory cytokines, including IFN-l1, IL-6, TNF, and IL-12, than were observed with WT parental cells (Fig.  5B). Thus, deficiencies of the OAS-RNase L pathway also result in exaggerated inflammatory responses to intracellular dsRNA stimulation in primary mononuclear phagocytes, or at least in monocytes and MDDCs.

Enhanced myeloid cell activation by SARS-CoV-2 in patient PBMCs
We studied the impact of OAS-RNase L deficiencies on the responses of the various PBMC populations to SARS-CoV-2 by performing single-cell RNA sequencing (scRNA-seq) on PBMCs from P1 (OAS1), P2 (OAS2), P3 (OAS2), and P5 (RNase L) and comparing the results with those for healthy controls. Regardless of genotype, 6 hours of stimulation with SARS-CoV-2 induced a strong immune response across all five major immune cell types including myeloid, B, CD4 + T, CD8 + T, and natural killer (NK) cells (Fig. 5C), with 1301 unique differentially expressed genes (DEGs) (data S1). OAS-RNase L deficiency significantly changed the response of 48 to 94% of the DEGs in each lineage, with myeloid cells being the most affected. Cellular responses were generally stronger in the OAS-RNase L-deficient patients and were essentially limited to the IFN-a and IFN-g response pathways. Myeloid cell responses were characterized by a distinct proinflammatory component, such as IL1B and CCL3, that was stronger in OAS-RNase L-deficient cells ( Fig.  5D and data S2). We then calculated pseudobulk estimates by cell type. Consistent with the single-cell observations, genes strongly up-regulated by SARS-CoV-2 in OAS-RNase Ldeficient myeloid cells were enriched in types I and II IFN signature genes and TNF signature genes, whereas those strongly up-regulated in CD4 + T cells were enriched in type I IFN signature genes (Fig. 5E). Thus, there is an exaggerated inflammatory response to intracellular dsRNA or extracellular SARS-CoV-2 stimulation in primary monocytes and other mononuclear phagocytes with deficiencies of the OAS-RNase L pathway cultured alone or with other PBMC populations. This provides a plausible pathogenic mechanism for MIS-C, in which this condition is driven by the exacerbated activation of mononuclear phagocytes. This hypothesis is also supported by scRNA-seq on PBMCs from P5 (RNase L deficient) collected during MIS-C and the convalescence period. Enhanced expression levels were observed for IFN-a, IFN-g, or TNF signature genes in monocytes, myeloid dendritic cells (mDCs), B lymphocytes, plasmacytoid dendritic cells (pDCs), and activated T cells of P5 relative to healthy pediatric controls (Fig. 5, F and G, and fig.  S8, A to D). Quantitatively inferred cell-cell communications (74) revealed that MIS-C in the RNase L-deficient patient was probably driven by a signal from hyperactivated monocytes and mDCs directed at CD8 + ab T cells (Fig. 5, H and I, and fig. S8, E to G). This situation differs from that observed in patients with COVID-19 pneumonia without MIS-C but is similar to reports for previously described MIS-C patients ( fig. S9) (33,34,36), identifying exaggerated myeloid cell activation due to OAS-RNase L deficiency as the core driver of the immunological and clinical phenotypes of MIS-C in our patients.

Discussion
We report AR deficiencies of OAS1, OAS2, and RNase L as genetic etiologies of MIS-C in five unrelated children, corresponding to ∼1% of the international cohort of patients studied. OAS-RNase L-deficient monocytic cell lines, monocyte-derived dendritic cells modeling patient genotypes, and primary monocytes from patients displayed excessive inflammatory responses to intracellular dsRNA, SARS-CoV-2, SARS-CoV-2-infected cells, and their RNA, providing a plausible mechanism for MIS-C. In these patients, MIS-C may result primarily from an excessive response of monocytes and other mononuclear phagocytes to SARS-CoV-2 dsRNA intermediates or by-products, followed by the presentation of a viral superantigen to T cells, resulting in the activation and expansion of Vb21.3 + CD4 + and CD8 + T cells. The molecular basis of the exacerbated inflammatory response to SARS-CoV-2 due to OAS-RNase L deficiency in mononuclear phagocytes involves an impairment of the activation of RNase L by the dsRNA-sensing molecules OAS1 and OAS2, probably resulting in defective posttranscriptional RNase L activity (67, 68) and the unchecked RIG-I/MDA5-MAVS-mediated production of inflammatory cytokines. Alternative molecular mechanisms cannot be excluded (64,75). The SARS-CoV-2-related RNA products that trigger phagocyte activation, the viral superantigen(s) that activate T cells, and the human leukocyte antigen (HLA) restriction elements all remain to be discovered. Our findings also do not exclude the possibility that AR OAS-RNase L deficiency additionally affects antiviral responses in cells of other tissues injured during MIS-C, such as cardiomyo-cytes, enterocytes, and endothelial cells. The role of this pathway in T cells themselves merits further investigation. MIS-C in other patients may result from IEIs that may or may not be related to the OAS-RNase L pathway. Our findings also suggest that other forms of Kawasaki disease may be caused by other virus-specific IEIs in other patients (15). The notion that the OAS-RNase L pathway is essential for antiviral immunity in mononuclear phagocytic cells was first proposed nearly 40 years ago (60). Intriguingly, the OAS-RNase L pathway is apparently dispensable for protective immunity to SARS-CoV-2 in the respiratory tract. None of the five MIS-C patients had a pulmonary phenotype, and no viral replication was detectable in the upper respiratory tract of any of the five children at the onset of MIS-C. Nevertheless, genomewide association studies have suggested that common variants in the vicinity of OAS1 may be weakly associated with COVID-19 severity (10,11,53,(76)(77)(78)(79). Our finding that the human OAS-RNase L pathway is crucial for regulation of the mononuclear phagocyte response to SARS-CoV-2, but not for SARS-CoV-2 restriction in the respiratory tract, suggests that the main protective action of this pathway is mediated by the control of phagocyte-driven systemic inflammation at a later stage of disease rather than viral restriction in the respiratory tract early on. These findings are also consistent with the discovery of germline gain-of-function OAS1 mutations in humans with an autoinflammatory syndrome involving myeloid cells (80,81).
The five patients, now aged 1 to 15 years, are normally resistant to diseases caused by other common viruses. Since the discovery of the OAS-RNase L pathway in the 1970s (65,82,83), this pathway has been one of the most intensively studied type I IFN-inducible pathways (42,84). Biochemically, the three OASs have different subcellular distributions and different dsRNA optima for activation, they synthesize 2-5A of different lengths (42,85), and they appear to have antiviral activity against different viruses (86)(87)(88). The only well-established function of 2-5A is the activation of RNase L (66), and any of the three OASs appears to be sufficient for the biochemical activation of RNase L in human cells in vitro. RNase L has been shown to have antiviral activity against certain viruses (dengue virus and Sindbis virus), but not others (Zika virus), in murine and human cells in vitro (85,89). In vivo RNase L deficiency in mice drives susceptibility to various viruses (e.g., encephalomyocarditis virus, coxsackievirus B4, murine coronavirus, etc.) (45,85). Our data suggest that human OAS1, OAS2, and RNase L are each essential for the correct regulation of immunity to SARS-CoV-2 but are otherwise largely redundant in natural conditions of infection. It is also clear that the RNase L-dependent functions of OAS1 and OAS2 are crucial for the regulation of immunity to SARS-CoV-2 within the same cells, as the genetic deficiency of any of these three components results in the same immunological and clinical phenotype, namely MIS-C.

Materials and methods Patients
We enrolled an international cohort of 558 MIS-C patients (aged 3 months to 19 years, 60.4% boys and 39.6% girls) originating from Europe, Africa, Asia, and America and living in 16 different countries. All patients met the WHO diagnostic criteria for MIS-C (52). We focus here on five of these patients (P1 to P5). Written informed consent was obtained Lee    The five patients with MIS-C and AR deficiencies of the OAS-RNase L pathway-two boys and three girls-ranged in age from 3 months to 14 years at the time of diagnosis and all fulfilled the WHO criteria for MIS-C (Table 2) (52). They originated from the Philippines (P1), Spain (P2), Turkey (P3 and P4), and Canada (of French descent) (P5) and lived in Spain, Turkey, and Canada. P1 (OAS1 mutation) (29), P3 (OAS2), and P4 (OAS2) had a severe course of MIS-C, with coronary aneurysm, myocarditis, and polyneuropathy, respectively. P2 (OAS2) and P5 (RNASEL) had a milder course of MIS-C, with a typical Kawasaki disease presentation. None of these patients presented any clinical or radiological evidence of pneumonia. Cytokine profiling of serum obtained from P1, P2, and P5 during MIS-C revealed high levels of IFN-g, soluble CD25, IL-18, IL-1RA, and MCP1 (CCL2) (Fig. 1G), consistent with previously published immune profiles of MIS-C and in contrast to those for pulmonary COVID-19 (21). Bulk mRNA sequencing (RNA-seq) of whole-blood RNA from P1 and P2 collected during the MIS-C phase revealed transcriptomic signatures clearly different from those of healthy controls and a pediatric case of acute COVID-19 pneumonia, but similar to those of previously reported MIS-C patients (Fig. 1H) (33). T cell receptor Vb repertoire analysis confirmed the expansion of TRBV 11-2 (encoding Vb21.3) in one of the three MIS-C-phase samples available (P5, with AR RNase L deficiency) (Fig. 1I). The clinical and immunological features of the five patients were, therefore, consistent with those previously reported for other MIS-C patients (21,22,(26)(27)(28)(29)(30)(31)(32)(33)(34)(35)(36).
Whole-exome, whole-genome, and Sanger sequencing Genomic DNA was extracted from whole blood. Whole-exome sequencing (WES) or wholegenome sequencing (WGS) was performed at several sequencing centers, including the Genomics Core Facility of the Imagine Institute (Paris, France), the Yale Center for Genome Analysis (USA), the New York Genome Center (NY, USA), the American Genome Center (TAGC, Uniformed Services University of the Health Sciences, Bethesda, USA), and the Genomics Division-Institute of Technology and Renewable Energies (ITER) of the Canarian Health System sequencing hub (Canary Islands, Spain). More technical details are provided in the supplementary materials. For the Sanger sequencing of OAS1, OAS2, and RNASEL variants, the relevant regions of OAS1, OAS2, and RNASEL were amplified by PCR, purified by ultracentrifugation through Sephadex G-50 Superfine resin (Amersham-Pharmacia-Biotech), and sequenced with the Big Dye Terminator Cycle Sequencing Kit on an ABI Prism 3700 apparatus (Applied Biosystems).

Whole-exome sequencing data analysis
We performed an enrichment analysis focusing on the three candidate genes in our cohort of 558 MIS-C patients and 1288 children and adults with asymptomatic or paucisymptomatic SARS-CoV-2 infection (controls). We considered variants that were predicted to be loss-of-function or missense, with a highest population MAF < 0.01, not included in segmental duplication regions (gnomAD v2.1.1). We considered genes corresponding to the Gene Ontology term "response to virus" (GO:0009615), with a gene damage index of <13.83 (41), corresponding to the 90% least-damaged genes. We searched for all homozygous variants in MIS-C patients, SARS-CoV-2-infected controls, and the gnomAD database. We compared the proportions of patients and controls carrying experimentally confirmed deleterious homozygous variants by means of a logistic regression model, accounting for the ethnic heterogeneity of the cohorts by including the first five principal components of the principal components analysis (PCA), and for data heterogeneity (WGS and WES with various kits and calling processes) by including the two first PCs of a PCA on individual sequence-quality parameters, as previously described (9). The PCA for ethnic heterogeneity was performed with PLINK (v1.9) on WES and WGS data, with the 1000 Genomes Project phase 3 public database as a reference, using >15,000 exonic var-iants with a MAF > 0.01 and a call rate > 0.99. The PCA for data heterogeneity was performed with the R FactoMineR package and the following individual sequence quality parameters calculated with bcftools stats: number of alleles, number of ALT alleles, number of heterozygous variants, Ts/Tv ratio, number of indels, mean depth of coverage, number of singletons, and number of missing genotypes. We also compared the frequency of experimentally confirmed deleterious homozygous variants of the three genes between our MIS-C cohort and gnomAD using Fisher's exact test.

Cell culture
Primary cultures of human fibroblasts were established from skin biopsy specimens from patients or healthy controls. They were transformed with an SV40 vector, as previously described (56), to create immortalized SV40fibroblast cell lines. SV40-fibroblasts, human embryonic kidney 293T (HEK293T) cells, and A549 cells were cultured in Dulbecco's modified essential medium (DMEM; GIBCO) with 10% fetal bovine serum (FBS) (GIBCO). THP-1 cells were cultured in RPMI 1640 medium (GIBCO) with 10% FBS. For the generation of phorbol-12-myristate-13-acetate (PMA)-primed THP-1-derived macrophages, THP-1 cells were incubated with 50 ng/ml of PMA for 48 hours then left without PMA overnight before stimulation. PBMCs were cultured in RPMI 1640 medium (GIBCO) with 10% FBS. For intracellular poly(I:C) or SARS-CoV-2 stimulation of the PBMCs, blood samples were obtained from the OAS-RNase L-deficient patients 2 months to 1 year after acute-phase MIS-C and from five healthy controls with (two pediatric controls and one adult control) or without (one pediatric control and one adult control) prior asymptomatic or mild SARS-CoV-2 infection ∼6 months before sample collection. For the differentiation of monocyte-derived dendritic cells, monocytes were isolated from PBMCs with the Pan Monocyte Isolation kit (Miltenyi Biotec) and cultured with 50 ng/ml of recombinant human granulocyte-macrophage colony-stimulating factor (GM-CSF; PeproTech) and 20 ng/ml of recombinant human IL-13 (PeproTech) for 7 days before cell stimulation experiments.

Plasmids
For overexpression studies in HEK293T cells, WT cDNAs for OAS1 and RNASEL in a pCMV6 backbone were purchased from Origene. For rRNA degradation assays, human OAS1 (GenBank accession no. BC071981.1), OAS2 (GenBank accession no. BC049215.1), OAS3 (GenBank accession no. BC113746), and RNASEL (GenBank accession no. L10381.1) cDNAs were inserted into p3X-FLAG-CMV-10 (Sigma) as previously described (75,88). Patient-specific variants or variants from the gnomAD database were generated by site-directed mutagenesis PCR with the Super Pfx DNA Polymerase (CWbio). For stable lentivirus-mediated transduction with ACE2 and RNASEL, cDNAs for WT and patient-specific ACE2 or RNASEL variants were inserted into pTRIP-SFFV-CD271-P2A, a modified pTRIP-SFFV-mtagBFP-2A (Addgene 102585) in which mtagBFP is replaced with CD271, with InFusion (Takara Bio), according to the manufacturer's instructions. We used the XhoI and BamHI restriction sites. For stable lentivirus-mediated transduction with OAS1 and OAS2, cDNAs for WT and patient-specific OAS1 or OAS2 variants were inserted into a modified pSCRPSY vector (KT368137.1) with a PaqCI cutting site expressing blue fluorescent protein (BFP). The PaqCI site was used for cDNA insertion with InFusion. We checked the entire sequences of the OAS1, OAS2, OAS3, and RNASEL cDNAs in the plasmids by Sanger sequencing.

Cell-free system assays of OAS and RNase L activity
Assays for OAS and RNase L activity were performed with a modified cell-free system assay based on HeLa M cells (49,50). The HeLa M cells were cultured in DMEM with 10% FBS, and their identity was confirmed by the presence of short tandem repeat loci with a 94.12% match to HeLa cells (ATCC CCL2, Genetica, Burlington, NC). We previously reported that HeLa M cells have no RNase L expression (51). Cells were plated in 24-well dishes (6 × 10 4 cells per well) with empty vector (p3X-FLAG-CMV-10) or vector containing WT or mutant human OAS1 (GenBank accession no. BC071981.1), OAS2 (GenBank accession no. BC049215.1), OAS3 (GenBank accession no. BC113746), or RNASEL (GenBank accession no. L10381.1) cDNAs. HeLa M cells were cotransfected with cDNAs in the presence of Lipofectamine 2000 for 20 hours. Conditions were optimized for each type of enzyme assayed. RNase L assays were performed on cells cotransfected with 300 ng of WT or mutant RNASEL cDNA and 100 ng of WT OAS3 cDNA. OAS1 assays were performed with 300 ng of OAS1 cDNA and 100 ng of RNASEL cDNA. OAS2 assays were performed with 300 ng (condition 1) or 600 ng (condition 2) of OAS2 cDNA and 100 ng of RNASEL cDNA, and OAS3 assays were performed with 300 ng of OAS3 cDNA and 100 ng of RNASEL cDNA. The lysis-activation-reaction (LAR) buffer contained 0.1% (by volume) Nonidet P-40, 50 mM Tris-HCl pH 7.5, 0.15 M NaCl, 2 mM EDTA, 10 mM MgCl 2 , 2 mM ATP, 400 U/ml of RNaseOUT (Thermo Fisher Scientific), and 2.5 mg/ml of poly(I): poly(C) (Millipore catalog no. 528906). LAR buffer (75 ml) was added to each well of cells on ice and the contents of the wells were then transferred to tubes on ice. The lysates were then incubated at 30°C for 30 min, except in OAS2 assays, for which lysates were incubated at 37°C (condition 1) or 30°C (condition 2) for 40 and 50 min, respectively. Total RNA was isolated with RLT buffer supplemented with guanidinium isothiocyanate and the EZ-10 Spin Columns Total RNA Minipreps Super kit (BIO BASIC). RNA was separated on RNA chips with an Agilent Bioanalyzer 2000, from which images and RNA integrity numbers (RINs) were obtained. For immunoblots, aliquots of the lysates (10 mg of protein) were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE) in a 7% acrylamide gel. Immunoblots were probed with a monoclonal antibody against the Flag epitope or b-actin (Sigma-Aldrich).

FRET-based OAS enzyme assays
FRET assays of the amount of 2-5A synthesized by WT and mutant isoforms of OAS1 or OAS2 were performed with lysates of transfected HeLa M cells (90). Cells were plated in 24-well dishes (6 × 10 4 cells per well), cultured for 24 hours and transfected for 20 hours with Lipofectamine 2000 transfection reagent (Thermo Fisher Scientific) and 0.5 mg empty vector (p3X-FLAG-CMV-10), or 500 ng of vector containing WT or mutant OAS1 or OAS2. Cells were washed with cold PBS and then lysed with 100 ml of LAR buffer [containing ATP and poly(I:C)] per well on ice. The lysates were transferred to tubes on ice and incubated at 30°C for 50 min before heating at 95°C for 10 min (to stop the reaction and denature proteins) and vortexing twice. The lysates were centrifuged at 12,000g for 10 min. The supernatants were then collected and diluted 10-fold in H 2 O. Diluted samples (2 ml) were added to 45 ml of cleavage buffer (25 mM Tris-HCl, pH 7.4, 0.1 M KCl, 10 mM MgCl 2 , 50 mM ATP pH 7.4, and 7 mM b-mercaptoethanol) containing 40 nM RNase L and 135 nM FRET probe in 96-well plates. The probe used was a 36-nucleotide synthetic oligoribonucleotide probe with multiple RNase L cleavage sites, a fluorophore (6-FAM or 6-carboxyfluorescein) at the 5′ terminus, and the black hole quencher-1 (BHQ1) at the 3′ terminus (IDT, Inc.) (90). FRET assays were performed at room temperature, every 5 min, for 30 min. Fluorescence was measured in relative fluorescence units (RFU), with excitation at 485 nm and emission at 535 nm, with a Varioskan LUX multimode microplate reader and Skanit version 6.0.1 software (Thermo Fisher Scientific). There were six biological replicates for each treatment group. Standard curves were plotted in triplicate with 0.1 to 30 nM ppp5′A2′p5′A2′p5′A (trimer 2-5A) synthesized with isolated OAS1 and purified by high-performance liquid chromatography (HPLC) (70).

Cytokine quantification in plasma samples
Cytokine quantification in plasma samples was performed as previously described (32). Briefly, whole blood was sampled into EDTA tubes. The plasma concentrations of IFN-g, IL-1RA, IL-10, IL-18, IL-6, MCP-1, soluble CD25, and TNF were then determined with Simpleplex technology and an ELLA instrument (Protein Simple) according to the manufacturer's instructions. Plasma IFN-a concentrations were determined with a single-molecule array (Simoa) on an HD-1 Analyzer (Quanterix) with a commercial kit for IFN-a2 quantification (Quanterix). Blood samples from P1, P2, and P5 were obtained on days 7, 4, and 9 after symptom onset, respectively.

TRBV 11-2 relative expression levels
Whole blood was collected into PAXgene (BD Biosciences) or Tempus (Thermo Fisher Scientific) blood RNA tubes or EDTA tubes. RNA was extracted with the corresponding RNA extraction kits or with the Maxwell 16 LEV Blood RNA kit and a Maxwell extractor (Promega) and quantified by spectrometry (Nanovue). For P5, RNA was extracted from sorted T cells with the RNeasy Plus microkit (Qiagen). Relative expression levels were determined for TRBV 11-2 with nCounter analysis technology (NanoString Technologies), by calculating TRBV 11-2 mRNA levels relative to other TRBV mRNA levels and normalizing against the median value for the healthy volunteer group. Blood samples from P1, P2, and P5 were obtained on days 7, 4, and 9 after symptom onset, respectively.

Immunoblots
Total protein extracts were prepared by lysing cells in NP40 lysis buffer (150 mM NaCl, 50 mM Tris pH 8.0, and 1.0% NP40) supplemented with cOmplete Protease Inhibitor cocktail (Roche, Mannheim, Germany). Equal amounts of protein from each sample were subjected to SDS-PAGE, and the proteins were blotted onto polyvinylidene difluoride membranes (Bio-Rad). The membranes were then probed with the desired primary antibody followed by the appropriate secondary antibody. Primary antibodies against the following targets were used:

RT-qPCR
Total RNA was extracted from THP-1 cells and various other cell types with the Quick-RNA MicroPrep kit (Zymo Research). RNA was reverse-transcribed with random hexamers and the Superscript III first-strand cDNA synthesis system (Invitrogen). Quantitative real-time PCR was then performed with the TaqMan universal PCR master mix (Applied Biosystems). For gene expression assays, TaqMan probes for OAS1, OAS2, OAS3, RNASEL, IL6, and CXCL9 were used (Thermo Fisher Scientific). We used b-glucuronidase (GUSB) for normalization (Applied Biosystems). The results were analyzed with the DCt or DDCt method. For SARS-CoV-2 genomic RNA quantification, RNA was extracted from 3 × 10 5 THP-1 cells infected with SARS-CoV-2 for 24 hours. Cells were washed three times with PBS and lysed for RNA extraction. Equal amounts of total RNA were reverse-transcribed with random hexamers and the Superscript III first-strand cDNA synthesis kit (Invitrogen). Equal amounts of cDNA were used for the qPCR reaction. Primers and probes for the N gene (N2 region), the RNA-dependent RNA polymerase (RdRP) gene, and their respective standards were purchased from IDT technologies. All qPCR reactions were analyzed with the QuantStudio 3 system.

Gene knockdown
The  (200226261 and 200226578), or a nonsilencing control shRNA (RHS4346). Lentiviral particles encoding shRNA were generated by the transient transfection of HEK293T cells with lentiviral GIPZ vectors and a mixture of packaging plasmids with X-tremeGENE 9 transfection reagent, used according to the manufacturer's instructions. Briefly, HEK293T cells at 80 to 90% confluence in a six-well plate were transfected with 1.5 mg of the lentiviral vector GIPZ, 1 mg of the packaging plasmid (psPAX2, Addgene), and 0.5 mg of the envelope plasmid (pMD2G, Addgene). The medium was changed the following day, and the virus-containing supernatant was collected 48 hours after transfection, passed through a filter with 0.45-mm pores, and used directly for cell transduction or stored at −80°C.
For the transduction of THP-1 cells, the cells were incubated with supernatants containing the lentiviral particles. The medium was replaced with fresh medium the following day, and puromycin was added 3 days after transduction, to a final concentration of 2 mg/ml. Protein production was analyzed by immunoblotting after 4 days of selection. All the experiments were performed between days 7 and 14 after transduction.
For shRNA-mediated knockdown experiments in primary monocyte-derived dendritic cells (MDDCs), a high transduction efficiency (>60% GFP + cells) was achieved by cotransduction with shRNA-encoding lentiviral particles and virion-like particles (VLPs) carrying the SIV viral protein Vpx (VLP-Vpx). Vpx suppresses the SAMHD1-mediated restriction of lentiviral reverse transcription in myeloid cells. VLP-Vpx were produced by transfecting HEK293T cells with 1.5 mg of the packaging vector SIV3+ (derived from SIVmac251) and 0.5 mg of the envelope plasmid pMD2G with XtremeGENE9. Monocytes were isolated from PBMCs from healthy donors by negative selection with the Pan Monocyte Isolation Kit (Miltenyi Biotec). Freshly purified monocytes were transduced with shRNA-encoding lentiviral particles and VLP-Vpx in the presence of protamine (8 mg/ml). Transduced cells were allowed to differentiate into MDDCs in the presence of recombinant human GM-CSF (10 ng/ml) and IL-4 (25 ng/ml) for 5 days.

Lentiviral transduction
HEK293T cells were dispensed into a six-well plate at a density of 8 × 10 5 cells per well. The next day, cells were transfected with pCMV-VSV-G (0.2 mg), pHXB2-env (0.2 mg; NIH-AIDS Reagent Program; 1069), psPAX2 (1 mg; Addgene plasmid no. 12260), and either pTRIP-SFFV-CD271-P2A empty vector or encoding the protein of interest (1.6 mg) in Opti-MEM (Gibco; 300 ml) containing X-tremeGENE 9 (Sigma Aldrich; 10 ml), according to the manufacturer's instructions. After 6 hours, the medium was replaced with 3 ml of fresh culture medium, and the cells were incubated for a further 24 hours for lentiviral particle production. The viral supernatant was collected and passed through a syringe filter with 0.2-mm pores (Pall) to remove debris. Protamine sulfate (Sigma; 10 mg/ml) was added to the supernatant, which was then used immediately or stored at −80°C until use.
For the transduction of THP-1 cells with OAS1, OAS2, or RNASEL, the corresponding gene KO THP-1 cells were dispensed into a 12-well plate at a density of 1 × 10 6 cells per well, in 500 ml of culture medium per well. Viral supernatant was then added (500 ml per well) the next day. For the transduction of SV40fibroblasts with ACE2, healthy control or patient-specific SV40-fibroblasts were used to seed six-well plates at a density of 5 × 10 5 cells per well. Viral supernatant was added (500 ml per well) the next day. The cells were then incubated for a further 48 hours at 37°C. Transduction efficiency was evaluated by surface staining for CD271 (Miltenyi Biotec) for the pTRIP vector, or by flow cytometry to evaluate BFP expression levels for the pSCRPSY vector. MACS column separation was performed with selection beads for CD271-positive cells (Miltenyi Biotec) if the proportion of CD271-positive cells was <80%. Cells transduced with the pSCRPSY vector were selected with puromycin or by flow cytometry. Protein production was subsequently validated by immunoblotting.

SARS-CoV-2 infection
The SARS-CoV-2 NYC isolate was obtained from the saliva of a deidentified patient on 28 July 2020. The sequence of the virus is publicly available (GenBank OM345241). The virus isolate was initially amplified in Caco-2 cells (passage 1, or P#1 stock). For the generation of P#2 and P#3 working stocks, Caco-2 cells were infected with the P#1 and P#2 viruses, respectively, at a multiplicity of infection (MOI) of 0.05 plaque-forming units (PFU)/cell and incubated for 6 and 7 days, respectively, at 37°C. The virus-containing supernatant was then harvested, clarified by centrifugation (3000g for 10 min), and filtered through a disposable vacuum filter system with 0.22-mm pores. The P#3 stock used in this study had a titer of 3.4 × 10 6 PFU/ml determined on Vero E6 cells with a 1% methylcellulose overlay, as previously described (72).
A549 + ACE2/TMPSS2 cells, human SV40fibroblasts + ACE2, or THP-1 cells were used to seed 96-well plates at a density of 1.5 × 10 4 cells per well, 4 × 10 3 cells per well, and 1 × 10 5 cells per well, respectively, in the presence or absence of IFN-a2b at a concentration of 1000 IU/ml. The cells were infected with SARS-CoV-2 24 hours later by directly adding 10 ml of virus stock at various dilutions to the wells (final volume: 110 ml). Cells were infected for 24, 48, or 72 hours. The cells were fixed with neutral buffered formalin at a final concentration of 10% and stained for SARS-CoV-2 with an anti-N antibody (catalog no. GTX135357; GeneTex). An Alexa Fluor 488-or Alexa Fluor 647-conjugated secondary antibody (Invitrogen) was used. Plates were imaged with an ImageXpress micro XL and analyzed with MetaXpress (Molecular Devices).

Cell stimulation
THP-1 cells were used to coat a 96-well plate at a density of 1 × 10 5 cells per 100 ml of culture medium. For stimulations of PBMCs and MDDCs, we used 1 × 10 5 cells and 5 × 10 5 cells per 100 ml of culture medium, respectively. The cells were stimulated with the indicated stimulus at the specified concentrations, with or without lipofectamine 2000 (Invitrogen), according to the manufacturer's instructions. Poly(I:C), 5′ppp-dsRNA, 5′ppp-dsRNA control, ISD, ISD control, R848, CPG-ODN2006, and LPS were purchased from Invivogen. For exogenous 2′5′-linked oligoadenylate (2-5A) or dephosphorylated 2-5A, we used 20 mM of 2-5A for transfection in the presence of lipofectamine simultaneously with the other stimuli [poly(I:C), R848, or LPS]. Dephosphorylated 2-5A (A2′p5′A2′p5′A) was prepared by treating 2-5A with shrimp alkaline phosphatase (SAP) (Thermo Fisher Science) to remove the 5′-triphosphoryl group from 2-5A, rendering it unable to activate RNase L (69,70). The dephosphorylation reaction mixture contained 5 mM 2-5A incubated with five units of SAP at 37°C for 1 hour, according to the manufacturer's protocol. Samples were denatured by incubation at 95°C for 5 min. Supernatants containing dephosphorylated 2′,5′-A3 were removed after centrifugation at 18,000g for 15 min at 4°C. Dephosphorylated 2-5A was then validated by HPLC and FRET assays for RNase L activity. After cell stimulation, the cells or supernatants were harvested, and their cytokine mRNA and protein levels were assessed by RT-qPCR and with a multiplex bead assay (BioLegend), respectively.

Detection of secreted cytokines in a multiplex bead assay
The harvested supernatants of stimulated THP-1 cells, PBMCs, and other types of cells were prepared and used for the LEGENDplex multiplex bead assay (BioLegend), according to the manufacturer's instructions. Samples were analyzed by flow cytometry on an Attune NxT flow cytometer, according to the manufacturer's instructions. Data were analyzed with LEGENDplex Cloud-based Data Analysis Software.
Luciferase assay THP-1 cells expressing an ISRE-lucia luciferase reporter gene were purchased from Invivogen (THP1-Dual). Cells were stimulated according to the conditions specified above. The supernatant was collected and used for the luciferase assay in accordance with the manufacturer's instructions.

Coculture of THP-1 and SARS-CoV-2-infected cells
Vero cells were plated in a six-well plate and infected at a MOI of 0.05 (as determined by plaque assay on Vero E6 cells) for a total of 48 hours. The supernatant of the infected cells was carefully removed, and the infected cells were then transferred to fresh THP-1 culture medium. A fixed volume of the resulting cell suspension was then dispensed onto WT or RNase L KO THP-1 cells plated in a 96-well plate at a density of 1 × 10 5 cells in 100 ml. THP-1 cells stimulated with SARS-CoV-2 only were stimulated in parallel for 24 hours. THP-1 cells were stimulated for a total of 24 hours before collection of the supernatant for cytokine determinations and cells for total RNA extraction.

Transfection of THP-1 cells with RNA from SARS-CoV-2-infected cells
Total RNA was extracted from mock-infected Vero cells or Vero cells infected with SARS-CoV-2 at a MOI of 0.05 for a total of 72 hours. THP-1 cells were transfected with 2 mg/ml of total RNA extract for 8 hours. THP-1 cells were then collected for total RNA extraction.

Deep immunophenotyping by mass cytometry (CyTOF)
CyTOF was performed on whole blood with the Maxpar Direct Immune Profiling Assay (Fluidigm), according to the manufacturer's instructions, as previously described (7). Cells were frozen at −80°C after overnight staining to eliminate dead cells, and acquisition was performed on a Helios machine (Fluidigm). The antibodies used for staining are listed in table S3. All the samples were processed within 24 hours of sampling. Data analysis was performed with OMIQ software.
Bulk RNA sequencing (RNA-seq) Total RNA was extracted from THP-1 cells or sorted blood cell populations. Cells were left untreated or were stimulated with poly(I:C) in the presence of lipofectamine or infected with SARS-CoV-2. RNA was extracted with the Quick-RNA MicroPrep kit (Zymo Research) or the RNeasy Micro Kit (Qiagen) and treated with DNase I (Zymo Research and Qiagen) to remove residual genomic DNA. RNA-seq libraries were prepared with the Illumina RiboZero TruSeq Stranded Total RNA Library Prep Kit (Illumina) and sequenced on the Illumina NovaSeq platform in the 100 nucleotide, pairedend configuration. Each library was sequenced twice.
The RNA-seq FASTQ files were first inspected with fastqc to ensure that the raw data were of high quality. The sequencing reads of each FASTQ file were then aligned with the GENCODE human reference genome GRCh37.p13 with STAR aligner v2.6 and the alignment quality of each BAM file was evaluated with RSeQC. Reads were quantified with featureCounts v1.6.0 to generate gene-level feature counts from the read alignment, based on GENCODE GRCh37.p13 gene annotation. The gene-level feature counts were then normalized and log 2transformed with DESeq2, to obtain gene expression values for all genes and all samples. Differential gene expression analyses were conducted by contrasting the intracellular poly(I:C)-stimulated samples or the SARS-CoV-2-infected samples with the nonstimulated samples. For each gene expression analysis, we performed trimmed mean of M values (TMM) normalization and gene-wise generalized linear model regression by edgeR, and the genes displaying significant differential expression were selected according to the following criteria: FDR ≤ 0.05 and |log2(FoldChange)| ≥ 1. Differential gene expression was plotted as a heatmap with ComplexHeatmap, and genes and samples were clustered according to complete linkage and the Euclidean distances of gene expression values. GSEA was conducted with the fgsea package, by projecting the ranking of fold-change in expression onto the Hallmark gene sets (71).

Single-cell RNA sequencing of PBMCs
We performed scRNA-seq on SARS-CoV-2and mock-stimulated PBMCs sampled from four individuals with inborn errors of the OAS-RNaseL pathway (P1 with OAS1 deficiency, P2 and P3 with OAS2 deficiency, P5 with RNase L deficiency), three individuals with inborn errors of type I IFN immunity, and eight healthy donors-one pediatric control and one adult control with a history of past asymptomatic SARS-CoV-2 infection, and two pediatric controls and four adult controls with no history of prior SARS-CoV-2 infection. The cryopreserved PBMCs were thawed, stimulated, and processed for scRNA-seq. Across all samples, we captured 46,157 highquality single-cell transcriptomes that were classified into five major immune cell lineages: myeloid, B, CD4 + T, CD8 + T, and NK cells. The data were then analyzed as described in detail in the supplementary materials.
We also performed scRNA-seq on cryopreserved PBMCs from P5 (RNase L-deficient, aged 4 years) sampled during the acute (9 days after MIS-C onset) and convalescent (~1 month after onset) phases, together with cells from one healthy adult and two pediatric controls.
We compared the data obtained with a previously published dataset for patients with pediatric acute SARS-CoV-2 infection or MIS-C (33). Clustering analysis showed lower levels of monocytes and type 1 and type 2 conventional dendritic cells (cDCs) in these patients and an expansion of the activated T cell population strongly expressing MKI67 (Fig. 5F and fig. S8,  A and B). Other subsets were largely unaffected. Pseudobulk differential expression analysis was performed at the single-cell level for monocytes, mDCs, B lymphocytes, plasmacytoid dendritic cells (pDCs), and activated T cells. Bulk RNAseq was performed on sorted nonclassical monocytes and pDCs to further confirm the scRNA-seq findings. We also quantitatively inferred cell-cell communications with Cell-Chat (74) to identify the signal-outgoing and the signal-receiving cell subsets. The data generated during this study were analyzed in an integrative manner with historical controls from the laboratory (one pediatric and seven adult controls), publicly available control PBMC datasets downloaded from the 10X Genomics web portal (https://support.10xgenomics.com/ single-cell-gene-expression/datasets), and a previously published dataset for patients with acute SARS-CoV-2 infection and MIS-C (GEO accession: GSE167029), as described in detail in the supplementary materials. In addition, two other previously published sets of scRNAseq data for pediatric healthy controls and children with acute SARS-CoV-2 infection or MIS-C (GSE166489) (91) were used for an independent cohort analysis.

Statistical analysis
For experiments performed in vitro, quantitative data were obtained for cells carrying the different mutations and control cells, or cells treated with different stimuli, from at least three biological replicates. For each biological replicate, up to six technical replicates were performed and averaged for downstream analysis. Cytokine determinations were log-transformed after subtracting the limit of detection for the experiment concerned. Mean quantitative values were compared between cells carrying the various mutations and control cells or cells treated with different stimuli in unequalvariance t tests. Where relevant, statistical test results are indicated in the corresponding figures (ns, not significant; *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001).