Uncontrolled type I IFN activity has been linked to several human pathologies, but evidence implicating this cytokine response directly in disease has been limited. Here, Duncan et al. identified a homozygous missense mutation in STAT2 in siblings with severe early-onset autoinflammatory disease and elevated IFN activity. STAT2 is a transcription factor that functions downstream of IFN, and this STAT2R148W variant was associated with elevated responses to IFNα/β and prolonged JAK-STAT signaling. Unlike wild-type STAT2, the STAT2R148W variant could not interact with ubiquitin-specific protease 18, which prevented STAT2-dependent negative regulation of IFNα/β signaling. These findings provide insight into the role of STAT2 in regulating IFNα/β signaling in humans.
Excessive type I interferon (IFNα/β) activity is implicated in a spectrum of human disease, yet its direct role remains to be conclusively proven. We investigated two siblings with severe early-onset autoinflammatory disease and an elevated IFN signature. Whole-exome sequencing revealed a shared homozygous missense Arg148Trp variant in STAT2, a transcription factor that functions exclusively downstream of innate IFNs. Cells bearing STAT2R148W in homozygosity (but not heterozygosity) were hypersensitive to IFNα/β, which manifest as prolonged Janus kinase–signal transducers and activators of transcription (STAT) signaling and transcriptional activation. We show that this gain of IFN activity results from the failure of mutant STAT2R148W to interact with ubiquitin-specific protease 18, a key STAT2-dependent negative regulator of IFNα/β signaling. These observations reveal an essential in vivo function of STAT2 in the regulation of human IFNα/β signaling, providing concrete evidence of the serious pathological consequences of unrestrained IFNα/β activity and supporting efforts to target this pathway therapeutically in IFN-associated disease.
Type I interferons (including IFNα/β) are antiviral cytokines with pleiotropic functions in the regulation of cellular proliferation, death, and activation. Reflecting their medical importance, type I IFNs have been shown to be essential to antiviral immunity in humans (1), whereas their potent immunomodulatory effects have been exploited to treat both cancer and Multiple Sclerosis (2, 3).
IFNα/β also demonstrates considerable potential for toxicity, which became apparent in initial studies in rodents (4) and subsequent clinical experience in patients (5, 6). Thus, the production of and response to type I IFNs must be tightly controlled (7). Transcriptional biomarker studies increasingly implicate dysregulated IFNα/β activity in a diverse spectrum of pathologies ranging from autoimmune to neurological, infectious and vascular diseases (8–11).
The immunopathogenic potential of IFNα/β is exemplified by a group of monogenic inborn errors of immunity termed “type 1 interferonopathies,” wherein enhanced IFNα/β production is hypothesized to be directly causal (12). Neurological disease is typical of these disorders, which manifest as defects of neurodevelopment in association with intracranial calcification and white matter changes on neuroimaging, suggesting that the brain is particularly vulnerable to the effects of excessive type I IFN activity (9). A spectrum of clinical severity is recognized, from prenatal-onset neuroinflammatory disease that mimics in utero viral infection—Aicardi-Goutières syndrome (13)—to a clinically silent elevation of IFN activity (14).
However, the central tenet of the type I interferonopathy hypothesis, namely, the critical pathogenic role of type I IFNs (12), has yet to be formally established (15). Evidence for an IFN-independent component to disease includes (i) recognition that other proinflammatory cytokines are also induced by nucleic acid sensing, which might contribute to pathogenesis (16); (ii) imperfect correlations between IFN biomarker status and disease penetrance (14); (iii) the absence of neuropathology in mouse models of Aicardi-Goutières syndrome despite signatures of increased IFN activity (17); and (iv) the observation that crossing to a type I IFN receptor deficient background does not rescue the phenotype in certain genotypes (e.g., STING and ADAR1) (18, 19), although it does in others (e.g., TREX1 or USP18) (20, 21). Here, we provide concrete evidence of the pathogenicity of type I IFNs in humans, shedding new light on the critical importance of signal transducer and activator of transcription 2 (STAT2) in the negative regulation of this pathway.
Severe neurological and systemic inflammatory disease associated with increased IFN signature
We evaluated two male siblings, born in the United Kingdom to second cousin Pakistani parents. Briefly, patient II:3, born at 34 weeks + 6 days with transient neonatal thrombocytopenia, was investigated for neurodevelopmental delay at 6 months (which was attributed to compensated hypothyroidism). Aged 8 months, he presented with the first of three episodes of marked neuroinflammatory disease, associated with progressive intracranial calcification, white matter disease, and, by 18 months, intracranial hemorrhage (Fig. 1A). These episodes were associated with systemic inflammation and multiorgan dysfunction, including recurrent fever, hepatosplenomegaly, cytopenia with marked thrombocytopenia, raised ferritin, and elevated liver enzymes. Latterly, acute kidney injury with hypertension and nephrotic range proteinuria developed (see Table 1, Supplementary case summary, and table S1).Fig. 1 Neurological and systemic disease associated with excessive IFN activity.
(A) Neuroimaging demonstrating calcifications [brainstem/hypothalamus (proband II:3, top), cerebral white matter/basal ganglia/midbrain/optic tract (sibling II:4, top and middle)], hemorrhages [occipital/subdural/subarachnoid (proband II:3, middle)], and cerebral white matter and cerebellar signal abnormality with parenchymal volume loss (both, bottom), accompanied by focal cystic change and cerebellar atrophy (sibling II:4). (B) Whole blood RNA-seq ISG profiles: controls (n = 5); proband II:3 (n = 4); and patients with mutations in: TREX1 (n = 6), RNASEH2A (n = 3), RNASEH2B (n = 7), RNASEH2C (n = 5), SAMHD1 (n = 5), ADAR1 (n = 4), IFIH1 (n = 2), ACP5 (n = 3), TMEM173 (n = 3), and DNASE2 (n = 3). (C) IFN scores (RT-PCR) of patients, parents, and n = 29 healthy controls. ****P < 0.001, ANOVA with Dunnett’s posttest. (D) Renal histopathology in proband (×400 magnification) showing TMA with extensive double contouring of capillary walls (silver stain, top left); endothelial swelling, mesangiolysis, and red cell fragmentation (top right); arteriolar fibrinoid necrosis (bottom left); and myxoid intimal thickening of an interlobular artery (bottom right, all hematoxylin and eosin). (E) Transcriptional response to JAK inhibitor (JAKi) ruxolitinib in both patients (RT-PCR).
View this table:
This clinical phenotype was reminiscent of a particularly severe form of type I interferonopathy. In keeping with this observation, IFN-stimulated gene (ISG) transcripts in whole blood, measured by RNA sequencing (RNA-seq) and reverse transcription polymerase chain reaction (RT-PCR), were substantially elevated over multiple time points at similar magnitudes to recognized type I interferonopathies (Fig. 1, B and C), notably without evidence of concomitant induction of IFN-independent inflammatory pathways (fig. S1). disease in the proband, which not only met the diagnostic criteria for hemophagocytosis but also included features of a thrombotic microangiopathy (TMA) (Fig. 1D), was partially responsive to dexamethasone and stabilized with the addition of the Janus kinase (JAK) inhibitor ruxolitinib (Fig. 1E and fig. S2). Sadly, however, this child succumbed to overwhelming Gram-negative bacterial sepsis during hematopoietic stem cell transplantation.
Patient II:4, his infant brother, presented with abnormal neurodevelopment and neuroimaging in the neonatal period, characterized by apneic episodes from 3 weeks of age in conjunction with parenchymal calcifications and hemorrhage, abnormal cerebral white matter, and brainstem and cerebellar atrophy (Fig. 1A). Blood tests revealed an elevated ISG score (Fig. 1, B and C), anemia, elevation of D-dimers, and red cell fragmentation on blood film, together with proteinuria and borderline elevations of ferritin and lactate dehydrogenase; renal function was normal, and blood pressure was on the upper limit of the normal range for gestational age. Introduction of ruxolitinib led to prompt suppression of ISG expression in whole blood (Fig. 1E) and an initial reduction in apneic episodes, but neurological damage was irretrievable, and he succumbed to disease at 3 months of age. Mother’s pregnancy with patient II:4 had been complicated by influenza B at 23 weeks’ gestation.
A rare homozygous missense variant of STAT2 (Arg148Trp) segregates with disease
Whole-exome sequencing analysis of genomic DNA from the kindred, confirmed by Sanger sequencing (Fig. 2, A and B), identified an extremely rare variant in STAT2 (c.442C>T), which substituted tryptophan for arginine at position 148 in the coiled-coil domain (CCD) of STAT2 (p.Arg148Trp, Fig. 2C). The Arg148Trp variant was present in the homozygous state in both affected children and was heterozygous in each parent and one healthy sibling, consistent with segregation of an autosomal recessive trait (table S2). This variant was found in the heterozygous state at extremely low frequency in publicly available databases of genomic variation [frequency < 0.00001 in Genome Aggregation Database (22)], and no homozygotes were reported. A basic amino acid, particularly arginine, at position 148 is highly conserved (fig. S3). In silico tools predicted that this missense substitution was probably deleterious to protein function (table S2). STAT2 protein expression in patient cells was unaffected by the Arg148Trp variant, in contrast to the situation for pathogenic loss-of-expression STAT2 variants, which resulted in a distinct phenotype of heightened viral susceptibility (Fig. 2D) (23, 24). Filtering of exome data identified an additional recessive variant in CFH (c.2336A>G and p.Tyr779Cys; fig. S4) present in the homozygous state in II:3 but absent from II:4. We considered the possibility that this contributed to TMA in the proband, but functional studies of this variant showed negligible impact on factor H function (fig. S5).Fig. 2 A homozygous missense variant in STAT2 consistent with autosomal recessive inheritance.
(A) Pedigree, (B) capillary sequencing verification, (C) protein map, and (D) immunoblot (fibroblasts) showing normal expression of STAT2 protein. DBD, DNA binding domain; LD, linker domain; SH2, Src homology 2 domain; TAD, trans-activation domain.
STAT2R148W enhances cellular sensitivity to type I IFNs
The transcription factor STAT2 is essential for transcriptional activation downstream of the receptors for the innate IFN-α/β (IFNAR) and IFN-λ and their associated JAK adaptor proteins. In the current paradigm (25), STAT2 is activated by tyrosine phosphorylation, associated with IFN regulatory factor 9 (IRF9) and phosphorylated STAT1 (pSTAT1) to form the IFN-stimulated gene factor 3 (ISGF3) to effect gene transcription by binding to IFN-stimulated response elements in the promoters of ISGs. Although loss-of-function variants in STAT2 increase susceptibility to viral disease (23, 24), evidence here suggested pathological activation. Germline gain-of-function variants have been reported in STAT1 (26, 27) and STAT3 (28, 29) but not hitherto STAT2. Consistent with the apparent gain of IFN activity associated with mutant STAT2R148W, we observed in patient fibroblasts (Fig. 3, A and B) and peripheral blood mononuclear cells (PBMCs; fig. S6) the enhanced expression of ISG protein products across a range of IFNα concentrations. However, basal and induced production of IFNB mRNA by fibroblasts was indistinguishable from controls (Fig. 3C); nor was IFNα protein substantially elevated in patient samples of cerebrospinal fluid (II:3) or plasma (II:4) as measured by a highly sensitive digital enzyme-linked immunosorbent assay (ELISA) assay (30), albeit samples were acquired during treatment (table S3). Thus, the response to type I IFNs, but not their synthesis, was exaggerated. This heightened IFN sensitivity was accompanied by enhancement of key effector functions, as revealed by assays of IFNα-mediated viral protection (Fig. 3D) and cytotoxicity (Fig. 3E). Collectively, these data indicated that STAT2R148W was not constitutively active but rather resulted in an exaggerated response upon IFNα exposure. To confirm that the Arg148Trp variant was responsible for this cellular phenotype, we transduced STAT2-null U6A cells (31) and STAT2-deficient primary fibroblasts (23) with lentiviruses encoding either wild type (WT) or STAT2R148W, recapitulating the heightened sensitivity of cells expressing the latter to IFNα (Fig. 3, F and G, and fig. S7).Fig. 3 Heightened sensitivity to IFNα in cells bearing STAT2R148W.
Unless stated, all data are from patient II:3 and control fibroblasts. (A) ISG expression (immunoblot, IFNα for 24 hours) and (B) densitometry analysis (n = 3, t test). MX1, MX dynamin like GTPase 1; IFIT1, IFN-induced protein with tetratricopeptide repeats 1; RSAD2, radical S-adenosyl methionine domain containing 2. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (C) IFNB mRNA (RT-PCR) ± external polyinosinic:polycytidylic acid (poly I:C) treatment (25 μg/ml for 4 hours; n = 3, t test). US, unstimulated. (D) Antiviral protection assay (mCherry-PIV5). Twofold dilutions from IFNα (16 IU/ml), IFNγ (160 IU/ml) n = 7 replicates, representative of n = 2 experiments (two-way ANOVA with Sidak’s posttest). (E) Cytopathicity assay (IFNα for 72 hours; n = 3, t test). (F) As in (A), ISG expression in STAT2−/− U6A cells reconstituted with STAT2WT or STAT2R148W (immunoblot, IFNα for 24 hours). (G) As in (B), n = 3 to 4, t test. Data are presented as means ± SEM of repeat experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. n.s., nonsignificant.
STAT2R148W prolongs IFNAR signaling without affecting STAT2 dephosphorylation
To explore the underlying mechanism for heightened type I IFN sensitivity, we first probed STAT2 activation in IFNα-stimulated fibroblasts. In control lysates, levels of pSTAT2 had almost returned to baseline between 6 and 24 hours of treatment despite the continued presence of IFNα (Fig. 4, A and B). In contrast, pSTAT2 persisted for up to 48 hours in patient cells. This abnormally prolonged pSTAT2 response to IFNα was also observed in PBMCs of both patients (fig. S8). Consistent with immunoblot data, immunofluorescence analysis showed persistent (≥ 6 hours) nuclear localization of STAT2 in patient fibroblasts after IFNα treatment, at times when STAT2 staining was predominantly cytoplasmic in control cells (Fig. 4, C and D, and fig. S9). This was accompanied by continued expression of ISG transcripts for 36 hours after the washout of IFNα in patient cells as measured by RNA-seq and RT-PCR (Fig. 4, E and F). Thus, the type I IFN hypersensitivity of patient cells was linked to prolonged IFNAR signaling.Fig. 4 Prolonged STAT2 activation but no change to dephosphorylation rate.
All data are from patient II:3 and control fibroblasts. (A) pSTAT2 time course [immunoblot, IFNα (1000 IU/ml)] and (B) densitometry analysis (n = 5 experiments, two-way ANOVA with Sidak’s posttest). (C) Immunofluorescence analysis [IFNα (1000 IU/ml); scale bar, 100 μm; representative of n = 3 experiments] with (D) image analysis of STAT2 nuclear translocation (n = 100 cells per condition, ANOVA with Sidak’s posttest). A.U., arbitrary units. (E) RNA-seq analysis of IFN-regulated genes (n = 3 controls) with (F) validation by RT-PCR (n = 3, two-way ANOVA with Sidak’s posttest). CPM, read counts per million. (G) pSTAT2 decay (immunoblot). IFNα (1000 IU/ml; 30 min) followed by extensive washing and treatment with 500 nM staurosporine (STAU). Times relative to STAU treatment. (H) No significant differences by densitometry analysis (n = 3, t test). Data are presented as means ± SEM of repeat experiments. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.
The IFNAR signaling pathway is subject to multiple layers of negative regulation that target STAT phosphorylation directly—through the action of tyrosine phosphatases—or indirectly by disrupting upstream signal transduction (7). Prolonged tyrosine phosphorylation is reported with gain-of-function mutations in STAT1, in association with impaired sensitivity to phosphatase activity (27). By contrast, we observed no impairment of dephosphorylation of STAT2R148W in pulse-chase assays with the kinase inhibitor staurosporine (Fig. 4, G and H), implying instead a failure of negative feedback upon the proximal signaling events that generate pSTAT2.
Prolonged IFNAR signaling in STAT2R148W homozygous but not heterozygous cells
To localize this defect, we analyzed by phosflow and immunoblot the successive activation steps downstream of IFNAR ligand binding in Epstein-Barr virus (EBV)–transformed B cells from the proband (II:3) and a heterozygous parent (I:2). As was the case for STAT2 phosphorylation, we also observed prolonged phosphorylation of both JAK1 and STAT1 after IFNα treatment (Fig. 5, A to D). This points to a defect in regulation of the most proximal IFNAR signaling events, upstream of STAT2 (7). We observed no evidence of this phenotype in cells bearing STAT2R148W in the heterozygous state, consistent with autosomal recessive inheritance and the lack of clinical disease or up-regulation of IFN activity in heterozygous carriers. This genetic architecture provides a notable contrast to gain-of-function mutations affecting other STAT proteins, all of which are manifest in the heterozygous state (26–29).Fig. 5 Prolonged proximal IFNAR signaling in STAT2R148W homozygosity but not heterozygosity.
Time course of IFNα stimulation (1000 IU/ml) in EBV B cells from patient II:3 [homozygous (hom)], parent I:2 [heterozygous (het)], and n = 3 controls. (A) Immunoblot and (B) densitometry analyses. (C) Representative histograms (flow cytometry) and (D) mean fluorescence intensity (MFI). Data are means ± SEM of three repeat experiments (*P < 0.05, **P < 0.01, t test).
STAT2R148W specifically compromises negative regulation of IFNAR signaling
Known negative regulators of IFNAR signaling are suppressor of cytokine signaling (SOCS) 1 and SOCS3 (32) and the ubiquitin-specific protease 18 (USP18) (33). SOCS1 and SOCS3 participate in regulation of additional JAK-STAT signaling pathways, such as those activated by IFNγ and interleukin 6 (IL-6) (34, 35), whereas USP18 acts specifically upon IFNAR signaling (33). To better localize the molecular defect in patient cells, we examined the signaling responses to IFNγ (STAT1 phosphorylation) and IL-6 (STAT3 phosphorylation), based on the prediction that defects of SOCS1 or SOCS3 regulation would manifest under these conditions. These experiments revealed that regulation of STAT1 and STAT3 phosphorylation was normal in patient fibroblasts (fig. S10). Together with the absence of evidence of up-regulation of the IFNγ and IL-6 pathways in the analysis of whole blood RNA-seq data (fig. S1), these observations effectively ruled out the involvement of SOCS1 and SOCS3 in the clinical phenotype, leading us to suspect a defect of USP18 regulation.
To investigate this possibility, we primed patient and control cells with IFNα for 12 hours, washed them extensively, and rested and restimulated them with IFNα or IFNγ after 48 hours. In these experiments, IFNα-induced pSTAT2 and pSTAT1 were strongly inhibited by priming in control cells, consistent with desensitization, a well-established phenomenon of type I IFN biology (Fig. 6, A and B) (36). In marked contrast, the response to IFNα restimulation in patient cells was minimally suppressed, indicating a failure of desensitization. Desensitization has been shown to be exclusively mediated by USP18, an IFN-induced isopeptidase (37), through its displacement of JAK1 from the receptor subunit IFNAR2 (38, 39)—a function that is independent of its isopeptidase activity toward the ubiquitin-like protein ISG15 (33). STAT2 plays a critical role as an adaptor protein by supporting binding of USP18 to IFNAR2 (Fig. 6C) (40). Both the clinical and cellular effects of STAT2R148W resemble homozygous USP18 deficiency, which was recently described as the molecular cause of a severe pseudo-TORCH syndrome associated with elevated type I IFN expression (table S4) (41). Although this STAT2:USP18 interaction has been shown to be essential for negative regulation of type I IFN signaling in vitro (40), its significance in vivo has not previously been examined. Furthermore, the precise residue(s) of STAT2 that bind USP18 were unresolved, although this interaction had been localized to a region including the CCD and/or DNA binding domain(s) of STAT2 (40).Fig. 6 STAT2R148W fails to support desensitization through its impaired interaction with USP18.
(A) Desensitization assay (immunoblot, fibroblasts) with (B) pSTAT densitometry analysis (pSTAT/tubulin, ratio to unprimed; n = 4, ANOVA with Sidak’s posttest). (C) Schematic of USP18 mechanism of action and proposed model of STAT2R148W pathomechanism. (D) Modeling of exposed WT (R148)/mutant (W148) residue, demonstrating charge-change (blue, positive; red, negative) and possible steric restriction. (E) Coimmunoprecipitation of USP18 by STAT2 in U6A cells expressing STAT2WT or STAT2R148W with (F) densitometry analysis (USP18/STAT2, ratio to WT; one-sample t test). Data are means ± SEM (**P < 0.01, ****P < 0.0001). IB, immunoblot.
STAT2R148W impairs interaction with USP18
Because USP18 was induced normally in patient cells (Fig. 6, A and B) and in vivo (Fig. 1B), our data implied that STAT2R148W impedes the proper interaction of STAT2 with USP18, compromising its regulatory function (Fig. 6C). Molecular modeling of STAT2R148W placed the substituted bulky aromatic tryptophan, and resulting charge change, at an exposed site within the CCD (Fig. 6D). Consistent with our suspicion that this might impair the STAT2:USP18 interaction through electrostatic or steric hindrance, coimmunoprecipitation experiments in U6A cells stably expressing WT or STAT2R148W demonstrated a statistical significance reduction of USP18 pull down STAT2R148W compared with WT (Fig. 6, E and F), providing a molecular mechanism for the USP18 insensitivity of patient cells.
Although disruption to the STAT2R148W:USP18 interaction was the most plausible explanation for the clinical and molecular phenotype, we also considered the contribution of alternative regulatory functions of STAT2. Beyond the role of tyrosine phosphorylated STAT2 in innate IFN signal transduction, the unphosphorylated form of STAT2 (uSTAT2) has additional, recently described functions in the regulation of other cytokine signaling pathways. For example, uSTAT2 negatively regulates the activity of IFNγ (and other inflammatory cytokines that signal via STAT1 homodimers) by binding to uSTAT1 via its CCD (42). This interaction appears to limit the pool of STAT1 available for incorporation into transcriptionally active (tyrosine phosphorylated) STAT1 homodimers. Conversely, uSTAT2, induced by type I IFN signaling, has been reported to promote the transcriptional induction of IL6 through an interaction with the nuclear factor κB subunit p65 (43). To investigate the potential relevance of these regulatory functions of STAT2, we first examined the induction of IL6 by RT-PCR analysis of RNA isolated from whole blood of patients, their heterozygous parents, and healthy controls. We found no evidence of increased expression of IL6 or its target gene SOCS3 (fig. S11, A and B), consistent with our previous pathway analysis of RNA-seq data (fig. S1) and implying that STAT2R148W does not influence IL-6 induction. Next, to explore any impact on STAT2’s negative regulatory activity toward STAT1, we examined the transcriptional responses to IFNγ in patient fibroblasts and in U6A cells expressing STAT2R148W. Although we were able to reproduce the previously reported findings of heightened transcription of the IFNγ-regulated gene CXCL10 in U6A cells lacking STAT2, alongside a nonsignificant trend for IRF1 (fig. S12, A and B) (42), STAT2R148W did not enhance transcript levels of either CXCL10 or IRF1 above WT, in agreement with the data showing the preserved ability of STAT2R148W to bind STAT1 in a coimmunoprecipitation assay (fig. S12, C and D). Together, these studies effectively exclude a contribution of the USP18-independent regulatory functions of STAT2 to the disease phenotype.
Insensitivity to USP18 regulation in STAT2R148W patient cells
To conclusively demonstrate the impairment of STAT2:USP18-mediated negative regulation in patient cells, we tested the impact of overexpression or knockdown of USP18. First, we probed IFNAR responses in fibroblasts stably expressing USP18. As predicted, USP18 was significantly impaired in its ability to suppress IFNα signaling in patient cells, relative to controls, both in terms of STAT phosphorylation (Fig. 7, A and B) and STAT2 nuclear translocation (Fig. 7, C and D), recapitulating our prior observations with IFN priming (Fig. 6A). The reciprocal experiment, in which USP18 expression was stably knocked down using short hairpin RNA (shRNA), revealed significantly prolonged STAT2 phosphorylation in control cells at 24 hours, recapitulating the phenotype of patient cells (Fig. 7, E and F). In contrast, there was no effect of USP18 knockdown in patient cells, demonstrating that they are USP18 insensitive. Incidentally, we noted that the early peak (1 hour) of STAT2 phosphorylation in USP18-knockdown control fibroblasts was marginally reduced (Fig. 7E). This subtle reduction was also apparent in STAT2R148W patient fibroblasts (Fig. 4B), although not in EBV B cells (Fig. 5). We speculate that the cell type–specific induction of other negative regulator(s) of IFNAR signaling at early times after IFN treatment, such as SOCS1, might be responsible for this observation. RT-PCR analysis confirmed the increased expression of SOCS1 mRNA in whole blood of patients (fig. S11C), whereas examination of RNA-seq data from IFNα-treated fibroblasts revealed an eightfold enhancement of SOCS1 expression at 6 hours in patient cells as compared with controls (Padj = 0.0001; Fig 4E). Together, these data provide preliminary support for the hypothesis that alternative negative regulator(s) of IFNAR signaling may be up-regulated in patient cells. Nevertheless, such attempts at compensation are clearly insufficient to restrain IFNAR responses in the context of STAT2R148W, reflecting the nonredundant role of STAT2/USP18 in this process (39). Collectively, these data support a model in which the homozygous presence of the Arg148Trp STAT2 variant compromises an essential adaptor function of STAT2 toward USP18, rendering cells USP18 insensitive and culminating in unrestrained, immunopathogenic IFNAR signaling.Fig. 7 USP18 insensitivity in cells bearing STAT2R148W.
All data are from patient II:3 and control fibroblasts. (A) STAT phosphorylation in USP18 and vector expressing fibroblasts (immunoblot) with (B) pSTAT densitometry analysis (pSTAT/tubulin, ratio to unprimed; n = 3, ANOVA with Sidak’s posttest). (C) Immunofluorescence analysis of STAT2 nuclear translocation [IFNα (1000 IU/ml 30 min); representative of n = 3 experiments] with (D) image analysis (n = 100 cells per condition, ANOVA with Sidak’s posttest). (E) Time course of STAT phosphorylation upon IFNα stimulation (1000 IU/ml for 0, 1, 6, and 24 hours) of cells transduced with USP18 shRNA or nontargeting (NT) shRNA with (F) densitometry analysis of pSTAT2 (n = 3, t test). Data are means ± SEM (**P < 0.01, ***P < 0.001, ****P < 0.0001).
We report a type I interferonopathy, caused by a homozygous missense mutation in STAT2, and provide detailed studies to delineate the underlying molecular mechanism. Our data indicate the failure of mutant STAT2R148W to support proper negative regulation of IFNAR signaling by USP18—revealing an essential regulatory function of human STAT2. This defect in STAT2 regulation results in (i) an inability to properly restrain the response to type I IFNs and (ii) the genesis of a life-threating early-onset inflammatory disease. This situation presents a marked contrast with monogenic STAT2 deficiency, which results in heightened susceptibility to viral infection due to the loss of the transcription factor complex ISGF3 (23, 24). Thus, just as allelic variants of STAT1 and STAT3 are recognized that either impair or enhance activity of the cytokine signaling pathways in which they participate (44), we can now add to this list STAT2. Our findings also highlight an apparently unique property of human STAT2: That it participates directly in both the positive and negative regulation of its own cellular signaling pathway. Whether this is true of STAT2 in other species remains to be determined. Our findings also localize the interaction with USP18 to the CCD of STAT2, indicating a specific residue critical for this interaction. This structural insight may be relevant to efforts to therapeutically interfere with the STAT2:USP18 interaction to promote the antiviral action of IFNs.
This monogenic disease of STAT2 regulation provides incontrovertible evidence of the pathogenic effects of failure to properly restrain IFNAR signaling in humans. The conspicuous phenotypic overlap with existing defects of IFNα/β overproduction, particularly with regard to the neurological manifestations, provides compelling support for the type I interferonopathy hypothesis, strengthening the clinical rationale for therapeutic blockade of IFNAR signaling (15). JAK1/2 inhibition with ruxolitinib was highly effective in controlling disease in the proband; however, the damage that already accrued at birth in his younger brother was irreparable, emphasizing the importance of timely IFNAR blockade in prevention of neurological sequelae. A notable aspect of the clinical phenotype in patient II:3 was the occurrence of severe TMA. Our studies did not support a pathogenic contribution of the coinherited complement factor H variant in patient II:3. This evidence, together with clinical hematological and biochemical results suggestive of incipient vasculopathy in patient II:4—who did not carry the CFH variant—suggests that type I IFN may have directly contributed to the development of TMA. Although it is not classically associated with type I interferonopathies, TMA is an increasingly recognized complication of both genetic (41, 42) and iatrogenic states of IFN excess (43), consistent with the involvement of vasculopathy in the pathomechanism of IFN-mediated disease. The fact that STAT2R148W is silent in the heterozygous state at first sight offers a confusing contrast with “gain-of-function” mutations of its sister molecules STAT1 and STAT3, both of which produce autosomal dominant disease with high penetrance (26–29). However, the net gain of IFNAR signaling activity results from the isolated loss of STAT2’s regulatory function, which evidently behaves as a recessive trait. There are other examples of autosomal recessive loss-of-function disorders of negative regulators, including USP18 itself (41, 45); the unique aspect in the case of STAT2R148W is that the affected molecule is itself a key positive mediator within the regulated pathway.
In light of the intimate relationship between STAT2 and USP18 revealed by these and other recent data (40), it is reasonable to conclude that the clinical manifestations of human USP18 deficiency are dominated by the loss of its negative feedback toward IFNAR rather than the STAT2-independent functions of USP18 including its enzymatic activity (40, 46, 47). In mouse, white matter pathology associated with microglia-specific USP18 deficiency is prevented in the absence of IFNAR (21). There are now three human autosomal recessive disorders that directly compromise the proper negative regulation of IFNAR signaling and thus produce a net gain of signaling function: USP18 deficiency, which leads to embryonic or neonatal lethality with severe multisystem inflammation (41); STAT2R148W, which largely phenocopies USP18 deficiency; and ISG15 deficiency, in which there is a much milder phenotype of neurological disease without systemic inflammation (45). ISG15 stabilizes USP18, and human ISG15 deficiency leads to a partial loss of USP18 protein (41). Thus, a correlation is clearly evident between the extent of USP18 dysfunction and the clinical severity of these disorders, with STAT2R148W closer to USP18 deficiency and ISG15 on the milder end of the spectrum (table S4). Those molecular defects that result in a failure of negative regulation of IFNAR signaling (i.e., STAT2R148W and USP18−/−) lead to more serious and extensive systemic inflammatory disease than do defects of excessive IFNα/β production (41), suggesting that the STAT2:USP18 axis acts to limit an immunopathogenic response toward both physiological (48) and pathological (41) levels of IFNα/β. Thus, variability in the efficiency of this process of negative regulation might be predicted to influence the clinical expressivity of interferonopathies. Determining the cellular source(s) of “physiological” type I IFNs and the molecular pathways that regulate their production are important areas for future investigation.
Some limitations of our results should be acknowledged. Although strenuous efforts were made, we were only able to identify a single kindred, which probably reflects the rarity of this variant. As more cases are identified, our understanding of the clinical phenotypic spectrum will inevitably expand. Furthermore, for practical and cultural/ethical reasons, limited amounts of cellular material and tissues were available for analysis. As a result, we were unable to formally evaluate the relevance of STAT2 regulation toward type III IFN signaling; however, existing data suggest that USP18 plays a negligible role in this context (38). Together, our findings confirm an essential regulatory role of STAT2, supporting the hypothesis that type I IFNs play a causal role in a diverse spectrum of human disease, with immediate therapeutic implications.
MATERIALS AND METHODS
We investigated a kindred with a severe, early-onset, presumed genetic disease, seeking to determine the underlying pathomechanism by ex vivo and in vitro studies. Written informed consent for these studies was provided, and ethical/institutional approval was granted by the NRES Committee North East-Newcastle and North Tyneside 1 (ref: 16/NE/0002), South Central-Hampshire A (ref: 17/SC/0026), and Leeds (East) (ref: 07/Q1206/7).
Cells, cytokines, and inhibitors
Dermal fibroblasts from patient II:3 and healthy controls were obtained by standard methods and cultured in Dulbecco’s modified Eagle’s medium supplemented by 10% fetal calf serum and 1% penicillin/streptomycin (DMEM-10), as were human embryonic kidney 293 T cells and the STAT2-deficient human sarcoma cell line U6A (31). PBMCs and EBV-transformed B cells were cultured in RPMI medium supplemented by 10% fetal calf serum and 1% penicillin/streptomycin (RPMI-10). Unless otherwise stated, cytokines/inhibitors were used at the following concentrations: human recombinant IFN-α2b (1000 IU/ml; Intron A, Schering-Plough, USA); IFN-γ (1000 IU/ml; Immunikin, Boehringer Ingelheim, Germany); IL-6 (25 ng/ml; PeproTech, USA); and 500 nM staurosporine (ALX-380-014-C250, Enzo Life Sciences, NY, USA). Diagnostic histopathology, immunology, and virology studies were conducted in accredited regional diagnostic laboratories to standard protocols.
Whole-exome sequencing analysis was performed on DNA isolated from whole blood from patients I:1, I:2, II:3, and II:4. Capture and library preparation was undertaken using the BGI V4 exome kit (BGI, Beijing, China) according to manufacturer’s instructions, and sequencing was performed on a BGISEQ (BGI). Bioinformatics analysis and variant confirmation by Sanger sequencing are described in the Supplementary Materials.
RNA was extracted by lysing fibroblasts in TRIzol reagent (Thermo Fisher Scientific) or from whole blood samples collected in PAXgene tubes (PreAnalytix), as described previously (49). Further details, including primer/probe information, are summarized in the Supplementary Materials and table S5.
Whole-blood transcriptome expression analysis was performed using nine whole blood samples, from the proband taken before and during treatment, and five controls. In addition, the four patient II:3 samples taken before treatment and samples from six patients with mutations in TREX1, three with mutations in RNASEH2A, seven with mutations in RNASEH2B, five with mutations in RNASEH2C, five with mutations in SAMHD1, four with mutations in ADAR1, two with mutations in IFIH1, three with mutations in ACP5, three with mutations in TMEM173, and three with mutations in DNASE2 were analyzed, as described in the Supplementary Materials. RNA integrity was analyzed with Agilent 2100 Bioanalyzer (Agilent Technologies). mRNA purification and fragmentation, complementary DNA (cDNA) synthesis, and target amplification were performed using the Illumina TruSeq RNA Sample Preparation Kit (Illumina). Pooled cDNA libraries were sequenced using the HiSeq 4000 Illumina platform (Illumina). Fibroblasts grown in six-well plates were mock-treated or treated with IFNα for 6 or 12 hours, followed by extensive washing and 36-hour rest, before RNA extraction. The experiment was performed with patient II:3 and control cells (n = 3) in triplicate per time point. RNA was extracted using the ReliaPrep RNA Miniprep kit (Promega) according to manufacturer’s instructions and processed as described above, before sequencing on an Illumina NextSeq500 platform. Bioinformatic analysis is described in the Supplementary Materials. PMBC and fibroblast STAT2 patient and control data have been deposited in ArrayExpress (E-MTAB-7275) and Gene Expression Omnibus (GSE119709), respectively.
Details of lentiviral constructs, mutagenesis, and preparation are included in the Supplementary Materials. Cells were spinoculated in six-well plates for 1.5 hours at 2000 rpm, with target or null control viral particles, at various dilutions in a total volume of 0.5 ml of DMEM-10 containing hexadimethrine bromide [polybrene (8 μg/ml); Sigma-Aldrich]. Cells were rested in virus-containing medium for 8 hours and then incubated in fresh DMEM-10 until 48 hours, when they were subjected to selection with puromycin (2.0 μg/ml) or blastocidin (2.5 μg/ml) (Sigma-Aldrich). Antibiotic-containing medium was refreshed every 72 hours.
EBV B cells were seeded at a density of 8 × 105 cells/ml in serum-free X-VIVO 15 medium (Lonza, Basel, Switzerland) and stimulated with IFNα (1000 IU/ml) for the indicated times. After staining with Zombie UV (BioLegend, San Diego, CA, USA), cells were fixed using Cytofix buffer (BD Biosciences, Franklin Lakes, NJ, USA). Permeabilization was achieved by adding ice-cold PermIII buffer (BD Biosciences, Franklin Lakes, NJ, USA), and cells were incubated on ice for 20 min. After repeated washing steps with phosphate-buffered saline (PBS)/2% fetal bovine serum (FBS), cells were stained for 60 min at room temperature with directly conjugated antibodies (table S6). Samples were acquired on a Symphony A5 flow cytometer (BD Biosciences) and analyzed using FlowJo (FlowJo LLC, Ashland, OR, USA). The gating strategy is shown in fig. S13.
Immunoblotting was carried out as previously described (1) and analyzed using either a G:BOX Chemi (Syngene, Hyarana, India) charge-coupled device camera with GeneSnap software (Syngene) or a LI-COR Odyssey Fc (LI-COR, NE, USA). Densitometry analysis was undertaken using ImageStudio software (version 5.2.5, Li-COR). For complement studies, sodium dodecyl sulfate (SDS)–polyacrylamide gel electrophoresis (PAGE) under nonreducing conditions was performed on patient/parental serum [diluted 1:125 in nonreducing buffer (PBS)] or affinity-purified factor H (diluted to 200 ng in nonreducing buffer), separated by electrophoresis on a 6% SDS-PAGE gel, and transferred to nitrocellulose membranes for immunoblotting (antibodies in table S6). Blots were developed with Pierce ECL Western blotting substrate (Thermo Fisher Scientific) and imaged on a LI-COR Odyssey Fc (LI-COR).
U6A cells were lysed in immunoprecipitation buffer [25 mM Tris (pH 7.4), 1 mM EDTA, 150 mM NaCl, 1% Nonidet P-40, 1 mM sodium orthovanadate, and 10 mM sodium fluoride, with complete protease inhibitor (Roche, Basel, Switzerland)]. Lysates were centrifuged at 13,000 rpm at 4°C for 10 min. Soluble fractions were precleared for 1 hour at 4°C with Protein G Sepharose 4 (Fast Flow, GE Healthcare, Chicago, USA) that had been previously blocked with 1% bovine serum albumin (BSA) IP buffer for 1 hour. Precleared cell lysates were immunoprecipitated overnight with blocked beads that were incubated with anti-STAT2 antibody (A-7) for 1 hour and then washed three times in IP buffer before boiling with 4× lithium dodecyl sulfate buffer at 95°C for 10 min to elute the absorbed immunocomplexes. Immunoblot was carried out as described above.
Fibroblasts grown on eight-well chamber slides (Ibidi, Martinsried, Germany) were fixed with 4% paraformaldehyde in PBS for 15 min at room temperature before blocking/permeabilization with 3% BSA/0.1% Triton X-100 (Sigma-Aldrich) in PBS. Cells were incubated overnight with anti-STAT2 primary antibody (10 μg/ml; C20, Santa Cruz Biotechnology, Dallas, USA) at 4°C, and cells were washed three times with PBS. Secondary antibody [goat anti-rabbit Alexa Fluor 488 (1 μg/ml), Thermo Fisher Scientific] incubation was performed for 1 hour at room temperature, followed by nuclear staining with 4′,6-diamidino-2-phenylindole (DAPI; 0.2 μg/ml; Thermo Fisher Scientific). Cells were imaged with an EVOS FL fluorescence microscope with a 10× objective (Thermo Fisher Scientific). The use of STAT2-deficient cells (23) demonstrated the specificity and lack of nonspecific background of the staining approach. Image analysis was performed in ImageJ. The DAPI (nuclear) image was converted to binary, and each nucleus (object) was counted. This mask was overlaid onto the STAT2 image, and the mean fluorescence intensity of STAT2 within each nucleus was calculated (see also fig. S9). About n = 100 cells were analyzed per image.
The structure of human STAT2 has not been experimentally determined. We therefore used comparative modeling to predict the structure. The sequences of both the WT and mutant were aligned to mouse STAT2 (Protein Data Bank code 5OEN, chain B). For each sequence, 20 models were built using MODELLER (50), and the one with the lowest discrete optimized protein energy score was chosen. Protein structures and electrostatic surfaces were visualized with PyMOL (Schrodinger, USA).
Fibroblasts grown on 96-well plates were treated with IFNα (1000 or 10,000 IU/ml) or DMEM-10 alone for 72 hours. Cells were fixed in PBS containing 5% formaldehyde for 15 min at room temperature and then incubated with crystal violet stain. Plates were washed extensively then allowed to air dry. The remaining cell membrane-bound stain was solubilized with methanol and absorbance at 595 nm measured on a TECAN Sunrise plate reader (Tecan, Switzerland). Background absorbance was subtracted from all samples, and the results were expressed as a percentage of the absorbance values of untreated cells.
Antiviral protection assay
Fibroblasts grown on 96-well plates were pretreated in septuplicate for 18 hours with twofold serial dilutions of IFNα and IFNγ, followed by infection with mCherry-expressing parainfluenza virus 5 (PIV5) in DMEM/2% FBS for 24 hours. Monolayers were fixed with PBS containing 5% formaldehyde, and infection was quantified by measuring mean fluorescence intensity of mCherry (excitation, 580/9; emission, 610/20) using a TECAN Infinite M200 Pro plate reader (Tecan, Switzerland). Background fluorescence was subtracted from all samples, and the results were expressed as a percentage of the fluorescence values of untreated, virus-infected cells.
Unless otherwise stated, all experiments were repeated a minimum of three times. Data were normalized/log10-transformed before parametric tests of significance in view of the limitations of ascertaining distribution in small sample sizes and the high type II error rates of nonparametric tests in this context. Comparison of two groups used t test or one-sample t test if data were normalized to control values. Comparisons of more than one group used one-way analysis of variance (ANOVA) or two-way ANOVA as appropriate, with posttest correction for multiple comparisons. Statistical testing was undertaken in GraphPad Prism (v7.0). All tests were two-tailed with α ≤ 0.05.
Materials and Methods
Supplementary case summary
Fig. S1. Ingenuity pathway analysis of whole blood RNA-seq data.
Fig. S2. Longitudinal series of laboratory parameters.
Fig. S3. Multiple sequence alignment of STAT2.
Fig. S4. Factor H genotyping and mutant factor H purification strategy.
Fig. S5. Functional analysis of factor H Tyr779Cys variant.
Fig. S6. Immunoblot analysis of MX1 expression in PBMCs.
Fig. S7. Transduction of STAT2-deficient primary fibroblasts.
Fig. S8. Prolonged STAT2 phosphorylation in PBMCs.
Fig. S9. STAT2 immunofluorescence image analysis.
Fig. S10. STAT phosphorylation is not prolonged in patient cells in response to IFNγ or IL-6.
Fig. S11. RT-PCR analysis of gene expression in whole blood.
Fig. S12. STAT2R148W does not impair regulation of STAT1 signaling.
Fig. S13. Phosflow gating strategy.
Table S1. Laboratory parameters, patients II:3 and II:4.
Table S2. Rare variants segregating with disease.
Table S3. Digital ELISA detection of IFNα protein concentration.
Table S4. Phenotypes of monogenic defects of USP18 expression and/or function.
Table S5. RT-PCR primers and probes.
Table S6. Antibodies.
Data file S1. Raw data (Excel).
Acknowledgments: We are grateful to the patients and our thoughts are with their family. Funding: British Infection Association (to C.J.A.D.), Wellcome Trust [211153/Z/18/Z (to C.J.A.D.), 207556/Z/17/Z (S.H.), and 101788/Z/13/Z (to D.F.Y. and R.E.R.)], Sir Jules Thorn Trust [12/JTA (to S.H.)], UK National Institute of Health Research [TRF-2016-09-002 (to T.A.B.)], NIHR Manchester Biomedical Resource Centre (to T.A.B.), Medical Research Foundation (to T.A.B.), Medical Research Council [MRC, MR/N013840/1 (to B.J.T.)], MRC/Kidney Research UK [MR/R000913/1 (to Vicky Brocklebank)], Deutsche Forschungsgemeinschaft [GO 2955/1-1 (to F.G.)], Agence Nationale de la Recherche [ANR-10-IAHU-01 (to Y.J.C.) and CE17001002 (to Y.J.C. and D.D.)], European Research Council [GA 309449 (Y.J.C.); 786142-E-T1IFNs], Newcastle University (to C.J.A.D.), and ImmunoQure for provision of antibodies (Y.J.C. and D.D.). C.L.H. and R.S. were funded by start-up funding from Newcastle University. D.K. has received funding from the Medical Research Council, Wellcome Trust, Kidney Research UK, Macular Society, NCKRF, AMD Society, and Complement UK; honoraria for consultancy work from Alexion Pharmaceuticals, Apellis Pharmaceuticals, Novartis, and Idorsia; and is a director of and scientific advisor to Gyroscope Therapeutics. Author contributions: Conceptualization: C.J.A.D., S.H., and T.A.B. Data curation: C.F., G.I.R., A.J.S., J.C., A.M., R.H., Ronnie Wright, and L.A.H.Z. Statistical analysis: C.J.A.D., B.J.T., R.C., G.I.R., F.G., D.F.Y., S.C.L., V.G.S., A.J.S., L.A.H.Z., C.L.H., D.K., and T.A.B. Funding acquisition: C.J.A.D., D.D., Y.J.C., R.E.R., D.K., S.H., and T.A.B. Investigation: C.J.A.D., B.J.T., R.C., F.G., G.I.R., D.F.Y., Vicky Brocklebank, V.G.S., B.C., Vincent Bondet, D.D., S.C.L., A.G., M.A., B.A.I., R.S., Ronnie Wright, C.L.H., and T.A.B. Methodology: C.J.A.D., B.J.T., R.C., F.G., D.F.Y., A.J.S., D.D., K.R.E., Y.J.C., R.E.R., C.L.H., and D.K. Project administration: C.J.A.D., K.R.E., S.H., and T.A.B. Resources: S.M.H., Robert Wynn, T.A.B., J.H.L., J.P., E.C., S.B., K.W., and D.K. Software: C.F., A.J.S., M.Z., L.A.H.Z., and Ronnie Wright. Supervision: C.J.A.D., K.R.E., Y.J.C., D.D., C.L.H., R.E.R., D.K., S.H., and T.A.B. Validation: B.J.T., R.C., A.J.S., V.G.S., and C.L.H. Visualization: C.J.A.D., B.J.T., R.C., and S.C.L. Writing (original draft): C.J.A.D., with B.J.T., R.C., S.H., and T.A.B. Writing (review and editing): C.J.A.D., G.I.R., A.J.S., S.C.L., M.Z., S.M.H., K.R.E., R.E.R., D.K., S.H., and T.A.B. Competing interests: The authors declare that they have no competing interests. Data and materials availability: GEO accession: GSE119709. ArrayExpress accession: E MTAB-7275. Materials/reagents are available on request from the corresponding author(s). MBI6 is available from Claire Harris under a material agreement with Newcastle University. The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR, or the UK Department of Health.