Human USP18 deficiency underlies type 1 interferonopathy leading to severe pseudo-TORCH syndrome

We studied a Turkish consanguineous family (family A) with three children (patient 1 [P1], P2, and P3) affected by PTS with severe brain damage. We also analyzed a previously reported nonconsanguineous German family (family B) with two children (P4 and P5) affected by a similar disorder (Knoblauch et al., 2003; Fig. 1 A). Pre- and postnatal brain ultrasound and MRI scans showed an enlargement of lateral ventricles, calcifications, and hemorrhages (Fig. 1 B). Parents in both families are healthy and there was no medical indication to perform brain MRI scans.

In family A, the antenatal presentation of the condition in P1 led to a termination of pregnancy at 23 wk of gestation. In P2 and P3, presentation was neonatal and death occurred within two weeks. P2 and P3 had severe intracerebral hemorrhages and developed liver dysfunction, ascites, and lactic acidosis (Fig. 1 B, a–k). P2 also displayed severe thrombocytopenia with petechiae. Both siblings developed respiratory failure requiring ventilation, together with bradycardia, and they died at the ages of 7 and 17 d, respectively (see Supplemental Text for clinical details). Blood cells and cerebrospinal fluid were not available. In family B, two brothers (P4 and P5) presented a PTS-like phenotype with both neurological and systemic features, including thrombocytopenia, petechiae, ascites, hepatomegaly, and systemic calcifications, and were previously reported. (Knoblauch et al., 2003; Fig. 1 B, i). The clinical findings for both families are summarized in Table 1.

We performed haplotype linkage analysis assuming the autosomal recessive inheritance of a homozygous mutation in family A. We demonstrated only one overlapping area of homozygosity among affected individuals, on chromosome 22q11.1-22q11.21, with a suggestive LOD score of 2.83 (Lander and Kruglyak, 1995) (Fig. 1 C). The linked region spans the markers rs5746398-rs5757914 (chr22:17929926-22694656, GRCh37.p10) and contains 200 genes (Fig. 1 C and Fig. S1). Whole-exome sequencing failed to identify a putative pathogenic mutation, as the area was poorly covered (Tables S1 and S2). Capillary DNA sequencing of the missing exons in the linked region (Table S3) identified a homozygous mutation (c.652C>T) in the three patients that was present in the heterozygous state in the healthy brother and both parents (primers in Table S4). This variant mapped to exon 7 of the ubiquitin-specific protease 18 (USP18) gene (OMIM #607057) and was predicted to result in replacement of the glutamine 218 residue by a premature stop codon (p.Gln218X; Fig. 1 D). The variant was not found in 190 Turkish control chromosomes, or in the dbSNP or ExAC databases. Analysis of 952 DNA samples sequenced by Complete Genomics (CG; whole genome sequencing, containing 481 samples of Wellderly-aged Americans cohort, 46 public CG data from in-house database of the Erasmus Medical Center, and 425 samples from the 1,000 Genomes Project), which provide full coverage of USP18, showed no c.652C>T variant, thus indicating that this is an extremely rare variant (Fig. S2).

In family B, Sanger sequencing of USP18 in P4 and P5 revealed the same mutation (c.652C>T) present in the heterozygous state (Fig. 1 D). Haplotype analysis of 86 SNPs on Illumina cyto-snp-850K-B arrays covering the 400-kb area across the USP18 allele containing the c.652C>T change (chr 22: 18483880–18891398, hg19; USP18 spans bp 18632666–18660164) revealed a shared c.652C>T allele between the two families, consistent with remote common ancestry (Table S5). Capillary DNA sequencing of USP18 is laborious because of its high homology with the pseudogene USP41 (Fig. S3). Genomic sequencing with USP18-specific primers did not detect a second mutation in family B. Quantitative RT-PCR of USP18 mRNA in fibroblasts was performed using primers across exon 4 (forward) to exon 6 (reverse), before and after stimulation with IFN-β-1a (Fig. 1 E; primer sequences in Table S4). P1 and P2 showed low basal and induced transcript levels, whereas P4 and P5 showed an impaired induced transcript level. On the assumption of a cryptic mutation on the second allele in family B, we designed six different primer sets covering the whole cDNA (Fig. 2 A), including primers to amplify only the WT allele not carrying the c.652C>T mutation (marked with a red x in the schematic of Fig. 2 A) and performed RT-PCR.

RT-PCR amplifying exon 1–6 of USP18 yielded the expected 782-bp product in the control (C1), the healthy heterozygote sibling of family A (Het FamA), and P4 of family B, consistent with an intact 5′ portion of the transcript. The lower intensity band detected in homozygous P1, from family A, was consistent with nonsense-mediated RNA decay (Fig. 2 B, product A). RT-PCR amplifying exon 6–9 of USP18 yielded the expected 393-bp band in C1 and Het FamA, but a lower intensity band was detected in the homozygous P1, consistent with nonsense-mediated RNA decay. A normal product was detected in P4, consistent with heterozygosity at position c.652C>T (Fig. 2 B, product B). RT-PCR amplifying exon 6–11 of USP18 yielded the expected 863-bp product in C1 and Het FamA. In P1, this band was of lower intensity, consistent with nonsense-mediated decay, but importantly, P4 yielded an even more diminished 863-bp product, consistent with nonsense-mediated decay from one allele (c.652T) and a cryptic mutation after exon 9, possibly affecting splicing (Fig. 2 B, product C).

RT-PCR amplifying only the WT c.652C allele was also performed (red cross in schematic of Fig. 2 A). Amplifying exon 1–7 yielded the expected 830-bp product in C1, Het FamA, and P4, consistent with an intact 5′ portion of the transcript. The product was essentially absent in the homozygous P1, consistent with primer mismatch (Fig. 2 B, product D). Amplifying exon 7–9 yielded the expected 321-bp product in P4, similar to C1 and Het FamA, and no product in P1, again as a result of primer mismatch (Fig. 2 B, product E).

Amplifying exons 7–11 yielded the expected 791-bp product in C1 and Het Fam A, no product in P1, consistent with primer mismatch, and no product in P4 consistent with a cryptic mutation on the second allele (Fig. 2 B, product F). These experiments confirmed the presence of an altered transcript in family B.

Additional sequencing of the introns between exon 9 and 11 did not show any new variant in family B (primers in Table S4). This excluded a splicing error as explanation for the altered transcript.

Finally, 3′ RACE PCR was performed. For the first PCR, 5′ primers starting in exon 6 up to the poly-A tail were designed. The resulting product was amplified by nested PCR using 5′ primers in exon 7, annealing only to the WT c.652C sequence (Fig. 2 A, amplicons G and H). This experiment revealed four different products in C1, Het FamA, P1, and P4 (Fig. 2 C, band 1–4). Upon sequencing, the 1,200-bp product (Fig. 2 C, band 1) corresponds to the WT sequence of USP18 and is present in C1 and Het FamA, but is not detectable in P4. A 1,000-bp product (Fig. 2 C, band 2) is only present in P1. This product corresponds to an alternative transcript of USP41 fused with a part of the 3′ distal gene LOC105372866, a result of biallelic primer mismatch with the nested primer in USP18. An 850-bp product (Fig. 2 C, band 3) is present in C1, Het FamA, P1, and P4 and corresponds to an alternative transcript of USP41. A product of ∼800 bp is exclusively present in P4 (Fig. 2 C, band 4). Capillary sequencing of this product shows a repeat-rich sequence that maps distal to the 3′ UTR region of USP18, suggestive of a large deletion at the 3′ end of USP18 as the second mutation in family B (cryptic 3′ deletion).

Western blot analysis of lysates prepared from skin fibroblasts, with and without IFN-β stimulation, using monoclonal antibodies against the N-terminal part of USP18, revealed a complete lack of USP18 protein in the patients from both family A and B in contrast to healthy controls (Fig. 2 D). These results indicate that both the c.652C>T and the cryptic 3′ deletion in USP18 alleles produce no or abnormal transcripts, respectively, which are not translated into normal USP18 protein. Altogether these genomic, genetic, molecular, and biochemical analyses lead us to conclude that the complete USP18 deficiency results from autosomal recessive homozygous USP18 mutations in family A and from compound heterozygous mutations in family B.

USP18 is a type 1 IFN-stimulated gene (ISG) that encodes a protein that exerts dual functions: regulatory and enzymatic. USP18 is a negative feedback regulator of type I IFN signaling (Malakhova et al., 2006). USP18 is also an isopeptidase that cleaves the ubiquitin-like ISG15 protein from conjugated proteins (Zhang and Zhang, 2011). The study of Usp18^−/− mice highlighted a role of USP18 in the CNS. Yet, the Usp18^−/− phenotype appears to be dependent on the genetic background of mice. In the fully backcrossed FVB/N background, a majority of Usp18^−/− mice were reported to have a normal lifespan, in the C57BL/6 background Usp18^−/− mice died before E15.5, and in a mixed C57B/L6 × 129/Sv background they reached birth and showed an intermediate phenotype (Zhang and Zhang, 2011). All mice displayed varying degrees of neurological signs and brain inflammation, suggesting that the lack of USP18 has an impact on the CNS that differs in severity depending on the background (Ritchie et al., 2002; Goldmann et al., 2015). These CNS alterations were reminiscent of the sequelae of viral infection (Ritchie et al., 2002). Importantly, none of these diverse murine phenotypes were recognized as potential mimics of the human pseudo-TORCH syndrome.

In vivo brain imaging in USP18-deficient patients mostly showed severe hemorrhage, with findings reminiscent of those in other cases of perinatal hemorrhage, such as genetic porencephaly caused by COL4A1 and COL4A2 mutations (Meuwissen et al., 2015). Genetic testing in a cohort of 38 patients with isolated prenatal or infantile cerebral hemorrhage, where COL4A1 and COL4A2 mutations had been excluded, failed to identify additional USP18 mutations. We therefore deduce that USP18 mutations are infrequent.

Postmortem fetal brain pathology in P1 showed microencephaly, with a brain weight more than three standard deviations below the mean value for gestational weight (35 g), a thin cortex, and enlarged ventricles. At hematoxylin and eosin (H&E) staining, there was irregular ependymal layering with curvilinear shape (Fig. 2 E, a), columnar arrangement, and focal erosion (Fig. 2 E, b), compared with age-matched control (Fig. 2 E, c). There were trajectories of polymicrogyria reflecting defective neuronal migration (Fig. 2 E, d). In the white matter, heterotopic neural cell clusters were observed. In the layers below the ependyme, areas of hemorrhage with hemosiderin-laden macrophages were observed (Fig. 2 E, e, f, and g), with lysis, rarefaction, necrosis, and dystrophic calcification in the germinal matrix areas, the white matter, and the cortex. We next stained the brains of the patient and control with markers for vessels (CD34), microglia (Iba-1), MHC class II (HLA-DR), and astrocytes (GFAP; Fig. 2 F). In the patient brain, there is a strong induction of the brain innate immune system, as indicated by the robust presence of astrocytes, microglia, and up-regulation of MHC class II molecules. Positive staining for phosphorylated STAT1 (pSTAT1) in the patient versus the control brain (Fig. 2 F, pSTAT1) was suggestive of activation of type I IFN pathway.

USP18 functions as a negative feedback regulator of type 1 IFN signaling, down-regulating STAT-mediated ISG transcription. qRT-PCR was performed on cDNA from P1 and P2 and from control fibroblasts, with and without IFN-α stimulation, to determine the level of expression of several ISGs, IFIT1, MX2, and ISG15. Transcripts for the IFIT1, MX2, and ISG15 were much more abundant in the cells of P1 and P2 than in those of controls after stimulation with IFN (Fig. 3, A–C). Next, we monitored the levels of phosphorylated STAT2 and of the ISG-encoded proteins MX1, IFIT1, USP18, and ISG15 in patient and control cells (Fig. 3 D). Persistent STAT2 phosphorylation and high levels of induced MX1 and IFIT1 proteins were detected in patients’ cells as compared with WT controls. Analysis of the non–type I IFN-related gene product IL-6 showed no abnormal regulation (Fig. 3 E).

To ascertain that this enhanced type 1 IFN response with up-regulation of ISG expression was indeed caused by USP18 deficiency, we transduced USP18-deficient fibroblasts from P1 and P2 and control fibroblasts with lentiviral particles (LV)-expressing luciferase (Luc) or WT USP18. Unlike LV-Luc, LV-WT USP18 completely rescued the enhanced induction of ISGs at the mRNA level (Fig. 4, A–C). Similarly, at the protein level, unlike LV-Luc, LV-WT USP18 transduced patient fibroblasts exhibited reduced levels of phosphorylated STAT2, MX1, and IFIT1 (Fig. 4 D). ISGylated proteins are main targets of USP18 protease activity (Zhang and Zhang, 2011). We therefore determined global ISGylation levels and the amount of unconjugated ISG15. Consistently, fibroblasts from P1 and P2 stimulated with IFN-α displayed significantly greater amounts of ISGylated proteins than control cells (Fig. 4 E). As expected, LV-WT USP18 transduction, but not LV-Luc, decreased global protein ISGylation levels (Fig. 4 E). siRNA against USP18, but not control siRNA, in primary control fibroblasts completely phenocopied USP18-deficient cells when challenged with IFN (Fig. 4 F and G). Collectively, these rescue experiments indicate that the loss of USP18 is responsible for abnormal activation of IFN signaling. Moreover, our results demonstrate that, in humans, USP18-mediated regulation of IFN response is crucial for normal CNS development, and that its deficiency causes severe PTS. Yet, the diversity of phenotypes observed in the Usp18^−/− mice of different genetic background likely reflects different basal levels of IFNs, which itself points to the contribution of additional genetic or epigenetic factors.

Notably, the brain pathology of USP18-deficient patients (Fig. 2 E, d) is similar to that observed in cases of congenital infection with neurotropic viruses, such as cytomegalovirus and lymphocytic choriomeningitis virus, with activation of meningeal and hematopoietic-derived macrophages and microglia and the induction of cytokines including type 1 IFN (Bale, 2009; Bonthius, 2012; Sakao-Suzuki et al., 2014). Our observations are consistent with previous studies suggesting that, in some congenital viral infections, the brain damage may owe as much to the host response as to the infectious agent itself (Gresser et al., 1980). This is crucial for future considerations on management of congenital infections, as targeting only the virus itself will not be sufficient to avoid permanent damage to the fetus.

Interestingly, the mechanism behind type I interferonopathies described to date in AGS differs from the one operating in USP18 deficiency, as type I IFN dysregulation in USP18 deficiency is not a result of aberrant induction of these cytokines but is caused by an aberrant response to type I IFN. In support of this, although body fluids of the patients were not stored, IFNB mRNA and protein level were comparable to controls when patient cells were stimulated with Poly (I:C; Fig. 4, H and I).

Human ISG15 deficiency leads to milder USP18 dysfunction and causes cerebral abnormalities compatible with adult life, such as basal ganglia calcification and occasional seizures (Zhang et al., 2015). The cells of ISG15-deficient patients demonstrated reduced, albeit still detectable, amounts of USP18, and displayed defective down-regulation of type 1 IFN signaling. We speculate that biallelic hypomorphic USP18 mutations (i.e., leading to residual protein function) may cause milder cerebral abnormalities with prolonged survival, possibly overlapping the phenotypic spectrum of ISG15 mutations. Therapeutically, USP18 represents an attractive target, with agonists being of use in clinical settings where there is overabundance of type I IFNs and USP18 antagonists acting to prolong the beneficial effects of therapeutic IFNs, as in multiple sclerosis, hairy cell leukemia, and melanoma. Likewise, in genetic counseling and prenatal diagnosis of severe perinatal PTS, USP18 should be added to the list of genes to be tested. Collectively, our results establish USP18 deficiency as a novel genetic cause of severe PTS.

Written parental consent was obtained. Genetic tests were performed according to The Erasmus University Medical Center’s local ethics board approved protocol MEC-2012387.

Linkage analysis was performed on Illumina CytoSNP12v2.1 array data for the three affected siblings, their healthy brother, and both parents of family A. Genotype data were available for >300,000 SNPs per individual. Linkage analysis was performed with Merlin v1.0.1 (in the Easy Linkage Plus v5.08 package). An autosomal recessive model was specified, with an estimated allele frequency of 0.0001 and 100% penetrance. Closely spaced SNP markers were used for the analysis, so the genome analysis was performed with predefined spacings of 0.1 and 0.15 cM. Genomic regions with LOD scores >2.5 were considered to be candidate intervals. Haplotypes were constructed and graphic visualizations were performed with HaploPainter v029.5 (Hoffmann and Lindner, 2005) to facilitate inspection and analysis.

Whole-exome sequencing was performed on DNA extracted from the fibroblasts of P1 and peripheral leukocytes from the parents and P2 and P3 of family A with the Sure Select All Exome V4 capture kit (Agilent) on a HiSeq2000 machine (Illumina). The statistics are provided in Table S1. BWA version 0.5.9 (Li and Durbin, 2009) and NARWHAL software (Brouwer et al., 2012) were used to align reads with the reference genome (hg19). Variants were called with GATK (version 1.2–29; McKenna et al., 2010) and SAMtools (version 0.1.13). Variants were annotated with ANNOVAR software (revision 511; Wang et al., 2010) and filtered with TIBCO Spotfire version 4.5 (Tables S1 and S2).

Brain autopsy material was fixed in 4% paraformaldehyde and embedded in paraffin. Sections (4 µm thick) were cut and stained with H&E or Prussian blue, for iron visualization. Immunohistochemical staining was performed with a biotin-avidin and alkaline phosphatase/anti-alkaline phosphatase visualization. The primary antibodies used were as follows: anti-CD3 (T cells, rat monoclonal antibody; Serotec), anti-glial fibrillary acidic protein (anti-GFAP, rabbit polyclonal antibody; Dako), anti–HLA-DR (MHC class II, mouse monoclonal antibody, clone LN3; eBioscience), antiionized calcium-binding adaptor molecule (Iba)-1 antibody (microglia, macrophages; Wako), anti-CD34 (vessels, clone B-C34; Cell Sciences), and anti-pSTAT1 (clone 58D6, #9167; Cell Signaling Technology). The pictures were taken with a BZ-9000 Biorevo microscope (Keyence; Neu-Isenburg)

PCR and Sanger sequencing were conducted according to standard methods. The Sanger sequencing of USP18 was complicated by the presence of a pseudogene with a very similar sequence, USP41. Fig. S3 shows the alignment of these two genes, highlighting their similarity. Table S3 lists the exons for which there was little or no coverage, together with the primers used for Sanger sequencing. Table S4 lists the genomic primers for Sanger sequencing of USP18.

qRT-PCR was performed with a KAPA SYBR FAST qPCR kit (Kapa Biosystems) in the CFX96 Real-Time system (Bio-Rad Laboratories). Thermal cycling conditions were as follows: denaturation (95°C for 3 min), followed by 35 cycles of denaturation (95°C for 5 min), annealing, and extension (60°C for 30 min). Fluorescence detection and data analysis were performed with CFX Manager 2.0 (Bio-Rad Laboratories). Experiments were performed in triplicate, with RNF111 (Entrez Gene ID 54778 [OMIM 605840]) as the reference gene for the normalization of expression data (Bustin et al., 2009) and COPS5 (Entrez Gene ID 10987 [OMIM 604850]) as a secondary housekeeping gene. The primers used are also listed in Table S4. For qRT-PCR after transduction, see Stimulation and qRT-PCR.

Fibroblasts were routinely tested for mycoplasma infection and were cultured in 75-cm² culture flasks, in DMEM, to 80% confluence. Total RNA was extracted on RNeasy mini columns (QIAGEN) according to the manufacturer’s protocol. Reverse transcription was performed on 1 µg of RNA in a total volume of 20 µl, with the iScript cDNA Synthesis kit (Bio-Rad Laboratories) used according to the manufacturer’s instructions (Poulton et al., 2011). RT-PCR was performed using FASTStart Taq DNA polymerase (Sigma-Aldrich) according to manufacturer’s instructions. Primers for RT-PCR analysis for the experiments shown in Fig. 2 B are listed in Table S4.

Total RNA from control and patients fibroblasts was isolated using TRIzol (Ambion). 3′ RACE PCR was performed using the 3′ RACE System for Rapid Amplification of cDNA ends (Invitrogen), according to the manufacturer’s instruction.

3′ RACE Universal Amplification Primer (UAP), supplied in the kit, complementary to the adaptor primer, and a gene-specific primer designed in exon 6 of USP18 (5′-CAGAAACAGCAGCATGCTCACC-3′) were used for first amplification, followed by nested PCR with the UAP and a second gene-specific primer designed in exon 7 of USP18 (5′-CGCCCTGCACTGCTTCTTCC-3′).

Fibroblasts were stimulated for 12 or 24 h at 37°C with either 110 U/ml human IFN-β-1a (11415–1; pbl interferon source) or 10–1,000 U/ml IFN-α-2b (Merck IntronA) in complete medium. The IFN was removed by thorough washing with PBS, and the cells were either harvested or allowed to rest for 36 h at 37°C in normal growth medium. Cells were harvested for RNA isolation.

In Fig. 2 C, cell pellets were resuspended in a previously described buffer (Francois-Newton et al., 2012; 50 mM Tris/HCl, pH 8.0, 150 mM NaCl, 1% Triton X100, 0.5% sodium deoxycholate, 0.03% SDS, and 1 mM Na₃VO₄). We ran 40 µg of total protein on a precast Criterion XT 4–12% bis Tris gel (Bio-Rad Laboratories). The blots were incubated with anti-USP18 (1:1,000; rabbit monoclonal; Cell Signaling Technology #4813; recognizing residues surrounding amino acid Pro45), anti-ISG15 (1:500, rabbit polyclonal), or anti-GAPDH (mouse 1:1,000; monoclonal; Li-Cor Biosciences) primary antibodies; 680 nm goat anti–rabbit and 800 nm goat anti–mouse conjugates were used as secondary antibodies. In Figs. 3 C and 4 (D and E), cell lysates were prepared in modified RIPA buffer (50 mM Tris-HCl, pH 8.0, 200 mM NaCl, 1% NP-40, 0.5% DOC, 0.05% SDS, and 2 mM EDTA) with 1 mM Na₃VO₄ and a cocktail of proteases inhibitors (Roche). Rec IFN-α2 was provided by D. Gewert (Wellcome Trust, London, England). We used antibodies against STAT2 (EMD Millipore), phosphorylated Tyr689 STAT2 (EMD Millipore), USP18 (Cell Signaling Technology), β-tubulin (Takara Bio Inc.), ISG15 (gift from E.C. Borden, Cleveland Clinic, Cleveland, OH), IFIT1 (gift from G. Sen, Cleveland Clinic, Cleveland, OH), and Mx1 (MxA; gift from O. Haller, University of Freiburg, Freiburg, Germany). Signal was detected with enhanced chemiluminescence detection reagent (Western Lightning; PerkinElmer) and acquired with a Fuji ImageQuant LAS-4000 machine. Control primary human fibroblasts were silenced by incubation for 24 h with 25 nM USP18-siRNA (Dharmacon; pool-siRNA), followed by priming with IFN. siRNA transfections were performed with Lipofectamine RNAi max reagent (Invitrogen) according to the manufacturer’s instructions.

Lentiviral particles were produced by transfecting HEK-293T cells, in the presence of Lipofectamine-2000 (Invitrogen), with the following constructs: pCAGGS-VSV-G, pCMV-Gag/Pol, and pTRIP-hUSP18-IRES-RFP or pTRIP-luciferase-IRES-RFP. Supernatants were collected 36 h after transfection and clarified by centrifugation. The fibroblasts were covered with lentiviral inoculum and incubated at 37°C. Supplemental growth medium was added, and transduction efficiency was checked by fluorescence microscopy. Transduced fibroblasts were sorted (FACS Aria II; BD) into (−), low (Lo), medium (Med), and high (Hi) expression groups on the basis of RFP expression. We focused on the Lo expression group. Transduced cell populations were expanded and grown in the same conditions as untransduced cells, and the level of transduced gene expression was periodically confirmed by fluorescence microscopy.

Untransduced luciferase- and USP18-transduced fibroblasts were treated with 0 U/ml (mock), 10 U/ml, 100 U/ml, or 1,000 U/ml interferon α-2b (Merck IntronA) in normal growth medium for 12 h at 37°C. The IFN was removed by thorough washes with PBS, and the cells were allowed to rest for 36 h at 37°C in normal growth medium. The cells were then lysed and RNA was extracted with a kit, according to the manufacturer’s instructions (RNeasy; QIAGEN). RNA was reverse-transcribed according to the enzyme manufacturer’s instructions (ABI High Capacity Reverse transcription). The relative expression levels of multiple IFN-stimulated genes (ISGs; MX1, MX2, RSAD2, ISG15, IFI44L, and IFIT1) were analyzed by TaqMan quantitative real-time PCR on technical duplicates (TaqMan Universal Master Mix II w/UNG; LightCycler 480 II; Roche). The relative expression of the ISGs was calculated by the ΔΔCT method, with comparison to the mean value for the untreated controls.

Fig. S1 shows linkage data and haplotyping. Fig. S2 shows CG control alleles. Fig. S3 shows alignment of USP18 and USP41. Table S1 shows exome sequencing data. Table S2 lists filtering steps for family A. Table S3 lists the exons for which there was little or no coverage, together with the primers used for Sanger sequencing. Table S4 lists the genomic primers for Sanger sequencing of USP18. Table S5 lists haplotypes of chromosome 22q11 in family A and B using Illumina cyto-snp-850 kb arrays. Clinical data from family A is provided as supplemental text.

We thank Dr. Peter van der Spek for sharing the whole genome sequencing control data and Dr. Kathleen Millen (University of Washington, Seattle, WA) for helpful discussions. We also thank Tom de Vries Lentsch for the graphical work and Julie Sappa for copyediting the manuscript.

Financial support was obtained by NutsOhra Funds project 1203-030 to G.M.S. Mancini. D. Bogunovic is supported by the National Institute of Allergy and Infectious Diseases grant number R00AI106942-02. Z. Li and S. Pellegrini acknowledge institutional support from Institut Pasteur, Centre National de la Recherche Scientifique, and Institut National de la Santé et de la Recherche Médicale. Y.J. Crow acknowledges the European Research Council (GA 309449).

The authors declare no competing financial interests.