Characterization of novel TMEM173 mutation with additive IFIH1 risk allele

TMEM173 encodes for STING that is a transmembrane protein activated by pathogen or self-derived cytosolic nucleic acids causing its translocation from ER to Golgi, and further to vesicles. Monogenic STING gain-of-function mutations cause early-onset type I interferonopathy, with disease presentation ranging from fatal vasculopathy to mild chilblain lupus. Molecular mechanisms causing the poor phenotypegenotype correlation are presently unclear. Here we report a novel gain-of-function G207E STING mutation causing a distinct phenotype with alopecia, photosensitivity, thyroid dysfunction, and STING-associated vasculopathy with onset in infancy (SAVI) -features; livedo reticularis, nasal septum perforation, facial erythema, bacterial infections and skin vasculitis. Single residue polymorphisms in TMEM173 and an IFIH1 T946 risk allele modify disease presentation in the affected multigeneration family, explaining the varying clinical phenotypes. The G207E mutation causes constitutive activation of inflammation-related pathways in HEK cells, as well as aberrant interferon signature and inflammasome activation in patient PBMCs. Protein-protein interactions further propose impaired cellular trafficking of G207E mutant STING. These findings reveal the molecular landscape of STING and highlight the complex additive effects on the phenotype. BRIEF SUMMARY Novel gain-of-function mutation in TMEM173, associated with single residue polymorphisms in TMEM173 and IFIH1, causes a distinct clinical phenotype with some shared features of SAVI.


Introduction
Monogenic interferonopathies are characterized by increased type I interferon signaling leading to vasculopathy, autoinflammation and systemic lupus erythematosus (SLE)-like disease (1). In these, increased sensing of endo-and exogenous nucleic acids leads to hyperactive Toll-likereceptors (TLRs), RNA and DNA sensors, and to an enhanced type I interferon (IFN) response (2). The upregulation of type I IFNs is thought to maintain the persistent self-directed immune reaction (3).
One of the 18 currently known monogenic interferonopathies is STING-associated vasculopathy with onset in infancy (SAVI; MIM 615934) that is caused by mutations in TMEM173 (4,5). TMEM173 encodes for the stimulator of interferon genes (STING), a transmembrane protein residing in the endoplasmic reticulum (ER). It has an essential role in sensing cytosolic double stranded DNA (dsDNA) and directly binding to bacterial second messengers, such as cyclic dinucleotides (CDNs) c-di-GMP, c-di-AMP and 3'3'-cGAMP (6,7). The recognition of nucleic acids or cyclic nucleotides initiates the production of type I IFN and other inflammatory cytokines leading to nucleicacid driven inflammation.
STING has four amino-terminal transmembrane domains spanning the first 136 amino acid residues, followed by the helix α1 at residues 153-177 (8). Helix α1, also known as the dimerization domain, is essential for protein stability, intraprotein interactions and ligand binding (9). The CDN binding domain (residues 153-340) is part of the cytoplasmic carboxy-terminus, stretching over residues 138-379 and having multiple phosphorylation and downstream signaling interaction sites (8,10).
The first described constitutively active TMEM173 mutations are situated at or near the hydrophobic helix α1 dimerization domain and associated with early-onset vasculitis, autoinflammation and interstitial lung disease that define the SAVI phenotype (5,11,12). A recent study identified five patients with a GOF mutation affecting the dimerization domain (13). In contrast to the severely affected infants (5) these patients only presented with mild skin vasculitis and were diagnosed with familial chilblain lupus (13). In addition to the mutations in the dimerization domain, proximal substitutions affecting the CDN binding domain were reported in single patients presenting with variable phenotypes of STINGassociated autoinflammation (14,15). Overall, all of the reported TMEM173 mutations have been GOF, leading to elevated IFNB production which activates the JAK/STAT-pathway and generates a positive feedback loop (16). As there is poor correlation between genotype and clinical phenotype, additional, poorly understood intrinsic or environmental factors likely modify the disease outcome.
Another important interferonopathy gene is the IFN-induced helicase C domain-containing protein 1 (IFIH1, also known as MDA5). IFIH1 functions as a nucleic acid sensor in the cytoplasm, inducing transcription of type I IFN and IFN-regulated genes upon activation. GOF IFIH1 variants are seen in Aicardi-Goutières and Singleton-Merten syndromes, both of which feature prominent vascular inflammation (17)(18)(19). Also, polymorphism in IFIH1 has been linked to SLE (rs1990760, p.A946T) (20, 21) leading to a variable phenotype spectrum (22). The A946T GOF risk variant leads to increased production of type I IFN, promoting inflammation and elevating the risk of autoimmunity. The T946 allele also modifies the effects of other autoimmune risk alleles, which leads to variable disease severity (23, 24).
Here we report a multigenerational family of six affected individuals, presenting with several lupoid-like features (photosensitivity, alopecia, autoimmune thyroiditis) and features of SAVI (bacterial skin infections, early-onset livedo reticularis, nasal septum perforation and facial erythema). Affected individuals have a combined genotype of (1) a novel disease-causing TMEM173 mutation G207E together with (2) either arginine or histidine at a polymorphic position 232 in TMEM173, and (3) homo/heterozygosity for the IFIH1 T946 risk allele. We describe the additive effects of these multiplexed single residue polymorphisms using cell models and patient material, highlighting the effect of common genetic variation in rare disease presentations. The results broaden the spectrum of published TMEM173 mutation-associated phenotypes, provide insight into the activation of alternative NLRP3 inflammasome and reveal the STING interactome.

Clinical presentation and immune phenotype
We evaluated a multigenerational Finnish family with a few SAVI-like features, but presenting also multiple symptoms previously not associated with SAVI or chilblain lupus (Supplementary figure  1A). Of 10 affected family members, six were available for detailed clinical evaluation ( Figure 1A) and the main clinical findings are summarized in Table 1. Affected family members presented with novel STING mutation induced features of photosensitivity, alopecia ranging from areata to universalis and autoimmune thyroiditis. Prominent features included early-onset persistent livedo reticularis, skin vasculitis, recurrent and severe infections variably reported in SAVI (Figure 1B-E). None of the patients had typical SAVI associated interstitial lung disease, violaceus facial, nasal, auricular or acral patches progressing to ulcers or necrosis, febrile attacks, failure to thrive, elevated inflammatory markers or autoantibodies (Supplementary figure 1B and Supplementary table 1). The most severely affected family members suffered from UV exposure induced cutaneous vasculitis, nasal septum perforation, severe infections including necrotizing cellulitis and abscesses concomitant with SAVI ( Table 1 and Supplementary figure 1B).
All six evaluated patients had increased peripheral CD4+ T cell and naïve CD4+CCR7 + CD45RA + T cell count ( Table 2). The central memory CD4+ T cells numbers remained unaffected, but pronounced decreases were observed in the effector memory subset. These observations are in accordance with recent findings on patients carrying activating TMEM173 mutations (25). Among B cell populations, both naïve (CD27 − IgD + ) and immune-exhausted activated (CD38 low CD21 low ) B cells were increased in 4/6 patients ( Table 2). Immunologic workup of these six patients further showed normal acute phase reactants, blood cell counts, various autoantibodies, complement activation (Supplementary  table 1), proliferative lymphocyte responses to mitogens and antibody responses to both protein and polysaccharide vaccines (data not shown).

Identification and structural analysis of TMEM173 mutation
Based on the pedigree (Figure 1A), we assumed an autosomal dominant inheritance. We performed linkage analysis on 12 unaffected and 6 affected family members and whole genome sequencing on patients IV.1 and V.2. Considering all affected, Merlin non-parametric linkage analysis revealed no statistically significant linkage peaks. NPL scores of >1.7 were observed at four loci on chromosomes 3, 5, 9, and 12. In these, only one co-segregating rare variant negatively affected an evolutionary conserved residue. This heterozygous missense mutation Chr5 (GRCh38): g.139478409C>T results in a p.Gly207Glu (G207E) substitution in TMEM173 encoding for STING and was absent from major public and in-house genome databases. It localizes next to a previously described mutation p.Cys206Tyr (C206Y) (14) lying within the CDN binding domain. The G207E mutation locates in a fold that participates in substrate binding and STING folding, and changes the electrically neutral glycine to negatively charged glutamic acid (Figure 1F-G). Based on the protein structure prediction, the G207E mutation is expected to change the electric charge of the protein fold, destabilizing it and thus potentially altering its function ( Figure 1G and Table 3). For further structural comparison of C206Y and G207E, we predicted the effects of single residue changes on protein structure and stability using Site Directed Mutator (SDM) (26). SDM predicted that both C206Y and G207E reduce the stability of the protein (pseudo ∆∆G -1.27/-1.28 and -0.72/-0.63, respectively Table 3) and likely lead to protein malfunction and disease. The C206Y mutation with newly acquired aromatic side chain results in more drastic structural alteration than G207E. Interestingly, polymorphism at position 232 in TMEM173 influenced G207E's relative side chain solvent accessibility suggesting greater ligand accessibility and binding with variant R232 compared to H232. C206Y, as a buried residue, remained solvent inaccessible independent of position 232.

Identification of disease-modifying polymorphisms in TMEM173 and IFIH1
A non-synonymous variant of human STING encoding arginine (R) at position 232 has been identified as a major allele in the general population, in contrast to the histidine (H) 232 commonly listed as the wildtype allele (6). In accordance with the population data, the majority of the affected family members were heterozygotes for major allele R232 (Table 1). Additionally, single nucleotide polymorphism (rs1990760, Ala946Thr (A946T)) in IFIH1 causes constitutively active IFN signaling and predisposition to autoimmune diseases (20, 23, 24). In the affected family 4/6 analyzed members had at least one copy of the IFIH1 T946 risk allele ( Table 1). When the clinical presentations of affected family members were correlated with their combined genotypes of the monogenic TMEM173 mutation G207E, and R232H variants, as well as the presence of IFIH1 T946, a trend of additive effect jointly promoting disease was observed from 207E/H232/AA towards 207E/ R232/TT ( Table 1).

G207E mutation in STING constitutively activates IFN-β, STAT1/2, STAT3 and NFKB transcription
To study the effect of STING mutations on activation of IFNB, STAT1/2, STAT3 and NFKB pathways, we utilized luciferase-based reporter assays specific for each pathway. The constitutively active STING p.Ans154Ser (N154S) SAVI mutant was used as a positive control in all the assays (5). To mimic the combined genotypes of the affected family members, we combined the G207E mutation either with variant R232 (major population allele and genotype of patients IV.1, V.2, III.4, IV.6; herein R+207E) or H232 (minor allele and genotype of patients III.9, III.7; herein H+207E). Both combinations clearly induced IFNB, STAT1/2 and STAT3 pathway activation, even in basal unstimulated state (Figure 2A-C), thus indicating 207E as a clear GOF mutation.
The patients were susceptible to bacterial infections, but only a few remembered ever having viral infections such as common cold. Thus, we stimulated the transfected HEK293 cells with 2'3'-cGAMP mimicking the mammalian response to cytosolic viral DNA, in addition to bacterial second messengers c-di-GMP and 3'3'-cGAMP (6,27). After CDN stimulations variant R+207E presented slight increases of pathway activity in IFNB and STAT1/2 assays, while variant H+207E remained at constant level (IFNB) or showed diminished activity (STAT1/2) (Figure 2A-B). Both variants induced similar STAT3 and NFKB pathway activation upon stimulation (Figure 2C-D). As constitutive activation of STING has been thought to lead to IRF3 phosphorylation and subsequent expression of IFN-response genes (5, 28-30), we next analyzed IRF3 (Ser386) phosphorylation in stably expressing HEK293 cells ( Figure 2E). Interestingly, IRF3 phosphorylation remained at similarly low levels in all studied combinations and their single counterparts, while N154S showed markedly elevated phosphorylation ( Figure 2E). In concordance with our data, IRF3 independent disease development was recently reported in a STING N153S mouse model (31), demonstrating the involvement and activation of multiple pathways in STING-associated disease.
These results further confirm the crucial role of R at position 232 in cGAMP binding and IFN induction (32, 33), and show the constitutive activation of STING signaling pathways together with ligand-dependent hyperactivation of NFKBpathway in G207E mutant cells.

Baseline expression of interferon signature genes and JAK/STAT pathway components are elevated in patients with G207E mutation
To directly analyze baseline gene expression in patient PBMCs, we designed a custom Nanostring (34) gene set targeting multiple IFN-inducible and IFN signaling-related genes, as well as genes encoding components of the NFKB-regulated inflammasome pathway ( Figure 3A). The IFN pathway-related target genes were selected based on their involvement in SLE, SAVI and familial chilblain lupus (5,13,35,36). Strong upregulation of IFN-regulated genes was detected in an unrelated Singleton-Merten-patient with dominant GOF IFIH1 mutation, validating the gene set and analysis pipeline ( Figure 3A). The severely affected G207E patients carrying the IFIH1 risk allele (IV.1, V.2, III.4) displayed higher IFN-regulated mRNA levels (marked in red) than the patient with G207E mutation alone (III.7) ( Figure 3A).
For IFN signature analysis ( Figure 3B) we selected seven genes (IFIT1, ISG15, RSAD2, SIGLEC1, IFI27, IFI44, and IFI44L) based on previous publications on IFN signature in type I interferonopathies (13,36). For fold change calculation, the patient data was normalized to matched control. We noted pronouncedly different gene expression pattern and elevated IFN scores in patients with G207E mutation and IFIH1 risk allele ( Figure 3B). The patient without IFIH1 A946T variant did not show such a drastic response, but scored above the healthy control threshold (36).
The patient cells displayed no clear overall tendency of baseline mRNA levels of inflammasome-related genes, whereas JAK/STAT-pathway components that mediate IFN signaling were upregulated, and selected NFKB-related genes, including the key negative regulator NFKBIA, were downregulated (Figure 3C-E). The trend of additive effects of genetic variance was most pronounced with STING downstream targets JAK1, JAK3, STAT4, STAT6, and IL6, where patient V.2 presented the highest baseline gene expression levels ( Figure 3D-E).
Enhanced type I interferon response and aberrant inflammasome activation in patients with G207E mutation STING has been recently linked with activation of the NLRP3 inflammasome (37, 38) that controls the proteolytic maturation and secretion of cytokines IL1B and IL18. In monocytes the NLRP3 inflammasome can be activated either by the canonical, non-canonical or alternative pathway (39). To test whether inflammasome was hyperactive in the patients, we stimulated their PBMCs with LPS and a subsequent ATP pulse to activate the canonical inflammasome or with prolonged LPS alone to trigger the alternative pathway. The secretion of IL1B and IL18 was used as a proxy for inflammasome activation. Pam3Cys that activates NFKB pathway via Toll-like receptor (TLR) 1/2, but does not normally trigger inflammasome signaling or type I IFN response, was used as a control (40). In the patients' PBMCs LPS induced secretion of mature IL1B and IL18 via the alternative NLRP3 inflammasome pathway (Figure 4A-B) and, importantly, IL1B secretion was aberrantly triggered by Pam3Cys ( Figure 4A).
The activation of alternative NLRP3 inflammasome in monocytes involves TLR4-TRIF signaling that also mediates LPS-induced type I IFN response (40). Thus, we analyzed the expression of IFNB and selected IFN-response genes after LPS treatment in patients' PBMCs by RT-PCR that revealed elevated mRNA levels compared to controls ( Figure 4C). We also detected increases in pro-IL1B mRNA after LPS and Pam3Cys stimulation that may contribute to the elevated IL1B secretion, while mRNA levels of NLRP3 inflammasome components and IL18 did not show consistent alterations (Supplementary figure  2A-B). However, the patients' PBMCs could not release mature IL1B in response to ATP alone without pre-priming with LPS or Pam3Cys to induce pro-IL1B transcription ( Figure 4D). These results indicate that the patient genotype results in aberrant TLR-induced inflammasome responses mainly via increased NLRP3-mediated proteolytic maturation of the inflammasome target cytokines, rather than via increased cytokine mRNA expression.
Since the patient responses to ATP stimulation after short priming were normal ( Figure 4D and Supplementary figure 2C), we conclude that the mutant STING mainly heightens the alternative NLRP3 inflammasome signaling. As the patients also showed elevated TNFA response to LPS stimulation, the dysregulation of TLR responses likely extends beyond the alternative inflammasome pathway ( Figure 4E and Supplementary figure  2B). However, we did not see a general correlation between clinical disease severity and cytokine expression ( Figure 4A-B, E).

Molecular context of mutant STING
To further characterize the molecular context of STING at near physiological levels (41), we analyzed the interaction profiles of H232 (wild-type; wt), I200N (null (42, 43), N154S (5), and H+207E in inducibly expressing HEK293 cell lines using BioID proximity labeling. The BioID-method tags the protein of interest (bait) with a modified biotin ligase, which adds biotin to the proximal proteins (preys) (44). The biotinylated and thus closely interacting proteins are further affinity-purified and analyzed with quantitative mass spectrometry. In line with previous data (33), the wt STING preferentially interacted with Golgi, ER, endosomal and mitochondrial proteins ( Figure 5A). All studied STING mutants (I200N, G207E and N154S) had lost multiple protein-protein interactions compared to the wt STING. By using hierarchical clustering, the interactions of the mutants form a continuation from the null allele (I200N) to GOF SAVI mutant (N154S). The G207E mutation clusters between the wt and the SAVI mutant ( Figure 5A). This result is consistent with the reporter assay data and milder clinical phenotype in G207E carriers (Figure 2).
These observations led us to utilize mass spectrometry microscopy (MS microscopy) (45) on our steady state BioID data of wt and mutant STING. The resolution of MS microscopy outperforms fluorescent microscopy in refining the exact subcellular molecular location (45). MS microscopy showed I200N localization mainly to ER, while G207E, N154S and wt were predominantly localized to Golgi ( Figure 5B). Fascinatingly, the G207E localized more to the ER than wt or N154S (localization scores 0.86, 0.66 and 0.58, respectively) ( Figure 5B). As localization score reflects the subcellular localization and the time spent there, we speculate that the constantly highly active N154S mutant rapidly translocates from the ER to Golgi (lowest ER localization score) leading to hyperactivation of downstream signaling. Meanwhile, the G207E is less active, translocates more slowly, and spends more time in the ER.
To investigate the G207E mutation more closely, we analyzed its interactomes in both allelic backgrounds (H+207E and R+207E). Generally, the G207E mutation led to a reduction of proteinprotein interactions with the proteins residing in Golgi, ER, endosome and mitochondria, compared to the corresponding wt ( Figure 5C). Notably, the G207E mutant had weak interactions with all components of the coatomer complex I (COPI)machinery mediating retrograde transport from Golgi to the ER, and most strikingly with coatomer subunits A, B, B2, D, and E (COPA, COPB, COPB2, COPD, and COPE, respectively) (46, 47) ( Figure 5A, C). Also, multiple components of the coatomer complex II responsible for ER to Golgi transport, such as syntaxin 5 (STX5) and protein transport protein Secs were downregulated (Sec23A, Sec23B, Sec24A, Sec24B) ( Figure 5A, C). The interactions between GOGB1 and UBIA1 were stronger in the H+207E mutant (> 2-fold). UBIA1 localizes to ER-Golgi compartment, while GOGB1 mainly resides at the cis-Golgi participating in COPI-mediated transport (48-50). These data suggest that in addition to amplifying downstream signaling, the GOF mutations cause broad alterations in STING subcellular localization and function in the vesicular system, and reflect a plausible trafficking defect. The features overlap with other known monogenic autoimmune diseases that are caused by defects in the vesicular trafficking (51, 52). The alterations likely contribute to the autoimmune and infectious manifestations seen in the affected patients.
The clinical presentations suggested that the allelic variance at position 232 modifies the disease severity; therefore, we compared the protein interactions between variants R+207E and H+207E. These showed highly similar interac-tomes, with ~2-fold differences in five interactions ( Figure 5D). These genotype-specific interactions included lost association between R+207E and APMAP (adipocyte plasma membrane-associated protein with uncharacterized function), and increased interaction with PDXD1 (pyridoxaldependent decarboxylase domain-containing protein 1 residing at Golgi) and RFA1 (Replication protein A 70 kDa DNA/binding subunit). Slight decreases of interactions were observed with cytoplasmic NSF (vesicle-fusing ATPase) taking part in vesicle-mediated transport within the Golgi and from ER to Golgi, and endosomal SH3G1 (endophilin-A2), predicted to participate in endocytosis and recruitment of proteins to membranes.

DISCUSSION
We report a large family with a distinct somewhat lupoid-like and SAVI-like phenotype carrying a novel G207E mutation in the substrate binding domain of STING. We show that the varying disease severity in this family is due to coding polymorphisms in TMEM173 and IFIH1. The disease retained some typical clinical characteristics of SAVI such as cutaneous vasculitis, severe infections including necrotizing cellulitis, exacerbation of skin symptoms by cold and nasal septum defects (5,11,14). The condition was considerably milder and lacked lung manifestation, violaceus and ulcerative facial and acral skin lesions, elevated inflammatory markers, febrile attacks and autoinflammatory markers. (Supplementary figure 1A-B). The characteristic early-onset multisystemic inflammation was absent, and most patients did not develop life-threatening ulcers, did not require prolonged immunosuppressive medication and have had a normal lifespan. However, our patients presented with several previously unreported STING-associated lupoid-like features including alopecia, photosensitivity, and thyroid dysfunction. Most of them had a marked and widespread livedo reticularis from birth. Furthermore, their IFN score was less severely elevated.
Consistent with previous reports in SAVI patients (5,25), all our patients had expanded population of naïve CD4+ and CD8+ T-cells but low numbers of mature B and T cells. The skewed maturation might be due to the chronic inflammation and immune cell exhaustion. STING has also been proposed to act as an inhibitor of immune cell proliferation (25), an effect linked to NFKB activation and spatially limited to Golgi (25). This finding is supported by our data, which shows hyperactivation of NFKB in the patient cells, and also an altered localization of the G207E mutant within the Golgi/ER compartment.
Since we and others (5,11,(13)(14)(15) have seen considerable variation in disease severity among the TMEM173 mutant carriers, we further analyzed the effect of polymorphisms in TMEM173 (p.R232H) and IFIH1 (p.A946T) on disease phenotype. The patients carrying the R232 variant had more severe disease. In vitro, this polymorphism strengthened the constitutive activation of the mutant STING, leading to the overexpression of STING downstream targets IFN, IL1B, and IL18. The observation is consistent with previously reported effect of R232 on amplifying IFNB pathway activation in vitro (6). We also noted an additive clinical effect for the G207E mutant and the two polymorphisms, with the combination R232+G207E+IFIH1 T/T leading to a severe earlyonset phenotype that resembles the SAVI-features (Pat V.2). IFIH1 risk allele has been shown to promote disease through a combinatorial effect with other risk alleles (24). Although supported by experimental data, the additive effects will need validation in larger patient sets.
Understanding of the crosstalk between STINGpathway and the NLRP3 inflammasome controlling IL1B and IL18 secretion is rapidly expanding, and recently STING was shown to have a role in inflammasome activation (37, 38). We observed aberrant alternative NLRP3 inflammasome activation in response to bacterial TLR stimuli LPS and Pam3Cys in the patient PBMCs. The constitutively active STING-signaling in our patients' PBMCs could explain the elevated LPS-induced IL1B secretion and drive the observed aberrant inflammasome activation in response to Pam3Cys as LPS-TLR-signaling and the following inflammasome activation utilizes an adaptor protein TRIF, that also promotes STING signaling (40, 53). In agreement with our data LPS was reported to aberrantly induce IL1B secretion in TRIF -/monocytic cells artificially expressing a SAVI-associated V155M STING mutant (37). Activation of NLRP3 inflammasome has a key role in the pathogenesis of other autoinflammatory diseases, such as cryopyrinopathies (54). Our data suggest that the STING GOF patients' immune cells have a strong propensity for spontaneous inflammasome activation, provided that suitable inflammasome priming signals are present, and further imply that dysregulation of both interferons and IL1B may contribute to the disease phenotypes of interferonopathies.
To our knowledge, our study is the first to define the molecular interactions of STING. The majority of STING partners were Golgi, ER or endosomal proteins, consistent with reports of subcellular localization of STING (8,33). The monogenic GOF mutants N154S and G207E led to an overall reduction of interactions. The most distinct differences between the classical N154S SAVI and the novel G207E mutation are the lost interactions with COPI and COPII components. These two protein complexes are essential for protein transport between ER and Golgi (55), and loss of COPI also causes monogenic autoimmune disease (52). The increased G207E interactions included Sequestosome-1 (SQSTM), which is an autophagy receptor involved in endosome organization (56). The combined loss of COPI and gained SQSTM interactions together with aberrant subcellular localization, verified with MS microscopy, suggest impairment in mutant STING trafficking. Since STING has been shown to predominantly signal from endosomes, the impaired trafficking likely leads to prolonged signaling and activation of downstream pathways (55, 57). Interaction with AP-1 complex subunit beta-1 (AP1B1), a protein involved in cargo sorting at trans-Golgi network, was lost or diminished in both H+207E and R+207E, further supporting the model of defective transport.
The interactomes of G207E in polymorphic backgrounds (H232 and R232) were highly similar. The R+G207E variant, which is associated with severe clinical disease, further diminished the G207E interactions with the COPI proteins -likely exaggerating the trafficking defect, and further prolonging STING signaling from the endosomes. R+207E variant also associated more with ubiquitin carboxyl-terminal hydrolase CYLD, which regulates inflammation and immune responses via NFKB-inducing pathways (58-60).
To conclude, we describe in a large multigenerational family a novel STING mutation G207E as a cause for a combination of SAVI-associated vasculopathy and selected lupoid-like features. The clinical heterogeneity is explained by TMEM173 polymorphism p.R232H, and IFIH1 risk variant p.A946T, with the two common variants showing cumulative effect on disease severity. We also describe the interactomes of wt and mutant STING, demonstrating a localization defect in the latter, and report elevated NRLP3 inflammasome activation and IL1B secretion in patient PBMCs, potentially contributing to the disease phenotype. The results broaden the clinical spectrum of STING mutation phenotypes, and increase our understanding of the molecular mechanisms of nucleic acid-driven inflammatory diseases.

Patient information
From birth, the index case IV.1 (Figure 1) has had erythema on the cheeks and severe livedo reticularis on his body and extremities, most pronounced on his sides, thighs, upper arms, hips and buttocks. Later, remitting idiopathic trombocytopenia and unspecified nail changes were noted at the age of 7y, as well as alopecia areata. Later, alopecia universalis with sparing of genital, eyebrow and some scalp hair has developed. Repeated bone marrow aspirates and karyotype have been normal. He has nasal septal perforation and slightly dysmorphic features with hypoplastic alae nasi and deep philtrum. He is sensitive to UV radiation and burns extremely easily. He was admitted to the dermatology ward at age 29y due to painful necrotic ulcerations resembling vasculitis on the shin and ankle that had appeared after a sunburn. Skin biopsy confirmed cutaneous vasculitis but extended laboratory tests were normal, except for an elevated Factor VIII (192-211 %), considered an independent risk factor for venous thrombosis ( Figure 1E). His vasculitis was first treated with prednisolone (from 0.8mg/kg and tapered down) combined with azathioprine (150 mg, 1.5 mg/kg) or methotrexate (20 mg/week, 0.2 mg/kg), with poor response. Vasculitis then responded well to cyclosporine (maximum dose 250 mg/day, 2.5 mg/kg). A subsequent sunburn however induced re-flaring of previous ulcers. These subsided with topical corticosteroids. Additionally, he has had Streptococcus pyogenes periobital cellulitis and a single abscess on his inner thigh, but otherwise displays no susceptibility to infections.
IV.1 has three children, of whom daughter V.2 shares a very similar phenotype. Since birth, she has had cheek erythema and severe livedo reticularis of extremities and trunk, most severe on her thighs, upper arms and sides. She has had UV-induced urticaria. At age 6.8y, she developed alopecia areata, which progressed to universal alopecia with loss of all body hair except the eyebrows and eyelashes by age 7.8y. At age 12.5y she was diagnosed with autoimmune thyroiditis (TPOAb 370 IU/mL) and hypothyroidism. She has a nasal septal defect and similar dysmorphic features as her father. At age 3y, she had Streptococcus pyogenes sepsis and necrotizing cellulitis on her upper arm and abdomen. At age 10y, she developed an eschar on her abdomen, which developed rapidly into a large necrotic abscess with Streptococcus pyogenes ( Figure 1C). Surgical revision of the necrotic and infected tissue produced a large wound on the abdomen ( Figure  1D). Skin biopsy from the edge of the necrosis was indicative of small vessel vasculitis and underlying fasciitis. She was successfully treated with i.v. antibiotics, prednisolone (gradually tapered down from 0.5 mg/kg) and cyclosporine 100 mg/d (2.1 mg/kg). Treatment with negative pressure wound therapy successfully assisted wound closure. From birth, the mother of IV.1 (III.4), has had mild livedo reticularis on the buttocks, thighs and upper arms. At age 7y she developed alopecia areata and was diagnosed with hypothyroidism. She burns easily in the sun and has constant facial erythema. III.7 developed livedo reticularis first on the buttocks at age 8y, from where it spread to thighs and upper parts of shins, upper arms, shoulders and stomach, and most recently on her breasts. She also has facial erythema. The livedoid discoloration continuously spreads and increases. She burns easily and had UV intolerance as a child. She had early menopause at age 42y and since then has suffered from severe periodontitis. IV.6 has had mild livedo reticularis form early childhood on her upper arms and knees. She burns easily in the sun. At age 10y she developed total alopecia and, has since lacked all body hair while her scalp hair regrew within a few months. At age 12y, total alopecia reoccurred. She has occasionally also had partial loss of eyelashes. At age 13y, she was diagnosed with autoimmune thyroiditis and hyperthyroidism. Anti-thyroid peroxidase-antibody (TPOAb) titers reached 1855 IU/mL at age 21y. She was then also diagnosed with pituitary thyroid resistance and received radioiodine treatment. Thereafter, she has received thyroid hormone substitution. She has constantly elevated calcitriol and calcium levels with normal parathyroid hormone levels, not indicative of vitamin D resistance. However, after initiating calcium and vitamin D-replacement therapy, she has had minor regrowth of scalp hair. She has recurrent periodontitis and tooth enamel defects. Despite exercise-induced shortness of breath, lung function tests (spirometry and diffusion capacity) were normal.
From adulthood, III.9 has had severe livedo reticularis, which slowly progressed on her upper arms, posterior armpits, sides, lateral back region and buttocks extending to posterior thighs. At age 4y, she experienced total alopecia with subsequent hair regrowth. During her childhood and adolescence, she had recurrent, intermittent spells of alopecia areata. At age 30y, total alopecia with sparsening of eyelashes reoccurred and a vitiligo spot appeared on her cheek. At age 38y, she was diagnosed with autoimmune thyroiditis with TPOAb levels >1000 IU/mL. She has suffered from infertility that is unresponsive to treatment. She has a nasal septal defect, but is not sensitive to UV light. She has had recurrent sinusitis and sleep apnea, while lung function tests were normal.
A mild clinical immunodeficiency in the family was suggested by the severe and persistent periodontitis, recurrent upper respiratory tract and skin infections ( Table 1). Lupoid and autoimmune features included photosensitivity and erosion of the nasal septum (Table 1). However, no antinuclear, anti-DNA or other lupus autoantibodies were detected (Supplementary table 1). Most (5/6) patients had elevated serum IgE levels (Supplementary table 1).

Genotyping and linkage analysis
The DNA samples of 10 family members were run on the IlluminaHumanOmniExpressExome-8v1-2 array at the Science for Life Laboratory SNP&SEQ platform in Uppsala, Sweden, according to standard protocols (Illumina, San Diego, CA, USA) and the results were analyzed using the software GenomeStudio 2011.1 from Illumina Inc. Two controls were run in parallel. Genotyping was based on cluster files generated from the signal intensities from more than 800 DNA samples processed in parallel to this project.
The Rutgers genetic map interpolator Rutgers Map v2 (63) was used to attain genetic map information for the markers in the linkage analysis. Merlin (v1.1.2 (64)) was used to perform parametric and non-parametric linkage analysis, using the LD-pruned genotype dataset. The read mapping, variant calling and genome annotation were performed as described previously (65). We focused the search on the haplotype that segregated with the phenotype, prioritizing rare, heterozygous coding variants that negatively affect conserved residues and shared by both whole genome sequenced individuals. Only one such variant was identified, localized within the TMEM173 gene. The G207E variant was absent from all major public (dbSNP, Exome Aggregation Consortium) and in-house genome databases.

Targeted TMEM173 PCR amplicon sequencing
Sample preparation was performed according to an in-house targeted PCR amplification protocol. grafting P5/P7 sequence and the reaction mix was brought to a final volume with laboratory grade water. Samples were cycled as follows: initial denaturation at 98 °C for 2 minutes; 10 cycles at 98 °C for 20 seconds, at 65 °C for 30 seconds and at 72 °C for 30 seconds; final extension at 72 °C for 5 minutes, followed by a hold at 10 °C.
Following PCR amplification, samples were pooled in equal volumes. The sample pool was purified twice with Agencourt® AMPure® XP (Beckman Coulter, Brea, CA, USA) according to the manufacturer's instructions using equal volumes of the Agencourt® AMPure® XP and the sample pool. The purified sample pool was analyzed on Agilent 2100 Bioanalyzer using Agilent High Sensitivity DNA Kit (Agilent Technologies Inc., Santa Clara, CA, USA) to quantify amplification performance and yield.
Sequencing of PCR amplicons was performed using Illumina MiSeq instrument with MiSeq Control Software v2.5 (Illumina, Inc., San Diego, CA, USA). Samples were sequenced as 251 bp paired-end reads and two 8 bp index reads. The read mapping, variant calling and genome annotation were performed as described previously (65). Evaluation of T cell responses is described in detail elsewhere (66). For the assessment of T cell activation, fresh mononuclear cells were stimulated for 6 hours with anti-CD3, anti-CD28, and anti-CD49d (BD Biosciences, Franklin Lakes, NJ, USA). The cells were analyzed using a 4-or 6-color flow cytometry panel with mAbs against the antigens CD45, CD3, CD4, CD8, CD16, CD56, CD45, CD45RA, TCR-γ, CCR7, IFN-γ, and tumor necrosis factor (TNF).

Site-directed mutagenesis predictions
The 3D structures of STING (4LOI and 4KSY; (32, 67)) were extracted from Protein Data Bank (PDB; www.rcsb.org, (68)) and the effects on the protein stability of the individual mutants analyzed with SDM (26).

Construct generation
The TMEM173 cDNA was obtained as a gateway compatible entry-clone from human Orfeome collection (Horizon Discovery, UK). This entry-clone has the p.His232, which is the minor allele in population. All TMEM173 variant constructs were created using Q5 Site-directed Mutagenesis Kit (New England Biolabs, Ipswich, MA, USA). The mutagenesis primers are listed in Supplementary table 2. All the generated variant entry-constructs were confirmed with direct sequencing prior to introduction to C-MAC-Tag destination vector (45).

Nanostring analysis of patient PBMCs
Blood samples were collected from patients (V.2, IV.1, III.4, III.7) and matched controls in Vacutainer® CPT™-cell preparation tubes with sodium heparin N (BD, New Jersey, USA) and PBMCs were isolated as instructed by manufacturer. Approximately 10 million PBMCs for each patient were pelleted, snap frozen and stored at -80 °C until RNA extraction. RNA was extracted with the RNeasy Mini Kit (Qiagen, Hilden, Germany) according to manufacturer's protocol and concentration measured with a Qbit fluorometer (Thermo Fisher Scientific, WA, USA). For each sample, 100 ng RNA in 5 µl volume was pipetted into the NanoString nCounter gene expression platform (NanoString Technologies, Seattle, USA). Our custom gene set containing 50 genes includes IFN-regulated, inflammasomerelated, as well as JAK/STAT and NFKB signaling pathway genes. For overnight hybridization at 65 °C, RNAs were mixed with 5' reporter probes (tagged with fluorescent barcodes of targeted genes) and 3' biotinylated capture probes. Reactions were purified and immobilized on the sample cartidge surface by utilizing the Prep Station (NanoString Technologies), and the cartridge was scanned in triplicate using an nCounter Digital Analyzer (NanoString Technologies). Analysis of gene expression data was performed with the nSolver™ 4.0 analysis software (NanoString Technologies). As recommended by the manufacturer, we performed two-step normalization. First, background thresholding using negative control counts setting threshold to mean +2 standard deviations above the mean was used. Then, second, to remove technical variability we utilized positive control and CodeSet content (housekeeping gene) two-step normalization. Both normalizations were performed using default settings unless otherwise stated. The normalized values were utilized to calculate interferon scores. Interferon scores were calculated similarly to (36). Briefly, the fold change for each individual gene between patient and matched control was calculated, and the median fold change of all was used to calculate the interferon score for each patient.

Measurement of cytokine secretion
TNFA and the mature, cleaved forms of IL1B and IL18 were detected from PBMC culture media supernatants using Human TNFA DuoSet ELISA, Human IL1B/IL1F2 DuoSet ELISA, and Human Total IL18 DuoSet ELISA (all from R&D Systems).

Quantitative real-time PCR
PBMC RNA was isolated using the RNeasy Plus Mini Kit (Qiagen), followed by cDNA synthesis with iScript kit (Bio-Rad). Quantitative real-time PCR was run in duplicate reactions of 10 ng cDNA using LightCycler480 SYBR Green I master (Roche Diagnostics, Espoo, Finland) and a LightCycler96 instrument (Roche). See Supplementary table 2 for primer sequences. Relative gene expression was calculated using the 2 (-ΔΔCt) method. The data was normalized against geometric mean of expression of two housekeeping genes (ribosomal protein lateral stalk subunit P0 and beta-2-microglobulin).

Cell culture
To generate stable, isogenic, inducible cell lines Flp-In™ T-REx™ HEK293 cells (Invitrogen, Life Tech-nologies) were transfected with the generated STING-MAC-tagged constructs using FuGENE6 (Promega) as described earlier (69). Cells were cultured as instructed by manufacturer.
Five 14 cm plates for each construct and biological replicate were induced either with 1 µg/ ml tetracycline and 50 µM biotin 24 hours prior to harvesting. Harvesting, cell lysis, affinity-purification, mass spectrometry were performed as described (45, 69).

Mass spectrometry data analysis and filtering
The mass spectrometric data was searched with Proteome Discover 1.4 (Thermo Fischer Scientific) using the SEQUEST search engine against the Uni-ProtKB/SwissProt human proteome (http://www. uniprot.org/, version 2015-09). Search parameters were set as in (45). All data were filtered to highconfidence peptides according to Proteome Discoverer FDR 1%. The lists of identified proteins were conventionally filtered to remove proteins that were recognized with less than two peptides and two PSMs. The high-confidence interactors were identified using SAINT and CRAPome as in (45). Each sample's abundancy was normalized to its bait abundancy. These bait-normalized values were used for data comparison and visualization.

Mass spectrometry microscopy (MS microscopy)
The bait-normalized values from BioID data obtained from previous step were used as input for MS microscopy analysis (http://www.biocenter. helsinki.fi/bi/protein/msmic) (45). MS microscopy calculates a localization score that numerically describes the bait protein's dynamic localization in cell.

Statistics
Statistical analysis using 2-tailed Student's t-test was performed using Microsoft Excel. A p-value less than 0.05 was considered significant.