OTULIN deficiency in ORAS causes cell type‐specific LUBAC degradation, dysregulated TNF signalling and cell death

Abstract The deubiquitinase OTULIN removes methionine‐1 (M1)‐linked polyubiquitin signals conjugated by the linear ubiquitin chain assembly complex (LUBAC) and is critical for preventing TNF‐driven inflammation in OTULIN‐related autoinflammatory syndrome (ORAS). Five ORAS patients have been reported, but how dysregulated M1‐linked polyubiquitin signalling causes their symptoms is unclear. Here, we report a new case of ORAS in which an OTULIN‐Gly281Arg mutation leads to reduced activity and stability in vitro and in cells. In contrast to OTULIN‐deficient monocytes, in which TNF signalling and NF‐κB activation are increased, loss of OTULIN in patient‐derived fibroblasts leads to a reduction in LUBAC levels and an impaired response to TNF. Interestingly, both patient‐derived fibroblasts and OTULIN‐deficient monocytes are sensitised to certain types of TNF‐induced death, and apoptotic cells are evident in ORAS patient skin lesions. Remarkably, haematopoietic stem cell transplantation leads to complete resolution of inflammatory symptoms, including fevers, panniculitis and diarrhoea. Therefore, haematopoietic cells are necessary for clinical manifestation of ORAS. Together, our data suggest that ORAS pathogenesis involves hyper‐inflammatory immune cells and TNF‐induced death of both leukocytes and non‐haematopoietic cells.


Introduction
Timely activation and resolution of the innate immune response is essential for tissue homeostasis and host defence, and dysregulation of this response may cause autoinflammation (Nathan & Ding, 2010). Autoinflammatory diseases are a heterogeneous group of disorders characterised by spontaneous development of episodic or chronic inflammation in the absence of inflammatory provocation. Causative mutations have been identified in more than 25 genes, most of which regulate cytokine signalling, e.g. interleukin-1 (IL-1), type I interferon, or tumour necrosis factor (TNF) signalling, or activation of nuclear factor-jB (NF-jB) transcription factors (Manthiram et al, 2017).
Functionally, formation of RSCs and activation of the Ub-and kinase-dependent signalling cascades leads to activation of NF-jB transcription factors and pro-inflammatory transcriptional programmes (Jiang & Chen, 2012). However, for TNF-R1, it can also lead to cell death via formation of a TRADD-, FADD-, RIPK1/3-and caspase-8-containing death-inducing complex, complex-II (Fuchs & Steller, 2015;Kupka et al, 2016b). Recent research has shown that M1-linked Ub chains are key in determining the fate of a cell upon TNF stimulation. In the absence of LUBAC, TNF-mediated signalling is shifted from NF-jB activation towards induction of cell death via complex-II (Haas et al, 2009;Tokunaga et al, 2009;Gerlach et al, 2011;Ikeda et al, 2011;Peltzer et al, 2014Peltzer et al, , 2018. Mouse experiments have shown that the resulting TNF-induced cell death can be highly inflammatory as genetic ablation of LUBAC components leads to multi-organ inflammation (HogenEsch et al, 1993;Seymour et al, 2007;Gerlach et al, 2011;Ikeda et al, 2011;Tokunaga et al, 2011), in large part caused by TNF-induced cell death (Kumari et al, 2014;Rickard et al, 2014). Importantly, homozygous mutations in the LUBAC components HOIP and HOIL-1 have been identified in patients suffering from chronic autoinflammatory syndromes (Boisson et al, 2012(Boisson et al, , 2015. Deubiquitinases (DUBs) are proteases that cleave Ub modifications and functionally revert Ub signalling. Several DUBs, including A20 and CYLD, act as negative regulators of NF-jB signalling to balance activation and ensure resolution of inflammation (Harhaj & Dixit, 2012). OTULIN (OTU family deubiquitinase with linear Ub specificity, also known as FAM105B or Gumby) is the only DUB known to specifically cleave M1 linkages Rivkin et al, 2013). Several studies have shown that OTULIN binds directly to the LUBAC subunit HOIP (Elliott et al, 2014;Schaeffer et al, 2014;Takiuchi et al, 2014). In the absence of OTULIN, M1linked Ub chains accumulate in cells and receptor signalling complexes, enhancing NF-jB activation in response to stimulation of TNF-R1 and NOD2 Keusekotten et al, 2013;Rivkin et al, 2013;Draber et al, 2015;Damgaard et al, 2016;Hrdinka et al, 2016;van Wijk et al, 2017). Interestingly, similar to LUBAC components, homozygous mutations in OTULIN were recently found to cause autoinflammation in humans Zhou et al, 2016), showing that excessive M1-linked Ub signalling can be harmful. OTULIN-related autoinflammatory syndrome (ORAS) (also known as otulipenia or autoinflammation, panniculitis and dermatosis syndrome (AIPDS); OMIM ID: 617099) is a potentially fatal, TNF-driven autoinflammatory disease characterised by sterile systemic inflammation, recurrent high fevers, panniculitis, diarrhoea, arthritis and general failure to thrive Zhou et al, 2016). Key features of ORAS are recapitulated in OTULIN-deficient mice . However, the cellular defects underlying the development of inflammation in ORAS remain unclear.
Here, we report a new case of ORAS identified by whole-exome sequencing of a patient with neonatal-onset, severe inflammatory symptoms. Primary fibroblast cultures from this patient reveal a surprising phenotype of deregulated M1-linked Ub signalling, whereby OTULIN loss coincides with LUBAC downregulation to prevent intrinsic NF-jB activation. In contrast, OTULIN-deficient myeloid cells are hyper-inflammatory and spontaneously activate NF-jB and secrete TNF. Both cell types, however, are sensitised to cell death in response to TNF. Importantly, we find that haematopoietic stem cell transplantation (HSCT) leads to complete resolution of all inflammatory symptoms associated with ORAS. This shows that haematopoietic cells are crucial for the clinical manifestation of ORAS and indicates HSCT could be are curative treatment for the disease.

Results
Inflammatory symptoms in a patient with homozygous OTULIN mutations A female patient of Arab origin (patient III.2), the second of three children born to first-degree related parents (her grandfathers are identical twins; Fig 1A), developed severe inflammatory symptoms shortly after birth. From the age of 3 days, she developed severe idiopathic, systemic inflammation and had recurrent episodes of high fever in combination with widespread panniculitis (Fig 1B and Appendix Clinical Description). At the age of 7 months, her symptoms included high fevers, diarrhoea and panniculitis, and she was cachectic, weighing 3.4 kg (< 3 rd percentile; WHO Multicentre Growth Reference Study Group, 2006) and had severe splenomegaly and bilateral cataracts. Laboratory evaluation revealed elevated acute phase proteins, including C-reactive protein (CRP) and ferritin, elevated IL-6 and soluble IL-2 receptor (sIL-2R) in serum, severe anaemia, and leukocytosis with significant monocytosis in the absence of any evidence of infection (Fig 1B and Appendix Clinical Description).
The consanguineous relations in the family ( Fig 1A) indicated that patient III.2 suffered from an autosomal recessive disorder. Genetic testing excluded Mediterranean fever and chronic atypical neutrophilic dermatosis with lipodystrophy and elevated temperature (CANDLE) syndrome caused by mutations in MEFV and PSMB8, respectively. Whole-exome sequencing (WES) was performed to investigate patient III.2's underlying genetic defect. Filtering out all common and heterozygous variants, WES revealed a single, homozygous G-to-A substitution in OTULIN, c.841G>A; p.Gly281Arg, in patient III.2 (Figs 1A and C and Appendix Table S1). The parents of patient III.2 (II.1 and II.2) and her sister (III.1) were heterozygous for the substitution, whereas her brother (III.3) did not carry the mutation (Figs 1A and C). WES revealed no other homozygous or previously annotated pathogenic variants likely to cause the disease phenotype (Appendix Table S1).
Mutations in OTULIN have recently been described to cause ORAS, an autosomal recessive autoinflammatory disease Zhou et al, 2016). Patient III.2's symptoms, as well as the course of disease and effect of treatments (see below, Appendix Clinical Description), resembled those of other ORAS patients Zhou et al, 2016). Hence, patient III.2 was diagnosed with ORAS.
Mutation of Gly281 to Arg reduces OTULIN's activity towards M1-linked Ub by unfolding the substrate-binding site To date, three different mutations in OTULIN have been reported in families with patients affected by ORAS. In one family, a single nucleotide deletion introduces a premature stop codon, Gly174Aspfs*2, leading to deletion of roughly two-thirds of the catalytic OTU domain (Zhou et al, 2016) and thus likely a complete loss of function although the N-terminal fragment may be present in cells. Hypomorphic Leu272Pro and Tyr244Cys missense mutations were identified in two other families Zhou et al, 2016). Leu272 is located in the Ub-binding pocket in the OTU domain and the mutation to Pro impairs M1-linked diUb binding, catalytic efficiency and stability of OTULIN . Tyr244 is not found or predicted to make any contacts to Ub or be involved in catalysis , and the main effect of the Tyr244Cys mutation might be reduced protein stability (Zhou et al, 2016).
Similar to Leu272 and Tyr244, Gly281 is also located in the OTU domain . Gly281 is evolutionarily conserved and located in a loop between helices a11 and a12 (Fig EV1A; Keusekotten et al, 2013). Together with His282 and Thr283, Gly281 forms part of the Ub-binding site and makes direct contacts to the proximal Ub of M1-linked diUb bound across the active site (Figs 2A, centre panel and EV1A; Keusekotten et al, 2013). While Gly281 is in an important region, the effect of the mutation to Arg is difficult to predict. To understand the pathogenicity of the Gly281Arg mutation on OTULIN, we crystallised OTULIN's catalytic domain (OTULINcat, aa 80-352) containing the Gly281Arg ORAS mutation (OTULINcat G281R ) and determined its structure to 1.8 Å resolution (Figs 2A  and Appendix Table S2). Overall, the structure showed that OTULINcat G281R adopted an OTU fold similar to OTULINcat WT  without any major perturbations (Fig 2A). However, comparing the structure of OTULINcat G281R with the previously solved structure of OTULINcat in complex with M1-linked diUb  showed that the Gly281Arg mutation partly unfolds the substrate-binding site (Fig 2A, right panel). The introduced Arg side chain can no longer fit into the narrow pocket between helices a3 and a9 thereby flipping the Gly281-containing loop out of its normal binding groove (Figs 2A and B;Keusekotten et al, 2013). In the mutant protein, the loop has increased flexibility, as represented by the high crystallographic B-factor distribution across the loop compared to other regions within the OTU domain ( Fig EV1B) and also evidenced by lack of interpretable electron density for the side chains of Val280, mutated Arg281 and His282 ( Fig EV1C). The new conformation of the Gly281-containing loop indicated that His282 and Thr283 would clash with M1-linked diUb at the binding site and prevent substrate binding ( Fig 2B).
We probed the binding of M1-linked diUb to catalytically inactive OTULIN C129A or OTULIN C129A/G281R . As indicated from the structure, the Gly281Arg mutation strongly reduced binding of M1-linked diUb to OTULIN and the dissociation constant (K D ) of OTULIN C129A/G281R and M1-linked diUb was 9.3 lM compared with 85 nM in OTULIN C129A (Fig 2C).
The reduced substrate binding by OTULIN G281R was also reflected in catalytic activity. OTULIN G281R was~20-200 times less active towards M1-linked tetraUb compared with OTULIN WT (Fig 2D). At a 5 nM concentration, OTULIN WT cleaved~50% of the M1-linked tetraUb intro tri-, di-and mono-Ub within 5 min and all tetraUb was processed at 15 min. In contrast, between 100 nM and 1 lM of OTULIN G281R was needed to achieve the same catalytic efficiency ( Fig 2D). In addition to reduced Ub binding and activity, the Gly281Arg mutation caused a considerable intrinsic destabilisation of OTULIN. Monitoring protein unfolding over a thermal gradient, we observed that the melting temperature (T m ) was reduced from 56.5°C for OTULIN WT to 50.9°C for OTULIN G281R (Figs 2E and EV1D). This intrinsic loss of stability indicates a defect in folding of the mutant. Such defects can be exacerbated in cells where partially damaged or unfolded proteins trigger protein quality control mechanisms, including protein degradation.

M1-linked Ub accumulation and destabilisation of OTULIN G281R and LUBAC in patient fibroblasts
The cellular defects underlying the development of inflammation in ORAS are unclear. To study how OTULIN mutations impact on nonhaematopoietic cells and how these cells may contribute to the clinical aspects of ORAS, e.g. panniculitis, we established a primary dermal fibroblast culture from patient III.2. Surprisingly, immunoblot analysis showed that OTULIN G281R was virtually undetectable in the patient cells ( Fig 3A), suggesting that the partial unfolding of OTULIN G281R observed in vitro ( Fig 2E) indeed destabilises the protein. The Gly281Arg mutation did not affect detection of OTULIN by the antibodies used in this study, which both recognise OTULIN's N terminus (Fig EV2A), supporting the notion that OTULIN G281R is destabilised in cells. Treatment with the proteasome inhibitor MG132 substantially increased OTULIN G281R levels ( Fig 3B), and OTULIN transcript levels remained similar between healthy control and ORAS fibroblasts (Fig EV2B), strongly indicating that the reduced OTULIN G281R level is caused by proteasomal degradation.
Interestingly, we observed a substantial concomitant destabilisation of the LUBAC components HOIP and SHARPIN, whereas HOIL-1 remained largely stable except for a reduction in the level of a slower-migrating form (HOIL-1L) ( Fig 3A). We have previously reported a similar loss of LUBAC in murine B and T cells , and we also observed a loss of LUBAC in mouse embryonic fibroblasts (MEFs) with induced OTULIN deletion ( Fig EV2C), showing this phenomenon is not restricted to human fibroblasts. Like for murine T and B cells , transcript levels of HOIP (also known as RNF31), HOIL-1 (also known as RBCK1) and SHARPIN were largely unaltered between OTULIN G281R fibroblasts and controls ( Fig EV2B) and inhibition of autophagosomal degradation could not increase LUBAC protein levels ( Fig EV2D). Instead, we found that HOIP was polyubiquitinated in the OTULIN G281R fibroblasts ( Fig EV2E) and that HOIP levels could be rescued almost completely by proteasomal inhibition (Fig 3B), consistent with proteasome-mediated degradation of LUBAC, particularly HOIP, in the absence of OTULIN in the patient fibroblasts.
To assess how the reduction in OTULIN, HOIP and SHARPIN levels in OTULIN G281R fibroblast impacted the Ub system and the cellular composition of Ub chain types, we used a non-selective tandem Ub-binding entity (TUBE; Hjerpe et al, 2009) to purify polyUb chains from cells in a non-linkage-selective manner ( Fig 3C). The polyUb chains purified by the TUBE were analysed by absolute quantification (AQUA) tandem mass spectrometry (MS/ MS; Kirkpatrick et al, 2006; Fig 3C). K6, K48 and K63 linkages combined comprised~80-85% of the cellular polyUb chains and did not change, while K27, K29 and K33 linkages were undetectable or unquantifiable in the OTULIN G281R and control fibroblasts (Figs 3D and EV2F and G). Curiously, we noted a reduction in K11 linkages from~19% in healthy control fibroblasts to~10% in OTULIN G281R cells (Figs 3D and EV2F and G). Immunoblotting showed no apparent overall differences in the total levels of Ub conjugates or free Ub in whole-cell lysates between the patient fibroblasts and healthy control cells ( Fig 3E).
Strikingly, despite the reduction in LUBAC levels, the AQUA-MS/ MS analysis revealed a considerable increase in M1-linked Ub chains from~0.5% in healthy controls to~3.5% in OTULIN G281R fibroblasts (Figs 3D and EV2F and G). This implies that the remaining LUBAC is still active and may function unantagonised. The increase in cellular M1 Ub levels, however, was hardly detectable by immunoblotting, and only a very faint increase in the high molecular weight smear was observed in the OTULIN G281R fibroblasts ( Fig 3E). This indicates that the increase in M1 linkages might be relatively low compared to the levels observed in OTULIN-deficient cells that retain LUBAC expression (Draber et al, 2015;Damgaard et al, 2016;van Wijk et al, 2017). Consistently, OTULIN G281R fibroblasts exhibited no appreciable signs of spontaneous p65/RelA phosphorylation or IjBa degradation (Fig 3F and G). This suggests that, in the absence of OTULIN, LUBAC is downregulated via proteasomal degradation in the patient fibroblasts, potentially to reduce the levels of M1-linked Ub and prevent intrinsic activation of NF-jB signalling.

Signalling from the TNF-RSC is impaired in OTULIN G281R fibroblasts
Decreased OTULIN function increases M1-linked Ub conjugation and NF-jB activation in response to TNF stimulation in LUBACproficient cells Keusekotten et al, 2013;Damgaard et al, 2016;Hrdinka et al, 2016). Yet, it is unknown how ORAS cells with concomitant reduction of HOIP and SHARPIN levels respond to TNF-R1 engagement. We treated the primary OTULIN G281R patient fibroblasts with TNF and analysed NF-jB and MAP kinase activation by immunoblotting. Interestingly, in contrast to other OTULIN-deficient cells or cells with RNAi-mediated knock-down of OTULIN Keusekotten et al, 2013;Damgaard et al, 2016;Hrdinka et al, 2016), TNF signalling was decreased in OTULIN G281R fibroblasts, although not completely abrogated ( Fig 4A-C), similar to the response of LUBAC-deficient cells (Haas et al, 2009;Gerlach et al, 2011;Ikeda et al, 2011;Tokunaga et al, 2011;Peltzer et al, 2014). Stimulation with TNF led to decreased phosphorylation and degradation of IjBa as well as reduced phosphorylation of p65/RelA compared with healthy controls (Fig 4A and C). Similarly, phosphorylation and activation of the MAP kinases p38 and JNK was impaired in the OTULIN G281R fibroblasts ( Fig 4B). This correlated with a substantial reduction in IL-8 secretion from the patient cells in response to TNF (Fig 4D), showing that reduced NF-jB and MAP kinase signalling leads to functional impairment of innate immune signalling in the ORAS fibroblasts in response to TNF.
After all three LUBAC components, HOIP, SHARPIN and HOIL-1, together with high molecular weight M1-linked Ub conjugates, co-precipitated with TNF-R1 from healthy control fibroblasts ( Fig 4E). However, in the OTULIN G281R fibroblasts, recruitment of LUBAC to TNF-R1 was strongly reduced and hardly detectable ( Fig 4E). Interestingly, also HOIL-1, which remains relatively stable in the patient fibroblasts, could hardly be detected at TNF-RSC, further confirming that HOIL-1 and SHARPIN are recruited in a HOIP-dependent manner (Draber et al, 2015). Consistently, the amount of M1-linked Ub conjugates, but not total Ub conjugates, at the TNF-RSC was drastically reduced in the patient cells ( Fig 4E). Collectively, this suggests that the primary effect of OTULIN loss in the patient fibroblasts, namely the degradation of LUBAC components to prevent intrinsic signalling and autoactivation, results in a secondary effect: the failure to mount a sufficient LUBAC-dependent response to TNF stimulation.

NF-jB activation and TNF secretion in human OTULIN-deficient monocytes
Conditional deletion of OTULIN in myeloid cells recapitulates many aspects of ORAS in mice . This is associated with increased NF-jB activation and spontaneous TNF secretion from OTULIN-deficient BMDMs, which, in contrast to T   . The shOTULIN THP-1 cells showed substantial reduction in cellular OTULIN levels, although some OTULIN remained (Fig 5A). Similar to OTULIN-deficient murine BMDMs, LUBAC remained stable in shOTULIN cells and the levels of HOIP, HOIL-1 and SHARPIN were indistinguishable from those of shControl cells ( Fig 5A). Importantly, OTULIN deficiency in THP-1 monocytes led to spontaneous phosphorylation of p65/RelA ( Fig 5A) and drastically increased turnover of IjBa (Fig 5B), indicative of NF-jB activation in shOTULIN cells. Significantly, shOTULIN cells spontaneously secreted TNF in the absence of any stimulus ( Fig 5C). This shows that the shOTULIN cells have increased basal NF-jB activation and spontaneously secrete cytokines, similar to our previous report for OTULIN-deficient BMDMs . As the shOTULIN cells spontaneously secrete TNF, we examined their signalling after TNF stimulation. In response to TNF, the shOTULIN cells also exhibited a stronger NF-jB signalling response ( Fig 5D). Immunoblotting revealed increased IjBa phosphorylation and degradation as well as elevated p65/RelA phosphorylation after TNF stimulation in the shOTULIN cells compared with shControls (Figs 5D and E).
This shows that also human OTULIN-deficient myeloid cells have a TNF-associated hyper-inflammatory phenotype and indicates that myeloid cells may be fundamentally different from fibroblasts in their response to loss or reduction of OTULIN function.

OTULIN deficiency sensitises fibroblasts and monocytes to TNF-induced cell death
Dysregulation of TNF-induced NF-jB activation in various tissues and cell types can result in inflammation due to increased cell death (Pasparakis, 2009) Boisson et al, 2012Boisson et al, , 2015, leading to TNF-R1-dependent induction of (mainly) TRADD-, FADD-and caspase-8-dependent apoptosis (Gerlach et al, 2011;Ikeda et al, 2011;Peltzer et al, 2014Peltzer et al, , 2018, which is the main driver of inflammation (Kumari et al, 2014;Rickard et al, 2014). Additionally, knock-in mice expressing catalytically inactive OTULIN C129A die mid-gestation (~E10.5) due to aberrant cell death (Heger et al, 2018), similar to HOIP-or HOIL-1-deficient mice . We tested whether OTULIN G281R ORAS fibroblasts were sensitised to TNF-induced cell death. Interestingly, we found that OTULIN G281R fibroblasts, despite their reduction in LUBAC levels, were not sensitive to cell death induced by TNF (Fig 6A) or a range of other stimuli (Fig EV3A). This is in contrast to LUBAC-deficient cells (Gerlach et al, 2011;Ikeda et al, 2011;Peltzer et al, 2014Peltzer et al, , 2018 or cells expressing inactive OTULIN C129A (Heger et al, 2018), which are all sensitive to cell death induced by TNF alone. We found, however, that the OTULIN G281R fibroblasts were sensitive to cell death induced by treatment with TNF in combination with the protein synthesis inhibitor cycloheximide (CHX; Fig 6A), suggesting that under certain stress conditions the patient fibroblasts are sensitised to TNF-induced cell death. TNF-induced cell death may occur either via caspase-dependent apoptosis or RIPK1 kinase-dependent necroptosis (Fuchs & Steller, 2015). Treatment of the OTULIN G281R fibroblasts with the general caspase inhibitor Q-VD-OPh or the RIPK1 kinase inhibitor necrostatin-1 (Nec-1) showed that caspase inhibition rescued the majority of the TNF+CHX-induced cell death in the ORAS fibroblasts ( Fig 6B). Nec-1 treatment alone only conferred mild protection against cell death, but treatment of ORAS fibroblasts with a combination of Q-VD-OPh and Nec-1 led to an almost complete rescue (Fig 6B). This shows that the OTULIN G281R fibroblasts likely die by apoptosis and to a lesser extent by necroptosis, consistent with the mode of TNF-induced cell death observed in LUBAC-deficient cells in culture and in SHARPIN-deficient cpdm mice (Gerlach et al, 2011;Ikeda et al, 2011;Berger et al, 2014;Kumari et al, 2014;Rickard et al, 2014).
Interestingly, despite their normal LUBAC levels, the shOTULIN THP-1 cells were also sensitised to TNF+CHX-induced cell death, but not to cell death by TNF alone (Figs 6C and EV3B), similar to the OTULIN G281R fibroblasts. The cell death in shOTULIN cells could completely be prevented by treatment with Q-VD-Oph, and Nec-1 treatment had no significant effect, neither alone nor in combination with caspase inhibition (Fig 6D). This is consistent with a recent report showing that LUBAC, in addition to its pro-survival role at TNF-R1, also contributes to TNF-induced apoptosis by inducing cFLIP degradation (Tang et al, 2018). Inactivation of OTULIN by expression of OTULIN C129A in adult mice has recently been reported to cause systemic inflammation, which can be prevented by combined ablation of cell death pathways by deletion of caspase-8, RIPK3 and RIPK1 (Heger et al, 2018). Together with the increased cell death we observe in vitro, this suggests that cell death may contribute to ORAS pathogenesis. We analysed skin biopsies from ORAS patient III.2, taken at an inflammatory flare, and healthy control subjects for markers of cell death. Skin from the ORAS patient showed evidence of prominent inflammation with diffuse neutrophil infiltration, particularly around adipocytes, in the dermis and subcutaneous tissues, consistent with panniculitis ( Fig 6E and Appendix Clinical Description).   Figure 6. OTULIN deficiency sensitises cells to TNF-induced apoptosis.
A-D Viability of healthy control or OTULIN G281R primary fibroblasts (A, B) and shControl and shOTULIN THP-1 cells (C, D) after 24 h and 6 h, respectively, of treatment with TNF (100 ng/ml), CHX (50 lg/ml), Q-VD-OPh (10 lM) and Nec-1 (10 lM) as indicated was analysed using MTT reduction assays. Results were normalised to untreated samples. Each experiment was performed in duplicate. Bars represent mean AE SEM (A, n = 9; B, n = 6; C, n = 8; D, n = 4) and were analysed using the two-way ANOVA test of statistical significance with Sidak's correction for multiple comparisons. E Increased apoptotic cell death in the skin of ORAS patient III.2. Serial sections of normal skin (top panels) and a skin biopsy from patient III.2 taken at an inflammatory flare (bottom panels) were immunostained for cleaved caspase-3 (centre panels) or analysed by terminal deoxynucleotidyl transferase dUTP nick end labelling (TUNEL) assay (right panels). Arrowheads indicate cleaved caspase-3-positive mesenchymal cells. None of these markers were present in the healthy control skin. H&E, haematoxylin and eosin. Boxes in H&E panels indicate the areas magnified in the cleaved caspase-3 and TUNEL panels. Scale bars, 100 lm. Data are representative of stained sections from three healthy controls and one biopsy from ORAS patient III.2. This inflammation correlated with the presence of cleaved caspase-3-positive and TUNEL-positive cells in the dermis and subcutaneous tissues (Fig 6E), showing that cell death occurs in vivo during ORAS inflammation. Intriguingly, cleaved caspase-3-positive cells appeared to be mainly mesenchymal cells, including fibroblasts ( Fig 6E, centre panels, arrowheads), whereas TUNEL-positive cells appeared to be mainly infiltrating leukocytes (Fig 6E, bottom panels). These data show that cell death (including apoptosis) occurs in ORAS and that it may contribute to the TNF-driven pathogenesis of the disease.

Haematopoietic stem cell transplantation ameliorates ORAS
Previously, we have shown that deletion of OTULIN in all haematopoietic cell types in mice, or in myeloid cells alone, is sufficient to recapitulate key features of ORAS . But whether haematopoietic cells are necessary or sufficient to drive ORAS in humans is unknown. Importantly, we could use the case of ORAS in patient III.2 to address this retrospectively. As outlined, patient III.2 developed severe inflammatory symptoms within the first months of her life and she had a poor response to conventional anti-inflammatory treatment (Appendix Clinical Description). From the age of 8 months, she was treated with prednisone and colchicine, to which she only had a partial response (Fig 7A), similar to previous reports from ORAS patients Zhou et al, 2016). At the age of 11 months, the recombinant IL-1 receptor antagonist anakinra was added to the treatment regimen without any improvement of her symptoms. Because of this poor response to treatment and the impaired pre-transplant performance status of the patient, HSCT was performed at the age of 17 months (Fig 7A) with reduced intensity fludarabine-based conditioning that included fludarabine (30 mg/m 2 /day for 6 days), busulphan (3.2 mg/kg/day for 4 days) and thiotepa (10 mg/kg/day for 1 day) together with antithymocyte globulin (ATG) (10 mg/kg). The donor was the patient's HLA-matched father. The course of transplantation was uneventful, and 100% donor chimerism was achieved 17 days after HSCT (Fig 7B and Appendix Clinical Description). Remarkably, HSCT resulted in complete resolution of all inflammatory symptoms, including panniculitis, arthritis and diarrhoea, and the patient experienced complete clinical remission (Fig 7A and Appendix Clinical Description). As we and others have found signalling and cell death defects in OTULIN-deficient non-haematopoietic cells (Figs 3 and 4; Heger et al, 2018), complete remission after HSCT strikingly shows that OTULIN deficiency in haematopoietic cells is necessary for clinical manifestation of ORAS including the organ-specific symptoms panniculitis and diarrhoea (Fig 7D).
Nine months after HSCT, at the age of 26 months, patient III.2 relapsed with fevers, panniculitis and arthritis, albeit with less severe symptoms than before the transplantation (Fig 7A and C, and Appendix Clinical Description). The relapse correlated with a progressive decrease in bone marrow chimerism and regrowth of patient III.2's haematopoietic cells (Fig 7B). In mice, bone marrow chimerism with 50% OTULIN-deficient cells is sufficient to cause severe systemic ORAS-like inflammation , indicating that the decrease in chimerism observed in patient III.2 ( Fig 7B) is sufficient to cause the relapse.
After the relapse, treatment with prednisone was resumed until the age of 29 months where patient III.2 started treatment with the TNF-blocking agent etanercept (soluble TNF receptor fusion protein; 0.4 mg/kg, twice weekly; Fig 7A and Appendix Clinical Description). This led to immediate and complete resolution of her inflammatory symptoms, and corticosteroid treatment was discontinued. Now, at age~12 years, patient III.2 is treated with low dose etanercept (0.8 mg/kg) every 10 days and she remains asymptomatic as long as administration is regular (Fig 7C and Appendix Clinical Description).
Collectively, these data show that haematopoietic cells are required for clinical manifestation of ORAS and confirm that TNF is the driver cytokine for all inflammatory aspects of the disease , adding ORAS to the growing list of TNF-related pathologies (Kalliolias & Ivashkiv, 2016;Manthiram et al, 2017). Intriguingly, the dependence on OTULIN-deficient haematopoietic cells for ORAS implies that myeloid cells are the main drivers of the TNF-mediated inflammation (Figs 5 and 7D), as OTULIN deficiency in B or T cells alone in mice does not produce any inflammatory phenotypes . Moreover, it indicates that cell death in ORAS might be a key contributing and amplifying result of TNF-mediated inflammation (Figs 6E and 7D).

Discussion
Here, we report a new case of ORAS caused by homozygous hypomorphic c.841G>A; p.Gly281Arg mutations in a patient from a consanguineous family. To date, only three families (five patients) with OTULIN mutations and ORAS manifestations had been reported Zhou et al, 2016). Our report of a new case of ORAS caused by a novel Gly281Arg mutation adds a new family and patient that expands our understanding of the pathogenesis of the disease, its clinical aspects, as well as its treatment.
Correct regulation of M1-linked Ub conjugation by LUBAC and hydrolysis by OTULIN is critical to maintain immunehomeostasis in humans and mice. This is evident from clinical genetic studies (Boisson et al, 2012(Boisson et al, , 2015Damgaard et al, 2016;Zhou et al, 2016) and numerous mouse studies (reviewed in Hrdinka & Gyrd-Hansen, 2017). Interestingly, the clinical result of mutations in HOIL-1 or HOIP (i.e. reduced M1-linked Ub conjugation) or in OTULIN (i.e. reduced M1-linked Ub hydrolysis) is the same: inflammation. Clinically, HOIL-1 and OTULIN deficiency manifests similarly with neonatal-onset systemic inflammation, recurrent fevers and failure to thrive (Boisson et al, 2012;Damgaard et al, 2016;Zhou et al, 2016). However, some important differences between these diseases exist. HOIL-1 deficiency is associated with severe primary immunodeficiency, cardiomyopathy and muscle weakness (Boisson et al, 2012;Nilsson et al, 2013;Wang et al, 2013), of which there is no evidence in ORAS patients Zhou et al, 2016). In addition, TNF blockade is highly effective in all ORAS patients reported. Standard immunosuppression (e.g. corticosteroids) is, on the other hand, generally ineffective Zhou et al, 2016). In contrast, corticosteroid treatment is generally effective in HOIL-1 deficiency whereas anti-TNF treatment in these patients is not (Boisson et al, 2012). This suggests that the underlying defects causing these syndromes, while overlapping, may be different.
OTULIN-related autoinflammatory syndrome patients respond poorly to standard anti-inflammatory and immunosuppressive treatments Zhou et al, 2016)  We provide here the first report of the outcome of HSCT for ORAS. Remarkably, HSCT led to complete remission of the patient's severe ORAS symptoms, including panniculitis, arthritis and diarrhoea, showing that haematopoietic cells are crucial for manifestation of the disease. The patient's unfortunate relapse appears to be due to regrowth of her haematopoietic cells (Fig 7B). The decreased donor chimerism is most likely a consequence of the reduced intensity pretransplantation conditioning, which may have enabled proliferation of residual haematopoietic stem cells in the patient, leading to a drop in donor chimerism and a relapse of inflammatory symptoms. We therefore suggest the use of more intense, fully myeloablative conditioning regimes as preparation for HSCT if such regimes can be tolerated by the recipients. Despite the patient's relapse, this case supports further studies into HSCT as a potential treatment to ameliorate ORAS. The Gly281Arg mutation identified in this study has a drastic impact on OTULIN. Not only does it reduce binding of and catalytic activity towards M1-linked Ub chains, but it also destabilises the protein such that OTULIN G281R is hardly detectable in the absence of proteasome inhibitor in patient fibroblasts. Destabilisation of OTULIN by ORAS mutations has been described for two other point mutations, Leu272Pro and Tyr244Cys, as well as the premature stop mutant Gly174Aspfs*2 Zhou et al, 2016). The combination of reduced catalytic activity and very low protein levels of OTULIN G281R in cells indicates that the mutation might cause close to a complete loss of OTULIN function in vivo. The effect of the Gly281Arg mutation on OTULIN is similar in mechanism to the Leu272Pro mutation identified in a different ORAS family Zhou et al, 2016). Collectively, this strongly suggests that ORAS is a disease of loss or significant reduction of OTULIN activity and expression.
In cells, we observed a considerable increase in M1 Ub linkages in the OTULIN G281R patient fibroblasts, despite the drastically reduced HOIP and SHARPIN levels. At steady state, the cellular levels of M1-linked Ub are very low (Figs 3D and EV2F) Keusekotten et al, 2013;Draber et al, 2015), yet they accumulate transiently in receptor complexes to facilitate signalling ( Fig 4E) (Gerlach et al, 2011;Emmerich et al, 2013;Fiil et al, 2013;Keusekotten et al, 2013;Peltzer et al, 2014;Rickard et al, 2014;Draber et al, 2015;Hrdinka et al, 2016;Kupka et al, 2016a). This suggests that M1-linked Ub is a potent and highly regulated signal that is strictly controlled in cells to prevent intrinsic activation or dysregulation of signalling. In the OTULIN G281R fibroblasts, we observed a chronic and global accumulation of M1-linked Ub chains (Figs 3D and EV2F and G) together with a striking concomitant reduction in LUBAC levels and no evidence of spontaneous NF-jB activation (Figs 3A,F and G). This suggests that in certain cell types, e.g. T and B cells  and fibroblasts (Figs 3A and EV2C), the potentially uncontrolled M1-linked Ub accumulation in the absence of OTULIN may function as a "stress" signal that somehow feeds back and reduces LUBAC activity by downregulating the level of this complex via the proteasome, thereby maintaining homeostasis by preventing or reducing intrinsic activation of inflammatory signalling. However, downregulation of LUBAC also means that these cells cannot respond appropriately to external signals, including TNF, which may in turn contribute to ORAS pathogenesis (Fig 7D, see below).
Different cell types respond differently to loss of OTULIN. In murine models, LUBAC is stable in OTULIN-deficient myeloid cells, but not in B and T cells where HOIP and SHARPIN levels are reduced . Reduction in LUBAC levels, primarily in HOIP and SHARPIN, has also been reported in fibroblasts from three different ORAS patients (Zhou et al, 2016). Similarly, we observe here that HOIP and SHARPIN levels are reduced in OTULIN G281R fibroblasts while LUBAC remains stable in OTULIN-deficient THP-1 monocytes. This striking cell type-specific effect on LUBAC suggests that the phenotype of different cell types upon loss of OTULIN may depend on their LUBAC status. LUBAC exists, at least, as a trimeric HOIP-HOIL-1-SHARPIN complex (Gerlach et al, 2011;Ikeda et al, 2011;Tokunaga et al, 2011), but the stoichiometry of components and whether it varies between cell types is unknown. Specific subpopulations of LUBAC, e.g. dimeric HOIP-HOIL-1 or HOIP-SHARPIN complexes or complexes stably associated with other factors such as DUBs, remain a topic of active investigation. Recently, it was reported that OTULIN and CYLD, via its adaptor SPATA2, utilise the same mechanism to interact with HOIP, thereby forming mutually exclusive complexes (Draber et al, 2015;Elliott et al, 2016). This suggests that distinct pools of LUBAC-DUB complexes may exist (Hrdinka & Gyrd-Hansen, 2017). The apparently specific downregulation of HOIP and SHARPIN in ORAS patient fibroblasts (Fig 3A;Zhou et al, 2016) and in T and B cells  might suggest that OTULIN could be a part of and/or regulate a distinct HOIP-SHARPIN complex. However, this needs further investigation. Our data suggest that careful assessment of steady state levels of OTULIN and LUBAC, and possibly CYLD, is necessary for appropriate interpretation of experiments involving OTULIN, LUBAC and M1-linked Ub signalling.
OTULIN G281R fibroblasts are functionally LUBAC-deficient after TNF stimulation. LUBAC recruitment and M1-linked Ub conjugation at the TNF-RSC are reduced, accompanied by impaired TNF signalling, decreased IL-8 secretion and sensitisation to TNFinduced cell death in the presence of CHX (Fig 4). Interestingly, OTULIN-deficient THP-1 cells, which are hyper-inflammatory and hyper-signal in response to TNF (Fig 6), are also sensitive to this form of cell death, despite their normal levels of LUBAC. These observations suggest that ORAS may not only be caused by TNFinduced hyper-signalling and cytokine secretion , but that ORAS pathogenesis may also involve cell death ( Fig 7D). Our analysis of cell death in biopsies from ORAS patient III.2 supports this notion and provides clear evidence of apoptosis in the skin of this patient at the time of an inflammatory flare. This bears resemblance to the dermatitis phenotypes in LUBACdeficient mice and the cell death observed in LUBAC-deficient cells (Kumari et al, 2014;Peltzer et al, 2014Peltzer et al, , 2018Rickard et al, 2014;Taraborrelli et al, 2018). However, in contrast to LUBAC deficiency where cells are sensitised to cell death induced by TNF alone, OTULIN-deficient cells need additional perturbation (i.e. CHX) before they become sensitive to TNF-induced death. This could suggest that cell death of OTULIN-deficient cells may be a secondary effect of a "stressed" environment, such as an inflamed tissue. The dermatitis in SHARPIN-deficient cpdm mice is caused by cell or tissue intrinsic defects leading to TNF-induced cell death of keratinocytes (Kumari et al, 2014;Rickard et al, 2014) and it is not mediated by immune cells as allogeneic BM or splenocyte transplantations fail to transfer the disease to WT mice (Hogen-Esch et al, 1993;Rickard et al, 2014). In contrast, OTULIN deficiency in haematopoietic cells is necessary for manifestation of skin inflammation/panniculitis in ORAS patients as demonstrated here by HSCT (Fig 7A-C and Appendix Clinical Description). Yet, mouse models indicate that OTULIN deficiency in haematopoietic cells is not sufficient for development of the full spectrum of ORAS symptoms. Mice with genetic ablation of OTULIN either in all haematopoietic cell types or specifically in myeloid cells develop systemic ORAS-like inflammation, but the organ-specific symptoms that are hallmarks of the human disease (panniculitis, diarrhoea and arthritis) do not manifest spontaneously in these models . The full clinical disease in ORAS may therefore be a "compound" phenotype where hyper-inflammatory, TNF-secreting haematopoietic (myeloid) cells cause TNF-induced inflammation and cell death of both non-haematopoietic cells with LUBAC downregulation, e.g. dermal fibroblasts, and haematopoietic cells (Fig 7D), thereby giving rise the organ-specific hallmark symptoms panniculitis, diarrhoea and arthritis.
Together, these clinical and experimental data expand our understanding of the cellular mechanisms leading to ORAS development and indicate HSCT as a new potential treatment to ameliorate ORAS.

Whole-exome analysis
Exonic sequences from a DNA sample from patient III.2 were enriched using SureSelect Human All Exon 50 Mb V.5 Kit (Agilent Technologies, Santa Clara, CA) and sequenced on an Illumina HiSeq2500 (Illumina, San Diego, CA) with 100-bp paired-end reads. Data were analysed on the DNAnexus platform (DNAnexus, Inc., Mountain View, CA) using default parameters with the human genome assembly hg19 (GRCh37) as reference. For further details, see Appendix Supplementary Methods.

Expression and purification of recombinant proteins
OTULIN and TUBE proteins were purified from E. coli as described in (Elliott et al, 2014) and , respectively. For further details, see Appendix Supplementary Methods.

Qualitative DUB assay
Qualitative Ub cleavage assays were performed as previously described . For further details, see Appendix Supplementary Methods.

Protein crystallisation and structure determination
OTULINcat G281R crystals were grown by hanging drop vapour diffusion method and the structure solved by molecular replacement using PHASER (McCoy et al, 2007) with wild-type OTULINcat (PDB ID 3ZNV) as the search model. For further details, see Appendix Supplementary Methods.

Binding assays
Fluorescence anisotropy experiments were performed as previously described . For further details, see Appendix Supplementary Methods.

Differential scanning fluorimetry (DSF) thermal unfolding experiments
Nano-DSF measurements were performed as previously described . For further details, see Appendix Supplementary Methods.

Cell culture
A primary fibroblast culture from patient III.2 was established from a skin punch biopsy. Primary fibroblasts were cultured in Ham's F-10 nutrient mix + GlutaMAX TM (Life Technologies) supplemented with 20% iron-supplemented foetal calf serum (FCS) (#C8056, Sigma) (v/v) and penicillin/streptomycin (Life Technologies). Human THP-1 monocytes with stable short-hairpin RNA (shRNA)-mediated knock-down of OTULIN (shOTULIN) or a non-targeting control (shControl) were a kind gift from Dr. Mads Gyrd-Hansen (University of Oxford) . THP-1 cells were cultured undifferentiated in RPMI-1640 + GlutaMAX TM (Life Technologies) supplemented with 1 mM sodium pyruvate (Life Technologies), 10% FCS (Life Technologies) (v/v), 50 lM b-mercaptoethanol and penicillin/streptomycin (Life Technologies). All cells were kept at 37°C in a humidified atmosphere at 5% CO 2 . All cultures were found to be negative for mycoplasma contamination by regular PCR testing (Young et al, 2010).

Cell stimulations and signalling experiments
Primary fibroblasts or THP-1 cells were treated as indicated in figures and legends. For signalling experiments, cells were stimulated with recombinant human TNF (Life Technologies). For proteasomal inhibition, cells were treated with 10 lM MG132 (Sigma) in DMSO as indicated. For inhibition of autophagosomal degradation, cells were treated with 100 nM bafilomycin A1 from Streptomyces griseus (BafA) (Sigma) in DMSO as indicated. Cells were lysed either directly in sample buffer (50 mM Tris pH 6.8, 10% glycerol (v/v), 100 mM DTT, 2% SDS (w/v), bromophenol blue), sonicated using a microtip for 5 s at 30% amplitude on a Vibra-Cell TM VC750 (Sonics & Materials, Inc., Newton, CT) and boiled for 2 min, or lysed in RIPA buffer (50 mM Tris pH 7.4, 1% NP-40 (v/v), 0.5% deoxycholate (w/v), 0.1% SDS (w/v), 150 mM NaCl, 2 mM EDTA, 5 mM MgCl 2 ) for 20 min on ice, lysate cleared by centrifugation and supplemented with 4× sample buffer. For assessment of IjBa stability, THP-1 cells were washed in PBS and seeded in 96-well plates in complete medium~16 h before they were treated with 50 lg/ml cycloheximide (CHX; Santa Cruz Biotechnology, Santa Cruz, CA) in DMSO as indicated. IjBa stability in primary fibroblasts was assayed by seeding the fibroblasts in 12-well plates 18-24 h prior to treatment with CHX as indicated. After CHX treatment, all cells were washed in PBS and lysed in RIPA buffer.

Immunoprecipitation of the TNF-RSC
Immunoprecipitation of the TNF receptor signalling complex (TNF-RSC) was performed using 100 ng/ml FLAG-tagged TNF (Enzo Life Sciences, Farmingdale, NY) as previously described  Purification of endogenous polyUb conjugates by TUBE pull-down Endogenous polyUb conjugates were purified from primary fibroblasts using TUBE affinity reagents as described previously (Damgaard et al, 2012;Fiil et al, 2013). For further details, see Appendix Supplementary Methods.

Ub chain composition analysis (AQUA-MS)
Details on LC-MS/MS methods for analysis of the cellular composition of Ub chain linkages can be found in Appendix Supplementary Methods.

Immunoblotting
Samples were resolved on 4-12% Bis-Tris NuPAGE gels (Life Technologies) and transferred to nitrocellulose or polyvinylidene fluoride (PVDF) membranes. Blots were visualised using Clarity TM Western ECL Substrate (Bio-Rad) on a ChemiDoc TM Touch or ChemiDoc TM MP imager (Bio-Rad). All antibodies are listed in Appendix Table S3. Densitometry analysis of immunoblots was performed using the Fiji software (Schindelin et al, 2012). For further details, see Appendix Supplementary Methods.

ELISA
Cytokine and chemokine concentrations in cell culture medium were measured using the Human IL-8/CXCL8 Quantikine ELISA kit (D8000C) or the Human TNF-alpha Quantikine HS ELISA kit (HSTA00E) from R&D Systems, Minneapolis, MN, according to the manufacturer's instructions. For IL-8 measurements, equal numbers of primary fibroblasts from the OTULIN G281R patient or healthy control were seeded in flat-bottom 96-well cell culture plates 24 h before the experiments. Cells were washed, and fresh culture medium was added immediately before the cells were treated with 10 ng/ml recombinant human TNF (Life Technologies) as indicated. The cell culture medium was collected at the indicated times (for the "0" samples, medium from untreated cells was left on the cells for 24 h and collected at the same time at the 24 h TNF-treated samples) and subjected to ELISA analysis. For TNF measurements, equal numbers of shControl and shOTULIN THP-1 cells were washed in PBS and seeded in fresh medium in round-bottom 96-well plates. The cells were then left untreated, and the culture medium was collected at the indicated times and subjected to ELISA analysis.

Statistics
Data are presented as mean AE standard error of the mean (SEM) or AE standard deviation (SD) as indicated. Sample number (n) indicates the number of independent biological samples in each experiment. Sample numbers and experimental repeats are indicated in figures and figure legends. Data were analysed using the two-way analysis of variance (ANOVA) with Sidak's correction for multiple comparisons or the non-parametric Mann-Whitney U-test of the null hypothesis as indicated. For the two-way ANOVA tests, the Shapiro-Wilk test was used to assess normality of the data distributions. Generally, the datasets were normally distributed and as the variables analysed (cytokine secretion and cell viability) are generally accepted to be normally distributed, the use of the two-way ANOVA test was considered appropriate. Differences in means were considered statistically significant at P < 0.05. Analyses were performed using the GraphPad Prism software version 7.0b.

Study approval
Written informed consent was obtained from all subjects and family members.

Data availability
Due to restrictions from the patient consent approved by the research ethics committees, it is not be possible to deposit complete exome sequencing data in a public repository, but the data could be made available to interested researchers by contacting the corresponding author (Y.B.). Accession numbers, coordinates and structure factors for crystal structures of OTULIN G281R have been deposited within the protein data bank with accession code 6I9C. Mass spectrometry data include global ubiquitin linkage analysis by AQUA and been deposited to the Mass Spectrometry Interactive Virtual Environment (MassIVE) (ftp://massive.ucsd. edu/MSV000083154) at University of California San Diego, CA.
Expanded View for this article is available online.

Results
We identified a new case of ORAS in which a Gly281 to Arg mutation in OTULIN causes the disease. To assess the cellular impact of loss of OTULIN in ORAS, we analysed the inflammatory signalling response in patient-derived fibroblasts and in OTULIN-deficient human monocytes. This revealed that OTULIN loss in patient-derived fibroblasts leads to concomitant reduction of LUBAC and decreased TNF signalling, whereas OTULIN-deficient monocytes retain LUBAC expression and show increased signalling and TNF secretion. However, both cell types are sensitised to TNF-induced cell death. Remarkably, haematopoietic stem cell transplantation leads to complete resolution of all inflammatory symptoms associated with ORAS, showing that haematopoietic cells are necessary for development of ORAS inflammation. This indicates that ORAS pathogenesis involves hyperinflammatory immune cells and TNF-induced death of both immune cells and non-haematopoietic cells.
Impact OTULIN-related autoinflammatory syndrome is a potentially fatal, neonatal-onset autoinflammatory disease of unclear pathogenesis. Treatment with TNF inhibitors such as antibodies (e.g. infliximab) or TNF blockers (e.g. etanercept) effectively ameliorates inflammation, but there is currently no cure for ORAS. We show that haematopoietic cells are necessary for clinical manifestation of ORAS, suggesting that haematopoietic stem cell transplantation is a potentially curative treatment for ORAS.