A variant in IL6ST with a selective IL-11 signaling defect in human and mouse

Tobias Schwerd iD ,Freia Krause iD ,Stephen R.F.Twigg iD ,Dominik Aschenbrenner iD ,Yin-Huai Chen iD ,Uwe Borgmeyer iD ,Miryam Müller iD ,a,Santiago Manrique iD ,Neele Schumacher iD ,Steven A.Wall iD 7,Jonathan Jung iD ,b,Timo Damm iD ,Claus-Christian Glüer iD ,Jürgen Scheller,Stefan Rose-John ,E.Yvonne Jones ,Arian Laurence,Andrew O.M.Wilkie ,7,Dirk Schmidt-Arras and Holm H.Uhlig iD ,0,

INTRODUCTION

GP130 is the common receptor subunit for the family of interleukin(IL)-6 cytokines that includes IL-6,IL-11,IL-27,leukemia inhibitory factor(LIF),oncostatin M(OSM),ciliary neurotrophic factor(CNTF),cardiotrophin 1(CT1),and cardiotrophin-like cytokine(CLC).1To allow speci ficity of cytokine receptor binding in different cell types,GP130 forms a multimeric complex on the cell surface with cytokineselective receptor subunits that are either signaling-incompetent such as IL6RA,IL11RA,CNTFR,or signaling competent such as LIFR,OSMR,or IL-27Rfacilitating downstream activation of cytoplasmic tyrosine kinases which then phosphorylate STAT3 and STAT1 or activate the RAS/MAPKpathway.1

Diverse biologic functions of immune and non-immune cells depend on GP130-mediated signals.For example,bone formation and remodeling require IL-11 signaling through the GP130/IL11RA receptor complex.In mice,the absence of GP130 leads to skeletal abnormalities associated with defects in osteoblast and osteoclast function2–3and osteoblast-speci fic disruption of GP130-STAT3 pathway impairs bone formation.4Similarly,IL-11 receptor knockout(Il11ra-/-)mice show increased trabecular bone volume and synostosis of premaxillary sutures associated with a cellautonomous defect of osteoclast differentiation.5–6

Genetic defects in genes required for IL11RA-dependent STAT3 signaling cause skeletal abnormalities in humans.Skeletal and connective tissue abnormalities are commonly found in patientswith autosomal dominant hyper-IgEsyndrome(HIES)due to heterozygous mutations in STAT3,including craniosynostosis,varying degrees of scoliosis or retained primary teeth.7–13Patients with craniosynostosis and dental anomalies(CRSDA;MIM 614188)were found to carry recessive loss-of-function variants ofIL11RA.5,14–18The clinical disease phenotype is characterized by multi-suture craniosynostosis,maxillary hypoplasia,delayed or ectopic tooth eruption,supernumerary teeth and minor digit abnormalities.Craniosynostosis and delayed tooth eruption observed in individuals withIL11RAmutations,likely result from reduced bone resorption at sutures or in the jaw,as it has been shown that IL11RA-de ficient mice display decreased bone resorption in long bones.6We have recently described patients with severe immunode ficiency and skeletal abnormalities,such as severe craniosynostosis and progressive scoliosis caused by recessive partial loss-of-function variants inIL6ST,encoding GP130(hyper-IgE recurrent infection syndrome 4,autosomal recessive;MIM 618523).19–20Detailed functional studies demonstrated that different homozygous non-synonymous variants had relatively selective effects on the broad range of GP130-dependent cytokines with complete abrogation of IL-6 and IL-11 signals,reduction in OSM and IL-27 signaling but preserved LIF signaling.Similarly,autosomal-dominant variants in the cytoplasmic tail of GP130 cause a hyper-IgErecurrent infection syndrome due to defective IL-6 and IL-11 signaling(Beziat V.,et al.,JEM2020 in press).Complete abrogation of all GP130-dependent cytokine signaling due to biallelic essential loss-of-function variants in GP130 causes an extended Stüve–Wiedemann syndrome with neonatal lethality.21

Here,we describe a patient with a biallelic,non-synonymous variant in GP130 with a highly selective defect limited to IL-11 signal transduction.The patient’s phenotype was restricted to craniosynostosis and tooth abnormalities.We demonstrate that the variant has incomplete penetrance in humans and mice,suggesting a hypomorphic modi fier effect.

RESULTS

IL6STp.R281Q in a patient with craniosynostosis

We recently described loss-of-function variants inIL6STas a novel cause of autosomal recessive HIESwith skeletal abnormalities.19–20In order to understand the impact ofIL6STvariants in patients with skeletal abnormalities,we screened for homozygous or compound heterozygous variants inIL6STin a cohort of 467 unrelated patients with craniosynostosis,who were mutation negative after clinically driven genetic testing.19We identi fied a homozygous variant(c.842G＞A;p.R281Q)in a single patient of South Asian origin,hereafter referred to as PR281Q(Fig.1a–c).This individual presented at the age of 7 years with abnormal head shape associated with sagittal and bilateral lambdoid craniosynostosis and retained deciduous teeth,reportedly requiring the extraction of 14 teeth aged 8 years.Intracranial pressure monitoring was normal and on annual follow-up no clinical progression requiring surgical intervention for the craniosynostosis was observed.There were no infections or immune dysregulation problems up to the age of 18 years.Clinical genetic testing had been negative for the major causes of craniosynostosis,includingERF,FGFR1 exon 7,FGFR2(all exons associated with craniosynostosis),FGFR3exons 7 and 10,IL11RA,TCF12andTWIST1.

The amino acid R281 is conserved throughout evolution from amphibians to mammals(Fig.1d,Supplementary Fig.3a).The mutational impact of the p.R281Q substitution is predicted to be moderate;SIFT tolerated(0.162),PROVEAN neutral(-0.81),Poly-Phen2 probably damaging(0.999),and with a CADD score 9.782.

We initially classi fied this as a variant of unknown signi ficance since:(a)48(updated 20/11/2019)heterozygous individuals(but no homozygotes)are tabulated in gnomAD v3,(b)this variant is enriched in individuals of South Asian origin (minor allele frequency 0.001 5),(c)the mother of PR281Q,herself the offspring of consanguineous parents,was homozygous for the same variant(Fig.1a)but without any history of craniofacial or severe tooth abnormalities,and(d)the phenotype was only partially overlapping with the previously described humans withIL6STdefects(Supplementary Table 1).

p.R281Q causes defective IL-11 signal transduction while

maintaining normal signaling of other IL-6 family cytokines

To fully assess the functional consequences of the p.R281Q substitution,we used a previously described GP130-de ficient HEK293 cell line(HEK293 GP130-KO)generated by CRISPR/Cas9 technology.19This cell line does not phosphorylate STAT1 or STAT3 in response to stimulation with IL-6,IL-11,IL-27,OSM,or LIF,but has normal STAT3 signaling in response to type 1 interferon and normal STAT1 signaling in response to IFN-γ.Transfection with GP130 wild type(WT)restored GP130-dependent signaling(Fig.2).p.R281Q did not confer mRNA or protein instability(data not shown).Titration studies on transfected GP130-KO cells revealed that the p.R281Q substitution signi ficantly impaired STAT3 phosphorylation in response to IL-11 stimulation(Fig.2a),but had little effect on IL-6,IL-27,OSM,or LIF induced STAT3 phosphorylation(Fig.2b–e),as well as CT1,CLC,or CNTFinduced STAT3 phosphorylation(Supplementary Fig.1a–c).The p.R281Q substitution failed to rescue IL-11-induced STAT1 phosphorylation in a similarly selective manner(Fig.2f).

Next,we used confocal microscopy to assess the functional consequences of p.R281Q mutated GP130 on IL-11-mediated STAT3 nuclear translocation.In HEK293 GP130-KOcells transfected with p.R281Q,the impaired STAT3 phosphorylation in response to IL-11 was associated with defective cytoplasmic to nuclear STAT3 translocation(Fig.2g,h).Furthermore,we investigated the functional consequences of GP130 p.R281Q substitution in a STAT3 luciferase reporter assay.p.R281Q caused defective luciferase induction after IL-11 stimulation,whereas the IL-6-dependent STAT3 gene expression was not affected(Fig.2i).

IL-11 signaling defect in primary patient cells with endogenous GP130 p.R281Q

We next investigated GP130-dependent cytokines on primary T and Bcells but did not find a defect in IL-6 signaling(Fig.3a).We con firmed this finding with primary patient PR281Q-derived T lymphoblasts and an EBV-transformed lymphoblastoid cell line and found normal IL-27 signaling using Tlymphoblasts(Fig.3b,c).

Primary hematopoietic cells do not express the IL11RA and the individual PR281Qdeclined askin biopsy to establish fibroblast lines for IL-11 signaling studies.To overcome this problem and to measure IL-11 signaling in the available primary cells that endogenously express GP130,we transduced CD4+lymphocytes of the patient with a lentivirus that encodes human IL11RA and made those cells responsive to IL-11.We used co-expression of GFPto control for unequal transduction ef ficiencies(Supplementary Fig.2a,b).In contrast to primary CD4+T cells that do not respond to IL-11,Tcells transduced with WTGP130 became IL-11 responsive,whereas patient cellswith the p.R281Qvariant showed impaired responsiveness to IL-11(Fig.3d,e,Supplementary Fig.2c).We used GP130 p.N404Yexpressing primary Tcells as control,as these cells do not respond to either IL-6 or IL-11 stimulation.19This con firmed the selective IL-11 signaling defect caused by GP130 p.R281Q variant in primary patient cells.

Homology modeling of the GP130/IL11RA/IL-11 complex provides insight into the cytokine-selective effects of GP130 p.R281Q

We next investigated the structural effects of p.R281Q on GP130/IL11RA/IL-11 signaling.Since published cryo-electron microscopy data suggest a similar hexameric arrangement for the GP130/IL11RA/IL-11 and the GP130/IL6RA/IL-6 complex,22but no crystal structure is currently available for the former,we performed homology modeling.R281 is part of the solvent-excluded GP130/α-receptor interface between the domain D3 of GP130 and domain D3 of IL6RA/IL11RA(Supplementary Fig.3a,b).Amino acids at this interface show a high evolutionary conservation(Supplementary Fig.3a,b)and a low genetic variation in human populations as indicated by the minor allele frequencies in the ExACdatabase(Supplementary Fig.3c).The solvent-excluded area is larger in the GP130/IL6RA interface as compared to the GP130/IL11RA interface(Supplementary Fig.3b).Our data suggest that the GP130 Q281 amino acid disrupts all salt bridge and hydrogen bonding to IL11RA,including interactions with IL11RAY260,T281,and D282,suggesting a speci fic interaction defect with IL11RA(Fig.4a–c).

We next analyzed the WTGP130 and GP130 p.R281Q variant in molecular dynamics simulations.Side chain flexibility of R281 was low in both GP130/IL6RA/IL-6 and GP130/IL11RA/IL-11 complexes,consistent with the notion that R281 is engaged in salt bridges and hydrogen bonding in both(Fig.4d).However,Q281 in the mutant receptor complex displayed a signi ficantly higher degree of freedom(Fig.4d).Consequently,we observed an increased interdomain distance between GP130 domain D3 and IL11RA D3,but not the IL6RA D3(Fig.4e).

Given the importance of GP130 R281 interaction with Y260,D282,and T281 for IL11RA association,we assessed evolutionary conservation of the GP130 D3/α-receptor D3 interface.We observed that amino acids engaged in the solvent-excluded receptor interface,and in particular Arg(R)at the position corresponding to 281 in human GP130,are highly conserved from amphibians to mammals(Supplementary Fig.3a).The size,position,and residues of the corresponding IL11RA domain D3 receptor interface were also highly conserved.By contrast,IL6RA domain D3 displayed a higher degree of amino acid variation(Supplementary Fig.3a).These data suggest that IL11RA-containing receptor complexes,in contrast with the IL6RA-containing receptor complexes,are strongly dependent on D3 domain interactions,making them more susceptible to amino acid variations.

In order to analyze cytokine/α-receptor af finity to GP130 variants fused to the fluorescent protein YPet,we made use of Hyper-IL-6-Fc(Hyp IL-6-Fc)23and Hyper-IL-11-Fc(Hyp IL-11-Fc).24These are arti ficial fusion proteins of IL-6 or IL-11 to the soluble ectodomains of IL6RA or IL11RA,respectively,and an additional human Fc-tag,facilitating the isolation of GP130-containing receptor complexes at the plasma membrane(Fig.4f).We were able to immune-precipitate similar amounts of GP130 as detected by anti-GFP immunoblotting against the YPet-tag of the GP130 variants using Hyp IL-6-Fc in either WT-or R281Q-expressing cells(Fig.4f).However,when using Hyp IL-11-Fc,the amount of precipitated YPet-GP130 was signi ficantly reduced when cells expressed the R281Q variant as compared to the WT(Fig.4f).Therefore,these in vitro data con firm our in silico predictions that the af finity of interactions between GP130 and IL11RA but not IL6RA is signi ficantly lowered by the p.R281Q variant.

Amouse model withIl6stp.R279Qvariant is associated with facial synostosis and exhibits a IL-11 signaling defect

Loss of IL11RA signaling is associated with craniosynostosis and dental abnormalities.5,14–18We hypothesized that impaired IL-11 signaling of the p.R281Q variant accounts for the skeletal abnormalities in the patient.In order to test this hypothesis,we used CRISPR/Cas9 technology to generate two independent lines of mice(lines 4 and 6)that express GP130 p.R279Qcorresponding to human GP130 p.R281Q(Figs.1d and 5a).For this,an sgRNA targeting murineIl6stexon 8 and a single-stranded(ss)DNAdonor containing substitutions c.[835G＞A;836A＞G],i.e.the mouse equivalent to the human variant as well as two silent substitutions to destroy the PAM sequence used and to insert aBglIIrestriction endonuclease site for genotyping,were injected into one-cell stage mouse embryos(Supplementary Fig.4a)and implanted into foster mice.Offspring were identi fied by genotyping(Supplementary Fig.4b,c).

To analyze the functional consequences of GP130 p.R279Q in mice,we isolated primary skin fibroblasts from WTmice(R/R)and those with the p.R279Qvariant(heterozygous R/Qor homozygous Q/Q).We con firmed equal expression and surface localization of the GP130 variants in these cells(Supplementary Fig.4d).We then stimulated fibroblasts with different cytokinesand assessed STAT3 phosphorylation by immunoblotting and flow cytometry.STAT3 phosphorylation was signi ficantly impaired in Q/Q fibroblasts of both mouse lines,when stimulated with IL-11 or Hyper-IL-11,but not when stimulated with IL-6 or Hyper-IL-6 (Fig.5b,c,Supplementary Fig.4e).In contrast to the human variant,there was a partial reduction in LIFsignaling(Fig.5b).

We next investigated the phenotype of the animals.Crossing of GP130 p.R279Q heterozygous mice resulted in the expected Mendelian ratio of homozygotes,suggesting that the p.R279Q variant is not lethal(Supplementary Fig.4f).However,the litter size was signi ficantly reduced when mother animals carried two mutant alleles(Supplementary Fig.4g).This is reminiscent of the infertility phenotype observed in homozygous femaleIl11ra-/-mice.25Flow cytometry analysis of peripheral blood did not show any signs of abnormal hematopoiesis(Supplementary Fig.4h,i).

In contrast to our individual PR281Q,but similar to IL11RA-de ficient mice,we did not observe signs of abnormal cranial suture fusion in the skulls of Q/Q mice byμCT analysis(Supplementary Fig.5a).However,in a proportion of homozygous GP130 p.R279Q mice,we observed macroscopically visible facial malformations(Supplementary Fig.5b),reminiscent of IL11RA-de ficient mice.5,26Microcomputed tomography(μCT)revealed individual mice with sideward deviation of snout growth(Fig.5d),increased ossi fication of premaxillary sutures(Fig.5e),and dental malocclusion(Fig.5f).Although the penetrance of the phenotype was not complete,the number of mice with facial phenotype was signi ficantly elevated in homozygous Q/Qmice,asassessed by the degree of sideward deviation of snout growth(Fig.5g).We did not observe signsof increased deposition of bone matrix astrabecular parameters of tibiae were similar in both,R/R and Q/Q mice(Supplementary Fig.5c).

Taken together,our novel mouse model demonstrates that a selective IL-11 signaling defect of GP130 is suf ficient to induce facial synostosis reminiscent of mice withIl11rade ficiency.

DISCUSSION

The shared use of a common signal transducing receptor subunit by cytokines and their speci ficα-receptor subunits is observed in several cytokine receptor families including the IL-6 family cytokines.27–28Due to the combinatorial nature,variants in the common receptor chain likely affect multiple signaling pathways.We describe a biallelicIL6STvariant encoding a GP130 p.R281Q substitution that confers a IL-11-selective signaling defect,while signaling of other IL-6 family members remains intact.Albeit the signaling effects were also observed in patient-derived primary cells,we cannot exclude variations due to the use of non-isogenic primary cells.The craniosynostosis and dental abnormalities seen in the patient with homozygous GP130 p.R281Q are reminiscent of the phenotype of patients with craniosynostosis and dental anomalies caused byIL11RAmutations.5,14–18,29In light of the relatively high minor allele frequency of 0.0016 in the South Asian population and the incomplete penetrance in this consanguineous family,a causal relationship between the variant and the phenotype cannot be established with a single case.We therefore generated a mouse model by genomic engineering of theIl6stlocus,resulting in endogenous expression of a GP130 p.R279Q variant.These mice had facial synostosis mirroring the phenotype developed by homozygousIl11ranull mice,5in particular premature ossi fication of premaxillary sutures,along with abnormalities in nasal bone growth and dental malocclusion.Similar to IL11RA-de ficient mice,the facial phenotype of our GP130 p.R279Qmice was not completely penetrant.The apparent reduction in phenotype penetrance in humans and mice may be due to residual IL-11 signal transmission via GP130 p.R279Q seen at the highest concentrations of IL-11.It is likely that high IL-11 concentrations are available in the developing connective tissue explaining the incomplete phenotype of the variant.Similar,in contrast to IL11RA-de ficient mice,6we did not observe a signi ficant increase in trabecular bone volume in young GP130 p.R279Q mice.However,at present we cannot exclude that reduction in IL-11 signaling in GP130 p.R279Q mice might reduce loss of bone density in aged mice as has been previously observed in IL11RA-de ficient mice.6

Environmental factors may additionally in fluence phenotype penetrance.In humans multiparity,macrosomia,30–31in utero exposure to nicotine32–33,and alcohol34have been identi fied as potential risk factors for the development of craniosynostosis.It will be interesting to see whether changes in the microbiota could also contribute to the variable penetrance of skeletal abnormalities in the context of impaired IL-11 signaling.

Recently,a population of Gli1+cells within sutures of postnatal mice was identi fied as a major mesenchymal stem cell(MSC)population that gives rise to the osteogenic cells supporting craniofacial bone turnover.Ablation of these MSCs in adult mice resulted in craniofacial suture closure.35Interestingly,suture fusion occurred in mice with incomplete depletion,pinpointing to a certain threshold of MSC number that is needed to maintain suture patency.IL-11 promotes osteoblast differentiation into osteogenic cells in vitro36–38and supports osteoclastogenesis through induction of TNFSF11(previously termed as RANKL)expression in osteoblasts and also effects on osteoclasts.6,39It is therefore tempting to speculate that IL-11 signaling is responsible for an increase in craniofacial MSCturnover but is not necessary for MSC maintenance.This may explain why the phenotype ofIl11ra-/-and ourIl6stp.R279Q mice is not 100%penetrant as other signals in some sutures and animals might be suf ficient to keep the number of MSCs high enough to prevent suture closure.It may explain why in mice,in contrast to humans,loss of IL-11 signaling does not cause pathological suture fusion of the cranial vault.A similar species-dependent discrepancy in effects of mutant receptor signaling has been observed inFgfr3P244R-mutant mice that express theFGFR3-mutation associated with coronal synostosis in human Muenke syndrome.40

In addition to the connective tissue phenotype,the GP130 variant has an effect on reproduction in mice.In mice IL-11 signaling plays an important role for decidualization of endometrial stromal cells.Consequently,Il11ra-/-female mice fail to support proper embryonal implantation and are therefore infertile.25Consistent with impaired but not fully blunted IL-11 signaling,we observed a slight reduction in litter size in our GP130 p.R279Q mice,when female mice carried at least one mutantIl6stallele.Interestingly,litter size was further decreased if father animals also carried a mutantIl6stallele,suggesting additional effectsin the embryo itself.It wasdemonstrated that LIFpromotes human blastocyst formation,embryonal stem cell survival41and embryo implantation in mice.42A reduction in LIF signaling in GP130 p.R279Q mice that we observed in primary fibroblasts might therefore account for a decrease in blastocyst survival in these animals.The situation in humans might slightly differ,as females homozygous for inactivating IL11RA variants are able to reproduce and have healthy children.5,15This can be explained by species differences as in mice,decidualization of stromal cells is dependent on the presence of a blastocyst,while in the human female decidualization occurs spontaneously within the late secretory phase of the menstrual cycle.43

Our data provide a structural explanation for the cytokineselective effect of the variant.IL-6 and IL-11 are engaged in a hexameric complex consisting of cytokine,α-receptor and GP130.The receptor complex is formed by contact sites of the cytokine to α-receptor and GP130(site I,IIa,and III),44and further stabilized by a receptor interface(site IIb)built up between domains D3 of GP130,containing R281,and D3 of theα-receptor.The interface between GP130/IL11RA seems to be smaller than GP130/IL6RA as deduced from our in silico analysis.During evolution,site IIb in both GP130 and IL11RA is highly conserved in all vertebrate animal classes.The amino acid composition of IL6RA domain D3 was more variable suggesting that stability of the IL-6 receptor complex ismuch less dependent on site IIb than the IL-11 receptor complex.The IL-6 receptor complex might have evolved with a larger surface area at site I,IIa,and III compared to the IL-11 receptor complex.This might explain why the R281Qsubstitution lowers the af finity of IL-11/IL11RA but not IL-6/IL6RA interactions with GP130.Our analysis underlines that during evolution,GP130 has developed diverse cytokine/receptor interfaces with different characteristics meeting the needs of different cytokines in speci fic tissues.

Taken together,our analysis of GP130 p.R281Q clearly demonstrate that this variant(I)causes a selective defect in IL-11 signaling due to a decrease in GP130–IL11RA interaction,(II)occurs in an evolutionary conserved region of the protein,(III)is associated with craniosynostosis in one of two human subjects,and(IV)causes facial synostosis in a mouse model.According to previously published guidelines for single-patient genetic variants all criteria,except complete penetrance are ful filled.45Since the penetrance of the phenotype is incomplete in both,human and mice we only can suggest that based on our data the GP130 p.R281Qvariant is causally linked to the observed clinical phenotype.

To the best of our knowledge,the GP130 p.R281Q variant that we identi fied is the first described cytokine-selective loss-offunction variant in a common signal transducing receptor subunit that affects signal transmission of only one cytokine.Our findings help to understand the fundamental molecular aspectsunderlying cytokine-selective signaling involving common receptor signaling subunits.

MATERIALSAND METHODS

Case studies

The clinical studies were approved by Oxfordshire Research Ethics Committee B(reference C02.143)and London Riverside Research Ethics Committee(reference 09/H0706/20).The proband was enrolled into the craniosynostosis cohort based on referral to a craniofacial unit,with craniosynostosis proven on computed tomography(CT)of the skull.

DNA was extracted from either venous blood collected into EDTA or patient-derived lymphoblastoid cell lines(LCL).All DNA was extracted using the Nucleon Blood and Cell Culture(BACC)DNA extraction kit(Gen-Probe Inc.)according to the manufacturer’s instructions.

Healthy volunteer donors were recruited as part of the Oxford Gastrointestinal Illness Biobank(REC 11/YH/0020)or obtained as leukocyte cones from UKblood donor bank.Informed consent for participation in this study was obtained from healthy donors,patients,or their parents.

Targeted and Sanger sequencing

We used Fluidigm/Ion Torrent resequencing to screenIL6STin DNA panels from subjects with craniosynostosis who were negative for mutations in the major causative genes as described previously.19Amplicons that failed quality control were reanalyzed using molecular inversion probes.

Cell culture and cytokine stimulation

HEK293 cells were cultured in Dulbecco’s modi fied Eagle’s medium(DMEM)supplemented with 10%fetal calf serum(FCS).HEK293 GP130 knockout(KO)cell lines were generated using CRISPR/Cas9 following published protocols.19,46Primary cells and cell lines were stimulated with indicated concentrations of recombinant human IL-6,IL-21,IL-27,OSM,LIF(all Peprotech);IL-10 and IL-11(both R&D Systems);CT1,CNTF(both Miltenyi Biotec),and CLC(BioLegend).

Isolation and cultivation of primary mouse skin fibroblasts

For isolation of primary fibroblasts,about 1 cm²of freshly prepared murine ear tissue was incubated for 5 min in 70%ethanol and subsequently handled under sterile conditions.The tissue was air-dried,cut into pieces of a few millimeters,and digested with a mixture of collagenase A(2 mg·mL-1),collagenase D(2 mg·mL-1),and dispase(4 mg·mL-1)diluted in DMEM for 1 h under constant shaking at 37°C.Cells were isolated by grinding the tissue through cell strainers(70–100 μm),washed once,and then cultured in DMEM+20%FCS+1%penicillin/streptomycin under constant conditionsat 37°C,5%CO2,and 96%humidity.At 80%–90%con fluency cells were washed with PBSand detached applying 1× Trypsin EDTA for 10 min at 37 °C.Required cell numbers were seeded in either DMEM without serum or DMEM with 20%FCS.

Flow cytometry of primary mouse skin fibroblasts and peripheral blood

Primary fibroblasts were washed with PBS and incubated with accutase for 3 min at 37°C.Subsequently still attached cells were gently scraped from the plates and passed through 70 to 100μm cell strainers.Cell strainers were flushed with PBS and cells collected by centrifugation.Whole blood was collected in heparinized tubes and washed once with PBSprior to staining.

Fibroblasts were stained with anti-gp130-APC-antibody(R&D Systems;clone:#125623)or isotype control for 1 h at room temperature in the dark.Whole-blood cells were stained for 30 min at room temperature in the dark with different combinations of the following antibodies:anti-CD45-BV510(BioLegend;clone 30-F11),anti-CD3-FITC(BioLegend;clone145-2C11),anti-CD19-APC(BioLegend;clone HiB19),anti-CD4-APC/Cy7(BioLegend;clone RM4-5),anti-CD8-PE/Cy7(BioLegend;clone 53–6.7),anti-Ly6G-FITC(BioLegend;clone RB6-8C5),anti-Ly6C-APC/Cy7(BioLegend;clone HK1.4),anti-CD115-PE/Cy7(BioLegend;clone AFS98).Stained whole-blood cells were subsequently incubated with RBC Lysis and Fixation Solution(BioLegend)for 15 min at 37°Cand washed twice with PBS.Samples were acquired in FACS buffer[1%BSA in PBS]on a BD FACS Canto II flow cytometry system and analyzed with FlowJo Software(Version 10.5.3,Tree Star).

STAT3 and STAT1 phosphorylation assays

Unless indicated otherwise,phosphorylation of STAT1 or STAT3 transcription factors was assessed as described by Schwerd et al.19by phos flow,immuno fluorescence and confocal microscopy or luciferase STAT3 reporter assay.Phosphorylation assays were mainly performed in parallel to evaluation of patient PN404Ydescribed by Schwerd et al.19and results from healthy donors or WTcontrols are duplicated to ensure perfect comparability.

In addition to phos flow assessment,STAT3 phosphorylation in murine cells was analyzed by SDS-PAGEand immunoblotting.In brief,24h prior to analysis primary fibroblasts were seeded in sixwell plates.The next day cells were serum starved for 4h and stimulated with cytokines at the indicated concentrations for 10 min at room temperature.Stimulation was immediately followed by cell lysis for 15 min on ice with RIPA buffer[50 mmol·L-1Tris,150 mmol·L-1NaCl,0.1%SDS,0.3%sodium deoxycholate,1%Triton X-100]with freshly added protease and phosphatase inhibitors.Samples were separated by SDS-PAGEon 10%bis-tris-gels and subjected to immunoblotting using primary antibodies anti-gp130(R&D Systems;clone:#125623),anti-STAT3(Cell Signaling;clone 124H6),anti-phospho-STAT3(Cell Signaling;clone:D3A7),anti-beta-actin(Sigma Aldrich;clone AC-15).Signal intensity was determined by densitometry using ImageJ(version 1.52n).

Lentivirus production and IL11RAoverexpression in CD4+memory Tcells

The empty vector p CDH-EF1-MCS-T2A-cop GFP(CD526A-1)and the vector carrying humanIL11RAtranscript variant 3(NM_001142784.2)for ectopic expression were purchased from SBISystems Biosciences.

Lentiviral particles were produced by transiently transfecting HEK293 cells with the above described transfer vectors together with the ViraPower™lentiviral packaging mix(Invitrogen)in 150mm cell culture dishes(Corning).Brie fly,HEK293 cells were transfected with a cocktail of transfer vector and packaging mix in Opti-MEM(Gibco),using Lipofectamine®2000(Thermo Fisher)as a transfecting agent according to the manufacturer’s instructions.Culture supernatants containing viral particles were harvested at 72 h post-transfection and titers were determined by limiting dilution on HEK293 cells.

Resting memory CD4+T cell lines were transduced by spin infection(60 min;800g;32 °C)on anti-CD3(5 μg·mL-1;Biolegend;clone:OKT3)and anti-CD28(1 μg·mL-1;BD Biosciences;clone CD28.2)coated 24-well plates(1×106cells/well)in the presence of 5 μg·mL-1polybrene.The medium was then replaced and cells were cultured in IL-2-containing(500 U·mL-1)medium.After 48 h of culture,cells were transferred to uncoated plates.Following a minimum of 7 days of culture,expanded T cells were starved overnight in IL-2-free medium,washed extensively,and analyzed for IL-10 and IL-11 responsiveness after 30 min of stimulation by intracellular staining for phosphorylated STAT1 and STAT3.To exclude non-transduced T cells from the analysis gating was performed on the GFP+population.Flow cytometry data were analyzed with FlowJo(Version,Tree Star).

Receptor complex isolation and immunoblotting

GP130-de ficient HEK293 cells were transiently transfected with plasmids encoding for GP130 WT-YPet or GP130 R281Q-YPet variant using linear polyethyleneimine Max(Polysciences#24765)as transfection agent.Twenty-four hours post-transfection cells were starved overnight in DMEM with 0.5%FCS followed by incubation on ice for 30 min and subsequent stimulation with recombinant Hyper-IL-6-Fc or Hyper-IL-11-Fc for 30 min on ice.Cells were lysed at 4 °C in RIPA buffer(50 mmol·L-1Tris,150mmol·L-1NaCl,0.1%SDS,0.3%sodium deoxycholate,1%Triton X-100)supplemented with protease and phosphatase inhibitors.Lysates were cleared by centrifugation at 12 000gfor 15 min.

Lysates for immunoprecipitates were incubated with Protein A/G beads(Millipore)for 1 h at 4°C.Subsequently washed three times in RIPA Buffer supplemented with protease and phosphatase inhibitors.Beads were incubated for 5 min at 90 °C in 2×reducing Laemmli Buffer.

All samples were analyzed by 10%bis-tris SDS-PAGE and immunoblotting.The following antibodies were used:anti-GFP(Roche,11814460001),anti-beta-actin(clone:AC-15)and anti-IgGFc(R&D Systems,Cat.No.G-102-C).Secondary antibodies were anti-mouse-HRPor anti-rabbit-HRP(both Dianova).

Homology modeling and molecular dynamics(MD)simulations Due to the unavailability of IL11RA structure coordinate files,a model was generated using the structure coordinates of IL6RA(pdb entry 1P9M)as template and amino acids 112–320 of the IL11RA sequence deposited in UniProt Q14626 with matching alignment to the IL6RA.Modeling was performed using the MODELLERinterface of the UCSFChimera package.47WTR281 in GP130 was exchanged to Q and the corresponding rotamer was introduced searching the Dunbrack library.48The rotamer with the highestχ-angle probability score was chosen.

Molecular dynamics calculations were performed on trimeric complexes either consisting of GP130/IL6RA/IL-6 or of GP130/IL11RA/IL-11 using AmberTools17(ref.49)compiled for parallel computing on 32 cores.IL-11(4MHL.pdb)was fitted into the complex by superimposing it with IL-6 with the tool MatchMaker of UCSF Chimera.50Initially,the protein force field ff14SB was applied to the structure models and a solvate box was generated using the tip3p water model.51For the GP130 p.R281Q variant,overall charge was neutralized using Na+ions.To solve possible high-energy states,2 000 steps of minimization were performed prior to MD simulation.For the MD simulation,the system was heated to 300 K over 100 ps and hold at constant volume and temperature conditions(NVT)for additional 20ps.Subsequently,parameters were changed to 1 bar constant pressure and 300 K constant temperature(NPT)to resemble laboratory conditions and structural dynamics were analyzed for 1 ns in the NPTensemble.All calculations were performed by parallel computing on a NEC HPCLinux-Cluster at the CAU University Computer Center.

Trajectory filesfrom the MDsimulations were analyzed using UCSF Chimera.The center of mass for GP130 D3 was de fined of residues 198–298 and residues 215–317 for the IL11RA.To analyze rotamer flexibility throughout the 1ns simulation every 0.1ns PDBs were saved and subsequently superimposed using the MatchMaker tool of UCSF Chimera.Root mean square deviation(RMSD)of arginine or glutamine,respectively,was calculated as mean RMSD throughout the simulation relative to positions of arginine/glutamine at 0ns.

The mean buried area between GP130 D3 and the alpha receptor D3 was calculated from solvent-excluded surfaces from ten equally distributed time points of the 1 ns MD simulations of the IL-6 and the IL-11 complex,respectively.Information about highly contributing amino acids were extracted at the last frame of the simulations.

Analysis of evolutionary conservation

Conservational scores were calculated with the AL2CO algorithm in UCSFChimera on the basis of multiple sequence alignments.For conservation per class,170 sequences for GP130 were retrieved from the NCBI protein database and subjected to multiples sequence alignment using Clustal Omega.52Alpha receptor sequences were obtained by blast search with a threshold of e-4in the RefSeq-Database against the full-length alpha receptors(IL6RA:UniProt P40887,IL11RA:UniProt Q14626)using NCBIProtein BLAST.We only selected sequences annotated for IL6RAor IL11RAfrom species that also displayed an annotated database entry for GP130 and subjected them to multiples sequence alignments using Clustal Omega(Supplementary Table 2).AL2CO scores were mapped on a representative structure per classobtained by homology modeling of a class-speci fic sequence against the human ortholog.In the case of mammals,the human structures were used.Amino acids corresponding to pos.281 in human GP130,as well as amino acids corresponding to pos.269,281,282 in IL11RA,and the corresponding amino acids in IL6RA were labeled in the structures.Amino acids contributing to the buried interface were encircled.

Data on mutational hotspots in GP130,IL6RA,and IL11RA were obtained from the Exome Aggregation Consortium(ExAC).Results were converted into a format readable for mapping with UCSF Chimera and subsequently mapped onto the structures of human GP130 D3,IL6RA D3,and modeled IL11RA D3.

Mouse model generation

The sgRNA sequence targeting exon 8 of murineIl6stwas designed using the CRISPOR Program.53The template for transcription was derived by PCRusing Q5-Polymerase(Biolabs).Transcription was performed using the HiScribeT7 kit(Biolabs,E20140S) with subsequent puri fication of the transc ript with the MEGAClearTM kit(Fisher Scienti fic,AM1908),both according to the manufacturer’s instructions.

One-cell stage embryos derived from superovulated C57BL/6JUke mice were injected using 10 ng·μL-1sgRNA,20 ng single stranded repair template(Sigma)introducing NM_010560:c.[835G＞A;836A＞G]for the p.R279Q substitution and further silent mutations for aBglII restriction site and a degenerated PAM sequence.We selected the CAGcodon for glutamine,as the CAA codon is infrequently used in mice.The repair template(CTCCAGTAGCCCTTCCCACTGTCCTTAATGGACCGGATCCTAAACAC ATATTCTGTAAAAGGCTTGAGaTCtTGCACAGTGAAGGAAGTctGAGG AGACATTGTATCTTCAAGAGGGACC)was transfected jointly with 50 ng·μL-1Cas9 protein(IDT).Embryos were implanted into F1 foster mothers(C57BL6×CBA)and the resulting offspring was analyzed by PCRusing(Il6st-F:GGTCTGGTTCTTTAA GACAGG CTCTC,Il6st-rev:CACCACTTTTACGTATGTCTTCGTATGTG)andBglII digestion.Correct integration of the repair construct was veri fied by sequence analysis.Two independent lineages(termed lines 4 and 6 throughout the manuscript)withIl6stc.[835G＞A;836A＞G]mutation(p.R279Q)were obtained and further bred at the CAU Animal Facility.All experiments were performed in accordance with the local guidelines for animal care and protection.

Il11ra-de ficient mice54were on a C57BL/6 background and maintained as previously described.25

Analysis of craniofacial and skeletal phenotype in mice

Skulls were collected from WT,heterozygous p.R279Q mutant,homozygous p.R279Q mutant littermates of mouse lines 4 and 6,as well as mice and stripped of flesh and tendons.Skulls were fixed in 10%buffered formalin at 4°C for 1 to 2 days and subsequently stored in PBScontaining 0.05%sodium azide at 4°C.Images of cleaned and fixed skulls were taken and lateral twisting of snouts further analyzed.The angle between snout tip and sagittal suture was determined using ImageJ Software(Version 1.52n).

During macroscopic phenotype assessment,skulls were determined as phenotypically aberrant if one of the following criteria were met:sideward deviation of snout,shortening of snout,or downward deviation of snout.

Skulls from three animals and long bones from eight to ten animals per genotype were embedded in 1%agarose in dd H2O for μCT analysis.μCT scans were performed at the Molecular Imaging North Competence Center(MOIN CC),Department of Radiology and Neuroradiology,University Medical Center Schleswig-Holstein using a vivaCT 40(70 kVp,114μA,300 ms integration time,1 000 projections on 180°2048 CCD detector array,cone-beam reconstruction,ScancoMedical).All scans were done at an isotropic voxel size of 15.6μm.Images were further analyzed using ImageJ(Version 1.52n)and ParaView(Version 5.6.0,Kitware)software.Morphometric analysis of tibiae was performed ex-vivo using a VivaCT 80(Scanco AB,Brüttisellen,Switzerland)micro-CTscanner with 15.6μm isotropic voxel size(70kBp,114μA,31.9 mm FOV,300 ms integration time,software binning:1.5,no HWbinning,1 000 projections/180°,standard reconstruction with beam hardening correction,bone calibration in mgHA·cm–3).For segmentation and quanti fication of parameters,the manufacturer´s software was used.For trabecular measurements a volume of interest(VOI)of 2 mm(126 slices)axial length was selected starting 0.3 mm(20 slices)below the epiphyseal plate.Trabecular bone was contoured automatically in the diaphyseal area,but partially by manually drawing 2D regions of interest(ROIs)every 5–10 slice with geometric morphing approaching the thin cortex and more complex endosteal envelope near the proximal part of the tibiae(Scanco uct_evaluate V6.3–5).The images were binarized using a threshold of 250 mgHA·cm-³,resulting in a mask solely with bone and background voxels.Trabecular bone volume density(bone volume(BV)/total volume(TV)),trabecular number(Tb.N),trabecular thickness(Tb.Th),and trabecular separation(Tb.Sp)were calculated(IPL V5.15).

Protein sequence alignment

Multiple sequences were aligned using ClustalW2.55Data were obtained from the National Center for Biotechnology Information(NCBI).Sequence alignment is based on the following accession numbers: NP_001106976.1, NP_001124412.1, NP_990202.1,NP_034690.3, NP_002175.2, NP_004834.1, NP_000751.1,NP_002301.1,NP_003990.1,NP_005526.1 and NP_001550.1.

Statistical analysis

Results were analyzed with GraphPad Prism version 5.00(GraphPad software,Inc.,San Diego,CA)or Rstudio(version 1.2.1335).Signi ficance was determined by two-sided Mann–WhitneyUtest,one-way ANOVA,or Kruskal–Wallis with multiple comparisons post-test.Distribution of extreme snout deformation was analyzed using Fisher’s exact test.Pvalues below 0.05 were considered as signi ficant.

Online resources/URLs

The following online data sources have been accessed:

1 000 Genomes.http://www.1000genomes.org

dbSNP.http://www.ncbi.nlm.nih.gov/SNP

GenBank.http://www.ncbi.nlm.nih.gov

gnomAD.http://gnomad.broadinstitute.org/

PolyPhen.http://genetics.bwh.harvard.edu/cgi-bin/pph

SIFT.http://sift.jcvi.org

STRING.http://string-db.org

ExACbrowser.http://exac.broadinstitute.org

GDI-server. http://pec630.rockefeller.edu:8080/GDI/resultGene

Only.jsp

COSMIC.http://cancer.sanger.ac.uk/cosmic/gene/analysis?ln=IL6ST

Blast.https://blast.ncbi.nlm.nih.gov/Blast.cgiACKNOWLEDGEMENTS

We would like to thank all study pa rticipants and their families for their contribution to the study.We acknowledge the contribution of the Oxford Gastrointestinal Illness Biobank.We thank Dr.Jane Hurst for patient referral.We are grateful to Irm Hermans-Borgmeyer(Center for Molecular Neurobiology Hamburg(ZMNH),Hamburg)for support during generation of GP130 p.R279Q mice.We thank Tracy Putoczki and Michael Grif fin for discussions on the structural aspects of this work.We are grateful to Dr.Simone Knief(University Computing Center,CAU Kiel)for compilation of AmberTools17 for parallel computing on 32 cores and for further help with highpower computing.We thank Ilka Thomsen,Sarah Schacht,Monja Gandraß,Julia Bolik,Pit Christoffersen,Christian Bretscher,and Fabian Neumann (all Institute of Biochemistry,Kiel)for their valuable support during animal experimentation.We are grateful to Christoph Garbers(Institute of Pathology,University Hospital Magdeburg)for the provision ofIl11ra-/-mice.The views expressed are those of the authors and not necessarily those of the NIHRor the Department of Health and Social Care.This work was supported by funding from the Medical Research Council(MRC)through the WIMM Strategic Alliance(G0902418 and MC_UU_12025)and to E.Y.J.(G9900061),the Department of Health,UK,Quality,Improvement,Development and Initiative Scheme(QIDIS)(AOMW),and the Wellcome Trust(Project Grant 093329 to AOMW and SRFT;Investigator Award 102731 to AOMW;grant 090532/Z/09/Z supporting the Wellcome Trust Centre for Human Genetics).H.H.U.is supported by the Crohn’s&Colitis Foundation of America(CCFA),the Leona M.and Harry B.Helmsley Charitable Trust.This project was funded by the NIHROxford Biomedical Research Centre.T.S.was supported by the Deutsche Forschungsgemeinschaft(SCHW1730/1-1).F.K.,D.S.-A.,and S.R.-J.was supported by the Deutsche Forschungsgemeinschaft(DFG),Bonn(grant number SFB841 to F.K.,D.S.-A.,and S.R.-J.;SFB877 to S.R.-J.)and the Cluster of Excellence “In flammation at Interfaces”to S.R.-J.

AUTHORCONTRIBUTIONS

S.R.F.T.and A.O.M.W.initiated the project and discovered the genetic variant.T.S.,F.K.,D.A.,Y.-H.C.,A.L.,M.M.,and N.S.performed experiments.U.B.generated GP130 p.R279Q mice.S.A.W.and A.O.M.W.contributed to clinical data.S.M.and E.Y.J.performed structural analysis.T.D.and C.-C.G.performed imaging analysis.F.K.and J.J.performed molecular dynamics simulation and evolutionary analysis.S.R.-J.and J.S.provided reagents and contributed to manuscript writing.A.O.M.W.,D.S.-A.,and H.H.U.led the genetic and functional workup.

ADDITIONAL INFORMATION

The online version of thisarticle(https://doi.org/10.1038/s41413-020-0098-z)contains supplementary material,which is available to authorized users.

Competing interests:No con flict of interest related to this article.H.H.U.has received research support or consultancy fees from UCBPharma,Eli Lilly,Boehringer Ingelheim,P fizer,Celgene,MiroBio and AbbVie.H.H.U.,A.L.,and Y.-H.Care supported by a research collaboration with Celgene.