Absence of GP130 cytokine receptor signaling causes extended Stüve-Wiedemann syndrome

GP130 is a ubiquitously expressed cytokine receptor subunit and a key component of the IL-6 cytokine family receptors mediating signaling of 10 cytokines including IL-6, IL-11, IL-27, oncostatin M (OSM), cardiotrophin-1, cardiotrophin-like cytokine, neuropoietin, and leukemia inhibitory factor (LIF; Kang et al., 2019; Murakami et al., 2019). Collectively, these cytokines are important in inflammation, organogenesis, and cell survival. GP130 forms a hexameric receptor complex after binding the IL-6–activated IL-6R or IL-11–activated IL-11R and heterotrimeric GP130-LIF receptor (LIFR)-LIF or GP130-OSMR-OSM receptor complexes. Upon cytokine recognition, signal transduction via GP130 is mediated through the JAK-STAT pathway that principally results in STAT3 phosphorylation (pSTAT3) and activation (Kang et al., 2019; Murakami et al., 2019).

The importance of GP130-dependent signaling has been highlighted by gene deletion studies of Il6st or its cytokine-specific coreceptors (Akira, 2000). Homozygous deletion of Il6st is lethal in utero for mice, due to ventricular myocardial hypoplasia, disrupted placental architecture, and a loss of fetal liver hematopoiesis (Yoshida et al., 1996). This phenotype is similar to animals with targeted disruption of Stat3 (Takeda et al., 1997) or Lifr (Ware et al., 1995). Mice with conditional deletion of Il6st have elucidated the role of GP130 in other organs (Yoshida et al., 1996), including the lungs, central nervous system, immune system, and skeleton (Itoh et al., 2006). Studies of knock-in mouse mutants with impaired activation of either STAT or MAPK pathways by GP130 show that GP130 signaling ensures normal bone growth in a STAT-dependent manner (Sims et al., 2004).

We have previously shown that cytokine-selective defects in GP130 due to homozygous missense variants in IL6ST (c.1210A>T; p.Asn404Tyr and c.1493C>T; p.Pro498Leu) cause immunodeficiency with elevated IgE and craniosynostosis (Schwerd et al., 2017; Shahin et al., 2019). These variants cause selective defects in IL-6, IL-11, IL-27, and OSM signaling, while LIF signaling is preserved. Many features of the phenotype in those patients overlap with the autosomal dominant hyper-IgE recurrent infection syndrome (MIM #147060) that is associated with STAT3 mutations (Holland et al., 2007; Minegishi et al., 2007). It is thought that the skeletal phenotype is caused by defective IL-11 signaling, since patients with pathogenic IL11RA variants present with craniosynostosis and dental anomalies (MIM #614188; Kreiborg-Pakistani syndrome; Keupp et al., 2013; Nieminen et al., 2011). Based on the lethality in mice and the strong evolutionary selection against loss-of-function variants in human genomic databases, it has been assumed that complete absence of GP130 would be lethal in humans (Schwerd et al., 2017; Shahin et al., 2019).

Stüve-Wiedemann syndrome (STWS; MIM #601559) is a semilethal ultra-rare autosomal recessive condition with features of skeletal dysplasia (short bowed tubular bones, osteoporosis, and camptodactyly), respiratory distress, feeding difficulties, hyperthermic episodes, myotonia, and decreased sensitivity to pain. In most patients, STWS is caused by biallelic loss-of-function variants in the LIF receptor gene (LIFR; Dagoneau et al., 2004). Yet some STWS patients lack mutations in LIFR, suggesting genetic heterogeneity of this condition (Jung et al., 2010).

Here, we describe three unrelated families with seven affected individuals that presented with a lethal congenital disorder resembling an extended STWS (or STWS type II). In five patients with available DNA, the molecular diagnosis of homozygous loss-of-function variants in the IL6ST gene was established.

Family A is a consanguineous family from Afghanistan (Fig. 1 A; affected individuals are referred to as A-II-1 and A-II-2). The first fetus (A-II-1) had features of lethal skeletal dysplasia during routine ultrasound at gestational age (GA) 19+4 wk, and pregnancy was terminated (Table 1). Postmortem examination and skeletal radiographs confirmed severe bent bone dysplasia (Fig. 1, B and C). During the second pregnancy (A-II-2), ultrasound at GA 16+0 wk showed similar abnormalities as in the first fetus. The female baby was born via planned Caesarean section at GA 36+3 wk. She had a small thorax, short bowed extremities, and skeletal radiographical features consistent with STWS (Fig. 1, D and E), resembling the bone phenotype of A-II-1 (Fig. 1, B and C). Within a few hours of birth, thrombocytopenia was noted (A-II-2 platelet counts: day 1, 27 × 10⁹ platelets/liter; day 2, 57 × 10⁹/liter; normal range, 80–350 × 10⁹/liter). She developed progressive respiratory insufficiency, heart failure, and irreversible pulmonary hypertension, from which she died 2 d after delivery. Peripheral blood investigations identified normal C-reactive protein (A-II-2 C-reactive protein: day 1, <0.60 mg/liter; day 2, <0.60 mg/liter; normal range, <3.0 mg/liter) and low fibrinogen (A-II-2 fibrinogen: day 1, 0.5 g/liter; day 2, 1.8 g/liter; normal range, 2.0–4.0 g/liter). Postmortem examination confirmed extensive skeletal deformities with uneven bone trabecular pattern in the long bones of extremities (Fig. 1 F; representative normal control for A-II-1 is shown in Fig. S1 A), growth retardation, and membranous lung disease.

DNA of patient A-II-2 was investigated by prenatal array-CGH and gene panel sequencing for osteogenesis imperfecta genes, but no pathogenic variants were detected. Similarly, no pathogenic variants were found by whole genome sequencing (WGS) analyzed using a panel of known skeletal dysplasia genes (Genomics England PanelApp; https://panelapp.genomicsengland.co.uk/panels/309/). Subsequent WGS analysis on the two patients and their parents revealed a homozygous nonsense variant in both affected offspring in IL6ST (NM_002184.3: c.841C>T; p.Arg281*), confirmed by Sanger sequencing (Fig. 1 G). The parents and a third fetus (A-II-3, ongoing pregnancy) were heterozygous carriers (Fig. 1 A). The nonsense variant is predicted to delete the C-terminal half of GP130. Irrespective of whether nonsense-mediated decay occurs or a stable version of the truncated protein is expressed, a lack of the membrane-binding region would prevent signaling. The c.841C>T variant has not been observed in our in-house database (n = 1,432), gnomAD, or 1000 Genomes. No other genetic variants in family A explained the phenotype (Table S1).

Comparing bone histology from patient A-II-1 with a control showed grossly normal organization of the growth plate, with swollen chondrocytes and decreased levels of proteoglycans (Fig. S1 B). Less bone matrix was observed in the primary spongiosa (bone volume/total volume was 22.7 ± 1.2% in control versus 12.4 ± 2.0% in A-II-1). This was associated with similar numbers of DMP-1⁺ late osteoblasts but a 3.5-fold increase in DMP-1⁺ osteocytes in this region (the number of DMP-1⁺ cells/mm² was 1,390 ± 520 in the control versus 4,970 ± 820 in A-II-1). The size and the number of nuclei per cathepsin-K⁺ osteoclasts were greater in the primary spongiosa of A-II-1 compared with control (Fig. S1, C–E). All these changes suggest decreased activity of the growth plate. There was a substantial increase in diaphyseal bone volume (bone volume/total volume was 9.20 ± 2.36% in control versus 63.0 ± 5.2% in A-II-1), but no changes in the number of osteoblasts, osteocytes, or osteoclasts were detected between the patient and control samples.

We have subsequently identified a second consanguineous family (family B) from Saudi Arabia, with at least two fetuses affected by STWS-like skeletal dysplasia having the same homozygous mutation in c.841C>T; p.Arg281* in IL6ST as family A (Fig. 1 H). This family was previously included in a large-scale, first-tier clinical exome sequencing in a highly consanguineous population, where IL6ST was identified as a candidate gene but not further functionally characterized (Monies et al., 2019). The heterozygous parents had four pregnancies. Two fetuses (B-II-3 and B-II-4) were homozygous for c.841C>T; p.Arg281* (variants identified by exome and Sanger sequencing respectively; sequence of B-II-4 is shown in Fig. 1 L). B-II-3 was born spontaneously at term but died 2–3 h after delivery. He had short femora (less then fifth centile), bowed tibia (Fig. 1 I), multicystic dysplastic right kidney, dilated cisterna magna, and a small chest. Fetal ultrasound of B-II-4 showed normal nuchal translucency scan at 12 wk and short bowed bones (Fig. 1, J and K), and he died in utero. The same couple had two other pregnancies: the first fetus (B-II-1) died in utero with multiple anomalies and bowed bone deformities and was likely affected with the same condition, but no DNA sample was available. The second pregnancy was a miscarriage (B-II-2). Analysis of single nucleotide polymorphism data and runs of homozygosity regions between the two families, A and B, carrying the mutation c.841C>T; p.Arg281*, suggested that the two families were not related.

The third family (family C) is nonconsanguineous, of Dutch origin (Fig. 1 M). This family had two affected children (C-II-1 and C-II-3) and one healthy child (C-II-2). The first child was born at term with bowed femora and tibia (Fig. 1 N) as well as camptodactyly. He died 2 d after birth due to respiratory failure. The second affected child (C-II-3) was born at term and had mildly bowed femora (Fig. 1 O), camptodactyly, hyperhidrotic hands, tented upper lip, and high palate. The patient had infantile sucking, swallowing, and feeding difficulties but survived 6 yr. She had short stature, scoliosis, and motor delay and presented with multiple bouts of hyperthermia that on one occasion caused a febrile convulsion. C-II-3 had normal dentition but was troubled by eczematoid dermatitis from the age of 1 yr onward. At the age of 6 yr, she had scoliosis surgery and died 2 d later due to cardiorespiratory arrest. In family C, no mutations in LIFR were identified by Sanger sequencing (Jung et al., 2010). Subsequently, whole exome sequencing identified a homozygous intronic splice site variant within the IL6ST locus (NC_000005.10:g.55247753A>G; Hg19, c.1699+4A>G) in C-II-3. Among all homozygous or compound heterozygous variants (Table S1), the IL6ST variant was considered the top candidate. The main features of the patients are summarized in Table 1.

Sanger sequencing confirmed the homozygous c.1699+4A>G variant in DNA from EBV-transformed lymphoblastoid cell lines (EBV-LCLs) of C-II-3 (Fig. 1 P). The family has been lost to follow-up, and no DNA sequencing or clinical data are available from the parents, C-II-1, or unaffected C-II-2. To confirm that the apparently homozygous variant in the unrelated family is not caused by a compound heterozygosity with a deletion of the corresponding allele, we performed multiplex ligation-dependent probe amplification (MLPA) with four probes in the IL6ST gene (exons 4, 8, 13, and 17) and two probes in the two genes adjacent to IL6ST (ANKRD55 and IL31RA). No deletion was identified (Fig. S1 F).

The c.1699+4A>G variant is located at the splice donor site of intron 13 of IL6ST (Fig. 1 Q). Natural donor splice site variants at +4 are functionally relevant and highly conserved (Caminsky et al., 2014). The variant is predicted to prevent intron splicing (MaxEntScan tool drop in the score from 4.88 down to 0.52 caused by the variant). The transcript inferred pathogenicity score indicated a damaging splicing effect, with a score of 0.955 (cutoff is 0.4; highest is 1). This site is highly conserved among vertebrates (Fig. 1 Q) and in humans, as no variation was observed in 220,000 alleles of gnomAD, 1000 Genomes, or in-house databases.

To investigate the in vitro effects of the p.Arg281* variant, we used a GP130-deficient HEK293 cell line in transfection experiments (Fig. 2). As the nonsense variant is predicted to delete the C-terminal half of GP130 (mutation site is in exon 8), we first measured the expression of IL6ST exons 4–5 by quantitative RT-PCR. Exons 4–5 of IL6ST were comparably detectable in GP130-deficient HEK293 cells transfected with plasmid encoding WT GP130 and p.Arg281*, confirming the presence of the truncated GP130 receptor mRNA (Fig. 2 A). Transfection with WT GP130 restored GP130 surface expression, whereas it remained absent in the p.Arg281* transfected cells (Fig. 2 B). Functionally, transfection with WT GP130 plasmid restored pSTAT3 in response to the GP130-dependent cytokines IL-6, IL-11, IL-27, OSM, and LIF, whereas cells transfected with the p.Arg281* variant remained unresponsive (Fig. 2, C–G). By contrast, the GP130-independent pSTAT3 response to IFN-α was unaffected in all conditions (Fig. 2 H).

Consistent with homozygous nonsense variant, expression of GP130 in femur bone sections of A-II-1 was absent, whereas staining for the osteoprogenitor marker osterix was comparable to controls (Fig. 2 I). Next, we measured the surface expression of GP130 in patient-derived amniocytes by flow cytometry. GP130 protein expression was absent in amniocytes from A-II-1 and A-II-2 but present in healthy controls (Fig. 2 J). Although OSM, LIF, and IL-11 stimulation resulted in variable intensities of pSTAT3 in control amniocytes, the patient-derived amniocytes consistently displayed a complete loss of signaling (Fig. 2, K–M). In agreement with the minimal IL6RA expression (Kang et al., 2015; Fig. S1 G), IL-6-induced pSTAT3 was undetectable in both control and patient amniocytes. We noticed higher pSTAT3 baseline levels in A-II-1 and A-II-2 amniocytes compared with controls (Fig. 2, K–M). This may be explained by a feedback mechanism similar to the negative regulation of GP130 signaling by SOCS3 (Babon et al., 2014; Milner et al., 2015). Consequently, we found a reduction in SOCS3 mRNA expression in unstimulated amniocytes from A-II-2 compared with controls (Fig. 2 N). To confirm the causal relationship between the IL6ST variant and the defective GP130-dependent cytokine response, we performed a rescue experiment via lentiviral transduction of WT GP130. Transduction of WT GP130 partially restored GP130 expression as well as GP130-dependent OSM signaling, as indicated by the increase in pSTAT3 in patient-derived amniocytes (Fig. 2 O).

To investigate the genotype-phenotype segregation in the healthy parents of family A, we measured GP130 expression and pSTAT3 in response to IL-6 stimulation from CD4⁺, CD8⁺ T cells, and CD14⁺ monocytes (Fig. S1, H–M). Both father and mother displayed a mild reduction of GP130 protein expression in primary T cells and monocytes compared with healthy control individuals (Fig. S2, H–J). Monocytes from both parents showed a mild reduction of pSTAT3 in response to IL-6 (Fig. S2, K–M).

We next investigated the consequences of the intronic pathogenic c.1699+4A>G variant in family C. The variant is located at the splice donor site of intron 13 of IL6ST (Fig. 1 Q). To investigate whether exon 13 is expressed, we first measured mRNA expression of IL6ST exons 4–5 and exons 12–13, respectively. Exons 4–5 of IL6ST were detectable in EBV-LCLs of patient C-II-3 and three controls (Fig. 3 A), while expression of exons 12–13 of IL6ST was absent in C-II-3 (Fig. 3 B), suggesting skipping of exon 13. By fragment size analysis in agarose gel and Sanger sequencing, the PCR generated a longer product in healthy EBV-LCLs, compared with that from patient C-II-3 (Fig. 3 C, lane 4). cDNA sequencing of C-II-3 using primers on exons 11 and 14–15 (primer pair locations underlined in Fig. 3 D) showed absence of exon 13 sequence. A deletion of exon 13 is in-frame and is predicted to lead to a truncated protein retaining its membrane spanning and intracellular domains. Nested RT-PCR confirmed there was no full-length splice product from C-II-3, as indicated by lack of exon 13 amplification (Fig. S1, O and P; primer pair 3). Therefore, at the age when the cells from C-II-3 were obtained, the mutation was not leaky.

Consequently, we detected the presence of GP130 protein on the surface of EBV-LCLs of C-II-3, but with a 20–30% reduction compared with three healthy control EBV-LCLs (Fig. 3 E). The presence of GP130 raised the possibility of residual function. We stimulated EBV-LCLs from C-II-3 with IL-6 or IL-21 and recorded the presence of pSTAT3 in response to IL-21 but not IL-6 (Fig. 3, F and G).

To confirm the functional consequence of c.1699+4A>G mutation on GP130-dependent cytokines in particular LIF signaling, we transfected GP130-deficient HEK293 cells with plasmids encoding WT GP130 and c.1699+4A>G (Fig. 4). GP130 surface expression was detectable but reduced in c.1699+4A>G transfected cells, as well as in EBV-LCLs from C-II-3 (Figs. 4 A and 3 E). Functionally, cells transfected with plasmids encoding c.1699+4A>G did not respond to GP130-dependent cytokines, including LIF (Fig. 4, B–F). Stimulation with IFN-α resulted in comparable pSTAT3 response regardless of the transfection conditions (Fig. 4 G). We conclude that the c.1699+4A>G variant leads to a truncated extracellular GP130 receptor that is expressed but nonfunctional.

To examine the structural consequences of the c.1699+4A>G variant, we modeled the truncated protein using available x-ray crystal and EM images (Boulanger et al., 2003; Skiniotis et al., 2005, 2008; Xu et al., 2010). We mapped the peptide sequence coded by exon 13 (domain 6) onto the structure (Fig. 4 H). The structural analysis suggests that cytokine binding could occur in the mutant receptor (and the antibody epitopes are largely intact, explaining the detection on the surface) but predicts defective signal transduction via domain 5 and 6 interactions.

To confirm that any residual pSTAT3 response in both the p.Arg281* and c.1699+4A>G variants was not missed at our 15-min measurement point due to delayed kinetics, we stimulated transfected cells with indicated cytokines over a time course (Fig. S2, A–F). GP130-deficient HEK293 cells transfected with p.Arg281* and c.1699+4A>G variants consistently failed to respond to GP130-dependent cytokine stimulations at all time points tested, whereas pSTAT3 response over IFN-α stimulation decreased over time, mirroring the WT.

Many cytokine families have evolved highly conserved common receptor chains that define them, while cytokine-specific receptors show less evolutionary conservation. By nature of this combinatorial biology, mutations in the common chain have detrimental consequences, since they affect signaling of multiple cytokines. Genetic defects in IL2RG cause X-linked severe combined immunodeficiency due to defects in the entire IL-2 cytokine family (Leonard et al., 2019; Liao et al., 2011), and mutations in IL10RB result in infantile-onset inflammatory bowel disease, largely due to defective IL-10 signaling, while also associated with defective IL-22, IL-26, and IL-28 signaling (Ouyang and O’Garra, 2019).

Here we report a syndrome that is caused by the complete absence of signaling by the IL-6 cytokine family. We conclude the following. (a) Deleterious variants in IL6ST lead to complex phenotype with lethal cardiopulmonary defects, skeletal abnormalities, autonomous dysfunction and additional variable clinical features, an extended STWS phenotype (or STWS type II). (b) Patients with IL6ST loss-of-function variants can survive the intrauterine period and, in a few rare cases, the first few years of life. This is in contrast with embryonically lethal phenotype in gp130 knockout mice that may be explained by differences in pluripotent stem cell development (Onishi and Zandstra, 2015; Schnerch et al., 2010) and/or differences in placental support (Malassiné et al., 2003). (c) The absence of symptoms in the parents with heterozygous IL6ST and 20–44% reduction in GP130 surface expression suggest that one defective copy of IL6ST is not compensated for by the remaining WT allele, and that haploinsufficiency is not deleterious. (d) The similarities of the phenotype of our patients with IL6ST protein-truncating variants and those of classical STWS caused by biallelic LIFR mutations suggest that most of symptoms of our reported patients were due to defective LIF signaling (Nicola and Babon, 2015). Patient C-II-3 survived until 6 yr, in agreement with the fact that some patients with classical STWS survive during childhood, since their dysautonomia improves after the first year of life.

It is likely that different parts of defective cytokine signaling through GP130 explain the phenotypic components of the condition. In addition to the pathognomonic features of STWS caused by defective LIF signaling, this may explain the kidney abnormalities seen in B-II-3, since LIFR deficiency was similarly associated with kidney malformations in patients and in vivo mouse models (Kosfeld et al., 2017). Defective LIF signaling explains why six of seven patients with confirmed skeletal abnormalities were unable to survive the intrauterine or neonatal period. This confirms the phenotypic similarity of our patients with those with classic STWS due to LIFR deficiency and contrasts with previously reported GP130 missense variants (p.Asn404Tyr and p.Pro498Leu). These two variants have defective IL-6 and IL-11 signaling, reduced IL-27 and OSM signaling, and largely normal LIF signaling. We can only speculate as to the selectivity of these previously reported mutations in contrast with the phenotype of our patients. Both p.Asn404Tyr and p.Pro498Leu mutations are predicted to affect the secondary and tertiary structure of the domain 4 and 5 of the GP130 receptor. This flexibility could be more important for the hexameric complexes that form the IL-6 and IL-11 receptor complexes compared with the trimeric IL-27, OSM, and LIF receptor complexes. Defective IL-11 and OSM signaling may contribute to skeletal abnormalities (Sims, 2016). Defective IL-6 signaling may explain dermatitis and lack of acute-phase response, as indicated by the low fibrinogen and normal C-reactive protein despite a neonatal stress response. Our patients did not survive long enough to develop a phenotype similar to IL-6R defects (Spencer et al., 2019). Loss of OSM signaling may explain the severe congenital thrombocytopenia in A-II-2. Genetic ablation of either OSM (Sato et al., 2014) or OSM receptor (Tanaka et al., 2003) in mice is associated with thrombocytopenia, and a first-time-in-human study of an anti-OSM monoclonal antibody (GSK2330811) was associated with thrombocytopenia (Reid et al., 2018). Patient C-II-3 survived infancy and developed additional immunologic symptoms of eczematoid dermatitis reminiscent of defective IL-6/GP130/STAT3 signaling, as observed in previous GP130 homozygous patients with missense variants in IL6ST (Schwerd et al., 2017; Shahin et al., 2019).

In summary, our findings explain a complex human syndrome due to the functional defect of a whole class of IL-6 family cytokines. The phenotypic overlap between the conditions with global IL6ST deficiency, cytokine selective IL6ST defects, and those caused by isolated cytokine defects caused by genetic or therapeutic targeting of the IL-6, IL-11, OSM, and LIF signaling axis is summarized in Fig. S2 G.

The study was approved by the Ethics Committee of the Karolinska Institutet (2014/983-31/1); the patients of family A were clinically examined at Lund University, Sweden. The molecular analyses for family A were performed at the Departments of Clinical Genetics, Lund University, and Karolinska University Hospital on genomic DNA from amniocytes of A-II-1 and from blood of the parents and A-II-2. The patients from Saudi Arabia (family B) were recruited with informed consent under an institutional review board–approved research protocol (KFSRHC RAC #2121053). The patients from the Netherlands (family C) initially participated in a French study, protocol number IRB0000388. Control individuals were recruited via the Oxford GI illnesses biobank (REC 16/YH/0247). Control fetal samples were obtained from Biobank at the Department of Laboratory Medicine, Division of Pathology, Karolinska Institutet, Huddinge, Sweden. Control amniocytes were obtained commercially from Coriell Institute.

Genomic DNA was extracted according to standard protocols from peripheral blood from A-II-2 and parents and from amniocytes of A-II-1. CytoScan HD Array was used (Thermo Fisher Scientific) according to manufacturer’s instructions. In a normal setting (parents and A-II-2), the resolution is ≈12 kb for deletions and ≈36 kb for duplications. In a prenatal setting (A-II-1), we only report >1 Mb for deletions and >2 Mb for duplications.

For family A, clinical WGS was performed as previously described (Stranneheim et al., 2014). A negative clinical gene-screening panel, targeting 377 known skeletal disease genes, was performed at Science for Life laboratory, and data were analyzed at the Department of Clinical Genetics, Stockholm, Sweden, for A-II-2. Next, family members were added to the same analysis: first parents, and then the DNA from the affected fetus (A-II-1). WGS was performed using HiSeq X (Illumina) with a TruSeq DNA PCR-free library preparation at Clinical Genomics (SciLifeLab, Stockholm, Sweden) to yield minimum 30× coverage. The variants were ranked according to expected disease-causing potential, their presence in coding exons ±20 base pairs intronic sequence, and a minor allele frequency of <0.005 according to public databases (1000 Genomes, dbSNP, and ExAC/gnomAD) and our local in-house database. Family analysis of parents and the affected patients revealed a short list of very rare variants (Table S1).

Exome sequencing in family B and positional mapping were performed as previously described (Monies et al., 2019). Exome sequencing in family C was performed at the genomic platform of the IMAGINE Institute (Paris, France) with the SureSelect Human All Exon kit (Agilent Technologies). Agilent SureSelect Human All Exon (V4) libraries were prepared from 3 µg of genomic DNA according to instructions of the manufacturer. Pooled barcoded exome libraries were sequenced using a HiSeq2500 (Illumina) to generate paired-end reads. After demultiplexing, sequences were mapped to the human genome reference (NCBI build37/hg 19 version) with Burrows-Wheeler Aligner (Li and Durbin, 2010). The mean depth of coverage obtained for each sample was ≥80×, with 95% of the exome covered at least 15×. Variant calling was performed with the Genome Analysis Toolkit (GATK; McKenna et al., 2010), Sequence Alignment/Map tools (Li et al., 2009), and Picard Tools. Single nucleotide variants were called with GATK UnifiedGenotyper, whereas indel calls were made with the GATK IndelGenotyper_v2. All variants with a read coverage ≤2× and a Phred-scaled quality of ≤20 were filtered out. Variants were annotated and filtered using an in-house annotation software system (Polyweb, unpublished). The analyses focused on nonsynonymous variants, splice variants, and coding indels. Variant pathogenicity was evaluated using the prediction algorithms SIFT (sorting intolerant from tolerant; cutoff ≤0.05), PolyPhen2 (HumVar scores, cutoff ≥0.447), and MutationTaster (cutoff: qualitative prediction as pathogenic). The variant frequency in control datasets was assessed, including the ExAC database, dbSNP129, 1000 Genomes, ClinVar, HGMD, and in-house exome data. IL6ST variants were confirmed by Sanger sequencing, and segregation was verified.

The following primers were used. For family A: IL6ST forward primer, 5′-GTT​GTG​TTG​AGC​CAA​GAT​CG-3′, and IL6ST reverse primer, 5′-AAA​ACC​AGA​ACC​AGT​TCA​CAT​C-3′. For family B: IL6ST forward primer, 5′-CAC​CTG​TAA​TCC​CAG​CCA​CT-3′ and IL6ST reverse primer, 5′-GCA​TCT​GGC​CTC​CAT​GAT​AA-3′. For family C: IL6ST forward primer, 5′-TGT​AAA​ACG​ACG​GCC​AGT​GTG​GGA​CCA​ACT​TCC​TGT​TG-3′, and IL6ST reverse primer, 5′-CAG​GAA​ACA​GCT​ATG​ACC​AAC​CAT​AAG​GCA​AAT​TCT​CAG​G-3′. The variants have been reported to ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/): SCV000925665 for c.841C>T and SCV000924636 for c.1699+4A>G.

The following genetic databases were used: gnomAD (https://gnomad.broadinstitute.org); 1000 Genomes (https://www.internationalgenome.org); and ClinVar.

Splice variant analysis was performed using MaxEnt (http://genes.mit.edu/burgelab/maxent/Xmaxentscan_scoreseq.html) and the transcript-inferred pathogenicity score tool (http://trap-score.org/Search?query=5-55247753-T-C).

Healthy amniotic fluid–derived amniocytes were derived from Coriell Institute (GM00472 and GM00957) and patients A-II-1 and A-II-2 were obtained from the Department of Clinical Genetics, Lund University Hospital, Sweden. Amniocytes were maintained in AmnioMAX-II complete medium (Thermo Fisher Scientific) at a 1:2 passage ratio. EBV-LCLs from the patient carrying c.1699+4A>G variant and three healthy controls were generated from peripheral blood mononuclear cells with supernatant from the EBV-producing marmoset cell line B95-8 according to standard protocols. EBV-LCLs were maintained in RPMI 1640 (Sigma-Aldrich), whereas HEK293 and GP130-deficient HEK293 cell lines were cultured in DMEM (Sigma-Aldrich), both supplemented with 10% FCS.

Staining of surface GP130 was performed using anti-GP130-biotin (AN-G30; eBioscience) followed by staining with streptavidin-PE CF594, and expression was determined by flow cytometry. Cells were stained with Fixable Viability Dye eFluor 780 (eBioscience), and dead cells were excluded.

Cells were stimulated with indicated concentrations of recombinant human IL-6, OSM, LIF, IFN-α, IL-11, IL-27, and IL-21 (Miltenyi Biotec, R&D Systems, and PeproTech).

GP130 reconstitution in amniocytes was performed essentially as previously described (Schwerd et al., 2017). Sequence encoding WT IL6ST was cloned into the pcDNA3.1(+) vector without C- or N-terminal tags (GenScript). The WT IL6ST gene was custom cloned into pLENTI7.3V5-DEST (Geneart; Thermo Fisher Scientific). Lipofection on HEK293T cells was performed using plasmid encoding WT IL6ST and Lipofectamine 2000 (Thermo Fisher Scientific). Lentiviral supernatant was generated from HEK293T cells and filtered using a 0.45-µm filter before adding to amniocytes on 24-well plates. The transduction efficiency of amniocytes was >40% as confirmed by GFP fluorescence measurements by flow cytometry.

GP130-deficient HEK293 cell line was generated using CRISPR/Cas9 as previously described (Schwerd et al., 2017). GP130-deficient HEK293 cells were transiently transfected with plasmids encoding empty vector, WT, or mutant p.Arg281* or c.1699+4A>G GP130 (mutagenesis created by GenScript; pcDNA3.1(+)) using Lipofectamine 2000 (Thermo Fisher Scientific). To fully access IL-6 and IL-11 signaling, cells were cotransfected with IL-6Rα and IL-11Rα expression vectors. Cotransfection with GFP-encoding plasmid allowed gating on successfully transfected cells. After 24 h, cells were serum starved for 2 h before stimulation with indicated cytokines at 100 ng/ml. Phosflow STAT3 assay was performed as described below.

Phosphorylation of STAT3 (pSTAT3) was determined by Phosflow. Analysis of primary lymphocytes and monocytes was performed using peripheral blood. Peripheral blood mononuclear cells were separated by Lymphoprep and density gradient. Cells were subsequently stained with anti-CD3-FITC (clone: UCHT1; BioLegend), anti-CD4-BV605 (clone: OKT4; BioLegend), anti-CD8-PE CF594 (clone:RPA-T8; BD Biosciences), anti-CD14-BV650 (clone: M5E2; BioLegend), and anti-CD19-BV711 (clone: SJ25C1; BD Biosciences) for 20 min on ice. 5 min through surface staining, cells were stimulated for 15 min with 100 ng/ml IL-6 before fixation. For the analysis of amniocytes, cells were plated in 96-well plates and serum starved overnight to reduce background. Next day, cells were washed three times with serum-free DMEM (Sigma-Aldrich) and further serum-starved for 2 h. Cells were subsequently stained with Fixable Viability Dye eFluor 780 (eBioscience) and stimulated with 100 ng/ml OSM, 100 ng/ml IL-11, and 100 ng/ml LIF for 15 min before detachment with trypsin-EDTA solution and fixation in prewarmed Cytofix buffer. Levels of pSTAT3 in EBV-LCLs were determined in serum-starved cells, followed by stimulation with 50 ng/ml IL-6 or IL-21 before fixation. After fixation, amniocytes, HEK293 cell lines and EBV-LCLs were washed and permeabilized with ice-cold Perm Buffer III (BD Biosciences) and stained with anti-STAT3 (pY705)-Alexa Fluor 647 (clone 4/P-STAT3) at room temperature. LSRII or LSR Fortessa was used to acquire samples, and data were analyzed using Flowjo software (v.10.4.0).

Total RNA was extracted from EBV-LCLs, amniocytes, or transfected GP130-deficient HEK293 cells using the RNeasy mini kit (Qiagen) accordingly to the manufacturer’s instructions, and cDNA was synthesized using the High-Capacity cDNA Reverse Transcription kit (Applied Biosystems). Gene expression was determined by Taqman assays from Applied Biosystems: SOCS3 (Hs02330328_s1), IL6ST exons 4–5 (Hs01006736_g1), and exons 12–13 (Hs01006729_g1). Quantitative RT-PCR was performed in triplicate using a Bio-Rad machine. Changes in gene expression were normalized using an endogenous control RPLPO (Hs00420895_gH), and data were expressed in accordance with the 2^−ΔCt method.

PCR was performed using a standard protocol for Phusion High Fidelity DNA Polymerase (New England BioLabs). The designed primer pair with a combination of IL6ST_cDNA_exon11_forward (5′-TGC​TTT​GGG​TGG​AAT​GGA​CT-3′) and IL6ST_cDNA_exon 14–15_reverse (5′-TCC​TTG​AGC​AAA​CTT​TGG​GGT-3′) were used for PCR (Thermo Fisher Scientific). Thermocycling conditions were denaturation at 98°C for 3 min, 35 cycles of repeated 98°C for 10 s, 60°C for 30 s, and 72°C for 30 s, followed by final extension at 72°C for 10 min. The DNA fragments were subsequently run on a 1.5% agarose gel and purified using a QIAquick Gel Extraction Kit according to manufacturer’s instructions (Qiagen) before being Sanger sequenced (Source Bioscience).

cDNA derived from C-II-3 was used to perform nested PCR using a number of IL6ST primer pairs. Primer pair 1: exon 11_forward (5′-TGC​TTT​GGG​TGG​AAT​GGA​CT-3′) and exon 14–15_reverse (5′-TCC​TTG​AGC​AAA​CTT​TGG​GGT-3′). Primer pair 2: exon 12_forward (5′-ATG​CTG​ATG​GAC​CAG​GAA​GC-3′) and exon 16_reverse (5′-GAG​TGT​GAG​GTG​ACC​ACT​GG-3′). Primer pair 3: exon 12_forward (5′-CAG​GAA​GCC​CTG​AAT​CCA​TA-3′) and exon 14_reverse (5′-TTC​TGT​GTG​GGA​AGA​ATC​CA-3′). Primer pair 4: exon 11_forward (5′-TGC​TTT​GGG​TGG​AAT​GGA​CT-3′) and exon 16_reverse (5′-GAG​TGT​GAG​GTG​ACC​ACT​GG-3′). Thermocycling conditions for both rounds of nested RT-PCR were denaturation at 95°C for 5 min, 35 cycles of repeated 95°C for 45 s, 58°C for 30 s, and 72°C for 45 s, followed by final extension at 72°C for 10 min. The DNA fragments were subsequently run on a 1% agarose gel.

DNA was isolated from EBV-LCLs using the DNeasy Blood & Tissue kit (Qiagen) according to the manufacturer’s instructions. The variants in IL6ST (exons 4, 8, 13, and 17) and genes adjacent to IL6ST (ANKRD55 and IL31RA) were amplified with PCR primers, and multiple ligation-dependent probe amplification was performed according to the manufacturer’s protocol (MRC Holland). The sequences of the probes are as follows: GABRA4_fam-pilot_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​ACA​GCC​TGT​TGT​CAT​AAC​CAT​CG-3′; GABRA4_fam-pilot_B, [Phos] 5′-AGC​AAA​CTG​TCC​AGG​ATG​CGT​CTA​GAT​TGG​ATC​TTG​CTG​GCA​C-3′; IL6ST_ex4_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGT​GAA​CTT​CTA​GAT​CCA​TGT-3′; IL6ST_ex4_B, [Phos] 5′-GGT​TAT​ATC​AGT​CCT​GAA​TCT​CCA​GTC​TAG​ATT​GGA​TCT​TGC​TGG​CAC-3′. ANKRD55_downst_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGA​CTT​GGA​ATT​GCA​GGA​AGC​ACT​CG-3′; ANKRD55_downst_B, [Phos] 5′-GTT​TGT​CCG​CAG​AGG​TTC​TCC​AGT​CTA​GAT​TGG​ATC​TTG​CTG​GCA​C-3′. IL6ST_ex8_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​ACT​GTC​CAA​GAC​CTT​AAA​CCT​TTT​ACA-3′; IL6ST_ex8_B, [Phos] 5′-GAA​TAT​GTG​TTT​AGG​ATT​CGC​TGT​ATG​TCT​AGA​TTG​GAT​CTT​GCT​GGC​AC-3′. IL6ST_ex17_3UTR_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGC​TGC​GAC​TGA​TGA​AGG​CAT​GCC​TAA​AAG-3′; IL6ST_ex17_3UTR_B, [Phos] 5′-TTA​CTT​ACC​ACA​GAC​TGT​ACG​GCA​AGG​CTC​TAG​ATT​GGA​TCT​TGC​TGG​CAC-3′. IL31RA_upstr_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGC​TAA​ACC​TGA​AGG​AGT​CTG​ATG​ACT​CTG​T-3′; IL31RA_upstr_B, [Phos] 5′-GAA​CAC​AGA​AGA​CAG​GAT​CTT​AAA​ACC​ATG​TCT​AGA​TTG​GAT​CTT​GCT​GGC​AC-3′. Stern PCLN13q_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGA​CAC​AAG​GGT​GTA​AAA​TGC​ACG-3′; Stern PCLN13q_B, [Phos] 5′-TTT​CAG​GGT​GTG​TTT​GCA​TAT​GAT​TTA​ATC​AAT​CAG​TAT​GTC​TAG​ATT​GGA​TCT​TGC​TGG​CAC-3′. IL6ST_ex13_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​ACA​CCT​TCC​AAA​GGA​CCT​ACT​GTT​CGG​ACA​AAA​AAA​GTA​G-3′; IL6ST_ex13_B, [Phos] 5′-GGA​AAA​ACG​AAG​CTG​TCT​TAG​AGT​GGG​ACC​AAC​TTC​CTC​TAG​ATT​GGA​TCT​TGC​TGG​CAC-3′. RB1ex23_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGT​CAC​CAA​TAC​CTC​ACA​TTC​CTC​GAA​GCC​CTT​ACA​AGT​TTC​CT-3′; RB1ex23_B, [Phos] 5′-AGT​TCA​CCC​TTA​CGG​ATT​CCT​GGA​GGG​AAC​ATC​TAT​ATT​TCA​CCT​CTA​GAT​TGG​ATC​TTG​CTG​GCA​C-3′. MRPL41_e1_132_A, 5′-GGG​TTC​CCT​AAG​GGT​TGG​AGA​CCC​TGA​CAA​CCT​GGA​AAA​GTA​CGG​CTT​CGA​GCC​CAC​ACA​GGA​G-3′; MRPL41_e1_132_B, [Phos] 5′-GGA​AAG​CTC​TTC​CAG​CTC​TAC​CCC​AGG​AAC​TTC​CTG​CGC​TAG​CTG​TCT​AGA​TTG​GAT​CTT​GCT​GGC​AC-3′.

3-µm sections from paraffin-embedded tissue from epiphyses of lower femur were stained with H&E and examined by a pediatric histopathologist. Variations in the intensities of H&E staining can be attributed to protocols being conducted at different centers. Safranin O/Fast green staining was performed as previously described (Newton et al., 2015). The primary spongiosa was defined as being 50 µm from the growth plate.

Slides were dewaxed at 54°C for 1 h and rehydrated from xylene through an ethanol gradient to distilled water. Antigen retrieval was performed by submerging slides in Target retrieval solution (DAKO) and heating to 2 psig (103.6°C) in a pressure cooker (All American 1925x). After rinsing in distilled water, a peroxidase quench was performed using 3% hydrogen peroxide (Merck) in methanol. The slides were rinsed in distilled water and blocked with 3% normal horse serum (Vector Laboratories) in PBST (PBS containing 0.1% Tween 20 [Merck]) for 1 h. Slides were incubated overnight with primary antibodies to IL6ST (Sigma-Aldrich; HPA010558, 1:50, gently rocking), osterix (Abcam; ab22552, 1:50, gently rocking), DMP-1 (Abcam; ab103203, 1:50, at 2–8°C), or Cathepsin K (a gift from Goran Andersson [Karolinska Institutet, Stockholm, Sweden], previously developed and validated; Kamiya et al., 1998; Ljusberg et al., 2005; 1:100, at 2–8°C). Slides were then rinsed with PBST and incubated with biotin-conjugated goat anti-rabbit antibodies (Jackson ImmunoResearch Laboratories; 1:400), with gentle rocking for 1 h. Rinsing was performed with PBST, and slides were incubated with ABC reagent (Vector Laboratories), with gentle rocking for 1 h. After PBST rinsing, slides were exposed to TSA solution (Perkin Elmer) and gently rocked for 15 min, and rinsed again with PBST before incubation with ABC reagents (Vector laboratories) with gentle rocking for a further 1 h. Development with DAB solution (DAKO) for test and control slides was performed for the same duration. After dehydration and mounting with Pertex (Histolab), sections were imaged using an Axioskop 2 Plus microscope with Axiovision software (Zeiss).

A structural model of GP130 nonfunctional variant was built based on the crystal structure and EM images of the full ectodomain and Fibronectin III domains of GP130 (PDB accession nos. 3L5H, 3L5I, 3L5J, 3E0G, 1P9M, and 1PVH). Rigid body structural alignments were performed using the crystal structure of the IL6-IL6Rα-GP130 domain D1–D3 hexameric complex as a template and aligning the full ectodomain structure. Then the peptide sequence that corresponds to exon 13 was deleted. GP130 adopts a C-shaped conformation with an Ig-like domain (D1) at the top and three membrane-proximal FNIII domains (D4–D6) supporting the two membrane-distal ligand-binding regions (D2 and D3). Sequence alignment and structural modeling were performed using BioLuminate and Maestro software packages (2019-1 release, Schrödinger). Structures were prepared and minimized following Protein Preparation Wizard workflow in Maestro with the OPLS3e force field. All protein structures were modeled, analyzed, and rendered in Maestro (release 2019-1, v.11.9.011, MMshare Version 4.5.011, Platform Linux-x86_64). Coordinate files were obtained from PDB (accession no. 1P9M for IL-6/IL-6Rα/GP130; accession no. 3L5H for GP130 D1-D6).

Results were analyzed using Prism 8 (GraphPad Software). Significance of differences was determined by Mann–Whitney U test, and P values < 0.05 were considered significant. The bone histology values presented in the text are mean ± SEM (to three significant figures) of values collected from six discrete visual fields, analyzed for each patient.

Fig. S1 shows the bone phenotype, MLPA analysis of the IL6ST gene, RNA-seq expression data of multiple cytokine receptor units in control amniocytes, the parental analysis of GP130 and pSTAT3 response, and PCR analysis. Fig. S2 shows that the complete loss of GP130-dependent IL-6, IL-11, IL-27, OSM, and LIF signaling in p.Arg281* and c.1699+4A>G variant is not due to delayed kinetics and a schematic summary of GP130 deficiency, and the relationship between loss of cytokine signaling and clinical symptoms. Table S1 lists the rare sequence variants identified using whole genome variant analysis in family A (quadro) and whole exome variant analysis in family B and C (trios).

We thank the families and blood donors for participation in this study. Whole exome sequencing analysis for family B was done at the Imagine Institute (France). We acknowledge GeneMatcher Exchange for enabling the collaboration and Saskia van der Velde-Visser and Dr. Petr Jira for their support. We thank Anette Niklasson and Annica Westlund at Fetal Pathology laboratory at Karolinska University Hospital for help with sectioning the samples for histological analysis.

We acknowledge the contribution of the Oxford Gastrointestinal Biobank, which is supported by the National Institute for Health Research Oxford Biomedical Research Centre. This project was supported through the regional agreement on medical training and clinical research between Stockholm County Council and Karolinska Institutet (ALF20150143 and 20180131), by grants from the Swedish Research Council (G. Grigelioniene and A.S. Chagin), H.K.H. Kronprinsessan Lovisas Förening för Barnasjukvård/Stiftelsen Axel Tielmans minnesfond, Stiftelsen Sällskapet Barnavård (A. Hammarsjö and P.T. Newton), Stiftelsen Promobilia (G. Grigelioniene), and Stiftelsen Frimurare Barnhuset i Stockholm (G. Grigelioniene and P.T. Newton). Bone analysis was supported by a grant from the Russian Science Foundation (19-15-00241). Computational infrastructure and assistance in massively parallel sequencing were provided from Clinical Genomics, Science for Life Laboratory. H.H. Uhlig is supported by the Biomedical Research Centre Oxford.

Author contributions: Y.-H. Chen and G. Grigelioniene contributed equally and should be considered shared first authors. F.S. Alkuraya, A. Laurence, and V. Cormier-Daire contributed equally. G. Grigelioniene, Y.-H. Chen, J. Brodszki, J. Gullander, K. Jedrycha, M. Elving, D. Gisselsson, A. Hammarsjö, H.S. Alsaif, W.I.Y. Kurdi, F. Abdulwahab, F.S. Alkuraya, S. Bacrot, C. Huber, B. Hamel, B. Gürtl-Lackner, and V. Cormier-Daire performed genetic and clinical evaluations. Y.-H. Chen, A. Laurence, D. Aschenbrenner, P.T. Newton, D. Batkovskyte, and H.H. Uhlig performed functional analysis. G. Nishimura is responsible for interpretation of radiographical diagnosis for family 1. N Papadogiannakis., A.S. Chagin, and P.T. Newton performed and interpreted histology and immunohistochemistry analyses. V. Shanmugasundaram performed structural remodeling. L. Devey developed the hypothesis and discussed the experimental data. G. Grigelioniene, V. Cormier-Daire, and H.H. Uhlig directed research. G. Grigelioniene, Y.-H. Chen, J. Gullander, M. Elving, P.T. Newton, and H.H. Uhlig wrote the manuscript. All authors have revised and approved the manuscript.