Defective Gp130-Mediated Signal Transducer and Activator of Transcription (Stat) Signaling Results in Degenerative Joint Disease, Gastrointestinal Ulceration, and Failure of Uterine Implantation

The activities of a family of cytokines that comprises IL-6, IL-11, leukemia inhibitory factor (LIF), oncostatin M (OsM), ciliary neurotrophic factor (CNTF), and cardiotrophin-1 are mediated by the common signal transducing receptor β chain gp130 1,2. LIF/IL-6 cytokines exert multiple activities on target organs such as liver, heart, immune, hematopoietic, and nervous systems that are pivotal during embryogenesis, morphogenesis, and inflammation. Aberrant expression of LIF/IL-6 family cytokines has been associated with autoimmune disease, septic shock, and neoplasia 2,3. Somewhat surprisingly, mice with engineered null mutations for individual LIF/IL-6 family ligands or receptor α subunits develop normally and exhibit rather mild phenotypes. IL-6–deficient (IL-6^−/−) mice, for instance, showed impaired acute phase protein production and decreased high affinity immune responses 4,5, while mice deficient for LIF or the IL-11 receptor α subunit (IL-11Rα1) were unable to support pregnancies due to defective blastocyst implantation 6,7 and decidua formation 8,9, respectively. This finding has been attributed to partial functional redundancy between family members due to overlapping expression patterns and the sharing of common receptor β chains. Predictably, receptor β chain deficiency resulted in more severe and global effects. Depending on the genetic background, gp130^−/− mice died in utero or shortly after birth due to myocardial hypoplasia and reduced hemopoiesis in the fetal liver 11,12, while the LIF receptor β chain (LIF-Rβ)^−/− mice died perinatally due to degeneration of motor neurons and astrocytes 10.

The LIF/IL-6 family of cytokines recruits gp130 according to two principal mechanisms. In the case of IL-6 and IL-11, signaling is initiated by binding to ligand-specific receptor α subunits followed by recruitment and homodimerization of two gp130 receptor β chains. In contrast, binding of LIF, OsM, CNTF, or cardiotrophin-1 induces the formation of β chain heterodimers consisting of gp130 and one of the structurally and functionally related β chains, either the LIF-Rβ or OsM receptor 1,2. Thus, IL-6 and IL-11 are totally dependent on gp130 for signal transduction, while signals emanating from LIF, OsM, CNTF, and cardiotrophin-1 may be transduced, at least in part, by LIF-Rβ or OsM receptor. The cytoplasmic domains of all three β chains are characterized by conserved modular structures with each module thought to connect to a particular signal transduction pathway. First, the membrane-proximal homology region associates with cytoplasmic Jak kinases, which, upon receptor dimerization, phosphorylate tyrosine (Y) residues within the β chains 1. Second, a YxxV-phosphotyrosine motif (where Y is followed by any two amino acids and valine [V]) at positions Y₇₅₇ and Y₇₆₉ in mouse gp130 13 and LIF-Rβ 14, respectively, is required for recruitment of the phosphatase src homology 2 domain–bearing protein tyrosine phosphatase (SHP)-2, subsequent activation of the Erk mitogen-activated protein kinase (MAPK) and transduction of mitogenic signals 15. Third, the membrane distal portion of the β chains, which contain several YxxQ-phosphotyrosine motifs, with a glutamine (Q) in position +3 relative to Y_{765/812/904/914} of mouse gp130 and Y_{976/996/1,023} of mouse LIF-Rβ, mediates binding and activation of the latent transcription factors signal transducer and activator of transcription (STAT)1 and STAT3 16. This region, and in particular the YxxQ motifs, is necessary and sufficient to control differentiation and apoptosis of cells in vitro and to induce acute phase proteins in cultured hepatocytes 17,18,19,20. Moreover, the duration and intensity of the signals emanating from gp130 and LIF-Rβ appear to be limited by transcriptional induction of negative regulatory proteins of the SOCS family which impair Jak-mediated substrate phosphorylation 21,22.

To identify the biological responses to LIF/IL-6 cytokines specifically mediated by STATs in vivo, we generated a gp130 “knock-in” mouse containing a COOH-terminal truncation mutation in gp130 (gp130^ΔSTAT) which deleted all STAT binding sites in a manner likely to be functionally equivalent to phenylalanine substitutions (Y→F) in all YxxQ motifs 17,18,19.

A mouse genomic 129/Sv phage library was screened with a cDNA fragment encoding the cytosolic domain of gp130, yielding a 14-kb genomic clone that contained the final two coding exons encompassing amino acids 645–671 and 672–917 13 and a 5.4 kb of 3′ flanking sequence. PCR-mediated mutagenesis was used to introduce the gp130^ΔSTAT truncation mutation by introducing the Y₇₆₅F (TAC to TTC), Q₇₆₈A (CAG to GCC), and V₇₆₉stop (GTC to TAG) substitutions which were encoded by the oligonucleotide 5′-TTC.AGG.CAC.GCC.TAG.GAC-3′. The newly introduced stop codon was part of an AvrII site (underlined) and was fused to the fourth nucleotide (bold A) of the 3′ untranslated region of the endogenous g130 gene. A cDNA encoding neomycin resistance preceded by an internal ribosomal entry site (IRES) was introduced into the AvrII site as a XbaI/NheI fragment, thereby generating gp130-neo dicistronic RNA. The targeting vector also contained 0.8 kb and 4.3 kb of sequence homologous to the final intron and the 3′ flanking sequence, respectively.

From 25 × 10⁶ W9.5 embryonic stem (ES) cells (129/Sv) electroporated with the SalI linearized targeting construct, 12/120 neomycin resistant colonies were correctly targeted as judged by the hybridization pattern of BglII-digested DNA hybridized with a genomic fragment mapping upstream of the 5′ end of the targeting vector. Allele-specific amplification with PCR primers P1 and P2 was used to confirm the nucleotide sequence of the mutated exon in the targeted allele (see Fig. 1). Two recombinant ES cell lines (129/Sv) were injected into C57BL/6 blastocysts and the germline chimeras were mated with C57BL/6 females. Colonies homozygous for the mutation (gp130^{ΔSTAT/ΔSTAT}) were established and subsequently maintained by breeding gp130^{ΔSTAT/ΔSTAT} males with gp130^ΔSTAT/wt females, and their offspring were routinely genotyped at weaning age by PCR analysis using primers P1 and P2 of DNA prepared from tail biopsies.

Primary cultures of synovial fibroblasts were established by sequential collagenase digestion (collagenase type II CLS-4, Worthington; 165 U/ml in DMEM) of clinically normal knee joints taken from gp130^{ΔSTAT/ΔSTAT} mice and their gp130^wt/wt littermate controls as well as compound homozygous SOCS-1^−/−, INF-γ^2/− mice 23, and SOCS-1^+/+; INF-γ^2/− 24 controls were established as described previously 25. Cell populations with the highest intensity of vascular cell adhesion molecule 1 staining, a marker for synovial fibroblasts 26, were maintained in DMEM supplemented with 20% FCS and assayed at passage 5.

The mitogenic activity of mouse LIF (ESGRO; AMRAD) or recombinant human IL-6 on synovial cells was assessed by culturing 2 × 10⁴ cells per well in 96-well plates. After serum starvation (0.5% FCS) for 24 h, quadruplicate cultures were stimulated for 24 h with the indicated concentration of LIF or IL-6 and pulsed for the last 4 h with 0.5 μCi per well methyl-[³H]thymidine (Dupont NEN) before harvesting and measuring incorporated radioactivity.

For transient transfection assays, 5 × 10⁶ cells were coelectroporated (500 μFarad; 270V) with 20 μg of a luciferase reporter construct containing the MAPK responsive region (−404 to +41) of the human c-fos promoter (27), 20 μg of the Srα-β–galactosidase plasmid and increasing concentrations of a Flag epitope-tagged SOCS-1 expression construct (pSOCS-1; reference 28). Cells in 24-multiwell plates were maintained in 0.5% FCS for 48 h in the presence of IL-6 plus sIL-6R (500 ng/ml each), saline, or FCS (10%) before lysis with 40 μl of reporter lysis buffer (Promega). Lysates were assayed for luciferase activity which was normalized according to the β-galactosidase activity. Expression levels from the transfected SOCS-1 plasmid were monitored by anti-Flag immunoblotting 28.

Serum-starved cultures of synovial fibroblasts were stimulated with IL-6 plus sIL-6R (both at 500 ng/ml) for the indicated period of time. 1 mg of cell lysate was immunoprecipitated with an antiserum against Jak-2 (UBI) and used for in vitro phosphorylation experiments in the presence of γ-[³²P]ATP as described previously 19. SHP-2 immunoprecipitates were resolved by SDS-PAGE and blots were probed with antiphosphotyrosine (UBI) and SHP-2 (Santa Cruz Biotechnology, Inc.) antibodies to assess protein loading 19. For detection of active Erk MAPK, 50 μg of total cell lysate were resolved by SDS-PAGE and transferred proteins were reacted with an antibody that recognizes phosphorylated Erk1/2 (New England Biolabs, Inc.). The level of Erk protein was subsequently visualized by reaction with anti-Erk1/2 antibodies (Santa Cruz Biotechnology, Inc.). STAT3 phosphorylation in tissues was monitored after STAT3 immunoprecipitation (Santa Cruz Biotechnology, Inc.) of 2 mg of tissue lysate obtained from mice 20 min after a single intravenous injection of either 5 μg IL-6, LIF, IL-11, or saline and detection with antiphosphotyrosine and anti-STAT3 (Transduction Laboratories) antibodies, respectively 19.

Poly(A)⁺ RNA was prepared from livers of mice 40 min (for SOCS induction) or 4 h (for acute phase protein) after a single intravenous injection of 5 μg of LIF, IL-6, or saline according to standard protocols. Poly(A)⁺ RNA was also isolated from cultures of primary synovial cells stimulated with IL-6 plus sIL-6R (both at 500 ng/ml) for 45 and 180 min, respectively. 3 μg of poly(A)⁺ RNA were hybridized with cDNAs encoding either cytokine-inducible src homology 2 protein (CIS), SOCS-1, SOCS-2, or SOCS-3, gp130 or neomycin, and serum amyloid A or haptoglobin. All RNA blots were finally hybridized for glyceraldehyde 3-phosphate dehydrogenase (GAPDH) to assess the amounts of RNA analyzed.

For each limb, clinical scores were allocated for erythema, swelling, and flexibility of joints twice per week for up to 25 wk as described previously 29. The scoring system for each limb was as follows: 0, normal; 1, moderately affected; and 2, severely affected. Soft tissues were fixed in formalin for 24 h, while joints were fixed in 10% formalin for 48 h, decalcified, and embedded in paraffin 29. Sections were stained with hematoxylin and eosin. Histopathological features associated with the large appendicular joints (shoulder, elbow, wrist, hip, knee, ankle) were scored as 0, normal; 1, minor synovial changes without degenerative alterations or cartilage metaplasia; 2, mild degenerative joint disease with cartilage deposits with or without synovial hyperplasia; and 3, severe degenerative joint disease with marked cartilage metaplasia with or without synovial hyperplasia. All clinical and histological assessments were performed blind.

Mice were immunized by injecting ∼1 μl of OVA, (1 mg/ml in PBS, pH 7.4, emulsified with Freund's Complete Adjuvant) into each Peyer's patch 30. Mice were boosted intraduodenally 14 d later with OVA and euthanized 5 d after the secondary challenge. The proteinacious fraction of fecal pellets, obtained from intestinal washes, was prepared as described previously 30 and the small intestine was collected, fixed in cold ethanol, embedded in paraffin, and sectioned (5 μm) for immunohistochemistry of OVA-containing cells 5. OVA-specific ELISA assays were carried out for IgA, IgG, and IgM using Ig-specific biotin conjugated goat anti–mouse antibodies (Zymed Laboratories). Total isotype-specific Ig levels in sera of unchallenged mice were determined by ELISA using commercially available kits according to the manufacturer's recommendations (Bethyl Laboratories, Inc.).

Blastocysts were flushed from uteri on day 4 of gestation (day 1 is day of plug) and transferred to day 3 pseudopregnant recipients as described previously 6. All offspring were genotyped by Southern blotting of DNA from tail biopsies.

To inactivate the four YxxQ STAT-binding sites in gp130 at positions Y_{765/812/904/914}, we replaced exon 18, encoding amino acids 672–917, with a version truncated at the COOH-terminal side of amino acid 768 and containing the Y₇₆₅F and Q₇₆₈A substitutions. After homologous recombination in ES cells, the presence of the expected mutations was confirmed by sequencing allele-specific PCR amplicons generated from the targeted gp130 allele with primers P1 and P2 (Fig. 1 A). Transcription of the mutated gp130 allele generated a dicistronic gp130-neo mRNA species ∼1.4 kb larger than wild-type (wt) gp130 mRNA which encoded a truncated gp130^ΔSTAT protein (Fig. 1 C). Interbreeding of gp130^ΔSTAT/wt mice yielded homozygous offspring close to the expected Mendelian ratio (Fig. 1 B and Table). However, subsequent matings of gp130^ΔSTAT/wt females with gp130^{ΔSTAT/ΔSTAT} males yielded significantly fewer gp130^{ΔSTAT/ΔSTAT} than gp130^ΔSTAT/wt offspring at weaning age (Table), although mice of the two genotypes were recovered at equal ratios during embryonic development (E18, and data not shown). At birth, gp130^{ΔSTAT/ΔSTAT} mice appeared superficially healthy, but in comparison to gp130^wt/wt and gp130^ΔSTAT/wt littermates, they showed a 20–30% reduction in body weight (1.93 g ± 0.38 g vs. 2.93 g ± 0.12 g; P < 0.05, n = 20) and trunk length, which persisted throughout adulthood. The average life span of gp130^{ΔSTAT/ΔSTAT} mice was reduced (6 ± 1 mo for males, 8 ± 2 mo for females) compared with gp130^ΔSTAT/wt littermates (15 ± 2 mo for both sexes) which were indistinguishable from gp130^wt/wt mice in all aspects analyzed.

While gp130^{ΔSTAT/ΔSTAT} male mice sired litters of equal size to those of gp130^ΔSTAT/wt or wt males, female gp130^{ΔSTAT/ΔSTAT} mice mated with males of any genotype showed no external signs of pregnancy (Table). Histological analysis of uteri from gp130^{ΔSTAT/ΔSTAT} mice 5.5 d after coitum revealed the presence of blastocysts that showed no overt signs of implantation at a time when embryos in uteri of gp130^ΔSTAT/wt and gp130^wt/wt mice had implanted and were surrounded by the secondary decidua (Fig. 2 A). However, blastocysts of all genotypes recovered from gp130^{ΔSTAT/ΔSTAT} mice developed to term when surgically transferred to day 3 pseudopregnant wt recipient mice, while again no signs of pregnancy were observed when transferred to gp130^{ΔSTAT/ΔSTAT} recipient mice (Table). These results unequivocally show a maternal defect at the time of embryo implantation that is strikingly similar to the defect described in mice deficient for LIF 6,7 .

To determine whether impaired STAT signaling in the uteri of gp130^{ΔSTAT/ΔSTAT} mice was the likely reason for the failure of blastocyst implantation and subsequent decidua formation, we monitored STAT3 phosphorylation in the liver and uterus after single bolus injections of IL-6, IL-11, or LIF. In both tissues, STAT3 phosphorylation in gp130^{ΔSTAT/ΔSTAT} mice was markedly reduced in response to LIF and completely abrogated in response to IL-6 or IL-11, compared with identically treated heterozygous control mice (Fig. 2B and Fig. D). This observation was further pursued in liver by measuring the induction of hepatic class I acute phase genes which are regarded as classical STAT-mediated responses 20. Induction of serum amyloid A and haptoglobin mRNAs in gp130^{ΔSTAT/ΔSTAT} mice was clearly diminished in response to LIF and absent in response to IL-6 (Fig. 2 C). Collectively, these results suggest that the mutant receptor β chain heterodimers between gp130^ΔSTAT and LIF-Rβ (which retains the 3 YxxQ motifs in LIF-Rβ) induced by LIF, while less efficient than a wt gp130/LIF-Rβ complex containing 7 YxxQ motifs, retains some STAT-mediated signaling capacity. Meanwhile homodimers formed between two truncated gp130^ΔSTAT proteins in response to IL-6 and IL-11 failed to promote STAT-mediated signaling.

Under specific pathogen–free conditions, approximately half of all gp130^{ΔSTAT/ΔSTAT} mice of 4 mo of age showed occult traces of fecal blood and ulceration of the gastric pylorus (Fig. 3 a). We also observed ulceration of the anorectal region which was invariably associated with rectal prolapse. Notably, the ulceration at both sites was confined to the vicinity of the sphincters suggesting defective epithelial repair in response to sustained mucosal injury. Mutant mice kept under conventional conditions also developed caecal ulcers (41%; Fig. 3 b) and acute inflammation at additional mucosal sites including the conjunctivae (41%) and the nasolacrimal ducts (29%). Furthermore, in comparison to gp130^ΔSTAT/wt mice, gp130^{ΔSTAT/ΔSTAT} mice showed a much higher load of Giardia muris which is an endemic parasite in the conventional facility in which the mice were housed.

Since IL-6 is required for differentiation of B cells to high affinity IgA- and IgG-producing plasma cells 4,5,31, we sought to determine mucosal immune responses of gp130^{ΔSTAT/ΔSTAT} mice. In the absence of deliberate immunization and compared with heterozygous controls, mutant mice showed reduced levels of circulating IgA (gp130^{ΔSTAT/ΔSTAT}, 0.62 ± 0.05 mg/ml, n = 7 vs. gp130^ΔSTAT/wt mice, 0.85 ± 0.10 mg/ml, n = 6), IgG₁ (0.17 ± 0.07 mg/ml vs. 0.35 ± 0.11 mg/ml), IgG_2a (0.18 ± 0.13 mg/ml vs. 0.53 ± 0.12 mg/ml), and IgG_2b (0.38 ± 0.15 mg/ml vs. 0.87 ± 0.17 mg/ml). In contrast, the levels of circulating IgM were similar for mice of both genotypes. The capacity of gp130^{ΔSTAT/ΔSTAT} mice to mount a specific intestinal IgA response after local immunization of Peyer's patches with OVA was grossly impaired when assessed by OVA-specific Ig isotype level in fecal pellets and serum (Fig. 3 e). Similarly, the OVA-specific IgG response in the intestine was also reduced and there were significantly less anti-OVA–containing cells in the small intestine (Fig. 3c and Fig. d), mesenteric lymph nodes, and spleen of gp130^{ΔSTAT/ΔSTAT} mice (Fig. 3f) reminiscent of the findings reported previously in IL-6–deficient mice 30. Collectively, these data imply that gp130 transduces the IL-6–mediated final maturation of B cells to high affinity Ig-secreting plasma cells via its STAT-binding sites in the COOH-terminal domain.

Unexpectedly, swelling of one or more of the large weight-bearing joints was observed in 60% of adult gp130^{ΔSTAT/ΔSTAT} mice (n = 49, mean age 89 d ranging from 56–412 d), which resulted in flexion deformities of the hind limbs and reluctance of the mice to stand (Fig. 4, a and b). Joints of gp130^{ΔSTAT/ΔSTAT} mice were often affected bilaterally but only rarely showed signs of erythema. Systematic clinical assessment showed that all gp130^{ΔSTAT/ΔSTAT} mice showed stiffness of one or more limbs with restricted flexion and extension of the ankles as well as dorsiflexion at the wrist joints (Fig. 4 b to e). Notably, some degree of joint restriction was already evident in one or more limbs of P7-neonatal gp130^{ΔSTAT/ΔSTAT} mice but did not increase significantly with age and only rarely affected previously healthy joints (Table). By contrast, none of the gp130^ΔSTAT/wt control mice showed any clinical evidence of swelling or restricted joint movement at any age. Abnormal joints were also observed in gp130^{ΔSTAT/ΔSTAT} mice maintained under specific pathogen–free conditions, indicating that this pathology was unlikely to be secondary to infections at mucosal sites.

The joints of 49 clinically affected mice were examined histologically and graded as described in Materials and Methods. Of the 292 large appendicular joints examined (16 hips, 75 knees, 67 ankles, 40 shoulders, 50 elbows, 44 wrists), 76% were scored as 0 or 1 while 24% were graded as 2 or 3. Curiously, joints showing degenerative changes (grade 2 or 3) were twice as common in male (27.8%, n = 65) than female (15.1%, n = 8) gp130^{ΔSTAT/ΔSTAT} mice. Histologically, these joints showed cartilaginous nodules along the joint capsule extending from the edge of the articular cartilage. In the knee joints, these nodules invariably replaced the menisci and supporting connective tissue (Fig. 5, a and b) and occasionally the joint spaces contained dislodged and rounded cartilaginous nodules (Fig. 5 e). These changes were often accompanied by enlargement and deformation of the joints which occasionally affected the growth plate (Fig. 5 b). In addition, joints of older gp130^{ΔSTAT/ΔSTAT} mice often showed erosion of the articular surface and mild synovial hyperplasia associated with the tendon sheaths and nonswollen joints. Since histological examination of clinically affected joints of newborns revealed considerable structural variability, we were unable to unambiguously ascribe histological abnormalities to these joints. However, no primary irregularities were detected in the growth plate of young adult gp130^{ΔSTAT/ΔSTAT} mice. Occasionally, marked synovial hyperplasia was observed in the recesses of the joint cavities with expansion of the subsynovial connective tissue adjacent to the articular surfaces (Fig. 5c and Fig. d). However, foci of metaplastic cartilage were rarely observed within this de novo–formed connective tissues suggesting that the cartilaginous nodules were composed of newly formed cartilage and not disordered menisci or articular cartilage. Inflammatory cells, primarily lymphocytes and macrophages, were present in low numbers in the synovium and subsynovial tissues of pathological joints; however, neutrophils were usually absent. Collectively, these observations suggest that the joint pathology in gp130^{ΔSTAT/ΔSTAT} mice was primarily a result of synovial hyperplasia with progressive chronic inflammation and secondary cartilaginous metaplasia restricting joint movement and culminating in the degeneration of the articular cartilage of the major weight-bearing joints in older mice (Fig. 5f and Fig. g).

Since the LIF/IL-6 family cytokines have been implicated in promoting synovial hyperplasia 32,33, we investigated whether altered signaling through gp130^ΔSTAT receptors affected the mitogenic response of mutant synovial cells in vitro. Primary cultures of synovial fibroblasts, derived from clinically unaffected knees of gp130^{ΔSTAT/ΔSTAT} mice, showed an enhanced rate of [³H]thymidine incorporation in response to IL-6 or LIF when compared with cells obtained from gp130^wt/wt mice (Fig. 6 A). Furthermore, mutant cells responded to 100- to 1,000-fold lower concentrations of cytokines than wt cells and showed sustained IL-6–dependent activation of gp130-mediated signaling, as demonstrated by autophosphorylation of Jak-2, tyrosine phosphorylation of SHP-2, and phosphorylation of MAPK isoforms Erk1/2 (Fig. 6 B). These in vitro data indicate increased sensitivity and proliferative response of mutant synovial fibroblasts to IL-6 associated with sustained gp130 signaling through the SHP-2/ras/Erk pathway.

One possible mechanism accounting for sustained activation of the SHP-2/ras/Erk pathway in response to IL-6 could be that impaired STAT activation retarded the termination of gp130-mediated signaling. We hypothesized that this might result from impaired induction of SOCS/CIS family gene expression 21,22. As depicted in Fig. 7 A, the expression of CIS, SOCS-1, SOCS-2, and SOCS-3 mRNA were readily induced in the livers of wt mice after a single injection of either IL-6 or LIF. IL-6 treatment also induced CIS and SOCS-2 mRNA in the livers of gp130^{ΔSTAT/ΔSTAT} mice, however, we observed almost complete absence of SOCS-1 mRNA induction and reduced SOCS-3 mRNA induction. Notably, the induction of SOCS-1 mRNA was also markedly reduced in synovial cells from gp130^{ΔSTAT/ΔSTAT} mice compared with cells from wt mice (Fig. 7 B). These data support previous observations suggesting that IL-6–dependent activation of SOCS-1 and SOCS-3 genes is mediated, at least in part, by STATs 34. The reduced induction of SOCS-1 mRNA observed in the livers of LIF-treated gp130^{ΔSTAT/ΔSTAT} mice (Fig. 7 A) most likely resulted from diminished STAT signaling through the less effective gp130^ΔSTAT/LIF-Rβ heterodimers.

To explore the concept of enhanced gp130-mediated SHP-2/ras/Erk signaling in the absence of SOCS-1, we analyzed IL-6-signaling in SOCS-1–deficient synovial cells. These cells were derived from adult SOCS-1^−/− mice established on an IFN-γ–null background to rescue the perinatal lethality associated with SOCS-1 deficiency 23. As revealed in Fig. 8 A, compound mutant SOCS-1^−/−, IFN-γ^2/− cells, like gp130^{ΔSTAT/ΔSTAT} cells, showed enhanced [³H]thymidine incorporation in response to nanomolar concentrations of IL-6, compared with cells prepared from control mice of the corresponding genetic background (SOCS-1^+/+, IFN-γ^2/−, and gp130^ΔSTAT/wt, respectively). Compared with SOCS-1^+/+ IFN-γ^2/− control cells, we also observed sustained IL-6–mediated Erk1/2 phosphorylation in SOCS-1^−/− IFN-γ^2/− cells (Fig. 8 B). Thus, the enhanced proliferative responsiveness to IL-6 was also a feature of SOCS-1^−/− IFN-γ^2/− cells, however, this was less marked than in gp130^{ΔSTAT/ΔSTAT} cells.

If impaired or reduced induction of SOCS-1 (and SOCS-3) expression accounted for sustained activation of the Jak/SHP-2/ras/Erk cascade in response to IL-6 in gp130^{ΔSTAT/ΔSTAT} synovial cells, then overexpression of SOCS-1 should revert the hyperresponsive phenotype of these cells. To explore this notion, we exploited the fact that the minimal c-fos promoter is activated through the ras/Erk pathway 27. As expected, IL-6 stimulation of gp130^{ΔSTAT/ΔSTAT} cells, transiently transfected with a c-fos-luciferase reporter construct, yielded higher luciferase activity than IL-6 stimulation of wt cells (Fig. 8 C). Furthermore, increasing the cellular concentration of SOCS-1 protein, after cotransfection with an epitope tagged SOCS-1^Flag expression plasmid (pSOCS-1), reduced IL-6-responsiveness in gp130^{ΔSTAT/ΔSTAT} cells to levels observed in wt cells. Finally, a complete block of IL-6–mediated luciferase activity in gp130^{ΔSTAT/ΔSTAT} cells required greater amounts of transfected pSOCS-1 protein than in wt cells. Collectively, these observations are consistent with the notion that elevated levels of recombinant SOCS-1^Flag protein were required to terminate IL-6 signaling in gp130^{ΔSTAT/ΔSTAT} mutant cells to compensate for impaired induction of endogenous SOCS-1 and SOCS-3 expression. Since recombinant SOCS-1^Flag expression did not suppress c-fos-luciferase activity induced by serum, overexpression of SOCS-1^Flag was not directly interfering with activity of the c-fos-luciferase reporter plasmid.

Taken together, these data strongly suggest that the hyperresponsiveness of synovial cells from gp130^{ΔSTAT/ΔSTAT} mice arose from defective or impaired STAT-mediated SOCS induction which resulted in sustained activation of the SHP-2/ras/Erk pathway, leading to synovial hyperproliferation and cartilaginous metaplasia and ultimately the degeneration of the joint.

At the molecular level, the multiple and tissue-specific activities of cytokines are explained, in part, by activation of multiple signaling cascades emanating from discrete intracellular receptor domains which regulate distinct (sets of) target genes. While complete inactivation of multidomain receptors will result in phenotypes reflecting inactivation of all associated signaling pathways, the selective deletion of one particular domain provides the opportunity to associate a specific signaling cascade with a particular physiological or pathological process. Here we used a knock-in strategy to mimic the effect of a nonsense mutation that specifically deletes the COOH-terminal domain of gp130, along with its associated STAT activation sites, but still permits activation of the SHP-2/ras/Erk pathway 15,19. Consequently, the phenotype in gp130^{ΔSTAT/ΔSTAT} mice is largely a direct consequence of the absence (in the case of gp130^ΔSTAT homodimerization) or diminution (in the case of LIF-Rβ/gp130^ΔSTAT heterodimerization) of STAT-mediated cellular responses. Furthermore, we also identified the importance of STAT signals in promoting inhibitory feedback signals on gp130 and the associated SHP-2/ras/Erk pathway, thereby reinforcing a concept that the two pathways are under reciprocal negative control 35.

The phenotypes of mice with gp130 impairment mutations affecting either the STAT (gp130^{ΔSTAT/ΔSTAT} mice reported here) or the SHP-2/ras/Erk signaling cascade (gp130^757F/757F mice; reference 35) are dramatically different from the myocardial hypoplasia and associated embryonic lethality observed in gp130^−/− mice 11. Similarly, no profound effects on embryonic development were reported in gp130^FxxQ/FxxQ mice, where STAT signaling was abolished after replacement of the cytoplasmic domain of gp130 with a human cDNA encoding phenylalanine substitutions in all YxxQ motifs 36. However, the latter mutation resulted in perinatal death associated with the apparent failure of milk uptake and perturbed CNTF-mediated astrocyte differentiation 35. Since the phenotype of gp130^FxxQ/FxxQ mice is somewhat reminiscent of that observed in LIF-Rβ^2/− 10 and CNTF-Rα^2/− mice 36, but entirely different to that associated with IL-6^−/− 4,5 or IL-11Rα1^−/− mice 8,9, signaling through the heterodimeric LIF-Rβ/gp130 rather than through homodimeric gp130 appears critical after birth. Therefore, it is likely that STAT signaling through LIF-Rβ/gp130^FxxQ receptor heterodimers is rate limiting in gp130^FxxQ/FxxQ mice, while signaling through LIF-Rβ/gp130^ΔSTAT heterodimers, although decreased when compared with wt LIF-Rβ/gp130 dimers, reaches the threshold that enables postnatal survival of most of the gp130^{ΔSTAT/ΔSTAT} pups. The “magnitude” of the STAT signal may be influenced by the rate of internalization of ligand occupied LIF-Rβ/gp130 complexes, which in turn is determined by the collective activity of individual dileucine internalization signals (L₁₀₆₄I in LIF-Rβ and L₉₈₄L in gp130; references 37 and 38). Thus, one could speculate that LIF-Rβ/gp130^FxxQ dimers are more efficiently internalized than LIF-Rβ/gp130^ΔSTAT dimers, and thus signal from the LIF-Rβ/gp130^ΔSTAT dimers may be sustained for longer.

The most striking observation in gp130^{ΔSTAT/ΔSTAT} mice was restricted joint movement which was apparent from an early age and consistently affected older mice albeit with variable severity. While early stages of the joint disease bear hallmarks of inflammation, the intermediate stages were reminiscent of rheumatoid arthritis with the overgrowing pannus-like tissue at the edge of the articular surface. Likewise, the formation of cartilage in these pannus-like tissues resembled osteoarthritis in which cartilaginous and osseous osteophytes formed along the rim of the eroded cartilage plate. The advanced stage with its overgrowth of cartilage along the synovial lining, however, resembled neither pathology but rather the rare human condition synovial chondromatosis 39, although the characteristic multiple “loose” bodies of hyaline cartilage within the joint space were only infrequently observed in gp130^{ΔSTAT/ΔSTAT} mice.

Many studies have implicated the LIF/IL-6 family cytokines and their soluble receptors as proinflammatory mitogens for synovial fibroblasts in inflammatory joint diseases including rheumatoid arthritis 32,33. Our model, in which signaling through gp130^ΔSTAT resulted in impaired or reduced induction of SOCS-1 and SOCS-3, implicates activation of the SHP-2/ras/Erk pathway as the likely molecular mechanism promoting hyperproliferation of synovial cells by increasing their sensitivity to physiological levels of LIF/IL-6 cytokines. In turn, the hypersensitivity of these cells may trigger a sequence of changes including chronic synovitis that results in cartilaginous metaplasia and culminates in a degenerative process affecting the major weight-bearing joints. However, we cannot categorically exclude the possibility that synovial hyperplasia and the formation of exogenous cartilage occurred independently of each other.

Since our data provide strong genetic evidence for a critical involvement of gp130, STAT, and SOCS proteins in maintaining the integrity of joints of the appendicular skeleton, why do other mouse strains with mutations in either the LIF/IL-6 cytokine signaling cascade or in SOCS genes lack overt joint pathology? This apparent anomaly can be explained by extrapolating from observations by Ohtani et al. 35 who showed that the gp130-dependent STAT and SHP-2/ras/Erk pathways play reciprocal roles in the immune response by providing a balance between positive and negative signals. This delicate balance between the two pathways at the level of gp130 is further emphasized through their reciprocal negative regulation 35,40,41. Hence, balanced alterations of the signaling output from gp130-dependent STAT and SHP-2/ras/Erk pathways, either by attenuation (for instance in ligand and receptor null mutants) or augmentation (for instance in mice transgenic for the ligands), resulted in a number of phenotypes which were not observed in mice with an imbalance between the two pathways (for instance in gp130^{ΔSTAT/ΔSTAT} or gp130^757F/757F mice; reference 35). Similarly, the simultaneous and sustained activation of the SHP-2/ras/Erk and STAT pathways in response to IL-6 in synovial cells from SOCS-1^−/− IFN-γ^2/− compound mice did not result in cartilaginous overgrowth or destruction of the large weight-bearing joints. However, SOCS-1^−/− IFN-γ^2/− mice developed an acute synovitis of the tendons and small joints of distal limbs (unpublished data), presumably resulting from the activity of dysregulated T cells 42.

Since IL-6 and IL-11 are the only cytokines that signal exclusively through gp130, we predicted that these cytokines would fail to elicit STAT-mediated responses in gp130^{ΔSTAT/ΔSTAT} mice. This was clearly demonstrated by failure to induce type I acute phase proteins in the liver which is a STAT-mediated response in cultured hepatocytes 20. Furthermore, gp130^{ΔSTAT/ΔSTAT} mice, like mice either deficient for IL-6 4,5,30, or containing a conditional gp130-null allele 31, or mice transgenic for a truncated and dominant interfering gp130^Δ703 protein 43, displayed impairment of T cell–dependent high affinity immune and mucosal IgA responses. By contrast, constitutive activation of the STAT and the SHP-2/ras/Erk signaling cascades, as observed in IL-6 transgenic mice, resulted in IgG1 plasmacytosis 44, a phenotype that was never observed as a result of sustained activation of the SHP-2/ras/Erk pathway in gp130^{ΔSTAT/ΔSTAT} mice. Collectively, these observations implicate STAT signaling, rather than activation of the SHP-2/ras/Erk pathway, as the driving force for IL-6–mediated plasma cell maturation.

Intriguingly, spontaneously occurring ulcerative lesions have not been previously observed as a consequence of the genetic manipulation of the IL-6/LIF cytokine axis. Notably, no intestinal phenotype was observed in IL-6^−/− or IL-11Rα1^−/− mice, although a role for STAT3 has been proposed during cell migration in epidermal wound healing 45. Thus, it is conceivable that intestinal ulceration in gp130^{ΔSTAT/ΔSTAT} mice may result from a combination of STAT-dependent reduction in IL-6–mediated first line defense against pathogens at mucosal surfaces and impairment of epithelial homeostasis, which is maintained in the intestine, at least in part, through IL-11–mediated signaling 46.

The failure of blastocyst implantation in gp130^{ΔSTAT/ΔSTAT} mice resembled in all aspects the defect observed in LIF^−/− mice 6,7 and suggests a pivotal role for the COOH-terminal region of gp130 and associated STAT signaling during blastocyst implantation and possibly decidua formation. Since STAT signaling from the mutant gp130^ΔSTAT/LIF-Rβ receptor heterodimer was compromised in all tissues examined, the implantation defect is the likely consequence of a failure to reach a threshold level of LIF-induced STAT signaling in the uterus of gp130^{ΔSTAT/ΔSTAT} mice. However, we cannot exclude the possibility that the gp130^ΔSTAT mutation may also affect the implantation capacity of the uterine epithelium in response to simultaneous stimulation by IL-6 and IL-11. Although blastocyst implantation in mice individually deficient for either IL-6 or IL-11 signaling was not affected 8,9, increased IL-6 expression is observed in the epithelial cells of the endometrium during the “implantation window” 47. The generation of IL-6^−/− IL-11Rα^2/− mice will establish whether the simultaneous action of these two cytokines may play a hitherto unknown indispensable role during the process of uterine implantation.

This study genetically implicates gp130-mediated STAT signaling in a number of physiological host responses affecting uterine implantation, type I acute phase response, humoral and mucosal immune response, and epithelial homeostasis in the gastrointestinal tract. Surprisingly, blunting of STAT signaling in the joint compartment resulted in the mitogenic hyperresponsiveness of the synovial cells thereby triggering a series of pathological responses which ultimately culminated in the degeneration of the joints. This observation is consistent with a requirement for the balanced activation of the SHP-2/ras/Erk and STAT pathways to maintain synovial homeostasis in response to gp130 signaling and highlights the crucial role for the LIF/IL-6 cytokine family in retaining the integrity of joint architecture and function. To date, no disorder has been associated with impaired gp130 signaling in humans, however, germline mutations yielding a premature stop codon have been found in the related receptor for granulocyte colony stimulating factor in a subset of patients with severe congenital neutropenia 48. On the assumption that gp130 truncation mutations in humans would phenocopy at least some of the abnormalities observed in gp130^ΔSTAT mice, one might speculate that a subset of patients with degenerative joint diseases, female infertility, or impaired mucosal immunity may result from nonsense mutations in gp130 or other genetic lesions that impair gp130-mediated STAT signaling.

We are grateful to Drs. Richard Simpson for providing IL-6 and sIL-6R and Robyn Starr for making available the pSOCS-1 expression construct. Dr. Lorraine Robb is thanked for advice on embryo implantation, Dianne Grail for expert animal husbandry, Val Feakes for histological support, and Janna Stickland for photography.