C57BL/6 mice were obtained from Janvier (Le Genest-St-Isle, France). IL-33 KO mice (B6.129Sv-Il33) were generated at Amgen Inc. (Thousand Oaks, CA, USA) by targeting of the Il33 gene in 129Sv ES cells resulting in the deletion of six out of the seven coding exons 24. Genotyping was performed by a 3-primer PCR combining a common forward primer (5'-TGC TGA ATT TTA TTC TCC CCC C-3') with reverse primers specific for the wild-type (WT) (5'-GCC CGT CTT CAT GTT GAA ATA-3') or the KO (5'-GCT CAT TCC TCC CAC TCA TGA-3') allele. IL-33 KO mice were backcrossed to the C57BL/6 background for six generations by speed congenics and considered to be 100% congenic based on analysis of a 377 SNIP panel (Taconic, Hudson, NY, USA). ST2 KO C57BL/6 mice (Il1rl1^tm1Anjm, 25) were obtained from the MRC Laboratory of Molecular Biology (Cambridge, UK). For arthritis experiments, local colonies of WT, IL-33 KO and ST2 KO mice were established at the Centre Médical Universitaire in Geneva. KRN T cell receptor transgenic mice 26 were provided by the Institut de Génétique et de Biologie Moléculaire et Cellulaire (Strasbourg, France) and maintained on a C57BL/6 background (K/B). NOD/Lt mice were purchased from Charles River (L'Arbresle, France). Mice were housed under conventional or low barrier conditions. Institutional approval was obtained for all animal experiments (Geneva Veterinarian Office, licenses 31.1.1005/3402/2, 3402/2-C and 3402/2-R).

Blocking anti-ST2, anti-IL-1RI and isotype-matched control antibodies were generated at Amgen Inc. Cell culture media were obtained from Invitrogen Life Technologies (Basel, Switzerland). Recombinant mouse IL-33 was purchased from Enzo Life Sciences (Lausen, Switzerland), human IL-1β and mouse IL-18 from R&D Systems (Abingdon, UK), and purified lipopolysaccharide (LPS) from Fluka (Escherichia coli 055:B5, Buchs, Switzerland). Murine IL-36β was produced at Amgen Inc. as an N-terminal truncation variant 27.

Arthritic K/BxN mice were generated by crossing K/B mice with NOD/Lt mice, adult arthritic K/BxN mice were bled, and the sera were pooled. Age-matched female recipient WT, IL-33 KO or ST2 KO C57BL/6 mice were injected with pooled serum (200 μl, i.p.) on days 0 and 2. Alternatively, WT C57BL/6 mice were injected with pooled serum (200 μl, i.p.) on days 0 and 2, as well as with a monoclonal murinized immunoglobulin G (IgG1)-blocking anti-ST2 antibody (150 μg/mouse), a monoclonal murinized IgG1-blocking anti-IL-1R1 antibody (150 μg/mouse), or a monoclonal mouse IgG1 antibody directed against an irrelevant human antigen that was used as an isoytpe-matched control antibody (150 μg/mouse), on day 0 (4 h before injection of K/BxN serum) and on day 3 (experiment 1), or on days 0 and 2 (4 h before each injection of K/BxN serum; experiment 2). Efficacy of the blocking anti-ST2 antibody was demonstrated previously 2428. The development of arthritis was assessed daily and the severity of arthritis was scored in a blinded fashion for each paw on a 3-point scale, in which 0 = normal appearance, 1 = localized edema/erythema over one surface of the paw, 2 = edema/erythema involving more than one surface of the paw, 3 = marked edema/erythema involving the whole paw. The scores of all four paws were added for a composite score. Mice were sacrificed on day 6.

Serum of arthritic K/BxN mice (20 ml) was purified on protein A/G Sepharose 4 Fast Flow (GE Healthcare Life Sciences, Glattbrugg, Switzerland) and dialyzed into PBS 29. Recipient WT, IL-33 KO or ST2 KO mice were injected with the purified IgG fraction (450 μl i.p., corresponding to 200 μl of the initial volume of K/BxN serum) on days 0 and 2. Mice were assessed for arthritis severity daily and sacrificed on day 6.

At sacrifice, the right ankle joints were fixed in 10% buffered formalin, decalcified in 15% EDTA, and embedded in paraffin. Sections were stained with hematoxilin and eosin or toluidine blue and graded by one pathologist (CAS) in a blinded manner. The severity of the synovial inflammation including synovial hyperlasia and the % of polymorphonulcear (PMN) cell infiltration, as well as the degree of cartilage erosion and bone destruction were evaluated utilizing a scoring system ranging from 0 to 4 (0 = normal, 1 = minimal, 2 = moderate, 3 = severe, 4 = very severe) 30.

Left ankles were harvested at sacrifice and RNA was extracted with Trizol (Invitrogen Life Technologies). Total RNA (1 μg) was treated with RQ1 DNAse (Promega, Madison, WI, USA) and reverse transcribed using SuperScript II Reverse transcriptase (Invitrogen Life Technologies). Cytokine mRNA levels were assessed by RT-qPCR using appropriate primers (Table 1) and iQ SYBR Green Supermix in the iCycler iQ™ Real-Time PCR Detection System (Bio-Rad Laboratories Inc., Hercules, CA, USA). The annealing temperature was 60°C. Non-reverse-transcribed RNA samples and water were included as negative controls. RNA expression levels were calculated using the comparative method (2^-ΔCt) for relative quantification by normalization to Gapdh gene expression.

Serum and right wrist joints were harvested at sacrifice. Joints were homogenized in 500 μl of lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, complete EDTA-free protease inhibitor cocktail (Roche Diagnostics AG, Rotkreuz, Switzerland)). Lysates were cleared by centrifugation and total protein content was determined with the DC protein assay kit (Bio-Rad Laboratories Inc.). IL-33 was quantified using a mouse IL-33 Milliplex cytokine magnetic beads assay (Millipore AG, Zug, Switzerland; detection limit 1.61 and 4.3 pg/ml in two different kits used). IL-6 levels were quantified using a DuoSet ELISA Development System (R&D Systems; detection limit 16 pg/ml).

IL-33 protein expression was examined in knee joints of WT and ST2 KO mice using a polyclonal goat anti-IL-33 antibody (AF3626, R&D Systems). Formalin-fixed, decalcified, paraffin-embedded sections were deparaffinized and antigen retrieval was performed in a pressure chamber (Pascal; Dako, Baar, Switzerland) in 0.1 M Tris-HCl, pH 9.0, 0.01 M EDTA, at 125°C for 30 sec. Slides were blocked for endogenous peroxidase activity and incubated with anti-IL-33 antibody (1 μg/ml) in antibody diluent (number S2022, Dako) overnight at 4°C. Subsequently, slides were incubated with an anti-goat HRP antibody (1:500; number SC2304, Santa Cruz Biotechnology, Inc., Heidelberg, Germany) in antibody diluent and developed with diaminobenzidine (Dako). To assess staining specificity, negative controls were performed using knee sections of IL-33 KO mice.

Bone marrow-derived dendritic cells (BMDC) were generated from WT, IL-33 KO and ST2 KO C57BL/6 mice as previously described 31. On day 7, cells were seeded into triplicate wells of 96-well plates at 10⁵ cells/ml and cultured without or with 100 ng/ml recombinant IL-33, IL-1β, IL-36β or IL-18 or LPS for 72 h. IL-6 production in culture supernatants was assessed by ELISA. Cells from triplicate wells were pooled and used for total RNA extraction using the RNeasy kit (Qiagen, Valencia, CA, USA). Expression of IL-1R family receptors was quantified by RT-qPCR using appropriate sets of primers (Table 1).

IL-33 KO, ST2 KO and WT mice (n = 3 per genotype) were randomly selected at the end of the experiment shown in Figure 1 and genotyped for 55 microsatellite markers as described 32. In addition, all ST2 KO mice used in the arthritis experiments described herein were screened for D1Mit3, D1Mit211, D1Mit75a, D1Mit303, D6Mit166, D6Mit159, D6Mit102, and D15Mit193 genotypes using appropriate primers (32; see 33 for additional primer information).

To assess differences in the longitudinal evolution of arthritis outcomes, we used generalized linear mixed models for repeated measures 34. Tests were conducted at error alpha level of 0.05, two-sided and performed with STATA v. 11 for Windows (StataCorp LP, College Station, TX, USA). Differences in arthritis severity at the end of the follow-up were evaluated using the Kruskal-Wallis test. Significance of differences in histological scores, cytokine measurements, and mRNA expression was assessed by ANOVA or Kruskal-Wallis test, as appropriate.

We examined incidence and severity of K/BxN serum transfer-induced arthritis in IL-33 KO, ST2 KO and WT C57BL/6 mice. Clinical scoring showed reduced incidence and severity of arthritis in ST2 KO mice as compared to WT controls (Figure 1A-C). In contrast, arthritis development and severity in IL-33 KO mice was similar to the disease observed in WT mice. Histological scoring of arthritic ankles on day 6 confirmed a trend toward reduced inflammation and PMN cell infiltration, as well as reduced cartilage and bone erosion in ST2 KO paws (Figure 1D).

On day 6 after induction of arthritis, serum IL-33 levels were similar in WT and ST2 KO mice (WT, mean: 310 ± 192 pg/ml, range < 1.61 to 1258 pg/ml; ST2KO, 447 ± 287 pg/ml, range < 1.61 to 2142 pg/ml) and comparable to IL-33 levels measured in naïve WT C57BL/6 mice (n = 17; 294 ± 191 pg/ml, range < 1.61 to 3062 pg/ml) housed in our conventional mouse facility. IL-33 was undetectable in the serum of IL-33 KO mice. IL-33 mRNA (Figure 2A) and protein (Figure 2B) expression in synovial tissues was similar in arthritic WT and ST2 KO mice, and absent in IL-33 KO mice. Expression of IL-6 mRNA (Figure 2A) and protein (Figure 2B) in arthritic paws, monitored as a marker of inflammation, paralleled severity scores and tended to be lower in ST2 KO as compared to WT mice. Strong nuclear IL-33 immunostaining was observed in the synovial tissue of WT and ST2 KO mice, in cells morphologically consistent with fibroblasts, as well as in other cell types, which we have not formally identified (Figure 2C). Interestingly, and in contrast to human RA synovium 1617 synovial endothelial cells did not express IL-33. A similar pattern of IL-33 protein expression was observed in the synovium of WT and ST2 KO mice, while no IL-33 staining was detected in IL-33 KO mice.

We wondered whether differences observed between IL-33 and ST2 KO mice regarding the severity of arthritis might be related to the existence of confounding variables in our experiment. Among these, we wanted to exclude a potential effect of IL-33 present in the injected serum, which might affect arthritis severity in IL-33-deficient, but not in ST2 receptor-deficient mice. The amount of IL-33 detected in different KBxN serum pools (n = 4) used for arthritis induction was 77 ± 31 pg/ml (range < 4.3 to 139 pg/ml). Therefore, to avoid co-injection of 'contaminating' IL-33 contained in the serum as a potential confounding factor, we thus used total IgG purified from K/BxN serum 29 to induce arthritis. IL-33 levels were below the detection limit of the Milliplex assay (4.3 pg/ml) in this purified IgG fraction. Similarly to the experiment performed with complete K/BxN serum, clinical scoring showed significantly reduced incidence and severity of arthritis in ST2 KO mice as compared to WT controls after injection of purified IgG (Figure 3A-C). In contrast, arthritis development and severity in IL-33 KO mice was still similar to that observed in WT controls. Therefore, a contribution of IL-33 contained in the serum of K/BxN donor mice to arthritis severity in IL-33 KO recipient mice can be excluded.

In an attempt to determine whether reduced arthritis severity in ST2 KO mice might relate to a true IL-33 independent effect of ST2 rather than to the existence of confounding variables unrelated to ST2, we used a blocking anti-ST2 antibody 2428 to inhibit ST2 signaling during K/BxN serum transfer-induced arthritis (Figure 4). Treatment of the mice with the anti-ST2 antibody before and during development of arthritis did not reduce disease severity in two independent experiments, while blocking of the type 1 IL-1R with an isotype-matched anti-IL-1R1 blocking antibody essentially abrogated disease (Figure 4A-D). Histological severity scores confirmed clinical severity data (data not shown).

We next wondered whether the decreased severity of arthritis observed in ST2 KO mice might be related to off-target effects of the KO construct, rather than to deletion of ST2 itself. The ST2 KO mice used in this study were generated by replacement of exons 4 and 5 of the ST2 gene with a neomycin selection cassette, which was left in the locus 25. There are many examples where selection markers remain within targeted loci interfere with the expression of neighboring genes or even entire loci 3536, so that nowadays more refined KO strategies, allowing for the subsequent removal of selection markers, are generally preferred. The mouse Il1rl1 gene encoding the ST2 receptor is located on chromosome 1, within a cluster of genes encoding other receptors of the IL-1R family, namely the type II decoy receptor for IL-1 (Il1r2), the type I signaling IL-1R (Il1r1), the receptor for IL-36 (IL-36R, Il1rl2), as well as the receptor and co-receptor for IL-18 (IL-18R, Il18r1 and IL-18R accessory protein, Il18rap) (Figure 5A). Interference of ST2 targeting with expression of any of these genes might in principle affect arthritis severity. IL-1 activity, in particular, is known to be absolutely required for joint inflammation in this model 37. We thus examined expression of the signaling IL-1, IL-36 and IL-18 receptors (Figure 5B) and the responsiveness to cognate IL-1 family cytokines (Figure 5C) in vitro in cultured BMDC derived from ST2 KO mice, as compared to BMDC derived from WT and IL-33 KO mice. BMDC were chosen as target cells because they respond to all IL-1 family cytokines 31. No significant differences in IL-1, IL-36 and IL-18 receptor expression were observed in BMDC derived from IL-33 KO or ST2 KO mice, as compared to WT cells (Figure 5B). Furthermore, the response of IL-33 KO and ST2 KO cells to IL-36 and IL-18, as assessed by induction of IL-6 production, was comparable to that of WT cells, while the response to IL-1β was even slightly enhanced in ST2 KO cells. The response to IL-33 was of course lost in ST2 KO cells and appeared to be also slightly reduced in IL-33 KO cells (Figure 5C). LPS, used as an unrelated positive control, elicited similar IL-6 production in IL-33 KO, ST2 KO and WT BMDC. These data do not point toward any major defect in expression of other IL-1R family receptors or in IL-1 family cytokine signaling in ST2 KO cells.

IL-33 KO and ST2 KO mice were generated in 129Sv, respectively 129P2/OlaHsd ES cells and backcrossed to a C57BL/6 background. It has become increasingly clear that despite extensive backcrossing, genetic material of the original mouse strain, and in particular regions closely linked to the targeted loci, remains associated to KOs in recipient strains, sometimes leading to confounding effects mistakenly attributed to the disrupted gene 3839. We thus wondered whether carryover of 129 genetic material might have influenced the outcome of our experiments. IL-33 KO mice were considered 100% C57BL/6 congenic at the end of their backcross based on screening of a 377-marker panel. In addition, at the end of the experiment shown in Figure 1, we screened three mice per genotype using a 55-marker panel 32 to obtain a picture of the genetic background in our local mouse colonies. We confirmed 98% purity of the C57BL/6 background in WT (one heterogeneous marker: D4Mit166) and 100% purity in IL-33 KO mice. Screening of ST2 KO mice revealed lower (89%) purity (six heterogeneous markers: D1Mit211, D6Mi166, D6Mit159, D6Mit102, D11Mit224, D15Mit193). We further examined two additional markers close to the Il1rl1 gene on chromosome 1 (D1Mit3 and D1Mit75a, Figure 6A, Table 2), one of which (D1Mit3) was also non-C57BL/6 in 17 out of 20 ST2 KO mice tested (Figure 6B and data not shown). In contrast, the D1Mit75a and D1Mit303 markers, located downstream of the ST2 locus were C57BL/6 in all ST2 KO mice examined. Finally, we screened the 20 ST2 KO mice used in arthritis experiments (Figure 1 and 3) for the presence of either C57BL/6, or 129 alleles at the D1Mit211, D6Mi166, D6Mit159, D6Mit102, and D15Mit193 loci and correlated their genotypes with arthritis severity of individual mice (Figure 6C and data not shown). With the limited number of mice included in these studies, we observed no significant correlations. However, arthritis severity in mice carrying two C57BL/6 alleles at a given locus often tended to be lowest (D1Mit211, D6Mit166; D6Mit159).