Mice lacking IRAK3 (B6.129S1-IRAK3^tmlFlv/J) with a C57BL/6 background, were kindly provided by Hans-Joachim Anders courtesy of Koichi Kobayashi and Richard Flavell; the genetic modification introduces a stop mutation resulting in a significantly truncated mutant transcript (29). Knockout (KO) status was screened using the following primers: forward primer 5′-CGTTCCATAACACACCTCTCTGC-3′; reverse primer 5′-TTCTATCGCCTTCTTGACGAGTTC-3′. WT status was determined using the following primers: forward primer 5′-GCCAGAAGAATACATCAGACAGGG-3′; reverse primer 5′-TGTTTCGGGTCATCCAGCAC-3′. C57BL/6 IRAK3^−/− mice (which express the MHC-II allele A^b) were crossed with the congenic C57BL/6NQ mice (expressing the MHC-II allele A^q) to increase susceptibility to arthritis (30); C57BL/6NQ mice were generously provided by Johan Bäcklund (Karolinska Institute). Mice were crossed to create heterogeneous breeding pairs, which were further crossed and their progeny screened for H2^q homozygotic mice that were homozygous for a lack of Irak3; see Supplementary Table 1 for primers. All mice were housed in pathogen-free conditions with ad libitum food and water at the Kennedy Institute of Rheumatology (KIR). All procedures were conducted in accordance with UK Home Office regulations and guidelines.

Male C57BL/6NQ IRAK3^−/− and C57BL/6NQ WT mice were immunized with bovine type II collagen in complete Freud’s adjuvant as previously described (31). After mice developed clinical arthritis, the disease severity in the paws was scored as follows: 0, normal; 1, slight swelling and/or erythema; 2, clear swelling; and 3, pronounced edematous swelling/ankylosis. Clinical parameters were measured for 10 days after disease onset in each animal, and then animals were euthanized; blood plasma, spleens, inguinal lymph nodes, and paws were harvested. For pre-onset studies, tissue harvests took place 14 days after immunization.

To analyze Th lymphocyte subsets, cells from the spleen and LN were isolated as previously described (28). Leucocytes were directly stained with viability dye (Zombie NIR Fixable Viability dye, BioLegend, San Diego, CA, USA), and antibodies for CD4 (RM4-5, BioLegend) and CD25 (PC61, BioLegend); then fix/permeabilized and stained intra-cellularly for FoxP3 (236A/E7, eBioscience, San Diego, CA, USA), Gata3 (TWAJ, eBioscience), RORγt (Q31-378, BD Biosciences, San Jose, CA, USA), Tbet (4BD10, BioLegend), and Helios (22F6, BioLegend); and analyzed using the FACSDIVA software (BD Biosciences) and FlowJo (TreeStar). To quantify T-cell subsets, live cells were gated into Tregs (CD4⁺CD25⁺FoxP3⁺), natural Tregs (CD4⁺CD25⁺FoxP3⁺Helios⁺), Th2 (CD4⁺Gata⁺), Th1 (CD4⁺Tbet⁺), and Th17 (CD4⁺RORγt⁺) cells.

Blood was collected via cardiac puncture into microcentrifuge tubes containing heparin. Samples were then centrifuged at 13,000 rpm for 10 minutes at 4°C. Plasma was then stored at −80°C until the measurement of anti-collagen responses as previously described (32). Inflammatory mediators in murine plasma were measured using the Mouse TH1/TH2 9-Plex Ultra-Sensitive Kit (MesoScale Discovery, Rockville, MD, USA; catalogue number K15013C-1) according to the manufacturer’s instructions.

Paws were snap-frozen in liquid nitrogen and processed as previously described (31). Briefly, tissue was pulverized with the BioPulverizer™ (BioSpec, Bartlesville, OK, USA). Paw powder was then homogenized in 500 μL of TRIzol reagent (Invitrogen, Carlsbad, CA, USA) using the Sample Grinding Kit (GE Healthcare, Chicago, IL, USA). Chloroform (100 μL) was added to the tube, and the lysate was mixed and then centrifuged to separate the mixture into a lower phenol phase and upper aqueous separated by an interphase. The aqueous phase of the phenol/chloroform extraction was mixed with an equal volume of ethanol and then added to an RNA extraction column (miRNeasy Mini Kit, Qiagen, Valencia, CA, USA), and mRNA extraction was completed according to the manufacturer’s instructions.

cDNA was reverse transcribed from 500 ng of RNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Foster City, CA, USA), according to the manufacturer’s protocol. The expression of target genes was determined using TaqMan assays (Thermo Fisher Scientific, Waltham, MA, USA) and qPCR master mix (Takyon Low ROX Probe 2x dTTP blue, Eurogentec, Seraing, Belgium) and was expressed relative to Gapdh gene expression using the ΔΔCT approximation method; see Supplementary Table 2 for TaqMan assays. Murine Mφ from mice lacking either Tnfrsf1a or Tnfrsf1b were derived from bone marrow as previously described (33) and stimulated for up to 12 hours with Lipopolysaccharide (LPS) (10 ng/mL) before being processed for RNA extraction, reverse transcription, and qPCR.

“Classical” full-length IRAK3 (producing NP_009130.2) and “alternative” death domain-truncated IRAK3 (producing NP_001135995.1) were cloned by traditional molecular biology techniques and transfected into HEK293 cells using Lipofectamine 2000 (Thermo Fisher Scientific). After 24 hours, cells were lysed for extraction using the Proteome Profiler Human NFkB Pathway Array kit (R&D Systems, Minneapolis, MN, USA). Human monocyte-derived Mφ (MDM) were derived from monocytes and stimulated as previously described (33). Briefly, MDM were differentiated from monocytes for 5 days in 10% foetal bovine serum supplemented Roswell Park Memorial Institute 1640 medium (FBS RPMI) (10⁷ cells/10 mL/10-cm dish) supplemented with 50 ng/mL macrophage-colony stimulating factor (M-CSF) (PeproTech, Cranbury, NJ, USA); MDM were re-plated into 12-well plates, rested overnight, and then stimulated with medium alone, LPS (10 ng/mL, Merck, Darmstadt, Germany), LPS + IFNγ (10 ng/mL + 10 ng/mL, Merck/PeproTech), TNF (50 ng/mL, PeproTech), granulocyte macrophage-colony stimulating factor (GM-CSF) (50 ng/mL, PeproTech), IL-4 (50 ng/mL, PeproTech), or TGF-β1 (100 ng/mL, PeproTech) for 20 hours; protein and RNA were extracted using an Isolate II RNA/DNA/Protein kit (Bioline, Luckenwalde, Germany). Protein concentration was determined (Pierce™ BCA Protein Assay Kit, Thermo Fisher Scientific), and 4 µg of protein was boiled in Laemmli buffer (Alfa Aesar), run on a polyacrylamide gel (NuPAGE™ 4%–12% Bis-Tris Mini Protein Gels, Thermo Fisher Scientific), and then transferred to a nitrocellulose membrane for immunoblotting. IRAK3 was detected using the antibody clone SAB3500193 (Merck) or AF6264 (R&D Systems) according to the manufacturer’s instructions; β-actin was detected using AC-74 (Merck).

Synovial explants from RA and OA patients were processed as previously described (34). Briefly, synovial membrane tissue was obtained from RA and OA patients’ joints, dissected from the surrounding tissues, and digested in vitro with Liberase TL (Merck) and DNAseI (Merck) diluted in RPMI. After incubation at 37°C for 1–2 hours, the digestion was halted by the addition of RPMI containing 10% FBS, and the digestion passed through a cell strainer and washed, and erythrocyte lysis was performed. The cells were then washed and pelleted for RNA extraction and reverse transcription. Gene expression was measured by standard curve qPCR using a linearized plasmid containing the “classical” IRAK3 isoform and housekeeping gene (HPRT1); tissues were collected in compliance with approval from the Riverside Research Ethics Committee. The gene expression of IRAK3 in MDM was determined by ΔΔCT approximation.

Data were analyzed using Excel (Microsoft Ltd.), Prism (GraphPad Software Ltd.), and MultiExperiment Viewer (TM4 software). Student’s t-test or one-way ANOVA with post-hoc analysis was used to test statistical significance. A p-value <0.05 was considered statistically significant.

We and others have demonstrated the differential expression of splice variants with different functional properties in inflammatory conditions (33, 35). Although most research has focused on the larger variant of IRAK3, which has the full-length death domain necessary for associating with the Myddosome complex (6), we demonstrated that two commercially available antibodies were able to detect both isoforms ( Figures 1A–C ) before determining whether mature Mφ primed with different inflammatory stimuli (including TNF) expressed different proportions of each isoform. Only the larger isoform was observed in human MDM primed with different stimuli ( Figure 1D , see Supplementary Figure 1 for quantification). As previously observed (2), the expression of IRAK3 is induced by TNF ( Figures 1E, F ); however, murine Mφ from Tnfrsf1^−/− mice had significantly reduced the expression of IRAK3 compared to WT or Tnfrsf2^−/− mice ( Figure 1G ). In synovial explants from OA and RA patients, the expression of IRAK3 was measured by standard curve qPCR; IRAK3 was found to be higher in RA patient samples ( Figure 1H ).

As IRAK3 variants have an association with accelerated and exacerbated disease activity in humans, we bred a transgenic Irak3^−/− mouse onto a background suitable for CIA in order to determine how the lack of this gene altered experimental disease. The murine MHC-II A^q molecule has a high degree of similarity to HLA-DRB1*0101 and *0401, which confer susceptibility to human RA (30). Most genetically modified mouse strains have a C57BL/6 (H2^b) background, which is resistant to the development of autoreactive T-cell responses to type II collagen in CIA. Hence, it has been proposed that C57BL/6NQ mice (expressing A^q) serve as an international standard in studies of T cell-driven immunopathology (30). We crossed IRAK3^−/− mice onto C57BL/6NQ mice after developing a PCR method for genotyping H2^b and H2^q mice; this necessitated the sequencing of relevant genomic regions of each strain ( Supplementary Figure 2A ) and a comparison of genotyping by flow cytometry and PCR ( Supplementary Figures 2B–D , 3 ). The IRAK3^−/− NQ mice were used for CIA, with disease induction being performed as previously described (33). As shown in Figure 2A , the progression of the disease in IRAK3^−/− NQ mice was significantly more rapid, with the average time to arthritis onset being 26 versus 38 days in the WT group ( Figures 2A, B ). Disease activity was similar, although there was a trend towards increased severity in IRAK3^−/− mice, as indicated by greater paw swelling at later time-points ( Figures 2C–E ); this was consistent with significantly higher levels of IL-1β and lower levels of IL-5 detected in the plasma of IRAK3^−/− NQ mice ( Figure 3A ).

To further investigate the differences in kinetics and severity of disease between WT and KO mice, an analysis of auto-antibodies and Th lymphocyte subsets was performed to assess systemic changes between strains. Anti-collagen auto-antibodies were measured in the plasma of arthritic animals to determine humoral responses and infer cytokines present in proximity to auto-reactive B cells; IgG1 and IgG2a anti-collagen antibodies were measured to determine class switching. Surprisingly, while the level of IgG1 antibodies was virtually the same in both strains, the EC₅₀ for a serial dilution analysis was higher for IgG2a antibodies in the IRAK3^−/− group, indicating lower titers for this Ig class ( Figure 3B ).

In spleens and lymph nodes, no difference between the two strains was found for Th1 and Th17 cell activity; however, the proportion of Tregs was reduced in IRAK3^−/− spleens and lymph nodes; Th2 cells were reduced in spleens ( Figures 3C, D ). The ratio of Treg:Th17 cells was also significantly reduced in spleens ( Figure 3D ) and nominally lower in LN ( Figure 3C ). These data were consistent with mean fluorescence intensity of FoxP3 staining in FoxP3⁺ cells that was almost threefold higher in WT versus KO lymph nodes (2,914.50 ± 298.9 vs. 1,159.9 ± 114.9, p = 1.6 × 10⁻⁶) and spleens (3,015.4 ± 332.3 vs. 1,081.9 ± 96.9, p = 7.4 × 10⁻⁷) (data not shown). As disease progression was accelerated in IRAK3^−/− mice, we measured the proportions of T-cell subsets ( Supplementary Figures 4A, B ) and anti-collagen antibody titers in the pre-onset period ( Supplementary Figure 4C ); a similar reduction in Tregs was observed in IRAK3^−/− mice, but anti-collagen antibody titers were lower and not dissimilar between strains at this time. In the spleens of naïve IRAK3^−/− mice, there were minor increases in CD4⁺ cells co-expressing Tbet, Gata3, or RORγt ( Supplementary Figures 4D, E ). These data indicate significant systemic differences, particularly in regard to T helper subsets, between the strains in the course of experimental arthritis.

As there were systemic differences between strains, we then compared gene expression localized to the affected joints to determine in finer detail the differences beyond the nominal differences in paw swelling.

Despite IRAK3^−/− and WT mouse joints showing relatively minor clinical differences in CIA, a clear difference between the groups was observed when a panel of genes specific for the Th17 and Treg sub-populations was measured ( Figure 3E , Supplementary Figure 5 ). IRAK3^−/− CIA paws expressed higher levels of Cd3e and Cd4, suggesting increased numbers of CD4⁺ lymphocytes in the paws of IRAK3^−/− mice compared with WT mice. Concomitantly, IRAK3^−/− paws expressed higher levels of Stat3 and Ahr, two markers associated with the differentiation of Th17 cells, and significantly higher levels of FoxP3, Il2ra, Ctla4, and Tnfrsf1b, which are expressed by Tregs; higher levels of Il12b, Ilb, Il6, and Il17 were also measured, although the difference did not reach statistical significance. Tnf and Tgfb1 gene expression levels were also significantly higher in the IRAK3^−/− group.

The influence of IRAK3 on the response to therapy has been observed in RA, in which an IRAK3 polymorphism has been shown to be predictive of response to anti-TNF (36). IRAK3 dysregulation is also associated with accelerated and exacerbated asthma (9) and infectious diseases (37), and increased IRAK3 expression in peripheral blood leucocytes is associated with reduced disease activity in RA (17). Animal models using IRAK3 knockout mice have confirmed its role in inhibiting inflammatory responses, but few studies have compared IRAK3 isoforms (6, 38); the disruption of the death domain has been demonstrated to prevent the immunomodulatory activity of IRAK3 (6). We and others have previously demonstrated relatively high levels of inflammatory gene expression (including TNF, which induces IRAK3) in affected joints in arthritis (31). As blocking TNF in vitro has been shown to prevent the induction of IRAK3 (2), this study investigated the role of IRAK3 in an animal model of RA, a disease which is amenable to TNF blockade. We began by determining the isoforms of IRAK3 present in mature Mφ exposed to different inflammatory stimuli. As it lacks most of the death domain, the capacity of the alternative isoform to attenuate TLR/IL-1β signaling is likely to be reduced; however other putative actions, such as guanylate cyclase activity, may still be intact (7). We observed that only the longer classical form of IRAK3 was expressed at the protein level.

A deficit of Tregs or Treg activity is a feature of the failure of tolerance in autoimmune disease in multiple disease models (39–41). Changes to cell composition in spleens and lymph nodes of IRAK3^−/− animals, as has been observed previously (22), may have contributed to reduced antibody class switching. While FoxP3 is a phenotype-defining transcription factor for Tregs, the two main sub-types, natural and inducible, were reduced. Most relevant to the current study is the observation that IRAK3 reduces the severity of disease in experimental autoimmune encephalomyelitis (EAE) by Lui B et al. (38); in that study, IRAK3 deficiency accelerated disease progression and increased the ratio of Th17 to Tregs, which we also observed during arthritis. Reduced CD3⁺FoxP3⁺ T cells were also observed in the spleens of IRAK3-deficient mice in a colorectal cancer model (22). An increase of CD4⁺GATA⁺ lymphocytes was measured in the spleens of WT mice, which also had increased plasma levels of IL-5, which is secreted by GATA3⁺ Th2 lymphocytes (42).

A panel of inflammatory genes was assessed for the IRAK3^−/− and WT CIA paws. There was a significant difference in the production of pro-inflammatory mediators in the paws of the IRAK3^−/− group. In particular, pro-inflammatory cytokine genes such as Tnf, Il12b, Il1b, Il6, and Il17 were all expressed at higher levels in the IRAK3^−/− mice, although the difference was not always statistically significant; the expression of Tgfb1 and Il10 was also increased. When genes specific for different T-cell subsets were assessed in arthritic paws, an increased expression of both Th17 and Treg markers was observed. IRAK3^−/− arthritic paws expressed higher levels of pan-lymphocytic markers Cd3 and Cd4. Smad3 and Foxp3, which are Treg-associated transcription factors, were higher in IRAK3^−/− paws as were Stat3 and Il23r, which are more specific for the Th17 subset. Interestingly, Ctla4 appeared to be expressed in the IRAK3^−/− joints at higher levels. CTLA-4 binding to the B7 family members blocks the co-stimulatory signal represented by CD28-B7; this mechanism is used by nTregs to control T-cell activity and as such is suggested to represent a marker for nTregs. Based on these data, we could hypothesize that Th17 and Tregs are increased in the IRAK3^−/− paws. The perturbation of T helper subsets in LN and spleen, and increased T cell-associated gene expression in paws, is suggestive of differences in the mobilization and trafficking of T-cell subsets in the absence of IRAK3.

Fewer significant differences were measured in pre-onset lymph nodes and spleens; however, there were a reduced number of FoxP3⁺Helios⁺ cells in IRAK3^−/− LN and spleens, and a lower ratio of Tregs to RORγt⁺ lymphocytes in spleens. This indicates a cellular re-distribution of Th sub-populations between different compartments during the pre- and post-onset periods, although an indirect effect of prolonged and exacerbated inflammation on T-cell FoxP3 expression cannot be discounted. The paws as the primary site of the inflammation would contain more pathological cells, such as Th17, and this phenotype may have been stabilized by the higher levels of IL-6 and TGF-β1.