Simultaneous inhibition of JAK and SYK kinases ameliorates chronic and destructive arthritis in mice

DBA/1 mice (six w.o. males from Janvier, France) were immunized subcutaneously (s.c.) (100 μl at each side of the base of the tail) with 400 μg recombinant human (rhu) glucose-6-phosphate isomerase (G6PI) emulsified in complete Freund’s adjuvant (CFA, Sigma-Aldrich, St. Louis, MO, USA) on day 0, as previously described [20]. The indicated amount of antigen was mixed with CFA in a 1:1 ratio (v/v) and emulsified with a Polytron. When specified, regulatory T cells (Tregs) were depleted injecting intraperitoneally (i.p.) 400 μg anti-CD25 Ab (PC61.5, BioXcell, West Lebanon, NH, USA) 11 and 8 days before immunization [21].

The mouse weight was recorded and the clinical score was evaluated over time. Each paw section was scored separately, and these scores were all added together as follows:

Total score per mouse = (Sum of scores of 2 wrists + 2 ankles (i.e., max 12)) + (Sum of scores of 2 metacarpals + 2 feet (i.e., max 12)) + (Number of inflamed fingers (max 8) + toes (max 10)/2 (i.e., max score of 9)).

For each paw section, a score of 0 to 3 was assigned, where 0 indicates no clinical signs of arthritis (healthy state), 1 and 2 indicate mild/intermediate swelling and redness of the paw, and 3 indicates massive swelling, redness and burst skin.

CD-1 mice (six to eight w.o. females from Janvier, France) were immunized on day 0 with 300 μg of keyhole limpet hemocyanin (KLH, Sigma-Aldrich, subplantar injection of 30 μl). On day 14, mice were sacrificed and blood, plasma and spleens were obtained.

The kinase inhibitors Tofacitinib (JAKi) and PRT-062607 (SYKi) were synthesized by Aptonchem Co. Ltd. (Hangzhou, China) or in house (Draconis Pharma), respectively. These compounds were diluted in water, sonicated for 5 minutes and given orally (p.o.) (20 mg/kg JAKi and 30 mg/kg SYKi) once daily (QD) starting on day 0 or 12, unless otherwise specified. We corrected the amount of each compound used by the molecular weight (MW) of the free amine: MW of JAKi = 504.49 g/mol (free amine 312.37 g/mol) and MW of SYKi = 453.50 g/mol (free amine 393.45 g/mol). Prednisolone (Sigma-Aldrich, 3 mg/kg) was also diluted in water and given p.o. QD. The soluble dimeric human TNFR p75–IgG-Fc fusion protein (etanercept, Enbrel®, Pfizer, Thousand Oaks, CA, USA) was diluted in PBS and given s.c. (10 mg/kg) every third day.

Samples were obtained at the indicated time points. Inguinal lymph nodes and spleens were weighted and single cell suspensions were made in Roswell Park Memorial Institute (RPMI)1640 medium (Life technologies, ThermoFisher Scientific, Eugene, OR, USA, supplemented with 10 % FBS, glutamine, 100 U/ml penicillin, 100 mg/ml streptomycin) by mechanical disruption of the tissue with a syringe plunger on a 40-μm cell strainer. Peripheral blood was collected in tubes with 0.5 % K₂-EDTA, hematological parameters analyzed by Celltac E analyzer (Nihon Khoden, Tokyo, Japan) and plasma stored at −80 °C for cytokine and Ig quantification. Ankle swelling was measured using a dial-gauge caliper (Peacok, Ozaki MFG Co. Ltd., Tokyo, Japan); the average thickness of both hind legs was used. The right hind limb of each mouse was fixed in 10 % formalin and then decalcified with Osteomoll (Merck Millipore) for histopathological assessment.

Splenocytes were plated in triplicates (8 × 10⁵ cells/well in 96-well flat-bottom plates) and re-stimulated with either 20 μg/ml rhu G6PI or PBS in RPMI medium supplemented with 10 % FBS, 100 U/ml penicillin and 100 μg/ml streptomycin at 37 °C in 5 % CO₂ for 72 h. When indicated, splenocytes from non-treated arthritic mice were also exposed to 0.5 μM of JAKi and 0.5 μM of SYKi for 72 h. Supernatants were collected and murine IFNγ, IL-17A and IL-6 were measured by ELISA (BD, Franklin Lakes, NJ, USA,  and Affymetrix, Santa Clara, CA, USA and eBioscience, San Diego, CA, USA), according to manufacturer’s protocols.

Plasma levels of G6PI- and KLH-specific antibodies were measured by ELISA. Briefly, 96-well flat-bottom plates were coated with 5 μg/ml of G6PI or 10 μg/ml of KLH in carbonate buffer (0.1 M Na₂CO₃ and NaHCO₃, pH 9.5) overnight at 4 °C, washed with PBS containing 0.05 % Tween-20 (Sigma-Aldrich) and incubated with a blocking buffer (2 % BSA (Sigma-Aldrich) in PBS) for 1 h to saturate non-specific binding sites. Serial 5-fold or 10-fold dilutions of the plasma in PBS + 0.05 % Tween + 2 % BSA were added for 2 h followed by 1 h incubation with isotype-specific Ab with peroxidase (goat anti-mouse Fcγ- or Fcμ-specific Ab-HRP, Jackson ImmunoResearch, Suffolk, UK). After washing, the substrate 3,3′-5,5′ tetramethylbenzidin (TMB) (BD) was added, the reaction was stopped with 1 M H₂SO₄ and the absorbance was determined at 450 nm with a reference wavelength of 570 nm.

Cell surface phenotyping was performed by using anti-mouse antibodies to CD3, CD4, CD8, CD19, CD11c, CD11b, MHCII, CD86, CD49b, CD25, CD44, CD138 and/or fixable live/dead. For intracellular markers, cells were fixed, permeabilized and stained with anti-Foxp3, CD40L, IFNγ, IL-2, TNFα, 5-Bromo-2′-deoxyuridine (BrdU), ovalbumin (OVA), κ and/or λ. Cells were analyzed with a FACS Calibur or a LSR Fortessa flow cytometer (BD Biosciences) and data were analyzed using FlowJo software.

Murine splenocytes (effector cells, EC) were isolated from treated and untreated mice. Target cells (TC, YAC-1 cell line) were labeled with 2.5 μM CFSE fluorescent dye (Life technologies) to separate them from the EC population. Untreated splenocytes were treated in vitro for 1 h at 37 °C with JAKi + SYKi or anti-TNF, as indicated. Then, labeled TC were co-incubated with EC (ratio of 1 TC:10 EC) for 4 h and the dye PI (Life technologies) was added to discriminate live and dead cells.

Peripheral blood mononuclear cells (PBMCs) were isolated from human buffy coats by Ficoll density gradient. Cells (2 × 10⁶ cells/tube) were pre-treated with JAKi and/or SYKi for 1 h. Media were removed and replaced with 1 mg/ml Red pHrodo^TM BioParticles® suspension (Life Technologies) and incubated for 1.5 h at 37 °C. Phagocytosis of particles was quantified by flow cytometry.

Total RNA was purified from splenocytes using the PureLink RNA Mini kit (Ambion, ThermoFisher Scientific, Carlsbad, CA, USA) and the concentration and purity were assessed using a Safire spectrophotometer (Tecan, Männedorf, Switzerland). RNA (1 μg) was retro-transcribed to cDNA using High-Capacity cDNA Reverse Transcription kit (Life technologies) following manufacturer’s instructions. RANKL, IFNγ and IL-6 expression were evaluated by quantitative real-time PCR using TaqMan Universal Master Mix and a 7500 Real Time PCR System (Applied Biosystems, ThermoFisher Scientific). Relative quantification was determined using the comparative method: 18S ribosomal RNA was used as the housekeeping gene (Mm04277571_s1). Taqman probes for RANKL, IFNγ and IL6 were Mm00441906_m1, Mm00801778_m1 and Mm00446190_m1, respectively (Applied Biosystems).

Tissue samples were paraffin-embedded and longitudinal microsections were stained with H&E (Merck Millipore), according to standard procedures. A score of 0 (normal) to 5 (strongly affected) was given based on the degree of inflammation, bone erosion, cartilage damage and pannus formation in tarsus and phalanges, as previously described [22, 23]. Experienced pathologists at AnaPath (Spain) processed the tissues and performed the blinded histopathological assessment.

Small joints from Treg-depleted mice at day 56 post-immunization were dissected and digested in 1 mg/ml collagenase type IV solution for 2 h (Worthington Biochemical Corp., Lakewood, NJ, USA 250 U/mg). FLS were cultured in DMEM (Sigma-Aldrich) supplemented with 10 % FBS, 100 U/ml Penicillin/Streptomycin (Jena Bioscience, Jena, Germany) and 10 mM Hepes (Serva, Heidelberg, Germany). All the FLS used were between passages 4 and 5. FLS were plated (5 × 10⁴ cells/well in 24-well plates) and stimulated with 10 ng/ml rmTNF-α and 50 ng/ml of rmIL-17A (both Peprotech, London, UK) or left unstimulated. All cells were also treated with 0.5 μM JAKi and/or SYKi; dimethyl sulfoxide (DMSO) only was used as a control. After 24 h of incubation, the supernatants were harvested and IL-6, matrix metalloproteinase-3 (MMP-3) and MMP-9 were quantified by ELISA (reagents from eBioscience, San Diego, CA, USA and Peprotech for IL-6; commercial DuoSet kits from R&D Systems, Minneapolis, MN, USA for MMPs).

To test FLS invasiveness, a transwell system was used. ThinCerts™ inserts with 8-μm pores (Greiner bio-one, Frickenhausen, Germany) were coated with a bovine collagen solution (PureCol ®, Advanced BioMatrix, Carlsbad, CA, USA; 10 × MEM, Gibco, ThermoFisher Scientific, Waltham, MA, USA; sodium bicarbonate 7.5 %, Gibco) and put in 24-well plates. FLS (2 × 10⁴ cells/well) resuspended in culture medium without FBS in the presence of 0.5 μM of JAKi and/or SYKi or DMSO only, were seeded on top of the collagen matrix. Culture medium with 10 % FBS was used as chemoattractant. After 48 h of incubation, the cells that did not migrate were removed from the upper part of the insert with a cotton swab and the lower part of the insert was stained with crystal violet. The percentage of invaded area was calculated using the program Fiji, as described [24].

Bone marrow cells were collected after flushing out of femurs and tibiae, subjected to gradient purification using ficoll-paque (GE Healthcare, Little Chalfont, Buckinghamshire, UK), plated in 96-well plates at a density of 6 × 10⁴ cells/well and cultured in αMEM medium (Gibco) containing 10 % FBS supplemented with 40 ng/ml rank ligand (RANKL) (R&D Systems) and 25 ng/ml M-CSF (R&D Systems) [25] in the presence or absence of CP and PRT compounds for 5 days. To visualize osteoclasts, cell cultures were stained with tartrate-resistant acid phosphatase (TRAP), using an acid phosphatase leukocyte (TRAP) kit (Sigma-Aldrich).

To assess osteoclast activity, bone marrow cells were seeded on 650-μm-thick bovine cortical bone slices (10⁵ cells/slice) in αMEM supplemented with 5 % FBS, 20 ng/ml RANK-L and M-CSF (30 ng/ml). Medium was refreshed after 3 days. Differentiated osteoclasts were treated with DMSO or 0.5 μM of JAKi and/or SYKi on day 5 and bone resorption was assessed 48 h later. Osteoclasts were lysed with water and mechanically removed by sonicating the bone slices in 10 % NH₃ for 20 minutes. Finally, bone slices were washed, incubated for 10 minutes with 10 % hydrated potassium aluminium sulfate, washed again and stained with Coomassie blue (PhastGel Blue R-350; GE Healthcare Bioscience, Uppsale, Sweden) in order to visualize the pits of resorption. Five micrographs per slice were acquired with a digital camera mounted on an inverted light microscope and the percentage of eroded area was quantified using the program Leica Application Suite (version 4.3.0).

BALB/c mice (Charles River, Germany) were immunized i.p. with 100 μg OVA (Sigma-Aldrich) in 100 μl alum (ThermoFisher Scientific) at day 0. On day 21, mice were boosted with 100 μg OVA in 100 μl alum. From day 20 to 35, mice received 1 mg/ml BrdU (Sigma-Aldrich) dissolved in drinking water to label newly generated plasma cells (i.e., plasma cells generated after boost). From day 35 to 46, mice were treated with 20 mg/kg CP and/or 30 mg/kg PRT (p.o. QD). On day 46, single cell suspensions were generated from spleen and bone marrow and stained for flow cytometry. For the ex vivo re-stimulation assay, cells were plated in complete RPMI medium in 12-well plates (10⁷ cells/well) and re-stimulated for 6 h with 100 μg OVA. Brefeldin A (BioLegend, London, UK) was added after the first 4 h and finally stained for flow cytometry.

Data are expressed as mean ± standard error of the mean and were analyzed and graphed with Prism software version 5 (GraphPad, La Jolla, CA, USA). Statistical analysis was calculated using Student’s t test (unpaired, two-tailed) and one-way or two-way analysis of variance (with the Dunnett or Bonferroni post hoc test). Differences were considered statistically significant when the p value was <0.05.

We used tofacitinib (CP-690,550, designed by Pfizer) as a selective pan-JAK inhibitor (JAKi) and PRT062607 (Portola Pharmaceuticals Inc. and Biogen Idec Inc.) to specifically inhibit SYK (SYKi), or their combination (JAKi + SYKi). In Additional file 1, we show data obtained to determine the doses of inhibitors to be further used for in vivo studies. Briefly, we first confirmed the potency and selectivity of these inhibitors in biochemical and cellular assays (half maximal inhibitory concentration (IC₅₀) (JAK) approximately 0.02 μM and IC₅₀ (SYK) approximately 0.5 μM). We also determined their efficacy in standard acute arthritis models in mice; namely the collagen-induced arthritis (CIA) model, which strongly relies on T cell function and therefore eases the study of JAK inhibition (ED₅₀ (JAK) approximately 11 mg/kg QD), and the collagen antibody-induced arthritis (CAIA) model for SYK inhibition, as it requires processing of anti-type II collagen antibodies and immune complexes via FcR. Based on these results, we decided to use a dosage of 20 mg/kg of tofacitinib (JAKi) and 30 mg/kg of PRT062607 (SYKi) QD. A single oral dose of tofacitinib or PRT062607 in naive mice resulted in a rapid increase of plasma levels (t_max = 1–3 h), reaching maximal concentrations that did not compromise kinase selectivity, and were completely cleared daily. Therefore, the selected doses guaranteed high efficacy while maintaining selectivity.

To test the efficacy of dual JAK + SYK inhibition, we chose a novel and well-accepted murine model of arthritis induced by the systemic antigen G6PI. Immunization is achieved by administering G6PI along with CFA subcutaneously. The resulting acute, self-limited type of arthritis [20] can be switched deliberately to a non-self-remitting and destructive type by transiently depleting Treg cells prior to disease induction [21]. This chronic model is characterized by exacerbated clinical signs of inflammation (redness and paw swelling), joint deformations and higher levels of T helper (Th1) and Th17 cytokines compared to the acute type (see [21] and Additional file 2). Here, we were interested in mimicking severe and non-self-remitting disease, as it better represents human disease; therefore, we used the chronic G6PI model.

In order to study the impact of kinase inhibition on arthritis development, daily oral treatment with JAKi and/or SYKi was started upon immunization (see a diagram of this experimental protocol in Additional file 2). Figure 1a shows that mice receiving SYKi developed moderate and persistent arthritis compared to vehicle controls and ended with 40 % lower clinical scores. Mice treated with JAKi had mild inflammation of the paws, except for the toes and fingers, for a week only, as clinical signs were resolved in all mice after 17 days of treatment. As a measure for the aggregate effect of these treatments over time, the average area under the clinical score curve decreased by 94 % with JAKi, but by only 33 % with SYKi. The greatest effects were observed with the dual inhibition of JAK + SYK, which prevented all mice from developing arthritis. None of the mice had inflammation of the toes/fingers and only 30 % were scored with incipient swelling of the wrists/ankles for at most 4 days. Accordingly, lymph node weights were significantly decreased compared to the progressively bigger ones in arthritic controls over time (Fig. 1b). In the spleen, the CD11c^hi MHCII⁺ dendritic cell pool (not shown), and consequently the number of activated (CD86⁺) DCs (Fig. 1c), were reduced in mice treated with JAKi + SYKi, likely contributing to minimize antigen-specific immune responses.

Parameters of T and B cell activity support a critical role for JAK signaling in the induction phase of the disease. Splenocytes from treated mice were stimulated in vitro with G6PI for 72 h and cytokine production was quantified by ELISA. Levels of IFNγ, IL-17A and IL-6 were comparable in vehicle controls and SYKi-treated mice (Fig. 1d and not shown). In contrast, JAKi decreased cytokine production and a greater reduction was observed upon dual JAKi + SYKi, confirming milder Th1/Th17 responses in these groups. Similarly, CD86⁺-activated B cells and levels of G6PI-specific IgGs were significantly lower in mice receiving JAKi or JAKi + SYKi (Fig. 1e and f).

A detailed histopathological evaluation of the hind limbs of arthritic mice revealed severe lesions in both tarsal (Fig. 1g-i) and phalangeal (not shown) joints with abundant immune and inflammatory cells (primarily lymphocytes, macrophages and neutrophils), pannus formation with neovascularization, marked cartilage and bone erosion, loss of trabeculae and presence of more fibroblasts and osteoclasts. Dual JAK + SYK inhibition prevented paw inflammation and bone/cartilage damage. Most of these effects were driven by JAK inhibition, which completely cleared an initial mild inflammatory response and prevented bone and cartilage erosion. With SYKi the histopathology score was reduced by 30-40 % at day 31.

We next evaluated the therapeutic potential of dual JAK + SYK inhibition in a curative protocol, which is a more strenuous measure of activity and is clinically more relevant. In this case, oral daily treatment was started at day 12 post-immunization, when clinical signs, primarily inflammation in proximal and distal joints of paws, were present. The clinical score of vehicle control mice progressively increased, becoming stable around day 20 (Fig. 2a). Inhibition of JAK, but not SYK, was able to stop disease progression, and after 2 weeks of treatment started to mildly reduce arthritis severity. By day 31, mice treated with JAKi or prednisolone (3 mg/kg QD), a corticosteroid widely used as anti-inflammatory agent, had a reduction in overall clinical score of 58 % and 74 %, respectively. In contrast, dual JAKi + SYKi immediately prompted a gradual decrease in the clinical score for all joints (Fig. 2b) from all mice and over 70 % of these mice were totally cured (score of 0) after 3 weeks of treatment.

Upon in vitro G6PI stimulation of splenocytes from these mice, IL-6 production was only significantly reduced in the dual treatment group (Fig. 2c), confirming attenuation on antigen-specific responses. Th1 (IFNγ) and Th17 (IL-17A and IL-6) cytokine expression was also decreased (Fig. 2d and not shown). Furthermore, combined JAKi + SYKi was more effective at limiting B cell activation and function, as assessed by the levels of serum G6PI-specific IgG, relative to either inhibitor alone (Fig. 2e and f).

In terms of histopathological changes in tarsal joints (Fig. 2g and h), single JAKi or SYKi decreased around 50 % of joint inflammation, while dual treatment induced a reduction of over 90 %. Of interest, only simultaneous inhibition of JAK + SYK was able to prevent bone and cartilage damage, showing a completely normal joint structure in 60 % of mice. Therefore, concurrent JAK + SYK inhibition ameliorates clinical, immunological and histological parameters of arthritis.

We have showed greater efficacy of dual versus single JAK and SYK inhibition in a chronic model of arthritis. Next, we compared the efficacy of dual JAKi + SYKi with anti-TNF therapy, a standard anti-rheumatic biological approach used in the clinics. Previous reports showed that anti-TNF prevents the development of very early acute G6PI arthritis [20, 26]. Here, we started treatment of Treg-depleted G6PI immunized mice when arthritic symptoms were moderate to severe. Interestingly, both TNF blockade (10 mg/kg, every third day) and dual JAKi + SYKi stopped disease worsening but only dual treatment was able to progressively decrease the clinical score, which had decreased by over 60 % at the endpoint (Fig. 3a). The average area under the clinical score curve decreased by 42 % with dual treatment, while it only decreased by 25 % with etanercept (not shown). Accordingly, lymph node weight (Fig. 3b) and ankle thickness (Fig. 3c) were significantly lower in all mice treated with JAKi + SYKi but not with anti-TNF. Moreover, 66 % of dual-treated mice experienced complete recovery of phalangeal joints after 3 weeks of treatment (not shown). Blinded histologic observations also confirmed reduced recruitment of immune and inflammatory cells and joint damage in tarsal joints of mice treated with JAKi + SYKi but not in anti-TNF-treated mice (Fig. 3d). Abnormal layers of granulation and fibrovascular tissue (pannus) were abundant in joints of vehicle-treated mice, which coincides with chronic and severe stages of arthritis. Dual inhibition significantly reduced pannus progression in both tarsal and phalangeal joints, while anti-TNF treatment only had a partial effect in phalanges. Furthermore, Fig. 3e shows that dual kinase inhibition reversed the pathological changes in mice with mild and moderate disease and decreased symptoms in mice with very severe disease. In contrast, blockade of TNF stopped, without improving, disease progression in all groups. Thus, dual JAK + SYK inhibition also presents greater efficacy than anti-TNF treatment in this experimental model of severe arthritis.

We next evaluated whether the benefit of greater efficacy of dual JAKi + SYKi in arthritis was limited by increased adverse effects imputable to target inhibition. To this end, mouse weight as well as hematological and immunological parameters were examined in mice immunized either with G6PI + CFA (Figs. 4a-c and 5a) or with KLH (Fig. 4d-f) and treated for extended periods of time, i.e., 3 or 4 weeks, respectively. KLH is an immunogen routinely used in the rodent TDAR model, which allows detection of potential immunotoxicity, especially immunosuppression. As shown in Additional file 3a and b, the body weight from mice treated with JAKi/SYKi remained comparable to vehicle-treated animals. As expected, the control immunosuppressant drug prednisolone significantly reduced body weight, circulating white blood cells, spleen weight and both CD4⁺ T cell and B cell counts in G6PI- and KLH-immunized mice. Kinase inhibitors alone or in combination and anti-TNF treatment did not alter red blood cell or white blood cell numbers. Dual JAKi + SYKi and single JAKi decreased spleen weight compared to G6PI-immunized controls but never below naive levels, and without affecting CD4⁺ T cell and B cell counts. In the TDAR model, dual JAK + SYK inhibition, but not etanercept, decreased B cell counts but never below baseline, and consequently, KLH-specific antibody production was also only mildly reduced.

Preclinical studies and clinical trials with tofacitinib have highlighted an impact on Treg cells. Here, we also observed a significant decrease in Treg counts in the spleen of mice treated with tofacitinib and, consequently, also with dual JAKi + SYKi, but not SYKi alone (Fig. 5a). Similar diminished Treg numbers were identified in prednisolone-treated mice.

We also examined potential effects of JAKi, SYKi and TNF blockade on cytotoxic cell populations (Fig. 5a). CD8⁺ T cell counts were similar in all groups. In contrast, dual inhibition, but not anti-TNF (not shown) significantly impacted on natural killer (NK) cell numbers, likely due to a direct effect of JAK inhibition on IL-15 signaling. Under these conditions, we observed a fully functional cell-mediated cytotoxic response of in vivo treated splenocytes on target YAC-1 cells upon co-culture (Fig. 5b left). Comparable results were obtained in another cytotoxic assay with murine splenocytes from naive mice but, this time, treated in vitro with physiologically relevant (0.5 μM), and ten times higher (5 μM), concentrations (Fig. 5b right). Therefore, reduced NK cell numbers upon dual inhibition did not impact overall cytotoxic activity. In a separate experiment, we confirmed that tofacitinib and PRT062607 were not cytotoxic at the assayed doses (Additional file 3C).

In terms of potential effects on the phagocytic cell compartment, human PBMCs showed intact phagocytosis of Escherichia coli bioparticles when exposed to 0.5 μM of JAKi and/or SYKi for 1 h (Fig. 5c). At the high concentration of 5 μM, dual inhibition decreased almost 60 % the phagocytic activity of PBMCs, which can be explained by the role of SYK on actin cytoskeleton reorganization during phagocytic vesicle closure. The flow cytometric analysis revealed CD14⁺ cells and granulocytes to be the main phagocytes involved. Again, these adverse effects have a limited impact in vivo, as high plasmatic concentrations are reached only for short periods of time. Overall, our data show that dual JAK + SYK inhibition is efficacious and has a good safety profile, and therefore can be seen as a candidate treatment strategy in arthritis.

We next stopped treatment and ten days later 75 % of arthritic mice treated with both inhibitors either maintained or continued to ameliorate clinical signs (Fig. 5d). In sharp contrast, discontinuation of anti-TNF treatment led to a worsening of paw inflammation in all mice. Indeed, an average 17 % increase in the clinical score was observed in this group, reaching a score comparable to vehicle-treated arthritic mice. Therefore, the benefits of dual JAK + SYK inhibition were perpetuated longer than anti-TNF. Therapy discontinuation also led to immune recovery. The reduced Treg and NK cell counts after daily JAKi + SYKi were quickly restored to normal levels (Fig. 5e), which minimizes potential adverse effects.

Resident cells of the joint are major contributors to arthritis pathology. In particular, FLS and osteoclasts represent a link between joint inflammation and structural damage, which is critical in the functional disability of RA patients. Here, we studied the importance of the JAK and SYK pathways on their function in order to understand the mechanism leading to increased efficacy of dual JAKi + SYKi. FLS were isolated from small joints of mice with chronic arthritis (Treg-depleted). Stimulation of FLS with TNF and IL-17A, which are elevated in arthritic mice (Fig. 1 and Additional file 2) and in the synovium of RA patients [27, 28], induced IL-6 and MMP production (Fig. 6a). Dual JAKi + SYKi markedly reduced IL-6 secretion to a similar extent to JAKi. Milder effects were observed in MMP9 and MMP3 levels (not shown). In a transwell system with collagen-coated inserts, JAKi, SYKi and dual JAKi + SYKi decreased the invasive capacity of FLS by 13, 27 and 50 %, respectively (Fig. 6b). Therefore, JAK mediated cytokine production, and both JAK and SYK additively contributed to collagen invasiveness, a critical step in cartilage destruction.

In bone destruction, osteoclasts are the primary resorbing cells. Here, we found that G6PI-immunized mice treated with JAKi + SYKi had baseline counts of osteoclast precursors (CD11b^hiGr1⁻ cells) and levels of RANKL expression in spleens (Fig. 6c and d). Furthermore, in vitro osteoclastogenesis from bone marrow cells was completely prevented with SYKi and dual JAKi + SYKi exposure (Fig. 6e). We also exposed mature osteoclasts to these inhibitors and observed that SYK, but not JAK, is critically required for bone resorption activity (Fig. 6f). Thus, dual JAKi + SYKi treatment interferes at multiple stages in osteoclast development and function, primarily via SYK.

The JAK/STAT pathway is known to mediate T cell differentiation by interfering with cytokine signaling. Our data show that dual JAKi + SYKi prevented T cell proliferation upon unspecific stimulation in vitro (Fig. 7a). Antigen-specific responses, primarily IFNγ and IL-6 production, were also markedly decreased in splenocytes from arthritic mice treated in vitro with both inhibitors (Fig. 7b).

Cells of the immunological memory, in particular memory T cells and long-lived plasma cells are implicated in the chronification of disease and have proven to be refractory to conventional immunosuppression [29]. In order to investigate the role of JAK and SYK on memory cell maintenance, OVA-immunized mice were treated with inhibitors starting 2 weeks after challenge, a time when memory cells are already generated. The number of memory CD4⁺CD44^hi T cells was similarly decreased in mice treated with JAKi and dual JAKi + SYKi (Fig. 7c). Upon OVA re-exposure in vitro, JAKi + SYKi markedly reduced antigen-specific (CD40L⁺) CD4⁺ T cells and their ability to produce IFNγ, IL-2 and TNFα (Fig. 7d and e). OVA-specific memory plasma cells (CD138⁺κλ^hi) were also significantly reduced by an additive effect on JAK and SYK signaling (Fig. 7f). Thus, both JAK and SYK mediate memory cell pool maintenance and, thus, their inhibition could maintain disease remission and prevent arthritis recurrence.