Impact of JAK-inhibition on pain and biphasic P2X7R expression on CD4+ T cells in experimental arthritis Authors: Jérôme Biton, Roxane Hervé, Magali Breckler, Delphine Lemeiter, Patrice Decker, Johanna Sigaux, Eric Assier, Luca Semerano, Natacha Bessis, Marie-Christophe Boissier, Elodie Rivière Published: 2026 ABSTRACT Pain improvement is a major unmet need for patients with rheumatoid arthritis (RA). The purinergic receptor P2 X7R may play a central role in inflammatory pain. We aimed to assess the differential effect of Jak and TNF-α inhibition on pain in a collagen-induced arthritis (CIA) mouse model and to decipher mechanisms, focusing on P2X7R. Either a Jak-inhibitor (baricitinib, BARI) or an anti-TNF-α (etanercept, ETN) were administered to CIA mice in early or late phase of arthritis. Pain was evaluated with three independent tests. We assessed P2X7R on leucocytes and IFN-γ, IL-1β, TNF-α, IL-10, IL-6 and IL-17 A concentrations. BARI and ETN decreased joint inflammation and ameliorated histologic destruction and inflammation. When administered in the late phase, ETN and BARI exerted an analgesic effect. P2X7R expression on blood CD4+ T cells was increased in early phase BARI treated mice, and decreased in late phase. When treatments were administered early, percentage of P2 × 7R+ CD4+ T cells inversely correlated with clinical and histological scores, and with IL-6, IL-17 A and IL-1β levels. In conclusion, both BARI and ETN improved pain in a CIA model. The differential expression of P2X7R on CD4+ T cells according to the time of BARI administration may contribute to a better understanding of pain mechanisms in arthritis.Supplementary InformationThe online version contains supplementary material available at 10.1038/s41598-026-54494-0. Introduction Inflammatory rheumatic diseases are frequent pathologic conditions characterized by chronic inflammation, painful disabilities of joints and various systemic features. Among them, rheumatoid arthritis (RA) is the most frequent, with a prevalence of 0.51–0.56%1. During RA, joint involvement consists of chronic inflammation, bone, and cartilage destruction, resulting in pain and disability. A major breakthrough was the development of treatments that target central mediators of inflammation, such as TNF-α, IL-6, or IL-1, or immune cells involved in the disease (activated T cells, B cells). These treatments have demonstrated efficacy in reducing inflammation, protecting against joint destruction and represent the standard of care for RA treatment2. More recently, the Janus-associated kinase (JAK) signaling pathway was highlighted as a major target for chronic inflammatory and autoimmune diseases treatment3. Several JAK inhibitors (jakinhibs) have been approved for the treatment of several chronic inflammatory diseases, such as RA in early and long-standing phase2. From the patient’s perspective suffering from inflammatory arthritis, pain is the main symptom requiring treatment4. Pain negatively affects the quality of life. Despite recent pharmacologic advances, pain which is mediated by inflammatory and non-inflammatory mechanisms, remains frequently uncontrolled5. Thus, there is a need for developing therapeutic tools acting on both pain and inflammation. Clinical observations and treatment outcomes show that direct inflammatory pathways seem to predominate in early-stage arthritis. By contrast, chronic long-standing arthritis might involve a neurogenic participation6. Interestingly, in the context of intestinal inflammation, post-inflammatory hypersensitivity symptoms have been described. The mechanisms of post-inflammatory persistent visceral pain involve peripheral and central neuroplastic changes, low-grade chronic inflammation that sensitizes visceral afferent pathways and sensitization of non-neuronal resident cells of the gut7. Inflammatory cytokines signaling via JAK pathways, such as IL-6, play a major role in pain mechanisms8. Moreover, a clinical study suggests that therapeutic inhibition of either JAK or TNF-α might exert distinct effects on pain improvement in RA9. Interestingly, inhibition of the JAK/STAT signaling pathway with specific JAK inhibitors has the potential to modulate pain in patients with RA. Baricitinib (BARI) is a JAK1/JAK2 inhibitor, that acts as a competitive inhibitor of adenosine triphosphate, transiently preventing the phosphorylation and activation of JAK, therefore, interrupting signal transducer and activator of transcription (STAT) downstream signaling. BARI competes with ATP at the JAK kinase domain, not at ATP-gated receptors (such as P2X7R). Because multiple cytokines are involved in pain signaling directly and indirectly through the JAK/STAT pathway, JAK inhibition might be more effective in reducing pain than would be a single cytokine inhibitor10. Randomized controlled trials and real-world settings on a large number of patients with RA evaluating tofacitinib or baricitinib have shown a rapid effect on the pain component11,12. A post-hoc analysis of the phase 3 clinical trial (RA-BEAM)13 showed that patients with RA treated with a JAK inhibitor achieved significantly greater improvements in pain than those treated with an anti-TNF-α. Moreover, in another study, patients treated with BARI reported greater and more rapid pain relief than patients treated with MTX alone over 1 year14. JAK inhibitors may have a dual therapeutic role by modulating inflammation and nociception, which leads to clinical benefits including reduction of pain beyond that related to inflammation15. Recently, Makabe et al. showed that BARI ameliorates inflammatory and neuropathic pain in collagen antibody-induced arthritis mice by modulating the IL-6/JAK/STAT3 pathway and CSF-1 expression in dorsal root ganglion neurons16. P2X 7R (Purinergic Receptor P2 × 7) is a purinergic receptor implicated in neuropathic pain, expressed by most immune cells, including the monocyte-derived cell lineages, T and B cells and their precursors17. Moreover, fibroblasts have also been shown to express P2X7R18. P2X7R is mainly activated by extracellular ATP. P2X7R activation triggers the production and release of pro-inflammatory cytokines from macrophage, microglia, dendritic cells and monocytes. P2X7R also mediates neuropathic pain19,20. Some experiments suggest that the critical role of the P2X7R in the enhanced nociceptive transmission during neuropathic (and also inflammatory) pain is associated with microglial activation and secretion of IL-1β in the dorsal horn of the spinal cord21. P2X7R is involved in numerous pathways leading to NLRP3 inflammasome activation, cytokine and chemokine release, T cells survival and differentiation, transcription factor activation, and cell death22. Activation of P2X7R promotes the release of pro-inflammatory factors, such as IL-1β, IL-6, CCL2, TNF-α, CCL3, IL-18 and nitric oxide, suggesting that P2X7R might play a pivotal role in processes of inflammation and pain23. In a rat model, inhibition of P2X7R led to reduced articular inflammation and erosive progression24. In addition, the P2X7R antagonist OXATP alleviates pain in arthritis model induced by complete Freund’s adjuvant (CFA) and decreases edema in the inflamed paw25. In the present study, we aimed to assess the consequences of JAK inhibition using BARI, a potent JAK1/2 inhibitor, and of TNF-α inhibition, using etanercept (ETN), in a murine collagen-induced arthritis model. Two different experiments were conducted: the “late experiment” (BARI or ETN administration after the onset of arthritis) and the “early experiment” (BARI or ETN administration before the onset of arthritis). We showed that BARI and ETN decreased clinical joint inflammation and ameliorated histological scores of joint destruction and inflammation, both when administered in the early and the late phase of arthritis. ETN and BARI exerted an analgesic effect, assessed by pinch algometer and postural pain methods when administered in the late phase. Lastly, P2X7R expression at the surface of blood CD4+ T cells was different according to the phase of BARI administration. Results Baricitinib and etanercept both reduce arthritis, pain and joint destruction when administered in late phase arthritis As expected, arthritis occurred in 90 to 100% of mice in each group. Arthritis occurred between day 22 and 40 and were fully established when treatments were started (Fig. 1A). Individual analysis of each day of arthritis scoring has shown a late improvement of arthritis in BARI as well as ETN treated mice compared to control (Fig. 1A). At D60, the comparison of the arthritis clinical scores showed lower clinical scores in the groups treated with BARI and ETN compared to controls (Fig. 1B). The same comparison at D67 showed lower arthritis clinical scores in the group treated with BARI compared to controls (Fig. 1C). Fig. 1Clinical and histological effects of ETN and BARI treatments when administered at the late phase of collagen-induced arthritis. (A) Arthritis clinical scores all along the experiment in mice treated late with BARI, ETN or vehicle. Statistically significant differences were observed between BARI and vehicle, and between ETN and vehicle. (B) Clinical arthritis score at D60 in the 3 groups of treatment. (C) Clinical arthritis score at D67 in the 3 groups of treatment. (D) Histological characterization of posterior paws from representative mice (haematoxylin-eosin staining) in the different conditions of treatment. (E) Score of inflammation including synovitis, synovial proliferation, and cell infiltration evaluation at D68 in the 3 groups of treatment. (F) Score of destruction, including bone and cartilage thickness, and erosions at D68 in the 3 groups of treatment. In A-C, E and F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. Clinical and histological effects of ETN and BARI treatments when administered at the late phase of collagen-induced arthritis. (A) Arthritis clinical scores all along the experiment in mice treated late with BARI, ETN or vehicle. Statistically significant differences were observed between BARI and vehicle, and between ETN and vehicle. (B) Clinical arthritis score at D60 in the 3 groups of treatment. (C) Clinical arthritis score at D67 in the 3 groups of treatment. (D) Histological characterization of posterior paws from representative mice (haematoxylin-eosin staining) in the different conditions of treatment. (E) Score of inflammation including synovitis, synovial proliferation, and cell infiltration evaluation at D68 in the 3 groups of treatment. (F) Score of destruction, including bone and cartilage thickness, and erosions at D68 in the 3 groups of treatment. In A-C, E and F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. Histological evaluation showed statistically significant differences between control groups and BARI or ETN groups in terms of inflammation and joint destruction (Fig. 1D, E and F). Moreover, histological inflammation (D68) strongly correlated with histological destruction (r = 0.96, p < 0.0001). Latest clinical evaluation of arthritis score (D67) correlated with histological inflammation (r = 0.58, p < 0.01) and destruction (r = 0.548, p < 0.05) (Supp Fig. 3). We evaluated pain using pinch algometer to measure hyperalgesia. ETN and BARI treatments exerted an analgesic effect observed at the late phase (D67, i.e., 21days after starting the treatment) compared to control (Fig. 2A). Postural pain was also improved by both treatments. This was statistically significant for BARI at D67 on arthritic paws (Fig. 2B). Lastly, allodynia, evaluated using Von Frey filaments, was not modified by any treatment (data not shown). The hyperalgesia scores evaluated using the pinch algometer correlated with arthritis clinical score, and histological destruction scores (Fig. 2C and D). Fig. 2Pain evaluation in the late phase experiment. (A) Analysis of the hyperalgesia score of arthritic paws at D67, normalized to D46 measured scores (starting time of the treatment). (B) Analysis of the postural pain score of arthritic paws at D67, normalized to D46 measured scores (starting time of the treatment). The lowest values correspond to the highest pain. (C-D) Correlation between global hyperalgesia (meaning that the pain score of each mouse is the mean of the 4 paws) and clinical arthritis score (C), and destruction histological score (D). In A-B data are expressed as mean ± SEM. A non-parametric test (Mann-Withney) was used to compare data across the different groups of treatment. In C-D correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05. Pain evaluation in the late phase experiment. (A) Analysis of the hyperalgesia score of arthritic paws at D67, normalized to D46 measured scores (starting time of the treatment). (B) Analysis of the postural pain score of arthritic paws at D67, normalized to D46 measured scores (starting time of the treatment). The lowest values correspond to the highest pain. (C-D) Correlation between global hyperalgesia (meaning that the pain score of each mouse is the mean of the 4 paws) and clinical arthritis score (C), and destruction histological score (D). In A-B data are expressed as mean ± SEM. A non-parametric test (Mann-Withney) was used to compare data across the different groups of treatment. In C-D correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05. Baricitinib and etanercept administered in the early phase of collagen-induced arthritis decrease clinical arthritis scores but have no effect on pain evaluation scores We then evaluated the influence of TNF-α blockade or Jak inhibition in the early phase of CIA. Treatments (or vehicule) were administered from D11 after CII priming. As expected, ETN and BARI exerted anti-inflammatory effect as shown by a significant reduction of the clinical arthritis scores (Fig. 3A), of the maximal clinical arthritis score reached by animals (Fig. 3B) and a significant reduction of the area under curve calculation of clinical score (Fig. 3C). At this early phase of collagen-induced arthritis treatment, no effect of the different treatments was observed on pain evaluation (data not shown). However, the histological evaluation showed statistically significant differences between control groups and BARI or ETN groups in terms of inflammation and joint destruction (Fig. 3D, E and F). Lastly, plasma detection of anti-CII antibodies was decreased in the ETN and BARI treated groups compared to the control group. This difference was not observed in the late phase experiment (Supp Fig. 4). Fig. 3Clinical and histological effects of ETN and BARI treatments when administered at the early phase of collagen induced arthritis. (A) Arthritis clinical scores evolution from the day of first immunization to the end of the experiment. (B) Maximal clinical arthritis score (Amax) reached by each mouse in the 4 different groups. (C) Comparison of the AUC of clinical scores between the 4 groups of treatment. (D) Histological characterization of posterior paws from representative mice (haematoxylin-eosin staining). (E) Score of inflammation including synovitis, synovial proliferation, and cell infiltration evaluation at D39 in the 4 groups of treatment. (F) Score of destruction, including bone and cartilage thickness, and erosions at D39 in the 4 groups of treatment. In A-C and E-F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. Clinical and histological effects of ETN and BARI treatments when administered at the early phase of collagen induced arthritis. (A) Arthritis clinical scores evolution from the day of first immunization to the end of the experiment. (B) Maximal clinical arthritis score (Amax) reached by each mouse in the 4 different groups. (C) Comparison of the AUC of clinical scores between the 4 groups of treatment. (D) Histological characterization of posterior paws from representative mice (haematoxylin-eosin staining). (E) Score of inflammation including synovitis, synovial proliferation, and cell infiltration evaluation at D39 in the 4 groups of treatment. (F) Score of destruction, including bone and cartilage thickness, and erosions at D39 in the 4 groups of treatment. In A-C and E-F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. Fig. 4Expression of P2X7R by immune cells. (A-C ) Percentage of P2X7R+ cells among CD3+ (A), CD4+ T cells (B) and monocytes (C) from blood at D68 (late phase experiment). (D-F) MFI of P2X7R in CD3+ T cells (D), CD4+ T cells (E) and monocytes (F) from blood at D68. (G) Correlation between the percentage of P2X7R+ among CD4+ cells and the arthritis score. (H) Correlation between the MFI of P2X7R+ among CD4+ cells and the arthritis score (H). In A-F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. In G-H correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01. Expression of P2X7R by immune cells. (A-C ) Percentage of P2X7R+ cells among CD3+ (A), CD4+ T cells (B) and monocytes (C) from blood at D68 (late phase experiment). (D-F) MFI of P2X7R in CD3+ T cells (D), CD4+ T cells (E) and monocytes (F) from blood at D68. (G) Correlation between the percentage of P2X7R+ among CD4+ cells and the arthritis score. (H) Correlation between the MFI of P2X7R+ among CD4+ cells and the arthritis score (H). In A-F data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. In G-H correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01. P2X7R is differentially expressed on CD4+ T cells depending on the time of BARI administration Flow cytometry was performed on blood cells collected at D68. We showed that the mean fluorescence intensity (MFI) of P2 × 7R among CD3+ T cells and CD4+ T cells was significantly reduced in BARI treated animals (Fig. 4D and E). ETN failed to influence the expression of P2X7R on CD4+ T cells when administered at this phase (percentage and MFI). Expression of P2X7R on monocytes (Ly6C+ cells) was not modified with any of these therapeutic interventions (Fig. 4C and F). The percentage of P2X7R+ cells among CD4+ T cells as well as P2X7R MFI on CD4+ T cells correlated with the arthritis clinical score (r = 0.42, p < 0.05 and r = 0.64, p < 0.001, respectively) (Fig. 4G and H). In a second experiment, in which the treatments were administered at the early phase, we assessed P2X7R surface expression on immune cells, in blood and spleen at D39. We showed a significant effect of JAK inhibition by BARI on P2X7R expression at the surface of immune cells in comparison with control group (vehicle-treated mice). In the blood, there was an increase of P2X7R expression on monocytes (Ly6C high, Ly6G neg), as well as on CD4+ T cells in the BARI treated group compared to the ETN treated group and the control group (Fig. 5A). There was no difference for P2X7R expression between the groups in CD8+ T cells, B cells and neutrophils (Fig. 5A) (data not shown). In the spleen, P2X7R expression was increased on CD4+ T cells in the BARI treated group compared to control and to ETN groups (Fig. 5B). Interestingly, the frequency of P2X7R+ cells among CD4+ T cells inversely correlated with clinical arthritis scores (r= -0.47, p < 0.01), and with histological scores of inflammation (r= -0.64, p < 0.0001) and destruction (r= -0.58, p < 0.001) (Fig. 5C, D, E, respectively). In spleen and blood, no correlation was observed between the frequency of P2X7R+ cells among CD4+ T cells and pain evaluation scores (data not shown). Fig. 5P2X7R expression on immune cells from blood (A) and spleen (B) at D39 (early phase) in CD4+ T cells, CD8+ T cells and monocytes (Ly6C+ cells). Correlation between the frequency of P2 × 7R+ cells among CD4+ T cells and arthritis clinical score (C), histological inflammation score (D) and histological destruction scores (E). In A-B data are expressed as mean ± SEM. Anova comparison test was used to compare the different groups of treatment. In C-E correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. P2X7R expression on immune cells from blood (A) and spleen (B) at D39 (early phase) in CD4+ T cells, CD8+ T cells and monocytes (Ly6C+ cells). Correlation between the frequency of P2 × 7R+ cells among CD4+ T cells and arthritis clinical score (C), histological inflammation score (D) and histological destruction scores (E). In A-B data are expressed as mean ± SEM. Anova comparison test was used to compare the different groups of treatment. In C-E correlations used the non-parametric Spearman test. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. Modulation of cytokines by ETN and BARI and correlation with clinical and biological factors Plasma levels of several cytokines including IFN-γ, IL-1β, TNF-α, IL-10, IL-6 and IL-17 A were assessed in the late (Fig. 6A) and the early phase experiments (Fig. 6B). BARI increased IFN-γ level in both experiments, whereas ETN had no effect on IFN-γ level. In the early phase experiment, BARI and ETN were also responsible for a decrease of IL-1β, while only BARI induced a decrease of IL-1β in the late phase experiment. TNF-α was increased in sera from mice treated with ETN in the early phase experiment. Lastly, IL-6 was decreased in mice treated with ETN in the late phase experiment and there was a non-significant trend for a decrease induced by BARI. Fig. 6IFN-γ, IL-1β, TNF-α, IL-10, IL-6 and IL-17 A assessment in sera from CIA mice treated with ETN or BARI in the late (A) and early (B) phases experiments. In A-B data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. IFN-γ, IL-1β, TNF-α, IL-10, IL-6 and IL-17 A assessment in sera from CIA mice treated with ETN or BARI in the late (A) and early (B) phases experiments. In A-B data are expressed as mean ± SEM. According to data distribution, a parametric test (ANOVA) or a nonparametric test (Kruskal–Wallis) were used to compare data across the different groups of treatment. * corresponds to p ≤ 0.05, ** corresponds to p ≤ 0.01, ***corresponds to p ≤ 0.001. When treatments were administered early, the frequency of P2X7R+ cells among CD4+ T cells inversely correlated with IL-6 (r= -0.44, p < 0.05), IL-17 A (r= -0.41, p < 0.05) and IL-1β (r= -0.37, p < 0.05) plasma levels (Fig. 7A, B, C). IFN-γ plasma level positively correlated with the frequency of P2X7R+ cells among CD4+ T cells (r = 0.45, p = 0.01) (Fig. 7D). No correlation was observed between P2 × 7R+ among CD4+ T cells and cytokines in the late phase experiment. Fig. 7Correlation between the percentage of P2 X7R+ cells among CD4+ T cells and IL-6 (A), IL-17 A (B), IL-1β (C) and IFN-γ (D) cytokines MSD dosage at early phase experiment. Correlations used the non-parametric Spearman test. Correlation between the percentage of P2 X7R+ cells among CD4+ T cells and IL-6 (A), IL-17 A (B), IL-1β (C) and IFN-γ (D) cytokines MSD dosage at early phase experiment. Correlations used the non-parametric Spearman test. Discussion In this study, we showed the improvement of inflammatory arthritis after BARI administration in CIA. This effect was comparable to those observed with ETN, even when JAK inhibitor was given after the onset of the disease (late phase treatment). We also demonstrated that BARI, as ETN administration, resulted in pain improvement at this late phase of treatment. However, no effect was observed on pain improvement when treatments were administered early. Lastly, P2X7R was differentially expressed at the surface of CD4+ T cells after treatment with BARI depending on the phase of treatment administration. In the early phase experiment, P2X7R expression was increased by BARI on CD4+ T cells from blood and from the spleen, whereas in the late phase experiment, P2 × 7R expression was decreased by BARI on CD4+ T cells from blood. There was no effect of ETN on P2 X7R expression. At cytokine level, BARI induced, both in the early and the late phase experiments an increase of IFNγ and a decrease of IL-1β. Pain in RA involves multiple driving mechanisms that may differentially occur over time. Quantitative assessment of pain in animal models of arthritis is technically and conceptually challenging26. In this study, we developed evaluation tools dedicated to quantify pain parameters in a mouse experimental model of RA. We used the CIA model because it involves the actors that are known to be involved in human RA pathogenesis such as neutrophils, dendritic cells, T and B cells and inflammatory cytokines including TNF-α. Interestingly, we showed in this study a strong correlation between the clinical arthritis scores and the histological parameters including inflammation and destruction. Such results are in line with previous studies. For example, Tanimoto et al. showed in a rat collagen-induced model that the administration of JTE-052, a JAK inhibitor, ameliorated articular inflammation and joint destruction even when administered after the onset of arthritis27. We also showed a correlation between histological score of inflammation or destruction and pain evaluation according to the pinch algometer measure at the late phase experiment. Interestingly a recent study from Simon et al. showed in a collagen antibody-induced arthritis (CAIA) model that BARI reduced joint inflammation, with most pronounced antinociceptive effects during the late phase of the CAIA-model, when etanercept was ineffective. Moreover, unlike ETN, BARI reduced pain-like behaviour and synovial hyper-innervation when administered exclusively in the late phase. These results suggest that BARI has antinociceptive effects by targeting not only immune cells but also neurons and glia cells via inhibition of signaling pathways linked to chronic pain28. P2X7R may play an important role in the development of chronic pain and inflammation. Here, we showed a biphasic expression of P2X7R on CD4+ T cells, depending on the phase of BARI treatment administration (late phase vs. early phase). Further studies are required to better understand the mechanisms and consequences at cellular level, as well as to precise whether observed effects are secondary to reduced inflammation or reflect a direct anti-nociceptive effect. Of note, P2X7R is cleaved from the surface of activated T cells by metalloproteases (MMP)29. Given that baricitinib inhibits JAK1/2 signaling and thereby suppresses cytokine-driven T-cell activation30, this treatment might indirectly reduce MMP activity downstream of inflammatory signaling. Given this hypothesis, the apparent increase in surface P2X7R observed during the early phase of baricitinib treatment might not reflect true upregulation of receptor expression, but rather reduced receptor shedding due to diminished T-cell activation. Thus, decreased receptor shedding under JAK inhibition could represent an alternative explanation for the early increase in surface P2X7R. Molecular mechanisms involved in P2X7R regulation remain to be identified, and surface P2X7R expression may not accurately reflect receptor function. This should be of interest in the perspective of targeting this receptor to control pain in arthritis. Interestingly, Fan et al. showed in a CIA mouse model that blocking P2X7R signaling using P2X7R antagonist (Suramin or A-438079) attenuated CII -mediated joint damage31. They also showed in CII-treated dendritic cells (DCs) and DC/CD4+ T cell co-culture system, that pretreatment with pharmacological antagonists of P2X7R inhibited the production of Th17-promoting cytokines (IL-1β, TGF-β1, IL-23p19 and IL-6), but not of Th1 promoting cytokines (IL-12p40 and IL-12p35)28. This is in line with the results of our study, showing an inverse correlation between P2 × 7R expression on CD4+ T cells and IL-1β, IL-6 and IL-17 levels, whereas P2X7R expression on CD4+ T cells positively correlated with IFN-γ concentration in the early phase experiment. However, CD4⁺ T cell subset composition was not characterized in this study and may represent a potential confounding factor. Of note, P2 X7R is one of the most potent activators of the NLRP3 inflammasome, and therefore of caspase-1 cleavage and of mature IL-1β release20. Li et al. showed that Th17 cells and P2X7R expression was loosely correlated with IL-1β in patients with systemic lupus or RA32. We observed a link between P2X7R expression on CD4+ T cells and IL-17 only in the early phase experiment. IL-17 A has been shown to be present in the inflamed joints of patients with RA and to exert pro-inflammatory effects. IL-17 A may play distinct or differential roles in early versus late disease in experimental arthritis models. While Th17 cells may be present even before onset of disease, IL-17 A may play a more prominent role in the erosive stages of disease33,34. However, selective P2X7R antagonists have failed to demonstrate clinical efficacy in human rheumatoid arthritis. Both AZD9056 and CE-224,535 showed no significant clinical benefit in controlled RA trials despite a preclinical rationale35,36. We observed an increase of IFN-γ level in the mice treated with BARI but not in the ETN treated mice. This result could appear as counterintuitive, given that BARI has been shown to be effective in the treatment of auto-inflammatory interferonopathies37. However, a biphasic effect of IFN-γ has already been showed in a previous study showing that IFN-γ may play a critical role during both the induction and the course of CIA, first enhancing the immune response, and then regulating the arthritis process38. Here also, as for IL-17 A, the correlation between IFN-γ level and P2X7R expression by CD4+ T cells was observed only in the early phase experiment. Moreover, although BARI blocks IFN-γ signaling but does not necessarily reduce IFN-γ production; paradoxically, network feedback mechanisms may even lead to increased circulating IFN-γ levels despite attenuated downstream signaling. Lastly, the treatment with BARI in the early phase experiment induced a decrease of CII antibodies compared to ETN treatment or controls. This was not observed when ETN or BARI were administered after the complete immunization of the mice. This result suggests an effect of BARI in the early phase of immunization, preventing the anti-CII antibodies in this model35. In conclusion, pain remains frequently uncontrolled in RA. JAK or TNF-α inhibitors might exert distinct effects on pain improvement. P2X7 receptor-driven inflammatory mediators play a key role in the development of pain. This study showed in a collagen induced arthritis mouse model, that JAK inhibition, but not TNF-α inhibition, is associated with modifications of P2X7R expression at the surface of blood CD4+ T cells, depending on the phase of treatment administration (before or after the onset of arthritis). All together, these results suggest potential cellular mechanisms involved in response to treatments, involving P2X7R, although this study did not include functional assessments beyond surface expression. Methods Mice 60 male DBA/1mice (8 weeks old) were provided by Janvier Labs (Le Genest-Saint-Isle, France). The local Ethics Committee on Animal Care and Experimentation (#25279) approved all experimental procedures related to animals. All in vivo experiments were conducted in accordance with the recommendations for animal experimentation issued by the Institutes of Laboratory Animal Resources committee and the ARRIVE guidelines. Arthritis was induced with subcutaneous injection of a 1:1 emulsion of bovine type II collagen (CIIb, 50 µg/mice, MD Bioproducts, Switzerland) and complete Freund’s adjuvant (MD Bioproducts, Switzerland) at day 0 (total volume injected in the tail: 100 µl)39. A booster subcutaneous injection was given on day 21 with 100 µl of a 1:1 emulsion of CIIb (50 µg/mice) and incomplete Freund’s adjuvant (IFA, BD, France). Mice were weighted once a week. Prior to the experiments, the mice had a mean weight of 20.4 g (range: 18–24 g). In the case of a weight loss of more than 15% of their initial weight over two consecutive measurements, mice were euthanized. At the end of the procedures, the animals received an injection of the anesthetic ketamine (80 mg/kg)/xylazine (12 mg/kg) at a dose of 100 µl/10 g of body weight, followed by an intracardiac puncture of about 1 ml of blood. Euthanasia was carried out by cervical dislocation. Clinical evaluation A blinded procedure was used to monitor clinical arthritis three times per week. Arthritis was monitored in all four paws. For each mouse, the clinical severity of arthritis was scored as 0: normal; 1: erythema; 2: swelling; 3: deformity; and 4: ankylosis in 10 joints, or groups of joints: three joints of the two hind limbs (toes, tarsus, ankle) and two joints of the two forelimbs (digits, wrist)40. The maximum score for each of the 10 joints was 4, so the maximum individual arthritis clinical score on a given day was 40. The mean arthritic score on each day of clinical observation was calculated for each group of treatment. Pain evaluation Pain was evaluated for each mouse once a week, since the beginning of the experiment until euthanasia. Three methods were used to assess pain level: Pinch algometer was used to measure hyperalgesia. This test involves application of an increasing mechanical force to the surface of the paw. Mechanical pressure is applied focally to surface of the hind paw. The pressure is then increased at a constant rate until a nociceptive behavioral response is observed. The withdrawal response is detected visually by the researcher. One limit is the subjective measurement of the threshold. A result showing lower values means a lower threshold, i.e. more painful paw. The hyperalgesia score was calculated as 1/force applied. Thus, hyperalgesia appears as a higher value.Static Weight Bearing Test to evaluate the spontaneous pain. The mouse is placed on an inclined holder with the hind paws resting on two separate pressure sensors. The weight distribution between the hind paws is recorded. A result showing lower values means more painful instability resulting from pain. Analgesic effect appears as a higher value with a certain degree of bias in this model (bilateral hind paw arthritis results in unpredictable effects on static Weight Bearing).Von Frey filaments test to evaluate allodynia41. The mouse is placed in a small cage with a penetrable bottom. A monofilament is applied perpendicularly to the plantar surface of the hind paw until it buckles, delivering a constant pre-determined force. A response is considered positive if the animal exhibits any reaction to pain behaviors, including brisk paw withdrawal, licking, or shaking of the paw during application of the stimulus. A result showing lower values means a pain induction with smaller filament. Analgesic effect appears as a higher value. Pinch algometer was used to measure hyperalgesia. This test involves application of an increasing mechanical force to the surface of the paw. Mechanical pressure is applied focally to surface of the hind paw. The pressure is then increased at a constant rate until a nociceptive behavioral response is observed. The withdrawal response is detected visually by the researcher. One limit is the subjective measurement of the threshold. A result showing lower values means a lower threshold, i.e. more painful paw. The hyperalgesia score was calculated as 1/force applied. Thus, hyperalgesia appears as a higher value. Pinch algometer was used to measure hyperalgesia. This test involves application of an increasing mechanical force to the surface of the paw. Mechanical pressure is applied focally to surface of the hind paw. The pressure is then increased at a constant rate until a nociceptive behavioral response is observed. The withdrawal response is detected visually by the researcher. One limit is the subjective measurement of the threshold. A result showing lower values means a lower threshold, i.e. more painful paw. The hyperalgesia score was calculated as 1/force applied. Thus, hyperalgesia appears as a higher value. Static Weight Bearing Test to evaluate the spontaneous pain. The mouse is placed on an inclined holder with the hind paws resting on two separate pressure sensors. The weight distribution between the hind paws is recorded. A result showing lower values means more painful instability resulting from pain. Analgesic effect appears as a higher value with a certain degree of bias in this model (bilateral hind paw arthritis results in unpredictable effects on static Weight Bearing). Static Weight Bearing Test to evaluate the spontaneous pain. The mouse is placed on an inclined holder with the hind paws resting on two separate pressure sensors. The weight distribution between the hind paws is recorded. A result showing lower values means more painful instability resulting from pain. Analgesic effect appears as a higher value with a certain degree of bias in this model (bilateral hind paw arthritis results in unpredictable effects on static Weight Bearing). Von Frey filaments test to evaluate allodynia41. The mouse is placed in a small cage with a penetrable bottom. A monofilament is applied perpendicularly to the plantar surface of the hind paw until it buckles, delivering a constant pre-determined force. A response is considered positive if the animal exhibits any reaction to pain behaviors, including brisk paw withdrawal, licking, or shaking of the paw during application of the stimulus. A result showing lower values means a pain induction with smaller filament. Analgesic effect appears as a higher value. Von Frey filaments test to evaluate allodynia41. The mouse is placed in a small cage with a penetrable bottom. A monofilament is applied perpendicularly to the plantar surface of the hind paw until it buckles, delivering a constant pre-determined force. A response is considered positive if the animal exhibits any reaction to pain behaviors, including brisk paw withdrawal, licking, or shaking of the paw during application of the stimulus. A result showing lower values means a pain induction with smaller filament. Analgesic effect appears as a higher value. Histological evaluation For histology, hind paws of each mouse were collected, fixed, decalcified, dehydrated, and transferred to paraffin blocks. Slides of 4 μm thickness were stained with hematoxylin and eosin before optical microscopy observation. At least 3 fields per section were performed with blinding to treatment. In each joint, two variables were separately assessed on a 4-point scale (0–3, 0 indicating a normal joint, and 3, maximally severe arthritis). The first variable was inflammation, as reflected by synovial membrane thickness (synovial proliferation) and inflammatory cell infiltration. The second variable was joint destruction (bone erosions, cartilage thickness, and cartilage unevenness)40. Treatments with etanercept, baricitinib and control (DMSO vehicle) Treatments were administered as follows: BARI was diluted in 10% DMSO, and administered at 20 mg/kg, using sub-cutaneous injection, six times a week. The dissolution was achieved after careful vortexing, and the solution was injected immediately after dissolution. ETN was diluted in PBS, and administered at 30 mg/kg, using intra-peritoneal injections, two times a week. The control group received injections of 10% DMSO diluted in PBS, using sub-cutaneous injections, six times a week. In the “late phase” experiment, the treatments were administered from D46 until D67 (Supp Fig. 1A). Each group of treatment was composed by 10 animals. Blood sampling for cytokines assessment on plasmas and P2X7R expression on immune cells were collected at D68. Hind paws sampling was collected for histology at D68. In the “early phase” experiment, the treatments were administered from D11 until D38 (Supp Fig. 1B). Each group of treatment was composed by 10 animals. Three additional mice, which were not immunized to induce CIA, constituted the naive group. These mice were used to assess the habituation (advance reactions) of mice to pain measuring devices in the absence of pathologic condition. Blood sampling for cytokines assessment on plasmas and P2X7R expression on immune cells were collected at D39. Spleen was also collected for P2 × 7R assessment on immune cells. Hind paws sampling was collected for histology at D39. Flow cytometry Blood with heparin was collected for each mouse by intracardiac puncture at the end of the experiment. Leukocytes from the spleen were prepared using a homogenizer. Freshly isolated mononuclear cell viability was systemically checked using a Luna-FL cell counter (Logos Biosystems), a cell viability threshold higher than 98% was used to perform the flow cytometry experiments. Cells were incubated for 30 min at 4 °C in PBS 0.5 mM of EDTA containing 2% of human serum to block the Fcg receptors. Cell surface were stained with appropriate dilutions of various monoclonal antibodies. Flow cytometry acquisition was performed with the BD FACSCanto (BDbiosciences). The expression of P2X7R (PE anti-mouse P2X7R Antibody, clone 1F11, Biolegend) was analyzed on CD3+ (BD) positive, CD4+ (Biolegend) T cells and monocytes (Ly6C highly positive cells, BD) in the late experiment. FMO controls were used to define P2X7R positivity. All experiments were performed on ice, with staining and centrifugation steps carried out at 4 °C. In the early phase experiment, CD4+ T cells, CD8+ (Biolegend) T cells, B cells (CD19+ cells, BD), monocytes (Ly6C highly positive cells (BD) and neutrophils (Ly6G+ cells, BD) were analyzed. These analyses were performed in blood for the late phase experiment and in blood and spleen for the early phase experiment. The gating strategy is presented in Supp Fig. 2. Plasma anti-collagen II antibodies and cytokine assays Blood was collected for each mouse by intracardiac puncture. Plasmas were aliquoted and frozen for future analysis, including cytokine assays. Anti-CII antibodies were assessed using the enzyme linked immunosorbent assay (ELISA) method after dilution 1:10 000. The results are expressed as the ratio of the tested sample to a positive control. Experiments were performed according to the usual ELISA protocol42. Briefly, after CII adsorption on microtitration plates, mice sera, diluted at 1/10,000 was incubated on coated plates. Anti-cCII IgG antibodies were detected with alkaline phosphatase‐linked rat anti‐mouse IgG antibody (Sigma‐Aldrich, Saint‐Quentin Fallavier, France) after incubation with its para‐nitrophenylphosphate substrate (Sigma‐Aldrich) and read at 405 nm. Sera of CIA mice with severe disease from previous experiments was used as a positive control. IFN-γ, IL-1β, TNF-α, IL-10, IL-6 and IL-17 A plasma levels were assessed using Mesoscale Discovery technology (U-PLEX Mouse Custom Biomarker Group 1 (ms) Assays). Experiments were performed according to the manufacturer recommendations. Statistical analysis According to data distribution, a parametric test (ANOVA, Student test) or a nonparametric test (Kruskal–Wallis, Mann–Whitney U) with appropriate post hoc comparisons were used to compare data across the different groups. Clinical scores curves were compared with ANOVA. A p-value of < 0.05 was considered statistically significant. Correlations used the non-parametric Spearman test. All statistical analyses were performed using Prism software. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1 Supplementary Material 1