The “evolving landscape” of Janus kinase (JAK) inhibitors has positioned them as a transformative therapeutic class for SpA spectrum disorders 1. As summarized in rheumatology reviews such as UpToDate, these agents are effective by broadly targeting cytokine signaling crucial for inflammation. Their clinical efficacy, demonstrated in numerous trials and real-world studies, including those in Taiwanese populations with ankylosing spondylitis 2, is well-established. A pivotal new study by Giryes et al. in Arthritis & Rheumatology now provides a deep, mechanistic explanation for this therapeutic “pro” 3. The authors confirm that JAK1 inhibition (via upadacitinib) powerfully suppresses the adaptive immune axis. The evidence is specific: it effectively blocks IFNγ-mediated STAT1 phosphorylation and, most critically, “strongly blocked” the production of pathogenic cytokines TNFα, IL-17A, and IL-17F from activated entheseal T cells 3. This finding provides a clear and direct molecular basis for the success of JAK inhibition in T-cell-driven articular inflammation. However, the study's primary importance lies in its discovery of a significant and paradoxical “con”: the simultaneous and potent activation of the innate immune system. Giryes et al. present striking evidence of this “paradoxical activation.” When entheseal myeloid cells were stimulated with LPS (a TLR4 agonist simulating bacterial challenge), pretreatment with upadacitinib led to an “exaggerated” pro-inflammatory surge 3. The quantitative evidence is unambiguous, showing an approximately 10-fold increase in IL-23 production (p < 0.001) and a similar 10-fold increase in TNFα (p < 0.0001) 3. This is not a minor fluctuation but a robust and potentially detrimental off-target effect. The authors meticulously dissect the mechanism behind this “con,” tracing it to the blockade of a critical homeostatic pathway: Interleukin-10 (IL-10) signaling 3. IL-10 is a vital anti-inflammatory cytokine that functions as a “brake” on myeloid cell activation, and it requires JAK1 and Tyk2 for its signal transduction. The study provides elegant proof: this paradoxical inflammation was induced by both the JAK1 inhibitor (upadacitinib) and the Tyk2 inhibitor (deucravacitinib), but was not seen with the JAK3 inhibitor (ritlecitinib) 3. In effect, by inhibiting JAK1 or Tyk2, these drugs inadvertently cut the brake line, removing the negative feedback IL-10 provides and allowing innate inflammatory responses to run unchecked. This mechanistic dichotomy—a beneficial suppression of T-cells (the “pro”) coupled with a detrimental activation of myeloid cells (the “con”)—offers a powerful explanation for the divergent clinical outcomes seen across the SpA spectrum. In the joint, where T-cell pathology is dominant, the “pro” of adaptive immune suppression likely outweighs the “con,” resulting in clinical success. However, in the gut, an organ highly dependent on innate immunity and IL-10-mediated tolerance, this equation is reversed. The “con” of blocking IL-10 and unleashing myeloid-driven IL-23 production could be disastrous. This mechanism provides a compelling molecular hypothesis for the high-profile clinical failure of the Tyk2 inhibitor deucravacitinib in phase 2 trials for ulcerative colitis 4. The Giryes et al. study is a critical reminder that JAK inhibitors are not a uniform panacea. Their effects are tissue-specific, and their therapeutic window is defined by a complex balance between adaptive and innate pathways. This highlights an urgent need for novel research strategies to dissect these complex, tissue-specific responses. Leveraging the complementary strengths of real-world evidence from large-scale platforms like TriNetX 5 and applying more rigorous causal inference methodologies 6 will be essential to truly understand and optimize the use of these powerful therapies for the correct patient and the correct disease. Hung-ke Lin: conceptualization, writing – original draft. Tang-Kai Cheng: conceptualization, writing – original draft. Meng-Che Wu: conceptualization, supervision, writing – review and editing. The authors declare no conflicts of interest. Data sharing not applicable to this article as no datasets were generated or analysed during the current study.
Lin et al. (Fri,) studied this question.