Abstract Background The number of advanced therapies (AdTs) for steroid-refractory ulcerative colitis (UC) has been increasing, but clear selection criteria remain lacking. To achieve personalized medicine, treatment decisions should be guided by each patient’s cytokine profile. This study aimed to develop a therapeutic selection model based on cytokine profiling using two novel serum biomarkers: leucine-rich α2 glycoprotein (LRG; a sensitive biomarker produced by diverse cytokines) and prohaptoglobin (proHp; a newly developed intestinal mucosa–derived biomarker). Methods The secretion mechanisms of LRG and proHp were investigated using patients’ peripheral blood mononuclear cells (PBMCs) and Caco-2 and THP-1 cell lines. A total of 272 UC patients enrolled in a prospective multicenter cohort study were analyzed, and serum cytokines, LRG, and proHp levels were measured. Cytokine-gradient mapping based on generalized additive models was applied to estimate cytokine activity on the LRG–proHp plots. Among 226 patients who received AdTs, the cohort was randomly divided (3:1) into training and validation sets. A dual-biomarker therapeutic map was constructed using a logistic regression model to predict treatment response to each AdT class, defined as a partial Mayo score ≤2 with all subscores ≤1 at Week 8. Predictive performance was assessed by comparing the model-matched versus unmatched groups using odds ratio (OR) analysis. Results In Caco-2 cells, proHp secretion was induced solely by IL-22 and was undetectable in THP-1 cells or PBMCs under any cytokine stimulation, indicating that proHp is an IL-22–inducible, epithelium-derived marker. In contrast, LRG was induced by distinct cytokines across all three cell types. In the LRG–proHp map of the training cohort, cytokine-gradient analysis showed that IL-8 and IL-17A formed diagonal gradients spanning both biomarkers, whereas IL-6 and IL-12/23 p40 aligned along the LRG axis and MCP-1 aligned along the proHp axis. The LRG-dominant region was associated with favourable response to anti-IL-12/23 agents, the proHp-dominant region to anti-integrin therapy, and the dual-high region to JAK inhibitors. In the treatment-selection model, the model-matched group achieved a significantly higher remission rate than the unmatched group (83.3% vs 55.7%; OR = 3.97, P = 0.003). Similar results were observed in the validation cohort (84.2% vs 50.0%; OR = 5.33, P = 0.012). Conclusion The combination of LRG and proHp provides a two-dimensional inflammatory signature that stratifies systemic and mucosal immune activity and aligns with treatment-responsive pathways. This dual-axis biomarker framework enables mechanism-informed treatment selection for UC and offers a practical step toward cytokine-guided precision medicine. Conflict of interest: Dr. Yoshihara, Takeo: Grants: Mochida Pharmaceutical Co., Ltd. Personal fees: Abbvie Inc. Sakon, Daisuke: No conflict of interest Asakura, Nobuhiko: No conflict of interest Amano, Takahiro: No conflict of interest Shinzaki, Shinichiro: Shinichiro Shinzaki has received grant support from AbbVie, EA Pharma, Nippon Kayaku, advisor fees from AbbVie, EA Pharma, Gilead Sciences, Janssen Pharmaceutical, Sekisui Medical and Takeda Pharmaceutical and honoraria from AbbVie, EA Pharma, Gilead Sciences, Janssen Pharmaceutical, Mitsubishi Tanabe Pharma, Nippon Kayaku, Pfizer, Sekisui Medical, and Takeda Pharmaceutical. Sawa, Hiroki: No conflict of interest Ogawa, Minyon: No conflict of interest Tsujii, Yuri: No conflict of interest Otake-Kasamoto, Yuriko: No conflict of interest Sakakibara, Yuko: No conflict of interest Yamada, Takuya: No conflict of interest Hiyama, Satoshi: No conflict of interest Nishida, Tsutomu: No conflict of interest Murayama, Yoko: No conflict of interest Uema, Ryotaro: No conflict of interest Tsujii, Yoshiki: No conflict of interest Inoue, Takahiro: No conflict of interest Kondo, Jumpei: No conflict of interest Iijima, Hideki: No conflict of interest Hayashi, Yoshito: No conflict of interest Suzuki, Takashi: No conflict of interest Miyoshi, Eiji: No conflict of interest
Yoshihara et al. (Thu,) studied this question.