Abstract Background Crohn’s disease (CD) is a highly heterogeneous disorder, and primary non-response to anti-TNF therapy remains a major clinical challenge. Reliable plasma biomarkers capable of predicting treatment outcome are lacking, and it is unclear whether different clinical subtypes of CD exhibit distinct molecular features associated with anti-TNF response. This study aimed to identify plasma biomarkers that predict primary non-response to anti-TNF therapy across CD subtypes. Methods We investigated a longitudinal cohort comprising 89 patients with Crohn’s disease and 30 non-IBD controls. A total of 162 plasma samples collected before and after anti-TNF therapy underwent 4D-label-free proteomic and untargeted-metabolomic profiling. Patients with Crohn’s disease were further stratified according to the Montreal classification (B1, B2, L1, L3, pCD) to enable subtype-specific analyses. Predictive models for therapeutic response were constructed using LASSO regression based on the integrated multi-omics features, and key biomarkers were validated by immunofluorescence. Results Thirteen candidate markers differentiated PNR from responders in the overall CD cohort (Fig. 1A-C). A seven-marker model (FCN1, TTR, CHGB, KHSRP, CIP2A, VASN and PFOA) achieved robust predictive performance with an AUC of 0.9232 (Fig. 1D). Immunofluorescence staining confirmed that CHGB, KHSRP, CIP2A, and VASN were significantly upregulated in non-responders (Fig. 1E). Stratified analyses according to the Montreal classification (B1, B2, L1, L3, pCD) revealed patterns of both shared and subtype-specific predictors. Some markers, such as CHGB, showed elevation across multiple subtypes, whereas several non-response-associated features exhibited subtype-specific enrichment (Fig. 2A-C), including NIF3L1, LZIC, and UCHL3 in L1; CALM1 and ESD in L3; PF4V1, CALR, and SAA4 in B1; BIN1 and COL4A2 in B2; and NEB and PCDHGC1 in pCD (Fig. 2C). Conclusion Primary non-response to anti-TNF therapy in CD is driven by a combination of subtype-specific molecular mechanisms and shared pathways. These findings underscore the need for subtype-tailored predictive biomarkers while recognizing common targets that may inform cross-subtype therapeutic strategies. Conflict of interest: Tian, Chuwen: No conflict of interest Huang, Lingjie: No conflict of interest Cao, Qian: No conflict of interest
Tian et al. (Thu,) studied this question.