Abstract Background Ulcerative colitis (UC) therapies lead to variable remission in participants in clinical trials likely due to interindividual variability, differences in active biological pathways, feedback, and/or resistance mechanisms. We sought to characterise these differences using mucosal biopsy transcriptomics datasets from two recent UC clinical trials. Methods Two clinical trial datasets including patients with moderate to severe UC with mucosal biopsy RNA-Sequencing analysis were used: a phase 2/3 study of andecaliximab (anti-matrix metalloproteinase-9, NCT02520284) and a phase 3 study of ustekinumab (anti-interleukin-12/23, UNIFI, NCT02407236). Samples were scored for enrichment of ~5200 MSigDB signatures using Geneset Variation Analysis and were evaluated for correlation to the sample Robarts Histopathology Index (RHI) (Figure). Results From the andecaliximab baseline and follow-up samples, 11 Reactome pathways were specifically selected that were moderately correlated with RHI (r=~0.4) and had low correlation to each other (r<0.7). The 11 genesets, called Metabolism and Response to Stress (MARS) signatures, can generally be sorted into 2 categories: 5 metabolism-related and 6 related to stress response. Clustering of baseline andecaliximab samples scored with MARS signatures revealed 3 major sample groups (baseline and follow-up samples). Group 1 had low metabolism/high stress scores, group 2 had high metabolism/low stress scores, and group 3 had a mixture of samples that had high metabolism/low stress and low metabolism/high stress. Group 2 was associated with a lower proportion of current smokers (p=.04), and group 3 had a higher proportion of immunomodulator failure (p=.03), but not associated with disease duration or prior biologic use. Group 2 had lower Geboes score for epithelial neutrophils (p=.02), lamina propria neutrophils (p=.002), and inflammatory infiltrate (p=.03), while eosinophils increased (p=.01). To evaluate prediction of response to therapy, we evaluated the UNIFI dataset baseline samples using the MARS signatures and identified 4 groups. Group 2 had low metabolism/high stress response, group 3 had high metabolism/low stress response, and groups 1 and 4 had a mixture. The mucosal healing response rate was 3- to 4-fold lower for group 2 than other groups (5.3% [group 2] and 19%, 23%, and 21% for other groups, p=.0009). Conclusion We describe the MARS signatures which characterise the heterogeneity of participants with UC clinical trials and identify participants most likely to respond to ustekinumab at baseline. These signatures may be generally useful to predict patient response, match therapeutics to patient profiles, or identify pathways to target in difficult-to-treat patients.
Read full abstract