DOP31 Gene correlation network analyses reveal transcript modules and pathogenic mechanisms associated with resistance and response to ustekinumab therapy in Ulcerative Colitis

Abstract

Abstract Background Precision medicine in ulcerative colitis (UC) remain elusive. The heterogeneity in patients’ mucosal transcriptomic profiles suggests that stratification based on molecular mechanisms might be possible, which may link to treatment response. We interrogated a dataset of ustekinumab-treated patients treated to determine transcript modules (TMs) in active UC, investigate pathogenic pathways and consider which cells might be driving ustekinumab resistance. Methods Pre-treatment mucosal transcriptomes and outcome data were available for 358 ustekinumab-treated patients enrolled in the UNIFI programme. Modules of co-expressed genes were defined with weighted gene correlation network analysis (WGCNA) and their functional roles and regulatory interactions inferred with Fisher tests. Multivariate regression was used to test the ability of clinical and transcriptomic data to predict treatment resistance. Further insights into the cellular and molecular features of ustekinumab resistance were gauged from the analysis of single cell data (Single Cell Portal accession SCP259). Analyses were performed in R 4.2.3 (Vienna, Austria). Results WGCNA generated 23 transcript modules (TMs) whose correlation with mucosal healing ranged from -0.26 to 0.25. The three TMs most negatively correlated with response comprised genes enriched in neutrophil degranulation, extracellular matrix activity and endoplasmic reticulum function. The modules correlating highest with response were enriched in genes involved in mitochondrial function. The constituents of the ustekinumab resistance-related modules are regulated by a hierarchy of transcription factors downstream of NFKB1, AHR, JUN and RELA. They are also densely connected by regulatory protein-protein interactions (Figure 1). A multivariate model including TM1 enrichment score and clinical variables predicts resistance with area under the curve 0.76 (95% confidence interval: 0.69–0.82). Notably, TM1 included TREM1 and OSM, which were previously associated with anti-TNF resistance. Surmising that cell populations co-expressing these genes might be key to ustekinumab resistance too, we identified TREM1+ OSM+ inflammatory monocytes (IMs). These showed higher expression of genes such as IL1B, IL8 and CXCL2 compared to other IMs, and increased activity in pathogenic processes including interleukin (IL)10, IL4 and IL13 signalling and G-protein-coupled receptor signalling. Conclusion Gene co-expression analysis identified transcripts and molecular mechanisms associated with response and predictive of resistance to ustekinumab. TREM1+ OSM+ IMs may be key to driving resistance, suggesting an important role for innate immunity. These findings are important for developing precision medicine approaches in UC.