DOP54 Integrated network analysis using patient-specific single-nucleotide polymorphism profiles uncovers new pathways involved in ulcerative colitis pathogenesis

D. Modos1,2,3, J. Brooks2,3,4, P. Sudhakar2,3,5, B. Verstockt5,6, B. Alexander-Dann1, A. Zoufir1, D. Fazekas2,7, S. Vermeire5,6, T. Korcsmaros2,3, A. Bender1

1Department of Chemistry, University of Cambridge, Cambridge, UK, 2Earlham Institute, Norwich, UK, 3Gut Microbes and Health Programme, Quadram Institute, Norwich, UK, 4Department of Gastroenterology, Norfolk and Norwich University Hospitals, Norwich, UK, 5Department of Chronic Diseases, Metabolism and Ageing, KU Leuven, Leuven, Belgium, 6Department of Gastroenterology and Hepatology, University Hospitals Leuven, Leuven, Belgium, 7Department of Genetics, Eötvös Loránd University, Budapest, Hungary

Background

Genome-wide association studies have deciphered the single nucleotide polymorphisms (SNPs) which are responsible for ulcerative colitis (UC) susceptibility. However, to understand how these SNPs are involved in UC, additional methods are necessary. One such approach is in silico network propagation modelling, which can discover how the effects of SNPs in UC can affect the whole cell. A complementary approach is weighted gene co-expression network analysis (WGCNA), where co-regulated genes are identified using transcriptomic data. Integrating these two methods can shed light on how SNPs are affecting the transcriptome of UC patients.

Methods

We used immunochip profiles of 941 UC patients and focussed on UC-associated SNPs altering regulatory regions. Based on these regions, we identified affected genes. To understand how their corresponding proteins rewire transcriptional regulation, we predicted the path between these proteins and relevant transcription factors (TF) using the OmniPath signalling network (http://omnipathdb.org). From the TFs, we propagated the signal further to target genes using TFlink (https://tflink.net) and GTRD (http://gtrd.biouml.org). To evaluate the predicted network propagation signal, we conducted WGCNA with transcriptomics data from 46 matching patients’ (GEO ID: GSE48959). To interpret the results, we used Gene Ontology Biological Process annotations of the target genes, and we compared the function and regulation of affected genes and the determined WGCNA modules.

Results

We found 9 predominant signalling pathways, some already known from other studies to be involved in UC pathogenesis, including NFkB signalling, chemokine signalling, Notch pathway, JAK/STAT signalling. Downstream of these pathways we identified potential key TFs regulate the UC phenotype, for example NFKB1, GATA3, GTF2I. The targets of these TFs were enriched in the WGCNA modules of the patients. The WGCNA modules and the transcriptionally affected genes had enriched processes including cell migration, TGF-β signalling, exocytosis, adaptive T- and B-cell-specific immune responses and tight junctions. We also found myogenetic development specific TFs affected transcriptionally such as MyoD, MEF2A, MEF2D. We are currently validating these results through patient-specific biopsies.

Conclusion

In silico methods bring us closer to understanding UC pathogenesis. Our results suggest that in a well-defined set of patients, weakened tight junctions and insufficient immune response can lead to dysfunctional epithelial barrier, resulting in poor wound healing in UC. We hope the developed workflow will provide novel diagnostic and therapeutic options in UC.