Meta-analysis of gene expression disease signatures in colonic biopsy tissue from patients with ulcerative colitis

Bryan Linggi,Azucena Salas,Mark S Silverberg,Guangyong Zou,Vipul Jairath,Lisa M Shackelton,Dermot P B Mcgovern,Niels Vande Casteele,Bram Verstockt,Shadi Nayeri,Brian G Feagan

doi:10.1038/s41598-021-97366-5

Abstract

Publicly available ulcerative colitis (UC) gene expression datasets from observational studies and clinical trials include inherently heterogeneous disease characteristics and methodology. We used meta-analysis to identify a robust UC gene signature from inflamed biopsies. Eight gene expression datasets derived from biopsy tissue samples from noninflammatory bowel disease (IBD) controls and areas of active inflammation from patients with UC were publicly available. Expression- and meta-data were downloaded with GEOquery. Differentially expressed genes (DEG) in individual datasets were defined as those with fold change > 1.5 and a Benjamini–Hochberg adjusted P value < .05. Meta-analysis of all DEG used a random effects model. Reactome pathway enrichment analysis was conducted. Meta-analysis identified 946 up- and 543 down-regulated genes in patients with UC compared to non-IBD controls (1.2 and 1.7 times fewer up- and down-regulated genes than the median of the individual datasets). Top-ranked up- and down-regulated DEG were LCN2 and AQP8. Multiple immune-related pathways (e.g., ‘Chemokine receptors bind chemokine’ and ‘Interleukin-10 signaling’) were significantly up-regulated in UC, while ‘Biological oxidations’ and ‘Fatty acid metabolism’ were downregulated. A web-based data-mining tool with the meta-analysis results was made available (https://premedibd.com/genes.html). A UC inflamed biopsy disease gene signature was derived. This signature may be an unbiased reference for comparison and improve the efficiency of UC biomarker studies by increasing confidence for identification of disease-related genes and pathways.

Highlights

Available ulcerative colitis (UC) gene expression datasets from observational studies and clinical trials include inherently heterogeneous disease characteristics and methodology
Datasets were excluded for the following reasons: they did not contain both UC patients and non-inflammatory bowel disease (IBD) controls, they included only pediatric patients, there were fewer than 10 combined UC patients and non-IBD controls, samples were only taken from uninflamed mucosa or were included in other datasets, data was not expressed in intensity values or was z-score transformed, or samples were not processed at the same time
A total of 8 microarray datasets deposited in the NCBI Gene Expression Omnibus (GEO) database between 2009 and 2018 that were derived from intestinal tissue RNA from various cohorts and that included at least 10 patients with UC and non-IBD controls combined, and which originated from a range of institutions and microarray platforms (Table 1) were identified, for a total of 251 samples from patients with UC and 94 samples from non-IBD controls

Summary

Introduction

Available ulcerative colitis (UC) gene expression datasets from observational studies and clinical trials include inherently heterogeneous disease characteristics and methodology. Eight gene expression datasets derived from biopsy tissue samples from noninflammatory bowel disease (IBD) controls and areas of active inflammation from patients with UC were publicly available. Global transcriptome (RNA) analysis has been a powerful and unbiased tool to aid in understanding of disease etiology, pathology, diagnosis, and subtypes, as well as for the identification of predictive markers of drug efficacy, molecular surrogates of disease activity, and/or pharmacodynamic m arkers[3]. Despite this abundant potential, the application of transcriptomics to clinical trial data has been frequently limited by small sample size, especially when subgroups of interest are considered (e.g., treatment responders versus non-responders). Several aspects of tissue acquisition and processing may influence the results of transcriptomics analysis, including biopsy location within the colon and relationship to endoscopically visible active disease[7]

Methods

Results

Conclusion