Characterisation of the Circulating Transcriptomic Landscape in Inflammatory Bowel Disease Provides Evidence for Dysregulation of Multiple Transcription Factors Including NFE2, SPI1, CEBPB, and IRF2

Abstract

Aim: To assess the pathobiological and translational importance of whole-blood transcriptomic analysis in inflammatory bowel disease [IBD].

Methods: We analysed whole-blood expression profiles from paired-end sequencing in a discovery cohort of 590 Europeans recruited across six countries in the IBD Character initiative (newly diagnosed patients with Crohn's disease [CD; n = 156], ulcerative colitis [UC; n = 167], and controls [n = 267]), exploring differential expression [DESeq2], co-expression networks [WGCNA], and transcription factor involvement [EPEE, ChEA, DoRothEA]. Findings were validated by analysis of an independent replication cohort [99 CD, 100 UC, 95 controls]. In the discovery cohort, we also defined baseline expression correlates of future treatment escalation using cross-validated elastic-net and random forest modelling, along with a pragmatic ratio detection procedure.

Results: Disease-specific transcriptomes were defined in IBD [8697 transcripts], CD [7152], and UC [8521], with the most highly significant changes in single genes, including CD177 (log2-fold change [LFC] = 4.63, p = 4.05 × 10-118), MCEMP1 [LFC = 2.45, p = 7.37 × 10-109], and S100A12 [LFC = 2.31, p = 2.15 × 10-93]. Significantly over-represented pathways included IL-1 [p = 1.58 × 10-11], IL-4, and IL-13 [p = 8.96 × 10-9]. Highly concordant results were obtained using multiple regulatory activity inference tools applied to the discovery and replication cohorts. These analyses demonstrated central roles in IBD for the transcription factors NFE2, SPI1 [PU.1], CEBPB, and IRF2, all regulators of cytokine signalling, based on a consistent signal across cohorts and transcription factor ranking methods. A number of simple transcriptome-based models were associated with the need for treatment escalation, including the binary CLEC5A/CDH2 expression ratio in UC (hazard ratio = 23.4, 95% confidence interval [CI] 5.3-102.0).

Conclusions: Transcriptomic analysis has allowed for a detailed characterisation of IBD pathobiology, with important potential translational implications.

Keywords: Crohn’s disease; Inflammatory bowel disease; transcription factor; transcriptome; ulcerative colitis.

Graphical Abstract

Figure 1.

Overview of the study.

Figure 2.

Volcano plots showing differential expression of whole-blood transcripts between inflammatory bowel disease [IBD] and subtypes of IBD vs. controls [A–C] and CD vs. UC [D] in the discovery cohort. Only transcripts with a false-discovery rate <0.05 are shown. Genes with log2-fold change [LFC] >2 are indicated in red. Genes with the top 10 most significant p-values and top 10 LFC are labelled [the two lists may overlap]. The genes labelled in panel D may be considered overexpressed in CD relative to UC. CD, Crohn’s disease; UC, ulcerative colitis.

Figure 3.

The inflammatory bowel disease [IBD] differential expression consensus shortlist obtained by intersecting data from the discovery [IBD Character] and replication cohorts [Ostrowski et al.]. Flowchart illustrating the intersection of differential expressed transcripts in the discovery and the replication data. Mean log2-fold changes [LFC] and Bonferroni-corrected p-values [p_Bonf.] for genes overexpressed in the discovery and replication cohorts. Only genes with absolute LFC >1 and p_Bonf. <0.05 in both datasets were selected. This selection was performed separately in the IBD, Crohn’s disease [CD], and ulcerative colitis [UC] data.

Figure 4.

Most significant gene ontology sets [A] and reactome pathways [B] enriched by genes overexpressed in transcriptomes from the discovery [IBD Character] and replication [Ostrowski et al.] cohorts. Ratio represents the fraction of genes from the target dataset which were identified among the overexpressed genes. FDR, false-discovery ratio; IBD, inflammatory bowel disease.

Figure 5.

The intersection of EPEE and ChEA3 results in the discovery and replication data. Rectangles filled in blue indicate the presence of a transcription factor among the most significant results in the given analysis. Contrasts between inflammatory bowel disease [IBD], Crohn’s disease [CD], ulcerative colitis [UC], and controls were explored. DoRothEA2 was used to confirm the findings [thus filtering out IRF1, which is not shown]. SPI1 encodes PU.1.

Figure 6.

Heatmap illustrating Pearson’s correlations between transcription factor expression and LM22-predicted cell type abundance in the whole blood of patients with inflammatory bowel disease [IBD] and in controls, in the discovery cohort. A stronger absolute correlation suggests more consistent expression in the cell type. Apart from B cells, γδ T cells, and M0 macrophages, the absolute differences between the two groups were small [<0.3].

Figure 7.

Ulcerative colitis treatment escalation depending on the CLEC5A/CDH2 ratio [low-risk <3]. The log-rank test p-value is shown. Hazard ratio: 23.4 (95% confidence interval [CI] 5.3–102.0). Patients with escalation after 1 year and censored within the first year [excluded from modelling] are also included for illustration.