Search

P180 Association between serum and tissue protein profiles in patients with active inflammatory bowel disease: the ASCERTAIN study

T. Stevens MD1, M. Wildenberg1, M. Koželj2, G. Novak2, K. Gecse1, M. Duijvestein1, M. Löwenberg1, D. Drobne2, G. D’Haens1

1Department of Gastroenterology and Hepatology, Amsterdam UMC – Location AMC, Amsterdam, The Netherlands, 2Department of Gastroenterology and Hepatology, University Medical Center Ljubljana, Ljubljana, The Netherlands

Background

To facilitate personalised treatment of inflammatory bowel disease (IBD), biomarkers for disease entity and disease severity in easily accessible samples are required. Therefore, we aimed to (i) identify proteins that discriminate Crohn’s disease (CD) from ulcerative colitis (UC), (ii) assess the association of proteomic profiles with disease severity and (iii) assess the potential to use serum samples as a proxy for intestinal biopsies.

Methods

In this prospective cross-sectional study, IBD patients with active endoscopic disease (presence of ≥1 ulcer (CD) or Mayo score ≥1 (UC)) and age matched non IBD controls (normal endoscopy) were included. In IBD, biopsies were taken from inflamed tissue; in CD from the edge of an ulcer and in UC and controls between 20 and 25 cm from the anal verge. Disease severity was dichotomised in mild vs. severe defined as SES CD <10 vs. ≥11 in CD and Mayo endoscopy score 1/2 vs. 3 in UC. Ninety-two inflammatory proteins were measured in biopsies and serum using a proximity extension assay (OLINK). The elastic net algorithm was used to select proteins associated with disease severity. Multiplicity adjustments were done using the Benjamini–Hochberg approach.

Results

Forty-one CD, 39 UC and 10 controls were included. Median SES-CD score [IQR] was 9 [7–16] in CD. In UC, 20 (51.3%), 15 (38.5%) and 4 (10.3%) patients had an endoscopic Mayo score of 3, 2 and 1, respectively. Three proteins differed significantly between UC and CD in tissue (IL-17A, IL13 and CCL4). However, this was not reflected in serum levels of these molecules. For prediction of disease severity in CD, elastic net regression identified a set of 11 serum proteins that discriminated mild from severe disease (accuracy 0.667). Top five proteins were IFNγ (OR 3.03), CCL23 (OR 1.28), Cystatin D (OR 0.61), TNF (OR 1.28) and HGF (OR 1.17) (Figure 1). In tissue, no discriminatory profile was identified. Conversely, in serum of UC, no proteins were selected, while a set of three tissue proteins was associated with disease severity (accuracy 0.725). These proteins were osteoprotegerin (OPG) (OR 1.66), IL8 (OR 1.09) and CCL25 (OR 0.97). The discrepancy between serum and tissue was further supported by correlation analyses. Of all 92 proteins, only three showed significant correlation between tissue and serum: IL17A (r = 0.44), TGF-α (r = 0.42) and IL6 (r = 0.39).

Conclusion

Proteomic profiles between tissue and serum in active IBD correlated poorly. Only in CD, serum markers differentiated mild from severe disease. In UC, no proteins in serum were selected whereas in tissue, expression of OPG, IL-8 and CCL25 was associated with disease severity. These data again emphasise the difference between CD and UC on a molecular level.