A discriminative metabolic profile in the sera of Crohn's disease patients with fibrostenosis

Introduction: Crohn’s disease (CD) patients are at high risk of developing fibrotic strictures, which significantly affects the patient’s quality of life. The accumulation of fibrotic tissue and progression of stricture formation is difficult to assess, leading to late awareness of stricture formation and surgical resection. Therefore, fibrostenosis-specific biomarker profiles are highly needed. Aim: Given the increasing evidence of metabolic alterations in activated fibroblasts, we aimed to identify discriminating metabolic markers in the serum of CD patients with and without fibrostenosis. Methods: In this retrospective study, samples of 66 CD patients with (n=28) and without (n=38) ileal fibrotic strictures at the time of sampling were selected from a local biobank (UZ Gent, BB190100). Fibrostenosis was defined as a narrowing of lumen and prestenotic dilation on CT/MRI at the time of serum collection. Both groups included an equal number of patients in remission or with active disease, based on imaging and/or CRP levels (cut-off at 10mg/L) and were age- and gender-matched. Metabolomics analysis was performed applying UHPLC-Q-Orbitrap-HRMS. The in-house method for metabolite extraction and mass spectrometry analysis was validated for serum compatibility. Statistical analysis of the untargeted MS data was performed using SIMCA 15.0 and MetaboAnalyst 4.0, allowing multivariate statistical modelling through, amongst others, Principal Component Analysis (PCA) and sparse Partial Least-Squares Discriminant Analysis (sPLS-DA). Results: Age at diagnosis, exposure to anti-TNF drugs, disease location and disease activity were similar between both groups. Validation of metabolomics analysis of serum samples, including instrumental precision, intra-assay, and inter-day analyses, showed excellent coverage of the measured metabolites with respective coefficients of variance 75% of metabolites complied) and <30% for untargeted analysis (compound compliance rate of 80% in positive and 90% in negative ionisation mode). The extraction and analysis of the serum samples yielded a total of 5,959 features. sPLS-DA models were used to determine which features were most discriminating between groups, and 1,000 features were retained. A valid Orthogonal PLS-DA model based on these 1,000 features had R2Y of 0.99 and Q2 of 0.83, suggesting excellent predictivity and fitting of the existent data. The top differentiating features, 47 in total, were retained after further filtering, based on a Variable Influence on Projection score higher than 1, Jack-knifed confidence interval higher than 0 and S-plot with p(corr)-values lower than -0.3 or higher than 0.3. Conclusions: To our knowledge, this is the first study using a comprehensive metabolomics approach by which we unveiled a discriminative metabolic fingerprint in the serum of fibrostenotic CD patients. Further MS2 measurements to allow the identification of these metabolites and validation are on-going and will allow to prove biomarker potential and gain insight in the functional pathways involved.