P1042 Serum N-glycan Biomarkers Predict Patient Response to Biologics for Crohn’s Disease

Abstract

Abstract Background Several therapeutic antibodies are approved to treat Crohn’s Disease (CD), for example, Vedolizumab (anti-α4β7), Ustekinumab (targets IL-12 and IL-23), Adalimumab and Infliximab (anti-TNF). Glycosylation alterations have been associated with CD, but have not been thoroughly investigated as potential predictors of disease remission¹. A comparison of serum N-glycan profiles between responders and non-responders prior to each therapy was used to detect potential biomarkers to response. The N-glycan composition after treatment was used to distinguish the different mechanisms of the biologics. In addition, a comparison of glycomic profiles across the different therapies could highlight the distinct biological pathways involved in response. Methods Serum samples, before (t1) and after (t2) treatment, were taken from 166 patients, containing the four sub-cohorts of patients under different biologics with CD. Treatment response, observed in 100 patients, was assessed by a combination of clinical and endoscopic outcomes. Firstly, the Serum N-glycans were detected by hydrophilic interaction liquid chromatography to produce a 62-peak chromatogram. Followed by quantifying the relative area of 62 glycan peaks via HappyTools². Logistic regression was used to determine which glycan peaks at t1 are predictive of response (p-value <0.05) and which are markers of response (p-value<0.05). 5 fold cross-validation tested on a variety of classifiers was used to derive the optimal predictor of response. The glycomic profiles across the four biologics, at both t1 and t2, were compared through the Kruskal test. The significant peaks (p-value < 0.05) were then combined to generate a multiclass model to predict response in each therapeutic. Results The N-glycan classification models for all four sub-cohorts under different biologics, Ustekinumab (0.78 AUC), Adalimumab (0.82 AUC), Infliximab (0.75 AUC), Vedolizumab (0.75 AUC), performed moderately well at predicting therapy response prior to treatment(figure 1A). In turn, patients could be identified as responders after treatment by N-glycan based classification models, Ustekinumab (0.70 AUC), Adalimumab (0.86 AUC), Infliximab (0.85 AUC), Vedolizumab (0.84 AUC). Prior to treatment, patients who responded were strongly distinguished by the therapy type (0.96 AUC) through a combination of 13 significant glycan peaks (figure 1B). Conclusion This study has demonstrated multiple N-glycans are potential predictors of biological response, with the possibility to guide clinicians in the most appropriate form of treatment. These results require the integration of mass spectrometry and validation by larger cohorts.