PERIPHERAL BLOOD EXTRACELLULAR VESICLE LIPIDS AS BIOMARKERS FOR HUMAN INFLAMMATORY BOWEL DISEASE

Abstract BACKGROUND The current paradigm in inflammatory bowel disease (IBD) diagnostics is based on peripheral biomarkers such as C-reactive protein and fecal calprotectin that achieve low sensitivity and specificity for intestinal inflammation. Extracellular vesicles (EVs) are lipid-enveloped particles involved in inter tissue and cell-cell interactions. They also have unique properties reflecting the metabolic and phenotypic nature of the producer cells. Recent data revealed that surface proteins on intestinal epithelial cells-derived EVs can be detected in the peripheral blood. We posit that lipid profiling of circulating EVs (PBEs) can be used to discriminate active IBD from healthy subjects and further classify different stages of IBD (PCT Patent Pending: 13177N/2194P5). METHODS Patients diagnosed with Ulcerative Colitis (UC, n=50) or normal controls (n=50) were recruited at the UK IBD clinic or colonoscopy suite. During their colonoscopy or regular outpatient labs, 2 tubes of blood (20-30 mL) were drawn into K2-EDTA tubes. The blood samples were immediately stored on ice and then centrifuged at 1,500 g for 15 minutes at 4 °C within 30 minutes. The clarified plasma was immediately aliquoted and frozen in liquid N2. PBEs were isolated from plasma using size-exclusion microcolumns. Isolated exosomal preparations were lysed in cold acetonitrile, followed by extraction using a modified Folch method. The lipid fraction was carefully aspirated and dried in a Vacufuge before reconstitution in a chloroform/methanol mixture containing butylated hydroxytoluene. The extracted lipids from active UC and healthy control patient plasma were analyzed using direct infusion ultrahigh resolution Orbitrap mass spectrometry. We used statistical tools LASSO and Random Forest to select informative lipid features and build classification models. The performance of the classifiers was quantified by the receiver operating characteristic (ROC) curve and area under the ROC curve (AUC) based on 10-fold cross-validation. RESULTS Three PBE lipids identified by LASSO were also identified using the Random Forest method along with 7 additional lipids. Discriminating lipid classifiers between active UC and normal patients identified by both LASSO and Random Forest included phosphatidylcholines, plasmalogens, and sphingolipids. An AUC of 0.86 discriminated active UC from normal patients using the Random Forest method and 0.80 using the LASSO method. CONCLUSION These results are the fir