Background: Proximity extension assays (PEA) can compare concentrations of multiple proteins across biological samples and utilises the specificity of antibody proximity and the sensitivity of polymerase chain reaction to detect proteins of interests. As part of IBD Character, we performed high-throughput prospective case-control serum profiling to identify proteins that can predict Inflammatory Bowel Disease (IBD) and its disease course.
Methods: Serum profiling was performed in treatment naïve newly diagnosed IBD and Non-IBD (symptomatic and healthy controls) using PEA panels (Olink Proteomics). Phenotypic data were obtained for all patients and follow up outcome data were captured for the Edinburgh and Oslo IBD cohorts. Treatment escalation was defined as the need for surgery and/or biologic therapies after initial induction of remission. Linear models were created for each protein including age and sex as covariates. Statistical analysis was performed using R.
Results: Protein profiles were available in 635 patients (152 CD, 159 UC, 26 IBD-U, 298 non-IBD). 61 protein markers were significantly associated with IBD including MMP12 (Holm-adjusted p=4.1×10–26). Mapping the top markers to the cell-specific FANTOM 5 , several differentially expressed proteins originate from innate and adaptive immune cells such as dendritic cells.
As diagnostic markers, 5 proteins differentiate UC from CD including MMP-12 (p= p=4.6×10–4)
Follow up data were available for 206 patients. A total of 25 (32%), 21 (18%) and 2 (18%) patients required treatment escalation in the CD, UC and IBDU respectively. The data were randomly split into a testing (n=124) and a validation cohort (n=82). Using multivariable analyses with age, sex and follow up time as covariates, 9 proteins survived Holm adjustment and 8 of these proteins significantly predicted escalation in the validation cohort.
1000 iterations of unsupervised linear discriminant consensus clustering were performed using 7 randomly selected top protein probes. This identified 2 patients groups that had significantly different disease courses:logrank p=2.2×10–10, HR 5.6 (2.0–15.6), outperforming conventional biomarkers in predicting treatment escalation (hsCRP >4 mg/L, HR 3.2 (1.7–5.8), logrank p=0.0003 and Alb <36 g/L, HR 2.7 (1.4–5.2), p=0.0004).
Conclusions: We have identified immune cell-specific PEA-based serum proteins that can diagnose IBD and predict disease course. These data demonstrate the translational potential of a PEA based technology in IBD
 FANTOM Consortium and the RIKEN PMI and CLST (DGT), et al., (2014), A promoter-level mammalian expression atlas, Nature Publishing Group, Nature, 507:462–70