Torres J, Petralia F, Sato T, et al.
Inflammatory Bowel Disease is a chronic relapsing-remitting, immune-mediated condition with increasing prevalence globally . Despite novel agents targeting different disease pathways, the likelihood of achieving sustained clinical remission and mucosal healing remains low . One of the potential reasons may be that patients seek help and clinicians treat IBD once the disease is in its clinical phase. A sub-clinical phase of variable length may precede the symptoms that lead to a diagnosis and perhaps contribute to tissue damage which, once established, is difficult to reverse with currently available medical treatments.
In this study, Torres and colleagues set out to test the hypothesis that a pre-clinical phase of IBD may well be present and could be identified by proteomic markers .
The investigators performed a nested case-control study, selecting 599 individuals [Crohn’s Disease (CD), n=200; Ulcerative Colitis (UC), n=199; healthy controls (HC), n=200] who were participants in the US Department of Defence Serum Repository (DoDSR) and who provided at least four samples up to seven years prior to a diagnosis of IBD. They tested two platforms: (1) a panel of circulating antibodies – perinuclear anti-neutrophil cytoplasmic antibodies (pANCA), anti-Saccharomyces cerevisiae antibodies (ASCA) IgA and IgG (directed against a cell wall polysaccharide of the yeast), anti-Escherichia coli outer membrane porin C (OmpC), and anti-flagellin antibodies (anti-CBir1, anti-flagellin 2 and anti-flagellin X) (all provided by Prometheus Labs) and (2) a multiplex platform profiling 1129 protein biomarkers representing a broad range of biological functions (provided by SomaLogic).
By splitting the cohort into a training and a validation set, the authors were able to use a number of dimension reduction approaches and machine learning methodology to identify clusters of biomarkers that could differentiate between conditions, test their value as predictive markers and then validate their findings. Performing pathway enrichment analysis based on the SomaLogic proteomics platform, they were able to identify biological functions that were enriched in the pre-clinical phase of disease.
Overall, 51 SomaLogic biomarkers were predictive of CD development. The predictive performance of these combined markers ranged from an area under the curve (AUC) of 0.76 at 5 years to 0.88 at 1 year before a clinical diagnosis was made. Among the selected markers there were six [C-reactive protein (CRP), C5, trypsin 2 (PRSS2), serum amyloid-P (APCS), osteomodulin (OMD), and aggrecan core protein (ACAN)] which were associated with an AUC greater than 0.70 across all years, ranging from 0.75 at 5 years to 0.81 at 1 year. Among the antimicrobial antibody markers, anti-flagellin X and ASCA-IgA were predictive of CD across all timepoints. The predictive performance (AUC) of the model based on the antimicrobial markers varied from 0.69 at 5 years to 0.76 at 1 year. The integration of the antibody markers with the proteomic biomarkers did not result in better predictive performance.
The combination of antimicrobial markers for UC provided lower predictive performance, with the AUC ranging from 0.57 at 5 years to 0.61 at 1 year before a clinical diagnosis was made. In addition, the overall proteomics predictive performance was suboptimal, with the AUC ranging from 0.49 at 5 years to 0.68 at 1 year.
The authors also explored the risk of developing CD based on a positive test for six biomarkers (CRP, C5, PRSS2, APCS, OMD, ACAN) and ASCA IgA up to 5 years prior to diagnosis. For each year prior to a clinical diagnosis the mean probability of developing CD was between 0.29% and 0.35% for those testing positive (the baseline prevalence of CD in the US Army is 0.14%).
As expected, from the most predictive proteomic biomarkers for CD identified, the biological pathways highlighted as being most enriched in the pre-clinical phase were associated with innate immune responses, including the complement cascade and glycosaminoglycan metabolism.
The investigators concluded that CD may be predicted years before clinical symptoms develop and lead to a clinical diagnosis of IBD. In contrast, their approach was not successful in identifying predictors for UC. They stress the preliminary nature of their findings and that they should be validated in independent, prospective cohorts. They hope that in the future, timely identification of individuals at high risk for IBD development may permit early interventions that could delay, attenuate or even stop the development of clinical disease.
Polychronis Pavlidis – Short Biography
Polychronis Pavlidis is a Clinical Lecturer in Gastroenterology based at the IBD Centre, St Thomas’ Hospital, London, UK and the School of Immunology and Microbial Sciences at King’s College London. His translational research programme focusses on advancing personalised medicine approaches in IBD and identifying mechanisms driving non-response to available treatments.