P0250 Circulating blood matrix metalloproteinase (MMP)-10 and 12 can accurately predict mucosal healing in Inflammatory Bowel Disease (IBD) in a multi-centred prospective longitudinal cohort study (2020-2024)
P Cartlidge, R Whelan, P Siu-Ying Lau, R Hall, C S Chuah, L Shafi, M Hodge, S Ong, A Le Saint-Grant, E Brownson, I Campbell, A Oddy, B Poulose, C Mowat, J MacDonald, J P Seenan, G T Ho, R KallaAbstract
Background
Complete mucosal healing (CMH) is a key prognostic indicator of long-term remission in IBD. There are however, no early biomarkers of mucosal healing to allow for deeper biological studies and potential for treatment stratification. Here, we prospectively investigate the clinical utility of a novel custom-designed 21-proteomic PEA-Olink* Flex panel in blood to identify proteins with the propensity for mucosal healing.
Methods
In the multi-centre, prospective, longitudinal IBD study (www.musicstudy.uk), we used a custom selected 21-protein marker panel (based on roles in resolution of inflammation; Olink Flex [Figure 1]) in blood plasma (1µl), collected at timepoints T1, T2 and T3 (months 0, +3 and +6 respectively). This uses a proximity extension assay, allowing for highly specific and sensitive absolute protein quantification. All subjects had prospective endoscopic assessment T1→T2→T3 (follow-up mean 5.2 months). CMH was defined by a Mayo score of 0 in UC and SES-CD of 0 in CD. Statistical analysis was performed using R, adjusting for age and gender. A machine learning (ML) model using SHAP (SHapley Additive exPlanations) analysis was then applied to the data.
Results
A total of 239 blood samples from 81 IBD patients were analysed (median 37 years, 59% male; 37% UC) – all with paired endoscopic mucosal healing data. 27/81 patients (33%) achieved CMH at follow-up. Of 21 blood proteins, multivariate analysis showed MMP10 and MMP12 concentrations to be significantly associated with CMH at timepoint 3 (FDR p=0.02 for both) with AUC 0.68 and 0.64 respectively; comparable to faecal calprotectin (AUC 0.69) in predicting mucosal healing. Both proteins were significantly lower in the UC CMH group vs no CMH (p=0.01 respectively), with only MMP10 (p=0.02) being associated with CMH in CD alone. In UC, endoscopic severity score UCEIS at follow-up endoscopy, correlated strongly with MMP10 and MMP12 levels combined (p<0.001, r=0.62). In UC, ΔMMP10 levels over time were highly significant in predicting improvement in endoscopic scores; but not CD. Random Forest ML model (incorporating 80 clinical variables including drug treatment) to the 21 proteins from T1→T2→T3 predicted CMH with an AUC of 0.69. SHAP analysis identified MMP12 in the top 3 variables impacting the model’s prediction.
Conclusion
Our study identifies tissue remodelling proteins MMP-10 and 12 as biomarkers for CMH on par with faecal calprotectin. The ΔMMP10 expression in UC shows dynamic change, predicting mucosal healing with more significance over time. MMP-10 and 12 are involved in the tissue degradation and remodelling process of inflamed mucosa which may explain the reduction in blood levels as mucosal inflammation resolves. Further large-scale validation work is planned.
References
*PEA-Olink: Proximity Extension Assay