Predictive Biomarkers for Bronchiolitis Obliterans

Abstract

Purpose Chronic Allograft Dysfunction (CLAD) with Bronchiolitis obliterans (BOS) phenotype is a major limitation for long term survival after lung transplantation (LT). Predictive biomarkers for BOS are unavailable. Purpose of our study was to establish a tractable system to evaluate the effects of pre-transplant antibodies to self-antigens (AutoAbs) and to examine specific patterns that correlate with BOS. Methods Serum samples collected pre-transplant and stored in HLA lab were retreived after IRB approval. Pre-existing AutoAbs in sera were measured using a multiplexed protein array bearing 375 auto antigens developed by Microarray core lab. Microarray data was analyzed using GLMNET in R package. Machine learning program was used to select panel of AutoAbs with best predictive value for BOS. Survival analysis was done using Kaplan-Meier estimation or cox proportional hazards model . Results 41 recipients who met inclusion criteria were grouped into low BOS (BOS grade 0 and 1, n=20) and high BOS (BOS grade 2, 3; n=21). Survival analysis showed worse survival in patients with higher BOS grade (figure 1A). 75 AutoAbs were significantly higher in high BOS compared to low BOS in pre transplant samples (figure 1B). Top 5 significantly elevated AutoAbs in high BOS were: anti-peptidyl arginine deiminases , anti-endothelial cell extract, anti-MPO, anti beta 2 glycoprotein 1 and anti-ribosomal phosphoprotein1. Using the machine learning program, 15 AutoAbs (figure 1C)were selected for best prediction of BOS with area under curve 0.91, CI (95%) 0.83-0.99. Sensitivity of prediction with the panel is 90% and specificity 85% (figure 1D). Conclusion Our study points to a panel of 15 AutoAb that can predict BOS before LT. Novel findings reported in our study are being confirmed in a replication cohort. Prediction models with a panel of AutoAb can significantly transform management of LT recipients for improved survival.