Proteomic Profiling of Cold Storage Preservation Solution to Identify Signatures of Primary Graft Dysfunction Following Heart Transplantation

Abstract

Purpose There currently exist no validated biomarkers of donor primary graft dysfunction (PGD) risk. We hypothesized that proteomic profiling of cold-storage preservation solution may identify novel biomarkers of post-transplant allograft dysfunction. Methods 169 samples of HTK preservation solution were obtained between 2016 and 2018 at Duke University and analyzed using LC-MS. After QC, proteins quantified by two or more unique precursors were retained, and protein levels were log-transformed. Univariate logistic regression was performed to identify proteins associated with ISHLT-defined moderate-severe PGD (p<0.05). Significant proteins were included in an iterative Monte Carlo LASSO regression model; proteins retained in at least 50% of the 1000 total iterations were selected as a parsimonious model. We used AUC analysis to assess the predictive performance of the protein model compared to a clinical model including donor age, ischemic time and predicted heart mass. Results Out of 1340 proteins, 213 were nominally associated with PGD in univariate analysis. Sixteen of these proteins were retained in at least 50% of iterations of the LASSO model (Table 1). All of these proteins remained significant in multivariable analysis after adjustment for the aforementioned clinical covariates. AUC of the 16-protein model was 0.90 as compared to the clinical model at 0.68 (p<0.001). Two proteins (FMOD and CSN5) were modestly correlated with ischemic time (R=0.28 and R=-0.15, p<0.05) and four proteins (CMA1, CNDP1, CSN5, and HV323) were modestly correlated with donor predicted heart mass (R=-0.24 - 0.2, p<0.05). Conclusion Using non-targeted proteomic profiling of traditional cold storage preservation solution, we have identified a set of novel biomarkers that are associated with PGD. Key biologic candidates within this set include MYL4, FMOD, ELOC, and CMA1. Future studies are needed to validate these signals and elucidate their role in clinical care. There currently exist no validated biomarkers of donor primary graft dysfunction (PGD) risk. We hypothesized that proteomic profiling of cold-storage preservation solution may identify novel biomarkers of post-transplant allograft dysfunction. 169 samples of HTK preservation solution were obtained between 2016 and 2018 at Duke University and analyzed using LC-MS. After QC, proteins quantified by two or more unique precursors were retained, and protein levels were log-transformed. Univariate logistic regression was performed to identify proteins associated with ISHLT-defined moderate-severe PGD (p<0.05). Significant proteins were included in an iterative Monte Carlo LASSO regression model; proteins retained in at least 50% of the 1000 total iterations were selected as a parsimonious model. We used AUC analysis to assess the predictive performance of the protein model compared to a clinical model including donor age, ischemic time and predicted heart mass. Out of 1340 proteins, 213 were nominally associated with PGD in univariate analysis. Sixteen of these proteins were retained in at least 50% of iterations of the LASSO model (Table 1). All of these proteins remained significant in multivariable analysis after adjustment for the aforementioned clinical covariates. AUC of the 16-protein model was 0.90 as compared to the clinical model at 0.68 (p<0.001). Two proteins (FMOD and CSN5) were modestly correlated with ischemic time (R=0.28 and R=-0.15, p<0.05) and four proteins (CMA1, CNDP1, CSN5, and HV323) were modestly correlated with donor predicted heart mass (R=-0.24 - 0.2, p<0.05). Using non-targeted proteomic profiling of traditional cold storage preservation solution, we have identified a set of novel biomarkers that are associated with PGD. Key biologic candidates within this set include MYL4, FMOD, ELOC, and CMA1. Future studies are needed to validate these signals and elucidate their role in clinical care.