Senescence Biomarkers Independently Predict Disease-Related Mortality in Chronic Kidney Disease

Chronic kidney disease (CKD) staging systems predict kidney failure progression but fail to provide clinicians with patient level estimates of mortality risk, unlike similar tools that exist in cardiovascular disease, COPD and liver disease. Cellular senescence and its associated SASP have been implicated in CKD progression and comorbidity burden, yet whether these circulating SASP components independently predict long-term mortality has yet to be proven. Using plasma proteomics from a CKD diagnosed UKBiobank cohort, we identify EDA2R, NT-pro-BNP, CTSZ and REN as senescence associated biomarkers that strongly associate with 9-year disease-related mortality and estimated glomerular filtration rate (eGFR). A support vector machine (SVM) trained on nine proteomic features achieved an AUC of 0.832 during training and 0.711 in a held-out test set, outperforming both a clinical covariate model and the Charlson Comorbidity Index (CCI), the gold standard risk-adjustment tool for predicting mortality. Cox proportional hazard models confirmed that EDA2R, NT-pro-BNP, CTSZ and REN independently predict long-term disease related mortality, post-adjustment for age, sex, cardiovascular disease, diabetes and eGFR. Proteomic aging clocks (PACs), which utilized EDA2R, NT-pro-BNP and REN among other proteins, proved that individuals experiencing 9-year disease-related mortality were often more biologically aged, with proteomic accelerated aging being statistically significant between survivors and non-survivors, regardless of disease severity. Lastly, we found that EDA2R, NT-pro-BNP, CTSZ and REN expression was inversely correlated with renal function, while also confirming their direction of expression was consistent across our discovery cohort in the UKBiobank and in our own independently recruited CKD cohort. These findings promote senescence-associated plasma proteins as non-invasive, independent predictors of mortality in CKD, and nominate EDA2R and CTSZ as previously uncharacterised candidates for senomorphic interventions.