Serum Proteomics-Based Outcome Prediction in ADPKD

Journal of the American Society of Nephrology, vol. 36

Abstract

Background: Autosomal dominant polycystic kidney disease (ADPKD) is the most common genetic cause of kidney failure. Predicting key outcomes such as the time to kidney failure is crucial to patient counselling and treatment decisions. However, currently available prediction algorithms require MRI images or genetic exams and have limited predictive accuracy. Thus, new approaches based on simple blood values are an important goal. We built proteome-based models to predict yearly eGFR decline. Methods: Proteomic analysis was performed on blood samples collected from 3 ADPKD cohorts: Screening Cohort (SC, 214 patients, serum), Internal/Temporal Cohort (ITC, 408 patients, serum), and External Cohort (EC, 173 patients, plasma). eGFR-slope-associated proteins were identified using LIMMA and LASSO. LASSO proteins were used to build proteome-based linear models. ITC and EC were used for model validation. Results: We identified 29 proteins that are significantly associated with eGFR slope (LIMMA), and 6 proteins that are optimal to build prediction models (LASSO). The Proteome Model (PM, adj. R2 0.31) containing only these 6 proteins was superior to a clinical model (age, sex, eGFR and Mayo Imaging Class (MIC), adj. R2 0.23). Adding the protein data to the clinical model showed independent added value (Combined Model (CM, adj. R2 0.34)). The proteome-based model was confirmed in both validation cohorts, but the use of plasma instead of serum decreased the predictive capacity. Lastly, we calculated future eGFR to compare prediction performance of our models to the previously established MIC Model. Both PM and CM performed better than the MIC Model, which underestimated the future eGFR in SC and ITC. In EC, our models overestimated while the MIC model underestimated future eGFR. Conclusion: Our study shows the great potential for the identification of biomarkers in ADPKD. Importantly, the independent value of a panel of 6 proteins could be confirmed in two validation cohorts. Prospective validation using absolute quantification will now be crucial towards clinical implementation. Beside biomarker identification, the proteome data provides important added information towards molecular processes contributing to disease progression.

Authors 7

1. Roman-Ulrich Mueller

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Universitatsklinikum Koln, Cologne, NRW, Germany;

   University of Cologne Center for Molecular Medicine Cologne, Cologne, NRW, Germany;

2. Hande Özge Aydogan Balaban

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Universitatsklinikum Koln, Cologne, NRW, Germany;

   University of Cologne Center for Molecular Medicine Cologne, Cologne, NRW, Germany;

3. Sita Arjune

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Universitatsklinikum Koln, Cologne, NRW, Germany;

   University of Cologne Center for Molecular Medicine Cologne, Cologne, NRW, Germany;

4. Franziska Grundmann

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Universitatsklinikum Koln, Cologne, NRW, Germany;

   University of Cologne Center for Molecular Medicine Cologne, Cologne, NRW, Germany;

5. Jan‐Wilm Lackmann

   University of Cologne · Cologne Excellence Cluster on Cellular Stress Responses in Aging Associated Diseases

   Affiliation as printed

   Exzellenzcluster CECAD in der Universitat zu Koln, Cologne, NRW, Germany;

6. Thomas Rauen Aachen

   Universitätsklinikum Aachen

   Affiliation as printed

   Universitatsklinikum Aachen, Aachen, NRW, Germany

7. Philipp Antczak

   University of Cologne · University Hospital Cologne

   Affiliation as printed

   Universitatsklinikum Koln, Cologne, NRW, Germany;

   University of Cologne Center for Molecular Medicine Cologne, Cologne, NRW, Germany;