Kidney Week

Abstract: TH-PO794

Integrative Computational Characterization of Membranous Nephropathy from Genome-Wide Transcriptome Profiles

Session Information

• Cellular Crosstalk in Glomerular Diseases - I
  October 25, 2018 | Location: Exhibit Hall, San Diego Convention Center
  Abstract Time: 10:00 AM - 12:00 PM

Category: Glomerular Diseases

• 1201 Glomerular Diseases: Fibrosis and Extracellular Matrix

Authors

• Sealfon, Rachel S., Flatiron Institute, New York, New York, United States

Rachel S. Sealfon,

• Mariani, Laura H., University of Michigan, Ann Arbor, Michigan, United States

Laura H. Mariani,

• Nair, Viji, University of Michigan, Ann Arbor, Michigan, United States

Viji Nair,

• Funk, Julien, Flatiron Institute, New York, New York, United States

Julien Funk,

• Wong, Aaron, Flatiron Institute, New York, New York, United States

Aaron Wong,

• Avila-Casado, Carmen, University Health Network, University of Toronto, Toronto, Ontario, Canada

Carmen Avila-Casado,

• Troyanskaya, Olga, Flatiron Institute, New York, New York, United States

Olga Troyanskaya,

• Kretzler, Matthias, University of Michigan, Ann Arbor, Michigan, United States

Matthias Kretzler,

• Beck, Laurence H., Boston University Medical Center, Boston, Massachusetts, United States

Laurence H. Beck,

Background

Genome-wide expression data from kidney tissue of patients with nephrotic syndrome provide an opportunity to elucidate mechanisms of injury specific to membranous nephropathy (MN). NEPTUNE and ERCB are prospective, multicenter cohort studies of patients with kidney disease profiled for whole-genome transcriptomes from microdissected biopsy tissue. These datasets provide rich, complementary resources to study the molecular characteristics of MN.

Methods

We examine whole-genome expression profiles of patients with MN and with other causes of nephrotic syndrome (e.g. FSGS, MCD, IgAN) from the NEPTUNE and ERCB cohorts. We construct machine learning classifiers to distinguish patients with each disease based on gene expression and evaluate performance using five-fold crossvalidation. Using Significance Analysis of Microarrays, we identify genes differentially expressed in MN and common to both cohorts. To identify functional modules within the MN-specific gene set, we perform community clustering of these genes in a kidney functional network which integrates thousands of transcriptome-wide assays using tissue-specific regularized Bayesian integration (Greene et al. 2015; Krishnan et al. 2016).

Results

We find that glomerular gene expression can separate patients with MN from other nephrotic patients with high accuracy (AUC=0.9). The genes most predictive of MN diagnosis in NEPTUNE are also differentially expressed in MN patients in the independent ERCB cohort. We identify 220 genes differentially expressed in MN patients common to both cohorts. These genes are enriched in multiple Gene Ontology terms, including anatomical structure morphogenesis, biological adhesion, and basement membrane, and also significantly enriched in homologs of mouse podocyte markers (Park et al. 2018). Clustering the MN-specific genes in a kidney functional network uncovers multiple modules, including gene clusters enriched in ion transport and the extracellular matrix.

Conclusion

This work identifies an expression profile specific to MN patients, enriched in podocyte-specific genes, and shared across patients in two independent cohorts. Such work may facilitate transcriptome-based disease classification, insight into disease pathophysiology, and targeted therapeutics.

Funding

• NIDDK Support