FC043: Calibration of GTEX Samples using Curated Primary Human Kidney Tissues Improves Identification of Medulla-Specific Gene Expression Patterns

Abstract BACKGROUND AND AIMS The parenchyma of the kidney is divided into two major structures, the outer cortex and the inner medulla. The medulla is a key regulator of salt and water balance and is commonly affected by congenital anomalies of the kidney and urinary tract. Its cell types represent the origin of several monogenic kidney diseases such as ADTKD-UMOD and Bartter-SLC12A1. Currently, there is a lack of human medulla-specific gene expression data from sufficiently large sets of primary human kidney samples, which has hampered our ability to identify regulatory mechanisms and medulla-specific diseases. In this study, we generated gene expression data from macro-dissected, primary kidney tissue samples and derived medulla- and cortex-specific patterns of gene transcription. We utilized this information to evaluate and correct the assignment of kidney tissue samples to cortex and medulla in the Genotype-Tissue Expression Project (GTEx) project, and performed a combined analysis with kidney samples from over 90 individuals in order to identify medulla-specific genes, transcription factors and enriched molecular functions and pathways. METHOD Macro-dissected matched pairs of healthy cortex and medulla tissue were sampled from kidneys of three tumor nephrectomy patients. Tissue identity was confirmed post facto histologically. RNA-sequencing was performed on an Illumina HiSeq with ∼50 M paired-end reads/sample. Differentially expressed genes (DEGs, P-adj < 0.01, log2-fc > 1) in cortex and medulla were identified with DESeq2. Additional kidney gene expression data with tissue assignments and pathology notes were downloaded from the GTEx V8 portal and DEG analyses were performed as described above. Principal component analysis (PCA) of gene count data was used to identify and compare sample clusters in both datasets. Reproducibly up-regulated DEGs in either cortex or medulla in both datasets were further analysed. RESULTS A PCA of our samples revealed clear separation into two clusters, matching the assignment to cortex and medulla. Conversely, two clusters identified by PCA from all kidney samples in the GTEx database each contained a mix of samples assigned to cortex and medulla (Figure 1a). We projected the GTEx samples along the principal components of our dataset, showing the two GTEx clusters correspond to medullary and cortical tissue (Figure 1b). Nine GTEx samples were assigned to the wrong tissue subtype, which was confirmed for six cases by review of GTEx pathology notes that mention contamination with the respective other kidney tissue subtype. After reclassification, nine medullary and 80 cortical GTEx samples were compared by DEG analysis and 3926 DEGs identified. Intersection with 2372 DEGs from our samples showed 402 reproducible, protein coding genes with higher medullary expression in both datasets (653 for cortex, respectively). These reproducible DEGs have highly correlating log2-fold changes between the two datasets (Spearman's ρ = 0.91, Figure 1c). Results showed high biological plausibility: First, medullary DEGs with high average expression included genes with well-established roles in diseases originating in the medulla such as UMOD and SLC12A1 (Table 1). Second, reproducible medulla DEGs were overrepresented in gene ontology terms related to urogenital system or kidney development and the organization of extracellular matrix. Conversely, the respective cortical genes were overrepresented for terms related to metabolic processes of small molecules handled by the proximal tubules (Figure 1d). CONCLUSION Here we show how a small set of carefully curated samples can be used to calibrate larger publicly available datasets such as GTEX to identify medulla-specific genes and pathways with increased resolution. This valuable resource will help us better understand diseases and conditions originating in the medullary portion of the human kidney.