Glucocorticoid- and pioglitazone-induced proteinuria reduction in experimental NS both correlate with glomerular ECM modulation

Abstract

Idiopathic nephrotic syndrome (NS) is a common glomerular disease. Although glucocorticoids (GC) are the primary treatment, the PPARγ agonist pioglitazone (Pio) also reduces proteinuria in patients with NS and directly protects podocytes from injury. Because both drugs reduce proteinuria, we hypothesized these effects result from overlapping transcriptional patterns. Systems biology approaches compared glomerular transcriptomes from rats with PAN-induced NS treated with GC vs. Pio and identified 29 commonly regulated genes-of-interest, primarily involved in extracellular matrix (ECM) remodeling. Correlation with clinical idiopathic NS patient datasets confirmed glomerular ECM dysregulation as a potential mechanism of injury. Cellular deconvolution in silico revealed GC- and Pio-induced amelioration of altered genes primarily within podocytes and mesangial cells. While validation studies are indicated, these analyses identified molecular pathways involved in the early stages of NS (prior to scarring), suggesting that targeting glomerular ECM dysregulation may enable a future non-immunosuppressive approach for proteinuria reduction in idiopathic NS.

Keywords: Gene network; Nephrology; Systems biology; Transcriptomics.

Graphical abstract

Figure 1

Glucocorticoids and pioglitazone both ameliorate NS-associated glomerular gene expression profile dysregulation (A) UPCR of individual animals on Day 11 post-PAN, PAN + MP, PAN + Pio, or PBS control; n = 4 rats per group ∗p < 0.05. Line with bars represents mean ± SEM. (B) Principal-component analysis (PCA) plot representing unsupervised clustering of RNA-seq data. Colored bubbles (n = 4rats/group) represent the confidence interval of the combined four points, and each point within the bubble represents the transcriptome (∼16,915 annotated genes) from each rat’s pooled glomeruli. (C) Heatmap representation of 1,872 differentially expressed genes (DEGs). Red to blue scale denotes Z score.

Figure 2

Commonly regulated DEGs and pathways in GC and Pio-treated PAN-NS rats (A) Venn diagram representing the number of genes either upregulated or downregulated in PAN vs. Control (FDR <0.05) and in treatment groups vs. PAN (FDR <0.05). BLUE outer oval denotes downregulated genes, and ORANGE outer oval denotes upregulated genes. FDR is defined as false discovery rate, adjusted for multiple testing with the Benjamini-Hochberg procedure. (B) Line plot representing the fold change (Log2FC) of 319 and 126 DEGs significantly induced and suppressed by PAN, respectively (FDR <0.05, PAN vs. Control), but significantly reversed by treatment groups (FDR<0.05, PAN + MP, PAN + Pio vs. PAN). BLACK line denotes PAN. RED line denotes PAN + MP. GREEN line denotes PAN + Pio. (C) Bar graph representing the functional annotation of 319 PAN-induced DEGs and 126 PAN-suppressed DEGs plotted based on enrichment scores using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) functional annotation analysis platform. Enrichment score ranks the biological significance of gene groups based on the overall Fisher exact score of all enriched annotation terms.

Figure 3

Drug-nuclear receptor-DEG interaction network-based analysis of PAN-induced and PAN-suppressed DEGs revealed 20 genes-of-interest (Method 1) (A) An ingenuity pathway analysis (IPA)-derived interaction network map of 319 PAN-induced and 126 PAN-suppressed genes with both nuclear receptors (glucocorticoid receptor [NR3C1; shown in YELLOW] and PPARγ receptor [PPARG; shown in BLUE]) and their respective agonists (methylprednisolone; shown in BROWN; dexamethasone, shown in PINK, and pioglitazone; shown in GREEN). The interaction network map was generated via the curated IPA knowledge base. Genes in BOLD denote the 20 genes-of-interest that were commonly altered by both nuclear receptors (NR3C1 and PPARγ) and/or the receptors’ agonists. (B) The corresponding protein-protein interactions of the 20 genes-of-interest determined by the STRING platform. Four clearly segregated clusters became apparent based on their respective biological annotation (biological processes, molecular function, cellular compartment). GRAY nodes represent proteins, and the line thickness between nodes signifies the strength of data supporting the association (i.e., thicker lines denote more available data sources showing the interaction, based on the STRING database). Interaction data sources included text mining, experiments, and databases. Minimum required interaction scores were set on medium confidence at 0.400. Clustering was performed using k-means clustering. Disconnected nodes were removed from the network.

Figure 4

GSEA enriched gene sets encoding genes involved in extracellular matrix remodeling and cell-cycle regulation (Method 2) Gene set enrichment analysis (GSEA) shows enrichment of (A) NABA_MATRISOME_ASSOCIATED, (B) NABA _ECM_REGULATORS, (C) NABA_CORE_MATRISOME, and (D) SA_REG_CASCADE_OF_CYCLIN_EXPR related genes among the PAN, PAN + MP, and PAN + Pio genes that were ranked by normalized enrichment signal (NES) and q-value (FDR). The heatmap represents gene expression values of core-enriched genes, which account for the enrichment signal and thus represent the small subset of all the genes that participate in a biological process. Gene sets with a false discovery rate (FDR) value <0.05 after 1,000 permutations were considered significant.

Figure 5

Drug-nuclear receptor-DEG interaction network-based analysis of PAN-induced and PAN-suppressed DEGs revealed 20 genes-of-interest (Method 2) (A) Ingenuity pathway analysis (IPA)-based interaction network formed between nuclear receptor (NR3C1: YELLOW, PPARG: BLUE) and drug (dexamethasone: PINK, methylprednisolone: BROWN, pioglitazone: GREEN)—targets from the core-enriched genes from GSEA (shown in GRAY color). Genes in BOLD denote genes-of-interest that commonly interact with NR3C1 and PPARG receptors and/or their respective agonists. (B) The corresponding protein-protein interactions of the 14 genes-of-interest determined by the STRING platform. Two clearly segregated clusters became apparent based on their respective biological annotation (biological processes, molecular function, cellular compartment). GRAY nodes represent proteins, and the line thickness between nodes signifies the strength of data supporting the association (i.e., thicker lines denote more available data sources showing the interaction, based on the STRING database). Interaction data sources included text mining, experiments, and databases. Minimum required interaction scores were set on medium confidence at 0.400. Clustering was performed using k-means clustering. Disconnected nodes were removed from the network.

Figure 6

Altered ECM- and growth-factor-activity-related glomerular gene expression patterns in rats correlate extensively with those in patients with glomerular disease (A) Rat glomerular RNAseq results of ECM-related genes with bar graphs mean ± SEM with ∗p < 0.05 vs. Control; #p < 0.05 vs. PAN only, §p < 0.05 vs. PAN + Pio. (B) Hodgin MCD and FSGS glomerular data and Woroniecka DKD (DN) glomerular data from the Nephroseq database. Heatmap data are expressed as the log2-median centered intensity with fold changes relative to normal condition and significant changes noted as ∗p < 0.05 vs. Normal Kidney/Healthy Donor. ANGPTL4 was not included in the DKD (DN) dataset.

Figure 7

Altered cytoplasmic-organization- and lipid metabolism-related glomerular gene expression patterns in rats correlate broadly with those in patients with glomerular disease Rat glomerular RNAseq results of (A) cytoplasmic-organization- and (B) lipid metabolism-related genes with bar graphs mean ± SEM with ∗p < 0.05 vs. Control; #p < 0.05 vs. PAN only, §p < 0.05 vs. PAN + Pio. (C) Expression trends in PAN-NS rats, MCD and primary FSGS glomeruli, DKD (DN) glomeruli, and MP/Pio treated rats. ∗p < 0.05 vs. Control/Normal Kidney/Healthy Donor; #p < 0.05 vs. PAN only, §p < 0.05 vs. PAN + Pio.

Figure 8

Altered DNA binding- and cell-cycle-related glomerular gene expression patterns in rats moderately with those in patients with glomerular disease (A) Rat glomerular RNAseq results of DNA-binding- and cell-cycle-related genes with bar graphs mean ± SEM with ∗p < 0.05 vs. Control; #p < 0.05 vs. PAN only, §p < 0.05 vs. PAN + Pio. (B) Expression trends in PAN-NS rats, MCD glomeruli, primary FSGS glomeruli, DKD (DN) glomeruli, and MP/Pio treated rats. ∗p < 0.05 vs. Control/Normal Kidney/Healthy Donor; #p < 0.05 vs. PAN only, §p < 0.05 vs. PAN + Pio. AIFM3 was not included in the DKD (DN) dataset.

Figure 9

Glucocorticoids and pioglitazone both reverse PAN-induced mRNA changes in podocyte- and mesangial-cell-specific, but not endothelial-cell-Specific, gene clusters Heatmap representations (A–C) of treatment-induced gene expression changes in podocyte-specific genes [(A): podocyte cluster, n = 66 genes], mesangial-cell-specific genes [(B): mesangial cluster n = 43 genes], and endothelial-cell-specific genes [(C): endothelial cluster n = 42 genes]. The heatmap scale is based on raw Z scores calculated from the normalized read counts. The BLUE color denotes lower expression (i.e., downregulation), whereas the RED color denotes higher expression (i.e., upregulation). The normalized count cutoff for the selection of genes for each cluster was set as ≥500. (D) Podocyte-specific, (E) mesangial-cell-specific, and (F) endothelial-cell-specific PCA plots with ellipses as confidence intervals of four points per group combined, where each point of a sample group denotes 66 genes for podocyte cluster, 43 genes for mesangial cluster, and 42 genes for endothelial cluster. The heatmap compares the (G) gene function dynamics and (H) toxicities for PAN vs. PAN + MP vs. PAN + Pio treatments in respect to control, derived from IPA analyses of the combined podocyte + mesangial cell gene clusters. The color spectrum ranges from negative (inhibition) Z scores (shown in BLUE) to positive (activation) Z scores (shown in ORANGE).