Prioritization of novel ADPKD drug candidates from disease-stage specific gene expression profiles

Abstract

Background: Autosomal Dominant Polycystic Kidney Disease (ADPKD) is one of the most common causes of end-stage renal failure, caused by mutations in PKD1 or PKD2 genes. Tolvaptan, the only drug approved for ADPKD treatment, results in serious side-effects, warranting the need for novel drugs.

Methods: In this study, we applied RNA-sequencing of Pkd1cko mice at different disease stages, and with/without drug treatment to identify genes involved in ADPKD progression that were further used to identify novel drug candidates for ADPKD. We followed an integrative computational approach using a combination of gene expression profiling, bioinformatics and cheminformatics data.

Findings: We identified 1162 genes that had a normalized expression after treating the mice with drugs proven effective in preclinical models. Intersecting these genes with target affinity profiles for clinically-approved drugs in ChEMBL, resulted in the identification of 116 drugs targeting 29 proteins, of which several are previously linked to Polycystic Kidney Disease such as Rosiglitazone. Further testing the efficacy of six candidate drugs for inhibition of cyst swelling using a human 3D-cyst assay, revealed that three of the six had cyst-growth reducing effects with limited toxicity.

Interpretation: Our data further establishes drug repurposing as a robust drug discovery method, with three promising drug candidates identified for ADPKD treatment (Meclofenamic Acid, Gamolenic Acid and Birinapant). Our strategy that combines multiple-omics data, can be extended for ADPKD and other diseases in the future.

Funding: European Union's Seventh Framework Program, Dutch Technology Foundation Stichting Technische Wetenschappen and the Dutch Kidney Foundation.

Keywords: 3D cyst assay; Autosomal dominant polycystic kidney disease; Cheminformatics; Drug repurposing; RNA-Sequencing.

Fig. 1

Kidneys taken out at various disease stages show differences in expression profiles. (a) Boxplot representation of the 2KW/BW values for groups of Pkd1cko mice representing different phases of ADPKD with increasing disease severity. (b) Results from principal component analysis of the Pkd1cko samples. Shown are the loadings of plot of pc1 (x-axis) and pc3 (y-axis) of all samples. In the panel the samples are colored based on their 2KW/BW value. (c) Boxplots of the top 10 most up-regulated (left part) and the 10 most down-regulated genes (right part) during disease progression, as extracted from the loadings of the genes on pc1. Expression data are given as log2 (counts per million).

Fig. 2

12 distinct expression patterns are associated with PKD progression. (a) The different expression patterns observed in Pkd1cko mice representing the progression of ADPKD towards end-stage renal disease (weeks after tamoxifen induction). For each cluster, mean log-transformed gene expression levels relative to the control mice that did not receive tamoxifen are plotted. The top panel represents the early dysregulated clusters, the middle panel represents the clusters dysregulated in the moderate to advanced stage and the bottom panel the clusters associated with the advanced stage of the disease. (b) Replication of expression profiles in an independent study. For each cluster, a representation factor reflecting the gene overlap of each cluster with the expression signatures from the five different mouse groups defined in the study by Menezes et al. is given in a color representation. A representation score > 1 reflects enrichment. (c) Correlation of gene expression with disease progression. For each cluster, the average Spearman's correlation coefficient between the expression values of the genes in a cluster and the 2KW/BW ratio was calculated. Green represents a negative correlation while red reflects a positive correlation. Clusters that were dysregulated in an early stage have the lowest correlation with the 2KW/BW increase, suggesting they follow a different trend in disease progression. (d) Association of cluster with drug response. A bar chart representation for each cluster showing the proportion of the 2731 genes that were also affected by one of the drug treatments: sActRIIB-Fc early (Act Early) and late (Act Late), curcumin, rapamycin short (Rapa Short) and long (Rapa Long). The x-axis represents the % of genes that were significanltly dysregulated (P < 0.05) due to the drug treatments per cluster per drug treatment.

Fig. 3

Pathways associated with disease progression and drug response. (a) A heatmap representation of the molecular pathways significantly enriched in the different stages of PKD (left). For each cluster category from Fig. 2A, the significantly enriched Wikipathways were obtained (FDR < 0.05) and plotted in the heatmap. Color scale reflects the representation factor in the different phases of ADPKD (Early, Moderate and Advanced). On the right, a heatmap representation of the pathways that are enriched with significantly dysregulated genes after drug treatment. Enrichment was established based on the representation (RF) factor calculation, where pathways that had RF >= 1 are considered significant. (b) A schematic representation of the different pathways involved in ADPKD's progression. Pathways are selected based on their Wikipathways significance (FDR) across the different disease stages in part A.

Fig. 4

3D-cysts assay of candidate compounds. (a) Top: Quantification of cyst size of the tested compounds normalized to forskolin induced swelling. Reference compounds rapamycin (0.01 µM) and staurosporin (0.25 µM) reduce cyst size, as well as brinapant, Gamolenic Acid, icosapent and Meclofenamic Acid at highest tested concentration of 100 µM, 500 µM, 500 µM and 100 µM respectively (N = 4 wells). Bottom: Assessment of staurosporin-like induction of toxicity. Graphs representing average nuclei area, nuclei roundness and the fraction of nuclei that are apoptotic show changes for reference compound staurosporin and for icosapent. (b)Representative images of positive and negative control and two of the test compounds at highest tested dose; 100 µM for brinapant and 500 µM for Gamolenic Acid. Each scalebar is 400 μM.