Pathway identification through transcriptome analysis

Abstract

Systems-based, agnostic approaches focusing on transcriptomics data have been employed to understand the pathogenesis of polycystic kidney diseases (PKD). While multiple signaling pathways, including Wnt, mTOR and G-protein-coupled receptors, have been implicated in late stages of disease, there were few insights into the transcriptional cascade immediately downstream of Pkd1 inactivation. One of the consistent findings has been transcriptional evidence of dysregulated metabolic and cytoskeleton remodeling pathways. Recent technical developments, including bulk and single-cell RNA sequencing technologies and spatial transcriptomics, offer new angles to investigate PKD. In this article, we review what has been learned based on transcriptional approaches and consider future opportunities.

Figure 1.. Sex differences in mouse kidney transcriptome.

A) Kidney to body weight ratios (KBW) in mutant mice plotted over time starting 50 days after Pkd1 inactivation show that cystic disease progresses faster in males. B) Representative images of adult kidneys of wild type (top row) and mutant (lower row) kidneys of female (left column) and male (right column) mice. The number of differentially expressed genes is shown inside each box for the corresponding comparisons. Note that there are roughly as many differentially expressed genes between wild type female vs. male kidneys (3,502) as there are between grossly cystic mutant and control male kidneys (3,721). C) Heatmap plot showing genes that are differentially expressed between male and female control kidneys. Each column is a sample and each row a gene. This set of genes is over-represented in fatty acid oxidation (up-regulated [green] in males) and lipid transport (down-regulated [red] in males) pathways. D) Number of differentially expressed genes between mutant and control kidneys in females (red) or males (blue) of different ages that roughly correspond to mild to severe kidney phenotypes. Note that pre-cystic samples (younger females) have very few differentially expressed genes. Figure modified from Menezes et al.[18].

Figure 2.. Genes differentially expressed in the early and late onset models show relatively similar average fold change in both models.

A fold-change plot showing all genes that were differentially expressed between mutant and control kidneys (adj. p value < 0.05) in early onset model (green), late onset model (blue) and both models (red). This nearly diagonal distribution of the genes shows that genes that were differentially expressed in either model tend to show similar average fold change in the two models, suggesting that similar pathways may be implicated in the early and late onset models.

Figure 3.. Confounders in PKD cell line transcriptome.

RNAseq data of immortalized cell lines derived from kidney proximal tubules of Pkd1 conditional mice. For each cell line, Pkd1 was inactivated in vitro, yielding a mutant and corresponding control pair. Cells were then propagated in cell culture. Different batches correspond to different passage number. A) Principal component plot (PCA) showing that most of the variance in gene expression patterns can be attributed to cell line identify (PC1; 53%). Note that the second highest source of variation (PC2; 29%) is due to passage/batch of a mutant cell line. B) Number of differentially expressed gene (fdr-p value < 0.01) comparing two different mutant vs. control cell lines (red), different batches of the same cell line (green), or different cell lines (blue). Note that cell line and passage number had a greater effect than Pkd1 genotype in gene expression pattern (T. Terabayashi, unpublished data).

Figure 4.. scRNAseq data can inform interpretation of bulk RNAseq data.

A) GSE107585 kidney single-cell RNAseq data showing predicted cell identities in an unsupervised t-distributed stochastic neighbor embedding (tSNE) map (Park et al.[64]). B-D) Cells in tSNE map colored based on expression levels of specific genes. The genes shown were prioritized by a screen to identify novel ADPKD therapies in a mouse model of inducible Pkd1 deletion in epithelial cells of the distal nephron (E-MTAB-8086; Malas et al.[45]). B) Pbld2 was one of the top down-regulated genes. The expression pattern suggests that its down-regulation could be explained by proximal tubule drop-out. C) Spp1 was one of the top differentially expressed genes. Its increase could be explained by higher expression in mutant tissue, or by an increase in the proportion of distal segments as cystic disease progresses. D) Alox5ap was one of the differentially expressed genes selected for drug screening in an in vitro epithelial cell model. As the gene is mostly expressed in fibroblasts, a negative result in an in vitro system of epithelial cells may not exclude drug efficacy in altering fibrosis, for example.