Formal Meta-Analysis of Hypoxic Gene Expression Profiles Reveals a Universal Gene Signature

Abstract

Integrating transcriptional profiles results in identifying gene expression signatures that are more robust than those obtained for individual datasets. However, a direct comparison of datasets derived from heterogeneous experimental conditions is problematic, hence their integration requires applying of specific meta-analysis techniques. The transcriptional response to hypoxia has been the focus of intense research due to its central role in tissue homeostasis and prevalent diseases. Accordingly, many studies have determined the gene expression profile of hypoxic cells. Yet, despite this wealth of information, little effort has been made to integrate these datasets to produce a robust hypoxic signature. We applied a formal meta-analysis procedure to datasets comprising 430 RNA-seq samples from 43 individual studies including 34 different cell types, to derive a pooled estimate of the effect of hypoxia on gene expression in human cell lines grown ingin vitro. This approach revealed that a large proportion of the transcriptome is significantly regulated by hypoxia (8556 out of 20,888 genes identified across studies). However, only a small fraction of the differentially expressed genes (1265 genes, 15%) show an effect size that, according to comparisons to gene pathways known to be regulated by hypoxia, is likely to be biologically relevant. By focusing on genes ubiquitously expressed, we identified a signature of 291 genes robustly and consistently regulated by hypoxia. Overall, we have developed a robust gene signature that characterizes the transcriptomic response of human cell lines exposed to hypoxia in vitro by applying a formal meta-analysis to gene expression profiles.

An exploration of vascular cell development in mouse embryoid bodies under hypoxia through single cell RNA-seq

“…As a general rule, the number of genes differentially expressed has a strong positive correlation with sample size (i.e., the number of cells in the cluster), while the opposite happens with the magnitude of changes considered significant (median -LFC-, for instance), so the number of DEGs cannot be taken as an indication of the responsiveness of particular cell types to hypoxia. These effects were expected and have also been observed in a meta-analysis of bulk RNA-seq experiments [40]: the sum of DEGs increased with the number of studies included in the meta-analysis, while the mean effect size decreased. Thus, bearing in mind the limitations of a binary classification, we found a low overlap in the identities of genes differentially expressed due to hypoxia between the different clusters (Fig.…”

“…A-B: For each cluster, percentage of genes that are up-regulated (A) or down-regulated (B) in each or both time conditions. C-F: For each time condition, we divided the genes expressed in all clusters in two categories depending on their inclusion in a hypoxia driven up-regulation geneset[40]. For each set of genes, we have measured the coefficient of determination (R 2 ) between log2(Fold Change) values in each pair of clusters.…”

“…divided in two categories: I) the members of a meta-analysis derived gene set composed of genes widely up-regulated by hypoxia[40](FDR < 0.01, LFC > 0), which we will use as a proxy for the direct response to hypoxia, and II): the rest of commonly expressed genes. The coefficient of determination (R 2 ) was then calculated for the log 2 (Fold Change) values of each gene set and pair of clusters.…”

Bhlhe40 limits proliferation and prevents excessive angiogenesis in mouse embryoid bodies under hypoxia

“…In our recent meta-analysis of transcriptomic studies, which included 430 RNA-seq samples from 43 individual studies across 34 different cell types, we assessed the impact of hypoxia on the transcription of 20918 genes identified across the datasets [6]. This analysis yielded a hypoxic transcriptomic profile, assigning the estimated effect size of hypoxia (log 2 -Fold Change (log 2 FC)) and statistical significance for the change in expression to each of the 20918 genes.…”

“…Although HIF directly controls gene upregulation, gene downregulation in hypoxia is controlled by indirect mechanisms that remain incompletely understood [3–5]. By integrating data from 43 individual RNA-sequencing (RNAseq) studies conducted across 34 distinct cell types, we have recently discovered a signature comprising 291 ubiquitously expressed genes that are consistently and robustly (FDR < 0.01 and |log 2 FC| > 0.7) regulated by hypoxia [6]. Notably, the transcriptional repressor basic-helix-loop-helix family member e40 (Bhlhe40) emerged as one of the most consistently upregulated genes within this signature.…”