Authors:
Aitken S, Magi S, Alhendi AM, Itoh M, Kawaji H, Lassmann T, et al.
Authors notes:
PLoS computational biology. 2015;11(4):e1004217
Keywords:
Immediate-early response, immediate-early genes, cell fate, cancer, non-coding RNA, microRNA
Abstract:
The immediate-early response mediates cell fate in response to a variety of extracellular stimuli and is dysregulated in many cancers.
However, the specificity of the response across stimuli and cell types, and the roles of non-coding RNAs are not well understood.
Using a large collection of densely-sampled time series expression data we have examined the induction of the immediate-early response in unparalleled detail, across cell types and stimuli.
We exploit cap analysis of gene expression (CAGE) time series datasets to directly measure promoter activities over time.
Using a novel analysis method for time series data we identify transcripts with expression patterns that closely resemble the dynamics of known immediate-early genes (IEGs) and this enables a comprehensive comparative study of these genes and their chromatin state.
Surprisingly, these data suggest that the earliest transcriptional responses often involve promoters generating non-coding RNAs, many of which are produced in advance of canonical protein-coding IEGs.
IEGs are known to be capable of induction without de novo protein synthesis.
Consistent with this, we find that the response of both protein-coding and non-coding RNA IEGs can be explained by their transcriptionally poised, permissive chromatin state prior to stimulation.
We also explore the function of non-coding RNAs in the attenuation of the immediate early response in a small RNA sequencing dataset matched to the CAGE data:
We identify a novel set of microRNAs responsible for the attenuation of the IEG response in an estrogen receptor positive cancer cell line.
Our computational statistical method is well suited to meta-analyses as there is no requirement for transcripts to pass thresholds for significant differential expression between time points, and it is agnostic to the number of time points per dataset.