The relative contributions of transcriptional and post-transcriptional regulation to steady-state messenger RNA levels

The regulation of gene expression in eukaryotes is a complex process balancing two opposing schemes into one regulatory network. Stable maintenance of gene expression patterns is as important as quick adaption to intrinsic and extrinsic stimuli. Over the past years it has emerged that gene regulation is a multistep process occurring at many levels. On the level of DNA and chromatin it is determined how efficiently a gene is transcribed by RNA polymerase II (RNAP II) in the first place. Influenced by many processing steps, which are mediated amongst others by RNA binding proteins (RBPs), only a fraction of a respective gene arrives to the cytoplasm, where more regulatory processes alter the lifetime of messenger RNA (mRNA), during which it is available for translation into protein. Due to the local separation of nucleus and cytoplasm in eukaryotes it is intuitive to imagine a stepwise process, which can be split up in transcriptional regulation in the first place and subsequent post-transcriptional regulation.
At the beginning of my PhD high resolution genome-wide data of chromatin modifications and transcription became available, which allowed a global correlation of mRNA expression with chromatin features. Also supported through RNA sequencing data, more small regulatory RNAs were discovered and their expression linked to specific cell types. Both, histone marks influencing the chromatin environment and post-transcriptional processes operating on RNA level, have a contribution to the final mRNA concentration per gene in a cell. It was still largely unknown if these processes are separable and how much each process contributes to the final mRNA expression.
Therefore we set out to define the relative contributions of transcriptional and post-transcriptional regulation which shape the mRNA profile in a cell. To this end we obtained all necessary data from murine embryonic stem cells, which are differentiated into neurons in cell culture. Modifications at histone H3 (di-methylation of lysine 4 at histone tail H3 (H3K4me2), tri-methylation of lysine 27 at histone tail H3 (H3K27me3) and tri-methylation of lysine
36 at histone tail H3 (H3K36me3)) and RNAP II occupancy were derived by chromatin immunoprecipitation (ChIP) followed by deep sequencing to predict transcription rate. In addition we measure mRNA decay rates of protein coding genes both, by transcription arrest and pulse labeling and infer expression profiles of micro RNAs (miRNAs) during neuronal differentiation by small RNA sequencing.
Our integrative analysis in ESC revealed that chromatin marks are very good predictors of steady-state mRNA level. Especially, H3K36me3, which is a co-transcriptional histone mark, is highly correlated with mRNA abundance when integrated over the whole gene body. This is in contrast to two other studies, which also use histone marks to predict mRNA expression, however because their analysis is restricted to regions around the TSS, they do not use the full predictive power of the H3K36me3. Here we show that with H3K36me3, additional two promoter proximal histone marks and RNAP II occupancy, we can explain most of the variance in mRNA levels (?85%). Based on this result we went on to ask which regulatory mechanism could explain the additional variance in transcript levels, and investigated the contribution of mRNA de- cay to steady-state levels in general and in particular focus on miRNA-mediated degradation of transcripts.
This analysis, integrating mRNA half-life of each transcript in a model together with transcription-relevant measures, shows, that degradation has a minor quantitative impact on mRNA levels (<2%). This is in accordance with two recent publications in murine fibroblast and dendritic cells, which show, by measuring mRNA transcription rate and modeling RNA decay, a similar ratio of transcriptional and post-transcriptional regulation to quantify mRNA levels. Furthermore, we were interested in the quantitative contribution of mRNA degradation, which is mediated by miRNAs specifically. To this end we established weighted miRNA-target connections by combining a posterior probability score of interaction with experi- mentally inferred miRNA expression data. On a subset of likely miRNA target genes we can see a small effect of miRNA-mediated post-transcriptional decay, however on a genome- wide level the quantitative contribution of this regulatory layer is too small to be detectable.
Together, our findings establish a chromatin-based quantitative model for the contribution of transcriptional and post-transcriptional regulatory processes to steady-state levels of messenger RNA and support the recent notion that the lion share of mRNA expression regulation is happening at the level of transcription.