Inference of gene regulation functions from dynamic transcriptome data.

Patrick Hillenbrand,Ulrich Gerland,Patrick Cramer,Kerstin C Maier

doi:10.7554/elife.12188

Patrick Hillenbrand, Ulrich Gerland + Show 2 more

Open Access

https://doi.org/10.7554/elife.12188

Copy DOI

Abstract

To quantify gene regulation, a function is required that relates transcription factor binding to DNA (input) to the rate of mRNA synthesis from a target gene (output). Such a 'gene regulation function' (GRF) generally cannot be measured because the experimental titration of inputs and simultaneous readout of outputs is difficult. Here we show that GRFs may instead be inferred from natural changes in cellular gene expression, as exemplified for the cell cycle in the yeast S. cerevisiae. We develop this inference approach based on a time series of mRNA synthesis rates from a synchronized population of cells observed over three cell cycles. We first estimate the functional form of how input transcription factors determine mRNA output and then derive GRFs for target genes in the CLB2 gene cluster that are expressed during G2/M phase. Systematic analysis of additional GRFs suggests a network architecture that rationalizes transcriptional cell cycle oscillations. We find that a transcription factor network alone can produce oscillations in mRNA expression, but that additional input from cyclin oscillations is required to arrive at the native behaviour of the cell cycle oscillator.

Highlights

Much of the topology of the yeast transcriptional network is known from functional genomics approaches such as chromatin immunoprecipitation and mRNA expression data (Lee et al, 2002; Harbison et al, 2004; Tsai et al, 2005; Wu et al, 2006; Hu et al, 2007)
After selecting a target gene of interest, we compile a list of known input factors and focus on those that display a significant fold-change in mRNA level over the time course of the experiment
We assume that their dynamics can rationalize the output dynamics (Figure 1B) via a smooth input-output relation, the gene regulation function’ (GRF)

Summary

Introduction

Much of the topology of the yeast transcriptional network is known from functional genomics approaches such as chromatin immunoprecipitation and mRNA expression data (Lee et al, 2002; Harbison et al, 2004; Tsai et al, 2005; Wu et al, 2006; Hu et al, 2007). For a single-cell organism such as yeast, the reconstruction of GRFs must instead rely on the temporal variation of input factors. A major obstacle for inferring GRFs is that GRFs describe gene activity but expression data typically provide only total mRNA levels. This limitation is overcome by ‘Dynamic Transcriptome Analysis’ (DTA), which uses nonperturbing metabolic RNA labeling to obtain the amounts of newly synthesized mRNA, which can be used as a proxy for RNA synthesis rates and gene activity (Miller et al, 2011; Sun et al, 2012)

Methods

Results

Conclusion