Application of Gene Expression Trajectories Initiated from ErbB Receptor Activation Highlights the Dynamics of Divergent Promoter Usage

Abstract

Understanding how cells use complex transcriptional programs to alter their fate in response to specific stimuli is an important question in biology. For the MCF-7 human breast cancer cell line, we applied gene expression trajectory models to identify the genes involved in driving cell fate transitions. We modified trajectory models to account for the scenario where cells were exposed to different stimuli, in this case epidermal growth factor and heregulin, to arrive at different cell fates, i.e. proliferation and differentiation respectively. Using genome-wide CAGE time series data collected from the FANTOM5 consortium, we identified the sets of promoters that were involved in the transition of MCF-7 cells to their specific fates versus those with expression changes that were generic to both stimuli. Of the 1,552 promoters identified, 1,091 had stimulus-specific expression while 461 promoters had generic expression profiles over the time course surveyed. Many of these stimulus-specific promoters mapped to key regulators of the ERK (extracellular signal-regulated kinases) signaling pathway such as FHL2 (four and a half LIM domains 2). We observed that in general, generic promoters peaked in their expression early on in the time course, while stimulus-specific promoters tended to show activation of their expression at a later stage. The genes that mapped to stimulus-specific promoters were enriched for pathways that control focal adhesion, p53 signaling and MAPK signaling while generic promoters were enriched for cell death, transcription and the cell cycle. We identified 162 genes that were controlled by an alternative promoter during the time course where a subset of 37 genes had separate promoters that were classified as stimulus-specific and generic. The results of our study highlighted the degree of complexity involved in regulating a cell fate transition where multiple promoters mapping to the same gene can demonstrate quite divergent expression profiles.

Fig 1. Outline of the trajectory models and experimental design of the data.

A. Schematic view of the gene expression trajectory models for MCF-7 cells undergoing proliferation or differentiation in response to EGF or HRG, respectively. B. Experimental design of the time course experiments where time points were selected to cover early to late stages of the cell fate transition.

Fig 2. Examples of promoters with generic and stimuli-specific expression profiles.

A. the promoter region maps to gene SEMA3F and has a generic expression profile across the EGF and HRG profiles. B. the promoter mapping to SULF2 also has a generic profile. C. the stimuli-specific promoter maps to gene FHL2 and D. the stimuli-specific promoter maps to gene FLNA.

Fig 3. Expression activity of the generic and stimulus-specific promoters across the time course.

Heat maps showing the expression levels of promoters in the A. generic and B. stimulus-specific groups under EGF or HRG treatments. Expression levels have been scaled by row (promoter). Generic promoters drive predominantly immediate early expression, while stimulus-specific ones drive mid-delayed expression; expression changes are mainly driven by the HRG treatment, being much less obvious and dramatic after EGF stimulation.

Fig 4. Time-course expression profiles of stimulus-specific promoters associated with significant fold changes of HRG-induced expression versus EGF-induced expression.

A. and B. show examples of promoters with an overall increase in expression. C. and D. show promoters with an overall decrease in expression.

Fig 5. Protein-protein interaction networks (PPI) of genes mapped by the generic and stimulus-specific promoters.

A. PPI sub-network of shortest paths connecting the 38 transcription factors controlled by generic promoters (32 controlled by generic promoters only, and 6 controlled by alternative promoters classified as either generic or stimulus-specific; note that ZBTB42 (Zinc Finger And BTB Domain Containing 42) does not appear, as it is not listed in iRefIndex). B. PPI sub-network connecting the 68 transcription factors controlled by stimulus-specific promoters (62 controlled by stimulus-specific promoters only, and 6 controlled by alternative promoters classified as either generic or stimulus-specific; note that FOXQ1 (forkhead box Q1) does not appear, as it is not listed in iRefIndex). Larger nodes denote TFs, smaller nodes represent non-TF interactors. Green and yellow nodes represent genes mapped by generic promoters and stimulus-specific promoters respectively. Red nodes denote genes that map to both generic and stimulus-specific promoters. Square nodes, like EGF1, identify genes that map to at least one promoter showing significant increments in expression in the HRG over EGF time course.