Limits...
Transcriptional dynamics reveal critical roles for non-coding RNAs in the immediate-early response.

Aitken S, Magi S, Alhendi AM, Itoh M, Kawaji H, Lassmann T, Daub CO, Arner E, Carninci P, Forrest AR, Hayashizaki Y, FANTOM ConsortiumKhachigian LM, Okada-Hatakeyama M, Semple CA - PLoS Comput. Biol. (2015)

Bottom Line: 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.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.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.

View Article: PubMed Central - PubMed

Affiliation: MRC Institute of Genetics and Molecular Medicine, University of Edinburgh, Edinburgh, United Kingdom.

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.

No MeSH data available.


Related in: MedlinePlus

Kinetic signatures for IEGs.(A) Kinetic signatures are defined as piece-wise exponential (peak and dip), simple exponential (decay) or linear functions.(B) CAGE clusters associated with known IEGs show significant expression at time 0 (left; median 14.7 TPM). The maximum log2 fold change at any point in the time course over expression at time 0 is typically less than 2 (right; median 1.64). Histograms show data from all four data sets for 194l known IEGs. (C) Kinetic signatures fitted to the CAGE time course of EGR1 in EGF treated MCF7 cells yield values for the fit (log Z) and estimates for parameter moments. Plots show the kinetic signature function using computed parameter means (blue) and confidence intervals (red) for peak (left) and linear (right) kinetic signatures. In this case, log Z for the peak signature (-27.2) is greater than that for the linear model (-35), indicating a significantly better explanation of the data. Data values are plotted as circles (median value is filled). (D) CAGE time course data and best-fitting kinetic signature for IEGs JUN, FOS, EGR1 and DUSP1 (colours as in (C)). The vertical green lines indicate the mean switch time tS and one standard deviation above and below.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4401570&req=5

pcbi.1004217.g001: Kinetic signatures for IEGs.(A) Kinetic signatures are defined as piece-wise exponential (peak and dip), simple exponential (decay) or linear functions.(B) CAGE clusters associated with known IEGs show significant expression at time 0 (left; median 14.7 TPM). The maximum log2 fold change at any point in the time course over expression at time 0 is typically less than 2 (right; median 1.64). Histograms show data from all four data sets for 194l known IEGs. (C) Kinetic signatures fitted to the CAGE time course of EGR1 in EGF treated MCF7 cells yield values for the fit (log Z) and estimates for parameter moments. Plots show the kinetic signature function using computed parameter means (blue) and confidence intervals (red) for peak (left) and linear (right) kinetic signatures. In this case, log Z for the peak signature (-27.2) is greater than that for the linear model (-35), indicating a significantly better explanation of the data. Data values are plotted as circles (median value is filled). (D) CAGE time course data and best-fitting kinetic signature for IEGs JUN, FOS, EGR1 and DUSP1 (colours as in (C)). The vertical green lines indicate the mean switch time tS and one standard deviation above and below.

Mentions: Four kinetic signature functions were defined as illustrated in Fig 1A (see Materials and methods for details). These patterns were intended to capture mRNA transcription in response to a stimulus. Such exponential kinetics are characteristic of formalised systems biology models (comparable with observed and modelled mRNA and pre-mRNA expression in [4, 9]), and may reflect changes in both transcription and degradation rates over time [20]. The genome-wide CAGE data considered here necessarily included transcripts whose functions are unknown thus we began by hypothesising the possible kinetics they may display, rather than by constructing a detailed, interconnected systems model. Kinetic signatures serve as prototypical patterns reflecting changes in regulation, and are used here as a means to categorise time course responses for each transcript present. We focused on four particular time series datasets: human aortic smooth muscle cells (AoSMC) treated with FGF2 and with IL-1β (9 time points from 0 to 360 min; 3 replicates per treatment; IL-1β will be referred to as IL1b hereafter), as well as human MCF7 breast cancer cells treated with EGF and HRG (16 time points from 0 to 480 min; 3 replicates per treatment). Aortic smooth muscle cells are primary cells which are components of blood vessels. They are normally growth-quiescent in the normal adult vessels, but are activated by injury, or exposure to growth factors (including FGF2) and pro-inflammatory cytokines (including IL1b). These cues are sensed by these cells through changes in immediate-early gene expression, and can lead to increased proliferation and migration.


Transcriptional dynamics reveal critical roles for non-coding RNAs in the immediate-early response.

Aitken S, Magi S, Alhendi AM, Itoh M, Kawaji H, Lassmann T, Daub CO, Arner E, Carninci P, Forrest AR, Hayashizaki Y, FANTOM ConsortiumKhachigian LM, Okada-Hatakeyama M, Semple CA - PLoS Comput. Biol. (2015)

Kinetic signatures for IEGs.(A) Kinetic signatures are defined as piece-wise exponential (peak and dip), simple exponential (decay) or linear functions.(B) CAGE clusters associated with known IEGs show significant expression at time 0 (left; median 14.7 TPM). The maximum log2 fold change at any point in the time course over expression at time 0 is typically less than 2 (right; median 1.64). Histograms show data from all four data sets for 194l known IEGs. (C) Kinetic signatures fitted to the CAGE time course of EGR1 in EGF treated MCF7 cells yield values for the fit (log Z) and estimates for parameter moments. Plots show the kinetic signature function using computed parameter means (blue) and confidence intervals (red) for peak (left) and linear (right) kinetic signatures. In this case, log Z for the peak signature (-27.2) is greater than that for the linear model (-35), indicating a significantly better explanation of the data. Data values are plotted as circles (median value is filled). (D) CAGE time course data and best-fitting kinetic signature for IEGs JUN, FOS, EGR1 and DUSP1 (colours as in (C)). The vertical green lines indicate the mean switch time tS and one standard deviation above and below.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4401570&req=5

pcbi.1004217.g001: Kinetic signatures for IEGs.(A) Kinetic signatures are defined as piece-wise exponential (peak and dip), simple exponential (decay) or linear functions.(B) CAGE clusters associated with known IEGs show significant expression at time 0 (left; median 14.7 TPM). The maximum log2 fold change at any point in the time course over expression at time 0 is typically less than 2 (right; median 1.64). Histograms show data from all four data sets for 194l known IEGs. (C) Kinetic signatures fitted to the CAGE time course of EGR1 in EGF treated MCF7 cells yield values for the fit (log Z) and estimates for parameter moments. Plots show the kinetic signature function using computed parameter means (blue) and confidence intervals (red) for peak (left) and linear (right) kinetic signatures. In this case, log Z for the peak signature (-27.2) is greater than that for the linear model (-35), indicating a significantly better explanation of the data. Data values are plotted as circles (median value is filled). (D) CAGE time course data and best-fitting kinetic signature for IEGs JUN, FOS, EGR1 and DUSP1 (colours as in (C)). The vertical green lines indicate the mean switch time tS and one standard deviation above and below.
Mentions: Four kinetic signature functions were defined as illustrated in Fig 1A (see Materials and methods for details). These patterns were intended to capture mRNA transcription in response to a stimulus. Such exponential kinetics are characteristic of formalised systems biology models (comparable with observed and modelled mRNA and pre-mRNA expression in [4, 9]), and may reflect changes in both transcription and degradation rates over time [20]. The genome-wide CAGE data considered here necessarily included transcripts whose functions are unknown thus we began by hypothesising the possible kinetics they may display, rather than by constructing a detailed, interconnected systems model. Kinetic signatures serve as prototypical patterns reflecting changes in regulation, and are used here as a means to categorise time course responses for each transcript present. We focused on four particular time series datasets: human aortic smooth muscle cells (AoSMC) treated with FGF2 and with IL-1β (9 time points from 0 to 360 min; 3 replicates per treatment; IL-1β will be referred to as IL1b hereafter), as well as human MCF7 breast cancer cells treated with EGF and HRG (16 time points from 0 to 480 min; 3 replicates per treatment). Aortic smooth muscle cells are primary cells which are components of blood vessels. They are normally growth-quiescent in the normal adult vessels, but are activated by injury, or exposure to growth factors (including FGF2) and pro-inflammatory cytokines (including IL1b). These cues are sensed by these cells through changes in immediate-early gene expression, and can lead to increased proliferation and migration.

Bottom Line: 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.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.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.

View Article: PubMed Central - PubMed

Affiliation: MRC Institute of Genetics and Molecular Medicine, University of Edinburgh, Edinburgh, United Kingdom.

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.

No MeSH data available.


Related in: MedlinePlus