Dual transcriptional activator and repressor roles of TBX20 regulate adult cardiac structure and function

Abstract

The ongoing requirement in adult heart for transcription factors with key roles in cardiac development is not well understood. We recently demonstrated that TBX20, a transcriptional regulator required for cardiac development, has key roles in the maintenance of functional and structural phenotypes in adult mouse heart. Conditional ablation of Tbx20 in adult cardiomyocytes leads to a rapid onset and progression of heart failure, with prominent conduction and contractility phenotypes that lead to death. Here we describe a more comprehensive molecular characterization of the functions of TBX20 in adult mouse heart. Coupling genome-wide chromatin immunoprecipitation and transcriptome analyses (RNA-Seq), we identified a subset of genes that change expression in Tbx20 adult cardiomyocyte-specific knockout hearts which are direct downstream targets of TBX20. This analysis revealed a dual role for TBX20 as both a transcriptional activator and a repressor, and that each of these functions regulates genes with very specialized and distinct molecular roles. We also show how TBX20 binds to its targets genome-wide in a context-dependent manner, using various cohorts of co-factors to either promote or repress distinct genetic programs within adult heart. Our integrative approach has uncovered several novel aspects of TBX20 and T-box protein function within adult heart. Sequencing data accession number (http://www.ncbi.nlm.nih.gov/geo): GSE30943.

Figure 1.

Impact of Tbx20 ablation on the adult mouse cardiac transcriptome. (A) The scatter plot of expression levels per gene in RPKM. To enable calculation of log values of genes with 0 RPKM, 1 was added to all data points. Correlation between two RNA-seq control samples (cyan) was higher than control and Tbx20 conditional knockout samples, reflecting differential expression. Green, down-regulated genes; red, up-regulated. (B) Genes differentially expressed in Tbx20 knockout hearts and TBX20 ChIP-seq target genes at adjusted P-value of <0.001, >1.5-fold expression difference. The pie chart depicts the number of differentially expressed genes, according to RNA-seq data (numbers in blue). White ovals represent genes with a nearby TBX20 ChIP peak (ChIP-seq data). Numbers in orange represent the intersection between each set of differentially expressed genes and genes with a nearby TBX20-binding region, representing putative TBX20 direct targets. All genes with no nearby TBX20 ChIP peak were considered putative indirect targets. The fold-difference between the observed overlap with the ChIP set and number expected (based on the size of each category) is shown near each intersection.

Figure 2.

Distinct molecular functions are associated with TBX20 activation and repression of transcription. (A) The log ratio of the abundance of functional categories (see Materials and Methods) between down- and up-regulated genes (RNA-seq data only). (B) The log ratio of the abundance of functional categories between genes with a nearby TBX20-binding region (putative TBX20 direct targets) that are down- and up-regulated (RNA-seq and ChIP-seq data combined). (C) Comparison of GO term fold-enrichment among (i) genes with a nearby TBX20 ChIP peak, (ii) all down-regulated, (iii) down-regulated with a nearby TBX20 ChIP peak (putative direct targets) and (iv) down-regulated genes with no nearby TBX20 ChIP peak (putative indirect targets). Arrows show instances in which enrichment is lower in direct targets than in indirect targets, suggesting that these categories are enriched for genes that are indirectly regulated in the absence of TBX20.

Figure 3.

Comparison of genomic features among peak sets grouped by gene expression pattern. (A) The frequency of TBX20 DNA-binding motif identified de novo from ChIP-seq data (14). (B) The peak height was determined by the peak calling program QuEST (45) and is a proxy for the binding frequency or strength. The higher the peak height, more frequent binding was observed. (C) The frequency of peaks overlapping (±200 bp) with transcription start sites (TSS). (D) TBX20-binding region frequency per gene locus (regions/bp). (E) Conservation as measured by the phastCons algorithm (48). (F) Overlap with repeats. Asterisks indicate statistically significant differences (P < 0.05), compared with all expressed genes (A–D) or to sequences randomly selected from the genome (E and F). A description of the data presented in each panel and the statistical tests used are provided in Materials and Methods. Not expressed consists of the ‘no reads’ category. Results for the not expressed set with background levels were similar and are not shown for clarity sake (Supplementary Material, Fig. S6).

Figure 4.

Overlap of TBX20-binding regions with p300, GATA4, NKX2–5 and TBX5-binding regions. The set of genes with a nearby TBX20 ChIP-seq peak has different proportions of differentially expressed genes, compared with all genes (Fig. 1). Overlap of TBX20 ChIP peaks with (A) p300 (20), (B) GATA4, (C) NKX2–5 and (D) TBX5 (19) ChIP peaks further biases these proportions, suggesting that these proteins co-regulate gene expression with TBX20. X-axis: fold-difference between (i) the proportion of genes in a given expression category with a nearby TBX20 peak overlapping another ChIP peak and (ii) the proportion of genes in the same category with any nearby TBX20 peak (TBX20 only). Asterisks indicate statistically significant differences (P < 0.05), compared with TBX20 only. A description of the statistical test used is provided in Materials and Methods. Not expressed consists of the ‘no reads’ category. Results for the not expressed set with background levels were similar and are not shown for clarity sake (Supplementary Material, Fig. S6).

Figure 5.

Comparison of T-box DNA-binding motifs and -binding affinities. (A) TBX20 DNA-binding motif obtained de novo from ChIP-seq data (14) and (B) EOMES DNA-binding motif determined by protein-binding microarrays (31). Representative sequences from each motif used to determine their binding affinities and their relative K_d's are shown on the right.