Targeting transcription in heart failure via CDK7/12/13 inhibition

Abstract

Heart failure with reduced ejection fraction (HFrEF) is associated with high mortality, highlighting an urgent need for new therapeutic strategies. As stress-activated cardiac signaling cascades converge on the nucleus to drive maladaptive gene programs, interdicting pathological transcription is a conceptually attractive approach for HFrEF therapy. Here, we demonstrate that CDK7/12/13 are critical regulators of transcription activation in the heart that can be pharmacologically inhibited to improve HFrEF. CDK7/12/13 inhibition using the first-in-class inhibitor THZ1 or RNAi blocks stress-induced transcription and pathologic hypertrophy in cultured rodent cardiomyocytes. THZ1 potently attenuates adverse cardiac remodeling and HFrEF pathogenesis in mice and blocks cardinal features of disease in human iPSC-derived cardiomyocytes. THZ1 suppresses Pol II enrichment at stress-transactivated cardiac genes and inhibits a specific pathologic gene program in the failing mouse heart. These data identify CDK7/12/13 as druggable regulators of cardiac gene transactivation during disease-related stress, suggesting that HFrEF features a critical dependency on transcription that can be therapeutically exploited.

Fig. 1. CDK7/12/13 inhibition attenuates hallmark features of pathologic cardiomyocyte remodeling in vitro.

A Chemical structure of THZ1. B Representative image of NRVM treated ± THZ1 (at indicated concentrations) and PE (100 µM) for 48 h with cell area quantification (n = 50–55 randomly selected cells per condition). Scale bar: 30 μm. Bars denote mean ± SD. ***p = .0002. C Quantification of expression levels of indicated genes from real-time PCR (n = 3 for THZ1 1 nM + PE, and THZ1 10 nM ± PE groups; n = 2 for THZ1 0 nM ± PE and THZ1 1 nM − PE). Experiment was repeated three independent times with similar results. D Representative Western blots of NRVM with indicated treatment for specific targets. N = 3 independent experiments. Pol II phosphoforms and total Pol II were each probed from a separately run gel and membrane, with each lane loaded with an equivalent amount of total protein and lysate volume, derived from a common master stock of protein lysate for each designated condition. The loading control (α-tubulin) was probed on the same blot with Pol II Ser7P. For these Western blots, NRVM were harvested 30 min after exposure to PE. E Representative images of NRVM treated ± si-control (si-cntrl) or si-Cdk7/12/13 (TKD) and PE (100 µM) for 48 h with cell area quantification (n = 55 randomly selected cells per condition). Scale bar: 30 μm. Bars denote mean ± SD. *p = 0.0127. F Representative Western blots of NRVM with indicated treatment for specific targets. Pol II phosphoforms and total Pol II were each probed from a separately run gel and membrane, with each lane loaded with an equivalent amount of total protein and lysate volume, derived from a common master stock of protein lysate for each designated condition. The loading control (TBP) was probed on the same blot with Pol II Ser7P. For these Western blots, NRVM were harvested 30 min after exposure to PE. Experiment was repeated three independent times with similar results. G Densitometry of Western blots in Fig. 1F (n = 4 per condition). Box plots show center line as median, whiskers show maxima and minima, and box limits show upper and lower quartiles. ***p = 0.0002 for Ser2P: scr vs. PE, ***p = 0.0004 for Ser2P: PE vs. siTKD + PE, **p = 0.0048 for Ser5P: scr vs. PE, *p = 0.0131 Ser5P: PE vs siTKD + PE, ***p = 0.0006 for Ser7P: scr vs PE, ***p = 0.0005 for Ser7P: PE vs siTKD + PE. H Quantification of expression levels of indicated genes from real-time PCR (n = 4). I, J Representative hiPSC-CM images (alpha-actinin staining in green and DAPI in blue) and quantification of cell size for indicated treatments (n = 10 for all conditions except ET-THZ1–250nM for which n = 9). Each data point represents the mean cell area for a single cell-culture chamber. For each chamber, 25 cells were randomly selected for area quantification. Scale bar: 30 μm. Bars denote mean ± SD. Significance values represent comparisons against the group that was exposed only to ET-1 (10 nM). *p = 0.0395 for ET vs. ET-THZ1–5nM, *p = 0.0133 for ET vs. ET-THZ1–10nM. K Gene expression levels of NPPB in hiPSC-CM with indicated treatments (n = 10). Significance values represent comparisons against the group that was exposed only to ET-1 (10 nM). L Protein concentration of secreted NT-proBNP in hiPSC-CM in the culture medium (n = 10). Significance values represent comparisons against the group that was exposed only to ET-1 (10 nM). *p = 0.0281 for ET vs ET-THZ1–10nM. Data are shown as means ± SEM unless otherwise noted. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for all indicated comparisons. One-way ANOVA with Tukey’s multiple comparisons test was used for all statistical analyses. Exact p values are noted when possible. Source data are provided as a Source Data file.

Fig. 2. THZ1 blocks stress-responsive gene programs and Pol II enrichment in cultured cardiomyocytes.

A Schema of the experimental design. ChIP-seq for total Pol II was performed for 3 conditions: baseline, PE, and PE + THZ1. B A heatmap showing clustered, row-normalized global gene expression profile in each sample in the indicated treatment group (n = 3) with representative gene callouts. C Representative gene tracks of Pol II enrichment for canonical hypertrophic marker genes (Nppb and Nppa). D Accumulative Pol II occupancy on all promoters. E Empirical cumulative distribution plots of Pol II traveling ratios (TR) for genes that are differentially expressed in response to PE stimulation (PE vs baseline). TR is defined as the ratio of (Pol II signal at TSS) / (Pol II signal at gene body). ****p < 0.0001 for indicated comparisons. Source data are provided as a Source Data file.

Fig. 3. THZ1 ameliorates cardinal features of adverse cardiac remodeling and HFrEF pathogenesis in mice.

A Schema of the experimental design. B Two-dimensional echocardiographic quantification of left ventricle area fractional shortening in indicated groups (n = 8). Echocardiography for the sham groups (Sham-Veh and Sham-THZ1) were performed at the end of the study (week 8) and are plotted alongside the week 8 TAC data in Supplementary Fig. 4A–C. *p = 0.0339, **p = 0.0039. Two-way ANOVA with Holm–Sidak multiple comparisons correction was used for statistical analysis. C Representative M-mode echocardiographic images obtained at mid-papillary muscle level from parasternal short-axis view (8 weeks post-TAC). D, E Representative heart images and quantification of heart weight/tibial length (HW/TL) ratio at 8 weeks post-TAC. Scale bar: 3 mm. F Representative images of myocardium visualized by wheat germ agglutinin staining (red) and quantification of cardiomyocyte cross-sectional area in indicated groups at 8 weeks post-TAC (pooled analysis of 150 individual cardiomyocytes randomly selected from hearts of 3 independent animals per experimental condition). Scale bar: 20 μm. Data presented as standard box-and-whisker plots showing median cardiomyocyte area (horizontal line). Upper and lower quartiles are designated by the box and whiskers designate upper and lower extremes. G Representative heart cross sections stained with picrosirius red with quantification of fibrosis area in indicated groups (n = 5; 8 weeks post-TAC). **p = 0.0027. Scale bar: 200 μm. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001 for all indicated comparisons. Data are shown as means ± SEM unless otherwise noted. One-way ANOVA with Tukey’s multiple comparisons test was used for all statistical analyses unless noted. Exact p values are noted when possible. Source data are provided as a Source Data file.

Fig. 4. THZ1 ameliorates specific pathological gene programs in diverse cardiac cellular compartments in failing mouse hearts.

A Schema of the experimental design. B Principal component (PC) analysis showing global gene expression in indicated samples. C A heatmap showing clustered, row-normalized stress-induced gene expression profile in each sample in indicated group (n = 4). D Gene expression levels of indicated representative genes (n = 4) calculated from normalized DeSeq2 counts. Significance levels were based on the adjusted p-values reported by DeSeq2. Bars denote mean ± SD. *p = 0.0171 for Nppb: TAC-Veh vs TAC-THZ1, *p = 0.0178 for Ctgf: TAC-Veh vs TAC-THZ1. E Venn diagram demonstrating number of differentially expressed genes in each indicated compartment (cutoffs are specified in Methods section). *p < 0.05, ****p < 0.0001 for all indicated comparisons. F UMAP (Uniform Manifold Approximation and Projection) projection of cardiac cell compartments in mouse heart identified by integrating two public mouse heart datasets (n = 32,407 cells). Each dot represents a single cell. Different cell-type clusters are color-coded. G Gene expression distribution of cell-type enriched genes whose expression is increased with TAC and suppressed by THZ1 (in E) across cell compartments in adult mouse heart. Genes that were induced with TAC and suppressed by THZ1, as determined by bulk RNA sequencing, were cross-referenced with the integrated single cell dataset (in F) to determine the most likely cell-type compartment from which each transcript originated. The size of the dot indicates the percentage of cells with at least one compartment-defining transcript detected and the color of the dot represents the scaled average expression level of expressing cells. Representative heatmaps for differentially expressed genes emanating from each identified cell compartment are shown in Supplementary Fig 6. One-way ANOVA with Tukey’s multiple comparisons test was used for all statistical analyses. Exact p values are noted when possible. Source data are provided as a Source Data file.