Widespread Translational Control of Fibrosis in the Human Heart by RNA-Binding Proteins

Abstract

Background: Fibrosis is a common pathology in many cardiac disorders and is driven by the activation of resident fibroblasts. The global posttranscriptional mechanisms underlying fibroblast-to-myofibroblast conversion in the heart have not been explored.

Methods: Genome-wide changes of RNA transcription and translation during human cardiac fibroblast activation were monitored with RNA sequencing and ribosome profiling. We then used RNA-binding protein-based analyses to identify translational regulators of fibrogenic genes. The integration with cardiac ribosome occupancy levels of 30 dilated cardiomyopathy patients demonstrates that these posttranscriptional mechanisms are also active in the diseased fibrotic human heart.

Results: We generated nucleotide-resolution translatome data during the transforming growth factor β1-driven cellular transition of human cardiac fibroblasts to myofibroblasts. This identified dynamic changes of RNA transcription and translation at several time points during the fibrotic response, revealing transient and early-responder genes. Remarkably, about one-third of all changes in gene expression in activated fibroblasts are subject to translational regulation, and dynamic variation in ribosome occupancy affects protein abundance independent of RNA levels. Targets of RNA-binding proteins were strongly enriched in posttranscriptionally regulated genes, suggesting genes such as MBNL2 can act as translational activators or repressors. Ribosome occupancy in the hearts of patients with dilated cardiomyopathy suggested the same posttranscriptional regulatory network was underlying cardiac fibrosis. Key network hubs include RNA-binding proteins such as Pumilio RNA binding family member 2 (PUM2) and Quaking (QKI) that work in concert to regulate the translation of target transcripts in human diseased hearts. Furthermore, silencing of both PUM2 and QKI inhibits the transition of fibroblasts toward profibrotic myofibroblasts in response to transforming growth factor β1.

Conclusions: We reveal widespread translational effects of transforming growth factor β1 and define novel posttranscriptional regulatory networks that control the fibroblast-to-myofibroblast transition. These networks are active in human heart disease, and silencing of hub genes limits fibroblast activation. Our findings show the central importance of translational control in fibrosis and highlight novel pathogenic mechanisms in heart failure.

Keywords: RNA-binding proteins; TGF-beta1; dilated cardiomyopathy; fibrosis; ribosome profiling.

Figure 1.

Time-resolved stimulation of the fibrotic response. Primary human cardiac fibroblasts were isolated from the atrial biopsies of 4 individuals and stimulated with TGFβ1 (transforming growth factor β1) (5 ng/mL). (A) Microscopic images show fibroblasts at 5 timepoints (B, Baseline, 45 minutes, 2 hours, 6 hours, 24 hours after TGFβ1 stimulation) with immunostaining for Collagen I, α-smooth muscle actin (ACTA2) and periostin (POSTN). Scale bar equals 50 µm. (B) High-resolution fluorescence imaging with immunostaining of nuclei (DAPI, blue), ACTA2 (purple), and F-actin (phalloidin, cyan) showing TGFβ1 activates fibroblast stress fiber formation. Scale bars indicate 30 µm. (C through E) Fluorescence was quantified on the Operetta high-content imaging platform after immunostaining for Collagen I (C), ACTA2 (D), and POSTN (E) (28 measurements across 4 wells) and normalized for cell count (D) or cell area, I/A, intensity/area (C and E). Total secreted collagen (F), concentration (conc) of MMP2 (G), and TIMP-1 (H) in the supernatant of TGFβ1-stimulated cardiac fibroblasts (n=3, biologically independent samples) was quantified by Sirius red collagen assay (F) and by ELISA (G and H) respectively. P values were determined by one-way ANOVA and corrected for comparisons to the same sample (Baseline) using Dunnett’s test. *P<5×10^-2, **P<10^–4, ***P<10^–8, ****P<2×10^-16. (I) Western blotting of phosphorylated protein (P-) expression of SMAD (small mothers against decapentaplegic)2 and extracellular signal-regulated kinase (ERK) signaling molecules showed rapid activation. B, baseline (0 minutes).

Figure 2.

Ribosome profiling of TGFβ1 (transforming growth factor β1) stimulated primary human cardiac fibroblasts. (a) Sample level periodicity: Distribution of inferred peptidyl-site (P-site) locations (+12 offset) for each sample (4 patients over 5 time points) at annotated translation start sites reveals ribosomes located on the canonical start codon (AUG) and majority of the P-sites downstream of the start codon located in-frame. (b) The 3-nt (3-nucleotide)-periodicity for all 20 samples (P1–P4 patients, 5 time points) is >86%, indicating most reads represent actively translating ribosomes. R represents random 3-nt-periodicity of 33%. (c) Gene-level periodicity: P-site location across all annotated expressed (transcript per million mapped reads, TPM>1) genes (combined data from 4 patients over 5 time points) shows efficient ribosome drop-off at the canonical stop codon (UGA/UAG/UAA, represented by *). Rel. indicates relative.

Figure 3.

Genome-wide temporal transcriptional and translational landscape in cardiac fibrosis. (a) Western blots showing ribosome occupancy determining changes in protein levels independent of mRNA changes for translationally exclusive genes, FTL, FTH1, ITGA3. Control: GAPDH. B, basal. (b) Log-fold changes in the mRNA and ribosome occupancy at 45 minutes after TGFβ1 (transforming growth factor β1) stimulation. DTEG, differential translational-efficiency genes; DTG, differentially transcribed genes. (c) The interplay between DTGs and DTEGs showing several categories of gene expression regulation. Forwarded genes, where the occupancy changes are explained by the mRNA changes; Exclusive, where changes occur exclusively in TE without underlying mRNA changes; Buffered and Intensified, where both the TE and the mRNA are changing. (d and e) Western blotting of genes detected as buffered, PRKG1 (d) and intensified, HES1 (e). (f)Forwarded gene clusters (F2, F26, F11, F21) with transient changes in expression following TGFβ1 stimulation. FC, fold change; n, number of genes in the cluster.

Figure 4.

Posttranscriptional regulators in fibroblast activation. (a) RBP target overrepresentation test in the regulatory groups within DTGs and DTEGs. Z-score is the effect size of overrepresentation. Regulators per member represent the number of RBPs overrepresented per member of the group. (b)RBP overrepresentation test (FDR<1%) for regulatory patterns separate translationally activated and repressed clusters. RBP expression in response to TGFβ1 (transforming growth factor β1) is determined using significant RPF changes. RBPs that are not overrepresented for their targets in any translational regulated gene cluster are not shown. Cluster IDs are denoted by their regulatory groups and cluster number. B, buffered; B’, completely buffered (special case); and E, exclusive. Clusters with less than 50 genes, or with no RBP overrepresentation are not shown. DTEG indicates differential translation-efficiency gene; DTG, differentially transcribed genes; FC, fold change; FDR, false discovery rate; RBP, RNA-binding protein; RPF, ribosome protected fragment; and TE, target's translation efficiency.

Figure 5.

Posttranscriptional regulators in fibrosis and dilated cardiomyopathy. (a) Periostin (POSTN) and latent TGFβ binding protein 2 (LTBP2) are upregulated in DCM patients. (Fold change, FDR). (b) Cardiac expression (transcripts per million mapped reads, TPM) and differential expression (fold change, P-adjusted: P value corrected by Benjamini-Hochberg) of top 10 RNA binding proteins in DCM patients compared to nondiseased donors. (c) RBPs with significantly more correlated targets than expected by chance (RBP RPF vs target TE, FDR values are false discovery rates calculated by the Benjamini-Hochberg method), indicating translational control also in DCM patients. DCM indicates dilated cardiomyopathy; RBP, RNA-binding protein; RPF, ribosome protected fragment; and TE, target's translation efficiency.

Figure 6.

Posttranscriptional regulatory network in dilated cardiomyopathy. (a) Exemplars of RBP-target pairs correlated in the DCM heart. Cor, Spearman ranked correlation value; p, P value for correlation test. (b) RBP-target network in disease based on permutation tests ( ρ  ≥ 0.45 for visualization). (c) Patient stratification based on severity of fibrosis assessed using marker gene expression (only low and high severity groups shown; full clustering in Figure VIIb in the online-only Data Supplement). (d) RBP expression differences between patients with low and high severity of fibrosis. In red, P value for 2-tailed t test. DCM indicates dilated cardiomyopathy; QKI, Quaking; RBP, RNA-binding protein; RPF, ribosome protected fragment; TE, target's translation efficiency; and TPM, transcripts per million mapped reads.

Figure 7.

Effect on fibrotic phenotype after siRNA (short interfering RNA) knockdown of RNA-binding proteins. Knockdown of Quaking protein (QKI) and Pumilio RNA binding family member 2 (PUM2) followed by stimulation with TGFβ1 (transforming growth factor β1) (5 ng/mL) on fibroblasts from 3 cardiac patients. NT, nontargeting control siRNA. (a) Microscopic images show fibroblasts with immunostaining for α-smooth muscle actin (ACTA2). Scale bar equals 50 µm. TGFβ1 stimulation for 6 hours. (b and c) Fold changes were calculated for %ACTA2^+ve cells and MMP2 concentration with respect to NT siRNA in baseline fibroblasts. Fluorescence quantified on the Operetta high-content imaging platform after immunostaining for ACTA2 (28 measurements across 4 wells) and normalized for cell count (b). Total concentration of MMP2 in the supernatant of TGFβ1-stimulated cardiac fibroblasts (n=3, biologically independent samples) was quantified by ELISA (c). P values were determined by 1-way ANOVA and corrected for comparisons to the same sample (NT+TGFβ1) using Dunnett’s test. *P<1.5×10^-3; **P<2×10^-15; ***P< 2.22×10^-16.