Advances in reprogramming enable the switch of one cell fate to another with potential use in regenerative therapy. Cardiomyocyte (CM)-like cells can be reprogrammed from somatic fibroblasts by overexpression of cardiac genes in vitro (1–6) and in vivo (5, 7–10). However, efficiently transdifferentiating human non-cardiac cells into highly functional CMs remains a significant challenge (1, 4, 6). In contrast to conventional reprogramming by genetic methods, a chemical reprogramming approach introduces small molecules that interact with and modulate endogenous factors in the starting cell type (e.g., fibroblast) in the absence of target cell type–specific proteins. Small molecules have certain advantages: they are convenient to use, can be efficiently delivered into cells, provide greater temporal control, are nonimmunogenic and more cost-effective. Moreover, their effects can be fine-tuned by varying their concentrations and combinations. Here, we report identification of a combination of small molecules that enable reprogramming of human fibroblasts into chemically-induced functional CMs (ciCMs) that uniformly display contractile properties.
To induce cardiac reprogramming of human fibroblasts, we used an established human foreskin fibroblast (HFF) line that contains no cardiac cells, as assayed by quantitative RT-PCR (qPCR), flow-cytometry, and immunofluorescence analyses (fig. S1). HFFs were virally transduced with an alpha myosin heavy chain (αMHC)-GFP reporter (11) to specifically label CMs. Based on the cell-activation and signaling-directed/CASD reprogramming paradigm (12,13), we treated cells with compounds that induce or enhance cellular reprogramming to alternative fates (cell activation) in conjunction with cardiogenic molecules (signaling-directed) to induce cardiogenesis. We initially screened a collection of 89 small molecules known to facilitate reprogramming (table S1). Each compound was added to a baseline cocktail containing SB431542, CHIR99021, parnate, and forskolin, which enabled cardiac reprogramming in an earlier study when combined with a single gene, Oct4 (13). Cells were treated with the various small molecule cocktails for 6 days and then cultured for 5 days in an optimized cardiac induction medium (CIM) containing cardiogenic molecules including activin A, bone morphogenetic protein 4 (BMP4), vascular endothelial growth factor, and CHIR99021 (fig. S2A). Positive hits that enhanced cardiogenic gene expression, as assayed by qPCR, after CIM treatment were identified, and further tested in different combinations. Cells were then replated in human CM–conditioned medium, which may provide cardiac paracrine signals that mimic the in vivo environment for further maturation (14). After iterative rounds of testing, we found that a combination of total 15 compounds (15C) generated about 5 αMHC-GFP-positive clusters that beat spontaneously from 3x105 replated cells at day 30 (Fig. 1A), indicating reprogramming toward the cardiac fate.
To identify indispensable factors in 15C, we removed compounds one by one and treated cells with the remaining compounds. The number of beating clusters was significantly reduced by removal of CHIR99021, A83-01, BIX01294, AS8351, SC1, Y27632, or OAC2 (Fig. 1B). Together, these seven compounds (7C) were sufficient and necessary to efficiently induce cardiac reprogramming (Fig. 1C). We then screened a larger in-house generated library containing ~300 modulators of signaling pathways, including known inhibitors of kinases, phosphatases, and other signaling receptors with 7C as the baseline. Two inhibitors of the platelet-derived growth factor pathway—SU16F and JNJ10198409 (JNJ)—accelerated the down-regulation of fibroblast genes (fig. S3) and increased the yield of beating clusters (Fig. 1D). The two inhibitors were added to 7C (9C) for subsequent assays after optimization of dosage and treatment duration (table S2 and fig. S4). At day 30, 6.6 ± 0.4% of the 9C-treated cells expressed the CM marker cardiac troponin T (cTNT), with a yield of up to 1.2 cTNT+ CMs per input HFF. Removal of any of the nine compounds markedly reduced the induction efficiency (Fig. 1E and fig. S2B). Similarly, 9C reprogrammed human fetal lung fibroblasts (HLFs) into ciCMs with comparable efficiency (fig. S5).
Cardiogenesis involves sequential induction of mesoderm, cardiac progenitor cells (CPCs), and CMs (15). This developmental sequence, which was apparent in differentiation of human pluripotent stem cells (hPSCs) (16–18), was observed during 9C-induced conversion of HFFs into ciCMs. After exposure to CIM, up to 27.9 ± 4.0% of the 9C-treated cells started to express a key mesoderm marker, KDR (fig. S6). These cells initiated a cardiogenic program by sequentially expressing mesoderm, CPC, and CM genes (figs. S7 and S8), and finally became beating ciCMs (fig. S6). Expression of CPC genes, particularly markers of second but not first heart field progenitors, was further confirmed by qPCR and immunofluorescence analysis in 9C-treated cells (figs. S8A and S9). Small contracting clusters of ciCMs began to appear around day 20, continued to beat in culture (movies S1 and S2), and expressed specific CM markers (figs. S5 and S8B). Thus, they closely resembled CMs derived from hPSCs (hPSC-CMs) (fig. S1G).
Although new reprogramming protocols are improving cell yield, cardiac cells generated by genetic methods are heterogeneous with only about 0.1% of genetically reprogrammed human iCMs achieving a high degree of reprogramming as characterized to beat spontaneously, display cardiac action potentials (APs), and express multiple cardiac genes uniformly in vitro (10). In contrast, we found that more than 97% of the ciCMs spontaneously beat and uniformly expressed multiple cardiac structural proteins, similar to hPSC-CMs (fig. S10). The homogeneity of ciCMs was further confirmed by single-cell qPCR analyses, revealing little, if any, difference in the expression levels of cardiac genes between individual ciCMs (fig. S11). Collectively, these results confirm that ciCMs are highly reprogrammed and largely homogeneous.
Next, we further characterized ciCMs reprogrammed from HFFs (HFF-CMs) and HLFs (HLF-CMs). Immunofluorescence and transmission electron microscopy revealed that ciCMs exhibited well-organized sarcomere organization closely resembling hPSC-CMs (Fig. 2A and fig. S12). ciCMs also highly expressed genes involved in CM function, including atrial natriuretic factor, connexin 43, Cav3.2, HCN4, and Kir2.1 (Fig. 2A and fig. S13). Intracellular electrical recordings from ciCMs at reprogramming day 45–50 revealed robust APs that were synchronized 1:1 with rhythmic Ca2+ transients (Fig. 2B and fig. S14, A and B), similar to hPSC-CMs (19), suggesting ciCMs are highly reprogrammed and have a comparable maturity to hPSC-CMs. Most ciCMs exhibited ventricular-like APs and expressed ventricular but not atrial CM markers (Fig. 2B and fig. S14). Moreover, ciCMs responded to caffeine (a ryanodine receptor agonist), isoproterenol (a β-adrenergic agonist), and carbachol (a muscarinic agonist) (fig. S15). Thus, ciCMs are functional in vitro and possess electrophysiological features similar to hPSC-CMs.
We next compared the transcriptomes of ciCMs, their parental fibroblasts, hPSC-CMs, purified human fetal CMs (1), and primary human hearts (20) by microarray analyses. ciCMs, hPSC-CMs, and primary CM/hearts displayed transcriptional profiles that clearly differed from those of the fibroblasts (Fig. 3 and fig. S16, A and B). Genes most significantly up-regulated in ciCMs were related to CM function and heart development; genes for fibroblast function, such as cell proliferation and motility, were down-regulated in ciCMs (fig. S16, C and D, and fig. S17).
We examined the expression of several maturation-related genes that are distinctly expressed during cardiogenesis (20). ciCMs closely resembled hPSC-CMs in a hierarchical clustering analysis (fig. S18A) and both CM types had a similar expression pattern of α-smooth muscle actin and ventricular myosin light chain 2v, two maturation-related markers (21) (fig. S14, C and D, and fig. S18B), indicating ciCMs acquire maturity similar to hPSC-CMs. To determine whether ciCMs had an established CM-like chromatin state, we analyzed histone and DNA methylation status in the promoter regions of several fibroblast and cardiac genes and found that ciCMs gained key epigenetic features similar to hPSC-CMs (fig. S19). Furthermore, ciCMs were directly reprogrammed without going through a PSC-like state (fig. S20) and maintain genomic stability when compared to their parental fibroblasts (fig. S21).
Next, we investigated whether the diseased-heart niche would support the generation of ciCMs. HFFs harboring the αMHC-GFP reporter were treated with 9C for 6 days and then with CIM for 5 days and transplanted into the infarcted hearts of immunodeficient mice. HFFs not treated with 9C served as the negative control. Two weeks after transplantation, 9C-treated HFFs (identified by human-specific lamin A/C staining) robustly expressed CM markers, exhibited well-organized sarcomeres, and partially re-muscularized the infarcted area (Fig. 4). These results suggest that 9C-treated cells are compatible with the diseased-heart environment to further mature into CMs in vivo.
We hypothesized that 9C promotes an epigenetic state characterized by open chromatin, which renders cells responsive to extrinsic cardiogenic signals. Indeed, we observed a decrease in the number of heterochromatin foci (densely stained for H3K9me3 and HP1γ) in 9C-treated HFFs (Fig. 5A and fig. S22). Thus, 9C treatment appears to de-condense closed chromatin regions in HFFs, possibly creating a more euchromatic structure at loci important for cardiogenesis. To assess this mechanism further, we analyzed the genome-wide epigenetic changes by chromatin immunoprecipitation–sequencing (ChIP-seq) analysis of H3K4me3 and H3K27ac (active chromatin marks) and H3K27me3 (inactive chromatin mark) at reprogramming D6 and D11. We observed a dynamic loss of most H3K27me3 and specific gain of H3K4me3 among a subset of genomic loci during reprogramming (Fig. 5B). These genes were frequently related to developmental processes and cell differentiation (fig. S23). More specifically, we observed a significantly increased enrichment of H3K4me3 and H3K27ac on a cohort of heart developmental genes during reprogramming, whereas the deposition of H3K27me3 was downregulated (Fig. 5C and fig. S24). Consistently, 9C enabled the binding of β-catenin and Smad1 (effectors of major cardiogenic signal Wnt and BMP, respectively) to core promoter/enhancers of key genes important for heart development (Fig. 5, D and E). Thus, 9C appeared to convert the passive chromatin state of fibroblasts into a more euchromatic state, gaining chromatin accessibility on core cardiogenesis gene loci, and thereby facilitating cardiac reprogramming (Fig. 5F).
We next investigated the role of the reprogramming compound, AS8351. AS8351 and its functional analogs affect epigenetic modifications (22, 23) by competing with α-ketoglutarate (α-KG) for chelating iron/Fe(II) in certain epigenetic enzymes, such as the JmjC-domain-containing histone demethylases (JmjC-KDMs) that require α-KG and iron as co-factors (24). We hypothesized that AS8351’s effects on cardiac reprogramming might in part be mediated via modulation of a specific JmjC-KDM. To test this hypothesis, we abrogated each of the 22 genes in the JmjC-KDM family by small hairpin RNAs and found that only knockdown of KDM5B, or using a KDM5B inhibitor, PBIT could “phenocopy” AS8351 in generating ciCMs (fig. S25), suggesting that it might be a target of AS8351. KDM5B catalyzes the demethylation of tri-, di-, and mono-methylation states of H3K4 and facilitates heterochromatin formation (24). Consistent with the overall effect of 9C on re-opening the closed chromatin structure, inhibition of KDM5B may facilitate this process and sustain the active chromatin marks (i.e., H3K4 methylation) at specific genomic loci.
The present study shows that reprogrammed and functional lineage-specific cells can be generated from human fibroblasts by defined small molecules/growth factors. This study not only achieves significantly higher quality human iCMs than previously reported, but also provides a chemical approach free of foreign genetic material that may be adapted toward generating multiple cell types. ciCMs are functionally comparable to PSC-CMs, albeit generating fully-mature CMs remains a critical challenge for cardiac regenerative therapies. An important feature of ciCM induction is that, in response to signals mimicking the paracrine environment in the in vivo heart, 9C-treated cells are induced to become CMs. This finding may lay a foundation for ultimate in situ repair of the heart by targeting endogenous cardiac fibroblasts with small molecules. However, many critical challenges (e.g., reprogramming efficiency and tissue-specific delivery of multiple drugs in an efficient and controllable manner) need to be resolved before this strategy can be considered for in vivo therapeutic applications. Additional studies are needed to determine whether unintended genomic changes occur in ciCM subpopulations, as well as increase the maturity of ciCMs. Furthermore, we need a better understanding of the underlying mechanisms for this reprogramming.
Materials and Methods
Figs. S1 to S25
Tables S1 to S6
Movies S1 and S2