Main Text

Lower organisms, such as zebrafish, possesses a robust cardiac regeneration capacity through CM dedifferentiation characterized by reactivation of embryonic cardiogenic genes and disassembly of their sarcomeric structures (Kikuchi & Poss, 2010; Lepilina et al, 2006). In contrast, limited cardiac regeneration was only permitted from postnatal day 1 to day 6 (PN1 to PN6) in mice with sarcomeric disassembly and proliferation of CMs in the resected section of heart (Porrello et al, 2011), which is no longer observed beyond this narrow time window. Therefore, discovery of small molecule drugs enabling induction of regenerative capacity in mammalian adult endogenous CMs by recapitulating the regenerative cellular state presented in zebrafish will shed mechanistic lights on the understanding of roadblocks preventing CMs from dedifferentiation in mammalian hearts, and facilitate development of regenerative medicine to treat various heart diseases caused by CM loss.

During mammalian heart development, cardiomyocytes (CMs) mainly originate from first heart field (FHF) and second heart field (SHF) (Srivastava, 2006). In particular, SHF cells marked by the pioneer transcription factor ISL1 can proliferate and differentiate into three major cardiovascular cell types, CMs, SMCs and ECs during embryonic heart development (Bu et al, 2009; Moretti et al, 2006). Consistently, a small number of ISL1+ cells residing in the neonatal mouse heart between postnatal Day 1 and Day 5 were reported to expand and differentiate into CMs under appropriate in vitro conditions (Laugwitz, 2005). Since ISL1 transcriptionally governs the expression of a cohort of cardiac genes, e.g., NKX2-5 (Ma et al, 2008), FGF10 (Watanabe et al, 2012), essential for embryonic cardiogenesis, it is reasonable to hypothesize that induced expression of ISL1 in CM may be a determinative step in the induction of regenerative cardiac cells (RCCs).

Toward induction of regeneration capacity in CMs, we conducted a screening for small molecules that can induce the expression of ISL1 in CMs. Our combinatorial screening identified a novel combination of CHIR99021 and A-485/2C that could unprecedentedly and efficiently induce dedifferentiation of human CMs into RCCs. These RCCs exhibited disassembled sarcomeric structures, high expression of embryonic cardiogenic genes, and increased number of CMs through re-differentiation in vitro. Further studies showed that 2C could robustly generate RCCs in hearts of adult mice and improve cardiac function in the mice with myocardial infarction. Our proof-of-concept discovery demonstrated that a simple combination of small molecule drugs can grant regenerative capacity to the heart by reprogramming CMs into RCCs.

2C treatment efficiently induces dedifferentiation of hESC-derived CMs

To obtain an adequate number of CMs for small molecule screening, we differentiated hESCs into CMs (SI Appendix, Fig. S1A and Movie S1) following a well-established step-wise protocol (Lian et al, 2013). Consistent with previous reports, nearly homogenous contracting CMs were observed on day 10 of differentiation. After purification with glucose-depleted medium containing abundant lactate (SI Appendix, Fig. S1B), highly pure TNNT2+ CMs were obtained, and then subsequently dissociated and seeded into 96-well plates. When CMs resumed contraction, they were treated with individual small molecules from a collection of over 4,000 compounds for 3 days (SI Appendix, Fig. S1C and Table S1), and then fixed and immunostained with ISL1. Using a high-content imaging (HCI) and analysis system, five compounds were initially identified to potentially induce regenerative capacity in CMs, indicated by induced sarcomere disassembly and ISL1 expression (SI Appendix, Fig. S1D). After further comparing the effects from various combinations, the unique combination of CHIR99021 and A-485 (2C) was found to most efficiently induce ISL1 expression with healthy cell state (SI Appendix, Fig. S1E).

When CMs with well-organized sarcomeres were treated with 2C in vitro, cells started to exhibit dedifferentiation-associated phenotype, such as reduction of cell size after 24 hours and growing as clusters after 48 hours and forming colonies after 60 hours (Fig. 1A and SI Appendix, Fig. S2 A-C). During this time, the expression of TNNT2 and MYL2 were gradually downregulated while ISL1 expression and the percentage of ISL1+ cells were gradually increased (SI Appendix, Fig. S2 D-G). Compared to untreated CMs, significant sarcomere disassembly and reduction of cytoplasmic/nuclear area were observed in cells treated with 2C for 60 hours (Fig. 1 B-D). In parallel, the number and percentage of TNNT2+ cells remained constant while ISL1+ cells increased remarkably (Fig. 1 E-G), accompanied by a ∼3-fold increase ISL1 expression at both mRNA and protein levels (Fig. 1 H-J). Of note, other typical genes involved in early embryonic cardiogenesis, including MESP1, LEF1 (Klaus et al, 2007), FUT4 (Wang et al, 2019), and NR2F2 (Churko et al, 2018) were also induced by 2C to express at high levels (Fig. 1J and SI Appendix, Fig. S2H). In addition, treatment with 2C resulted in significantly decreased expression of CM-specific genes, such as TNNT2, MYL2, MYL7, and MEF2C, while the pan-cardiac transcription factors, GATA4, TBX5, and NKX2-5, showed no obvious expression changes (Fig. 1J and SI Appendix, Fig. S2H). Furthermore, when we generated mature CMs from hESCs by treating differentiating cells with ZLN005 following a previously reported method (Liu et al, 2020), 2C could induce ISL1 expression in these mature CMs (SI Appendix, Fig. S2I). These results suggested that CMs might be converted by 2C into a state with regenerative capacity.

2C treatment induced dedifferentiation of hESC-derived CMs toward ISL1-expressing cells.

Cells induced from hESC-derived CMs by treatment with DMSO (NC) or 2C for 60 hours. A, Phase contrast images showing cell morphology. B, Immunofluorescence staining of the ISL1 (ISL1, green), and the CM marker cardiac troponin T (TNNT2, red) in the cells. Nuclei were stained by DAPI (4′,6-diamidino-2-phenylindole) and presented in DNA blue. (C and D), Cytosolic (C) and nuclear (D) areas of the cells. E, The number of total cells, TNNT2+ cells, and ISL1+ cells. F, Cell percentage of TNNT2+ cells. G, Fraction of ISL1+ cell in TNNT2+ cells. Error bars represent SD. ns, not significant (P > 0.05); **P < 0.01; ***P < 0.001. (H and I) Western blot (H) and quantitative analysis (I) of ISL1 expression in DMSO (NC) or 2C-treated CMs for 60 hours. Error bars represent SD. *P < 0.05; J, Heatmap illustration showing the fold-changes of indicted marker genes’ expression by 2C treatment, which were measured using qRT-PCR.

2C induced cells possess regenerative capacity

To confirm the regenerative capacity of 2C-induced cardiac cells, we investigated cell proliferation using a BrdU incorporation assay. Immunostaining clearly showed the co-localization of ISL1 and BrdU labeling in cells (Fig. 2A), and subsequent statistical analysis revealed a significant decrease in the nuclear area and increase in the cell percentage of ISL1+/BrdU+ positive cells following 2C treatment (Fig. 2 B and C). We next examined the potential for re-differentiation of 2C-induced cardiac cells into CMs. As expected, upon withdrawal of 2C and subsequent culture in CM media for 3 days (60h+3d), we observed the emergence of spontaneously contracting cells exhibiting CM-specific morphology with typical size of cytoplasmic/nuclear area (Fig. 2 D-F and Movie S2), downregulated expression of early embryonic cardiogenesis genes and upregulated expression of CM-specific genes, and the expression of pan-cardiac genes didn’t show any noticeable changes throughout this process (SI Appendix, Fig. S3A). Remarkably, there was a 1.4-fold increase in the number of re-differentiated CMs with clear TNNT2 staining relative to the initial CMs before 2C treatment (Fig. 2 G and H), demonstrating a regenerative potential in 2C-induced cardiac cells. In addition to CMs, 2C-induced cardiac cells were capable of differentiation towards SMA+ SMCs or CD31+/VE-Cadherin+ ECs in the presence of PDGF-BB and TGF-β1 or VEGF, bFGF and BMP4, respectively (Fig. 2I and SI Appendix, Fig. S3 B-D). In contrast, the expression of those genes was not detectable in DMSO-treated cells. Collectively, our findings indicate that treatment with 2C facilitates the conversion of CMs into cardiac cells with regenerative capability, including proliferative potential and ability to differentiate into the three major cardiovascular cell types, which were therefore named as regenerative cardiac cells (RCCs).

Regenerative ability of 2C-induced RCCs.

(A-C), Immunofluorescence staining (A) and statistical analysis (B and C) of the ISL1 (green) and BrdU (red) double positive RCCs induced from CMs by treatment with DMSO (NC) or 2C for 60 hours. DAPI (4′,6-diamidino-2-phenylindole) staining labeled nuclei as blue. Error bars represent SD. **P < 0.01. D, Phase contrast images of hESC-derived CMs treated by DMSO (NC) or 2C for 60 hours (60h) and subsequently cultured in the absence of 2C for another 3 days (60h+3d). (E and F), Cytosolic (E) and nuclear (F) areas of the cells under the same condition in (D). Error bars indicate SD. ns, not significant (P > 0.05); ***P < 0.001. G, Immunostaining showed the expression of ISL1 (green) and TNNT2 (red) in the cells under the same condition in (D). H, The fold change of TNNT2+ cells under the same condition in (D). Error bars represent SD. ns, not significant (P > 0.05); **P < 0.01. Immunostaining showed the expression of EC markers (CD31, green and CD144, red), SMC marker (SMA, red), and CM marker (TNNT2, green).

Lineage tracing demonstrated that 2C induced RCCs dedifferentiated from TNNT2+ CMs

Although TNNT2+ CMs purified by lactate-based culture medium were almost homogeneous populations, we noticed that these were still a low percentage (<4%) of purified cells still expressed ISL1 (SI Appendix, Fig. S1B). Therefore, we performed lineage-tracing experiments to further confirm that the 2C-induced RCCs were actually dedifferentiated from TNNT2+ CMs rather than proliferation of those residual ISL1+ cells. Upon directed cardiac differentiation of an ISL1mCherry/+ H9 hESC (K9) line established through CRISPR-based knock-in (SI Appendix, Fig. S4 A-D), mCherry expression faithfully reflected endogenous ISL1 expression (Fig. 3 A-C). After purification of K9-derived CMs, mCherry-negative cells were selected by FACS (Fig. 3 D and E), all of which were confirmed to be TNNT2+ CMs (SI Appendix, Fig. S4E). Those mCherry-negative CMs were then treated with 2C for 60 hours. The expression of TNNT2 and MYL2 were dramatically down-regulated, whereas the expression of ISL1 and mCherry was significantly up-regulated (Fig. 3 F and K). We found the induced mCherry+/ISL1+ RCCs exhibited significant morphological changes accompanied by a dramatic decrease in TNNT2 expression (Fig. 3F). In contrast, the mCherry-negative CMs remained high expression of TNNT2 with the typical sarcomere structure (Fig. 3F). FACS analysis further confirmed that 2C was capable of K9-derived mCherry-negative CMs into mCherry+/ISL1+ cells (Fig. 3 G and H). Consistently, similar results were observed using HUES7 hESC line with ISL1-mCherry knock-in reporter (K7) (Fig. 3 I, J, L, and SI Appendix, Fig. S5).

Lineage tracing demonstrated 2C induced dedifferentiation of TNNT2+ CMs to ISL1-expressing RCCs.

A, Immunofluorescence images showing expression of endogenous ISL1 (green) and ISL1-mCherry (red) reporter in the cells differentiated from K9 hESC KI reporter line at day 6 (D6). (B and C) Flow cytometry analysis of the percentage of mCherry+/ISL1+ cells in the cells differentiated from K9 at D6. (D and E) Flow cytometry analysis of the percentage of mCherry-negative cells at selection day 4 (SD4) in lactate purification medium. Error bars represent SD. (F-H) Cells induced from mCherry-negative CMs by treatment with or without 2C for 60 hours. Images showing the expression of mCHERRY (red) and TNNT2 (green) in the cells (F) and flow cytometry analysis of the percentage of mCherry-positive cells (G and H). Error bars represent SD. (I and J) Flow cytometry analysis of the percentage of mCherry-positive cells dedifferentiated from K7 hESC KI reporter line derived mCherry-negative CMs. Error bars represent SD. (K and L), Relative gene expression of ISL1, mCHERRY, LEF1, TNNT2 and MYL2 in K9- and K7-derived mCherry-negative CMs treated with DMSO (NC) or 2C for 60 hours, respectively. Error bars represent SD. *P < 0.05; **P < 0.01; ***P < 0.001. (M and N) Flow cytometric plots showing EGFP-labeled CMs by lineage-tracing of K9-derived mCherry-negative CMs (M), and bar graph showing the percentage of mCherry-negative CMs expressing EGFP (N). Error bars represent SD. O, Images showing the expression of ISL1 (red) and EGFP (green) in the cells induced from EGFP-positive/mCherry-negative CMs in (M) by treatment with or without 2C for 60 hours.

Furthermore, we used a lineage-tracing system to unequivocally examine the conversion of CMs to ISL1-expressing RCCs by 2C treatment. In this assay, CMs were transfected with plasmids encoding CreERT2 under the control of TNNT2 promoter and an EGFP reporter following flox-stop-flox cassette (SI Appendix, Fig. S6A). After optimizing virus titer and infection time (SI Appendix, Fig. S6 B-D), ∼0.6 of K9-derived mCherry-negative CMs were permanently labeled with EGFP following 6 days of treatment with tamoxifen (Fig. 3 M and N). Then, those EGFP-labeled mCherry-negative CMs were sorted by flow cytometry and subsequently treated with 2C for 60 hours. As expected, the expression of ISL1 was observed in these EGFP+ CMs (Fig. 3O). Collectively, these results robustly demonstrated that RCCs with ISL1 expression were truly generated by 2C treatment from TNNT2+ CMs.

2C-induced dedifferentiation of CMs provided a protective effect on cardiac infarction

In order to examine 2C’s reprogramming effect on CMs in vivo, we first tested 2C on primary neonatal rat CMs (Sakurai et al, 2014), on which 2C was found to consistently induce RCCs generation with corresponding ISL1 induction and morphological changes (SI Appendix, Fig. S7A). Then, we sought to examine whether 2C was capable of reprograming endogenous CMs to RCCs in vivo. For this purpose, neonatal SD rats were intraperitoneally administrated with 20 mg/kg of CHIR99021 and 10 mg/kg of A-485 daily for 5 days (SI Appendix, Fig. S7B). Compared to the vehicle/DMSO controls (NC), ISL1 was robustly induced in the cells from hearts of rats treated with 2C without apparent effects on body weight (SI Appendix, Fig. S7 C and D). These induced ISL1-expressing cells were widely distributed within the region 600 µm to 3000 µm down from the base of the hearts, including aorta (Ao), left/right atria, and both ventricles (SI Appendix, Fig. S7E). Notably, all of these ISL1+ expressed TNNT2 as well, strongly indicating 2C-induced RCCs were actually originated from CMs in vivo. More importantly, neither administration of 20 mg/kg CHIR99021 nor 10 mg/kg A-485 alone observed ISL1+ cells in ventricles, powerfully illustrating the combined effect of 2C on the induction of RCCs (SI Appendix, Fig. S8).

Similarly, RCCs were also efficiently induced in the Ao root and RA regions of the heart dissected from adult 129SvJ mice administrated with 2C (20 mg/kg of CHIR99021 and 10 mg/kg of A-485) for 5 consecutive days as well (Fig. 4A). Intriguingly, we sought to explore whether 2C-induced RCCs could rescue cardiac function in mice subjected to myocardial infarction (MI). In a prophylactic setting, mice were pre-treated with either 2C or vehicle/DMSO for 5 days (Fig. 4B) and then inducing MI by ligation of left anterior descending (LAD) artery. When measuring cardiac function with magnetic resonance imaging (MRI) on days 1, 8, 25 and 35 post-MI, we found pretreatment with 2C was able to significantly improve both cardiac function and survival rate in the mice, without affecting their body weight (Fig. 4 C-E). Next, when we treated mice with 2C or vehicle/DMSO post-MI in a therapeutic setting (Fig. 4F), remarkably mice with 2C treatment displayed a recovered cardiac function as assessed by both left ventricular ejection fraction (LVEF) and echocardiography (Fig. 4 G and H). Consistently, cardiac fibrosis was largely ameliorated in 2C-treated mice that displayed significantly smaller scar size in the hearts compared to control mice (Fig. 4I). Those in vivo studies collectively indicate that 2C or further improved small molecule drugs with similar mechanisms may have therapeutic utility to repair or regenerate hearts after cardiac injury. Altogether, these in vivo studies indicate that 2C-induced RCCs possess the same regenerative capacities comparable to those generated in vitro, which will shed light on development of small molecule drugs with therapeutic utility to repair or regenerate hearts after cardiac injury.

Heart regeneration via 2C-induced dedifferentiation of CMs.

A, Immunofluorescence staining of ISL1 (green) and TNNT2 (red) in cross-sectioned hearts from 2C or vehicle-treated (NC) adult 129SvJ mice. Ao, aorta. PA, pulmonary artery. LA, left atrial. RA, right atrial. IAS, interatrial septum. B, Schematic illustration of the method used to examine the prophylactic effect of 2C in 129SvJ mice post MI. C, Body weight of mice pre-treated with vehicle (DMSO) or 2C as shown in (B) at Day 12, Day 16, Day20, and Day35 after MI (1dpi). Error bars represent SD. ns, not significant (P > 0.05). D, Survival curve of sham-operated mice and mice pre-treated with vehicle (DMSO) or 2C as shown in (B), at indicated time points before or after MI. E, Ejection fraction (EF) of sham-operated mice and mice pre-treated with vehicle (DMSO) or 2C as shown in (B), before MI (baseline) or at Day 1, Day8, Day25, and Day35 after MI (1dpi). Error bars represent SD. ns, not significant (P > 0.05); **P < 0.01; ***P < 0.001. F, Schematic illustration of the method used to examine therapeutic effect of 2C in the 129SvJ mice post MI. G, Echocardiography of sham-operated mice and mice treated with vehicle (DMSO) or 2C as shown in (F) at Day 35 post MI. H, Serial fMRI measurements showing the cardiac function from sham-operated mice and mice treated with vehicle (DMSO) or 2C at as shown in (F), at Day 35 post MI. Error bars represent SD. *P < 0.05. I, Masson staining of serial transverse sections of hearts from sham-operated mice and mice treated with vehicle (DMSO) or 2C as shown in (F), at Day 35 post MI.

Both CHIR99021 and A-485 are required to induce dedifferentiation of CMs to RCCs

To gain insights into mechanisms involved in 2C-induced dedifferentiation of CMs to RCCs, we performed bulk RNA-seq analysis of K9-derived ISL1/mCherry-negative CMs treated with DMSO (NC) or 2C for 60 hours (Fig. 5A). Analysis of differentially expressed genes (DEGs) revealed a remarkable change in the transcriptome upon 2C treatment (Fig. 5B). In detail, 2C treatment led to increased expression of a large set of genes known to be associated with embryonic cardiogenic genes (e.g., MSX1, BMP4, TCF4, and LEF1) and suppression of a significant list of genes associated with CMs (Fig. 5C). Consistently, gene ontology (GO) analysis further revealed that down-regulated genes were mainly involved in cardiac maturation and muscle contraction (Fig. 5D), while up-regulated genes were enriched in cell junction and catenin complex (Fig. 5 E and F). Those embryonic cardiogenic genes and other genes significantly up-regulated (e.g., MSX2, NKD1, PDGFC, CTNNA2) after 2C treatment were confirmed by qPCR in K9-derived mCherry-negative CMs (Fig. 5G).

Bulk RNA-seq of analysis of 2C-treated ISL1/mCherry-negative CMs.

A, Scheme of bulk RNA-seq analysis of K9-derived mCherry-negative CMs with DMSO (NC) or 2C treatment for 60 hours. B, Heatmap of differentially expressed genes (DEGs) in ISL1/mCherry-negative CMs treated with DMSO (NC) or 2C for 60 hours. C, Volcano plot showing genes significantly changed by 60 hours of 2C treatment. (D and E) Gene Ontology (GO) analysis of downregulated (D) and upregulated (E) genes in ISL1/mCherry-negative CMs by 2C treatment for 60 hours, compared to DMSO (NC) treated cells. F, Plotting GO terms of upregulated genes by 2C treatment with cnetlpot. G, Relative expression fold-changes of indicated genes in K9-derived ISL1/mCherry-negative CMs by 60 hours of DMSO (NC) or 2C treatment. Error bars indicate SD. **P < 0.01; ***P < 0.001.

To investigate the individual effects of CHIR99021 and A-485 on induction of RCCs, we compared H9 hESC-derived CMs treated with CHIR99021 and A-485 individually or in combination. When measuring cytoplasmic/nuclear area, we found CHIR99021 significantly reduced both cytoplasmic and nuclear area, a distinguishable morphological feature of RCCs, in comparison to A-485 that only reduced cytoplasmic area (SI Appendix, Fig. S9 A-C). Consistently, disassembly of sarcomere structures was more frequently observed in CMs treated with CHIR99021 than in CMs treated with A-485 (SI Appendix, Fig. S9E). Although reduced expression of TNNT2 and MYL2 as well as increased expression of ISL1 can be induced individually by CHIR99021 and A-458 (SI Appendix, Fig. S9 D and E), CHIR99021 was able to uniquely promote the expression of a cohort of genes involved in embryonic cardiogenesis, including BMP4, NKD1, MSX2, PDGFC, LEF1 and TCF4 (SI Appendix, Fig. S9D). Upon withdrawal of CHIR99021, A-485 or 2C, the cell percentage of ISL1+ cell was similarly and significantly reduced (SI Appendix, Fig. S9F). Of note, a drastic increase in the number of TNNT2+ CMs was only observed in redifferentiation of 2C-treated cells, as confirmed by statistical analysis (SI Appendix, Fig. S9G). These findings suggested that CHIR99021 played a leading role in 2C-induced de-differentiation of CMs to RCCs, A-485 is also indispensable.

Reprogramming of CMs to RCCs by 2C went through an intermediate cell state

Next, we performed scRNA-seq of K9-derived mCherry-negative CMs treated with DMSO (NC) or 2C for 60 hours. Utilizing UMAP analysis, we observed the proportion of cells within three (cluster 0, 2, 3) out of 7 identified clusters increased after 2C treatment (Fig. 6 A and B). Among these three clusters, cells in cluster 2 highly expressed RCC characteristic genes, including those essential for embryonic cardiogenesis (e.g., ISL1, BMP4 and FGF20 (a key gene involved in expansion of early embryonic progenitor cells (Cohen et al, 2007)), and cell proliferation marker gene MKI67 (Fig. 6 C and D). By contrast, cells in cluster 0, 3, 4, and 5 expressed CMs specific markers (e.g., MYH6 and MYL2) at high levels (Fig. 6 C and D), indicating they were CMs. Meanwhile, cells in cluster 1 and 6 were considered as intermediate cells (ICs) during 2C-induced reprogramming of CMs to RCCs, because of expression of ACTA2, a marker gene reported to express in CM dedifferentiation, and genes associated with CM development, including COL1A1, COL1A2, and COL3A1 (Cui et al, 2020; Mononen et al, 2020) (Fig. 6 C and D).

Single-cell RNA-seq of 2C-treated mCherry-negative CMs.

A, UMAP analysis showing 7 clusters in cells induced from K9-derived mCherry-negative CMs by treatment with DMSO (NC) and 2C for 60 hours. B, The percentage of cells in the 7 indicated clusters, following DMSO (NC) or 2C treatment. C, Heatmap showing the differentially expressed genes in the cells from 7 indicated clusters. The representative marker genes of 7 indicated clusters were listed on the right. D, Violin plots showing the expression levels of marker genes of CMs (MYL2, MYH6), ICs (COL1A1, ACTA2), and RCCs (ISL1, BMP4, FGF20) among cells from 7 indicated clusters. E, UMAP analysis showing the second-level clustering of cluster 2 into 4 subclusters (left), which exhibited dramatic distinction under condition of 2C or NC (right). F, The percentage of cells in the 4 indicated subclusters within cluster 2, following DMSO (NC) or 2C treatment. G, Heatmap showing the differentially expressed genes among cells from 4 subclusters of cluster 2. Genes related to RCCs are highlighted in the green blocks on the right. (H and I) Pseudotime trajectory showing changes across various cell states upon 2C treatment, which were presented with different developmental pseudotime points (H) and cell states (I). J, Curves showing the dynamic expression of representative genes of RCCs (ISL1, FGF20), ICs (COL1A1), and CMs (TNNT2) along indicated pseudotime points.

When further examining cluster 2 with 435 cells from DMSO-treated cells and 410 cells from 2C-treated cells by second-level clustering analysis, we identified subcluster 0 (408, or 48.3% out of 845 cells) in cluster 2 which was exclusively observed in 2C-treated cells (Fig. 6 E and F). Furthermore, cells in subcluster 0 expressed a series of RCC-related genes essential for embryonic cardiogenesis, including MSX1, LEF1, BMP4, MSX2, and HAND1 (Fig. 6G), as well as a series of genes co-expressed with ISL1 in SHF progenitors in vivo, such as NR2F2, TBX5, ALCAM (Ghazizadeh et al, 2018), and CXCR4 (Andersen et al, 2018) (Fig. 6G), strongly indicating cells in subcluster 0 represent RCCs induced by 2C treatment. Thus, a unique gene set that included LIX1, NKD1, PDGFC, ARL4A, AXIN2, FZD7, ISL1, MSX1, MSX2, BMP4, LEF1, and HAND1 was found to determine RCC state during the reprogramming of CMs by 2C, and hence designated as RCC genes. In addition, the pseudo-time analysis that clearly revealed 2C-induced reprogramming of CMs (highly expressing TNNT2) went through ICs (highly expressing COL1A1), and ultimately towards RCCs (highly expressing ISL1 and FGF20) (Fig. 6 H-I).

RCC and CM genes were epigenetically regulated by 2C during reprogramming of CMs into RCCs

As a p300 acetyltransferase inhibitor, A-485 was previously reported to specifically inhibit acetylation of H3 at lysine 27 (H3K27Ac) but not at lysine 9 (H3K9Ac) in the PC3 cells (Lasko et al, 2017). Consistently, the percentage of H3K27Ac+ cells but not that of H3K9Ac+ cells were specifically and significantly reduced in CMs were treated with A-485 alone (SI Appendix, Fig. S10). In contrast, neither CHIR99021 nor 2C treatment resulted in noticeable alterations in the percentage of H3K27Ac+ and H3K9Ac+ cells (SI Appendix, Fig. S10). Intriguingly, further analyses revealed a remarkable reduction in the levels of H3K27Ac and H3K9Ac in TNNT2+ cells following A-485 treatment (SI Appendix, Fig. S10 F and L), suggesting that A-485 simultaneously inhibit both H3K27Ac and H3K9Ac on CM-specific genes during the dedifferentiation of CMs.

To gain more insights on the mechanism underlying 2C-induced silencing of CM genes and activation of RCC genes during reprogramming of CMs to RCCs, we performed chromatin immunoprecipitation-sequencing (ChIP-seq) analyses of H3K27Ac and H3K9Ac in DMSO, A-485, CHIR99021 and 2C-treated H9 hESC-derived CMs (Fig. 7A). We identified active promoters as proximal regions (±4 kb from the transcription start sites, or TSS) exhibiting H3K4me3 enrichment in the cells treated with DMSO or 2C (Fig. 7 B, and C). Treatment with 2C led to a striking increase in H3K9Ac and H3K27Ac enrichment around TSS regions on a whole-genome-wide scale (Fig. 7 B and C). Consistently, 2C treatment induced a significantly enhanced enrichment of H3K9Ac in 4,485 genes and H3K27Ac enrichment in 2,560 genes, whereas a decreased of H3K9Ac and H3K27Ac enrichment was detected in 346 and 1,834 genes respectively (SI Appendix, Fig. S11 A-F). Of note, a significant part of genes decreased H3K9Ac (268, or 58.1% out of 461) and H3K27Ac (777, or 68.6% out of 1132) enrichment in A485-treated cells were also observed among genes that H3K9Ac (268, or 77.5% out of 346) and H3K27Ac (777, or 42.4% out of 1834) enrichment following 2C treatment. Further GO analysis revealed that those genes decreased H3K9Ac and H3K27Ac enrichment in their promoters commonly seen in repose to A-485 or 2C treatment, preferentially exhibited cardiac specific functions, such as cardiac muscle contraction, sarcomere organization, and heart contraction (SI Appendix, Fig. S11 G-L). Compared to individual treatment with A-485 or CHIR99021, 2C treatment resulted in an increased H3K9Ac and H3K27Ac enrichment in the genes associated with cell cycle and cell division (SI Appendix, Fig. S11 K and L), strongly supporting the combined effects of 2C on RCC cell state establishment.

Chromatin immunoprecipitation-sequencing (ChIP-seq) analyses of chemical-treated H9 hESC-derived CMs.

A, Schematic illustration of ChIP-seq analysis of H9-derived CMs subjected to DMSO, A-485, CHIR99021, or 2C treatment for 60 hours. B, Average ChIP-seq signal profiles showing the indicated histone modifications around the TSS in the input and ChIP samples prepared from DMSO and 2C-treated cells. C, Heatmap showing the whole-genome wide distribution of H3K9Ac, H3K27Ac, and H3K4me3 peaks within a range of ±4kb from TSSs in the cells treated with DMSO or 2C for 60 hours. D, H3K9Ac and H3K27Ac peaks surrounding CM genes (TNNT2 and TNNI1) and RCC genes (LEF1 and ISL1) in the cells treated with DMSO, A-485, CHIR99021 or 2C for 60 hours, and H3K4me3 peaks surrounding the same genes in the cells treated by DMSO or 2C. E, Veen diagram showing the number of annotated genes with H3K9Ac or H3K27Ac enrichment in the cells treated with DMSO, A-485, CHIR99021 or 2C for 60 hours. Red circles indicate the number of genes with unique H3K9Ac or H3K27Ac enrichment induced by 2C treatment; black circles indicate the number of genes with H3K9Ac or H3K27Ac enrichment unaffected by any chemical treatment. (G-J) The annotated genes with the most significant changes in H3K9Ac enrichment following treatment with DMSO (G) or 2C (I) were ranked by Log10LR and analyzed by GOs (H and J), respectively. K, ISL1 binding motifs identified from the cells treated with A-485 or 2C.

Next, we specifically examined H3K9Ac and H3K27Ac peaks in the promoters of critical CM genes and RCC genes in the cells treated with or without A-485, CHIR99021 or 2C. We observed that both A-485 and CHIR99021 effectively reduced the peaks of H3K9Ac and H3K27Ac in the promoters of CM genes, including TNNT2, TNNI1, MYL7, MYH6, and MYH7. These reductions were further intensified with 2C treatment (Fig. 7D and SI Appendix, Fig. S12A). Interestingly, increased peaks of H3K9Ac and H3K27Ac in promoters of the RCC genes, such as LEF1, AXIN2, BMP4, LIX1, MSX1, MSX2, and NKD1, was only observed in condition with CHIR99021 available (Fig. 7D and SI Appendix, Fig. S12B). Other RCC genes, including ISL1, ARL4A, FZD7, HAND1, and PDGFC, exhibited a significantly elevated peaks of H3K9Ac and H3K27Ac by treatment with 2C and much higher than those treated with CHIR99021 alone (Fig. 7D and SI Appendix, Fig. S12C). Therefore, while CHIR99021 was able to activate the expression of RCC genes and simultaneously repressed CM genes partially through changing H3K27Ac and H3K9Ac during 2C-induced CM dedifferentiation to RCC, A-485 can inhibit CM genes directly through reducing of H3K27Ac and H3K9Ac modifications in the promoters of CM-specific genes, thereby assisting CHIR99021 to establish RCC state.

In addition, we observed that different small molecules had varied impacts on distribution of H3K9Ac and H3K27Ac enrichment during 2C-induced dedifferentiation of CM into RCC. As indicated by Veen diagram, distributions of H3K9Ac and H3K27Ac enrichment significantly diverged among cells treated with DMSO, A-485, CHIR99021, and 2C (Fig. 7E and F). For genes with H3K9Ac enrichment, 2716 (or 46.1%) out of 5891 annotated genes were only enriched in the cells treated with 2C, while only 527 (or 4.9%) out of 10785 annotated genes with H3K27Ac enrichment was exclusively observed (Fig. 7E and F). Thus, the process of 2C-induced dedifferentiation of CM to RCC is accompanied by vigorous alterations in H3K9Ac, whereas H3K27Ac is remain relatively stable, indicating that the genome-wide distribution of H3K9Ac exhibited drastic distinction between CMs and RCCs. To further confirm this notion, we performed a comparative analysis of annotated genes between the cells treat with 2C or DMSO to validate the most significant differential genes (i.e., Log10LR>3) under the respective condition. In comparison to 430 genes with H3K9Ac enrichment observed in DMSO-treated cells, which were mainly involved in cardiac muscle contraction (Fig. 7G and H), 2C treatment markedly enhanced the enrichment of H3K9Ac in 123 genes associated with the transcription factor complex and Wnt signaling pathway (Fig. 7 I and J). In particular, 12 CM genes, including TNNT2, were ranked in top 90 out of 430 annotated genes with H3K9Ac enrichment. In contrast, 7 RCC genes, including ISL1, were ranked in the top 50 out of 123 annotated genes with H3K9Ac enrichment. Moreover, DNA-binding motif analysis indicated that the ISL1 binding motifs only exhibited accessibility in the cells treated with A-485 or 2C (Fig. 7K), underscoring the vital role of A-485 in 2C treatment to activate the expression of RCC genes. Taken together, these findings revealed that CHIR99021 is responsive for transcriptional and epigenetic activation of genes essential for cardiac embryonic development, while A-485 promotes conversion of CMs to RCCs by epigenetic downregulation of H3K27Ac and particularly H3K9Ac in CMs, which together facilitate the induction of RCCs from CMs.


It is worth noting that the RCCs were only generated in hearts of 2C-treated postnatal rats, but not in rats after treatment with only 20 mg/kg CHIR99021 or 10 mg/kg A-485 alone (SI Appendix, Fig. S8), indicating the reprogramming of endogenous CMs to RCCs depends on the cooperative effect from these two molecules. CHIR99021 has been well characterized as a GSK3 inhibitor to activate the Wnt signaling pathway. Previous studies have elucidated the role of CHIR99021 in initiating proliferation of hPSC-derived CMs (Buikema et al, 2020; Quaife-Ryan et al, 2020), accompanied by induction of H3K27Ac on promoter regions of multiple cardiac lineage genes, including TNNT2, MYH6, and ISL1 (Quaife-Ryan et al., 2020). However, this proliferation was only achieved in neonatal CMs in vivo (Buikema et al., 2020; Quaife-Ryan et al., 2020). Consistently, we also found the enhanced levels of H3K27Ac in TNNT2+ CMs by treatment with CHIR99021 alone or 2C (Fig. 7D and SI Appendix, Fig. S10F). In addition, we found that the p300/CBP inhibitor A-485 facilitate the reprogramming of CMs into RCCs by epigenetically inhibiting expression of CM-specific genes such as TNNT2 (Fig. 7D and SI Appendix, Fig. S9D), while enhancing the modifications of H3K9Ac and H3K27Ac on RCC-specific genes such as ISL1 in combination with CHIR99021 (Fig. 7 D and K). These findings promoted us to postulate that A-485 and its mediated mechanism may be used to modulate the cell fate of other terminally differentiated cells by epigenetically suppressing the expression of corresponding lineages specific genes.

2C-induced RCCs exhibited potential to differentiate into three cardiac-lineages of the heart in vitro, which might contribute additional beneficial effects in heart repair and regeneration. Furthermore, RCCs expressed a broad set of genes essential for embryonic cardiogenesis (Fig. 1J and Fig. 5G), including ISL1, and a series of potentially important genes identified in this study (Fig. 6G), all of which were highly expressed in ISL1+ SHF cardiac progenitors during mammalian heart development. However, RCCs had a restricted proliferation capacity, although they were able to be passaged 3-5 times under 2C conditions (data not shown), which is distinct to ISL1+ cardiac progenitors (Bu et al., 2009; Moretti et al., 2006). These findings collectively indicated that 2C-induced RCCs dedifferentiated from CMs may be expandable under an appropriate condition and deserve further investigation.


This work is supported by the National Natural Science Foundation of China (32030031 to S.D.), Beijing Natural Science Foundation (JQ22016 to T.M.), the National Key R&D Program of China (2022YFA1103704 to S.D.; 2022YFA1104503 to Y. N.), Center for Life Sciences (to S.D.). We thank the Center for Pharmaceutical Technology, Tsinghua University for the activity screening platform, Biomarker Technologies Corporation, Beijing, China and BeiJing Geek Gene Technology Co Ltd for technical support, and support from Tsinghua-Peking Center for Life Sciences.

Author Contributions

Conceptualization, S.D. and W.Z.; Methodology, W.Z., H.L., T.H.M. and Y.N.; Investigation, W.Z., C.Y.W., K.Z.H., P.Q.W., B.W.W., and H.G.; Visualization, W.Z., P.Q.W., B.W.W., H.G., and D.W.; Funding Acquisition, S.D., T.H.M., and Y.N.; Project Administration, W.Z., H.L., T.H.M., and Y.N.; Supervision: S.D., and W.Z.; Writing – Original Draft, W.Z., T.H.M., and S.D.; Writing – Review & Editing: W.Z., T.H.M., Y.N., S.D.

Competing Interests

The authors declare no competing interests.

Data and Materials availability

All targeted amplicon sequencing data have been deposited in the Sequence Read Archive of the NCBI under the BioProject accession number PRJNA903530. All data are available in the main text or the supplementary materials.