High-efficiency reprogramming of fibroblasts into cardiomyocytes requires suppression of pro-fibrotic signalling

Direct reprogramming of fibroblasts into cardiomyocytes by forced expression of cardiomyogenic factors, GMT (GATA4, Mef2C, Tbx5) or GHMT (GATA4, Hand2, Mef2C, Tbx5), has recently been demonstrated, suggesting a novel therapeutic strategy for cardiac repair. However, current approaches are inefficient. Here we demonstrate that pro-fibrotic signalling potently antagonizes cardiac reprogramming. Remarkably, inhibition of pro-fibrotic signalling using small molecules that target the transforming growth factor-β or Rho-associated kinase pathways converts embryonic fibroblasts into functional cardiomyocyte-like cells, with the efficiency up to 60%. Conversely, overactivation of these pro-fibrotic signalling networks attenuates cardiac reprogramming. Furthermore, inhibition of pro-fibrotic signalling dramatically enhances the kinetics of cardiac reprogramming, with spontaneously contracting cardiomyocytes emerging in less than 2 weeks, as opposed to 4 weeks with GHMT alone. These findings provide new insights into the molecular mechanisms underlying cardiac conversion of fibroblasts and would enhance efforts to generate cardiomyocytes for clinical applications.

Intense investigation has focused on enhancing cardiomyogenic reprogramming, since GMT-mediated reprogramming is slow and inefficient 18 . For example, GMT plus a microRNA, miR-133a, was shown to induce B16 beating cells (B4 beating cells cm À 2 ) from mouse embryonic fibroblasts (MEFs) in a well of a 12-well plate after 30 days 19 . Stoichiometry expression of G, M and T protein affects the efficiency of GMT-mediated cardiac reprogramming 20 . Fusion of the MyoD transactivation domain to MEF2C with GHT (MM 3 -GHT) enhanced conversion efficiency 21 . When MEFs transduced with MM 3 -GHT were treated with GSK126, an inhibitor of EZH2 methyltransferase or UNC0638, a potent inhibitor of G9a histone methyltransferase, B20% of MEFs spontaneously contracted on day 14 post treatment 22 . In another study, Nkx2-5 was added to GHMT, and MEFs were subsequently treated with SB431542, an inhibitor of the TGF-b type I receptor ALK5 (Tgfbr1), and after 14 days of treatment, B16% of these cells displayed calcium transients 23 . Thus, despite extensive efforts, the efficiency of direct reprogramming of embryonic fibroblasts (MEFs) into cardiomyocytes is yet to exceed 20%. More disappointingly, the efficiency of direct reprogramming of adult fibroblasts, including adult CFs (ACFs) and tail-tip fibroblasts (ATTFs), into beating cardiomyocytes is less than 0.1% (refs 13,19-23). The inefficiency of direct reprogramming of fibroblasts, especially adult fibroblasts, into functional cardiomyocytes has led many in the field to question the clinical translatability of this method.
Here we demonstrate that pro-fibrotic events are concomitantly activated during GHMT-mediated cardiac reprogramming, leading to inhibiting conversion of fibroblasts into beating cardiomyocytes. Suppression of pro-fibrotic signalling using small molecules that target pro-fibrotic networks provides a remarkably efficient and rapid approach to reprogramming. These findings would facilitate efforts to employ therapeutic cardiac reprogramming and also provide significant and novel molecular insights into the mechanisms underlying reprogramming of fibroblasts into functional cardiomyocytes.

Results
Dynamics of pro-fibrotic events during reprogramming. After expression of GHMT, the fraction of cells positive for cardiac troponin T (cTnT), a cardiomyocyte marker, increased with time and reached B37% on day 6 ( Supplementary Fig. 1a,b). After 1 month, we observed B200 (n ¼ 6) beating cells cm À 2 , which is less than 5% of total cells in dishes (Supplementary Movie 1). GHMT retroviruses infected more than 80% of MEFs ( Supplementary Fig. 1c), but less than 5% of infected cells were reprogrammed into beating cardiomyocytes, suggesting endogenous barriers to cardiac reprogramming.
To search for endogenous barriers, we performed RNA sequencing (RNA-Seq) to define genes regulated by GHMT overexpression on day 7. We performed gene ontology (GO) enrichment analysis on genes that were upregulated in GHMTinfected cultures. Cardiomyocyte GO terms were enriched among genes upregulated in GHMT-infected cultures (Table 1). Surprisingly, pro-fibrotic/ECM ontology terms 24 were also significantly enriched among genes upregulated in GHMT-infected cultures on day 7 (Table 1). To confirm RNA-Seq data, we examined fibrotic markers in GHMT-infected MEFs using quantitative RT-PCR (qPCR) or western blot. Expression of ECM genes such as Fn-EDA and Collagen I (Col1a1) was upregulated after GHMT infection, but downregulated 12 days post infection (Fig. 1a). Immunoblotting also revealed dynamic changes of aSMA expression (Fig. 1b,c and Supplementary Fig. 2). During the first week of induction, pro-fibrotic gene expression in GHMTinfected cultures was higher than that in GFP-infected cultures ( Supplementary Fig. 3a). However, by day 14, expression of fibrotic genes in GHMT-infected cultures was B50% of that in GFP-infected cultures ( Supplementary Fig. 3a). These data suggest that genes involved in fibrotic events are upregulated in GHMT-infected cells during the first week post infection, and expression of these genes wanes after 12 days post-GHMT infection.
TGF-b signalling is an important pathway controlling fibrotic events [5][6][7] . TGF-b family cytokines, such as TGF-b1/2, bind to type I and type II TGF-b receptors (Tgfbr1 and Tgfbr2), and activate target gene expression by phosphorylating Smad transcription factors, Smad2 and Smad3 (ref. 25). To address whether increasing ECM expression during the early stages of reprogramming correlates with the activation of TGF-b signalling, we analysed the active form of Smad2 with a phospho-specific antibody. Compared with GFP-infected fibroblasts, higher levels of phospho-Smad2 were detected in GHMT-infected cultures on day 5 and day 7 ( Fig. 1b and Supplementary Fig. 2). However, phospho-Smad2 and aSMA in GHMT-MEFs were downregulated on day 14, compared with days 5 and 7 ( Fig. 1c and Supplementary Fig. 2). In addition, expression of TGF-b signalling components, Tgfb2 and Tgfbr1, was significantly upregulated in GHMT cultures on days, 3, 5 and 7 ( Supplementary Fig. 3b). In addition to profibrotic genes, TGF-b signalling regulates downstream targets such as epithelial-to-mesenchymal transition markers, Snail and Slug transcription factors 26 . Overexpression of GHMT downregulated Snail and Slug by day 7, at both the protein and mRNA levels (Fig. 1b, Supplementary Figs 2 and 3c). Taken together, these data suggest that TGF-b signalling and ECM expression are activated in GHMT-infected fibroblasts during the early stages, but are downregulated during later stages of reprogramming. We therefore hypothesized that activation of ECM expression by pro-fibrotic signalling in GHMT-infected fibroblasts suppresses conversion of the cells into cardiomyocytes.
H3K4me2 marks the cluster of miR-1-2 and miR-133a-1. In addition to searching for potential barriers to GHMT-mediated reprogramming, we attempted to look for enhancers of the process. Enhancement of reprogramming of fibroblasts into iPS cells has been facilitated by genome-wide studies 27,28 . To gain insights into global genome changes, we performed chromatin immunoprecipitation followed by deep sequencing (ChIP-Seq) to mark H3K4 dimethylation (H3K4me2) in GHMT-infected MEFs. H3K4me2 marks both promoter and enhancer regions of active genes 29 .
H3K4me2 peaks shifted from a fibroblast towards a cardiomyocyte state post expression of GHMT ( Supplementary Fig. 4a). On day 7, 47% of a total of 16,296 H3K4me2 peaks identified in GHMT-expressing cultures were shared with peaks from primary cardiomyocytes. These shifts indicate cardiomyocyte identity caused by GHMT and suggest that expression of GHMT at early stages promotes chromatin changes required for cardiac reprogramming.
miR-1 and miR-133 enhance cardiac reprogramming. To determine whether miR-1 and miR-133 are able to enhance reprogramming, we examined expression of cTnT using flow cytometry 7 days post infection with retroviruses carrying GHMT and/or the microRNAs. GHMT induced expression of cTnT in B54% of MEFs (Fig. 2a,b). We divided cTnT þ cells into two classes, cTnT low and cTnT high . When miR-1 and miR-133 (2m) were delivered alone, less than 1% of cells became positive for cTnT. However, adding 2 m into GHMT (GHMT2m) significantly increased the cTnT high fraction from B32 to B42% ( Fig. 2a,b). Next, we examined effects of 2 m on the assembly of sarcomeres. Immunocytochemistry was performed with antibodies against cTnT and another sarcomeric protein, a-actinin. Two weeks post infection, GHMT produced cells with strong immunostaining of a-actinin and cTnT (Fig. 2c). More positive cells were observed in GHMT2m-infected cultures (Fig. 2c), and expression of cardiomyocyte genes, such as Actc1, Myh6 and Ryr2, was also enhanced by 2 m (Supplementary Fig. 5a). These data suggest that miR-1 and miR-133 promote expression of cardiac genes and assembly of sarcomeres.
We next assessed the relative importance of miR-1 and miR-133 for enhancing GHMT-mediated reprogramming. Depletion of either miR-1 or miR-133 from GHMT2m did not significantly reduce the percentage of cTnT high cells in cultures ( Supplementary Fig. 5b,c), suggesting that either miR-1 or miR-133 is sufficient to enhance expression of cTnT. Next, we examined spontaneously beating cells. GHMT produced iCMs that started to beat by 3 weeks. GHMT plus miR-1 (GHMTm1) or GHMT plus miR-133 (GHMTm133) produced iCMs that started to beat by 2 weeks and GHMT2m produced iCMs that began beating by 8 days (Fig. 2d and Supplementary Movie 2). More beating iCMs were observed in GHMT2m-infected cultures than with the other conditions (Fig. 2d). By 4 weeks, GHMT, GHMTm1, GHMTm133 and GHMT2m produced B200, B1,100, B700 and B1,400 beating iCMs per cm 2 , respectively ( Fig. 2d and Supplementary Movie 3). On day 25, GHMT2m induced B12% of cultures to spontaneously contract, which is more than beating fractions induced by GHMT, GHMTm1 or GHMTm133 (Fig. 2e). These data suggest that both miR-1 and miR-133 are required for generating maximal beating iCMs in GHMT-infected cultures.
To examine effects of miR-1 and miR-133 on global gene expression, we performed RNA-Seq in GHMT-and GHMT2minfected MEFs on day 7. Cardiac development ontologies were  Table 1). These data suggest that miR-1 and miR-133 globally enhance cardiac gene expression and inhibit pro-fibrotic gene expression.
Negative effect of pro-fibrotic signalling on reprogramming.
To test the hypothesis that pro-fibrotic signalling serves as a barrier to cardiac reprogramming, we overactivated pro-fibrotic signalling in reprogramming cells. Treatment of fibroblasts with TGF-b1 activates pro-fibrotic signalling cascades [5][6][7] . TGF-b1 treatment increased phosphorylation of Smad2, expression of Fn-EDA and Col1a1 and formation of aSMA stress fibres in GHMT-or GHMT2m-infected cultures on day 7, compared with non-treated cultures (Fig. 3a-c and Supplementary Fig. 2). These data confirm that TGF-b1 treatment stimulates pro-fibrotic events in GHMT-infected fibroblasts. Next, we sought to determine whether overactivation of profibrotic signalling suppresses cardiac reprogramming. GHMT induced B45 or B35% of fibroblasts to be positive for cTnT or a-actinin, respectively, by day 14. Treatment of GHMT-infected cultures with TGF-b1 significantly reduced the cTnT þ and a-actinin þ populations ( Fig. 3d-g). Although treatment with TGF-b1 did not significantly decrease the fraction of cTnT þ cells in GHMT2m cultures on day 14 (Fig. 3d,e), expression of other cardiomyocyte markers, including a-actinin, Myh6 and Actc1, was significantly downregulated by treatment with this pro-fibrotic stimulus ( Fig. 3f-h). Next, we examined the effect of TGF-b1 treatment on beating iCMs in GHMT-and GHMT2m-infected cultures. GHMT produced B1,000 beating iCMs per cm 2 by 1 month; however, this was reduced by TGF-b1 treatment (Fig. 3i). By 1 month, GHMT2m produced B2,000 beating iCMs per cm 2 . TGF-b1 treatment decreased beating iCMs in GHMT2m-infected cultures by 100-fold (Fig. 3i). Taken together, these data demonstrate that overactivation of pro-fibrotic signalling attenuates reprogramming of fibroblasts into beating cardiomyocytes.
ROCK inhibitors enhance cardiac reprogramming. We next examined whether inhibiting pro-fibrotic events enhances cardiac   reprogramming. First, we addressed the impact of blocking ROCK signalling on cardiac reprogramming of fibroblasts. In response to increased mechanical tension, Rho triggers the formation of stress fibres and stimulates pro-fibrotic events via activation of its downstream effector, ROCK. The ROCK inhibitor Y-27632 decreased expression of Fn-EDA and aSMA but did not change Snail and Slug expression in GFP-, GHMT-and GHMT2m-infected cultures (Supplementary Fig. 2 and Supplementary Fig. 6a-c). These data confirm that Y-27632 suppresses pro-fibrotic genes in reprogramming cells. Next, we examined the effect of Y-27632 on cardiac reprogramming. Y-27632 treatment enhanced assembly of myocyte sarcomeres ( Fig. 4a,b) and cardiomyocyte-specific gene expression ( Supplementary Fig. 6d). GHMT induced B200 beating iCMs per cm 2 by 4 weeks. Adding Y-27632 (30 mM) produced contracting iCMs by 12 days and increased the number of beating iCMs by sevenfold by 4 weeks (Fig. 4c). By day 8, GHMT2m produced B160 beating iCMs per cm 2 and GHMT2m plus Y-27632 induced B700 beating iCMs per cm 2 (Fig. 4d). By day 12, we observed B1,100 beating cells per cm 2 and B11,000 total cells per cm 2 in GHMT2m-infected cultures. However, we observed 2,300 beating cells per cm 2 and 7,000 total cells per cm 2 in GHMT2m/Y-27632 cultures. Therefore, Y-27632 treatment increased beating cell population from B10 to B30% in GHMT2m cultures (Fig. 4e). Changes in cell numbers occur during the reprogramming process, which may contribute to the overall efficiency. We calculated the beating iCM yield as the percentage of beating cells in relation to the initial number of plated fibroblasts right before viral infection, as suggested 35 . We observed B5,500 fibroblasts right before viral infection. Using this calculation, Y-27632 treatment increased the beating iCM yield from B20 to B45% in GHMT2m cultures by day 12 (Fig. 4e).
To exclude the possibility of Y-27632 off-target effects, we tested more ROCK inhibitors. Other ROCK inhibitors, including Thiazovivin 36 and SR-3677 (ref. 37), also enhanced cardiac reprogramming by promoting formation of spontaneously beating iCMs (Fig. 4f). Taken together, these results suggest that small molecules that inhibit pro-fibrotic genes by targeting ROCK signalling significantly enhance reprogramming of MEFs into beating cardiomyocytes.   Inhibition of TGF-b signalling enhances reprogramming. In addition to the ROCK pathway, the TGF-b pathway plays a critical role in pro-fibrotic gene expression [5][6][7] . Treatment of MEFs with A83-01, a selective inhibitor of TGF-b signalling 38 , decreased phosphorylation of Smad2, and inhibited expression of Fn-EDA, Col1a1 and Col3a1, and formation of aSMA þ stress fibers in GHMT-and GHMT2m-infected cultures ( Supplementary Figs 2 and 7a-c). We also observed that 2 m significantly decreased expression of Fn-EDA, Col1a1 and Col3a1 ( Supplementary Fig. 7b). Among the four groups of cultures, GHMT, GHMT with A83-01, GHMT2m and GHMT2m with A83-01, the lowest expression of these pro-fibrotic genes was detected in GHMT2m cultures treated with A83-01 ( Supplementary Fig. 7b), suggesting that two microRNAs and A83-01 synergistically suppress pro-fibrotic gene expression in reprogramming cells.
Given the ability of A83-01 to inhibit pro-fibrotic signalling, we next examined effects of the compound on cardiac reprogramming. A83-01 alone did not induce cTnT þ cells, as determined using flow cytometry. In contrast, A83-01 increased cTnT high population from B28.1 to B30.7% in GHMT-infected cultures and from B38.8 to B45.6% in GHMT2m-infected cultures on day 7 ( Supplementary Fig. 7d,e). More significant increases occurred on day 14. Immunostaining analysis revealed that A83-01 significantly increased the percentages of cTnT þ cells in GHMT-and GHMT2m cultures from B45%, B60% to B58%, B67%, respectively (Fig. 5a,b). A83-01 also significantly increased the percentages of a-actinin þ cells in GHMT-and GHMT2m cultures from B34%, B42% to B57%, B64%, respectively (Fig. 5a,b). GHMT2m plus A83-01 induced 460% of fibroblasts to be positive for cTnT or a-actinin (Fig. 5b). A83-01 treatment also enhanced a programme of cardiomyocyte gene expression in reprogramming cells ( Supplementary Fig. 7f). The percentage of b-myosin heavy chain (b-MHC) in total MHC (a-MHC þ b-MHC) is a critical indicator of cardiomyocyte maturation. There is lower percentage of b-MHC in adult cardiomyocytes than that in neonatal cardiomyocytes ( Supplementary Figs 2 and 7g). GHMT2m plus A83-01 generated  Figs 2 and 7g), suggesting that induced cardiomyocytes are not as mature as adult cardiomyocytes. By 2 weeks, immunostaining analysis showed that expression of myosin light chain-2a (Myl7), an atrial-specific marker, was expressed in B70% (n ¼ 3) of cTnT þ cells, whereas the ventricular-specific marker Myl2 was detected in only a few of cells ( Supplementary Fig. 7h), suggesting that induced cardiomyocytes are composed of different subtypes of cardiomyocytes. The dramatic effect of blocking TGF-b signalling was also observed at the level of cell contraction. In the absence of A83-01, beating cells were observed 3 weeks post expression of GHMT and reached B200 cells per cm 2 by 4 weeks (Fig. 5c). Treatment of GHMT-infected cultures with A83-01 led to cell beating by 11 days, which culminated in dramatic and spontaneous contraction of B4,400 cells per cm 2 by 4 weeks (Fig. 5c and Supplementary Movie 4). Even more pronounced changes were observed in GHMT2m-infected cultures. By day 8, GHMT2m reprogrammed a fraction of MEFs into beating cells (B300 cells per cm 2 ) and A83-01 treatment increased the number of beating cells by eightfold ( Fig. 5d and Supplementary Movie 5). On day 11, we observed B1,000 beating cells per cm 2 and B10,000 total cells per cm 2 in GHMT2m-infected cultures. However, we observed B7,000 beating cells per cm 2 and B11,500 total cells per cm 2 in GHMT2m/A83-01 cultures. Therefore, A83-01 treatment increased beating cell fraction from B10 to B60% ( beating iCM yield. The initial number of fibroblast per cm 2 before viral infection was B5,500. Therefore, we obtained the beating iCM yield of B120% for GHMT2m plus A83-01 on day 11 (Fig. 5e). This represents the highest efficiency of reprogramming and correlates with the most significant loss of pro-fibrotic gene expression ( Supplementary Figs 2 and 7a-c). Addition of A83-01 and Y-27632 together produced B69% spontaneously beating cells in GHMT2m cultures, which is not significantly different from the fraction of beating cells induced by A83-01 alone (Fig. 5e). These data suggest that inhibition of pro-fibrotic events by multiple compounds is not able to further increase the reprogramming efficiency and it will be necessary to manipulate other barriers to further optimize cardiac reprogramming.
To globally compare the similarity of gene expression in induced cardiomyocytes and primary cardiomyocytes, we performed RNA-Seq on MEFs, neonatal mouse cardiomyocytes (NMCMs), GHMT-infected cultures and GHMT2m-infected cultures treated with dimethylsulphoxide or A83-01 on day 7. Heat map indicates that GHMT2m plus A83-01 not only increases the efficiency but also increases the similarity between iCMs and primary cardiac myocytes, compared with GHMT (Fig. 5g). To gain insights into mechanisms of A83-01 action, we analysed gene expression profiles in GHMT2m cultures treated with or without A83-01 on day 7 by RNA-Seq. The top 10 GO terms that were enriched among genes upregulated in A83-01treated cultures belong to cardiac development ontologies, whereas the top eight GO terms enriched among genes downregulated in A83-01-treated cultures belong to pro-fibrotic or ECM ontologies (Table 2). These data suggest that small molecules that suppress pro-fibrotic events by targeting TGF-b signalling dramatically decrease pro-fibrotic events and enhance reprogramming of MEFs into beating cardiomyocytes.
Induced cardiomyocytes display functional properties. A gap junction protein, connexin-43 (Cx43), is responsible for electrical coupling and intercellular communication of cardiomyocytes. Immunostaining revealed that Cx43 was detected along the periphery of iCMs (Fig. 6a), indicating development of gap junction channels.
We next sought to characterize functional properties of these induced cardiomyocytes by electrophysiological analysis. Action potentials (APs) were recorded from single spontaneously beating cells on day 9 of reprogramming. Induced cardiomyocytes fired spontaneous APs (Fig. 6b). Frequency and shapes of APs varied, consistent with a wide range of development or differentiation on day 9 of reprogramming ( Fig. 6b,c). APs of induced cardiomyocytes mimic those of fetal (or nodal) cardiomyocytes by showing high rate of spontaneous firing, short AP durations and slow upstroke velocity (Fig. 6c). During cardiac development in vitro and in vivo, early differentiated cardiomyocytes fire nodal-like APs that specialize into atrial-, ventricular-and nodal-like APs at later differentiation stages 42,43 . Therefore, iCMs formed on day 9 appear to mimic early development of cardiomyocytes.
To determine whether iCMs were developing functional excitation-contraction (EC) coupling, we imaged spontaneous calcium transients from Fura-2 AM-loaded iCMs. Spontaneous calcium transients were detected in spontaneously beating cells on days 10-12 (Supplementary Movie 7 and Fig. 6d-g). Addition of nifedipine, a blocker of L-type calcium channels, significantly decreased the frequency of calcium transients, indicating that L-type calcium channels contribute to EC coupling in the iCMs as they do primary cardiomyocytes (Fig. 6d,e). To determine whether iCMs are able to appropriately respond to hormone stimulation, a critical function of normal cardiomyocytes, we measured calcium transients in response to the b-adrenergic agonist isoproterenol (Iso). Addition of 1 or 2 mM Iso significantly increased frequency of cell contraction and spontaneous calcium transients (Fig. 6f,g). In the presence of Iso, some cells that did not beat began to spontaneously contract (Fig. 6f). These data suggest that the development of functional EC coupling machinery and b-adrenergic signalling components in iCMs.

Reprogramming mediated by adeno-associated viral vectors.
Clinical trials using adeno-associated virus (AAV) for gene therapy have shown safety and efficacy. Thus, we sought to determine whether AAV-mediated delivery of reprogramming factors to fibroblasts would lead to conversion to cardiomyocytes. Expression of exogenous GFP delivered by AAV in MEFs was transient, compared with that by retrovirus ( Supplementary Fig. 8a). AAV-GHMT2m plus A83-01 induced a small fraction of fibroblasts to be positive for cTnT, a-actinin and to spontaneously contract by day 12 (Supplementary Fig. 8b    the efficiency of AAV-mediated reprogramming is not comparable to retrovirus-mediated reprogramming, our data suggest that it will be possible to develop an episomal vector system to reprogramme fibroblasts into functional cardiomyocytes, thereby circumventing safety concerns associated with the use of retroviral and lentiviral vectors in humans.
Reprogramming of adult fibroblasts into beating iCMs. Adult fibroblasts are less amenable to reprogramming than embryonic fibroblasts 44 . The efficiency of direct reprogramming adult fibroblasts into beating cardiomyocytes is less than 0.1%. We therefore examined whether inhibiting pro-fibrotic events could enhance reprogramming of adult fibroblasts, including ACFs and ATTFs, into functional cardiomyocytes.
Thus, inhibition of pro-fibrotic signalling also enhances reprogramming of adult cardiac and dermal fibroblasts into functional cardiomyocytes.

Discussion
Direct conversion of fibroblasts into cardiac muscle represents a promising approach for cardiac regeneration and would benefit patients with ischaemic heart disease 45,46 . Here we reveal an effective method for reprogramming fibroblasts into functional cardiomyocytes and demonstrate that pro-fibrotic signalling serves as a major barrier to cardiac lineage reprogramming (Fig. 8). Overactivation of pro-fibrotic signalling by TGF-b1 attenuates cardiac reprogramming. Conversely, inhibition of profibrotic events using small molecules that inhibits either RhoA-ROCK or TGF-b signalling dramatically enhanced the ability of cardiogenic factors to reprogramme mouse fetal and adult fibroblast into beating cardiomyocytes (Fig. 8). These findings suggest potential clinical applications and also provide a novel and facile system to study cardiomyogenesis. Activation of lineage-specific genes by transcription factors largely depends on chromatin alterations. During differentiation of embryonic stem (ES) cells into cardiomyocytes, promoters of cardiomyocyte genes are gradually marked by H3K4me3 associated with gene activation and expression 47,48 . Cardiomyogenesis is regulated by a transcriptional cascade, and the first wave of cardiogenic factors induces the second wave of factors 49 . We speculate that overexpression of GHMT leads to active histone mark(s) at regulatory regions of the key factors promoting cardiomyogenesis in fibroblasts. ChIP-Seq revealed that active histone mark, H3K4me2, significantly increased at the regulatory region of the cluster of miR-1-2 and miR-133a-1. MiR-1 and miR-133 promote cardiac muscle formation by several mechanisms, including inhibiting aSMA expression [30][31][32] . Our data reveal that miR-1 and miR-133 enhance cardiac reprogramming, at least in part, by attenuating pro-fibrotic gene expression (Supplementary Table 1 and Fig. 8).
Fibrosis is induced by sequential regulation of signalling pathways 50 . Activation of TGF-b signalling occurs early, whereas the delayed activation of the ROCK pathway helps to maintain myofibroblast phenotype. We observed that Y-27632, a ROCK inhibitor, inhibited expression of pro-fibrotic markers less efficiently than A83-01, a TGF-b-signalling inhibitor (Supplementary Figs 6a,c and 7b,c), and this correlated with the generation of induced cardiomyocytes.
TGF-b signalling plays a critical role in diverse biological processes including cell growth, differentiation and development.
Here we demonstrate that TGF-b signalling inhibits GHMTmediated cardiac reprogramming by promoting fibrotic events. Our findings are likely related to the fact that TGF-b signalling inhibits specification and differentiation of cardiomyocytes from mesoderm cells derived from ES cells 51 . Thus, cardiomyogenesis occurring during differentiation of ES cells or GHMT-mediated reprogramming of fibroblasts likely occurs via common mechanisms.
Application of small molecules and iPS-reprogramming factors efficiently reprogrammes mouse fibroblasts into beating cardiomyocytes 52,53 . Sequentially adding the JAK inhibitor, JI1 and BMP4, four iPS-reprogramming factors, Oct4, Sox2, Klf4 and c-Myc, induced B257 contracting colonies from 100,000 MEFs on day 21 (ref. 52). More recently, in presence of small molecules, SB431542, CHIR99021, parnate and forskolin, Oct4 induced B100 contracting colonies from 10,000 MEFs or 50 beating colonies from 10,000 TTFs on day 30 (ref. 53). Here we show that cardiomyocyte lineage factors GHMT and A83-01 induced B4,400 beating cells from 5,000 MEFs by week 4. The combination of GHMT2m and A83-01 induced B7,000 contracting cells from 5,000 MEFs on day 11, or B300 beating cells from 5,000 ACFs/ATTFs by 1 month. In our studies, spontaneously contracting cells did not form colonies, suggesting that GHMT/GHMT2m might induce beating cardiomyocytes from fibroblasts through a path different from pluripotent factormediated cardiac reprogramming. Therefore, our studies provide an additional approach to efficiently convert fibroblasts into spontaneously beating cardiomyocytes.
Loss of cardiomyocytes, cardiac hypertrophy and fibrosis are major factors contributing to pathological ventricular remodelling in patients post-MI. Treatment with GW788388 significantly decreased TGF-b activity, cardiac fibrosis and ventricular remodelling with attenuation of systolic dysfunction in rats following MI 41 . The ROCK inhibitor Y-27632 decreased infarct size and remodelling in mice following ischaemia/reperfusion injury 54 . Our findings demonstrate that these small molecules enhance GHMT/GHMT2m-mediated cardiomyocyte conversion by blocking pro-fibrotic signalling (Figs 4 and 5 and Table 2), and highlight the potential of a strategy that combines reprogramming factors with anti-fibrotic compounds as a means of regenerating cardiac tissue post-MI.

Methods
Animals. All research involving animals complied with protocols approved by the Institutional Animal Care and Use Committee of the University of Colorado.
Preparation of MEFs. C57BL/6 pregnant mice at E13 were purchased from Charles River's laboratory. Embryos at E14.5 were harvested and their internal organs and head were removed. The body below the liver was minced to fine pieces. Derivation of adult fibroblasts. Hearts for mouse ACFs or skinned tails for ATTFs from adult C57BL/6 mice were cut into small pieces with B1 mm diameter. The biopsies were seeded on a 10-cm culture dish and incubated with 10 ml of growth medium (DMEM/Hi glucose (Hyclone), 10% FBS (Gemini), 1.1% Penicillin-Streptomycin (Gibco) and 1.1% GlutaMAX supplement (Gibco) at 37°C with 5% CO 2 . Fibroblasts migrated out of the biopsies after 3 days. The media were changed every other day. Once the fibroblasts have reached a 70-80% confluence, the cells were harvested and stored for future use.
Immunocytochemistry. Cells were fixed in 2% paraformaldehyde for 15 min on ice and then washed with DPBS (Gibco) three times at room temperature (RT). Counting of beating cells. Cells in 60-mm dishes were examined under an EVOS FL Color Imaging System (Life Technologies) at 25°C. Beating cells were counted at a field with an area of 0.89 mm 2 . Ten to fifteen fields were randomly chosen for each dish.
Recording videos of beating cells. Beating cells were visualized on an inverted microscope (Nikon Diaphot 300), and videos were obtained using a CCD (chargecoupled device) camera (Point Grey Research, Flea 2) at 25°C.
Recording of spontaneous APs. Fibroblasts were plated on fibronectin-coated glass coverslips (BD BioCoat #354088) followed by viral infection and drug treatment. Fibroblasts were sparsely plated for electrophysiological recordings and isolated cells were selected for patching to minimize AP waveform distortions arising from intercellular electrical coupling. Fragments of glass coverslips plated with iCMs on day 9 were transferred to a recording chamber (200 ml) on the stage of an inverted microscope. Cells were constantly perfused (1-2 ml min À 1 ) with normal Tyrode's solution at 35 ± 1°C (in mM, 140 NaCl, 5.4 KCl, 1.2 KH 2 PO 4 , 5 HEPES, 5.55 glucose, 1 MgCl 2 , 1.8 CaCl 2 ; pH-adjusted to 7.4 with NaOH). Spontaneous APs were recorded from spontaneously beating cells in the amphotericin perforated-patch configuration in current-clamp mode (Axopatch 200B, Molecular Devices). The pipette solution contained (in mM) 135 KCl, 0.1 CaCl 2 , 1 MgCl 2 , 5 NaCl, 10 EGTA, 4 Mg-ATP and 10 HEPES, with pH adjusted to 7.2 with KOH and Amphotericin-B (Fisher Scientific, BP928) added at a final concentration of 100 mg ml À 1 . APs were analysed off-line using the ClampFit software (Molecular Devices). AP firing rates for each cell are reported as the average instantaneous firing rate measured during 15-30 s recording windows. AP waveform parameters were calculated for each cell from average waveforms from 5-s recording windows as we have previously described 61 .
Measurements of calcium transient. For calcium imaging, beating iCMs or MEFs were loaded with 5 mM Fura-2 AM (Life Technologies) together with 0.1% Pluronic F-127 (Life Technologies) in modified Tyrode's solution (140 mM NaCl, 5 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 10 mM glucose and 10 mM HEPES, pH 7.4) containing 0.1% BSA and 1% pyruvate for 30 min at 37°C while shielded from light. Before imaging, the cells were washed and allowed to de-esterify the Fura-2 AM for 30 min in Tyrode's solution at RT. Ca 2 þ imaging was performed at RT (B25°C) using Slidebook 5.5 Ca 2 þ Imaging System (3I Marianas Spinning Disk) with an automated fluorescence microscope and a CCD camera at the Advanced Light Microscopy Core, University of Colorado Anschutz Medical Campus. When it is applied, 1 or 2 mM isoproterenol (Sigma) or 10 mM Nifedipine (Sigma) was applied locally to the cell. Calcium transients in individual spontaneous beating cell were calculated by the ratio of fluorescence intensity at 340 nm to that at 380 nm.
Statistical analysis. Statistical analysis was performed using the two-sided Student's t-test. Po0.05 was considered statistically significant.