Overcoming Intrinsic Barriers in Myofibroblasts Permits Efficient Cardiac Reprogramming After Infarction

Abstract

BACKGROUND:

Direct reprogramming of cardiac fibroblasts (CFs) into induced cardiomyocytes (iCMs) holds promise as a therapeutic strategy for heart regeneration. After myocardial infarction (MI), resident quiescent CFs (QCFs) activate and differentiate into myofibroblasts (MFs) in the infarcted region that drive pathological cardiac fibrosis. Converting these injury-activated MFs into iCMs could simultaneously alleviate fibrosis and replenish lost cardiomyocytes. However, whether the marked heterogeneity of CFs in the infarcted heart, and in particular the activation of QCFs into MFs, creates an intrinsic barrier that limits the reprogramming of these injury-activated MFs remains unknown.

METHODS:

To define the molecular basis of this heterogeneity, we purified PDGFRα⁺ CFs and used Postn lineage tracing to distinguish injury-activated MFs from quiescent CFs, enabling matched comparison of their reprogramming competence and single-cell RNA sequencing (scRNA-seq) profiling of post-MI CF subpopulations. A targeted in vitro shRNA screen was conducted against 9 MF-enriched TFs as candidate molecular barriers for cardiac reprogramming. The lead candidate was validated in mouse MFs, human iPSC-derived MFs, primary human MFs, and in vivo using dual-recombinase-mediated lineage tracing. Mechanistic insights were gained through integrated bulk RNA-seq, scRNA-seq and Cleavage Under Targets and Tagmentation (CUT&Tag), together with functional assays including DNA-binding–deficient and domain-swap MEOX1 mutants.

RESULTS:

We identified the upregulated transcription factor MEOX1, a known fibrosis determinant downstream of the key post-MI cytokines transforming growth factor-beta 1 and interleukin-1 beta, as the principal molecular barrier responsible for the profound reprogramming resistance of MFs. MEOX1 knockdown markedly enhanced reprogramming efficiency in both mouse and human MFs and enabled GATA4-free reprogramming combinations. This inhibition depended on the transcriptional activation activity of MEOX1, as disrupting its DNA-binding domain or fusing it to a repressor domain rescued reprogramming. Integrated scRNA-seq and CUT&Tag analyses revealed that MEOX1 binds and stabilizes a fibrotic, MF-defining transcriptional program that antagonizes the cardiogenic program while also modulating the inflammatory response; its knockdown disrupted this fibrotic network to favor iCM fate acquisition. Using stringent dual-recombinase lineage tracing, we demonstrated that MEOX1 knockdown enables highly efficient in vivo MF-to-iCM conversion, leading to significant reductions in cardiac fibrosis and substantial improvement in cardiac function after MI.

CONCLUSIONS:

Our study identifies pathological MEOX1 upregulation as a key mechanism underlying the reprogramming resistance of post-MI mouse MFs and activated human MFs. Overcoming this barrier achieves unprecedented, lineage-confirmed in vivo reprogramming efficiency, thereby addressing a significant obstacle for the clinical translation of in situ reprogramming therapies.

Clinical Perspective.

What Is New?

• Our study reveals that reprogramming efficiency is different across cardiac fibroblast subtypes after myocardial infarction, and pinpoints myofibroblasts (MFs) as the most abundant yet most reprogramming-resistant population.

• It defines the transcription factor MEOX1, a convergence point for transforming growth factor-beta 1/interleukin-1 beta signaling, as the master regulator of this resistance. Its knockdown, validated by stringent lineage tracing, unlocks robust in vivo MF-to-induced cardiomyocyte conversion.

• It also elucidates the mechanism whereby MEOX1 enforces a fibrotic gene program that antagonizes the cardiogenic network, directly linking the pathophysiology of fibrosis to impaired regeneration.

What Are the Clinical Implications?

• Converting MFs into induced cardiomyocytes simultaneously achieves dual benefits: reversal of fibrosis and regeneration of functional myocardium.

• Knockdown of the MF endogenous barrier gene MEOX1 enables streamlined reprogramming cocktails, reducing the number of exogenous transcription factors required, a critical advance for feasible in vivo gene delivery in patients.

• Targeting MEOX1 presents an ideal strategy, simultaneously addressing pathological fibrosis and unlocking the reprogramming potential of MFs.

Cardiovascular diseases are the leading cause of death worldwide. Extensive myocardial injury, such as myocardial infarction (MI), leads to death of cardiomyocytes, scar formation, and cardiac remodeling, which, together, impede cardiac function and can eventually lead to acute heart failure.¹ Cardiac reprogramming represents an attractive approach for heart regeneration. In this strategy, cardiac fibroblasts (CFs) are induced to transdifferentiate into cardiomyocytes to regenerate damaged heart tissue.² Direct intramyocardial injection of viruses expressing the cardiac transcription factors Gata4, Mef2c, and Tbx5 (GMT) or GMT plus Hand2 (HMGT) after coronary ligation results in conversion of nonmyocytes into induced CMs (iCMs) in vivo, ultimately decreasing scar size and improving cardiac function,^3,4 and thus showing considerable promise for in situ cardiac reprogramming strategy in regenerative medicine applications.

Recent advancements in cardiac reprogramming have focused on identifying both booster factors and barriers to improve the efficiency and functionality of iCMs. The supplementation of transcription factors such as MYOCD/SRF,⁵ NKX2-5/HAND2,⁶ MYOCD/SALL4,⁷ AKT1,⁸ ZNF281,⁹ PHF7,^9–11 and ASCL1^12,13 has been shown to significantly enhance reprogramming efficiency. Adjusting the expression levels of MEF2C, GATA4, and TBX5 further boosts efficiency.¹⁴ Concurrently, several intrinsic barriers, including BMI1,¹⁵ IFNAR2,¹⁶ YBX1,¹⁷ and NR4A3,¹⁸ have been identified. Additional enhancement has been achieved using specific microRNAs^19,20 and small-molecule inhibitors targeting key signaling pathways such as transforming growth factor-beta (TGF-β; SB431542 and A83-01),^21–24 ROCK (Y-27632),²¹ JAK (baricitinib and ruxolitinib),^22,23,25 Wnt (XAV939),²⁴ and Notch (DAPT).²⁶

Methodological innovations have also propelled the in vivo study of cardiac reprogramming. Dual-recombinase systems have provided stringent lineage tracing tools for accurately tracking cell fate conversion.^22,27,28 Furthermore, employing a knock-in strategy to overexpress MEF2C/GATA4/TBX5/HAND2 specifically in Tcf21-CreER–traced CFs has demonstrated therapeutic efficacy, improving cardiac function in models of chronic MI²⁹ and heart failure with preserved ejection fraction (HFpEF).³⁰ This benefit is achieved through a dual mechanism: inducing new cardiomyocyte formation while concurrently restraining excessive fibroblast activation, highlighting the multifaceted therapeutic potential of direct cardiac reprogramming strategies.

Cardiac myofibroblasts (MFs) derive mainly from the activation of Tcf21⁺ resident fibroblasts.³¹ This phenotypic switch upon injury leads to their activation and expansion as the predominant CF population,³² which, in turn, contributes critically to pathological cardiac fibrosis.³³ Activated CFs are highly proliferative in the early stage after MI, and differentiate into “mature” MFs that express α-smooth muscle actin (α-SMA) and secrete abundant extracellular matrix proteins, such as collagen, fibronectin and periostin (POSTN), as damage progresses.³⁴ Lineage tracing studies have shown that POSTN expression is restricted to MFs.³¹

These characteristics make MFs an attractive target for reprogramming, offering the dual benefit of ameliorating fibrosis while replenishing lost myocardium. However, most previous studies reporting enhanced reprogramming efficiency have relied on noninjured fibroblast sources, such as mouse embryonic fibroblasts (MEFs), neonatal CFs (nCFs), or healthy adult CFs (ACFs), as starting cells. Although established reprogramming factor combinations (eg, MGT) can efficiently convert nCFs or ACFs into iCMs, fibroblasts from adult MI hearts exhibit profound resistance to reprogramming. This resistance can be partially overcome by strengthening the exogenous cocktail, for example, with MYOCD/SALL4-based combinations.⁷ However, such approaches require delivering additional transcription factors, thereby increasing the genetic payload and complicating clinical translation. Moreover, because such studies used fibroblasts isolated from infarcted hearts without purifying MFs or removing non-MF populations such as macrophages, the low reprogramming efficiency they observed could reflect extrinsic inhibitory signals rather than an MF-intrinsic barrier,^16,35 leaving unresolved whether activated MFs themselves possess endogenous obstacles to reprogramming. Thus, to advance cardiac reprogramming within the infarct zone, it is essential to understand and overcome the heterogeneity in reprogramming potential across distinct CF subtypes.

In this study, using PDGFRα-based sorting combined with Postn-MCM;Rosa-tdTomato lineage tracing, we purified injury-activated CFs (ie, MFs) from infarcted hearts and found that, compared with quiescent CFs under matched reprogramming conditions, MFs displayed markedly lower cardiac reprogramming efficiency in vitro. Through single-cell RNA sequencing of post-MI CFs, we selected a set of MF-enriched transcription factors as candidate factors that might intrinsically suppress cardiac reprogramming of MFs. Among them, the MF fate-related transcription factor MEOX1^36,37 was identified as a primary endogenous barrier to reprogramming in MFs from MI model mice, both in vitro and in vivo, as well as in human MFs in vitro; notably, direct MEOX1 knockdown (KD) was sufficient to overcome this intrinsic resistance and restore efficient MF-to-iCM conversion, in line with previous reports that forced MEOX1 expression suppresses cardiac reprogramming.²⁹ Mechanistically, MEOX1 functioned as an active transcriptional barrier rather than merely as a marker of activated MFs. Its anti-reprogramming effect was markedly attenuated when the DNA-binding domain was disrupted or when MEOX1 was converted into a transcriptional repressor, indicating that transcriptional activation is required for MEOX1-mediated resistance. Our transcriptomic, chromatin-binding and single-cell fate-axis analyses support a model in which MEOX1 acts as a transcriptional barrier by helping preserve an injury-activated MF state, thereby limiting cardiogenic network activation during MF-to-iCM reprogramming, in line with previous observations linking MEOX1 to CF activation.^36,37 Thus, Meox1 KD appears to weaken the transcriptional architecture that sustains MF identity, thereby rendering injury-activated MFs more permissive to iCM fate conversion. Using stringent dual-recombinase lineage tracing, we demonstrate that MEOX1 KD rescues the inefficient reprogramming typical of the infarct zone and achieves highly efficient in vivo MF-to-iCM conversion, thereby advancing this therapeutic strategy toward clinical translation.

METHODS

Detailed experimental procedures and materials are described in the Supplemental Methods. All animal experiments were conducted in accordance with institutional guidelines and were approved by the institutional animal care and use committee of Peking University. The data, analytical methods, and study materials will be made available to other researchers for the purpose of reproducing the results or replicating the procedures.

Statistical Analyses

All data are presented as the mean±SEM. Differences between groups were examined for statistical significance using the Student t test or ANOVA. Data were analyzed using GraphPad Prism 10 (GraphPad Software, Inc.). P values of <0.05 were regarded as significant (ns, nonsignificant; *P<0.05, **P<0.01, ***P<0.001, ****P<0.0001).

Data Availability

All scRNA-seq, RNA-seq, ChIP-seq, and Cleavage Under Targets and Tagmentation (CUT&Tag) data that support the findings of this study have been deposited in the Gene Expression Omnibus (GEO) under accession code GSE241144. Previously published data that were re-analysed here are available under accession codes ChIP-seq (GSE155882) and single-cell RNA-seq (ArrayExpress database, E-MTAB-7376). Source data are provided with this paper. All other data supporting the findings of this study are available from the corresponding author upon reasonable request.

RESULTS

The Capacity for Reprogramming Is Impaired in Activated Cardiac MFs After MI

In order to study the difference in reprogramming potential between QCFs and injury-induced cardiac MFs, we crossed a Postn (encoding Periostin) knock-in mouse line carrying the MerCreMer cDNA cassette (Postn-MCM) with Rosa26 (R26) loxP-inactivated tdTomato (R26-lsl-tdTomato) mice to permanently label Postn-expressing MFs with tdTomato after tamoxifen treatment (Figure 1A).³¹ We then induced myocardial infarction in the Postn-MCM^+/-;R26-lsl-tdTomato^+/- mice and treated them with tamoxifen every other day after surgery. For in vitro experiments, to reduce contamination by macrophages and other immune cells,¹⁶ CFs were isolated by flow cytometry at 7 days after injury using the CF marker, CD140a (PDGFRα). PDGFRα⁺ CFs accounted for ≈85% of the isolated cells, approximately half of which expressed Postn-tdTomato (Figure 1B). Cardiac reprogramming was then induced using a polycistronic MGT lentivirus,¹⁴ with Myocd and Sall4 (referred to as MGTMS hereafter).⁷ Potential iCMs were then detected by immunofluorescence staining for cardiac troponin I (cTnI). On reprogramming days 7 and 14, QCFs (ie, tdTomato^- CFs) showed higher cardiac reprogramming efficiency (≈3–8-fold) than Postn-tdTomato⁺ MFs (Figure 1C and 1D; Figure S1A and S1B). qPCR-based transcriptional analysis indicated that the 5 reprogramming factors were expressed at comparable levels (Figure S1C); we also observed that QCFs and MFs exhibit similar infection efficiencies when simultaneously exposed to lentivirus expressing enhanced green fluorescent protein (EGFP; Figure S1D and S1E). These results indicated that the difference in reprogramming efficiency was not attributable to virus infection. In light of this finding, we hypothesized that MFs are more resistant to reprogramming and some endogenous factors in MFs hindered cardiac reprogramming.

Figure 1.

The capacity for reprogramming is impaired in activated cardiac fibroblasts (CFs) after myocardial infarction (MI). A, Schematic diagram of the Postn genetic locus with a tamoxifen-regulated MerCreMer cDNA cassette, which was crossed with Rosa26-tdTomato reporter mice containing 2 loxP sites to allow for Cre-dependent lineage tracing. B, Cell sorting of quiescent CFs (QCFs) and injury-activated cardiac myofibroblasts (MFs) by fluorescence-activated cell sorting (FACS), QCFs were PDGFRα⁺Postn^-; MFs were PDGFRα⁺Postn⁺. C, Immunostaining of induced cardiomyocytes (iCMs) produced by polycistronic MGT with Myocd and Sall4 (MGTMS) from QCFs (tdTomato^-) and MFs (tdTomato⁺) 2 weeks after infection. Postn-tdTomato is shown in red. cTnI staining is illustrated in green. Nuclei are shown in blue. Scale bar=200 µm. D, Quantification of cTnI+ cells indicated in C (QCFs and MFs; n=3). Data are presented as the mean±SEM. **P≤0.01 vs the relevant control using the Student t test. E, Uniform Manifold Approximation and Projection (UMAP) visualization of CFs and subpopulations separated according to experimental condition (healthy, MI-D3: 3 days after myocardial infarction). F, UMAP visualization of CF marker genes. G, UMAP visualization of fibrosis marker genes. H, Trajectory analysis (Palantir algorithm) identified the differentiation pathway of CFs 3 days after MI injury. I, Pseudotime analysis and feature plot showing the expression of indicated genes. J, Dot plot showing expression of 9 candidate transcription factors (TFs) for each cluster of CFs. K, Candidate-centered activation-axis correlation analysis of the 9 selected MF-enriched TFs. Regulatory TFs positively or negatively correlated with each candidate TF were projected onto the QCF-to-MF activation axis by Spearman correlation; boxplots summarize activation-axis correlation coefficients. Curve color indicates Spearman rho to the corresponding candidate TF. P values and false discovery rates (FDRs) are provided in Table S3.

To identify differentially expressed genes between QCFs and MFs that could potentially block their reprogramming ability, we performed single-cell RNA sequencing (scRNA-seq) using CFs on MI Day-3 (MI-D3) for comparison with CFs isolated from healthy mice. Subsequent UMAP analysis of transcriptomic signatures identified 5 fibroblast subpopulations (Figure 1E; Figure S1F), all of which expressed canonical fibroblast markers Col1a1, Pdgfra, and Tcf21 at varying levels (Figure 1F and 1I). Based on previous studies, QCFs could be divided into 3 populations under healthy conditions.^32,38 Among them, Fibroblast-Sca1-high (F-SH) and Fibroblast-Sca1-low (F-SL) populations were distinguished by differences in their expression of the cell surface stem/progenitor cell marker, Ly6a (Sca1).³⁹ The third population of healthy CFs, comprising Fibroblast-Wnt–expressing (F-WntX) cells, was distinguished by its high differential expression of Wif1 (Figure 1F), a canonical secreted WNT pathway inhibitor⁴⁰ and connective tissue growth factor (CTGF) signaling antagonist.⁴¹

After cardiac injury, MFs entered either activated (F-Act) or proliferative (Fibroblast-cycling; F-Cyc) states. These MFs exhibited reduced expression of quiescent CF markers (Tcf21, Ly6a, and Pdgfra), but increased expression of fibrosis-related genes (Figure 1G), such as extracellular matrix markers (Postn; Collagen Triple Helix Repeat Containing 1, Cthrc1) and contractile proteins (α-SMA; gene name: Acta2).³² In contrast with these signature genes of the F-Act subpopulation, the F-cyc subpopulation expressed Birc5 and Ccnb2 at high levels (Figure 1F), indicating their proliferative state. To better understand the associations between identified clusters and the fibroblast activating process, trajectory analysis was then performed using the Palantir algorithm (Figure 1H).⁴² F-SH was identified as the root of this process, characterized by high expression of stemness-associated gene Ly6a³⁹ and the resident CF gene Tcf21,⁴³ with low expression of CF activation regulators Meox1³⁶ and activation marker Postn³¹ (Figure 1I). Pseudo-time analysis revealed that differentiation in fibroblasts after myocardial injury progresses from F-SL to F-Act, with F-Act entering a proliferative state (F-Cyc) and contributing to extracellular matrix secretion (Figure 1H).

To identify potential molecular barriers of MF reprogramming, we examined the expression of transcription factors (TFs) characteristic of each MF subpopulation and selected candidate barrier genes based on their differential expression (Figure 1J and Figure S1G). In order to select potentially important genes, we also re-analyzed published data from a CF scRNA-seq study that examined sham, MI-D3 and MI-D7 CF populations (Figure S1H). Combining our scRNA-seq data with analysis of TF expression profiles for different MF subpopulations, we selected 9 candidate genes to test their roles in regulating cardiac reprogramming (see Figure 1J and Figure S1G for genes detected in our data or Figure S1I for genes detected in published data sets).

To further assess the regulatory context of the selected MF-enriched TFs, we performed a candidate-centered activation-axis correlation analysis. For each of the 9 candidate TFs, regulatory TFs positively or negatively correlated with the candidate were projected onto the QCF-to-MF activation axis. The x-axis position of each regulatory TF indicates its association with the QCF or activated MF side, whereas the curve color indicates its correlation with the corresponding candidate TF. Most candidate-associated regulatory neighborhoods were shifted toward the activated MF side to varying degrees (Figure 1K), supporting the rationale for functionally screening these TFs as candidate barriers to MF reprogramming. These results provide a further layer of validation for our candidate selection, indicating that the 9 TFs identified in the preceding step are indeed likely to be key transcription factors shaping MF cell fate.

A Targeted TF KD Screen Identifies Meox1 as an Intrinsic Barrier for Cardiac Reprogramming of MFs

To test whether these endogenous TFs indeed functioned as barriers for reprogramming in MFs, we constructed a small shRNA library targeting the above 9 candidate TFs to evaluate whether and how KD of each TF affected cardiac reprogramming (Figure 2A). Each gene was targeted with 2 to 5 short hairpin RNAs (Table S1). The KD efficiency of each shRNA was evaluated by qPCR, and the shRNA with the highest KD efficiency was subsequently selected for cardiac reprogramming experiments (Figure S2A). We isolated MFs from mice at 5 or 7 days after myocardial infarction and individually infected shRNA-expressing virus together with the MGT plus Myocd (MGTM) combination. Immunofluorescent (IF) staining for cTnI indicated that shRNAs targeting Meox1 resulted in significantly higher (≈3.5-fold) numbers of reprogrammed cells compared with that in the scrambled shRNA control groups (Figure 2B through 2D; Figure S2B). Besides, Meox1 KD increased the percentage (≈3-fold) of cTnT⁺ iCMs evaluated by FACS (Figure 2G and 2H) and the proportion of cTnI⁺ (≈5-fold) iCMs (Figure 2E and 2F). Furthermore, this Meox1 KD-associated increase in reprogramming efficiency was further verified by the elevated expression of CM functional markers, including sarcomere (Actc1 and Myh6)-, contractility (Tnni3)-, and ion channel-associated (Ryr2) genes compared with that in the shRNA scramble controls (Figure 2I). More importantly, calcium flux experiments showed a >3-fold increase in the number of cells exhibiting periodic calcium oscillations (Figure 2J and 2K; Movie S5 and S6), and we observed a nearly 4-fold increase of spontaneously beating iCMs (Figure 2L; Movie S1 and S2) in the shMeox1-treated group compared with the control groups. Electrophysiological analyses also revealed that the iCMs with Meox1 KD could exhibit action potentials (Figure S2C). iCMs generated upon Meox1 KD exhibited better-organized sarcomere structures, as assessed by α-Actinin immunofluorescence. In addition, these iCMs co-expressed multiple cardiomyocyte markers: shMeox1 markedly increased the proportion of cTnI+ cells co-stained with α-Actinin, RYR2, MYL2, or TTN, confirming iCM identity by multiple markers (Figure 2P and 2Q).

Figure 2.

Targeted TF knockdown screen identifies Meox1 as a cardiac reprogramming barrier. A, Schematic representation of the shRNA screen to identify transcriptional barriers of induced cardiomyocyte (iCM) reprogramming. B, Quantification of cTnI⁺ cells induced by polycistronic MGT with Myocd (MGTM) plus scramble shRNA control (shCtrl) or shRNA with the highest knockdown (KD) efficiency of each candidate gene at day 14 (n=3). C and D, Immunostaining (C) and quantification (D) for cTnI⁺ iCMs 14 days after MGTM and shMeox1, or shCtrl transduction on myofibroblasts (MFs). Scale bar=100 µm (n=6). E and F, The entire well view (24-well plate) of cTnI+ cells 14 days after MGTM and shMeox1, or shCtrl transduction on MFs (E) and quantification (F) of cell percentage. Scale bar=1200 µm (n=3). G and H, Representative flow plots (G) and quantification (H) for cTnT+ cells 14 days after MGTM and shMeox1, or shCtrl transduction on MFs (n=3). I, mRNA expression of CM marker genes in infected MFs (2 weeks after transduction) across 3 groups: enhanced green fluorescent protein (EGFP) with shCtrl, MGTM with shCtrl, and MGTM with shMeox1 (all using the pLKO-U6 vector). J and K, Representative image (J) and quantification (K) of iCMs displaying calcium transients after 4 weeks of induction (n=4). Each data point represents an image window for each sample. The solid white lines show the areas of highest (Ca²⁺ max) and lowest (Ca²⁺ min) calcium transient intensity within the same cell. L, Quantification of spontaneously contracting iCMs with indicated viral infection 6 weeks after reprogramming by MFs, corresponding to Movie S1 and Movie S2 (n=3). M and N, Immunostaining (M) and quantification (N) for cTnI+ iCMs 14 days after MGTM and vector control, or Meox1 overexpression on MFs. Scale bar=100 µm (n=3). O, mRNA expression of Meox1 and CM marker genes in vector control or Meox1 overexpression in MFs 2 weeks after MGTM transduction. P and Q, Representative immunostaining (P) and quantification (Q) of cTnI co-stained with α-Actinin, RYR2, MYL2, or TTN in iCMs 14 days after MGTM with shCtrl or shMeox1 transduction on MFs. White boxes indicate areas shown at higher magnification in the adjacent panels. Scale bar=200 µm (n=3). R through T, Heatmap comparing differentially expressed transcripts on day 14 after MGTM with shCtrl- or shMeox1-infected MFs identified by RNA-seq (R). Upregulation (red), downregulation (blue). Color scale represents Z-score (n=3 biologically independent samples per group; false discovery rate [FDR]<0.1, fold change [FC]>2). Gene ontology (GO) analysis illustrating biological processes associated with genes downregulated (S) or upregulated (T) as assessed using RNA-seq. U, One-sided preranked gene set enrichment analysis (GSEA) across GO GSEA of selected collagen- and integrin-associated GO terms in MGTM+shMeox1 vs MGTM+shCtrl RNA-seq data (n=3 per group). Genes were ranked by −log2 fold change, with shMeox1-downregulated genes positioned on the left. GO biological process and cellular component gene sets containing 10–500 genes were tested using fgsea with scoreType = “pos.” Normalized enrichment score (NES) and FDR are shown. V, Expression of the fibrosis related genes (differentially expressed genes [DEGs]) on day 14 after MGTM with shCtrl- or shMeox1-infected MFs identified by RNA-seq as visualized by a heatmap. All data are presented as the mean±SEM. *P≤0.05, **P≤0.01, ***P≤0.001, and ****P≤0.0001 vs the relevant control using the Student t test (D, F, K, L, N, O, and Q) and 1-way ANOVA followed by Tukey’s multiple-comparisons test (B, H, and I). ns indicates not significant. SMC indicates smooth muscle cell.

To assess the enhancement of cardiac reprogramming with Meox1 KD, we used the polycistronic MGT construct to induce reprogramming in MFs in vitro, with low basal reprogramming efficiency.⁷ In agreement with our above results, with Meox1 KD, reprogramming efficiency was improved by 4–5-fold in the numbers of cTnI⁺ cells and spontaneously beating cells (Figure S2D through S2G; Movie S3 and S4). Subsequent RNA-seq analysis of Meox1 KD MFs treated with the MGTM combination showed that the most significant differentially expressed genes (DEGs) were related to muscle system processes and heart contraction (Figure 2R through 2T). In addition, Meox1 KD also led to repression of smooth muscle cell differentiation-related genes, downregulated genes were also enriched in extracellular matrix organization and leukocyte migration.

A one-sided preranked gene set enrichment analysis (GSEA) across gene ontology (GO) biological process and cellular component terms further supported enrichment of collagen- and integrin-associated extracellular matrix programs toward the shMeox1-downregulated side of the transcriptome, including integrin-mediated signaling pathway, collagen biosynthetic process, and collagen metabolic process (Figure 2U). These results are consistent with the attenuation of ECM-associated transcriptional programs after Meox1 KD during MGTM-mediated reprogramming.

Notably, key fibrosis-related genes, including Postn, Col8a1, Fn1, Acta2, and Ccn2, exhibited marked downregulation (Figure 2V). These findings suggest that Meox1 KD suppresses the fibroblast-associated program.

To further assess the repressive effects of Meox1 on cardiac reprogramming, the Meox1 gene was overexpressed in MFs together with MGTM cardiac reprogramming factors. Immunofluorescent staining for cTnI and cTnT showed that cardiac reprogramming efficiency was significantly lower in cells overexpressing Meox1 compared with that in vector control MFs (Figure 2M and 2N; Figure S2H and S2I). qPCR-based relative expression analysis indicated that transcript levels of cardiac marker genes were also downregulated in Meox1 overexpressing cells (Figure 2O). Further bulk RNA-seq comparing MFs with and without Meox1 overexpression after 6 days of reprogramming with MGTM indicated that Meox1 expression could induce extensive transcriptomic changes (Figure S2J), with characteristic downregulation of genes encoding mature cardiac structural proteins, sarcomeres, and contractile fibers, such as Actc1, Tnni3, Myh6, and Tnnt2, among others (Figure S2K). GO term annotations of upregulated genes showed enrichment for leukocyte migration, Collagen-related extracellular matrix and cytokine-cytokine receptor interactions, indicating that an MF-associated program was stimulated under Meox1 overexpression (Figure S2L). To validate the specificity of shRNA targeting, we constructed an shRNA-resistant Meox1 expression vector harboring a synonymous mutation that blocked shRNA binding. shMeox1 resulted in significant suppression of Meox1 expression in the wild-type (WT) Meox1 overexpression groups, but not mutant Meox1 groups (Figure S2M). Moreover, in the presence of MGTM, the enhanced cardiac reprogramming observed under Meox1 KD was reversed by overexpressing the shRNA-resistant Meox1 mutant (Figure S2N and S2O). These results thus validated the specificity of the shRNA KD and further supported the likelihood that Meox1 served as a barrier for cardiac reprogramming.

We also noted that Meox1 was expressed at a relatively low level in quiescent fibroblasts in published bulk RNA-seq dataset (Figure S2P).⁴⁴ In line with this finding, Meox1 KD also led to higher cardiac reprogramming efficiency in QCFs (Figure S2Q and S2R), although the fold change of reprogramming efficiency under Meox1 KD was more pronounced in MFs (Figure S2S and S2T).

Meox1 Impedes MF-to-iCM Reprogramming by Activating a Fibrotic Transcriptional Program

As a homeodomain (HD) family transcription factor, Meox1 has been shown to transcriptionally activate the BAPX1 gene in the sclerotome via binding to its promoter region.⁴⁵ In addition, MEOX1 was also found to activate p16^INK4a in a DNA binding dependent manner.⁴⁶ In order to determine whether the suppression of cardiac reprogramming by MEOX1 was dependent on DNA binding, we generated the MEOX1^Q219E variant with abolished DNA binding function through a glutamine to glutamate conversion at residue 50 of the homeodomain (Figure 3A).^46,47 Unlike the inhibitory effect of WT MEOX1, MEOX1^Q219E overexpression did not decrease reprogramming efficiency in MFs expressing MGTM, as indicated by cTnI and cTnT immunostaining (Figure 3B through 3F; Figure S3A), indicating that MEOX1 plays a suppressive role in cardiac reprogramming via its transcriptional activities.

Figure 3.

Meox1 transcriptionally activates and stabilizes a fibrosis-specific program of cardiac fibroblasts. A, Schematic representation of the MEOX1 protein constructs used in this research. MEOX1^Q219E represents the DNA binding-deficient mutation, which included a substitution of glutamate for glutamine at position 50 of the homeodomain. B and C, Immunostaining (B) and quantification (C) of cTnI+ induced cardiomyocytes (iCMs) 14 days after MGT with Myocd (MGTM) with vector control, wild-type, or mutant MEOX1^Q219E overexpression in myofibroblasts (MFs). Scale bar=100 µm (n=3). D, mRNA expression of CM marker genes in wild-type MEOX1 or MEOX1^Q219E overexpression group relative to vector Ctrl in MFs 2 weeks after MGTM transduction (n=3). E and F, Immunostaining (E) and quantification (F) of cTnT⁺ iCMs 14 days after MGTM with vector control, wild-type, or mutant MEOX1^Q219E overexpression in MFs. Scale bar=100 µm (n=3). G through J, Immunostaining and quantification of cTnI⁺ (G and H) and cTnT⁺ (I and J) iCMs 14 days after MGTM and vector-control (Ctrl), wild-type MEOX1, or MEOX1 fusion protein in which the ZIM3 KRAB transcription repression domain is fused to the C-terminus of MEOX1 (MEOX1-TRD) in MFs. Scale bar=100 µm (n=3). K, mRNA expression of Meox1 and CM marker genes in wild-type MEOX1 or MEOX1-TRD overexpression group relative to Ctrl in MFs 2 weeks after MGTM transduction (n=3). L, mRNA expression of fibrosis marker genes in wild-type MEOX1 or MEOX1-TRD overexpression group relative to Ctrl in MFs 2 weeks after MGTM transduction (n=3). M and N, Immunostaining (M) and quantification (N) of cTnI⁺Ki67⁺ iCMs 14 days after MGTM and shMeox1 or shCtrl transduction on MFs. Scale bar=100 µm (n=3). O, mRNA expression of proliferation-related genes in shCtrl- or shMeox1-infected MFs 2 weeks after MGTM transduction (n=3). All data are presented as the mean±SEM. *P≤0.05, **P≤0.01, ***P≤0.001, and ****P≤0.0001 vs the relevant control using the Student t test (N and O) and 1-way ANOVA followed by Tukey’s multiple-comparisons test (C, D, F, H, J, K, and L). ns indicates not significant.

To further explore the possible mechanisms by which Meox1 blocks cardiac reprogramming, we analyzed the gene sets most correlated with Meox1 expression in Healthy and MI-D3 scRNA-seq data and found that these genes were mainly associated with extracellular matrix assembly and collagen fibril organization (Figure S3D and S3E). We then reanalyzed MEOX1 occupancy of promoter regions in previously reported ChIP-seq data.³⁶ GO analysis of MEOX1-bound genes confirmed that its regulatory targets were enriched in processes associated with fibroblast activation, such as regulation of cell migration, motility and extracellular matrix (Figure S3F and S3G).

In order to test whether Meox1 impeded cardiac reprogramming through positive regulation of a fibrosis-associated program, we fused the transcriptionally repressive ZIM3 Krüppel-associated box (KRAB) domain to the C-terminus of MEOX1 (MEOX1-TRD) and expressed this fusion protein in MFs on MGTM condition. We found that reprogramming efficiency was enhanced in cells expressing MEOX1-TRD compared with that in cells expressing WT MEOX1 (Figure 3G through 3K). Besides, overexpression of MEOX1-WT significantly increased the expression levels of Postn and Ctgf. In contrast, MEOX1-TRD effectively inhibited the activation of these genes by MEOX1 (Figure 3L), indicating that MEOX1 mainly exerted its function through transcriptional activation. To test this possibility, we fused the MYOD transactivation domain (TAD)^48,49 to the C-terminus of MEOX1, generating MEOX1-TAD, and examined whether this chimeric protein could activate cardiomyocyte gene loci. However, MEOX1-TAD did not increase cardiomyocyte gene expression compared with WT MEOX1-expressing cells (Figure S3K through S3O). These results suggest that MEOX1 blocks cardiac reprogramming mainly through its transcriptional activation activity, likely by activating a fibrosis-enriched program that antagonizes cardiac reprogramming, rather than through direct repression of cardiomyocyte gene loci.

In light of the above evidence, we next sought to determine whether the enhanced iCM reprogramming observed under Meox1 KD was mediated through the regulation of cell proliferation. To this end, we first analyzed the difference in cell counts between shMeox1-infected CFs and shRNA scramble control CFs, both expressing the MGTM combination, and found no significant difference in CF numbers between Meox1 KD and control groups (Figure S3B). In addition, Ki67 immunofluorescent staining also showed no difference between shRNA scramble control and Meox1 KD cells (Figure 3M and 3N; Figure S3C), whereas qPCR assays indicated that the expression levels of the proliferation-related gene Aurkb was comparable between groups. Although the expression level of Ki67 in the Meox1 KD group showed a significant change, the extent of this change was only ≈0.13-fold (Figure 3O). These findings indicate that the increase in cardiac reprogramming efficiency induced by Meox1 KD does not result from accelerated cell proliferation.

We performed CUT&Tag sequencing to identify MEOX1 binding sites under MGTM reprogramming conditions (Figure S3H). Approximately 31% of these binding sites were located within gene promoter regions (Figure S3I), emphasizing the pivotal role of Meox1 in transcriptional regulation. To further explore its functional impact, we intersected MEOX1-bound genes with genes directionally regulated after Meox1 KD (|log2 fold change [FC]|>0.2; Figure 4A). This analysis revealed that a subset of genes was downregulated, primarily associated with RNA polymerase transcription regulation, inflammation, and extracellular matrix assembly. Notably, under MGTM reprogramming conditions, MEOX1 binding was enriched at the promoter regions of key fibrosis-related genes, including Postn and Ctgf, and KD of Meox1 significantly reduced their expression. Additionally, MEOX1 was found to bind to its own promoter region, indicating a potential self-stimulating mechanism (Figure 4B). These findings are consistent with previous studies highlighting the critical role of this regulatory network in cardiac fibrosis.^36,37,50

Figure 4.

CUT&Tag and scRNA-seq analyses reveal that Meox1 stabilizes a fibrosis-associated transcriptomic program that antagonizes cardiac reprogramming. A, Heatmap of MEOX1-bound genes responsive to Meox1 KD in RNA-seq analysis (adjusted P<0.1 and |log2 fold change [FC]|>0.2). Representative downregulated and upregulated genes and enriched gene ontology (GO) biological process terms are shown. Heatmap colors represent row-scaled Z-scores. GO enrichment was assessed by hypergeometric over-representation analysis with multiple-testing correction. B, Integrative Genomics Viewer (IGV) tracks showing the distribution of MEOX1-binding peaks by Cleavage Under Targets and Tagmentation (CUT&Tag) at Postn, Ctgf, and Meox1 gene locus between vector control and MEOX1-OE groups under polycistronic MGT with Myocd (MGTM) reprogramming condition. C, Schematic diagram of the experimental design for single-cell RNA-seq of myofibroblasts (MFs) on day 6 after MGTM infection. D, UMAP analysis of single-cell RNA-seq data from MGTM-reprogrammed MFs. Left, UMAP visualization of the cell clusters defined as Fib0, Fib1, Fib2, Fib3, Fib4, induced cardiomyocyte (iCM), and before iCM. Right, UMAP visualization of Meox1. E, Dot plot showing the relative expression levels of representative cardiac, fibroblast, and endocardium genes across the clusters. F, Trajectory analysis (Palantir algorithm) identified the differentiation pathway of MFs in MGTM-mediated cardiac reprogramming. G, Meox1-centered fate-axis correlation analysis of MEOX1-bound, Meox1-KD-responsive genes during MF-to-iCM reprogramming. Each dot represents one gene and is positioned by its Spearman correlation with the iCM-fate axis; negative values indicate MF-like association, whereas positive values indicate iCM-fate association. Curve color indicates the Meox1 effect score, and boxplots compare KD-down bound and KD-up bound genes. P value was calculated by 2-sided Mann-Whitney U test. H, Pseudotime analysis and feature plot showing the expression of indicated genes.

To further investigate dynamic changes with Meox1 KD during cardiac reprogramming, we isolated CFs from mouse hearts on day 5 after MI, subjected these cells to reprogramming under MGTM conditions, and collected cell samples on Day 6 during reprogramming for scRNA-seq (Figure 4C). The analysis revealed that a subset of fibroblasts had begun transitioning into less mature pre-iCMs and iCMs, characterized by elevated expression levels of cardiomyocyte marker genes (Tnnt2, Tnni3, Ryr2; Figure 4D and 4E; Figure S3J). Notably, Meox1 expression was markedly reduced in both pre-iCMs and iCMs, consistent with the findings of Tani et al²⁹ (Figure S3J). Trajectory analysis further revealed 3 distinct paths of fibroblast fate during reprogramming (Figure 4F and 4H):

1. Progressive maturation path: Some fibroblasts followed a trajectory from CFs to pre-iCMs and then to iCMs (Fib0→Fib2→Fib4→pre-iCM→iCM), representing the gradual maturation process of reprogramming.

2. Direct reprogramming path: Another subset of fibroblasts downregulated CF-related genes and directly differentiated into relatively mature iCMs (Fib0→Fib1→iCM), representing the shortest differentiation path during reprogramming.

3. Reprogramming failure path: A group of fibroblasts failed to reprogram and retained high expression of activated CF markers, such as Postn and Meox1 (Fib0→Fib3). We speculate that high Meox1 expression may play a critical role in the reprogramming failure of these cells.

We next asked whether MEOX1-bound genes affected by Meox1 KD were associated with the MF or iCM state during reprogramming. We intersected MEOX1 CUT&Tag-bound genes with Meox1-KD-responsive genes under MGTM conditions and projected these genes onto the MF-to-iCM reprogramming axis. In this analysis, each gene was positioned according to the Spearman correlation between its single-cell expression and the reprogramming axis; negative values indicate association with the MF-fate side, whereas positive values indicate association with the iCM-fate side. The curve color indicates the Meox1 effect score derived from bulk RNA-seq, with positive values representing genes downregulated after Meox1 KD and negative values representing genes upregulated after Meox1 KD. KD-down MEOX1-bound genes were preferentially associated with the MF-fate side of the trajectory, whereas KD-up MEOX1-bound genes were shifted toward the iCM-fate side (Figure 4G). These results support a model in which Meox1 maintains a fibrotic MF-associated program that is diminished during successful iCM fate acquisition.

Therefore, Meox1 may facilitate the formation of a fibroblast-associated program, whereas Meox1 downregulation disrupts this gene regulatory network and thereby favors iCM fate determination.

Meox1 KD Is Sufficient to Restore High Reprogramming Efficiency Suppressed by TGF-β1 and Interleukin-1 Beta

After examining its effects on downstream targets, we then explored the upstream signaling responsible for regulating Meox1 expression after cardiac injury, which could include genes involved in transducing environmental cues that impede in situ cardiac reprogramming. In particular, TGF-β and interleukin-1 beta (IL-1β) have been previously identified as upstream regulators of Meox1 during fibroblast activation.^36,37,51 Moreover, overactivation of pro-fibrotic signaling, especially TGF-β1, is known to attenuate CF reprogramming into iCMs.²¹ As a shared downstream target of these 2 key inflammatory cytokines in the post-MI microenvironment, we asked whether Meox1 KD could enhance reprogramming efficiency in their presence.

Consistent with previous reports,³⁶ TGF-β1 treatment significantly increased both mRNA and protein levels of Meox1 in MFs. As expected, Meox1 KD attenuated this upregulation. In line with earlier studies, overexpression of Gata4 also downregulated Meox1 expression (Figure 5A, 5B, and 5D). Interestingly, in our system using post-MI MFs, which exhibit elevated basal Meox1 levels (Figure 1I and 1J), Meox1-KD further reduced its expression even in the presence of GATA4-OE (Figure 5A, 5B, and 5D). In addition, TGF-β1–induced cell contraction was attenuated in Meox1-KD MFs (Figure S3P and S3Q).

Figure 5.

TGF-β1 and IL-1β repress cardiac reprogramming specifically through Meox1 regulation. A and B, mRNA (A) and protein (B) levels of Meox1 under vehicle control or transforming growth factor-beta 1 (TGF-β1; 10 ng/mL) treatment in myofibroblasts (MFs) infected with lentivirus (n=4). C, mRNA expression of fibrosis marker genes in MFs 3 days after MGT with Myocd (MGTM) and shCtrl or shMeox1 transduction under solvent control or TGF-β1 (10 ng/mL) treatment (n=3). D, Representative immunoblot showing MEOX1 protein levels under vehicle control or TGF-β1 (10 ng/mL) treatment in MFs infected with lentivirus. E and F, Immunostaining (E) and quantification (F) of cTnI+ induced cardiomyocytes (iCMs) 2 weeks after MGTM and shCtrl or shMeox1 transduction in MFs under vehicle control or TGF-β1 (10 ng/mL) treatment. Scale bar=200 µm (n=3). G and H, Immunostaining (G) and quantification (H) of cTnI+ iCMs 2 weeks after MGTM and shCtrl or shMeox1 transduction in MFs under vehicle control or interleukin-1 beta (IL-1β; 10 ng/mL) treatment. Scale bar=100 µm (n=3). I, mRNA levels of Meox1 under vehicle control or IL-1β (10 ng/mL) treatment in MFs infected with lentivirus (n=4). J and K, Quantification (J) and immunostaining (K) of cTnI⁺cTnT⁺ iCMs 2 weeks after MGTM and shCtrl or shMeox1 transduction on MFs under dimethyl sulfoxide (DMSO) or SB431542 (2 µM) treatment. Scale bar=100 µm (n=4). All data are presented as the mean±SEM. *P≤0.05, **P≤0.01, ***P≤0.001, and ****P≤0.0001 vs the relevant control using 1-way ANOVA followed by Tukey’s multiple-comparisons test. ns indicates not significant.

Moreover, fibrosis associated genes were also significantly reduced by Meox1 KD (Figure 5C). In comparison, MGTM expression induced cTnI expression in 5% of fibroblasts 2 weeks after transduction, whereas TGF-β1 treatment significantly reduced cTnI⁺ populations to approximately 1% of the cultured iCM population. In Meox1 KD cells treated with TGF-β1, cardiac reprogramming efficiency was partially restored to levels comparable with that of the vehicle controls (Figure 5E and 5F). Similarly, IL-1β treatment also upregulated Meox1 and inhibited reprogramming, and this inhibition was likewise rescued by Meox1 KD (Figure 5G through 5I). These consistent results demonstrate that Meox1 acts as a common downstream effector through which distinct inflammatory cytokines impair cardiac reprogramming. Moreover, under treatment with the TGF-β pathway inhibitor, SB431542, Meox1 KD had only minor positive effects on cardiac reprogramming induced by MGTM infection (Figure 5J and 5K).

These results indicate that Meox1, as a shared downstream node of key post-MI cytokines, represents a feasible target; its KD rescues reprogramming efficiency under conditions that mimic the MI microenvironment, thereby informing the design of effective in situ reprogramming strategies.

Meox1 KD Together With Mef2c and Myocd Expression Is Sufficient to Induce High-Efficiency Cardiac Reprogramming

To further explore whether all reported factors shown to produce iCMs were also required under Meox1 KD, we screened variations of the MGTM combination by removing one or more factors in each treatment of cells expressing shMeox1. Immunostaining for cTnI and cTnT indicated that removal of Mef2c or Myocd led to dramatically lower reprogramming efficiency, whereas removing Gata4 or Tbx5 had no obvious negative effects on reprogramming efficiency (Figure 6A through 6F). As previously reported, Myocd is a potent booster of cardiac reprogramming,⁷ and has been shown to upregulate cardiac marker expression on mouse embryonic fibroblasts (MEFs) or murine neonatal CFs (nCFs) co-expressing Mef2c and Tbx5.⁵² In MFs expressing the Mef2c, Myocd, plus shMeox1 (MMshM) combination, without Gata4 and Tbx5, reprogramming efficiency was the highest among the 2 TF combinations (Figure 6G through 6K). FACS assessment showed that approximately 17% of MMshM-treated cells were cTnT⁺ iCMs (Figure 6N and 6O). Moreover, the MMshM conditions successfully induced cTnT, cTnI, and α-Actinin triple-positive iCMs (Figure 6L and 6M). However, we did not observe spontaneously beating cells in MMshM-produced iCMs, suggesting that MMshM promotes the initiation of CM cell fate but is not sufficient to generate functional iCMs. Taken together, our results demonstrate that, with removal of the endogenous barrier by Meox1 KD, Mef2c, and Myocd are sufficient to reprogram CFs into CM cell fate.

Figure 6.

Meox1 knockdown (KD) together with Mef2c and Myocd expression (MMshM) is sufficient to induce cardiac reprogramming. A through F, Immunostaining (A through E) and quantification (F) of cTnI⁺ and cTnT⁺ induced cardiomyocytes (iCMs) on shCtrl- or shMeox1-infected MFs expressing Gata4, Mef2c, Tbx5, and Myocd combinations, and with individual factors omitted from the MGT with Myocd (MGTM) cocktail (scale bar=100 µm; n=3). G through K, Immunostaining (H through K) and quantification (G) of cTnI⁺ and cTnT⁺ iCMs on shCtrl- or shMeox1-infected MFs expressing Gata4 + Myocd, Mef2c + Myocd, Tbx5 + Myocd, and Mef2c + Tbx5. Scale bar=100 µm (n=3). L and M, Immunostaining (L) and quantification (M) of cTnI⁺, cTnT⁺ and α-Actinin⁺ iCMs on shCtrl- or shMeox1-infected MFs expressing Mef2c and Myocd. Scale bar=100 µm (n=3). N and O, Representative flow plots (N) and quantification (O) for cTnT⁺ cells 14 days after Mef2c + Myocd and shMeox1 or shCtrl transduction on MFs (n=3). All data are presented as the mean±SEM. **P≤0.01 and ****P≤0.0001 vs the relevant control using the Student t test (M), 1-way ANOVA followed by Tukey’s multiple-comparisons test (O), and 2-way ANOVA followed by Bonferroni’s multiple-comparisons test (F and G). ns indicates not significant.

Meox1 KD Enables Robust Cardiac Reprogramming of MFs In Vivo Assessed by a Dual-Recombinase Lineage-Tracing System

Based on our above findings that Meox1 indeed functions as a molecular barrier for MF-to-CM reprogramming in vitro, we next examined the possible effects of applying shMeox1 to promote cardiac reprogramming in mice in vivo. For this purpose, we generated Tnni3-Dre;IR1;Postn-MCM^27,31 dual recombinase tracing mice to selectively label activated MFs with ZsGreen driven by the Postn promoter (Figure 7A). During heart development, resident cTnI⁺ cardiomyocytes are efficiently labeled with tdTomato via Dre recombinase under the control of the Tnni3 promoter. This process permanently removes LoxP-flanked ZsGreen elements, preventing any subsequent leakage of ZsGreen expression in resident cardiomyocytes upon Cre activation.²⁷ Subsequent tamoxifen (TAM)-induced Cre-LoxP recombination led to permanent labeling of Postn⁺ MFs and their descendants, whereas Rox ablation prevented tdTomato expression in response to Dre, even after MFs reprogramming into iCMs. Consistent with previous findings, few ZsGreen⁺ MFs can be observed in uninjured area of hearts³¹ (Figure S4H). We demonstrated that in situ injection of lentivirus (Fu-tet-EGFP and Fu-UbC-rtTA) into the infarcted region efficiently infects ZsGreen⁺ MFs (Figure 7C), indicating the effectiveness of in situ gene overexpression. Moreover, no ZsGreen⁺tdTomato^-cTnI⁺ CMs were detected in the MI and EGFP+shCtrl group (Figure 7D and 7E), indicating no spontaneous MF-to-CM transdifferentiation after MI. Additionally, the numbers of ZsGreen⁺tdTomato⁺cTnI⁺ fused CMs among MI, MGTM+shCtrl and MGTM+shMeox1 groups were comparable (Figure S4I and S4J).

Figure 7.

Dual recombinase lineage tracing demonstrates improved cardiac reprogramming efficiency under Meox1 knockdown (KD). A, Strategy for labeling resident cardiomyocytes and myofibroblasts (MFs) using Tnni3-Dre;IR1;Postn-MCM transgenic mouse. Resident Tnni3⁺ cardiomyocytes were efficiently labeled with tdTomato via Dre recombinase under the control of the Tnni3 promoter. Subsequent tamoxifen-induced Cre-LoxP recombination permanently labeled Postn⁺ MFs with ZsGreen. B, Schematic diagram illustrating experimental procedures. Dox indicates doxycycline; MI, myocardial infarction; and TAM, tamoxifen. SR, SB431542 (4 mg/kg/day), ruxolitinib (1 mg/kg/day). C, FU-tet-o-EGFP lentivirus injected into the border zone of an MI heart. Mice were euthanized 7 days after MI. Immunostaining for tdTomato, enhanced green fluorescent protein (EGFP), and ZsGreen in EGFP-overexpressing heart sections from a Tnni3-Dre;IR1;Postn-MCM mouse. Different pseudo-colors indicate second antibody signals separately: white, Alexa555 (anti-tdTomato); green, Alexa405 (anti-EGFP); and red, Alexa647 (anti-ZsGreen). White box in the merged picture indicating the area of high magnification images shown in the right panel. Scale bar=100 µm. D, Immunostaining of tdTomato, ZsGreen, and cTnI in MI heart sections from Tnni3-Dre;IR1;Postn-MCM mice. White boxes in the merged pictures indicate areas of high magnification images shown in the right panels. Scale bar=100 µm. E, Quantitative analyses of ZsGreen⁺tdTomato^- induced cardiomyocyte (iCM) numbers per section (n=5 independent biological replicates). F and G, Immunostaining of tdTomato, ZsGreen, and cTnI in MGT with Myocd (MGTM)+shCtrl (F), and MGTM+shMeox1 (G) heart sections from Tnni3-Dre;IR1;Postn-MCM mice. White boxes in the merged pictures indicate areas of high magnification images shown in the right panels. Scale bar=100 µm. H, Immunostaining of tdTomato, ZsGreen, and cTnI in MGTM+shMeox1 heart section from Tnni3-Dre;IR1;Postn-MCM mice. The white box in the merged picture indicates the area of high magnification shown in the right panel. Scale bar=200 µm. I, Schematic illustration of the echocardiography experimental procedures. J, Representative image of echocardiography showing left ventricular function 8 weeks after myocardial infarction. K, Left ventricular ejection fraction (LVEF) was quantified serially by echocardiography in mice treated with EGFP shCtrl (n=11), MGTM+shCtrl (n=12), and MGTM+shMeox1 (n=12) at 0 days, 5 days, 2 weeks, 4 weeks, and 8 weeks after MI. L, Quantification of LVEF in mice treated with EGFP+shCtrl (n=11), MGTM+shCtrl (n=12), and MGTM+shMeox1 (n=12) after 8 weeks (each point indicates independent biological replicates). M through O, Masson-trichrome staining of heart sections from EGFP+shCtrl (n=6), MGTM+shCtrl (n=6), and MGTM+shMeox1 (n=6) treated mice 6 weeks after MI surgery. The experimental timeline is shown in M; representative sections from each treatment group are shown in N; and quantification of scar size is shown in O. Heart fibrosis was evaluated at 4 levels (L1–L4) by Masson-trichrome staining. Scale bar=1 mm. All data are presented as the mean±SEM. *P≤0.05, **P≤0.01, ***P≤0.001, and ****P≤0.0001 vs the relevant control using 1-way ANOVA followed by Tukey’s multiple-comparisons test (E and L) and 2-way ANOVA followed by Tukey’s multiple-comparisons test (K) or Dunnett’s multiple-comparisons test (O). DAPI indicates 4’,6-diamidino-2-phenylindole; ns, not significant.

We next injected lentiviruses expressing different genes to test the efficiency of in vivo cardiac reprogramming from MFs. Lentivirus-containing solution was injected intramuscularly along the boundary of ischemic zone (Figure 7B). In the MGTM+shCtrl group, only a small number of ZsGreen⁺tdTomato^-cTnI⁺ iCMs were observed, indicating low efficiency of in situ MF-to-CM reprogramming (≈10.80 ZsGreen⁺tdTomato^-cTnI⁺ CMs/section; Figure 7E and 7F). In contrast, abundant clusters of ZsGreen⁺tdTomato^-cTnI⁺ iCMs were observed throughout the infarct and border zone of injured hearts of the MGTM + shMeox1 group, with an ≈5-fold increase in ZsGreen⁺tdTomato^-cTnI⁺ iCM counts, or ≈51.72 ZsGreen⁺tdTomato^-cTnI⁺ CMs/section (Figure 7E, 7G, and 7H). A combination of SB431542 (S; an ALK5 inhibitor) and ruxolitinib (R; a pan-JAK inhibitor) was used to enhance iCM reprogramming efficiency in vivo, as previously reported.²² Meox1 KD in the MGTM background significantly improved in vivo reprogramming efficiency. However, the addition of the SR combination led to a substantial further improvement in reprogramming efficiency (Figure S4L and S4M).

Because Postn⁺ MFs represent a subset of nonmyocytes, we further used Tnni3-Dre;IR1;R26-iCre marker-free lineage tracing to specifically label all nonmyocytes in the hearts (Figure S4A and S4B). Similarly, in situ injection of EGFP virus efficiently infects noncardiomyocytes (Figure S4C). No ZsGreen⁺tdTomato^-cTnI⁺ CMs were detected in MI and EGFP+shCtrl group (Figure S4C, S4D, and S4G). In the MGTM+shCtrl group, only a small number of ZsGreen⁺tdTomato^-cTnI⁺ iCMs were observed (≈21.6 ZsGreen⁺tdTomato^-cTnI⁺ CMs/section; Figure S4E). By contrast, Meox1 KD significantly improved reprogramming efficiency of MGTM, indicated by notably higher iCM cell counts, with 85.5 ZsGreen⁺tdTomato^-cTnI⁺ CMs/section, ≈4-fold more iCMs than that in the MGTM+shCtrl group (Figure S4F and S4G). qPCR analysis of infarcted tissue revealed that MGTM+shMeox1 treatment significantly increased the expression of cardiomyocyte-related genes and decreased fibrosis-related gene expression, compared with MGTM+shCtrl (Figure S4K). These data suggested that Meox1 was an important barrier for in vivo cardiac reprogramming.

We conducted echocardiography to determine whether cardiac reprogramming with Meox1 KD could improve heart function after MI. Serial echocardiography was performed on all mice before MI surgery and at 5 days, 2 weeks, 4 weeks, and 8 weeks after surgery. Measurements and analyses were performed in a blinded manner to ensure objectivity (Figure 7I). Echocardiography revealed that MGTM+shMeox1 significantly ameliorated left ventricular ejection fraction (LVEF) after standardized MI surgery (Figure 7J). MGTM+shMeox1 improved LVEF to ≈34.4% (≈23.3% in MGTM+shCtrl group) at 8 weeks after MI (Figure 7K and 7L). After blinded quantification on serial Masson’s trichrome-stained sections from 4 layers (L1–L4) spanning the infarct left ventricles (Figure 7M), we found that fibrosis area was reduced in both MGTM+shCtrl- and MGTM+shMeox1-treated mice, with a more pronounced reduction in the MGTM+shMeox1 group (Figure 7N and 7O). Additionally, a reduction in the ZsGreen⁺ cTnI^- cell area was observed under the MGTM+shCtrl/shMeox1 reprogramming condition (Figure S4R and S4S). Moreover, KD of Meox1 enabled highly efficient in situ reprogramming using only 3 transcription factors: Mef2c, Tbx5, and Myocd (Figure S4T and S4U). This achievement provides pivotal support for developing a minimized and more deliverable reprogramming factor cocktail for future clinical translation. Collectively, these results indicate that improving cardiac function after MI via cardiomyocyte reprogramming confers dual benefits: enhanced remuscularization and reduced fibrosis.

To assess in vivo Meox1 KD efficiency, we generated a lentivirus vector containing a U6-driven shRNA and a UbC-driven mScarlet-H2B reporter, allowing selective labeling of infected CFs (Figure S4N and S4O). Five days after MI, mScarlet⁺ fibroblasts were isolated by FACS (Figure S4P), and qPCR analysis revealed significantly reduced Meox1 mRNA levels in shMeox1-expressing cells (Figure S4Q), confirming effective in vivo KD.

In summary, Meox1 KD overcomes the reprogramming resistance in MFs, enabling robust in vivo MF-to-CM conversion even after MI, a finding stringently validated by dual-recombinase tracing. The resultant improvement in cardiac function and attenuation of fibrosis with MGTM+ shMeox1 treatment underscores its direct translational potential.

MEOX1 KD Greatly Improves Cardiac Reprogramming Efficiency in Human Fibroblasts

To investigate the role of MEOX1 in human cardiac reprogramming, we employed iPSC-derived CFs (iPSC-CFs), a model that recapitulates key features of primary human CFs.⁵³ We first induced iPSC differentiation into epicardial cells (EPCs),⁵⁴ and then switched to the fibroblast differentiation medium including FGF2 (fibroblast growth factor) and TGF-β inhibitor SB431542 (Figure 8A). As reported,⁵³ iPSC-CFs showed a typical fibroblast morphology and expressed CF marker proteins (Figure S5A) and mRNAs (Figure S5C). Most importantly, immunostaining showed that iPSC-CFs did not express cardiomyocyte markers (Figure S5B), indicating the absence of cardiomyocyte contamination in differentiated CFs, and further highlighting that iPSC-CFs may serve as a relatively convenient source of cells suitable for human cardiac reprogramming. To model injury-induced activation, we treated iPSC-CFs with TGF-β1. RNA-seq comparing iPSCs, iPSC-CFs, and activated iPSC-MFs confirmed that TGF-β1 upregulates fibrosis-related genes in iPSC-CFs (Figure 8B). GO analysis showed that genes upregulated in iPSC-MFs were mainly enriched in extracellular matrix organization, vasculature development, cell adhesion, cell migration, and response to wounding, all of which are related to profibrotic phenotypes. In addition, typical fibrosis-related genes (COL1A1, POSTN, ACTA2, and FN1) were also highly expressed in iPSC-CFs after stimulation by TGF-β1 (Figure 8C). Consistent with the murine data, MEOX1 KD in iPSC-MFs significantly suppressed fibrosis-associated genes (Figure 8D), which in turn enhanced cardiac reprogramming efficiency under MGTM conditions (Figure 8E and 8F).

Figure 8.

MEOX1 knockdown (KD) improves cardiac reprogramming efficiency in human fibroblasts. A, Schematic diagram showing the protocol for differentiating cardiac fibroblasts (CFs) and myofibroblasts (MFs) from induced pluripotent stem cells (iPSC-CFs/iPSC-MFs). B, Heatmap showing differentially expressed genes (DEGs) identified through RNA-seq analysis in iPSC, iPSC-CF, and iPSC-MF. Representative genes and gene ontology terms enriched in iPSC (blue), iPSC-CF (green) and iPSC-MF (pink) were listed. C, Expression levels of fibrosis marker genes in RNA-seq data from iPSCs, iPSC-CFs, and iPSC-MFs (n=3). D, qPCR of MEOX1 and fibrosis gene expression in iPSC-CFs infected with MEOX1 shRNA or scramble control (shCtrl) under transforming growth factor-beta 1 (TGF-β1; 10 ng/mL) treatment. Expression value in shCtrl infected cells was set to 1 (n=3). E and F, Immunostaining (E) and quantification (F) of cTnI⁺cTnT⁺ induced cardiomyocytes (iCMs) on shCtrl- or shMEOX1-infected iPSC-MFs expressing MGT with Myocd (MGTM). Scale bar=200 µm (n=3). G, qPCR of MEOX1 gene expression in human CFs under control (Ctrl) or TGF-β1 (10 ng/mL) treatment. Expression value in Ctrl cells was set to 1 (n=3). H, qPCR of MEOX1 gene expression in human CFs infected with shCtrl or shMEOX1 under TGF-β1 (10 ng/mL) treatment. Expression value in shCtrl infected cells was set to 1 (n=3). I and J, Immunostaining (I) and quantification (J) of cTnI⁺ iCMs on shCtrl- or shMEOX1-infected human CFs expressing MGT+miR-133a. Scale bar=100 μm (n=3). K, Schematic of the model for the roles of MEOX1 on CF activation and cardiac reprogramming under MI surgery conditions. MI injury induces TGF-β1 and IL-1β secretion, which leads to the activation of CFs and the upregulation of MEOX1, converting them into an MF state. This fibroblast activation hinders cardiac reprogramming, whereas MEOX1 KD removes this barrier and enables more efficient cardiac reprogramming, or reprogramming with fewer TFs. MGT, polycistronic Mef2c, Gata4 and Tbx5. MGTM indicates polycistronic MGT with Myocd; MM, Mef2c and Myocd; and MTM, Mef2c, Tbx5, and Myocd. All data are presented as the mean±SEM. *P≤0.05, **P≤0.01, ***P≤0.001, and ****P≤0.0001 vs the relevant control using the Student t test. ns indicates not significant.

To more precisely evaluate the effects of MEOX1 KD in a clinically relevant model, we performed cardiac reprogramming experiments using primary human cardiac MFs under MGT+miR-133a conditions. Notably, MEOX1 KD significantly enhanced reprogramming efficiency (Figure 8G through 8J), demonstrating its conserved role in modulating reprogramming outcomes across species. These data indicate that, as in mice, human MFs also exhibit reprogramming resistance, which can be reversed by MEOX1 KD. This establishes MEOX1 as a cross-species barrier to cardiac reprogramming, highlighting its potential as a therapeutic target for clinical translation.

DISCUSSION

This study sought to resolve a central challenge: how to achieve high-efficiency cardiac reprogramming within the infarcted heart. During this exploration, we uncovered a critical impediment: the most desirable cellular source for reprogramming, MFs, ironically exhibited the strongest resistance to conversion. To overcome this impediment, we elucidated the molecular basis of this resistance and identified a key endogenous barrier whose removal enables efficient in situ cardiac regeneration. We pinpointed MEOX1 (Mesenchyme homeobox 1), a known fibrosis-driving transcription factor downstream of TGF-β/IL-1β signaling,^36,37 as a primary endogenous barrier that blocks iCM reprogramming from injury-activated MFs, both in vitro and in vivo (Figure 8K).

To overcome the reprogramming resistance of MFs, we systematically evaluated MEOX1 KD. This intervention substantially enhanced cardiac reprogramming efficiency in vitro and, critically, in vivo. Using stringent dual-recombinase lineage tracing, we demonstrated that Meox1 KD significantly boosts MF-to-iCM conversion within the infarct, leading to improved cardiac function and attenuated fibrosis. These findings represent the first report of successful MEOX1 loss-of-function in this context. Our results also confirm that MEOX1 overexpression potently suppresses reprogramming in vitro, consistent with a previous report.²⁹

These previous observations provide an important foundation for our study, but they address a related rather than identical question. The observation that forced MEOX1 expression suppresses cardiomyocyte marker induction demonstrates that high MEOX1 levels are incompatible with efficient reprogramming. However, it does not establish whether endogenous, disease-associated MEOX1 induction is required for the low reprogramming competence of injury-activated MFs, nor whether Meox1 KD is sufficient to improve MF-to-iCM reprogramming in vivo. In our system, even in fibroblasts isolated from infarcted hearts, MEOX1 overexpression increased Meox1 mRNA expression by ≈98-fold (Figure 2O), whereas TGF-β stimulation increased Meox1 expression by only ≈17-fold under matched conditions (Figure 5A). This distinction indicates that forced overexpression and endogenous pathological induction represent different biological contexts. Building on previous work, we therefore tested whether purified injury-activated MFs are intrinsically less permissive than quiescent CFs under matched reprogramming conditions, whether endogenous MEOX1 contributes to this resistance, and whether direct MEOX1 KD is sufficient to overcome this barrier and achieve efficient MF-to-iCM conversion in the infarcted heart.

These findings position our work as an extension of, rather than an alternative to, previous strategies aimed at improving cardiac reprogramming in the infarcted heart. Previous studies have enhanced the reprogramming of MI-derived CFs by strengthening exogenous cardiogenic inputs, including MYOCD/SALL4-based factor combinations.^7,22 Other work has shown that IFN signaling from macrophages and T cells in the infarcted region acts as an extrinsic microenvironmental barrier to CF-to-iCM reprogramming.^16,35 Our study addresses a distinct, cell-intrinsic layer: the transcriptional state of the target fibroblast population itself. Injury-activated MFs are not merely passive substrates exposed to insufficient cardiogenic inputs or extrinsic immune inhibition; they actively retain disease-associated regulatory programs that stabilize MF identity and oppose cardiogenic fate acquisition. Efficient in vivo cardiac reprogramming may therefore require a combined strategy that simultaneously strengthens pro-cardiogenic inputs, alleviates inhibitory microenvironmental signals, and relieves the endogenous barriers that maintain the activated MF state. By identifying MEOX1 as one such intrinsic barrier and showing that its KD restores MF reprogramming competence, our work contributes the barrier-relief arm of this framework.

Beyond substantially improved reprogramming efficiency, MEOX1 KD enabled successful reprogramming with simplified cocktails of factors, even in vivo. Previous work has shown that GATA4 overexpression can attenuate TGF-β1–mediated CF activation and suppress Meox1 expression.³⁰ Notably, in post-MI MFs, we found that MEOX1 KD via shRNA further reduced its mRNA and protein levels even in the presence of overexpressed GATA4, leading to a marked increase in reprogramming efficiency. Importantly, after MEOX1 KD, we achieved robust reprogramming using fewer factors: only two transcription factors (Mef2c+Myocd) in vitro and 3 (Mef2c+Tbx5+Myocd) in vivo. These findings offer a highly promising solution to a major obstacle in clinical translation, the challenge of gene delivery, by demonstrating that removing endogenous barriers can reduce the number of required transcription factors, thereby minimizing the total genetic payload. Future investigation on delivery methods and long-term safety studies will be essential for successful clinical translation.

Our findings highlight that MEOX1 serves as a node between the mutually antagonistic processes of cardiac fibrosis and cardiac reprogramming. scRNA-seq analysis of transcriptional networks between fibrotic MFs and iCMs revealed that MEOX1 is a potent stabilizer of MF identity that must be overcome to induce MF-to-iCM conversion. These results align well with a recent report that showed delivery of cardiac reprogramming factors (ie, GATA4) in vivo leads to MEOX1 downregulation³⁰ and anti-fibrotic effects.⁵⁵ Thus, targeting MEOX1 represents an ideal strategy that concurrently mitigates pathological fibrosis and unlocks the reprogramming potential of MFs. However, it should be emphasized that the timing of anti-fibrotic and reprogramming interventions requires careful consideration, as depleting activated fibroblasts too early after MI may increase the risk of ventricular rupture and mortality.³¹

Our study specifically addressed the unique post-MI microenvironment, seeking to identify and overcome the reprogramming barriers imposed by fibroblast heterogeneity. This work identifies Meox1 as the principal mediator through which TGF-β and IL-1β exert their inhibitory effects on MF-to-iCM reprogramming. However, the infarct microenvironment is complex and likely governed by a broader array of factors beyond these cytokines, potentially including various MI-associated soluble factors and metabolites.^56,57 Future work will require extensive analysis to determine whether and how other microenvironment factors may influence in situ reprogramming efficiency, such as stress-associated Calr expression and the downstream calcium signaling activity, as revealed by Perturb-seq in vivo.⁵⁸ It was previously reported that the combination of injury-enriched microenvironmental factors including FGF2,⁵⁹ VEGF⁶⁰ and cardiac repair factor FGF10⁶¹ (FFV combination) can significantly enhance cardiac reprogramming in vitro,⁶² whereas inflammation-associated factors, such as C-C motif chemokine ligand 3 (CCL3), CCL6, and CCL17, inhibit iCM formation in vitro,⁶³ indicating that exogenous factors may activate conflicting regulatory mechanisms in the injury microenvironment. Macrophage-derived interferon-β activates IFN-β–IFNAR–STAT1 signaling in MFs to impair GATA4 chromatin binding, and macrophage-directed T-cell recruitment further augments this IFN-mediated suppression of cardiac reprogramming.^16,35 These findings highlight a synergistic interplay between CF-to-iCM reprogramming and microenvironmental optimization. Alleviating inhibitory signals within the microenvironment not only improves reprogramming efficiency but also facilitates the restoration of cardiac function.

Whether MEOX1 KD could also enhance the reprogramming efficiency of MFs in other organs remains an open and compelling question for future investigation. MEOX1, a mesodermal homeodomain transcription factor,⁶⁴ plays a well-established role in somitogenesis during embryogenesis⁶⁵ and contributes to endothelial cell senescence by transcriptionally activating p16(INK4a) and p21(CIP1/WAF1).⁴⁶ It also acts as a positive regulator of TGF-β–induced smooth muscle cell differentiation.⁵¹ Recent studies, including our own data (Figure 5), firmly establish MEOX1 as a master regulator of CF activation, during which its KD attenuates TGF-β/IL-1β–driven differentiation into MFs.³⁶ In light of findings that TGF-β treatment leads to MEOX1 upregulation in human lung, liver, and kidney fibroblasts and in human idiopathic pulmonary fibrosis,^36,66 future work will necessarily examine whether Meox1 functions in maintaining MF identity under pathological conditions, thereby conferring resistance to in situ direct reprogramming in other organs.

This study addresses a key aspect of improving in situ cardiac reprogramming efficiency: whether and why MFs exhibit strong reprogramming resistance and how to overcome it. Furthermore, future work is needed to develop more efficient and specific delivery systems for targeting cardiac MFs, as well as optimizing the dosage and timing of reprogramming factor expression to maximize tissue repair outcomes.

ARTICLE INFORMATION

Acknowledgments

The authors thank Li Qian (University of North Carolina, Chapel Hill) for providing the pMXs-MGT plasmid and the detailed cardiac reprogramming protocol. They thank Dr Jeffery D. Molkentin in Cincinnati Children’s Hospital Medical Center for providing Postn-MerCreMer mice as a gift. The authors thank Nan Sheng, Yang Liu, Ruilin Xu, Xin Dang, Hao Wang, Hong Zhao, Yanmeng Tao, Junbo Yang, Xinxin Nian, Tianyu Zhang, and all members in Yang Zhao’s laboratory for technical assistance. They thank the National Center for Protein Sciences at Peking University in Beijing, China, for assistance with Nikon A1R/AXR confocal microscope and for providing shRNA plasmids. They also thank the flow cytometry Core at National Center for Protein Sciences at Peking University, particularly Huan Yang for technical help.

Disclosures

Y.Z. is a shareholder of Plastech Pharmaceutical Technology Co, Ltd.

Supplemental Material

Supplemental Methods

Tables S1–S3

Figures S1–S5

Figure Legends S1–S5

Movies S1–S6

References 67–86

Supplementary Material

cir-154-1806-s001.pdf ^{(8.9MB, pdf)}

cir-154-1806-s002.pdf ^{(104.8KB, pdf)}

Download video file ^{(15.4MB, mov)}

Download video file ^{(9.7MB, mov)}

Download video file ^{(9.9MB, mov)}

Download video file ^{(7.5MB, mov)}

Download video file ^{(46.9KB, mp4)}

Download video file ^{(36.4KB, mp4)}

Funding Statement

This work is supported by the National Natural Science Foundation of China (Y.Z.; 32470882), the National Natural Science Foundation of China (J.W.; 82400328), the China Postdoctoral Science Foundation (J.W.; 2024M760091 and BX20240021), the National Key Research and Development Program of China (Y.Z.; 2018YFA0800504 and 2025YFA0922000), Clinical Medicine Plus X - Young Scholars Project, Peking University, the Fundamental Research Funds for the Central Universities (J.W.) and funding provided by Plastech Pharmaceutical Technology Co, Ltd. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.