Regeneration of infarcted mouse hearts by cardiovascular tissue formed via the direct reprogramming of mouse fibroblasts

Abstract

Fibroblasts can be directly reprogrammed into cardiomyocytes, endothelial cells or smooth muscle cells. Here, we report the reprogramming of mouse tail-tip fibroblasts simultaneously into cells resembling these three cell types via the microRNA-mimic miR-208b-3p, ascorbic acid and bone morphogenetic protein 4, as well as the formation of tissue-like structures formed by the directly reprogrammed cells. Implantation of the formed cardiovascular tissue into the infarcted hearts of mice led to the migration of reprogrammed cells into the injured tissue, reducing regional cardiac strain and improving cardiac function. The migrated endothelial cells and smooth muscle cells contributed to vessel formation, and the migrated cardiomyocytes, which initially displayed immature characteristics, became mature over time and formed gap junctions with host cardiomyocytes. Direct reprogramming of somatic cells to make cardiac tissue may aid the development of applications in cell therapy, disease modelling and drug discovery for cardiovascular diseases.

One-sentence editorial summary (to appear right below the title of your Article on the journal’s website): Mouse fibroblasts can be simultaneously directly reprogrammed into cardiomyocytes, endothelial cells and smooth muscle cells to form cardiac-tissue structures that improve cardiac function on implantation into infarcted mouse hearts.

Heart disease is the leading cause of mortality worldwide. After injury, human postnatal hearts do not recover appropriately, as CMs have limited proliferation capacity¹. As an emerging cardiac regenerative therapy, direct reprograming of somatic cells into CM-like cells was developed by introducing a combination of cardiac transcription factors (TFs) or cardiac microRNAs (miRNAs)^2–6. These cells are referred to as induced CMs (iCMs) or cardiac-like myocytes (iCLMs). While direct viral delivery of TFs or miRNAs into heart was shown to be effective for cardiac repair^2,4–6, the use of retro- or lenti-viruses for reprogramming limits their clinical application. Thus, we thought that cell therapy with directly reprogrammed cells generated with non-integration vectors would be a better option. However, this possibility has not been investigated previously.

In this type of cell therapy strategy, poor cell survival may limit the therapeutic effects, as studies have shown that cell engraftment and survival are low in any cells transplanted into the heart^7–9. Thus, to enhance the survival of the implanted cells and improve cardiac repair, various biomaterials have been used in the form of an injectable hydrogel¹⁰, or engineered tissue or patch¹¹. However, most of the tissue-patch approaches showed little or no migration of implanted cells from the patch into the host heart^11–13. As another approach, adding other cells such as ECs, SMCs, or fibroblasts to the transplanted stem-cell-derived CMs was shown to improve cell survival and enhance cardiac regenerative effects^14,15. To date, the effects of transplantation of induced- or directly reprogrammed cells together with biomaterial or co-transplantation of multiple directly reprogrammed cells on cardiac regeneration has not been explored.

In this study, we sought to address the effects of cell therapy on cardiac regeneration using multiple cardiovascular cells together with extracellular matrix (ECM) generated by a direct reprogramming approach. We first aimed to reprogram fibroblasts into CMs with cardiac microRNAs miR-1, miR-208a, miR-208b, and miR-499, which are known to regulate cardiac development¹⁶. We used miRNAs due to their potential advantages over TFs. First, as reprogramming involves changes in expression of numerous genes, and miRNAs regulate a large group of mRNAs simultaneously, miRNAs would be more efficient reprogramming inducers. Second, we expected potential multicellular reprogramming effects for the above miRNAs, as they are involved in cardiac, not only CM, development. Finally, the availability of miRNA mimics avoids viral use. As such, cultured fibroblasts were treated with individual or all the above miRNA mimics to generate CMs. However, we found miRNA alone was insufficient to generate functional CMs. To enhance reprogramming, we added AA and BMP4, which were reported to augment reprogramming and induce cardiac differentiation^17–20, to the single best miRNA selected from our experiments, miR-208b-3p. These regimens generated not only functional CM-like cells but also EC- and SMC-like cells, and produced extracellular matrices (ECMs) as well, forming three-dimensional (3D) tissue-like structures. We referred to these structures as reprogrammed cardiovascular tissue (rCVT) and the contained cells as reprogrammed CMs (rCMs), ECs (rECs), and SMCs (rSMCs). Implantation of this rCVT onto the infarcted mouse heart reduced regional cardiac strains and improved cardiac function. Histological examination showed migration of reprogrammed cells from rCVT into the infarcted hearts. Migrated rECs and rSMCs contributed to vessel formation together with host vascular cells. Migrated rCMs initially displayed immature characteristics but became mature over time and formed gap junctions with host CMs, becoming indistinguishable from host CMs.

Here, we report direct reprogramming of adult somatic cells into multiple cardiovascular cells and even into cardiac tissue-like structures, rCVT, and the favorable effects of this rCVT on cardiac repair after transplantation onto infarct heart. This direct tissue-reprogramming strategy would be useful for regenerative medicine by providing essential cardiovascular cells and natural matrix at the same time. This approach can circumvent the complicated processes required for stem cell generation, their differentiation into essential target cells, and addition of biomaterials.

Results

Generation of a tissue-like structure containing CM-like cells by reprogramming of fibroblasts with miRNA 208b, AA, and BMP4

To investigate whether cardiac miRNAs are able to directly reprogram fibroblasts toward CM-like cells, we administered miRNA mimics miR-1a-2–5p, miR-208a-3p, miR-208b-3p, miR-208b-5p, and miR-499–5p, either alone or all, into cultured fibroblasts and examined expression of CM genes (Supplementary Fig. 1a and Supplementary Table 1). Quantitative real-time RT-PCR (qRT-PCR) analyses demonstrated that each of the miRNA mimics increased mRNA expression of CM genes including Mef2c, Nkx2–5, Gata4, and Myh6, but not a mesodermal gene, Mesp1. A combination of all five miRNAs did not have an additive effect on gene expression, and miR-208b-3p alone was the most effective. Immunocytochemistry confirmed expression of a CM marker, sarcomeric α-actinin (ACTN2) in miRNA-transfected fibroblasts at a low frequency (Supplementary Fig. 1b). Moreover, ACTN2 expression was diffuse and sarcomeres were poorly organized in any miRNA-transfected MTTFs. These results suggested that cardiac miRNA alone can induce CM gene and protein expression at low levels, but is insufficient to generate functional CMs.

To enhance the reprogramming efficiency and generate functional CMs, we selected a single miRNA, miR-208b-3p, which resulted in the best gene expression in the above experiments, and added AA and BMP4, which are known to augment reprogramming and/or cardiomyogenic differentiation of stem cells^17–20, one day after miR-transfection. We referred to this combined treatment of miR-208b-3p, AA, and BMP4 as “MAB.” Next, we determined the reprogramming capability of MAB at day 6 after the combined treatment (Fig. 1a–f and Extended Data Fig. 1). For these experiments, we used MTTFs derived from MYH6 promoter-driven mCherry transgenic mice (MYH6-mCherry-MTTFs) (Fig. 1a) to visualize cell conversion into CM-lineages. In the MTTFs treated with MAB, bright mCherry⁺ cells were observed, suggesting conversion of MTTFs into CM-lineage cells. However, in MTTFs cultured under basal conditions or treated with scramble-miR, AA, and BMP4, mCherry⁺ cells were not seen. qRT-PCR analyses confirmed increased mRNA expression of cardiac transcription factors Gata4, Nkx2–5, Hand1/2, and Tbx20, and structural/contractile proteins including Tnni3, Myl2 and Myh6/7, but not Myl7 (an atrial marker) (Extended Data Fig. 1). AA and/or BMP4 alone also increased expression of a few CM genes, but a combination of MAB was the most consistent and effective for the induction of CM gene expression. The ratio of Myh6 to Myh7 measured by qRT-PCR was lower in MAB-treated MTTFs compared to both adult and fetal hearts (Supplementary Fig. 1c). Collectively, these data indicated low maturation status of reprogrammed cells. Next, we investigated the mRNA expression of fibroblast marker genes including Thy1, Vim, S100a4 and Ddr2 before and after MAB treatment. qRT-PCR analysis showed reduced expression of all these fibroblast genes after MAB treatment (Supplementary Fig.1d). The percent of fibroblasts expressing VIM and THY1 was reduced from 99% and 98% in MTTFs to 36% and 46% in rCVT, respectively (Supplementary Fig. 1e–f). These data highlight suppression of fibroblast marker expression during the reprogramming toward CMs. Neither MAB nor individual components of MAB increased mRNA expression of pluripotent stem cell markers (Pou5f1, Sox2 and Nanog) (Supplementary Fig. 1g).

Fig. 1 |. Reprogramming of fibroblasts into cardiomyocyte-like cells and a tissue-like structure.

a, Fluorescence images of MYH6-mCherry-MTTFs following no treatment (1^st row); treatment with scramble-miR, AA and BMP4 (2^nd row); and MAB (3^rd row). b-c, Immunofluorescence staining for TNNT2 in WT-MTTFs (b) and ACTN2 in Nkx2–5-GFP-MTTFs (c) after MAB treatment. d, Action potentials (APs) in spontaneously contracting cells at day 6. N = 19; Resting membrane potential = −55.3 ± 1.0 mV; Amplitude of action potential = 86.3 ± 2.8 mV; Maximum upstroke slope = 201.0 ± 26.3 mV/ms. e, Calcium oscillations measured by Fluo-4 (GFP) in physically connected CM-like cells. (See also Supplementary Movie 2). f, Transmission electron microscopic (TEM) images of MTTFs and rCMs. rCMs showed thick sarcomeric structures. Z, A, I and M indicate Z-lines, A-band, I-band and mitochondria, respectively. g, Representative photographs of MTTFs before and after MAB treatment. The picture on the right shows an enlarged image of rCVT which was detached and moved to a smaller dish. h, Immunofluorescence images of TNNT2⁺ cells at day 10. i-j, Flow cytometry analyses of TNNT2⁺ cells in MTTFs(i) and rCVT(j). Cells incubated with IgG-Isotype were used as negative controls. Data are mean ± SEM. n = 3.

Immunofluorescence staining for MAB-treated fibroblasts confirmed expression of TNNT2 and showed organized sarcomeres (Fig. 1b). Nkx2–5 and ACTN2 expression was clearly observed in MTTFs derived from Nkx2–5 promoter-driven GFP transgenic mice (Nkx2–5–MTTFs) (Fig. 1c). Immunofluorescence staining for TNNT2 revealed various morphologies of rCMs (Supplementary Fig. 2a–b). rCMs were categorized into six types according to the cell size, ratio of length to width, and presence of sarcomeric striations (Supplementary Fig. 2a). Two types of rCMs, designated as irregular and non-striated round types, expressed TNNT2, but not in a striated pattern. The other four types showed sarcomeric striations. Among all rCMs, only non-striated round type, the smallest cells, expressed Ki67, but the proliferation capacity appeared limited as they did not form a colony-like structure.

Moreover, spontaneously contracting cells began to emerge between 4–6 days after MAB treatment (Supplementary Movie S1). Single cells were dissociated enzymatically at day 6, and were patch-clamped to examine their action potential (AP) morphology. Most spontaneously beating cells exhibited APs that were commensurate with their immature state with depolarized maximum diastolic potentials and nodal-like AP morphology (Fig. 1d). Similarly, the dissociated single cells exhibited intracellular Ca²⁺ transients that appeared to be of immature CMs with slow rise in Ca²⁺ transients (Fig. 1e, Supplementary Movie 2). Repeated attempts to dissociate and record APs and Ca²⁺ transients from single cells at a later stage of reprogramming failed to yield viable CMs, likely due to the extensive development of the extracellular matrix and compaction of the de novo tissue construct as mentioned below. However, transmission electron microscopy (TEM) demonstrated cardiomyogenic features of reprogrammed cells such as sarcomeric structures with clear Z-lines, distinguishable A- and I-bands, and abundant mitochondria aligned between sarcomeres (Fig. 1f). With these CM-like cell features, we referred these cells as rCMs.

Between days 6 to 10, we found a growing deposition of extracellular matrices (ECMs), converting monolayer cells to a three-dimensional (3D) tissue-like patch (Fig. 1g). We referred to this structure as reprogrammed cardiovascular tissue (rCVT) because it includes three-dimensional ECM (Supplementary Fig. 3a) and other cardiovascular cells such as ECs and SMCs as shown below. This rCVT remained attached unless detached intentionally with tweezers (Fig. 1g, right panel) and had tight integration of cells and ECM. In the rCVT, clearly striated TNNT2⁺ cells were observed after immunostaining (Fig. 1h). TNNT2⁺ cells comprised ~17% of the cells within rCVT by flow cytometric analysis, which was performed after digestion of rCVT (Fig. 1i–j). To analyze ECM components, rCVT was decellularized at day 10 and subjected to liquid chromatography-mass spectrometry (LC-MS) (Supplementary Fig. 3b). In our rCVT, ~51% of the matrix consisted of collagen isoforms with collagen type I, and Prolargin, a collagen type 1 binding protein, being the most abundant. Formation of 3D structure increased cell density but decreased cell proliferation capacity (Supplementary Fig. 3c–e). These data demonstrated that miR-208b-3p, together with AA and BMP, reprogrammed MTTFs toward a cardiac tissue-like structure, rCVT, containing functional CM-like cells, rCMs, and ECMs.

Simultaneous reprogramming of fibroblasts toward vascular cells

During the culture of fibroblasts with MAB, we also observed the emergence of multi-branched structures, similar to vascular networks (Fig. 2a). To examine the vascular nature, we stained for CDH5 (VE-CADHERIN), an EC marker, and confirmed its expression (Fig. 2b and Supplementary Fig. 4a). We then examined mRNA expression of EC markers Cdh5, Pecam1, Nos3, Kdr, and Tek (Fig. 2c) in rCVT and confirmed increased expression of all the examined EC genes compared to the control fibroblasts. However, AA and BMP4 without miR-208b showed increased expression of only Tek and Nos3, and miR-208b alone, only Tek (Fig. 2c). Immunofluorescence staining of rCVT further confirmed expression of PECAM1 (Fig. 2d) and TEK (Fig. 2e). The EC-like cells in rCVT also took up Dil-conjugated acetylated low density lipoprotein (Dil-Ac-LDL) (Fig. 2f) and reacted with DAF-FM diacetate (Fig. 2g), an indicator of nitric oxide production. Flow cytometry analyses of dissociated rCVT showed that ~ 29% of cells reacted with DAF-FM diacetate (Supplementary Fig. 4b) and ~ 28% of the cells expressed PECAM1 (Fig. 2h–i). Thus, we referred to these EC-like cells as reprogrammed ECs (rECs). These networks composed of rECs emerged distinctively from rCMs. These results suggest simultaneous reprogramming toward cardiomyogenic and EC-lineage cells (Supplementary Fig. 4c–d). Next, we investigated whether the rCVT also contained SMC-like cells. qRT-PCR analysis showed increased mRNA expression of SMC markers such as Myocd, Myh11, and Acta2. Treatment of MTTFs with miR-208b alone increased only Acta2 expression (Fig. 3a). Immunostaining for ACTA2 (or α-smooth muscle actin), confirmed its expression in network structures (Fig. 3b). Flow cytometry analyses further demonstrated ~19% SMTN⁺ cells (Fig. 3c–d). Taken together, these data indicate that MAB induced reprogramming of fibroblasts into rECs and rSMCs, which formed vessel-like structures in rCVT.

Fig. 2 |. Reprogramming toward endothelial cells.

a, A phase contrast image showing vascular structures in MTTFs treated with MAB for 10 days. b, A confocal three-dimension fluorescence image of rCVT stained for CDH5 and DAPI, showing vascular networks. c, mRNA expression of EC genes after the indicated treatment. Ten independent experiments, each with technical replicates. Expression in MTTFs was set as 1. ***P < 0.001: MAB vs MTTF, AA, BMP4, AA/BMP4. Statistical analysis was performed using one-way ANOVA with Bonferroni’s multiple comparison test. Data are mean ± SEM. d-e, Fluorescence microscopic images of rCVT for PECAM1 (d) and TEK (e). f-g, Phase-contrast and fluorescence microscopic images of MTTFs and rCVT after incubation with Dil-Ac-LDL (f) or DAF-FM Diacetate (NO indicator) (g). h-i, Flow cytometry analyses of PECAM1⁺ cells in MTTFs (h) and dissociated rCVT-derived cells (i). Non-stained cells were included as controls. Data are mean ± SEM. n = 3.

Fig. 3 |. Reprogramming toward smooth muscle cells.

a, mRNA expression of SMC genes after the indicated treatment. Ten independent experiments, each with technical replicates. Expression in MTTFs was set as 1. ***P < 0.001: MAB vs MTTF, AA, BMP4, AA/BMP4 and miR-208b. Statistical analysis was performed using one-way ANOVA with Bonferroni’s multiple comparison test. Data are mean ± SEM. b, Phase-contrast and fluorescence microscopic images of ACTA2 in rCVT. c-d, Flow cytometry analyses of SMTN⁺ cells in MTTFs (c) and dissociated rCVT-cells (d). IgG-Isotype treated cells were used as controls. Data are mean ± SEM. n = 3. e, Double immunofluorescence staining for SMC (ACTA2) and EC (PECAM1) markers. f-g, Double immunofluorescence staining for a SMC marker (SMTN or SM22α) and an EC marker (PECAM1 or CDH5).

Interestingly, we also found that some cells expressed markers of more than two cell types. Double immunofluorescence staining demonstrated that some of the TNNT2⁺ rCMs exhibited low expression of CDH5 (Supplementary Fig. 4d, bottom). We also found some cells expressing PECAM1 and ACTA2 (Fig. 3e and Supplementary Fig. 4e). The ACTA2 expression in such rECs appeared speckled and not as typical striated SMC-like patterns. Again, some cells expressed both SMC (SMTN and SM22α) and EC markers (PECAM1⁺ or CDH5) ), in which low expression of SMC markers were found (Fig. 3f–g). However, we did not observe any cells expressing all three cell markers (TNNT2, PECAM1 and ACTA2) (Supplementary Fig. 4f).

Transcriptome analysis showing reprogramming of fibroblasts into cardiac cells

To investigate global transcriptome changes during reprogramming of fibroblasts to CM-, EC-, or SMC-like cells, we conducted RNA-seq using total RNAs from MTTFs extracted at days 0 (D0), 6 (D6) and 10 (D10) after MAB treatment and fetal heart. These data were also compared with public data sets of primary EC (SRP285778)²¹, CM (SRP057984)²² and SMC (SRP253689). Principal component analysis (PCA) demonstrated that the reprogrammed cells acquired similar gene expression patterns with primary EC, CM, and SMC (Extended Data Fig. 2a). To analyze the activation of cell type-specific genes and heart-related genes in the reprogrammed cells, we set gene signatures of the heart, EC, SMC and CM as those genes exhibiting 2-fold gene expression changes (Supplementary Table 2) and performed gene set enrichment analysis (GSEA). The results showed significant enrichment of gene signatures for heart, EC, CM, and SMC in D6 and D10 reprogrammed cells relative to D0 fibroblasts (FDR < 0.001), suggesting simultaneous reprogramming of fibroblasts toward cardiac cells (Extended Data Fig. 2b, Supplementary Table 2). We then examined changes of fibroblast genes during the reprogramming. While most of the general or cardiac fibroblast genes were downregulated including Ddr2, Wt1, Thy1, Vim, and S100a4, some genes known to be specific for cardiac myofibroblasts (or activated fibroblasts) were upregulated such as Acta2, Agtr1a, Tnc, Tgfb2, Cdh11, Postn, and Col1a1 (Extended Data Fig. 2c), suggesting potential activation of the cardiomyofibroblast gene program in reprogrammed cells^23,24.

We further performed single-cell RNA-seq (scRNA-seq) on MTTFs (11,580 cells) and MTTFs treated with MAB for 6 days (12,261 cells) to explore detailed transcriptome changes during the transition from fibroblasts toward cardiac cells including CM-, EC- and SMC-like cells. In MTTFs, the mean reads and median genes per cell were 53,107 and 5,096, respectively; in reprogrammed cells (Day 6), the mean reads and medium genes per cell were 100,515 and 6,164, respectively. Uniform Manifold Approximation and Projection (UMAP) clustering showed ten clusters of MTTFs (Fig. 4a), showing heterogeneity in gene expression patterns (Supplementary Fig. 5 and b), and eighteen clusters of reprogrammed cells (Fig. 4b). After reprogramming, we used the following gene sets for identifying cardiac cell types: Hand2, Nkx2–5, and Tnnt2 for CMs; Hand2, Nkx2–5, and Tnnt2 for SMCs; Pecam1, Cdh5, Nos3, and Angptl7 for ECs (Fig. 4c). Based on these gene expression, four clusters (1, 3, 10, and 17) were grouped as rCMs, four clusters (4, 6, 13, and 14) as rSMCs, and six clusters (2, 5, 11, 12, 15, 16, and 18) as rECs (Fig. 4b–c). rCM clusters (1, 3, 10, 17) did not express Mesp1, Isl1, and Fut4 (or SSEA1) but expressed Hand2, Tnnt2, and Nkx2–5 (Fig. 4d), suggesting that reprogramming did not involve cardiac mesodermal or early cardiac progenitor stages. Among rCMs, two clusters expressed proliferation markers (Fig. 4d). Some CM genes were increased in rSMC clusters (4 and 6) and some SMC markers such as Hexim1, Smtn, Cnn1 and Emilin2 were expressed in rCMs (Fig. 4d–e), suggesting partial reprogramming and/or intermediate lineage cells. Nonetheless, most SMC markers were expressed exclusively in rSMC clusters (4, 6, 13 and 14) (Fig. 4e). Among EC clusters, one cluster (16) robustly expressed all mature EC markers such as Pecam1, Cdh5 and Vwf already at day 6 (Fig.4f). In reprogrammed cells, there were 3 clusters (7, 8 and 9) showing expression of fibroblast genes without expressing CM, EC and SMC markers (Fig. 4b). Theses clusters were left as unclassified due to the lack of expression patterns of known cell-specific marker genes (Supplementary Table 3). In cluster 7 of reprogrammed cells, 12 differentially expressed genes were found (P < 0.1, Benjamini-Hochberg correction for multiple tests), and those genes are related to mitochondrial metabolism including ATP synthesis and electron transport chain in GO biological process (Supplementary Table 3). The differentially expressed genes in cluster 8 (263 genes) are also involved in mitochondrial metabolism as well as regulation of cell migration. In cluster 9, only one gene, AY036118, with statistical significance (P < 0.05) was found. AY036118 is ETS-related transcription factor, Erf1, which suppresses ets-associated tumorigenesis and is regulated by phosphorylation during the cell cycle and mitogenic stimulation25. These unclassified clusters appear to be related to an intermediate stage of becoming rCMs, given their enrichment of mitochondrial-related genes. After reprogramming, no cell clusters showed the same fibroblast profiles as the unreprogrammed MTTFs, while various fibroblast genes were scattered among all clusters with more heterogeneity (Supplementary Fig. 5). Representative fibroblast genes commonly expressed among MTTFs were downregulated in reprogrammed cells (Supplementary Fig. 5a, c). Expression of cancer-associated fibroblast genes such as Mmp1126, Tgfbi ²⁷ and Efemp1²⁸ were reduced in reprogrammed cells (Supplementary Fig. 5c). Taken together, RNA sequencing studies demonstrate successful and simultaneous reprogramming of MTTFs by MAB into CMs, ECs and SMCs. While fibroblast gene expression was globally reduced, some cardiac myofibroblast signatures were increased, and reprogrammed cardiac cells expressed certain levels of fibroblast genes at day 6 after reprogramming.

Fig. 4 |. Transcriptome analysis.

a-b, UMAP showing clusters of unreprogrammed MTTFs (a) and reprogrammed cells (MTTFs treated with MAB for 6 days) (b). The cell number of each cluster and potential cell types are described in the boxes. c, Identification of cell clusters enriched with CM-, EC- and SMC-specific marker genes. d-f, Violin plots showing gene expression of CPC, CM, and proliferation markers (d), SMC markers (e), and EC markers (f).

Favorable therapeutic effects of rCVT on experimental myocardial infarction

To investigate the therapeutic potential of rCVT, we transplanted rCVT onto the infarcted mouse heart (Fig. 5a and Supplementary Fig. 6a). Mice were randomly assigned to five groups: sham, MI without any treatment (MI), MI followed by intramyocardial injection of 5 ×10⁵ MTTFs (MI + MTTF), MI followed by transplantation of decellularized rCVT (MI + d-rCVT) or with rCVT (MI + rCVT). rCVTs were generated from GFP-MTTFs, detached from the culture dish at day 10 before the surgery, and sutured directly onto the left ventricular (LV) anterior wall and covering the apex of the heart after permanent ligation of the left anterior descending (LAD) coronary artery (Supplementary Fig. 6a). For consistency, the CVTs were prepared to weigh an average of 15 mg and placed directly above the region of the infarct (Supplementary Fig. 6b). To examine the non-cell autonomous effects of rCVT, decellularized rCVT (d-rCVT) was prepared by removing the cells with Triton-X and EDTA²⁹ (Supplementary Fig. 6c). Cardiac function was evaluated by echocardiography at weeks 0, 1, 4, and 12 after the surgery. Compared to the two MI control groups (MI and MI + MTTF), rCVT transplanted MI mice, but not d-rCVT transplanted mice, showed smaller LV end-diastolic dimension (LVEDD) and LV end-systolic dimension (LVESD) as well as greater fractional shortening (FS) and ejection fraction (EF) as early as week 1 and throughout the 12-week study period (Fig. 5b and Supplementary Fig. 6d).

Fig. 5 |. Transplantation of rCVT on MI heart.

a, Surgical views of MI heart before (upper) and after (lower) rCVT transplantation. b, Echocardiographic analyses before and 1, 4, and 12 weeks after the surgery (Sham (n = 7); MI (n = 11); MI + MTTFs (n = 9); MI + d-rCVT (n = 8); MI + rCVT (n = 11)). LVEDD: LV end-diastolic dimension; LEVSD: LV end-systolic dimension; FS: fractional shortening; EF: ejection fraction. Two-way ANOVA with Tukey’s multiple comparison test. Data are mean ± SEM. c, Daily quantification of PVCs for 4 weeks. Data are mean ± SEM. n = 5. d-e, Representative cross-sectional images of Masson’s trichrome-stained hearts at weeks 1, 4, and 12 (d), and quantitative analyses of the circumferential fibrosis area at week 12 (e). One-way ANOVA with Bonferroni’s multiple comparison test. Data are mean ± SEM. d-rCVT: decellularized rCVT.

Transplantation of immature cardiomyocytes with spontaneous electrical activity and isotropic cell alignment poses risks for arrhythmias, particularly during the first 1–2 weeks of cell transplantation^30,31. We quantified the types and frequency of ventricular arrhythmias with real-time, 24/7 measurements of the animals’ electrocardiogram for four weeks by implantation of telemetry two days after the surgery (Supplementary Fig. 6e). Telemetry was implanted at day 2 to lessen the surgical trauma to the acute MI animals at day 0. Occasional premature ventricular contractions (PVCs) were observed in MI mice transplanted with either rCVT (n=5) or d-rCVT (n=5) at frequencies too low to be consequential (Fig. 5c and Extended Data Fig. 3a). No sustained or non-sustained ventricular tachyarrhythmias were observed in either group. In the absence of spontaneous tachyarrhythmias, the animals were subjected to a programmed electrical stimulation (PES) protocol to gauge arrhythmia inducibility. The majority (n = 4 out of 5) of the mice could not be induced into arrhythmias upon PES in either group. PES led to ventricular tachycardia in one animal from each group (Extended Data Fig. 3b–c). Collectively, these data demonstrated that implanted rCVT did not induce ventricular arrhythmias in the MI mice. Masson’s trichrome staining at weeks 1, 4 and 12 showed engrafted rCVTs onto the infarct hearts. Thinning of the LV wall and compensatory enlargement of the ventricles were significantly reduced in the rCVT-transplanted group compared to the MI, MI + MTTF, and MI + d-rCVT groups (Fig. 5d). The percent circumferential fibrosis at week 12 was significantly smaller in the rCVT-group compared to all other groups (Fig. 5e). Taken together, these results demonstrated favorable therapeutic effects of rCVT on acute MI.

Migration of rCVT-derived cells into host myocardium

Next, we investigated migration of rCVT-cells into the infarcted hearts. Since the size of the engrafted rCVT was progressively reduced over time after transplantation (Fig. 5d), we suspected that a portion of the cells within the rCVT had migrated into the infarcted heart. To examine cell migration, we transplanted GFP-mouse-derived rCVT onto infarcted wild type (WT) mice. In detail, we isolated MTTFs from constitutively GFP-expressing transgenic mice (GFP-MTTFs) and generated rCVT (GFP-rCVT) (Supplementary Fig. 7a). These GFP-MTTFs showed robust expression of GFP (~97%) in flow cytometric analysis (Supplementary Fig. 7b). We sutured this rCVT onto the heart of WT mice after creating MI (Supplementary Fig. 7c). Confocal microscopic examination at 1 week showed distinct GFP-rCVT on the infarcted heart (Supplementary Fig. 7d). While some GFP⁺ cells were seen on the cardiac side, most of the GFP⁺ cells were still located on the rCVT side (Supplementary Fig. 7d). At week 2, cell migration from GFP-rCVT (GFP⁺ cells) toward host hearts were visible (Supplementary Fig. 8a). Many GFP⁺ cells were found in the infarcted area at week 4 and the adjacent and remote areas at week 16 (Supplementary Fig. 8b). Immunohistochemical staining for the GFP signal with a 3, 3’-diaminobenzidine (DAB) reagent further confirmed these results (Supplementary Fig. 8c–d). We further investigated potential migration of GFP-rCVT cells by examining mRNA expression of GFP in various organs including heart, lung, liver, pancreas, stomach, kidney, brain and intestine (Supplementary Fig. 8e); however, qRT-PCR showed no remote migration of rCVT-cells into other organs. These data present evidence for migration of rCVT-derived cells in the host infarcted heart.

Neovascularization by rCVT Transplantation

We next investigated neovascularization, or new vessel formation, in the rCVT-transplanted heart by confocal microscopy. We considered several aspects in the analyses: the contribution of reprogrammed cells in rCVT to new vessels as vasculogenesis and coaptation, the contribution of host vessels to new vessels as angiogenesis, and functionality, serial changes, and regional variation (central or border zone) of rCVT-derived vessels.

By one week, immunostaining for PECAM1 showed newly formed vessels from migrated rECs (GFP⁺PECAM1⁺) and host ECs (GFP⁻PECAM1⁺) on the infarcted area (Supplementary Fig. 9a). These rapidly formed vessels must have supported endangered cell survival in the early phase of MI. At week 4, many GFP⁺ cells expressed PECAM1 and formed more mature vascular structures in the border zones (upper) and MI (lower) areas (Fig. 6a), suggesting potent vasculogenesis from rECs. Morphologically, rECs (GFP⁺PECAM1⁺) formed vascular structures not only by themselves (pink arrows), but also by coaptation with host ECs (GFP⁻PECAM1⁺) forming hybrid vessels (white arrows) (Fig. 6a–b). In the infarcted area, GFP⁺PECAM1⁺ rECs made up 5.1 ± 0.7% of the total cells and 43.6% of the total GFP⁺ cells (data not shown).

Fig. 6 |. Functional vessel formation by rCVT at week 4.

MI hearts transplanted with GFP-rCVT. a-b, Immunostaining for PECAM1 showing an MI border area (a) and an MI area (b). Pink arrows indicate vessels formed by rECs (GFP⁺PECAM1⁺) alone and white arrows indicate vessels formed by rECs with host ECs (GFP⁻PECAM1⁺). b, Confocal microscopic images of a BLS1-perfused heart at an MI border area. d, Serial confocal images of a hybrid vessel formed by coaptation of rECs (GFP⁺ILB4⁺) and host ECs (GFP⁻ILB4⁺). The magnified image shows the junction of a rEC and host EC. e-f, Immunofluorescence staining for ACTA2 in serial confocal images (e) and magnified images (f). Reprogrammed SMCs (GFP⁺ACTA2⁺, green arrows) and host SMCs (GFP⁻ACTA2⁺, white arrows) surround a BSL1⁺ EC (red) in a large vessel.

To explore the functionality of blood vessels formed by rECs, we perfused hearts with rhodamine-conjugated bandeiraea simplicifolia lectin 1 (BSL1) (Fig. 6c, e–i and Supplementary Fig. 9b–c) or isolectin B₄ (ILB4) (Fig. 6d). At week 4, in the infarcted (Supplementary Fig. 9b) and border (Fig. 6c) areas, numerous BSL1-perfused vessels were seen and many of them were GFP⁺, indicating their origin from rECs and functional connection to the systemic circulation. We also observed many GFP⁻BSL1⁺ host ECs contributing to formation of various-sized vessels, showing strong angiogenesis (Supplementary Fig. 9b). Intriguingly, as seen in Figure 6d, we found functional hybrid vessels (coaptation) consisting of host ECs and transplanted rECs. In the remote zones, we observed contribution of GFP⁺ cells to neovascularization through vasculogenesis including coaptation, although the extent was not as robust as in infarct or border zones (Supplementary Fig. 9c). To further examine the functionality of the new vessels, the existence of red blood cells (RBCs) was explored by immunostaining for TER-119 in BSL1 perfused heart at week 4. We found that GFP⁺BSL1⁺ new vessels contained TER-119⁺ RBCs (Supplementary Fig. 10), indicating connections of new vessels to the systemic circulation. We then investigated the contribution of rSMCs to the formation of vessels. Immunostaining for ACTA2 showed ACTA2-expressing GFP⁺ cells, which surrounded GFP⁺BSL1⁺ rEC-derived vessels (Fig. 6e). There were even hybrid vessels surrounded by host SMCs (GFP⁻ACTA2⁺, white arrows) and transplanted rSMCs (green arrows) (Fig. 6f). These findings indicated that rSMCs also contributed to the formation of mature and muscular vessels. When we examined the vasculature at week 16 after BSL1 perfusion, in the infarcted area, various-sized vessels consisting of rECs alone (white arrows), rECs coaptated with host ECs (yellow arrows), and host ECs alone (red arrows) were still observed, but most of the rEC-derived vessels were large and tubular (Fig. 7a). Vessels fully composed of or coaptated with GFP⁺BSL1⁺ rECs comprised 21.0 ± 4.2% of the total BSL1⁺ vessels in the infarcted area. In the remote zones, solely rEC-derived small vessels were found (Fig. 7b). Again, we observed ensheathment of rSMCs (GFP⁺ACTA2⁺) over hybrid vessels consisting of host ECs (GFP⁻PECAM1⁺) and rECs (GFP⁺PECAM1) in 2D (Fig. 7c) and 3D images (Fig. 7d–e and Supplementary Movie 3).

Fig. 7 |. Functional vessel formation by rCVT at week 16.

a-b, Infarcted (a) and remote (b) areas of BSL1-perfused hearts at week 16. White, red, and yellow arrows indicate vessels formed by rECs alone (GFP⁺BSL⁺), host-derived vessels (GFP⁻BSL⁺), and hybrid vessels, respectively. c-e, 2D (c) and 3D (d-e) images of hybrid vessels in BSL1 perfused hearts after staining for ACTA2. rSMCs (GFP⁺ACTA2⁺, yellow arrows) surround BSL1⁺ EC in a vessel (b). Individual images of each fluorescence at two different angles (d) and merged images (e) showing a vessel composed of host EC/rEC and host SMC/rSMC. Related to Supplementary Movie 3.

Taken together, these findings demonstrated that rCVT-derived rECs and rSMCs significantly contributed to vessel formation in the central and border zones of infarct hearts through vasculogenesis, including coaptation, and angiogenesis. Over 16 weeks, these reprogrammed cells and rCVT-derived vessels underwent dynamic migration and remodeling, forming functional and mature vessels.

Cardiomyogenesis by rCVT transplantation

To investigate cardiomyogenesis by the migrated GFP-rCVT cells, we conducted immunostaining for ACTN2, TNNT2 and GJA1 with 4 and 16 week samples. At week 4, immunohistochemical staining for the GFP signal with DAB reagent showed migrated GFP-rCVT cells in the border and infarcted area (Fig. 8a). GFP⁺ACTN2⁺ cells were found in small groups in infarcted host heart (Supplementary Fig. 11a–b). rCMs at 4 weeks were coarsely striated and were smaller than the host CMs, suggesting an immature state. At week 16, rCMs were rarely found in the infarct area, but were observed more frequently in the border zone. Some mature rCMs were morphologically indistinguishable from host CMs (Fig. 8b–c). These rCMs expressed TNNT2 (Fig. 8b) and ACTN2 (Fig. 8c), and shared GJA1 expression with host CMs, suggesting a mature form of CMs. Fluorescent staining with rhodamine-wheat germ agglutinin (WGA) further demonstrated T-tubules in rCMs as well as in host CMs (Fig. 8d–e). The T-tubules, invagination of sarcolemma^32,33, appear as dots arranged in a linear fashion within the cells. Approximately 16% of rCMs were bi-nucleated, which was not seen by 4 weeks (Fig. 8f–g). To explore the potential for fusion between the rCMs and host CMs, we co-cultured GFP-rCVT with adult CMs from MYH6 promoter-driven mCherry mice in vitro. Immunostaining at 2 weeks showed a rare incidence of fusion-like phenomena (GFP⁺mCherry⁺ double positive cells) (Supplementary Fig. 11c), measuring 0.5% by flow cytometric analyses (Supplementary Fig. 11d). We further found that rCMs (GFP⁺TNNT2⁺) did not express Ki67 or phospho-histone H3 (PH3) at week 12, suggesting loss of proliferative capacity in the heart (Supplementary Fig. 11e).These data suggest that increased cell size and bi-nucleation of rCMs resulted from maturation of rCMs in vivo. Although the total number of rCMs (GFP⁺ACTN2⁺) was reduced from 4 to 16 weeks, they comprised 3.7 ± 1.0% of the total ACTN2⁺ CMs in the border zone. In total, our data indicated that rCVT transplantation into infarct hearts induced migration of rCMs into the infarcted and border zones, and allowed them to mature over 16 weeks in the border zone and contribute to new CM generation.

Fig. 8 |. Cardiomyogenesis induced by rCVT.

a, Microscopic images of GFP-positive cells stained by DAB in GFP-rCVT transplanted MI hearts at week 4. b-e, Confocal microscopic images of GFP-rCVT-transplanted MI hearts at week 16. Immunofluorescence staining for GJA1 together with TNNT2 (red) (b) or ACTN2 (c). Representative images following immunofluorescence staining of hearts with rhodamine-WGA and TNNT2, showing cell boundaries and T-tubules in GFP⁺ cells (d) and GFP⁺TNNT2⁺ cells (e). White arrows indicate some T-tubules inside GFP⁺ rCM and host CMs. f, Confocal images of bi-nucleated rCMs at week 16. White interrupted lines and arrows indicate bi-nucleated rCM and nuclei, respectively. g, The percentage of bi-nucleated rCMs at weeks 1, 4 and 16 among total rCMs in the MI border area. Data are mean ± SEM. U indicates undetected.

Cell fate of fibroblasts in transplanted rCVT

Since half of rCVT-cells are fibroblasts (Supplementary Fig. 1e–f), we next examined the fate of fibroblasts derived from rCVT. Immunostaining showed that in the MI area, GFP- and either VIM- or THY1-double positive cells (GFP⁺VIM⁺ or GFP⁺THY1⁺) were rarely found at weeks 1, 4, and 16 (Supplementary Fig. 12a–d). These results suggest that most fibroblasts derived from rCVT did not survive well or were further reprogrammed toward the three cardiac cell types in the heart. The very low survival of transplanted fibroblasts has been reported previously³⁴.

Scaffolding effects of the rCVT on infarct heart

Since the therapeutic effects were observed as early as one week, which is before many rCVT cells were migrating into the heart side (Supplementary Fig. 7d), we suspected that other effects of transplanted rCVT, such as scaffolding effects, were involved as early therapeutic mechanisms. Thus, we investigated regional cardiac strains using high-fidelity speckle-tracking imaging in the longitudinal axis in the MI, MI + d-rCVT, and MI + rCVT groups 3 and 7 days after the surgery. Global longitudinal strain (GSL) and the time-to-peak to reach the highest strain were calculated in six segments in the longitudinal axis (base, mid, and apex of anterior and posterior endocardium) (Supplementary Fig. 13a–b). Compared to normal mice, GLS was decreased in all MI mice at day 3: it was more significantly decreased in the d-rCVT-transplanted group compared to the rCVT-transplanted group and there was no significant difference between the MI and MI + rCVT groups (Supplementary Fig. 13a). At day 7, GLS was further decreased in the MI and MI + d-rCVT groups, but not in the MI + rCVT group. Moreover, abnormal strain patterns such as severe heterogeneous time-to-peak (Supplementary Fig. 13b) and increased longitudinal strain in the opposite direction compared to the normal strain (Supplementary Fig. 13c) were seen in the MI and MI + d-rCVT groups, but less in the MI + rCVT group. These data demonstrate that implanted rCVT improved regional cardiac strains within 7 days and preserved cardiac function, establishing mechanical supportive effects of rCVT and the requirement for cells within rCVT for appropriate therapeutic function.

Paracrine effects of rCVT

To explore the mechanisms underlying the early therapeutic effects of rCVT, we also investigated the paracrine effects. We examined mRNA expression of Vegfa, Fgf2, Angpt1, Mmp2, Mmp3, Mmp9, Igf1, and Hgf by qRT-PCR, comparing untreated MI hearts and rCVT-transplanted MI hearts at 1 and 4 weeks (Supplementary Fig. 14). At week 1, Mmp3, Igf1 and Hgf levels were higher in rCVT-MI hearts compared to untreated MI hearts. Igf1 and Hgf are well known cell survival, angiogenic, and cardiomyogenic factors^35,36. At week 4, all examined genes were higher in rCVT-MI hearts than in untreated MI hearts. These data revealed that paracrine effects also play a role in the beneficial effects of rCVT on cardiac repair after MI.

Discussion

This study demonstrates the proof-of-concept for direct reprogramming of non-cardiac, adult fibroblasts into de novo cardiovascular tissue constructs. The rCVT recapitulates major composition of the native chamber myocardium, integrating cardiomyocytes, endothelial cells, smooth muscles cells and fibroblasts embedded in the cardiac extracellular matrix. Direct epicardial transplantation of an rCVT patch on infarcted hearts significantly improved post-MI cardiac function through scaffolding effects, paracrine actions, enhanced cell survival, neovascularization, and cardiomyogenesis. Long-term follow-up studies showed maturation of rCVT-derived CMs, ECs and SMCs.

Conceptually, this study presents the possibility of direct reprogramming into tissue from single adult somatic cells, a previously unknown path, and its utility in regenerative medicine. A single cardiac miRNA mimic, miR 208b, together with AA and BMP4 generated a cardiovascular patch, rCVT, which included rCMs, rECs, rSMCs, fibroblasts, and ECM, within 10 days of culture. This multi-lineage “tissue” reprogramming can offer an advance in cardiac regeneration approaches by satisfying some major unmet needs of the current strategies. One of the most critical barriers for cell therapy for cardiac regeneration is low cell retention after cell transplantation, and various approaches have been attempted. Among them, the most successful approach is biomaterial-mediated cell delivery^10,37. The biomaterial provides a protective environment for transplanted cells and enhances cell survival. Another approach is addition of supportive cells such as ECs, SMCs, and/or fibroblasts to stem cell-derived CMs¹⁵. The supportive cells augment cellular cross-talk, paracrine actions, vascularization, and cell survival and function in the infarcted area³⁸. There are also reports that a combination of biomaterial and these supportive cells can further improve the effects of cell therapy^11,12,39. In fact, rCVT provides these cell types and natural matrix at the same time, bypassing the complicated processes needed for artificial tissue construction using individual elements. The other important and novel finding of our study is migration of cells from the transplanted rCVT toward the infarcted host heart. In previous studies, when a patch composed of cells and biomaterials was transplanted onto the infarct heart, there was little or no migration of cells from the patch into the host heart, and some studies showed formation of a non-cardiomyocyte layer between the exogenous patch and the heart, which restricted cell migration^11–13. However, rCVT did not induce such barrier formation, allowing migration of cells. Such characteristics provide neovascularizing and cardiomyogenic potential within the host heart in addition to its early scaffolding effects. This is probably because while other patches include artificial or foreign ECMs, rCVT includes its own naturally formed matrices and highly expresses matrix metalloproteinases (Supplementary Fig. 3a–b).

We also found a novel role of miRNA for reprogramming toward ECs and SMCs. In fact, our method is the first one to demonstrate simultaneous reprogramming of somatic cells into vascular cells as well as CMs. In our pilot study, we employed five miRNA mimics, miR-1a-2–5p, miR-208a-3p, miR-208b-3p, miR-208b-5p and miR-499–5p, alone or together. Each of these miRNAs induced expression of cardiomyocyte genes. Unexpectedly, miR-208b-3p additionally induced expression of EC- and SMC-specific genes. These findings are not surprising because miR-1 was suggested to be involved with vessel formation⁴⁰ and overexpression of miR-208a in myoblasts was reported to upregulate collagen I and endoglin (Eng)⁴¹, which are critical EC specification genes⁴².

This study also demonstrated indispensable functions of AA and BMP4 for reprogramming of fibroblasts into cardiovascular cells and tissue generation. Addition of AA and BMP4 to miR-208b-3p not only reprogrammed MTTFs toward CMs, but also into rECs and rSMCs. AA and BMP4 also stimulated fibroblasts to secrete ECMs to form 3D structures. Addition of both, but neither alone, induced these effects. AA stimulates biosynthesis of collagen and other ECM components and their cross-linking¹⁸, enhances CM differentiation of ESCs¹⁹, and functions as an epigenetic modifier²⁰. BMP4 signaling plays a role in cardiac specification¹⁷ and CM differentiation from human ESCs⁴³. Its function in vascular cell differentiation of ESCs was also reported⁴⁴. Our study further showed that these two molecules synergistically facilitated direct reprogramming and tissue generation.

Transplanted rCVT induced cardiac regeneration through versatile mechanisms. Since the therapeutic approach in this study is tissue transplantation with multiple cellular elements and ECMs, it is difficult to exactly define the underlying mechanisms. However, our data support cell retention, mechanical support, paracrine factor secretion, neovascularization, and cardiomyogenesis as all contributing to cardiac repair. In the early phase, cell retention, paracrine effects, and scaffolding effects play major roles. The first four weeks after MI is a critical window for future cardiac function, as most cardiac cell death happens during this period⁴⁵. rCVT, by forming a tissue-like structure, promotes survival of transplanted cells and thus increases and prolongs their therapeutic function. Particularly at an early phase, paracrine effects are prominent, salvaging endangered CMs. Paracrine factors also induce angiogenesis (growth of the host vessels), protecting both host myocardium and transplanted cells. Moreover, at an early phase, this tissue-like structure provides scaffolding effects on infarcted walls and prevents cardiac deformation and further deterioration of function.

In the next phase, new vessel formation through rECs and rSMCs plays a prominent role. Cell migration from the rCVT into host myocardium was clearly visible between 1 and 3 weeks after the transplantation. These migrated rECs and rSMCs formed new vessels and connected to host vessels, which further help salvage the endangered myocardium and prevent ongoing myocardial damage. Numerous hybrid vessels formed between rECs/rSMCs and host vascular cells were observed, suggesting clear association between transplanted cells and the host vascular system. Between 4 to 16 weeks, many vessels were reorganized: reactive vessels in the infarct area disappeared and more rCVT-derived vessels were seen at the border and remote zones at 16 weeks. To our knowledge, such robust and mature vessel formation contributed by transplanted cells in hearts has not been reported.

Cardiomyogenesis happened slowly over 16 weeks. At week 4, many rCMs were localized in the central infarcted area in immature forms, and some were in the border zone in a more organized and mature form. At 16 weeks, most rCMs were found in the border zone in a mature form, histologically indistinguishable from host CMs, and formed gap junctions with host CMs. The increased number and maturation of rCMs in the border zone suggests their continuous role in preventing adverse cardiac remodeling and promoting cardiac regeneration over 16 weeks. In addition, the paracrine role of transplanted rCMs per se cannot be disregarded in either the early or later phases, as CMs are a major depot of VEGF, the most important angiogenic factor⁴⁶.

There are limitations and remaining questions in this study. This study is a proof of concept study demonstrating the potential generation and utility of a directly reprogrammed tissue-like structure, but rCVT does not exactly mimic heart tissue. For example, rCMs did not exhibit electrophysiological characteristics of mature CMs in vitro; however, when transplanted, rCMs became mature CMs in the host heart. scRNA-seq was performed only at day 6 because the increase in cell size with longer culture periods prevented larger rCMs (> 50 μm in diameter, Supplementary Fig. 2) from being captured on the microfluidics-based scRNA-seq platform. Bulk RNA-seq at D10 as well as scRNA-seq at D6 indicated that residual fibroblast gene profiles were observed in the reprogrammed cells despite suppression of the overall fibroblast gene expression with a parallel increase in the heterogeneity of cell populations. Populations of reprogrammed cells exhibited gene signatures of multiple cell types. The presence of cell populations with more than one cell lineage may reflect partially reprogrammed cell populations that would eventually achieve a final cell type, or a true intermediate cell population. Recent single-cell/nuclear studies reported that the adult mammalian heart harbors cell populations that exhibit markers of more than one cell type including endothelial and CM lineages and endothelial and fibroblast lineages^47,48, which supports the case for an intermediate population. Even adult human hearts include various types and maturity levels of CMs as well as CMs expressing SMC genes, showing significant heterogeneity and maturity⁴⁹. Not surprisingly, retention of fibroblast gene profiles and cellular immaturity are shared features of other directly reprogrammed or induced cells^50–52. In our study, while individual cells in rCVT were immature, the cells underwent maturation and became functional cells in situ upon transplantation. This is in line with human PSC-derived CMs, in which immature hPSC-CMs became mature over time when transplanted in vivo^53,54. Studies with human PSC-derived hepatocytes also showed that true maturation happens only when cells are transplanted in vivo⁵⁵. Due to the complex nature of reprogramming into multiple cell types and ECM deposition, and the complexity of data derived from more powerful technologies such as single cell sequencing, the mechanisms of the reprogramming studies will take considerable resources and time, and thus will be explored in future studies.

Despite these limitations and questions, this study presents previously unreported potential for tissue reprogramming from differentiated adult somatic cells by non-genetic introduction of miRNAs with AA and BMP4, bypassing stem cell stages. This study demonstrates that rCVT can serve as a valuable platform for cardiac regeneration and may also prove useful for drug discovery and disease modeling.

Methods

Animal Models

All animal experiments were approved by the Institutional Animal Care and Use Committees of the Emory University and Yonsei University College of Medicine, and were performed in accordance with federal guidelines. C57BL/6J wild-type (#000664) or transgenic mice carrying a GFP reporter driven by human ubiquitin C (UBC-GFP, #004353) promoters or an mCherry reporter driven by the MYH6 promoter (MYH6-mCherry, #021577) were purchased from Jackson Laboratory. Nkx2–5-GFP transgenic mice were kindly given by Dr. Richard P. Harvey (Victor Chang Cardiac Research Institute)⁵⁶.

Cell culture and fibroblast reprogramming in vitro

Mouse tail-tip fibroblasts (MTTFs) isolated from 1–4 month old C57BL/6J wild-type or transgenic mice carrying a GFP reporter driven by the Nkx2–5 (Nkx2–5–GFP) or UBC-GFP promoters, or a MYH6-mCherry promoter, were cultured in DMEM/F-12 containing 10% FBS, 1% Glutamax, 1% non-essential amino acids and 1% Antibiotic-Antimycotic. After 3–5 passages, MTTFs were frozen until use. For reprogramming purposes, MTTFs at 70–80% confluence were transfected with a synthetic miRNA mimic (20 nM): miR-1a-2–5p (5’-ACAUACUUCUUUAUGUAGUACCCAUA-3’, Qiagen (MSY0017047)); miR-208a-3p (5’-AUAAGACGAGCAAAAAGCUUGU-3’, Qiagen (MSY0000520)); miR-208b-3p (5’-AUAAGACGAACAAAAGGUUUGU-3’, Exiqon (472072–001), Qiagen (MSY0004939)); miR-208b-5p (5’-AAGCUUUUUGCUCGCGUUAUGU-3’, Qiagen (MSY0017280); miR-499–5p (5’-UUAAGACUUGCAGUGAUGUUU-3’, Qiagen (MSY0003482); or Scramble-miR (Dharmacon, CN-001000–01-05) using Lipofectamine 3000, and cultured in advanced DMEM/F12 containing 10% FBS, 1% Glutamax, 1% non-essential amino acids, 1% Antibiotic-Antimycotic, acetic acid (0.01%), AA (ascorbic acid, 0.25 mM) and BMP4 (Peprotech, 10ng/ml). Cells on a culture dish spontaneously aggregated, forming an rCVT at day 10. When no patch was formed at day 10, the border of the culture dish was scratched and cells were lifted up with a pipette to expedite cell aggregation forming an rCVT.

Quantitative RT-PCR

Total RNA was extracted using the RNeasy or RNeasy fibrous tissue kits (Qiagen) according to the manufacturer’s protocol, and cDNA was synthesized from 0.5 μg of RNA with a Taqman cDNA synthesis kit (Life Technologies). Quantitative RT-PCR was performed using SYBR Green PCR Master Mix, the 7500 fast real-time PCR system, and primers designed to examine expression of markers for pluripotent stem cells (Pou5f1, Sox2 and NANOG), CMs (Mesp1, Nkx2–5, Gata4, Mef2c, Tbx20, Hand1, Hand2, Tnni3, Myl2, Myl7, Myh6 and Myh7), fibroblasts (Thy1, Vim, S100a4 and Ddr2), ECs (Kdr, Tek, Cdh5, Nos3 and Pecam1), and SMCs (Acta2, Myocd and Myh11), angiogenic paracrine factors (Vegfa, Fgf2, Angpt1, Mmp2, Mmp3, Mmp9, Igf1 and Hgf), and GFP (Supplementary Table). All annealing steps were carried out at 60°C. Relative mRNA expression of target genes was calculated with the comparative CT method using Gapdh as an internal control. Differences in CT values (ΔCT = CT₁ (gene of interest) – CT₂ (Gapdh in experimental samples)) and relative mRNA expression (2^-ΔCT) were calculated for each sample and target gene.

Immunocytochemistry & immunohistochemistry

The cells and hearts were fixed with 4% paraformaldehyde for 10 and 60 minutes, respectively. Fixed tissue samples were incubated in 30% sucrose overnight and cryo-sectioned. Fixed cells and cryo-sectioned tissues were blocked with PBS containing 3% bovine serum albumin and 1% Triton X-100 for 60 min at room temperature (RT), and incubated with antibodies against ACTN2 (Sigma-Aldrich, 1:100 (Cell) or 1:30 (Tissue)), TNNT2 (Thermo Fisher Scientific, 1:100 (Cell) or 1:30 (Tissue)), GJA1 (Sigma-Aldrich, 1:200), PECAM1 (BD Biosciences, 1:100), CDH5 (BD Biosciences, 1:100), Ki67 (Abcam, 1:50), ACTA2 (Abcam, 1:100), SMTN (Santa Cruz Biotechnology, 1:100), SM22α (Abcam, 1:100), TEK (Santa Cruz Biotechnology, 1:100) or TER-119 (Abcam, 1:200) for 3 hours at RT or overnight at 4°C. The cells or tissues were washed with PBS for 10 minutes three times and incubated with anti-rabbit IgG–Alexa Fluor 405, 488, 555 or 647 (Thermo Fisher Scientific, 1:100), anti-mouse IgG–Alexa Fluor 488, 594 or 647 (Thermo Fisher Scientific, 1:100) or anti-rat IgG–Alexa Fluor 594 or 647 (Thermo Fisher Scientific, 1:100) for 3 hours at RT or overnight at 4°C. After washing with PBS for 20 minutes three times, the cells or tissues were mounted with VECTASHIELD^® Mounting Media with DAPI (Vector Labs). For T-tubule staining, rhodamine-conjugated WGA (Vector Labs, RL-1022, 1:200) was used. Immunofluorescence signals were detected with a fluorescence microscope (Nikon) or a Zeiss LSM 510/780/980 confocal laser scanning microscope (CLSM, Carl Zeiss). The percentage of host and engrafted ECs and CMs and the size of engrafted rCMs were quantified from confocal images of three separate sections from MI hearts transplanted with GFP-rCVT using ImageJ (NIH). Confocal images were analyzed for fluorescence intensity using Zen Blue (Zeiss) or were processed for 3D imaging and movies with Imaris software (Oxford Instrument). Masson’s trichrome staining (Leica) and 3,3’-diaminobenzidine (DAB) staining (Thermo Fisher Scientific) were performed according to the manufacturer’s instructions, and samples were mounted with VectaMount™ Mounting Medium (Vector Labs) after dehydration steps (2 × 90%, 2 × 100% ethanol and xylene for 1 minute).

Calcium transients and action potentials

For calcium transient measurements, cultured cells were incubated in Opti-MEM containing 1 μg/ml of Fluo-4 AM (Thermo Fisher Scientific) for 20 minutes at 37°C. After the incubation, cells were washed with DPBS and incubated in fresh Opti-MEM (Life Technologies) for 10 minutes for de-esterification of intracellular AM esters. Green fluorescence sparks indicating spontaneous calcium transients were monitored with an Epi-fluorescence microscope (Nikon) and movie files taken with image software NIS-Elements AR 3.0 (Nikon) were subjected to quantitative analysis by Pulse Video Analysis software⁵⁷. Whole-cell recordings were performed on spontaneously contracting cells. Reprogrammed cells were transferred into a bath mounted on the stage of an inverted microscope (Ti2, Nikon). The whole-cell patch was achieved by rupturing the membrane after giga ohm sealing. The bath solution was perfused at 5 mL/min and the voltage and current recordings were performed at 37°C. Patch pipettes with a resistance of 2–4 MΩ were connected to the head stage of a patch clamp amplifier (Axopatch-200B, Molecular Devices, Sunnyvale, CA, USA). The bath solution contained (in mM) 140 NaCl, 5.4 KCl, 1.8 CaCl₂, 2 MgCl₂, 10 glucose, and 10 HEPES (pH 7.4 titrated with NaOH). The pipette solution contained (in mM) 120 KCl, 10 EGTA, 1 MgCl₂, 3 MgATP, and 10 HEPES (pH 7.2 titrated with KOH). To acquire data and apply command pulses, pClamp 10.2 and Digidata 1550B (Molecular Devices) were used. Voltage and current traces were stored and analyzed using pClamp 10.2 and Origin 8.0 (OriginLab Corp., Northampton, MA, USA). The currents were filtered at 5 kHz and sampled at 1 kHz.

Transmission electron microscopy of rCMs

To examine formation of sarcomere structures in rCMs, cells on culture dishes were fixed at day 6 with 2.5% glutaraldehyde fixative and kept overnight at 4°C. Cells were washed with 0.1 M cacodylate buffer (EMS) at RT for 30 minutes before and after fixation with 0.1% osmium fixative (0.2 M cacodylate buffer: 6% potassium ferrocyanide (Sigma): 4% osmium tetroxide (EMS) = 2:1:1) for 1 hour at RT. After dehydration of cells by soaking in 25%, 50%, 70%, 95%, and 100% ethanol, cells were infiltrated and embedded in Eponate 12 resin (Ted Pella). TEM images were acquired with a JEOL JEM_1400 transmission electron microscope at the Robert P. Apkarian integrated electron microscopy core facility at Emory University⁵⁸.

Uptake of acetylated low density lipoprotein

To examine the functionality of reprogrammed EC-like cells, cells were incubated in Opti-MEM containing Dil-labeled acetylated low density lipoprotein (Dil-Ac-LDL, Thermo Fisher Scientific (L-3484), 10μg/ml) for 2 hours at 37°C. Uptake of Dil-Ac-LDL was monitored under epi-fluorescence microscopy and MTTFs treated with Dil-Ac-LDL were used as negative controls.

Detection of intracellular nitric oxide (NO) production

MTTFs w/wo reprogramming treatments were incubated with Opti-MEM containing an NO indicator, DAF-FM Diacetate (Thermo Fisher Scientific (D-23844), 1 μM) for 20 minutes at 37°C. After washing with DPBS, cells were incubated with culture medium and FITC-fluorescence was monitored under epifluorescence microscopy 3 hours later.

Flow cytometry

rCVTs were dissociated with Accutase (Invitrogen) for 10 minutes in a cell incubator and 20 minutes in a 37° water bath. After washing out the enzyme solution, cells were passed through a cell strainer to eliminate cell clumps and were suspended in MACS running buffer. To detect surface markers, cells were incubated with FITC- or APC-conjugated anti-THY1 (Biolegend) or anti-PECAM1 antibody (Biolegend) for 1 hour in a 37° incubator. To detect intracellular proteins, cells were fixed with 4% paraformaldehyde, and permeabilized/blocked in PBS containing 3% BSA and 2% Triton-X. To detect expression of CM, SMC, fibroblast, or proliferation markers, cells were incubated with mouse anti-TNNT2, rabbit anti-smoothelin (SMTN, Santa Cruz Biotechnology), rabbit anti-vimentin (VIM, Abcam) or rabbit anti-Ki67 (Abcam) overnight, respectively. After washing out primary antibody with PBS, cells were further incubated with anti-mouse–Alexa Fluor 647 or anti-rabbit–Alexa Fluor 647 (Thermo Fisher Scientific) for 30 minutes⁵⁹. For isotype controls, cells were incubated with mouse (BD Bioscience) or rabbit (Abcam) IgG overnight or anti-mouse–Alexa Fluor 647 for 30 minutes. Flow cytometric data were acquired with an Accuri C6 Flow Cytometer and BD LSRFortessa X-20 (BD Biosciences) and analyzed with FlowJo. The schematic gating strategy is shown in Supplementary Figure 7b.

Decellularization of rCVT

rCVTs were incubated in 0.25% trypsin for 2 hours in a 37°C water bath and stirred for 30 minutes in PBS containing 3% triton-X and 0.05% EDTA. Decellularized rCVTs were washed in PBS two times and drained in cell strainers (BD Falcon) for 1 hour to wash out cell debris. Decellularized rCVTs were stained with 4, 6-diamidino-2-phenylindole (DAPI) and monitored with epifluorescence microscopy to confirm successful removal of cells.

Total RNA transcriptome analysis

Total RNA was extracted at days 0 (D0), 6 (D6) and 10 (D10), and the converted cDNAs from isolated RNAs and input were used to generate DNA libraries following the Illumina protocol. 38-cycle single end sequencing was performed. Image processing and sequence extraction were done using the standard Illumina Pipeline. Two independent libraries and runs (one lane per library) for each biological replicate were generated. FASTQ sequence files were aligned to the mm10 reference genome using HISAT1 (v2.1.0) with default parameters. RNA-seq data for primary EC (SRP285778)²¹, SMC (SRP253689) and CM (SRP057984)²² were downloaded from the NCBI SRA database. All RNA-seq reads were aligned to the mm10 mouse genome by Tophat (v2.1.1) with default parameters⁶⁰. The read count in each gene was calculated by HTSeq (v0.12.4) using Refseq genes as reference gene annotation⁶¹. The variation in each dataset was normalized by RUVSeq (v1.24.0) in R (v4.0.2). Briefly, the gene-sample count matrix was first normalized by upper-quartile. Subsequently, the top 500 less variable genes were identified as empirical negative controls and used for RUVg normalization with k = 3 options. PCA was implemented with the normalized count matrix by prcomp function. Gene signatures (unique gene sets for each cell type) of the heart, EC, SMC, and CM used for GSEA were defined by 2-fold gene expression change. Enrichment of these gene signatures in D6 and D10 reprogrammed cells relative to D0 was evaluated by GSEA software (v4.1.0) with 1,000 permutations to gene set, no dataset collapse and weighted enrichment statistics⁶².

Single-cell transcriptome analysis

MTTFs with or without MAB treatment for 6 days were dissociated with Accutase (STEMCELL) and filtered to remove aggregated cells or oversized cells. Library construction (10X Chromium Single Cell 3’ Reagent Kits v3.1), sequencing (Illumina NovaSeq 6000 platform) and conversion of preliminary results to FASTQ files (Cell Ranger v3.1) were performed sequentially. After alignment of FASTQ files to the mouse reference genome (mm10)⁶³, Cell Ranger was used for data analysis and generation of a file containing a barcode table, a gene table and a gene expression matrix⁶⁴. To visualize scRNA-seq results, cloupe files generated with Cell Ranger were analyzed using Loupe Browser, and gene expression-based clustering information and gene expression information for cells were acquired.

Transplantation of rCVT or decellularized-rCVT on infarcted heart

MTTFs derived from GFP mice were reprogrammed into rCVT for 10 days by MAB treatment. rCVT generated from a 10 cm culture dish was detached from the dish by lifting up with forceps, washed in DPBS twice, cut in half, and kept in fresh DPBS on ice until needed. We used rCVT weighing ~15 mg. Three month-old male C57BL/6 mice were anesthetized with 2% isoflurane, and 2% isoflurane mixed with oxygen was provided constantly through a ventilator connected to the intubation catheter. After opening the chest wall, myocardial infarction (MI) was induced by ligating the left anterior descending coronary artery located 2 to 3 mm distal to the left atrial appendage with an 8–0 prolene suture^{10,59,65–67}. rCVT (n = 11) or decellularized-rCVT (d-rCVT, n = 8) was sutured on the apex and a part of the anterior wall of the mouse heart with a 10–0 nylon suture. As a control, fibroblasts (1 × 10⁵ cells), the source cells, suspended in 30 μl of PBS were injected with a 30 G needle at two sites in the MI border zone (n = 9). As another control, MI without any treatment was included (n = 11). The mice were sacrificed 1, 4, 12, or 16 weeks after the surgery for immunohistological and molecular analyses. Rhodamine-conjugated bandeiraea simplicifolia lectin 1, BSL1, or APC-conjugated isolectin B₄ was diluted in DPBS (1 mg/ml), and 100 μl of the diluted solution was injected intraperitoneally before sacrifice to stain vascular endothelial cells.

Real-time electrocardiogram (ECG) measurements and arrhythmia inducibility

Continuous, real-time ECGs were acquired from conscious and ambulatory MI mice with rCVT (n = 5) or d-rCVT (control group, n = 5). Two days after creation of MI and transplantation of the rCVT constructs, mice were anesthetized and a midline incision was performed on the sternum. A rodent telemeter was implanted (ETA-F10, Data Sciences International, St. Paul, MN) by placing the electrodes for lead II measurement with the negative electrode fixed on the right upper border of the sternum, and the positive electrode on the left lower border of the sternum (Supplementary Fig. 6e). The telemeter body was inserted through the subcutaneous tunnel and placed in the right lateral side of the thoracic cage. Premature ventricular complexes were quantified by manual inspection of the raw ECG traces as well as by the arrhythmia analysis feature of Ponemah software (DSI, St. Paul, MN). At 4 weeks after transplantation, arrhythmia inducibility was examined on each mouse with programmed electrical stimulation as previously described⁶⁸. Upon anesthesia, a custom-designed stimulating electrode was inserted through the intercostal space to directly contact the cardiac apex, epicardially. A series of 10 stimuli was applied with a basic cycle length of 100 ms followed by a coupled extra-stimulus (S2). A train of three extra stimuli was applied after S1 (S2-S4) when double extra stimuli failed to induce arrhythmia. If ventricular arrhythmia was not induced after 3 extra stimuli (S1-S4), it was considered that arrhythmia was not inducible in the animal.

Echocardiographic analyses and determination of fibrosis

Echocardiography was performed before the surgery, and 1, 4, and 12 weeks after the surgery. LV end-diastolic dimension (LVEDD), LV end-systolic dimension (LVESD), fractional shortening (FS), and ejection fraction (EF) were measured with two-dimensional M-mode images using the Vevo 770 TM Imaging System (Visual Sonics, Inc.) as previously described^. Speckle-tracking echocardiography was additionally performed to investigate regional and global cardiac dynamics specifically at the early time points (3 and 7 days after the surgery) using Vevo 3100. For the immediate functional decline at the acute MI, more precise and sensitive measurements were needed to detect the therapeutic effects of rCVT at early time points. Cardiac cycles were acquired digitally from the long-axis view with a heart rate > 350 BPM (average 428 BPM) and a fixed frame rate at 307. The LV endocardium was mapped using Vevolab software. The heart was divided into 6 segments (base, mid and apex of anterior and posterior) and global longitudinal strain, segmental strain, and time-to-peak of each segment were automatically calculated. To measure fibrosis area, the hearts harvested at weeks 1, 4 and 12 were cross-sectioned and stained with Masson’s trichrome staining (Leica) according to the manufacturer’s instructions. The percentage of circumferential fibrosis area was calculated with ImageJ as we previously described.

Quantification and statistical analysis

Data were presented as mean ± SEM. Significance was determined by one-way ANOVA followed by multiple comparisons with the Tukey or Bonferroni method, two-way ANOVA followed by multiple comparisons with the Bonferroni method, or two-tailed t-test. GraphPad Prism 7 was used for the analyses, and values of P < 0.05 were considered to denote statistical significance.

Reporting Summary.

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

The main data supporting the results in this study are available within the paper and Supplementary Information. The raw and analysed datasets generated during the study are too large to be publicly shared, yet they are available for research purposes from the corresponding author on reasonable request. Some exemplary raw data are available from figshare with the identifier: The accession number for the RNA-seq data reported in this paper is GEO:GSE96617; the accession numbers for single-RNA seq are GEO:GSE179589, GSM5420997 and GSM5420998.

Extended Data

Extended Data Fig. 1. Increased mRNA expression of CM genes in reprogramming fibroblasts.

mRNA expression of CM genes after the indicated treatment. Expression in MTTFs was set as 1 and U indicates undetected. Ten independent experiments, each with technical replicates. ***P < 0.001: MAB vs MTTF, AA, BMP4, AA/BMP4 and miR-208b. Statistical analysis was performed using one-way ANOVA with Bonferroni’s multiple comparison test. Data are mean ± SEM.

Extended Data Fig. 2. Transcriptome changes during the reprogramming of fibroblasts in total RNA-seq analysis.

a, 3D PCA plot of fibroblasts (D0), reprogrammed cells (D6 & D10), fetal heart, ECs, SMCs and CMs. b, Relative enrichment of gene signatures for heart, ECs, SMCs and CMs in D6 and D10 reprogrammed cells. Genes are sorted by relative expression to D0 (fibroblasts). The presence of gene signatures is shown by black bars. Normalized enrichment score (NES); statistical significance (FDR). c, Heatmap showing expression of fibroblast markers.

Extended Data Fig. 3. No arrhythmogenic potential of rCVT in infarcted hearts.

a, Representative electrocardiograms of premature ventricular contractions (PVCs, arrows) in MI mice transplanted with d-rCVT (n = 5) or rCVT (n = 5). Telemetry was implanted at day 2 to lessen the surgical trauma to the acute MI animals at day 0. b-c, Representative electrocardiograms of MI mice transplanted with d-rCVT (b) or rCVT (c) upon programmed electrical stimulation after ECG monitoring for 30 days. VT: ventricular tachycardia.