Critical Factors for Cardiac Reprogramming

Cellular reprogramming achieved by somatic nuclear transfer or cell fusion has long been recognized¹. The potency of specific transcription factors as cell fate determinants was first demonstrated by the discovery of MyoD, a master regulator for skeletal muscle differentiation, and by the subsequent identification of several genes as lineage-converting transcription factors in blood cells^{2, 3}. These pioneering works led to the landmark study by Yamanaka and colleagues that demonstrated the generation of induced pluripotent stem cells (iPSCs) from fibroblasts by transducing four stem cell-enriched transcription factors, Oct4, Sox2, Klf4, and c-Myc^{4, 5}. Numerous subsequent improvements in techniques and additional factors have increased the efficiency and robustness of the technology and such enhancements continue, as do analyses of the similarities and differences of iPSCs to embryonic stem cells (ESCs). Increasingly efficient differentiation protocols now permit us to make significant quantities of many individual cell types from iPSCs.

More recently, a “next generation” of cellular reprogramming has involved the direct conversion of one adult cell type into another by combinations of lineage-specific transcription factors or microRNAs, without passing through a pluripotent stem cell state. Direct lineage reprogramming, also known as transdifferentiation, can yield a diverse range of medically relevant cell types, including pancreatic β cell-, neuron-, neural stem cell-, cardiomyocyte-, and hepatocyte-like cells from somatic cells^6–10. In some cases, the resulting cellular phenotype has been more mature and adult-like than the corresponding pluripotent stem cell-derived phenotype, likely reflecting the lack of an embryonic intermediate during the reprogramming process.

A limitation of direct reprogramming is that cells appear to quickly exit the cell cycle as they adopt a unique fate. Thus, the utility of reprogramming in vitro is limited given the inability to generate larger numbers of cells for potential therapeutic or investigative use. However, leveraging the in vitro system as a screening tool to identify combinations of factors that could more fully reprogram cells in an in vivo setting to stimulate organ regeneration could be powerful. A greater efficiency of cellular reprogramming in the vivo microenvironment, likely involving spatial cues, extracellular matrix proteins, tensile forces, among other factors, was highlighted by work from Melton and colleagues. In their work, a combination of transcription factors enriched in pancreatic β-cells could convert pancreatic exocrine cells into insulin-secreting endocrine-like cells in vivo, but not fibroblasts in vitro⁶. In 2010, we reported the in vitro reprogramming of fibroblasts to cardiomyocyte-like cells by various combinations of core cardiac developmental transcription factors. We found a minimum cocktail of Gata4, Mef2c and Tbx5 was sufficient to broadly reset of gene transcription toward a cardiomyocyte-like state⁷. Although global transcriptome changes were observed in pooled cells that activated cardiac reporter gene expression (~15% of cells), only ~0.5% of these became fully reprogrammed with the ability to contract - cells we termed induced cardiomyocytes (iCMs). This efficiency was similar to that achieved by the original set of iPS reprogramming factors, but unlike iPSCs, iCMs cannot be expanded in vitro. More recently, two other groups have reported the ability to reprogram cardiac fibroblasts in vitro into cardiomyocyte-like cells with GMT, while indicating improved reprogramming with either GMT plus addition of Hand2 (GMHT)¹¹, or by replacement of Gata4 with Myocardin¹². A third group described in vitro reprogramming into cardiomyocyte-like cells with a combination of cardiac-enriched miRNAs and a small molecule inhibitor of Jak¹³. Combinations of transcription factors, miRNAs and small molecules are currently being tried by many to identify the preferred combination of stimuli for cardiac reprogramming, as should be expected for any new technology.

Successful reprogramming of fibroblast cells to cardiomyocyte-like cells eluded the field for several decades, so it is no surprise that the conditions conducive to reprogramming are difficult to engineer. The approaches described to date require significant optimization of myriad experimental details, with many pitfalls that may lead to failure of reprogramming. Highly standardized conditions that make the process more efficient and more easily transferable among different laboratories will undoubtedly be developed, thereby facilitating successful entry into the field with greater ease. The difficulty for some labs to successfully achieve cardiac reprogramming is highlighted in the manuscript by Chen et al, published in this issue of Circulation Research¹⁴. Chen and colleagures report their difficulty in expressing levels of Gata4, Mef2c and Tbx5 sufficiently high to promote robust reprogramming, and correspondingly only observe only a few changes toward a cardiomyocyte-like phenotype and a minimal shift in the overall transcriptome of the overall population of transduced cells in vitro.

It is worthwhile to consider the technical differences between the reprogramming achieved by Ieda et al. and Song et al., compared to the less successful results described by Chen et al., with the objective of clarifying aspects of reprogramming that remain challenging and that should be addressed by future research. First, the starting cell type and the condition of the cells are pivotal for successful reprogramming. Chen et al. utilized 3–6 week old tail-tip fibroblasts (TTFs) or cardiac fibroblasts (CFs) for their studies. In our experience, neonatal CFs were more amenable to reprogramming than TTFs, with a greater percentage of α-myosin heavy chain (αMHC)-GFP⁺ cells expressing cardiac troponin T (cTnT) and more complete alteration of epigenetic marks at specific loci. We did not see any TTF that were more fully reprogrammed with the ability to contract. Similarly Protze et al. had more success with GMT in neonatal CFs compared to TTFs, although Song et al. report the emergence of some beating cells from TTF with GMHT after prolonged periods in culture of greater than one month. In addition to the source of fibroblasts, the health and senescent state of primary fibroblasts had major effects on reprogramming. We found α-MHC-GFP induction by GMT was best in fresh, primary fibroblasts without passage, and the efficiency progressively decreased by approximately 50% with each passage of primary fibroblasts.

Second, high expression levels and proper stoichiometry of reprogramming factors are necessary for success. In our hands, cells reprogrammed to activate the α-MHC-GFP reporter using retroviral vectors, had 6–8 fold more expression of all three factors than neonatal cardiomyocytes, which was much greater than fibroblasts; fibroblasts cells that were infected by the viral vectors but failed to reprogram had significantly lower levels of expression. Chen et al. utilized lentiviral vectors but never achieved high levels of expression of all three reprogramming genes in the same cell population. In their hands, Mef2c levels were only 10-fold greater in transduced TTFs than in non-transduced CFs; Gata4 levels were only 8-fold greater in transduced CFsthan fibroblasts. One would not expect significant reprogramming with the levels of expression achieved by Chen et al., as greater levels of each factor would be required.

Third, the reporter system used for screening will affect the results. We generated transgenic mice with the α-MHC promoter driving GFP and screened through many mice before selecting one that had the most reproducibly high expression of GFP in cardiomyocytes with the greatest specificity. Chen et. al utilized Cre-dependent reporter systems, seeing no activation of the α-MHC-Cre or Nkx2.5-Cre reporters, but robust activation of the cTnT-Cre reporter. Careful validation of the reporter systems and of the percentage of cardiomyocytes expressing the reporter is important, as is monitoring expression of several markers, such as the expression of α-MHC and cTnT in the same cell, among others, as we had reported.

Finally, for whole transcriptome analysis, meaningful experiments require selection of the subset of cells that have begun to reprogram. In our studies, a reporter system was used to select the fraction of cells that appeared to be shifting in gene expression from the much larger population that received the reprogramming factors, but failed to respond. In this setting, we observed a broad resetting of gene expression. While the failure to express high levels of the reprogramming factors by Chen et al. likely precluded broad alterations in gene expression, even partial resetting would have been undetected by their approach, since all cells that expressed the Tbx5-expressing virus were used for microarray analysis. For reprogramming studies, there is limited value in gene expression analysis without purifying cell types of interest.

The paper in this issue of Circulation Research highlights the fact that with current technology, direct cardiac reprogramming in vitro is challenging and requires scrupulous attention to a great variety of technical details. Already, several groups are improving upon the original recipe and as the field develops, other small molecules, secreted proteins, or miRNAs will likely improve the efficiency in vitro. However, the clinical potential of this technology for cardiac regeneration lies not in the ability to reprogram cells in culture, but rather in harnessing the vast pool of resident cardiac fibroblasts for in situ reprogramming into cardiomyocytes that could integrate with pre-existing cardiomyocytes and contribute to force generation of the intact heart. To this end, the recent papers from our group and from the Olson laboratory indicate that the plausibility of this approach^11,15. In these studies, lineage tracing experiments labeling non-myocytes in the heart demonstrated the emergence of new cardiomyocyte-like cells from the non-myocyte population upon expression of GMT or GMHT. Most importantly, both studies suggest that the in vivo microenvironment, likely enriched by secreted signals, components of the extracellular matrix, and mechanical forces, significantly enhances the degree of cardiac reprogramming by GMT or GMHT. The percent of reprogrammed cells that contract with electrical stimulation increased to around 50%¹¹. Furthermore, new cardiomyocyte-like cells reprogrammed from fibroblasts resident within the intact heart could electrically couple with pre-existing cardiomyocytes. Most importantly, adding the reprogramming factors by a gene therapy approach resulted in improved cardiac function and a decrease in total scar area after myocardial infarction in a murine model. Thus, while it remains important to use an cell culture systems to refine cardiac reprogramming technology, it may not be necessary to obtain the optimal type of cardiomyocyte-like cells in culture if the ultimate use of the “recipe” will be in vivo.

To achieve the promise of in vivo reprogramming of tissues to the desired cell types, many challenges remain. In addition to identifying the optimal conditions for generating cardiomyocyte-like cells, efficacy and safety issues will be important to resolve in larger animals. While gene therapy approaches to deliver reprogramming factors may have a viable regulatory path, attempts to replace transcription factors with small molecules and/or secreted proteins would be valuable. Experience from the iPSC field suggests that at least some reprogramming factors can be substituted with small molecules, and that epigenetic regulators may enhance efficiency. As this young field develops, there will be many challenges to overcome, but the recent reports of successful in vitro and in vivo reprogramming by many groups firmly establish the conceptual advance that non-myocytes can be transdifferentiated to cardiomyocyte-like cells capable of contractile performance. Future studies in human cells, development of safe and efficient systems for delivery of reprogramming factors into cells of the heart in situ, and understanding the molecular mechanisms involved in direct cardiac reprogramming are necessary to advance this technology for future clinical applications.

Table. Comparing Methods for Cardiac Reprogramming.

Ieda et al. used retroviral or Dox-inducible lentiviral vectors to express Gata4/Mef2c/Tbx5 (GMT), but performed most experiments and in depth analyses in neonatal cardiac fibroblasts (CFs) with retroviruses⁷. Song et al. used retroviral vectors for Gata4/Mef2c/Hand2/Tbx5 (GMHT) overexpression in adult CFs or tail tip fibroblasts (TTFs)¹¹. Chen et al. used adult CFs/TTFs with lentiviral vectors expressing GMT¹⁴.

	Ieda et al. (GMT)		Song et al. (GMHT)	Chen et al. (GMT)
Vector	Retrovirus	Dox-on lentivurus	Retrovirus	Dox-on lentivirus
Transgenic mouse	αMHC-GFP	αMHC-GFP	αMHC-GFP	αMHC-Cre
				Nkx2.5-Cre
				cTnT-Cre
Cell type	Neonatal/adult CF	Neonatal TTF	Adult TTF/CF	Adult TTF/CF
	Neonatal TTF
Transduction efficiency	>95% (TTF/CF)	80% (TTF)	N.D.	N.D.
		40% (CF)
GMT expression	6–8 fold greater than cardiomyocytes		N.D.	10–1000 fold more than CFs
	60–80 fold greater than CFs
Toxicity	No	Yes	No	N.D.
Cardiac induction	After 1 week by FACS (α-MHC-GFP)		After 1 week by FACS (α-MHC-GFP)	After 3 weeks by FACS
αMHC-GFP	15%	0.5–1%	15–18%	0% (αMHC-Cre reporter)
cTnT	5%	N.D.	9%	35% (cTnT-Cre reporter)
Gene expression analyses
Cell type	αMHC-GFP+ cells	N.D.	CF transduced with GHMT	Tbx5-expressing cells
Array data	Similar to neonatal CMs	N.D.	Cardiac gene upregulation	Similar to TTF/CF
Immunohistochemistry	α-Actinin+	α-Actinin+	α-Actinin+	N.D.
	cTnT+		cTnT+
	ANP+		cTnI+
Function	Action potential	N.D.	Action potential	Ca++ channel–mediated depolarization
	Cell contraction		Cell contraction
	Ca++ transient		Ca++ transient
Cell transplantation	αMHC-GFP+	N.D.	N.D.	Cell death
	α-Actinin+

CM, cardiomyocytes; N.D., not determined.