In this study, we have described the hESC-CM transcriptome within the spectrum of changes that occur between undifferentiated hESCs and fetal heart cells, and used molecular imaging to follow their survival and engraftment in the heart. Global gene expression profiling of hESC differentiation thus enables a systems-based analysis of the biological processes, networks, and genes that drive hESC fate decisions. This systems biology approach has obvious benefits over traditional PCR-based methods, which measure only a limited number of transcripts and so cannot define the complex regulatory networks of genes and pathways important for hESC differentiation.
Previous studies have also analyzed the transcriptional profiles of hESC-CMs 
, and we found a high degree of similarity in the significant gene lists between our results and theirs. For example, we also noted upregulation of cardiac markers (e.g. MYH6, MYL4, TNNT2), cardiac transcription factors (e.g. TBX5, MEF2C, GATA4), as well as phospholamban (PLN). Interestingly, the two previous studies compared only spontaneously beating clusters of non-purified cardiomyocytes (what we refer to as “beating embryoid bodies” in this study). In our experience, only 2–5% of the cells within these clusters actually express the cardiac-specific marker cardiac troponin-T. Because of concern that non-cardiac and non-mesodermal cell types will obscure the hESC-CM molecular signatures, we used Percoll density gradient separation to achieve cardiac troponin-T-enriched populations ranging from 40–45%. Analysis of the expression differences between purified and non-purified CMs revealed considerable downregulation of early mesodermal and homeobox genes, and upregulation of cardiovascular and structural genes such as actins and extracellular collagens. Given this purified population of CMs, we were also able to perform robust bioinformatics analysis of these cells and compare them with fetal heart cardiomyocytes. This is important since we would like to establish how closely our derived cells compare with a gold standard. When looking at the GO processes that are more active in the FH cells, we found a pattern suggesting increased metabolic activity but not structural protein biogenesis. We believe these energy-related pathways are likely necessary for the cardiomyocyte to contract forcefully in the in vivo
environment. This finding suggests that derived hESC-CMs have not adequately matured, at least as far as energy metabolism, and so may benefit from exogenous mechanical or electrical stimulation in order to upregulate energy-related pathways prior to transplantation.
Using purified hESC-CMs, our comprehensive, systems-based approach to transcriptional analyses supports the case that each of the stages of differentiation and selection results in a significant enrichment in cells of the cardiomyocyte lineage, expresses appropriate stage specific genes, and turns on appropriate biological processes corresponding to these stages. Given the robustness of our differentiation method, we believe the hESC-CM population would be an ideal source of replacement cells in the in vivo setting. We also demonstrate that hESC-CMs can successfully engraft in the ischemic heart for an extended duration that result in improved cardiac function, though only transiently. This latter finding may be partly attributed to the activation of paracrine signaling mechanisms by transplanted cells on host cells and themselves, which then attenuates after acute donor cell death. Lastly, we show that cardiac differentiation prior to transplantation can prevent teratoma formation, which remains a major safety concern for investigators exploring the therapeutic uses of hESCs.
In our microarray analysis, we observed high expression of pluripotency-related genes involved in the core hESC regulatory circuitry, including OCT4, SOX2, and NANOG, as well as CRYPTO 1 and 3, LCK, and HESX1. Differentiation into beating EBs was accompanied by mesodermal differentiation and dramatic activation of TWIST1, TBX5, and MEOX transcription, as well as the very clear induction of nearly all of the early cardiogenic genes, including FOXC1, ISL2, HAND1, GATA4, 5, and 6, FOXH1, and MEF2C. While it is clear that other developmental lineages are still present in the EB population, it is also clear from the high levels of cardiac gene expression that this population is significantly enriched for the cardiac lineage even at an early stage. The transcriptional analysis of the final differentiation and selection of the hESC-derived CMs indicates that this enrichment continues, with the CM population expressing differentiated cardiomyocyte genes at levels similar to our more advanced FH cells. Importantly, because of the cell type heterogeneity in the fetal heart, we specifically isolated cardiomyocytes from the fetal left ventricles for microarray analysis.
We now briefly discuss the four major trends in the microarray data seen in the K-means clustering analysis (), which will allow us to explore the major themes within an enormous amount expression data. Cluster 1 is composed of 1775 genes whose expression increases at each stage from hESC to EB to hESC-CM to FH ( and Table S3 (B1)
). Overrepresentation analysis of this cluster of genes shows that the GO processes to which these genes contribute include many basic differentiated cell functions such as the establishment of cellular transport and secretory processes, regulation of cell localization, response to cellular stresses and hypoxia, cytoskeletal biogenesis, control of apoptosis, and interestingly, cardioblast cell fate commitment (Table S4 (B5)
). This cluster of genes has a substantial overlap with the component genes of principal component 1 from the PCA analysis (), demonstrating how two analytic approaches can result in similar significant findings.
The converse expression pattern is seen in the 2,453 genes composing cluster 2, which are sequentially downregulated across the groups from hESC to EB to hESC-CM to FH (Table S3 (B2)
). The processes overrepresented in this cluster primarily involve nucleic acid synthesis, DNA replication and chromatin maintenance, cell cycle, and transcription in general (Table S4 (B6)
). This theme is consistent with patterns seen in normal embryonic development in both drosophila and mouse 
, and reflects the fact that earlier undifferentiated cells are undergoing rapid replication and production of broad ranges of transcripts, while cell cycling slows dramatically later in development as cells begin to express a more limited number of genes that are appropriate for the differentiated state.
Cluster 3 is comprised of 1,009 genes whose expression increases at each stage from hESC to EB to hESC-CM, but which are expressed significantly less in the FH cells (Table S3 (B3)
). The overrepresented processes in this cluster correspond to non-cardiac cell differentiation pathways, particularly neuroectodermal differentiation, that compose a portion of the hESC-CM population which we have differentiated and purified it from hESC precursors, but are not present in the harvested fetal heart cells (Table S4 (B7)
). The final interesting cluster of genes is cluster 4 (Table S3 (B4)
), representing genes which generally increase in expression across all stages from hESC to EB to hESC-CM to FH, but which are expressed at considerably higher levels in the two older FH samples, 3 and 4 (at 20 and 21 weeks, respectively), than in FH1 and 2 (19 weeks each). The processes overrepresented in this gene group are heavily weighted toward cardiac muscle contraction, muscle development, heart development and other cardiac specific processes, and the genes contributing to these processes include dozens of cardiac structural proteins such as cardiac myosin heavy and light chains, cardiomyocyte potassium channels such as KCNE1 
, KCNQ1 
and KCNH2 
, cardiomyocyte troponins including T2 and C1, as well as cardiac phospholamban and cardiac actin 1 (Table S4 (B8)
). Thus the hESC-CM population's expression of terminal cardiac differentiation markers at a level intermediate between younger and older FH cardiomyocytes suggests that this population is sufficiently advanced developmentally to serve as a potential replacement population for cells lost to ischemia.
With these very interesting and detailed gene expression studies, we began focusing on cellular transplantation to the ischemic heart. We observed significant improvements in echocardiographic metrics when comparing treated and control animals. Histologic analysis revealed reduced scar formation, but there was underwhelming evidence of functional myocardium regeneration, confirming previous reports 
. To explain this disparity, the improvement in cardiac function may be due to paracrine factors, as suggested by Dzau and colleagues 
. Our own studies indicate some increased cytokine signaling in hypoxic hESC-CMs (Figure S7
). However, if paracrine signaling is the primary mechanism of improvement, then long-term generation of these factors from sufficient numbers of transplanted hESC-CMs may be required for a sustained improvement in cardiac function. Until now no studies have analyzed the overall survival and growth kinetics of transplanted hESC-CMs in ischemic myocardium.
To address this lack of understanding, we employed molecular imaging technology for understanding the fate of cells following transplantation 
. Longitudinal imaging of transplanted hESC-CMs exposes the limitations of cardiac stem cell therapy, as ~90% of cells die within the first three weeks of delivery. Though we did not address the specific mechanism of death in this study, poor cell survival is likely due to widespread apoptosis and anoikis of cells injected into an inhospitable environment. Improving cell survival by subjecting hESC-CMs to the appropriate anti-apoptotic and pro-survival cues may alleviate some of the survival issues, and efforts to this end have been reported since completion of this work 
. Other methods that take advantage of tissue engineering technologies in which biopolymers and synthetic tissue constructs are used to organize and support transplanted cells may offer another means for increasing cell survival 
. Delivery techniques other than intra-myocardial injection, such as intracoronary or retrograde coronary venous, may also improve cell survival 
. Another confounding factor is the host immune response, which we did not address in this study (as SCID mice were used). With a functioning host immune system, we would expect to see a further reduction in cell survival. Nevertheless, it is important to note that even in our SCID mice, transplantation of Fluc+/eGFP+
hESC-CMs did not
form teratomas in the post-transplantation period. The lack of teratoma formation emphasizes the robustness of our hESC-CM purification protocol in removing undifferentiated cell contaminants.
In summary, hESC-CMs hold potential promise for treatment of cardiovascular disease. The molecular processes that control stem cell pluripotency, differentiation, and proliferation are complex, justifying the need for a broad investigation that integrates systems biological tools for transcriptome analysis with molecular imaging tools for confirmation of survival, engraftment and functional benefit in the in vivo setting. We found that the enriched hESC-CMs expresses cardiomyocyte genes at levels similar to 20-week fetal heart cells, making this population a good source of potential replacement cells in the in vivo setting. Beyond a characterization of the overall transcriptional characteristics of our differentiated cells, we have also identified a large number of potentially important new genes that are expressed at high levels at distinct stages and that may play roles in the cardiogenic developmental program. These genes may also act as specific markers of cell differentiation in addition to being inducers of cardiogenic differentiation, thus opening new avenues of investigation into the basic biology of cardiovascular development. However, understanding the molecular networks of differentiation is not enough to predict the fate of differentiated cells once transplanted in a living host. To address this lack of knowledge, we have shown molecular imaging to be a powerful method for assessing cellular localization, engraftment, survival, and proliferation in vivo. Taken together, gene expression and molecular imaging studies such as this will serve as a crucial foundation for future clinical applications of stem cell therapies.