Global transcriptome analysis of murine embryonic stem cell-derived cardiomyocytes

Microarray analysis reveals that the specific pattern of gene expression in cardiomyocytes derived from embryonic stem cells reflects the biological, physiological and functional processes occurring in mature cardiomyocytes.


Background
Heart failure caused by loss of functional cardiomyocytes represents one of the most common cardiovascular diseases. Elucidation of the genetic networks and intracellular mechanisms that underlie cardiomyocyte development from ES cells is a prerequisite for future cell replacement therapies in heart failure [1,2]. Recently, genetic strategies for differentiation of stem cells and nonmuscle cells through expression of developmental control genes that specify cardiac cell identity have been favoured in cell replacement therapies to regenerate heart muscle tissue [3]. However, a prerequisite for these strategies is identification and an understanding of cardiac cell-specific biologic, physiologic, and molecular processes. To this end, signaling pathways and gene signatures characteristic of cardiomyocytes must be deciphered in order to characterize the cardiomyocytes derived from embryonic stem (ES) cells.
Mouse ES cells can proliferate indefinitely without senescence in vitro in their undifferentiated state in the presence of leukemia inhibitory factor or on a layer of mitotically inactivated mouse embryonic fibroblasts (MEFs) [4]. ES cells can be genetically manipulated with reporter and selection markers to identify and select cardiomyocytes from differentiating ES cells [5][6][7][8]. Most often, protocols to enrich cardiomyocytes from transgenic cardiac cell lines were optimized for ES cell lines such as the D3 cell line cultivated on MEFs. It is well known that several, as yet uncharacterized factors from MEFs have an influence on the differentiation processes of ES cells, necessitating the use of MEF-free ES cells in differentiation studies [9]. Recently, we clearly demonstrated that the first contact with MEFs contaminates ES cells even if they are subsequently cultivated in the absence of MEFs, and the gene expression profile of MEFs interferes with those of ES cells and embryoid bodies (EBs). Even 9-day-old EBs are still contaminated by MEFs, and MEF-specific gene expression is still detectable [9]. Therefore, consistent gene expression and developmental studies on ES cells require MEF-free ES cells.
Although MEF-dependent, ES cell derived cardiomyocytes have been well characterized electrophysiologically [5][6][7][8], the cardiac-specific gene signatures and signaling cascades had not until now been characterized in detail. Even though several attempts have been made, a comprehensive transcriptome analysis of MEF-free murine ES cell derived pure cardiomyocytes is not yet available.
We recently reported an optimized CGR8 ES cell model that permits consistent gene expression and facilitates studies of the early embryonic development [9]. In order to identify all signal transduction pathways and biologic processes in cardiomyocytes, we generated a transgenic cardiomyocyte-specific cell line from CGR8 mouse ES cells and isolated pure cardiomyocytes. Thereafter, large-scale expression studies were performed using Affymetrix expression microarrays covering all known transcripts. Here we report, for the first time, a transcriptome analysis of pure cardiomyocyte preparations from MEF-free ES cells.
Similar to findings mature cardiomyocytes, we demonstrate that cardiomyocytes derived from ES cells strongly express classic genes that are required to accomplish their physiologic function. Interestingly, the genes required for cell proliferation and apoptotic processes are significantly downregulated in ES-derived cardiomyocytes. We may conclude that the identification of 'gene signatures' and signal transduction pathways that are specifically expressed in the α-myosin heavy chain (MHC)-positive cell population will significantly contribute to an understanding of cardiomyocyte-specific physiologic processes.

Results and discussion
Isolation of highly purified α-MHC + cardiomyocytes from the transgenic α-MHC embryonic stem cell line We first generated cardiomyocytes with high purity from a transgenic α-MHC ES cell line. When EBs were formed using the conventional hanging drop method (Figure 1a) during the course of differentiation, the EGFP fluorescence increased significantly after 7 days and the EGFP-expressing cells were first detectable microscopically within the EBs. After 24 hours, the 8-day-old EBs were treated with 4 μg/ml puromycin for a further 7 days. During puromycin treatment the nonpuromycin-resistant cells died, and beating clusters of puromycin-resistant 15-day-old EGFP-expressing α-MHC + cells were progressively enriched (Figure 1a and Additional data files 1 and 2).
Reverse transcription (RT)-polymerase chain reaction (PCR) analysis indicated maximal expression of the α-MHC + gene in the 7-day-old EBs ( Figure 1b; for RT-PCR conditions and primers, see Additional data file 3). The purity of the cardiomyocytes in the 15-day-old untreated EBs (hereafter referred to as 'control EBs') and in the 15-day-old α-MHC + cardiomyocyte EBs was determined by fluorescence-activated cell sorting analysis after dissociation of the cells with trypsin and calculated to be 16.7% (Figure 1c) and 91.2% (Figure 1d), respectively.
The ES cell derived cardiomyocytes exhibited a multi-angular (Figure 1e subpanels a and b), more rectangular (Figure 1e, subpanels c and d), and a triangular morphology (Figure 1e, subpanels e and f). Detection of cardiac α-actinin by immunocytochemistry ( Figure 1e, subpanels b, d and f) clearly indicated the Z-disc specific protein and the characteristic striations of sarcomeric structures of the cardiac cells. The gap junction protein connexin-43 is highly expressed in heart and was detected by immunocytochemistry (Figure 1e, subpanels g and h). Connexin-43 is distributed in the cytosol and in the outer membranes in the cell border regions (Figure 1e, subpanels g and h). expression of single α-MHC + cells with different morphologies (subpanels a, c, and e). Detection of α-cardiac actinin (subpanels b, d, and f) and connexion-43 (subpanels g and h) was performed using anti-cardiac actinin (1:400) and anti-connexin-43 (1:400). Secondary detection was performed with antimouse-IgG 1 -AlexaFluor555 and anti-rabbit-Ig-AlexaFluor647. Hoechst dye was used to stain nuclei. Bars in panel e (subpanels a to f) are 50 μm; bar in panel e (subpanel g) is 20 μm; and bar in panel (subpanel h) is 7.5 μm.

Electrophysiological characterization of the cardiomyocytes
Functional characterization of the α-MHC + cardiac cells was performed by measuring their typical spontaneous action potentials (APs). Spontaneous APs were measured in single α-MHC + cardiomyocytes (n = 32) as well as in multicellular α-MHC + aggregates (n = 24). All APs exhibited parameters characteristic of cardiac APs. The minimal diastolic potential was -60.2 ± 1.1 mV; membrane potentials normally showed a diastolic depolarization, leading to a spontaneous AP frequency of 125.9 ± 8.0 min. The maximal upstroke velocity was 22.9 ± 2.2 V/s, pointing to a contribution of voltage-activated sodium currents, which was confirmed by voltage clamp measurements (data not shown). APD 90 , APD 50 and APD 20 (AP duration from maximum to 90%, 50% and 20% repolarization) were 96.4 ± 4.2 ms, 71.1 ± 3.9 ms and 41.3 ± 2.6 ms, respectively. APs exhibited a variety of morphologies, including pacemaker-like, atrial-like, and ventricular-like APs ( Figure 2). In most cases, however, morphologic properties did not match any type of specific differentiation. These unspecified APs mostly possessed a plateau phase, but had a much shorter APD 90 than ventricular APs, which are characterized by a long APD 90 of about 200 ms [7].
To characterize the hormonal regulation of α-MHC cardiomyocytes, carbachol (an agonist of m-cholinoceptors) and isoproterenol (an agonist of β 1 adrenoceptors) were applied (Figure 2b,c). Carbachol at 1 μmol/l decreased the AP frequency significantly, to 44.8 ± 7.5% of control values (n = 20; the frequency under control conditions was determined for each recording and set to 100; Figure 2d). Isoproterenol at 1 μmol/l evoked a significant increase in frequency to 238.58 ± 23.7% of control values (n = 19; Figure 2e). Intracellular recordings of spontaneous APs revealed typical cardiac AP parameters and morphologies, confirming the cardiac differentiation and functionality of puromycin-selected α-MHC + cells. Muscarinic and adrenergic regulation of the AP frequency, which is estabished for ES cell derived cardiomyocytes [10] as well as for native murine cardiomyocytes at early developmental stages [11], further supports a physiologic cardiac differentiation of α-MHC cells. As described previously [12,13], APs at the intermediate developmental stage exhibited diastolic depolarizations and diverse shapes. APs with a distinct plateau phase were frequent but considered to be unspecific rather than ventricular-like in the majority of cases, because the APD 90 was much shorter than reported for early-stage murine ventricular cardiomyocytes as well as for murine ES cell derived ventricular-like cardiomyocytes [7]. Because most APs had unspecific morphologic properties, a general classification into pacemaker-like, atrial-like, and ventricular-like APs could not be done, which accords well with previous findings from intermediate-stage ES cell derived cardiomyocytes [12,13]. Only few APs exhibited typical morphologic features of the respective differentiation types.
It was recently reported that, in ES cell derived cardiomyocytes expressing green fluorescent protein under control of the α-MHC promotor, green fluorescence is restricted to pacemaker-like and atrial-like cells [7]. Because we found puromycin-purified α-MHC + cardiomyocytes with a ventricular-like AP morphology in few cases, our data suggest that α-MHC expression is not completely absent in ES cell derived ventricular-like cardiomyocytes. This apparent discrepancy might arise from the complex stage-dependent expression pattern described for α-MHC in murine EBs [14] and murine embryonic ventricles [15], because a different developmental stage of ES cell derived cardiomyocytes was investigated in the present study (15-day-old cardiomyocytes) as compared with that in the study conducted by Kolossov and coworkers [7] (9-day-old to 11-day-old cardiomyocytes).

Validation of the microarray data by quantitative realtime PCR and semiquantitative RT-PCR analyses
RNA from α-MHC ES cells, 15-day old α-MHC + cells, and control EBs was used as a template for hybridizations to Affymetrix MG 430 v2.0 oligonucleotide microarrays (RNA was obtained from three independent experiments) (Affymetrix UK Ltd., High Wycombe, UK). Raw expression data were RMA normalized [16]. We verified the Affymetrix data by examining the expression levels of five randomly chosen representative genes (Nanog, T Brachyury, Bmp2, Sox17, and α-MHC) applying the quantitative real-time PCR (qPCR) method ( Figure 3a). Additionally, expression levels of randomly chosen genes, such as Troponin T, Myocardin, α-MHC, Mef2C, Nkx2.5, MLC-2v, and AFP were verified by semiquantitative RT-PCR analysis ( Figure 3b). As indicated, the expression levels of the late cardiomyocyte markers α-MHC and MLC2v and the early cardiac marker Nkx2.5 were higher in the 15-day-old α-MHC + cardiomyocytes as compared with cells in the 15-day-control EBs. Not surprisingly, expression of α-fetoprotein (a marker of cell types of endodermal origin, such as liver cells) was absent in the cardiomyocyte clusters but not in the 15-day-old control EBs. As indicated in Figure 3 panels a and b, results from the Affymetrix analyses clearly correspond to the results obtained from the qPCR and semiquantitative RT-PCR analyses, respectively. Note that RNA used in Figure 3a for qPCR validation was isolated from set of experiments other than that used in Figure 3b.

Selected Gene Ontology Biologic Process annotations of genes differentially expressed in α-MHC + cardiomyocytes
Pair-wise comparisons between experimental conditions were performed on RMA-normalized data using Student's ttest (unpaired, assuming unequal variance). In order to identify transcripts with an α-MHC + cell specific expression pattern, a three-condition comparative analysis of the α-MHC + cells versus control EBs and versus α-MHC ES cells was made (intersection of genes differentially expressed between undifferentiated α-MHC ES cells and α-MHC + EBs, as well as dif-    Two subclusters were identified. Subcluster A (196 genes) includes genes with low expression level in undifferentiated cells, moderate expression in the control EBs, and high expression levels in the α-MHC + cardiomyocytes. Cluster B (455 genes) includes genes with low expression in both control EBs and undifferentiated ES cells but with higher expression in α-MHC + cells. Interestingly, the expression level of a subset of genes (highlighted in Figure 5b) was higher in undifferentiated cells as compared with control EBs but lower than that in α-MHC + cells.
Validation of Affymetrix data by quantitative real-time PCR and semi-quantitative PCR analyses Figure 3 (see previous page) Validation of Affymetrix data by quantitative real-time PCR and semi-quantitative PCR analyses. (a) Validation of Affymetrix data by quantitative real-time polymerase chain reaction (PCR) analyses. The fold change was calculated by using the following formula: fold-change = . ΔC t of the gene in the sample in which it is expressed lowest is taken as ΔC t gene2 to calculate the fold change using the above formula. The resulting fold change is expressed as percentage of the maximum fold change (= 100%) for that particular gene in every assay. Values are expressed as mean ± standard deviation (n = 3; technical replicates). (b) Additional validation of Affymetrix data by semi-quantitative reverse transcription (RT)-PCR analyses. Randomly chosen candidate genes to validate Affymetrix data by semi-quantitiative RT-PCR analyses and their relative expression values expressed as percentage of maximum expression for every gene, as obtained from Affymetrix profiling, are given in the table.
The 'oxidative phosphorylation' KEGG pathway is associated with aerobic energy production (also see below) whereas calcium is the second messenger regulating several physiologic processes such as contractility in cardiomyocytes [20]. Additional data file 6 (part a) shows the gene expression level changes of transcripts belonging to the oxidative phosphorylation KEGG pathway and the corresponding signal transduction scheme. Additional data file 6 (part b) shows the upregulated probe sets of the GO categories that participate in aerobic energy production and their increase in expression level. In general, the dependence of cardiac homeostasis on mitochondria is primarily attributed to the ATP derived from oxidative phosphorylation for maintaining myocardial con-tractility [21]. Genes in these categories are 'classical' for cardiomyocytes and essential for aerobic oxygen dependent energy production for intact heart function. Mammalian heart muscle cells fail to produce enough energy under anaerobic conditions to maintain essential cellular processes. Because the mammalian heart is an obligate aerobic organ that consumes oxygen intensively [22], a constant supply of oxygen is indispensable for sustaining cardiac function and viability. This notion is well elucidated by our analysis, indicating that genes involved in aerobic energy production are upregulated in the α-MHC + cardiomyocytes.
We also found the fatty acid metabolism GO category to be enriched in the genes over-expressed in the α-MHC + cardiomyocytes (Additional data file 6 [part c]). These findings are consistent with the fact that β-oxidation of fatty acids in mitochondria accounts for the vast majority of ATP generation and therefore is the preferred substrate in the adult myocardium, which supplies about 70% of total ATP (for review, see Huss and Kelly [23]). Defects in mitochondrial fatty acid transport and fatty acid oxidation result in sudden cardiac death, bioenergetic dysfunction, cardiac arrhythmias, and cardiomyopathy [21].
Genes that are specifically expressed in α-MHC + cells participate in multiple signal transduction pathways. Additional data file 7 (parts a and b) show that the genes belonging to the 'enzyme linked receptor protein signaling pathway' and to 'protein kinase activity' categories are over-expressed in ES cell derived cardiomyocytes. Among these genes, the phosphatidylinositol 3-kinase (Additional data file 7 [part a]) and the Wnt inhibitory factor 1 (Additional data file 7 [part b]) and several other kinases participate in key biologic signal transduction pathways.
Interestingly, five genes belonging to the category 'negative regulation of Wnt receptor signaling pathway' and four genes belonging to 'p38 mitogen-activated protein kinase signaling' were found to be upregulated in the α-MHC + cells (see Additional data file 7 [parts c and d]). In recent years, regulation of Wnt signal transduction has been discussed as an important event that initiates cardiac development (for review, see Eisenberg and Eisenberg [24]). It has been demonstrated that the Wnt inhibitors Dkk1 and crescent induce cardiogenesis, suggesting that Wnts actively inhibit cardiogenesis.

Gene Ontology enrichment analysis of the downregulated genes in α-MHC + cardiomyocytes
In order to identify transcripts that are specifically downregulated in α-MHC + cells, a three-condition comparative analysis of the α-MHC + cells versus control EBs and versus α-MHC ES cells was made (intersection of genes differentially expressed between undifferentiated α-MHC ES cells and α-MHC + EBs as well as differentially expressed between 15-dayold control EBs and α-MHC + EBs at t-test P < 0.01, fold change <2 for both comparisons).
Downregulated genes are grouped into three main subclusters. In all subclusters the gene expression level is lower in α-MHC + cells (last three lanes) compared with those in the other two cell populations. Subcluster A contains probe sets with a higher expression level in undifferentiated ES cells compared with the control EBs, whereas cluster C shows the genes with a similar expression level in the two cell populations ( Figure 6; gene names are given in Additional data file 8). Only a small cluster contains genes with a higher expression level in control EBs compared with the undifferentiated ES cells (Figure 6b). Table 2 shows the GO categories, and the KEGG and Biocarta pathways that are enriched among the probe sets downregulated in the α-MHC + cardiac cell population compared with the control EBs and compared with the undifferentiated α-MHC ES cells. Genes belonging to eight KEGG and 10 Biocarta pathways were identified as being downregulated in the ES cell derived cardiomyocytes (Table 2). Of particular interest are the genes that are associated with cell proliferation belonging to the KEGG pathway 'cell cycle' (32 genes) and to the 10 Biocarta pathways, as well as genes belonging to the GO category 'positive regulation of programmed cell death', because both the proliferation and apoptosis process are of great physiologic relevance. In accordance with the findings from the KEGG and Biocarta pathways, the GO category 'regulation of cell cycle' (also includes genes of the mitotic cell cycle and the M-phase GO categories) is enriched among the transcripts downregulated in the ES derived cardiomyocytes. This category contains genes that are involved in regulating cell proliferation.
Additional data file 9 (part a) indicates the downregulated genes that belong to the KEGG pathway 'cell cycle' and to the GO category 'cell cycle'. As indicated several cyclins (cyclin A 1 , cyclin A 2 , cyclin B 1 , cyclin B 2 , cyclin D 1 , cyclin D 2 , and cyclin E 1 ) are downregulated in the ES cell derived cardiomyocytes. Fully differentiated cardiac myocytes are postmitotic cells, and regulation of cell size is their main mechanism of adaptation to variations in workload in the heart [27]. Increase in cell size of the cardiomyocytes due to increased protein synthesis without cell division is a characteristic phenomenon for cardiac cells and is referred as myocyte hypertrophy. These findings are consistent with the finding that 'cell growth' genes are upregulated in the cardiac cell population (Additional data file 7 [part f]). In agreement with these cardiac cell specific mechanisms, we found that -compared with the mixed cell population -genes belonging to the 'positive regulation of programmed cell death' (apoptosis) and to the 'apoptosis' GO categories are downregulated in the cardiac cell population (Additional data file 9 [part c]). Markedly, the expression level of the apoptosis-inducing gene (mainly in cancer cells) breast cancer 1 [28] was 13-fold lower in α-MHC + cardiomyocytes than in the mixed cell population of control EBs. In addition, it is well known that cardiac myocytes have developed remarkable mechanisms of cytoprotection and cell survival, also because of the absence of cell division [27].
To determine whether the identified gene signatures are specific to α-MHC + cardiomyocytes and are not due to puromycin treatment, we generated a transgenic β-actin ES cell line expressing both puromycin resistance and EGFP cassettes under the control of the β-actin promoter. After application of the differentiation protocol used for isolation of the α-MHC + cardiomyocytes (Figure 1), RNA was isolated from 15-day-old

Functional annotations enriched among transcripts that are downregulated in α-MHC positive cells
EBs either in the presence or in the absence of puromycin. Because β-actin is constitutively highly expressed in all cell types, the β-actin EBs were resistant to puromycin. Accordingly, no morphologic alterations were observed compared with 15-day-old control untreated EBs.
Analysis of the differentially expressed genes in the 15-dayold β-actin EBs in the absence of puromycin in comparison with the 15-day-old β-actin EBs treated with 4 μg/ml puromycin resulted in 554 differentially expressed probe sets (at ttest P < 0.01, fold change >2). Among the 554 probe sets, 31 are found also to be differentially expressed in the α-MHC + cardiomyocytes. Interestingly, all of the 31 probe sets are downregulated in the 15-day-old β-actin EBs (Additional data file 10). Although 20 of these probe sets are downregulated in the 15-day-old β-actin EBs because of puromycin treatment, they are upregulated in the α-MHC + cardiomyocytes enriched by puromycin treatment. Therefore, these 20 genes can be considered to be cardiomyocyte specific.
A group of 11 probe sets corresponding to 10 genes is downregulated in puromycin treated β-actin EBs and in cardiomyocytes. However, in comparison with 1,845 probe sets specifically regulated in cardiomyocytes, 10 genes account for a negligible fraction. Hence, our expression profiling remains of importance within the context of defining the cardiomyocyte specific transcriptome.

Conclusion
After the completion of several mammalian genome sequencing projects, the next challenge will be to identify the complete cell specific transcriptome. Identification of the complete transcriptome of the ES cell derived somatic cells will contribute to our understanding of the developmental processes that lead from undifferentiated ES cells to tissuespecific cells. Here, we identified all α-MHC + cardiomyocyte specific gene expression signatures, including signal transduction pathways that are characteristic for cardiomyocytes. From our results, we conclude that -similar to mature cardiomyocytes -cardiomyocytes derived from ES cells strongly express genes required for regulation of respiratory energy metabolism, cytoskeleton, and ion channels; these genes are required for cardiomyocytes to fulfil their physiologic function. Interestingly, as in postmitotic adult cardiomyocytes, genes required for cell proliferation are significantly downregulated, whereas genes involved in the process of cell growth were significantly upregulated in the ES cell derived cardiomyocytes. Moreover, genes that are involved in the process of apoptosis were significantly downregulated in the ES cell derived cardiomyocytes. Finally, identification of the gene signatures and signal transduction pathways that are specifically expressed in the α-MHC + cell population will contribute to establishment of a gene expression atlas for cardiomyocytes and will be useful for further studies to elucidate their role during developmental processes.

Embryonic stem cell culture and differentiation of embryoid bodies
CGR8 ES cells (ECACC 95011018) were cultured without feeder cells in Glasgow minimum essential medium (GMEM) supplemented with 10% fetal bovine serum, 2 mmol/l Lglutamine, 100 units/ml leukemia inhibitory factor, and 50 μmol/l β-mercaptoethanol (ME) in 0.2% gelatine coated flasks, as described previously [9]. The cells were passaged every other day and care was taken to maintain the confluency below 70%. The cells were passaged at least four times before use in the experiments to attain a stable gene expression profile. To induce differentiation, the hanging drop protocol was used, as described previously [9]. Briefly, an ES cell suspension of 2.5 × 10 4 cells/ml was prepared in Iscove's modified Dulbecco's Medium (IMDM) supplemented with 20% fetal calf serum, 1% non-essential amino acids (vol/vol), 2 mmol/l L-glutamine, and 100 μmol/l β-ME. Of this ES cell suspension, 20 μl was spotted on the inside of the upper lid of a 10 cm bacteriologic dish and then covered over its bottom dish containing 5 ml phosphate-buffered saline. On day 2, the formed multicellular aggregates (EBs) were transferred to suspension in a new dish with 20 ml IMDM supplemented with 20% fetal calf serum, 1% non-essential amino acids (vol/ vol), 2 mmol/l L-glutamine, and 100 μmol/l β-ME. On day 8, contracting clusters were observed. At this stage, cultures were either treated with puromycin (4 μg/ml) or left without treatment for 7 days with a medium change on every other day. On day 15, the beating clusters were trypsinized and used for fluorescence-activated cell sorting analysis or for RNA extraction.

Vector constructs and cell line generation
pIRES2 EGFP was purchased from Clontech (Heidelberg, Germany). The human cytomegalovirus (CMV) immediate early promoter and enhancer were removed by a double digestion with NheI and AseI and then subsequently blunt end ligated to obtain pIRES2 EGFP Δ CMV. Puromycin Nacetyl transferase cDNA was flanked on both ends by BamHI restriction sites, which was PCR-amplified from pIRESPuro3 by Pfu DNA polymerase (Promega, Germany) and inserted at the SmaI site of the pIRES2 EGFP Δ CMV construct to obtain pPuro IRES2 EGFP. The 5.5 kilobase α-MHC promoter construct was a kind gift from Dr J Robbins (Cincinnati, OH, USA) [29]. The promoter was klenowed on one end and the other end digested with SacI.

Quantitative real-time PCR
Validation of the Affymetrix data was performed by qPCR analysis with the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA, USA). A total of 100 ng RNA of α-MHC ES cells, 15-day-old control Ebs, and α-MHC + cells were reverse transcribed with ThermoScript™ Reverse Transcriptase (Invitrogen). Then, real-time PCR was perfomed in triplicates for every sample using TaqMan Gene Expression Assays (Applied Biosystems). The Gene Expression Assays used for validation were Brachyury (T) expression is used as ΔC t gene2 to calculate the fold change using the above formula. The resulting fold change is expressed as percentage of the maximum fold change.

Flow cytometry
A single cell suspension was prepared by trypsinization. Cell clumps were removed by passing through cell strainer cap of a round bottom tube from Falcon ® (Becton Dickinson GmbH, Heidelberg,, Germany). Propidium iodide (PI) staining (Sigma) was included to exclude dead cells. Acquisition of 10,000 live (PI-negative) cells was made with FACscan (BD Biosciences) and the data were analyzed using the CellQuest software (BD Biosciences). The respective wild-type EBs on the same day as the sample EBs were used as controls.

Affymetrix analysis
Total RNA was extracted from undifferentiated ES cells and EBs using the RNeasy total RNA isolation kit (Qiagen GmbH, Hilden, Germany). The preparation quality was assessed by agarose-formaldehyde gel electrophoresis. Three independent total RNA preparations, each 1 μg from the α-MHC + cells, from the mixed cell population and from the undifferentiated ES cells, were labelled with the One-Cycle Target Labeling and Control Reagent package (Affymetrix), as described in the manufacturer's instructions. Briefly, double-stranded cDNA was synthesized using the one-cycle cDNA synthesis module. Biotinylated cRNA was synthesized using the IVT labeling kit and cleaned up using the sample cleanup module.
After fragmenting of the cRNA for target preparation using the standard Affymetrix protocol, 15 μg fragmented cRNA was hybridized for 16 hours at 45°C to Mouse Genome 430 2.0 arrays, which carry probes representing 45,101 probe sets. Following hybridization, arrays were washed and stained with streptavidin-phycoerythrin in the Affymetrix Fluidics Station 450 and further scanned using the Affymetrix Gene-Chip Scanner 3000 7G. The image data were analyzed with GCOS 1.4 using Affymetrix default analysis settings. After RMA normalization [31], three pair-wise comparisons were performed using Student's t-test (unpaired, assuming unequal variances). A Student's t-test P value < 10 -2 and a fold change >2 between two conditions was used to define differential expression. An intersection of upregulation in α-MHC + cardiomyocytes (twofold; t-test P value < 0.01) compared with control cells in the 15-day-old EBs (d15) and compared with undifferentiated α-MHC ES cells (d0) was used to identify transcripts upregulated in α-MHC + cardiomyocytes. For downregulation, an intersection of downregulation in α-MHC + cardiomyocytes (twofold, t-test P value < 0.01) compared with control cells in the 15 day control EBs (d15) and compared with undifferentiated α-MHC ES cells (d0) was used.
Hierarchical clustering was done using Cluster version 3.0 [32,33] applying mean centering and normalization of genes before average linkage clustering with uncentered correlation.

Functional annotation
Differentially expressed genes were analyzed according to predefined pathways and functional categories annotated by KEGG [34], Biocarta [35], and GO [36] using the DAVID bioinformatic resource [37]. For an over-represented GO or KEGG pathway, a cutoff P value of 0.01 has been selected. In general, it should be noted that one gene can participate in more than one KEGG or Biocarta pathway and GO category.

Additional data files
The following additional data are available with the online version of this paper. Additional data file 1 provides a video clip showing the 15-day-old untreated EBs. Additional data file 2 provides a video clip showing the 15-day-old puromycin treated EBs. Additional data file 3 provides the RT-PCR conditions and primers used in the RT-PCR experiments. Additional data file 4 provides the lists of probe sets used for the subclusters A and B, as identified in the hierarchical clustering of probe sets upregulated in α-MHC + cells ( Figure 5). Additional data file 5 summarizes genes of various GO categories that are upregulated in the α-MHC + cardiomyocytes compared with control cells in the 15-day-old EBs and compared with undifferentiated α-MHC ES cells. Additional data file 6 summarizes genes belonging to the KEGG pathway 'oxidative phosphorylation' and various GO categories that are upregulated in α-MHC + cardiomyocytes as compared with control cells in the 15-day-old EBs and compaed with undifferentiated α-MHC ES cells; also provided is a schematic of the KEGG oxidative phosphorylation pathway. Additional data file 7 summarizes genes belonging to various GO categories and the Biocarta pathway 'p38 mitogen-activated protein kinase signaling' that are upregulated in the α-MHC + cardiomyocytes compared with control cells in the 15-day-old EBs and compared with undifferentiated α-MHC ES cells. Additional data file 8 provides lists of probe sets for the subclusters A, B, and C, as identified in the hierarchical clustering of probe sets downregulated in α-MHC + cells ( Figure  6). Additional data file 9 summarizes genes belonging to the KEGG pathway 'cell cycle', various GO categories, and Biocarta Pathways 'G1/S checkpoint' and 'G2/M checkpoint' that are downregulated in α-MHC + cardiomyocytes compared with control cells in the 15-day-old control EBs and compared with undifferentiated α-MHC ES cells. Additional data file 10 lists probe sets both differentially regulated by puromycin treatment in 15-day-old β-actin EBs and transcripts upregulated in α-MHC + cells compared with the control cells in the 15-day-old EBs and compared with undifferentiated α-MHC ES cells. Additional data file 11 lists probe sets differentially regulated in the puromycin-treated 15-day-old β-actin + EBs. Additional data file 12 provides raw data for the α-MHC experiments (part 1). Additional data file 13 provides raw data