Genome-Wide Transcriptional Profiling of ... - Wiley Online Library

STEM CELL GENETICS AND GENOMICS Genome-Wide Transcriptional Profiling of Human Embryonic Stem Cells Differentiating to Cardiomyocytes ABDELAZIZ BEQQALI,a JANTINE KLOOTS,a DORIEN WARD-VAN OOSTWAARD,a,b CHRISTINE MUMMERY,a,b ROBERT PASSIERa,b a

Hubrecht Laboratory, Utrecht, The Netherlands; bInteruniversity Cardiology Institute of the Netherlands, Utrecht, The Netherlands

Key Words. Human embryonic stem cells • Cardiomyocytes • Microarrays • Transcriptional profiling • Gene expression Differentiation

ABSTRACT Mammals are unable to regenerate their heart after major cardiomyocyte loss caused by myocardial infarction. Human embryonic stem cells (hESCs) can give rise to functional cardiomyocytes and therefore have exciting potential as a source of cells for replacement therapy. Understanding the molecular regulation of cardiomyocyte differentiation from stem cells is crucial for the stepwise enhancement and scaling of cardiomyocyte production that will be necessary for transplantation therapy. Our novel hESC differentiation protocol is now efficient enough for meaningful genome-wide transcriptional profiling by microarray technology of hESCs, differentiating

toward cardiomyocytes. Here, we have identified and validated time-dependent gene expression patterns and shown a reflection of early embryonic events; induction of genes of the primary mesoderm and endodermal lineages is followed by those of cardiac progenitor cells and fetal cardiomyocytes in consecutive waves of known and novel genes. Collectively, these results permit enhancement of stepwise differentiation and facilitate isolation and expansion of cardiac progenitor cells. Furthermore, these genes may provide new clinically relevant clues for identifying causes of congenital heart defects. STEM CELLS 2006;24: 1956 –1967

INTRODUCTION

transcriptomes of mESCs with hESCs [7, 8], multiple hESC lines [9 –13], and undifferentiated hESCs (UH) with differentiated cells in “embryoid bodies” (EBs) [14 –18]. Although these studies showed that established stem cell markers, such as OCT4, NANOG, TDGF1, and UTF1, were upregulated in UH, other upregulated genes associated with UH showed relatively little overlap. This has been the reason for a recent International Stem Cell Initiative [19] to compare up to 75 cell lines under standard growth conditions using standard assays. Only one study to date has described the transcriptome of hESCs undergoing directed differentiation specifically to hepatic and neuronal lineages [15]. This is likely due to the relatively poor efficiency of many differentiation protocols. We [20 –23] and others [22, 23] have described various methods for inducing differentiation of hESCs to cardiomyocytes, which all lead to the characteristic formation of beating cell aggregates. In general, however, the proportion of cardiomyocytes within these aggregates is either low or not determined. Recently, we demonstrated significantly improved cardiomyocyte differentiation

Embryonic stem cells (ESCs) are pluripotent cells derived from the inner cell mass of embryos at the blastocyst stage of development. They are characterized by their capacity to self-renew and potential to differentiate into all cells of the adult individual. Since derivation of the first human embryonic stem cell (hESC) lines in 1998 [1], many hESC lines have been isolated into culture [2, 3]. Although they have many features in common with each other and with mouse ESCs, particularly with respect to specific transcription factors, they are clearly not identical in terms of their differentiation capacity in vitro [4]. To identify stem cell-specific features and create a molecular signature for “stemness,” transcriptional profiling has been carried out on different stem cell populations. Two independent studies first compared the transcriptional profiles of mouse embryonic stem cells (mESCs), mouse neural cells, and hematopoietic stem cells [5, 6]. Although more than 200 stem cellenriched genes were found, only six genes overlapped between the two studies. Since then, several studies have compared the

Correspondence: Christine Mummery, Ph.D., Hubrecht Laboratory, Uppsalalaan 8, 3584 CT Utrecht, The Netherlands. Telephone: ⫹31-30-2121800; Fax: ⫹31-30-2516464; e-mail: [email protected] Received January 26, 2006; accepted for publication April 28, 2006; first published online in STEM CELLS EXPRESS May 4, 2006. ©AlphaMed Press 1066-5099/2006/$20.00/0 doi: 10.1634/stemcells.2006-0054

STEM CELLS 2006;24:1956 –1967 www.StemCells.com

Beqqali, Kloots, Ward-van Oostwaard et al.

1957

by lowering the concentration of fetal calf serum (FCS) [24] in a culture system in which hESCs are cocultured with a visceral endoderm-like cell line, END-2. The number of beating areas in the hESC-END-2 cocultures was 24-fold higher in the absence of FCS than in the presence of 20% FCS and up to 20% of the cells within the aggregates were hESC-derived cardiomyocytes (hESC-CMs). This improved efficiency made it then feasible to study the transcriptome of hESCs differentiating toward the cardiac lineage. Here, we describe the temporal gene expression profiles of hESCs during their directed differentiation toward cardiomyocytes by whole-genome microarray analysis. mRNA from human fetal heart (HFH) of 16 weeks of gestation (after voluntary abortus provocatis) was included as a reference source, allowing us to compare gene expression profiles of hESC-CMs with that of HFH. By microarray cluster analysis, genes were identified that were enriched in UH, transiently expressed during early differentiation, and associated with mesoderm or endoderm differentiation, or genes were identified that were enriched in hESC-CMs and HFH. The specificity of the candidate genes was confirmed by reverse transcription-polymerase chain reaction (RT-PCR), real-time RT-PCR, and whole-mount in situ hybridization (ISH) on hESC colonies or on hESC-CMcontaining aggregates, mouse embryos, and HFHs. This study demonstrates that directed differentiation of hESCs toward cardiomyocytes can be made sufficiently efficient to provide an excellent model for understanding commitment to the mesoderm or endoderm lineage and the early steps in human cardiac cell differentiation. Furthermore, it is an approach to data mining to identify new genes that may be associated with (abnormal) human heart development.

Primary Human Fetal Cardiomyocytes

MATERIALS

Probe Preparation and Hybridization of Agilent DNA Microarrays

AND

METHODS

Cell Culture END-2 cells and HES-2 cells were cultured as described previously [21]. To initiate cocultures, END-2 cell cultures treated for 3 hours with mitomycin C (Sigma, St. Louis, http://www. sigmaaldrich.com) (10 ␮g/ml) replaced mouse embryonic fibroblasts (MEFs) as feeders for hESCs. hESC-END-2 cocultures are grown in 12-well plates in Dulbecco’s modified Eagle’s medium containing L-glutamine, insulin-transferrin-selenium, nonessential amino acids, 90 ␮M ␤-mercaptoethanol, and penicillin/streptomycin in the absence of fetal calf serum [24]. At 1, 3, 6, 9, and 12 days (d) after the start of hESC-END-2 coculture (passages 79 and 80), differentiated hESCs were carefully dissected from the END-2 cell layer to avoid contamination of END-2 cells. Beating areas normally appear from days 6 –7 onwards and were included in the 9- and 12-day time points. Samples were collected for RNA isolation. For the isolation of undifferentiated hESCs (UH), only undifferentiated HES-2 colonies based on morphology (passages 46, 79, and 82) were removed from MEFs by dispase (10 mg/ml) treatment. After 2 minutes of incubation, HES-2 colonies were lifted and washed twice in phosphate-buffered saline (PBS) and collected for RNA isolation. For the verification of the microarray data by RT-PCR or real-time quantitative PCR (qPCR), different independent samples were used. www.StemCells.com

Primary tissue was obtained after abortion after individual permission had been obtained by use of standard informed consent procedures and approval of the ethics committee of the University Medical Center, Utrecht. Human fetal hearts (HFHs) (11–16 weeks of development) were directly frozen in liquid nitrogen and fixed overnight in 2% paraformaldehyde or perfused by Langendorff’s method, followed by culture of cardiac cells. After 4 days in culture, fetal cardiac cells were washed with PBS and collected for RNA isolation.

RNA Isolation and Sample Preparation Total RNA was isolated using TRIzol (Invitrogen, Carlsbad, CA, http://www.invitrogen.com), following the manufacturers’ protocol. The quality of all RNA preparations was confirmed prior to microarray hybridization with an Agilent Bioanalyzer 2100 (Agilent Technologies, Palo Alto, CA, http://www.agilent. com) using the RNA 6000 Nano Labchip kit (Agilent Technologies). Equal amounts of total RNA from UH, all time points from hESC-END-2 differentiation (1, 3, 6, 9, and 12 days), and HFH were combined to create a common reference pool (CRP). All microarray hybridizations were performed against this CRP. The use of a CRP has several advantages. First, by comparing each sample directly to the CRP, it is possible to compare relative gene expression in all other samples. Second, it prevents loss of data for genes that are expressed in one sample but not in another. Third, it is possible to extend the data set at a later date by hybridizing a new sample to the CRP, allowing comparisons for the expression of the whole genome to all other samples. Furthermore, we hybridized UH with 12-day differentiated hESCs and performed dye swap for this experiment.

Agilent 44K Whole Human genome arrays (G4112A; Agilent Technologies) were prepared and hybridized with linearly amplified and labeled total RNA following the manufacturer’s protocol. Briefly, probes were prepared from 500 ng of total RNA by linear amplification and cyanine-3 (cy-3) or cyanine-5 (cy-5) labeling of cRNA using an Agilent Low Input Linear Amplification and Labeling kit. cy-3-CTP (10 mM) and cy-5CTP (10 mM) were obtained from PerkinElmer/NEN (PerkinElmer Life and Analytical Sciences, Boston, http://www. perkinelmer.com). Fluorescent labeled probes were purified using Qiagen RNeasy Mini kit (Qiagen, Hilden, Germany, http:// www1.qiagen.com) as described by the manufacturer. Dye incorporation was confirmed using UV spectrophotometer analysis. According to Agilent’s ISH Kit Plus protocol, we used 0.75 ␮g of labeled cRNA per sample for 17 hours of hybridization at 60°C. After the coverslips were removed in 6⫻ sodium chloride/sodium phosphate/EDTA (SSPE) ⫹ 0.005% N-laurylsarcosine, the slides were washed in 6⫻ SSPE ⫹ 0.005% Nlaurylsarcosine for 1 minute, washed in 0.06⫻ SSPE ⫹ 0.005% N-laurylsarcosine for 1 minute, and washed in Agilent stabilization and drying solution for 30 seconds. This wash is to delay degradation of cy-5 dye, which is susceptible to ozone and conveniently dries the slides. Arrays were scanned using the Agilent DNA microarray scanner.

hESC Differentiation Toward Cardiomyocytes

1958

Data Analysis/Statistics DNA microarray images are processed using Feature Extraction software. The processing that Feature Extraction does involves finding the spots, analyzing the spots (pixel population statistics), analyzing the spot population (feature population statistics), background subtraction, dye normalization (combination of linear and locally weighted linear regression), and then calculating ratios from the processed signals with error values and p values. The signal intensity from the red channel (cy-5) is divided by the signal of the green channel (cy-3), and the scale is compressed by applying log to base 10 (log10) for easier visualization. Luminator from Rosetta Biosoftware (Seattle, http://www.rosettabio.com) was used to analyze the microarray data by clustering using an agglomerative algorithm, trending, and comparing profiles at a p value of ⱕ0.01. Additional microarray software analysis was performed using Genespring version 7.2 software (Agilent Technologies).

RT-PCR

Total RNA (1 ␮g) was transcribed by reverse transcriptase (SuperScript II; Invitrogen) and used for PCR using Silverstar DNA polymerase (Eurogentec, Seraing, Belgium, http://www. eurogentec.be). For each PCR, 1 ␮l of cDNA was used in a reaction volume of 50 ␮l. The cycling parameters were 94°C for 15 seconds; 58°C for 30 seconds; and 72°C for 45 seconds, for 35 cycles. The PCR cycles were preceded by an initial denaturation of 2 minutes at 94°C and followed by a final extension of 7 minutes at 72°C. Primer sequences for PCR are given in supplementary online Table 1A and 1B. Products were analyzed on ethidium bromide-stained 1.5% agarose gel. Human acidic ribosomal phosphoprotein (hARP) was used as RNA input control. HARP levels did not significantly change during differentiation (data not shown). As a negative control, total RNA was used directly for PCR.

Real-Time qPCR Real-time PCR was performed according to standard protocols on a MyIQ single-color real-time PCR detection system (BioRad, Hercules, CA, http://www.bio-rad.com). Briefly, 1 ␮g of total RNA was DNase-treated and transcribed to cDNA. Ten ␮l of a 1/10 dilution of cDNA was then added to 12.5 ␮l of the 2⫻ SYBR green PCR master mix (Applied BioSystems, Foster City, CA, http://www.appliedbiosystems.com) and 500 ␮M of each primer. PCR cycles were 3 minutes at 95°C, followed by 40 cycles of 15 seconds at 95°C, 30 seconds at a specific annealing temperature, and 45 seconds at 72°C. The thermal denaturation protocol was run at the end of PCR to determine the number of products. Samples were run on a 2% agarose gel to confirm the correct size of the PCR products. All reactions were run in triplicate. As negative controls, PCR was performed on water and on RNA without reverse transcription. The cycle number at which the reaction crossed an arbitrarily placed threshold (C␶) was determined for each gene. The relative amount of mRNA levels was determined by the formula 2⫺⌬C␶. Relative gene expression was normalized to ARP expression. Primer sequences and annealing temperatures used for real-time PCR are shown in supplementary online Table 2.

In Situ Hybridization Whole-mount in situ hybridization was performed as described before [25]. Probe template primers were designed with the T3-promoter sequence at the 5⬘ end of the forward primer and

T7-promoter at the 5⬘ end of the reverse primer. Digoxigeninlabeled probes were generated using the purified PCR product as a template for probe transcription with either T3 RNA polymerase or T7 RNA polymerase (Promega, Madison, WI, http:// www.promega.com). Human and mouse probe template primers are shown in supplementary online Table 3. The T3 promoter sequence was 5⬘-ATACAATTAACCCTCACTAAAGGG-3⬘. The T7 promoter sequence was 5⬘-ATAGGTAATACGACTCACTATAGGGC-3⬘.

RESULTS Temporal Gene Expression During Differentiation of hESCs Whole-genome microarray analysis was carried out on total RNA from UH (HES-2 cells), hESCs isolated after coculture with END-2 cells for various times (HES-2-END-2 cocultures: 1–12 days), and whole HFH (16 weeks of gestation). From all samples, equal amounts of RNA were pooled to create a CRP. Each separate sample was then hybridized against the CRP on Agilent whole-genome microarray chips (Fig. 1A). The use of a CRP allows comparison of relative gene expression levels of each sample with all other samples. Dye-swap experiments confirmed the reproducibility of the microarray data (data not shown). Since we previously have shown that hESC-CMs resemble human fetal cardiomyocytes [21], RNA from HFH was used as a reference source for comparison to differentiating hESCs. Additional microarray experiments were performed by hybridizing UH directly against hESCs cocultured for 12 days with END-2 (12 days). Again, dye-swap control experiments were performed (Tables 1, 2). Differentially expressed genes were identified using a statistical cutoff value of p ⱕ .01 (Rosetta Luminator). Figure 1B shows the number of significantly up- and downregulated genes of all samples during hESC differentiation toward cardiomyocytes and HFH. A steady increase in differentially expressed genes is seen as differentiation proceeds. The relatively high number of downregulated genes in the UH and 1-day samples compared with the CRP is likely due to differentiation being initiated but not yet to distinct lineages. The results of all microarray experiments are provided in supplementary online Data 1A–1J.

hESCs Preferentially Differentiate Toward Mesoderm and Endoderm Previously, we demonstrated that after 12 days of hESC-END-2 coculture, various cell types formed, in addition to cardiomyocytes [21, 24]. The majority of these noncardiomyocytes were positive for Troma-1/cytokeratin 8 [24], expressed in endoderm and trophectoderm. The formation of cardiomyocytes, which are derived from mesoderm, was preceded by transient upregulation of Brachyury T, a marker of nascent mesoderm (Fig. 1C). This suggested that hESCs grown on END-2 cells differentiated preferentially toward mesoderm and endoderm derivatives. This was confirmed by calculating the percentage of tissue-enriched genes that were ⱖ3-fold upregulated at 12 days compared directly to UH. Genes were indicated as tissue-enriched according Gene Ontology Tree Machine classifications, representing genes with a p value ⱕ10⫺10. Since most genes are significantly expressed in multiple tissues and therefore classified in different tissues, the sum of the percentages of all tissue-enriched genes


1959

Figure 1. Whole-genome temporal gene expression profiling of hESCs differentiating toward cardiomyocytes. (A): Experimental approach in which RNA of UH, differentiating human embryonic stem cells (hESCs) from hESC-END-2 coculture (1–12 days), and HFH were first isolated, followed by creating a CRP. Each sample was hybridized with the CRP on a 44K Agilent Microarray, and a direct comparison of UH and 12-day differentiated hESCs was done. The first appearance of beating areas (average starting point) is indicated by a red circle on the time line. (B): Number of significantly up- and downregulated genes of UH and all timepoints during hESC differentiation toward cardiomyocytes (p ⬍ .01). The number of upregulated genes is displayed in red, and the number of downregulated genes is displayed in blue. (C): Quantitative real-time PCR analysis of the mesoderm marker brachyury T. (D): Percentage of tissue-enriched genes ⱖ3-fold-upregulated at 12 days compared directly to UH. Genes are indicated as tissue-enriched according Gene Ontology Tree Machine classifications, representing genes with a p value ⱕ10⫺10. Abbreviations: CM, cardiomyocyte; CRP, common reference pool; d, day(s); END, endoderm; HESC, human embryonic stem cells; HFH, human fetal heart; UH, undifferentiated human embryonic stem cells.

Table 1. Human embryonic stem cell (hESC)-enriched genes, including seven genes of unknown function that are rapidly downregulated during hESC-END2 differentiation Fold change GenBank accession no. ENST00000297751 NM_018308 NM_002701 AF268617 NM_019079 NM_024865 NM_003212 NM_018228 NM_015973 BC013923 NM_175850 NM_024504 NM_002164 NM_020997 AK097672

Gene name

Gene description

12d vs. UH

Dye-swap

Unknown FLJ11042 POU5F1 POU5FLC12 FLJ10884 NANOG TDGF1 C14ORF115 GAL SOX2 DNMT3B PRDM14 INDO LEFTB LOC284344

Unknown Acyl-coenzyme A oxidase-like POU domain, class 5, transcription factor 1 POU 5 domain protein Hypothetical protein FLJ10884 Nanog homeobox Teratocarcinoma-derived growth factor 1 Chromosome 14 open reading frame 115 Galanin Sex-determining region Y-box 2 DNA cytosine-5-methyltransferase 3␤ PR domain containing 14 Indoleamine-pyrrole-2,3-dioxygenase Left-right determination factor 1 Hypothetical protein LOC284344

⫺15.49 ⫺17.74 ⫺13.96 ⫺38.40 ⫺19.85 ⫺39.70 ⫺30.30 ⫺22.34 ⫺12.33 ⫺7.75 ⫺17.44 ⫺6.37 ⫺17.59 ⫺18.75 ⫺5.91

18.43 26.25 24.81 57.59 20.86 65.08 42.25 18.63 19.94 25.99 16.11 11.64 23.18 24.92 3.21

Fold change is shown for the direct comparison between UH and 12d and their dye-swap controls. Abbreviations: 12d, 12 days of differentiation; UH, undifferentiated human embryonic stem cells.

exceeds 100%. As expected, the highest percentage (62%) was observed for heart-enriched genes (Fig. 1D). In addition, liver (58%) (endoderm-derived) and kidney-enriched (47%) (mesoderm-derived) genes were also highly represented, whereas www.StemCells.com

brain (12%) and skin-enriched (11%) (both ectoderm-derived) genes showed the lowest percentage. Thus hESC-END-2 cocultures differentiate predominantly toward mesoderm and endoderm, with little ectoderm differentiation taking place.


1960

Table 2. Cardiomyocyte-enriched genes that are upregulated at 12d and human fetal heart, including two genes of unknown function (ADPRHL1 and SYNPO2L) Fold change GenBank accession no.

Gene name

NM_006308 BC009755 NM_014424

HSPB3 ADPRHL1 HSPB7

NM_006789

APOBEC2

NM_003280 NM_024875 NM_003476

TNNC1 SYNPO2L CSRP3

NM_001103 NM_016599 NM_002667 NM_021223 NM_013292 NM_014332 X52005 X79859

ACTN2 MYOZ2 PLN MYL7 HUMMLC2B SMPX MYL4 TNNT2

Gene description

12d vs. UH

Dye-swap

Heat shock 27-kDa protein 3 ADP-ribosylhydrolase-like 1 Heat shock 27-kDa protein family, member 7 (cardiovascular) Apolipoprotein B mRNA-editing enzyme, catalytic polypeptide-like 2 Troponin C, slow Synaptopodin 2-like Cystein and glycin-rich protein 3 (cardiac LIM protein) Actinin, ␣2 Myozenin 2 Phospholamban Myosin light, polypeptide 7, regulatory Myosin light chain 2 Small muscle protein, X-linked Myosin light, polypeptide 4, alkali; atrial, embryonic Troponin T2, cardiac

11.89 15.22 9.80

⫺14.00 ⫺13.41 ⫺9.14

25.79

⫺22.02

10.63 12.04 16.88

⫺12.01 ⫺12.45 ⫺12.79

19.24 1.52 30.37 38.95 30.52 21.61 3.17 11.95

⫺17.73 ⫺1.27 ⫺31.85 ⫺30.72 ⫺35.12 ⫺31.64 ⫺25.04 ⫺12.71

Fold change is shown for the direct comparison between UH and 12d and their dye-swap controls. Abbreviations: 12d, 12 days of differentiation; UH, undifferentiated human embryonic stem cells.

hESC-Enriched Genes It is to be expected that genes important for maintenance or self-renewal of hESCs would be rapidly downregulated upon differentiation. By cluster analysis, we identified a group of 15 genes (cluster 1a) that displayed rapid downregulation (Fig. 2A, 2B) and included well-established hESC-markers such as OCT4 and NANOG. Other genes in this cluster were TDGF1, GAL, LEFTB, DNMT3B, and SOX2, also described as being enriched in UH [15], indicating that the approach used could identify genes that may be involved in hESC self-renewal. Seven genes in this group were novel with unknown function but with expression trend-lines during differentiation similar to those of OCT4 and NANOG (Fig. 2B). A less restrictive cluster (cluster 1b) with similar downregulating gene patterns identified a group of 145 gene entries with potential roles in self-renewal of hESCs (supplementary online data 2). To verify the microarray results, expression levels of a subset of known and unknown genes from both clusters were examined by RT-PCR in two independent UH and 12-day differentiated samples (Fig. 2C). For four unknown genes (ENST00000297751, THC1818772, FLJ10884, and C14ORF115), we carried out whole-mount ISH on UH colonies, cultured on MEFs, and compared expression with that of OCT4 (Fig. 2D). High expression of the selected genes was observed in the parts of the colonies with characteristic undifferentiated morphology, whereas the differentiated parts and the MEFs remained negative. This confirmed association of these genes with the pluripotent state.

Activated and Repressed Signaling Pathways During hESC Differentiation The earliest appearance of beating areas in hESC-END-2 cocultures was consistently at 6 –7 days. Before the appearance of beating areas, cells are committed to form cardiomyocytes, and

regulatory as well as structural genes are activated. Accordingly, expression of the earliest mesoderm marker Brachyury T is transiently upregulated in 3-day hESC-END-2 cocultures, 3 or 4 days before the onset of beating, as described above. Clustering analysis revealed a group of 25 genes (Table 3, cluster 2a), transiently co-upregulated with Brachyury T on 3 days of hESC-END-2 coculture (Fig. 3A, 3B). Interestingly, in addition to eight unknown genes, the clustered group contained genes shown previously to be important for early embryonic development and for mesoderm formation in particular. The Notch signaling pathway, with ligands ␦-like 1 (DLL) and DLL3, upregulated at 3 days, has pleiotropic functions during development, including a role in differentiation of paraxial mesoderm, which later gives rise to skeletal muscle [26, 27]. The WNT family plays an important role in early embryogenesis. Dickkopf-1 (DKK1), an inhibitor of canonical (␤-catenindependent) WNT signaling has been demonstrated to induce cardiomyocyte differentiation [28, 29]. For Wnt3A, both inducing [29] and inhibiting effects [30] on cardiomyocyte differentiation have been described. Other genes in this cluster were the transcription factors FOXC1 and MESP1 and the transcriptional repressor SNAI1, which are all expressed in mesoderm and important for cardiac development. The majority of these genes also play roles in epithelial-mesenchymal-transformation (EMT), important for many processes during development, including that of the heart. The transient expression of these genes during differentiation was confirmed by quantitative real-time PCR (Fig. 3C). Interestingly, for DKK1, WNT3A, FOXC1, and SNAI1 a biphasic expression pattern was observed with peaks at 3 and 12 days, suggesting a role in both early and late phases of differentiation. A larger cluster with a similar pattern (cluster 2b) is available as supplementary online Data 2. Besides activation of genes involved in mesoderm formation, genes associated with endoderm formation, such as FOXA2 and SOX17, displayed a similar transient upregulation with a peak at 3 days (data not shown).


1961

Figure 2. Human embryonic stem cell (hESC)-enriched genes. (A): Cluster of 15 unique genes (cluster 1a) that are rapidly downregulated upon differentiation, of which seven are unknown. Red indicates upregulation versus common reference pool (CRP); blue indicates downregulation versus CRP. *, since multiple probes are present on the microarray chip, some genes are represented multiple times in one cluster. (B): Genes of cluster 1a shown in a time-course graph with a reference line at y ⫽ 0 where the expression ratio is equal to 1. (C): Reverse transcription-polymerase chain reaction confirmation of various hESC-enriched genes, including three unknown genes. (D): In situ hybridizations on hESC colonies for four unknown hESC-enriched genes. OCT4 was used as a positive control, and OCT4 sense probe was used as a negative control. To visualize nonstained hESC colonies, different transmission light settings were used. Arrowheads indicate differentiated areas within hESC colonies. Scale bar ⫽ 1 mm for OCT4, OCT4 sense, ENST00000297751, FLJ10884, and C14ORF115. Scale bar ⫽ 500 ␮m for THC2173136. Abbreviations: d, day(s); RT, reverse transcription; UH, undifferentiated human embryonic stem cells.

We next identified a cluster of genes (cluster 3) with expression transiently increased at 6 days and enhanced expression in HFH. This cluster consisted of 204 gene entries, including the transcription factors LBX1, IRX3, IRX4, FOXC2, HES7, and SRF, all of which have been implicated in cardiac development (supplementary online Data 2). In addition, MSX1 and ISL1, homeodomain transcription factors, also increased from 3–12 days and from 6 –12 days, respectively (supplementary online Fig. 1). MSX1 plays a role in EMT and cardiac development, whereas ISL1 has recently been described as a marker for a cardiac progenitor cell (CPC) population [31]. The WNT, fibroblast growth factor (FGF), transforming growth factor ␤ (TGF␤), and bone morphogenetic protein (BMP) signaling pathways have been shown to play important www.StemCells.com

roles during embryonic development and early cardiogenesis. The early transient increase of members of the WNT family, DKK1, DKK3, and WNT3a have been discussed above, but other WNT family members, such as FRZ10, LEF1, WNT5A, WNT5B, and WNT11, showed a steady increase in expression during differentiation, whereas secreted FRZ1, FRZ2, Cerberus, WNT8B, FRZ5, and SRY17 were downregulated during differentiation (Fig. 3D). For the TGF␤ and FGF families, FGF9, FGF receptor 2, TGF␤1, TGF␤2, TGF␤3, SMAD3, TGF␤ receptor III, and Activin A receptor type I and II were upregulated, whereas nodal, FGF receptor 1, and FGF2 (basic FGF) were downregulated. Interestingly, BMP10, enriched in heart tissue and essential for cardiogenesis in mouse [32], was highly expressed in UH, rapidly downregulated after initiation of


1962

Table 3. Genes transiently upregulated at 3d of differentiation include established mesoderm markers GenBank accession no. Gene name NM_175903 NM_032944 NM_005097 NM_033131 NM_013435 NM_203486 NM_018670 NM_014420 NM_002546 NM_000552 NM_000104 NM_021990 NM_006605 NM_147164 NM_014379 NM_173515 NM_005161 NM_005985 NM_018723 NM_012242 NM_001453 AK091351 U85992 AB023144 NM_005332

Fold change Gene description

LOC284033 STK31 LGI1 WNT3A

Hypothetical protein LOC284033 Serine/threonine kinase 31, transcript variant 2 Leucine-rich, glioma-inactivated 1 Wingless-type MMTV integration site family, member 3A RAX Retina and anterior neural-fold homeobox DLL3 Delta-like 3 (Drosophila), transcript variant 2 MESP1 Mesoderm posterior 1 DKK4 Dickkopf homolog 4 TNFRSF11B Tumor necrosis factor receptor superfamily, member 11b (osteoprotegerin) VWF Von Willebrand factor CYP1B1 Cytochrome P450, family 1, subfamily B, polypeptide 1 GABRE Gamma-aminobutyric acid (GABA) A receptor, epsilon, transcript variant 4 RFPL2 Ret finger protein-like 2 CNTFR Ciliary neurotrophic factor receptor transcript variant 1 KCNV1 Potassium channel, subfamily V, member 1 MAGI1 Membrane-associated guanylate kinase interacting protein-like 1 AGTRL1 Angiotensin II receptor-like 1 SNAI1 Snail homolog 1 A2BP1 Ataxin 2-binding protein 1 transcript variant 4 DKK1 Dickkopf homolog 1 FOXC1 Forkhead box C1 SAMD3 Sterile ␣ motif domain containing 3 BMPER Human clone IMAGE:35527 unknown protein mRNA, partial cds SEZ6L Seizure-related 6 homolog (mouse)-like HBZ Hemoglobin, ␨

UH

1d

3d

6d

9d

12d

0.846 0.858 0.403 0.535

0.803 0.578 0.428 0.561

4.501 3.913 3.077 3.367

0.845 0.756 0.580 0.570

0.717 0.566 0.744 0.839

0.481 0.127 0.301 0.460

0.853 0.329 0.230 0.709 0.723

0.718 0.419 0.520 0.380 0.810

2.217 3.600 3.566 2.892 1.942

1.117 0.324 0.300 0.173 0.723

0.815 0.377 0.120 0.219 0.594

0.213 0.246 0.118 0.091 0.378

0.809 0.888 2.427 0.786 0.916 1.144 0.796 0.668 2.634 0.585 0.714 0.554 0.672 0.730 2.407 0.679 0.693 0.819 0.889 0.605 2.864 0.621 0.593 0.480 0.573 0.734 2.655 1.093 0.838 0.936 0.498 0.401 2.429 0.795 0.944 0.887 0.611 0.784 3.036 0.900 1.124 1.239 0.414 0.512 0.354 0.665 0.334 0.780 0.270

0.554 0.549 0.541 0.492 0.483 0.848 0.706

2.327 2.060 1.916 2.706 2.575 3.356 2.160

1.781 1.049 0.546 0.578 0.473 0.662 1.102

0.538 0.743 0.791 0.749 0.647 0.592 0.681

0.403 0.784 0.683 0.711 1.048 0.882 0.563

0.368 0.978 3.899 0.386 0.584 0.280 0.913 0.499 2.391 0.915 0.954 1.367

Transiently upregulated genes. Abbreviations: 1d–12d, 1–12 days of differentiation; UH, undifferentiated human embryonic stem cells.

differentiation, and increased again at 9 and 12 days. BMP4, also essential for cardiogenesis [33], and BMP5, expressed in cardiac development [34], were increased during differentiation (supplementary online Data 3; supplementary online Fig. 2).

Cardiac-Enriched Genes Since the number of beating areas are increasing from day 9 to day 12 of differentiation, genes responsible for cardiomyocyte differentiation and function are expected to be upregulated during this time period. As reference source, HFH was used to increase the possibility that upregulated genes may be associated with human cardiac development. A cluster of 18 genes (cluster 4a), upregulated at 9 or 12 days of differentiation and in HFH was identified (Fig. 4A, 4B). This group consisted of known cardiac genes, such as troponin T 2 (TNNT2), myosin light chain 4 (MYL4), ␣-actinin-2 (ACTN2), and myosin light chain 2B (HUMMLC2B), but also four novel genes and several genes that have been annotated but not previously associated with cardiac development. A larger cluster of 145 gene entries with similar expression pattern (cluster 4b) is represented in supplementary online Data 2. Verification of a subset of the cardiac-enriched genes was performed by RT-PCR (Fig. 4C). Many transcription factors known to play a role in cardiac

development were upregulated during differentiation. Real-time RT-PCR for cardiac transcription factors MEF2C, TBX2, and TBX5 demonstrated increased expression at 9 and 12 days of differentiation. By comparison, ␣-actinin, a sarcomeric protein, was only increased after 12 days of differentiation (Fig. 4D). These data demonstrate distinct molecular regulation during each step of differentiation (Fig. 4E). Furthermore, ISH for a set of the cardiac-enriched genes, including phospholamban and two novel genes annotated as SRD5A2L2 and SYNPO2L (from clusters 4a and 4b), on microdissected beating areas from 12day hESC-END-2 cocultures showed strong expression (Fig. 5A–5C). Expression of these genes was not detected in the nonbeating areas derived from the same hESC-END-2 cocultures. Mouse orthologs were found for both SRD5A2L2 and SYNPO2L and whole-mount ISH on mouse embryos at different development stages showed cardiac-restricted/cardiac-enriched expression in early stages of cardiac development for both genes. At E8.5 and E9.5, SRD5A2L2 was specifically expressed in the sinus venosus, giving rise to the inflow tract, and at lower levels in the ventricle (Fig. 5D–5F). SYNPO2L was detected in the heart from E8.5 and in myotomes, which give rise to skeletal muscle, at E10.5 (Fig. 5G–5I). In addition, whole-mount ISH on intact HFH (Fig. 5J) or hearts cut in half (Fig. 5K–5M) showed


1963

Figure 3. Transiently upregulated genes during human embryonic stem cell (hESC) differentiation. (A): Cluster of 25 unique genes (cluster 2a) that are transiently upregulated at day 3 of differentiation. Among these genes, we found mesoderm markers and genes that are important in early cardiac development. *, since multiple probes are present on the microarray chip, some genes are represented multiple times in one cluster. (B): Genes of cluster 2a are shown in a time-course graph with a reference line at y ⫽ 0 where the expression ratio is equal to 1. (C): Quantitative real-time PCR also demonstrates a transient upregulation of genes implicated in mesoderm and heart formation. (D): Schematic representation of the regulation of genes of the WNT family during hESC differentiation. Transiently upregulated expression is indicated by blue-red-blue coloring, downregulation by blue, and upregulation by red. (BRA-T) is downstream of LEF1 and is important for mesoderm formation, whereas SOX17 may interact with LEF1 for the formation of endoderm. Abbreviations: BRA-T, Brachyury T; ␤-CAT, ␤-catenin; d, day(s); UH, undifferentiated human embryonic stem cells.

that the unknown genes THC2339346, SRD5A2L2, and SYNPO2L were all expressed as the positive control phospholamban in the fetal heart. As in mouse embryos, SRD5A2L2 was detected predominantly in the inflow tract of HFHs.

DISCUSSION In the present study, we described the temporal gene expression of hESCs during differentiation toward the cardiac lineage. Besides information on genes that may play roles in self-maintenance or self-renewal of hESCs, genes important during the different stages of cardiac development were also identified. It www.StemCells.com

is expected that genes with important roles in self-renewal of ESC would be rapidly downregulated upon differentiation. For this reason, several studies have compared the transcriptome of UH with differentiated hESCs [15, 17, 18]. In most studies, differentiation of hESCs occurs through the formation of EBs. However, information on changes in gene expression of hESCs during the differentiation process itself has, so far, been very limited. In one study [18], microarray analysis was performed on 2-, 10-, 20-, and 30-day differentiated EBs from hESCs. However, 10-day EBs had essentially unchanged levels of OCT4 expression, suggesting that differentiation by formation

1964


Figure 4. Cardiomyocyte-enriched genes at 12 days of differentiating hESCs. (A): Cluster of 15 unique genes (cluster 4a) that are upregulated at 12 days as well as in HFH of which three genes of unknown function. *, since multiple probes are present on the microarray chip, some genes are represented multiple times in one cluster. (B): Genes of cluster 4a in a time-course graph, including HFH for comparisons with 12 days with a reference line at y ⫽ 0 where the expression ratio is equal to 1. (C): Reverse transcription-polymerase chain reaction confirmation of a subset of genes (from clusters 4a and 4b) of known and unknown function that are expressed at 12 days and HFH but not or at low levels in UH. ARP was used as an internal control. (D): Quantitative real-time polymerase chain reaction of cardiac transcription factors MEF2C, TBX2, and TBX5 and cardiac marker ACTN2. (E): Schematic representation of hESCs differentiating toward the cardiac lineage compared with embryonic development. UH resemble the epiblast stage during embryonic development. Shortly after the start of hESC-END-2 coculture, (3-day) gene expression profiles indicate the onset of gastrulation with a preference for endoderm and mesoderm formation, followed by the probable formation of cardiac progenitor cells (CPCs) between 3 and 6 days, and optimal differentiation to CMs at 12 days. At the bottom, the contributions of genes from the different clusters during the different phases of differentiation are indicated. The shape of the triangles represents the relative expression of the genes within that cluster during differentiation, with two examples of genes in each triangle. The first appearance of beating areas (average starting point) is indicated by a red circle on the time line. Abbreviations: ARP, acidic ribosomal protein; CL, cluster; CM, cardiomyocyte; CPC, cardiac progenitor cell; d, day(s); hESCs, human embryonic stem cells; HFH, human fetal heart; RT, reverse transcription; UH, undifferentiated human embryonic stem cells.

of EBs was relatively inefficient and slow. We demonstrated here that temporal transcriptional profiling of hESCs can be a rich source of information on the molecular mechanisms underlying human development, provided differentiation is directed and efficient, as in this study for the cardiac lineage. By cluster analysis of microarray data, we identified genes that were rapidly downregulated upon differentiation, which included OCT4 and NANOG, as well as multiple unknown genes, and confirmed their expression independently in UH colonies. Whether these genes play roles in the self-renewal of stem cells or are simply stem cell markers has to be determined in functional studies.

Previously, we have shown that hESCs in coculture with END-2 cells differentiated predominantly to endoderm and mesoderm derivatives [21, 24]. Here, this was confirmed by microarray analysis with transient expression of the early mesodermal marker brachyury T and the endodermal markers FOX2A and SOX17, peaking after 3 days of differentiation (3– 4 days before the appearance of the first beating areas). Other mesodermal markers or transcription factors, such as SNAI1, MSX1 MESP1, and FOXC1, were transiently upregulated, with peaks at 3 days. The detailed temporal gene profiling of differentiating hESCs presented here may lead to improvements in stepwise directed differentiation toward mesoderm and


1965

Figure 5. In vitro and in vivo expression analysis of novel cardiac specific genes by whole-mount in situ hybridization. (A): Expression of PLN in an excised beating area at 12 days. The inset shows a nonbeating area at 12 days that shows no expression of PLN. (B): Expression of SRD5A2L2 in a 12-day beating area, whereas the nonbeating area (inset) shows no expression. (C): Expression of SYNPO2L in an excised beating area at 12 days. Nonbeating (inset) area showed no expression. (D): Expression of mouse SRD5A2L2 at day 8.5 postfertilization is restricted to the heart, with strongest expression in the ift. Dorsal to the right, anterior to the top. (E): Transverse section of an E9.5 mouse embryo with strong expression of SRD5A2L2 in both horns of the sinus venosus, lower expression in other parts of the heart. Dorsal to the top. (F): Magnification of right horn of sinus venosus showing strong expression of SRD5A2L2 at E9.5. Dorsal to the top. (G): Expression of mouse SYNPO2L at E9.5 is restricted to the heart and somites. Dorsal to the right, anterior to the top. (H): Transverse section of E8.5 mouse embryo showing SYNPO2L expression in the heart tube and developing somites. (I): Transverse section of an E10.5 mouse embryo showing SYNPO2L expression in the myotome. (J): SYNPO2L expression in human fetal heart at 11 weeks of gestation. Expression of SYNPO2L in both ventricles and stronger in both atria but absent in aorta and pulmonary vein. (K): Expression of SRD5A2L2 predominantly in the vena cava and pulmonary vein of a 14-week human fetal heart. (L): Ubiquitous expression of THC2339346 in a 17-week human fetal heart. (M): Phospholamban expression in a 20-week human fetal heart. Scale bars ⫽ 500 ␮m (A–D, G), 300 ␮m (E, H, I), 150 ␮m (F), and 2 mm (J–M). Abbreviations: ao, aorta; avc, atrioventricular canal; E, embryonic day; ht, heart tube; ift, inflow tract; la, left atrium; ls, left horn sinus venosus; lv, left ventricle; lvw, left ventricular wall; m, myotome; nf, neural fold; nt, neural tube; oft, outflow tract; PLN, phosholamban; pv, pulmonary vein; ra, right atrium; rs, right horn sinus venosus; rv, right ventricle; rvw, right ventricular wall; s, somite; sp, septum; SRD5A2L2, steroid reductase 5-␣2-like 2; st, septum transversum; svc, superior vena cava; SYNPO2L, synaptopodin 2-like; W, weeks.

www.StemCells.com

1966

endoderm-derived cell-types or tissues, if specific signaling pathways are further activated or enhanced at the times indicated. Many studies have demonstrated the importance of endoderm for the induction of cardiac differentiation [35]. The presence of endoderm as well as mesoderm markers in differentiating hESCs suggests that differentiation to cardiomyocytes in hESC-END-2 cultures may be either direct, by the END-2 cells, and/or indirect, by the formation of endoderm derived from hESCs, inducing adjacent mesodermal cells to become cardiomyocytes. Proteins from the WNT, TGF␤, BMP, and FGF signaling pathways have been implicated in both self-renewal of stem cells as well as cardiac differentiation. Basic FGF, important for self-renewal [36], was rapidly downregulated upon differentiation. DKK1, WNT11, and WNT5A and downstream transcription factor LEF1 increased during differentiation, and all have been shown to play a role in cardiac development. For other factors, such as WNT3A, both cardiac-inducing and cardiac-inhibiting effects have been described [29, 30]. In addition, WNT3A has been suggested to play a role in the self-renewal of stem cells [37], although some controversy surrounds this report. Overall, however, this suggests that the net effect on differentiating cells depends on the state of the cell when targeted; the concentration, duration, or exposure; and the presence of other factors. This was recently shown by Yuasa et al. [38], who demonstrated that the BMP inhibitor Noggin only induced cardiomyocyte differentiation of mESCs when applied in a very narrow time window during ESC culture. Microarray studies have been performed on mouse P19 embryonal carcinoma cells differentiating to cardiomyocytes and during early mouse cardiogenesis, using an Nkx2.5 enhancer to select cardiac progenitor cells [39]. The human heart, however, is not accessible at the first crucial stages of development, 4 weeks of gestation being the earliest we have obtained when the heart already has four defined chambers. Here, we

REFERENCES 1

Thomson JA, Itskovitz-Eldor J, Shapiro SS et al. Embryonic stem cell lines derived from human blastocysts. Science 1998;282:1145–1147.

2

Reubinoff BE, Pera MF, Fong CY et al. Embryonic stem cell lines from human blastocysts: Somatic differentiation in vitro. Nat Biotechnol 2000; 18:399 – 404.

hESC Differentiation Toward Cardiomyocytes have described for the first time the transcriptome of hESCs differentiated toward the cardiac lineage. Using HFH as a reference source, we were able to identify genes that were both upregulated during differentiation of hESC-CMs and in HFH. In addition to the presence of known cardiac sarcomeric genes, several unknown genes and genes not previously associated with cardiomyocytes were identified (clusters 4a and 4b). Two of these unknown genes, SRD5A2L2 and SYNPO2L, displayed cardiac expression during mouse embryonic development and in HFHs. In addition, THC2339346, THC1564329, and THC1452070, for which no mouse ortholog could be found, were expressed in HFH. These findings indicate that in vitro differentiation of hESCs as used in this study is a very powerful model for human cardiac development. Using hESCs as a model for human cardiac development will likely result in the discovery of genes not previously associated with cardiomyocyte differentiation or function and may lead to a better understanding of abnormal cardiac development. Furthermore, improved cardiomyocyte differentiation will lead to a higher production and purer populations of cardiomyocytes, facilitating entry into regenerative medicine for heart patients.

ACKNOWLEDGMENTS We thank Linda W. van Laake and Rutger J. Hassink for the isolation of the human fetal hearts and Jeroen Korving for histology. We thank ServiceXS (Leiden, The Netherlands) for help and use of Rosetta Luminator software for data analysis. Furthermore, we thank Hans C. Clevers and Chris N. Denning for critical reading of the manuscript. This study was supported by ES Cell International (J.K.) and the Interuniversity Cardiology Institute of the Netherlands (D.W.-v.O., R.P., and C.M.).

DISCLOSURES J.K. has performed contract work for ES Cell International within the last 2 years.

10 Carpenter MK, Rosler E, Rao MS. Characterization and differentiation of human embryonic stem cells. Cloning STEM CELLS 2003;5:79 – 88. 11 Zeng X, Miura T, Luo Y et al. Properties of pluripotent human embryonic stem cells BG01 and BG02. STEM CELLS 2004;22:292–312. 12 Abeyta MJ, Clark AT, Rodriguez RT et al. Unique gene expression signatures of independently-derived human embryonic stem cell lines. Hum Mol Genet 2004;13:601– 608.

3

Cowan CA, Klimanskaya I, McMahon J et al. Derivation of embryonic stem-cell lines from human blastocysts. N Engl J Med 2004;350:1353–1356.

4

Passier R, Mummery C. Origin and use of embryonic and adult stem cells in differentiation and tissue repair. Cardiovasc Res 2003;58:324 –335.

13 Bhattacharya B, Miura T, Brandenberger R et al. Gene expression in human embryonic stem cell lines: Unique molecular signature. Blood 2004;103:2956 –2964.

5

Ramalho-Santos M, Yoon S, Matsuzaki Y et al. “Stemness”: Transcriptional profiling of embryonic and adult stem cells. Science 2002;298: 597– 600.

14 Sato N, Sanjuan IM, Heke M et al. Molecular signature of human embryonic stem cells and its comparison with the mouse. Dev Biol 2003;260:404 – 413.

6

Ivanova NB, Dimos JT, Schaniel C et al. A stem cell molecular signature. Science 2002;298:601– 604.

7

Ginis I, Luo Y, Miura T et al. Differences between human and mouse embryonic stem cells. Dev Biol 2004;269:360 –380.

15 Brandenberger R, Wei H, Zhang S et al. Transcriptome characterization elucidates signaling networks that control human ES cell growth and differentiation. Nat Biotechnol 2004;22:707–716.

8

Wei CL, Miura T, Robson P et al. Transcriptome profiling of human and murine ESCs identifies divergent paths required to maintain the stem cell state. STEM CELLS 2005;23:166 –185.

9

Sperger JM, Chen X, Draper JS et al. Gene expression patterns in human embryonic stem cells and human pluripotent germ cell tumors. Proc Natl Acad Sci U S A 2003;100:13350 –13355.

16 Calhoun JD, Rao RR, Warrenfeltz S et al. Transcriptional profiling of initial differentiation events in human embryonic stem cells. Biochem Biophys Res Commun 2004;323:453– 464. 17 Miura T, Luo Y, Khrebtukova I et al. Monitoring early differentiation events in human embryonic stem cells by massively parallel signature sequencing and expressed sequence tag scan. Stem Cells Dev 2004;13: 694 –715.


1967

18 Dvash T, Mayshar Y, Darr H et al. Temporal gene expression during differentiation of human embryonic stem cells and embryoid bodies. Hum Reprod 2004;19:2875–2883.

29 Nakamura T, Sano M, Songyang Z et al. A Wnt- and beta-catenindependent pathway for mammalian cardiac myogenesis. Proc Natl Acad Sci U S A 2003;100:5834 –5839.

19 Andrews PW, Benvenisty N, McKay R et al. The International Stem Cell Initiative: Toward benchmarks for human embryonic stem cell research. Nat Biotechnol 2005;23:795–797.

30 Schneider VA, Mercola M. Wnt antagonism initiates cardiogenesis in Xenopus laevis. Genes Dev 2001;15:304 –315.

20 Mummery C, Ward D, van den Brink CE et al. Cardiomyocyte differentiation of mouse and human embryonic stem cells. J Anat 2002;200: 233–242.

31 Laugwitz KL, Moretti A, Lam J et al. Postnatal isl1⫹ cardioblasts enter fully differentiated cardiomyocyte lineages. Nature 2005;433:647– 653.

21 Mummery C, Ward-van Oostwaard D, Doevendans P et al. Differentiation of human embryonic stem cells to cardiomyocytes: Role of coculture with visceral endoderm-like cells. Circulation 2003;107:2733–2740.

32 Pashmforoush M, Lu JT, Chen H et al. Nkx2–5 pathways and congenital heart disease; loss of ventricular myocyte lineage specification leads to progressive cardiomyopathy and complete heart block. Cell 2004;117: 373–386.

22 Xu C, Police S, Rao N et al. Characterization and enrichment of cardiomyocytes derived from human embryonic stem cells. Circ Res 2002; 91:501–508.

33 Schultheiss TM, Burch JB, Lassar AB. A role for bone morphogenetic proteins in the induction of cardiac myogenesis. Genes Dev 1997;11: 451– 462.

23 He JQ, Ma Y, Lee Y et al. Human embryonic stem cells develop into multiple types of cardiac myocytes: Action potential characterization. Circ Res 2003;93:32–39.

34 Somi S, Buffing AA, Moorman AF et al. Dynamic patterns of expression of BMP isoforms 2, 4, 5, 6, and 7 during chicken heart development. Anat Rec A Discov Mol Cell Evol Biol 2004;279:636 – 651.

24 Passier R, Oostwaard DW, Snapper J et al. Increased cardiomyocyte differentiation from human embryonic stem cells in serum-free cultures. STEM CELLS 2005;23:772–780.

35 Lough J, Sugi Y. Endoderm and heart development. Dev Dyn 2000;217: 327–342.

25 Nieto MA, Bennett MF, Sargent MG et al. Cloning and developmental expression of Sna, a murine homologue of the Drosophila snail gene. Development 1992;116:227–237. 26 Bettenhausen B, Hrabe de Angelis M, Simon D et al. Transient and restricted expression during mouse embryogenesis of Dll1, a murine gene closely related to Drosophila Delta. Development 1995;121:2407–2418. 27 Jen WC, Wettstein D, Turner D et al. The Notch ligand, X-Delta-2, mediates segmentation of the paraxial mesoderm in Xenopus embryos. Development 1997;124:1169 –1178. 28 Marvin MJ, Di Rocco G, Gardiner A et al. Inhibition of Wnt activity induces heart formation from posterior mesoderm. Genes Dev 2001;15: 316–327.

36 Xu RH, Peck RM, Li DS et al. Basic FGF and suppression of BMP signaling sustain undifferentiated proliferation of human ES cells. Nat Methods 2005;2:185–190. 37 Sato N, Meijer L, Skaltsounis L et al. Maintenance of pluripotency in human and mouse embryonic stem cells through activation of Wnt signaling by a pharmacological GSK-3-specific inhibitor. Nat Med 2004; 10:55– 63. 38 Yuasa S, Itabashi Y, Koshimizu U et al. Transient inhibition of BMP signaling by Noggin induces cardiomyocyte differentiation of mouse embryonic stem cells. Nat Biotechnol 2005;23:607– 611. 39 Masino AM, Gallardo TD, Wilcox CA et al. Transcriptional regulation of cardiac progenitor cell populations. Circ Res 2004;95:389 –397.

See www.StemCells.com for supplemental material available online.

www.StemCells.com