In Vitro Chamber Specification during Embryonic Stem Cell ~ Cardiogenesis EXPRESSION OF THE VENTRICULAR MYOSIN LIGHT CHAIN-2 GENE IS INDEPENDENT HEART TUBE FORMATION*

The molecular cues that control patterning of the heart tube during early cardiogenesis are largely un-known. The present study has explored the embryonic stem (ES) cell differentiation system to determine if this in vitro model could be useful in studying the process of regional specification of cardiac muscle cells at the earliest possible stages. As assessed by polym- erase chain reaction, ribonuclease protection, in situ hybridization, and immunohistochemical analyses, ES cell differentiation into embryoid bodies is characterized by the transcriptional and translational activation of the ventricular regulatory (phosphorylatable) myosin light chain gene, demonstrating that ventricular specification occurs during ES cell cardiogenesis. The finding of a ventricular-specific marker in an in vitro system in the absence of an intact heart tube provides evidence for cardiac regional specification independent of positional cues or physiologic stimuli. The tem- poral expression of the myogenic regulatory factors, myogenin and MyoD, suggests activation of the skele- tal muscle program following cardiac myogenesis in vitro, indicating temporal fidelity to the progression of in vivo myogenesis. These data establish the mouse embryonic stem cell system as a model for cardiac chamber specification and suggest a promising approach in the study of regional specification in geneti- cally engineered cardiac muscle cells.

eventually drive the pulmonary and systemic circulations in the mature heart. The acquisition of the divergent morphological, electrophysiological, and biochemical properties of the specialized cell types that constitute the different cardiac chambers is largely due to the activation of regional specific programs of muscle gene expression during cardiac development. Although basic region helix-loop-helix (bHLH)' myogenic regulators, MyoD and related family members, for skeletal muscle cell lineages have been extensively studied (for review, see Olson (1993)), these genes are not expressed in cardiac muscle, and there are currently no known cardiacspecific bHLH proteins. In fact, increasing evidence suggests that divergent programs may control the different muscle specificities of a single gene that is expressed in both cardiac and skeletal muscle (Lee et al., 1992). Although a small number of transcriptional factors that may participate in regulating cardiac muscle genes have been identified (Pollock and Treisman, 1991;Farrance et al., 1992;Yu et al., 1992;Adolf et al., 1993;Zhu et al., 1993), a clear view of how cardiac specificity is conferred is lacking. Accordingly, the molecular and positional cues that control patterning of the heart tube or regional specification of muscle cells during early embryogenesis are poorly understood.
Our previous studies have established the gene that encodes the ventricular isoform of the cardiac regulatory myosin light chain (MLC-2V) as a valuable model for the elucidation of molecular mechanisms that underlie the regulation of muscle gene expression during cardiac growth and development . Within the normal adult mammalian myocardium, this gene is expressed exclusively in the ventricular chamber, with negligible expression in the atrium. In contrast to other genes, which are co-expressed in all cardiac chambers throughout the early looped heart and acquire regional specificity relatively late during the course of cardiogenesis (post-septation) or postnatal period (Lyons et al., 1990), the MLC-2V gene displays positional specification to the ventricular segment of the primitive linear heart tube, preceding cardiac septation and the development of distinct cardiac chambers (O'Brien et al., 1993).
The lack of continuous cell lines that can mimic the tran-

In Vitro
Model for Ventricular Specification 25245 sition from a pluripotent stem cell to specific cardiac muscle cell lineages (atrial, ventricular, conduction system) has represented a major impediment for coupling molecular and genetic approaches to studying the sequential steps toward regional specification in the embryonic heart. To date, studies of the developmental regulation of the cardiac muscle gene program have predominantly necessitated the use of primary cardiomyocytes derived from embryonic or neonatal myocardium (Thompson et al., 1991;Zhu et al., 1991;Navankasattusas et al., 1992;Molkentin et al., 1993), cells which are already committed to a specific cardiac muscle cell lineage.
In this regard, the recent availability of cell culture models of differentiation using murine embryonic stem (ES) cells, which can activate cardiac myogenesis during in vitro differentiation to embryoid bodies, provides an attractive system in which to address gene regulation during cardiac growth and development (Doetschman et al., 1985;Robbins et al., 1990;Sanchez et al., 1991). Pluripotent ES cells, derived from the inner cell mass of mouse blastocysts, can be maintained for many generations in the undifferentiated state in culture, retaining their embryonic phenotype (Evans and Kaufman, 1981;Martin, 1981). When ES cells are removed from the differentiation inhibitory influence of feeder cells or their equivalent, the cells spontaneously give rise to structures resembling those found in an embryo. The "embryo-like'' structures or embryoid bodies express markers of endodermal, ectodermal, and mesodermal differentiation and have been used as in vitro model systems to study hematopoiesis, vasculogenesis, and angiogenesis (Risau et al., 1988;Lindenbaum and Grosveld, 1990;Schmitt et al., 1991;Wiles and Keller, 1991;Wang et al., 1992;Keller et al., 1993). In addition, embryoid bodies express a number of muscle markers, including myosin heavy chain and tropomyosin (Robbins et al., 1990(Robbins et al., , 1992Wieczorek et al., 1990;Sanchez et al., 1991;Muthuchamy et al., 1993), suggesting that this system can serve as a model to study early aspects of cardiogenesis.
To determine if embryonic stem cells could be useful in studying the process of regional cardiac muscle cell specification in vitro, the current study has utilized the cardiac MLC-2V gene as a genetic marker of the process of ventricular specification. A combination approaches based on polymerase chain reaction (PCR), ribonuclease (RNase) protection, and in situ hybridization documents ventricular specification during ES cell cardiogenesis. The results demonstrate that the activation of myogenic determination/differentiation genes (myogenin and MyoD) lags behind cardiac myogenesis, thereby displaying temporal fidelity with the onset of expression of these muscle markers during normal embryogenesis. Accordingly, these studies suggest that genetic manipulation of ES cells will now allow genetic approaches to study the complex process of cardiac chamber specification. In addition, these results in an in vitro system suggest that the activation of ventricular MLC-2 gene expression is not dependent upon positional cues that arise from the formation of an intact heart tube, as occurs during cardiogenesis in vivo.

EXPERIMENTAL PROCEDURES
Cell Culture-The mouse blastocyst-derived D3 embryonic stem cell line (Doetschman et al., 1985) was generously provided by Dr. T. C. Doetschman (University of Cincinnati). ES cells were propagated in high glucose Dulbecco's modified Eagle medium (Life Technologies, Inc.) supplemented with 15% heat-inactivated fetal calf serum (Sigma), 2 mM L-glutamine and 0.1 mM P-mercaptoetbanol (Sigma). Cells were maintained in the undifferentiated state either by culture on confluent feeder layers of mitomycin C-treated primary mouse embryonic fibroblasts or by the addition of purified recombinant mouse leukemia inhibitory factor (LIF; Esgro, Life Technologies, Inc.) at 1,000 units/ml to the culture media. Under these conditions, the majority of the ES cells (exceeding 95%) displayed an undifferentiated phenotype, as assessed by visual inspection under phasecontrast microscopy. Cells were maintained at 37 "C in a humidified atmosphere of 10% C02 in air. Monolayers were passaged by trypsinization at 70-80% confluence.
In uitro differentiation was initiated by harvesting ES monolayers with trypsin-EDTA and transferring dissociated single cell suspensions to bacterial Petri dishes containing growth media as described above excluding LIF. In the case of ES cells grown in the presence of fibroblasts, differential cell culture substrate attachment was performed to remove feeder cells (Sanchez et al., 1991). To ensure a single cell suspension, the harvested cells were passaged through a small bore plastic pipette; in some cases, additional trituration by passage through a fine bore Pasteur pipette was required. On the 4th day of culture, the serum concentration was raised to 20%. The media in the differentiation cultures was changed every other day, with replacement of half of the media on the in-between days. Under these conditions cells aggregate to form three-dimensional structures or "embryoid bodies," which increase in complexity as differentiation progresses. Several batches of fetal calf serum were prescreened that would allow for a relatively high proportion of beating cell aggregates in the suspension cultures and, at the same time, would also optimally support ES cell growth and minimize their differentiation. In each differentiation culture, 40-70% of embryoid bodies exhibited spontaneous contractile activity, usually first identified after a week in culture.
RNA Isolation and cDNA Synthesis-Total cellular RNA was isolated from feeder-layer fibroblasts, ES cells, or ES cell-derived embryoid bodies by the RNAzol method (Cinna Biotecx). First-strand cDNA synthesis was performed using 5 pg of total RNA, Moloney murine leukemia virus RNase Hreverse transcriptase, and 100 ng of random hexamers, according to the manufacturer's recommendations (Life Technologies, Inc.) in a final reaction volume of 40 pl. Following termination of the first-strand synthesis reaction, the samples were incubated with RNase H at 37 "C for 20 min. The same first-strand preparation was used for analyzing each of the gene products by PCR.
Polymerase Chain Reaction and Analysis of Amplified Products-PCR amplifications were performed with 5% of the first-strand reaction, 2.5 units of Taq polymerase (Life Technologies, Inc.), and 50 pmol of the appropriate primers in a reaction volume of 50 pl. The reactions were carried out in an automated thermal cycler (Ericomp). The amplification sequence consisted of an initial denaturation at 94 "C for 4 min, followed by 5 cycles of denaturation at 94 "C for 45 s, annealing at 55-65 "C for 1 min (depending upon the melting temperature of the oligonucleotide primers), and extension at 72 "C for 2 min, and 30 cycles of 94 "C for 45 s, 45-55 "C for 1 min, and 72 "C for 2 min. A final extension at 72 "C for 10 min was performed. The authenticity of the amplified products was determined by size analysis (15 pl of the reaction) on agarose gels (2%).
Ribonuclease Protection Studies-RNase protection assays were performed as previously described (Zhu et al., 1991). For the assessment of the expression of the mouse ventricular regulatory light chain gene, a 460-bp probe (406-bp protected fragment) was generated by XbaI restriction of a plasmid containing the MLC-2V cDNA (mMLC-2V-510). To assay levels of atrial natriuretic factor (ANF) mRNA, a rat ANF cDNA representing the entire coding region cloned into the PstI site of pGEM-1 was used (Knowlton et al., 1991). Digestion of the plasmid with XhoI and transcription with T7 RNA polymerase generated a 141-bp probe (95-bp protected hybrid with test RNA). Mouse alkali atrial (MLC-1A) and ventricular light chain (MLC-IV) riboprobes were derived by EcoRI restriction of appropriate plasmids subcloned into the Bluescribe+ vector and transcription with T7 polymerase (Barton et al., 1988;Lyons et al., 1990). The RNA probes were labeled with [32P]CTP and purified on an 8 M urea, 6% polyacrylamide gel. Five to 15 pg of total RNA was hybridized with 50,000-150,000 cpm of the respective purified probe at 45 "C overnight. The unprotected RNA was subsequently digested with RNase A at 25 "C for 60 min. The reaction was terminated with SDS and proteinase K, and the reaction mixture was phenol-extracted and precipitated in ethanol. The RNase-resistant hybrids were analyzed by gel electrophoresis on a denaturing polyacrylamide/urea gel. The gel was dried, and autoradiography was performed for various time periods until signal was detected.
In Situ Analyses of Embryoid Bodies-In situ hybridization studies were performed essentially as described in detail by Lyons et al. (1990). Briefly, embryoid bodies were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS), dehydrated, and infiltrated with paraffin. Serial sections (5-7 pm in thickness) were mounted on gelatinized slides, deparaffined in xylene, and rehydrated. Following proteinase K digestion, the sections were post-fixed, treated with triethanolamine/acetic anhydride, washed, and dehydrated. In order to distinguish between transcripts within the same multigene family, riboprobes were derived from the 3'-noncoding untranslated regions (UTR) of the sequences of interest. The following probes were used 3' UTR of the mouse MLC-1A mRNA (Barton et al., 1988;Lyons et al., 1990), 3' UTR of the mouse MLC-1V mRNA (Lyons et al., 1990), 3' UTR of the rat ventricular myosin heavy chain (MHC p) mRNA (Mahdavi et al., 1982;Lyons et al., 1990), 3' end of the mouse atrial myosin heavy chain (MHC a) mRNA (Weydert et al., 1985;Lyons et al., 1990), 3' end of the mouse MLC-2V mRNA (Lee et al., 1992), and 3' end of the rat ANF mRNA (Knowlton et al., 1991). Sections were hybridized overnight at 52 "C with 50,000-75,000 cpm/pl 35S-labeled cRNA probe. The tissues were subjected to stringent washes (2 X ssc (1 X SSC is 0.15 M NaCl, 0.015 M sodium citrate solution), 50% formamide, 10 mM dithiothreitol at 65 "C) and RNase A treatment (20 pg/ml) at 37 "C for 30 min. The slides were dehydrated, immersed in photographic emulsion, and exposed for a week in light-tight boxes. Following photographic development, slides were analyzed using both light and dark field optics.
Immunohistochemistry-Single cell suspensions were prepared from entire differentiation cultures containing 40-70% of beating embryoid bodies (days 11-13 in suspension cultures). Cells were dispersed by gentle mechanical agitation in a spinner flask in the presence of 0.25% trypsin-EDTA (Life Technologies, Inc.), plated in plastic chamber slides (Nunc), and maintained in growth media (as for embryoid bodies) for a period of 48 h. Contractile activity was evident in single cells or groups of cells following their attachment to the tissue culture substrate. Indirect immunofluorescence studies were performed according to a previously described method . Briefly, fixation was carried out in 3% paraformaldehyde and following incubation in 50 mM NH&1 for 10 min, the cells were permeabilized with 0.2% Triton X-100 in PBS. Following a step to block nonspecific binding sites (1% bovine serum albumin, for 10 min), the monolayers were incubated at 37 "C for 60 min with the appropriate primary antibody, polyclonal rabbit antibody against the ventricular myosin light chain-2 (Iwaki et al., 1990) and/or monoc!onal mouse antibody against atrial natriuretic factor (generously provided by Dr. C. C. Glembotski, San Diego State University). Secondary antihodies were applied as follows: goat anti-rabbit IgG conjugated to fluorescein (Jackson ImmunoResearch Laboratories, Inc.) to recognize MLC-2V and biotinylated sheep anti-mouse immunoglobulin (Jackson ImmunoResearch Laboratories, Inc.), followed by Texas Red-conjugated strepavidin (Amersham Corp.) for ANF. The slides were rinsed in PBS, mounted with propyl gallate, and examined with both a Nikon microscope equipped with epifluorescence optics and a Bio-Rad 4000 series laser scanning confocal microscope.

ES Cell
Culture and Embryoid Body Differentiation-In order to inhibit ES cell differentiation and to promote proliferation and retention of a normal diploid karyotype, defined culture conditions are required. One approach is to seed ES cells over a layer of fibroblasts previously treated with irradiation or drugs to inhibit further cell division (Doetschman et al., 1985). An alternative culture method that obviates the need for feeder cells is the maintenance of ES monolayers in the presence of leukemia inhibitory factor, known to inhibit ES cell differentiation (for review, see Metcalf (1991)). We compared gene expression in both ES cells and derived embryoid bodies grown under these two culture conditions and were not able to detect significant differences in the temporal expression of muscle genes or the level of expression of these markers (data not shown).
Significant heterogeneity in the size and morphology of the cell aggregates is apparent throughout the course of ES cell differentiation in suspension cultures. In consideration of the potential influence of this variability on the onset and extent of gene transcription, culture conditions were manipulated to optimize the homogeneity of the starting cell aggregates and result in a more synchronous differentiation system. This was accomplished by ensuring that the differentiation program was initiated in fully dissociated ES cell suspensions (for details, see "Experimental Procedures"). This technique is not only known to improve the reproducibility of the timing of gene expression, it also results in higher overall levels of expression (Lindenbaum and Grosveld, 1990).* These conditions gave rise to cell aggregates, which typically displayed spontaneous foci of rhythmic contractions by day 8 in suspension culture.
Amplification of Gene Products Expressed in ES Cells and Embryoid Bodies-The kinetics of induction of a profile of endogenous muscle genes during ES cell differentiation was investigated using the reverse transcriptase-dependent polymerase chain reaction (RT-PCR). In order to perform a semi-quantitative analytical comparison between the amplified samples for a particular transcript on the same time course, cDNA samples were adjusted to yield a relatively equal amplification of a ubiquitously expressed standard, P-tubulin, prior to their analyses (Fig. L4). This was performed by quantitation of incorporated radiolabeled CTP into the Ptubulin standard and ethidium bromide staining following calibration of cDNA samples. Oligonucleotide-specific primers designed to assess mRNA levels throughout embryoid body development generally included the 3' UTR and, when possible, spanned at least one intron of the associated gene ( Table I). The amplification reactions were performed in parallel with appropriate controls (excluding RNA template or reverse transcriptase in the cDNA synthesis reaction) to discriminate that the amplified fragments were derived from mRNA and not from genomic DNA or other contaminants, which could potentially influence the PCR results. The temporal expression of the gene products of interest was determined in pools of embryoid bodies, each different time point representing additional days in suspension culture. The time point at which a particular transcript undergoes significant expression would suggest when that gene begins to play a role in the differentiation process. Representative amplifications of four to six independent RNA series are shown.-Muscle Gene Expression during ES Cell Differentiation-To determine the kinetics of muscle development during the differentiation of ES cells into embryoid bodies, the expression pattern of a subset of muscle mRNA sequences known to be developmentally regulated in the mouse heart was assessed. Analysis of myosin heavy chain transcript expression reveals the up-regulation of both a-and P-cardiac isoforms during embryoid body development (Fig. 1B). @-Cardiac MHC transcripts were detected from day 6 of differentiation and continued to be evident throughout the developmental period examined (up to 21 days in suspension culture). a-Cardiac MHC gene products were present from day 9 through day 21 of differentiation. Atrial alkali myosin light chain (MLC-1A) transcripts were initially evident between days 6 and 9 of embryoid body differentiation, with expression continuing as differentiation proceeded (Fig. 1B). The time course of MLC-1V gene expression demonstrated earlier transcriptional activation. A faint signal was detected in fibroblasts and embryonic stem cells and during the first few days of suspension cultures. A definite increase in the amplified mRNA signal was seen in embryoid bodies between days 6 and 20 of differentiation (Fig. 1B). This  between days 8 and 10 of differentiation by PCR analysis, and expression was maintained throughout the course of differentiation (Fig. IC).
In consideration of the exquisite sensitivity of PCR technology and its potential for the scoring of a positive signal on a physiologically insignificant amount of gene transcription, parallel RNase protection studies were performed to validate the amplification results. The kinetic pattern of MLC-1A gene expression as studied by RNase protection analysis confirmed the PCR findings, with transcripts detected between days 9 and 20 of differentiation (Fig. 2 A ) . The relatively earlier transcriptional activation of the ventricular isoform of the alkali myosin light chain was also confirmed by RNase protection assays. However, the low levels of MLC-1V gene expression identified as weak PCR signals in fibroblasts and ES cells were not evident (Fig. 2 A ) , probably a reflection of the greater sensitivity of the PCR amplification to detect trivial amounts of gene transcription. The temporal expression of the ANF gene was also supported by corresponding RNase protection studies (Fig. 2B). Thus, these analyses parallel the PCR data regarding the temporal expression of muscle gene transcripts during ES cell differentiation, documenting the appropriate, stage-specific pattern of expression of these genes in the differentiation cultures. A combined summary of the PCR and RNase protection analyses is presented in Table 11.
Expression of the MLC-2V Gene, a Molecular Marker for Ventricular Specification, during in Vitro Cardiogmesis-Recent studies in our laboratory regarding ventricular MLC-2 mRNA expression during murine cardiogenesis have documented early positional specification of this gene in the primitive heart tube exclusively in regions destined to become ventricle, providing evidence for regional specification of the ventricular muscle gene program at the earliest stages of mammalian cardiogenesis (O'Brien et al., 1993). In view of this property, which characterizes the cardiac MLC-2V gene as a genetic marker for ventricular specification, we wished to address whether in uitro cardiogenesis is also highlighted by the transcriptional activation of this muscle sequence. The data shown in Fig. 1C document the expression of ventricular regulatory light chain gene transcripts throughout the differentiation of ES cells into embryoid bodies. Amplification of MLC-2V mRNA yields a low level of gene expression initially identified between days 8 and 10 of embryoid body differentiation, concomitant with the appearance of spontaneous contractile activity in the cultures by microscopic observation. MLC-2V message continues to be evident with advancing days in culture.
RNase protection studies were also conducted to validate the MLC-2V amplification results. These analyses were performed using an antisense mouse cardiac MLC-2 riboprobe encompassing nearly the entire coding region. A protected fragment representing RNase-resistant hybrids is initially detected between 8 and 10 days of embryoid body differentiation (Fig. 2B). These results confirm the RT-PCR findings regarding the temporal expression of the MLC-2V gene in this model system.
Expression of Helix-Loop-Helix Transcription Factors in Developing Embryoid Bodies-During the last decade significant progress has been made in the field of muscle developmental biology by the identification and characterization of a gene family encoding master transcriptional factors that are able to mediate the myogenic programming of mesodermal cells (for review, see Weintraub et al. (1991) and Olson (1993)). The basic helix-loop-helix family of myogenic regulators that control establishment of the differentiated muscle phenotype includes rnyf-5, myogenin, MyoD, and MRF-4/ herculinlmyf-6. However, despite the shared expression of numerous contractile protein genes by striated muscles common to both skeletal and cardiac types, particularly throughout early development, none of the previously characterized myogenic regulatory genes has been identified in the embryonic, neonatal, or adult heart. Since the activation of the skeletal muscle program is characterized by the up-regulation of these sequences, we were interested in defining the onset of skeletal myogenesis in this in uitro system as identified by the expression of three myogenic regulatory factors, myf-5, myogenin, and MyoD. These data would also define whether the transcriptional activation of other muscle genes examined during ES cell differentiation was associated with skeletal uersus cardiac myogenesis. PCR analysis documented the expression of rnyf-5 transcripts in ES cells and throughout their differentiation (Fig. 3). An increase in myf-5 mRNA I n Vitro Model for Ventricular Specification were performed with singlestranded RNA probes generated in T3 and T7 vectors and labeled with ?' P (panel A, panel B,. Following overnight hybridization, the samples were treated with RNase A and the hybrids were separated by acrylamide gel electrophoresis. levels was seen with additional days in culture (days 12-20). Myogenin and MyoD transcripts were evident during days 16-20 of differentiation (Fig. 3). We were also interested in assessing the expression of the inhibitory helix-loop-helix protein, Id, in this in vitro system. Id is expressed in a wide variety of proliferating cell types, and mRNA levels appear to be high in proliferating myoblasts in culture with downregulation occurring during differentiation (Benezra et al., 1990). The present study demonstrates high levels of Id-1 expression in undifferentiated ES cells, which persist throughout embryoid body development (Fig. 3). No significant changes in Id-1 mRNA signal were observed as differentiation progressed in the embryoid bodies.

Spatial Pattern of Muscle Gene Expression in Beating Em-
bryoid Bodies-Developing embryoid bodies exhibit foci of spontaneous rhythmic contractile activity, the extent of this activity being primarily influenced by culture conditions. Light microscopy examination usually reveals either single

Summary of RT-PCR amplifications and RNase protection analyses
Table shows visual scoring of signal detected by a combination of PCR amplification and RNase protection analyses (-,  or, in some cases, multiple regions of contraction within a single embryoid body. To determine the spatial pattern of distribution of muscle gene transcripts within beating embryoid bodies and to address whether the same pattern of cardiac-specific gene expression observed in vivo is recapitulated in vitro, in situ hybridization with muscle-specific antisense probes was performed on serial sections of embryoid bodies a t various time points after differentiation was initiated. Figs. 4 and 5 display two different embryoid bodies cultured for 21 days and hybridized with riboprobes to myosin light chains, myosin heavy chains, and atrial natriuretic factor These findings are consistent with in situ hybridization studies in vivo, which document that MLC-2V and ANF mRNAs are localized in a subpopulation of myocytes in the developing heart tube (O'Brien et al., 1993). In contrast, MHC cy, MLC-lA, and MLC-1V transcripts are detected in most if not all cardiac myocytes from days 8-11 post coitum (Lyons et al., 1990).

Muscle Gene Expression in Embryoid Body-derived Cardiac
Muscle Cells-To investigate the expression of muscle gene translational products in single cells derived from embryoid bodies, immunofluorescence studies were performed following the dispersion of beating cell aggregates. The single cell level analysis using antibodies directed against the regulatory ventricular light chain and atrial natriuretic factor revealed cells that are positive for both MLC-2V and ANF expression. The staining pattern of the MLC-2V antibody, identified striations characteristic of the sarcomeric structures, confirming the muscle nature of the differentiated cells (Fig. 6A). The typical perinuclear location of ANF gene expression in the same field of cells is also shown (Fig. 6B). This finding provides evidence that embryoid body-derived muscle cells express genes that are restricted to specific cardiac cell types.
Further inspection of the stained cells reveals a t least four cellular subpopulations, as assessed by distinct immunofluorescence phenotypes. These include cells that do not recognize either MLC-2V or ANF antibodies, cells that express both gene products, and cells that only express either MLC-2V or ANF. Three of these cellular phenotypes are depicted in Fig.   6 (panels A, B, and E). Although the precise nature of these various cells is unclear, it could be speculated that the ANFpositive, MLC-2V-negative cells may represent cardiac muscle progenitors or ventricular cells at an early stage in the specification program and not yet expressing the regulatory ventricular myosin light chain. More interestingly, we suggest that these cells may be atrial cells, which characteristically express ANF a t high levels and negligible amounts of ventricular MLC-2. The identification and characterization of additional atrial-specific genetic markers should provide further information regarding the nature of this particular cellular subtype.

DISCUSSION
The Temporal Activation of Muscle Genes during in Vitro ES Cell Myogenesis Mimics Embryonic Myogenesis-Despite significant recent accomplishments in the area of myocardial growth and developmental regulation, defining the molecular basis underlying these events continues to present a fundamental challenge in the study of the developmental biology and molecular genetics of the cardiovascular system . The embryonic stem cell system provides a promising approach in the investigation of gene regulation during cardiac growth and development. Previous studies have demonstrated that hematopoietic differentiation in ES cell-derived embryoid bodies follows a well defined temporal pattern of gene expression, similar to that involved in the establishment of the hematopoietic system in the normal embryo (Lindenbaum and Grosveld, 1990;Schmitt et al., 1991;Keller et al., 1993). It has also been shown that embryoid bodies transcribe muscle genes (myosin heavy chain and tropomyosin isoforms) in an appropriate tissue-and developmental stage-specific pattern (Robbins et al., 1990;Muthuchamy et al., 19931, suggesting that developmental stages in ES cell culture reflect normal embryonic transitions. Likewise, the present study has exploited the features of this system that mimic in vivo embryogenesis to demonstrate that the developmental gene program activated during establishment and differentiation of muscle lineages in vitro parallels the sequence of events occurring in the developing mouse embryo. Although myf-5 gene expression is evident at earlier time points, this may be consistent with recent data suggesting that myf-5 up-regulation may not be unique to the establishment of the muscle lineage, since products of a transgene linked to the myf-5 control region have also been identified in the developing murine nervous ~y s t e m .~ I n myf-5-expressing embryoid bodies, neurofilament mRNA and protein have also been det e~t e d ,~.~ suggesting that myf-5 transcriptional activation during early stages of embryoid body development may be the result of concomitant activation of the neural program. The temporal pattern of expression of the myogenic factors, myogenin and MyoD, provides evidence for activation of the skeletal muscle program subsequent to the development of beating myocardium. This progression of in vitro myogenesis documents that the ES cell differentiation system has temporal fidelity to in vivo myogenesis during murine embryogenesis. This feature allows the opportunity for the study of the cardiac developmental program in a time window where there is absence of interference from skeletal muscle. The Mouse Embryonic Stem Cell Differentiation System as an in Vitro Model System for Cardiac Regional Specification-Cardiogenesis involves the expression of many genes that encode proteins required for muscle activity. Until recently, all the known mammalian cardiac chamber-specific protein isoforms were known to be co-expressed throughout the early looped heart, with establishment of regional specificity relatively late during cardiogenesis and, in some instances, following parturition (DeGroot et al., 1989;Lyons et al., 1990). Such is the case, for example, for the alkali myosin light chains, myosin heavy chains, and atrial natriuretic factor genes. However, recent evidence documenting early restriction of MLC-2V gene expression to the ventricular myocardium of the primitive heart tube prior to septation and the development of distinct cardiac chambers suggests that the molecular cues that lead to regional specification of some cardiac muscle cells may be operative very early in development (O'Brien et al., 1993). This pattern of restricted expression of the cardiac MLC-2V gene provides a potential marker for examining patterning of the heart tube during the early stages of cardiogenesis.
During their differentiation in culture, ES cells express transcriptional and translational products of the MLC-2V gene, demonstrating that cardiac cell type specification occurs during in vitro cardiogenesis. Although the MLC-2V gene is expressed at high levels in both cardiac and slow skeletal muscles in mice, the demonstration that cardiac development precedes skeletal myogenesis in vitro supports the view that MLC-2V expression in beating bodies without detectable myogenin or MyoD mRNA is most likely derived from developing cardiac muscle. The co-expression of MLC-2V and ANF genes in individual cells following the dispersion of early stage differentiated aggregates, in addition to recent electrophysiological data documenting the expression of tetrodotoxinresistant Na+ channels,* provide further compelling evidence for the cardiac-like nature of these cells.
The demonstration of the expression of a ventricular-specific muscle marker in an in vitro system that lacks a primor- dial heart tube, in the present study, provides evidence that regional specification can occur independently of positional cues or physiologic stimuli associated with the development of functional cardiac chambers. This finding suggests that the mechanisms that restrict gene expression to specific muscle cells may not be totally dependent upon formation of a heart tube and the associated hemodynamic influences. The question remains as to whether specification of the MLC-2V gene during the earliest stages of cardiogenesis results from the direct commitment of mesodermally derived precursor cells to the ventricular muscle lineage without an intermediary cardiac muscle stem cell, or whether a common cardiac muscle progenitor cell exists and gives rise to both atrial and ventricular myocytes.
The analysis of single cells derived from pools of beating embryoid bodies demonstrates at least four subpopulations of cells based on MLC-2V and ANF gene expression. We have recently developed techniques for the harvesting and purification of cardiac muscle cells from beating, differentiated embryoid bodies based on modification of previous protocols for the isolation and culture of primary cardiomyocyte~.~ These cells also display the same immunofluorescence phenotypes, thereby confirming these findings. Under these conditions, the various cell types are detected at the following frequency: MLC-2V-positive/ANF-positive, 7%; MLC-2Vpositive/ANF-negative, 44%; MLC-2V-negative/ANF-positive, 13%; and MLC-2V-negative/ANF-negative, 36%. These results further establish the mouse ES cell system as a model for cardiac chamber specification and suggest the possibility of tracking effects on these individual cell populations following genetic alterations of the parental ES cell lines.
Conclusions-In summary, our findings demonstrate that the embryonic stem cell system can model certain aspects of early cardiac muscle commitment and differentiation. The finding of a ventricular-specific marker in differentiated ES cells suggests that this may serve as a promising approach in the elucidation of the regulatory mechanisms that determine cell fate and control the expression of specific genes in specialized cell types during cardiogenesis. Modification of specific candidate genes via homologous recombination in ES cells should allow the opportunity to explore potential influences of targeted genes on the cardiac muscle program and cardiac chamber specification in vitro. This system should also provide a valuable approach for exploring genetic manipulations within an in vitro cardiac context that may otherwise result in embryonic lethality in transgenic animals. The in vitro differentiation of mouse embryonic stem cells into cardiac muscle cells with atrial-or ventricular-specific properties may ultimately allow for studies of chamber specification in genetically engineered cardiac muscle cells, thus providing a useful tool in the dissection of developmental cardiogenesis. In addition to the powerful features described above, this system may provide for the generation of large amounts of embryonic material and the biochemical purification of factors with potential influences on cardiac growth and differentiation. Ultimately, this system may also allow the opportunity for the identification and isolation of cardiac cell progenitors.
Bruce Micales, the generous use of the microscopy facilities at the San Diego Microscopy and Imaging Resource under the direction of Dr. Mark Ellisman (National Institutes of Health Grant RR04050), a n d the invaluable technical assistance provided by Thomas Deerinck.