Modulation of L-type Calcium Channel Expression during Retinoic Acid-induced Differentiation of H9C2 Cardiac Cells*

The molecular mechanisms underlying the developmental regulation of L-type voltage-dependent Ca2+ channels (VDCCs) are still unknown. In this study, we have characterized the expression patterns of skeletal (α1S) and cardiac (α1C) L-type VDCCs during cardiogenic differentiation in H9C2 cells that derived from embryonic rat heart. We report that chronic treatment of H9C2 cells with 10 nm all-trans-retinoic acid (all-trans-RA) enhanced cardiac Ca2+ channel expression, as demonstrated by reverse transcription-polymerase chain reaction, immunoblotting, and indirect immunofluorescence studies, as well as patch-clamp experiments. In addition, RA treatment prevented expression of functional skeletal L-type VDCCs, which were restricted to myotubes that spontaneously appear in control H9C2 cultures undergoing myogenic transdifferentiation. The use of specific skeletal and cardiac markers indicated that RA, by preventing myogenic transdifferentiation, preserves cardiac differentiation of this cell line. Altogether, we provide evidence that cardiac and skeletal subtype-specific L-type Ca2+ channels are relevant functional markers of differentiated cardiac and skeletal myocytes, respectively. In conclusion, our data demonstrate that in vitro RA stimulates cardiac (α1C) L-type Ca2+ channel expression, therefore supporting the hypothesis that the RA pathway might be involved in the tissue specific expression of Ca2+ channels in mature cardiac cells.

It is accepted that adult cardiac myocytes exhibit tissuespecific expression of L-type voltage-dependent calcium channels (VDCCs), 1 i.e. the class C (␣ 1C ) channel subtype (1). However, despite the large body of information on the properties of these VDCCs in adult mammalian heart (1,2), little is known concerning the regulation of the functional expression of these channels in the developing heart. No triggering or regulatory mechanism for cardiac L-type VDCC expression has been clearly identified to date, although there is some evidence that cardiac embryonic differentiation may be controlled by specific inducers, such as transforming growth factor-␤ 1 (3), triiodothyronine (4), and retinoic acid (5).
The skeletal muscle, in which myoblasts fuse to form mature multinucleated myotubes, is a good model to address basic mechanisms involved in muscle differentiation (6). It is now clear that the expression of the skeletal L-type VDCC subtype specifically correlates with skeletal muscle differentiation (7). In the cardiac muscle lineage, cell proliferation and differentiation are not mutually exclusive events, and embryonic cardiac myocytes undergo maturation during development. The appearance of L-type VDCCs occurs at early stages of cardiogenesis (8,9). However, the study of L-type VDCC expression during cardiogenic differentiation is a difficult task, mainly because of the laboriousness of obtaining embryonic cardiac cells from early embryos.
The development of in vitro models that recapitulate molecular events underlying the establishment of the cardiac muscle phenotype is an important step toward understanding the mechanisms involved in the regulation of cardiac ion channel gene expression (9). An interesting experimental model is the clonal cardiac cell line H9C2 derived from embryonic rat heart (10). This cell line has been studied for its electrophysiological properties, and it was demonstrated that H9C2 cells exhibit L-type VDCCs with cardiac-specific characteristics (11), although concomitant expression of skeletal L-type VDCCs occurs in H9C2 cells (12).
Here, we present evidence that the H9C2 cell line is valuable for the investigation of the ontogenic properties of these two Ca 2ϩ channel isotypes. We have conducted experiments to characterize the changes in the expression of VDCCs during differentiation of H9C2 cells. We show that upon a chronic treatment with 10 nM all-trans-retinoic acid (all-trans-RA), H9C2 cells exhibit enhanced cardiac Ca 2ϩ channel expression as a consequence of the maintenance of their cardiac phenotype. By contrast, myogenic transdifferentiation is inhibited. The data unambiguously demonstrate that the functional expression of muscle L type VDCC isotypes (␣ 1C versus ␣ 1S ) correlates well with specific differentiation processes, cardiogenesis versus myogenesis.

EXPERIMENTAL PROCEDURES
Cell Culture-H9C2 cells obtained from the European Collection of Cell Cultures were grown as a stock of cells in culture flasks in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum (FCS) (Eurobio), glutamine (2 mM), penicillin (100 IU/ml), and streptomycin (100 g/ml) under standard culture conditions (37°C, 5% CO 2 , and water-saturated atmosphere). Before confluence, cells were split, plated at a lower density in Petri dishes in 10% FCS culture medium and cultured for 1 day. Cells were then cultured in Dulbecco's modified Eagle's medium supplemented with 1% FCS. The culture medium was replaced every 2 days. Stimulation with retinoic acid analogs was performed daily. All-trans-RA, as well as 9-cis-RA (Sigma) were diluted in the dark in Me 2 SO (1 mM stock solution) and stored at Ϫ20°C. Aliquots were used only once and diluted extemporaneously in * This work was supported in part by the Association Française contre les Myopathies, Association pour la Recherche contre le Cancer Grant ARC9011, Ministère de la Recherche Grant ACC-9, Sciences du Vivant, and the Program Génome du CNRS. A travel grant was provided by the Center Européen de Bioprospectives and Dr. B. Pourrias. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
RT-PCR Experiments-Multiplex RT-PCR detection of transcripts encoding ␣ 1C and ␣ 1S was chosen for its high sensitivity allowing adequate semiquantitation of these transcripts. Total RNAs were prepared by using the guanidinium thiocyanate-phenol-chloroform procedure (14,15) and stored in sterile water. First strand cDNA synthesis was carried out for 1 h at 37°C in 50 l containing 5 g of DNase treated-RNAs, 1 l of oligo(dT) primers (500 g/ml), and Superscript II reverse transcriptase (50 units, Life Technologies, Inc.), according to the manufacturer's instructions. The enzyme was then heat-inactivated (5 min at 95°C).
PCR primers that were designed according to Mejia-Alvarez et al. (12) to amplify transcripts encoding ␣ 1C and ␣ 1S were first assayed using full-length cDNAs of these subunits (kindly supplied by Dr T. P. Snutch and B. Fontaine, respectively) as templates. Multiplex RT-PCR conditions allowing the simultaneous amplification of a 377-base pair fragment related to ␣ 1C (forward primer, 5Ј-GGAGAGTTTTCCAAA-GAGAGG-3Ј) and a 201-base pair fragment related to ␣ 1S (forward primer, 5Ј-CGCGAGGTCATGGACGTGGAG-3Ј) were optimized to allow maximal yield using a common reverse primer, 5Ј-GATCACCAGC-CAGTAGAAGAC-3Ј. One-fifth (10 l) of each PCR was electrophoretically separated on 2% agarose gels and stained with ethidium bromide. Gels were then scanned, and a densitometric analysis was performed using ImageQuant (Molecular Dynamics). Gel figures were produced using Adobe Photoshop (version 4.0) and printed on a Kodak Colorease thermal sublimation printer. cDNA sequences were further verified by sequencing.
Characteristics of the Primary Antibodies-A monoclonal antibody developed by Morton and Froehner (16) that recognizes an extracellular epitope of the ␣ 1S subunit (mab427; Chemicon) was used at a 1/100 dilution for indirect immunofluorescence experiments. A polyclonal antibody (pA1C) recognizing the ␣ 1C subunit was produced in rabbit (Eurogentec). It was directed against a peptide from the intracellular II-III loop (819-TKINMDDLQPSENEDKS-835) coupled to KLH. Working dilutions of the crude serum for immunofluorescence experiments were in the range of 1/400 to 1/200. As a myogenic marker, we used a monoclonal antibody specific of rat myogenin: the F5D hybridoma, kindly supplied by Dr. Wright. The supernatant was used undiluted. In addition, a monoclonal antibody specific of skeletal muscle troponin T (clone JLT-12; Sigma) was used. In this case, the mouse IgG1 fraction was used at a 1/200 dilution in our immunofluorescence experiments. As a cardiac marker, we used a polyclonal antibody that specifically recognized myosin light chain 2v (MLC-2v), kindly supplied by Drs. S. Kubalac and K. Chien. Working dilutions for immunofluorescence experiments were in the range of 1/100. A polyclonal antibody that specifically recognizes the glyceraldehyde-3-phosphate dehydrogenase, kindly supplied by J. M. Blanchard (17), was used as internal control in Western blotting experiments.
Western Blotting and Immunoblotting Analyses-Membrane extracts were prepared from 2-month-old Wistar rats. Heart and skeletal muscles were quickly removed after death. Tissues were disrupted at room temperature in the extraction buffer prepared extemporaneously (50 mM Tris-HCl, pH 8, 0.1% saponin; 1 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml leupeptin, 0.1 mg/ml soybean trypsin inhibitor) using a polytron homogenizer. Homogenates were centrifuged for 5 min at 3500 ϫ g, and supernatants were collected and boiled for 5 min in a denaturing solution containing ␤-mercaptoethanol (5%) and bromphenol blue. Samples were then analyzed by SDS-polyacrylamide gel electrophoresis followed by Western blot immunodetection. Briefly, after separation on SDS-polyacrylamide 7.5% gels, proteins were electrotransferred onto nitrocellulose membranes (40 mA, overnight). Membranes were immersed overnight at 4°C in TBE buffer (90 mM Trisborate, 2 mM EDTA, pH 8) supplemented with 5% skim milk and 0.05% Tween. Membrane strips were then incubated for 1 h at room temperature in primary antibody solutions diluted in TBE containing 0.1% Tween and 5% skim milk (mab427, 1/300; pA1C, 1/500) and rinsed three times for 10 min each with TBE 0.05% Tween. Blots were subsequently incubated for 30 min at room temperature with horseradish peroxidase-conjugated secondary antibody diluted at 1/5000 and washed three times before visualizing the immunoreactive bands using ECL reagents (Amersham Pharmacia Biotech), used according to the manufacturer's protocol. For the detection of ␣ 1S in skeletal muscle extracts with mab427, similar results were obtained using purified goat anti-mouse IgG coupled to alkaline phosphatase (1/5000, Jackson Im-munoResearch Laboratory). Membranes were then stripped in 100 mM ␤-mercaptoethanol, 2% SDS, 62.5 mM Tris-HCl, pH 6.8, at 50°C for 30 min and reexposed to the glyceraldehyde-3-phosphate dehydrogenase antibody (17).
Indirect Immunofluorescence Experiments-Cells were rinsed with PBS, fixed overnight at 4°C in a paraformaldehyde solution (4% in PBS), and then washed three times for 5 min each with PBS. Blocking of nonspecific sites was performed by incubating cells in a PBS-bovine serum albumin (3%) solution for 30 min at 37°C. When required, cell permeabilization was performed by treating with 0.03% Triton in PBS for 5-10 min. Cells were then incubated for either 1 h at 37°C or overnight at 4°C with the first antibody at the desired dilution in PBS-bovine serum albumin. Cells were washed three times with PBS and incubated for 45 min at 37°C with goat anti-rabbit or goat antimouse secondary antibodies, fluorescein isothiocyanate-conjugated (Cappel). They were then washed three times with PBS buffer and distilled water before labeling nuclei by a 1-min treatment with Hoescht 33258 dye (Sigma). Coverslips were washed in distilled water and mounted in a mixture containing 15% (w/v) Airvol 205, 33% glycerol dissolved in Tris-HCl buffer. Digital images acquired on a microscope (Leica) using a Kodak DCS420 camera were analyzed with no or a minimal image treatment on a Silicon Graphics workstation with the Imgwork and Iris Showcase 3.2 software packages (Silicon Graphics) or Adobe Photoshop (version 4.0). Data were printed using a Kodak Colorease thermal sublimation printer. rapidly and exhibited an L-type Ba 2ϩ current (Fig. 1B), as described for cardiac myocytes (2). The density of this current, named I CARD , was 3.2 Ϯ 0.7 pA/pF, n ϭ 31 (Fig. 1C). I CARD was detected in all mononucleated cells, whatever their morphology. This current was sensitive to both dihydropyridine antagonists and agonist and was up-regulated by the ␤-adrenergic agonist isoproterenol (not shown), as described previously (11). A low voltage activated Ba 2ϩ current (T-type), observed in very few cells (2 out of 80), was not characterized further.

L-type Ca 2ϩ Currents in H9C2 Myoblasts
I SKM -When H9C2 cells reached confluence, they were switched to a 1% FCS medium, and several elongated and multinucleated cells were observed (Fig. 1, D and G). These cells exhibited a specific labeling for myogenin as well as for troponin T and MyoD (18) and thus could be identified as skeletal muscle myotubes. H9C2 myotubes displayed a very slow activating and inactivating L-type Ba 2ϩ current (Fig. 1, E and H). This current, named I SKM , resembled the one recorded in skeletal muscle myotubes and in a variety of differentiated skeletal muscle cell lines and primary cultures (19). The distinct kinetics of I SKM and I CARD allowed us to easily identify cells that exhibit the two Ba 2ϩ current subtypes (Fig. 1E), as well as to measure their respective amplitudes. Indeed, we found that in small myotubes (2-5 nuclei, n ϭ 12), I CARD (2.5 Ϯ 0.8 pA/pF) was generally associated with I SKM (3.1 Ϯ 0.7 pA/ pF) (Fig. 1, E and F). These two currents, despite their difference in activation kinetics, were clearly high voltage activated and dihydropyridine-sensitive. In addition, we never identified low voltage activated T-type Ba 2ϩ current in such myotubes (n ϭ 50). In larger myotubes with up to 5 nuclei (Fig. 1G), I SKM was of higher density (I SKM , 8 Ϯ 2.3 pA/pF; n ϭ 6), and no more I CARD was detected (Fig. 1, H and I). These data indicate that the functional expression of skeletal muscle L-type Ca 2ϩ channels occurs concomitantly with the fusion of the myoblasts. Our electrophysiological experiments, conducted on a large number of cells (up to 150) have revealed that the presence of I SKM is exclusive of myotubes. However, it is important to note that I CARD was still detected in mononucleated cells present in confluent cultures, with an average Ba 2ϩ current density of 1.8 Ϯ 1.1 pA/pF (n ϭ 25).

Enhanced Cardiac Differentiation in RA-treated H9C2 Cells
The H9C2 cell line was derived from rat cardiac ventricule embryonic cardiocytes (10). Freshly plated cells were strictly mononucleated myoblasts, as judged from DNA staining by Hoescht reagent. We took advantage of this cellular model, capable of expressing multiple isoforms of the L-type Ca 2ϩ channels, to seek factors that would modulate muscle differentiation together with Ca 2ϩ channel expression. We describe here the effects of a chronic exposure of differentiating H9C2 cells to RA (Fig. 2). Treatment of H9C2 cells with RA analogs (10 nM all-trans-or 9-cis-) for 5-7 days markedly reduced proliferation, as judged from the lower percentage of BrdUrdlabeled nuclei after short exposure (30 min) of the cells to this reagent (not shown). RA-treated cells exhibited morphological changes with rather large and rounded cells ( Fig. 2A), distinct from the elongated myotubes present in control cultures (Fig.  2B) or in Me 2 SO-or cAMP-treated cultures (not shown). Large cells found in both RA-exposed cultures (Fig. 2C) and the myotubes (Fig. 2D) were multinucleated. These various parameters define the morphological phenotype of RA-exposed H9C2 cells, which requires a low FCS concentration (1%).
Biochemical differentiation of H9C2 cells was characterized using skeletal and cardiac protein markers. RA treatment of differentiating H9C2 cells markedly altered myogenic differentiation because none of the multinucleated cells were positive for the expression of myogenin (Fig. 2E) and troponin T (Fig.  2G). By contrast, myotubes present in control differentiated cultures (control, 1% FCS, 1% Me 2 SO, and 1 M 8-Br cAMP) showed typical staining for myogenin (Fig. 2F), MyoD, and troponin T (Fig. 2H). After 4 -5 days of culture in 1% FCS, 30 -40% of troponin T-positive cells could be observed. Differentiating H9C2 cells were then checked for the presence of the ventricular MLC-2v, a well characterized marker of cardiac differentiation (20,21). Fig. 2I shows that H9C2 cells treated with RA exhibited an intense staining for MLC-2v, typical of that obtained with cultured cardiac cells. By contrast, H9C2 cells cultivated in myogenic differentiation conditions displayed only a faint staining for MLC-2v (Fig. 2J). At the myoblast stage, H9C2 cells were positive for both MLCv-2a and MLC-2v cardiac markers (not shown), an observation that is similar to that reported for newborn rat cardiac myocytes in primary culture (21). In contrast, and as reported previously (18), these cells were negative for myogenic differentiation markers, such as myogenin, MyoD, and troponin T.

Ca 2ϩ Channel Expression in RA-treated H9C2 Cells
Using RT-PCR, we have identified a change in Ca 2ϩ channel ␣ 1C and ␣ 1S subunit mRNA distribution at a semiquantitative level. Multiplex PCR, performed with a common antisense primer for the amplification of both ␣ 1C and ␣ 1S specific frag- ments (see under "Experimental Procedures"), was conducted by using reverse-transcribed products from mRNAs collected at days 1, 2, 3, 5, and 10 in control (Fig. 3A) and 10 nM all-trans-RA-treated (Fig. 3B) cultures. Similar results were obtained from three independent batches of culture, thus indicating that the relative changes in amounts of PCR fragments presented in Fig. 3 truly reflected the effect of RA treatment. These experiments showed that the amount of mRNA encoding ␣ 1S present in old control cultures that contain large myotubes was increased (Fig. 3C). For ␣ 1C , we noticed that the amount of PCR products was transiently enhanced at day 3 but declined in older cultures. By contrast, the amount of ␣ 1C mRNA was increased in older cultures (Fig. 3D), whereas that of ␣ 1S remained unchanged during cardiogenic differentiation of H9C2 cells.

Immunoblotting Analysis of ␣ 1C and ␣ 1S Proteins
At the protein level, the pore-forming subunits ␣ 1C and ␣ 1S are the signature of the cardiac and skeletal L-type channel expression, respectively. In our experiments, a polyclonal antibody that recognized ␣ 1C (pA1C) and a monoclonal antibody recognizing ␣ 1S (mab427) were used (see under "Experimental Procedures"). mab427 (1/300) and pA1C (1/500) antibodies reacted specifically with a 170-kDa protein (Fig. 4A) and a 195-kDa protein (Fig. 4B), respectively, when used in Western blotting experiments with skeletal muscle and heart membranes purified from a 2-month-old rat. The pA1C reactivity, which appeared as a single band, could be blocked by preincubating the antibody with 3 M of free peptide (Fig. 4B), thus confirming the specificity of pA1C for ␣ 1C .
Immunodetection of ␣ 1S and ␣ 1C proteins with mab427 and pA1C was further performed using Western blots generated with purified membrane protein preparations from control and RA-treated H9C2 cells. These experiments were conducted several times, and typical results are presented in Fig. 4, C and D.
In RA-treated cultures, mab427 revealed a faint band (Fig. 4C), whereas the signal observed with pA1C was stronger (Fig. 4D), compared with control cultures. Densitometric analysis revealed a 3-fold increase (3.1 Ϯ 0.2) in ␣ 1C expression, whereas ␣ 1S exhibited a 2-fold decrease (2.1 Ϯ 0.1) at the protein level. Similar results were obtained with two other antibodies: (i) a polyclonal antibody that recognizes ␣ 1C (kindly supplied by Dr. T.P. Snutch), and (ii) a polyclonal antibody raised against purified skeletal muscle Ca 2ϩ channels that identifies ␣ 1S (Upstate Biotechnology) (data not shown). Immunoblotting data thus demonstrated that H9C2 cells expressed ␣ 1S and ␣ 1C proteins typical of those present in rat skeletal and cardiac muscles and that RA treatment of differentiating cells enhanced ␣ 1C expression while decreasing the amount of ␣ 1S protein.

Immunostaining of ␣ 1C and ␣ 1S Proteins in Differentiating H9C2 Cells
Immunostaining patterns of ␣ 1S and ␣ 1C proteins presented in Figs. 5 and 6 were obtained using the indirect immunofluorescence technique performed with the mab427 and pA1C antibodies from a representative batch of differentiated control and RA-treated H9C2 cultures. Various controls including the immunostaining of several cardiac and skeletal markers and nuclei visualization (with Hoescht 33258) were concomitantly performed. The results described in this section were reproduced in more than ten several independent experiments.
In preliminary experiments performed with recombinant channel proteins expressed in fibroblasts, we verified that both antibodies recognized efficiently and specifically ␣ 1C and ␣ 1S proteins (not shown). H9C2 myoblasts showed a positive staining with pA1C (1/200) (Fig. 5A), in good agreement with electrophysiological data described above. The specific staining observed with pA1C was blocked upon incubation with 3 M of the corresponding ␣ 1C peptide (Fig. 5B). By contrast, the monoclonal antibody mab 427 (1/100) generated only a background of labeling in H9C2 myoblasts, indicating that there was no expression, or very little, of the ␣ 1S isoform at the myoblast stage (Fig. 5C). A positive staining could be observed with mab427 in two-nuclei myotubes (Fig. 5C, arrowhead), indicating that ␣ 1S was expressed concomitantly to myoblast fusion.
Immunofluorescence studies were then performed on differentiated control and RA-treated cells (10 nM all-trans-RA). Large myotubes found in control cultures maintained for up to 7 days in differentiating medium (1% FCS) were markedly stained with mab427 (Fig. 5D), similarly to that described for cultured myotubes stained with antibody (16,22). These myotubes exhibited a typical punctuate pattern with membrane "hot spot" for ␣ 1S immunostaining. Hot spot staining for ␣ 1S could be preferentially observed at the edge of the membrane area of the myotubes as well as in the close vicinity of the nuclei (Fig. 5D, arrowheads). In most RA-treated cells, ␣ 1S immunostaining was close to the background level in most cells, although a small percentage of these cells (Ͻ10%) did show a positive mab427 labeling (Fig. 5E). This staining was weak in intensity and different from the typical ␣ 1S immunostaining described for myotubes because it was rather homogeneous throughout the membrane surface in these cells. Although this observation is in good agreement with the immunoblotting data, it is nevertheless intriguing from a functional point of view, because none of the cells present in RA-treated cultures that we examined by electrophysiology exhibited a slow Ba 2ϩ current similar to I SKM found in control myotubes (see below).
In RA-treated cultures, we observed a stronger staining with the pA1C antibody, compared with control H9C2 cells (Fig. 5, F and G). H9C2 cells in control cultures, even large myotubes, displayed a weak pA1C staining (Fig. 5F). However, we have evidence that the labeling of large myotubes is less specific than that in mononucleated cells because preimmune serum showed similar labeling of these cells (not shown). When H9C2 cells were cultivated in the presence of RA during differentiation, pA1C staining was enhanced, revealing a marked increase in ␣ 1C protein expression that was observed both in mononucleated and multinucleated cells (Fig. 5G). Clearly, the pattern of ␣ 1C staining in RA-treated cells was different from the pattern of ␣ 1S staining in myotubes. Staining for ␣ 1C is more diffuse and can be observed all over the membrane surface with some circular spots of higher intensity (Fig. 5G,  arrowheads).
To further characterize the ␣ 1C and ␣ 1S stainings, we performed double labeling experiments with the pA1C and mab427 antibodies in differentiated H9C2 cells. Results presented in Fig. 6 show that ␣ 1C and ␣ 1S expressions are mutually exclusive in differentiated cells, under both untreated and RA-treated culture conditions. In a control myotube exhibiting typical hot spot and punctate pattern for ␣ 1S staining, only a background of pA1C labeling could be observed (Fig. 6, A-C). In contrast, in cells present in cultures treated with 10 nM alltrans-RA, a weak ␣ 1S staining was observed (Fig. 6D), whereas pA1C labeling revealed an intense staining with a typical spot of ␣ 1C staining (Fig. 6E).

Functional Expression of L-type Ca 2ϩ Channels in RAtreated H9C2
Cardiac-type Ba 2ϩ currents (I CARD ) were recorded both in multinucleated and mononucleated cells present in differentiating H9C2 cultures treated with RA analogs (Fig. 7). In multinucleated cells, I CARD exhibited typical properties of a cardiac L-type current (Fig. 7A). Interestingly, the inactivation kinetics was faster for Ba 2ϩ currents recorded on multinucleated cells ( fast ϭ 47 Ϯ 12 ms at 10 mV; n ϭ 6) than with mononucleated cells (Fig. 7B). In both multinucleated and in mononucleated cells, I CARD was modulated by dihydropyridine molecules, as illustrated by the block observed following application of 1 M PN 200 -110 (Fig. 7, A and B, open symbols). These two types of cells exhibited I CARD with similar currentvoltage relationships, as illustrated in Fig. 7C. No I SKM was detected in multinucleated cells (n Ͼ 40) in RA-treated cultures (Fig. 7, A and D), unlike multinucleated cells (myotubes) in control cultures or cultures treated with Me 2 SO (1%) or 8Br-cAMP (1 M) that showed comparable density for I SKM (Fig.  7D). Furthermore, we report here that the cardiac current density was significantly higher in multinucleated cells (5.9 Ϯ 1.8 pA/pF; n ϭ 25) than in mononucleated cells from RAtreated cultures (1.4 Ϯ 0.9 pA/pF; n ϭ 8) or in myoblasts (2.5 Ϯ 0.9 pA/pF; n ϭ 7) and differentiating myotubes cultivated in control conditions (Fig. 7D). Comparable results were obtained using 100 nM 9-cis-RA instead of 10 nM all-trans-RA (Fig. 7D). Direct modulation of cardiac or skeletal L-type Ca 2ϩ channels by RA could be ruled out in our experiments because no change in L-type current properties were observed in cells acutely treated with 0.1 M RA. DISCUSSION The important finding of this study is that expression of cardiac (␣ 1C ) L-type voltage-dependent Ca 2ϩ channels is up-regulated during retinoic acid-induced differentiation of embryonic cardiac H9C2 cells. The data also provide strong evidence that myogenic transdifferentiation of H9C2 cells is suppressed by RA treatment, further supporting the possibility that RA favors establishment of the cardiac phenotype in H9C2 cells. Altogether, our analysis of Ca 2ϩ channel activity in differentiating H9C2 cells supports the hypothesis that ␣ 1C and ␣ 1S channel proteins are phenotypic markers of cardiogenic and myogenic differentiation, respectively.
Despite their cardiac origin, defined by the expression of cardiac markers such as MLC-2a and MLC-2v at the myoblast stage (20), differentiating H9C2 cells cultivated in a low concentration of serum transdifferentiate and acquire a skeletal muscle phenotype. Evidence for skeletal muscle differentiation of H9C2 cells is based on several criteria, including myoblast fusion; expression of specific markers, such as myogenin, troponin T, and MyoD; and nicotinic receptors (10). Expression of myogenin and MyoD, two specific markers of skeletal muscle differentiation, has not been described in cardiac cells (23). In contrast, skeletal muscle differentiation can be induced in cardiac cells when myogenic factors are overexpressed (24). Myogenic differentiation of H9C2 cells is not caused by a constitutive expression of myogenic factors but most likely results from a transdifferentiation process responsible for a phenotypic switch from cardiac to skeletal muscle.
Chronic treatment with RA can prevent transdifferentiation of H9C2 cells because no typical myotubes could be observed in these cultures. We describe the presence of multinucleated cells in RA-treated cultures, but we have no evidence to date that myoblast fusion occurs in RA-treated H9C2 cultures, and we suggest that an alternative mechanism is involved. It has been shown that DNA replication followed by karyokinesis without cell division occurs in cardiac myocytes giving rise to adult binucleated cells, and cardiomyocytes can become multinucleated when cultured (25). The latter mechanism, although not yet proven, seems likely to operate in RA-treated H9C2 cultures. RA has been shown to affect myogenic differentiation in different ways; it inhibits myogenic differentiation of embryonic muscle cells in primary cultures (26) while enhancing myogenin and MyoD expression and myoblast fusion in skeletal muscle cell lines (C2 and L6) (27). H9C2 cells exposed to RA are negative for skeletal markers, indicating, therefore, that RA inhibits myogenic differentiation of these cells.
Our data show that chronic treatment of H9C2 cells with RA analogs maintains their cardiac phenotype. The expression of MLC-2v even suggests that RA-treated cells exhibit enhanced cardiac differentiation that occurs with a ventricular specification (20,28). Recently, the role of RA in heart development has been assessed in genetically manipulated mouse lines (29,30). Targeted disruption of the gene encoding RXR␣ subunits resulted in mice showing ventricular chamber defects reminiscent of vitamin A deficiency syndrome (31)(32)(33). However, it remains unclear whether a direct RA-dependent pathway is necessary for proper heart development (34). In contrast, in vitro studies have indicated essential involvement of the RAdependent pathways in the establishment and maintenance of cardiac differentiation in stem cell-derived cardiomyocytes and neonatal rat cardiac myocytes (20,35,36). Consistent with these observations, treatment of P19 embryonic carcinoma cells with RA favored cardiac commitment (37). In a similar manner, the requirement for RA in the maintenance of the cardiac phenotype of several cell lines such as QCE-6 (38) and HL-1 (39) has been shown. Because H9C2 cells are easy to manipulate, more and more investigations are performed with this cell line. We have shown that H9C2 cell line is a suitable cardiac cellular model when exposed to RA, and therefore, our study is an important contribution that delineates the experimental requirements for the maintenance of the cardiac phenotype of this cell line. Consequently, we suggest that H9C2 cells can serve as a model system for a variety of investigations that are difficult to handle on primary cardiac myocytes or related to cardiac maturation.
An up-regulation of cardiac/␣ 1C L-type Ca 2ϩ channel expression is observed in association with RA treatment. The increase in ␣ 1C expression is most likely a consequence of the sustained cardiac phenotype because it correlates with MLC-2v expression. Similarly, increased expression of functional cardiac Ltype Ca 2ϩ channels was observed in developing cardiomyocytes derived from embryoid bodies (9), further indicating that this channel is an hallmark of early cardiogenesis. We have demonstrated that an important consequence of both myogenic and cardiogenic differentiation is the triggering and tuning of the expression of functional L-type channel isotypes. Untreated H9C2 myoblasts or RA-treated cells exhibit typical cardiac L-type Ca 2ϩ channels. In contrast, H9C2 cells that have undergone transdifferentiation to a myogenic phenotype express skeletal muscle L-type Ca 2ϩ channels. Overall, our data reconcile some of the apparent discrepancies between previous studies, which reported the simultaneous presence of cardiac and skeletal isoforms of the L-type Ca 2ϩ channel in this cell line (11,12). Both biochemical and patch-clamp experiments were necessary to delineate the expression patterns of these channel isotypes during differentiation. Altogether, our data lead us to conclude that the electrophysiological properties of these two channel isotypes in H9C2 cells are similar to those found in previous studies (2,40). H9C2 cells can therefore be exploited for further channel investigations, such as subunit modulation, pharmacology, and regulation pathways, known to modulate cardiac or skeletal Ca 2ϩ channels. In addition, it will be important to determine whether Ca 2ϩ channel expression, using antisense or overexpression strategies, influences the differentiation pathways of H9C2 cells.
The identification of a cardiac-to-skeletal transdifferentiation process in vitro provides an opportunity to study the specificity of cardiac and skeletal muscle differentiation programs. Cardiac development proceeds from the initial commitment of mesodermally derived progenitors, as for skeletal muscle. The transition of embryonic to adult cardiomyocytes is accompanied by changes in the expression patterns of a variety of contractile protein isoforms. Several fetal isoforms, including skeletal muscle ␣-actin (41), are identified as skeletal muscle isoforms. Whether cardiomyocyte maturation is accompanied by changes in Ca 2ϩ channel subunit or isotype expression has not been studied. Whether developing cardiomyocytes that derive from embryoid bodies (9) have the potential to undergo a myogenic transdifferentiation and/or express several isotypes of Ca 2ϩ channels is also unknown and remains important to evaluate. In this respect, it is important to note that newborn cardiac myocytes in primary culture express functional ␣ 1S that generates contractile activity (42). Conversely, skeletal muscle cells in vitro, as well as developing muscle cells in vivo, do express cardiac ␣ 1C mRNA (43,44). Altogether, our data provide new insights into the tissue-specific expression of L-type Ca 2ϩ channel isotypes and further suggest that the expression of functional skeletal L-type Ca 2ϩ channels can occur within the cardiac cell lineage through a de-differentiation process demonstrated in vitro in a cardiac cell line model. It is now important to search for the Ca 2ϩ channel isotypes, including ␣ 1S , that are expressed in early stages of cardiac development. Our study should also fosters in vivo investigations, such as in animal models that have been genetically manipulated for their RXR␣ pathway (31,34), to probe whether Ca 2ϩ channel expression is altered.
De-differentiation that occurs during cardiac pathologies might favor de novo expression of channel proteins, such as T-type Ca 2ϩ channels (45) and pacemaker channels (46). Preliminary experiments also indicate the presence of a skeletal L-type Ca 2ϩ channel in cardiomyocytes obtained from diseased human hearts, because expression of the ␣ 1S subunit can be identified at both transcript and protein levels (47). These results raise the question of whether expression of skeletal L-type Ca 2ϩ channels in heart is functionally relevant. Such a channel activity has not been described to date in hypertrophy models (48). However, functional expression of ␣ 1S -related Ca 2ϩ channels that act mainly as a voltage sensor is expected to be difficult to identify in cardiac myocytes, and combined patch-clamp, contraction, and Ca 2ϩ imaging experiments will certainly be necessary to resolve this issue. In addition, it will be now important to determine whether the RA treatment can improve alterations of Ca 2ϩ handling properties identified in cardiac pathologies, such as hypertrophy and congestive heart failure.