Differential Regulation of Exonic Regulatory Elements for Muscle-specific Alternative Splicing during Myogenesis and Cardiogenesis*

Muscle-specific isoform of the mitochondrial ATP synthase γ subunit (F1γ) was generated by alternative splicing, and exon 9 of the gene was found to be lacking particularly in skeletal muscle and heart tissue. Recently, we reported that alternative splicing of exon 9 was induced by low serum or acidic media in mouse myoblasts, and that this splicing required de novo protein synthesis of a negative regulatory factor (Ichida, M., Endo, H., Ikeda, U., Matsuda, C., Ueno, E., Shimada, K., and Kagawa, Y. (1998) J. Biol. Chem. 273, 8492–8501; Hayakawa, M., Endo, H., Hamamoto, T., and Kagawa, Y. (1998)Biochem. Biophys. Res. Commun. 251, 603–608). In the present report, we identified a cis-acting element on the muscle-specific alternatively spliced exon of F1γ gene by an in vivo splicing system using cultured cells and transgenic mice. We constructed a F1γ wild-type minigene, containing the full-length gene from exon 8 to exon 10, and two mutants; one mutant involved a pyrimidine-rich substitution on exon 9, whereas the other was a purine-rich substitution, abbreviated as F1γ Pu-del and F1γ Pu-rich mutants, respectively. Based on an in vivo splicing assay using low serum- or acid-stimulated splicing induction system in mouse myoblasts, Pu-del mutation inhibited exon inclusion, indicating that a Pu-del mutation would disrupt an exonic splicing enhancer. On the other hand, the Pu-rich mutation blocked muscle-specific exon exclusion following both inductions. Next, we produced transgenic mice bearing both mutant minigenes and analyzed their splicing patterns in tissues. Based on an analysis of F1γ Pu-del minigene transgenic mice, the purine nucleotide of this element was shown to be necessary for exon inclusion in non-muscle tissue. In contrast, analysis of F1γ Pu-rich minigene mice revealed that the F1γ Pu-rich mutant exon had been excluded from heart and skeletal muscle of these transgenic mice, despite the fact mutation of the exon inhibited muscle-specific exon exclusion in myotubes of early embryonic stage. These results suggested that the splicing regulatory mechanism underlying F1γ pre-mRNA differed between myotubes and myofibers during myogenesis and cardiogenesis.

Alternative pre-mRNA splicing is a fundamental process in eukaryotes that contributes to tissue-specific and developmen-tally regulated patterns of gene expression at the posttranscriptional level. Recently, both cis-acting elements and transacting factors have been reported to varying degrees (1)(2)(3). Significant progress has been made in identifying the alternative splicing mechanism involved in the Drosophila sex determination pathway (4 -7).
Exonic and intronic cis-acting regulatory elements for RNA splicing have been reported in a number of mammalian genes located near weak 5Ј and 3Ј splice sites, and these elements appear to be involved in the control of stage-or tissue-specific splicing events (5, 8 -13). For example, a majority of exonic splicing enhancer (ESE) 1 elements have been reported to be abundant in purine nucleotides (14,15), and these elements have been shown to bind the splicing factor for control of alternative splicing. Serine/arginine-rich (SR) proteins, a group of splicing factors, play important roles both in ESE-independent and ESE-dependent splicing (8,9,16,17). SR proteins show RNA binding activity to each cis-acting element, as well as protein-protein interaction activity in the formation of a commitment complex of spliceosome (16). SF2/ASF and SC35, two SR proteins, differ in their ability to commit different pre-mRNAs to the splicing pathway (18). SF2/ASF, but not SC35, recognized sequences very similar to purine-rich elements found in various natural splicing enhancers (9,19). SRp40, another member of the SR protein group, binds specific ESE, and one study showed that phosphorylation of the arginrine/serine-rich domain is necessary for sequence-specific binding (20). Research has shown that tissue-specific distribution of SR proteins varies to a certain degree and that overexpression of SR proteins in vivo alters the splicing pattern. Hence, the likelihood is that either tissue-specific or development stagespecific alternative splicing is controlled by changes in the amount of SR proteins or by covalent modification of these SR proteins, as in phosphorylation control (16).
Previously, we cloned the mitochondrial ATP synthase ␥-subunit (F 1 ␥) gene and reported that its tissue-specific isoforms were generated by alternative splicing in the human, cow, and mouse (21)(22)(23). RNA transcripts excluding exon 9 are specifically expressed in heart and skeletal muscle tissue (21,22). F 1 ␥ mRNA contains exon 9 in mouse myoblast C2C12 cells, but F 1 ␥ mRNA showed exclusion of exon 9 in myotubes (23). This muscle-type mRNA is cell-specifically and reversibly induced by acidic stimulation in human fibrosarcoma, human rhabdomyosarcoma, and mouse myoblast cells (23,24). Fur-thermore, this process of muscle-specific exon selection was inhibited by cycloheximide, a protein synthesis inhibitor, and the protein kinase C inhibitor H-7 in these cell lines (23,24). Other muscle-specific alternative splicing in the neural celladhesion molecules, ␤-tropomyosin, MEF2A, and MEF2D genes was induced by this acidic stimulation (23). The above muscle-specific alternative splicing likely requires de novo protein synthesis of a splicing regulatory factor. In fact, we found evidence using an in vitro splicing system that a negative regulatory factor for muscle-specific exon exclusion of human F 1 ␥ pre-mRNA existed in nuclear extracts from acid-stimulated human fibrosarcoma cells (25). However, we were unable to determine the exact regulatory mechanism underlying muscle-specific alternative splicing in muscle tissue.
In the present study, we identified the cis-acting regulatory element for alternative splicing of F 1 ␥ pre-mRNA and showed that the action of this element in myotubes and mature muscle fibers differed. Using in vivo splicing analysis of mouse F 1 ␥ minigenes, we determined that a cis-acting regulatory element on the alternatively spliced exon was important for both exon inclusion and exclusion in culture cells. Purine-rich mutation on the exon of F 1 ␥ minigene, which likely destroys the negative regulatory element, blocked muscle-specific exon exclusion in myotubes. In contrast, the purine-rich mutation of the F 1 ␥ minigene failed to block exon exclusion in skeletal and heart muscle tissues in transgenic mice. This difference between the splicing regulation that occurs during myogenesis and cardiogenesis will be discussed.

MATERIALS AND METHODS
Cell Culture-C2C12, mouse myoblast cells, were obtained from the American Type Culture Collection. L929, mouse fibrosarcoma cells, were provided by Dr. T. Kirikae (Jichi Medical School, Tochigi, Japan). The C2C12 cells and L929 cells were grown in DMEM (Dulbecco's modified Eagle's medium) containing 10% fetal bovine serum (Life Technologies, Inc.) as growth medium at 37°C under 5% CO 2 . The cells were plated on 10-cm diameter tissue culture dishes. Once the C2C12 cells had grown to semi-confluence, the growth medium was replaced with two different media. The differentiation medium (low serum medium) was DMEM with 2% heat-inactivated horse serum (Irvine Scientific), while the acidic medium was DMEM with 10% fetal bovine serum and 2.7 mM NaHCO 3 (final pH 6.6 under 5% CO 2 ). Once the L929 cells had grown to semiconfluence, the growth medium was replaced with acidic medium and cultured for 48 h as acidic stimulation. Primary culture of myoblasts prepared from skeletal muscle of embryonic day 19 transgenic mice was plated at 1 ϫ 10 6 /60-mm diameter tissue culture dish.
Mutations on the alternatively spliced exon of pcDEB SR␣ -F 1 ␥ wildtype minigene was created using two-piece PCR cloning strategy (13) using pBS-F 1 ␥ plasmid DNA. Mutant overlapping primer pairs (MU1 and MU2; MU3 and MU4) were generated for purine-rich and pyrimidine-rich mutants on exon 9, respectively. Primers M13 and M13RV for pBluescript II were used as the common outside oligonucleotides. MU1 with M13RV and MU2 with M13 were used to generate individual mutated ends of the clone of interest. The two PCR products were then separated on 1% agarose gel and eluted. A third round of PCR using M13 and M13RV with the two PCR products was used to generate the full-length mutated insert. Two full-length inserts were resubcloned into dideoxy T vector of pBluescript II SKϩ and confirmed by sequencing on both strands around the mutated regions. Finally, these F 1 ␥ mutants were subcloned into pcDEB SR␣ , as described above. Pyrimidine-rich F 1 ␥ mutant minigene and purine-rich F 1 ␥ mutant minigenes were termed F 1 ␥ Pu-del and F 1 ␥ Pu-rich minigenes, respectively. Sequences of the oligonucleotides were as follows: DNA Transfection to Cells-After C2C12 myoblasts and L929 fibrosarcoma cells had been cultured with growth medium to semiconfluence, the cells were transfected with 5 g each of pcDEB SR␣ -F 1 ␥ wildtype and two mutant minigenes using the polycationic liposome method (DMRIE-C, Life Technologies, Inc.). After transfection, the cells were cultured with growth medium, acidic medium, and differentiation medium for 3 days, and then harvested.
Generation of Transgenic Mice-Transgenic mice were established following the standard protocol described previously (27). To construct the expression plasmid for transgenic mouse, two F 1 ␥ mutant minigenes were subcloned into the pCAGGS expression vector (28) provided by Dr. J. Miyazaki (Osaka University, Osaka, Japan). pCAGGS-F 1 ␥ mutants were digested with SalI and BamHI. The DNA fragments were separated on 1% agarose gel, eluted, and extracted by phenol/chloroform/isoamylalcohol. The purified DNA fragment was dissolved to a final concentration of 500 copies/pl with 10 mM Tris-HCl (pH 7.5), 0.25 mM EDTA for microinjection. C57BL/6J mice were used for microinjection, and ICR mice were used as foster mothers.
We simultaneously co-injected green fluorescent protein (GFP) gene constructed in the same pCAGGS expression vector, pMG2 (29) (provided by Dr. K. Kohno, Nara Institute of Science and Technology, Nara, Japan). pMG2 were digested with SalI and BamHI. The purified GFP DNA fragments were mixed with the F 1 ␥ mutant minigene DNA fragments (ratio ϭ 1:1, the final concentration 500 copies/pl) subjected to microinjection.

RESULTS
In Vivo Alternative Splicing System of F 1 ␥ Minigene in C2C12 Myoblasts-To investigate the role of exon recognition for muscle-specific alternative splicing, we constructed a mouse F 1 ␥ minigene and developed an in vivo alternative splicing system by inducing differentiation in C2C12 mouse myoblasts (23). The F 1 ␥ wild-type minigene was constructed in pcDEB SR␣ mammalian expression vector (26) (pcDEB SR␣ -F 1 ␥ wild-type minigene), which contained a full-length mouse F 1 ␥ genomic gene corresponding to human F 1 ␥ gene from exon 8 to exon 10 (GenBank TM /EMBL/DDBJ Data Bank accession no. U43893) (Fig. 1a). Muscle-specific alternative splicing of endogenous F 1 ␥ pre-mRNA in C2C12 mouse myoblasts is induced in two different ways; one is replacement from growth medium to differentiation medium, and the other is replacement to acidic medium (23). After transfection of pcDEB SR␣ -F 1 ␥ wild-type minigene to C2C12 cells, culture medium was changed to growth, differentiation, and acidic media. Three days later, all cells were subjected to RT-PCR in order to detect splicing patterns of both endogenous and exogenous transcripts. As shown in Fig. 1 (b and c), muscle-specific exclusion of exon 9 in both endogenous F 1 ␥ pre-mRNA and exogenous pcDEB SR␣ -F 1 ␥ wild-type minigene pre-mRNA was observed after induction by differentiation medium or acidic medium. This minigene con-tains all genomic sequences from exon 8 to exon 10 of mouse F 1 ␥ gene, and the spliced transcripts of the minigene was independently regulated by both induction systems in a similar manner as endogenous induction. These results indicated that the F 1 ␥ wild-type minigene contained all cis-acting regulatory elements involved in muscle-specific alternative splicing.
cis-Acting Element Exists on the Alternatively Spliced Exon in Culture Cells-Based on a comparison of the genomic sequences around exon 9 of F 1 ␥ gene in the human, cow, and mouse, a relatively purine-rich sequence was found at 5Ј region of exon 9 in all three animal models (Fig. 2). Previously, exonic splicing enhancer containing purine-rich sequences was reported to play an important role in splice site selection (10). To determine whether this sequence is a cis-acting element involved in the regulation of alternative splicing of F 1 ␥, we constructed two types of mutant minigenes derived from the pc-DEB SR␣ -F 1 ␥ wild-type minigene (Fig. 2b). One type was a F 1 ␥ Pu-del mutant, made by introducing an 11-nucleotide stretch of poly(U) into the exon of wild-type minigene in place of the relatively purine-rich sequence at the 5Ј region of the exon. The other was a F 1 ␥ Pu-rich mutant, made by introducing a purinerich sequence, AAAGAAGAAGAA, into the middle region of the exon. This purine-rich sequence was derived from the 5Ј region of exon 10 of F 1 ␥ gene and resembled the relatively purine-rich sequence at the 5Ј region of exon 9 (Fig. 2b).
In vivo splicing assay using F 1 ␥ Pu-del mutant minigene transfected to C2C12 myoblasts revealed that no inclusion of the Pu-del mutant exon had occurred in the C2C12 cells cultured in growth medium, despite the fact that the wild-type exon of endogenous F 1 ␥ pre-mRNA was included (Fig. 3a, lane  4). In addition, Pu-del mutant exon was also excluded from C2C12 cells cultured with differentiation and acidic media (Fig.  3a, lanes 5 and 6). A similar experiment revealed that, as in the myoblasts, the Pu-del mutant exon was not included in mouse fibrosarcoma L929 cells (Fig. 3b, lanes 5 and 6). The transcripts of F 1 ␥ wild-type minigene in L929 cells included exon 9 in both growth and acidic media (Fig. 3b, lanes 1 and 2), as observed in endogenous F 1 ␥ transcripts in the cells (data not shown). These results indicated that the presence of purine nucleotides on the wild-type exon was very important in splice site selection of this alternatively spliced exon.
The other mutant minigene, F 1 ␥ Pu-rich minigene constructed in pcDEB SR␣ , was also tested. The alternative exon of F 1 ␥ Pu-rich minigene pre-mRNA was selected in C2C12 cells cultured not only in growth medium but also in differentiation and acidic media (Fig. 3a, lanes 1-3). In addition, the Pu-rich exon was primarily included in L929 cells cultured in growth and acidic media, whereas the wild-type exon of F 1 ␥ minigene was partially excluded from cells with acidic medium (Fig. 3b,  lanes 3 and 4). These data suggested that the purine-rich mutant sequence enhanced the inclusion of an alternative exon of F 1 ␥ pre-mRNA in myoblast, myotubes, and non-muscle cells.
In summary, the moderately purine-rich sequence at the 5Ј region of wild-type exon played a role in the selection of exon 9 as an exonic splicing enhancer, and the Pu-del mutant sequence disrupted this putative enhancer sequence and its activity. On the other hand, muscle-specific exon exclusion was inhibited in Pu-rich mutant minigene. Based on these results, we constructed two types of mutant minigenes; one was a Pu-del mutant minigene, the exon of which is constitutively excluded, while the other was a Pu-rich mutant minigene, the exon of which is included in every culture condition.
Gene Expression Pattern of CAG Promoter in Various Organs of Transgenic Mice Using GFP as a Useful Coinjection Marker-In order to determine whether this exonic cis-acting element was important for alternative splicing of F 1 ␥ pre-mRNA FIG. 1. Splicing patterns of F 1 ␥ minigene pre-mRNA in C2C12 myoblasts. a, mouse F 1 ␥ wild-type minigene in pcDEB SR␣ expression vector. Mouse F 1 ␥ genomic gene from exon 8 to exon 10 was ligated into pcDEB SR␣ mammalian expression vector. SR␣, SR␣ promoter; An. Sv, SV40 poly(A) site; Ptk, thymidine kinase promoter; An. Tk, thymidine kinase gene poly(A) site; HygroB R , hygromycin B-resistant gene; OriP, cis-acting element of Epstein-Barr virus genome that allows recombinant plasmids to replicate and to be maintained; Amp R , ampicillinresistant gene. b and c, splicing patterns of endogenous and exogenous F 1 ␥ transcripts analyzed by RT-PCR. C2C12 myoblasts were transiently transfected with pcDEB SR␣ including F 1 ␥ minigene and cultured in growth medium (G). After cells had grown to semiconfluence, medium was changed to differentiation medium (D) and acidic medium (A), and the cells were then cultured for 72 h. A 10-g aliquot of total RNAs from C2C12 myoblasts was subjected to RT-PCR using two pairs of primers for endogenous and exogenous F 1 ␥ transcripts as described under "Materials and Methods." The size of the PCR products was determined by 3% agarose gel electrophoresis, and the products were stained with ethidium bromide. even in various tissues in vivo, we produced transgenic mice carrying either a Pu-del or Pu-rich mutant minigene, both of which were re-subcloned into a pCAGGS mammalian expression vector (28) (Fig. 4a). Both of the minigenes in the pCAGGS vector which had transfected into myoblasts were confirmed to work well, in the same way as minigenes in the pcDEB SR␣ vector (data not shown). Three lines of transgenic mice containing F 1 ␥ Pu-del mutant minigenes were produced from the 42 eggs surviving after microinjection, while four lines of transgenic mice containing F 1 ␥ Pu-rich mutant minigene were produced from 77 eggs (Table I).
At the same time, we co-injected a modified green fluorescent protein (GFP) gene in the same expression vector, pMG2 (29), in order to allow for: 1) selection of transgenic mice using GFP and 2) detection of expression patterns of the promoter activity in various tissues of transgenic mice. GFP has been shown to be useful as a marker in transgenic mice given its lack of toxicity, high sensitivity, and reproducibility (31,32). We co-injected two types of DNA fragments, F 1 ␥ mutant minigene and GFP cDNA. The GFP used in this study was a double mutant (S65T/ S147P) that is strongly fluorescent and heat-stable (29). The majority of GFP-carrying F0 transgenic pups, which were detected by UV irradiation (wavelength 360 nm), carried both minigenes and expressed their mRNA (Table I). F1 transgenic mice were obtained from all F0 mice expressing GFP and F 1 ␥ mutant minigenes, using simple selection of transgenic pups by GFP (Fig. 4b).
pCAGGS mammalian expression vector has a cytomegalovirus enhancer and chicken ␤-actin promoter (28). The fluorescent signals of GFP expressed in various tissues of transgenic mice were observed using a CCD camera under 488 nm excitation light (fluorescent microscopy MZ FL III, Leica). As shown in Fig. 4c, skeletal muscle, heart, liver, and brain were visualized as strongly green under blue light. On the other hand, very weak fluorescence was observed in the spleen.
These results indicated that the chicken ␤-actin promoter showed activity in various tissues of transgenic mice. Thus, we demonstrated that GFP is a useful marker for selection of living transgenic mice and for simple detection of promoter activity.
Splicing Patterns of Pu-del or Pu-rich Mutant Minigenes in FIG. 3. Splicing patterns of F 1 ␥ mutant minigenes in C2C12 myoblasts and L929 cells. Mouse F 1 ␥ mutant minigene was ligated into pcDEB SR␣ mammalian expression vector. C2C12 myoblasts and L929 cells were transiently transfected with pcDEB SR␣ including F 1 ␥ mutant minigene and were cultured in growth medium (G). After cells had grown to semiconfluence, medium was changed to differentiation medium (D) and acidic medium (A), and the cells were then cultured for 72 h. A 10-g aliquot of total RNAs from these cells was used for RT-PCR. Primers for PCR were described in the legend for Fig. 1. The size of the PCR products was determined by 3% agarose gel electrophoresis, and the products were stained with ethidium bromide. a, splicing pattern in C2C12 myoblasts; b, splicing pattern in L929 cells. wt represents F 1 ␥ wild-type minigene. Various Organs of Transgenic Mice-As the promoter of pCAGGS was shown to be expressed in various tissues, we investigated the splicing pattern of F 1 ␥ mutant minigenes in the organs of transgenic mice. Total RNA was prepared from several organs of 4-week-old transgenic mice and then subjected to RT-PCR analysis (Fig. 5). Splicing patterns of endogenous mouse F 1 ␥ pre-mRNA in various tissues revealed that exclusion of exon 9 was specific to heart and skeletal muscle tissues (Fig. 5a). The alternative exon of F 1 ␥ Pu-del pre-mRNA was excluded not only in skeletal and heart muscle tissues, but also in non-muscle organs such as the liver, brain, spleen, and kidney (Fig. 5b). These data indicated that purine nucleotides at the 5Ј region of wild-type exon were necessary for exon inclusion to occur in cultured myoblast cells and non-muscle organs, suggesting that this region is an exonic splicing enhancer in non-muscle organs as same as in culture cells. pMG2, GFP with double mutants (S65T/S147P) constructed in the same vector, was provided by Dr. Kohno (28). pCAGGS-F 1 ␥ mutants and pMG2 were digested with SalI and BamHI. Equivalent amounts of two different purified DNA fragments, F 1 ␥ mutant minigene and GFP cDNA, were co-injected to generate transgenic mice. b, ease of selection of transgenic mice using GFP. F1 transgenic pups expressing GFP and F 1 ␥ mutant minigene were detected under UV irradiation (wavelength 360 nm) (right panel). GFP and F 1 ␥ mutant minigenes were not separated in every pup, and thus these two transgenes were likely located on the same chromosome. c, gene expression pattern of CAG promoter in various tissues of transgenic mice. The fluorescent signals of GFP expressed in various tissues of transgenic mice and wild mouse were observed using a CCD camera under 488 nm excitation light (fluorescent microscopy; MZ FL III, Leica). Each organ on the right was obtained from transgenic mice, whereas those on the left were from control mice.
Next, we analyzed the splicing patterns of Pu-rich mutant minigene transcripts in various tissues. As shown in Fig. 5c, interestingly, the Pu-rich mutant exon of exogenous F 1 ␥ minigene was excluded in skeletal and heart muscle tissues, although this exon was included even in myotubes (Fig. 3a). The difference between the splicing patterns in cultured myotubes and those in mature muscle fibers suggested that differential regulation is involved in the muscle-specific alternative splicing for each stage of muscle development.
Different Alternative Splicing Patterns of F 1 ␥ Pu-rich Pre-mRNA Were Observed during Myogenesis and Cardiogenesis in Uteri-To confirm the possibility of differential regulation of alternative splicing in F 1 ␥ pre-mRNA during myogenesis and cardiogenesis, we analyzed the splicing pattern of F 1 ␥ Pu-rich and endogenous F 1 ␥ pre-mRNAs in the heart and forelimbs of transgenic mice at each embryonic day. Heart and skeletal muscle tissues of 4-week-old transgenic mice were used as controls. Fetal hearts from embryonic day (E) 9 -15 were prepared under a stereoscopic microscope. The E9 heart had already begun to beat at that point. As shown in Fig. 6a, the endogenous exon 9 of F 1 ␥ pre-mRNA was primarily excluded in heart from E9 to adult. In contrast, the Pu-rich mutant exon was shown to be partially excluded in fetal hearts; the ratio was approximately 1:1. This Pu-rich mutant exon was mostly excluded in adult heart, and thus a differential regulation of muscle-specific splicing likely exists between fetal and adult heart muscles.
Forelimbs were prepared from E10 to E15 transgenic mice containing both muscle and non-muscle tissues. In general, myoblasts appeared in forelimb from E10 to E11 and myotubes from E11 to E12 (33,34). RT-PCR analysis revealed that E10 forelimb expressed only the non-muscle type of both endogenous and exogenous F 1 ␥ pre-mRNAs (Fig. 6b). Exclusion of the endogenous wild-type exon was detectable in E11 and E12 forelimbs, whereas no exclusion of Pu-rich mutant exon was detected in the same samples. Exogenous Pu-rich mutant exon of F 1 ␥ minigene was shown to be excluded from E13 forelimbs. At the early stage of myotube formation in forelimbs, exon exclusion was inhibited in Pu-rich mutant minigene transcripts. These data indicated that muscle-specific alternative splicing of F 1 ␥ pre-mRNA was differentially regulated during myogenesis and cardiogenesis.
Primary Cultures of Myoblasts from Transgenic Mice Showed Inclusion of Pu-rich Mutant Exon-To establish that a difference exists between the alternative splicing regulations of F 1 ␥ pre-mRNA in myotubes and that in mature muscle fibers, we analyzed primary cultures of myoblasts and myotubes from skeletal muscle of transgenic mice using a F 1 ␥ Pu-rich mutant minigene. Primary cultures were contaminated with a number of different fibroblasts, although myotube formation was observed after treatment with differentiation medium (Fig. 7a). From RT-PCR analysis of primary culture cells, muscle-specific exon exclusion of endogenous F 1 ␥ pre-mRNA was observed after treatment with acidic or differentiation medium (Fig. 7b,  lanes 5 and 6). In contrast, no exclusion of Pu-rich mutant exon was observed after treatment with either medium (Fig. 7b,  lanes 2 and 3). In these cells, the non-muscle type of F 1 ␥ FIG. 5. Splicing patterns of F 1 ␥ mutant minigenes in transgenic mice. A 10-g aliquot of total RNAs from tissues of transgenic mice was subjected to RT-PCR. The two pairs of PCR primers for endogenous and exogenous transcripts were described in the legend for Fig. 1. The size of the PCR products was determined by 3% agarose gel electrophoresis. a, tissue distribution of splicing pattern of endogenous F 1 ␥ mRNA. b, splicing pattern in transgenic mice with F 1 ␥ Pu-del mutant minigene. c, splicing pattern in transgenic mice with F 1 ␥ Pu-rich mutant minigene. H, heart; M, skeletal muscle; L, liver; B, brain; S, spleen; K, kidney; C1, muscle-type F 1 ␥ cDNA as a control; C2, nonmuscle-type F 1 ␥ cDNA as another control. splicing pattern was predominant, so that an abundance of non-myoblast cells, such as fibroblasts, was found in the primary cultures. Although they were very faint, bands of muscletype endogenous F 1 ␥ transcripts were detected after induction (Fig. 7b, lanes 5 and 6). These results from primary cultures of myoblasts and myotubes supported the finding that musclespecific alternative splicing of F 1 ␥ pre-mRNA was differentially regulated between myotubes and mature muscle tissues.

DISCUSSION
The present study demonstrated that the cis-acting elements play an essential role in exonic alternative splicing of F 1 ␥ pre-mRNA both during myogenesis and cardiogenesis. First, in vivo splicing analyses in mouse myoblasts and in a transgenic mouse system using F 1 ␥ mutant minigenes revealed that purine nucleotides at the 5Ј region of the alternatively spliced exon were important for exon inclusion in non-muscle cells and tissues. Second, a Pu-rich mutation at the middle region of the exon inhibited muscle-specific exon exclusion in myotubes, but did not inhibit in heart and skeletal muscle tissues of transgenic mice. This discrepancy in the splicing patterns of the mutant exon between myotubes and mature muscle fibers suggests that there is a difference of the alternative splicing of F 1 ␥ in each step of muscle development.
The exonic cis-acting positive regulatory element of F 1 ␥, we demonstrated here, located on the 5Ј region of the alternatively spliced exon. In general, splicing of eukaryotic pre-mRNAs requires the consensus sequences in 5Ј and 3Ј splice sites, branching point, and polypyrimidine tract. In cases where the polypyrimidine tract is weak, splicing enhancer sequences are needed, such as ESEs (9,10). ESE-dependent splicing reaction requires SR proteins, for example SF2/ASF and SC35, which are needed at an early stage of spliceosome assembly. These SR proteins are able to bind to each specific RNA cis-element (9,19), and these elements have been primarily reported to be purine nucleotide-rich sequences. An exon inclusion activity of the enhancer element of F 1 ␥ was clearly proved by in vivo splicing assay using a Pu-del mutant minigene in myoblasts and in non-muscle tissues of transgenic mice (Figs. 3 and 5b). The enhancer element was rich in purine nucleotide and was not identical to well known ESE sequences. This exonic enhancer sequence would require a sequence-specific splicing factor such as SR proteins. Previously, we reported on a reversible induction system of muscle-specific alternative splicing of F 1 ␥ using human fibrosarcoma and mouse myoblast cells and revealed that the non-muscle type of exon inclusion is a default type and that the muscle-specific exon exclusion required de novo protein synthesis of an intracellular protein factor (23,24). Considering above, it is likely that this exonic enhancer element is necessary for constitutive splicing of F 1 ␥ in nonmuscle tissues, and that a splicing factor for this enhancer element, such as SR proteins, is also required (Fig. 8, left  column).
The negative regulation for selection of exon 9 in the F 1 ␥ gene was induced in low serum-or acid-stimulated cells (23,24). We also established that the nuclear extracts from postinduced cells contained a sequence-specific negative regulatory factor of exon selection by in vitro splicing assay (25). Based on these findings, a negative regulatory factor was induced, which in turn inhibited the constitutive splicing in F 1 ␥-wt minigene in both myotubes and post-induced cells (Fig. 1c). When the Pu-rich mutation was introduced at the middle region of the wild-type exon 9, induction of exon exclusion by acidic stimulation or differentiation medium was blocked (Fig. 3). From the result, this mutation likely affected the negative regulatory element on the exon. Alternatively, the Pu-rich mutation would make a new exonic splicing enhancer sequence. However, the Pu-rich mutant sequence resembles, but is not identical to, SF2/ASF-binding consensus element (41). In addition, this mutant sequence did not inactivate the enhancer element on the alternatively spliced exon, but inhibit the silencer activity in myotubes. Considering this, it is likely that the negative regulatory element located on the middle region of the exon (Fig. 8,  middle column).
This Pu-rich mutation on the exon did not work in skeletal and heart muscle tissues of the transgenic mice (Fig. 5c). However, this mutation was observed to inhibit the exon exclusion not only in primary cultures of myoblasts from the transgenic mice ( Fig. 7) but also in premature heart and skeletal muscle FIG. 7. Splicing patterns of F 1 ␥ Purich minigene in primary culture of myoblasts from transgenic mice. a, primary culture of myoblasts prepared from transgenic mice. Muscle tissues from limbs of 2-day-old F 1 ␥ Pu-rich transgenic mice were used for primary culture. Cells were cultured in differentiation medium for 2 days. Myotubes appeared in the upper panel. These myotubes showed stronger fluorescence of GFP than other cells, which was observed using a CCD camera under 488 nm (PXL1400, Photometrics) (lower panel). b, splicing patterns of F 1 ␥ Pu-rich minigene in primary culture of myoblasts from transgenic mice. Primary cultures of myoblasts were grown under the three types of medium: growth medium (G), differentiation medium (D), and acidic medium (A). The cells were cultured for 72 h. Total RNAs from the cells were used for RT-PCR analysis. The size of the PCR products was determined by 3% agarose gel electrophoresis. Primer sequences were described in the legend of Fig. 6c. during fetal development in uteri (Fig. 6). These data indicated that there was a differential regulation of muscle-specific alternative splicing between myotubes and mature muscle fibers. When the Pu-rich mutation made the exonic silencer element inactivated, there were two possibilities to explain the difference of the splicing regulation between myotubes and mature muscle fibers. First, a negative regulatory factor that acts to the negative regulatory element would be different quantitatively or qualitatively between myotubes and myofibers. Second, a positive regulatory factor for the enhancer element would differ quantitatively or qualitatively. In fact, the amount of SR protein has been reported to differ among various tissues (38,39). For example, tissue distribution of rat SF2/ASF and hnRNPA1 proteins have been reported (40). The above study found that the rat heart muscle contained only trace amounts of SF2/ASF. It is then likely that the protein level of a constitutive splicing factor is only decreased in mature muscle fibers, which in turn results in exon exclusion, even when the silencer element is inactivated (Fig. 8, right column).
There are many muscle-specific alternative splicings in several genes, e.g. ␣or ␤-tropomyosin, neural cell-adhesion molecule, myocyte-specific enhancer proteins, and F 1 ␥. Although the development stage of myotubes is earlier rather than that of myofibers, most of muscle-specific alternative splicings were reported to appear at myotubes (23). Our Pu-rich mutant minigene succeeded in clearly demonstrating the difference in splicing regulation between myotubes and mature muscle fibers.
We have identified the splicing enhancer sequence of alternative splicing of F 1 ␥ pre-mRNA in vivo. This enhancer element is rich in purine nucleotides, exists on the alternatively spliced exon, and works to include this exon in myoblasts and non-muscle tissues. Muscle-type F 1 ␥ mRNA was created by the induction of a negative regulatory factor (25), whereas a differential regulation of muscle-specific exon exclusion was present between myotubes and mature muscle fibers. To elucidate the exact mechanism underlying alternative splicing during myogenesis, muscle-specific splicing regulatory factors must be identified. FIG. 8. Model for muscle-specific splicing in ATP synthase ␥-subunit pre-mRNA in myoblasts, in myotubes, and in mature muscle fibers. The exonic cis-acting positive regulatory element (black box) located on the 5Ј region of exon 9. This enhancer element is necessary for constitutive splicing in nonmuscle tissues and myoblasts, and a constitutive splicing factor is also required. The Pu-del mutation (gray box) at the 5Ј region affected exon inclusion in any tissue and cell (left column). A negative regulatory factor was induced and then inhibited the constitutive splicing in myotubes. The Pu-rich mutation (horizontal box) at the middle region of exon 9 inhibited induction of exon exclusion (middle column). The Pu-rich mutation (horizontal box) did not work in skeletal and heart muscle tissues. In this model, the protein level of a constitutive splicing factor is decreased in mature muscle fibers, which in turn results in exon exclusion, even when the silencer element is inactivated (right column). See text details.