Splicing of 5′ Introns Dictates Alternative Splice Selection of Acetylcholinesterase Pre-mRNA and Specific Expression during Myogenesis*

Splicing of alternative exon 6 to invariant exons 2, 3, and 4 in acetylcholinesterase (AChE) pre-mRNA results in expression of the prevailing enzyme species in the nervous system and at the neuromuscular junction of skeletal muscle. The structural determinants controlling splice selection are examined in differentiating C2-C12 muscle cells by selective intron deletion from and site-directed mutagenesis in the Ache gene. Transfection of a plasmid lacking two invariant introns (introns II and III) within the open reading frame of the Ache gene, located 5′ of the alternative splice region, resulted in alternatively spliced mRNAs encoding enzyme forms not found endogenously in myotubes. Retention of either intron II or III is sufficient to control the tissue-specific pre-mRNA splicing pattern prevalent in situ. Further deletions and branch point mutations revealed that upstream splicing, but not the secondary structure of AChE pre-mRNA, is the determining factor in the splice selection. In addition, deletion of the alternative intron between the splice donor site and alternative acceptor sites resulted in aberrant upstream splicing. Thus, selective splicing of AChE pre-mRNA during myogenesis occurs in an ordered recognition sequence in which the alternative intron influences the fidelity of correct upstream splicing, which, in turn, determines the downstream splice selection of alternative exons.

Formation of a contiguous open reading frame in eukaryotic mRNA by elimination of intervening sequences (introns) through splicing is a critical step in biosynthesis of functional proteins (1)(2)(3). Alternative splicing of pre-mRNA in various cell types generates protein isoforms with distinct activities or tissue-specific distribution patterns during development (4,5). About 15% of mammalian gene mutations associated with disease states are reported to affect RNA splicing signals (2). However, the molecular mechanisms underlying regulation of pre-mRNA splicing, especially in mammalian cells, are not well understood (1)(2)(3). One such example is the selective splicing of acetylcholinesterase (AChE) 1 pre-mRNA during myogenesis.
The primary function of AChE is to terminate the action of the released neurotransmitter acetylcholine in the central and peripheral nervous systems. AChE exists in multiple molecular forms whose catalytic subunits are encoded by alternatively spliced mRNAs from a single gene (6 -9). The alternative gene products differ in sequence only at their very carboxyl termini, resulting in enzyme forms that differ in amphiphilic character, extent of oligomerization, and cellular disposition, but exhibit identical catalytic parameters. Within the open reading frame of the mammalian Ache gene are found three invariant exons (exons 2-4) that encode the amino-terminal 536 amino acids common to all forms of the enzyme. These invariant exons are alternatively spliced to one of three downstream sequences encoding distinct carboxyl termini and generating tissue-specific molecular forms of AChE (10). Exon 4 to 6 spliced mRNA, encoding catalytic subunits that assemble as monomers, dimers and tetramers, is the major species found in mammalian skeletal muscle and brain (10,11). The second splice option is extension of exon 4 to its 3Ј intron, yielding a monomeric, hydrophilic species found in embryonic skeletal muscle and cells of hematopoietic origin (6,10,12). The third splice option is the exon 4 to 5 splice that encodes an amphiphilic, glycophospholipid-linked form of AChE typically expressed in hematopoietic cells as well as pituitary and certain neuronal cells in culture (10).
Splicing of pre-mRNA involves the binding of small nuclear ribonucleoprotein particles (snRNPs) to conserved intronic sequences on pre-mRNA. The binding of U1 snRNP and U2AF to the 5Ј splice site and the polypyrimidine tract, respectively, promotes the binding of U2 snRNP to the branch point. The next step is the formation of spliceosome complex through the binding of the U4/U5/U6 snRNPs followed by a catalytic reaction resulting in a lariat formation between the 5Ј splice site and the adenosine residue at the branch point. This, in turn, activates the cleavage of 3Ј splice site and brings the spliced exons together (3,13,14). Thus, intronic sequences play an important role in molecular interactions during constitutive splicing.
One mechanism for the control of alternative splicing involves interaction of tissue or cell-specific factors with specific cis-elements in pre-mRNA (14). This interaction may influence the relative strength and, consequently, recognition of a given splice site as demonstrated in regulated splicing in vitro (13,14). A large body of evidence indicates that flanking intronic sequences are important in such an interaction (1,15). One such example is the splicing of rat ␤-tropomyosin in which downstream intron splicing events are critical in upstream alternative splicing regulation (16). Since a single gene encodes all tissue-specific forms of AChE, tissue-specific factors interacting with specific cis-elements can be expected to regulate alternative splicing.
Within the Ache gene, three invariant introns are found upstream of the alternative splice region where three splice options occur from a donor site. Two of the upstream invariant introns are located within the open reading frame of the gene (Fig. 1). As the first step toward understanding the alternative splicing of AChE pre-mRNA, we examined the role of these introns in the regulation of AChE pre-mRNA splicing during myogenesis. Fortuitously, the mammalian Ache gene is relatively compact with only 6 kilobase pairs separating the CAP site for transcription and the first polyadenylation signal. This enabled us to transfect the entire Ache genomic DNA with its accompanying promoter as well as specified deletion constructs directly into differentiating myocytes.

EXPERIMENTAL PROCEDURES
Materials-Cell culture supplies were from Fisher, and components of culture medium were from Life Technologies, Inc. [ 32 P]UTP (specific activity: 800 Ci/mmol) was purchased from NEN Research Products (Wilmington, DE). Sucrose (ultrapure) was from ICN Biomedicals (Aurora, OH). Other chemicals were from Sigma.
Tissue Cultures-Mouse myoblast C2-C12 cells and human embryonic kidney cells (HEK-293) were from American Type Culture Collection. C2-C12 cells were cultured at 37°C, with 5% CO 2 in Dulbecco's modified Eagle's medium (DMEM) containing 20% fetal bovine serum, 0.5% chick embryo extract and 1% penicillin, streptomycin, and amphotericin B stock solution (Antibiotic-Antimycotic, Life Technologies, Inc.). Cells were passed either two or three times before plating onto 100-mm culture dishes. Differentiation from myoblasts to myotubes was induced at about 70% confluence by replacing the high serum medium with DMEM containing 2% horse serum. HEK cells were maintained under the same conditions as the C2-C12 cells, except that the culture medium contained 10% fetal bovine serum.
Transient Transfections-Constructions of Ache genomic expression plasmids with the endogenous promoter and their branch point mutants are described in the legends to Figs. 1 and 3. cDNA expression constructs under CMV promoters were described previously (10). Proliferating C2-C12 myoblasts transfect poorly, so to optimize transfection efficiencies, cells were transfected 1 day after induction of differentiation. Transfection of 10 g/100-mm plate of the designated expression plasmids was performed by standard calcium phosphate procedures followed by glycerol shock 4 -5 h later (17). In some experiments, LipofectAMINE (Life Technologies, Inc.)-mediated transfection with 5 g of DNA/plate was utilized. Control cells were mock-transfected with the plasmid vectors without the Ache inserts. Cells were allowed to differentiate for 2-3 additional days in normal differentiation medium. To measure AChE secreted into the medium, cells were differentiated in serum-free DMEM after transfection. To correct differences in transfection efficiency, 2-5 g of a pcDNA3 expression vector containing the Escherichia coli lacZ gene (18) were co-transfected, and ␤-galactosidase activity was assayed spectrophotometrically. HEK cells were transfected by standard calcium phosphate method without glycerol shock.
RNA Extraction and RNase Protection Assays-Total RNA was extracted from previously transfected myotubes as described previously (19) or with TRIZol reagent (Life Technologies, Inc.), treated with 10 units of RQ1 DNase (Promega) for 30 min at 37°C, and stored at Ϫ20°C. AChE mRNA species were quantified by RNase protection assays as described (11). For detecting exon 4 to 6 spliced mRNA, a mouse Ache cDNA containing the sequence of exons 4 and 6 was subcloned in a Bluescript SK II plasmid and linearized with XhoI. After in vitro transcription with [ 32 P]UTP, a 458-bp labeled antisense cRNA probe was used for RNase protection. Similarly, for detecting exon 4 and its retained 3Ј intron, a XhoI to ApaI fragment of Ache genomic DNA was subcloned into Bluescript SK II and linearized with XhoI. Upon transcription it gives rise to a 464-bp antisense probe. For detecting exon 4 to 5 spliced species, an antisense probe was transcribed from a MscI linearized mouse AChE cDNA containing the exon 4 to 5 spliced sequence in an expression vector (pRc/CMV, Invitrogen, San Diego, CA). To normalize for transfection efficiencies, an antisense probe of ␤-galactosidase was transcribed from a BamHI to SstI fragment of the E. coli lacZ gene, subcloned into Bluescript SK II, and linearized with Eco47III. A tRNA lane was included in each RNase protection assay to ensure the complete digestion of the free probes. Molecular masses of the protected probes were estimated by electrophoresis on polyacrylamide gels, and protected bands were exposed to BioMax films (Kodak) and quantified by densitometry (UltroScan XL, Amersham Pharmacia Biotech).
Determination of AChE Activity-AChE was extracted from rinsed C2-C12 myotubes in 0.01 M sodium phosphate buffer (pH 7.0) containing 1 M NaCl, 0.01 M EGTA, 1% Triton X-100, and a spectrum of protease inhibitors (20). Culture media were centrifuged at 1000 ϫ g for 30 min to remove cell debris. AChE in the media was concentrated over 10-fold in Centriprep 30 concentrators (Amicon, Inc., Beverly, MA). Enzyme activity was determined at room temperature in 0.1 M sodium phosphate buffer (pH 7.0) using 0.05-0.1 ml of cell extract (21). Exogenous AChE activity was estimated by subtracting the activity in mocktransfected cells from the total activity in transfected cells. Differences in transfection efficiency were corrected from the ratios of AChE to ␤-galactosidase activities.
Sedimentation Analysis-AChE species were distinguished on the basis of their sedimentation coefficients in 5-30% sucrose gradients as described previously (22,23). Briefly, cell extracts or concentrated media were layered onto sucrose gradients containing 0.01 M sodium phosphate buffer (pH 7.0), 1 M NaCl, 0.01 M EGTA, and 1% Triton X-100 or Brij-96 (v/v). Carbonic anhydrase (3.3 S), alkaline phosphatase (6.1 S), and ␤-galactosidase (16 S) were added to the extracts as sedimentation markers. A 0.2-ml cushion of 50% sucrose was layered at the bottom of the gradients. The gradients were centrifuged at 41,000 rpm for 24 h at 4°C in a SW41 Ti rotor (Beckman Instruments, Palo Alto, CA) and upward fractionated. Aliquots of each fraction were assayed for AChE and marker activities using a 96-well microtiter plate reader.

Invariant Introns II and III Dictate mRNA Alternative Splicing Preferences-To examine the role of invariant introns of the
Ache gene in pre-mRNA splice selection, we constructed from genomic DNA five mouse AChE expression plasmids containing the endogenous promoter (Fig. 1). The genomic DNA construct (gDNA), upon transfection into C2-C12 cells, should express a pattern of AChE mRNA species similar to the endogenous AChE transcripts. A second construct lacks both invariant introns II and III within the open reading frame of the Ache gene (⌬i2-3,3-4), but contains the entire alternative splicing region 3Ј of exon 4. The other three expression constructs contain the gDNA with one of the three 5Ј introns in the gene deleted.
As shown in Fig additional AChE transcripts correctly. However, when the same plasmid DNA was transfected into human embryonic kidney cells, only small fraction of the AChE transcripts was spliced from exon 4 to 6, indicating that selective exon 4 to 6 splicing may be intrinsic to certain differentiated cells, such as C2-C12 muscle cells. When the transfected cells were maintained in an early stage of differentiation by returning to high serum conditions immediately after transfection, AChE mRNA transcribed from the transfected gDNA was only 37% (average of two independent transfections) of the level found in fully differentiated myotubes ( Fig. 2A). This suggests that, similar to the endogenous AChE mRNA (19,24), mRNA from the transfected gene underwent stabilization during myogenesis.
By contrast, transient transfection of ⌬i2-3,3-4 resulted in additional protected bands corresponding to the individual sizes of exon 4 (E4) and 6 (E6) in the probe. This severing of the exon 4 to 6 linkage indicates that, in addition to exon 4 to 6 splice, exon 4 with retained 3Ј intron or/and exon 4 to 5 spliced species are expressed. These alternatively spliced species are confirmed by two probes shown in Fig. 2 (B and C). The fulllength protected probes indicate the presence of mRNA with exon 4 either linked to its 3Ј intron (E4-RI in Fig. 2B) or spliced to exon 5 (E4 -5 in Fig. 2C), respectively. These alternatively spliced species were also confirmed by the protected exon 5 sequences (E5) in these probes reflecting exon 4 to 5 splice ( Fig.  2B) or exon 4 linked to its 3Ј intron (Fig. 2C). Protected exon 4 sequences (E4) in these probes appearing in the absence of protected E5 demonstrates the presence of only the exon 4 to 6 splice ( Fig. 2, B and C). These data show that removal of introns II and III results in expression of alternatively spliced AChE mRNA species, which are normally not seen in myotubes.

Retention of Either Intron II or III Is Sufficient to Control
Exon 4 to 6 Splicing-To examine the influence of individual intron II or III on pre-mRNA splicing, AChE expression constructs devoid of either intron II (⌬i2-3) or III (⌬i3-4) were transfected into differentiating C2-C12 cells. Total AChE mRNA levels accumulating in differentiated myotubes were lower than that in myotubes transfected with either gDNA or ⌬i2-3,3-4 (Table I). However, the majority of the transcripts were spliced between exon 4 and 6, as seen for the endogenous mRNA (Fig. 2, B and C, and Table I). Thus, the presence of FIG. 3. Intron IV and partial intron III deletion constructs and branch point mutations in shortened intron III. The exon 4 to 5 (⌬i4 -5) or exon 4 to 6 (⌬i4 -6) spliced plasmids were constructed by swapping the genomic DNA fragments for mouse cDNA fragments in which exon 4 was spliced either to exon 5 or to exon 6. The P3-4⌬i2-3 plasmid was constructed by taking a 58-bp intronic fragment from the 5Ј end of intron III and a 30-bp intronic fragment from the 3Ј end of intron III. The 5Ј and 3Ј splice site and branch point sequences are kept intact in this construct. Two branch point mutants made from this construct are indicated. The uppercase letters represent the intronic sequence, while the lowercase letters represent the exonic sequences. Intronic sequences derived from the 5Ј end of intron III are indicated in the open boxes. Consensus sequences around the branch point are underlined. The branch point adenosine (A) in construct SA/C was mutated to cytosine (C) by site-directed mutagenesis. In the QUA/C construct, 4 adenosine residues were mutated to cytosines. All mutations were confirmed by sequencing. The consensus sequences around 5Ј and 3Ј splice sites and the branch point in vertebrate pre-mRNAs are shown below the constructs (44). N, any nucleotide; Pyr, pyrimidine. These deletion or mutation constructs containing the AChE promoters were ligated into pBluescript II SK plasmids.

TABLE I
Enhanced mRNA expression in C2-C12 myotubes after transfection of the respective plasmids into myoblasts Genomic DNA encoding AChE and its respective deletion constructs are described in Fig. 1. Protected antisense mRNA probes were analyzed by densitometric analysis. After subtracting the density of the endogenous AChE bands quantified from mock transfected myotubes, band densities are normalized to the densities of protected ␤-galactosidase probes. Uracil content in each protected probe of designated length was used to normalize for the differences in sizes of protected bands. Total mRNA levels were determined by averaging the sum of AChE mRNA levels detected by three probes. The levels of different mRNA species were determined by protected bands from the respective probes and shown as percentage of the total AChE mRNA detected by the same probe. Values reported are the means Ϯ S.E. averaged from four to six independent transfections.
either intron II or III is sufficient to direct exon 4 to 6 splicing during myogenesis and yield a pattern approaching that of endogenous pre-mRNA splicing. Cumulative data on the pattern of spliced AChE mRNA species in transfected C2-C12 myotubes are summarized in Table I. Intron I Is Necessary for AChE Expression-To examine the influence of the first intron in the Ache gene on its expression and splicing, differentiating C2-C12 cells were transiently transfected with an Ache expression construct devoid of this intron (⌬i1-2). However, differentiated myotubes did not express the transfected plasmid at levels above the endogenous AChE mRNA ( Fig. 2A, representative data from three independent transfection experiments), indicating that the first intron residing 5Ј of the ATG start site is necessary for Ache gene transcription or RNA stabilization.
Splicing Events That Flank the Alternative Splice Region Occur in an Orderly Fashion, and Intron IV Is Critical in Controlling Upstream Splicing-The following experiments were directed to resolving the sequence of splicing events occurring around the alternative splice region of AChE pre-mRNA. Helfman et al. (16) reported that downstream splicing in rat ␤-tropomyosin pre-mRNA is critical to upstream splice site selection. To examine the order of intron splicing around the alternative splice region and the influence of downstream intron on upstream splicing, we used the probe shown in Fig. 4 to detect unspliced and spliced transcripts, and the presence of splicing intermediates in cells transfected with gDNA or expression constructs in which the upstream invariant exons were spliced to either exon 5 (⌬i4 -5) or exon 6 (⌬i4 -6) (Fig. 3). Transfection of gDNA resulted in expression of a 540-bp species representing unspliced pre-mRNA, a 480-bp splice intermediate devoid of intron III, and a 171-bp species representing exon 4 to 6 splice (Fig. 4, A and B; note that retention of intron IV and exon 4 to 5 splicing should yield a protected exon 5 species). We did not see a protected species that would indicate production of splice intermediates with only intron III attached after gDNA transfection, suggesting that, at steady state, splicing of intron III precedes splicing of intron IV. However, transient transfection of ⌬i4 -5 and ⌬i4 -6 resulted in expression of a 231-bp species representing an unspliced transcript or an aberrant mRNA species spliced from a cryptic splice site 5Ј of exon 4 (Fig. 4B). These data indicate that intron IV is critical in controlling the efficiency of correct upstream splicing and further suggest that splicing of the invariant intron occurs prior to splicing of intron IV, the site of alternative splicing.
Elimination of Upstream Splicing, but Not Changes in pre-mRNA Structure, Interrupts Downstream Splice Site Selection-The influence of upstream introns on downstream splice selection could result from a secondary structure in the pre-mRNA or be linked to upstream splicing events. To distinguish these possibilities, we deleted a large part of intron III from the construct ⌬i2-3, but kept the splicing machinery including the splice junctions and the branch point intact (P3-4⌬i2-3 in Fig. 3). Expression of this deletion construct should result in more dramatic changes in pre-mRNA structure, but not upstream splicing, compared with the expression of ⌬i2-3.
If the secondary structure of pre-mRNA plays a major role on downstream splicing selection, expression should result in a splicing pattern similar to that of ⌬i2-3,3-4. By contrast, if splicing events play a predominate role in downstream splice selection, expression should result in a splicing pattern similar to that of ⌬i2-3, which is close to the splice pattern of the genomic construct. As indicated in Fig. 4A, transfection of P3-4⌬i2-3 resulted in a 513-bp unspliced species, shorter than the 540-bp unspliced species seen in gDNA-transfected cells as expected from deletion of the intronic sequences. The absence of a protected exon 5 species in P3-4⌬i2-3-transfected cells, similar to the gDNA, indicates that neither the intron IV retained species nor the exon 4 to 5 spliced mRNA is expressed. Therefore, in gDNA and P3-4⌬i2-3-transfected cells, the 480and 171-bp protected species represent a splicing intermediate devoid of intron III and exon 4 to 6 spliced mRNA, respectively. Accordingly, partial deletion of intron III in the ⌬i2-3 plasmid resulted in a splicing pattern similar to that of gDNA, rather than the splicing pattern seen for ⌬i2-3,3-4.
To confirm the linkage of upstream splicing to downstream splice selection, the branch point adenosine in P3-4⌬i2-3 was mutated to cytosine (single A 3 C mutation, SA/C). Since such a change in mammalian branch sites may activate abnormal branch formation through cryptic branch points (25)(26)(27), four adenosines proximal to the branch point in the intronic sequence were also mutated to cytosines (four adenosine mutations, QUA/C) (Fig. 3). If upstream splicing is linked to downstream splice selection, elimination of splicing in these constructs should render a splicing pattern similar to that of ⌬i2-3,3-4. As shown in Fig. 4, transfection of SA/C resulted in a splice pattern similar to that of the parental construct P3-4⌬i2-3 with a slight reduction in the density of the 480-bp band and appearance of an aberrant mRNA species. The presence of the 480-bp species indicates that intron III can still be spliced out despite mutation of the branch point. The presence of unspliced 513-bp species and the absence of a 501-bp species expected from the single point mutation suggest that the hybridized probe RNA with the single point mismatch is not sensitive to RNase digestion.
By contrast, transfection of QUA/C resulted in diminished upstream splicing as indicated by the disappearance of the 480-bp splice intermediate, and the appearance of the 499-bp unspliced species containing sequences of intron III downstream of the point mutations. This interruption of upstream splicing resulted in aberrant splicing at the alternative splice region, as indicated by the lack of protected exon 4 sequences over the endogenous level and the appearance of aberrant mRNAs. The similar densities of the 171-and 191-bp species (over the endogenous level) suggest a low level of exon 4 to 5 splicing, but we cannot rule out the possibility that the 191-bp species results from aberrant splicing. Using the exon 4 to 6 spliced probe shown in Fig. 2A, we confirmed that normal exon 4 to 6 splicing is diminished in QUA/C-transfected cells (data not shown). The co-existence of the 513-and 499-bp species suggests that only partial digestion occurred in the region of single point mutations. Thus, the 499-bp species, based on the assumption that the probe was excised at the first mismatch immediately upstream of the 3Ј splice site, could be a mixture of 499-, 501-, and 507-bp species derived from the three point mutations within the probing sequences of intron III.
The Abundance of Alternatively Spliced mRNA Species Reflects the Expression Pattern of Active Enzymes-To determine whether the respective alternatively spliced mRNA species result in translation of active enzyme forms, we examined the population of AChE molecular species in transfected myotubes. The endogenous Ache gene in differentiating C2-C12 cells expresses exon 4 to 6 spliced mRNA (Fig. 2) (12,19), and generates mainly monomers and tetramers, both secreted into the media and retained in the cells (Figs. 6 and 7 and Table II) (28). The exogenous enzyme species from transfected gDNA show a distribution pattern (cell-associated or secreted into the medium) similar to that of the endogenous enzyme. By contrast, a predominance of secreted AChE over cellular AChE is observed in ⌬i2-3,3-4-transfected myotubes (Table II).
Expression from cDNAs-The molecular forms of AChE were further characterized by sedimentation in 5-30% sucrose gradients containing either Triton X-100 or Brij 96. Initially, differentiating C2-C12 cells were transiently transfected with Ache cDNAs containing the invariant exons extended to intron IV without splicing, spliced to exon 5, or spliced to exon 6, respectively. Since mRNAs encoded by these three cDNAs lack the alternative splice options, the molecular forms of AChE separated in the gradients are translated from a single mRNA species, thus providing a frame of reference for identifying all of the gene products encoded by alternatively spliced mRNA species.
Monomers associated with mock-transfected cells predominantly shift from about 3.3 S in gradients containing Triton X-100 to about 1.5 S in gradients containing Brij 96, consistent  with the characteristics of amphiphilic monomers described previously (22) (Fig. 6, A and B). The amphiphilic character of the monomers is consistent with the prediction that the unique sequence at the carboxyl terminus encoded by exon 6 (Fig. 5) can form an amphipathic helix with the hydrophobic amino acid residues facing the outside surface of the protein (29). Detergent could bind to this surface owing to the amphiphilic character of the monomeric and dimeric species. Alternatively, the cysteine residue at the carboxyl terminus could be acylated by palmitate, contributing to a detergent binding domain (30). The insensitivity of tetramers to a reduction in S values in gradients containing Brij 96 (Figs. 6 and 7) likely results from occlusion of this amphipathic region from solvent in the assembled tetramers. A small amount of secreted monomers, whose sedimentation coefficient is not reduced by Brij 96, is also present in the medium (Fig. 6, C and D). Transient transfection of exon 4 to 6 spliced cDNA resulted in enhanced expression of enzyme forms similar to the endogenous species (data not shown). Transfection of exon 4 to 5 linked cDNA resulted in expression of a 5.3 S dimeric species associated with the cell. A change in detergent results in marked shift in the S value from 5.3 S in gradients containing Triton X-100 to 3.7 S in gradients containing Brij 96 (Fig. 6, A  and B). This shift in S values is characteristic of amphiphilic AChE forms containing the carboxyl-terminal glycophospholipid anchor and a cysteine residue responsible for dimer formation (Fig. 5) and is consistent with the findings for expression of a cDNA encoding an analogous form in Torpedo (22). Transfection of the cDNA containing exon 4 and its 3Ј intron resulted in enhanced expression of a 3.3 S monomeric AChE species in the cellular fraction, but with most of the hydrophilic monomers being secreted into the medium (3.8 S). These monomers sedimented at about 4.8 -5.1 S in gradients containing Brij 96, values consistent with hydrophilic monomers (Fig. 6) containing a cluster of primarily polar residues at the carboxyl terminus and devoid of the cysteine residue critical for dimerization (Fig. 5) (31, 32). The detergent-sensitive shift of S values for these secreted monomers in EGTA containing phosphate buffer is greater than seen in the absence of EGTA where the differences are only 0.1-0.2 Svedberg units (22,33). 2 Expression from Genomic DNA Constructs-Transient transfection of gDNA in differentiating muscle cells resulted in enhanced expression of cell-associated and secreted monomers, dimers and tetramers ( Fig. 7 and Table II). These species have the same hydrodynamic properties as the endogenous species. Expression of these cDNA constructs is under the control of CMV promoters. Cellular AChE was extracted, and medium AChE was concentrated from cultures of transfected C2-C12 myotubes. AChE molecular forms were separated in 5-30% sucrose gradients and assayed. The sedimentation coefficients of marker enzymes are indicated by arrows from left to right: carbonic anhydrase (3.3 S), alkaline phosphatase (6.1 S), and ␤-galactosidase (16 S). At least two separate transfections for each condition were performed with similar results. A, solubilized cellular AChE forms in Triton-containing sucrose gradients; B, solubilized cellular AChE forms in Brij 96-containing sucrose gradients; C, secreted AChE forms in Triton-containing sucrose gradients; D, secreted AChE forms in Brij 96-containing sucrose gradients. OE, mock-transfected cells; q, cells transfected with a cDNA construct containing exon 4 and its retained 3Ј intron; E, cells transfected with a cDNA construct containing the exon 4 to 5 splice. The expression pattern of a cDNA containing the exon 4 to 6 splice is not shown, but parallels that of the endogenous gene.
We also observed secretion into the medium of a 1.5 S species in gradients containing Brij 96, representing amphiphilic monomers similar to the cell-associated ones. Transient transfection of ⌬i2-3,3-4 resulted in enhanced expression of secreted 3.8 S as well as cell-retained 3.3 S monomers. These monomers sedimented at 5.1 S and 4.8 S in gradients containing Brij 96, respectively, consistent with the hydrophilic species encoded by the cDNA containing exon 4 linked to its 3Ј intron (compare Fig. 7, C and D with Fig. 6, C and D). Some cell-retained monomers shifted to a lower S value in gradients containing Brij 96 (1.5 S), consistent with amphiphilic monomers encoded by exon 4 to 6 spliced mRNA. The lack of a discrete peak in gradients corresponding to the species encoded by exon 4 to 5 spliced mRNA indicates a low expression level of the glycophospholipid-linked AChE species. This is consistent with the observation that expression level of exon 4 to 5 spliced mRNA is the lowest of the splicing alternatives (Table I). Thus, small amounts of amphiphilic dimers encoded by this mRNA species may well be buried in the shoulders of major peaks in both Triton-and Brij 96-containing gradients. Hence, the intron-deleted constructs give rise to a distribution of gene products corresponding to their mRNA abundance.

DISCUSSION
Alternative pre-mRNA splicing from a single Ache gene is the initiating event giving rise to the structural divergence in AChE species (6). Selective splicing of invariant exons to exon 6 generates a single AChE species essential for cholinergic neurotransmission in the central nervous system and at the neuromuscular junction. In this study, we examined the influence of introns and their splicing on AChE pre-mRNA splice selection and show that selective splicing of AChE pre-mRNA during myogenesis is governed by splicing of 5Ј introns, which, in turn, is dependent on maintaining the alternative splicing region intact.
Our data indicate that the first intron, just 5Ј of the translation start site, is critical to AChE expression at the level of transcription or regulation of RNA stability (Fig. 2). Interestingly, when introns II and III were eliminated and the synthetic gene was transiently expressed in differentiating skeletal muscle cells, all three potential splicing options ( Fig. 2 and Table I) and gene products ( Fig. 7 and Table II) were realized, two of which are not normally found in skeletal muscle cells. These data indicate that the deleted introns 5Ј of the donor site in the alternative splicing region of the Ache gene are critical in directing AChE pre-mRNA splicing preferences during muscle differentiation.
The selective splicing of AChE pre-mRNA is highly regulated and the absence of splicing control in ⌬i2-3,3-4-transfected cells is not due to the inability of the cell to process correctly a nascent transcript from a transfected DNA. These conclusions FIG. 7. Representative sucrose gradient profiles showing the AChE molecular forms expressed in C2-C12 myotubes transfected with genomic or the ⌬i2-3,3-4 deletion Ache constructs. Expression of these DNA constructs is under the control of the endogenous Ache promoters. At the initiation of differentiation, C2-C12 myoblasts were transfected and allowed to differentiate for 2 days. AChE in the cells was extracted, and the culture medium was concentrated. AChE molecular forms were separated in 5-30% sucrose gradients and assayed as described under "Experimental Procedures." Sedimentation coefficients of marker enzymes indicated by arrows are identical to Fig. 6. Similar results were obtained from six separate transfections and are summarized in Table II are supported by the following observations. Transfection of gDNA, similar to the endogenous species, resulted in differentiation-induced expression of primarily exon 4 to 6 spliced mRNA ( Fig. 2A and Table I) and the corresponding molecular forms of the active enzyme ( Fig. 7 and Table II). By contrast, only small portion of the AChE pre-mRNA was spliced from exon 4 to 6 when gDNA was transfected into human embryonic kidney cells ( Fig. 2A), indicating that, like the endogenous Ache gene, regulation of preferential expression of the transfected DNA in C2-C12 cells is also species-and/or tissue-specific. Finally, all alternatively spliced mRNA species appear to be translated into functional proteins (Fig. 7).
Since retention of either intron II or III is sufficient to express exon 4 to 6 spliced AChE mRNA with a fidelity approaching that of gDNA, these data are inconsistent with a model where a secondary structure requiring both introns affects AChE pre-mRNA splicing preference. Rather, our data support a model where cis-elements in introns 5Ј of the splice region influence the splice selection. Furthermore, the mRNA splicing pattern in ⌬i2-3-transfected myotubes is closer to the splicing pattern in gDNA-transfected myotubes than that in ⌬i3-4transfected myotubes (Fig. 2, B and C, and Table I). This suggests that the proximal intron 5Ј upstream to the alternative splicing region has greater influence than distal introns in controlling subsequent pre-mRNA splicing. Elimination of the proximal intron enlarges the adjacent exon and thereby increases the bridging distance between neighboring introns. This would decrease the probability of interactions between splicing factors bound to the 3Ј and 5Ј end of neighboring introns (34). In Drosophila, introns flanking an alternatively spliced region are subject to additional constraints during intron evolution, perhaps due to specific sequence requirements for control of alternative pre-mRNA splicing (35).
The lack of selective splicing in ⌬i2-3,3-4-transfected cells indicates that pre-mRNA structure derived from correctly spliced upstream exons is not sufficient to control downstream splice selection, and suggests that splicing events and/or RNA structure derived from upstream individual introns may be critical to splice site selection. This hypothesis is supported by the findings that retention of either upstream intron II or III results in a splice pattern close to that seen for gDNA (Fig. 2, B and C, and Table I), and that splicing of upstream intron III occurs prior to splicing of intron IV (Fig. 4). However, more dramatic changes in the secondary structure of AChE pre-mRNA in P3-4⌬i2-3-transfected cells failed to mimic the splicing pattern seen in ⌬i2-3,3-4-transfected cells, but yielded a splice pattern similar to that seen in gDNA-transfected cells (Fig. 4A). These data argue against a major role of a secondary structure of AChE pre-mRNA derived from a single upstream intron on its downstream splice selection, and are consistent with the general observations that most of the sequence within introns is not essential for splicing regulation (36).
The hypothesis that upstream splicing affects downstream splice selection was tested directly in P3-4⌬i2-3 by mutating the branch point to eliminate upstream splicing events. Transfection of SA/C resulted in a splice pattern similar to that of its parental construct P3-4⌬i2-3, indicating that a single branch point mutation is not sufficient to affect appreciably the splicing of intron III. This is consistent with observations showing neighboring adenosine usage when the preferred branch point is not available (26,27,37). Even though the branch point adenosine is 100% conserved in genes examined in different species (38), the branch site sequences in mammalian genes, including the Ache gene, are much less conserved than that in yeast. In addition, none of the natural mutations of mammalian genes that have been characterized affects the branch site sequences, suggesting the lack of sequence specificity in lariat formation (36).
When four consecutive adenosine residues were mutated near the branch site in the QUA/C mutant, splicing of the shortened intron III was eliminated. Interestingly, elimination of the upstream splicing interrupts downstream selective exon 4 to 6 splice, indicating a sequential splicing mechanism, i.e. a critical event occurs during upstream splicing of AChE pre-mRNA that dictates downstream alternative splice selection. This finding is similar to alternative splicing of preprotachykinin (39) and ␤-tropomyosin pre-mRNAs (16), where downstream splicing controls upstream selective splicing. The differences between alternative splicing in ⌬i2-3,3-4-transfected cells and the absence of regulated splicing in QUA/Ctransfected cells suggest that pre-mRNA structure derived from correctly spliced upstream exons may be sufficient to allow splicing to occur, but splice selection of the correct fidelity is controlled by upstream splicing events. The close to normal splicing pattern in differentiated muscle cells transfected with the ⌬i2-3, ⌬i3-4, and P3-4⌬i2-3 deletion constructs, all of which retain the splicing of at least one upstream intron and correctly spliced upstream exons, supports this conclusion.
The influence of intron IV on upstream splicing is consistent with the "exon definition" described by Berget and colleagues (34,40). In this model, downstream 5Ј splice site is a critical determinant in the recognition and splicing of the upstream intron. Binding of U1 snRNP at this site stabilizes the binding of a splicing factor U2AF at the upstream 3Ј splice site by an exon-spanning network interaction, thus, controlling the splicing efficiency of upstream introns (41,42). It is likely that deletion of intron IV interrupts the binding of splicing factors such as U1 and subsequently the interactions across the exon. This, in turn, results in abnormal upstream splicing.
Since only a single gene encodes all tissue-specific forms of AChE, tissue-specific factors are likely involved in regulation of alternative pre-mRNA splicing. Specific interaction of these factors with conserved cis-elements in the Ache gene can explain the findings in this study. For example, mutation at the branch point will abolish or decrease the binding of splicing factors to the branch point, thus ablating the selectivity of the splicing process. In addition, it is known that splicing of invariant introns is usually constitutive while alternative splicing is highly regulated by specific factors (14). It is possible that exon 4 to 6 splicing is a default mechanism while other splicing options may require activators during differentiation-induced expression of AChE. Under conditions leading to differentiation, introns II and III may serve as "traptors" and allow activators to bind. This will decrease the local concentration of activators so that the default splicing becomes prevalent. Deletion of introns II and III may result in increased local concentrations of activators allowing alternative splicing to occur. Identification of these factors and the cis-elements in the gene should provide further insight into the detailed mechanism of AChE pre-mRNA splicing.
Based on our findings, we propose a mechanism of selective splicing of AChE pre-mRNA during myogenesis. As shown in Fig. 8, splicing near the alternative splice region occurs sequentially from the 5Ј to the 3Ј direction. The presence of intron IV controls the efficiency of upstream splicing since binding of splicing factors to 5Ј splice site of intron IV may facilitate upstream splicing complex formation. Lariat formation and subsequent upstream splicing event commit the partially processed pre-mRNA to the prevailing downstream splice selection in differentiating muscle cells.