INTRODUCTION

Myosin is a family of mechanochemical proteins that demonstrates force-generating ATPase activity when interacting with actin filaments. Myosin is found in all eukaryotic cells and appears to be involved in diverse cellular motile processes, such as muscle contraction, cytokinesis, and cell shape change (1). Conventional myosin (myosin II) molecules consist of a pair of heavy chains (~200 kDa) and two pairs of light chains (15-28 kDa). Chemical, immunological, and molecular cloning studies have demonstrated the presence of multiple isoforms of the myosin II heavy chain (MHC)¹ (2).

For vertebrate nonmuscle cells, cDNA cloning revealed the existence of at least two nonmuscle MHC genes, referred to as MHC-A and MHC-B (3, 4, 5, 6). The two nonmuscle MHC mRNAs are expressed in a variety of tissues, but the relative amounts of the two mRNAs vary among different tissues. In addition to the ubiquitously distributed form of MHC-B, neuron-specific forms of MHC-B, which encode cassettes of inserted nucleotide sequences in two different locations of the MHC head region, have also been demonstrated by cDNA cloning (5, 7). One insertion of 30 nt (encoding 10 amino acids) occurs near the ATP binding domain of the MHC and the other of 63 nt (encoding 21 amino acids) near the actin binding domain. The inserted 10 amino acids include a phosphorylation site for proline-directed kinases (8). Expression of the two different inserts appears to be regulated differently by agonist stimulation in cultured cells as well as during brain development (7). cDNA sequences of these inserted forms of MHC-B suggest that they are most likely generated by alternative splicing of a single MHC-B pre-mRNA. Alternative exons of different sequences and sizes, but identical in position to the 30-nt insert, have been reported for the vertebrate smooth muscle MHC and Drosophila nonmuscle MHC (9, 10).

Alternative splicing of pre-mRNA is a fundamental mechanism for regulating eukaryotic gene expression. In many cases, alternative RNA splicing contributes to developmentally regulated and cell type-specific patterns of gene expression. Although a great deal of information is available concerning the general constitutive splicing reaction of simple transcription units, the molecular basis for alternative splice site selection in vertebrates is poorly understood (for review, see Refs. 11 and 12). A number of features in the pre-mRNA have been shown to influence alternative splice site selection. These include the relative strength of 5- and 3-splice sites, intron size, exon size, branch point sequence or location, exon sequence, and intron sequences (13, 14, 15, 16, 17, 18, 19). To address the mechanisms for regulated alternative splicing, I have made use of the human nonmuscle MHC-B gene as a model system for studying neuron-specific alternative splicing.

MATERIALS AND METHODS

Genomic DNA from human peripheral lymphocytes was amplified by PCR using primers that annealed to the 30-nt insert and 50 nt 5 to the insert in human nonmuscle MHC-B cDNA. The resulting PCR product of ~650 bp, which consists of part of exon 5 and exon N30 (see Fig. 1) and the intron between these two exons, was then used to screen a human genomic DNA library constructed in a FIXII vector (Stratagene) as described previously (20).

The plasmid pET01 (U. S. Biochemical Corp.), which contains the Rous sarcoma virus long terminal repeat promoter and part of the rat preproinsulin gene with a polyadenylation signal, was used as a host vector for the minigene construction. The 13-kb fragment of the human nonmuscle MHC-B gene was inserted into a multiple cloning site located in an intron of the rat preproinsulin gene (see Fig. 2C). Deletions D1-D13 (see Fig. 3A) were generated by removing appropriate restriction fragments or by exonuclease Bal31 digestion and religating the remaining DNA. When religating the linearized DNA, an MluI linker was inserted to mark the joining position and to facilitate introduction of DNA fragments into this position. For construction of D4a-D4i, defined DNA fragments (a-i) (see Fig. 3B) were generated by PCR. All primers contained an MluI site at their 5 ends, and PCR products were inserted into minigene D4 at an MluI site. For C and Cm0-Cm8, two fragments of the MHC-B gene, one starting 241 nt upstream from the 5 end of exon 5 and ending 92 nt downstream from the 3 end of exon N30 (1029 nt, C5 in Fig. 1) and the other starting 261 nt upstream from the 5 end of exon 6 and ending 140 nt downstream from the 3 end of exon 6 (494 nt, C3 in Fig. 1) were joined by recombinant PCR (21). An MluI site was introduced at the junction of two fragments (i.e. between exons N30 and E6). The recombined fragment was inserted into pET01 at a multiple cloning site, and the resulting minigene was designated C. The mutated fragments m1-m8, which were produced by recombinant PCR (21), as well as wild-type fragment m0 were inserted into minigene C at the MluI site (Cm0-Cm8). All constructs were verified by digestion with multiple restriction enzymes and/or sequencing.

Y79()

The human retinoblastoma cell line Y79 was obtained from Dr. Gerald Chader (NEI) and maintained in suspension in RPMI 1640 medium with 10% fetal bovine serum (FBS). The cells were transfected with DNA using Lipofectin (Life Technologies, Inc.). Typically, a mixture of 1-1.5 µg of minigene constructs and 1 µg of pRSV-luciferase (20) and 7-8 µl of Lipofectin was used for transfection of 2 × 10⁶ cells in a 6-well plate. pRSV-luciferase was used to monitor transfection efficiency. For proliferating and undifferentiated stages, transfected cells were continued in suspension culture for 48 h and harvested. For differentiated cells, 1 day after transfection, cells were replated on a poly-D-lysine-coated 6-well plate and maintained in 2 mM sodium butyrate containing culture medium for 6-8 days before being harvested (7, 22). HeLa and NIH 3T3 cells were maintained in minimum Eagle's medium with nonessential amino acids and 10% FBS, and Dulbecco's modified Eagle's medium with 10% FBS, respectively. Transfection of DNA into these cells was performed using calcium phosphate coprecipitation.

Total RNA was prepared from transfected cells as described previously (20). Total RNA was reverse transcribed using random hexamers, and the resulting cDNA was amplified by PCR. The primers for PCR were as follows: upstream primer for both the endogenous gene and minigene transcripts was 5-AGGAAGAAAGGACCATAATATTCC-3 and downstream primers for the endogenous gene and minigene transcripts were 5-CCTGTAGTTATTAAATCCTTCAAG-3 and 5-CCTAGTTGCAGTAGTTCTCCAG-3, respectively. For quantification of the relative ratio of N30 included and excluded products, 5 end-labeled primers were added to the PCR. Following agarose gel electrophoresis, the appropriate bands stained with ethidium bromide were excised from the gels, and radioactivity was determined by Cerenkov counting.

RESULTS AND DISCUSSION

The human genomic clones that encode the region corresponding to the neuron-specific inserted sequences as well as flanking sequences of nonmuscle MHC-B were isolated from a  phage library. Fig. 1 shows the exon-intron organization of the 13-kb fragment of genomic DNA. Exon 5 consists of 103 nt, and exon 6 encodes 93 nt (exon numbers are tentative and based on the numbering for the homologous smooth muscle MHC gene (23)). Between exons 5 and 6 (~8 kb), two alternative exons, one of 30 nt (N30) and a second one of 18 nt (R18), were identified (Fig. 1). This exon-intron arrangement indicates that alternative splicing occurs by either exon inclusion or exon skipping (cassette type) (24).

In mammals and birds, expression of the N30 exon is restricted to neuronal cells, although the nonmuscle MHC-B gene is expressed in most cells (5, 7). Exon R18 expression has been found along with N30 only in butyrate-treated cultured human retinoblastoma cells but not in the retina, which splices in 90-95% of exon N30, or any other neuronal and non-neuronal cells to date. Moreover, the expression of exon R18 alone without N30 has not been found. Therefore, the present study focuses on the regulation of N30 exon splicing.

The original nonmuscle MHC-B minigene construct used for this study is diagrammed in Fig. 2C. The entire 13-kb fragment of the human nonmuscle MHC-B gene was inserted between the rat preproinsulin gene exons. Following transfection, mRNA from this minigene was analyzed by RT-PCR using primers, one of which is located in the human nonmuscle MHC-B gene exon 5 and the other in the rat preproinsulin exon (Fig. 2C). This primer set allows amplification of the mRNA from the minigene construct but not the mRNA from the endogenous host cell gene. The mRNA from the endogenous gene can also be amplified specifically using primers, one of which is located in exon 5 and the other 3 to exon 6. These primers amplify both the N30 included and excluded mRNAs, since they are located upstream and downstream of N30. Moreover, the ratio of the PCR products of included N30 versus excluded N30 reflects the ratio of the starting concentration of the two cDNAs prior to amplification (7, 20, 25).

For minigene transfection, human retinoblastoma Y79 cells (neuronal cells) and HeLa and mouse NIH 3T3 cells (both non-neuronal cells) were chosen as host cells. First, the endogenous nonmuscle MHC-B mRNA was analyzed by RT-PCR, and the results are shown in Fig. 2A. All cell lines analyzed express the endogenous nonmuscle MHC-B mRNA. In addition, Y79 cells treated with butyrate, which cease dividing and are differentiated, express 82.9 ± 5.2% (n = 7) of their total MHC-B mRNA to include the N30 exon or the N30 plus R18 exons, whereas untreated Y79 cells, which are in proliferating stages, express 6.4 ± 1.9% (n = 6) of their total MHC-B mRNA to include the N30 exon. HeLa cells and NIH 3T3 cells exclude both N30 and R18 exons almost completely.

The expression of the nonmuscle MHC-B minigene construct transiently transfected into the same cells is shown in Fig. 2B. The Y79 cells treated with butyrate can process the minigene transcript to include the N30 exon to 32.0 ± 4.5% (n = 8) of the total minigene mRNA. All other cells exclude the N30 exon.

Since the original minigene construct maintained the neuron-specific alternative splicing of the N30 exon, this construct was further modified to test the effects of deletions and mutations on the regulation of N30 exon splicing. Deletion constructs shown in Fig. 3A lack various portions of the ~7.4-kb intron between N30 and E6. These constructs were transiently transfected into Y79 cells, which were made to differentiate by treatment with butyrate. The mRNAs generated from the minigene constructs were analyzed by RT-PCR. As shown in Fig. 3A, the presence or absence of the 605-bp region between the two vertical broken lines, which lies ~1.5 kb downstream from N30, correlates with the inclusion or exclusion of the N30 exon.

To confirm that this 605-bp region contains positive element(s) required for N30 inclusion, the 605-bp fragment (Fig. 3B, i) was reinserted into minigene D4, which has the largest deletion among the D series constructs. The presence of this fragment in the D4 construct in the right orientation restores N30 inclusion (Fig. 3B, D4i). Placement of the fragment in the opposite orientation (Fig. 3B, D4 anti-i) does not restore N30 inclusion. None of the D series deletion constructs as well as the D4i construct affect the complete N30 skipping pattern exhibited by transfected HeLa and NIH 3T3 cells (data not shown). To further narrow down the sequences required for inclusion of the N30 exon, overlapping ~130-bp fragments (Fig. 3B, a-h) from the 605-bp region (Fig. 3B, i) were also placed into minigene D4. As shown in Fig. 3B, fragments c and d, which overlap by 69 bp, also restore N30 exon inclusion, but other fragments do not.

To localize further the putative cis-acting element(s) involved in the neuron-specific N30 exon recognition, a series of minigenes containing clustered point mutations in the 69-bp region described above were tested. The effect of deletion of the same nucleotides was also determined. Since the RNA cis-elements can function through either nucleotide sequence itself or the RNA secondary structure, or both, the secondary structures of the RNA transcript corresponding to the fragments c, d, and c plus d (see Fig. 3B) were first predicted by Zuker's algorithm (26). The RNA structure of the 201-nt transcript of the c plus d fragment (designated m0) is shown in Fig. 4A. The middle 42-nt region, which corresponds to most of the overlapping 69-bp sequence between fragments c and d, forms stem and loop structures, named loop 1 and loop 2 and stem 1-3. Mutations were created at each loop and stem in the 201-nt m fragment. The mutated sequences are shown in Fig. 4A. Mutations m2-m4 and m7 preserve their secondary structure the same as the wild-type m0, whereas m5, m6, and m8 destroy the original secondary structure. m1 lacks stem 1 and loop 1, but the secondary structure of the other nucleotides is the same as the wild type.

To reduce the possible effects of other intronic sequences, minigene C was constructed by joining fragments C5 and C3 shown in Fig. 1. Without an m fragment, the C minigene transcript includes the N30 exon in 13.0 ± 0.9% (n = 2) of the total mRNA (C in Fig. 4B). The inclusion of N30 in the absence of the m fragment in minigene C may be due to the effect of changes in splicing site proximity that have been shown to affect splice site selection (27). Alternatively, it may be due to the presence of negative element(s) in the intronic sequence present in the D4 construct but not in the C construct. Introduction of the m0 fragment into the C construct causes 84.3 ± 6.1% (n = 2) of mRNA to include the N30 exon (Cm0). Thus, the majority of the N30 exon inclusion is still dependent on the presence of the m0 fragment. Mutation of stem 1 with (Cm2) and without (Cm4) mutation of loop 1 results in a decrease in N30 exon recognition to near the background level (C). The same result is obtained with the deletion of loop 1 and stem 1 (Cm1). Mutation of loop 1 (Cm3) shows a partial decrease in N30 recognition (47.3 ± 2.6%), but it is still significantly decreased compared with Cm0 (84.3 ± 6.1%). Three nt mutations of stem 1 (m5 or m6) change the RNA secondary structure, but Cm5 and Cm6 retain N30 exon recognition, suggesting that the sequence itself is more important than its secondary structure. Moreover, of note is the presence of a repeat of the 7-nt sequence, UGCAUGU, within the stem 1 and 2 and loop 1 sequences (see underline in Fig. 4A). Cm5 and Cm6 contain mutations in either one of the two repeat sequences, whereas Cm4 contains mutations in both of the two repeats. Thus, the results above with Cm4-Cm6 also suggest the functional importance of the 7-nt repeat sequence. Mutation of loop 2 and stem 2 (Cm7 and Cm8) as well as stem 3 (data not shown) does not affect the m0 fragment's activity. These results suggest that the stem 1 and loop 1 sequences are critical for N30 exon recognition.

These mutant minigene constructs were also tested in HeLa and NIH 3T3 cells. All of the C series constructs yield 5-10% mRNA that included N30, regardless of the presence or absence of the m fragments (data not shown). This is similar to the effect seen following transfection of Y79 cells with construct C in the absence of m0. The study presented here establishes that the neuron-specific N30 inclusion requires a cis-acting RNA sequence that lies ~1.5 kb downstream of the N30 exon in the primary transcript. The 19-nt sequence in this region (UGCAUGUCGUACUGCAUGU) is critical for its cis-regulatory function.

A number of examples of neuron-specific alternative splicing have been reported (24). For many of these, however, minimal essential elements that affect splicing have yet to be defined. Recently, several regulatory sequences have been identified, including splice sites deviating from the consensus splice sites and elements in the flanking introns (13, 14, 15). Of particular note is the 33-nt intronic sequence, which is required for the neuron-specific N1 exon recognition for the mouse c-src gene (15). When the 19-nt sequence identified here is compared with the 33-nt sequence for the c-src gene, the 7-nt sequence, UGCAUGU, is found to be identical between two elements. As noted above, two copies of this 7-nt sequence exist in the nonmuscle MHC-B gene transcript (Fig. 4A), and mutation analysis suggests that at least one copy of this repeat sequence is required for regulated N30 recognition. Interestingly, however, this 7-nt sequence consists of the hexanucleotide UGCAUG, which has been identified as a cis-element required for alternative splicing of the EIIIB exon in the rat fibronectin gene (19). But alternative splicing of the EIIIB exon is not regulated in a neuron-specific manner. Thus, it is possible that this hexamer sequence present in the MHC-B transcript may contribute to alternative splicing in a general rather than in a neuron-specific mechanism. Recently, the report by Min et al. (28) demonstrated that both neuron-specific and generally expressed factors form a complex to bind to a neuron-specific cis-element in the c-src transcript. Identification and characterization of trans-acting factor(s) for this element and reproducing the 19-nt-dependent N30 splicing in vitro should offer further insights into mechanisms of neuron-specific alternative splicing.