Tissue-selective Expression of α-Dystrobrevin Is Determined by Multiple Promoters*

α-Dystrobrevin, the mammalian orthologue of theTorpedo 87-kDa postsynaptic protein, is a dystrophin-associated and dystrophin-related protein. Knockout of the gene in the mouse results in muscular dystrophy. The control of the α-dystrobrevin gene in the various tissues is therefore of interest. Multiple dystrobrevin isoforms differing in their domain content are generated by alternative splicing of a single gene. The data presented here demonstrate that expression of α-dystrobrevin from three promoters, that are active in a tissue-selective manner, also plays a role in the function of the protein in different tissues. The most proximal promoter A is active in brain and to a lesser extent in lung, whereas the most distal promoter B, which possesses several Sp1 binding sites, is restricted to brain. Promoter C, which contains multiple consensus myogenic binding sites, is up-regulated during in vitro myoblast differentiation. Interestingly, the organization and the activity of the α-dystrobrevin promoters is reminiscent of those in the dystrophin gene. Taken together we suggest that the multipromoter system, distributed over a region of 270 kilobases at the 5′-end of the α-dystrobrevin gene, has been developed to allow the regulation of this gene in different cell types and/or different developmental stages.

␣-Dystrobrevin, the mammalian orthologue of the Torpedo 87-kDa postsynaptic protein, is a dystrophin-associated and dystrophin-related protein. Knockout of the gene in the mouse results in muscular dystrophy. The control of the ␣-dystrobrevin gene in the various tissues is therefore of interest. Multiple dystrobrevin isoforms differing in their domain content are generated by alternative splicing of a single gene. The data presented here demonstrate that expression of ␣-dystrobrevin from three promoters, that are active in a tissue-selective manner, also plays a role in the function of the protein in different tissues. The most proximal promoter A is active in brain and to a lesser extent in lung, whereas the most distal promoter B, which possesses several Sp1 binding sites, is restricted to brain. Promoter C, which contains multiple consensus myogenic binding sites, is up-regulated during in vitro myoblast differentiation. Interestingly, the organization and the activity of the ␣-dystrobrevin promoters is reminiscent of those in the dystrophin gene. Taken together we suggest that the multipromoter system, distributed over a region of 270 kilobases at the 5-end of the ␣-dystrobrevin gene, has been developed to allow the regulation of this gene in different cell types and/or different developmental stages.
Dystrobrevin, a dystrophin related protein, was originally identified from the Torpedo californica electric organ as an 87-kDa phosphoprotein associated with the cytoplasmic face of the postsynaptic membrane (1,2). The protein is expressed in the electric organ, skeletal muscle, and brain, and it has been postulated to play a role in synaptic structure and function, because it copurifies with the acetylcholine receptors and rapsyn from the electric organ membranes. The protein is concentrated with the acetylcholine receptors at the synapse, but it is also found extrasynaptically at the sarcolemma of both Torpedo electric organ and vertebrate muscle. Furthermore, dystrobrevin is also found in association with dystrophin and the 58-kDa syntrophins in the Torpedo electric organ (3,4).
By contrast to the single known Torpedo dystrobrevin molecule, cDNAs encoding several different isoforms differing in their domain content and tissue distribution have been identified in both human and mouse (5)(6)(7). The longest isoform, ␣-dystrobrevin-1 (94 kDa), has a tyrosine kinase substrate domain similar to the Torpedo protein (2,5), in addition to a ZZ- (8) and two predicted ␣-helical-coiled coil domains (9), which it shares with ␣-dystrobrevin-2, while ␣-dystrobrevin-3 has simply the ZZ-domain. The genetic basis of this isoform diversity and additional alternative splicing was resolved by the determination of the genomic organization of the coding region of a single gene on mouse chromosome 18 (6). The conservation of the genomic organization between dystrophin and ␣-dystrobrevin is maintained across the homologous CRCT, 1 indicating that both genes evolved from an ancestral duplication event (10).
Considerable evidence supports an association of dystrophin and ␣-dystrobrevin. Dystrophin and ␣-dystrobrevin colocalize in skeletal muscle, copurify biochemically, and associate directly in vitro via the coiled-coil region of dystrophin (11)(12)(13). The expression pattern of the ␣-dystrobrevin gene also closely parallels that of dystrophin, where a set of diverse isoforms are generated in vivo by alternative splicing in brain and muscle. Dystrophin expression in muscle and brain results in three 14-kb transcripts (muscle-type, brain-type, and Purkinje celltype), controlled by promoters at the 5Ј-end of the gene (14,15). In addition, a whole family of smaller mRNAs are transcribed from promoters lying between exons within the rod domain of the gene. The first of these transcripts designated Dp71 (apodystrophin-1) and apodystrophin-3, encode the CRCT, or the first part of the CRCT, respectively, and are expressed predominantly in brain and non-muscle tissues from a housekeeping like promoter (16,17). Other smaller transcripts are expressed in a tissue-specific manner, such as in peripheral nerve (Dp116, Ref. 18), the retina, brain, and cardiac muscle (Dp260, Ref. 19), and throughout the central nervous system (Dp140,Ref. 20).
In view of the similarity between dystrophin and ␣-dystrobrevin, and the recent evidence that mice null for ␣-dystrobrevin suffer from muscular dystrophy and impaired aggregation of acetylcholine receptors (21), it was of interest to determine the regulation of this gene. Our preliminary evidence indicated a minimum of four 5Ј-UTR exons, suggesting that the regulation of the ␣-dystrobrevin gene might be as complex as dystrophin. While in the process of investigating whether particular ␣-dystrobrevin isoforms are associated with a specific 5Ј-UTR region, we identified additional 5Ј-UTR exon sequences, consistent with a regulation of expression by multiple promoters.
Here we present evidence that the mouse 5Ј-region of the gene is composed of seven 5Ј-UTR exons covering 270 kb of genomic DNA and demonstrate that ␣-dystrobrevin is expressed from three promoters that are active in a tissue-selective manner. Our data suggest that this multipromoter system of the ␣-dystrobrevin gene has been developed to allow the regulation of this gene in different cell types and/or different developmental stages.
RNA Extraction, Northern Blotting, and RNase Protection-RNA extraction, Northern blots, and RNase protection assays were performed as described previously (22). Genomic regions spanning the most 5Ј-end exons A, B, and C, and their adjacent 5Ј-flanking sequences were amplified by PCR using the primer sets RnpAf/RnpAr for exon A, RnpBf/RnpBr for exon B, and RnpCf/RnpCr for exon C, subloned into pGEM-T, and used to generate cRNA probes.
5Ј-RACE-5Ј-RACE was carried out using the 5Ј AmpliFINDER RACE kit (CLONTECH). 1 g of poly(A) ϩ RNA isolated from mouse tissues was reverse transcribed using a primer derived from exon 6 (DB6r, 5Ј-TGCAGAAGAGGCAGCCATACC-3Ј). The anchor ligated cDNA was then PCR amplified using the anchor primer and the internal primer derived from exon 3 (DB3r2,). PCR-products were cloned into pGEM-T vector (Promega), and the largest were sequenced using Sequenase v2.0 (U. S. Biochemical Corp.) The sequences of 20 exon A-, B-, or C-containing clones were aligned using PILE UP (Genetics Computer Group, version 8.0; Madison, WI).
Cell Culture, Transfection Procedures, and Luciferase and ␤-Galactosidase Assays-H-2K b -tsA58 cells (23) were propagated at 33°C in Dulbecco's modified Eagle's medium supplemented with 20% fetal calf serum, 2% chicken embryo extract, and 20 units/ml mouse recombinant interferon-␥ (Life Technologies, Inc.) (growth medium). Myoblast fusion was induced by culturing the cells in Dulbecco's modified Eagle's medium supplemented with 5% horse serum and 1% chicken embryo extract (differentiation medium) at 39°C in 10% CO 2 . H-2K b -tsA58 myoblasts at approximately 40% confluence were transiently transfected with 2 g of luciferase transporter construct and 0.5 g of pSV-␤-galactosidase plasmid (Promega) using Superfect protocol (Qiagen). The cells to be maintained as myoblasts were incubated in growth medium throughout the experiment, whereas cells to be differentiated into myotubes were switched to differentiation medium after reaching 80% confluence (usually within 24 h). Cells were harvested 72 h after the start of transfection. NIH3T3 cells were transfected in the same manner and harvested 60 h after the start of transfection. Each transfection was performed in triplicate and repeated twice. Cells were washed twice with phosphate-buffered saline, harvested into 200 l of Reporter Lysis Buffer (Promega). Cell extract (20 l) was mixed with 100 l luciferase assay reagent (Promega) and light production quantified using a Turner Designs Model 20 luminometer. ␤-Galactosidase activity was measured using an enzyme assay system (Promega) and then analyzed using a spectrophotometer at 420 nm.

5Ј-UTR Exons and
Alternative Splicing-We have described multiple transcripts for mouse ␣-dystrobrevin that encode three protein isoforms (5,6,11). Some of these transcripts differed in their 5Ј-UTRs, suggesting that they may be by alternative splicing of different 5Ј-exons, indicative of transriptional regulation by multiple promoters. To determine the origin of these transcripts and the mechanism of ␣-dystrobrevin expression, we characterized genomic fragments covering the 5Ј-end of the gene. We have reported previously the exon-intron boundaries of four UTR exons A-D, derived from different cDNA clones. Here, 5Ј-UTR sequences have been extended to their 5Ј-ends by sequencing of RACE products of the dystrobrevin mRNAs from mouse heart and brain. A PCR approach on a YAC vectorette library of the YAC clone ICRFy902M0312Q positive for the dystrobrevin gene was used to clone three additional 5Ј-UTR exons and determine their exon-intron boundaries (Fig. 1). Exon sizes and splice junctions sequences of all seven 5Ј-UTR exons are summarized in Table I. The splice junctions have the conserved GT and AG dinucleotides present at the 5Ј splice donor and 3Ј splice acceptor sites (25). The three most upstream exons, exons A, B, and C, are all of similar size and are larger than the internal exons D-G. Mapping of the 640-kb YAC clone ICRFy902M0312Q indicated that the maximum distance between the UTR exons is approximately 270 kb. To determine the order and estimate the distance between the UTR exons, a mouse BAC genomic library was screened using primer pairs for all 5Ј-UTR exons and the coding exon 3. This resulted in the isolation of three overlapping BACs, two of which (BAC 343 and BAC 547) had insert sizes of 170 and 140 kb and contained the first coding exon, UTR exons A and D. The third BAC, with an insert size of 160 kb, contained the other five UTR exons: exons C, G, E, F, and B. Fig. 2A shows the approximate positions of the seven UTR exons spanning a distance of approximately 270 kb relative to the 25 coding exons of the gene.
Comparison of the nucleotide sequences of the seven 5Ј-UTR exons to the sequences of 5Ј-RACE products from brain and heart revealed a complex pattern of alternative splicing. As indicated in Fig. 2B, exon A was spliced directly to exon 1, the first protein-coding exon, while both exon B and exon C were separated from exon 1 by a further exon, exon D. In a minority of RACE clones, one or the other of three additional 5Ј-UTR exons, exons E, F, and G, were spliced between exons B and D. All RACE clones were identical in sequence downstream of exon 1. These results suggest that the different dystrobrevin transcripts are generated by differential usage of three promoters.
Mapping of Transcriptional Initiation Sites-Twenty RACE clones for each type of end-terminal UTR exon (exons A, B, and C) were sequenced and provide an indication of the location of transcriptional start sites. The majority of the exon A-positive clones terminated 239 bp upstream of the 3Ј-end of exon A. By contrast, RACE clones positive for exon B and exon C ended at several different points, but in each case there was a cluster of clones. Thus for exon B clones terminated at nucleotide Ϫ95, Ϫ110, Ϫ121, Ϫ128, Ϫ134, Ϫ142, Ϫ148, Ϫ197, and Ϫ200, where the 3Ј-end of the exon is designated as Ϫ1. Similarly transcription initiation sites were identified at nucleotide positions Ϫ121, Ϫ146, Ϫ171, Ϫ185, amd Ϫ209 for exon C. In all three populations of RACE clones we did not find any clones that were spliced within these regions. Moreover, there are no multiple AG sequences to serve as splice acceptor sites.
To confirm the location of the cap sites for the various ␣-dystrobrevin mRNAs, three genomic probes were constructed and used in the RNase protection assay. Probe A was a 299-bp PCR fragment of exon A and its 5Ј-flanking region. Using this probe, a major 243-bp fragment was protected by mRNA from mouse brain, heart, lung, and skeletal muscle. Additionally, a second product at 170 bp was also evident, albeit at very low intensity, indicating that this fragment might reflect a minor transcription start site (Fig. 3). By contrast probe B, a 268-bp PCR fragment of exon B and its 5Ј-flanking region, revealed multiple protected products ranging in size from 223 to 109 bp. None of these fragments were protected by mRNA from lung or muscle. Using a 258-bp PCR fragment of exon C and its 5Јflanking region revealed major protected products at 209, 195, 164, and 135 bp. Interestingly, the intensity of the 135-bp protected fragment was significantly higher in skeletal muscle than in all the other tissues. This protected fragment probably reflects a preferentially used cap site for ␣-dystrobrevin mRNAs in skeletal muscle. Weak muscle-specific products at 183 and 175 bp were also detected. In summary, multiple transcription start sites for all three promoters were detected.
Nucleotide Sequence of Regions Upstream of 5Ј-UTR Exons A, B, and C-To identify putative regulatory promoter elements that flank the most upstream 5Ј-UTR exons, we sequenced a 858-, 528-, and 994-bp upstream of exons A, B, and C, respectively (Fig. 4). In common with other genes with multiple transcription start sites, no typical TATA or CAAT box was present in the most distal promoter B region. However, we found the motif GCTCCC downstream of the ϳ100-bp multiple start site window, which is identical to a conserved downstream element defining a new subclass of RNA-polymerase II promoters (26). Computer-assisted analysis revealed several putative transription factor binding sites including Sp1 and Ap2. The promoter C sequence also contained no TATA box in the first 100 bp, although a CAAT box consensus was observed at nucleotide Ϫ337. Interestingly, we identified multiple sequences within a 370-bp region known to mediate skeletal and cardiac muscle expression of a number of genes. These include four conserved MEF-1 motifs (27) at Ϫ294, Ϫ234, Ϫ150, and Ϫ48, a TGCCTGG motif at Ϫ363 (28), and a M-CAT motif at Ϫ30 (29), where the most distal cap site is designated as ϩ1. Examination of the most proximal promoter A sequence did not reveal a consensus TATA box or binding sites for common transcription factors such as Sp1 or CAAT factors, and the sequence was not GC-rich. The major initiation site is located within a pyrimidine-rich sequence, TTTTGTCAGTCTTTT (cap site underlined). This sequence is similar to the initiator (Inr) consensus sequence, 5Ј-PydPyd-CA-PydPydPydPydPyd, that can direct specific transcription initiation in TATA less and non-GC-rich promoters (30). Several putative transcription factor sites, inclucding AP1 and PuF, were found within the promoter A sequence.
Promoter Activities in the 5Ј-Flanking Regions of Upstream Exons-Genomic fragments containing the putative promoter regions A, B, and C (Fig. 4) were cloned upstream of the luciferase gene. These constructs were then transiently transfected into the mouse fibroblast NIH3T3 and mouse myogenic H2K-tsA58 cell lines. All constructs directed luciferase expression in NIH3T3 cells (Fig. 5A). The level of luciferase activity produced by the constructs containing sequences of the 5Јflanking regions of exon B and exon C was substantially higher in forward (Bf and Cf) than in reverse orientation (Br and Cr). The promoter strength of the construct Bf was comparable with the activity of the herpes simplex virus thymidine kinase promoter used as a control. The construct Cf was about 3-fold stronger than this promoter. The 5Ј-flanking region of exon A TAGGGTTGgtaggtag. . . showed orientation-independent expression in this cell line and was excluded from further analysis.
To investigate the effect of myoblast differentiation on ␣-dystrobrevin expression, we transfected the aforementioned constructs into H2K-tsA58 myoblasts and analyzed luciferase activity in total cell extracts prepared from H2K-tsA58 muscle cells at different time points during myoblast fusion. The construct Cf exhibited higher luciferase activity in myotubes than in undifferentiated myoblasts, with a ϳ200-fold increase in differentiated H2K-tsA58 myotubes over that obtained with the promoter construct Bf (Fig. 5B), which showed similar low levels of activity in myoblasts and myotubes. Luciferase activities in extracts of cells transfected with a construct containing the same region in reverse orientation (Cr and Br) and a promoterless pGL-3 basic vector are shown as negative controls.
In order to determine the endogenous transcription pattern of ␣-dystrobrevin in the H2K-tsA58 cell line, total RNA was isolated from myoblast cultures at various stages of differentiation. First, the differential usage of 5Ј-UTR sequences was investigated by RT-PCR using UTR exon-specific forward primers and a common reverse primer in exon 3. PCR products were detected by hybridization to an oligonucleotide probe derived from exon 1. Little or no exon A or exon B containing transcripts could be detected in this cell line at the different time points tested (Fig. 6A). By contrast, levels of exon C-containing transcripts increased steadily from very low levels at the nonconfluent myoblast stage to higher levels in cultures containing differentiated multinucleated myotubes. A 370-bp amplification product of utrophin was present at all different time points at similar levels, indicating that its expression was not affected during myogenesis (11).
To determine whether exon C-containing transcripts encode all three ␣-dystrobrevin isoforms, we performed RT-PCR using a common forward primer and isoform-specific reverse primers. Southern blot analysis of the PCR products using a radiolabeled m32 cDNA probe (5) revealed single bands for all three dystrobrevin isoforms in proliferating myoblasts (Fig. 6B). After switching to differentiation medium, the level of ␣-dystrobrevin-3 (Fig. 6B, ␣-db-3) increases and then stabilizes as myoblast fusion proceeds. In the case of ␣-dystrobrevin-1 (Fig. 6B, ␣-db-1) and -2 (Fig. 6B, ␣-db-2), an additional upper band appears at increased levels, whereas the lower band seen in proliferating myoblasts remains at almost at the same level. This amplification product probably reflects an alternatively spliced form of ␣-dystrobrevin-1 and -2 containing the vr3 sequence (5).
Western blot analysis of protein extracts prepared from H2K cells at different time points during myoblast fusion using an antibody against ␣-dystrobrevin-1 detects two proteins of similar relative mobility (Fig. 6C). Consistent with our detection of ␣-dystrobrevin-1 mRNA (Fig. 6B, ␣-db-1), the protein is found at low levels in undifferentiated H2K-tsA58 myoblasts. After switching to differentiation medium, ␣-dystrobrevin-1 becomes more abundant and remains expressed at constant levels as myoblast fusion proceeds. At this time, the muscle-specific splice variant of ␣-dystrobrevin-1 containing the vr3 sequence is also detected (5). Similar results were obtained for ␣-dystrobrevin-2 from expression studies in the myogenic C2C12 cell line (11). Taken together these results indicate that in the H2K-tsA58 cell line the transcriptional activation of the mouse ␣-dystrobrevin gene occurs upon differentiation of myoblasts into multinucleated myotubes. Promoter C, which contains multiple consensus myogenic binding sites, is active, whereas promoters A and B are not. However it is formally possible that a fourth as yet unidentified promoter is active as well.
Tissue-selective Expression of ␣-Dystrobrevin mRNAs Containing Different 5Ј-UTR Exons-To examine the expression FIG. 3. Mapping of transcription initiation sites in the ␣-dystrobrevin gene. 40 g of total RNA from the indicated tissues was hybridized to antisense cRNA probes shown at the bottom. An RNase protection assay was then performed as described under "Experimental Procedures." The following fragments were used as templates to synthesize cRNA probes: in A a 299-bp PCR fragment of exon A and its 5Ј-flanking region; in B a 268-bp fragment of exon B and its 5Ј-flanking region; in C a 258-bp fragment of exon C and its 5Ј-flanking region; H, heart; L, lung; B, brain, S, skeletal muscle; Y, yeast tRNA; DNA size markers are designated by number of base pairs. pattern of UTRs originating from the three putative dystrobrevin promoters, RT-PCR on the same amount of total RNA isolated from mouse brain, lung, skeletal, and cardiac muscle was performed using UTR exon-specific forward primers and a common reverse primer in exon 3. Fig. 7 indicates that exon A was predominantly expressed in brain and to a lower extent in lung. By comparison, exon C was highly expressed in muscle tissues, with significant but lower expression in brain and lung, and expression of exon B was only found in the brain. Primers for the exons E, F, and G produced products of the expected size in brain, confirming that they are expressed and are not artifacts of the RACE protocol. However, there was no evidence for the presence of exon F and exon G in mRNAs derived from exon B transcripts in brain, indicating that additional promoters might be present within this huge control region of the ␣-dystrobrevin gene.
To determine whether mRNA species encoding the three major ␣-dystrobrevin isoforms are associated with a particular promoter, we hybridized 5Ј-UTR exon-specific probes to multiple tissue Northern blots. The hybridization patterns are sum-marized in Fig. 8. Hybridization with a common probe spanning exons 1-6 illustrates the five predominant transcripts encoding the three major isoforms, which are estimated to be 7.5 and 4.0 kb (␣-dystrobrevin-1), 5.0 and 3.6 kb (␣-dystrobrevin-2), and 1.7 kb (␣-dystrobrevin-3) in size as described previously. The main difference in the length of these transcripts is due to differential splicing of three exons containing stop codons and alternative usage of polyadenylation sites (except for ␣-dystrobrevin-3). All three UTR exon probes hybridized to the dystrobrevin-1 and dystrobrevin-2 transcripts in brain. Both exon A and exon C sequences were found in the full-length dystrobrevin-1 transcript expressed in lung and in a 3.8-kb transcript that has not been assigned to any of the three isoforms (6). The 1.7-kb transcript corresponding to the recently described muscle expressed ␣-dystrobrevin-3 isoform (11) was detected by the exon C probe in muscle, but also hybridized, albeit weakly, with the exon A probe in brain and possibly lung. These results indicate that the promoters are active in a tissueselective rather than a tissue-specific manner and that the formation of individual dystrobrevin variants is independent of promoter activity and is probably the result of a post-transcriptionally regulated process. DISCUSSION In the present study we have identified multiple promoters that control the expression of the ␣-dystrobrevin gene. The exon-intron arrangement of the unusually long 5Ј-end of the gene explains the origin of the different mRNA species that we have characterized (Ref. 5 and Fig. 2B). It is clear from the structure of the gene that both the alternative use of three promoters and the differential splicing of the resulting transcripts is involved in the generation of the multiple ␣-dystrobrevin mRNAs. The most distal brain promoter is separated by 120 kb of genomic sequence from the muscle promoter. A third promoter located 70 kb further downstream is separated by another 80 kb of genomic sequence from exon 1, which contains the translational start site. Remarkably, the region containing the seven small UTR exons A-G spans 270 kb of genomic DNA. Considering that the ␣dystrobrevin gene is organized into 25 coding exons contained within a genomic interval of 170 kb, the total size of the gene can now be estimated as at least 440 kb (6).
The ␣-dystrobrevin gene exhibits a structural arrangement similar to a number of other genes in which multiple promoters generate tissue-specific mRNAs that differ only at their 5Јuntranslated region. Four short 5Ј-noncoding exons of the rat gene for brain-derived neurotrophic factor can be spliced to a common coding exon and are each regulated by separate promoters (31). These promoters confer tissue-specific, axotomyand neuronal activity-induced expression in transgenic mice (32). Our results show that all three promoters of the ␣-dystrobrevin gene are active in brain. It is therefore possible that the ␣-dystrobrevin transcripts detected in the brain are transcribed in a region or cell specific manner from these promoters. Interestingly, the arrangement of the control region of the ␣-dystrobrevin gene is very similar to that described for the dystrophin locus, where tissue-specific promoters regulate expression of full-length dystrophin isofoms (14,33). One of these promoters, which is located upstream of the muscle promoter, regulates the expression of dystrophin in the cortex and hippocampus (34), while a third promoter, active in brain Purkinje cells, has been identified between the muscle promoter and the second exon of dystrophin (15). However, whereas the dystrophin isoforms have different first coding exons, the dystrobrevin transcripts described utilize the same first coding exon and H2K-tsA58 cells were either allowed to remain myoblasts or were induced to form myotubes; L-MB, late H2K-tsA58 myoblasts; MT, differentiated H2K-tsA58 myotubes (day 2 after switching to differentiation medium). TK refers to a control plasmid (pXp2luc) containing the herpes simplex virus thymidine kinase promoter fused to the luciferase gene. pGL3 represents a pGL3 basic vector without an insert. The luciferase activity of each construct was normalized in comparison with co-expressed ␤-galactosidase activity. The relative luciferase activities are shown as a percentage of the activity of either pXp2luc (in NIH3T3 cells) or Cf constructs (in H2K-tsA58 myoblasts), which were arbitarily set to 100%. Error bars on graphs and numerical errors on overall percentages are shown.

ATG.
In cases in which a promoter switch does not affect the coding region, as in the ␣-dystrobrevin gene, the translation efficiency of the mRNAs can be affected (35). We found that the long 5ЈUTRs of the ␣-dystrobrevin mRNAs contain small partially overlapping upstream open reading frames that precede the major translation initiation site, which are highly conserved between human and mouse (data not shown). It has been suggested that such an arrangement might be particularly suitable for translational regulation (36). The mouse gene for the retinoic acid receptor-␤2 has a similar complex 5Ј-UTR exon organization. Mutation in the small open reading frames of the 5Ј-UTR affected expression of the downstream major open reading frame, resulting in an altered regulation of gene expression in vivo (37).
The regions upstream of the 5Ј-UTR exons A and B of the ␣-dystrobrevin gene exhibited none of the consensus features that define proximal promoter regions of tissue-specific promoters. The presence of multiple transcriptional start sites could be expected, since these regions lack a canonical TATA box-like element conferring a unique transcription start site (38). This situation is reminiscent of housekeeping genes (30). Despite the fact that the discovery of TATA-less promoters is steadily increasing, there is little information how this multiple selection process occurs. Recently, a protein binding sequence GCTCCC (MED-1; multiple start element downstream) was found to be positionally conserved in a number of promoters that initiate at multiple unclustered start sites (ϳ100-bp window) and was shown to be involved in the regulation of this FIG. 6. Expression of ␣-dystrobrevin isoforms in H2K-tsA58 myoblasts. Total RNA was prepared from H2K-tsA58 muscle cells at different time points during myoblast differentiation. MB, early (E) proliferating and late (L) confluent myoblasts; MT, different time points (in days, d) during myotube formation. A, autoradiographs of Southern blots of PCR products (20 cycles), which were amplified from same amounts of cDNA using forward primers specific for the 5Ј-UTR exons A-C and a common reverse primer. An amplification product of utrophin was used as a positive control. The blot was hybridized with an internal specific oligonucleotide derived from exon 1. The sizes of the PCR products in bp are indicated on the right of the figure. B, autoradiographs of Southern blots of PCR products (20 cycles), which were amplified from same amounts of cDNA using a common forward primer derived from exon 8 and reverse primers specific for ␣-dystrobrevin-1, ␣-dystrobrevin-2, and ␣-dystrobrevin-3, respectively. The blot was hybridized with a radiolabeled 416-bp NsiI/HindIII restriction fragment of the m32 cDNA clone (5). C, total cell extracts were prepared from H2K-tsA58 cells at time points indicated above. 20 g of protein were separated on 8% SDS-polyacrylamide gels, blotted, and probed with an antibody against ␣-dystrobrevin-1.

FIG. 7. Tissue-selective usage of ␣-dystrobrevin 5-UTR exons.
Autoradiographs of Southern blots of PCR products (20 cycles), which were amplified from same amounts of cDNA using primers specific for 5Ј-UTR exons A-G and a common reverse primer derived from exon 3. The blot was hybridized with a common oligonucleotide derived from exon 1. The sizes of the PCR products in bp are indicated on the left. H, heart; L, lung; B, brain, S, skeletal muscle. multiple initiation process (26). We found an identical sequence stretch immediately downstream of the multiple cap sites in exon B, and therefore the exon B promoter might be a member of this new subclass of TATA-less promoters.
Our RT-PCR studies clearly indicate that in H2K-tsA58 myoblast the transcriptional activation of the ␣-dystrobrevin gene occurs upon differentiation of myoblasts into multinucleated myotubes. Functional promoter studies show that the 1169 bp promoter C fragment is able to direct luciferase expression during myoblast differentiation. The proximal part of this region is composed of cis-acting sequences that have been implicated in muscle specific expression. Four copies of the CANNTG motif present in the MyoD target sequence and characteristic of many muscle-specific regulatory regions (27,39) and a "M-CAT" similarity and a TGCCTGG sequence, both of which have been proposed as muscle-specific motifs (33,34), are present within a 370-bp region upstream of the transcription initiation site. Taken together we suggest that the multipromoter has been developed to allow the regulation of the ␣-dystrobrevin gene in different cell types or at different developmental stages at the level of both transcription and translation.