Disruption of Splicing Regulated by a CUG-Binding Protein in Myotonic Dystrophy

See allHide authors and affiliations

Science  01 May 1998:
Vol. 280, Issue 5364, pp. 737-741
DOI: 10.1126/science.280.5364.737


Myotonic dystrophy (DM) is caused by a CTG expansion in the 3′ untranslated region of the DM gene. One model of DM pathogenesis suggests that RNAs from the expanded allele create a gain-of-function mutation by the inappropriate binding of proteins to the CUG repeats. Data presented here indicate that the conserved heterogeneous nuclear ribonucleoprotein, CUG-binding protein (CUG-BP), may mediate the trans-dominant effect of the RNA. CUG-BP was found to bind to the human cardiac troponin T (cTNT) pre–messenger RNA and regulate its alternative splicing. Splicing of cTNT was disrupted in DM striated muscle and in normal cells expressing transcripts that contain CUG repeats. Altered expression of genes regulated posttranscriptionally by CUG-BP therefore may contribute to DM pathogenesis.

Myotonic dystrophy (DM) is a dominantly inherited neuromuscular disorder caused by a trinucleotide (CTG) expansion in the 3′ untranslated region (UTR) ofDM, the gene that encodes DM protein kinase (1). The severity of symptoms correlates with the length of the expansion (2). It has been proposed that the CUG repeats in theDM RNA create a gain-of-function mutation that disrupts processing or transport of heterologous RNAs in a trans-dominant manner (3-5). This hypothesis is supported by the observation that transcripts of the expanded DM allele accumulate in the nucleus (6). One possible mediator of a trans-dominant effect is CUG-BP (or hNab50), a conserved heterogeneous nuclear ribonucleoprotein (hnRNP) that binds specifically to RNA that contains CUG repeats (7, 8). CUG-BP is ubiquitously expressed, distributes to both the nucleus and cytoplasm, and exists in two isoforms that differ by the extent of phosphorylation (7, 9). The hypophosphorylated isoform accumulates in the nucleus in tissues and cultured cells from DM patients (7,9). Altered posttranscriptional processing of transcripts that require CUG-BP has been hypothesized to contribute to DM pathogenesis; however, target genes for CUG-BP activity have not been identified.

We noticed that CUG repeats occur within muscle-specific splicing enhancers (MSEs) of the chicken cTNT gene (10) (Fig.1, A and B). Chicken cTNT is expressed in embryonic and adult cardiac muscle and embryonic skeletal muscle, and alternative splicing of exon 5 is regulated so that the exon is included in mRNAs produced during early development of heart and skeletal muscle but not in adult heart. The gene is transcriptionally silent in adult skeletal muscle (11). The MSEs (10) and the CUG repeats within the MSEs (12) are required for the enhanced inclusion of ccTNT exon 5 in embryonic striated muscle. Regulation of cTNT alternative splicing is conserved in chickens and humans (13), and a similar CUG motif occurs in human cTNT, 31 nucleotides (nt) downstream of exon 5 (Fig. 1B).

Figure 1

A conserved muscle-specific splicing element downstream of exon 5 in human cTNT. (A) Diagram of the ccTNT genomic fragment necessary and sufficient for regulated splicing of exon 5 (black box) in striated muscle. Four MSEs (white boxes) are required for enhanced exon inclusion in striated muscle (10). (B) A common sequence in chicken cTNT MSEs and human (hcTNT) intron 5 [highlighted in black in (A) and (B)] contains CUG repeats (underlined).

To investigate whether cTNT splicing patterns are altered in DM, we analyzed cTNT transcripts in cardiac tissue from two unrelated DM patients by reverse transcriptase–polymerase chain reaction (RT-PCR) (Fig. 2, A and B). In both patients, the fraction of cTNT mRNAs that included the alternative exon was greatly increased compared with that in normal adult cardiac tissue or several cardiomyopathies. Primary cell lines derived from DM skeletal muscle also showed an abnormal shift to the embryonic splicing pattern (Fig. 2B) (14). Altered splicing of cTNT in DM striated muscle was not due to nonspecific changes in splicing efficiency, as splicing of the hnRNP A2/B1 alternative exon, which is not regulated in striated muscle and is not flanked by CUG-containing elements, was unaffected (Fig. 2, A and C).

Figure 2

RT-PCR analysis of cTNT and hnRNP A2/B1 alternative splicing in DM and non-DM cardiac and skeletal muscle cells. (A) Variable regions of the human cTNT and hnRNP A2/B1 mRNAs are illustrated along with relative positions of the oligonucleotides used for RT-PCR. cTNT exon 4 (15 nt) and exon 5 (30 nt) are included or excluded as cassettes. cTNT exon 5 also contains an internal alternative 3′ splice site, which removes 3 nt. hnRNP A2/B1 exon 2 is spliced as a cassette. RT-PCR products from the cTNT mRNAs labeled A to D are 134, 131, 104, and 89 nt, respectively. RT-PCR products for hnRNP A2/B1 mRNAs A and B are 170 and 134 nt, respectively. (B) RT-PCR analysis of endogenous cTNT mRNA from heart tissue and primary skeletal muscle cell lines. DM1 was homozygous for the expanded repeat and presented as a classic case of DM (3, 25). DM2 had an expansion of >1600 repeats in one allele and a normal (12 repeats) second allele. Control heart tissue samples include three samples from normal patients (N), seven samples from patients with left ventricular hypertrophy (LH), and one patient with right ventricular hypertrophy (RH) secondary to pulmonary hypertension. The patients' ages (years) were as follows: N1 (58), N2 (63), N3 (20–78, pooled), LH1 (65), LH2 (54), LH3 (69), LH4 (69), LH5 (83), LH6 (66), LH7 (81), RH (53), DM1 (21), DM2 (16 days). DM skeletal muscle cultures were from DM1. Reverse transcription was done on 4 μg of total RNA with 10 ng of the reverse primer (12). One-quarter of the cDNA was added to 40-μl PCR mixtures containing Taq polymerase buffers (Promega), 2 ng of the forward primer 32P-labeled by polynucleotide kinase, 120 ng each of unlabeled forward and reverse primers, and 2.5 units of DNA Taq polymerase. Twenty cycles were carried out at annealing and extension temperatures of 70° and 72°C, respectively. Forward and reverse oligonucleotides for cTNT were ATAGAAGAGGTGGTGGAAGAGTAC and GTCTCAGCCTCTGCTTCAGCATCC, respectively. PCR products were resolved on 5% nondenaturing polyacrylamide gels. Bands were quantitated with a PhosphorImager (Molecular Dynamics). The percent of exon 5 inclusion is calculated as {[counts per minute (cpm) of bands A + B]/(cpm of bands A + B + C + D) × 100}. Small amounts of band B are detectable in some lanes. Because the results are expressed as a ratio, lane-to-lane variation in the amount of mRNA is internally controlled. (C) RT-PCR analysis of hnRNP A2/B1 mRNA. Forward and reverse oligonucleotides were CTGAAGCGACTGAGTCCGCG and ACAGTCTGTAAGCTTTCCCC, respectively.

To explore the basis for the aberrant splicing, we inserted a genomic fragment from the human cTNT gene containing exon 5 into the intron of a constitutively spliced minigene (Fig.3A) (10). The hcTNT minigene was transiently transfected into normal and DM primary skeletal muscle cell lines (15), and regulation of exon 5 splicing was assayed by RT-PCR. Substantially more minigene transcripts containing exon 5 were present in DM than in normal skeletal muscle cells (Fig. 3B). Thus, the alteration in cTNT splicing in DM cells is mediated by factors acting in trans.

Figure 3

Transient transfection analysis of human cTNT alternative splicing. (A) Structure of hcTNT and hcTNT CUG → CAG minigenes. Five vertical lines represent the proximal CUGs that are mutated in the hcTNT CUG → CAG minigene. Arrows indicate positions of the oligonucleotides used for RT-PCR analysis. The enhancer and transcription initiation site are from Rous sarcoma virus (RSV). (B) Transient transfection of hcTNT and hcTNT CUG → CAG minigenes into primary skeletal muscle cell lines from normal individuals (N) and DM patients. For the experiment shown, the percent of spliced mRNAs that include exon 5 is indicated below each lane. The averaged results from five or more independent experiments (lanes 1 to 4) are 41 ± 6.7%, 89 ± 4.5%, 41 ± 9.3%, and 50 ± 19%. (C) Response of hcTNT (circles) or hcTNT CUG → CAG (boxes) to increasing amounts of cotransfected CUG-BP expression vector in normal human primary skeletal muscle cell lines (n = 5 independent transfections). Full-length CUG-BP cDNA replaced the chloramphenicol acetyltransferase (CAT) gene in the pOPRSVICat expression vector (Stratagene). The increase in percent inclusion of exon 5 was calculated as the difference in the percent of spliced mRNAs that include exon 5 with and without the cotransfected CUG-BP expression vector. (D) hcTNT and hcTNT CUG → CAG minigenes were transiently transfected into the QT35 quail fibroblast cell line (lanes F) and primary chicken embryo skeletal muscle cultures (lanes M) as described (10). Enhanced exon inclusion in chicken muscle cultures requires positive-acting factors that are induced as part of the myogenic program (10). The averaged results from five independent experiments (lanes 1 to 4) are 26 ± 4.6%, 57 ± 3.9%, 11 ± 3.0%, and 23 ± 2.0%. (E) Expression of an expanded repeat reproduces the trans-dominant effect. hcTNT and hcTNT CUG → CAG minigenes were transiently transfected (15) with the DM (3 μg) and tRNAMeti (1 μg) expression vectors containing the number of CTG repeats indicated above each lane. Averaged results from four (DM) and three (tRNAMeti) independent experiments (left to right) are 49 ± 14%, 63 ± 14%, 79 ± 9%, 33 ± 6%, 44 ± 12%, 46 ± 12%, 50 ± 10%, 79 ± 7%, 29 ± 4%, and 23 ± 5%. RT-PCR analysis detected large amounts of DM transcripts in all samples shown.

To determine whether the proximal CUG repeats in the downstream intron were targets for these factors, we introduced 5-nt substitutions into the hcTNT minigene to create hcTNT CUG → CAG (Fig.3A). The CUG → CAG mutation had no effect in normal cells (16) but nearly eliminated the enhanced exon 5 inclusion in DM cells (Fig. 3B). This indicates that the effect on cTNT alternative splicing in DM cells is mediated, at least in part, through the proximal downstream CUG repeats.

To determine whether CUG-BP plays a role in the regulation of cTNT alternative splicing, we cotransfected the hcTNT minigene with a cDNA expression vector encoding human CUG-BP into normal human skeletal muscle cells. The amount of exon 5 inclusion directly correlated with the expression of CUG-BP and the minigene containing the five CUG → CAG substitutions did not respond to CUG-BP overexpression (Fig. 3C). Immunoblot analysis confirmed that the amounts of CUG-BP protein were increased in these experiments (17). We conclude that CUG-BP activates inclusion of the cTNT alternative exon 5 and that this activation requires conserved CUG repeats in the downstream intron.

We next investigated whether the CUG repeats are required for muscle-specific splicing of cTNT exon 5 by testing the hcTNT and hcTNT CUG → CAG minigenes in a transient transfection system (10). Inclusion of cTNT exon 5 was enhanced in differentiated embryonic skeletal muscle cells compared with fibroblasts and the five CUG → CAG mutations nearly eliminated enhanced exon 5 inclusion in muscle cultures (Fig. 3D) (18). These results indicate that the CUGs are required for enhanced splicing of cTNT exon 5 in embryonic striated muscle.

To determine whether CUG-BP binds directly to the CUG repeats, we did ultraviolet (UV) cross-linking experiments with recombinant CUG-BP expressed as a fusion protein with maltose-binding protein (MBP). The cTNT genomic fragment that responded to CUG-BP overexpression (Fig. 3A) was subcloned downstream of the T7 promoter and transcribed in vitro as several uniformly labeled RNAs (Fig.4, RNAs A to H). Recombinant MBP–CUG-BP did not bind to RNA containing the upstream intron and the alternative exon (Fig. 4A, RNA A), but it did bind to the downstream intron (RNA B). Recombinant MBP–CUG-BP binding localized to a 30-nt RNA (Fig. 3A, positions +17 to +46 of intron 5) containing the five CUG repeats (RNA G). Specific rMBP–CUG-BP binding to this segment was eliminated by the CUG → CAG mutations (RNA H). CUG-BP binding to RNA F was inhibited by RNA G but not by the identical RNA fragment H that contained the CUG → CAG substitutions (Fig. 4B). Thus, CUG-BP binds exclusively to the CUG motifs located 19 to 43 nt downstream of the alternative exon.

Figure 4

Binding of CUG-BP to cTNT pre-mRNA. (A) The indicated PCR-generated fragments from the human cTNT gene were transcribed in vitro (27) from the T7 promoter by using guanosine [32P]triphosphate. For UV cross-linking, 800 ng of bacterial rMBP–CUG-BP was preincubated with 3 μl (20 μg of protein) of HeLa cell nuclear extract (28) at 30°C for 10 min (29). The labeled RNA was added with a reaction mixture [containing 3.3 mM MgCl2, 2.5 mM adenosine triphosphate, 25 mM creatine phosphate, and buffer D (28)] to 10 μl and incubated at 30°C for 10 min. Reaction mixtures were placed on an aluminum block prechilled in ice water and UV irradiated 4 cm from a germicidal lamp (Philips G15T8) for 8 min. Samples were digested with ribonuclease (RNase) A (0.5 μg) and RNase T1 (0.5 μg) at 37°C for 20 min, mixed with an equal volume of protein loading buffer, denatured at 90°C, and subjected to electrophoresis on an 11% SDS-polyacrylamide gel. (B) Binding of rMBP–CUG-BP to uniformly labeled RNA F in the presence of increasing amounts of unlabeled G or H RNAs.

To determine whether the trans-dominant effect could be reproduced in normal muscle cells, we cotransfected the hcTNT and hcTNT CUG → CAG minigenes with DM minigenes that contained 40, 480, or 1440 CTG repeats in the 3′ UTR (19). Coexpression of DM minigenes containing 480 and 1440 repeats with the hcTNT minigene resulted in increased exon 5 inclusion (Fig. 3E). Furthermore, the amount of exon 5 inclusion appeared to correlate with the length of the expanded repeat. Splicing of exon 5 was essentially unaffected when the hcTNT minigene lacked the CUG-BP binding site (Fig. 3E). We also inserted CTG repeats into a human tRNAMeti gene that was modified to accumulate tRNA transcripts in the nucleus (20). Coexpression of tRNAs containing 960 repeats promoted exon inclusion for the wild-type but not the CUG → CAG hcTNT minigene mRNAs (Fig. 3E). Thus, transcripts containing at least 480 CUGs have a trans-dominant effect on splicing of cTNT exon 5 and this effect requires the conserved CUG-BP binding site. In addition, the trans-dominant effect is independent of gene context and whether the RNA is transcribed by polymerase (Pol) II or Pol III.

Our results support the hypothesis that nuclear accumulation of CUG-BP in DM cells alters expression of genes that require CUG-BP for posttranscriptional processing. CUG-BP activity is increased in DM cells as assayed by the enhanced inclusion of cTNT exon 5, suggesting that the trans-dominant effect is not simply due to sequestration of CUG-BP on the expanded DM transcripts as originally proposed (3, 7). Increased CUG-BP activity could be due to the increased nuclear concentration of CUG-BP. It remains to be determined whether CUG-BP colocalizes with transcripts from the expanded allele. It is also likely that the phosphorylation state of CUG-BP plays a role in its nuclear-cytoplasmic distribution as well as its intrinsic splicing activity, as demonstrated for other splicing factors (21). Altered ratios of CUG-BP phosphorylated isoforms with different splicing activities could have a strong effect on overall splicing activity.

The hnRNPs such as CUG-BP perform multiple posttranscriptional regulatory functions in the nucleus and cytoplasm (22). Thus, a wide array of aberrant RNA processing phenotypes could result from an overabundance of CUG-BP in the nucleus and a deficiency in the cytoplasm. Because CUG-BP is ubiquitously expressed, its role in muscle-specific splicing is most likely as a constitutive component of a muscle-specific splicing complex. A common requirement for CUG-BP for tissue-specific splicing or other posttranscriptional functions in different tissues could account for the multisystemic nature of DM symptoms.

Overexpression of the fetal cTNT isoform may contribute to the reduced myocardial function and conduction abnormalities observed in DM patients (23). Mutations in the cTNT gene are responsible for dominantly inherited familial cardiomyopathies (24) and small increases in the fetal cTNT isoforms have been described in some cardiomyopathies (13). Because the cTNT mRNA was not detected in normal or DM adult skeletal muscle tissue (17), altered splicing of cTNT in skeletal muscle cultures is an indicator of altered CUG-BP activity rather than a cause of DM pathogenesis in skeletal muscle. Introns that flank alternative exons in other muscle-specific genes also contain CUG repeats (12). We predict that processing of transcripts from one or more of these genes is disrupted in DM and contributes to pathogenesis.

  • * To whom correspondence should be addressed. E-mail: tcooper{at}


View Abstract

Navigate This Article