Report

RNA Leaching of Transcription Factors Disrupts Transcription in Myotonic Dystrophy

See allHide authors and affiliations

Science  16 Jan 2004:
Vol. 303, Issue 5656, pp. 383-387
DOI: 10.1126/science.1088679

Abstract

Myotonic dystrophy type 1 (DM1) is caused by a CUGn expansion (n ≈ 50 to 5000) in the 3′ untranslated region of the mRNA of the DM protein kinase gene. We show that mutant RNA binds and sequesters transcription factors (TFs), with up to 90% depletion of selected TFs from active chromatin. Diverse genes are consequently reduced in expression, including the ion transporter CIC-1, which has been implicated in myotonia. When TF specificity protein 1 (Sp1) was overexpressed in DM1-affected cells, low levels of messenger RNA for CIC-1 were restored to normal. Transcription factor leaching from chromatin by mutant RNA provides a potentially unifying pathomechanistic explanation for this disease.

Myotonic dystrophy type 1 (DM1) is an autosomal dominant disorder linked to a mono-allelic expansion of the CTGn repeat in the 3′ untranslated region of the DM protein kinase gene (DMPK); healthy individuals have repeats of n = 5 to 37, whereas affected individuals have repeats of n = 50 to 5000 (1). The mechanism of DM1 pathogenesis and its multisystem presentation has spawned many hypotheses (28), but a satisfyingly unifying concept has yet to emerge.

We hypothesized that DMPK mutant RNA might exert its deleterious effects through a transcriptional mechanism by direct binding of basic transcription factors (TFs). Because mutant RNA is known not to transport to the cytoplasm but to coalesce into ribonucleoprotein (RNP) foci in the nucleus (9, 10), this association had the potential to divert these factors from their essential transcriptional functions.

If TFs are selectively sequestered by mutant RNA in DM1-affected cells, it should be possible to show mutant RNA but not other RNAs in complex with the affected TFs in vivo (11). As a cell source, we applied the widely used model of MyoD-generated “myocytes” from normal and DM1 subjects, which leads to equivalent muscle-specific DMPK gene induction in control and mutant cells: Control cells express only wild-type DMPK mRNA; DM1 cells express both wild-type and mutant RNAs (10). DM1 cells (CTG100) showed selective DMPK mutant RNA coprecipitation (dual bands; Fig. 1) with TFs Sp1 and retinoic acid receptor gamma (RARγ) and, as a positive control, CUG-binding protein 1 (CUGBP1), for its known affinity for mutant versus wild-type DMPK mRNA in vivo (4). α-Actin and β-actin mRNAs were not detectable in any of the complexes. In contrast, DMPK mutant RNA was not coprecipitated with nuclear pore proteins complex, nuclear pore component NUP153, or platelet-derived growth factor (PDGF) membrane receptor. With control cells, no DMPK wild-type mRNA (single band) was recovered bound to any of the proteins. Equivalent results were obtained with a second DM1 cell line, GM03132 (CTG2000) (12). These data demonstrate that TFs are selectively complexed in vivo by mutant, but not by wild-type, DMPK mRNA or heterologous (α- or β-actin) mRNAs.

Fig. 1.

Nuclear TFs selectively bind mutant mRNA in vivo. (A) Sp1. Nuclear extracts from DM1 and normal cells underwent PCR after immunoprecipitation (“IP”) or not (“No IP”), per nRIP protocol (see Methods). DMPK mutant mRNA in precipitate revealed by dual bands. (B) RARγ. (C) Interpretation of bands. Primer pair 101-102 flanks CUGn. Wild-type DMPK mRNA generates single band [150 nt (⚫)] and mutant DMPK mRNA dual bands [150 nt (⚫) and 450 nt (◯)] of approximately equal intensity (150 > 450), with the smaller band arising by template sliding across expanded CUGn during PCR (17, 30) (fig. S1). PCR efficiency is poorer for mutant than wild-type templates: equimolar mutant and wild-type DMPK in DM1 cellular DNA [lane 9 in (A)] shows strong dominance of low-molecular-mass PCR band (⚫) (150 ≫ 450), also evident in “No IP” DM1 mRNA [lane 7 in (A)]. Primer pair 8DM-9DM spans exons 14 to 15 with same size (△) and amplification efficiency for mutant and wild-type RNAs. Actin control [♦ in (B)] was not present in any IP.

To address the core mechanistic element of this hypothesis, i.e., that TFs are depleted from their sites of action in mutant RNA–expressing cells, we examined whether mutant RNA binding of TFs correlated with a disturbance to their normal distribution among nuclear compartments. For TFs, we studied representatives of three classes, associated nominally with cell maintenance (Sp family, Sp1 and Sp3); activation (signal transducer and activator of transcription family, STAT1 and STAT3); and differentiation (RARγ). For 4 weeks after MyoD induction, control cells maintained their stable distribution, with RARγ residing mainly in chromatin instead of RNP (Fig. 2, A and B). In contrast, over the same period, DM1-affected cells showed a progressive decline in the ratio of TF in chromatin versus that in RNP, such that RNP became the dominant site. At 4.5 weeks, all four DM1-affected lines showed a similar, pronounced fivefold redistribution of RARγ toward the RNP (Fig. 2, C and D, top).

Fig. 2.

Nuclear compartment distribution of TFs is altered in DM1-affected cells. (A) Redistribution of RARγ over time. Cells were induced for DMPK expression and assayed at times specified for RARγ. Derived from Western blot data of (B) as ratio of band intensities for chromatin- and RNP-associated TF. (A) Normal (NHDF adult), ♦; DM1 (GM04033), ◼. (C) Redistribution of RARγ in four different DM1 lines at 4.5 weeks (GM03991, GM03987, GM04033, and GM03132). Derived from Western blot data of (D, top) as normalized chromatin/RNP RARγ ratio in DM1 versus normal counterpart's ratio, assigned as 100%. (D, bottom). Chromatin-RNP fractions for Sp- and STAT-family TFs at 3 weeks of culture. (E) Reduction of RARγ in DM1 chromatin during cultivation. Normal (NHDF adult), DM1 (GM04033). (F) At a fixed time (4.5 weeks), TF is lower in DM1 chromatin (GM03987 and GM03991) than in normal (CCD-986SK and CCD-919SK). (G) Chromatin and RNP loading. Coomassie staining of chromatin and RNP from normal (NHDF adult) and DM1 (GM04033). Top arrows, deoxyribonuclease I (DNase I) to solubilize chromatin samples; bottom arrows, ribonuclease (RNase) to solubilize RNP samples. Brackets mark histones H1 (top) and H3, H2B, H2A, and H4 (bottom). Equivalent histone bands prove uniformity of chromatin extractions.

The four other TFs from the Sp and STAT families (Fig. 2D, bottom) were all nearly absent from RNP before mutant RNA induction, which precluded a derived baseline chromatin/RNP ratio (>10). After 3 weeks of mutant RNA expression, however, the measured fractions of TF in chromatin in the DM1 cells shifted dramatically for Sp1, STAT1, and STAT3, with levels of 30, 14, and 32%, respectively, versus ∼100% in chromatin for control. Sp3 was only minimally affected; 75% of it remained in chromatin.

Did the “redistribution” ratio data in Fig. 2, A to D, represent an actual depletion from chromatin or merely an accumulation in the RNP and protection of unbound TF that would normally have been turned over? Chromatin isolations were repeated under conditions controlled for equivalent absolute numbers of viable myocytes. Stable levels of TF in chromatin were seen in normal cells at 4 weeks as compared with 0 weeks of DMPK expression, but a marked drop in DM1-affected cells was apparent over the same period (Fig. 2E). Two additional DM1 cell lines similarly showed markedly lower RARγ levels in chromatin after 4.5 weeks of mutant RNA expression (Fig. 2F). The net amount of RARγ remaining in chromatin was 9, 15, and 11% of control after mutant RNA induction. Hence, these tests demonstrate an authentic leaching of TFs out of the chromatin and into the RNP fraction that is specifically mediated in mutant RNA–expressing DM1 cells. However, these measured depletions of TF from chromatin are even more severe than the maximum changes of three- to fivefold expected from the redistribution model (Fig. 2C).

Loss of TFs from chromatin in DM1 means loss from their sites of action. We therefore examined whether gene expression was broadly affected in concert with the TF depletion. Whereas TFs are themselves dependent on TFs for their own transcription (13), the same TFs could also be subject to transcriptional depression. Northern blotting of cellular RNA 4.5 weeks after DMPK induction revealed the following: a moderate reduction of glyceraldehyde-3-phosphate dehydrogenase (GAPDH) mRNA (∼35% decrease) in DM1 versus normal cells; no change in FCGRT (Fc IgG receptor for transport, Brambell receptor) mRNA (∼5% increase) or in α-actin or β-actin mRNAs; and significantly lower mRNA levels for TFs Sp1, Sp3, and RARγ (∼50% reduction) and lesser reduction of Sp2 (∼20%) (Fig. 3A).

Fig. 3.

Reduced mRNA levels of selected genes in DM1-affected cells. (A) Northern blot analysis. mRNA of specific genes before and after DMPK gene induction. (B) Relative mRNA levels in normal and DM1. Graphed values calculated as mean of four independent cell line pairs by TaqMan (DM1/normal) ± SD, normalized to 18S per extract, except Sp2, which was estimated by Northern blot phosphorescence imaging only. (C) CLCN1 mRNA in DM1 cells after Sp-family overexpression. Baseline mRNA relative to paired normal was 32% for CLCN1 and 115% for FCGRT. (D) TaqMan gene-specific amplification plots such as used for (B). (E) TaqMan analysis of mRNA in DM1-affected cells after Sp family overexpression. MyoD-transduced GM04033 cells at 4 weeks were untransfected or transfected with plasmids for Sp1 or Sp3, with elevated TF expressions by TaqMan (12). Changes in (E) are graphed relative to baseline (Affected) after Sp1 and Sp3 overexpression as shown in (C).

As corroboration, multiplex reverse transcription polymerase chain reaction (RT-PCR) was conducted to quantify RNA levels (Fig. 3B). Before DMPK expression (–MyoD), DM1 and control cells began with equal levels for all RNAs. After DMPK induction (+MyoD), changes were pronounced for GAPDH, CLCN1, Sp1, Sp3, and RARγ, with average mRNA suppression in the four DM1 mutant RNA–expressing cell lines of –41, –70, –46, –50, and –44% relative to wild-type cells, respectively. No significant change was observed in FCGRT, α-actin, or β-actin gene expression or in 18S RNA (Fig. 3, B and C). All values were close to the corresponding numbers estimated by Northern blot phosphorescence imaging (Fig. 3A).

Hence, maldistribution (leaching) of TFs can lead to reduced expression of selected genes, including that of TFs themselves. When TF transcription is also reduced, it compounds the impact of mutant RNA leaching to deplete TFs from chromatin. This accords with our RARγ results that showed greater net suppressions of TF in chromatin (Fig. 2, E and F) than that predicted by redistribution alone (Fig. 2C). If mutant RNA–induced redistributions by factors of 2 to 5 and reductions in mRNA by of as much as a factor of 2 for selected TFs are considered nominal, then overall reductions by a factor of 5 to 10 (to 10 to 20% of normal) in chromatin may be commonplace among TFs in DM1 setting.

We examined CLCN1, the gene encoding CIC-1, the skeletal muscle chloride channel whose disruption has been implicated in DM1 myotonia (6, 7). The impact of DM1 on CLCN1 mRNA levels in humans has been controversial (7, 14). In muscle of long-repeat (CTG250) transgenic mice, reductions by factors of 2.5 to 3 (to 31 to 43% of normal) were observed (6). Our own tests of CLCN1 mRNA in DM1-affected cells showed suppression in transcript levels by factors of 3 to 6 across four cell lines (to 15 to 35% of normal) and support a common phenomenon in human and mouse models.

A search for TF binding motifs within the 300-nt CLCN1 promoter region (15) revealed 18 sites with similarity higher than 0.7 for Sp1 consensus, in addition to the three putative E-box sites for TFs of the helix-loop-helix (HLH) family, class I: E47 and class II: MyoD, which bind as heterodimers (16). We confirmed (17) indicated the Sp1-binding activity of such sites and their structural compatibility for assembly into canonical transcription initiation complexes (supporting on-line materials; fig. S2). Because Sp1 is one of the TFs most affected by mutant RNA binding, Sp1 depletion could be a proximate cause of depressed CLCN1 transcription in DM1-affected cells. Therefore, we tested whether “replenishment” of Sp1 in the same cells could reverse this finding.

Under conditions of induced DMPK, the GM0433 cell line displayed a factor of 3 depression (–68%) of CLCN1 mRNA, to 32% of control. When these affected cells were transduced with a high-expressing Sp1 plasmid to restore Sp1 “sufficiency,” the depressed CLCN1 mRNA was increased 2.8-fold (+280%), which restored it to normal levels (Fig. 3, C and E). A nonspecific effect of Sp1 overexpression on all genes was ruled out by lack of impact on negative control gene FCGRT in the same cells, which was not depressed in DM1-affected cells. Overexpression of Sp3, the TF that was most resistant to mutant RNA–induced redistribution in vivo, had no effect on CLCN1 or on FCGRT mRNA levels in mutant RNA–expressing cells. These data are compatible with an important role for Sp1 in CLCN1 regulation and, correspondingly, support the potential of diverse TF depletions in DM1-affected cells to depress transcription of CLCN1 and other genes in vivo.

CLCN1 mRNA has been shown in DM1 to exhibit an increase in aberrant splicing of pre-mRNA that serves to further depress levels of functional CIC-1 protein beyond that due to low total mRNA (7). Reduced total CLCN1 mRNA levels were initially proposed to result from nonsense-mediated decay (NMD) to rapidly clear misspliced CLCN1 transcripts (6, 7), although a role for NMD was not directly demonstrated [e.g., in the manner of (18)]. In light of our data, it may alternately be suggested that low CLCN1 mRNA levels in DM1 may derive from depressed CLCN1 transcription as a dominant primary effect, on which splicing disruptions may secondarily be imposed.

Our results support the hypothesis that mutant RNA binds and sequesters nuclear TFs in DM1 myocytes and leads to disrupted gene expression patterns. This binding induces a redistribution of TFs away from active chromatin and toward an RNP fraction of the nuclear matrix for four out of five general transcription factors. Inasmuch as nuclear structural and cytoplasmic membrane proteins were not similarly bound by mutant RNA, the basis for this interaction between TF and mutant RNA appears to reside in the proteins' normal capacity for nucleic acid (double-stranded DNA) binding, possibly with analogous double-stranded regions in the mutant RNA itself (19). However, the TF–mutant RNA binding lacks particular sequence specificity that is the hallmark of the TF-DNA interactions.

We also show that maldistribution of TFs can lead to reduction in total TF transcription, the combined effects of which can be a net factor of 5 to 10 (80 to 90%) depletion of select chromatin-associated TFs. These data thus present a scenario with the potential for massive derangement of transcription in affected cells. Mutant RNA–induced leaching of TFs from chromatin led to significant (>30%) reductions in mRNA expression (20) in several of the genes examined. However, the disruption was not general: An equal number maintained normal or near-normal transcript levels under the same conditions. With the addition of seven further genes recently examined (21), the fraction of genes with significantly reduced mRNA levels is 8/16 or 50% (95% confidence limits: 27 to 73%).

This effect on TFs in DM1 suggests comparisons with Huntington's disease (HD) in which huntingtin selectively binds and disrupts Sp1 (22, 23). However, this protein-protein interaction in HD is more restricted than the RNA-protein interaction in DM1, which sequesters many TFs concurrently. Correspondingly, the extent of derangement of gene transcription in HD models (estimated at ∼2%) (24) is far less than seen in our DM1 sample (50%).

A common requirement for basic transcription factors in different tissues that express DMPK could account for the multisystemic and multisymptomic nature of DM1 findings (e.g., myotonia, myopathy, diabetes, testicular atrophy). Notably, the clinical analogies between DM1 and DM2, in which the latter has CCUG mRNA expansions (25), may conceivably derive from a common trans-acting mechanism with trapping and depletion of similar or identical TFs.

Other examples of transcriptional dysregulation in genetic disorders are the trichothiodystrophies, a group of autosomal recessive diseases (26), and Cockayne syndrome (27). All result from phenotype-specific mutations in genes encoding TFIIH, a multicomponent complex that has been implicated in transcription and DNA excision repair. Whereas repair defects explain the skin photosensitivity seen in a subset of patients, other disease phenotypes have been attributed to transcriptional defects (26). Deletion of the variable-number tandem-repeat structure that binds transcription repressor YY1 leads to transcriptional derepression in the respective chromosomal band, to result in facioscapulo-humeral muscular dystrophy (28). This short list of genetic disorder transcription syndromes (29) can now perhaps be extended.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1088679/DC1

SOM Text

Figs. S1 and S2

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article