Expanded CUG Repeats Dysregulate RNA Splicing by Altering the Stoichiometry of the Muscleblind 1 Complex*

To understand the role of the splice regulator muscleblind 1 (MBNL1) in the development of RNA splice defects in myotonic dystrophy I (DM1), we purified RNA-independent MBNL1 complexes from normal human myoblasts and examined the behavior of these complexes in DM1 myoblasts. Antibodies recognizing MBNL1 variants (MBNL1CUG), which can sequester in the toxic CUG RNA foci that develop in DM1 nuclei, were used to purify MBNL1CUG complexes from normal myoblasts. In normal myoblasts, MBNL1CUG bind 10 proteins involved in remodeling ribonucleoprotein complexes including hnRNP H, H2, H3, F, A2/B1, K, L, DDX5, DDX17, and DHX9. Of these proteins, only MBNL1CUG colocalizes extensively with DM1 CUG foci (>80% of foci) with its partners being present in <10% of foci. Importantly, the stoichiometry of MBNL1CUG complexes is altered in DM1 myoblasts, demonstrating an increase in the steady state levels of nine of its partner proteins. These changes are recapitulated by the expression of expanded CUG repeat RNA in Cos7 cells. Altered stoichiometry of MBNL1CUG complexes results from aberrant protein synthesis or stability and is unlinked to PKCα function. Modeling these changes in normal myoblasts demonstrates that increased levels of hnRNP H, H2, H3, F, and DDX5 independently dysregulate splicing in overlapping RNA subsets. Thus expression of expanded CUG repeats alters the stoichiometry of MBNL1CUG complexes to allow both the reinforcement and expansion of RNA processing defects.

Myotonic dystrophy I (DM1) 4 is a multi-system disorder resulting from the expansion of a CTG repeat sequence in the 3Ј-untranslated region of DMPK, located on chromosome 19 (1)(2)(3)(4). A key molecular consequence of expanded CUG repeat expression is the abnormal splicing of physiologically important RNAs. Specifically, abnormal splicing of the chloride channel (ClC-1) and the insulin receptor (IR) RNAs have been implicated in the development of myotonia and insulin resistance, respectively, in DM1 patients (5)(6)(7)(8). In DM1 cells, expression of expanded CUG repeats results in the formation of aberrant nuclear CUG RNA aggregates or foci (9). The significance of nuclear CUG aggregate formation in the development of aberrant splicing and DM1 pathology is reflected in a recent study, which demonstrates that antisense RNA-mediated silencing of the mutant DMPK RNA and the consequent decrease of CUG foci allow the correction of splice defects in DM1 animal models (10). Significantly, CUG focus formation requires the binding of the alternative splice factor, muscleblind 1 (MBNL1) to the expanded CUG repeat expansions (11,12). The importance of the aberrant MBNL1-CUG interaction in DM1 is underscored by the use of pentamidine and morpholino antisense oligonucleotides that dislodge MBNL1 from expanded CUG RNA, restore free MBNL1 levels, and rescue splice defects in DM1 mouse models (13,14). Consistent with these observations, inactivation of Mbnl1 in mice recapitulates a large fraction of the splice defects and several key features of DM1 pathology (15,16). Thus based on these data we hypothesized that examination of the behavior of MBNL1 proteins that can sequester in CUG foci in DM1 cells should yield key insights into the development of DM1 splice defects. MBNL1, codes for four zinc finger motifs that play an important role in RNA target identification (11,17). Specifically, MBNL1 has been shown to recognize and bind to the stem of hairpin structures that form in both toxic expanded CUG RNAs and in normal splicing targets (18). Although nine MBNL1 splice variants have been described, only a subset of MBNL1 variants that encode both pairs of zinc finger motifs separated by an intervening linker sequence bind to expanded CUG repeat sequences in a yeast three-hybrid assay system (17). Thus to biochemically isolate and study MBNL1 variants that can sequester in CUG foci, we utilized MB1a monoclonal antibodies (MB1a mAbs) that recognize the linker sequence found between the two pairs of zinc finger motifs (19,20), to purify RNA-independent MBNL1 CUG protein complexes from normal human myoblasts. Examination of the behavior of these complexes demonstrates that expression of expanded CUG repeats alters MBNL1 CUG complex stoichiometry resulting in elevated steady state levels of MBNL1 CTG partners, which serve to both independently reinforce splicing abnormalities in overlapping sets of RNA targets and to potentially increase the number of RNA processing defects in DM1.

Cell Culture
One normal human myoblast culture and two DM1 myoblast cultures were a gift from Dr. Charles Thornton (University of Rochester Medical Center, Rochester, NY). The second normal human myoblast culture (skeletal muscle cells (SkMC); catalogue number CC-2661) was purchased from Lonza Inc. Normal and DM1 myoblasts were immortalized by infection with the SV40 virus. Detailed characterization of these cell lines have been previously described (21). Briefly, these DM1 myoblast lines have CTG tracts of ϳ8 kb and show CUG foci and aberrant splicing. Fibroblast contamination in DM1 and normal myoblasts lines is minimal (Յ5%). Myoblast cultures were maintained in SkGM medium (Lonza Inc.; catalogue number 3160) containing 10% fetal bovine serum. Cos7 cells were maintained in DMEM containing 10% fetal bovine serum. siRNAs siRNA oligonucleotides were synthesized by Dharmacon Inc. The oligonucleotides were deprotected, and the complementary strands were annealed. The sequences of the siRNAs used in this study are: scrambled siRNA, 5Ј-GCGCGCUUU-GUAGGAUUCGdTdT-3Ј; MBNL1, 5Ј-CACUGGAAGUAU-GUAGAGAdTdT-3Ј (22); and PKC␣, 5Ј-AAAGGCUGAG-GUUGCUGAUdTdT-3Ј (23).
siRNA-mediated Depletion of Proteins-SkMC (1 ϫ 10 6 ) were plated on 100-mm dishes overnight, and siRNAs (100 nM) were transfected using Oligofectamine (Invitrogen) according to the manufacturer's protocol. Prior standardization experiments showed that maximum silencing is achieved 4 -5 days post-transfection in these cells.

Whole Cell Extracts and Subcellular Fraction Preparations
To determine steady state protein levels, equal numbers of normal and DM1 myoblasts were plated and sampled 24 h after plating. The cells were harvested and suspended in SDS-PAGE sample buffer (Laemmli sample buffer; Bio-Rad; catalogue number 161-0737) and then boiled for 5 min. Equal volumes of the extracts were subjected to Western blot analysis to determine the steady state levels of proteins using the antibodies listed in supplemental Table S1. For Cos7 cells expressing exogenous CTG repeats, the harvested cells were lysed as described above and used for Western blot analyses to determine steady state protein levels. Protein levels were measured by densitometry using GAPDH levels as a loading control. Mass spectrometry analysis was carried out at the Scripps Center for Mass Spectrometry.
Whole cell extracts were prepared using a two-step lysis procedure: first, the cells were lysed with a cytoplasmic extraction buffer (25 mM Tris-HCl, pH 7.6, 10 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, 2.0 mM MgCl 2 , and protease inhibitors; Sigma; catalogue number P-8340). Cytoplasmic extracts were separated by centrifugation at 14,000 rpm for 20 min. Second, the resultant pellets were suspended in nuclear lysis buffer (25 mM Tris-HCl, pH 7.6, 500 mM NaCl, 0.5% Nonidet P-40, 2 mM EDTA, 2 mM MgCl 2 , and protease inhibitors), and the nuclear extracts were separated by centrifugation at 14,000 rpm for 20 min. The nuclear extracts were combined with the cytoplasmic extracts, and the resultant extracts were denoted as whole cell extracts. For subcellular fractions, cytoplasmic extracts and nuclear extracts were analyzed separately. Protein concentrations were measured using the Bio-Rad protein assay (catalogue number 500-0006) with BSA as a standard. Equal amounts of protein were subjected to Western blot analysis to determine protein levels. The antibodies used are listed in supplemental Table S1. To ensure that the signals in Western blot analysis were not saturated, prior standardization experiments were carried out as described in Ref. 22.

Immunoprecipitations
Normal myoblasts were cultured and harvested followed by washing with cold PBS. Whole cell extracts were prepared using the two-step lysis procedure described above. Specifically, the nuclear extracts were combined with the cytoplasmic extracts and denoted as whole cell extracts. NaCl concentration was adjusted to 150 mM at this point. Whole cell extracts were treated with 1.0 mg/ml RNase A (Amersham Biosciences) for 15 min at 37°C and incubated with MB1a mAb immobilized beads for 3 h. After several washes with a washing buffer (10 mM Tris-HCl, pH 7.6, 10 mM HEPES, 200 mM NaCl, 2.0 mM MgCl 2 , 0.5 mM EDTA, 0.05% Nonidet P-40, 10% glycerol, 1.0 mM DTT, and 0.1 mM PMSF), the bound proteins were eluted from the beads with an elution buffer (10 mM Tris-HCl, pH 7.6, 1.0 M NaCl, 1.0 mM EDTA, 5% glycerol, 1.0 mM DTT, and 0.1 mM PMSF). The eluted proteins were analyzed by SDS-PAGE followed by silver staining and Western blots where indicated.

Western Blot Analyses
Western blot analysis was carried out as described in Ref. 22. Primary and secondary antibodies used are described in supplemental Table S1. The relative band intensities were measured by densitometry analyses using Gene Tool (Syngene Inc.).

PCR Analyses of the Steady State Expression Levels of RNA and Alternative RNA Splicing
RT-PCR and real time PCR analysis were performed as described in Ref. 24. Where indicated, real time RT-PCR was performed using cDNAs (5 ng) derived from two normal and two DM1 myoblasts, to measure the RNA steady state expression levels of MBNL1 CUG -interacting proteins. The primer sequences used for the amplification of RNAs in these analyses are listed in supplemental Table S2. To study alternative RNA splicing, cDNA (150 ng) derived from total RNA was subjected to PCR analysis. The primer sequences and conditions used for the amplification of MBNL1, MBNL2, IR, cTNT, Zasp, FN1, and GAPDH RNAs are listed in supplemental Table S3. In all cases, the PCR products were cloned and verified by sequencing. The relative band intensities were measured by densitometry analyses using Gene Tool (Syngene Inc.). To ensure that the signals in the PCR analyses were not saturated, prior standardization experiments were carried out as described in Ref. 22. The percentage of exon inclusion (or exclusion) was calculated as [exon inclusion (or exclusion)/(exon inclusion ϩ exon exclusion)] ϫ 100.

Fluorescence in Situ Hybridization (FISH)
Detection of CUG transcripts was carried out primarily as described (12,22). Briefly, DM1 myoblasts were plated on chamber slides overnight and then fixed in 4% paraformaldehyde/PBS for 20 min at room temperature. The cells were then permeabilized with 70% ethanol and stored at 4°C. In other experiments, the cells were cultured on chamber slides and transiently transfected with plasmids containing CTG repeats. For FISH studies, a Cy3-conjugated (CAG) 10 oligonucleotide probe (IDT Inc.) was used to detect CUG repeat expansions, and the nuclei were stained with DAPI (Sigma). After hybrid-ization, the slides were mounted with mounting medium, Vectashield (Vector Inc.), and the cells were visualized using Nikon Eclipse E600 or LSM 510 confocal microscopy.

FISH/Immunofluorescence
Myoblasts were plated on chamber slides and transfected with GFP-tagged constructs. Subsequently, the transfected cells were fixed in 4% paraformaldehyde/PBS for 20 min at room temperature. The cells were permeabilized with 70% ethanol and stored at 4°C. For FISH studies, a Cy3-conjugated (CAG) 10 oligonucleotide probe was used to detect CUG repeat expansions as described above. To detect endogenous proteins and examine whether they demonstrate colocalization with CUG foci, the cells were blocked with 10% BSA/PBS for 2 h at room temperature. The slides were incubated with primary antibodies (supplemental Table S1) in 10% BSA/PBS for 2 h at room temperature or overnight at 4°C, washed three times with PBS, and incubated with the corresponding secondary antibodies conjugated with Alexa Fluor 488 (supplemental Table S1) in 10% BSA/PBS at room temperature for 2 h. Following incubation, the cells were washed three times with PBS, and the nuclei were stained with DAPI. The slides were mounted with mounting medium, and the cells were visualized as noted above.

MB1a mAbs Recognize and Immunoprecipitate Endogenous MBNL1 Variants That Can Colocalize with CUG Foci (MBNL1 CUG ) in DM1 Myoblasts-To assess whether MB1a
antibodies specifically identify MBNL1, we measured the endogenous levels of MBNL1 from cell extracts of normal human myoblasts (SkMC) treated with scrambled siRNAs or in which MBNL1 was depleted by the use of cognate siRNAs (Fig.  1A). Subsequently, MB1a mAbs were tested for their ability to immunoprecipitate MBNL1 from human myoblasts. MB1a mAb, but not preimmune IgG, allowed precipitation of MBNL1 from human myoblasts (Fig. 1B). On immunostaining, MB1a mAbs detect MBNL1 variants that colocalize with CUG foci (Fig. 1C). Thus these data demonstrate that MB1a mAbs detect and allow immunopurification of MBNL1 variants that can colocalize with CUG foci (MBNL1 CUG ) in DM1 patient cells.

MBNL1 CUG Variants Associate in an RNA-independent Manner with 10 Proteins Involved in RNA Metabolism in Normal Human
Myoblasts-Proteins that interact in an RNA-independent manner with MBNL1 CUG variants were purified using MB1a mAb immunoprecipitation from SkMC cell extracts treated with RNase A. RNase A-treated whole cell extracts were incubated with MB1a mAb immobilized on beads, and the immunoprecipitates were washed with a buffer containing 200 mM NaCl. The bound proteins were eluted from the beads with a buffer containing 1 M NaCl and examined by SDS-PAGE followed by silver staining. Immunoprecipitation experiments were carried out in triplicate, and a representative panel is shown ( Fig. 2A). Western blot analyses of the eluted proteins show coimmunoprecipitation of hnRNP H, which we have previously shown to interact with MBNL1 in an RNAindependent manner, but not of hnRNP A1, a protein that does not interact with MBNL1 in vivo (data not shown and Ref. 22). NOVEMBER 4, 2011 • VOLUME 286 • NUMBER 44

JOURNAL OF BIOLOGICAL CHEMISTRY 38429
To determine the identity of the proteins that coimmunoprecipitate with MBNL1 CUG , the eluted proteins were subjected to tryptic digestion, and the mixture of peptides was separated by liquid chromatography and analyzed by tandem mass spectrometry. Mass spectrometry analyses were carried out in triplicate using elutes from three independent immunoprecipitations. Subsequent data base searches revealed the identity of several hnRNP proteins including hnRNP H, H2, H3, F, A2/B1, K, and L and three RNA helicases, DDX5, DDX17, and DHX9.
To test whether the proteins identified by mass spectrometry are bona fide MBNL1 CUG interactors, we carried out Western blot analysis of MBNL1 CUG immunoprecipitates to validate these interactions. RNase A-treated SkMC whole cell extracts were subjected to MB1a mAb immunoprecipitation, and after washing, the bound proteins were eluted with a high salt buffer. The MB1a mAb immunoprecipitates were analyzed by SDS-PAGE followed by Western blot analysis using antibodies directed against the putative MBNL1 CUG -interacting proteins. These data demonstrate coprecipitation of all identified MBNL1 CUG -interacting proteins. All of the interactions were validated at least three times, and a representative panel is shown (Fig. 2B). Next, we examined whether MBNL1 CUG interactions could be confirmed by using anti-FLAG antibodies. SkMC were transduced with recombinant adenoviruses expressing a FLAG-MBNL1 CUG variant that encodes both pairs of zinc fingers and the intervening linker. Subsequently, cell extracts were immunoprecipitated using anti-FLAG antibodies. FLAG-MBNL1 CUG showed coimmunoprecipitation of the three MBNL1 CUG interactors tested (Fig. 2C). These data demonstrate that MBNL1 CUG -interacting proteins identified by mass spectrometry are authentic interactions and do not reflect nonspecific interactions with the MB1a mAb.
To examine the relative strength of these interactions, RNase A-treated SkMC whole cell extracts were subjected to MB1a mAb immunoprecipitation, and the immunoprecipitates were washed with wash buffers containing 200, 400, or 600 mM NaCl. The bound proteins were eluted with a high salt buffer and analyzed by SDS-PAGE followed by silver staining. Silver stain-FIGURE 1. MB1a mAb detect and immunoprecipitate MBNL1 CUG variants from SkMC. A, SkMC cells were transfected with scrambled siRNAs or siRNAs directed against MBNL1, and 5 days post-transfection, the protein extracts were subjected to Western blot analysis for MBNL1 using MB1a mAb. B, SkMC whole cell extracts (Ext.) were subjected to MB1a mAb immunoprecipitation (IP), and the products were analyzed by Western blot using MB1a mAbs. C, MB1a mAbs detect MBNL1 variants that can sequester in CUG foci (MBNL1 CUG ). DM1 myoblasts were transfected with scrambled siRNAs or siRNAs directed against MBNL1 and immunostained for MBNL1 using Mb1a mAb. Endogenous MBNL1 stained with Mb1a mAb is visualized as a green signal (panels a, e, and i). CUG RNA foci were detected by FISH (red signal; panels b, f, and j). Merged images demonstrate that Mb1a mAbs detect MBNL1 variants that can sequester in CUG foci (MBNL1 CUG ) (panels d, h, and l).
ing of the eluted proteins demonstrates that subsets of the interactions are retained under stringent wash conditions of 400 and 600 mM NaCl (supplemental Fig. S1). Thus a significant number of MBNL1 CUG -interacting proteins demonstrate strong interactions with MBNL1 CUG in vivo. DM1 Myoblasts Show Altered Stoichiometry of MBNL1 CUG Complexes and Exhibit an Increase in the Steady State Levels of MBNL1 CUG -interacting Proteins-To examine the behavior of MBNL1 CUG complexes in DM1 myoblasts encoding CTG tracts of ϳ8 kb, we measured the steady state protein levels of hnRNP H, hnRNP H2, hnRNP H3, hnRNP F, hnRNP A2/B1, hnRNP K, hnRNP L, DDX5, DDX17, and DHX9, which interact with MBNL1 CUG in an RNA-independent manner (Fig. 3, A and B), and CUG-BP1, which interacts in an RNA-dependent fashion with MBNL1 (22,25) in two normal and two DM1 myoblast lines. As previously noted (21), these four myoblast lines have similar and low levels of fibroblast contamination of Յ5%. Therefore these lines can be directly compared. Because differential solubility of the CUG foci can confound the measurement of steady state levels of these proteins, we solubilized proteins directly from the cells by using SDS-PAGE sample buffer followed by Western blot analyses to measure the steady state levels of the protein components of MBNL1 CUG complexes. The steady state levels of all MBNL1 CUG protein interactors identified, with the exception of hnRNP L, were elevated on average 2-3.5-fold in the two DM1 myoblast lines when compared with the two normal myoblast lines (Fig. 3, A and  B). A similar increase (ϳ3.5-fold) in CUG-BP1 levels was observed in the two DM1 myoblast lines studied when compared with the normal myoblast lines. Levels of other RNA regulatory proteins, including hnRNP A1 (Fig. 3, A and B), SF2/ASF, and polyadenylate binding protein nuclear 1 (PABPN1), which do not interact with MBNL1 (22), were not significantly different in normal and DM1 myoblasts (data not shown). Similar results were also obtained when nuclear extracts were independently studied, with the exception of DDX17, which demonstrated elevated levels in DM1 cytoplasmic but not nuclear extracts when compared with controls (supplemental Fig. S2).

MBNL1 CUG Protein Partners
Colocalize with Ͻ10% of CUG RNA Foci-Because MBNL1 plays a key role in CUG aggregate formation in DM1 cells (12), we tested whether MBNL1 CUG interactors colocalize with DM1 foci in DM1 myoblast line 1. Localization of MBNL1 CUG and each of the MBNL1 CUG partners with mutant DMPK RNA in DM1 cells was studied by FISH using (CAG) 10 -Cy3 probes to locate and identify CUG foci in conjunction with immunofluorescence using antibodies against the endogenous proteins. As previously described, on average ϳ80% of CUG foci colocalized with MBNL1 CUG in DM1 myoblast line 1 (Ref. 24, Table 1, and supplemental Fig.  S3). In contrast, immunostaining of each of the MBNL1 CUG interactors by antibodies against the endogenous proteins demonstrates that MBNL1 CUG -interacting proteins colocalize with Ͻ10% of the foci ( Table 1 and supplemental Fig. S3). To further confirm this observation, we studied the cellular localization of GFP-tagged versions of both a MBNL1 CUG variant and MBNL1 CUG -interacting proteins in DM1 myoblasts (images are not shown). A GFP-MBNL1 CUG variant, encoding both pairs of zinc fingers and the linker region, colocalized strongly with the CUG foci (83%) ( Table 1); however, colocalization of the GFP-tagged MBNL1 CUG -interacting proteins with the CUG foci was significantly lower (3-7%) in DM1 myoblasts (Table 1). Thus these data demonstrate that MBNL1 CUG are the primary proteins that are recruited to CUG foci and that the MBNL1 CUG protein partners localize with CUG aggregates to a much lower extent. These data are therefore consistent with an increase in the unsequestered levels of MBNL1 CUG partners in DM1 myoblasts.

Alteration of MBNL1 CUG Complex Stoichiometry Is Not a Consequence of Altered Steady State RNA Levels or PKC␣
Activation-To test whether the elevated levels of the MBNL1 CUG interactors were consequences of altered steady state RNA levels, we measured the RNA levels of MBNL1 CUG protein interactors by real time PCR analyses. The steady state RNA levels of MBNL1 CUG partners did not correspond to the pro-FIGURE 3. MBNL1 CUG complexes show altered stoichiometry in DM1 myoblasts. A, protein extracts from two normal and two DM1 myoblast lines were subjected to Western blot analysis to measure the steady state levels of proteins identified in MBNL1 CUG complexes using the indicated antibodies. The blots were reprobed for GAPDH as an internal loading control. B, quantitation of the steady state levels of the MBNL1 CUG partner proteins in normal and DM1 myoblasts is tabulated. tein levels observed in the two DM1 myoblast lines when compared with the two normal myoblast lines (supplemental Fig. S4). Therefore MBNL1 CUG complex stoichiometry is altered at the translational or post-translational level in DM1 myoblasts.
As aberrant activation of PKC␣ has been hypothesized to play a key role in increasing the steady state levels of CUG-BP1, a protein that interacts with MBNL1 in an RNA-dependent manner (22), in DM1 (26), we measured the steady state levels of total PKC␣ and phospho-PKC␣ in Cos7 cells expressing DMPK 11-15(CTG) 5 or 11-15(CTG) 300(i) constructs (Fig. 4, A  and B) and in two normal and two DM1 myoblast lines (Fig.  4C). Consistent with previous observations (26), the steady state levels of phospho-PKC␣ were elevated ϳ4.8-fold in Cos7 cells expressing expanded CUG repeats and ϳ3.7and ϳ2.9fold in the two DM1 myoblast lines when compared with the normal myoblast lines (Fig. 4, B and C). To test whether PKC␣ function is required in DM1 myoblasts to maintain altered MBNL1 CUG complex stoichiometry, we functionally inactivated PKC␣ using the cognate siRNAs in DM1 myoblasts. Inactivation of PKC␣ did not reverse the altered stoichiometry of MBNL1 CUG complexes or the elevated levels of CUG-BP1 and did not rescue the splice defects (Fig. 4, D-F).
Altered Stoichiometry of the MBNL1 CUG Complex Is Directly Linked to the Expression of   300(i)

in Cos7
Cells-To test whether the alteration of MBNL1 CUG complex stoichiometry is directly linked to the expression of expanded CUG repeat RNA, we measured the steady state levels of MBNL1 CUG interactors in cell extracts derived from Cos7 cells transiently (48 h) transfected with DMPK 11-15(CTG) 5 or DMPK 11-15(CTG) 300(i) . Western blot analysis revealed that expression of DMPK 11-15(CTG) 300(i) repeats in Cos7 cells was sufficient to recapitulate the elevated steady state levels of MBNL1 CUG -interacting proteins when compared with controls (Fig. 5). As previously reported, expression of expanded CUG repeats also resulted in elevated CUG-BP1 levels in Cos7 cells (26). Thus these data demonstrate that the alteration of the steady state levels of MBNL1 CUG -interacting proteins results as a direct consequence of expanded CUG repeat expression.
Elevated Levels of MBNL1 CUG -interacting Proteins Result in Aberrant Splicing of Overlapping RNA Targets-To test the contributory role of elevated steady state levels of MBNL1 CUG interactors in the development of the aberrant RNA splice patterns observed in DM1, we tested the effect of increased steady state levels of each MBNL1 CUG -interacting protein in normal human myoblasts on six sample RNAs: IR, cTNT, Zasp, MBNL1, MBNL2, and FN1, which are known to be spliced abnormally in DM1 myoblasts. Each protein was overexpressed in a range that was ϳ2-4-fold higher than control transfections to mimic the changes that are observed in DM1 myoblasts (Fig.  6A). Elevated expression of five MBNL1 CUG interactors (hnRNP H, hnRNP H2, hnRNP H3, hnRNP F, and DDX5) individually dysregulated splice site selection in distinct subsets of RNAs (Fig. 6, B-D). Increased levels of CUG-BP1, a protein that interacts with MBNL1 in an RNA-dependent manner (22), resulted in aberrant splicing of two (IR and cTNT) of five RNAs tested (supplemental Fig. S5). Elevated levels of hnRNP L, hnRNP A2/B1, DHX9, hnRNP K, and DDX17 did not alter the splice pattern of the 6 RNAs tested. These results demonstrate that altered stoichiometry of MBNL1 CUG protein partners serve to independently regulate and reinforce aberrant splice site choice in overlapping sets of DM1 RNA targets.

DISCUSSION
Several lines of evidence demonstrate that MBNL1 CUG plays a key role in the development of splice defects and pathology in DM1 (12,15,16,22,27). We therefore attempted to gain mechanistic insights into the role of MBNL1 CUG variants in causing DM1-specific splice defects by studying the behavior of MBNL1 CUG complexes in normal and DM1 myoblasts. As a first step towards this goal, we utilized MB1a mAbs to isolate

Expanded CUG Repeats Alter MBNL1 Complex Stoichiometry
RNA-independent MBNL1 CUG complexes from normal myoblasts. In normal myoblasts, members of this complex include seven hnRNP proteins, hnRNP H, hnRNP H2, hnRNP H3, hnRNP F, hnRNP A2/B1, hnRNP K, hnRNP L, and three RNA helicases, DDX5, DDX17, and DHX9. Our experiments, however, do not distinguish between the possibility that MBNL1 CUG variants directly interact with each of these proteins or alternatively if indirect interactions occur with one or more proteins of this complex via other partner proteins. Nine of ten MBNL1 CUG partner proteins, with the exception of hnRNP L, demonstrate increased (2-3.5-fold) steady state levels as a consequence of expanded CUG repeat expression, and the elevated levels of five proteins, hnRNP H, H2, H3, F, and DDX5, alter splice site selection in overlapping sets of six DM1 target RNAs examined. We have previously demonstrated that MBNL1 interacts in an RNA-dependent manner with CUG-BP1 (22), an alternative splice factor that demonstrates elevated levels in DM1 (12,28) and which facilitates aberrant splicing in two of five RNAs tested in this study. Thus, taken together, these results demonstrate that the pattern of splicing achieved upon expression of expanded CUG repeats is an amalgam that reflects the deregulation of the steady state levels of several independent members of the MBNL1 CUG complex. Notably, the splice defects studied in DM1 myoblasts are similar to that observed in normal myoblasts lacking MBNL1 (data not shown). However, in the DM1 myoblasts examined in this study, the amount of MBNL1 CUG sequestered in CUG foci is relatively minor (ϳ8%; Ref. 24). Therefore, the role of MBNL1 in the development of splice defects in these patient cells has been unclear. Here, our data show that CUG repeat expression can result in DM1-specific splice defects by dysregulating the stoichiometry of the MBNL1 CUG complex, to B and C, endogenous phospho-PKC␣ and PKC␣ levels were measured in the protein extracts from Cos7 cells expressing DMPK 11-15(CTG) 300(i) (B) and two normal and two DM1 myoblasts cells (C) by Western blot analyses. The blots were reprobed for GAPDH as an internal loading control. D and E, DM1 myoblasts were transfected with siRNAs directed against PKC␣, and the cells were harvested as two aliquots 5 days post-transfection. D, protein extracts from one aliquot were analyzed by Western blots for PKC␣ and MBNL1 CUG partners using the antibodies indicated. The blots were reprobed for GAPDH as an internal loading control. E, total RNA was isolated from the second aliquot, and synthesized cDNAs (150 ng) were subjected to exon splicing analysis. GAPDH RNA was amplified in parallel as an internal control. F, RNA splicing defects observed upon depletion of PKC␣ levels in DM1 myoblasts are tabulated. result in an increase in the levels of a panel of MBNL1 CUG interactors, which serve to create a pattern of splice abnormalities that mimic those resulting from MBNL1 loss.
The role of the toxic CUG RNA and its ability to sequester MBNL1 CUG in DM1 etiology is of interest and suggests that although suboptimal MBNL1 CUG sequestration results in splice defects occurring as a consequence of the alteration in complex stoichiometry, over time, as the foci sequester larger amounts of MBNL1 CUG , a precipitous drop in the free MBNL1 CUG reservoir results in a development a MBNL1 CUG deficiency, which in turn may complement and potentially expand the changes that have occurred because of the dysregulation of the steady state levels of its partner proteins. Thus this gradual loss of free MBNL1 CUG may serve to exacerbate the DM1 phenotype with time. The identity of MBNL1 CUG protein partners and the cellular processes that are dysregulated as a consequence of CUG tract expression may, however, vary from tissue to tissue in patients.
The identification of binding sites of several splicing regulators including Nova, Fox2, and PTB demonstrates a correlation between the location of the binding site and exon splicing outcome (28 -30). In these cases, the positional effects are hypothesized to alter the competitiveness of a neighboring site to bind a basal splice factor. Here we potentially observe another level of regulation, conferred by the alteration of the stoichiometry of splice regulator complexes. Previous experiments in DM1 myoblasts demonstrate that overexpression of MBNL1 CUG can reverse DM1 splice defects (12). Thus elevated levels of MBNL1 CUG overcome the independent opposing effects resulting from the increased levels of its partner proteins. In addition, coexpression of MBNL1 CUG and hnRNP H in normal myoblasts significantly reverses the splice defect resulting from elevated levels of hnRNP H. In this regard it is of interest to note that Warf et al. (31) have demonstrated that MBNL1 controls the splicing of exon 5 of cTNT by altering the binding of the splicing factor U2AF65 to the 3Ј end of intron 4. This inhibition appears to be a consequence of these two proteins binding to mutually exclusive RNA structures of the same sequence. Specifically, MBNL1 binds to a portion of the intron that can form a stem loop, whereas U2AF65 binds to the same region as a single-stranded structure (31). In DM1 myoblasts, an increase in MBNL1 levels could not only increase the stability of such a stem loop structure in the cTNT RNA and further decrease the chance of U2AF65 binding but also result in the sequestration of its partner proteins away from their target RNAs, thus influencing their function in splicing. In this model, the ability of splicing regulators to interact with partners that can independently regulate splicing of target RNAs provides a potential mechanism whereby changes in complex stoichiometry could have an enhanced effect on splicing outcomes because of the ability of the members not only to regulate splicing on their own but also to influence the activities of partner splicing regulators by establishing physical interactions.
Lastly, these data further demonstrate that silencing of MBNL1 CUG partners is not a feasible therapeutic strategy for DM1 because several proteins overlapping in their function exist, and their coordinate elimination may be impossible unless early events that regulate changes in the stoichiometry of the MBNL1 CUG are understood. Disruption of MBNL1 CUG binding to its partner proteins is, however, unlikely to be a general mechanism that serves to alter MBNL1 CUG complex stoichiometry, because our previous studies have demonstrated that siRNA-mediated inactivation of MBNL1 does not result in elevated levels of either hnRNP H or CUG-BP1, two proteins that bind MBNL1 in an RNA-independent and RNA-dependent manner, respectively (12,22). To begin to understand the mechanism by which MBNL1 CUG complex stoichiometry is altered, we carried out explorative experiments. We observe that ϳ48 h after expanded CUG tracts are expressed in Cos7 cells, a shift occurs in the stoichiometry of MBNL1 CUG complexes, which leads to the increased levels of a majority of the MBNL1 CUG protein partners identified. Thus the stoichiometry of the MBNL1 CUG complex changes as a direct consequence of expanded CUG repeat expression. The steady state RNA levels of the members of the MBNL1 CUG complex do not however correspond to their protein levels. Thus the changes in MBNL1 CUG complex stoichiometry reflect alterations in either protein synthesis or stability. To distinguish between these two possibilities, we measured the stability of MBNL1 CUG -interacting proteins over a period of 24 h subsequent to cyclohexamide treatment in normal and DM1 myoblasts. These experiments demonstrate that although the half-life of p21 is ϳ5 h in both cell types, MBNL1 CUG interactors are quite stable at 24 h after cyclohexamide treatment in both normal and DM1 myoblasts. Longer time frames of cyclohexamide treatment are not possible in DM1 myoblasts, because these cells show significant cell death at 24 h of cyclohexamide treatment (supplemental Fig. S6). Thus, given the relatively long half-lives of these proteins and the sensitivity of DM1 myoblasts to toxins, this system does not allow us to easily distinguish between these two possibilities.
Importantly, the changes in protein stability do not appear to be global in nature, because levels of other RNA regulatory proteins, including hnRNP L, hnRNP A1 (Fig. 2B), SF2/ASF, and polyadenylate binding protein nuclear 1 (PABPN1), are not significantly different in normal and DM1 myoblasts (data not shown). These results suggest that expression of expanded CUG repeats may perturb one or more signaling pathways, which may in turn be responsible for the modification and subsequent alteration of the steady state levels of MBNL1 CUG complex members. Because Kuyumcu-Martinez et al. (26) have hypothesized that activation of PKC␣ plays a role in the phosphorylation and stabilization of CUG-BP1, a protein that interacts with MBNL1 in an RNA-de-FIGURE 6. Overexpression of MBNL1 CUG -interacting proteins in SkMC results in aberrant splicing of overlapping RNA targets. A and B, SkMC were transfected with 30 g of plasmids encoding FLAG-tagged cDNAs of MBNL1 CUG -interacting proteins. 48 h post-transfection, the cells were selected using G418 for 5 days, harvested, and divided into two aliquots. A, protein extracts (5 or 10 g) prepared from one aliquot were analyzed by Western blots to measure the relative expression of the FLAG-tagged proteins. The blots were reprobed for GAPDH as an internal loading control. B, total RNA was isolated from the second aliquot, and cDNAs (150 ng) were used for exon splicing analysis. C, relative expression of FLAG-tagged recombinant proteins and their effect on RNA splicing is shown. D, a summary of the results is shown. pendent fashion, we examined the potential role of PKC␣ in altering the stoichiometry of MBNL1 CUG complexes and splicing. Consistent with previously reported results, we observe PKC␣ activation in DM1 cells. Results from Kuyumcu-Martinez et al. (26) are consistent with the hypothesis that PKC␣ activation increases the phosphorylation and stabilization of CUG-BP1, as detailed by the behavior of chemical activators and inhibitors of PKC␣. However, because these experiments do not establish a causal relationship between PKC␣ and elevated CUG-BP1 levels or aberrant RNA splicing in DM1, we carried out a mirror image experiment to test the potential role of PKC␣ function in both facilitating the change in the steady state levels of members of MBNL1 CUG complexes and aberrant RNA splicing in DM1. siRNA-mediated inactivation of PKC␣ was not sufficient to normalize the altered levels of the members of MBNL1 CUG complexes, reverse elevated levels of CUG-BP1, or rescue aberrant RNA splicing in DM1 myoblasts. Therefore, inactivation of PKC␣ is insufficient to reverse the altered stoichiometry of MBNL1 CUG complexes or the aberrant RNA splicing observed in DM1 patient cells. However, it is possible that activation of PKC␣ may have deleterious effects on other cellular processes in ways that are yet to be fully understood. Our current experiments are therefore focused on understanding the nature of the early events that occur upon expanded CUG repeat expression and their effect on the stability and/or synthesis of members of MBNL1 CUG complexes.
In this study our focus was on deciphering the role of MBNL1 CUG complexes in the development of DM1 splice defects. Nonetheless, it is of interest to note that the members of MBNL1 CUG complexes identified in this study have been shown to play a variety of roles in remodeling ribonucleoprotein complexes, including microRNA processing, RNA transport, translation, and the coactivation of transcription factors (32)(33)(34)(35)(36)(37)(38). Thus elevated levels of individual members of MBNL1 CUG complexes may alter other key processes and contribute in additional important ways to DM1 pathology.