Intrinsically cell-penetrating multivalent and multitargeting ligands for myotonic dystrophy type 1

Significance

Multivalent ligands have tremendous potential as therapeutic agents; however, their efficacy is limited by delivery issues, poor cell permeability, and toxicity. We report here a strategy wherein multivalent ligands are designed to be intrinsically cell-penetrating, allowing them to target the expanded trinucleotide repeat sequences of DNA and RNA that cause myotonic dystrophy type 1 (DM1). The multivalent ligand studied shows cell permeability and low toxicity both in cells and in mice. Importantly, the ligand reduced or eliminated DM1 defects in DM1 cells and in vivo, validating the multivalent strategy. The approach should be broadly applicable to other repeat expansion diseases and to any multivalent oligomeric therapeutic agent whose activity can accommodate structural elements that mimic cell-penetrating peptides.

Abstract

Developing highly active, multivalent ligands as therapeutic agents is challenging because of delivery issues, limited cell permeability, and toxicity. Here, we report intrinsically cell-penetrating multivalent ligands that target the trinucleotide repeat DNA and RNA in myotonic dystrophy type 1 (DM1), interrupting the disease progression in two ways. The oligomeric ligands are designed based on the repetitive structure of the target with recognition moieties alternating with bisamidinium groove binders to provide an amphiphilic and polycationic structure, mimicking cell-penetrating peptides. Multiple biological studies suggested the success of our multivalency strategy. The designed oligomers maintained cell permeability and exhibited no apparent toxicity both in cells and in mice at working concentrations. Furthermore, the oligomers showed important activities in DM1 cells and in a DM1 liver mouse model, reducing or eliminating prominent DM1 features. Phenotypic recovery of the climbing defect in adult DM1 Drosophila was also observed. This design strategy should be applicable to other repeat expansion diseases and more generally to DNA/RNA-targeted therapeutics.

Multivalency is an effective strategy for the development of potent and selective ligands, especially at cell surfaces. The thermodynamic and kinetic advantage of multiple binding moieties interacting with many receptor sites is attributed to several possible factors, including steric stabilization, the chelation effect, and concentration effects (1–3). For the maximum benefit of multivalency to accrue, there must be a suitable spatial arrangement of the ligands and a strain-free, ligand-receptor “fit” upon complexation. Dimeric ligands can be viewed as simplified versions of multivalent ligands, and there is ample evidence that optimization of the linkage between two binding motifs is necessary to obtain significant affinity improvement (4–6). However, rather than using an optimized dimeric linkage, multivalent ligands are often prepared by covalent attachment of ligands to readily available extrinsic scaffolds, such as a PNA or dendrimer. In these cases, the ease of synthesis often comes with loss of tunability and spatial optimization (7–9). A separate consideration comes into play when the target is intracellular. Namely, the increase in mass can reduce cell permeability, although the use of a cell-penetrating peptide (CPP) as the scaffold can help (9).

Trinucleotide repeat expansion diseases (TREDs) are outstanding targets for exploring intracellular multivalent approaches, because the pathogenic targets involve regular and repeating sequences of DNA and RNA. For example, myotonic dystrophy type 1 (DM1), an incurable TRED, originates in expanded CTG repeats [d(CTG)^exp] in the 3′-UTR of the DMPK gene on chromosome 19 (Fig. 1). An RNA gain-of-function model for DM1 involves r(CUG)^exp transcripts altering the level of splicing regulators, such as muscleblind-like 1 (MBNL1) and CUG binding protein 1 (CUGBP1), leading to misregulation of pre-mRNA splicing (10–13).

Fig. 1.

DM1 pathogenesis and therapeutic strategy. CTG expanded repeats in 3′-UTR of the DMPK gene is transcribed to r(CUG)^exp, forming a hairpin secondary structure with U–U mismatches. The structured RNA sequesters MBNL1 and causes splicing misregulation and other DM1 defects. A designed multivalent ligand has a multitargeting ability toward d(CTG)^exp and r(CUG)^exp. The oligomeric inhibitor blocks transcription of d(CTG⋅CAG)^exp by d(CTG)^exp binding and intervenes the r(CUG)^exp–MBNL1 interaction. These strategies are expected to decrease toxic r(CUG)^exp levels and restore MBNL1 levels, respectively, thereby recovering DM1 defects.

Recent reports shed light on additional pathobiology of DM1, such as microRNA dysregulation (14) and repeat-associated non-ATG (RAN) translation (15, 16). Nonetheless, the sequestration of MBNL1 by r(CUG)^exp, and their coaggregation into distinctive nuclear foci is considered as a hallmark of DM1 (17). Importantly, MBNL1 knockout mouse model studies have reported that >80% of DM1 missplicing defects and related symptoms are directly associated with the loss of MBNL1 by r(CUG)^exp (18, 19). Thus, therapeutic efforts focus on inhibiting the transcription of d(CTG)^exp to reduce r(CUG)^exp levels (20–22) or on inhibiting the interaction of r(CUG)^exp with MBNL1 (23–27). Here, we report an oligomeric ligand that targets both d(CTG)^exp and r(CUG)^exp in a multivalent fashion and contains a structure designed to mimic CPPs for inherent cell permeability (Fig. 1). The mixture of oligomers reduces the disease foci and toxic r(CUG)^exp level in cellular assay and shows promising in vivo activity using a known adult DM1 Drosophila phenotypic assay and a liver-specific DM1 mouse model.

Results

Rational Design of a Multivalent Ligand.

We recently reported a small-molecule inhibitor of the r(CUG)^exp–MBNL1 interaction (26) whose structure (1 in Fig. 2) features a bisamidinium core as a general RNA groove binder and two triaminotriazine units designed for selective recognition of U–U mismatches. Although the synthesis was not trivial, we were able to prepare dimeric ligand 2 using the alkyne-azide copper-assisted (CuAAC) click reaction and thus a triazole linker between two monomer units (Fig. 2A) (5). Dimer 2 inhibited formation of the MBNL1–r(CUG)₁₆ complex with a K_i = 25 ± 8 nM that was nearly 1,000-fold lower than that measured for monomer 1 and r(CUG)₁₂ (K_i = 16 ± 6 μM). However, 2 did not perform well in DM1 cellular assays compared with 1,000-fold enhanced activity in vitro.

Fig. 2.

Rational design and synthesis of multivalent ligands. (A) Structure of ligand 1 with two triaminotriazine U–U or T–T recognition moieties (green hexagons) connected through a groove-binding bisamidinium linker (optimized linker A). Dimeric ligand 2 was previously prepared by click chemistry (triazole linker B). Ligand 3 shows an alternative design for a dimer, linking two monomers using a groove binder linker A. This design provides a more regular structure, enabling further elongation to an oligomeric ligand 4. (B) Synthesis of oligomeric ligand 4 by polycondensation reaction. (C) MALDI-TOF characterization of oligomer 4 shows the formation of an oligomeric mixture ranging from 4 to 8 mer (average molecular mass of ∼2 kDa). See text for explanation of R end groups. DHB was used as the matrix. The spectrum was collected by using reflective mode. Intens., intensity.

Several lines of evidence suggest that the structure of 2 is particularly unsuitable for cell-membrane penetration. First, its large size (molecular mass = 1,166.4 Da) (5) places it well outside of the Lipinski rule of 5 (Ro5), where compounds with molecular mass < 500 are most likely to show excellent passive diffusion across the cellular membrane. Although some clinically useful and orally bioavailable drugs have molecular masses between 500 and 1,000, a range known as “beyond the Ro5” (bRo5), these agents are usually cyclic peptides, often with N-methyl groups. Invariably, a steep drop-off in membrane permeability occurs at molecular mass > 1,000, and an upper size limit appears to be molecular mass ∼ 1,000 for passive diffusion of small molecules across cell membranes (28–30). Beyond the high molecular mass, dimer 2 violates at least two other Ro5s with a large number of N–H and hydrogen-bond acceptor groups. The final challenge for 2 to enter the cell comes from the need to move four cations (amidinium groups) across the membrane. Interestingly, previous studies with oligoarginines and aromatic amide foldamers found that four positive charges were not enough to efficiently translocate into cells, whereas eight cationic groups were found to be the most efficient (31, 32). These CPPs and their mimics frequently have molecular masses >1,000, yet readily enter cells through either energy-dependent or -independent mechanisms.

Given the considerations above, it appears that ligand 2 is too large and too charged to passively penetrate the cell membrane as a small molecule and yet too small and insufficiently cationic to enter the cells through CPP-type mechanisms. Thus, we sought to connect four to six monomeric units of 1 to create a CPP mimic with 8–12 cationic groups. Ligand 1 is designed to recognize (CUG)₃ with the bisamidinium linker, designated linker A in Fig. 2A, binding the groove of the central CUG and the triaminotriazine units recognizing the terminal U–U mismatches. In linking monomeric units of 1, it can be critical to select the appropriate linker. The monomeric units of dimer 2 are linked by a triazole group (linker B, Fig. 2A). In this study, we considered using the optimized linker A, as seen in ligand 3. The central bisamidinium in 3 provides two more positive charges, as well as additional groove binding, potentially improving cell permeability and affinity for the target r(CUG)^exp. The regular alternating structure of 3 further suggested oligomeric or polymeric ligand 4 with multivalent binding capability (Fig. 2A).

To validate our molecular design, we performed a preliminary molecular modeling study using molecular dynamics (MD) simulations. Following the same approach we reported (26) for the MD simulation of ligand 1 bound to [r(CUG)₆]₂, the oligomeric ligand 4 was manually docked into [r(CUG)₁₅]₂. The ligand used had three bisamidinium moieties and four triaminotriazine units, corresponding to a degree of polymerization (DP; i.e., number of repeat units) of 3.5. A 10-ns MD simulation suggested stable major groove binding with the triaminotriazine units recognizing every other U–U mismatch along the sequence (SI Appendix, Fig. S1). The triaminotriazine units formed multiple hydrogen bonds to and were positioned roughly coplanar with one of the uracil groups, whereas the other was flipped out. This simple computational study supported our rational design wherein oligomeric ligand 4 can selectively recognize r(CUG)^exp in a multivalent fashion.

Preparation of Oligomer 4 by Polycondensation and Interaction with r(CUG)₁₆.

The alternation of triaminotriazine and bisamidinium moieties in 4 suggested a single-step synthesis involving the polycondensation of diamine 5 and bisimidate 6 (Fig. 2B), precursors that can be prepared in one or two steps from commercially available compounds (SI Appendix, SI Materials and Methods). The polycondensation reaction was performed in DMF at 35 °C, reliably producing low-molecular-mass oligomeric products. Temperature control was important to restrict the DP.

The small-molecule by-products and the smallest oligomers were removed by dialysis, giving a mixture of oligomers that exhibited a single HPLC peak (SI Appendix). The MALDI-MS indicated formation of the desired oligomeric products ranging from 4 to 8 mers with an average molecular mass of ∼2 kDa (DP ∼ 5; Fig. 2C). Thus, two prominent mass series are seen in the MALDI-MS, each separated by m/z 396, the molecular mass of the repeat unit. The cluster of peaks in each mass series in the MALDI-MS spectrum suggests that different end groups are present on the oligomers. ¹H NMR indicates the presence of both ester groups from hydrolysis of imidate ester 6 and aminobutyl groups expected for 5 as terminal groups (SI Appendix). The more prominent mass series (green arrows, Fig. 2C) can be assigned to the oligomeric mixture 4 with ester groups at both ends. The second mass series (magenta arrows, Fig. 2C) is consistent with two possible structures. The first involves 5 as one terminal group and a 4-cyanobenzamidinium as the other. The cyano group may arise from incomplete conversion of 1,4-dicyanobenzene into bisimidate 6. The second possibility involves dual MALDI-MS fragmentation at the benzamidinium unit, consistent with the known fragmentation of arginine. The mass series with blue arrows likely involves one ester end group and fragmentation at the benzamidinium unit (additional details on the MALDI-MS can be found in SI Appendix, Characterization) (33). The presence of amine end groups in 4 derived from 5 is supported by the ability of the oligomer to be fluorescently labeled (see below).

Given that 4 is an oligomeric mixture, in vitro binding studies with a discrete r(CUG)_n will lead to a complex mixture of bound structures. Indeed, the binding of 4 with r(CUG)₁₆ was studied by using isothermal titration calorimetry (ITC) (SI Appendix, Fig. S2A), which showed a strongly enthalpic interaction and a U-shaped curve consistent with positive cooperativity. In comparison, the ITC curve for monomer 1 was consistent with a 1:1 complex and significantly weaker binding (SI Appendix, Fig. S2B). Dynamic light scattering (DLS) studies were also performed at similar concentrations to examine the nature of the complexation. The data suggested large polyplex formation with increased concentration of oligomer 4 and r(CUG)₁₆ (SI Appendix, Fig. S3). The DLS also indicated that oligomer 4 is monomeric at 20 μM and below. It needs to be noted that the ITC and DLS experiments require a concentration considerably higher than the expected working concentration in biological environments.

Cell Permeability of Oligomer 4.

To investigate the cell permeability of oligomer 4, we used confocal microscopy to visualize oligomer 4 inside live cells. The oligomer was fluorescently labeled with fluorescein (FAM) or Rhodamine B (RhB) by using isothiocyanate-amine conjugation chemistry. Either RhB– or FAM–oligomer 4 was incubated with HeLa cells. A significant amount of fluorescence was observed inside the cells and nucleus, confirming the cell permeability of oligomer 4 (SI Appendix, Fig. S4A). Preliminary studies were undertaken with Lysotracker Red to shed light on the mechanism of the cell penetration. The Lysotracker Red fluorescence was observed to overlap significantly with the FAM fluorescence from oligomer 4, consistent with an endocytotic uptake mechanism (SI Appendix, Fig. S4 B and C). However, nonoverlapping FAM fluorescence was also observed, indicating either escape from the endosome or a direct penetration mechanism. No disruption of the cell surface was observed in any of the experiments.

rCUG-Targeting in DM1 Model Cell Culture and Patient-Derived Myoblasts.

Given that the key cellular feature of DM1 is the presence of ribonuclear foci, formed between r(CUG)^exp and MBNL1, we examined whether oligomer 4 could inhibit their formation. Initial studies were performed with a standard DM1 model cell culture produced by transfecting HeLa cells with a plasmid containing GFP and DMPK exons 11−15, with 960 CTG repeats in exon 15 (DT960–GFP) (34). The r(CUG)₉₆₀ and MBNL1 in the foci were separately visualized, the former by using fluorescence in situ hybridization (FISH) with Cy3-(CAG)₁₀ and the latter by using immunofluorescence with an anti-MBNL antibody (17). Untreated cells showed significant areas of nuclear foci observed as yellow punctate within the nucleus, demonstrating MBNL1 protein (green)–r(CUG)₉₆₀ RNA (red) colocalization (SI Appendix, Fig. S5A, row 1 merge). Treatment with oligomer 4 at various concentrations for 48 h significantly suppressed the foci formation (SI Appendix, Fig. S5A). Quantitative analysis indicated a small dose-dependent response from 100 to 500 nM of oligomer 4, with 68–76% reduction in the foci area (SI Appendix, Fig. S5B). It is noteworthy that monomeric ligand 1 showed similar activity at ∼1,000-fold higher concentration (100 µM) (26), consistent with an oligovalent advantage conferred on 4.

More interestingly, we used FAM-conjugated oligomer 4 in an effort to track its location within the DM1 model cells. At a concentration of 200 nM, colocalization of the FAM signal and r(CUG)₉₆₀ were observed within the nucleus. This observation provides evidence for 4 penetrating both the nuclear and cell membranes and directly engaging r(CUG)₉₆₀ (SI Appendix, Fig. S6). At a higher concentration of 500 nM, the FAM–oligomer 4 appeared as cytoplasmic punctate, suggesting lack of endosomal release, limited nuclear membrane penetration, or saturation of r(CUG)₉₆₀ binding sites at the higher concentration (SI Appendix, Fig. S6). This would explain the modest dose dependence for foci disruption upon increasing the concentration of 4 from 100 to 500 nM and the inability to fully dissolve all foci at the highest concentrations (SI Appendix, Fig. S5B).

For a more disease-relevant cellular assay, we used DM1 patient-derived myoblasts, which were prepared by successful differentiation of DM1 fibroblasts (GM03987; Coriell Institute) using MyoD-containing retrovirus (SI Appendix) (35, 36). Whereas the DM1 model cell culture used above produces r(CUG)₉₆₀ with a doxycycline (Dox)-inducing expression system (22, 34), the DM1 myoblasts spontaneously generate heterogeneous transcripts up to r(CUG)₅₀₀. Treatment of the patient cells with 200 nM 4 for 48 h led to a 30% reduction in foci area (Fig. 3A and SI Appendix, Fig. S5C). The smaller reduction in the DM1 myoblasts may indicate a lower level of uptake in DM1 myoblasts compared with in HeLa cells. It is important to note that no sign of cytotoxicity was observed throughout the working concentration ranges both in HeLa cells and DM1 patient-derived fibroblasts (GM03987) (SI Appendix, Fig. S7).

Fig. 3.

Ligand 4 as an inhibitor of r(CUG)^exp–MBNL1 and d(CTG⋅CAG)^exp transcription. (A) Reduction of disease foci in DM1 patient-derived myoblasts upon the treatment of oligomer at 200 nM. Foci areas are indicated with white arrows. (Scale bars, 5 μm.) (B) Schematic description of splicing pattern of IR pre-mRNA and the result of splicing assay in a presence of oligomer 4. Full recovery of the IR splicing defect was obtained at 200 nM. P values were calculated against data points of the diseased sample. *P < 0.05. No significant differences between normal cells and each of all the treated samples (P > 0.05, Student’s t test, two-tailed). (C) A plot of normalized RNA transcript level against the oligomer 4 concentration. RNA transcript level was normalized to the level of a loading reference RNA, HIV fs RNA, for quantitative comparison between the loading samples. Oligomer 4 inhibited d(CTG)^exp transcription in a dose-dependent manner (IC₅₀ = 100 nM), but no significant inhibition was detected with the control DNA template throughout all the tested concentrations. Error bars indicate standard error of the mean from three independent experiments.

Inspired by enhanced activity of oligomer 4 in foci disruption relative to monomer 1, we further evaluated an important downstream feature, the misregulation of alternative splicing of pre-mRNAs. IR pre-mRNA was chosen for this study because its missplicing is relatively difficult to correct (37). The IR pre-mRNA undergoes two possible splicing pathways dependent on the exclusion or inclusion of exon 11, which produces isoform A or B, respectively (Fig. 3B). The DM1 model cell culture produced by transfecting HeLa cells with plasmids containing DT960 and IR minigenes is able to reproduce the IR splicing pattern observed in normal and disease states (13, 36). Oligomer 4 showed full rescue of IR missplicing at a concentration of 200 nM, although the recovery rate diminished slightly and leveled off at higher concentrations, similar to our observation with foci disruption. (Fig. 3B). Again the strength of oligovalent approach was evident when comparing to monomeric ligand 1, which under similar conditions provided only 43% missplicing rescue at a 500-fold higher concentration (100 µM) (26).

dCTG-Targeting by Oligomer 4 and Inhibition of Toxic RNA Synthesis.

As explained, recent studies showed that r(CUG)^exp can produce additional toxicity, which adds to the complexity of the DM1 disease pathobiology. For example, the r(CUG)^exp transcript can disrupt the translation of MEF2 protein, leading to microRNA dysregulation in DM1 heart tissue (14). Additionally, r(CUG)^exp undergoes RAN translation, producing homopeptides that may be toxic (15, 16). For these reasons, we tested the potential of oligomer 4 to regulate the levels of toxic r(CUG)^exp. Although in our previous studies, we did not see d(CTG)^exp binding by monomeric ligand 1, N-substituted analogs that add additional cationic groups are able to target both the DM1 DNA and RNA expansions (22).

We first checked whether oligomer 4 can inhibit the transcription of d(CTG)^exp in vitro using a (CTG)₇₄-containing DNA template. A strong dose-dependent inhibition effect of 4 was observed toward the (CTG)₇₄ template (IC₅₀ = 100 nM), whereas no inhibition was apparent for the transcription of a non-repeat-containing control template (Fig. 3C and SI Appendix, Fig. S8A). This finding is consistent with oligomer 4 selectively and potently targeting d(CTG)^exp transcription, possibly by stabilizing an analogous hairpin secondary structure formed in transcription bubble (38). The potential of 4 to act as a transcription inhibitor and down-regulate the toxic RNA levels in cells was tested. After treatment of the DM1 model cell culture with 4 for 72 h, the level of r(CUG)₉₆₀ was found to be reduced by 20–50%, depending on the oligomer concentration. These results indicate that oligomer 4, by inhibiting d(CTG)^exp transcription, can suppress the level of r(CUG)^exp, the DM1 causative agent (SI Appendix, Fig. S8B).

Recovery of Phenotypic Defect in DM1 Drosophila Adult Climbing Assay.

The promising cell-permeability and multitargeting features of 4 produced significant improvement of cellular DM1 phenotypes, which motivated us to further study its activity in vivo. We used a phenotypic assay with adult DM1 Drosophila that takes advantage of their innate climbing behavior (39). Because DM1 is a neuromuscular disease affecting both the central nervous system and the motor system, DM1 flies show age-dependent progressive climbing defects (39, 40). By aging DM1 adult flies for several days or even months, more r(CUG)^exp is accumulated, allowing for disease symptoms that range from a mild to a severe climbing defect. In turn, this tunability allows screening of therapeutic ligands with widely different activities. Moreover, the climbing assay allows fast screening using a larger number of adult flies, generating statistically important data faster and more conveniently.

With the aged adult DM1 flies created with (CTG)480 as reported, but with an Elav–GAL4-inducing system (40), we evaluated head to head the ability of monomeric ligand 1 and oligomer 4 to reverse the climbing failure (SI Appendix, Fig. S9). For normal flies with CTG60, treatment with 4 (80 µM) and monomer 1 (400 µM) did not affect their climbing ability, suggesting minimal toxicity of compounds. When it came to DM1 flies (CTG480), almost 80% of the flies failed to climb, but treatment with oligomer 4 at a concentration of 20–80 µM rescued their climbing ability in a dose-dependent fashion. At the highest dose of 80 µM, the failure rate was dropped to 37%, representing a 54% recovery. In contrast, monomer 1 decreased the failure rate to 55%, a 31% recovery, at a fivefold higher concentration (400 µM). The phenotypic improvement in DM1 adult flies demonstrated the in vivo activity of 4, although the molecular mechanism of action was not investigated.

In Vivo Activities of Multivalent Ligands in a Liver-Specific DM1 Mouse Model.

As a result of the multisystemic nature of DM1, symptoms other than muscle wasting and myotonia are often exhibited by patients. These symptoms include cardiac conduction defects, cataracts, testicular atrophies, dysphagic symptoms, and other gastrointestinal-tract-related symptoms (41, 42). Abnormal elevation of liver enzymes such as aspartate aminotransferase, alanine aminotransferase, alkaline phosphatase, and lactate dehydrogenase along with nonalcoholic fatty liver disease and gallbladder dysfunction are also common in DM1 patients (43–45), and even though such liver-associated pathologies may be the first symptom observed, they remain understudied. Therefore, we developed a liver-specific DM1 mouse model by expressing 960 CUG repeat-containing DMPK RNAs in a tetracycline-inducible (46) and hepatocyte-specific manner (TRE-960I;ApoE-rtTA mouse model; SI Appendix, Fig. S10A). The liver-specific CUG₉₆₀ mouse model served as an attractive model for screening in vivo efficacy of oligomer 4, considering the ease of delivery of compounds to the liver via i.p. injections (47). An 8-d pathogenic r(CUG)₉₆₀ RNA induction and treatment scheme was used in this study (Fig. 4A). Efficient r(CUG)₉₆₀ RNA expression was observed in the mouse livers upon feeding a 6 g/kg Dox-containing diet, as seen by quantitative RT-PCR (qRT-PCR) (Fig. 4B). Continued injections with oligomer 4 led to a significant reduction in the level of r(CUG)₉₆₀ RNA, likely due to its inhibitory effect on d(CTG)^exp transcription, as observed with in vitro and DM1 cell studies. Dox diet and oligomer 4 treatment as mentioned above had no significant impact on the body weight or liver-to-body-weight ratios (SI Appendix, Fig. S11 A and B). Histological analysis showed no overt signs of liver toxicity—such as inflammation, injury, or fibrosis—in oligomer 4-treated mice (SI Appendix, Fig. S11F). Furthermore, the effects of oligomer 4 in reducing r(CUG)₉₆₀ RNA load were specific, as the mRNA levels of three other nontargeted genes, Mapkap1, Pcolce, and Mllt3—which contain 26, 12, and 8 CUG repeats, respectively—were not reduced after treatment (SI Appendix, Fig. S11C).

Fig. 4.

In vivo activities of oligomer 4 in a liver-specific CUG₉₆₀ transgenic mouse model. (A) Schematic of disease induction and 4 administration in the livers of TRE-960I;ApoE-rtTA mice. (B) qRT-PCR result shows efficient expression of r(CUG)^exp in TRE-960I;ApoE-rtTA mice upon Dox treatment. Oligomer 4 efficiently reduces r(CUG)^exp expression. Data are normalized to β-actin. (C) Quantitative analysis of colocalized disease foci area per cell. Significant amount of disease foci with MBNL1 was released with 4 treatment. (D) Foci disruption upon 4 treatment in the liver tissue. Colocalization of r(CUG)^exp and MBNL1 was detected by FISH combined with immunofluorescence. (Scale bars, 5 μm.) (E) Rescue of splicing defects caused by expression of r(CUG)^exp. Mean (±SD) of percent spliced in (PSI) is plotted for different events from n = 3, 4, and 4 mice for control, untreated, and oligomer 4-treated mice, respectively. *P < 0.05; ns, not significant.

We then studied the most prominent molecular features of DM1 to understand the effects of the oligomer. To analyze the characteristic disease ribonucleic foci, FISH and immunofluorescence were performed on frozen liver-tissue sections (17). Upon Dox induction, RNA foci were clearly visible in the hepatocyte nuclei, which were significantly reduced after treatment with oligomer 4 (Fig. 4D and SI Appendix, Fig. S11D). Colocalization of MBNL1 and CUG RNA also exhibited a marked reduction, demonstrating reduced sequestration of MBNL1 (Fig. 4 C and D). In addition, we identified several developmentally regulated splicing events that are sensitive to MBNL1 levels in the liver by screening their splicing patterns in wild-type fetal and adult as well as MBNL1-knockout mice (SI Appendix, Fig. S10 B and C) (18, 48).

When induced with Dox, TRE-960I;ApoE-rtTA mouse livers consistently reproduced the alternative splicing defects evident in the livers of MBNL1-knockout mice, which were significantly reversed with oligomer 4 treatment (Fig. 4E and SI Appendix, Fig. S11E). As expected, the developmentally regulated splicing events (Add1_34 and Usp4_141) that are insensitive to MBNL1 levels in the liver showed no change upon r(CUG)^exp induction or after oligomer 4 treatment (Fig. 4E and SI Appendix, Fig. S11E). Together, these results demonstrated that systemic administration of oligomer 4 to a liver-specific transgenic r(CUG)₉₆₀ mouse model can reduce the burden of r(CUG)^exp RNA levels, disperse the ribonuclear foci-releasing MBNL1 proteins, and correct the splicing defects for a number of pre-mRNAs that are regulated by MBNL1 in the liver.

Discussion

DM1 serves as an excellent disease platform for testing intracellular multivalent approaches because the pathogenic agents d(CTG)^exp and r(CUG)^exp involve regularly repeating sequences. Current approaches for multivalency heavily depend on linker optimization between monomeric units or incorporation of binding moieties onto a delivery vehicle such as CPPs or their mimics (4, 5, 9, 24). The former strategy often requires laborious in vitro optimization, yet the increased size and low cell permeability frequently limits the potency in cellular assays or in vivo. In the latter approaches, it is often difficult to control the positions and spatial arrangement of binding ligands on the carrier scaffold. In this study, we introduced multivalent ligands that are inherently cell-permeable. The alternating structure of triaminotriazines and bisamidinium recognition units contains the CPP features of cationic charges and amphipathicity.

Despite its large molecular mass, oligomer 4 significantly outperformed monomer 1 in each cellular and in vivo assay. The oligomer exhibited excellent multitargeting activity binding both d(CTG)^exp and r(CUG)^exp, thereby inhibiting the DM1 pathogenesis. These activities include reducing the toxic r(CUG)^exp level by inhibiting transcription and dispersing disease foci (rCUG^exp-MBNL1 coaggregates) in both DM1 model cells and DM1 patient myoblasts. The difference in the foci-dissolving activity observed in DM1 model cell culture and differentiated DM1 myoblasts may reflect a different extent of uptake. Membrane permeability can be modified by abnormal protein–membrane interactions, loss of integral proteins, etc., which are possible differences between DM1 model cells and differentiated DM1 myoblasts (49, 50).

More interestingly, the same activities of correcting DM1 molecular defects were also observed in vivo in a liver-specific CUG₉₆₀ mouse model, fully rescuing multiple MBNL1-regulated splicing defects. Our liver-specific r(CUG)₉₆₀ mouse model is expected to be a useful model to understand pathogenic mechanisms of DM1 in the liver and related symptoms, which have not been examined before. Also, considering a facile delivery of compounds to the liver via i.p. injections, our mouse model can serve as an initial in vivo platform for potential DM1 therapeutic agents. Importantly, oligomer 4 showed no sign of toxicity at the working concentrations in HeLa cells, DM1 patient cells (GM03987), and both in vivo models. The maximum tolerated dose (MTD) was determined to be 40 mg/kg in C57BL/6 mice, which is significantly higher than that for dimeric ligand 2 (3 mg/kg) (5).

Our oligomeric ligand inherently mimics the structure of CPPs such as oligoarginines or Tat. In this regard, we believe that this approach represents a paradigm in multivalent ligand design, where the direct assembly of binding moieties serves as their own delivery agent. The ability to prepare oligomer 4 by a condensation polymerization streamlines the synthesis, but moving forward it will be important to prepare homogenous oligomers for testing. Nonetheless, our general design principle should be applicable to other TREDs, such as Huntington’s disease and spinocerebellar ataxias, considering the similarities in their pathogenesis.

Materials and Methods

All reagents were purchased from standard suppliers and used without further purification unless otherwise noted. r(CUG)₁₆ was purchased from Integrated DNA Technologies or synthesized on an Applied Biosystems 3400 DNA Synthesizer, followed by HPLC purification. NMR spectra and mass spectral analyses were performed by using instrumentation available in the School of Chemical Sciences, University of Illinois. Details of the synthesis, compound, and oligomer characterization, as well as additional experimental details and additional figures, can be found in SI Appendix. We strictly followed the National Institutes of Health guidelines for use and care of laboratory animals, and all experimental protocols were approved by the Institutional Animal Care and Use Committee at the University of Illinois, Urbana–Champaign.