Structure of the myotonic dystrophy type 2 RNA and designed small molecules that reduce toxicity

Abstract

Myotonic dystrophy type 2 (DM2) is an incurable neuromuscular disorder caused by a r(CCUG) expansion (r(CCUG)(exp)) that folds into an extended hairpin with periodically repeating 2×2 nucleotide internal loops (5'CCUG/3'GUCC). We designed multivalent compounds that improve DM2-associated defects using information about RNA-small molecule interactions. We also report the first crystal structure of r(CCUG) repeats refined to 2.35 Å. Structural analysis of the three 5'CCUG/3'GUCC repeat internal loops (L) reveals that the CU pairs in L1 are each stabilized by one hydrogen bond and a water-mediated hydrogen bond, while CU pairs in L2 and L3 are stabilized by two hydrogen bonds. Molecular dynamics (MD) simulations reveal that the CU pairs are dynamic and stabilized by Na(+) and water molecules. MD simulations of the binding of the small molecule to r(CCUG) repeats reveal that the lowest free energy binding mode occurs via the major groove, in which one C residue is unstacked and the cross-strand nucleotides are displaced. Moreover, we modeled the binding of our dimeric compound to two 5'CCUG/3'GUCC motifs, which shows that the scaffold on which the RNA-binding modules are displayed provides an optimal distance to span two adjacent loops.

Figure 1

The design of small molecules that reduce r(CCUG)^exp toxicity in cellular model systems of DM2. A, a resin-based selection completed in our laboratory determined that 2×2 nucleotide pyrimidine-rich internal loops bind to kanamycin A with the highest affinity. These loops are similar to those found in r(CCUG)^exp. B, DM2 is caused by an expanded r(CCUG) repeat (r(CCUG)^exp) found in intron 1 on the ZNF9 pre-mRNA. r(CCUG)^exp binds and inactivates MBNL1, leading to dysregulation of alternative pre-mRNA splicing. The kanamycin A-like module (K) identified from our resin-based selection binds the repeating motif in r(CCUG)^exp, 5’CCUG/3’GUCC, with high affinity and selectivity. We previously showed that displaying the K module multiple times on a peptoid scaffold potently inhibits the r(CCUG)^exp-MBNL1 complex in vitro. Herein, we demonstrate that the optimal compound identified from in vitro studies, 3K-4, improves DM2-associated defects in cellular model systems. C, the enhanced bioactivity of 3K-4 is due to a hinge effect. That is, multivalency affects k_on and k_off when binding to the RNA target.

Figure 2

A designed, modularly assembled compound improves dysregulation of alternative pre-mRNA splicing in a DM2 model cellular system. A, schematic of the BIN1 mini-gene used to study dysregulation of pre-mRNA splicing due to r(CCUG)^exp. The construction of the mini-gene has been previously described. B, K, 2K-4, and 3K-4 improve BIN1 alternative splicing defects to varying extents. 3K-4 is the most potent restoring BIN1 splicing patterns to wild type at 10 M. Splicing patterns were determined by RT-PCR. C, Importantly, K, 2K-4, and 3K-4 do not affect the splicing of BIN1 pre-mRNA in the absence of r(CCUG)^exp. Moreover, 3N-4, a peptoid similar to 3K-4 but displays a neamine-like (N) module, does not improve DM2-associated splicing defects. 3N-4 does not bind r(CCUG)^exp and is a poor inhibitor of the r(CCUG)^exp-MBNL1 complex in vitro. D, K, 2K-4, and 3K-4 do not affect the alternative splicing of SMN2 mRNA, which is regulated by Sam68, indicating that the compounds exhibit specificity in vivo. Error bars represent the standard deviation of at least three measurements. A “*” indicates p<0.05 while “**” indicates p<0.01 as determined by a two-tailed student t-test in which the treated samples were compared to untreated samples derived from cells that express r(CCUG)^exp.

Figure 3

Crystal structure of r(CCUG) repeat-containing RNA. (A) Secondary structure of the RNA that was subjected to crystallization. Tetraloop-tetraloop receptor elements are colored blue and CCUG repeats are colored orange with each loop numbered L1, L2, and L3. (B) Overall crystal structure of the RNA showing only the r(CCUG) repeats and closing pairs (left) and a standard A-form RNA with the same length and base pair composition except CU pairs are replaced with 5’CU/3’GA (right). The repeats are shown as orange sticks with transparent surfaces while the closing pairs are shown as grey sticks. (C) Electropotential surfaces of r(CCUG) repeats and closing pairs in the crystal structure (A). The insets show the electropotential surfaces of L1 (left) and L3 (right). The electropotential was contoured at ± 20 KT/e.

Figure 4

Hydrogen bonding pattern and base stacking of CU pairs in the crystal structure. Interatomic bond distances (cyan dashes) and non-bonded distances (yellow dashes) between subject atoms of L1 (A), L2 (B), and L3 (C) are shown. The C1’-C1’ distances (black dash lines) are indicated under each CU pair. 2F_o-F_c electron density map covering each CU pair is contoured at 1 σ.

Figure 5

Model systems (A) 2×CCUG and (B) 2×CCUG_inf used in MD simulations, and two most frequently seen conformations, (C) CU_a and (d) CU_b, in the MD trajectories. 2×CCUG and 2×CCUG_inf are solvated in truncated octahedral and cubic boxes, respectively. In (A) and (B), RNA molecules are shown in new ribbon representations with the CU/UC loops shown in red. Oxygen atoms of water molecules in (A) and (B) are shown as red spheres while Na⁺ and Cl⁻ ions are shown as blue and cyan spheres, respectively. In (B), the RNA sequence and system were designed in order to study an infinitely long RNA molecule with CU/UC loops when periodicity was turned on in the MD simulations. In (B), the unit cell is shown in opaque colors while the nearest neighboring cells are shown in transparent colors to illustrate how RNA appears in the MD simulation. In (C) and (D), different colors were used to emphasize structural transformations observed in RMSD plots (Figure S-5). In (C) and (D), CU base pair conformations interacting with Na⁺ and/or water molecules are shown next to the structure. All conformations of the CU pairs observed in the MD trajectories are shown in Figure S-5 (Table S-8). Note that in (C) and (D) the top and bottom parts of each CU conformation represent the minor and major grooves, respectively.

Figure 6

Binding modes of K-acetyl (K-alkyne mimic) and 2K-4. (A) Lowest free energy binding mode of K-acetyl-r(5’CCCCUGGG/3’GGGUCCCC). Only the 2×2 CU/UC loop region is displayed. The RNA molecule is shown in new ribbon and CPK representations to distinguish it from K-acetyl. Dashed lines represent attractive electrostatic interactions between K-acetyl and loop residues. (B) The binding mode of 2K-4 to 2×CCUG modeled from the lowest free energy binding mode of K-acetyl-5’CCUG/3’GUCC displayed in (A). RNA loop residues and Watson-Crick GC base pairs are highlighted in orange and silver, respectively. The two K RNA-binding modules in 2K-4 are displayed in molecular surface representations. Note that the peptoid backbone is linear in geometry.