Computational and Structural Biotechnology Journal 19 (2021) 51–61 journal homepage: www.elsevier.com/locate/csbj Design of novel small molecule base-pair recognizers of toxic CUG RNA transcripts characteristics of DM1 Raul Ondono a, Ángel Lirio a, Carlos Elvira a, Elena Álvarez-Marimon b, Claudia Provenzano c, Beatrice Cardinali c, Manuel Pérez-Alonso d,e, Alex Perálvarez-Marín b, José I. Borrell a, Germana Falcone c, Roger Estrada-Tejedor a,⇑ a IQS School of Engineering, Universitat Ramon Llull, Barcelona, Spain b Biophysics Unit, Department of Biochemistry and Molecular Biology, School of Medicine, Universitat Autònoma de Barcelona, Cerdanyola del Vallès, Spain c Institute of Biochemistry and Cell Biology, National Research Council, Monterotondo, Rome, Italy d Translational Genomics Group, Incliva Health Research Institute, Valencia, Spain e Department of Genetics and Interdisciplinary Research Structure for Biotechnology and Biomedicine, University of Valencia, Valencia, Spain a r t i c l e i n f o a b s t r a c t Article history: Myotonic Dystrophy type 1 (DM1) is an incurable neuromuscular disorder caused by toxic DMPK Received 20 September 2020 transcripts that carry CUG repeat expansions in the 30 untranslated region (30 UTR). The intrinsic com- Received in revised form 25 November 2020 plexity and lack of crystallographic data makes noncoding RNA regions challenging targets to study in Accepted 28 November 2020 the field of drug discovery. In DM1, toxic transcripts tend to stall in the nuclei forming complex inclu- Available online 06 December 2020 sion bodies called foci and sequester many essential alternative splicing factors such as Muscleblind- like 1 (MBNL1). Most DM1 phenotypic features stem from the reduced availability of free MBNL1 and Keywords: therefore many therapeutic efforts are focused on recovering its normal activity. For that purpose, Myotonic dystrophy Molecular modelling herein we present pyrido[2,3-d]pyrimidin-7-(8H)-ones, a privileged scaffold showing remarkable bio- RNA targeting logical activity against many targets involved in human disorders including cancer and viral diseases. Small molecule Their combination with a flexible linker meets the requirements to stabilise DM1 toxic transcripts, Base recognition and therefore, enabling the release of MBNL1. Therefore, a set of novel pyrido[2,3-d]pyrimidin-7- (8H)-ones derivatives (1a-e) were obtained using click chemistry. 1a exerted over 20% MBNL1 recov- ery on DM1 toxic RNA activity in primary cell biology studies using patient-derived myoblasts. 1a promising anti DM1 activity may lead to subsequent generations of ligands, highlighting a new afford- able treatment against DM1. Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC BY-NC-ND license (http://creative- commons.org/licenses/by-nc-nd/4.0/). 1. Introduction an incurable neuromuscular disorder characterized by symptoms such as progressive muscle weakening, myotonia, cardiac Trinucleotide repeat expansion diseases (TREDs) are character- arrhythmias, cognitive dysfunction, and cataracts [3]. To date, ized by an expansion into a disease-causing range of a short homo- there is no effective treatment against this disease. This fact, along geneous sequence of nucleotides [1]. These microsatellite with the increasing number of cases due to the improvement of expansions may not be toxic per se. In some cases they are located diagnostic tools, is raising awareness of DM1-like diseases [4]. in intronic or intergenic loci; however, they usually undergo tran- The expansion of dCTG repeats within the 30 untranslated (30 - scription giving rise to aberrant transcripts and, in some cases, if UTR) region of the dystrophia myotonica protein kinase (DMPK) these mRNAs are successfully exported outside the nuclei, they gene constitutes the molecular basis of the disease [5]. Somatic may trigger repeat-associated non-AUG (RAN) translation [2]. instability related to this expansion causes progressive phenotypic One representative case of TRED is myotonic dystrophy type 1 worsening as the pathogenic CTG expansion tends to grow with (DM1). DM1 is the most common adult-onset muscular dystrophy, age [6]. Upon that, mutant DMPK transcription produces aberrant non-coding CUG trinucleotide pre-mRNAs [r(CUG)EXP]. These tran- ⇑ Corresponding author. scripts fail to exit to the cytosol and tend to accumulate in the E-mail address: roger.estrada@iqs.url.edu (R. Estrada-Tejedor). nuclei and sequester many essential RNA-binding proteins (RBPs), https://doi.org/10.1016/j.csbj.2020.11.053 2001-0370/Ó 2020 The Authors. Published by Elsevier B.V. on behalf of Research Network of Computational and Structural Biotechnology. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 such as Muscleblind-like 1 (MBNL1) [7], forming complex inclusion 2. Materials and methods bodies called ribonuclear foci [8]. These gel-like structures are able to arrest different RBP when forming foci or being in their diffuse 2.1. Computer-aided molecular design state. Concomitant to MBNL1 sequestration, CUGBP Elav-like family member 1 (CELF1), an MBNL1 antagonist, is upregulated 2.1.1. Receptor preparation in DM1 patients [9]. Both MBNL1 and CELF1 proteins are essential The receptor used in the current study is a CUG16 RNA model alternative splicing factors and their alterations in DM1 maintain a previously developed [28]. The flexible behavior of the system fetal splicing pattern of several muscle transcripts in adults. There- was assessed by molecular dynamics simulations. The system fore, the fetal isoforms are produced instead of the adult ones. was initially heated to 300 K within 500 ps restraining the position Hence, the unbalance in splicing pattern originates disease symp- of the RNA with a 2.0 kcal/molÅ2 force constant. These restraints toms. Most phenotypic features are explained through MBNL1 pro- were gradually reduced along the system’s equilibration stage, per- tein sequestration in the foci, albeit DM1 is a multifactorial disease formed at constant temperature and pressure (1.0 bar, 300 K corre- [10]. Therefore, relieving MBNL1 off the foci, and thus recovering spondingly). A final NPT production stage was conducted during the normal splicing pattern, is a remarkable goal in drug develop- 100 ns without restraints. The Particle Mesh Ewald method was ment against DM1. used for electrostatics interactions under periodic boundary condi- To recover MBNL1 and balance the splicing pattern, many tions and SHAKE [29] algorithm was applied for hydrogen atoms. strategies have been described: [11,12] including gene editing, The time step was fixed to 2 fs and simulations were conducted transcription repression or chemotherapy (i.e.: antisense oligonu- using AMBER18 software (University of California, San Francisco, cleotides, peptides, and small molecules) [13–18]. Currently, the CA). Upon that, a set of frames exhibiting opening values similar small molecule approach is undoubtedly the preferred pharmaco- to the ones expected from alternate UU mismatches were selected logical strategy to treat any kind of disease due to its robustness using the cpptraj [30] module. and cost-effectiveness ratio. In the last few years, some molecular entities have been described to exert a significant activity against 2.1.2. Pharmacophore model and docking DM1. They can be classified into few major groups based on their A pharmacophore model was developed using Molecular Oper- characteristic mechanisms of action: compounds able to bind ating Environment (MOE) software (Chemical Computing Group, DNA and repress transcription, inhibitors able to recover missplic- Montreal, QC), to evaluate the ability of the ligands to establish a ing events by CELF1 repression, and compounds able to recover Janus-Wedge interaction pattern with two UU mismatches. The MBNL1 from the foci by targeting RNA [18–20]. Focusing on the generated model involved three consecutive UU mismatches, foci, only three strategies have led to compounds able to achieve each of them led to 5 pharmacophore features (three H-bond selectivity and potency: groove binding, base intercalation, and donors and two H-bond acceptors) (Fig. S1). Thereafter, pharma- base recognition [18,19,21]. Among them, and given the charac- cophore guided docking was performed using MOE under induced teristic regular structure of DM1 transcripts, base recognizers rep- fit conditions. We considered two different score functions (Lon- resent the most revolutionary strategy in terms of having a don dG and GBVI/WSA dG) for the docking study. The best confor- comprehensive action mechanism. They act mimicking nucleo- mation for each ligand was selected for the subsequent tides and stablishing hydrogen bond interactions with the target calculations. nucleic acids. This leads to a secondary structure stabilization in double-stranded conformation of CUG. As previously described, stabilization of r(CUG)EXP triggers MBNL1 recovery as MBNL1 2.1.3. Molecular dynamics and data analysis colocalizes with CUG in foci [22]. Such an accomplishment has Ligands were prepared in UCSF-Chimera [31], calculating AM1- also been reported via uridine replacement, antisense oligonu- BCC atomic charges and using the GAFF [32] forcefield. Complex cleotides treatment [15]. structures were generated using tleap, defining the Amber14 and In order to adapt base recognition strategy to exogenous drug- Rochester torsions forcefields [33] for the RNA. Complexes were like structures a common paradigm is used. Candidates must be neutralized and solvated using the OPC water model in a truncated able to bridge non-canonical Watson-Crick base pairs, the so- octahedral box [34]. All molecular dynamics simulations were con- called Janus-Wedge interaction (JW) [23]. The interaction of these ducted using pmemd.cuda module in Amber18.[35–37] and run molecules with the foci can turn the unstructured gel-like inclu- through the following steps. The system was initially heated to sion bodies to regular double stranded conformation (dsCUG). An 300 K within 500 ps restraining the position of the RNA with a 2. example of base recognizers is the Zimmerman’s bisamidinium lin- 0 kcal/molÅ2 force constant. These restraints were gradually ker (Z), which is included in this study as reference [24]. The target reduced along the system’s equilibration stage, performed at con- nucleobases to fit in the recognizers are the UU mismatches that stant temperature and pressure (1.0 bar, 300 K correspondingly). are generated when dsCUG conformation is achieved [25]. By sta- A final NPT production stage was conducted during 100 ns without bilizing the irregular structure of the foci, the toxic r(CUG)EXP may restraints. The Particle Mesh Ewald method was used for electro- return to its soluble and non-pathogenic form, relieving MBNL1. statics interactions under periodic boundary conditions and SHAKE We applied a de novo computer-aided drug design procedure to [29] algorithm was applied for hydrogen atoms. The time step was predict the ability of novel compounds to bind to r(CUG)EXP. The set to 2 fs. RMSD analysis for the full-length trajectories was con- consideration of Janus-Wedge interactions as the base mechanism ducted using ccptraj module and MMPBSA binding free energies for CUG recognition led us to the identification of a novel active were calculated for the last 20 ns of simulation. RNA structural core based on pyrido[2,3-d]pyrimidine derivatives. Pyrido[2,3-d] parameters analysis was performed using the X3DNA software pyrimidine is a privileged scaffold with notable biological activity [38]. Data analysis and plotting were performed using R (R Founda- in fields such as cancer or viral diseases [26,27]. Herein, we report tion for Statistical Computing, Vienna, AT). a set of new multivalent compounds, composed by two pyrido[2,3- d]pyrimidin-7-(8H)-ones units linked by an aliphatic spacer 2.2. Chemistry (Fig. 1a), able to target two UU mismatches in r(CUG)EXP struc- tures via Janus-Wedge interactions (Fig. 1b). The effect of the 2.2.1. General considerations spacer length on the ability to rescue MBNL1 was assessed by both All solvents and chemicals were reagent grade. Unless other- computational and biological techniques. wise mentioned, all solvents and chemicals were purchased from 52 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 Fig. 1. General structure of ligands proposed on current work. a) All structures are characterized for bearing a pyrido[2,3-d]pyrimidine scaffold at both terminal ends separated by a series of aliphatic spacers of different lengths. Both terminal scaffolds are bound to the central spacer with amide bond. b) Janus Wedge interaction mechanism described between an UU mismatch and the introduced pyrido[2,3-d]pyrimidine scaffold, allowing base recognition of singular nucleic acids such as dsCUG. commercial vendors (Sigma Aldrich, ABCR, Fluorochem, Apollo sci- was dried over MgSO4, filtered, and evaporated under reduced entific and ACROS Organics) and used without further purification. pressure to afford 6 (0.61 g, 2.7 mmol, 87% yield) as a brown oil. 1 H and 13C and spectra were recorded on a Varian 400-MR spec- 1 H NMR (400 MHz, CDCl3): d 4.23 (dd, J = 9.8, 5.7 Hz, 1H), 3.74 trometer (1H NMR at 400 MHz and 13C NMR at 100.5 MHz). Chem- (s, 3H), 3.70 (s, 3H), 3.09 – 2.99 (m, 1H), 2.74 (ddt, J = 22.8, 17.2, ical shifts were reported in parts per million (d) and are referenced 6.0 Hz, 2H), 2.50 (ddd, J = 13.9, 10.2, 5.7 Hz, 1H), 2.20 (ddd, to the residual signal of the solvent DMSO d6 2.50 ppm or in 1H J = 13.9, 9.8, 4.0 Hz, 1H); 13C NMR (100.6 MHz, CDCl3): d 172.5, NMR spectra and to the residual signal of the solvent DMSO d6 171.0, 112.5, 112.3, 52.8, 52.3, 38.0, 35.5, 32.1, 21.3; IR (KBr) 39.5 ppm in 13C NMR). Coupling constants are reported in Hertz mmax(cm1): 2959, 2928, 1735, 1440, 1376, 1214, 1170. (Hz). Standard and peak multiplicities are designed as follows: s, methyl 2-(2,4-diamino-7-oxo-5,6,7,8-tetrahydropyrido[2,3-d] singlet; d, doublet; dd, doublet of doublets; ddd, doublet of dou- pyrimidin-6-yl)acetate (4). Compound 3 (1.6 g, 10.3 mmol), blets of doublets; t, triplet; tt, triplet of triplets; br, broad signal. malononitrile (0.35 g, 5.2 mmol) and guanidine carbonate IR spectra were recorded in a Thermo Scientific Nicolet iS10 FTIR (0.93 g, 5.2 mmol) were added to a microwave vial with 20 mL spectrophotometer with Smart iTr. Wavenumbers (m) are reported of anhydrous MeOH. The mixture was heated at 140 °C under in cm1. MS data (m/z (%), EI, 70 eV) were obtained by using an Agi- microwave irradiation for 15 min. The resultant solid was collected lent Technologies 5975. HRMS data were obtained by using a Bru- by filtration, washed with H2O and Et2O and oven dried to afford ker micrOTOF (ESI-FIA-TOF). Elemental microanalyses were 0.93 g (3.7 mmol, 71% yield) of 4 as yellow solid. obtained on a EuroVector Instruments Euro EA 3000 elemental Or alternatively, guanidine carbonate (1.08 g, 6.02 mmol) and analyzer. The melting points were determined with a Büchi- sodium methoxide (650 mg, 12.0 mmol) were refluxed in 4.5 mL Tottoli 530 capillary apparatus and are uncorrected. Automatic of anhydrous MeOH under Ar atmosphere for 15 min. The mixture flash chromatography was performed in an Isco Teledyne Combi- was filtered to eliminate the formed sodium carbonate. To the flash Rf medium pressure liquid chromatograph with RediSepÒ sil- resulting solution, 6 (900 mg, 4.0 mmol) was added. Then, the mix- ica gel columns (35–70 mm) using a suitable mixture of solvents as ture was refluxed under Ar atmosphere overnight. The formed eluent. Microwave irradiation experiments were carried out in an solid was filtered and washed with water, EtOH and Et2O. Methyl Biotage InitiatorTM microwave apparatus, operating at a frequency 2-(2,4-diamino-7-oxo-5,6,7,8-tetrahydropyrido[2,3-d]pyrimidin- of 2.45 GHz with continuous irradiation power from 0 to 400 W. 6-yl)acetate (4) (520 mg, 2.1 mmol, 52% yield) was obtained as a Reactions were carried out in 5, or 20 mL glass tubes, sealed with yellow solid. mp = 276.8 °C; 1H NMR (400 MHz, DMSO d6): d aluminum/Teflon crimp tops. Compound 3 and Z were prepared 10.29 (s, 1H), 6.17 (s, 2H), 5.85 (s, 2H), 3.62 (s, 3H), 2.89–2.70 according to literature [24,39]. (m, 3H), 2.47–2.43 (m, 1H), 2.26 (dd, J = 15.3, 13.0 Hz); 13C NMR dimethyl 2-methylenesuccinate (3)0.4 purified by flash chro- (100.6 MHz, DMSO d6): d 172.4, 172.0, 161.7, 156.4, 83.7, 51.4, matography (Cy: AcOEt, 80:20) 2.94 g (19 mmol, 48% yield) was 36.8, 34.0, 23.0; IR (KBr) mmax(cm1): 3381, 3346, 3189, 1717, afforded as a colorless oil. 1H NMR (CDCl3, 400 MHz): d 6.18 (s, 1683, 1634, 1570, 1493, 1445, 1391, 1364, 1296, 1237, 1223, 1H), 5.60 (s, 1H), 3.66–3.61 (m, 3H), 3.59–3.54 (m, 3H), 3.21 (s, 1159; MS (EI) m/z: 251.1 (7%) [M + 1]+, 178.1 (100%). 2H). Spectroscopic data are consistent with those reported in 2-(2,4-diamino-7-oxo-5,6,7,8-tetrahydropyrido[2,3-d]pyrimidin- reference. 6-yl)acetic acid (5). Compound 4 (0.500 g, 2.0 mmol) was dissolved dimethyl 2-(2,2-dicyanoethyl)succinate (6). NaH (0.13 g, in NaOH 0.5 M (7 mL) and the solution was stirred for 21 h. Reac- 5.41 mmol) was dissolved in anhydrous THF (5 mL) and the result- tion mixture was acidified with conc. HCl, then, the mixture was ing mixture was stirred for 10 min at room temperature. Then, filtered and the resulting solid was washed with H2O and oven malononitrile (0.21 g, 2.14 mmol) was dissolved in anhydrous dried to afford 2-(2,4-diamino-7-oxo-5,6,7,8-tetrahydropyrido[2,3 THF (5 mL) and added to the mixture that was further stirred for -d]pyrimidin-6-yl)acetic acid (0.298 g, 1.26 mmol, 63% yield) as a 20 min at room temperature. After that, 4 (0.50 g, 3.16 mmol) white solid. mp > 300 °C; 1H NMR (400 MHz, DMSO d6): d 12.25 was dissolved in anhydrous THF (5 mL) and added to the mixture (s, 1H), 10.74 (s, 1H), 7.74 (s, 2H), 7.53 (s, 2H), 2.94–2.80 (m, 2H), that was stirred for 2 h at 40 °C. After this period, water was added. 2.71 (dd, J = 16.9, 4.9 Hz, 1H), 2.43 (dd, J = 17.0, 6.6 Hz, 1H), 2.35 The resulting mixture was acidified with concentrated acetic acid (dd, J = 14.8, 12.9 Hz, 1H); 13C NMR (100.6 MHz, DMSO d6): d and extracted with dichloromethane (3  15 mL). The organic layer 178.2, 175.6, 159.3, 154.7, 83.1, 41.1, 37.5, 24.9; IR (KBr) 53 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 mmax(cm1): 3444, 3368, 3160, 1698, 1646, 1397; HRMS (m/z): 2.3. Biochemistry [M + 1]+ calcd. for C9H12N5O3: 238.0935; found: 238.0935. 2.3.1. Protein purification Competent BL21(DE3)pLysS E. coli (Invitrogen, Thermo Scien- tific, Carlsbad, CA) were transformed with the pGEX-6P-MBNL1- 2.2.2. General procedure for the synthesis of ligands 1a-e D105-His plasmid (GE Healthcare & Invitrogen) [40]. A single 1 eq 5, 1 eq of the corresponding aliphatic diamine, 2.2 eq colony of the transfected BL21 was picked into 10 mL of LB with EDCHCl and 2.2 eq DMAP are dissolved in anhydrous DMF 100 lg/mL ampicillin and grown for 16 h at 37 °C and 200 rpm. (2.5 mL) and the mixture is stirred for 24 h at room temperature. 1 mL of the resulting culture mixture was inoculated to 1 L of LB After that, the reaction mixture is concentrated under reduced with 100 lg/mL ampicillin and incubated at 37 °C and 150 rpm pressure. The resulting solid is resuspended in HCl 0.5 M to afford in a shaking incubator until OD600nm reached 0.5–0.6. The cell cul- the desired diamides by filtration. ture was then induced with 500 lM IPTG for 3 h at 30 °C. There- N,N’-(pentane-1,5-diyl)bis(2-(2,4-diamino-7-oxo-5,6,7,8-tetrahy after, the culture media was centrifuged at 4000 xg and 4 °C for dropyrido[2,3-d] pyrimidin-6-yl)acetamide) (1a). 60 mg (0.11 mmol, 30 min*. The supernatant was discarded and the pellet was resus- 44% yield). mp > 300 °C; 1H NMR (400 MHz, DMSO d6): d 10.48 (s, pended in 50 mL of cell lysis buffer (150 mM NaCl, 50 mM Tris-HCl 2H), 7.98 (t, J = 5.5 Hz, 2H), 7.36 (s, 4H), 7.07 (s, 4H), 3.10–2.99 (m, (pH 8.00), 5 mM imidazole, 0.1% IGEPAL CA-630, 5% glycerol and 4H), 2.92–2.84 (m, 2H), 2.73 (dd, J = 15.6, 7.0 Hz, 2H), 2.60 (dd, freshly added: 0.1 M PMSF, 5 mM benzamidine, 1 EDTA-free com- J = 15.4, 4.5 Hz, 2H), 2.30–2.19 (m, 4H), 1.44–1.37 (m, 4H), 1.44– plete protease inhibitors tablet, 1 mM DNAse and 1 mg/mL lyso- 1.23 (m, 2H); 13C NMR (100.6 MHz, DMSO d6): d 172.8, 169.7, zyme). The cell culture was lysed by alternating 15 cycles of 154.9, 83.5, 38.6, 36.2, 35.2, 28.8, 23.9, 22.2; IR (KBr) mmax(cm1): sonication (1 min sonication – 1 min icing). To remove cell debris, 3361, 2932, 1648, 1569, 1495, 1462, 1395, 1329, 1274, 1226, the sample was centrifuged at 10,000 xg and 4 °C for 50 min and 1164, 1088, 762, 600; HRMS (m/z): [M + 1]+ calcd. for the supernatant was collected for purification. The supernatant C23H33N12O4, 541.2742; found, 541.2743. was added to 2 mL of a previously equilibrated TALONÒ Metal N,N’-(heptane-1,7-diyl)bis(2-(2,4-diamino-7-oxo-5,6,7,8-tetrahy Affinity Resin (Takara Bio, Kusatsu, JP) and incubated at 4 °C for dropyrido[2,3-d] pyrimidin-6-yl)acetamide) (1b). 80 mg (0.14 mmol, 1 h. The resin was washed three times with 10 column volumes 57% yield). mp > 300 °C;1H NMR (400 MHz, DMSO d6): d 10.63 (s, (30 mL  3 times) of washing buffer (150 mM NaCl, 50 mM Tris- 2H), 7.88 (t, J = 5.6 Hz, 4H), 7.56 (s, 4H), 7.36 (s, 4H), 3.10–2.96 HCl (pH 8.00), 25 mM imidazole, 0.1% IGEPAL CA-630, 5% glycerol). (m, 4H), 2.91–2.85 (m, 2H), 2.74 (dd, J = 15.4, 6.9 Hz, 2H), 2.60 The protein was eluted with 5 mL of elution buffer (150 mM NaCl, (dd, J = 15.5, 4.5 Hz, 2H), 2.32–2.19 (m, 4H), 1.40–1.37 (m, 4H), 50 mM Tris-HCl (pH 8.00), 250 mM imidazole, 0.1% IGEPAL CA-630, 1.25 (s, 6H); 13C NMR (100.6 MHz, DMSO d6): d 172.8, 169.6, 5% glycerol). The eluates were added to PD-10 Desalting Columns 153.5, 83.4, 39.1, 36.0, 35.0, 29.1, 28.5, 26.4, 22.0; IR (KBr) (GE Healthcare, Chicago, IL) that was previously preequilibrated mmax(cm1): 3383, 2927, 2853, 1642, 1497, 1387, 1318, 1270, using fluorescence test buffer (25 mM HEPES pH 7.4, 110 mM 1226, 1162, 1086, 818, 761, 534; HRMS (m/z): [M + 1]+ calcd. for KCL, 10 mM NaCl, 1 mM MgCl2, 15 lM ZnCl2, 0.02%Tween-20). C25H37N12O4, 569.3055; found, 569.3052. Protein purity was assessed via Comassie staining of SDS-PAGE N,N’-(octane-1,8-diyl)bis(2-(2,4-diamino-7-oxo-5,6,7,8-tetrahy electrophoresis 10% polyacrylamide gel (Figure S4). Finally, GST- dropyrido[2,3-d] pyrimidin-6-yl)acetamide) (1c). 76 mg (0.13 mmol, MBNL1-His6 was quantified via direct absorbance spectroscopy at 52% yield). mp > 300 °C; 1H NMR (400 MHz, DMSO d6): d 10.14 (s, 280 nm using the ExPASy Protparam (protein parameter) tool to 2H), 7.82 (t, J = 3.8 Hz, 2H), 6.14 (s, 4H), 5.82 (s, 4H), 3.09–2.99 (m, assess extinction coefficient (57300 M1cm1) [41]. GST-MBNL1- 4H), 2.85–2.78 (m, 2H), 2.70–2.59 (m, 4H), 2.26–2.13 (m, 4H), His6 samples were stored at 80 °C with 50% glycerol. Prior to each 1.40–1.38 (m, 4H), 1.25 (s, 8H); 13C NMR (100 MHz, DMSO d6) d use, samples were ultracentrifuged at 150,000 xg, 4 °C for 1 h and 173.3, 170.0, 161.8, 161.6, 156.3, 83.7, 38.6, 36.9, 35.7, 29.1, 28.8, quantified to avoid aggregation issues. 26.5, 23. IR (KBr) mmax (cm1): 3372, 3192, 2925, 1643, 1598, 1569, 1506, 1458, 1387, 1277, 1225, 1093, 1015, 815, 760, 678, 2.3.2. AID 2675 536, 436; HRMS (m/z): [M + 1]+ calcd. for C26H39N12O4, The previously described assay was adapted to RNase pre- 583.3212; found, 583.3211. treated 96-well plates (Greiner, Kremsmünster, AT). Reagents were N,N’-(nonane-1,9-diyl)bis(2-(2,4-diamino-7-oxo-5,6,7,8-tetrahy premixed and added to the wells containing compounds solution dropyrido[2,3-d] pyrimidin-6-yl)acetamide) (1d). 80 mg (0.14 mmol, (20 nM Biot-(CUG)6, 20 nM MBNL1-His6, 0,11 ng/mL anti-His-Tb, 57% yield). mp > 300 °C; 1H NMR (400 MHz, DMSO d6): d 10.20 (s, 10 mM SA-XL665 and 0,1 mM tested compound). All reagents were 2H), 7.80 (t, J = 5.5 Hz, 2H), 6.15 (s, 4H), 5.85 (s, 4H), 3.10–2.98 (m, diluted in binding buffer composed of 25 mM HEPES pH 7.4, 4H), 2.87–2.79 (m, 2H), 2.67–2.55 (m, 2H), 2.26–2.12 (m, 4H), 110 mM KCl, 10 mM NaCl, 1 mM MgCl2, 15 lM ZnCl2, 0.02% 1.40– 1.38(m, 4H), 1.25 (s, 10H); 13C NMR (100 MHz, DMSO d6) d Tween-20, 0.1% BSA and 5 mM freshly added DTT. After 60 min 172.8, 169.6, 164.1, 83.5, 38.6, 36.1, 35.1, 29.1, 28.8, 26.5; IR incubation at room temperature, the fluorescence was measured (KBr) mmax (cm1): 3338, 2925, 2852, 1646, 1565, 1494, 1465, using a BMG Labtech Fluostar Optima microplate reader. Samples 1389, 1320, 1269, 1226, 1165, 1087, 818, 762, 668, 535; HRMS were irradiated at 340 nm, and fluorescence was read at 545 and (m/z): [M + 1]+ calcd. for C27H41N12O4, 597.3368; found, 597.3370. FRET at 665 nm with a time gap of 30 ms and an integration time N,N’-(decane-1,10-diyl)bis(2-(2,4-diamino-7-oxo-5,6,7,8-tetrahy of 1500 ms. dropyrido[2,3-d] pyrimidin-6-yl)acetamide) (1e). 80 mg (0.14 mmol, 57% yield). mp > 300 °C; 1H NMR (400 MHz, DMSO d6): d 10.54 (s, 2.4. Cell biology 2H), 7.88 (t, J = 5.6 Hz, 2H), 7.34 (s, 4H), 7.10 (s, 4H), 3.12–2.95 (m, 4H), 2.90–2.84 (m, 2H), 2.76–2.70 (m, 2H), 2.63–2.58 (m, 2H), 2.4.1. Cell culture 2.28–21.8 (m, 4H), 1.39 (t, J = 6.8 Hz, 4H), 1.25 (s, 12H); 13C NMR Immortalized human myoblasts, control and DM1 derived from (100 MHz, DMSO d6) d 173.0, 169.7, 155.8, 83.5, 38.7, 36.3, 35.2, primary dermal fibroblasts of control and DM1 affected individuals 29.1, 29.0, 28.8, 26.5, 22.3; IR (KBr) mmax (cm1): 3382, 2923, respectively, were used as described in previous work [13]. Fibrob- 2851, 1641, 1566, 1461, 1387, 1320, 1271, 1224, 1161, 1087, lasts were sequentially infected with retroviral vectors carrying 816, 762, 534; HRMS (m/z): [M + 1]+ calcd. for C28H43N12O4, TERT and Hygromycin selection (Addgene #1773, Cambridge, 611.3525; found, 611.3525. MA) [42], and with retroviruses carrying estrogen-inducible mouse 54 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 Myod1 and puromycin selection (Addgene #13494) [43]. Cells were then exposed to primary antibodies overnight at 4 °C were propagated in dishes coated with collagen from rat tail at (MBNL1, CUGBP1, myogenin, myosin). Subsequently, membranes 5 mg/cm2 (Gibco, Thermo Fisher Scientific, Waltham, MA), using were washed 3 times in TBST and incubated in horseradish-perox DMEM without phenol red (Gibco, Thermo Fisher Scientific, Wal- idase-conjugated anti-mouse antibody solution at room tempera- tham, MA) supplemented with 15% FBS (Sigma-Aldrich, St. Louis, ture for 1 h. Finally, Chemiluminescence detection was achieved MO) and incubated for 48 h. Differentiation to myotubes was using ETA C 2.0 and NOVA 2.0 reagents (Cyanagen, Bologna, Italy) induced by growing cells to confluency and replacing the prolifer- in a ChemiDoc XRS+ (Bio-Rad, Hercules, CA). Imaging and quantita- ation medium with differentiation medium consisting of DMEM tion of the bands were carried out by the ChemiDoc XRS Western without phenol red supplemented with 10-7 M b-estradiol Blot Imaging System using the ImageLab 4.0 software (Bio-Rad, (Sigma-Aldrich) and 2% horse serum and incubating for 24 h. All Hercules, CA). Equal loading of proteins onto the gel was confirmed cells were incubated under a 5% CO2 atmosphere at 37 °C. For by immunodetection of vinculin using Monoclonal Ab aVinculin PCR and WB experiments 35 mm dishes were seeded at 1.3103 (Sigma-Aldrich, St. Louis, MO) at room temperature for 1 h. cells/cm2 whereas for FISH and IF cells where seeded on glass cov- erslips at 3.2103 cells/cm2. DM1 cells were treated with the com- 2.4.4. PCR pounds at 10 mM for 48 h in all experiments. RNAs from both control and patient cell lines were extracted with TRIzol reagent (Invitrogen, Thermo Scientific, Carlsbad, CA). 2.4.2. FISH IF Cells were washed twice with PBS and frozen at 80 °C for Treated cells were washed twice in PBS for 5 min. After that, 30 min. After that, TRIzol reagent was added and cells were cells were fixed with 2% formaldehyde for 30 min at 4 °C. After fix- scraped off the plates, transferred into Eppendorf tubes and incu- ation, coverslips were washed 3 times in PBS, and permeabilized in bated at room temperature for 5 min. Chloroform was added and 0.4% TritonTM X-100 (Sigma-Aldrich) in PBS for 10 min. After wash- the resulting mixture was vigorously shaken and incubated ing with PBS 3 times, coverslips were incubated in 40% formamide 3 min at room temperature. The aqueous phase was separated by and 2x SSC for 10 min at room temperature and then hybridized centrifugation, and RNA precipitated by adding IPA. The sample with (CAG)6 probe labeled with Texas Red at the 50 end (IDT, Cor- was centrifuged at 12,000 xg for 10 min at 4 °C. The supernatant alville, IA) (1 ng/lL for 2 h at 37 °C in 30% formamide, 2x SSC, 0.02% was discarded, and the RNA was washed twice with ethanol 80% BSA, 67 ng/lL yeast tRNA, 2 mM vanadyl ribonucleoside complex). prechilled at 20 °C. After drying, samples were dissolved in Cells were then washed first in 40% formamide and 2x SSC at 45 °C RNAse free water, incubated 5 min at 60 °C, cooled in ice and quan- for 30 min. After that, coverslips were washed sequentially in 2x tified using Nanodrop ND-1000 (Nanodrop Technologies, Wilming- SSC and 1x SSC at 45 °C for 15 min each. Then, samples were incu- ton, DE). After that, RNA samples were retro transcribed with the bated with anti-MBNL1 3A4 antibody (Santa Cruz Biotechnology) GoScript Reverse Transcription System (Promega, Fitchburg, WI). solution in PBS at room temperature for 1 h. After that samples Oligo (dT) and random primers were preincubated with RNA for were washed 3 times with PBS and incubated using goat anti- 5 min at 70 °C. The resulting mixture was chilled in ice for mouse IgG-Alexa Fluor 488 (Invitrogen, Thermo Scientific, Carls- 5 min. Next, reactions were prepared by adding 3 mM MgCl2, bad, CA). After that, samples were incubated with Hoechst PCR Nucl. Mix, Rnasin, Reverse Transcriptase and placed in the 33,258 (Invitrogen) in PBS at room temperature for 10 min. Finally, thermal cycler at 42 °C for 1 h, 70 °C for 15 min and finally chilled. coverslips were washed 3 times in PBS, water and mounted on The resulting cDNA samples were amplified with GoTaq Flexi DNA glass slides with ProLongTM Gold Antifade Mountant (Invitrogen). Polymerase (Promega, Fitchburg, WI) for 35 cycles using the speci- Samples were examined with an Olympus AX70 immunofluores- fic primers (Table S2). Upon reaction ending, samples were loaded cence microscope. Images were recorded on an Olympus XM10 on agarose gel containing Midori staining (Nippon Genetics, Düren, camera and processed using the Olympus CellSens Standard 1.8.1 GE). Quantitation of amplified bands on gel images was performed software. Cell quantification was conducted using ImageJ software. using UVITec1D software. Exon inclusion was quantified as the percentage of the total intensity of both isoforms. 2.4.3. Western blot Treated cells were washed twice with PBS for 5 min. After that, 2.4.5. Statistics cells were ice cooled and lysed by adding RIPA buffer (50 mM Tris All biological experiments were performed at least in triplicate, HCl pH 7.4, 150 mM NaCl, 0.5% Sodium deoxycholate, 0.1% SDS, 1% and pictures represent typical examples. Data are presented as NP40, 1 mM EDTA, 5 mM NaF, 5 mM Na3VO4 and Protease inhibi- mean ± standard deviation (SD). A two-tailed unpaired t-test was tor cocktail (Sigma-Aldrich, St. Louis, MO). After 5 min of incuba- applied for assessment of the statistical significance of differences tion at 0 °C, cells were scraped, and the lysate was transferred to between two groups. A p-value lower than 0.05 was deemed statis- an Eppendorf tube. The extract was centrifuged at 14,000 xg, 4 °C tically significant. Statistical analyses were conducted using for 10 min to remove cell debris. The supernatant was transferred GraphPad Prism version 7.01 (GraphPad Software, Inc., La Jolla, to a new Eppendorf tube and total protein concentration was CA, USA). assessed using the Micro BCA Protein Assay Kit (Thermo Scientific, Waltham, MA). Homogeneous protein samples mixed with Laemmli sample buffer and Novex Nupage sample reducing agent 3. Results (Invitrogen, Thermo Scientific, Carlsbad, CA) were loaded and sep- arated using Bolt 4–20% Tris-glycine SDS-PAGE (Invitrogen, 3.1. Rational design Thermo Scientific, Carlsbad, CA). Gels were run in Novex Nupage MOPS buffer (Invitrogen, Thermo Scientific, Carlsbad, CA) setting The binding mechanism of 1a-1e compounds was assessed by voltage to 220 V for 40 min. After gel running, proteins were trans- combining molecular docking and molecular dynamics simula- ferred to PVDF or nitrocellulose membranes using a stack soaked in tions, using a short hairpin CUG16 (shCUG) model as receptor transfer buffer (25 mM Tris, 190 mM glycine, 10% methanol) and [28]. Although allowing moderate motion of the side chains, flexi- setting current at 350 mA for 90 min at 4 °C. After checking correct ble docking does not guarantee the generation of Janus-Wedge transferring using ponceau staining, membranes were blocked for arrangement. Given the difficulty in dealing with static structures, 1 h with 5% non-fat dry milk in Tris-buffered saline containing molecular dynamics simulations were applied to generate an 0.05% Tween 20 (TBST). After 3 washes with TBST, membranes ensemble of shCUG conformations. The length of the hairpin model 55 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 ensures a sufficiently large double-stranded conformation for biva- been evaluated against DM1, it is desirable to assess whether the lent ligands to bind consecutive or non-consecutive uridine units. primary scaffolds are active or not, without any further interac- It is also short enough to allow long production times in MDs, tions from the moiety core. For this reason, linkers were main- which is a critical point in this study. Dynamics of non-canonical tained as simple as possible although it is a drawback in terms of UU pairs were therefore analyzed to capture the preferred confor- total binding energy. mation able to stabilize JW interactions. Docking was The analysis of structural parameters in the hairpin base-pairs pharmacophore-guided considering that base recognizers tend to evidenced that most of the compounds under study increase the fall in base pairs via major groove (Fig S1) [24]. The pyrido[2,3-d] stability of the system by reducing the dispersion of structural pyrimidine unit fulfils the requirements to act as uracil recognizer, parameters in the UU mismatch and its vicinities (Fig. S3), partic- establishing the desired hydrogen bond (Fig. 2a). ularly of the opening value (Fig. 2c). From all the structural param- The stability of docked complexes was assessed by molecular eters considered, opening singularly represents the instability of a dynamics (MD, Fig. S2). The spacer length determines the binding generated UU mismatch and it gives valuable data when a Janus- mechanism of pyrido[2,3-d]pyrimidine derivatives and their ability Wedge interaction mechanism is established (inducing an opening to act as bivalent ligands (targeting two UU pairs); shorter linkers value shift about 60°). would not be long enough to reach both mismatches while longer ones could trigger swapping the interaction to a more distant base 3.2. Synthesis pair. The stability of all complexes was assessed by studying the structural parameters and quantified through MMPBSA analysis. A standard procedure to synthesize the pyrido[2–3-d]pyrim- Variations in the binding mechanism would explain the differ- idine scaffold [44] has been adapted to fulfil the structural features ence in Poisson-Boltzmann Gibbs free energy of binding (D°GPB), required for Janus-Wedge interaction: (a) leave the pyrimidinic predicted by MMPBSA (Fig. 2b). Results show a clear dependence amino groups unsubstituted (since these amines are meant to play of the spacer length, highlighting 1a and 1e (-37 ± 5 kcal/mol a critical role in hydrogen bonding between ligand and receptor) and 50 ± 4 kcal/mol respectively). The pharmacophore allows and (b) to be able to derivate the core with a reactive group, to bind both consecutive and non-consecutive binding for the ligands. the linker, without compromising the amines. We chose to intro- Short ligands (1a) can achieve higher binding energies via consec- duce a carboxylic acid group at the pyridone ring due to the widely utive mismatch interaction whereas longer ones reach alternate reported robustness of the click-chemistry reactions between car- mismatches (1d, 1e). The nature of the spacer would account for boxylic acids and aliphatic amines even in complex systems as our que significant energy gap between pyrido[2,3-d]pyrimidine extremely polar pyrido[2–3-d]pyrimidines. Upon that, a full novel derivatives (1a-1e) and the positive control (Z). The bisamidinium synthetic pathway was designed and performed to achieve the linker gave not only the desired distance between active sites of desired compounds (Fig. 3). the molecule but also a boost in unspecific binding energy while The above-mentioned synthesis leads to compounds 1a-e in a linker moieties used in the current work lack of any functional 4-step synthesis. Starting from the commercially available itaconic groups. As this was the first time that this set of structures have acid (2) a Fischer esterification using MeOH and H2SO4 renders the Fig. 2. Molecular mechanics results summary. a) Representative snapshot taken from MD showing the interaction mechanism between 1a compound and shCUG b) MM- PBSA values of free Gibbs energy binding calculated over 20 ns of each ligand in their respective MD. c) Structural parameter analysis of opening values. Full-length MD is displayed in boxplot graph format. 56 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 Fig. 3. General synthetic pathway to attain structures 1a-e. a) H2SO4/MeOH reflux 21 h. b) Guanidine carbonate, malononitrile in MeOH MW 140 °C 15 min. c) NaOH 0.5 M RT 21 h + conc HCl. d) aliphatic diamine, EDCHCl, DMAP, and K2CO3 in DMF RT 24 h. methyl diester 3 in 48% yield. The next step is to build the hetero- the complex is formed (Fig. 4a). If the complex is broken by a can- cyclic core using our two-step synthetic protocol. First a Michael didate, fluorescence intensity falls as the fluorophores are no addition of malononitrile to the a, b-unsaturated ester 3 and the longer near enough. Thus, FRET does not occur. resulting adduct is reacted with guanidine to obtain the corre- To obtain the recombinant MBNL1, transfected E. coli were sponding pyridopyrimidine 4 (45% aggregate yield). However, the grown, and the purified recombinant protein was harvested prior a, b-unsaturated ester 3 is also suitable to use our microwave- its use in the test. Minor changes to the original protocol allowed assisted multicomponent reaction to obtain pyrido[2,3-d]pyrim- achieving the final recombinant protein with fewer impurities idines. Thus, 3 is placed in a sealed vial together with 0.5 eq. of and to avoid aggregation issues (Fig. S4). The whole preparation malononitrile and 0.5 eq of guanidine carbonate in anhydrous process was conducted in RNase free conditions and all reagents methanol and brought to 140 °C for 15 min to give pyrido[2,3-d] were RNase free grade to avoid synthetic oligo CUG to degrade pyrimidine 4 in 71% yield. Once the main base-pair recognizing before measuring the protein displacement. moiety is formed, the next steps are to prepare and bind the alipha- Compound concentration was set at a very low range to avoid tic linker. First, the carboxymethyl group is saponified to achieve fluorescent interactions and possible precipitation issues. How- the corresponding carboxylic acid. Thus, 4 is dissolved in 0.5 M ever, it is interesting to see that even in these conditions structure aqueous NaOH and stirred at room temperature for 21 h, and the 1a can bind tightly to RNA displacing recombinant MBNL1 and reaction mixture is acidified to afford 5 in 63% yield. Ligands and therefore reducing the FRET intensity a 33% compared to the neg- linkers are bound via Steglich amidation using EDCHCl as coupling ative control (Fig. 4b). The rest of the candidates also exhibit some agent and DMAP as catalyst. Given the fact that amino groups at C2 MBNL1 displacing potency. These results encouraged us to further and C4 of the pyrimidine ring of 5 are much less nucleophilic than study these candidates in patient-derived cells. the aliphatic primary amines, the only expected by-products of the coupling were mono-aminated derivatives. Pyridopyrimidines 1a-e hamper classical purification tech- 3.4. Cell biology niques (i.e. flash chromatography) due to its high MW and polarity. The sidechain unsubstituted diamino pyridopyrimidines tend to In order to determine the drug efficacy and gain understanding form stable hydrogen bonds with other pyridopyrimidines giving of the behavior of compounds 1a-e in a complex system, we per- stable microcrystalline structures that are barely soluble in water. formed the biological characterization of the candidates. Dermal For this reason, most reagents and reaction by-products must be fibroblasts of both healthy and DM1 affected individuals were used eliminated via washes instead of classical flash column chromatog- to run the experiments. Both cell lines were previously immortal- raphy. This plays a critical role in a double amidation where poten- ized by infection a with retroviral vector carrying the human tial multiple species can be obtained (product, by-products, excess telomerase (hTERT) gene. Moreover, cells were also retrovirally of reactants). Despite that, compounds 1a-e were successfully syn- infected with vector producing an inducible myogenic differentia- thetized, and their structure confirmed by NMR and high- tion 1 (MYOD1) transcription factor fused to the estrogen receptor resolution mass spectrometry in 44–57% yields (see Supporting (ER) hormone-binding domain (MYOD1-ER) [43]. On the one hand, Information). TERT ensures a homogeneous proliferation over cycles of duplica- tion with no evidence of genetic affection related upon infection. On the other hand, MYOD1 allows, via estradiol addition to the cul- 3.3. Biochemistry ture media, the expression of myogenic factors leading to differen- tiation into myoblasts and finally forming mature myotubes. All One of the main challenges of dealing with a rare disease in the characteristic events of DM1 such as CUG nuclear foci, splicing drug development stages is the lack of any commercial kits or sup- factor sequestration and missplicing events were maintained in the pliers for assessing the potency of novel drug candidates. For this DM1 patient-derived cells in culture and absent in control cells reason, we chose to reproduce the already described Förster Reso- derived from healthy individuals [13]. Original fibroblasts from nance Energy Transfer (FRET) based AID 2675 assay [40]. This flu- DM1 patient had 290 repeats. Nonetheless, further expansion orescent test can measure the ability of candidates to free MBNL1 occurs in culture. Thus, and despite the cell population’s hetero- of an DM1-like oligo-RNA. The ability of MBNL1 (and the recombi- geneity, the most abundant length is in PCR amplification is larger nant form used) to recognize and bind to small GC rich structures is than 1000 CTG repeats. The biological activity of the compounds on the key factor of the AID 2675 molecular basis. Two fluorophores these DM1 features was tested on these cell models. Being skeletal bind to both protein and RNA (via antibody recognition and avi- muscle the most affected tissue in DM1 patients, the use of myo- din–biotin interaction respectively) making FRET possible when genic cell models offers the advantage to test the effect of the 57 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 Fig. 4. a) Graphical representation of AID2675 FRET test molecular basis and its principal components: 50 -biotinylated RNA oligo Biot-(CUG)12 and the recombinant GST- MBNL1-His6 form a stable complex. FRET donor anti-His6-Terbium cryptate and FRET acceptor streptavidin-conjugated XL665 bind to His tag and Biotin, respectively. While no active compound is present in the reaction media, 340 nm beam light hits Tb conjugate and this is close enough to transfer the energy to XL665 conjugate emitting at 665 nm. If a compound can free the protein no FRET occurs, and Tb fluorophore emits at 545 nm. b) Normalized FRET intensity measured as a relationship between the fluorescence intensity of the signals of 545 and 665 nm (average of three measures). Final concentrations of studied candidates were set at 0.1 mM and 20 nM of protein-RNA complex. Z is included for comparison. designed drugs on both DM1 hallmarks and differentiation capac- consistent with this prediction. Nevertheless, a reduction in the ity of these cells, in the perspective of a human therapy. To evalu- total number of nuclear foci was not expected as that means tran- ate 1a-1e and Z in cell culture some considerations were made. scription inhibition. The staining with the (CAG)6 RNA-FISH probe Some of these candidates underwent crystallization in the culture (red) showed no significant difference between untreated and trea- media at concentrations over 50 mM in previous studies. For this ted cells in the foci count per cell ratio for 1a-1e compounds. This reason, and to make all candidates comparable, their concentration is consistent with the prediction since the specific binding of these was set to 10 mM. These test concentrations are lower than the ligands to CUG motifs is not expected to affect the expression level ones used in previously reported studies [24]. Nevertheless, it is of the mutant transcripts. However, some compounds were able to compatible with a possible in vivo use of the candidates. reduce the amount of colocalized MBNL1 in those nuclei. That was In order to directly see whether MBNL1 was sequestered in CUG observed via IF using anti-MBNL1 mAb (green). Interestingly, 1a foci, RNA Fluorescent In Situ Hybridization (RNA-FISH) analysis induces an important depletion of MBNL1 sequestered in nuclear combined to immunofluorescent staining (IF) of MBNL1 protein foci of above 23%, which means that this compound can selectively with a specific antibody were performed as described previously attach to CUG and competitively displace MBNL1, eventually (Fig. 5, extended in Fig. S5) [45]. Ligands 1a-e were designed to reducing the number of nuclear foci containing the protein in a sig- specifically bind to CUG motifs and obtained FISH results are nificant manner (Fig. 5b). Fig. 5. a) FISH images using a (CAG)6 probe labelled with Texas Red at the 50 end in combination with immunofluorescence staining. To verify the co-localization of MBNL1 in ribonuclear inclusions, following the last post-hybridization wash, cells were stained sequentially with antibodies to MBNL1 and with goat anti-mouse antibody conjugated with Alexa Fluor 488. Nuclei were visualized with Hoechst 33,258 dye. Insets show zoomed-in images (scale bars 20 mm). b) Quantitative FISH analysis of total foci per cell and colocalized MBNL1 in foci. Results are presented as mean (±SD) of n = 3 experiments. Statistical significance was assessed from a two-tailed Student’s t-test (**p < 0.01, *p < 0.05). 58 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 Another important event in DM1 is missplicing deficiencies of a and triggering MBNL1 recovery in vitro (AID2675 test). It is worth large number of gene transcripts. To study possible recoveries in of attention that 1a also exhibited these features in DM1 patient- the splicing pattern we chose two gene transcripts known to be derived myoblasts, reducing the amount of observed MBNL1 in affected in DM1: the sarcoplasmic/endoplasmic reticulum calcium FISH-IF microscopy. These findings not only provide valuable data ATPase 1 (SERCA1) and the insulin receptor (INSR). The inclusion or about the length needed to reach two consecutive UU mismatches not of exon 22 of SERCA1 and exon 11 of INSR was monitored via for the pyrido[2,3-d]pyrimidines scaffold suggested in this study. semiquantitative RT-PCR to check whether missplicing events Moreover, results confirm the usefulness of molecular modelling were modulated by test compounds present in the media. The for the rational design of JW base recognizers with potential results showed that at the studied conditions (see Supporting anti-DM1 activity. information) the missplicing of these genes was not significantly These results allow a better understanding of the biological affected by the addition to the culture media of either ligands interaction of pyrido[2,3-d]pyrimidine structures and set a starting 1a-e or Z (Fig. S6). point in future drug discovery studies. To check if structures 1a-e affected the expression levels of DM1-related splicing factors or myogenic differentiation markers, Funding Western blot analysis was performed. The effect of compounds was studied on 4 protein levels: MBNL1, CELF1, myosin and myo- This work was supported by the Plan Estatal de I+D+i 2013– genin (Fig. S7). Ligands 1a-e were observed not to affect the total 2016, and the Instituto de Salud Carlos III - Subdirección General amount of MBNL1 or CELF1, which is in fact a positive result, as de Evaluación y Fomento de la Investigación (FIS13-0386, includ- these proteins are essential and any anti-DM1 drug candidate aims ing funds from FEDER). Roger Estrada-Tejedor thankfully acknowl- to recover the natural balance between these antagonists. On the edge the financial support of l’Obra Social ‘‘La Caixa”. other hand, myosin and myogenin are crucial muscle differentia- tion proteins [46,47]. A decrease in their total cell concentration CRediT authorship contribution statement may alter cell differentiation. Therefore, looking at the state of dif- ferentiation is an important indication of a possible toxic effect of a Raul Ondono: Conceptualization, Methodology, Formal analy- tested compound. Our findings with Western blot analysis show sis, Investigation, Writing - original draft, Visualization. Ángel that only in the case of ligand 1e a visible decrease of Myogenin Lirio: Methodology, Software, Investigation, Data curation. Carlos levels is observed while no effect at all was observed for the Elvira: Methodology, Investigation, Supervision. Elena Álvarez- remaining compounds. Marimon: Methodology, Investigation, Supervision. Claudia Provenzano: Methodology, Investigation, Supervision. Beatrice Cardinali: Methodology, Investigation, Supervision. Manuel 4. Discussion and conclusion Pérez-Alonso: Supervision, Funding acquisition. Alex Perálvarez- Marín: Supervision. José I. Borrell: Supervision. Germana Falcone: Base recognition has been one of the main trends in small mole- Methodology, Investigation, Supervision, Writing - review & edit- cule drug development in the DM1 field. This is due to the imple- ing. Roger Estrada-Tejedor: Conceptualization, Supervision, Fund- mentation of a comprehensive structure–activity relationship ing acquisition, Project administration, Writing - review & editing. [18,48]. The triangular shape of pyrido[2,3-d]pyrimidin-7(6H)- ones, combined with the spatial disposition of hydrogen bond Declaration of Competing Interest donors and acceptors in such core, resulted in a potential UU rec- ognizer interaction. The authors declare that they have no known competing finan- Computer-aided rational design studies have been described in cial interests or personal relationships that could have appeared the literature to study ligand-RNA interaction in recent years [48– to influence the work reported in this paper. 50]. It is worth noting that studies describing a rational approach, leaned on ligand-based drug design strategies, are commonly Acknowledgements applied on commercially available chemical libraries [51]. This study provides tools to design novel compounds with structure- Authors thank professor R. Artero from University of Valencia based methods, using the Janus-Wedge recognition as strategy to for his valuable contribution in revising the manuscript. stabilize CUG in its non-toxic conformation. Selectivity is one of the key points when designing small molecules to target the Appendix A. Supplementary data DM1 pathogenic RNA transcripts. Around 5% of DM1 patients do not have a pure CTG repeat, including interrupting CCG or CGG tri- Supplementary data to this article can be found online at plets, which may alter RNA’s secondary structure [52]. Thus, the https://doi.org/10.1016/j.csbj.2020.11.053. molecular modeling approach herein described could be adapted to embrace other kind of transcripts. References A set of 5 novel chemical entities based on two pyrido[2,3-d] pyrimidin-7(6H)-one scaffold separated by an aliphatic linker of [1] Hummerich H, Lehrach H. Trinucleotide repeat expansion and human disease. variable length was studied (1a-e). These candidates would exhibit Electrophoresis 1995;16(1):1698–704. https://doi.org/10.1002/ different interaction mechanisms with CUG repeats depending on elps.11501601282. [2] Cleary JD, Ranum LPW. Repeat-associated non-ATG (RAN) translation in their length, according to their predicted Gibbs free energy of bind- neurological disease. Hum Mol Genet 2013;22(R1):R45–51. https://doi.org/ ing. Simulations point to the 5-carbon length linker for an optimal 10.1093/hmg/ddt371. recognition of consecutive UU mismatches the 10-carbon length [3] Thomas JD, Oliveira R, Sznajder ŁJ, Swanson MS. Myotonic dystrophy and developmental regulation of RNA processing. Compr Physiol 2018;8:509–53. linker and for alternate mismatches the 10-carbon linker. More- https://doi.org/10.1002/cphy.c170002. over, the analysis of base pair opening showed a reduction in data [4] Chong-Nguyen C, Wahbi K, Algalarrondo V, Bécane HM, Radvanyi-Hoffman H, variance for 1a stating that the candidate can stabilize the CUG Arnaud P, Furling D, Lazarus A, Bassez G, Béhin A, Fayssoil A, Laforêt P, environment and maintain the JW interaction during the whole Stojkovic T, Eymard B, Duboc D. Association Between Mutation Size and Cardiac Involvement in Myotonic Dystrophy Type 1: An Analysis of the DM1- simulation. Interestingly, this observation was validated in Heart Registry. Circ Cardiovasc Genet 2017;10(3). https://doi.org/10.1161/ biological studies that proved the ability of 1a for CUG binding CIRCGENETICS.116.001526. 59 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 [5] André LM, Van Cruchten RTP, Willemse M, Wansink DG. (CTG)n repeat- [26] Camarasa M, Puig de la Bellacasa R, González ÀL, Ondoño R, Estrada R, Franco mediated dysregulation of MBNL1 and MBNL2 expression during myogenesis S, Badia R, Esté J, Martínez MÁ, Teixidó J, Clotet B, Borrell JI. Design, synthesis in DM1 occurs already at the myoblast stage. PLoS One 2019;14:1–17. and biological evaluation of pyrido[2,3-d]pyrimidin-7-(8H)-ones as HCV https://doi.org/10.1371/journal.pone.0217317. inhibitors. Eur J Med Chem 2016;115:463–83. https://doi.org/10.1016/j. [6] Morales F, Couto JM, Higham CF, Hogg G, Cuenca P, Braida C, et al. Somatic ejmech.2016.03.055. instability of the expanded CTG triplet repeat in myotonic dystrophy type 1 is a [27] Recasens-Zorzo C, Cardesa-Salzmann T, Petazzi P, Ros-Blanco L, Esteve-Arenys heritable quantitative trait and modifier of disease severity. Hum Mol Genet A, Clot G, Guerrero-Hernández M, Rodríguez V, Soldini D, Valera A, Moros A, 2012;21:3558–67. https://doi.org/10.1093/hmg/dds185. Climent F, González-Barca E, Mercadal S, Arenillas L, Calvo X, Mate JL, [7] Jain A, Vale RD. RNA phase transitions in repeat expansion disorders. Nature Gutiérrez-García G, Casanova I, Mangues R, Sanjuan-Pla A, Bueno C, Menéndez 2017;546(7657):243–7. https://doi.org/10.1038/nature22386. P, Martínez A, Colomer D, Tejedor RE, Teixidó J, Campo E, López-Guillermo A, [8] Mankodi A, Urbinati CR, Yuan Q-P, Moxley RT, Sansone V, Krym M, et al. Borrell JI, Colomo L, Pérez-Galán P, Roué G. Pharmacological modulation of Muscleblind localizes to nuclear foci of aberrant RNA in myotonic dystrophy CXCR4 cooperates with BET bromodomain inhibition in diffuse large B-cell types 1 and 2. Hum Mol Genet 2001;10:2165–70. https://doi.org/10.1093/ lymphoma. Haematologica 2019;104(4):778–88. https://doi.org/10.3324/ hmg/10.19.2165. haematol.2017.180505. [9] Philips AV, Timchenko LT, Cooper TA. Disruption of Splicing Regulated by a [28] González ÀL, Teixidó J, Borrell JI, Estrada-Tejedor R. On the applicability of CUG-Binding Protein in Myotonic Dystrophy. Science (80-) 1998;280:737–41. elastic network models for the study of RNA CUG trinucleotide repeat https://doi.org/10.1126/science.280.5364.737. overexpansion. PLoS One 2016;11:1–20. https://doi.org/10.1371/journal. [10] Kanadia RN, Shin J, Yuan Y, Beattie SG, Wheeler TM, Thornton CA, Swanson MS. pone.0152049. Reversal of RNA missplicing and myotonia after muscleblind overexpression in [29] Ryckaert J-P, Ciccotti G, Berendsen HJC. Numerical integration of the cartesian a mouse poly(CUG) model for myotonic dystrophy. Proc Natl Acad Sci equations of motion of a system with constraints: molecular dynamics of n- 2006;103(31):11748–53. https://doi.org/10.1073/pnas.0604970103. alkanes. J Comput Phys 1977;23(3):327–41. https://doi.org/10.1016/0021- [11] Overby SJ, Cerro-Herreros E, Llamusi B, Artero R. RNA-mediated therapies in 9991(77)90098-5. myotonic dystrophy. Drug Discovery Today 2018;23(12):2013–22. https://doi. [30] Roe DR, Cheatham III TE. PTRAJ and CPPTRAJ: Software for Processing and org/10.1016/j.drudis.2018.08.004. Analysis of Molecular Dynamics Trajectory Data. J Chem Theory Comput [12] Konieczny P, Selma-Soriano E, Rapisarda AS, Fernandez-Costa JM, Perez- 2013;9(7):3084–95. https://doi.org/10.1021/ct400341p. Alonso M, Artero R. Myotonic dystrophy: candidate small molecule [31] Pettersen EF, Goddard TD, Huang CC, Couch GS, Greenblatt DM, Meng EC, therapeutics. Drug Discovery Today 2017;22(11):1740–8. https://doi.org/ Ferrin TE. UCSF Chimera?A visualization system for exploratory research and 10.1016/j.drudis.2017.07.011. analysis. J Comput Chem 2004;25(13):1605–12. https://doi.org/10.1002/ [13] Provenzano C, Cappella M, Valaperta R, Cardani R, Meola G, Martelli F, jcc.20084. Cardinali B, Falcone G. CRISPR/Cas9-Mediated Deletion of CTG Expansions [32] Wang J, Wolf RM, Caldwell JW, Kollman PA, Case DA. Development and testing Recovers Normal Phenotype in Myogenic Cells Derived from Myotonic of a general amber force field. J Comput Chem 2004;25(9):1157–74. https:// Dystrophy 1 Patients. Mol Ther Nucleic Acids 2017;9:337–48. https://doi. doi.org/10.1002/jcc.20035. org/10.1016/j.omtn.2017.10.006. [33] Aytenfisu AH, Spasic A, Grossfield A, Stern HA, Mathews DH. Revised RNA [14] Lo Scrudato M, Poulard K, Sourd C, Tomé S, Klein AF, Corre G, Huguet A, Furling Dihedral Parameters for the Amber Force Field Improve RNA Molecular D, Gourdon G, Buj-Bello A. Genome Editing of Expanded CTG Repeats within Dynamics. J Chem Theory Comput 2017;13(2):900–15. https://doi.org/ the Human DMPK Gene Reduces Nuclear RNA Foci in the Muscle of DM1 Mice. 10.1021/acs.jctc.6b00870.s002. Mol Ther 2019;27(8):1372–88. https://doi.org/10.1016/j.ymthe.2019.05.021. [34] Izadi S, Anandakrishnan R, Onufriev AV. Building Water Models: A Different [15] Jauvin D, Chrétien J, Pandey SK, Martineau L, Revillod L, Bassez G, Lachon A, Approach. J Phys Chem Lett 2014;5(21):3863–71. https://doi.org/10.1021/ MacLeod AR, Gourdon G, Wheeler TM, Thornton CA, Bennett CF, Puymirat J. jz501780a. Targeting DMPK with Antisense Oligonucleotide Improves Muscle Strength in [35] Salomon-Ferrer R, Götz AW, Poole D, Le Grand S, Walker RC. Routine Myotonic Dystrophy Type 1 Mice. Mol Ther Nucleic Acids 2017;7:465–74. Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 2. https://doi.org/10.1016/j.omtn.2017.05.007. Explicit Solvent Particle Mesh Ewald. J Chem Theory Comput 2013;9 [16] Garcia-Lopez A, Llamusi B, Orzaez M, Perez-Paya E, Artero RD. In vivo (9):3878–88. https://doi.org/10.1021/ct400314y. discovery of a peptide that prevents CUG-RNA hairpin formation and reverses [36] Le Grand S, Götz AW, Walker RC. SPFP: Speed without compromise—A mixed RNA toxicity in myotonic dystrophy models. Proc Natl Acad Sci 2011;108 precision model for GPU accelerated molecular dynamics simulations. Comput (29):11866–71. https://doi.org/10.1073/pnas.1018213108. Phys Commun 2013;184(2):374–80. https://doi.org/10.1016/j. [17] Rzuczek SG, Colgan LA, Nakai Y, Cameron MD, Furling D, Yasuda R, Disney MD. cpc.2012.09.022. Precise small-molecule recognition of a toxic CUG RNA repeat expansion. Nat [37] Götz AW, Williamson MJ, Xu D, Poole D, Le Grand S, Walker RC. Routine Chem Biol 2017;13(2):188–93. https://doi.org/10.1038/nchembio.2251. Microsecond Molecular Dynamics Simulations with AMBER on GPUs. 1. [18] Lee JY, Bai Y, Chembazhi UV, Peng S, Yum K, Luu LM, Hagler LD, Serrano JF, Generalized Born. J Chem Theory Comput 2012;8(5):1542–55. https://doi. Chan HYE, Kalsotra A, Zimmerman SC. Intrinsically cell-penetrating org/10.1021/ct200909j. multivalent and multitargeting ligands for myotonic dystrophy type 1. Proc [38] Lu X-J, Olson WK. 3DNA: a versatile, integrated software system for the Natl Acad Sci USA 2019;116(18):8709–14. https://doi.org/10.1073/ analysis, rebuilding and visualization of three-dimensional nucleic-acid pnas.1820827116. structures. Nat Protoc 2008;3(7):1213–27. https://doi.org/10.1038/ [19] Coonrod LA, Nakamori M, Wang W, Carrell S, Hilton CL, Bodner MJ, Siboni RB, nprot.2008.104. Docter AG, Haley MM, Thornton CA, Berglund JA. Reducing Levels of Toxic RNA [39] Gualandi A, Mazzarella D, Ortega-Martínez A, Mengozzi L, Calcinelli F, with Small Molecules. ACS Chem Biol 2013;8(11):2528–37. https://doi.org/ Matteucci E, Monti F, Armaroli N, Sambri L, Cozzi PG. Photocatalytic Radical 10.1021/cb400431f. Alkylation of Electrophilic Olefins by Benzylic and Alkylic Zinc-Sulfinates. ACS [20] Palomo V, Perez DI, Roca C, Anderson C, Rodríguez-Muela N, Perez C, Morales- Catal 2017;7(8):5357–62. https://doi.org/10.1021/acscatal.7b01669.s001. Garcia JA, Reyes JA, Campillo NE, Perez-Castillo AM, Rubin LL, Timchenko L, Gil [40] Chen CZ, Sobczak K, Hoskins J, Southall N, Marugan JJ, Zheng W, Thornton CA, C, Martinez A. Subtly Modulating Glycogen Synthase Kinase 3 b: Allosteric Austin CP. Two high-throughput screening assays for aberrant RNA–protein Inhibitor Development and Their Potential for the Treatment of Chronic interactions in myotonic dystrophy type 1. Anal Bioanal Chem 2012;402 Diseases. J Med Chem 2017;60(12):4983–5001. https://doi.org/10.1021/acs. (5):1889–98. https://doi.org/10.1007/s00216-011-5604-0. jmedchem.7b00395.s002. [41] Artimo P, Jonnalagedda M, Arnold K, Baratin D, Csardi G, de Castro E, Duvaud S, [21] Jahromi AH, Nguyen L, Fu Y, Miller KA, Baranger AM, Zimmerman SC. A Novel Flegel V, Fortier A, Gasteiger E, Grosdidier A, Hernandez C, Ioannidis V, CUG exp MBNL1 Inhibitor with Therapeutic Potential for Myotonic Dystrophy Kuznetsov D, Liechti R, Moretti S, Mostaguir K, Redaschi N, Rossier G, Xenarios Type 1. ACS Chem Biol 2013;8(5):1037–43. https://doi.org/10.1021/ I, Stockinger H. ExPASy: SIB bioinformatics resource portal. Nucleic Acids Res cb400046u. 2012;40(W1):W597–603. https://doi.org/10.1093/nar/gks400. [22] deLorimier E, Hinman MN, Copperman J, Datta K, Guenza M, Berglund JA. [42] Counter CM, Meyerson M, Eaton EN, Ellisen LW, Caddle SD, Haber DA, Pseudouridine Modification Inhibits Muscleblind-like 1 (MBNL1) Binding to Weinberg RA. Telomerase activity is restored in human cells by ectopic CCUG Repeats and Minimally Structured RNA through Reduced RNA expression of hTERT (hEST2), the catalytic subunit of telomerase. Oncogene Flexibility. J Biol Chem 2017;292(10):4350–7. https://doi.org/10.1074/jbc. 1998;16(9):1217–22. https://doi.org/10.1038/sj.onc.1201882. M116.770768. [43] Hollenberg SM, Cheng PEIF, Weintraub H. Trans-Activation and Muscle [23] Branda N, Kurz G, Lehn J-M. JANUS WEDGES: a new approach towards Determination. Proc Natl Acad Sci U S A 1993;90:8028–32. nucleobase-pair recognition. Chem Commun 1996(21):2443. https://doi.org/ [44] Mont N, Teixidó J, Kappe CO, Borrell JI. A one-pot microwave-assisted 10.1039/cc9960002443. synthesis of pyrido[2,3-d]pyrimidines. Mol Divers 2003;7(2-4):153–9. [24] Wong C-H, Nguyen L, Peh J, Luu LM, Sanchez JS, Richardson SL, Tuccinardi T, https://doi.org/10.1023/B:MODI.0000006808.10647.f8. Tsoi Ho, Chan WY, Chan HYE, Baranger AM, Hergenrother PJ, Zimmerman SC. [45] Cardani R, Mancinelli E, Sansone V, Rotondo G, Meola G. Biomolecular Targeting Toxic RNAs that Cause Myotonic Dystrophy Type 1 (DM1) with a identification of (CCTG)n mutation in myotonic dystrophy type 2 (DM2) by Bisamidinium Inhibitor. J Am Chem Soc 2014;136(17):6355–61. https://doi. FISH on muscle biopsy. Eur J Histochem 2004;48:437–42. https://doi.org/ org/10.1021/ja5012146. 10.4081/918. [25] Yuan Y, Compton SA, Sobczak K, Stenberg MG, Thornton CA, Griffith JD, et al. [46] Sharma A, Agarwal M, Kumar A, Kumar P, Saini M, Kardon G, et al. Myosin Muscleblind-like 1 interacts with RNA hairpins in splicing target and Heavy Chain-embryonic is a crucial regulator of skeletal muscle pathogenic RNAs. Nucleic Acids Res 2007;35:5474–86. https://doi.org/ development and differentiation. BioRxiv 2018:261685. https://doi.org/ 10.1093/nar/gkm601. 10.1101/261685. 60 R. Ondono, Á. Lirio, C. Elvira et al. Computational and Structural Biotechnology Journal 19 (2021) 51–61 [47] Flynn JM, Meadows E, Fiorotto M, Klein WH. Myogenin regulates exercise RNA Complex in Myotonic Dystrophy Type 1 (DM1) Leads to the Unexpected capacity and skeletal muscle metabolism in the adult mouse. PLoS One 2010;5. Discovery of a DNA-Selective Binder. ChemBioChem 2012;13:2505–9. https:// https://doi.org/10.1371/journal.pone.0013535. doi.org/10.1002/cbic.201200602. [48] Li J, Nakamori M, Matsumoto J, Murata A, Dohno C, Kiliszek A, et al. A Dimeric [51] Gonzalez ÀL, Konieczny P, Llamusi B, Delgado-Pinar E, Borrell JI, Teixidó J, et al. 2,9-Diamino-1,10-phenanthroline Derivative Improves Alternative Splicing in In silico discovery of substituted pyrido[2,3-d] pyrimidines and pentamidine- Myotonic Dystrophy Type 1 Cell and Mouse Models. Chem - A Eur J like compounds with biological activity in myotonic dystrophy models. PLoS 2018;24:18115–22. https://doi.org/10.1002/chem.201804368. One 2017;12:1–23. https://doi.org/10.1371/journal.pone.0178931. [49] Velagapudi SP, Gallo SM, Disney MD. Sequence-based design of bioactive small [52] Braida C, Stefanatos RK, Adam B, Mahajan N, Smeets HJ, Niel F, et al. Variant molecules that target precursor microRNAs. Nat Chem Biol 2014;10(4):291–7. CCG and GGC repeats within the CTG expansion dramatically modify https://doi.org/10.1038/nchembio.1452. mutational dynamics and likely contribute toward unusual symptoms in [50] Wong CH, Richardson SL, Ho YJ, Lucas AMH, Tuccinardi T, Baranger AM, et al. some myotonic dystrophy type 1 patients l. Hum Mol Genet. 2010;19:1399– Investigating the Binding Mode of an Inhibitor of the MBNL1{dot operator} 412. doi: 10.1093/hmg/ddq015 61