Systematic evaluation of 2′-Fluoro modified chimeric antisense oligonucleotide-mediated exon skipping in vitro

Antisense oligonucleotide (AO)-mediated splice modulation has been established as a therapeutic approach for tackling genetic diseases. Recently, Exondys51, a drug that aims to correct splicing defects in the dystrophin gene was approved by the US Food and Drug Administration (FDA) for the treatment of Duchenne muscular dystrophy (DMD). However, Exondys51 has relied on phosphorodiamidate morpholino oligomer (PMO) chemistry which poses challenges in the cost of production and compatibility with conventional oligonucleotide synthesis procedures. One approach to overcome this problem is to construct the AO with alternative nucleic acid chemistries using solid-phase oligonucleotide synthesis via standard phosphoramidite chemistry. 2′-Fluoro (2′-F) is a potent RNA analogue that possesses high RNA binding affinity and resistance to nuclease degradation with good safety profile, and an approved drug Macugen containing 2′-F-modified pyrimidines was approved for the treatment of age-related macular degeneration (AMD). In the present study, we investigated the scope of 2′-F nucleotides to construct mixmer and gapmer exon skipping AOs with either 2′-O-methyl (2′-OMe) or locked nucleic acid (LNA) nucleotides on a phosphorothioate (PS) backbone, and evaluated their efficacy in inducing exon-skipping in mdx mouse myotubes in vitro. Our results showed that all AOs containing 2′-F nucleotides induced efficient exon-23 skipping, with LNA/2′-F chimeras achieving better efficiency than the AOs without LNA modification. In addition, LNA/2′-F chimeric AOs demonstrated higher exonuclease stability and lower cytotoxicity than the 2′-OMe/2′-F chimeras. Overall, our findings certainly expand the scope of constructing 2′-F modified AOs in splice modulation by incorporating 2′-OMe and LNA modifications.

DMD is a severe and fatal muscle wasting genetic disorder mainly affecting newborn boys [26][27][28] . DMD is caused by one or more mutations in the dystrophin gene that ablate the expression of functional dystrophin proteins required for protecting muscle fibers from eccentric contraction and movement 29,30 . Recently, AO-mediated exon skipping has been established as one of the most promising therapeutic strategy for treating DMD 28,[31][32][33][34][35][36] . Skipping the mutation containing exons can restore the dystrophin reading frame and rescue the production of the truncated but partially functional dystrophin protein. PMO and 2′-OMe-PS-modified AOs have been investigated in phase-3 clinical trials for DMD 27,28,[32][33][34][35][36][37][38] . In 2016, the PMO-based AO drug (Exondys51) has been granted accelerated approval by the US FDA 3 . In contrast, the 2′-OMe-PS-based candidate (drisapersen) was rejected mainly due to safety issues and lack of efficacy 39 . Although PMO-modified oligonucleotides showed excellent safety profile, it is not compatible with standard oligonucleotide synthesis chemistries in order to synthesise as mixmers with other well-known nucleotide analogues and large-scale production of PMOs is challenging due to distinctive synthesis procedure. Therefore, it is necessary to evaluate alternative nucleic acid chemistries that can be used for AO drug development.
Towards this, Kawasaki et al. introduced 2′-F as an attractive ribonucleotide analogue for constructing AO. In fact, 2′-F-PS ( Fig. 1) AOs showed higher target binding affinity, and nuclease stability 8 . Previous studies have also revealed its enhanced capability of inducing exon skipping in vitro compared to 2′-OMe-PS ( Fig. 1) AOs [8][9][10][11] , which may be due to the recruitment of interleukin enhancer binding factors 2 and 3 (ILF2/3) by 2′-F AO/pre-mRNA duplex which may result in improved steric block efficiency [9][10][11] . However, 2′-F-modified AOs did not reach clinical evaluation and the scope of 2′-F-modified AOs needs to be improved by novel design approaches. LNA ( Fig. 1) is another prominent RNA analogue that has been successfully investigated in recent studies to induce exon skipping in the dystrophin gene transcript 16,40 . In this study, for the first time, we herein report the design, synthesis and evaluation of 2′-F-modified exon skipping AOs to induce exon-23 skipping in DMD mouse myotubes in vitro.

Results
In the present study, we used a previously reported fully modified 2′-OMe-PS 18-mer AO sequence, DmdE23D (+1-17) which was designed to induce exon-23 skipping in mouse Dmd transcript 16 , as a positive control (Table 1). Based on this AO, we systematically designed and synthesised a fully modified 2′-F AO on a PS backbone, three 2′-OMe/2′-F-PS chimeric AOs, and three LNA/2′-F-PS chimeric AOs which include a gapmer and two mixmer designs (Table 1). Two-step systematic evaluation was performed in vitro in mouse myotubes differentiated from H-2K b -tsA58 (H2K) mdx myoblasts. Initial evaluation was conducted for all AOs at 12.5 nM, 25 nM, and 50 nM concentrations while secondary evaluation was performed at lower concentrations (2.5 nM, 5 nM, and 12.5 nM) for chimeric AOs. In general, H2K mdx myoblasts were plated on 24-well plates and incubated for 24 h for differentiation. The differentiated myotubes were then transfected with different concentrations of the above-mentioned AOs by Lipofectin transfection reagent using a ratio of 2:1 (Lipofectin: AO). Twenty-four hours after transfection, cells were collected followed by total cellular RNA extraction, and reverse transcription polymerase chain reaction (RT-PCR) to amplify the dystrophin transcripts across exons 20-26 as reported previously 41 . Next, 2% agarose gel electrophoresis and densitometry (using Image J software) were performed to quantify the PCR products. The actual percentages of full length (901 bp), exon-23 skipping (688 bp), and exon-22/23 dual skipping (542 bp) products are presented based on the total amount of the dystrophin transcripts. Systematic exon skipping evaluation was performed in duplicates.  (Figs 2 and 3). In line with previous report 16 , the 2′-OMe-PS control AO induced efficient exon-23 skipping by yielding the skipped product of 688 bp at all concentrations (43% at 12.5 nM, 47% at 25 nM, and 51% at 50 nM; Fig. 3A). Interestingly, the fully modified 2′-F-PS AO showed higher exon-23 skipping at 12.5 nM (50%) compared to the control AO (43%), but the efficiency reduced to 44% at 25 nM and remained 51% at 50 nM, respectively. But, 2′-F-PS AOs showed the undesired exon-22/23 dual skipping product of 542 bp in higher yield at 25 nM (35%) compared to 2′-OMe-PS AO (27%) (Fig. 3A).

evaluation of in vitro nuclease stability of the 2′-F modified AOs.
To gain more insight into the AO stability, we then performed the nuclease degradation assay of all the 2′-F modified AOs in comparison to the fully 2′-OMe-PS control. In short, AOs were incubated with Phosphodiesterase I from Crotalus adamanteus venom which possesses very high exonuclease activity, at 37 °C for various incubation periods including 0, 1, 2, 4, and 6 h. Samples were collected at the desired timepoints and quenched with formamide loading buffer, followed by denaturation at 95 °C for 5 min. Next, 20% denaturing polyacrymide gel analysis was performed and the results were analysed by gel imaging. All 2′-F modified AOs demonstrated high stability under the applied conditions compared to the fully 2′-OMe-PS control (Fig. 7). Not surprisingly, all the LNA/2′-F-PS chimeras showed higher nuclease resistance than the other AOs without LNA modification (Fig. 7). In addition, the gapmer chimeras showed higher stability than the mixmer chimeras.

Discussion
Therapeutic potential of AOs was first demonstrated by Zamecnik et al. in 1978 42 . Stemming from this initial work, AOs have been extensively explored as a potential gene-targeting approach for the treatment of various genetic diseases. In line with this, splice-switching AOs have been developed, firstly by Dominski et al. in 1993, and later became promising therapies towards tackling genetic diseases caused by mutations such as DMD and SMA 43 . Along this line, chemically-modified nucleic acid analogues play a crucial role in the successful clinical www.nature.com/scientificreports www.nature.com/scientificreports/ translation of any AO-based drug, given the PMO-based Exondys51 was granted conditional approval in 2016, while the 2′-OMe-PS-based drisapersen was rejected in the same year due to lack of efficacy and life-threatening side effects. Although a PMO-based AO is relatively non-toxic, PMO chemistry has its disadvantages due to limitation of large-scale synthesis and lack of compatibility with other chemistries. Towards exploring chemically modified AOs targeting DMD, Aartsma-Rus and colleagues evaluated AOs constructed by fully modified 2′-F-PS and LNA-PS 10,11,44 , while our group has investigated LNA, HNA, CeNA, ANA, and MNA monomers 16,22,23 .
The first attempt to study 2′-F modified AO was reported in 1993 when Kawasaki et al. found that 2′-F modification enhanced the AO's target binding affinity to their complementary RNA compared to 2′-OMe modified AO, and showed excellent nuclease stability 8 . Two decades later, Rigo and colleagues discovered a unique property of the 2′-F modified AO/target pre-mRNA duplex that is able to recruit the ILF2/3 proteins, resulting in exon-7 skipping of SMN2 mRNA in a SMA model system 9 . Based on this finding, Aartsma-Rus and coworkers compared the exon skipping capability of the fully 2′-F-PS and fully 2′-OMe-PS AOs in targeting DMD, and showed that 2′-F-PS AO induced higher human exon-53 and mouse exon-23 skipping in vitro 10,11 , which was not surprising as 2′-F modification has many advantages as proven by Kawasaki et al. 8 and Rigo et al. 9 , however, 2′-F-PS AO was less efficient than 2′-OMe-PS in vivo and indicated toxicity in mice 11 . Thus, their results did not support clinical use of 2′-F-PS AOs 11 .
In an attempt to improve the therapeutic potential of 2′-F modified AO, we incorporated 2′-OMe-PS and LNA-PS nucleotides into an 18 mer 2′-F-PS AO sequence that contained either four 2′-OMe or LNA nucleotides designed to target exon-23 of mdx mouse myotubes ( Table 1). The efficacies of the AOs were first evaluated at higher (12.5 nM, 25 nM, 50 nM), and then lower concentrations (2.5 nM, 5 nM, 12.5 nM); in addition to performing cytotoxicity and nuclease stability analysis.
In addition, cell viability assay performed to assess the cytotoxicity of the AOs showed that all 2′-F modified chimeric AOs achieved comparable cytotoxicity profiles in comparison to their fully 2′-OMe-PS control and the fully modified 2′-F-PS AO. Notably, the LNA/2′-F-PS chimeras showed better safety profile than the 2′-OMe/2′-F-PS chimeras, and the 2′-OMe/2′-F-PS mixmer 2 appeared to be the most toxic AO compared to others. This suggests that nucleotide positioning can be important in optimizing AO toxicity, although further evaluation in in vivo models are required. Nuclease stability assay demonstrated that fully 2′-F-PS AO and 2′-OMe/2′-F-PS chimeras possess similar stability as their fully 2′-OMe-PS control in vitro. Not surprisingly, LNA/2′-F-PS chimeras were more stable than the other AOs without LNA modification.
In conclusion, 2′-F-PS modified AOs induce higher Dmd exon-23 skipping efficiency than fully 2′-OMe-PS AO. Introduction of LNA nucleotides into 2′-F-PS sequence further improved exon-23 skipping efficiency, while not compromising cell viability and nuclease stability, in comparison to the 2′-OMe/2′-F-PS chimeras. In addition, mixmer designs of 2′-OMe/2′-F chimeras achieved higher efficiency of exon-23 skipping than their gapmer counterpart, while gapmer design of LNA/2′-F chimeras achieved higher efficiency of exon-23 skipping than their mixmer counterparts. Collectively, our findings expand the scope of utilizing 2′-F modified AOs in splice modulation application by constructing 2′-OMe and LNA-modified 2′-F-PS chimeras. Specifically, we suggest that development of LNA modified 2′-F-PS mixmer or gapmer chimeric AOs, may present a promising therapeutic strategy for DMD.

RNA isolation and reverse transcription polymerase chain reaction (RT-PCR).
Total RNA was extracted from transfected mouse myotubes using Direct-zol ™ RNA MiniPrep Plus with TRI Reagent ® (Zymo Research, supplied through Integrated Sciences; cat#: R2052) as per the manufacturer's instructions. The dystrophin transcripts were then analysed by RT-PCR using SuperScript ™ III Reverse Transcriptase III (ThermoFisher Scientific; cat#: 12574026) across exons-20 to 26 as described previously 38 . PCR products were separated on a 2% agarose gel in Tris-acetate-EDTA buffer and the images were captured on a Fusion Fx gel documentation system (Vilber Lourmat). Densitometry was performed by Image J software 47 . To quantify the actual exon skipping efficacy induced by AOs, the amount of full length (901 bp), exon-23 skipping (688 bp), and exon-22/23 dual skipping (542 bp) products are expressed as percentages of total dystrophin transcript products.
In vitro nuclease stability analysis of AOs. Stability of all the AOs (Table 1) against 3′ end to 5′ end exonuclease degradation was evaluated using 0.08 units/reaction of Phosphodiesterase I from Crotalus adamanteus venom (Sigma; cat#: P3134-100MG) in a buffer of 10 mM Tris-HCL, 100 mM NaCl, and 15 mM MgCl 2 in a final volume of 45 μL. Briefly, samples were incubated at 37 °C and 7.5 μL of reaction aliquots were removed at different time points (0, 1, 2, 4, and 6 h) and an equal volume of 80% formamide containing bromophenol blue and xylene cyanol gel tracking dyes was then added, followed by heating for 5 min at 95 °C. Next, the products were separated on a 20% denaturing polyacrylamide gel. Quantitation was performed on a Fusion Fx gel documentation system (Vilber Lourmat).