Intramuscular Evaluation of Chimeric Locked Nucleic Acid/2′OMethyl-Modified Antisense Oligonucleotides for Targeted Exon 23 Skipping in Mdx Mice

Duchenne muscular dystrophy (DMD) is a fatal disorder characterised by progressive muscle wasting. It is caused by mutations in the dystrophin gene, which disrupt the open reading frame leading to the loss of functional dystrophin protein in muscle fibres. Antisense oligonucleotide (AON)-mediated skipping of the mutated exon, which allows production of a truncated but partially functional dystrophin protein, has been at the forefront of DMD therapeutic research for over two decades. Nonetheless, novel nucleic acid modifications and AON designs are continuously being developed to improve the clinical benefit profile of current drugs in the DMD pipeline. We herein designed a series of 15mer and 20mer AONs, consisting of 2′O-Methyl (2′OMe)- and locked nucleic acid (LNA)-modified nucleotides in different percentage compositions, and assessed their efficiency in inducing exon 23 skipping and dystrophin restoration in locally injected muscles of mdx mice. We demonstrate that LNA/2′OMe AONs with a 30% LNA composition were significantly more potent in inducing exon skipping and dystrophin restoration in treated mdx muscles, compared to a previously tested 2′OMe AON and LNA/2′OMe chimeras with lower or higher LNA compositions. These results underscore the therapeutic potential of LNA/2′OMe AONs, paving the way for further experimentation to evaluate their benefit-toxicity profile following systemic delivery.


Introduction
Duchenne Muscular Dystrophy (DMD) is an X-linked recessive disorder characterised by progressive muscle wasting, loss of ambulation in early adolescence and premature mortality due to cardiorespiratory complications. The disorder is mainly caused by nonsense or frame-shift mutations in the dystrophin gene, which disrupt the open reading frame, of a tcDNA/PS AON in mdx mice resulted in high restoration of dystrophin expression in skeletal muscle, heart and diaphragm tissues, and to a lower extent in the brain, leading to both functional and neurobehavioral improvement. Notably, tcDNA/PS was more efficient than the PMO and 2 OMePS AONs and demonstrated an encouraging safety profile with minimal histological changes and slight differences in serum and urinary levels of renal toxicity biomarkers [37,38].
LNA nucleoside analogues consist of a methylene bridge joining the 2 -oxygen to the 4 -carbon, resulting in a locked C3 -endo conformation that closely resembles A-form RNA. This conformational feature essentially pre-organizes LNA-modified AONs for RNA binding [39]. LNA nucleotides impart impressive thermodynamic stability, the highest among all other chemistries, increasing the Tm of a duplex by approximately 5.6 • C per insert [40]. Though much more potent, LNA-modified AONs are more tolerant to mismatches increasing the risk of unintended, off-target interactions. In a comparative study of chemically modified AONs targeting human exon 46, a 14mer all-LNA AON induced the highest levels of exon skipping in both unaffected control-(85%) and patient-derived (98%) myotubes. By contrast, a 20mer 2 OMe/PS AON was remarkably more efficient in the DMD versus control myotubes (75 versus 20%, respectively), a finding that questioned the sequence specificity of LNA versus 2 OMe AONs. Indeed, while demonstrating higher skipping efficiencies, LNA AONs containing up to three mismatches showed much less sequence specificity compared to mismatched 2 OMe AONs [41]. Consequently, recent studies have focused on addressing the optimal design of splice-switching LNA AONs, by altering the number of incorporated LNAs as well as the length of the oligomer sequence [42,43]. A 16mer LNA/DNA AON with an LNA incorporation of 60% efficiently induced skipping of exon 51 and restored dystrophin expression in the plasma membrane of patient-derived myotubes, thus highlighting the value of short LNA AONs for splice-switching therapy [43].
Attempting to limit sequence-dependent and sequence-independent off-target interactions, associated with the LNA and PS modifications, respectively, Le and colleagues designed short mixmer AONs incorporating a combination of 2 OMe and LNA nucleotides on a full PS backbone. The introduction of four LNAs in a 14mer 2 OMe/PS AON, showing no exon skipping activity, significantly improved the yield of exon 23-skipped RNA in cultured H2K mdx mouse myotubes, underscoring the value of short LNA/2 OMe AONs for DMD therapy [44]. Recently, chimeric LNA/2 OMe AONs (14-18mer) have also been evaluated for Myotonic Dystrophy type 1 (DM1) therapy, demonstrating strong steric-blocking activities in cultured human DM1 myoblasts and in locally injected muscles of DM1 mice [45]. To further evaluate the optimal length and percentage chemistry composition of LNA/2 OMe chimeras for DMD therapy, we herein designed a series of 20mer LNA/2 OMe AONs by introducing an increasing number of LNA nucleotides to a previously published 20mer 2 OMePS sequence targeting exon 23 of the mouse dystrophin pre-mRNA [46,47]. Truncated (15mer) sequence analogues with similar LNA compositions were also created and compared to a corresponding 2 OMe AON. The AONs were delivered intramuscularly into the tibialis anterior (TA) muscle of mdx mice and evaluated for targeted exon 23 skipping and rescued dystrophin protein expression. This study represents the first in vivo evaluation of the LNA/2 OMe AON design for DMD therapy.

Evaluation of Exon Skipping Efficiency of Chimeric LNA/2 OMe AONs after Intramuscular Delivery in Mdx Mice
In order to address the optimal design of chimeric LNA/2 OMe AONs for exon skipping activity in vivo, we created a series of 15mer (sLNA-3, sLNA-5 and sLNA-6) and 20mer (LNA-4, LNA-6 and LNA-8) LNA/2 OMe oligomers exhibiting an LNA incorporation of 20, 30 and 40% respectively, for each group, and compared their efficiencies to a 15mer and 20mer 2 OMe-modified AON. Both 2 OMe and chimeric LNA/2 OMe AONs were synthesized on a fully modified PS backbone. An equimolar dose of each AON was injected into the TA muscle of 8-week-old mdx mice and evaluated for exon 23 skipping by RT-PCR, two weeks after treatment.
Kruskal-Wallis analysis rendered an overall p-value of 0.00028, demonstrating statistically significant difference(s) among the efficiencies of the various AONs. Overall, 20mer 2 OMe and 20mer LNA/2 OMe AONs (LNA-4, LNA-6 and LNA-8) induced higher levels of exon 23 skipping compared to their respective 15mer analogues (s2 OMe, sLNA-3, sLNA-5 and sLNA-6; Figure 1A). AONs LNA-6 (20mer) and sLNA-5 (15mer), exhibiting an LNA incorporation of 30%, yielded the highest levels of exon 23 skipped RNA (10 and 8% of total dystrophin RNA, respectively), followed by 20mer AON LNA-8 consisting of 40% LNAs (5.3% exon skipping). The efficiencies of LNA-6 and sLNA-5 were found to be statistically significantly different when compared to 2 OMe, s2 OMe or sLNA-3 separately, using Dunn's pairwise test for multiple comparisons. While the truncation of LNA-6 (30% LNA) to its shorter analogue sLNA-5 resulted in a relatively small decrease in exon skipping activity (10 versus 8%, respectively), truncation of LNA-8 (40% LNA) to generate AON sLNA-6 considerably reduced the percentage of exon skipped RNA by more than half (5.3 versus 2.2%, respectively). Taken together, these findings demonstrate that a relatively high incorporation of LNA nucleotides (≥40%) in a short antisense construct significantly reduces AON potency. Nonetheless, a 20% inclusion of LNA nucleotides to the 2 OMe and s2 OMe sequences showed minimal to no improvement in AON activity, respectively, demonstrating that an LNA incorporation of 30%, evenly distributed at every third or fourth position, is the optimal design to achieve maximal exon skipping efficiency.
To confirm on-target specificity, a scrambled LNA/2 OMe sequence with similar percentage nucleotide composition and 33% LNA inclusion was also injected into the TA muscle of mdx mice under the same experimental conditions. In contrast to our LNA-6 and sLNA-5 AONs, no skipping of mouse dystrophin exon 23 was observed following treatment with the scrambled sequence ( Figure 1B). Furthermore, to determine the half maximal effective concentration (EC50) of our most potent AON, isolated primary mdx myoblasts were treated with increasing concentrations of LNA-6 (10, 30, 50 and 300 nM) and evaluated for exon 23 skipping 24 h post transfection. LNA-6 induced skipping of exon 23 in a dose-dependent manner, reaching highest efficiency (82%) at 50 nM. The EC50 value for LNA-6 was calculated at 13.9 nM ( Figure 1C). Quantification of exon 23-skipped mRNA is expressed as the percentage of total dystrophin mRNA. n = 3-4 mice per group; values are mean ± SD. Square brackets and asterisks indicate statistically significant differences using Dunn's multiple comparisons test followed by Benjamini-Hochberg adjustment. * denotes p ≤ 0.05 and ** denotes p ≤ 0.01. (B) Detection of exon 23-skipped dystrophin mRNA in TA muscles of mdx mice injected with an equimolar dose of a scrambled LNA/2 OMe sequence (n = 2). The contralateral TA was injected with saline only. Positive exon skipping control AONs (LNA-6 and sLNA-5) and a WT control mouse are also shown. (C) Percentage exon 23 skipping in isolated primary mdx myoblasts treated with increasing concentrations of 20mer LNA-6 AON (30% LNAs). RNA extraction was carried out 24 h post transfection. Values are mean ± SD from three independent transfection experiments. We subsequently evaluated the efficiency of LNA/2 OMe AONs to induce dystrophin restoration in vivo, by immunofluorescence staining of TA cryosections using a polyclonal antibody against the C-terminal domain of dystrophin. Dystrophin restoration was measured as the number of dystrophin-positive fibres in the whole muscle section and reported as a percentage of the WT ( Figure 2). Kruskal-Wallis analysis rendered an overall p-value of 0.0001, demonstrating statistically significant difference(s) among the efficiencies of the various AONs. In agreement with the exon skipping data, 20mer LNA-6 and 15mer sLNA-5 AONs, both consisting of 30% LNAs, showed the highest percentages of dystrophin positive fibres (22 and 19% of WT, respectively), while 15mer AONs s2 OMe and sLNA-3 (20% LNAs) induced significantly lower dystrophin restoration corresponding tõ 5% of the WT (adjusted p-values ≤ 0.01 for each pairwise comparison: LNA-6-s2 OMe, LNA-6-sLNA-3, sLNA-5-s2 OMe and sLNA-5-sLNA-3). LNA-8 and sLNA-6, exhibiting the highest percentage of LNA incorporation, were less efficient compared to their respective AONs containing 30% LNAs, underscoring again the negative impact of an increased LNA inclusion on AON activity. Notably, while inducing minimal (2%) skipping of exon 23 in TA muscles, AON LNA-4 containing 20% LNAs showed a 14% increase in dystrophin positive fibres relative to the WT. This result reflects on the interpretive bias of the immunofluorescence analysis since the number of dystrophin positive fibres is not necessarily proportional to the levels of dystrophin expression; i.e., a larger number of dystrophin positive fibres may respectively express lower levels of the protein and, conversely, a smaller number of dystrophin positive fibres may respectively express higher levels of the protein.

Mass Spectrometry Quantification of Dystrophin
To confirm dystrophin expression, we carried out a targeted mass spectrometry (MS)based protein quantification assay using aliquots of 100 ng total protein extract from WT control and AON-treated mdx mice. Protein quantification was carried out using two custom dystrophin peptides: IFLTEQPLEGLEK, used as quantifier, and LLAEELPLR, used as qualifier. To assess linearity of our assay, three WT concentration standards were prepared (10, 50 and 100% of WT extract) and used for the relative quantification of dystrophin levels in AON-treated mdx samples. As evident in Figure 3A, our assay shows linearity over a wide dynamic range (10 to 100%): peak area values for each product ion of the quantifier peptide increase proportionally to the input amount of WT protein, i.e., a 5-fold increase in peak area is observed from 10 to 50% of input protein and a 2-fold increase from 50 to 100%. Therefore, dystrophin levels in AON-treated mdx samples were determined using the sum peak area of all three product ions of the quantifier peptide. (B) Quantification of the number of dystrophin positive fibres, expressed as a percentage of the WT, in TA muscle cross sections from mdx mice treated with an intramuscular injection of the indicated AONs. n = 3 mice per group; values are mean ± SD. Square brackets and asterisks indicate statistically significant differences using Dunn's multiple comparisons test followed by Benjamini-Hochberg adjustment. * denotes p ≤ 0.05 and ** denotes p ≤ 0.01. As expected, AONs LNA-6 and sLNA-5 consisting of 30% LNAs induced the highest levels of dystrophin restoration (4.2 and 4% of WT, respectively; Figure 3B, Figure S1), followed by AONs LNA-8 and sLNA-6 with 40% LNA inclusion and 20mer LNA-4 containing 20% LNAs; all of the latter three showed similar levels of restored dystrophin expression corresponding to~1.8% of the WT (overall p-value of 0.009 using Kruskal-Wallis analysis). Notably, efficiency differences among LNA/2 OMe chimeras were more pronounced for exon 23 skipping ( Figure 1A) compared to rescued dystrophin expression: at the RNA level, increasing the percentage of LNA composition from 20 to 30% resulted in a 5-and 16-fold increase in AON activity for the 20mer and 15mer AONs, respectively, compared to a 2-and 3-fold rise when evaluating protein restoration by MS. Similarly, while 20mer and 15mer LNA/2 OMe AONs with 20 and 40% LNA compositions showed similar levels of rescued protein expression, exon skipping activity of LNA-8 (20mer with 40% LNAs) and sLNA-6 (15mer with 40% LNAs) increased by 2.5-and 5-fold, respectively, compared to their corresponding AONs with 20% LNA composition. These findings reflect highly on the sensitivity of the techniques and how RNA data may correlate with corresponding data at the protein level.

Western Blot Quantification of Dystrophin
Dystrophin restoration was further evaluated by chemiluminescent Western blotting using a multi-point standard curve to assess for potential oversaturation on the gels, which would lead to overestimation of protein levels at the low end of the range. A four-point standard curve was constructed by spiking quantified aliquots of a WT protein extract (2.5, 5, 10 and 20 µg) into mdx lysates to produce 2.5, 5, 10, and 20% concentration standards, respectively ( Figure 4A). Data were analysed by linear regression, generating a coefficient of determination (R2) value of 0.9726, and the identified equation (y = 0.1043x) was used to calculate the amount of restored dystrophin in AON-treated mdx samples. The standard curve demonstrates good linearity of the chemiluminescent signal over the indicated concentration range (2.5 to 20%), which corresponds to the percentage range of dystrophin positive fibres observed in muscles sections of AON-treated mice ( Figure 3B). In contrast to the protein expression profiles obtained by immunofluorescence and targeted MS analysis, chemiluminescent detection of Western blot products revealed a 2.8-and 11-fold decrease in the efficiencies of 15mer LNA/2 OMe AONs consisting of 30 or 40% LNAs, respectively, compared to their corresponding 20mer analogues ( Figure 4B; overall p-value of 0.0026 using Kruskal-Wallis analysis). Nonetheless, in agreement with the exon skipping data, 20mer and 15mer LNA/2 OMe AONs consisting of 30% LNAs were more potent than their respective 2 OMe/PS and LNA/2 OMe counterparts exhibiting a higher (40%) or lower (20%) percentage LNA composition.

Discussion
AON-mediated exon skipping has been in the forefront of DMD therapeutic research for the past 20 years. Two chemically modified oligonucleotide designs, a PMO and a 2 OMe/PS AON, have entered clinical trials, with the former receiving provisional FDA approval based on minor increases in rescued dystrophin expression. By contrast, the 2 OMe/PS AON was associated with adverse systemic and injection-site reactions and was therefore withdrawn from consideration. Attempting to limit the toxicity associated with the PS modification, Le et al. designed a series of truncated 2 OMe/PS AONs, relying on the introduction of LNA nucleotide monomers to retain a high RNA binding affinity. Indeed, the introduction of four LNAs in a 14mer 2 OMe/PS AON, showing no exon skipping activity, significantly improved the yield of exon 23-skipped RNA in H2K mdx mouse myotubes [44]. To further evaluate the optimal length and percentage chemistry composition of LNA/2 OMe chimeras for DMD therapy, we created a series of 15mer and 20mer AONs, with variable LNA compositions (20-40%), and compared their efficiencies in inducing exon 23 skipping and rescued protein expression in locally injected muscles of mdx mice.
When optimising novel oligonucleotide designs for clinical therapy, the sensitivity of the various methodologies is crucial for determining subtle differences in the efficiency of the tested AONs. In the present study, AON efficacy was assessed at RNA level by densitometric analysis of PCR products and at protein level by semi-quantitative immunofluorescent analysis, quantitative MS analysis and quantitative chemiluminescent WB. While comparing the data obtained by the various methodologies, we observed that efficiency differences among LNA/2 OMe chimeras with variable LNA compositions were significantly more prominent for exon skipping percentages analysed by RT-PCR and rescued protein levels determined by WB. Considering that dystrophin gene expression is rather low in healthy tissues, accurate quantification of low exon-skipping samples would be particularly challenging. The highest exon skipping percentage reported in this study was 10%, for 20mer LNA-6 (30% LNAs), and the lowest was 0.4%, for 15mer AONs s2 OMe and sLNA-3 (20% LNAs). In a comparative study of different methodologies to quantify exon skipping percentages in AON-treated mdx muscle tissues, samples with exon skipping percentages between 4 and 10%, as determined by an absolute quantification method, showed a 30-50% value variation following densitometric analysis of ethidium bromide-stained PCR products [48]. Notably, this was not observed for samples with reference exon skipping values between 28 and 30%. Therefore, while relatively accurate, densitometric analysis of PCR products could lead to an under-or over-estimation of the real values in samples with low exon skipping percentages.
While LNA-6 (20mer with 30% LNAs) and sLNA-5 (15mer with 30% LNAs) AONs showed similar levels of rescued dystrophin expression by MS analysis, WB quantification of dystrophin levels revealed a 2.8-fold decrease in the efficiency of the sLNA-5 AON compared to its longer analogue. WB followed by chemiluminescent detection has served as a golden standard for dystrophin quantification in both clinical and pre-clinical studies; yet, the robustness of this method is limited by poor transfer qualities, saturation issues and poor reproducibility. During an assessment of dystrophin levels in patients treated with eteplirsen, researchers identified an average coefficient of variation of 16% while performing sample duplicates in separate gels [49]. This highlights the importance of performing at least two technical replicates per biological replicate, something that was not carried out in the present study. Furthermore, while MS shows higher reproducibility compared to WB, it has been reported to demonstrate low resolution for dystrophin values under 5% of WT (dystrophin values obtained by MS analysis in the present study were between 0.5 and 4.5%) [49]. Considering all of the aforementioned limitations, and based on the results of the present study, we recommend that dystrophin restoration should always be quantified using all three methodologies, to allow for a more robust correlation between RNA and protein expression profiles and thus a more accurate comparison of AON efficiencies during pre-clinical assessment.
Collectively, the data obtained demonstrate that 15mer and 20mer LNA/2 OMe AONs consisting of 30% LNAs, evenly distributed at every third or fourth position, are significantly more potent in inducing exon 23 skipping and restoring dystrophin expression in locally injected mdx muscles, compared to a previously tested 2 OMe AON and LNA/2 OMe chimeras with lower or higher LNA compositions. In agreement with current findings, recently published work from our laboratory has shown that 14-18mer LNA/2 OMe AONs with a 33% LNA composition and a full PS backbone, targeting the expanded CTG-repeat mutation that causes DM1, were more efficient in correcting DM1 molecular defects in vitro compared to an 18mer LNA/2 OMe AON containing half the number of LNA nucleotides and LNA/DNA mixmers of similar or higher percentage LNA composition [45]. More importantly, systemic delivery of the 14mer and 18mer LNA/2 OMe chimeras in DM1 mice resulted in relatively high levels of AON accumulation in skeletal muscles, diaphragm and heart tissue, with no profound histological toxicity.
Efficient delivery of AONs to the cardiac tissue is of great therapeutic relevance, as almost all DMD patients will manifest signs of cardiomyopathy by the age of 18. Consequently, failure of dystrophin restoration in cardiac tissue, in the presence of rescued protein expression in skeletal muscles, may exacerbate any subclinical heart failure due to increased voluntary muscle activity [50,51]. Delivery to the heart and skeletal muscle tissues has been improved by conjugating positively charged arginine-rich peptides to charge neutral PMOs [52][53][54]. Therefore, future systemic evaluation of our LNA/2 OMe chimeras may be further exploited using a combination of muscle-homing, hydrophobic ligands that would potentially enhance AON accumulation and efficacy in all target tissues. While intramuscular delivery may be sufficient to screen a larger number of AON designs with different chemical compositions, it has limited capacity in providing solid evidence on the therapeutic efficacy of an antisense sequence. Systemic delivery of our two most potent AONs, LNA-6 and sLNA-5, is further warranted to evaluate the benefit-toxicity profile of the respective LNA/2 OMe sequences for DMD therapy.

Antisense Oligonucleotides
All AONs used in the study, listed in Table 1, were designed and produced by the laboratory of Professor Jesper Wengel. We have chosen the most efficient 2 OMe/PS AON sequence from previous studies, which targets the donor splice site of exon 23 in the mouse dystrophin pre-mRNA [46,47]. For simplicity of nomenclature, the AON was named 2 OMe. A varying number of LNA nucleotides was introduced throughout the 20mer 2 OMe sequence to generate chimeric LNA/2 OMe AONs with LNA percentage incorporations of 20, 30 and 40%. These were named LNA-4, LNA-6 and LNA-8, respectively, to refer to the number of LNA nucleotides introduced in the sequence. Truncated, 15mer analogues of the 2 OMe and LNA/2 OMe oligonucleotides were also generated, distinguished by the letter 's' at the beginning of the AON name. All AONs were kept on a fully modified PS backbone. An 18mer CAG-repeat LNA/2 OMe AON with a full PS backbone and an LNA incorporation of 33% was also included in the study to serve as a control scrambled sequence.

RNA Extraction and RT-PCR
Frozen tissues were cryosectioned in the transverse plane at 5 µm thickness and 100 µm intervals. Intervening muscle sections were collected for immunofluorescence, Western blot and targeted mass spectrometry analysis. Sections were homogenized in TRIzol reagent using a Precellys tissue homogenizer and total RNA was then extracted according to the manufacturer's instructions. RT-PCR was carried out using 300 ng of total RNA in a 50 µL reaction, using the One-step RT-PCR kit (Qiagen, Hilden, Germany) and external primers Ex20F (5 -CAGAATTCTGCCAATTGCTGAG-3 ) and Ex26R (5 -TTCTTCAGC TTGTGTCATCC-3 ). The cycling conditions were 95 • C for 1 min, 55 • C for 1 min and 72 • C for 2 min for 30 cycles. PCR products were examined by electrophoresis on a 2% agarose gel and densitometric analysis was carried out using Image J. The percentage of exon skipping was calculated as the percentage of transcripts in which exon 23 is skipped relative to the sum of non-skipped and exon 23-skipped dystrophin transcripts. The RT-PCR experimental procedure was implemented according to Spitali et al., 2010: "Accurate quantification of dystrophin mRNA and exon skipping levels in Duchenne muscular dystrophy" [48]. To detect exon skipping in AON-transfected primary mdx myoblasts (see section below), total RNA was extracted by TRIzol reagent and 300 ng were subjected to RT-PCR analysis using 27 cycles of amplification instead of 30.

Isolation of Primary Mdx Myoblasts and Transfection
The protocol for isolating satellite cells from single muscle fibres was adapted from Rosenblatt and colleagues [55]. In brief, extensor digitorum longus (EDL) muscles were isolated from 8-week-old mdx mice and incubated in 0.2% type I collagenase (Sigma-Aldrich, St. Louis, MS, USA) in Dulbecco's Modified Eagle Medium (DMEM, high glucose, pyruvate, no glutamine; Thermo Fisher Scientific, Waltham, MA, USA), supplemented with 1× penicillin-streptomycin (PS; Thermo Fisher Scientific) and 1× glutamax solution (Thermo Fisher Scientific), at 37 • C for 1 h. Following digestion, single muscle fibres were liberated by repeatedly triturating the EDL muscle using a wide-mouth glass pipette. Viable single muscle fibres were carefully transferred to a new culture dish, in three rounds of washing, in order to discard hypercontracted fibres and tissue debris. Finally, viable muscle fibres were transferred to a new culture dish coated with 1 mg/mL Matrigel (Corning) in DMEM (1× PS/1× glutamax) and incubated in 'plating medium' (DMEM supplemented with 1× PS, 1× glutamax, 10% horse serum (HS; Thermo Fisher Scientific), 0.5% chick embryo extract (CEE; MP Biomedicals)) at 37 • C for 3 days, to allow myogenic satellite cells to dissociate from their fibre. Three days after plating, the muscle fibres were removed from the dish and the 'plating medium' was replaced with 'proliferation medium' consisting of 20% fetal bovine serum (FBS; Thermo Fisher Scientific), 10% HS, 1% CEE and 2.5 ng/mL recombinant murine fibroblast growth factor-basic (FGF-b; PeproTech, London, UK) in DMEM (1× PS/1× glutamax). Myogenic cells were allowed to grow in 5% CO 2 and 37 • C for 2-3 days, before plating for transfection. For the dose-response study, primary mdx myoblasts were seeded in 12-well plates, at 60% confluency, and transfected with increasing concentrations of AON LNA-6 (10, 30, 50 and 300 nM), complexed with Lipofectamine 3000 reagent in Opti-MEM Reduced Serum Medium. Exon skipping efficiency was measured 24 h post transfection by RT-PCR for exon 23 skipping. Gel bands were measured using Image J and half-maximal effective concentration (EC50) was calculated using the sigmoid Emax dose-response model in the DoseFinding R package (Date: 5 July 2019).

Immunofluorescence
Transverse cryosections of 7 µm thickness were examined for dystrophin expression using a rabbit polyclonal anti-dystrophin antibody (1:400 dilution, RB-9024-P Thermo Scientific), which was detected using the Alexa Fluor 594 goat anti-rabbit antibody (molecular probes ref 37117). Dystrophin-labelled sections were imaged at low magnification (5×) and composite images of the entire transverse section were constructed from overlapping images. The number of dystrophin positive fibres in the entire muscle cross section was then counted using the cell count function in Image J (3 sections per TA muscle were counted). The percentage of dystrophin positive fibres for each AON was calculated relative to the average number of dystrophin positive fibres in WT control sections.

Sample Preparation for Mass Spectrometry
Intervening cryosections from WT control and AON-treated mdx muscle tissues were lysed in protein homogenization buffer (refer to Section 4.8) and aliquots of total protein extract were further processed using a modified filtered-aided sample preparation (FASP) protocol [56]. Briefly, 100 ng of protein was transferred to a centrifugal filter (Pall Nanosep, 30 kDa MWCO), washed with urea buffer, then reduced with 100 µL DTT (8 mM in 50 mM ammonium bicarbonate) for 15 min at 60 • C and alkylated with 100 µL iodoacetamide (50 mM in 50 mM ammonium bicarbonate) for 20 min in the dark, at room temperature. Finally, the samples were washed with 100 µL of 50 mM ammonium bicarbonate (×3) and digested with 2 µg trypsin (proteomics grade, Roche Diagnostics GmbH, Mannheim, Germany) at 37 • C for 18 h. Peptides were collected by centrifugation for 10 min at 11,000× g, acidified with TFA to a final concentration of 0.2% and desalted using reverse phase solid phase extraction cartridges (Sep-Pak C18, Waters, UK). Eluates were lyophilized using a centrifugal vacuum concentrator. Peptide pellets were re-dissolved in 1% acetonitrile, 0.1% formic acid to yield an approximate concentration of 250 ng/µL (determined by NanoDrop measurement at 280 nm).

Mass Spectrometry Analysis
Custom dystrophin peptides LLAEELPLR and IFLTEQPLEGLEK were purchased from GenScript Biotech (The Netherlands). Peptides were suspended in 1% LC-MS grade acetonitrile containing 0.1% formic acid, pooled together and further diluted to a final concentration of 10 fmol/µL. To determine the retention time of each peptide, 5 µL of the peptide mix were loaded onto a C18 column (Acquity UPLC M-Class Peptide CSH, 75 µm × 250 mm, 1.7 µm) and separated at a flow rate of 300 nL/min using a nanoAcquity UPLC system (Waters, UK). The mobile phases consisted of 0.1% formic acid in water (mobile phase A) and 0.1% in acetonitrile (mobile phase B). Peptides eluted with a linear gradient from 5% B to 31% B over 57 min (elution profile is shown in Supplementary Table S1). An HD-MRM (High-Definition Multiple Reaction Monitoring) MS assay, which is based on ion mobility mass spectrometry technique, was developed. The method employs additional separation of peptide fragments based on their collisional cross section (CCS), followed by high resolution MS detection and is implemented on an ion mobility-enabled quadrupole time-of-flight hybrid mass spectrometer [57]. The targeted MS assay was implemented on a Synapt G2Si HDMS instrument equipped with the NanoLockSpray source and operated on the HD-MRM acquisition mode. Fragmentation was performed on the Trap T-Wave  Table S2. Data analysis and interpretation were performed using Skyline [58].

Western Blot
Intervening cryosections from untreated and AON-treated mdx TA muscles were collected in Precellys CK14 tubes and lysed in 120 µL of homogenization buffer containing 1.25 mM Tris-HCl pH 6.8, 10% Glycerol, 4% SDS, 4 M Urea, 10% b-mercaptoethanol and 200 µL protease inhibitor (25x). Protein extracts were then denatured at 95 • C for 5 min and centrifuged to collect the supernatant. A sample of the supernatant mixture was acetonetreated and used to determine protein concentration using the Bradford assay (Biorad). In total, 100 µg of protein extract from treated and untreated TA muscles were loaded on 6% SDS-PAGE gels. The samples were electrophoresed for 3.5 h at 80 V and blotted to a PVDF membrane at 30 V overnight at 4 • C. The membrane was probed overnight at 4 • C with 1:500 dilution of rabbit anti-dystrophin antibody (Thermo fisher 9024-P1) in 5% skimmed milk. The bound antibody was detected by incubation with a horseradish peroxidise conjugated goat anti-rabbit antibody (1:5000 dilution, Jackson laboratories) in 5% milk in PBS-T and the Lumi Sensor chemiluminescent HRP Substrate kit (L00221V300). The membranes were also probed overnight at 4 • C with rabbit anti-vinculin (1:1000 dilution, Abcam ab73412) in 5% skimmed milk, which served as a loading control. Band intensities were analysed using Image J and dystrophin values were normalised with respect to vinculin.
To construct a standard curve for WT dystrophin, aliquots of a pooled (4 TA muscles) WT protein extract (2.5, 5, 10 and 20 µg) were spiked into aliquots of a pooled mdx extract (97.5, 95, 90 and 80 µg, respectively) to maintain an equal loading of protein (100 µg). Protein homogenization, sample preparation and Western blotting were carried out as described above. Dystrophin band signals were normalized to the intensity of the respective vinculin signal and data were analysed by linear regression. The identified equation of the linear regression line was used to calculate the amount of restored dystrophin in AON-treated mdx samples.

Statistical Analysis
All data are expressed as mean ± standard deviation (SD). Statistical significance was determined using Kruskal-Wallis one-way analysis of variance followed by Dunn's (1964) pairwise test for multiple comparisons. p-values were adjusted using the Benjamini-Hochberg (false discovery rate) method.

Conclusions
In vivo screening of different LNA/2 OMe antisense designs for targeted exon 23 skipping in mdx TA muscles revealed that AONs consisting of 30% LNAs are more potent in inducing exon 23 skipping and restoring dystrophin protein expression, compared to a 2 OMe-modified AON and LNA/2 OMe chimeras with lower (20%) or higher (40%) LNA compositions. Our data further demonstrate the importance of using parallel techniques (i.e., immunohistochemistry, mass spectrometry and Western blotting) for the quantification of dystrophin protein restoration, to allow for a more robust correlation between RNA and protein expression profiles and a more accurate comparison of AON efficiencies during pre-clinical assessment. Systemic evaluation is now warranted to prove the therapeutic potential of our most potent LNA/2 OMe design, comprising of 30% LNAs, evenly distributed at every third of fourth position of the AON sequence.

Conflicts of Interest:
The authors declare no conflict of interest.