Bi-specific splice-switching PMO oligonucleotides conjugated via a single peptide active in a mouse model of Duchenne muscular dystrophy

The potential for therapeutic application of splice-switching oligonucleotides (SSOs) to modulate pre-mRNA splicing is increasingly evident in a number of diseases. However, the primary drawback of this approach is poor cell and in vivo oligonucleotide uptake efficacy. Biological activities can be significantly enhanced through the use of synthetically conjugated cationic cell penetrating peptides (CPPs). Studies to date have focused on the delivery of a single SSO conjugated to a CPP, but here we describe the conjugation of two phosphorodiamidate morpholino oligonucleotide (PMO) SSOs to a single CPP for simultaneous delivery and pre-mRNA targeting of two separate genes, exon 23 of the Dmd gene and exon 5 of the Acvr2b gene, in a mouse model of Duchenne muscular dystrophy. Conjugations of PMOs to a single CPP were carried out through an amide bond in one case and through a triazole linkage (‘click chemistry’) in the other. The most active bi-specific CPP–PMOs demonstrated comparable exon skipping levels for both pre-mRNA targets when compared to individual CPP–PMO conjugates both in cell culture and in vivo in the mdx mouse model. Thus, two SSOs with different target sequences conjugated to a single CPP are biologically effective and potentially suitable for future therapeutic exploitation.


INTRODUCTION
Splice-switching oligonucleotides (SSOs) are currently very promising for therapeutic use in both Duchenne muscular dystrophy (DMD) through exon skipping and for Spinal muscular atrophy (SMA) through promotion of exon inclusion. Such SSOs are designed to sterically block splice sites or specific binding motifs for splicing machinery in order to promote exon inclusion or exclusion. For DMD, targeting of the dystrophin pre-mRNA with a SSO is used to 'skip' an exon that contains a nonsense coding mutation, or to remove an exon neighbouring an out-of-frame genomic deletion, so as to restore the mRNA reading frame. This allows synthesis of an internally deleted dystrophin protein that retains the elements crucial for function (1)(2)(3).
Proof-of-principle of this approach has been demonstrated in murine and canine animal models of DMD (4-7) and more recently in human phase II/III clinical trials (8)(9)(10)(11)(12). Various SSO chemistries have been developed for exon skipping including phosphorodiamidate morpholino oligonucleotides (PMO), peptide nucleic acids (PNA), locked nucleic acid (LNA), 2 -O-methoxyethyl phosphorothioate oligonucleotides (2 -MOE/PS), 2 -Omethyl phosphorothioate oligonucleotides (2 -OMe/PS) and tricyclo oligonucleotides (see (2) for a recent review). The two major SSO chemistries that have been used more extensively and which are in current clinical trials in DMD patients are 2 -OMe/PS (9,11) and PMO (8,10,12). In these trials, production of dystrophin protein has been demonstrated, albeit at low levels, and rather disappointingly a recent phase III trial with 2 -OMe-PS SSO was not able to meet its primary endpoint of a statistically significant improvement in the 6 min walk test (13). Nevertheless, preliminary results with PMO chemistry are promising and further studies are planned to target additional exons (8,9). To date, clinical trials have focused on targeted removal of exon 51, since this would benefit the largest patient pool, but SSOs are also being developed to target other exons between 45 and 55.
The most significant barrier to success for spliceswitching therapies has been effective delivery. In the case of PMO, and indeed other chemistries, the level of exon skipping, and hence the amount of functional dystrophin restored in muscle, is poor unless high doses are used. This is thought to be due mostly to their rapid clearance from the body following systemic administration, as well as their poor ability to penetrate cellular barriers and reach their nuclear target site. One approach to address this has been to attach cell-penetrating peptides (CPPs) that can effectively carry charge neutral PMO cargos across cell membranes to their pre-mRNA target site in the nucleus. In particular, PMOs conjugated to arginine-rich CPPs (known as P-PMOs) have been shown to enhance dystrophin production in muscle following systemic administration in the mdx mouse model of DMD (14)(15)(16). We have previously developed a series of novel arginine-rich CPPs known as PNA/PMO internalization peptides (Pips), comprised of two arginine-rich sequences separated by a central short hydrophobic core sequence. These Pip peptides were designed to improve serum stability whilst maintaining a high level of exon skipping, initially by attachment to a PNA cargo (17). Further derivations of these peptides were designed as conjugates of PMOs, which were shown to lead to high body-wide skeletal muscle dystrophin production, and importantly also including the heart, following systemic administration (5). A later version (Pip6a-PMO) proved to be an even more efficient conjugate in mediating dystrophin production in the mdx mouse (18).
Whilst restoration of the absent dystrophin protein is the primary goal for genetic therapies for DMD, consideration of complementary therapies to reduce pathological features of disease or to improve muscle function are also very important. One such strategy would be to promote muscle growth through targeting of the myostatin pathway. Myostatin, or growth/differentiation factor-8 (GDF8), is a member of the transforming growth factor-␤ (TGF-␤) family, which is involved in muscle homeostasis and acts to inhibit muscle growth (19). Myostatin is involved in control of myogenesis through binding to activin type IIB receptor (AcvRIIb) (20,21) which recruits and activates activin type I receptor (ALK4 or ALK5) (22). Activin receptor activation results in phosphorylation of intracellular signalling mediators Smad2 and Smad3 that translocate to the nucleus to affect gene transcription. There is evidence that inhibition of the myostatin pathway has the potential for clinical benefit in DMD. Transgenic knockout models for both myostatin and dystrophin demonstrate increased musculature due to fibre hypertrophy as well as reduced fibrosis and fat deposition, compared to mdx mice alone (23). Similarly, when mdx mice were treated with anti-myostatin antibodies this resulted in enlarged muscles concurrent with improved muscle function and strong reduction in diaphragm fibrosis (24). Based on these improved muscle features, the concept of myostatin down-regulation concurrent with dystrophin restoration has been investigated. Dumonceaux et al. reported the use of Adeno-associated virus (AAV) constructs to combine RNAi-mediated down-regulation of AcvRIIB with a U7-based small RNA exon skipping technique to restore dystrophin (25). Whilst concurrent treatment did not improve muscle mass, absolute and specific forces were much greater compared to either individual strategy. A subsequent study by Hoogaars et al., utilized soluble AcvRIIB decoy receptors in combination with AAV-U7 mediated dystrophin restoration to treat mdx mice (26). Treatment with decoy AcvRIIB increased body weight, with morphometric measurements of muscle fibres suggesting that muscle growth was due to hypertrophy (20,21,27). Based on the potential success of this approach, we sought to simultaneously target both the dystrophin and the myostatin pathway as a molecular model to evaluate the efficacy of using bi-specific PMO compounds.
A splice-switching approach has been used previously to target and down-regulate expression in the myostatin pathway, with SSOs developed to target both the Mstn (28) and Alk5 (29) transcripts. Whilst the principle of multiple-exon targeting for both dystrophin restoration and myostatin depletion has been demonstrated before using a cocktail of individual SSOs (30)(31)(32), there are advantages in development of a bi-specific compound to target two pre-mRNAs. First, the outcome of use of a bi-specific SSO is that both SSOs must enter the same cell, whereas for a cocktail of SSOs there will likely be a mixed population of cells where either none, one or both genes are targeted. More importantly, since cell and in vivo toxicity of P-PMOs is thought to be predominantly due to that of the peptide, the use of a single peptide to deliver both PMOs halves the total peptide requirement compared to use of a cocktail of two separate P-PMOs, and this may help to reduce the potential for peptide-mediated toxicity. Note also that two 2 -OMe/PS SSOs that target both dystrophin and myostatin have been joined together recently to make a bi-specific construct, but without any attached delivery peptide, but in this case activity in mdx mice was not seen for the dystrophin SSO (33).
In our study, we selected Pip6a (18) as the CPP to simultaneously deliver two different PMOs as a bi-specific conjugate and to develop the chemistries for their attachment. Three different conjugation chemistries including amide, disulfide and triazole bonds were utilized to allow orthogonal conjugation. The first PMO targets exon 23 of the dystrophin gene to correct the mdx genotype and the second to target removal of exon 5 of the Acvr2b gene so as to produce an internally deleted protein that lacks the crucial trans-membrane domains. Several different bi-specific conjugate designs were investigated whereby two PMOs were joined either at one end of the Pip6a peptide or with one PMO at either N-or C-termini. The activities of these conjugate constructs were assessed in mouse mdx cells and the most active bi-specific conjugates (D2 and D3), which had both PMOs attached at the C-terminus of the CPP, were shown to have closely comparable Dmd exon skipping activity to the single Pip6a-PMO targeting Dmd. D2 and D3 conjugates were also assessed in the same cells for targeting of Acvr2b and both conjugates demonstrated only very slightly reduced exon skipping activity compared to Pip6a-PMO targeting Acvr2b. Importantly, the cell viability using a bi-specific compound was significantly better than for a mixture of the two individual Pip6a-PMOs. We furthermore assessed the potential of this approach in an in vivo environment through intramuscular administration and demon-Nucleic Acids Research, 2015, Vol. 43, No. 1 31 strated that there were no significant differences in exon skipping activities for both Dmd and Acvr2b targets between bi-specific conjugates (D2 and D3) and a cocktail of the individual P-PMO equivalents.

Materials
Fmoc-protected amino acids, coupling reagents (HBTU and PyBOP) and the Fmoc-␤-Ala-OH preloaded Wang resin (0.19 mmol g −1 ) were obtained from Merck (Hohenbrunn, Germany). Fmoc-azido-L-lysine-OH was from IRIS Biotech GMBH (Deutschland, Germany). Fmoc-Lbis-homopropargylglycine-OH (Bpg) was purchased from Chiralix (Nijmegen, The Netherlands). Chicken embryo extract (CEE) and horse serum (HS) for cell culture were obtained from Sera Laboratories International Ltd (West Sussex, UK). ␥ -Interferon was obtained from Roche Applied Science (Penzberg, Germany). All other reagents were obtained from Sigma-Aldrich (St Louis, MO, USA) unless otherwise stated. MALDI-TOF mass spectrometry (Table 1) was carried out using a Voyager DE Pro BioSpectrometry workstation. A stock solution of 10 mg ml −1 of ␣-cyano-4-hydroxycinnamic acid or sinapinic acid in 60% acetonitrile in water was used as matrix. The measurements have an accuracy level of ±0.1%.

Peptide synthesis
Peptides were synthesized by standard Fmoc chemistry (34) using a CEM Liberty TM microwave peptide synthesizer (Buckingham, UK). Peptides were assembled on Fmoc-␤-Ala-OH preloaded Wang resin on a 0.1 mmol scale with excess of Fmoc-protected amino acids, PyBOP and DIPEA (5:5:10). The N␣-Fmoc protecting groups were removed by treating the resin with piperidine in DMF (20% v/v) at 75 • C twice, once for 30 s and then for 3 min. The coupling reactions were carried out at 75 • C for 5 min. In order to prevent racemization, Fmoc-cysteine (Trt)-OH was coupled at 50 • C for 10 min at 60 W microwave power. All amino acids were single coupled except for the arginines, which were double coupled. The Fmoc-L-bis-homopropargylglycine-OH was coupled manually using a 2-fold excess and the coupling success was checked using a TNBS test (35). After completion of peptide assembly, the resin bound peptide was cleaved off by treating the resin with a cocktail of TFA:DoDt:H 2 O:TIPS (94:2.5:2.5: The peptide was precipitated by addition of ice-cold diethyl ether and washed three times. The crude peptides were analysed and purified to >95% by reversed-phase HPLC (RP-HPLC). The peptide mass characterization was carried out using a MALDI-TOF mass spectrometry (ABI Voyager DE Pro) and an ␣-cyano-4-hydroxycinnamic acid matrix made up in 70% acetonitrile containing 0.05% TFA.

Functionalization of PMO
The PMO sequence for exon 23 skipping of Dmd pre-mRNA (5 -GGCCAAACCTCGGCTTACCTGAAAT) was either unmodified (standard morpholino with a secondary amine at the 3 end) or functionalized with a disulfide at its 3 -end. The PMO targeting exon-5 of Acvr2b was unmodified (5 -GCCTCGTTTCTCGGCAGCAATGAAC-3 ). All PMOs were purchased from Gene Tools LLC (Philomath, USA). 3 -Unmodified PMO was functionalized with an azido group by coupling the free 3 -secondary amine group with Fmoc-azido-L-lysine-OH ( Figure 1A). The coupling was carried out by activating the carboxyl group of the amino acid derivative using HBTU (2.5 eq.) and HOAt (2 eq.) in NMP in the presence of 2.5 eq. of DIEA before addition of the PMO dissolved in dimethylsulfoxide (DMSO). The Fmoc-azido-L-lysine-PMO conjugate was then purified using RP-HPLC followed by Fmoc deprotection and purification. In the case of PMO with a disulfide bond at its 3 end ( Figure 1B), the disulfide bond was reduced to give a free sulfhydryl group using a 10-fold excess of tris (2-carboxyethyl)phosphine hydrochloride (TCEP·HCl) in water for 1 h followed by filtration to remove the excess TCEP. The PMO with a free sulfhydryl group was then activated using a 2.5-fold molar excess of 2,2 -dithiobis (5-nitropyridine) (DTNP) in DMSO: acetonitrile (0.1% TFA):H 2 O (0.1% TFA) with (1:1:3) ratios (36). The reaction mixture was stirred at room temperature for 2 h and the NPys-activated PMO was purified by RP-HPLC.

Peptide-PMO conjugations
Conjugations of peptides to PMOs ( Figure 2) were carried out in solution using a 2.5-fold excess of peptide using similar conditions to the coupling of Fmoc-azido-L-lysine-OH to PMO. The conjugation of the second PMO was carried out using either copper (I) mediated alkyne-azide click chemistry between the alkyne-functionalized P-PMO (Figure 2A) and the azide-functionalized second PMO or by forming a disulfide bond between the 3 NPys-activated second PMO and a free cysteine thiol of the P-PMO (Figure 2B). The alkyne-azide click reaction between azidefunctionalized PMO and alkyne functionalized P-PMO was carried out by dissolving the P-PMO in water followed by addition of azido-functionalized PMO (1.2 eq.). Sodium ascorbate (10 eq. as a 20-mM solution) was added and the reaction mixture was vortexed thoroughly followed by addition of copper (II)-TBTA (12 eq. as a 20 mM solution) (37). The click reaction was carried out at room temperature for 6h or at 40 • C for 30 min. The conjugation of 3 NPysactivated PMO to P-PMO was carried out by first dissolving the Npys-activated PMO in ammonium bicarbonate solution (pH 8) followed by addition of the P-PMO dissolved in water. The reaction mixture was stirred at room temperature for 1 h.
The single and dual P-PMO conjugates were purified on a high-resolution (HR)-16 cation-exchange column (GE Healthcare, USA) using 25 mM sodium phosphate buffer (pH 7.2) containing 25% acetonitrile. The conjugates were eluted using a 1 M NaCl solution in the same buffer at a flow rate of 6 ml min −1 . The excess salts were removed by centrifugation using an Amicon R Ultra-15 3K centrifugal filter device. The conjugates were characterized using MALDI-TOF MS as mentioned above. They were dissolved in ster-  ile water and filtered through a 0.22 m cellulose acetate membrane (Costar) before use.

Cell culture and transfection
Mouse H2K/mdx myoblasts were plated at a density of 5 × 10 5 cells per well in a gelatin (0.01%) pre-coated 24well plate. H2K/mdx myoblasts were grown in high-glucose Dulbecco's modified Eagle's medium (DMEM) supplemented with 20% foetal calf serum (FCS), 2% CEE and 0.002% of interferon-␥ at 33 • C. The myoblast cells were differentiated into myotubes for the exon-skipping assay. Myoblasts were differentiated for 4 days in DMEM supplemented with 5% HS at 37 • C prior to transfection of P-PMOs in serum-free Opti-MEM for 4 h at 37 • C. The transfection medium was then replaced with DMEM/5% HS and cells incubated for a further 20 h at 37 • C.

Intramuscular administration
Experiments were carried out in the Biomedical Sciences Unit, University of Oxford according to procedures authorized by the UK Home Office. Experiments were carried out in mdx mice (C57BL/10ScSn-Dmd mdx /J) (38). Intramuscular (IM) injections (n = 3 per treatment) were carried out on 24-week-old mdx mice under general anaesthesia. 0.5 nmol peptide-PMO in 30 l 0.9% saline volume was injected into tibialis anterior (TA) muscle. Two weeks postadministration animals were sacrificed by rising CO 2 inhalation and tissues snap-frozen in a dry ice cooled isopentane bath and stored at −80 • C.

RT-PCR and qPCR analysis
RNA was extracted from either H2Kmdx cell pellets or from TA tissue sections by mechanical disruption; and subsequently processed using Trizol according to manufacturer's instructions (Life Technologies). RT-PCR analysis of exon skipping levels was carried out with 400 ng of total RNA used as a template in a 50 l RT-PCR using the GeneAmp RNA PCR kit (Applied Biosystems, Warrington, UK). RT-PCR amplification of the dystrophin Dmd transcript was carried out under the following conditions: 95 • C for 20 s, 58 • C for 60 s and 72 • C for 120 s for 30 cycles using the following primers: DysEx20Fo (5 -CAGAATTCTGCCAATTGCTGAG) and DysEx26Ro (5 -TTCTTCAGCTTGTGTCATCC). Two microlitres of this reaction was used as a template for nested amplification using Amplitaq Gold (Applied Biosystems, Warrington UK) under the following conditions: 95 • C for 20 s, 60 • C for 4 s and 72 • C for 120 s for 22 cycles using the following primers: DysEx20Fi (5 -CCCAGTCTACCACCCTATCAGAGC) and DysEx26Ri (5 -CCTGCCTTTAAGGCTTCCTT). Acvr2b RT-PCR amplification was carried out under the following conditions: 95 • C for 20 s, 60 • C for 45 s and 72 • C for 60 s for 35 cycles using the following primers: Acvr2bEx4F (5 -CTGCGTTTGGAAAGCTCAGCTCAT) and Acvr2bEx9R (5 -AAGGGCAGCATGTACTCATCGACA). PCR products were analysed on 2% agarose gels. For quantitative analysis of exon skipping levels, 1 g of RNA was reverse transcribed using the High Capacity cDNA RT Kit (Applied Biosystems, Warrington, UK) according to manufacturer's instructions. qPCR analysis was carried out using 25 ng cDNA template and amplified with Taqman Gene Expression Master Mix (Applied Biosystems, Warrington, UK) on a StepOne Plus Thermocycler

MTS cytotoxicity assay
The levels of cytotoxicity of P-PMOs were assessed in human hepatocytes

Bi-specific P-PMO design and synthesis
Novel bi-specific PMO compounds were developed that involved use of standard peptide synthesis methods for synthesis of the functionalized peptide component. The PMO components were obtained using initially unmodified PMO that was then functionalized at its 3 -end with an azido group ( Figure 1A) to enable 'click' conjugation. Alternatively, a 3 -disulfide functionalized PMO was used to prepare a 3 -NPys-activated PMO ( Figure 1B). The yield of 3 -Npys PMO was 55%, whereas the 3 -azido PMO was in lower yield of 19.5% because of the two-step RP-HPLC purification used.
The CPP chosen for the constructions was Pip6a (Figure 2A). Two different types of bi-specific compounds were designed. In the first, the two PMO oligonucleotides were each conjugated to a different end of the Pip6a peptide ( Figure 2A) (designated D1) or in the second where the PMO oligonucleotides were both attached at the carboxyterminal end of Pip6a (designated D2 and D3). In the case of click chemistry conjugation to PMO, Pip6a was synthesized with an alkyne group either at its N-terminus for D1, or at its C-terminus for D2 and D3 (Figure 2A). In D1, PMO (Dmd) was conjugated to the C-terminus through an amide bond and the 3 -azido-PMO (Acvr2b) was conjugated at the N-terminus through a triazole bond. For bi-specific conjugate D2, the Acvr2b PMO was click conjugated, whereas in conjugate D3 the Dmd targeting PMO was click conjugated. For click conjugations at the C-terminus of Pip6a, one ␤-alanine (B) and one aminohexanoic acid (X) spacer residue were incorporated on ei-ther side of the Bpg alkyne derivative, whereas no spacer was used for N-terminal click conjugations (Figure 2A). Bispecific conjugate D4 was prepared with a similar spacing to D3, but where a disulfide bond replaced the triazole bond through synthesis of Pip6a having a C-terminal X-Cys-B extension ( Figure 2B). The assembly of these bi-specific conjugates required synthesis of three different derivatives of Pip6a, which were synthesized on solid phase using Fmoc peptide chemistry and purified by RP-HPLC to >95% purity in yields of 50-60%.
For each construct, conjugation of the first PMO to each of the three Pip6a derivatives was carried out through amide bond formation between the C-terminal carboxylic acid group of the peptide to the secondary amine at the 3end of the PMO, similarly to the synthesis of Pip6a-PMO (Dmd) (18) (Figure 2A and B). The conjugations were carried out in solution and purifications were carried out by ion exchange HPLC. Isolated yields of P-PMO were 35-40% (based on the amount of starting PMO). The 3 -azido PMO was coupled to the alkyne-P-PMO using copper (I)mediated alkyne-azide click chemistry resulting in a yield of 42% for conjugate D1 and 38% for conjugates D2 and D3 (Figure 2A). The click reaction for syntheses of D2 and D3 was sluggish at room temperature and after 6 h only a small amount of bi-specific conjugate was formed ( Figure 3A). However, heating the reaction mixture to 40 • C significantly improved the reaction rate and after 30 min the reaction had proceeded to near completion as determined by ion exchange HPLC ( Figure 3B). The 3 -Npys PMO (Dmd) was conjugated to P-PMO (Acvr2b) to give a disulfide bond in bi-specific conjugate D4 ( Figure 2B) in a yield of 58%. Mass   Table 1.

Evaluation of dystrophin and activin exon-skipping activities of bi-specific P-PMOs
The efficacies of the bi-specific P-PMO constructs were assessed in an initial screening step by RT-PCR using Dmd exon 23 skipping in H2K mdx cells ( Figure 4A). For all conjugates tested, high levels of exon 23 skipping were found in a dose-dependent manner (with some exons 22-23 double skipping which is frequently observed in this test system). The bi-specific conjugates D2 and D3, where both PMOs are conjugated to the C-terminus of Pip6a, exhibited better exon-skipping activity than for D1 and D4. Since bispecific conjugate D4 was less effective than D3, this suggests that there is no advantage to use of a cleavable disulfide linkage over use of stable click chemistry for addition of the second PMO. The control singly conjugated Pip6a-PMO (Dmd) demonstrated only slightly higher exon skipping compared to conjugates D2 and D3.
Bi-specific conjugates D2 and D3, which showed the best Dmd exon skipping levels, were examined further for their exon-skipping efficacies in Acvr2b ( Figure 4B). The results for Acvr2b mirrored that of Dmd targeting. Only very slightly higher exon skipping was observed with the control singly conjugated Pip6a-PMO compared to bi-specific counterparts D2 and D3. In addition, the general level of exon 5 skipping in the Acvr2b gene was found to be lower than that for Dmd. This is potentially because disruptive exon skipping by oligonucleotides for this gene is harder to achieve than skipping of Dmd exon 23 and has not been fully optimized as yet. Bi-specific conjugate D3 was marginally more efficient than for D2.

Evaluation of dystrophin and activin exon-skipping activity in mdx mice
Since both D2 and D3 conjugates demonstrated skipping activity of both genes in cells and had the highest levels of Dmd exon 23 skipping, they were further evaluated in vivo in the mdx mouse model of DMD. Intramuscular administration of 0.5 nmol of either the D2, D3 bi-specific conjugates was carried out into the TA muscle of mdx mice and compared with a 1:1 molar cocktail of singly conjugated Pip6a-PMOs. Analysis of splice-switching activity was carried out 2 weeks post-administration. Each of the CPP-PMO conjugates demonstrated robust splice-switching activity, with higher splice-switching activity evident for Dmd targeting compared to Acvr2b, as was also seen in the in vitro cell culture studies. Since no significant differences between the constructs could be seen following gel analysis by RT-PCR in the case of Dmd gene targeting ( Figure 5A), quantitative analysis of splice switching was carried out using qPCR primers to determine the reduction in the level of transcripts containing the target exons. In each case 35-40% of exon 23 skipped Dmd transcripts were found, when normalized to Dmd transcripts from non-injected control muscle, with no statistically significant differences seen between P-PMO treated mice for both singly conjugated and bi-specific conjugates. Unsurprisingly, the pattern of exon skipping was also maintained for Acvr2b gene targeting, where there were no significant differences seen between D2 or D3 conjugates or the singly conjugated Pip6a-PMO counterpart ( Figure 5B).

Cell viability
Cell and in vivo toxicities of P-PMOs are known to be predominantly a function of the peptide component and are dose-dependent (39). Therefore, the cell viability of the D2 bi-specific conjugate was assessed in human hepatocytes (Huh7) cells and compared to that of Pip6a-PMO (Dmd) and a mixture of the two Pip6a-PMOs against the two different targets Dmd and Acvr2b, in each case using a high equimolar concentration based on total PMO. Thus 20 M of bi-specific conjugate D2 was compared with a mixture of 10 M each of Pip6a-PMO (Dmd) and Pip6a-PMO (Acvr2b) and the percentage of cell survival measured (Figure 6). These results showed significantly higher cell viabil-  ity for the bi-specific D2 conjugate compared to a mixture of both Pip6a-PMOs for the two individual targets.

DISCUSSION
The promise of SSOs as therapeutic agents is being realized, with a number of clinical trials for DMD in progress to assess the efficacy of targeting a single exon (exon 51) to by-pass disease causing mutations (8)(9)(10). Further clinical trials are also being undertaken to target other exons, notably in the region covering exons 45-55 (40). Multiple simultaneous exon skipping using a cocktail of SSOs has been suggested as an approach to target a majority of patients (40,41). A proof-of-concept for multiple exon skipping was demonstrated in the golden retriever model of DMD (42) as well as in human patient cells lines (43). The concept was further extended to the targeting of exons 45-55 using cocktails of various SSOs (44). Unsurprisingly, the levels of exon skipping found to date have been low. In an attempt to improve this efficacy, studies were undertaken in the mdx52 mouse model using a cocktail of 10 different 'vivo-morpholino' PMOs to delete the entire stretch of exons 45-55 (45). Although a skipped transcript could be detected in these mice, the significant likely toxicity of delivering into mice 10 individual PMOs that are functionalized with guanidinium groups might hinder their clinical development.
Instead of the use of a cocktail of P-PMOs and with the need in mind to minimize toxicity, we sought to develop proof-of-concept for use of bi-specific PMO SSOs that could simultaneously target two different exons in different genes rather than in the same gene. We chose for these initial studies simultaneous targeting of exon 23 of the Dmd gene and exon 5 of the Acvr2b gene, using a single CPP as PMO delivery agent and to observe whether exon skipping could be maintained for both targets. Dumonceaux et al. showed that a combination of restoration of dystrophin and simultaneously inhibition of the myostatinsignalling pathway results in a significant improvement in muscle growth and force in dystrophic mdx mice (25). In a similar study, myostatin knockdown in conjunction with dystrophin restoration using an exon-skipping approach using a cocktail of two separate Pip6a-PMOs resulted in significantly increased mouse muscle mass (32). Thus targeting two genes in this way might be expected to have clinical relevance.
Nucleic Acids Research, 2015, Vol. 43, No. 1 37 We used our well-characterized Arg-rich CPP Pip6a as the model CPP because of the known high level of exon skipping observed for PMO conjugates in the DMD model (18) and we designed orthogonal conjugation chemistries using amide, disulfide and triazole bonds. The first conjugation of PMO to Pip6a was carried out in all cases through formation of a stable amide bond between the 3 -secondary amine of the PMO to the C-terminal carboxyl group of a synthetic Pip6a derivative. The second PMO conjugation was then effected either using an alkyne group on the Pip6a to an azide-functionalized PMO to give a stable triazole linkage or with a Cys residue on the Pip6a to an activated thiol group on the PMO to form a reversible disulfide linkage (Figures 1 and 2). Bi-specific conjugate D1 has one PMO at each Pip6a terminus whilst D2 and D3 have both PMOs at the C-terminus of Pip6a. The more sluggish click conjugation of the second PMO to the C-terminus of Pip6a in D2 and D3 at room temperature was presumably due to poorer accessibility than in the case of D1, but heating to 40 • C readily facilitated triazole bond formation (Figure 3). Conjugation of the second PMO using a disulfide bond (D4) did not require heating.
In mdx muscle cells, the lowest exon-skipping efficacy was observed for bi-specific conjugate D1 that had a PMO conjugated to each end of the Pip6a, which suggests that the ability of Pip6a to deliver PMO through the endosomal pathway and into the nucleus is inhibited by placing a bulky PMO at its N-terminus. By contrast, when both PMOs were placed at the C-terminus of the Pip6a (D2 and D3), the Dmd skipping activity was restored to close to the level observed with Pip6a-PMO (Dmd) ( Figure 4A), which confirms that CPP delivery is optimal when the CPP N-terminus is not blocked by a bulky substituent. Both conjugates D2 and D3 also demonstrated high levels of Acvr2b exon skipping in cells and close to that of the single Pip6a-PMO (Acvr2b) ( Figure 4B). Further, both D2 and D3 conjugates showed exon skipping of both the Dmd and Acvr2b targets by intramuscular delivery ( Figure 5) at levels unaltered from that of individual Pip6a-PMOs. These results show conclusively that there is no sequestration of one PMO at its own target that prevents the action of the other PMO at its own target. Thus, there must be a sufficient on-off equilibrium established for each target release and for target accessibility. This is an important finding that validates the use of bi-specific P-PMO.
Interestingly bi-specific P-PMO D4, in which the PMO (Dmd) was conjugated through a reducible disulfide bond, was less active than stably conjugated D2 and D3 constructs ( Figure 4). One explanation for this could be due to partial reduction of the disulfide bond upon cell entry and liberation of free Dmd PMO from the bi-specific conjugate before the Pip6a can deliver it into the cell nucleus. However, note that no difference was found in splicing redirection in a HeLa705 cell model using either disulfide-linked or stably linked R 6 -Penetratin PNA (46). By contrast in a recent study, bi-specific 2 -O-methyl phosphorothioate SSOs targeting both the Mstn and Dmd genes and delivered using a cationic transfection reagent were more effective when linked through a cleavable disulfide linker than through a non-cleavable hydrocarbon linker (33). The differential effect of use of a cleavable disulfide linkage probably reflects a different cell and nuclear uptake mechanism for SSOs delivered by a transfection agent compared to covalent CPP delivery of uncharged PMO or PNA. Detailed mechanistic studies would be needed to confirm this, but it should be noted that uptake of Pip6a-PMO (Dmd) into mdx skeletal muscle cells has been shown recently to be predominantly caveolae-mediated, whereas in cardiomyocytes uptake was mostly clathrin-mediated (47), suggesting that uptake routes are also cell dependent in addition to being dependent on oligonucleotide type and delivery method.
For the future it should be noted that the efficiency of exon 5 skipping for PMO (Acvr2b) is not as high as for exon 23 skipping for PMO (Dmd). PMO (Acvr2b) will therefore need further optimization to improve efficiency on this target. Alternatively, one might prefer to use a combination of PMO (Dmd) and a PMO that targets myostatin directly (32) to make a bi-specific PMO rather than targeting its receptor, and thus this second PMO needs to be optimized before embarking on lengthy intravenous delivery studies in mdx mice where larger scale synthesis would be needed. Clearly such studies need to investigate physiology benefits in addition to exon skipping (32). However, the increase in cell viability in human hepatocyte cultures for bi-specific conjugate D2 compared to the same PMO concentration for a mixture of two Pip6a-PMOs ( Figure 6) is encouraging and if this lower toxicity is maintained in systemic delivery, the clinical importance could be significant when considering multiple targeting (whether targeting different exons in the same gene or in two different genes).
In summary, we have developed new synthetic methodology for conjugation of two PMO SSOs to a single Pip6a CPP and shown efficient targeting in cells and in vivo of two separate genes with retained potency for each SSO. This work should enable further systemic studies on multiple SSO targeting in both DMD and also potentially in other neuromuscular disease models.