**3. Antisense oligonucleotide-mediated exon skipping**

### **3.1 Introduction antisense oligonucleotides**

Antisense oligonucleotides (AONs) are small synthetic pieces of DNA or RNA (15-30 bp), which are complementary to their target mRNA. Initially, DNA oligos were used for the specific knockdown of gene expression. These DNA oligos bind to the RNA to form DNA-RNA hybrids which activate RNase H. This enzyme cleaves the double-stranded mRNA, thereby preventing the translation into protein, thus decreasing protein expression. DNA oligos are fast degraded by endonucleases, therefore oligos with a phosphorothioate instead of a phosphodiester backbone (PS DNA oligos) were developed, which are more endonuclease-resistant. These led to very efficient expression knockdown of for example genes (UL36 or IL2) involved in CMV retinitis (85-95%) (Baker & Monia, 1999). In addition to activation of RNase H, AONs can also down regulate gene expression by inducing translational arrest through steric hindrance of ribosomal activity, interference with mRNA maturation by inhibiting splicing or destabilisation of pre-mRNA in the nucleus (Chan et al., 2006). Later, 2'O modified RNA oligos were developed, which have a higher affinity for mRNA and turned out not to induce RNase H-dependent cleavage (Sproat et al., 1989). The activation of RNase H is useful when down regulation of gene expression is required, but not when AONs are used for modulation of pre-mRNA splicing.

#### **3.1.1 Antisense-mediated exon skipping for DMD**

AON-mediated exon skipping for DMD is based on the reading frame rule (fig. 2), which underlies the phenotypic differences between DMD and BMD. Furthermore, in some DMD

The Golden retriever muscular dystrophy (GRMD) dog is a spontaneously occurring canine model for Duchenne muscular dystrophy. These dogs have a single base substitution in the 3' consensus splice site of intron 6, resulting in skipping of exon 7, thereby introducing a premature stop codon in exon 8. The course of the disease is more comparable to human patients than that of the *mdx* mouse. The dogs display rapid and fatal muscular dystrophy, characterised by muscle atrophy, myofibre degeneration, replacement by fibrotic and adipose tissue and cardiomyopathy (Sharp et al., 1992). Most affected animals die within a few years, mainly due to degeneration of the cardiac muscle (Howell et al., 1997). Although phenotypically the GRMD dog seems a better model for DMD, it shows a lot of interindividual variation in the severity of the pathology. Some animals die within days after birth, whereas others appear almost normal and live for years (Ambrosio et al., 2008). This makes the dogs less suitable for experimental use, due to standardisation problems.

Because of the large size of the golden retriever, the GRMD dog has been bred with a much smaller beagle to generate the canine X-linked muscular dystrophy (CXMDj) model. These dogs have a milder phenotype compared to GRMD dogs and therefore have a longer life

In addition to the above mentioned large phenotypical variation, experiments with dogs are very costly. Dogs have a long breeding time and the availability is low (a heterozygous breeding program is needed, due to the severity of the phenotype). Furthermore, for

Antisense oligonucleotides (AONs) are small synthetic pieces of DNA or RNA (15-30 bp), which are complementary to their target mRNA. Initially, DNA oligos were used for the specific knockdown of gene expression. These DNA oligos bind to the RNA to form DNA-RNA hybrids which activate RNase H. This enzyme cleaves the double-stranded mRNA, thereby preventing the translation into protein, thus decreasing protein expression. DNA oligos are fast degraded by endonucleases, therefore oligos with a phosphorothioate instead of a phosphodiester backbone (PS DNA oligos) were developed, which are more endonuclease-resistant. These led to very efficient expression knockdown of for example genes (UL36 or IL2) involved in CMV retinitis (85-95%) (Baker & Monia, 1999). In addition to activation of RNase H, AONs can also down regulate gene expression by inducing translational arrest through steric hindrance of ribosomal activity, interference with mRNA maturation by inhibiting splicing or destabilisation of pre-mRNA in the nucleus (Chan et al., 2006). Later, 2'O modified RNA oligos were developed, which have a higher affinity for mRNA and turned out not to induce RNase H-dependent cleavage (Sproat et al., 1989). The activation of RNase H is useful when down regulation of gene expression is required, but

AON-mediated exon skipping for DMD is based on the reading frame rule (fig. 2), which underlies the phenotypic differences between DMD and BMD. Furthermore, in some DMD

therapeutic studies the size of the dogs requires large amounts of compound.

**3. Antisense oligonucleotide-mediated exon skipping** 

not when AONs are used for modulation of pre-mRNA splicing.

**3.1.1 Antisense-mediated exon skipping for DMD** 

**3.1 Introduction antisense oligonucleotides** 

**2.3.2 Canine models for DMD** 

span (Shimatsu et al., 2003).

patients rare, dystrophin-positive (so-called "revertant" fibres) were found, which are the result of spontaneous exon skipping or secondary mutations restoring the reading frame in these fibres and allowing dystrophin production. Therefore it was hypothesised that using AONs to induces skipping of specific exons could lead to the restoration of the reading frame and thereby production of slightly shorter dystrophin proteins, as found in BMD and revertant fibres (fig. 3) (van Ommen et al., 2008). This approach is mutation-specific and a large variety in mutations exists among DMD patients. Fortunately, 2 "hotspots" (a major around exon 43 to 53 and a minor spanning exons 2 to 20) exist, comprising a large proportion of the mutations (Aartsma-Rus et al., 2006b). In this Chapter we will describe the development of this therapeutic approach. We are aware that many excellent papers about exon skipping for DMD exist. Due to space constraints it was not feasible to cover them all. For a recent overview see Aartsma-Rus, *RNA Biology* 2010 (Aartsma-Rus, 2010).

In DMD mutations in the *DMD* gene lead to a disruption of the open reading frame (in this example a deletion of exon 50), thereby preventing production of a functional dystrophin protein. Binding of an exon-specific AON (in this example against exon 51) hides the exon from the splicing machinery. The exon will be 'skipped' and not incorporated in the mRNA. Thereby the reading frame is restored and translation of a shorter, but still largely functional dystrophin protein can occur, which is similar to the proteins found in BMD.

Fig. 3. Antisense oligonucleotide-mediated exon skipping

#### **3.2 Backbone chemistries**

To prevent activation of RNase H the 2'-*O* position of the ribose was modified (2'-*O*-methyl (2OME) or 2'-*O*-methoxyethyl (2OMOE)). Furthermore, various chemical modifications (fig. 4) have been developed, which differ in sugar and backbone chemistry and have different

AON-Mediated Exon Skipping for Duchenne Muscular Dystrophy 63

PS DNA 2OMePS RNA PNA

Base

Base

2OMOEPS RNA

Base

O

O H

S O - O

Vivo-PMO

O

Phosphorothioate (PS) DNA, 2'-*O*-methyl phosphorothioate (2OMePS) RNA, 2'-*O*-methoxyethyl phosphorothioate (2OMOEPS) RNA, peptide nucleic acid (PNA), locked nucleic acid (LNA), ethylene bridged nucleic acid (ENA), phosphoroamidate morpholino (PMO) and dendrimeric octaguanidine

Target sites for exon skipping AONs are splice sites (SS), exonic splicing enhancer (ESE) sites or exon inclusion sequences (EIS). Splice sites are required for the correct identification of exons by the spliceosome, a catalytic complex that coordinates the splicing process and consists of 5 small nuclear ribonucleoproteins (snRNP) and hundreds of other splicing factors. The 5' (donor) splice site (beginning of an intron), the branch point (just upstream of the acceptor splice site) and the 3' (acceptor) splice site (end of an intron) contain consensus sequences that are bound by snRNPs and splicing factors to bring about the removal of introns and ligation of exons. Blockage of splice sites or the branch point prevents incorporation of the exon in the mRNA. Exon recognition is further facilitated by ESE sites, which are exonic sequence motives to which certain splicing factors (Ser-Arg-rich (SR) proteins) can bind. AONs targeting ESEs have been proposed to sterically hinder the binding of SR proteins (Aartsma-Rus et al., 2005; Aartsma-Rus et al., 2009b; Kole et al., 2004;

N

N

N

N

N

O

N

O

P O Base

CH3

N

Base O

N{(CH2)2OCN[(CH2)6NHCNH2]2}2

O NH

O NH

N{(CH2)2OCN[(CH2)6NHCNH2]2}2

O NH

LNA/ENA

**3.3 AON design and targets** 

Tanaka et al., 1994).

P O O O-

O

O H

O

O O

P O S O - Base

Base

O

O O

<sup>S</sup> <sup>O</sup> - CH3

O

N

O P O NO

P O

PMO

conjugated phosphoroamidate morpholino (Vivo-PMO).

Fig. 4. Chemical structure of different antisense oligonucleotides

biophysical, biochemical and biological properties. For a more detailed review see Chan et al., *Clin.Exp.Pharmacol.Physiol* 2006 (Chan et al., 2006). The 2OMePS chemistry has an increased affinity for RNA and nuclear uptake. Disadvantages are that the phosphorothioate backbone is toxic to some extent and some sequences elicit an immune response. This is partly counteracted by the 2OMe modification.

Peptide nucleic acids (PNA) contain a flexible, uncharged, achiral *N*-(2-aminoethyl)glycine backbone to which nucleobases are attached via methylenecarbonyl linkages in stead of the phosphodiester backbone of DNA oligos. PNAs have a high affinity for RNA, are not toxic even at high concentrations, are peptidase-and nuclease-resistant and have a high sequencespecificity. A disadvantage is the insolubility of PNAs, due to their hydrophobic nature, which makes transfection difficult. This can be solved by the attachment of carrier groups, which easily bind to the peptide backbone, or addition of cationic lysine residues to the Cterminus. Another disadvantage is the rapid clearance of PNAs *in vivo*. Their mechanism of action is mainly by steric hindrance (Larsen et al., 1999).

Locked nucleic acid (LNA) DNA oligos contain a 2'-*O*, 4'-*C*-methylene bridge in the β-Dribofuranosyl configuration. They have a high hybridisation affinity towards target mRNA or DNA, thereby forming stable duplexes. This is an advantage, but also a disadvantage, since LNAs longer than 15 base pairs show self-annealing and are not very sequencespecific, which increases the chance of unwanted side effects (Aartsma-Rus et al., 2004b). However, currently mainly LNA/2'-O-methyl oligonucleotide mixmers are used, which show much more sequence-specificity (Fabani & Gait, 2008). LNAs have a good nuclear uptake and are nuclease-resistant.

Ethylene bridged nucleic acids (ENA) contain an ethylene bridge between the 2'-*O* and the 4'-*O*-C of the ribose. They have similar properties to LNAs, but have a higher affinity to RNA, are very stable and more nuclease-resistant (Morita et al., 2002; Yagi et al., 2004).

Phosphoroamidate morpholino oligomers (PMO) have a six-membered morpholino ring instead of the ribose sugar and the phosphodiester bond is replaced by a phosphoroamidate linkage. They do not activate RNase H, are very resistant to nucleases and are non-toxic. Furthermore, they are uncharged, which prevents undesired binding to proteins. However, this also results in limited nuclear uptake, where pre-mRNA splicing takes place. Their neutral charge also makes them hard to transfect in cell cultures, but *in vivo* PMOs can be taken up by tissues after local injection. This is probably due to the fact that the neutral nature does not form interactions with other cellular components. In general, PMOs are often a bit longer than 2OMePS AONs (25 nucleotides or more compared to around 20 nucleotides for 2OMePSs). They primarily act by steric prevention of ribosomal assembly (Aartsma-Rus et al., 2004b; Chan et al., 2006; Heemskerk et al., 2009b). PMOs have been linked to arginine-rich cell-penetrating peptides (pPMOs) to increase uptake and efficiency. These conjugates indeed have higher efficacy, but there are toxicity concerns and the peptide might evoke an immune response (Moulton & Moulton, 2010), though this has not yet been observed. Conjugation of PMOs with a dendrimeric octaguanidine polymer (vivomorpholino) improves the delivery of the compound *in vivo*. Since this polymer is not a peptide, the risk of an immune response is small and has not been observed so far (Wu et al., 2009), though the polymer is toxic at higher concentrations as well.

biophysical, biochemical and biological properties. For a more detailed review see Chan et al., *Clin.Exp.Pharmacol.Physiol* 2006 (Chan et al., 2006). The 2OMePS chemistry has an increased affinity for RNA and nuclear uptake. Disadvantages are that the phosphorothioate backbone is toxic to some extent and some sequences elicit an immune response. This is

Peptide nucleic acids (PNA) contain a flexible, uncharged, achiral *N*-(2-aminoethyl)glycine backbone to which nucleobases are attached via methylenecarbonyl linkages in stead of the phosphodiester backbone of DNA oligos. PNAs have a high affinity for RNA, are not toxic even at high concentrations, are peptidase-and nuclease-resistant and have a high sequencespecificity. A disadvantage is the insolubility of PNAs, due to their hydrophobic nature, which makes transfection difficult. This can be solved by the attachment of carrier groups, which easily bind to the peptide backbone, or addition of cationic lysine residues to the Cterminus. Another disadvantage is the rapid clearance of PNAs *in vivo*. Their mechanism of

Locked nucleic acid (LNA) DNA oligos contain a 2'-*O*, 4'-*C*-methylene bridge in the β-Dribofuranosyl configuration. They have a high hybridisation affinity towards target mRNA or DNA, thereby forming stable duplexes. This is an advantage, but also a disadvantage, since LNAs longer than 15 base pairs show self-annealing and are not very sequencespecific, which increases the chance of unwanted side effects (Aartsma-Rus et al., 2004b). However, currently mainly LNA/2'-O-methyl oligonucleotide mixmers are used, which show much more sequence-specificity (Fabani & Gait, 2008). LNAs have a good nuclear

Ethylene bridged nucleic acids (ENA) contain an ethylene bridge between the 2'-*O* and the 4'-*O*-C of the ribose. They have similar properties to LNAs, but have a higher affinity to RNA, are very stable and more nuclease-resistant (Morita et al., 2002; Yagi et al., 2004).

Phosphoroamidate morpholino oligomers (PMO) have a six-membered morpholino ring instead of the ribose sugar and the phosphodiester bond is replaced by a phosphoroamidate linkage. They do not activate RNase H, are very resistant to nucleases and are non-toxic. Furthermore, they are uncharged, which prevents undesired binding to proteins. However, this also results in limited nuclear uptake, where pre-mRNA splicing takes place. Their neutral charge also makes them hard to transfect in cell cultures, but *in vivo* PMOs can be taken up by tissues after local injection. This is probably due to the fact that the neutral nature does not form interactions with other cellular components. In general, PMOs are often a bit longer than 2OMePS AONs (25 nucleotides or more compared to around 20 nucleotides for 2OMePSs). They primarily act by steric prevention of ribosomal assembly (Aartsma-Rus et al., 2004b; Chan et al., 2006; Heemskerk et al., 2009b). PMOs have been linked to arginine-rich cell-penetrating peptides (pPMOs) to increase uptake and efficiency. These conjugates indeed have higher efficacy, but there are toxicity concerns and the peptide might evoke an immune response (Moulton & Moulton, 2010), though this has not yet been observed. Conjugation of PMOs with a dendrimeric octaguanidine polymer (vivomorpholino) improves the delivery of the compound *in vivo*. Since this polymer is not a peptide, the risk of an immune response is small and has not been observed so far (Wu et al.,

2009), though the polymer is toxic at higher concentrations as well.

partly counteracted by the 2OMe modification.

action is mainly by steric hindrance (Larsen et al., 1999).

uptake and are nuclease-resistant.

Phosphorothioate (PS) DNA, 2'-*O*-methyl phosphorothioate (2OMePS) RNA, 2'-*O*-methoxyethyl phosphorothioate (2OMOEPS) RNA, peptide nucleic acid (PNA), locked nucleic acid (LNA), ethylene bridged nucleic acid (ENA), phosphoroamidate morpholino (PMO) and dendrimeric octaguanidine conjugated phosphoroamidate morpholino (Vivo-PMO).

Fig. 4. Chemical structure of different antisense oligonucleotides

#### **3.3 AON design and targets**

Target sites for exon skipping AONs are splice sites (SS), exonic splicing enhancer (ESE) sites or exon inclusion sequences (EIS). Splice sites are required for the correct identification of exons by the spliceosome, a catalytic complex that coordinates the splicing process and consists of 5 small nuclear ribonucleoproteins (snRNP) and hundreds of other splicing factors. The 5' (donor) splice site (beginning of an intron), the branch point (just upstream of the acceptor splice site) and the 3' (acceptor) splice site (end of an intron) contain consensus sequences that are bound by snRNPs and splicing factors to bring about the removal of introns and ligation of exons. Blockage of splice sites or the branch point prevents incorporation of the exon in the mRNA. Exon recognition is further facilitated by ESE sites, which are exonic sequence motives to which certain splicing factors (Ser-Arg-rich (SR) proteins) can bind. AONs targeting ESEs have been proposed to sterically hinder the binding of SR proteins (Aartsma-Rus et al., 2005; Aartsma-Rus et al., 2009b; Kole et al., 2004; Tanaka et al., 1994).

AON-Mediated Exon Skipping for Duchenne Muscular Dystrophy 65

would largely increase its therapeutic applicability for a number of different mutations. Indeed the same result could be induced in patient cells with an exon 48-50 deletion (Aartsma-Rus et al., 2004a). Skipping of other large stretches of exons (multiple exon skipping) however turned out to be technically challenging and has had limited success so far (Aartsma-Rus et al., 2006a). The use of several ratios of 45AON and 55AON in both control as patient cell lines resulted in undetectable to very low exon 45-55 skipping

Exon skipping is in theory useful for the majority of patients. Exceptions are mutations that involve regions in the gene that are essential for the function of the dystrophin protein: all actin-binding N-terminal parts, the cysteine-rich C-terminal part (binding to the DGCcomplex), the promoter region or the first exon. Furthermore it is not applicable to translocations. Fortunately these kind of mutations make up only a small part (~8%) of all known mutations (Aartsma-Rus et al., 2009a). The largest part of mutations is made up by deletions and small mutations. A minor part consists of exon duplications (double or multiple). In the case of single duplications, skipping of one of these exons would in theory generate wild-type dystrophin transcripts. However, this turned out to be challenging. In cells with an exon 45 duplication this was indeed possible, but in other cases the skipping was so efficient that both exons were skipped, leading to an out-of-frame transcript (Aartsma-Rus et al., 2007). Skipping of an additional exon could restore the reading frame again. For example for an exon 18 duplication, successful skipping of exon 17 and both exon 18s resulted in restoration of the reading frame (Forrest et al., 2010). Successful skipping of multiple exon duplications has not yet been achieved (Aartsma-Rus et al., 2007). In total 6%

of patients could benefit from single or multiple skipping of exon duplications.

After the promising *in vitro* results, AONs have been tested *in vivo* in animal models. As mentioned before, the *mdx* mouse is most widely used. The target site for exon 23 was first optimised in *mdx* myotube cell cultures. This resulted in a 5' splice site targeting AON with a 2'-*O*-methoxyethyl phosphorothioate (2OMePS) backbone, which was tested locally in the *mdx* mouse. A single intramuscular injection of this compound in the tibialis anterior of both young (2 or 4 weeks old) and aged (6 months old) mice resulted in marked dystrophin expression 2 weeks after injection, which persisted up to 3 months after injection. The functionality of the dystrophin protein was suggested by the re-expression of dystroglycans, sarcoglycans and nNOS at the membrane. It also resulted in partial restoration of physiological function, maximum isometric titanic force, of the treated muscles. Importantly no auto-immune response against the newly formed dystrophin protein was observed (Lu et al., 2003). Of course, since DMD affects body-wide musculature, including heart and diaphragm, injection of every muscle separately is not feasible and systemic treatment is required. Three intravenous injections at weekly intervals resulted in dystrophin expression, highest in gastrocnemius, intercostal muscles and the diaphragm, without signs of toxicity or damage to other organs. However dystrophin could not be detected in the cardiac muscle (Lu et al., 2005). To optimise delivery and efficiency, different administration routes have

**5. Antisense-mediated exon skipping** *in vivo* **in animal models** 

frequencies (van Vliet L. et al., 2008).

**5.1 AONs in mouse models for DMD 5.1.1 AONs in the** *mdx* **mouse model** 
