Enhancement of SMN2 Exon 7 Inclusion by Antisense Oligonucleotides Targeting the Exon

Several strategies have been pursued to increase the extent of exon 7 inclusion during splicing of SMN2 (survival of motor neuron 2) transcripts, for eventual therapeutic use in spinal muscular atrophy (SMA), a genetic neuromuscular disease. Antisense oligonucleotides (ASOs) that target an exon or its flanking splice sites usually promote exon skipping. Here we systematically tested a large number of ASOs with a 2′-O-methoxy-ethyl ribose (MOE) backbone that hybridize to different positions of SMN2 exon 7, and identified several that promote greater exon inclusion, others that promote exon skipping, and still others with complex effects on the accumulation of the two alternatively spliced products. This approach provides positional information about presumptive exonic elements or secondary structures with positive or negative effects on exon inclusion. The ASOs are effective not only in cell-free splicing assays, but also when transfected into cultured cells, where they affect splicing of endogenous SMN transcripts. The ASOs that promote exon 7 inclusion increase full-length SMN protein levels, demonstrating that they do not interfere with mRNA export or translation, despite hybridizing to an exon. Some of the ASOs we identified are sufficiently active to proceed with experiments in SMA mouse models.


Introduction
Spinal muscular atrophy (SMA), the most common genetic cause of infant mortality, is an autosomal recessive neuromuscular disease characterized by progressive loss of a-motor neurons in the anterior horns of the spinal cord, leading to limb and trunk paralysis and atrophy of voluntary muscles. Based on the severity and age of onset, SMA is clinically subdivided into types I, II, and III (MIMs 253300, 253550, and 253400), with type I being the most severe [1]. SMA has an incidence of approximately one in 6,000 live births, and a carrier frequency of one in 40. Loss of function of the survival of motor neuron 1 (SMN1) gene is responsible for SMA [2]. Mice with homozygous SMN1 disruption display massive cell death during early embryogenesis [3]. SMN protein is ubiquitously expressed and is mainly localized in the cytoplasm and in nuclear ''gems'' [4]. Multiple SMN-interacting partners have been identified, suggesting the involvement of SMN in various cellular processes, such as transcription, mRNA transport, and assembly of ribonucleoprotein particles (RNPs), including small nuclear RNPs (snRNPs), small nucleolar RNPs (snoRNPs), and stress granules [5][6][7][8].
Humans have an extra SMN gene copy, designated SMN2. Both SMN genes reside within a segmental duplication on Chromosome 5q13 as inverted repeats [2]. SMN1 and SMN2 are almost identical, except for 11 nucleotide substitutions: seven in intron 6, two in intron 7, one in coding exon 7 (a translationally silent C to T transition, relative to SMN1), and one in non-coding exon 8 [9]. SMN2 is a prototypical example of alternative splicing caused by a single nucleotide substitution in the affected exon. Exon 7 is efficiently included in spliced mRNA from the SMN1 gene; however, the silent C6T transition in SMN2 exon 7, which weakens the recognition of the upstream 39 splice site [10], results in significant skipping of this exon during pre-mRNA splicing. The C6T transition abrogates a splicing factor 2/alternative splicing factor (SF2/ ASF)-dependent exonic splicing enhancer (ESE); this SF2/ASF heptamer motif is present in SMN1 exon 7 (nucleotides þ6 to þ13, CAGACAA) and supersedes the inhibitory effect of heterogeneous nuclear ribonucleoprotein (hnRNP) A1 [11][12][13]. As a result, SMN2 encodes mostly the exon 7-skipped protein isoform (SMND7), which is unstable, mislocalized, and at best, only partially functional [14][15][16]. Although the small amount of full-length SMN protein derived from SMN2 is not sufficient to fully compensate for loss of SMN1, it is essential for viability in the absence of SMN1, and is an important disease modifier: in both SMA patients and mouse models, there is an inverse relationship between SMN2 copy number and disease severity [17,18]. These properties define SMN2 as an ideal therapeutic target for potential treatment of SMA.
Antisense technology was initially employed to downregulate gene expression by targeting mRNA to induce its degradation or block its translation [19]. Advances in antisense chemistry allowed this technology to be applied as a powerful tool to manipulate pre-mRNA splicing. Modification of the base, sugar, or phosphodiester structure of oligonucleotides generates highly stable molecules with not only high affinity for RNA targets, but also resistance to various nucleases, including RNase H (which cleaves RNA in RNA/DNA hybrids). The Kole laboratory pioneered the use of antisense oligonucleotides (ASOs) to correct aberrant splicing by targeting pre-mRNA, demonstrating both in vitro and in vivo that ASOs can restore correct expression of a defective b-globin gene by blocking the cryptic splice sites generated by intronic mutations that cause b-thalassemia [20,21]. Recently, ASO applications to redirect pre-mRNA splicing were extended to genes with therapeutic relevance to cancer or other diseases [22][23][24]. ASOs have been employed to promote skipping of non-essential exons, so as to restore the translational reading frame and suppress the effects of various deletion, nonsense, and frameshifting alleles of the dystrophin gene (DMD) [25][26][27]. Although it is now relatively straightforward to block the use of specific exons by targeting the corresponding 59 or 39 splice sites or empirically selected sequences within the exon, it remains much more challenging to find suitable antisense targets for promoting exon inclusion.
Several antisense approaches have been documented to promote SMN2 exon 7 splicing. One approach is based on the notion that there is a competition between the 39 splice sites of exons 7 and 8 for pairing with the 59 splice site of exon 6 [10], so impairing the recognition of the 39 splice site of exon 8 should favor exon 7 inclusion. Increased SMN2 exon 7 inclusion and full-length SMN protein were indeed observed in HeLa cells transfected with a modified U7 small nuclear RNA (snRNA) complementary to the intron 7/exon 8 junction [28]. Another approach, called TOES (targeted oligonucleotide enhancer of splicing) [29], relies on bifunctional ASOs composed of one segment complementary to exon 7 and a non-complementary tail consisting of RNA sequences with ESE motifs recognized by a serine/arginine-rich (SR) protein.
One such bifunctional ASO stimulated SMN2 exon 7 inclusion and led to significant restoration of gem number-an indicator of SMN protein increase-in SMA-patient fibroblasts [30]. Moreover, an in vivo delivery system was recently developed for expression of bifunctional ASOs in an adeno-associated virus vector [31].
Strong intronic splicing silencers (ISSs) represent ideal antisense targets; however, if present, they are usually inconspicuous within large introns. An ISS in intron 6 was identified in the context of an SMN2 minigene, and targeting this inhibitory element with an ASO increased exon 7 inclusion in transiently transfected COS-7 cells [32]. An intron 7 ISS, located immediately downstream of the 59 splice site of exon 7, was recently reported; an ASO against this silencer efficiently restored exon 7 inclusion of SMN2 and increased SMN protein levels in SMA-patient-derived cells [33].
The ESSENCE (exon-specific splicing enhancement by small chimeric effectors) method, which was developed to rescue disease-associated exon skipping or modulate alternative splicing, employs small chimeric effectors designed to emulate SR proteins [34,35]. A synthetic ESSENCE molecule comprises two portions: an antisense moiety complementary to a target exon; and a minimal RS domain peptide similar to the splicing activation domain of SR proteins. We described several ESSENCE compounds that significantly promote SMN2 exon 7 inclusion in a cell-free splicing assay [34]. Interestingly, the antisense moiety alone also significantly stimulates exon 7 inclusion, although with approximately 8fold less potency compared to ESSENCE compounds, suggesting the existence of an exonic splicing silencer (ESS) in exon 7 [34]. This observation highlighted the possibility of further optimization of the antisense moiety of ESSENCE compounds to enhance their potency. Here, we used two-step ASO walks along SMN2 exon 7 to define position-dependent effects of ASO-binding sites. With the first coarse whole-exon 7 walk, we identified two ASO target regions separated by a central core sequence, and with the second high-resolution walks, we identified optimal ASOs against each target region. The two best ASOs strongly promoted exon 7 inclusion during SMN2 pre-mRNA splicing in three different splicing assays, and increased SMN protein levels in cultured cells, including SMA type I fibroblasts. Thus, appropriate ASOs targeting a coding exon can increase protein levels by promoting exon inclusion and not interfering with translation. Most importantly, the effective ASOs should be suitable for studies in animals and to further optimize bifunctional or peptide-linked ASOs for better efficacy in rescuing SMN2 splicing and full-length protein expression.

ASO Walk along SMN2 Exon 7
Previously we reported that one PNA (peptide nucleic acid) ASO complementary to positions þ7 to þ18 of SMN2 exon 7 stimulates exon 7 inclusion during splicing in vitro [34]. This unexpected observation suggested that the ASO blocked one or more putative ESSs-either a specific motif or a secondary structure that impairs the recognition of exon 7. To pinpoint the exact locations of these presumptive elements, we used a systematic ASO walk along the entire length of SMN2 exon 7. Oligonucleotides with a 29-O-methoxy-ethyl ribose (MOE) phosphodiester backbone were used. This backbone modification imparts a very high affinity for targeted mRNA, resistance to both exo-and endonucleases, and does not support cleavage of hybridized mRNA by RNase H [36,37]. We synthesized nine MOE 15-mer ASOs complementary to the 54nucleotide (nt) exon 7; beginning with the first position of the exon, the overlapping ASOs provide coverage in 5-nt increments from 59 to 39, with a last step of 4 nt to the last position

Author Summary
Spinal muscular atrophy (SMA) is a severe genetic disease that causes motor-neuron degeneration. SMA patients lack a functional SMN1 (survival of motor neuron 1) gene, but they possess an intact SMN2 gene, which though nearly identical to SMN1, is only partially functional. The defect in SMN2 gene expression is at the level of pre-mRNA splicing (skipping of exon 7), and the presence of this gene in all SMA patients makes it an attractive target for potential therapy.
Here we have surveyed a large number of antisense oligonucleotides (ASOs) that are complementary to different regions of exon 7 in the SMN2 mRNA. A few of these ASOs are able to correct the pre-mRNA splicing defect, presumably because they bind to regions of exon 7 that form RNA structures, or provide protein-binding sites, that normally weaken the recognition of this exon by the splicing machinery in the cell nucleus. We describe optimal ASOs that promote correct expression of SMN2 mRNA and, therefore, normal SMN protein, in cultured cells from SMA patients. These ASOs can now be tested in mouse models of SMA, and may be useful for SMA therapy.
We first evaluated these ASOs by in vitro splicing in HeLa cell nuclear extract with a radiolabeled SMN2 minigene transcript [11]. Although the precise extent of exon 7 inclusion varies between extracts or depending on the precise reaction conditions, the difference between SMN1 and SMN2 is highly reproducible. We tested eight different concentrations of each ASO, from 1 to 400 nM, compared to no ASO. As shown in Figure 1B, the nine ASOs had different effects on splicing of the SMN2 pre-mRNA. Four ASOs, 01-15, 16-30, 21-35, and 26-40, strongly inhibited exon 7 inclusion; ASOs 11-25 and 31-45 slightly inhibited exon 7 inclusion; and ASO 40-54 and the control oligonucleotide 00-00 had no effect on alternative splicing of SMN2 exon 7. Interestingly, we found two ASOs, 06-20 and 36-50, that significantly stimulated exon 7 inclusion in the cell-free splicing assay, starting at a concentration of approximately 50 nM.
ASO 01-15 targets the exonic portion of the 39 splice site, whereas ASOs 11-25, 16-30, 21-35, and 26-40 target a central region that comprises a transformer 2 beta 1 protein (Tra2b1)-binding motif in SMN exon 7 [38]. Both elements are required for exon 7 inclusion, which can account for the negative effects of these ASOs. ASO 31-45 has a 10-nt overlap with the stimulatory ASO 36-50, yet it had a negative effect on splicing of exon 7, suggesting that there is a motif upstream of the ASO 36-50-binding site that is also important for exon 7 inclusion. ASO 40-54 targets the exonic portion of the exon 7 59 splice site; however, it is essentially neutral. We offer two explanations for this observation: first, the exonic portion of this 59 splice site deviates from the consensus sequence, and thus ASO 40-54 might not affect U1 snRNA base-pairing; second, ASO 40-54 has a strong predicted stem-loop structure (unpublished data) and might not bind efficiently to its complementary sequence. It is not surprising that ASO 06-20 had a positive effect, because it encompasses the sequence of the antisense PNA we reported previously [34]; however, ASO 36-50, which is complementary to nucleotides þ36 to þ50 of exon 7, defines an additional region in SMN2 exon 7 that can be targeted to increase exon 7 inclusion.
To determine whether the ASO-mediated stimulation of exon 7 inclusion can take place in cells, we co-transfected plasmid pCI-SMN2 with each of the nine MOE ASOs into HEK293 cells by electroporation. At the same time, we cotransfected plasmid pBabe Puro and selected the transfected cells by treatment with puromycin. Two days after transfection, we analyzed the transiently expressed RNA by RT-PCR ( Figure 2A). Each of the ASOs gave similar effects on splicing of exon 7 in the minigene pre-mRNA as observed by in vitro splicing ( Figure 1B), except for two ASOs, 11-25 and 31-45. These two ASOs, especially the latter one, resulted in robust inhibition of exon 7 inclusion in vivo, whereas only slight effects were observed in vitro.
We next tested the effect of the exon 7 ASOs on splicing of transcripts from the endogenous SMN genes in HEK293 cells. This time, only the ASOs and plasmid pBabe Puro were electroporated into the cells. The transfected cells were puromycin-selected as above, and RNA samples were collected 2-d post-transfection. After RT-PCR, we digested the cDNA with DdeI to distinguish SMN2 mRNA from SMN1 mRNA [39,40]. As expected, the effects of these MOE ASOs on splicing of transcripts from the endogenous SMN2 gene were very similar to those we observed with the SMN2 minigene ( Figure 2B). The only notable difference between the two assays was in the relative efficacy of the two positive ASOs, 06-20 and 36-50. ASO 06-20 was less efficient in stimulating exon 7 inclusion than ASO 36-50 in the in vitro splicing and minigene splicing assays, but it was reproducibly more efficient than ASO 36-50 in the endogenous SMN2 splicing assay.
On the basis of the above results obtained with the three splicing assays, we conclude that exon 7 comprises three distinct regions: a core sequence that is essential for exon 7 inclusion, and two flanking regions that correspond to putative ESSs. These two inhibitory regions correspond to the binding sites for ASOs 06-20 and 36-50, and are designated region A and region B, respectively.

High-Resolution Microwalks within Regions A and B
Having roughly delineated regions A and B as effective antisense targets to stimulate exon 7 inclusion, we sought to define their boundaries more precisely by identifying ASOs with optimal sequences and lengths, and thus maximize exon 7 inclusion. To this end, we designed 39 new ASOs: 17 ASOs of length 12, 15, or 16 nt, spanning 21 nt within or overlapping region A (microwalk A), and 22 ASOs of length 12, 15, or 18 nt, spanning 22 nt within or overlapping region B (microwalk B) ( Figure 3A). The sequences of all 39 ASOs are shown in Table 1. We first screened all these MOE ASOs using the in vitro splicing assay and a concentration of 100 nM for each ASO.
The 15-mer ASO 07-21 was identified as the most effective one in the case of microwalk A ( Figure 3B). Its target site is  The effects of the two groups of microwalk ASOs were further examined in vivo, first with the minigene pCI-SMN2 and then with the endogenous SMN1/2 genes, as described above. Each ASO, at a concentration of 10 lM, with or without the SMN2 minigene plasmid, was electroporated into HEK293 cells ( Figure 5A). ASO 07-21 stimulated exon 7 inclusion during splicing of the SMN2 minigene pre-mRNA more effectively than ASO 06-20. As expected, ASO 07-21 also potently stimulated exon 7 inclusion during splicing of the endogenous SMN2 pre-mRNA ( Figure 5B). Similarly, ASO 34-48 robustly increased the inclusion of exon 7 for both the SMN2 minigene and the endogenous SMN2 gene transcripts ( Figure 5C and 5D). The effects of all the antisense molecules on exon 7 splicing were generally consistent between in vivo minigene and endogenous gene assays. Compared to the in vitro splicing assay, the two in vivo splicing assays appeared more sensitive in terms of the stimulatory or inhibitory effects on exon 7 inclusion caused by the tested ASOs. In addition, some ASOs gave inconsistent effects between the in vitro assay and the two in vivo assays: three ASOs in microwalk A (08-22, 02-16, and 09-20) and three ASOs in microwalk B (33-47, 34-45, and 33-44) all slightly promoted exon 7 inclusion in vitro, but more or less inhibited exon 7 inclusion in vivo; ASOs 12-23, 35-52, 34-51, 33-50, and 32-49 showed neutral or negligible effects on exon 7 inclusion in vitro, but inhibited exon 7 inclusion in vivo to some extent; and finally, ASO 38-49 showed no effect on exon 7 inclusion in vitro or with the endogenous gene, but promoted exon 7 inclusion in the minigene co-transfections.
The microwalk data obtained from the two in vivo assays not only verified the presence of a central core sequence, but also established the precise boundaries between the core sequence and regions A and B on either side of it. In microwalk A, the target site for the improved ASO 07-21 extends to three consecutive As of the proposed Tra2b1-recognition sequence (AAAGAAGGA) [38], suggesting that this portion of the Tra2b1-dependent ESE is not critical for binding of the protein. This interpretation is consistent with the finding that mutating the triple As to triple Us does not abrogate Tra2b1 recognition [38]. The triple As mark the 39 boundary of region A; extending the walk farther downstream causes strong exon 7 skipping, as in the cases of ASOs 08-22, 12-23, and 11-22 ( Figure 5A and 5B), presumably because of interference with Tra2b1 binding. In microwalk B, ASOs that bind farther upstream of the improved ASO 34-48, such as 33-47, 32-46, and 33-44, caused a strong reduction in exon 7 inclusion ( Figure 5C and 5D). These three ASOs are complementary to exon 7 with a 59-most boundary at þ33C and þ32G, respectively, suggesting that this GC dinucleotide is part of an element or structure important for exon 7 recognition. This observation is also consistent with the in vivo effect of ASO 31-45 in the initial exon 7 walk (Figure 2). The combined data from the coarse-and high-resolution antisense walks point to the existence of a core sequence in the middle of exon 7, from þ22 to þ33, that is essential for exon 7 inclusion.
The microwalk data obtained from the two in vivo assays are also useful for mapping both the 59 boundary of region A and the 39 boundary of region B. In microwalk A, ASO 04-18 displayed the least stimulatory effects among the 15-mer ASOs; other ASOs that target further upstream sequences were either neutral or inhibitory. These results define the boundaries of region A as þ4 to þ21. In microwalk B, the 15mer ASOs 39-53 and 38-52, whose 39-most complementarysequence boundaries are þ53G and þ52G, respectively, were essentially neutral; all other 15-mer ASOs, except 33-47 and 32-46, which partly target the central core sequence, had stimulatory effects. Therefore, nucleotides þ34 to þ51 define the boundaries of region B. The results obtained with 12-mer ASOs are generally consistent with those obtained with 15mer ASOs, except for ASO 04-15, which targets region A, but was unexpectedly inhibitory with respect to exon 7 inclusion, and ASO 09-20, which also targets region A, but was inhibitory in the in vivo assays. The two high-resolution walk exon 7 ASOs were further characterized with respect to their effects on splicing of the endogenous SMN1/2 genes in HEK293 cells. We first conducted dose-response experiments with each ASO, using a concentration range from 0.2 to 20 lM for electroporation. Although we do not know the intracellular concentration of each ASO after electroporation, inclusion of exon 7 in both SMN1 and SMN2 transcripts showed a clear dependence on the dose of each ASO ( Figure 6A and 6B). For both ASOs, a significant increase in exon 7 inclusion was already noticeable at a starting concentration of 0.2 lM; at 20 lM ASO, SMN2 exon 7 inclusion reached 87%-89%, compared to 86%-87% SMN1 and 50%-51% SMN2 exon 7 inclusion in the untreated cells. After treatment with ASO 07-21 or ASO 34-48, the exon 7 inclusion level of the SMN1 gene transcripts rose to 93%-95%, indicating that these two ASOs affect the Next, we analyzed the effects of the two ASOs over time, as a measure of their intracellular stability. We transfected HEK293 cells by electroporation with each ASO at 10 lM concentration, and total RNA samples from parallel transfections were collected daily at 1-5 d post-transfection. Robust effects were already apparent on the first day and reached a plateau by the second day ( Figure 6C and 6D). The strong effects persisted for at least 5 d, though a slight decline of exon 7 inclusion level was observed, which might partly reflect dilution of the ASOs as the cells divided.

Effect of ASO 07-21 and ASO 34-48 at the Protein Level
Both ASO 07-21 and ASO 34-48 strikingly increased the amount of full-length mRNA expressed from the endogenous SMN1/2 genes. However, because the ASOs might remain associated with exon 7 in the spliced mRNA, they could potentially interfere with mRNA export and/or translation. Therefore, we sought to determine whether the stimulatory effect of these ASOs on SMN1 exon 7 inclusion results in increased full-length SMN protein. We could not accurately determine this in HEK293 cells, because of the high levels of identical SMN protein expressed from the SMN1 gene. Therefore, we generated another SMN2 minigene construct, designated pEGFP-SMN2. This minigene consists of an Nterminal enhanced green fluorescent protein (EGFP), a short cloning-site sequence, a hemagglutinin epitope tag (HA tag), SMN exon 6, a shortened intron 6 (same as in minigene pCI-SMN2), exon 7, intron 7, and the 59 end of exon 8 (75 nt) at the C-terminus ( Figure 7A). When the pre-mRNA from this minigene is spliced, it generates two mRNAs via exon (C) The proteins expressed from the reporters were detected by Western blotting with a monoclonal antibody against the HA tag; a-tubulin was detected with a monoclonal antibody as a loading control. Although exon-7-excluded mRNA expressed from the pEGFP-SMN2 plasmid was detected (B), the corresponding protein was apparently unstable (C). doi:10.1371/journal.pbio.0050073.g007 inclusion or skipping. Translation of the exon 7-included mRNA gives a larger protein of 308 amino acids (designated P-Incl), comprising EGFP, seven amino acids (SGLRSRE) derived from the cloning site, the HA tag, and exons 6 and 7 of SMN. The exon 7-skipped mRNA is translated into a smaller protein of 296 amino acids (designated P-Excl), comprising EGFP, SGLRSRE, the HA tag, exon 6, and four amino acids (EMLA) from exon 8 ( Figure 7A). Both proteins can be detected with anti-HA antibody. Before proceeding to the protein-level analysis, we first verified the increase in exon 7 inclusion for the minigene pEGFP-SMN2 after electroporation with ASO 07-21 or 34-48. Compared with the endogenous SMN2 gene, minigene pEGFP-SMN2 favors the exon 7-skipped mRNA isoform. The extent of exon 7 inclusion was about 28% without any treatment; however, after treatment with 10 lM ASO 07-21 or 34-48, exon 7 inclusion rose to 74% or 85%, respectively ( Figure 7B). Protein samples were generated 3 d after the cells were cotransfected via electroporation with the minigene construct and either ASO at 10 or 30 lM concentration. Surprisingly, using mouse monoclonal anti-HA antibody, we only detected the exon-included protein product, P-Incl, with an apparent mass of approximately 38 kDa ( Figure 7C). The effects of both ASOs were dose dependent, and after the treatment with 30 lM ASO, P-Incl reached almost the same level as observed with the SMN1 minigene.
We were not able to detect any P-Excl protein produced from the SMN2 D7 mRNA transcript; we believe this is due to protein instability. SMND7, the unstable protein isoform derived from the SMN1/2 genes, and P-Excl share the same Cterminus, corresponding to the peptide coded by exon 6 plus four amino acids derived from exon 8; it is possible that these sequences, when present close to the C-terminus, induce degradation of the protein. To overcome this problem, we constructed another SMN2 minigene, pEGFP-SMN2D6; this construct is similar to pEGFP-SMN2, but lacks most of exon 6, except for the last 9 nt, so as to maintain an intact natural 59 splice site ( Figure 7A). This minigene encodes two protein isoforms: the exon 7-included one (P-InclD6; 274 amino acids) and the exon 7-excluded one (P-ExclD6; 262 amino acids). Using this minigene, we again tested ASO 07-21 or 34-48 in HEK293 cells; significant increases in both exon 7included mRNA and protein were observed ( Figure 7B and 7C). This time, using anti-HA antibody, we detected two polypeptides from pEGFP-SMN2D6, with sizes of about 32 kDa and 30 kDa, and with the latter giving a stronger signal. After treatment with the ASOs, the P-ExclD6 band disappeared, whereas the P-InclD6 band became considerably stronger and comparable in intensity to the SMN1 minigene product. To control for sample loading differences, the same blots were reprobed with anti-a-tubulin monoclonal antibody. These results clearly demonstrate that ASOs targeting a coding exon are able to promote exon inclusion at the mRNA level, and the resulting mRNA can be translated into protein.

Effects of ASO 07-21 and ASO 34-48 in SMA-Patient Fibroblast Cells
To test whether the in vivo results in human HEK293 cells can be extended to SMA-patient cells, SMA type I 3813 fibroblasts were transfected with either ASO at 100 nM concentration using Lipofectin; this method of delivery was chosen because it was more efficient than electroporation with these primary fibroblast cells, consistently giving approximately 50% transfection efficiency with an EGFP plasmid (unpublished data). At 48 h after transfection, the cells were harvested to prepare mRNA and protein samples. The 3813 cells treated with the control ASO 00-00 or buffer alone gave 31% exon 7 inclusion. After treatment with ASO 07-21 or 34-48, the apparent level of exon 7 inclusion increased to 58% and 61%, respectively ( Figure 8A, left); based on the standard deviations (unpublished data; n ¼ 3), this corresponds to a 1.9 6 0.04 fold increase for ASO 07-21, and a 2.0 6 0.04 fold increase for ASO 34-48, relative to the treatment with the control ASO 00-00. The increase in the percentage of exon 7 inclusion could potentially reflect reduced exon skipping alone, without the desirable concomitant increase in exon inclusion. To rule out this possibility, we used GAPDH mRNA as a loading control, and calculated the ratio of SMN full-length mRNA to GAPDH mRNA, normalized to that observed with the control ASO 00-00 ( Figure 8A, right). The ratio increased by approximately 1.6fold in ASO 07-21-treated cells and approximately 1.8-fold in ASO 34-48-treated cells. To detect changes at the protein level, we used a-tubulin as a loading control. The normalized ratio significantly increased by approximately 1.8-fold and approximately 2.1-fold in ASO 07-21-and ASO 34-48treated samples, respectively, compared to ASO 00-00 ( Figure 8B). Assuming a delivery efficiency of approximately 50%, the actual effects should be approximately 2-fold greater in the cells that took up the ASOs.
Gems are nuclear bodies in which SMN protein accumulates [4]. A correlation between the number of gems in patient-derived fibroblasts and SMA clinical severity has been demonstrated [41]. To assess the effects of ASOs 07-21 and 34-48 on nuclear gem counts, we used indirect immunofluorescence with an anti-SMN monoclonal antibody, after transfection of 3813 cells with either ASO. The cells treated with the unrelated control oligonucleotide contained an average of 10 gems per 100 cells; however, when the cells were treated with ASO 07-21 or 34-48, the average gem number increased to 26 or 31 per 100 cells, respectively. The carrier 3814 fibroblast cells had approximately 50 gems per 100 cells ( Figure 8C-8E). In addition, after treatment with ASO 07-21 or 34-48, the number of cells containing multiple gems increased, and the number of cells containing no gems decreased. These results confirm that ASOs 07-21 and 34-48 significantly promote full-length SMN protein expression.

Discussion
Multiple strategies aimed at correcting exon 7 splicing of SMN2, a modifying gene for SMA, have been investigated. Mechanistic studies to understand how SMN2 exon 7 is alternatively spliced led to the discovery of several ciselements and trans-acting factors that can be targeted to stimulate exon 7 inclusion [42]. A number of drugs, including synthetic compounds that can modify SMN2 splicing, have been identified using cell-based high-throughput drug screening [42]. On the basis of the current knowledge of splicing mechanisms, we and others have employed antisensebased technologies to stimulate exon 7 inclusion. In recent years, a growing number of studies have demonstrated that MOE-modified ASOs (with a phosphodiester or a phosphorothioate backbone) and antisense PNAs can be valuable tools, Figure 8. Effect of ASOs 07-21 and 34-48 in SMA Type I Patient 3813 Fibroblasts Endogenous SMN2 mRNA, SMN protein levels, and gem number were analyzed after each ASO at a concentration of 100 nM was transfected into patient fibroblasts using Lipofectin. Two days later, RNA or protein samples were collected, or indirect immunofluorescence microscopy was carried out. Carrier 3814 fibroblasts were used as a positive control, and ASO 00-00 and buffer treatments were used as negative controls. (A) SMN2 mRNA was analyzed by radioactive RT-PCR, using GAPDH mRNA as a loading control. The percent of SMN2 exon 7 inclusion in each lane is shown below the autoradiogram. (B) SMN protein was detected by Western blotting, and the same blot was re-probed with antibody against a-tubulin as a protein-loading control. The histograms on the right of (A) or (B) show the corresponding quantitation from three independent experiments; the ratios are normalized to that observed with the buffer control, and error bars show the standard deviations. not only for dissecting gene function, but also for clinical applications [19]. Our previous work showed that ESSENCE compounds have the potential to treat diseases caused by exon skipping resulting from the loss of ESEs in mutant genes, including SMN2 [34]. A critical parameter in ESSENCE effectiveness is the selection of the optimal binding site along a target exon. In principle, the best target sequences are those that comprise negative exonic splicing signals or secondary structures, such as ESSs.
In this study, we systematically analyzed SMN2 exon 7 with a large number of MOE ASOs to identify putative ESSs. We employed a two-step ASO walk method, with an initial coarse ASO walk in 5-nt steps along the entire exon, followed by high-resolution single-nucleotide walks within the regions identified in the first step. With the first-step walk, we identified two ASO targets of potential therapeutic importance. The second-step microwalks optimized the ASOs and defined the apparent boundaries of the two ESS-containing regions. Using three independent splicing assays, we identified two potent ASOs that have significant therapeutic potential for SMA treatment. Our data also suggest that the two-step ASO walk is a powerful general method that can be used for screening inhibitory or stimulatory splicing regions present in the target exon(s) and the surrounding intron sequences of any gene.
Our data revealed an essential core sequence in the center of the 54-nt exon 7 from þ22 to þ33 (GAAGGAAGGTGC), which is surrounded by two separate inhibitory regions (A and B) containing negative splicing signals, with region A extending close to the upstream 39 splice site (þ4 to þ21) and region B extending close to the downstream 59 splice site (þ34 to þ51) (Figure 9). Blocking any part of the central core sequence promoted efficient skipping of exon 7, whereas blocking either of the two inhibitory regions promoted exon 7 inclusion. The 59 portion of the central core sequence comprises a previously identified Tra2b1-binding motif (GAAGGA) [38]; the 39 part (AGGTGC) may represent another cis-acting element that is also crucial for exon 7 recognition. A relatively long conserved sequence in the middle of exon 7, surrounded by two short inhibitory sequences, was previously identified by an iterative in vitro selection method [43]. Through an analysis of mutability, three segmental sequences with negative or positive cis-acting elements were inferred: the most conserved residues are located in the middle and form a conserved tract (þ16 to þ44), whereas the highly mutable upstream nucleotides form an inhibitory region (þ3 to þ15), and seven highly mutable downstream nucleotides (þ45 to þ51) form another negative element [43]. Our data sharpen the boundaries for these three segments, as verified by three different splicing assays and several ASOs, particularly the two most effective ASOs, 07-21 and 34-48. The target sequence of the 15-mer ASO 07-21 overlaps by six nucleotides with the previously reported conserved tract, and 11 out of 15 target nucleotides of ASO 34-48 are part of this conserved track. Nucleotides A36 and C37, which fall within the inhibitory region B, have been suggested to be part of a stimulatory motif, as the double mutation A36U/C37U abolishes exon 7 inclusion in the SMN1 gene [43]. However, the double mutation might create a stronger inhibitory motif, or strengthen a secondary RNA structure that impairs exon 7 recognition.
We searched exon 7 for putative ESSs with the Web servers PESX (http://cubweb.biology.columbia.edu/pesx) and ACES-CAN2 (http://genes.mit.edu/acescan2/index.html). PESX found no ESS motifs, whereas ACESCAN2 found three putative ESS motifs: GGTTTT (þ1 to þ6), TTTTAG (þ3 to þ8) and TTCCTT (þ39 to þ44). The first two motifs partially overlap with the inhibitory region A, and the last motif resides within the inhibitory region B, so these motifs might contribute to the inhibitory properties of regions A and B.
We note that both optimized ASOs are 15-mers, whereas all five of the 18-mer ASOs (36-53, 35-52, 34-51, 33-50, and 32-49) displayed more or less inhibitory effects on SMN2 exon 7 inclusion. The longer 18-mer ASOs are likely to overlap two or more binding sites, resulting in complex effects. The strong inhibitory effect of ASO 32-49 and the relatively weak inhibitory effect of ASO 33-50 might be explained by their interference with the positive regulation conferred by the essential core sequence, because the 59-most boundary nucleotide of their target sequence is 33G and 34C, respectively. The other three ASOs, whose target sequences lie next to the exon-intron junction, might interfere with recognition of the 59 splice site.
A small fraction of the ASOs had inconsistent effects on exon 7 splicing in the different splicing assays. One type of inconsistency occurred between the minigene splicing assays (both in vitro and in vivo) and the endogenous gene assay. The other type of inconsistency occurred between the in vitro splicing assay and the in vivo splicing assays. The first type of inconsistency may reflect the differences in the pre-mRNA substrates, although SMN2 exon 7 was identical in all the assays. In both the in vitro splicing assay and the in vivo minigene splicing assay, a small minigene was used. Higherorder structure differences between the small minigene pre-mRNA and the much larger endogenous pre-mRNA might affect the accessibility of some ASOs to their target sequences in exon 7, resulting in splicing differences. The second type of inconsistency might reflect differences in the splicingreaction environments between the cell-free splicing assay and the cell-based splicing assays, even though in general, the in vitro splicing reactions accurately reproduce cellular splicing events. For example, it is possible that the effects of some ASOs are more sensitive to changes in the concentration of certain splicing factors resulting from the nuclear extraction procedure. Another possibility is that the discrepancies result from kinetic differences and the coupling between transcription and mRNA processing that normally occurs in vivo, but not in standard in vitro splicing reactions [44]; moreover, transcription from different promoters-as in the minigene and the endogenous gene-can also affect use of alternative cassette exons in vivo [45].
Why ASOs 07-21 and 34-48 are more effective than the original ones, which bind to their target sequences only one or two nucleotides upstream or downstream, is currently under investigation. Because the ASOs have the same lengths and the same or similar G/C contents, annealing kinetics is probably not the cause, assuming equal accessibility to their target sites in exon 7. Two possible explanations of the effects of ASOs 07-21 and 34-48 are that they precisely block one or more splicing silencers, or efficiently disrupt an inhibitory RNA secondary structure. It has been reported that hnRNP A1 can inhibit SMN2 and SMN1 exon 7 inclusion [12,13]. In region A of SMN2, SF2/ASF presumably fails to block the cooperative propagation of hnRNP A1, due to the loss of an SF2/ASF motif caused by the C6T transition [13]. ASO 07-21 binds a 15-nt sequence that spans the mutant SF2/ASF motif; therefore, it is reasonable to assume that one of the roles played by ASO 07-21 is similar to a proposed function of SF2/ASF in SMN1 exon 7, i.e., blocking the propagation of hnRNP A1. ASO 07-21 might also block other inhibitory motifs residing in region A, for which an extended inhibitory context has been proposed [46]. We have unexpectedly observed in microwalk A that the 12-mer ASO 04-15 inhibited exon 7 inclusion; ASO 09-20 was likewise inhibitory, at least in vivo. These results suggest that the inhibitory effect of region A is the result of interplay among multiple trans-acting factors. It is likely that ASOs 04-15 and 09-20 can effectively block an undefined weak stimulatory signal, but cannot efficiently block the propagation of hnRNP A1 and possibly other repressors. A terminal-stem loop has been predicted in the less-studied region B, extending to two intron 7 nucleotides [47]. Though it is possible that the inhibitory property of region B is attributable to the proposed RNA secondary structure itself, other possibilities cannot be ruled out, including the existence of one or more splicing repressors that recognize a specific motif or RNA structure. Additional experimental studies will be required to elucidate the precise splicing-stimulatory mechanism of ASO 34-48.
We cannot exclude the possibility that the stimulatory ASOs act in part by stabilizing the pre-mRNAs and/or mRNAs to which they hybridize. Because only some of the exon 7 ASOs stimulate exon 7 inclusion, such a stabilization effect would presumably involve masking of specific instability elements. However, the effect of the ASOs is primarily, if not exclusively, at the level of splicing, because an effect at the level of mRNA half-life should result in the same proportional increase in SMN1 and SMN2 full-length mRNAs, yet the effects on SMN1 were much smaller, consistent with a switch in alternative splicing from the low basal level of exon 7skipped mRNA. In addition, we did not observe any ASOmediated stabilization of the labeled pre-mRNA in the in vitro splicing experiments.
We used three different assays to demonstrate that the ASOs that promote exon inclusion are compatible with synthesis of full-length SMN protein, indicating that they do not block mRNA export or translation. First, we showed that ASOs 07-21 and 34-48 promote efficient expression of reporter proteins corresponding to exon 7 inclusion in HEK293 cells. Second, we showed that the same ASOs promote expression of full-length SMN protein in primary fibroblasts from a type I SMA patient; though significant, this effect appears subtle because of the relatively low transfection efficiency with these cells. Finally, we showed that the same ASOs promote an increase in the number of nuclear gems detected with anti-SMN antibody in the patient fibroblasts; this single-cell immunofluorescence assay obviates the problem of transfection efficiency and provides an indirect readout for SMN protein levels [4,41].
Our initial goal in this study was to define optimal antisense targets to maximize ESSENCE effectiveness. However, our optimized antisense targets should also be useful with bifunctional ASOs [30,31]. The antisense portions of the bifunctional ASOs used in two earlier studies were not optimized. The first study used a 15-mer ASO targeting the exon 7 sequence þ2 to þ16 [30], i.e., the same target sequence as that of our ASO 02-16. However, we found that this ASO had a slightly negative effect on exon 7 inclusion in vivo with both the SMN2 minigene and the endogenous SMN2 gene, though it had a slightly stimulatory effect on splicing in vitro ( Figures 3B, 5A, and 5B). ASO 02-16 was one of few ASOs that gave inconsistent effects among the different splicing assays we used. The second study used a 20-mer ASO targeting the exon 7 sequence þ6 to þ25 [31], which partially overlaps the central core sequence that is essential for exon 7 inclusion. On the basis of our results, we predict that more-effective bifunctional ASOs targeting exon 7 can be designed, using the ASO 07-21 or 34-48 sequences as the antisense moiety.
Although we expect that even more-potent effectors can be designed by attaching ASOs 07-21 or 34-48 to activationdomain peptides or SR protein-binding sites-as in the ESSENCE [34] and TOES [29] methods-these ASOs appear to be sufficiently effective on their own to be tested in animal models of SMA. ASOs with the same chemistry were recently demonstrated to be active modulators of splicing in mice [48]. More generally, the antisense walk strategy described here should be applicable to identifying exon-targeting ASOs that modulate alternative splicing of normal or mutant genes. At the same time, these ASOs can facilitate the identification of important cis-regulatory sequences that can be further investigated to elucidate splicing mechanisms, and to identify trans-acting factors that may be useful molecular targets for disease therapy.

Materials and Methods
Oligonucleotide synthesis. Synthesis and purification of chimeric 29-O-methoxyethyl-modified oligonucleotides with a phosphodiester backbone were performed using an Applied Biosystems 380B automated DNA synthesizer (Applied Biosystems, Foster City, California, United States) as described previously [49]. The oligonucleotides were dissolved in water. The sequences of all the oligonucleotides are indicated in Table I.
Constructs. SMN minigene constructs used for both the in vitro splicing assay and the in vivo minigene splicing assay in HEK293 cells were pCI-SMN1 and pCI-SMN2, which are derivatives of pCISMNxD6-WT and pCISMNxD6-C6T, respectively [11,50]. The SMN minigenes comprise the 111-nt-long exon 6, a 200-nt shortened intron 6, the 54-nt exon 7, the 444-nt intron 7, the first 75 nt of exon 8, and a consensus 59 splice site (CAGGTAAGTACTT) to promote exon definition in vitro. We amplified the minigenes with primer set T7-F2 (59-TAC TTA ATA CGA CTC ACT ATA GGC TAG CCT CG-39) and SMN8-75-59R-SalI-NotI (59-GAA GCG GCC GCG TCG ACA AGT ACT TAC CTG TAA CGC TTC-39), and inserted them into vector pCI-neo (Promega, Madison, Wisconsin, United States) using XhoI and NotI. Two restriction sites, SalI and NotI, were included in the reverse primer. Four minigene constructs were used for protein expression: constructs pEGFP-SMN2 and pEGFP-SMN1 have the same minigenes as in pCI-SMN2 and pCI-SMN1, but without the consensus 59 splice site. Minigene inserts were amplified with primer set Xho-HA-E6-F (59-GAT CTC GAG AGT ACC CAT ACG ACG TAC CAG ATT ACG CTA TAA TTC CCC CAC CAC CTC CC-39) and Pr-E8-75-BamH-R (59-CGG GAT CCT AAC GCT TCA CAT TCC AGA TC-39), digested with XhoI and BamHI and subcloned into the XhoI and BamHI sites in pEGFP-C1 (Clontech, Mountain View, California, United States). Constructs pEGFP-SMN2D6 and pEGFP-SMN1D6 are similar to pEGFP-SMN2 and pEGFP-SMN1, but most of exon 6 was deleted, with only the last nine nucleotides retained to keep the natural context of the 59 splice site of exon 6. Minigene inserts were amplified with Pr-HA-I6-Xho-F (59-ATC TCG AGA GTA CCC ATA CGA CGT ACC AGA TTA CGC TTA TTA TAT GGT AAG TAA TCA CTC-39) and Pr-E8-75-BamH-R, and were subcloned into XhoI and BamHI double-digested pEGFP-C1. One copy of the HA epitope tag sequence was included in primers Xho-HA-E6-F and Pr-HA-I6-Xho-F, so that all protein products expressed from the four EGFP-fused minigene constructs have an N-terminal HA epitope tag and can be detected with anti-HA antibody.
In vitro splicing. To generate in vitro splicing substrates, we linearized plasmids pCI-SMN1 and pCI-SMN2 with SalI, and transcribed them with T7 RNA polymerase (Promega) in the presence of a-32 P-UTP and 7Me GpppG cap analog [11,51]. Substrates were purified by denaturing polyacrylamide gels and spliced in HeLa cell nuclear extract, as described [52]. We incubated 8 fmol of transcript in 10-ll splicing reactions containing 3 ll of nuclear extract and 1.6 mM MgCl 2 . After incubation at 30 8C for 4 h, we extracted the RNA and analyzed it on 8% denaturing polyacrylamide gels, followed by autoradiography and phosphorimage analysis with an Image Reader FLA-5100 (FujiFilm Medical Systems, Stamford, Connecticut, United States). We calculated exon 7 inclusion as a percentage of the total amount of spliced mRNA, i.e., included mRNA 3 100/(included mRNA þ skipped mRNA). The signal intensity of each mRNA isoform band was normalized according to its U content.
Cell culture and transfection. HEK293 cells, SMA type I homozygous and carrier fibroblasts (3813 and 3814; Coriell Cell Repositories, Camden, New Jersey, United States) were cultured in Dulbecco's modified Eagle's medium (DMEM; Invitrogen, Carlsbad, California, United States) containing 10% (v/v) fetal bovine serum and antibiotics (100-U/ml penicillin G and 100-lg/ml streptomycin). For transfection of MOE oligonucleotides and plasmids into HEK293 cells, electroporation was used, followed by puromycin selection (Figures 2 and 5-7). Briefly, 90 ll of freshly-prepared cells (1 3 10 7 /ml) suspended in Opti-MEM medium was mixed with 10 ll of ASO/DNA mixture in a 1-mm cuvette. A total of 2.5 lg of the plasmid pBabe Puro [53] was included in the ASO/DNA mixture for selection. For the in vivo minigene splicing assay and the minigene protein assays, 5 lg of each minigene plasmid was used. For electroporation, 80 volts and 500 lF were applied with a Gene Pulser II apparatus (Bio-Rad, Hercules, California, United States). Shocked cells were then cultured in 60-mm dishes in normal medium. Untransfected cells were killed by treatment with 2-lg/ml puromycin for 20 h. For transfection of MOE oligonucleotides into 3813 cells, Lipofectin (Invitrogen) was used according to the manufacturer's instructions.
Plotting dose-response curves and kinetic curves. All curves were plotted with Prism software (Graphpad Software, http://www. graphpad.com). To compare the effects of ASOs 07-21, 34-48, 06-20, and 36-50 on exon 7 inclusion in vitro, sigmoidal curve-fitting was used, with X values plotted on a logarithmic scale. The EC50 value of each ASO was calculated by the software. Sigmoidal curves were also plotted for the effects of ASOs 07-21 and 34-48, respectively, on endogenous SMN genes in HEK293 cells. The time-course analysis of ASOs 07-21 and 34-48 in HEK293 cells was plotted using exponential-decay curve-fitting.
Immunofluorescence and gem counting. The 3813 and 3814 fibroblasts were plated on coverslips and grown in DMEM. After overnight growth, the cells were transfected with 100 nM oligonucleotides. At 48 h post-transfection, the cells were washed with phosphate-buffered saline (PBS) and fixed with 3.7% paraformaldehyde in PBS for 20 min. After washing with PBS, the cells were permeabilized in 0.1% Triton X-100 in PBS for 3 min, blocked for 30 min in blocking buffer (5% goat serum, 3% IgG-free bovine serum albumin [BSA]), and incubated for 3 h in blocking buffer containing mouse monoclonal anti-SMN antibody. The cells were then washed three times with PBS and incubated for 2 h in blocking buffer containing goat anti-mouse secondary antibody Alexa Fluor 594 (Molecular Probes/Invitrogen). The cells were again washed with PBS three more times, and coverslips were mounted with Prolong Gold mounting solution containing DAPI (Molecular Probes/Invitrogen) for nuclear staining. Cells were analyzed using an Axioplan 2i fluorescence microscope (Carl Zeiss, Thornwood, New York, United States) equipped with Chroma filters (Chroma Technology, Rockingham, Vermont, United States). OpenLab software (Improvision, http://www.improvision.com) was used to collect digital images from a CCD camera (Hamamatsu, Bridgewater, New Jersey, United States). To count gems, first a field was randomly selected using the DAPI channel, the cell number was determined based on the DAPI counterstaining, and then the gems were counted in the red channel. At least three different fields were counted, and each field contained more than 40 cells.