|Home | About | Journals | Submit | Contact Us | Français|
Spinal muscular atrophy (SMA) is an autosomal recessive neuromuscular disorder and is the leading genetic cause of infant mortality. SMA is caused by the loss of survival motor neuron-1 (SMN1). In humans, a nearly identical copy gene is present called SMN2, but this gene cannot compensate for the loss of SMN1 because of a single silent nucleotide difference in SMN2 exon 7. This single-nucleotide difference attenuates an exonic splice enhancer, resulting in the production of an alternatively spliced isoform lacking exon 7, which is essential for protein function. SMN2, however, is a critical disease modifier and is an outstanding target for therapeutic intervention because all SMA patients retain SMN2 and SMN2 maintains the same coding sequence as SMN1. Therefore, compounds or molecules that increase SMN2 exon 7 inclusion hold great promise for SMA therapeutics. Bifunctional RNAs have been previously used to increase SMN protein levels and derive their name from the presence of two domains: an antisense RNA sequence specific to the target RNA and an untethered RNA segment that serves as a binding platform for splicing factors. This study was designed to develop negatively acting bifunctional RNAs that recruit hnRNPA1 to exon 8 and block the general splicing machinery from the exon 8. By blocking the downstream splice site, this could competitively favor the inclusion of SMN exon 7 and therefore increase full-length SMN production. Here we identify a bifunctional RNA that stimulated full-length SMN expression in a variety of cell-based assays including SMA patient fibroblasts. Importantly, this molecule was also able to induce SMN expression in a previously described mouse model of SMA and demonstrates a novel therapeutic approach for SMA as well as a variety of diseases caused by a defect in splicing.
Spinal muscular atrophy (SMA) is an autosomal recessive neuromuscular disorder and is the leading genetic cause of infant mortality. SMA has a carrier frequency of nearly 1 in 35 and an incidence of 1 in 6000 live births (Pearn, 1980; Melki, 1997). SMA is caused by the loss of survival motor neuron-1 (SMN1). In humans, a nearly identical copy gene, called SMN2, is present, but this gene cannot compensate for the loss of SMN1 because of a single silent nucleotide difference in SMN2 exon 7 (Lefebvre et al., 1997; Lorson et al., 1999; Monani et al., 1999). This single-nucleotide difference attenuates an exonic splice enhancer and at the same time creates an exonic splice silencer, resulting in the production of an alternatively spliced isoform lacking exon 7 (Cartegni and Krainer, 2002; Kashima and Manley, 2003; Kashima et al., 2007). The majority of SMN2-derived transcripts lack exon 7, and these transcripts produce an unstable protein called SMNΔ7, which cannot provide protection from SMA development when SMN1 is not present (Lorson and Androphy, 2000).
SMN exon 7 is highly regulated by a complex cohort of positive and negative regulatory cis and trans factors. SF2/ASF, a serine–arginine (SR)-rich protein, has been shown to interact with an enhancer that overlaps the critical C/T transition in SMN exon 7, and the protein is unable to bind the region in the SMN2 context (Cartegni and Krainer, 2002). The C/T transition not only disrupts the SF2/ASF enhancer, but it creates a novel hnRNPA1-dependent splicing silencer (Kashima et al., 2007). Other SR or SR-like proteins such as hTra2β1, SRp30c, RBMY, and hnRNP-G have all been shown to associate either directly or indirectly with SMN exon 7 (Hofmann et al., 2000; Hofmann and Wirth, 2002; Young et al., 2002). There are also several cis-acting negative regulatory regions that have been identified that surround SMN exon 7, including a negative element upstream and several negative elements including the extended inhibitory context (Miyajima et al., 2002; Singh et al., 2004a–c, 2007).
Increasing expression from SMN2, which is retained in essentially all SMA patients, may be a viable treatment option because none of the nonpolymorphic nucleotide differences between SMN1 and SMN2 alters the coding sequence when properly spliced. Several strategies have been employed toward this end. Stimulating the SMN2 locus to increase total transcripts would increase SMNΔ7 and full-length transcripts. Increasing both transcripts would be beneficial from a therapeutic perspective because the SMNΔ7 protein product lessens the disease phenotype (Le et al., 2005). Pharmacologically increasing full-length SMN expression from SMN2 or altering SMN2 splicing to increase exon 7 inclusion has also been reported for a variety of compounds, including valproic acid, phenylbutyrate, polyphenol-based botanical compounds, hydroxyurea, suberoylanilide hydroxamic acid (SAHA), and quinazoline-based compounds (Andreassi et al., 2001, 2004; Chang et al., 2001; Lunn et al., 2004; Brahe et al., 2005; Grzeschik et al., 2005; Jarecki et al., 2005; Kernochan et al., 2005; Brichta et al., 2006; Avila et al., 2007; Sakla and Lorson, 2008; Thurmond et al., 2008). In addition, aminoglycosides, which can promote translational read-through of cellular stop codons, may prove beneficial for therapy by stabilization of the SMN2-derived SMN protein (Wolstencroft et al., 2005; Mattis et al., 2006).
Several nucleic-acid based SMA therapeutics have been developed that are designed to increase production of functional SMN protein by promoting inclusion of SMN2 exon 7 (Hua et al., 2007, 2008). Oligonucleotide therapies that have been designed to block the exon 8 splice site increase splicing to exon 7 (Lim and Hertel, 2001; Madocsai et al., 2005). Novel peptide–nucleic acids that comprise an exon 7 antisense domain and a synthetic splicing-activation domain peptide were shown to increase SMN2 exon 7 inclusion (Cartegni and Krainer, 2003). Similar methods using bifunctional RNAs have also been previously shown to increase full-length SMN transcripts as well as SMN protein levels and derive their name from the presence of two domains: an antisense RNA sequence specific to the target RNA and an untethered RNA segment that serves as a binding platform for splicing factors (Skordis et al., 2003; Baughan et al., 2006). In other genetic models, negatively acting bifunctional RNAs that recruit negatively acting proteins to an intron/exon boundary successfully redirected pre-mRNA splice site decisions (Villemaire et al., 2003; Gendron et al., 2006).
This study aimed to discover whether a negatively acting bifunctional RNA, designed to recruit hnRNPA1 to the exon 8 splice site, could block the general splicing machinery from the exon 8 boundary, thereby competitively favoring the slightly weaker SMN exon 7 splice site. A bifunctional RNA was identified that specifically bound the SMN exon 8 splice boundary and stimulated full-length SMN protein in a variety of cell-based assays, including SMA patient fibroblasts. Furthermore, SMN protein levels were elevated in vivo after administration of the bifunctional RNA in a mouse model of SMA.
Bifunctional clones were generated with annealed complementary pairs of DNA oligonucleotides (Integrated DNA Technologies, Coralville, IA) that were cloned into the previously published pMU2 vector between the BamHI and SpeI sites (Baughan et al., 2006). A KpnI site was engineered as the spacer region between the antisense and tandem repeats of the splicing factor, targeting for ease of transition between different antisense and targeting domains. Sequences used for the exon8-hnRNPA1 bifunctional RNA are as follows: top, 5′-GAT CCA GCA TTT CCT GCA AAT GAG GGT ACC TAG GGA TAG GGA TAG GGA TTT TTT A-3′; bottom, 5′-CTA GTA AAA AAT CCC TAT CCC TAT CCC TAG GTA CCC TCA TTT GCA GGA AAT GCT G-3′. Plasmids were sequenced for verification. The SMN intron 7/exon 8 sequence targeted by the antisense RNA within the bifunctional RNA and the antisense sequence used previously (Madocsai et al., 2005) is as follows: 5′-tttctcatttgcaggaattctggcatagag-3′.
Subconfluent HeLa cells were plated in 12-well dishes and cotransfected with a previously described SMN2-luciferase plasmid (Zhang et al., 2001) and the exon8-hnRNPA1 bifunctional plasmid, using the Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA). pCI-SMN1-Luc and pCI-SMN2-Luc were transfected at 1μg and, to equalize DNA content, were also transfected with 3μg of sheared salmon sperm DNA. The experimental cells were transfected with 1μg of pCI-SMN2-Luc and 3μg of pMU2-exon8-hnRNPA1. Transfected cells were collected with a luciferase assay system (Promega, Madison, WI) and luminescence was measured with a luminometer.
Sequences of the 2′-O-methyl RNAs (Integrated DNA Technologies) are as follows: Exon8-hnRNPA1, 5′-CCA GCA UUU CCU GCA AAU GAG GGU ACC UAG GGA UAG GGA UAG GGA-3′; G2, 5′-CCU UUA CGA CCG UAU CUC-3′; D2-2, 5′-AAG AUU AAA GAG UAA ACG UCC-3′; D2-2 hnRNPA1, 5′-AAG AUU AAA GAG UAA ACG UCC GGU ACC UAG GGA UAG GGA UAG GGA-3′; and hnRNPA1-act, 5′-UAG GGA UAG GGA UAG GGA-3′. The D2-2 sequence was based on a previously published sequence that was unable to increase splicing to exon 7 (Madocsai et al., 2005). Subconfluent type I patient fibroblast cells (GM03813 cells; Coriell Cell Repositories, Camden, NJ) were transfected with 100ng of each indicated RNA, using Lipofectamine 2000 reagent (Invitrogen).
For all immunofluorescence staining, subconfluent GM03813 cells were transfected for 48hr with either exon8-hnRNPA1-pMU2 or 2′-O-methyl RNAs, using Lipofectamine 2000 reagent (Invitrogen). Transfected cells were fixed with an acetone–methanol (1:1) solution and washed with phosphate-buffered saline (PBS). Cells were blocked with 5% bovine serum albumin (BSA) in PBS and subsequently washed with PBS. A previously described pooled collection of mouse anti-SMN monoclonal antibodies (Wolstencroft et al., 2005) was added, diluted 1:10 in 1.5% BSA in PBS. After PBS washing, the secondary antibody, anti-mouse antibody conjugated to either Texas red (Jackson ImmunoResearch Laboratories, West Grove, PA) or fluorescein isothiocyanate (FITC) (Sigma-Aldrich, St. Louis, MO), was added diluted 1:200 in 1.5% BSA in PBS. After another wash, 4′,6-diamidino-2-phenylindole (DAPI) was added to each sample for 5min and the samples were washed again. Coverslips were then mounted to samples, using mounting medium (1,4-diazabicyclo[2.2.2]octane [DABCO, 2.3% (w/v)], 10% PBS, 87.7% glycerol), and sealed with nail polish. Microscope images were captured with a Nikon Eclipse E1000 using Meta-Morph software. Gem counts were performed in triplicate, counting 500 cells for each treatment.
Patient fibroblast cells transfected with 2′-O-methyl RNAs were harvested 48hr posttransfection. Cells were added to 10μl of sodium dodecyl sulfate (SDS) loading dye, boiled for 5min, and resolved on an SDS–10% polyacrylamide gel. The gel was transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, Billerica, MA) and an SMN immunoblot was performed with 4F11 monoclonal SMN antibody (Wolstencroft et al., 2005). Horseradish peroxidase-conjugated goat anti-mouse antibody was used as the secondary antibody and was detected by enhanced chemiluminescence (Pierce Biotechnology, Rockford, IL). Blots were H2O2 inactivated (H2O2/PBS, 1:1) and reprobed with anti-actin antibodies and developed with anti-rabbit secondary antibody by similar methods. All Western blots were performed in triplicate or quadruplicate and representative images are shown.
Injected mouse tissues were harvested 24hr postinjection for brain samples and 48hr postinjection for kidney and liver samples and snap frozen in liquid nitrogen. Samples were then thawed on ice and prepared for Western blot as previously described. All Western blot images were captured with a Fuji Imager and Multi-Gage V2.3 system. Each Western blot was performed in triplicate or quadruplicate and representative results are shown.
All animals were housed and treated in accordance with animal care and use committee guidelines. SMN2+/+, SMNΔ7+/+, and Smn−/− mice (Le et al., 2005) were genotyped on the day of their birth, designated day 0, and injected on day 1. Intravenous injection was performed as previously described (Sands and Barker, 1999), using 2′-O-methyl RNA at 12.5mg/kg. Intraventricular injection was performed as previously described (Passini and Wolfe, 2001), using 2μg of 2′-O-methyl RNA per ventricle. Neonates were then returned to their mothers for 24–48hr, and then anesthetized with isoflurane and killed by decapitation.
Antisense RNAs that recruit negatively acting proteins have been successfully developed to modulate alternative pre-mRNA splicing in other systems (Villemaire et al., 2003; Gendron et al., 2006). To determine whether negatively acting bifunctional RNAs could be developed that stimulated full-length SMN expression, we used a previously published antisense region (Madocsai et al., 2005) and three tandem repeats of the high-affinity hnRNPA1 sense region (Burd and Dreyfuss, 1994) (Fig. 1A). The SELEX-determined hnRNPA1-binding site was used because this binding motif would have a high probability of recruiting hnRNPA1. To initially screen these bifunctional RNAs, cassettes encoding the bifunctional RNAs were cloned into a previously published delivery vector, pMU2 (Baughan et al., 2006). The bifunctional RNA is driven by a polymerase III-dependent U6 promoter and a separate polymerase II-dependent cytomegalovirus (CMV) promoter drives expression of green fluorescent protein (GFP), thus allowing detection of transfected cells.
To determine whether the exon8-hnRNPA1 bifunctional RNA could alter SMN2 exon 7 splicing, a previously described SMN2 luciferase reporter plasmid was used to analyze the exon8-hnRNPA1 bifunctional RNA (Zhang et al., 2001). The SMN2-luciferase vector is designed to produce luciferase only when exon 7 is in frame; therefore, compounds or RNAs that induce exon 7 inclusion should result in an increase in luciferase activity. When the SMN1 construct was transiently expressed in HeLa cells, expected high levels of luminescence were detected, whereas the SMN2 construct expressed lower levels of luminescence (Fig. 2). Cotransfection of the plasmid expressing the exon8-hnRNPA1 bifunctional RNA, exon8-hnRNPA1-pMU2, resulted in a significant increase in luciferase activity compared with SMN2 alone (Fig. 2), whereas cotransfection of the SMN2 luciferase plasmid and the parental pMU2 plasmid that expresses an RNA of similar size did not increase luminescence (Fig. 2).
SMA type I patient fibroblasts (GM03813 cells) are an excellent cell-based system with which to study potential SMA therapies because these cells lack SMN1 and retain two copies of SMN2 (Coovert et al., 1997). These cells contain low levels of SMN protein and few SMN-positive nuclear foci, termed “gems,” which are a biomarker for disease severity as carriers and unaffected cell lines have higher gem numbers than do SMA cells (Coovert et al., 1997). Because the luciferase assay demonstrated an increase in splicing to SMN2 exon 7 in exon8-hnRNPA1-transfected cells, we next sought to determine whether exon8-hnRNPA1-pMU2 induced increased SMN levels in a more complex and disease-relevant cellular context. To this end, GM03813 cells were transfected with the plasmid expressing the exon8-hnRNPA1 bifunctional RNA. Plasmid DNA transfection levels of GM03813 cells, which are primary patient fibroblasts, are low, as expected. However, transfected cells are easily identified because of the plasmid-derived GFP moiety encoded by a separate promoter within the pMU2 derivatives. After transient transfection of the exon8-hnRNPA1-expressing plasmid for 48hr, SMN-positive gem numbers were increased severalfold above those of untransfected or parental pMU2-transfected cells, as determined by indirect immunofluorescence (Fig. 3). The parental vector, pMU2, expresses a similarly sized scrambled RNA that has no homology to the SMN sequences.
As it is not currently practical to deliver plasmid-derived bifunctional RNAs as an SMA therapeutic, the previous qualitative analysis of bifunctional RNAs served as a screen to determine whether negatively acting bifunctional RNAs could increase SMN levels. 2′-O-Methyl RNAs are a more tractable therapeutic candidate and are more easily transfected into primary patient fibroblasts, and were therefore selected as the chemistry to analyze in the second generation of bifunctional RNAs. The addition of the 2′-O-methyl moiety stabilizes the RNA and allows for sustained expression in vivo. To evaluate the ability of the exon8-hnRNPA1 bifunctional 2′-O-methyl RNA to induce SMN levels, several additional RNAs were designed: “ex8 antisense” corresponds exclusively to the antisense segment within the exon8-hnRNPA1 bifunctional RNA and is a previously published antisense that increased splicing to exon 7 (Madocsai et al., 2005); “hnRNPA1-act” is an RNA that lacks any antisense component and is composed solely of the hnRNPA1-binding motifs present in the bifunctional RNAs, and can be thought of as the activation (act) alone; “int7/ex8” is a previously published antisense RNA that targets the intron 7/exon 8 boundary and was only minimally able to increase splicing to exon 7 (Madocsai et al., 2005); “int7/ex8-hnRNPA1” is a bifunctional RNA with the same antisense domain as “int7/ex8” and the same hnRNPA1-binding motifs as “exon8-hnRNPA1” (Fig. 1B).
To determine whether these RNAs increase total SMN levels, the RNAs were transfected into GM03813 cells. RNA transfection levels in these cells are significantly higher than plasmid transfection levels, and can range from 10 to 30% (data not shown). After 48hr, cells were analyzed by indirect immunofluorescence to identify SMN-positive gems (Fig. 4A). In cells transfected with the exon8-hnRNPA1 bifunctional RNA, gem numbers were significantly elevated when 500 randomly selected cells were analyzed (Fig. 4B). In this system, it was not possible to identify the transfected cells because the RNA molecules were not tagged with a fluorescent moiety. The additional 2′-O-methyl RNAs were analyzed similarly, and were shown to have a modest impact on gem numbers, albeit greater than untransfected GM03813 levels. However, gem numbers from the additional RNAs were significantly lower than those associated with the exon8-hnRNPA1 bifunctional RNA (Fig. 4B).
As the previous results demonstrated an increase in SMN-containing gems, we next sought to determine whether bifunctional RNAs altered steady state levels of total SMN protein in GM03813 cells. GM03813 cells were transfected with the panel of 2′-O-methyl RNAs, extracts were generated from transfected cells, and SMN levels were measured by Western blotting with an anti-SMN monoclonal antibody. In extracts from cells transfected with the exon8-hnRNPA1 bifunctional RNA, total SMN levels were consistently increased, resulting in levels of SMN similar to those detected in extracts from unaffected 3814 cells. Additional RNAs that resulted in lower levels of gems were less robust inducers of SMN (Fig. 5).
To determine whether the exon8-hnRNPA1 bifunctional RNA could elevate SMN levels in an SMA-specific context, the facial vein of 1-day-old SMA mice was injected with either the exon8-hnRNPA1 bifunctional RNA or int7/ex8 antisense RNA. Facial vein delivery in neonates is an effective method to achieve systemic delivery (Sands and Barker, 1999). The SMA mouse model used in these experiments is well characterized, lacks endogenous murine Smn protein, and has the following genotype: SMN2+/+; SMNΔ7; Smn−/−. Importantly, the genomic SMN2 is present, which allows alternative splicing and the examination of potential therapeutics; therefore, increases in SMN protein are due to an induction of the human SMN2 transgene. Forty-eight hours postinjection, SMN levels were measured in kidney and liver, as these are the tissues most likely to be exposed to the RNAs by this type of delivery protocol. Consistent with the cellular analysis of the exon8-hnRNPA1 bifunctional RNA, Western blot analysis demonstrated that SMN levels were increased in the kidney and liver of exon8-hnRNPA1 bifunctional RNA-injected animals (Fig. 6A). SMN levels were poorly elevated in one mouse, potentially because of an injection error or because of a lack of responsiveness to the exon8-hnRNPA1 bifunctional RNA either at the uptake level or intracellularly. In addition, one int7/ex8-injected mouse showed a small SMN increase in liver. Although this latter result is incongruous with the previous cell-based analysis, it is not completely unexpected as the antisense sequence within the int7/ex8 bifunctional RNA corresponds to a previously published antisense sequence that was capable of modestly altering SMN2 pre-mRNA splicing (Madocsai et al., 2005). In this previously published study, however, SMN protein levels were not examined in cells or in vivo.
Because SMA is a neurodegenerative disease, SMA mice were used to determine whether the exon8-hnRNPA1 RNA could increase SMN levels in central nervous system tissue. To deliver the RNAs to the central nervous system, intraventricular injections of exon8-hnRNPA1 and int7/ex8-hnRNPA1 bifunctional RNAs were performed on 1-day-old SMA mice. Twenty-four hours postinjection, brain tissue was harvested and Western blotting of total protein extracts was performed to determine total SMN protein levels in the brain. Brain extract from five exon8-hnRNPA-injected SMA mice was examined. Consistent with the analysis of the intravenous injections, mice that received the intraventricular exon8-hnRNPA1 RNA injection exhibited a consistent increase in total SMN protein levels compared with the PBS-injected animal (Fig. 6B). D2-2 and hnRNPA1-act exhibited only modest levels of activation compared with PBS-treated samples. Taken together, these results identify negatively acting bifunctional RNAs as an efficient means to increase SMN protein levels in primary patient fibroblasts, as well as in the more complex in vivo context of the SMA mouse model.
SMA represents a unique genetic context for therapeutic development for several reasons. Approximately 97% of SMA cases are caused by the loss of SMN1, but retain at least one copy of SMN2 (Wirth et al., 1999). As SMN2 is a nearly identical copy of SMN1 and has the ability to encode the same protein as SMN1 when exon 7 is included in the fully spliced transcript, this makes the disease extremely amenable to therapies designed to promote full-length SMN mRNA. In addition, transcripts derived from SMN2 appear relatively stable (Heier et al., 2007), and similarly abundant compared with SMN1 transcripts.
The results presented in this study include the first description of a 2′-O-methyl bifunctional RNA functioning in vitro, and more importantly, in vivo in the mouse model of SMA. These RNAs supply in trans a binding platform for regulatory splicing factors, in this case hnRNPA1. The binding site for this study used an in vitro-determined hnRNPA1-binding sequence; however, it is possible that an empirically determined binding site may stimulate greater levels of full-length SMN production. The antisense region may contribute to the production of full-length SMN because the same targeting region was previously shown to increase exon 7 inclusion in cells transfected with synthetic antisense RNAs with U7 moieties (Madocsai et al., 2005). However, in these experiments, the antisense alone was less effective in producing endogenous SMN, demonstrating that the combined activity of the antisense and trans-factor-binding sites contribute to SMN production.
The results described herein show that a negatively acting bifunctional RNA designed to recruit hnRNPA1 to the SMN intron7/exon8 boundary is able to increase splicing to exon 7. As the exon8-hnRNPA1-based RNA blocks the exon 8 splice site decision, it is possible that although the splicing to exon 7 is increased via the SMN2-luciferase assay, this may not accurately represent the extent to which the bifunctional RNA is working, because if intron 7 is not removed, the luciferase cassette will be out of frame. It is possible, therefore, that the increase in SMN protein we observed does not directly relate to the increase in luminescence observed. In addition, it is important to examine SMN2 expression in a variety of experimental contexts because it appears that some of the oligonucleotides induced SMN levels to various degrees, depending on the cellular context. Although these are important considerations, the results presented here demonstrate that the leading candidate, exon8-hnRNPA1, was robust in all of the assays examined, ranging from reporter assays to a complex in vivo disease context within the central nervous system.
The ideal time during development that an SMA therapy must be administered in order to be effective is still unclear. This time line would need to be evaluated clearly in order to have an effective therapy. Another challenge facing SMA therapy is the necessity that the therapy cross the blood–brain barrier. As direct delivery to the ventricles may not be ideal for a human treatment, it is possible to express these small RNAs using recombinant adeno-associated virus (rAAV). This virus is an ideal candidate for gene therapy as it is replication deficient and certain AAV serotypes have been shown to have high tropism for both muscles and neurons and can use retrograde transport to travel from muscle to neurons in vivo. Previous studies have successfully used the retrograde transport ability of a pseudotyped lentivirus to deliver SMN cDNA (Azzouz et al., 2004). Although these data have direct implications for SMA therapy, they could also be transitioned to be applicable to many disease states in which it would be desirable to skip an exon, such as in muscular dystrophy, where skipping an exon or set of exons may return the proper reading frame.
The authors thank John Marston and Jacqueline King for technical assistance. A.D. was partially funded by a University of Missouri Life Sciences Fellowship. This work was funded by grants from FightSMA (C.L.L.) and the National Institutes of Health (C.L.L, R01 NS41584; R01 HD054413).
There are no competing or financial interests to disclose.