Antisense Oligonucleotide-Mediated Terminal Intron Retention of the SMN2 Transcript.

The severe childhood disease spinal muscular atrophy (SMA) arises from the homozygous loss of the survival motor neuron 1 gene (SMN1). A homologous gene potentially encoding an identical protein, SMN2 can partially compensate for the loss of SMN1; however, the exclusion of a critical exon in the coding region during mRNA maturation results in insufficient levels of functional protein. The rate of transcription is known to influence the alternative splicing of gene transcripts, with a fast transcription rate correlating to an increase in alternative splicing. Conversely, a slower transcription rate is more likely to result in the inclusion of all exons in the transcript. Targeting SMN2 with antisense oligonucleotides to influence the processing of terminal exon 8 could be a way to slow transcription and induce the inclusion of exon 7. Interestingly, following oligomer treatment of SMA patient fibroblasts, we observed the inclusion of exon 7, as well as intron 7, in the transcript. Because the normal termination codon is located in exon 7, this exon/intron 7-SMN2 transcript should encode the normal protein and only carry a longer 3' UTR. Further studies showed the extra 3' UTR length contained a number of regulatory motifs that modify transcript and protein regulation, leading to translational repression of SMN. Although unlikely to provide therapeutic benefit for SMA patients, this novel technique for gene regulation could provide another avenue for the repression of undesirable gene expression in a variety of other diseases.

Introduction

With a frequency of 1 in 10,000 live births, the neurodegenerative disease spinal muscular atrophy (SMA) is the leading genetic cause of infant death. SMA arises from inadequate levels of the survival motor neuron (SMN) protein that ultimately results in the death of motor neurons. While the survival motor neuron 1 (SMN1) gene is missing in most SMA patients, copies of the homologous gene, SMN2, potentially compensate for SMN production; however, a C > T base change in SMN2 exon 7 results in exclusion of the exon from 90% of neuronal SMN2 transcripts.3, 4 To date, the main RNA therapeutic focus for SMA has been the use of antisense oligonucleotides (AOs) to enhance SMN2 exon 7 inclusion and increase SMN levels (for review, see Porensky and Burghes). In particular, a 2′O-methoxyethyl (MOE) AO covering the ISS-N1 splicing domain (Anti-ISS-N1) has shown promise in clinical trials6, 7, 8 and has recently received approval by the U.S. Food and Drug Administration. However, the therapy is by no means definitive, with unknown consequences of long-term AO exposure and further improvements in AO efficacy needed before this therapy can be considered a qualified success. While other studies have focused on targeting AOs to intronic splice silencing motifs to enhance exon 7 inclusion,8, 10, 11, 12 AO-mediated splice modification has broader potential.

The strategy described here was focused on targeting AOs to the last exon in an attempt to slow transcription rates and concurrent pre-mRNA processing to temporarily stall the spliceosome machinery. Others have shown that a slow RNA polymerase II elongation rate during transcription can increase the “window of opportunity” for upstream splicing events, with alternative exons more likely to be included in the mature transcript.13, 14 To determine whether slower transcription elongation could be induced by an AO, we targeted AOs to SMN2 exon 8 in an attempt to increase the inclusion of SMN2 exon 7 in the transcript.

Unexpectedly, AOs targeting SMN2 exon 8 induced the retention of exon 7 and intron 7 in the mature transcript. Interestingly, an AO covering the exon 8 acceptor site has been reported by others to induce exon 7 and intron 7 retention, yet this work was not pursued further. Because the normal termination codon is located within exon 7, this induced transcript should therefore encode the normal full-length protein; however, the size of the 3′ UTR is increased. It is well documented that the length of the 3′ UTR can affect transcript stability and protein translation, with longer 3′ UTRs having more opportunity for the binding of microRNAs and regulatory elements (for review, see Barrett et al.). However, the consequences of intron retention within the mature transcript, and more specifically within the 3′ UTRs, are a more recently explored and less well understood area.

A study by Braunschweig and colleagues reported that three-quarters of mammalian multi-exon genes exhibit intron retention within the mature transcript as a result of alternative splicing events. While 6%–16% of 3′ UTRs are suggested to contain introns, it is unclear at this stage what percentage of these have the propensity to retain an intron within the mature message. While transcripts containing introns within the 3′ UTR were once believed to be non-functional due to nonsense-mediated decay,19, 20 there is now evidence to show that intron retention within the mature message is an important mechanism for transcript and protein regulation (for review, see Bicknell et al. and Ge and Porse). Tissue-specific transcript regulation by intron retention is particularly common in neuronal cells during differentiation and maturity, and recent studies have revealed a role for intron retention in hematopoietic cellular differentiation. Furthermore, intron retention within the 3′ UTR has been shown to play a role in transcript autoregulation to maintain protein homeostasis, a mechanism that is particularly common in proteins involved in forming the spliceosome and in regulating pre-mRNA processing.23, 24

A number of factors have been reported to regulate splicing events resulting in intron retention, with a correlation observed between intron retention and the presence of certain regulatory cis elements. Of particular interest, intron retention has been suggested to be the result of stalling of the RNA polymerase II elongation due to poor splicing factor recruitment and weakened splicing in non-essential transcripts.17, 25 Other factors influencing this mechanism include the position of the intron within the transcript, reduced intron length, an increase in G/C content within the intron, and weak splice site strength.

While factors that determine intron retention have been studied in canonical splicing events, it is unknown what role they play in mediating AO-induced intron retention and transcript expression. Consequently, this study focused on gaining a further understanding of the mechanisms influencing AO-induced intron retention and, furthermore, investigating how it can impact transcript and protein expression as a potential strategy in treating genetic disease.

Results

Targeting AOs to Exon 8 Results in Exon 7 and Intron 7 Retention in SMN2 Transcripts

SMA type I fibroblasts (Coriell GM03813) were transfected with 2′O-methyl AOs targeting SMN2 exon 8 (for binding coordinates and AO sequences, see Table 3) at 300, 150, and 75 nM and incubated (37°C) for 48 hr. RT-PCR analysis (Figure 1A) of SMN2 showed an increase in abundance of an approximately 850-bp product, which was confirmed by sequencing (Figure 1B) to be the SMN2 transcript retaining exon 7, as well as intron 7 (848 bp). This product is referred to as exon/intron 7-SMN2 and is labeled ex/in7 in the figures. Because the stop codon is located within exon 7, the addition of an extra 444-bp intronic sequence should encode the same protein as SMN1, but increases the length of the 3′ UTR (Figure 1C). These results were reproducible in two unrelated SMA patient primary cell strains (data not shown), including an SMA type II patient (prepared in-house) and an SMA type I patient with only one copy of SMN2 (Coriell GM00232). Two additional bands were observed at approximately 100 bp above and 100 bp below the exon/intron 7-SMN2 transcript. The larger band was deemed to be a PCR artifact because it was unable to be re-amplified and disappeared following increasing primer annealing temperature. The lower band was confirmed by sequencing to be the naturally occurring Δ5-SMN2 transcript containing intron 7 (data not shown).

Table 3

AO Sequences and SMN1 and SMN2 Binding Coordinates

Synthesis Round	AO Number	Annealing Coordinates	Sequence 5′ → 3′	Length (bp)
1: Initial screening	1	hSMN8A(−18+7)	GCA UUU CCU GCA AAU GAG AAA UUA G	25
	2	hSMN8A(−12+13)	AUG CCA GCA UUU CCU GCA AAU GAG A	25
	3	hSMN8A(−7+18)	GCU CUA UGC CAG CAU UUC CUG CAA A	25
	4	hSMN8A(−2+23)	GUG CUG CUC UAU GCC AGC AUU UCC U	25
	5	hSMN8A(+4+28)	AUU UAG UGC UGC UCU AUG CCA GCA U	25
	6	hSMN8A(+9+33)	GUG UCA UUU AGU GCU GCU CUA UGC C	25
	7	hSMN8A(+14+38)	UAG UGG UGU CAU UUA GUG CUG CUC U	25
	8	hSMN8A(+24+48)	AUC GUU UCU UUA GUG GUG UCA UUU A	25
	9	hSMN8A(+34+58)	GAU CUG UCU GAU CGU UUC UUU AGU G	25
	10	hSMN8A(+59+83)	AUC UUC UAU AAC GCU UCA CAU UCC A	25
	11	hSMN8A(+84+108)	AUA UUU UGA AGA AAU GAG GCC AGU U	25
	12	hSMN8A(+152+178)	CAU AAC UUU UAA UCA AGA AGA GUU ACC	27
	13	hSMN8A(+24+43)	UUC UUU AGU GGU GUC AUU UA	20
	14	hSMN8A(+44+63)	UUC CAG AUC UGU CUG AUC GU	20
	15	hSMN8A(+64+83)	AUC UUC UAU AAC GCU UCA CA	20
	16	hSMN8A(+29+48)	AUC GUU UCU UUA GUG GUG UC	20
	17	hSMN8A(+35+53)	GUC UGA UCG UUU CUU UAG UG	20
	18	hSMN8(+39+58)	GAU CUG UCU GAU CGU UUC UU	20
2: Microwalking	19	hSMN8A(−12+18)	GCU CUA UGC CAG CAU UUC CUG CAA AUG AGA	30
	20	hSMN8A(−10+15)	CUA UGC CAG CAU UUC CUG CAA AUG	24
	21	hSMN8A(+39+63)	UUC CAG AUC UGU CUG AUC GUU UCU U	25
	22	hSMN8A(+44+68)	UCA CAU UCC AGA UCU GUC UGA UCG U	25
	23	hSMN8A(+55+79)	UCU AUA ACG CUU CAC AUU CCA GAU C	25
	24	hSMN8A(+57+81)	CUU CUA UAA CGC UUC ACA UUC CAG A	25
	25	hSMN8A(+59+78)	CUA UAA CGC UUC ACA UUC CA	20
Control AOs	ISS-N1	hSMN2.7D(−10-29)	AUU CAC UUU CAU AAU GCU GG	20
	Skipping	mSmn7A(+7+36)	UGA GCA CUU UCC UUC UUU UUU AUU UUG UCU	30
	Sham	scrambled sequence	GCT ATT ACC TTA ACC CAG	18

Figure 1

SMN Transcript Analysis following 2′OMethyl AO Transfection

Analysis of SMA fibroblasts following transfection with exon 8 targeting 2′O-methyl AOs, showing (A) RT-PCR analysis of SMN2 transcripts from transfected fibroblasts (300, 150, 75 nM). Anti-ISS-N1 was used as a positive control for transfection efficiency, and a sham control AO was used as a transfection negative control. A 100-bp marker was used for comparison, and an RT-PCR no-template negative control was loaded in the final lane, (B) sequencing chromatogram across SMN2 exon 7 to intron 7 and intron 7 to exon 8, confirming the presence of intron 7 within the mature transcript, and (C) schematic showing SMN2 transcripts identified within fibroblasts transfected with exon-8-targeting AOs, including full-length SMN (FL-SMN), the Δ7-SMN transcript missing exon 7, and the exon/intron 7 retained transcript (ex/in7-SMN) with the extended 3′ UTR.

The initial screening of AO sequences 1–18 is shown in Figure S1. Following preliminary screening, additional AOs were designed by microwalking around promising AO target sites, shifting up or downstream of the original sites (Table 3). Analysis of SMN2 transcripts following transfection with refined AO sequences showed an improvement in AO-induced exon/intron 7 retention (Figure 1A). A clear dose response was observed in all AO-treated cells, with AOs 10, 18, 24, and 25 consistently inducing the highest levels of inclusion across experiments (n = 6). These promising AOs were therefore selected for further evaluation, including protein analysis.

Splice Site Analysis Shows a Weak Exon 7 Donor Splice Site

To further investigate the exon/intron 7-SMN2 transcript induced by AOs targeting exon 8, we analyzed splice site scores (Table 1) using the online Human Splicing Finder 3.0 website. SMN2 exon 7 was predicted to have a very strong acceptor site with a score of 98.2 out of a possible 100, while the donor splice site was weaker, scoring 82.81 out of 100. The exon 8 acceptor splice site had a predicted score of 91.9 out of 100. While these splice site scores are only a predicted measurement of the likelihood of the site being recognized by the splicing machinery, the comparatively weaker exon 7 donor splice site could lead to reduced splicing at the exon/intron 7 junction when the intron 7/exon 8 junction is further compromised following AO treatment.

Table 1

Exons 7 and 8 Splice Site Predictions Using Human Splicing Finder 3.0

Exon	Splice Site Type	Splice Site Motif	Consensus Value (0–100)
7	acceptor	ttttccttacagGG	98.2
7	donor	GGAgtaagt	82.81
8	acceptor	tctcatttgcagGA	91.9

PMO Delivery by Electroporation Improves Exon/Intron 7 Inclusion, Inducing a Decrease in SMN Protein

Previously identified optimal 2′O-methyl AO sequences 10, 18, 24, and 25 were resynthesized as phosphorodiamidate morpholino oligomers (PMOs) by Genetools (Philomath, OR, USA), and are now cited as PMOs 10, 18, 24, and 25. PMOs were administered to cells using nucleofection for optimal delivery at 1 and 0.5 μM for SMN transcript and protein analysis by RT-PCR and western blot, respectively. Nucleofection of PMOs showed increased levels of exon/intron 7 retention in the mature transcript compared with the same sequences tested as 2′O-methyl AOs, with a clear reduction in the levels of FL-SMN and Δ7-SMN transcripts. In particular, PMO-10 induced almost 100% exon/intron 7 inclusion as determined by RT-PCR (Figure 2A).

Figure 2

SMN Transcript and Protein Analysis in SMA Fibroblasts following PMO Nucleofection

SMN transcript and protein levels in SMA fibroblasts transfected with PMOs by nucleofection at 1 and 0.5 μM, showing (A) RT-PCR analysis of SMN2 products confirming exon/intron retention, (B) western blots showing SMN protein levels compared with β-tubulin levels, and (C) densitometric analysis showing changes in SMN protein levels normalized against β-tubulin. SMN levels in transfected fibroblasts are shown as an n-fold change compared with those in samples from untreated cells. Error bars represent the SEM.

Interestingly, western blot analysis of SMN protein levels revealed a significant decrease in the amount of SMN detected in samples transfected with exon-8-targeting PMOs (Figures 2B and 2C). PMO-10 and PMO-24 were the most effective compounds inducing a respective 50% (p = 0.022) and 33% (p = 0.027) decrease in SMN protein when compared with the level observed in untreated fibroblasts (n = 4). The Anti-ISS-N1 PMO sequence was transfected as a positive control and was shown to increase SMN levels by up to 80% compared with that in untreated SMA patient fibroblasts (p = 0.032).

PMO-Induced SMN Knockdown Is Reproducible in Unaffected Fibroblasts

PMO-24 and PMO-25 were evaluated in non-SMA fibroblasts to determine the effects of intron 7 retention on SMN protein levels in cells with a higher baseline of SMN. PMOs were transfected by nucleofection at 1 and 0.5 μM, and incubated for 3 days prior to western blot analysis. RT-PCR analysis of the total SMN transcripts confirmed that exon-8-targeting AOs induce almost 100% exon/intron 7 retention, and hence this must represent both the SMN1 and the SMN2 transcripts (Figure 3A). Consistent with the findings in SMA patient fibroblasts, western blot analysis (Figures 3B and 3C) demonstrated that PMO-24 and PMO-25 effectively decreased the levels of SMN protein in non-SMA fibroblasts by 55% (p = 0.041) and 38% (p = 0.072), respectively (n = 3). Anti-ISS-N1 was transfected as a positive control and was shown to increase the levels of SMN protein by up to 35% as seen in non-SMA cells transfected with the low AO dose; however, this was not statistically significant. Furthermore, an AO designed to induce exon 7 skipping was transfected into non-SMA fibroblasts as a positive control for downregulating SMN levels. Fibroblasts transfected with this PMO show a 76% decrease in SMN levels compared with sham control and untreated fibroblasts (p = 0.030).

Figure 3

SMN Transcript and Protein Analysis in Unaffected Fibroblasts following PMO Nucleofection

SMN transcript and protein levels in unaffected fibroblasts transfected with PMOs by nucleofection at 1 and 0.5 μM, showing (A) RT-PCR analysis of SMN products confirming exon/intron retention, (B) western blots showing SMN protein levels compared with β-tubulin levels, and (C) densitometric analysis showing changes in SMN protein levels normalized against β-tubulin. SMN levels in transfected fibroblasts are shown as an n-fold change compared with those in samples from untreated cells. Error bars represent the SEM.

Functional SMN protein forms aggregates with the gemin proteins that fluoresce as bright sparkling foci, reminiscent of gems after antibody staining. Therefore, the presence of gems within the cell indicates functional localization of SMN protein. The percentages of fibroblast nuclei staining positive for gems were counted (Figure 4A). PMOs 10, 18, 24, and 25 were all transfected into non-SMA cells by nucleofection at 1 μM and incubated for 3 days prior to fixation and immunofluorescent staining. Interestingly, the sham control PMO induced an increase in nuclei containing gems from 16.3% in untreated fibroblasts to 20.6% following control AO transfection. However, PMOs 10, 18, 24, and 25 all decreased the number of nuclei containing gems, with as low as 7.3% of fibroblasts containing gems following transfection with PMO-10. In comparison, 25.9% of fibroblasts transfected with the Anti-ISS-N1 PMO sequence express gems, while only 3.3% of those transfected with the exon skipping control PMO express gems. Representative images of fibroblasts transfected with each PMO are shown in Figure 4B.

Figure 4

Immunofluorescence Staining and Analysis of SMN Protein following PMO Nucleofection

Immunofluorescence analysis of SMN shown as gems in PMO nucleofected unaffected fibroblasts (1 μM), compared with untreated fibroblasts. (A) Graph displaying the percentages of cell nuclei containing gems, as indicated above each bar, and (B) representative images showing anti-SMN- (green) and Hoechst (blue)-stained SMA fibroblasts following PMO nucleofection. Typical gems stained within the cell nucleus are indicated with white arrows. Images were taken at 20× objective. Scale bar, 25 μm.

PMOs targeting exon 8 induce more efficient retention of exon and intron 7 in both the SMN1 and the SMN2 transcripts compared with the 2′O-methyl AOs of the same sequence, and analysis of the SMN protein by western blot and immunofluorescence shows a further decrease in SMN expression following transfection. While the exon/intron 7-SMN transcript occurs naturally at low levels in untransfected cells, it appears that the extended 3′ UTR introduces a number of new regulatory mechanisms into the transcript that negatively impact on protein expression.

Intron Retention Introduces Negative Regulatory Elements to the 3′ UTR

In silico analysis of the extended 3′ UTR was carried out using the online tools UTRscan, miRBase, Polyadq, and DNA Functional Site Miner (DNA FS Miner). Table 2 lists the potential regulatory elements identified within intron 7 from each of these databases, and Figure 5A illustrates the location of these elements within intron 7. In silico analysis of the extended 3′ UTR by the online tool UTRscan drew attention to two possible regulatory motifs known as the bearded (BRD) box and the K box. Each of these motifs has been shown by others to disrupt translation of neuronal gene transcripts during Drosophila development by recruiting microRNAs.31, 32 The BRD box consensus sequence is AGCUUUA and for K box is UGUGAU. A search for microRNA recognition sites using miRBase revealed three potential microRNA binding sites within intron 7, with E values below 10, that suggests that these sites are active. The microRNAs hsa-miR-3118, hsa-miR-3976, and hsa-miR-5580-3p all have complementary bases to the SMN intron 7 sequence within the seed region. It is therefore possible that these microRNAs or the BRD and K box motifs could disrupt SMN translation.

Table 2

In Silico Analysis of Potential Regulatory Elements within SMN Intron 7

The prediction value or score indicates the strength of the regulatory site within the SMN sequence. For poly(A) signals, scores >0.5 are true predictions for Polyadq and scores >0.6 are true predictions for DNA FS Miner. For miRBase, E values <10 may have binding potential.

NA, not applicable.

Figure 5

Analysis of Predicted SMN Intron Regulatory Elements

Analysis of predicted SMN intron 7 regulatory elements, showing (A) a schematic of the in silico analysis of regulatory elements within intron 7, (B) a schematic of the location of potential poly(A) signal and cleavage sites and the primer binding sites, (C) RT-PCR across SMN and Cyclin-D (housekeeping gene) transcripts in fibroblasts nucleofected with 1 μM (1) PMO-24, (2) PMO-25, (3) control PMO, and (4) an untreated sample, using indicated primers, and (D) sequencing chromatogram of the RT-PCR product amplified using the exon 6/7 forward and poly(A)R2 reverse primers.

The exon/intron 7-SMN transcript was further analyzed for polyadenylation [poly(A)] signals using two online tools, Polyadq and DNA FS Miner. Each tool identified two potential poly(A) sites with corresponding CA cleavage sites within intron 7. A potential poly(A)-1 (ATTAAA) signal was identified at 132 bases into intron 7, and a potential poly(A)-2 (AATAAA) signal was identified at 238 bases into intron 7. To determine whether early polyadenylation could destabilize the extended SMN transcript following PMO treatment, we designed specific poly(A) primers to target the predicted cleavage sites and downstream sequences to amplify polyadenylated products (Figure 5B). Following nucleofection of PMOs into unaffected fibroblasts, RNA was collected at multiple time points including 12, 24, 48, and 72 hr. Samples were DNase treated and RNA was amplified using the exon/intron 7 forward primer with the specific poly(A)-R1 and R2 primers (Figure 5C). No differences were observed within each treatment group over the 72-hr duration of the time course.

Specific poly(A)-R1 primer binding to the first ATTAAA poly(A) site amplified a faint product in some samples, suggesting this site could initiate polyadenylation. RT-PCR using the specific poly(A)-R2 primer directed to the second AATAAA site resulted in two products, a stronger amplicon amplified by the second cleavage site, as well as a fainter non-specific amplification of the first cleavage site. The stronger amplicon was sequenced and confirmed to have a poly(A) tail extending past the primer annealing site (Figure 5D). This result suggests that early polyadenylation is occurring at this second AATAAA site within intron 7, and as a result could destabilize the exon/intron 7-SMN transcript and therefore result in decreased protein levels.

To compare the stability and cleavage of the exon/intron 7-SMN and FL-SMN transcripts, we designed primers to target downstream of the poly(A) sites, and we tested them with a forward primer targeting the exon/intron 7 boundary (Figure 5C). Interestingly, both primer sets produce a strong amplicon extending beyond the poly(A) signal and do not show diminished expression of the transcript following AO treatment. Taken together, these results show that while early polyadenylation appears to occur at the second poly(A) signal, the transcript level remains stable, suggesting that transcription may be occur at a faster rate than polyadenylation and cleavage, or that the early polyadenylation is not destabilizing the transcript.

Discussion

The original intent of this study was to influence the rate of SMN2 transcription by targeting AOs to the terminal exon in an attempt to increase exon 7 inclusion. However, another splice-switching mechanism for manipulating expression was revealed. Selected AOs targeting SMN2 exon 8 promoted exon 7 and intron 7 inclusion in the mature SMN message, revealing a novel AO application: inducing terminal intron retention. In silico analysis of the SMN2 exon 7 splice sites suggests that this action may be the result of a strong acceptor splice site (scoring 98 out of 100) and a weaker donor splice site (scoring 83 out of 100). The weaker donor site might be subject to poor recognition by the splicing machinery, abrogating splicing between exon 7 and the following exon, especially when the splicing of exon 8 is compromised by AO binding.

The correlation in splice-switching efficacy between 2′O-methyl and PMO compounds of the same sequence is well established for induced exon skipping in dystrophin transcripts as a therapy for Duchenne muscular dystrophy, as well as for the efficacy of the Anti-ISS-N1 sequence. Furthermore, the PMO chemistry has been shown to be safe, with no significant adverse events reported in a 5-year study of Duchenne muscular dystrophy patients receiving weekly treatment with Exondys51, a PMO drug granted accelerated approval by the U.S. Food and Drug Administration (FDA) in 2016 for amenable dystrophin mutations.35, 36

Consistent with the results of exon 8 targeting 2′O-methyl AOs, PMOs of the same sequence were effective at inducing exon/intron 7 retention. Interestingly, while PMOs targeting exon 8 increased the levels of the exon/intron 7-SMN transcript, these PMOs also induced a 50% decrease in SMN protein as assessed by western blot. This result was reproducible between SMA patient fibroblasts and unaffected fibroblasts following AO transfection. Similarly, immunofluorescence staining of the transfected fibroblasts showed fewer nuclei containing functional SMN in the form of “gems” when compared with sham-control PMO transfected cells. It is likely that the observed decrease in SMN protein could be due to the longer than normal 3′ UTR within the exon/intron 7-SMN transcript, and therefore many regulatory factors could contribute to protein downregulation.

Endogenous intron retention has been shown by others to act as a form of gene repression, often through the introduction of a premature termination codon, rendering the transcript susceptible to nonsense-mediated decay.17, 37 In the study presented here, nonsense-mediated decay is unlikely to be the cause of SMN downregulation due to the retained intron occurring after the normal termination codon. However, as a result of the extended 3′ UTR, influences of downstream sequences could lead to this transcript not being efficiently translated.

The length of the 3′ UTR can be a critical factor in regulating transcript stability and protein expression. A longer 3′ UTR can increase the opportunity for sequence-specific recognition motifs to recruit regulatory factors, including microRNAs. Furthermore, the possibility of altered mRNA secondary structures can influence the availability of such sequences to these factors. A number of online databases can identify potential microRNA and regulatory factor binding sites in NCBI documented transcripts. However, there is a lack of appropriate resources whereby an altered 3′ UTR sequence can be analyzed, limiting the search possibilities for this study. It is probable that there are many factors influencing the translational knockdown observed for the exon/intron 7-SMN transcript.

The microRNA prediction database miRBase allows the user to input an mRNA sequence for analysis, and analysis of the SMN intron 7 sequence revealed three potential microRNA binding sites. Of these, miR-3976 is a validated microRNA and is reportedly overexpressed in pancreatic cancer. However, how miR-3976 impacts translation, and whether it is expressed in the dermal fibroblast used in this study, is yet to be determined. Further in silico analysis of the sequence using the online UTRscan database drew attention to a number of motifs, including the BRD and K box motifs, as well as two alternative polyadenylation signals. The BRD box and K box motifs have been shown to recruit microRNAs that act as translational inhibitors of certain proteins during Drosophila development.31, 32, 40 The functionality of these motifs in human sequences is unknown and, therefore, any influence on SMN translation is only speculative.

Others have tested the use of AOs to prevent microRNAs from binding to a transcript, by either binding of the microRNA as an antagomir, or to act as a decoy, binding directly to the transcript of interest. To further investigate the mechanism of translational knockdown presented here, future studies could use AOs targeting the microRNAs themselves, their binding sites, as well as the BRD and K box motifs. Examining SMN expression in such studies could indicate whether these regulatory elements play a role in inhibiting SMN translation in the exon/intron 7-SMN transcript. However, it is unlikely that this will be of clinical benefit to SMA patients.

Further in silico analysis of the SMN2 intron 7 sequence identified two potential polyadenylation signals with corresponding CA cleavage sites that could prematurely cleave the mRNA, potentially resulting in transcript and protein destabilization. Premature polyadenylation and mRNA cleavage have been shown by others to cause less efficient processing of transcripts when compared with those cleaved at the distal 3′ or canonical poly(A) site. In this study, primers were designed to anneal to the sites identified within intron 7, and should amplify a product only if polyadenylation has occurred. The primer designed to anneal to the ATTAAA site failed to generate a product, while the primer annealing to the AATAAA site amplified a clean, consistent product whose levels increased with an increase in intron 7 retention.

It appears that the early AATAAA site within intron 7 initiates polyadenylation prior to use of the canonical poly(A) site within exon 8, and as such negatively affects SMN translation. However, primers amplifying downstream of the early poly(A) site produced a strong and consistent product that is not diminished following AO treatment at any time point, indicating that mRNA cleavage does not occur downstream of either of the poly(A) signals. Taken together, these results show that while early polyadenylation appears to occur at the second poly(A) signal, the transcript level remains stable, suggesting that transcription in this case may occur at a faster rate than polyadenylation and cleavage.

Interestingly, the product amplified by the specific poly(A)-R2 primer was also observed in untreated samples, suggesting polyadenylation may be a natural mechanism for controlling SMN levels. Furthermore, it has been reported that the canonical poly(A) signal and cleavage site in SMN exon 8 is inefficient at recruiting cleavage factors, and consequently SMN polyadenylation is subjected to additional regulation by U1A, a component of the U1 snRNP. The study showed that overexpression of U1A can inhibit SMN polyadenylation and cleavage, decreasing the levels of SMN protein. Given the inefficiency of the canonical poly(A) site, it is probable that the intron 7 poly(A) signal could be more favorable for initiating polyadenylation and cleavage. However, the presence of U1A may still inhibit cleavage at this site, and therefore explain the lack of mRNA cleavage observed in the RT-PCR experiments within the current study.

The presence of polyadenylation signals in terminal introns could contribute to the process of protein regulation by intron retention. To examine this theory further, future studies should assess the occurrence of polyadenylation in alternatively spliced 3′ UTR-intron retention transcripts. It would be interesting to investigate whether this effect of “premature polyadenylation” occurs in other gene transcripts.

AOs targeting the terminal exon to induce intron retention could be useful in those diseases where protein repression is essential for treatment, including many types of cancer. We speculate that in the study presented here, the extended SMN transcript is regulated by a number of motifs within intron 7 that are involved in translational repression due to the retention of this sequence in the 3′ UTR. However, this mechanism may only apply to a select number of genes. Alternatively, if the stop codon were to be in the final exon, intron retention could disrupt the reading frame or introduce a premature termination codon. To identify genes where intron retention could be applied, it will be necessary to look at a number of factors within the gene, most importantly the splice site scores for the flanking exons. If the donor site is strong, then intron retention may not be possible.

Aside from weakened splice sites of retained introns, additional cis and trans factors suggested to influence natural intron retention include the position of the intron in the transcript, an increase in G/C content, and reduced intron length. Interestingly, while the G/C content of SMN intron 7 is only 33.8%, being positioned adjacent to an alternatively spliced exon increases the probability of the intron being retained. Furthermore, at 444 nt long, intron 7 is relatively short compared with the median human intron length of 1,334 nt in the coding region and 1,303 nt within the 3′ UTR. It will be interesting in future studies to compare the relevance of these factors across transcripts to identify markers that could predict the likelihood of effective AO-induced intron retention.

In this study we present a novel application for splice-switching PMOs in initiating terminal intron retention. It is unfortunate that this model is unlikely to provide therapeutic benefit to SMA patients, yet further work could see intron retention being applied to a number of other diseases. This study is reflective of the ever-expanding complexity of gene regulation and undoubtedly sheds new light on splicing and AO mechanisms that may offer new avenues of therapy.

Materials and Methods

AO Design and Synthesis

AOs were designed to target the exon 8 acceptor splice site and exon splice enhancers (ESEs) as predicted by the online SpliceAid prediction tool, available at http://www.introni.it/splicing.html. AO nomenclature was based on that described by Mann et al. All 2′O-methyl PS-AOs were synthesized in-house on an Expedite 8909 nucleic acid synthesizer with a phosphorothioate backbone. Following identification of optimal 2′O-methyl AO sequences, these AOs were prepared as PMOs, purchased through Genetools (Philomath, OR, USA). Table 3 lists the details of all AOs used in this study.

2′O-Methyl AO Transfection

SMA type I patient fibroblasts (GM03183; Coriell Cell Repositories, Camden, NJ, USA) and normal human dermal fibroblasts prepared in-house (Murdoch University Human Research Ethics Committee Approval #2013/156) were proliferated and seeded in 10% fetal bovine serum (FBS) DMEM and incubated at 37°C for 24 hr prior to transfection. All 2′O-methyl PS-AOs were transfected using Lipofectin (Life Technologies, Melbourne, Australia) at a 2:1 ratio of lipofectin to total AO, according to manufacturer’s protocols, and incubated for 48 hr.

Nucleofection of PMOs

PMO delivery by nucleofection was performed using a Nucleofection X unit with the Nucleofection P2 kit, using the CA-137 program (Lonza, Melbourne, Australia). PMOs were transfected at 1 and 0.5 μM, as determined by the final transfection volume, supplemented with 5% FBS DMEM and incubated for 72 hr.

RNA Extraction and PCR

RNA was extracted using the MagMAX-96 Total RNA Isolation Kit, including a DNase treatment (Life Technologies), according to the manufacturer’s instructions. RT-PCRs were performed using the One-Step SuperScript III RT-PCR kit with Platinum Taq Polymerase (Life Technologies) according to manufacturer’s instructions. All primer sequences used in this study are detailed in Supplemental Materials and Methods. Products were amplified with the temperature profile, 55°C for 30 min, 94°C for 2 min, followed by 25–30 cycles of 94°C for 30 s, 56°C for 30 s, and 68°C for 1 min. Amplicon sequences were identified by Sanger sequencing at the Australian Genome Research Facility (AGRF, Perth, Australia).

Western Blot Analysis

Cell lysates were prepared in 125 mM Tris/HCl (pH 6.8), 15% SDS, 10% glycerol (v/v), 1.25 μM PMSF (Sigma-Aldrich), protease inhibitor cocktail (3 μL/100 μL; P8340; Sigma-Aldrich), 0.004% bromophenol blue, and 2.5 mM dithiothreitol. Pellets were sonicated six times for 1-s pulses and samples denatured at 94°C for 5 min.

Approximately 10 μg of total protein (as determined by BCA assay) was loaded per sample on a NuPAGE Novex 4%–12% BIS/Tris gel (Life Technologies). Proteins were transferred onto a Pall Fluorotrans polyvinylidene fluoride (PVDF) membrane at 350 mA for 1 hr in western transfer buffer. MANSMA7 (1:1,000; Developmental Studies Hybridoma Bank) and β-tubulin (1:20,000; DSHB) monoclonal primary antibodies were incubated overnight at 4°C prior to detection using a Western Breeze Chemiluminescent Immunodetection System (Life Technologies), according to the manufacturer’s instructions. Western blot images were captured on a Vilber Lourmat Fusion FX system using Fusion software, and Bio-1D software was used for densitometry analysis. All p values were calculated using a paired two-tailed t test, and SE bars were used to represent the SEM.

Immunofluorescence

Cells on coverslips were fixed using ice-cold acetone-methanol (1:1), then blocked in 10% filtered goat serum in PBS containing 0.2% Triton-X (PBT). SMN was detected with MANSMA1 (1:100; DSHB) antibody, incubated overnight at 4°C in PBT. Cells were stained with Hoechst (Sigma-Aldrich) for nuclei detection (1 mg/mL diluted 1:125), and the MANSMA1 primary antibody was detected using Alexa Fluor 488 (1:400; Thermo Fisher Scientific). Photos were overlaid, and the number of gems per nuclei was counted and recorded as a percentage of total nuclei, with at least 300 cells counted per slide.

In Silico Analysis

A number of online databases were used to analyze the extended 3′ UTR sequence to identify potential regulatory elements. Splice site scores were analyzed by Human Splice Finder version 3.0 available at http://www.umd.be/HSF3/. Regulatory element binding was predicted using UTR Scan available at http://itbtools.ba.itb.cnr.it/. Polyadenylation signals were analyzed using Polyadq available at http://rulai.cshl.edu/tools/polyadq/polyadq_form.html and DNA FS Miner available at http://dnafsminer.bic.nus.edu.sg/.

Author Contributions

Conceived and designed experiments: L.L.F., C.M., S.F., and S.D.W. Performed experiments: L.L.F., C.M., and I.L.P. Wrote and edited manuscript: L.L.F., C.M., I.L.P., S.F., and S.D.W.