Spinal muscular atrophy (SMA) is a progressive neurodegenerative disease associated with low levels of the essential survival motor neuron (SMN) protein. Reduced levels of SMN is due to the loss of the SMN1 gene and inefficient splicing of the SMN2 gene caused by a C>T mutation in exon 7. Global analysis of the severe SMNΔ7 SMA mouse model revealed altered splicing and increased levels of the hypoxia-inducible transcript, Hif3alpha, at late stages of disease progression. Severe SMA patients also develop respiratory deficiency during disease progression. We sought to evaluate whether hypoxia was capable of altering SMN2 exon 7 splicing and whether increased oxygenation could modulate disease in a severe SMA mouse model. Hypoxia treatment in cell culture increased SMN2 exon 7 skipping and reduced SMN protein levels. Concordantly, the treatment of SMNΔ7 mice with hyperoxia treatment increased the inclusion of SMN2 exon 7 in skeletal muscles and resulted in improved motor function. Transfection splicing assays of SMN minigenes under hypoxia revealed that hypoxia-induced skipping is dependent on poor exon definition due to the SMN2 C>T mutation and suboptimal 5′ splice site. Hypoxia treatment in cell culture led to increased hnRNP A1 and Sam68 levels. Mutation of hnRNP A1-binding sites prevented hypoxia-induced skipping of SMN exon 7 and was found to bind both hnRNP A1 and Sam68. These results implicate hypoxic stress as a modulator of SMN2 exon 7 splicing in disease progression and a coordinated regulation by hnRNP A1 and Sam68 as modifiers of hypoxia-induced skipping of SMN exon 7.
Proximal spinal muscular atrophy (SMA) is a progressive neurodegenerative disease associated with reduced level of the essential survival motor neuron (SMN) protein. SMN protein is encoded by two genes in humans: Survival Motor Neuron 1 and 2 (SMN1 and SMN2). Mutation or deletion of the SMN1 gene is the cause of SMA in humans, and the low SMN protein levels in SMA are generated by SMN2 expression (1–4).
Although SMN2 encodes SMN protein, only low levels of functional protein are generated due to inefficient splicing of exon 7 (5,6). Exon definition during pre-mRNA splicing is regulated by the strength of splice sites and can be modulated by the use of enhancer or silencer elements within both exonic and intronic sequences of the pre-mRNA. RNA-binding proteins act in the regulation of splicing efficiency and often bind splicing enhancer and silencer elements. Exon 7 of the SMN genes is innately poorly definition due to suboptimal 3′ and 5′ splice sites (7,8). However, the splicing defect of SMN2 exon 7 is primarily due to a silent C>T mutation at position 6 of exon 7 (SMNc.840C>T) that further reduces exon 7 definition (5,6). The C>T mutation disrupts an exonic splicing enhancer (ESE) site for SF2/ASF binding (SRSF1) and generates an exonic splicing silencer for hnRNP A1 binding (9–12).
Poor definition of exon 7 is modified by many regulatory elements within exon 7 and the surrounding intronic regions. The SMN2 C>T mutation has also been shown to bind the negatively acting splicing factor Sam68, which increases exon 7 skipping (13). The positively acting Tra2beta (SRSF10) splicing factor in complex with other splicing factors, hnRNP G and SRp30c, binds an ESE in the center of exon 7 (14–16). Also residing in the center of exon 7 is an hnRNP A1-binding site that functions in increasing exon 7 skipping (17). In addition to exonic regulatory elements, intronic regulatory elements also play a major role in exon 7 splicing. Within intron 7, a potent intronic splicing silencer, ISS-N1, has been shown to increase exon 7 skipping and bind to hnRNPs (hnRNP A1 and hnRNP A2/B1) (18,19). ISS-N1 is a strong inhibitor of exon 7 inclusion and is currently a primary target for therapeutic antisense oligonucleotide (ASO) splicing correction of SMN2 in SMA mouse models (20–23). Another hnRNP A1-binding site in intron 7 that leads to exon 7 skipping is also present in SMN2 due to an A>G nucleotide difference between SMN1 and SMN2 (24). Together, the interaction of both positively and negatively acting splicing factors modulates the efficiency of SMN exon 7 splicing.
The evaluation of severe mouse models of SMA at the RNA level using gene expression and splicing microarrays has identified transcriptional and splicing changes during the progression of disease. The observed changes increase nearing disease end-stage and include perturbations in myelination, growth factor signaling, cellular damage and hypoxia gene targets (25–27). One gene that undergoes both an increase in levels and alteration in splicing is the hypoxia-inducible gene, hypoxia inducible factor 3 alpha (Hif3alpha) (25). Proper regulation of hypoxia response has also been shown to be essential for motor neuron maintenance, as mice lacking the hypoxia-responsive element in the Vegf promoter develop late-onset neurodegeneration associated with motor neuron loss (28). Additionally, respiratory deficiencies were reported during the progression of disease in an SMA mouse model with three copies of SMN2 (29). Respiratory support is capable of extending survival in severe SMA patients (30). These data raise the questions of whether stress resulting from hypoxic conditions is a component of SMA disease progression and whether hypoxia can modulate splicing of the SMN2 gene.
Here we report increased SMN2 exon 7 skipping of endogenous SMN2 transcripts and reduced levels of SMN protein in cell lines treated with hypoxia. In the severe SMNΔ7 mouse model, SMA mice exhibit increased skipping of SMN2 compared with non-SMA controls, and when treated with higher oxygen levels they show signs of improved motor function as measured by righting reflex. The improved motor function correlates with increased inclusion of SMN2 exon 7 in skeletal muscle. Hypoxia-induced exon 7 skipping can be recapitulated in minigene transfection splicing assays, in which mutational analysis indicates that hypoxia-induced skipping is mediated by poor exon definition due to the C>T mutation and a suboptimal 5′ splice site (ss). Furthermore, mutation of hnRNP A1 sites in exon 7 and ISS-N1 can prevent exon 7 skipping under hypoxia. As both hnRNP A1 and Sam68 were elevated under hypoxia, we evaluated the binding of these factors to the hnRNP A1 sites under hypoxia which showed increased binding of these factors to the regulatory elements crucial for hypoxia-induced skipping. These data implicate hypoxic stress and the negatively acting splicing regulatory factors hnRNP A1 and Sam68 as co-regulators of SMN2 splicing efficiency under hypoxia that increases SMN2 exon 7 skipping and reduces SMN protein levels.
Hypoxia induces SMN2 exon 7 skipping in cell culture
To determine whether hypoxic stress can alter the regulation of SMN exon 7 splicing, we treated cells in culture with physiological hypoxia (1% O2) or normoxia (21% O2). Total RNA was then isolated and subjected to RT-PCR, followed by restriction enzyme digestion to differentiate between transcripts generated by SMN1 and SMN2. A unique DdeI restriction site is present in exon 8 of transcripts from SMN2 that can be used to differentiate spliced product generated from SMN1 and SMN2 (5,31).DdeI digestion of SMN RT-PCR products generates unique products for SMN1 full-length (507 bp), SMN1 exon 7 skipped (453 bp), SMN2 full-length (392 bp) and SMN2 exon 7 skipped (338 bp). A variety of cell lines were tested for hypoxia-induced splicing changes in SMN exon 7: MCF7 (breast cancer), HeLa S3 (spinner-adapted cervical cancer), Weri-1 (retinoblastoma, neuronal), GM03814 (SMN1+/− carrier SMA fibroblasts) and GM03813 (SMN1−/− type I SMA fibroblasts) were treated for 24 h under normoxic (21% oxygen) or hypoxic (1% oxygen) conditions. RT-PCR analysis of endogenous SMN1 and SMN2 exon 7 splicing was evaluated by agarose electrophoresis and densitometry. Only a minor increase in SMN1 exon 7 skipping was observed and was not significantly changed to warrant further evaluation; however, all cell types showed an increase in SMN2 exon 7 skipping after 24 h hypoxia, though the basal level and induction of exon 7 skipping varied between cell types (Fig. 1A). Radioactive low-cycle RT-PCR was performed to confirm that the hypoxia-induced skipping of SMN2 exon 7 seen in the RT-PCR of SMN splicing was consistently in the linear range of the reaction (Supplementary Material, Fig. S1). The most significant changes were observed in HeLa S3 and MCF7 cells, and whereas the SMN2 transcript showed altered splicing under hypoxia treatment, the stress-responsive gene p53 did not undergo alternative splicing (data not shown).
The two cell lines that showed the greatest amount of hypoxia-induced SMN2 exon 7 skipping were further evaluated for the timing of hypoxia-induced SMN2 exon 7 skipping. HeLa S3 and MCF7 cells were cultured under normoxia (24 h) or hypoxia for 2, 6, 12, 18 and 24 h. RNA was isolated for RT-PCR to evaluate SMN2 exon 7 skipping. HeLa S3 cells exhibited increased SMN2 exon 7 skipping within 6 h following hypoxia treatment (Fig. 1B), and MCF7 cells also induced SMN2 exon 7 skipping within 6 h following hypoxia treatment and showed a statistically significant induction within 12 h (P = 0.0355, two-tailed t-test; n = 2) (data not shown). Hypoxia induction was verified by RT-PCR for the expression of the hypoxia-inducible VEGF165 transcript, which was increased within 12 h of hypoxia treatment in HeLa S3 cells (Fig. 1B). Hypoxia induction as observed by VEGF165 expression correlates well with hypoxia-induced skipping of SMN2 exon 7 skipping. Furthermore, the skipping of SMN2 exon 7 increases and persists during the duration of hypoxia treatment to 24 h.
Hypoxia leads to reduced SMN protein levels in cell culture
The induced skipping of SMN2 exon 7 caused by hypoxia treatment would predict a reduction of total SMN protein. To evaluate the impact of hypoxia treatment on levels of SMN protein, we treated HeLa S3, SMA carrier (SMN1+/−) fibroblasts and SMA patient (SMN1−/−) fibroblasts for 48 h under normoxia or hypoxia. Total RNA was isolated from the three cell types to verify sustained induction of SMN2 exon 7 skipping. Consistent with the 24 h treated data, HeLa S3 cells, SMA carrier (SMN1+/−) fibroblasts and SMA patient (SMN1−/−) fibroblasts all retained increased levels of SMN2 exon 7 skipping in hypoxia-treated samples (Fig. 2A). Whole-cell protein lysates were isolated from 48 h treated cell lines and analyzed by western blot for SMN protein levels normalized to the internal control Actin. Following 48 h of hypoxia treatment, a reduction of SMN protein was observed in HeLa S3 cells, SMA carrier (SMN1+/−) fibroblasts and SMA patient (SMN1−/−) fibroblasts (Fig. 2B). Therefore, hypoxia-induced skipping of SMN2 exon 7 is sustained over an extended treatment of hypoxia, and the induced skipping leads to a reduction in the level of SMN protein in cell culture.
SMNΔ7 mice exhibit improved motor function following hyperoxia treatment
Global analysis utilizing gene expression and splicing-sensitive microarrays was performed previously to identify gene expression and splicing changes that occur during disease progression in severe SMA mouse models. From these analyses, induction of stress-responsive genes and perturbations in splicing were observed late in disease progression, but not in the presymptomatic SMA mice (25–27). The induction of hypoxia-induced stress-responsive genes such as Hif3alpha, which has both increased expression and exhibits altered splicing late in disease, implicates hypoxia in disease progression in the SMNΔ7 severe mouse model of SMA (25). As severe SMA patients also develop respiratory deficiency during disease progression, we sought to evaluate whether hypoxia contributes to SMA disease by altering SMN2 exon 7 splicing and whether increased oxygenation could modulate disease severity in a severe SMA mouse model. We used the SMNΔ7 SMA mouse model that produces litters of SMA pups that are null for the Smn (−/−) gene and non-diseased control littermates (non-SMA) that express the Smn wild-type allele and do not develop SMA. The SMNΔ7 mice carry both the SMN2 BAC transgene and the SMNΔ7 transgene homozygously, which allows for the analysis of SMN2 splicing in the presence and absence of disease (32). We first analyzed SMN2 exon 7 skipping from the SMN2 BAC transgene by radioactive RT-PCR in neuronal tissue (brain) and skeletal muscle (quadriceps and gastrocnemius) from postnatal day (PND) 10.5 SMNΔ7 SMA and non-SMA control mice. Consistent with previous analysis of spinal cord and spinal cord motor neurons (33,34), we observed an increase in SMN2 exon 7 skipping in SMA mice in both brain and skeletal muscles when compared with control mice (Fig. 3A). To assess the levels of the hypoxia-inducible transcript, Hif3alpha, we analyzed brain tissue from control and SMA mice before disease onset (PND5.5) and at disease end-stage (PND13.5–14.5). The levels of Hif3alpha showed no difference in the brain of SMNΔ7 non-SMA control and SMA mice at the presymptomatic (PND5.5) time point as assayed by RT-PCR. The expression of Hif3alpha was low at PND5.5, which required a higher cycle number to assess transcript levels when compared with samples from PND13.5–14.5 mice. However, in the brain of PND13.5–14.5 SMNΔ7 SMA mice, higher levels of Hif3alpha expression were observed when compared with the non-SMA controls (Fig. 3B). These data are in agreement with previously reported elevation of Hif3alpha expression in the end stage of diseased- SMNΔ7 mice as well as the most severe Line89 mice (25,35).
To address the effect of oxygenation on disease progression in severe SMA mice, we housed SMNΔ7 mice under normoxia (∼21% O2), hypobaric hypoxia (∼12% O2) or hyperoxia (∼50% O2) starting on PND5.5. The hypoxia level utilized for cell culture experiments is too extreme for use in mice, thus hypobaric hypoxia is utilized to approximate low oxygen levels (36). Mice were analyzed for motor function and weight daily during a 1 h period in which the mice were removed from their respective chambers/cages. When comparing the weights of the SMNΔ7 control non-SMA mice, we found that there was no difference between the weight gain of the pups under normoxia or hyperoxia; however, hypoxia treatment impaired pup weight gain. This observation was even more pronounced in the SMNΔ7 SMA mice, as the SMA mice were dramatically smaller compared with normoxia- or hyperoxia-treated SMA pups (Supplementary Material, Fig. S2A).
Survival was measured for each treatment group but no significant change in overall survival was observed [average survival: normoxia 13.36 ± 4.706 (n = 7), hyperoxia 14.75 ± 1.909 (n = 8), hypoxia 13.83 ± 2.582 (n = 6), mean survival and standard deviation in days]. To address the impact of hypoxia and hyperoxia treatment on motor function, we assessed the pups using the complex motor function righting reflex assay. Pups were placed on their back, and the time to righting was measured as the time to rotate onto their ventral side. The SMNΔ7 non-SMA control littermates were able to perform this act within 1 s by PND8.5 irrespective of treatment, whereas SMA mice had a delay in time to righting. We hypothesized that hypoxia would negatively affect motor function, whereas hyperoxia treatment would improve motor function compared with normoxia treated mice. Indeed, the hyperoxia-treated SMNΔ7 SMA mice had a reduced latency to righting, indicating an improvement in motor function under hyperoxia treatment (Fig. 3C). However, we were unable to accurately compare the hypoxia-treated SMA mice with the normoxia- and hyperoxia-treated SMA mice due to the severe growth deficit of the SMA mice under hypoxia treatment (Supplementary Material, Fig. S2A).
To evaluate the impact of the level of oxygen on splicing of SMN2 exon 7 in the SMNΔ7 mice, we isolated tissues from normoxia-, hypoxia- and hyperoxia-treated SMNΔ7 mice. RNA was isolated from PND10.5 SMNΔ7 SMA mice, and radioactive RT-PCR was performed to evaluate SMN2 exon 7 skipping, which can be evaluated from the SMN2 BAC transgene present in both the SMNΔ7 SMA and non-SMA control mice. In the brain, the level of SMN2 exon 7 skipping did not differ between the three treatment groups (normoxia average skipping 74.97 ± 8.76%, n = 3; hypoxia average skipping 78.76 ± 3.25%, n = 3; and hyperoxia 77.54 ± 4.12% n = 5) and this was recapitulated in the spinal cord of these mice (normoxia average skipping 61.93 ± 16.7%, n = 3; hypoxia average skipping 68.69 ± 3.96%, n = 3; and hyperoxia 71.49 ± 2.65% n = 5) (Supplementary Material, Fig. S2B). However, skeletal muscles isolated from the three treatment groups exhibited trends in SMN2 exon 7 skipping consistent with the anticipated treatment. Hyperoxia treatment improved motor function and increased exon 7 inclusion, whereas hypoxia treatment increased exon 7 skipping (Fig. 3D and Supplementary Material, Fig. S2B). As expected, the greatest difference in SMN2 exon 7 skipping was between the hypoxia- and hyperoxia-treated SMA mice. We assessed SMN2 exon 7 splicing in the SMNΔ7 non-SMA control mice that also have the SMN2 BAC transgene, and we did not see a similar change in exon 7 skipping (data not shown).
To determine whether the change in SMN2 exon 7 splicing changed the overall SMN protein levels, we performed western blot analysis of the muscle from mice in these three groups, which showed the most change in splicing. The relative levels of SMN protein were not substantially different between the treatment groups which may in part be due to low levels of SMN in muscle or the moderate change in splicing (data not shown). However, hyperoxia treatment of the SMNΔ7 SMA mice improved motor function, and a modest increase in SMN2 exon 7 inclusion in muscle tissue was observed.
SMN minigenes recapitulate hypoxia-induced exon 7 skipping
Understanding the splicing regulatory factors and elements involved in exon 7 splicing provides for directed correction of the splicing defect of SMN2 as a therapeutic approach for the treatment of SMA (reviewed in 37). In order to evaluate the regulatory elements important for hypoxia-induced skipping, we first sought to identify a hypoxia-inducible SMN minigene for cell transfection assays. We utilized the previously characterized SMNx minigenes comprised of exon 6, a truncated intron 6 containing 5′ and 3′ ends of intron 6 and the entire sequence of exon 7 through the end of exon 8 (10).
HeLa S3 cells were transfected with GFP (to evaluate transfection efficiency and serve as a negative control for RT-PCR amplification), SMNx-wt (SMN1) or SMNx-6ct (approximating SMN2 by the presence of the exon 7 C>T mutation). Transfected cells were split over two plates within 24 h after transfection and cultured under normoxic or hypoxic conditions for an additional 24 h. Radioactive RT-PCR utilizing minigene-specific primers was performed and RT-PCR products were resolved on a 6% polyacrylamide denaturing gel. The percentage of exon 7 skipping from the SMNx minigenes was in agreement with the anticipated splicing for SMNx-wt and SMNx-6ct, where more exon 7 skipping was observed in the C>T mutant (Fig. 4A, normoxia lanes). Moreover, a trend of hypoxia-induced skipping of the SMNx-wt minigene and a statistically significant induction of SMNx-6ct exon 7 skipping was observed (Fig. 4A). These results recapitulate our observations from the endogenous SMN transcripts as SMN1 showed low levels of hypoxia-induced skipping (though not quantifiable off endogenous transcripts), and SMN2 is capable of a greater response to hypoxic stress in terms of increased exon 7 skipping. The SMNx minigene can thus serve as a model for hypoxia-induced skipping of exon 7 and allow for mutational analysis of regulatory elements within the pre-mRNA to identify regulatory elements that make SMN2 susceptible to hypoxia-induced skipping.
Hypoxia-induced skipping of exon 7 is due to poor exon definition
The SMNx minigenes were generated from the SMN1 gene and the only difference between the SMNx-wt (SMN1) and SMNx-6ct (SMN2) minigenes is the presence of the C>T mutation. Thus, the increased induction of exon 7 skipping under hypoxia in SMN2 is likely due to reduced exon definition by the C>T mutation, which alters splicing regulatory elements. To evaluate the significance of the C>T mutation, we utilized the SMNx-6/11 minigene that has the C6T mutation and compensatory mutation A11G (SMNc.845A > G) in exon 7, which reintroduces the SF2/ASF-binding site (10). We utilized SMNx-6ct as a control to verify hypoxia-induced skipping of exon 7 in paired observations of the SMNx-6/11 minigene transfections in HeLa S3 cells. The SMNx-6/11 minigene exhibited increased inclusion of exon 7 under normoxia, and hypoxia failed to induce exon 7 skipping. The inability of hypoxia to increase exon 7 skipping underscores the importance of the SF2/ASF-binding site in preventing hypoxia-induced exon 7 skipping. The primary impact of the SF2/ASF ESE is an improvement in exon 7 definition which is innately poor due to the suboptimal 5′ss. To evaluate whether exon definition independent of reintroducing SF2/ASF ESE could prevent hypoxia-induced exon 7 skipping, we strengthened the 5′ss of the SMNx-6ct minigene by mutation of the last nucleotide of exon 7 from A>G (SMNc.888A>G) (SMNx-5ssA>G) (8). Strengthening the 5′ss improved exon 7 inclusion resulting in nearly all full-length transcripts in both normoxia and hypoxia, and prevented exon 7 skipping under hypoxia treatment. Therefore, hypoxia-induced skipping of SMN exon 7 is due to weak exon definition due to the C>T mutation and suboptimal endogenous 5′ss. However, the SMNx-wt minigene also has the same suboptimal 5′ss and still showed a trend toward hypoxia-induced exon 7 skipping. Therefore, there may be regulatory elements in addition to the C>T mutation that contribute to the hypoxia-induced skipping of exon 7 in the SMN2 transcripts.
Mutation of hnRNPA1-binding sites prevent hypoxia-induced exon 7 skipping of SMN minigenes
Many of the regulatory cis elements and trans factors that govern the splicing of SMN exon 7 have been identified (recently reviewed in 37,38). To identify the additional regulatory elements that modulate hypoxia-induced SMN2 exon 7 skipping, we first evaluated the relative levels of several known exon 7 regulatory splicing proteins following 48 h of hypoxia treatment in HeLa S3 cells. There was minimal change in the positively acting splicing regulators SF2/ASF and Tra2Beta under hypoxia treatment. Conversely, a significant increase in the negatively acting regulatory proteins, hnRNP A1 and Sam68, was observed (Fig. 5A and Supplementary Material,Fig. S3A). We also evaluated the levels of hnRNP A1 and Sam68 protein in fibroblasts isolated from human SMA patients, wherein an increase in hnRNP A1 was observed under hypoxia (1.38 ± 0.134-fold, P = 0.0195, Supplementary Material, Fig. S3B), but no change was seen for Sam68 (Supplementary Material, Fig. S3B). We additionally assessed the levels of hnRNP A1 and Sam68 in the muscle of SMNΔ7 SMA mice from the normoxia and hypoxia groups. We did not see a statistically different change in either protein upon comparing normal with hypoxia-treated SMA mice (Supplementary Material,Fig. S3B), indicating that the function of these proteins could be altered by mechanisms in addition to changes in protein level.
The increase in hnRNP A1 and Sam68 following hypoxia treatment in HeLa S3 cells, both of which bind the C>T point mutation of SMN2, may explain the sensitivity of SMN2 to increased exon 7 skipping under hypoxia. Additional hnRNPA1-binding sites in the SMN pre-mRNA reside within exon 7 and its downstream intron and may also play a role in hypoxia-induced regulation of SMN splicing. An hnRNP A1-binding site in exon 7 was recently identified by a G25C (SMNc.859G>C) patient mutation in exon 7 that disrupts the hnRNP A1 site and increases SMN2 exon 7 inclusion (17,39). The G25C patient mutation in SMN2 was first predicted to generate an SF2/ASF-binding site (39). However, subsequent analysis by RNA-binding assays showed that the G25C mutation actually led to reduced binding of hnRNP A1, further supporting the role of hnRNP A1 as a negative regulator of exon 7 splicing in the normal SMN2 gene (G25) (17). hnRNP A1-binding sites in intron 7 include two sites in the potent intronic silencer element ISS-N1, and an SMN2 exclusive binding site at position 100 of intron 7 due to an A>G mutation (SMNc.888+100A>G) (18,19,24) (Fig. 5B). Additionally, it has previously been shown that blocking ISS-N1 function by mutation or ASOs dramatically increases exon 7 inclusion and is a primary target of splicing correction in pre-clinical therapeutic development for SMA models (20–23).
To evaluate the role of hnRNP A1 splicing regulatory elements in hypoxia-induced SMN2 exon 7 skipping, we performed mutational analysis of hnRNP A1-binding sites in the hypoxia-responsive SMNx-6ct minigene. The hnRNP A1 site at position 100 of intron 7 is not present in the SMNx-6ct minigene as it was generated from the SMN1 gene, which lacks the A>G mutation required to generate the hnRNP A1-binding site. To evaluate the other intron 7 hnRNP A1-binding sites, the SMNx-6ct minigene was mutagenized to introduce point mutations that disrupt the two weak hnRNP A1-binding sites in ISS-N1, mutations termed A12C (SMNc.888+12A>C), A23C (SMNc.888+23A>C), and double-mutant 2A-2C (SMNc.888+12A>C; +23A>C) (18). The ISS-N1 mutant SMNx-6ct minigenes were transfected into HeLa S3 cells and cultured for 24 h under normoxia or hypoxia conditions. The percentage of exon 7 skipping under normoxia and hypoxia was compared with the SMNx-6ct minigene transfections under normoxia and hypoxia. The A12C mutant and A23C mutant both improved exon 7 inclusion in the normoxia samples as predicted by mutation of the ISS-N1 silencer (18). However, under hypoxia treatment, neither the A12C nor the A23C mutant prevented hypoxia-induced skipping of exon 7 (Fig. 5C).
We further evaluated the ISS-N1 hnRNP A1-binding sites by generating a double-mutant 2A-2C minigene to test hypoxia-induced skipping of exon 7. The ISS-N1 double-mutant 2A-2C further improved exon 7 inclusion under normoxia, but the 2A-2C mutant also failed to prevent hypoxia-induced skipping (Fig. 5D). All three mutant hnRNP A1 ISS-N1 minigenes retained the capacity to increase exon 7 skipping after hypoxia treatment, and this argues that ISS-N1 disruption alone is not sufficient to prevent hypoxia-induced skipping of exon 7. Thus, although elevated levels of hnRNP A1 under hypoxia may correlate with induced exon 7 skipping, their function is not solely through the splicing silencer ISS-N1.
We extended our mutational analysis of hnRNP A1-binding sites to include the exonic splicing regulatory element that resides between nucleotides 16 and 39 of exon 7, which is independent of the splicing regulatory element disrupted by the C>T mutation (17). We generated SMNx-6ct minigenes containing the G25C mutation alone (SMNx-G25C) or in combination with the ISS-N1 double-mutant 2A-2C (SMNx- G25C 2A-2C) using the SMNx-6ct and SMNx-2A-2C minigenes for site-directed mutagenesis, respectively. The SMNx-G25C mutant minigene under normoxia increased exon 7 inclusion but retained hypoxia-induced skipping of exon 7 (Fig. 5E). Therefore, the exonic hnRNP A1 site was also insufficient to prevent hypoxia-induced skipping of exon 7 when mutated alone. However, when the exonic hnRNP A1 site and intronic ISS-N1 hnRNP A1 sites were mutated in combination, the SMNx-G25C 2A-2C minigene showed a decrease in exon 7 skipping under normoxia that prevented hypoxia-induced skipping of exon 7 under hypoxia (Fig. 5E). Thus, although independent mutation of hnRNP A1-binding sites did not prevent hypoxia-induced skipping of exon 7, the disruption of both exonic and intronic hnRNP A1-binding sites resulted in an SMNx minigene that no longer induced skipping of exon 7 under hypoxia. These data implicate a functional role of multiple hnRNP A1-binding sites, both exonic and intronic, in sensitizing exon 7 to hypoxia-induced skipping.
hnRNP A1 and Sam68 bind to regulatory elements in SMN necessary for hypoxia-induced skipping
As both exonic and intronic hnRNP A1-binding sites were found to be necessary for hypoxia-induced skipping of exon 7, we evaluated these hnRNP A1-binding sites for increased binding under hypoxia treatment. Additionally, both hnRNP A1 and Sam68 are upregulated in response to hypoxia. Therefore, we sought to evaluate whether Sam68 also bound to either of these hnRNP A1 sites and whether binding of Sam68 and/or hnRNP A1 is enhanced under hypoxic conditions.
To test the binding of hnRNP A1 and Sam68 regulatory splicing factors, we generated splicing-competent nuclear extracts from HeLa S3 cells treated under normoxia or hypoxia (1% O2) for 12 h for in vitro RNA-binding assays (as previously described in 40). RNA oligos were generated for the hnRNP A1-binding sites in the exon [wild-type (G25) and mutant (G25C)] and the intron (wild-type ISS-N1 and double-mutant 2A-2C) (17,18). The RNA oligos were conjugated to agarose beads and incubated in the normoxia and hypoxia nuclear extracts under splicing conditions. In vitro binding to the RNA oligos in HeLa nuclear extracts was assessed by western blot for the binding of hnRNP A1 and Sam68. We first verified that hnRNP A1 bound to these regulatory elements in the normal nuclear extract and that binding was reduced in the case the exonic G25C mutation or nearly lost in the double-mutant ISS-N1 2A-2C. These data are in agreement with the previously reported binding of hnRNP A1 to these regulatory elements (17,18). RNA affinity pulldowns using the hypoxia nuclear extract showed increased binding of hnRNP A1 to both the exonic (ESS-A1S) and intronic (ISS-N1) elements, which is congruent with the role of hnRNP A1 in increasing exon 7 skipping under hypoxia (Fig. 6). To evaluate the potential binding of Sam68 to the hnRNP A1-binding sites, we performed western blots of Sam68 using the in vitro RNA-binding assay samples. Sam68 was bound to the RNA of the exonic element in both normal and hypoxia nuclear extracts, whereas no appreciable binding of Sam68 was identified on ISS-N1 (Fig. 6). hnRNP A1 and Sam68 have previously been shown to also bind to the C>T mutation in SMN2 exon 7 (11,13). In addition to the C>T mutation, both hnRNP A1 and Sam68 bind to the additional ESS (ESS-A1S) in exon 7. The increased levels of hnRNP A1 and Sam68 under hypoxia are congruent with the function of these two regulatory splicing factors binding to and promoting skipping of SMN2 exon 7. Furthermore, as hnRNP A1 shows increased binding of both the G25 and ISS-N1 regulatory elements under hypoxia, we propose that hypoxia induces a region of splicing repression that prevents exon 7 definition and increases skipping under hypoxia.
SMA patients and mouse models of SMA are associated with respiratory deficiency, in which patients can be addressed by supportive breathing intervention. The success of non-invasive respiratory support in severe SMA patients can result in survival extension (30). In the SMNΔ7 severe SMA mouse model, induction of stress genes such as the hypoxia-inducible factor Hif3alpha is observed during the progression of disease. In this paper, we have shown that SMN2 exon 7 skipping is increased under hypoxic stress, and following chronic exposure to hypoxia, SMN protein levels are subsequently reduced. The splicing change is observed in a variety of cell types to varying degrees, including neuronal cells and SMA patient fibroblasts. Differential splicing ratios have been reported in SMA patient blood and muscle samples and may indicate cell or tissue type changes in splicing efficiency of SMN2 exon 7 (41). Furthermore, the reduction of SMN protein is evident in SMA patient fibroblasts. SMA is caused by low levels of SMN protein, and reduced SMN levels are also associated with increased exon 7 skipping via a negative feedback loop (33,34). The primary function of SMN is the maturation of snRNPs, which form the spliceosome and catalyze the splicing reaction (42–44). These observations of increased skipping in SMN-depleted cells and SMA mouse models predict that reduced SMN protein reduces splicing due to lower levels of functional splicesomal complexes (33,34). Here we have shown that hypoxia treatment also reduces SMN protein levels, and therefore the skipping of SMN2 exon 7 may, in part, be due to reduced levels of functional splicing complexes. We also observe an increase in SMN2 exon 7 skipping in the SMNΔ7 SMA mice compared with non-SMA controls, both in the brain and in skeletal muscles, consistent with recent reports of increased skipping of SMN2 exon 7 in spinal cord and spinal cord motor neurons in SMA mouse models (33,34). Motor neurons are the primary cell type affected in SMA and were also shown to have a higher level of exon 7 skipping compared with non-motor neurons in SMNΔ7 mice (34). The increase in exon 7 skipping of motor neurons may explain the selective vulnerability of motor neurons in SMA, and hypoxic stress may further exacerbate exon 7 skipping during SMA disease progression (34). Our observations provide for the modulation of SMN2 exon 7 splicing under hypoxia, however, the capacity to induce exon 7 skipping may not be unique to hypoxia. Additional stresses associated with neuromuscular diseases such as excitotoxicity or oxidative stress (45) may also modulate stress-induced skipping of SMN2 exon 7, especially as disease symptoms progress (25–27).
Respiratory support is one of the most common treatment options for severe SMA patients as respiratory deficiencies increase with disease progression. Indeed, respiratory assistance has a large impact on survival extension in SMA patients (30). Sleep disordered breathing in SMA patients has also been reported and can include nocturnal hypoxemia (low pulse oximetry) and hypercapnia (high transcutaneous CO2 levels) (46). Furthermore, disordered breathing during sleep can be improved or eliminated in SMA patients by the use of non-invasive ventilation during sleep. The improvement of nocturnal breathing disorders was also associated with improvement in daytime indicators of respiratory function and improvement in the quality of sleep that translated into improved quality of life (47). Therefore, as indicators of daytime respiratory deficiencies are becoming more defined, they may serve as criteria to identify SMA patients who could benefit from nocturnal ventilation to improve quality of life and reduce nocturnal breathing disorders and hypoxemia (46). We show that hyperoxia treatment was able to improve motor function in the severe SMNΔ7 SMA mice compared with normoxia treatment.
Furthermore, a trend of increased SMN2 exon 7 inclusion was observed under hyperoxia treatment, whereas hypoxia treatment increased skipping. These splicing changes were observed in the skeletal muscles of the SMNΔ7 SMA mice but not in the CNS (brain or spinal cord). This may, in part, be explained by the physiological response to periods of hypoxia, wherein the brain increases oxygen tension at the expense of peripheral tissues (48). Conversely, we cannot rule out the possibility that the absence of splicing changes may be associated with an inherent reduced sensitivity to oxygenation changes in the CNS at the level of SMN2 exon 7 splicing. However, the level of hypoxia and hyperoxia that can be achieved in mice is less than that used in cell culture, and the marginal changes in splicing in the hypoxia and hyperoxia treatment groups may be due to the already high degree of exon 7 skipping observed in the late-stage diseased SMA mice at PND10.5 (72–82% exon 7 skipped, Fig. 3A). It is of note that survival extension in the SMNΔ7 mice does not compare to the dramatic survival extension seen in severe SMA patients that receive supportive breathing intervention. This may be a technical limitation of the experiment as the mice were treated by alterations to the oxygen level rather than providing assisted breathing, which is the conventional means of respiratory assistance for SMA patients by way of BiPAP. The respiratory failure in the SMA mice cannot therefore be overcome by oxygen supplementation alone.
The functional improvement reported here may only indicate in part the potential impact of oxygenation/respiratory support in the treatment of SMA by splicing modulation of SMN2. Overall, the improvement in motor function correlates with increased exon 7 inclusion and demonstrates that oxygen treatment can modulate disease progression in severe SMA mice. Additionally, the benefits of oxygen treatment in the SMNΔ7 SMA mice outside of the change in SMN exon 7 splicing cannot be ruled out and may contribute to the phenotype observed. The corresponding in vitro data supporting the in vivo observations implicate hypoxia as a consideration in the splicing of the SMN gene and in the SMA disease.
The interplay between regulatory elements and trans-splicing factors provides for a wealth of regulatory modulation of SMN exon 7 splicing. Using SMN minigenes that recapitulate exon 7 skipping, we show that hypoxia-induced skipping is due to poor exon 7 definition, which can be prevented by correcting the C>T mutation or improving splice site strength. By evaluating the relative level of exon 7 splicing regulatory factors under hypoxia, we found an increase in inhibitory splicing factors that bind at the SMN2 C>T mutation, hnRNP A1 and Sam68. Mutation of hnRNP A1-binding sites in exon 7 and intron 7 prevented hypoxia-induced skipping. The significance of the hnRNP A1 sites in regulating hypoxia-induced skipping of exon 7 can be explained by two possible mechanisms. The first is that the multiple hnRNP A1-binding sites in exon 7 and intron 7 sequences produce a general region of splicing inhibition that is strengthened under hypoxia by an increase in hnRNP A1 levels. The alternative mechanism still implicates hnRNP A1 sites in sensitizing exon 7 to skipping; however, when multiple hnRNP A1 sites were mutated, the overall efficiency of exon 7 splicing improved as marked by increased exon 7 inclusion. As the sensitivity of exon 7 to hypoxia-induced skipping is dependent upon poor exon definition, the failure to induce skipping under hypoxia may be due to improved exon definition. However, as the endogenous SMN gene has the hnRNP A1-binding sites intact, the change in the relative levels of regulatory splicing factors in response to hypoxia or other stresses could permit differential splicing efficiency of exon 7. To test these potential mechanisms, we performed RNA-binding assays of the hnRNP A1-binding sites necessary for hypoxia-induced skipping and found increased binding of hnRNP A1 on both sites under hypoxia, as well as binding of Sam68 to the exonic G25 element under normoxia and hypoxia. These data underscore the importance of hnRNP A1 and Sam68 in regulating hypoxia-induced skipping of SMN exon 7. As both hnRNP A1 and Sam68 bind to the regulatory elements required for hypoxia-induced skipping, we propose that increased binding of hnRNP A1 under hypoxia in combination with Sam68 binding prevents exon 7 definition and increases exon 7 skipping. In this paper, we have focused on the importance of both hnRNP A1 and Sam68 on the splicing of SMN2 exon 7. As both hnRNP A1 and Sam68 relocalize to stress granules under both, p38-induced stress and oxidative stress, they may have additional functions in RNA metabolism that is important in hypoxic stress (49,50). The impact of altering the repertoire of exon 7 regulatory splicing factors under hypoxia, by changes in level or localization, may include additional regulatory splicing factors and thus the evaluation of the impact of other exon 7 regulatory splicing factors that are changed under hypoxia is warranted to understand the interplay of these splicing factors on stress-mediated splicing of SMN2 exon 7 and to identify a novel intervention point for SMA therapy.
In summary, our results indicate that hypoxia can modulate the splicing of SMN2 exon 7, and treatment with hyperoxia in severe SMA mice can improve motor function and increase SMN2 exon 7 inclusion. Respiratory care is a major point of intervention in SMA and has shown dramatic effect in extending survival in type I SMA patients. The therapeutic benefit of respiratory assistance in the treatment of SMA may encompass preventing periods of low oxygenation that would otherwise increase SMN2 exon 7 skipping and reduce SMN levels. Early treatment in pre-clinical therapy development has been shown to have greatest impact on survival extension; the corollary for early treatment by way of non-invasive respiratory care may also improve survival extension and quality of life (20–23,47,51,52). Daytime indicators of nocturnal hypoxemia and hypercapnia may serve as measures to include SMA patients in earlier respiratory support and therefore improve quality of life and/or survival.
MATERIALS AND METHODS
Cell culture and transfections
Cell lines were cultured according to manufacturers' instructions. All cells were grown with 10% fetal bovine serum (HyClone, Thermo) and antibiotic mixture of penicillin–streptomycin (Cellgro) unless otherwise listed. MCF7 cells were grown in DMEM (Gibco), HeLa S3 cells were grown in RPMI-1640 (Cellgro) and Weri-1 cells were grown in RPMI-HEPES (Gibco). SMA type I patient fibroblasts GM03813 (SMN1−/−) and carrier fibroblasts GM03814 (SMN1+/−) were grown in AMEM (Cellgro), 15% fetal bovine serum, penicillin/streptomycin. Cells grown under normoxia conditions (21% O2) were grown at 37°C, 5% CO2, in a humidified incubator. Cells grown under hypoxic conditions (1% O2) were cultured at 37°C, 5% CO2, in a humidified incubator (IN VIVO2, Baker Biosystems). In all cell culture experiments, cells were seeded to ensure subconfluent growth at the time of harvesting to reduce other stresses (overcrowding, nutrient deprivation, etc.). HeLa S3 cells were seeded in 6 cm plates and transfected at ∼60–70% confluency with SMN minigenes. The SMN1 (SMNx-wt), SMN2 (SMNx-6ct) and C>T compensatory mutant (SMNx-6/11) minigenes were provided by Dr Luca Cartegni. The SMNx-wt, SMNx-6ct and SMNx-6/11 minigenes differ at the C6T mutation or at A11G compensatory mutation, and all contain a C14U mutation in intron 7, which does not alter splicing ratios (18). ISS-N1 mutations A12C, A23C or double-mutant 2A-2C were generated by site-directed mutagenesis (Agilent Technologies) (mutagenesis primers in Supplementary Material, Table S2). Transfected cells were split over two 6 cm plates at 20–22 h after transfection and cultured under normoxia or hypoxia (1% O2) at 24 h after transfection for an additional 24 h. Minigene-specific radioactive RT-PCR was performed using T4 polynucleotide kinase (Roche), 32P-radiolabeled T7 forward primer 5′-TAATACGACTCACTATAGG-3′ and exon 8 reverse primer hSMN X8 RV 5′-ACCGAATTCCACATACGCCTCACATACA-3′. RT-PCR (Platinum Taq, Invitrogen) products (full-length, 668 nt, and skipped, 614 nt) were separated on a 6% polyacrylamide/8 m urea denaturing gel and quantified by densitometry (ImageQuant TL, GE Healthcare Life Sciences).
Total RNA was isolated from tissue culture samples using RNeasy (Qiagen), and cDNA (Transcriptor RT, Roche) was generated from 1–5 μg of total RNA using random hexamers (Roche). RT-PCR (Sigma) to detect exon 7 splicing of SMN1 and SMN2 transcripts was performed using primers in exon 6 (hSMN X6s3) 5′-CCCCCACCACCTCCCATATG-3′ and exon 8 (hSMN X8as) 5′-CCCTTCTCACAGCTCATAAAATTAC-3′ at the following conditions: annealing temperature 62°C, 30 s; 72°C extension, 2 min—for 35 cycles. To differentiate between SMN1 and SMN2 transcripts, RT-PCR products were digested with the SMN2 exon 8-specific DdeI restriction site. Digested RT-PCR products were resolved on 2.5% agarose gels and product sizes identified (SMN1 full-length: 507 bp, SMN1 skipped: 453 bp, SMN2 full-length: 392 bp, SMN2 skipped: 338 bp). Endogenous human p53 exon 7 splicing was evaluated by RT-PCR by nested PCR at an annealing temperature of 55°C; the first primer pair was Hp53 X6F BH 5′-CTGGCCCCTCCTCAGCATCTTATCC-3′ and Hp53exas 5′-CTGAAGGGTGAAATATTCTCCATCC-3′ and the second primer pair was Hp53 X6s 5′-GTGGTGGTGCCCTATGAGCCG-3′ and Hp53 X9as 5′-CTCCATCCAGTGGTTTCTTCTTTG-3′ with full-length (331 bp) and skipped (221 bp) PCR products. Splicing ratios were determined by densitometry (Image Quant TL, GE Healthcare Life Sciences), and statistical analysis (two-tailed t-test) was performed using the Graphpad Prism software. RT-PCR conditions of Hif3alpha, Actin, 18S rRNA and VEGF can be found in Supplementary Material, Table S1.
Total protein lysates were isolated from tissue culture samples or mouse tissues in RIPA buffer. Protein samples were run on 10% SDS–PAGE gels. Levels of SMN (BD Transduction Laboratories, 1:5000 dilution) and loading control Actin (AC-15) (Sigma Aldrich) at 1:20 000–60 000 were performed for western analysis using a Versa Doc scanner and quantified using the Image Quant TL software (GE Healthcare Life Sciences). The following SMN exon 7 splicing factors were detected by western blotting: SF2/ASF (Zymed, 1:1000 dilution), hnRNP A1 (4B10) (Santa Cruz Biotechnology, 1:5000), Tra2beta (Abcam, 1:1000 dilution), Sam68 (C-20) (Santa Cruz Biotechnology, 1:1000 dilution). Regulatory splicing factor levels were normalized to Actin (AC-15) for SF2/ASF, hnRNP A1 and Sam68, and to beta-tubulin (E7) (hybridoma supernatant, 1:50 dilution) for Tra2Beta.
Mouse treatment and tissue analysis
All experiments were performed according to institute standards and were approved by the Institute Animal Care and Use Committee (IACUC) at The Research Institute at Nationwide Children's Hospital. SMNΔ7 heterozygous carrier mice (Smn+/−, SMN2+/+, SMNΔ7+/+) were obtained from Jackson Laboratories (stock number 005025) (32). The SMNΔ7 heterozygous carrier mice were interbred to generate severe SMA mice and non-SMA control littermates. Neonatal pups were genotyped and litters were culled to eight pups when needed at the time of treatment. The mother and pups were housed under normal, hypobaric hypoxia (450 Torr or ∼12% O2) or hyperoxia (50% O2/50% air mixture) starting on PND5.5. Animals were removed from treatment for 1 h during which weights and motor function were measured by the righting reflex assay. Pups were placed on their backs and time to rotating onto their ventral side was measured for up to 30 s; pups were evaluated five times or up to three failed attempts of 30 s each. Motor function was graphed and analyzed by one-way ANOVA, Newman-Keuls multiple comparisons test (GraphPad Prism). Tissue harvests from mice treated starting on PND5.5 under normoxia, hypoxia or hyperoxia were performed directly after the removal from the treatment chamber or cage, and tissues were snap-frozen on liquid nitrogen. Total RNA was isolated from tissues by homogenization in TRI reagent (Invitrogen) and purification by RNeasy (Qiagen). Total RNA was utilized to make cDNA, using random hexamers (Roche), and radioactive RT-PCR was performed using radiolabeled forward primer (hSMN2 E6 FW) 5′-AGATTCTCTTGATGATGCTGATG-3′ and reverse primer (hSMN2 E8 RV) 5′-TTATATACTTTTAAACATATAGAAGATAG-3′. PCR products of full-length (685 bp) and skipped (631 bp) were run on a 6% polyacrylamide/8 m urea denaturing gel and splicing ratios were determined by densitometry (ImageQuant TL, GE Healthcare Life Sciences).
RNA affinity chromatography
RNA oligonucleotides were purchased from Invitrogen for the hnRNP A1-binding sites termed ESS-A1S (wt) (CAAAAAGAAGGAAGGUGCUCACAU), G25C (mutant) (CAAAAAGAACGAAGGUGCUCACAU), ISSN1 (wt) (CCAGCAUUAUGAAAGU) and ISSN1 2A-2C (mutant) (CCCGCAUUAUGAACGU). Two nanomoles of RNA was suspended in a 400 ml reaction mixture containing 100 mm sodium acetate, pH 5.0, and 5 mm sodium m-periodate (Sigma) for 1 h in the dark at room temperature. After ethanol precipitation, the RNA was resuspended in 50 µl of 0.1 m sodium acetate, pH 5.0. A 200 µl aliquot of adipic acid dihydrazide agarose bead 50% slurry (Sigma) was washed four times in 5 ml of 0.1 m sodium acetate, pH 5.0, and pelleted after each wash at 100 g for 3 min in a clinical centrifuge. After final wash, the beads were resuspended in 150 µl of 0.1 m sodium acetate, pH 5.0. Fifty microliters of RNA was mixed with 150 µl of beads and rotated at 4°C overnight, then pelleted and washed three times in 1 ml of 2 m NaCl and then three times in 1 ml of buffer D (20 mm HEPES-KOH, pH 8.0, 20% glycerol, 0.1 m KCl, 0.2 mm EDTA, 0.5 mm DTT) and resuspended in 125 µl of buffer D. Two 150 µl of in vitro splicing reaction mixtures, normal or hypoxia, were made containing 60 µl of the nuclear extract and 62.5 µl of the RNA bound beads. The reaction was incubated at 30°C for 40 min by gently mixing every 5–10 min, then the protein-bound RNA/beads were washed three times in buffer D with 100 mm KCl. The beads were resuspended in 40 µl of 2× SDS buffer, heated at 100°C for 5 min quickly, spun down, collected and run on a 10% SDS–PAGE. Blots were probed using anti-SF2/ASF (1:500 dilution; Zymed) and anti-hnRNPA1 (1:200 dilution; Abcam). Splicing-competent nuclear extracts were generated from HeLa S3 cells cultured for 12 h at normoxia (∼21% O2) or hypoxia (1% O2) according to standard protocols (53).
Splicing changes are graphically represented as mean and SEM. Change of percent skipping is listed as average change and standard deviation. Graphical representation and statistical analyses were performed using the GraphPad Prism software. Statistical significance levels where identified by asterisks follow normal P-value cutoffs: *P < 0.05, **P < 0.01 and ***P < 0.005.
Funding for this research was provided by The Research Institute at Nationwide Children's Hospital and the Jeffrey J. Seilhamer Cancer Foundation (to T.W.B.); The Association Francaise Contre les Myopathies (AFM; contract grant number 14278 to D.S.C.); NIH (NIH; contract grant number 1R21NS054690 to D.S.C.).
We would like to thank Dr Luca Cartegni, who graciously provided the SMNx minigene constructs, as well as Dr Louis Chicoine, who kindly provided technical assistance and the use of the hypobaric hypoxia and hyperoxia chambers for in vivo mouse experiments. In addition, we would like to thank Chandler Lab members Aishwarya Jacob and Daniel Comiskey for their assistance in reading and preparing the revised manuscript, as well as Casey Gentis for her editing expertise.
Conflict of Interest statement. None declared.