Abstract
Spinal muscular atrophy (SMA) is a neuromuscular disease caused by reduced expression of survival of motor neuron (SMN), a protein expressed in humans by two paralogous genes, SMN1 and SMN2. These genes are nearly identical, except for 10 single-nucleotide differences and a 5-nucleotide insertion in SMN2. SMA is subdivided into four main types, with type I being the most severe. SMN2 copy number is a key positive modifier of the disease, but it is not always inversely correlated with clinical severity. We previously reported the c.859G > C variant in SMN2 exon 7 as a positive modifier in several patients. We have now identified A-44G as an additional positive disease modifier, present in a group of patients carrying 3 SMN2 copies but displaying milder clinical phenotypes than other patients with the same SMN2 copy number. One of the three SMN2 copies appears to have been converted from SMN1, but except for the C6T transition, no other changes were detected. Analyzed with minigenes, SMN1C6T displayed a ∼20% increase in exon 7 inclusion, compared to SMN2. Through systematic mutagenesis, we found that the improvement in exon 7 splicing is mainly attributable to the A-44G transition in intron 6. Using RNA-affinity chromatography and mass spectrometry, we further uncovered binding of the RNA-binding protein HuR to the -44 region, where it acts as a splicing repressor. The A-44G change markedly decreases the binding affinity of HuR, resulting in a moderate increase in exon 7 inclusion.
Introduction
Spinal muscular atrophy (SMA) is a devastating inheritable disorder, characterized by loss of spinal cord α-motor neurons, which results in widespread atrophy of skeletal muscles, especially those of the limbs and trunk (1). SMA is caused by reduced expression of survival of motor neuron (SMN), a housekeeping protein, whose best known function is assisting the assembly of spliceosomal U snRNPs and other RNPs (2–4). SMN may have additional important cellular functions, as it physically interacts with many diverse and functionally-distinct proteins (5). SMN is expressed by two paralogous genes, SMN1 and SMN2, which are nearly identical except for 11 sequence differences, including 10 single-nucleotide changes—6 in intron 6, 1 in exon 7, 2 in intron 7, and 1 in the noncoding exon 8—and a 5-nucleotide insertion in SMN2 intron 6 (6). The C6T change in exon 7 from SMN1 to SMN2, though translationally silent, results in predominant skipping of the exon during SMN2 pre-mRNA splicing, giving rise to an unstable, mislocalized, and dysfunctional truncated protein, SMNΔ7 (6–10). The limited amount of full-length SMN expressed by the SMN2 gene is insufficient to compensate for mutations or loss of SMN1 in SMA patients, though it is essential for their survival.
SMA is classified into four main types, based on the clinical severity, maximum function achieved, and age of onset (11,12). Type I SMA is a severe infantile form, also called Werdnig-Hoffmann disease, manifesting generalized muscle weakness and hypotonia during the first three months after birth. Children with the type I form usually die before age two, due to respiratory failure. Type I SMA is sometimes subdivided into Ia, Ib and Ic, with subtype Ia (also known as type 0) being the most severe, with onset at or before birth (12). Type II SMA is an intermediate form, also called Dubowitz disease; affected children are able to sit but are unable to stand or walk unaided. Type III SMA is a milder juvenile form, also called Kugelberg-Welander disease, with onset after age one; affected children are able to walk without support at some point in their life. Type II and Type III SMA are also subdivided into a and b subtypes. Type IV is the adult-onset form of SMA. The copy number of SMN2 is a major factor in determining the clinical severity of the disease (13,14); this correlation is also recapitulated in SMA mouse models (15,16). In general, most patients with type I SMA have 1 or 2 copies of SMN2; most patients with the intermediate type II form have 3 copies of SMN2; and most patients with the type III form have 3–4 copies of the gene. However, there are cases in which the SMN2 copy number is not inversely correlated with clinical severity; such discordance indicates the existence of disease modifiers that significantly contribute to the phenotypic outcome.
Identification of SMA modifiers would not only help to better understand the molecular mechanisms of SMA pathogenesis, but may also suggest new approaches for therapeutic intervention. So far, however, only a few modifiers have been reported in SMA patients. In 2008, Oprea et al. reported that high levels of plastin 3 (PLS3) fully protect a group of SMN1-deleted females carrying 3–4 copies of SMN2 (17). In contrast, their siblings carrying the same SMN1/2 alleles were not spared from the disease, correlating with low expression of PLS3. Interestingly, transgenic expression of PLS3 in a severe SMA mouse model (SMNΔ7) did not ameliorate motor function and lifespan (18). A recent study suggested that the protective effect of PLS3 requires less severe SMA models, in which peripheral tissues are not critically impaired (19). PLS3 promotes axonogenesis, neuromuscular transmission and endocytosis, by elevating the F-actin level (17,19,20). More recently, Neurocalcin delta (NCALD), a negative regulator of endocytosis was identified as a disease modifier in five SMN1-deleted individuals who carried 4 copies of SMN2 but were asymptomatic (21).
PLS3 and NCALD represent modifiers whose mechanisms of action do not rely on the upregulation of SMN expression. Several signaling proteins, such as RHOA, PTEN and CTNNB1 (beta-catenin) have been reported to modulate the phenotypic severity in SMA mouse models (22–25). In particular, depletion of the tumor suppressor PTEN extended the median survival by 3-fold in the severe SMNΔ7 model (24). Proteins involved in these signaling pathways can be considered potential SMN-independent modifiers in patients.
The SMN2 gene has been the main target for therapeutic intervention to treat SMA. While reagents that substantially increase SMN expression through SMN2, particularly by promoting exon 7 splicing, are potential drugs, other factors that influence SMN levels are potential modifiers of the disease. SMN-dependent modifiers could in principle include, e.g., trans-acting splicing activators or repressors. Many splicing factors have been reported to modulate SMN2 exon 7 splicing; however, to date, none has been identified to naturally modify SMA severity in patients. In a recent study using SMNΔ7 mice, knockdown of SAM68, a splicing repressor of SMN2 exon 7, improved innervation of neuromuscular junctions (NMJs) and extended the mean survival by 5 days, indicating that expression of splicing factors can be manipulated to modify the SMA severity (26).
SMN-dependent modifiers can also be splicing-affecting mutations or SNPs. Though infrequent, mutations seen in patients may disrupt a splicing silencer or fortuitously create a splicing enhancer, resulting in an increase in exon 7 inclusion. We previously showed that the SMN2 c.859G > C variant promotes inclusion of the exon by 20%, and accounts for the much milder phenotype seen in three patients who carry two copies of SMN2 (11). This mutant allele was later found in more chronic patients; haplotype analysis suggests that the mutation originated from a common ancestor (27). ESEfinder3.0 predicts that the c.859G > C substitution creates an SRSF1-dependent splicing enhancer; a subsequent publication reported that a weak hnRNP A1-dependent silencer is disrupted by this mutation (11,28).
In this study, we investigated 53 patients with very mild SMA who carry three copies of SMN2 but did not carry the c.859G > C substitution. Genomic sequencing revealed that one of the three SMN2 copies in 5 patients was apparently converted from SMN1, but the conversion is confined to the C6T transition, with other sequences remaining as the SMN1 version. Using SMN1/2 minigenes comprising all sequence differences in exon 7 and its flanking introns, we compared splicing of SMN1 C6T versus SMN2, and surprisingly found that the former minigene exhibited ∼20% greater extent of exon 7 inclusion in various cell types. To pinpoint which sites account for the difference, we systematically examined various minigenes with combinations of mutations. A-44G in intron 6 emerged as a major positive splicing modifier. We further uncovered that the -44 region is a HuR-dependent splicing silencer; the A-44G transition reduces the binding affinity of HuR, resulting in a moderate improvement of exon 7 splicing. Therefore, we identified a novel mechanism—the A-44G substitution—that modifies SMA phenotype. Considering that most patients with type II or milder forms carry ≥3 copies of SMN2, this mechanism may account for a larger percentage of patients who manifest milder clinical presentations.
Results
SMN1 C6T pre-mRNA splices exon 7 more efficiently than SMN2 pre-mRNA
Patients with 3 copies of SMN2 are found in all types of SMA. As determined by us, ∼6% of type I, ∼81% of type II, and ∼48% of type III patients carry three copies (Prior TW, et al., unpublished data). We are interested in the clinical variability in these patients, particularly those manifesting very mild phenotypes. In this study, we followed a set of 53 patients with 0 copies of SMN1 and 3 copies of SMN2, who were older than 20 years at the time of molecular diagnostic testing. The age range was 25 to 65 years, with a mean of 34.8. These patients were all negative for the 859G > C change, which improves SMN2 exon 7 inclusion and results in a milder phenotype (11). Genomic sequencing revealed one conversion (SMN1 to SMN2) event in 5/53 patients, i.e., in 5 out of 159 total alleles. The ages of these 5 patients were 32, 32, 38, 42 and 45. We compared these findings to a more severe group of 90 SMA patients younger than 6 months at the time of testing. The genotypes of these patients were the following: 7 patients had 1 SMN2 copy, 78 had 2 SMN2 copies, and 5 had 3 SMN2 copies. The population genotypes of the 90 patients are highly consistent with those of type I patients (29–33), and all 178 alleles were negative for the conversion and also for the 859 change.
Since the conversion events occurred only in mild SMA patients, but not in type I SMA patients, we further examined the conversion alleles. Interestingly, we found that all 5 alleles were actually partially converted: except for the C6T change, all other sites that differ between SMN1/2 genes remained as the SMN1 version (Fig. 1A). Although the T at position 6 in exon 7 symbolizes SMN2, and is recognized as the decisive factor in causing exon 7 skipping, the other 10 nucleotide changes have not been thoroughly studied. It is possible that one or more of the other changes in SMN2 moderately inhibit exon 7 inclusion. To explore this possibility, we constructed new SMN1/2 minigenes with a more complete version of intron 6, based on the minigenes we previously reported (34,35). The new minigenes comprise all 10 nucleotide differences that naturally exist in exon 7 and both flanking introns between the two endogenous genes, including G-980A, -908AGGCA (5-nt insertion in SMN2), A-849C, G-549A, T-478C, T-255C, and G-44A in intron 6, C6T in exon 7, and A100G and A215G in intron 7 (Fig. 1B). Exon 7 splicing of the new SMN1/2 minigenes was analyzed after transient transfection into HEK293 cells by fluorescence-labelled RT-PCR (36); we observed similar splicing patterns as those of their old minigene counterparts (Fig. 1C). Surprisingly, however, similar to the c.859G > C variant, the SMN1 C6T mutant (thereafter referred to as SMN1 C6T) displayed a ∼20% (ranging from 17 to 23% in different experiments) increase in exon 7 inclusion compared to SMN2 (Fig. 1C). This finding suggests that the incomplete conversion from SMN1 to SMN2 accounts for the phenotypic discrepancy in milder patients who carry 3 copies of SMN2.
Diagrams of a partially converted SMN2 allele and the new SMN1/2 minigenes, and splicing analysis of the minigenes and SMN1 C6T mutant. (A) Diagram of part (from exon 6 to exon 8) of an allele of SMN2, showing incomplete conversion from SMN1. The natural sequence differences between SMN1 and SMN2 are listed below the diagram. This allele was identified in 5 out of 53 three-SMN2-copy patients with mild clinical phenotypes. (B) Diagram of the new SMN1/2 minigenes cloned in pCI-neo. The new minigenes comprise all the sequences that naturally differ between the two genes except the G236A change in exon 8. The genomic DNA is shown for comparison, and the intron and exon sizes are indicated. (C) Exon 7 splicing analysis of the SMN1/SMN2 minigenes, and their mutants SMN1 C6T and SMN2 c.859G>C. Each plasmid was transfected into HEK293 cells, and splicing was analyzed as described in Materials and Methods. As shown in the gel and histogram, the C6T mutation in exon 7 in the SMN1 minigene setting decreased exon 7 inclusion by 47% (from 97 to 50%), but this is still better than SMN2 (50% versus 29%) and similar to the c.859G>C mutation in the SMN2 minigene setting (50% versus 53%). Inc%: percentage of exon 7 inclusion. **P<0.001 (n = 4).
The A-44G change is responsible for the moderate improvement of exon 7 splicing in SMN1 C6T
To explore which nucleotide difference(s) between SMN1 C6T and SMN2 contribute to the improvement in exon 7 splicing, we used the SMN1 C6T minigene to generate all nine possible double mutants (Fig. 2A). Exon 7 splicing of these mutants was examined in transfected HEK293 cells. Interestingly, the double mutant SMN1 C6T/G-44A displayed a similar level of exon 7 inclusion as the SMN2 minigene (Fig. 2A). The exon 7 inclusion percentage of SMN1 C6T/A100G was also significantly decreased, compared to SMN1 C6T (by ∼ 9%). In contrast to these two double mutants, the others displayed very subtle changes that were not statistically significant.
The G-44A transition accounts for the splicing deterioration in SMN2 compared to SMN1 C6T. (A) Screening of double mutations in the SMN1 C6T setting. The second mutation was either G-980A, Ins5nt (5-nt AGGCA insertion at -908), A-849C, G-549A, T-478C, T-255C, G-44A, A100G, or A215G, and was introduced in the parental plasmid, pCI-SMN1 C6T. SMN1, SMN2, SMN1 C6T, and SMN2 T6C were used as controls. Each plasmid was transfected into HEK293 cells, and the effect of each mutation was analyzed. Two mutations resulted in significant splicing suppression: G-44A in intron 6 reduced exon 7 inclusion by ∼20%, and A100G in intron 7 reduced it by 9%. **P < 0.001 versus SMN1 C6T (n = 4). (B) Mutations analyzed in the SMN1 C6T/G-44A setting. Each mutation, including G-980A, Ins5nt, A-849C, G-549A, T-478C, T-255C, A100G, and A215G was introduced in the parental double-mutant plasmid pCI-SMN1 C6T/G-44A, and analyzed in HEK293 cells. The effect of A100G was weaker than in the double-mutant context (Panel A), but there was a significant reduction in exon 7 inclusion (∼3%). On the other hand, some mutations, such as G-980A and G-549A, which had no effect in the double-mutant context, significantly improved splicing in the triple-mutant context (4–5% increase in exon 7 inclusion). *P < 0.01 versus parental SMN1 C6T/G-44A (n = 5). (C) SMN1 C6T and SMN1 C6T/G-44A (+G-44A) mutants were analyzed in five different cell types, including neural cells: HEK293, HeLa, COS1, NSC34, and SH-SY5Y. The splicing patterns in all cell types are consistent with a 15–20% decrease in exon 7 inclusion caused by the G-44A substitution. **P < 0.001 versus SMN1 C6T (n = 4).
We then generated all eight triple mutants based on SMN1 C6T/G-44A, and tested them in HEK293 cells. Interestingly, A100G in this setting displayed a very subtle but significant inhibitory effect (∼3% change). One the other hand, G-980A and G-549A displayed a slim (4–5% increase) but significant stimulatory effect on exon 7 splicing (Fig. 2B). These data suggest that the degree of effect of a splicing regulatory cis-element is context-dependent: splicing enhancers are more effective in a weak splicing setting, whereas splicing silencers are more effective in a more favorable splicing setting.
Based on the above mutagenesis analysis, we conclude that the G-44A substitution is the major determinant of splicing improvement in SMN1 C6T compared to SMN2, whereas A100G makes a more subtle contribution. To ensure that the G-44A effect in modulating exon 7 splicing is a general phenomenon, rather than HEK293 cell-specific, we tested the SMN1 C6T/G-44A mutant in various cell types, including human epithelial HeLa cells, monkey kidney fibroblast-like COS1 cells, human neuroblast-like SH-SY5Y cells, and mouse motor-neuron-like NSC34 cells; in all these cell types, G-44A resulted in a ∼20% decrease in exon 7 inclusion (Fig. 2C).
An RNA secondary structure is not the cause of the G-44A effect
A potential stem-loop structure around the -44 position in SMN1/2 intron 6 is predicted by UNAFold (Fig. 3A) (37). The G-44A transition changes a G-U pair in the stem to an A-U pair, slightly stabilizing the secondary structure (Fig. 3B). To test if stabilization of the structure accounts for the moderate inhibition of exon 7 splicing, we introduced several mutations in SMN1 C6T or SMN1 C6T/G-44A that either strengthen or weaken the putative structure, and tested them in HEK293 cells (Figs 3B and C). Three of the mutations were at position -44, either changing G into C or T, or deleting the G to weaken the structure; none of them resulted in an increase in exon 7 inclusion, compared to SMN1 C6T, despite a slight increase compared to SMN1 C6T/G-44A. Another mutation, A-46G, was generated in both SMN1 C6T and SMN1 C6T/G-44A contexts, changing a putative U-A pair to a U-G pair; both mutants resulted in only a slight increase in exon 7 inclusion. In the context of SMN1 C6T/G-44T, we also mutated position -57 from T to A to restore an A-U pair in the stem; however, exon 7 splicing remained unchanged. We additionally mutated position -57 from T to C in SMN1 C6T to restore base pairing at the same position as created by the G-44A change; however, the T-57C mutation had a much weaker effect in reducing exon 7 inclusion compared to the G-44A change (∼9% versus 22% decrease). Finally, we mutated position -42 from C to A, adding 3 more base-pairs to the putative stem. Though the C-42A mutation should strengthen the putative stem-loop structure, it resulted in only a ∼13% decrease in exon 7 inclusion in both SMN1 C6T and SMN1 C6T/G-44A contexts. Note, however, that the small splicing decrease caused by the C-42A mutation may or may not be solely attributable to the structural change. Indeed, analysis of the free energy of the predicted stem-loop structure at -44 in the SMN1/2 minigenes/mutants and their exon 7 inclusion percentages revealed no correlation between them (Fig. 3D).
RNA secondary structure is not involved in splicing repression caused by the G-44A substitution. (A) An RNA secondary structure was predicted by UNAFold at a region from position -41 to -60 in SMN1 intron 6. The arrow indicates position -44. (B) List of mutations in the SMN1 C6T setting for structure analysis, corresponding sequences, predicted minimum free energy (ΔG, Kcal/mol) of the secondary structure, and exon 7 inclusion of each mutant (Inc%). The underlined sequence is the loop, and mutations are shaded. Minimum free energy values were calculated by UNAFold. Some mutations strengthen the putative stem-loop structure, whereas others destabilize it. Exon 7 inclusion was analyzed as in Panel C; mean values are shown (n = 4). (C) All mutant plasmids in the SMN1 C6T setting were analyzed in HEK293 cells. SMN1 and SMN2 wild types and SMN1 C6T were used as controls. % exon inclusion is shown below this representative gel. (D) The percentage of exon 7 inclusion of each tested mutation, as well as SMN1 C6T, is presented as a scatter plot against the calculated minimum free energy (listed in Panel B). No correlation was observed by SPSS Pearson analysis (R2=0.11).(E) Deletion mutants Del1, Del2, Del3 and Del4 were created in the SMN1 C6T setting, and their splicing was analyzed in HEK293 cells. SMN1 WT, SMN1 C6T and SMN1 C6T/G-44A were used as controls. The histogram shows the quantitation of splicing from four experiments similar to the one shown on the left. Four mutants displayed exon 7 inclusion levels of 40, 73, 83 and 60%, respectively, all of which are significantly better than SMN1 C6T/G-44A. **P < 0.001 versus SMN1 C6T (n = 4).
The G-44A mutation moderately increases binding of HuR to the RNA
After ruling out that a change in the above stem-loop structure accounts for the G-44A effect, we hypothesized that this site may be part of a splicing regulatory element, and the G-44A change increases the binding of a splicing repressor to this element. We generated several deletion mutants around position -44, all of which improved exon 7 inclusion by various extents (Fig. 3E). To identify which protein(s) might bind to the -44 region, we conducted RNA-affinity chromatography. A 16mer RNA sequence (-49 to -34 in intron 6) with the G-44A mutation was obtained commercially and covalently linked to agarose beads via the 3′ end. The wild type (WT) sequence was used as a control (Fig. 4A). After incubation of these RNAs with HeLa cell nuclear extract under splicing conditions, the beads were extensively washed with buffer containing 100 mM KCl, and proteins that remained bound to the RNA were analyzed by SDS-PAGE and Coomassie-blue staining. As shown in Figure 4B, a few prominent bands around 36–37 kDa were present in both the G-44A and control samples, but appeared moderately stronger in the G-44A sample. The bands were excised from gels and analyzed by mass spectrometry. Four proteins stood out, with total peptide-spectrum matches (PSMs) ≥4 in one or both of the samples, with hnRNP A1 and hnRNP A2 being the most abundant, followed by HuR and WDR5 (Fig. 4C). It is not surprising that hnRNP A1 and A2 were efficiently pulled down by the two RNA species, because both RNAs have a UAG trinucleotide, a critical core of the binding motifs of hnRNP A/B family proteins. However, HuR was enriched in the G-44A pull-down protein sample, compared to the control. To validate this result, we performed western blotting with protein samples eluted from the two agarose bead-linked RNAs. We observed a 37% increase in HuR bound to the G-44A RNA, compared to the WT control (Fig. 4D). There was no statistically significant increase in hnRNP A1 and hnRNP A2 bound to the G-44A RNA. HuR is an abundant RNA-binding protein with multiple cellular functions, including splicing regulation mediated by binding to AU-rich or U-rich RNA motifs (38,39). The sequence surrounding the -44 site is an AU-rich stretch, and the G-44A change increases the A/U content; thus HuR represents a plausible candidate for the splicing repressor we were searching.
Analysis of proteins bound to the intronic splicing silencer around position -44 by RNA-affinity chromatography. (A) Two 16-nt RNA oligonucleotides with WT (corresponding to SMN1) and G-44A mutant (corresponding to SMN2) sequences were used for RNA-affinity chromatography. (B) Agarose beads covalently linked to the RNAs shown in (A) were incubated with HeLa cell nuclear extract (NE) under splicing conditions, and the beads were washed four times with Buffer D containing 100 mM KCl. Captured proteins were eluted with SDS sample buffer, separated by SDS-PAGE and stained with Coomassie Blue. A reaction without RNA (No RNA) was used as a control. Protein bands around 36-37 kDa were prominent on the gel in both WT and G-44A samples, with the bands from the G-44A sample appearing stronger. Bands were excised from the gel and subjected to mass spectrometry. (C) Mass spectrometry analysis revealed multiple proteins bound to the two RNA species. Four proteins with sizes between 35 kDa and 38 kDa had PSMs over 4. Among them, HuR had the largest increase in PSMs from the WT to G-44A samples (2.5 fold). The mass spectrum of a peptide derived from HuR, identified in the pulldown, is shown below. (D) Western blot analysis of the eluted proteins with anti-hnRNP A1, anti-hnRNP A2 and anti-HuR antibodies. 7.5% input of NE was loaded on the right lane. Band intensities were measured using Image J software. Normalized band intensity was displayed in the histogram. HnRNP A1, hnRNP A2 and HuR had a fold increase of 1.04, 1.15, and 1.37, respectively, in binding to the G-44A mutant RNA, compared to the WT RNA. **P < 0.001 (G-44A versus WT RNA, n = 3)
HuR represses splicing when bound to the -44 region
To explore if HuR influences exon 7 splicing when bound to the -44 region, we employed an MS2 tethering splicing assay (40). In this method, a protein of interest is fused to bacteriophage MS2 coat protein (CP) and co-expressed in cultured cells with a pre-mRNA substrate that contains one or more copies of the MS2 hairpin sequence at a position of interest, so that the effects of the test protein on pre-mRNA splicing of the substrate can be analyzed. Plasmid pCl-SMN1-MS2 was generated with one copy of the 19-nt MS2 hairpin sequence inserted at the -44 site in the SMN1 minigene (Fig. 5A); the new MS2 SMN1 minigene displayed a similar exon 7 splicing efficiency (96% inclusion) as the parental minigene (98%). Plasmid that expresses MS2 CP - HuR fusion protein, designated T7-CP-HuR, was generated in the pCGT7 vector. Since hnRNP A1 and hnRNP A2 also bind to or near the -44 region, their fusion proteins were also generated and designated T7-CP-A1 and T7-CP-A2, respectively. T7-tagged coat protein (T7-CP) was used as a control. All proteins had an N-terminal T7 tag. We co-transfected the MS2 SMN1 construct (1 μg) into HEK293 cells with a construct that expresses either T7-CP or one of the MS2 CP fusion proteins (100 ng). Empty vector pCGT7 (T7-Empty) that expresses no CP or its fusion proteins was used as a control. Before splicing analysis, western blot analysis with anti-T7 monoclonal antibody (mAb) confirmed that all proteins were properly expressed in transfected cells (Fig. 5B). Unexpectedly, T7-CP caused a 33% decrease in exon 7 inclusion (from 96% in T7-Empty to 63%); such a strong splicing repression by CP in an MS2-tethering system has never been documented. Fusion of hnRNP A1, hnRNP A2 and HuR to CP further reduced the percentage of exon 7 inclusion to 14, 7 and 16%, respectively (Fig. 5B). We used a low amount of each expression plasmid (100 ng) in cell transfection with the intention to minimize off-target effects caused by binding of the hnRNP A1, hnRNP A2, or HuR portion of each fusion protein to sites other than the -44 region. With this low amount of transfected plasmid, the extent of splicing repression by non-fused HuR, hnRNP A1, or hnRNP A2 was small (decreased by 1–8% exon 7 inclusion, three rightmost lanes in Fig. 5B). To further confirm that the robust repression of exon 7 splicing by hnRNP A1, hnRNP A2 and HuR fusion proteins is because they are fused to MS2 CP, T7-CP construct (50 ng) and each of the three constructs that express non-fused hnRNP A1, hnRNP A2 or HuR (50 ng each) were co-transfected into HEK293 cells, together with the MS2-SMN1 minigene plasmid (1 μg). As shown in Figure 5B, co-expression of two separate proteins resulted in a similar effect on exon 7 splicing or a slight decrease, compared to T7-CP alone. Taken together, these data demonstrate that hnRNP A1, hnRNP A2 and HuR are all strong splicing repressors when tightly tethered to the -44 region.
MS2 tethering assay revealed HuR as a splicing repressor when bound to the -44 site. (A) Diagram of the SMN1 minigene with an MS2 hairpin (MS2 SMN1). The sequence from −47 to − 36 in intron 6 (TATGTCTATATA) was deleted and replaced with an MS2-binding sequence (5’ -ACATGAGGATCACCCATGT -3’). The pre-mRNA transcribed from the MS2 SMN1 minigene forms a stem-loop structure that binds to MS2 CP. (B) The effects of various expressed proteins on exon 7 splicing were analyzed in HEK293 cells. The MS2 SMN1 minigene (1 μg) and various expression plasmids (100 ng each for transfection with one expression plasmid or 50 ng each for co-transfection with 2 expression plasmids) were co-transfected into cells. T7-Empty: pCGT7 empty vector was transfected as a blank control. T7-CP: T7-tagged coat protein; T7-CP-A1: hnRNP A1 fused to T7-CP; T7-CP-A2: hnRNP A2 fused to T7-CP; T7-CP-HuR: HuR fused to T7-CP. T7-A1: T7-tagged hnRNP A1; T7-A2: T7-tagged hnRNP A2. Western blot (WB) analysis with an anti-T7 antibody showed proper expression of all proteins. Cy5-labeled RT-PCR analysis of exon 7 splicing was quantitated and shown in the histogram. **P<0.001 versus T7-CP (n = 3). *Nonspecific band.
RRM1 and RRM2 mediate splicing repression activity by HuR
HuR and its homologs comprise three highly conserved RNA-recognition motifs (RRMs 1–3) and a less conserved hinge region (Hinge) between RRM2 and RRM3 (38). To identify which domains of HuR are required for repressing splicing of SMN2 exon 7, we generated a series of deletion constructs (Fig. 6A) from the parental pCGT7-CP-HuR plasmid. Each deletion plasmid (100 ng) was co-transfected with pCl-SMN1-MS2 (1 μg) into HEK293 cells, and both protein expression of the mutants and their effects on exon 7 splicing were analyzed. T7-CP and T7-CP-HuR were used as controls. Compared to T7-CP-HuR, deletion of RRM3 had no effect on exon 7 splicing, and further deletion of Hinge only slightly increased exon 7 inclusion (from 18 to 25%), indicating that RRM1 and RRM2 are sufficient for splicing repression (Fig. 6B). Interestingly, deletion of either RRM1 or RRM2 resulted in significant reductions in their cellular protein levels, as well as abrogation of their splicing-inhibitory activities. It was previously reported that RRM1 and RRM2 of Hu family proteins are essential for protein homodimerization and RNA binding, and that removal of RRM3 and Hinge does not affect the overall stability of HuR protein folding (41,42). Our data are consistent with these findings, and suggest that loss of the homodimerization ability compromises protein stability of these mutants.
RRM1 and RRM2 of HuR are sufficient for splicing repression. (A) Diagram of the primary structure of HuR and its deletion mutants. HuR has three RNA recognition motifs (RRM1, RRM2 and RRM3) and a hinge region between RRMs 2 and 3. The residue numbers at domain boundaries are indicated. The T7 tag is denoted by a hexagon. (B) The effects of the indicated HuR mutants on exon 7 splicing were analyzed in HEK293 cells. The MS2 SMN1 minigene (1 μg) and each protein expression plasmid (100 ng) were co-transfected into cells. Western blot (WB) analysis using anti-T7 antibody was performed to examine protein expression. Splicing analysis with Cy5-labeled RT-PCR was performed, and quantitation of exon 7 inclusion is shown in the histogram. **P < 0.001, #P>0.05 versus T7-CP-HuR (n = 3). *Nonspecific band.
Effects of HuR overexpression and knockdown
Surrounding position -44 are two trinucleotide UAGs at positions -35 to -37 and -53 to -55 in intron 6, which are presumptive hnRNP A1/A2-dependent splicing silencers. Between the two UAG motifs from -53 to -38, 11 out of 15 nt are As or Us. When the sequences of the above minigene mutants were aligned, a clear negative correlation between A/U content and exon 7 inclusion percentage was observed (Fig. 7A). To confirm that HuR binding to the -44 region is A/U-dependent, we analyzed the effect of HuR overexpression on the SMN1 minigene and its mutants Del4 (most of the A/U stretch was deleted, as shown in Fig. 3E), G-44A and G-44A/C-42A (see Fig. 3), which have different numbers of As and Us in the A/U stretch. Plasmids were co-transfected into HEK293 cells and RNA samples were collected 36-h post-transfection for splicing analysis. As shown in Figure 7B, HuR displayed a stronger inhibitory effect when the minigene harbored more A/T in the stretch and the least effect when the A/T stretch was mostly deleted (Del4).
Effects of HuR overexpression and knockdown. (A) A negative correlation between the A/T content and exon 7 splicing was observed in SMN1 C6T, SMN1 C6T/G-44A and SMN1 C6T/G-44A/C-42A (see Fig. 3). The number of A or T residues within a 15-nt segment is indicated for each construct. (B) Effect of HuR overexpression on exon 7 splicing, analyzed with SMN1 WT, SMN1 G-44A, SMN1 G-44A/C-42C and SMN1 Del4 (see Fig. 3). In the SMN1 setting, the -44 specific and A/T content-dependent effects of HuR could be distinguished with these WT and mutant minigenes. Each minigene plasmid (1 μg) and pCGT7-HuR (1 μg) were co-transfected into HEK293 cells, and exon 7 splicing was analyzed with Cy5-labeled RT PCR; the histogram on the right shows the quantitation of exon 7 inclusion. Western blot (WB) analysis using an anti-T7 antibody confirmed expression of T7-HuR. **P<0.001, T7-HuR versus T7-Empty (n = 3). (C) Effect of HuR knockdown on exon 7 splicing, analyzed with SMN1 C6T and SMN1 C6T/G-44A. Each siRNA (100 nM) and one minigene plasmid (1 μg) were co-transfected into HEK293 cells. HuR was detected with a monoclonal anti-HuR antibody, with β-tubulin as a loading control. Two siRNAs (siRNA-1 and siRNA-2) robustly reduced HuR expression (to 52% and 48%, respectively) compared to a control siRNA (siNC). After HuR knockdown, exon 7 splicing in SMN1 C6T/G-44A showed a greater increase than in SMN1 C6T (18–20% versus 7–10%). **P < 0.001 compared to siNC (n = 3).
We further performed HuR knockdown in HEK293 cells using two independent siRNAs to determine whether it causes an increase in exon 7 inclusion with the minigene mutants SMN1 C6T and SMN1 C6T/G-44A. Based on western blotting analysis, HuR expression levels were reduced to ∼ 50% after siRNA treatment (Fig. 7C). As expected, knockdown of HuR enhanced exon 7 inclusion in both mutants, and the effect on mutant SMN1 C6T/G-44A (more HuR binding) was much more pronounced than that on SMN1 C6T (less HuR binding). Previously, HuR was reported as a weak repressor of exon 7 splicing of the endogenous SMN2 (43); our overexpression and knockdown data further demonstrate that the inhibitory splicing effect on exon 7 splicing by HuR is at least partially mediated by the A/U stretch around the -44 region.
Discussion
SMA is an SMN-deficiency disease, and thus the SMN2 copy number, reflected in the SMN expression levels, is inversely correlated with SMA severity. However, in the clinic, we and others frequently observed cases that do not follow the SMN2 copy number rule, including rare SMA discordant families, in which siblings with the same SMN1 mutations and SMN2 alleles can be asymptomatic or affected (11,17,44). One interpretation is the existence of a modifier gene(s) that acts downstream in the disease pathway, partially compensating for the low level of SMN, such as PLS3 (17). It was recently reported in a severe mouse model that overexpression of PLS3 combined with treatment with a therapeutic ASO restores endocytosis in NMJs, a defect considered by the authors to be critical in SMA pathogenesis (19). Another mechanism underlying the phenotypic discordance is SMN2 itself, i.e. that SMN2 alleles are not identical. Such a concept was proved by our earlier discovery of the splicing-improving SMN2 c859G > C allele in three patients who carry 2 copies of this allele and display a very mild phenotype (11). To date, PLS3, NCALD and c859G > C are the only three SMA modifiers identified in humans, which together account for a limited number of cases that do not follow the SMN2 copy-number rule. Thus, many additional cases remain to be elucidated.
In this study, we identified a single allele conversion event in 5/53 mild SMA patients who carry 3 copies of SMN2. The converted SMN2 allele is equivalent to the SMN1 C6T mutant. Using a minigene system, we surprisingly found that SMN1 C6T splices exon 7 more efficiently than SMN2, and the extent of splicing improvement is similar to that caused by the c859G > C substitution (11). Furthermore, using mutational and splicing analysis, we identified A-44G in intron 6 as the major contributor, and A100G in intron 7 as a minor contributor, to the improved splicing of SMN1 C6T. Finally, using RNA-affinity chromatography, mass spectrometry, an MS2-tethering splicing assay, as well as HuR overexpression and knockdown, we showed that the AU-rich stretch around the -44 site is a HuR-dependent splicing silencer, and the G-44A transition improves HuR binding, resulting in moderate inhibition of exon 7 splicing. Thus, we identified a novel mechanism that accounts for the discordance with the SMN2 copy-number rule.
Lorson et al. first used minigenes to explore the effects of five natural nucleotide changes between SMN1 and SMN2 (G-44A in intron 6, C6T in exon 7, A100G and A215G in intron 7, and G236A in exon 8) on exon 7 splicing (7). Later, Monani et al. evaluated the effect of the remaining nucleotide differences in intron 6 (6); these studies led to the key finding that the C6T substitution in exon 7 dictates the distinct splicing patterns of SMN1/2, which laid the foundation for SMN2-splicing-based therapeutics. Our present work highlights the existence of another natural nucleotide transition between SMN1/2 that plays an important detrimental role in exon 7 splicing, though the extent of decrease in exon 7 inclusion caused by G-44A is not as large as that caused the C6T change (∼20% versus ∼50%). Previously, Kashima et al. reported that the A100G change in intron 7 between SMN1/2 creates an hnRNP A1-dependent splicing silencer (from UAA to UAG) and exerts an inhibitory effect on exon 7 splicing (45). Our mutagenesis analysis confirmed that the A100G transition contributes to the splicing deterioration in SMN2, though its contribution is subtle.
The ELAV/Hu family proteins include four members: HuR (also called HuA), HuB, HuC and HuD, which play important roles in posttranscriptional regulation of gene expression. HuR is ubiquitously expressed and involved in diverse cellular processes, including the cell cycle, cell proliferation, cell survival, angiogenesis, etc., whereas the other three members show neuron-specific expression and are involved in neuronal development, maintenance and plasticity (38,39). The Hu proteins are highly similar in structure, and their best-known function is to regulate mRNA stability. They bind to AU-rich elements in the 3’ UTR to prevent degradation of target mRNAs, through competition with either destabilizing proteins or miRNAs that also bind to AU-rich elements or nearby sequences. In the cytoplasm, Hu proteins also regulate protein expression positively or negatively at the level of translation, through various mechanisms (38). In the nucleus, Hu proteins regulate polyadenylation and inhibit splicing of certain alternative cassette exons (38). Wee et al. recently examined the effects of multiple splicing factors on exon 7 splicing of the endogenous SMN2 gene, using siRNA knockdown and overexpression, and reported that HuR acts as a weak repressor of exon 7 inclusion (43). Our data further pinpointed a site to which HuR binds to exert its splicing-repressing activity. When bound upstream of an exon, Hu proteins are thought to decrease U2AF binding to the 3’ splice site; when bound downstream of an exon, they inhibit U1 snRNP binding to the 5’ splice site by competing with the positive splicing regulators TIA-1/TIAR, which also recognize AU-rich sequences (46).
HuR can multimerize when bound to AU-rich mRNA sequences (41). Thus HuR may repress splicing by blocking the binding of splicing factors not only at the AU-rich stretch, but also at neighboring sequences, through cooperative binding along the pre-mRNA, in this respect resembling hnRNP A1 (39,47). This multimerization mechanism is consistent with our HuR-deletion-mutant analysis, which revealed that RRM1 and RRM2, which are essential for HuR dimerization, are necessary and sufficient for its repressing effect on exon 7 splicing. However, we note that all the HuR deletion mutants that failed to repress splicing were also unstable; thus, it remains to be determined whether splicing repression by HuR requires a specific protein motif or relies only on steric blockage.
Interestingly, Hu family proteins have been reported to stimulate exon inclusion (48). Using RNA-protein crosslinking (PAR-CLIP), Lebedeva et al. identified ∼15,000–20,000 HuR binding sites in HeLa cells, about one third of which mapped to introns near the splice sites and showed high sequence conservation, suggesting that HuR is a broad regulator of pre-mRNA splicing (49). The authors uncovered more events in which HuR functions as a splicing activator than a repressor, indicating that HuR may utilize multiple mechanisms to regulate pre-RNA splicing. One of the mechanisms could involve interaction with chromatin. Zhou et al. reported that HuR induces local histone hyperacetylation by inhibiting histone deacetylase 2 when associated with nascent pre-mRNA (50). Therefore, HuR may affect many alternative splicing events by altering transcription and/or the RNA polymerase II elongation rate.
We noticed that the effect of HuR on exon 7 splicing of the endogenous SMN2 gene was not as pronounced as we observed with the minigenes (43). One possible explanation for the discrepancy is that HuR may modulate exon 7 splicing positively when binding to other regions that are not included in the minigenes; as a result, the effect of HuR at the -44 position would be partially neutralized. Alternatively, the discrepancy could be caused by the role of HuR in histone modifications. Histone deacetylase inhibitors promote SMN2 exon 7 inclusion and have been explored as SMA therapeutics (51). Therefore, the repressing effect of HuR at the -44 site on SMN2 exon 7 splicing might be counterbalanced by its effects on transcription-coupled splicing of the endogenous SMN2 gene through chromatin changes.
Aside from its splicing effect, HuR has also been reported to regulate SMN1/2 mRNA stability by binding to the 3’ UTR region, which is thought to engage the P38 pathway to increase SMN levels (52). In addition, HuD is an SMN-interacting partner, and is required for proper localization of HuD-RNA complexes in neuronal RNA granules; SMN-deficient motor neurons have low levels of HuD and poly(A) mRNAs in their axons, and HuD overexpression can rescue axonal defects in SMN-deficient motor neurons (53–55). Taken together with our data, these findings implicate Hu family proteins in multiple aspects of SMN1/2 expression and SMN protein functions.
Materials and Methods
Plasmid construction
SMN1/2 minigene constructs pCl-SMN1 and pCI-SMN2 were similar to those previously described (34), but with a longer truncated intron 6 to include all sequence differences from intron 6 to intron 7 that naturally occur between the endogenous SMN1 and SMN2 genes. Briefly, an 1186-nt fragment was inserted into the original SMN1 minigene using the SLIC method (56), such that the new shortened intron 6 is 1386-nt long, retaining 61 nt immediately downstream of exon 6, and 1325 nt (1186 + 139) immediately upstream of exon 7, with the middle segment being deleted. The SMN2 minigene and SMN1/2 mutants were generated from this SMN1 minigene by site-specific mutagenesis with partially-overlapping primer pairs.
Plasmid pCl-SMN1-MS2 was generated by replacing the AU-rich sequence from -47 to -36 with a 19-nt MS2 hairpin sequence (5’-ACAUGAGGAUUACCCAUGU-3’) (57). All expression plasmids were constructed with vector pCGT7, and thus all expressed proteins have an N-terminal T7 tag. Plasmids pCGT7-hnRNPA1 and pCGT7-hnRNPA2 were previously generated (35). Plasmid pCGT7-HuR has a cDNA inserted between the XbaI/KpnI sites with primers HuR-F (5’-AAATCTAGAATGTCTAATGGTTATGAAG-3’) and HuR-R (5’-ATGGGTACCTTATTTGTGGGACTTGTTGG-3’). For generating plasmid pCGT7-CP that expresses bacteriophage MS2 CP, CP cDNA was amplified from the previously described plasmid MS2-Fox1C with primers MS2CP-F (5’- ACGTCTAGAGCCTCCAACTTCACC-3’) and MS2CP-R (5’-ACGGTACCTACGTCGACGTAGATGCCGGAGTTG-3’), and inserted between the XbaI/KpnI sites (40). Plasmid pCGT7-CP has a SalΙ restriction site at the 3’ end of the CP cDNA for further cloning of inserts to express CP fusion proteins. Plasmids pCGT7-CP-HuR, pCGT7-CP-A1 and pCGT7-CP-A2 that express three CP fusion proteins, T7-CP-HuR, T7-CP-hnRNP A1 (or T7-CP-A1) and T7-CP-hnRNP A2 (or T7-CP-A2), respectively, were generated by inserting the corresponding cDNAs into pCGT7-CP at the SalI/BamHI or SalI/KpnI sites with primer sets hnRNPA1-F (5’-ACGTCGACTCTAAGTCAGAGTCTCC-3’) and hnRNPA1-R (5’-ACGGATCCTTAAAATCTTCTGCCACTGC-3’), hnRNPA2-F (5’-ACGTCGACGAGAGAGAAAAGGAACAG-3’) and hnRNPA2-R (5’-ACGGTACCTCAGTATCGGCTCCTC-3’), or HuR-F (5’-GTCGACATGTCTAATGGTTATG-3’) and HuR-R (5’- GGGTACCTTATTTGTGGGACTT -3’). HuR deletion mutants were generated in the parental plasmid pCGT7-CP-HuR by replacing HuR with the corresponding fragments using restriction sites and/or the SLIC method.
Cell culture, transfection and siRNA knockdown
HEK293, HeLa, COS1 and SH-SY5Y cells were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Invitrogen) supplemented with 10% (v/v) fetal bovine serum (FBS) and antibiotics (100 U/ml penicillin and 100 μg/ml streptomycin) at 37°C in a humidified 5% CO2 atmosphere. For splicing analysis of SMN1/2 WT and mutant minigenes, 105 cells per well were seeded in six-well plates in DMEM with 10% FBS; the next day, 1 μg of each minigene plasmid together with or without a total of 0.1 μg (or as indicated) protein-expression plasmid(s) was delivered to cells using branched polyethylenimine reagent (Sigma). The siRNAs to knock down HuR were siHuR-1 (sense: 5’-CAGUUUCAAUGGUCAUAAATT-3’; antisense: 5’- UUUAUGACCAUUGAAACUGGT-3’) and siHuR-2 (sense: 5’-GAGGCAAUUACCAGUUUCATT-3’; antisense: 5’-UGAAACUGGUAAUUGCCUCTT-3’), purchased from Genepharma (Shanghai), and the negative control was siNC (sense: 5’-UUCUCCGAACGUGUCACGUTT-3’; antisense: 5-’ACGUGACACGUUCGGAGAATT-3’); siRNAs were transfected using Lipofectamine LTX (Life Technologies).
Fluorescence-labelled RT-PCR
Cells were harvested 36-h post-transfection, and total RNA was isolated with TRizol reagent (Tiangen); 1 μg of each RNA sample was used per 20-μl reaction for first-strand cDNA synthesis with oligo (dT)18 and M-MLV reverse transcriptase (Vazyme). Splicing products were amplified semi-quantitatively using 26 PCR cycles (94°C for 30 s, 55°C for 30 s, and 72°C for 25 s) with forward primer T7-F2 (5’-TACTTAATACGACTCACTATAGGCTAGCCTCG-3’) and Cy5-labeled reverse primer Ex8-29to52-R (5’-Cy5-TCTGATCGTTTCTTTAGTGGTGTC-3’). Cy5-labeled PCR products were separated on 6% native polyacrylamide gels, followed by fluorescence imaging with G:BOX Chem XL (Syngene); signals were quantitated by GeneTools software (Syngene) and exon 7 inclusion was expressed as a percentage of the total amount of spliced transcripts.
RNA-affinity chromatography
RNA oligonucleotides WT (5’-UCUAUGUCUAUAUAGC-3’) and G-44A (5’-UCUAUAUCUAUAUAGC-3’) were purchased from Genscript (Nanjing, China). RNA-affinity chromatography was performed as described (35,58). In brief, 20 μg of each RNA was oxidized in a 50-μl reaction containing 100 mM sodium acetate pH 5.0 and 5 mM sodium m-periodate (Sigma) for 1 h in the dark, and then covalently attached to adipic acid dihydrazide agarose beads (Sigma) by rotating the mixture in 500 μl of 0.1 M sodium acetate (20% slurry) for 12 h at 4°C. The RNA-conjugated beads in buffer D (20 mM HEPES-KOH, pH 7.6, 20% (v/v) glycerol, 0.1 M KCl, 0.2 mM EDTA, 1 mM dithiothretol, and 0.5 mM PMSF) were incubated with HeLa-cell nuclear extract in a 500-μl reaction under in vitro splicing conditions for 30 min at 30°C. The beads were then washed four times with 1 ml buffer D. The proteins bound to the immobilized RNA were eluted by the addition of 50 μl of 1× Laemmli buffer and heated for 10 min at 90°C.
Mass spectrometry
Proteins captured by RNA-affinity chromatography (see above) were separated by SDS-PAGE and stained with Coomassie Blue. The bands of interest were excised and processed for in-gel trypsin digestion and LC-MS/MS analysis, according to a published procedure (59).
Western blotting
Protein samples separated by 12% SDS-PAGE were electroblotted onto PVDF membranes (Millipore). The blots were then probed with primary mAbs or polyclonal antibodies (pAbs), followed by secondary IRDye® 680RD goat anti-mouse or goat anti-rabbit antibody (LI-COR). Anti-T7 mAb was generated at CSHL; anti-HuR and anti-β-tubulin mAbs were purchased from Santa Cruz and MultiScience; anti-hnRNP A1 and anti-hnRNP A2/B1 pAbs were purchased from Protein Tech. Protein signals were detected with an Odyssey Infrared Imaging System (LI-COR).
Statistical analysis
Data are presented as mean ± standard deviation. Statistical significance was analyzed by Student’s t-test and one-way ANOVA with software SPSS 16.0.
Acknowledgements
The authors thank Professor Guoqiang Xu and his graduate students Xun Lin and Yang Zhang at the College of Pharmaceutical Sciences, Soochow University, for performing the mass spectrometry analysis. Y.H. gratefully acknowledges support from the National Natural Science Foundation of China. A.R.K. acknowledges support from NIH.
Conflict of Interest statement. None declared.
Funding
National Natural Science Foundation of China (grants 81271423, 81471298 and 81530035), Priority Academic Program Development of Jiangsu Higher Education Institutions and NIH grant GM42699.
