An intronic element contributes to splicing repression in spinal muscular atrophy

Abstract

The neurodegenerative disease spinal muscular atrophy is caused by mutation of the survival motor neuron 1 (SMN1) gene. SMN2 is a nearly identical copy of SMN1 that is unable to prevent disease, because most SMN2 transcripts lack exon 7 and thus produce a nonfunctional protein. A key cause of inefficient SMN2 exon 7 splicing is a single nucleotide difference between SMN1 and SMN2 within exon 7. We previously provided evidence that this base change suppresses exon 7 splicing by creating an inhibitory element, a heterogeneous nuclear ribonucleoprotein (hnRNP) A1-dependent exonic splicing silencer. We now find that another rare nucleotide difference between SMN1 and SMN2, in intron 7, potentially creates a second SMN2-specific hnRNP A1 binding site. Remarkably, this single base change does indeed create a high-affinity hnRNP A1 binding site, and base substitutions that disrupt it restore exon 7 inclusion in vivo and prevent hnRNP A1 binding in vitro. We propose that interactions between hnRNP A1 molecules bound to the exonic and intronic sites cooperate to exclude exon 7 and discuss the significance of this exclusion with respect to SMN expression and splicing control more generally.

Alternative splicing of mRNA precursors is an important mechanism that regulates gene expression and generates increased protein diversity from a limited number of genes. Indeed, expression of 70% or more of human genes is now estimated to involve alternative splicing (1, 2). In higher eukaryotes, alternative splicing plays an essential role in diverse cellular functions, contributing to such basic processes as cell growth, differentiation, and cell death (3, 4). Several types of cis-acting RNA elements and trans-acting protein factors help to regulate alternative splicing and do so by diverse mechanisms. In general, however, non-spliceosomal proteins that participate in regulating alternative splicing associate with regulatory elements in exons and/or introns. Serine/arginine-rich proteins (5) are typically involved in positive regulation of splicing, stimulating splicing by interacting with exonic splicing enhancer elements (ESEs) or intronic splicing enhancer elements. In contrast, negative regulation is promoted most frequently by heterogeneous nuclear ribonucleoproteins (hnRNPs) (6), which function by binding sequences known as exonic splicing silencers (ESSs) or intronic splicing silencers.

Several hnRNP proteins have been identified as key splicing repressors. Among these, the abundant hnRNP A1 protein has been extensively characterized. The first hnRNP A1-dependent ESS was identified in studies of HIV tat exon 2 repression (7–9). This ESS was found to bind hnRNP A1, and mutations disrupting the ESS prevented hnRNP A1 binding and allowed enhanced exon 2 splicing. Several mechanisms have been proposed to explain hnRNP A1-mediated splicing repression. In one, hnRNP A1 binds to an ESS and through direct protein-protein interactions, recruits more hnRNP A1 molecules, assembling a complex that inhibits spliceosome formation (10). In this case, no additional specific sequence elements are required; a single ESS triggers exon repression in cooperation with A1 self-assembly. In a second scenario, an hnRNP A1 binding site is located adjacent to the branch point sequence near the 3′ splice site, and A1 directly blocks U2 small nuclear ribonucleoprotein (snRNP) complex formation (11). However, even in this case, both exonic and intronic A1 binding sites are necessary for full repression, perhaps reflecting cooperative interactions between A1 molecules. A third mechanism was elucidated by studies of alternative splicing of the hnRNP A1 transcript itself (12, 13). The presence of hnRNP A1 binding sites in both introns surrounding alternative exon 7b is necessary for exclusion of this exon. Cooperation between the two A1 complexes on these sites is suggested to promote “looping-out” of the intervening RNA, including exon 7b, thereby inhibiting splicing. Finally, in Drosophila, hrp48, a homolog of hnRNP A1, contributes to splicing repression of the P-element third intron by binding to an ESS in the upstream exon, and this repression interferes with 5′ splice-site recognition (14, 15). Splicing repression by hnRNP A1 and the existence of A1-dependent ESSs have been documented in a number of other instances in humans (16–19), but in these cases, the mechanism and possible requirement of additional sequences elements that cooperate with the ESSs have not been investigated.

Mutations in regulatory elements in premRNAs that affect alternative splicing can give rise to genetic disorders (20). Although this frequently involves disruption of positive elements, e.g., ESEs, some mutations also affect splicing by creating negative elements, such as ESSs (21). Splicing of exon 7 in SMN1 and SMN2 transcripts constitutes an excellent and clinically important model to investigate molecular mechanisms of splicing regulation and specifically the roles of ESEs and ESSs. Homozygous loss of SMN1 is responsible for the hereditary disease spinal muscular atrophy (22). Although the SMN1 and SMN2 genes are almost identical, the presence of SMN2 does not prevent development of spinal muscular atrophy. A C → T transition at position +6 of exon 7 is largely responsible for alternative exon 7 splicing (95% of SMN1 transcripts include exon 7, and 80–95% of SMN2 mRNAs lack exon 7; see refs. 23 and 24). This base change, which is one of only four in >1.25 kb encompassing exon 7 and much of the flanking introns (23), does not affect either the exon 7 5′ splice site or an ESE in the middle of the exon dependent on the splicing regulator Tra2 (25, 26). Krainer and colleagues have presented evidence that the +6 base change disrupts an ESE in SMN1 that depends on a specific serine/arginine-rich protein, ASF/SF2, which they suggest results in SMN2 exon 7 exclusion (27, 28). In contrast, Kashima and Manley (29) provided data indicating that the C → T change creates an hnRNP A1-dependent ESS that recruits hnRNP A1 to SMN2 but not SMN1 premRNA, thereby repressing SMN2 exon 7 splicing. Consistent with the view that a negative element exists in the vicinity of +6, Singh et al. (30) showed that sequences flanking this position exert an inhibitory effect on SMN2 exon 7 inclusion.

Here we provide additional unexpected insights into the mechanism by which SMN2 exon 7 splicing is repressed. We first identify another single nucleotide difference, at position +100 in intron 7, that has the potential to create a second hnRNP A1 binding site specific to SMN2. We then provide evidence that this intronic A1 binding site is necessary for efficient exon 7 exclusion in vivo. We also show that this site specifically and strongly binds hnRNP A1 in vitro and that mutations that disrupt the A1 binding motif prevent protein binding and increase exon 7 inclusion. Our data indicate that the mechanism of SMN2 exon 7 exclusion is more complex than previously thought and provide new insights into splicing control more generally.

Results

Our previous data showed that the C → T single nucleotide difference at position +6 in SMN2 exon 7 creates an hnRNP A1-dependent ESS and that this sequence strongly and specifically binds hnRNP A1 (29). The presence of the ESS is a key mechanism in suppression of exon 7 splicing in SMN2 transcripts. Although our data are consistent with the view that this is the sole factor responsible for SMN2 exon 7 exclusion (ref. 29; unpublished data), it remains possible that an alternative mechanism, such as loss of ESE function (27, 28), also contributes to SMN2 exon 7 exclusion. However, we do not address this possibility here, but rather describe additional experiments designed to elucidate how the SMN2-specific ESS functions.

Deletion of Most Intronic Sequences Affects Exon 7 Splicing Only Modestly.

Alternatively spliced exons are frequently located near long introns with nonconsensus 5′ splice sites and/or a short polypyrimidine tract at the 3′ splice site (21). SMN exon 7 is typical of this finding, because it is located next to a long intron (intron 6 encompasses ≈6.0 kb) and is flanked by both weak 3′ and 5′ splice sites (25, 31). We first wished to determine whether the lengths of the flanking introns influence exon 7 splicing. To this end, we gradually deleted internal sequences in expression plasmids containing exons 6–8 of both SMN1 and SMN2 (see Fig. 1A) and then tested the effect of these deletions on SMN2 exon 7 splicing, using RT-PCR after transfection in HEK293 cells. The shortest constructs, designated SMN1_E and SMN2 _E, deleted 5.7 kb of intron 6 and approximately half (217 bp of 444 bp) of intron 7 but displayed similar splicing patterns compared with the corresponding full-length transcripts (Fig. 1B). Nearly 95% of both full-length and truncated SMN1 transcripts retained exon 7, whereas with SMN2, exon 7 was ≈90% excluded with the full-length transcript and 70% excluded with the truncated version. Thus, significant shortening of the long intron 6 had no affect on splicing, but shortening of the 217 bases intron 7 led to a slight decrease in exon 7 exclusion (see also Fig. 2F, lanes 2 and 8). These results indicate that most of the intronic sequences are not essential for determining exon 7 inclusion or exclusion.

Fig. 1.

Deletion of internal intron 6 and intron 7 sequences minimally affects SMN exon 7 splicing. (A) Schematic diagram of internal deletions of introns 6 and 7 in SMN1/2 plasmids. Black boxes indicate exons, horizontal thick lines indicate introns, and the vertical white line denotes the position of the C → T transition in SMN2 exon 7. Restriction enzyme sites that were used for cloning are indicated. Sizes of exons and introns in the shortest constructs, specifically SMN1/2_E, are indicated. (B) RT-PCR of RNA isolated from transfected 293 cells with the indicated deletion constructs. Positions of full-length and exon 7-excluded (Δ7) PCR products are indicated on the right. Both spliced products from the full-length SMN1 and SMN2 constructs migrated slightly slower than the corresponding RNAs from the mutants; this slower migration was due to different cloning sites at 3′ ends. (C) Schematic diagram shows wild type, ΔInt6, and ΔInt7 constructs in SMN1 and SMN2 backgrounds. Open boxes indicate exons, thick lines indicate introns, and the vertical line in exon 7 denotes the position of the ESS. (D) SMN1/2_HP and SMN1 and SMN2 Δint6 and Δint7 plasmids (4 μg) were transfected into 293 cells. Total RNA was prepared after 48 h and analyzed by using RT-PCR. The position of PCR products corresponding to precursor (unspliced) RNA, full-length, and exon 7-excluded (Δ7) mRNA are indicated on the right. Asterisk indicates an apparent aberrant splicing product.

Fig. 2.

Disruption of consensus hnRNP A1 binding sites in intron 7 rescues exon 7 splicing. (A) Location and sequence of consensus hnRNP A1 binding sites (TAGNNA/T) in SMN1/2_E constructs. SMN1/2-common consensus A1 sites are indicated above the diagram, and SMN2-specific consensus sites are indicated below the diagram. (B) Position of single nucleotide differences between SMN1 and SMN2 in the SMN_E constructs. The exonic +6 C → T and intronic +100 A → G transitions are indicated below the diagram. Other nucleotide differences are shown above the diagram. (C) The schematic diagram shows the positions of the T → C mutations in the consensus hnRNP A1 sites in intron 7 of the SMN1/2_E derivatives. Open boxes indicate exons, solid lines indicate introns, and the vertical thick line in exon 7 denotes the ESS. Crosses indicate positions of mutated consensus A1 sites. (D) RNA was prepared after 48 h transfection of 293 cells and analyzed by quantitative RT-PCR in the presence of [³²P]dCTP. The position of full-length (FL) and exon 7-excluded (Δ7) PCR products are indicated on the right. The percentage of exon 7 skipping was calculated from the total of exon 7 inclusion and exclusion products, measured by PhosphorImager, and is indicated below. (E) Diagram of mutant and intron 7 “swap” constructs. Open boxes indicate exons, dotted lines indicate SMN1 introns, solid lines indicate SMN2 introns, and the vertical line in exon 7 denotes the position of ESS. The cross indicates TAG mutations at position +100 in intron 7 in the SMN2_E construct. (F) SMN1/2_E, SMN1/2_HP, and derivative plasmids (4 μg) were transfected into 293 cells. RNAs was prepared after 48 h and analyzed by quantitative RT-PCR. Positions of full-length (FL) and exon 7-excluded (Δ7) PCR products are indicated on the right. The percentage of exon 7 skipping for each sample is indicated below.

Intron 7 Contains Sequences That Are Necessary for Exon 7 Exclusion.

We next wished to determine whether intronic sequences have any essential role in modulating exon 7 splicing. That is, does complete deletion of intron 6 (or intron 7) affect the ability of exon 7 to be spliced to exon 8 (or exon 6)? And does the exon 7 +6 base change influence splicing under these conditions? To investigate, we first precisely deleted either intron 6 (Δint6) or intron 7 (Δint7) in SMN1/2_HP (see Fig. 1C), transfected the plasmids into 293 cells, and again measured exon 7 splicing, using RT-PCR. Deletion of intron 6 resulted in very inefficient splicing of exons 7 and 8 in both SMN1 and SMN2 (Fig. 1D, lanes 4 and 5), suggesting that intron 6 contains (a) positive element(s) necessary for efficient splicing of exons 7 and 8. Importantly, though, deletion of intron 7 resulted in efficient splicing of exon 6 to exon 7 in SMN2 and SMN1 (lanes 6 and 7). This finding suggests that intron 7 contains an element required for SMN2 exon 7 exclusion and motivated us to investigate whether specific intron 7 sequences are necessary for SMN2-specific exon 7 exclusion.

Identification of Consensus hnRNP A1 Binding Sites in Intron 7 and Their Role in Exon 7 Exclusion.

Intronic and exonic hnRNP A1 binding sites function together to suppress HIV tat exon 3 splicing (11, 32). We therefore next searched for the presence of the consensus hnRNP A1 binding motif, UAGNNA/U (29, 33), in intron 6 and especially intron 7 of SMN1_E and SMN2_E (Fig. 2A). There are four consensus A1 binding sites in intron 6 and two/three in intron 7. Remarkably, SMN2 intron 7 contains a unique consensus A1 binding site, reflecting an A → G transition at position +100 (relative to the 5′ splice site). Given that there are only four nucleotide differences between SMN1 and SMN2 in the 1.25 kb SMN1/2_E constructs (Fig. 2B) (23, 24), it is striking that two of these base changes create consensus hnRNP A1 binding motifs.

We next wished to determine whether any of the putative hnRNP A1 binding sites in intron 7 play a role in SMN2 exon 7 exclusion. First, all potential A1 binding sites in intron 7 were mutated (T → C, at position +1 of the consensus) in both SMN1_E and SMN2_E and the mutated constructs were transfected into 293 cells (Fig. 2C). Splicing was again measured by RT-PCR, but in this case by using ³²P-dCTP to facilitate quantitation (Fig. 2D). Significantly, exon 7 splicing was enhanced (30% exon 7 inclusion in SMN2_E compared with 70% exon 7 inclusion in SMN2_E-7m) in the SMN2_E context (Fig. 2D, lanes 3 and 5), whereas no effect was observed with SMN1_E (lanes 2 and 4). These data suggest that the presence of one or more consensus hnRNP A1 binding sites in intron 7 is necessary for optimal exon 7 exclusion, at least in the SMN2_E context.

The SMN2-Specific Intron 7 A → G Transition at Position +100 Is Necessary for Efficient Exon 7 Exclusion.

We next wished to determine which of the consensus hnRNP A1 binding sites is necessary for exon 7 exclusion in SMN2_E, and especially whether the SMN2-specific site at position +100 is important. To begin to address this question, we first prepared a chimeric construct in which SMN2_E intron 7 was precisely replaced with SMN1_E intron 7 (SMN2-1_E). This process in effect creates a single-base transition, converting SMN2 +100 G to SMN1 +100 A. Plasmids were transfected into 293 cells and exon 7 splicing again measured by quantitative RT-PCR (Fig. 2E). The results revealed that this change dramatically reduced exon 7 exclusion, from 71% in SMN2_E to 8% in SMN2-1_E (compare lanes 3 and 4). To extend this result, we constructed two SMN2_E mutant derivatives with single base changes predicted to disrupt the +100 hnRNP A1 binding site (Fig. 2E) (14, 26). Both mutations (TAG → TCG or TAG → TAC; see Fig. 2E) significantly enhanced SMN2 exon 7 inclusion (Fig. 2F, compare lane 3 with lanes 5 and 6), although somewhat less effectively than the intron 7 swap (30/35% vs. 8% exclusion, respectively). Thus, our results show that four separate base changes altering the core UAG motif of the SMN2-specific hnRNP A1 consensus in intron 7 each enhanced exon 7 inclusion.

We next wished to examine whether intron 7 sequences affect exon 7 exclusion in the context of full-length intron 7. Recall that we showed that deleting ≈50% intron 7 resulted in a slight decrease (from 96% to 71%) in SMN2 exon 7 exclusion (see Figs. 1B and 2D, lanes 8 and 3, respectively), suggesting that the deleted sequences in some way enhance exon 7 exclusion. Furthermore, previous studies have suggested that SMN2-specific intron 7 sequences do not have a significant effect on exon 7 splicing (refs. 23 and 24; see also ref. 34 and Discussion). We therefore constructed an SMN2 intron 7 swap derivative, using the SMN1/2_HP plasmids, which contain full-length intron 7 (see Fig. 2E). SMN1_HP, SMN2_HP and the corresponding SMN2_HP derivative containing SMN1 intron 7 (SMN2-1_HP) were transfected into 293 cells and exon 7 splicing was measured by using quantitative RT-PCR (on the right side of Fig. 2F). Consistent with the semiquantitative RT-PCR in Fig. 1B, exon 7 was included in essentially all of the SMN1 transcripts and was excluded from ≈96% of the SMN2 transcripts (lanes 7 and 8). Most importantly, replacing SMN2 intron 7 with SMN1 intron 7 again significantly enhanced exon 7 inclusion, from ≈4% to 40% inclusion (lanes 8 and 9). Although less than the increased inclusion observed with the short intron 7-containing transcripts (which went from ≈29% to 92%; see Fig. 2F, lanes 4 and 5), these results confirm that the SMN2 intron 7-specific base change at +100 indeed plays a significant role in exon 7 exclusion.

SMN2 Intron 7 Specifically Interacts with hnRNP A1 in HeLa Nuclear Extract (NE).

We next wished to examine whether hnRNP A1 in fact binds to the SMN2-specific site in intron 7. To this end, we first used ³²P-labeled intron 7 (plus the two terminal nucleotides of exon 7) RNAs prepared from SMN1_E and SMN2_E in UV crosslinking assays with HeLa NE (ref. 29; see Methods). Wild-type SMN1/2_E RNAs and the mutant derivatives containing U → C base changes in the putative hnRNP A1 consensus sites (see Fig. 2 A and B) were analyzed (Fig. 3A). With all four RNAs (Fig. 3B, lanes 2–5), the prominent crosslinked species was the size expected of hnRNP A1 (≈37 kDa) (lanes 2–5), and the identity of the protein was confirmed by immunoprecipitation with anti-hnRNP A1 antibodies (lanes 7–10). Notably, crosslinking with the SMN2_E intron 7 RNA (lanes 3 and 8) was much stronger than with any of the other RNAs. This finding did not reflect anything unusual about this RNA, because crosslinking of background bands was similar with all four RNAs. This result strongly supports the notion that the +100 A → G transition creates a strong hnRNP A1 binding site.

Fig. 3.

hnRNP A1 in HeLa NE UV-crosslinks preferentially to SMN2 intron 7 RNA. (A) Diagram of intron 7 RNAs used in UV crosslinking and immunoprecipitation assays. Open boxes indicate exons and solid lines indicate intron 7. Solid triangles indicate consensus A1 sites, and open triangles indicate locations of TAG → CAG mutations. (B) (Left) SMN1 (lane 1) and SMN2 (lane 2) wild type and mutant (lanes 3 lane 4, respectively) intron 7 RNAs were incubated with HeLa NE, crosslinked, and analyzed by SDS/PAGE. (Right) UV crosslinked proteins were immunoprecipitated with anti-hnRNP A1 antibodies (lanes 6–10) and analyzed by SDS/PAGE. The position of hnRNP A1 is indicated on the right, and the position of size marker proteins is indicated on the left.

The SMN2-Specific Intron 7 hnRNP A1 Site Strongly and Specifically Binds hnRNP A1.

To extend the results in Fig. 3, we next prepared several ³²P-labeled short RNAs (≈23 nt) that contained single potential A1 binding sequences from intron 7 (see Fig. 4A) and tested these in UV crosslinking assays. Strikingly, the RNA that contained the +100 SMN2-specific site, but none of the others, gave rise to strong crosslinking (Fig. 4B Left). Crosslinking was abolished by an excess of cold SMN2-specific RNA, but not by the corresponding sequence from SMN1 (Fig. 4B Right), confirming that the enhanced crosslinking indeed reflected stronger RNA binding. Extending these results, the two point mutations in the core UAG sequence that increased SMN2 exon 7 inclusion in vivo (see Fig. 2 E and F) greatly reduced crosslinking (Fig. 4C). The identity of the crosslinked species was again confirmed by immunoprecipitation with anti-hnRNP A1 antibodies (Fig. 4C). Taken together, our results indicate an SMN2-specific single nucleotide change creates a high-affinity hnRNP A1 binding motif that plays a significant role in bringing about maximal exon 7 exclusion.

Fig. 4.

The SMN2-specific intronic A1 consensus site binds hnRNP A1 strongly and specifically. (A) Schematic diagram shows locations of all consensus hnRNP A1 binding sites in SMN1/2 intron 7. Partial boxes indicate exonic fragments and solid lines indicate introns. Filled diamonds indicate positions of consensus hnRNP A1 sites. Arrows with filled diamonds indicate RNAs containing a consensus A1 site, and the arrow without a diamond indicates the SMN1-specific RNA that lacks a consensus site. (B) (Left) UV crosslinking assay with indicated RNAs and HeLa NE. RNAs were incubated with NE, crosslinked, and analyzed by SDS/PAGE. The position of hnRNP A1 is indicated on the right. Positions of size marker proteins are indicated on the left. (Right) competition UV crosslinking assay. Incubation of labeled RNA (50 fmol) with NE was carried out in the presence of 10-μM or 0.1-μM concentrations of the indicated unlabeled RNAs. The position of hnRNP A1 is indicated on the right, and the size marker proteins are identical to the gel in Left. (C) (Left) RNAs containing the indicated sequences were incubated in NE, crosslinked and analyzed by SDS/PAGE. (Right) UV crosslinked proteins were immunoprecipitated with anti-hnRNP A1 antibodies (lanes 6–10) and analyzed by SDS/PAGE. The position of hnRNP A1 is indicated on the right.

Discussion

In this study, we have provided evidence that an intronic hnRNP A1 binding site, in addition to the previously characterized exonic site, is necessary for maximal SMN2 exon 7 exclusion. Combined with our previous data (29), this finding indicates that at least two sequence-specific interactions of hnRNP A1 with splicing silencer-type elements are necessary for full repression of exon 7 splicing in SMN2 transcripts. Given that both the exonic and intronic sequences were independently shown to interact strongly and specifically with hnRNP A1, we suggest that these two hnRNP A1 protein-RNA interactions cooperate to suppress SMN2 exon 7 splicing. Below, we discuss possible mechanisms by which the two protein-RNA interactions lead to exon 7 exclusion and the implications of our results with respect to SMN expression and to splicing control more generally.

A key mechanistic implication of our findings is that two distinct, sequence-specific protein-RNA interactions, involving high-affinity hnRNP A1 binding sites in exon 7 and intron 7, cooperate to suppress SMN2 exon 7 splicing. hnRNP A1 has been observed to form homodimers by biochemical (35, 12) and structural (36, 37) studies. We suggest that the ability of hnRNP A1 to dimerize facilitates formation of a loop structure in SMN2 premRNA that suppresses exon 7 splicing. This model is related to the original loop-out model proposed by Chabot and colleagues (12, 13). In the SMN2 exon 7 case, however, one exonic and one intronic high-affinity hnRNP A1 site are necessary for looping, instead of two intronic sites in the case of hnRNP A1 exon 7b. In both instances, though, an RNA loop formed by interaction between distant hnRNP A1 molecules (separated by 148 bases in SMN2 premRNA) is suggested to facilitate exon exclusion.

How does the proposed RNA structure repress exon 7 splicing? The hnRNP A1 binding sites do not affect U1 snRNP complex assembly at the exon 7 5′ splice site (T.K. and J.L.M., unpublished data), suggesting that the mechanism of exon 7 exclusion involves a step subsequent to early spliceosome assembly. This finding is consistent with studies with hnRNP A1 premRNA (38, 12) and with other studies proposing a similar mechanism for PTB-dependent exon suppression (39). Thus, looping out of RNA between occupied hnRNP1 A1 binding sites likely does not affect early snRNP-RNA interactions but rather subsequent steps, such as cross-talk between U1 and U2 snRNPs.

Cooperation between an hnRNP A1-dependent ESS and an intronic site has been described in HIV tat exon 3 splicing (11, 32). However, in this case the intronic site is located within the polypyrimidine tract at the 3′ splice site, which suggests that repression is due to competition between hnRNP A1 and U2AF binding to the polypyrimidine tract. In SMN2 exon 7 splicing, there are no hnRNP A1 binding sites in the vicinity of either proximal 5′ or 3′ splice sites, consistent with the idea that the mechanism of inhibition is distinct from that is used in tat exon 3 repression. Our results thus illustrate an important principle: Repressive elements can be localized at diverse positions relative to splicing signals and lead to inhibition by distinct mechanisms.

Studies have provided evidence that intronic elements are involved both positively and negatively in control of SMN exon 7 splicing (34, 40, 43). However, none of these elements involve SMN2-specific sequences, and thus they likely contribute to the inherent strength of the exon 7 5′ and 3′ splice sites. For example, the element intronic splicing silencer (ISS)-N1, identified by Singh et al. (40), was suggested to interfere with the function of TIA-1, a factor that helps U1 snRNP assembly at weak 5′ splice sites (41, 42). This sequence and two other less characterized elements (43, 34) appear to play a more general role in facilitating the efficiency of exon 7 splicing. For example, they may constitute binding sites for other hnRNP proteins. In this regard, it is intriguing that hnRNP F/H binding sites near 5′ splice sites can cooperate with hnRNP A1-dependent ESSs to repress a targeted exon (44, 45). All these studies together suggest that the splicing efficiency of SMN 1/2 exon 7 is determined by multiple protein–RNA interactions.

Our data has provided evidence that a second rare nucleotide difference between SMN1 and SMN2 contributes to SMN2 exon 7 exclusion. Two studies examined the role of intron 7 sequences by analyzing splicing of transcripts that contained SMN1/SMN2 intron 7 swaps and reached the opposite conclusion (23, 24). However, these experiments used nonquantitative RT-RCR and did not consider smaller effects that could be detected by quantitative analyses. In fact, a third study that used a more quantitative RT-PCR analysis did reveal a slight increase in exon 7 inclusion when SMN2 intron 7 was replaced with its SMN1 counterpart (34). It is noteworthy, however, that we observed the strongest effect of the SMN2 +100 site in the context of a shortened intron 7, which also resulted in slightly less efficient exon 7 exclusion (compared with the derivative containing full-length intron 7). A reasonable explanation for both of these observations is that the 225 nucleotides deleted in the shortened intron 7 contains sequences that contribute to exclusion, both increasing exclusion in the full-length context and partially compensating when the +100 hnRNP A1 site is mutated. We suggest that the sequence elements include additional hnRNP A1 binding sites. Indeed, this region contains four potential A1 binding sites, and three of these are capable of binding hnRNP A1, two with high affinity (T.K. and J.L.M., unpublished data). In any event, our data has established that intron sequences, and especially the SMN2-specific +100 transition, contribute to full exon 7 exclusion.

Recent bioinformatics studies have shown that UAG motifs, the core of the hnRNP A1 binding site, are frequently found in pseudo exonic regions within introns in the human genome, leading to the suggestion that this motif plays a significant role in constitutive splicing, for example in intron definition (46, 47). This idea is supported by recent studies showing that hnRNP A1 binding motifs are more frequently found in noncoding regions than in coding regions, perhaps playing a role in compaction of long intronic sequence by forming sequential loop structures (48). Consistent with the idea that UAG motifs play important roles in the fidelity of intron recognition and in ensuring efficient splicing, this sequence is frequently lacking in Fugu introns, which offers an explanation of why Fugu introns cannot be spliced when introduced into human cells (49). Our results extend these findings by showing that an intronic UAG-based site can cooperate with a similar exonic sequence to bring about exon exclusion. We suggest that this reflects the incorporation of the exon into an intronic “network” that is compacted and defined based on binding and interaction of multiple hnRNP A1 molecules (and perhaps other hnRNPs; see ref. 48).

In summary, our data has identified a second SMN2-specific base change that creates an hnRNP A1 binding site and provided unexpected evidence that it plays an important role in exon 7 exclusion. That two of the rare base changes between SMN1 and SMN2 are located in relatively close proximity and create high-affinity hnRNP A1 binding sites is intriguing. It is conceivable that the human-specific SMN2 gene provides a buffer that allows up-regulation of SMN protein in cells or tissues with reduced hnRNP A1 levels. Although studies examining SMN protein levels in different human tissues have not been reported, it is known that there are variations in hnRNP A1 protein concentration (50–52). More generally, our results have highlighted the importance of interactions between regulatory elements across intron/exon boundaries and provided new insight into how hnRNP A1 can modulate splicing.

Methods

Plasmids.

All SMN1 and SMN2 derivatives were made from the original full-length (exons 6–8) genomic SMN1 and SMN2 constructs described in ref. 29. All deletion constructs (shown in Fig. 1A) were produced by using standard procedures. SMN1/2_E constructs were used for most experiments. Multiple point mutations (TAG → CAG) in intronic sequences were introduced as described in ref. 53, and single point mutations were also created as described (29). Precise intron deletion mutations (Δint6 and Δint7) were created by back-to-back PCR (29). Briefly, PCR was performed with a reverse primer for exon 6 and a forward primer for exon 7 to create Δint6 and a reverse primer for exon 7 and forward primer for exon 8 to create Δint7, by using Vent DNA polymerase, and blunt-ended PCR products were ligated with T4 DNA ligase. To prepare intron 7 swap constructs, we PCR-amplified full-length SMN1 intron 7 or the short version from SMN1_E intron 7, then ligated them with the SMN2 Δint7 PCR products described above. All SMN1/2-derived constructs were digested with XhoI and XbaI and subcloned into pcDNA3 (Invitrogen, Carlsbad, CA).

We generated all template plasmids for in vitro transcription by PCR with a proof-reading DNA polymerase (Pfx DNA polymerase; Invitrogen) and amplified intron 7 of SMN1/2_E and their derivatives by PCR to add restriction enzyme sites at both ends (HindIII at 5′ and XbaI at 3′). These were digested with both enzymes, gel-purified and cloned into pGEM4Z (Promega, Madison, WI). For short RNA preparation, we slowly annealed double-strand DNA oligonucleotides (Invitrogen) that contained potential hnRNP A1 binding motifs and HindIII and XbaI sites and cloned them into pGEM4Z. All mutations were verified by sequencing (GENEWIZ, South Plainfield, NJ).

Transfection and Splicing Assays.

We transiently transfected SMN1/2-derived plasmids into 293 cells with Lipofectamine 2000 (Invitrogen) as described (29). After 48 h, we isolated total RNA and reverse-transcribed 1-μg RNA, using 0.5 μg of oligo (dT)_18–20 primer and SuperScript II RT (Invitrogen). Resultant cDNAs were PCR amplified with an SMN-specific forward primer and a plasmid-specific reverse primer, and reactions were terminated during the linear phase. PCR products were resolved on 2% agarose gels and visualized by using ethidium bromide staining. For quantitative PCR assays, we amplified 2 μl of cDNAs in 50 μl of standard reaction mixtures containing 5 μCi (1 Ci = 37 GBq) [α-³²P]dCTP (3,000 Ci/mmol; Amersham, Pittsburgh, PA) for 20 cycles. PCR products were resolved on gels buffered with 5% polyacrylamide and 90 mM Tris/89 mM boric acid/2 mM EDTA, pH 8.3, exposed to x-ray film, and quantitated by using a PhosphorImager (Molecular Dynamics, Piscataway, NJ).

UV Crosslinking and Immunoprecipitation Assays.

We carried out UV (UV) crosslinking assays as described in ref. 29, with some modifications. Intron 7 or short RNAs were synthesized with [³²P]CTP and [³²P]UTP from XbaI-linearized plasmids and gel-purified them. We incubated 40–50 fmol (5 × 10⁵ c.p.m.) of RNA with 5 μl of HeLa NE plus 100 μg of tRNA in 10 μl of reaction mixtures under splicing conditions, omitting ATP and creatine phosphate. Reaction mixtures were routinely incubated at 30°C for 30 min or for an additional 10 min with 1% Empigen BB when long RNAs (intron 7, 269 nucleotides) were used. Samples were then irradiated with UV light by using a Stratalinker (Stratagene, La Jolla, CA), treated with RNase A (10 μg/ml USB; Sigma, St. Louis, MO), and proteins resolved by 10% SDS/PAGE. For competition assays, we prepared large amounts of unlabeled RNA by in vitro transcription with T7 RNA polymerase in 200 μl of reaction mixtures and gel-purified them (54). The indicated amounts of cold RNA were added to NE, followed by the ³²P-labeled RNA. Immunoprecipitation was performed as described (29) with the following antibodies: 12CA5 for hemagglutinin epitope and 9H10 for hnRNP A1 (Immuquest, Ingleby Barwick, U.K.) (33).

Abbreviations

ESE

exonic splicing enhancer element

ESS

exonic splicing silencer

hnRNP

heterogeneous nuclear ribonucleoprotein

NE

nuclear extract

snRNP

small nuclear ribonucleoprotein.