Determinants of exon 7 splicing in the spinal muscular atrophy genes, SMN1 and SMN2

Abstract

Spinal muscular atrophy is a neurodegenerative disorder caused by the deletion or mutation of the survival-of-motor-neuron gene, SMN1. An SMN1 paralog, SMN2, differs by a C-->T transition in exon 7 that causes substantial skipping of this exon, such that SMN2 expresses only low levels of functional protein. A better understanding of SMN splicing mechanisms should facilitate the development of drugs that increase survival motor neuron (SMN) protein levels by improving SMN2 exon 7 inclusion. In addition, exonic mutations that cause defective splicing give rise to many genetic diseases, and the SMN1/2 system is a useful paradigm for understanding exon-identity determinants and alternative-splicing mechanisms. Skipping of SMN2 exon 7 was previously attributed either to the loss of an SF2/ASF-dependent exonic splicing enhancer or to the creation of an hnRNP A/B-dependent exonic splicing silencer, as a result of the C-->T transition. We report the extensive testing of the enhancer-loss and silencer-gain models by mutagenesis, RNA interference, overexpression, RNA splicing, and RNA-protein interaction experiments. Our results support the enhancer-loss model but also demonstrate that hnRNP A/B proteins antagonize SF2/ASF-dependent ESE activity and promote exon 7 skipping by a mechanism that is independent of the C-->T transition and is, therefore, common to both SMN1 and SMN2. Our findings explain the basis of defective SMN2 splicing, illustrate the fine balance between positive and negative determinants of exon identity and alternative splicing, and underscore the importance of antagonistic splicing factors and exonic elements in a disease context.

Figure A1

Weight matrices and pictograms representing the relative frequency of nucleotides in the hnRNP A1 high-affinity motif derived by Burd and Dreyfuss (1994). a, Weight matrix values derived from the nonredundant set of hnRNP A1 SELEX winner motifs, calculated as by Liu et al. (1998). Positive values are in bold. b, Pictogram representations of the two versions of the hnRNP A1 motif. The height of each letter represents the relative frequency of the corresponding nucleotide at that position; colored and gray letters indicate higher-than-background and lower-than-background frequencies, respectively. T is used instead of U, for convenience.

Figure A2

SF2/ASF and hnRNP A1 matrix scores. SF2/ASF scores were calculated using ESE finder 2.0 (a) or after adjustment for the compositional bias of the functional SELEX SF2/ASF winner pool (b). hnRNP A1 scores were calculated either as matches to the Burd and Dreyfuss consensus (c) or by use of the position weight matrices shown in figure A1a, which were adjusted for the compositional bias of the hnRNP A1 SELEX winner pool (d) or were assumed to have a neutral composition (e). In each case, the highest scoring value for each sequence is shown. SMN1 constructs have a C, and SMN2 constructs a blue T, at position +6. Mutations are indicated in red lowercase, and insertions are indicated in green lowercase. Statistical analysis was performed using Prism 4.0 (GraphPad software) with the settings given in the table. a and c, Scores corresponding to the scores given in table 1 in the main article.

Figure A3

Analysis of the SMN exon 7 mutations described by Singh et al. (a). a, 66 independent mutants described by Singh et al. (a). Mutants 25–66 are SELEX winner clones, of which 33/42 maintain a high-score SF2/ASF motif (30 at the original position). These mutations impinge only on the first position of the original SF2/ASF heptamer, where either C or G maintains a high-score motif, and, therefore, only 21/42 clones would be expected to retain high scores at the original position by chance; the calculated binomial probability for ⩾30 clones is P<.004. b, Nucleotide sequence around the ESE/ESS motifs. The ESEfinder 2.0 SF2/ASF top-scoring motif for each sequence is shaded. The best fit to the hnRNP A1–binding consensus (TAGGGA/T) is underlined. SMN1 constructs have a C, and SMN2 constructs a blue T, at position +6. Mutations are indicated in red lowercase. Insertions are indicated in italics. c, Top SF2/ASF score. High-score motifs are shown in bold (threshold = 1.96). d, Number of nucleotide matches to the hnRNP A1 hexamer consensus, shown in bold when >3. e, Average inclusion percentages. w = SELEX winner sequence not directly tested (all 14 tested winner sequences show >94% inclusion [Singh et al. 2004a]).

Figure A3

Analysis of the SMN exon 7 mutations described by Singh et al. (a). a, 66 independent mutants described by Singh et al. (a). Mutants 25–66 are SELEX winner clones, of which 33/42 maintain a high-score SF2/ASF motif (30 at the original position). These mutations impinge only on the first position of the original SF2/ASF heptamer, where either C or G maintains a high-score motif, and, therefore, only 21/42 clones would be expected to retain high scores at the original position by chance; the calculated binomial probability for ⩾30 clones is P<.004. b, Nucleotide sequence around the ESE/ESS motifs. The ESEfinder 2.0 SF2/ASF top-scoring motif for each sequence is shaded. The best fit to the hnRNP A1–binding consensus (TAGGGA/T) is underlined. SMN1 constructs have a C, and SMN2 constructs a blue T, at position +6. Mutations are indicated in red lowercase. Insertions are indicated in italics. c, Top SF2/ASF score. High-score motifs are shown in bold (threshold = 1.96). d, Number of nucleotide matches to the hnRNP A1 hexamer consensus, shown in bold when >3. e, Average inclusion percentages. w = SELEX winner sequence not directly tested (all 14 tested winner sequences show >94% inclusion [Singh et al. 2004a]).

Figure A4

Correlation between exon 7 inclusion and SF2/ASF or hnRNP A1 motifs. Scores derived from figure A3 are plotted against the percentage of SMN exon 7 inclusion for the 30 mutants (fig. 1), by use of the different scoring methodologies: a, SF2/ASF ESEfinder 2.0 scores; b, SF2/ASF matrix corrected for the winner pool background; c, matches to the hnRNP A1 consensus TAGGGA/T; d, hnRNP A1 matrix corrected for the winner pool background; e, hnRNP A1 matrix with the assumption of neutral background. The data points for SMN1 and SMN2 in each plot are indicated.

Figure A5

Putative hnRNP A1 low- and high-affinity sites. The orange boxes represent hnRNP A1 positive scores within the last 2,000 nt of SMN2; the height of each box indicates the motif score (calculated using the composition-adjusted matrix in fig. A1), and the width shows the position of the hexamer motif along the sequence. The yellow box indicates the score of the low-affinity site in SMN1. The green boxes indicate the position of SMN2 exons, and the red line indicates the minimum value (3.43) of a 5/6 match to the TAGGGA/T consensus sequence.

Figure 1

Mutational analysis of SMN exon 7 splicing. A, Radioactive RT-PCR analysis of RNA from SMN mutant minigenes expressed in 293-HEK cells. Transcripts that either include or skip exon 7 are indicated. Check marks indicate the presence of a high-score SF2/ASF motif (⩾1.96), as predicted by ESEfinder 2.0 (Cartegni et al. 2003), or a ⩾4 of 6 match to the hnRNP A1 consensus motif (Burd and Dreyfuss 1994). Lane numbers correspond to the 30 minigene construct numbers listed in table 1. B, Results of multiple experiments (table 1) quantitated by phosphorimage analysis and presented as scatterplots. The percentage of exon 7 inclusion was plotted against SF2/ASF ESEfinder 2.0 scores (left) or against matches to the hnRNP A1 consensus binding motif (right). Thresholds are set at 1.96 for SF2/ASF, at between 3 and 4 matches for hnRNP A1, and at 90% for exon inclusion. The white and black symbols represent SMN1 and SMN2, respectively. Quadrants where the ESE-loss (left) and ESS-gain (right) models predict that the points should fall are shaded.

Figure 2

Binding specificity of SF2/ASF and hnRNP A1. Western blot of proteins recovered from agarose beads covalently linked to RNAs corresponding to the first 17 nt of SMN1 or SMN2. The RNA beads were incubated with identical aliquots of HeLa cell nuclear extract under splicing conditions; they were then divided into three aliquots for washing at the indicated salt concentrations, and the proteins that remained bound were recovered by heating in SDS-sample buffer. Blots were probed with monoclonal antibodies against SF2/ASF (top) or hnRNP A1 (bottom). Supernatant (sup) and bound samples shown correspond to 1/100th and 1/16th of the initial binding reactions, respectively.

Figure 3

Analysis showing that depletion of hnRNP A1/A2 in vivo relieves exon 7 skipping, irrespective of the C→T transition. A, HeLa cells were treated with siRNAs against hnRNP A1, A2, or both and then were transfected with SMN minigene plasmids. Endogenous levels of hnRNP A1 and A2 and tubulin control were measured by western blotting 24 h after transfection (upper images). Radioactive RT-PCR analysis was performed, and spliced products were visualized by ethidium-bromide staining (lower images). B, Sequences of the relevant portions of intron 6 (lowercase) and exon 7 (uppercase) in the minigenes. Mutations are underlined, and the −8 and +6 positions are boxed. C, RT-PCR products shown in panel A were quantitated on a phosphorimager. Note that the SMN1.t3 and SMN1.t4 mutants gave higher levels of inclusion in HeLa cells than in 293-HEK cells (compare with fig. 1A and table 1). D, Overexpression of SF2/ASF promotes exon 7 inclusion in SMN1.PyD. MEF cells were cotransfected with an SF2/ASF expression plasmid or vector alone (C, lane 1) and the indicated SMN minigenes. RT-PCR analysis and quantitation were performed as in panels A and C.

Figure 4

Analysis showing that overexpression of hnRNP A1 in vivo inhibits SMN1 and SMN2 exon 7 inclusion. Top, Radioactive RT-PCR analysis was performed with RNA from control mock-transfected 293-HEK cells (C, lane 1) or cells expressing T7-tagged hnRNP A1 (lane 2). To distinguish between SMN1 and SMN2 endogenous mRNAs, the cDNAs were digested with DdeI to specifically cleave SMN2 in exon 8. Products were separated on a 6% native polyacrylamide gel. The white box represents the 3′ end of exon 8 that was cleaved from SMN2 spliced products. The proportion of SMN1 and SMN2 transcripts lacking exon 7 (% skip) is shown below the autoradiogram. Bottom, Western blot of lysates from mock-transfected cells or cells expressing T7-tagged hnRNP A1. Blots were probed with anti-T7 antibody and an antibody against U2B′′ as a control.

Figure 5

Analysis showing that hnRNP A1 inhibits SMN1 and SMN2 exon 7 splicing in vitro. A, Western blot of hnRNP A1 in 1 μl of HeLa nuclear extract (NE), mock-depleted extract (m), extract depleted of hnRNP A/B proteins (ΔA1), and 50 ng of recombinant hnRNP A1 (rA1). B, In vitro splicing of SMN1 and SMN2 transcripts in 3 μl of mock- or hnRNP A/B–depleted extract supplemented with 0 (−), 170, or 340 ng of recombinant hnRNP A1. Marker lane is indicated (M).

Figure 6

Models for the effects of the SMN2 C→T transition on exon 7 splicing. Multiple positive and negative determinants influence SMN1 and SMN2 splicing. In SMN1 (A), efficient splicing is driven by the interactions of SF2/ASF with the ESE in the +6c region and of other factors with a separate enhancer downstream. There may also be an interplay between the various elements and additional, perhaps redundant, factors (“SR?”) might be involved. In this strong exon definition context, a possible hnRNP A1/A2 role and/or the action of a putative repressor (“R?”), are not sufficient to inhibit splicing. The loss of the SF2/ASF–dependent ESE in SMN2 (B) allows a secondary, non–SMN2-specific inhibitory action of hnRNP A1/A2, possibly in collaboration with “R?”. Positive interactions downstream cannot counteract this inhibitory activity. Exon 7 inclusion in SMN2 can be rescued to SMN1-like levels by other means (white arrows), such as reconstitution of the SF2/ASF–dependent ESE (or other ESEs) (C), overexpression of some splicing activators (e.g., Tra2β1) (D), down-regulation of hnRNP A1/A2 by RNAi (E), or the presence of compounds that functionally replace SR proteins or ESEs (e.g., ESSENCE or TOES) (F). Fluctuations in different cells and tissues in the relative levels of SF2/ASF, hnRNP A1/A2, and/or the additional splicing factors involved would therefore determine the precise pattern of SMN2 exon 7 splicing. Green boxes represent known ESEs. Red and orange boxes represent putative ESS and ISS elements, respectively.