An Apparent Pseudo-Exon Acts both as an Alternative Exon That ...

MOLECULAR AND CELLULAR BIOLOGY, Mar. 2006, p. 2237–2246 0270-7306/06/$08.00⫹0 doi:10.1128/MCB.26.6.2237–2246.2006 Copyright © 2006, American Society for Microbiology. All Rights Reserved.

Vol. 26, No. 6

An Apparent Pseudo-Exon Acts both as an Alternative Exon That Leads to Nonsense-Mediated Decay and as a Zero-Length Exon Sushma-Nagaraja Grellscheid† and Christopher W. J. Smith* Department of Biochemistry, University of Cambridge, 80 Tennis Court Rd., Cambridge CB2 1GA, United Kingdom Received 29 September 2005/Returned for modification 26 October 2005/Accepted 21 December 2005

Pseudo-exons are intronic sequences that are flanked by apparent consensus splice sites but that are not observed in spliced mRNAs. Pseudo-exons are often difficult to activate by mutation and have typically been viewed as a conceptual challenge to our understanding of how the spliceosome discriminates between authentic and cryptic splice sites. We have analyzed an apparent pseudo-exon located downstream of mutually exclusive exons 2 and 3 of the rat ␣-tropomyosin (TM) gene. The TM pseudo-exon is conserved among mammals and has a conserved profile of predicted splicing enhancers and silencers that is more typical of a genuine exon than a pseudo-exon. Splicing of the pseudo-exon is fully activated for splicing to exon 3 by a number of simple mutations. Splicing of the pseudo-exon to exon 3 is predicted to lead to nonsense-mediated decay (NMD). In contrast, when “prespliced” to exon 2 it follows a “zero length exon” splicing pathway in which a newly generated 5ⴕ splice site at the junction with exon 2 is spliced to exon 4. We propose that a subset of apparent pseudo-exons, as exemplified here, are actually authentic alternative exons whose inclusion leads to NMD. lation of potentially harmful truncated proteins that may have dominant-negative activity. More recently, it has become evident that NMD may have been co-opted as a component of normal gene regulation. Bioinformatic analyses indicated that ca. 35% of human alternative splicing events result in RNA isoforms that are predicted NMD substrates (21). Since NMD substrates are necessarily under-represented in expressed sequence tag (EST) databases, this represents a conservative estimate. A rationale for splicing leading to NMD is provided by examples of autoregulatory alternative splicing events that can prevent overexpression of splicing factors by inducing splicing events that lead to NMD (37, 43). We are interested in the possibility that this mechanism of gene regulation may account for some conserved “pseudo-exons.” Intronic sequences that are flanked by apparently respectable splice site sequences but that are not observed in spliced mRNA are referred to as pseudo-exons. In some cases several independent mutations are required before they are spliced (36). For this reason they have been viewed as a conceptual challenge to our understanding of how the splicing machinery distinguishes authentic exons from pseudo-exons. Recent global characterizations of exon splicing enhancer and silencer (ESE and ESS) motifs has indicated that pseudo-exons can be discriminated from genuine exons on the basis of their low ESE and high ESS contents (33, 40, 44). Nevertheless, some disease-causing point mutations activate splicing of pseudoexons (see, for example, references 7, 10, 14, 15, 25, 27, and 38), indicating that some pseudo-exons are poised to splice efficiently. Since many of these pseudo-exons contain PTCs, a plausible biological function would be quantitative gene control by alternative splicing leading to NMD. In this scenario, some apparent pseudo-exons would be authentic alternative exons that are absent from EST databases because mRNAs containing them are degraded efficiently by NMD. We have investigated one such example in the ␣-tropomyosin (TM) gene, where an apparent pseudo-exon can be spliced in conjunction with a pair of tissue specific mutually exclusive exons.

Alternative splicing of pre-mRNA is a fundamental mechanism for the regulation of gene expression in higher eukaryotes (1, 4, 19, 22, 24, 35) and is now accepted to be the rule rather than the exception, with one-half to two-thirds of all genes estimated to be alternatively spliced (26). Alternative splicing allows for the generation of more than one RNA isoform from the same gene, thus helping to bridge the gap between the number of genes in the genome and the much larger number of proteins in the proteome. Alternative splicing is often regulated in a tissue-specific or developmental manner, resulting in different transcripts being generated from the same gene in different tissue types or developmental stages. Alternative splicing also plays a less appreciated role in the quantitative regulation of gene expression by the deliberate generation of mRNA isoforms that are targeted for degradation by nonsense-mediated decay (NMD) (19, 21). NMD is a eukaryotic mRNA surveillance mechanism that detects and degrades mRNAs with premature termination codons (PTCs). Mammalian stop codons are recognized as premature if they lie more than 50 to 55 nucleotides (nt) upstream of an exonexon junction (23, 39). A multiprotein complex called the exon junction complex (EJC) is deposited about 20 nt upstream of the exon junction during splicing (20). EJCs are removed by translating ribosomes during a pioneer round of translation unless they are located more than ⬃30 nt downstream of a stop codon (18). When a stop codon is encountered, the ribosome stalls, and a termination complex is formed. If any EJCs are present downstream of this termination complex, the mRNA is marked for degradation. NMD plays an important role in maintaining the fidelity of gene expression by preventing trans-

* Corresponding author. Mailing address: Department of Biochemistry, University of Cambridge, 80 Tennis Court Rd., Cambridge CB2 1GA, United Kingdom. Phone: 44-1223-333655. Fax: 44-1223-766002. E-mail: [email protected]. † Present address: Institute of Human Genetics, International Centre for Life, Newcastle upon Tyne NE1 3BZ, United Kingdom. 2237

2238

GRELLSCHEID AND SMITH

MOL. CELL. BIOL.

FIG. 1. TM pseudo-exon. (A) Exons 2 and 3 of the TM gene are mutually exclusive. Exon 3 is selected in most cells, but in smooth muscle cells exon 3 is repressed leading to exon 2 inclusion. This repression is mediated by a number of regulatory elements, including a pyrimidine tract, referred to as DY, located 170 to 225 nt downstream of exon 3 (black rectangle). Just downstream of DY is the apparent pseudo-exon characterized in the present study (P). (B) Multiple sequence alignment (CLUSTAL; asterisks denote fully conserved positions) of intronic sequences reveals that conservation extends beyond the DY pyrimidine tract (in boldface in the rat sequence) for another 120 nt. This conserved region encompasses a suboptimal GAG 3⬘ss, immediately followed by a 5⬘ss in a “zero-length exon” arrangement (in boldface). Toward the end of the conserved region is a second 5⬘ss (AAG兩GT[C/T]TGT in boldface) defining a pseudo-exon, which is 107 nt in the rat. In all species, inclusion of the pseudo-exon to either exon 2 or 3 would introduce in-frame stop codons (boldface TGA) leading to NMD.

TM exons 2 and 3 are mutually exclusive (34, 41) (Fig. 1A). Exon 3 is spliced in many cell types, but exon 2 is only selected to high levels in smooth muscle tissues (9, 41). Selection of exon 2 results from repression of exon 3, mediated by negative regulatory elements on each side of exon 3. A pyrimidine tract lying 170 to 225 nt downstream of exon 3 and referred to as the DY element binds the splicing repressor PTB. No sequences further downstream are required for regulation (12, 13), and yet sequence conservation persists for a further ⬃120 nt (Fig. 1B). The DY element and adjacent upstream sequences were previously shown to be actively used as a 3⬘ splice site (3⬘ss) when moved to a location immediately upstream of exon 3 (13), with five A’s just upstream of DY used as branch points (Fig. 1) (C. Gooding and C. W. J. Smith, unpublished observations). The conserved sequences downstream of DY show an interesting arrangement. A suboptimal GAG 3⬘ss is followed immediately by a potential 5⬘ss (GAG兩GUGGGU). Such “zero-

length exon” (ZLE) arrangements allow for recursive splicing and are proposed as a mechanism that facilitates the removal of long introns in a series of steps (3, 17). There is another conserved 5⬘ss 107 nt downstream of the ZLE, defining a pseudo-exon. Sequence conservation drops significantly beyond this pseudo-exon (Fig. 1B). This conserved sequence arrangement therefore has the potential to splice to either TM exons 2 and 3 and then to continue to splice as a ZLE, using the 5⬘ss regenerated at its junction with exon 2 or 3. Alternatively, the entire pseudo-exon could be spliced to exon 4, in which case the inclusion of PTCs would lead to NMD. We have investigated the activity of this conserved element. We have detected cellular RNA in which the element has been spliced to TM exons 2, 3, and 4. A large number of simple mutations are able to fully activate its splicing to exon 3. Further, “presplicing” of the element to either TM exon 2 or 3 leads to distinct outcomes. When spliced to exon 2 it can be

VOL. 26, 2006

PSEUDO-EXON ACTS AS ALTERNATIVE AND ZERO-LENGTH EXON

“re-spliced” as a ZLE between exons 2 and 4, but when spliced to exon 3 the full 107-nt pseudo-exon is included, which would lead to NMD of the product. Our results suggest that this apparent pseudo-exon may be representative of a class of authentic alternative exons that are misclassified as pseudo-exons because their splicing leads to degradation of the products containing them.

MATERIALS AND METHODS Constructs. All constructs were based on pTS23D (12), which contains exons 1 to 4 of the TM gene, but with a large deletion in intron 3, which is ⬎12 kb in the native gene. Expression of the plasmid in mammalian cells is driven by the simian virus 40 early promoter and enhancer. Mutations were created by PCR mediated site-directed mutagenesis using Stratagene native Pfu polymerase. Mutations were verified by DNA sequencing, and a fragment containing the mutation was then subcloned back into the parental vector to ensure that there were no secondary mutations elsewhere in the plasmid. Cell culture and transfections. HeLa cells and PAC-1 rat pulmonary artery vascular smooth muscle cells were grown under standard conditions in Dulbecco modified Eagle medium (Gibco-BRL) supplemented with 10% (vol/vol) fetal calf serum (Sigma) and 2 mM L-glutamine (Gibco-BRL). Transient transfections were performed as described previously (42). Puromycin treatment of cells in culture was carried out by incubating at 37°C for 5 h in complete medium with 300 ng of puromycin (Sigma) ml⫺1. Rat cardiac myocytes were maintained in a medium containing 135 mM NaCl, 5.4. mM KCl, 2 mM MgCl2, 1 mM CaCl2, 10 mM glucose, and 10 mM HEPES (pH 7.35). Analysis of cellular RNA. RNA was harvested by using TRI-reagent (Sigma) and analyzed by reverse transcription-PCR (RT-PCR) as previously described (42). Rat tissue RNA was extracted from tissues frozen in liquid nitrogen by using TRI-Reagent (Sigma). PCR primers were as follows. Expression from transiently transfected cells was detected by using a plasmid-specific forward primer SV5⬘2 (5⬘-GGC CTA GGC TTT TGC AAA AAG) and TM4, a reverse primer in exon 4 (5⬘-AGA TGC TAC GTC AGC TTC AGC), in 1 mM Mg2⫹ PCR buffer (50 mM KCl, 10 mM Tris-HCl [pH 8.3], 1.0 to 2.5 mM MgCl2, 0.1% [wt/vol] gelatin) at an annealing temperature of 62°C for 30 cycles. Exon 2- and exon 3-specific PCRs were carried out with TM2F (5⬘-GGT GCT GGA GGA GCT GCA C) and TM4 in 1 mM MgCl2 PCR buffer at an annealing temperature of 60°C for 30 cycles and with E3F (5⬘-GAA GAT GAG CTG GTG TCA CTG) and TM4 in 1 mM MgCl2 PCR buffer at an annealing temperature of 60°C for 30 cycles, respectively. Endogenous RNAs with spliced junctions between the pseudo-exon and exon 2 or 3 were detected by two rounds of PCR. The first round used a forward primer in exon 1, TM1 (5⬘-GCA GAG CAG GCG GAG) and a reverse primer within the pseudo-exon, chetR1 (5⬘-GGC GAA GTT CAC TTC CAG CAG T), and the reaction was carried out in 1.5 mM MgCl2 PCR buffer at an annealing temperature of 60°C for 30 cycles. The second round was performed with exon 2- and exon 3-specific primers as described above and a nested reverse primer, chetR2 (5⬘-GAT GAC ATA CAC TTC AGC GG), in 1.5 mM MgCl2 PCR buffer at an annealing temperature of 58°C for 25 cycles. Endogenous RNAs with spliced junctions between the pseudo-exon and TM exon 4 were detected by using a forward primer in the pseudo-exon, PF8 (5⬘-CGT AAT ACG ACT CAC TAT AGG CTG GGG AAC ACG GCC TCTG) and a reverse primer in exon 4, TM4 (CAG AGA TGC TAC GTC AGC TTC AGC) in 1 mM MgCl2 PCR buffer at an annealing temperature of 58°C for 30 cycles. RNA levels were monitored by ␤-actin primers (F [5⬘-GTG GGT ATG GGT CAG AAG GAC TC] and R [5⬘-GAG CCA GGG CAG TAA TC]) in 1.5 mM MgCl2 PCR buffer at 57°C for 25 cycles or GAPDH (glyceraldehyde-3phosphate dehydrogenase) primers (F [5⬘-GAT CCA CTG GTG CTG CC] and R [5⬘-AAT TCA TTG GGG GTA GGA]) in 1.5 mM MgCl2 PCR buffer at 60°C for 25 cycles. PTB primers used for detecting the efficacy of puromycin treatment were (F [5⬘-AGC CGA GAC TAC ACA CGC] and R [5⬘-GCT TTG GGG TGT GAC TCT]) in 1.5 mM MgCl2 PCR buffer at an annealing temperature of 58°C for 25 cycles. Endogenous TM was detected by using primers TM1 and TM4, in 1 mM MgCl2 PCR buffer at an annealing temperature 60°C for 25 cycles. Since exons 2 and 3 are the same size, they were discriminated by using the restriction enzyme XhoI that cleaves in exon 2 and PvuII that cleaves the junction between exons 1 and 3. Computational analysis of putative ESS/ESE content of pseudo-exon. We used the database of pseudoexons and real internal noncoding exons and the list of putative ESS and ESE octamers as described in reference 44. This was obtained online (http://www.columbia.edu/cu/biology/faculty/chasin/xz3/).

2239

The ESE or ESS count for each sequence was determined as follows. We searched for the occurrence of all octamer motifs above the threshold z-score of 2.62. Since the occurrence of several motifs dispersed in the sequence has a different pattern for several overlapping motifs occurring in a cluster, we scored each nucleotide of the cluster as one-eighth, so that a cluster with a single octamer got a score of 0.125 ⫻ 8 ⫽ 1. The scores of all clusters in the sequence were added, and this raw count was normalized to 140 nt, the length of an average exon, to give the ESE and ESS counts for each sequence in the database. We used a Perl script written by D. Grellscheid to search each sequence for ESE/ESS motifs and compute the count. The data were plotted by using Gnuplot as shown in Fig. 6. The Perl script, as well as the ESE and ESS counts, are available on request.

RESULTS The conserved element can splice as a ZLE or a pseudoexon. We initially set out to detect cellular RNAs in which TM exon 2 or 3 had spliced to the conserved element. RT-PCR was carried out with RNA from various rat muscle tissues by using nonquantitative nested PCR (Fig. 2). Products of exon 2 splicing to the element were observed in the smooth muscle tissues aorta and uterus but not in skeletal muscle, where exon 2 selection is negligible (Fig. 2A, lanes 1 to 3). Products of exon 3 splicing to the element were seen in all three tissues tested (Fig. 2A, lanes 4 to 6), although these were at relatively low levels compared to the larger band corresponding to unspliced pre-mRNA (upper band, lanes 4 to 6). Splicing of exons 2 or 3 to the conserved element regenerates a 5⬘ss at the new junction (AAG兩GUGGGU for exon 2 and GAU兩GUGGGU for exon 3). In order to test whether these intermediates can resplice using the regenerated 5⬘ss, we made constructs from a parental minigene containing TM exons 1 to 4, where either exon 2 or 3 was “prespliced” to the conserved element (Fig. 2B). In transfected HeLa cells, the exon 2 prespliced construct (2-P) spliced exclusively using the regenerated 5’ss, generating a PCR product of the same size as that from the wild-type construct (Fig. 2C, lanes 1 and 2). In contrast, the exon 3 prespliced construct spliced exclusively using the downstream 5⬘ss, thereby including the entire 107-nt pseudo-exon. These data suggest that the conserved element can act either as a ZLE or as a pseudo-exon, depending upon the identity of the upstream mutually exclusive exon to which it is first spliced. We next sought to detect endogenous RNAs containing the pseudo-exon spliced to exon 4 (P-4). The P-4 RT-PCR product was initially detected in PAC-1 cells (Fig. 3A). Upon incubation of the cells with puromycin for 5 h to inhibit NMD, we observed a threefold increase in the P-4 product relative to ␤-actin (Fig. 3A, lanes 3 and 4). As a control for NMD inhibition, we also carried out RT-PCR on PTB RNA to look for stabilization of exon 11 skipped RNA (trPTB), which is stabilized ⬃20-fold by Upf-1 knockdown (43). Puromycin produced a 3.8fold increase in trPTB RNA in the PAC-1 cells (Fig. 3A, lanes 5 and 6). This suggests that we are probably underestimating the extent to which the P-4 product is degraded by NMD. We next analyzed RNA from various rat tissues and isolated cardiac myocytes for pseudo-exon splicing to exon 4, a splicing event which involves removal of an ⬃12-kb intron (Fig. 3B). P-4 product was readily observed in heart and skeletal muscle and cardiac myocytes (lanes 3, 6, and 7). Although products of exon 2 splicing to the pseudo-exon were observed in smooth muscle tissues such as aorta and uterus (Fig. 2A, lanes 2 and 3),

2240


FIG. 2. Dual-function ZLE and pseudo-exon. (A) Endogenous splicing intermediates were detected by nonquantitative RT-PCR from rat skeletal muscle (S), aorta (A), and uterus (U) RNA. Primers used to detect 2-P and 3-P products are indicated above. Products of pseudo-exon splicing to exon 2 (lanes 1 to 3) and exon 3 (lanes 4 to 6) and unspliced RNA are labeled alongside the gel. The asterisk indicates a PCR artifact that appears to arise by amplification of unspliced RNA with a subsequent 423-bp deletion between repeated GTCC motifs probably arising by template slippage. The upper band in lane 3 did not appear reproducibly and has not been identified. Lane M, size markers (100, 200, 300, 400, and 500 bp); lanes “⫺,” PCR no-template control. (B) Tropomyosin reporter constructs. pTS23D is wild type, and 2-P and 3-P have exons 2 and 3, respectively, “prespliced” to the ZLE/ pseudo-exon. The observed splicing patterns of these constructs are indicated by the dashed lines. (C) Constructs were transfected into HeLa cells, and splicing was analyzed by RT-PCR with the primers SV5⬘2 and TM4. Spliced products are indicated schematically at the side, and the splice patterns are summarized in panel B).

no product of pseudo-exon splicing to exon 4 was detected in gut, which is also a smooth muscle-rich tissue (Fig. 3B, lane 2). This is consistent with resplicing as a ZLE when spliced to exon 2, but full incorporation as a pseudo-exon when spliced to exon 3, as indicated by the prespliced 2-P and 3-P constructs (Fig. 2B). Point mutations activate the pseudo-exon. The most obvious feature of the pseudo-exon that might limit its splicing efficiency is the suboptimal GAG 3⬘ss. The consensus 3⬘ splice site is PyAG, with a preference at the ⫺3 position of C ⬎ U ⬎ A ⬎ G. We tested the effects of mutating the pseudo-exon 3⬘ss from GAG to CAG, AAG, or UAG. All constructs were tested for splicing in HeLa cells. The CAG and UAG mutations

MOL. CELL. BIOL.

resulted in a complete shift to use of the pseudo-exon (Fig. 4A, lanes 2 and 4 compared to lane 1), whereas AAG showed an intermediate effect (lane 3). Exon 3- and exon 2-specific PCRs on the GAG and CAG samples suggested that pseudo-exon activation also led to preferential selection of exon 3 (lane 7) and that exon 2 products were not detectable in the CAG mutant (lane 10). We next tested mutants in which G-to-A point mutations introduced new UAG motifs 3 or 6 nt upstream of the pseudo-exon. In both cases, pseudo-exon splicing was highly activated (Fig. 4B), a finding consistent with the introduction of a stronger 3⬘ss. In addition to creating stronger 3⬘ss, the ⫺3 and ⫺6 TAG mutations also separated the 3⬘ss sequence from the ZLE 5⬘ss sequence. Activation of a pseudo-exon in the human ATM gene occurs by mutations that disrupt a U1 snRNP binding site immediately downstream of the activated 3⬘ss (27). We therefore tested whether the 5⬘ss-like sequence (GUGGGU) at the 5⬘ end of the pseudo-exon interfered with recognition of the 3⬘ss. We made two mutants in the ZLE 5⬘ss (Fig. 4C). The first (⌬5⬘) had four base changes, while mutant “CT” had two changes in the ZLE 5⬘ss. Both mutations led to complete inclusion of the pseudo-exon in HeLa cells, even though the 3⬘ss was still the suboptimal GAG (Fig. 4C, compare lane 1 with lanes 2 and 3). The pseudo-exon-containing band was found by restriction digests and sequencing to be purely exon 3-based (data not shown). In another approach to understanding whether the proximity of the ZLE 5⬘ss sequence to the 3⬘ss influenced pseudoexon splicing, we tested a construct with 6nt spacer element (ATCGAT, a ClaI site) between the 3⬘ss and the ZLE 5⬘ss. This construct was originally designed to monitor the possible extent of ZLE splicing by converting a zero-length exon to 6 nt. However, a substantial shift to pseudo-exon inclusion was observed (Fig. 4D, compare lanes 1 and 2). Digestion of PCR products with ClaI, and sequencing of both bands showed that the two spliced products were either normal 1-3-4 product or a product with the expanded pseudo-exon spliced between exons 3 and 4. The effect of the insertion was therefore to activate the pseudo-exon. To distinguish whether the effect of the ⫹6 mutation was due to the separation of the ZLE 3⬘ and 5⬘ss or due to its weakening effect on the ZLE 5⬘ss (altered to gAt兩GTGGGT), we made two further constructs in which the ZLE 5⬘ss was maintained as gAG兩GTGGGT. In mutant ⫹6G the sequence ATCGAG was inserted between the 3⬘ss and ZLE 5⬘ss, while in ⫹9G ATCGATGAG was inserted (Fig. 4D). In both cases, restoring the 5⬘ss suppressed inclusion of the pseudo-exon (Fig. 4D, lanes 3 and 4), suggesting that the main effect of the ⫹6 mutation was in weakening the ZLE 5⬘ss. Another ⫹6 construct also indicated the importance of the ZLE 5⬘ss as a silencer element. The clone ⫹6G4 contains a T-to-G change at position 6 of the ZLE 5⬘ ss within the context of the ⫹6 mutation. This additional mutation further weakens the ZLE 5⬘ss and leads to greater inclusion of the pseudoexon (Fig. 4D, compare lanes 2 and 5). Sequencing of PCR products and ClaI digestion of products generated from ⫹6, ⫹9, and ⫹6G4 indicated that in no case were the 6- and 9-nt spacers spliced into RNA as discrete 6- or 9-nt exons, even though they are flanked by the ZLE splice sites (data not shown). A final pair of constructs illustrated the sensitivity of the pseudo-exon to activating mutations away from the splice sites.

VOL. 26, 2006


2241

FIG. 3. Pseudo-exon splicing to exon 4 leading to NMD. (A) Rat smooth muscle (PAC-1) cells were incubated in the presence (lanes 1, 3, and 5) or absence (lanes 2, 4, and 6) of puromycin. Pseudo-exon- to exon 4-spliced products were detected by semiquantitative RT-PCR with primer 107F in the pseudo-exon and primer 4R in exon 4 (lanes 3 and 4). RNA input levels were normalized by using beta-actin PCRs (lanes 1 and 2), indicating a threefold increase in P-4 product upon puromycin treatment. An increase of 3.2-fold was observed in an independent repeat. The effectiveness of the puromycin treatment was also monitored by observing the levels of PTB exon 11 skipping (lanes 5 and 6). The exon 11 skipped product trPTB has been shown to accumulate to 20-fold when the NMD machinery was inactivated by Upf-1 knockdown by RNAi (43). PCR primers were in PTB exons 8 and 12, so the products of exon 9 inclusion (PTB4) and skipping (PTB1) are also evident. The levels of exon 11-skipped products (trPTB1 and 4) as a proportion of total PTB products increased by 3.8-fold upon puromycin treatment and by 4.5-fold in an independent repeat. (B) Detection of pseudo-exon usage in various rat tissues. RT-PCR was carried out with primer 107F in the pseudo-exon and primer 4R in exon 4 with RNA from rat brain (B), gut (G), heart (H), liver (L), kidney (K), skeletal muscle (S), and cardiac myocytes (CM) (lanes 1 to 7, respectively). Pseudo-exon products were highest in heart (H), skeletal muscle (S), and cardiac myocytes (CM). RNA input levels were checked by using GAPDH PCRs (lower panel).

In order to circumvent the influence of NMD upon pseudoexon containing transcripts, we generated constructs with a G or C insertion between nt 11 and 12 of the pseudo-exon (Fig. 4E, ⫹12C and ⫹12G). Both insertions should render pseudoexon containing products immune to NMD. Surprisingly, the ⫹12C insertion activated pseudo-exon splicing very strongly (Fig. 4E, lanes 1 and 2), whereas the ⫹12G insertion had a negligible effect on the pseudo-exon band (lane 3). These effects are most likely explained by differential effects upon ESEs and/or ESS’s. In summary, the data of Fig. 4 demonstrate that the pseudoexon can be fully activated by a variety of simple point mutations. Taken together, with its conservation (Fig. 1), this suggests that it may be an authentic regulated exon rather than a pseudo-exon. Regulation of pseudo-exon splicing. The preceding transfection experiments were all carried out in HeLa cells, where exon 3 is uniformly selected at very high levels. In these cells the pseudo-exon was very strongly activated in the ⫹6 construct compared to wild type (Fig. 4D, lane 2). We also tested the ⫹6 construct a number of times in PAC-1 smooth muscle cells. Analysis in these cells was complicated by the fact that they splice to produce a mixture of 1-3-4 and 1-2-4 (12, 13). Moreover, the state of differentiation of the cells cannot be controlled well, and the levels of exon 2 selection vary unpredictably between 20 and 50%. However, comparison of data from a number of independent transfections of PAC-1 cells with the ⫹6 construct revealed an inverse correlation between exon 2 selection and pseudo-exon splicing. Figure 5A illustrates the splicing pattern of endogenous TM in poorly differentiated (lane 2) and well-differentiated (lane 4) cells. Analysis of ⫹6 splicing in the same cells revealed that pseudo-exon inclusion was negligible in the highly differentiated cells (lane 8), intermediate in the poorly differentiated cells (lane 6), and highest

in HeLa cells (lane 10). This inverse correlation could be explained if repression of the pseudo-exon accompanies repression of exon 3. This can readily be rationalized because the polypyrimidine tract of the pseudo-exon is the DY pyrimidine tract that represses TM exon 3 by binding PTB (13, 28). Digestion of PCR products with ClaI again indicated that only bands containing the complete pseudo-exon were ClaI sensitive and that a 6-nt exon was not included in normal-sized products containing exon 2 or 3 (data not shown). Moreover, the pseudo-exon was only included in conjunction with exon 3 and was never observed to be spliced to exon 2, a finding consistent with the possibility of resplicing at the 2-P junction (Fig. 2). As an initial test of the hypothesis that PTB might play a role in repressing the pseudo-exon, we cotransfected the ⫹6 construct with expression constructs for hnRNPG, hnRNPA1, hnRNPC, Raver1, and PTB (Fig. 5B). Cotransfection of PTB, but none of the other factors, led to repression of pseudo-exon splicing. The expression vectors did not contain a common epitope tag, so we were unable to ascertain whether the factors were expressed to comparable levels. Nevertheless, the experiment demonstrates that PTB overexpression leads to pseudoexon skipping, suggesting a basis for the coregulation of exon 3 and the pseudo-exon. The pseudo-exon has a putative ESE/ESS profile of a real exon. The data in Fig. 4 suggested that the TM pseudo-exon may be a genuine alternative exon. We therefore decided to analyze the pseudo-exon using the data set of octamers representing potential ESE and ESS motifs derived by Zhang and Chasin (44). The relative enrichment of predicted ESE and ESS motifs was shown to be able to distinguish authentic exons from pseudoexons (44). We produced a scatter plot of real noncoding exons and pseudo-exons with the ESS score on the y axis and the ESE score on the x axis (Fig. 6; see also Materials and Methods). As

2242

MOL. CELL. BIOL.


FIG. 4. The pseudo-exon is easily activated by mutations. Various mutations were generated in the tropomyosin reporter construct pTS23D and mutant constructs transfected into HeLa cells. Splicing was analyzed by RT-PCR with primers SV5⬘2 and TM4 unless otherwise stated. (A) 3⬘ss mutants. Lane 1, GAG (wild type); lane 2, CAG; lane 3, AAG; lane 4, TAG; lane 5, no PCR template. Exon-specific PCRs were also carried out for the GAG (wild type) and CAG mutant using forward primers in exon 3 (lanes 6 to 8) or exon 2 (lanes 9 to 11). (B) Mutations generating AG dinucleotides 3 and 6 nt upstream of the pseudo-exon. Lane 1, wild type; lane 2, ⫺6 TAG; lane 3, ⫺3 TAG; lane 4, no template control. (C) Mutation of the ZLE 5⬘ss. The wild-type (wt) and mutant 5⬘ss sequences are shown. (D) Effect of spacers between the ZLE 3⬘ and 5⬘ splice sites. A 6-nt spacer was introduced into the ZLE in the pTS23D construct to generate the ⫹6 construct. The constructs ⫹6G, ⫹9, and ⫹ 6G4 were derived from the ⫹6 construct as indicated. (E) Effect of single base insertion mutations ⫹12C (lane 2) and ⫹12G (lane 3).

expected, there was a clear separation of authentic and pseudoexons, with the authentic exons having a low putative ESS score and higher ESE scores. The TM pseudo-exon was clearly located in a region of the plot that was highly enriched in genuine exons. Moreover, this property was conserved with all tested TM pseudo-exons (human, chimp, rat, mouse, and dog) clustering together. We also plotted two other pseudo-exons that are activated in human disease: one from the ATM gene (27) and one from the GHR gene (25). Although both of these had a relatively low ESS count, they also had a low ESE content and, in contrast to the TM pseudo-exon, did not lie in a region that was enriched in real exons. Two additional online resources, ESEfinder (5) and RescueESE (11), also indicated that the TM

pseudo-exon is enriched in ESEs comparing with the neighboring intron sequence. The conserved profile of potential ESE and ESS motifs provides additional support to the contention that the TM pseudo-exon is in fact an authentic exon. DISCUSSION The dilemma of how the splicing machinery discriminates between authentic and pseudo-exons has in large part been resolved by the observed relative enrichment of pseudo-exons with ESS as opposed to ESE elements (40, 44). We suggest that for some apparent pseudo-exons the original dilemma does not apply at all; rather, some genuine alternative exons

VOL. 26, 2006


2243

FIG. 5. The pseudo-exon is coregulated with exon 3. (A) Levels of pseudo-exon splicing in PAC-1 smooth muscle cells and HeLa cells transfected with the wild-type pTS23D (lanes 5, 7, and 9) and ⫹6 (lanes 6, 8, and 10) constructs. Endogenous TM exon 2 inclusion in PAC-1 cells was determined by RT-PCR with TM1 and TM4 primers, followed by digestion with PvuII (lanes 2 and 4) which cuts in exon 3. PAC-1 cells include exon 2 to different levels depending on the differentiation status of the cells (compare lanes 2 and 4). HeLa cells do not express TM, so no panel is shown (N/A), but they splice constructs almost exclusively to include exon 3. Construct RNA was analyzed by RT-PCR with SV5⬘2 and TM4 primers (lanes 5 to 10). Lanes 5 and 6 result from transfection of the same cells shown in lanes 1 and 2, while lanes 7 and 8 are transfections of the same cells shown in lanes 3 and 4. (B) HeLa cells were transfected with 0.5 ␮g of wild-type pTS23D (lane 1) or ⫹6 construct (lane 2) and cotransfected with 0.5 ␮g of pGem4Z (lanes 1 and 2) or plasmids expressing hnRNPG (lane 3), hnRNPA1 (lane 4), hnRNPC (lane 5), Raver1 (lane 6), and PTB (lane 7).

may be misclassified as pseudo-exons because when spliced into mRNA they lead to NMD and hence are not observed in cDNAs or EST databases. A number of lines of evidence suggested that the TM pseudoexon might be a genuine exon. It is conserved (Fig. 1), it has a conserved profile of predicted ESEs and ESSs consistent with it being a genuine exon (Fig. 6), and it can be activated by multiple independent mutations (Fig. 4). Naturally occurring mutations of this sort would create a null allele, so the conservation suggests that the apparent pseudo-exon has a function. RNA with spliced junctions between the pseudo-exon and the flanking exons could be detected by RT-PCR from cell and tissue RNA (Fig. 2 and 3). Moreover, transgenic mice expressing a TM-based minigene with no mutations in the pseudoexon showed tissue specific variations in levels of pseudo-exon inclusion, with the highest levels in the heart (9). On the basis of our observations we would predict that a small subset of classified pseudo-exons may actually be genuine exons. Although the pseudo-exons from the ATM and GHR genes did not have a clear profile of genuine exons (Fig. 6), it would be interesting to analyze other reported human disease associated pseudo-exons (e.g., (7, 10, 14, 15, 38) to see whether they might be good candidate alternative exons. The TM pseudo-exon displays interesting bifunctionality de-

pending upon whether it is first spliced to TM exon 2 or 3. It is included as a complete exon when spliced to exon 3 (Fig. 2, 4, and 5) but respliced as a ZLE when first spliced to exon 2 (Fig. 2). The ability to resplice is consistent with the match to consensus of the different regenerated 5⬘ splice sites as reflected in Shapiro and Senepathy scores (32; data not shown) and with the higher exon splicing enhancer activity of TM exon 2 compared to exon 3 (8, 29). Although experiments with “prespliced” constructs 2-P and 3-P are suggestive of the possibility of recursive splicing, it is important to note that the same junctions created by a natural splicing event may not be available for resplicing in the same way. ZLEs have been proposed to play a role in splicing very long introns by a process of “ratcheting” (3, 17). This is unlikely to be the role of the TM ZLE/pseudo-exon. Although it is 12 kb upstream of exon 4, it is within 600 nt of both upstream exons. One possibility is that splicing via the ZLE might help to enforce exon 2 selection in SM cells, since this splicing event removes exon 3 from the pre-mRNA well before exon 4 is transcribed. A problem with this suggestion is that splicing of exon 2 arises from repression of exon 3, which involves PTB binding to the DY element that acts as the polypyrimidine tract for the pseudo-exon. Thus, under conditions where exon 2 is spliced, the pseudo-exon/ ZLE is expected to be repressed.

2244


MOL. CELL. BIOL.

FIG. 6. The pseudo-exon has the profile of a real exon. A scatter plot of ESS (y axis) and ESE (x axis) scores for 462 pseudo-exons (red crosses) and 502 noncoding exons (green crosses) was prepared (44). A randomness of 0.1 to 0.3 was introduced for the points at zero to reduce overlap. The white areas running immediately parallel to the axes are because the ESE/ESS score can be zero or between 1 and n. Pseudo-exons from the human TM (box 1), growth hormone (box 2), ATM (box 3), chimp TM (box 4), mouse TM (box 5), rat TM (box 6), and dog TM (box 7) are indicated. The ESE and ESS counts for these test pseudo-exons are shown in the table. Note that the TM pseudo-exons from all species lie in an area that is highly enriched for genuine exons.

The true extent of pseudo-exon splicing in endogenous TM genes is difficult to estimate directly. The level of ZLE splicing is even more difficult to assess since, by definition, it leaves no trace in the final RNA product. In the cells and tissues that we examined the steady-state levels of pseudo-exon containing products were sufficiently low that pseudo-exon product could not be detected using primers based in exons 3 and 4. This could be because NMD disposes of these products with very high efficiency. As an alternative approach to assessing the degree of pseudo-exon splicing, we have used morpholino oligonucleotides targeted at its 5⬘ss. Morpholinos can base pair to target RNA and influence splicing without causing degradation

(30, 31). In PAC1 cells the morpholinos caused alterations in the ratios of exon 2- and exon 3-containing products which were consistent with a proportion of exon 3 products usually splicing to the pseudo-exon and then being degraded (unpublished observations). In the future it will be important to carry out such analyses more quantitatively by using real-time RTPCR to determine the absolute alterations in the levels of exon 3 (and exon 2)-containing products under conditions where the pseudo-exon has been inhibited. This will also allow these experiments to be carried out in striated muscle cell lines, which more closely resemble the tissues in vivo that appear to have the highest levels of pseudo-exon splicing (Fig. 3B) (9).

VOL. 26, 2006


Even if we assume that pseudo-exon splicing does occur to a significant extent in some tissues, the function of such splicing would remain open to question. One speculative possibility is that its splicing might be activated by signaling pathways that respond to an excess of tropomyosin over actin. This would provide a feedback pathway, allowing coordinated production of contractile proteins, which represent a substantial biosynthetic cost in striated muscles. In the longer term the best way to test the physiological importance of this splicing pathway will be to generate knock-in alleles in mice in which splicing of the pseudo-exon has been inactivated, while the conventional protein-coding splicing pathway remains intact. Activation of pseudo-exon splicing by improvement of its 3⬘ss (Fig. 4A and B) has an obvious mechanistic basis, but the other mutations (Fig. 4C to E) suggest the presence of silencer elements. In particular the ZLE 5⬘ss acts as a silencer of pseudoexon splicing. In the human ATM gene, U1 snRNP binding to such a 5⬘ss-like sequence represses a pseudo-exon and a mutation that impairs this inhibitory U1 influence causes disease by activating the pseudo-exon (27). We were unable to suppress a number of our ZLE 5⬘ss mutants by coexpressing mutant U1 snRNAs designed to base-pair with the mutated sequences (unpublished observations). Thus, the possibility remains that the ZLE 5⬘ss might act as a splicing silencer by binding factors other than U1 snRNP. Given the overall Grichness of the 5⬘ end of the pseudo-exon, another interesting candidate is hnRNP F/H (2, 6, 16). We are currently carrying out systematic mutagenesis through the pseudo-exon in order to characterize the full complement of ESEs and ESSs that may influence its activity inclusion. Computational analysis of ESTs suggested a surprisingly high proportion of human alternative splicing produces substrates for NMD (21). The example reported here, in which a previously unrecognized exon can splice to produce an NMD substrate, emphasizes that the EST-based inferences of the extent of alternative splicing leading to NMD may indeed be touching only the tip of an iceberg. ACKNOWLEDGMENTS We thank Larry Chasin and X. H. Zhang for the octamer data set of ESE/ESS values, Clare Gooding for helpful discussions throughout the course of this project and for comments on the manuscript, Andrew Proven and Martin Bootman for the cardiac myocytes, and David Elliott for the kind use of his laboratory to undertake some of the final experiments. We are very grateful to David Grellscheid for invaluable assistance in computational analysis. This study was supported by Wellcome program grant 059879 to C.W.J.S. REFERENCES 1. Black, D. L. 2003. Mechanisms of alternative pre-messenger RNA splicing. Annu. Rev. Biochem. 72:291–336. 2. Buratti, E., M. Baralle, L. De Conti, D. Baralle, M. Romano, Y. M. Ayala, and F. E. Baralle. 2004. hnRNP H binding at the 5⬘ splice site correlates with the pathological effect of two intronic mutations in the NF-1 and TSH␤ genes. Nucleic Acids Res. 32:4224–4236. 3. Burnette, J. M., E. Miyamoto-Sato, M. A. Schaub, J. Conklin, and A. J. Lopez. 2005. Subdivision of large introns in Drosophila by recursive splicing at non-exonic elements. Genetics 170:661–674. 4. Caceres, J. F., and A. R. Kornblihtt. 2002. Alternative splicing: multiple control mechanisms and involvement in human disease. Trends Genet. 18: 186–193. 5. Cartegni, L., J. Wang, Z. Zhu, M. Q. Zhang, and A. R. Krainer. 2003. ESEfinder: a web resource to identify exonic splicing enhancers. Nucleic Acids Res. 31:3568–3571.

2245

6. Chen, C. D., R. Kobayashi, and D. M. Helfman. 1999. Binding of hnRNP H to an exonic splicing silencer is involved in the regulation of alternative splicing of the rat beta-tropomyosin gene. Genes Dev. 13:593–606. 7. Coutinho, G., J. Y. Xie, L. T. Du, A. Brusco, A. R. Krainer, and R. A. Gatti. 2005. Functional significance of deep intronic mutation in the ATM gene and evidence for an alternative exon 28a. Hum. Mutat. 25:118–124. 8. Dye, B. T., M. Buvoli, S. A. Mayer, C. H. Lin, and J. G. Patton. 1998. Enhancer elements activate the weak 3⬘ splice site of alpha-tropomyosin exon 2. RNA 4:1523–1536. 9. Ellis, P. D., C. W. Smith, and P. Kemp. 2004. Regulated tissue-specific alternative splicing of enhanced green fluorescent protein transgenes conferred by alpha-tropomyosin regulatory elements in transgenic mice. J. Biol. Chem. 279:36660–36669. 10. Eng, L., G. Coutinho, S. Nahas, G. Yeo, R. Tanouye, M. Babaei, T. Dork, C. Burge, and R. A. Gatti. 2004. Nonclassical splicing mutations in the coding and noncoding regions of the ATM gene: maximum entropy estimates of splice junction strengths. Hum. Mutat. 23:67–76. 11. Fairbrother, W. G., G. W. Yeo, R. Yeh, P. Goldstein, M. Mawson, P. A. Sharp, and C. B. Burge. 2004. RESCUE-ESE identifies candidate exonic splicing enhancers in vertebrate exons. Nucleic Acids Res. 32:W187–W190. 12. Gooding, C., G. C. Roberts, G. Moreau, B. Nadal Ginard, and C. W. J. Smith. 1994. Smooth muscle-specific switching of alpha-tropomyosin mutually exclusive exon selection by specific inhibition of the strong default exon. EMBO J. 13:3861–3872. 13. Gooding, C. G., G. C. Roberts, and C. W. J. Smith. 1998. Role of an inhibitory pyrimidine-element and general pyrimidine-tract binding proteins in regulation of ␣-tropomyosin alternative splicing. RNA 4:85–100. 14. Gorman, L., D. R. Mercatante, and R. Kole. 2000. Restoration of correct splicing of thalassemic beta-globin pre-mRNA by modified U1 snRNAs. J. Biol. Chem. 275:35914–35919. 15. Gupta, P. A., D. D. Wallis, D. Guo, T. Guidry, V. Tran-Fadulu, and D. M. Milewicz. 2003. Analysis of a fibrillin 1 pseudoexon and its effects on nonsense mediated degradation. Am. J. Hum. Genet. 73:2441. 16. Han, K., G. Yeo, P. An, C. B. Burge, and P. J. Grabowski. 2005. A combinatorial code for splicing silencing: UAGG and GGGG motifs. PLoS Biol. 3:e158. 17. Hatton, A. R., V. Subramaniam, and A. J. Lopez. 1998. Generation of alternative ultrabithorax isoforms and stepwise removal of a large intron by resplicing at exon-exon junctions. Mol. Cell 2:787–796. 18. Ishigaki, Y., X. Li, G. Serin, and L. E. Maquat. 2001. Evidence for a pioneer round of mRNA translation: mRNAs subject to nonsense-mediated decay in mammalian cells are bound by CBP80 and CBP20. Cell 106:607–617. 19. Lareau, L. F., R. E. Green, R. S. Bhatnagar, and S. E. Brenner. 2004. The evolving roles of alternative splicing. Curr. Opin. Struct. Biol. 14:273–282. 20. Le Hir, H., E. Izaurralde, L. E. Maquat, and M. J. Moore. 2000. The spliceosome deposits multiple proteins 20–24 nucleotides upstream of mRNA exon-exon junctions. EMBO J. 19:6860–6869. 21. Lewis, B. P., R. E. Green, and S. E. Brenner. 2003. Evidence for the widespread coupling of alternative splicing and nonsense-mediated mRNA decay in humans. Proc. Natl. Acad. Sci. USA 100:189–192. 22. Maniatis, T., and B. Tasic. 2002. Alternative pre-mRNA splicing and proteome expansion in metazoans. Nature 418:236–243. 23. Maquat, L. E. 2004. Nonsense-mediated mRNA decay: splicing, translation and mRNP dynamics. Nat. Rev. Mol. Cell. Biol. 5:89–99. 24. Matlin, A., F. Clark, and C. W. J. Smith. 2005. Understanding alternative splicing: toward a cellular code. Nat. Rev. Mol. Cell. Biol. 6:386–398. 25. Metherell, L. A., S. A. Akker, P. B. Munroe, S. J. Rose, M. Caulfield, M. O. Savage, S. L. Chew, and A. J. Clark. 2001. Pseudoexon activation as a novel mechanism for disease resulting in atypical growth-hormone insensitivity. Am. J. Hum. Genet. 69:641–646. 26. Modrek, B., and C. Lee. 2002. A genomic view of alternative splicing. Nat. Genet. 30:13–19. 27. Pagani, F., E. Buratti, C. Stuani, R. Bendix, T. Dork, and F. E. Baralle. 2002. A new type of mutation causes a splicing defect in ATM. Nat. Genet. 30:426–429. 28. Perez, I., C. H. Lin, J. G. McAfee, and J. G. Patton. 1997. Mutation of PTB binding sites causes misregulation of alternative 3⬘ splice site selection in vivo. RNA 3:764–778. 29. Roberts, G. C., C. Gooding, and C. W. J. Smith. 1996. Smooth muscle alternative splicing induced in fibroblasts by heterologous expression of a regulatory gene. EMBO J. 15:6301–6310. 30. Sazani, P., S. H. Kang, M. A. Maier, C. Wei, J. Dillman, J. Summerton, M. Manoharan, and R. Kole. 2001. Nuclear antisense effects of neutral, anionic and cationic oligonucleotide analogs. Nucleic Acids Res. 29:3965–3974. 31. Schmajuk, G., H. Sierakowska, and R. Kole. 1999. Antisense oligonucleotides with different backbones: modification of splicing pathways and efficacy of uptake. J. Biol. Chem. 274:21783–21789. 32. Shapiro, M. B., and P. Senapathy. 1987. RNA splice junctions of different classes of eukaryotes: sequence statistics and functional implications in gene expression. Nucleic Acids Res. 15:7155–7174. 33. Sironi, M., G. Menozzi, L. Riva, R. Cagliani, G. P. Comi, N. Bresolin, R.

2246

34. 35. 36. 37. 38. 39.


Giorda, and U. Pozzoli. 2004. Silencer elements as possible inhibitors of pseudoexon splicing. Nucleic Acids Res. 32:1783–1791. Smith, C. W. J., and B. Nadal-Ginard. 1989. Mutually exclusive splicing of alpha-tropomyosin exons enforced by an unusual lariat branch point location: implications for constitutive splicing. Cell 56:749–758. Smith, C. W. J., and J. Valca ´rcel. 2000. Alternative pre-mRNA splicing: the logic of combinatorial control. Trends Biochem. Sci. 25:381–388. Sun, H. Z., and L. A. Chasin. 2000. Multiple splicing defects in an intronic false exon. Mol. Cell. Biol. 20:6414–6425. Sureau, A., R. Gattoni, Y. Dooghe, J. Stevenin, and J. Soret. 2001. SC35 autoregulates its expression by promoting splicing events that destabilize its mRNAs. EMBO J. 20:1785–1796. Tuffery-Giraud, S., C. Saquet, S. Chambert, and M. Claustres. 2003. Pseudoexon activation in the DMD gene as a novel mechanism for Becker muscular dystrophy. Hum. Mutat. 21:608–614. Wagner, E., and J. Lykke-Andersen. 2002. mRNA surveillance: the perfect persist. J. Cell Sci. 115:3033–3038.

MOL. CELL. BIOL. 40. Wang, Z., M. E. Rolish, G. Yeo, V. Tung, M. Mawson, and C. B. Burge. 2004. Systematic identification and analysis of exonic splicing silencers. Cell 119: 831–845. 41. Wieczorek, D. F., C. W. J. Smith, and B. Nadal Ginard. 1988. The rat alpha-tropomyosin gene generates a minimum of six different mRNAs coding for striated, smooth, and nonmuscle isoforms by alternative splicing. Mol. Cell. Biol. 8:679–694. 42. Wollerton, M., C. Gooding, F. Robinson, E. Brown, R. Jackson, and C. W. J. Smith. 2001. Differential alternative splicing activity of isoforms of polypyrimidine tract binding protein. RNA 7:819–832. 43. Wollerton, M. C., C. Gooding, E. J. Wagner, M. A. Garcia-Blanco, and C. W. J. Smith. 2004. Autoregulation of polypyrimidine tract binding protein by alternative splicing leading to nonsense-mediated decay. Mol. Cell 13:91– 100. 44. Zhang, X. H., and L. A. Chasin. 2004. Computational definition of sequence motifs governing constitutive exon splicing. Genes Dev. 18:1241–1250.