A class of human exons with predicted distant branch points revealed by analysis of AG dinucleotide exclusion zones
- Clare Gooding†1,
- Francis Clark†2,
- Matthew C Wollerton1,
- Sushma-Nagaraja Grellscheid1,
- Harriet Groom1 and
- Christopher WJ Smith1Email author
© Gooding et al.; licensee BioMed Central Ltd. 2006
Received: 26 July 2005
Accepted: 28 November 2005
Published: 13 January 2006
The three consensus elements at the 3' end of human introns - the branch point sequence, the polypyrimidine tract, and the 3' splice site AG dinucleotide - are usually closely spaced within the final 40 nucleotides of the intron. However, the branch point sequence and polypyrimidine tract of a few known alternatively spliced exons lie up to 400 nucleotides upstream of the 3' splice site. The extended regions between the distant branch points (dBPs) and their 3' splice site are marked by the absence of other AG dinucleotides. In many cases alternative splicing regulatory elements are located within this region.
We have applied a simple algorithm, based on AG dinucleotide exclusion zones (AGEZ), to a large data set of verified human exons. We found a substantial number of exons with large AGEZs, which represent candidate dBP exons. We verified the importance of the predicted dBPs for splicing of some of these exons. This group of exons exhibits a higher than average prevalence of observed alternative splicing, and many of the exons are in genes with some human disease association.
The group of identified probable dBP exons are interesting first because they are likely to be alternatively spliced. Second, they are expected to be vulnerable to mutations within the entire extended AGEZ. Disruption of splicing of such exons, for example by mutations that lead to insertion of a new AG dinucleotide between the dBP and 3' splice site, could be readily understood even though the causative mutation might be remote from the conventional locations of splice site sequences.
Pre-mRNA splicing is an essential step in eukaryotic gene expression as well as an important regulatory point via the process of alternative splicing [1–4]. Removal of introns and splicing together of exons is essential for the generation of functional mRNAs from pre-mRNAs. The importance of splicing is attested to by the observation that at least 15% of human genetic diseases are caused by mutations within the consensus sequence elements at the exon-intron boundaries, which are important for specifying the splice sites [5–7]. The 5' splice site consists of a nine-nucleotide consensus containing the invariant GU dinucleotide at the start of the intron. At the 3' end of the intron, usually within about 40 nucleotides upstream of the exon, there are three elements (in 5' to 3' order): a branch point sequence (BPS); a polypyrimidine tract (PPT); and the 3' splice site itself, which consists of the invariant AG dinucleotide at the end of the intron, usually preceded by a pyrimidine residue. Recognition of these consensus elements by various trans-acting protein and RNA splicing factors leads to assembly of the spliceosome, within which the two chemical steps of splicing occur . In the first step the 2'-OH group of the branch point adenosine attacks the 5' splice site, leading to formation of the 5' exon and the intron lariat intermediates. In the second step, the 3'-OH of the 5' exon attacks the 3' splice site, leading to production of the spliced RNA and the excised intron, still in the lariat configuration.
Although the consensus splice site elements are essential, they are degenerate in many positions, and have insufficient information content to specify correctly the ends of long metazoan introns . This deficit is partly addressed by the presence of auxiliary splicing enhancer sequences, commonly found within exons (exonic splicing enhancers), which activate splicing of adjacent splice sites [9, 10]. A number of RNA binding (for example ) and functional SELEX (selective evolution of ligands by exponential enrichment) experiments [12–14], as well as computational analyses [15, 16], have been used to identify various classes of exonic splicing enhancers (see Matlin and coworkers  for a discussion).
We devised a simple algorithm that can be used to locate putative dBPs. First, we define the AGEZ upstream of each exon by conducting a 3' to 5' search from the 3' splice site for the first upstream AG. In the small number of cases in which AG dinucleotides exist before -12, we ignore them and continue the search for the first AG beyond -12. We then search for probable candidate BPs in a region defined by the AGEZ but also including a further approximate 15 nucleotides upstream. This additional 15 nucleotides is also considered because AGs very close to the BPS can be bypassed by the spliceosome during step 2 of splicing [28, 30]. Candidate dBPs are identified by consensus sequence (see Materials and methods, below) and by the presence of an adjacent PPT, and are often close to the 5' end of the AGEZ. Here, we have applied this approach globally by classifying human exons according to the size of their AGEZ. We find that there is an excess of exons with large AGEZ, and that putative dBP exons exhibit a higher than average prevalence of alternative splicing.
Analyzing introns for AG exclusion zones
Comparison of the mod-EZ1 and mod-EZ2 profiles revealed the curves for each dinucleotide to be (visually) very similar in all cases except for AG (Figure 2; compare blue and red lines). For the AG dinucleotides there was a readily identifiable shoulder on the mod-EZ1 distribution at higher values (≥ 100 nucleotides) compared with the mod-EZ2 distribution. There are 279 mod-AGEZ1 exons at 100 nucleotides or greater compared with 148 for the mod-AGEZ2 curve, giving a χ2 value of 116 (P ≈ 0). This confirms the visual impression that the mod-AGEZ1 and mod-AGEZ2 distributions are significantly different. The excess of exons with large AGEZ1 represents a group of potential dBP exons. We note that some other dinucleotides also exhibit lesser but still statistically significant differences under equivalent analysis (in particular bias toward TC and CT in the mod-EZ1 region; further details may be found under Materials and methods, below). As an initial test of whether the excess of the exons with mod-AGEZ1 ≥ 100 are associated with dBPs, we repeated the analysis of mod-AGEZ1 distributions having first split the intron data-set into two groups according to whether or not they had an AG dinucleotide between -12 and -25 with respect to the 3' splice site. The expectation is that exons with an AG between -12 and -25 (the 'plus' group) cannot have a dBP (otherwise the additional AG would be used as the 3' splice site). Consistent with this expectation, the percentage of mod-AGEZ1 values ≥ 100 nucleotides was 0.68% for introns without an AG in the -12 to -25 region (the minus group), but only 0.23% for those with an AG. With a null hypothesis that the minus group should generate the same statistics as the plus group, we observe the null hypothesis to be false, with a χ2 of 356 (P ≈ 0).
In the data set proper there are 838 exons with AGEZ ≥ 100. We estimate that between one-half and one-fifth of these indicate dBPs (see details under Materials and methods, below); with 838 of 67,334 introns having an AGEZ1 ≥ 100, we expect that approximately 1/160 to 1/400 introns have dBPs. Taking an 'average' human gene to have eight introns, this reduces to between 1/20 and 1/50 genes having at least one dBP (as defined here).
To facilitate manual examination of large AGEZ exons, we restrict consideration to those exons with AGEZ ≥ 150 (165 cases). Our data are available online , with separate files for the starting data set, and for exons with AGEZ ≥ 150. Among the exons with AGEZ ≥ 150 were exon 11 of the human PTB gene (AGEZ = 381, IDB1087423.10917), for which we have some in vitro evidence for use of a dBP . Likewise, the equivalent exon from the neuronal specific paralog nPTB/brPTB [34, 35] was identified (AGEZ = 438, predicted branch point at -389, IDB1145220.85254). Other dBP exons from the α-tropomyosin and α-actinin genes, which were experimentally verified for the rat genes and appear to be conserved, were not in the current build of AltExtron. Many of the exons with large AGEZ (≥ 150) had a clear potential dBP located toward the upstream end of the AGEZ with no obvious candidate BPS close to the 3' splice site. For example, IDB1152764.11013 has an AGEZ of 220 with a TACTAAC sequence at -214 and an adjacent PPT. The mouse ortholog has an AGEZ of 247 and a consensus TACTAAC BPS at -214. In other cases, large AGEZs did not appear to be related to splicing, with no obvious candidate BPS/PPT toward the 5' end of the AGEZ, whereas good candidates were in the conventional location. For example, IDB1079466.8106 has an AGEZ of 369 nucleotides. However, this appears to be due to a repetitive element upstream of the 3' splice site. Because this element lacks AG dinucleotides there is a large AGEZ, and AGs further upstream are still widely spaced. The only good candidate BPS is at -17. Instructively, the mouse orthologous exon has an AGEZ of only 31 nucleotides and a predicted BPS at -24. Intermediate between these extremes are multiple examples that might have dBPs, but that will require careful experimental verification. A striking example is tyrosine phosphatase sigma (IDB1087363.1770), which has an AGEZ of 1126 (the entire intron is only 1132 nucleotides) and potential dBPs at -1079, -829 and -288. The closest potential BPS that scores above threshold is at -171. The mouse orthologous exon has an AGEZ of 229 nucleotides with a predicted dBP at -192.
Testing predicted distant branch points
Prevalence of alternative splicing in candidate distant branch point exons
All experimental examples of dBP exons are alternatively spliced [17–21], and it is our expectation that a dBP is likely to indicate that an exon is alternatively spliced under at least some circumstances. We therefore analyzed the prevalence of observed alternative splicing (as seen in the AltExtron data set) as a function AGEZ size.
As a comparison, we observed the level of acceptor site exon modification (extension or truncation at the 3' splice site) type alternative splicing events versus AGEZ (Figure 7b). The median level was around 8% and was fairly uniform. Exons with large AGEZ did not exhibit elevated levels of this type of alternative splicing event. However, the group of exons with shortest AGEZ had a 15% observed level of alternative splicing. This spike at low AGEZ values had been considerably more pronounced prior to the following: ignoring any AGs in the last 12 nucleotides of an intron in the determination of the AGEZ; and the exclusion from the analysis (for Figure 7) of acceptor sites ≤ 40 nucleotides downstream of another acceptor site (data not shown). These filtering steps removed a large number of acceptor site isoforms involving small truncations or extensions, including the class of so-called NAGNAG splicing events [31, 41] that result from competition between closely spaced AGs during step 2 of splicing [28, 42]. It is noteworthy that, even after restricting the analysis in this way, there remained a modest spike at low AGEZ values. Further examination of this phenomenon is beyond the scope of this report but will be examined thoroughly in future work.
Mutations within the AG dinucleotide exclusion zones
Characterization of exons by upstream AGEZs provides a novel perspective for branch point prediction. This approach contrasts with conventional methods, which usually search for probable branch points within a fixed distance of the 3' splice site, sometimes using a 3' to 5' polarity for the search (for example ). Although the number of exons with very large AGEZ is relatively small (165 with AGEZ of 150 nucleotides or greater in our data set), there is a much larger number of exons with AGEZ of 80 nucleotides and more (2,264 cases), which is likely to include many exons with dBPs well beyond the conventional 40 nucleotides distance from the 3' splice site.
Some dBPs can be predicted by an almost mechanical application of a 5' to 3' search from the 5' end of the AGEZ (Figures 4, 5, 6). This was the case with GABBR1 exon 23, for which the AGEZ was 287 nucleotides and a high scoring dBP, subsequently verified by mutagenesis, was located at -275 (Figure 5). PTB exon 11 was slightly more complex in that the AGEZ is 380 nucleotides and, in addition to the verified dBP at -351, there were two other high scoring potential dBPs at -384 and -369, and the latter even had an adjacent PPT (Figure 4). However, the -351 BPS was higher scoring than either of the upstream candidates and its adjacent PPT is extensive and uridine rich, whereas the predicted PPT adjacent to -369 has a number of purine interruptions. The PTB, GABBR1, and IDB1088375 systems provide an attractive illustration of the applicability of the AGEZ approach to identifying dBP exons. However, many of the other large AGEZ exons do not have such readily predictable dBPs. In some cases there are multiple potential dBPs, and in others there are few or no obvious candidates.
One of our aims in future work will be to improve the computational prediction of dBPs taking into account additional information relating to the quality of the branch point and PPT sequence, and the distance separating possible branch point and PPT elements. Some of these approaches have already been adopted . However, further improvements in prediction should be facilitated by the experimental verification of some of the more 'difficult' dBP exons. Another useful factor to consider is phylogenetic conservation. The BPS of human-mouse orthologous pairs have been found to be more highly conserved for alternative than constitutive exons . Comparison of mouse orthologs of the human exons whose dBPs we verified here (Figures 4, 5, 6) suggests that conservation of a large AGEZ can help to focus in on a dBP even when basic local alignment search tool (BLAST) alignments do not detect significant sequence matches. For example, BLAST detected only a 24 nucleotides match immediately upstream of GBBR1 exon 23, even though the mouse had an AGEZ of 264 and predicted dBP at -225 (compared with 287 and -275 for human). Another striking example, as we previously noted , is the Fugu PTB exon 11. Its AGEZ of 590 and predicted dBP at -566 is remarkable in an organism noted for its compact genome.
We have focused on the use of AGEZs to identify unusually distant BPS. However, this approach may be a generally useful first step in prediction of all BPS. Previous BPS prediction approaches have typically used an arbitrary distance upstream of the 3' splice site within which to search for potential BPS. For example, both AltExtron [31, 32] and the successful BPS procedure described by Ast and coworkers  restricted their searches to 100 nucleotides upstream of the exon. Defining the AGEZ as the first step in BPS prediction may help to focus the search zone to a much shorter region in many cases, in addition to the obvious advantage of locating dBPs that would otherwise be missed.
The significance of the group of probable dBP exons that we identified is twofold. First, we identified a group of exons with an increased probability of being alternatively spliced (Figure 7a). In contrast to computational identification of alternative splicing events by EST alignments , our approach is expected to identify some alternative splicing events for which there may be no existing experimental data. This is analogous to the use of extended regions of flanking conserved sequence as an indicator of alternative splicing [46–48]. For example, alternative splicing of PTB exon 11 was not recognized for a long time because the exon skipping event leads to NMD of the spliced product . Characterization of the probable dBP arrangement gave us an early suggestion that exon 11 may indeed be a genuine alternatively spliced exon. We expect that the initial identification of some exons as having a probable dBP may provide an initial prediction of their alternative splicing, and that as more data becomes available the proportion of dBP exons known to be alternatively spliced will approach 100%.
The second significant point is that the dBP exons are expected to be vulnerable to mutations within the entire AGEZ. As we showed, mutations that introduce AG dinucleotides at multiple locations in the AGEZ can have highly disruptive effects. At a minimum, additional amino acids would be inserted. More catastrophically, the reading frame can be disrupted. Even in cases in which newly inserted sequence does not alter the reading frame of the spliced mRNA, the newly retained intron sequences can apparently lead to degradation. Interestingly, although mutant 3b (Figure 8) is predicted to lead to NMD, mutant 3a is not, and so degradation may result directly from the presence of the usually intronic sequences in the mRNA product. In addition, the regions between dBPs and their exons are often occupied by regulatory elements. Mutations that did not introduce AG dinucleotides could have more subtle effects by altering the appropriate regulation of exon selection, which can itself be a molecular cause of pathology [5, 6]. Notably most of the experimentally verified cases of dBPs have regulatory elements between the 3' splice site and the dBP and PPT [20, 21, 23–26]. We have shown effects upon levels of exon inclusion for multiple mutations in the extended AGEZ of α-tropomyosin exon 3 [49, 50]. Moreover, single nucleotide polymorphisms (SNPs) that affect the BPS have been shown to have a dramatic influence on the degree of exon inclusion or skipping . Given the sensitivity of dBP exons to mutation within their AGEZ, it is interesting to note that many of the exons with AGEZ ≥ 150 are within genes that are either already known to be disease associated or are in some other way of biomedical interest. So far, we are not aware of any disease causing mutations within the AGEZs of dBP exons. However, there are a number of intronic SNPs within some of them (for example, two within the AGEZ of GABBR1 exon 23; Figure 5), and it is possible that some of these could modulate alternative splicing of their associated exons. Indeed, awareness of the possibility of dBPs, as suggested by the presence of a large AGEZ, might help to improve the design of diagnostic scans. For example, exons 3, 4 and 5 of the serotonin 5-HT4 receptor gene (HTR4) have AGEZs of 149, 291, and 221 nucleotides (IDB1090103.1894, IDB1090103.27415, and IDB1090103.40737), an arrangement that is conserved in the murine ortholog. Polymorphisms in HTR4 have been associated with bipolar disorder and schizophrenia [52, 53]. However, the PCR primers used to detect polymorphisms were proximal to the exons and would have missed potentially interesting SNPs further upstream within the extended AGEZs.
An interesting feature of the regulation of dBP exons is that the small group that have been analyzed experimentally are all regulated by PTB [20, 21, 24, 54, 55]. It will be of interest to determine whether this is a general feature of dBP exons or is merely a coincidence, and also to investigate whether the dBP organization is associated with particular types of tissue specificity of regulation. The collection of extended AGEZs should also provide an enriched source of sequence elements involved in splicing regulation.
We have characterized a group of human exons based upon the large size of the AGEZ immediately upstream. We have verified the location of the dBP toward the 5' end of some of these large AGEZs. Exons with large AGEZs have a higher incidence of computationally observed alternative splicing. If the common rationale for the dBP arrangement is to have regulatory elements located between the dBP and exon, then it is likely that many or most of the dBP exons will ultimately prove to be alternatively spliced, and that initial characterization of a large AGEZ may be a predictor of alternative splicing. These exons are also of interest because they would be vulnerable to mutations within their entire AGEZ that could lead to modification or even loss of gene function. We plan to develop our data set of dBP exons further with the aims of improving our predictions for the likely location of dBPs, and of improving the annotation of the database entries to include evidence of alternative splicing, locations of known SNPs, or mutations, and the consequences of these known sequence variants.
Materials and methods
Computational base data
The altExtron data set of transcript confirmed introns and exons was used as base data (altExtron version 3; based on GenBank version 147) [31, 32, 56]. This provides a cleaned data set of transcript confirmed introns and exons in a convenient flat-file format. From these data we extracted 67,334 human introns (excluding AT-AC introns), belonging to 10,527 distinct genes. Here we are considering acceptor (3') splice sites, and refer to the splice site, downstream exon, or upstream intron as best suits the context. For each intron/exon we also extracted from altExtron its status regarding observed alternative splicing, indicating whether the exon had been observed as absent in some transcripts (a cassette exon type alternative splicing event), and/or whether there was any observed alternative acceptor splice sites (leading to exon truncation or extension).
Definition of the AGEZ, the region of interest (ROI), and modified exclusion zone values
For each exon/intron under consideration, the AGEZ was defined as the distance from the acceptor splice site to the first upstream AG, ignoring any AG found in the first 12 nucleotides (as explained under Background, see above). Note that AG dinucleotides are usually absent from this region (in >90% of cases) compared with equivalent regions downstream of acceptor sites or on either side of randomly sampled AGs within the pre-mRNA sequences (for all of which absence of flanking AGs occurs at around 40%). We also scan further upstream for the second and third AG, and downstream for the first and second downstream AG dinucleotides. A ROI was defined as spanning from the third AG upstream to the second AG downstream. Note that the ROI extends to the third upstream AG for two reasons. First, AGs that are not used as 3' splice sites can be located within a short distance downstream of the BPS [28, 30, 45]. In addition, if the second and third AGs are also widely separated, then this may indicate that the large AGEZ is not associated with a dBP (see Results, above). We proceeded to build a flat-file that, for each intron/exon, included the sequence of the ROI (broken into the upstream and downstream components), the positions of the identified AG dinucleotides, and the positions of putative PPT and U2 BPS (described below). This flat-file forms the base data set for ongoing computational work into the sequence elements that define and constrain acceptor splice sites.
For the purposes of constructing and analyzing Figure 2, modified exclusion zone (mod-EZ) values were used. For each dinucleotide, we defined mod-EZ1 by searching upstream from position -25 (relative to the 3' splice site) for the first occurrence of the dinucleotide. A further shift of -25 from the AG that terminated the corresponding mod-EZ1 was performed before commencing the search to define the mod-EZ2. This definition acts to exclude the region immediately upstream of the 3' splice site within which the PPT is most often found, and hence minimize bias in the EZ1 distributions caused by this pyrimidine-rich region. Furthermore, we have observed that the occurrence of an AG (and all other dinucleotides) is not an independent event in that the observed probability that a dinucleotide under consideration is an AG is greater if there is a nearby AG than otherwise (data not shown). Thus, by including the -25 shift at the start of the mod-EZ2 search, we treat the EZ1 and EZ2 searches equally in this regard. Note that use of these mod-EZ values is expected to be conservative in demonstrating the postulated differences between the AGEZ1 and AGEZ2 distributions. Finally, and again just for the purpose of constructing Figure 2, introns of length less than 350 nucleotides were excluded for the following reasons: first, we observe overall an increased frequency of AG dinucleotides in exons compared to introns (by close to 10%; data not shown); and, secondly, the last two nucleotides of an exon are AG in around 50% of cases. Hence, we do not want the EZ2 to extend into exonic regions, which is avoided in all but the most extreme cases by application of this length restriction. Note that 50% of mod-AGEZ1 values are ≤ 13 nucleotides and 90% are ≥ 40 nucleotides.
Base composition issues, interpretation of the mod-AGEZ1 shoulder, and predicted frequency of distant branch points
Heterogeneity in the base composition between sequences is an important and often problematic aspect of analyses of the sort presented here. In addition to the consequences arising from the dynamics of overall compositional biases, there may also be subsets of the data in which aspects of the composition are under quite specific selective pressures; indeed, the behavior of AG dinucleotides that we are examine here is precisely such an effect and it might be that there are numerous other phenomena of this flavor affecting the composition of subsets of the data. As a first step to ameliorating the consequences of such effects, we designed the analysis around the comparison of EZ1 and EZ2 distributions, in which each intron contributes one datum to each of these distributions (for each dinucleotide under consideration).
Initially we attempted to model the curves as resulting from a process analogous to a series of coin tosses, in which the chance of observing the terminating dinucleotide at each step in the scan was a constant, like that of obtaining a head in the toss of a fair coin, which remains one half irrespective of what has happened previously. This assumption allowed us to model these EZ curves as negative binomial distributions and, we thought, this would lead to straightforward quantification of the differences between the pairs of EZ1 and EZ2 distributions. We were unable to obtain good and robust fits of this model to the data, which led to exploration of the underling base composition dynamics. Importantly, we found that (for all dinucleotides) the overall chance of terminating at the next step in the scan decreased substantially as the distance scanned increased (data not shown). We came to understand this effect as resulting from heterogeneity across the gene sequences; that is, as the region scanned becomes long it becomes more probable that the gene/intron being examined has base composition dynamics that tend to exclude the dinucleotide under consideration. Although these facts prevented our analysis from proceeding along the path of fitting distributions to the data, Figure 2 suggests that - at a gross level - these dynamics affect the EZ1 and EZ2 distributions equally.
We also specifically examined the overall observed probability of an AG occurring close to either a 3' splice site AG or an AG randomly selected from within the pre-mRNA sequence. A randomly selected AG was seen to have a 60% chance of having another AG in the 12 nucleotides immediately upstream, as compared with only around 10% for a 3' splice site AG. In both cases the region 12 nucleotides immediately downstream also had a 60% chance of containing an AG. Thus, the regions immediately upstream of 3' splice site have a greatly decreased occurrence of AG, leading to a concern that comparison of simple EZ distributions (in which the EZ2 starts immediately where the EZ1 ends) was not a fair comparison. That is, the EZ2 values would tend to be shorter than they might otherwise be because scanning starts in a region that has an increased chance of containing the dinucleotide under consideration. A further related concern is that, although heterogeneity between the sequences in the data set may be contained, heterogeneity within an intron could affect our analysis. The positional bias in pyrimidine composition within introns is an obvious candidate for introducing this sort of bias. It was for these reasons that the mod-EZ definition was developed, whereby the inequality between the AGEZ1 and AGEZ2 distributions arising from starting the AGEZ2 scan immediately where the AGEZ1 terminated was avoided, and whereby much of the bias that might arise from the presence of PPTs immediately upstream of 3' splice site was also avoided.
Analysis of the curve pairs from Figure 2
Examination of Table 1 revealed a series of biases that strongly suggest that the presence of PPT sequences in some mod-AGEZ1 regions were acting to generate for GA, GG and AA lesser biases of the same sort observed for AG. For instance, if our model had been that scanning took place for a GA, then we might conclude that a GAEZ1 of ≥100 nucleotides had an approximately 158/211 (75%) observed probability of arising by chance alone, and the complementary 25% observed probability of indicating a dBP. We have no reason to suppose there is any such scanning for GA, and every reason to suppose that this difference between the mod-GAEZ1 and mod-GAEZ2 curves is a straightforward consequence of PPT sequences in some mod-GAEZ1 regions acting to bias the mod-GAEZ1 distribution to higher values. Should we attribute, for AG, this same effect to some part of the 47% that we have provisionally attributed to dBPs? For the AA, GA, and GG binucleotides there are tail mod-EZ1 excesses as for AG of 33%, 25%, and 25%, respectively, averaging to 28%; if a correction were to be applied for AG, then it would discount the 47% by this amount to 19%.
A further simple analysis sheds some light on this question. We broke the mod-AGEZ1 distribution into two parts on the basis of there being an AG in the region -12 to -25 (the 'plus' group), and the complementary 'minus' group without an AG in this region. We then looked at the fraction of the mod-AGEZ1 values ≥100 nucleotides for each of these two groups and found for the plus group 22/10,330 (0.21%) of the mod-AGEZ1 values at ≥ 100. In contrast, if there were no AG in the region -12 to -25 (the minus group), then we see 257/39,546 (0.65%) with mod-AGEZ1 ≥ 100. With a null hypothesis that the minus group should generate the same statistics as the plus group, we see the null hypothesis to be false with a χ2 of 356 (P ≈ 0). This compares with tails ≥ 100 for the mod-AGEZ1 and mod-AGEZ2 curves shown in Figure 2 of 0.57% and 0.30%, respectively. That the magnitude of the mod-AGEZ1 tail is less than that for the 'minus' group above (at 0.58% compared to 0.65%) is expected because the mod-AGEZ1 includes both the plus and minus groups (and is thus a conservative measure). That the magnitude of the mod-AGEZ2 tail is greater than the 'plus' group (at 0.30% versus 0.21%) is not statistically significant (P = 0.2), and in any case is expected because the presence of an AG in -12 to -25 increases the chance of observing an AG shortly after -25 compared with the minus group. This analysis confirms that the shoulder seen for mod-AGEZ1 is not a general feature of the sequence composition at the 3' ends of introns independent of the splicing signals, but rather is a consequence of dBPs.
There are at least two ways to think about the difference between the mod-AGEZ1 and mod-AGEZ2 curves in Figure 2. On the one hand it may seem the curves demonstrate that if an AGEZ value of 100 nucleotides or more is observed, then there is an approximately 50% (148/279 = 0.53) probability that this has arisen by chance alone, and a complementary 50% probability that a dBP is the causative factor. On the other hand it may seem that the presence of PPTs at the 3' end of introns can push the mod-AGEZ1 distribution toward higher values and thus introduce a bias that is not fully accounted for (as above). If it is accepted that the scanning model is true, along with the implication that AG dinucleotides will only in rare circumstances be found in the region from about 15 nucleotides downstream of the BPS to about 12 nucleotides upstream of the 3' splice site, then the fact that PPTs influence the base composition in this region is a consequence of the position of the BPS, and thus is immaterial in relation to interpretation of the shoulder observed for mod-AGEZ1 in comparison with mod-AGEZ2 in Figure 2. This is what we contend. If our contention is not accepted, then it is necessary to discount the portion of AGEZ1 values ≥ 100, indicating a dBP from around 50% to around 20% (as above).
Identification of putative polypyrimidine tract and branch point sequence signals
Derived weight matrix for identifying and scoring human branch point sequences
Weight matrix from 19 experimentally determined human branch point sequences
The flat files of acceptor splice sites, with information about putative PPT and U2 BPS is available online  and contains entries of the form:
GB_MAP: IDB1072296 = A06939.1 (1..5322)
ROI: 1181..1265 -> -50..34
AG: -50, -47, -29, -2, 3, 35,
PPT: -42..-33, -23..-5, 5..14,
U2BP: -31 [5.49], -24 [5.2], 11 [4.58],
'IDB1072296.1230', in this case, is the altExtron identifier for the gene (IDB1072296), with a transcript confirmed intron having a 3' splice site position (1230) being the position of the final intronic nucleotide. 'GB_MAP' gives the mapping to the GenBank entry from which this gene was derived as the accession, version, and (region). If the gene is on the complement strand in GenBank (always sense in altExtron), then the mapping will be labelled as complement. 'PROD' is the gene product as parsed from the GenBank flat files in the construction of altExtron. 'AGEZ' gives the AG exclusion zone. 'ROI' gives the ROI (see above), first in gene coordinates and then relative to the 3' splice site (with no position 0). 'AG' lists the relative positions of the AG nucleotides in the ROI, including the splice site itself at -2. 'PPT' gives the relative positions of putative PPTs in the ROI. 'U2BP' gives the relative positions of putative U2 BPS, with the bracketed number being the bit score from the weight matrix analysis. 'SEQ1' gives the intronic sequence part of the ROI with the putative PPTs in upper case (this may wrap over several lines). 'SEQ2' gives (as for SEQ1) the exonic sequence part of the ROI. Finally, 'END' is a tag helpful in file parsing that indicates the end of the record.
Frequency of alternative splicing versus AG dinucleotide exclusion zone
We examined the level of observed alternative splicing as a function of the AGEZ. Transcript confirmed introns/exons were seen to undergo alternative splicing when they were overlapped by other transcript confirmed introns/exons, and this information was derived from the altExtron flat files. We considered only those observed alternative splicing events that unambiguously fitted into one of two classes of alternative splicing: cassette exon usage (of an exon adjacent to the 3' splice site under consideration), and exon modification at the acceptor site (extension or truncation by use of competing 3' splice site). In both cases we excluded from consideration (for Figure 7) any acceptor site where another acceptor site was observed ≥ 40 nucleotides upstream. This prevents such an upstream 3' splice site defining the AGEZ in all but extreme cases, and acts to select the upstream AG in cases where an isoform pair share a BPS and PPT but differ in the use of two closely spaced AGs for the 3' splice site. Acceptor sites were grouped according to AGEZ value, initially into decade bins, with these then combined as necessary to ensure a minimum of 50 entries per group; as plotted in Figure 7, the fraction of each group with observed alternative splice isoforms was calculated. The standard error was calculated as sqrt(r·(n - r)/n), with n being the total number of introns in the group, and r being the number of these seen to undergo alternative splicing of the defined type.
For the purposes of statistically testing the hypothesis that exons with large AGEZ values are observed to be alternative exons at a higher frequency than exons with lesser AGEZ values, a cutoff of 100 nucleotides was used to define 'large'. Above this cutoff, 68 out of 235 exons were seen to be cassette exons, as compared with an expected value of 47 (on the basis of the overall average of 19.8%); this leads to a χ2 value of 9.4, giving a P value of 0.002.
Molecular and cell biology
PAC clones, RPI-271M21 (IDB1089010) and RP5-1009E24 (IDB1088375), were obtained from The Sanger Institute Clone Resources Group. PAC clones were tested for bacteriophage contamination by standard procedures . The ROI was PCR amplified using Pfu polymerase, treated with Taq polymerase to add an A overhang, and ligated into pGEM-Teasy (Promega, Madison, WI, USA) as a shuttle vector.
PCR primers for IDB1089010 were as follows: forward 5'-CCTCTAGTAGTCAACACTCACAGCAGC; and reverse 5'-GGATAGCATGTTCTTCCCAGCTGG. PCR primers for IDB1088375 were as follows: forward 5'-CCCAAAGTGTTGGGATTACAGG; and reverse 5'-CGGACGAATTCTGTCTGCGTTGAC.
Branchpoint and upstream AG mutations were carried out using QuikChange Site-Directed Mutagenesis (Stratagene, La Jolla, CA, USA) in the Teasy clones. Wild-type and mutant clones were subcloned using standard cloning techniques  from the shuttle vector either as a NotI fragment into pCAGGsEGFP [21, 59] or as an EcoRI fragment into pTS3St [26, 60]. All subcloned fragments were sequenced in their entirety to ensure that no constructs had secondary mutations.
Cell culture, transfection, and analysis of cellular RNA
HeLa cells were grown in Dulbecco's modified Eagles medium containing 10% foetal calf serum. Transient transfection was carried out using Lipofectamine (Invitrogen, Carlsbad, CA, USA), total RNA was isolated using TRI reagent (Sigma, Poole, Dorset, USA), and RT-PCR was carried out as previously described . Primers for pCAGGsEGFP based constructs were as follows: RT primer 3'CGRT, 5'-TAGTTGTACTCCAGCTT; forward 5'CGTM, 5'-GGCAAAGAATTCGCCACCA; and reverse 3'CGTM, 5'-GGGTGTCGCCCTCGAACTT. Conditions for the PCR were 30 cycles of 94°C at 30 s, an annealing temperature of 58°C for 30 s, followed by an extension at 72°C for 1 minute using a MgCl2 concentration of 1.5 mmol/l.
Primers for pTS3St based constructs were as follows: RT primer SV3'RT, 5'-GCAAACTCAGCCACAGGT; forward SV5'2, 5'-GGAGGCCTAGGCTTTTGCAAAAAG; reverse SV3'1, 5'-ACTCACTGCGTTCCAGGCAATGCT. Conditions for the PCR were 30 cycles of 94°C for 30 s, an annealing temperature of 62°C for 30 s, followed by an extension at 72°C for 1 minute using a MgCl2 concentration of 2.5 mmol/l.
Additional data files
The following additional data are included with the online version of this article: A data file containing the complete set of sequences used in the analysis (Additional data file 1) and a data file with the subset of entries with AGEZ> = 150 nt (Additional data file 2) Data are also available online . Scripts are available on request.
This work was funded by programme grant 059879 from the Wellcome Trust to C.W.J.S. We thank the Mapping Core group at the Sanger Institute for PAC clones, Alphonse Thanaraj for suggesting the collaboration between C.W.J.S and F.C., and Igor Vorechovsky for helpful comments on the manuscript. F.C. thanks the Australian Academy of Science for a travel fellowship to support a visit to the UK.
- Black DL: Mechanisms of alternative pre-messenger RNA splicing. Annu Rev Biochem. 2003, 72: 291-336. 10.1146/annurev.biochem.72.121801.161720.PubMedView ArticleGoogle Scholar
- Caceres JF, Kornblihtt AR: Alternative splicing: multiple control mechanisms and involvement in human disease. Trends Genet. 2002, 18: 186-193. 10.1016/S0168-9525(01)02626-9.PubMedView ArticleGoogle Scholar
- Matlin AJ, Clark F, Smith CW: Understanding alternative splicing: towards a cellular code. Nat Rev Mol Cell Biol. 2005, 6: 386-398. 10.1038/nrm1645.PubMedView ArticleGoogle Scholar
- Maniatis T, Tasic B: Alternative pre-mRNA splicing and proteome expansion in metazoans. Nature. 2002, 418: 236-243. 10.1038/418236a.PubMedView ArticleGoogle Scholar
- Faustino NA, Cooper TA: Pre-mRNA splicing and human disease. Genes Dev. 2003, 17: 419-437. 10.1101/gad.1048803.PubMedView ArticleGoogle Scholar
- Garcia-Blanco MA, Baraniak AP, Lasda EL: Alternative splicing in disease and therapy. Nat Biotechnol. 2004, 22: 535-546. 10.1038/nbt964.PubMedView ArticleGoogle Scholar
- Pagani F, Baralle FE: Genomic variants in exons and introns: identifying the splicing spoilers. Nat Rev Genet. 2004, 5: 389-396. 10.1038/nrg1327.PubMedView ArticleGoogle Scholar
- Burge C, Tuschl T, Sharp P: Splicing precursors to mRNAs. The RNA World. Edited by: Gestetland R, Cech T, Atkins J. 1999, Cold Spring Harbor: Cold Spring Harbor Laboratory Press, 525-560. 2Google Scholar
- Cartegni L, Chew SL, Krainer AR: Listening to silence and understanding nonsense: exonic mutations that affect splicing. Nat Rev Genet. 2002, 3: 285-298. 10.1038/nrg775.PubMedView ArticleGoogle Scholar
- Blencowe BJ: Exonic splicing enhancers: mechanism of action, diversity and role in human genetic diseases. Trends Biochem Sci. 2000, 25: 106-110. 10.1016/S0968-0004(00)01549-8.PubMedView ArticleGoogle Scholar
- Tacke R, Manley JL: The human splicing factors ASF/SF2 and SC35 possess distinct, functionally significant RNA binding specificities. EMBO J. 1995, 14: 3540-3551.PubMedPubMed CentralGoogle Scholar
- Liu HX, Zhang M, Krainer AR: Identification of functional exonic splicing enhancer motifs recognized by individual SR proteins. Genes Dev. 1998, 12: 1998-2012.PubMedPubMed CentralView ArticleGoogle Scholar
- Coulter LR, Landree MA, Cooper TA: Identification of a new class of exonic splicing enhancers by in vivo selection. Mol Cell Biol. 1997, 17: 2143-2150.PubMedPubMed CentralView ArticleGoogle Scholar
- Tian H, Kole R: Selection of novel exon recognition elements from a pool of random sequences. Mol Cell Biol. 1995, 15: 6291-6298.PubMedPubMed CentralView ArticleGoogle Scholar
- Fairbrother WG, Yeh RF, Sharp PA, Burge CB: Predictive identification of exonic splicing enhancers in human genes. Science. 2002, 297: 1007-1013. 10.1126/science.1073774.PubMedView ArticleGoogle Scholar
- Zhang XH, Chasin LA: Computational definition of sequence motifs governing constitutive exon splicing. Genes Dev. 2004, 18: 1241-1250. 10.1101/gad.1195304.PubMedPubMed CentralView ArticleGoogle Scholar
- Helfman DM, Ricci WM: Branch point selection in alternative splicing of tropomyosin pre-messenger RNAs. Nucleic Acids Res. 1989, 17: 5633-5650.PubMedPubMed CentralView ArticleGoogle Scholar
- Goux-Pelletan M, Libri D, d'Aubenton-Carafa Y, Fiszman M, Brody E, Marie J: In vitro splicing of mutually exclusive exons from the chicken β-tropomyosin gene: role of the branch point and very long pyrimidine stretch. EMBO J. 1990, 9: 241-249.PubMedPubMed CentralGoogle Scholar
- Smith CW, Nadal-Ginard B: Mutually exclusive splicing of alpha-tropomyosin exons enforced by an unusual lariat branch point location: implications for constitutive splicing. Cell. 1989, 56: 749-758. 10.1016/0092-8674(89)90678-8.PubMedView ArticleGoogle Scholar
- Southby J, Gooding C, Smith CW: Polypyrimidine tract binding protein functions as a repressor to regulate alternative splicing of alpha-actinin mutually exclusive exons. Mol Cell Biol. 1999, 19: 2699-2711.PubMedPubMed CentralView ArticleGoogle Scholar
- Wollerton MC, Gooding C, Wagner EJ, Garcia-Blanco MA, Smith CW: Autoregulation of polypyrimidine tract binding protein by alternative splicing leading to nonsense-mediated decay. Mol Cell. 2004, 13: 91-100. 10.1016/S1097-2765(03)00502-1.PubMedView ArticleGoogle Scholar
- Reed R: The organization of 3' splice-site sequences in mammalian introns. Genes Dev. 1989, 3: 2113-2123.PubMedView ArticleGoogle Scholar
- Libri D, Goux-Pelletan M, Brody E, Fiszman MY: Exon as well as intron sequences are cis-regulating elements for the mutually exclusive alternative splicing of the beta tropomyosin gene. Mol Cell Biol. 1990, 10: 5036-5046.PubMedPubMed CentralView ArticleGoogle Scholar
- Mulligan GJ, Guo W, Wormsley S, Helfman DM: Polypyrimidine tract binding protein interacts with sequences involved in alternative splicing of beta-tropomyosin pre-mRNA. J Biol Chem. 1992, 267: 25480-25487.PubMedGoogle Scholar
- Gallego ME, Balvay L, Brody E: cis-acting sequences involved in exon selection in the chicken beta-tropomyosin gene. Mol Cell Biol. 1992, 12: 5415-5425.PubMedPubMed CentralView ArticleGoogle Scholar
- Gooding C, Roberts GC, Moreau G, Nadal-Ginard B, Smith CW: Smooth muscle-specific switching of alpha-tropomyosin mutually exclusive exon selection by specific inhibition of the strong default exon. EMBO J. 1994, 13: 3861-3872.PubMedPubMed CentralGoogle Scholar
- Smith CW, Porro EB, Patton JG, Nadal-Ginard B: Scanning from an independently specified branch point defines the 3' splice site of mammalian introns. Nature. 1989, 342: 243-247. 10.1038/342243a0.PubMedView ArticleGoogle Scholar
- Smith CW, Chu TT, Nadal-Ginard B: Scanning and competition between AGs are involved in 3' splice site selection in mammalian introns. Mol Cell Biol. 1993, 13: 4939-4952.PubMedPubMed CentralView ArticleGoogle Scholar
- Liu ZR, Laggerbauer B, Luhrmann R, Smith CW: Crosslinking of the U5 snRNP-specific 116-kDa protein to RNA hairpins that block step 2 of splicing. RNA. 1997, 3: 1207-1219.PubMedPubMed CentralGoogle Scholar
- Chua K, Reed R: An upstream AG determines whether a downstream AG is selected during catalytic step II of splicing. Mol Cell Biol. 2001, 21: 1509-1514. 10.1128/MCB.21.5.1509-1514.2001.PubMedPubMed CentralView ArticleGoogle Scholar
- Clark F, Thanaraj TA: Categorization and characterization of transcript-confirmed constitutively and alternatively spliced introns and exons from human. Hum Mol Genet. 2002, 11: 451-464. 10.1093/hmg/11.4.451.PubMedView ArticleGoogle Scholar
- The altExtron dataset. [http://bioinformatics.org.au/altExtron/]
- A class of human exons with predicted distant branch points revealed by analysis of AG dinucleotide exclusion zones. [http://bioinformatics.org.au/dBP/]
- Markovtsov V, Nikolic JM, Goldman JA, Turck CW, Chou M-Y, Black DL: Cooperative assembly of an hnRNP complex induced by a tissue-specific homolog of polypyrimidine tract binding protein. Mol Cell Biol. 2000, 20: 7463-7479. 10.1128/MCB.20.20.7463-7479.2000.PubMedPubMed CentralView ArticleGoogle Scholar
- Polydorides AD, Okano HJ, Yang YYL, Stefani G, Darnell RB: A brain-enriched polypyrimidine tract-binding protein antagonizes the ability of Nova to regulate neuron-specific alternative splicing. Proc Natl Acad Sci USA. 2000, 97: 6350-6355. 10.1073/pnas.110128397.PubMedPubMed CentralView ArticleGoogle Scholar
- Ruskin B, Krainer AR, Maniatis T, Green MR: Excision of an intact intron as a novel lariat structure during pre-mRNA splicing in vitro. Cell. 1984, 38: 317-331. 10.1016/0092-8674(84)90553-1.PubMedView ArticleGoogle Scholar
- Mullen MP, Smith CW, Patton JG, Nadal-Ginard B: Alpha-tropomyosin mutually exclusive exon selection: competition between branchpoint/polypyrimidine tracts determines default exon choice. Genes Dev. 1991, 5: 642-655.PubMedView ArticleGoogle Scholar
- Modrek B, Lee C: A genomic view of alternative splicing. Nat Genet. 2002, 30: 13-19. 10.1038/ng0102-13.PubMedView ArticleGoogle Scholar
- Penalva LO, Lallena MJ, Valcarcel J: Switch in 3' splice site recognition between exon definition and splicing catalysis is important for sex-lethal autoregulation. Mol Cell Biol. 2001, 21: 1986-1996. 10.1128/MCB.21.6.1986-1996.2001.PubMedPubMed CentralView ArticleGoogle Scholar
- Lallena MJ, Chalmers KJ, Llamazares S, Lamond AI, Valcarcel J: Splicing regulation at the second catalytic step by Sex-lethal involves 3' splice site recognition by SPF45. Cell. 2002, 109: 285-296. 10.1016/S0092-8674(02)00730-4.PubMedView ArticleGoogle Scholar
- Hiller M, Huse K, Szafranski K, Jahn N, Hampe J, Schreiber S, Backofen R, Platzer M: Widespread occurrence of alternative splicing at NAGNAG acceptors contributes to proteome plasticity. Nat Genet. 2004, 36: 1255-1257. 10.1038/ng1469.PubMedView ArticleGoogle Scholar
- Scadden ADJ, Smith CWJ: Interactions between the terminal bases of mammalian introns are retained in inosine-containing pre-mRNAs. EMBO J. 1995, 14: 3236-3246.PubMedPubMed CentralGoogle Scholar
- Spritz RA, Jagadeeswaran P, Choudary PV, Biro PA, Elder JT, deRiel JK, Manley JL, Gefter ML, Forget BG, Weissman SM: Base substitution in an intervening sequence of a beta+-thalassemic human globin gene. Proc Natl Acad Sci USA. 1981, 78: 2455-2459.PubMedPubMed CentralView ArticleGoogle Scholar
- Zeniou M, Gattoni R, Hanauer A, Stevenin J: Delineation of the mechanisms of aberrant splicing caused by two unusual intronic mutations in the RSK2 gene involved in Coffin-Lowry syndrome. Nucleic Acids Res. 2004, 32: 1214-1223. 10.1093/nar/gkh272.PubMedPubMed CentralView ArticleGoogle Scholar
- Kol G, Lev-Maor G, Ast G: Human-mouse comparative analysis reveals that branch-site plasticity contributes to splicing regulation. Hum Mol Genet. 2005, 14: 1559-1568. 10.1093/hmg/ddi164.PubMedView ArticleGoogle Scholar
- Philipps DL, Park JW, Graveley BR: A computational and experimental approach toward a priori identification of alternatively spliced exons. RNA. 2004, 10: 1838-1844. 10.1261/rna.7136104.PubMedPubMed CentralView ArticleGoogle Scholar
- Sorek R, Ast G: Intronic sequences flanking alternatively spliced exons are conserved between human and mouse. Genome Res. 2003, 13: 1631-1637. 10.1101/gr.1208803.PubMedPubMed CentralView ArticleGoogle Scholar
- Sorek R, Shemesh R, Cohen Y, Basechess O, Ast G, Shamir R: A non-EST-based method for exon-skipping prediction. Genome Res. 2004, 14: 1617-1623. 10.1101/gr.2572604.PubMedPubMed CentralView ArticleGoogle Scholar
- Gromak N, Smith CW: A splicing silencer that regulates smooth muscle specific alternative splicing is active in multiple cell types. Nucleic Acids Res. 2002, 30: 3548-3557. 10.1093/nar/gkf480.PubMedPubMed CentralView ArticleGoogle Scholar
- Gromak N, Rideau A, Southby J, Scadden AD, Gooding C, Huttelmaier S, Singer RH, Smith CW: The PTB interacting protein raver1 regulates alpha-tropomyosin alternative splicing. EMBO J. 2003, 22: 6356-6364. 10.1093/emboj/cdg609.PubMedPubMed CentralView ArticleGoogle Scholar
- Kralovicova J, Houngninou-Molango S, Kramer A, Vorechovsky I: Branch site haplotypes that control alternative splicing. Hum Mol Genet. 2004, 13: 3189-3202. 10.1093/hmg/ddh334.PubMedView ArticleGoogle Scholar
- Suzuki T, Iwata N, Kitamura Y, Kitajima T, Yamanouchi Y, Ikeda M, Nishiyama T, Kamatani N, Ozaki N: Association of a haplotype in the serotonin 5-HT4 receptor gene (HTR4) with Japanese schizophrenia. Am J Med Genet B Neuropsychiatr Genet. 2003, 121: 7-13. 10.1002/ajmg.b.20060.View ArticleGoogle Scholar
- Ohtsuki T, Ishiguro H, Detera-Wadleigh SD, Toyota T, Shimizu H, Yamada K, Yoshitsugu K, Hattori E, Yoshikawa T, Arinami T: Association between serotonin 4 receptor gene polymorphisms and bipolar disorder in Japanese case-control samples and the NIMH Genetics Initiative Bipolar Pedigrees. Mol Psychiatry. 2002, 7: 954-961. 10.1038/sj.mp.4001133.PubMedView ArticleGoogle Scholar
- Grossman JS, Meyer MI, Wang YC, Mulligan GJ, Kobayashi R, Helfman DM: The use of antibodies to the polypyrimidine tract binding protein (PTB) to analyze the protein components that assemble on alternatively spliced pre-mRNAs that use distant branch points. RNA. 1998, 4: 613-625. 10.1017/S1355838298971448.PubMedPubMed CentralView ArticleGoogle Scholar
- Wollerton MC, Gooding C, Robinson F, Brown EC, Jackson RJ, Smith CW: Differential alternative splicing activity of isoforms of polypyrimidine tract binding protein (PTB). RNA. 2001, 7: 819-832. 10.1017/S1355838201010214.PubMedPubMed CentralView ArticleGoogle Scholar
- Thanaraj TA, Stamm S, Clark F, Riethoven JJ, Le Texier V, Muilu J: ASD: the alternative splicing database. Nucleic Acids Res. 2004, D64-D69. 10.1093/nar/gkh030.Google Scholar
- Gene Service Phage Testing Assay. [http://www.geneservice.co.uk/products/clones/phage_assay.jsp]
- Sambrook J, Russell D: Molecular Cloning. A Laboratory Manual. 2001, Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press, 3Google Scholar
- Okabe M, Ikawa M, Kominami K, Nakanishi T, Nishimune Y: 'Green mice' as a source of ubiquitous green cells. FEBS Lett. 1997, 407: 313-319. 10.1016/S0014-5793(97)00313-X.PubMedView ArticleGoogle Scholar
- Gooding CG, Roberts GC, Smith CWJ: Role of an inhibitory pyrimidine-element and general pyrimidine-tract binding proteins in regulation of α-tropomyosin alternative splicing. RNA. 1998, 4: 85-100.PubMedPubMed CentralGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.