|Home | About | Journals | Submit | Contact Us | Français|
Recent ChIP experiments indicate that spliceosome assembly and splicing can occur cotranscriptionally in S. cerevisiae. However, only a few genes have been examined, and all have long second exons. To extend these studies, we analyzed intron-containing genes with different second exon lengths, by ChIP as well as by whole-genome tiling arrays (ChIP-CHIP). The data indicate that U1 snRNP recruitment is independent of exon length. Recursive splicing constructs, which uncouple U1 recruitment from transcription, suggest that cotranscriptional U1 recruitment contributes to optimal splicing efficiency. In contrast, U2 snRNP recruitment as well as cotranscriptional splicing is deficient on short second exon-genes. We estimate that ≥90% of endogenous yeast splicing is post-transcriptional, consistent with an analysis of post-transcriptional snRNP-associated pre-mRNA.
Eukaryotic pre-mRNAs undergo a number of processing steps prior to export from the nucleus to the cytoplasm. Covalent modifications include 5′ capping, pre-mRNA splicing, and polyadenylation (for review, see Bentley, 2005). Some of these, capping for example, are intimately connected to transcription. Splicing, in contrast, has a more uncertain relationship to transcription, especially in yeast (Bentley, 2005; Neugebauer, 2002).
Intron removal by pre-mRNA splicing is carried out by the spliceosome, a macromolecular complex containing five small ribonucleoprotein particles (U snRNPs) and multiple non-snRNP auxiliary factors. Spliceosome assembly is directed by the pre-mRNA 5′ splice site, branch point, and 3′ splice site, after which two coupled transesterification reactions are catalyzed (Brow, 2002). Based largely on in vitro experiments, yeast spliceosome assembly begins with U1 snRNP association/base pairing with the 5′ splice site and Mud2p-BBP association with the branchpoint-3′ss region to form commitment complex (CC). In the first ATP-dependent step, U2 snRNP undergoes a base pairing interaction with the branch point in the pre-spliceosome. The U5.U4/U6 tri-snRNP then joins this U1–U2 complex to form the complete spliceosome. A number of major rearrangements ensue, which result in destabilization and release of U1 and U4 snRNPs from the spliceosome with concomitant activation (for reviews, see Brow, 2002; Burge et al., 1999; Jurica and Moore, 2003; Villa et al., 2002).
Many aspects of in vivo spliceosome assembly, splicing, and mRNP processing occur cotranscriptionally. In higher eukaryotes, RNA Pol II may directly promote cotranscriptional spliceosome assembly. The C-terminal domain (CTD) of Pol II consists of 52 repeats of a YSPTSPS heptad, which are differentially phosphorylated at serine-2 and serine-5 during transcription. This domain directly interacts with splicing factors and is probably responsible for coordination between transcription and splicing (for reviews, see Kornblihtt et al., 2004; Maniatis and Reed, 2002; Proudfoot et al., 2002). Recently, two groups described in vitro transcription-splicing systems that suggest Pol II-synthesized substrates are spliced much more efficiently than pre-synthesized or T7-derived transcripts (Das et al., 2006; Ghosh and Garcia-Blanco, 2000; Hicks et al., 2006). These reports demonstrate a functional coupling between transcription and splicing, i.e., Pol II transcription is required for efficient shunting of the pre-mRNA into the spliceosome assembly pathway. Although the coupling mechanism(s) between transcription and splicing is less well understood in yeast, there are some positive indications, and introns appear to stimulate gene expression (Furger et al., 2002; Howe et al., 2003; Juneau et al., 2006; Lacadie et al., 2006).
Recent in vivo studies of yeast spliceosome assembly have benefited from the application of chromatin immunoprecipitation (ChIP) to assess U snRNP association with nascent pre-mRNAs (Gornemann et al., 2005; Kotovic et al., 2003; Lacadie and Rosbash, 2005). The approach monitors where along a gene snRNPs are recruited to nascent pre-mRNA and associates distance along a gene with assembly time, i.e., where = when. Like the in vitro studies, the in vivo experiments depict a step-wise assembly pathway, at least CC formation followed by association of the other splicing snRNPs (Gornemann et al., 2005; Lacadie and Rosbash, 2005; Tardiff and Rosbash, 2006). Although preformed penta-snRNP complexes have been purified under low salt conditions in vitro (Stevens et al., 2002), this particle does not seem to be required in vivo and may not participate in cotranscriptional spliceosome assembly (Tardiff and Rosbash, 2006). Studies in HeLa nuclear extracts give rise to similar conclusions (Behzadnia et al., 2006). We have also recently observed cotranscriptional splicing by using a split MS2 ChIP assay (Lacadie et al., 2006).
The data describing cotranscriptional assembly and splicing in yeast is based on a limited number of intron-containing genes. Moreover, these studies have focused on genes with relatively long (>1 kb) second exons, which facilitate the ChIP analysis. However, a significant number of yeast intron-containing genes have second exons considerably smaller than 1 kb (see below). These include many ribosomal protein genes, which comprise ~40% of intron-containing yeast genes. Ribosomal proteins are generally small, and their genes therefore have short open reading frames. They are also highly transcribed and therefore generate a considerable fraction of cellular mRNA (Ares et al., 1999). This second exon size difference is non-trivial, because our studies suggested that cotranscriptional splicing occurs about 1 kb after the 3′ss (Lacadie and Rosbash, 2005). Perhaps there are specific mechanisms to ensure that spliceosome assembly is cotranscriptional. For example, Pol II elongation rates could decrease, allowing more time for the recruitment and function of splicing factors. To address this directly, we assayed cotranscriptional spliceosome assembly and splicing of genes with various second exon lengths. ChIP experiments demonstrate that cotranscriptional U1:U2 ratios are well-correlated with second exon length, suggesting a kinetic competition between spliceosome assembly and 3′ end formation. Post-transcriptional splicing complexes corroborate cotranscriptional recruitment, namely, shorter second exon genes are released from the site of transcription at an earlier stage of spliceosome assembly. Whole-genome tiling arrays provide high resolution recruitment patterns as well as more detailed information about splicing kinetics on a large number of intron-containing genes. Finally, a ChIP assay on a set of recursive splicing constructs suggests that cotranscriptional U1 snRNP recruitment contributes to splicing efficiency, whereas post-transcriptional U2/tri-snRNP recruitment is as efficient as cotranscriptional recruitment.
Recent in vivo experiments on yeast cotranscriptional spliceosome assembly and splicing have examined genes with relatively long (>1 kb) second exons, e.g., ACT1 and ECM33 (Figure S1). To examine the extent to which these are representative, we generated a histogram describing the distribution of second intron lengths of 238 annotated yeast intron-containing genes (Figure S1A). The most populated group contains second exons between 300 and 400 bp, and the median second exon length is 434 bp. These lengths are much shorter than the positions along second exons where spliceosome assembly and splicing appear to take place, i.e., maximal U2/U5 levels occur ~500 bp past the 3′ ss, and splicing occurs ~1 kb past the 3′ ss (Gornemann et al., 2005; Kotovic et al., 2003; Lacadie and Rosbash, 2005; Lacadie et al., 2006). The distinction is even more striking when relative transcription rates are considered (Figure S1B). This is largely due to ribosomal protein genes, which are intensely transcribed (Ares et al., 1999; Holstege et al., 1998; Warner, 1999) and also have short second exons. These observations suggest the possibility that only a small fraction of intron-containing nascent pre-mRNAs are sufficiently long to support cotranscriptional splicing (Figure S1B). Alternatively, short second exon genes could have altered transcription or spliceosome assembly to allow cotranscriptional splicing.
To distinguish between these two possibilities, we used a standard ChIP approach and compared U snRNP recruitment to ACT1 (long second exon: 1117 bp) and RP51A (short second exon: 407 bp). These are two well-studied intron-containing genes and have been used for in vivo as well as in vitro experiments.
On ACT1, U1 snRNP (U1C-TAP) signal peaks near the 3′ ss, whereas U2 snRNP (Lea1-TAP) peaks ~400 bp after the 3′ ss; the overall U1:U2 peak ratio is ~1:1 (Figure 1A), consistent with published results (Lacadie and Rosbash, 2005; Tardiff and Rosbash, 2006). In contrast, snRNP recruitment to RP51A is different. Although U1 peaks within the intron similar to ACT1 recruitment, U2 snRNP levels are comparatively much lower, resulting in a U1:U2 peak ratio of ~3:1 (Figure 1B). Note that it is this relative difference between U1 and U2 levels rather than the absolute enrichment that is most relevant. Absolute levels are also influenced by transcription; RP51A is transcribed nearly 2-fold higher than ACT1 (Holstege et al., 1998; data not shown). One interpretation of this difference in U1:U2 ratio is that the second exon of RP51A is too short to recruit maximal levels of U2 snRNP. This hypothesis predicts that cleavage and polyadenylation would release the RNP from the transcription site, resulting in a significant fraction of post-transcriptional RP51A pre-mRNAs associated with U1 snRNP. We therefore immunoprecipitated U1 snRNP and compared the relative association between ACT1 and RP51A pre-mRNAs by RT-PCR. To ensure that the pre-mRNA was post-transcriptional and had undergone polyadenylation, reverse transcriptase was primed with oligo dT. Data were normalized to endogenous DBP2 pre-mRNA to control for experimental variation and the intronless gene, PMA1, is shown as a negative control. Poly(A)+ RP51A pre-mRNA is ~4 fold more highly associated with U1 snRNP than ACT1 (Figure 1C), suggesting that many RP51A pre-mRNAs are released from Pol II at an early stage of spliceosome assembly. Interestingly, there are also significant levels of post-transcriptional ACT1 pre-mRNA-U1 snRNP complexes despite robust cotranscriptional U2 snRNP recruitment (see Discussion).
The ACT1 and RP51A ChIP and snRNP IP data suggest that exon length may define a limited time window during which nascent spliceosome assembly can occur. However, the altered snRNP patterns could also result from other differences between genes. For example, specific secondary structures within yeast introns (Goguel and Rosbash, 1993; Newman, 1987; Parker and Patterson, 1987) or different promoters (Cramer et al., 1999; Kadener et al., 2001; Kadener et al., 2002) could alter snRNP recruitment in a gene-specific fashion. To minimize gene-specific differences, we created different second exon lengths within a single gene. The constructs are based on HZ18, which expresses a galactose-driven RP51A-LacZ fusion protein and has been assayed for snRNP recruitment by ChIP (Lacadie et al., 2006; Teem and Rosbash, 1983). The well-characterized ADH2 3′ UTR (Hyman et al., 1991) was inserted into LacZ of HZ18 to generate constructs with second exon lengths of approximately 350, 600, 1200, and 2300 bp (lengths include ~100 bp of 3′ UTR; Figure 2A).
The U1 snRNP recruitment results reflect those of ACT1 and RP51A, namely, higher U1 levels on HA-350 and HA-600, whereas HA-1200 and HA-2300 have lower levels (Figure 2B). Conversely, U2 snRNP levels are lower on HA-350 and HA-600 and higher on HA-1200 and HA-2300 (Figure 2C). This indicates that the short second exon genes have a limited capacity to recruit U2 snRNP/tri-snRNP and dislodge U1 cotranscriptionally. Moreover, this suggests that the differences between ACT1 and RP51A snRNP recruitment are principally due to exon length rather than gene-specific features. We interpret the earlier “peak” in U2 snRNP values (Figure 3C, HA-350 and HA-600) to reflect incomplete nascent snRNP recruitment due to premature cleavage and polyadenylation/transcription termination. Indeed, the values at the second primer pair prior to the polyA site of HA-350 and HA-600 are nearly identical for all four constructs.
Based on the differences between U1/U2 levels, we suspected that splicing occurs cotranscriptionally on the two longer constructs but not on the two shorter constructs. To test this prediction, we imported into the four constructs a recently developed assay for cotranscriptional splicing (Abruzzi et al., 2004; Lacadie et al., 2006). In this ChIP-based system, an RNA stem loop that binds to the MS2 (fused to HA epitope) phage coat protein is split by an intron (called split MS2). Upon intron removal, the stem loop forms, binds the MS2 protein and thereby monitors cotranscriptional splicing by ChIP with an anti-HA antibody. Previous results showed significant splicing by ~1 kb past the 3′ ss of the HZ18 construct (Lacadie et al., 2006).
Consistent with the U snRNP recruitment patterns, there is no signal with the shorter two constructs, whereas robust, comparable splicing signals are observed with the longer two constructs (Figure 2E). These results suggest that splicing and therefore complete assembly requires more than 600 bp but less than 1.2 kb of second exon. The split MS2 ChIPs correlate with the U2 recruitment patterns (Figure 2C), suggesting that they are a reasonable surrogate for cotranscriptional splicing. This notion also fits with the observation that U2 and U5 recruitment patterns are generally very similar (Lacadie and Rosbash, 2005) and that the MS2 splicing pattern immediately follows the U2 and U5 patterns (Lacadie et al., 2006).
If cotranscriptional U2/U5 recruitment is important for efficient splicing, the shorter constructs might have an increased pre-mRNA:mRNA ratio. An increase in pre-mRNA might also be due to release of U1 snRNP containing pre-mRNA by polyadenylation, even if post-transcriptional splicing is as efficient as cotranscriptional splicing. We therefore assayed steady-state RNA levels by primer extension and quantitative RT-PCR (Figure 3A).
All four constructs gave rise to indistinguishable pre-mRNA levels (Figure 3A, upper right panel). However, the two shorter constructs do show reproducibly higher mRNA levels than the two longer constructs (Figure 3A, lower right panel), which alters the pre-mRNA percentages. Given the cotranscriptional splicing results (Figure 2), this could indicate that the predominantly post-transcriptional splicing of these two constructs is more efficient than cotranscriptional splicing or that the smaller mRNAs could have longer cytoplasmic half-lives. In either case, the suggestion is that cotranscriptional U2/U5 recruitment and splicing do not increase splicing efficiency. This suggests that, like RP51A, the shorter exon pre-mRNAs are more associated with U1 post-transcriptionally (Figure 1C). The four pre-mRNAs were therefore assayed for post-transcriptional U1 association, and the data were normalized to endogenous DBP2 pre-mRNA to control for experimental variation.
As expected, a bias was found in favor of the shorter second exon constructs (Figure 3B, compare with 1C): HA-350 (3-fold) and HA-600 (2-fold) showed enhanced association of U1 snRNP with poly(A)+ pre-mRNA than HA-1200 or HA-2300 (Figure 3B). As an additional control, GAL10 mRNA (intronless) association with U1 snRNP was assayed in parallel and was approximately 10-fold lower than even the longest HZ18-ADH2 pre-mRNAs. Consistent with a post-transcriptional association of some ACT1 pre-mRNA with U1 snRNP (Figure 1C), there are significant levels of post-transcriptional U1-containing complexes even for HA-1200 and HA-2300. This is despite robust cotranscriptional U2 snRNP recruitment and splicing (see Discussion).
Whole-genome tiling arrays are a powerful method of mapping DNA binding protein sites (Mockler et al., 2005). More recently, this strategy has been used to map RNA binding protein locations (Swinburne et al., 2006). We therefore sought to complement our limited sample of cotranscriptional recruitment patterns and characterize U snRNP recruitment to most and perhaps all intron-containing genes. To this end, we performed ChIPs against U1, U2, and U5 snRNPs and analyzed immunoprecipitated DNAs on high-density oligonucleotide tiling arrays; these were printed with 50-mer probes at 32 bp spacing. This arrangement should also increase the resolution of U snRNP recruitment, at least to intron-containing genes that are transcribed at sufficiently high levels.
A representative view of chromosome 16 contains 4 annotated intron-containing genes and many intronless genes; U snRNP signals are strongly apparent on 3 of the 4 genes (Figure 4A). Examining the entire genome indicates that credible U snRNP signals are highly specific for intron-containing genes and reflect the lower resolution patterns previously observed for ACT1 and RP51A (compare Figures 1A and 1B with Figures 5A and 5E). ~98% of ribosomal protein intron-containing genes were enriched for U snRNPs (99 of 101), whereas only ~25% of non-ribosomal protein intron-containing genes were enriched for U snRNPs (34 of 138). This difference is likely due to the high transcription rates of ribosomal protein genes (Warner, 1999).
To generalize the relationship between U1:U2 ratios and second exon lengths (Figures 1 and and2),2), we calculated the U1:U2 ratio for all 133 intron-containing genes that gave significant signal (Figure 4B). Consistent with the reporter constructs and the ChIP data on ACT1 and RP51A (Figures 1 and and2),2), shorter second genes have a higher U1:U2 ratio. This relationship is strongly exon length dependent and independent of known intron features, such as intron length and consensus splicing sequences (data not shown). We find this relationship quite remarkable, considering the diversity in intron sequence and structure.
Although these ratios provide a rough indication of spliceosome progression, we exploited the high resolution to more closely examine spliceosome assembly patterns. To this end, we compared U1, U2, and U5 snRNP recruitment patterns of representative genes as a function of second exon length (Figure 5A to 5F). There are three main patterns. (1) As exemplified by ACT1 (long 2nd exon), there are separate U1 and U2/U5 peaks with a gradual decrease of U1 followed by a subsequent gradual decrease of U2/U5 (Figure 5A). This is similar to the previously published ChIP patterns (Lacadie and Rosbash, 2005). (2) RPS4B and RPS0B also have separate U1 and U2/U5 peaks; however, U2/U5 decreased more abruptly at the 3′ end of the ORFs (Figure 5B and 5C). (3) RPL16A, RP51A, and RPL30 have less distinct U1 and U2/U5 peaks, and both U1 and U2/U5 decrease simultaneously at the 3′ end of the ORF (rather than first U1 and then U2/U5) (Figure 5D–F). These patterns are consistent with cotranscriptional splicing (5A), robust recruitment of late spliceosomal complexes but transcript release because of cleavage/termination before complete cotranscriptional splicing (5B–C), and incomplete U2/U5 recruitment with release of both U1 and U2/U5 complexes because of cleavage/termination (5D–F). These data are in good agreement with those of Neugebauer and colleagues (Kotovic et al., 2003), who assayed U1 recruitment with ChIP-CHIP and conventional microarrays. Long second exon genes were less enriched, presumably because the microarray probes are biased against the 5′ ends of genes.
The genes with patterns (2) and (3) are the likely principal source of post-transcriptional splicing complexes containing U1 snRNP and CBC, the nuclear cap-binding complex. Indeed, CBC associates more highly with U1 snRNP than with the other four splicing snRNPs: U2, U4, U5, and U6. Moreover, proteomic analysis of tagged Cbp20p complexes indicates that U1 snRNP is the most abundant CBC-associated snRNP. As the CBC-U1 snRNP association is transcription-dependent (Figure S2; see Discussion), it suggests that these complexes are bridged by pre-mRNA and are in vivo commitment complexes.
Our current results indicate that U1 snRNP is robustly recruited to all intron-containing yeast genes. This is because maximal U1 recruitment occurs near the 3′ ss and is independent of second exon length. Moreover, numerous experiments indicate that U1 snRNP is important for subsequent U2 and U5 cotranscriptional recruitment. However, the importance of cotranscriptional versus post-transcriptional U1 recruitment is unknown. To address this issue, we employed a recursive splicing assay first described by the Séraphin group (Lopez and Seraphin, 2001). In the experimental design, a first intron is imbedded within the 5′ splice site of a second intron. The first intron must be accurately removed to unmask the 5′ ss of the second intron, which is subsequently spliced. Lopez and Séraphin (2001) observed no detectable deficit in splicing efficiency (by β-galactosidase activity) when an intron was imbedded within the 5′ ss, suggesting that transcription and splicing are not tightly coupled. Because the LacZ open reading frame is long, we generated four recursive constructs with decreasing second exon lengths; all have the ACT1 intron interrupting the 5′ ss of the RP51A intron (Figure 6A).
Based on our knowledge of splicing kinetics and U snRNP recruitment, we surmised that U1 snRNP recruitment to the smallest masked 5′ss (recHA-350; note that the true second exon size includes the RP51A intron and is ~750 nt) might be substantially post-transcriptional. If there is any positive influence of cotranscriptional U1 recruitment on splicing, this construct might then accumulate pre-mRNA (ACT1 intron removed) relative to pre-pre-mRNA (completely unspliced); the latter should be similar for all four constructs. Indeed, pre-pre-mRNA levels are indistinguishable between the four constructs, by primer extension (Figure 6B) and quantitative PCR (Figure 6C). However, there is a roughly 2.5-fold increase in the pre-mRNA to pre-pre-mRNA ratio for recHA-350 as compared to recHA-2300 (Figure 6C).
Does this pre-mRNA increase correlate with altered spliceosome assembly? Interestingly, both splicing events are evident by the snRNP ChIP assays. All four constructs have a similar U1 signal (both level and location), indicative of the first 5′ss/recruitment event (Figure 6D). However, the second U1 recruitment event is less uniform; the smallest second exon construct (recHA-350) shows no cotranscriptional U1 recruitment to the unmasked 5′ss. We surmise that 750 bp is insufficient distance/time to remove the first intron and recruit U1 snRNP to the unmasked 5′ ss. However, recHA-600 shows maximal U1 recruitment, indicating that 1000 bp is enough distance/time to splice and recruit maximal U1 with this particular construct. Note that cotranscriptional recruitment of U1 snRNP to the masked intron is an assay for cotranscriptional splicing of the first intron.
For U2 snRNP recruitment to the first intron, all four constructs follow the U1 pattern and have similar U2 snRNP recruitment signals and patterns (Figure 6E). For the second splicing event, U2 snRNP recruitment is also similar to the U1 patterns to this intron, namely, high U2 levels on recHA-1200 and recHA-2300, partial U2 recruitment to recHA-600, and no U2 recruitment to recHA-350 (Figure 6E). The data suggest that the decreased splicing efficiency of recHA-350 is due to the lack of efficient cotranscriptional U1 snRNP recruitment to the unmasked intron. In contrast, the reduced cotranscriptional U2 snRNP recruitment to recHA-600 does not result in a large decrease in splicing efficiency. Cotranscriptional U1 snRNP recruitment may commit the intron to efficient post-transcriptional splicing factor recruitment and splicing.
Recent work shows that cotranscriptional spliceosome assembly and splicing occur in yeast (Gornemann et al., 2005; Kotovic et al., 2003; Lacadie and Rosbash, 2005; Lacadie et al., 2006). However, these analyses used a limited set of endogenous genes and reporter constructs; all have relatively long second exons (>1 kb). Here we present a more complete view of cotranscriptional spliceosome assembly by combining ChIP analyses with whole-genome tiling arrays. U1 snRNP recruitment is independent of second exon length and makes a contribution to splicing efficiency. U2 and U5 snRNP recruitment, in contrast, is dependent on exon length. As a consequence, genes with short second exons undergo predominantly post-transcriptional splicing, which occurs efficiently.
How much yeast splicing then takes place cotranscriptionally? The U snRNP and split MS2 ChIPs indicate that splicing begins when Pol II is beyond 600 bp past the 3′ ss and that maximal cotranscriptional splicing requires ~1 kb past the 3′ ss (Figure 2). Assuming an average transcription rate of ~1.5 kb/min and that cotranscriptional splicing is complete on HA-1200, spliceosome assembly and splicing occurs within ~1 minute of transcribing the 3′ ss. This estimate is similar to that of Neugebauer and colleagues, which is based on the ChIP recruitment of Prp19p, a spliceosome activation complex component (Gornemann et al., 2005). One minute is marginally faster than previously observed in metazoans (1) by electron microscopy of early Drosophila embryo genes, in which splicing occurs ~3 min after transcription of the 3′ ss; (Beyer and Osheim, 1988) (2) from β-globin transcripts in a cell culture system (Audibert et al., 2002), or (3) from an analysis of C. tentans (Kiseleva et al., 1994; Wetterberg et al., 2001). Nonetheless, we do not know exactly what fraction of splicing on long second exon genes is cotranscriptional. As we observe some post-transcriptional U1 snRNP association of these pre-mRNAs (Figure 2 and and3B),3B), a fraction of all nascent pre-mRNAs probably fails even to undergo cotranscriptional U1 snRNP recruitment and then engages the entire splicing machinery post-transcriptionally. We suspect, however, that most splicing on a long second exon gene is cotranscriptional. Evidence for this includes comparable ChIP signal from a 5′ UTR-located MS2 stem loop and the split MS2 stem loop (Lacadie et al., 2006; data not shown).
In contrast, most splicing on short second exon genes is apparently post-transcriptional. The abrupt decreases in U2/U5 signal at the end of smaller ORFs (Figure 5) indicate that 3′ UTRs are short and do not substantially impact this conclusion. This is also consistent with recent tiling array results (David et al., 2006). The abrupt decreases also suggest relatively little cotranscriptional splicing of these genes - such as RPS4B (771 bp, Figure 5B) and RPS0B (668 bp, Figure 5C). By incorporating transcription rate data (Figure S1B), we conservatively predict that ~90% of yeast splicing is post-transcriptional (700 bp cut off). If patterns such as RPS4B reflect full assembly without splicing, the estimate rises to ~98% (800 bp cut off)
Several groups have reported associations of splicing factors with RNA Pol II in higher eukaryotes (for reviews, see Bentley, 2005; Kornblihtt et al., 2004). Moreover, the mammalian CTD is required for efficient splicing of all genes tested (Fong and Bentley, 2001; McCracken et al., 1997) and contributes to alternative splicing regulation (de la Mata and Kornblihtt, 2006). Nonetheless, it is uncertain whether yeast RNA Pol II contributes directly to splicing factor recruitment. It may be relevant that the yeast CTD contains roughly half the number of heptad repeats of mammalian CTD and seems to be dispensable for splicing (Licatalosi et al., 2002). There is one report that yeast Prp40p, a U1 snRNP component, associates with the CTD (Morris and Greenleaf, 2000; Phatnani et al., 2004). However, another group failed to confirm this observation through a different experimental approach (Wiesner et al., 2002). A recent mammalian study identified an important association between mammalian 3′ end formation factors and U2 snRNP (Kyburz et al., 2006). Perhaps a similar relationship in yeast promotes faster assembly than would occur otherwise. However, all data to date are consistent with simple kinetic competition between yeast splicing and 3′ end formation; if there is not enough time/distance, cotranscriptional assembly and splicing will not occur. The is based in part on the lack of robust U2 snRNP recruitment by genes with short second exons, which suggests a lack of robust coupling between Pol II and U2/tri-snRNP recruitment. The recruitment data include the tiling arrays (Figure 4), which show a coherent spliceosome assembly picture independent of intron features (e.g., size and structure). Finally, the recursive splicing experiments of Lopez and Séraphin (2001) also suggest that transcription and splicing are not tightly coupled.
Use of this recursive splicing strategy does provide insight into a potential contribution of cotranscriptional U1 snRNP recruitment to splicing efficiency (Figure 6). In the recursive splicing of the recHA-350 construct, substantial U1 recruitment appears to be pushed post-transcriptionally; this may contribute to less efficient splicing. This is in contrast to the longest construct, recHA-2300, in which assembly and splicing of the second intron occurs cotranscriptionally and more efficiently. However, the (second) pre-mRNA substrate is considerably different than a canonical nascent pre-mRNA. For example, the 5′ ss “appears” much later than usual; in fact, the branch point is probably available before the 5′ ss. As we previously showed that the branch point alone can recruit U1 snRNP (Lacadie and Rosbash, 2005; Lacadie et al., 2006), it is possible that the second branch point (of RP51A) helps recruit U1 to the second intron. This consideration suggests that the two-fold decrease in splicing efficiency may be an underestimate of the post-transcriptional versus cotranscriptional splicing efficiency ratio. In any case, we suggest that the recursive 5′ ss of recHA-350 experiences less than optimal splicing, possibly because of enhanced competition from RNA binding proteins away from the site of transcription. Relevant to this idea are recent experiments showing that nonspecific RNP complexes reduce spliceosome formation efficiency on pre-mRNAs not coupled to transcription (Das et al., 2006).
Given the paucity of known direct transcription-splicing contacts in yeast, cotranscriptional splicing events may benefit from more indirect interactions. For example, recent experiments suggest that TFIIS promotes cotranscriptional splicing without directly recruiting splicing factors (Lacadie et al., 2006). Another possibility is a higher local U1 snRNP concentration near active chromatin. As recent in vitro coupled transcription-splicing experiments suggest competition between splicing and degradation by nucleases (Hicks et al., 2006), coupling between U1 recruitment and transcription may help avoid degradation pathways. Interestingly, transcription-dependent CBC-U1 complexes are highly abundant and stable in vivo (Figure S2). Pre-mRNAs bridging these in vivo commitment complexes are most likely derived from short second exon genes (Figure 1C and and3B3B).
Pre-mRNA microarray experiments show that introns are not all affected similarly by the same mutant, e.g., ts splicing factors (Burckin et al., 2005; Clark et al., 2002). Whole genome-tiling ChIP-CHIPs from various splicing mutant strains should be able to complement these observations, particularly from mutants that probably affect early assembly steps. For example, some differences between intron-containing genes may result from splicing factors focused on post-transcriptional events. In higher eukaryotes as well, tiling arrays would significantly increase our understanding of alternative splicing as well as the binding and regulatory sites of various splicing factors. Indeed, recent ChIP experiments with mammalian splicing factors (Listerman et al., 2006) reinforce this optimism and suggest that the same tiling approach taken here with yeast should be applicable to metazoans.
Yeast strains were manipulated according to standard procedures (Guthrie and Fink, 1991). Cultures were maintained in 2% galactose 0.05% glucose for all experiments with HZ18 derivatives. For ChIP-CHIP experiments, cells were grown in 2% glucose. TAP-tagged strains were obtained from Open Biosystems.
Construction of HZ18 derivatives with varying second exon lengths was achieved by either Quick change PCR or standard cloning. For Quick change, oligos were designed to amplify ADH2 3′ UTR and poly(A) site as defined previously (Hyman et al., 1991). 2.5 μg of this PCR product (~300 base pairs of ADH2 with 40 bases upstream and downstream for HZ18 targeting) was then used as a primer for quick change according to manufacturer’s protocol (Stratagene). For standard cloning, the same ADH2 sequence was PCR amplified with oligos containing either ClaI or SacI sites. Digested PCR products were then cloned into ClaI or SacI sites in HZ18 using standard methods. Split MS2 constructs were generated by cloning a BamHI/AclI digested PCR product amplified from the split MS2 vector into BamHI/AclI sites in the HZ18-ADH2 derivatives. Recursive constructs were generated by quick change PCR. The ACT1 intron was PCR amplified with oligos containing RP51A sequence splitting the 5′ ss. The PCR product was then used as a primer for quick change and reaction were performed as described above. Plasmids and strains can be found in Supplemental Table 1. Oligos used for Quick change and cloning are located in Supplemental Table 2.
Chromatin Immunoprecipitations and quantitative PCR were performed as previously described (Komarnitsky et al., 2000; Lacadie and Rosbash, 2005). Real time PCRs were performed on a Rotor-Gene 3000 PCR machine (Corbett Research). ACT1 ChIP signals are expressed as the IP to input ratio for ACT1 over the IP to input ratio for the intronless PMA1 gene. Reported ACT1/PMA1 values are the averages of three independent experiments and error bars represent average deviation. For ChIPs to HZ18-ADH2 constructs, U1 snRNP IPs are normalized to the last primer pair and U2 snRNP IPs are normalized to the first primer pair. Split MS2 ChIPs are normalized to the first primer pair. All data represent the average of at least two independent experiments and error bars report average deviation. Sequences of oligos used in this study can be found in Supplemental Table 2.
U1 snRNP IPs for detecting post-transcriptional pre-mRNA association were performed as described (Tardiff and Rosbash, 2006). Input and IP RNAs were phenol:chloroform extracted, ethanol precipitated, and cDNA made by oligo dT-primed reverse transcription according to manufacturer’s protocol (Invitrogen). For U1 snRNP IPs of RP51A, ACT1, and the HZ18-ADH2 constructs, pre-mRNA IP:IN ratios were normalized to DBP2 pre-mRNA IP:IN ratios as a positive control and then set to 1.0 for an intronless mRNA negative control. All data represent the average of at least two independent experiments and error bars report average deviation.
Tiling ChIP-CHIP was performed by NimbleGen Systems, Inc as part of a ChIP Array Service (http://www.nimblegen.com). ChIP was performed as described above. The input (~20 ng) and the IP were amplified by LM-PCR as previously described (Ren et al., 2000). Input (4 μg) and IP (4 μg) of LM-PCR product ranging from 200 to 600 nucleotides was provided. Input (Cy3-labeled) and IP (Cy5-labeled) were hybridized on S. cerevisiae whole-genome tiling arrays with 50-mer probes at 32 bp spacing. For each hybridization, NimbleGen Systems returned raw data signal intensities for specific IP and input DNA, plus a log2 ratio of the IP to input signals. Signals for intron-containing genes were considered if U1, U2, and U5 signals were at least 2-fold over background and showed separable U1 from U2/U5 peaks.
RNA was purified from yeast strains harboring recursive constructs by hot acid phenol extraction (Ausubel et al., 1994). Total RNA (2 μg) was treated with DNase and extended with RB1 (Pikielny and Rosbash, 1986) using Superscript III reverse transcriptase (Invitrogen). Quantitative PCR was then performed with oligos specific to each pre-mRNA. Reported values are the average of three independent experiments and error bars represent average deviation.
We thank our local colleagues, especially K. Abruzzi and S. Kadener, as well as K Neugebauer for suggestions and comments on the manuscript. Thanks also to B. Séraphin, B. Graveley, and A. Berglund for helpful discussions. This work was supported in part by a grant to M.R. from the National Institutes of Health (NIH) (R01 GM23549) and the NIH Genetics Training Grant (5T32 GM07122).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.