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 (). 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 ( and ), 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 () 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, ) and RPS0B
(668 bp, ). 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 (), 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 (). 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 ( and ).
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.