Splicing-coupled 3′ end formation requires a terminal splice acceptor site, but not intron excision

Splicing of human pre-mRNA is reciprocally coupled to 3′ end formation by terminal exon definition, which occurs co-transcriptionally. It is required for the final maturation of most human pre-mRNAs and is therefore important to understand. We have used several strategies to block splicing at specific stages in vivo and studied their effect on 3′ end formation. We demonstrate that a terminal splice acceptor site is essential to establish coupling with the poly(A) signal in a chromosomally integrated β-globin gene. This is in part to alleviate the suppression of 3′ end formation by U1 small nuclear RNA, which is known to bind pre-mRNA at the earliest stage of spliceosome assembly. Interestingly, blocks to splicing that are subsequent to terminal splice acceptor site function, but before catalysis, have little observable effect on 3′ end formation. These data suggest that early stages of spliceosome assembly are sufficient to functionally couple splicing and 3′ end formation, but that on-going intron removal is less critical.


INTRODUCTION
Most human pre-mRNAs contain multiple introns that are removed by splicing. The splicing process has two catalytic steps with the first releasing the 5 0 exon and the second joining the two exons together with consequent release of an intron lariat (1). It is performed by a set of five small nuclear RNAs (snRNAs) in concert with 100 s of protein factors. Intronic sequences, such as the 5 0 and 3 0 splice site, intervening branch-point and polypyrimidine tract, play a vital role (2). The function of snRNAs in splicing has been well characterized through decades of study using both in vitro and in vivo systems. First, U1 snRNA (U1) contacts the 5 0 splice site after which U2 snRNA (U2) is recruited to the branch-point following recognition of the 3 0 splice site and polypyrimidine tract by U2AF35 and 65, respectively. U4, U5 and U6 snRNAs then join, and two-step catalysis occurs after U1 and U4 are ejected from the spliceosome (2). Splicing of most introns in higher eukaryotes is thought to occur by definition of the comparatively shorter exons, across which 3 0 and 5 0 splice sites are recognized by U2 and U1, respectively (3).
The majority of pre-mRNAs undergo cleavage and polyadenylation at their 3 0 ends. This process requires a poly(A) (pA) signal within the RNA, which is composed of a hexameric sequence, usually AAUAAA, followed by a U or G/U rich sequence (4,5). Cleavage of the transcript occurs between these two elements and is carried out by the CPSF73 endonuclease, which is part of the Cleavage and Polyadenylation Specificity Factor (CPSF) complex (6). Other protein complexes including Cleavage Stimulation Factor (CstF), Cleavage Factors I and II (CFIm and CFIIm) are also required (7). The upstream cleavage product is polyadenylated to form the mRNA, whereas the downstream product is rapidly degraded in a process linked to termination of RNA polymerase II (Pol II) (8,9).
First observed 30 years ago (10), it is now established that there is reciprocal functional coupling between splicing and 3 0 end processing of RNA (11,12). This is through definition of the last exon, which requires a terminal splice acceptor (tSA) site and a pA signal. Mutation of either sequence disrupts both splicing and 3 0 end formation as well as the linked process of transcriptional termination (12). In addition, trans-acting factors play a role by forming bridging contacts across the terminal exon. Prominent among these are U2AF65, which binds cleavage factor Im and poly(A) polymerase (13,14), and SF3b components, which interact with the CPSF complex (15). A large-scale purification of the 3 0 end processing complex identified other splicing factors that may function analogously (16).
3 0 end formation is also negatively regulated by certain splicing factors. Early experiments showed that 5 0 splice site sequences are involved in silencing some viral pA signals through the recruitment of U1 snRNA (17)(18)(19)(20). More recently, U1 was found to suppress the processing of a huge number of pA signals embedded within human pre-mRNAs (21,22). There must therefore be a mechanism of co-ordinating this suppression with activation by later splicing factors to establish coupling with 3 0 end formation.
The complex regulation of 3 0 end formation by splicing should be considered in the context of transcription, to which both reactions are coupled in vivo (23). This is achieved by ternary interactions between processing factors, the pre-mRNA transcript and the C-terminal domain of the large subunit of Pol II. Consequently, cotranscriptional splicing of introns is common (24)(25)(26)(27). However, introns near to the 3 0 end of a transcript frequently display a lower incidence of co-transcriptional removal (25,28,29).
We have used several different approaches to determine how and when splicing is coupled to 3 0 end formation. We mutated the tSA site in a chromosomally integrated bglobin gene. This inhibited splicing and 3 0 end formation in a manner involving U1 snRNA and was associated with the loss of U2AF recruitment to the tSA site. To further analyse the requirements for tSA site function, we blocked splicing at subsequent stages using the small molecule Spliceostatin A (SSA) and by inhibiting U4 snRNA. These approaches reduced splicing but surprisingly had little effect on 3 0 end formation. Thus, a tSA site establishes coupling between splicing and 3 0 end formation, which is sufficient for cleavage and polyadenylation even when intron removal is inefficient.

Plasmids
The b-globin plasmid for Figure 5 (bÁ5-7) has been described previously (30) as has the Tat plasmid (31). The VA plasmid contains the adenovirus VAI gene and was a gift from Nick Proudfoot. For the bWT cell line, b-globin and a portion of 3 0 flank was amplified from bÁ5-10 (30) using primers bE1F and bFR. This was inserted into the pcDNA5/FRT/TO vector that was initially linearized by PCR with primers pcF and pcR. This new vector (bWT) was used as a template for site-directed mutagenesis with primers tSAmF and tSAmR for the btSA cell line and with bpAmF and pAmR for the pAmF line. bWTCoTC was made in the same way as bWT, but the insert was generated by PCR amplification of the previously described bÁ5-7 (30) plasmid with bE1F and bFR.

Cell culture
Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% foetal calf serum. SSA was used at 50 ng/ml unless otherwise stated. Actinomycin D (ActD) was used at 10 mg/ml. Electroporation was performed on a confluent 10-cm diameter dish of cells using 10-15 mM anti-sense morpholino (AMO) in 400 ml of Dulbecco's modified Eagle's medium using a 4-mm gap cuvette (960 mF, 280 v in a Biorad gene pulser). RNA was isolated 3 h after electroporation. For transient transfection, 2.5 mg of reporter and 0.5 mg VA were introduced using Xtreme gene (Roche). For anti-sense oligonucleotide (ASO) experiments, 500 pmol of oligo was introduced into a 24-well dish of cells using Lipofectamine RNAiMax (Life Technologies). For all plasmid and ASO transfections, RNA was isolated after 24 h. For stable integration, a 6-cm dish of Flp-IN HEK cells was transfected with 1 mg of the appropriate b-globin plasmid in addition to 3 mg of pOG44 using JetPrime (Polyplus) transfection reagent. Forty-eight h post-transfection, cells were transferred to a tissue culture flask and grown in the presence of Hygromycin (100 mg/ml) and Blasticidin (10 mg/ml) until resistant cells emerged. Transcription was induced by overnight growth in media supplemented with 1 mg/ml tetracycline (tet).

RNA isolation
Chromatin-associated and nucleoplasmic RNA was isolated as previously described (32,33). RNA was treated twice with Turbo DNase (Ambion). Fractionation efficiency was determined by detection of endogenous U1. Total RNA was isolated using Trizol (Life Technologies). For real-time PCR analysis, 1 mg of RNA was reverse transcribed using Inprom II (Promega). Parallel reactions were performed in the absence of reverse transcriptase. In all, 1/20th of the cDNA mix was used for real-time PCR using 5-10 pmol of forward and reverse primer and Brilliant III SYBR mix (Agilent Technologies) in a Qiagen Rotorgene machine. Differences were calculated using comparative quantitation.
Chromatin and RNA Immunoprecipitation and S1 nuclease analysis Detailed protocols are provided in the supplementary material.

Primer sequences
Primer, AMO and ASO sequences are provided in the supplementary material.

RESULTS
A tSA site is required for maximal 3 0 end formation in a chromosomal context To analyse coupling between splicing and 3 0 end formation in a chromosomal context, we constructed two HEK cell lines containing either a single stably integrated wildtype (bWT) b-globin gene or one with a mutated tSA site (btSA) ( Figure 1A). Transcription was under the control of a tet-inducible CMV promoter. To verify the two cell lines, total RNA was isolated from bWT or btSA cells after overnight growth in media supplemented or not with tet. Random hexamer generated cDNA was analysed by PCR to detect unspliced (Ex2-In2) and spliced (Ex2-Ex3) b-globin transcripts ( Figure 1B). Unspliced RNA was detected in both the bWT and btSA lines, and levels were increased following tet induction. There was a higher level of unspliced intron 2 in btSA samples in line with the expected splice inhibitory effect of the mutation. Correctly spliced b-globin was detected only in samples from bWT cells, again displaying a tet-dependent increase. Endogenous b-globin is not expressed in HEK cells (34).
To test whether the btSA mutation affected 3 0 end formation in the chromosomal context as it does in plasmids (12), we performed 3 0 RACE analysis on total RNA from tet-induced bWT and btSA cells ( Figure 1C). A tagged oligo-dT was used for reverse transcription, and subsequent PCR was done with primers directed to the tag and to exon 3 to detect all polyadenylated species. A strong band was detected in the bWT sample, but a much weaker band was detected in btSA cells indicating little processing in that case (8-fold less than bWT). We also analysed random hexamer generated cDNA using primers spanning the pA site to detect b-globin transcripts that were not 3 0 end processed ( Figure 1D). The tSA site mutation resulted in a 3-4-fold accumulation of this species compared with bWT samples, which is again consistent with reduced 3 0 end formation. This was also associated with a mild reduction in the recruitment of CstF77 pA-processing factor to the btSA gene (Supplementary Figure S1).
As pre-mRNA processing defects often reduce transcription (35)(36)(37), we tested whether this was the case on btSA. There was little difference in the abundance of unspliced intron 1 in bWT and btSA cell lines ( Figure 1E), which we used as a proxy for transcriptional activity. We also analysed Pol II loading on the bWTand btSA-mutated b-globin genes using chromatin immunoprecipitation (ChIP) ( Figure 1F). This was assessed at three positions across the gene using an antibody that detects all forms of Pol II (N20). No significant differences were observed over any of the regions between the two cell lines, indicating that similar levels of Pol II were present. Collectively, these data demonstrate that the tSA site mutation inhibits splicing and 3 0 end processing, which is not mediated through obvious reductions in transcription in this context.
Mutation of a tSA site is also predicted to prevent the recruitment of relevant 3 0 splice site-binding factors that participate in terminal exon definition. We tested this by performing RNA immunoprecipitation (RIP) to analyse binding of the 65 kDa component of U2AF, U2AF65, to the tSA site of bWT or btSA transcripts. Immunoprecipitated RNA was reverse transcribed with random hexamers and then PCR amplified with primers across the tSA site. Histone H1E mRNA was used as a negative control, as U2AF65 is not expected to bind this transcript. A 4-5-fold greater proportion of bWT tSA sitecontaining transcripts were recovered in comparison ( Figure 1G). In contrast, equivalent btSA transcripts were close to this background level, suggesting that U2AF is poorly recruited in the absence of a tSA site. This is in line with previous observations that the splice site AG dinucleotide is important for its efficient association with RNA (38,39). Its absence from btSA transcripts correlates with their poor processing and with data from several laboratories showing U2AF65 to promote cleavage and polyadenylation (13,40,41).
3 0 end processing of btSA pre-mRNA is activated by inhibition of U1 snRNA In addition to blocking the recruitment of factors that may act positively in 3 0 end formation, a tSA site mutation may also leave the pA signal vulnerable to negative effects. In particular, U1 inhibits cleavage and polyadenylation at pA sites when bound to nearby 5 0 splice sites (17,21,22,42), which are all intact in btSA transcripts. To test whether this was the case, we inhibited the ability of U1 to interact with pre-mRNA using an ASO that binds to its 5 0 end (43,44). Treatment with the ASO is therefore predicted to block the earliest steps in spliceosome assembly. Transfection of the U1 ASO into bWT and btSA HEK cells induced premature 3 0 end formation within the NR3C1 pre-mRNA, which is a known consequence of U1 inhibition and confirms the effectiveness of the approach (22) (Figure 2A).
To determine the effect of U1 inhibition on b-globin splicing, reverse transcription and real-time PCR were used to quantitate the level of unspliced intron 2 (Ex2-In2), intron 2 lariat, spliced (Ex2-Ex3) and total (Ex3) bglobin transcripts ( Figure 2B and C). For bWT cells, U1 ASO caused a significant increase in unspliced intron 2 and a decrease of lariat intron 2 consistent with splicing inhibition. Little change in total or spliced bWT RNA was seen following U1 ASO treatment potentially because these species are stable or because some splicing still occurs. For btSA cells, U1 ASO treatment caused an increase in amounts of unspliced intron 2 and exon 3, but spliced transcripts remained at low levels as expected.
To test for the effect of depleting U1 binding on 3 0 end formation, we quantitatively analysed unspliced pre-mRNA that was 3 0 end processed ( Figure 2D). To assay RNAs for which splicing was inhibited, RNA from control or U1 ASO-treated cells was reverse transcribed with a tagged oligo-dT probe and then amplified using the tag primer and a primer within intron 2. Only authentic 3 0 end-processed RNA is detected by this approach, as it is absent when the pA signal is mutated (Supplementary Figure S2). U1 inhibition caused a 2-fold increase of this species in bWT cells, suggesting that 3 0 end formation occurs when U1 is not recruited to the natural intron. Interestingly, a similar effect on 3 0 end processing was observed in btSA cells. Furthermore, an AMO targeting U1 gave the same result in btSA plasmid transfected HeLa cells treated for only 3 hours ( Figure 2E). This enhancement of 3 0 end formation of btSA transcripts strongly suggests that bound U1 normally plays a role in suppressing cleavage and polyadenylation in this situation. This is consistent with the confirmed function of U1 bound to 5 0 splice site sequences, which are all intact in btSA, in inhibiting cleavage and polyadenylation (21,22,42).
3 0 end formation occurs in the presence of SSA in a tSA site-dependent manner The data so far show that 3 0 end formation requires a tSA site and that, in part, this is to alleviate the negative effects of U1. We were next interested in testing whether a block to splicing that occurs after tSA site recognition would impair 3 0 end formation. SSA was used, as it blocks splicing after U1 and U2 have bound the RNA but before catalysis (45)(46)(47). Unlike the tSA site mutation, SSA does not affect the recruitment of U2AF65 (47,48), suggesting that recognition of the 3 0 splice site still occurs. As SSA blocks splicing before catalysis, this experiment also tests whether intron excision plays a role in coupling splicing and 3 0 end formation.
We first determined the effect of SSA on b-globin splicing using both the bWT and btSA HEK cell lines. Total RNA was isolated from cells treated with SSA or its methanol (MeOH) solvent. Treatment was limited to 3 h to limit potential off-target effects of splicing inhibition. Real-time PCR was then used to quantitate the level of unspliced intron 2 (Ex2-In2), intron 2 lariat, spliced (Ex2-Ex3) and total (Ex3) b-globin transcripts ( Figure 3A and B). In bWT cells, SSA increased the level of intron 2 and reduced the amount of lariat intermediate, demonstrating a strong inhibition of splicing before catalysis. By this measure, inhibition was more efficient than with the U1 ASO. Intron 2 levels were also increased in btSA cells treated with SSA owing to inhibition of a less frequent cryptic splicing event (Supplementary Figure S3). Little SSA-dependent change in total or spliced bWT RNA was seen potentially because stable processed transcripts are present from before this short splicing block. As expected, total and correctly spliced btSA transcripts were present at low levels or absent, respectively. To analyse 3 0 end formation, we detected b-globin transcripts that were not cleaved at the pA site ( Figure 3C). We also analysed 3 0 flanking RNA, as transcriptional read-through is a feature of improper pA site function. SSA treatment resulted in a modest increase in uncleaved bWT pA sites (1.6-fold) and 3 0 flanking (2-fold) Ex2-Ex3 1.9 kb transcripts indicating a potential effect on 3 0 end formation. However, an SSA-dependent effect of similar magnitude was observed in btSA samples. Given that pA cleavage is substantially less efficient for btSA ( Figure 1C), it is unlikely that a 3 0 end processing defect is responsible for this.
To further determine the effect of SSA on b-globin 3 0 end formation, we used another cell line (bWTCoTC). This is the same as bWT but has a sequence (the CoTC element) in the 3 0 flank, which promotes transcriptional termination in both plasmid and chromosomal environments (30,34). An efficient termination process would reduce the synthesis of any read-through RNA, which might be somehow stabilized following SSA treatment. More importantly, however, if the effect of SSA is on processing at the b-globin pA site, this would still be seen, as this sequence is identical to bWT. Figure 3D shows that SSA inhibits terminal intron splicing on bWTCoTC RNA as for bWT, as intron 2 and lariat signals are, respectively, increased and decreased by similar amounts. However, in bWTCoTC cells, SSA did not affect the level of non-pA cleaved or flank RNA. This result does not support substantial inhibition of b-globin 3 0 end formation by SSA. In line with this conclusion, an independent study showed a modest reduction in b-globin read-though, following 24 h of SSA treatment, using a similar stable integrate (48). If 3 0 end formation does continue in the presence of SSA, cleaved and polyadenylated transcripts that are also unspliced will accumulate. To test this and any function for a tSA site, we used our 3 0 RACE approach to detect these transcripts ( Figure 3E). SSA induced a 2-fold increase in unspliced 3 0 end processed RNA recovered from bWT and bWTCoTC cells. Importantly, however, there was a lack of SSA effect on 3 0 end formation of btSA transcripts, which is in contrast to what is seen by U1 inhibition. Thus, a tSA site is required for 3 0 end formation when splicing is impaired by SSA, but not when prior steps in spliceosome assembly are blocked by inhibiting U1. This is consistent with the idea that recruitment of U1 inhibits cleavage and polyadenylation until a functional tSA site is encountered. However, although a tSA site is required for 3 0 end formation in the presence of SSA, subsequent intron removal is less critical, as unspliced 3 0 end processed bWT and bWTCoTC RNA accumulates.

U1 snRNA inhibition and SSA do not affect 3 0 end formation of P27 and Myc transcripts
We next wanted to analyse the effects of U1 inhibition and SSA on transcripts synthesised from genes in their natural chromosomal context. P27 and Myc transcripts were chosen, as they have previously been used to study effects of SSA (48,49). Moreover, they have no obvious sites for premature cleavage and polyadenylation, which is a widespread consequence of U1 inhibition (22). We first tested the effects of U1 inhibition by electroporating HeLa cells with a control AMO or one that binds the pre-mRNA interacting region of U1 (22). An AMO was used, as it can be electroporated into HeLa cells facilitating a more rapid inhibition of splicing than ASO transfection (we were unable to introduce AMOs into our HEK cells). Accordingly, total RNA was isolated after 3 h and reverse transcribed. We then used primer sets to detect unspliced, spliced and non-pA cleaved P27 or Myc RNA ( Figure 4A). Consistent with splicing inhibition, we observed a substantial increase in unspliced RNA for both transcripts, with a more modest decrease in spliced RNA, likely owing to its stability. There was no change in the level of non-pA cleaved RNA indicating that 3 0 end processing remains functional in the absence of U1 and early spliceosome assembly. Finally, U2AF65 RIP analysis revealed that it associates with P27 and Myc transcripts with similar efficiency in control and U1 AMOtreated cells ( Figure 4B). Thus, although U1 AMO inhibits U1 binding and pre-mRNA splicing, the intact 3 0 splice site can still recruit some U2AF. This might be due to terminal exon definition requiring an intact 3 0 splice site and pA signal (11,50), which are present in this situation. In sum, these data show that 3 0 end processing can proceed when splicing is impaired by preventing U1 from interacting with pre-mRNA. This is similar to our b-globin observations and reports that many pA signals are activated when U1 is inhibited in this way (21,22). We next analysed the effects of blocking later steps in spliceosome assembly using SSA. Total RNA was isolated from cells treated with MeOH or SSA. Following reverse transcription, cDNA was analysed with primers to detect unspliced, non-pA cleaved and spliced RNA from Myc and P27 genes ( Figure 4C). In both cases, SSA treatment caused a large accumulation of unspliced transcripts when compared with the MeOH control. SSA also caused a strong reduction in the level of spliced Myc and P27 RNA consistent with a substantial inhibition of splicing. Despite this, there was little effect of SSA on the level of non-pA cleaved RNA arguing that, although splicing was inhibited, 3 0 end formation was still occurring. Consistently, an anchored oligo-dT RT-PCR revealed little effect of SSA on the total level of 3 0 end processed P27 and Myc RNA, despite the spliced versions of each being depleted ( Figure 4D). These effects were also observed when lower concentrations of SSA were used and when shorter time points were used (Supplementary Figure S4). This is again consistent with our b-globin analyses.

A U4 AMO is a rapid and potent splicing inhibitor
We found it surprising that SSA-induced splicing inhibition did not observably impact on 3 0 end formation. To be more confident about this finding, we sought an alternative means of inhibiting a similar step of splicing and used an AMO directed to U4 ( Figure 5A). The region of U4 targeted by our AMO is required to form a stem with U6 that is essential for active splicing (51). Pre-mRNA injected into Xenopus Oocytes that contain U4 lacking this region can only associate with U1 and U2 (52). Consistently, treatment with our U4 AMO causes a substantial reduction in the level of chromatin-associated U4, U5 and U6 but has less effect on the recovery of U1 and U2 (Supplementary Figure S5).
We validated the effectiveness of the U4 AMO in vivo by electroporating it into HeLa cells that were transiently transfected with a b-globin expression plasmid ( Figure 5B). Total RNA was isolated after 3 h to avoid off target effects, and cDNA was generated with random hexamers. PCR was performed with primers in b-globin exons 1 and 2 or in exons 2 and 3. In control AMO samples, these primers detected mostly spliced RNA with a fainter band corresponding to unspliced transcripts. In the U4 AMO sample, we observed a reduction in spliced RNA and a large increase in unspliced precursor. Splicing is inhibited before catalysis because accumulation of lariat from intron 2 was strongly reduced by the U4 AMO (Supplementary Figure S6A).
We next tested whether the reduced b-globin splicing caused corresponding effects on 3 0 end formation. We performed S1 nuclease analysis on the same RNA samples as Figure 5B (Figure 5C). Transcripts were probed with endlabelled DNA specific to the b-globin 3 0 end or the transfection control VA RNA. Levels of VA RNA are similar in both samples indicating equivalent transfection. Strikingly, there was also little difference in the level of RNA cleaved at the pA site. Thus, although splicing is severely compromised by U4 inhibition, 3 0 end processing still occurs. As with SSA, unspliced 3 0 end processed RNA was produced following U4 inhibition in a manner requiring a tSA site (Supplementary Figure S6B and C). Analysis of prematurely processed NR3C1 revealed that the ability of U1 to bind pre-mRNA and inhibit premature 3 0 end formation is largely unaffected by U4 AMO treatment (Supplementary Figure S7).
It is well established that there is an intimate relationship between pA signal function and transcriptional termination. Therefore, as a final measure of the effect of U4 inhibition on 3 0 end formation, we checked its effect on transcriptional termination ( Figure 5D). RNA was isolated from transfected control or U4 AMO-treated cells, and qRT-PCR was performed to detect non-pA cleaved transcripts and read-through RNA from three different positions 3 0 of the pA signal. Signals were normalized following adjustment to co-transfected VA RNA levels. There was no increase in read-through RNA at any of the positions tested, suggesting that termination continues to occur on the b-globin plasmid following U4 AMO treatment. Termination is therefore independent of intron removal when splicing is blocked through inhibition of U4.

U4 AMO inhibits P27 and Myc splicing, but not 3 0 end formation
We next analysed the effect of U4 inhibition on endogenous P27 and Myc transcripts by performing qRT-PCR to detect unspliced, spliced and non-pA cleaved sampled in control or U4 AMO-treated cells ( Figure 6A). We observed a large increase in unspliced RNA levels with a reduction of spliced RNA on U4 AMO treatment demonstrating effective splicing inhibition. Consistent with a lack of effect on 3 0 end formation, U4 inhibition did not increase the amount of non-pA cleaved RNA that  Values are from SSA samples and expressed as fold change relative to those obtained in MeOHtreated cells following normalization to U1 levels. Error bars show SD from at least three biological repeats. Asterisk denotes P < 0.05 for differences following splicing inhibition.
was recovered. As with SSA, this result was maintained at lower concentrations of AMO, which makes it unlikely that saturation of a key factor is responsible for the observation (Supplementary Figure S8). A lack of effect on processing at the P27 and Myc pA sites was also confirmed using an anchored oligo-dT reverse transcription and real-time PCR approach (Supplementary Figure S9). We also performed U2AF65 RIP to assess its binding to the tSA site of Myc and P27 transcripts in control and U4 AMO samples ( Figure 6B). A similar proportion of tSA containing transcripts were bound by U2AF65 in control and U4 AMO samples. However, unlike with U1 AMO, the association of U1 snRNA with pre-mRNA remains efficient under these conditions (Supplementary Figure S7). This is consistent with the prediction, from previous data and from the classical splicing pathway (52), that inhibition of U4 in this way blocks splicing following the recruitment of U1 and U2 to the 5 0 and 3 0 splice sites.

U4 AMO does not affect recruitment of Pol II and CstF77 to the Myc gene
Having analysed the pre-mRNA levels and 3 0 splice site occupancy by U2AF65, we next tested the effects of U4 inhibition on Pol II density across the Myc gene. This is important as changes in the level of various pre-mRNA species may also result from differences in the abundance of transcribing Pol II. Indeed, splicing perturbation is sometimes associated with changes in Pol II loading on genes (35,36,53). We performed Pol II ChIP (using the N20 antibody) in control and U4 AMO-treated cells using primer pairs along the Myc gene. Primer sets were also located within the 3 0 flanking region to analyse transcriptional termination ( Figure 6C).  the Myc gene. For many such factors, this is readily observed by ChIP. For instance, members of the CstF complex are found recruited at and beyond the pA signal consistent with their biological function in 3 0 end formation (54). We therefore assayed the recruitment of CstF77 in both control and U4 AMO-treated samples using ChIP ( Figure 6D). In control cells, CstF77 signal was strongest at positions downstream of the pA site, which is consistent with previous observations (54). Importantly, a similar profile was observed in U4 AMOtreated samples indicating that, even when splicing is impaired in this manner, the 3 0 end processing complex is present to potentially execute rapid 3 0 end formation.
U4 AMO causes release of pre-mRNA into the nucleoplasm and enhances its stability Mutations that prevent pre-mRNA splicing cause aberrant transcripts to be retained at their site of synthesis (35,36,53,55,56). However, it is notable that cis-mutations block splicing at earlier points to where the U4 AMO is predicted to act. With this in mind, we assessed the intranuclear distribution of unspliced pre-mRNA following control or U4 AMO treatment. A well-established technique was used to isolate chromatin-associated and nucleoplasmic RNA (33). We then used qRT-PCR to detect unspliced and non-pA cleaved P27 and Myc transcripts ( Figure 7A). U4 AMO treatment caused an increase in the level of chromatin-associated unspliced P27 and Myc transcripts consistent with inhibition of co-transcriptional splicing. Strikingly, U4 AMO treatment also caused a strong accumulation of unspliced transcripts in the nucleoplasm, indicating that they are released from chromatin. In contrast, non-pA cleaved RNA was largely confined to the chromatin fraction in control and U4 AMO samples consistent with co-transcriptional 3 0 end formation in both conditions. This experiment suggests that, in contrast to RNA containing cis-mutations that inactivate early steps in splicing, the U4 AMO results in the accumulation of many unspliced transcripts in the nucleoplasm. This is in line with the independent observation that SSA induces the accumulation of polyadenylated RNA in nucleoplasmic foci (45). Aberrant pre-mRNAs are often rapidly degraded as a consequence of their retention at transcription sites (35). We therefore wished to test the hypothesis that the 3 0 end processing and release of pre-mRNAs into the nucleoplasm might stabilize them instead. To do so, we treated cells with control and U4 AMOs before supplementing media with actinomycin D (ActD) to inhibit transcription. The levels of unspliced, spliced, total and non-pA cleaved transcripts were then measured at various time points thereafter using RT-PCR on total RNA. Figure 7B shows the results obtained for the Myc transcripts in this experiment. In control AMO samples, total and spliced transcripts are relatively stable up until 40 min, and then they decay. In contrast, both unspliced and non-pA cleaved transcripts begin to disappear after only 10 min, as a consequence of their rapid processing or degradation. In U4 AMO-treated samples, unspliced, spliced and total RNAs were highly stable. However, the non-pA cleaved transcripts behaved identically to those in control samples. These data provide further support for our model that U4 inhibition does not impair pA site cleavage.

DISCUSSION
Our findings are consistent with the following model for the interconnections between splicing and 3 0 end processing. Assembly of U1 on the 5 0 splice site of the terminal intron occurs first and serves to inhibit cleavage and polyadenylation. To overcome this inhibition, a tSA site is required to promote terminal exon definition in combination with a pA signal. Our data indicate that these early interactions are vital to couple splicing and 3 0 end formation, as a tSA site mutation inhibits both processes. Although splicing catalysis and intron excision occur shortly after terminal exon definition is established under normal circumstances, our results strongly suggest that splicing catalysis is dispensable for coupled 3 0 end formation. This is shown by the apparently efficient 3 0 end formation that takes place in the presence of SSA or U4 AMO, both of which block splicing before any catalytic steps.
Our data indicate that 3 0 end formation is linked to splicing once U1 is recruited to the transcript because, once this occurs, cleavage and polyadenylation is inhibited unless a functional 3 0 splice site is encountered. Such a mechanism is consistent with observations of widespread 3 0 end formation within introns on inhibition of U1 (17,21,22,57). This function of U1 must somehow be negated at gene ends. Although removal of the intron is one way that this could happen, our data suggest that 3 0 end formation can still occur when splicing catalysis is blocked by SSA or U4 AMO. Under these conditions, U1 is still able to inhibit pA site processing, suggesting that it remains bound to pre-mRNA [(22) and Supplementary Figure S7]. Therefore, the tSA site must play a role in buffering its negative effects or promoting cleavage and polyadenylation. This would presumably involve the recruitment of factors with this function.
U2AF65 is involved in 3 0 splice site recognition, terminal exon definition and has been shown to stimulate cleavage and polyadenylation. Interestingly, U2AF65 can promote b-globin 3 0 end processing when tethered artificially to the transcript, suggesting that this function does not require removal of an intron (40). Our finding that reduced 3 0 end formation on tSA mutation is also associated with a lack of U2AF65 recruitment is correlative with a potential positive function in 3 0 end formation. Other factors involved at the 3 0 splice site may also play a role here. In particular, a previous study identified SF3b components as being important for coupling splicing to 3 0 end formation in vitro (15). SF3b is also needed for branch-point recognition by U2, disruption of which shows a mild 3 0 end processing defect for some transcripts (Supplementary Figure S10). More generally, most splicing factors described as having potential function in cleavage and polyadenylation are associated with either U1 or U2 recruitment (15,40,58). This is consistent with our model that early events in splicing are important for coupling.
Global analyses of nascent transcripts show cotranscriptional removal of a large number of introns (24,26). However, co-transcriptional splicing is reported as being less efficient for introns located nearer to the end of pre-mRNAs (25,28). In those instances, 3 0 end formation might occur more rapidly than splicing. Indeed, available analyses indicate that some pA sites are processed in a matter of seconds with splicing occurring in minutes (59,60). Similarly, our actinomycin time course indicates that unspliced Myc transcripts are processed more slowly than non-pA cleaved RNA ( Figure 7B). Our data indicate that splicing and 3 0 end formation can be coupled in these situations owing to recruitment of U1 to the 5 0 splice site and U2AF65 (and presumably U2) to the 3 0 splice site. This is likely to be a rapid process, as both factors are present on transcribing Pol II (41,61,62).
It is well documented that many splice defective transcripts in lower and higher eukaryotes are rapidly degraded (36,53,63). Although this is also true of some transcripts following SSA treatment, other pre-mRNAs accumulate and are stabilized (35). We observe a similar accumulation and stabilization of P27 and Myc transcripts following U4 AMO treatment. This correlates with their release into the nucleoplasm, which we propose is facilitated by the continued cleavage and polyadenylation that occurs under these splice inhibitory conditions. Consistently, immunofluorescence experiments demonstrated that SSA or U4 treatment causes a large accumulation of polyadenylated RNA in nucleoplasmic foci that co-localize with the splicing speckle marker, SC35 (45) and our unpublished results. Previous work showed that unspliced P27 pre-mRNA can be translated following SSA treatment (45), which may implicate the nonsensemediated decay pathway in the turnover of these and similar transcripts. However, it is not known how unspliced pre-mRNA that accumulates in intra-nuclear speckles is eventually exported to the cytoplasm.
We have shown that the tight co-transcriptional coupling between splicing and 3 0 end formation is established early during terminal exon definition following U1 recruitment to the 5 0 splice site and recognition of the tSA site. These early events seem sufficient to permit 3 0 end formation, which continues following blocks to later steps in the splicing process.