Newly Initiated RNA Encounters a Factor Involved in Splicing Immediately upon Emerging from within RNA Polymerase II*

We employed RNA-protein cross-linking to map the path of the nascent RNA as it emerges from within RNA polymerase II. A UV-cross-linkable uridine analog was incorporated at two positions within the first five nucleotides of the transcript. Only the two largest subunits of RNA polymerase II cross-linked to the transcript in complexes containing 17–24-nucleotide (nt) RNAs. Extension of the RNA to 26 or 28 nt revealed an additional strong cross-link to the splicing factor U2AF65. In U17 complexes, in which the RNA is still contained within the polymerase, U2AF65 is tightly bound. In contrast, U2AF65 is more loosely bound in C28 transcription complexes, in which about 10 nt of transcript have emerged from the RNA polymerase. Cross-linking of U2AF65 to RNA in a C28 complex was eliminated by the addition of an excess of an RNA oligonucleotide containing the consensus U2AF65 binding site, but U2AF65 was not displaced by a nonconsensus RNA. These findings indicate that U2AF65 shifts from protein-protein to protein-RNA interactions as the RNA emerges from the polymerase. During transcription of one particular template at low UTP concentration, RNA polymerase II pauses just after synthesizing a transcript segment that is a U2AF65 binding site. Dwell time of the polymerase at this pause site was significantly and specifically reduced by the addition of recombinant U2AF65 to the transcription reaction. Therefore, the association of U2AF65 with RNA polymerase II may function not only to deliver U2AF65 to the nascent transcript but also to modulate efficient transcript elongation.

We employed RNA-protein cross-linking to map the path of the nascent RNA as it emerges from within RNA polymerase II. A UV-cross-linkable uridine analog was incorporated at two positions within the first five nucleotides of the transcript. Only the two largest subunits of RNA polymerase II cross-linked to the transcript in complexes containing 17-24-nucleotide (nt) RNAs. Extension of the RNA to 26 or 28 nt revealed an additional strong cross-link to the splicing factor U2AF65. In U17 complexes, in which the RNA is still contained within the polymerase, U2AF65 is tightly bound. In contrast, U2AF65 is more loosely bound in C28 transcription complexes, in which about 10 nt of transcript have emerged from the RNA polymerase. Cross-linking of U2AF65 to RNA in a C28 complex was eliminated by the addition of an excess of an RNA oligonucleotide containing the consensus U2AF65 binding site, but U2AF65 was not displaced by a nonconsensus RNA. These findings indicate that U2AF65 shifts from protein-protein to protein-RNA interactions as the RNA emerges from the polymerase. During transcription of one particular template at low UTP concentration, RNA polymerase II pauses just after synthesizing a transcript segment that is a U2AF65 binding site. Dwell time of the polymerase at this pause site was significantly and specifically reduced by the addition of recombinant U2AF65 to the transcription reaction. Therefore, the association of U2AF65 with RNA polymerase II may function not only to deliver U2AF65 to the nascent transcript but also to modulate efficient transcript elongation.
It is now appreciated that transcription is regulated during promoter clearance and transcript elongation as well as at transcription complex assembly (recently reviewed in Refs. [1][2][3][4][5][6]. Also, it is becoming increasingly apparent that transcription and RNA processing are both interconnected and interdependent (recent reviews include Refs. [7][8][9]. It is thus of considerable importance to carefully characterize the interaction of the transcription and processing machineries. We now have a detailed picture of RNA polymerase II structure (10,11), but how factors interact with this structure to modulate elongation and RNA processing is still very poorly understood. In particular, it is not certain how the transcript itself interacts with RNA polymerase II as transcription proceeds.
Immediately upstream of the point of bond formation in RNA polymerase II, nascent RNA remains in hybrid with the DNA template for 8 -9 bp. Structural information is available for the transcript in this region (12,13), but beyond the RNA-DNA hybrid, the location of the RNA within the transcription complex is unknown. A considerable segment of RNA remains inside the polymerase upstream of the point at which the transcript and template strand separate, since 17-19 nt of RNA within the transcription complex are protected from nuclease digestion and oligonucleotide hybridization probes (14 -16). Two possible exit paths have been proposed for the RNA (17,18). These models place the emerging transcript on opposing sides of the dock region of Rpb1, 1 the largest polymerase subunit. On path 1, the RNA passes around the base of the clamp domain toward subunit Rpb7, which contains both a ribonucleoprotein fold and an oligonucleotide binding fold (19,20). The Rpb4/Rpb7 heterodimer was shown to bind singlestranded DNA and RNA, and it has been suggested that Rpb7 may bind the emerging RNA transcript (10,11,13,19,20). On the alternative path, path 2, the RNA passes around the opposite face of RNA polymerase toward subunit Rpb8. Westover et al. (13) point out that in bacterial RNA polymerase only path 2 is electrostatically favorable. The simple extension of the RNA in their structure points toward this second groove (13). Several findings suggest that the transcript remains in contact with the polymerase for some time after emerging from within the active center. Our earlier results (16) are consistent with a stabilizing interaction between the 5Ј-end of the RNA and the RNA polymerase at the point where the nascent transcript is 30 -40 nt long. Hanna and Meares (21) showed that the 5Ј-end of transcripts generated with Escherichia coli RNA polymerase can cross-link to the polymerase up to a chain length of 94 nt.
In order to better understand the exit path of the transcript from RNA polymerase II, we synthesized RNAs 17 nt and longer, which contained both cross-linkable residues and radioactive labels at their 5Ј-ends. We were surprised to discover that the emerging RNA immediately contacts U2AF65, a protein known to be involved both in splicing (reviewed in Ref. 22) and in nuclear export (23,24). We show that the presence of U2AF65 can assist RNA polymerase in recovery into transcriptional competence at pause sites, and we suggest that the close association of U2AF65 and polymerase is important in facilitating the participation of U2AF65 in RNA processing events.

MATERIALS AND METHODS
Reagents-We obtained reagents from the following sources: fast protein liquid chromatography-purified NTPs from Amersham Biosciences, 32 P-labeled NTPs from PerkinElmer Life Sciences, Bio-Gel A1.5m from Bio-Rad, oligo(dT)-cellulose and DeepVent DNA polymerase from New England Biolabs, 5-iodo-UTP from Sigma, DNase I from Invitrogen, and streptavidin-coated paramagnetic beads from Promega. CpA dinucleotide was obtained as a custom synthesis from Dharmacon. DNA oligonucleotides were synthesized by Integrated DNA Technologies, Inc. The RNA oligonucleotides RNA2 and RNA3 used in Fig. 4C were synthesized with T7 RNA polymerase and subsequently purified on a 15% denaturing polyacrylamide gel. HeLa cells for nuclear extract preparation were purchased from the National Cell Culture Center. The 8WG16 antibody against RNA polymerase II was obtained from Abcam and the MC3 antibody against U2AF65 was a gift from Dr. Margarida Gama-Carvalho.
Transcription Templates-Templates for transcription were all based on the adenovirus major late promoter. The construction of the pML20-42, pML20-40(6G), and pML20-40(17G) plasmids is described elsewhere (25,26). Plasmid pML220 was constructed by PCR amplifying a 120-bp segment of pGR220 (27) with the following primers: 5Ј-CCGCTCGAGCACAAACGGCAAGATCGAGGG-3Ј and 5Ј-CCC-AAGCTTGCGGATAACAATTTCACACAGGAAACAGCTATGACC-3Ј. The PCR product was cut with restriction enzymes XhoI and HindIII and was cloned into pML20-42 between the XhoI and HindIII sites. This places the sequence CTTTTGTTCCCTTT 81 base pairs downstream from the transcription start site. To generate the pML222 plasmid, this sequence was altered by changing the underlined G to a T residue through the use of the Stratagene QuikChange TM site-directed mutagenesis kit.
The DNA templates for transcription, which ranged in size from 190 -298 bp, were produced by PCR. In all cases, the transcription start site was 96 bp from the upstream end of the fragment. For attached template experiments, the 5Ј primer was biotinylated. DNA was purified using the Concert Rapid PCR Purification System (Marligen Bioscience) according to the manufacturer's instructions.
Protein Purification-Recombinant human U2AF (consisting of U2AF35 and U2AF65) was expressed in and purified from baculovirusinfected High-Five cells (Invitrogen) as described previously (28) with some modifications. The insect cells (0.74 g) were resuspended in 6 ml of lysis buffer (50 mM Tris, pH 7.9, 500 mM NaCl, 10% glycerol, 5 mM imidazole, 5 mM ␤-mercaptoethanol, and complete EDTA-free protease inhibitor mix, from Roche Applied Science) and sonicated two times for 10 s on ice. Nonidet P-40 was added to a final concentration of 1%. The lysate was cleared by centrifugation, and the supernatant was mixed with 0.5 ml of 50% Ni 2ϩ -nitrilotriacetic acid-agarose (equilibrated with lysis buffer) for 1 h at 4°C. The resin was washed twice with 10 ml of lysis buffer plus 1% Nonidet P-40 and twice with 10 ml of wash buffer (50 mM Tris, pH 7.9, 500 mM NaCl, 20 mM imidazole, protease inhibitors, and 5 mM ␤-mercaptoethanol). The bound protein was eluted with two 1-ml aliquots of elution buffer (50 mM Tris, pH 7.9, 500 mM NaCl, 200 mM imidazole, protease inhibitors, and 5 mM ␤-mercaptoethanol) and was dialyzed overnight against BC100 (20 mM Hepes, pH 7.9, 100 mM KCl, 20% glycerol, 0.2 mM EDTA) plus 1 mM phenylmethylsulfonyl fluoride and 5 mM ␤-mercaptoethanol. In addition to the full-length U2AF65 and U2AF35, a 55-kDa fragment of U2AF65 was also visible by Coomassie staining after resolving the U2AF preparation on an SDS gel.
Depletion of U2AF from Nuclear Extract-HeLa cell nuclear extract was depleted of U2AF by passage through an oligo(dT)-cellulose column at 1 M KCl as described by Valcarcel et al. (29). A mock depletion reaction was carried out by making nuclear extract 1 M in KCl without exposure to oligo(dT)-cellulose, followed by dialysis against a 100 mM KCl buffer.
Assembly and Purification of Ternary Transcription Complexes-Preinitiation complexes were assembled by incubating a mixture containing 50% HeLa nuclear extract and 14 g/ml template DNA fragment at a final concentration of 75 mM KCl and 8 mM MgCl 2 at 30°C for 20 min, followed by gel filtration on Bio-Gel A1.5m to remove contaminating NTPs. The 1.9-ml gel filtration columns used BC100 (with 20 mM Tris-HCl, pH 7.9, instead of 20 mM Hepes) as the running buffer. Complexes U21 (pML20 -40(6G) template) and U17 (pML20 -40(17G) template) were generated by incubating preinitiation complexes with 1 mM CpA (initiating at position Ϫ1), 10 M 5-iodo-UTP, 1 M [␣-32 P]CTP, and 50 M dATP as the energy source at 30°C for 5 min. This produced initial transcripts of 6 nt (pML20 -40(6G) template) or 7 nt (pML20 -40(17G) template). These transcripts were then chased for 5 min at 30°C by the addition of 200 M CTP, UTP, and GTP (to generate U21 complexes) or 200 M CTP, UTP, and ATP (to generate U17 complexes). The U21 and U17 complexes were then purified by the addition of Sarkosyl to 1% followed by gel filtration in MEM buffer (30 mM Tris-HCl, pH 7.9, 10 mM ␤-glycerophosphate, 62.5 mM KCl, 0.5 mM EDTA, and 1 mM dithiothreitol). This Sarkosyl rinsing procedure is described in detail in Ref. 30. The transcripts in Sarkosyl-rinsed complexes were elongated by incubation for 5 min at room temperature with a 50 M concentration of the appropriate NTPs as indicated in the figure legends.
The experiment in Fig. 5 used templates attached to beads, which were assembled into preinitiation complexes as described (31). A23 complexes were generated by incubation with 100 M ATP, 10 M UTP, and 1 M [␣-32 P]CTP at 30°C for 5 min, followed by the addition of 10 M CTP for an additional 5 min. After Sarkosyl rinsing, the A23 complexes were chased with all four NTPs at 30°C as indicated in the legend to Fig. 5.
Cross-linking-Samples were placed in microtiter plates in an ice water bath and irradiated for 20 min at 312 nm as described by Bartholomew et al. (32) using a Fisher Biotech transilluminator in the presence of aprotinin and leupeptin at 10 g/ml. They were then treated with DNase I (0.3 units/l) and, where indicated, RNase A (50 ng/l) for 15 min at 37°C. An aliquot from each sample was extracted with phenol and chloroform, ethanol-precipitated, and resolved on a 13% denaturing polyacrylamide gel to check the RNA content. The rest of the samples were dissolved in SDS loading buffer and resolved on 4 -20% gradient polyacrylamide-SDS gels (Invitrogen). Gels were visualized on a STORM Imager (Amersham Biosciences).
Immunoprecipitations-Cross-linked complexes were immunoprecipitated under nondenaturing conditions or after dissociation of the proteins under denaturing conditions. Complexes shown in Fig. 2 (left) were denatured in the presence of 1% SDS and 10 mM dithiothreitol by heating at 95°C for 5 min and were subsequently diluted to a final concentration of 50 mM Hepes, pH 7.5, 150 mM NaCl, 1% Triton X-100, 0.1% SDS. Samples in the middle panel of Fig. 2 were in 50 mM Hepes, pH 7.5, 150 mM NaCl, and 1% Triton X-100, and samples on the right were in the buffer reported by Robert et al. (33). All samples contained aprotinin and leupeptin at 10 g/ml and were DNase I-digested (and RNase A-treated for samples in the right panel) before immunoprecipitation overnight at 4°C with the 8WG16 anti-Rpb1 antibody (0.5 g/ml) or MC3 anti-U2AF65 antibody (5 l/ml of reaction).
Gel Filtration-For the experiments in Fig. 4, A and B, 400-l samples of transcription complexes were loaded onto a 40-ml (0.9 ϫ 57-cm) gel filtration column (Bio-Gel A1.5m) that was equilibrated with MEM buffer. For Sarkosyl and salt rinsing, the column was preloaded with 400 l of 400 mM KCl and 1% Sarkosyl in minimal essential medium before loading the sample that was also brought up to a final 400 mM KCl and 1% Sarkosyl. The excluded volume was collected in 750-l fractions. Samples were supplemented with 8 mM MgCl 2 and 10 g/ml aprotinin and leupeptin. For the experiment in Fig. 4A, transcription complexes were advanced to U32 by the addition of 50 M CTP, UTP, and GTP at room temperature for 5 min. Cross-linking took place as described above for 20 min. Samples were DNase I-digested (and, in A, RNase A-digested) before concentration by trichloroacetic acid precipitation.

RESULTS
Structural studies have shown that the catalytic center of RNA polymerase II resides in a cleft formed by the two largest subunits, Rpb1 and Rpb2 (recently reviewed in Ref. 34). In order to investigate the path of the nascent RNA as it emerges from this central cleft, we have carried out RNA-protein crosslinking experiments in which UV cross-linkable residues were located near the 5Ј-end of the RNA. An earlier study (32) demonstrated UV cross-linking of 19 -40-nt RNAs containing 4-thiouridine to the two largest subunits of RNA polymerase II. We found that whereas 4-thio-UTP was an acceptable substrate for extension of relatively long RNAs (ϳ150 nt) by RNA polymerase II, the polymerase would not incorporate 4-thio-UTP in the initial 5 nt of the transcript (data not shown). We therefore employed 5-iodo-UTP as our cross-linkable nucleotide. 5-Iodouridine shows specific photoreactivity with aromatic and sulfur-containing residues in proteins (35).
Cross-linking of the 5Ј Region of the Transcript to Components of the Transcription Complex-The template employed in the initial phases of our study, called pML20 -40(6G), is derived from the adenovirus major late promoter. The first 30 nt of the transcript from this template are shown in Fig. 1A. RNA polymerase II transcription complexes were assembled by incubating pML20 -40(6G) DNA in HeLa nuclear extracts. Transcription was initiated with the dinucleotide primer CpA in the presence of low concentrations of 5-iodo-UTP and radioactive CTP. This resulted in RNA synthesis up to the G stop at position ϩ5. After the addition of GTP and high concentrations of UTP and nonlabeled CTP, transcription proceeded to position ϩ20. These complexes contained two 5-iodouridine residues and two neighboring radiolabeled cytidines at the 5Ј-end, but no cross-linkable or labeled residues were present beyond the sixth base (see Fig. 1A). We will use the convention of designating transcripts and the associated complexes by the length of the RNA and the final base incorporated; thus, pausing of CpA-primed transcription at ϩ20 on pML20 -40(6G) generates U21 complexes.
U21 complexes were partially purified by Sarkosyl rinsing, which involves transient exposure to the detergent Sarkosyl during gel filtration (see "Materials and Methods"). As shown in Fig. 1B, transcripts in the Sarkosyl-rinsed U21 complexes could be extended very efficiently to generate A24, G26, or C28 complexes. Reactions containing U21, A24, G26, and C28 complexes were exposed to UV light, and the proteins that were labeled by cross-linking to the RNA were visualized after SDS-PAGE (Fig. 1C). Cross-linking of the U21 and A24 reactions resulted in the labeling of two major protein bands, of roughly 220 -240 and 135 kDa (Fig. 1C, lanes 2 and 3). These are, as expected, the two largest subunits of RNA polymerase II, Rpb1 and Rpb2, based on their molecular weights and the reactivity of the upper band with anti-Rpb1 antibodies (Fig. 2) (see also Ref. 32). Reactions in which the transcript was extended only 2 or 4 bases further downstream, to G26 or C28, showed an additional band upon UV cross-linking (Fig. 1C, lanes 4 and 5). After RNase A treatment to truncate the cross-linked RNA, this protein displayed an apparent molecular mass of 65 kDa (Fig. 2, right). There was no significant cross-linking in control reactions, which were not exposed to UV (Fig. 1C, lane 1) or in which no 5-iodo-UTP was used (not shown).
Identification of Proteins Cross-linked to the Transcript-Reactions containing U21 or C28 complexes were UV-crosslinked and then immunoprecipitated with the anti-Rpb1 antibody 8WG16 (Fig. 2). In control reactions, which were denatured before precipitation (lanes 1-4), only the largest of the three labeled bands was recovered, confirming that this protein is the largest subunit of RNA polymerase II. When the precipitations were done under native conditions, all three bands were recovered (lanes [5][6][7][8], indicating that the 65-kDa protein is part of the transcription complex and not simply cross-linked to free RNA that might have been released during the transcription reaction. We were surprised at the size of the 65-kDa cross-linked protein. We had expected from our detergent rinsing protocol that only core subunits of RNA polymerase II would be present in the transcript elongation complexes. However, no subunit of RNA polymerase II has a molecular mass close to the observed size. (Human Rpb2 and Rpb3 are 134 and 31 kDa, respectively; see Ref. 36). We considered the possibility that, contrary to expectation, general transcription factors might remain in our complexes. The 62-kDa subunit of TFIIH and the largest subunit (58 kDa) of TFIIF have roughly the correct molecular weight, but we could find no evidence that either protein is responsible for the 65-kDa band (data not shown). A critical clue was provided by the observations of Robert et al. (33), who showed that the splicing factor U2AF65 copurifies with RNA polymerase II through a TFIIS affinity column. The identity of the 65-kDa band was established by the fact that it could be recovered from cross-linking reactions using a monoclonal an-  1-4) or under nondenaturing condition (lanes 5-12). 10 -30% of the input (I) was also loaded on the gels. Rpb1 and U2AF65 (left) were identified by precipitation with the appropriate antibody. The identification of Rpb2 (left) is inferred from the apparent molecular weight and co-precipitation with anti-Rpb1. The migration positions for two molecular weight markers are shown on the right. Samples in lanes 9 -12 were treated with RNase A to allow accurate comparison with the markers. tibody (37) against U2AF65 (Fig. 2, lanes 9 -12).
U2AF65 (U2 small nuclear ribonucleoprotein auxiliary splicing factor, large subunit), together with U2AF35, forms the essential splicing factor U2AF (reviewed in Ref. 22). U2AF65 recognizes the polypyrimidine tract of introns and recruits U2 small nuclear ribonucleoprotein to the branch point, whereas U2AF35 specifies the AG dinucleotide step at the 3Ј splice site (22,38). U2AF65 contains an N-terminal RS domain and three C-terminal RNA recognition motifs (RRM domains) (38). Biochemical and structural studies suggest that only the first two RRM domains are involved in RNA recognition, and the third RRM represents a novel class of protein recognition motifs (39). U2AF65 has also been shown to be involved in mRNA export (23). Recent studies on the Drosophila U2AF large subunit (U2AF50) implicate dU2AF50 in the expression and nuclear export of transcripts from both intron-containing and intronless genes (24).
The minimal length of RNA which allows cross-linking to U2AF65 is informative. As noted above, about 17 bases of the nascent transcript are contained within RNA polymerase II, as judged by protection against both nuclease and oligonucleotide probes (14 -16). U2AF65 protects 17 bases of a consensus RNA binding site from nuclease attack (see Ref. 40 and references therein); however, at another site, only a short polypyrimidine sequence of 6 nt is protected (41). Since U2AF65 recognizes its 17-nt site through two RRM domains (40), a single binding domain would recognize 8 -9 nt, consistent with the protection seen at the nonconsensus site (41). Thus, our ability to detect cross-linking beginning with a 26-nt RNA suggests that U2AF65 binds to RNA immediately upon its emergence from within the RNA polymerase, probably through the use of a single RRM domain.
Transcription in a U2AF-depleted Nuclear Extract-In order to investigate the role U2AF65 plays during transcription, U2AF65 and its associated factor U2AF35 were depleted from HeLa nuclear extract by passage through an oligo(dT)-cellulose column at 1 M KCl followed by dialysis back to 100 mM KCl (29). A control extract was cycled through the same salt concentration changes without the addition of oligo(dT)-cellulose. Transcription complexes were efficiently assembled in both extracts. These complexes could be walked from U21 to C28 as shown in the left panel of Fig. 3. Cross-linking reactions with C28 complexes assembled with the depleted extract did not show the 65-kDa band (lane 11), but this band was restored by the addition of recombinant U2AF65 and U2AF35 (lane 12). This result confirms the assignment of the 65-kDa band to U2AF65 cross-linking. We observed a new band, marked by an asterisk in Fig. 3, just above the 55-kDa marker in those reactions (lanes 9 and 12) that received recombinant U2AF65. We believe this band results from cross-linking of RNA to an N-terminal fragment of U2AF65, since our recombinant protein preparation contains an additional band with the same apparent molecular weight that is recognized by both the anti-U2AF65 monoclonal antibody and by an anti-His tag antibody (data not shown). The presence of this band even in C28 complexes assembled with the control extract and supplemented with recombinant U2AF65 (lane 9) suggests that U2AF65 is in equilibrium between the transcription complex and free solution.
U2AF65 Is Tightly Associated with the Early Transcription Complex but Switches from Protein-Protein to Protein-RNA Interactions as the Transcript Emerges-Coulombe and co-workers (33) demonstrated that an anti-U2AF65 antibody will coprecipitate U2AF65 and RNA polymerase II from a HeLa cell extract that was extensively RNase-treated. This provides evidence for a direct protein-protein interaction between U2AF65 and free RNA polymerase II. We wanted to determine how strongly U2AF65 is bound in our transcription complexes and whether this interaction depends on the length of the transcript. We were initially surprised to find U2AF65 in our transcription reactions, because the Sarkosyl rinsing procedure was expected to remove all proteins other than RNA polymerase subunits. However, our usual protocol involves a rapid gel filtration step performed at relatively low resolution. We therefore decided to test whether U2AF65 remained in association with our complexes after transient exposure to detergent and high salt concentrations followed by high resolution gel filtration. We used a variant of our original template, pML20 -40(17G), for this experiment (Fig. 4A). Transcription of pML20 -40(17G) in the presence of CpA, radiolabeled CTP, and 5-iodo-UTP resulted in the synthesis of a 7-nt RNA. Subsequent chase with ATP, UTP, and nonlabeled CTP generated a 17-nt transcript. As noted above, this RNA is not long enough to be accessible to external probes. The U17 complexes were Sarkosyl-rinsed by the usual rapid protocol. We set aside 20% of this reaction as a control. The remainder was brought to 1% Sarkosyl and 400 mM KCl, followed by fractionation on a high resolution 40-ml gel filtration column (see "Materials and Methods"). The excluded volume containing the transcription complexes was incubated with GTP, CTP, and UTP to generate U32 complexes, followed by UV cross-linking. The control sample was divided into two parts and then treated similarly, except that half of the reaction was diluted 7.5-fold to achieve similar dilution conditions to the gel filtration sample. All of the cross-linked reactions were concentrated by precipitation before analysis by SDS-PAGE. The results (Fig. 4A) show that a significant amount of U2AF65 remained with the transcription complex during gel filtration, even after exposure to 1% Sarkosyl and 400 mM KCl. (Note that we confirmed that free U2AF65 was well separated from the void volume on this column; data not shown.) Since U2AF65 cannot be interacting with RNA in the U17 complex, this indicates a strong proteinprotein interaction between U2AF65 and RNA polymerase II in this ternary transcription complex.
We repeated the gel filtration experiment with C28 complexes, in which the RNA is accessible to U2AF65. In this case, the gel filtration was performed under native conditions, without the addition of Sarkosyl or 400 mM salt. As shown in Fig.  4B, in this case, all of the U2AF65 was removed from the complex by the gel filtration column. Based on the results FIG. 3. Nuclear extract depleted of U2AF is functional in transcription. Mock-depleted (Mock Depl.) and U2AF-depleted nuclear extract were prepared as described under "Materials and Methods" and used to assemble either U21 or C28 complexes as indicated. Recombinant U2AF65 (final concentration of 7 ng/l) was added to the indicated lanes before cross-linking. RNA samples from the reactions (lanes 1-6) were resolved on a 15% polyacrylamide gel. Cross-linked proteins were resolved on a 4 -20% SDS gel (lanes 7-12). The migration positions of molecular weight markers are shown on the right. The additional band indicated by the asterisk in lanes 9 and 12 results from the presence of an N-terminal fragment of U2AF65 in our recombinant preparation (see "Results").
shown in Fig. 4, A and B, it appears that once RNA has emerged from within the polymerase, U2AF65 binds more loosely to the transcription complex, perhaps because its binding is now divided between RNA and protein sites. To address the question of whether any RNA can dissociate U2AF65 from the transcription complex, we incubated C28 complexes prior to cross-linking with an excess of either an RNA oligonucleotide containing the consensus binding site of U2AF65 (RNA3) (see Ref. 42) or a control RNA, RNA2, whose sequence does not resemble a U2AF65 binding site. As shown in Fig. 4C, RNA3 eliminated the cross-link to U2AF65 in C28 complexes, but there was no effect of RNA2. This indicates that the interaction between the 5Ј-end of the transcript and U2AF65 can be competed away by a consensus RNA but not by any RNA sequence.
U2AF65 Can Assist RNA Polymerase II in Crossing Pause Sites-The results in Fig. 4C indicate that U2AF65 should bind strongly to a consensus U2AF65 site within the nascent RNA as soon as that site emerges from within the RNA polymerase. This could have important functional consequences for transcript elongation, since the reverse threading of the transcript that accompanies extended pausing and arrest by RNA polymerase (14,43,44) should be prevented by a tight protein-RNA interaction at the mouth of the RNA exit channel (see Ref. 15). During the course of our studies, we were made aware of unpublished results from the Conaway laboratory, 2 which suggested such a function for U2AF65. These workers showed that RNA polymerase II pauses extensively at low UTP concentrations within a pyrimidine-rich region (nontemplate strand) on the pGR220 template (27). This pause was relieved by a chromatographic fraction enriched for U2AF65. We constructed a new template with a 120-bp region around the pGR220 pausing site cloned downstream of the adenovirus major late promoter. This sequence encodes a possible U2AF65 binding site, starting at ϩ81 (CTTTTGTTCCCTTT) in our construct. In preliminary tests, we observed pausing of RNA polymerase II just downstream of this sequence under limiting UTP conditions. The pausing was partially relieved in a dose-dependent fashion by the addition of recombinant U2AF65 (data not shown). To further investigate this effect, we have modified the original template with a single base change to create an uninterrupted pyrimidine stretch on the nontemplate strand; the corresponding RNA transcript of this sequence more closely resembles a consensus U2AF65 binding site (see Fig. 5A). A U2AF-depleted nuclear extract was used to assemble radiolabeled A23 transcription complexes. After Sarkosyl rinsing, these complexes could be chased to the end of the template with 1 mM NTPs (Fig. 5A, lane 1). If the UTP concentration in the chase was lowered to 10 M, complexes paused before the addition of each U residue (lanes 2-6 and 7-11). In the presence of recombinant U2AF, complexes generally proceeded through these pauses slightly faster than in its absence. The most significant effect of the U2AF addition was observed at positions 106 -108 (marked by the dashed rectangle in Fig. 5A), which is just downstream of the U2AF65 binding site. The intensities of the bands corresponding to these pauses were quantitated and plotted as a function of time (Fig. 5B). The presence of U2AF does not prevent the initial pausing event (10-min points), but it does allow the polymerase to escape the pause much more rapidly than in its absence (20-and 30-min points). This is consistent with a model in which U2AF65 scans the RNA as it is being synthesized by the polymerase. U2AF65 should bind with high affinity to its consensus RNA binding site. If this binding is competitive with backtracking by the polymerase, it would prevent prolonged reverse threading of the transcript and facilitate resumption of transcription. For the template used in Fig. 5, a polymerase paused at the central base of the pause site would protect about 17 bases upstream, which would place just over half of the polypyrimidine segment of the RNA outside of the polymerase. U2AF binding to this emerging consensus site should block backtracking by the polymerase. DISCUSSION In the course of mapping the path of the 5Ј-end of the RNA during its exit from RNA polymerase II, we had expected to detect interactions of the transcript with subunits other than Rpb1 and Rpb2, which contain the polymerase active site and enclose the RNA-DNA hybrid (12,13). However, we discovered that as the nascent RNA emerges from within RNA polymerase II, it immediately encounters U2AF65, a factor involved in splicing and in nuclear export. In HeLa whole cell extracts, U2AF65 is primarily associated with the hypophosphorylated form of RNA polymerase II (33), which is the form competent to assemble into preinitiation complexes. In contrast, RNA polymerase in the transcription complexes we studied should be phosphorylated in the C-terminal domain (CTD) of the largest 2 J. Conaway, and R. Conaway, personal communication.
FIG. 4. U2AF65 shifts from protein-protein to protein-RNA interactions as the RNA emerges from the polymerase. In all panels, samples were resolved on 4 -20% SDS-polyacrylamide gels. A, the initial transcript sequence from the pML20 -40(17G) template is shown. In the initial step of the transcription reaction, RNA polymerase II incorporated 5-iodouridine residues (rectangles) and 32 P-labeled cytidine residues (asterisks). U17 complexes were assembled on this template and processed as shown in the flow diagram (described in detail under "Materials and Methods"). Half of the control reaction was diluted 7.5-fold to achieve a dilution similar to that which occurred during gel filtration. All samples were walked to U32 with the addition of the appropriate NTPs, cross-linked, treated with DNase I and RNase A, and trichloroacetic acid-precipitated. B, Sarkosyl-rinsed U21 complexes were made on template pML20 -40(6G) as described in the legend to Fig. 1, walked to C28, and then processed as shown in the flow diagram (described in detail under "Materials and Methods"). Cross-linked samples were DNase I-digested and trichloroacetic acid-precipitated. C, C28 complexes on the pM20 -40(6G) template were incubated with RNA2, RNA3, or no RNA before cross-linking (RNAs were added to a final concentration of 3 M). The sequence of RNA3 is GGACUUUUUUUC-CCUUUUUUUCC, which contains the consensus U2AF65 binding site (42). The sequence of RNA2 is GGACAAAAAAACCCAAAAAAACC.
subunit. In order to explore the possible relationship of CTD phosphorylation and U2AF65 association with the transcription complex, we treated our transcription reactions prior to cross-linking with protein phosphatase-1, which will dephosphorylate the polymerase II CTD (45). To ensure that the CTDs of transcriptionally active RNA polymerases were substantially dephosphorylated, we checked the mobility of the labeled Rpb1 subunit after cross-linking. As expected, the mobility of this band increased after phosphatase-1 treatment. An initial titration allowed us to determine the amount of phosphatase sufficient to complete the shift in mobility of the Rpb-1 band. When reactions were treated with this level of phosphatase-1 and then cross-linked, there was no change in the intensity of the U2AF65 band (data not shown). This suggests that U2AF65 is not retained within the transcription complex via the phosphorylated CTD.
It is important to note that the Rpb7 subunit of RNA polymerase II is located near the mouth of proposed RNA exit channel 1 (10,11). This subunit contains potential RNA binding domains (19,20). Also, the association of yeast Rpb7 with RNA polymerase II can be completely disrupted by rinsing with a mild denaturant (46). One might therefore suspect that Rpb7 was removed from our complexes by the Sarkosyl rinsing procedure, thereby allowing access of U2AF65 to the transcript. To guarantee the presence of Rpb7 in our experiments, we supplemented our reactions after the Sarkosyl rinsing step with an excess of recombinant human Rpb4 and Rpb7 (gifts from Finn Werner). We observed the same level of cross-linking of U2AF65 to the nascent RNA regardless of whether Rpb4 and Rpb7 were added (data not shown). Thus, it seems unlikely that Rpb7 competes with U2AF65 for association with the polymerase or binding to the RNA.
We demonstrated (Fig. 4A) that U2AF65 is present in early transcription complexes when the RNA is still contained within the polymerase. This indicates that U2AF65 can initially interact with the transcription complex through protein-protein interactions. Note that U2AF contains two protein interaction regions, one of which (the third RRM domain) belongs to a novel class of protein recognition motifs (39). All of our results, taken together with those of Robert et al. (33), suggest that U2AF65 is recruited to RNA polymerase II before the start of transcription. This positions U2AF65 to interact immediately FIG. 5. The addition of U2AF helps RNA polymerase II cross a pause site just downstream of a U2AF65 binding site. A, Sarkosyl-rinsed A23 transcription complexes were assembled on the pML222 template using U2AF-depleted nuclear extract as described under "Materials and Methods." In lane 1, these complexes were run off the template by the addition of 1 mM NTPs for 10 min. In lanes 2-11, complexes were incubated with 1 mM ATP, CTP, and GTP and 10 M UTP, with or without the addition of U2AF (29 ng/l), for the indicated times. RNA ladders (lanes 12-15) were synthesized on the same template using reactions limiting in each of the NTPs (31). RNAs were resolved on an 8% denaturing polyacrylamide gel. The sequence of the transcript from ϩ81 to ϩ109 is shown on the right. The putative U2AF65 binding sequence is indicated by a solid box, and the downstream RNA polymerase pause site is shown by a dashed box. B, the number of transcription complexes paused at positions 106 -108 were quantitated, and values were plotted as a function of time. Open squares, complexes without U2AF65; filled squares, complexes with the addition of U2AF65. with the emerging RNA, even if the transcript sequence is not a close match to the U2AF65 consensus binding site. For example, we demonstrated that U2AF65 in C28 complexes can be cross-linked to an RNA with the initial transcript sequence of 5Ј-CAGUGUGUU (data not shown). We propose that after transcript initiation, U2AF65 retains its interaction with the RNA polymerase, allowing it to scan the newly synthesized RNA for the presence of polypyrimidine tracts without the requirement for high affinity for all RNA sequences. Retention of U2AF65 in the transcription complex may also be important for those transcription units that lack introns, since U2AF65 is involved in nuclear export of both intron-containing and intronless transcripts (24).
It would be particularly disadvantageous for splicing if RNA polymerase II paused near the 3Ј intron-exon junction, which in the vast majority of cases is located just downstream of a polypyrimidine tract. A paused RNA polymerase would obstruct recognition of this junction and assembly of the spliceosome. Zhang et al. (44) have shown that extended pausing by RNA polymerase II results in some backtracking of the polymerase along the template, which must be reversed for transcription to resume. The interaction of an RNA-binding protein with the transcript at the point of exit from the polymerase should resist reverse threading of the RNA and the associated polymerase backtracking. Consistent with this expectation, we showed ( Fig. 5) that U2AF65 can assist polymerase in returning to transcriptional competence when the polymerase pauses immediately downstream of a polypyrimidine tract. Therefore, U2AF65 could play a role in facilitating splice site recognition by guaranteeing the continuous forward movement of RNA polymerase II at these sites. This idea agrees with other recent observations that demonstrate the influence of the splicing machinery on transcription (8,47,48).
In summary, we propose that U2AF65 is recruited to transcription complexes before RNA synthesis begins via interaction with RNA polymerase II. As the transcript is elongated, U2AF65 scans the emerging RNA without losing contact with the RNA polymerase. This positions U2AF65 to effectively search for its binding sites during transcription despite competition from RNA packaging factors. It also allows U2AF65 to affect transcriptional pausing by opposing the reverse threading of the transcript. When a consensus binding site is synthesized, U2AF65 presumably shifts completely to protein-RNA interaction and binds tightly to the RNA.