Interaction between Transcription Elongation Factors and mRNA 3′-End Formation at the Saccharomyces cerevisiae GAL10-GAL7 Locus*

Spt6 is a conserved transcription factor that associates with RNA polymerase II (pol II) during elongation. Spt6 is essential for viability in Saccharomyces cerevisiae and regulates chromatin structure during pol II transcription. Here we present evidence that mutations that impair Spt6, a second elongation factor, Spt4, and pol II can affect 3′-end formation at GAL10. Additional analysis suggests that Spt6 is required for cotranscriptional association of the factor Ctr9, a member of the Paf1 complex, with GAL10 and GAL7, and that Ctr9 association with chromatin 3′ of GAL10 is regulated by the GAL10 polyadenylation signal. Overall, these results provide new evidence for a connection between the transcription elongation factor Spt6 and 3′-end formation in vivo.

Previous work has shown that Spt6 is broadly utilized in pol II transcription (17,18), and data from several laboratories implicates Spt6 as a pol II elongation factor (16, 18 -20). Spt6 has been shown to interact with histones and is involved in the organization of chromatin structure over transcribed regions (21)(22)(23). Spt6 has also been implicated in RNA processing, because Drosophila Spt6 is associated with the nuclear exosome (24). Additional roles for Spt6 in the transcription cycle probably also exist.
The Paf1 complex is a multisubunit complex that associates with pol II and localizes to transcribed regions (14,16,(25)(26)(27)(28)(29). Although all of the roles of the Paf1 complex remain to be elucidated, recently, some complex members have been shown to be required for cotranscriptional ubiquitylation of histone H2B at position 123 (H2B Lys-123) and methylation of histone H3 lysines at positions 4 (Lys 4 ), 36 (Lys 36 ), and 79 (Lys 79 ) (30 -34). Thus, the Paf1 complex appears to coordinate histone modifications with transcription. Mutations in genes encoding Paf1 complex members also exhibit a wide spectrum of genetic interactions with mutations in elongation factor genes such as SPT4, SPT5, SPT16, and POB3 (14,35,36). As pol II goes through the cycle of transcription initiation, elongation, and termination, it is presented with different tasks. Many of these tasks are coordinated through the phosphorylation of the largest subunit of pol II on its conserved C-terminal domain, which in yeast consists of 26 repeats of consensus Tyr 1 -Ser 2 -Pro 3 -Thr 4 -Ser 5 -Pro 6 -Ser 7 (reviewed in Refs. 2 and 37-39). Upon initiation of transcription, unphosphorylated pol II becomes phosphorylated on serine 5 of the C-terminal domain repeats, allowing recruitment of the mRNA capping machinery, and the nascent mRNA is capped as it emerges from the pol II complex. During elongation, serine 5 phosphorylation decreases and serine 2 becomes hyperphosphorylated (40). Phosphorylation of serine 2 has been implicated in recruitment and regulation of mRNA 3Ј-end formation factors as well as termination by pol II (41,42). In addition, elongating pol II has the ability to monitor transcript composition, because splicing and export factor recruitment can be enhanced by the presence of introns in a transcript (43). Thus, the pol II holoenzyme is dynamic, undergoing a characteristic sequence of reversible modifications and associating with a variety of accessory factors during the course of a single round of transcription.
Following transcription of a DNA-encoded polyadenylation signal by pol II, the nascent RNA is cleaved and polyadenylated. Concurrent or subsequent to these events, pol II terminates transcription at a site downstream. The relationship between mRNA 3Ј-end formation and transcription termination is complex. Recent evidence shows that termination and cleavage may be uncoupled, although both require transcription of a polyadenylation signal (44 -47).
In this work, we present analysis of the role of Spt6 at the well studied GAL10-GAL7 genes. We present evidence that spt6 mutations, as well as spt4 mutations and mutations that impair pol II, can alter mRNA 3Ј-end formation at GAL10. In addition, our results show that both Spt6 and a member of the Paf1 complex, Ctr9, are cotranscriptionally recruited to GAL10-GAL7. However, whereas Spt6 is physically associated with the intergenic region 3Ј of the GAL10 polyadenylation signal, Ctr9 levels are significantly lower in this region. Fi-nally, our results suggest a functional interaction between Spt6 and Ctr9 based on both chromatin immunoprecipitation and genetic experiments. Taken together, these results suggest a reorganization of the pol II holoenzyme in response to 3Ј-end formation signals during transcription termination in vivo.

EXPERIMENTAL PROCEDURES
S. cerevisiae Strains, Plasmids, and Media-All S. cerevisiae strains (Table I) are isogenic to a GAL2 derivative of S288C (48). DNA encoding a FLAG epitope was fused to the 5Ј-end of the SPT6 open reading frame in plasmid pCC11 (49), using standard techniques (50,51). This SPT6-FLAG genomic fragment was then subcloned into pRS306 (52), and this plasmid, pCK40, was used to integrate SPT6-FLAG in place of SPT6, using a standard two-step gene replacement method. Thus, SPT6-FLAG is the sole source of Spt6 in strains with SPT6-FLAG. Strains with this fusion are wild type with respect to all spt6 mutant phenotypes tested, including growth, temperature sensitivity, and suppression of insertion mutations (data not shown). RPB3-HA1::LEU2 (53) was crossed from GHY645, a kind gift from G. Hartzog, University of California, Santa Cruz, CA. Plasmids made using PCR were verified by DNA sequencing. Sequence encoding nine copies of the MYC-epitope tag was amplified along with a kanMX4 cassette from pCK185 and integrated 3Ј of CTR9, creating CTR9 -9MYC::kanMX4. pCK185 is a kanMX4 derivative of pWZV87 (54) with a kanMX4 cassette replacing a BglII fragment of the Kluyveromyces lactis TRP1 gene. The genes encoding each of these epitope-tagged proteins are integrated into the genome, expressed from their natural promoter, and are the sole source of each particular protein in the cell. The epitope tags do not confer any apparent phenotypes to the wild type strain, and any differences between wild type and spt6 -1004 are presumed to rest solely with the spt6 -1004 mutation. A set of spt6 internal deletions were constructed by standard methods using PCR, subcloned into pCK40, and then used to replace genomic SPT6 as described above. spt6 -1004 encodes a version of Spt6 lacking residues 931-994. For characterizing genetic interactions between SPT6 and PAF complex genes, spt6 -1004, spt6 -1006, and spt6 -1002 were crossed to strains CKY87 (paf1⌬::kanMX4), CKY91 (cdc73⌬::kanMX4), or CKY95 (ctr9⌬kanMX4). The resulting diploids were transformed with pCK25 (SPT6-FLAG CEN URA3) so that, in case of double mutant inviability, double mutant spores would be complemented by SPT6 on the plasmid. After sporulation, at least 18 tetrads were analyzed from each cross. paf1⌬, cdc73⌬, and ctr9⌬ were followed in the crosses by resistance to G418 (PerkinElmer Life Sciences) conferred by the kanMX4 cassette. spt6 alleles were scored by strong suppression of lys2-128␦ (growth on medium lacking lysine) after selection against pCK25 on 5-fluoroorotic acid plates. If a spore was 5-fluoroorotic acid-sensitive, the presence of spt6 -1004 was inferred by segregation analysis. The gal7⌬6 mutation (constructed similarly as in Ref. 55) was constructed using a QuikChange kit from Stratagene. The major polyadenylation signal of GAL10 was also deleted by this strategy, creating gal10⌬56, which is similar to gal10⌬55 (55). gal10⌬56 removes nucleotides from ϩ74 through ϩ128 downstream of the GAL10 stop codon and appears to create a cryptic polyadenylation signal at the deletion junction (data not shown). The gal10 and gal7 mutations were integrated into the genome by standard two-step gene replacement. For experiments involving GAL gene expression, strains were grown in YP (1% yeast extract (Difco), 2% peptone (Difco)) supplemented with 2% raffinose and 2% galactose for at least 8 h and harvested at a concentration of 1-2 ϫ 10 7 cells/ml. For examination of the kinetics of GAL gene induction, cells were grown for at least 8 h in YP plus 2% raffinose and were then induced by the addition of 2% galactose. RNA was prepared from cells at the indicated times after induction. All other media were made as described previously (56).

This study
Western Blotting and ChIP Analysis-Protein extracts were prepared by growing cells to mid-log phase in the appropriate medium, followed by glass bead disruption in lysis buffer (25 mM NaPO 4 , pH 7.5, 200 mM KOAc, 2 mM MgOAc, 10% (v/v) glycerol, 1 mM phenylmethylsulfonyl fluoride) (20). For Western and ChIP analysis, antibodies used were: M2 (Sigma), mouse anti-FLAG IgG; 12CA5 (ascites, kind gift of E. Harlow, Harvard Medical School), mouse anti-HA IgG; and A14 (Santa Cruz Biotechnology), rabbit anti-MYC. ChIP analysis was performed as described (57,58) with slight alterations. Briefly, 200 ml of mid-log phase cultures were cross-linked with 1% (v/v) formaldehyde for 20 min. Cross-linking was then quenched by addition of 30 ml of 2.5 M glycine for 5 min. Cells were washed with TBS (100 mM NaCl, 20 mM Tris-HCl, pH 7.5) and disrupted with glass beads in lysis buffer. Crosslinked material was then washed twice with 1 ml of FA buffer (solution as described before (59) but with 300 mM NaCl, no SDS) and resuspended in 1 ml of FA buffer. Cross-linked material was solubilized by sonication to shear chromatin to ϳ400-bp fragments. Sonication conditions were 2 ϫ 20 s at setting 4, 4 ϫ 20 s at setting 8, and 2-4 ϫ 20 s at setting 10, as needed (Fisher Scientific Sonic Dismembrator Model 50). Approximately 5-10% of this sonicated material and 2 l of the appropriate antibody were incubated with rocking overnight at 4°C. Immuno-chromatin complexes were isolated by addition of 50 l of 1:1 slurry Protein G-Sepharose (Amersham Biosciences) in FA buffer (no SDS) for 1-2 h at room temperature. Sepharose beads were washed as described, except FA buffer for the first two washes was 300 mM NaCl, no SDS, and FA buffer for the third wash was 500 mM NaCl. Chromatin was eluted in a single step elution by incubating beads with 50 l of elution buffer (59) for 20 min at 65°C. PCR conditions were as follows: 5 l of diluted chromatin containing either 1% or 0.5% eluted material (IP) or 0.05% or 0.025% input chromatin (Input) was added to 10 l of PCR mixture and amplified using standard conditions (94°C 3 min followed by 26 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 30 s, followed by 72°C for 5 min). Final concentrations of PCR reagents were: 200 M dNTPs (Amersham Biosciences), 1ϫ TaqPCR buffer (Invitrogen), 1.5 mM MgCl 2 , 1% Me 2 SO, 0.067 unit/l TaqDNA polymerase (Invitrogen), 0.033 Ci/l [ 32 P]dATP. Reactions were separated via 6% 29:1 acrylamide:bis-acrylamide gels in 0.5ϫ TBE (45 mM Tris-borate, 1 mM EDTA) (50).
Northern Blotting, RNA Isolation, and RT-PCR-RNA was isolated as described (60). 20 g of total RNA/lane was separated via formaldehyde/MOPS gels as previously described (61). For analysis of GAL10 and ACT1 transcripts, DNA fragments covering the GAL10 or ACT1 ORFs were amplified by PCR and radiolabeled via random priming with the Klenow fragment of Escherichia coli DNA polymerase, as described previously (51). For analysis of GAL10-GAL7 fusion transcripts, a fragment of the GAL10-GAL7 intergenic region covering ϩ137 to through ϩ443 3Ј from the GAL10 stop codon was amplified by PCR and cloned into pSP64 (Stratagene), creating pCK195. pCK195 was linearized with BamHI (Roche Applied Science) and was used as a template for in vitro transcription by SP6 RNA polymerase (Roche Applied Science) as directed by the manufacturer. This GAL10-GAL7 intergenic antisense RNA probe was radiolabeled during in vitro transcription and used in Northern analysis. Total RNA for RT-PCR was isolated as above and treated with RNase-free DNase to remove residual genomic DNA. Oligonucleotide primers for RT-PCR were kind gifts of Dr. Jerry Kaplan (University of Utah) and fit criteria for efficiency and primer-dimer formation specified in Ref. 62. For RT-PCR using oligo(dT)-primed cDNA, 5 g of total RNA per reaction isolated as above were added to Ready-To-Go You-Prime First-Strand Beads (Amersham Biosciences), and cDNA was synthesized using oligonucleotide CKO99, according to the manufacturer's instructions. Sequences of oligonucleotides are available upon request.

Mutation of SPT6 Causes Reduced
Read-through Transcription Past GAL10 -Previous studies have shown that yeast mutants believed to have transcription elongation defects display delayed and decreased induction of highly transcribed genes, including the GAL1, -7, and -10 genes (63-66). Therefore, we examined the requirement for Spt6 in transcription of GAL10. In these experiments, we used the mutation spt6 -1004, which contains an internal deletion of the sequence encoding the helix-hairpin-helix motif of Spt6 (under "Experimental Procedures") (67)). Strains containing this mutation are temperature-sensitive for growth at 37°C and have a strong Spt Ϫ phenotype at the permissive temperature (30°C). 2 Northern analysis ( Fig. 1) showed that, compared with a wild-type strain, the spt6 -1004 mutant has slower induction kinetics for GAL10 mRNA and modestly lower steady-state levels of GAL10 (ϳ40 -50% of wild type) (Fig. 1) when cells are grown at permissive temperature (30°C). Thus, an spt6 mutation causes defects similar to what has been previously observed for other putative elongation mutants (63)(64)(65)68).
We next asked if spt6 -1004 affected transcription readthrough of the normal GAL10 polyadenylation site. The GAL10 and GAL7 genes are adjacent and in the same transcriptional orientation (diagrammed in Fig. 2). Occasionally, transcripts that originate from the GAL10 promoter read-through GAL10 polyadenylation signals and end at the normal GAL7 3Ј-end, creating a large GAL10-GAL7 fusion transcript (55,69). Our results show that the spt6 -1004 mutant produces lower levels of the GAL10-GAL7 fusion transcript (Fig. 1, lower panel). The decrease in the GAL10-GAL7 fusion transcript is greater than the defect in the normal GAL10 transcript (quantified in Fig. 3, A and D, below). Thus the decrease in the fusion transcript in the spt6 -1004 mutant is not likely due to decreased initiation from the GAL10 promoter. These experiments suggest that read-through transcription may be specifically defective in spt6 -1004.
Spt6 and Ctr9 Association with GAL10-GAL7 Correlates with pol II Association-To characterize the role of Spt6 in GAL10-GAL7 transcription by a different approach, we measured the level of Spt6 and pol II across the GAL10-GAL7 locus in wild-type and spt6 -1004 strains using chromatin immunoprecipitation (ChIP) (Fig. 2). In addition, we measured the level of Ctr9, a member of the Paf1 complex, believed to play a role in transcription elongation (14,16,25,26,35,70). Paf1 and Spt6 may be functionally related as mutations in genes encoding Paf1 members have genetic interactions with spt6 mutations (Table II), and both are present in pol II complexes purified via association with Spt5 (15). For both wild-type and spt6 -1004 strains, measurements for pol II (using the Rpb3 subunit), Spt6, and Ctr9 association with GAL10-GAL7 were performed in the same strain by using different epitope tags ("Experimental Procedures"). The levels of the tagged proteins in SPT6 and spt6 -1004 strains are shown in Fig. 2A. Spt6 and Ctr9 have been previously shown to associate with the open reading frames of the S. cerevisiae ADH1 and PMA1 genes (16,29).
Our results show that both Spt6 and Ctr9 are cotranscriptionally recruited to GAL10-GAL7 (Fig. 2, B and C). In wildtype cells grown under conditions in which GAL10 and GAL7 are not transcribed (2% raffinose as the carbon source), there is little or no detectable Rpb3, Spt6, or Ctr9, associated with these genes. In contrast, when GAL10 and GAL7 are highly induced by galactose, Rpb3, Spt6, and Ctr9 are physically associated across GAL10 and GAL7. Rpb3 was detected at all positions tested, although at significantly lower levels over the GAL10 UAS. When the Spt6 and Ctr9 ChIP levels are com-pared with the Rpb3 levels over the GAL10-GAL7 region (Fig.  2C), the levels correlate over most of the region, indicating that the levels of Spt6 and Ctr9 over GAL10 and GAL7 may be determined by pol II levels. However, one exception to this correlation is the decreased level of Ctr9 over the GAL10-GAL7 intergenic region. From these results, we conclude that Spt6 and Ctr9 are physically associated with GAL10-GAL7 in a cotranscriptional fashion, although Ctr9 does not strongly associate with the GAL10-GAL7 intergenic region. While this work was in progress, another study also provided evidence for reduced association of the Paf1 complex over 3Ј noncoding regions (29).
ChIP analyses of the mutant grown at 30°C show both similarities and differences with the wild type strain (Fig. 2, B-D). The ChIP signal of Rpb3 associated over the GAL10-GAL7 locus is similar in the spt6 -1004 mutant and in wild type cells. This result is in contrast to Northern analysis that indicates that the GAL10 mRNA level is reduced in spt6 -1004 (Fig.  1). In spt6 -1004 cells, the levels of association of Ctr9 and the Spt6 -1004 mutant protein over GAL10-GAL7 are significantly reduced. This decrease in ChIP is explained by the apparent reduction in levels of Spt6 and Ctr9 proteins in the spt6 -1004 mutant ( Fig. 2A). We note, however, that in some experiments with the spt6 -1004 strain, the levels of Ctr9, and to a lesser extent, Spt6 -1004, were not significantly reduced ( Fig. 2A). In all experiments in spt6 -1004 mutants, however, Ctr9 and Spt6 -1004 associations with GAL10-GAL7 were reduced (Fig.  2, C). 3 These results suggest that Spt6 may play a role in either the recruitment and/or stability of Ctr9. The decreased level of Ctr9 association with GAL10-GAL7 in spt6 -1004 mutants suggested that spt6 -1004 strains might also be defective in Paf1 complex-dependent histone modification. We examined methylation of H3 lysine 4 (Lys 4 ) in spt6 -1004 strains and saw no affect on global levels H3 Lys 4 tri-or di-methylation (data not shown).
Positive Roles for Spt6, Spt4, and Rpb2 and a Negative Role for TFIIS in mRNA 3Ј-End Formation at GAL10 -The reduced level of the GAL10-GAL7 fusion transcript in the spt6 -1004 mutant ( Fig. 1) suggests that a normal function of Spt6 is to regulate elongation past polyadenylation signals. To measure the level of the GAL10-GAL7 fusion transcript without interference from the stronger GAL10 mRNA signal, a probe specific for the GAL10-GAL7 intergenic region was used that would detect longer GAL10 transcripts without detecting mature GAL10 mRNA ("Experimental Procedures"). The advantage of this probe is shown by comparing Fig. 3A, lanes 1, 15, and 16, which illustrate that the high levels of GAL10 almost obscure the GAL10 -7 read-through products if a GAL10 probe is used to detect them.
To determine if the read-through defect observed for spt6 -1004 is either allele-specific or gene-specific, the level of the GAL10-GAL7 fusion RNA was measured in three additional spt6 mutants and in several other elongation mutants. Our results (Fig. 3A) show that GAL10-GAL7 fusion transcript levels were reduced in three of the four spt6 mutants tested and in spt4 -289, a presumed null allele of SPT4 (71). This analysis also revealed that the read-through defect applies to other species of GAL10 transcripts that are processed 3Ј of the major GAL10 cleavage site. Interestingly, the spt6 -1004 defect for production of longer GAL10 transcripts, including the GAL10-GAL7 fusion transcript, is suppressed by a dst1⌬ mutation (Fig. 3A, compare lanes 5 and 6). DST1 encodes the S. cerevisiae homologue of the elongation factor TFIIS. A dst1⌬ mutation causes few phenotypes on its own, but it has been shown to cause more severe mutant phenotypes when combined with mutations in several genes that encode elongation factors (14,20,72,73), including spt6 -1004 and other spt6 alleles. 2 Therefore, the suppression of the spt6 -1004 defect in production of longer GAL10 transcripts by dst1⌬ relative to total GAL10 transcription in the double mutant stands in contrast with the enhancement of spt6 phenotypes by dst1⌬ (Ref. 20 and data not shown). These results implicate Spt6, Spt4, and TFIIS in playing roles in the processing decision at the major GAL10 polyadenylation site.
The correlation of changes in GAL10-GAL7 fusion transcript levels with perturbation of transcription elongation was extended by examination of pol II mutants (Fig. 3B). Three pol II mutants, one in the gene encoding the largest subunit RP021 (here called RPB1 to decrease confusion with alleles of RPO21 named as rpb1 alleles), and two in the gene encoding the second largest subunit, RPB2, were examined. Each mutant has phenotypes indicative of elongation defects. The rpb1-221 and rpb2-10 mutations show allele-specific interactions with spt5 mutations (20). Additionally, rpb2-10 has an elongation defect in vitro, and in vivo it is moderately sensitive to 6-azauracil (6-AU), an inhibitor of GTP synthesis that is thought to increase pausing by pol II (74 -76). Frequently, elongation factor mutants are hypersensitive to 6-AU, presumably because they are less tolerant to 6-AU-induced pausing. An rpb2-7 mutant is strongly sensitive to 6-AU but shows phenotypes distinct from rpb1-221 and rpb2-10 (74,76,77). Fig. 3B shows that, in addition to spt6 and spt4 mutants, the rpb2-10 mutant also exhibits defects in GAL10-GAL7 fusion production. No defects were observed for rpb1-221 or rpb2-7. This defect of rpb2-10 supports the hypothesis that processing at GAL10 is sensitive to defects in transcription elongation. These results, then, provide support for an in vivo elongation defect for rpb2-10, which increases pausing by pol II in vitro. Our results for spt4, rpb2-10, and rpb2-7 are similar to a recent report from the Denis laboratory (77).
Interestingly, although the most 5Ј GAL10 processing site is favored in spt6 -1004, spt4 -289, and rpb2-10 mutants, there are no apparent effects when levels of the GAL10-GAL7 fusion transcript are compared with levels of the minor, longer GAL10 transcripts (Fig. 3C). These results suggest that the altered processing efficiency in these mutants is specific for the most 5Ј GAL10 processing site, because there is no alteration in processing efficiency of the GAL10 minor sites (just 3Ј of the major site). These results cannot be explained by a simple model suggesting that elongation mutants promote upstream processing events simply by pausing more and thus allowing more time for upstream processing events to occur. If this were the case, one might expect elongation mutants to show increased defects in read-through to further downstream polyadenylation sites. At GAL10, elongation mutants only appear to affect the choice between the major upstream processing site and the minor GAL10 processing site, a few hundred bases downstream, but not between the two downstream polyadenylation signals, which are separated by over a kilobase.
We utilized two alternate methods to examine transcription at the 3Ј-end of GAL10 (Fig. 3D). We examined polyadenylated transcripts in the GAL10 -7 intergenic region by RT-PCR using oligo(dT)-primed cDNA, and we examined all transcripts using region-specific cDNA generated by a thermostable DNA polymerase. The results for RT-PCR of spt6 -1004, rpb2-10, rpb2-7, and rpb1-221 are similar to those from the Northern analysis (Fig. 3, A and B). Taken together, these results indicate that the effects observed for specific transcripts in the GAL10 -7 intergenic region are applicable to all detectable transcripts in the GAL10 -7 region.
Mutation of the GAL10 Polyadenylation Signal Increases Spt6 and Ctr9 Association Downstream of GAL10 -The physical association of Ctr9 with transcribed regions requires transcription and wild type Spt6 function or levels (Fig. 2) Additionally, Ctr9 association with chromatin appears specifically decreased in the GAL10-GAL7 intergenic region, suggesting the possibility that the GAL10 polyadenylation signal or cleavage of the GAL10 transcript may initiate a reorganization of elongation factors associated with pol II, including release of Ctr9. To examine this possibility, we constructed a strain with a deletion of the major GAL10 polyadenylation signal so that we might examine the role of mRNA 3Ј-end formation in Ctr9 and Spt6 association with chromatin. This deletion, gal10⌬56, is a 55-bp deletion altering function of the major GAL10 polyadenylation signal that causes a Gal Ϫ phenotype and increases read-through of the GAL10 polyadenylation signal to about 10 -20% of total GAL10 transcription (data not shown). Construction of this deletion was based on previous studies of GAL10 3Ј-end formation (55). Although efficiency of mRNA 3Ј-end formation at the major GAL10 polyadenylation signal is decreased and production of GAL10-GAL7 read-through transcript is greatly increased, spt6 -1004 still shows enhancement of GAL10 processing in gal10⌬56 (data not shown). In addition to the gal10⌬56 mutation, we included a deletion of the GAL7 TATA element (gal7⌬6), to allow easier detection of GAL10 the GAL10 coding region probe, illustrating the level of the GAL10 -7 read-through product relative to the level of the GAL10 transcript. B, mutations in RPB1 and RPB2 have different effects on production of GAL10-GAL7 transcripts relative to GAL10. Data from at least three independent experiments are shown in the graph and represented as described above. rpb2-10 is defective for production of the GAL10-GAL7 RNA, rpb2-7 appears to have higher read-through than wild type (compare lanes 3 and 1, and lanes 5 and 4), and rpb1-221 has no effect (compare lane 2 to lane 1). The difference in GAL10 RNA levels between lanes 1-3 and lanes 4 and 5 indicates only different exposure levels of independent Northern blots. Lane 1-3 are from one experiment, whereas lanes 4 and 5 are from another. Each mutant sample is normalized to wild type samples from the same experiment. C, graph of ratio GAL10-GAL7 fusion transcript relative to GAL10 longer, minor transcript. Data show the average of at least three experiments Ϯ S.D. D, RT-PCR of elongation mutants confirming results of A. RT-PCR was performed on total RNA using methodology of (62) (left panel) or on oligo(dT)-primed cDNA (right panel) for regions shown in the schematic. Region 1 is within the GAL10 ORF at the 5Ј-end, region 2 spans the major GAL10 cleavage site (shown as asterisk), and region 3 is at the extreme 3Ј-end of the GAL10-GAL7 intergenic region and does not amplify GAL7 cDNA. Values are the average of a minimum of three independent experiments Ϯ 95% confidence interval of the mean. transcription events by reducing the levels of pol II and associated factors recruited to the GAL7 ORF by the GAL7 UAS.
We performed ChIP analysis for Rpb3, Spt6, and Ctr9 in these modified GAL10-GAL7 strains, and the results are shown in Fig. 4. Fig. 4A shows the Rpb3 ChIP in gal7⌬6 cells, and Fig.  4B shows the Rpb3 ChIP in gal7⌬6 gal10⌬56 cells. These results show that spt6 -1004 does not strongly affect pol II density over GAL10-GAL7, with the exception of possibly de-creasing pol II density in the GAL10-GAL7 intergenic region, consistent with an enhancement of processing at the major GAL10 polyadenylation site seen in spt6 -1004 (Fig. 4, A and  B). As shown in Fig. 4C, Ctr9 association was reduced downstream of GAL10, whereas Spt6 association downstream of GAL10 did not decrease as quickly as Ctr9. If Ctr9 and Spt6 chromatin association are normalized to Rpb3 association over the GAL10-GAL7 region, it is clear that Ctr9 association with GAL10 decreases 3Ј of the GAL10 ORF while Spt6 association is maintained (Fig. 4, C and D). Fig. 4D specifically compares ChIP of Spt6 and Ctr9 in a gal7⌬6 strains to that in a gal10⌬56 gal7⌬6 strain, where GAL10 3Ј-end formation has been compromised. ChIP of Spt6 increases over the GAL10 3Ј-end and GAL10-GAL7 intergenic regions when GAL10 mRNA 3Ј-end formation is compromised. ChIP of Ctr9 also increases over this region, although the increased level appears to extend further than observed for Spt6. These data suggest that Ctr9 is preferentially lost from elongation complexes downstream of a functional polyadenylation signal.
Genetic Interactions between spt6 -1004, Paf1 Complex Genes, and pol II Mutants-Given the molecular relationships between Spt6, pol II, and Ctr9, we tested for genetic interactions by double mutant analysis. First, we examined genetic interactions between spt6 -1004 and pol II mutants. For example, does spt6 -1004 interact more strongly with rpb2-10 because both show apparent enhanced processing at GAL10? What we observed was that spt6 -1004 causes a strong double mutant phenotype with both rpb2-10 and rpb1-221 with respect to growth on three carbon sources, glucose, raffinose, and galactose (Fig. 5A). rpb1-221 and rpb2-10 do share some phenotypes in that they both suppress specific spt5 alleles, suggesting that this phenotype could be a predictor of genetic interaction between spt6 -1004 and these two pol II alleles, and not alteration in processing efficiency. Second, we tested genetic interactions between three different spt6 alleles and the Paf1 complex null mutations paf1⌬, ctr9⌬, and cdc73⌬. In each case, the spt6 mutations caused double mutant lethality when combined with the Paf1 complex deletions (Table II). This suggests that the reduced level and reduced chromatin association of Ctr9 that was observed in an spt6 -1004 mutant did not completely abolish Ctr9 function. In addition, this result is consistent with recent results from the Jaehning laboratory (78) that show loss of Paf1 complex chromatin association in a cdc73⌬ strain, a mutant that shows much milder phenotypes than deletions of core members of the complex, Paf1 or Ctr9. This result was interpreted as evidence for the Paf1 complex having additional functions beyond those associated with recruitment to chromatin. Thus, it is proposed that spt6 mutants are enhanced by defects in these other putative Paf1 complex functions. Overall, these genetic interactions provide evidence that Spt6 and Paf1 contribute to a common essential process, presumably pol II elongation.
Functional Consequences of Elongation Factor Mutation on GAL10-GAL7 Interference-Proudfoot and coworkers (55,79) have demonstrated that transcription from GAL10 interferes with transcription from the downstream GAL7 gene, especially when GAL10 3Ј-end formation is compromised. Defects in GAL7 transcription lead to galactose toxicity (Gal-sensitive (Gal s ) phenotype) under permissive conditions due to the inability to metabolize galactose 1-phosphate, a toxic intermediate in the Gal pathway (55,80). This phenotype is observed in the gal10⌬56 mutation (Fig. 5B, compare growth on raffinose to galactose/raffinose). To gain additional evidence that spt6 -1004 impairs read-through transcription past GAL10, we tested whether spt6 -1004 suppresses the Gal s phenotype caused by gal10⌬56. Our results (Fig. 5B) show that, indeed, spt6 -1004 does suppress gal10⌬56. These results provide independent evidence for enhanced 3Ј-end formation caused by spt6 -1004. In addition, rpb2-10 as expected, suppresses gal10⌬56 (Fig. 5C). Surprisingly, rpb2-7 weakly suppresses gal10⌬56 as well, suggesting that there may be additional ways to overcome the Gal s phenotype induced by defective GAL10 processing.

DISCUSSION
In this work we have provided evidence that the transcription elongation factors Spt6 and Spt4 promote transcription elongation past the GAL10 3Ј-end. In these studies, we have shown that spt6 and spt4 mutations, as well as an allele of RPB2, rpb2-10, impair transcription read-through at GAL10. Importantly, spt6 -1004 and rpb2-10 also suppress transcriptional interference at GAL7 when GAL10 mRNA 3Ј-end formation is compromised, suggesting that the effect of these mutations on the levels of the longer GAL10 mRNAs is via reduced read-through and not indirect, via effects on mRNA stability. Furthermore, we present evidence that Spt6 is required for the association of Ctr9 with chromatin and that Ctr9 is selectively lost from elongation complexes downstream of the GAL10 major polyadenylation signal. Our results with respect to Ctr9 are similar to recent studies presented while this work was in preparation (29).
A recent study showed that elongation factor defects can strains lacking gal10⌬56 are shown as controls. Spots are 10-fold serial dilutions of each strain. Growth on glucose and raffinose plates is shown as a control for general growth defects. Growth on galactose indicates suppression of gal10⌬56 Gal Ϫ phenotype. Growth on raffinose/galactose indicates suppression of gal10⌬56 Gal s phenotype, suggesting a decrease in interference at GAL7. Glucose plate is 3 days of growth. All other plates are 5 days of growth. C, suppression of gal10⌬56 phenotypes in rpb2-7 and rpb2-10 strains. Experiment is as in B except galactose and raffinose/galactose plates are 7 days of growth. enhance the use of internal polyadenylation sites and cause a preference for the most upstream polyadenylation site for genes that have multiple 3Ј processing sites (77). Our results, coupled with those of the Denis laboratory (77), suggest that the phenomenon we observe at GAL10 may be applicable to other genes as well. Those results led to the model that increased pausing of pol II in elongation mutants leads to increased processing at any given polyadenylation signal, leading to an upstream bias in polyadenylation site usage. Although this model is generally supported by our findings, it does not entirely explain our results at GAL10. At GAL10, the upstream polyadenylation site is favored in elongation mutants, but it is favored equally over both downstream sites, suggesting that there is no effect on 3Ј-end formation choice between the two downstream elements. A simple pol II-pausing model predicts a gradient of decreasing usage of polyadenylation sites from 5Ј to 3Ј, causing greater defects in read-through to subsequent polyadenylation sites. The choice of 3Ј-processing sites may involve multiple factors, including the position of the site, the strength of the polyadenylation signal, and an interaction between pol II and particular DNA sequence elements that can affect the elongation rate or the degree of pausing by pol II. The role that we have observed for Spt6, Spt4, and pol II in readthrough at GAL10 may reflect an important role for these factors in promoting transcription read-through of cryptic or weak polyadenylation signals within coding regions, thereby promoting pol II processivity.
The composition of the pol II elongation complex appears to be dynamic and is likely influenced by both DNA and RNA signals. Here we provide evidence that Spt6 affects association of Ctr9 with GAL10 either by affecting Ctr9 recruitment directly or by regulating Ctr9 levels. The entire Paf1 complex is most likely not recruited, because Ctr9 has been shown to be integral for wild type levels of other members of the complex (78). Previous studies have shown that some mRNA-processing factors are recruited by the pol II large subunit C-terminal domain phosphorylated on serine 2 of the repeat and influenced by a polyadenylation signal (Refs. 41, 42, and 81-84 and references therein). Our results and those of others (29) have now also shown the loss of certain factors such as Ctr9. The functional consequences of Ctr9 loss downstream of the GAL10 polyadenylation signal are unclear, because we did not observe changes in GAL10 read-through products in Paf1 complex mutants. However, Jaehning and coworkers have identified alterations in mRNA polyadenylation in Paf1 complex mutants, suggesting some function for the Paf1 complex at the 3Ј-end of genes (78).
The interplay between transcription elongation and termination is likely quite complex. Understanding better the interactions between factors that promote transcription elongation, such as Spt6 and Spt4, and polyadenylation signals that effectively inhibit elongation by promoting transcript cleavage, most likely will require a dissection of the termination mechanism itself, which remains elusive.