Functions for S. cerevisiae Swd2p in 3′ end formation of specific mRNAs and snoRNAs and global histone 3 lysine 4 methylation

INTRODUCTION

Eukaryotic genomes encode different classes of genes, which are transcribed by RNA polymerase II (RNAP II). The major class of protein encoding genes serves as templates for the synthesis of messenger RNA precursors (pre-mRNAs). A second class of RNAP II transcripts are derived from small nucleolar RNA (snoRNA) genes and function as components of ribonucleoprotein particles in pre-rRNA processing and ribosomal RNA modification (Kiss 2002). All known primary transcripts of RNAP II require processing to be biologically functional. For pre-mRNAs this includes capping, splicing, and 3′ end formation. snoRNA transcripts generally have to be processed by exonucleolytic trimming to generate the mature RNA.

Pre-mRNA 3′ end formation requires site-specific endonucleolytic cleavage of the RNA, followed by polyadenylation of the upstream cleavage product (for review, see Zhao et al. 1999; Edmonds 2002). In Saccharomyces cerevisiae, biochemical fractionation of whole-cell extracts initially defined the cleavage factors IA and IB (CF IA and CF IB) and the cleavage factor II (CF II) as sufficient to catalyze the cleavage step (Chen and Moore 1992); polyadenylation was suggested to require CF IA, polyadenylation factor I (PF I), and poly(A) polymerase (Pap1p). Subsequently, it was found that CF II, PF I, and Pap1p are associated in vivo in the so called cleavage and polyadenylation factor CPF (Ohnacker et al. 2000). The polypeptide composition of these factors is complex (see Fig. 4D below). CF IA is a heterotetrameric protein (Kessler et al. 1996; Minvielle-Sebastia et al. 1997) and CPF consists of at least 15 different polypeptides (Dichtl et al. 2002b; Gavin et al. 2002; Dheur et al. 2003; He et al. 2003; Nedea et al. 2003). Seven CPF subunits (Pta1p, Ref2p, Pti1p, Swd2p, Ssu72p, Syc1p, and Glc7) have been suggested to form the APT subcomplex (associated with Pta1p) that together with core-CPF (previously called PF I) forms holo-CPF (Nedea et al. 2003).

Maturation of independently transcribed snoRNAs requires the production of an entry site for 3′ → 5′ exonucleases that eventually form the correct 3′ end. In one group of snoRNAs this entry site is produced by Rnt1p (yeast RNase III), which recognizes specific RNA hairpin structures (Chanfreau et al. 1998). A second group of snoRNAs, which lacks Rnt1p recognition sites, requires endonucleolytic cleavage by the 3′ end formation machinery uncoupled from the polyadenylation step (referred to as polyadenylation-independent 3′ end formation; Fatica et al. 2000). The latter genes carry specific cis-acting elements in proximity of the cleavage site (Morlando et al. 2002; Steinmetz and Brow 2003) that differ from typical polyadenylation signals found in mRNAs (Graber et al. 1999; van Helden et al. 2000). Cleavage uncoupled from polyadenylation requires components of CF IA (Fatica et al. 2000; Morlando et al. 2002) and certain subunits of CPF, which may all be contained within the APT complex (Dheur et al. 2003; Ganem et al. 2003; Nedea et al. 2003; Steinmetz and Brow 2003). Strikingly, mutations in core-CPF subunits have been found not to affect polyadenylation-independent 3′ end formation (Morlando et al. 2002). In addition, the CTD interacting protein Nrd1p and the RNA binding protein Nab3p have been shown to be involved in this process (Steinmetz et al. 2001).

Progress in the fields of RNAP II transcription and RNA processing revealed that these processes are intimately connected (for recent reviews, see Bentley 2002; Neugebauer 2002; Proudfoot et al. 2002). In vivo 3′ end formation is coupled to transcription by RNAP II (Proudfoot et al. 2002). Like other pre-mRNA processing factors, the 3′ end formation machinery is thought to associate with the carboxy terminal domain (CTD) of the largest subunit of RNAP II during transcription elongation (Proudfoot and O’Sullivan 2002). Site-specific phosphorylation and dephosphorylation of the CTD controls binding of CTD interacting proteins and integrates pre-mRNA processing with the transcription cycle. Transcription termination depends on functional polyadenylation signals and the activity of certain pre-mRNA processing factors (Proudfoot and O’Sullivan 2002). Several studies indicate that 3′ end formation factors can act during termination independent of their roles in cleavage and polyadenylation (Osheim et al. 1999; Tran et al. 2001; Orozco et al. 2002; Osheim et al. 2002; Kim and Martinson 2003; Sadowski et al. 2003).

Posttranslational modifications of histone tails play important roles in the regulation of gene expression (Jenuwein and Allis 2001). Modifications include phosphorylation, acetylation, ubiquitination, and methylation and their combinatorial occurrence was suggested to translate into a histone code (Jenuwein and Allis 2001; Turner 2002). The S. cerevisiae genome encodes six SET domain proteins (for review, see Kouzarides 2002). Of those, Set1p is associated with seven additional proteins (Bre2p, Swd1p, Swd2p, Swd3p, Sdc1p, Spp1p, Shg1p) in SET1C and methylates H3K4 (Briggs et al. 2001; Miller et al. 2001; Roguev et al. 2001; Krogan et al. 2002; Nagy et al. 2002). H3K4 tri-methylation is associated with actively transcribed genes (Santos-Rosa et al. 2002) and was suggested to act as cellular memory for recent gene expression (Krogan et al. 2003; Ng et al. 2003).

Here, we analyzed S. cerevisiae Swd2p that is physically associated with both CPF and SET1C. We provide evidence that Swd2p is required for 3′ end formation of specific mRNAs and snoRNAs. Furthermore, the protein is necessary for SET1C methyltransferase activity on H3K4.

RESULTS

Swd2p carries seven WD-40 repeat motifs and is conserved within eukaryotes

Proteomic analysis of polypeptides associated with CPF and SET1 revealed Swd2p as a common component of both complexes (Miller et al. 2001; Roguev et al. 2001; Dichtl et al. 2002b; Nagy et al. 2002; He et al. 2003). We searched databases and identified Swd2p homologs in a large number of eukaryotes (Fig. 1A; data not shown; see Materials and Methods). Schizzosaccharomyces pombe, Arabidopsis thaliana, Caenorhabditis elegans, and Drosophila melanogaster have two Swd2p homologs each whereas most other species have only one. It should be noted, however, that the Swd2 family is not sharply delineated from the larger superfamily of WD-40 proteins and we cannot rule out that more distantly related proteins also belong to the Swd2 family. Standard protein motif prediction tools (SMART, PFAM; see Materials and Methods) detected up to three WD-40 repeat sequences in Swd2p (repeats 3, 5, and 6 in Fig. 1A). WD-40 repeat proteins form a large protein family with diverse biological functions (Smith et al. 1999). The majority of these proteins form seven-bladed β-propeller-like structures although structures with four, five, and six blades have also been described. Because many WD-40 repeats are poorly predicted with the Pfam and SMART tools, we subjected the Swd2 family to sensitive profile–profile dot plots (Thompson et al. 1994). As shown in Figure 1B, there are six distinct tiers of off-diagonal signals, strongly suggesting that the Swd2 family has a seven-bladed β-propeller structure.

Swd2p is required for 3′ end formation of specific mRNAs and snoRNAs

To functionally analyze SWD2 we generated temperature-sensitive alleles (see Materials and Methods). Subunits of yeast CPF have been implied in transcription termination at protein coding genes and snoRNA genes. The association of Swd2p with CPF suggested that it might also function in transcription termination. To test this, we analyzed steady-state levels of several snoRNAs and mRNAs by Northern blotting of total RNA extracted from wild-type, swd2-2 and swd2-6 strains grown at 23°C and after shift to 37°C (Fig. 2). A ssu72-2 mutant strain was analyzed in parallel. Figure 2A shows that swd2-2 and swd2-6 mutants accumulated an extended snR33 transcript following growth at 37°C. A RNA of the same length was observed in the ssu72-2 mutant, mostly at restrictive temperature. This suggested that Swd2p was required to prevent transcriptional read-through at the snR33 terminator. To prove this, we probed for the product of the YCR015c gene that lies immediately downstream of the snR33 gene. swd2 mutants strongly accumulated a snR33-YCR015c mRNA at 37°C. This mRNA was also readily detected in the ssu72-2 strain. Strikingly, levels of these transcripts were strongly increased in both swd2 and ssu72 mutants, because the endogenous YCR015c mRNA was barely detectable under these conditions in the wild type. Furthermore, we observed accumulation of an additional RNA that was previously shown to occur in ssu72-69 cells from read-through at the YCR015c terminator and to correspond to a snR33-YCR015c-POL4 transcript (Ganem et al. 2003). This RNA accumulated in swd2 mutants and ssu72-2 cells at 37°C. We could not detect read-through RNAs with oligonucleotide probes directed against snR13 and snR71 snoRNAs (Fig. 2B,C); however, read-through products containing snR13 and snR71 were prominent in the ssu72-2 mutant. In contrast, the use of probes against TRS31 and REC104 mRNAs that are transcribed immediately downstream of the snR13 and snR71 genes, respectively, revealed transcriptional read-through also in the swd2 mutant strains. Strikingly, levels of these transcripts were strongly increased in the ssu72-2 mutant but not in the swd2 mutants. We therefore conclude that the failure to detect snR13 and snR71 read-through RNAs with the use of oligonucleotide probes was due to the lower abundance of these transcripts compared to the ssu72-2 mutant strain where these genes were either more heavily transcribed or more stable. Note that oligonucleotide probes used to detect snoRNAs carry only one radioactive label at the 5′ end compared to many radioactive labels in the random-primed probes that were used to detect mRNAs. We observed no read-through RNAs with the swd2 mutants when snR45 and the downstream ASN1 mRNA were tested, or when the NRD1 mRNA and the downstream MRPL17 mRNA were tested (Fig. 2D; data not shown). In contrast, the ssu72-2 mutant gave strong read-through at both the snR45 and the NRD1 genes; in many cases read-through in the ssu72-2 mutant resulted in the accumulation of intermediate-sized RNAs (indicated by asterisks) that may be caused by processing at cryptic cleavage and polyadenylation sites. Figure 2E shows that SWD2 mRNA levels were increased in swd2 mutants and extended transcripts appeared. These extended SWD2 transcripts were not observed in the ssu72-2 mutant. Other mRNAs tested (SET1, CYH2, ADH1, and ACT1; Fig. 2F; data not shown) did not reveal read-through transcripts in swd2 and ssu72-2 mutants; however, some reduction of these mRNAs was observed following growth at 37°C.

To test for transcription termination by transcriptional run-on analysis (TRO), we analyzed a plasmid-borne CYC1 gene that was under control of a minimal CUP1 promoter (pCUP-CYC1; Fig. 3A). Wild-type and mutant strains carrying pCUP-CYC1 were grown in glucose-based synthetic medium containing 1 mM CuSO₄. Northern analysis confirmed that expression of the pCUP1-CYC1 gene depended on the presence of CuSO₄ in the medium (data not shown). TRO was performed as described (Birse et al. 1998). We found that wild-type, swd2-2, and swd2-6 strains terminated transcription efficiently following growth at 25°C and following a shift to 37°C for 1 h. These results indicated that swd2 mutant strains were not deficient in termination of the pCUP-CYC1 gene.

As swd2 mutants displayed transcriptional read-through at certain genes, we asked whether the protein was required for 3′ end cleavage in vitro. Processing extracts were produced from wild-type and mutant strains (swd2-1, swd2-2, swd2-4, swd2-5, swd2-6, swd2-7, swd2-8) and tested for cleavage of a synthetic CYC1 pre-mRNA substrate. In vitro cleavage occurred efficiently in wild-type and all analyzed swd2 mutant extracts (Fig. 3B; data not shown). The efficiency of processing was not reduced when the extracts were preincubated at 37°C for 5 min and when the assay temperature was elevated from 30°C to 37°C. We conclude that defects in 3′ end formation of some mRNAs and snoRNAs in swd2 mutant strains did not result from a general loss of cleavage activity.

Taken together, the results in Figures 2 and 3 suggest that mutations in SWD2 result in defective 3′ end formation at some genes, whereas other genes are not or are at least less strongly affected. The reasons for these apparently gene-specific effects remain unclear, but may relate to differences in 3′ end formation signals (see Discussion).

Swd2p bridges the APT, core-CPF, and CF IA subcomplexes of the 3′ end formation machinery

WD-40 repeat proteins are frequently associated with mul-tiprotein complexes and are implied in mediating protein–protein contacts (Smith et al. 1999). To identify the interactions that are mediated by Swd2p we performed GST pull-down experiments with bacterially expressed GST-Swd2p and in vitro translated [³⁵S]-methionine-labeled 3′ end formation factors. Figure 4A, lanes 8, 9, 18, and 19, shows representative experiments in which GST-Swd2p bound efficiently to the core-CPF components Yhh1p/ Cft1p, Ysh1p/Brr5p, Pfs2p, and Pta1p. Other core-CPF components (Ydh1p/Cft2p, Mpe1p, Pap1p, Fip1p) did not bind. Of the APT proteins, Pta1p, Glc7p, and Pti1p showed an interaction with GST-Swd2p (Fig. 4B, lanes 29,39,40); Ref2p, Ssu72p, and Syc1p did not interact significantly. Finally, the CF IA subunit Pcf11p bound to GST-Swd2p; no binding was observed with other CF IA components (Rna15p, Rna14p, Clp1p; Fig. 4C, lanes 48–50), CF IB (Nab4p; Fig. 4C, lane 52), or the poly(A) binding protein Pab1p (Fig. 4A, lane 10). All observed interactions were specific for Swd2p because GST alone did not result in binding. Furthermore, because the binding reactions were done in the presence of RNase A, it is unlikely that RNA mediated the interactions. We propose that Swd2p contributes specific interactions between subunits of the APT, core-CPF, and CF IA subcomplexes of the 3′ end formation machinery (see Fig. 4D) and may therefore function in the assembly of 3′ end formation complexes (see Discussion).

Swd2p is required for SET1C activity

Because Swd2p is a component of SET1C (Miller et al. 2001; Roguev et al. 2001; Nagy et al. 2002) we asked whether it is also required for SET1C activity, that is, methylation of H3K4. Western analysis was performed with whole-cell extracts derived from wild-type and swd2 mutant strains grown at 25°C or following growth at 37°C for 2, 4, and 8 h. For control we also analyzed extracts from strains lacking the SET1 gene (ΔSET1), from H3K4A, H3K9A mutant strains and corresponding isogenic wild-type strains. Figure 5A shows a representative experiment. Antibodies directed against di- and tri-methylated H3K4 recognized proteins in wild-type extracts that were absent in ΔSET1 and H3K4A strains and therefore corresponded to the modified forms of H3K4. swd2-2 and swd2-3 extracts showed no or only trace amounts of di- and tri-methylation in multiple experiments; these deficiencies were observed at either permissive or restrictive temperatures. Some di-methylated and tri-methylated H3K4 was detected in swd2-6 extracts at permissive temperature, which was lost upon a shift to 37°C. Cross-reacting bands that were observed served as control for loading of comparable amounts of whole-cell extract in each lane (indicated by asterisks). These data suggested a requirement for Swd2p as component of SET1C in global methylation of H3K4.

In addition to methylation of H3K4, components of SET1C are required for the maintenance of correct telomere length, and this requirement is based at least partly on the histone methyltransferase activity of SET1C (Roguev et al. 2001). A Southern blot strategy was used to test whether mutations in SWD2 affected telomere length (Fig. 5B; Roguev et al. 2001). Wild-type and swd2 mutant strains were analyzed following growth at 25°C and after an 8-h shift to 37°C. We reproducibly found that the swd2-2 mutant had shortened telomeres compared to the wild type. The swd2-3 strain appeared to be affected most severely after growth at 37°C; the swd2-6 strain displayed a weaker reduction of telomere length at both temperatures. Taken together these results suggested a requirement for Swd2p for a functional SET1C complex.

To reveal the molecular basis of loss of SET1C function we analyzed the levels of mutant Swd2p proteins. Because no specific antibody against the protein was available we used strains that expressed wild-type and mutant proteins as N-terminal ProteinA fusions. We found that the levels of protA-swd2-2p, protA-swd2-3p and protA-swd2-6p proteins were stable at permissive and nonpermissive temperatures (Fig. 5C). In contrast, Set1p levels were strongly reduced in mutant extracts; Set1p levels were similarly reduced in nontagged swd2 mutant strains (data not shown). Because SET1 mRNA levels were stable in swd2 mutants at permissive temperature (see Fig. 2F) underaccumulation of Set1p was likely due to a posttranscriptional defect. We suggest that SET1C might be destabilized in swd2 strains causing a reduction of Set1p levels and consequently a defect in methylation of H3K4. In contrast to Set1p, none of the tested CPF subunits (Pta1p, Ysh1p, Glc7p, Ssu72p) was significantly reduced in the swd2 mutant strains (Figs. 5C, 6).

To test whether the transcriptional read-through observed with swd2 mutants (Fig. 2) related to its association with SET1C, we performed Northern analysis with RNAs extracted from various strains lacking SET1C components (ΔSET1, ΔBRE2, ΔSWD1, ΔSWD3, ΔSDC1, ΔSHG1) and the H3K4A mutant strain. For better comparison all strains were grown under the same conditions (at 30°C) in this experiment; note that this temperature was semipermissive for swd2 mutants. Under these conditions the swd2-2 strain displayed strong accumulation and the swd2-6 strain weak accumulation of extended YCR015C, TRS31, and SWD2 transcripts (Fig. 5D; data not shown). We did not observe extended transcripts for any strain that lacked a component of SET1C or with the H3K4A strain. Increased levels of YCR015C and SWD2-containing transcripts were observed exclusively for the swd2 mutant strains. We conclude that transcriptional read-through defects of swd2 mutants were independent of functional SET1C and H3K4.

Swd2p is independently associated with CPF and SET1C in extracts

The association of Swd2p with CPF and SET1C may indicate a physical link between both complexes. To test this we initially analyzed the size distribution of cellular Swd2p-containing complexes by gel filtration of wild-type and mutant extracts derived from strains expressing the ProteinA fusions. We used the core-CPF component Ysh1p and the Set1p methyltransferase as markers for the elution profile of both complexes (Fig. 6). SET1C displayed a higher molecular weight (peak at fraction 10) than CPF (peak at fraction 15), consistent with the previous prediction that SET1C may occur as an heterooligomeric factor in vivo (Roguev et al. 2003a). Small amounts of Ysh1p were also detected in high molecular weight fractions; however, no Set1p was detectable in CPF peak fractions. ProtA-Swd2p (from cells grown at 25°C) eluted in two major complexes, which overlapped with Set1p and with Ysh1p, consistent with its association with both complexes. Mutant ProtA-swd2-2p (from 25°C extracts) clearly shifted in its mobility to the lower molecular weight CPF complex. This is most likely due to the fact that underaccumulation of Set1p (see Fig. 5C) disrupted most of SET1C. Interestingly, the distribution of wild-type ProtA-Swd2p clearly shifted to the higher molecular weight SET1C following growth at 37°C whereas the distribution of Ysh1p was the same as in extract from cells grown at 25°C. The reasons for this observation remain unclear but may indicate a cellular response to heat treatment. Mutant extracts prepared after growth at 37°C displayed reduced ProtA-swd2-2p associated with SET1C while most of the protein co-eluted with Ysh1p. These observations are consistent with the interpretation that Swd2p was associated with two separable high molecular weight complexes in extracts and that mutant ProtA-swd2-2p failed to associate efficiently with SET1C but not with CPF. The possibility remained that subpopulations of SET1C and CPF were physically associated. To further test this we probed for the presence of Set1p in CPF fractions purified from ProtA-Ydh1p- and ProtA-Pfs2p-expressing strains by Western blot. However, we could not detect Set1p above background levels (data not shown), indicating that SET1C and CPF are not stably associated with each other in extracts under nonstringent salt conditions (150 mM KCl).

DISCUSSION

Intimate connections exist between the transcription apparatus and pre-mRNA processing factors. To further understand the complex interactions between these processes we analyzed the function of the Swd2p protein, which associates with the CPF and SET1C complexes that are involved in 3′ end formation and in transcription elongation, respectively. We found that Swd2p has essential functions as component of both complexes. It is necessary to prevent transcriptional read-through at specific genes and it is required for global H3K4 methylation.

Our analysis indicated that swd2 mutants are defective in 3′ end formation of specific mRNAs and snoRNAs. Recent results on other subunits of CPF revealed similarities with the phenotypes of swd2 mutants observed here. Swd2p was suggested to associate with the APT subcomplex of CPF that includes also Pta1p, Ref2p, Pti1p, Glc7p, Ssu72p, and Syc1p (Nedea et al. 2003). A common phenotype of mutants in Ref2p, Pti1p, Ssu72p, and Swd2p is transcriptional read-through at snoRNA genes (Dheur et al. 2003; Ganem et al. 2003; Nedea et al. 2003; Steinmetz and Brow 2003). These phenotypes strongly contrast results obtained with mutants in core-CPF subunits that did not affect snoRNA termination in a snR13-based reporter assay (Morlando et al. 2002). Moreover, transcription run-on analysis revealed deficient termination on a plasmid-borne CYC1 gene in yhh1/cft1 mutants (a component of core-CPF; Dichtl et al. 2002a) but not in ssu72-2 (Dichtl et al. 2002b; He et al. 2003) and swd2 mutant strains (Fig. 3A). These analyses point to distinct requirements for specific subunits of CPF for 3′ end formation at different sets of genes.

An important open question is whether distinct 3′ end formation complexes exist in vivo, which could function at specific genes. Termination depends on the cotranscrip-tional recognition of poly(A) signal sequences on pre-mRNAs. An important function of CPF lies in the recognition of poly(A) site sequences within pre-mRNAs and the RNA-binding proteins involved (Yhh1p/Cft1p, Ydh1p/Cft2p, and Yth1p) were so far exclusively assigned to core-CPF (Barabino et al. 2000; Dichtl and Keller 2001; Dichtl et al. 2002a). Poly(A) signals in pre-mRNAs and cleavage signals in snoRNAs are clearly distinct. The recognition of the respective sequences by RNA-binding components of CPF might therefore contribute to the gene-specific requirement for CPF subunits. Transcriptional read-through in swd2 and other APT mutants might reflect a deficiency in the recognition of termination signals or a defect in the transduction of the signals to RNAP II. Moreover, a deficiency in the recognition of termination signals or 3′ end formation signals, respectively, may result in failure to cleave specific transcripts at the 3′ end although the catalytic endonucleolytic activity per se may generally be unaffected. However, the phenotypes of mutants in APT subunits and core-CPF subunits cannot clearly be categorized as snoRNA and mRNA specific because APT mutants also affect termination at selected mRNA genes. Interestingly, the APT subunits Pti1p and Ref2p were suggested to uncouple cleavage from polyadenylation at 3′ ends of snoRNA genes (Dheur et al. 2003). It was proposed that the exclusion of poly(A) polymerase in Pti1p- and Ref2p-containing CPF complexes (within the so-called small nucleolar cleavage factor, snCF) may contribute to polyadenylation-independent 3′ end formation. It will be important for future analyses to test a representative set of genes (or the entire genome) for the potentially differential association of specific 3′ end formation factors, for example, APT proteins and poly(A) polymerase.

How does Swd2p function as a component of CPF and SET1C? Our bioinformatic analysis of Swd2p proteins suggests a seven-bladed β-propeller structure. Many members of this large protein family have been found to associate with multiprotein complexes, including numerous factors involved in pre-mRNA processing (Smith et al. 1999). Our protein interaction studies indicate that Swd2p can engage in multiple interactions with other 3′ end formation components. Interestingly, we identified interaction partners within the core-CPF, the APT, and the CF IA subcomplexes of the 3′ end formation machinery. Thus, it is tempting to speculate that Swd2p may contribute to the assembly of 3′ end formation complexes, which may be required primarily at selected genes. Alternatively, Swd2p may play solely a structural role within CPF. However, mutant Swd2p proteins did not appear to destabilize the complex significantly, as no reduction of any of the tested CPF components could be observed. In contrast, mutations within Swd2p resulted in strong underaccumulation of Set1p, most likely due to a posttranscriptional defect. This may reflect a crucial structural role for Swd2p in SET1C. Alternatively, SET1C assembly may be affected. We do not know if the same part of Swd2p is responsible for the interaction with the two complexes. Because no obvious common domains have been found in SET1C and CPF, it is possible that the Swd2p interaction depends on a short motif found in two different types of proteins in the two complexes.

Despite its bifunctional nature, the roles of Swd2p as a component of CPF and SET1C appear to occur independently of each other. This conclusion is based on three important observations: First, CPF and SET1C do not detectably interact in extracts; however, the possibility remains that a putative interaction of both complexes may be very labile or even transient. Second, phenotypes of mutations in Swd2p in 3′ end formation are duplicated by other mutations in CPF but not by mutations in SET1C or H3K4. Third, the bifunctional nature of Swd2p is not conserved in evolution, as two different Swd2p proteins associate with CPF and SET1C in S. pombe (Swd2.1 and Swd2.2, respectively; Roguev et al. 2003b). Thus, the functions that are played by a single Swd2p protein in S. cerevisiae have been separated in S. pombe. It will be very interesting to analyze the proteomic distribution of human Swd2p. Although there may be more than one Swd2p protein encoded in the human genome, this requires experimental verification. However, none of the known components of the mammalian 3′ end formation machinery displayed clear homology to S. cerevisiae Swd2p in our database searches. The association of Swd2p with both CPF and SET1C remains intriguing, as both of these complexes interact with elongating RNAP II. SET1C was suggested to associate with genes primarily at 5′ regions (Krogan et al. 2003; Ng et al. 2003), whereas CPF associates with genes increasingly toward their 3′ end (Komarnitsky et al. 2000; Licatalosi et al. 2002; Nedea et al. 2003). Moreover, recent reports suggested a requirement for SET1C in 3′ end formation/termination on certain genes that may depend on the interaction of the chromatin remodeling ATPase Isw1p with di- and tri-methylated H3K4 (Santos-Rosa et al. 2003). Isw1p itself was previously implied in RNAP II termination and may function to establish chromatin boundaries at 3′ ends of genes (Alen et al. 2002). However, Isw1p has crucial functions already at the beginning of genes, where it controls the association of the transcription machinery by nucleosome remodeling and influences the phosphorylation status of the RNAP II CTD during distinct phases of the transcription cycle (Morillon et al. 2003). The common denominator for CPF and SET1C function may therefore be the interaction with the elongating RNAP II complex, which ultimately controls termination of transcription and the formation of 3′ ends.

MATERIALS AND METHODS

Yeast growth, strain construction, and plasmids

Strains used in this study are listed in Table 1. SET1C deletion strains (ΔSET1, ΔBRE2, ΔSWD1, ΔSDC1, and ΔSHG1) were generously provided by A. Roguev and F. Stewart (Technical University of Dresden, Germany) and will be described elsewhere. Yeast were grown in rich YPD medium (2% glucose, 2% bactotryptone, 1% yeast extract) or synthetic drop-out medium (2% glucose, 0.67% yeast nitrogen base, 1× amino acids) as indicated in the figure legends. Temperature-sensitive SWD2 alleles were isolated by in vivo gap repair (Muhlrad et al. 1992). The SWD2 gene (nt -817–1932 relative to the A of the translational start codon, which corresponds to +1) was cloned as a SalI/BamHI fragment into pRS416 (pBD236). A 2.3-kb SalI/PstI fragment from pBD236 carrying SWD2 was cloned into pRS415ΔSpeI (generated by SpeI digest, Klenow treatment, and religation) to generate pBD238. The SWD2 gene was amplified by Taq Polymerase under error-prone conditions (Ohnacker et al. 2000) with primers swd2mut2 (TTG GTATGATACGCAAGGTGATG) and swd2mut3 (TGCTGTATT GAGATTGATCTCTTC). PCR products and linear SpeI-digested pBD238 (removing nt -100–1207 of SWD2 sequence) were co-transformed into YWK164. LEU⁺ transformants were first replica-plated on 5-FOA; after subsequent replica-plating on YPD growth was tested at 15°C, 23°C, 30°C, 34°C, and 37°C. Plasmid isolation from candidate mutant strains and reintroduction into a swd2 deletion background identified nine strains, which were nonviable at 37°C. Detailed analyses were performed with swd2-2, swd2-3, and swd2-6 strains. Amino acid changes in the mutant alleles were inferred by DNA sequencing of plasmids carrying mutant alleles. The following mutations were identified: swd2-2, F14P, C27G, F41I, K185E, S253L, C257R; swd2-3, G185E, S253E, C257R; swd2-6, L12P, K120E, F324E. Plasmids encoding proteinA-tagged fusions of wild-type and mutant SWD2 alleles were generated by PCR amplification of the respective open reading frame and cloning into pNOP1 (Senger et al. 1998) using primer-encoded NcoI and BamHI restriction sites: pBD239 (ProtA-SWD2-LEU2); pBD377 (ProtA-swd2-2-LEU2); pBD378 (ProtA-swd2-2-LEU2); pBD379 (ProtA-swd2-2-LEU2). Plasmids were transformed into strain YWK164, followed by plasmid shuffling on 5-FOA.

pCUP-CYC1 (pBD352) used for TRO was produced by replacing the EcoRI/BglII fragment containing the GAL1/10 promoter from pGAL-CYC1 (Birse et al. 1998) with a PCR fragment spanning bp −400 to 37 (relative to the A of the translational start codon) of the CUP1 promoter using primer-encoded EcoRI (CATAGGGAATTCTTAAAACACTTTTGTATTATTTTT) and BglII (TAAATGAGATCTGACCTTCATTTTGGAAGTTAAT TA) restriction sites. GST-Swd2p (pBD241) was produced by PCR amplification of the SWD2 open reading frame with primers encoding NcoI (CATGATCCATGGGTATGACCACCGTGTCCATCAAT) and BamHI (ATGCGGGATCCTCATTCATCGTACACGTAAAAATC) restriction sites and cloning into pGDV1 (Dichtl and Keller 2001). H3K4 and H3K9 mutations were produced by site-directed mutagenesis on plasmid pJH18 (CEN ARS TRP1 HHT1-HHF1; details available upon request; Hsu et al. 2000). The resulting plasmids pBD364 (H3K4A) and pBD365 (H3K9A) were transformed into the shuffling strain MX1-4C (Hsu et al. 2000). Following selection for transformants, URA3-marked H3K4 plasmids were shuffled out by plating on 5-FOA medium.

RNA analyses

Northern analyses were done as described (Dichtl et al. 2002a). PCR products covering the following sequences (relative to the A of the translational start codon) served as templates for probes: PGK1 (nt 1–1251), CYH2 (nt 1–450), ADH1 (nt 1–1049), CUP1 (nt 1–186), ACT1 (nt 1–677), anti-TRS31 (nt 34–849), anti-YCR015c (nt 60–654), anti-MRPL15 (nt 28–838), anti-NRD1 (nt 3–706), anti-SWD2 (nt 1–990) and anti-SET1 (nt 1–772). Oligonucleotide probes were labeled with γ[³²P]ATP and T4 polynucleotide kinase: anti-18S (CAGACAAATCACTCCA), anti-snR13 (GGTCAGATAAAAGTAAAAAAAGGTAGC), anti-snR33 (CCGGTGTAAAAGCTAGGCTTCAATC), anti-snR45 (ATCGCT CCGAGAAGAATTGTTCGAT), and anti-snR71 (TCTGAGTGAG CTGAGAAGGTTATCA). Transcriptional run-on analysis was performed according to Birse et al. (1998).

Protein analyses and in vitro 3′ end processing

Protein extracts for analysis of histone proteins were prepared from 50-ml cultures of exponentially growing yeast by glass-bead homogenization [50% (v/v) glass beads; 800 μL total volume] in homogenization buffer A (20 mM HEPES-KOH at pH 8.0, 350 mM KCl, 10% glycerol, 0.1% Tween 20, 1 mM PMSF) at 4°C. Crude extracts were cleared by three consecutive 30-min centrifugation steps at maximum speed in a table-top centrifuge at 4°C. Concentration of total protein was determined by Bradford analysis and 50 μg total protein were analyzed from each sample by SDS-PAGE on 15% gels. Antibodies for detection of di- and tri-methylated H3K4 were obtained from Abcam. Protein extracts for gel-filtration analysis were prepared as described for histone extracts described above, but in homogenization buffer B (150 mM KCl, 20 mM Tris-HCl at pH 7.9, 10% glycerol, 0.01% NP-40, 1 mM DTT, 1 mM PMSF). Three hundred fifty micrograms total protein were applied to a SMART Superdex 200 column, equilibrated in homogenization buffer B. Eighty-milliliter fractions were collected and 20 μL were used for SDS-PAGE on 12% gels. Molecular weight standards (Blue dextran: 2000 kD; thyroglobin: 669 kD; ferritin: 440 kD; catalase: 232 kD; ovalbumin 43 kD) were resolved under identical conditions. Production of GST-fusion proteins, in vitro translation, and GST pull-down assays were performed as described (Dichtl et al. 2002b). Preparation of extracts and in vitro 3′ end processing was done as described (Ohnacker et al. 2000).

Database searches and sequence analysis

Database searches in the UniProt database (Apweiler et al. 2004) were performed with Blastp (Altschul et al. 1997) as implemented at ExPASy (URL: http://au.expasy.org/tools/blast/). Searches were performed with S. cerevisiae Swd2 and S. pombe Swd2.1 and Swd2.2 as queries using default parameters. Top scoring sequences from a representative set of organisms were collected and multiply aligned using Clustal_X (Thompson et al. 1997) and manually edited to optimize minor poorly aligned regions. The C. elegans sequence Ce_C33H5.6 (trembl:Q18404) was obviously mispredicted and was corrected aided by profile alignment to the genomic sequence (embl:CEC33H5) using PairWise (Birney et al. 1996) and the Swd2 family profile, which led to the correction of exon 4 to position 3292–3582 and a new last exon at 4016–4252. The profile was generated from the multiple alignment using PairWise and used to generate the profile–profile dotplot using Pro-Plot (Thompson et al. 1997).