Previous work links histone methylation by Set2 with transcriptional elongation. yFACT (Spt16–Pob3 and Nhp6) reorganizes nucleosomes and functions in both transcriptional initiation and elongation. We show that growth defects caused by spt16 or pob3 mutations can be suppressed by deleting SET2, suggesting that Set2 and yFACT have opposing roles. Set2 methylates K36 of histone H3, and K36 substitutions also suppress yFACT mutations. In contrast, set1 enhances yFACT mutations. Methylation at H3 K4 by Set1 is required for set2 to suppress yFACT defects. We did not detect an elongation defect at an 8 kb ORF in yFACT mutants. Instead, pob3 mutants displayed reduced binding of both pol II and TBP to the GAL1 promoter. Importantly, both GAL1 transcription and promoter binding of pol II and TBP are significantly restored in the pob3 set2 double mutant. Defects caused by an spt16 mutation are enhanced by either TBP or TFIIA mutants. These synthetic defects are suppressed by set2, demonstrating that yFACT and Set2 oppose one another during transcriptional initiation at a step involving DNA binding by TBP and TFIIA.
Eukaryotic DNA is packaged into highly compacted chromatin, limiting the accessibility of transcription factors to DNA. DNA sequences within chromatin can be made accessible to transcription factors in several ways. First, ATP‐dependent chromatin remodeling factors can use the energy of ATP to move nucleosomes and expose the DNA sequences for transcription factor binding (Cairns, 2005). Second, variant histones can be inserted into nucleosomes, resulting in chromatin with altered properties (Kamakaka and Biggins, 2005). Third, nucleosomes can be altered by post‐translational modifications, including acetylation, methylation and ubiquitylation of lysine residues and phosphorylation of serine residues (Strahl and Allis, 2000; Peterson and Laniel, 2004). These modifications may directly change the properties of chromatin, thereby aiding factor binding, and they may create recognition sites for other factors such as bromodomain‐ and chromodomain‐containing proteins that recognize acetylated and methylated lysines, respectively.
We have been studying a fourth mechanism that enhances accessibility of binding sites, the ATP‐independent reorganization of nucleosomes by yFACT (Formosa, 2003). Mammalian FACT contains two subunits, hSpt16 and SSRP1, which are homologous to Spt16 and Pob3 in yeast, except that SSRP1 has a single HMGB DNA‐binding motif at its C‐terminus that is missing from Pob3 (Wittmeyer and Formosa, 1997; Orphanides et al, 1999). The yeast Nhp6 protein, essentially a single HMGB domain, supports the ability of Spt16–Pob3 to function as yFACT both in vitro and in vivo (Brewster et al, 2001; Formosa et al, 2001). There is substantial evidence linking yFACT to transcriptional elongation. For example, several yFACT mutants are sensitive to the elongation inhibitor 6‐azauracil (6‐AU) (Orphanides et al, 1998; Formosa et al, 2001) and show genetic interactions with known elongation factors (Formosa et al, 2002; Squazzo et al, 2002). In addition, yFACT physically associates with several elongation factors (Krogan et al, 2002; Squazzo et al, 2002; Simic et al, 2003), human FACT facilitates RNA polymerase II (pol II) elongation through a chromatin template in vitro (Orphanides et al, 1998), and ChIP and immunolocalization studies show association of FACT with elongating pol II (Mason and Struhl, 2003; Saunders et al, 2003).
In addition to a role in transcript elongation, experiments also suggest that yFACT has a role in regulating transcriptional initiation. An spt16 mutation can change the site of transcriptional initiation (Malone et al, 1991), and Drosophila FACT associates with the GAGA factor and stimulates chromatin changes at promoters (Shimojima et al, 2003). Spt16 inactivation in yeast results in reduced binding of TBP and TFIIB at promoters, spt16 mutants show strong genetic interactions with mutations affecting TBP and TFIIA, and yFACT facilitates TBP and TFIIA binding to nucleosomal binding sites in vitro (Mason and Struhl, 2003; Biswas et al, 2005). yFACT can, therefore, enhance the accessibility of DNA sequences in chromatin (Formosa et al, 2001), and this is an important component of transcriptional regulation both during initiation and elongation. yFACT also interacts with DNA polymerase α, and with MCM proteins, and it plays an important role in DNA replication (Wittmeyer et al, 1999; Gambus et al, 2006; VanDemark et al, 2006).
Histone proteins are methylated by SET domain‐containing proteins, and histone methylation can regulate transcription (Lee et al, 2005; Martin and Zhang, 2005). In yeast, histone H3 is methylated at K4, K36 and K79 by the Set1, Set2, and Dot1 histone methyl transferases, respectively. It has been suggested that K4 methylation by Set1 facilitates transcriptional elongation, as K4 methylation is enriched in the transcribed regions of actively transcribed genes (Liu et al, 2005; Pokholok et al, 2005), and Set1 is recruited to elongating RNA polymerase complexes (Ng et al, 2003). These observations suggest that Set1 is a positive elongation factor.
Di‐ and tri‐methylation at K36 by Set2 is also found at transcribed open reading frames (Liu et al, 2005; Pokholok et al, 2005; Rao et al, 2005). Set2 also associates with the elongating form of RNA polymerase (Krogan et al, 2003; Li et al, 2003; Xiao et al, 2003). Moreover, set2 mutants show synthetic growth defects with genes implicated in elongation, consistent with Set2 also being a positive elongation factor (Krogan et al, 2003; Li et al, 2003). However, several observations are more consistent with a negative role for Set2 in regulating transcription initiation. K36 methylation by Set2 is required to recruit the Rpd3S histone deacetylase complex through its Eaf3 chromodomain subunit, and deacetylation by Rpd3S may be required to restore chromatin to the repressed pretranscribed state (Carrozza et al, 2005; Joshi and Struhl, 2005; Keogh et al, 2005). Set2 represses transcription when tethered to a heterologous promoter, indicating a direct negative effect on initiation (Strahl et al, 2002). Additionally, expression of a mutant GAL4 promoter lacking its UAS element is very low, but can be increased either by a set2 mutation or a histone H3 K36R substitution, suggesting that modification of H3 by Set2 inhibits initiation (Landry et al, 2003). Importantly, although several studies have shown greater K36 methylation at open reading frames (Pokholok et al, 2005; Rao et al, 2005), it is clear that K36 methylation also occurs in promoter regions (Xiao et al, 2003). Set2 therefore has complex and perhaps opposite effects on different stages of transcriptional regulation.
In this report, we show that yFACT and K36 methylation by Set2 have opposing roles in regulating transcription. Surprisingly, we find no evidence that mutations affecting yFACT and Set2 influence transcriptional elongation rates. Instead, our results show that yFACT and histone methylation by Set2 regulate, in opposite ways, binding of both pol II and TBP to promoters.
Set1 methylation at histone H3 K4 supports the function of yFACT
SPT16 and POB3 are essential genes, and mutant alleles with distinct phenotypes have been isolated (Malone et al, 1991; Rowley et al, 1991; Schlesinger and Formosa, 2000; Formosa et al, 2001). We chose the spt16‐11 and pob3(L78R) alleles for these studies because they display the Spt‐phenotype from inappropriate TATA element usage, and they are sensitive to elevated temperatures, to the dNTP synthesis inhibitor hydroxyurea (HU), and to the transcription elongation inhibitor 6‐AU. Thus, the phenotypes of the spt16‐11 and pob3(L78R) alleles suggest that they have defects in transcriptional initiation, transcriptional elongation, as well as in replication of DNA.
We previously showed that some yFACT mutations are synthetically lethal with some mutations in histone H3 and H4, including deletions of the N‐terminal tails and mutations of certain acetylatable lysine residues (Formosa et al, 2002; VanDemark et al, 2006). Here, we look for genetic interactions between yFACT mutations and H3 mutations in methylated lysines K4 and K79, and acetylated site K23. Strains with deletions of both sets of chromosomal genes encoding histone H3 and H4, and carrying a YCp‐URA3 plasmid with the wild‐type HHT2–HHF2 genes were constructed. Plasmids with either wild‐type or mutant HHT2‐HHF2 alleles were introduced into these strains by transformation, and the ability of transformants to grow on media with FOA was assessed. URA3+ strains cannot grow on FOA, and thus growth demonstrates that the wild‐type histone genes on the YCp‐URA3 plasmid can be lost with the introduced plasmid supporting viability.
As shown in Figure 1A, introducing plasmids with wild‐type histones, H3(K4R), H3(K23R), or H3(K79R) into a wild‐type strain results in healthy growth, while the empty vector does not. We conclude that these H3 mutations support viability in a wild‐type strain, the H3(K23R) mutation shows a modest growth defect in combination with either an spt16‐11 (Figure 1B) or a pob3(L78R) (Figure 1C) mutation, and H3(K79R) does not affect growth of these mutants. The H3(K4R) mutation has a more striking effect, showing a strong synthetic defect when combined with either spt16 or pob3. Lysine 4 of histone H3 is methylated by the Set1 enzyme (Briggs et al, 2001), and thus we predict a similar effect from a set1 mutation. We constructed an spt16 set1 double mutant and found it to be viable at 25°C, but lethal at 33°C (Figure 1D). We were unable to construct a pob3 set1 double mutant, as it was lethal at all temperatures tested. We conclude that the function of yFACT is strongly dependent on methylation of histone H3 at K4 by Set1.
Absence of Set2 methylation at histone H3 K36 suppresses temperature sensitivity caused by yFACT mutations
In contrast with our results with the K4R mutation, we found that mutations at histone H3 K36 suppress growth defects associated with yFACT mutations. The spt16‐11 mutant does not grow at 35°C, as evidenced by its failure to grow on FOA when containing a plasmid with wild‐type histone genes (Figure 2A). However, the spt16 mutant grows on FOA if the plasmid contains either a K36R or a K36A mutation in histone H3. Similarly, a strain with the pob3(L78R) allele is unable to grow at 30°C, but the H3 K36R or K36A mutations suppress this growth defect (Figure 2B). To verify that this apparent growth suppression was not an artefact of growth on FOA‐containing medium, we used a plasmid shuffle at 25°C to evict the YCp‐URA3 plasmid, obtaining strains with a YCp‐TRP1 plasmid with either wild‐type histone genes or a derivative with the H3(K36A) substitution. The spt16 mutant with wild‐type histone H3 is unable to grow at 35°C, but the K36A substitution suppresses (Figure 2C). Similarly, histone H3(K36A) suppresses the pob3 mutant (Figure 2D).
The Set2 enzyme methylates K36 of histone H3 (Strahl et al, 2002), so if methylation at this site by Set2 is the cause of the suppression, a set2 mutation should have a similar phenotype to an H3 K36 mutation. As shown in Figure 2E and F, a set2 gene deletion also suppresses the temperature sensitivity of the spt16 and pob3 strains. These observations clearly indicate that Set2 methylation of histone H3 at K36 has an opposing role to that of yFACT in supporting viability.
Opposing roles of Set1 and Set2 methyltransferases
The temperature‐sensitive growth defect in an spt16 mutant is affected by set1 and set2 mutations, but in opposite directions. To examine the epistasis relationships, we constructed an spt16 set1 set2 triple mutant. The results in Figure 3A show that the triple mutant has a marked growth defect, although not quite as severe as the spt16 mutant or the spt16 set1 double mutant. We next constructed a plasmid with both histone H3 K4R and K36R mutations and tested it in the spt16 and pob3 mutants. The results in Figure 3B show that the H3(K4R,K36R) double mutant is synthetically lethal with both spt16 and pob3, the same phenotype seen with K4R. Thus, the K4R mutation is epistatic to K36R. We conclude that the absence of Set2 methylation at H3 K36 can only suppress the yFACT defects when methylation by Set1 occurs at H3 K4.
Histone mutations affect growth in nhp6ab mutant strains
In addition to Spt16 and Pob3, yFACT contains Nhp6, a small HMG protein required for nucleosomal binding by Spt16–Pob3 (Formosa et al, 2001). Nhp6 is encoded by two redundant genes, NHP6A and NHP6B, and the nhp6ab double mutant shows growth defects. We next examined the effect of histone mutations on an nhp6ab strain. Similar to the results with spt16 and pob3 mutants, the H3(K4R) mutation caused a severe synthetic defect with nhp6ab (Figure 3C) and H3(K79) mutations had no effect (not shown). However, the H3(K36) substitutions had a markedly different effect in the nhp6ab mutant; instead of the suppression seen with spt16 and pob3, the K36 mutations inhibited growth of the nhp6ab strain. A synthetic growth defect is also seen when a set2 mutation is introduced into a nhp6ab strain (data not shown). In addition to its role in yFACT, Nhp6 has been shown to interact with other chromatin proteins, including Swi/Snf, RSC, and Ssn6/Tup1 (Szerlong et al, 2003; Biswas et al, 2004; Fragiadakis et al, 2004) and to play a role in transcription by RNA polymerase III (Kassavetis and Steiner, 2006). These additional roles for Nhp6 could explain why the H3(K36) mutations have such markedly different effects in the nhp6ab strain compared to the spt16 and pob3 mutants.
A variety of yFACT mutant defects are suppressed by set2
spt16‐11 mutants are defective for growth on media containing 6‐AU (Formosa et al, 2001). 6‐AU is a uracil analog that causes imbalances in the pools of rNTPs (Exinger and Lacroute, 1992; Shaw and Reines, 2000), and many strains with defects in transcriptional elongation are sensitive to 6‐AU. We determined whether the 6‐AU sensitivity caused by spt16 could be suppressed either by a histone mutation or by a set2 mutation. As shown in Figure 4A, the 6‐AU sensitivity of the spt16 strain (line 4) is suppressed by a K36A substitution in histone H3 (line 6). A set2 mutation similarly suppresses the 6‐AU sensitivity caused by spt16 (Figure 4B). We note that set2 mutants display slightly higher 6‐AU resistance than wild‐type strains, consistent with previous reports (Adhvaryu et al, 2005; Keogh et al, 2005; Kizer et al, 2005). Thus, spt16 and set2 mutants have opposite responses to 6‐AU, suggesting that yFACT and Set2 have opposing roles in transcriptional elongation. However, sensitivity to 6‐AU does not necessarily demonstrate a role in transcriptional elongation; a mutation in the SNR6 promoter that reduces expression of the U6 small nuclear RNA causes 6‐AU sensitivity (Eriksson et al, 2004).
Based on the observation that a set2 mutation suppresses the temperature‐ and 6‐AU‐sensitive phenotypes associated with yFACT mutations, we tested whether set2 can also suppress other synthetic defects observed with yFACT mutants. An spt16 mutation displays synthetic defects with mutations in either the GCN5 or the ELP3 histone acetyltransferase genes, or with the nhp6ab double mutant (Formosa et al, 2002), and a set2 mutation suppresses all of these synthetic growth defects (Supplementary Figure S1). It has been proposed that the Isw1 chromatin remodeling factor has a role in transcriptional elongation (Morillon et al, 2003). A spt16 isw1 double mutant shows a synthetic growth defect at 33°C, and this is suppressed by set2 (Figure 4C). These suppression results support the idea that yFACT and Set2 have opposing roles in multiple functions.
The H2A.Z histone variant of H2A in yeast, encoded by HTZ1 gene, has diverse functions (Dryhurst et al, 2004). We constructed an spt16 htz1 double mutant strain, and observed synthetic lethality at 33°C (Figure 4D). This growth defect is also suppressed by a set2 mutation. Importantly, genome wide studies show that the Htz1 protein localizes preferentially at the promoter regions of genes (Li et al, 2005; Raisner et al, 2005; Zhang et al, 2005). This promoter localization of Htz1, and the genetic interactions seen here, suggest that Set2 and yFACT might have opposing roles at promoter regions, in addition to their proposed elongation functions.
Specificity of set2 suppression
We also tested whether set1 or set2 mutations also affect other factors thought to be involved in transcriptional elongation. We constructed double mutant strains, combining either a set1 or a set2 mutation with disruptions in PAF1, CDC73, DST1, SPT4, or ELP3. The double mutants with set1 or set2 were examined for growth phenotypes, including sensitivity to temperature and 6‐AU. There are some instances of suppression and some of synthetic defects (Supplementary Figure S2A). However, these elongation mutants do not all show suppression with set2 and synthetic defects with set1, and thus the effect appears to be specific to spt16 and pob3.
spt16 mutants also cause an Spt‐phenotype, altering transcription start sites from the his4‐912δ and lys2‐128δ alleles and conferring a His+ Lys+ phenotype (Malone et al, 1991). Interestingly, spt16 set2 strains are still Spt‐, and thus set2 does not suppress this phenotype (data not shown). Similarly, the synthetic growth defect seen in a spt16 rpd3 double mutant (Formosa et al, 2002) is not suppressed by set2 (data not shown). However, we do see suppression by set2 of the Spt‐phenotype seen in a gcn5 mutant with the lys2‐173R2 allele (Supplementary Figure S1D). A set2 deletion, therefore, does not suppress all elongation defects, and it also does not suppress all of the defects caused by yFACT mutations.
Pob3 and Set2 regulate GAL1 induction in opposing ways
Mason and Struhl (2003) used a GAL1‐YLR454w reporter, with the GAL1 promoter inserted upstream of the nonessential 8 kb YLR454w gene, to show that yFACT associates with open reading frames during transcription. We constructed pob3 and set2 strains with this GAL1‐YLR454w allele and measured YLR454w mRNA levels by S1 nuclease protection assays following induction of the GAL1 promoter. The results in Figure 5A show a rapid increase in YLR454w mRNA in the wild‐type strain, as expected. The pob3 mutant strain is markedly defective in inducing YLR454w mRNA from the GAL1 promoter (Figure 5A), and a similar defect is seen in induction of the native GAL1 gene in the pob3 mutant (Figure 5B). Importantly, deletion of SET2 ameliorates the transcriptional defect significantly at GAL1‐YLR454w (Figure 5A), and completely at GAL1 (Figure 5B) (compare pob3 and pob3 set2). Additionally, we examined other genes and found either an spt16 or pob3 mutation could reduce expression, and a set2 mutation partially suppresses these defects (Supplementary Figure S3A). These observations imply that Set2 and yFACT have opposing roles in regulating transcription.
While a set2 mutant caused decreased induction of GAL1‐YLR454w, it displayed increased expression before induction (Figure 5C). Set2 is, therefore, needed both for repression of the GAL1 promoter in this context and for normal induction of expression, underscoring the complexity of the role of Set2 in transcription. Notably, Set2 is not needed for full induction of the native GAL1 message (Figure 5B). It is possible that the large size of the 8 kb YLR454w transcription unit places a larger demand on Set2 function than the 1.6 kb GAL1 gene.
Defect in pol II binding to the GAL1 promoter in pob3 mutants
We used the GAL1‐YLR454w reporter to assess the rate of pol II elongation in wild‐type and mutant strains in several ways. In the first assay, we took RNA samples every 10 min following galactose induction, and used probes for S1 nuclease protection assays specific to the 5′ and 3′ ends of the gene. The time between appearances of mRNA sequences corresponding to the 5′ and 3′ ends gives an indication of how long it takes for pol II to traverse the 8 kb gene. We do not see a significant difference in this time in the wild‐type, set2, pob3, and pob3 set2 strains (Supplementary Figure S4). In a second assay to measure elongation rates, glucose was added to cells growing in galactose to shut down expression of GAL1‐YLR454w, and chromatin immunoprecipitation (ChIP) was used to measure Pol II occupancy during the last wave of transcription down the gene (Mason and Struhl, 2005). This assay did not show any elongation defect in pob3 mutants (Supplementary Figure S5). Thus, the marked defect in GAL1‐YLR454w expression in pob3 mutants is not due to an elongation defect. Of course, pob3 mutants may be defective for elongation at other genes.
We also used ChIP assays to measure pol II levels along the YLR454w gene following galactose induction. Figure 6A shows a map of the GAL1‐YLR454w reporter, and four regions amplified with specific primers corresponding to the GAL1‐YLR454w promoter, 1 kb downstream of start codon, the middle of the YLR454w ORF (+3600), and the 3′ end of the gene (+7800). At time intervals following galactose induction, samples were harvested, treated with formaldehyde to crosslink, and processed for ChIP. The results of the pol II ChIP from the wild‐type strain are shown in Figure 6B. As expected, pol II occupancy increases throughout the gene during induction. Pol II binding to the promoter is not saturated during this time course, unlike previous results (Bryant and Ptashne, 2003); there are important differences in media and cell concentration between the two studies. A delay is seen in pol II occupancy at +1000 and downstream, reflecting the amount of time required to initiate and progress through the ORF. For example, occupancy reaches a 30‐fold ratio after about 8 min at the promoter, 14 min at +1000, and 22 min at +7800. Notably, the time course is similar for a pob3 mutant, although the overall occupancy is much lower (Figure 6C). This is consistent with similar elongation rates at GAL1‐YLR454w in wild‐type and pob3 strains; similar results are seen with set2 and pob3 set2 strains (data not shown).
Examination of the wild‐type strain shows that occupancy at the promoter is much higher than it is at any point within the ORF at all time points. Plotting the Pol II occupancy at the 50 min time point clearly shows this difference (Figure 6D). Greater occupancy at the promoter suggests that recruitment is faster than the transition from initiation to elongation under these strongly inducing conditions. Pol II occupancy at all sites decreases in both set2 and pob3 mutants, so Set2 and yFACT are both involved in recruiting Pol II to this promoter. Interestingly, among the mutants, only the pob3 set2 strain has slightly higher pol II occupancy at the promoter than at the ORF. Loss of the higher accumulation of Pol II at the promoter compared to the ORF in the mutants might indicate that Set2 and yFACT cooperate to impose a barrier or selective elongation switch between recruitment and initiation or elongation by Pol II. This functional cooperation between Set2 and yFACT contrasts with the opposing roles detected above.
Importantly, while a pob3 strain has a severe defect in pol II occupancy at the at GAL1‐YLR454w promoter, the set2 pob3 double mutant has a less severe defect, similar to the set2 single mutant (Figure 6D). The same effect is seen at the native GAL1 promoter, where the set2 mutation does not by itself cause diminished pol II occupancy (Figure 6E). Deletion of SET2, therefore, at least partially restores the ability of a pob3 mutant to recruit pol II to a promoter.
Defective TBP binding in pob3 mutants is suppressed by set2
Pol II is typically recruited to promoters by the TATA‐binding protein TBP. We have previously shown that yFACT can play a role in transcriptional initiation through regulation of TBP binding (Biswas et al, 2005), and we therefore examined TBP binding to the GAL1‐YLR454w promoter in these mutants. As shown in Figure 6F, TBP binding to GAL1‐YLR454w is essentially eliminated in the pob3 mutant but is largely restored in the pob3 set2 double mutant strain. We also found TBP binding at other promoters is reduced in a pob3 mutant, and increased TBP binding is seen in the pob3 set2 strain compared to the pob3 single mutant (Supplementary Figure S3B). Thus, TBP binding to the GAL1 promoter is stimulated by yFACT, and the pob3 defect can be suppressed by a set2 mutation.
Pol II binding at GAL1‐YLR454w is markedly reduced in the set2 and pob3 set2 strains, while TBP binding is at wild‐type levels. Thus TBP binding is not sufficient to direct pol II recruitment, and other factors such as TFIIA or TFIIB may be involved.
The synthetic lethality between spt16 and either TBP or TFIIA mutations is suppressed by set2
We previously demonstrated genetic interactions between spt16 and TBP mutations (Biswas et al, 2005). We showed that combining mutant Spt16 and mutant TBP proteins in the same cell results in lethality, using a plasmid shuffle assay to introduce mutant alleles into a strain lacking both SPT15 (which encodes TBP) and SPT16.
We have repeated this experiment, now including an spt15Δ spt16Δ set2 YCp‐URA3‐TBP‐Spt16 strain. As shown in Figure 7A, certain combinations of Spt16 and TBP result in lethality in SET2 strains, but a set2 mutation allows these combinations of Spt16 and TBP to be viable. For example, cells with spt16‐11 and TBP(E93G) cannot lose the YCp‐URA3‐TBP‐Spt16 plasmid with the wild‐type genes, as evidenced by the failure to grow on FOA, demonstrating the synthetic lethality. In contrast, the set2 mutant with spt16‐11 and TBP(E93G) can grow on FOA. Thus, set2 suppresses the synthetic lethality between spt16 and TBP.
We also showed synthetic lethality between spt16 and TFIIA mutations (Biswas et al, 2005). Yeast TFIIA protein is composed of two subunits, Toa1 and Toa2. Based on the TBP–TFIIA crystal structure, Ozer et al (1998) generated toa2 mutations at the TBP–TFIIA interface. Although these mutations in TOA2 eliminate TBP–TFIIA interactions with in vitro binding assays, the toa2 mutants are viable, presumably because other factors present in cells facilitate TBP–TFIIA interaction and DNA binding. For example, Toa2(W76) is required for cooperative DNA binding with TBP in vitro, but a toa2(W76A) mutant is viable in an otherwise wild‐type strain (Ozer et al, 1998; Biswas et al, 2005). The toa2(W76A) allele is lethal in an spt16 mutant (Figure 7B). However, toa2(W76A) is viable in the spt16 set2 strain, and thus set2 suppresses the synthetic lethality between spt16 and TFIIA mutations.
Suppression of the synthetic lethality between spt16 and TBP or TFIIA mutants by deletion of SET2 strongly supports our hypothesis that yFACT and Set2 have opposing roles in supporting binding of TBP and TFIIA to promoters.
Spt16 and Pob3, along with Nhp6, comprise the yFACT complex that can reorganize chromatin structure. SPT16 and POB3 are essential genes, and spt16 and pob3 point mutants have been isolated with phenotypes including temperature‐sensitive growth, sensitivity to 6‐AU, and synthetic lethality with certain transcription factor mutations. We find that a set2 deletion, eliminating the enzyme that methylates lysine 36 on histone H3, suppresses all of these phenotypes. Additionally, a histone H3 mutation, replacing the lysine at position 36 with either an alanine or an arginine residue, also suppresses spt16 and pob3 mutations. We conclude that methylation on histone H3 at K36 by Set2 acts in opposition to the chromatin changes facilitated by yFACT. spt16 and pob3 mutants also cause initiation at aberrant sites, leading to the Spt‐ phenotype. This defect was not suppressed by set2, indicating that not all functions of yFACT are opposed by Set2.
Both Set2 and yFACT have been implicated in regulating transcriptional elongation. Both localize preferentially to transcribed regions compared to promoters (Krogan et al, 2003; Mason and Struhl, 2003; Saunders et al, 2003), Set2 associates with hyperphosphorylated pol II (Li et al, 2003; Xiao et al, 2003), and several studies have shown greater K36 methylation at open reading frames compared to promoters (Xiao et al, 2003; Pokholok et al, 2005; Rao et al, 2005). Additionally, both spt16 and set2 mutations show genetic interactions with known elongation factors (Formosa et al, 2002; Squazzo et al, 2002; Krogan et al, 2003; Li et al, 2003). We looked for defects in transcriptional elongation in a pob3 mutant, using a strain with the GAL1 promoter inserted upstream of the nonessential 8 kb YLR454w gene. We found no evidence for a transcriptional elongation defect at GAL1‐YLR454w in pob3 mutants, but pob3 mutants may affect elongation at other genes. We did find reduced expression of both GAL1 and GAL1‐YLR454w in the pob3 mutant. Importantly, expression was restored in the pob3 set2 double mutant.
ChIP experiments indicate that yFACT and Set2 act at the GAL1 promoter at the level of transcriptional initiation. Following galactose induction, there is reduced binding of both pol II and TBP at the GAL1 promoter in the pob3 mutant; binding of both factors is increased in the pob3 set2 strain. These experiments suggest that yFACT facilitates TBP binding, at least at some promoters, and Set2 opposes this effect. How does Set2 inhibit TBP binding at GAL1? One possibility is that histone methylation by Set2 inhibits binding of a transcriptional coactivator complex that stimulates TBP binding. The SAGA complex is recruited to GAL1 and promotes TBP binding (Dudley et al, 1999). We have analyzed SAGA binding at the GAL1 UAS by ChIP, but we did not observe any significant difference in SAGA binding in strains differing at the POB3 and SET2 loci (data not shown). Chromodomains bind to methylated histone residues, and it is possible that methylated K36 residues at the GAL1 promoter encourage binding of a chromodomain‐containing transcription factor that regulates TBP binding.
Our results strongly support a role for K36 methylation by Set2 in decreasing TBP binding to promoters. There are two earlier studies that support a repressive role for Set2 at promoters. A Set2–LexA fusion protein strongly represses transcription when tethered to a promoter with a lexA‐binding site (Strahl et al, 2002). This repression by the Set2–LexA fusion protein was reduced by point mutations in the Set2 catalytic domain. In a separate study, the weak expression from a GAL4 promoter lacking its UAS element could be suppressed by either a set2 mutation or a K36R substitution in histone H3 (Landry et al, 2003). This suggests that Set2 can repress transcription from promoters.
Set2 and yFACT are believed to function in transcriptional elongation, although we did not see an effect of mutations on the elongation rate at one specific gene. Our results strongly suggest that both Set2 and yFACT function at initiation of transcription by regulating DNA binding by TBP, at least at the GAL1 promoter. We previously demonstrated that spt16 mutants are synthetically lethal with point mutations in either TBP or TFIIA (Biswas et al, 2005). Importantly, a set2 mutation can suppress this synthetic lethal interaction. Additionally, in vitro studies show that yFACT can facilitate binding of TBP and TFIIA to a nucleosomal TATA site that is normally refractory to binding (Biswas et al, 2005). Biochemical studies have shown that the TFIIS factor (encoded by DST1 in yeast) facilitates elongation by pol II. However, it was recently shown that TFIIS binds to the GAL1 promoter (Prather et al, 2005). Additionally, a dst1 mutation affected the kinetics of GAL1 induction, and reduced the association of both TBP and pol II to the GAL1 promoter (Prather et al, 2005), and a further decrease in promoter occupancy by basal factors is seen when a dst1 mutant is treated with 6‐AU (Mason and Struhl, 2005). It has been suggested that a decrease in elongation slows promoter clearance, leading to destabilization of the preinitiation complex (Mason and Struhl, 2005). It is intriguing that TFIIS, Set2 and yFACT, all proposed as elongation factors, regulate binding of TBP at GAL1.
Although a set2 mutation suppresses defects caused by yFACT mutations, we find that combining a set1 mutation with either spt16 or pob3 results in synthetic defects. Set1 and Set2 methylate histone H3 at different residues, K4 and K36, respectively. A histone H3 K4 substitution enhances yFACT defects, while a substitution at K36 suppresses these defects. We examined the epistasis relationships by constructing an spt16 set1 set2 triple mutant, and by testing a histone H3(K4R, K36R) double mutant in spt16 and pob3 strains. The results show that the absence of K36 methylation is not sufficient to suppress the yFACT mutants. It is the combination of methylated K4 and unmethylated K36 at histone H3 that suppresses the yFACT defects.
How do the presence or absence of methylated lysine residues on histone H3 produce such marked effects on growth of cells with a partially defective yFACT complex? Chromodomain‐containing proteins can bind to methylated lysine residues. One group has reported that yeast Chd1 binds to methylated K4 (Pray‐Grant et al, 2005), while another group finds that human, but not yeast Chd1, is capable of binding to methylated K4 (Sims et al, 2005). It is possible that in the spt16 H3(K4)R mutant, the lack of binding of a chromodomain protein, possibly Chd1, is toxic in the presence of the defective yFACT complex. The Eaf3 chromodomain protein has been shown to bind methylated H3 K36 (Carrozza et al, 2005; Joshi and Struhl, 2005; Keogh et al, 2005). Eaf3 is present in two complexes, NuA4 and Rpd3S, and thus it is possible that the absence of one of these complexes and their associated enzymatic activities suppresses the growth defects of the spt16 and pob3 mutant strains. It has also been shown that K36 methylation by Set2 recruits the Rpd3S histone deacetylase complex to the 3′ portions of coding regions. In set2 or H3(K36R) mutants, there could be a redistribution of Rpd3S from coding regions to promoters, and thus the effect of set2 on TBP binding could be indirect. Further work will be needed to decipher the molecular mechanisms of how loss of H3 K36 methylation suppresses yFACT mutants.
Materials and methods
Yeast strains are listed in Supplementary Table S1. Cells were grown in YPD medium (Sherman, 1991) at 30°C, except where other temperatures are noted, or in synthetic complete medium (Sherman, 1991) with 2% glucose and supplemented with adenine, uracil and amino acids, as appropriate, to select for plasmids. For the galactose induction experiments, cells were grown at 25°C in YP medium supplemented with 2% raffinose to mid log, shifted to 30°C for growth for 2 h, and then galactose was added to a final concentration of 2%. Plasmids are listed in Supplementary Table S2. The histone mutations were generated by site‐directed mutagenesis, and the sequences verified.
RNA levels were determined with S1 nuclease protection assays as described (Bhoite and Stillman, 1998) using probes listed in Supplementary Table S3. Chromatin immunoprecipitations were performed as described (Ausubel et al, 1987) using the 8WG16 monoclonal antibody against the pol II C‐terminal repeat, and a polyclonal anti‐TBP sera generously provided by Tony Weil, (Schroeder et al, 2000). Real‐time PCR and calculations were performed as described (Eriksson et al, 2004), using the ORF‐free chromosome I region (Mason and Struhl, 2005) as the internal control.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Supplementary Table S1
Supplementary Table S2
Supplementary Table S3
We thank Fred Winston for providing strains, Kevin Struhl for providing the GAL1‐YLR454w allele, Tony‐Weil for anti‐TBP sera, and Grant Hartzog for comments on the paper. This work was supported by grants awarded to TF, DJS, and BDS from the National Institutes of Health.
- Copyright © 2006 European Molecular Biology Organization