Skip to main content
NCBI home page
The EMBO Journal logoLink to The EMBO Journal
. 2004 Sep 30;23(21):4243–4252. doi: 10.1038/sj.emboj.7600433

Evidence for distinct mechanisms facilitating transcript elongation through chromatin in vivo

Arnold Kristjuhan 1, Jesper Q Svejstrup 1,a
PMCID: PMC524397  PMID: 15457216

Abstract

The mechanism and kinetics of RNA polymerase II transcription and histone acetylation were studied by chromatin immunoprecipitation in yeast. Our results indicate that a significant fraction of polymerases starting transcription never make it to the end of a long GAL-VPS13 fusion gene. Surprisingly, induction of GAL genes results in substantial loss of histone–DNA contacts not only in the promoter but also in the coding region. The loss of nucleosomes is dependent on active transcript elongation, but apparently occurs independently of histone acetylation. In contrast, histones in genes previously shown to require the histone acetyltransferases GCN5 and ELP3 for normal transcription do not lose DNA contacts, but do become acetylated as a result of transcription. Together, these results suggest the existence of at least two distinct mechanisms to achieve efficient transcript elongation through chromatin: a pathway based on loss of histone–DNA contacts, and a histone acetylation-dependent mechanism correlating with little or no net loss of nucleosomes.

Keywords: ELP3, GAL1, GCN5, histone acetylation, transcript elongation

Introduction

The eukaryotic genome is packaged into chromatin, which presents a constant barrier to transcription and other cellular processes that require access to DNA. In order to contend with chromatin, the basal transcription machinery is assisted by numerous factors that increase the plasticity of nucleosomal DNA. The first group of chromatin-remodelling factors are enzymes that covalently modify the N-termini of histones (histone acetyltransferases (HATs), histone methyltransferases, etc.) to help make nucleosomes mobile or support the binding of other factors, while proteins in the second group, the ATP-dependent chromatin remodellers, change the positions of nucleosomes to expose DNA for the transcription machinery (Workman and Kingston, 1998; Narlikar et al, 2002). Acetylation of histones is one of the best-characterized general mechanisms for increasing the accessibility of chromatin and allowing factors to gain access to DNA. For example, it has been shown that many transcriptional activators recruit HATs to increase acetylation of histones at promoters (Hassan et al, 2001a; Kadam and Emerson, 2002), and we have previously shown that the steady-state level of H3 acetylation in the coding region of genes has a positive correlation with gene activity (Kristjuhan et al, 2002). Although hyperacetylation of histones generally indicates higher transcriptional activity, not all active promoters are hyperacetylated and the need for acetylation differs substantially between genes (Deckert and Struhl, 2001; Kristjuhan et al, 2002; Kurdistani et al, 2004). In addition to ‘targeted' recruitment to genes, HATs and histone deacetylases (HDACs) also have a role in the maintenance of a global histone acetylation–deacetylation equilibrium in cells (Wittschieben et al, 2000). This mechanism has been proposed to be responsible for rapid restoration of a normal histone acetylation pattern when the signal for targeted acetylation or deacetylation has been removed (Vogelauer et al, 2000; Katan-Khaykovich and Struhl, 2002).

Acetylated histone tails can be recognized by chromatin remodelling complexes, which in turn can lead to a full opening of chromatin (Hassan et al, 2001a, 2001b). For example, nucleosomes in the PHO5 and the PHO8 promoter are first hyperacetylated by Gcn5, and then remodelled by the SWI/SNF complex to achieve full activation of the gene (Reinke et al, 2001; Reinke and Horz, 2003). Theoretically, remodelling of chromatin can occur in two ways: nucleosomes can be pushed away from the target DNA sequence, assuming that they can be accommodated in neighbouring regions of DNA (remodelling in cis). Alternatively, they may be temporally removed from DNA altogether (remodelling in trans). Recent studies of the PHO genes have shown that removal of nucleosomes from a promoter in response to induction of transcription can indeed take place (Boeger et al, 2003; Reinke and Horz, 2003), and that this removal occurs in trans, with the help of chromatin assembly factor Asf1 (Adkins et al, 2004; Boeger et al, 2004). The extent and generality of nucleosome loss in response to transcriptional activation has not been reported, and whether it is restricted to promoters is also unknown. However, in the case of PHO5 and PHO8, only promoter nucleosomes, not coding region nucleosomes, are lost (Adkins et al, 2004).

Transcription-related histone chaperones such as Spt6 and FACT are capable of binding histones and assembling chromatin (Belotserkovskaya et al, 2003; Kaplan et al, 2003), suggesting that they may be involved in disassembling and reassembling chromatin during transcript elongation. Nevertheless, the basic mechanisms employed by the transcription machinery to contend with nucleosomes in the coding region of activated genes in vivo are still poorly understood (Svejstrup, 2003).

In this study, we followed the dynamics of transcription and histone acetylation over the entire locus of two galactose-inducible genes, as well as three constitutive genes whose normal expression requires the HATs Gcn5 and Elp3. Our results shed new light on the mechanisms underlying transcriptional activation and transcript elongation through chromatin.

Results

Transcriptional induction of a GAL-VPS13 fusion gene

The Saccharomyces cerevisiae genome contains about 6000 genes with an average size of less than 1500 bp (Cherry et al, 1997). The very compact genome and short genes mean that discriminating between chromatin modification events originating from promoters and coding regions is difficult. To try and overcome the problem of too short genes, we constructed a model yeast strain where a galactose-inducible promoter was inserted in front of the long (9.5 kb) VPS13 gene at its natural locus (Figure 1A). In this strain, transcription of the synthetic GAL-VPS13 gene can be induced by galactose, and the transcriptional unit itself is long enough to distinguish between chromatin modification caused by transcriptional initiation and elongation, without considerable influence from events elsewhere in the gene or at other genes nearby.

Figure 1.

Figure 1

Open in a new tab

Kinetics of RNAPII recruitment to GAL-VPS13 and GAL1 upon galactose induction. (A) Schematic representation of the synthetic GAL-VPS13 locus on chromosome XII. Vertical arrows indicate the locations of promoter (prom) and coding region (2.2, 5.5, and 8.3 kb) PCR probes that were used in ChIP assays. (B) The amount of RNAPII on the promoter and in the coding region of GAL-VPS13 was analysed by ChIP. Co-immunoprecipitated DNA was quantified by real-time PCR with the appropriate primers. (C) Kinetics of VPS13 mRNA accumulation upon induction of the GAL-VPS13 gene measured by real-time RT–PCR with primers close to the 5′ end of the transcript. (D) As (C), but with primers specific for the 3′ end (8.3 kb) of the VPS13 transcript. (E) Recruitment of RNAPII to the indigenous GAL1 promoter was analysed by ChIP and real-time PCR. All samples were also analysed with primers specific for the constitutively expressed FBA1 gene and this result was used as a ‘loading control' to normalize results from the GAL-VPS13 and GAL1 genes. RNAPII level (panels B and E) or VPS13 mRNA level (panels C and D) in uninduced conditions (time point 0 min) was set to 1 and other time points show the fold of difference relative to this uninduced sample.

Cells carrying the GAL-VPS13 gene were grown to mid-log phase in YP-raffinose and induced by adding galactose to the growth medium. Samples were taken at different time points up to 2 h of induction and acetylation of histones was analysed by chromatin immunoprecipitation (ChIP) assays. We also used the ChIP assay to study the kinetics of transcriptional induction by measuring RNA polymerase II (RNAPII) density in the promoter and coding region of the GAL-VPS13 gene. An RNAPII ChIP approach has previously been used to study transcription and it has been shown that the level of RNAPII density in a gene is in good correlation with the rate of its transcription (Kulish and Struhl, 2001; Kristjuhan et al, 2002; Pokholok et al, 2002; Bryant and Ptashne, 2003).

RNAPII was already detectable on the promoter of GAL-VPS13 4 min after adding galactose, but it took about 1 h until recruitment of polymerase reached its maximal level (Figure 1B). The kinetics of TBP recruitment to the GAL-VPS13 promoter was similar to that of RNAPII (data not shown). Accumulation of RNAPII at the end of the gene (8.3 kb from the transcriptional start site, and more than 2 kb from the beginning of the next gene) followed the same kinetics as observed on the promoter, although the rate of increase in density, as well as the final density level, was lower. This may suggest that several of the polymerases loaded onto this long gene never make it to the end. This possibility was further supported by the observation that the level of RNAPII occupancy decreased gradually with increased distance from the transcription start site as measured by ChIP of RNAPII also at 2.2 and 5.5 kb from the transcriptional start site (Figure 1B).

In order to further investigate this hypothesis, we used reverse transcriptase–PCR (RT–PCR) on RNA from formaldehyde crosslinked cells to monitor the production of 5′ and 3′ sequences of the VPS13 transcript during the time course of induction (Figure 1C and D). The overall kinetics of accumulation of the transcript precisely reflected the kinetics of RNAPII appearance on the gene (compare Figure 1B with Figure 1C and D), confirming the reliability of RNAPII recruitment by ChIP as an indicator of ongoing transcription. More importantly, as predicted from the RNAPII recruitment data, a substantial amount of ‘aborted transcription' was indicated also at the mRNA level. At full induction, the amount of 5′ sequences of the VPS13 transcript increased about 120-fold, while full-length mRNA (measured by the accumulation of 3′ sequence of VPS13 transcript) increased only about 30-fold compared to noninduced conditions (compare Y-axis in Figure 1C and D). This suggests that a considerable fraction of polymerases loaded onto the GAL-VPS13 gene never reach the end of the coding region.

We noticed that full induction of the GAL-VPS13 promoter was considerably slower than reported previously for a similar type of induction of the GAL1 promoter (Bryant and Ptashne, 2003). To investigate whether the kinetics of induction of GAL-VPS13 was specific for this long fusion gene or simply representative of GAL activation in our strain background, we also tested the kinetics of induction of the indigenous GAL1 gene (Figure 1E, and data not shown). Analysis of RNAPII ChIP samples with GAL1-specific probes did not reveal any significant difference in kinetics or fold of activation between the endogenous GAL1 gene and the long GAL-VPS13 fusion gene (compare with Figure 1B).

Nucleosomes are lost from GAL-VPS13 and GAL1 loci upon induction

As a consequence of activation-induced chromatin remodelling, the number of nucleosomes in the promoter region of the PHO5 gene is reduced dramatically (Boeger et al, 2003; Reinke and Horz, 2003). Because removal of histones could be a general mechanism to achieve high levels of transcription, we tested whether a loss of nucleosomes was also detectable upon induction of GAL genes. A yeast strain expressing FLAG-tagged histone H2B was induced with galactose and the relative amounts of histones H2B and H3 on the GAL1 locus were determined by ChIP assays using anti-FLAG and anti-H3 C-terminal antibodies, respectively. As observed at the PHO5 promoter (Boeger et al, 2003; Reinke and Horz, 2003), histones lost contact with the GAL1 promoter in response to gene activation (Figure 2A). Surprisingly, histones also lost contact with DNA in the coding region of the GAL1 gene (Figure 2B). Because both H2B and H3 were lost in a coordinated manner, we favour the idea that entire nucleosomes, rather than individual histone dimers, were ‘lost' from the locus. In support of this idea, loss of histone DNA contacts was also observed in the GAL-VPS13 gene where transcriptional activation led to a reduction of H3 levels in the coding region similar to what was found for GAL1 (compare Figure 2D and B). At fully induced conditions (1–2 h after adding galactose), the detectable amount of H3 dropped up to four-fold compared to uninduced conditions, and the loss of H3 signal was in good correlation with the increase of RNAPII transcription through the region (compare Figure 1B with Figure 2D, for example). GAL1 seemed to lose histone H3–DNA contacts to an even higher degree than GAL-VPS13 in the promoter region, although the difference was minor (compare Figure 2A and C). The reason for this difference is unknown, but it could be envisaged that the small intergenic region between the divergently transcribed, highly active GAL1 and GAL10 genes would be even more prone to nucleosome loss than the unidirectional promoter region of the GAL-VPS13 fusion gene.

Figure 2.

Figure 2

Open in a new tab

Loss of histone–DNA contacts at the active GAL1 and GAL-VPS13 loci. The amount of histone H3 (grey bars) and H2B (white bars) on the GAL1 promoter (A) or coding region (B) was determined by ChIP assays with anti-FLAG and anti-H3 C-terminal antibodies, respectively. Co-immunoprecipitated DNA was quantified by real-time PCR and all samples were normalized according to the amount of histones H3 and H2B in a nontranscribed region on chromosome IV (see Materials and methods). The level of histones under uninduced conditions (time point 0 min) was set to 1 and other time points show the amounts of histones relative to the uninduced sample. (C, D) The relative amount of histone H3 at the promoter (C) or in the coding region (D) of GAL-VPS13 was determined by ChIP assays with H3 C-terminal antibody.

These data, combined with those previously reported for the PHO5 promoter (Boeger et al, 2003; Reinke and Horz, 2003), suggest that loss of histone–DNA contact may be a common consequence of gene activation. Significantly, our results also indicate that histones can lose contact with DNA not only in promoters but also in the coding regions of actively transcribed genes.

Transcription-dependent nucleosome loss in the coding region of GAL1

One of the hallmarks of chromatin remodelling at the PHO5 promoter is that it occurs in a transcription-independent manner (Fascher et al, 1993). Similarly, loss of histone–DNA contacts in the GAL-induced genes might conceivably be brought about by a transcription-independent mechanism. To address whether the loss of histone–DNA contacts in the GAL1 gene required transcription, we used a strain expressing a temperature-sensitive allele of the largest RNAPII subunit, Rpb1. At the restrictive temperature, the rpb1-1 strain rapidly ceases transcription due to loss of RNAPII polymerization activity (Nonet et al, 1987), as well as loss of recruitment of the polymerase to promoters (Schroeder et al, 2000). The rpb1-1 strain and its wild-type counterpart were induced at 25 and 37°C, and the dynamics of RNAPII recruitment and loss of H3 was monitored by ChIP assays. As shown in Figure 3A, both wild-type and rpb1-1 cells activated GAL1 transcription at 25°C, but only wild-type cells were able to do so at 37°C. Importantly, the loss of histone–DNA contacts was dependent on active RNAPII transcription, because only wild-type but not the rpb1-1 strain lost histone H3 at 37°C in the GAL1 coding region (Figure 3B; similar data for the promoter are not shown). These data indicate that the high level of transcription at the GAL1 gene displaces nucleosomes, or at least leads to a substantial loss of histone–DNA contacts.

Figure 3.

Figure 3

Open in a new tab

Loss of histone–DNA contacts in the coding region of GAL1 requires active transcription. Wild-type (WT) and rpb1-1 strains were induced at 25 and 37°C and the kinetics of RNAPII recruitment (A) or loss of histone H3 (B) were determined by ChIP before (0 min) and after (15 and 90 min) galactose induction. The results were normalized as described for Figure 2.

Acetylation of histones upon galactose induction of the GAL-VPS13 gene

Transcription of genes is facilitated by protein complexes that covalently modify chromatin. HATs are the best characterized of such factors. In order to evaluate the role of histone acetylation in transcription of the long fusion gene, we analysed the status of H3 and H4 acetylation at GAL-VPS13 during induction (Figure 4). Because of the observed loss of histone–DNA contacts, the results from ChIP assays with acetyl-H3- and acetyl-H4-specific antibodies were normalized according to the total level of H3 present at the corresponding time point (see Figure 2C and D). Upon induction of transcription, acetylation of histone H3, but not H4, increased at the promoter of GAL-VPS13 throughout the time course and reached a maximal level (three- to four-fold above uninduced conditions) within 15 min of induction (Figure 4A).

Figure 4.

Figure 4

Open in a new tab

Histone H3 and H4 acetylation at the GAL-VPS13 locus during activation. The amount of acetyl-H3 (grey bars) and acetyl-H4 (white bars) at the promoter (A) or in the coding region (B) of the GAL-VPS13 gene was determined by ChIP. The results were first normalized according to the amount of acetyl-H3 or acetyl-H4 in a nontranscribed region on chromosome IV, and then normalized to reflect the number of nucleosomes present at the corresponding position of the GAL-VPS13 gene at the relevant time point of induction (adjustment for the loss of nucleosomes).

Surprisingly, we did not observe any significant increase in histone acetylation in the coding region of GAL-VPS13 during gene induction (Figure 4B). This indicates that H3 and H4 are either not acetylated in a transcription-dependent manner in the coding region, or that if they are, such acetylation is so transient that we cannot detect it by the ChIP assay. In the latter case, one might assume that inactivation of HDACs would help to preserve a putative hyperacetylated state of histones. To address this possibility, we used HDAC inhibitors (TSA and sodium butyrate) but failed to observe an effect on histone acetylation levels (data not shown). We also analysed histone acetylation in cells lacking specific HDACs (using rpd3, hos2, and rpd3 hos2 strains). Although the steady-state levels of H3 and H4 acetylation in these strains were generally higher than in wild-type cells, we were still unable to detect any transcription-dependent increase in H3 or H4 acetylation in the coding region of the GAL-VPS13 gene (data not shown).

To reveal the HAT(s) responsible for H3 acetylation in the GAL-VPS13 promoter, we constructed a strain carrying the inducible GAL-VPS13 gene in combination with deletion of the genes encoding two transcription-related HATs, GCN5 and ELP3. Gcn5 is the catalytic subunit of SAGA, a major transcription-related HAT complex in yeast (Grant et al, 1997), while Elp3 is the catalytic subunit of Elongator, a HAT complex originally isolated with hyperphosphorylated, elongating RNAPII (Otero et al, 1999). Histone H3 is the primary target for both of these complexes and significant redundancy in their function has been demonstrated (Grant et al, 1997; Wittschieben et al, 2000; Kristjuhan et al, 2002, 2003; Winkler et al, 2002). Induction of GAL-VPS13 in elp3, gcn5, and elp3 gcn5 strains showed nearly wild-type levels of RNAPII recruitment both to the promoter and to the end of coding region (Figure 5A), with only a slight delay in the kinetics observed in the mutant strains, indicating no dramatic requirement for these factors for the induction of the gene. The loss of nucleosomes also occurred to a similar extent both in wild-type and the mutant strains (Figure 5B). As significant histone H4 acetylation was not detected in the locus in wild-type cells, and the gcn5 and elp3 gcn5 strains were deficient for H3 acetylation in the promoter of the gene (Figure 5C), the loss of histone–DNA contacts detected in this region may have occurred in an acetylation-independent manner. Another possibility was that the lack of GCN5 and hence lack of H3 acetylation resulted in compensatory acetylation of H4. However, no evidence for increased histone H4 acetylation at the gene in gcn5 or gcn5 elp3 cells was obtained (data not shown).

Figure 5.

Figure 5

Open in a new tab

Induction of GAL-VPS13 transcription in elp3, gcn5, and elp3 gcn5. (A) Recruitment of RNAPII to the promoter (pr.) and the end of the coding region (8.3 kb) of GAL-VPS13 was analysed in WT, elp3, gcn5, and elp3 gcn5 cells after 0, 15, and 90 min of induction. (B) As in (A), but histone H3. (C) Acetylation of histone H3 at the GAL-VPS13 locus in WT, elp3, gcn5, and elp3 gcn5 strains was determined at 0, 15, and 90 min time points of galactose induction. The results were normalized as described for Figures 2 and 4.

Taken together, these data show that Gcn5 is the major HAT involved in acetylation of H3 in the promoter of the GAL-VPS13 gene. Nevertheless, Gcn5 HAT activity (and the increase of promoter H3 acetylation in general) is apparently not important for efficient transcription of the gene, or for the observed loss of histone–DNA contacts.

Transcription-dependent changes in histone–DNA contacts and histone acetylation at GCN5/ELP3-dependent genes

The experiments presented above point to the removal of nucleosomes as a mechanism to achieve high levels of transcription. These data raised the question whether other genes, such as the majority of constitutively expressed genes, might also be subject to this type of chromatin reorganization as a consequence of transcription. Relevant in this regard, we have previously shown that the global level of histone H3 acetylation is extremely low in elp3 gcn5, and that expression of several genes is dramatically reduced in this strain. Most significantly, the expression levels of dramatically downregulated genes correlate very well with the level of H3 acetylation in their coding regions (Kristjuhan et al, 2002). The mechanism allowing transcription of these genes, in contrast to the GAL-induced genes, might thus conceivably involve acetylation of resident histones rather than, or in addition to, nucleosome eviction.

To test these possibilities experimentally, we first investigated whether the low level of expression of several genes in the elp3 gcn5 strain resulted in a higher nucleosome density in the coding region of these genes in the mutant compared to wild type. The amounts of RNAPII and nucleosomes in the coding region of BAT1, GDH1, and YGP1 in wild-type and elp3 gcn5 cells were compared (Figure 6). As reported previously (Kristjuhan et al, 2002), there was a significantly reduced level of transcribing RNAPII in these genes in elp3 gcn5 cells compared to the wild-type strain (Figure 6; ‘RNAPII'). In contrast, ChIP analysis of the histone H3 levels indicated that the nucleosome density in the middle of the coding region remained the same (Figure 6; ‘Histone H3'). Two conclusions can be drawn from this result. Firstly, nucleosome levels in active genes are not necessarily directly correlated with the level of transcriptional activity, that is, transcription does not necessarily lead to loss of histone–DNA contacts. And secondly, the reduced gene expression observed in gcn5 elp3 cannot be caused by an abnormally high density of nucleosomes in the coding region of the tested genes.

Figure 6.

Figure 6

Open in a new tab

Nucleosome density at BAT1, GDH1, and YPG1 is unaffected by gcn5 elp3 mutation. The amount of RNAPII and histone H3 in the middle of the coding region of BAT1, GDH1, and YGP1 in WT and gcn5 elp3 cells was determined by ChIP. The graphs represent the relative levels of RNAPII and histone H3 in the gcn5 elp3 strain with the corresponding values in the WT strain set as 1 (stippled line). Co-immunoprecipitated DNA was quantified by real-time PCR and all samples were normalized according to the amount of RNAPII and histone H3 in a nontranscribed region on chromosome IV (see Materials and methods).

To further investigate nucleosome density in the coding region of the GCN5/ELP3-dependent constitutive genes, we also made deletions of the BAT1, GDH1, and YGP1 promoters to reduce their transcription levels. Quantification of the corresponding transcripts confirmed that the production of mRNA from these genes was effectively shut down by deletion of the respective promoters (Figure 7A). As can be seen in Figure 7B, however, the level of histone H3–DNA contacts did not change significantly in the coding region of these genes in response to the abrogation of promoter-directed transcription.

Figure 7.

Figure 7

Open in a new tab

Transcription-associated histone H3 acetylation, but little nucleosome loss at the GCN5/ELP3-dependent genes. (A) The level of BAT1, GDH1, and YGP1 mRNAs as determined by real-time RT–PCR in strains where the promoters of the indicated genes had been deleted. mRNA level in the wild-type strain was set as 1 (stippled line). Note that these genes are expressed at different levels compared to each other in wild-type cells (BAT1=15 transcripts/cell; GDH1=10 transcripts/cell; YGP1=6 transcripts/cell) (Velculescu et al, 1997). (B) The amount of histone H3 at BAT1, GDH1, and YGP1 in wild-type and promoter deletion strains was determined by ChIP. The level of histone H3 in the wild-type strain was set as 1 (stippled line). (C) The amount of histone H3 in the middle of the coding region of BAT1, GDH1, and YGP1 in wild type and rpb1-1 at 25 and 37°C was determined by ChIP. The fraction of histone H3 observed in the absence of transcription compared to in the presence of transcription (37°C/25°C ratio) is shown. No significant change occurred after temperature shift in the wild-type cells, and therefore the ratio observed in this strain was 1 (stippled line). (D) The amount of histone H3 acetylation in the middle of the coding region of BAT1, GDH1, and YGP1 in wild type and rpb1-1 at 25 and 37°C was determined by ChIP. The fraction of histone H3 acetylation in the absence of transcription compared to in the presence of transcription (37°C/25°C ratio) is shown, with the ratio obtained in the wild-type strain set as 1 as in (C) (stippled line). The results were normalized as described for Figures 2 and 4.

Finally, we also investigated chromatin structure in the genes after abrogation of transcription via the temperature-sensitive rpb1-1 allele, allowing measurements of nucleosome density and modification levels without altering the primary sequence of the DNA in the region. This approach also minimized the effect of transcription/modification at neighbouring genes. This complete transcriptional shut-off resulted in a slight but repeatedly observed increase in histone–DNA contacts over each of the three genes (Figure 7C), suggesting that some loss of histone–DNA contact may be a more widespread consequence of ongoing transcription. Significantly, cessation of transcription correlated with a more than three-fold decrease in histone H3 acetylation at YGP1, and approximately two-fold decreases at BAT1 and GDH1 (Figure 7D), indicating that transcription-associated histone acetylation indeed occurs in the coding region of these genes.

Discussion

In this study, we followed the dynamics of transcription and chromatin modification during the time course of gene activation. The key results were the following. Firstly, by comparing ‘snapshots' of RNAPII density and mRNA production in the beginning and end of a long test gene, we found that a significant proportion (at least half) of RNAPII molecules initiating transcription never make it to the end of this gene. Secondly, nucleosomes not only in the promoter but also in the coding region of GAL genes lose contact with DNA in the course of gene induction, and this ‘nucleosome loss' requires transcription. Thirdly, histone H3 rapidly becomes hyperacetylated in the GAL promoter, but not the coding region, as a result of activation. In contrast, no evidence of histone H4 acetylation was obtained. Fourthly, histone H3 promoter acetylation was dependent on GCN5. Nucleosome loss was observed also in cells lacking GCN5 (and even in gcn5 elp3), suggesting that the loss of histone–DNA contacts in GAL genes is independent of H3 acetylation. Finally, histone loss and modification in response to transcription was studied in genes previously shown to require Gcn5 and Elp3 for normal transcription. Interestingly, these genes did not dramatically lose histone–DNA contacts in response to transcription, but the histones were acetylated in a transcription-dependent manner. Together, these data suggest that efficient transcription through chromatin can occur by distinct mechanisms.

Time course of transcriptional activation in a GAL-VPS13 fusion gene

The use of a GAL-VPS13 fusion gene made it possible to clearly distinguish between transcription-coupled events driven by transcriptional initiation and elongation. While this study was in progress, results obtained by Struhl and colleagues with a similar fusion gene were reported (Strasser et al, 2002; Mason and Struhl, 2003). Although the speed of GAL activation and repression is known to occur with different speeds in different strain backgrounds, it was somewhat surprising that activation of the GAL-VPS13 fusion gene in our strains occurred in a more or less linear manner over a period of more than 1 h. This is in contrast to results reported by Bryant and Ptashne (2003), who found that recruitment of RNAPII to the GAL1 gene in their strains reached plateau levels already 15–18 min after galactose addition. The reason for the slower activation kinetics observed in our strain background (the same kinetics were also observed at the indigenous GAL1 gene) is unknown and unfortunately precluded rigorous measurements of the speed of elongation through the long VPS13 coding region.

However, our experiments with the fusion gene yielded the interesting result that the level of polymerase detected at the promoter was significantly higher than that detected further into the gene (at 2.2, 5.5, and 8.8 kb), suggesting either that RNAPII–DNA crosslinking efficiency is higher at the promoter than for elongating polymerases further into the gene, or that a substantial fraction of polymerases loaded onto the gene never make it to the end, presumably because they abort transcription prematurely. Quantitative RT–PCR measurements of RNA produced from the locus showed that up to four-fold more transcript RNA was produced in the 5′ end of the gene than at the 3′ end, supporting the idea that substantial abortive transcription occurred at this locus. This idea was further supported by the observation that substantial smearing below the full-length VPS13 transcript was detected by Northern blot (data not shown). The idea that abortive transcription (in the beginning of the gene prior to promoter clearance) is a natural part of the transcription process has long been generally accepted based on in vitro experiments (Dvir, 2002), but data from living cells to support it have been lacking. Our data clearly indicate that even transcripts of substantial length can be aborted during transcription of certain genes. The process underlying such transcriptional abortion is unknown but is of great relevance to an understanding of basic transcription processes and their regulation.

Loss of histone–DNA contacts during activation

The chromatin transactions taking place at the yeast PHO and GAL genes have been the subject of several studies over the years (Almer and Horz, 1986; Almer et al, 1986; Cavalli and Thoma, 1993; Fascher et al, 1993). However, only with the availability of techniques such as ChIP has it become possible to make judgment on the fate of nucleosomes in the promoter of PHO5 and PHO8, for example (Boeger et al, 2003; Reinke and Horz, 2003; Adkins et al, 2004; Boeger et al, 2004). Thus, traditional tools for studying chromatin structure, such as nuclease digestion of chromatin in isolated nuclei, are capable of assessing whether chromatin structure has been perturbed, but not of distinguishing between loss of nucleosome positioning and loss of histone–DNA contacts. Indeed, others have reported substantial changes in chromatin structure at the GAL1-10 locus after gene activation, as assayed by micrococcal nuclease digestion (Li and Smerdon, 2002). Our ChIP data indicate that these dramatic changes are due to histones losing contact with DNA in both the promoter and coding region of the genes. Interestingly, this is in apparent contrast to PHO5 and PHO8 where nucleosomes in the promoter, but not the coding region, are lost (Adkins et al, 2004). Moreover, the loss of histone–DNA contacts we observed is dependent on active transcription. This also appears to be in contrast to the dramatic change in chromatin structure in the promoter of PHO5, which occurs even when the TATA box of the gene is removed so that transcription is abolished (Fascher et al, 1993). In support of our data, loss of histone–DNA contacts in the coding region of GAL genes has also been observed by Bentley and co-workers (L Zhang and D Bentley, personal communication).

Cotranscriptional changes in histone acetylation

Studying the pattern of histone acetylation during activation of GAL-VPS13 indicated that histone H3, but not H4, is hyperacetylated in the promoter in response to transcriptional induction. Deletion of the transcription-related HATs ELP3 and GCN5 revealed that Gcn5 is the major HAT responsible for this acetylation. Fully induced levels of H3 acetylation in the promoter were achieved within 15 min, whereas final levels of RNAPII recruited were achieved much later, suggesting that recruitment of SAGA (Gcn5) and of RNAPII were independent events, as previously observed by others (reviewed by Cosma, 2002). Histone H3 acetylation in the promoter was not crucial for activation, however, as gcn5 cells showed wild-type levels of gene activation despite a lack of increased promoter acetylation.

In contrast to the acetylation pattern in promoter regions, we did not detect any significant increase of H3 or H4 acetylation in the coding region of the genes. This result was surprising, because several observations predict a role for histone acetylation and deacetylation activities in the coding regions of the GAL genes during transcription. For example, the Elongator HAT complex and the histone deacetylase Hos2 are indeed present in the coding region of GAL1 during activation (Wang et al, 2002; Gilbert et al, 2004), indicating a role for HATs and HDACs in transcript elongation. It is important to note, however, that although we were unable to detect changes in the level of histone acetylation in the coding region during activation, it is possible that these modifications are very dynamic and that hyperacetylation of histones occurs only transiently in coding regions (Svejstrup, 2003).

Genes do not invariably lose nucleosomes in a transcription-dependent manner

We previously reported that several genes are severely downregulated in gcn5 elp3 cells and that reduced transcription of these genes correlates with low histone H3 acetylation in their coding region (Kristjuhan et al, 2002). We were therefore somewhat surprised that activation of the long GAL-VPS13 fusion gene showed no requirement for GCN5/ELP3. The same lack of GCN5/ELP3 requirement was observed for the GAL1 gene (data not shown). In an attempt to shed light on potential differences in basic mechanisms underlying transcription through chromatin, we also studied transcription-dependent nucleosome density and acetylation in the GCN5/ELP3-dependent BAT1, GDH1, and YGP1 genes. Because these genes are constitutively expressed at different levels, we compared their histone H3 density under normal conditions (where it would be presumed to be low if transcription generally leads to loss of histone–DNA contacts) to conditions where transcription was significantly reduced (by gcn5 elp3 mutation, deletion of the TATA box or inactivation of RNAPII). Remarkably, none of these genes appear to lose nucleosomes as a consequence of transcription. However, they are acetylated in a transcription-dependent manner.

These data point to the existence of distinct mechanisms for enabling transcription through coding regions packed in nucleosomes. In the first, histones lose contact with DNA, and in the other, little or no net loss of histone–DNA contact occurs. These mechanisms correlate with transcription levels and histone H3 acetylation in a potentially telling manner: firstly, transcription levels at the GAL genes are very high, with a new polymerase initiating transcription approximately every 6–8 s (Iyer and Struhl, 1996). This in all likelihood results in a large number of polymerases on the coding region at the same time (theoretically one every 150–200 bp, if the rate of transcription is presumed to be 1500 bases/min). Indeed, RNAPII ChIP–reChIP–reChIP experiments with strains concomitantly expressing three different, distinctly tagged versions of Rpb3 indicate that at least three polymerases can be present per 300–600 bp GAL coding region fragment under fully induced conditions (A Kristjuhan, unpublished). Such high levels of transcription/RNAPII density might thus not permit efficient re-establishment of histone–DNA contacts after polymerase passage, resulting in a net loss of nucleosomes.

However, transcription of the majority of genes occurs with a much lower density of transcribing polymerases, and transcription of such genes might thus occur without substantial loss of nucleosomes. Our data suggest that at least the GCN5/ELP3-dependent BAT1, GDH1, and YGP1 genes fall into this category. The fact that these genes in contrast to GAL1 and GAL-VPS13 are acetylated in a transcription-dependent manner is in agreement with the idea that when nucleosome eviction does not occur, histone acetylation can become very important for transcription through chromatin.

Materials and methods

Yeast strains

All S. cerevisiae strains were congenic with strain W303 (Thomas and Rothstein, 1989). To create a GAL-VPS13 fusion gene at the VPS13 locus in chromosome XII, the GAL10 promoter (500 bp fragment upstream from the GAL10 ORF) was inserted in front of the VPS13 gene by homologous recombination in vivo. The resulting GAL-VPS13 locus contains all the regulatory elements of the GAL10 promoter (the Gal4 binding sites, TATA box, transcriptional start site, and the 5′ noncoding stretch of mRNA) followed by the 9435 bp long ORF of VPS13 (Figure 1A). Growth of yeast strains is neither affected by the loss of VPS13 function nor by its overexpression. The genotypes of the strains were as follows: MATa GAL-VPS13∷TRP1 TBP-3xHA∷URA3 (JSY 679), MATa GAL-VPS13∷TRP1 TBP-3xHA∷URA3 elp3Δ∷LEU2 (JSY 681), MATa GAL-VPS13∷TRP1 TBP-3xHA∷URA3 gcn5Δ∷HIS3 (JSY 683), MATa GAL-VPS13∷TRP1 TBP-3xHA∷URA3 elp3Δ∷LEU2 gcn5Δ∷HIS3 (JSY 685), MATα GAL-VPS13∷TRP1 bat1Δpromoter∷URA3 (JSY1008), MATα GAL-VPS13∷TRP1 gdh1Δpromoter∷URA3 (JSY1009), and MATα GAL-VPS13∷TRP1 ygp1Δpromoter∷URA3 (JSY1010). The JR5-2A strain (W303, but htb1-1 htb2-1), expressing a plasmid-borne copy of FLAG-tagged H2B protein (Robzyk et al, 2000), was kindly provided by Dr Mary Ann Osley, while the strain carrying an rpb1-1 mutation and its wild-type counterpart (Nonet et al, 1987) was kindly provided by Dr Richard Young.

Galactose induction of transcription

Cells were grown for 12–15 h in yeast extract-peptone (YP) medium containing 2% raffinose as a carbon source. Before induction with galactose, 25 ml of mid-log phase cells (density 0.5–1 × 107 cells/ml) were fixed in 1% formaldehyde as a 0 min time-point sample. Cells in mid-log phase were induced in a large batch (200 ml) by adding galactose to the growth medium (final concentration 2%). At different time points after adding galactose, 25 ml aliquots were taken and fixed in 1% formaldehyde for ChIP assays.

Cells carrying the rpb1-1 mutation were grown in YP-raffinose at 25°C to mid-log phase and shifted to 37°C for 30 min before galactose induction.

Chromatin immunoprecipitation assay and RT–PCR

ChIP assays were performed essentially as described previously (Kristjuhan et al, 2002). Antibodies directed against RNAPII (4H8), diacetyl-H3, and tetraacetyl-H4 were purchased from Upstate, anti-FLAG M2 antibody was purchased from Sigma, and antibody recognizing the C-terminus of histone H3 was a generous gift from Dr Alain Verreault. Co-precipitated DNA was analysed by quantitative real-time PCR using the ABI Prism® 7000 Sequence Detection System in standard conditions (40 cycles; 95°C 15 s+60°C 1 min). ABsolute™ QPCR reagents (ABgene) were used in combination with FAM-labelled TaqMan® fluorogenic quantitation probes to detect sequences in the promoter or in the coding regions of the GAL-VPS13 gene. Sequences of the GAL1 gene were analysed by real-time PCR with ABsolute™ QPCR SYBR Green reagents (ABgene) without fluorescent probe. GAL-VPS13 promoter-specific PCR primers amplify the sequence about 150 bp downstream from the TATA box and contain sequences unique to the VPS13 gene in order to distinguish the GAL-VPS13 promoter from the GAL10 promoter. The coding region primers are specific for the sequences at 2.2, 5.5, and 8.3 downstream from the promoter of GAL-VPS13, and 1.2 kb downstream from the promoter of GAL1 genes. Two control regions in the genome were used to compensate for possible fluctuations arising either from different amounts of starting material or possible loss of material during handling of samples. RNAPII ChIP of the ubiquitously and highly expressed FBA1 gene was used for normalization of all RNAPII ChIPs. A probe in the middle of a 3.2 kb long ORF-free region on chromosome IV (between the NRG1 and HEM13 genes) was used to normalize all histone-related ChIPs. Because we observed a loss of histone H3 signal from GAL-VPS13 and GAL1 genes upon galactose induction, we also normalized all acetyl-histone-specific ChIP results (shown in Figures 4, 5, and 7) according to the total H3 signal (as detected with C-terminus-specific antibody) at the genes. Experiments were typically repeated 3–4 times; error bars in figures show standard deviations.

For detection of mRNA, cells were treated as described for RNA-immunoprecipitation assays (Gilbert et al, 2004). The amount of VPS13 and FBA1 mRNAs was then quantified by one-step real-time RT–PCR with ABsolute™ QPCR SYBR Green reagents (ABgene).

The sequences of all primers and fluorogenic probe oligos used in this study are available upon request.

Acknowledgments

We thank Drs Mary Ann Osley and Richard Young for their kind gifts of yeast strains, and Dr Alain Verreault for his kind gift of antibodies. We thank Chris Gilbert for technical assistance. This project was supported by grants from Cancer Research UK and the European Union (HPRN-CT-2000-00087) to JQS.

References

  1. Adkins MW, Howar SR, Tyler JK (2004) Chromatin disassembly mediated by the histone chaperone Asf1 is essential for transcriptional activation of the yeast PHO5 and PHO8 genes. Mol Cell 14: 657–666 [DOI] [PubMed] [Google Scholar]
  2. Almer A, Horz W (1986) Nuclease hypersensitive regions with adjacent positioned nucleosomes mark the gene boundaries of the PHO5/PHO3 locus in yeast. EMBO J 5: 2681–2687 [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Almer A, Rudolph H, Hinnen A, Horz W (1986) Removal of positioned nucleosomes from the yeast PHO5 promoter upon PHO5 induction releases additional upstream activating DNA elements. EMBO J 5: 2689–2696 [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Belotserkovskaya R, Oh S, Bondarenko VA, Orphanides G, Studitsky VM, Reinberg D (2003) FACT facilitates transcription-dependent nucleosome alteration. Science 301: 1090–1093 [DOI] [PubMed] [Google Scholar]
  5. Boeger H, Griesenbeck J, Strattan JS, Kornberg RD (2003) Nucleosomes unfold completely at a transcriptionally active promoter. Mol Cell 11: 1587–1598 [DOI] [PubMed] [Google Scholar]
  6. Boeger H, Griesenbeck J, Strattan JS, Kornberg RD (2004) Removal of promoter nucleosomes by disassembly rather than sliding in vivo. Mol Cell 14: 667–673 [DOI] [PubMed] [Google Scholar]
  7. Bryant GO, Ptashne M (2003) Independent recruitment in vivo by Gal4 of two complexes required for transcription. Mol Cell 11: 1301–1309 [DOI] [PubMed] [Google Scholar]
  8. Cavalli G, Thoma F (1993) Chromatin transitions during activation and repression of galactose-regulated genes in yeast. EMBO J 12: 4603–4613 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cherry JM, Ball C, Weng S, Juvik G, Schmidt R, Adler C, Dunn B, Dwight S, Riles L, Mortimer RK, Botstein D (1997) Genetic and physical maps of Saccharomyces cerevisiae. Nature 387: 67–73 [PMC free article] [PubMed] [Google Scholar]
  10. Cosma MP (2002) Ordered recruitment: gene-specific mechanism of transcription activation. Mol Cell 10: 227–236 [DOI] [PubMed] [Google Scholar]
  11. Deckert J, Struhl K (2001) Histone acetylation at promoters is differentially affected by specific activators and repressors. Mol Cell Biol 21: 2726–2735 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Dvir A (2002) Promoter escape by RNA polymerase II. Biochim Biophys Acta 1577: 208–223 [DOI] [PubMed] [Google Scholar]
  13. Fascher KD, Schmitz J, Horz W (1993) Structural and functional requirements for the chromatin transition at the PHO5 promoter in Saccharomyces cerevisiae upon PHO5 activation. J Mol Biol 231: 658–667 [DOI] [PubMed] [Google Scholar]
  14. Gilbert C, Kristjuhan A, Winkler GS, Svejstrup JQ (2004) Elongator interactions with nascent mRNA revealed by RNA immunoprecipitation. Mol Cell 14: 457–464 [DOI] [PubMed] [Google Scholar]
  15. Grant PA, Duggan L, Cote J, Roberts SM, Brownell JE, Candau R, Ohba R, Owen-Hughes T, Allis CD, Winston F, Berger SL, Workman JL (1997) Yeast Gcn5 functions in two multisubunit complexes to acetylate nucleosomal histones: characterization of an Ada complex and the SAGA (Spt/Ada) complex. Genes Dev 11: 1640–1650 [DOI] [PubMed] [Google Scholar]
  16. Hassan AH, Neely KE, Vignali M, Reese JC, Workman JL (2001a) Promoter targeting of chromatin-modifying complexes. Front Biosci 6: D1054–D1064 [DOI] [PubMed] [Google Scholar]
  17. Hassan AH, Neely KE, Workman JL (2001b) Histone acetyltransferase complexes stabilize swi/snf binding to promoter nucleosomes. Cell 104: 817–827 [DOI] [PubMed] [Google Scholar]
  18. Iyer V, Struhl K (1996) Absolute mRNA levels and transcriptional initiation rates in Saccharomyces cerevisiae. Proc Natl Acad Sci USA 93: 5208–5212 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Kadam S, Emerson BM (2002) Mechanisms of chromatin assembly and transcription. Curr Opin Cell Biol 14: 262–268 [DOI] [PubMed] [Google Scholar]
  20. Kaplan CD, Laprade L, Winston F (2003) Transcription elongation factors repress transcription initiation from cryptic sites. Science 301: 1096–1099 [DOI] [PubMed] [Google Scholar]
  21. Katan-Khaykovich Y, Struhl K (2002) Dynamics of global histone acetylation and deacetylation in vivo: rapid restoration of normal histone acetylation status upon removal of activators and repressors. Genes Dev 16: 743–752 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Kristjuhan A, Walker J, Suka N, Grunstein M, Roberts D, Cairns BR, Svejstrup JQ (2002) Transcriptional inhibition of genes with severe histone h3 hypoacetylation in the coding region. Mol Cell 10: 925–933 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kristjuhan A, Wittschieben BO, Walker J, Roberts D, Cairns BR, Svejstrup JQ (2003) Spreading of Sir3 protein in cells with severe histone H3 hypoacetylation. Proc Natl Acad Sci USA 100: 7551–7556 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Kulish D, Struhl K (2001) TFIIS enhances transcriptional elongation through an artificial arrest site in vivo. Mol Cell Biol 21: 4162–4168 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Kurdistani SK, Tavazoie S, Grunstein M (2004) Mapping global histone acetylation patterns to gene expression. Cell 117: 721–733 [DOI] [PubMed] [Google Scholar]
  26. Li S, Smerdon MJ (2002) Nucleosome structure and repair of N-methylpurines in the GAL1-10 genes of Saccharomyces cerevisiae. J Biol Chem 277: 44651–44659 [DOI] [PubMed] [Google Scholar]
  27. Mason PB, Struhl K (2003) The FACT complex travels with elongating RNA polymerase II and is important for the fidelity of transcriptional initiation in vivo. Mol Cell Biol 23: 8323–8333 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Narlikar GJ, Fan HY, Kingston RE (2002) Cooperation between complexes that regulate chromatin structure and transcription. Cell 108: 475–487 [DOI] [PubMed] [Google Scholar]
  29. Nonet M, Scafe C, Sexton J, Young R (1987) Eucaryotic RNA polymerase conditional mutant that rapidly ceases mRNA synthesis. Mol Cell Biol 7: 1602–1611 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Otero G, Fellows J, Li Y, de Bizemont T, Dirac AMG, Gustafsson CM, Erdjument-Bromage H, Tempst P, Svejstrup JQ (1999) Elongator, a multisubunit component of a novel RNA polymerase II holoenzyme for transcriptional elongation. Mol Cell 3: 109–118 [DOI] [PubMed] [Google Scholar]
  31. Pokholok DK, Hannett NM, Young RA (2002) Exchange of RNA polymerase II initiation and elongation factors during gene expression in vivo. Mol Cell 9: 799–809 [DOI] [PubMed] [Google Scholar]
  32. Reinke H, Gregory PD, Horz W (2001) A transient histone hyperacetylation signal marks nucleosomes for remodeling at the PHO8 promoter in vivo. Mol Cell 7: 529–538 [DOI] [PubMed] [Google Scholar]
  33. Reinke H, Horz W (2003) Histones are first hyperacetylated and then lose contact with the activated PHO5 promoter. Mol Cell 11: 1599–1607 [DOI] [PubMed] [Google Scholar]
  34. Robzyk K, Recht J, Osley MA (2000) Rad6-dependent ubiquitination of histone H2B in yeast. Science 287: 501–504 [DOI] [PubMed] [Google Scholar]
  35. Schroeder SC, Schwer B, Shuman S, Bentley D (2000) Dynamic association of capping enzymes with transcribing RNA polymerase II. Genes Dev 14: 2435–2440 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Strasser K, Masuda S, Mason P, Pfannstiel J, Oppizzi M, Rodriguez-Navarro S, Rondon AG, Aguilera A, Struhl K, Reed R, Hurt E (2002) TREX is a conserved complex coupling transcription with messenger RNA export. Nature 417: 304–308 [DOI] [PubMed] [Google Scholar]
  37. Svejstrup JQ (2003) Transcription. Histones face the FACT. Science 301: 1053–1055 [DOI] [PubMed] [Google Scholar]
  38. Thomas BJ, Rothstein R (1989) Elevated recombination rates in transcriptionally active DNA. Cell 56: 619–630 [DOI] [PubMed] [Google Scholar]
  39. Velculescu VE, Zhang L, Zhou W, Vogelstein J, Basrai MA, Bassett DE Jr, Hieter P, Vogelstein B, Kinzler KW (1997) Characterization of the yeast transcriptome. Cell 88: 243–251 [DOI] [PubMed] [Google Scholar]
  40. Vogelauer M, Wu J, Suka N, Grunstein M (2000) Global histone acetylation and deacetylation in yeast. Nature 408: 495–498 [DOI] [PubMed] [Google Scholar]
  41. Wang A, Kurdistani SK, Grunstein M (2002) Requirement of Hos2 histone deacetylase for gene activity in yeast. Science 298: 1412–1414 [DOI] [PubMed] [Google Scholar]
  42. Winkler GS, Kristjuhan A, Erdjument-Bromage H, Tempst P, Svejstrup JQ (2002) Elongator is a histone H3 and H4 acetyltransferase important for normal histone acetylation levels in vivo. Proc Natl Acad Sci USA 99: 3517–3522 [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Wittschieben BO, Fellows J, Du W, Stillman DJ, Svejstrup JQ (2000) Overlapping roles for the histone acetyltransferase activities of SAGA and Elongator in vivo. EMBO J 19: 3060–3068 [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Workman JL, Kingston RE (1998) Alteration of nucleosome structure as a mechanism of transcriptional regulation. Annu Rev Biochem 67: 545–579 [DOI] [PubMed] [Google Scholar]

Articles from The EMBO Journal are provided here courtesy of Nature Publishing Group

RESOURCES