The nucleosome remodeling complex, Snf/Swi, is required for the maintenance of transcription in vivo and is partially redundant with the histone acetyltransferase, Gcn5

Priya Sudarsanam, Yixue Cao, Lena Wu, Brehon C. Laurent, Fred Winston

Author Affiliations

  1. Priya Sudarsanam1,
  2. Yixue Cao2,
  3. Lena Wu3,
  4. Brehon C. Laurent2 and
  5. Fred Winston*,1
  1. 1 Department of Genetics, Harvard Medical School, Boston 200 Longwood Avenue, MA, 02115, USA
  2. 2 Department of Microbiology and Immunology, State University of New York, Brooklyn, NY, 11203, USA
  3. 3 Present address: Millennium Pharmaceuticals, Cambridge, MA, 02139, USA
  1. *Corresponding author. E-mail: winston{at}
View Full Text


Snf/Swi, a nucleosome remodeling complex, is important for overcoming nucleosome‐mediated repression of transcription in Saccharomyces cerevisiae. We have addressed the mechanism by which Snf/Swi controls transcription in vivo of an Snf/Swi‐dependent promoter, that of the SUC2 gene. By single‐cell analysis, our results show that Snf/Swi is required for activated levels of SUC2 expression in every cell of a population. In addition, Snf/Swi is required for maintenance of SUC2 transcription, suggesting that continuous chromatin remodeling is necessary to maintain an active transcriptional state. Finally, Snf/Swi and Gcn5, a histone acetyltransferase, have partially redundant roles in the control of SUC2 transcription, suggesting a functional overlap between two different mechanisms believed to overcome repression by nucleosomes, nucleosome remodeling and histone acetylation.


The mechanism by which transcriptional activators and enhancers facilitate transcription from their target promoters in vivo remains unclear. Two models have been proposed for how activators and enhancers control transcription—gradient and binary (Walters et al., 1996). In the gradient model, activators control the level of transcription from each promoter template. In the binary model, activators increase the probability that a promoter will be active rather than controlling the level of transcription from each activated promoter. A large body of evidence has suggested that most activators affect transcription in a gradient manner (Lewin, 1990). These studies have used conventional techniques to study populations of cells. However, most single‐cell assays have suggested that mammalian activators may control transcription in a binary manner (Weintraub, 1988; Fiering et al., 1990; Walters et al., 1995, 1996; Bagga and Emerson, 1997). For example, studies have shown that some mammalian enhancers increase the number of active promoters rather than the level of expression from each promoter. In addition, these results have suggested that transcription can be maintained in a stable manner in the absence of an enhancer, albeit in a small number of cells (Weintraub, 1988; Fiering et al., 1990; Walters et al., 1995, 1996; Bagga and Emerson, 1997). In contrast, there has been a recent report of a graded transcriptional response in a eukaryotic system (Kringstein et al., 1998). In this study, the expression of a reporter gene increased in a gradient manner, proportional to increasing concentrations of the transactivator, in every cell of the population. Taken together, these results suggest that neither a binary nor a gradient response is a universal phenomenon in transcriptional activation and different mechanisms may operate depending on the particular enhancers and activators.

In our work, we have established an in vivo system at the endogenous SUC2 locus in Saccharomyces cerevisiae to distinguish between the binary and gradient models for transcriptional activation by the nucleosome remodeling complex Snf/Swi. Snf/Swi is conserved throughout eukaryotes (Kingston et al., 1996) and, in S.cerevisiae, it is required for the transcription of a diverse set of genes, including SUC2, HO, INO1 and Ty elements (Winston and Carlson, 1992). Considerable evidence strongly suggests that Snf/Swi controls transcription by its nucleosome remodeling activity (Peterson and Tamkun, 1995; Steger and Workman, 1996; Cairns, 1998). Snf/Swi in S.cerevisiae is an ∼2 mDa complex containing eleven proteins, including Snf2, Snf5 and Swi1 (Cairns et al., 1994; Côté et al., 1994; Peterson et al., 1994). Previous work has demonstrated that Snf/Swi is required for the SUC2 promoter to have an active chromatin conformation (Hirschhorn et al., 1992; Matallana et al., 1992). SUC2 encodes invertase, an enzyme required for growth on media containing the carbon sources sucrose and raffinose. SUC2 transcription is repressed in high glucose (2%) and there is an ∼100‐fold derepression of SUC2 transcription in media containing low glucose (0.05%), sucrose, or raffinose. In a snf/swi mutant such as snf2Δ, which lacks Snf/Swi nucleosome remodeling activity, derepression is defective and only 10% of the wild‐type level of SUC2 mRNA or invertase is expressed (Winston and Carlson, 1992).

In this work, we have addressed three issues concerning the mechanism of Snf/Swi function in vivo by studying the effect of snf/swi mutations on SUC2 expression. First, we demonstrate that Snf/Swi is required for activated levels of SUC2 expression in every cell of a population. Secondly, we provide evidence that Snf/Swi is continuously required for SUC2 transcription. Finally, we show that Snf/Swi is partially redundant with the histone acetyltransferase Gcn5 in controlling SUC2 transcription. Taken together, the continuous requirement for Snf/Swi in maintaining SUC2 transcription provides new evidence for the dynamic nature of chromatin in every cell of a population.


Snf/Swi controls the level of expression from each SUC2 template

We wanted to determine which model of transcriptional activation, binary or gradient, applies to Snf/Swi activation of SUC2. The gradient model predicts that every cell in a snf2Δ population would show the same reduced level of SUC2 expression, that is 10% of the wild‐type level. In contrast, the binary model predicts that there would be two populations in a snf2Δ mutant: 10% of the population with full activity and the remaining 90% of the population with no activity at all. Either model would be consistent with the observed 10% invertase activity in a population of snf2Δ cells.

To distinguish between these two models, we established an in vivo system in S.cerevisiae to analyze expression of the endogenous SUC2 locus in individual cells of a population. Saccharomyces cerevisiae strains were constructed in which the SUC2 open reading frame (ORF) is replaced by the ORF for green fluorescent protein (GFP) (Prasher, 1995; Tsien, 1998). Northern analyses of this allele (suc2Δ10::GFP) demonstrate that it derepresses at a similar rate as wild‐type SUC2 and that its expression is Snf/Swi dependent (Figure 1).

Figure 1.

Northern analysis of suc2Δ10::GFP expression. SNF2+ SUC2+ (FY86), SNF2+ suc2Δ10::GFP (FY1755) and snf 2Δ suc2Δ10::GFP (FY1756) cells were grown in YPD (2% glucose) and shifted to YEP plus 0.05% glucose. RNA was isolated and analyzed from samples taken at the indicated times after the shift. TUB2 mRNA was measured as a loading control. The top pair of panels shows the previously characterized derepression of wild‐type SUC2 RNA (Winston and Carlson, 1992). The middle and bottom pairs of panels compare the expression of the suc2Δ10::GFP allele in SNF2+ and snf 2Δ genetic backgrounds, respectively. The higher level of suc2Δ10::GFP mRNA compared with SUC2 mRNA is probably caused by differential stability of the mRNAs and different specific activities of the 32P‐labeled probes used in these experiments.

To examine expression from the SUC2 promoter in individual cells in a population, both SNF2+ and snf2Δ cells containing suc2Δ10::GFP were subjected to fluorescence‐activated cell sorting (FACS) analysis after growth in either repressing or derepressing conditions (Figure 2). In these analyses, the repressed/derepressed ratios of GFP fluorescence observed in SNF2+ and snf2Δ are not identical to the ratios observed when measuring SUC2 mRNA in the same genetic backgrounds. These differences probably reflect the inherent differences in measuring SUC2 mRNA versus GFP protein. Nevertheless, the suc2Δ10::GFP fusion appears to be an accurate reporter of Snf/Swi control of the SUC2 promoter. As expected, SNF2+ cells did not show any green fluorescence in repressing conditions but did show a high level under derepressing conditions (Figure 2). Also as expected, snf2Δ cells did not express any green fluorescence under repressing conditions. However, under derepressing conditions, snf2Δ cells showed a single, uniform population with only a small increase in GFP fluorescence (Figure 2). FACS analysis of a swi1Δ suc2Δ10::GFP mutant gave similar results (data not shown). In control experiments mixing SNF2+ derepressed cells with SNF2+ repressed cells, we detected the GFP‐expressing cells as a distinct peak when they were as low as 2.5% of the total population (data not shown). Therefore, the presence of a single, uniform population of GFP‐expressing snf2Δ cells suggests that Snf/Swi increases the level of SUC2 expression in every cell and strongly supports the gradient model, and not the binary model, for Snf/Swi activity at SUC2. These results strongly suggest that Snf/Swi is required for maintenance of the activated state and do not rule out the possibility that it is also required for establishment. However, these data exclude the possibility that Snf/Swi is required solely for establishment of the activated state. Lastly, the small increase in green fluorescence in snf2Δ and swi1Δ cells under derepressing conditions suggests that other factors besides Snf/Swi may contribute towards SUC2 transcription.

Figure 2.

Snf/Swi increases the level of SUC2 expression in every cell of the population. (A) FACS profiles of GFP fluorescence in SNF2+ suc2Δ10::GFP (FY1755) and snf 2Δ suc2Δ10::GFP (FY1756) repressed and derepressed cells. Cells were grown in YPD (repressed samples) and shifted to YEP plus 0.05% glucose (derepressed samples) for 2 h and 45 min before undergoing FACS analysis for green fluorescence. SNF2+ SUC2+ cells gave similar levels of green fluorescence as SNF2+ suc2Δ10::GFP repressed cells in both repressing and derepressing conditions. The slight increase in green fluorescence of snf 2Δ cells in repressing conditions is probably due to the larger cell size of the snf 2Δ cells. (B) The median fluorescence of each sample in both repressing and derepressing conditions. The median fluorescence for the total population of cells is shown, as calculated by the CellQuest program.

Snf/Swi is continuously required for SUC2 transcription

To study the role of Snf/Swi in maintaining SUC2 transcription, SUC2 mRNA levels were measured in a snf5 temperature sensitive mutant, snf5‐51. snf5‐51 causes no growth defect at 30°C but prevents growth on raffinose media at 37°C. Analysis of a snf5‐51 mutant at 37°C has shown that SUC2 chromatin is in a ‘repressed’ state, similar to that seen in snf5Δ and snf2Δ mutants (Y.Cao and B.C.Laurent, in preparation). To examine the need for Snf/Swi after SUC2 derepression, the snf5‐51 mutant was grown at 25°C in raffinose to fully derepress SUC2 and then shifted to 37°C. SUC2 mRNA levels were measured over time by Northern analysis. Under the conditions of this experiment, the snf5‐51 cells did not divide following the shift to raffinose. SUC2 mRNA levels significantly decreased in the snf5‐51 mutant after a shift to 37°C and were close to that of a snf5Δ mutant after 4 h (Figure 3). The kinetics of the decrease are consistent with the relatively long half‐life of SUC2 mRNA in media containing a poor carbon source such as raffinose or glycerol (Cereghino and Scheffler, 1996). Our studies indicate that the half‐life of SUC2 mRNA in the snf5‐51 mutant is ∼3 h, consistent with SUC2 transcription being significantly reduced after the shift to 37°C (see Figure 3 legend). Similar results have been obtained by Biggar and Crabtree (1999) by analyzing a snf2Δ mutant under conditions in which the half‐life of the SUC2 mRNA is much shorter. Both sets of results strongly suggest that Snf/Swi is continuously needed to maintain a normal level of SUC2 transcription in vivo.

Figure 3.

Snf/Swi is required for maintenance of SUC2 transcription. SNF5+ (FY22), snf5‐51 (FY1757) and snf5Δ2 (FY1658) cells were grown in YPD at 25°C and then shifted to YPRaffinose for 2 h and 45 min to derepress SUC2. After derepression, cells were shifted to 37°C. RNA was isolated and analyzed from samples taken at the time points indicated. Expression of SUC2 and the loading control, SPT15, was measured by Northern analyses. The relative intensities of the SUC2 and SPT15 bands were quantitated using a PhosphorImager (Molecular Dynamics). The SUC2/SPT15 ratio for each strain was normalized to the SUC2/SPT15 ratio for the wild‐type strain at 0 h, 37°C. The average values from two experiments are presented at the bottom.

Gcn5 is partially redundant with Snf/Swi for SUC2 transcription

The low level of SUC2 expression observed in snf/swi mutants suggested that factors besides Snf/Swi could activate SUC2 transcription to a limited degree. Recent genetic studies have suggested that Gcn5, a histone acetyltransferase (Brownell et al., 1996; Kuo et al., 1998; Wang et al., 1998), could be one of those other factors. Double mutant analysis demonstrated that a combination of a null mutation in GCN5 with null mutations in different SNF/SWI genes caused double mutant sickness (Roberts and Winston, 1997; Recht and Osley, 1999). In addition, gcn5Δ and snf2Δ have shown double mutant lethality in other genetic backgrounds (Pollard and Peterson, 1997). One interpretation of these double mutant phenotypes is that Gcn5 is partially redundant with Snf/Swi for transcription.

To investigate whether Gcn5 was responsible for the low level of SUC2 expression observed in a snf2Δ mutant, two tests were performed. First, a gcn5Δ snf2Δ double mutant was compared with gcn5Δ and snf2Δ single mutants for their ability to grow on media containing sucrose and raffinose as a carbon sources (Figure 4A). As sucrose is a less stringent carbon source than raffinose, the SUC2 expression in either a snf2Δ mutant or a gcn5Δ mutant is sufficient to allow those strains to grow on sucrose plates but not on raffinose plates. However, a gcn5Δ snf2Δ double mutant was unable to grow even on sucrose plates, similar to a suc2Δ, suggesting that SUC2 expression is decreased further in the gcn5Δ snf2Δ double mutant compared with the snf2Δ single mutant. Secondly, we measured SUC2 mRNA levels by Northern analysis under repressing and derepressing conditions in gcn5Δ, snf2Δ, and gcn5Δ snf2Δ mutants (Figure 4B). SUC2 mRNA levels were not significantly affected in a gcn5Δ mutant and were greatly reduced in a snf2Δ mutant under derepressing conditions (Figure 4B). In a gcn5Δ snf2Δ double mutant, the level of SUC2 mRNA was reproducibly below that seen in a snf2Δ single mutant, close to repressed levels. The remaining signal detected in the gcn5Δ snf2Δ double mutant under derepressing conditions and under repressing conditions is probably the unregulated 1.8 kb constitutive, cytoplasmic SUC2 RNA (Carlson et al., 1981). Our analysis of the gcn5Δ snf2Δ effects agree with those obtained by Biggar and Crabtree (1999). These results strongly suggest that both Snf/Swi and Gcn5 contribute to transcriptional activation of SUC2.

Figure 4.

Snf/Swi and Gcn5 are partially redundant for transcriptional activation of SUC2. (A) gcn5Δ snf 2Δ cells cannot grow on YEP plus 2% sucrose (YPSucrose). Wild‐type (FY114), gcn5Δ (FY1354), snf 2Δ (FY1360), and gcn5Δ snf 2Δ (FY1352) cells were spotted onto YPD, YPSucrose and YPRaffinose plates and incubated for 3 days at 30°C. (B) The gcn5Δ snf 2Δ double mutant has lower levels of SUC2 RNA than the gcn5Δ or snf 2Δ single mutants. Wild‐type (FY114), gcn5Δ (FY1354), snf 2Δ(FY1360) and gcn5Δ snf 2Δ (FY1352) cells were grown in YPD (repressing, R) and shifted to YEP plus 0.05% glucose (derepressing, DR) for 2 h and 45 min. mRNA levels of SUC2 and the loading control, TUB2, were measured by Northern analyses. The relative intensities of the SUC2 and TUB2 bands were quantitated using a PhosphorImager (Molecular Dynamics). The SUC2/TUB2 ratio for each strain was normalized to the SUC2/TUB2 ratio for the wild‐type strain derepressed for SUC2 expression. The average values from two experiments are presented at the bottom.


Our studies have investigated the mechanism of Snf/Swi activity in vivo at the SUC2 promoter and have provided three new results. First, we have conducted the first study of transcriptional activation at the level of single cells in S.cerevisiae. Using a GFP‐based reporter integrated at the endogenous SUC2 locus and FACS analysis, we have shown that Snf/Swi is required for activated levels of SUC2 expression in every cell in the population. Secondly, our results strongly suggest that Snf/Swi is continuously required to maintain SUC2 transcription in vivo and not merely required for establishing an active chromatin structure at the locus. Thirdly, we have demonstrated a functional redundancy between Snf/Swi and the histone acetylatransferase, Gcn5.

The use of a GFP reporter has allowed us to analyze the role of Snf/Swi in transcriptional control at the level of individual cells. In the case that we have studied, all cells in the population had a similar level of transcription and did not show any bimodal expression patterns. Among single‐cell studies using mammalian cells, both gradient and binary responses have been obtained (Weintraub, 1988; Fiering et al., 1990; Walters et al., 1995, 1996; Bagga and Emerson, 1997; Kringstein et al., 1998) suggesting that mechanisms for transcription will differ among various activators and loci. Given the relative ease of this type of analysis in yeast, it should be straightforward to understand the role of any transcriptional activator at this level in S.cerevisiae. In addition, each promoter can be studied at its endogenous genomic location avoiding artifacts resulting from changes in position of the reporter.

The role of Snf/Swi in maintaining transcription has been an open issue. Several in vitro studies have examined the need for continuous Snf/Swi activity for nucleosome remodeling, both on mononucleosomes and on nucleosomal arrays. Studies performed on mononucleosomes and some plasmid templates have suggested that Snf/Swi, or some aspect of Snf/Swi activity, may not be continuously required in vitro to maintain an altered nucleosomal structure (Imbalzano et al., 1996; Côté et al., 1998). Other studies, with nucleosomal arrays, have suggested that Snf/Swi‐dependent remodeling is reversed upon removal of Snf/Swi, implying a need for continuous chromatin remodeling (Owen‐Hughes et al., 1996; Logie and Peterson, 1997). Our results suggest that, in vivo, continuous remodeling by Snf/Swi is necessary to maintain the transcriptionally active state of promoters like SUC2.

There are several possible reasons why Snf/Swi might be continously needed for SUC2 transcription. Snf/Swi may facilitate increased rates of RNA polymerase II reinitiation, as has been observed for other activators (Ho et al., 1996). Alternatively, Snf/Swi may play a role in transcription elongation at SUC2. In support of this possiblity, Snf/Swi has been shown to help overcome the inhibitory effect of nucleosomes on elongation (Brown et al., 1996). Finally, Snf/Swi may be continously required because, in vivo, other factors may remodel chromatin back to an inactive state. By this model, chromatin structure at Snf/Swi‐dependent genes would be in a dynamic equilibrium between a remodeled ‘open’ state and a non‐remodeled ‘closed’ state, constantly requiring Snf/Swi activity for transcription. This picture of transcription is further complicated by recent studies showing that human Snf/Swi and the related yeast complex, Rsc, can cycle nucleosomes between different conformations in vitro. These work suggest that, in addition to other factors, Snf/Swi itself may also contribute to modulating this dynamic equilibrium in vivo (Lorch et al., 1998; Schnitzler et al., 1998).

Lastly, we have shown that the remaining SUC2 transcription in snf/swi mutants is dependent on the action of Gcn5. The genetic interactions between Gcn5 and Snf/Swi and their effects on SUC2 expression suggest that their roles in transcriptional control overlap. The comparison of the phenotypes of the snf2Δ and gcn5Δ single mutants to those of the gcn5Δ snf2Δ double mutant, taken in combination with their effects on SUC2 expression and other genes (Biggar and Crabtree, 1999), suggests that both Snf/Swi and Gcn5 contribute to the activation of a common set of genes or to an overlapping but non‐identical set of genes. Measurement of all mRNA levels in these mutants (Schena et al., 1995; Chee et al., 1996; Velculescu et al., 1997) should help to distinguish between these possibilities. Based on this type of mRNA analysis for cells grown under one set of conditions, Snf/Swi and Gcn5 clearly do not control an identical set of genes (Holstege et al., 1998). However, mRNA analysis in the gcn5Δ snf2Δ double mutant would be required to truly clarify these possibilities.

It is unclear how the the Snf/Swi and Gcn5 activities might be coordinated to facilitate transcription from the SUC2 promoter. The relative effects of snf2Δ and gcn5Δ mutations suggest that chromatin remodeling by Snf/Swi is very important for SUC2 transcription whereas histone acetylation by Gcn5 plays a more minor role. Since the most striking effect of a gcn5Δ mutation is observed in a snf2Δ background, Gcn5 may prefer to acetylate histones at SUC2 when they have not been remodeled by Snf/Swi. Analyzing acetylation states at SUC2 in both wild‐type and snf/swi mutant strains will be important to understand the functional relationships between these factors. These studies should enhance our understanding of how different chromatin modifying factors control and maintain a dynamic chromatin structure at a particular locus.

Materials and methods

Yeast strains and methods

The S.cerevisiae strains used in these studies were constructed by standard methods (Rose and Winston, 1990). They are FY22 (MATa his3Δ200 ura3‐52), FY86 (MATα his3Δ200 ura3‐52 leu2Δl), FY114 (MATa ura3‐52 lys2‐173R2), FY1352 (MATa his3Δ200 ura3‐52 leu2Δ1 lys2‐173R2 gcn5Δ::HIS3 snf2Δ::LEU2), FY1354 (MATα his3Δ200 ura3‐52 leu2Δ1 lys2‐173R2 gcn5Δ::HIS3), FY1360 (MATa his3Δ200 ura3‐52 leu2Δ1 lys2‐173R2 snf2Δ::LEU2), FY1658 (MATa his3Δ200 ura3‐52 lys2‐128δ snf5Δ2), FY1755 (MATα his3Δ200 ura3‐52 leu2Δ1 suc2Δ10::GFP), FY1756 (MATα his3Δ200 ura3‐52 snf2Δ::HIS3 suc2Δ10::GFP), and FY1757 (MATa his3Δ200 ura3‐52 snf5‐51). Media was prepared according to standard methods (Rose and Winston, 1990).

In the suc2Δ10::GFP allele, the coding region of SUC2 is replaced with the ORF of yeast‐enhanced GFP (yEGFP3) (Cormack et al., 1997). This allele was constructed in two steps. First, the URA3 ORF was inserted into the BamHI site within the SUC2 coding region. Then, a PCR fragment containing yEGFP3 flanked by 41 bp of 5′ and 3′ sequence immediately adjacent to the SUC2 ORF was used to transform the suc2::URA3 strain. Selecting for 5‐fluoroorotic acid resistance enriched for cells in which the GFP‐encoding sequence had integrated at SUC2, precisely replacing the SUC2 ORF.

FACS analysis

Cells were grown in yeast extract/peptone/2% glucose (YPD) to a density of 1−2×107 cells/ml. They were washed in water, and either resuspended in YPD (repressed sample) or shifted to yeast extract/peptone (YEP) plus 0.05% glucose (derepressed sample). Cells were then grown at 30°C for 2 h and 45 min, washed once and resuspended in water before FACS analysis. Thirty thousand cells were scanned for green fluorescence in a Becton Dickinson FACScan and results were analyzed using the CellQuest 3.1 program (Becton Dickinson).

Northern hybridization analysis

Strains were grown in YPD (2% glucose) to ∼8×106 cells/ml. They were washed in water and resuspended in either YPD (repressed sample), YEP plus 0.05% glucose (derepressed sample), or YEP plus 2% raffinose (YPRaffinose; derepressed sample). Ten milliliter samples were taken at the times indicated in the figures. RNA was prepared and analyzed as described previously (Swanson et al., 1991). RNA levels were quantitated by PhosphorImager analysis (Molecular Dynamics).


We thank S.R.Biggar and G.R.Crabtree for communicating unpublished results. We thank Glenn Paradis, Mike Jennings and Juanita Campos Torres for expert technical assistance with FACS analysis. We thank Barak Cohen, Aimée Dudley, Grant Hartzog and Robert Kingston for comments on the manuscript. This work was supported by a grant from the National Institutes of Health to F.W.


View Abstract