INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

Chromatin structure inhibits gene transcription by blocking access of components of the transcriptional machinery and gene-specific activator proteins to DNA recognition sequences (18, 22, 33). The structural changes in chromatin that accompany transcriptional activation of a gene often require multiprotein factors that remodel nucleosomes. The conserved Saccharomyces cerevisiae Snf-Swi complex is the prototype of one class of eukaryotic factors that restructures chromatin by a process requiring ATP hydrolysis (7, 31, 46, 66). Histone acetyltransferases represent a second class of multiprotein complexes that modulates nucleosome structure by acetylating the amino-terminal tails of the core histones (2, 21, 58, 70). Genetic and biochemical experiments suggest that these two classes of remodeling factors function interdependently or in parallel in transcription at certain promoters (3, 12, 32, 37, 49, 51, 59, 61, 67). ATP-utilizing factors can also function with histone deacetylases in transcriptional repression (see reference 63).

Other Snf-Swi family complexes have been isolated from human (hSWI/SNF), Drosophila melanogaster (brm), and yeast (RSC) cells (34, 70), indicating the evolutionary importance of the nucleosome-remodeling activity. hSWI/SNF appears to regulate cell cycle progression (73) and cellular differentiation (6). RSC is essential for progression through the mitotic cell cycle (9, 10, 16, 62). Despite functional and compositional differences, several subunit polypeptides of the Snf-Swi complexes are highly conserved. Moreover, three conserved human Snf-Swi members, Snf2p (also known as BRG1 or hBrm), Snf5p (also known as INI1), and Swi3p (also known as BAF155 or BAF170), constitute a core set of Snf-Swi factors (see reference 48). Importantly, INI1 may function as a tumor suppressor (65).

Snf-Swi complexes disrupt histone-DNA contacts in mononucleosomes and nucleosome arrays in reactions requiring ATP hydrolysis in vitro (see reference 66). Possible remodeling mechanisms include the generation of superhelical torsion, the creation of chromatin loops, the generation of activated nucleosome intermediates, and the sliding or transfer of histone octamers (26, 66).

The yeast SNF and SWI genes, first identified as mutants defective in the expression of the SUC2 (snf, for sucrose nonfermenting) and HO (swi, for mating-type switching) genes (43, 56), have now been shown to affect the transcription of many other genes (28, 35, 60). A functional link to chromatin was established by the identification of mutations in genes encoding histones and other chromatin assembly factors as snf and swi suppressors (27, 36, 44, 50, 51). The Snf-Swi proteins are assembled into a large 2-MDa complex comprised of 11 polypeptides that is important but not essential for mitotic growth of cells (8, 13, 45).

In vivo analysis of the chromatin structures of the SUC2 and PHO8 promoters provided evidence that Snf-Swi-dependent chromatin-remodeling activity is required for transcriptional initiation (19, 20, 27, 39, 71). Snf-Swi associates directly with target promoters (12, 14), and both transcriptional activators and repressors have been implicated in the targeting mechanism(s) (see reference 47).

With the exception of the conserved Snf2p-Swi2p ATPase subunit required for nucleosome perturbation, the mechanism of action of other Snf-Swi component proteins has not been addressed. In this study, we have investigated the in vivo roles of Snf5p, a core subunit of Snf-Swi, through genetic and biochemical characterization of two conditional snf5-ts regulatory mutants. We present evidence that Snf5p is involved in maintaining Snf-Swi integrity and in postrecruitment chromatin remodeling in vivo.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

Yeast strains and genetic methods. Yeast strains BLY1 (MAT his3-200 lys2-801 ura3-52 SUC2), BLY3 (MAT snf5-2 his3-200 ura3-52 ade2-101), BLY35 (MAT snf2-2::URA3 his3-200 ura3-52 ade2-101), BLY61 (MAT snf5-51ts his3-200 ura3-52 ade2-101), and BLY169 (MAT snf5-83bts his3-200 ura3-52 ade2-101) are isogenic derivatives of S288C, and BLY54 (MATa leu2 ura3-1 trp1-1 his3 6lexAOp-LEU2 GAL⁺) (17) (gift of E. Golemis) is derived from W303. The following media were used: YPD medium (YEP [1% yeast extract and 2% peptone] containing 2% dextrose), YPR medium (YEP supplemented with 2% raffinose [1 µg of antimycin A/ml]), and YPGal medium (YEP supplemented with galactose). Synthetic complete (SC) medium is SC medium containing yeast nitrogen base supplemented with 2% sugars (dextrose or raffinose) and a drop-out mixture of amino acids and bases (52). SD-inositol medium contains inositol-free yeast nitrogen base (Bio-101) and 2% dextrose. Glucose-repressed cultures were grown in YPD at 30°C. Glucose-derepressed cells were grown first to mid-logarithmic phase in YPD, washed twice with water, and transferred to YEP plus 0.05% glucose for the times indicated at either 30 or 37°C. Standard genetic procedures were followed (52).

Plasmids. LexA hybrid plasmids are derivatives of pSH2-1 (24) and express, from the constitutive ADH1 promoter, the amino-terminal 87 residues of the LexA protein fused to the indicated Snf5p residues. pLY50 carries the wild-type SNF5 gene and was created by cloning the 4.8-kb EcoRI-BamHI fragment of pJW34 (1) into pRS316. Site-directed mutagenesis of SFH1 was carried out with pIN18, a derivative of pRS316 (CEN6 URA3) carrying the 1.9-kb XhoI-AseI fragment of pYC5H (SFH1) (10). Oligonucleotide sequences and plasmid construction details will be provided upon request.

Enzyme assays. -Galactosidase activity was assayed in permeabilized cells (23, 40). Secreted invertase activity was assayed in whole cells as previously described (64).

Isolation of snf5 temperature-sensitive alleles. pLY50 (SNF5 CEN6 URA3) was mutagenized in vitro as previously described (16), and the mutagenized plasmid DNA was used to transform strain BLY3 (snf5) to uracil prototrophy. Approximately 4,500 Ura⁺ transformants were patched onto YPD plates and then replicated to two sets of YPR plates. One set was incubated at 30°C and the other at 37°C. Plasmid DNA was recovered from those colonies that grew at 30°C but not at 37°C. Mutations located within SNF5 were confirmed by restriction fragment swapping. Five temperature-sensitive mutants were identified, and the alleles were sequenced. The nucleotide changes (and predicted amino acid changes) of the three alleles described here are as follows. snf5-51ts and snf5-65ts (snf5-51ts and snf5-65ts are identical and are hereafter referred to as snf5-51ts) contained a nucleotide change of G-1744 to A (E582K), and snf5-83ts contained nucleotide changes of G-1066 to A (E356K), C-1085 to T (P362L), and G-1423 to A (D475N) and contained a silent mutation at Y360.

Immunological procedures. Recombinant Snf5 (amino acids 1 to 193) and Snf2 (amino acids 1256 to 1703) proteins fused to glutathione S-transferase in pGEX-3X and pGEX-2T (Pharmacia), respectively, were purified from Escherichia coli cells and used to immunize rabbits as previously described (25). Immunoblot analysis was performed as described previously (10) using anti-Snf5p (1:2,000), anti-Snf2p (1:1,000), anti-Swi3p (1:1,000; gift of C. L. Peterson), and anti-LexA (1:2,000; gift of R. Brent) polyclonal antibodies and developed with the nitroblue tetrazolium-BCIP (5-bromo-4-chloro-3-indolylphosphate) color reagents except where noted.

Transcriptional activation of lexA-LEU2 Growth of cells expressing the wild-type and mutant Snf5 LexA fusion proteins was first compared on SC-His plates containing 0, 100, 300, or 900 µg of leucine. Mutant Snf5 proteins exhibited temperature-sensitive growth only on plates containing 300 µg of leucine (limiting leucine). Therefore, the transcriptional activation assays were carried out on leucine (900 µg) and limiting leucine (300 µg) plates at 30 and 37°C.

RNA analysis. Total RNA was prepared as previously described (38), and SUC2 and U6 (as a control) RNA transcript levels were measured by primer extension analysis using specific primers as previously described (50).

Gel filtration of Snf-Swi complex. Log-phase cultures of the snf5-ts or SNF5 (BLY1) cells grown in YPD were derepressed at either 30 or 37°C for 2 h. Whole-cell protein extracts were prepared and analyzed on a fast protein liquid chromatography Superose 6 gel filtration column (Pharmacia) as previously described (45). Proteins were trichloroacetic acid precipitated, separated on sodium dodecyl sulfate-6% polyacrylamide gels, and analyzed by Western blot analysis.

Protein immunoprecipitation. Volumes containing 1.2 mg of protein from whole-cell lysates (51) were incubated with 1 µl of anti-Snf5p antibody for 2 h at 4°C in 1.0 ml of immunoprecipitation buffer (51). Immune complexes were collected using protein A-coupled agarose beads and washed as previously described (51).

Chromatin structure analysis. Nuclei were isolated according to methods described by Roth and Simpson (53) from cells grown under glucose-repressing or -derepressing conditions at 30 or 37°C for 2 h. Micrococcal nuclease (Sigma) digestion, isolation of DNA from nuclei, and primer extension analysis were carried out as previously described (19). Primer extension analysis at the SUC2 locus was performed with Taq DNA polymerase (Fisher Scientific) and oligonucleotide primer F1 (a gift from R. T. Simpson and I. Gavin), which corresponds to SUC2 base pairs 784 to 755 (19).

Chromatin immunoprecipitations. Approximately 3 × 10⁸ cells were fixed in 1% formaldehyde for 15 min at room temperature. Cross-linked cells were lysed by glass bead breakage in lysis buffer as previously discussed (57). Chromatin was solubilized by sonication to an average DNA fragment size of 0.4 kb. For immunoprecipitation, 1 µl of anti-Snf5p or anti-Snf2p antibody was incubated with 0.8 mg of extract in 1.0 ml of lysis buffer overnight at 4°C (57). Immune complexes were collected and washed as previously described (57), and PCR was performed on extracted DNA with SUC2 gene and reference primer pairs. PCR products were separated on 8% polyacrylamide gels, and photoprocessing was carried out using a Foto Eclipse (Fotodyne) digital imaging system.

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

The functional domain of Snf5p contains two repeat motifs and is conserved throughout eukaryotes. Snf5p is an essential and highly conserved component of the multiprotein yeast Snf-Swi complex. The sequences between Snf5p amino acids 455 and 676 are 32 to 46% identical to proteins encoded by open reading frames in human (30, 42), Drosophila (15), Danio rerio (zebrafish) (EMBL accession no. AJ249795.1), Caenorhabditis elegans (EMBL accession no. R07E5.3), Schizosaccharomyces pombe (EMBL accession no. C2F7.08C), S. cerevisiae (10), Arabidopsis thaliana (4), and Tetraodon fluviatilis (puffer fish) (72) cells, suggesting that an activity essential for basic cellular processes has been conserved during evolution. Notably, the highly conserved domain of the indispensable yeast Sfh1p protein, the sole yeast homologue of Snf5p, is sufficient for wild-type SFH1 function (10). In addition, truncating mutations of the conserved domain of human Snf5p are associated with oncogenesis (65).

455-678

View larger version (27K): [in this window] [in a new window]   FIG. 1.   The functional domain of Snf5p. The first and last amino acids of Snf5p fused to LexA, or the amino acids deleted (), are indicated. Invertase activity was measured in strain BLY3 cells (snf5) carrying the indicated plasmids grown under glucose-repressing (R) and glucose-derepressing (D) conditions and is expressed as micromoles of glucose released per minute per 100 mg (dry weight) of cells. -Galactosidase (-Gal) activity was measured in strain BLY1 cells (SNF5) carrying the indicated lexA-SNF5 plasmids and target plasmids with a single (1Op) or six overlapping (6Op) lexA operators upstream of the GAL1-lacZ reporter gene, and the results are expressed in Miller units. The glutamine-rich N terminus (Q) is stippled, three proline-rich regions (P) are hatched, a highly charged central region is filled, and Rep1 and Rep2 are the direct imperfect repeats at positions 457 to 498 and 541 to 601, respectively. Invertase and -galactosidase values represent the averages of four independent isolates. Errors were <12% for values >2.

Conditional temperature-sensitive alleles of SNF5 alter amino acids within evolutionarily conserved repeat motifs. For a more detailed investigation of the role of Snf-Swi in transcriptional regulation in vivo, a genetic screen was initiated to isolate conditional loss-of-function mutations in SNF5. Temperature-sensitive mutations in SNF5 would permit a temporal examination of the physiological impact of loss of Snf-Swi activity. Conditional mutants were identified on the basis of thermolabile growth on raffinose media. Each of the alleles to be described, snf5-51ts and snf5-83bts, is recessive and was complemented by plasmids containing the wild-type SNF5 gene. Snf5 protein levels were comparable in the snf5-ts mutants and SNF5 cells grown at permissive (30°C) or nonpermissive (37°C) temperatures (Fig. 2A).

View larger version (29K): [in this window] [in a new window]   FIG. 2.   Characterization of snf5-ts mutations. (A) Snf5 protein expression levels are unaltered in the snf5-ts mutants. Log-phase cultures of BLY1 (SNF5 [WT]), BLY61 (snf5-51ts), and BLY169 (snf5-83bts) cells were derepressed for 2 h at 30 or 37°C. Proteins from whole-cell lysates (10) were immunoblotted with anti-Snf5p antiserum. (B) The snf5-ts mutants show distinct growth phenotypes at nonpermissive temperature. The same strains used in panel A and BLY3 (snf52) were grown in patches on a YPD plate and then replica plated onto YPD, YPR, SD-inositol, or YPGal media and incubated for 2 days at 30 or 37°C. (YPD patches were taken from a different region of the same master plate.) (C) The LexA-Snf5ts fusion proteins fail to activate transcription in vivo. Threefold serial dilutions of log-phase BLY54 (lexAop-LEU2) cells expressing LexA-Snf5p, LexA-Snf5-83bp, LexA-Snf5-51p, or LexA were spotted onto SC (leucine) or SC limiting leucine plates and incubated for 2 to 4 days at 30 or 37°C. (D) The snf5 temperature-sensitive mutations alter highly conserved amino acids. Schemes of Snf5p and Sfh1p are shown drawn to scale, with the conserved domains of each being shaded and Rep1 and Rep2 shown as filled boxes. The positions of mutations and the amino acid changes are indicated. (E) The corresponding mutations in SFH1 confer a temperature-sensitive phenotype. Cells of strains BLY210 (SFH1), BLY213 (sfh1-D219N), BLY214 (sfh1-D219K), BLY215 (sfh1-D317N), and BLY217 (sfh1-D317K), which carry the indicated SFH1 allele on plasmids in the sfh1 background, were streaked onto YPD plates and incubated for 2 days at 30 or 37°C.

snf5-ts mutations regulate transcription of SUC2 To study the consequences of the conditional inactivation of SNF5 at a target promoter in more detail, invertase activity and SUC2 RNA levels were compared in SNF5, snf5, and snf5-ts cells under derepressing conditions at 30 and 37°C (Table 1; Fig. 3). SUC2 is a glucose-repressible gene whose expression is repressed in the presence of high glucose and induced 100-fold by growth in low glucose (29).

View larger version (29K): [in this window] [in a new window]   FIG. 3.   snf5-ts mutations regulate SUC2 transcription. Primer extension analysis was carried out with total RNA prepared from repressed cells (0 h) or cells derepressed at 30 or 37°C for the indicated times. Primers specific for SUC2 and U6 transcripts were used for each reaction.

Snf-Swi assembly is perturbed in the snf5-51ts mutant and only moderately altered in the snf5-83bts mutant. The transcriptional defects in the snf5-ts mutants could be explained by disassembled Snf-Swi complexes or by assembled but functionally inactive complexes. Therefore, the integrity of Snf-Swi in the snf5-ts mutants was examined by gel filtration (Fig. 4A). Whole-cell extracts were applied to a Sepharose 6 gel filtration column, and elution of Snf-Swi polypeptides was monitored by immunoblot analysis. The fractionation of SNF5 whole-cell extracts showed coelution of the Snf2p, Swi3p, and Snf5p polypeptides as peaks in fraction 19, suggesting an assembled Snf-Swi complex as shown previously (45), and this pattern was unaffected at 37°C. In contrast, in the fractionation of snf5 cells, the elution of Snf2p and Swi3p was altered significantly (Fig. 4A) as shown previously (45), suggesting partial disassembly of Snf-Swi. The elution pattern of proteins upon fractionation of snf5-51ts extracts indicated that assembly of Snf-Swi was perturbed at both 30 and 37°C. In contrast, fractionation of whole-cell extracts derived from snf5-83bts cells suggested that the integrity of Snf-Swi was only moderately affected. At 30°C, Snf2p, Swi3p, and Snf5p coeluted exactly as they eluted in SNF5 cells. At 37°C, significant amounts of these polypeptides coeluted and peaked in fraction 19 (Fig. 4A), an elution pattern resembling that of extracts prepared from the Snf-Swi ATPase mutant swi2K798A (45).

View larger version (37K): [in this window] [in a new window]   FIG. 4.   Assembly of the Snf-Swi complex is aberrant in the snf5-51ts mutant but only mildly altered in the snf5-83bts mutant. (A) Whole-cell protein extracts prepared from the same strains used in Fig. 2 grown under derepressing conditions for 2 h at 30 or 37°C were fractionated on Superose 6, and fractions were assayed for Snf2p, Swi3p, and Snf5p by immunoblot analysis. Swi3p immunoblots and the Snf2p immunoblot prepared from wild-type cells at 30°C were followed by chemiluminescence; all other immunoblots were followed colorimetrically. Proteins with molecular masses of 669 and 443 kDa should have eluted in fractions 25 and 28, respectively (45). L, load. (B) Immunoprecipitations were carried out as described in Materials and Methods with whole-cell lysates prepared from the same strains described for panel A under the same growth conditions. Anti-Snf5p-precipitated proteins were separated on sodium dodecyl sulfate-4 to 15% polyacrylamide gels and probed with anti-Snf5p or anti-Snf2p antibodies.

snf5-ts mutants fail to remodel chromatin structure of SUC2 promoter. The chromatin structure of the SUC2 promoter changes dramatically upon derepression in an Snf-Swi-dependent, transcription-independent manner (19, 27). To study whether the snf5-ts mutant Snf-Swi complexes are competent for nucleosome remodeling, we analyzed the chromatin organization of a region of the SUC2 promoter sensitive to mutations in SNF/SWI, 678 to 519 (19, 71), by primer extension methodology (Fig. 5).

View larger version (132K): [in this window] [in a new window]   FIG. 5.   The nucleosome structure of the SUC2 UAS region is not remodeled in the snf5-ts mutants. Primer extension analysis (using the F1 primer) was carried out with chromatin isolated from cells grown under repressing (R) or derepressing conditions at 30°C (D30°C) or 37°C (D37°C) digested previously with increasing amounts of micrococcal nuclease. Schematic features of the SUC2 upstream region are shown on the left (19). The presumed positions of the nucleosomes are indicated by ellipses. Brackets denote regions containing derepression-sensitive nuclease cutting sites. Arrows mark the major sites of digestion in the repressed SUC2 promoter; numbers indicate the distance from the initiating A residue. Triangles indicate increasing concentrations of micrococcal nuclease as previously described (19). The naked DNA samples (N) were digested with micrococcal nuclease as a control. Slight differences in the mobility of DNA fragments in the four strains are due to differences in gel electrophoresis. Strains used were the same as those in Fig. 4.

Snf-Swi occupancy at SUC2 promoter. The inability of the snf5-51ts and snf5-83bts mutant Snf-Swi complexes to remodel nucleosomes or activate transcription at target chromosomal loci suggests that these complexes are compromised either in their association with chromatin or in a subsequent chromatin-remodeling step(s). To test these possibilities, we measured the presence of Snf5p and Snf2p in vivo at the SUC2 promoter in wild-type (derepressed and repressed) and mutant cells by chromatin immunoprecipitation. At the same time, chromatin immunoprecipitation assays carried out with extracts prepared from wild-type glucose-repressed and glucose-derepressed cells affords a test of the model that transcription of SUC2 is controlled by promoter recruitment of Snf-Swi. Polyclonal anti-Snf5p and anti-Snf2p antibodies were incubated with formaldehyde cross-linked chromatin, and the immune complexes were collected by binding to protein A-Sepharose beads. Levels of SUC2 promoter DNA and two reference DNAs in the immunoprecipitates were compared using PCR. None of the DNAs was precipitated in the absence of Snf5p or Snf2p, in mock immunoprecipitations using preimmune sera or in immunoprecipitations with non-cross-linked chromatin (Fig. 6A and B).

View larger version (31K): [in this window] [in a new window]   FIG. 6.   In vivo association of Snf5p and Snf2p with the SUC2 promoter. Chromatin obtained from 2-h glucose-derepressed (except as noted) strains BLY1 (WT), BLY3 (snf5), BLY35 (snf2), BLY61 (snf5-51ts), and BLY169 (snf5-83bts) was immunoprecipitated with anti-Snf5p, anti-Snf2p, or preimmune sera (pre-I). DNA isolated from immunoprecipitates or from whole-cell extracts (WCE) was amplified by PCR with primers specific for the SUC2 UAS (SUC2; 250 bp) combined with two pairs of reference primers for promoter regions of flanking genes 1.7 kb downstream (228-bp product) and 2.4 kb upstream (325-bp product) of SUC2, simultaneously. The PCR products were separated on an 8% polyacrylamide gel. (A) Association of Snf5p with the SUC2 UAS. Lane 1, anti-Snf5p immunoprecipitation with non-cross-linked SNF5 chromatin (No-X); lane 2, preimmune serum immunoprecipitation of SNF5 chromatin; lane 3, anti-Snf5p immunoprecipitation of chromatin from glucose-repressed SNF5 (WT-R) cells; lanes 4 to 9, anti-Snf5p immunoprecipitation of snf2, snf5, snf5-51ts, snf5-83bts, and wild-type chromatin (lane 8, undiluted wild type; lane 9, 1:3 dilution) from derepressed cells at 30°C; lanes 10 to 13, anti-Snf5p immunoprecipitation of snf5, snf5-51ts, snf5-83bts, and wild-type chromatin prepared from derepressed cells at 37°C; lanes 14 to 16, threefold serial dilutions of total input DNA. (B) Association of Snf2p with the SUC2 UAS. The same chromatin solutions used in panel A were immunoprecipitated with anti-Snf2p antibody or preimmune serum.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Here, we present genetic and biochemical evidence that supports distinct roles for the conserved core subunit, Snf5p, in Snf-Swi assembly and nucleosome-remodeling activities. These data provide new insights into in vivo mechanisms of chromatin targeting and remodeling by Snf-Swi complexes.

In this work, critical roles of Snf5p in Snf-Swi function have been revealed by characterization of two snf5-ts mutants harboring single alterations of invariant or conserved amino acids that reside within two repeat sequences, Rep1 and Rep2. The possibility that these repeats carry out distinct functions is suggested by the finding that the corresponding hSNF5 repeat motifs show differential properties in binding human immunodeficiency virus integrase and c-MYC (11, 41). We found that, under nonpermissive conditions, both mutations in the Snf5p repeat motifs compromise Snf-Swi chromatin remodeling and transcriptional activation but through distinct mechanisms. The snf5-51ts mutant demonstrates that Snf5p is essential for the assembly and promoter targeting of Snf-Swi. In contrast, the snf5-83bts mutant uncovers a critical role for Snf5p in one or more postrecruitment remodeling functions.

The snf5-51ts mutation (an E582K substitution in Rep2) severely perturbs Snf-Swi assembly, even at permissive temperature, and further dissociates Snf5p from Snf2p at nonpermissive temperature (Fig. 4). This result extends the previous whole-cell extract fractionation studies, which showed that Snf-Swi polypeptides no longer copurify in the absence of Snf5p (45), by demonstrating that a single amino acid substitution within Rep2 abolishes the architectural function of Snf5p. Furthermore, in the snf5-51ts mutant at both temperatures, association of Snf5p and Snf2p with the SUC2 promoter is impaired, arguing that an intact Snf-Swi complex is necessary for promoter recruitment.

We observed substantial derepression of the SUC2 gene at 30°C in the snf5-51ts cells despite defective nucleosomal remodeling (compare Table 1 and Fig. 3 to Fig. 4). One possibility is that transient chromatin remodeling at earlier time points was missed, since Snf-Swi recruitment and chromatin-remodeling assays were carried out only at the 2-h time point. However, our preliminary results showed that the chromatin organization of the SUC2 UAS region in snf5-51ts cells derepressed for 30 min is no different than that at 2 h (data not shown). Therefore, we favor the idea that the low amount of Snf-Swi complex present at the UAS region in derepressed snf5-51ts cells (comparable to that in repressed SNF5 cells) is sufficient to allow efficient transcription but fails to stably support transcription (Fig. 3). Snf-Swi could activate transcription by disrupting higher-order chromatin structure, facilitating histone acetylation (3, 35, 51, 59), or interacting with the RNA polymerase holoenzyme (68), any of which might not lead to detectable changes in the remodeling assay.

An alternative possibility is that the snf5-51ts mutant Snf-Swi remodels other regions of the SUC2 promoter, distinct from the UAS, to permit transcription. Although the SUC2 UAS region is essential for gene derepression (5, 54, 55), other SUC2 promoter sequences are also remodeled in an Snf-Swi-dependent manner (5, 19, 27, 39, 71). In addition, Snf-Swi can be differentially recruited to distinct regions of the HO promoter (12). Therefore, it is tempting to speculate that the snf5-51ts mutation affects Snf-Swi remodeling activity at distinct promoter elements (such as the UAS and TATA regions) differently. Experiments are under way to resolve these important issues.

In contrast to the snf5-51ts mutation, the snf5-83bts mutation (a D475N substitution in Rep1) affects neither Snf-Swi assembly nor its recruitment to SUC2, indicating that the remodeling and transcriptional defects at nonpermissive temperature are caused by blockage of one or more postrecruitment Snf-Swi remodeling step(s). Interestingly, hSNF5 has been shown to moderately stimulate the nucleosome-remodeling activity of the human Snf2p homologue, BRG1, in an in vitro reconstitution experiment (48). Snf5p could have a similar stimulatory effect on remodeling, although the effect is expected to be more robust in vivo. Another possibility is that Snf5p plays a role in the coordination of Snf-Swi and other remodeling factors required for chromatin remodeling (20, 61).

The activity of Snf-Swi in vivo is controlled by both chromosomal targeting (see reference 47) and postrecruitment events (20, 61). Our results reveal essential roles for Snf5p in both processes. We propose that Snf5p integrates important protein-protein interactions for Snf-Swi assembly and coordinates promoter recruitment and chromatin remodeling.