The Histone Chaperone Asf1 Increases the Rate of Histone Eviction at the Yeast PHO5 and PHO8 Promoters*

Eukaryotic gene expression starts off from a largely obstructive chromatin substrate that has to be rendered accessible by regulated mechanisms of chromatin remodeling. The yeast PHO5 promoter is a well known example for the contribution of positioned nucleosomes to gene repression and for extensive chromatin remodeling in the course of gene induction. Recently, the mechanism of this remodeling process was shown to lead to the disassembly of promoter nucleosomes and the eviction of the constituent histones in trans. This finding called for a histone acceptor in trans and thus made histone chaperones likely to be involved in this process. In this study we have shown that the histone chaperone Asf1 increases the rate of histone eviction at the PHO5 promoter. In the absence of Asf1 histone eviction is delayed, but the final outcome of the chromatin transition is not affected. The same is true for the coregulated PHO8 promoter where induction also leads to histone eviction and where the rate of histone loss is reduced in asf1 strains as well, although less severely. Importantly, the final extent of chromatin remodeling is not affected. We have also presented evidence that Asf1 and the SWI/SNF chromatin remodeling complex work in distinct parallel but functionally overlapping pathways, i.e. they both contribute toward the same outcome without being mutually strictly dependent.

The DNA of eukaryotic cells is compacted in the nucleus into a complex structure called chromatin. The first level of chromatin organization is formed by the nucleosome, which consists of a histone octamer core organizing ϳ1.7 turns of double-stranded DNA around its surface (1). DNA that is wound around a histone octamer in a canonical nucleosome is much less accessible for most DNA-interacting factors than DNA in the linker regions between nucleosomes.
It is now widely accepted not only that nucleosomes serve a structural role for the compaction of eukaryotic DNA but also that the obstructive nature of the nucleosomal histone-DNA interactions is a means to regulate the expression of genetic information (2)(3)(4). This mode of regulation involves changes in chromatin structure at, for example, promoter or enhancer regions. A hallmark of such regulatory changes is the switch of DNA regions from a state that is protected from nucleases to a state that is sensitive, or even hypersensitive, to nucleases.
To understand the process of regulation through chromatin structure it is therefore crucial to study the molecular mechanisms that lead to the inducible generation of hypersensitive sites. To this end, the PHO5 promoter in yeast became a classical model system (5). In its repressed state this promoter region is organized into four positioned nucleosomes with a short hypersensitive site in the middle. Upon activation by phosphate starvation this characteristic chromatin organization becomes remodeled into an extended hypersensitive region (6). The promoter nucleosomes in this induced state are completely disassembled as assayed by the loss of histone DNA contacts (7)(8)(9). Recently, we and others showed that this loss of nucleosomal organization corresponds to a movement of histones away from the promoter in trans (10,11), raising the mechanistic question of where the histones go.
Free histones together with DNA are notoriously aggregation prone, and it is assumed that they occur in the cell mainly in complex with nucleic acids or histone chaperones (12)(13)(14). Histone chaperones are a diverse family of proteins that interact with various kinds of histones and probably serve mainly as histone donors and acceptors during nucleosome assembly or disassembly processes. In addition, they can have regulatory functions as in the case of the Hir proteins or Asf1 (15,16). Histone chaperones appear to form a redundant network as yeast strains deleted in multiple histone chaperone genes are viable (13). Nonetheless, histone chaperones can also be specific for certain histone variants (17) or for certain processes, e.g. for replication-dependent or -independent chromatin assembly (18, 19). Chromatin remodeling complexes may work more or less specifically with, or may even contain, histone chaperones (14, 20 -23). Therefore, histone chaperones are likely candidates for the histone acceptor in the process of histone eviction in trans upon remodeling of PHO5 promoter chromatin.
In this study we investigated the effect of mutations in genes coding for various histone chaperones on the induction of the PHO5 gene. Of all five tested histone chaperones, only the lack of Asf1 showed an effect on PHO5 induction. The rate of chromatin remodeling was delayed, but, importantly, the final extent of chromatin remodeling and gene expression was not compromised.
We also extended our study to the PHO8 gene that codes for an alkaline phosphatase and is coregulated with PHO5 by the same transactivator Pho4 (24, 25). Here as well, chromatin remodeling upon induction led to histone loss from the promoter region and Asf1 contributed to the rate of histone eviction. Because PHO8 induction is strictly dependent on Snf2 (26), we conclude that Asf1 may cooperate with the remodeling complex SWI/SNF. However, Asf1 and Snf2 do not exclusively work in the same pathway, as we observed a synthetic effect of combined snf2 and asf1 mutations for the case of PHO5 induction.
Chromatin Analysis-Nuclei preparation and chromatin analysis by restriction enzymes or DNaseI as well as indirect end labeling, gel electrophoresis, and blotting procedures were as described (6,31,32). Quantification of restriction enzyme accessibility assays was done using phosphorimaging (Fuji FLA3000, AIDA TM software). The probe for the PHO5 locus corresponds to the ApaI-BamHI fragment upstream of the PHO5 promoter (probe D in Ref. 6), and ApaI (for DNaseI mapping) or HaeIII (for restriction enzyme accessibility) was used for secondary cleavage. The probe for the PHO8 locus corresponds to the PvuII-XhoI fragment downstream of the PHO8 promoter (25), and BglII (DNaseI mapping) or BglII and EcoRV (restriction analysis) were used for secondary cleavage.
Chromatin Immunoprecipitation-Yeast cultures of a density of 1-2 ϫ 10 7 cells/ml were treated with 1% formaldehyde for 20 min at room temperature. Cross-linking was quenched by adding glycine to 125 mM final concentration. Cells were washed twice in ice-cold 0.9% NaCl, resuspended in HEG150 buffer (150 mM NaCl, 50 mM HEPES, pH 7.6, 10% glycerol, 1% Triton X-100, 1 mM EDTA, 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride) and treated with a French Press three times at 1100 psi. This step lysed the cells and sheared the chromatin to an average size of 450 bp. Alternatively, glass beads were added and cells were lysed by shaking for 2 h in an Eppendorf shaker at 4°C. Subsequently the glass beads were removed, and chromatin was sheared by sonication to a similar average fragment size as with the French press. Immunoprecipitation was done as described (33). Antibodies against the C termini of histones H3 and H4, respectively, were gifts from A. Verreault. The anti-Pho4 antibody was commercially prepared (Abcam) using purified Pho4. Immunoprecipitated DNA was quantitated in duplicates by the ABI PRISM 7000 sequence detection system using the following amplicons:

RESULTS
The Kinetics of PHO5 Induction Are Delayed in the Absence of the Histone Chaperone Asf1-In the course of our mechanistic studies of chromatin opening at the PHO5 promoter, we examined a possible role of histone chaperones in the process that leads to histone eviction in trans (10,11). Therefore we tested whether induction of Pho5 activity, i.e. secreted acid phosphatase activity, was impaired in strains that are singly disrupted in various of the known histone chaperone genes in yeast: hir1, hir2, hir3, cac1, or asf1, as well as the double mutants asf1,hir3 and asf1,cac1 (12)(13)(14). All these strains could be induced to wild type levels of Pho5 activity by overnight incubation in phosphatefree medium ( Fig. 1 and data not shown). As complete remodeling of the PHO5 promoter chromatin is a prerequisite for induction of Pho5 activity (34), this result shows that neither of the tested histone chaperones is essential for chromatin remodeling at the PHO5 promoter.
The histone chaperone Asf1 was recently reported to be required for remodeling at the PHO5 promoter (35). Therefore we compared the chromatin structure of asf1 and wild type (wt) 2 strains by DNaseI mapping and restriction enzyme digestion, both under repressive conditions and after overnight incubation in phosphate-free medium. Under both conditions the asf1 mutant showed the same results as the wt strain (data not shown). We conclude that Asf1 is necessary neither for the assembly of the repressed state nor for chromatin remodeling upon induction.
We and others observed in the past that the absence of chromatinrelated factors like Gcn5 or Snf2, although not essential for chromatin opening at the PHO5 promoter, could lead to a delay in the induction process 3 (7,30,36,37). We also tested this possibility for the above histone chaperone mutant strains by monitoring induction kinetics. The strain deleted in the ASF1 gene showed a marked delay in PHO5 induction (Fig. 1). An equivalent delay was observed in two other strain backgrounds if deleted for ASF1 (W303 and BY4741, data not shown). The cac1 mutant also showed a significant delay in PHO5 induction. However, by monitoring restriction enzyme accessibility during induction kinetics (see below), we found that the delay in induction was not because of retardation on the level of chromatin opening (data not shown) as was analogously observed by Adkins et al. (35) in histone-loss kinetics during induction of a cac2 mutant. Accordingly, the induction of phosphatase activity in the asf1,cac1 double mutant was more delayed than in either single mutant but did not show a significant further delay on the level of chromatin opening as compared with the asf1 single mutant (data not shown). All other tested mutants did not significantly affect PHO5 induction (data not shown), and we focused our further studies on the role of Asf1.
The Absence of Asf1 Delays the Kinetics of PHO5 Induction on the Level of Chromatin Opening-We tested whether the kinetic effect of the asf1 mutation as measured on the level of PHO5 activity was because of a kinetic delay at the chromatin transition step. The course of chromatin remodeling at the PHO5 promoter during induction kinetics in asf1 and wt strains was monitored by measuring the accessibility of the intranucleosomal ClaI site in the PHO5 promoter region (6,32). A clear delay in nucleosome remodeling was observed ( Fig. 2A). In addition, we followed the loss of histones from the PHO5 promoter region by ChIP assay either with antibodies directed against the C terminus of histone H3 in the CY337 background (Fig. 2B) or with antibodies directed against the C terminus of histone H4 in the W303 background (data not shown). The ChIP assay was internally controlled for region specificity as amplicons corresponding to an open reading frame upstream ("upstream ORF") of PHO5 (Fig. 2B) did not detect significant changes in histone levels (7). This chromatin assay also showed a clear delay for the asf1 strain in both cases, but after prolonged incubation in phosphate-free medium no difference between the wt and asf1 strains was observed (Fig. 2B). Therefore, both quantitative chromatin assays confirmed that the rate of chromatin opening, rather than only other steps downstream in the gene expression process, is affected in the absence of Asf1.
It was shown previously that stable binding of Pho4 to its intranucleosomal binding site UASp2 requires chromatin opening (38). We therefore hypothesized that the observed delay in chromatin remodeling should also cause a delay in Pho4 binding to the promoter. Indeed, by ChIP assay using anti-Pho4 antibodies, we found in asf1 strains that Pho4 recruitment to the promoter was equally delayed as chromatin remodeling (Fig. 2C). Equivalent results were obtained in a W303 background (data not shown).
The Kinetic Delay in asf1 Strains Is Not Due to Effects Upstream of the Chromatin Remodeling Process-Previously, we took advantage of the PHO5 promoter variant 33 (P pho5v33 ), in which both Pho4 binding sites are replaced by Gal4 binding sites, to control for effects that a mutation might have upstream of the chromatin remodeling step at the PHO5 promoter (30,39). In such a variant, the same chromatin transition and resulting promoter activation take place upon galactose induction as with the wt promoter upon phosphate starvation (40). Therefore, any mutation that is thought to directly affect the chromatin remodeling step should also affect the chromatin transition in this variant.
The induction of the promoter P pho5v33 through the GAL pathway was similarly delayed in an asf1 strain (Fig. 3A) as the induction of the wt promoter through the PHO pathway (Fig. 1). This delay was not because of possible effects on the GAL signaling pathway or on Gal4 activity as the induction kinetics of the GAL1 promoter was unaffected by the asf1 mutation (Fig. 3B). The asf1 strain even showed a much higher extent of final induction for the P gal1 -lacZ construct than the wild type strain (data not shown). We note that asf1 strains tend to yield higher final values than the wt strains for basal and fully induced levels of PHO5 induction (Fig. 1) or for final Pho4 binding (Fig. 2C). However, such higher expression levels are not associated with changes in the chromatin structure (data not shown) and are probably due to downstream effects on gene expression.
The Combined Disruption of the SNF2 and ASF1 Genes Has a Synthetic Effect on the Induction Kinetics of PHO5-So far, no chromatin remodeler has been identified that is essential for opening PHO5 promoter chromatin. However, the absence of Snf2 leads to a marked kinetic delay in PHO5 induction 3 (7,36,37), indicating a significant role of the SWI/SNF complex in chromatin remodeling at the PHO5 promoter. We wanted to test whether the SWI/SNF complex cooperates with Asf1 in this remodeling process and, if so, whether they function  MARCH 3, 2006 • VOLUME 281 • NUMBER 9 exclusively in the same pathway or whether Asf1 also has a role in the alternative pathway through which PHO5 promoter chromatin is remodeled in the absence of Snf2.

Asf1 Function at the Yeast PHO5 and PHO8 Promoters
We constructed an asf1,snf2 double disruption strain and compared PHO5 induction kinetics in this strain with the kinetics in the single mutants (Fig. 4A). While the kinetics of PHO5 induction were strongly delayed in the snf2 strain, the additional absence of Asf1 caused a significant further delay, showing that Asf1 contributes to the rate of the alternative, Snf2-independent remodeling process. After overnight induction, however, the final activity level of the double mutant was similar to that of the single mutants (data not shown), arguing for full opening of PHO5 promoter chromatin even in the absence of both Snf2 and Asf1.
Deletion of the negative regulator Pho80 has previously been shown to lead to full opening of the chromatin structure at the PHO5 promoter even under otherwise repressive high phosphate conditions (41,42). Under such conditions the absence of Asf1 had no effect, or even slightly increased Pho5 activity, and the absence of Snf2 led to a modest decrease in activity. However, the level of induction of the asf1,snf2 double mutant was strongly reduced in comparison to that of the snf2 mutant (Fig. 4B). This demonstrated again a synthetic effect of this double deletion, suggesting that under such conditions the contribution of Asf1 to chromatin remodeling becomes important only in the absence of Snf2, i.e. as part of an alternative pathway. Nonetheless, the double mutant was still able to significantly induce PHO5 in a pho80 background under high phosphate conditions as compared with the corresponding repressed PHO80 strain (compare asf1,snf2,pho80 to asf1,snf2 in Fig. 4B).

The Kinetics of PHO8 Induction Are Also Delayed at the Level of Chromatin Remodeling in the Absence of Asf1-Induction of the PHO8
promoter is triggered by the transactivator Pho4 as well and also leads to a pronounced chromatin transition (25). Intriguingly, chromatin remodeling and the induction of this promoter depend critically on the activities of Snf2 and Gcn5 (26), much in contrast to the chromatin transition at the PHO5 promoter.
Adkins et al. (35) reported that Asf1 was essential for induction of PHO8. We measured PHO8 promoter activity by using a PHO8 promoter-driven lacZ reporter gene because alkaline phosphatase activity in yeast is due not only to the PHO8 but also to the PHO13 gene product (43). The PHO8 induction kinetics in phosphate-free medium were delayed in an asf1 strain, whereas the final level of activity was close to wild type levels (Fig. 5A).
We have shown here that induction of the PHO8 gene also led to histone depletion in the promoter region in agreement with recently published data (35). Furthermore, the kinetics of histone loss in asf1 strains were significantly delayed in comparison to wt strains (Fig. 5B), corresponding with the delay in promoter induction. The same immunoprecipitated DNA as in Fig. 2B was used, making additional control regions not necessary. Normalization of the data in Figs. 2B and 5B to the pre-induction values allows one to compare more directly the kinetic delay of histone loss at the PHO5 and PHO8 promoter regions (Fig. 5C). For unknown reasons, the histone occupancy at the PHO8 promoter prior to induction as measured by ChIP started off with lower

Asf1 Function at the Yeast PHO5 and PHO8 Promoters
absolute values in asf1 strains either by using antibodies directed against the H3 (Fig. 5B) or the H4 C terminus (data not shown).
Further support for a delay of PHO8 induction on the level of chromatin opening came from restriction enzyme accessibility measurements. Using the same nuclei preparations as in Fig. 2A we also tested the accessibility of restriction enzymes in the PHO8 promoter region that are known to change due to the chromatin transition upon induction (25). After 2 h of induction all three restriction sites tested were less accessible in asf1 than in wt cells (Fig. 5D). Nonetheless, we confirmed in several strain backgrounds by HpaI restriction site accessibility that remodeling in asf1 cells, even though delayed after 2 h of induction in phosphate-free medium, reached wild type levels after overnight induction (Fig. 5E). In addition, the constitutive induction of an asf1,pho80 double mutant in high phosphate medium led to the same open chromatin structure. Furthermore, the DNaseI pattern of the asf1 mutant after full induction was indistinguishable from the wild type pattern (Fig. 5F).
As was the case for the PHO5 promoter, at the PHO8 promoter no significant effect on the level of chromatin opening was observed with the hir3 or cac1 histone chaperone mutant strains (data not shown). Altogether, these data demonstrate that the rate of chromatin remodeling at the PHO8 promoter was compromised in the absence of Asf1, although not as strongly as at the PHO5 promoter (compare Figs. 2A and 5E and Figs. 2B and 5B), and that full remodeling was achieved after prolonged incubation.
Asf1 Is Critically Required for PHO5 Induction Only under Submaximal Inducing Conditions-As already mentioned, Adkins et al. (35) reported that Asf1 is essential for PHO5 and PHO8 induction, whereas we have shown here that full induction of both promoters was achieved in asf1 strains. Adkins et al. use phosphate-depleted medium for induction, and it is known that the phosphate depletion protocol (44) generates low phosphate medium that contains residual amounts of phosphate, in contrast to the no phosphate synthetic medium used in our study. Therefore, the apparent discrepancy between the results from our and the Tyler group could be due to the different induction media used.
Dhasarathy and Kladde (36) showed recently that the cofactors Gcn5 and Snf2, which are not necessary for complete PHO5 induction and chromatin opening under fully inducing conditions 3 (7,30,36,37), become essential under submaximal inducing conditions. The submaximal induction conditions were generated by allowing various amounts of residual phosphate in the induction medium. We used the same approach to examine the Asf1 requirement for PHO5 induction at various low phosphate concentrations. By increasing the phosphate concentration the induction of PHO5 was more severely affected in the asf1 mutant than in the wt cells (Fig. 6). At certain amounts of phosphate the induction of the asf1 mutant was entirely repressed, whereas induction of the wt remained mostly unaffected.

DISCUSSION
In this study we have shown that the histone chaperone Asf1 increases the rate of histone eviction during induction of the yeast PHO5 and PHO8 promoters. At both promoters the kinetics of chromatin remodeling and consequent induction were delayed in asf1 strains. Such a phenotype is reminiscent of earlier findings with PHO5 induction in strains deleted in the genes for the histone acetyltransferase Gcn5 or the ATPase subunit Snf2. Both deletions also lead to a kinetic delay in PHO5 induction on the chromatin level 3 (7,30,36,37). Interestingly, in the case of the coregulated PHO8 promoter that is critically dependent on Snf2 and Gcn5 (26) no mutations causing a kinetic delay are known so far. The asf1 mutation described here is the first of this kind.
Importantly, prolonged induction of the asf1 mutant under fully inducing conditions, i.e. in the absence of phosphate, led to the full extent of remodeling, and the resulting final activity levels were as high FIGURE 5. The kinetics of PHO8 induction are also delayed at the level of chromatin remodeling in the absence of Asf1. A, ␤-galactosidase activity was measured for the wt (CY337; closed circles) and isogenic asf1 strain (open circles), both carrying the plasmid pP pho8 -lacZ, during induction kinetics in phosphate-free medium. Error bars indicate the variation of two independent experiments. B, histone H3 occupancy at the PHO8 promoter during induction kinetics for wt (closed circles) and asf1 strain (open circles). The same immunoprecipitated DNA as in Fig. 2B was used for ChIP analysis but with primers for an amplicon in the PHO8 promoter region. Equivalent results were obtained using the W303 background and anti-histone H4 antibodies (not shown). C, the data for the PHO5 promoter in Fig. 2B and for the PHO8 promoter in panel B were normalized to the respective values at 0-h induction. D, using a nuclei preparation after 2 h of induction as in Fig or even higher than in wt strains. Therefore there is an alternative, Asf1-independent pathway for chromatin remodeling at the PHO5 and PHO8 promoters. This is not unexpected as histone chaperones constitute a notoriously redundant system in yeast (13). However, the normal mechanism of chromatin remodeling at both promoters in the wt situation appears to have some kind of specificity for Asf1 as the absence of four other histone chaperones; i.e. Hir1, Hir2, Hir3, or Cac1, did not show any effect.
A role for a histone chaperone in remodeling of PHO5 promoter chromatin agrees with the prediction that a remodeling mechanism leading to histone eviction in trans (10,11) should, in some way or other, involve a histone acceptor in trans as well. An involvement of histone chaperones would be less likely if histones were to leave the promoter region by a mechanism in cis. In turn, we use this argument to suggest that the observed contribution of the histone chaperone Asf1 to PHO8 promoter chromatin remodeling speaks for a mechanism of histone eviction in trans at this promoter as well.
Interestingly, induction of the GAL10 promoter was also shown to lead to the depletion of histones (45,46), but we have shown here that Asf1 does not influence the rate of induction of the coregulated GAL1 promoter. This may suggest that histone eviction at the GAL promoters does not occur via a trans pathway or that there is a specificity for a different histone chaperone. It clearly shows that Asf1 does not necessarily have a role in transcriptional activation in general.
Recently, another study also addressed the role of Asf1 in the induction of PHO5 and PHO8 (35). The authors came to the conclusion that neither the PHO5 nor the coregulated PHO8 gene could be induced in asf1 strains at all and therefore considered Asf1 as essential for transcriptional activation of these genes. As outlined above, our results are in agreement with the general conclusion that Asf1 plays a role in chromatin remodeling and the resulting activation of both genes, but we have shown here that chromatin remodeling at both genes can be achieved to wild type levels in the absence of Asf1. The apparent discrepancy between the two data sets can be explained by differences in the induction conditions used. It is reported that the extent of induction of the PHO regulon can be controlled by the phosphate concentration in the medium and that submaximally inducing conditions (low instead of no phosphate) lead to more pronounced dependences of the PHO5 gene on cofactors like, for example, Snf2 or Gcn5 (36). We obtained analogous results comparing the induction of PHO5 in an asf1 and wt strain at various levels of residual phosphate in the induction medium. There were intermediary induction conditions in which a wt strain is substantially induced while an asf1 mutant is uninduced. Adkins et al. (35) made use of induction medium that is prepared by a chemical depletion protocol (44). In contrast to phosphate-free synthetic medium, such phosphate-depleted medium contains residual amounts of phosphate and may even allow slow but continued cell proliferation. Therefore, such medium probably corresponds to intermediary inducing conditions as described above and by two other groups (36,47) and result in the more stringent dependence on Asf1. In our experiments, we examined the requirement for Asf1 during induction in phosphate-free medium that let cells go into growth arrest after overnight incubation (6), or we used the pho80 mutation that allows constitutive PHO induction even for continuously dividing cells in high phosphate media (41,42). Both protocols led to complete opening of the PHO5 and PHO8 promoters even in the absence of Asf1.
Our findings are in agreement with the recent results of a genomewide screen for yeast mutants defective for PHO5 regulation (48). This screen identified nine new genes that were not known before to be involved in PHO5 regulation. However, also in this study phosphatefree synthetic medium was used and the ASF1 gene was not found to be essential for PHO5 induction, either in the screen or when tested individually and in various strain backgrounds.
It is reported and discussed that chromatin remodeling complexes, for example the SWR or Ino80 complexes that catalyze the exchange of histones, work in concert with histone chaperones (14,21,22). The same is likely to be true for chromatin remodelers that catalyze the eviction of histones in trans. This eviction mechanism has not yet been properly reconstituted in vitro; however, there is ample evidence for its relevance in vivo (45,49), and chromatin remodeling at the PHO5 and PHO8 promoters were among the first examples (7, 9 -11).
The induction of both PHO5 and PHO8 is mediated by Snf2 with PHO8 induction being totally abolished (26) and PHO5 induction severely compromised in the absence of Snf2 3 (7,36,37). Here we have provided evidence that the SWI/SNF complex can work together with the histone chaperone Asf1, suggesting that the SWI/SNF complex is an example of a chromatin remodeler that can catalyze histone eviction in trans in vivo. However, our results indicate that the role of Asf1 in nucleosome eviction is not exclusively connected with the remodeling function of the SWI/SNF complex, i.e. that the SWI/SNF complex and the histone chaperone Asf1 work in distinct parallel pathways that functionally overlap. The SWI/SNF complex and Asf1 can cooperate with alternative histone chaperones or remodelers, respectively. The argument is as follows. Chromatin remodeling at the PHO8 promoter has to proceed via a SWI/SNF-dependent pathway. This pathway is delayed in asf1 strains, implying that Asf1 usually contributes to the outcome of this pathway but can be replaced by some other factor. Conversely, chromatin remodeling at the PHO5 promoter usually involves a SWI/ SNF-dependent pathway as well but can also be achieved in the absence of Snf2 through another pathway, albeit at a slower rate. This alternative pathway in turn can also cooperate with Asf1, as a deletion of Asf1 in a snf2 background leads to a synthetic kinetic effect with even slower induction rates and even lower steady state induction levels during induction via the pho80 mutation in high phosphate media. Strictly speaking, we cannot be sure whether the alternative pathways make use of alternative remodelers or histone chaperones at all. However, the alternative pathways in each case led to the same final outcome of remodeled chromatin structure, suggesting that the main features of the overall mechanism, e.g. histone eviction in trans through the combined FIGURE 6. Asf1 becomes critically required for PHO5 induction only under submaximal inducing conditions. YPDA, or phosphate-free medium supplied with phosphate (KH 2 PO 4 ) to the indicated final concentrations, was inoculated with logarithmically growing cultures of wt (CY337) or asf1 strains to a final A 600 of 0.01 and further incubated at 30°C overnight. Acid phosphatase activity was determined for each culture, and the error bars indicate the variation between two independent cultures. action of a chromatin remodeler and a histone chaperone, are maintained.