GAL4/GAL80-dependent nucleosome disruption/deposition on the upstream regions of the yeast GAL1-10 and GAL80 genes.

Metabolic reactivation (incubating spheroplasts with galactose and casamino acids) causes disruption of nucleosomes from the upstream regions of the yeast GAL1, GAL10, and GAL80 genes. The disruption is specific. It depends on the transcription activator Gal4; it only occurs in galactose-reactivated chromatin from galactose-grown cells; it only affects upstream region, gene-proximal nucleosomes. Due to this specificity and because some of the same regions have shown induction-dependent changes by in vivo analyses, we suggest that the nucleosome-disrupted structure produced by reactivation is the authentic chromatin structure for these regions under conditions of galactose-induced GAL1-10 and GAL80 expression. It is necessary to carry out a spheroplast reactivation treatment in order to observe this disrupted structure in nuclear chromatin because nucleosomes are redeposited onto these regions during the preliminary steps of nuclear isolation (cell harvest/spheroplast preparation) probably in response to the nonphysiological conditions associated with these steps. However, during the same isolation procedures in cells lacking Gal80 protein, there is no nucleosome deposition on these regions, and the in vivo disrupted structure remains present in the nuclear chromatin. Therefore, the nucleosome deposition process that operates in wild-type cells is dependent on Gal80 protein, defining another activity of this negative regulator.

Metabolic reactivation (incubating spheroplasts with galactose and casamino acids) causes disruption of nucleosomes from the upstream regions of the yeast GAL1, GAL10, and GAL80 genes. The disruption is specific. It depends on the transcription activator Gal4; it only occurs in galactose-reactivated chromatin from galactosegrown cells; it only affects upstream region, gene-proximal nucleosomes. Due to this specificity and because some of the same regions have shown induction-dependent changes by in vivo analyses, we suggest that the nucleosome-disrupted structure produced by reactivation is the authentic chromatin structure for these regions under conditions of galactose-induced GAL1-10 and GAL80 expression. It is necessary to carry out a spheroplast reactivation treatment in order to observe this disrupted structure in nuclear chromatin because nucleosomes are redeposited onto these regions during the preliminary steps of nuclear isolation (cell harvest/ spheroplast preparation) probably in response to the nonphysiological conditions associated with these steps. However, during the same isolation procedures in cells lacking Gal80 protein, there is no nucleosome deposition on these regions, and the in vivo disrupted structure remains present in the nuclear chromatin. Therefore, the nucleosome deposition process that operates in wild-type cells is dependent on Gal80 protein, defining another activity of this negative regulator.
GAL structural gene expression is tightly regulated at the transcriptional level by carbon source (1). GAL1-10, -7, and -2 are induced to very high levels of expression in galactose, via the activator Gal4p, 1 but are completely inactive in other carbon sources. The GAL4 gene and GAL80, the gene that encodes the GAL-specific negative regulator, are expressed in all carbon sources (1) as might be expected of regulatory genes, but their expression levels do vary with carbon source. For example, GAL80 expression occurs at a low basal level in glycerol but is induced 5-10-fold higher, via Gal4p, in galactose (2).
This gene family provides excellent opportunities to study the relationship of chromosome structure to gene regulation. Such studies have shown, for example, that under conditions of induced expression (galactose) Gal4p strongly protects the ma-jor promoter element, the UAS G , on GAL1-10 (3)(4)(5) and on GAL80 (6). Gal4p protection of the GAL1-10 UAS G is also strong in glycerol (3)(4)(5), even though the genes are completely inactive. This reflects the distinctive poised-for-expression status of the structural genes in this carbon source (7). Gal4p protection of the single UAS G on GAL80 is weaker in glycerol (6), perhaps to minimize interference with the basal promoter that drives GAL80 expression in glycerol (7).
The UAS G elements on GAL1-10 and GAL80 lie within constitutively nonnucleosomal, chromatin hypersensitive regions (4,6,8). In the uninduced expression state (glycerol, glucose), positioned nucleosomes are located between these hypersensitive regions and the genes, in the chromosomal copy (6, 8 -10) as well as in GAL-containing plasmids (10,11). The GAL10 TATA, the GAL1 TATA/transcription start site, and the GAL80 TATA/transcription start site are contained in these nucleosomes. In vivo analysis detected structural changes in two of the GAL1-10 intergenic nucleosomal regions upon galactoseinduction of expression (5). However, no such changes were detectable in nuclear chromatin isolated from induced cells (9). In this paper, we describe experiments that allow us to detect intergenic nucleosome changes in nuclear chromatin, thus reconciling the in vivo and nuclear results. In addition, these experiments yield some unexpected insights on nucleosome disruption/deposition processes taking place on gene control regions.

MATERIALS AND METHODS
Strain 21R yeast cells were grown as described previously (9) in glucose/glycerol/ethanol (D); galactose/glycerol/ethanol (G); glycerol/ ethanol (g). These media allow growth of any cells in any medium, even 4 D cells (disrupted GAL4) in galactose, and thus allow the most controlled comparison of the various strains. Yeast spheroplasts were made as described in Ref. 12. Metabolic reactivation of spheroplasts consists of adding a carbon source to 2%, usually galactose in these studies, and casamino acids (to 0.5%) to the spheroplasts during the last few minutes of the ϳ25-min incubation with the cell wall lytic enzyme Oxalyticase (12). For our strains, the typical time exposed to reactivation conditions is 10 -12 min. If reactivation media are added too early in the incubation with Oxalyticase, spheroplasting efficiency is affected. If the total incubation with Oxalyticase is allowed to proceed too long, spheroplasts begin to lyse. 2 The choice of typical conditions (10 min of reactivation, 25 min of total incubation with Oxalyticase) for this set of strains is sufficient for producing striking effects on chromatin (see below) while avoiding significant spheroplast lysis.
Nuclei were isolated by method I (12), and DNase I and MNase digestions were performed on the nuclei as described therein. DNase I and MNase cleavage sites were located by an indirect end label approach (13,14). The GAL1-10 patterns in Figs. 1, 2, and 5 were obtained by mapping from the essentially coincident TaqI (see Fig. 1) or EcoRI (see Figs. 2 and 5) sites, which lie within the coding sequence of GAL10, ϳ170 bp downstream of the GAL10 transcription start site (cf. Ref. 9). The patterns in Fig. 4 were obtained by mapping from the EcoRI site within the coding sequence of GAL1, which lies ϳ1100 bp downstream from the GAL1 transcription start site (cf. Ref. 8). The patterns in Fig. 6 were obtained by mapping from the EcoRI site within the coding sequence of GAL80, which lies ϳ400 bp downstream of the GAL80 transcription start site (cf. Ref. 6). DNA from the nuclear digests was purified (cf. Ref. 9). Samples with similar and suitable digestion extents were chosen by analysis of the bulk patterns on mini-gels, cut with the appropriate restriction enzyme, and electrophoresed on denaturing gels (DNase I nuclear footprints) or native gels (MNase profiles) as described previously (9). DNA was electrophoretically transferred to DBM paper using a homemade electroblot apparatus. For all hybridizations, small (50 -115 bp) probes abutting the mapping restriction sites were used so that probe homology ends below the smallest DNA bands on any of the patterns shown in this work. The probe was radiolabeled by random priming (15) so both strand patterns are detected in all profiles. The DBM paper was washed and exposed to Kodak XAR film as described previously (9).

Why Is the Chromatin Structure of the GAL1-10 Intergenic
Region Not Expression-sensitive in Nuclear Chromatin as It Is in Vivo?-GAL1-10 intergenic chromatin consists of an ϳ170-bp nonnucleosomal, UAS G -containing hypersensitive region (4,8), which is surrounded by positioned nucleosomes (8 -11). The DNase I nuclear cleavage site pattern (Fig. 1) shows generally strong cleavage within the hypersensitive region (4, track 3), except for the UAS G elements, which are protected by Gal4p; the nucleosomal regions, upstream of GAL10 and between the HR border and the GAL1 TATA, are relatively protected (E, track 3). These chromatin features are not present in naked DNA digests (Fig. 1, tracks 1 and 8) and show little dependence on digestion extent (cf. Fig. 1, tracks 3  and 4). The MNase cleavage site pattern ( Fig. 2A) is a high intensity ladder of bands at ϳ170-bp intervals across the intergenic region (␣, ␤, ␥, ␦, ⑀; tracks 2 and 4), with little or no cleavage between these bands. It has been shown that the lack of cleavage in the interband regions ␣-␤, ␥-␦, and ␦-⑀ is due to nucleosome protection (8,9). Regions ␣-␤ and ␥-␦ correspond to the E regions in Fig. 1. Region ␤-␥, which is the nonnucleosomal hypersensitive region, is simply not cut by MNase in limited digests, even as naked DNA (8,9).
As reported previously (9), the intergenic region nuclear chromatin profiles from induced cells, which actively express the GAL1-10 genes, are basically the same as profiles from cells in which GAL1-10 are inactive, whether the inactivity is due to growth in a noninducing carbon source (Fig. 1, compare tracks 2 and 3) or to the absence of the transcription activator Gal4p ( Fig. 2A, compare tracks 4 and 2). The single difference around the GAL1 5Ј end in MNase profiles will be discussed below. This basic similarity would suggest that the chromatin structure of the intergenic region is insensitive to gene activity (9). However, in vivo analysis (5) detected induction-dependent structural changes within two nucleosomal portions of the intergenic region (ϫ to the right of Fig. 1, track 5, and Fig. 2A, track 3). Since nucleosome structure on upstream regions may be an aspect of gene control (16,17), we wanted to determine why these nucleosomes do not seem to be sensitive to gene activity in our nuclear chromatin. This issue was particularly perplexing because intergenic nucleosome changes were observed in a previous nuclear chromatin analysis (8), which used a yeast strain of uncharacterized GAL genotype. The strains used in Ref. 9 (and below) comprise an isogenic series with a well characterized GAL pedigree and are widely used in GAL analysis. Note that in vivo and nuclear chromatin approaches only differ for the nucleosomal regions; the two approaches give similar results concerning UAS G protection and the absence of induction-dependent change in the nonnucleosomal hypersensitive region.
Galactose-and Gal4p-dependent Intergenic Nucleosome Changes Are Observed in Nuclear Chromatin When Nuclei Are Isolated from Metabolically Reactivated Spheroplasts-Isolating yeast nuclei involves a number of cell washing, harvesting, and spheroplasting steps that require 1-2 h at 0 -4°C in nonphysiological solutions (cf. Ref. 12). This corresponds to starvation conditions for yeast cells. It was shown some time ago that incubation of spheroplasts in nutrients (a carbon source plus casamino acids) just prior to their lysis enhances the transcription level in nuclei isolated from the treated spheroplasts (18). This treatment restores the spheroplasted cells to a metabolically active state, allowing them to recover from the starvation during the preliminary steps. To test whether this spheroplast reactivation treatment, which is not ordinarily used, might affect GAL1-10 chromatin structure, we have carried out comparative experiments in which cells harvested from a single culture were divided just prior to the final step in making spheroplasts, treatment with lytic enzyme. One aliquot was reactivated (see "Materials and Methods") during the lytic treatment; the other was not. Nuclei were isolated from both sets of spheroplasts and digested with nucleases in parallel. The two sets of chromatin digests only differ in the presence of nutrients for a few minutes during spheroplast formation, thus FIG. 1. DNase I digestion patterns from the GAL1-10 intergenic region. DNase I cleavage sites were mapped from the EcoRI site within GAL10 by indirect end label analysis on 4.7% polyacrylamide, 0.6% agarose, 7 M urea denaturing gels, as described under "Materials and Methods." The tracks show from left to right: track 1, a naked DNA digest; track 2, a digest from glycerol/ethanol-grown wild-type cells; tracks 3 and 4, two digests (of differing digestion extents) from wild-type galactose-grown cells; track 5, a digest from 80 D (disrupted GAL80) galactose-grown cells; tracks 6 and 7, two digests (of differing digestion extents) from wild-type galactose-grown cells in which spheroplasts were reactivated with galactose (G R ); track 8, a higher exposure of the lower part of the naked DNA profile shown in track 1. DNA sizes were determined by comparison to the mobilities of pBR322/MspI marker fragments run on the gel (numbers to the right of track 8): 1, 622 nucleotides; 2, 527 nucleotides; 3, 404 nucleotides; 4, 307 nucleotides; 5, 242 nucleotides). The map to the left shows the DNA sequence organization on the GAL1-10 intergenic region: UAS G elements (brackets); the TATA boxes (T); GAL1 and GAL10 transcription start sites (wavy lines). The lower case t to the right of track 6 are TATA-like motifs. The location of the hypersensitive region (HR) is shown to the left of track 2. Other symbols are discussed in the text. excluding chromatin differences due to variations in cell growth or in day to day nuclear isolation/nuclease digestion conditions.
Including a spheroplast reactivation step in the isolation of nuclei from galactose-induced cells results in exposure of the E protected chromatin regions to DNase I cleavage ( Fig. 1, tracks 6 and 7), and of the ␣-␤, ␥-␦, and ␦-⑀ interband regions to MNase cleavage ( Fig. 2A, track 3). The exposure of DNA that is nucleosome-protected in unreactivated chromatin indicates that reactivation causes a significant nucleosome structural alteration, either unfolding or nucleosome loss. Reactivation also causes a loss of the high intensity MNase ladder pattern, due to reduced cleavage at sites ␤, ␥, and ␦ (Fig. 2B, tracks 2-4 and 7). All of the reactivation-induced changes depend on functional Gal4p (Fig. 2B, tracks 4 and 5) and can only be triggered when cells are grown and spheroplasts are reactivated in galactose. Even spheroplasts from glycerol-grown cells show no evidence of nucleosome alterations (Fig. 2B, track 6), even if reactivated with galactose for up to 25 min (not shown), a period of time that is longer than the time required for galactose to induce full GAL1-10 expression in cells growing in glycerol. Reactivating spheroplasts from induced cells with nongalactose carbon sources also fails to produce any GAL1-10 chromatin changes (not shown). Thus, when isolated from reactivated spheroplasts, nuclear chromatin demonstrates galac-tose-and GAL4-dependent change in the GAL1-10 intergenic nucleosome regions.
We can exclude a number of artifactual explanations for the DNA exposure produced by reactivation. Endogenous nuclease activity is negligible in both types of nuclei (not shown) and thus cannot explain the results. Spheroplast preparation can cause a heat shock-like response, which can be prevented by azide treatment of cells prior to harvest 3 ; reactivation effects are the same with or without this treatment (not shown). Treatment with cycloheximide just prior to harvest does not alter the reactivation effect (not shown), so protein synthesis is apparently not required. Most importantly, histone profiles of nuclei from reactivated or unreactivated spheroplasts are very similar (Fig. 3, tracks 1 and 2). Thus, the reactivation-induced DNA exposure does not reflect a proteolysis artifact.
Reactivation Affects all Three GAL1-10 Intergenic Nucleosomes, Probably Equally; Reactivation-induced Changes Are Restricted to These Intergenic Nucleosomes-The response of the three intergenic nucleosome regions to reactivation appears to vary in extent. However, we suggest that reactivation causes a similar and significant change on all three regions because reactivation causes each of the three to become similar 3 W. Garrard, personal communication. Other symbols are described in the text. Note that the intensity response in these blots is size dependent (19). This causes ␣-␤ region intensity to be generally diminished relative to the other interband regions so that longer exposures are required to visualize ␣-␤ interband intensities in the reactivated profiles.
to its corresponding naked DNA profile in both relative intensity level and cleavage pattern. For example, in reactivated chromatin profiles interband region ␥-␦ has the highest intensity, as it does in naked DNA profiles, and as in the naked profiles, this high intensity is limited to the GAL1-proximal half of the region ( Fig. 2A, tracks 1 and 3). On the other hand, ␦-⑀ region intensity is lower but spread through the whole region. In the ␣-␤ region, both DNase I (Fig. 1, tracks 6 and 7 versus 8), and MNase (not shown) reactivated chromatin cleavage patterns resemble naked patterns. Note particularly the region around the GAL10 TATA (Fig. 1). The similarities of reactivated and naked DNA profiles and the fact that changes are always observed in each of the three regions suggest that the same change occurs on all three nucleosomes and that the change involves major or complete nucleosome loss, but the possibility that any or all of these nucleosome changes are only partial or reflect only nucleosome unfolding cannot be excluded. We will refer to these changes as nucleosome disruption. The ␥-␦ nucleosome was suggested by in vivo approaches to be disrupted (20).
Reactivation produces a strong MNase band at the GAL1 5Ј end. This band has similar mobility to a band in the naked DNA profile and is a doublet like the naked band (Fig. 2B,  tracks 1, 3, and 7). It probably reflects exposure of the DNA at the GAL1 5Ј end due to the reactivation-induced loss of the ␦-⑀ nucleosome that covered this site. The GAL1 5Ј end region is extensively protected in unreactivated profiles from induced cells, over an even larger region than the nucleosome protection seen in the inactive state ( Fig. 2A, tracks 2 and 4). This extensive protection is the one feature that differs between MNase digests of inactive and induced (unreactivated) chromatin. It may reflect the presence of stalled transcription-associated complexes near the GAL1 5Ј end due to the nonphysiological conditions during harvest/spheroplast pretreatment (below).
Reactivation produces no detectable change in the UAS G footprint or within the hypersensitive region (Fig. 1, tracks 6  and 7) and, except for the extreme 5Ј end, has no effect on the GAL1 intragenic chromatin pattern. As shown previously (8, 10), galactose-induction of expression abolishes the very regular intragenic nucleosome pattern of inactive GAL1 (Fig. 4, track 4), resulting in more heterogeneous nuclease cutting and new bands ( , Fig. 4, track 5). However, reactivated and unreactivated patterns from galactose-induced cells are quite similar (Fig. 4, tracks 5 and 6). In these same profiles, the striking reactivation-dependent changes on the GAL1-10 intergenic region can be seen clearly (Fig. 4, tracks 2 and 3). Thus, reactivation only affects the intergenic nucleosomes.

The GAL1-10 Intergenic Nucleosomes Are Disrupted without Reactivation in Cells That Lack Gal80p, Providing Evidence That Gal80p Plays a Role in Nucleosome Occupancy on This
Region-Nuclear chromatin isolated from galactose-induced cells that lack functional Gal80p (80 D cells) shows the same features that are produced in wild-type chromatin by reactivation: nuclease accessibility of the DNA within the E protected regions (Fig. 1, track 5) and the ␣-␤, ␥-␦, ␦-⑀ interband regions (Fig. 5, track 5) and weak cleavage at sites ␤, ␥, and ␦. However, these features are present in unreactivated chromatin from cells that lack Gal80p. Moreover, reactivation of these spheroplasts does not produce any further chromatin change (Fig. 5,  track 4). Thus, the nucleosome-disrupted structure produced by reactivation in wild-type is present without reactivation in nuclear chromatin isolated from cells lacking Gal80p.
The chromatin regions affected by reactivation in wild-type spheroplasts include most of the sites affected in vivo by the induction of GAL1-10 expression (ϫ, Figs. 1 and 2). Because of this similarity and the association of the reactivation changes with conditions specific to induced GAL1-10 expression (Gal4p, galactose), we suggest that the nucleosome-disrupted structure produced by reactivation is the authentic and in vivo active chromatin structure for the GAL1-10 intergenic region. Why then are nucleosomes found on this region in unreactivated nuclear chromatin from wild-type induced cells? The most likely explanation is that these nucleosomes were deposited on the region during the 1-2 h of cell harvest/spheroplast pretreatment, probably in response to the nonphysiological (0°C/ starvation) conditions associated with these steps. Reactivation then causes disruption of these deposited nucleosomes, restoring the in vivo structure. On the other hand, the nucleosomedisrupted structure is maintained through these same steps when cells lack Gal80p. Thus, nucleosome deposition does not occur on the intergenic regions during the isolation steps when Gal80p is absent. Since the same isolation protocol is used for both wild-type and 80 D and they are isogenic strains, the absence of nucleosome reprotection in 80 D cells indicates that the nucleosome deposition process that occurs in wild-type during isolation depends on Gal80p, directly or indirectly. This explains why GAL80-dependent structural features were observed in nuclear chromatin (9) but not in vivo (5).
Reactivation Causes Disruption of the Gene-proximal Upstream Nucleosome on the GAL80 Gene-As is the case on GAL10 and GAL1, the UAS G on GAL80 lies within a constitutive, nonnucleosomal HR (Fig. 6, track 1) that is immediately flanked on the gene-proximal side by a positioned nucleosome (D, Fig. 7). This nucleosome contains the TATA and transcription start sites of GAL80. It protects the region in chromatin from cells grown in noninducing carbon sources or in unreactivated chromatin from galactose-induced cells (Fig. 6, tracks 1,  3, and 4). However, reactivation (Fig. 6, tracks 2, 5, and 8) results in strong nuclease cutting at the GAL80 transcription start site (ϩ, track 2), which lies near the center of this nucleosome and is completely protected in unreactivated/uninduced digests, and at a site between the UAS G and TATA, which lies near the nucleosome edge and which is only weakly cleaved in unreactivated/uninduced chromatin. These enhanced cleavages indicate that reactivation causes a significant nucleosome alteration, i.e. disruption, which exposes the DNA in this region to cleavage. That only limited changes are observed probably reflects the lack of strong naked DNA cutting sites within this nucleosomal region (Ref. 6; Fig. 6, track 0). As on GAL1-10, reactivation-induced changes are GAL4-dependent (Fig. 6,  tracks 7 and 8) and only occur in galactose-grown and galactose-reactivated chromatin (Fig. 6, track 6). The similarities to GAL1-10 suggest that this GAL80 upstream nucleosome is disrupted under induced conditions in vivo and replaced on the TATA/5Ј region during cell harvest/pretreatment by a GAL80dependent process. We cannot test the Gal80p-dependence of this replacement because the disruption site used to make 80 D cells lies within this nucleosome. However, GAL80 expression is known to be regulated by Gal80p (2), so it is at least plausible that Gal80p could affect nucleosome presence on its own upstream region. Note that the GAL80 profiles were obtained from some of the same DBM papers, and chromatin digests, used to obtain the GAL1-10 profiles in Figs. 2 and 4.
Reactivation also results in some protection within the hy- persensitive region, near the GAL80 UAS G (Fig. 6, tracks 2, 5,  and 8). This protection is GAL4-dependent (Fig. 6, tracks 7 and  8), but whether it is due to Gal4p or to other elements of the transcription apparatus is not clear. Gal4p protects the GAL80 UAS G strongly, with or without reactivation, in chromatin from induced cells, (Ref. 6 and data not shown).

DISCUSSION
Nucleosome Disruption-Metabolic reactivation with galactose results in exposure of GAL1, GAL10, and GAL80 DNA sequences that are nucleosome-protected in unreactivated, or inactive, chromatin (Fig. 7). Thus, reactivation causes a disruption of the structure of these nucleosomes. The nucleosomes affected lie between the transcription units and the UAS Gcontaining hypersensitive regions and contain the GAL1, GAL10, and GAL80 TATA and the GAL1 and GAL80 transcription start sites. One of the GAL1-10 nucleosomes, B (Fig. 7), was shown by photofootprinting in vivo to undergo disruption when GAL1 expression is induced by galactose (20). For this reason and because the reactivation-induced changes depend on Gal4p and are specific to galactose-grown/galactose-reactivated chromatin, it is likely that the disrupted structure produced by reactivation is the authentic chromatin structure for these upstream sequences under conditions of galactose-induced expression and that the nucleosome-disruption process triggered by galactose reactivation in spheroplasts is a process that normally acts in cells during induction of GAL1-10 and GAL80.
A chromatin response to spheroplast treatments is not limited to GAL genes nor to gene activation. Incubation of spheroplasts with phosphate results in specific nucleosome deposition on the PHO5 promoter region, assayed, as here, in nuclei isolated from treated spheroplasts (21). The presence of these nucleosomes is associated with the inactive state of PH05, and phosphate is the normal inactivation signal. Thus, on PH05 as on the GAL genes, a metabolic signal can exert its gene-specific chromatin effect in spheroplasts. Such responsiveness should not be unexpected; spheroplasts are intact cells and on solid media can regenerate a cell wall and resume normal growth. Spheroplast treatment should thus be included in all studies of gene-specific yeast chromatin structure.
Nucleosome Redeposition-We suggest that it is necessary to carry out a spheroplast reactivation treatment in order to observe the authentic (nucleosome-disrupted) GAL1-10/GAL80 upstream structure in nuclear chromatin from induced cells because nucleosomes are deposited back onto these regions during the steps of cell harvest/spheroplast preparation, probably in response to the nonphysiological conditions (0°C/starvation) encountered during these steps. However, under the same conditions, such nucleosome deposition does not occur in cells that lack Gal80p, and the in vivo nucleosome-disrupted structure remains present through the isolation. Thus, the nucleosome deposition that occurs in wild-type cells during these steps depends on Gal80p. The structure produced by this "isolation-induced" nucleosome deposition is quite similar to the inactive chromatin structure produced by GAL-specific regulatory processes, as seen for example in chromatin from glycerol-or glucose-grown cells. Thus, the same Gal80p-dependent process may also function in vivo to place nucleosomes on these upstream regions in response to GAL-specific inactivation. In this regard, it is noteworthy that the GAL1-10 intergenic region is never as completely nucleosome-protected in cells that lack Gal80p function, even in glucose, as it is in wild-type FIG. 7. Reactivation-induced changes in the GAL1-10 and GAL80 upstream regions. The sequence organization of the GAL1-10 and GAL80 upstream regions are shown on the center lines. The transcription start sites (arrows), the TATA (T) and UAS G are located to scale on these lines. The location of upstream nucleosomes whose presence is altered by spheroplast reactivation are shown as stippled strips below the center lines, and labeled, A-D. The GAL1 and GAL80 start sites lie within nucleosomes C and D, respectively. Above the central lines, the MNase cutting sites (␣, ␤, ␥, ␦, ⑀) in unreactivated or inactive chromatin (i/u) are shown by filled arrowheads. The GAL1-10 sequences exposed to DNase I (D) or MNase (M) by reactivation are located by regions of ϩ separated by dashes. The specific GAL80 sequences exposed to MNase by reactivation are located by ϩ, the sequences which become more protected by E. inactive chromatin (9). Gal80p loss also permits low level GAL1 expression in glucose (22). This expression could be the result of incomplete nucleosome occupancy of the intergenic region; in wild-type cells, partial nucleosome loss produced by histone depletion allows a roughly similar, low level of GAL1 expression in glucose (23).
The Role of the GAL-specific Factors-The observations in this work describe a link between upstream region nucleosome occupancy and GAL gene regulation, implemented via the GAL-specific regulatory factors. Upstream nucleosome removal is associated with gene activity and depends on the Gal4p activator; upstream nucleosome replacement, which is associated with gene inactivity, depends on the negative regulator Gal80p. Gal4p and Gal80p are in contact in all carbon sources (7), including galactose (24). Moreover, the contact site of Gal80p with Gal4p lies in the Gal4p transcription activation region (1, 7), the region whose activity is associated with the nucleosome removal process. The structural organization of these two factors is thus consistent with the suggestion that they function in reciprocal nucleosome disruption/deposition processes. The limitation of these reciprocal processes to the upstream/5Ј region nucleosomes is also consistent with this suggestion because Gal4p-Gal80p are probably restricted to these regions. The occurrence of these processes on both GAL1-10 and GAL80 suggests they may be a general GAL feature. Gene expression also affects the nucleosomes within genes (cf. Ref. 25), including GAL1 (8, 10). However, GAL1 intragenic patterns are not sensitive to reactivation, and therefore, the intragenic nucleosomes must not be subject to a GAL80-dependent reorganization process like that acting on the intergenic nucleosomes. Intragenic nucleosome changes may be due to polymerase passage (25) and thus reflect a different process than the upstream nucleosome changes described in this work. Cavalli and Thoma (10) have suggested that the loss of the GAL1 intragenic pattern in galactose reflects mainly nucleosome rearrangements, with at most partial nucleosome loss.
The GAL4-dependence of nucleosome disruption and GAL80dependence of nucleosome deposition could reflect the direct action of these factors or indirect action mediated via other factors. For example, Gal4p might act via Swi/Snf (26), which have been implicated in the disruption of TATA-bound nucleosomes in vitro (27). Concerning the nucleosome target, it has been observed that removal of histone H4 N-terminal tails results in a significant reduction in the in vivo level of GAL1induced expression (28). This reduction might reflect involvement of the H4 tails in mediating the Gal4p-dependent nucleosome disruption associated with induction. Conversely, removal of histone H3 N-terminal tails results in an increased level of GAL1 induced expression (29), roughly similar to that produced by removal of Gal80p function (22). Perhaps H3 tails mediate Gal80p-dependent nucleosome deposition and loss of these tails (or of Gal80p) hinders this process, thereby affecting expression.
A Model for Upstream Region Nucleosome Occupancy-These upstream nucleosome disruption/deposition processes can apparently take place rapidly. Disruption effects can be detected by 3 min and are well in place by 9 min of reactivation treatment (Fig. 2B), while deposition in our experiments occurs sometime within the 1-2 h of cell harvest/spheroplast preparation. However, nucleosome deposition may also be quite rapid; on PHO5 it occurs within minutes (21). This rapidity and the involvement of opposing regulatory functions in these reciprocal disruption/deposition processes suggest that nucleosome presence on the TATA/start site regions might involve an ongoing competition of Gal4p-dependent nucleosome disrup-tion and Gal80p-dependent nucleosome deposition. Activation conditions (signals) favor disruption; inactivity, or simply a lack of activation signals, favors deposition. Such a competition could facilitate rapid changes in expression state and fits the observation that the precise degree of disruption of the ␥-␦ nucleosome correlates with the precise strength of the Gal4p activator (20). The increased level of GAL1-induced expression when Gal80p function is removed indicates that Gal80p normally modulates GAL1 expression in the induced state (7). This modulation might involve Gal80p maintenance of at least partial nucleosome occupancy on the GAL1 upstream region in galactose, thus reflecting the operation of this competition even in the active state.
Nucleosome loss on upstream regions in association with gene expression is common on yeast and other eukaryotic genes (16,17). On many genes, the nucleosomes are lost from genespecific promoter elements. However, on GAL1-10 and GAL80, the specific promoter elements, the UAS G , lie in constitutively nonnucleosomal regions (4,6,8). The disrupted nucleosomes come from the TATA/transcription start site regions, and their disruption/deposition occurs without any effect on activator-UAS G interactions. Thus on GAL genes, the activator-specific promoter interaction and disruptable nucleosome occupation probably reflect distinct levels of control. This can allow cells to respond to fluctuating carbon sources or growth conditions in stages. In noninducing carbon sources like glycerol, the activator is bound to the UAS G , but subsequent stages of expression, which involve upstream nucleosome disruption, are not implemented. If galactose becomes available, the inhibition at these subsequent stages can be quickly released. Transcription-associated DNA melting occurs ϳ20 bp downstream of the GAL1 and GAL10 TATA (30), in regions that are exposed by the upstream nucleosome disruption. In addition to making this DNA more accessible, nucleosome disruption might also aid the DNA melting process by liberating the constrained negative supercoiling residing in these upstream nucleosomes. Negative supercoiling is generated behind a transcribing polymerase (31). Negative supercoiling favors nucleosome formation and thus would favor the redeposition of nucleosomes on these TATA/start site sequences after a polymerase initiates transcription. Transcription initiation might therefore involve a cycle of upstream nucleosome disruption/redisposition. Such a cycle is consistent with the structural view presented above, of a balanced competition of Gal4p-dependent nucleosome disruption and Gal80p-dependent nucleosome deposition processes operating on the upstream nucleosomes, and may play a facilitating role in the events that take place at transcription initiation of GAL genes. We also note that the need for metabolic reactivation in the disruption of GAL nucleosomes is consistent with the recent demonstration that nucleosome disruption is energy requiring (32).