(Received for publication, January 16, 1997, and in revised form, February 20, 1997)
From the Molecular Genetics Program, Wadsworth Center, New York State Department of Health and State University of New York School of Public Health, Albany, New York 12201-2002
We examine the generality of transcription factor-mediated chromatin remodeling by monitoring changes in chromatin structure in a yeast (Saccharomyces cerevisiae) episome outside of the context of a natural promoter. The episome has a well defined chromatin structure and a binding site for the transcription factor GAL4 but lacks a nearby functional TATA element or transcription start site, so that changes in chromatin structure are unlikely to be caused by transcription. To separate changes caused by binding and by activation domains, we use both GAL4 and a chimeric, hormone-dependent activator consisting of the GAL4 DNA-binding domain, an estrogen receptor (ER) hormone-binding domain, and a VP16 activation domain (Louvion, J.-F., Havaux-Copf, B. and Picard, D. (1993) Gene (Amst.) 131, 129-134). Both GAL4 and GAL4·ER·VP16 show very little perturbation of chromatin structure in their nonactivating configurations. Substantial additional perturbation occurs upon activation. This additional perturbation is marked by changes in micrococcal nuclease cleavage patterns, restriction endonuclease accessibility, and DNA topology and is not seen with the nonactivating derivative GAL4·ER. Remodeling by GAL4·ER·VP16 is detectable within 15 min following hormone addition and is complete within 45 min, suggesting that replication is not required. We conclude that activation domains can exert a major influence on chromatin remodeling by increasing binding affinity and/or by recruitment of other chromatin remodeling activities and that this remodeling can occur outside the context of a bona fide promoter.
Transcriptional induction in eukaryotes requires binding of activators to promoter elements termed enhancers (called upstream activator sequences (UASs)1 in yeast). Binding of a transcriptional activator in its activating form results in recruitment or reconfiguration of a preinitiation complex in a form competent to allow initiation and elongation by RNA polymerase II. These events occur in the context of chromatin, which means that histone proteins are available to compete with activators and components of the preinitiation complex for binding to DNA. One function of transcriptional activators may be to help overcome nucleosomal repression of transcription, particularly at the site of preinitiation complex formation (1).
Gene activation is often accompanied by perturbations in chromatin
structure. For some genes, this perturbation appears to be essential
for proper transcriptional regulation (2-4). In some instances,
specific effects of activation domains have been observed. The yeast
activator GAL4 can bind to promoter sites in both activating and
nonactivating forms; this has allowed the detection of subtle effects
on chromatin structure in the GAL1-10 promoter that require
an unmasked activation domain (2, 5-6). Another transcriptional
activator, PHO4, perturbs chromatin structure at the PHO5
promoter in an activation domain-dependent manner (7). The
thyroid hormone receptor-retinoid X receptor heterodimer binds without
perturbing chromatin structure in its unliganded state; the addition of
hormone results in a local perturbation of chromatin structure in the
TRA gene promoter (8).
The above examples of chromatin remodeling by transcription factors occur at natural promoters. Although chromatin remodeling under activating conditions can occur in nonfunctional promoters having mutated TATA boxes (3, 9), other cis-acting promoter elements could still be involved in this process. Perturbation of a positioned nucleosome by GAL4 has been examined outside of the context of a natural promoter in a plasmid episome in yeast (10). In this study, the GAL4 site was near the center of a positioned nucleosome (when GAL4 was absent); GAL4 expression resulted in micrococcal nuclease (MNase) cleavage sites, indicative of nucleosome perturbation. This perturbation appeared stronger, as assessed by the intensities of the MNase cleavages, under activating conditions.
The mechanism by which an activation domain causes chromatin remodeling
remains unknown. To examine the generality of transcription factor-mediated chromatin remodeling and to increase our understanding of this important step in transcriptional activation, we have examined
the ability of GAL4 and a chimeric activator, GAL4·ER·VP16, to
perturb chromatin structure in the yeast episome TALS, which has an
exceptionally well defined chromatin structure. In yeast haploid cells, the repressor complex
2·MCM1 helps to incorporate TALS into
strongly positioned nucleosomes (11). One of these nucleosomes contains
a GAL4 binding site, which derives from the GAL3 promoter
(12). We show that both GAL4 and GAL4·ER·VP16 perturb TALS
chromatin most strongly in their activating configurations, and we
discuss possible mechanisms by which an unmasked transcriptional activation domain could remodel chromatin in vivo.
Introduction of TA1780 into yeast has been
described (10). TALS (11) and TALS4 (13), in which the
UASGAL has been removed by mutagenesis (generous gift of M. Kladde), were excised from pUC19 sequences by digestion with
HindIII and religated before introduction into yeast.
GAL4·ER·VP16 was expressed from the plasmid pHCA, here referred to
as pRS313GAL4·ER·VP16 (Ref. 14; generously provided by D. Picard)
or from pRS416GAL4·ER·VP16, which was constructed by ligation of
the ClaI-SpeI fragment encompassing the
GAL4·ER·VP16 gene from pRS313GAL4·ER·VP16 into pRS416 (Ref.
15). The expression vector for GAL4·ER was constructed by deletion of
the SacI fragment bearing the VP16 coding sequence from the
expression vector for GAL4·ER·VP16. For experiments in which
perturbation of chromatin by GAL4 was monitored, yeast cells contained
the multicopy GAL4 expression vector pRS425GAL4 (10). The plasmid pCL1
(16) was used to express GAL4 from the ADH1 promoter. The
reporter construct used for
-galactosidase assays was
pRS314-17
80lacZ (10).
Plasmids were transformed into yeast cells (strain YNN282
(MAT trp1-
his3-
200 ura3-52
lys-801a ade2-10) (11), strain
FY24 (MAT
ura 3-52 trp1
63 leu 2
1) (17), or the
gal80- strain KY243 (MAT
gal80
his4-917
lys2-173R2 leu2
1 ura3-52 trp1
63); generous gift of Karen
Arndt) by a modification (18) of the method of Ito et al.
(19). Cells were grown in dropout media (Bio 101) in the presence of
2% glucose, 1.5% raffinose, or 2% galactose.
-Estradiol (Sigma)
was added from a 5 mM stock in ethanol to 0.1 µM 3-4 h before harvesting cells or assaying for
-galactosidase activity.
-Galactosidase activity was assayed as
described (20, 21), and units were expressed as 1000 × A420/(A600 × time (min) × volume (ml)).
Yeast cells (0.4-1.0 liter)
were grown at 30 °C to an optical density at 600 nm between 0.6 and
1.6. Yeast nuclei (22) or spheroplast lysates (23) were prepared and
digested as described previously with micrococcal nuclease for 10 min
at 37 °C at concentrations varying from 0 to 50 units/ml. Cleavage
patterns did not vary within this range of concentrations. For
SstI digestion, the same buffer was used with the addition
of volume of 10 × enzyme buffer supplied by the
manufacturer. Control samples were incubated for 30 min without enzyme;
these did not show the specific SstI cleavage seen when
enzyme was added. TALS chromatin prepared under identical conditions
from FY24 and YNN282 strains showed identical MNase cleavage patterns
and topoisomer distributions. For digestion of naked DNA, samples were
first treated with proteinase K, extracted with phenol and chloroform, and then taken up in 300 µl of 150 mM NaCl, 5 mM KCl, 1 mM EDTA, 20 mM
Tris·HCl, pH 8.0, 2 mM CaCl2, 5 mM MgCl2, or in 10 mM HEPES, pH
7.5, 2 mM CaCl2, 5 mM
MgCl2 and digested as above. Digestions were stopped by the
addition of 55 µl of 5% SDS, 5 mg/ml proteinase K to 300-µl
reactions, and after >2 h at 37 °C, samples were extracted with
phenol and chloroform and the DNA was precipitated. Pellets were taken
up in 100 µl of 10 mM Tris·HCl, pH 8.0, 1 mM EDTA, and treated with RNase A (40 µg); samples from
spheroplast lysates were cleaned over 1-ml G50 spin columns (24). For
analysis of micrococcal nuclease cleavage patterns by indirect end
labeling (25, 26), aliquots of 25-40 µl were digested with
EcoRV for 3-4 h, and the DNA was precipitated and then
electrophoresed on 1.2% agarose gels at 4 V/cm for 5-6 h.
SstI-cut samples were analyzed similarly but on smaller gels
run for shorter times. Bromphenol blue was omitted from sample buffers.
DNA was transferred to nylon membranes (Duralon UV; Stratagene) by
capillary action and hybridized with random primer-labeled probes as
described (24). Hybridization and washing were as described (27).
Probes used were the 200-bp EcoRV-XbaI fragment
(see Figs. 2, 3, 5, and 6) and the 231-bp EcoRV-HindIII fragment (Fig. 1) from TRP1ARS1 or
identical fragments prepared by polymerase chain reaction. Marker lanes
contained 0.5 ng of a
X/HaeIII digest, and labeled
X/HaeIII DNA was included with the hybridization probe.
Chromatin was prepared from at least two independent clones for all
MNase analyses in at least three independent preparations (total).
SstI digestion experiments were done with at least three
independent chromatin preparations for all conditions examined.
Nucleosome perturbation in the yeast episome
TA1780 by ligand-activated and nonactivated GAL4 ER·VP16.
Upper panel, nuclei were prepared from yeast cells (strain
FY24) harboring TA17
80 and either pRS425GAL4 (left part,
lanes 1-4) or pRS416GAL4·ER·VP16 (right
part, lanes 5-9) and digested with micrococcal
nuclease. Cells were grown in glucose (lanes 1-2 and
5-9) or galactose (lanes 3-4) in the absence
(lanes 1-7) or presence (lanes 8-9) of
-estradiol, as indicated. Micrococcal nuclease cleavage sites were
mapped clockwise relative to the EcoRV site as indicated.
Samples were digested with increasing amounts of micrococcal nuclease
from the outer lanes of each panel toward the
centers. Bands indicated by stars are
cleaved in naked DNA (data not shown and Ref. 10) and protected by
nucleosomes I and II (lanes 1 and 2); the
filled stars correspond to bands that are cleaved more strongly than the empty
star. Cleavage at these sites indicates that nucleosome positioning has been perturbed. The cleavage site corresponding to the
distal border of nucleosome II is marked with a filled circle. Locations of nucleosomes I and II in cells grown in
glucose are indicated; the box attached to nucleosome I
represents the GAL4 binding site. Lane 8 was taken from a
different exposure than lanes 5-7 and 9.
Lower panel, densitometric scans of the salient regions of
lanes 1 and 4 (upper traces, labeled
glucose and galactose, respectively) and
lanes 7 and 8 (lower traces, labeled
E2 and +E2, respectively). The filled
circle and starred peaks correspond to those in the
upper panel, and the regions corresponding to nucleosomes I
and II are indicated; the black box attached to nucleosome I
represents the GAL4 binding site.
Topoisomer Analysis
DNA was prepared from at least three independent clones for topoisomer analysis as described (28) from 10-ml cultures grown to an A600 of 0.6-1.2 by rapid glass bead lysis. Purified DNA was electrophoresed on 1.5% agarose gels containing 40 µg/ml chloroquine diphosphate (Sigma) at 2 V/cm for 18-20 h, transferred to nylon membranes, and hybridized using probes specific for TRP1ARS1 sequences as described above. Gaussian centers of the topoisomer distributions were calculated as described previously (28), using images scanned on a Molecular Dynamics PhosphorImager.
Previously we showed that GAL4 can bind to a nucleosomal recognition site in yeast (10). Although this result shows that GAL4 can outcompete histones for occupancy of its site in vivo, it was not determined whether binding could take place to a stable, fully assembled nucleosome or whether it occurred during replication, when nucleosome structure is transiently perturbed. To attempt to address this question, we obtained an expression vector for the chimeric protein GAL4·ER·VP16, which consists of the GAL4 DNA-binding domain, the human estrogen receptor hormone-binding domain, and the viral VP16 activation domain (Ref. 14; generously provided by Dr. D. Picard). This protein, when expressed in yeast, activates transcription from promoters having a GAL4 site or sites in a strictly hormone-dependent fashion (Ref. 14; Table I). We hoped that the GAL4·ER·VP16 protein might also bind DNA in a hormone-dependent manner, so that we could examine its binding to sites in chromatin at various points in the cell cycle.
|
To determine whether GAL4·ER·VP16 binds to chromatin in
vivo in the absence of hormone, we compared transcriptional
activation of a UASGAL-lacZ reporter by GAL4 in
the presence and absence of the GAL4·ER·VP16 expression vector.
-Galactosidase activity induced by GAL4 (via growth of cells in
galactose) was decreased in the presence of GAL4·ER·VP16 (Table I),
suggesting that GAL4·ER·VP16 could compete with GAL4 for site
occupancy in the absence of hormone. To determine whether
GAL4·ER·VP16 could bind to a nucleosomal site in the absence of
hormone, we examined its effect on the chromatin structure of TA17
80
(10). TA17
80 is a TRP1ARS1-based plasmid in which a single strong
GAL4 binding site has been introduced such that the site is
incorporated near the center (i.e. near the pseudodyad) of a
positioned nucleosome in cells grown in glucose, conditions under which
GAL4 is expressed at extremely low levels (29). Yeast cells harboring
TA17
80 were grown in the presence of GAL4·ER·VP16 with and
without hormone and in the absence of GAL4·ER·VP16 in glucose (GAL4
nearly absent) and in galactose (GAL4 present in its activating
configuration). Nuclei were prepared and treated with micrococcal
nuclease (MNase), and the cleavage sites were mapped relative to an
EcoRV restriction site (25, 26). Regions of about 150 base
pairs that are protected from MNase digestion in chromatin but are
cleaved as naked DNA are diagnostic of positioned nucleosomes (30, 31);
in the case of TA17
80 and its progenitor TRP1ARS1, the assignment of
positioned nucleosomes I and II (Fig. 1) is also based
on a substantial body of earlier work (10, 30, 31).
In cells grown in glucose without the GAL4·ER·VP16 expression
vector, GAL4·ER·VP16 is absent and GAL4 is expressed at very low
levels, and positioned nucleosomes are present as expected (Fig. 1,
lanes 1 and 2). In galactose medium, GAL4
expression is induced, and MNase cleavage sites are generated in the
regions of both nucleosomes I and II, as observed previously (Fig. 1, lanes 3-4; Ref. 10). These new cleavages are also seen in
naked DNA treated with MNase (10) and indicate that GAL4 is binding to
its site and perturbing nucleosome positioning. The same cleavage sites
are seen, albeit more weakly, when GAL4·ER·VP16 is expressed in
cells grown in glucose medium and hormone is not added (Fig. 1,
lanes 5-7), indicating that GAL4·ER·VP16 can bind under
nonactivating conditions. Interestingly, the addition of 0.1 µM -estradiol caused enhanced cleavage in the regions
of nucleosomes I and II in TA17
80 (compare the relative intensities
of cleavage sites marked by stars with the site marked by a
filled circle, lanes 5-9), suggesting that
binding of GAL4·ER·VP16 was increased or altered by the addition of
hormone.
These results are consistent with previous findings that binding of
GAL4 to TA1780 under activating conditions (in galactose) allowed
strong MNase cleavage in the regions of nucleosomes I and II, whereas
ectopic expression of GAL4 in glucose, in which it is repressed by
GAL80, or expression of the very weakly activating derivative,
GAL4(1-147)H, allowed only weak MNase cleavage (10). We wondered
whether a GAL4 binding site embedded in chromatin in a different
context would show a similar effect of an unmasked activation domain,
so we examined another yeast episome having a GAL4 binding site in a
positioned nucleosome, TALS. The TALS episome is packaged into strongly
positioned nucleosomes in yeast
cells by the
2·MCM1 complex in
conjunction with SSN6 and TUP1 (Refs. 11 and 32; see Fig.
2). Nucleosome IV of TALS, immediately adjacent to the
2·MCM1 operator, contains a strong binding site for GAL4, which
derives from the GAL3 promoter (12) and is in a region of
nucleosome IV inaccessible to Escherichia coli Dam methyltransferase expressed in yeast (33). As in TA17
80, the GAL4
binding site in this plasmid is removed from its natural context
(because the 3
portion of the GAL3 promoter, including the
TATA box, is not present in TALS, and sequences farther than 60 base
pairs upstream of the UASGAL do not contribute to
GAL3 expression (12)). Nearly the entire TALS plasmid is
packaged into strongly positioned nucleosomes in yeast
cells, so we
thought it would provide a potentially interesting template to examine the effects of GAL4 binding and activation domain unmasking on chromatin structure.
Nuclei from yeast cells harboring TALS were prepared and digested with
MNase, and cleavage sites were mapped counterclockwise from the
EcoRV site (Fig. 2). Comparison of the digestion patterns of
chromatin from cells lacking the GAL4·ER·VP16 expression vector and
grown in glucose (lane 9) with naked DNA (lane
11) shows that TALS is packaged into strongly positioned
nucleosomes, as observed previously (11). (Although the sample in
lane 9 is heavily digested, the cleavage sites are identical
to those seen in lighter digests (Ref. 11 and data not shown).) When
GAL4·ER·VP16 is expressed, a new cleavage site is generated in the
region of nucleosome IV, close to the GAL4 binding site, suggesting
that GAL4·ER·VP16 can bind and perturb nucleosome positioning in
TALS even in its nonactivating form (Fig. 2, lanes 2-4; see
the band marked by an arrowhead, which is absent
from lane 9). A parallel culture incubated for 4 hours
in the presence of 0.1 µM -estradiol showed further
changes in chromatin structure (Fig. 2, lanes 5-7); the
cleavage indicated by the arrowhead is enhanced relative to
the site below it, and a new site is generated in the vicinity of
nucleosome III (filled circle). A weak cleavage site is also
generated near the center of the region protected by nucleosome IV,
which is also cut only weakly in naked DNA (lane 11).
GAL4 in galactose perturbs the MNase cleavage pattern of TALS chromatin similarly to GAL4·ER·VP16 in the presence of hormone. GAL4 can also bind to DNA in a nonactivating configuration in the absence of galactose, when its activation domain is masked by the repressor protein GAL80 (34, 35). Binding of nonactivating GAL4 in vivo has been detected by dimethyl sulfate footprinting (36), photofootprinting (37), and perturbation of positioned nucleosomes (10). When galactose is present, a conformational change unmasks the GAL4 activation domain (38). To determine whether GAL4, like GAL4·ER· VP16, would perturb TALS chromatin differently in its activating and nonactivating configurations, we performed MNase digests of TALS chromatin from cells grown in glucose (GAL4 nearly absent), raffinose (GAL4 present in its nonactivating configuration), and galactose (GAL4 present in its activating configuration). We also analyzed chromatin from cells in which GAL4 was ectopically expressed from the strong ADH1 promoter in glucose. The high levels of GAL4 expressed from the ADH1 promoter titrate GAL80 (35), but most (>75%; data not shown) of the GAL4 molecules are still repressed by GAL80 under these conditions.
Fig. 3 shows changes in the MNase cleavage pattern of TALS in galactose compared with glucose (lanes 2 and 3; see bands marked by dots) which are similar to those seen in the presence of ligand-activated GAL4·ER·VP16 compared with in its absence (Fig. 2, compare lanes 5-7 with lane 9; see bands marked by arrowhead and dot). When GAL4 is expressed in its nonactivating configuration by growth of cells in raffinose (Fig. 3, lane 1) or under conditions where activation is weak by ectopic expression in glucose (lanes 6 and 7), the MNase cleavage pattern appears intermediate between that seen in glucose and galactose. Thus, similar to what was observed for GAL4·ER· VP16, binding of GAL4 and unmasking of its activation domain have distinct effects on the MNase cleavage pattern of TALS chromatin.
Topological Perturbation of TALS by Activating and Nonactivating GAL4 and GAL4·ER·VP16Close comparison of the MNase cleavage
patterns of naked TALS DNA and of TALS chromatin in the presence of
hormone-activated GAL4·ER· VP16 (Fig. 2, lanes 5-7 versus
lane 11) suggests that nucleosome IV may be lost entirely, since
the cleavage pattern in this region is very similar to that of naked
DNA. A new MNase cleavage site is also generated in the vicinity of
nucleosome III upon hormone administration (Fig. 2, lanes 5-7
versus lanes 2-4, filled circle), but this site is
close to the edge of nucleosome III, and its cleavage may reflect only
a subtle change in nucleosome positioning. To examine the extent of
nucleosome loss from TALS caused by binding of activated
GAL4·ER·VP16, we measured plasmid topology of TALS isolated from
cells harboring GAL4·ER·VP16 in the presence and absence of
-estradiol. Since each nucleosome introduces one negative supercoil
into plasmid DNA (39, 40), any loss of nucleosomes accompanying binding
of GAL4·ER·VP16 in either its activating or nonactivating forms
should change plasmid topology.
DNA was rapidly isolated from yeast cells harboring TALS under
conditions that rapidly inactivate topoisomerase, so plasmid topology
reflects that present in vivo (28, 41, 42). Samples were
electrophoresed on chloroquine-containing gels to resolve individual
topoisomers, blotted, and hybridized with a probe specific for TALS.
Under the conditions used (40 µg/ml chloroquine diphosphate), TALS
topoisomers migrated as positively supercoiled molecules, so that
faster migrating bands represent more positively supercoiled species.
Fig. 4 shows that in the absence of GAL4·ER·VP16,
administration of 0.1 µM -estradiol had no effect on
TALS topology (lanes 1 and 2). Expression of
GAL4·ER·VP16 in the absence of hormone had no effect on TALS
topology (lane 3), despite its effect on the MNase cleavage
pattern (Fig. 2, lanes 2-4 compared with lane
9). Hormone administration caused a loss of nearly one negative
supercoil (Fig. 4, lane 4; Table II),
suggesting loss of one nucleosome in a majority of the TALS
minichromosomes. Similarly, growth in raffinose, in which GAL4 is
expressed but does not activate transcription due to its repression by
GAL80, resulted in minimal topological perturbation (compare lane
5 with lane 1), whereas growth in galactose, where GAL4
is present in its activating form, caused a loss of nearly one negative
supercoil (lane 6). When the same experiment was done in a
gal80
strain, loss of negative supercoiling
was seen in cells grown in raffinose compared with cells grown in
glucose, and no change was seen between cells grown in raffinose and
galactose (data not shown). GAL4 expressed ectopically in glucose
medium, in which it is repressed by GAL80, had only a minimal effect on
TALS topology, as did the very weakly activating derivative
GAL4(1-147)H, which lacks the GAL4 activation domain (data not shown).
A derivative of TALS lacking the GAL4 binding site, TALS4, showed no
topological shift between glucose and galactose (Fig. 4, lanes
9 and 10) and was not affected by hormone addition in
the presence of GAL4·ER·VP16 (data not shown). Results from several
such experiments are tabulated in Table II, providing quantitative
support for the visual evidence of Fig. 4.
|
If nucleosome IV of
TALS is lost from a majority of minichromosomes upon binding of
GAL4·ER·VP16 in the presence of -estradiol, but only from a
minority in the absence of hormone, we would expect restriction
endonuclease sites in this region to be only moderately more accessible
in the presence of GAL4·ER·VP16 without hormone than in its absence
and considerably more accessible following hormone addition. We have
used an SstI site that is close to the center of the region
protected by nucleosome IV to test this hypothesis. Yeast nuclei from
cells harboring TALS were treated with SstI, and aliquots
were stopped after 15- and 30-min incubation with enzyme. Cleavage of
the nucleosome IV SstI site was increased only slightly
after 30 min compared with after 15 min (Fig. 5), and a
control plasmid with an SstI site added at 15 min was nearly completely cleaved by 30 min, demonstrating that SstI
activity was not lost during the incubation (Fig. 5, lane 13 and data
not shown). Together these results suggest that almost all of the accessible SstI sites in the TALS minichromosomes were
cleaved by the end of the 30-min reaction.
TALS minichromosomes from cells grown in glucose and lacking
GAL4·ER·VP16 were cleaved very little by SstI (Fig. 5,
lanes 10 and 11). Cleavage was increased little
if at all by expression of GAL4·ER·VP16 (lanes 1 and
2) but was increased substantially and reproducibly by
incubation with -estradiol (lanes 4 and 5). Similarly, expression of GAL4 under nonactivating conditions by growth
of cells in raffinose caused a very modest increase in SstI
accessibility (Fig. 5, lanes 12 and 13), whereas
expression of the activating form of GAL4 by growth in galactose
consistently increased accessibility (Fig. 5, lanes 14 and
15). Thus, restriction enzyme accessibility is consistent
with increased disruption of nucleosome IV accompanying unmasking of
the activation domains of GAL4 or GAL4·ER·VP16.
The data presented so far suggest that the activation domains of GAL4 and GAL4·ER·VP16 affect their interaction with TALS chromatin. However, it has been reported that DNA binding of intact ER and derivatives is affected by hormone addition (13, 43, 44). To separate effects exerted by the hormone-binding domain of GAL4·ER·VP16 on the DNA-binding domain from those due to unmasking of the activation domain, we decided to examine the effect of eliminating the activation potential of GAL4·ER·VP16. We therefore excised the VP16 coding region from our expression vector to create an expression vector for GAL4·ER. GAL4·ER is unable to activate transcription from the simple UASGAL-lacZ reporter used in our assay (Table III), in agreement with Louvion et al. (14). It does, however, measurably inhibit transcription of the UASGAL-lacZ reporter by GAL4 (Table III), suggesting that it is expressed at levels similar to that of GAL4·ER·VP16. GAL4·ER slightly affects TALS chromatin in the absence of hormone, as assayed by micrococcal nuclease digestion (Fig. 6, lanes 9 and 10; compare with Fig. 2, lane 9, and Fig. 3, lane 3), similar to what was observed for GAL4·ER·VP16 (Fig. 2). However, in contrast to GAL4·ER· VP16, hormone addition in the presence of GAL4·ER does not further perturb the MNase cleavage pattern (Fig. 6, lanes 9-12 versus lanes 3-6). Consistent with this lack of effect, hormone addition does not alter supercoiling (Fig. 4, lanes 7-8) or appreciably increase restriction enzyme accessibility (Fig. 5, lanes 6-9). We conclude that it is not a hormone-induced conformational change in GAL4·ER·VP16 per se but unmasking of the VP16 activation domain upon ligand binding by the ER hormone binding domain that results in chromatin perturbation.
|
The
topological analysis presented above entails rapid isolation of DNA
from small (10-ml) volumes of yeast cells. We took advantage of this
protocol to examine the kinetics of TALS chromatin remodeling by
GAL4·ER·VP16 following hormone administration. Yeast cells
harboring TALS and the expression vector for GAL4·ER·VP16 were
grown, and DNA was isolated from 10- or 20-ml aliquots immediately before (t = 0) and at intervals following the addition
of -estradiol to 0.1 µM. Topological distributions of
TALS from one representative experiment are shown in Fig.
7A, and results from three independent experiments are combined graphically in Fig. 7B. A
substantial change in TALS topology is seen 15 min after hormone
addition, and the topological shift is essentially complete after 30 min. Since the cell doubling time under the experimental conditions used was more than 2 h, we conclude that replication is not
required for chromatin remodeling by GAL4·ER·VP16 in yeast.
In this paper we have used the yeast episomes TA1780
and TALS as reporters to monitor changes in chromatin structure caused by binding of GAL4 and GAL4·ER· VP16 in their activating and
nonactivating configurations. The chromatin remodeling we have observed
in both of these plasmids occurs outside the context of natural
promoters. TALS (11) contains a single GAL4 binding site from the
GAL3 promoter (UASg1 in Ref. 2) but lacks
downstream promoter elements, including the TATA element and
transcription start site. Furthermore, chromatin structure in TALS in
yeast
cells is dominated by the
2 operator (11), and in the
region perturbed by GAL4 and GAL4·ER·VP16, which corresponds to the
region upstream of the GAL4 binding site in the GAL3
promoter, it is very different in TALS than in the endogenous
GAL3 promoter whether GAL4 is present or
absent.2 Thus, the results reported here
suggest that the chromatin remodeling observed in natural promoters (2,
5-8) reflects a general property of at least some transcriptional
activators that does not depend on collaboration of particular
cis-acting promoter elements. The precise nature of the remodeling,
however, depends on the chromatin structure that is perturbed. For
example, substantial perturbation by GAL4 is seen with TALS and
TA17
80, both of which have nucleosomal GAL4 binding sites, but only
modest perturbation is seen in the GAL1 promoter, in which
the GAL4 binding sites are nonnucleosomal (5, 6).
The GAL4 binding site in TALS is located 33-49 base pairs from the
edge of nucleosome IV, which is adjacent to the 2·MCM1 nucleosome-positioning element (22). This region is inaccessible to
E. coli Dam methyltransferase expressed in yeast (33).
Nevertheless, the GAL4 DNA-binding domain can evidently gain access to
this region, although whether it does so in the presence of a fully formed nucleosome or at a point in the cell cycle when nucleosome structure is perturbed, such as during DNA replication, has not been
determined. Similarly, a GAL4 binding site near the center of
nucleosome I of the TRP1ARS1 derivative TA17
80 can be accessed by
GAL4 (Ref. 10; Fig. 1), although this region is also inaccessible to
E. coli Dam methyltransferase (33).
Some perturbation of the chromatin structure of TA1780 (Fig. 1) and
TALS (Fig. 2) is caused by both GAL4 and GAL4·ER·VP16 in their
nonactivating configurations, as evident by a slightly altered MNase
digestion pattern. Binding by nonactivating GAL4·ER·VP16 and
GAL4·ER is also demonstrated by interference with GAL4 activation (Tables I and III). This may represent transient or unstable binding under these conditions. It seems reasonable that some interaction must
occur between the factor and its binding site even without an
activation domain, since otherwise it is difficult to understand how
the factor in its activating configuration makes its initial interaction (i.e. it seems unlikely that a single factor
such as GAL4 would cause a global alteration in chromatin structure that allows the factor to interact with its binding sites in
chromatin). Unmasking of the GAL4 activation domain by growth in
galactose results in altered micrococcal nuclease cleavage of TALS
chromatin, increased restriction enzyme accessibility to the
SstI site in nucleosome IV, and a loss of negative
superhelicity. The addition of hormone leads to similar changes
mediated by GAL4·ER·VP16. Similarly, chromatin structure in
TA17
80 shows increased perturbation by GAL4 and GAL4·ER·VP16 in
their activating compared with their nonactivating forms (Fig. 1; Ref.
10). The effect of hormone addition on TALS chromatin is not seen with
GAL4·ER, showing that this effect is not a consequence of unmasking
an occluded DNA-binding domain.
Two pathways by which GAL4 or GAL4·ER·VP16 could alter TALS
chromatin structure in an activation domain-dependent
manner are by increasing factor binding and by recruiting a chromatin
remodeling activity. Increased or more stable binding could result in
increased loss/perturbation of nucleosome IV in TALS and nucleosomes I
and II in TA1780, because the GAL4 binding site is nucleosomal. This mechanism is suggested by experiments showing that occupancy of low and
moderate affinity GAL4 binding sites is increased in the presence of a
nearby TATA box in yeast, suggesting direct or indirect cooperative
interactions between the GAL4 activation domain and TBP (45), and by
in vivo footprinting data showing that an Oct-2 POU
DNA-binding domain occupies binding sites in yeast more efficiently when it carries an activation domain (46). Similarly, GAL4 binding in
mammalian cells is enhanced by the presence of an activation domain
(47). This increased binding could be caused by an interaction between
the activation domain and a protein or proteins involved in
transcriptional activation, such as TBP, TFIIB, or the RNA polymerase
II holoenzyme (48-50). The interacting protein could contact DNA
sequences near the GAL4 binding site and increase the effective
affinity of GAL4 through cooperative interactions. Both the
interactions between the contacted protein(s) and DNA and the increased
effective affinity of the activator for its site could inhibit histones
from occupying these sequences and thereby alter chromatin structure.
One possible difficulty with this mechanism is that chromatin
remodeling in TALS occurs outside of the context of a natural promoter,
and therefore, presumably, high affinity binding sites for components
involved in transcription are lacking. A near consensus TATA box,
TATATA, is present 330 base pairs from the
UASGAL, distal to the
2·MCM1 operator; however, activation domain-dependent changes in TALS chromatin
structure are still observed when this sequence is mutated (data not
shown). It is conceivable that nonspecific interactions between
proteins recruited by the activation domain and sequences near the GAL4 binding site are sufficiently avid to increase GAL4 or GAL4·ER·VP16 binding. Since TBP has a high affinity for nonspecific sites in DNA
(51) and interacts with the GAL4 (48, 50) and VP16 (49) activation
domains in vitro, it remains a good candidate for such interactions.
This first mechanism does not involve any interactions specifically directed toward chromatin; rather, protein-DNA interactions are postulated to be sufficient to alter the ability of histones to occupy particular sequences. However, strong evidence exists for proteins that have evolved to contend with chromatin templates (52), and it is possible that an unmasked activation domain acts to recruit such proteins. For example, the GAL4 DNA binding domain has been shown to interact with ADA2 in vitro (48), which is part of a complex in yeast containing GCN5 (53). Since yeast GCN5 is homologous to Tetrahymena histone acetyltransferase A and itself possesses histone acetyltransferase activity (54), recruitment of the ADA·GCN5 complex could lead to alterations in chromatin structure. Another likely candidate for this role is the SWI·SNF complex (52), which in one in vitro study has been shown to cooperate with GAL4 derivatives to alter nucleosome structure and enhance GAL4 binding in a manner dependent on the strength of the activation domain attached to the GAL4 DNA-binding domain (55). It is of course also possible that both pathways contribute to the activation domain-dependent changes in chromatin structure we have reported here. For example, an activation domain may help basal transcription factors compete with histones for occupancy of sequences near the activator binding site (1), and this effect may also depend on auxiliary factors such as the SWI·SNF complex (56).
Whatever the mechanism by which the GAL4·ER·VP16 activation domain remodels chromatin, it is apparently able to do so in the absence of replication (Fig. 7). Chromatin remodeling by PHO4 (57), by the glucocorticoid receptor (58), and by the thyroid hormone receptor-retinoid X receptor heterodimer (8) also occurs independent of replication. The ability to remodel chromatin independent of replication may prove to be a general property of transcriptional activators.
Implications for the Mechanism of Hormone Activation of Estrogen ReceptorOur results suggest that GAL4·ER·VP16 can bind to a UASGAL in the absence of hormone, albeit weakly, but is still unable to activate transcription (Table I; Figs. 1 and 2). This suggests that the ER hormone-binding domain can mask activator function even when bound at a promoter, as also proposed for intact ER expressed in yeast (59) and mammalian cells (60). Remarkably, this occurs with a heterologous activation domain. Hormone administration induces substantial changes in chromatin structure, consistent with previous observations of hormone-induced alterations in chromatin structure mediated by ER derivatives expressed in yeast (44, 61). We suggest that upon hormone addition, unmasking the VP16 activation domain results in increased binding to chromatin or in recruitment of chromatin remodeling activity. An alternative explanation is that the unliganded hormone-binding domain of the ER partially masks the attached DNA-binding domain, as has been suggested for the intact ER (13, 44), and that hormone binding relieves this masking. However, if this were the case, the GAL4·ER chimera should have shown increased perturbation of chromatin upon hormone addition, and this was not observed.
Implications for Other SystemsAs discussed in the introduction, activation domain-dependent perturbation of chromatin structure has been observed at the GAL1 and PHO5 promoters (2, 7). In both cases, perturbation is seen even when the TATA box is mutated. This could be explained by recruitment of factors that can modify chromatin components, as discussed above. It could also be explained by nonspecific interactions between TBP and DNA or by interactions between proteins recruited by the activation domain and sequences outside the TATA box. These latter interactions may be strong; Fedor and Kornberg (5) reported that coding sequences linked to the GAL1-10 promoter were transcribed under inducing conditions even in the absence of the GAL1 TATA and initiation sequences.
It has also been reported that PHO4 can bind to the nucleosomal site in the PHO5 promoter in the absence of the binding site located in the nearby linker region between nucleosomes (7, 62). This binding requires overexpression of PHO4 and requires PHO4 to carry an activation domain; it also results in disruption of all four positioned nucleosomes in the PHO5 promoter. This would again be consistent with an increased affinity of PHO4 for its binding site caused by recruitment of another DNA-binding protein by the PHO4 activation domain or by recruitment of chromatin-modifying machinery.
One important difference between the two mechanisms that we have suggested for activation domain-dependent chromatin remodeling is that one is passive in the sense that histones "get out of the way" as a consequence of binding of other factors (e.g. TBP or TFIIB), without invoking any special chromatin remodeling activity, whereas the second explicitly invokes such an activity. Determining what proteins are required for the activation domain-dependent changes in chromatin structure we have reported here should shed light on this problem and should also provide new insight into mechanisms of transcriptional activation in vivo.
We thank Robert Simpson (National Institutes of Health) for generous support and encouragement, Michael Kladde for valuable discussions, Sharon Roth for a critical reading of the manuscript, and Rachael Jones and Peter Bocala for excellent technical help. Didier Picard, Michael Kladde, and Karen Arndt are thanked for providing strains and/or plasmids. We gratefully acknowledge the use of the Wadsworth Center's molecular genetics core facility.