(Received for publication, December 11, 1996, and in revised form, February 19, 1997)
From the Department of Biochemistry and Molecular Biology and The Center for Gene Regulation, The Pennsylvania State University, University Park, Pennsylvania 16802-4500
To investigate the potential mechanisms by which
the SWI/SNF complex differentially regulates different genes we have
tested whether transcription factors with diverse DNA binding domains were able to exploit nucleosome disruption by SWI/SNF. In addition to
GAL4-VP16, the SWI/SNF complex stimulated nucleosome binding by the
Zn2+ fingers of Sp1, the basic helix-loop-helix
domain of USF, and the rel domain of NF-B. In each case SWI/SNF
action resulted in the formation of a stable factor-nucleosome complex
that persisted after detachment of SWI/SNF from the nucleosome. Thus,
stimulation of factor binding by SWI/SNF appears to be universal. The
degree of SWI/SNF stimulation of nucleosome binding by a factor appears to be inversely related to the extent that binding is inhibited by the
histone octamer. Cooperative binding of 5 GAL4-VP16 dimers to a 5-site
nucleosome enhanced GAL4 binding relative to a single-site nucleosome,
but this also reduced the degree of stimulation by SWI/SNF. The SWI/SNF
complex increased the affinity of 5 GAL4-VP16 dimers for nucleosomes
equal to that of DNA but no further. Similarly, multimerized NF-
B
sites enhanced nucleosome binding by NF-
B and reduced the
stimulatory effect of SWI/SNF. Thus, cooperative binding of factors to
nucleosomes is partially redundant with the function of the SWI/SNF
complex.
In eukaryotic cells the DNA is packaged into nucleosomes, the primary order of chromatin structure that provides an inherent barrier for activator proteins and the basal transcription machinery. There are several mechanisms that help transcription factors contend with nucleosomal DNA (reviewed in Refs. 1 and 2). One such mechanism involves energy-dependent remodeling of chromatin, which allows transcription factors to access otherwise inaccessible promoters. Such activities have been purified from yeast (3-5), Drosophila (6), and humans (7-9). The first of these activities to be identified was the yeast SWI/SNF complex.
SWI/SNF was originally identified in yeast as a set of positive
regulators of the HO gene (mating type switch, SWI) and the SUC2 gene (sucrose non-fermenting, SNF) (reviewed in Ref.
10). The connection to chromatin was revealed when mutant suppressors were identified in histones as well as in other chromatin proteins (11,
12). These SWI/SNF proteins were biochemically isolated as an
11-subunit complex of approximately 2 megadaltons. Members of this
complex include SWI1, SWI2-SNF2, SWI3,
SNF5, SNF6 (3, 4), SWP73 (13),
SNF11 (14), and TFG3 (TAF30) (15),
although the latter two do not demonstrate a swi
phenotype. The SWI2-SNF2 subunit contains a conserved ATPase domain and has been shown to function as a DNA-dependent
ATPase. This domain is required for SWI/SNF function in vivo
(16). Interestingly, SWI/SNF homologs have also been identified in
mammals and have a conserved function. In contrast to yeast, however,
multiple forms of the complexes can be found within the same cell (17). Two SWI2-SNF2 homologs (brm and BRG1) have been identified in mouse and
human systems. Both are found in large (2-megadalton) complexes, but
each is present in separate complexes (8). In vitro binding
experiments showed that ySWI/SNF can enhance the binding of GAL4
derivatives to nucleosomes in an ATP-dependent reaction
(18). This enhancement required only the GAL4 DNA binding domain and
was accompanied by an ATP-dependent disruption of
histone-DNA contacts. The human SWI/SNF complex was also shown to
enhance the binding of GAL4 derivatives (7, 8) as well as TBP to nucleosomes (19).
The SWI/SNF complex is able to disrupt nucleosome structure in a sequence-independent manner (20, 21) suggesting that it may be able to function at diverse promoters. Consistent with this possibility is the fact that transcription of a set of diversely regulated genes was found to require SWI/SNF. These genes include HO, SUC2, ADH1-2, INO1, Ty elements, and to some degree the GAL1-10 genes (10, 22). In addition, certain yeast activators as well as the function of several heterologous activators were shown to require the SWI/SNF complex in yeast (22-24). However, transcription of several other yeast genes does not depend on the SWI/SNF complex in vivo. For example, transcription of the PHO5 gene is SWI/SNF-independent (13). Moreover, transcription from several yeast promoters or from artificial promoters driven by heterologous activators in yeast is differentially affected by mutations in the SWI/SNF subunit Swp73 (13).
The differential requirement for SWI/SNF function in different promoter contexts might be explained in part by at least two possibilities. First, SWI/SNF function may be available to only a subset of transcriptional activators. In other words, the disruption of DNA-histone interactions by the SWI/SNF complex may be effectively exploited by activators with particular DNA binding domains but insufficient to enhance the affinity of others. Second, SWI/SNF function may be redundant with alternative mechanisms mediating access of transcription factors to binding sequences in chromatin. To test these possibilities we have analyzed the ability of the SWI/SNF complex to stimulate nucleosome binding by transcription factors with diverse DNA binding domains and whether stimulation by SWI/SNF was affected by cooperative binding of activators to nucleosomes.
To generate plasmid pUGB-NFX1, a
20-bp1 oligonucleotide
(5-CGTAGGGGACGTCCCCGTAT-3
) harboring a high affinity NF-
B site was inserted into a unique BstBI site between the USF and GAL4
sites of pUSFGALBend (25). Plasmid pGUB-NFX3 was similarly generated by
inserting three copies of the same oligonucleotide into the BstBI site in pGALUSFBend (25). Plasmid pG5SUE4T
was created by the insertion of a 37-bp oligonucleotide
(5
-ACGGGGCGGGGCGGTTACCTTCGAACCACGTGGCCGT-3
) containing an Sp1 site, a
USF site, and XbaI cohesive ends into pG5E4T
(26) at the unique XbaI site.
All of the probe DNA fragments were labeled by Klenow reaction after
the first enzyme digest and then cleaved with the second enzyme. The
149-bp USF site probe with a USF site centered 31 bp from the end of
the fragment was generated by cutting pGALUSFBend (25) with
SpeI followed by MluI. pUGB-NFX1 was digested
with XmaI and BamHI to create a 156-bp fragment
with an NF-B site 30 bp from the end and a consensus GAL4 site 58 bp
from the end. A 167-bp probe harboring an Sp1 site 35 bp from the end
was cleaved from pG5SUE4T with enzymes HindIII
and BamHI. The 5 GAL4 site probe was obtained from
pG540HSP70CAT (27) by cutting with NheI followed by
PstI giving rise to a 156-bp fragment with the first GAL4
site centered at 17 bp from the end. Plasmid pGUB-NFX3 was digested
with XmaI followed by BstEII to generate a 166-bp
molecule containing 3 NF-
B sites centered at 13, 33, and 53 bp from
the end. All probe DNA fragments were isolated from 8% polyacrylamide (1 × Tris borate/EDTA) gels.
Recombinant proteins USF and NF-B
were overexpressed in Escherichia coli using the pET system
(Novagen) and purified as described by Pognonec et al. (28).
The form of NF-
B used for this study was a 42-kDa truncated form of
the p50 derivative (29). The GAL4-VP16 fusion protein was expressed in
E. coli and purified according to the procedure of Chasman
et al. (30). The purified Sp1 used in this study was
purchased from Promega. The SWI/SNF complex was purified from yeast as
described in Côté et al. (18) except that the
elution from Ni2+ agarose was done with 300 mM
imidazole. HeLa oligonucleosomes were purified as described previously
(20, 31).
Nucleosomes
were reconstituted by octamer transfer from HeLa nucleosomes as
described previously (20, 31). Gel shift binding reactions included
12.5 ng of total nucleosomes (12.5 nM) in 10 µl with
final conditions of 20 mM HEPES, pH 7.5, 3 mM
MgCl2, 1 µM ZnCl2, 0.1 mg/ml
bovine serum albumin, 5% glycerol, 2 mM dithiothreitol, 0.2 mM phenylmethylsulfonyl fluoride. The concentration of
nucleosomes and/or DNA fragments actually bearing the transcription
factor binding sites was less than 0.1 nM in all instances.
Each transcription factor was serially diluted in its stock buffer as
follows: USF dilution buffer (10 mM HEPES, pH 8.0, 1 mM EDTA, pH 8.0, 3 mM MgCl2, 0.1%
(v/v) Nonidet P-40, 1 mg/ml bovine serum albumin, 20% glycerol, 5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride), NF-B dilution buffer (20 mM Tris-Cl, pH 7.8, 1 mM EDTA, pH 8.0, 100 mM KCl, 0.1% (v/v)
Nonidet P-40, 20% glycerol, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride), GAL4 dilution buffer (20 mM HEPES, pH 7.5, 100 mM KCl, 10 µM ZnCl2, 1 mg/ml bovine serum albumin, 20%
glycerol, 10 mM
-mercaptoethanol), and Sp1 dilution
buffer (12 mM HEPES-KOH, pH 7.7, 50 mM KCl, 6 mM MgCl2, 5 µM ZnSO4,
0.05% Nonidet P-40, 50% glycerol, 1 mM dithiothreitol). The yeast SWI/SNF complex was added at a concentration of 10 nM (0.8:1 molar ratio of SWI/SNF to nucleosomes) for all
gel shift experiments except the experiment illustrated in Fig. 6 where 20 nM SWI/SNF was added. MgATP (1 mM) was
included where indicated in the figure legends. Binding proceeded for
30 min at 30 °C. Interaction of ySWI/SNF was competed with 250 ng of
cold HeLa oligonucleosomes (25-fold molar excess) and 250 ng of calf
thymus DNA and incubated for an additional 30 min at 30 °C.
DNase I binding reactions were performed similarly except 25 ng of nucleosomes were present in a 20-µl final volume (12.5 nM) and 1 mM ATP was included in all lanes. Yeast SWI/SNF was included at 20 nM where indicated for a molar ratio of SWI/SNF to nucleosomes of 1.6. After 30 min of binding at 30 °C, DNase I (Boehringer Mannheim) was added (0.5 units to nucleosomes, 0.05 units to DNA) and incubated for 1 min at room temperature. The reaction was stopped with 1 volume of DNase I stop buffer (20 mM Tris-Cl, pH 7.5, 50 mM EDTA, 2% SDS, 0.2 mg/ml proteinase K, 0.25 mg/ml yeast tRNA) and incubated at 50 °C for 1-3 h. Deproteinized samples were precipitated with ammonium acetate and washed with 80% ethanol. Pellets were resuspended in 2 µl of double distilled H2O and 3 µl of formamide dye (95% formamide, 10 mM EDTA, 0.1% xylene cyanol, 0.1% bromphenol blue), incubated at 90 °C, and placed on ice. Samples were loaded and run at 60 watts of constant power on an 8% polyacrylamide (19:1 acrylamide to bisacrylamide), 8 M urea sequencing gel. Gels were subjected to autoradiography. Results were quantitated on a Molecular Dynamics laser densitometer.
When bound to nucleosomes, the SWI/SNF complex alters the path of DNA bending around the nucleosome core in an ATP-dependent reaction that does not by itself displace the underlying histones (18, 21). While the structural basis of this alteration is unknown, it results in the enhanced binding of GAL4 derivatives (7, 18) and TBP to nucleosome cores (19). In principle, GAL4 and TBP could occupy nucleosomal DNA without major steric clashes with the histone octamer (19, 32). Thus, SWI/SNF-induced alteration of the path of DNA bending around the histone octamer might suffice to stimulate binding of these factors by providing optimum exposure of their binding sites on the nucleosome surface. A priori, it would seem less likely that this mechanism would effectively stimulate the binding of factors with DNA binding domains that surround the DNA helix.
To determine whether different types of DNA binding domains affected a
factor's ability to exploit SWI/SNF disruption of nucleosomes, we
tested 3 different transcription factors for ySWI/SNF stimulation of
nucleosome binding. These included Sp1, which has three
Zn2+ fingers in its DNA binding domain interacting with 9 consecutive base pairs essentially covering a turn of the DNA helix
(33), USF, a basic helix-loop-helix transcription factor that binds to
DNA as a dimer with the two basic regions of the monomers bound to the
major groove on opposite sides of the helix essentially grasping DNA
like a scissors (34), and NF-B, which belongs to the rel domain
family of proteins of which the entire rel homology region is required
for DNA binding. The crystal structure of the p50 homodimer form of
NF-
B demonstrated a new mode of DNA binding that is very different
from those of previously characterized transcription factors. The p50
homodimer wraps itself around DNA almost completely surrounding the DNA
helix (35, 36). Nucleosome-length DNA fragments bearing DNA binding
sites for each of these respective factors were reconstituted into
nucleosome cores. Fig. 1A shows a mobility
shift assay demonstrating that each of the DNA fragments used in the
following experiments were 90-95% reconstituted into nucleosomes.
DNase I footprinting was employed to assay the ability of SWI/SNF to stimulate binding to nucleosomes by each of these factors. An analysis of Sp1 binding is shown in Fig. 1B. Reconstitution of the DNA fragment bearing the Sp1 recognition sequence resulted in a pattern of DNase I cutting that was distinct from that of naked DNA (compare lanes 1 and 5). When the nucleosomes were treated with the SWI/SNF complex, the DNase cutting pattern was perturbed (compare lanes 5 and 9), illustrating SWI/SNF interaction with these nucleosomes. While Sp1 readily bound the mock-reconstituted naked DNA control resulting in a clear footprint at 1 nM Sp1 (lane 2), binding was not observed to the reconstituted nucleosomes at even 100 nM Sp1 (lane 8). However, Sp1 binding was observed when the nucleosomes were also treated with the SWI/SNF complex (compare lanes 9 and 12).
The effect of SWI/SNF treatment on nucleosome binding by USF and
NF-B is shown in Fig. 1, C and D,
respectively. Each of these factors also bound the mock-reconstituted
naked DNA probes with nanomolar affinities (lanes 1-4 of
each panel). In the case of USF, at high concentrations of
the factor alone some protection of the USF site from DNase I cleavage
was observed on the nucleosome template (Fig. 1C,
lanes 5-8). However, in the presence of the SWI/SNF complex
nucleosome binding by USF was stimulated greater than 10-fold (Fig.
1C, lanes 9-12). There was no detectable binding of NF-
B to the nucleosome template at 110 nM factor
(Fig. 1C, lanes 5-9). By contrast, when the
NF-
B site nucleosome was treated with the SWI/SNF complex a clear
footprint was observed at 11 nM (lane 10)
indicating greater than 10-fold stimulation of NF-
B binding. Similar
to the Sp1 site nucleosome (Fig. 1B) the action of SWI/SNF
also altered the DNase I digestion pattern of the USF site and NF-
B
site nucleosomes (compare lanes 5 and 9 of Fig. 1, C and D, respectively). Thus, as with GAL4
derivatives the disruption of nucleosomal DNA by the SWI/SNF complex
also resulted in enhanced affinity of Sp1, USF, and NF-
B for
nucleosomal DNA.
Since the nucleosomes in Fig. 1 did not display a strong rotational phasing pattern by DNase I (i.e. a 10-bp ladder), it is likely that a mixed population of rotational orientations of the DNA on nucleosomes were present. Thus, in these cases the precise orientation of the binding sites with respect to the histone octamer was heterogeneous. However, stimulation of factor binding by SWI/SNF was observed in all cases indicating that stimulation by SWI/SNF was not dependent on a particular rotational phasing of the DNA relative to the histone octamer.
The data in Fig. 1 illustrate that the SWI/SNF complex is able to stimulate nucleosome binding by transcription factors with diverse DNA binding domains. However, the SWI/SNF complex also binds to DNA (37) and nucleosomes2 with nanomolar affinities. Thus, in the DNase I footprinting experiments shown in Fig. 1 as well as in footprinting experiments from previous reports (7, 8, 18, 19) the enhanced affinity of the factors represents a comparison of its affinity to a SWI/SNF-nucleosome complex versus that of a nucleosome alone. Therefore, it is useful to determine whether the observed enhanced binding is dependent on the continued binding of the SWI/SNF-complex or whether it persists once SWI/SNF has detached from the nucleosome. Indeed, only GAL4 derivatives have been shown (i.e. by gel shift analysis) to form a stable ternary complex with nucleosomes as a result of SWI/SNF action that persists following detachment of the SWI/SNF complex (18, 21).
While the SWI/SNF complex binds to nucleosomes with nanomolar affinity,
it can be competed off of nucleosomes to which it is bound by the
addition of cold nucleosomes and/or naked DNA (21). Thus, to determine
whether the SWI/SNF-induced nucleosome binding by Sp1, USF, or NF-B
resulted in a stable ternary factor-nucleosome complex we competed
SWI/SNF off the target nucleosomes following the binding reactions
(Fig. 2). Following this competition we tested for the
presence of stable factor-nucleosome complexes by gel shift analysis.
This analysis is shown in Fig. 3. Fig. 3A
shows mobility shift of Sp1 site nucleosomes that were incubated with
SWI/SNF in the presence or the absence of ATP and Sp1 followed by
competition of the SWI/SNF complex. In the absence of ATP very little
Sp1-nucleosome complex was observed (lanes 3 and
5). By contrast, when ATP was present to allow SWI/SNF
function there was a clear stimulation of Sp1-nucleosome complex
formation (compare lane 3 to 4 and lane
5 to 6). The loss of a fraction of the Sp1-nucleosome complexes into Sp1 aggregates (not resolved on the gel) due to the
self-association properties of Sp1 accounts for the reduction of total
radioactivity in lanes 4 and 6. While this effect
and nonspecific competition of Sp1 by the added competitor chromatin and nonspecific DNA reduced the apparent amount of Sp1 binding relative
to the footprinting reactions above, it is clear that Sp1-nucleosome
ternary complexes persisted following the detachment of SWI/SNF. Thus,
the ATP-dependent Sp1-nucleosome complexes formed by the
action of SWI/SNF did not require the continued presence of
SWI/SNF.
Both USF and NF-B also formed stable factor nucleosome complexes as
a consequence of SWI/SNF action (Fig. 3, B and
C). Fig. 3B clearly indicates that stable USF
binding to nucleosomes was also stimulated by the SWI/SNF complex in
the presence of ATP. The formation of a stable USF-nucleosome complex
that persisted after competition of the SWI/SNF complex was stimulated
by the presence of ATP (compare lane 3 to 4 and
5 to 6). Fig. 3C shows a similar
analysis using NF-
B. In the presence of SWI/SNF and ATP the
formation of a stable NF-
B/nucleosome complex was enhanced (Fig.
3C, compare lanes 3 and 4, and
5 and 6). Thus, the data in Fig. 1 illustrate
that the SWI/SNF complex stimulates the binding of disparate
transcription factors to nucleosomes. The data in Fig. 3 illustrate
that SWI/SNF action also results in the formation of transcription
factor-nucleosome complexes, which persisted independent of continued
SWI/SNF binding.
The experiments described above argue that SWI/SNF stimulation of transcription factor binding to nucleosomes is universal and therefore unlikely to provide distinction between SWI-dependent and -independent promoters. We therefore considered the alternative possibility that SWI/SNF function in stimulating factor binding is in part redundant with alternative mechanisms of nucleosome binding. For example, many promoter/enhancer regions contain multiple transcription factor binding sites, which may lead to cooperative binding in chromatin. Cooperative binding of 5 GAL4 dimers to nucleosomes has been shown to occur in vitro (27, 38). The observed cooperativity is in response to inhibition from the histone octamer and stimulates the binding of multiple GAL4 dimers relative to binding of individual GAL4 dimers to a single GAL4 site (especially near the center of the nucleosome). We thus tested whether cooperative binding of GAL4-VP16 to nucleosome cores would enhance or reduce the stimulation observed by the SWI/SNF complex.
Analysis of GAL4-VP16 binding to DNA or nucleosome cores bearing 5 GAL4
sites is illustrated Fig. 4. When increasing amounts of
GAL4-VP16 were added to the mock-reconstituted naked DNA (lanes 1-6), complete protection of the 5 sites was observed at 14 nM protein (lane 4). When reconstituted into a
nucleosome, this DNA fragment generates a distinct 10-11-bp repeating
pattern of hypersensitive cutting by DNase I indicating that this
fragment has a preferred path of DNA bending around the histone octamer
(lane 7). When GAL4-VP16 was titrated into reactions
containing the reconstituted 5-site nucleosome, complete protection of
the GAL4 sites was apparent at 140 nM protein (lane
12). Thus, a 10-fold repression of GAL4-VP16 binding to the 5 GAL4
site probe resulted from the presence of the histone octamer. When
SWI/SNF was also included, the DNase I digestion pattern of the 5-site
nucleosome was disrupted (lane 14). Concurrent with
nucleosome disruption by SWI/SNF, the affinity of GAL4-VP16 was
enhanced allowing full protection of the GAL4 sites at 14 nM GAL4-VP16 (lanes 14-20).
The data in Fig. 4 illustrate that under conditions of cooperative
nucleosome binding the SWI/SNF complex stimulated the binding of
GAL4-VP16 to nucleosomes 10-fold. By contrast, stimulation of GAL4-VP16
binding to a nucleosome bearing only a single GAL4 site was greater.
Fig. 5 illustrates the binding of GAL4-VP16 to a
mock-reconstituted naked DNA fragment (lanes 1-6) or the same fragment after reconstitution into nucleosome cores (lanes 7-13). Greater than 50% protection of the GAL4 site on the naked DNA was observed at 4.5 nM GAL4-VP16 with complete
protection at 14 nM (lanes 3 and 4). When
binding to the nucleosome core, however, half-protection of the GAL4
site required 1400 nM GAL4-VP16 (lane 12)
indicating that GAL4-VP16 binding to this single site in the nucleosome
was inhibited approximately 300-fold relative to naked DNA. However,
when the SWI/SNF complex was included, half-protection of the GAL4 site
occurred between 14 and 45 nM GAL4-VP16 (lane
16). Thus, the SWI/SNF complex stimulated the binding of a single
GAL4 dimer to a nucleosome core by approximately 100-fold.
The data in Figs. 4 and 5 illustrate several important conclusions regarding GAL4-VP16 binding to nucleosome cores and the function of the SWI/SNF complex. 1) Five GAL4-VP16 dimers bound nucleosomes with an affinity that was more than 10-fold greater than the binding of a single GAL4-VP16 dimer. Indeed, the cooperative binding of 5 dimers was only inhibited by the presence of the histone octamer 10-fold relative to GAL4-VP16 binding to naked DNA. 2) The SWI/SNF complex simulated the binding of GAL4-VP16 to the single-site nucleosome 100-fold but only stimulated the binding of GAL4-VP16 to the 5-site nucleosome 10-fold. Thus, SWI/SNF exerted the most dramatic effect where inhibition of GAL4-VP16 binding to nucleosomes was greatest. 3) The SWI/SNF complex did not enhance the affinity of GAL4-VP16 for nucleosomes beyond its affinity for naked DNA. The binding of 5 GAL4-VP16 dimers to nucleosomes in the presence of SWI/SNF occurred with an affinity equal to that of naked DNA.
To determine whether cooperative binding might similarly reduce the
extent of SWI/SNF stimulation of different factors we have also tested
SWI/SNF stimulation of NF-B binding to nucleosomes containing either
1 or 3 NF-
B sites. As we were unable to obtain a DNase I footprint
of NF-
B binding to a single-site nucleosome in the absence of
SWI/SNF we utilized the gel shift assay where partial occupancy could
be measured. Fig. 6 shows a mobility shift analysis of
binding to a 3 NF-
B site (NFX3) and a single-site (NFX1) nucleosome. Mock-reconstituted DNA lanes
(1, 2, 6, and 7) were
incubated with 3.3 nM NF-
B and used to illustrate the migration of NF-
B/DNA complexes (lanes 2 and
7). Binding of NF-
B to the first site of the 3-site
nucleosome occurred easily due to its proximity to the edge of the
nucleosome (13 bp from the end). However, binding to the next two sites
(33 and 53 bp into the nucleosome, respectively) occurred in a
"cooperative" manner as very few two NF-
B nucleosome complexes
were observed relative to three NF-
B complexes (lane 4).
In the presence of 330 nM NF-
B, 16.9% of the total
nucleosomes were bound by 3 NF-
B molecules (lane 4). When
the SWI/SNF complex was added to the reaction mixture, this binding was
stimulated 2.9-fold resulting in the occupancy of 49.7% of the
nucleosomes with 3 NF-
B proteins (lane 5). In the case of
the nucleosome with a single NF-
B site 30 bp into the nucleosome
core, NF-
B binding was more greatly inhibited by the histone octamer
than binding to the 3-site nucleosome indicating that multiple sites
increased NF-
B binding (compare levels of the 3 NF-
B-nucleosome
complex in lane 4 to the levels of 1 NF-
B-nucleosome complex in lane 9). Addition of 330 nM NF-
B
resulted in NF-
B binding to only 4.6% of the single-site
nucleosomes (lane 9). However, upon incubation with SWI/SNF
binding of NF-
B to the single-site nucleosome was increased to
36.5%, a 7.9-fold increase. Thus, the extent of SWI/SNF stimulation of
NF-
B binding to a nucleosome bearing 3 sites relative to a single
site was reduced approximately 3-fold.
The data in Figs. 4, 5, 6 illustrate that multimerization of binding sites
for either GAL4-VP16 or NF-B enhanced the extent of binding to sites
well within the nucleosome and in turn reduced the extent of SWI/SNF
stimulation observed. These observations indicate that degree of
SWI/SNF stimulation of factor binding to nucleosomes is inversely
related to the degree of inhibition by the histone octamer. Thus,
parameters that reduce nucleosome-mediated inhibition of factor
binding, such as cooperative binding, are partially redundant with the
function of the SWI/SNF complex in facilitating binding of
transcription factors to nucleosomal DNA.
The SWI/SNF complex participates in the activation of several
yeast genes in vivo (3); however, many inducible genes do not require SWI/SNF, even some that have been shown to undergo chromatin remodeling upon induction, such as PHO5 (13, 39, 40). We have tested the possibility that the differential effect of
SWI/SNF might reflect a selective ability of different transcription activators to exploit SWI/SNF-mediated disruption of nucleosomes to
gain access to their binding sites in nucleosomes. By biochemical analysis we have found that the purified SWI/SNF complex stimulated nucleosome binding by three different transcription factors, Sp1, USF,
and NF-B concomitant with the ATP-dependent SWI/SNF
disruption of histone-DNA contacts. Moreover, the function of SWI/SNF
resulted in the formation of stable factor nucleosome complexes
containing any of these factors that persisted after detachment of the
SWI/SNF complex. The fact that these factors all contain very different types of DNA binding domains indicates that the stimulation of stable
transcription factor binding to nucleosomes by SWI/SNF is a general
mechanism. Most DNA binding proteins are inhibited by the presence of
nucleosomes. Our data illustrate that SWI/SNF perturbation of
histone-DNA interactions involves contacts which are generally
important for repression of transcription factor binding. Thus, the
effect of SWI/SNF action on activator binding is universal and likely
does not play a major role in determining which genes are
SWI-dependent.
A common quality of many promoter/enhancer regions is the occurrence of
multiple binding sites for upstream activators which may lead to
cooperative binding to nucleosomes (25). The presence of multiple
adjacent GAL4 sites within a nucleosome leads to cooperative binding by
GAL4 derivatives in vitro. (27, 38). This cooperativity led
to only a 10-fold repression of GAL4-VP16 binding to nucleosomes relative to naked DNA, which was completely reversed by the SWI/SNF complex (Fig. 4). By contrast, SWI/SNF was able to stimulate the binding of a single GAL4-VP16 dimer by over 100-fold since the repression of GAL4-VP16 binding for a single dimer was greater than
300-fold. Thus, the requirement for SWI/SNF function in the in
vitro binding of GAL4-VP16 to nucleosome cores is partially redundant with cooperative nucleosome binding. Furthermore, SWI/SNF had
a greater stimulatory effect on the binding of a single NF-B to its
nucleosomal recognition site as compared with its effect on NF-
B
binding to 3 sites in a nucleosome. The fact that similar effects are
seen with a different type of transcription factor supports the general
idea that reducing the repressive effect of nucleosomes in turn reduces
the stimulation effect of the SWI/SNF complex. This observation raises
the possibility that SWI/SNF function may be redundant with cooperative
binding and other mechanisms that enhance the binding of transcription
factors to nucleosomes in vivo. In this manner, SWI/SNF may
act relatively promiscuously in the yeast genome, but is required only
at the genes which lack other sufficient mechanisms to contend with
chromatin structures.
Two other nucleosome remodeling activities have been biochemically identified that are distinct from SWI/SNF but have some similar biochemical activities. The NURF (ISWI) complex was purified from Drosophila (6, 41) and the RSC (STH1) complex from yeast (5). ISWI is the catalytic subunit of NURF for which homologs have been identified in yeast as well as humans (42). These complexes may also function by antagonizing the repressive effects of chromatin and thus might in principle be functionally redundant with SWI/SNF. It will be interesting to see if these distinct nucleosome disrupting activities provide overlapping functions in the yeast nucleus.
We thank members of the Workman laboratory for assistance and advice during the course of these studies.