Mutations in the AF-2/Hormone-binding Domain of the Chimeric Activator GAL4 z Estrogen Receptor z VP16 Inhibit Hormone-dependent Transcriptional Activation and Chromatin Remodeling in Yeast*

GAL4 z estrogen receptor z VP16 (GAL4 z ER z VP16), which contains the GAL4 DNA-binding domain, the human ER hormone binding (AF-2) domain, and the VP16 activation domain, functions as a hormone-dependent transcriptional activator in yeast (Louvion, J.-F., Havaux-Copf, B., and Picard, D. (1993) Gene ( Amst .) 131, 129–134). Previously, we showed that this activator can remodel chromatin in yeast in a hormone-dependent manner. In this work, we show that a weakened VP16 activation domain in GAL4 z ER z VP16 still allows hormone-depend-ent chromatin remodeling, but mutations in the AF-2 domain that abolish activity in the native ER also eliminate the ability of GAL4 z ER z VP16 to activate transcription and to remodel chromatin. These findings suggest that an important role of the AF-2 domain in the native ER is to mask the activation potential of the AF-1 activation domain in the unliganded state; upon ligand activation, a conformational change releases AF-2-medi-ated repression and transcriptional activation ensues. We also show that the AF-2 domain, although inactive at simple promoters on its own in yeast, can enhance transcription by the MCM1 activator in hormone-dependent manner, consistent with its having a role in activation as well as repression in the native ER.

The estrogen receptor (ER) 1 is a member of a superfamily of nuclear receptors, which are ligand-activated transcription factors. These proteins have independent modular domains capable of DNA binding, ligand binding (LBD), and transcriptional activation (1). Two distinct domains have transactivation capacity: a hormone-independent transactivation function (AF-1) resides in the N terminus, while the hormone-dependent AF-2 lies within the hormone-binding domain (2,3). These domains appear to interact functionally and probably physically. For example, when expressed as separate peptides in mammalian cells, the two domains synergize to activate a reporter plasmid only in the presence of hormone (4). The AF-1 domain displays constitutive activity in some cells (3) and is active in yeast cells, and both the AF-1 and the AF-2 domains can behave as transcriptional activators when fused to a heterologous DNA-binding domain (5,6). Additionally, coactivators for nuclear receptors interact with AF-2 in mammalian cells (7).
With the determination of the partial crystal structure (LBD) of four nuclear receptors, including the liganded estrogen receptor (8 -11), the mechanism of action of this superfamily is coming into focus. The ligand-binding domain of nuclear receptors appears to act as a switch or a "mouse-trap." The ligand binds to a pocket in the receptor, tripping a conformational change that rearranges helix 12 to form a "lid" over the agonist. Helix 12 contains the AF-2 core, and Wurtz et al. (12) suggest that realignment may create a new surface for interactions with co-activators and/or break connections with repressors. Brzozowski et al. (8) have shown that an estrogen antagonist prevents the alignment of helix 12, providing further evidence that activation requires this new surface.
However, it is still not clear how tripping the mouse-trap of the LBD leads to activation of the receptor. It is certain that the AF-2 core is vital for productive interaction with agonist. Danielian et al. (13) demonstrated that mutations in this core significantly reduce ligand-dependent transcription, without affecting steroid or DNA binding. The co-activators RIP 140 and TIF2 likewise require AF-2 core activity to interact with nuclear receptors (14,15), while SRC-1 also requires a lysine residue in helix 3 (16).
Experiments utilizing the modular domains of nuclear receptors are instructional in elucidating the mechanisms of receptor activation and transcription. Previously, using a chimeric transcription factor, GAL4⅐ER⅐VP16, we demonstrated that disruption of chromatin structure required an unmasked activation domain (17). GAL4⅐ER⅐VP16 is composed of the yeast GAL4 DNA-binding domain, the human estrogen receptor hormone binding/AF-2 domain, and the viral VP16 activation domain (18). This hormone-dependent transcriptional activator perturbed chromatin in a yeast episome, outside of the context of a bona fide promoter, only when both hormone and the VP16 activation domain were present. Likewise, the yeast GAL4 protein significantly disrupted chromatin structure only when in an active form.
Here, we further show that a weakened VP16 domain in GAL4⅐ER⅐VP16 does not abrogate its ability to disrupt chromatin. Conversely, mutations that abolish AF-2 activity without compromising hormone binding or binding to DNA in the intact estrogen receptor (13) also eliminate the ability of GAL4⅐ER⅐VP16 to activate a lacZ reporter and disrupt chromatin structure without interfering with the ability of the chimeric receptor to bind hormone or to bind to DNA. These results suggest that the AF-2 domain has a repressive function in the unliganded ER, which can act even on a heterologous activation domain, and that mutations in the AF-2 domain can prevent ligand from "unlocking" this configuration. Additionally, these results support the tight correlation between an intact activation domain and chromatin remodeling observed previously (17, 19 -21). We also show that, while eliminating the VP16 moiety results in a factor incapable of independently activating transcription or remodeling our chromatin reporter, GAL4⅐ER can still synergize with a proximal activator.
Analysis of Plasmid Chromatin-Yeast cells (1 liter) were grown at 30°C to A 600 between 0.6 and 1.6. Yeast nuclei were prepared as described previously (17,30) and digested with varying concentrations of micrococcal nuclease (MNase) (Worthington) for 10 min at 37°C. Cleavage patterns were consistent over a range of 0 -50 units/ml. Naked DNA controls were purified from nuclei preparations prior to digestion with MNase. Cleavage patterns were visualized by indirect end labeling (31,32). Following clean-up with phenol and chloroform, aliquots were treated with RNase and digested with EcoRV. The samples were electrophoresed along with HaeIII-digested ⌽X markers in a 1.2% agarose gel at 4 V/cm for 5-5.5 h. The DNA was transferred to nylon membranes (Duralon UV, Stratagene) and Southern analysis performed. Probes were EcoRV to XbaI fragments from TALS prepared by polymerase chain reaction. Indirect end label analysis was done in two independent experiments for each of the AF-2 mutants examined here, as well as for the parent F442P receptor.
Topoisomer Analysis-DNA was prepared by pelleting 10 ml of cell cultures grown to A 600 ϭ 0.6 -1.2, resuspending in 500 l of 10 mM Tris, 1 mM EDTA, and rapidly lysing with glass beads in the presence of 100 l of 5% SDS, 5 mg/ml proteinase K. Purified DNA was separated on 1.5% agarose gels with 40 g/ml chloroquine diphosphate (Sigma) in both gel and buffer at 2.5 V/cm for 18 -20 h. The gel was blotted and probed as above. Quantitation of topoisomers was performed by Phos-phorImager analysis (Molecular Dynamics) and the Gaussian centers of distribution calculated.

RESULTS
A Slight Reduction in the Activating Potential of GAL4⅐ER⅐VP16 Does Not Affect Its Ability to Remodel Chromatin-We previously used the chimeric transcription factor GAL4⅐ER⅐VP16 to demonstrate activation domain-dependent chromatin disruption in the yeast episome TALS, which contains a strong binding site for GAL4 within a positioned nucleosome (17). Since it has been reported that the DNA binding activity of the intact estrogen receptor is affected by hormone addition (33-35), we wished to separate possible effects of the hormone-binding domain of GAL4⅐ER⅐VP16 on its DNA-binding domain from those due to unmasking of the activation domain. We first examined the effect of reducing the activation potential of GAL4⅐ER⅐VP16 by introducing a point mutation into the VP16 activation domain, F442P, which reduces activity of GAL4-VP16 at a CYC1-lacZ reporter driven by the GAL1-10 UAS by two-thirds (36). Transcription was monitored using a lacZ construct containing a single strong GAL4 binding site (UAS 17 ) in the presence and absence of hormone. Transcription by the mutant GAL4⅐ER⅐VP16(F442P) showed a modest reduction, to about 60% of the activity induced by GAL4⅐ER⅐VP16 (Fig. 1). (Although the standard errors for these measurements were fairly large, we consistently observed lower activity of the mutant than of the parental GAL4⅐ER⅐VP16 when the two were measured simultaneously. Furthermore, Student's t test yields p Ͻ 0.001 for the null hypothesis that the two activators are equally active.) We next assessed the ability of GAL4⅐ER⅐VP16(F442P) to remodel chromatin structure in the TALS reporter plasmid. The TALS minichromosome is packaged into strongly positioned nucleosomes in yeast ␣ cells by the ␣2⅐MCM1 protein complex in conjunction with Ssn6p and Tup1p (37). We assayed chromatin remodeling in TALS by following hormone-dependent changes in MNase cleavage patterns in the presence of GAL4⅐ER⅐VP16(F442P). Isolated nuclei were subjected to different concentrations of MNase for 10 min, followed by indirect 2 S. Hanes, unpublished results.

FIG. 1. Transcriptional activation by GAL4⅐ER⅐VP16 derivatives.
Activation of pRS314 -17⌬80lacZ, a CYC1-lacZ reporter gene with a single strong GAL4 binding site (UAS 17 ), by derivatives of GAL4⅐ER⅐VP16 was monitored in the absence or presence of ␤-estradiol (Ϯ E2). The VP16 activation domain was either absent or present as wild type or with the F442P mutation, as indicated, and the AF-2 domain was either present as wild type or mutant, or absent, as indicated. The first two columns show measurements done in the absence of GAL4⅐ER⅐VP16. All measurements were performed with cells grown in glucose medium, and are averages of at least five independent determinations; standard errors are indicated or else are too small to be visible. end label analysis (31,32). The pattern of MNase cleavages in TALS chromatin induced by GAL4⅐ER⅐VP16(F442P) changed upon the addition of ␤-estradiol (Fig. 2, left panel). The lowest arrowhead marks a site of enhanced cleavage seen when cells are grown with hormone. The upper two arrowheads mark cleavage sites seen only in the presence of hormone. These hormone-dependent alterations in TALS chromatin demonstrate perturbation of nucleosome IV, which contains the GAL4 binding site, and the adjacent nucleosome III, and are indistinguishable from those seen previously with GAL4⅐ER⅐VP16 (17).
Perturbation of chromatin structure in a closed circular minichromosome can result in a change in the distribution of supercoiled topoisomers, since each nucleosome in the plasmid constrains one negative supercoil (38). To investigate the effect of GAL4⅐ER⅐VP16(F442P) on TALS topology, DNA from cells expressing the chimeric factor was rapidly harvested and treated to inactivate topoisomerases (so that the distribution of supercoiled plasmids reflected the in vivo distribution). We previously demonstrated that GAL4⅐ER⅐VP16 in the absence of ␤-estradiol did not alter plasmid topology as compared with DNA from cells lacking GAL4⅐ER⅐VP16. Cells harboring GAL4⅐ER⅐VP16 and grown in the presence of hormone exhibited a shift in the center of distribution of the topoisomers equivalent to the loss of almost one negative turn (17). Hormone induction of GAL4⅐ER⅐VP16(F442P) likewise resulted in the loss of negative supercoiling in TALS (Fig. 3, lanes 1 and 2). These topoisomer distributions conform to Gaussian distributions, which allows their centers to be measured precisely from the relative intensities of individual topoisomers (39). Consequently, differences in topology can also be measured with precision. Quantitation yielded a value for the loss of negative supercoiling in TALS in the presence of hormone-activated GAL4⅐ER⅐VP16(F442P) of 0.6 Ϯ 0.2 (Table I), similar to the value seen with GAL4⅐ER⅐VP16 (0.7 Ϯ 0.1) (17). Thus, the modest reduction in activation potential caused by the F442P mutation did not detectably diminish the ability of GAL4⅐ ER⅐VP16 to remodel TALS chromatin.
A Functional AF-2 Core Is Required for Both Transcriptional Activation and Chromatin Disruption by GAL4⅐ER⅐VP16 -The hormone-binding domain of the estrogen receptor also contains an activation domain (AF-2). The AF-2 domain can activate transcription in the absence of the AF-1 or any other transcriptional activation domain in mammalian cells (2). In yeast, the AF-2 domain apparently lacks a universal ability to activate transcription on its own; although it can activate transcription from a complex promoter, it does not activate trancription from a simple promoter lacking binding site(s) for other activators (6,17,18).
Either of two pairs of mutations within AF-2 (L539A/L540A and M543A/L544A of the human ER) almost totally abolishes transcriptional activation by the estrogen receptor in mammalian cells, without affecting DNA or ligand binding (13). In order to assess the requirements for AF-2 activity in the context of the chimeric activator, we introduced these mutations into GAL4⅐ER⅐VP16(F442P) and expressed the resulting mutant proteins in yeast.
To demonstrate that the mutant GAL4⅐ER⅐VP16 receptors still effectively bound hormone, we first attempted binding assays using radiolabeled ␤-estradiol with both intact yeast cells (40) or yeast cellular extracts (41). Unfortunately, although binding was easily measured and could be competed with unlabeled ␤-estradiol, identical results were obtained whether or not the yeast cells expressed GAL4⅐ER⅐VP16. As an alternative, we therefore performed an in vivo competition assay. YJ0 cells harboring an expression vector for LexA⅐ER⅐ VP16 and a ␤-galactosidase reporter containing eight LexA binding sites were transformed with expression plasmids for mutant and wild type GAL4⅐ER⅐VP16(F442P) or an empty vector control. LexA⅐ER⅐VP16 is identical to GAL4⅐ER⅐VP16, except it contains the LexA DNA-binding domain in place of that for GAL4, and functions as a hormone-dependent transcriptional activator via LexA binding sites. 3 At limiting hormone concentrations, we expected that the GAL4⅐ER⅐VP16 chimeras might compete with LexA⅐ER⅐VP16 for the available ␤-estradiol and hence decrease transcription of the ␤-galactosidase reporter gene. As shown in Table II, ␤-galactosidase activity induced by LexA⅐ER⅐VP16 in the presence of 2.5 nM ␤-estradiol was indeed reduced by about 40% in cells harboring pRS414GAL4⅐ER⅐VP16(F442P) compared with cells harboring the empty vector pRS414. Similarly, the chimeric GAL4⅐ER⅐ VP16(F442P) receptors containing mutated AF-2 domains also reduced ␤-galactosidase activity induced by LexA⅐ER⅐VP16 by about 50% (Table II). In contrast, ␤-galactosidase activity induced by LexA⅐ER⅐VP16 was reduced only slightly by either the wild type or mutant GAL4⅐ER⅐VP16(F442P) receptors at 250 nM ␤-estradiol (Table II), supporting the interpretation of the results at 2.5 nM ␤-estradiol as being due to competition for limiting hormone. Thus, these results indicate that, as for the intact ER (13), the L539A/L540A and M543A/L544A mutations in the AF-2 domain do not affect hormone binding in the GAL4⅐ER⅐VP16 chimeras.
When GAL4⅐ER⅐VP16(F442P) harboring either the L539A/ L540A or M543A/L544A mutation were assayed for transcriptional activity using the UAS 17 -lacZ reporter, both were found to be completely inactive (Fig. 1). The L539A/L540A and M543A/L544A mutations also result in nearly complete inactivation of GAL4⅐ER⅐VP16 (i.e. with completely active VP16) ( Fig. 1 and data not shown). The data of Table II suggest that the mutated proteins were expressed at levels comparable to the parent GAL4⅐ER⅐VP16(F442P). To provide further evidence for their expression and to show that the AF-2 mutants were capable of binding to a GAL4 binding site, we examined the ability of unliganded GAL4⅐ER⅐VP16, GAL4⅐ER⅐VP16(F442P), and the AF-2 mutants derived from the latter to interfere with transcriptional activation by native GAL4 in cells grown in galactose-containing media (Fig. 4). Both unliganded mutant and wild-type receptors interfered with activation by GAL4, indicating that they occupy the UAS 17 to similar extents. As a control, we examined transcription of a LexA-lacZ reporter by a fusion protein containing the LexA DNA-binding domain and the GAL4 activation domain and found that transcription was unaffected by the presence of unliganded GAL4⅐ER⅐VP16 (data not shown). Thus, even in the context of a heterologous activation domain [VP16 or VP16(F442P)], the chimeric receptor requires an active AF-2 function to activate transcription.
We next assessed the effect of the AF-2 mutation on the ability of the chimeric activator to remodel chromatin. In contrast to GAL4⅐ER⅐VP16(F442P), GAL4⅐ER(L539A;L540A)⅐VP16 (F442P) does not alter the MNase cleavage pattern of TALS chromatin upon addition of ␤-estradiol (Fig. 2). Consistent with this result, the distribution of supercoiled TALS topoisomers is not affected by hormone induction in the presence of the AF-2 mutants (Fig. 3, lanes 3 and 4; Table I). We conclude that the AF-2 mutations, which do not affect DNA binding or hormone binding (13), abolish the ability of GAL4⅐ER⅐VP16(F442P) both to activate transcription and to remodel chromatin.
Hormone-dependent Synergy between the AF-2 Domain and the Yeast Activator MCM1-To investigate the effect of a nonactivating GAL4⅐ER⅐VP16 derivative with an intact regulatory (ER) domain, we excised the VP16 moiety to recover GAL4⅐ER. GAL4⅐ER is unable to activate transcription from the UAS 17 -lacZ reporter in our assay ( Fig. 1 and Ref. 17). It does, however, measurably inhibit transcription by endogenous GAL4, showing that it is expressed and capable of DNA binding (17). GAL4⅐ER binding slightly affects TALS chromatin, but addition of hormone does not induce any further changes, as assayed by MNase digestion, restriction enzyme accessibility, or plasmid topology (17).
To discern whether the presence of hormone could increase DNA binding of GAL4⅐ER, we utilized a different lacZ reporter, pRS314␣2GAL4lacZ⌬Nco. In addition to a single GAL4 binding site in the promoter, this reporter has a binding site for the ␣2 and MCM1 proteins further upstream. In yeast a cells, which lack ␣2 protein, MCM1 binds to this site and activates transcription ( Fig. 5; Ref. 42). Non-activating proteins binding at the GAL4 site are expected to interfere sterically with activation by MCM1 (43). Indeed, GAL4⅐ER measurably reduced lacZ activity (Fig. 5) in the absence of hormone, as did GAL4⅐ER(L539A;L540A) and GAL4⅐ER(M543A;L544A). However, the addition of hormone did not result in a greater reduction of lacZ activity by GAL4⅐ER, as would be expected if 3

AF-2 Mutations in GAL4⅐ER⅐VP16
DNA-binding of transcriptionally inert GAL4⅐ER were increased by hormone binding. Instead, lacZ activity was increased nearly 2-fold (Fig. 5). This level is higher than that seen in the absence of any GAL4⅐ER, indicating that it cannot be attributed to loss of GAL4⅐ER binding, but rather must reflect enhanced transcription from the combined effects of MCM1 and ligand-bound GAL4⅐ER. We observed a similar effect in a different yeast strain, YJ0 (data not shown). This increase in activity required a functional AF-2 core, as neither GAL4⅐ER(L539A;L540A) nor GAL4⅐ER(M543A;L544A) increased MCM1-activated transcription in the presence of ␤-estradiol (Fig. 5).

DISCUSSION
The results presented here support the view of the hormone binding/AF-2 domain of the human estrogen receptor as a modular entity that is capable of regulating the activity of a heterologous activation domain, in agreement with previous results (Ref. 18 and references therein). We further show that the hormone-dependent release of the heterologous VP16 activation domain from ER-mediated repression is abolished by mutations in the AF-2 domain (Fig. 1), which do not affect DNA binding or ligand binding (Figs. 4 and 5; Table II; Ref. 13), suggesting that these mutations lock the AF-2 domain in a repressive configuration. These findings suggest that an important role of the AF-2 domain in the native ER is to mask the activation potential of AF-1 in the unliganded state; upon ligand activation, a conformational change releases AF-2-mediated repression and transcriptional activation ensues. Transcriptional activation by the native ER and other nuclear hormone receptors in their native, physiological contexts is clearly more complicated than this, with ligand binding result-ing in release of co-repressors and recruitment of co-activators (7). These additional complexities are absent in the heterologous system studied here, and their relative importance therefore cannot be assessed; nevertheless, the work reported here indicates a role for the unliganded AF-2 domain in directly preventing activation by a linked activation domain that is likely to be relevant to its normal function in the intact ER.
We also show that the AF-2 domain, although inactive at simple promoters on its own in yeast (Refs. 17 and 18; Fig. 1), can increase transcription by the MCM1 activator in a hormonedependent fashion (Fig. 5). This would be consistent with the AF-2 domain contributing to activation as well as repression in the native ER, as suggested by previous work (2)(3)(4). Finally, we demonstrate that the AF-2 mutations which abolish transcriptional activation by GAL4⅐ER⅐VP16 also abolish chromatin remodeling ( Figs. 2 and 3), supporting our previous conclusion that chromatin remodeling by GAL4⅐ER⅐VP16 and by GAL4 requires an unmasked transcription activation domain (17).
Considerable progress has been made toward understanding the mechanism whereby the binding of a ligand to the estrogen receptor promotes transcription. Structural studies have shown that ligand binding causes a conformational change within the ligand-binding domain, repositioning several helices, most importantly helix 12 (8 -11). It has been proposed that the proper placement of helix 12 creates a surface that interacts with coactivators (12). Although the evidence for coactivators in mammalian cells is overwhelming (7), any mechanism for ER activation must also include AF-1, which can function alone (3,5) and which synergizes with AF-2, even when expressed as a separate protein (4). The activation potential of AF-1 must therefore somehow be nullified in the absence of hormone in the intact ER. Furthermore, GAL4⅐ ER⅐VP16 is virtually inactive in the absence of hormone (17,18), indicating that AF-2 can prevent even a heterologous activation domain from functioning in the absence of ligand.  17 ) and an ␣2/MCM1 operator and expression vectors for the indicated GAL4⅐ER derivatives were grown in glucose medium, in the presence or absence of hormone, and ␤-galactosidase activity was measured. Values are averages of at least three independent measurements, and standard errors are indicated. Note that GAL4⅐ER does not activate transcription with or without hormone in the absence of an MCM1 binding site (Fig. 1).
The unliganded AF-2 domain could act as a modular repressor domain by interacting with another protein(s) that could prevent the receptor from binding to its cognate site. Steroid hormone receptors can complex with many proteins, including hsp90, which could prevent the estrogen receptor from binding to promoter sites (1). However, Lee et al. (44) have demonstrated that this mechanism cannot provide a complete explanation by designing chimeric receptors that do not bind hsp90, yet still display hormone dependence. They fused a VP16 activation domain N-terminal to the GAL4 DNA-binding domain and several truncations of the ER ligand-binding domain. These chimeras displayed 15-35% of the activity of VP16⅐GAL4 in the absence of hormone, but activity was enhanced 4-fold by estradiol. These results indicate that at least part of the hormone dependence of the AF-2 domain is independent of sequestration by hsp90, consistent with previous suggestions (45,46). Even more directly, we have found that GAL4⅐ER⅐VP16 and GAL4⅐ER bind DNA in the absence of ␤-estradiol, as inferred from their ability to inhibit activation by GAL4 ( Fig. 4; Ref. 17) and by MCM1 via steric interference (Fig. 5), as well as the ability of unliganded GAL4⅐ER⅐VP16 to perturb chromatin structure, albeit weakly (17).
The AF-2 domain of the unliganded receptor could interact with a repressor. For example, the thyroid hormone receptor binds the repressor SMRT, although its exact mechanism of action is unknown (47). However, the presence of a repressor in yeast that would act on the AF-2 of the mammalian receptor seems unlikely, although certainly not impossible (48,49). More tellingly, we do not see evidence of trans-repression (e.g. of MCM1), as we would expect from a repressor that contacts general transcription factors or modifies chromatin (as expected for a histone deacetylase; Ref. 50).
We propose that the unliganded AF-2 domain presents a surface that binds nearby activation domains (AF-1 or VP16), preventing them from contacting normal targets. If the unliganded AF-2 surface resembled such a target, it would explain why AF-2 inhibits a heterologous activator. This is also consistent with the differences between GAL4⅐ER⅐VP16, which is essentially inactive in the absence of hormone, and VP16⅐ GAL4⅐ER, which retains 15-35% of its hormone-stimulated activity even in the absence of hormone (44); the conformation of the activation domain with respect to AF-2 determines the extent of inhibition. We suggest that, in yeast, the ligandbinding domain principally regulates the linked activation domain, which, when unmasked, recruits general transcription factors and/or chromatin remodeling complex(es). When the core AF-2 is mutated, the second activation domain cannot be released from its repression by AF-2, and no remodeling or transcription is seen. This hypothesis does not dismiss the role of co-activators like SRC-1 and RIP-140 in the native context (7,14). Rather, it adds another level of control over the estrogen receptor.
The chromatin remodeling activity tightly correlates with the transcriptional potential of the receptor, consistent with findings with the thyroid hormone receptor and the yeast transcriptional activators Gal4p and Pho4p, as well as with our previous work (17, 19 -21). GAL4⅐ER was not able to activate transcription or cause chromatin remodeling (17), and the AF-2 mutants used in this paper cause simultaneous extinction of transcriptional activation and chromatin remodeling by GAL4⅐ER⅐VP16(F442P). Thus, chromatin remodeling appears generally to be intimately linked to transcriptional activation. Uncoupling these functions, other than by the trivial routes of disabling transcription by crippling critical promoter elements or downstream components of transcription (as with ␣-amanitin), is likely to require considerable ingenuity (51,52).
The increase in MCM1-activated transcription caused by ligand-bound GAL4⅐ER was unexpected, and suggests that there can exist a threshold below which an "activator" cannot function on its own, although it can still synergize with another (weak) activator. Similar results have been reported using the vitellogenin A1 io promoter (53). This promoter contains estrogen response elements which cannot activate the promoter on their own, but which contribute to transcription by Sp1. One mechanism suggested to account for this synergy was that Sp1 could interact with the same transcriptional machinery as the liganded estrogen receptor at the downstream estrogen response elements. In our system, it may be that the AF-2 domain alone does not interact strongly enough with the transcriptional machinery in yeast to activate transcription in isolation, but does interact sufficiently to stabilize interactions created by another activator (e.g. MCM1) and thus to increase levels of transcription. This mechanism does not differ in principle from mechanisms proposed to account for synergy between activators, but adds the proviso that one activator can be so weak that it cannot activate transcription by itself and yet still can enhance transcription if another, stronger activator acts at the same promoter.