An antiestrogen-responsive estrogen receptor-alpha mutant (D351Y) shows weak AF-2 activity in the presence of tamoxifen.

Antiestrogens, including tamoxifen and raloxifene, block estrogen receptor (ER) action by blocking the interactions of an estrogen-dependent activation function (AF-2) with p160 coactivators. Although tamoxifen does show some agonist activity in the presence of ERalpha, this stems from a distinct constitutive activation function (AF-1) that lies within the ERalpha N terminus. Previous studies identified a naturally occurring mutation (D351Y) that allows ERalpha to perceive tamoxifen and raloxifene as estrogens. Here, we examine the contributions of ERalpha activation functions to the D351Y phenotype. We find that the AF-2 function of ERalpha D351Y lacks detectable tamoxifen-dependent activity when tested in isolation but does synergize with AF-1 to allow enhanced tamoxifen response. Weak tamoxifen-dependent interactions between the ERalpha D351Y AF-2 function and GRIP1, a representative p160, can be detected in glutathione S-transferase binding assays and mammalian two-hybrid assays. Furthermore, tamoxifen-dependent AF-2 activity can be detected in the presence of ERalpha D351Y and high levels of overexpressed GRIP1. We therefore propose that the D351Y mutation allows weak tamoxifen-dependent AF-2 activity but that this activity is only detectable when AF-1 is strong, and AF-1 and AF-2 synergize, or when p160s are overexpressed. We discuss the possible structural basis of this effect.

Antiestrogens, including tamoxifen and raloxifene, block estrogen receptor (ER) action by blocking the interactions of an estrogen-dependent activation function (AF-2) with p160 coactivators. Although tamoxifen does show some agonist activity in the presence of ER␣, this stems from a distinct constitutive activation function (AF-1) that lies within the ER␣ N terminus. Previous studies identified a naturally occurring mutation (D351Y) that allows ER␣ to perceive tamoxifen and raloxifene as estrogens. Here, we examine the contributions of ER␣ activation functions to the D351Y phenotype. We find that the AF-2 function of ER␣ D351Y lacks detectable tamoxifen-dependent activity when tested in isolation but does synergize with AF-1 to allow enhanced tamoxifen response. Weak tamoxifen-dependent interactions between the ER␣ D351Y AF-2 function and GRIP1, a representative p160, can be detected in glutathione S-transferase binding assays and mammalian two-hybrid assays. Furthermore, tamoxifen-dependent AF-2 activity can be detected in the presence of ER␣ D351Y and high levels of overexpressed GRIP1. We therefore propose that the D351Y mutation allows weak tamoxifen-dependent AF-2 activity but that this activity is only detectable when AF-1 is strong, and AF-1 and AF-2 synergize, or when p160s are overexpressed. We discuss the possible structural basis of this effect.
Estrogens act by binding two specific intracellular receptor proteins (ERs 1 ␣ and ␤, hereafter ER␣ and ER␤; reviewed in Refs. [1][2][3][4]. Both receptors are conditional transcription factors that work either by binding to specific estrogen response elements (EREs) within the promoters of estrogen-regulated genes or by enhancing the activity of heterologous transcription factors such as the AP-1 (Jun⅐Fos) complex (5,6). Like other nuclear receptors, the ERs consist of three distinct domains, an N-terminal domain, a centrally located DNA binding domain (DBD), and a C-terminal ligand binding domain (LBD). Specific DNA recognition is mediated by the DBD, and transcriptional enhancement is mediated by the synergistic action of two separate activation functions, AF-1 and AF-2, that lie within the N-terminal domain and LBD, respectively. The overall process of transcriptional enhancement has two distinct hormone-dependent components. First, estrogens promote ER␣ dissociation from a heat shock protein-chaperonin complex that serves to restrict its activity. Second, AF-2 absolutely requires estrogens for its activity.
The ER activation functions work by recruiting coactivator proteins (7)(8)(9)(10). AF-2 binds to members of the p160 coactivator family, including GRIP1 (TIF2/NCoA2), SRC-1 (NCoA1), and ACTR (pCIP/RAC3/AIB1/TRAM1), which bind to the histone acetyltransferases CBP/p300 and pCAF. The ER AF-2 functions also bind to other coactivators, including TRAP220 (11,12), a component of the TRAP-DRIP-ARC-SMCC complex, PGC-1 (13), E6-AP (14), and others (reviewed in Ref. 10), although the significance of many of these interactions is not yet clear. In each case, AF-2 binds short amphipathic ␣-helices with the consensus LXXLL, termed nuclear receptor boxes (NR boxes), which are often reiterated several times within each coactivator. The AF-2 surface is composed a hydrophobic cleft made up of residues from helices 3-5 and 12 (15). A recent ER␣ co-crystal with a GRIP1 NR box peptide has revealed that there are two components of ER␣/NR box recognition (16). First, lysine and glutamic acid residues, which lie within ER␣ helices 3 and 12, respectively, form a charge clamp that stabilizes the carbamyl backbone of the NR box peptide. Second, residues within the hydrophobic cleft interact with the NR box leucines. ER␣ AF-1, which is cell type-specific and constitutive (17), also binds to p160s but does not bind NR boxes (18 -20). Instead, AF-1 recognizes a distinct surface that lies within the p160 C terminus (18).
Because estrogens stimulate the growth of about 50% of human breast tumors and also play a role in tumor incidence, drugs that antagonize estrogen action have found favor as breast cancer treatments and preventatives (21). Each of the available antiestrogens (including tamoxifen, raloxifene, and ICI 182,780) allows the ERs to bind to DNA but inhibits the ability of AF-2 to bind to coactivators (22). Tamoxifen and raloxifene do show some agonist activity at classical EREs, but this activity stems from AF-1 and not from residual AF-2 activity (17,(23)(24)(25). ER-tamoxifen and ER-raloxifene complex crystal structures have revealed the molecular basis of antiestrogen action (16,26,27). Unlike estrogens, which are completely buried within the LBD hydrophobic core, tamoxifen and raloxifene both possess a bulky side chain extension that protrudes through the LBD surface near the base of helix 12. This extension displaces helix 12, which rotates 110°and folds back into the remainder of the hydrophobic cleft thereby occluding the coactivator binding surface. Thus, the tamoxifen and raloxifene side chain extension plays a key role in the antiestrogenicity of both compounds. Interestingly, an aspartic acid residue, which lies at the base of ER␣ helix 3 (Asp-351), forms hydrogen bonds with a tertiary amine group in the tamoxifen and raloxifene extensions (16,26). Moreover, Asp-351 was later found to be mutated to tyrosine in an MCF-7 breast tumor cell variant whose growth was stimulated, rather than inhibited, by tamoxifen, and the D351Y mutant allowed increased tamoxifen and raloxifene agonist activity at ERE-responsive genes (28 -31). It was therefore proposed that Asp-351 plays an important role in securing the position of the tamoxifen and raloxifene side chain extensions and that the D351Y mutation allowed ER␣ to perceive tamoxifen and raloxifene as estrogens.
The molecular basis of the D351Y phenotype is not yet known. One possible explanation for the enhanced tamoxifen responses is that the D351Y mutant might allow AF-2 activity in the presence of both estrogens and antiestrogens (28,29), although this has not been directly confirmed. Another possible explanation is that the D351Y mutant might indirectly enhance AF-1 activity. In this study, we examine the role of ER␣ activation functions in the D351Y mutant phenotype. We demonstrate that the enhanced tamoxifen responses that are characteristic of this mutant stem from synergy between AF-1 and very weak (Ͻ1% maximal) tamoxifen-dependent AF-2 activity. We discuss possible structural explanations for this effect.
The expression vector for the Gal-GRIP1 NR box fusion protein was generated by PCR using pSG5-GRIP1 as a template. The upper strand oligonucleotide contained an EcoRI site and began at the nucleotide corresponding amino acid 618. The lower strand oligonucleotide contained a SalI site and began at the nucleotide corresponding to amino acid 778. The amplified product was digested with EcoRI and SalI and cloned in frame into the pM Gal4DBD expression vector (CLONTECH, Palo Alto, CA) which had been linearized with appropriate restriction enzymes.
The expression vector for the VP16-ER␣LBD fusion protein was generated from a mammalian expression vector for the ER␣ LBD amino acids 282-595 (HE14G) (37). The LBD coding sequences were excised as a BamHI fragment and cloned into the corresponding site in the pACT VP16 AD expression vector (CLONTECH, Palo Alto, CA). The resulting clone was then put into the appropriate reading frame by standard PCR-based point mutagenesis (Quickchange kit, Stratagene) to remove a single base between the VP16-AD and ER␣-LBD coding regions and thereby destroy the pACT SmaI site.
ER␣ D351Y mutations were introduced into the full-length ER␣, DBD-LBD region and VP16-LBD fusion protein by PCR-based point mutagenesis (Quickchange kit, Stratagene). In all cases, the resulting clones were subjected to sequence analysis.
Transfections-MDA-MB-453, HeLa, and MCF-7 cells were transfected by electroporation (5, 38). 2-3 million cells were trypsinized, resuspended in 0.5 ml of PBS supplemented with 10% glucose and 10 g/ml BioBrene (Applied Biosystems, Foster City, CA) in a single cuvette. Cells were electroporated at 0.24 kV, 960 microfarads in a Bio-Rad Gene Pulser apparatus (Bio-Rad). Generally, transfections included 2 g of luciferase reporter and 1 g of actin-␤-galactosidase internal control. Other expression vectors were included, as appropriate, at levels detailed in the figure legends. Following electroporation the cells were resuspended in growth medium, plated onto 12-well dishes, and treated with ligands (ICI 182,780 (0.1 M), raloxifene (0.1 M), tamoxifen (5 M), estradiol (10 nM), or diethylstilbestrol (DES) (10 nM)) or ethanolic vehicle. Cells were harvested 20 -24 h after transfection, except for assays performed with the AP-1-responsive reporter which were harvested 36 -40 h after transfection. The transfected cells were washed in cold PBS, lysed for 10 min at 4°C in 200 l of lysis buffer containing 100 mM Tris-HCl, pH 7.8, 0.1% Triton X-100. Luciferase and ␤-galactosidase activities were then measured using standard luciferase (Promega, Madison, WI) and ␤-galactosidase Galacto-Light assay systems (Tropix, Bedford, MA). Luciferase activities were normalized for ␤-galactosidase activities, although the variation in transfection efficiency was usually relatively small (Ͼ20%). Individual transfections (each containing data from triplicate wells) were repeated three times.
In Vitro Protein Binding Assays-To produce labeled proteins, 1 g of coupled transcription-translation vector was incubated with a coupled transcription-translation kit (Promega, Madison, WI) in the presence of [ 35 S]methionine. The GST-GRIP1 fusion proteins were expressed in Escherichia coli BL21. Bacteria were grown at 37°C, and 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added at A 600 ϭ 0.7, and cultures were then induced for 4 h. Bacteria were then pelleted, resuspended in IPAB-80 (20 mM Hepes, 80 mM KCl, 6 mM MgCl 2 , 10% glycerol, 1 mM dithiothreitol, 1 mM ATP, 0.2 mM phenylmethylsulfonyl fluoride, and protease inhibitors, pH 7.9), and then sonicated mildly. The lysate was cleared by centrifugation at 12,000 rpm for 1 h in an SS34 rotor. The cleared supernatant was then incubated for 2 h at 4°C with 500 l of glutathione-Sepharose 4B beads, which had been prewashed with 5 volumes of PBS, 0.2% Triton X-100 and equilibrated with 5 volumes of IPAB-80 at 4°C. The protein/bead mixture was then washed with 5 volumes of PBS, 0.05% Nonidet P-40 and resuspended in 1 ml of IPAB-80. Protein preparations were stored at Ϫ20°C.
For binding assays a volume of bead suspension containing 3 g of GST fusion protein was incubated at 4°C with 1-2 l of 35 S-labeled protein in IPAB-80 supplemented with 20 g/ml bovine serum albumin, either 5 M tamoxifen, 100 nM estradiol, or vehicle, in a total final volume of 150 l. After a 90-min incubation the beads were washed four times in IPAB-150 (identical to IPAB-80, but including 150 mM instead of 80 mM KCl). The beads were then resuspended in standard SDS-PAGE gel loading buffer. Bound proteins were then analyzed by SDS-PAGE and visualized by autoradiography, along with input protein controls.

ER␣ D351Y-dependent Tamoxifen Responses
Are Cell-specific-To investigate the contribution of ER activation functions to the D351Y phenotype, we first examined the behavior of ER␣ D351Y in cell types that naturally allow distinct levels of AF-1 activity. In MDA-MB-453 breast tumor cells, which allow relatively strong AF-1 activity (18), 2 ER␣ enhanced ERE-dependent transcription even in the absence of exogenous ligand, and addition of estradiol or the synthetic estrogen DES gave further stimulation (Fig. 1A). Whereas the antiestrogens all suppressed the ER␣-dependent constitutive activity, they did so to different degrees. Significant residual activity was retained in the presence of tamoxifen, modest activity in the presence of raloxifene, and none in the presence of the pure antiestrogen ICI 182,780 (hereafter, ICI). An ER␣ G400V mutant (34), which lacks constitutive activity, gave comparable levels of transcriptional activity to wild type ER␣ in the presence of each ligand.
In parallel, and in agreement with previous results (31), ER␣ D351Y showed no constitutive activity but did give estrogen activation that was comparable to wild type ER␣. Here, however, tamoxifen gave significantly higher transcriptional activity. Raloxifene also gave higher transcriptional activity in the presence of ER␣ D351Y, although these effects were weaker than tamoxifen. No ICI activation was detected. Thus, our results confirm that ER␣ D351Y allows enhanced tamoxifen and raloxifene agonist activity in breast cells (28 -31).
We then examined the behavior of the D351Y mutant in HeLa cells, which only allow low AF-1 activity (18, 23, 33, 39) (Fig. 1B). Once again, wild type ER␣ gave significant constitutive activity and further estradiol and DES response. Tamoxifen, however, only gave weak residual agonist activity, and the other antiestrogens showed none. ER␣ G400V and ER␣ D351Y again gave similar levels of estrogen activation to wild type ER␣, and ER␣ G400V again showed comparable levels of tamoxifen activation. While ER␣ D351Y did show slightly enhanced tamoxifen response, the overall level remained low. Thus, the extent of ER␣ D351Y-dependent tamoxifen activation is cell type-specific and correlates with AF-1 activity.
The D351Y Mutant Phenotype Requires AF-1 and Does Not Stem from Strong Tamoxifen-dependent AF-2 Activity-We next examined the effect of the ER␣ D351Y mutant upon ER␣ truncations that only contained isolated ER activation functions. An ER␣ truncation that only contained AF-1 (AB-DBD) elicited modest constitutive activity that was comparable to the overall level of tamoxifen activation with wild type ER ( Fig.  2A). An ER␣ truncation that only contained AF-2 (DBD-LBD) elicited a strong estrogen response and no tamoxifen response.
In parallel, ER␣ D351Y again showed comparable estrogen response and enhanced tamoxifen response. However, a DBD-LBD truncation that contained the D351Y mutation showed only modest estrogen response and no tamoxifen response. Thus, AF-1 is absolutely required for the D351Y phenotype and the D351Y mutation does not allow strong tamoxifen-dependent AF-2 activity in vivo.
We next examined the interactions of ER␣ D351Y with GRIP1, a representative p160, in vitro (Fig. 2B). As expected, wild type ER␣ bound strongly to a bacterially expressed GST-GRIP1 fusion protein overlapping the NR box region in the presence of estradiol but only showed weak residual binding in the presence of tamoxifen. In parallel, ER␣ D351Y showed reduced binding to GRIP1 in the presence of estradiol. Moreover, although tamoxifen-liganded ER␣ D351Y did show a slight, but reproducible, increase in binding to GRIP1 (taken up below), this still represented a very weak interaction. We therefore conclude that ER␣ D351Y lacks strong tamoxifen-dependent AF-2 activity and that, in fact, ER␣ D351Y behaves as a partial AF-2 mutant even in the presence of estradiol.
The D351Y Mutant Phenotype Is Not Related to Reduced AF-2 Activity-Given that ER␣ D351Y showed both enhanced tamoxifen response and reduced AF-2 activity, we confirmed that the enhanced tamoxifen response was not a general con- The experiment was carried out as described for Fig. 1A. B, interactions of ER␣ D351Y with GRIP1 in vitro. GST beads bound to GST alone or a GST-GRIP1 NR box region fusion protein (GRIP1, amino acids 563-1121) were incubated with labeled wild type or mutant ERs in the presence of ethanolic vehicle, tamoxifen, or estradiol. After washing, the products of the binding reaction were separated on SDS-PAGE gels and analyzed by autoradiography. sequence of reduced AF-2 activity. Fig. 3 shows that ER␣ D351Y again showed enhanced tamoxifen response relative to either wild type ER␣ or ER␣ G400V in HeLa cells. ERs bearing mutations in the AF-2 charge clamp (ER␣ K362A and ER␣ E542K), or the AF-2 hydrophobic cleft (ER␣ V376R) showed the expected reduction in estrogen activation but no increase in tamoxifen response. Thus, the ER␣ D351Y phenotype is specific to this mutant and not related to reduced AF-2 activity.
Each of the ER AF-2 mutants did show reduced constitutive activity. Coupled with previous observations of reduced constitutive activity in another AF-2 mutant (18), in ER␣G400V (34), in ERs bearing other mutations at position Asp-351 (31), and in ERs bearing mutations in the LBD dimerization interface, 3 our results suggest that this effect is not specific to the D351Y mutant and is probably a general consequence of disturbances of the ER␣-LBD surface.
Tamoxifen-dependent Interaction of ER␣ D351Y with LXXLL Motifs-Although the D351Y mutant did not allow strong tamoxifendependent AF-2 activity, it was noteworthy that tamoxifen activation in the presence of the ER␣ D351Y mutant exceeded the level of constitutive activation in the presence of isolated AF-1 and that tamoxifen-liganded ER␣ D351Y showed a slight increase in weak residual interactions with the GRIP1 NR box region (see Fig. 2). Together, these results pointed to an active role for the ER␣-LBD in the D351Y mutant phenotype and suggested that the D351Y mutant might allow very weak tamoxifen-dependent AF-2 activity. We therefore examined ER␣ D351Y binding to GRIP1 in mammalian two-hybrid assays, which are sensitive enough to detect relatively weak interactions. Fig. 4 shows a Gal fusion protein containing the GRIP1 NR box region efficiently recruited an VP16-ER␣LBD fusion protein to the promoter in the presence of estradiol. Moreover, the GRIP1 NR box region completely failed to recruit the VP16-LBD fusion in the presence of tamoxifen (Fig. 4, inset), even in the presence of high levels of the VP16-LBD fusion protein. In parallel, the GRIP1 NR box region also recruited a similar VP16-LBD D351Y mutant fusion protein in the presence of estradiol. The overall level of estrogen-dependent recruitment ranged from 10 to 20% of wild type ER␣-LBD, consistent with the notion that D351Y behaves as a partial AF-2 mutant. More importantly, the GRIP1 NR box region now also recruited the mutated VP16-LBD fusion protein weakly in the presence of tamoxifen (up to 2-3 times over background, see Fig. 4, inset). This suggests that the D351Y mutant permits low levels of tamoxifen-dependent AF-2 activity. We estimate that this level of AF-2 activity is less than 1% of wild type ER␣ in the presence of estradiol.
We then asked whether the D351Y mutant might also affect ER␣ interactions with other types of coactivators. The LXXLL motifs of different ER␣ coactivators fall into one of three homology groups, which differ according to their receptor interaction preferences (11,36,41). Class I includes p160s, and class II includes TRAP220, and class III includes PGC-1. In accordance with previous results (11), the VP16-LBD fusion protein interacted strongly with idealized class I (D2) and class III peptides (Phe-6) and more weakly with a class II peptide (D47/ F6) in the presence of estradiol but not tamoxifen (Fig. 5). In each case, the D351Y mutant reduced estrogen-dependent ER␣ interactions with each peptide and allowed, at best, weak tamoxifendependent interactions (Fig. 5, inset). Thus, the D351Y mutant does not alter the overall spectrum of ER␣/coactivator recognition, but rather acts as a generalized partial AF-2 mutant that allows very weak interactions with LXXLL motifs in the presence of tamoxifen.

GRIP1 Enhances AF-2 Activity in the Presence of Estrogens and Antiestrogens in the Context of the ER␣ D351Y Mutant-To
confirm that the weak tamoxifen-dependent ER␣-LBD/NR box interactions played a role in the D351Y phenotype, we examined the effects of GRIP1 overexpression upon isolated ER␣ AF-2. Fig. 6 shows that wild type GRIP1 enhanced the overall level of estrogen response with the isolated ER␣ DBD-LBD region but gave no tamoxifen or raloxifene response. In parallel, a GRIP1 NR box mutant showed markedly reduced potentiation of estrogen activation. This is consistent with the notion that GRIP1 potentiation of AF-2 activity requires intact NR boxes (35).
Wild type GRIP1 also enhanced the overall level of estrogendependent AF-2 activity in the presence of ER␣ DBD-LBD D351Y mutant. Here, however, significant tamoxifen and raloxifene activation were also detected. This suggests that antiestrogen-dependent AF-2 activity is obtained in the presence of overexpressed GRIP1. Moreover, the GRIP1 NR box mutant failed to potentiate the ER␣ D351Y-dependent antiestrogen responses. This suggests that the antiestrogen-depend-3 P. Webb, unpublished data . ent AF-2 activity requires ER␣ interactions with the GRIP1 NR boxes.
We then confirmed that antiestrogen-dependent AF-2 activity played a role in the context of the full-length ER␣ D351Y mutant. Fig. 7 shows that GRIP1 enhanced the overall levels of estrogen and tamoxifen response in the presence of wild type ER␣. In parallel, the GRIP1 NR box mutant enhanced tamoxifen response even better than wild type GRIP1. This is consistent with the notion that GRIP1 enhances tamoxifen responses at classical EREs by boosting AF-1 activity (18). Wild type GRIP1 also enhanced the overall levels of estrogen re-sponse in the presence of ER␣ D351Y and also gave very potent enhancement of both tamoxifen and raloxifene response. In parallel, the GRIP1 NR box mutant gave significantly weaker enhancement of both the estrogen responses and the tamoxifen and raloxifene responses, consistent with the notion that AF-2/NR box interactions play a role in the D351Y phenotype. Taken together, our results suggest that the ER␣ D351Y mutant allows AF-2 activity in the presence of antiestrogens and that this AF-2 activity requires GRIP1 NR boxes. Our results also suggest that the strong antiestrogen responses that are characteristic of the D351Y mutant require both this weak AF-2 activity and either strong AF-1 activity, or p160 overexpression.
D351Y Suppresses Antiestrogen Effects at AP-1 Sites-We have previously demonstrated that ER␣ also enhances gene expression by enhancing AP-1 activity, via protein-protein interactions, and that this effect is stimulated both by estrogens and antiestrogens (5,6,32). Antiestrogen activation occurs through a pathway that is suppressed by the ER␣ activation functions. Fig. 8 shows that an ER␣ truncation that lacks AF-1 (DBD-LBD) elicited strong ICI and raloxifene responses, and more modest tamoxifen responses, from an AP-1-responsive reporter gene in MCF-7 breast cells. A similar truncation containing the D351Y mutation retained the strong ICI responses but completely lacked raloxifene and tamoxifen responses. Because active AF-2 suppresses antiestrogen action at AP-1 sites, this result supports the notion that the ER␣ D351Y mutant allows AF-2 activity in the presence of tamoxifen and raloxifene but not ICI.

DISCUSSION
Antiestrogens work by interacting with the ER ligand-binding pocket, thereby competing with endogenous estrogens and inactivating AF-2 (10,16,22,26,27,42). A key determinant of antiestrogenicity is the presence of a bulky side chain extension on the antiestrogen molecule. Recent crystal structures of the ER␣-LBD complexed with either tamoxifen or raloxifene (16,26), and the ER␤-LBD complexed with raloxifene (27), have revealed that the extension protrudes through the ER surface and displaces helix 12, which occludes the coactivator binding surface. The ER␣ crystal structures have also revealed that Asp-351 formed hydrogen bonds with the tamoxifen and raloxifene extension, leading to the suggestion that Asp-351 might play a key role in the behavior of antiestrogens. The discovery that a tamoxifen-stimulated MCF-7 variant cell line contained a mutated ER␣ with a tyrosine substitution at residue 351 strengthened this view and suggested that it was important to understand the role of this residue in antiestrogen action (28 -31).
In this study, we have examined the contribution of ER␣ activation functions to the D351Y mutant phenotype. It is well established that tamoxifen activation at classical ERE-responsive reporters stems from AF-1 activity (17). We found that the strength of ER␣ D351Y-dependent tamoxifen activation correlates with the strength of AF-1 in different cells and that tamoxifen activation could not be observed with ER␣ D351Y truncations that lack AF-1. These results indicate that AF-1 also plays an important role ER␣ D351Y-dependent tamoxifen responses.Moreover,ER␣D351Ydoesnotallowstrongtamoxifendependent AF-2 activity in vivo, and we and others (31) also found that ER␣ D351Y does not allow strong tamoxifen-dependent interactions between ER␣ and GRIP1 in vitro. Thus, the D351Y mutation does not allow ER␣ to perceive tamoxifen exactly as wild type ER␣ perceives estrogen. Nonetheless, several lines of evidence did point to an unusual role for AF-2 in the D351Y phenotype. First, ER␣ D351Y gives higher levels of tamoxifen activation than would be expected from AF-1 alone. Second, ERa D351Y shows very slightly increased binding to the GRIP1 NR box region in the presence of tamoxifen in vitro. Third, tamoxifen does not allow interactions between the wild type ER␣-LBD and the GRIP1 NR box region in mammalian two-hybrid assays but does allow interactions between the equivalent ER␣ D351Y mutant and the GRIP1 NR boxes. Fourth, weak tamoxifen-dependent AF-2 activity occurs in the presence of ER␣ D351Y and overexpressed GRIP1. Fifth, ER␣ D351Y-dependent tamoxifen responses show dependence upon the GRIP1 NR boxes, which bind AF-2. Finally, tamoxifen and raloxifene effects at AP-1 sites, which are suppressed by AF-2, are both abolished by the D351Y mutant. Our results therefore support the previous suggestion of Jordan and co-workers (28,29) that the D351Y mutant allows AF-2 activity in the presence of antiestrogens. However, our results also suggest that this antiestrogen-dependent AF-2 activity is relatively weak and is only detectable when AF-1 is strong, and AF-1 and AF-2 synergize, or when p160s are overexpressed. It is therefore not surprising that the D351Y mutant phenotype should be especially prominent in breast cells, which show relatively strong AF-1 activity and also contain elevated levels of AIB1 protein, one of the p160 coactivators (43).
We stress that the actual levels of ER␣ D351Y tamoxifen-dependent AF-2 activity are very small (Ͻ1% wild type). Although it may seem paradoxical that such low levels of AF-2 activity are sufficient for strong tamoxifen-dependent transcriptional activation, similar behaviors have been noted before. AF-1 often completely masks the phenotype of partial AF-2 mutants in the context of full-length ER␣ (44) and even partially masks the phenotype of some strong AF-2 mutants (11). Indeed, the fact that the D351Y mutant shows reduced binding to GRIP1 in vitro, but allows normal levels of estrogen response in vivo, illustrates this principle. It is also well established that GRIP1 overexpression suppresses the phenotypes of many partial ER␣ AF-2 mutants and can even partially mask the phenotype of strong AF-2 mutants (15,45).
It will be interesting to ask exactly how the D351Y mutant allows AF-2 activity in the presence of tamoxifen. Presently, we favor the idea that the Asp-351 residue helps secure the position of the tamoxifen side chain extension and that this, in turn, helps secure helix 12 in the inactive position. We speculate that the D351Y mutant allows increased mobility of the tamoxifen extension and that this, in turn, allows increased mobility of helix 12. In principle, the D351Y mutation might allow helix 12 of the antiestrogen-liganded ER␣ D351Y mutant to adopt a position that resembles the estrogen-liganded ER␣, at least for some of the time. Alternatively, the D351Y mutation might lead to complete displacement of helix 12 in the presence of antiestrogens, and thereby promote inefficient interactions between the helix 3,4,5 region of the hydrophobic cleft and the p160 NR box. It is even conceivable that the substituent tyrosine residue itself could make novel stabilizing contacts with p160s in either of these configurations.
We, and others (31), have also shown that ER␣ D351Y shows markedly reduced constitutive activity. This is a common phenotype that is also observed in ER␣ G400V (34), in ER␣ AF-2 mutants (Fig. 3 and Ref. 18), in other ER␣ Asp-351 mutants (31) and in ERs bearing mutations in the LBD dimerization interface. 3 It is known that the ER␣ G400V phenotype stems from increased association with inhibitory heat shock proteins (46), and because heat shock proteins bind solvent exposed hydrophobic regions, it is likely that lack of constitutive activity is indicative of exposure of hydrophobic residues upon the ER␣-LBD surface. Nonetheless, ER␣ D351Y is well expressed, with normal affinity for estradiol and antiestrogens, suggesting that its overall conformation is relatively normal (28,29,31). We therefore speculate that an altered position of helix 12 could expose the AF-2 hydrophobic cleft and target the D351Y mutant to the heat shock complex in the absence of hormone.
Finally, our studies also allow us to draw some conclusions about the behavior of wild type ER␣. First, we have been unable to detect any association between the tamoxifen-liganded wild type ER with LXXLL motifs whatsoever. We have also confirmed that the overall level of tamoxifen response correlates with AF-1 activity. Thus, our results agree with the idea that tamoxifen agonist activity stems AF-1 and not from weak residual AF-2 activity. Second, we have confirmed that the ER␣ D351Y mutant allows increased agonist activity in the pres- FIG. 8. Effect of D351Y on ER␣ action at AP-1 sites. Transcriptional activity of the AP-1-responsive collagenase promoter-driven reporter (2 g) was determined in MCF-7 breast cells in the presence of 5 g of empty pSG5 vector or equivalent expression vectors for the ER␣ DBD-LBD region or its D351Y mutant equivalent. ence of tamoxifen and raloxifene but not ICI (28,29,31). This suggests that tamoxifen and ICI must work by different mechanisms. Third, ER␣ D351Y also behaved as a partial AF-2 mutant. Our unpublished studies 4 also reveal that the equivalent ER␤ mutant (D303Y) also behaves as a partial AF-2 mutant and, given that a vitamin D receptor bearing a mutation in the equivalent residue also behaves as a partial AF-2 mutant (47), we suggest that the same residue could help stabilize helix 12 positioning in many nuclear receptors. Finally, our results also address the mechanism of estrogen-dependent regulation of cell division. It is known that ER␣ D351Y allows tamoxifen to mimic the stimulatory effects of estrogens upon MCF-7 cells and also the inhibitory effects of estrogens upon MDA-MB-231 cells (28,29). Since ER␣ D351Y works by promoting tamoxifen-dependent association of ER␣ AF-2 with p160s, then we can conclude that ER␣ interactions with p160s must be the key step in both ER␣-dependent stimulation of cell growth and ER␣-dependent inhibition of cell growth. It will be interesting to ask how the same ER␣/coactivator interactions lead to opposite growth responses in closely related cell types.