The nuclear receptor (NR) family comprises 48 structurally related transcription factors, many of which require their cognate ligand for activity [13]. The NRs regulate transcription by binding to response elements in the promoters of target genes and acting as scaffolds for the assembly of large coactivator and corepressor complexes [4]. NR coactivators include the p160s (including GRIP1/TIF-2, SRC-1 and AIB1/ACTR/pCIP). The p160s enhance transcription by binding histone acetyl-transferases such as p300/CBP and pCAF [5]and methyl-transferases such as CARM1 and PRMT [6] which, in turn, enhance transcription by modification of chromatin. Other NR coactivators include TRAP220[7], which is part of a larger complex (TRAP/DRIP/SMCC/mediator) that contacts the basal transcription machinery and PGC-1[8], a cold inducible coactivator that binds CBP and SRC-1 and proteins involved in RNA processing [9]. NR corepressors include NR corepressor (N-CoR) and silencing mediator of retinoid and thyroid responsive transcription (SMRT) [4]. Both N-CoR and SMRT repress transcription, at least in part, by binding to histone de-acetylases (HDACs) either directly or indirectly through other corepressor complex components. Other known NR corepressors include RIP140 [10], Hairless [11], short heterodimer partner (SHP) [12] and DAX [13], and receptor specific corepressors such as the estrogen receptor (ER) interacting proteins REA and HET-SAFB [14, 15].

Generally, NR transcriptional activity is dictated by the balance between coactivator and corepressor recruitment, and one of the most important factors that influences this balance is the absence or presence of agonist ligands (reviewed in [4]). Unliganded NRs such as thyroid (TRs) and retinoid receptors (RARs) bind corepressors, and ligand promotes release of corepressor and subsequent binding of coactivators. The mechanism of this coregulator exchange is well understood. NRs consist of three domains, the N-terminal domain (which contains a context-specific activation function AF-1), the central DNA binding domain (DBD) and the C-terminal ligand binding domain (LBD), which contains a hormone-dependent activation function, AF-2. The unliganded LBD recognizes hydrophobic motifs, termed interaction domains (IDs), which are reiterated three times in N-CoR and twice in SMRT and conform to the consensus L/IXXIIXXXL [4](see also[16]). By contrast, the liganded LBD binds shorter hydrophobic motifs termed NR boxes that are reiterated several times within each coactivator and conform to the consensus LXXLL. The LBD utilizes a large hydrophobic cleft composed of residues along H3 and H5 to bind IDs [17], and a smaller hydrophobic cleft that is composed of residues in the upper part of H3 and H5 and H12 (and corresponds to AF-2) to bind NR boxes [18]. Thus, agonists promote coregulator exchange by promoting the packing of H12 over the lower part of the ID binding region, an event that simultaneously completes the coactivator binding surface. In other cases, however, the balance of coactivator and corepressor recruitment is regulated by direct competition for the AF-2 surface, rather than ligand-dependent coregulator exchange. RIP140, Hairless and DAX possess NR boxes that interact with AF-2 [11, 13, 19] and these corepressors act as negative regulators of the activity of the liganded NR.

The NR family contains two related ERs (ERα and ERβ) that conform to the typical three domain NR structure and share extensive sequence homology in the DBD and LBD region[20, 21]. Analysis of the function of the individual ERs in mouse knockout models suggests that the major proliferative effects of estrogen are mediated by ERα and not by ERβ, which seems to play an inhibitory role in proliferation in some studies[22, 23]. The ligand-binding properties of the ERs are different, with ERβ often exhibiting stronger binding to plant-derived phytoestrogens [24]. More importantly, the ERs exhibit isoform-specific effects on gene expression. Both ERs enhance transcription from genes with classical estrogen response elements (EREs), but ERα requires less ligand to obtain maximal activation than ERβ [25, 26]. Likewise, both ERs suppress the activity of the TNFα promoter in response to estrogens, but ERβ is a more potent repressor than ERα [27]. However, some of the most striking isoform-specific differences in gene regulation are observed at promoters, such as that of cyclin D1, which contain AP-1 sites or related cyclic AMP response elements (CREs). ERα enhances AP-1 activity in response to estrogens, [28, 29] but ERβ inhibits AP-1 activity in response to estrogens [2931]. ERβ also completely suppresses ERα activity at the cyclin D1 promoter and even suppresses the activity of an ERα mutant that is selectively superactive at AP-1 sites and CREs [29]. Finally, ERβ shows a unique capacity to enhance AP-1 activity in response to selective estrogen receptor modulators (SERMs) such as raloxifene, tamoxifen and ICI 182,780/Faslodex (ICI) [3032]. Together, these observations suggest that there are fundamental differences in the way that the ERs bind unspecified cofactors that modulate gene expression, and that some of these cofactors must play a role in differential ER activity at AP-1 sites.

Although the poorly conserved NTD region clearly plays an important role in isoform-specificity [32, 33], it is also likely that there are differences in the better conserved LBD region that contribute to differential ERα and ERβ activities. Phage display techniques have revealed that ERα and ERβ show different preferences in LXXLL binding (reviewed in [34, 35]). Moreover, some cofactors that contain LXXLL motifs show differential binding to LBDs of the ER isoforms (reviewed in [36]). SHP binds ERα preferentially [37], and represses ERα activity more strongly than that of ERβ. The PGC-1 related protein PERC also binds ERα preferentially, and potentiates ERα activity more strongly than that of ERβ [38]. We recently observed that ERα binds the C-terminal NR interacting regions of N-CoR and SMRT in the presence of SERMs but not estrogens [39]. In this study, we report that ERβ interactions with N-CoR and SMRT are promoted by agonists and inhibited by SERMs. Thus, the ERs show completely opposite ligand preferences of interaction with corepressors. We discuss the potential significance of these different modes of ER interaction with N-CoR in terms of known isoform-specific behaviors.


Agonist Dependent ERβ Interactions with N-CoR and SMRT

To investigate ERβ interactions with corepressors, we examined the interactions of full length ERβ (amino acids 1–530) with bacterially expressed C-terminal NR interacting domain of N-CoR (amino acids 1944–2453) in vitro (Fig. 1A). Fig. 1B reveals, surprisingly, that ERβ bound N-CoR in the absence of hormone and in the presence of agonist ligands (E2, DES) and phytoestrogens (genistein, coumestrol). Moreover, these interactions were suppressed by SERMs (ICI, raloxifene and tamoxifen). ERβ bound to the coactivator GRIP1 more strongly than N-CoR, but did so with an identical ligand preference. Similar agonist-dependent interactions could be observed between ERβ and the alternate NR corepressor SMRT in vitro (Fig. 1C). Control binding experiments performed in parallel confirmed that ERα bound to N-CoR in the presence of SERMs, and not estradiol (Fig. 1D) and that TRβ bound N-CoR in the absence of hormone, and was released in the presence of T3, whereas TRβ only bound GRIP1 in the presence of T3 (Fig. 1E).

Figure 1
figure 1

ERβ Binds N-CoR in the Presence of Agonists. (A) Schematic of N-CoR primary structure. Silencing domains are indicated with hatched bars. NR binding regions (ID motifs) are illustrated with solid bars. (B) Radiolabeled ERβ retained by GST-N-CoR (amino acids 1944–2453) or GST-GRIP1 (amino acids 563–1121) after separation by SDS-PAGE. The amounts of bound proteins are compared to 10% of the input protein used in the binding assay in this experiment, and all results shown in the paper. (C) Radiolabeled ERβ retained by GST SMRT (amino acids 987–1491). (D) Radiolabeled ERα retained on GST-N-CoR beads in parallel. (E) Radiolabeled TRβ retained on GST-N-CoR and GST-GRIP1 beads in parallel.

To examine interactions between ERβ and N-CoR in mammalian cells we performed two-hybrid assays using a GAL4 DBD/N-CoR C-terminus fusion protein as bait and a VP16-ERβ LBD fusion as the prey. Fig. 2 shows that the ERβ-LBD bound N-CoR in the presence of agonists and phytoestrogens, but not SERMs. Control two-hybrid assays confirmed that a VP16-TRβ LBD fusion protein bound N-CoR in the absence of hormone, but not in the presence of T3. E2 dependent binding of ERβ to N-CoR was dose dependent (Fig. 2B) with an EC50 (0.3 nM) that resembled that of ERβ binding to the GRIP1 NR box region (amino acids 610–770). Thus, ERβ binds the N-CoR C-terminal NR interacting region in the presence of agonists, but not SERMs, and does so in vitro and in mammalian cells. Moreover, results from the two-hybrid assay indicate that the ERβ LBD is sufficient to obtain estrogen-dependent interactions with N-CoR.

Figure 2
figure 2

ERβ Binds N-CoR in the Presence of Agonists in Mammalian Cells. (A) Two-hybrid assays. Components of the two-hybrid assay are shown in schematic at top. Results of a representative assay are shown below (with error bars representing the standard error of values determined from three separate wells). Ligand concentrations were: ICI, raloxifene, Genistein, Coumestrol; 1 uM, Tamoxifen; 5 uM, estradiol DES 100 nM. (B) Estradiol dependence of ERβ interactions with N-CoR and GRIP1 fusion proteins in mammalian cells. A representative experiment is shown. Error bars represent standard deviations from four wells.

ERβ Interactions with N-CoR are Dependent on AF-2 and require H12

Unliganded NRs usually bind ID motifs (consensus L/IXXIIXXXL) in the N-CoR C-terminus. To ask whether ERβ might bind these motifs in the presence of estradiol, we examined the ability of peptides containing known NR interacting motifs to compete for the interaction (Fig. 3A). A peptide overlapping to the N-CoR ID1 motif (amino acids 2265–2291) that competes for N-CoR binding to unliganded TR and RAR [16]failed to compete for agonist-dependent ERβ interactions with N-CoR. By contrast, a peptide corresponding to GRIP1 NR box 2 did compete for this interaction [40, 41]. This finding suggests that agonist-bound ERβ does not recognize ID motifs, and that ERβ interactions with N-CoR more closely resemble those with GRIP1.

Figure 3
figure 3

ERβ AF-2 is required for Corepressor and Coactivator Binding. (A) Peptide competition for ERβ binding to cofactors. Binding of ERβ to GST-N-CoR was assessed in the presence of 10μg of designated competitor peptide. (B) Binding of VP16-ERβ LBD mutants to Gal-N-CoR and Gal-GRIP1 was analyzed. Luciferase activity obtained with Gal-fusion and VP16-ERβ in the presence of estradiol (100 nM) over four experiments was corrected for background and set to 100%. Activity obtained with ERβ mutants was compared to this value. Standard errors are shown.

NR interactions with N-CoR are usually mediated by a hydrophobic cleft that spans residues from H3 and H5 and includes residues that lie under H12 in the liganded configuration [4, 17]. These interactions are either independent of, or inhibited by, NR H12 [17, 39, 42]. By contrast, NR interactions with coactivators are mediated by residues from the upper part of H3–H5 and also require H12 itself [18]. Fig. 3B shows that a mutation in a conserved residue on H12 that is required for coactivator binding (E493K) abolished the interaction of ERβ with both N-CoR and GRIP1. Moreover, other mutations in the upper part of the H3–H5 region that comprises the AF-2 surface (D303Y, I310R and K314A on H3; V328R and L331R on H5) abolished ERβ interaction with both cofactors. Control mutations in other regions of the ERβ surface left its interactions with N-CoR and GRIP1 either slightly reduced or intact (these are L301R at the base of H3, V361R in the S-bends, M379R in H8; L426R, T434R in the H10 and Y488S in the H11–H12 loop). Thus, ERβ interactions with N-CoR are dependent on the AF-2 surface (including H12) and, in this regard, resemble those of ERβ and GRIP1.

ERβ Binds an NR Box-Like Motif in the N-CoR C-terminus

To map the region of N-CoR that interacted with ERβ, we examined ERβ binding to a series of rationally designed smaller fragments of the N-CoR C-terminus (Fig. 4). ERβ did not bind two of these smaller fragments of N-CoR (1944–2033; 2230–2322) that contain known ID motifs (IDs 3 and 1 [16]). ERβ bound weakly to two regions of N-CoR (2033–2123; 2123–2230), one of which (2033–2123) contains an ID motif (ID2), but did so in a ligand-independent fashion. However, ERβ did bind to a fragment that spanned the extreme C-terminus (2322–2453) and did so in a manner that was promoted by E2 and suppressed by ICI, much like the interactions of ERβ with the entire N-CoR nuclear receptor interacting region.

Figure 4
figure 4

ERβ binds the N-CoR C-terminus. Binding of ERβ to short bacterially produced GST-N-CoR fragments was assessed as in Fig. 1B. The approximate positions of N-CoR IDs are indicated with black bars, the putative ERβ interacting motif with a grey bar. Note that all of the binding experiments with shorter N-CoR fragments were performed in parallel, and are compared to the same input ERβ protein (10% of total) displayed alongside the 1944–2033 fragment. A binding experiment that was performed in parallel utilizing the entire GST-N-CoR C-terminus (1944–2453) is shown for comparison at top. The likely reasons that ERβ binding to the shorter N-CoR fragment (2322–2453) is stronger than that obtained with the intact N-CoR C-terminus are discussed in the text.

The interaction of ERβ with the small N-CoR C-terminal fragment (amino acids 2322–2453) was stronger than that observed with the intact C-terminus (amino acids 1944–2453). This apparently improved binding is likely to be a consequence of our methodology (West et al. Unpublished data). In general, expression of large fragments of the N-CoR C-terminus in E. Coli yields a mix of full length protein along with truncated products. To create the expression vectors for the smaller fragments, truncated N-CoR polypeptides that were obtained in E. Coli extracts were subjected to protein sequence analysis and cDNA fragments that coded for the major truncated products were prepared. Each of the resulting polypeptides was expressed very efficiently in E. Coli. The end product that was obtained after GST purification essentially consisted of a single short polypeptide as judged by Coomassie stain. Binding of ERβ to N-CoR (2322–2453) is probably very efficient for two reasons. First, equal amounts of GST fusion protein were used as baits for the translated ERβ protein in this series of experiments (3μg per assay). Thus, N-CoR (2322–2453) is present in molar excess over N-CoR (1944–2453). Second, as developed above, preparations of N-CoR (1944–2453) generally contain truncated products, so sequences corresponding to the extreme N-CoR C-terminus (which binds ERβ) is markedly under-represented. In any case, the fact that ERβ binds weakly or not at all to the three N-CoR ID motifs that mediate interactions with TRs and RARs and, instead, binds in an agonist-dependent fashion to a region in the C-terminus of N-CoR that has not previously been implicated in NR interactions indicates that ERβ recognizes a novel protein sequence motif within N-CoR.

The N-CoR C-terminus contains the sequence PLTIRMLS (β-box, amino acids 2399–2406; Fig. 5). This sequence does not exactly conform to the LXXLL consensus, but contains features (underlined) that resemble the ERβ H12 region (LLLEML), and artificial ERβ interacting LXXLL peptides (293, PNLISLLS; D47, PLLLSLLS), both of which bind to the ERβ AF-2 surface [4347]. Moreover, the presence of a proline residue amino-terminal to the hydrophobic groups is typical of so-called class II LXXLL motifs which are found in ERβ interacting cofactors such as TRAP220 and RIP140[45]. Finally, the unusual C-terminal hydrophobic pair (ML) has been observed in ERα and ERβ H12[43, 44, 48], and in RIP140 NR boxes [19].

Figure 5
figure 5

Sequence comparison of putative NR box-like motif (β-box) with known ER AF-2 interacting peptides. Homologies between the β-box and ERβ H12 or the artificial ERβ interacting peptide 293 are indicated with thick (homologous) or thin (conserved nature of side chain) lines.

We investigated the significance of the β-box in ERβ interactions with N-CoR. As Fig. 6A shows, a synthetic β-box peptide competed for binding to N-CoR, albeit somewhat less efficiently than native GRIP1 NR box 2. Similar results were obtained in competition experiments that used GST-GRIP1 instead of GST-N-CoR (data not shown). The isolated β-box also acted as bait for a VP16-ERβ fusion protein in mammalian cells, and did so with similar efficiency to other known ERβ interacting peptides (Fig. 6B). Finally, mutations within the β-box (especially M2405A, L2406A) disrupted ERβ interactions with N-CoR in mammalian two-hybrid assays, but did not affect TRβ interactions (Fig. 6C). Thus, the β-box is sufficient to bind ERβ and is essential for agonist-dependent ERβ interactions with the N-CoR C-terminus.

Figure 6
figure 6

The N-CoR β-box binds ERβ. (A) Peptide competitions. Binding of ERβ to either bacterially expressed GST-N-CoR (amino acids 2239–2453) was determined in the presence of increasing doses of competitor peptide (amounts shown at top refer to μg of peptide, plus signs refer to the presence of estradiol in the assay). (B) Interaction of ERβ with different peptides in mammalian two-hybrid assays. A representative experiment is shown. Errors are derived from three separate wells. (C) Analysis of ERβ interaction with mutant Gal-N-CoR fusion proteins. Values were determined as in Fig. 3.

Next, we examined whether the β-box would bind other NRs. The Gal-β-box fusion failed to recruit the ERα, TRβ or RARβ LBDs in mammalian two-hybrid assays (Fig. 7A). Moreover, while the β-box and GRIP1 NR box 2 peptides both competed for ERβ interactions with GRIP1, only the NR box 2 peptide competed for ERα interactions with GRIP1 (Fig. 7B). Thus, the N-CoR β-box is, at least to some degree, ERβ specific. Mutation of N-CoR to obtain a β-box sequence that more closely resembled a conventional LXXLL motif (T2402L) led to enhanced hormone-dependent interactions with ERβ and permitted novel hormone-dependent interactions with ERα (Fig. 7C). Thus, some of the observed ERβ specificity is probably a consequence of an unexpected ability to tolerate the absence of a leucine residue at the N-terminus of the LXXLL motif. Together, our results indicate that ERβ has the potential to utilize its AF-2 surface to bind NR boxes within coactivators or an NR box-like sequence in the C-terminus of N-CoR.

Figure 7
figure 7

The β-box is an ERβ-specific NR box. (A) Interactions of VP16-NR fusions with the Gal-β-box fusion or a GAL-GRIP1 (amino acids 610–770) control were determined in HeLa cell transfections. A representative experiment is shown. (B) Peptide competitions for ERα and ERβ binding to GST-GRIP1. Experiments utilized 10μg of competitor (C) Mutation of the β-box to resemble a consensus LXXLL motif. The panel shows the results of a mammalian two-hybrid assay in which binding of Gal-N-CoR or Gal-N-CoR T2402L to VP16-ERβ or VP16-ERα was assayed.

A HDAC Repressor Enhances ERβ Activity

Since ERβ bound N-CoR and SMRT in the presence of estrogens, we investigated the possible involvement of corepressors in the actions of agonist-bound ERβ in vivo. To perform this experiment, we examined the effect of the HDAC inhibitor trichostatin A (TSA) on ERβ activity in transiently transfected HeLa cells. Fig. 8A confirms that ERα shows stronger transcriptional activity than ERβ at a simple ERE responsive reporter gene. TSA enhanced the basal activity of the ERE-TK reporter gene by about fifteen-fold in the absence of ER (see inset). However, TSA also equalized the relative transcriptional activity of both ERs. Fig. 8B shows that the isolated ERα LBD exhibited more potent transcriptional activity than the ERβ LBD. However, both LBDs showed similar transcriptional activity in the presence of TSA. Thus, corepressor complex HDACs must play an unspecified role in restricting the transcriptional activity of both ERβ and, in particular, the ERβ-LBD. This is consistent with the notion that corepressors restrict the activity of agonist-bound ERβ-LBD.

Figure 8
figure 8

TSA Enhances ERβ-LBD activity. (A) Results of a HeLa cell transfection in which ERα and ERβ activity was compared at a simple ERE responsive reporter gene. The value obtained in the absence of ER and ligand was set at 1 (irrespective of the presence or absence of TSA, 100 nM). Other values were calculated relative to this value. A single representative experiment is shown with errors derived from three wells. (B) Results of transfection in which ERα and ERβ LBD activities were compared in the absence (left) or presence (right) of TSA (100 nM); note differences in scale. A single representative experiment is shown. Standard errors are derived from three wells.


NRs generally interact with the corepressors N-CoR and SMRT either in the absence of ligand, or in the presence of receptor antagonists, and agonists promote corepressor release [4]. In this study, we demonstrated that ERβ binds to N-CoR in the presence of ER agonists such as estradiol and DES and the phytoestrogens genistein and coumestrol, but not in the presence of SERMs. Moreover, this interaction is dependent upon ERβ AF-2, including H12, and is competed by NR box peptides but not ID peptides. The hormone-dependent component of the ERβ /N-CoR interaction maps to the extreme C-terminus of N-CoR, which has not been previously implicated in NR interactions, and requires a sequence that resembles an ERβ-specific NR box (PLTIRMLS, β-box). In this regard, ERβ differs from ERα, which probably binds ID motifs in a SERM-dependent fashion [49, 50] and shows reduced binding to N-CoR in the presence of estradiol [39]. ERβ also differs from many other NRs, which either bind N-CoR in the absence of ligand and are released in the presence of ligand or interact with N-CoR in the presence of antagonists but not agonists [4].

The fact that the mode of ERβ interaction with N-CoR resembles that of NRs with coactivators [4], or with corepressors that modulate the activity of liganded NR complexes, such as RIP140 [10], raises the possibility that ERβ may be able to recruit N-CoR and SMRT to estrogen-regulated promoters in response to agonists and that the balance of overall ERβ activity in the presence of estrogens may be regulated by competition between p160s and corepressors for the same ERβ AF-2 surface. We recognize that our studies do not directly address this issue. Our attempts to identify ERβ mutants that differentiate between GRIP1 and N-CoR binding to analyze the role of agonist-dependent corepressor binding have not yet been successful (probably because ERβ utilizes the same surface to bind both cofactors). Moreover, transfection of N-CoR or various mutated N-CoR derivatives did not significantly affect ERβ activity at EREs or AP-1 sites (data not shown). We do not understand why, but in our hands, transfected N-CoR also fails to affect TR or ERα activity (data not shown), despite the fact that it clearly interacts with both NRs. Nevertheless, we suspect that estrogen-dependent N-CoR binding may represent an important component of the regulation of ERβ activity. As described in the Introduction, ERα and ERβ must interact differentially with factors that modulate ER activity in the presence of estrogens. The finding that estrogens suppress N-CoR binding to ERα [39], but promote N-CoR binding to ERβ represents the first demonstration of a corepressor that shows completely distinct modes of hormone-dependent interaction with the ER isoforms. Thus, N-CoR and SMRT and their associated HDACs are excellent candidates to explain some of the differential behaviors of the ER isoforms. Consistent with this notion, the apparent weak transcriptional activity of the ERβ LBD is a consequence of corepressor HDAC activity at some level (Fig. 8). Full verification of the importance of ERβ interaction with N-CoR will await demonstration that ERβ recruits N-CoR and SMRT to estrogen-regulated promoters in vivo, and that this event is related to modulation of estrogen response.

While the ER isoforms have contrasting effects on AP-1 activity in the presence of estrogens, ERα truncations that lack the NTD and ERβ both enhance AP-1 activity in the presence of SERMs [30, 31]. Mutational analysis of ERα action at AP-1 sites suggests these effects may be related to N-CoR binding [39], and we have proposed that SERM action at AP-1 sites may therefore involve contacts with corepressors [31, 51]. The fact that ERα and ERβ show completely different ligand preferences of interaction with N-CoR suggests that the target for SERM activation at AP-1 sites may not be N-CoR in both cases. Thus, this finding complicates our attempts to explain this unusual phenomenon. Perhaps the ER isoforms enhance AP-1 activity by superficially similar mechanisms that involve different cofactors. Alternatively, ERα and ERβ action at AP-1 sites could, in fact, be mediated by SERM-dependent contacts with a common cofactor that is, as yet, unidentified. This common factor may yet prove to be N-CoR if ERβ interactions with the β-box were somehow masked in vivo.

What features of the β-box contribute to ERβ specificity? Intriguingly, the β-box contains N-terminal proline and C-terminal serine residues that extend the homology of this region to an artificial ERβ-specific peptide [45]. However, the β-box also lacks the first Leu of the consensus LXXLL. A mutation (PLTIRML>PLLIRML) that restores the LXXLL consensus increases ERβ binding to N-CoR and permits ERα to bind to N-CoR in the presence of estrogens in mammalian two-hybrid assays. Thus, the unusual sequence of the β-box contributes to ERβ specificity and ERβ can tolerate the absence of a conserved N-terminal leucine in LXXLL motifs. ERβ might bind to yet more cofactors that contain variant NR boxes that resemble the β-box. Other aspects of ERβ interactions with corepressors warrant further study. It will be interesting to understand whether the weaker ERβ interactions with other regions of N-CoR (which are insensitive to ICI) play a role in ERβ binding (Fig. 4). Finally, SMRT also binds ERβ in the presence of estrogens, but we have not explored the structural features that promote this interaction. Intriguingly, human SMRT contains a sequence insertion at the position of the hydrophobic pair in the N-CoR β-box, which apparently leads to deletion of both residues (N-CoR/RMLS>SMRT/RLqagvmaS) [52]. Perhaps SMRT contains a different NR interacting motif or the N-CoR NR box sequence may be more complex than we have initially reported here.



Estradiol, diethylstilbestrol (DES), tamoxifen, genistein, coumestrol, thyroid hormone, retinoic acid and trichostatin A (TSA) were purchased from Sigma (St. Louis, MO). ICI 182,780 was a gift from Alan Wakeling (Astra/Zeneca Pharmaceuticals, Macclesfield UK). Raloxifene was a gift from Stefan Nilsson (KaroBio AB, Huddinge, Sweden). Peptides were synthesized at the Biomolecular Resource Center at UCSF.

The following plasmids (pSG5-ERα, pSG5-ERβ (amino acids 1–530) [31], pGEX-N-CoR and pGEX-SMRT[39], VP16-TRβ and Gal-N-CoR[16], GST-N-CoR fusions[17], ERE-LUC, GK1/Gal4 responsive reporter and Gal-ERα LBD [53], pM-D2, pM-D47, pM-F6[46]) have been previously described. VP16-ERβ LBD and Gal-ERβ LBD contain human ERβ sequences and were gifts from Dr. Dale Leitmann (University of California, San Francisco). VP16-RAR-LBD was a gift from Dr. David Moore, Baylor, Houston, Texas. Gal-GRIP1 NR box (1,2,3) fusion (amino acids 610–770) was prepared by PCR amplification of the appropriate region of GRIP1 (primers obtained from Biomolecular Resource Center, UCSF) containing EcoRI and SalI sites, the PCR fragment was digested with these enzymes and subcloned into the pM GAL4 expression vector (Clontech Laboratories, Inc. Palo Alto, CA). VP16-ERβ mutations and Gal-N-CoR mutations were prepared using standard PCR-based site directed mutagenesis (Quickchange Kit, Stratagene, La Jolla, CA) and confirmed by sequencing. The GAL4-β box fusion was prepared by synthesizing oligonucleotides corresponding to the β-box sequence with engineered EcoRI and SalI restriction sites. Annealed and phosphorylated double stranded oligonucleotide was subcloned into the appropriate sites in the PM vector.

Bacterial Protein Expression and GST Pulldown Assays

GST-fusions were expressed in E. Coli BL21 [28]. Cultures were grown to OD600 1.5 at room temperatures (approximately 22°C) and protein production was initiated by addition of IPTG to 1 mM. After four hours, bacterial pellets were obtained, resuspended in 20 mM HEPES pH 7.9/ 80 mM KCl/6 mM MgCl2/1 mM Dithiothreitol/1 mM ATP/0.2 mM phenylmethylsulfonyl fluoride and protease inhibitors and sonicated. Debris was pelleted by centrifugation in an ss34 rotor for 1 hour at 12,000 rpm. The supernatant was incubated with glutathione sepharose 4B beads (Amersham Pharmacia Biotech AB, Uppsala, Sweden) and washed as previously described. Protein preparations were stored at -20°C in 20% glycerol.

Labeled ERs were produced using coupled in vitro transcription-translation (TNT kit, Promega, Madison, Wisconsin). Assays were carried out in a volume of 150μl that contained 137.5μl of ice-cold protein binding buffer (PBB) along with 10μl of GST-bead slurry corresponding to 3μg of fusion protein, 1μl of in vitro translated protein and 1.5μl of ligand or vehicle and/or peptides or vehicle. PBB was freshly prepared in 24 ml aliquots composed of 20 ml A-150 (20 mM Hepes, 150 mM KCl, 10 mM MgCl2 and 1% glycerol), and 2 ml each of phosphate buffered saline supplemented, respectively, with 1% Triton X-100 and 1% NP-40. PMSF, DTT, BSA and protease inhibitor cocktail (Novagen) were added to 0.1 mM, 1 mM, 2μg/ml and 1/1000 dilution respectively. The mix was incubated for two hours in the cold room with gentle agitation, the beads were pelleted by spinning briefly on a bench top Eppendorf centrifuge, washed four times with PBB containing no BSA, and the pellet was dried under vacuum for twenty minutes. Labeled protein was subjected to SDS-polyacrylamide gel electrophoresis and autoradiography.


HeLa cells were grown in DME/F-12 Ham's 1:1 mix, without phenol red (Sigma) containing 10% iron supplemented calf serum (Sigma) and pen-strep. Cells were transfected by electroporation [28]. Transfections contained 2μg of luciferase and actin-β-galactosidase reporters and, where indicated, 1μg of ER, VP16-fusion protein or GAL4-fusion protein expression vectors or empty vector controls. Luciferase and β-galactosidase activities were measured using luciferase (Promega, Madison, WI) and Galacto-Light assay systems (Tropix, Bedford, MA).

Authors Contributions

PW Conceived and directed the studies, wrote the paper, performed plasmid constructions and transfection studies.

CV Performed transfection, mutagenesis and GST pulldown experiments under guidance of PW.

PN Performed transfection, mutagenesis and GST pulldown experiments under guidance of PW.

RHP Jr. Provided intellectual input in the study design.

AM. Expressed GST N-CoR fragments for pulldown studies with ERβ.

BW Designed bacterial expression vectors for N-CoR fragments, cloned cDNA fragments for GST-N-CoR expression vectors and provided intellectual input into the study design.

JDB. Directed studies of NR interactions with N-CoR and provided intellectual input into study design.

PJK. Co-directed studies along with PW and provided intellectual input into study design.

All authors read and approved the final manuscript.