A mutant form of ERα associated with estrogen insensitivity affects the coupling between ligand binding and coactivator recruitment

Abstract

A homozygous missense mutation in the gene encoding the estrogen receptor α (ERα) was previously identified in a female patient with estrogen insensitivity syndrome. We investigated the molecular features underlying the impaired transcriptional response of this mutant (ERα-Q375H) and four other missense mutations at this position designed to query alternative mechanisms. The identity of residue 375 greatly affected the sensitivity of the receptor to agonists without changing the ligand binding affinity. Instead, the mutations caused changes in the affinity of coactivator binding and alterations in the balance of coactivator and corepressor recruitment. Comparisons among the transcriptional regulatory responses of these six ERα genotypes to a set of ER agonists showed that both steric and electrostatic factors contributed to the functional deficits in gene regulatory activity of the mutant ERα proteins. ERα–coregulator peptide binding in vitro and RIME (rapid immunoprecipitation mass spectrometry of endogenous) analysis in cells showed that the degree of functional impairment paralleled changes in receptor-coregulator binding interactions. These findings uncover coupling between ligand binding and coregulator recruitment that affects the potency rather than the efficacy of the receptor response without substantially altering ligand binding affinity. This highlights a molecular mechanism for estrogen insensitivity syndrome involving mutations that perturb a bidirectional allosteric coupling between ligand binding and coregulator binding that determines receptor transcriptional output.

INTRODUCTION

Estrogen insensitivity syndrome (EIS) or estrogen resistance is caused by a defective estrogen receptor (ER) (1), most commonly due to a mutation in the gene encoding ERα that prevents estrogen from exerting its biological effects (2). Estrogens, including the endogenous hormone estradiol (E₂), are key regulators of growth, differentiation, and various physiological functions. E₂ acts through the two ERs, ERα and ERβ (3–6), both of which belong to the nuclear receptor (NR) superfamily of ligand-inducible transcription factors (7).

A homozygous missense ERα mutant, ESR1-Q375H, was identified in an 18-year-old female and led to profound estrogen insensitivity resulting in primary amenorrhea, continued linear growth into adulthood, delayed bone age, osteopenia, incomplete epiphyseal closure into adulthood, absent breast development, enlarged bilateral multicystic ovaries, and hypoplastic uterus with no endometrial stripe (8). The markedly high endogenous concentrations of E₂ in this patient (3500 pg/ml, compared to the normal range 15 to 350 pg/ml) are due to the lack of negative feedback on the hypothalamic-pituitary-adrenal axis, and she proved unresponsive to high doses of E₂ (8, 9). Although this clinical presentation is similar to phenotypes of the ERα knockout (αERKO) female mouse (3), the mouse KO phenotypes are due to the absence of ERα, whereas the patient phenotypes are due to the mutant Q375H ERα protein being markedly impaired in its response to estrogen. In the initial report on this patient (8), Q375H was found to be much less sensitive to E₂, requiring ca. 200-fold higher concentrations of E₂ in a cellular assay to achieve a transcriptional response equivalent to that of wild-type (WT) ERα.

Like other NRs, ERα has a distinct domain structure (10), with the A/B domain containing the hormone-independent activation function 1 (AF-1) and the C domain, which mediates DNA binding. The D domain or hinge region has nuclear localization sequences, and the E/F domain is the ligand-binding domain (LBD), which includes the ligand-dependent activation function, AF2, in which residue 375 is located (7, 11). In the “classical” genomic mechanism through which estrogens influence cell biology, estrogens are bound by nuclear ERα, and the nuclear estrogen-ER complex then binds either directly or indirectly to estrogen response element (ERE) sequences in the regulatory region of estrogen-responsive genes. This interaction is followed by the recruitment of a series of coregulatory proteins (coactivators or corepressors), resulting in changes in gene expression that induce physiological responses (12, 13).

Coregulators interact with NRs to alter chromatin accessibility and facilitate gene activation (14). More than 400 NR coregulators have been identified and characterized (14), the first being members of the steroid receptor coactivator (SRC) gene family (15). For many NRs, transcriptional activity derives from the AF2 region within the LBD and involves the interaction between a conserved hydrophobic cleft on the surface of the LBD and short, leucine-rich hydrophobic motifs (NR boxes, consensus LxxLL motif) reiterated within each coactivator (16). In addition, NR-interacting coactivators themselves have, or recruit other nuclear proteins that have, enzymatic activities crucial for efficient gene expression (17). It is notable that the Q375H mutation is on the coactivator LxxLL binding surface of the AF2 transactivation domain of ERα. However, the molecular mechanism by which Q375H renders ERα so insensitive to E₂, requiring much higher E₂ concentrations to become activated, remains unclear.

In this study, we probed the molecular features that underlie the impaired transcriptional response of this naturally occurring Q375H mutant ERα to estrogens. Using six estrogen ligands, we determined the effects of the Q375H mutation on ligand binding, coactivator interactions, and gene regulation. We also explored the effect of four synthetic mutations at this position that were designed by modeling to distinguish among possible alternative mechanisms. Using a cell system stably expressing either WT or Q375H mutant ERα, we performed quantitative polymerase chain reaction (qPCR) analysis to identify differences between them in the regulation of a few well-known endogenous ER target genes. We also performed RIME (rapid immunoprecipitation mass spectrometry of endogenous) analysis comparing cellular proteins bound to WT and Q375H ER in the absence and presence of E₂. Furthermore, we profiled the recruitment of coregulator-derived peptides by the ligand complexes with WT ERα or the five position 375 mutants using a fluorescence-based binding assay with a coregulator motif peptide chip. Insensitivity of the ERα-Q375H mutant to activation by estrogens resulted from an impairment in coregulator binding and not from a reduction in ligand binding affinity, thereby highlighting a bidirectional allosteric coupling between ligand binding and coregulator binding that is integral to the functional transcriptional output of the receptor.

RESULTS

Selection of substitution mutations at position 375 by molecular dynamics simulations

In the human ERα LBD, residue 375 is located near the start of helix 5 (h5), which is far from the ligand binding pocket but close to the top of the AF2 hydrophobic groove that accommodates the LxxLL motif of coactivators (Fig. 1A). In WT ERα, the glutamine residue, which is very highly conserved throughout the NR superfamily (18), is directed away from the coactivator groove itself (Fig. 1, A and B, and movie S1); this places Gln³⁷⁵ so that it is deeply engaged in productive hydrogen bonds with nearby residues (Fig. 1C). These interactions appear to stabilize the h3–4-5 turn motif of this region of the LBD, which is part of the “signature sequence,” the most highly conserved structural feature of NR LBDs (18). The 375 site is close in space to Lys³⁶², which is one of the two charge clamp residues and is positioned to stabilize the C terminus of the LxxLL α helix (Fig. 1, A and B, and movie S1).

Fig. 1. Location of ERα Gln³⁷⁵ in the AF2 core of the LBD and structural changes in the ERα-Q375H mutant.

(A) Ribbon drawing of a portion of ERα LBD showing Gln³⁷⁵ (residue in atom colors) in helix 5 (h5) at the top of the coregulator binding groove (AF2); Gln³⁷⁵ is far from E₂ (gray and red atom colors), which is situated in the ligand binding pocket below the groove and behind h12. The coregulator binding groove (AF2) contains the coactivator helix LxxLL sequence (green cylinder) and is stabilized by the N-terminal charge clamp residue Glu⁵⁴² (red residue in h12) and the C-terminal charge clamp residue Lys³⁶² (blue residue in h3); Lys³⁶² is close in space to Gln³⁷⁵. (B) Orthogonal (90°) views of the ERα LBD illustrating the position of the bound E₂ (atom coloring toward the bottom), the LxxLL coactivator peptide (green helix), the four helices that make up the coregulator binding groove (h-3–4-5 and h12), the location of Gln³⁷⁵ (colored side chain near the h4–5 turn), and the two charge clamp residues (Lys³⁶², blue; Glu⁵⁴², red). These structural features— and the relationships among them—are clearer in an animated version of this structure (movie S1). (C) Residues in h4 and the h4–5 loop that are near Gln³⁷⁵ and with which it can hydrogen bond. (D to F) Measured distances during MD simulation between the hydrogen bond donor of residue 375 side chain and the backbone carbonyl oxygen of residues 367, 368, and 370. The hydrogen bond donors are the amide nitrogen of Gln (D), Nδ of His (E), and Nε of His (F). (G) The change in side-chain conformation of residue 375 in the ERα-Q375H mutant compared to WT ERα perturbs the coactivator binding cleft. (H to J) Occupancy isosurfaces (magenta, 0.5; cyan, 0.75) for the side-chain heavy atoms of residue 375 [the amide nitrogen of Gln (H), Nδ of His (I), and Nε if His (J)] and the nearby charge clamp residue, Lys³⁶², projected onto the x-ray crystal structure.

In search of insight into the molecular mechanism driving the reduced activity of Q375H ERα, we used molecular dynamics (MD) simulations to assess structural changes between the WT and Q375H mutant receptor. The simulation trajectories of WT ER suggested high conformational stability of the glutamine side chain, specifically through hydrogen bonding interactions formed with the backbone carbonyls of residues 367, 368, and 370, and quantifying these distances for each simulation confirmed a stable interaction profile for Gln³⁷⁵ (Fig. 1D). Although the histidine substituted at this site contains two possible hydrogen bond donors available to form similar interactions (Nδ and Nε), substantially longer distances were observed for both in Q375H ERα (Fig. 1, E and F), indicative of a histidine side-chain conformation that has rotated away from the three backbone carbonyl interaction sites of Gln³⁷⁵ in WT ERα (Fig. 1G).

To understand how a change in side-chain conformation at position 375 might affect the shape of the hydrophobic binding cleft, occupancy maps were computed for this position and the nearby charge clamp residue Lys³⁶². An occupancy map is a representation that captures the volume occupied within the ensemble of conformations, with the frequency specified as an isosurface (Fig. 1, H to J). Visualizing the occupancy isosurfaces at 0.75 (cyan) and 0.5 (magenta) revealed that the histidine side chain rotated downward to adopt a conformation that appears to either directly interact with Lys³⁶² (His-Nδ, Fig. 1I) or occupies a position that occludes entry to the coactivator binding cleft (His-Nε, Fig. 1J).

On the basis of these data, we imagined two mechanisms by which the Q375H substitution could impede binding of the coactivator LxxLL motif: an “ionic-lock” whereby His³⁷⁵ forms a hydrogen bond with Lys³⁶², thus disrupting its role in the charge clamp, or a “steric-block” whereby the side chain directly occupies a portion of the binding cleft. To distinguish between these alternatives, we introduced four other mutations at position 375 (Fig. 2A): Q375L, which would only block coactivator binding through the steric mechanism; Q375K, which would introduce an electrostatic interaction that might support the Lys³⁶² charge clamp; Q375E, which would introduce an electrostatic interaction that might interfere with the Lys³⁶² charge clamp; and Q375A, a control substitution having neither steric nor electrostatic effects. Our strategy was to compare the effect of these four synthetic mutations with that of the natural Q375H mutant to characterize the contribution of these two mechanisms to the altered phenotype of the patient with Q375H ERα.

Fig. 2. The ERα Gln³⁷⁵ mutants and the chemical structures of the six ligands used in this study.

(A) Patient mutant ERα-Q375H and four other ERα synthetic mutants: Q375L, Q375K, Q375E, and Q375A. (B) Chemical structures of the ligands used in this study.

To investigate the responsiveness of all six ERα 375 genotypes to different ligands, we selected six well-known estrogen agonists of high potency (Fig. 2B). These included three steroidal ligands: E₂, 11β-chloromethylestradiol [CME (19), an Organon compound also known as Org 4333 (20)], and 11β-ethylestradiol [Et-E₂, an analog of CME (21)], as well as three nonsteroidal ligands: diethylstilbestrol (DES), meso-hexestrol (Hex), and trans-diethyl-tetrahydrochrysene [THC (22)].

The six ERα 375 genotypes show a wide spectrum in the transcriptional response sensitivity to agonists

To investigate ERE-mediated transcriptional activation by the six ERα 375 genotypes, we used a luciferase reporter assay in HepG2 cells individually transiently transfected with expression plasmids encoding each of the six mutants to compare the responses of WT, Q375H, and the four other Q375 mutants to increasing concentrations of the six ligands (Fig. 3, A and B). First, we confirmed that the amounts of each ERα protein expressed by the cells were comparable by Western blot analysis (fig. S1). For the steroidal ligands, E₂ began to activate WT ERα at low concentrations (0.1 nM) and activated ERα-Q375K at even lower concentrations (Fig. 3A). Activation of the natural ERα-Q375H mutant, by contrast, required higher E₂ concentrations than did WT, and activation of ERα-Q375A and ERα-Q375L required even higher concentrations (≥10 nM). The two other steroidal ligands were overall more potent than E₂, with CME significantly activating WT, ERα-Q375K, and ERα-Q375A even at the lowest concentration tested (0.01 nM) (Fig. 3A). Et-E₂ began to activate WT and ERα-Q375K at 0.01 nM, but the other four mutants required much higher concentrations (Fig. 3A). For the nonsteroidal ligands, DES activated WT, ERα-Q375K, and even ERα-Q375A at the lowest concentration (0.01 nM), with activation of ERα-Q375E being observed at 0.1 nM, ERα-Q375H at 1 nM, and ERα-Q375L at 10 nM (Fig. 3B). The pattern was relatively similar for Hex, but THC overall required higher concentrations (Fig. 3B).

Fig. 3. ERE-mediated transcriptional activity of WT ERα, the natural ERα-Q375H mutant, and the four designed mutants in HepG2 cells in response to the six ligands.

Cells were transiently transfected with 3xERE-luc reporter and pRL-TK plasmids encoding WT ERα or mutant ERαs Q375H, Q375K, Q375E, Q375A, or Q375L and later treated with vehicle or (A) the steroidal ligands 0.01 to 100 nM E₂, CME, or Et-E₂ or (B) the nonsteroidal ligands DES, Hex, or THC. ERE-mediated activity was detected by luciferase reporter assay. Data shown are representative of three independent experiments, and reporter expression was calculated relative to the vehicle control ± SEM, using the two-way ANOVA with Tukey’s multiple comparison test (^aP < 0.0001, ^bP < 0.001, or ^cP < 0.01).

Thus, all six of the ERα 375 genotypes were functional in this assay of transcriptional activity. They have comparable efficacies (or intrinsic activities), meaning that the WT and mutant proteins all had robust and equivalent or nearly equivalent maximal responses when given sufficiently high concentrations of agonist. On the other hand, they show marked differences in their sensitivities to ligand-induced activation based on the nature of the 375 residue. The rough rank order of ERα 375 genotype ligand sensitivity (or ligand potency) was Q375K ≈ WT and, then more variably, > Q375E ≈ Q375A ≥ Q375H > Q375L. The relative sensitivity of each ERα 375 genotype to the six ligands was similar; overall, CME and Et-E₂ were more potent than E₂, and DES and Hex were more potent than THC. None of the ligands activated Q375H at 0.1 nM, but CME, Et-E₂, DES, and Hex were more potent than E₂ and THC. Nevertheless, ligands that were more potent on ERα-Q375H than E₂ were also uniformly more potent on the other five ERα genotypes. Hence, none of the ligands seemed specific in rescuing the phenotype of the Q375H mutant ERα; they had similar, relative potencies on all the ERα mutants. The full set of median effective concentration (EC₅₀) values (table S1), which represents the sensitivity of the six ERα 375 genotypes to transcriptional activation by the six ligands, is a useful reference for comparison with the other characteristics of these receptors, such as ligand binding or coregulator interactions (see below).

Ligand binding affinity is largely unaffected by substitutions at position 375

In seeking to explain the reduced ligand sensitivity of some of the 375 mutant ERα in the transcriptional assay, we assayed the binding affinity of WT ERα and the five Gln³⁷⁵ mutants for E₂ and the five other ligands by direct or competitive radiometric assays, respectively, using [³H]E₂ as the tracer. Binding affinities were determined as the K_d (dissociation constant) values for E₂ and the K_i (inhibition constant) values for the other ligands (Table 1). All of the mutants bound E₂ with at most only a few-fold reduction in affinity compared to that of WT ERα, and aside from a few cases, the other ligands had similarly small reductions in binding affinity for the mutants compared to WT (Table 1). Thus, 375 mutations, even those that greatly reduced transcriptional responsiveness to estrogens (such as Q375H, Q375L, and Q375E), had little effect on ligand binding affinity. For each genotype, there were only small changes in the affinity of the six ligands (Table 1), consistent with the observation that none of the ligands had any particular ability to “rescue” ligand-induced transcriptional activation by any of the mutants or showed any notable allele-specific binding.

Table 1.

Binding affinities of the six ligands for WT ERα and the five Gln³⁷⁵ mutants.

		Ligand Ki or Kd ± SD (nM)* [Fold reduction in ligand binding affinity to 375 mutant vs. WT]† <Fold increase in ligand binding affinity vs. E2>‡
ER	WT-Q	Q375H	Q375L	Q375K	Q375E	Q375A
Steroidal ligands
E2	0.16 ± 0.03	0.67 ± 0.02	0.25 ± 0.01	0.36 ± 0.01	0.34 ± 0.04	0.14 ± 0.04
	[1]	[4.2]	[1.6]	[2.3]	[2.1]	[0.88]
	<1>	<1>	<1>	<1>	<1>	<1>
CME	0.11 ± 0.04	0.58 ± 0.14	0.22 ± 0.05	0.41 ± 0.10	0.35 ± 0.08	0.09 ± 0.02
	[1]	[5.4]	[2.0]	[3.7]	[3.2]	[0.82]
	<1.5>	<1.2>	<1.1>	<0.88>	<1.0>	<1.6>
Et-E2	0.13 ± 0.04	0.74 ± 0.04	0.16 ± 0.04	0.29 ± 0.02	0.27 ± 0.06	0.08 ± 0.03
	[1]	[5.7]	[1.2]	[2.2]	[2.1]	[0.62]
	<1.2>	<0.91>	<1.6>	<1.2>	<1.3>	<1.8>
Nonsteroidal ligands
DES	0.07 ± 0.03	0.39 ± 0.05	0.09 ± 0.01	0.16 ± 0.02	0.21 ± 0.05	0.08 ± 0.02
	[1]	[5.6]	[1.3]	[2.3]	[3.0]	[1.1]
	<2.3>	<1.7>	<2.8>	<2.3>	<1.6>	<1.8>
Hex	0.06 ± 0.02	0.29 ± 0.07	0.07 ± 0.01	0.29 ± 0.05	0.16 ± 0.02	0.06 ± 0.02
	[1]	[4.8]	[1.2]	[4.8]	[2.7]	[1.0]
	<2.7>	<2.3>	<3.6>	<1.2>	<2.1>	<2.3>
THC	0.09 ± 0.03	0.88 ± 0.1	0.43 ± 0.01	0.57 ± 0.03	0.42 ± 0.05	0.15 ± 0.06
	[1]	[9.8]	[4.8]	[6.3]	[4.7]	[1.7]
	<1.8>	<0.76>	<0.58>	<0.63>	<0.81>	<0.93>

^*K_d values for E₂ were determined directly with [³H]E₂ by Scatchard analysis. K_i values for the other five compounds were determined by competitive assays with [³H]E₂ as tracer. K_i = K_d^estradiol/RBA, where the RBA value was IC₅₀^compound/IC₅₀^estradiol. Both the RBA and the affinity measurements were done at 0°C. Values represent mean ± SD calculated from three independent assays. Statistical analysis by two-way ANOVA with Dunnett’s multiple comparison test, P < 0.0001 for the entire interaction, as well as for the column and row factors.

^†For each ligand (row), numbers in [brackets] represent fold reduction in ligand binding affinity of 375 mutant versus WT ER: [K_i/d^Q/K_i/d^mutant], read horizontally. Larger numbers indicate greater reduction in ligand binding by the 375 mutant ER compared to WT-ER.

^‡For each genotype (column), numbers in <carets> represent fold increase in affinity of ligand versus E₂: <K_d^E2/K_i^compound>, read vertically. Larger numbers represent greater increase in the affinity of ligand compared to E₂.

SRC coactivator binding affinities for the ERα Gln³⁷⁵ mutants vary widely

To investigate the correlation between the transcriptional response and coactivator binding, we first examined the affinities with which the NR interaction domains of SRCs SRC1 (also called NCOA1) and SRC3 (also called NCOA3) bound to WT ERα and the five Gln³⁷⁵ mutants when they were fully saturated with each of the six estrogens. A Förster resonance energy transfer (FRET) binding assay composed of a FRET donor–labeled ERα LBD and a FRET acceptor–labeled NR interaction domain of the coactivator (ca. 200 amino acids) was run in coactivator titration mode to quantify binding affinities (23).

SRC1 bound with similar high affinity to WT ERα complexes with five of the ligands, but with lower affinity to the THC complex (Fig. 4 and Table 2). SRC1 showed essentially no binding to the natural ERα-Q375H mutant or to ERα-Q375L and substantially lower affinities for the ERα-Q375A and ERα-Q375E mutants, but the affinity of SRC1 for liganded ERα-Q375K complexes was very similar to that for WT ERα (Fig. 4). Although SRC1 is the SRC coactivator most closely associated with reproductive function (24), we found that SRC3, a related coactivator associated with breast cancer function, showed the same pattern of dependence on the residue at position 375 as did SRC1; however, overall SRC3 bound somewhat less well than SRC1 (table S2).

Fig. 4. SRC1 coactivator binding affinity for the ligand complex with WT ERα and the five Gln³⁷⁵ mutants.

Binding curves for the titration of an SRC1-derived, ERα-interacting peptide with the LBDs of the six ERα Gln³⁷⁵ genotypes, measured by a time-resolved FRET assay. n = 3 independent experiments.

Table 2.

Kd values for SRC1 binding to WT ERα and five Q375 mutants in the ligand complex.

		Kd ± SD (nM)* [Fold reduction in SRC binding to mutant ER vs. WT ER]† <Fold increase in SRC binding with ligand vs. E2>‡
SRC1
Ligand/mutant	WT-Q	H	L	K	E	A
E2	6.4 ± 2.0			2.1 ± 0.8	± 5	± 11
	[1]	nb§	nb§	[0.33]	[3.4]	[8.0]
	<1>			<1>	<1>	<1>
CME	3.0 ± 0.3			1.6 ± 0.2	± 3	± 3
	[1]	nb	nb	[0.53]	[8.7]	[16]
	<2.1>			<1.3>	<0.85>	<1.1>
Et-E2	3.4 ± 0.4			1.7 ± 0.4	± 1	± 1
	[1]	nb	nb	[0.5]	[11]	[17]
	<1.9>			<1.2>	<0.61>	<0.86>
DES	6.8 ± 2.0			2.2 ± 0.7	± 2	± 15
	[1]	nb	nb	[0.32]	[2.8]	[8.8]
	<0.94>			<0.95>	<1.2>	<0.85>
Hex	3.5 ± 0.4			1.5 ± 0.2	± 11	± 7
	[1]	nb	nb	[0.43]	[7.4]	[21]
	<1.8>			<1.4>	<0.85>	<0.68>
THC	± 8			± 4	± 17	± 11
	[1]	nb	nb	[0.62]	[4.2]	[5.5]
	<0.30>			<0.16>	<0.25>	<0.44>

^*The K_d values were obtained from coactivator titration curves with SRC1. All experiments were done with 1 nM ER, 1 μM ligand, and increasing amounts of SRC1. Values represent mean ± SD calculated from three independent assays. Statistical analysis by two-way ANOVA with Dunnett’s multiple comparison test, P < 0.0001 for the entire interaction, as well as for the column and row factors.

^†For each ligand (row), numbers in [brackets] are fold decrease in SRC1 binding affinity relative to WT ER [K_d^mutant/K_d^Q], read horizontally. Larger numbers indicate greater reduction in SRC binding affinity by the mutant.

^‡For each ER genotype (column), numbers in <carets> are fold increase in SRC1 binding affinity with ligand versus E₂: <K_d^E2/K_d^compound>, read vertically. Larger numbers represent greater increase in SRC binding affinity to ER with bound ligand compared to bound E₂.

^§nb, no binding; for the Q375H and Q375L mutants, the level of SRC binding was insufficient to determine a K_d value.

The Q375H ERα mutant has defects in E₂-responsive gene expression and interactions with chromatin-associated proteins

To explore the hypothesis that impaired activity of the natural Q375H mutant changes endogenous gene expression and global protein interactome profiling, we established HepG2 cells stably expressing WT ERα or the ERα-Q375H mutant using a lentiviral infection system. The ERα transgenes were stably integrated, and expression of the endogenous ERα target gene ESR1 and expression of the ERE luciferase were confirmed in the infected cells (fig. S2). HepG2 cells expressing the ERα-Q375H protein (HepG2/Q375H cells) had higher amounts of ERα protein (fig. S2A) and basal ESR1 gene expression (fig. S2B) compared to cells expressing the WT protein (HepG2/ WT cells). Nevertheless, despite the greater abundance of the ERα-Q375H mutant in HepG2/Q375H cells, the transcriptional activity of the ERE-reporter gene at 1 and 10 nM E₂ was markedly less in HepG2/Q375H cells compared to HepG2/WT cells (fig. S2C), which is similar to the results in the transiently transfected HepG2 (Fig. 3). The effect of differing expression levels of ERα-Q375H versus WT ERα on gene responsiveness has no influence, as shown by a previous study in which we made stable cell lines in MDA-MB-231 cells, some having considerably higher levels of ERα-Q375H than WT ERα (25). Even with the high ERα-Q375H levels, there was no evidence that the pattern and level of gene expression was altered in any substantial way in the cell lines with different ER levels, suggesting to us that squelching due to different ER expression levels is not likely to be an issue. GREB1 (gene regulated by estrogen in breast cancer 1), TFF1 (Trefoil factor 1, also known as PS2), and WISP2 (WNT1-inducible-signaling pathway protein 2, also known as CCN5) are ER targets stimulated by receptor activation (26). We focused on these three genes to examine the responses to E₂ in the HepG2/WT and HepG2/ Q375H cells. The expression of all three genes was significantly induced by 10 nM E₂ in HepG2/WT cells, but there was no significant induction of GREB1 or WISP2 in HepG2/Q375H cells and a lesser induction of TFF1 compared to HepG2/WT cells (Fig. 5A). These effects on E₂-induced endogenous gene regulation suggest again that impaired estrogen-responsive gene expression is a key factor that results in reduced functionality by the natural mutant ERα-Q375H.

Fig. 5. Changes in the expression of ERα target genes and RIME analysis for WT and ERα-Q375H.

(A) Expression of GREB1, TFF1, and WISP-2 by qPCR analysis in HepG2 cells stably expressing WT or ERα-Q375H treated with vehicle or the indicated concentrations of E₂. The samples (n = 4) were normalized to human ACTB expression. The fold changes by log₂ transform were calculated relative to the HepG2/WT ERα vehicle ± SEM, using the two-way ANOVA with Tukey’s multiple comparison test (*P < 0.05, **P < 0.01, ***P < 0.001, or ****P < 0.0001). (B) Comparisons of protein interactomes of WT ERα and ERα-Q375H (QH) ERα. Venn diagrams show unliganded interactions and liganded interactions. The interactions unique to WT and ERα-Q375H are listed. (C) Interactions of ERα, GREB1, HDAC1, and GADD45G with WT ERα or Q375H mutant ERα by RIME analysis of chromatin extracts from HepG2 cells stably expressing WT or ERα-Q375H and treated with vehicle or 10 nM E₂. RIME assay was carried out using an ERα antibody to identify the proteins that interacted with ERα using mass spectrometry.

RIME is a method for characterizing protein complexes, especially chromatin and transcription factor complexes (27). To determine whether WT ERα and the Q375H mutant have different interactomes in a cellular context, we performed RIME analysis in HepG2/WT and HepG2/Q375H cells, in the absence and presence of E₂. The ERα interactomes in the four groups of samples revealed common and distinct members in each group (data file S1 and Fig. 5, B and C).

Under basal conditions (the absence of ligand), we found that 80 proteins interacted with WT ERα (data file S1, S#1), whereas only 50 proteins interacted with Q375H mutant ERα (data file S1, S#2) (Fig. 5B, left). By overlapping the two lists, we found that 44 proteins were shared between WT ERα and ERα-Q375H mutant (table S3 and data file S1). On the basis of spectral counts, most protein interactions with WT ERα were stronger than with ERα-Q375H, but there were a few interactions that showed a much stronger interaction with ERα-Q375H (data file S1).

Under conditions of agonist stimulation, we found 63 proteins that interacted with WT ERα (data file S1) and 54 proteins that interacted with Q375H mutant ERα in the presence of E₂ (data file S1 and Fig. 5B). Comparing the two lists showed that 35 of the proteins were shared between WT ERα and Q375H mutant ERα after E₂ treatment (table S4 and data file S1). As expected, both WT ERα and ERα-Q375H proteins were pulled down with the ERα antibody whether or not they were E₂-liganded (Fig. 5C); this interaction was fundamental to the RIME analysis in these cells. The interaction of most of the shared proteins was stronger with WT ERα than with ERα-Q375H when both were E₂-liganded (data file S1), consistent with protein interactions with the unliganded ERs.

To further investigate differences between the interactomes of WT and Q375H mutant ERα, we focused on proteins that were uniquely associated with either WT or Q375H ERα, with or without E₂ (Fig. 5B). We identified 36 protein ligand-independent interactions that were specific to WT ERα and 6 that were specific to ERα-Q375H (Fig. 5B). There were 28 ligand-dependent protein interactions that were specific to WT ERα and 19 that were specific to ERα-Q375H (Fig. 5B). We selected three proteins—GREB1, HDAC1, and GADD45G—to highlight different patterns of interaction with WT and Q375H ERα. GREB1 interacted with WT ERα in a ligand-dependent manner but had no interaction with ERα-Q375H, whether E₂-liganded or not (Fig. 5C). GREB1 is both an estrogen-responsive transcriptional target and a coregulator of E₂-stimulated WT ERα (26, 28, 29). Two other proteins had completely different interaction profiles between WT ERα and the ERα-Q375H mutant (Fig. 5C): HDAC1, a histone deacetylase associated with NR function (30), and GADD45G, a NR coregulator (31). Together, these results show that WT and ERα-Q375H interacted with some of the same proteins whether they are unliganded or E₂-liganded. A greater number of proteins, however, interacted only with WT ERα, and fewer interacted exclusively with ERα-Q375H. These findings indicate that the Q375H mutation in ERα has altered the manner in which this receptor interacts with other endogenous cellular proteins, alterations that might be associated with the loss of gene regulatory function by ERα-Q375H.

Peptide interaction profiling reveals marked differences in preferential interactions of WT ERα and Gln³⁷⁵ mutants in the absence and presence of ligands

We used a high-throughput assay to further investigate mechanisms and interactions that underlie the altered cellular activity and binding behavior of the Gln³⁷⁵ mutants versus WT ERα. Western blot analysis of human embryonic kidney (HEK) 293 cells individually transfected with expression plasmids for WT ERα and the five Gln³⁷⁵ mutants showed that the amounts of protein expressed for the six ERα genotypes were comparable (fig. S3). These six groups of cells were then subjected to microarray analysis for real-time coregulator-NR interaction (MARCoNI) under basal (ligand-free) conditions and when bound by each of the six ligands. In this assay, the binding profile of the ERα to an array of peptides representing coactivator and corepressor NR interaction sequences (NR and CoRNR boxes, respectively) reflects the conformation of the receptor (32, 33). The modulation of ERα conformation, as induced by the Gln³⁷⁵ mutation and/or ligand, is expressed by a modulation index (MI), which is the log-fold alteration of binding liganded versus apo WT ERα and is calculated for all coregulator peptides on the array.

The effects of ERα mutation and/or ligand binding on receptor conformation reflected by the MI values for the coregulator interactions are summarized in a global heatmap (fig. S4). Unsupervised clustering segregated the conformation of unliganded WT ERα and Q375K from the other genotypes and, further, in the presence of any ligand, quite neatly subclustered the six genotypes into groups. Similar unsupervised clustering divided the coregulator motifs into two major classes (described below). To better illustrate our observations, we extracted from this global heatmap a series of subsidiary heatmaps showing MI values for the six unliganded ERαs (Fig. 6A), as well as those that best distinguished among the agonist-liganded ERαs (Fig. 6B). In these heatmaps, we have maintained the classification of the coregulator motifs (columns) obtained in the global heatmap, whereas genotypes and/or ligands are arranged in the order used in presentation of the data from our other assays.

Fig. 6. Comparison of peptide binding to WT and ERα-Q375H.

We measured binding of WT and ERα-Q375H to a PamChip containing 154 coregulator-derived motif peptides. Triplicates were used to calculate mean binding and log₁₀ fold change of binding [modulation index (MI)] values versus WT apo (unliganded) ERα (control) and to determine the significance of this modulation by Student’s t test with post hoc false discovery rate correction, P < 0.05. Hierarchical clustering was applied to sort compounds and motifs and visualize similarity or dissimilarity by dendrograms. The MI values were normalized to those for WT ERα (MI = 0). The red color shows increased ERα binding (MI > 0), and the blue color shows decreased ERα binding (MI < 0) by the ligands. (A) Heatmap for basal ERα binding in the absence of ligand and lists of the subset of potential coregulators represented by peptides that bound more strongly to ERα-Q375H than to WT ERα (Q375H > WT) and those that interacted more strongly with WT ERα than to ERα-Q375H (WT > Q375H). (B) Heatmap for peptides interacting with ligand-bound WT and ERα-Q375H. (C) Binding of coactivator peptides (SRC1, SRC2, NRIP1, and PRGC1) to the WT ERα-ligand complex relative to the ERα-Q375H ligand complex for all six tested ligands (E₂, CME, Et-E₂, DES, Hex, and THC).

Hierarchical clustering of MI values for peptide binding under basal conditions (the absence of ligand) differentiated the peptides into two groups: those that showed stronger binding to ERα-Q375H than to WT ERα (Q375H > WT) and those that showed the reverse (WT > Q375H) (Fig. 6A). These groups were considered to be emblematic of coregulators associated with the low-activity natural Q375H mutant ERα versus those supporting the high activity of the WT ERα, respectively. In summary, we found that a subset of peptides (16 peptides from 12 coregulators) had stronger binding to ERα-Q375H than to WT ERα (Fig. 6A and table S5), and another subset of peptides (51 peptides from 22 coregulators) showed stronger binding to WT ERα than to ERα-Q375H (Fig. 6A and table S6).

The 16 peptides that preferentially bound to the low-activity ERα-Q375H represented mainly potential coregulators but also corepressors such as CHD9, NOT1H, and NCOR1, whereas the coregulators represented by the 51 peptides favoring binding to the high-activity WT ERα (WT > Q375H) were highly enriched for coactivators such as SRCs, RIP140, and PGC1α. The two groups of peptides favoring WT ERα or ERα-Q375H were shared with the other four genotypes to a degree that mirrored their respective functional capacities (Fig. 6A). Thus, even with ERs in their unliganded state, this assay detected differences in the basal recruitment of coregulators by WT ERα and the five Gln³⁷⁵ mutants that reflected remarkably well how all six behaved in the agonist-liganded state in the gene expression assays (Fig. 3 and table S1) and in the quantitative SRC3 binding assays (Fig. 4 and table S2), with Q375K being similar to WT ERα, the Q375H and Q375L mutants being the most different, and Q375E and Q375A being intermediate.

We next performed hierarchical clustering of MI values for peptide binding to WT ERα and ERα-Q375H in the presence of E₂ and the other five ligands (Fig. 6B). Again, as in the analysis of the apo WT ERα, the peptides clustered two groups: those that showed stronger binding to ERα-Q375H than to WT ERα (Q375H > WT) in the presence of every ligand and those that showed the reverse (WT > Q375H). With each of the six genotypes, one can observe the impact of agonist-ligand binding best by focusing on the coregulators preferred by WT ERα (Fig. 6B). Compared to the apo-ERs, agonist binding markedly enhanced recruitment of these peptides in WT ERα and ERα-Q375K; ERα-Q375E and ERα-Q375A show a switch from reduced binding in the apo state to more modestly promoted binding when agonist bound, and with agonist binding, ERα-Q375H and ERα-Q375L show a partial reversal of the reduced binding of these peptides. The recruitment of the coregulators preferred by ERα-Q375H (Fig. 6B) shows a pattern that is essentially the reciprocal of the degree to which the WT ER–preferring group of peptides was recruited; this is most evident in the WT ERα and ERα-Q375K compared to the ERα-Q375H and ERα-Q375L mutants. Thus, the pattern and degree of recruitment of members of these two coregulator peptide classes to all six agonist-liganded ERα Gln³⁷⁵ genotypes match that found in the basal recruitment by the apo-ERs (Fig. 6A), and they also align with the sensitivity of the transcriptional response in luciferase reporter assays (Fig. 3) and the quantitative binding of the coregulator SRC1 (Fig. 4). In the RIME assay, there were several proteins that interacted with both E₂-liganded and unliganded receptors, whether WT ERα or ERα-Q375H mutant (data file S1).

To further explore coactivator and corepressor recruitment by the WT and mutant ERα proteins, we selected peptides from four well-known coactivators (SRC1, SRC2, NRIP1, and PRGC1) and four corepressors (CNOT1, NCOR2, NELFB, and MGMT) for additional evaluation (table S7) and calculated the MI values for these eight peptides, relative to apo WT ERα as control, with all six ERα genotypes for all six ligands (Fig. 6C and fig. S5). For the coactivator peptides, WT ERα had the highest binding values for all four peptides with all six ligands (Fig. 6C). A very similar binding profile in the five Gln³⁷⁵ mutants was observed with SRC1 and SRC2, with binding values as follows: WT ≈ Q 375K > Q375E > Q375A > Q375H > Q 375L. The NRIP1 binding values of the five mutants were only modestly different: WT > Q375E > Q375K = Q375A > Q375H > Q375L, and with a few deviations, those for PRGC1 were the same as for the SRCs. There were few notable differences in the binding patterns for the different ligands except for Et-E₂, which showed only negative (weak) MI with all of the mutant genotypes, in agreement with its effect on reporter gene transcription (Fig. 3).

The corepressor peptides showed a pattern that was essentially the reciprocal of the coactivator peptide binding (fig. S5). The WT ERα–ligand complexes had the lowest MI (weak corepressor peptide binding), followed most frequently by increased MI in ERα-Q375K and then more roughly by Q375E < Q375L < Q375H < Q375A. With all five Gln³⁷⁵ mutants, Et-E₂ had the highest MI (strongest binding) for four corepressor peptides compared to the other five ligands. Overall, both the reciprocal recruitment of coactivators and corepressors by WT ERα and the Gln³⁷⁵ mutants and the differential strength of their interactions reflect the sensitivity of the different Gln³⁷⁵ genotypes to the ligands in the transcription assay (Fig. 2 and table S1), with the most responsive WT and Q375K ERs having strong interactions with coactivators but not corepressors, and the four other less responsive ERα proteins generally favoring corepressor over coactivator binding.

DISCUSSION

Here, we have investigated the functional consequences of a naturally occurring, homozygous Q375H mutation in ERα that rendered a patient essentially unresponsive to administration of high concentrations of estrogens (9). To better understand the mechanistic basis for the impaired sensitivity of ERα-Q375H to E₂, we compared the response of this ERα genotype to that of WT ERα and of ERα with four other Gln³⁷⁵ mutations designed to probe alternative mechanisms that might underlie the reduced sensitivity to estrogens. These mutational changes had little effect on ligand binding affinity, agonist efficacy, or intrinsic activity, and full activation of all of the mutants could be reached with sufficiently high ligand concentrations. However, the mutations greatly affected coregulator binding affinity, resulting in marked reductions in agonist ligand potency. The functional deficit in the patient-derived mutant ERα-Q375H appeared to result from its greatly impaired ability to recruit coactivators necessary for ER transcriptional regulatory activity, as reflected by differences in its protein interactome compared with that of WT ERα.

The comparative impact of the ERα 375 genotype on transcriptional sensitivity to ligands and on the binding affinity of ligands and coregulators can be visualized by a graphical summary (Fig. 7), in which each panel represents one ligand, and each plot displays four different parameters for the six genotypes, quantified on a log scale relative to WT ERα: (i) ligand binding affinity (Table 1), (ii) SRC1 binding affinity (Fig. 4 and table S2), (iii) overall index of coactivator interaction (Fig. 6B), and (iv) transcriptional responsiveness (table S1). These “genotype profiles” illustrate that alterations in the sensitivity of ERα 375 genotypes to agonist ligand activation (ligand potency in transcription assays) closely resemble the patterns of indices that summarize coactivator binding (MI) as well as SRC1 binding affinity, whereas ligand relative binding affinity (RBA) is much more minimally altered by 375 genotype. The genotype profile of the corepressor interaction is essentially the reverse or mirror image that of the coactivator interaction index (fig. S6).

Fig. 7. Changes in ligand binding, transcriptional responses, coactivator peptide binding, and SRC1 binding for each of the ERα mutant genotypes relative to WT ERα.

Genotype profile plots show relative ligand binding affinities (RBA, blue, Table 1: log₁₀ of K_i^Q/K_i^mutant), modulation indices of coactivator peptide binding profiles (MI, red, average values from Fig. 6B), transcriptional response (yellow, table S1: log₁₀ of EC₅₀^Q/EC₅₀^mutant), and SRC1 binding affinity (SRC1, green, Table 2: log₁₀ of K_d^Q/K_d^mutant) for WT ERα and the five ERα mutants Q375H, Q375L, Q375K, Q375E, and Q375A. The one-letter amino acid abbreviations on the x axis indicate the identity of the amino acid at position 375. Because MI values from the MARCoNI assay represent the log fold change in ERα-peptide interaction for WT ERα and E₂ due to ligand and/or ERα genotype and are not strictly K_d or EC₅₀ values, the scale for presenting MI values (right y axis) has been expanded to highlight the close match in profile between the genotype profiles for MI and for transcription response sensitivity and SRC1 binding. All the values for WT ER are zero because the values for the changes in these four parameters for the five mutants are expressed as a ratio relative to WT ER on log₁₀ scales.

Gln³⁷⁵ in ERα is a critical nexus for LBD structure and interaction with coregulators

In WT ERα, Gln³⁷⁵ is located in a region of the LBD that is highly conserved in nuclear hormone receptors (NRs), with glutamine at this position being invariant across the entire NR superfamily (18). In crystal structures of the ERα LBD, Gln³⁷⁵ is engaged in organizing an extensive network of hydrogen bonds that stabilize an important tertiary structural motif, the h3–4-5 turn (Fig. 1, A and B, and movie S1), which forms the fixed end of the AF2 hydrophobic groove for coregulator binding; Gln³⁷⁵ is also close to the Lys³⁶² charge clamp residue that stabilizes coactivator binding. Hence, one would expect mutational changes at 375 to affect ER function by altering coregulator binding rather than ligand binding, consistent with what we have found.

Although the profound loss of ERα function in the Q375H mutant might be due to a general disruption of the h3–4-5 turn motif, our MD simulations suggested that histidine in this position might be interfering with coactivator binding by either a steric or a charge mechanism. The four additional synthetic mutations that we made at the 375 position were designed to probe these two alternatives, and both appear to be operative. Steric obstruction by the bulky Leu³⁷⁵ mutant residue has an effect essentially equivalent to that of His³⁷⁵ in the natural mutation. The charge clamp activity of Lys³⁶² was assisted by like-charged Lys³⁷⁵ mutant residue, resulting in an ER with activity rivaling—and sometimes even exceeding—that of WT ERα; by contrast, very low activity resulted from the opposite-charged Glu³⁷⁵ mutant residue, which would neutralize the charge clamp Lys³⁶². The Q375A mutation was intended to be sterically and electronically neutral, but alanine lacks the hydrogen bonding capabilities of glutamine, and its significant, though more modest, effects likely represent a loss of the intrinsic stabilizing effect of Gln³⁷⁵ in WT ERα. Thus, given the very high conservation of glutamine at this position across the whole NR superfamily (18), it is likely that any mutational change at this or even nearby sites could have other, perhaps more subtle effects on the shape and flexibility of the coregulator binding groove, resulting in alterations in the balance of coactivator and corepressor binding affinity and preference that might lead to other forms of hormone insensitivity syndromes (HISs).

How mutational changes at Gln³⁷⁵ that do not affect ligand binding might alter ER sensitivity to agonists

Given the location Gln³⁷⁵ in the upper reaches of the coregulator binding groove, far from the ligand binding pocket, the fact that substitutions at this site have more pronounced effects on the binding of coregulators rather than the binding of ligands is not unexpected. What is puzzling, however, is that these alterations in coregulator binding and coactivator versus corepressor recruitment preferences have their principal effect not on the efficacy of agonists but on their potency. How the effect of this remote-site mutation is transduced into altered receptor sensitivity to ligands in mediating ER responses is the crux of this conundrum.

Because alterations of the WT Gln³⁷⁵ genotype had very little effect on agonist binding affinity, one can presume that the very poorly responsive ERα-Q375H and ERα-Q375L will become saturated with agonist at much lower concentrations than that required for full activation of the response. For equivalent ligand saturations to result in different dose-response profiles, one can offer two possible mechanisms: First, there may be an “activity-limiting step” in the signal transduction cascade for ERα activation, downstream of agonist-liganded ERα, which becomes saturated by low concentrations of the ternary ligand-receptor-coactivator complex. ERs that avidly bind coactivators would reach maximum activation at low ligand concentrations, giving the high apparent potencies for E₂ with WT ERα or ERα-Q375K. By contrast, ERs such as Q375H and Q375L, which bind coactivators much less well, would require higher fractional saturation by ligand to reach concentrations of the E₂-ERα-coactivator ternary complex needed for full response activation; this would result in the low apparent potency for E₂ with these ERα genotypes. With certain G protein–coupled receptors, such a “spare receptor hypothesis” has been used to explain cases where the dose-response EC₅₀ is substantially left-shifted with respect to the ligand-binding K_d values (34). Evidence supporting this mechanism for ERα can also be gleaned from the very low concentrations of E₂ needed to stimulate proliferation of MCF-7 breast cancer cells (35), as well as the very low doses of E₂ needed to maximally stimulate certain physiological responses, such as the increase in uterine weight in immature rats (36).

A second possibility involves more general allosteric mechanisms whereby the global dynamics of the ERα LBD can reciprocally couple the interaction of coregulators with the potency with which agonists activate the receptor. Previous studies in our own laboratories as well as in those of others support such bidirectional allosteric mechanisms. For example, the binding of coactivator peptides to the ERα LBD can profoundly affect agonist-binding kinetics, markedly reducing dissociation rates (37); ligand stabilization of the ERα LBD can be further enforced by coactivator peptide binding (38); both ligand and coactivator levels affect the sensitivity of ERα activity to agonists (23); and a bidirectional relay between coactivator and ligand binding in the androgen receptor (AR) affects the response to ligand (39).

The bidirectional allosteric interactions operative in the ERα LBD are related to a number of other features of these proteins: Both NR LBDs and coregulators are intrinsically disordered proteins that become more structured upon interaction with other proteins (40, 41), features that enable flexibility in coregulator interactions with many transcription factors (42) and for NR LBDs to be efficient in searching for binding partners (43). ERα LBD mutations found in ERα-positive but endocrine therapy–resistant breast cancers reveal how mutations can overcome forces involved in ensuring that WT ERα is off in the absence of an agonist ligand (44) and illustrate that the coregulator-mutant ERα interaction landscape is allele selective (45). Furthermore, a recent study highlights the dynamic assembly and activation of ERα enhancers through coregulator switching using selective loss- or gain-of-function ERα mutants (46). A better understanding of the diverse ways by which mutations in the ERα LBD can affect allosteric coupling mechanisms will require more detailed structural and computational energetics and dynamics analyses in the future that ideally will include study of coregulator recruitment and histone modifications at the chromatin level.

ESR1 mutations associated with EIS cause loss of estrogen responsiveness and differential interactions with endogenous proteins

A different homozygous missense mutation in ERα, this one located at a critical site of ligand binding stabilization (R394H), was recently identified in three siblings (two females and a male), originating from a consanguineous Algerian family (47). Recently, we have investigated the whole transcriptome profiles of these two clinical ERα mutations, Q375H and R394H, in stably expressing the MDA-MB-231 cells (25). Both clinical mutants show a different gene expression profile compared to WT ERα, consistent with a loss of estrogen responsiveness. We also generated a transgenic mouse model Esr1-Q, harboring the human Gln to His mutation, using CRISPR-Cas9 genome editing. Both female and male Esr1-Q mice are infertile and have phenotypes similar to αERKO mice. The overall phenotype of female Esr1-Q mice corresponds to that observed in the Q375H patient. In the current study, using RIME assays in HepG2 cells stably expressing WT ERα or the Q375H mutant ERα, we found that the Q375H mutation also interacted differently with endogenous proteins compared to WT ERα. These findings highlight the molecular mechanisms underlying the impaired activity of clinical ERα mutations at the genome-wide level in vitro and provide a better understanding of their phenotypes and biological function in vivo.

Different modes by which mutations in NRs can lead to HISs

Last, we consider how our findings fit within the larger context of HISs due to mutations in nuclear hormone receptors. This physiological phenomenon encompasses a wide variety of conditions that result in an individual’s inability to exhibit an expected response to circulating steroid hormone, and it is well represented by numerous mutations throughout the gene encoding the AR, which cause androgen insensitivity syndrome or androgen resistance (48–50). By contrast, mutations in the gene encoding the ER are thought to be rare, in all cases only arising due to mutational homozygosity within consanguineous pedigrees. The three that have been reported have distinctly different origins and different effects on the ER, with one mutation causing ER truncation, one disrupting ligand binding, and one disrupting coactivator binding. The truncation was identified in an estrogen-insensitive 28-year-old man with a disruptive homozygous nonsense mutation (R157X) that produced a premature stop codon in the N terminus of ERα, resulting in no ER protein expression (2). The mutation that disrupts the hormone-receptor interaction (R394H) was identified in three siblings mentioned above, with the two sisters presenting with skeletal defects of retarded bone age and low bone mineral density (BMD), lack of breast development, endometrial atrophy, and cystic ovaries, and the brother having similar skeletal defects with unilateral cryptorchidism (46). The mutation that we have studied here (Q375H) (8) has only minimal effect on ligand binding but results in profound resistance to estrogens by altering the coactivator LxxLL binding surface of the AF2 transactivation domain, revealing a bidirectional allosteric communication between ligand binding and coactivator binding that determines the ligand sensitivity of the ERα transcriptional response. One may presume that additional molecular mechanisms, beyond the ones we have thus far characterized, may underlie other examples of EIS that are likely to be found in the future, each of which may reveal other features of the ERα LBD that are needed for agonist binding to be transduced into functional receptor transcriptional activity.

MATERIALS AND METHODS

Chemicals

E₂, DES, and Hex were purchased from Sigma-Aldrich. CME, Et-E₂, and THC were prepared in the Katzenellenbogen laboratory (19, 21, 22). All 10 mM stock solutions were made in dimethyl sulfoxide (DMSO) and kept at −20°C.

Plasmids

The expression vector pcDNA3 was purchased from Invitrogen. An internal control plasmid for transfection efficiency, pRL-TK renilla luciferase (pRL-TK Luc), was purchased from Promega. The synthetic vitellogenin 3xERE-TATA fused to a luciferase reporter gene plasmid, 3xERE Luc, and a full-length human ERα expression plasmid, pcDNA/WT ERα, and pcDNA/SRC-2 expression plasmids have been described previously (51). All the Q375 mutant expression plasmids used were generated by Celplor Inc. (www.celplor.com) using the pENTR system.

Cell lines and tissue culture

The human hepatocellular cancer cell line HepG2 and HEK cell line HEK293 were purchased from the American Type Culture Collection. HepG2 cells were maintained in phenol red–free minimum essential medium (MEM) (Invitrogen), and HEK293 cells were maintained in phenol red–free Dulbecco’s modified Eagle’s medium (DMEM) (Invitrogen) supplemented with 10% fetal bovine serum (FBS; Gemini Bio-Products) and 4 mM l-glutamine (Invitrogen). For cell-starving conditions, 10% charcoal/dextran stripped FBS (sFBS; Gemini Bio-Products) was substituted for FBS in the medium.

Transient transfection and luciferase reporter analysis

HepG2 cells were seeded in 24-well plates with 10% sFBS medium overnight. A total of 0.5 μg of DNA, including 0.2 μg of WT ERα or Q375 mutants, 0.2 μg of 3xERE Luc, and 0.1 μg of pRL-TK Luc plasmids, was transfected into the cells using Effectene transfection reagent (Qiagen). After 6 hours, cells were changed to fresh 10% sFBS medium overnight and then treated with vehicle control (DMSO, final concentration < 0.01%), E₂, CME, Et-E₂, DES, Hex, or THC (0.01 to 100 nM) for 18 hours. Luciferase assays were performed using the Dual Luciferase Reporter Activity System (Promega). Transfection efficiency was normalized against the renilla luciferase. Fold changes were calculated relative to vehicle controls. All experiments were repeated at least three times. Data shown are the average of triplicate determinations in a representative experiment. Values were calculated relative to vehicle control and presented as ±SEM.

Protein extraction and Western blot analysis

HepG2 or HEK293 cells were transiently transfected with WT or Q mutant ER expression vectors for 28 to 30 hours. Whole-cell lysates were prepared by using the BD TransFactor Extraction Kit (BD Biosciences). For Western blot, the fresh-made cell lysate (40 μg) was loaded on an SDS–polyacrylamide gel electrophoresis gel and separated by electrophoresis. The proteins were electrotransferred onto nitrocellulose membranes, and membranes were subsequently blocked in phosphate-buffered saline (PBS) with 5% nonfat milk for 2 hours. The blots were incubated with primary antibody (hERα, clone HC-20, catalog no. sc-543, Santa Cruz Biotechnology Inc.) diluted 1:500 at 4°C overnight, washed with PBS with 0.1% Tween 20 (PBS-T), and then incubated with anti-rabbit IRDye 800CW secondary antibody (catalog no. 926–32211, LI-COR Biosciences) at room temperature for 1 hour. The immunoreactive products were detected by Fc ODYSSEY image system (LI-COR Biosciences). Anti–β-ACTIN was used as a loading control.

Ligand binding affinity assay

Scatchard analysis and competitive radiometric binding assays were performed on 96-well microtiter filter plates (Millipore), using the six ERα LBD genotypes expressed in Escherichia coli and purified with a his-tag, with tritiated E₂ as tracer, as previously described (52, 53). After incubation on ice for 18 to 24 hours, ERα-bound tracer was absorbed onto hydroxyapatite (Bio-Rad), washed with buffer, and measured by scintillation counting. RBA values are the average ± SD of two to three determinations.

Coactivator binding affinity assay

SRC1 or SRC3 NR interaction domains (ca. 200 amino acids) were titrated into a fixed amount of ERα-LBD-biotin preparations of the six 375 genotypes (1 nM) and mixed with 25 nM streptavidin-terbium (Invitrogen, Grand Island, NY) and 25 μM of ligand, on 96-well black microplates (Molecular Devices, Sunnyvale, CA) following previously determined methods (23, 53). The time-resolved FRET measurements were performed with a Victor X5 plate reader (Perkin Elmer, Shelton, CT). Diffusion-enhanced FRET was determined by a parallel incubation without biotinylated ER-LBD and subtracted as a background signal. The data, representing two to three replicate experiments, each with duplicate points, were analyzed using GraphPad Prism 4 and are expressed as K_d ± SD in nM.

Production of lentivirus and HepG2 stable cell lines

All lentiviruses were packaged in HEK293T/17 cells according to published protocols (54). Briefly, 293T cells were transiently transfected with psPAX2, MD2.G, and pDEST673 vector, pDEST673/ WT ERα, or pDEST673/Q375H mutant carrying the neomycin resistance gene using Lipofectamine 2000. Supernatant was collected 48 hours after transfection and concentrated by centrifugation at 50,000g for 2 hours over a 20% sucrose cushion. Pellets were resuspended in PBS and used for infection. Titers were determined using qPCR to measure the number of lentiviral particles integrated into the transduced HEK293T genome. Multiplicity of infection (MOI) ranging from 25 to 180 was used for infection of HepG2 cells. After 3 days of infection, cells were selected with geneticin (1.2 mg/ml, Invitrogen, #11811–031) and a stably pooled population of cells was obtained after 2 weeks. Stable integration of ERα protein was detected by Western blot (fig. S2A). Despite our best efforts, we were only able to isolate HepG2/Q375H tranfectants that expressed higher levels of mutant ERα than WT ERα.

Treatment, RNA isolation, and qPCR

HepG2 stable cells (HepG2/vector, HepG2/WT ERα, and HepG2/Q375H lines) were seeded in six-well plates at 5 × 10⁵ cells per well in phenol red–free MEM with 10% FBS overnight. The cells were then starved in phenol red–free MEM with 5% sFBS for 48 hours and then treated with vehicle control (<0.1% DMSO) and 1 or 10 nM E₂ for 18 hours. Total RNA was extracted from HepG2 cells using the RNeasy Mini Kit (Qiagen). First-strand cDNA synthesis was performed using SuperScript reverse transcriptase according to the manufacturer’s protocol (Invitrogen). The mRNA levels of ESR1 and ER-target genes (GREB1, TFF1, and WISP2) were measured using SYBR green assays (Applied Biosystems). Cycle threshold (C_t) values were obtained using the ABI PRISM 7900 Sequence Detection System and analysis software (Applied Biosystems). Each sample was normalized to human β-ACTIN (ACTB) expression, and fold changes were calculated relative to the HepG2/vector vehicle control C_t values. The sequences of qPCR primers have been described previously (26).

Treatments and RIME analysis

HepG2 stable cells (HepG2/WT ERα and HepG2/Q375H lines) were seeded in 150 mm × 25 mm dishes at 1 × 10⁷ cells per dish (total 10 dishes for each sample) in phenol red–free MEM with 10% FBS overnight. The cells were starved in phenol red–free MEM with 5% sFBS for 48 hours and then treated with vehicle control (<0.1% DMSO) and 10 nM E₂ for 18 hours. Cell fixation was performed following Active Motif’s protocol. Briefly, cells were fixed with ¹/₁₀ volume formaldehyde solution [16% methanol-free formaldehyde, 0.1 M NaCl, 1 mM EDTA (pH 8.0), and 50 mM Hepes (pH 7.9)]. Cell plates were placed on an orbital shaker and agitated at medium speed at room temperature for 8 min. After the 8-min agitation, ¹/₂₀ volume 2.5 M glycine solution was added, and the plate was incubated at room temperature for 5 min. After centrifugation at 800g for 10 min at 4°C, the cell pellets were immediately resuspended in ice-cold 0.5% Igepal-PBS. After centrifugation at 800g for 10 min at 4°C, cell pellets were transferred to a 1.5-ml tube. The cells were immediately stored in −80°C. A total of four groups of samples (WT Veh, WT E₂, Q375H Veh, and Q375H E₂) was submitted to Active Motif for RIME analysis. RIME was carried out using an antibody against ERα (Millipore, catalog no. 06–935) and 150 μg of chromatin from HepG2 stable cells to identify proteins that interact with ERα using mass spectrometry. For the data analysis, duplicate run samples (R1 and R2) for each group were performed for completion of the RIME analysis (data file S1). The common enriched protein lists were used in further analysis. These protein lists were uploaded into Partek Genomics Suite 7.0 for comparing the overlap proteins among the four group samples.

Cell lysis preparation and MARCoNI assay

HEK293 cells were seeded in 100 mm × 20 mm dish at 2 × 10⁷ cells per dish overnight. A total of 2 μg of WT ERα or Q375 mutant expression plasmids was transfected into the cells using Effectene transfection reagent (Qiagen). The expression levels of the six ERα genotype proteins were detected by Western blot (fig. S3). After 36 hours, cell pellets were collected and suspended in ice-cold lysis buffer (32, 33, 51) supplemented with protease inhibitor cocktail (Pierce Biotechnology) (150 μl per 1 × 10⁷ cells). The cell suspension was further snap-frozen, lysed, and centrifuged to obtain clear supernatant for the analyses later.

Crude lysate of ERα genotype–expressing cells was incubated with or without ligand on a PamChip array with 154 coregulator motifs (PamGene, #88101). Binding was detected using fluorescent ERα antibody (45) and quantified as arbitrary unit’s fluorescence using BioNavigator Software (PamGene). Each combination of genotype and ligand was tested using three arrays and thus resulted in three binding values for each peptide. These triplicates were used to calculate a mean binding and log₁₀-fold change of binding (MI) versus WT apo (unliganded) ERα (control) and to determine the significance of this modulation by Student’s t test with post hoc false discovery rate correction. Coregulator motifs that did not display significant binding or significant modulation versus control (P > 0.05) were removed from the dataset, thus reducing the number to 93 coregulator motifs. (Dis)similarities in receptor conformation relative to apo WT ERα, each reflected by a 93-point MI profile, were visualized by hierarchical clustering using Euclidean distance and Ward’s clustering. Similarity between coregulator motifs was determined similarly. Results are shown in the heatmap in fig. S1. A red color reflects genotype/ligand-induced enhancement of binding versus WT ERα, while blue indicates decreased binding. Significance of the genotype/ligand effect is indicated by asterisks (*P < 0.05, **P < 0.01, and ***P <0.001). For data and statistical analysis, BioNavigator, Microsoft Excel (version 14.0.7106.5003; Microsoft Corporation), and R [V2.15.2, R Core Team (2013), a language and environment for statistical computing (Vienna, Austria; www.R-project.org/)] were used.

MD modeling

All molecular systems were prepared and simulated according to previously described protocols (53) and are only briefly described here. Atomic models of ERα complexed to E₂ and NCOA2 (SRC2) peptide were downloaded from the Protein Data Bank (PDB) (PDB code: 1GWR) (55). Missing atoms (loops, side chains, and explicit hydrogens) were added using the protein preparation module of the Molecular Operating Environment (MOE) (56). All subsequent preparatory steps were performed in VMD (57). The molecular topology was constructed using PSFGen; neutral caps were applied to the N and C termini (N-methyl amido and acetyl, respectively), and point mutations (i.e., Q375H, both HSD and HSE tautomers) were introduced using the “mutate” statement. The explicit water molecules were added using the Solvate plugin with 20-Å thickness surrounding the protein. Sodium and chloride ions were added via the Autoionize plugin to neutralize the system and yield a final salt concentration of 0.1 M.

All simulations were performed in NAMD2 (58), using the CHARMM family of force fields to describe the motions of protein (59, 60), E₂ (61–63), ions, and water (56). Simulation conditions included constant pressure [1.0 atm, Langevin piston (64): period, 100 fs; decay, 50 fs; damping coefficient, 0.5 ps⁻¹] and temperature (310 K, Nosé-Hoover thermostat). Periodic boundary conditions were used with nonbonded interactions truncated via a switching function from 10.0 to 12.0 Å, computed at every step, and long-range electrostatics were evaluated using the particle mesh Ewald (PME) method (65), computed every other step. All molecular systems were first minimized, and a 1-ns simulation was constrained to equilibrate components modeled during system preparation (i.e., loops, mutated residues, solvent, and ions). Production simulations were performed under equilibrium conditions for a total of 100 ns each.

Simulation trajectories were analyzed in VMD. Hydrogen bonding distances (heteroatom-heteroatom) were measured between the side chain of residue 375 (as the donor) and backbone carbonyls of residues 367, 368, and 370 (as acceptors). The ensemble of conformations for the side-chain heavy atoms of residue 375 (Gln/HSD/HSE) and Lys³⁶² was computed for the last 50 ns of the simulation trajectory as occupancy maps using the VolMap plugin.

Statistical analysis

For Figs. 3 and 5A and fig. S2C, the two-way analysis of variance (ANOVA) with Tukey’s multiple comparison test (** ^{or c}P < 0.01, *** ^{or b}P < 0.001, or **** ^{or a}P < 0.0001) was performed using GraphPad Prism version 8.2.1. For Table 1 and table S2, the one-way ANOVA with Dunnett’s multiple comparison test (*P < 0.05 and **P < 0.01) was performed using GraphPad Prism version 8.2.1. For Table 2 and table S2, the two-way ANOVA with Dunnett’s multiple comparison test, P < 0.0001 for the entire interaction, as well as for the column and row factors, was performed using GraphPad Prism version 7.04; the P values are noted in the legends to these tables.

Supplementary Material

Supplementary figures

Fig. S1. Western blot for WT ERα and the five Gln³⁷⁵ mutants in HepG2 cells after transient transfection with expression plasmids.

Fig. S2. Generating HepG2 cells stably expressing WT ERα or ERα-Q375H.

Fig. S3. Western blot for WT and mutant ERαs in HEK293 cells.

Fig. S4. Global heatmap of the effect of mutation and/or agonist binding on receptor conformation reflected by the MI values for the coregulator interactions.

Fig. S5. Corepressor peptides bound to the mutant receptor-ligand complex stronger than to the WT ERα-ligand complex.

Fig. S6. Genotype profiles showing the corepressor and coactivator interactions of WT and ERα-Q375H.

Table S1. The EC₅₀ values for transcriptional activation of WT ERα and the five Q375 mutants by the six ligands.

Table S2. K_d values for SRC3 binding to liganded WT and mutant ERα.

Table S3. Comparison of the relative ligand-independent interactions of peptides with WT ERα and ERα-Q375H.

Table S4. Comparison of the relative ligand-dependent interactions of peptides with WT ERα and ERα-Q375H.

Table S5. Comparison of binding values for peptides that bound more strongly to ERα-Q375H than to WT ERα across all five Gln³⁷⁵ mutants.

Table S6. Comparison of binding values for peptides that bound more strongly to WT ERα than to ERα-Q375H across all five Gln³⁷⁵ mutants.

Table S7. Summary of the peptides representing eight selected coregulators.

Supplementary data file S1

Data file S1. RIME analysis datasets for the proteins interacting with WT ERα and ERα-Q375H.

Movie S1

Movie S1. ERα LBD, with Gln³⁷⁵, E₂ ligand, and other features highlighted.