ESR1 Mutations Associated With Estrogen Insensitivity Syndrome Change Conformation of Ligand-Receptor Complex and Altered Transcriptome Profile

Abstract

Estrogen insensitivity syndrome (EIS) arises from rare mutations in estrogen receptor-α (ERα, encoded by ESR1 gene) resulting in the inability of estrogen to exert its biological effects. Due to its rarity, mutations in ESR1 gene and the underlying molecular mechanisms of EIS have not been thoroughly studied. Here, we investigate known ESR1 mutants, Q375H and R394H, associated with EIS patients using in vitro and in vivo systems. Comparison of the transcriptome and deoxyribonucleic acid methylome from stable cell lines of both Q375H and R394H clinical mutants shows a differential profile compared with wild-type ERα, resulting in loss of estrogen responsiveness. Molecular dynamic simulation shows that both ESR1 mutations change the ERα conformation of the ligand-receptor complexes. Furthermore, we generated a mouse model Esr1-Q harboring the human mutation using CRISPR/Cas9 genome editing. Female and male Esr1-Q mice are infertile and have similar phenotypes to αERKO mice. Overall phenotypes of the Esr1-Q mice correspond to those observed in the patient with Q375H. Finally, we explore the effects of a synthetic progestogen and a gonadotropin-releasing hormone inhibitor in the Esr1-Q mice for potentially reversing the impaired female reproductive tract function. These findings provide an important basis for understanding the molecular mechanistic consequences associated with EIS.

Clinical estrogen insensitivity syndrome (EIS) or estrogen resistance results from a defective estrogen receptor (1), most commonly caused by a mutation in the ERα protein (encoded by the ESR1 gene). These types of defects result in the inability of estrogen to exert its receptor-mediated biological effects (2). Estrogens are key regulators of cell growth, cell differentiation, and a variety of physiological functions in both males and females (3, 4).

EIS is a very rare disease as there have only been 3 recorded cases of patients with the ESR1 mutation, all arising from consanguineous relations. Human ESR1 R175X, a disruptive homozygous mutant, was the first case of EIS found in 1994 in a 28-year-old male with open epiphyses, low bone mineral density, and elevated serum estrogens and gonadotropins (2). Recently, a homozygous missense ERα mutant, ESR1 Q375H, was identified in an 18-year-old female that led to profound estrogen insensitivity, resulting in delayed puberty, primary amenorrhea, continued linear growth, incomplete epiphyseal closure into adulthood, low bone mineral density, lack of breast development, enlarged bilateral multicystic ovaries, and hypoplastic uterus with no endometrial stripe (5). More recently, a familial homozygous missense ERα mutant, ESR1 R394H, was identified in 2 sisters and 1 brother, originating from a consanguineous Algerian family (6). They also had delayed puberty and presented with delayed bone maturation, consistent with estrogen insensitivity, and both sisters had enlarged multicystic ovaries (6). Their clinical phenotypes were similar to those of the ERα knock-out (αERKO) mouse model previously described (7, 8). Female αERKO mice are infertile and anovulatory, resulting in multicystic ovaries without corpora lutea, hypoplastic uteri, and no pubertal mammary gland development (9).

The biological effects of estrogens are mediated through ERα or ERβ that belong to the nuclear receptor superfamily of ligand-inducible transcription factors (7, 10-12). The major mechanisms of ER-mediated transcriptional gene regulation are either ligand-bound ER directly interacting with deoxyribonucleic acid (DNA) estrogen response elements in the regulatory regions of the ER target genes (genomic mechanism) or the ER interacting with other transcription factors/coregulators to regulate gene expression (tethered mechanism) (13, 14). Like other nuclear receptors, ERα has a distinct domain structure, including the A/B domain (the hormone-independent activation function domain), C domain (DNA-binding domain), D domain (a hinge region with nuclear localization sequences domain), and the E/F domain (the ligand-binding domain, LBD, a hormone-dependent activation function domain) (15). Generally, coactivator proteins such as the steroid receptor coactivator family interact with ERα through a conserved hydrophobic cleft comprised of a short motif (LxxLL motif) on the surface of the LBD (16). Structurally, the ESR1 Q375H mutation is located in the coactivator LxxLL binding surface (5), and R394H is located in the ligand-binding pocket of the ERα-LBD (6). Mechanistically, both locations control gene regulation and the ERα function. Recently, we reported ERα mediates DNA methylation and transcriptome aberrations in mouse tissues following developmental diethylstilbestrol (DES, a synthetic estrogen) exposure, and these findings suggest ERα can play a role in epigenetic reprograming (17).

To test the functionality of the ESR1 Q375H or R394H mutations, we integrated transcriptome and DNA methylome profiles to characterize the effects of estrogen responsiveness using an in vitro cell system. We evaluated the effects of the mutations on receptor protein and investigated whether the 2 natural mutants change the ERα conformation in the ligand complex. Furthermore, we generated a mouse model, Esr1-Q379H (Esr1-Q), which harbors the human ESR1 Q375H mutation, using the CRISPR/Cas9 genome-editing system and characterized resulting physiological effects in both female and male mice. Finally, we explored the effects of norethindrone (NET, a synthetic progestogen) and Firmagon (FMG, a gonadotropin-releasing hormone [GnRH] inhibitor) in the Esr1-Q mice for potentially testing the reversal of the impaired female reproductive tract function.

Materials and Methods

Reagents

17β-estradiol (E2) and diethylstilbestrol (DES) were purchased from Sigma-Aldrich. Norethindrone (NET) was purchased from Innovative Research of America. Firmagon (FMG)/degarelix acetate was purchased from Bachem. Mannitol was purchased from Millipore Sigma.

Cell line and tissue culture

The human breast cancer cell line MDA-MB-231 was purchased from American Type Culture Collection and maintained in phenol red-free Dulbecco's modified Eagle medium:F12 medium (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.

Plasmid constructions

The expression vector plasmid, pcDNA3-DEST-hERa (595 aa) was plasmid was generated by Celplor Inc (www.celplor.com) by cloning the complementary DNA into pENTR-3C (Invitrogen) at BamHI/EcoRI. The pcDNA3-DEST-hERα-Q375H (595 aa) and pcDNA3-DEST-hERα-R394H (595 aa) were generated by Celplor Inc (www.celplor.com) via polymerase chain reaction (PCR)-based site-directed mutagenesis [CAG > CAT] ((5) Q clinical study MS). These plasmids were cloned into pcDNA3-DEST by gateway cloning (LR reaction) at attR1 and attR2 and then cloned into the lenti viral expression vector pDEST-673 (National Institute of Environmental Health Sciences Protein Core).

Production of lentivirus and stable cell lines

All lentiviruses were packaged in HEK293T/17 cells according to published protocols (18). Briefly, 293T cells were transiently transfected with psPAX2, MD2.G, and pDEST673 carrying the neomycin-resistance gene and the desired ERα mutant using Lipofectamine 2000. Supernatant was collected 48 hours post transfection and concentrated by centrifugation at 50 000 X g for 2 hours over a 20% sucrose cushion. Pellets were resuspended in phosphate-buffered saline and used for infection. Titers were determined using quantitative PCR to measure the number of lentiviral particles integrated into the transduced HEK293T genome. The multiplicity of infection ranging from 25 to 180 was used for infection of MDA-MB-231 cells. After 3 days of infection, cells were selected with Geneticin (1.2 mg/mL, Invitrogen, #11811-031) and a stable pooled population of cells was obtained after 2 weeks. Stable integration of ERα was verified by Western blot (Supplementary Fig. 1). All supplementary material and figures are located in a digital research materials repository (19).

E2 or DES treatment and microarray assays

Cells were seeded in 60 × 15 mm dishes with 5% sFBS medium and starved for 2 days. After changing to fresh 5% sFBS medium, cells were treated with vehicle control (dimethyl sulfoxide, final concentration < 0.01%), 100 nM E2, or DES for 18 hours. Gene expression analysis was conducted using Affymetrix Human Clariom D arrays (Affymetrix, Santa Clara, CA). Three hundred nanograms (300 ng) of total ribonucleic acid (RNA) were amplified and labeled as directed using the Affymetrix WT Plus Reagent Kit (WT Plus Kit). A total of 5.5 μg of amplified biotin-cDNAs was fragmented and hybridized to each array for 16 hours at 45°C in a rotating hybridization oven. Array slides were stained with streptavidin/phycoerythrin utilizing a double-antibody staining procedure and then washed for antibody amplification according to the GeneChip Hybridization, Wash, and Stain Kit and user manual following protocol FS450-0001. Arrays were scanned in an Affymetrix Scanner 3000, and data were obtained using the GeneChip Command Console software.

DNA extraction and bisulfite treatment and EPIC BeadChip DNA methylation microarray assays

Genomic DNA (400-500 ng) was extracted from MB231 stable cells using a tissue blood kit (Qiagen, Valencia, CA) according to the manufacturer’s protocol. Bisulfite treatment was performed using the EZ DNA Methylation-Gold Kit (Zymo Research, Irvine, CA) following the manufacturer’s instructions. DNA methylation analysis was conducted using Infinium Methylation EPIC BeadChip arrays (Illumina, San Diego, CA) following the Illumina InfiniumHD methylation protocol. Starting with 200 ng of bisulfite-converted DNA, samples were amplified for 22 hours at 37°C using random priming. The amplified DNA was then fragmented, precipitated, and resuspended prior to hybridization to BeadChips for 17 hours at 48°C. The slides were washed and stained as per the standard Illumina HD methylation protocol. Finally, the arrays were scanned with an Illumina iScan.

Microarray data analysis

For GeneChip transcriptome analysis, the CEL files generated from GeneChip were used to identify differentially expressed genes (DEGs) by the Genomics Suite Gene Expression workflow of Partek software package version 7.17 (Partek Inc., St. Louis, MO, USA). The Robust Multichip Analysis algorithm with quantile for normalization and log2 transformation was applied to generate signal values of all samples. DEGs were defined using the filters of analysis of variance unadjusted P value <0.05 and absolute fold change >1.5. All experiments were performed in triplicate with independent pools of RNA. Transcriptome array data have been deposited in the Gene Expression Omnibus (GEO, accession GSE139489). Hierarchical clustering and heatmap of DEGs were generated by the Genomics Suite of Partek software package version 7.17 (Partek Inc., St. Louis, MO, USA). The expression values of each gene were shifting to mean of zero and scaled to standard deviation of 1. Functional analysis of DEGs was performed using the Ingenuity Pathway Analysis (IPA, www.ingenuity.com) based on the content of 2019-06-16 version 48207413. For a given biological category in IPA, the Fisher exact test was used to measure the probability (P value) that the category was randomly associated. The categories with P values < 0.05 were defined as significantly enriched.

For EPIC BeadChip DNA methylome analysis, the methylation Pipeline (ChAMP) (version 2.10.1) was applied to analyze the Infinium Human Methylation EPIC BeadChip (Illumina) in R environment (R version 2.5.1). Raw IDAT data were first loaded into R objects using the scripts adapted from the ChAMP pipeline (20). Probes with the detected P value >0.01 in more than 1 sample or a bead count < 3 in more than 5% of samples were considered poor quality and removed from further analysis. All the non-CpG probes, nonspecific probes, single nucleotide polymorphism-related probes and probes located in sex chromosomes were also excluded from downstream analysis. One sample with >10% failed probes was also removed from the analysis. As a result, there were a total 649 726 CpG probes included in the downstream analyses. Normalization was performed using the beta mixture quantile dilation approach (21), and batch effect was adjusted using the ComBat method (22). DNA methylation array data have been deposited in the Gene Expression Omnibus (GEO, accession GSE139687). To test the differential methylation probes, the probe-wise differential methylation analysis was carried out using the limma algorithm (23), which computes the P value for differential methylation using a linear model. The P value of <0.01 was used to define probes with significantly differential methylation. The differential methylation regions (DMRs) were called using the bumphunter method (24) with default parameters. The bumphunter algorithm uses smoothed methylation levels to detect DMRs. It first groups all probes into small clusters (or regions) and then applies a random permutation method to estimate candidate DMRs. The DMRs were considered significant if they reached a minimal false discovery rate value < 0.05.

Molecular dynamics simulations

Using the x-ray crystal structures of E2/ERα (pdb ID 3UUD (25), DES/ER (25) ID 4ZN7 (26)), and inactive ERα (pdb ID 1A52 (27), starting structures of the ligand-binding domains of ERα were created for molecular dynamics (MDs). Missing residues (462-464 in 3UUD and 462-471 in 4ZN7) were introduced using the program Modeller (version 9.14) (28) and missing side chains (305, 306, 397, 467-477, 492, 531, and 545 in 3UUD; 314, 472, 531 in 4ZN7) were reconstructed using the program Coot (version 0.8.2) (29). Q375H and R394H mutations were introduced using Coot. Fifteen different systems were considered (E2/ERα, DES/ERα, inactive ERα, E2/inactive ERα, DES/inactive ERα, and each of the 5 with Q375H or R394H mutation). Protons were added, followed by the addition of counter ions, and the resulting complex structures were then solvated in a box of water for each system. The amount of water molecules varied from 26 958 to 28 898 for these systems. Prior to equilibration, each system was subjected to the following steps: (1) more than a nanosecond of constrained MDs with the peptide and the ligand constrained to the original position with a force constant of 10 kcal/mol/nm, (2) minimization, (3) low-temperature, constant-pressure MD simulation to assure a reasonable starting density, (4) minimization, (5) stepwise heating MD at constant volume, and (6) a constant-volume MD run for another 9 nanoseconds. All final unconstrained trajectories were calculated at 300 K under constant volume MD (300 ns total dynamics with 1 fs time step) using PMEMD from the Amber 16 program (30) to accommodate long-range interactions. The amino acid parameters were taken from the Amber SB14 force field. The charges of atomic positions of the ligands were derived using the ChelpG scheme with the 6-31g(d) basis, set at the B3LYP level using the program Gaussian 09.D01 (31). The other ligand force field parameters were generic parameters listed in the Amber force field. Using molecular mechanics/ poisson-boltzmann surface area (MM/PBSA) protocol (32) of Amber, binding free energies of E2 and DES were calculated for the 100 configurations extracted from the last 100 ns of each trajectory. The ionic strength for MM/PBSA calculations was selected to be 0.1M.

Animal and CRISPR/Cas9 genome editing for generating Esr1-Q mice

All animal studies were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. The mouse Esr1 Q379H locus was generated via the CRISPR/Cas9 genome-editing system. Briefly, cytoplasmic microinjection of single-cell C57BL/6J embryos with Cas9 messenger RNA (mRNA) (200 ng/µL; 7-methylguanosine cap and poly[A] tail), locus-specific Cas9 scaffold guide RNA (20 ng/µL; CGAGAAGGTGGACCTGATCANGG), and single-stranded 200 bp oligo (100 ng/µL) served as a homology-directed repair template (IDTDNA, Ultramer). The repair template contained a 95 bp 5' homology arm and 92 bp 3' homology arm with a genetic payload of 13 bp that contained the Q379H mutation and 4 synonymous mutations in the flanking codons (WT: T GAT CAG GTC CAC; mutant: c GAc CAt GTg CAt). The founders (first generation mice) were screened by direct PCR amplicon sequencing (777 bp amplicon; fwd: 5'-AGACAGTTAAGTGGGTTGTCTGA-3'; Rev: 5'-CGTGACTGATTGTAAAGGCACTA-3'). ESR1*Q379H mutation-positive founders were bred to C57BL/6J mice, and the germline transmission of the ESR1*Q379H mutation sequence was confirmed in the F1 offspring.

Tissue protein extraction and Western blot

Cytoplasmic and nuclear fractions of samples were extracted from frozen tissues using Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher) as per the manufacturer's protocol. Proteins from each of the fractions collected were quantified using a protein assay on a Qubit fluorometer (Invitrogen). Ten micrograms of cytoplasmic and nuclear-extracted proteins were loaded into a premade 4% to 20% gradient acrylamide gel (Bio-Rad) and run for 40 minutes at constant 200 V. Gels were transferred onto nitrocellulose membranes using an iBlot transfer system (Invitrogen). Blots were blocked using 5% milk in tris-buffered saline and polysorbate 20 (TBS-T) for an hour at room temperature. After blocking, the blots were incubated overnight with primary ERα antibody (RRID:AB_631470) (33) or glyceraldehyde 3-phosphate dehydrogenase antibody (RRID:AB_10167668) (34) in 5% milk/TBS-T at 4°C. The blots were washed using TBS-T and then incubated with IRdye 800CW goat, anti-rabbit (RRID:AB_2782998 or RRID:AB_621843) (35, 36) in 5% milk/TBS-T buffer for 1 hour at room temperature. The blots were washed and the signals imaged using the Li-Cor Odyssey Fc system. ERα protein expression was quantified using the image studio 5.2 program. The percentage of the band intensity was compared with glyceraldehyde 3-phosphate dehydrogenase

RNA extractions and qPCR assays

Total RNA samples were extracted from frozen tissues of individual mice using Trizol or from cells using RNeasy Mini Kit (Qiagen). For qPCR, first-strand complementary DNA synthesis was performed using Superscript reverse transcriptase (Invitrogen). The mRNA levels of genes were measured using SYBR green assays (Applied Biosystems). Cycle threshold values were obtained using the ABI PRISM 7900 Sequence Detection System and analysis software. The experiments were repeated 3 times, and results are presented as the fold increase calculated relative to the vehicle of WT ERα ± standard error (SEM).

Serum steroid hormone assays

All hormone assays were carried out on serum from individual animals. Serum luteinizing hormone (LH) levels were measured in singlicate from a 20-μL sample per animal using the sensitive dissociation-enhanced lanthanide fluoroimmunoassay, a modified method of sandwich enzyme-linked immunosorbent assay. Briefly, the LH in the sample was captured by the biotinylated anti-mouse LH antibody then trapped on the streptavidincoated plate; the captured LH on the plate was detected by the lanthanide-labeled anti-mouse LH antibody. The anti-mouse LH antibody used for dissociation-enhanced lanthanide fluoroimmunoassay was provided by Dr Deborah Best (Environmental

Protection Agency, Durham, NC). Serum E2 levels were analyzed by the Ligand Assay and Analysis core at the University of Virginia School of Medicine (Charlottesville, VA).

Hematoxylin and eosin staining

Mouse tissues were placed immediately into 10% neutral-buffered formalin after euthanasia. Tissues were transferred to 70% ethanol and then processed, and paraffin was embedded once the solution remained clear in 70% (w/v) ethanol for several hours. Tissue was sectioned (7 µm) and placed onto charged-glass microscope slides and stained with hematoxylin and eosin for histopathology analysis.

Uterine bioassay in ovariectomized adult mice

Adult female mice were ovariectomized at approximately 10 weeks of age and housed for 2 weeks to eliminate endogenous ovarian steroids before the study, and then 12-week-old female mice were treated with oil (vehicle) or 100 μg/kg DES in oil daily for 3 days. On the day after the last injection, necropsies were performed, and reproductive tract tissue was collected.

Treatment of NET and FMG/degarelix acetate

For NET treatment, 4-week-old female mice were treated with placebo or 1.5 mg NET for 3 weeks by subcutaneous injection. All tissues were collected at week 7. For FMG treatment, 4-week-old female mice were injected with 100 mL saline containing 60 μg FMG and 70 μg mannitol and then injected with 100 μL saline containing 30 μg FMG and 35 μg mannitol after 2 weeks. All tissues were collected at week 8. The vehicle group was treated with 100 μL saline containing 70 μg mannitol.

Ovarian cyst quantification

Five sections of one ovary for each sample were stained by hematoxylin and eosin, scanned, and then uploaded to eSlide Manager system. The slide scans were randomized, stripped of identifiers, and total number of ovarian cysts were counted per slide. The samples were then matched to their respective cyst counts and sorted based on genotype of sample and treatment. The data was transferred to GraphPad Prism 7 software where statistical significance was calculated.

Results

ESR1 mutants alter basal transcription and exhibit mutation-specific loss of estrogen signaling

To understand the mechanism underlying the dysfunctional ESR1 mutants, E375H and R394H, from patients with EIS, we tested in an ERα– human breast cancer cell line, MB-231 cells, and we generated cells stably expressing either WT ERα, Q375H- or R394H-mutant ERα (Supplementary Fig. S1) (19). Whole transcriptome expression was profiled using GeneChip arrays. The differentially expressed probes were analyzed by examining fold changes (±1.5 fold cutoff) from the 9 group comparisons (Q-veh vs WT-veh, R-veh vs WT-veh, Q-veh vs R-veh, WT-E2 vs WT-veh, Q-E2 vs Q-veh, R-E2 vs R-veh, WT-DES vs WT-veh, Q-DES vs Q-veh, and R-DES vs R-veh) for whole transcriptome, noncoding RNA, micro RNA, and small RNA summarized in Supplementary Fig. S2 (19). We found that both Q375H and R394H mutants altered the whole transcriptome at the basal level (Supplementary Fig. S2A) (19). This was also the case when specific types of RNA were compared independently (noncoding RNA, micro RNA or small RNA, Supplementary Fig. S2A, B, and C) (19). For coding genes, a summary of the number of DEGs that were mapped from coding differentially expressed probes is shown in Supplementary Table S1 (19). The GeneChip data were validated by quantitative PCR (qPCR) analysis for the 3 comparisons (Supplementary Table S2) (19). A comparison of the basal expression changes between Q375H vs WT ERα or R394H vs WT ERα are shown in Fig. 1. There were more DEGs identified from both mutants samples than in WT (Fig. 1A, comparisons 1 and 2). Interestingly, there were also DEGs identified between the activities of the Q and R mutants (Fig. 1A, comparison 3). IPA analysis provided identification of the top 5 diseases and biological functions and upstream regulators between Q375H vs WT and R394H vs WT (Fig. 1B).

Figure 1.

Transcriptome of the basal levels in MB-231 cells stably expressing WT ERα, Q375H, and R394H mutations. (A) Comparisons of the basal levels. (B) IPA analysis of Q or R DEGs in comparing with WT ERα basal levels. Top biological functions and upstream regulators were reported. DEG, differentially expressed gene; ERα, estrogen receptor-a; IPA, Ingenuity Pathway Analysis; Q, glutamine; R, WT, wild type.

To determine if ligand responses differed between the Q375H and R394H mutants, we analyzed E2-mediated and DES-mediated mRNA transcriptome profiles. The profiles showed that there were more than 1000 DEGs in all 6 comparisons (Fig. 2, E2 treatment and Fig. 3, DES treatment). However, only about 10% or less of the DEGs overlapped between WT ERα and Q375H or WT and R394H. These findings indicate that both Q375H and R394H ESR1 mutations result in alterations in the WT ERα basal expression levels as seen from the whole transcriptome and result in loss of estrogen signaling linked to reproductive-related diseases.

Figure 2.

E2- mediated transcriptional aberrance in MB-231 cells stably expressing WT ERα, Q375H, and R394H mutations. (A) Venn diagrams of the comparison of E2-altered DEGs. (B) Overlapping E2-altered DEGs. (C) Top 5 diseases and function, and upregulators of E2-altered DEGs. DEG, differentially expressed gene; E2, estradiol; ERα, estrogen receptor-α; WT, wild type.

Figure 3.

DES-mediated transcriptional aberrance in MB-231 cells stably expressing WT ERα, Q375H, and R394H mutant. (A) Venn diagrams of the comparison of DES-altered DEGs. (B) Overlapping DES-altered DEGs. (C) Top 5 diseases and function, and upregulators of DES-altered DEGs. DEG, differentially expressed gene; DES, diethylstilbestrol; E2, estradiol; ERα, estrogen receptor-α; WT, wild type.

ESR1 mutants differentially alter the DNA methylome

To further investigate the mechanism impacting aberrant gene expression associated with the natural mutants ESR1 Q375H and R394H, DNA methylation EPIC BeadChip analysis was performed to evaluate if any differential profiles existed in the same stable cell systems used in the GeneChip array. A P value < 0.01 was used to define probes with significantly DMRs with a false discovery rate < 0.05. The 6 comparisons of DMRs are summarized in Supplementary Fig. S3 (19). We found that both theQ375H and R394H mutants had different DNA methylome profiles in the basal levels (comparison 1 and comparison 2) when compared with WT. The DMRs were also found in the basal levels between R and Q mutants (comparison 3). E2 altered the DNA methylome patterns in WT, Q, and R mutants (comparison s 4-6, Supplementary Fig. S3A) (19). A heatmap combining the differential methylation probes of all 3 phenotypes after E2 treatments is shown in Supplementary Fig. S3B (19). These results suggest that ESR1 mutations can result in alteration of the DNA methylome in cells, which could be one reason for the changes detected in the transcriptome resulting in varying degrees for estrogen responsiveness.

ESR1 mutations change the ERα structural conformation of the ligand-receptor complex

As depicted in Figure 4A, the residue Q375 is located in the N-terminal segment of helix3, which, along with helix5 and helix12, forms the coregulator binding surface. In contrast, residue R394 is involved in a strong salt bridge interaction with E353 located within the base of the ligand-binding pocket, and the salt bridge is in direct contact with the phenolic hydroxyl group of E2 or DES. MD simulations have extensively been used in characterization of mutant proteins and especially in the cases where mutant residues are in functionally important regions such as ligand-binding sites or protein-interacting surfaces. Since from each mutant and the resulting experimental and clinical phenotypes under the present conditions there appear to be in a functionally important location, MD analysis should result in projecting conformations that can be used to further characterize functionality.

Figure 4.

The locations of Q375H and R394H mutants in ERα-LBD and dynamics cross-correlation maps. (A) Locations of Q375H and R394H mutants. (B) Dynamics cross-correlation map for ligand free. (C) Dynamics cross-correlation map for E2-bound. (D) Dynamics cross-correlation map of DES-bound. The top triangle of each figure corresponds to the mutant system and the bottom triangle to WT ERα system. DES, diethylstilbestrol; ERα, estrogen receptor-α; E2, estradiol; LBD, ligand-binding domain; WT, wild type.

In order to characterize the effects of the Q375H and R394H mutations on ligand binding, we performed several MD trajectory calculations on various ERα conformations of the mutants and compared these with WT ERα. To establish reference configurations in their dynamical states of WT ERα, forms of inactive ligand-free as well as E2- and DES-bound agonist conformations were subjected to solution simulations. For comparison, both Q375H and R394H mutants were also simulated in the same inactive conformation as the wild type with the E2- and DES-bound agonist conformations. Time-dependent root-mean-square deviations (RMSDs) of ERα quantify how much each configuration along the calculated MD trajectory deviates from the starting crystal conformations. Overall features of RMSD values can be used to show relatively stable ER structures in all cases considered (ligand-free and ligand-bound as well as wild-type and mutants, Supplementary Fig. S4) (19). RMSD values of the ligand-free systems show significant variations when helix12 is included in the calculation. Once this region that has not yet adapted the agonist or antagonist conformation is removed from the calculations, the RMSD values fall back to the normal range of a stable protein conformation. Such calculations would reflect the fact that helix12 is highly mobile in all 3 unliganded systems. Note that the RMSDs are global variables, and they can only predict whether the protein conformations remain stable or not during dynamics. RMSDs do not reflect any local structural changes that can be significant impediments to the function. Local structural variations can be estimated using root-mean-square fluctuations (RMSF) of individual residues of ERα and currently, the values displayed are calculated using the structures collected over the last 100 ns of each simulation. Large RMSF values (Supplementary Fig. S5) (19) in the unstructured segment between helix2 and helix3 are common to all simulations. Large RMSFs are observed when both E2 or DES bound with the Q375H mutant in and around the mutated site. That region is part of the helix triad (helix 3, 5, and 12) that becomes the base for coregulator interactions. Such variations can virtually modulate the binding around this area, thereby impacting gene regulation. Similar fluctuations in this region are also observed in the R394H system. Even if both E2 and DES act as agonists for ERα, it is clearly observable from our simulation results that surface residue fluctuations in various regions may vary in the mutant ERα association. In addition, the binding free energies of E2 and DES shown in Table 2 were calculated using the MM/PBSA protocol, and both E2 and DES ligands consistently show higher binding free energies (> 2 kcal/mol in each case of ligand) when interacting with WT ERα compared with the ERα mutants.

Table 2.

Binding Free Energies (in kcal/mol) and Their Components of EST and DES on ER and Its Mutants in Inactive and Active Conformations

	∆Ggas	∆Gsolv	∆Gtot
ER inactive + E2	–50.4 ± 3.7	44.4 ± 3.4	–6.0 ± 3.8
ER inactive (Q375H) + E2	–49.6 ± 2.7	45.1 ± 2.7	–4.5 ± 3.4
ER inactive (R394H) + E2	–47.4 ± 3.3	44.2 ± 3.3	–3.2 ± 3.1
ER active + E2	–51.0 ± 2.2	44.5 ± 3.5	–6.5 ± 3.1
ER active (Q375H) + E2	–50.2 ± 3.2	46.0 ± 3.2	–4.2 ± 3.9
ER active (R394H) + E2	–46.7 ± 2.4	43.3 ± 2.5	–3.4 ± 2.6
ER inactive + DES	–51.4 ± 3.2	41.4 ± 4 2.6	–10.0 ± 3.4
ER inactive (Q375H) + DES	–48.0 ± 5.9	40.9 ± 4.6	–7.1 ± 4.0
ER inactive (R394H) + DES	–49.9 ± 5.3	43.0 ± 4.4	–6.9 ± 4.4
ER active + DES	–50.9 ± 1.5	40.3 ± 2.9	–10.6 ± 3.2
ER active (Q375H) + DES	–49.6 ± 3.1	42.6 ± 4.5	7.0 ± 5.1
ER active (R394H) + DES	–49.1 ± 2.2	41.8 ± 2.0	7.3 ± 2.4

∆Gtot is estimated as the sum of its gas-phase energy, ∆Ggas and the solvation-free energy, ∆Gsolv.

Abbreviations: DES, diethylstilbestrol; E2, estradiol; ER, estrogen receptor.

The dynamic cross-correlation maps (DCCM) reveal dynamic correlative motions between any residue pair that occur during the last 100 ns of MD trajectories and are shown in Figures 4B, 4C, and 4D). Positive values correspond to residue pairs moving in the same direction, and the negative values correspond to residue pairs moving in the opposite directions. In each figure, out of the 2 triangles created by the diagonal passing through the bottom-left corner and the top-right corner, the bottom triangles correspond to the reference DCCMs of WT ERα systems (Fig. 4B, ligand-free ERα; 4C, E2-bound ERα; 4D, DES-bound ERα). DCCMs corresponding to Q375H and R394H mutants are in left and right sets, respectively. If the mutants exhibit similar correlations, one should notice a rather symmetric distribution around the diagonals of the triangles. Varying degrees of DCCM differences were observed when the WT and mutant systems were compared, and the region between residues 350 and 400 that marks coregulator interacting surfaces and their vicinity is specifically noted.

Esr1-Q379H (Esr1-Q) mice have similar phenotypes to αERKO mice and overall mimic the patient with EIS

To investigate the biological characterization of the natural human mutation in vivo, we generated a mouse model, Esr1-Q. In this mouse model, the glutamine (Q) at codon 397 corresponding to the human mutation was changed to a histidine to mimic the ESR1-Q375H mutation found in 1 of the clinical patients (Supplementary Fig. S6) (19). Both male and female Esr1-Q mice were characterized at age 6 months. In female Esr1-Q mice, ERα protein expression was detected in both cytosol and nuclear fractions extracted from uterine and ovarian tissue in adult mice (Fig. 5A). However, the expression level of the mutant ERα protein in the Esr1-Q mice was much lower than found in WT mice (Fig. 5A). There was no significant difference in uterine or ovarian ERα mRNA levels between WT and Esr1-Q mice (Fig. 5B). Female Esr1-Q mice are infertile and have similar phenotypes as the αERKO mice, including hypoplastic uteri and reduced uterine weight (Fig. 5C), hemorrhagic cystic ovaries (Fig. 5D), increased body weight (Fig. 5E), rudimentary mammary ducts (Supplementary Fig. S7A) (19), and elevated levels of LH and estradiol (E2) (Supplementary Fig. S7B) (19). In males, the Esr1-Q mice showed infertility, an increased body weight, and induction of seminal vesicle weight (Supplementary Fig. S8A) (19) as well as seminiferous tubule disruption in the testes, which were similar phenotypes to the αERKO male mice (Supplementary Fig. S8B) (19).

Figure 5.

The general characterization of female Esr1-Q mice. All tissues were collected from the adult mice (n = 5 or n=10). (A) ERα protein expression. Cytoplasmic and nuclear fractions of samples were extracted from frozen tissues using Nuclear and Cytoplasmic Extraction Reagents. Ten micrograms of cytoplasmic or nuclear extracted proteins were loaded into a 4% to 20% gradient acrylamide gel and run for 40 minutes. Gels were transferred onto nitrocellulose membranes. After blocking using 5% milk/TBS-T buffer for an hour, the blots were incubated overnight with primary ERα antibody (1:1000; MC-20) or GAPDH antibody (1:2000; LF-335) in 5% milk/TBS-T buffer at 4°C. The blots were washed incubated with IRdye 800CW goat anti-mouse or anti-rabbit fluorescent secondary antibody (1:10 000) in 5% milk/TBS-T buffer for an hour at room temperature. The signals were imaged using the Li-Cor Odyssey Fc system. (B) RNA expression. Total RNA samples were extracted from frozen tissues using Trizol. The mRNA levels of genes were measured using SYBR green assays. The experiments were repeated 3 times, and results are presented as the fold increase calculated relative to the vehicle of WT ERα ± SE. (C) Uterine weight. (D) Histology. Mouse tissues were placed immediately into 10% neutral-buffered formalin after euthanasia and then processed and paraffin was embedded once the solution remained clear in 70% (w/v) ethanol for several hours. Tissue was sectioned (7 µm) and placed onto charged-glass microscope slides and stained with hematoxylin and eosin for histopathology analysis. (E) Body weight. αERKO, ERα knock-out; cyto, cytoplasmic; ERα, estrogen receptor-α; ext, extraction; Nuc, nuclear; RNA, ribonucleic acid; SE, standard error; TBS-T, tris-buffered saline and polysorbate 20; UT, uterine; WT, wild type.

To test the responsiveness to estrogen in the Esr1-Q mice, a 3-day bioassay was performed using the synthetic estrogen, DES (Fig. 6A). WT uterine weight increased, and serum LH decreased after DES treatment compared with the vehicle; however, Esr1-Q uterine weight did not change with DES treatment (Fig. 6B and C). DES induced 3 well-characterized estrogen-responsive genes (Ltf, C3, and Greb1) in WT mice but not in the Esr1-Q mice, suggesting that this natural mutant lost ERα function (Fig. 6D). Overall, these results indicated that the Esr1-Q mice have similar phenotypes to αERKO mice and mimic the patient with EIS.

Figure 6.

The 3-day uterine bioassay with DES treatment in WT and Esr1-Q mice. (A) The timeline of the treatment and collection. (B) Histology of the uterus. (C) Uterine weight and serum LH level. The serum samples were collected from individual mice (n = 4 or n=5). Serum LH levels were measured in singlicate from a 20-μL sample per animal using the sensitive dissociation-enhanced lanthanide fluoroimmunoassay. (D) Changes of estrogen-responsive genes. Total RNA samples were extracted from frozen tissues using Trizol. The mRNA levels of genes were measured using SYBR green assays. The experiments were repeated 3 times, and results are presented as the fold increase calculated relative to the vehicle of WT ERα ± SE. DES, diethylstilbestrol; ERα, estrogen receptor-α; LH, luteinizing hormone; mRNA, messenger RNA, RNA, ribonucleic acid; SE, standard error; WT, wild type.

NET and FMG treatments reduced ovarian cysts in the Esr1-Q mice

A clinical study showed that NET, a synthetic progestogen treatment, can reduce ovarian cyst size in the patient with Q mutation(5). To investigate possible treatments to reverse the impaired female reproductive tracks in the Esr1-Q mice, we exposed WT and Esr1-Q mice to NET or the GnRH inhibitor, FMG. The Esr1-Q mice showed a significant reduction in the total number of cysts in response to the NET treatment (Fig. 7A, B, and C). In addition, NET treatment partially restored uterine weight in the Esr1-Q mice, although it did not reach the WT level (Fig. 7D). A similar effect was also observed when the Esr1-Q mice were treated with FMG (Supplementary Fig. S9) (19). Results using this model indicate that both NET and FGM had some potential to prevent ovarian cyst development.

Figure 7.

NET reduces the cyst numbers of ovaries in the Esr1-Q mice. (A) The timeline of the NET treatment and collection. (B) Histology of ovaries (n = 2). Mouse ovaries were placed immediately into 10% neutral-buffered formalin after euthanasia and then processed and paraffin was embedded once the solution remained clear in 70% (w/v) ethanol for several hours. Tissue was sectioned (7 µm) and placed onto charged-glass microscope slides and stained with hematoxylin and eosin (H&E) for histopathology analysis. (C) Qualification of the numbers of ovarian cysts and uterine weight (n = 4). Five sections of 1 ovary for each sample were stained by H&E, scanned, and then uploaded to the eSlide manager system. The slide scans were randomized, stripped of identifiers, and total number of ovarian cysts were counted per slide. (D) Uterine weight (n = 4). NET, norethindrone; WT, wild type.

Discussion

In this study, we characterized that the molecular mechanism of 2 clinical ESR1 mutations, Q375H (5) and R394H (6), associated with patients with EIS, results in impaired activity using in vivo and in vitro model systems. We showed that whole transcriptome and DNA methylome profiles using a stably transfected cell model show that both Q375H and R394H mutations result in aberrations in the whole transcriptome and DNA methylome, resulting in loss of normally altered ERα estrogen-response genes. ERα structural conformation analysis shows that both ESR1 mutants result in a change of the ERα conformation involving the ligand complexes. We generated Esr1-Q mice (a model harboring the human ESR1 Q375H mutation) and found that both female and male mice have similar phenotypes to αERKO mice, and the overall phenotypes of Esr1-Q mice are consistent with those of the patient with Q375H.

Estrogen signaling is an important mechanism for many biological functions and processes. Many studies show that uterine proliferation is regulated by estrogens such as E2 and DES in an ERα-dependent manner, and the lack of uterine stimulation and mitotic growth responses in αERKO uteri can be shown by using the classical uterine bioassay (37, 38). In this study, we indicated that DES did not stimulate Esr1-Q uterine growth and estrogen-responsive genes, which confirms the Q375H mutant results in dysfunctional ERα. To understand the molecular mechanism of those ESR1 mutations, the 2 ERα– cell models, COS-7 (5) and HEK293T (6) cell models, were used in functional analysis using transient transfection with those mutants. Here, we established the MDA-MB-231 cell system stably expressing WT ERα and the different mutants. In the same cell background, we were able to directly compare any differences between WT and the mutants. Whole transcriptome profiling can give us a better understanding of the extent of changes resulting from the mutations in the ESR1 gene and possible underlying molecular mechanisms. More interestingly, the transcriptome profiling indicates reproductive system disease appears to be one of the top diseases from E2-altered Q and R mutation profiles. However, this disease example did not appear in the E2-WT ER profile, suggesting the value of this in vitro analysis in cellular and molecular level translational relevance to clinical investigation in patients with EIS. Our analyses demonstrate that only one single point mutation in ERα such as those of Q375H or R394H could result in the whole transcriptome aberrations observed in the in vitro cell system. The estrogen-responsive gene profiles (E2 or DES treatment) appeared to only share very low percentages of DEGs between WT ERα and Q375H or WT ERα and R394H. The canonical pathway analyses show that E2-altered genes in both mutants are related to reproductive tract system diseases, but this was not found in WT ERα. These findings can provide insights into a possible molecular mechanism for those ESR1 mutations and direct relevance to clinical investigation and phenotypes. Generally, E2 and DES have similar gene profile stimulation in mouse uteri, which is different to what we observed from the in vitro cell system. But, since gene profiles using the cell model appear to be consistent with a hypothesis that Q375H and R394H mutations found in patients with EIS result in loss of ER function, this in vitro cell model would be a good tool for the mechanistic studies.

DNA methylation is a well-characterized epigenetic modification and plays a critical role in gene regulation, transcriptional silencing, development, and tumorigenesis (39, 40). Our recent study shows that DNA methylation dynamically changes during development in male tissues of WT and αERKO mice, and the expression levels of more than 60% of DES-altered genes are associated with their DNA methylation status (17). EPIC DNA methylome obtained from this study shows that both Q375H and R394H mutations have differential DNA methylation profiles compared with the normal ERα, suggesting that DNA methylation aberrations may directly contribute to loss of normal estrogen-responsive gene expression from the ESR1 mutants found in patients with EIS. Relationships in ER structural dynamics display variations for the 2 mutant receptors when compared with the wild-type ER, in its ligand-free form and with the 2 agonist ligand-bound forms. The mutant proteins have stable structures, but the variations in dynamical properties observed in the MD simulations, specifically in the regions responsible for coregulator interactions, may be an explanation for the variations in the receptor-mediated regulation of estrogen-responsive genes.

Based on experimental models and listed clinical cases, the principal mechanism of estrogen action is through ERα expression in predominantly estrogen target organs (41). To understand the physiological roles of ERs, our research group developed and characterized multiple mouse models, including αERKO mice (homozygous null for ERα) (42), AF2ER mice (homozygous knock-in of 2-point mutation in the LBD of ERα) (43), and H2NES mice (homozygous knock-in of multiple-point mutations in the D domain of ERα) (44). Here, we developed the first mouse model corresponding to a natural clinical point mutation (in the LBD of ERα) using the CRISPR/Cas9 genome editing. In the ER target organs, Esr1-Q mice show similar phenotypes such as hypoplastic uteri, hemorrhagic cystic ovaries, and rudimentary mammary ducts to those of the αERKO mice (Table 1, Esr1-Q mice and αERKO mice). The similarities of the phenotypes are also found in AF2ER mice (43) and H2NES mice (44), suggesting dysfunctional ERα results in common phenotypes that mirror those observed in αERKO mice even though they still express a form of the ERα protein. The same phenotypes are also found in another example from a collaborative study of a DBD-mutant mouse model (EAAE) (45). The clinical studies show that patients with both ESR1 Q375H and R394H had 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 uteri with no endometrial stripe (5, 6). Those patients have markedly high endogenous levels of serum E2, LH, and follicle-stimulating hormone (Table 1, Q375H patient and R394H patient), which match the phenotypes of Esr1-Q mice compared with αERKO mice.

Table 1.

The Comparison of the Phenotypes in the Esr1-Q Mice, αERKO Mice, and Patients with Q375H and R394H

	Esr1-Q Mice	αERKO Mice	ESR1 Q375H Patient (Reference 5)	Patient With ESR1 R394H (Reference 6)
Age	Adult (>10 Weeks)	Adult (>10 Weeks)	18 years	25 years
Body weight (based on BMI)	Overweight	Overweight	Underweight	Overweight
Ovarian cysts	Present	Present	Present	Present
Hypoplastic uterus	Present	Present	Present	Present
Acne	N/A	N/A	Present	Present
E2 levels	Elevated (~7.5 pg/mL; observed WT ~4 pg/mL)	Elevated (~7 pg/mL; observed WT ~4 pg/mL)	Elevated (3500 pg/mL; normal range 15-350 pg/mL)	Elevated (9476 pg/mL; normal range 15-350 pg/mL)
Gonadotropin (LH and FSH) levels	Elevated (LH ~5 ng/mL; observed WT ~0.5 ng/mL)	Elevated (LH ~4 ng/mL; observed WT ~0.5 ng/mL)	Slightly elevated (LH 13I U/L, normal range 2-8 IU/L; FSH 6.7 IU/L, normal range 2-10 IU/L)	Elevated (24 IU/L, normal range 2-8IU/L; FSH 13 IU/L, normal range 2-10 IU/L)
Estrogen treatment	Nonresponsive (DES)	Nonresponsive (DES)	Nonresponsive (E2)	Nonresponsive (E2)
Norethindrone treatment	Reduction in ovarian cyst size	N/A	Reduction in ovarian cyst size	N/A

Abbreviations: αERKO, ERα knock-out; BMI, body mass index; DES, diethylstilbestrol; E2, estradiol; ERα, estrogen receptor-α;ESR1, estrogen receptor 1; FSH, follicle-stimulating hormone; LH, luteinizing hormone; N/A, not applicable; WT, wild type.

A homozygous nonsense ESR1 C157T mutation was identified from a 28-year-old male patient with EIS resulting in a premature stop codon and the lack of synthesis or expression of ERα protein (2). His main phenotypes were osteopenia, incomplete epiphyseal closure, a tall stature, and abnormal gonadotropin secretion. The affected brother from the R394H family of patients with EIS has similar phenotypes as ESR1 C157T mutation (6). Overall, the main phenotypes of the male patients are all in accordance with Esr1-Q and αERKO male mice.

A clinical study of the patient with ESR1 Q375H mutation showed that the administration of NET (a synthetic progestogen) resulted in a temporary reduction in the serum LH level and the volume and number of ovarian cysts (5). NET-treated Esr1-Q female mice also responded with a reduction in the size and number of ovarian cysts, suggesting that aspects of negative feedback mediated by progesterone receptor are intact in this mouse model (46). More recently, we reported that the severe hemorrhagic ovarian phenotype found in PitERtgKO mice could be prevented by a GnRH inhibitor (FMG/degarelix acetate) (47). Our preliminary data indicate that FGM treatment can reduce the size and number of ovarian cysts in Esr1-Q mice. However, FGM treatment has some adverse side effects, and little is known about the effect on ERα function. Further studies could be considered for potential treatment toward prevention or reduction of cystic ovaries found in patients with EIS.

In summary, the mechanistic studies, including genetic and epigenetic profiles, and the consequences of ERα structural analyses can provide an important basis for understanding the molecular mechanism of EIS. CRISPR/Cas9 genome editing makes it feasible to mimic rare clinical mutations such as the mouse model, Esr1-Q, presented here. It may offer a potential new tool to study the impact and spectrum of effects from rare genetic mutations in humans in hopes of developing diagnostic endpoints and potential therapeutic treatment in a shorter period of time.

Glossary

Abbreviations

aERKO

ERa knock-out

DCCM

dynamic cross-correlation map

DEG

differentially expressed gene

DES

diethylstilbestrol

DMR

differential methylation region

DNA

deoxyribonucleic acid

E2

estradiol

EIS

estrogen insensitivity syndrome

ER

estrogen receptor

ESR1

estrogen receptor 1

FBS

fetal bovine serum

FMG

Firmagon

GnRH

gonadotropin-releasing hormone

IPA

Ingenuity Pathway Analysis

LBD

ligand-binding domain

LH

luteinizing hormone

MD

molecular dynamic

mRNA

messenger RNA

NET

norethindrone

Q

glutamine

qPCR

quantitative polymerase chain reaction

RMSD

root-mean-square deviation

RMSF

root-mean-square fluctuation

RNA

ribonucleic acid

SE

standard error

sFBS

stripped FBS

TBS-T

tris-buffered saline and polysorbate 20

Additional Information

Disclosure Statement : The authors have no competing interests and have nothing to disclose All authors have submitted the ICMJE form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Data Availability: All data generated or analyzed during this study are included in this published article or in the data repositories listed in References.