Poor Endometrial Proliferation After Clomiphene is Associated With Altered Estrogen Action

Abstract

Context

Suboptimal endometrial thickening is associated with lower pregnancy rates and occurs in some infertile women treated with clomiphene.

Objective

To examine cellular and molecular differences in the endometrium of women with suboptimal vs optimal endometrial thickening following clomiphene.

Methods

Translational prospective cohort study from 2018 to 2020 at a university-affiliated clinic. Reproductive age women with unexplained infertility treated with 100 mg of clomiphene on cycle days 3 to 7 who developed optimal (≥8mm; n = 6, controls) or suboptimal (<6mm; n = 7, subjects) endometrial thickness

underwent preovulatory blood and endometrial sampling.

The main outcome measures were endometrial tissue architecture, abundance and location of specific proteins, RNA expression, and estrogen receptor (ER) α binding.

Results

The endometrium of suboptimal subjects compared with optimal controls was characterized by a reduced volume of glandular epithelium (16% vs 24%, P = .01), decreased immunostaining of markers of proliferation (PCNA, ki67) and angiogenesis (PECAM-1), increased immunostaining of pan-leukocyte marker CD45 and ERβ, but decreased ERα immunostaining (all P < .05). RNA-seq identified 398 differentially expressed genes between groups. Pathway analysis of differentially expressed genes indicated reduced proliferation (Z-score = –2.2, P < .01), decreased angiogenesis (Z-score = –2.87, P < .001), increased inflammation (Z-score = +2.2, P < .01), and ERβ activation (Z-score = +1.6, P < .001) in suboptimal subjects. ChIP-seq identified 6 genes bound by ERα that were differentially expressed between groups (P < .01), some of which may play a role in implantation.

Conclusion

Women with suboptimal endometrial thickness after clomiphene exhibit aberrant ER expression patterns, architectural changes, and altered gene and protein expression suggesting reduced proliferation and angiogenesis in the setting of increased inflammation.

Adequate endometrial estrogen action during the follicular phase of the cycle is necessary to establish normal endometrial receptivity to embryo implantation (1). The essential action of estrogen offers a logical foundation for the strong, positive association between measurements of endometrial thickness and pregnancy rates during infertility treatment, including in vitro fertilization (2-4), unstimulated cycles (5), and insemination cycles (6). Clomiphene is a selective estrogen receptor modulator that acts centrally as an estrogen antagonist to induce pituitary gonadotropin synthesis and release. For more than 50 years, clomiphene has been the medication used most commonly for the treatment of infertility worldwide (7). However, clomiphene also has antiestrogenic effects within the female reproductive tract (8-10), which may impair endometrial growth and embryo implantation, thereby decreasing cycle fecundity (11-13).

Whereas most women receiving clomiphene treatment exhibit normal endometrial proliferation, a significant minority do not (14-19). Poor endometrial growth during treatment has been widely assumed to result from increased endometrial sensitivity to peripheral clomiphene estrogen antagonism, although the mechanism(s) responsible have not been investigated in detail (8-11, 20, 21).

Unexplained infertility is a diagnosis that generally includes unrecognized endometriosis, implantation defects, and, in theory, luteal phase defects or dysfunction (LPD). Whereas the importance and prevalence of LPD is unknown because there is no validated method for diagnosis (22, 23), decreased endometrial receptivity (24) and altered expression of receptivity markers (25-28) have been observed among women with unexplained infertility.

In recent years, clinicians have used letrozole in place of clomiphene. Compared with the central estrogen antagonistic effects and slow metabolism of clomiphene, letrozole decreases ovarian estrogen production directly by inhibiting aromatase activity and has a shorter half-life (29). Consequently, letrozole is not thought to have a direct effect on endometrium and thus should be less likely to adversely affect estrogen-stimulated endometrial growth than clomiphene (30). However, limited clinical studies observed no significant difference in average peak endometrial thickness among infertile women receiving clomiphene or letrozole (14, 30, 31). These observations suggest that the endometrium of women who exhibit poor endometrial thickening following clomiphene may be intrinsically different in a way that is revealed by such treatment (5). They also raise the intriguing possibility that an altered response to estradiol and clomiphene may contribute to otherwise unexplained infertility.

We hypothesized that poor endometrial proliferation is a consequence of inadequate or aberrant estrogen actions that may result from altered estrogen receptor expression and/or postreceptor responses. We recruited women with unexplained infertility receiving a fixed regimen of clomiphene treatment and identified those with robustly normal endometrial growth (endometrial thickness ≥8 mm) and others with clearly suboptimal growth (endometrial thickness <6 mm) (5), and compared serum sex hormone concentrations, tissue morphometrics, global gene expression, mRNA and protein expression, and genome-wide analysis of estrogen receptor binding in the 2 groups.

Materials and Methods

Human Subjects and Ethics Approval

Poor Endometrial Proliferation after Clomiphene (PEPC) was a translational prospective cohort study conducted at the University of North Carolina at Chapel Hill from 2018 to 2020. All infertile women aged 18-45 years receiving clomiphene citrate treatment (100 mg cycle days 3-7) for a diagnosis of unexplained infertility were screened for eligibility. All couples were English speaking and had a diagnosis of unexplained infertility in accordance with published best practices guidelines (32, 33), including objective evidence of normal ovulatory function (in natural cycles or after ovulation induction), a normal uterine cavity and bilateral tubal patency (documented by hysterosalpingogram), and a partner with normal semen quality (ejaculate containing ≥10 million total motile sperm). Women with any known endometriosis, adenomyosis, intrauterine lesion or congenital uterine malformation, recurrent pregnancy loss, undiagnosed vaginal bleeding or abnormal uterine bleeding, history of uterine surgery (operative hysteroscopy, myomectomy, or Cesarean delivery) or pelvic infection, any uncontrolled intercurrent illness, a body mass index (BMI) >38 kg/m², or current tobacco use were excluded. Women who took clomiphene in the preceding month were not excluded due to the short half-life of the biologically active isomer (29), but women who had endometrial biopsy or diagnostic hysteroscopy within 2 months of enrollment were excluded due to potential functional effects related to tissue injury (34, 35).

This study was carried out in accordance with federal regulations governing human subjects research. All procedures were reviewed and approved by the Committee for the Protection of Human Subjects at the University of North Carolina at Chapel Hill under file #17-3106 prior to study initiation. Informed consent was obtained from all subjects before their participation in the study. Separate consents were obtained for consent to undergoing the endometrial biopsy procedure and for the storage of specimens.

Group Assignment

At the time of a clinically scheduled transvaginal ultrasound examination performed between stimulation cycle days 12 to 14 that demonstrated 1 or more mature ovarian follicles (mean diameter ≥19 mm²), women were assigned to 1 of 2 exposure groups based on their observed endometrial response, defined as “optimal” controls (endometrial thickness ≥8 mm) or “suboptimal” subjects (endometrial thickness <6 mm). Those women exhibiting an endometrial thickness measuring from 6.0 to 7.99 mm were excluded. Endometrial thickness measurements were performed in the midsagittal plane, in accordance with standard procedure at our facility (36). Assessment of endometrial grade was visually determined by LHB and confirmed by an additional experienced observer (S.L.Y.).

Intervention

Investigations of the relationship between endometrial thickness and pregnancy rates suggest that endometrial thickness has greatest predictive value when measured in the late proliferative phase just before ovulation (37-39). We collected serum and endometrium in the late proliferative phase, defined as the day on which at least 1 mature ovarian follicle was observed. Continued participation required that subjects have a serum estradiol concentration >200 pg/mL and a serum progesterone concentration <1.5 ng/mL, as measured in our clinical assay (Siemens Immulite 1000 Analyzer, Siemens Medical Solution, Malvern, PA; estradiol 17β: solid-phase, 2-site chemiluminescent immunometric assay (40); progesterone: solid-phase, ligand-labeled, competitive chemiluminescent immunoassay (41)).

Endometrial samples were obtained using an aspiration sampling device (M0015 Preferred Curette SureFlex Endo Sampler, Bioteque America, San Jose, CA). Tissue specimens were divided into 2 portions; 1 was fixed in formalin for histopathologic examination and the second was snap frozen in liquid nitrogen and stored at −80°C for later RNA and cistrome analysis.

Sample Size Justification

Previous research from our group suggested that a sample size of 6 patients per group would yield useful results for analysis of both endometrial protein and gene expression (42). Although no relevant empiric data on this subject were available, we hypothesized that if inadequate endometrial proliferation after clomiphene treatment was associated with a change in endometrial receptivity sufficient to adversely affect pregnancy rates, the difference should be detectable even with this small sample size.

Histopathology

Sections of human endometrial tissue stained with hematoxylin and eosin (H&E) were scanned on an Aperio ScanScope XT (Leica Biosystems). Images were uploaded to eSlide Manager and visualized with ImageScope 12.4 (Leica Biosystems) to permit whole slide imaging analysis (43-45). Analysis was performed with Genie pattern recognition software (Aperio Technologies) as previously described (46-48). Briefly, the Genie algorithm was trained to segment tissue into different regions of interest (ROIs): glandular epithelium, luminal epithelium, stroma, artifact, and glass. Each scanned image was annotated manually for analysis by a third party (B.R.M.) to confirm appropriate ROIs and exclude areas of glass or poor sample quality. The final Genie classifier then calculated the total tissue area and the area percentages for each of the ROIs (49).

Immunohistochemistry and Multiplex Immunofluorescence

Immunohistochemical staining was performed using established protocols at the Translational Pathology Laboratory at UNC. Tissue slides were analyzed using proliferating cell nuclear antigen (PCNA; Leica Microsystems Inc, Wetzlar, Germany, PA0042, 1:4800 dilution), Ki67 (Biocare Medical, Pacheco, CA, monoclonal CRM325A, 1:100 dilution), platelet endothelial cell adhesion molecule (PECAM-1; Leica Microsystems Inc, Wetzlar, Germany, monoclonal PA0250, RTU), CD45 (Leica Microsystems Inc, Wetzlar, Germany, monoclonal X16/99, RTU), CD138 (Leica Microsystems Inc, Wetzlar, Germany, monoclonal MI15, RTU), ESR1 antibody (Cell Marque/Sigma Aldrich, Rocklin, CA, monoclonal 249R-16-ASR, 1:200 dilution), ESR2 antibody (BioGenex, San Roman, CA, polyclonal AR385-5R, 1:2 dilution), and a second ESR2 antibody (Invitrogen, Monoclonal IgG2b Clone # PPZ0506 (50), 1:100 dilution).

To ensure specificity of binding, samples were stained with Rabbit or Mouse IgG, as appropriate, as controls for nonspecific secondary antibody binding. Acknowledging known challenges with anti-estrogen receptor (ER) β antibodies (51), we used both a polyclonal antibody and the most specific and well-performing ERβ antibody (50).

Multiplex immunofluorescence was performed with ERα and ERβ antibodies to permit comparison in precisely the same tissue section. For immunofluorescence, we used ERα antibody (Cell Marque/Sigma Aldrich, Rocklin, CA, monoclonal 249R-16-ASR, 1:100 dilution) and ERβ antibody (Invitrogen, Monoclonal Mouse IgG2b Clone # PPZ0506 (50), 1:100 dilution).

For analysis, genie pattern recognition software was trialed but algorithm accuracy appeared poor at tissue edges in some samples and decision was made to proceed with histologic scoring. Previous work demonstrated low intra- and interobserver variability of the HSCORE system applied to immunostaining of endometrium (52). A single experienced observer (L.H.B.) blinded to the identity of the slides by a third party (J.L.P.) used the semiquantitative digital histologic scoring (dSCORE) system and Fiji (ImageJ) software (53), which has previously been validated for image analysis in endometrial assessment (54, 55).

Data Analysis

Statistical analysis software (Stata 15.1, SAS Institute Inc., Cary, NC) was used for all data analysis. Dot plots were generated using GraphPad Prism software (Version 8.1.2, GraphPad Software, La Jolla, CA). Assumption of non-normal distribution was planned to minimize risk of Type I error in the setting of a small sample size and use of human tissue samples, which introduce more intrinsic variability (56-58). Clinical and demographic variables were interrogated using the Wilcoxon rank sum test for continuous variables and the Mantel–Haenszel test for dichotomous and count variables where appropriate. The Wilcoxon rank sum test generated P values comparing semiquantitative dSCOREs between groups.

RNA Extraction

RNA was extracted using Trizol and RNEasy® (Qiagen) and reverse transcribed using a Transcriptor First Strand cDNA Synthesis Kit (Roche; 04379012001) as previously described by our group (59).

RNA Sequencing Analysis

RNA sequencing was performed as we have previously described (59). Six optimal and 5 suboptimal endometrial specimens were subjected to the RNA-seq assay. The libraries prepared from endometrial specimens were sequenced with approximately 200 millions of 75 bp paired-end reads per sample. The raw reads were processed by filtering with average quality scores greater than 20. The reads passing the initial processing were aligned to the human reference genome (hg19; Genome Reference Consortium Mouse Build 37) using TopHat version 2.0.4 (60). Expression values of RNA-seq were expressed as FPKM (fragments per kilobase of exon per million fragments) values. Sequencing data are deposited in the National Center for Biotechnology Information Gene Expression Omnibus with accession number GSE164768 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164768). Differential expression was calculated using Cufflink version 2.2.1 (61). Differentially expressed genes (DEGs) were defined as absolute fold change ≥1.2, FPKM > 1 in at least 1 of the samples, and P < .05 (62, 63). As suggested by Jeffrey and colleagues as best practice, substantive conclusions were restricted to DEGs with P < .001 (64).

DEGs generated from RNA sequencing data were analyzed using several software programs to study their biological functions (65-67). Ingenuity Pathway Analysis software (IPA, www.ingenuity.com) and the Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.8. were used to predict molecular activity in upstream regulator pathways (65, 66). Hierarchical clustering and heat maps of DEGs were generated by the Partek^© Genomics Suite software version 6.6. Heat maps were generated from statistically significant upstream regulator pathways denoted by IPA. FPKM values were log2 transformed, shifted to a mean of 0, and rescaled to a standard deviation of 1 for display on heat maps.

ERα Chromatin Immunoprecipitation Sequencing

Frozen endometrial biopsy tissue specimens were analyzed by the Active Motif company (Carlsbad, CA) for chromatin immunoprecipitation sequencing (ChIP-seq) analysis using the ERα antibody (Millipore/Merck, Darmstadt, Germany, rabbit polyclonal #06935) from a subset of patient samples (total n = 4; optimal controls n = 2, suboptimal subjects n = 2) as described by our research group (68). Limited human endometrial tissue, particularly in the suboptimal group whose endometrium was <6 mm in thickness at the time of biopsy, as well as cost limitations restricted our use to 4 samples. Peak calling was performed with MACS2 by the Active Motif data analysis service (69). Area-proportional Venn diagrams were generated comparing ERα occupancy intervals and DEGs bound by ERα in suboptimal and optimal groups (70). Sequencing data is deposited in the National Center for Biotechnology Information Gene Expression Omnibus with accession number GSE164764 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164764).

Validation of RNA Sequencing Data by Reverse Transcription Polymerase Chain Reaction

Quantitative real-time reverse transcription polymerase chain reaction (RT-PCR) was performed on total RNA, isolated as above, in technical duplicate, using probe–primer sets specific for genes of interest to provide further confirmation of RNA-seq data. As previously described by our group (59), DNA was generated using the Transcriptor First Strand cDNA Synthesis Kit (Roche Life Science, Penzberg, Germany) in accordance with the manufacturer’s manual. PCR was conducted using SsoAdvanced Universal SYBR Green Supermix (1725270, Bio-Rad, Hercules, CA) or SsoAdvanced Universal Probes Supermix (1725280, Bio-Rad, Hercules, CA) on the CFX Connect Real-Time PCR Detection System (1855201, Bio-Rad, Hercules, CA) per manufacturer’s instructions.

Probe–primer sets cross introns and, therefore, provide a specific signal from mRNA rather than genomic DNA. Primers for HMOX1, PRSS35, and ADAMT5 were designed using ProbeFinder version 2.53 for Humans (Roche Diagnostics). The primer sequences used in this study were as follows (from 5′ to 3′, F = forward and R = reverse): HMOX1—TGAACTCCCTGGAGATGACTC (F) and CCTGCAACTCCTCAAAGAGC (R); PRSS35— GTTCAAGCGATCCTGCTGA (F) and TCAACCAAAG CAGCATATTTTC (R); ADAMTS5- TGGCAGCACC AACACAAC (F) and AATGATGCCCACATAAATCCTC (R). Gene expression was normalized to 18s rRNA by the ΔΔCT method. The 2-tailed Student’s t-test with equal variance was used for the quantitative qPCR assays.

Results

Thirteen whole endometrial tissue samples from the late proliferative phase, after clomiphene exposure, were collected for examination of molecular differences between endometrium of optimal (n = 6) and suboptimal, (n = 7) thickness. Specimens from 2 women who were found to have serum progesterone levels >1.5 ng/mL at the time of endometrial biopsy were excluded from subsequent analysis, leaving a total of 11 samples (optimal: n = 6, suboptimal: n = 5) analyzed. The minimum criterion was met by a Shapiro–Wilkes test of normalcy (P = .06) but nonparametric analyses were performed as planned in accordance with expert consensus (56-58).

Serum estradiol was not different between groups (P = .56), nor were cycle day or number of mature follicles measured on the day of tissue sampling (Table 1). Clinical and demographic variables were similar between groups with the exception of dysmenorrhea, which was more commonly reported among suboptimal subjects (40% vs 0%, P = .10), and body mass index (kg/m²), which was higher among suboptimal subjects (32 vs 23, P = .01, Table 1). Normal, trilaminar endometrial architecture was similarly frequent in both groups (P = .36). In subsequent cycles, a higher proportion of suboptimal subjects underwent a letrozole treatment cycle than optimal controls (P = .045). Notably, among women who subsequently underwent a letrozole treatment cycle, endometrial thickness following letrozole was similar between groups (P = .48, Table 1).

Table 1.

Clinical and demographic variables

Variable	Optimal (n = 6)	Suboptimal (n = 5)	P
Age, years	31 (29-36)	33 (27-36)	0.46
BMI, kg/m2	23 (21-25)	32 (29-34)	0.01*
Gravidity	0 (0-1)	1 (0-1)	0.23
Parity	0 (0-1)	0 (0-0)	0.64
AMH (ng/mL)	3.8 (2.7-6.3)	1.9 (0.8-2.3)	0.05
PCOS	0 (0%)	1 (20%)	0.27
LPD	1 (14%)	0 (0%)	0.68
Dysmenorrhea	0 (0%)	2 (40%)	0.10
Cycle day	13 (12-13)	13 (13-13)	0.61
Estradiol (pg/mL)	678 (317-1138)	792 (679-875)	0.58
Progesterone (ng/mL)	0.7 (0.5-0.7)	0.5 (0.5-0.6)	0.14
No. of eggs ≥14 mm	2 (1-3)	2 (1-3)	0.79
ES (mm) at time of biopsy	9.7 (8.9-10.4)	5.5 (5-5.6)	0.006*
Trilaminar endometrial grade at biopsy	5 (83%)	5 (100%)	0.36
Letrozole usea	1 (16.7)	4 (80.0)	0.045*
ES (mm) on Letrozolea	9.4 (9.4)	8.1 (6.9-9.3)	0.48
Prior clomiphene use 1 month before biopsy	2 (33%)	2 (40%)	0.12
Prior clomiphene use ever	3 (50%)	3 (60%)	0.76

Presented as N (%) or median (interquartile range: quartile 1-quartile 3).

Mann–Whitney used for continuous variables, M-H odds used for categorical or count variables. *Statistical significance (P < .05).

Abbreviations: BMI, body mass index; AMH, anti-Mullerian hormone; PCOS, polycystic ovarian syndrome; LPD, luteal phase defect; ES, endometrial stripe thickness.

^a One optimal control and 4 suboptimal subjects used letrozole.

Tissue Level Differences

H&E staining of sections was performed using Aperio ImageScope with Genie automated software analysis to determine cell type composition as described above. Results demonstrated a lower percentage of glandular epithelium in suboptimal subjects than in optimal controls (16.1% vs. 24.2%, P = 0.01, Fig. 1). The proportion of tissue occupied by stromal cells and luminal epithelium as well as the total tissue area were similar between the groups (P > .05 for all). No H&E-stained sections demonstrated increased plasma cell numbers that would support a histologic diagnosis of chronic endometritis.

Figure 1.

Endometrial tissue level changes by histopathology examination of infertile women with optimal and suboptimal endometrial thickening following clomiphene citrate treatment. (A) Hematoxylin and eosin staining of whole slide tissue image and filter from optimal controls and (B) suboptimal subjects (filter: yellow = stromal cells, green = glandular epithelium, blue = luminal epithelium). Scale = 100µm. (C) Dot-plots depicting differences in tissue composition area (mm²) and percentages (%) in glandular epithelium, luminal epithelium, stroma, and total tissue area. Wilcoxon rank-sum generated P values.

Comparative Gene Expression

Whole endometrial RNA Sequencing analysis identified over 20 000 genes that had least 1 sample with FPKM > 1. Comparison of endometrium between optimal and suboptimal groups identified 398 DEGs (P < .05) (Fig. 2A).

Figure 2.

Comparative gene expression between infertile women with optimal and suboptimal endometrial thickening following clomiphene citrate treatment. (A) Volcano plot of 398 differentially regulated genes identified by RNA-seq whereby |fold change| > 1.2, P < .05 and genes had at least 1 sample with FPKM > 1. Red denotes increased expression and blue denotes decreased expression in suboptimal vs optimal controls. (B) Heat map of IL2 and VEGF target gene expression in infertile women exposed to clomiphene citrate who developed an optimal vs. suboptimal endometrial thickness (x-axis) by predicted molecular activity (y-axis). Red denotes increased expression and blue denotes decreased expression in suboptimal subjects vs optimal controls.

Upstream regulator pathway analysis of DEGs via IPA identified significant differences between groups for predicted molecular activity in proliferation, angiogenesis, and inflammation pathways (Table 2). Compared with optimal controls, suboptimal subjects had reduced predicted molecular activity in cellular proliferation (FOXM1 Z-score = –2.2, P < .01) and angiogenesis (VEGF Z-score = –2.87, P < .0001) pathways and increased predicted molecular activity in inflammatory pathways (interleukin [IL]-2 Z-score = +2.2, P < .01, Table 2 and Fig. 2B). Suboptimal subjects also demonstrated a decrease in predicted molecular activity of the anti-inflammatory drug dexamethasone pathway (Z-score = –3.6, P < .0001). Heat map representation of gene expression levels determined by RNA-seq identified alterations in pathways related to inflammation and angiogenesis (Fig. 2B). Individual genes are presented elsewhere (Table 1 (71)).

Table 2.

Upstream regulators identified using ingenuity pathway analysis

Upstream regulator	Molecule type	Activation Z-score	P of overlap
FOXM1	Transcription regulator	–2.23	1.71 × 10–3
VEGF	Group	–2.87	3.46 × 10–5
IL2	Cytokine	2.22	2.53 × 10–3
ESR1	Ligand-dependent nuclear receptor	–0.57	5.08 × 10–5
ESR2	Ligand-dependent nuclear receptor	1.60	3.0 × 10–5
Beta-estradiol	Hormone	0.63	3.20 × 10–3
Dexamethasone	Chemical drug	–3.58	1.06 × 10–10
SMAD7	Transcriptional regulator	3.10	9.22 × 10–5
ERBB3	Kinase	2.50	6.0 × 10–4
ERBB4	Kinase	2.06	8.87 × 10–5
TGFB1	Growth Factor	–3.48	4.22 × 10–13

+Z-score implies activation of downstream molecular pathway in the suboptimal group

Immunohistochemical Staining

Immunohistochemical staining was performed to evaluate differences in proliferation, angiogenesis and inflammation at the protein level. Suboptimal subjects demonstrated evidence of reduced proliferation by decreased glandular and stromal PCNA immunostaining (dSCORE = 74 vs 101, P < 0.01; 27 vs 82, P <0.05, respectively, Fig. 3A) and decreased glandular ki67 immunostaining (dSCORE = 38.3 vs 65.0, P = .03, Fig. 3B). Evidence of inflammation was assessed by immunostaining for bone marrow–derived immune marker CD45 (cluster of differentiation 45) and showed increased stromal staining in suboptimal compared with optimal subjects (dSCORE = 31 vs 12, P < .01, Fig. 3C). Immunostaining for the plasma cell marker CD138 (syndecan-1) was similar between groups in both glandular (P = .93) and stromal (P = .13) compartments. Immunohistochemical staining for an angiogenic marker, PECAM-1 (CD31), was reduced among suboptimal compared to optimal subjects (dSCORE = 77 vs 95, P = .03, Fig. 3D).

Figure 3.

Protein level changes by immunohistochemical staining in late proliferative endometrium from women with optimal compared to suboptimal endometrial thickening following follicular phase exposure to clomiphene citrate. (A) Tissue staining in representative samples and dot-plots depicting semiquantitative analyses of digital H-score (dSCORE) from all samples stained for PCNA (scale = 100µm), (B) ki67, (scale = 100μm) (C) pan-leukocyte inflammatory marker CD45 (scale = 200µm), and (D) angiogenesis marker PECAM-1 (scale = 100µm). Dot-plots represent medians and interquartile ranges (Q1-Q3). Wilcoxon rank-sum generated P values.

Estrogen Signaling

As estrogen is known to promote endometrial proliferation and enhance both angiogenesis and inflammation, initial attention was focused on evaluating for differences in estrogen signaling between groups. Pathway analysis, using IPA software, demonstrated altered ER subtype activity between groups with suboptimal subjects demonstrating increased ERβ expression (Z-score = +1.6, P < .0001) compared with optimal controls and slightly decreased predicted ERα activity (Z-score = –0.57, P < .0001). Heat map representation of gene expression levels demonstrated visually distinct profiles between groups for predicted molecular activity of ERα and ERβ downstream pathways (Fig. 4A).

Figure 4.

Examination of estrogen signaling and estrogen action between infertile women with optimal compared to suboptimal endometrial thickening after clomiphene citrate. (A) Heat map of differentially expressed ERα and ERβ target gene expression in infertile women exposed to clomiphene citrate who developed optimal vs suboptimal endometrial thickening (x-axis) by predicted molecular activity (y-axis). Red denotes increased expression and blue denotes decreased expression in suboptimal subjects vs optimal controls. (B) Immunohistochemical staining of ERα and ERβ showing altered estrogen receptor subtype abundance between suboptimal and optimal controls. (C) Dot-plots depicting semiquantitative analysis of digital H-score (dSCORE) from all samples. (D) Variation in ER subtype abundance by multiplex immunofluorescence analysis. Scale = 100µm.

To further examine the potential role of altered estrogen signaling, we examined the localization and intensity of ERα and ERβ immunostaining. Subjects with a suboptimal endometrial thickness demonstrated decreased immunostaining of ERα and increased immunostaining of ERβ compared with optimal controls in both glands (P = .01) and stroma (P = .04, Fig. 4C). Experiments were repeated with a highly specific ERβ antibody (Monoclonal Mouse IgG2b Clone # PPZ0506) (50) and yielded highly similar results (data not presented due to lack of antibody validation in the application of immunohistochemical staining (50)).

Since analysis of tissue composition (vide supra) demonstrated a higher proportion of glandular epithelium in optimal subjects, multiplex immunofluorescence was performed to permit visualization of ERα and ERβ immunostaining in the same tissue section. Results demonstrated colocalization of ERα and ERβ in the glandular nuclei (Fig. 4D).

ERα ChIPSeq

To further evaluate altered estrogen receptor function between groups, we examined genome-wide alterations in ERα binding using ERα ChIP-seq in a subset of samples (total n = 4; optimal controls n = 2, suboptimal subjects n = 2). The majority of ESR1 target genes bound by ERα in suboptimal samples were also bound by ERα in optimal controls, whether comparing target genes bound in both endometrial samples per group (Fig. 5A) or at least 1 sample per group (Fig. 5B).

Figure 5.

ERα ChIP-seq results and confirmation of differentially expressed genes identified by RNA-seq. (A) Venn Diagram representation of ChIP-Seq results overlapped with differentially expressed genes (DEGs) identified by RNA-seq comparing genes bound by ERα in both samples in the optimal and suboptimal groups. (B) Venn diagram depicting DEGs bound by ERα in at least 1 sample per group. (C) Dot-plots depicting evaluation of gene expression by RT-PCR in 3 DEGs identified on RNA-seq: HMOX1, PRSS35, and ADAMTS5.

A total of 53 target genes were bound by ERα on ChIPSeq and were also DEGs using our RNA sequencing data (Table 3). Of the 53 differentially expressed ERα target genes, 6 genes were found to have an absolute fold change of >1.5, P < .001 (64), and were bound by ERα on ChIPSeq in at least 1 sample per group: ADAM metallopeptidase with thrombin type motif 5 (ADAMTS5; fold change from suboptimal to optimal of –1.77, P < .0001), Laminin subunit alpha 4 (LAMA4; fold change –1.72, P < .0001), Myotilin (MYOT; fold change +1.587, P < .0001), Ribosomal Protein S6 Kinase Like 1 (RPS6KL1; fold change +1.61, P < 0.001), Sphingomyelin phosphodiesterase 3 (SMPD3; fold change +1.71, P < .001), and Thrombospondin 2 (THSB2; fold change –1.68, P < 0.001) (Table 1 (71)).

Table 3.

Differentially expressed genes (RNA-Seq) that are bound by ERα (ChIP-seq)

Gene symbol	Fold change**	P	Gene description
ABCA3	1.338	3.4 × 10–02	ATP binding cassette subfamily A member 3
*ADAMTS5	–1.773	3.1 × 10–05	ADAM metallopeptidase with thrombospondin type 1 motif 5
ADAMTS8	–1.361	3.6 × 10–02	ADAM metallopeptidase with thrombospondin type 1 motif 8)
ADAMTSL5	1.41	.000117	ADAMTS-Like Protein 5
AFAP1L2	–1.522	.00175	Actin filament associated protein 1 like 2
AKAP1	1.49	.000435	A-Kinase Anchoring Protein 1
ALOX5	–1.3	.0078	arachidonate 5-lipoxygenase
ANXA5	–1.351	.001	Thromboplastin inhibitor
AOC3	0.6	3.0 × 10–04	Amine Oxidase Copper Containing 3
ARHGAP26	1.567	.0004	Rho GTPase activating protein 26
BCL6	–1.325	.016	B cell CLL/lymphoma 6
C1orf198	–1.381	.003	chromosome 1 open reading frame 198
C1QTNF7	–1.41	.028	C1q and TNF related 7
C8orf48	–1.299	.0282	chromosome 8 open reading frame 48
CCBE1	–1.385	.0273	collagen and calcium binding EGF domains 1
CDH2	–1.302	.0468	cadherin 2
CISH	–1.222	.046	Cytokine Inducible SH2 Containing Protein
COL18A1	–1.272	.0396	collagen type XVIII alpha 1 chain
COL21A1	–1.422	.0245	collagen type XXI alpha 1 chain
DAB2	–1.355	3.6 × 10–03	Differentially Expressed Protein 2
EDN2	–1.511	.0105	endothelin 2
FAT4	–1.451	.0212	FAT atypical cadherin 4
FMO1	–1.488	.0109	flavin containing monooxygenase 1
GABRR2	1.281	.0287	gamma-aminobutyric acid type A receptor rho2 subunit
GRK5	–1.259	.0299	G protein–coupled receptor kinase 5
HAPLN2	–1.387	.0419	hyaluronan and proteoglycan link protein 2
HSD17B2	–1.439	.0176	hydroxysteroid 17-beta dehydrogenase 2
ITGB5	–1.253	.0371	integrin subunit beta 5
ITPK1	1.255	.0365	inositol-tetrakisphosphate 1-kinase
KDM4B	1.2	.03	Lysine Demethylase 4B
KLK2	1.31	.03	Kallikrein Related Peptidase 2)
*LAMA4	–1.715	.0000705	laminin subunit alpha 4
LOC389895	–1.335	.0454	chromosome 16 open reading frame 72-like
MN1	1.345	.0166	MN1 proto-oncogene, transcriptional regulator
MYB	1.379	.0102	MYB proto-oncogene, transcription factor
*MYOT	1.587	.00042	myotilin
NXNL2	–1.348	.03	Nucleoredoxin-Like Protein 2
PKDCC	–1.289	.0338	protein kinase domain containing, cytoplasmic
PLCE1	–1.316	.0145	phospholipase C epsilon 1
PLCG2	1.369	.0151	phospholipase C gamma 2
PXK	–1.41	.00239	PX domain containing serine/threonine kinase like
RBM47	1.275	.047	RNA binding motif protein 47
*RPS6KL1	1.609	.000411	ribosomal protein S6 kinase like 1
RSPO1	–1.447	.0174	R-spondin 1
RUNX1	1.327	.0127	runt related transcription factor 1
RXRA	1.245	.0295	retinoid X receptor alpha
SLC7A5	1.37	.03	Solute Carrier Family 7 Member 5)
*SMPD3	1.708	.000176	sphingomyelin phosphodiesterase 3
SPARCL1	–1.389	.027	Proliferation Inducing Protein 33
*THBS2	–1.684	.000563	thrombospondin 2
TSPAN5	–1.33	.0115	tetraspanin 5
USP31	1.29	.038	Ubiquitin Specific Peptidase 31
WWC1	1.319	.0132	WW and C2 domain containing 1

*Absolute fold change >1.2, P < .05, and bound by ERα in at least 1 sample per group. **Positive fold change denotes increased expression in the Suboptimal group.

ERα genome occupancy profiles were also compared and found to be similar between groups, with 86% of occupancy intervals bound by ERα in suboptimal subjects also bound by ERα in optimal specimens (Fig. 1A (71)). ERα-occupied regions exhibited similar contexts between groups, with similar proportions of occupied intervals occurring in exons, intergenic, and promoter regions in optimal (Fig. 1B (71)) and suboptimal (Fig. 1C (71)) groups.

To estimate which endometrial ERα-binding intervals occur in the stromal compartment, we compared ERα-binding intervals with a recently published ESR1-binding profile in stroma (72) (Fig. 1A (71)). A total of 36 stromal ESR1 occupancy sites were shared between groups and 14 are associated with known genes (Table 3 (71)), although none were differentially expressed between groups with P < .001.

RT-PCR Confirmation of DEGs

Differences in RNA abundance in specific genes involved in proliferation (HMOX1 (73, 74)), angiogenesis (ADAMTS5 (75), HMOX1 (76)), and inflammatory pathways (PRSS35 (77)) were validated via RT-PCR. The expression of HMOX1, ADAMTS5, PRSS35 was +3.6, +3.1, and +6.6-fold higher, respectively, in optimal vs suboptimal samples, consistent with results from our RNA-seq analysis (Fig. 5B). These findings further support the concept of reduced proliferation and angiogenesis combined with increased inflammation in subjects exhibiting poor endometrial thickening after clomiphene.

Discussion

Poor endometrial growth correlates with worse treatment outcomes (3-5). However, the molecular biology defining the mechanisms causing poor endometrial growth among some women receiving clomiphene treatment has not been investigated. We hypothesized that women with poor endometrial growth after clomiphene treatment have aberrant endometrial ER signaling. We performed a set of complementary experiments to identify unique tissue and molecular differences between women with optimal or suboptimal endometrial thickening after clomiphene treatment, including morphologic examination, immunostaining, mRNA expression patterns, and the ERα cistrome. Key findings included reduced biomarkers of proliferation and angiogenesis, increased markers of inflammation, and aberrant estrogen receptor expression in patients with poor endometrial thickening.

Hypothesized Downstream Pathways of Altered ER Subtype Abundance and Activity

We found that most women with suboptimal endometrial thickness following clomiphene went on to develop normal midcycle endometrial thickness following letrozole treatment, suggesting that alterations in cellular pathways rather than intrinsic endometrial abnormality drives an abnormal response to clomiphene in these women. A higher proportion of women in the suboptimal group subsequently used letrozole, which is consistent with most patients’ preference to trial letrozole before proceeding with in vitro fertilization, which is not covered by insurance in our nonmandated state, after clomiphene treatment resulted in the development of a suboptimal endometrial thickness. A similar incidence of trilaminar endometrial architecture in both groups after clomiphene treatment implies that differences in endometrial thickening are not driven by differences in endometrial grade (2).

Proliferation, angiogenesis, and inflammation pathways differed by endometrial thickness after clomiphene treatment and appeared to reflect differences in estrogen receptor expression. Women with a suboptimal endometrial thickness demonstrated a lower proportion of glandular epithelium on H&E staining despite a similar overall area of tissue in the sections. Recognition that glandular hyperplasia characteristic of the late proliferative phase is an estrogen-dependent process (1, 78-81) offers mechanistic insight into how the development of a suboptimal endometrial thickness following clomiphene might represent a suboptimal response to clomiphene’s selective agonism and antagonism of the estrogen receptor. Evidence of reduced proliferation in suboptimal subjects also was demonstrated by decreased immunostaining of proliferation markers PCNA and ki67. Results from RNA-seq and pathway analysis corroborated this finding with reduced FOXM1 pathway activity in subjects with a suboptimal growth response to clomiphene. The decreased proliferation observed in some women after clomiphene treatment may be a downstream effect of alterations in estrogen signaling, as endometrial proliferation is largely driven by estradiol in a process characterized by an increase in the number and size of endometrial glands (2, 4, 5). Endometrial glands are hypothesized to play an important role in normal uterine receptivity to implantation (82) and, thus, the reduced abundance of glandular cells might herald altered endometrial function.

Angiogenesis normally occurs during the proliferative phase, is promoted by estrogen, and is critical to endometrial receptivity and fetal survival (83, 84). Specifically, VEGF has been identified as an essential protein for implantation (85). Our findings of decreased immunostaining of the blood vessel marker PECAM-1, reduced predicted molecular activity of the VEGF pathway, and lower levels of HMOX-1 transcript in RT-PCR in patients with poor endometrial proliferation suggest a defect in endometrial angiogenesis at the protein and mRNA transcript levels. Angiogenesis itself might also drive endometrial epithelial proliferation (86), implying that alterations in angiogenesis may be fundamental to endometrial dysfunction.

Activation of IL-2 signaling mediators and reduction in anti-inflammatory signaling via the dexamethasone pathway along with findings of increased CD45 immunostaining suggest an association between a proinflammatory environment and poor endometrial growth after clomiphene. IL-2 signaling can affect the differentiation and survival of both effector and memory T cells (87, 88) as well as Th17 regulatory cells, both of which are thought to be important to fetal survival (89). Endometrial inflammation is evident in women with endometriosis by evaluation of eutopic endometrium (90, 91) and, as such, our findings and a higher prevalence of dysmenorrhea in the suboptimal group could suggest undiagnosed endometriosis or other causes of inflammation among suboptimal subjects. Though angiogenesis and inflammation are typically mutually dependent processes (92), defects in angiogenesis and decidualization occur in the setting of increased inflammation in the eutopic endometrium of women with endometriosis (90) and additionally suggest that development of a suboptimally thin endometrium after clomiphene may identify early or clinically unrecognized endometriosis.

In the present study, women with poor endometrial proliferation had a significantly higher BMI than women with optimal endometrial thickness. Although small sample size precludes reliably meaningful correlations with demographic variables, the documented association of obesity with increased endometrial inflammation provides a biologically plausible explanation (93-95). Although women in our study were neither clinically hyperandrogenic nor anovulatory (96), it is possible that women with a higher BMI experience subclinical hyperandrogenism (97) that impacts their endometrium. This speculation may be supported by elegant experiments in rhesus monkeys showing that chronic hyperandrogenism and high-calorie, high-fat diets led to the development of a thinner endometrium (98), infertility (99), and placental dysfunction (100).

Differences in ER Subtype Abundance

Women who exhibited poor endometrial growth after clomiphene demonstrated decreased ERα and increased ERβ gene expression at the transcript and protein levels. Women with endometriosis demonstrate increased ERβ gene expression in endometriotic lesions (101, 102) and mechanistic investigations suggest that increased ERβ gene expression drives a decrease in ERα gene expression (103). ERβ has been shown to inhibit ERα-induced cell proliferation through formation of ERα/ERβ heterodimers with different molecular activity (104), and other nongenomic mechanisms (105).

ERα Cistrome Results

Key findings from ChIPSeq analysis included significant overlap between groups. ERα occupancy profiles were largely similar and the majority of genes bound by ERα in subjects with poor endometrial growth were also bound by ERα in normal controls. In addition, most DEGs identified by RNA sequencing were not differentially bound between groups.

Our observations have numerous potential explanations. Downstream modifications including post-transcriptional or post-translational alterations may affect differences in ER subtype abundance and localization, which would be consistent with our findings on immunohistochemical staining. It is also possible that the activity of coactivators or corepressors, rather than ERα binding itself, drives differential ER effects on gene expression, or that nongenomic functions of ER vary between groups (106). The relative affinity of ERα binding to target genes may not be captured by ChIPSeq, such that while the threshold for binding was met in both groups, differential affinity for binding might exist above the ChiPSeq detection threshold. Lastly, our observations may be driven by differences in ERβ gene expression, which has been shown to alter ERα activity and expression (102, 104, 105).

ChIP-seq identified 6 genes bound by ERα that were also identified by RNA-seq as differentially regulated between groups: ADAMTS5, LAMA4, THSB2, SMPD3, MYOT, RPS6KL1. ADAMTS5 expression is negatively regulated by cytokines in human decidual stromal cells (107) and in our study was lower in suboptimal subjects, consistent with higher levels of inflammation in this group. Suboptimal subjects had lower levels of LAMA4, an extracellular matrix glycoprotein that has been associated with pre-eclampsia (108, 109) and decreased VEGF gene expression (110), which may relate to altered angiogenesis in these women. THBS2, a glycoprotein that mediates cell-to-cell and cell-to-matrix interactions, is associated with angiogenesis in a receptor-dependent fashion and was lower in suboptimal subjects (111, 112). SMPDE3 was increased in suboptimal compared with optimal subjects and is a potential biomarker for endometriosis that is upregulated in the eutopic endometrium of women with endometriosis (113). Expression of myotilin, a protein necessary for microfilament assembly (114), was increased in suboptimal subjects, as was a protein kinase RPS6KL1 (115). No specific role for either myotilin or RPS6KL1 in endometrial function or pathology has been previously reported.

To date, 1 previous investigation has assessed the ERα cistrome in the stromal endometrium of healthy women undergoing hysterectomy (72). These women were not known to be infertile and endometrial tissue was not collected at a specific time in the menstrual cycle. Yet, as the only existing investigation of ERα occupancy intervals in endometrium, we compared our data directly with these. We did not expect overlap due to baseline differences and did not identify significant differences.

Strengths/Limitations

Our study is a novel, mechanistic investigation of a commonly encountered clinical observation of poor endometrial proliferation after clomiphene treatment. Rigorous inclusion criteria were used to identify subjects with the diagnosis of unexplained infertility without history of uterine procedure or pathology. Study subjects had the same diagnosis, were subjected to the same treatment dose and duration and underwent tissue sampling in the same (late proliferative) phase of the endometrial cycle. Serum sex hormone measurement confirmed typical and similar late follicular phase estradiol concentrations in both groups and demonstrated that neither ovulation nor premature luteinization had occurred. Transvaginal ultrasound permitted assessment of follicle sizes, measurement of endometrial thickness, and assessment of endometrial grade.

To the best of our knowledge, there has not been a previously reported mechanistic investigation of endometrial tissue in women who exhibit poor endometrial proliferation after clomiphene exposure or in any other context. ERα ChIP-seq has not been performed on human endometrial tissue from women with infertility. ERα ChIP-seq in murine models (116) may not be informative in humans due to species conservation in mouse–human comparison of just 6% of predicted estrogen response elements (117).

The greatest insight into whether the extent of endometrial growth after clomiphene treatment impacts implantation mechanisms might be gained by study of tissues obtained during the secretory phase, but endometrial tissue sampling after ovulation may disrupt implantation or early pregnancy and thus is not acceptable in women actively seeking pregnancy. Previous investigations suggest that the late proliferative phase is the next-most ideal time to identify key differences in endometrial gene expression in patients with unexplained infertility (118-120). Intra- and interobserver variability from the recording of measurements by multiple providers may have introduced bias, although research demonstrates excellent reproducibility of measurements within 1 mm and specifically when using a cutoff value of 6 mm (36, 121). To mitigate potential bias, we excluded women with endometrial thickness measurements from 6.0 to 7.99 mm in efforts to ensure enrollment of only those women with either clearly suboptimal (<6 mm) or clearly optimal (>8 mm) endometrial thickening following clomiphene.

Our results are limited by small sample size. To account for this, conservative statistical methods, including use of nonparametric tests, were employed to mitigate the risk of Type I error. The limited amount of available human tissue and cost constraints restricted evaluation of ChIP-seq to 4 samples, which limits our insight into causal genomic mechanisms and generalizability of findings. At present, no ERβ antibody has been validated for ChIP-seq and this complicated our ability to more thoroughly evaluate the impact of altered ER subtype expression on gene expression. Considering controversy over specificity and sensitivity of ERβ antibodies, we used 2 different antibodies for immunostaining, 1 of which has previously been shown to be highly specific for ERβ in endometrium (50). While we focused our immunohistochemical analysis on protein targets with known roles in endometrial proliferation, Immunohistochemical analysis of those transcripts identified as differentially regulated by RNA-seq may provide additional insight.

Conclusion

We investigated mechanisms that might explain the common clinical observation of poor endometrial thickening after clomiphene treatment in some infertile women. Women with inadequate endometrial growth demonstrated reduced proliferation, increased inflammation and altered estrogen signaling. Altered ER subtype abundance and activity suggest an intrinsic endometrial mechanism for the altered action of estrogen after clomiphene in some women. We speculate that these endometrial abnormalities may underlie a distinct cause of infertility among patients currently deemed unexplained.

These foundational data allow for an improved understanding of why a sizable minority of women with infertility exhibit inadequate endometrial proliferation. Ongoing efforts to obtain endometrium from the same patient groups in natural (unmedicated) cycles, organoid culture and, when feasible, assessing the ERβ cistrome would provide additional foundational information on the molecular regulation of poor endometrial proliferation after clomiphene treatment.

Glossary

Abbreviations

BMI

body mass index

CD

cluster of differentiation

ChIP-seq

chromatin immunoprecipitation sequencing

DEG

differentially expressed gene

ER

estrogen receptor

FPKM

fragments per kilobase of transcript per million mapped reads

H&E

hematoxylin and eosin

IL

interleukin

IPA

Ingenuity Pathway Analysis

LPD

luteal phase defects or dysfunction

PCNA

proliferating cell nuclear antigen

ROI

region of interest

RT-PCR

reverse transcription polymerase chain reaction

Additional Information

Disclosures: L.H.B. is a shareholder in 2minutemedicine®. All other authors have nothing to disclose.

Data Availability

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

Supplemental tables and figures can be found at: https://doi.org/10.5061/dryad.9kd51c5gk

Gene expression data can be found at: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164764

https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE164768