Molecular Evidence for Differences in Endometrium in Severe Versus Mild Endometriosis

Study Participants

The study was approved by the Committee on Human Research of the University of California−San Francisco (UCSF) and the Stanford University Committee on the Use of Human Subjects in Medical Research. Samples were obtained from the National Institute of Health Specialized Cooperative Centers Program in Reproduction and Infertility Research (NIH SCCPRR) Human Endometrial Tissue and DNA Bank at UCSF. Endometrial tissue was obtained from 12 participants without endometriosis undergoing endometrial biopsy or hysterectomy for benign disorders not related to endometrial pathology (used in immunohistochemistry experiments) and from 63 participants with endometriosis undergoing endometrial biopsy for infertility evaluation or hysterectomy for treatment of severe pelvic pain and extensive endometriosis (Table 1 ). All participants were documented not to be pregnant and not to have had hormonal treatment for at least 3 months before surgery. Staging of endometriosis was performed according to the revised American Fertility Society classification system.^3,4 Of the 63 participants, 19 had mild and 44 had severe endometriosis. The majority of endometriosis samples were obtained by endometrial biopsy, whereas the majority of control (no endometriosis) samples were obtained after hysterectomy (Table 1). (Note our previous studies demonstrated that sampling technique does not affect the endometrial transcriptome.¹⁶) The mean ages (years) of patients were: mild endometriosis group, 35.7 ± 1.4; severe endometriosis group, 35.2 ± 1.2l (P > .05). Menstrual cycle phase was determined based on histological evaluation of the tissue by 3 independent readers and according to the Noyes' criteria.¹⁷

Isolation of RNA and Preparation for Hybridization

Each endometrial tissue specimen was processed individually for microarray hybridization, as described earlier.¹³ Briefly, total RNA was extracted from whole-tissue specimens using the Trizol reagent (Invitrogen, Carlsbad, California), subjected to DNase treatment, and purified using the RNeasy Plus Kit (QIAGEN, Valencia, California). RNA purity was assessed by the A260/A280 ratio, and quality and integrity were assessed using the Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, California), with all samples having high-quality RNA (RNA Integrity Number (RIN) = 9.7-10).

RNA samples were prepared for microarray analysis according to the Affymetrix protocol (Affymetrix, Inc, Santa Clara, California), as described earlier.^13,16 Briefly, for each sample, 5 µg of total RNA were reverse transcribed to complementary DNA (cDNA). Second strand DNA was generated using DNA polymerase, followed by overnight in vitro transcription to generate cRNA. After chemical fragmentation, biotinylated cRNAs were ready for hybridization. Quality of the final product was assessed in the Agilent Bioanalyzer. Each sample was hybridized to HU133 Plus 2.0 high-density oligonucleotide array (Affymetrix), with 54 600 genes and expressed sequence tags (ESTs), at the UCSF Genomic Core Facility. The data were scanned according to the protocol described in Assay Manual from Affymetrix.

Microarray Data Analysis

The .cel data files were imported into GeneSpring GX 10.0 software (Agilent Technologies) and processed using the robust multiarray analysis (RMA) algorithm for background adjustment, normalization, and log2-transformation of perfect match (PM) values.¹⁶ The data during each menstrual cycle phase (PE, ESE, and MSE were compared between the severe and mild endometriosis groups. The generated gene lists included only genes with >2.0-fold change (FC) and P < .05 by 1-way analysis of variance (ANOVA) with Tukey post hoc test and Benjamini-Hochberg multiple testing correction for false discovery rate.

Principal Component Analysis and Hierarchical Clustering

Principal component analysis (PCA) and hierarchical clustering were performed as described.^15,16 Principal component analysis is an unbiased analysis performed in GeneSpring with all samples, using all 42 203 genes and 12 397 ESTs on Affymetrix Human HU133 Plus 2.0 arrays to look for similar expression patterns and underlying cluster structures. Hierarchical cluster analysis of differentially expressed genes from all samples was conducted using the smooth correlation distance measure algorithm (GeneSpring) to identify samples with similar patterns of gene expression. Compared to PCA, hierarchical clustering uses only informational genes—that is are differentially expressed among all experimental conditions.

Ingenuity Pathway Analysis

Gene symbols and FCs of the up- and downregulated genes in each pairwise comparison were imported into Ingenuity Pathway Analysis (IPA, Ingenuity Systems, Redwood City, California), as described earlier.¹⁵ For each comparison, associated top significantly regulated molecular and biological networks and canonical molecular pathways were identified. Only networks with the highest score were selected for the analysis. This was followed by functional analysis on the data set level and canonical pathway analysis. The significance of the association between the genes from the data set and the canonical pathway (in the IPA library) was presented as a ratio of the number of genes from the data set in a given pathway divided by the total number of molecules that make up the canonical pathway (Fisher exact test was used to calculate a P value). Pathways with P < .05 and ratio >0.05 were considered significant.

Microarray Validation by Real-Time Reverse Transcriptase−Polymerase Chain Reaction

Real-time reverse transcriptase−polymerase chain reaction (RT-PCR) was performed in duplicate using the SYBR Green PCR Mix (Fermentas Inc, Glen Burnie, Maryland), according to the manufacturer’s instructions. The housekeeping gene RPL19 was used as the normalizer. Numbers of mild endometriosis samples used for validation were n = 5, n = 3, and n = 5 for PE, ESE, and MSE, respectively, and in the severe endometriosis group, n = 6, n = 5, and n = 6 for PE, ESE, and MSE, respectively (Table 1 ). The following primer sequences were used: thyroxine deiodinase 2 (DIO2) sense 5′- TTGTACTTACTCTAAATTTCCCAAGG-3′ and antisense 5′-CATTGCCACTGTTGT CACCT-3′; insulin-like growth factor binding protein 5 (IGFBP5) sense 5′-TGCACCTGAGATGAGACAGG-3′ and antisense 5′-GCTTCATCCCGTACTTGTCC-3′; somatostatin (SST) sense 5′-CCCAGACTCCGTCAGTTTCT-3′ and antisense 5′-ATCATTCTCCGTCTGGTTGG-3′; transgelin (TAGLN) sense 5′-TTAGCTTTCCCCAGACATGG-3′ and antisense 5′-CGGTAGTGCCCATCATTCTT-3′; versican (VCAN) sense 5′-CCAGCCCCCTGTTGTAGAAA-3′ and antisense 5′’-ATTGAATTGTCCTTT GCTGATG-3′; solute carrier family 1, member 1 (SLC1A1) sense 5′-AACACTGCCTGTCACCTTCC-3′ and antisense 5′-GCACTCAGCACAATCACCAT-3′; epidermal growth factor receptor (EGFR) sense 5′-GAATGCATTTGCCAAGTCCT-3′ and antisense 5′-CGTCTATGCTGTCCTCAGTCA-3′; and RPL19 sense 5′-GCAGAT AATGGGAGGAGCC-3′ and antisense 5′- GCCCATCTTTGATGAGCTTC-3′. Polymerase chain reactions were run on the Mx4000 and Mx3005 quantitative real-time reverse transcriptase−polymerase chain reaction (QPCR) Stratagene systems (Agilent Technologies), using thermal cycling conditions, as described.^15,18 Statistical analysis for the QRT-PCR results was performed using the nonparametric Mann-Whitney test. Significance was determined at P ≤ .05.

Immunohistochemistry

Immunostaining was performed for VCAN and EGFR using 4 µm thick paraffin-embedded endometrial tissue sections from women with mild and severe endometriosis: PE, n = 4 and n = 5, respectively; ESE, n = 3 and n = 7, respectively; and MSE, n = 3 and n = 5, respectively), as well as women without endometriosis (n = 4 in all phases). The samples were de-paraffinized in Xylene (Sigma-Aldrich, St Louis, Missouri) and rehydrated in decreasing concentrations of ethanol. All slides were incubated for 15 minutes in H₂O₂ (3% in methanol) to block endogenous peroxidase activity after antigen retrieval by boiling slides in citrate buffer (pH = 6.0). Thereafter the slides were blocked with normal horse serum for 45 minutes, followed by incubation with the primary antibody: overnight 4°C incubation with the rabbit polyclonal anti-VCAN antibody (Versican V0/V2 Neo, ThermoScientific, Waltham, Massachusetts) at 5 µg/mL concentration, and 1 hour at room temperature for the rabbit anti-human EGFR antibody (Santa Cruz Biotechnology, Inc, Santa Cruz, California, kind gift from Dr M Hsieh, UCSF) at 1:50 dilution.

In negative control slides, the primary antibody was replaced with nonimmune immunoglobulin G (IgG) of equivalent concentration from the same species. All slides were incubated with universal goat anti-rabbit/mouse secondary antibodies (Vector Laboratories Inc, Burlingame, California) for 30 minutes at room temperature. A freshly prepared diaminobenzidine-hydrogen peroxide solution (ImmPACT DAB kit, Vector Laboratories) was added to the slides, which were thereafter rinsed with distilled water. The slides were counterstained with haematoxylin (Vector Laboratories) and mounted with Clarion mounting medium (SigmaAldrich). A Leica microscope was used to visualize the immunostaining and to photograph the results. Sections of mouse ovarian and lung tissue (a kind gift from Dr Marco Conti, UCSF) were used as positive controls for VCAN immunostaining.^19,20 Sections of 12-week human placental tissue served as a positive control for EGFR staining²¹; myometrium served as an internal positive control.^22,23

Cluster Analysis

Principal component analysis of all genes showed that mild and severe endometriosis samples cluster according to their cycle phase rather than the disease stage (Figure 1A), confirming previous observations of phase-dependent segregation when analyzing endometrial tissue or isolated cells.^13,15 However, PCA followed by subsequent analysis of disease stage demonstrated that severe endometriosis samples cluster separately from mild endometriosis, regardless of cycle phase, although there was some overlap (Figure 1B).

Unsupervised hierarchical clustering analysis was conducted using the profiles of significantly regulated genes in each study group (Figure 1C, clusterogram). Severe endometriosis samples clustered together and separately from mild endometriosis samples. Early secretory endometriosis from the mild endometriosis group clustered close to the mild PE group. Remarkably, even though clustering analysis of mild and severe endometriosis samples showed that they clustered separately from each other, signifying the difference between these 2 stages of endometriosis, PE as well as ESE and MSE samples from all groups demonstrated branching from the same stem, supporting the conclusion that cycle phase has greater impact than disease stage in sample clustering (Figure 1C).

Endometrial Transcriptome

Gene Ontology Categories in Severe Versus Mild Endometriosis Throughout the Menstrual Cycle

The most common gene ontology (GO) biological process groups in all comparisons were transcription, transport, cell adhesion, nuclear messenger RNA (mRNA) splicing, proteolysis, translation, and cell cycle, with angiogenesis and apoptosis processes having significant representation in the secretory (ESE and MSE) phase (Table 2). The main cellular components involved were nucleus, cytoplasm, extracellular and intracellular regions, and membranes (Table 2), demonstrating the ubiquitous participation of all cellular components in molecular differences between severe and mild endometriosis. The main GO molecular function categories included nucleotide binding, protein and DNA binding, receptor activity, actin binding, signal transducer activity, and others.

Microarray Validation by Real-Time RT-PCR

Some of the highly up- or downregulated genes, as well as genes dysregulated in all cycle phases between severe and mild endometriosis groups were selected randomly for validation using real-time RT-PCR: DIO2, IGFBP5), VCAN, SLC1A1, SST, TAGLN, and EGFR (Figure 2 ; Supplemental Tables 1-3). Most of the validated genes (VCAN, IGFBP5, SST, DIO2) follow the trend differences of the microarray results.

Analysis of Networks and Canonical Pathways Regulated in Severe Versus Mild Endometriosis

Ingenuity pathway analysis of gene expression profiles revealed several associated network functions identified as different between severe versus mild endometriosis in proliferative, early secretory, and mid-secretory phase samples (Table 3). As expected, several genes are involved in more than 1 network/pathway. Comparison of canonical pathways regulated in severe versus mild endometriosis revealed large differences in eutopic endometrium in these 2 disease stages, as shown by the high number of regulated genes in the pathways. The major canonical pathways regulated are presented in Table 4 .

Major differences in neuregulin signaling, which involves members of the EGFR family, were observed in the proliferative and mid-secretory phases between severe versus mild endometriosis (Figure 3 ). Epidermal growth factor receptor mRNA was upregulated in severe versus mild endometriosis in PE and ESE (Figure 2B), indicating its involvement in severe disease, and confirming our earlier reports of the involvement of EGF family in the pathophysiology of severe endometriosis.^13,24

Epidermal Growth Factor Receptor Protein Immunoreactivity

As presented in Figure 4 and Table 5 , EGFR protein was expressed throughout the menstrual cycle in women with mild as well as severe endometriosis. Interestingly, the most dramatic difference in EGFR protein immunoreactivity was observed in the early secretory phase, where strong stromal expression was observed in severe compared to mild endometriosis, consistent with the real-time RT-PCR data (Figure 2B, Figure 4E,H, Table 5). There was a slight increase in epithelial EGFR immunostaining in the proliferative phase of severe endometriosis samples (Figure 4D,G, Table 5), whereas immunostaining was similar in MSE samples, regardless of the endometriosis stage (Figure 4F, I, Table 5). Immunohistochemical analysis of EGFR in endometrial samples from women without endometriosis throughout the menstrual cycle revealed weak expression of this protein in epithelial and/or stromal compartments (in particular, in ESE stroma), compared to the endometrium from women with endometriosis (Figure 4A-C, Table 5).

Versican Protein Immunoreactivity

In the proliferative phase, there was a strong stromal and epithelial VCAN immunostaining in severe and mild endometriosis, respectively (Figure 5D,G , Table 5). In ESE, similar VCAN immunostaining was observed in the stroma and epithelial compartments (Figure 5E,H, Table 5), and epithelial VCAN immunostaining in MSE tended to be stronger in the severe endometriosis samples (Figure 5F,I, Table 5). Of note, diffuse stromal reactivity of VCAN in both cellular and extracellular compartments was observed. Remarkable was the VCAN immunoreactivity in the vasculature—in the smooth muscle layer and endothelial cells, regardless of disease status, stage, or cycle phase. Versican immunostaining in endometrial tissue from women without endometriosis throughout the menstrual cycle demonstrated overall weaker expression compared to samples from women with mild or severe endometriosis. This was particularly evident in PE epithelium and stroma in patients with severe and mild endometriosis, respectively, as well as MSE epithelium in severe endometriosis samples (Figure 5A-C, Table 5). Positive and negative controls demonstrated specificity of the observed results (Figure 5J-L).

General Comments

The main finding of this study is the demonstrated difference in global gene expression in eutopic endometrium from participants with severe versus mild endometriosis, throughout the menstrual cycle. These 2 endometriosis stages are distinct in their clinical presentation, as well as therapeutic and surgical management, although the corresponding scientific literature is limited. The data herein underscore significant molecular and signaling pathway differences between these 2 stages of endometriosis in distinct hormonal milieu, suggesting that eutopic endometrium in severe versus mild endometriosis has different functional capacities.

Comparison of severe versus mild endometriosis samples by cycle phase revealed the dysregulation of several cyclic adenosine monophosphate (cAMP) and/or progesterone regulated gene, such as downregulation of IHH, SST, and TAGLN in ESE and upregulation of DKK1, MAO, IL15, and IL1R1 in MSE. Upregulation of DIO2 and downregulation of TRH transcripts between severe and mild endometriosis samples indicate potential involvement of thyroid hormone homeostasis and metabolism in the pathophysiology of this endometrial disorder.

Women with severe endometriosis experience higher rates of implantation failure during IVF treatment cycles. Of the 25 human receptivity-related genes identified by analysis of endometrial tissue from healthy fertile women,²⁵ TAGLN and calponin 1 transcripts were dysregulated in MSE from women with severe versus mild endometriosis, suggesting their potential role in the impaired implantation process in women with severe disease.

Dysregulation of Neuregulin Signaling and EGFR in Severe Versus Mild Endometriosis

Of interest is the association of neuregulin signaling with endometriosis. Neuregulin signaling involves ligands for the transmembrane tyrosine kinase receptors ERBB1 (EGFR), ERBB2, ERBB3, and ERBB4—members of the EGFR family.²⁶ Ligand binding activates intracellular signaling cascades and the induction of cellular responses including proliferation, migration, differentiation, and survival or apoptosis in different organs and systems.²⁷ Neuregulin genes, though not regulated themselves in the current study, influence proliferation, migration, and differentiation of epithelial, neuronal, glial, cardiac, and other types of cells.^28–30 This canonical pathway was highly regulated herein in PE and MSE between severe and mild endometriosis samples. Neuregulin (also known as heregulin) signals through HER3 and HER4 receptors; although no changes were observed herein between disease stages or in different cycle phases.

Epidermal growth factor receptor is the major player of neuregulin-signaling pathway. Epidermal growth factor receptor (ERBB1) expression in normal eutopic endometrium on the mRNA and protein levels during different menstrual cycle phases was demonstrated herein and confirms earlier reports.^31–33 We have observed that EGFR gene expression is increased in eutopic endometrium of women with severe endometriosis compared to women without disease in ESE and MSE, but not PE, and is not regulated in mild endometriosis versus nonendometriosis samples throughout the cycle (Aghajanova et al, unpublished data). Herein, we have found that EGFR is dysregulated in severe versus mild endometriosis throughout the menstrual cycle on both mRNA and protein levels, with the most dramatic difference (upregulation) in ESE. Furthermore, transducer of ERBB2 was upregulated in ESE. Thus, the present study demonstrates differences in EGFR expression between different stages of endometriosis and also supports earlier studies noting involvement of EGF family members in the pathophysiology of endometriosis.^13,25,34 Interestingly, EGFR is a tumor marker, particularly for epithelial tumors such as colon cancer, lung cancer, prostate cancer, breast cancer, or other solid tumors.^26,35 Whether it is a marker of a severity of endometriosis remains to be determined.

Dysregulation of ECM Molecules in Severe Versus Mild Endometriosis

This is the first study to demonstrate mRNA expression and immunoreactivity of the ECM proteoglycan VCAN in human endometrium. In participants without endometriosis, VCAN immunoreactivity was weak, especially in PE; whereas, strong immunoreactivity was observed in endometrial stroma from women with severe endometriosis and in epithelium of samples from women with mild disease. Versican can bind to integrins on the cell surface,³⁶ stimulating cell proliferation and inhibiting apoptosis.³⁷ Versican has multiple functions and interactions in different model systems. For example, overexpression of VCAN in a pheochromocytoma cell line upregulates EGFR.³⁸ Also, VCAN expression is increased in endothelial cells with increased migrating capacity.³⁹ Thus, high levels of VCAN may promote an invasive phenotype of endometrial cells in endometriosis by affecting their proliferation, apoptosis, adhesion, and migration, and also may participate in or be causative of the upregulation of EGFR in endometriosis. These functions in endometriosis await further investigation.

Dysregulation of MicroRNAs in Severe Versus Mild Endometriosis

MicroRNA (miRNA) 21 (MIR21) was found to be upregulated on the array, herein, in eutopic endometrium throughout the menstrual cycle in severe versus mild endometriosis. It has recently been shown to be upregulated in eutopic endometrium of women with versus without endometriosis.⁴⁰ Some of the predicted target genes for this miRNA are the tumor-suppressor gene PTEN (downregulated 2.18- and 2.3-fold in severe vs mild endometriosis in PE and MSE, respectively), PDCD4, E2F1, and TGFBRII.^41–43 Interestingly, downregulation of MIR21 inhibits expression of EGFR in human glioblastoma cells.⁴⁴ Whether there is such a mechanism operating in human endometrial stromal fibroblasts (hESF) remains to be determined.

Herein, we observed the upregulation of DICER1 in ESE and MSE from severe versus mild endometriosis (Supplement Tables 2 and 3). The transcript for DICER1 (dicer1, ribonuclease type III), which is a repressor of gene expression due to its involvement in the biogenesis of microRNAs and small interfering RNAs, demonstrates cyclic variation throughout the normal human menstrual cycle.⁴⁵ Female mice with a conditional knockout of Dicer1 in mesenchyme-derived cells of the oviducts and uterus are sterile, in part, due to uterine defects.⁴⁶ Although endometrial stromal Dicer1 expression was absent, the decidualization process was not compromised,⁴⁶ consistent with the recent finding that DICER1 knockdown in hESF does not affect decidualization.⁴⁵ Increased expression of DICER1 in secretory endometrium from women with severe versus mild endometriosis may lead to downregulation of apoptosis-associated genes and dysregulation of adhesion molecules, leading to resistance to apoptosis and increased migratory functions in endometrial cells, as observed with endothelial cells.⁴⁷

Dysregulation of Canonical Pathways in Severe Versus Mild Endometriosis

Several pathways regulated between severe and mild endometriosis are of interest. Severe endometriosis samples exhibited dysregulation of second-messenger signaling pathways, including PI3K/AKT, JAK/STAT, SPK/JNK, and MAPK, confirming recent reports.⁴⁸ Regulation of neurotrophin/TRK (neurotrophic tyrosine kinase) signaling in PE and MSE and axonal signaling in ESE are consistent with the presence of nerve fibers in eutopic endometrium and perhaps their role in the pathogenesis of endometriosis-associated pain.^49,50 However, other functions of these pathways (and their members) may be operating in endometrium, not related to pain. The current study demonstrates differences in these pathways between severe and mild stages of endometriosis, although pain and stage are not necessarily correlated.^51,52

Of note is the involvement of cancer-associated pathways, such as prostate, endometrial, bladder, colorectal, pancreatic cancer, and basal cell carcinoma signaling, suggesting commonalities in the pathophysiology between severe endometriosis and epithelial cancers.

Wnt signaling, NRF2-mediated oxidative stress response signaling (nuclear factor [erythroid-derived 2]-like 2, involved in apoptosis and the oxidative stress response), and retinoid X receptor (RXR) signaling were significantly regulated in secretory endometrium (ESE and MSE; Supplememtal Table 2), probably indicating the differences in the endometrial response to progesterone between severe and mild endometriosis.