Endometrial Stromal and Epithelial Cells Exhibit Unique Aberrant Molecular Defects in Patients With Endometriosis

Abstract

Context:

Endometriosis is a chronic inflammatory disease that causes pain and infertility in women of reproductive age.

Objective:

To investigate the pathologic pathways in endometrial stromal and epithelial cells that contribute to the manifestation of endometriosis.

Design:

In vitro cellular and molecular analyses of isolated eutopic endometrial stromal and epithelial cells.

Methods:

Eutopic stromal and epithelial cells from endometriotic and normal patients were isolated by fluorescence-activated cell sorting for paired sibling RNA sequencing and microRNA microarray. Aberrant pathways were identified using ingenuity pathway analysis networks and confirmed with in vitro modulation of the affected pathways in stromal and epithelial cell cultures.

Results:

Both stromal versus epithelial cell types and paired endometriotic versus normal samples exhibited distinct hierarchical clustering. Compared to normal samples, there were 151 and 215 differentially expressed genes in the endometriotic stromal and epithelial populations, respectively, and concomitantly 9 and 16 differentially expressed microRNAs. Overall, endometriotic stromal and epithelial cells revealed distinct defects. In endometriotic stromal cells, key decidualization genes Zinc finger E-box Binding protein 1 (ZEB1), Heart And Neural crest Derivatives expressed 2 (HAND2), WNT4, and Interleukin 15 (IL-15) were found to be downregulated and Periostin (POSTN) and Matrix Metallopeptidase 7 (MMP7) were upregulated. Specifically, ZEB1 was downregulated in stromal cells by aberrant elevation in miR-200b. In contrast, ZEB1 was found to be upregulated in endometriotic epithelial cells through associated upregulation of transforming growth factor β1 (TGFβ1), inducer of the TGFβ1–Bone Morphogenetic Protein 2 (BMP2)–MMP2–Prostaglandin-endoperoxide Synthase 2 (COX2)–ZEB1 pathway, which activates epithelial–mesenchymal transition.

Conclusion:

Manifestation of endometriosis involves dysregulation of unique molecular pathways within the diseased endometrial stromal and epithelial cells in the endometrium. Targeting the cell type–specific defects may offer a novel approach to treating endometriosis.

Introduction

The extrauterine invasion and growth of estrogen-dependent endometrial cells is the clinical hallmark of patients with endometriosis.^1–3 It is an inflammatory disease that affects approximately 10% of women of reproductive-age,^4,5 with symptoms such as chronic pelvic pain, dysmenorrhea, and infertility, which often precede the diagnosis of endometriosis by an average of 11 years.^3,6,7 Even when diagnosed, there is no current effective and definitive treatment beyond temporizing surgical or medical intervention.

Retrograde menstruation of eutopic endometrial tissues, stem cell origin, coelomic metaplasia, or developmental origins during fetal development have all been hypothesized to be potential causes of endometriosis; however, the precise pathophysiology of endometriosis remains to be understood.^8–16 Recently, molecular profiling studies examining the disease from a variety of approaches have offered critical insights into the basic understanding of endometriosis.^17–22 Microarray studies have revealed differential gene expression in the eutopic endometrium between endometriotic and normal, between ectopic and eutopic tissues, and comparisons made between the phases of the menstrual cycle.^17,23–27 Similarly, epigenetic mechanisms, which often mediate environmental factors,^28–31 such as DNA methylation,^32–37 histone modification,^38–41 long noncoding RNA,^22,42 and microRNA,^43–49 have also been investigated.

However, there are inconsistencies and a general lack of consensus among studies for specific genes and microRNAs involved in the reported aberrant molecular pathways.^17,45 Perhaps this lack of clarity is due to the complex nature of this disease and the diversity of the endometrial tissues used in these studies.¹⁷ Specifically, molecular studies often use tissues collected from different phases of the menstrual cycle and predominantly utilize whole endometrial biopsies consisting of mixed populations of endothelial, stromal, and epithelial cells for analyses.^{25,46,50–52} While informative, it is challenging to resolve the specific impact of the menstrual cycle and how individual cell types within the endometrium contribute to the overall pathophysiology of the disease.^53–55 Since endometrium from patients was shown to share similar defects to ectopic endometrium (although there are some differences), the studies of isolated stromal and epithelial cells from the eutopic endometrium could minimize the heterogeneity seen among current studies.^23,51

Endometrial stromal and epithelial cells have separate yet important interacting functions. Stromal cells are fibroblast-like cells that differentiate into larger, rounder decidual cells during the secretory phase of the menstrual cycle in order to be receptive and nourish an embryo ready to implant inside the endometrium.^56,57 Failure of stromal cells to decidualize properly has consequences such as impaired fertility, placenta accreta, poorly developed placenta, or miscarriage.^58–60 Meanwhile, epithelial cells form a protective lining of the lumen interface of the endometrium and glandular ducts and are the point of attachment for an embryo to implant.⁶¹ A 2-way paracrine interaction exists between stromal and epithelial cells, in particular, retinoic acid signaling from stromal to epithelial cells, and conversely, the Hedgehog pathway paracrine signaling from epithelial cells to stromal cells that regulates WNT family member 4 (WNT4), a critical protein for decidualization.^62–66

In this study, paired eutopic endometrial stromal and epithelial cells from diseased (endometriotic) and normal endometrial biopsies were used to investigate the cell type–specific aberrant molecular pathways involved in the pathogenesis of endometriosis.

Materials and Methods

Tissue Collection

Endometrial tissue was collected, after informed written consent, from normal patients and patients with endometriosis (Supplementary Table 1) under a protocol approved in writing by the Committee on Human Research of the University of California, San Francisco (UCSF). All patients had regular menstrual cycles and were in the proliferative phase. Endometriotic tissue was confirmed by laparoscopy, with histologic evaluation of normal and stage I to IV endometriosis. Staging of endometriosis was defined according to the revised American Fertility Society classification system. Exclusion criteria were tissues from patients with fibroids, endometrial polyps, pregnancy, or any hormonal medications known to interfere with the menstrual cycle either directly on the uterus or centrally for at least 3 months before tissue collection.

Tissue Digestion

Endometrial tissue from more than 29 patients was digested according to established protocol. The fresh tissue was minced and digested with 3.15 mg/mL collagenase I (Worthington Biochemical, Lakewood, New Jersey) and 40 µg/mL deoxyribonuclease I (DNase I; Sigma Aldrich, St Louis, Missouri) in digestion media (Dulbecco’s minimum essential medium [DMEM]/F12 phenol red-free supplemented with 10% fetal bovine serum [FBS]; Gibco, Life Technologies, Carlsbad, California) and 1% antimycotic (penicillin, streptomycin, fungizone [amphotericin B]; Cell Culture Facility, UCSF) on a rotator at 37°C for 1 hour. The digestion was filtered through a 40-μm cell strainer. Both the flow-through fraction and the backwash fraction were collected in 10 mL digestion media and centrifuged. The first fraction pellet was reconstituted and the second fraction incubated with 0.8 mg/mL collagenase II (Worthington) and 40 µg/mL DNase I at 37°C for 20 minutes and centrifuged, after which the pellet was incubated with 2 mL Accutase (Life Technologies) at 37°C for 10 minutes. This fraction was filtered through a 70-μm cell strainer, centrifuged, and the pellet reconstituted.

Fluorescence-Activated Cell Sorting of Human Endometrial Stromal and Epithelial Cells

To sort into stromal fibroblast and epithelial cells, digested cells were labeled with antibody staining, incubated for 30 minutes at 37°C, and sorted by fluorescence-activated cell sorting (FACS) AriaII UV, with FACSDiva software (BD Biosciences, San Jose, California; DRC Center Grant [NIH P30 DK063720] Parnassus Flow Cytometry Core, UCSF). Otherwise, all cells were kept on ice during processing and after the FACS sorting. Antibody staining to isolate stromal cells was CD90 (BioLegend, San Diego, California) and epithelial cells was EpCAM PerCP-Cy5.5 (CD326, cat #347199; Becton Dickinson, Franklin Lakes, New Jersey). All cells were labeled with FITC CD146 (cat # 560846; BD Pharmingen, San Jose, California) to remove endothelial cells, Pacific Blue CD45 (cat # 304029; BioLegend) to remove leukocytes, and Sytox Blue (Life Technologies) to remove dead cells. The FACS cell sort was analyzed using FlowJo v9.6 (FlowJo, Ashland, Oregon). (Fluorescence-activated cell sorting selects large quantities of cells purely compared to laser capture of cells).

Isolation of Stromal and Epithelial Cells

To compare diseased versus normal endometrial stromal and epithelial cells, these 2 cell types were isolated from endometrial biopsies from both diseased and normal patients (Figure 1A). Only live nonleukocytic endometrial cells were gated for initial analysis. Subsequently, only enriched stromal fibroblast cells (CD90⁺) and epithelial cells (EPCAM⁺) were collected after the exclusion of endothelial cells (Figure 1B). Immunocytochemistry confirmed the purity of stromal and epithelial cells with vimentin staining of stromal cells (C9080; Sigma Aldrich) and pan-cytokeratin of epithelial cells (M0821; DakoCytomation, Santa Clara, California). RNA was collected from the sibling stromal and epithelial population from each patient sample, and each RNA sample was further divided equally for paired RNA and microRNA analyses with RNA sequencing (RNA-Seq) and microRNA microarray, respectively (Figure 1A).

Figure 1.

A, Isolation of stromal and epithelial cells from endometriotic and normal tissue for sibling-paired RNA-Seq and micro-RNA microarray analyses; each sample was independently sorted and analyzed. B, Isolation of stromal fibroblasts (CD90⁺) and epithelial (EpCAM⁺) cells by fluorescence-activated cell sorting (FACS) and analyzed by FlowJo v9.6. Only single, CD45⁻/CD146⁻ live cells were gated for stromal fibroblast and epithelial cell isolation (leukocytes and endothelial cells were excluded). FSC-A indicates forward scatter area; SSC-A, side scatter area.

Cell Culture

Stromal cells were cultured in DMEM/F12 phenol red-free 10% FBS media (Gibco, Life Technologies) with 1% antimycotic (penicillin/streptomycin/fungizone). Epithelial cells were cultured in defined keratinocyte serum-free media supplemented with l-glutamine and initially 28% FBS (Life Technologies). All cells were incubated at humidified 37°C and 5% CO₂. Cell cultures of both stromal (n = 4) and epithelial cells (n = 6) were used for transient transfection studies of microRNA and their target genes.

RNA Isolation

Total RNA was isolated for RNA-Seq and microRNA microarray by RNeasy Plus Mini Kit (Qiagen, Valencia, California). High-quality RNA for both assays was confirmed on a Bioanalyzer 2100 (Bioanalyzer; Agilent Technologies, Santa Clara, California), and the samples had a minimum RNA integrity number of 8.8. Otherwise, total RNA for quantitative polymerase chain reaction (qPCR) was isolated by TRI-reagent (Sigma-Aldrich) following the manufacturer’s instructions.

RNA Sequencing Expression Profiling

Endometrial cells from 3 patients with endometriosis and 3 normal participants were sorted by FACS into stromal and epithelial cells (Figure 1A). Cell type–specific diseased and normal samples were processed individually during RNA isolation. Each sample was divided equally for RNA-Seq and microRNA microarray. TopHat and Cufflinks were used to process genome-wide sequencing data and Cuffdiff to obtain values for FPKM (fragments per kilobase of exon per million fragments mapped). Significant messenger RNA (mRNA) differential expression was defined as >2-fold change and Cuffdiff P < .05.

Pathway Analysis of Different Gene Networks in Endometriotic Stromal and Epithelial Cells

Pathway analysis of the differentially expressed genes profiled by RNA-Seq from stromal and epithelial cells was performed using ingenuity pathway analysis (IPA; Qiagen, Redwood City, California, www.qiagen.com/ingenuity). The DAVID Bioinformatics Resource (v6.7) was also used to analyze the stromal and epithelial data sets (Database for Annotation, Visualization and Integrated Discovery. [NIAID] NIH. Huang et al 2009).

microRNA Microarray Expression Profiling

The microarray was conducted on an Affymetrix GeneChip miRNA 3.0 Array (Affymetrix, Santa Clara, California). Unique reads were aligned to human microRNA sequences from miRBaseGv17 (www.mirbase.org). The microarray detected more than 1300 microRNAs (Supplementary Table 2). Significant microRNA differential expression was defined as ≥1.5-fold change and Students t test P ≤ .05.

Real-Time qPCR for mRNA and microRNA

Validation of RNA-Seq and microarray was performed by qPCR of stromal cell differential mRNA (n = 4), microRNA (n = 3), and epithelial mRNA (n = 5), and microRNA (n = 6). Total RNA was converted to complementary DNA (cDNA) by qScript SuperMix (Quanta Biosciences, Gaithersburg, Maryland) for mRNA expression and qScript microRNA cDNA Synthesis (Quanta Biosciences) for microRNA by following the manufacturer’s instructions. FastStart SYBR-Green ROX (Roche Diagnostics, Indianapolis, Indiana) was used for mRNA expression and PerfeCTa SYBR Green Supermix Low-ROX (Quanta Biosciences) used for microRNA expression. Quantitative PCR was performed in an Applied Biosystems ViiA7 real-time PCR system (Life Technologies). Primers used for qPCR are listed in Supplementary Table 3. The ΔΔCT method was used to calculate the relative quantity of transcripts. The reference genes for mRNA qPCR were selected as suitable for 2 different cell types from our previous studies: RNA18S5 for stromal, Glyceraldehyde 3-Phosphate Dehydrogenase (GAPDH) for epithelial cells, and Small Nucleolar RNA C/D box 44 (SNORD44) for microRNA qPCR.

microRNA Target Genes

The selection of predicted mRNA target genes of miR-200b and miR-204 for stromal cells and miR-504 and miR-1827 for epithelial cells from our microarray data was based on DIANA-lab MicroT-CDS (www.diana.imis.athena-innovation.gr/DianaTools/index.php) and TargetScanHuman v6.2 (www.targetscan.org). The predicted targets of microRNAs were matched with our RNA-Seq data set of differentially expressed mRNA (Supplementary Table 4). Transfection conditions for mimics (overexpression) and antagomirs (inhibition) of selected microRNAs (Supplementary Table 5) were 15 nM for mimics, incubated for 48 hours, and 50 nM for antagomirs, incubated for 72 hours. The transfection reagent DharmaFECT 1 (T-2001-03; GE Dharmacon, Layfayette, Colorado) was applied at 1 µL per transfection in 6-well plates. Nontargeting mimic and hairpin inhibitor were negative controls.

Statistical Analysis

Data were analyzed and graphed using GraphPad Prism v6.0.5 for Windows (GraphPad Software, La Jolla, California, www.graphpad.com). Results are expressed as mean (standard deviation). Statistical significance was determined by Student unpaired, 2-tailed t test and for groups of 3 or more by 1-way analysis of variance with Benjamini-Hochberg multiple testing correction for false discovery rate: for RNA-Seq and microarray assays; n ≥ 3 samples for each experiment.

Results

Differential Expression of Genes in Endometriotic Stromal and Epithelial Cells

To determine the differential gene expression between endometriotic and normal stromal cells, and similarly for epithelial cells, RNA-Seq was performed on paired sibling samples of stromal and epithelial cells. On average, there were 51 501 530 and 45 364 780 total reads of >23 000 genes from stromal and epithelial samples, respectively. Hierarchical clustering analyses revealed that there were distinct, differentially expressed genes between stromal and epithelial cells, regardless of whether they were endometriotic or normal (Figure 2A). Additionally, distinct clustering was also observed between endometriotic and normal stromal and epithelial endometrial cells (Figure 2A). Endometriotic stromal and epithelial cells had 151 (52 upregulated) and 215 (80 upregulated) differentially expressed genes, respectively (Figure 2B; Table 1). There were only 5 common differentially regulated genes between diseased stromal and epithelial cells (upregulated: Colony Stimulating Factor 3 (CSF3), Serpin family A member 1 (SERPINA1), ASH1 Like histone lysine methyltransferase (ASH1L)⁶⁷; downregulated: Phosphoinositide-3-Kinase Regulatory subunit 1 (PIK3R1), latent TGFβ1-binding protein 4 (LTBP4); Figure 2B). Zinc finger E-Box Binding homeobox 1 (ZEB1), Heart And Neural crest Derivatives expressed 2 (HAND2), and C-X-C motif Chemokine Ligand 12 (CXCL12) were confirmed to be downregulated, while Periostin (POSTN), SERPINA1, and CSF3 were upregulated in diseased stromal cells by qPCR, consistent with RNA-Seq data (Figure 2C). The qPCR also confirmed upregulation of ZEB1, Integrin subunit beta 6 (ITGB6), CSF3, Bone Morphogenetic Protein 2 (BMP2), and Fms related tyrosine kinase 1 (FLT1) and downregulation of PIK3R1 in diseased epithelial cells (Figure 2D). Interestingly, ZEB1 was downregulated in diseased stromal cells, while it was upregulated in diseased epithelial cells.

Figure 2.

A, Dendrogram of hierarchical clustering analysis (HCA) of messenger RNA (mRNA) data from RNA-Seq; q1-q4 are sample queries. B, Venn diagram of mRNA expression from RNA-Seq of endometriotic compared to normal stromal and epithelial cells; selection by >2-fold change, CuffDiff, P < .05. C, Endometriotic stromal genes from RNA-Seq data set validated by quantitative polymerase chain reaction (qPCR), normalized with 18SRNA, n = 4. D, Endometriotic epithelial genes from RNA-Seq data set validated by PCR, normalized with glyceraldehyde 3-phosphate dehydrogenase (GAPDH), n = 5. Mean (Standard deviation). *P ≤ .05, **P ≤ .01.

Table 1.

Differentially Expressed Genes in Endometriosis From RNA-Seq: 151 Endometriotic Stromal Genes and 215 Epithelial Genes.^a

Stromal Genes	Log2 Change
ABCA8	−2.593
ABCC8	−3.511
ADAMTS15	−2.024
ADAMTS5	−1.818
ADCY1	−1.983
AHNAK2	−1.364
ALDH1A3	−1.548
ALDH1L2	−1.405
ALPK3	2.736
ANKRD13B	−2.648
AP2A1	−1.517
APBB1IP	−3.899
ASH1L	2.699
ATP8A2	−2.044
BCAT1	−1.321
BMPER	−2.231
BRINP1	−3.483
BTBD2	−1.770
BTBD3	−1.359
C10orf128	4.444
C1orf95	−3.335
C2	−2.545
C20orf27	−2.313
C3	−1.644
CBS	−2.425
CD93	1.683
CDH6	1.873
CFD	−4.148
CILP	−3.583
CLDN11	2.847
CLEC3B	−3.926
COL17A1	−6.048
COL24A1	1.537
COL4A5	−1.956
COL4A6	−4.199
COLEC11	−2.198
CPE	2.253
CRABP2	1.616
CREB5	1.143
CSF3	2.672
CXCL12	−1.642
CYP1A1	4.355
DEPTOR	−3.093
DIO3	−3.555
DLL4	1.922
DUSP10	1.735
EMILIN2	−1.241
EMR2	1.452
EPHA7	1.962
ETV5	1.466
FAM134B	−1.702
FBLN2	−1.631
FERMT1	−2.387
FGF10	−3.295
FMNL3	1.528
FNDC1	2.163
FOXF1	1.999
G0S2	1.782
GABRA2	−2.857
GALNT13	−2.554
GFRA2	3.334
GLB1L2	−4.410
GNG4	−1.477
GPBAR1	−2.955
GPR116	2.633
GPR133	−2.057
GPR64	−2.186
GRIA4	−2.102
H6PD	−1.248
HAND2	−1.388
HSDL2	−1.681
HSPB6	−2.728
IGBP1P1	4.444
IGFBP2	−1.445
IL15	−2.158
ITGB8	−1.196
KAL1	−2.639
KCND3	−3.827
KCNH1	−2.149
KCNK3	−2.210
LAPTM5	2.705
LGI2	−2.298
LIMCH1	−1.817
LRRC15	−2.763
LTBP4	−1.659
LYNX1	−3.025
MAOB	−1.831
MICAL1	−1.303
MMP7	2.081
MPZL2	3.159
MROH6	−2.577
MTR	−1.293
MYD88	−1.295
MYLK3	−1.826
MYOCD	−1.867
NMB	1.925
NPW	−3.533
NRARP	2.329
NRG1	1.751
NRXN1	−2.261
OAS2	−1.956
PAPLN	−1.291
PCDH20	−3.091
PDE3A	1.825
PDGFA	1.847
PDLIM3	2.191
PFKFB4	2.181
PI15	3.574
PIK3R1	−1.145
PILRA	−3.592
PIM2	2.249
PLA2G7	−2.310
PLCB1	3.325
PLCD1	−2.120
PLCL1	−1.446
PLXND1	−1.507
POSTN	1.973
PPM1H	−1.617
PPP1R14A	−2.849
PPP4R4	3.947
PROK1	−5.321
PTCH1	−1.619
PTGDS	−1.830
PTPRB	1.853
PYGL	−1.741
RAPGEF4	2.370
RASD1	−1.462
RASGRP3	2.266
REXO1	3.576
RPRM	−3.173
S100A7	4.444
SCARA5	−6.773
SDK1	−1.382
SELE	1.910
SERPINA1	3.799
SHB	1.492
SLC2A8	−2.747
SLC40A1	−1.515
SMIM3	1.509
SPATA18	−2.621
STEAP1	2.559
SULF2	−1.292
TMEM132B	−1.692
TMEM196	−3.447
TMEM37	−2.290
UBA7	−5.969
UCN2	1.774
WNT4	−2.081
XIRP1	2.145
ZC3H12C	1.229
ZNF703	−1.630
Epithelial genes	Log2 Change
ABCA6	−2.541
ABCD1	−2.951
ABL2	1.543
ACPP	−2.450
ACTG2	−3.924
ADAMTS1	−1.221
ADAMTS8	−6.935
AKR1C2	3.146
ALDH3B2	−4.806
ALOX5	1.839
ALPL	−1.593
ANO6	1.160
AOX1	−2.827
AQP9	2.792
ARHGAP15	−4.444
ARNTL2	1.206
ASH1L	2.792
ASS1	1.359
ATP10B	−4.313
ATP6V0E2	−2.908
AXL	2.646
BMP2	3.687
C14orf183	4.444
C1orf64	−4.444
C2orf54	2.703
CABYR	−2.156
CAMP	−4.444
CAPN14	2.627
CAPN6	−2.525
CBX4	−1.776
CCBL1	−3.111
CCDC144B	1.692
CD36	−4.850
CD40	3.180
CDK6	2.611
CENPF	−2.349
CES4A	−2.668
CHST10	−2.045
CIC	−1.832
CLGN	−1.896
CMYA5	−2.098
COL1A2	−3.050
COL27A1	2.500
COL6A2	3.924
COL6A3	2.396
CREB3L1	−2.739
CSF3	2.628
CST6	2.768
CXCL5	2.128
CXorf27	4.444
CYP26A1	−4.647
DFNB31	−1.719
DLX5	−1.397
DNAJA4	−1.319
DPP4	−1.609
DPYSL3	2.766
DUOX1	−2.608
DUOXA1	−2.589
DYNC1I1	−2.271
ECI2	−2.016
EDNRB	−2.450
EEF2K	−1.403
ELK3	1.259
ELMSAN1	−1.384
ELP3	−2.000
EML5	−2.105
EMP1	1.247
EMP3	2.144
EPN3	−3.353
ETV1	−2.130
FAM129A	1.802
FAM171B	−2.418
FAM172A	3.709
FGB	2.721
FGF2	2.273
FILIP1	−3.850
FJX1	2.023
FLT1	2.257
FMOD	2.986
FN1	2.144
GCNT1	1.529
GJB4	3.413
GLYATL2	−5.294
GLYATL3	−4.444
GMPR	−2.515
GPR141	4.444
GRAMD1C	−1.629
GREM2	−2.130
HEY1	−2.369
HGD	−2.045
HPGD	−1.925
HS6ST1	−1.659
HSD11B2	−2.567
HSD17B2	−2.184
HSPA6	−1.878
HYAL1	−2.535
IDH1	−2.126
IGF2	2.123
IGFBP5	2.024
IL1B	2.942
IL1R2	2.076
IL1RN	1.675
IL20RA	−2.481
ITGB6	1.352
KAT2B	−1.876
KCTD12	2.115
KLF2	−2.404
KRT13	2.134
KRT17	1.444
KRT7	1.294
LSAMP	2.916
LTBP4	−2.107
MAOA	−1.668
MAP2K6	−2.859
MARCKS	1.322
MESP1	−4.444
MMP2	2.872
MMP26	−1.956
MSX1	−1.273
MT1G	−2.416
MT1H	−3.334
MT1M	−2.949
MTSS1	−2.023
MUC5B	3.391
MXRA5	1.560
NAV3	2.895
NDC80	−3.144
NEK6	−1.323
NLRP5	−4.444
NME4	−1.641
NOS2	4.861
NPR3	−4.013
NQO1	−1.331
NRP2	2.272
OGDHL	−2.263
OPRK1	−2.528
PAPSS1	−1.229
PAX2	−1.976
PCSK5	1.925
PEBP4	−4.444
PER1	−1.715
PIK3R1	−1.931
PIK3R2	−1.558
PKHD1L1	−2.633
PLLP	−1.501
PLXNA4	−4.199
PRICKLE1	−2.119
PROL1	−4.444
PSG9	−5.337
PTGS2	1.792
RAB20	−2.004
RAB31	1.925
RAP1GAP	−2.171
RBP1	1.920
RNASE1	−3.146
RYR3	−2.010
SAPCD2	−2.872
SCGB1D2	−2.090
SCGB1D4	−4.296
SCNN1B	2.356
SERPINA1	1.201
SERPINA5	−3.252
SERPINB2	5.797
SERPINB5	3.364
SERPINE2	2.326
SFTPA2	−4.444
SHISA2	2.114
SHISA8	−5.245
SLC15A2	−2.851
SLC25A1	−1.564
SLC25A15	−3.775
SLC25A38	−1.395
SLC25A48	−4.444
SLC26A9	2.495
SLC39A8	−2.066
SLC43A1	−3.262
SLC46A2	−4.623
SLC47A1	2.041
SLC6A14	2.618
SLC7A4	−3.340
SLC7A8	−1.826
SLCO2A1	2.901
SMAD9	−1.745
SPARC	2.106
SPDEF	−2.606
SPOCD1	2.612
SPOCK1	1.774
SPON1	−1.533
SQLE	−1.358
ST3GAL3	−2.045
ST6GALNAC1	−2.143
STEAP4	−1.486
STRA6	3.899
SUFU	−2.606
TCEA3	−3.842
TEX36	4.444
TGFBI	4.850
TGM2	1.806
THEM4	−1.444
TMEM101	−1.617
TMEM132C	−2.257
TMEM252	−3.294
TMEM42	2.413
TOP2A	−2.279
TPD52L1	−1.420
TRPC6	−2.375
TRPM8	−2.202
TSPAN12	−1.461
TTLL1	−3.297
UBAP2	3.475
USP54	−1.223
WFS1	−1.482
ZEB1	2.103
ZNF319	−2.254
ZNF589	−1.718

^aSelection by >2-fold change, P < .05.

Differential Expression of microRNAs in Endometriotic Stromal and Epithelial Cells

To determine the specific microRNA expression between diseased and normal stromal and epithelial cells, the same RNA samples that were used for RNA-Seq were also used for microRNA microarray for paired microRNA–mRNA analyses. Hierarchical clustering of these data revealed that differentially expressed microRNAs in stromal cells were distinct from microRNAs in epithelial cells and that differentially expressed microRNAs in endometriotic samples were distinct from microRNAs in normal samples (Figure 3A). Compared to normal samples, there were 9 (6 upregulated) and 16 (8 upregulated) differentially expressed microRNAs in diseased stromal cells and epithelial cells, respectively (Figure 3B; Table 2). Specifically, miR-200b and miR-204 were confirmed to be upregulated in diseased stromal cells, while miR-1827 was confirmed to be upregulated and miR-504 downregulated in diseased epithelial cells by qPCR (Figure 3C).

Figure 3.

A, Dendrogram of hierarchical clustering analysis of microRNA data from microRNA microarray, q1 to q4 are sample queries. B, Venn diagram of microRNA expression from microRNA microarray of endometriotic stromal and epithelial cells; selection by >1.5-fold change, P ≤ .05. C, MicroRNAs from microRNA microarray of endometriotic stromal (n = 3) and epithelial cells (n = 6) validated by quantitative polymerase chain reaction (qPCR), normalized with SNORD44. D, Ingenuity pathway analysis (IPA) of Akt/ERK/P38MAPK/TNF/NF-κB network of endometriotic stromal messenger RNA (mRNA). E, Venn diagram of mRNA expression from RNA-Seq of endometriotic stromal and epithelial genes with subsets of predicted microRNA target genes. F, The IPA analysis of microRNA predicted target gene network in endometriotic stromal cells. G, The IPA analysis of mRNA transforming growth factor β1 (TGFβ1)/interleukin 1B (IL-1B) network in endometriotic epithelial cells. H, The IPA analysis of microRNA predicted target gene network in endometriotic epithelial cells. Shaded genes are statistically significant. Shading intensity of a gene corresponds with the degree of up- or downregulation in endometriosis. Red-colored genes are upregulated in endometriosis and green-colored genes are downregulated. Mean (standard deviation). *P ≤ .05.

Table 2.

microRNA Differential Expression in Endometriotic Stromal and Epithelial Cells.^a

Stromal		Epithelial
Up	Fold Change	Up	Fold Change
miR-200b	2.93	miR-936	1.57
miR-204	2.19	miR-1827	1.98
miR-629	1.77	miR-3141	3.76
miR-3926	1.49	miR-3938	1.51
miR-3928	1.69	miR-4266	1.59
miR-4670	1.66	miR-4450	1.83
		miR-4522	1.64
		miR-4792	2.27
Down		Down
miR-548al	0.28	miR-375	0.21
miR-1297	0.52	miR-504	0.53
miR-4795	0.57	miR-548n	0.54
		miR-601	0.61
		miR-603	0.56
		miR-3136	0.51
		miR-3176	0.59
		miR-4476	0.49

^aSelection by >1.5-fold change, P ≤ .05.

Aberrant Pathways Involved in Endometriosis

Potential molecular pathways involved in both endometriotic stromal and epithelial cells were evaluated by IPA software and by DAVID Bioinformatics (Table 3). Based on the differentially expressed genes between the endometriotic and the normal stromal populations, IPA identified 44 genes involved in the Protein Kinase B (Akt)/Extracellular signal-Regulated Kinases (ERK)/Mitogen Activated Protein Kinase 14 (P38MAPK)/Tumor Necrosis Factor (TNF)/Nuclear Factor kappa B (NF-κB) decidualization pathways, such as WNT4, HAND2, POSTN, Interleukin 15 (IL-15), Complement Factor D (CFD, SERPINA1, and Matrix Metallopeptidase 7 (MMP7) (Figure 3D). The IPA identified the differentially expressed genes within the diseased stromal population that function to interfere with normal immune cell trafficking (z = −2.190), with myeloid differentiation primary response 88 (MYD88) as the master upstream regulator together with IL-15, Complement C3 (C3), Plexin D1 (PLXND1), and CXCL12. Additionally, the differentially expressed MMP7, POSTN, SERPINA1, and Insulin like Growth Factor Binding Protein 2 (IGFBP2) genes also function to modulate cell migration (2.186; Table 3). DAVID classified the main altered functional groups of stromal cells into signal (64 genes), extracellular (29 genes), cell adhesion (17 genes), vasculature development (12 genes), and neuropeptide signal (7 genes) (Table 3; Supplementary Table 6). Diseased stromal cells have 5 differentially expressed microRNAs that have 33 predicted target mRNAs within the same differentially regulated genes found in RNA-Seq (Supplementary Table 4), represented as a subset in the Venn diagram (Figure 3E). The predicted TNF/ERK/p38MAPK/NF-κB pathways based on the differentially expressed microRNAs and their target mRNAs in endometriotic stromal cells (Figure 3F) suggests that upstream microRNA regulation plays an important role in impaired decidualization, with key genes Myosin Light chain Kinase 3 (MYLK3), ITGB8, Patched 1 (PTCH1), and WNT4 being downregulated. WNT4 regulates the critical Forkhead box O1 (FOXO1) transcription factor via the canonical β-catenin signaling pathway.⁶⁶

Table 3.

Network and Gene Analysis of Differentially Expressed mRNA, from RNA-Seq of Endometriotic Stromal and Epithelial Samples, by Ingenuity Pathway Analysis (IPA) and DAVID.

Ingenuity Pathway Analysis
z Score	Function	Genes
Stromal
−2.190	Immune cell trafficking decrease	MYD88 upstream
−2.366	Antigen-presenting cell movement decrease	C3, CXCL12, IL15, MYD88, PLXND1, PTGDS, SELE
−2.000	Mast cell migration decrease	C3, CXCL12, IL15, PIK3R1
−2.216	Macrophage cell movement decrease	C3, CXCL12, IL15, PLXND1, PTGDS
−2.177	Phagocyte maturation decrease	CSF3, IL-15, MYD88, PIK3R1, SERPINA1
−2.219	Inflammatory disease decrease	C3, CXCL12, DLL4, MYD88, PIK3R1
2.321	Cell morphology increases cell size	CBS, DEPTOR, PIM2, PLCB1, UCN2, IGFBP2
2.186	Muscle cell migration increase	IGFBP2, MMP7, MYOCD, POSTN, SERPINA1
Epithelial
2.268	Cancer increase	TGFβ1 upstream
2.537	Tumor cell proliferation increase	AXL, CST6, IL1RN, MMP2, NOS2, PTGS2 (COX2), ZEB1
2.420	Vascularization increase	ADAMTS1, ALOX5, FGF2, IGF2, IL1B, NOS2, NRP2, SPARC
2.803	Cell migration increase	ALOX5, AXL, CDK6, CHST10, CXCL5, FGF2, FMOD, GCNT1, HEY1, IGF2, IL1B, NOS2, NQO1, NRP2, RAP1GAP, SPARC, SPDEF, TGM2
2.608	Leukocyte migration increase	ALOX5, AQP9, CXCL5, FGF2, GCNT1, IGF2, IL1B, NOS2, NQO1, NRP2, RAP1GAP, SPARC, SPDEF, TGM2
DAVID
Function	Gene Number	Enrichment	P Value	Benjamini
Stromal	150 genes
Signal	64	8.34	2.3 × 10−13	5.9 × 10−11
Extracellular region	29	5.03	4.9 × 10−8	4.6 × 10−6
Cell adhesion	17	3.54	4.5 × 10−4	7.4 × 10−2
Vasculature development	12	2.66	1.4 × 10−5	4.2 × 10−3
EGF-like region	10	2.17	7.5 × 10−4	1.5 × 10−1
Neuropeptide signal	7	1.88	1.7 × 10−4	4.1 × 10−2
Carbohydrate binding	11	1.83	7.2 × 10−4	1.2 × 10−1
Hormone regulation	8	1.78	3.9 × 10−4	7.4 × 10−2
Epithelial	213 genes
Extracellular	35	5.41	8.1 × 10−8	1.9 × 10−5
Signal	48	5.37	1.6 × 10−9	5.6 × 10−7
Vasculature development	11	2.44	1.1 × 10−3	1.3 × 10−1
Oxidoreductase	19	2.27	4.7 × 10−5	3.3 × 10−3

Of the 215 differentially regulated genes from the diseased epithelial cells, there were 87 genes identified from the transforming growth factor β1 (TGFβ1) family, involved in the TGFβ1/interleukin 1B (IL-1B) pro-inflammatory pathways (Figure 3G). Additionally, IPA identified an overall increase associated with cancer development (z = 2.268), with TGFβ1 as an upstream regulator, which encompassed tumor cell proliferation (2.537), leukocyte migration (2.608), cell migration (2.803), and vascularization (2.420; Table 3). The critical genes TGFβ1, ZEB1, BMP2, MMP2, and Prostaglandin-endoperoxide Synthase 2 (PTGS2, COX2), which were all upregulated in RNA-Seq, activate the epithelial–mesenchymal transition (EMT) pathway (Table 1). DAVID classified the 215 differentially regulated genes into the main functional gene groups of signal, (48 genes) extracellular, (35 genes) and vasculature development, (11 genes) which is similar to stromal cells but with a different set of genes (Table 3; Supplementary Table 6). The differentially expressed microRNAs in the diseased epithelial cells shared 97 predicted target genes found to be differentially expressed based on RNA-Seq analyses of the diseased epithelial cells (Supplementary Table 4). Of these 97 target genes, 24 target genes represent the main functional groups from the IPA analysis of the TGFβ1/IL-1B/Akt/ERK pathway (Figure 3H).

Target Genes of miR-200b and miR-204 in Stromal Cells

To investigate and confirm the target genes of differentially expressed microRNAs involved in the aberrant pathways described, the levels of miR-200b and miR-204 within the diseased and normal stromal cells were modulated with specific mimics and antagomirs by transient transfection. The qPCR confirmed that mimics and antagomirs could modulate the levels of miR-200b and miR-204 within the stromal cells in vitro as expected (Figure 4A). Upregulation of miR-200b by mimic downregulated the expression of target genes ZEB1, PTCH1, Forkhead box F1 (FOXF1), and CAMP Responsive Element Binding protein 5 (CREB5); however, the sequestration of miR-200b through antagomir upregulated only ZEB1 (Figure 4B). Conversely, Aldo-Ketose Reductase family 1 member C1 (AKR1C1) was upregulated by miR-200b mimic and downregulated by miR-200b antagomir. The miR-204 mimic and antagomir downregulated PTCH1 and upregulated CREB5 expression, respectively (Figure 4B).

Figure 4.

A, miR-200b and miR-204 expression after transfection of mimics and antagomirs into endometriotic and normal stromal cells validated by quantitative polymerase chain reaction (qPCR), normalized with SNORD44. B, miR-200b and miR-204 target gene expression analyzed by qPCR for endometriotic compared to normal stromal cells, normalized with 18SRNA, n = 4. C, miR-504 and miR-1827 expression after transfection of mimics and antagomirs into endometriotic and normal epithelial cells validated by qPCR, normalized with SNORD44. D, miR-504 and miR-1827 target gene expression analyzed by qPCR for endometriotic compared to normal epithelial cells, normalized with glyceraldehyde 3-phosphate dehydrogenase (GAPDH), n = 6; Mean (standard deviation). *P ≤ .05, **P ≤ .01. antagomir Endo indicates antagomir endometriosis; antagomir N, antagomir normal; mimic N, mimic normal; mimic Endo, mimic endometriosis; AKR1C1, Aldo-Keto Reductase family 1 member C1; CREB5, CAMP Responsive Element Binding Protein 5; ELMSAN1, ELM2 And Myb/SANT Domain containing 1; FOXF1, Forkhead box F1; NQO1, NAD(P)H Quinone dehydrogenase 1; NRP2, Neuropilin 2; PLXNA4, Plexin A4; PTCH1, Patched 1; ZEB1, Zinc Finger E-Box Binding homeobox 1.

Target Genes of miR-504 and miR-1827 in Epithelial Cells

Similarly, miR-504 and miR-1827 mimics and antagomirs were transiently transfected into diseased and normal epithelial cells, and qPCR confirmed their altered expression following modulation (Figure 4C). Selected predicted target genes of miR-504 and miR-1827 in the epithelial cell population were analyzed (Figure 4D). The miR-504 mimic downregulated Neuropilin 2 (NRP2) in endometriotic and normal epithelial cells. The miR-1827 mimic downregulated ELM2 And Myb/SANT domain containing 1 (ELMSAN1) and PLXNA4 in normal epithelial cells and NAD(P)H Quinone dehydrogenase 1 (NQO1) in endometriotic epithelial cells. Conversely, miR-1827 antagomir upregulated ELMSAN1 in normal epithelial cells.

Discussion

This study demonstrates that endometriotic stromal and epithelial cells are associated with uniquely aberrant molecular pathways that were regulated by microRNAs. In endometriotic stromal cells, these associated aberrant pathways interfere with immune cell trafficking and impair decidualization through the Akt/ERK/P38MAPK/TNF/NF-κB pathway, and specific target gene members were modulated by miR-200b and miR-204. In contrast, dysregulation of the IL-1B/Akt/ERK/NF-κB/Vascular Endothelial Growth Factor (VEGF) and TGFβ1/BMP2/MMP2/COX2/ZEB1 pathways associated with endometriotic epithelial cells may induce inflammation and potential activation of EMT, respectively (Figure 5A).

Figure 5.

A, Transforming growth factor β1 (TGFβ1)/ZEB1/miR-200b feedback loop for endometriotic epithelial cells in EMT. B, miR-200b/ZEB1 double-negative feedback loop for endometriotic stromal cells in MET. C, miR-1827/miR-504 model of target genes and pathways in endometriotic epithelial cells. BMP2 indicates Bone Morphogenetic Protein 2; COX2 (PTGS2), Prostaglandin-endoperoxide Synthase 2; DNTTIP1, deoxynucleotidyltransferase terminal interacting protein 1; ELMSAN1, (MIDEAS) ELM2 and Myb/SANT domain containing 1; EMT, epithelial–mesenchymal transition; FGF2, Fibroblast Growth Factor 2; FOXO, Forkhead Box; HDAC1, histone deacetylase 1; IL-1B, Interleukin 1B; KLF2/4, Kruppel Like Factor 2/4; MET, mesenchymal–epithelial transition; MMP2, Matrix Metallopeptidase 2; NQO1, NAD(P)H Quinone dehydrogenase 1; NRP2, neuropilin 2; OCT4 (POU5F1), POU class 5 homeobox 1; PLXNA4, plexin A4; SEMA, semaphorin; SOX2, SRY-box 2; MMP2, Matrix Metallopeptidase 2; TGFβ1, transforming growth factor beta 1; VEGF, Vascular Endothelial Growth Factor; ZEB1, Zinc finger E-box Binding protein 2.

Impaired decidualization and aberrant immune cell trafficking of stromal cells in endometriosis often result in infertility or poor embryo implantation.^68–70 Diseased stromal cells potentially modulate immune cell trafficking mainly by downregulation of the cytokines IL-15, CXCL12 and the mediator of IL-1, MYD88, that decrease movement of mast and antigen-presenting dendritic and macrophage cells (Table 3).^71–73 Decreased immune cell trafficking has 2 consequences, first, reduced inflammatory response to infectious pathogens, such as bacteria,⁷⁴ and second, the reduced influx of dendritic, uterine natural killer (uNK), mast, and macrophage cells leads to several reproductive disorders, such as embryo implantation failure, recurrent miscarriages, and preeclampsia.^69,75 Interleukin 15 both responds to infectious pathogens and stimulates proliferation and differentiation of uNK cells in the endometrium during decidualization.^76,77 Upon decidualization, dendritic and uNK cells crucially remodel endometrial spiral arteries ready for trophoblast invasion that will convert the arteries into wide-diameter, high-volume blood vessels for placental development.^78,79 Decidual cells act as a physical barrier to prevent the excessive invasion of the columns of extravillous trophoblast cells.^59,80,81 However, if decidualization is impaired, then trophoblasts will invade too deeply, and the life-threatening disorders of choriocarcinoma or placenta accreta may occur.^58,60,82

Key decidualization genes were found to be aberrantly expressed and may contribute to impaired decidualization in endometriotic stromal cells, such as upregulation of POSTN, SERPINA1, MMP7, and miR-200b and downregulation of HAND2, IL-15, ZEB1, and WNT4 (Figure 2C).^66,83,84 Although the stromal cells analyzed came from the proliferative phase, it has been demonstrated by other studies that aberrant expression of key decidualization genes in the proliferative phase, such as POSTN, FK506 Binding Protein 4 (FKBP4), progesterone receptor chaperone, Corticotropin Releasing Hormone (CRH), and Urocortin (UCN), have persisted into the secretory phase and adversely affect decidualization by delaying the phase transition due to progesterone resistance of endometriosis.^24,84–87 During decidualization, the miR-200 family is downregulated; however, miR-200b is upregulated in endometriotic stromal cells during the proliferative phase (Figure 3C), potentially affecting decidualization by downregulating CREB5, Sulfatase 2 (SULF2), and ZEB1, which are normally upregulated during decidualization.^84,88 CREB5 is a member of the important cyclic adenosine monophoshate/protein kinase A induction pathway of decidualization.⁸⁹ Impaired decidualization presages that the embryo will have difficulty implanting into the endometrium and, if successful, may result in poor placental development or miscarriage.^90,91 A poorly developed placenta can lead to preeclampsia, preterm birth, or intrauterine growth restriction.^82,92,93

Interestingly, in endometriotic stromal cells, the downregulation of ZEB1 through aberrant upregulation of miR200b (Figure 4B)^94–97 potentially activates a mesenchymal–epithelial transition (MET) pathway by a double-negative feedback loop.^98–100 Double-negative feedback loops consist of 2 molecular interactions that reciprocally repress each other, depending on their relative levels, and act as a reversible molecular switch that can maintain bistable cellular states, for example, in this case, either stromal or epithelial cells, by switching between either EMT or MET depending on the prevailing molecular factors. Activation of MET has been demonstrated by Wang et al to mediate early-stage induction of fibroblast dedifferentiation into induced pluripotent stem cells (Figure 5B).¹⁰¹ It is not clear what properties of MET-transformed stromal cells contribute to endometriosis other than, perhaps, being prone to form lesions ectopically.¹⁰²

Progesterone-resistant dysregulation of decidualization has been shown to play a significant role in the pathogenesis of endometriosis.^50,103–105 Increased POSTN upregulates Akt (protein kinase B), which downregulates the progesterone receptors (PRA/B), so that increased Akt activation in endometriosis inhibits decidualization.^106,107 Decreased HAND2 in endometriotic stromal cells, demonstrated here, may impact proper metabolism of estradiol and instead downregulate critical genes for decidualization of IL-15, TIMP Metallopeptidase Inhibitor 3 (TIMP3), and FOXO1. ¹⁰⁸ Progesterone resistance is also due to aberrant expression of steroid biosynthesis enzymes in endometriotic stromal and epithelial cells.^63,105 The aldo-keto reductases AKR1C1/2/3 are progesterone and steroid metabolizers that are upregulated in endometriosis.¹⁰⁹ Increased AKR1C1 (20α-hydroxysteroid dehydrogenase) inactivates progesterone, and in stromal cells, miR-200b overexpression upregulated AKR1C1 (Figure 4B). AKR1C2 inactivates potent 5-alpha-Dihydrotestosterone (5α-DHT) and is upregulated in endometriotic epithelial cells (Table 1).¹⁰⁹ Other metabolizing enzymes, such as 17β-hydroxysteroid dehydrogenase 2 (HSD17B2), metabolize estradiol to weakly estrogenic estrone and HSD11B2, which inactivates cortisol. HSD17B2 and HSD11B2 were both downregulated in endometriotic epithelial cells (Table 1).⁶³ HSD17B2 in epithelial cells is regulated by stromal paracrine factors, especially retinoic acid, which in normal stromal cells is upregulated by progesterone.^65,110 However, in endometriotic stromal cells, progesterone resistance disrupts the paracrine retinoic acid signaling and downregulates HSD17B2 in epithelial cells so that estradiol is not properly metabolized and abnormally accumulates.^63,65,106

Progesterone resistance from stromal cells also disrupts the 2-way paracrine interaction between stromal and epithelial cells.^64,111 Decreased progesterone signaling in the stromal cells downregulates Indian Hedgehog (IHH) in epithelial cells, which is downregulated in endometriosis.^64,112,113 Not only is IHH downregulated in diseased epithelial cells with decreased signal to stromal cells, but also the Hedgehog receptor, PTCH1, in stromal cells, is downregulated by miR-200b and miR-204 (Figure 4B). Hedgehog signaling in stromal cells also involves CREB5 and FOXF1, the target genes of miR-200b and miR-204, and NRP2, the target gene of miR-504 in epithelial cells.^114,115 The Hedgehog pathway and miR-204 in endometriotic stromal cells separately downregulate WNT4, essential for decidualization.⁶⁶ So the dysregulated Hedgehog signaling pathway from endometriotic epithelial cells impairs decidualization in stromal cells.⁶²

Furthermore, in diseased epithelial cells, inflammation-related genes are upregulated, such as TGFβ1, IL-1B, BMP2, PTGS2 (COX2), SERPINA1, and CSF3. ¹¹⁶ (SERPINA1 and CSF3 are in common with diseased stromal cells as well as downregulated LTBP4, which releases increased TGFβ1.) Increased TGFβ1 expression activates the ZEB1 molecular pathway to EMT and may increase pluripotency and stemness (Figure 5A) as well as increase the homing of mononuclear leukocytes (Table 3).^94,117–121 An increased expression of TGFβ1, MMP2, COX2, and BMP2 upregulates ZEB1.^122–124 ZEB1 overexpression promotes metastasis in cancer and marks aggressive endometrial carcinoma.^{117,125–127} ZEB1 activates EMT by repressing E-cadherin (CDH1), which dissociates cell adhesion and disrupts epithelial cell polarity.⁹⁶ ZEB1 represses stemness-inhibiting microRNAs, especially miR-200b, to induce stem cell-like properties in epithelial cells, notably in cancer, including endometrial carcinoma (Figure 5A).^{95,117,128,129} Although the downregulation of miR-200b was not discernible in this study of endometriotic epithelial cells, other studies have found that miR-200b is downregulated in endometriotic endometrium.^18,21,130 Perhaps the role of miR-200b in the ZEB1/miR-200b feedback loop in endometriotic epithelial cells is for attachment at an ectopic site, so that upregulated miR-200b inhibits ZEB1, reverses EMT, and induces MET.¹²⁹ (Similarly, downregulated ZEB1 in diseased stromal cells could activate an MET pathway in eutopic endometrium.) The EMT in endometriotic epithelial cells might also be activated by Hedgehog signaling indirectly via Fibroblast Growth Factor 2 (FGF2), NOTCH, and TGFβ1 signaling pathways or by TGFβ1, ZEB1, and miR-504 upregulating NRP2 (Figures 3H and 5C).^114,131 In turn, EMT strongly induces AXL receptor tyrosine kinase (AXL), which was upregulated in endometriotic epithelial cells in this study (Table 1). AXL increases cell migration and metastasis and is a stem cell regulator.^132–136

When potentially invasive, endometriotic, stem cell–like EMT cells are refluxed during menstruation and attach in ectopic sites, these EMT cells are ready primed by IL-1B for neovascularization with altered expression of Nitric Oxide Synthase 3 (NOS), FGF2, MMP2, TGFβ1, BMP2, and FLT1 to enable invasion.^118,128 The target genes of miR-504 and miR-1827, NRP2 and NQO1, increase vascularization, leukocyte, and cell migration in endometriotic epithelial cells (Table 3; Figure 5C). NRP2 and PLXNA4 are neovascularization pathfinders for ectopic sites.¹³⁷ However, ectopic lesions, compared to eutopic endometrium, can have reversed molecular expression of critical genes. Other studies have shown that when the refluxed EMT cells are implanted at an ectopic site, ZEB1 is downregulated, E-cadherin and miR-200b are upregulated, the cells transform back into epithelial cells by MET, and these cells establish, vascularize, and grow into lesions.^102,138

In conclusion, aberrant molecular pathways associated with diseased stromal cells decrease immune cell trafficking and impair decidualization, while diseased epithelial cells are associated with TGFβ1 and IL-1B inflammation-driven pathways that increase vascularization, cell proliferation, leukocyte and cell migration, and pathological EMT. The target genes of the differentially expressed microRNAs, such as miR-200b, miR-204, miR-504, and miR-1827, play a prominent role in the main aberrant functions associated with endometriotic stromal and epithelial cells. Understanding these cell type–specific microRNA—mRNA molecular pathways and interactions will contribute to the targeted treatment of this complex and devastating disease.