MicroRNA-Regulated Pathways Associated with Endometriosis

Abstract

Endometriosis is a prevalent gynecological disease characterized by growth of endometriotic tissue outside the uterine cavity. MicroRNAs (miRNAs) are naturally occurring posttranscriptional regulatory molecules that potentially play a role in endometriotic lesion development. We assessed miRNA expression by microarray analysis in paired ectopic and eutopic endometrial tissues and identified 14 up-regulated (miR-145, miR-143, miR-99a, miR-99b, miR-126, miR-100, miR-125b, miR-150, miR-125a, miR-223, miR-194, miR-365, miR-29c and miR-1) and eight down-regulated (miR-200a, miR-141, miR-200b, miR-142-3p, miR-424, miR-34c, miR-20a and miR-196b) miRNAs. The differential expression of six miRNAs was confirmed by quantitative RT-PCR. An in silico analysis identified 3851 mRNA transcripts as putative targets of the 22 miRNAs. Of these predicted targets, 673 were also differentially expressed in ectopic vs. eutopic endometrial tissue, as determined by microarray. Functional analysis suggested that the 673 miRNA targets constitute molecular pathways previously associated with endometriosis, including c-Jun, CREB-binding protein, protein kinase B (Akt), and cyclin D1 (CCND1) signaling. These pathways appeared to be regulated both transcriptionally as well as by miRNAs at posttranscriptional level. These data are a rich and novel resource for endometriosis and miRNA research and suggest that the 22 miRNAs and their cognate mRNA target sequences constitute pathways that promote endometriosis. Accordingly, miRNAs are potential therapeutic targets for treating this disease.

This study uses microarray technology, qRT-PCR and in silico analyses, to identify miRNA-regulated molecular pathways that are likely to contribute to the pathogenesis of endometriosis.

Endometriosis is a debilitating disease, causing chronic pelvic pain, painful periods, and infertility, that affects 10–15% of reproductive aged women. The disease is attributed to retrograde passage of endometrial fragments in menstrual fluid through the fallopian tubes, which then implant at ectopic sites, usually in the pelvis (1). There is evidence that the hormonal milieu, a genetic predisposition, an aberrant immune response, and environmental factors also contribute to endometriosis susceptibility and progression. Current medications ameliorate symptoms but do not cure endometriosis and cause significant side effects such as osteoporosis (reviewed in Ref. 2). Additionally, many women require extensive and repeated surgery for this condition. There is a need for novel therapeutic agents that suppress endometriotic lesion establishment and growth to reduce the high burden conferred by this disease (3).

Microarray analyses have identified a large number of mRNA transcripts differentially expressed in the eutopic endometrium of women with and without endometriosis (4, 5) and between ectopic and eutopic endometrial samples from the same women (6, 7, 8, 9). A species-specific microarray analysis of xenografts from the nude mouse model of endometriosis suggested the importance of the cellular and molecular dialogue between endometrial and peritoneal tissues in endometriotic lesion development. This cross talk appeared to be mediated by four key molecular networks involving the proteosome, nuclear factor κB (NFκB), TGFβ, and Ras. A comparison with differentially expressed transcripts in ectopic vs. eutopic endometrium in women with endometriosis indicated that these molecular networks may also underpin human disease (10).

However, an in silico study suggested that these data are incomplete, because many transcripts known to be associated with endometriosis are not identified in the published microarray studies of endometriosis (11). One explanation might be that alterations in mRNA transcript abundance reflect only a subset of the pathogenic mechanisms, and that posttranscriptional regulatory molecules such as microRNAs (miRNAs) also play an important role.

miRNAs are highly conserved, 20- to 24-nucleotide single-stranded RNAs that bind to single or multiple 7- to 8-mer motifs situated predominantly, but not exclusively, in the 3′-untranslated regions of target mRNAs. In mammalian cells, this generally leads to mRNA destabilization and/or translational repression, depending on the degree of sequence homology between the miRNA and target RNA transcript. Two recent proteomic studies showed that a single miRNA might curtail the synthesis of proteins encoded by hundreds of genes (12, 13). There may be up to 1000 miRNAs (14) targeting approximately 30% of the human genome and thereby regulating transcripts from as many as 8000 genes (15).

miRNAs are potent negative feedback regulators in a broad array of cellular processes that occur in endometriosis, such as cell differentiation and proliferation (16, 17, 18, 19), cell migration (20, 21, 22, 23), and myogenesis (16, 19, 24). Aberrant miRNA expression has been associated with complex disorders such as hematopoietic disease (18), heart failure (25), and several cancers including estrogen-regulated ovarian (26) and breast cancer (27). Therapeutic antagonism of miRNAs has been successful in nonhuman primates (28) and may therefore become a novel means of treating miRNA-associated human disease.

We hypothesized that miRNAs mediate posttranscriptional regulation in endometriotic lesions and regulate transcripts in previously identified endometriosis-associated molecular networks (10). Therefore, miRNA expression in paired eutopic and ectopic endometrial biopsies from patients with endometriosis was compared. Using microarray, quantitative RT-PCR (qRT-PCR), and bioinformatic techniques, we have identified both novel and previously characterized miRNA-regulated molecular pathways that are likely to contribute to the pathogenesis of endometriosis. These findings constitute a comprehensive foundation for future research into the role of miRNAs in this disease.

Results

miRNAs are differentially expressed in ectopic endometrial tissue

To determine the miRNA profile in endometrial tissue, miRNA microarray analysis was performed on sets of paired samples of eutopic and peritoneal ectopic endometrial tissue from seven patients with endometriosis. Three distinct computational analyses, Independent Component Analysis (29) (ICA, supplemental Fig. 1A published as supplemental data on The Endocrine Society’s Journals Online web site at http:// mend.endojournals.org), ANOVA, and Linear models for microarray data (Limma) (30) all identified 22 miRNAs that were differentially expressed with a fold change ±≥1.5 at P ≤ 0.05 in ectopic vs. eutopic endometrium (supplemental Fig. 1B). Fourteen miRNAs were up-regulated, and eight were down-regulated (Table 1) in ectopic tissue.

Table 1.

Differentially expressed miRNAs in paired ectopic vs. eutopic endometrial tissue from patients with endometriosis

Human (hsa-) miRNA gene	Fold Change	Accession no.	Chr	Chromosomal location	miRNA family	miRNA cluster
miR-145	4.47	MIMAT0000437	5	Intergenic	mir-145	1
miR-143	2.84	MIMAT0000435	5	Intergenic	mir-143	1
miR-99a	2.46	MIMAT0000097	21	Overlapping transcripts	mir-99
miR-99b	2.43	MIMAT0000689	19	Intergenic	mir-99	2
miR-126	2.33	MIMAT0000445	9	Overlapping transcripts	mir-126
miR-100	2.25	MIMAT0000098	11	Intergenic	mir-99
miR-125b	1.96	MIMAT0000423	11,21	intergenic	mir-125
miR-150	1.95	MIMAT0000451	19	Overlapping transcripts	mir-150
miR-125a	1.79	MIMAT0000443	19	Intergenic	mir-125	2
miR-223	1.72	MIMAT0000280	X	Intergenic	mir-223
miR-194	1.67	MIMAT0000460	1,11	Overlapping transcripts	mir-194
miR-365	1.65	MIMAT0000710	16,17	Overlapping transcripts	mir-365
miR-29c	1.63	MIMAT0000681	1	Intergenic	mir-29
miR-1	1.53	MIMAT0000416	20,18	Overlapping transcripts	mir-1
miR-196b	−1.60	MIMAT0001080	7	Overlapping transcripts	mir-196
miR-20a	−1.80	MIMAT0000075	13	Overlapping transcripts	mir-17
miR-34c	−1.91	MIMAT0000686	11	Intergenic	mir-34
miR-424	−2.27	MIMAT0001341	X	Overlapping transcripts	mir-322
miR-142–3p	−2.27	MIMAT0000434	17	Intergenic	mir-142
miR-200b	−2.28	MIMAT0000318	1	Intergenic	mir-8	3
miR-141	−2.34	MIMAT0000432	12	intergenic	mir-8
miR-200a	−2.67	MIMAT0000682	1	Intergenic	mir-8	3

miRNA genes significantly (P ≤ 0.05) differentially regulated (fold change ± ≥1.5) determined by ANOVA from microarray analysis of ectopic compared with eutopic endometrial tissue from seven patients with endometriosis are listed in order of their fold change. The accession number is for the mature miRNA sequence; Chr, Chromosome.

An ANOVA analysis on the miRNA microarray data according to menstrual cycle phase was also performed using Limma. This failed to show any significant differences in miRNA profiles between tissues taken from women during proliferative (n = 4) and secretory (n = 3) phases (data not shown).

qRT-PCR analysis of endometrial miRNA expression

To validate the microarray methods, three up-regulated (miR-99a, miR-126, miR-145) and three down-regulated (miR-141, miR-200b, and miR-424) miRNAs were quantified using qRT-PCR analysis (Fig. 1). In concordance with the microarray, qRT-PCR showed significant differential expression of these miRNAs in ectopic vs. eutopic endometrium (P ≤ 0.05). miR-145 and miR-141 were the most differentially expressed of the six miRNAs (P ≤ 0.001), up- and down-regulated, respectively. The qRT-PCR ΔCt values were normalized to miR-let-7a and miR-let-7d, which were both highly abundant and invariably expressed (fold change = 1.09 and 1.05, respectively; P = 0.5) according to the microarrays.

Fig. 1.

qRT-PCR analysis of miRNA expression in endometriosis. miRNA expression in paired ectopic and eutopic endometrial samples from women in endometriosis, as determined by microarray (n = 7, black) and qRT-PCR (n = 8). In qRT-PCR analysis, miRNA expression was normalized to hsa-miR-let-7a (dark gray) or hsa-miR-let-7d (light gray). Significant fold changes in ectopic vs. eutopic endometrium are marked by *, P ≤ 0.05; **, P ≤ 0.01; or ***, P ≤ 0.001. Data are mean ± sem.

miRNA gene families and clusters

The mean microarray fold changes, chromosomal locations, family memberships and cluster affiliations for the 22 differentially expressed miRNAs are presented in Table 1. These include both inter- and intragenic miRNAs. Two up-regulated (miR-125 and miR-99) and one down-regulated (miR-8) miRNA gene families, and three clusters (defined as groups of miRNAs located within 10 kb of each other) were identified among the differentially regulated miRNAs.

Identification of miRNA targets in endometriosis

The predicted target mRNAs of the 22 differentially expressed miRNAs were identified in silico. A conservative approach was undertaken, limiting the targets to those predicted by two algorithms, TargetScanS (31) and four-way PicTar (32), hereafter referred to as the two-way intersection. Although no targets were listed for miR-365 and miR-424 in the four-way PicTar algorithm, 3851 target transcripts corresponding to 2340 unique official gene symbols were present in this intersection for the remaining 20 miRNAs (supplemental Table 1 published as supplemental data on The Endocrine Society’s Journals Online web site at http://mend.endojournals.org). The number of mRNA targets predicted for each of the differentially expressed miRNAs varied substantially (Fig. 2A).

Fig. 2.

Functional analysis of all predicted mRNA targets of the 22 differentially expressed miRNAs in endometriosis. A, The number of mRNA transcripts predicted to be targeted by the 22 differentially expressed miRNAs in ectopic vs. eutopic endometrium from women with endometriosis. Seven miRNAs may each repress more than 200 mRNA transcripts, whereas five miRNAs had less than 30 anticipated targets. B, One of the molecular networks from IPA of all predicted targets of the 22 miRNAs incorporated molecules involved in cardiovascular system development and function, cellular movement, and cell morphology, including the validated miRNA target MEF2C. C, TGFβ2, SPARC, COL1A1, and COL1A2 are confirmed miRNA targets involved in extracellular matrix remodeling and cell invasion in an IPA network converging on interferon-γ. The gene names in these networks may be identified on the HUGO Gene Nomenclature Committee home page (http://www.genenames.org).

Functional analysis of miRNA targets in endometriosis

The functions of the 3851 predicted miRNA target transcripts and the molecular pathways they potentially constitute were assessed using Ingenuity Pathway Analysis (IPA) software. The predicted targets were significantly enriched for several biological functions known to be involved in endometriosis, including: cellular movement, assembly and organization; cell-to-cell signaling and interaction; nervous system development and function; embryogenesis; cell cycle; cell death; connective tissue development and function and endocrine system development and function (supplemental Table 2).

IPA identified 39 molecular networks constituted by the 3851 predicted targets, exemplified by two networks with well-characterized miRNA targets in Fig. 2.

The network in Fig. 2B includes myocyte enhancer factor 2C (MEF2C), a confirmed target (18) of one of the up-regulated miRNAs in ectopic endometrium in this study (miR-223). IPA suggested that 22 of the 3851 predicted targets of the differentially expressed miRNAs had molecular associations with MEF2C in the published literature (supplemental Fig. 2A). It is possible that groups of molecular associations such as these occur by chance. Therefore, as proof of principle, a permutation study was performed. This showed that only 4.4% of 1000 randomly selected lists of 22 miRNAs had more predicted targets associated with MEF2C in the IPA database than the 22 differentially expressed miRNAs (supplemental Fig. 2B). A corresponding permutation analysis focused on c-Jun, another target of interest in our study, showed similar results (supplemental Fig. 2, C and D).

Another molecular network constituted by the 3851 predicted targets converged on interferon-γ and included TGFβ2, secreted protein, acidic, cysteine-rich, osteonectin (SPARC), and collagens COL1A1 and COL1A2, which are all confirmed targets of the ectopically dysregulated miR-141 (33), and miR-29c (23), respectively. Mechanisms mediated by these molecules have been implicated in endometrial lesion development (10, 34), and the network in Fig. 2C suggests that these pathways may also be regulated by miRNAs.

Gene Ontology (GO) analysis suggested that the predicted targets of the 22 differentially expressed miRNAs were relatively enriched for GO paths associated with transcription. However, targets of five randomly chosen sets of 22 miRNAs were equally enriched for this category of GO paths (data not shown).

mRNAs differentially expressed in endometriosis are predicted targets of the 22 endometrial miRNAs

After identification of the 22 differentially expressed miRNAs and their predicted targets, we further restricted our analysis to a subset of putative miRNA targets that were also differentially expressed at the mRNA level (summarized in supplemental Fig. 3). In an mRNA microarray study of paired eutopic and peritoneal ectopic endometrium from women with endometriosis (10), hereafter referred to as the peritoneal comparison, 4034 transcripts were differentially expressed (fold change ±≥1.5, P ≤ 0.05). Of these transcripts, 673 (572 unique official gene symbols) were both differentially expressed and predicted to be targeted by at least one, but often several, of the 22 differentially expressed miRNAs (supplemental Table 3). IPA suggested that the 673 transcripts were significantly enriched for functional annotations associated with endometriosis, including: cell death; connective tissue, nervous and muscular system development and function; cellular movement; cell proliferation and the cell cycle; angiogenesis and reproductive/endocrine system disorders (supplemental Table 4).

These 673 mRNA transcripts were overrepresented in 12 IPA networks converging on 1) JUN; 2) platelet-derived growth factor β polypeptide)/chemokine (C-X-C motif) ligand 12 (PDGF BB/CXCL12); 3) CREBBP; 4) CCND1; 5) calmodulin; 6) NFkB; 7) MAPK; 8) caveloin 1/vinculin (CAV1/VCL); 9) c-AMP dependent protein kinase (PKA); 10) phosphoinositide-3-kinase (PI3K); 11) protein kinase C (PKC); and 12) AKT (Fig. 3, A–C; supplemental Fig. 4, A–I; and supplemental Table 4). Although each of these was originally identified as a distinct network, there was cross-interaction between the networks, exemplified by JUN in Fig. 3A, which is a transcriptional target of MEF2C in Fig. 3B. Overlap was also found between networks of all predicted targets of the 22 miRNAs and networks of the differentially expressed predicted targets of the 22 miRNAs, illustrated by TGFβ2, SPARC, COL1A1, and COL1A2 in Figs. 2C and 3A.

Fig. 3.

Functional analysis of differentially expressed miRNA targets in endometriosis. A–C, Networks identified by IPA of differentially expressed mRNAs that are also predicted targets of the 22 differentially expressed miRNAs in paired peritoneal ectopic vs. eutopic endometrial tissues from women with endometriosis. A, c-Jun is central in a network comprising extracellular matrix proteins that are confirmed targets of miRNAs differentially expressed in endometriosis. This network overlapped considerably to analysis of all predicted targets of the 22 miRNAs (Fig. 2C), suggesting a role for miRNAs in regulating cell migration during endometrial lesion development. B, MEF2C, histone deacetylase 4 (HDAC4), and DNA (cytosine-5-) methyltransferase 3α (DNMT3A) are validated miRNA targets represented in a network converging on CREBBP, proposing a role for miRNAs in the regulation of myogenesis, angiogenesis, and epigenetic modulation of gene expression in endometrial tissue remodeling. C, The IPA network converging on AKT indicated a role for miRNAs in regulating retinoic acid signaling, the cell cycle, and cellular growth and proliferation. D, IPA network from functional analysis of a second microarray data set of differentially expressed miRNA targets in paired ovarian ectopic vs. eutopic endometrial tissue from women with endometriosis. The network exemplifies the substantial overlap between the ovarian and the peritoneal comparisons, illustrated in orange and blue nodes. Red and green nodes depict up- and down-regulated miRNA targeted transcripts in ectopic vs. eutopic endometrial tissue, respectively. Lines represent the biological relationship between two nodes. The gene names in these networks may be identified on the HUGO Gene Nomenclature Committee home page (http://www.genenames.org).

To verify that the endometriosis-specific functions and networks identified here were not limited to the specific endometrial lesion types and patients examined, we also defined the intersection of miRNA data with a second mRNA microarray study of paired eutopic and ovarian ectopic endometriotic tissues (9), hereafter referred to as the ovarian comparison. This revealed 684 transcripts that were both differentially expressed (fold change ±≥1.5, P ≤ 0.05) and putatively targeted by one or more of the 22 differentially expressed miRNAs (supplemental Fig. 5 and supplemental Table 5).

The IPA networks identified for these 684 ovarian endometrial transcripts showed a remarkable overlap with those associated with the peritoneal comparison as exemplified in Fig. 3D, including those converging on AP1 (c-JUN/c-FOS), PI3K/AKT, CCND1, NFκB, and PDGF/BB (data not shown).

c-Jun mRNA expression

c-Jun is centrally located in one of the IPA networks (Fig. 3A) and is up-regulated at the mRNA level in the ectopic peritoneal endometrial tissue study (10). We used qRT-PCR to confirm the significant up-regulation of c-Jun mRNA (P = 0.015) in six of the same RNA samples used in the miRNA microarray and qRT-PCR analyses. Median normalized c-Jun expression in ectopic and eutopic tissue was 2.03 and 0.11, respectively (Fig. 4).

Fig. 4.

c-Jun mRNA expression was significantly increased in endometriosis. c-Jun expression was significantly increased (P = 0.015) in ectopic (Ec) vs. eutopic (Eu) endometrium from six women with endometriosis, as determined by qRT-PCR. Data are median and range of normalized c-Jun mRNA expression from three independent experiments for each patient.

Discussion

This study used microarray technology, qRT-PCR, and in silico analyses in a novel strategy to assess miRNA expression in peritoneal endometrial tissues from women with endometriosis. Twenty-two miRNAs that were differentially expressed in paired ectopic vs. eutopic endometrium were identified that putatively regulate the expression of 2340 genes. A subset of these target transcripts was previously found to be differentially expressed between endometrial ectopic and eutopic endometrium in two separate mRNA microarray studies (9, 10). The specific expression profiles of the 22 miRNAs and their predicted targets may cause alterations to molecular pathways that are pivotal in the endometriotic process. Therefore, the inhibition or reactivation of these pathways by synthetic antisense antagomiRs or over expression of these miRNAs are potential therapeutic interventions in endometriosis.

Because ectopic endometrial tissue is less susceptible to steroid hormone-driven fluctuations than eutopic endometrium from women with endometriosis (reviewed in Ref. 35), miRNA analysis in different phases of the cycle would be expected to mainly reflect alterations in eutopic tissue. Therefore, miRNA expression was studied in endometrial tissues during both secretory (n = 3) and proliferative (n = 4) phases. The 22 differentially expressed miRNAs are hence consistently associated with endometriosis across the menstrual cycle. No effect on menstrual cycle phase on endometrial miRNA expression could be demonstrated in the present study. Cycle-related effects may be revealed by analysis of a larger study group in the future.

The dataset of 22 differentially regulated miRNAs in endometriosis was defined conservatively by the concordance of three methods of analysis. The combination of two parametric statistical methods (linear models that use the pooled correlation of replicate spots on the arrays, as well as a nested linear model design) with an independent nonparametric method (ICA) as well as qRT-PCR confirmation of three up- and three down-regulated miRNAs lends weight to the validity of this dataset.

Although all the 22 miRNAs have overlapping target lists, seven had the potential to individually target up to 494 mRNAs. Each of the 22 miRNAs was predicted to suppress, on average, 175 target mRNAs, with some mRNAs predicted to be coregulated by multiple miRNAs. This indicates that miRNAs, like transcription factors, may be considered master regulators in cellular processes in endometriosis.

Proximally paired miRNA genes located up to 50 kb apart are frequently coexpressed (36). The finding of three miRNA clusters among the differentially expressed miRNAs provides further confidence in the results of this analysis. miRNAs in the same cluster have common predicted mRNA targets and might consequently have an additive repressive effect on target expression. The clustered and multitargeting miRNAs may be especially powerful regulators in ectopic endometrial lesions.

Approximately 80% of miRNA genes are located within introns (37), and these often have expression profiles that correlate with their host genes (36). We identified significant up-regulation in ectopic endometrial tissue of both miR-126 and epidermal growth factor (EGF)-like-domain, multiple 7 (EGFL7), a gene expressed in endothelial cells promoting vascular tubulogenesis. Because miR-126 is embedded in the EGFL7 gene, it is likely that these two genes are coexpressed in endometriosis, as previously described for miR-10a and homeobox B4 (38).

Correct identification of miRNA targets is central to determining the biological function of miRNAs. In this study, miRNA targets were identified at a conservative two-way intersection of TargetScanS and PicTar (four-way), as recommended in a recent consensus review (39). These are sensitive algorithms with substantial overlap in their predicted targets (40). miRNA targets with binding sites that were not conserved across species are excluded in this intersection, and we acknowledge that the impact of these miRNA targets in endometriosis was not explored here. However, in vitro studies suggested that this two-way intersection had a true positive rate of up to 90% (41) and that these two prediction models showed superior performance in a high-throughput examination of miRNA repression at protein level (12).

Frequent revision of miRNA target algorithms as well as target mRNA ENSEMBL annotations can dramatically alter predicted target transcript lists over time; therefore experimental confirmation of miRNA targets remains the gold standard. Despite the lack of large-scale screening methods for target verification, the number of validated miRNA targets is escalating, and there is growing evidence that miRNA targets can be confidently predicted in silico.

A large number of the mRNA targets of the 22 endometrial miRNAs discussed here are already experimentally confirmed in cellular systems similar to those involved in endometriosis. In addition, many of the molecular pathways these targets constitute are known to play important roles in the pathological processes involved in this condition. Therefore, we are confident that the molecular networks identified here represent conservative and well-founded hypotheses on which to base further experiments. Future studies are also necessary to determine the specific miRNA profiles of stromal and glandular epithelial cells as well as populations of infiltrating leukocytes, because the ratio of these cell types are altered in ectopic compared with eutopic endometrial tissue.

To appraise the general principle of identifying molecular networks using a systems biology database such as IPA, the associations with MEF2C and JUN of predicted targets of random miRNA datasets were assessed. These individual permutation studies suggested that, at least for these two genes, the results of the functional analysis were unlikely to occur by chance.

Functional analysis of miRNA targets was limited to the convergence of miRNA and mRNA microarray data. This method focuses the analysis not only on the potential actions of miRNAs in translational repression, but also on transcripts with a demonstrated regulation of abundance in endometriotic disease. We acknowledge that this approach may omit important molecules that are posttranscriptionally regulated by the 22 miRNAs, but that are not differentially expressed in endometriosis. Nevertheless we prefer to use this conservative approach to reduce false discovery. Predicted miRNA targets in two microarray studies (9, 10) identified pathways previously found to be critical in endometriosis (5, 6, 10). These two different mRNA data sets examined different sites of endometriosis and revealed distinct, but overlapping, transcript lists. The substantial correlation observed between these two independent comparisons lends confidence to the methodology used in evaluating the function of differentially expressed miRNAs in endometriosis.

Two distinct patient groups provided tissue for the mRNA and miRNA microarray analyses; therefore, it is possible that these individuals had distinct mRNA profiles. qRT-PCR of c-Jun confirmed that, at least for this transcript, the two independent tissue sets showed similar extents of dysregulation.

This is the first study demonstrating increased c-Jun mRNA expression by RT-PCR in ectopic vs. eutopic endometrial tissue, supporting our previous findings in a mRNA microarray study (10). Others identified reduced c-Jun mRNA levels in eutopic endometriotic compared with normal endometrial tissue (42). Endometrial c-Jun expression is induced by estrogen (43, 44, 45), a hormone required for endometriotic lesion development, whereas Danazol, an ethisterone derivative used to treat endometriosis, down-regulates c-Jun transcript levels in the lesions (46). Altogether this suggests a functional role for c-Jun at ectopic sites during endometriosis.

Functional analysis identified several molecular networks containing many transcripts linked to endometriosis. Whereas the cellular functions of these networks are consistent with the current view on endometriotic lesion pathogenesis, a comprehensive discussion of these pathways is beyond the scope of this paper. Instead, examples of the most prominent molecular networks are presented in Table 2 along with confirmed regulatory functions of the differentially expressed miRNAs, including regulation of cell migration and proliferation, wound contraction, extracellular matrix remodeling, and angiogenesis.

Table 2.

Experimentally confirmed regulatory functions of the 22 differentially expressed miRNAs in endometriosis

IPA Network (figure)	miRNA			miRNA target			Regulatory feedback loops	Confirmed biological functions	Proposed role in endometriosis	Ref. nos.
		Gene	FC	mRNA	Reg
CREBBP (Fig. 2B)	miR-223	↑	MEF2C	↑	MEF2C→ miR-1→| HDAC4→| MEF2C	miR-1 controls myogenesis and vascular development in association with MyoD family members	Cell migration, wound contraction, neovascularization in endometrial stromal myofibroblasts and pericytes	10 16 18 19 24 25
	miR-1	↑	HDAC4	↑				10 16
	miR-29c	↑	DNMT3A	↓		DNA methylation, tumorigenicity in lung cancer	DNA methylation, epigenetic regulation, gene expression	10 54
JUN (Fig 2A)	miR-29c	↑	SPARC, COL1A1, COL1A2	↑		Cell migration cancer metastasis	Extracellular matrix remodeling, cell invasion	10 23
	miR-141	↓	TGFβ2	↓		TGFβ-induced EMT	TGFβ-regulated stromal proliferation	10 33 55 56 57
CCND1 (S Fig. 3B)	miR-141, miR-200b	↓	ZEB1		ZEB1→| miR-141 and miR-200b→| ZEB1	TGFβ-induced EMT in cancer cell lines via targeting of TGFβ2, ZEB1 and ZEB2	Stromal fibroblast migration and invasion	20 21 22 33 58 59
	miR-141, miR-200b	↓	ZEB2	↑				10 20 21 22 59
	miR-20a	↓	CCND1	↑	miR-20a→| CCND1→ miR-20a		Epithelial cell proliferation	10 60
	miR-20a	↓	E2F3	↑	E2F3→ miR-20a→| E2F3		Cell proliferation and apoptosis	10 61
	miR-20a	↓	RUNX1	↑	miR-20a→| RUNX1→| miR-20a	Monocytopoiesis	Monocytic differentiation, proliferation, angiogenesis	10 62
PDGFBB/CXCL12 (S Fig. 3A)	miR-1	↑	FN1	↑		Cardiac hypertrophy, myocyte proliferation	Cell proliferation and migration, wound healing	10 63
NFκB (S Fig. 3D)	miR-20a	↓	CTGF	↑		Neovascularization in cancer	Neovascularization	10 64
	miR-1	↑	HSPD1	↓		Cardiomyocyte apoptosis	Apoptosis	10 65

S, Supplemental; C, fold change; Reg, miRNA regulation; ZEB1 and ZEB2, zinc finger E-box binding homeobox 1 and 2; E2F3, E2F transcription factor 3; RUNX1, runt-related transcription factor 1; FN1, fibronectin 1; CTGF, connective tissue growth factor; HSPD1, heat shock 60-kDa protein 1; EMT, epithelial to mesenchymal transition; ↑, up-regulated; ↓, down-regulated; →, positive regulation; →|, inhibition.

To our knowledge only one other study has explored miRNAs in endometriosis. Pan et al. (47) identified 48 miRNAs that were differentially expressed in eutopic and ectopic endometrium from a small study group of women with and without endometriosis; however, a direct comparison between endometriotic ectopic and eutopic tissues was not undertaken. Nevertheless, despite differences in study design, methods of analysis, measures of stringency, and RT-PCR endogenous controls, eight miRNAs concurred with the 22 miRNAs described in our analysis. This concordance underlines the likely importance of these miRNAs in the pathogenesis of endometriosis.

In summary, this study suggests that both miRNAs and the molecular pathways they target are differentially expressed between ectopic endometrial lesions and eutopic endometrium in women with endometriosis. Many of the miRNA-regulated molecular networks and biological processes identified here corroborate known endometriosis associations in the literature. These findings propose an intricate control of gene expression in endometriosis both transcriptionally and by miRNAs at the posttranscriptional level. The microarray data and molecular pathways presented here constitute a comprehensive resource on which to base future investigations into the role of specific miRNAs in endometriosis. In particular, the endometriosis-associated miRNAs and the molecular pathways they target should be evaluated as new candidates for the diagnosis and treatment of endometriosis.

Materials and Methods

Tissue collection and ethics approval

All procedures for collection of tissue were approved by the Central and Northwestern Human Ethics Committee, South Australia, or the Children’s Youth and Women’s Human Research Ethics Committee, South Australia. Eutopic and ectopic endometrial tissue was obtained with written informed consent from eight women with AFS stage II–IV endometriosis during laparoscopic surgical procedures. Endometrial biopsies were collected using Pipelle suction curettes (Pipelle de Cornier, Laboratoire C.C.D, Paris, France). All women had regular menstrual cycles (28–30 d) and were not taking any medication. Four were in the follicular phase of the cycle (d 9-d 14) and four in the secretory phase (d 17-d 28). The cycle phase of all biopsies was histologically confirmed using Noyes criteria (48). Samples from seven women were used for the miRNA microarray analysis, and an additional sample was used in the miRNA qRT-PCR analysis. All biopsies were immediately placed in RNA Later (Invitrogen, Carlsbad, CA) and stored at −80 C.

RNA extraction

Total RNA was extracted from endometrial tissues homogenized in Trizol solution (Invitrogen) according to the manufacturer’s instructions. RNA concentration was assessed using a Nanodrop spectrophotometer, accepting a ratio of 2.0 for sample absorbance at 260/280.

Microarray hybridization

The microarrays consisted of 377 antisense miRNA oligonucleotide probes (mirVana miRNA probe set; catalog no. 1564V1, Applied Biosystems, Foster City, CA) printed in triplicate onto epoxide-coated microarray slides (Corning, Inc., Corning, NY) with a VersArray ChipWriter Pro system (Bio-Rad Laboratories, Inc., Hercules, CA) using tungsten pins (PointTech, Barreal de Heredia, Costa Rica). Included on the array were three negative control probes, which were small-interfering RNAs of bacterial origin. For detection on the array, 5 μg total RNA was labeled by the ligation of a fluorescently modified RNA dimmer as previously described by Thomson et al. (49). Competitive hybridizations were performed, in each case comparing RNA extracted from eutopic vs. ectopic tissues from the same woman. To reduce effects of dye bias, dye swaps were performed, with RNA from each tissue labeled in both combinations (50). RNA was hybridized to the arrays for 2 h at 42 C using LifterSlips (Erie Scientific, Northbrook, IL) in 25 μl of hybridization buffer (400 mm Na₂HPO₄, pH 7.0; 0.8% BSA; 5% sodium dodecyl sulfate; 12% formamide). Slides were placed in Corning hybridization chambers, incubated in the dark for 2 h, then washed twice in 2× standard sodium citrate (SSC)/0.025% sodium dodecyl sulfate and three times in 0.8× SSC at room temperature, and finally twice in 0.4× SSC at 4 C. Slides were scanned at 10 μm resolution with a Genepix 4000B Scanner (Molecular Devices, Sunnyvale, CA) using photomultiplier tube (PMT) settings that approximately balanced the channel intensity histograms.

Computational microarray analysis

Median spot pixel intensity values in scanned images were determined using the Spot version 3 plugin (Commonwealth Scientific and Industrial Research Organization, Campbell Australian Capital Territory, Australia) for the statistical environment R. After subtraction of median background intensities and Loess normalization, mean intensities were log2 transformed, and ratios (Cy5/Cy3) were obtained. To select differentially expressed miRNAs, these data were analyzed using three different statistical methods. 1) A pooled correlation of triplicate spots for each miRNA was calculated for each chip using linear models and empirical Bayesian moderation of standard errors (Limma R package, Walter and Eliza Hall Institute, Parkville Victoria, Australia). 2) Alternatively, the three measurements for each transcript on each chip were analyzed using a nested ANOVA approach without empirical Bayesian moderation. 3) Data for the three measurements for each transcript on each chip (generated in method 2 above) were reanalyzed using an Ensembl learning implementation of Independent Component Analysis (ICA), which summarizes the overall transcript abundance patterns in the data. The 1% of miRNAs that contributed most strongly to the component associated with eutopic vs. ectopic tissue differences were identified. miRNAs that differentially expressed between the paired eutopic and ectopic tissues (fold change ≥1.5, P ≤ 0.05) according to all three methods (1–3) above were identified by Venn diagram analysis. Heat map analysis and hierarchical clustering were performed using R.

qRT-PCR analysis of miRNA expression

Transcript levels of three up-regulated and three down-regulated miRNAs were measured by qRT-PCR in four proliferative and four secretory paired eutopic and ectopic endometrial total RNA samples (total n = 16).

Total RNA (5 μg) was reverse transcribed on a GeneAmp PCR System 9700 using the Taqman MicroRNA Reverse Transcription kit and primers from TaqMan MicroRNA Assays (Applied Biosystems). Water substituted for transcript and reverse transcriptase samples served as negative controls. Assay-specific cDNA and probes (TaqMan MicroRNA Assays) were combined with TaqMan 2× Universal PCR Master Mix, No AmpErase UNGb (Applied Biosystems), and real-time PCR amplification was performed in triplicates using an ABI Prism 7000 (Applied Biosystems).

ΔΔCt values were calculated, and the data were normalized to hsa-miR-let-7a, which has been used as a control previously (51, 52), and to hsa-miR-let-7d. Kolmogorov-Smirnov test followed by nonparametric Mann-Whitney U test (asymptomatic significance, correlated for ties) determined the significant up- or down-regulation of ΔCt values, respectively, in ectopic compared with eutopic endometrium. Mann-Whitney U test of Ct and ΔCt [Ct(let-7a)-Ct(let-7d) values] validated the use of hsa-miR-let7a and hsa-miR-let-7d as invariable and ubiquitous endogenous controls. The mean and sem of fold change from eutopic endometrium was determined and the data were presented by GraphPad Prism 5 (GraphPad Software, Inc., San Diego, CA).

miRNA target prediction

Target mRNAs of the 22 differentially expressed miRNAs were determined using the miRGen web tool (53) for the algorithms TargetScanS (release 3.1) (31) and PicTar (four-species conservation) (32). Targets predicted by the two-way intersection of these algorithms were further analyzed. The Ensembl gene identification nos. of putative targets retrieved from miRGen were referenced to Ensembl build 46. The convergence of these miRNA targets with mRNAs that were differentially expressed in a microarray study of ectopic and eutopic peritoneal endometrial tissue (10) was determined. According to quality control analysis in the DChip application (http://biosun1.harvard.edu/complab/dchip/), the mRNA microarray data from patient 7 contained relative outliers; therefore, this patient was excluded from further analysis. mRNA probe sets were considered to be differentially expressed at fold change ≥±1.5 at cyberT P ≤ 0.05. To control for potential confounding due to variation between the different patient groups contributing tissues for the miRNA and mRNA microarray analyses, a second mRNA data set comparing ovarian endometriosis with eutopic endometrium (9) was studied. mRNA targets of the 22 miRNAs that were differentially expressed in ectopic endometrium were identified in both data sets.

Permutation study of ingenuity pathway analysis of miRNA targets

Manual pathway analysis showed that 82 molecules are associated with transcription factor MEF2C in the IPA database (http://www.ingenuity.com; May 25, 2008). Of 82 molecules, 22 are also putative targets of the 22 differentially expressed miRNAs. A permutation analysis was performed to verify the high degree of association between MEF2C and other miRNA targets in IPA: 1000 random sets of 22 miRNAs were selected from the 281 miRNAs within the Ambion human microarray probe set that were human in origin, unique, and fully annotated, and their targets were identified using the two-way intersection method described above. The intersect of the targets of each of these 1000 randomly chosen lists of 22 miRNAs, and the 82 molecules related to MEF2C in the IPA database, were calculated and plotted as histograms using R. This analysis was repeated for JUN. IPA identified 632 molecules associated with JUN (November 11, 2008), of which 147 are also putative targets of the 22 differentially expressed miRNAs.

Functional analysis of predicted mRNA targets of the 22 miRNAs

IPA was performed to identify the molecular pathways and functional groupings based on published literature, to which predicted targets of the 22 differentially expressed miRNAs belong. The targets were uploaded into IPA and overlaid onto a global molecular network developed from information contained in the application. Networks of these genes were generated by IPA based on their connectivity, each ranked by a score. This score is based on the hypergeometric distribution, calculated with the right-tailed Fisher’s Exact Test, and corresponds to the negative log of this P value. Networks with a score more than or equal to 20 were selected for further analysis. Functional analysis in IPA identified the published biological functions that were most significantly associated with the genes in the network. Genes or gene products are represented as nodes, and the biological relationship between two nodes is represented as an edge (line). All lines are supported by at least one reference in literature, textbook, or from canonical information stored in the Ingenuity Pathways knowledge database.

The GOStat application (http://gostat.wehi.edu.au/cgibin/goStat.pl) was used to determine the biological processes annotated to the putative miRNAs targets in the GO database (http://www.geneontology.org). The GOA human database was used, only considering the biological processes GO hierarchy, and GO paths with length more than or equal to 3. As a cut-off for significance, Fisher’s Exact Test (P ≤ 0.01) with Benjamini and Hochberg False Discovery Rate assessment was used to estimate the proportion of GO terms that would be identified by chance alone.

qRT-PCR analysis of c-Jun expression

Standard SYBR Green qRT-PCR was carried out to detect c-Jun transcript levels in total cellular RNA from the eutopic and ectopic endometrial tissues. After treatment with ribonuclease-free deoxyribonuclease I (Ambion Turbo DNAse, Applied Biosystems; 60 min/37 C), 1 μg RNA of the individual samples was reverse transcribed using the Superscript III Reverse Transcriptase kit (Invitrogen). Real-time PCR was performed in a RotorGene 6000 (Corbett Life Science, Concorde, New South Wales, Australia; 95 C for 10 min, 95 C for 15 sec followed by 45 cycles at 60 C for 60 sec) on 2 μg cDNA with SYBR Green PCR Master Mix (Applied Biosystems) and 5 μg primers to c-Jun [sense (5′-GGAGAGGAAGCGCATGAGG-3′; and antisense 5′-CATGTTGGCCGTGGACG-3′]. These primers were specific for the published c-JUN cDNA sequence (GenBank accession no. NM_002228), designed using Primer Express version 2 software (Applied Biosystems) and optimized for use at a final concentration of 0.5 μm. The resulting 136-base amplicon spans nucleotides 1801–1943 and was analyzed by melt curve analysis to ensure lack of nonspecific products and primer dimmers; 18S rRNA was analyzed in the same run with primers as recommended by the manufacturer (Applied Biosystems).

Linear standard curve analysis for each primer pair at a maximum 1:4096 dilution of control pooled cDNA and primer amplification efficiency (E) greater than 0.96 was obtained for all PCR experiments. c-Jun data were normalized to 18S RNA and expressed in arbitrary mRNA units as median values of three independent PCR runs. The data were analyzed by two-tailed Mann-Whitney U test (95% confidence interval) in GraphPad software.