Differential Expression of MicroRNAs between Eutopic and Ectopic Endometrium in Ovarian Endometriosis

Endometriosis, defined as the presence of endometrial tissue outside the uterus, is a common gynecological disease with poorly understood pathogenesis. MicroRNAs are members of a class of small noncoding RNA molecules that have a critical role in posttranscriptional regulation of gene expression by repression of target mRNAs translation. We assessed differentially expressed microRNAs in ectopic endometrium compared with eutopic endometrium in 3 patients through microarray analysis. We identified 50 microRNAs differentially expressed and the differential expression of five microRNAs was validated by real-time RT-PCR in other 13 patients. We identified in silico their predicted targets, several of which match the genes that have been identified to be differentially expressed in ectopic versus eutopic endometrium in studies of gene expression. A functional analysis of the predicted targets indicates that several of these are involved in molecular pathways implicated in endometriosis, thus strengthening the hypothesis of the role of microRNAs in this pathology.


Introduction
Endometriosis, defined as the growth of endometrial tissue outside the uterine cavity, is a common gynecological disease often resulting in chronic pelvic pain and infertility. The pathogenesis of endometriosis is likely multifactorial and several hypotheses have been suggested to explain the presence of ectopic endometrial tissue and stroma, such as retrograde menstrual reflux [1], immune system defects [2][3][4][5][6][7][8][9][10], and ectopic presence of endometrial stem cells originating the disease [11]. In addition, there is a growing body of evidence indicating the involvement of genetic factors in the etiology of endometriosis, as it has been calculated that there is a 6-9-fold increased prevalence of this pathology among the 1st-degree relatives of women with endometriosis, compared to the general population [12][13][14][15][16][17][18]. Extensive investigations have been performed to characterize the differences between the eutopic and ectopic endometrium in order to better understand and define the molecular basis of the disease and, indeed, several studies have revealed a distinct pattern of gene expression in eutopic and ectopic endometrium [19][20][21][22][23][24]. The differences in gene expression reported in these works include genes encoding proteins involved in cell adhesion, extracellular matrix remodeling, migration, proliferation, immune system regulation, and inflammatory pathways, thus accounting for the multiple mechanisms hypothesized to be responsible for the establishment of ectopic endometrial implants, including the adhesion of endometrial cells to the pelvic peritoneum, invasion into the mesothelium, and survival and proliferation of the ectopic endometrial cells.
MicroRNAs (miRNAs), members of a class of small noncoding RNA molecules, have a critical role in posttranscriptional regulation of gene expression by repression of target mRNAs translation. Originally identified in Caenorhabditis elegans [25], miRNAs have been shown to operate in a wide range of species, including humans. Computational predictions indicate that up to 30% of human genes are potential targets of miRNAs and that miRNAs compose 1%-5% of animal genomes [26][27][28][29]. MiRNA expression is tissue-and cell-specific [30][31][32][33]. It has been demonstrated that miRNAs are important in developmental processes as well as for other cellular activities involving cell growth, differentiation, and apoptosis. Moreover, several genes encoding miRNAs have been located at chromosomal fragile sites or regions of cytogenetic abnormalities associated with cancer and other disorders. Interestingly, miRNAs altered expression has been associated with tumorigenesis, and several studies have described differential expression of miRNAs in neoplastic versus normal tissue [34][35][36][37][38].  Our study is aimed to investigate the differential expression of miRNAs in endometriosis by direct comparison between paired ectopic and eutopic endometrium samples. Once we identified the differentially expressed miRNAs, we validated 5 of them by an independent technique. Then, we identified in silico the predicted molecular targets of the differentially expressed miRNAs and we used a bioinformatics tool to investigate the molecular pathways in which these targets could be involved.

Materials and Methods
2.1. Tissue Collection. Subjects (n = 16) scheduled for surgery for chronic pelvic pain or infertility at the University of Piemonte Orientale-affiliated "Maggiore della Carità" Hospital were recruited to participate in this study. The study was approved by the "Maggiore della Carità" Hospital's Institutional Review Board and informed consent was obtained from all participants. None of the authors have any conflict Table 1: Differentially expressed miRNAs in ectopic versus eutopic endometrium. List of differentially expressed miRNAs whose expression value in ectopic endometrium was at least twofold higher or lower than in eutopic endometrium P < .01.  Targetscan: graphical representation of one of the network (P-value = 10E−37, focus molecules = 35) identified by IPA analysis of the predicted targets of the miRNAs whose differential expression in eutopic and ectopic endometrium was validated by real-time RT-PCR. Highlighted are the genes involved in endometriosis according to IPA knowledge base.
of interest with the study. Surgery was scheduled 6 to 12 days after the onset of menses. No patients were receiving hormone therapy at the time of the study or in the previous three months. The patients ranged in age from 24 to 48 years, with an average of 36 years. Endometriomas were removed at laparoscopy by excision of the entire cyst wall by stripping technique, preserving normal ovarian tissue. Hysteroscopy with directed biopsies, performed to obtain a sample of eutopic endometrium from the same patient, were carried out using a 4 mm Bettocchi Hysteroscope System with a 5 Fr operative channel (Karl Stortz GmbH & Co., Tuttlingen, Germany). Laparoscopy and hysteroscopy procedures were performed during the same surgical intervention. Freshly recovered tissues were rinsed in saline solution and divided in two parts. One half of the tissue was immediately snapfrozen and kept in liquid nitrogen for further processing, while the other was sent to the pathology laboratory. The endometriomas of 9 patients were classified as moderate, while 7 were classified as severe according to the ASRM guidelines [39].

RNA Isolation.
Total RNA was extracted from tissues with the miRNeasy kit (Qiagen, Valencia, CA, USA) according to the manufacturer's protocol and quantified by Quant-iT RNA Assay Kit with Qubit Fluorometer (Invitrogen, Carlsbad, CA, USA).

MicroRNA Microarray Assay and Analysis.
Microarray assay was performed using a service provider (LC Sciences). Ten μg of total RNA from eutopic and ectopic endometrium obtained from three patients were size fractionated using a YM-100 Microcon centrifugal filter (Millipore) and the small RNAs (<300 nt) isolated were 3 -extended with a poly(A) tail using poly(A) polymerase. An oligonucleotide tag was then ligated to the poly(A) tail for later fluorescent dye staining; two different tags were used for the two RNA samples in dualsample experiments. Hybridization was performed overnight on a μParaflo microfluidic chip using a microcirculation pump (Atactic Technologies) [40,41]. On the microfluidic chip, each detection probe consisted of a chemically modified nucleotide coding segment complementary to target 475 mature human miRNA probes (Sanger miRBase sequence database 9.1) or other RNAs for control and a spacer segment of polyethylene glycol to extend the coding segment away from the substrate. The detection probes were made by in situ synthesis using PGR (photogenerated reagent) chemistry. The hybridization melting temperatures were balanced by chemical modifications of the detection probes. Hybridization used 100 μL 6xSSPE buffer (0.90 M NaCl, 60 mM Na 2 HPO 4 , 6 mM EDTA, pH 6.8) containing 25% formamide at 34 • C. After RNA hybridization, tag-conjugating Cy3 and Cy5 dyes were circulated through the microfluidic chip for dye staining. Fluorescence images were collected using a laser scanner (GenePix 4000B, Molecular Device) and digitized using Array-Pro image analysis software (Media Cybernetics). Data from miRNA microarray were analyzed by the service provider first subtracting the background and then normalizing the signals using an LOWESS filter (Locally weighted Regression) [42]. The ratio of the two sets of detected signals (log2 transformed, balanced) and Pvalues of the t-test were calculated; differentially detected signals were those with less than .01 P-values. Multiple sample analysis involved normalization, data adjustment, ttest, and clustering. Normalization was carried out using a cyclic LOWESS. Data adjustment included data filtering, Log2 transformation, and normalization. The t-test was performed between "control" and "test" sample groups  Figure 5: The axon guidance pathway identified by Pathway-Express analysis. Pathway-Express analysis performed on the predicted targets of the 50 differentially expressed miRNAs identified, among the most significant KEGG pathways predicted to be relevant for endometriosis, the axon guidance pathway. In yellow are the predicted targets of the differentially expressed miRNAs. [43]. T-values were calculated for each miRNA, and Pvalues were computed from the theoretical t-distribution. miRNAs with P-values < .01 were selected for cluster analysis. The clustering was done using hierarchical method and was performed with average linkage and Euclidean distance metric [44] using TIGR MultiExperiment Viewer (http://www.tm4.org/mev.html). Data from real-time RT-PCR experiments are presented as the mean ± SEM. The variation among groups was compared by means of nonparametric Wilcoxon and Mann-Whitney U tests. Statistical significance was assumed for Pvalues < .05. Statistical analysis was performed with SPSS for Windows version 15.0 (SPSS; Chicago, IL).

MicroRNAs Differentially Expressed in Eutopic and
Ectopic Endometrial Tissue. In the present study, we used miRNA microarray technology to identify the pattern of miRNAs in paired eutopic/ectopic endometrium from the same patients, thus avoiding the variables attributable to heterogeneous genetic background between individuals and the effects of estrogenic stimulation during different phases of the menstrual cycle. Moreover, we considered the whole endometrial and endometriotic tissues in order to preserve the contribution of all the components of the tissues, including vascular and immune system components and to avoid potential changes in gene and miRNA expression due to cell isolation and manipulation. Microarray technology has allowed a global analysis of all miRNAs differentially expressed in ectopic versus eutopic endometrium. The initial analysis of miRNA expression in ectopic endometrium compared with eutopic endometrium of three patient samples generated a list of 84 miRNAs significantly differentially expressed (P-values < .01). The 50 miRNAs for which the expression value in ectopic endometrium was at least twofold higher or lower than in eutopic endometrium are listed in Table 1.

Real Time RT-PCR Analysis of miRNA Expression.
In order to confirm the results obtained with miRNA microarray, the expression analyses of 5 selected miRNAs was carried out by real-time RT-PCR on specimens from other 13 patients. These 5 miRNAs, namely, hsa-miR-200a, hsa-miR-200b, hsa-miR-200c, hsa-miR-182, and hsa-miR-202, were selected because their expression resulted to be highly altered in ectopic endometrium compared with the matched eutopic tissue. We verified the differential expression of the Table 2: Molecular networks constituted by the predicted miRNA targets. IPA analysis was performed in order to identify the molecular pathways and functions to which the predicted targets of the differentially expressed miRNAs belong. The networks are generated on the basis of the published literature and ranked by the P-value calculated by Fisher's exact Test. The biological processes in which the targets are involved are determined by IPA using the GOstat application P < .01.      GK, GPR64, GTF2H1, GTF2H2,  HMGN1, MLL2, MNAT1, NR2C2,  NT5E, PPARGC1A, RBBP5, SAFB,  SMAD6, TAF1, TAF2, TAF4,  TAF8, TAF9, TAF11, TAF15, TFF1,  TUBB, UTX   10E−2  1 5 Gene Expression, DNA Replication, Recombination, and Repair, Dermatological Diseases and Conditions  (Figures 1(a)-1(d)), while hsa-miR-202 expression in ectopic endometrium was increased up to 60 folds compared to eutopic endometrium (Figure 1(e)). The analysis of data according to the severity of the endometrioma, by means of nonparametric Wilcoxon and Mann-Whitney U tests, failed to reveal any significant differences in miRNA expression levels, although this may be ascribable to the group size. Further studies increasing the cohort will be necessary to completely address this issue.

Identification of Predicted miRNA Targets and In Silico
Functional Analysis. The predicted target mRNAs of the differentially expressed miRNAs common to two different search algorithms, TARGETSCAN (http://www.targetscan .org/) and PICTAR-VERT (http://pictar.mdc-berlin.de), were 3093. The functions of these predicted targets and the molecular pathways in which they could be involved were assessed using Ingenuity Pathways Analysis software (Ingenuity IPA 7.5). The predicted targets were uploaded in IPA, and the software identified 49 significant molecular networks to which the predicted targets of the differentially expressed miRNAs belong (Table 2). Among the biological functions reported to be statistically significant by IPA there were functions known to be involved in endometriosis such as gene expression, cellular growth and proliferation, cellular development, cellular movement, cell death, cell cycle, cancer, and reproductive system disorders. One of the subcategories of reproductive system disorders to be more represented, with P-value (calculated by Fisher's Exact test) of 6.1 · 10 −18 , was endometriosis with 119 molecules directly involved in this pathology (Table 3). An exemplificative network identified by IPA enriched for miRNA targets involved in endometriosis is shown in Figure 2. This network, converging on estrogen receptor 1 (ESR1), includes the DNA methyltransferases DNMT3A and DNMT3B that are validated targets of hsa-miR-29b and hsa-miR-29c, and of hsa-miR-29b, hsa-miR-29c, and hsa-miR-148a, respectively [45,46]. DNA methylation is an epigenetic modification that is involved in gene silencing, chromatin remodeling, and genome stability [47]. It has been demonstrated that DNMT1, DNMT3A, and DNMT3B are disregulated in endometriosis [48], and it has been suggested that aberrant methylation of HOXA10 and of the progesterone receptor PR-B may be responsible of the disregulation of their expression in endometriosis. Thus, this network strongly suggests a possible involvement of miRNAs in these mechanisms.    Journal of Biomedicine and Biotechnology To further analyze the possible role of these differentially expressed miRNAs in endometriosis, we performed a different analysis uploading the miRNAs directly in IPA. In this way, the software identified 6 networks, 3 of which are highly significant with known biological functions including genetic disorders, connective tissue disorders, skeletal and muscular disorders, cancer, and reproductive system disorders ( Table 4).
The difference in the number of networks identified by IPA is ascribable to the different database used by the software, as IPA uses the Argonaute 2 databases (http://www.ma.uni-heidelberg.de/apps/zmf/argonaute/) to analyse miRNAs and their known or predicted targets, and this database identified only 118 targets for the 50 miRNAs.
Next, we performed an IPA analysis on the 1203 predicted targets of the miRNAs whose differential expression between eutopic and ectopic tissue was confirmed by realtime RT-PCR. IPA software identified 49 networks and revealed that the predicted targets were enriched for biological functions such as cellular development, cell morphology, cell-mediated immune response, gene expression, cell cycle, cell death, cancer, and developmental disorders. The network with the highest score from this analysis, shown in Figure 3, includes molecules that have been implicated in endometriosis such as the TNF receptor, IL10, IL6, and FOXO1 [49][50][51][52][53][54][55].
Performing the analysis uploading directly the miRNAs in IPA, thus using the Argonaute2 database, the software identified only one network (Figure 4), the major biological Table 4: Molecular networks constituted by the predicted miRNA targets. The list of differentially expressed miRNAs was directly uploaded in IPA and an analysis was performed in order to identify the molecular pathways and functions to which the predicted targets of the differentially expressed miRNAs belong. The database used by IPA to analyze miRNAs and their targets is Argonaute2 (http://www.ma.uni-heidelberg.de/apps/zmf/argonaute/). P < .01. functions of which are cell cycle, cell death, and connective tissue disorders. This network contains PIK3R1, and its expression has been demonstrated to be upregulated in endometriosis, were it can play an essential role in TNFmediated antiapoptotic signaling [56]. Another interesting molecule present in this pathway is SIP1, a validated target of the miR-200 family, which is a factor implicated in epithelial to mesenchymal transition and tumor metastasis [57]. Thus, the observed downregulation of miR-200 family in the ectopic endometrium may have a role in the endometrial lesion development.

ID Molecules in Network
We further investigated the function of the predicted targets of the RT-PCR-validated miRNAs by using Onto-Express and Pathway-Express (http://vortex.cs.wayne.edu/) in order to categorize the targets according to Gene Ontology (GO) and KEGG pathways, respectively [58,59]. The predicted targets of the validated miRNAs were uploaded in Onto-Express and the list of the putative targets of the 475 miRNAs assayed was used as reference. Onto-Express calculates the mRNA targets in each GO category and compares it with the expected number of targets present in the GO category. Significant differences from Table 5: Gene Ontology analysis of the predicted target genes of 50 miRNAs differentially expressed. Onto-Express analysis on predicted targets of the differentially expressed miRNAs identified enrichment for biological process categories. The gene column indicates the number of predicted targets of the differentially expressed miRNAs upon the number of the targets of all miRNAs considered for the study. Significant differences from the number of targets in each GO category with the expected number of genes were calculated with the assumption of a hypergeometric distribution and P-values were adjusted with the false discovery rate (fdr) correction. P < .05.

Rank
Biological process category Genes Corrected P-value   the expected number of genes were calculated assuming a hypergeometric distribution, and P values were adjusted with the false discovery rate correction based on the number of GO categories tested. A corrected P value < .05 was considered statistically significant. Onto-Express analysis revealed enrichment for several biological processes known to be relevant in endometriosis, such as developmental process, cell death, cell cycle, and cell adhesion ( Table 5). Pathway-Express analysis identified 33 pathways significant at 5% level ( Table 6), most of which are coherent with the current knowledge on endometriosis. For instance, the most significant pathways putatively affected by the differential expression of miRNAs are MAPK and axon guidance the latter shown in Figure 5. While MAPK pathway, which is involved in several cellular functions, such as cell proliferation, migration, and differentiation, is clearly relevant for endometriosis, axon guidance, at first may appear unrelated to this pathology. However, nerves and blood vessels are highly interconnected, both physically and in their morphogenesis. Indeed, it has been demonstrated that several molecules involved in axon guidance, such as semaphorins, plexins, and neuropilins, are also strongly implicated in angiogenesis [60], a biological process essential for endometriosis. Intriguingly, this pathway contains ROBO1, and its expression, higher in ectopic endometrium compared to eutopic tissue, positively correlates with endometriosis recurrence [61], thus suggesting that miRNAs may take part in tuning ROBO1 expression and have a role in the recurrence of the pathology. Hypothesising that the genes differentially expressed common to both studies are likely those specific to endometriosis independently from the site of the lesion, we restricted the analysis to the differentially regulated genes in eutopic and ectopic endometrium common to the two studies that are also predicted targets of the 50 miRNAs (Table 7). IPA analysis identified 5 molecular networks, the most relevant functions of which being cancer, cell cycle, and reproductive system disease ( Table 8). The overlap of networks generated by IPA is shown in Figure 6. In this graphical representation the most relevant nodes are the transcription factor SP1, tumor necrosis factor (TNF), and SRC, in remarkable agreement with the nodes of the most significant networks obtained by IPA analysis performed on the distinct datasets of differentially expressed genes in ovarian and peritoneal endometriosis (data not shown). Table 7: Genes aberrantly expressed in ovarian and peritoneal endometriosis that are predicted targets of the differentially expressed miRNAs. The miRNAs predicted to regulate the expression of the genes known to be aberrantly up-(↑) or downregulated (↓) in both ovarian and peritoneal endometriosis were identified by TARGETSCAN and PICTAR algorithms. MicroRNAs whose regulation is in accordance with the resulting expression of their predicted target genes are reported in bold.

Conclusions
MicroRNAs are predicted to regulate a large fraction of protein-coding genes, as computational analysis reveals that an average miRNA could have as many as 100 or more target genes. On the other hand, a single gene may have target sites for several distinct miRNAs, allowing a fine tuning of gene expression by miRNAs.
In the present study, we used miRNA microarray technology to identify the miRNAs differentially expressed in paired eutopic/ectopic endometrium from the same patients and bioinformatics tools to identify their predicted targets as well as the molecular networks and the biological functions they may affect.
Comparing miRNA expression profiles among the different subjects, we identified 50 miRNAs differentially expressed in ectopic versus eutopic samples. Several of these miRNAs were also reported to be differentially expressed in two recent studies [62,63], although with a modulation occasionally discordant from our results. This, joint to a notable accordance between their predicted targets and the genes reported to be differentially expressed in two studies of gene expression [23,24], consolidates the hypothesis of a possible role of miRNAs in the pathogenesis of endometriosis.
The miRNAs-predicted targets were identified by the intersection of the results from two different search algorithms, and the biological functions the differentially expressed miRNA may affect were identified by Onto-Express and IPA software. Functional analysis, performed using IPA software, was carried out uploading either the predicted targets or the differentially expressed miRNAs, thus using different databases for miRNA targets. As expected, Table 8: Molecular networks constituted by the common differentially expressed transcripts in ovarian and peritoneal endometriosis predicted to be targets of the 50 miRNAs. Differentially expressed genes common to both ovarian and peritoneal endometriosis that are predicted targets of the 50 differentially expressed miRNAs were uploaded in IPA in order to identify the molecular networks and functions to which they belong. P < .01. Cell-To-Cell Signaling and Interaction, Cell-mediated Immune Response, Cellular Growth and Proliferation the different algorithms used to predict miRNA targets led to the identification of different molecular networks. Still, in both cases, the identified networks contained several transcripts known to be implicated in endometriosis and with their main biological functions linked to the disease. Since the targets of miRNAs are just predictions based on mathematical algorithms, the choice of the algorithm may radically modify on the whole the list of the predicted target genes and that of the molecular networks they belong to. For this reason, the validation of miRNA targets in vitro, in a cellular system, is essential to evaluate the contribution of each miRNA to the overall modulation of gene expression.