The molecular signature of endometriosis-associated endometrioid ovarian cancer differs significantly from endometriosis-independent endometrioid ovarian cancer

To identify possible differences in gene expression between EAOC and OC not associated with endometriosis, we used total RNA microarray technology and validated randomly selected genes by real-time PCR. There were no statistically significant differences in age, menopausal status ( Tables 1 and 2 ), and histologic stage between the samples used for microarray analysis and real-time PCR in each sample group. In addition, microarray analysis showed no difference in expression of the detected genes depending on menopausal status within one sample group. For an overview, we listed the differentially expressed genes in two main groups: genes up- or down-regulated in endometriosis and EAOC but not in OC and benign ovaries ( Table 4 ); and genes that are up- or down-regulated in EAOC and OC but not in ovarian endometriosis and benign ovaries ( Table 5 ). Representative genes in each group were validated by quantitative real-time PCR, including 25 additional samples to enlarge the number of patients (genes printed in Tables 4 and 5 in boldface). The first group of genes is equally regulated in endometriosis and EAOC but not in OC and benign ovaries, as validated by quantitative real-time PCR. The highest expression of the cytokine SICA2 (monocyte chemotactic protein 1 [MCP-1]) is found in EAOC compared with endometriosis (1,689 times higher; P =.00051), OC (425,854 times higher; P =.00017), and benign ovary (216,169 times higher; P =.003). There is no statistical difference in the very low expression of SICA2 in OC and benign ovary ( P =.84). Quantitative real-time PCR shows a trend for higher expression of SICA2 in endometriosis compared with OC ( P =.07) and benign ovary ( P =.12). Cytokine CCL14 expression is 89 times higher in endometriosis and 87 times higher in EAOC when compared to OC ( P =.0021) and benign ovary ( P =.038). There is no statistically significant difference in the elevated expression of CCL14 in endometriosis and in EAOC ( P =.98). CCL14 expression is equally low in OC and benign ovary ( P =.98). Conversely, the lowest expression of the growth factor TDGF1 is observed in endometriosis and EAOC compared to benign ovaries (22 times lower; P =.00039). There is no statistically significant difference in the expression of TDGF1 between endometriosis (11 times higher; P =.02) and EAOC (7 times higher; P =.59) compared with OC. The expression of small inducible cytokine subfamily A, member 14 (CCL14), small inducible cytokine A2 (SICA2/MCP-1), and TDGF1 was validated by quantitative real-time PCR with a total of 47 samples. The second group is comprised of genes that are up- or down-regulated in EAOC and OC but not in endometriosis and benign ovaries, as validated by quantitative real time PCR. We investigated the expression of SPINT1 and found it to be more highly expressed in EAOC and OC ( P =.92) compared with endometriosis (11 times higher; P =.01) or benign ovary (8 times higher; P =.02). In the same group, the expression of the intermediary filament Keratin 8 is higher in EAOC and OC ( P =.63) compared with endometriosis (18 times higher; P =.001) or benign ovary (6 times higher; P =.004). We found the expression of the transcription factor FoxM1B to be higher in EAOC and OC ( P =.91) compared with ovarian endometriosis (27 times higher; P =.001) or benign ovary (18 times higher; P =.03). Additionally, the expression of FOLR1 is higher in EAOC and OC ( P =.91) compared with ovarian endometriosis (20 times higher; P =.0005) or benign ovary (159 times higher; P =1.56E-5). The expression of FOLR1 is elevated in ovarian endometriosis compared with benign ovary (7 times higher; P =.01) as well. The expression of another gene, CRABP1, is higher in EAOC and OC ( P =.62) compared with endometriosis (107 times higher; P =.004) or benign ovary (158 times higher; P =.005). Finally, the expression of Claudin7 is elevated in OC compared with EAOC (3 times higher; P =.0001), ovarian endometriosis (40 times higher; p =1.22E-7), or benign ovary (45 times higher; P =.01). The second highest expression of Claudin7 is found in EAOC compared with endometriosis (12 times higher; P =.0001) or benign ovary (13 times higher; P =.04). StAR, however, is down-regulated in OC (196 times lower; P =.0008) and EAOC (40 times lower; P =.01) compared with benign ovaries and endometriosis ( P =.73). There is no statistically significant difference in the low expression of StAR in EAOC and OC ( P =.15).

We determined that SICA2, CCL14, and TDGF1 are equally regulated in endometriosis and EAOC compared with OC and benign ovaries. In this group of genes, cytokines in particular were up-regulated in ovarian endometriosis and EAOC. Cytokines are important mediators of intercellular communication within the immune system. The peritoneal fluid of women with endometriosis shows increased levels of many cytokines ( 10 , 11 ), supporting the theory that endometriosis is a chronic inflammatory process with an altered immune answer. Small inducible cytokine subfamily A, member 14 (CCL14), and SICA2/MCP-1 are members of the small inducible gene family and play a role in the recruitment of monocytes to sites of injury and infection. They are secreted by several cell types, such as endothelial cells, fibroblasts ( 12 ), and leukocytes ( 13 ). Their concentration is high in the peritoneal fluid of women with endometriosis and correlates with the severity of the disease ( 14 ). SICA2 and CCL14 seem to be specific markers for endometriosis and ovarian cancer associated with endometriosis. Although the direct or indirect effects of chemokines on tumor invasion are poorly evaluated, it is known that chemokines affect tumor development by attracting immuno-competent cells with pro- or antitumoral activities ( 15 ). Teratocarcinoma-derived growth factor (TDGF1, Cripto-1) was down-regulated in endometriosis and EAOC versus OC. Interestingly, the central portion of this protein is structurally similar to human transforming growth factor alpha and epidermal growth factor, both proteins involved in the pathogenesis of ovarian cancer. Strizzi et al. ( 16 ) reviewed the role of TDGF1 during embryogenesis, cell migration, invasion, and angiogenesis and concluded that these activities may relate to cellular transformation and oncogenesis. TDGF1 is expressed in 47% of ovarian cancer samples ( 17 ). The gene StAR is down-regulated and the genes SPINT1, Keratin 8, FoxM1B, FOLR1, CRABP1, and Claudin 7 are up-regulated in EAOC and OC but not in ovarian endometriosis or benign ovaries. We identified the steroidogenic acute regulatory protein (StAR) in the group of proteins down-regulated in EAOC and OC compared with ovarian endometriosis and benign ovary. StAR mediates the rapid increase in pregnenolone synthesis stimulated by tropic hormones and is mainly expressed in the adrenal cortex. Transcript variants of StAR were detected in ovary and testis ( 18 ). Avery recent study was able to demonstrate the expression of a few steroidogenic genes in endometriosis. StAR and aromatase were among the most important genes identified ( 19 ). Abd-Elaziz et al. ( 20 ) studied the expression of StAR in ovarian cancer and found a significant inverse relationship between FIGO stage and the residual size of the tumor. Serine protease inhibitor 1 (SPINT1) was up-regulated in EAOC and OC compared with ovarian endometriosis and benign ovary. Hepatocyte growth factor (HGF) stimulates tumor cell-cell interactions, matrix adhesions, migration, and angiogenesis. SPINT1 inhibits the generation of biologically active HGF ( 21 ). Although there are currently no publications that demonstrate an association between SPINT1 and ovarian cancer or endometriosis, there are several reports concerning the role of SPINT1 in cancer cells. Parr et al. examined the expression of SPINT1 in human cancerous and normal breast tissue ( 22 ). The breast cancer specimen expressed a significantly higher level of SPINT1 compared with the normal background tissues. We identified Keratin 8 in the group of genes highly up-regulated in EAOC and OC. Intermediary filaments, such as keratin 8, are involved in cell motility and cancer progression. Stimpfl et al. ( 23 ) studied the expression of keratin 8 in seven human ovarian cancer cell lines. Keratin 8 was detectable in all of the cell lines and was strongly expressed in four of them. As yet, there is no relevant literature about keratin 8 and endometriosis or ovarian cancer in vivo. A correlation between the expression of keratin 8 and the prognosis of certain human cancer seems to exist ( 24 ). Forkhead Box M1B transcription factor (FoxM1B) was found to be highly up-regulated in EAOC and OC but not in ovarian endometriosis or benign ovary. The transcription factor FoxM1B regulates expression of cell cycle genes essential for progression into DNA replication and mitosis. Although there is, as yet, no literature on the expression of FoxM1b in ovarian cancer or endometriosis, FoxM1b has been found to be expressed in a variety of distinct human cancers. Indeed, Kim et al. ( 25 ) studied the role of FoxM1B in human nonsmall cell lung cancers. Transient transfection experiments with lung adenocarcinoma cells showed that depletion of FoxM1B levels by short interfering RNA caused diminished DNA replication and mitosis and reduced growth of cell colonies. Another protein that is up-regulated in EAOC and OC but not in endometriosis or normal ovarian tissue is folate receptor 1 (FOLR1). FOLR1 has a high affinity for folic acid and for several reduced folic acid derivates. It mediates delivery of 5-methyltetrahydrofolate to the interior of cells. As early as 1991, FOLR1 was described as a marker for ovarian cancer ( 26 ). Hibbs et al. ( 27 ) found FOLR1 among the genes specifically up-regulated in ovarian carcinoma tissue compared with more than 350 other tissue samples. We detected the up-regulation of retinoic acid–binding protein type 1 (CRABP1) in EAOC and OC compared with ovarian endometriosis and benign ovary. Cellular retinoic acid binding is assumed to play an important role in retinoic acid–mediated differentiation and proliferation processes. A very early study investigated the concentration of CRABP in carcinomas of the ovary compared with normal ovary and found highly elevated levels in ovarian carcinomas ( 28 ). Hibbs et al. ( 27 ) studied gene expression in ovarian carcinomas, ovarian carcinoma metastases of the omentum, and in 50 normal ovaries by microarray analysis. CRABP1 was one of several genes found to be up-regulated in ovarian carcinomas. In the group of genes found to be up-regulated in EAOC and especially in OC, but not in benign ovaries or endometriosis, we detected Claudin 7. Claudins are involved in the formation of tight junctions between epithelial cells. Tight junctions restrict lateral diffusion of lipids and membrane proteins and thereby define the border between the apical and basolateral compartments of epithelial cells ( 29 ). Although Kominsky et al. ( 30 ) found that expression of claudin 7 was decreased in cancer such as invasive ductal carcinomas of the breast compared with normal breast epithelium, overexpression of other claudins, such as claudin 4, has been demonstrated in several types of cancer, including ovarian cancer ( 31 , 32 ). The results of the present study emphasize the usefulness of microarray analysis in elucidating the genetic profiles of ovarian endometriosis and ovarian cancer. We identified two main groups of genes. The first group was composed of the cytokines SICA2 and CCL14 and the growth factor TDGF1. These genes were equally regulated in endometriosis and EAOC but not in OC or benign ovaries ( Table 4 ). It seems likely that the regulation of the autoimmune system and of inflammatory cytokines is very important in the etiology of endometriosis and EAOC. The second group was composed of a diversity of genes: StAR, SPINT1, Keratin 8, FoxM1B, FOLR1, CRABP1, and Claudin 7 ( Table 5 ). They all play a central role in cell-cell interaction, differentiation, and cell proliferation and are equally regulated in EAOC and OC but not in ovarian endometriosis or benign ovaries. Therefore, these genes may be viewed as potential markers that could be used to distinguish malignant from benign ovarian tissue and to predict the development of ovarian cancer in women with endometriosis. Further studies are necessary to show if it will be possible to identify those endometriosis patients at risk for ovarian cancer who would benefit from prophylactic treatment.

We obtained tissue from patients undergoing either an ovarectomy for treatment of endometriosis, EAOC, or OC or a prophylactic ovarectomy of benign ovaries after informed preoperative consent. The study was approved by the respective local Institutional Review Board. Clinical data of the patients for microanalysis are listed in Table 1 and for real-time polymerase chain reaction (PCR) analysis (with additional samples) in Table 2 . The diagnosis of endometriosis, EAOC, OC, or benign ovaries was confirmed histologically. Total RNA was extracted from the samples with the RNeasyMini Kit from Qiagen (Hilden, Germany) following the manufacturer’s instructions for total RNA isolation from tissue. RNA concentration was quantified by UV-spectroscopy and its quality checked by analysis on a LabChip (BioAnalyzer; Agilent Technologies, Santa Clara, CA). The cDNA was synthesized from 5 µg total RNA starting with the annealing to a T7-(dT) 24 Primer. Reverse transcription and second-strand synthesis were carried out using a complete cDNA reaction mix. Synthesis of biotin-labeled cRNA was performed using the BioArray High Yield RNA Transcription kit (Enzo Diagnostics, Farmingdale, NY). The amplified cRNA was purified with an affinity resin column and fragmented. The cRNA amount was determined by UV-spectroscopy and distribution of fragment sizes analyzed on a LabChip (BioAnalyzer). The fragmented cRNA, taken from each patient separately, was hybridized to the HG U133A/MG U74Av2 array (Affymetrix, Santa Clara, CA). Hybridization was carried out at 45°C in a hybridization oven for 16 hours. Washing and staining of the array was performed using the GeneChip fluidics station protocol EukGE-WS2. Probe arrays were then scanned twice at 3 µm resolution using the GeneChip System confocal scanner (Hewlett-Packard, Santa Barbara, CA). Data analysis was performed using Microarray Suite Version 5.0 software and Data Mining Tool version 2.0 software (Affymetrix). Because the sample groups were of different sizes (five or seven samples per group), methods such as analysis of variance or significance analysis of microarrays for multiple correcting testing were not applicable. Therefore, classic filter technique of twofold or greater change of mean gene expression between groups was used in two substeps to identify genes of interest. Total RNA from tissues was subjected to a reverse-transcription (RT) reaction using Superscript II Reverse Transcriptase (Invitrogen, Karlsruhe, Germany). Quantitative real-time PCR was performed in a Light Cycler System using the Platinum Sybr Green qPCR Supermix UDG kit as recommended by the manufacturer (Invitrogen). Primer sets are shown in Table 3 . Quantification of the targets in the unknown samples was performed using a relative quantification method with external standard references. The target concentration is expressed relative to the concentration of a housekeeping gene, hypoxanthine guanine phosphoribosyltransferase 1 (HPRT-1). Among four different possible housekeeping genes (GAPDH, β-Actin, HPRT-1, and PGK-1), HPRT-1 showed the smallest differences between the samples of endometriosis, EAOC, OC, and benign ovaries in the microarray experiments. After each run, melting curve analysis was performed to verify the specificity of the PCR reaction. Calculation of 2 −δδCT was performed to determine fold change of gene expression.