Purpose: To describe the eicosanoid profile and differentially expressed eicosanoid and arachidonic acid pathway genes in tissues from patients with advanced epithelial ovarian cancer (EOC).
Experimental Design: We first employed electrospray tandem mass spectrometry to determine tissue-specific concentrations of the eicosanoids prostaglandin E2 (PGE2), the hydroxyeicosatetraenoic acids (12-HETE and 5-HETE), and leukotriene (LTB4), selected for tumor growth potential, and two other bioactive lipids (15-HETE and 13-HODE) with tumor cell proliferation interference potential. The cellular location of eicosanoid activity was identified by immunofluorescence antibody costaining and confocal microscopy. Differential analysis of eicosanoid and arachidonic pathway genes was done using a previously validated cDNA microarray platform. Tissues used included EOC tumor, tumor-free malignant peritoneum (MP), and benign peritoneum (BP) from patients with benign pelvic disease.
Results: (a) Eicosanoid products were detected in tumor, MP, and BP specimens. PGE2 levels were significantly elevated in tumors in an overall comparison with MP or BP (P < 0.001). Combined levels of PGE2, 12-HETE, 5-HETE, and LTB4 increased progressively from low to high concentrations in BP, MP, and tumors (P = 0.012). Neither 15-HETE nor 13-HODE showed a significant opposite trend toward levels found in BP. (b) Tissue specimens representing common EOC histotypes showed strong coexpressions of cyclooxygenases (COX-1) and prostaglandin E synthases (PGES-1) on tumor cells, whereas intratumoral or peritumoral MO/MA coexpressed COX-1 and COX-2 and PGES-1 and PGES-2, respectively. (c) cDNA microarray analysis of MP, BP, and tumor showed that a number of eicosanoid and arachidonic acid pathway genes were differentially expressed in MP and BP compared with tumor, except for CYP2J2, which was increased in tumors.
Conclusions: Elevated levels of eicosanoid metabolites in tumors and differential expression of eicosanoid and arachidonic acid pathway genes in the peritoneum support the involvement of bioactive lipids in the inflammatory tumor environment of EOC.
- ovarian cancer
- Gynecological cancers: ovarian
Eicosanoids is a collective term for oxygenated derivatives of different 20-carbon fatty acids such as leukotrienes and prostanoids (prostaglandins, prostacyclins, and thromboxanes). These bioactive lipids have important functional roles in regulating many physiologic processes and inflammatory responses (1). Eicosanoid production is a tightly regulated process that depends on (a) the acylation and transfer of arachidonic acid into specific phospholipid pools by arachidonic acid–selective acyltransferase and transacylase reactions and (b) the release of these pools by a variety of phospholipase A2 (PLA2) enzymes (2).
Arachidonic acid liberated by PLA2 is metabolized by cyclooxygenase (COX) and lipoxygenase (LOX) enzymes variously expressed in cells and tissues, and contribute to the production of specific metabolites (1, 3, 4). These bioactive lipids or eicosanoids then exert their biological effects in an autocrine or paracrine manner by binding to specific G-coupled receptors (5–8). In a previous study, we showed that sPLA2 (group 2a) gene transcripts in tumor-free specimens from the malignant peritoneum (MP) of patients with epithelial ovarian cancer (EOC) were differentially expressed when compared with those in tumor tissue (9). We also verified the expression of PLA2 at the protein level in MO/MA in ascitic specimens (10). MO/MA are the most prominent population of inflammatory cells observed in the tumor and peritoneal microenvironment of EOC (10, 11).
Despite the well-described roles of eicosanoids in cancer inhibition or proliferation, few studies have focused on tissue-specific eicosanoid metabolism, largely because methods for identifying and quantifying multiple COX- and LOX-derived products in specific tissues were lacking. One of the coauthors (R.A. Newman) has established an analytic procedure for this purpose based on liquid chromatography/electrospray tandem mass spectrometry analysis (12, 13). This lipidomics technique is currently being used to identify and measure specific changes associated with endogenous COX and LOX activities in cells and tissues; such changes are thought to reflect alterations in eicosanoid profiles between normal, inflamed, and malignant tissues. These studies have shown significant differences in patterns of eicosanoid metabolism between different types of malignant tissues; for example, it has been shown that colon cancer is regulated in part by the relative expression of 13-S-hydroxyoctadecadienoic acid (13-S-HODE; ref. 14), that glioblastomas invariably overexpress 5-LOX,6 and that prostate cancer is connected to an age-associated decline in the expression of 15-hydroxyeicosatetraenoic acid (15-HETE), a tumor suppressor (15), as well as up-regulation of 12-LOX which, in turn, inhibits Rb tumor suppressor activity (16).
In the current study, we employed electrospray tandem mass spectrometry analysis combined with gene expression analysis to produce the first description of specific variations in eicosanoid products and enzymes involved with eicosanoid and arachidonic acid pathways in ovarian tumor tissues. Our findings show that endogenous levels of prostaglandin E2 (PGE2), 5-HETE, and 12-HETE increase progressively from benign peritoneum (BP) tissue, through EOC peritoneum (MP), to EOC tumor. Distribution of the expressive enzymes involved in these pathways at the cellular and gene transcript levels is shown. Differential gene expression profiles involving eicosanoid and arachidonic acid metabolism pathways were shown by cDNA microarray analysis.
Materials and Methods
Electrospray tandem mass spectrometry analyses. Endogenous levels of key eicosanoids from cells and tissues were measured by validated, published methods (12, 17). Fresh-frozen tumor and generally matched peritoneum samples were obtained from patients with EOC and pelvic peritoneum samples from patients with benign pelvic disease and EOC tumors. All specimens were obtained under an Institutional Review Board–approved protocol. MP specimens with contaminating tumor as identified by a gynecologic oncology pathologist (M. Deavers) were excluded from eicosanoid and microarray analyses. Specimens for eicosanoid analyses were obtained in the operating room, snap-frozen directly, and stored at −80°C until analyzed. Endogenous eicosanoids were measured with a Quattro Ultima tandem mass spectrometer (Micromass) equipped with an Agilent HP1100 binary high-pressure liquid chromatography inlet. Eicosanoids were separated on a Luna 5μ Phenyl-Hexyl 2 × 150 mm column (Phenomenex) with a rapid linear gradient of 70% to 90% methanol in 4 min over 10 min. The mobile phase consisted of 10 mmol/L of ammonium acetate (pH 8.5) and methanol (30:70). Fragmentation of all compounds was achieved by using argon as the collision gas. This method produces excellent linearity and a lower limit of quantification for eicosanoids of 1 ng/mL, which is adequate to assess endogenous eicosanoid metabolism that commonly occurred as ng/mg of protein in 5 × 106 cell aliquots or tissue samples. We have used this method to characterize in detail eicosanoid metabolism in human lung, prostate, breast, and colon tissues as well as in numerous cell lines. The following eicosanoids were measured: PGE2 (product of cyclooxygenase pathway), 5-HETE (product of 5-lipoxygenase or 5-LOX), LTB4 (product of leukotrienes-A4 hydrolase and downstream product of 5-LOX), 12-HETE (product of 12-LOX), 15-HETE (product of 15-LOX2), and 13-HODE (product of 15-LOX1). Eicosanoid levels were compared between groups using ANOVA and t tests. Some analyses were repeated using Kruskal-Wallis and Wilcoxon tests; results were similar and are not reported in this article. Statistical significance was declared at P < 0.05. No adjustment was made for the multiplicity of testing.
Fluorescence labeled multi-antibody costaining and confocal microscopy visualization of peritoneal and tumor biopsy specimen cells. Our method to prepare tissues for staining and reading of stained slides are fully described elsewhere (10). A sequential staining technique was used as follows: 3 h incubation with the first primary antibody (red) at room temperature, overnight incubation with the second primary antibody (blue/green) at 4°C, Secondary antibodies were incubated with cryostat-prepared sections for 1 h after the incubations with the primary antibodies were complete. Nonspecific binding was blocked by adding 5% normal goat serum for 30 min. The primary antibodies used were COX-1 monoclonal mouse antibody (12E12), IgM, 1:25 dilution (GeneTex, Inc.); COX-2 mouse monoclonal antibody, IgG1, 1:50 dilution (Cayman Chemical Co.); prostaglandin E synthase-1 (PGES-1; microsomal) polyclonal rabbit antibody, IgG, 1:50 dilution (Cayman); prostaglandin E synthase-2 (PGES-2; microsomal) polyclonal rabbit antibody, IgG, 1:50 dilution (Cayman); 15-LOX2, 15-LOX2 polyclonal rabbit antibody, IgG, 1:50 dilution (Cayman); 5-LOX polyclonal rabbit antibody, IgG, 1:50 dilution (Cayman); mouse anti-human CD163, IgG1, 1:200 dilution (Serotec); mouse anti-human cytokeratin clone AE1/AE3, IgG1, κ, 1:100 dilution (DAKOCytomation). Secondary antibodies were Cy3 (red)-conjugated affinipure goat anti-mouse IgM, μ chain specific; Cy3 (red)-conjugated affinipure goat anti-mouse IgG, Fcγ subclass 1 specific; Cy3 (red) conjugated affinipure goat anti-rabbit IgG (H+L); Cy2 (green)-conjugated affinipure goat anti-mouse IgG1, Fcγ subclass 1–specific; Cy5 (blue)-conjugated affinipure goat anti-mouse IgG, Fcγ subclass 1 specific (all from Jackson ImmunoResearch Labs).
When two unconjugated primary antibodies from the same host species and the same class of immunoglobulin IgG1 were used, any open antigen binding sites on the first and secondary antibodies were saturated with 5% normal mouse serum. Mouse immunoglobulins are sterically covered with monovalent affinipure Fab fragment goat anti-mouse IgG (H+L), 1:65 dilution (Jackson ImmunoResearch Labs, Inc.). Negative controls employed secondary antibodies alone.
Tissue sections were mounted with Slow-Fade Gold Anti-Fade reagent (Molecular Probes) and viewed with an Olympus FV500 laser scanning confocal microscope; images were captured at 400× and 600× magnification by Fluoview software Version 4.3.
Differential gene expression analysis. Specimen collection and sample preparation were as described previously (9) but with a substantially increased sample size. Specimens for microarray analyses were transferred directly to the laboratory from the operating room in cold saline, then snap-frozen in RNAlate (Ambion, Inc.), stored at −80°C. Processing of frozen material and extraction of RNA with quality controls were previously described (9). Among the specimens, 15 of them were from peritoneum associated with benign pelvic disease, designated “BP”, 35 specimens from EOC or histopathologically related tumors, designated “Tu”, and 27 tumor-free peritoneal specimens designated “MP.” Specimens that had microscopic tumor invasion in subsequent histopathologic examination were excluded. Total RNA isolation, RNA amplification and array hybridization were done as described (18). Custom microarrays were printed at the Immunogenetics Section, Department of Transfusion Medicine, Clinical Center, NIH with a configuration of 32 × 24 × 23 and containing 17,500 elements. For a complete list of genes included in the Hs-CCDTM-17.5k-1px, printing is available at our web site.7 Genes involved in eicosanoid and arachidonic acid pathways from the cDNA microarray data set were selected according to the KEGG pathway finder8 from the complete cDNA microarray data set after normalization and background subtraction, using BRB array tool.9 Two-sample t tests were done between MP versus BP, MP versus tumor and BP versus tumor. In addition, F tests (n = 3) were also done with 10,000 random permutations using the BRB array tool. Genes with P < 0.05 were defined as significantly differentially expressed genes and were visualized by cluster and TreeView analysis. A selected gene group was used in the analysis.
Eicosanoid metabolites increased in EOC tumor, MP, and BP tissues. Eicosanoid metabolite levels were measured on frozen tumor specimens obtained from 12 chemotherapy-naïve patients with EOC. Eleven of 12 patients in the malignant group had stage III or IV carcinomas with serous components. Patients with benign pathology included entities such as benign ovarian epithelial or germ cell tumors (two patients), low malignant potential tumors of the ovary without invasive implants (four patients), chronic pelvic inflammatory disease (two patients). Median ages were 68 (range, 61-80) for the malignant group and 62 (range, 28-78) for the benign group. There were eight matched specimens of tumor and MP peritoneum and separate MP or tumor specimens from four patients, and BP was obtained from eight patients with a variety of benign pelvic conditions who were scheduled for pelvic abdominal surgery. Only MP specimens that were tumor-free were used.
Quantitative eicosanoid analysis of tumor, MP or BP tissues were analyzed for PGE2, 5-HETE, 12-HETE, LTB4, 15-HETE, and 13-HODE, and results are shown in Fig. 1 and Table 1 , expressed in nanograms of eicosanoid per milligram of tissue protein. PGE2, an eicosanoid commonly associated with malignant disease, was significantly elevated in tumor compared with either MP (P = 0.003) or BP (P = 0.004; Fig. 1A). When PGE2, 5-HETE, 12-HETE, and LTB4 (which are metabolites more clearly identified with tumor cell proliferation) are averaged, they show a significant linear trend upward from BP through EOC peritoneum and EOC tumor (P = 0.012; Fig. 1B). Other individual comparisons were not statistically significant (Table 1). Although mean values for the 15-LOX2 product, 15-HETE, and the 15-LOX1 product, 13-HODE, seemed to move in the opposite direction, the trends were not statistically significant (Fig. 1; Table 1).
Immunocostaining of eicosanoid pathway enzymes. We next examined coexpression of eicosanoid products by IIF using confocal microscopy on: tumor tissue, peritoneum of EOC patients, and peritoneum from patients with benign pelvic disease on two to three individual patients. We showed, as others have (19), that COX-1 expression was strongly expressed on the tumor cells, whereas COX-2 was only weakly expressed on tumor cell islets. In addition, expression of PGES-1 and PGES-2 paralleled the results for COX-1 and COX-2 with weak expression of PGES-2 (Fig. 2 ). In contrast, COX-1, COX-2, and PGES-1 and PGES-2 were each coexpressed on CD163+ MO/MA in the peritoneum of EOC patients (Fig. 3 ) although costaining for COX-1 and PGES-1 seemed more prominent. 15-LOX2, an enzyme responsible for 15-HETE production was coexpressed on tumor cells (Fig. 2) and on surface mesothelial cells, stromal cells, on a few of the resident MO/MA, and in the stroma of benign and EOC patients (Figs. 3 and 4 ). Similarly, 5-LOX, an enzyme responsible for synthesis of 5-HETE and which contributes to the downstream leukotriene pathway, was coexpressed on tumor cells and MO/MA (data not shown).
Eicosanoid pathway gene profiling. There were 36 patients with ovarian or Mullerian-type malignancies and 15 with benign ovarian or uterine pathology who provided specimens for gene profiling. The 36 patients in the malignant group included the following demographics: median age, 65 years (range, 34-80); histopathology, serous (22), serous/endometrioid, endometrioid (4 each), undifferentiated (3) mucinous (2), and mixed malignant Mullerian tumor (1); stage I (1), stage II (4), stage III (23), stage IV (8). Twenty-seven of 35 samples had tumor and MP specimens from the same patients (one patient had MP only).
The 15 patients in the benign group included: median age, 53 years (range, 28-82); histopathology, chronic pelvic inflammatory disease (PID) (two hydrosalpinges, one chronic salpingitis), three; cystadenofibroma (ovary), fibrothecoma (including two leiomyomatas), two each; cystadenofibroma (ovary and appendix), corpus luteum, fibromatous nodule (and leiomyoma), adenomyosis, cystic teratoma (and serous adenofibroma), mucinous tumor, low malignant potential (LMP) and endometriosis (and LMP/serous cystadenofibroma), one each. Genes differentially expressed (P < 0.05) among MP, BP, and tumor are shown in Table 2 and Fig. 5 . Ratios are shown according to the central method for display using a normalization factor (20).
Based on our previous finding that sPLA2 was differentially expressed in MP compared with tumor, we extended the comparison of tumor with MP as well as BP using an expanded list of selected genes associated with eicosanoid metabolites and arachidonic acid metabolism pathways. The results show that of the 49 genes that were analyzed, 17 genes were differentially expressed in MP compared with tumor and 15 genes in BP compared with tumor (Table 2; Fig. 5), the exception being cytochrome P450, family 2, subfamily J, polypeptide 2 (CYP2J2), also called arachidonic acid epoxygenase, which was increased in tumor relative to MP and BP. No multiple comparison correction analysis was done.
The list of genes that we found increased in BP or MP relative to tumor tissue include the following: AKRIC3 (catalyzes the reduction of prostaglandins—formation of 9α11βPGE2 from PGD2, PGH2, in the presence of NADPH); CBR3 (carbonyl reductase or PGE2 9 reductase)—metabolizes prostaglandins, steroids, and various pharmacologic agents; glutathiamine peroxidase (GPX3) inhibits 5-LOX—catalyzed with H2O2 and lipid hydroperoxide; LYPLA3—regulates phospholipids lysozymal enzyme and has Ca-independent PLA2 and transcyclase activity; FBXl7 involved with phosphorylation-dependent ubiquitination. In addition to the above enzymes, a number of PLA2s and a lysozymal phospholipases were increased in MP and often in BP tissues compared with tumor tissue (Table 2). In contradistinction to these genes, CYP2J2, arachidonic acid epoxygenase which catalyzes the reactions involved with drug metabolism, cholesterol, and other lipids, was decreased in MP and BP relative to tumor.
Our data show that PGE2 concentrations were increased in tumor tissue compared with MP or BP, as was a grouping of PGE2 and three other eicosanoids having the potential to drive proliferation of tumor and/or increase capillary permeability. The elevated eicosanoids within EOC tissue specimens seems consistent with the expression pattern seen for COX-1 and PGES-1, versus COX-2 and PGES-2, which are only poorly expressed. The latter is also in agreement with recent reports showing that COX-1 but not COX-2 was enhanced at both the RNA and protein levels in EOC (19), and that PGE2 seems to be regulated by COX-1 but not by COX-2 (21). Vascular endothelial growth factor, a promoter of angiogenesis in EOC, is also selectively inhibited by a COX-1 inhibitor—an effect reversed by PGE2 (19).
Using differential gene expression analysis, we showed that a number of genes involved with bioactive lipids within MP or BP specimens were up-regulated when compared with the gene expression within EOC tumor specimens. These genes are specifically linked to eicosanoid and arachidonic acid metabolism pathways. These findings seemed to contrast with our results showing elevated tissue levels of PGE2 and other eicosanoids compared with MP or BP in the tissue. However, several of the genes, such as AKR1C3, CBR3, GPX3, and LYPLA3, may actually serve to regulate levels of prominent eicosanoids. Moreover, 15-hydroxyprostaglandin dehydroxygenase was recently shown to contribute to the inactivation and degradation of prostaglandins in colon cancer (22), supporting a tumor suppressor role for this enzyme.
Increased phospholipase activity observed in MP and BP would likely represent an early step for the production of eicosanoids including PGE2, through the release of arachidonic acid from membranes. The presence and relatively higher expression of enzyme transcripts involved with inhibition or degradation alongside production, which suggests that production might be coordinately regulated in normal tissues, or even MP, is intriguing. Conversely, the loss of tight regulation in tumor tissue may be an etiologic factor in this disease.
Present data may implicate inflammation, perhaps sustained and produced by certain eicosanoids as potentially important in the development and progression of EOC. The role of other bioactive lipids typically associated with inhibition of proliferation also requires further examination. The tumor suppressor functions of 15-LOX2, for example, were previously shown in a prostate model (23). Here, we show that the peritoneum of normal subjects sometimes expressed 15-LOX2 and its product 15-HETE. These findings could suggest a protective role for 15-HETE in the absence of elevated levels of PGE2 in the normal peritoneum (22). Collectively, our data suggest that the presence of elevated levels of certain eicosanoids such as PGE2, 12-HETE, 5-HETE, and perhaps LTB4, might promote tumor progression, whereas others, such as 15-HETE and 13-HODE, might interfere with the progression to malignancy. This concept has also been considered by others (24) but not in EOC.
A number of genes that were differentially expressed in MP, BP, or tumor have previously been associated with cancer, if not specifically with EOC itself. For example, CYP2J2, which was differentially expressed in MP or BP, converts arachidonic acid to four regioisomeric epoxyeicosatrienoic acids. Although the exact biological role of epoxyeicosatrienoics is unclear, a recent report (25) has shown a strong presence of epoxyeicosatrienoics in human carcinoma, but not in normal tissues. This was accompanied by the activation of mitogen-activated protein kinases and phosphoinositide-3-kinase/Akt systems as well as the elevation of epithelial growth factor receptor phosphorylation, all of which suggest a role in promoting a neoplastic cellular phenotype.
Prostacyclin synthase (PTGIS) was differentially increased in MP or BP versus tumor. Inactivation of specific tumor suppressor genes by transcriptional silencing associated with hypermethylation of the promoter is common in cancer. Using reverse transcription-PCR, Frigola et al. (26) have shown that PTGIS was inactivated through hypermethylation of its promoter region. Prostacyclin seems to exhibit both antiproliferative effects (27) and chemopreventive properties (28). Down-regulation of the enzyme responsible for prostacyclin synthesis would contribute to loss of this important prostanoid compound at the tumor site.
Overexpression of prostaglandin D synthase in EOC has recently been reported (29), although we found prostaglandin D synthase to be differentially expressed in MP versus tumor. Expression of prostaglandin D synthase mRNA was found in tumor cells of all various types of EOC and relative staining intensity seemed to be selective for certain types of disease.
Thromboxane synthase metabolizes the cyclooxygenase product, prostaglandin H (2), into thromboxane A (2). Thromboxane synthase has been found to be weakly expressed or absent in normal differentiated or advanced prostate tumors, and markedly increased in tumors with perineural invasion (30, 31), and in adenocarcinoma and squamous cell carcinoma of the lung (32). The relative expression of thromboxane synthase in normal and tumor peritoneum from patients with cancer has not previously been reported but its overexpression in this disease might be a poor prognostic factor.
Prostaglandins typically require specific receptors to bring about their pharmacodynamic actions. It is interesting to note that prostaglandin E receptor gene, also differentially expressed in MP or BP versus tumor, is up-regulated by PGE2 (33). Although the consequence of increased gene expression in EOC is unclear, PGE2 stimulation of the prostaglandin E3 receptor in lung adenocarcinoma led to a direct increase in Src activity and is believed to be a contributing factor in tumor progression (34).
Peroxisome proliferator-activated receptor-γ is a ligand-activated transcription factor that, in addition to its well-established role in lipid and glucose metabolism, is known to control cell proliferation and differentiation in several tissues. Vignati et al. (35) recently showed that peroxisome proliferator-activated receptor-γ was expressed on epithelial ovarian tumor tissues but not in normal ovarian tissue. Our data shows higher differential expression in MP and BP versus tumor.
Because of their importance to the biology of cancer, certain eicosanoids including their enzymes and receptor targets might represent appropriate specific targets for therapy or diagnosis. New therapeutic strategies might include molecules that alter the ratio of different eicosanoids such as the ratio of PGE3 to PGE2 (36). Others might selectively block PLA2 enzymes, thereby limiting the release of arachidonic acid from cell surface membranes (37), or selectively block synthesis or inhibit production of PGE2, COX-1 and COX-2, 12-HETE, and 5-HETE. The studies reported above (19, 21) might also suggest a role for COX-1 inhibition in the control of vascular endothelial growth factor.
An inflammatory process is clearly a part of the tumor microenvironment of EOC (9, 10). There is increasing support for a linkage between the inflammatory process in EOC and eicosanoid and arachidonic pathways. Using a combination of electrospray mass tandem spectroscopy and cDNA microarray analysis, we have shown significant differences in the distribution of a number of these elements in the tumor and non–tumor-involved peritoneum and in the peritoneum of patients with benign disease. The emerging patterns could provide a basis for further studies aimed at establishing critical pathways and the strategies that could affect the prevention or treatment of EOC.