Endometrial cancer is the most common gynecologic malignancy, frequently arising in association with obesity and diabetes mellitus. To identify gene pathways contributing to endometrial cancer development, we studied the transcriptome of 20 endometrial cancers and 11 benign endometrial tissues using cDNA microarrays. Among the transcript changes identified in endometrial cancer were up-regulation of the nuclear hormone receptors peroxisome proliferator-activated receptors (PPAR) α and γ, whereas retinoid X receptor β was down-regulated. To clarify the contribution of PPARα to endometrial carcinogenesis, we did experiments on cultured endometrial carcinoma cells expressing this transcript. Treatment with fenofibrate, an activating ligand for PPARα, significantly reduced proliferation and increased cell death, suggesting that altered expression of nuclear hormone receptors involved with fatty acid metabolism leads to deregulated cellular proliferation and apoptosis. These results support further investigation of members of the PPAR/retinoid X receptor pathway as novel therapeutic targets in endometrial cancer.

Endometrial carcinoma is the most common gynecologic malignancy and comprises 97% of all uterine cancers (1). There is a peak incidence between ages 55 and 65 years, with <5% of endometrial cancers occurring below age 40 years (2). The majority are of an endometrioid histologic subtype and display an association with obesity and diabetes mellitus (2). There is a pressing need to better understand the molecular basis for this disease, as 25% of women present with extrauterine disease with 5-year survival rates of ∼31% and 10% for Federation Internationale des Gynaecologistes et Obstetristes stages 3 and 4 disease, respectively (2). An improved understanding of events at a molecular level is essential in the development of targeted therapy, with a view to improving survival and cure rates.

There are increasing efforts to gain a more global view of the multiple, interrelated molecular changes that occur during tumorigenesis (3–6). The gene microarray is a high-throughput technology able to interrogate multiple genetic changes within tissues and cells (7–9). Consequently, there has been a marked increase in the use of microarrays to interrogate cancers at the genomic level. In addition to screening for candidate genes, microarrays may provide molecular diagnoses, thus avoiding some of the weaknesses of conventional diagnostic techniques (4, 10).

Despite the increasing use of microarray technology in cancer research, there have been difficulties obtaining meaningful biological information. The cost of genome-wide, commercially available arrays may prohibit large experimental samples, and there are multiple sources of variation in experimental results complicating data analysis and interpretation (11). Large-scale gene expression analyses of endometrial cancer have mostly been confined to small sample sets and cell lines (12, 13) and have employed genome-wide, commercially available microarray systems (12). Previous microarray studies in endometrial cancer have highlighted differences in the abundance of individual genes between benign and malignant tissues (12, 13), although there has been little advance in the understanding of pathway-specific alterations that may contribute to endometrial tumorigenesis. Independent component analysis (ICA) is a sophisticated statistical method that aims to identify patterns of coregulated genes rather than individual transcript changes (14). We previously have applied high-density cDNA microarrays to determine gene transcript abundance in epithelial ovarian cancer (14).

Tumor Samples and RNA Preparation

Twenty frozen endometrial carcinoma tissues, three atypical complex hyperplasias, and eight postmenopausal benign endometrial control tissues (four atrophic and four benign polyps) were obtained from the Department of Obstetrics and Gynecology, Addenbrooke's Hospital NHS Trust (Cambridge, United Kingdom) with ethical approval. Samples were frozen immediately in liquid nitrogen after surgery and stored at −70°C. Total RNA was extracted from the frozen tissues using an acid-guanidium thiocyanate-phenol-chloroform method (15). Clinical information and histopathologic reports were obtained for all frozen tissues.

Hybridization to cDNA Microarrays

Tailored nylon microarrays comprising 1,056 cDNA clones and including 37 expressed sequence tags were used for transcriptome analysis. These were manufactured in the Departments of Obstetrics and Gynecology and Pathology, University of Cambridge (Cambridge, United Kingdom) and designed for the study of endometrial function and vascular biology (16) The entire gene list can be seen at http://www.obgyn.cam.ac.uk/genearray. Labeled cDNA was produced from total RNA preparations as described previously (16). ULTRArray hybridization buffer (Ambion, Austin, TX) was used, and hybridization was carried out in 15 × 4 cm roller bottles as described previously (16). Microarrays were subjected to high stringency washes in SSC and SDS, dried at 60°C, and exposed to low-energy storage phosphor screens (Molecular Dynamics, Sunnyvale, CA) for 64 hours. Arrays were scanned at high resolution using a Storm 860 PhosphorImager (Molecular Dynamics).

Data Analysis

Images were imported into Imagene 4.0 software (Biodiscovery Inc., Marina del Ray, CA) for calculation of signal intensity. Total signal for each spot was squared to correct for the square root transformation applied by the scanning software (17). Data were processed using a normalization algorithm method similar to that used in the Web-based SNOMAD program (18). Normalized data were analyzed by two complementary methods, the Web-based Cyber-T and ICA. Cyber-T uses a Bayesian version of the t test, which is suitable for the large-throughput analysis required for microarray data (19). A Bayesian P value < 0.05 was considered significant. ICA reduces the data to a set of components that explain the data. Analysis of those components that correlate closely with a factor(s) of interest allows extraction of gene expression changes contributing most strongly to that factor. The method of ICA used is an ensemble learning method that allows identification of components constituting the data while at the same time providing an estimate of the inherent “noise” (20). The ICA model also allows refinement or filtering of the data (20).

Quantitative, Real-time PCR (TaqMan)

Total RNA (2 μg) was incubated with oligo(dT) primers (1 μg) at 68°C and reverse transcribed at 42°C using Super Reverse Transcriptase (HT Biotechnology Ltd., Cambridge, United Kingdom). Total RNA samples used were those used in the microarray experiments described earlier.

cDNA (1 μL each) was amplified using PCR Master Mix (Applied Biosystems, Warrington, United Kingdom). All quantitative, real-time PCR experiments were done in the ABI PRISM 7700 Sequence Detector (Applied Biosystems) according to the manufacturer's instructions and were done in triplicate. The resultant data were averaged for each sample. No-template controls were included in each experiment. Specific oligonucleotide primers and probes were used. These were designed for each of five genes [cyclooxygenase-2 (COX-2), vascular endothelial growth factor-B (VEGF-B), PPARα, PPARγ, and retinoid X receptor β (RXRβ)] using Primer Express 1.5 software (Applied Biosystems). Sequences are given below:

  • (a) COX-2

    • 5′-TGATCCCCAGGGCTCAAA-3′ (forward primer),

    • 5′-ATCTGTCTTGAAAAACTGATGCGT-3′ (reverse primer),

    • 5′-6FAM-TGATGTTTGCATTCTTTGCCCAGCACT-TAMRA-3′ (probe);

  • (b) VEGF-B

    • 5′-AGCACCAAGTCCGGATG-3′ (forward primer),

    • 5′-GTCTGGCTTCACAGCACTG-3′ (reverse primer),

    • 5′-6FAM-AGATCCTCATGATCCGGTACCCGT-TAMRA-3′ (probe);

  • (c) PPARα

    • 5′-GACGTGCTTCCTGCTTCATAGA-3′ (forward primer),

    • 5′-CACCATCGCGACCAGATG-3′ (reverse primer),

    • 5′-6FAM-TGGAGCTCGGCGCACAACCA-TAMRA-3′ (probe);

  • (d) PPARγ

    • 5′-CAGAGCAAAGAGGTGGCCAT-3′ (forward primer),

    • 5′-GCTTTTGGCATACTCTGTGATCTC-3′ (reverse primer),

    • 5′-6FAM-CATCTTTCAGGGCTGCCAGTTTCGC-TAMRA-3′ (probe);

  • (e) RXRβ

    • 5′-CCATCCGCAAAGACCTTACATAC-3′ (forward primer),

    • 5′-GTTCCGCTGGCGCTTG-3′ (reverse primer),

    • 5-6FAM-TGCCGGGACAACAAAGACTGCACA-TAMRA-3′ (probe).

Results for gene abundance in each sample were normalized to abundance of an endogenous control gene. 18S rRNA was used as an endogenous control for all genes, with the exception of VEGF-B for which β-actin was used. Preliminary experiments to determine that each endogenous control sequence was amplified at the same rate as the target gene sequence identified that β-actin, but not 18S rRNA, was a suitable endogenous control for VEGF-B. Normalized log-transformed transcript levels were compared across samples using one-way ANOVA and unpaired Student's t tests. Mean target gene abundance for each tissue type was compared with the data obtained from microarray experiments.

Immunohistochemistry

Immunohistochemistry for PPARα protein was done on 6 μmol/L formalin-fixed, paraffin-embedded endometrial tissue sections identified from histopathologic archives. Six endometrioid cancers and four samples of benign endometrium were examined. Sections were dewaxed in xylene and rehydrated in graduated alcohols and antigens were retrieved with 0.1% trypsin for 15 minutes at 37°C. Nonspecific immunostaining was blocked with normal goat serum and sections were incubated with 4 μg/mL rabbit polyclonal anti-PPARα antibody (Santa Cruz Biotechnology, Santa Cruz, CA) overnight at 4°C. Rabbit IgG (4 μg/mL) provided a negative control. A biotinylated goat anti-rabbit secondary antibody (1:100 dilution) was used for 30 minutes at room temperature. Detection of the antibody reaction was carried out with the Vectastain Elite ABC reagent (Vector Laboratories, Burlingame, CA) combined with 3,3-diaminobenzidine color development (Sigma Chemical Co., MI). Endogenous peroxidases were blocked with 0.3% hydrogen peroxide in methanol. Results were analyzed using the two-tailed Fisher's exact probability test. Statistical significance was accepted as P < 0.05.

Cell Culture Assays

Ishikawa cells in the logarithmic growth phase were cultured in DMEM (Sigma Chemical) supplemented with glucose, l-glutamine, and FCS in 96-well plates. Cells were cultured for 24 hours at 37°C and 5% CO2 with varying doses of the PPARα agonist fenofibrate (Sigma-Aldrich Co. Ltd., Dorset, United Kingdom) dissolved in DMSO (Sigma-Aldrich). Control cells were treated with vehicle only. A minimum of five replicates was done at each dose. Total RNA was extracted from untreated Ishikawa cells, and quantitative, real-time PCR was done for PPARα (as above). In addition, cell proliferation in fenofibrate-treated and DMSO-treated cells was assessed by the uptake of 5-bromo-2′-deoxyuridine (BrdUrd) using the BrdUrd Labeling and Detection Kit III (Roche Diagnostic, East Sussex, United Kingdom) according to the manufacturer's instructions. Absorbance values for each well were measured on a microtiter plate reader (Anthos Labtech Instruments, Salzburg, Austria).

Apoptosis was measured using two different methods. Cells were grown in 96-well plates as above. At 24 hours, the medium was replaced with fresh complete culture medium and fenofibrate (10–100 μmol/L). Six replicates were done for each drug dose, and control cells were treated with vehicle (DMSO) only. Apoptosis was quantified using the APOPercentage Apoptosis Assay dye (Biocolor Ltd., Northern Ireland, United Kingdom) according to the manufacturer's instructions. Dye absorbance values were measured as above. In a second assay, the medium was exchanged after 24 hours for fresh medium containing 75 μmol/L fenofibrate or DMSO only. After further incubation at 37°C for 1 hour, cells were fixed and stained with Hoechst bisbenzamide HB 3258 (Calbiochem, La Jolla, CA) and viewed by fluorescent microscopy. Seven separate optical fields were counted for each well, and both total numbers of cells and number of apoptotic cells were counted. An apoptotic index was calculated using the following formula: apoptotic index = [(number of apoptotic cells) / (total cell count)] × 100.

Data were analyzed using the InStat 2.1 statistical program. For the BrdUrd and APOPercentage assays, differences between median values across multiple treatment groups were analyzed using the Kruskal-Wallis nonparametric ANOVA method. Comparison between median values from each treatment group and negative control group was made using the Mann-Whitney U test. Ratio data arising from the Hoechst nuclear staining assay were log transformed and analyzed by Student's t test. Statistical significance was accepted at P < 0.05 throughout.

Reporter Studies

Ishikawa cells cultured in six-well plates were transfected with 3 μg luciferase reporter plasmid using 4 μg LipofectAMINE 2000 (Invitrogen, Paisley, United Kingdom). The reporter plasmid comprised a triple repeat of the peroxisome proliferator response element in direct orientation with a basal TK-luc reporter construct (21). Cells were grown and treated, before confluence was reached, with varying doses of fenofibrate and the specific PPARα antagonist GW496471. At 24 hours, cells were harvested and luciferase activity was measured using the Ascent Fluoroscan luminometer. Normalization was done against total protein using the BCA protein assay (catalogue no. 23227, Pierce Chemical Co. Rockford, IL). A control plasmid, pRL-CMV (catalogue no. E2261, Promega, Southampton, United Kingdom), was initially used as a transfection control but was found to respond to PPAR agonists and subsequently noted to contain a putative peroxisome proliferator response element within the cytomegalovirus promoter region (data not shown). The reporter plasmid and PPARα antagonist were kind donations from Dr. E. Shoenmakers and Prof. K. Chatterjee (Cambridge Institute for Medical Research, Cambridge, United Kingdom).

Gene Microarrays Identify Differentially Expressed Transcripts in Endometrial Carcinoma

Transcript abundance, measured by mean signal intensity, was compared between benign and malignant tissues. The expression of 204 genes was altered in the cancers compared with the benign endometrium (P < 0.05) when the data were analyzed using Cyber-T. Of these 204 genes, 182 were up-regulated and 22 were down-regulated in the cancers compared with benign endometrium. The genes that were identified by the Cyber-T analysis were assigned to functional categories (Table 1). The “housekeeping genes” included on the arrays (e.g., actin and glyceraldehyde 3-phosphate dehydrogenase) were not significantly altered between the different tissue classes. Among the changes seen were significant alterations in angiogenesis-related genes. The VEGF-A splice variant (VEGF189) and angiopoietin-2 were up-regulated by 5.7- and 4.9-fold, respectively (P < 0.05). In addition, the VEGF-C and VEGF-D receptor tyrosine kinase, Fms-related tyrosine kinase 4 (VEGF receptor-3), was 5.4-fold more abundant in the endometrial cancers compared with benign postmenopausal endometrium. In contrast, the VEGF-related growth factor, VEGF-B and platelet-derived growth factor receptor β, were almost 2-fold down-regulated in the cancers (P < 0.001).

Table 1.

Selected examples of differentially expressed genes in endometrial carcinoma

Genbank Accession No.Gene NameP*Fold Change in Tumors
Transcription factors and nuclear receptors    
  U32849 N-myc (STAT) interactor (NMI0.006 8.2 
  L40904 PPARγ 0.039 6.1 
  U22431 Hypoxia-inducible factor 1α (HIF1α0.004 5.4 
  X06562 Growth hormone receptor (GHR0.030 5.1 
  Y07619 PPARα 0.026 4.3 
  V00568 v-myc myelocytomatosis viral oncogene homologue (MYC0.048 2.8 
  X07282 Retinoic acid receptor β (RARβ0.028 4.3 
  M84820 RXRβ 0.001 −1.4 
Cell cycle and growth regulators    
  X76132 Deleted in colorectal carcinoma (DCC0.018 6.2 
  M74091 Cyclin C 0.009 4.5 
  X05908 Annexin A1 (ANXA10.012 4.8 
  L27560 Insulin-like growth factor binding protein 5 (IGFBP50.025 2.2 
Angiogenic factors, receptors, and mediators    
  AB012911 VEGF189 0.013 5.7 
  AB009865 Angiopoetin-2 (ANGPT20.020 4.9 
  U43143 Fms-related tyrosine kinase 4 (FLT40.033 5.4 
  U48801 VEGF-B 1.48e−05 −1.9 
Cell adhesion, recognition, and cytoskeleton    
  Z13009 E-cadherin 0.014 4.2 
  J03209 Matrix metalloproteinase-3 (MMP-30.015 4.4 
  M14083 Plasminogen activator inhibitor type 1 (PAI-10.046 2.2 
  M13509 Matrix metalloproteinase 1 (interstitial collagenase; MMP-10.025 3.9 
Growth factors, cytokines, hormones, and receptors    
  J02958 HGF receptor (MET0.009 3.0 
  M31172 Transforming growth factor-α (TGFα0.045 4.8 
  X06374 Platelet-derived growth factor receptor α (PDGFα0.045 4.5 
  X00588 Epidermal growth factor receptor (c-erbB; EGFR0.026 5.1 
  X04434 Insulin-like growth factor receptor 1 (IGFR10.001 −1.9 
Intracellular signaling    
  X95282 Ras homologue gene family, member E (RhoE0.028 4.6 
  AB012911 Frizzled homologue 6 (Drosophila) (FZD60.011 3.4 
Enzymes/cellular metabolism    
  U04636 COX-2, prostaglandin-endoperoxide synthase 2 0.038 5.6 
  M14777 Glutathione S-transferase A1 (GSTA10.014 4.2 
  M61900 Prostaglandin D2 synthase (PTGDS2.45e−05 −2.1 
Genbank Accession No.Gene NameP*Fold Change in Tumors
Transcription factors and nuclear receptors    
  U32849 N-myc (STAT) interactor (NMI0.006 8.2 
  L40904 PPARγ 0.039 6.1 
  U22431 Hypoxia-inducible factor 1α (HIF1α0.004 5.4 
  X06562 Growth hormone receptor (GHR0.030 5.1 
  Y07619 PPARα 0.026 4.3 
  V00568 v-myc myelocytomatosis viral oncogene homologue (MYC0.048 2.8 
  X07282 Retinoic acid receptor β (RARβ0.028 4.3 
  M84820 RXRβ 0.001 −1.4 
Cell cycle and growth regulators    
  X76132 Deleted in colorectal carcinoma (DCC0.018 6.2 
  M74091 Cyclin C 0.009 4.5 
  X05908 Annexin A1 (ANXA10.012 4.8 
  L27560 Insulin-like growth factor binding protein 5 (IGFBP50.025 2.2 
Angiogenic factors, receptors, and mediators    
  AB012911 VEGF189 0.013 5.7 
  AB009865 Angiopoetin-2 (ANGPT20.020 4.9 
  U43143 Fms-related tyrosine kinase 4 (FLT40.033 5.4 
  U48801 VEGF-B 1.48e−05 −1.9 
Cell adhesion, recognition, and cytoskeleton    
  Z13009 E-cadherin 0.014 4.2 
  J03209 Matrix metalloproteinase-3 (MMP-30.015 4.4 
  M14083 Plasminogen activator inhibitor type 1 (PAI-10.046 2.2 
  M13509 Matrix metalloproteinase 1 (interstitial collagenase; MMP-10.025 3.9 
Growth factors, cytokines, hormones, and receptors    
  J02958 HGF receptor (MET0.009 3.0 
  M31172 Transforming growth factor-α (TGFα0.045 4.8 
  X06374 Platelet-derived growth factor receptor α (PDGFα0.045 4.5 
  X00588 Epidermal growth factor receptor (c-erbB; EGFR0.026 5.1 
  X04434 Insulin-like growth factor receptor 1 (IGFR10.001 −1.9 
Intracellular signaling    
  X95282 Ras homologue gene family, member E (RhoE0.028 4.6 
  AB012911 Frizzled homologue 6 (Drosophila) (FZD60.011 3.4 
Enzymes/cellular metabolism    
  U04636 COX-2, prostaglandin-endoperoxide synthase 2 0.038 5.6 
  M14777 Glutathione S-transferase A1 (GSTA10.014 4.2 
  M61900 Prostaglandin D2 synthase (PTGDS2.45e−05 −2.1 

NOTE: Selected examples of genes found to be differentially regulated in endometrial cancer. Microarray data from 20 endometrial cancers (17 endometrioid type and 3 papillary serous type) and 11 benign endometria were compared using Cyber-T. Fold change in the cancers compared with the benign tissues is shown with a minus sign indicating down-regulation. The P value for each transcript change is indicated. The genes are listed in potential functional categories with Genbank accession numbers.

*

Tumor versus normal as assessed by Cyber-T.

Fold changes <1.5 not included.

Only one independent component (the “cancer” component), resulting from ICA, showed statistical significance using the Mann-Whitney U test (P < 0.001) and was subjected to further analysis. Data for the 5% of genes contributing most highly to this component are represented in Fig. 1. Genes concerned with lipid metabolism were found to be significantly overrepresented in the ICA analysis (20). Of these, novel changes were identified in members of the nuclear hormone receptor superfamily. Genes encoding both PPARα (P < 0.05) and PPARγ (P < 0.05) were significantly up-regulated in the cancers compared with benign endometrium (4.3- and 6.1-fold, respectively). Conversely, the gene encoding a heterodimerization partner for these two nuclear receptors, RXRβ, was down-regulated 1.4-fold (P < 0.01).

Figure 1.

Molecular profiles of all 31 specimens. A, data filtered using ICA show separation of benign and malignant endometria. The transcripts are those identified by this method as demonstrating differential expression between benign and malignant endometria. Two-way hierarchical clustering is applied by experiments and genes, with subsequent ordering of the data to the clusters. B1-11, benign samples; M1-20, malignant samples. Data are signal increase (red) or decrease (green) from the control log mean. Signal changes >2 SDs show the brightest coloration (see color scale). B, microarray and quantitative PCR expression analysis of five genes differentially expressed between benign and malignant endometria.

Figure 1.

Molecular profiles of all 31 specimens. A, data filtered using ICA show separation of benign and malignant endometria. The transcripts are those identified by this method as demonstrating differential expression between benign and malignant endometria. Two-way hierarchical clustering is applied by experiments and genes, with subsequent ordering of the data to the clusters. B1-11, benign samples; M1-20, malignant samples. Data are signal increase (red) or decrease (green) from the control log mean. Signal changes >2 SDs show the brightest coloration (see color scale). B, microarray and quantitative PCR expression analysis of five genes differentially expressed between benign and malignant endometria.

Close modal

Five genes were chosen to confirm the microarray results using quantitative, real-time PCR. Four of these genes are components of the prostaglandin and fatty acid pathways (COX-2, PPARα, PPARγ, and RXRβ), whereas the fifth gene, VEGF-B, is an angiogenic growth factor. In addition to their biological interest, these genes were chosen based on their significant Bayesian P value (using Cyber-T) and high loading value within the ICA “cancer” component. The mean relative expression of these genes in cancers compared with benign tissues was calculated from normalized transcript abundance levels from real-time PCR and compared with the mean ratio of expression levels in cancer obtained from microarray experiments (Fig. 1B). Data from microarray experiments analyzed using Cyber-T indicated that transcripts for COX-2 (P < 0.01), PPARα (P < 0.01), and PPARγ (P < 0.05) were significantly up-regulated in the endometrial cancers examined compared with the benign endometrial samples (Table 1). Conversely, transcripts for VEGF-B (P < 0.01) and RXRβ (P < 0.01) were significantly down-regulated in the cancers compared with the benign tissues (Table 1). The up-regulation of COX-2, PPARα, and PPARγ and down-regulation of VEGF-B and RXRβ were confirmed using quantitative PCR (Fig. 1B). The direction and magnitude of fold changes was similar (Fig. 1B).

PPARα Protein Is Localized to Endometrial Carcinoma Cells but not to Stromal Cells

Paraffin-embedded tissue sections were used to confirm the expression of PPARα protein in endometrial cancers compared with benign (atrophic) endometrium. These specimens were unrelated to those used in the microarray experiments, thus providing independent confirmation. Specific immunostaining for PPARα protein was seen in five of six sections of endometrial cancer (Fig. 2) but in zero of four of the benign endometrial sections (P < 0.05). Negative controls did not show specific immunostaining (Fig. 2A). Immunostaining for PPARα protein was localized to the epithelial tumor cells and was cytoplasmic rather than nuclear (Fig. 2C). The intensity of staining was heterogeneous with some tumor cells demonstrating intense staining, whereas staining was weaker in others (Fig. 2B).

Figure 2.

PPARα protein is localized to the cytoplasm of endometrial carcinoma cells. Paraffin-embedded sections of benign endometrium and endometrial cancer were subjected to immunohistochemistry for PPARα. A biotinylated secondary antibody was used for detection. A, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrial carcinoma incubated with IgG (negative control) showing no specific immunostaining (magnification ×100). B, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrioid endometrial adenocarcinoma showing positive heterogeneous immunostaining for PPARα (magnification ×100). C, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrioid endometrial adenocarcinoma at high power showing cytoplasmic epithelial tumor cell staining but no staining of stromal cells (magnification ×400).

Figure 2.

PPARα protein is localized to the cytoplasm of endometrial carcinoma cells. Paraffin-embedded sections of benign endometrium and endometrial cancer were subjected to immunohistochemistry for PPARα. A biotinylated secondary antibody was used for detection. A, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrial carcinoma incubated with IgG (negative control) showing no specific immunostaining (magnification ×100). B, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrioid endometrial adenocarcinoma showing positive heterogeneous immunostaining for PPARα (magnification ×100). C, a Federation Internationale des Gynaecologistes et Obstetristes grade 2 endometrioid endometrial adenocarcinoma at high power showing cytoplasmic epithelial tumor cell staining but no staining of stromal cells (magnification ×400).

Close modal

PPARα Agonists Inhibit the Growth of Endometrial Cancer Cells

Ishikawa cells were shown to have levels of PPARα mRNA comparable with those seen in the cancer samples (Fig. 3). These were cultured with the PPARα agonist fenofibrate and showed a decrease in DNA synthesis in a dose-dependent manner after 24 hours of incubation (Fig. 4A). The difference in BrdUrd incorporation across the range of drug doses used was highly statistically significant (P < 0.001) when analyzed using the Kruskal-Wallis nonparametric ANOVA. Low doses of fenofibrate had no significant effect on DNA synthesis (Fig. 4A). However, a reduction in BrdUrd uptake (as measured by dye absorbance) was seen at doses of 50 μmol/L (P = 0.01, 95% confidence interval 0.20–0.26) and 75 μmol/L (P = 0.01, 95% confidence interval 0.20–0.26).

Figure 3.

Ishikawa cells express PPARα mRNA. Quantitative PCR for PPARα was done on RNA extracted from Ishikawa cells and compared with mean transcript abundance in benign (n = 8) and malignant (n = 17) endometria. Columns, mean PPARα transcript abundance relative to levels of 18S rRNA; bars, SE. All experiments were done in triplicate.

Figure 3.

Ishikawa cells express PPARα mRNA. Quantitative PCR for PPARα was done on RNA extracted from Ishikawa cells and compared with mean transcript abundance in benign (n = 8) and malignant (n = 17) endometria. Columns, mean PPARα transcript abundance relative to levels of 18S rRNA; bars, SE. All experiments were done in triplicate.

Close modal
Figure 4.

The PPARα agonist fenofibrate reduces proliferation and increases apoptosis in Ishikawa cells. A, PPARα-expressing Ishikawa cells in culture were treated with the PPARα agonist fenofibrate. After 24 hours, proliferation was assessed by BrdUrd uptake. Increasing doses of fenofibrate led to reduced proliferation compared with vehicle only. B, cultured Ishikawa cells were treated for 1 hour with fenofibrate, and uptake of dye into apoptotic cells was measured by dye absorbance. Apoptosis was increased in a dose-dependent manner by the addition of fenofibrate. C, in a reporter assay, cells were treated with DMSO or fenofibrate and 9-cis-retinoic acid (3 μmol/L) ± the PPARα antagonist GW496471 (3 μmol/L). Columns, luciferase activity (relative light units) relative to that seen in DMSO-treated cells in duplicate assays across two independent experiments; bars, SE.

Figure 4.

The PPARα agonist fenofibrate reduces proliferation and increases apoptosis in Ishikawa cells. A, PPARα-expressing Ishikawa cells in culture were treated with the PPARα agonist fenofibrate. After 24 hours, proliferation was assessed by BrdUrd uptake. Increasing doses of fenofibrate led to reduced proliferation compared with vehicle only. B, cultured Ishikawa cells were treated for 1 hour with fenofibrate, and uptake of dye into apoptotic cells was measured by dye absorbance. Apoptosis was increased in a dose-dependent manner by the addition of fenofibrate. C, in a reporter assay, cells were treated with DMSO or fenofibrate and 9-cis-retinoic acid (3 μmol/L) ± the PPARα antagonist GW496471 (3 μmol/L). Columns, luciferase activity (relative light units) relative to that seen in DMSO-treated cells in duplicate assays across two independent experiments; bars, SE.

Close modal

Cells were treated with fenofibrate for 1 hour at 37°C, and uptake of a labeling dye for fragmented DNA was measured in a microplate reader after lysing the cells. There was a significant and dose-dependent increase of cells undergoing apoptosis among those treated with fenofibrate compared with cells treated with vehicle (DMSO) alone (P < 0.0001; Fig. 4B). This effect was maximal at 75 μmol/L fenofibrate. In the second assay, cells were treated with either 75 μmol/L fenofibrate or vehicle only. Nuclear fragmentation and apoptotic bodies were frequently seen in the treated cells but infrequently seen in the control cells (Fig. 5A and B). However, mitotic figures were seen in control cells consistent with increased proliferation. The apoptotic index was ∼5 fold higher in treated compared with untreated cells (P < 0.05; Fig. 5C), confirming the results obtained from the BrdUrd proliferation assay.

Figure 5.

The PPARα agonist fenofibrate is proapoptotic for endometrial cancer cells. Cultured Ishikawa cells were treated with fenofibrate (75 μmol/L) or vehicle only for 1 hour, and apoptotic cells were counted after nuclear staining with Hoechst bisbenzamide. A, control cells show mitotic figures and few apoptotic cells. B, treated cells display nuclear fragmentation and apoptotic bodies (arrows). C, apoptosis is markedly increased in the presence of fenofibrate. Columns, mean of three replicates; bars, SD.

Figure 5.

The PPARα agonist fenofibrate is proapoptotic for endometrial cancer cells. Cultured Ishikawa cells were treated with fenofibrate (75 μmol/L) or vehicle only for 1 hour, and apoptotic cells were counted after nuclear staining with Hoechst bisbenzamide. A, control cells show mitotic figures and few apoptotic cells. B, treated cells display nuclear fragmentation and apoptotic bodies (arrows). C, apoptosis is markedly increased in the presence of fenofibrate. Columns, mean of three replicates; bars, SD.

Close modal

Using a peroxisome proliferator response element containing a luciferase reporter construct, we showed functional activity of endogenous PPARα in Ishikawa cells. Transfected cells treated with fenofibrate showed a significant increase in luciferase activity compared with untreated control cells (Fig. 4C). However, addition of the highly specific PPARα antagonist GW496471 (3 μmol/L) significantly reduced the increased luciferase activity induced by equimolar fenofibrate.

The use of ICA complemented Cyber-T analysis, enabling the identification of transcript patterns differentiating benign from malignant endometrium (20). Within these patterns, we showed up-regulation of two ligand-activated transcription factors belonging to the nuclear hormone receptor superfamily, PPARα and PPARγ. These receptors form heterodimeric complexes with RXRβ that bind to specific response elements within the promoter region of target genes (22). We have further shown that RXRβ is significantly down-regulated in endometrial cancer.

The PPARs exist as three isoforms, PPARα, PPARβ/δ, and PPARγ. Genes induced by PPARα are primarily concerned with β-oxidation in normal cells (22), whereas PPARγ regulates adipocyte differentiation and macrophage function and is linked with glucose homeostasis (23). PPARβ/δ shows widespread tissue expression and is particularly abundant during development (24), although its functions are not fully known. Endogenous ligands for PPARβ are prostaglandins and leukotrienes. In addition to these, ligands for PPARα include fatty acids (25). Ligands for the PPARs also regulate COX-2, which is critically involved in lipid metabolism, inflammation, and tumor promotion (26).

Our findings suggest involvement of the lipid metabolic and prostaglandin pathways in endometrial cancer development. The prostaglandin pathway has been implicated previously in carcinogenesis, with overexpression of COX-2 being described in several cancers including endometrium (27, 28), malignant mesothelioma (29), cervix (30), breast (31), and colon (32–34). In addition, there is epidemiologic evidence of a protective effect of nonsteroidal anti-inflammatory drugs against colon cancer (35). In the colon cancer model, it is suggested that prostanoids may promote tumor development via the PPAR receptors, although the data are not conclusive (36).

PPARα is highly expressed in the cancer cells of other malignancies [e.g., prostate (37) and bladder (38)]. Localization of PPARα protein to malignant endometrial cells suggests that up-regulation of the PPARα gene is also significant in endometrial carcinoma. Whereas PPARβ/δ is up-regulated in some endometrial cancers (27), neither PPARα nor PPARγ have been implicated previously in the development of this disease.

Although PPARα is up-regulated in endometrial carcinoma, proliferation was reduced by the addition of a PPARα-activating ligand. Conversely, apoptosis was markedly increased by a PPARα-activating ligand. We confirmed in reporter assays that addition of the same PPARα agonist was associated with an increase in PPARα activity. Therefore, increased apoptosis and reduction in DNA synthesis with fenofibrate administration are associated with an increase in PPARα activity. The downstream effectors that mediate these changes have not been identified. However, potential mechanisms include the generation of reactive oxygen species as a by-product of β-oxidation of fatty acids. PPARα regulates key enzymes within this metabolic pathway (25). We conclude therefore that aberrant function of the PPARα/RXR pathway occurs in endometrial carcinogenesis. Furthermore, we suggest that this pathway could potentially be targeted for therapeutic benefit. The underlying mechanism for the paradoxical effect of the PPARα agonist on cell growth is uncertain and warrants further investigation.

Endometrioid endometrial cancer is associated with obesity and metabolic abnormalities characterized by insulin resistance (2). Epidemiologic data suggest an association of dietary fat with increased endometrial cancer risk that may be independent of the association between increased fat intake and obesity (39). Earlier studies also revealed higher levels of sex hormone binding globulin in women eating a vegetarian diet (40), which could reduce the levels of free circulating estrogens. This raises the intriguing possibility of a link between fatty acid metabolism and the action of estrogens in endometrial cancer. There has been significant interest in the role of PPARs in metabolic disorders, although data on PPARs and lipid metabolism in endometrial cancer are scant. The data presented here suggest lipid metabolism and PPAR and RXR receptors as subjects for further study in endometrial cancer. PPARα activators are currently used for the treatment of hyperlipidemic disorders. The beneficial effects of a PPARα activator on the growth of endometrial cancer cells indicate that these drugs warrant further investigation as novel therapeutic options for the management of endometrial malignancy.

Grant support: Medical Research Council Clinical Training Fellowship G84/5733 (C.M. Holland), Medical Research Council Program grant G9623012 (S.K. Smith and D.S. Charnock-Jones), and Raymond and Beverly Sackler Award (C.M. Holland).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Rose PG. Endometrial carcinoma.
N Engl J Med
1996
;
335
:
640
–9.
2
Blake P, Lambert H, Crawford RAF. Molecular biology of gynecological cancer. In: Blake P, Lambert H, Crawford RAF, editors. Gynecological oncology. A guide to clinical management. Oxford: Oxford University Press; 1998. p. 137–52.
3
Golub TR, Slonim DK, Tamayo P, et al. Molecular classification of cancer: class discovery and class prediction by gene expression monitoring.
Science
1999
;
286
:
531
–7.
4
Khan J, Saal LH, Bittner ML, Chen Y, Trent JM, Meltzer PS. Expression profiling in cancer using cDNA microarrays.
Electrophoresis
1999
;
20
:
223
–9.
5
Dhanasekaran SM, Barrette TR, Ghosh D, et al. Delineation of prognostic biomarkers in prostate cancer.
Nature
2001
;
412
:
822
–6.
6
Beer DG, Kardia SL, Huang CC, et al. Gene-expression profiles predict survival of patients with lung adenocarcinoma.
Nat Med
2002
;
8
:
816
–24.
7
Duggan DJ, Bittner M, Chen Y, Meltzer P, Trent JM. Expression profiling using cDNA microarrays.
Nat Genet
1999
;
21
:
10
–4.
8
Brown PO, Botstein D. Exploring the new world of the genome with DNA microarrays.
Nat Genet
1999
;
21
:
33
–7.
9
Burgess JK. Gene expression studies using microarrays.
Clin Exp Pharmacol Physiol
2001
;
28
:
321
–8.
10
Bittner M, Meltzer P, Chen Y, et al. Molecular classification of cutaneous malignant melanoma by gene expression profiling.
Nature
2000
;
406
:
536
–40.
11
Nadon R, Shoemaker J. Statistical issues with microarrays: processing and analysis.
Trends Genet
2002
;
18
:
265
–71.
12
Smid-Koopman E, Blok LJ, Chadha-Ajwani S, Helmerhorst TJ, Brinkmann AO, Huikeshoven FJ. Gene expression profiles of human endometrial cancer samples using a cDNA-expression array technique: assessment of an analysis method.
Br J Cancer
2000
;
83
:
246
–51.
13
Matsushima-Nishiu M, Unoki M, Ono K, et al. Growth and gene expression profile analyses of endometrial cancer cells expressing exogenous PTEN.
Cancer Res
2001
;
61
:
3741
–9.
14
Martoglio AM, Tom BD, Starkey M, Corps AN, Charnock-Jones DS, Smith SK. Changes in tumorigenesis- and angiogenesis-related gene transcript abundance profiles in ovarian cancer detected by tailored high density cDNA arrays.
Mol Med
2000
;
6
:
750
–65.
15
Chomczynski P, Sacchi N. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction.
Anal Biochem
1987
;
162
:
156
–9.
16
Evans AL, Sharkey AS, Saidi SA, et al. Generation and use of a tailored gene array to investigate vascular biology.
Angiogenesis
2003
;
6
:
93
–104.
17
Ramdas L, Coombes KR, Baggerly K, et al. Sources of nonlinearity in cDNA microarray expression measurements.
Genome Biol
2001
;
2
:
0047.1.
18
Colantuoni C, Henry G, Zeger S, Pevsner J. Local mean normalization of microarray element signal intensities across an array surface: quality control and correction of spatially systematic artifacts.
Biotechniques
2002
;
32
:
1316
–20.
19
Baldi P, Long AD. A Bayesian framework for the analysis of microarray expression data: regularized t-test and statistical inferences of gene changes.
Bioinformatics
2001
;
17
:
509
–19.
20
Saidi SA, Holland CM, Kreil DP, et al. Independent component analysis of microarray data in the study of endometrial cancer. Oncogene. In press 2004.
21
Kliewer SA, Umesono K, Noonan DJ, Heyman RA, Evans RM. Convergence of 9-cis retinoic acid and peroxisome proliferator signaling pathways through heterodimer formation of their receptors.
Nature
1992
;
358
:
771
–4.
22
Berger J, Moller DE. The mechanisms of action of PPARs.
Annu Rev Med
2002
;
53
:
409
–35.
23
Escher P, Wahli W. Peroxisome proliferator-activated receptors: insight into multiple cellular functions.
Mutat Res
2000
;
448
:
121
–38.
24
Braissant O, Wahli W. Differential expression of peroxisome proliferator-activated receptor-α, -β, and -γ during rat embryonic development.
Endocrinology
1998
;
139
:
2748
–54.
25
Corton JC, Anderson SP, Stauber A. Central role of peroxisome proliferator-activated receptors in the actions of peroxisome proliferators.
Annu Rev Pharmacol Toxicol
2000
;
40
:
491
–518.
26
Ledwith BJ, Pauley CJ, Wagner LK, Rokos CL, Alberts DW, Manam S. Induction of cyclooxygenase-2 expression by peroxisome proliferators and non-tetradecanoylphorbol 12,13-myristate-type tumor promoters in immortalized mouse liver cells.
J Biol Chem
1997
;
272
:
3707
–14.
27
Tong BJ, Tan J, Tajeda L, et al. Heightened expression of cyclooxygenase-2 and peroxisome proliferator-activated receptor-δ in human endometrial adenocarcinoma.
Neoplasia
2000
;
2
:
483
–90.
28
Fujiwaki R, Iida K, Kanasaki H, Ozaki T, Hata K, Miyazaki K. Cyclooxygenase-2 expression in endometrial cancer: correlation with microvessel count and expression of vascular endothelial growth factor and thymidine phosphorylase.
Hum Pathol
2002
;
33
:
213
–9.
29
Edwards JG, Faux SP, Plummer SM, et al. Cyclooxygenase-2 expression is a novel prognostic factor in malignant mesothelioma.
Clin Cancer Res
2002
;
8
:
1857
–62.
30
Ferrandina G, Lauriola L, Distefano MG, et al. Increased cyclooxygenase-2 expression is associated with chemotherapy resistance and poor survival in cervical cancer patients.
J Clin Oncol
2002
;
20
:
973
–81.
31
Costa C, Soares R, Reis-Filho JS, Leitao D, Amendoeira I, Schmitt FC. Cyclo-oxygenase 2 expression is associated with angiogenesis and lymph node metastasis in human breast cancer.
J Clin Pathol
2002
;
55
:
429
–34.
32
Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, DuBois RN. Up-regulation of cyclooxygenase 2 gene expression in human colorectal adenomas and adenocarcinomas.
Gastroenterology
1994
;
107
:
1183
–8.
33
Sano H, Kawahito Y, Wilder RL, et al. Expression of cyclooxygenase-1 and -2 in human colorectal cancer.
Cancer Res
1995
;
55
:
3785
–9.
34
Kargman SL, O'Neill GP, Vickers PJ, Evans JF, Mancini JA, Jothy S. Expression of prostaglandin G/H synthase-1 and -2 protein in human colon cancer.
Cancer Res
1995
;
55
:
2556
–9.
35
Smalley WE, DuBois RN. Colorectal cancer and nonsteroidal anti-inflammatory drugs.
Adv Pharmacol
1997
;
39
:
1
–20.
36
Ota S, Bamba H, Kato A, Kawamoto C, Yoshida Y, Fujiwara K. Review article: COX-2, prostanoids and colon cancer.
Aliment Pharmacol Ther
2002
;
16 Suppl 2
:
102
–6.
37
Collett GP, Betts AM, Johnson MI, et al. Peroxisome proliferator-activated receptor α is an androgen-responsive gene in human prostate and is highly expressed in prostatic adenocarcinoma.
Clin Cancer Res
2000
;
6
:
3241
–8.
38
Fauconnet S, Lascombe I, Chabannes E, et al. Differential regulation of vascular endothelial growth factor expression by peroxisome proliferator-activated receptors in bladder cancer cells.
J Biol Chem
2002
;
277
:
23534
–43.
39
Littman AJ, Beresford SAA, White E. The association of dietary fat and plant foods with endometrial cancer (United States).
Cancer Causes & Control
2001
;
12
:
691
–702.
40
Armstrong BK, Brown JB, Clarke HT, et al.
J Natl Cancer Inst
1981
;
67
:
761
–7.