|Home | About | Journals | Submit | Contact Us | Français|
We used a functional complementation approach to identify tumor-suppressor genes and putative therapeutic targets for ovarian cancer. Microcell-mediated transfer of chromosome 18 in the ovarian cancer cell line TOV21G induced in vitro and in vivo neoplastic suppression. Gene expression microarray profiling in TOV21G+18 hybrids identified 14 candidate genes on chromosome 18 that were significantly overexpressed and therefore associated with neoplastic suppression. Further analysis of messenger RNA and protein expression for these genes in additional ovarian cancer cell lines indicated that EPB41L3 (erythrocyte membrane protein band 4.1-like 3, alternative names DAL-1 and 4.1B) was a candidate ovarian cancer-suppressor gene. Immunoblot analysis showed that EPB41L3 was activated in TOV21G+18 hybrids, expressed in normal ovarian epithelial cell lines, but was absent in 15 (78%) of 19 ovarian cancer cell lines. Using immunohistochemistry, 66% of 794 invasive ovarian tumors showed no EPB41L3 expression compared with only 24% of benign ovarian tumors and 0% of normal ovarian epithelial tissues. EPB41L3 was extensively methylated in ovarian cancer cell lines and primary ovarian tumors compared with normal tissues (P = .00004), suggesting this may be the mechanism of gene inactivation in ovarian cancers. Constitutive reexpression of EPB41L3 in a three-dimensional multicellular spheroid model of ovarian cancer caused significant growth suppression and induced apoptosis. Transmission and scanning electron microscopy demonstrated many similarities between EPB41L3-expressing cells and chromosome 18 donor-recipient hybrids, suggesting that EPB41L3 is the gene responsible for neoplastic suppression after chromosome 18 transfer. Finally, an inducible model of EPB41L3 expression in three-dimensional spheroids confirmed that reexpression of EPB41L3 induces extensive apoptotic cell death in ovarian cancers.
In 1978, Stanbridge and Wilkinson showed that the fusion between HeLa cervical carcinoma cells and normal diploid fibroblasts created somatic cell hybrids with a stable karyotype and a suppressed tumorigenic phenotype . The phenotypic changes were later attributed to the effects of normal copies of human chromosomes from the normal fibroblasts, functionally complementing the genetic background of the HeLa cells . These findings formed the basis of a methodological development, microcell-mediated chromosome transfer (MMCT), which enables the introduction of individual human chromosomes into cancer cells and, subsequently, the localization and identification of genes associated with a multitude of biologic mechanisms including cellular senescence, immortalization, and tumor suppression .
Since its development, MMCT has been used extensively as an approach to improve our understanding of tumor development. Investigators have found functional evidence for several loci scattered throughout the genome that induce neoplastic suppression in a variety of tumor types [4–8]. However, progress in taking these studies forward to the stage of identifying the genes responsible has been hampered by the limited resolution of themethodology. It has often required detailed genomic and functional mapping to reduce the chromosome of interest down to a few megabases, and this frequently proves to be a challenging and lengthy process. The advent of gene expression microarray technologies has enabled researchers to sidestep this bottleneck, and a few studies that combine MMCT analysis with gene expression profiling have now been published, leading to the identification of plausible functional candidate genes for different diseases [9–11].
Some studies have used MMCT to identify chromosomes and subchromosomal regions that cause neoplastic suppression in ovarian cancer cells [12–15]. However, these studies have yet to find definitive evidence of a role for any of the candidate genes subsequently identified in the development of ovarian cancers. Generally, the mechanisms that underlie ovarian tumor development remain poorly understood, and this continues to have a major impact on clinical intervention strategies for tackling the disease. Despite improvements in cytoreductive surgery and the initial good response of patients to platinum-based chemotherapies, there have been few improvements in the survival rates for patients diagnosed with ovarian cancer for more than three decades; approximately 65% of patients will die within 5 years of their diagnosis . There are many reasons for the poor survival rates. Significantly, most patients are diagnosed with advanced stage disease; but also, there are no reliable prognostic markers for predicting clinical response and guiding treatment and no novel molecular targets expressed in ovarian tumors that have led to the development of new therapies. To identify chromosome regions that are involved in the development of ovarian cancer, we recently reported using MMCT in ovarian cancer cell lines for seven chromosomes (4, 5, 6, 13, 14, 15, and 18) that we found to be frequently deleted in primary ovarian cancers [17,18]. We found that chromosome 18 functionally suppresses the neoplastic phenotype of ovarian cancer cell lines . Here we describe the identification and evaluation of 14 candidate genes located on chromosome 18 that are activated in MMCT+18/cancer cell line hybrids. For one of these candidate genes (erythrocyte membrane protein band 4.1-like 3, EPB41L3), further functional evaluation in primary ovarian tumors and in three-dimensional models of ovarian cancer provided additional evidence supporting a role of EPB41L3 in ovarian cancer development.
MMCT was performed in the TOV21G cancer cell line as previously described [18,19]. TOV21G is an epithelial ovarian cancer cell line derived from a clear cell ovarian carcinoma. Briefly, the recipient ovarian cancer cell line was fused with mouse (A9):human monochromosome donor cell lines carrying the selectable fusion gene marker hygromycin phosphotransferase. MMCT hybrids were selected from post fusion cells in medium supplemented with hygromycin B (Calbiochem, Merck Chemicals Ltd, Nottingham, UK). The in vitro phenotype of recipient-donor hybrids was evaluated using anchorage-dependent and -independent growth assays and invasion through a Matrigel . In vivo tumorigenicity was assayed after intraperitoneal injection of ~2.5 x 106 cells in immunosuppressed mice as previously described .
The chromosome 18 content of hybrids and recipient cell lines was evaluated using metaphase chromosome painting. Fluorescent whole chromosome paints (Q-Biogene, Cambridge, UK) were used to detect the transferred chromosome by in situ hybridization using standard protocols. A mouse pan-centromeric probe (Q-Biogene) was used to detect any mouse DNA transferred; none was present in any of the hybrids analyzed. Fifteen metaphase spreads were scored for each cell line. Microsatellite analysis of 18 markers spanning chromosome 18 (http://www.gdb.org/) was used to determine whether the transferred allele was present. Array comparative genomic hybridization analysis was performed using a whole genome tiling-path consisting of 32,450 BAC clones (http://www.instituteforwomenshealth.ucl.ac.uk/trl/microarray_facility.html). Fluorescently labeled DNA of chromosome hybrid cell lines was hybridized against the parental cell lines. DNA samples were labeled with Cy3 or Cy5 using the Bioprime Total Labeling Kit (Invitrogen, Paisley, UK). Slides were scanned using an Axon 4000B laser scanner (Genepix, Molecular Devices, Sunnyvale, CA) at a 5-µmresolution. Raw fluorescence data were extracted using “BlueFuse” software (BlueGnome, UK) and normalized using the MANOR and LIMMA packages . The log2 of fluorescence ratios (M) were plotted against genome location using BACclone locations derived from the National Center for Biotechnology Information Human Genome build 36 (HG18).
Total RNA isolated from cell lines was quality control tested using a Nano assay with an Agilent Bioanalyzer 2100 (Agilent Technologies, Inc, Santa Clara, CA). RNA was converted into digoxigenin-labeled complementary RNA and hybridized to a human genome microarray system (Human Genome Survey Microarray Version 2; Applied Biosystems, Foster City, CA), which contains 32,878 probes for the interrogation of 29,098 genes. Gene expression profiles were generated in triplicate for each cell line. Data analysis was performed using Applied Biosystem's 1700 ArrayExpress software, Spotfire DecisionSite for Functional Genomics software (Goteborg, Sweden), and R version 1.9.1. Probes that were deemed undetectable were excluded from the final analyses if they had a signal-to-noise ratio less than 3. An analysis of variance was used to generate P values for statistical differences between probes. P values were adjusted for multiple comparisons as described by Benjamini and Yekutieli . Genes were statistically different between groups if they had an adjusted P < .01 and an average fold change difference greater than 1.6. Gene ontology analysis was performed using the Panther classification system (http://www.pantherdb.org).
Total RNA was extracted and analyzed by real-time reverse transcription-polymerase chain reaction (PCR) using optimized TaqMan Gene Expression assays (Applied Biosystems). The efficacy of 18S as an endogenous control was examined using the 2-ΔCt method.
Immunoblot analysis was performed for cell lines using two primary monoclonal antibodies that recognize the EPB41L3 protein, anti-Dal-1 monoclonal antibody from Imgenex (San Diego, CA) and anti-Dal-1 monoclonal antibody from Abcam (Cambridge, UK). Cell lysates were separated by SDS-PAGE, transferred to Immobilon P membrane (Millipore, Watford Hertfordshire, UK), probed with the primary antibody and an HRP-conjugated secondary antibody. Immunoreactive bands were visualized with the enhanced chemiluminescence system (NEN Life Sciences, Boston, MA). For immunohistochemistry analysis, tissue sections were incubated with a rabbit anti-EPB41L3 polyclonal antibody (1:1500 dilution, kindly donated by Dr I. Newsham) using standard protocols . The malignant and nonmalignant tissues were scored for EPB41L3 by assessing the site of positive staining in the tissue. Semiquantitative scoring criteria were used for immunohistochemistry; both staining intensity and positive areas of staining were recorded. We scored EPB41L3 positive staining calculated from the proportion of immunopositive neoplastic cells in the specimen (0, negative [<5% of positive cells]; 1, weak [5%–20%]; 2, moderate [20%–50%]; 3, strong [>50%]).
Cell line analyses. Quantitative DNA methylation analysis was performed in ovarian cancer and normal ovarian epithelial cell lines using the MassARRAY EpiTYPER (Sequenom, San Diego, CA) after DNA bisulfite treatment and matrix-assisted laser desorption ionization time-of-flightmass spectrometry analysis. Mass spectra were acquired by using a MassARRAY Compact MALDI-TOF, and the spectra's methylation ratios were generated by the Epityper software v1.0 (Sequenom). The EPB41L3 promoter was analyzed in three PCR amplicons, designed using EpiDesigner software (Sequenom). In total, 154 CpG sites in EPB41L3 were analyzed.
Primary tissue analyses. A total of 45 invasive ovarian cancer tumor tissue samples (of various histologic subtypes) and 16 normal tissue samples (endometrial, peritoneal, and fallopian tube tissues) were analyzed for 20,000 probes using the Illumina Infinium (Illumina, San Diego, CA) “genome-wide” panel after extraction and bisulfite conversion of DNA. Tissue samples were collected from Duke University and the University of Southern California; the analysis had ethical committee approval. All data were analyzed using Genetrix/SB version 3.3 (Epicenter Software, Alhambra, CA). The mean methylation values between the tumor and normal samples for each of these probes were compared using a parametric t test.
For continuous expression, EPB41L3 complementary DNA (cDNA) was cloned into a pBMN expression vector and transfected into three epithelial ovarian cancer cell lines: TOV21G, SCOV3, and INTOV2 using FuGENE 6 Transfection Reagent (Roche, Welwyn, UK). Briefly, 106 cells were transfected with the gene expression vector or a control vector expressing the GFP reporter gene (Orbigen, San Diego, CA). Selection was performed using 500 µg/ml geneticin (Sigma-Aldrich, St Louis, MO), and cells expressing the plasmids were subcultured by ring cloning. Stably transfected cell lines expressing EPB41L3 were established as multicellular spheroids by culturing cells in 1% agarose-coated 24 multiwell plates. Spheroids were visualized using an inverted microscope to calculate spheroid size and volume. The projected area, A and perimeter P, for each spheroid were measured using IMAGE J software (http://rsbweb.nih.gov/ij/). Spherical volume and geometric mean diameter of each spheroid were calculated as previously described .
The Ecdysone muristerone-induced system (gift from Irene Newsham) was used to generate an inducible TOV21G-EPB41L3 expressing cell line. Transfection of the retinoid-X receptor containing vector (pVgRXR) was followed by the pIND- EPB41L3 vector, and the doubly transfected clones were selected with 35 µg/ml zeocin (Invitrogen) and 600 µg/ml geneticin (Sigma). Approximately 30 clones were screened by EPB41L3 specific fluorescence activated cell sorting (FACS) analysis for controlled inducible expression after 48 hours of 1 µM muristerone (Invitrogen) exposure. The clones generated from the inducible expression of EPB41L3 in ovarian cancer cells were screened for EPB41L3 protein expression by FACS analysis to identify a clone expressing high levels of EPB41L3 (>65% of cells expressed the protein) for use in downstream experiments. Multicellular spheroids were formed by culturing cells in poly-HEMA (Sigma)-coated vessels as previously described [25,26].
Apoptosis was measured using the Annexin-V-FLUOS staining kit (Roche) according to the manufacturers instructions. Briefly, 106 cells were washed with PBS and were stained with annexin Vand propidium iodide for 15 minutes at room temperature and analyzed using a flow cytometer (Becton and Dickinson, Oxford, UK). Live/dead assays were performed after induction of EPB41L3 expression using a commercially available live/dead viability and cytotoxicity assay (Molecular Probes, Invitrogen). Spheroids were stained with 2 mM calcein and 4 mM ethidium homodimer 1 for 30minutes at room temperature, and image capture was performed using a confocal microscope (Ultraview; Perkin Elmer, Cambridge, UK).
Spheroids were washed with phosphate-buffered saline and fixed with 2% paraformaldehyde, 1.5% glutaraldehyde in 0.1 M cacodylate buffer, pH 7.3, then washed in 0.1 M cacodylate buffer and postfixed with 1% osmium tetroxide in 0.1 M cacodylate buffer. Finally, spheroids were washed in 0.1 M cacodylate buffer before being stained with 0.5% uranyl acetate and dehydrated with ethanol. For transmission electron microscopy (TEM), spheroids were embedded in agar resin, sectioned, and examined on a JEOL 1010 TEM microscope ( Jeol Ltd, Tokyo, Japan). For scanning electron microscopy (SEM), samples were postfixed with 1% osmium tetroxide in 0.1 M cacodylate buffer pH 7.3 before dehydrating with ethanol, critical point drying, mounting on carbon stubs, and coating with gold before viewing under a JEOL 7401 series FEGSEM (Jeol Ltd).
We have previously shown that chromosome 18 induces neoplastic suppression of the epithelial ovarian cancer cell line TOV21G both in vitro and in vivo (Table W1) [18,27]. To identify candidate tumor-suppressor genes located on chromosome 18, we used gene expression microarrays to compare 29,098 different genes between the ovarian cancer cell TOV21G and two TOV21G+18 hybrids. The aim was to identify genes that showed an increase in expression in TOV21G+18 hybrids, which may suggest that they are functionally activated and responsible for the suppression phenotype. We found 14 chromosome 18 genes that were significantly overexpressed in both TOV21G+18 hybrids compared with the TOV21G cell line (Table 1). The locations of these genes with respect to mapping data for chromosome 18 in the hybrids are illustrated in Figure 1A. Semiquantitative RT-PCR analysis of 10 of these genes was used to validate the results of gene expression microarray analysis and showed the same trend in expression change for all 10 genes compared with the microarray expression data (data not shown).
For the 14 candidate genes, we reviewed the evidence of a role in cancer based on their known or putative function and any experimental data from previously published studies. On the basis of this analysis, we restricted the list of plausible candidates to 11 genes, which we evaluated further by analyzing gene and protein expression (if antibodies were available) in a panel of 19 ovarian cancer cell lines and 3 normal ovarian epithelial cell lines (Table 1). These analyses suggested that one candidate gene in particular, EPB41L3 (the erythrocyte membrane protein band 4.1-like 3 gene or alternative nomenclature DAL-1), at 18p11.32 is a candidate ovarian cancer-suppressor gene.
Expression microarray analysis indicated that EPB41L3 was more than 50-fold overexpressed in TOV21G+18 hybrids compared with its expression in TOV21G cells (Figure 1B). Microsatellelite analysis confirmed that this gene had been transferred into both TOV21G hybrids (data not shown). The high level of increased expression in the TOV21G+18 hybrids suggested that the gene had been activated as result of the chromosome transfer. Consistent with this, most genes flanking EPB41L3 showed only minor variation in the levels of differential gene expression between TOV21G and TOV21G+18 cells (Figure 1A).
Immunoblot analysis confirmed the absence of EPB41L3 protein in TOV21G cells and a strong expression in both of the TOV21G+18 hybrid cell lines (Figure 1C). Three normal ovarian surface epithelial cell lines also showed high levels of EPB41L3 expression.
EPB41L3 protein expression was further evaluated in ovarian cancer cell lines and tissues from primary invasive epithelial ovarian cancers, benign epithelial ovarian tumors, and normal ovarian epithelium by Western blot analysis and immunohistochemistry. EPB41L3 expression was absent in 15/19 ovarian cancer cell lines (79%; Figure 2A). Immunohistochemical staining of 794 invasive and 33 benign tumors distributed across tissue arrays from three different studies showed no identifiable EPB41L3 expression in 65% of all invasive tumors and minimal, partial or complete expression in 20%, 10%, and 5% of these tumors, respectively (P trend = .01). When tumors were stratified by histologic subtype, clear cell, serous, mucinous, and endometrioid subtypes showed the most frequent loss of EPB41L3 expression (69%, 66%, 61%, and 57%, respectively). In contrast, there was no evidence of loss of EPB41L3 expression in 55% of benign tumors and in all five normal ovarian epithelial tissue samples. These data are illustrated in Figure 2, B and C (see also Table W2).
Studies in other cancers suggest that methylation of the EPB41L3 promoter is a common mechanism by which the gene is downregulated during tumor development. Therefore, we examined the methylation status of EPB41L3 in 16 ovarian cancer cell lines and 8 normal ovarian epithelial cell lines using a semiquantitative approach that enabled us to examine the methylation status at all CpG sites for the gene (Supplementary Information 2). Most ovarian cancer cell lines, including TOV21G, showed extensive methylation at the EPB41L3 locus, whereas normal ovarian epithelial cells showed hypomethylation relative to cancer cells (Figure 2D). We also analyzed the methylation status of EPB41L3 and other genes in the flanking regions in 45 primary invasive ovarian tumors and 16 normal primary tissue samples (endometrium, peritoneum, and fallopian tube). This was part of a genome-wide methylation analysis of approximately 20,000 probes (data not shown). Of the 32 probes from the 7.9-Mb region surrounding EPB41L3, which contains 22 genes, only 2 probes showed a statistically significant difference in the mean methylation values between tumor and normal samples: RALBP1 (P = .016) and EPB41L3 (P = .00004). When all probes located on chromosome 18 (n = 304) were analyzed and a false discovery rate of 0.05 was applied (equivalent to P = .003), EPB41L3 was one of only 19 probes that showed a statistically significant difference in mean methylation values.
A full-length normal EPB41L3 cDNA, cloned into a pBMN expression vector, was transfected into three ovarian cancer cell lines: TOV21G (TOV21G+EPB41L3), SCOV3 (SCOV3+EPB41L3), and INTOV2 (INTOV2+EPB41L3). After confirming the reexpression of EPB41L3 in TOV21G, we evaluated the effects on anchorage-dependent and -independent growth and compared these phenotypes with those of the same cells transfected with an empty pBMN vector expressing GFP (TOV21G+GFP, SCOV3+GFP, and INTOV2+GFP). EPB41L3-expressing cells formed significantly fewer colonies than GFP-transfected cells and untransfected cells. The colonies that did form after EPB41L3 transfection were substantially more disparate and showed markedly decreased anchorage-dependent and -independent growth (Supplementary Information 3).
We established three-dimensional multicellular spheroids of TOV21G, INTOV2, and SCOV3 ovarian cancer cell lines, of the GFP- and EPB41L3-expressing forms of these cell lines and of TOV21G+18 hybrids (Supplementary Information 3). Macroscopically and microscopically, the spheroids that formed from TOV21G, INTOV2, and SCOV3 cell lines and from GFP-expressing cells shared similar phenotypic characteristics, with no significant differences in their size. In contrast, spheroids formed from TOV21G+18 hybrids and EPB41L3-transfected cancer cell lines were significantly smaller compared with parental cell lines, consistent with the postulated tumor-suppressive effects of EPB41L3 (P > .0001 for size and P > .0001 for volume; Figure 3, A and B). The structural and morphologic features of parental cell lines also changed substantially after EPB41L3 transfection, shown by SEM and TEM (Figure 3C). For example, TOV21G cells revealed structures consistent with their epithelial origin (secretion of an extracellular matrix; tight junctions, desmosomes, andmicrovilli; lumens forming within spheroids). However, spheroids established from TOV21G+18 and TOV21G+EPB41L3 cells showed less cell aggregation and membrane interactions (smooth contours), clear signs of increased apoptosis in surface cells, large cell surface protrusions (including filopodia and microspikes), and degenerating cellmembranes. FACS analysis for annexin V confirmed that the transfer of chromosome 18 or reexpression of EPB41L3 in TOV21G, INTOV2, and SCOV3 cells induces apoptosis (Figure 3D).
Finally, we evaluated the effects of inducing EPB41L3 expression after the formation of spheroids using an Ecdysone muristerone-inducible system. We generated an inducible TOV21G+EPB41L3/ECD cell line and used FACS analysis to confirm EPB41L3 expression after muristerone induction (data not shown). A fluorescent live/dead assay was used to evaluate apoptosis in spheroids. Ten days after EPB41L3 induction, the ratio of “live” to “dead” (apoptotic) spheroids was significantly lower than it was for the same cells without EPB41L3 induction (1:3 vs 6:1, P < .002; Figure 3E).
Using a functional complementation approach, we have identified EPB41L3 as a candidate tumor-suppressor gene for ovarian cancer. The gene is extensively methylated in ovarian cancer cell lines and primary tumor tissues compared with normal epithelial ovarian cells and primary normal tissues, EPB41L3 protein expression is completely lost in 65% of all primary invasive ovarian tumors and in 79% of ovarian cancer cells lines, and reexpression of the gene in three-dimensional ovarian cancer cell line models causes growth suppression and increased apoptosis. To our knowledge, this is the first report linking EPB41L3 to ovarian carcinogenesis.
In the past, MMCT has been successful at identifying chromosome and subchromosomal regions that are associated with a disease phenotype or functional mechanism but rarely has definitive evidence for a specific gene responsible for the observed phenotype been shown. Largely, this is because of the scale of the task in fine mapping the chromosome region of interest and then evaluating several putative candidate genes. The task is made greater still because there are several mechanisms by which the function of a gene can be abrogated (e.g., coding sequence mutation, promotor methylation, gene deletion), and evaluating each possible mechanism is a major challenge. In the current study, the approach we used identified a relatively small series of candidate genes that were associated with neoplastic suppression, caused by chromosome 18 transfer into an ovarian cancer cell line. Additional analysis based on known function and the molecular analysis of several ovarian cancer cell lines suggested that EPB41L3 was the strongest candidate; but it remains a possibility that one or more of the other identified genes, or an as yet unidentified gene, is the real cause of neoplastic suppression. Following up on one or several of the other candidate genes we identified will require additional and extensive functional analyses to establish if they have a role in ovarian carcinogenesis.
Some studies have successfully identified the gene responsible for the phenotype after MMCT, which shows the power of this methodological approach. For example, two recent MMCTstudies have found two genes, cblD on chromosome 2q23.2 and cblF on chromosome 6q13 that are involved in vitamin B12 metabolism [28,29]. In our study, the evidence that EPB41L3 is the chromosome 18 gene that causes neoplastic suppression in TOV21G is compelling. The phenotypic features of EPB41L3-expressing cells and TOV21G+18 hybrids are remarkably similar in the macroscopic and microscopic appearance of multicellular spheroids and in their growth characteristics and apoptotic phenotype. The analysis of EPB41L3 in cancer cell lines and primary tumors provides strong evidence that EPB41L3 is an ovarian cancer-suppressor gene. Immunohistochemistry analysis of more than 800 ovarian tumors found that EPB41L3 is downregulated in approximately two-thirds of all major histologic subtypes of ovarian cancer but is strongly expressed in normal ovarian epithelium. It is uncommon to find a molecular marker that shows a similar pattern of expression in all the major ovarian cancer subtypes. A recent study by Kobel et al.  analyzed 21 candidate tissue biomarkers in 500 invasive ovarian tumors and found extensive variation in the patterns of expression between high-grade serous, clear cell, endometrioid, and mucinous ovarian carcinomas for almost all markers tested. However, immunohistochemistry alone does not provide sufficient evidence that a gene is essentially involved in tumor development. It is possible that EPB41L3 is merely a marker for a functionally relevant biologic mechanism or pathway involved in ovarian cancer progression. Therefore, it was important to identify the mechanism by which EPB41L3 was downregulated. Previous studies have shown that EPB41L3 is hypermethylated in non-small cell lung cancers, hepatocellular carcinomas, meningiomas, renal clear cell carcinomas, and prostate carcinomas [31–34]. Another study suggests that functionally significant somatic mutations of EPB41L3 are probably rare . We tested and found evidence for extensive EPB41L3 promoter hypermethylation in ovarian cancer cell lines and primary ovarian tumors compared with normal cells, suggesting that this is the likely cause of EPB41L3 down-regulation in ovarian tumors.
EPB41L3 expression was more prominent in benign ovarian tumors, and all normal ovarian epithelial cells analyzed showed strong expression. This possibly indicates that loss of EPB41L3 expression is a late rather than an initiating event in ovarian cancer progression. This is supported by in vivo data, which show that mice deficient for EPB41L3 develop normally and are fertile. Rates of cellular proliferation and apoptosis in brain, mammary, and lung tissues from the EPB41L3 null mice were similar to those in wild-type mice, and there was no evidence that null mice were susceptible to tumor development . However, when EPB41L3-null mice are crossed with TRAMP mice, which express SV40 in the prostate, mice have a much greater propensity to develop aggressive, spontaneous prostate carcinomas compared with TRAMP mice alone . The known functions of EPB41L3 are also consistent with a role for the gene in the later stages of tumor development. EPB41L3 is a member of the band 4.1 family of cytoskeletal proteins, which includes ezrin, radixin, moesin, and merlin. These proteins participate in organizing the actin cytoskeleton and ensure stable cell-cell adhesion. It is well known that interrupting the mechanisms that regulate cell adhesion leads to increased growth, invasion, and metastasis in tumor cells; these features are characteristic of the phenotypic changes we observed in the current study after transferring chromosome 18 and, subsequently, EPB41L3 into ovarian cancer cells. Two other studies have shown that reexpressing EPB41L3 in the MCF7 breast cancer cell line both increases cell adhesion and induces apoptosis [38,39].
We used an in vitro three-dimensional modeling approach to test the functional effects of reexpressing EPB41L3 in ovarian cancer cell lines because we have previously shown that three-dimensional culture models are much more reflective of the in vivo phenotype than traditional two-dimensional monolayer cultures [25,26]. To our knowledge, this is the first time that such an approach has been used to study the functional effects of expressing individual genes in cancer cells. The data provide support for the suppressive effects of EPB41L3 reexpression in ovarian cancer cells and show that this expression induces apoptosis and restricts cell adhesion; spheroids generated from three EPB41L3-expressing ovarian cancer cell lines were visibly less compact than spheroids formed from the cancer cells without EPB41L3. These studies also suggest that this three-dimensional system could be suitable for evaluating novel therapeutic targets for ovarian cancer and that EPB41L3 may be one such target. An alternative (or combinational) therapeutic approach to the current broad-based regimens of platinum chemotherapies (carboplatin and paclitaxal) might be to use gene replacement therapy coupling tissue- and tumor-specific gene delivery to provide high levels of gene expression- and tumor-targeted cell death. Recent studies have used constructs that tightly control adenoviral- and lentiviral-driven transcription , and there is now promising evidence that this approach could be successful in treating tumors. For example, Ueda et al.  have shown that the expression of the nitrogen-permease-like 2 (NPRL2) tumor-suppressor gene is restored in vivo in an orthotopic human lung cancer model using the US Food and Drug Administration-approved plasmid vector backbone for human clinical application, and a phase 1 clinical trial has shown the effectiveness of targeting the FUS1 gene in stage IV lung cancer patients .
In conclusion, we have shown that a functional complementation approach in ovarian cancer cell lines has identified EPB41L3 as a candidate tumor-suppressor gene for invasive ovarian cancer. We were able to demonstrate that the abrogation of EPB41L3 in a large series of epithelial ovarian cancers was linked to promoter hypermethylation. We showed functional similarities between ovarian cancer cells in which chromosome 18 was transferred and EPB41L3-expressing cancer cells, suggesting that EPB41L3 is the target on chromosome 18 that is responsible for the neoplastic suppression we observed in TOV21G cancer cells. Finally, we show that EPB41L3 reexpression alone causes growth suppression and induces apoptotic cell death in three-dimensional models of ovarian cancer.
Base position of CpG sites on target sequence for EPB41L3.
Amplicon 1: 0–342 bp + strand
Chromosome 18, strand start: 5533584 - strand end: 5533925, span: 342 bp
Amplicon 2: 0–418 bp + strand
Chromosome 18, strand start: 5533901 - strand end: 5534318, span: 418 bp
Amplicon 3: 0–479 bp + strand
Chromosome 18, strand start: 5533590 - strand end: 5534068, span: 479 bp
The authors thank Maria Notaridou for the NOSE cell line DNA, Mark Turmane for assistance with the electron microscopy, Irene Newsham for providing the Ecdysone system and the EPB41L3 rabbit antibody. The authors also thank Ian Spendlove for use of ovarian cancer tumor tissue arrays and Claire Templeman, Michael Press, Sahar Hooshdaran, Dan Weisenberger, and Mihaela Campan for their contributions to the primary tumors methylation study.
1This work has been funded by the Eve Appeal Gynaecology Cancer Research Fund. This work was undertaken at the University College London Hospital/University College London that received a proportion of funding from the Department of Health's National Institute for Health Research Central Biomedical Research Centers funding scheme and from the Ovarian Cancer Research Fund (grant no. PPD/USC.06).