Variant HNF1 Modulates Epithelial Plasticity of Normal and Transformed Ovary Cells

Abstract

Ovarian carcinoma arises from the ovarian surface epithelium, which undergoes phenotypic changes characteristic of müllerian epithelium during the first stages of tumorigenesis. The variant isoform of the hepatocyte nuclear factor 1 (vHNF1) is a transcription factor involved in the development of tissues derived from the müllerian duct. Here, we show that vHNF1 knockdown in two ovarian carcinoma cell lines, SKOV3 and IGROV1, leads to reduced E-cadherin (E-cadh) expression and decreased proliferation rate. Accordingly, SKOV3 cells ectopically expressing a dominant-negative (DN) vHNF1 mutant undergo an epithelial-mesenchymal-like transition, acquiring a spindle-like morphology, loss of E-cadh, and disrupted cell-cell contacts. Gene expression profiling of DNvHNF1 cells on the basis of a newly compiled list of epithelial-mesenchymal transition-related genes revealed a correlation between vHNF1 loss-of-function and acquisition of the mesenchymal phenotype. Indeed, phenotypic changes were associated with increased Slug transcription and functionality. Accordingly, vHNF1-transfected immortalized ovarian surface epithelial cells showed down-regulation of Snail and Slug transcripts. In DNvHNF1-transfected SKOV3 cells, growth rate decreased, and in vHNF1-transfected immortalized ovarian surface epithelial cells, growth rate increased. By immunohistochemistry, we found a strong association of vHNF1 with E-cadh in clear cell and in a subset of serous carcinomas, data that could potentially contribute in distinguishing different types of ovarian tumors. Our results may help in understanding the biology of ovarian carcinoma, identifying early detection markers, and opening potential avenues for therapeutic intervention.

Introduction

The pathophysiology of epithelial ovarian cancers (EOCs) remains poorly defined. One widely supported hypothesis is that they are derived from inclusion cysts. These cysts originate from the ovarian surface epithelium (OSE), which is the monolayer of cells covering the ovaries [1,2]. Ovarian surface epithelium cells appear as a simple epithelium with some characteristics typical of mesenchymal cells. Ovarian surface epithelium cells remain plastic in short-term culture, expressing vimentin together with cytokeratins 8 and 18. Conversely, invaginations and inclusion cysts have properties characteristic of müllerian epithelium, including expression of the specific epithelial marker E-cadherin (cadh) at the cell-cell junctions. After transformation, EOC cells can coexpress E-cadh and the mesenchymal marker vimentin as well as epithelial cytokeratins [3]. Unlike the tumor suppressor function of E-cadh in breast, prostate, and colon carcinomas [4,5], expression of E-cadh in ovarian epithelium seems to be associated with the development of EOCs [6].Nonetheless, themechanismof E-cadh-associated malignant OSE transformation is controversial [7,8]. In some advanced-stage EOCs, the so-called mesenchymal-epithelial transition (MET), which occurs during the first stages of transformation, is followed by an epithelial-mesenchymal transition (EMT) with loss of E-cadh expression [9].

Epithelial-mesenchymal transition is required for morphogenesis during embryonic development but has also been implicated in the acquisition of invasiveness by end-stage tumors [10–12]. This conversion results in loss of expression of adhesion molecules, such as E-cadh, ZO-1, and occludin, with consequent loss of cell-cell contacts and extensive remodeling of the cytoskeleton. Loss of E-cadh during development and cancer progression in tumors, other than EOCs, is mainly caused by transcriptional repression resulting from interaction of regulators with specific E-boxes in the proximal promoter of Cdh1, the gene encoding E-cadh [13]. Most prominent in this respect are the Snail-related zinc-finger transcription factors Snail and Slug.

The variant isoform of the transcription factor HNF-1 (vHNF1) activates transcription on homodimerization or heterodimerization with its companion protein HNF1α [14]. A role for HNF1 proteins in tumors has not yet been defined. For HNF1α, a biallelic inactivation of the relevant gene has been found in 50% of human liver adenomas [15], and somatic mutations were observed in 11% of endometrial carcinomas but not in breast and ovarian carcinomas [16]. Regarding vHNF1, the complete inactivation by germ line mutation of TCF2, the gene encoding for vHNF1, seemed to be associated to renal cell carcinoma [17] hypothesizing a tumor suppressor function. More recently, two variants within TCF2 have been found to be associated to prostate cancer risk [18]. vHNF1 is involved in the development of tissues organized in tubules, such as the pancreatic exocrine ducts and the kidney tubules [19,20], and in müllerian duct-derived tissues [21]. The transcription of the FR gene, which encodes the folate receptor (FR) α, is strongly activated in EOCs. We recently showed that the FR gene is regulated by vHNF1 [22], which is expressed in ovarian tumor specimens but not in OSE cells or in specimens obtained from tumors of other oncotypes.

Here, we addressed the potential role of vHNF1 in the MET-like taking place during ovarian cell transformation. We used in vitro approaches to negatively or positively affect vHNF1 expression and/or functionality in ovarian normal and transformed cells. We found that vHNF1 expression and functionality are directly correlated with epithelial differentiation, positively associated with growth potential, and inversely correlated with expression and functionality of E-box-binding transcriptional repressors. Immunohistochemical analysis of normal and transformed ovarian tissues showed that vHNF1 is not expressed in OSE cells but is expressed in 33% of E-cadh-expressing EOCs independently of tumor grading. The overall results demonstrate that vHNF1 is a new player in the epithelial differentiation of a subset of normal and transformed ovary cells.

Materials and Methods

Cell Culture

The ovarian carcinoma cell lines IGROV1 and SKOV3 (American Type Culture Collection, Manassas,VA) were maintained in RPMI 1640 medium (Sigma, St. Louis, MO) supplemented with 10% FCS (Sigma) and 2 mM l-glutamine. hTERT-IOSE (hereafter designated IOSE), obtained as described [23], were maintained in 199-MCDB105 medium (Sigma) supplemented with 15% FCS, 2 mmol l-glutamine, 200 µg/ml G418, and 50 µg/ml hygromycin.

Reagents and Antibodies

Triton X-100 (TX-100) and MES were from Sigma-Aldrich Fine Chemicals (St. Louis, MO); geneticin sulfate (G418) was from Gibco BRL (Paisley, Scotland). The following primary antibodies (Abs) were used at the dilution recommended by the manufacturer: anti-vHNF1 (goat), anti-HNF1 (rabbit), anti-ZO-1, and anti.occludin 1 (Santa Cruz Biotechnology, Santa Cruz, CA); anti-E-cadh mAb (mouse; Transduction Laboratories, BD Biosciences Pharmingen, Palo Alto, CA); anti-S100A4 (rabbit; DakoCytomation, Glostrup, Denmark). Horseradish peroxidase-labeled secondary Abs were from Amersham Bioscience-GE Healthcare (Piscataway, NJ). Secondary fluorochrome-conjugated Alexa Fluor 488 (green) was from Molecular Probes (Eugene, OR).

Small interfering RNA Treatment

IGROV1 and SKOV3 cells (5 x 10⁵) were seeded in 24-well plates and transfected 24 hours later with 80 pmol/ml of small interfering RNA (siRNA) duplex against vHNF1 mRNA (SmartPool; Dharmacon, Lafayette, CO) or Luciferase siRNA as control (Quiagen-Xeragon, Germantown, MD). siRNA transfection was performed by using Lipofectamine 2000 (Invitrogen, Paisley, UK) according to the manufacturer's protocol. Cells were harvested 48 hours later and analyzed for RNA and protein expression by quantitative reverse transcription-polymerase chain reaction (RT-PCR) and Western blot analysis, respectively.

Construction of DNvHNF1 and vHNF1 Expression Vectors

vHNF1 cDNA was obtained from the vector RSV-LFB3 (kindly provided by C. Toniatti, IRBM,Merck Research Laboratories, Pomezia, Italy). Dominant-negative vHNF1 (DNvHNF1; nt 1-729 of the open reading frame) was obtained by standard PCR with sense and antisense primers containing HindIII and XbaI restriction sites, respectively (sense, 5′-AGGAGGTCTAGAATGGTGTCCAAGCTCACG-3′; antisense, 5′-AAGGGAAGCTTTCACCAGGCTTGTAGAGG-3′). The purified fragment was inserted into the HindIII and XbaI sites of the expression vector pcDNAIneo (Invitrogen). For the expression vector encoding vHNF1, the vHNF1 open reading frame was inserted into the HindIII and XbaI restriction sites of the pcDNA3.1/Hygro vector (Invitrogen). Before transfection, both vHNF1-pcDNA3.1/Hygro and DNvHNF1-pcDNAIneo were verified by sequencing.

Quantitative Real-time RT-PCR

Total RNA was isolated with the RNeasy Total RNA kit (Quiagen, Hilden, Germany) according to the manufacturer's instructions. One microgram of total RNA was reverse-transcribed using the ABI High Capacity cDNA Archiving Kit (Applied Biosystem, Foster City, CA). Three replicates were run for each gene in each sample in a 96-well format plate. The probes and primer sets were the following Assays on Demand: Ref Hs00170423_m1 for Cdh1, Hs00195591_m1 for Snail, Hs00161904_m1 for Slug, HS00170182_mi for PLAU, and Hs00277509_m1 for FN (Applied Biosystems). GADPH mRNA levels were used as a control for the RNA extraction and RT experiments. Data were analyzed with the Sequence Detector v1.9 software. Relative gene expression for each sample was determined using the formula 2^(-ΔCt) reflecting target gene expression normalized to GAPDH levels.

Cell Solubilization, Fractionation, and Western Blot Analysis

For preparation of total cell lysates, cells were washed with ice-cold PBS and lysed in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2.3% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.005% bromophenol blue). Proteins were separated on precast 4% to 12% SDS-PAGE (Invitrogen) and transferred onto nitrocellulose membranes (Amersham Bioscience-GE Healthcare) as described [24]. Visualization was by the enhanced chemiluminescence method (Amersham Biosciences) using a Chemidocxrs and the Quantity One software (Bio-Rad, Hercules, CA). For cellular fractionation, confluent cells were treated as described [24]. Protein concentration of the fractions was determined by the BCA protein reagent assay (Pierce, Rockford, IL).

Cell Transfection

IGROV1 and SKOV3 cells were transfected with the DNvHNF1 construct essentially as described [25] using Lipofectamine 2000 according to the manufacturer's suggestions (Invitrogen). Forty-eight hours after transfection, fresh medium containing 400 µg/ml G418 (Gibco BRL) was added to the cell culture. DNvHNF1-positive clones were identified by RT-PCR using oligonucleotides that amplify only DNvHNF1 but not wild type (wt) vHNF1 (data not shown). Stable clones were tested by Western blot analysis on total cell lysates using rabbit anti-HNF1 Ab, which recognized both wt and DNvHNF1 proteins (Santa Cruz Biotechnology).

Immunofluorescence

Immunofluorescence was performed essentially as described [24], 2 x 10⁴ cells seeded on glass coverslips were grown for 48 hours, washed with cold PBS, and fixed with cold methanol for 10 minutes before immunoreaction. Samples were mounted with Mowiol solution and examined with an Eclipse TE2000-S microscope with a 40x PanFluor objective (Nikon, Melville, NY). Images were acquired with ACT-1 software (Nikon) at a resolution of 2250 x 1800 pixels. All procedures were carried out at room temperature.

Electrophoretic Mobility Shift Assay

Preparation of nuclear extracts (NEs) and electrophoretic mobility shift assay were carried out essentially as described [22].

Microarray Analysis

Gene expression in DNvHNF1 and mock transfectants was compared in three different RNA preparations pooled for each cell line. Total RNA from transfected SKOV3 cultures was extracted, further purified on RNeasy Mini Kit (Qiagen, Valencia, CA) according to the manufacturer's protocols, and treated with DNase (RNase-free DNase Set; Qiagen). Targets were obtained by synthesizing cDNA from 15 µg of total RNA. To verify the reproducibility of the observations, we performed two separate retrotranscriptions from both cell types to obtain four separate targets for hybridization. Genome set Human U133 Plus 2.0 chips (Affymetrix, Santa Clara, CA) were used in duplicate. Data from the two biologic replicates for each target (DNvHNF1 and mock cells) were tested in duplicate chips (a total of four samples for target) after normalization. The statistical analyses of the microarray data were performed with GenePicker software designed by the IFOM Institute (Milan, Italy). This software allowed to set up analysis schemes and to search the data for regulated genes using t test and Change-Fold Change analysis.We performed the reported analysis selecting the probe sets with significant statistical analysis (P < .05 for t test) and a fold change >1.5 or <-1.5, obtaining a list of 621 probe sets.

Compilation of Gene Lists Associated with the Epithelial or Mesenchymal Phenotype

A list of EMT-related genes was compiled after a literature search for genes modulated during processes activated by EMT [10,12] and taking into consideration two studies of gene expression profiling: one on Ha-ras-transformed polarized mammary epithelial cell line EpH4 induced to EMT by TGFβ treatment [26] and another on Madison-Darby canine kidney epithelial cells expressing the E-box-binding repressors Snail, Slug, and E47 [27]. Table W1 shows a list of genes associated with epithelial (140) and mesenchymal (186) phenotypes and passing criteria as that reported above. The categorization reported in both Tables 1 and W1 was done according to the Gene Ontology categories.

Table 1.

Genes Associated with Epithelial and Mesenchymal Phenotypes^* and Found to Be Differentially Expressed in DNvHNF1 Versus Mock Cells.

Gene Symbol	GenBank ID	Gene Name	Fold Change	P
Epithelial genes (n = 134)†
Actin cytoskeleton organization (n = 6)
PODXL	NM_005397	podocalyxin-like	1.94	.0002
Cell adhesion/ECM-related (n = 22)
ANXA4	BC000182	annexin A4	-1.84	.0000
CD99	U82164	CD99 antigen	1.66	.0000
CDH1	NM_004360	cadherin 1, type 1, E-cadherin	-3.33	.0404
CLDN1	AF101051	claudin 1	-2.07	.0002
CLDN7	NM_020412	claudin 7	-7.26	.0000
EVA1	AF275945	epithelial V-like antigen 1	3.00	.0002
FBLP-1	AL133035	filamin-binding LIM protein-1	-3.01	.0131
ITGA3	NM_002204	integrin, alpha 3	1.62	.0002
ITGB6	AK026736	integrin, beta 6	2.50	.0002
PDZKIIPI	NM_005764	PDZK1-interacting protein 1	-2.03	.0000
NID2	NM_007361	nidogen 2	-1.91	.0005
OCLN	AI829721	occludin	-1.70	.0512
Cell cycle (n = 8)
DUSP1	NM_004417	dual-specificity phosphatase 1	-1.94	.0053
Cell growth/maintenance (n = 42)
BPAG1	AI798790	bullous pemphigoid antigen 1, 230/240 kDa	1.64	.0000
DEFB1	U73945	defensin, beta 1	-7.78	.0000
GDI2	D13988	GDP dissociation inhibitor 2	-1.86	.0000
SEMA3C	NM_006379	sema domain, immunoglobulin domain (Ig), secreted, (semaphorin) 3C	6.22	.0000
TACSTD2	J04152	tumor-associated calcium signal transducer 2	-4.03	.0325
Cell motility (n = 8)
F11R	AF154005	F11 receptor	-2.09	.0325
JAG1	U77914	jagged 1	2.19	.0000
Metabolism (n = 29)
CA2	M36532	carbonic anhydrase II	-17.04	.0000
CITED2	AF109161	Cbp/p300-interacting factor, with Glu/Asp-rich carboxy-terminal domain, 2	-2.02	.0340
EXT1	NM_000127	exostoses (multiple) 1	-1.63	.0005
Mesenchymal genes (n = 173)
Actin cytoskeleton organization (n = 9)
ACTN1	AI082078	actinin, alpha 1	1.9	.0005
PLEK2	NM_016445	pleckstrin 2	2.02	.0008
Cell adhesion/ECM-related (n = 21)
BGN	BC002416	biglycan	10.5	.0000
CD44	AF098641	CD44 antigen	2.10	.0000
COL5A1	NM_000393	collagen, type V, alpha 1	2.05	.0000
COL5A2	AL564683	collagen, type V, alpha 2	39.29	.0000
FN1	AK026737	fibronectin 1	-10.39	.0425
Lamb1	M20206	laminin, beta 1	2.91	.0000
Cell cycle (n = 10)
CDC2	NM_001786	cell division cycle 2, G1 to S and G2 to M	-1.69	.0523
Cell growth and/or maintenance (n = 59)
CXCL1	NM_001511	chemokine (C-X-C motif) ligand 1	-1.78	.0015
EPS8	NM_004447	epidermal growth factor receptor pathway substrate 8	2.07	.0000
FZD1	NM_003505	frizzled homolog 1	1.86	.0006
FZD2	L37882	frizzled homolog 2	1.66	.0001
HMGA2	NM_003483	high mobility group AT-hook 2	2.76	.0001
IGFBP1	NM_000596	nsulin-like growth factor binding protein 1	-9.94	.0067
IGFBP3	M31159	insulin-like growth factor binding protein 3	-2.90	.0006
KIFAP3	NM_014970	kinesin-associated protein 3	2.71	.0067
Met	BG170541	met proto-oncogene (hepatocyte growth factor receptor)	2.77	.0015
PMP22	L03203	peripheral myelin protein 22	2.05	.0013
PTPRM	NM_002845	protein tyrosine phosphatase, receptor type, M	-2.74	.0013
SLC29A1	AF079117	solute carrier family 29 (nucleoside transporters), member 1	1.90	.0003
Cell motility (n= 25)
MMP10	NM_002425	matrix metalloproteinase 10 (stromelysin 2)	3.59	.0000
MMP2	NM_004530	matrix metalloproteinase 2	4.10	.0000
MMP7	NM_002423	matrix metalloproteinase 7	7.51	.0000
PLAU	NM_002658	plasminogen activator, urokinase	2.90	.0000
PLAUR	X74039	plasminogen activator, urokinase receptor	4.02	.0000
S100A2	NM_005978	S100 calcium binding protein A2	2.20	.0000
S100A3	NM_002960	S100 calcium binding protein A3	2.04	.0004
S100A4	NM_002961	S100 calcium binding protein A4	2.05	.0000
S100A6	NM_014624	S100 calcium binding protein A6	1.64	.0001
S100P	NM_005980	S100 calcium binding protein P	-29.48	.0002
SERPINH1	BF316352	serine (or cysteine) proteinase inhibitor	-2.44	.0000
Development/differentiation (n = 23)
ID1	D13889	inhibitor of DNA binding 1, dominant negative helix-loop-helix protein	-1.94	.0000
ID3	NM_002167	inhibitor of DNA binding 3, dominant negative helix-loop-helix protein	-23.23	.0001
ID4	U16153	inhibitor of DNA binding 4, dominant negative helix-loop-helix protein	-3.89	.0001
LUM	NM_002345	lumican	18.65	.0001
SNAI2	AI572079	snail homolog 2 (Drosophila)	4.91	.0000
SPARC	NM_003118	secreted protein, acidic, cysteine-rich (osteonectin)	8.94	.0000
Metabolism (n = 22)
ACVR1	NM_001105	activin A receptor, type I	2.74	.0045
BHLHB2	NM_003670	basic helix-loop-helix domain containing, class B, 2	2.36	.0000
PTGIS	NM_000961	prostaglandin I2 (prostacyclin) synthase	-2.17	.0012
Biological process unknown (n = 3)
PSTPIP2	NM_024430	proline-serine-threonine phosphatase interacting protein 2	-2.67	.0003
UPP1	NM_003364	uridine phosphorylase 1	-2.80	.0000

^*The genes listed in the EMT-related gene database (Table W1) were extracted from the DNvHNF1/Mock data set generated using the GeneChip Human Genome U133 Plus 2.0 Array. Note that 307 (94%) of 326 EMT-related genes were present in the chip.

^†The total number of genes belonging to each category is shown in parentheses.

Luciferase Assay

Cells were transfected with TOP- and FOP-promoter-reporter gene constructs (Upstate Biotechnology, Lake Placid, NY) using Lipofectamine 2000 according to the manufacturer's suggestions (Invitrogen). Cotransfection with thymidine kinase-Renilla was performed to evaluate transfection efficiency. After 48 hours, cells were lysed and analyzed for promoter activity. The dual-luciferase assay was performed essentially as suggested by the manufacturer (Promega, Madison, WI).

Growth Potential Measurements

In vitro proliferation of stable transfected cells was measured with the CellTiter-Glo luminescent cell viability assay kit (Promega) according to the manufacturer's suggestions. Cells (4 x 10⁴ cells per well) were cultured in 96-well plates for up to 5 days. In vitro proliferation of siRNA-treated cells was evaluated as radiolabeled thymidine incorporation. Briefly, cells were plated in 96-well plates at a density of 1 x 10⁴ cells per well and transiently transfected with siRNA duplex against vHNF1 mRNA or control siRNA. Cells were pulsed for 4 hours with [methyl-³H]thymidine (Amersham; 1 µCi per well) and 24 (for IGROV1 cells) or 48 hours (for SKOV3cells) later washed twice with ice-cold PBS. After fixation with 100 µl of 10% trichloroacetic acid for 30 minutes at 4°C, cells were lysed with 100 µl per well 0.2 N NaOH and radiolabeled thymidine incorporation was measured by scintillation counting.

Immunohistochemistry

All clinical specimens were obtained with approval from the institutional review board and informed consent from all participating patients to use excess biologic material for investigative purposes. Immunohistochemistry (IHC) was performed by using routine tissue blocks and a commercially available tissue arrays (Ovary cancer, AccuMax Array, Petagen and CJ1 Human, Ovary cancer, and Super Bio Chips) essentially as described [28]. For antigen retrieval and primary Ab dilutions, see Supplementary data. Two observers (M.L.C. and A.T.) independently assessed positivity or negativity of staining on the basis of intensity and the percentage of positive cells.

Statistical Analyses

GraphPad Prism software (GraphPad Software, San Diego, CA) was used to analyze all data. Differences between mean values were determined by Student's t test, and Fisher's test was used to determine whether the percentage of EMT-related genes is different by chance. The correlation of vHNF1 and E-cadh expression levels in IHC was evaluated by χ² test. P values <.05 (2-sided) were considered significant.

Results

vHNF1 Silencing Impairs Epithelial Differentiation of Ovarian Tumor Cells

To identify the role of vHNF1 in ovarian carcinoma cells, IGROV1 and SKOV3 cells were transfected with vHNF1-specific siRNA or control siRNA (Figure 1A). The reduction of endogenous vHNF1 transcription and protein expression correlated directly with a decrease in the levels of E-cadh transcription and protein expression in both cell lines. The decreased intensity of the 120-kDa band corresponding to the full-size E-cadh protein in lysates of vHNF1 siRNA-treated cells (Figure 1B) was not caused by degradation [29] because proteins of lower molecular weight were equally abundant in all lysates.

Figure 1.

vHNF1 silencing impairs epithelial differentiation of ovarian tumor cells. IGROV1 and SKOV3 cells were treated with a vHNF1-specific siRNA. (A) Quantitative RT-PCR on total RNA extracts from cells treated with a control (light gray bar) or vHNF1-specific (dark gray bar) siRNA. Data represent mean (SD) for the vHNF1 and E-cadh genes normalized to the housekeeping gene GAPDH in at least six determinations. Asterisks indicate significant differences (P < .05). (B) In a parallel experiment, cells were lysed and analyzed by Western blot analysis with Abs against vHNF1 and E-cadh, respectively. Co siRNA indicates control siRNA. β-Actin was used for normalization of gel loading. One of three experiments is shown. (C) IGROV1 and SKOV3 cells were treated with a control (light gray bar) or a vHNF1-specific (dark gray bar) siRNA as in Figure 2, and proliferation was evaluated by incorporation of radiolabeled thymidine. Data are mean (SD) of six replicates; one of two experiments is shown. Asterisks indicate significant differences (P < .05).

Further, we measured the growth capability of IGROV1 and SKOV3 cells treated with vHNF1-specific siRNA (Figure 1C). By radio-labeled thymidine incorporation, we observed that reduction of endogenous vHNF1 expression led to a 25% (P = .026) and 45% (P = .018) decrease in proliferation of IGROV1 and SKOV3 cells, respectively.

vHNF1 Loss-of-Function Impairs Epithelial Differentiation of Ovarian Tumor Cells

o scrutinize the role of vHNF1 expression in ovarian carcinomas, we stably transfected IGROV1 and SKOV3 cells with an expression plasmid containing a DNvHNF1 cDNA encoding the truncated vHNF1 form that occurs naturally in pancreatic β-cells of patients bearing maturity-onset diabetes of the young type 5 (Figure 2A) [30].We were unable to obtain any stable clone despite repeated DNvHNF1 transfections in IGROV1 cells (data not shown), but we obtained several stable clones by transfecting SKOV3 cells. Western blot analysis of total cell lysates from DNvHNF1-transfected SKOV3 clones (hereafter designated DNvHNF1) using an Ab that detects both wt vHNF1 and DNvHNF1 revealed the truncated 27-kDa band only in DNvHNF1 cells (Figure 2B). Probing the same blots with an Ab against E-cadh showed a large decrease in the expression of this protein in all DNvHNF1 clones analyzed. Furthermore, we obtained the same results on transfection of MDCK cells with the same construct (data not shown). Clone 1 was further characterized.

Figure 2.

vHNF1 loss-of-function impairs epithelial differentiation of ovarian tumor cells. (A) Schematic representation of wt vHNF1, with its functional domains, and the truncated DNvHNF1. Note that DNvHNF1 only maintains the N-terminal dimerization and B domains. (B) Western blot analysis of total cell lysates from Mock cells and DNvHNF1 clones was performed using a rabbit anti-HNF1 Ab. The 15- to 24-kDa bands in DNvHNF1 lysates might represent shorter DNvHNF1 products; β-actin was used for normalization of gel loading. (C) Upper panel: Morphology of Mock and DNvHNF1 cells. Cells were grown to confluence in six-well plates, and images were obtained by phase-contrast microscopy with a 10x objective. Bar, 100 µm. Lower panel: IF was performed on methanol-fixed Mock and DNvHNF1 cells with Abs against the molecules indicated. Images were obtained with a 40x objective. (D) Electrophoretic mobility shift assay of NEs prepared from Mock and DNvHNF1 cells was performed using two oligonucleotides containing the HNF1 consensus DNA-binding site and corresponding to the proximal elements of the sequences of the albumin (high affinity) and FR (low affinity) promoters, respectively. Specific DNA-protein complexes were competed with a 100-fold molar excess of unlabeled probes (100x Comp), as indicated. (E) Quantitative RT-PCR of the FR transcript using total RNA extracted from Mock and DNvHNF1 transfectants. Data represent mean (SD) for FR expression normalized to the housekeeping gene GAPDH in at least six determinations. Asterisk indicates a significant difference (P < .01).

Consistent with the loss of E-cadh expression, in phase-contrast microscopy DNvHNF1 clones revealed a spindle-like shape and loss of defined cell-cell borders compared with the more epithelial morphology of mock-transfected cells (hereafter referred to as Mock cells; Figure 2C, upper panel). Immunofluorescence analysis of adherens junctions showed that E-cadh and β-catenin (ctn) expressions were confined to cell-cell contacts in Mock cells, whereas in DNvHNF1 cells, there was loss of E-cadh staining and discontinuous β-ctn staining at cell-cell contacts (Figure 2C, lower panel). Moreover, both ZO-1 and occluding 1, markers of tight junctions, were clearly present at cell-cell contacts in Mock cells, but they were mainly concentrated in the nuclei of DNvHNF1 cells, as previously shown in other cell systems [31,32]. Normal and transformed ovarian cells may display both epithelial and mesenchymal features [3], and accordingly, SKOV3 cells coexpress both cytokeratins and vimentin. Interestingly, DNvHNF1 cells maintained coexpression of cytokeratins and vimentin but failed to display the typical pattern of cell-cell junctional cytokeratin filaments observed in Mock cells. Similar expression patterns were seen in the other DNvHNF1 clones (data not shown).

Electrophoretic mobility shift assay was performed with NEs from DNvHNF1 cells using two different vHNF1-specific oligonucleotides corresponding to the proximal elements of the sequences of the albumin (high affinity) and FR (low affinity) promoters [22,33]. This analysis indicated that the ability of NEs from DNvHNF1 cells to form DNA complexes with the oligonucleotides was substantially reduced compared to those from Mock cells (Figure 2D). On the basis of our previous demonstration that vHNF1 binds to and activates the FR promoter in ovarian carcinoma cells, we evaluated the FR transcript levels in DNvHNF1 cells to confirm the effective down-regulation of endogenous vHNF1 transcriptional activity by expression of the DNvHNF1 protein. Indeed, real-time RT-PCR analysis revealed a fourfold lower level of the FR transcript in DNvHNF1 cells (Figure 2E).

vHNF1 Loss-of-Function Induces a Gene Expression Profile Resembling That of EMT

We compared the gene expression profiles of DNvHNF1 cells and Mock cells by the GeneChip Human Genome U133 Plus 2.0 Array. Using a cutoff of 1.5-fold, we identified 459 upregulated and 473 downregulated genes in DNvHNF1 cells. A preliminary analysis of the differentially expressed genes indicated a pattern suggestive of EMT (data not shown). Thus, we focused on EMT and compiled a list of specific epithelial (140) and mesenchymal (186) genes that are reported in Table W1 (for list compilation, refer to the Materials and Methods section). In the epithelial gene list, genes associated with cell growth/maintenance, metabolism, cell adhesion/extracellular cell matrix (ECM)-related, and development/differentiation were the largest classes; in the mesenchymal gene list, cell growth/maintenance and cell motility-associated genes formed the largest classes. We extracted the expression data for each gene of this EMT-related list from the DNvHNF1 versus Mock data sets. A search for variation of expression of these genes in DNvHNF1 cells was consistent with our initial observation: 22% (67/307) of EMT-related genes were expressed differently in the two cell lines (Table 1), compared to the 13% expected by chance (P < .0001). More specifically, 24 of 134 epithelial genes in DNvHNF1 cells were differentially expressed: 16 of them were down-regulated and 8 were upregulated, which is consistent with published EMT data [26,27] (Figure 3A). In addition, 43 of 173 mesenchymal genes were differentially expressed and 29 of them (67%) were upregulated in DNvHNF1 cells. Most modulated genes in DNvHNF1 cells were associated with cell adhesion/ECM (18 genes), cell growth/maintenance (16 genes), or motility (13 genes). Among the mesenchymal genes, we observed up-regulation of CD44, Met, PLAU, PLAUR, MMP2, MMP7, S100A4, HMG2A, SNAI2, and SPARC, all of which are expressed during EMT. Real-time RT-PCR for PLAU showed upregulation of these genes in DNvHNF1 cells (Figure 3B).Western blot analysis with anti-S100A4 Ab indicated increased expression of this protein in DNvHNF1 cells compared to Mock cells (Figure 3C). Among epithelial genes, we found down-regulation of CDH1, which encodes E-cadh, and OCLN, which encodes for occluding, that had been shown before to be downregulated in DNvHNF1 cells (Figure 2), and TACSTD2, encoding Ep-CAM protein, which is highly expressed in ovarian carcinoma cells [34].

Figure 3.

vHNF1 loss-of-function induces a gene expression profile resembling that of EMT. (A) Upper panel: we compiled a list of specific epithelial (134) and mesenchymal (173) genes that are reported in Table W1. The number of genes in the largest functional classes is reported. Lower panel: we extracted the expression data for each gene of this EMT-related list from the DNvHNF1 versus Mock data sets. Epithelial (dashed bars) and mesenchymal (dotted bars) genes differentially expressed in DNvHNF1 versus Mock data sets. (B) Quantitative RT-PCR for target mRNA was performed with total RNA extracted from Mock (light gray bars) and DNvHNF1 (dark gray bars) cells. Data represent mean (SD) for the relevant genes normalized to the housekeeping gene GAPDH in at least six determinations. Asterisks indicate significant difference (P < .02). (C) Western blot analysis of total cell lysates from Mock and DNvHNF1 cells was performed using a rabbit anti-S100A4 Ab. β-Actin was used for normalization of gel loading. One of three gels is shown.

Some apparent discrepancies with published data were found, such as the up-modulation of ITGB6 and SEMA3C, described before as down-modulated in the epithelial phenotype [26], and the down-regulation of typical mesenchymal genes, such as FN and IDs [9,35]. Nevertheless, real-time RT-PCR confirmed the down-regulation of FN in DNvHNF1 cells (Figure 3B).

These results strongly suggest that vHNF1 loss-of-function impairs cell-cell adhesion and leads to a more mesenchymal phenotype in the SKOV3 ovarian carcinoma cell line.

vHNF1 Loss-of-Function Leads to Slug Expression and Functionality

E-cadh down-regulation could have been caused by transcriptional repression by E-box transcription factors, so we performed quantitative real-time RT-PCR to evaluate the expression of E-cadh and its transcriptional regulators Snail and Slug. Compared to Mock cells, DNvHNF1 cells showed absence of E-cadh mRNA, 2-fold decreased Snail mRNA levels, and 1.8-fold increased Slug mRNA levels (Figure 4A). We then analyzed in these cell lines the effect of vHNF1 loss-of-function on Cdh1 promoter activity (Figure 4C) by transiently transfecting the promoter constructs depicted in Figure 4B. Luciferase reporter assay showed that the activity of the E-box-containing Cdh1 promoter significantly decreased about threefold in DNvHNF1 cells compared to Mock cells. The activities of the mutated E-box-containing construct were less derepressed in DNvHNF1 cells than in Mock cells, indicating that in DNvHNF1 cells, other mechanisms may contribute to the repression of the Cdh1 promoter in addition to the activity of E-box-binding proteins.

Figure 4.

vHNF1 loss-of-function leads to Slug expression and functionality. (A) Quantitative RT-PCR for E-cadh, Snail, and Slug transcripts was performed on total RNA extracted from Mock (light gray bars) and DNvHNF1 (dark gray bars). Data represent mean (SD) for the genes indicated, after normalization to the housekeeping gene GAPDH in at least six determinations. Asterisks indicate significant differences (P ≤ .02). (B) Schematic representation of Chd1 proximal promoter containing four putative E-box sequences cloned upstream of the luciferase gene and transiently transfected in Mock and DNvHNF1 cells. Mutations within the E-boxes are as indicated. (C) Luciferase-promoter gene assay of Mock (light gray bars) and DNvHNF1 (dark gray bars) cells transiently transfected with reporter plasmids containing the wt Cdh1 proximal promoter or the same promoter with mutated E-box sequences (mEbox) as reported in panel B. Data are mean (SD) normalized for transfection efficiency in three independent experiments performed in triplicate. Asterisk indicates a significant difference (P ≤ .01).

These results indicate that loss of E-cadh expression in DNvHNF1 cells might be caused by transcriptional repression partly mediated by Slug binding to specific E-boxes of the Cdh1 promoter.

Ectopic Expression of vHNF1 in IOSE Cells Is Sufficient to Induce Snail and Slug

The data presented above indicate that vHNF1 participates in determining the epithelial phenotype of ovarian cancer cells. To evaluate whether vHNF1 is sufficient to activate a differentiation program toward MET in normal ovary cells, we stably transfected hTERT-IOSE cells [23] with vHNF1 cDNA. Western blot analysis with anti-vHNF1 Ab revealed that two selected clones of vHNF1-transfected hTERT-IOSE (hereafter designated vHNF1-IOSE #1 and #2) express a 62-kDa vHNF1 protein not expressed by mock-transfected hTERT-IOSE (hereafter designated Mock-IOSE; Figure 5A). By phase-contrast microscopy, vHNF1-IOSE #2, which represents a clone grown in vitro for longer time than vHNF1-IOSE #1 cell line, appeared to lose the typicalmesenchymalmorphology,whereas both Mock- and vHNF1-IOSE #1 maintained a more spindle-like morphology (Figure 5B).

Figure 5.

Ectopic expression of vHNF1 in IOSE cells is sufficient to induce Snail and Slug. (A) Western blot analysis of total cell lysates from Mock and DNvHNF1-IOSE clones #1 and #2 was performed with a rabbit anti-vHNF1 Ab. (B) Morphology of Mock and DNvHNF1-IOSE cells. While Mock- and vHNF1-IOSE #1 showed a more spindle-like morphology, vHNF1-IOSE #2, which represents a clone grown in vitro for longer time than vHNF1-IOSE #1 cell line, appeared larger in size acquiring a more compacted morphology. Images were obtained by phase-contrast microscopy using a 10x objective. Bar, 100 µm. (C) Quantitative RT-PCR for Snail and Slug transcripts was performed on total RNA extracted from Mock (white bar) and vHNF1-IOSE #1 (light gray bars) and #2 (dark gray bars) cells. Data represent mean (SD) for the relevant genes normalized to the housekeeping gene GAPDH in at least six determinations. Asterisks indicate significant differences (P ≤ .05). (D) Luciferase-promoter gene assay using Mock (white bar) and DNvHNF1-IOSE #1 (light gray bars) and #2 (dark gray bars) cells transiently transfected with promoter reporter plasmids containing the wt Cdh1 proximal promoter or the same promoter with mutations in the E-box sequences (mEbox) as reported in Figure 4B.

By quantitative RT-PCR on total RNA, the E-cadh transcript was slightly detectable in all transfected IOSE (data not shown). Expression of Snail transcript in vHNF1-expressing clones was found to be downregulated 3- and 4-fold, and Slug transcript 2.5- and 3.5-fold, respectively (Figure 5C).

We then analyzed the effect of vHNF1 loss-of-function on the Cdh1 promoter activity in Mock and vHNF1-IOSE cells (Figure 5D) by transiently transfecting the promoter constructs (shown in Figure 4B). Luciferase reporter assay showed that the activity of the E-box-containing wt Cdh1 promoter increased approximately 2.5-fold in vHNF1-IOSE cells in comparison to Mock-IOSE cells. The activities of the mutated E-box1-3-4 construct was approximately twofold de-repressed in vHNF1-IOSE cells compared toMock-IOSE, suggesting that E-boxes 1, 3, and 4 are relevant for the CDH1 gene transcription in this type of cells.

These results together demonstrate that vHNF1 negatively regulates specific E-box-binding repressors in normal ovarian cells, which is in line with the data on ovarian carcinoma cells reported above.

vHNF1 Modulates the Proliferative Potential of Ovarian Cancer and Normal Cells

We aimed also to evaluate vHNF1-dependent growth potential in both DNvHNF1 and vHNF1-transfected-IOSE. In culture, DNvHNF1 cells grew slower than Mock cells, so that the DNvHNF1 cell density was 30% lower on day 5 after seeding (P = .0001; Figure 6A), consistent with the decrease in proliferation after transient vHNF1 silencing (Figure 1C). The slower growth rate was paralleled by reduced colony-forming capability (data not shown).

Figure 6.

vHNF1 modulates the proliferative potential of ovarian cancer and normal cells. Cells were seeded in 96-well plates, and growth was measured for up to 5 to 7 days with a CellTiter-Glo luminescent cell viability assay kit (Promega). (A) Mock (▪) and DNvHNF1 (▴): data represent mean (SD) of six determinations from three independent experiments. (B) Mock (○) and vHNF1-IOSE #1 (●) and #2 (▴): data represent mean (SD) of five determinations from one of three experiments. Asterisks indicate significant differences (P ≤ .05).

Growth potential was also evaluated in the IOSE transfected cells up to 7 days. It is noteworthy that ectopic vHNF1 expression caused a two- and three-fold increase of the growth rate of vHNF1-IOSE #1 and #2, respectively, in comparison to Mock-IOSE cells (Figure 6B).

These results are in line with those shown above (Figure 1A) and suggest a positive involvement of vHNF1 in controlling the proliferation of normal and transformed ovarian cells.

vHNF1 Is Expressed in a Subset of Normal and Transformed E-cadh-Expressing Ovarian Cells

We previously showed that vHNF1 is expressed only in ovarian carcinoma cell lines and not in short-term cultures of OSE cells [22]. Here, we used IHC to evaluate vHNF1 expression together with E-cadh in sections from four normal human ovaries, selected for having a normal monolayer epithelium or for presenting invaginations and inclusion cysts lined by a single layer of cells, and in samples from benign, low malignant potential (LMP) and malignant ovarian tumors of different histotypes (Figure 7). When detected, anti-vHNF1 staining was observed only in the nucleus or in the nucleus and the cytoplasm, whereas anti-E-cadh mainly stained the cell membrane. No vHNF1 expression was detected in OSE from different individuals. In 20% of cysts, the single, normal cell layer reacted with anti-vHNF1 together with anti-E-cadh (representative example in Figure 7). Cells from four benign tumors and four of six LMP tumors stained for vHNF1 and E-cadh (Table 2). One LMP tumor that did not stain with anti-vHNF1 was endometrioid. Among the 38 carcinomas tested, 18 reacted with anti-vHNF1: 7 of these were clear cell carcinomas, 5 were serous (representative examples in Figure 7A), and 1 was mucinous. Four of 7 endometrioid EOCs were vHNF1-positive.

Figure 7.

vHNF1 is expressed in a subset of normal and transformed E-cadh-expressing ovarian cells. Immunohistochemical analyses with anti-vHNF1 and -E-cadh Abs on paraffin-embedded normal and tumor-derived ovarian tissues. The immunohistochemical analysis is also reported in Table 2. (A) Representative examples of normal ovarian epithelium (a), an inclusion cyst (b), and two serous EOCs (c and d). Images were obtained with a 20x objective. Bar, 100 µm. (B) Epithelial ovarian carcinomas analyzed for vHNF1 and/or E-cadh expressions. Anti-E-cadh-negative samples comprise other cadh-expressing tumors. Total number of EOCs, n = 38; total number of serous EOCs, n = 19. Vertical bars, percentage of immunoreactive samples.

Table 2.

Expression of vHNF1 and E-cadh Detected by Immunohistochemistry in Ovarian Tumor Samples.

Ovarian Tumors*	No. of Cases (n = 49)	Presence of vHNF1/E-cadh by IHC
		+/+	+/-	-/+	-/-
Benign	5	4	—	1	—
LMP	6	4	—	1	1†
Carcinoma‡:
Serous	19	5	1	9	4
Mucinous	2	1	—	1	—
Clear cell	10	7	1	—	2
Endometrioid	7	—	3	3	1

^*Commercially available tissue arrays.

^†Endometrioid LMP tumor.

^‡Mucinous and endometrioid carcinomas comprise a grade I tumor; all other carcinomas were grades II and III.

Interestingly, 66% of EOCs tested were reactive with anti-E-cadh monoclonal Ab, and within these EOCs, 33% were positive for both vHNF1 and E-cadh (Figure 7B). Within serous histotype, which represents most EOCs, 73% expressed E-cadh and 26% together with vHNF1. Note that E-cadh-negative tumors could express N-cadh or cadh-11, as previously reported [23].

In conclusion, vHNF1 appears to be expressed in inclusion cysts, and it is clearly expressed in clear cell carcinomas and in some serous carcinomas, but it is not expressed in OSE. Interestingly, vHNF1 expression is significantly associated with E-cadh expression in a subset of samples comprising some cysts as well as in benign and malignant tumors of serous or clear cell histotype, whereas no coexpression was observed in endometrioid tumors (P = .0024).

Discussion

Here, we demonstrate that vHNF1 may act as an initial regulator of OSE plasticity and proliferation, thereby contributing significantly to the changes in differentiation of OSE cells during neoplastic transformation and progression. Indeed, a DN form of the transcription factor vHNF1 induces EMT when expressed in an ovarian carcinoma cell line (SKOV3), as confirmed by a change in mRNA expression profile resembling that of EMT and by a loss at the protein level of E-cadh and components of tight junctions. DNvHNF1 expression in SKOV3 ovarian carcinoma cells downregulated Slug expression and functionality, and conversely, vHNF1 ectopically expressed in hTERT-IOSE cells decreased Snail and Slug expression and functionality. Overall, our results uncover a novel role of vHNF1 in the epithelial differentiation of ovary cells.

The HNF transcription factors have been related mainly to hepatocyte and pancreatic β-cell differentiation, and vHNF1 in particular seems to be required for maintenance of the differentiation state and functional activity of mouse pancreatic β-cells. Mutations and/or deletions that impair vHNF1 functionality cause major alterations in expression of important metabolic genes. Indeed, vHNF1 knock-out mice die within a few days after gastrulation [19]. In adult mice, β-cells show impaired glucose tolerance and reduced insulin secretion when vHNF1 is selectively deleted using Cre recombinase [36]. The present results suggest for the first time that vHNF1 is one of the transcription factors governing the epithelial differentiation of OSE.

To evaluate the molecular signature associated with vHNF1 loss-of-function in SKOV3 cells, we performed an EMT-guided comparative expression profile analysis on the basis of a newly compiled EMT-related gene database (Table W1). The observed profile is in agreement with an EMT shift, with a few remarkable exceptions, such as down-modulation of FN1, ID1, and ID4 and up-regulation of ITGB6 and SEMA3C. This could simply be caused by cell line specificity of these molecules. Note that vHNF1 loss-of-function does not completely revert the malignancy of SKOV3 cells and that further analysis is needed to determine whether particular molecular mechanisms regulate the expression of these genes in a tumor type-specific way. Conversely, several genes characteristic of either EMT or MET but not yet associated with EOCs were identified as differentially expressed. Microarray and immunofluorescence analyses showed that vHNF1 functionality was associated mainly with genes that modulate cell adhesion and are ECM-related, which is consistent with a role for HNF transcription factors in regulating cell adhesion [37]. The potential usefulness of this information in the context of clinical screening markers and possible genetic or pharmacologic targeting awaits further validation of the role of vHNF1 and the newly identified associated genes, together with detailed gene expression comparisons between OSE and EOC samples.

Consistent with a previous report that only EOC with clear cell histotype expresses vHNF1 [38], our immunohistochemical analysis revealed vHNF1 expression in most clear cell EOCs, as well as in approximately 30% of serous EOCs, which are most EOCs. We detected heterogeneous expression of vHNF1 associated to E-cadh expression in 32% of the EOC samples, independently of tumor grading. These results confirm a role for vHNF1 in the epithelial phenotype of EOCs. An analysis to define the molecular signature that characterizes this subset of EOCs, which include mainly the clear cell and serous histotypes, is ongoing.

Previous studies indicated that the proximal vHNF1 promoter was methylated in 26% of the EOCs analyzed, but no such methylation was observed in OSE cells that do not express vHNF1 [39]. In renal cell carcinomas, TCF2 inactivation was caused by germ line mutations [17]. Therefore, in vHNF1-negative ovarian carcinomas, TCF2 gene could have either methylated promoter or inactivating mutations. We further hypothesize that cysts undergo transformation if they contain genetic mutations and possibly express vHNF1 conferring a further growth advantage. Epithelial ovarian carcinomas derived from those cysts continue to express vHNF1, but once tumors progress, epigenetic mechanisms such as methylation of the vHNF1 promoter might be activated, resulting in loss of vHNF1 expression. One key function of vHNF1 seems to be the negative modulation of EMT-inducing transcription factors such as Snail and Slug, leading to the positive modulation of E-cadh and other epithelial proteins. Previously, vHNF1 has been shown to be involved in METoccurring during kidney development, whereas kidney fibrosis has been associated with the binding of the E-box-binding repressors Snail or Slug to the promoters of vHNF1 and E-cadh encoding genes [40]. These results together with ours favor the hypothesis of a delicate reciprocal transcriptional regulation between E-box repressors and vHNF1. Recently another embryonic transcription factor, FOXC2, was identified as a central modulator of the EMT program in metastatic basal-like breast cancer [41]. These observations, together with ours, support the hypothesis that embryonic transcription factors are necessary for execution of transformation or invasion programs in different types of cancers. Of course, this hypothesis does not exclude that other mechanisms, such as those involving specific HOX genes, might contribute to determining the morphologic heterogeneity of EOCs [42].

In addition to its role in OSE cell plasticity, vHNF1 seems to contribute to the increased growth potential of normal and transformed ovarian cells. Indeed, siRNA-mediated silencing of vHNF1 or its inhibition by a DN mutant was associated with decreased growth proliferation, whereas de novo expression of vHNF1 increased proliferation. The increase in proliferation could be attributed to the modulation of cell cycle progression but did not confer unlimited growth potential to OSE cells (unpublished observation). We can also hypothesize that the vHNF1 confers a growth advantage in vitro so that ovarian carcinoma cell lines maintain vHNF1 expression, whereas in tumors the genomic modifications described above lead to the loss of vHNF1 expression.

Recently, a new model for the pathogenesis of EOCs has been proposed in which ovarian tumors are divided in two types [43]. Type I tumors, which include low-grade serous, mucinous, endometrioid, and clear cell carcinomas, are slowly growing and are generally confined to the ovary. Type II tumors are rapidly growing and highly aggressive. Despite considerable efforts, it is not yet possible to distinguish these different types of ovarian tumors at early stage and set up the most successful therapy. In this context, the strong association of vHNF1 with E-cadh in clear cell and in a subset of serous carcinomas could potentially contribute in distinguishing different types of ovarian tumors, on a more extensive molecular analysis. Epithelial-mesenchymal transition has been recognized as a potential mechanism for carcinoma progression. The mechanisms governing EMT in tumor progression recapitulate many of those identified in embryogenesis [10–12]. However, besides EOC with endometrioid histotype, EOCs seem to diverge in other ways from the general EMT scenario. For example, β-ctn does not detectably activate β-ctn/TCF-responsive genes on progression [44], whereas E-cadh expression is maintained in advanced EOCs [6,45]. Therefore, increasing knowledge of the molecular mechanisms of METoccurring at first stages of tumorigenesis and controlled by vHNF1 in EOCs may provide new and fascinating insights into the biology of this important disease and likely to identifying early detection markers and to opening potential avenues for therapeutic intervention.

Abbreviations

Ab

antibody

cadh

cadherin

ctn

catenin

DN

dominant-negative

EMT

epithelial-mesenchymal transition

EOC

epithelial ovarian carcinoma

FR

folate receptor

IHC

immunohistochemistry

LMP

low malignant potential

MET

mesenchymal-epithelial transition

NE

nuclear extract

OSE

ovarian surface epithelium

vHNF1

variant hepatocyte nuclear factor

wt

wild type