GILZ expression was assessed by immunohistochemical staining of sections isolated from three normal ovaries, seven benign EOC and 50 invasive EOC. GILZ was not detected on the surface epithelium of normal ovaries and in benign tumors. In contrast, among the invasive ovarian cancers, 40 (80%) expressed GILZ. GILZ immunoreactivity was detected in the four main histological subtypes, serous, clear cell, endometrioid and mucinous tumors. It was clearly confined to the cytoplasm of tumor cells and was weak in, or absent from the tumor stroma (Figure 1A and 1B). Using the same antibody, we detected GILZ protein by western blot. GILZ was revealed at 17 kDa, which is the size of the protein described by Ricardi and co-workers in 1997 5, in BG-1 cells transfected with the GILZ-encoding vector pcDNA3-GILZ (pGILZ BG-1 cells used as positive control), in epithelial cells from malignant ascites and in ovarian tumor samples for which frozen tissues were available, confirming the staining data (Figure 1C). Interestingly, the non-epithelial cells from malignant ascites do not express GILZ. GILZ 17 kDa protein was also detected in ovarian cancer cell lines SKOV-3, OVCAR-3 and BG-1; it was less abundant in BG-1 cell line than in SKOV-3 and OVCAR-3 (Figure 1C). BG-1 thus appeared as the best-fitted cellular model for processing up and down regulation of GILZ.

GILZ immunostaining was unevenly distributed in tumor cells, from no detectable staining to strong immunoreactivity. We asked if differences in GILZ expression levels are related to the expression of two markers, the proliferation marker Ki-67 used in routine diagnostics 23 and p-AKT used to characterize malignant ovarian tumors 24. Hyperactivation of AKT is frequently observed in ovarian neoplasms and is related to the control of cell proliferation in EOC 2526. Immunoreactivity of GILZ, Ki-67 and p-AKT was measured on serial sections of EOC (Figure 2A). GILZ and Ki-67 immunostainings were scored on a seven-point scale based on the staining intensity and the extent of staining. GILZ and Ki-67 expression scores were significantly correlated in the entire cohort (Spearman test in univariate analysis, P < 0.00001, r = 0.56) (Figure 2B). They were still correlated in serous carcinoma and non serous carcinoma as well (Spearman test in univariate analysis, P < 0.001, r = 0.61, n = 26 serous carcinoma, P < 0.02, r = 0.49, n = 24 non serous carcinoma). The expression of p-AKT in tumor cells was mostly cytoplasmic, although some nuclear staining was also detected (Figure 2A). Both nuclear and cytoplasmic staining patterns were considered to assess p-AKT immunoreactivity, scored as high or low. GILZ expression scores were significantly higher in p-AKT^high specimens (mean ± SE = 5.2 ± 0.2, n = 25) than in p-AKT^low specimens (mean ± SE = 2.6 ± 0.4, n = 25), (unpaired t test, P < 0.0001) (Figure 2C).

After applying a single cut-off on the entire cohort for identification of GILZ^high (scores 5-7, n = 24) and GILZ^low (scores 0-4, n = 26) cases, we found that high GILZ scores are associated with higher p-AKT staining and Ki-67 indexes (Fisher test, P < 0.0002 and P < 0.0008), (Table 1). In contrast, age at diagnosis and distribution of histological subtypes did not differ between the two groups (Table 1).

To study the effect of GILZ on cell proliferation in epithelial ovarian cancer, we generated BG-1 clones that stably and strongly express GILZ (pGILZ). As a control BG-1 cells were stably transfected with an empty vector (CTRL). pGILZ and CTRL clones were randomly selected for further experiments. The GILZ protein content was significantly higher in pGILZ clones than CTRL clones (Figure 3A). We then compared their spontaneous cell proliferation: it was significantly higher in pGILZ clones (P < 0.01) (Figure 3B). To confirm that GILZ overexpression increased the proliferation rate, CTRL and pGILZ clones were seeded at equal densities, and viable cells were counted over a 4-day period. Cells overexpressing GILZ grew faster than CTRL cells (Figure 3C). There was no difference in spontaneous apoptosis between pGILZ and CTRL clones [see Additional file 1].

We next investigated whether over-expression of GILZ affected AKT activation. p-AKT, currently the active form of AKT, was more abundant in pGILZ clones than in CTRL clones, whereas the status of phospho-ERK 1/2 remained unchanged (Figure 3D). In parallel, AKT activity was higher in pGILZ clones than in CTRL clones as assessed by testing for phosphorylation of glycogen synthase kinase (GSK-3α/β), a downstream target of AKT (Figure 3E). Thus GILZ overexpression induced an increase in p-AKT and an enhancement of AKT activity. AKT-binding proteins may cause structural changes and phosphorylations that activate AKT 2728. We also revealed the presence of GILZ-AKT complexes in BG-1 cells using immunoprecipitation experiments (Figure 3F).

We studied the effects of knocking down GILZ mRNAs in BG-1 cells on cell proliferation and AKT activation. Real-time PCR and western blot analyses revealed that siRNA duplexes efficiently and specifically inhibited the expression of GILZ (mRNA and protein abundance) more than 75% lower than in cells treated with irrelevant control siRNA (Figure 4A).

Silencing GILZ gene expression led to a marked inhibition of cell proliferation and AKT phosphorylation, without changing phospho-ERK1/2 status (Figure 4, B and 4C).

Down regulation of GILZ expression in OVCAR3 cells, an ovarian cancer cell line that contains high amount of GILZ, also resulted in a decrease of cell proliferation [see Additional file 2]. These various findings reveal a previously unappreciated role of GILZ in the regulation of proliferation and AKT activation.

The cyclin-dependent kinase inhibitor p21 and cyclin D1 are two AKT targeted proteins that negatively (p21) and positively (cyclin D1) control cell cycle progression and proliferation 29. Cyclin D1 activates cyclin-dependent-kinases (CDK4/6), leading to phosphorylation of retinoblastoma (Rb) with the resulting promotion of cell cycle progression 30. We found that the overexpression of GILZ caused the up-regulation of cyclin D1 (mRNA and protein) and increased the amount of phosphorylated Rb (p-Rb); in contrast, p21 was down-regulated (Figure 5A). At the opposite, down regulation of GILZ resulted in decreased amount of cyclin D1 gene products (mRNA and protein) and p-Rb whereas those of p21 increased (Figure 5B). Thus the effects of down regulation of GILZ mirrored those of overexpression.

GILZ caused changes in p21 and cyclin D1 expression in such a way that increases in GILZ expression would accelerate cell cycle progression. To confirm this prediction we analyzed the cell cycle distribution of synchronized cells following removal of the thymidine block. We found that pGILZ cells entered S phase earlier than CTRL cells (Figure 5C).

To investigate whether AKT activation is required for control of BG-1 cell proliferation, we used Triciribine, a specific pharmacological inhibitor of AKT phosphorylation. Triciribine treatment (5-20 μM) reduced p-AKT levels and in parallel decreased spontaneous proliferation of pGILZ and CTRL clones (Figure 6A). These findings indicate that AKT phosphorylation contributes to BG-1 cell proliferation. Further, Triciribine also caused an up-regulation of p21 expression in both CTRL and pGILZ clones. Interestingly, cyclin D1 expression remained unchanged. In addition, GILZ levels remained unchanged suggesting that p-AKT inhibition did not significantly affect GILZ expression (Figure 6B).

These various experiments show that AKT activation controls the expression level of p21 and contributes to cell proliferation in BG-1 cells. The enhancement of AKT activation by GILZ therefore accounts for GILZ effect, at least in part, on cell proliferation. In contrast, the enhancement of cyclin D1 promoted by GILZ is disconnected of its action on AKT.

Approval was obtained from the ethics commission of the Antoine Béclère Hospital (Clamart, France) for all analyses of tumor material from clinical samples and from archival material from patients with a diagnosis of invasive ovarian carcinoma. Immunohistochemical examination of GILZ, phosphorylated protein kinase B/AKT (p-AKT) and Ki-67 proliferation index was performed retrospectively on tissue specimens of primary invasive ovarian carcinomas taken for routine diagnostic and therapeutic purposes from 50 patients who were treated surgically following a diagnosis of ovarian tumor at Antoine Béclère Hospital between 1998 and 2007. Clinical and pathological characteristics of the patients are detailed in Table 2. None of the patients had received neo-adjuvant chemotherapy before surgery. Clinical stage was assigned according to the International Federation of Gynecology and Obstetrics staging system (FIGO); histological subtypes and grades were assigned according to the criteria of the World Health Organization (WHO) classification 47.

Tumor cell enrichment was based on the expression of CD326, a human epithelial antigen also known as EpCAM, which is broadly expressed on cells of epithelial origin and derived tumor cells 51. CD326+ cells were positively selected using autoMACs columns (Myltenyi Biotech, Paris, France) from ascites collected from patients diagnosed with EOC. The percentage of CD326+ cells in the positive fraction was more than 80% as assessed using a FACScan flow cytometer (BD Biosciences).

BG-1 cells derived from a solid tumor tissue of a patient with stage III ovarian adenocarcinoma (a kind gift from Dr G. Lazennec, U844 Inserm Montpellier, France) were maintained in Dulbecco's modified minimum essential medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 2 mmol/L L-glutamine, 0.1 mg/mL streptomycin and 100 U/mL penicillin (Invitrogen, Ilkirch, France). The human ovarian carcinoma cell lines SKOV-3 and OVCAR-3 (American Type Culture Collection) were maintained in RPMI 1640 medium containing 0.1 mg/mL streptomycin, 100 U/mL penicillin, 2 mmol/L L-glutamine and 10% FBS. Triciribine was purchased from Calbiochem (VWR International SAS, Fontenay-sous-bois, France).

BG-1 cells were transfected using jetPEI (Polyplus-transfection, France) according to the manufacturer's protocol, with the GILZ-encoding vector pcDNA3-GILZ or with the empty vector pcDNA3 as a control. Forty-eight hours after transfection, stably transfected cells were selected by a 2-week treatment with 500 μg/mL of Geneticin (G-418; Invitrogen) and cloned by limiting dilution. Clones were then screened for GILZ expression using quantitative real-time PCR and immunoblot assays. GILZ expression remained stable over 7 days in culture. We thus generated BG-1 clones that stably and strongly expressed GILZ, named pGILZ clones. BG-1 clones stably transfected with an empty vector, named CTRL clones, were used as control.

Small interfering RNA (siRNAs) duplexes were synthesized and tested for specific inhibition of GILZ expression as described previously 10. BG-1 cells (5 × 10⁵ cells/well) were transfected with 4 μg/well GILZ siRNA (siGILZ) or control siRNA (siCT), purchased from Qiagen (Courtaboeuf, France), by the lipofectamine method using X-tremeGENE siRNA transfection reagent according to the manufacturer's instructions (Roche Diagnostics, Meylan, France). Transfection efficiency was between 80% and 90%, as assessed by using fluorescent random siRNA: siRNA AlexaFluor 488 (Qiagen).

Total RNA was extracted using a RNeasy Mini kit (Qiagen). The RNA was transcribed into cDNA by reverse transcription with random hexamers (Roche Diagnostics) and Moloney murine leukemia virus (MMLV) reverse transcriptase (Invitrogen). GILZ mRNA was quantified by real-time PCR on a Light Cycler instrument (Roche Diagnostics) using the FastStart DNA Master SYBER Green kit (Roche Diagnostics) as described previously 910. Values were normalized to those for β-actin mRNA and are thus expressed as the GILZ/β-actin ratio. p21 and cyclin D1 mRNAs were assayed by semi-quantitative RT-PCR as described previously 52. The primers used were as following: GILZ (294 bp) sense 5'-TCTGCTTGGAGGGGATGTGG-3', antisense 5'-ACTTGTGGGGATTCGGGAGC-3'; Cyclin D1 (413 bp) sense 5'-TGCATGTTCGTGGCCTCTAA-3', antisense 5'-CAGTCCGGGTCACACTTGAT-3'; p21 (331 bp) sense 5'-CGACTGTGATGCGCTAATGG-3', antisense 5'-CCGTTTTCGACCCTGAGAG-3'; β-actin (237 bp) sense 5'-GGGTCAGAAGGATTCCTATG-3', antisense 5'-GGTCTCAAACATGATCTGGG-3'.

Cells were seeded in triplicate on 96-well plates at a density of 1 × 10⁴ cells/well in DMEM medium with 10% FBS. Twenty-four hours later, [³H] thymidine (0.5 μCi/well) (PerkinElmer, Boston, US) was added and the samples incubated overnight. The radioactivity incorporated was determined as described previously 52 and results are expressed as counts per minute (cpm).

Cells (2 × 10⁶) were lyzed as described previously 17. Equivalent amounts of proteins were separated by SDS-polyacrylamide gel electrophoresis (SDS-PAGE), transferred to nitrocellulose membranes (Hybond-ECL, Amersham, Saclay, France) and probed with rabbit polyclonal Abs recognizing GILZ (Santa Cruz), AKT phosphorylated at Ser⁴⁷³, total AKT, and ERK1/2 phosphorylated at Thr^202/204, and retinoblastoma phosphorylated at Ser^807/811 (all from Cell Signaling). Anti-cyclin D1 and anti-p21 mAbs were purchased from Cell Signaling. Loading controls used goat anti-β-actin Ab from Santa Cruz. Primary Abs were visualized using HRP-conjugated anti-rabbit, anti-mouse and anti-goat IgG (Santa Cruz) and enhanced chemiluminescence detection (Amersham). ScanAnalysis software (Biosoft, Cambridge, United Kingdom) was used for densitometric analysis.

Total protein lysates from BG-1 clones were immunoprecipitated with polyclonal anti-AKT Ab (Santa Cruz) overnight and then the immune complexes were precipitated with protein G bound to sepharose beads (Sigma-Aldrich). The immunoprecipitates were immunoblotted with anti-GILZ Ab to investigate the presence of GILZ.

The nonradioactive AKT kinase assay kit was used according to the manufacturer's instructions (Cell Signaling). Immobilized AKT mAb was used to immunoprecipitate AKT from cell lysates and the samples subjected to an in vitro kinase assay using GSK-3 fusion protein as a substrate. Phosphorylation of GSK-3 was measured by western blotting using phosphorylated GSK-3α/β (Ser^21/9) Ab and chemiluminescent detection.

BG-1 cells were synchronized by double thymidine block as described previously 53. After releasing the block in DMEM-10% FBS, cell cycles were analyzed using propidium iodide (PI) staining and fluorescence was measured using a FACScan flow cytometer. Cell cycle profiles were analyzed by ModFit Cell Cycle Analysis software.

StatEL statistical software (Adscience, Paris, France) was used. The Spearman test was used to analyze the relationship between GILZ and Ki-67 scores. The two-tailed unpaired Student's t test was used to compare two groups and the Kruskal Wallis test followed by Dunn's test was used to compare several groups. Fisher's exact test was used to compare the relationship between the expression levels of GILZ and Ki-67 and of GILZ and p-AKT. Significance was set at P < 0.05.