Three ovarian cancer populations were used in the present study. The training set consisted of 278 ovarian cancer cases was purchased from US Biomax Inc. (Rockville, MD, USA). The validation set consisted of 150 ovarian cancer cases and was purchased from Super BioTek (Shanghai, China). The clinical biodemography, pathological features and clinical outcomes of patients were collected by retrospective chart review (Table I). The third cohort of 95 patients with ovarian cancer was obtained from the Northwestern Memorial Hospital between 2002 and 2007. This cohort had follow-up information and the information was previously reported (20). Ethics approval for the study was obtained from the Institutional Review Board.

T29 were plated in 6-well plates in antibiotic-free medium. When the cells reached 70–80% confluence, each well was transfected with a mixture containing either 4 μg of pCMV6-STC2 or control pCMV6 and 10 μl of Lipofectamine 2000. Seventy-two hours after transfection, the cells were selected for 20–30 days in the presence of 600 ng/μl G418. Single colonies with STC2 overexpression were selected and confirmed by western blot analysis.

STC2 shRNA (pGPU6/GFP/Neo-shRNA-1604 with target sequence GGGCAAGTCATTCATCAAAGA) and STC2 control (the scrambled shRNA construct pGPU6/GFP/Neo-shNC) plasmids were transfected into cells using Lipofectamine 2000 as previously described (7). Single colonies with STC2 knockdown were selected and confirmed by western blot analysis.

The cells (9×10⁵) were seeded in 6-well plates and allowed to reach confluence overnight in culture medium. A linear scratch was created using a sterile pipette tip. The media were then changed to remove any detached cells and the scratched areas were photographed under an inverted microscope at indicated time points. At least three fields along each scratch were analyzed and the experiment was performed in triplicate.

Cells (1×10⁵/well) were suspended in 200 μl of serum-free medium and then seeded on the upper side of a 24-well Transwell migration chamber (Corning Inc., Corning, NY, USA). After 48 h, the cells retained on the upside of the membrane were removed by cotton swab. The cells that migrated onto the bottom side were fixed by 10% formalin, stained by 0.1% crystal violet and observed under microscope at a magnification of ×100. The Matrigel invasion assay was similar to the Transwell migration assay, except that the upside of the membrane was coated with the Matrigel.

After formalin-fixed and paraffin-embedded tissue samples were treated with xylene and ethanol, antigen retrieval was performed in 1 l of citrate buffer (0.01 mol/l, pH 6.0) at 95°C for 10 min. Endogenous peroxidase activity was inactivated with 3% H₂O₂ for 10 min at room temperature (RT) and slides were blocked in 10% goat serum at RT for 30 min. The sections were then incubated with rabbit polyclonal anti-HMGA2 (1:100; Bio-Check, Inc. (Foster City, CA, USA) or rabbit polyclonal anti-STC2 (1:100; Abcam, Cambridge, MA, USA) at 4°C overnight. Subsequently, the sections were treated with secondary antibody at RT for 1 h. The slides were incubated with diaminobenzidine (DAB) and counterstained with hematoxylin. Immunohistochemical staining scoring was semi-quantitatively evaluated by staining intensity and the percentage of positive cells. The percentage positivity was graded as 0 (<5%), 1 (5–25%), 2 (25–50%), 3 (51–75%), or 4 (>75%). The staining intensity was graded as 0 (no staining), 1 (weak staining), 2 (moderate staining) or 3 (strong staining). The two grades were added together to yield the immunoreactive score (IRS). Cases with discrepancies in IRS were discussed with other pathologists until consensus was reached. Evaluation of immunohistochemical staining was carried out by two pathologists (J.-J. Wei and M. Lai) blinded to the clinicopathological characteristics.

Equal amounts (25 μg) of total proteins were resolved by 12% SDS-PAGE and transferred to polyvinylidene fluoride (PVDF) membranes. The membranes were blocked with 5% non-fat milk for 1 h at RT and incubated with primary antibodies at 4°C overnight, including rabbit polyclonal anti-STC2 (1:200; Abcam) and mouse monoclonal anti-β-actin (1:1,000; Sigma, St. Louis, MO, USA). The secondary antibody was then detected by an enhanced chemiluminescence kit (Perkin-Elmer, Waltham, MA, USA).

The upstream regions of human STC2 from −290/+43, −647/+43, −1,313/+43, −1,313/−620 were generated by PCR (Table II) and cloned into the pGL3-basic plasmid. The 293T cells were seeded in 24-well plates at densities of 7×10⁴/well. After 24 h, 0.1 μg of pRL-TK, 0.5 μg of pGL3-STC2 and the indicated amounts (0, 0.2, 0.4 and 0.8 μg) of pcDNA3.1-HMGA2, along with varied amounts (0.8, 0.6, 0.4 and 0 μg) of blank pcDNA3.1 plasmid were transfected into 293T cells by Lipofectamine 2000. Forty-eight hours after transfection, the firefly luciferase activity was measured with the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA), and results were normalized by Renilla luciferase activity. The experiments were repeated at least three times with three replicates per sample.

The associations between expression status and clinicopathological characteristics were assessed using the χ² test. Overall survival curves were calculated for the expression groups using the Kaplan-Meier method. RNA and protein expression levels were presented as means ± standard deviation from at least three independent experiments. Data were analyzed by the Student's t-test in two groups and one-way ANOVA in multiple groups. SPSS 17.0 software was used for statistical analysis. P<0.05 was considered to indicate a statistically significant result.

To examine the roles of STC2 in cell mobility, a stable STC2 overexpression was established in an ovarian T29 surface epithelial cell line (T29-STC2) (Fig. 1A). Cells with a vector control (T29-pCMV6) were used. Introduction of STC2 overexpression enhanced cell metastasis and invasion in vitro. The wound-healing assay revealed that the closure rates of T29-STC2 at 12, 36 and 48 h (36.77±1.55, 82.59±1.50 and 97.33±1.15%) were significantly higher than those of T29-pCMV6 (25.69±2.22, 61.50±1.10 and 81.45±1.13%; P<0.01, Fig. 1B). The Transwell chamber assay indicated that migration of the T29-STC2 cells (301.33±9.29) was significantly higher than that of T29-pCMV6 (174.67±6.03, P<0.001, Fig. 1C) at 24 h. In the Matrigel invasion assay, we also observed a significant induction of the invasive potential of T29-STC2 (232.67±10.69) compared to that of the control (112.33±7.57, P<0.001, Fig. 1D). These results demonstrated that STC2 overexpression promoted the migration and invasion of ovarian surface epithelial cells in vitro.

To investigate whether blocking STC2 restored or reduced the cell aggressiveness, we used the ovarian cancer cell line, Caov-3. Caov-3 exhibited abundant endogenous STC2 expression. A stable knockdown cell (Caov-3-sh-STC2) and scrambled control cell (Caov-3-pGPU6) were prepared (Fig. 2A).

The wound closure rates of Caov-3-sh-STC2 at 12, 24 and 48 h (37.35±2.50, 47.79±2.20 and 62.48±4.89%) were significantly lower than those of Caov-3-pGPU6 (56.28±3.69, 74.33±4.91 and 87.10±4.35%, P<0.01, Fig. 2B), indicating that the downregulation of STC2 reduced the locomotion of the cancer cells. Similarly, the number of cells migrating through the Transwell membrane was significantly reduced in Caov-3-sh-STC2 cells (83±4), compared with the Caov-3-pGPU6 cells (151±5.57, P<0.001, Fig. 2C). The Matrigel invasion assay revealed that blocking STC2 expression inhibited the invasiveness of Caov-3 cells (41.33±3.21 for Caov-3-sh-STC2 cells vs. 96.33±6.02 for Caov-3-pGPU6 cells, P<0.001, Fig. 2D). These results indicated that knockdown of STC2 inhibited cell migration and invasion in ovarian cancer in vitro.

Gene profiling analysis revealed that ovarian epithelial cells with HMGA2 overexpression enhanced STC2 expression (7). To determine whether HMGA2 directly regulated STC2 expression, we examined HMGA2 and STC2 expression at transcription levels using RT-PCR in 11 randomly selected HGSC samples. As shown in Fig. 3A, there was a correlation between HMGA2 and STC2 expression. We then investigated whether the increased STC2 expression was directly regulated by HMGA2 at the transcription level. The sequence analysis revealed multiple HMGA2 AT binding domains (21) along the 1,200 bp region upstream from the 5′ transcription start site of STC2. To define the genomic regions that were possibly regulated by HMGA2, we prepared several luciferase reporter constructs representing different parts of the STC2 promoter (Fig. 3B). The genomic region immediately adjacent to the STC2 transcription start site, up to −600 bp, conferred a significant increase in luciferase expression when co-transfected with the HMGA2 expression vector (Fig. 3B). Elevation in luciferase activity in response to HMGA2 co-transfection was dose-dependent (Fig. 3B). By contrast, the upstream sequence beyond this region showed minimal activity in driving luciferase expression, indicating that the HMGA2 regulatory region was mostly confined to the +1 to −600 bp promoter region of STC2. The findings suggested that HMGA2-induced STC2 upregulation was likely mediated by transcription regulation.

The abovementioned results indicated that STC2 was associated with aggressive tumor growth and was regulated by HMGA2. To examine whether STC2 expression was correlated with HMGA2 expression at the protein level in epithelial ovarian cancer (EOC), immunohistochemical analysis (Fig. 3C) was employed for measurement of the protein expressions in two cohorts of the EOC cases. The results showed a moderate correlation between HMGA2 and STC2 expression in the training cohort (R=0.544, P<0.001, Fig. 3D) and validation cohort (R=0.286, P<0.001, Fig. 3E). These findings suggested that STC2 expression was mainly regulated by HMGA2 but may also be regulated by other genes or pathways.

The clinicopathological characteristics of the training and validation cohort are shown in Table I. As shown in Tables III and IV and Fig. 4, strong immunoreactivity for HMGA2 and STC2 staining were closely associated with tumor grade. In the training cohort, the patients with higher grade had a significantly higher HMGA2 (P<0.001; Table III, Fig. 4A and C). A similar trend was observed in STC2 expression (P<0.001; Table III, Fig. 4B and D). The validation cohort results were consistent with those of the training cohort. Increasing tumor grade showed a statistically positive association with a higher expression of HMGA2 (P=0.003; Table IV, Fig. 4A and C) and STC2 (P=0.013; Table IV, Fig. 4B and D).

A strong correlation between histologic subtypes and the levels of HMGA2 and STC2 expression was identified. A higher rate of HMGA2 was detected in serous tumors as compared with that in mucinous type in the training cohort (58.6% in serous type, 26.8% in mucinous type, P<0.001; Table III, Fig. 4A and C) and validation cohort (54.6% in serous type, 23.8% in mucinous type, P=0.001; Table IV, Fig. 4A and C). A similar trend was observed in STC2 expression. A higher rate of STC2 was detected in serous tumors as compared with that in mucinous type in the training cohort (59.5% in serous type, 26.8% in mucinous type, P<0.001; Table III, Fig. 4B and D) and validation cohort (64.8% in serous type, 42.9% in mucinous type, P=0.014; Table IV, Fig. 4B and D). Statistical analysis revealed that HMGA2 and STC2 expression was not significantly associated with age, stage, or lymph node and distant metastatic status.

In 95 ovarian cancer cases with clinical follow-up data, we examined the correlation of STC2 expression with the overall survival using the Kaplan-Meier method. We found that the elevated expression of STC2 was associated with a shorter overall survival rate (P=0.016, Fig. 4E). These results indicated that STC2 may be a valuable predictive marker for prognosis in EOC, with a high STC2 expression being associated with a poor overall survival.