Five well-characterized human GC cell lines (SNU-638, SNU-601, SNU-216, SNU-484, and SNU-668) were purchased from the Korean Cell Line Bank (Seoul, Korea). Cells were maintained in RPMI 1640 medium (Life Technologies, Grand Island, NY) containing 10% fetal bovine serum, incubated in 5% CO₂, and either 20% (normoxic) or 1% (hypoxic) O₂ atmosphere at 37°C.

Western blot analysis was performed as we described previously [10]. Cell lysates in sodium dodecyl sulfate (SDS) lysis buffer (125 mM Tris-HCl [pH 6.8], 4% SDS, 0.004% bromophenol blue, and 20% glycerol) were separated on 10% SDS–polyacrylamide gel and electrophoretically transferred to PVDF membranes (Millipore Co., Billerica, MA) blocked with 5% non-fat dry milk in phosphate buffered saline– Tween-20 (0.1%, vol/vol) for 1 hour. The membranes were then incubated with a primary antibody against FOXO1 (1:1,000, C29H4, Cell Signaling Technology, Beverly, MA), hypoxia inducible factor-1α (HIF-1α; 1:500, H1α67, BD Biosciences, San Jose, CA), vascular endothelial growth factor (VEGF; 1:1,000, C1, Santa Cruz Biotechnology, Santa Cruz, CA), SIRT1 (1:1,000, H300, Santa Cruz Biotechnology), or β-actin (1:1,000, C4, Santa Cruz Biotechnology). Horseradish peroxidase-conjugated anti-rabbit IgG (1:2,000, Santa Cruz Biotechnology) or anti-mouse IgG (1:2,000, Santa Cruz Biotechnology) was used as a secondary antibody. Enhanced chemiluminescence (Amersham, Arlington Heights, IL) was used to detect the immunoreactive proteins. Equal protein loading was confirmed by β-actin.

Lentiviral particles containing FOXO1 short hairpin RNA (shRNA) or non-targeting shRNA, control shRNA, were purchased (Sigma, St. Louis, MO). The sequence of FOXO1 shRNA was 5′-CCGGGCCTGTTAT CAATCTGCTAAACTC GAGTTTAGCAGATTGAT AACAGGCTTTTTG-3′. The control shRNA particles contain four base pair mismatches within the short hairpin sequence to any known human or mouse gene. Viral infection was introduced by incubating GC cells in the culture medium containing lentiviral particles for 12 hours in the presence of 5 μg/mL Polybrene (Santa Cruz Biotechnology). Pooled puromycin (2 μg/mL)-resistant cells were used for further analysis.

To determine FOXO1 nuclear DNA-binding activity, luciferase reporter assay was used as previously described with slight modifications [10,11]. Two oligonucleotides (GCAAAACAAACTTATTTTGAAGCAAAACAAACT TATTTTGA AGCAAAACA AACT TATTTTGAA and TTCA AAATAAGTTTGTTTTGCTTCAAAATAAGTTTGTTTTG TT CAAAATAAGTTTGTTTTGC) were annealed and ligated into the pGL4.27-Promoter vector (Promega, Madison, WI) to create 3XIRS-luciferase (Cosmo GENETECH, Seoul, Korea). This construct has three tandem repeats of a FOXO1 binding element, the insulin-responsive sequence (IRS), inserted upstream of the luciferase reporter gene. It is widely used as an indicator of FOXO1 transcriptional activity. GC cells were cotransfected transiently with 0.4 μg 3XIRS-luciferase vector and 0.4 μg pSV-β-galactosidase vector (Promega), an internal control, using Lipofectamine Plus (Life Technologies). Twenty-four hours after transfection, assays for luciferase and β-galactosidase were carried out using a Dual-Luciferase Reporter Assay System (Promega). FOXO1 luciferase activity was measured on an AutoLumat LB 9505c luminometer (Berthold Analytical Instruments, Nashua, Germany) and was normalized by β-galactosidase activity.

SNU-638 (2.5×10⁴ cells) and SNU-601 cells (2.5×10⁴ cells) were seeded into each well of 24-well plates and were allowed to grow for 0 to 24 hours. Cell numbers were measured indirectly using the method reported by Kim et al. [12]. Cells were stained with 0.2% crystal violet aqueous solution in 20% methanol for 10 minutes, dissolved in 10% SDS, transferred into 96-well plates, and the absorbance was measured at 570 nm using an enzyme-linked immunosorbent assay reader (Bio-Rad, Hercules, CA).

All animal procedures were performed in accordance with the procedures described in the Seoul National University Laboratory Animal Maintenance Manual (approval No. SNU-120412-5). Six-week-old male nude mice (BALB/cSlcn/n) were purchased from SLC Inc. (Hamamatsu, Shizuoka, Japan) and maintained under specific pathogen-free conditions. For subcutaneous implantation, tumors were established by injecting SNU-638 GC cells expressing either control shRNA or FOXO1 shRNA at the density of 5×10⁶ cells in 100 μL of Matrigel (R&D Systems, Minneapolis, MN) subcutaneously into the left flank of each mouse. Mice were divided into two groups by FOXO1 expression in GC cells: control shRNA or FOXO1 shRNA. Tumor volumes were measured every two days using a caliper, and were calculated by the equation V (mm³)=(length×width×height)× (π/6). Animals were sacrificed 48 days after cell implantation, and tumor xenografts were removed and prepared for immunohistochemistry or Western blot analysis.

Four hundred and seventy-one surgically resected human GC specimens were obtained from the Department of Pathology, Seoul National University College of Medicine from January 2 to December 29, 2006. Nine paraffin tissue array blocks were prepared as previously described [13]. This protocol was reviewed and approved by the Institutional Review Board of the Seoul National University (approval No. C-1006-035-320).

Tissue specimens from clinical GC samples and xenograft tumors derived from GC cells were fixed with 10% neutralbuffered formalin, and 4-μm paraffin sections were prepared. After rehydration, sections were stained with hematoxylin and eosin for histologic assessment, or were immunostained after antigen retrieval using a Bond-max automated immunostainer (Leica Microsystems, Newcastle upon Tyne, UK). The primary antibodies used were against FOXO1 (1:40, C29H4, Cell Signaling Technology), phospho-FOXO1^Ser256 (pFOXO1; 1:60, Cell Signaling Technology), CD31 (1:100, M20, Santa Cruz Biotechnology), HIF-1α (1:50, provided by Dr. Jong-Wan Park at Seoul National University), VEGF (1:200, C1, Santa Cruz Biotechnology), and SIRT1 (1:100, H300, Santa Cruz Biotechnology). Antibody binding was detected with the Bond Polymer Refine Detection Kit (Leica Microsystems). All immunostained sections were lightly counterstained with Mayer’s hematoxylin. Throughout the above analysis, negative controls were prepared by omitting the primary antibody. The results of immunostaining were evaluated by two pathologists (Y.K. and J.-S.P.), who were blinded to the origin of the samples. For statistical analysis, the results of immunostaining for proteins were considered positive if immunoreactivity was seen in ≥ 10% (cytoplasmic pFOXO1 and nuclear SIRT1) or ≥ 5% (nuclear HIF-1α) of the neoplastic cells.

Microvessel areas (MVAs) in subcutaneous xenograft tumors were determined by light microscopy/optical image analysis after immunostaining xenograft tumor sections with anti-CD31 as described previously [10]. The three most highly vascularized areas in areas of tumors near the tumor– normal tissue interface were selected. Photographs of CD31-immunopositive vessels in tumor sections were taken under a light microscope, and the cross-sectional areas of CD31-immunopositive structures (i.e., microvessel areas) were quantified by capturing images, converting them to grey scale, and analyzing CD31-stained areas using NIH Image Analysis software after setting one consistent intensity threshold for all slides. CD31-positive areas were expressed as pixels squared per high-power field and were measured for all tumors.

For cell culture and animal experiments, data were analyzed using GraphPad Prism software for Windows 7 (ver. 4, GraphPad Software, San Diego, CA). A two-tailed Student’s t test was used to rule out the role of chance in the results. For tissue array analysis, statistical analyses were conducted using SPSS ver. 19.0 (IBM Co., Chicago, IL), and the chi-squared test was used to determine the relationship between the expressions of pFOXO1^Ser256, SIRT1, and HIF-1α. Results were expressed as mean value±standard deviation. Null hypotheses of no difference were rejected if p-values were less than 0.05.

To investigate the molecular effects of FOXO1 in GC, we initially examined the expression level of FOXO1 in GC cell lines. The protein contents of FOXO1 varied in five different cell lines (Fig. 1A). To observe the effect of FOXO1 silencing in the GC cell, SNU-638 and 601 cell lines with high levels of FOXO1 expression were selected, and FOXO1 expression was suppressed using RNA interference. Western blot analysis (Fig. 1B) and luciferase reporter assay (Fig. 1C) verified that transfection of FOXO1 shRNA effectively inhibited FOXO1 expression and its transcriptional activity in both cell lines.

HIF-1α, which regulates many genes involved in adapting to a hypoxic environment, is important in promoting gastric tumor growth and angiogenesis [14]. Therefore, we performed cell culture experiments using GC cell lines SNU-638 and SNU-601 to examine the direct relationship between the expressions of FOXO1 and HIF-1α. GC cells expressing control shRNA or FOXO1 shRNA were cultured under normoxic or hypoxic conditions. By Western blot, FOXO1 down-regulation corresponded with increased HIF-1α protein expression under hypoxic conditions, but not under normoxic conditions, in both cell lines (Fig. 1D).

Fig. 1A shows that FOXO1 expression was higher in SNU-601 cells than in SNU-638 cells. In contrast, we found that cell growth rate was higher in SNU-638 cells than in SNU-601 cells, which became more evident under hypoxic conditions (Fig. 2A). To confirm the growth-inhibitory role of FOXO1, we compared the viability of GC cell lines under normoxic and hypoxic conditions. At 24 hours of hypoxic exposure, in both GC cell lines FOXO1 knockdown increased cell viability under hypoxic conditions, but not under normoxic conditions (p < 0.05) (Fig. 2B and C). These data suggest that the effect of FOXO1 on GC cell growth needs a hypoxic environment.

Fig. 2A shows that SNU-638 cells grow faster than SNU-601 cells. Thus, to examine the role of FOXO1 in GC growth, we injected SNU-638 cells rather than SNU-601 cells into nude mice. Stable SNU-638 cells expressing non-targeting (control) or FOXO1-targeting shRNA were used. We found that tumors were successfully formed in all mice. The tumor weights were markedly higher in mice injected with FOXO1 shRNA expressing cells than those injected with control shRNA expressing cells (p < 0.05) (Fig. 3A).

Next, tumor tissues were removed and prepared for immunohistochemistry (Fig. 3B, left) and Western blot analysis (Fig. 3C). FOXO1 shRNA tumors showed much lower FOXO1 protein expression than the control shRNA tumors (Fig. 3B, top row), verifying the direct inhibitory role of FOXO1 in tumor growth. Angiogenesis in xenografted tumors was assessed using immunohistochemistry for CD31, a specific marker of vascular endothelial cells. CD31 immunoreactivity increased in FOXO1 shRNA tumors compared to control shRNA tumors (Fig. 3B, second row). Optical image analyses showed that the CD31-positive MVA was noticeably larger in FOXO1 shRNA tumors (mean, 18.56×10³ pixels²) than in control shRNA tumors (mean, 3.25×10³ pixels²) (Fig. 3B, right). Thus, FOXO1 silencing increased MVA in GC xenografts. In addition, FOXO1 shRNA increased the expression of HIF-1α and VEGF, both of which promote GC tumor growth [15], angiogenesis [15] and metastasis [16] (Fig. 3B, third and fourth rows; Fig. 3C, second and third rows). These results strongly suggest that FOXO1 functions to inhibit tumor growth through negative regulation of tumor angiogenesis mediated by the HIF-1α– VEGF pathway.

In addition, the association found between FOXO1 and SIRT1 was observed in xenografted tumors. Western blot analysis showed that FOXO1 shRNA-tumors showed higher SIRT1 expression than control shRNA-tumors (Fig. 3C).

SIRT1 increases gastric tumor growth and angiogenesis [17]. Thus, we examined the relationship between the expressions of FOXO1 and SIRT1. In immunohistochemical tissue array analysis of human GC samples, pFOXO1 (Fig. 4A and D), SIRT1 (Fig. 4B and E), and HIF-1α (Fig. 4C and F) were expressed in both the nuclei and cytoplasm of tumor cells. Regarding pFOXO1 expression, cells showing distinct cytoplasmic staining, with or without the presence of nuclear staining, were considered to express the inactive form of FOXO1 constitutively. With SIRT1 and HIF-1α expressions, nuclear staining was considered to indicate activation of these two molecules. We analyzed the association between FOXO1 inactivation (pFOXO1 expression) and the activations of SIRT1 and HIF-1α in 471 human GC specimens. Table 1 shows that cytoplasmic pFOXO1 expression was found in 359 (77%), nuclear SIRT1 expression in 178 (38%) and nuclear HIF-1α expression in 108 of the 471 (23%) of human GC cases. This result shows positive associations between pFOXO1 and SIRT1 (p < 0.001) and HIF-1α (p < 0.001). Moreover, a positive association was found between SIRT1 and HIF-1α (p < 0.001).