Human GC SGC-7901 cells were purchased from the China Center for Type Culture Collection (CCTCC; Wuhan, China) and were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 100 IU/ml penicillin and 100 µg/ml streptomycin in a humidified incubator containing 5% CO₂ at 37°C. FBS, trypsin and DMEM were purchased from HyClone (Thermo Scientific, Waltham, MA, USA).

Parental cells (PCs) were collected and washed to remove serum, and were then suspended in serum-free stem cell conditional medium containing DMEM/F12 (Gibco-Invitrogen, Carlsbad, CA, USA) supplemented with 50X B27 (Invitrogen, Carlsbad, CA, USA), 20 ng/ml EGF, 20 ng/ml bFGF (both from eBioscience, Inc., San Diego, CA, USA), 4 µg/ml insulin, 100 IU/ml penicillin G and 100 µg/ml streptomycin. After that, the cells were plated in ultra-low adherence culture plates (6-wells) at a density of 10,000 cells/ml and maintained in a humidified incubator containing 5% CO₂ at 37°C. After 5 days of culture, the first generation of sphere-forming cells (SFCs) was obtained after trypsin-EDTA digestion. The first-generation SFCs were further cultured and expanded at a density of 10,000 cells/well in ultra-low adhesion 6-well culture plates to obtain the SFCs.

The PCs and third-generation SFCs were seeded in 6-well plates at a density of 4×10⁵/well in DMEM supplemented with 10% FBS. When the cells grew to 85% confluency, the wound was generated by scratching the surface of the plates in the central region with a 200 µl pipette tip. Washing was performed 2 times using phosphate-buffered saline (PBS) to remove floating cells and debris. The cells were incubated for 48 h, and were imaged at 0 and 48 h in the same location of the wound, respectively. The numbers of cells in the scratch area were counted, and the migratory rate of the cells was determined in relation to the migratory rate of the PCs considered as the standard migration rate (100%).

DMEM (1.0 ml) supplemented with 10% FBS as a chemical inducer was added to the 24-well cell culture plate which was then embedded in the Transwell chamber. A total of 10,000 PCs or SFCs were plated in the top chamber of the Transwell coated with Matrigel. After being cultured for 24 h, the cells that had not invaded through the pores of the insert were cleared with a sterile cotton swab and discarded. The cells that invaded to the lower chamber were fixed with methanol, stained with crystal violet and counted under an optical microscope with the migration rate of the PCs or SFCs treated with 0.1% dimethyl sulfoxide (DMSO) considered as the standard invasion rate (100%).

The cells were washed with PBS once, and lysed in 1 ml lysis enzyme buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.2 mM EDTA, 0.2% NP-40, 10% glycerol, 1 M β-Me, 1 µg/ml Trasylol, 0.5 µg/ml leupeptin, 0.1 mM Na₃VO₄, 0.5 mM 4-NPP, 0.5 mM NaF and protease inhibitors]. The cells were scraped and collected after incubation for 20 min at 4°C. The lysates was centrifuged at 13,200 rpm for 5 min at 4°C to prepare whole cell extracts. The Bradford assay (Bio-Rad Laboratories, Hercules, CA, USA) was used to determine the protein content. The proteins were separated and extracted using 10% SDS-polyacrylamide gel electrophoresis, and transferred to a polyvinylidene difluoride (PVDF) membrane (Millipore, Billerica, MA, USA). The membranes were detected using mouse antibodies against CD133 (Santa Cruz Biotechnology, Santa Cruz, CA, USA), CD44, ALDH1 (both from Cell Signaling Technology, Danvers, MA, USA), N-cadherin (Upstate Biotechnology, Inc., Lake Placid, NY, USA), E-cadherin (BD Transduction), FoxM1 (Santa Cruz Biotechnology), Twist1 (Cell Signaling Technology) and β-actin (Sigma, St. Louis, MO, USA), respectively.

Data are presented as the mean ± SE (mean ± SD) and were analyzed by SPSS 17.0 statistical software. Multiple comparisons were performed by one-way ANOVA and pair-wise comparison was conducted by the LSD t-test method. The Dunnett's method was used for unequal variances. P<0.05 as considered to indicate a statistically significant result.

Compound 1 was synthesized and obtained using methods from the patent application (31) and was identified by a combination of NMR and mass spectral data and by comparison of these to the published literature.

Compound 1, yellow powder; EI-MS, m/z 544.1; ¹H NMR (500 MHz, CDCl₃): 0.88–0.92 (6H, m), 1.26–1.55 (24H, m), 1.77–1.83 (2H, m), 1.90–1.96 (2H, m), 3.98 (2H, J=6.5 Hz), 4.06 (2H, J=6.5 Hz), 6.53 (1H, d, J=2.0 Hz), 6.65 (1H, t, J=72.5 Hz), 6.69 (1H, d, J=2.0 Hz), 6.94 (2H, d, J=8.5 Hz), 7.44 (2H, d, J=8.5 Hz), 7.77 (1H, s); ¹³C NMR (125 MHz, CDCl₃): 14.1, 22.6, 22.7, 25.8, 26.0, 28.8, 29.1, 29.2, 29.3, 29.4, 29.5, 31.8, 68.0, 69.9, 98.3, 99.1, 112.6, 113.2, 114.4, 115.2, 117.3, 123.6, 126.5, 130.3, 150.3, 154.8, 159.0, 159.1, 161.4 and 175.2. The ¹H-NMR data were consistent with the literature (31) and ¹³C-NMR data were reported for the first time. Thus, compound 1 was identified as 7-difluoromethoxyl-5,4′-di-n-octyl genistein and named DFOG (Fig. 1).

In order to enrich GCSLCs from human GC SGC-7901 cells, a stem cell conditioned medium suspension culture method was used. Under these conditions, the cells grew as non-adherent, three-dimensional sphere clusters. Fig. 2A shows the anchorage-independent spheres that formed in the SGC-7901 cells. After 5 days of incubation, the SFCs from the SGC-7901 cell line were found to generate more and larger spheroid colonies compared with that noted in the PCs (Fig. 2B and C).

Next, western blotting was performed to identify protein expression of the gastric CSC markers [cluster of differentiation CD133, CD44 and aldehyde dehydrogenase 1 (ALDH1)]. The results showed enrichment of CD133⁺, CD44⁺ and ALDH1-high populations in the SFCs derived from the SGC-7901 cells compared with the PCs (Fig. 2D). These results indicated that SFCs from the SGC-7901 cells cultured in stem cell-conditioned medium possessed GCSLC properties.

CSCs have higher migratory and invasion capacities, which facilitate metastasis and growth. The migration and invasion capabilities of GCSLCs and PCs were evaluated by scratch method and Transwell chamber invasion assay in vitro, respectively. The results demonstrated that GCSLCs showed increased migratory and invasive capabilities than these capacities noted in the PCs (Fig. 3A and B). CSCs are also thought to facilitate metastasis through EMT characteristics related to the mobility of cells. We evaluated the protein expression of a known mesenchymal phenotype cell biomarker (N-cadherin) and an epithelial phenotype cell biomarker (E-cadherin) by western blot analysis. The results demonstrated that the relative protein level of N-cadherin was highly expressed in the GCSLCs, while that of E-cadherin was weakly expressed (Fig. 3C).

Tumor sphere assay is used to identify stem cells in in vitro assays. We examined the tumor sphere formation capacity of SGC-7901 cells following treatment of DFOG. The results showed that DFOG (1.0, 3.0 and 10.0 µmol/l) reduced the number of SFCs derived from the SGC-7901 cells in a concentration-dependent manner (Fig. 4), indicating that DFOG is able to preferentially suppress the self-renewal of GCSLCs.

CD133, CD44 and ALDH1 are used as cancer stem cell markers in many types of tumor cells including GC. To investigate the effect of DFOG on GCSLC surface marker expression including CD44, CD133 and ALDH1, we incubated GCSLCs with DFOG (1.0, 3.0 and 10.0 µmol/l) and DMSO as a control. The expression levels of CD44, CD133 and ALDH1 in the GCSLCs were significantly suppressed following treatment with DFOG compared with the spheres that were untreated (Fig. 5A). A previous study demonstrated that overexpression of FoxM1 led to EMT by the acquisition of the EMT phenotype and downregulation of FoxM1 led to the inhibition of EMT in GC cell lines (21). Therefore, the present study compared the level of FoxM1 protein expression in the PCs and SFCs and evaluated the inhibitory effect of DFOG on FoxM1. The results revealed that FoxM1 expression was higher in the SFCs in comparison with that in the PCs (Fig. 5B). Furthermore, as shown in Fig. 5C, the expression of FoxM1 in the SFCs was significantly reduced by DFOG in a dose-dependent manner.

We next transfected FoxM1 siRNA into the GCSLCs to confirm the inhibitory effect of DFOG on FoxM1 expression in GC cells and the effect of FoxM1 on the characteristics of GCSLCs. The GCSLCs were transfected with either scramble siRNA or FoxM1 siRNA. The expression levels of FoxM1 and CSC markers including CD133, CD44 and ALDH1 were assessed using western blot analysis. The results showed that the protein expression level of FoxM1 was inhibited following FoxM1 knockdown (Fig. 6A). Furthermore, after transfection of FoxM1 siRNA, the inhibitory effects of DFOG on the expression of FoxM1 and CSC markers (CD133, CD44 and ALDH1) were markedly enhanced compared with the scramble siRNA control group (Fig. 6B and C). The knockdown of FoxM1 also suppressed the sphere-forming capability of the GCSLCs and suppressed the inhibition of the self-renewal of GCSLCs synergistically together with DFOG (Fig. 6D).

To investigate the effects of DFOG on the EMT process of GCSLCs, the protein levels of E-cadherin, N-cadherin and the Twist1 (EMT-related transcription factors) were measured. The results from the western blot analysis demonstrated that DFOG upregulated the protein level of E-cadherin and downregulated the protein level of N-cadherin (Fig. 7A). The protein expression level of Twist1 in the GCSLCs was higher than that in the PCs (Fig. 7B), while DFOG also dose-dependently suppressed the expression of Twist1 (Fig. 7C). These results clearly demonstrated that DFOG could reverse EMT, relying on inhibition of the EMT phenotypic biomarkers and Twist1. Migration and invasion abilities are important characteristics of CSCs responsible for tumor metastasis and growth. CSCs are assumed to have higher migration capacity than normal cancer cells and Twist1 is closely associated with cancer cell migration and invasion. Functionally, the relative migratory and invasive cell numbers of GCSLCs were significantly decreased when compared to these numbers in the negative control, suggesting that DFOG reduced cell migration and invasion in the GCSLCs (Fig. 7D and E). The above results showed that DFOG inhibited the expression of Twist1, leading to the inhibition of cancer cell migration and invasion.

To examine whether forced knockdown of FoxM1 can cooperate with DFOG to reverse EMT, the protein expression of EMT biomarkers (E-cadherin and N-cadherin) and Twist1 was measured by western blot analysis. As compared to the negative control groups, silencing by FoxM1 siRNA increased the protein expression of E-cadherin (Fig. 8A) and decreased the protein expression of N-cadherin and Twist1 (Fig. 8B and C), suggested that knockdown of FoxM1 could reverse EMT. Moreover, transfection of FoxM1 siRNA also enhanced the upregulation of E-cadherin protein expression (Fig. 7A) and the downregulation of N-cadherin and Twist1 protein expression (Fig. 8B and C) caused by DFOG, demonstrating that transfection of FoxM1 siRNA can cooperate with DFOG to reverse the EMT process. Following transfection of FoxM1 siRNA, the influence on the migratory and invasive capabilities of the GCSLCs were evaluated. The results showed that transfection of FoxM1 siRNA suppressed the migration and invasion of GCSLCs and also enhanced the inhibitory effect of DFOG on the migration and invasion of GCSLCs compared with the control group (Fig. 8D and E). These results confirmed that transfection of FoxM1 siRNA and DFOG can synergistically reverse the EMT process of GCSLCs.