Different human cancer cell lines LoVo, HCT-116, LS174T, SW-620, SW-480, U87, U251, KBV, A549, MDA-231, HeLa, CaSki, SiHa, and MG-63 were obtained from the Research Center of Clinical Medicine, Nanfang Hospital (Guangzhou, China) and were maintained in Dulbecco's modified Eagle's medium (DMEM) containing 10% fetal bovine serum (FBS) (both from HyClone, Logan, UT, USA) in a humidified incubator with 5% CO₂ at 37°C.

Colorectal cancer tissue specimens were obtained from 68 patients who received surgery for tumor resection in Nanfang Hospital between December 2011 and January 2013. These patients did not receive any treatments before surgery and all cases were histologically diagnosed and retrospectively reviewed by two pathologists in the present study. The present study was approved by the Human Ethics Committee of Nanfang Hospital. In the present study, we collected paired normal and colorectal cancer tissue specimens and clinicopathological data from each patient. Tissue samples were stored in liquid nitrogen until use according to our previous study (29). Among these patients, 24 patients had rectal cancer and 44 had colon cancer, while 40 were at early stages of disease without tumor metastasis while 28 had different degrees of tumor metastasis according to the standard of the National Comprehensive Cancer Network (NCCN). The clinicopathological characteristics of the patients are presented in Table I.

Different cancer cell lines were grown and then collected from cell culture plates after being washed with phosphate-buffered saline (PBS) 3 times and lysed in a lysis buffer containing 50 mM Tris (pH 7.4), 150 mM NaCl, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate and phosphorylase and metalloproteinase inhibitor (Beyotime, Inc., Beijing, China). Colorectal tumor tissues were grinded on ice and then lysed in the same lysis buffer. Both types of cell lysis were then centrifuged and the concentration of protein samples was measured using the BCA protein assay kit (Beyotime). The protein samples were then treated with 5X loading buffer and boiled at 100°C for 5 min, separated using sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) on 12% SDS-PAGE gels and transferred onto polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA). The membranes were then blocked in 5% skim milk solution in PBS and then incubated with different primary anti-human antibodies at 4°C overnight and the secondary antibody and enhanced chemiluminescence (ECL) kit according to a standard protocol. The primary antibodies were anti-GAPDH (1:1,000), anti-GMFG (1:1,000) (both from ProteinTech Group, Wuhan, China), anti-MMP2 (1:1,000), anti-E-Cad (1:1,000) and anti-vimentin (1:1,000) (all from Abcam, Cambridge, MA, USA).

Paraffin-embedded tissue specimens were sectioned into 4-µm thick sections and put onto 3-aminopropyltriethoxysilane (APES)-coated glass slides. For immunohistochemistry, the sections were deparaffinized in xylene and rehydrated in a series of ethanol solution and submerged in tap-water. The antigen retrieval was then performed using an antigen unmasking solution, which was preheated in a microwave for 5 min and the sections in the buffer were heated at highest energy levels for 8 min in a microwave and cooled down to room temperature. Potential endogenous peroxidase activity was blocked in 3% hydrogen peroxide for 20 min and the sections were incubated in 5% casein in PBS for 1 h to block non-specific antibody binding. The sections were then incubated at 4°C overnight with a rabbit anti-GMFG antibody (ProteinTech) at a dilution of 1:100 and subjected to a post primary block using the polymer penetration enhancer (Boshide, Wuhan, China) for 30 min. Sections were then washed with PBS 5 min each for 3 times and further incubated at room temperature for 30 min with an anti-rabbit IgG-Poly-HRP (Merck, Kenilworth, NJ, USA). The sections were visualized with 3,3′-diaminobenzidine (DAB) solution (Boshide) and briefly counterstained with hematoxylin solution.

The immunostained sections were reviewed and scored by two investigators under a light microscope for intensity and percentage of staining. Protein localization was evaluated and classified as nuclear, cytoplasmic and plasmalemma locations. The percentage of positively stained tumor cells was graded as 0–10% (1+), 11–50% (2++) and >50% stained cells (3+++), while the staining intensity was graded as 1 (weakly positive), 2 (positive) and 3 (strongly positive). Subsequently, an immunostaining index was calculated by multiplying the staining intensity with the percentage of positive cells to reach negative (0), weak (1+), medium (2++) and strong (3+++) expression of GMFG (Table I).

Total RNA was isolated from cells using TRIzol reagent (Invitrogen, Carlsbad, CA, USA) and reversely transcribed into cDNA using 5X PrimeScript RT Master Mix (Takara, Dalian, China) according to the manufacturers protocols. After that, these cDNA samples were subjected to qPCR amplification using 2X SYBR Premix Ex Taq (Takara) in the 7500 ABI Real-Time PCR System (Applied Biosystems, Foster City, CA, USA). The amplification conditions were 95°C for 30 sec, 95°C for 5 sec, and 60°C for 34 sec for 40 cycles and the relative level of mRNA expression was calculated using the 2^−ΔΔCt method after normalization to the level of GAPDH mRNA. The primers used were: GMFG, 5′-CGC GGG AAG TAA AAA CAGG C-3′ and 5′-GGT CTC GTT GAG GTC GTC TG-3′; and GAPDH, 5′-AGA AGG TGG GGC TCA TTT G-3 and 5-AGG GGC CAT CCA CAGT CTT C-3′.

To knockdown GMFG expression in colorectal cancer cells, we used GMFG siRNA purchased from Ruibo Co. (Guangzhou, China) and transfected GMFG siRNA or negative control (NC) siRNA into the colorectal cells for 24 h using Hylimax (Dojindo, Kumamoto, Japan) in Opti-MEM (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer's instructions and assayed. The NC siRNA was random siRNA duplex sequences.

For the migration assay, 5×10⁴ cells in a serum-free medium were seeded onto an uncoated filter with 8.0-µm pores (Corning, Corning, NY, USA), whereas 1×10⁵ cells were seeded onto a filter with 8.0 mm pores precoated the extracellular matrix (BD Biosciences, Franklin Lakes, NJ, USA) for the invasion assays. The bottom chambers were filled with 10% FBS and cultured for 24 h. In separated experiments, different concentrations of recombinant GMFG (0, 5, 10, 20 and 30 µg/ml) were added into the bottom chambers. At the end of the experiments, cells on the surface of the filters were removed using cotton swabs and PBS, while cells that had migrated or invaded into the low surface of the filters were fixed and stained with crystal violet (Beyotime). The numbers of migrated and invaded cells were counted under a microscope for 10 of 20x microscopic fields and averaged.

The level of GMFG in cell culture medium was measured using the GMFG ELISA kit (Cusabio Biotech Co., Ltd., Wuhan, China) according to the manufacturer's protocol. GMFG levels were expressed as mean ± SD of triplicate experiments and repeated at least 3 times.

Statistical analyses were performed using the SPSS 13.0 statistical software (SPSS, Chicago, IL, USA). The data are expressed as mean ± SD and were statistically analyzed using one-way ANOVA and χ² test. P<0.05 indicates statistically significant results.

In the present study, we first detected GMFG expression in 14 different cancer cell lines and found that GMFG was expressed in the majority of the cell lines (Fig. 1A). GMFG protein was preferentially expressed in U87, U251, KBV, MDA-231, SiHa and LoVo cells, while colorectal cancer cell line LoVo expressed the highest level of GMFG compared with the other cell lines (Fig. 1B).

We then analyzed GMFG expression in 68 colorectal cancer surgical samples and found that 42 samples (61.8%) expressed a high level of GMFG protein using western blotting compared with the adjacent non-tumor tissues, while 14 samples (20.6%) showed no obvious difference, whereas 12 samples (17.6%) had lower levels of GMFG compared with normal tissues (Fig. 2A).

Immunohistochemical data showed that GMFG was highly expressed in colorectal tissues compared to normal tissues (Fig. 2B). GMFG protein was mainly localized in mesenchymal cells of the non-tumor colorectal tissues and in the gland cells in the tumor tissues. We then associated GMFG expression with clinicopathological parameters from these patients and found that GMFG was associated with lymph node metastasis (85.7 vs. 57.5% in non-metastatic cases; Table I), but there was no statistically significant difference in GMFG expression associated with patient gender and age and tumor differentiation (Table I).

We then assessed the effects of GMFG knockdown in colorectal cancer cells using GMFG siRNA. Our data showed that GMFG siRNA significantly silenced GMFG expression in the LoVo cells (Fig. 3A and B). We then performed the Transwell assay to assay the altered ability of tumor cell migration and invasion, and found that both tumor cell migration and invasion capacities were reduced after knockdown of GMFG expression (Fig. 3C and D).

We further determined whether addition of exogenous GMFG could affect LoVo cell migration. As shown in Fig. 4A, the number of LoVo cells was increased after addition of exogenous GMFG and the maximal GMFG concentration was 20 µg/ml, although the level of LoVo cell migration did not obviously increase as GMFG concentration continued to rise.

As discussed in the Introduction section, GMFG can promote angiogenesis. We, thus, co-cultured LoVo cells with human umbilical vein endothelial cells (HUVECs). Our data showed that the migration capacity of the LoVo cells was increased after being co-cultured with the HUVECs (Fig. 4B). However, addition of the GMFG antibody into this co-culture downregulated LoVo cell migration (Fig. 4C). Moreover, knockdown of GMFG expression in HUVECs and then co-culture with LoVo cells showed a decrease in LoVo cell migration compared with the NC group (Fig. 4D). However, we failed to detect GMFG levels in cell culture medium in our cancer cell lines by using GMFG ELISA kit (data not shown).

Western blot analysis revealed that knockdown of GMFG expression downregulated expression of E-Cad and MMP2, but upregulated the vimentin level in the LoVo cells (Fig. 5), indicating that GMFG promoted tumor cell EMT while silencing of GMFG reversed the EMT phenotype in the LoVo cells.