Purified rabbit AMF was purchased from Sigma for exogenous AMF stimulation (cat. no P9544). Mouse monoclonal anti-AMF (ab66340), rabbit polyclonal anti-AMF (ab86950), E-cadherin (ab15148), vimentin (ab45939), Snail (ab180714), TGFBR1 (abab31013) and anti-β-actin (ab8227) for use in immunohistochemistry or western blot analyses were obtained from Abcam Ltd. (Hong Kong, China). U0126 (MAPK inhibitor) was purchased from Selleck Chemicals.

Ishikawa and HEC-1B cells were purchased from the American Type Culture Collection (ATCC; Manassas, VA, USA) and maintained according to the provider's instruction in DMEM/F12 media (Gibco, Auckland, New Zealand) supplemented with 10% fetal bovine serum (FBS; Gibco, Carlsbad, CA, USA). All cells were grown until confluent and were incubated in serum-free medium for 24 h before treatment with various experimental agents as described.

High-density tissue microarrays were constructed by Outdo Biotech Co., Ltd., (Shanghai, China) using clinical samples obtained from a cohort of 62 patients who underwent surgery at the Shanghai First People's Hospital. Each array includes 31 normal and 31 malignant tissue specimens of the endometrium (various grades and stages). Paraffin-embedded tissue sections (5 μm) were deparaffinized by xylene and rehydrated in a graded alcohol series (100, 95, 80 and 70%, 5 min each). Antigen retrieval was done by boiling the slides in a small beaker filled with 1 mM EDTA (pH 8.0). Endogenous peroxidase activity was quenched by a 10-min incubation in 3% hydrogen peroxide. After antigen retrieval, the slides were washed two times in a 0.1% Tween/1X TBS (0.1% TBST) bath (5 min each) at room temperature to remove non-specific background binding. Protein blocking was performed by incubating the specimens in 5% normal rabbit serum or normal horse serum in 0.1% TBS for 1 h. Primary antibodies against AMF, E-cadherin or Snail and vimentin were diluted 1:200 and 1:500, respectively and applied for 1 h at room temperature. After a series of TBST rinses as described above, bound antibody was subsequently detected using EnVision reagents (Wuhan Boster Biological Engineering Co., Ltd., Wuhan, China) according to the manufacturer's instructions.

Immunostained slides were scored under a microscope. The staining intensity was scored as 0 (negative), 1 (weak), 2 (medium) or 3 (strong). The extent of staining was scored as 0 (0%), 1 (1–25%), 2 (26–50%), 3 (51–75%) or 4 (76–100%), according to the percentage of the positively stained areas in relation to the whole tumor area. The sum of the intensity score and the extent score was used as the final staining score (0–7). Tumors with a final staining score of 4 or higher were considered ‘positive' (24). Results were assessed by two pathologists in a blinded manner with respect to the tissue source.

AMF shRNA constructs were cloned into pLKO.1 plasmid under the control of U6 promoter for stable expression (Sigma). Three pairs of annealed DNA oligonucleotides were inserted into the AgeI and EcoRI restriction sites of pLKO.1. The most effective pairs of sequence targeted to human AMF is: sense, 5′-CGCCATGTATGAGCACAAGAT-3′ and anti-sense 5′-GCGGTACATACTCGTGTTCTA-3′ (shAMF-1), as well as sense 5′-CCTGTCTACTAACACAACCAA-3′ and antisense 5′-GGACAGATGATTGTGTTGGTT-3′ (shAMF-2), respectively. Ishikawa and HEC-1B cells were infected with lentiviral vector pLKO.1 (mock) or AMF-specific shRNA lentiviral particles in 6-well plates in the presence of polybrene (6 mg/ml) and then treated with puromycin (2 mg/ml) to generate stable clones. Puromycin-resistant AMF knockdown clones were harvested by ring selection, and AMF gene expression and protein level were confirmed by qRT-PCR and immunoblotting.

Total RNA was isolated using TRIzol reagent (Invitrogen, Life Technologies; Shanghai, China) and reverse transcribed using a reverse transcriptase kit (Takara, Dalian, China). Gene expression was detected with SYBR-Green Master Mix (Takara) on an ABI Prism 700 Thermal Cycler (Applied Biosystems, Foster City, CA, USA). Gene expression was calculated using the 2^−ΔΔCt formula and normalized against β-actin. The oligonucleotide primers were 5′-CGCCCAACCAACTCTATTG-3′ (forward) and 5′-GAT GATGCCCTGAACGAAG-3′ (reverse) for human AMF detection; 5′-TTATGATTCTCTGCTCGTG-3′ (forward) and 5′-ATAGTCCTGGTCTTTGTCT-3′ (reverse) for human E-cadherin detection; 5′-AGGAGGAAATGGCTCGTCA-3′ (forward) and 5′-TGTAGGTGGCAATCTCAAT-3′ (reverse) for human vimentin detection; 5′-CGGCTCCTTCGTCCT TC-3′ (forward) and 5′-GCACCCAGGCTGAGGTATT-3′ (reverse) for human Snail detection; 5′-TGTGAAGCCTTGAG AGTAA-3′ (forward) and 5′-TGTTGACTGAGTTGCGATA-3′ (reverse) for human TGFBR1 detection; 5′-CAGCCATGT ACGTTGCTATCCAGG-3′ (forward) and 5′-AGGTCCAGA CGCAGGATGGCATG-3′ (reverse) for human β-actin detection (as a housekeeping gene). All experiments were performed independently in triplicate.

For whole-cell lysates, cells were washed twice with PBS and collected by scraping. Cell pellets were lysed in cold precipitation assay buffer [20 mM Tris-HCl (pH 7.4), 150 mM NaCl, 10 mM EDTA, 1% NP40, Triton X-100, sodium deoxycholate] containing 1 mmol/l DTT, 1 mmol/l phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin and 10 μg/ml aprotinin. Samples were separated by centrifugation (15,000 rpm in 4°C for 30 min). Lysate supernatants were 100-fold concentrated with Amicon Ultra (30,000 nominal molecular weight limit; Millipore).

The extracted protein concentration was measured with BCA Protein assay kit (Thermo Fisher Scientific). Equal amounts of the proteins (20 μg of cellular protein or 80 μg of secreted protein) were separated on 10% SDS-PAGE gels and transferred to 0.2-Am PVDF membrane (Osmonics, Inc.). Each membrane was then incubated overnight at 4°C with an appropriate diluted primary antibody. HRP-secondary anti-rabbit or anti-mouse antibodies (diluted 1:5,000 to yield 0.2 mg/ml; Abcam) were used to detect the bound primary antibodies. Blots were visualized using the Bioscience Odyssey Infrared Imaging System (LI-COR Biosciences), and band density was quantitated using ImageJ imaging analysis software (NIH, Bethesda, MD, USA). Data were normalized to β-actin expression by densitometry and statistical data from at least three experiments were recorded.

For detection of E-cadherin and vimentin, cells were seeded onto glass coverslips, fixed with 4% paraformaldehyde for 10 min and permeabilized with PBS Triton X-100 4% for 10 min. Cells were sequentially incubated with primary antibody (rabbit anti-E-cadherin and anti-vimentin 1:200) diluted in blocking buffer overnight at 4°C, which was followed by incubation with secondary antibodies conjugated with Alexa Fluor 568 (Molecular Probes-Invitrogen, Carlsbad, CA, USA) for 1 h at RT in the dark. After washing with PBS, the cells were stained with 4′,6′-diamidino-2-phenylindole (DAPI; targeting DNA in the cell nucleus) for 5 min. After extensive washing, the cells were mounted on a glass slide with 80% glycerol and fluorescent images were analyzed in an Olympus fluorescence microscope using a ×400 lens.

The Agilent SurePrint G3 Human Gene Expression microarray (8×60K) was used in the present study. This chip targets >30,000 genes with >50,000 probes and >12,000 lincRNA derived from a broad survey of well known sources such as RefSeq, Ensembl, UniGene and others. The resulting view of the human genome covers 30K unique genes and transcripts that have been verified and optimized by alignment to the human genome assembly and by the Agilent Empirical Validation process. Total RNA (>300 ng) was extracted from four independent cultures of Ishikawa, HEC-1B cells, shAMF-1 (Ishikawa) and shAMF-1 (HEC-1B) cells. Microarray hybridization, data collection and analysis were performed at Oebiotech Biotechnology Corp. (Shanghai, China). The threshold set for upregulated and downregulated genes was a fold-change ≥ 2.0. KEGG analysis was applied to determine the roles of these differentially expressed mRNAs.

Cells were seeded at 1.0×10⁵ cells/well of a 12-well plate, containing 1.2 ml regular medium. The media was then changed to phenol-red free medium with 0.5% stripped FBS for incubation at 37°C overnight. Immediately prior to treatment, the medium in the culture plates was aspirated, triply washed with PBS and replaced with fresh medium. U0126 dissolved in DMSO or PGI/AMF dissolved in PBS was subsequently added to each well. Concurrently, the same amount of DMSO or PBS was added to the control wells.

Continuous variables were recorded as mean ± SD and all statistical analyses were done using Statistical Package for the Social Sciences (SPSS) software version 17.0 (SPSS, Inc., Chicago, IL, USA). Volumetric data were assessed using an unpaired Student's t-test, or one-way ANOVA analysis followed by post-hoc LSD test or Dunnett's test for multiple comparisons. Concordance and correlation among antibodies were accessed by calculating Chi-square test and McNemar's statistical test. P-values <0.05 were considered statistically significant. All experiments were repeated independently at least three times.

AMF abnormal expression has been associated with tumor progression and EMT phenotype conversions in many human cancers (13,22,23). In our studies, immunochemistry staining showed that AMF protein was predominantly localized to the cytoplasm of endometrial epithelial cells. There was only weak or no staining in normal endometrium, whereas strong AMF immunostaining was found in endometrial carcinoma tissues, including endometrioid adenocarcinoma, uterine papillary serous carcinoma and endometrial clear cell carcinoma (Fig. 1A). The mean scores for AMF staining were 4.5 and 0.95 for cancer tissue and normal human endometrium, respectively (P<0.001) (Fig. 1B). Furthermore, we used a semi-quantitative analysis of IHC staining to evaluate AMF and EMT markers including E-cadherin, vimentin and Snail expression in the tissues microarrays. AMF levels positively correlated with vimentin (P=0.012) and Snail (P=0.021) levels, inversely correlated with the levels of E-cadherin (P=0.035) (Table I). These data indicated that AMF expression was much higher in EC tissue specimens and relevant to the levels of EMT marker protein.

Processes involved in the EMT are closely correlated with cancer metastasis. The expression of AMF and EMT markers in EC tissue specimens prompted us to examine whether AMF silencing could induce morphologic changes, loss of mesenchymal and/or gain of epithelial markers. The human EC cell lines Ishikawa and HEC-1B were stably expressed with either shAMF or control vectors. We microscopically examined AMF-silencing EC cells to determine the effects of AMF on cellular morphology. Ishikawa^shAMF and HEC-1B^shAMF cells were morphologically transformed toward epithelia compared with Ishikawa, HEC-1B and their control cells (Fig. 2A). This change was characterized a transition from the spindle-like morphology of the parental cells to a more differentiated keratinocyte-like morphology, suggesting a phenotypic transition from mesenchymal to epithelial. Thus, the expression of the intermediate filaments vimentin (a mesenchymal marker) and E-cadherin (an epithelial marker) was examined by immunofluorescent staining (Fig. 2B). Vimentin was prominently expressed throughout the cytoplasm of the parental Ishikawa and HEC-1B cells, whereas a significantly weaker vimentin signal was detected in their shAMF cell clone. Next, we tested whether the reduced vimentin expression was associated with a gain of epithelial markers. The cells were immunostained for the presence of the mesenchymal intermediate filament marker (i.e., E-cadherin). Either the parental Ishikawa or HEC-1B cells expressed E-cadherin proteins at a low level, while E-cadherin expression increased in the shAMF cells (Fig. 2B).

To further determine if this transformation represented an EMT [as has been reported (23)], we analyzed the levels of AMF and several characteristic epithelial and mesenchymal proteins by qRT-PCR and western blotting. We observed the expression of AMF was markedly decreased by stable transfection with shAMF (Fig. 2C and D). Furthermore, the levels of vimentin expression were reduced, while E-cadherin expression levels were raised in Ishikawa^shAMF and HEC-1B^shAMF cells (Fig. 2C and D). Taken together, these findings indicate that AMF silencing is accompanied by the loss of mesenchymal and the gain of epithelial markers, and show that AMF likely plays a crucial role in the regulation of the EMT in EC cells.

To investigate the functional significance of AMF involved in EMT in EC cells, we compared the expression profiles of the gene associated with EMT in Ishikawa/Ishikawa^shAMF and HEC-1B/HEC-1B^shAMF. Pairwise comparisons identified 46 differentially expressed genes that had changed by at least 2-fold after AMF silencing in Ishikawa. In contrast, only 34 differentially expressed genes were identified following AMF silencing of HEC-1B. We generated Venn diagrams by combining the separate gene lists obtained from the GeneSifter analysis, and then manually identified the common 15 genes that were regulated by AMF silencing in both Ishikawa and HEC-1B (Fig. 3A). To validate the array data, we performed qRT-PCR for two genes: Snail and transforming growth factor β receptor 1 (TGFBR1) whose expression was significantly downregulated by AMF silencing in both Ishikawa and HEC-1B. A comparison of the fold induction of each of the two genes as determined by microarray and qRT-PCR found that induction of gene expression as determined by qRT-PCR was consistently with that determined by microarrays (Fig. 3B). We further confirmed by western blotting that AMF silencing markedly downregulated the expression of the two genes in both Ishikawa and HEC-1B (Fig. 3C). These results demonstrated that AMF activated Snail and TGFBR1 regulated EMT in Ishikawa and HEC-1B cells.

To explore which pathways were activated during EMT mediated by AMF, we categorized the 15 differentially expressed genes commonly regulated by AMF silencing in both Ishikawa and HEC-1B using the KEGG pathway analysis tool provided in GeneSifter software. As shown in Fig. 4A, the total network comprised 9 distinct KEGG pathways, including pathways related to EMT (mitogen-activated protein kinase pathway and focal adhesion pathway) (19,20). In addition, two of these 15 differentially expressed genes were previously verified downregulated in AMF silencing cells including Snail and TGFBR1, reside in the MAPK pathway (25). We therefore assessed levels of phosphorylation of the MAPK pathway proteins ERK1/2 (at Thr202/Tyr204) in Ishikawa and HEC-1B cells. Interestingly, the phosphorylation levels of ERK were substantially inhibited by AMF silencing, whereas, exogenous PGI/AMF increased the phosphorylation levels of ERK in AMF silenced cells (Fig. 4B). To determine whether AMF-mediated regulation of the EMT phenotype resulted from the AMF ability to activate the MAPK pathway, U0126 (an inhibitor of MAPK pathway) was used to pretreat Ishikawa^shAMF and HEC-1B^shAMF cells. In Ishikawa^shAMF and HEC-1B^shAMF cells, exogenous PGI/AMF decreased the level of the epithelial marker E-cadherin, whereas not the levels of the mesenchymal marker vimentin. The EMT inducers Snail and TGFBR1 known as TGFβ receptor type I (15,26) were upregulated, compared with control groups (Fig. 4C and D). In contrast, the effects of exogenous PGI/AMF on enhancing EMT phenotype were eliminated in U0126 pretreated groups which MAPK pathway was inhibited (Fig. 4C and D). Our findings suggested that MAPK pathway activated by MAPK induced EMT in EC cells.