Samples were fully encoded to protect patient confidentially. The study and its protocols were approved by the Research Ethics committees of Shengjing Hospital Affiliated with China Medical University (2013PS66K). The ethics committee waived the need for patient consent, as the patient information was withheld.

All tissue samples used in this study (including the colon cancer sample) were obtained from the tissue bank at Shengjing Hospital of China Medical University. Between May, 2006 and July, 2010, 92 ovarian cancer patients were treated at the Department of Gynecology of Shengjing Hospital of China Medical University and were categorized according to the National Comprehensive Cancer Network (NCCN) guidelines, which uses the following definitions: Recurrence during the chemotherapy period (paclitaxel and carboplatin) or within 6 months after the end of chemotherapy is defined as drug resistance; recurrence between 6–12 months after the end of chemotherapy is defined as partial sensitivity; and recurrence beyond 12 months after the end of chemotherapy or no recurrence is defined as sensitivity. Based on these criteria, we divided all patients into the resistant group (n=37) and sensitive group (n=55). Lewis(y) expression in the ovarian cancer tissues was detected by immunohistochemistry and its association with the chemo-resistance status of the patients was examined. All the cases are primary, the information and follow-up data are complete, and chemical treatment was not used in any of the patients prior to surgery.

The following reagents were purchased from commercial sources: Mouse anti-human Lewis(y) monoclonal antibody (cat. no. F3, ab3359) from Abcam (Cambridge, UK); carboplatin as an apoptosis-inducing factor from Shandong Qilu Pharmaceutical Co., Ltd. (Jinan, China); DMEM and fetal bovine serum from HyClone (Logan, UT, USA); the streptavidin-biotin-peroxidase (SP) test kit from Boshide Biotech Co. (Wuhan, China). The protein content in the cell lysates was measured by BCA assay (Beyotime Institute of Biotechnology, Jiangsu, China); HRP-labeled secondary antibodies (goat anti-mouse IgG-HRP, cat. no. sc-2005, 1:2,000; goat anti-rabbit IgG-HRP, cat. no. sc-2004, 1:2,000) were purchased from Santa Cruz Biotechnology, Inc. (Dallas, TX, USA). For immunoblot analysis, the following antibodies were used: Annexin A4 (ANXA4) (cat. no. sc-28827, 1:200), Bax (cat. no. sc-20067, 1:500), p-Bad (cat. no. sc-166932, 1:500), Bcl-2 (cat. no. sc-509, 1:500), Bcl-xL (cat. no. sc-8392, 1:500) and β-actin (cat. no. sc-47778, 1:1,000) from Santa Cruz Biotechnology, Inc.; p-Akt (cat. no. 9271, 1:1,000) and caspase-3 (cat. no. 9662, 1:500) from Cell Signaling Technology, Inc. (Danvers, MA, USA). All other reagents were commercially available in China.

The RMG-I-H cell line, highly expressing the FUT1 gene and Lewis(y) antigen, was established by transfecting the pcDNA3.1(−)-HFUT-H expression vector (containing the FUT1 gene) into RMG-I cells (a human ovarian clear cell carcinoma cell line, donated by Professor Iwamori Masao of Tokyo University, Tokyo, Japan) (6). The SKOV3, A2780 and COC1 cells, and their cognate cisplatin-resistant cells SKOV3/DDP, A2780/DDP, COC1/DDP were obtained from the American Tissue Culture Collection (ATCC, Manassas, VA, USA).

The cells were maintained in DMEM supplemented with 10% fetal bovine serum at 37°C in a humidified 5% CO₂ atmosphere. Cells in the exponential growth phase were used in the subsequent experiments. In the case of carboplatin treatment, carboplatin was added to the culture medium at 60 mg/l and the cells were incubated for 48 h. For the inhibition assay, the final concentration of Lewis(y) antibody was 20 µg/ml. The duration of treatment was 1 h.

For SP immunohistochemistry, the following protocol was used. Tissues were fixed in 4% formaldehyde and embedded in paraffin. A colon cancer sample served as a positive control for Lewis(y) antigen. The group treated with phosphate-buffered saline (PBS) instead of primary antibody was used as a negative control. The working concentration of the primary antibody against Lewis(y) was 1:160. The empirical procedure was performed based on the kit instructions [streptavidin-biotin-peroxidase (SP) test kit].

The cells cultured in chamber slides were fixed with 4% of paraformaldehyde, then stained according to the SP test kit instructions. The primary antibody, mouse anti-human Lewis(y) antibody, was used at a 1:100 dilution. Lewis(y) immunostaining was performed using an avidin-biotin peroxidase complex kit and then slides were photographed. The presence of brownish-yellow granules in the cytoplasm and cell membrane were considered as a positive result.

The protocols strictly followed the instructions of the reagent suppliers. The ANXA4 antibody and Lewis(y) antibody were simultaneously added to monolayered cell slides prepared from the RMG-I-H cells. To these, the following secondary antibodies were applied: Fluorescein isothiocyanate green fluorescence-labeled mouse IgM fluorescence (cat. no. sc-2859; 1:8) and tetramethylrhoda-mine red fluorescence-labeled rabbit IgG (cat. no. sc-2492; 1:50) (both from Santa Cruz Biotechnology, Inc.). Cell nuclei were stained with 4′,6-diamidino-2-phenylindole. For negative controls, PBS replaced the primary antibodies. Double-labeled immunofluorescence samples were viewed under a fluorescence confocal microscope (C1-SI; Nikon, Tokyo, Japan).

Briefly, after the various treatments, the cells were washed twice with ice-cold PBS, scraped in lysis buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.5% NP-40, 100 mM NaF, 200 µM Na₃VO₄ and 10 µg/ml each aprotinin, leupeptin, phenylmethanesulfonyl fluoride and pepstatin], and incubated for 20 min at 4°C while rocking. Lysates were cleared by centrifugation (15 min at 13,000 × g, 4°C). For western blot analysis, 50 µg of total protein were resolved with 10% SDS-PAGE and transferred onto poly (vinylidene difluoride) membranes. The membranes were blocked with Tris-buffered saline Tween [25 mM Tris-HCl, 150 mM NaCl (pH 7.5), and 0.1% Tween-20] containing 5% non-fat milk and incubated overnight at 4°C with primary antibody in TBST/1% non-fat milk. The blots were washed in TTBS and incubated with the appropriate horseradish peroxidase-linked IgG at room temperature for 1 h, and immunoreactive proteins were visualized using an ECL detection system. The protein bands were visualized using the Molecular Imager system GDS8000b (UVP, Inc., Upland, CA, USA). Total protein levels were normalized to β-actin expression on the same membrane, and the bands were quantified using ImageJ software v1.8.0 (National Institutes of Health, Bethesda, MD, USA).

The cold PBS washed monolayer cells were lysed with 200 µl lysis buffer as described above. The protein content was measured using the protein assay BCA kit (Beyotime Biotechnology, Jiangsu, China). Following protein determination, cell lysate containing 500 µg protein was incubated with 5 µg of one of the following antibodies, and incubated at 4°C overnight. Protein G plus-agarose was added and the samples were incubated at 4°C for 3 h for immunoprecipitation. ANXA4 was subjected to 10% SDS-PAGE, then transferred onto a poly (vinylidene difluoride) membrane and treated with 1:500 diluted anti-Lewis(y) and anti-ANXA4 sera in Tris-buffered saline with 5% non-fat milk, followed by HRP-labeled secondary antibody at a 1:1,000 dilution. Finally, the proteins were visualized with ECL reagent (ECL Prime Western Blotting Detection Reagent, Amersham, Pittsburgh, PA, USA). The densitometric analysis of the protein bands was performed using ImageJ software v1.8.0 (National Institutes of Health, Bethesda, MD, USA).

Total RNA was extracted from the ovarian cancer cells using TRIzol reagent (Invitrogen, Shanghai, China). Complementary DNA (cDNA) was synthesized using reverse transcriptase from a Perfect Real-Time PrimeScript™ RT Reagent kit (Takara Bio, Inc.). The reaction conditions were as follows: 37°C for 15 min, 85°C for 5 sec, 4°C for 5 min. Real-time PCR and Ct value analysis were performed with a Roche LightCycler system. The primers of target genes were commercially synthesized (Table I). The real-time PCR reaction conditions were denature at 95°C for 30 sec, 40 cycles of 95°C for 5 sec and 60°C for 30 sec in a 20-µl reaction mixture containing SYBR^@ Premix Ex Taq™ (2X) 10 µl, PCR Forward Primer (10 µmol/l) 0.4 µl, PCR Reverse Primer (10 µmol/l) 0.4 µl, cDNA 2 µl, dH₂O 7.2 µl. GAPDH was used as an endogenous control. Relative quantification was performed according to the ΔΔCq method, and results were expressed in the linear form using the formula 2^−ΔΔCq (14). Results were considered significant when at least a 2-fold difference in expression levels was detected.

Using human whole genome oligonucleotide microarrays (Agilent whole human genome oligo microarray; Agilent Technologies, Santa Clara, CA, USA), gene expression profile alteration in response to Lewis(y) was investigated and 390 DEGs were identified and validated in the epithelial ovarian cell lines (RMG-I and COC1) compared with their sublines (RMG-I-H and COC1/DDP) with greater resistance to chemotherapy, in which 229 genes were upregulated and 161 genes were down-regulated (15). Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis were performed for the identified DEGs using the Database for Annotation, Visualization and Integrated Discovery (DAVID) database. A value of P<0.05 was set as the cut-off criterion.

The functional interactions between proteins can provide context for the molecular mechanism of cellular processing. In this study, a PPI network of DEGs was constructed using the Search Tool for the Retrieval of Interacting Genes (STRING, http://string.embl.de/) database and subsequently was visualized using Cytoscape. A confidence score ≥0.4 was set as the cut-off criterion. Subsequently, Molecular Complex Detection (MCODE) was performed to screen the modules of the PPI network with a degree cut-off of 2, a node score cut-off of 0.2, a k-core of 2, and a max. depth of 100. Finally, the functional enrichment analysis of genes in each module was performed using the DAVID database.

SPSS 17.0 statistical software (IBM, Armonk, NY, USA) was applied for statistical analysis. Statistical methods were varied based on the data type. Quantitative data are presented as the means ± SD. As for the analysis of the data from RT-qPCR, data are expressed as the means ± SEM. Positive ratio rates were evaluated using the Chi-square (χ²) test. A Student's t-test was employed for comparisons between 2 groups and one-way ANOVA with the LSD or Bonferroni post hoc test was used for comparisons between >2 groups. A P-value <0.05 was considered to indicate a statistically significant difference.

Our results revealed that in the 92 samples of ovarian cancer tissues, Lewis(y) antigen was mainly expressed on the cell membrane and to a lesser extent in the cytoplasm (Fig. 1A). The positive rate of Lewis(y) antigen in chemoresistant ovarian cancer tissues was 91.89%, which was significantly higher than that in the sensitive group (61.82%, P<0.05) (Table II). The expression intensity of Lewis(y) antigen in the resistant group was also significantly higher than that in the sensitive group. In the resistant group, 8 of the 34 positive cases exhibited a strong positive expression, while in the sensitive group, there were no strongly positive cases among the 34 positive cases (Table II).

In previous studies, we introduced the FUT1 gene into human ovarian carcinoma-derived RMG-I cells through gene transfection and established a cell model overexpressing the FUT1 gene and Lewis(y) antigen. Further experiments demonstrated that the FUT1-transfected cells exhibited a reduced sensitivity to common chemotherapeutic drugs, such as carboplatin, 5-FU and paclitaxel (6–8). In this study, to further examine whether the expression of Lewis(y) antigen was increased when the cells became resistant to chemotherapy, the expression of Lewis(y) antigen and FUT1 in ovarian cancer cells was detected by immunocytochemical staining and RT-qPCR, respectively. The results revealed that after the cells were challenged by chemotherapeutic drugs, the expression of Lewis(y) antigen in the cisplatin-resistant cell lines (SKOV3/DDP, COC1/DDP and A2780/DDP) was significantly higher than the expression in their parental non-resistant cell lines (SKOV3, COC1 and A2780); the antigen was also widely distributed on the cell membrane (Fig. 1B). The results of RT-qPCR confirmed that the mRNA expression level of FUT1 in the SKOV3/DDP cells increased by 3.17-fold in comparison to that in the SKOV3 cells, the mRNA expression level of FUT1 in the COC1/DDP cells increased by 2.42-fold in comparison to that in the COC1 cells, and the mRNA expression level of FUT1 in the A2780/DDP cells increased by 2.84-fold in comparison to that in the A2780 cells (all P<0.05) (Fig. 1C).

GO enrichment analysis revealed that the DEGs were mainly involved in the transmembrane receptor protein tyrosine kinase signaling pathway, the positive regulation of cell proliferation, the integrin-mediated signaling pathway, extracellular matrix organization and angiogenesis (Fig. 2 and Table III).

KEGG pathway analysis of the DEGs suggested that these genes were significantly enriched in antigen processing and presentation, fructose and mannose metabolism, and central carbon metabolism in cancer (Fig. 3 and Table IV).

We also constructed a PPI network to examine the mechanisms through which these genes interact with each other, as well as to identify the central node of the PPI network. The PPI network of the DEGs consisted of 323 nodes and 287 edges. A significant module was obtained from the PPI network of DEGs using MCODE, including 18 nodes and 45 edges (Fig. 4A and B). Functional enrichment analysis revealed that genes in this module were mainly associated with the regulation of the apoptotic process, the positive regulation of release of the cytochrome c from the mitochondria, and protection from natural killer cell-mediated cytotoxicity (Fig. 4C).

Four DEGs, ANXA4, BCL2 interacting killer (BIK), transmembrane 4 L six family member 4 (TM4SF4) and pleckstrin homology like domain family A member 1 (PHLDA1), were selected to validate the results of gene chip analysis. The results of RT-qPCR revealed that the mRNA levels of the 4 genes were significantly increased in the Lewis(y) highly-expressing chemoresistant ovarian cancer cells (RMG-I-H and COC1/DDP cells), exhibiting 2.82-(ANXA4), 4,21-(BIK), 9.04-(TM4SF4) and 3.38-fold (PHLDA1) increases over the expression levels in the non-resistant cells (RMG-I and COC1 cells) (P<0.01) (Fig. 5). This was consistent with the results of gene chip analysis. These results suggest that Lewis(y) causes cancer chemotherapeutic resistance may be due to the inhibition of apoptosis.

As ANXA4 is a membrane phospholipid-binding protein, we further investigated the protein expression of ANXA4 in ovarian cancer cells and the presence of Lewis(y) antigen in this protein. The results of western blot analysis revealed that the protein expression of ANXA4 was significantly increased in the ovarian cancer chemotherapy-resistant cells (RMG-I-H and COC1/DDP cells) expressing high levels of Lewis(y), with an expression level increase of 2.72-fold and 2.24-fold over the level in RMG-I cells and COC1 cells, respectively, exhibiting the same trend as the mRNA expression (Fig. 6A).

Co-immunoprecipitation and laser confocal microscopy were used to determine whether there was a Lewis(y) structure in the ANXA4 protein. The results revealed that ANXA4 co-immunoprecipitated with Lewis(y) antigen corresponding to the 36 kDa band (Fig. 6B). Dual-labeling immunofluorescence revealed red immunofluorescence, indicating that the Lewis(y) antigen was mainly distributed in the membrane, while green immunofluorescence indicated ANXA4 was also mainly distributed in the membrane, although it could also be observed in the cytoplasm. Moreover, red and green immunofluorescence were overlaid in the membrane, suggesting the co-localization of Lewis(y) antigen and ANXA4 (Fig. 6C).

We further examined the expression of apoptosis-related proteins, such as Bcl-2 in ovarian cancer cells. Following carboplatin treatment, the RMG-I-H cells exhibited an upregulated expression of the pro-apoptotic protein, Bax, by 57%, and the phosphorylation of the anti-apoptotic protein, Bad, markedly increased by 197%, compared with the pretreatment protein levels (Fig. 7A) (Bad itself is a pro-apoptotic protein that becomes anti-apoptotic upon phosphorylation by Akt.) The expression of other anti-apoptotic proteins, such as Bcl-2 and Bcl-xL, was not noticeably altered following carboplatin treatment. Before treatment, the Bax expression level in the RMG-I cells was higher than that in the RMG-I-H cells by approximately 3.95-fold. In addition, the post-treatment expression of Bax in the RMG-I cells increased by approximately 6.02-fold greater than the increase exhibited by the RMG-I-H cells. The phosphorylation levels of Bad, and the levels of Bcl-2 and Bcl-xL in the RMG-I-H cells significantly increased by 4.64-, 1.91- and 2.07-fold as much as in the RMG-I cells, respectively. Following carboplatin treatment, the phosphorylation level of Bad in the RMG-I-H cells was approximately 7.78-fold higher than that in the RMG-I cells. Following carboplatin treatment, there was no evident change in the levels of phosphorylated Bad, and the levels of Bcl-2, or Bcl-xL in the RMG-I cells.

We also examined the expression of caspase-3, and found that, in the absence of carboplatin, caspase-3 existed in its inactive, pro-enzyme form in the RMG-I and RMG-I-H cells, and thus no hydrolysis of caspase-3 was observed (Fig. 7B). The hydrolytic breakdown of caspase-3 into a small fragment subunit is required for its activation. The caspase-3 pro-enzyme level in the RMG-I cells was approximately 1.6-fold higher than that in the RMG-I-H cells. Treatment with carboplatin markedly increased the level of active caspase-3 in the RMG-I cells to a level approximately 4.03-fold as much as that in the RMG-I-H cells. However, the level of the caspase-3 pro-enzyme exhibited no apparent changes in response to carboplatin treatment in either cell line.

To further determine whether the Lewis(y) antigen is involved in the inhibitory effects on the apoptosis of ovarian cancer cells, we used Lewis(y) antibody to treat the cells before carboplatin treatment. The phosphorylation levels of Akt and Bad were decreased in both cell lines in response to Lewis(y) antibody pre-treatment, although this decrease was greater in the RMG-I-H cells, which overexpress Lewis(y) (Fig. 8). The active fragment of caspase-3 was increased in both cell lines, but more predominantly in the RMG-I-H cells (Fig. 8). High-and low-level differences between the RMG-I and RMG-I-H cells with regard to the levels of phosphorylated Akt, phosphorylated Bad and caspase-3 fragments were also suppressed after Lewis(y) antibody pre-treatment.