PC3, a highly metastatic prostate cancer cell line, was obtained from American Type Culture Collection (ATCC; Manassas, VA, USA). Cells were maintained in RPMI-1640 medium (Sigma-Aldrich, St. Louis, MO, USA) supplemented with 10% fetal bovine serum (FBS) (PAA Laboratories, Morningside, Australia). All biochemical reagents were purchased from Sigma-Aldrich, unless otherwise noted. Recombinant human tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) was purchased from R&D Systems, Inc. (Minneapolis, MN, USA). The following monoclonal antibodies to human antigens were used: anti-MUC1 (VU4H5; Cell Signaling Technology, Danvers, MA, USA) and phycoerythrin (PE)-labeled anti-CD56 (BD Biosciences, San Jose, CA, USA). The following polyclonal antibodies were used: anti-lysosome-associated membrane glycoprotein 1 (anti-LAMP1; Lifespan Biosciences, Seattle, WA, USA), anti-death receptor 4 (anti-DR4; Millipore, Temecula, CA, USA) and anti-actin (Sigma-Aldrich).

PC3 cells express high levels of C2GnT. PC3 cells with reduced expression levels of C2GnT were generated using shRNA technology as previously described (7). An shRNA expression plasmid was constructed using pBAsi-hU6 Neo DNA (Takara Bio Inc., Shiga, Japan). The shRNA sequence for C2GnT-1 was: GATCCGAATCCTAGTAGT GATATTTAGTGCTCCTGGTTGAATATCACTACTAGG ATTCTTTTTTA. The siRNA sequence is underlined. The sequence of the control shRNA containing the scrambled siRNA sequence was: GATCCGTCTTAATCGCGTATAAGG CTAGTGCTCCTGGTTGGCCTTATACGCGATTAAGAC TTTTTTA. The shRNA expression plasmids (knockdown and control constructs) were introduced into PC3 cells using Lipofectamine 2000. Drug-resistant colonies were selected in the presence of 200 μg/ml geneticin. Total RNA was prepared from each drug-resistant cell line using a FastPure RNA kit (Takara Bio Inc.). Quantitative real-time PCR was performed to quantify the C2GnT mRNA level of each drug-resistant cell line using a Prime Script RT-PCR kit and Thermal Cycler Dice Real Time system (Takara Bio Inc.). The sequences of the primer set used for C2GnT-1 were: 5′-GGATGTCACCTG GAATCAGCACTA-3′ and 5′-TTATCAGAGCTGCAACGG CATC-3′. The primers specific for glyceraldehyde-3-phosphate dehydrogenase (GAPDH) were: 5′-ATGACTCTACCCACG GCAAG-3′ and 5′-CATACTCAGCACCAGCATCAC-3′ and were used as the internal control. Two C2GnT-deficient clones (C2KD-1 and -2) were chosen based on their reduced mRNA expression levels. C2KD-1, C2KD-2 and one control clone (PC3) were used for the assays described in the present study.

Total lysates of cancer cells were prepared by solubilization in 50 mM Tris-HCl buffer, pH 7.5, containing 1% IgepalCA-630, 150 mM NaCl and proteinase inhibitors. The lysates were resolved by SDS-PAGE on an 8–16% gradient gel (Invitrogen Life Technologies, Carlsbad, CA, USA) and transferred to a polyvinylidene fluoride (PVDF) membrane. Western blotting analyses were performed using specific primary antibodies and a horseradish peroxidase-conjugated secondary antibody. Signals were visualized using the ECL PLUS detection system (GE Healthcare, Amersham, UK).

Lectin immunopreocipitation was performed as previously described (7). For lectin immunoprecipitation, lysates from prostate cancer cells were incubated with Lycoperiscon esculentum (tomato) lectin (LEL)-agarose (Vector Laboratories, Burlingame, CA, USA). The resin binding the immune complex was eluted with 1X Laemmli SDS-PAGE sample buffer.

Human primary NK cells were purified from human peripheral blood mononuclear cells using an NK cell isolation kit (Myltenyi Biotech, Auburn, CA, USA). The NK cells (1×10⁶ cells/ml) were cultured for 3 days in RPMI-1640 medium supplemented with 10% FBS in the presence of 1000 U/ml human recombinant IL-2 (Wako, Osaka, Japan).

Heterotypic cell conjugates were quantitatively determined by a double fluorescence assay (12) with certain modifications. Prostate cancer target cells (2×10⁶ cells/ml) were tranfected with a green fluorescence protein (GFP)-expression plasmid, pmaxGFP (Lonza Walkersville, Inc., Walkersville, MD, USA). After 1 h incubation of the GFP-expressing tumor cells with IL-2-activated NK cells at 37°C, the cells were stained with PE-labeled anti-CD56 antibody and then the number of conjugates were counted using a flow cytometer, FACScant II (BD Biosciences). Double-colored (green and red) conjugates were calculated as a percentage of the total GFP-positive cells. For the granzyme B secretion assay, the culture supernatant was collected from the co-culture of the target cancer cells with NK cells. The quantity of granzyme B in the supernatant was assayed by measuring serine proteinase activity with a colorimetric peptide substrate, IEPD-p-nitroanilide (Biomol International, Plymouth Meeting, PA, USA) as previously described (7). Specific granzyme B secretion was expressed as a percentage of the total cellular enzyme activity after subtracting the spontaneous release.

Cytotoxicity was measured using the Cytotox 96 Non-radioactive Cytotoxicity Assay kit (Promega Corporation, Madison, MI, USA). The target cells (tumor cells) were incubated with IL-2-activated NK cells for 4 h at 37°C. The release of lactate dehydrogenase from the lysed target cells was measured. All assays were performed in quadruplicate. Percentage cytotoxicity = (experimental lactate dehydrogenase release - effector spontaneous release - target spontaneous release)/(target maximum release - target spontaneous release) × 100.

For the cell viability assay, the cells (1×10⁴ cells/well in 96-well plates) were incubated with the indicated concentration of recombinant TRAIL at 37°C for the indicated time. The cell viability was assayed using the Cell Counting Kit-8 (Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer’s instructions.

We used the statistical program SPSS 12.0 (SPSS, Chicago, IL, USA). Statistically significant differences were determined using the Student’s t-test. P<0.05 was considered to indicate a statistically significant result.

C2GnT is responsible for the formation of core2 O-glycans (Fig. 1A). To test our hypothesis, we investigated the role of core2 O-glycans in the evasion of NK cell immunity by C2GnT-expressing prostate cancer. We used a prostate cancer cell line, PC3, which is derived from a malignant and metastatic prostate cancer and expresses C2GnT at a high level (13). We established two C2GnT-deficient cell lines (designated C2KD-1 and C2KD-2) and one control cell line (PC3) using PC3 cells. RT-PCR analysis of the C2GnT expression showed that the C2GnT expression levels were markedly reduced in the C2KD-1 and C2KD-2 cells compared with those in the PC3 cells (Fig. 1B). These cells exhibited no marked morphological differences (Fig. 1C). The results obtained for C2KD-1 and PC3 only are shown, since the two C2GnT-deficient clones yielded almost identical results in all assays.

We have previously demonstrated that patients with C2GnT-expressing prostate cancer have a significantly shorter survival time than patients with C2GnT-non-expressing prostate cancer (6), suggesting that C2GnT-expressing prostate cancer is highly metastatic. However, a growing body of evidence indicates that O-glycans in cell-surface mucins are important in numerous biological processes, including the protection of epithelial cell surfaces, immune response, cell adhesion, inflammation, tumorigenesis and tumor metastasis (14). These observations led us to postulate that core2 O-glycans carried by mucins play an important role in prostate cancer metastasis. We analyzed the MUC1 glycoproteins of the prostate cancer cells for O-glycosylation by western blotting. The MUC1 of the control PC3 cells exhibited a higher molecular weight than the MUC1 from C2GnT-deficient cells (C2KD-1; Fig. 2, lanes 1 and 2). By contrast, we observed no significant differences in the molecular weights of the non-O-glycosylated cell-surface protein, LAMP1, between the PC3 and C2KD-1 cells (Fig. 2, lanes 3 and 4). These results indicate that MUC1 glycoproteins from PC3 cells carry higher levels of core2 O-glycans than those from C2KD-1 cells due to their higher expression levels of C2GnT.

The core2 branch is a scaffold for the subsequent production of lactosamine disaccharide repeats, specifically poly-N-acetyllactosamine (Galβ1-4GlcNAc)n (Fig. 1A) (8). To determine whether MUC1 from PC3 cells carries poly-N-acetyllactosamine on its O-glycans, we first excluded N-glycans from prostate cancer cells by treatment with tunicamycin, an N-glycosylation inhibitor, and then analyzed the cell lysates by immunoprecipitation using LEL. LEL binds specifically to poly-N-acetyllactosamines with at least three lactosamine unit repeats. The LEL immunoprecipitates were subjected to western blotting with anti-MUC1 antibody. MUC1 was detected in the LEL immunoprecipitates from the PC3 cells, but the amount of MUC1 detected in the LEL immunoprecipitates from the C2KD-1 cells was markedly reduced (Fig. 2, lanes 7 and 8). This result indicates that MUC1 from C2GnT-expressing PC3 cells carries a larger amount of poly-N-acetyllactosamine on its core2 O-glycans than that from C2KD-1 cells.

NK cells play a critical role in tumor rejection responses in the host blood circulation. The attack on cancer cells by NK cells is initiated by the NK cell-cancer cell interaction mediated through the NK receptor-ligand interaction (15). To determine whether C2GnT expression affects the NK cell-prostate cancer cell interaction, we compared the conjugate formation of NK cells with PC3 and C2KD-1 cells. Following the incubation of the target cells with NK cells for 1 h, the NK cells formed a significantly higher number of conjugates with the C2KD-1 cells than with the PC3 cells (Fig. 3A). This result, taken together with Fig. 2, suggests that in the C2GnT-expressing cells (PC3), the poly-N-acetyllactosamine moieties carried by the MUC1 core2 O-glycans reduce the adhesive properties of MUC1 due to their bulkiness, resulting in reduced conjugate formation between the PC3 and NK cells (Fig. 3A).

NK cells are activated through the NK cell-target cell interaction and release their granular contents to kill the target cells. The granular contents include perforin and granzyme B which induce apoptosis of the target cells. We measured the secretion of granzyme B stimulated by the NK cell-prostate cancer cell interaction using an assay of the protease activity of granzyme B in the co-culture supernatant. The C2GnT-expressing PC3 cells induced the secretion of significantly lower levels of granzyme B than the C2GnT-deficient C2KD-1 cells (Fig. 3B). These results suggest that MUC1 glycoproteins carrying poly-N-acetyllactosamine attenuate the NK cell-prostate cancer cell interaction, resulting in decreased secretion of granzyme B (Fig. 3).

We then questioned whether C2GnT expression by prostate cancer cells affects the cytotoxic activity of NK cells, since the interaction of NK cells with C2GnT-expressing prostate cancer cells was impaired and the secretion of the target cell apoptosis-inducing substance, granzyme B was reduced (Fig. 3). To address this, we assayed the cytotoxicity of NK cells against prostate cancer cells. Fig. 4A shows that NK cells killed C2KD-1 cells more efficiently than PC3 cells (Fig. 4A), indicating that C2GnT-expressing prostate cancer cells (PC3) were more resistant to NK cell cytotoxicity than were C2GnT-deficient cells (C2KD-1).

NK cells use two main mechanisms to kill tumor cells. One is that NK cells are activated through the NK cell receptor-tumor ligand interaction to release cytotoxic granules, including perforin and granzymes. The other is that NK cells kill tumor cells using death ligands, including TRAIL (16). TRAIL induces target cell apoptosis through its interaction with death receptors, including DR4, on the surface of the target cell. We evaluated the expression of DR4 on the prostate cancer cells by western blotting. We observed no significant differences in the DR4 expression levels between PC3 and C2KD-1 (Fig. 4B, lanes 1 and 2). Unlike MUC1, no significant differences in the O-glycosylation levels of DR4 between PC3 and C2KD-1 cells were observed on western blots (Figs. 2 and 4B). To evaluate the effect of C2GnT expression on the sensitivity of prostate cancer cells to TRAIL, we measured TRAIL-induced cell death using soluble recombinant TRAIL. In the presence of TRAIL, the viability of the PC3 cells was significantly higher than that of the C2KD-1 cells (Fig. 4C), indicating that C2GnT-expressing prostate cancer cells are more resistant to TRAIL than are C2GnT-deficient cells.