RNA Interference Suppression of MUC1 Reduces the Growth Rate and Metastatic Phenotype of Human Pancreatic Cancer Cells

Materials and Methods

Materials. The S2-013 cell line is a cloned subline of human pancreatic tumor cell line (SUIT-2), which was derived from a liver metastasis (6). Restriction enzymes and ligase were purchased from New England Biolabs (Beverly, MA). Monoclonal antibodies were obtained from the following sources: B27-29 against MUC1 was kindly provided by Dr. Mark Reddish (Biomira, Inc., Edmonton, Alberta, Canada). Armenian hamster monoclonal antibody (mAb) CT-2 against MUC1CT was kindly provided by Dr. Sandra Gendler (Mayo Clinic, Scottsdale, AZ). The following sources were used for reagents: anti–carcinoembryonic antigen–related cell adhesion molecule (CEACAM) 5 + 6 mAbs from Abcam, Inc. (Cambridge, MA); mouse mAb anti-β-actin from Sigma (St. Louis, MO); peroxidase-conjugated goat anti-mouse IgG from Southern Biotechnology Associates (Birmingham, AL); peroxidase-conjugated rabbit anti-mouse IgG serum from Sigma; enhanced chemiluminescence kits from Amersham Life Sciences, (Piscataway, NJ) and Pierce (Rockford, IL); Lipofectin from Life Technologies, Inc. (Carlsbad, CA); cell culture medium from Celox (Oakdale, MN) or Life Technologies; and fetal bovine serum and geneticin (G418) from Life Technologies.

Cell culture. S2-013 cells were cultured in RPMI 1640 supplemented with 10% heat-inactivated fetal bovine serum. The cells were passaged at >80% confluence using 0.25% (w/v) trypsin solution containing 0.04% (w/v) EDTA.

siRNA design. Selection of the 19 nucleotide target sequences were based on the OligoEngine protocols (OligoEngine, Inc., Seattle, WA). 5′-ACCTCCAGTTTAATTCCTC-3′ (target 2), 3151 downstream of the initiation codon was selected (target 2 was kindly suggested by Dr. Sandra Gendler, Mayo Clinic, Scottsdale, AZ). These were separated by a nine-nucleotide noncomplementary spacer (TCTCTTGAA) from the reverse complement of the same 19-nucleotide sequence.

Preparation of recombinant plasmids. Oligonucleotides (64 bp) were ligated into the mammalian expression vector, pSUPER gfp-neo (OligoEngine) at the BglII and HindIII cloning sites. Recombinant MUC1-pSUPER gfp-neo constructs were used to transform Escherichia coli DH5α, which were selected on ampicillin-agarose plates and verified by sequencing.

Transfection and selection of clones. S2-013 cells were transfected with recombinant pSUPER gfp-neo by the Lipofectin method. Briefly, cells were plated in six-well culture plates and grown to ∼60% confluence. Growth medium was removed, the cells were washed twice in serum-free Opti-MEM, and then incubated 5 hours in 1 mL serum-free Opti-MEM with 10 μL Lipofectin reagent and 1 μg recombinant pSUPER gfp-neo with a target or pSUPER gfp-neo as control. At 24 hours after transfection, the medium was replaced with normal growth medium, and after 48 hours each well was passaged into a 10 cm plate with growth medium containing G418 at 600 μg/mL. After 7 to 10 days, the cells were passaged and plated at ∼30 to 50 per plate. Single colonies were selected and the cloning procedure was repeated. We established six clones of MUC1-suppressed cells, S2-013.MTII.C1-C6. The clonal cell lines were maintained in complete medium with G418 at 150 μg/mL.

Western blot analysis. Cell lysates were prepared by applying 400 μL lysis buffer [10 mmol/L Tris-HCl (pH 8.0), containing 150 mmol/L NaCl, 1 mmol/L phenylmethylsulfonyl fluoride, and 1% Triton X-100] to confluent cells grown in 60 mm dishes. Twenty-five micrograms of protein were loaded on 4% to 20% Novex Tris-Glycine gradient denaturing polyacrylamide gels (Invitrogen) in a 1× SDS-PAGE buffer (1 g/L SDS, 3 g/L Tris base, and 14.4 g/L glycine). Proteins were transferred to polyvinylpyrrolidine difluoride membranes electrophoretically and incubated overnight at 4°C in Blotto [5% dry milk in 1× TBS (0.9% NaCl, 10 mmol/L Tris (pH 7.4), and 0.5% MgCl₂)]. Membranes were incubated for 60 minutes at room temperature with primary antibody in Blotto, followed by three 10-minute washes with Blotto. After washing with Blotto, the membrane was incubated with a 1:4,000 dilution of horseradish peroxidase–linked anti-mouse secondary antibodies. The immune complexes were detected using ECL Western blotting detection reagents. The membranes were stripped of bound antibody and reprobed with an anti-β-actin antibody to confirm equal loading of the samples.

Flow cytometric analysis of MUC1 surface expression. S2-013 cells were harvested at a density of 8 × 10⁵ to 10 × 10⁵ cells per cm² with 0.25% (w/v) trypsin solution containing 0.04% (w/v) EDTA. All of the subsequent steps were carried out on ice. Cells were incubated with an appropriate dilution of a mouse mAb antibody, B27-29, against human MUC1 mucin for 30 minutes at 4°C. After washing twice with fluorescence-activated cell sorting buffer (1× PBS with 1% fetal bovine serum), the cells were incubated with a phycoerythrin-conjugated rabbit anti-mouse secondary antibody (Jackson ImmunoResearch, West Grove, PA) for 30 minutes at 4°C. The cells were washed twice and resuspended in fluorescence-activated cell sorting buffer, followed by analysis on a FACSCalibur (Becton Dickinson, Immunocytometry Systems, San Jose, CA) flow cytometer. Expression levels were determined as a percent positive above control cells not expressing MUC1. List mode data files containing multiparametric log data were analyzed using CellQuest (Becton Dickinson Immunocytometry Systems). Cells were gated using forward and side scatter variables to eliminate dead cells and debris.

Cell proliferation assay. Cell proliferation was evaluated by direct evaluation of change in cell number over time. For cell counting assays, triplicate aliquots of 1 × 10⁴ of each clone and a control cell line were seeded on 48-well plates. At 24-hour intervals, cells were harvested by trypsinization and counted on a cell counter. A profile of the cell size was determined at the same time.

Cell growth was transformed to log scale to better meet model assumptions. Regression modeling with a generalized estimating equations approach was used to examine differences in cell growth over time and account for the correlation over time. P < 0.05 was considered significant. All analysis was done using the SAS System for Windows, version 8.2 (SAS Institute, Cary, NC).

Analysis of DNA-cell cycle and apoptosis evaluation. Cells were prepared according to the method of Telford et al. (7). Briefly, the pellets of 1 × 10⁶ cells were fixed by 1 mL of 70% ethanol at 4°C for 1 hour. After washing the cells with PBS, 1 mL Telford reagent [EDTA: 16.81 g, RNase A (93 units/mg): 13.4 mg, propidium iodide: 25 mg, Triton X-100: 500 μL, PBS: 500 mL] was added and incubated at 4°C overnight. Data were collected on a Becton Dickinson FACSCalibur (Becton Dickinson Immunocytometry Systems) flow cytometer within 24 hours of staining. List mode data were subsequently analyzed using ModFit (Verity Software House, Topsham, ME). Control samples were used to select regions of the cell cycle. Apoptosis was reported as the percentage of events occurring in the sub-G₀-G₁ peak as determined by the software algorithm.

Adhesion assay. The cell adhesion assay was based on the method described by Laurie et al. (8). Ninety-six-well plates were coated with type I collagen, type IV collagen, laminin, or fibronectin (5 μg/cm²) overnight at 4°C and washed with PBS. One set of wells for each group was not coated as a control. Bovine serum albumin was added to the coated plates and also noncoated wells, which were incubated for 1 hour at 37°C to block remaining binding sites. Aliquots of suspended cells (5 × 10⁴) in culture medium containing 0.02% bovine serum albumin were added in triplicate to coated and noncoated wells, and incubated for 60 minutes at 37°C. At the end of the assay, the plate was washed thrice with PBS to remove nonadherent cells. Adherent cells were stained with crystal violet. Fifty microliters of 0.01% crystal violet solution were added to each well and incubated for 30 minutes at room temperature, followed by five washes with PBS. The crystal violet that absorbed onto the cells was solubilized with 0.1% Triton X-100 (Sigma). The absorbance was measured at 590 nm with a microplate reader. Results were expressed as a percentage of the noncoated control well. Nonparametric Kruskal-Wallis tests were used to compare differences in adhesion between cell lines for each coating type. If the overall Kruskal-Wallis test indicated a difference in cell lines, a Wilcoxon rank sum test was done for each pairwise comparison of interest. P < 0.05 was considered significant. All analyses were done using the SAS System for Windows, version 8.2 (SAS Institute).

Oligonucleotide array gene expression analysis. Human oligonucleotide probes (60-mers) were designed for each target gene (Compugen, Inc., Rockville, MD) and manufactured by Sigma-Genosys, Inc. (The Woodlands, TX). Oligonucleotides (12, 261) were resuspended (30 μmol/L) in 3× SSC and spotted onto poly-l-lysine–coated slides using a MagnaSpotter robot (BioAutomation Corp., Dallas, TX) with a 12-pin print head (Telechem, Sunnyvale, CA) configuration in a humidified, HEPA-filtered hood. After spotting, the DNA was cross-linked to the slides by UV irradiation (350 mJ/cm²) with a Stratalinker UV Crosslinker (Stratagene, Inc., La Jolla, CA), blocked by succinic anhydride treatment, and rinsed in ethanol. The printed arrays were boxed and stored desiccated at room temperature.

Total RNA was isolated from each cell line (S2-013.gfp-neo, S2-013.MTII.C1, S2-013.MTII.C2) using TRIzol reagent (Life Technologies) according to the protocol of the manufacturer. To generate fluorescently labeled single-stranded cDNA target, 40 μg total RNA was incubated with 2 μg anchored oligodeoxythymidylate primer in a total volume of 30 μL at 70°C for 10 minutes and chilled on ice. To each sample was added 10× first-strand buffer (6 μL), 0.1 mol/L DTT (6 μL), 20× aminoallyl-deoxynucleotide triphosphate mix (3 μL; 10 mmol/L dATP, 10 mmol/L dCTP, 10 mmol/L dGTP, 6 mmol/L dTTP, and 4 mmol/L aminoallyl-dUTP), RNasin (0.5 μL), and StrataScript reverse transcriptase (3 μL; 50 units/μL). After incubation at 48°C for 60 minutes, an additional 50 units (1 μL) of StrataScript was added to the samples and incubated for an additional 60 minutes. The reaction was stopped by adding 12 μL of 0.5 mol/L EDTA (pH 8). Residual RNA was hydrolyzed by adding 12 μL of 1 mol/L NaOH to the mixture followed by incubation at 65°C for 15 minutes, and cooling to room temperature. The reaction was neutralized with 16.8 μL of 1 N HCl. First-strand cDNA was purified from unicorporated amino-allyl-dUTPs on QIAquick PCR purification columns (Qiagen, Valencia, CA) according to protocol of the manufacturer, except that QIAquick wash buffer was replaced with 5 mmol/L potassium phosphate buffer (pH 8.5) containing 80% ethanol, and cDNA was eluted with 4 mmol/L potassium phosphate buffer (pH 8.5) and vacuum-dried. cDNA was resuspended in 10 μL of 0.05 mol/L Na₂CO₃ buffer (pH 9), mixed with either Cy3 or Cy5 monofunctional NHS-ester (Amersham Pharmacia), and incubated for 90 minutes in the dark at room temperature. Cy3- and Cy5-conjugated cDNA targets were then purified by QIAquick PCR purification columns, combined, vacuum-concentrated, and diluted to 55 μL with hybridization solution containing final concentrations of 50% formamide, 4.1× Denhardt's solution, and 4.4× SSC. To reduce nonspecific hybridization, the hybridization solution also contained final concentrations of 15 μg human Cot1 DNA (Invitrogen), 12 μg polydeoxyadenylate, and 5 μg yeast tRNA (Sigma). After clarification by centrifugation, the cDNA/hybridization solution was applied to DNA microarrays and incubated at 42°C for 16 to 20 hours. After hybridization, microarray slides were washed by immersion in 2× SSC, 0.1% SDS for 5 minutes at room temperature, 1× SSC, 0.01% SDS for 5 minutes at room temperature, and 0.2× SSC for 2 minutes at room temperature. The microarrays were dried by centrifugation for 5 minutes at 1,000 rpm and scanned immediately with a ScanArray 4000 confocal laser system (Perkin-Elmer, Wellesley, MA). Fluorescence intensities were background subtracted, and normalization and filtering of the data were done using the QuantArray software package (Perkin-Elmer). After normalization, expression ratios were calculated for each feature.

Real-time PCR. Three micrograms total RNA from each of S2-013.gfp-neo, S2-013.MTII.C1, and S2-013.MTII.C2 were prepared as described in gene expression analysis and used for the reverse transcription reaction. Briefly, total RNA were incubated with 0.5 μg oligodeoxythimidylate for 10 minutes at 70°C, after cooling on ice they were incubated with 10× PCR buffer, 25 mmol/L MgCl₂, 20× deoxynucleotide triphosphate mix, and 0.1 mol/L DTT for 5 minutes at 42°C, then incubated with StrataScript reverse transcriptase for 50 minutes at 42°C and 15 minutes at 70°C to stop the reaction. After cooling on ice, RNase H was added. Assays-on-Demand Gene Expression products consisting of primers and TaqMan MGB probes (Applied Biosystems) were used for quantitative PCR reactions. The ABI PRISM 7700 sequence detection system was used and the reaction was incubated for 2 minutes at 50°C followed by 10 minutes at 95°C, then the reaction was run for 40 cycles at 15 seconds, 95°C and 1 minute, 60°C per cycle. Duplicate reactions for each sample were done and the standard curve method was used for analyzing data.

Mice. Congenitally athymic female National Cancer Institute nude mice (NCr-nu/nu) were purchased from the National Cancer Institute (Bethesda, MD). Animals were maintained in pathogen-free conditions and were fed sterile water and food ad libitum. Mice were treated in accordance with the Institutional Animal Care and Use Committee guidelines.

Tumor challenge. On the day of tumor challenge, adherent control cells (S2-013 and S2-013.gfp-neo) and MUC1-suppressed cells (S2-013.MTII.C1, C2) were removed from the tissue flask with 0.25% (w/v) trypsin solution containing 0.04% (w/v) EDTA in PBS for 5 minutes at 37°C, counted, and resuspended in DMEM at a concentration of 1 × 10⁷ viable cells/mL. Mice received 1 × 10⁶ viable cells by s.c. injection between the scapulae. Tumor growth was evaluated every 2 to 3 days, and tumor diameter was measured using a caliper. Mice were euthanized when the tumor volume reached 1,000 mm³. At this time, tumors were harvested and prepared for transplantation to recipient animals. For statistical analysis of tumor challenge, the regression method generalized estimating equations was used to account for the correlated nature of the data.

Metastasis assay in vivo by cecum implantation. Two independent metastasis assays were conducted within 60 days of tumor challenge. Briefly, donor tumor specimens were excised and inspected for necrotic tissue. Necrotic and suspected necrotic tissues were removed. Portions of the remaining specimen were fresh frozen and formalin fixed to characterize the donor tumors. The remaining specimens were divided into equal parts, squares of 2 mm in length and width, and kept in HBSS at 25°C (Life Technologies) until implantation into recipient animals. Recipient animals were anesthetized with a mixture of xylazine (35 mg/kg) and ketamine (120 mg/kg) by i.p. injection. The abdomens of recipient animals were sterilized by painting with betadine (The Purdue Frederick Company, Norwalk, CT). A small incision (5 mm) was made on the lower left quadrant and the cecum was exteriorized. One 2 mm tumor specimen was implanted onto the distal portion of the cecum with one 5-0 chromic gut surgical suture (Ethicon, Somerville, NJ). The cecum was carefully returned to the abdominal cavity, and the abdominal wall was closed with 5-0 nylon surgical suture (Ethicon). Postsurgery, animals were kept warm and observed until they recovered from anesthesia. The animals were observed regularly for the next 2 days. Tumor xenotransplants were allowed to grow for 5 weeks or until the animal became moribund (whichever occurred first). At this time, animals were sacrificed and examined for tumor invasiveness and metastasis.

Tumor, cecum, lungs, kidneys, liver, brain, and lymph nodes were harvested from each animal and fixed in a buffered formalin solution [(100 mL formalin, 3.4 g NaH2PO4, and 10.3 g Na₂HPO₄) / 1,000 mL (pH 7.4)] and embedded in paraffin. Serial 5 μm sections were cut and mounted on slides, and stained with H&E using standard procedures. The slides were examined under a microscope (Nikon E400; Nikon, Tokyo, Japan). Images were captured using a digital camera (Nikon CoolPix 950; Nikon). To analyze differences in the metastasis of different cell lines, logistic regression was used for binary outcomes and linear regression was used for continuous outcomes. Outcomes included tumor size, lymph vessel invasion, lymph node metastasis, liver metastasis, lung metastasis, and peritoneum dissemination. P < 0.05 was considered significant. All analysis was done using the SAS System for Windows, version 8.2 (SAS Institute).

Comparison of MUC1 expression between primary lesions and metastatic lesions in lung and lymph node. Primary and metastatic lesions from lymph node and lung were evaluated for MUC1 expression. MUC1 expression on primary tumors and metastatic lesions in lung and lymph node was evaluated on 5-μm-thick, paraffin-embedded tissue sections that were assayed using VECTOR M.O.M. Immnunodetection kit (Vector Laboratories, Inc., Burlingame CA). Briefly, each section was deparaffinized with xylene and endogenous peroxidase was blocked by incubating the sections in 3% hydrogen peroxide in tap water for 5 minutes. The sections were washed in 0.01 mol/L PBS (pH 7.4) twice and incubated for 1 hour in working solution of M.O.M Mouse Ig Blocking Reagent. The sections were washed with PBS twice and incubated 5 minutes in working solution of M.O.M diluent. Sections were incubated for 30 minutes with diluted primary antibody B27-29. After washing sections with PBS twice, the working solution of M.O.M. Biotinylated Anti-Mouse IgG Reagent was applied for 10 minutes. The slides were then incubated for 5 minutes at room temperature with VECTASTAIN Elite ABC Reagent, rinsed with PBS for 5 minutes twice and incubated for 3 to 5 minutes with 3,3′-diaminobenzidine substrate (Vector Laboratories) observing closely for color to develop, then counterstained with Meyer's hematoxylin for 30 seconds and coverslips were applied. The results were evaluated by the percentage of positively stained neoplastic cells. To compare the expression of MUC1 in a more comprehensive manner, the degree of positive expression was graded as follows: “1+,” 0% to 25% of neoplastic cells stained; and “2+,” 25% to 50% of neoplastic cells stained; and “3+,” 50% to 75% of neoplastic cells stained; and “4+,” 75% to 100% of neoplastic cells stained. For evaluation of the incidence of expression of mucin antigens, statistical analysis using Wilcoxon's signed rank test was done. A probability of P < 0.05 was considered statistically significant.

Survival assay. To investigate survival, we observed some animals for 10 weeks after cecum implantation. The numbers of mice evaluated were as follows: S2-013 = 4; S2-013.gfp-neo = 13; S2-013.MTII.C1 = 12; S2-013.MTII.C2 = 7. The procedure was identical for the metastasis assay except for the time span after implantation (as indicated in Results). Duration of survival was measured from the date of implant surgery to the date of death. Overall survival curves were calculated using the Kaplan-Meier method, and the differences were analyzed using the log-rank test. StatView software (Abacus Concepts, Inc., Berkeley, CA) was used for all statistical analysis.

Metastasis assay in vivo by orthotopic injection. S2-013.gfp-neo and S2-013.MTII.C2 were each injected into the pancreas of 17 or 18 mice, respectively, mice for this study. Preparation of cells for injections, anesthesia, and suture methods for peritoneum and skin were using the same methods as were described for the tumor challenge and cecum implantation studies. For orthotopic injection, a small incision (5 mm) was made on the upper left quadrant and 1 × 10⁴ viable cells in a volume of 20 μL were injected into pancreas. Mice were euthanized 6 weeks after injections. At this time, tumors and organs were harvested and prepared for histologic analysis as described above for the cecum implantation studies.

Results

Suppression of MUC1 expression by RNAi

Verification of MUC1 suppression in S2-013 cells. We established six independent cloned cell lines of S2-013 cells in which MUC1 was suppressed, which were named S2-013.MTII clones C1-C6 (e.g., S2013.MTII.C1, S2-013.MTII.C2, etc.). Suppression of MUC1 at the cell surface was shown by flow cytometry with mAb B27-29. MUC1 was expressed at the cell surface in S2-013 and S2-013.gfp-neo, but was suppressed in clones expressing the RNAi construct: S2-013.MTII.C1, S2-013.MTII.C2, S2-013.MTII.C3, S2-013.MTII.C4, S2-013.MTII.C5, and S2-013.MTII.C6 (Fig. 1 ).

Western blotting. Western blot analysis was done to verify suppression of MUC1 in S2-013 by RNAi. mAb B27-29 recognizes the tandem repeat moiety on the extracellular domain, which is repeated ∼40 times in the allelic form expressed by the cells used here (hence, the antibody is up to 40 times more sensitive than antibodies that recognize single epitopes per molecule), and CT-2 recognizes a single epitope on the cytoplasmic tail on the surface-associated/cytoplasmic tail domain. The results revealed the predicted >250,000 Da band for the extracellular domain and a series of bands around 30,000 Da for the processed surface-associated/cytoplasmic tail domain of MUC1. Expression of MUC1 was significantly decreased in all MUC1-suppressed cell clones (Fig. 1). Clones S2-013.MTII.C5 and S2-013.MTII.C6 expressed low levels of intracellular MUC1; clone S2-013.MTII.C2 expressed a very small quantity of intracellular MUC1; and clones S2-013.MTII.C1, S2-013.MTII.C3, and S2-013.MTII.C4 did not express detectable MUC1. We selected for further characterization clones S2-013.MTII.C1 and S2-013.MTII.C2 as representative of the spectrum of suppression (no expression and barely detectable expression of MUC1, respectively).

Properties of MUC1-suppressed cell growth in vitro

Cell proliferation assays. S2-013.MTII.C1 and S2-013.MTII.C2 showed proliferation rates in vitro that were significantly less (≤50%) than S2-013.gfp-neo (Fig. 2A ). Analysis by generalized estimating equation indicated a significant difference between S2-013.MTII.C1 and S2-013.MTII.C2 compared with S2-013.gfp-neo (P < 0.001).

Cell cycle and apoptosis assay by flow cytometry analysis. There were no significant differences in cell cycle progression or rates of apoptosis (mean values of three independent analyses) among S2-013 gfp-neo and MUC1-suppressed cell lines (S2-013.MTII.C1 and S2-013.MTII.C2) cultured in vitro. The G₀-G₁ ratios (SD) for S2-013 gfp-neo, S2-013.MTII.C1, and S2-013.MTII.C2 were 53.8% (30.2), 72.2% (22.7), and 65.8% (28.6), respectively (P = 0.42 for S2-013.gfp-neo versus S2-013.MTII.C1, P = 0.68 in S2-013.gfp-neo versus S2-013.MTII.C2). G₂-M ratios were 18.4% (8.2), 3.3% (3.1), and 0.9% (1.2; P = 0.14 in S2-013.gfp-neo versus S2-013.MTII.C1 and C2), and S-phase ratios were 28.0% (23.6), 24.4% (24.0), and 33.2% (29.8; P = 0.68 in S2-013.gfp-neo versus S2-013.MTII.C1, P = 0.99 in S2-013.gfp-neo versus S2-013.MTII.C2). The apoptosis indices were 7.4% (5.2), 5.0% (6.9), and 4.1% (4.7; P = 0.42 in S2-013.gfp-neo versus S2-013.MTII.C1, P = 0.28 in S2-013.gfp-neo versus S2-013.MTII.C2).

Adhesion assay. As expected, suppressing expression of MUC1 altered the adhesive properties of pancreatic tumor cells to defined components of the extracellular matrix. The binding of S2-013.MTII.C1 and S2-013.MTII.C2 to type I collagen and laminin were decreased, but adhesion to type IV collagen and fibronectin were increased compared with S2-013 and S2-013.gfp-neo control cells (Fig. 2B). For adhesion to type I collagen and type IV collagen, nonparametric Kruskal-Wallis tests indicated differences among the cell lines (P = 0.025 for type I, P = 0.002 for type IV, P = 0.060 for laminin, and P = 0.084 for fibronectin). Wilcoxon rank sum tests showed significant differences in adhesion of S2-013 versus S2-013.MTII.C1 (P = 0.008), S2-013 versus S2-013.MTII.C2 (P = 0.008) to type I collagen; S2-013 versus S2-013.MTII.C1 (P = 0.016), S2-013 versus S2-013.MTII.C2 (P = 0.008), S2-013.gfp-neo versus S2-013.MTII.C2 (P = 0.008) for adhesion to type IV collagen.

Effect of suppression of MUC1 on tumor growth in vivo

Models of tumor growth and metastasis in vivo. We evaluated tumor growth and metastasis in vivo at three different sites: by s.c. implantation of 10⁶ tumor cells into nude mice, by tumor transplantation from s.c. tumors to the cecum, and by orthotopic injection of tumor cells directly into the pancreas.

Subcutaneous tumor growth. S2-013.MTII.C1 and S2-013.MTII.C2 tumors showed significantly slower growth rates than control cells (S2-013 and S2-013.gfp-neo) at all time points (Fig. 2C). The slower growth rates resulted in significantly prolonged survival (Fig. 2D) for mice challenged with MUC1-suppressed cells (S2-013.MTII.C1 and S2-013.MTII.C2) compared with control cells (S2-013 and S2-013.gfp-neo; P < 0.0001, P < 0.0005 in S2-013.MTII.C1 and S2-013.MTII.C2). There were no significant differences between S2-013 and S2-013.gfp-neo, or between S2-013.MTII.C1 and S2-013.MTII.C2. Suppression of MUC1 expression was verified by immunohistochemistry on samples of each tumor.

Cecum tumor growth. We evaluated tumor growth and invasion at 5 weeks following cecum implantation of tumor fragments. The largest area of tumor growth in the cecum was measured for each mouse. The mean (SD) areas were (in mm) as follows: S2-013, 163.1 (73.3); S2-013.gfp-neo, 149.4 (65.7); S2-013.MTII.C1, 84.8 (34.1); and S2-013.MTII.C2, 94.6 (54.0). S2-013 and S2-013.gfp-neo tumors were significantly larger than S2-013.MTII.C1 (P < 0.001) and C2 (P < 0.001). No significant differences were found between S2-013 and S2-013.gfp-neo (P = 0.72).

Tumors grown as s.c. xenografts and transplanted to the cecum in athymic nude mice retained characteristics of a moderate to well-differentiated pancreatic adenocarcinoma, but it was notable that MUC1-suppressed cells showed fewer characteristics of differentiated morphology (ductal and tubular structures) than S2-013 and S2-013.gfp-neo cells (Fig. 3 ). MUC1 was expressed on the cell surface of the majority of S2-013 and S2-013.gfp-neo cells, but was suppressed in S2-013.MTII.C1 and S2-013.MTII.C2, indicating that the effectiveness of RNAi was generally maintained in vivo (Fig. 3 and data not shown). Nonetheless, MUC1 reexpression was detected in some lesions, especially those associated with an inflammatory infiltrates (Fig. 4A and B ) and during vessel invasion (Fig. 4C). Figure 4C shows that the primary tumors from challenge with MUC1-suppressed cells were moderately differentiated and exhibited low to no MUC1 expression, but upon invasion of lymphatic vessels, the tumor cells were morphologically well differentiated and exhibited higher MUC1 expression.

There was a trend that the degree of lymphatic and blood vessel invasion was decreased in MUC1-suppressed cells (S2-013.MTII.C1 and S2-013.MTII.C2); however, this did not achieve statistical significance between control cells (S2-013, S2-013.gfp-neo) and MUC1-suppressed cells (S2-013.MTII.C1 and S2-013.MTII.C2; P = 0.55). Typical findings of vessel invasion are shown in Fig. 5A .

Orthotopic (pancreas) tumor growth. The largest area of tumor growth in the pancreas was determined for each mouse. The mean (SD) areas were (in mm²) as follows: S2-013.gfp-neo, 44.75 (31.81), n = 17; S2-013.MTII.C1, 10.36 (14.4), n = 18, which showed that knockdown of MUC1 significantly reduced growth of tumors implanted orthotopically at the pancreas (P = 0.0002, unpaired t test).

Metastasis assays. Peritoneal organs, lungs, mesenteric lymph nodes, intrathoracic lymph nodes, brain, and liver were examined for distant metastases 5 weeks after cecum implantation (Table 1 ). No metastases to the brain were observed. Liver metastasis (Fig. 5D) was rarely seen for the cell lines and sites of implantation evaluated here. Both the S2-013 and S2-013.gfp-neo had statistically significantly higher incidences of lymph node metastasis (Fig. 5B) than S2-013.MTII.C1 and S2-013.MTII.C2 (Table 1). The incidences of metastasis to intrathoracic lymph nodes were as follows: S2-013.gfp-neo, 50% (5 of 10); S2-013.MTII.C1, 22.2% (2 of 9); and S2-013.MTII.C2, 7.1% (1 of 14). S2-013 and S2-013 gfp-neo caused significantly higher incidences of lung metastasis (Fig. 5C) than S2-013.MTII.C1 and S2-013.MTII.C2. There were no significant differences in the incidence of metastases to lymph node and lung among S2-013 and S2-013.gfp-neo. Multivariate logistic regression analysis was used to evaluate the cell lines for differences or correlations in tumor size and the probability of metastasis to specific sites. For lung metastasis, the P value among all cell lines was not significantly different at 0.1576, but larger tumor size was a major factor with a significance of P = 0.0047. For lymph node metastasis, the P value among control and MUC1-suppressed cell lines was significant at P = 0.0009, but tumor size was not significant at P = 0.8303. For peritoneal metastasis, the P value among control and MUC1-suppressed cell lines was significant at P = 0.0163, but tumor size was not significant at P = 0.085. Thus, the major contributing factor to lung metastasis was tumor size, but the major contributing factor to lymph node and peritoneal metastasis was expression of MUC1.

Analysis of metastasis in mice challenged orthotopically. The incidence of intrathoracic lymph nodes metastasis was 47% (8 of 17) for S2-013.gfp-neo, which decreased to 16% (3 of 18) for S2-013.MTII.C2 (P = 0.0750, Fisher's exact test). The incidence of lung metastasis was 80% for S2-013.gfp-neo and 100% for S2-013.MTII.C2 (not significant). The incidence of metastasis to spleen was 42% (5 of 17) for S2-013.gfp-neo, 7% (1 of 14) for S2-013.MTII.C2 (P = 0.0652, Fisher's exact test). The incidence of metastases to peritoneal sites was 70% for S2-013.gfp-neo, which decreased to 16% (3 of 18) for S2-013.MTII.C2 (P = 0.0020, Fisher's exact test). The incidence of metastasis to liver was 6% (1 of 17) for S2-013.gfp-neo, and 6% (1 of 18) for S2-013.MTII.C2 (not significant). Thus, results at the orthotopic site showed that suppression of MUC1 decreased the incidence of metastasis to lymph nodes, spleen, and peritoneal sites.

Comparison of MUC1 expression between primary tumor and metastatic lung and lymph nodes. We compared relative levels of expression of MUC1 between the primary lesions and the metastatic lesions. For expression of MUC1 in metastatic lung lesions compared with that at the primary lesions in each mouse: S2-013.gfp-neo (decrease, seven cases; no change, none; increase, none; Wilcoxon's signed-rank test, P = 0.018); S2-013.MTII.C1 (decrease, one case; no change, three cases; increase, none); S2-013.MTII.C2 (decrease, two cases; no change, three cases; increase, none). The expression of MUC1 in metastatic lymph node lesions was also compared with that at the primary lesions for each mouse as follows: S2-013.gfp-neo (decrease, eight cases; no change, three cases; increase, five cases), S2-013.MTII.C1 (decrease, none; no change, four cases; increase, one case), S2-013.MTII.C2 (decrease, five cases; no change, three cases; increase, four cases). These results show a trend of decreased MUC1 expression in the metastatic lung lesions compared with primary lesions, but increased MUC1 expression in some metastatic lymph node lesions compared with primary lesions.

Survival and incidence of metastasis following cecum implantation. We evaluated a series of mice for up to 10 weeks postimplantation to determine survival and incidence of metastasis at end stage of disease: 4 mice implanted with S2-013, 13 mice implanted with S2-013.gfp-neo, 12 mice implanted with S2-013.MTII.C1, and 7 mice implanted with S2-013.MTII.C2. Survival curves were calculated using the Kaplan-Meier method, and the differences were analyzed using the log-rank test (data not shown). Mice implanted with S2-013.MTII.C1 had statistically improved survival compared with S2-013.gfp-neo (P < 0.05). Similar trends were seen for S2-013 and S2-013.MTII.C2, which showed worse and better survival, respectively, although statistical significance was not achieved because of the relatively small numbers of animals analyzed.

Virtually all of the mice with end-stage disease exhibited lung and lymph node metastasis. There were no statistical differences in incidence of metastases among MUC1-suppressed cells and control cells at end-stage disease.

Gene expression analysis of MUC1-suppressed S2-013 cells. DNA oligonucleotide microarrays representing 12,261 genes were used to identify potential genes influenced by MUC1 suppression. MUC1-suppressed cells, S2-013.MTII.C1 and S2-013.MTII.C2, were compared with a control-transfectant S2-013.gfp-neo. Table 2 shows the genes consistently differentially expressed between S2-013 MTII.C1, S2-013.MTII.C2, and S2-013.gfp-neo. The results showed that a number of gene products that influence differentiation (ciliary neutrophic factor receptor), cell division (cyclins), proliferation (ITGA2), and metastatic propensity (CEACAM6) were differentially expressed among these cell types.

We evaluated the expression of one gene that showed a high differential of expression and was known to influence metastatic propensity, CEACAM6, by quantitative real-time PCR and Western blotting analysis (data not shown). The results showed that CEACAM6 was strongly suppressed in S2-013.MTII.C1 and S2-013.MTII.C2 compared with S2-013.gfp-neo and S2-013, which were comparable with the results obtained by oligonucleotide microarray gene expression analysis.

Discussion

In this study, we established stable clones of the S2-013 pancreatic cancer cell line in which MUC1 expression was suppressed by RNAi, and investigated the effects of reducing or eliminating MUC1 on the malignant and metastatic potential of these pancreatic cancer cells. Current theory holds that MUC1 is involved in configuring the cell surface properties of epithelial cells, associating with receptors for growth factors and differentiation factors at the cell surface, and signaling or modifying signal transduction in the cell in response to different cell surface conditions (adhesion, cellular polarity, and cellular neighborhood) or stimulation with growth or differentiation factors (9). Moreover, overexpression of MUC1 in pancreatic cancer is associated with poor prognosis (10, 11). Thus, suppressing expression of MUC1 in pancreatic tumor cells would be predicted to affect different aspects of the malignant and metastatic properties of cancer, including proliferation, invasion, metastasis, or differentiation status.

One notable effect of reducing or eliminating MUC1 RNA and protein in S2-013 cells was a decrease in the proliferation rate in vitro and a reduction in the rate of tumor growth in vivo. Consistent with the results presented here, a previous study showed statistically significant (P < 0.0001) increases in the growth rates of S2-013 clones overexpressing a recombinant form of MUC1, S2-013.MUC1F, compared with clones expressing endogenous amounts of MUC1 (S2-013.NEO). These results suggest that MUC1 expression influences proliferation rates in S2-013 cells.⁴

One possible explanation for the finding that decreased MUC1 expression slowed tumor progression would be an increase in the steady-state levels of apoptosis of pancreatic tumor cells. There are reports showing that MUC1 activates antiapoptosis pathways (12, 13). In addition, Ren et al. (5) showed that MUC1 attenuates apoptotic responses to DNA damage in that MUC1-suppressed cells were more sensitive to DNA-damaging agents. A recent report showed that MUC1 can associate with p53 when cells are challenged with DNA-damaging agents, and that it up-regulates the p53-dependent growth arrest response and suppresses the p53-dependent apoptotic response to DNA damage (14). Oligonucleotide gene expression analysis presented here showed that several genes related to proliferation and apoptosis were up-regulated in MUC1-suppressed cells, including death-associated protein kinase-related 1, which is thought to be involved in triggering of apoptosis (15); cyclin B, which is believed to regulate the entry into mitosis (16); and cyclin D3, which is thought to be essential for G₁ progression and is believed to play a dual role in cell proliferation and in establishment and and/or maintenance of terminal differentiation (17). However, with decreasing levels of MUC1 in the pancreatic tumor cell lines, we did not observe an increase in the steady-state apoptosis index or in the expression of genes regulated by p53 in response to DNA damage (14). Moreover, the levels of cyclin B and D3 protein were not significantly altered in S2-013 control cells, S2-013.MUC1F, and S2-013.MTII.C1 and S2-013.MTII.C2 (data not shown). Thus, the decreased proliferation rates observed when MUC1 is suppressed in pancreatic tumor cells cannot be explained by our current understanding of the contribution of cyclin gene product expression or of the reported influence of MUC1 on antiapoptotic pathways. This leads us to conclude that MUC1 influences proliferation of pancreatic tumor cells by other as yet undefined mechanisms.

A second significant finding of this study is that down-regulation of MUC1 decreased the metastatic capacity of the S2-013 cell line in vivo. We observed suppression of metastasis to lymph nodes and lung in MUC1-suppressed cells implanted at the cecum (S2-013.MTII.C1 and S2-013.MTII.C2), and suppression of metastasis to the lymph node, spleen, and peritoneal cavity when S2-013.MTII.C2 was orthotopically implanted in the pancreas. There are at least two explanations for the deceleration of metastasis by MUC1 suppression: one possibility is that MUC1 contributes to the cell surface properties of tumor cells and influences their ability to adhere, migrate, and/or survive in different organ environments; a second possibility is that the decreased tumor size over time that results from a decrease in the proliferation rate yields fewer tumor cells that are capable of seeding metastases. These factors are not mutually exclusive and both could contribute to the metastatic process. The studies described here provide evidence that both processes influence metastasis of pancreatic tumor cells.

With respect to tumor size, we found that S2-013 tumors allowed to progress to end-stage disease (10 weeks) showed no differences in metastatic lesions in the peritoneal cavity and in the lungs. Thus, blocking MUC1 expression does not eliminate the metastatic potential of these cells as primary tumors progress to large volumes. Nonetheless, these animals showed significantly improved survival compared with animals challenged with tumors in which MUC1 was not suppressed. This is consistent with clinical findings, which suggest that increased tumor growth rates (18–20) or MUC1 expression (21, 22) are associated with poor prognosis in many types of cancers.

We also investigated whether eliminating MUC1 expression altered the binding properties of tumor cells to known extracellular ligands for tumor cells. Eliminating MUC1 altered the cell surface properties of the tumor cells. Adhesion to type IV collagen and fibronectin were increased in MUC1-suppressed cells, whereas adhesion to laminin and type I collagen were decreased. This is partially consistent with the results that Satoh et al. (23) reported with respect to the effects of MUC1 overexpression in MUC1F S2-013 compared with S2-013 controls. The results of adhesion of MUC1-suppressed cells to laminin and type I collagen are not consistent with the previous results, which, in contrast, showed increased binding to cells expressing lower levels of MUC1. Taniguchi et al. (24) reported that type I collagenolytic activity of SUIT-2 and its sublines was correlated with incidence of lung metastasis. In any case, suppressing MUC1 is predicted to have at least two consequences. The direct effect is that the binding properties of the tumor cells to extracellular matrices are altered, which would be predicted to affect the process of invasion and metastasis. A secondary effect would be that these alterations would affect the ability of laminin or other extracellular ligands to induce signal transduction through integrin or other cell surface receptors. Tarbe et al. (25) previously reported that expression of laminin receptors is one of the most remarkable features of metastatic lesions in orthotopic pancreatic tumor xenograft models using severe combined immunodeficient mice. Our oligonucleotide gene expression analysis showed that integrin α2 was down-regulated in our MUC1-suppressed cells.

We obtained additional evidence from oligonucleotide microarray analysis that there were alterations in the transcriptional profile of cells in which MUC1 was suppressed. These alterations are predicted to be downstream effects of altered signaling that is directly affected by MUC1, such as when the MUC1 cytoplasmic tail binds to β-catenin and affects the transcriptional coactivation of genes downstream of canonical Wnt signaling pathways (which include cyclin D1 as a target; ref. 26). Changes in gene expression could also result from alterations in other signaling mediated by molecules with which MUC1 interacts, or from signaling that results from other cell surface molecules that are blocked by MUC1 expression and that become accessible upon reduction of cell surface MUC1 expression (9).

Prominent among the alterations in gene expression profiles we observed was CEACAM6, which was consistently and strongly reduced in MUC1-suppressed cells (S2-013.MTII.C1 and S2-013.MTII.C2). This difference was confirmed by real-time PCR and Western blotting analysis (data not shown). CEACAM6 is a member of the CEA family and a glycosylphosphatidylinositol-linked cell surface protein (27). Recent reports show that this molecule is associated with a poor prognosis for pancreatic cancer and colon cancer in vitro and in vivo (28–32). Duxbury et al. (28) reported that posttranscriptional inhibition of CEACAM6 expression inhibits the ability of pancreatic adenocarcinoma cells to form experimental liver metastases in vivo and that systemic CEACAM6-specific siRNA could decrease proliferation and metastasis in murine models (32). We also observed suppression of expression of phosphatidylinositol synthase in MUC1-suppressed cells, an enzyme that creates the membrane anchor for CEACAM6 (27, 33). Additional investigation is required to determine the relationship between CEACAM6, phosphatidylinositol synthase, and MUC1 expression at the cell surface. We analyzed the promoter regions of CEACAM6 by MatInspector 6.0 (Genomatix Software, Munich, Germany). Seventy-eight transcription factor–binding sites were identified, among them TCF/LEF-1(34) and myoD (35), which are involved in Wnt signaling pathways. These results suggest that MUC1 may influence transcription of CEACAM6 by its interactions with β-catenin and affects on Wnt signaling pathways.

Recently, the problem of “off-target” effects from RNAi has been documented (36). In studies not presented here, we established a separate line of MUC1-suppressed S2-013 cells (S2-013.MTI.Sel.) with a different target (target 1: 5′-GTTCAGTGCCCAGCTCTAC-3′, 125 downstream of initiation codon). These cells showed similar properties to those reported here. Gene expression analysis of S2-013.gfp-neo versus S2-013.MTI.Sel showed that CEACAM6 was down-regulated in S2-013.MTI.Sel. These data strongly support our conclusion that the effects of this study were not caused by off-target effects, but were the result of altering MUC1 expression in these cells.

Our investigation of primary lesions and metastatic lesions by immunohistochemistry showed different staining patterns of MUC1 for lung metastases and lymph node metastases. Metastatic lung lesions showed low expression of MUC1 compared with primary tumors, although MUC1-suppressed groups showed lower incidences of lung metastasis. These results raised at least two possibilities: that lung metastasis is influenced by primary tumor growth rate and volume and/or that decreased MUC1 expression is favorable for metastasis to the lung. We therefore analyzed the factors that contribute to metastasis by multivariate logistic regression analysis. After controlling for tumor size, there were no differences among the cell lines with respect to lung metastasis. There was, however, an effect associated with tumor size: Larger tumors were associated with lung metastasis. Because metastasis to lung is believed to be primarily associated with a hematogenous route, we conclude that lung metastasis results from the increased incidence of hematogenous metastasis that accompanies the development of large tumors with extensive neovascularization.

In contrast, S2-013.gfp-neo metastases in lymph nodes expressed higher levels of MUC1, and MUC1 expression was restored in the few lymph node metastases that were detected for the MUC1-suppressed cell lines, although these differences did not quite achieve statistical significance because of the relatively small numbers of animals analyzed here. Multivariate logistic regression analysis showed clearly that MUC1 suppression reduced lymph node metastasis and peritoneal metastasis, but that metastases to these sites were not affected by tumor size. Taken together, the different expression patterns of MUC1 during lung metastasis and lymph node metastasis, and the association of tumor size with metastatic propensity to lung, suggest that different mechanisms regulate metastasis to lung and lymph nodes and that MUC1 plays a direct role in regulating lymph node metastasis.

One potentially important immunohistologic finding was that we detected low levels of MUC1 reexpression in MUC1-suppressed cells associated with inflammatory cellular infiltrates in the primary tumor and invasive sites. The expression levels of MUC1 at these sites were substantially less than controls. This finding raises the possibility that the mechanism by which RNAi down-regulates gene expression may be inhibited in vivo under conditions of inflammation or that inflammation induces MUC1 expression beyond levels that can be inhibited by RNAi. This point should be examined carefully in future studies because it may portend general effects that greatly affect the overall therapeutic use of RNAi strategies.

It was notable that primary tumors of MUC1-suppressed cells showed lower states of morphologic differentiation than controls, including S2013 or S2-013.gfp-neo. In some lesions, especially during vessel invasion, the suppressed cells showed MUC1 reexpression and commensurately higher states of morphologic differentiation.

Among pancreatic cancer cell lines, Panc1 and HGC25 are established as poorly differentiated adenocarcinoma cell lines (37, 38) that express low levels of MUC1. HPAF and Colo 357 are established as well-differentiated adenocarcinoma cell lines (39, 40) that express abundant MUC1 (41). The original report of cloning the SUIT cell lines (42) presented 28 clones that were established from the parent cell line, which differed in morphologic differentiation status from well to poorly differentiated. The differentiation grade of these cell lines was in direct proportion to levels of CA19-9 on the cell surface. CA19-9 recognizes the sialyl-Lewis A (sLe^a) epitope, which is attached to MUC1 (and other mucins) on the cell surface of pancreatic tumor cell lines. Thus, there is substantial evidence that MUC1 expression is associated with increased states of morphologic differentiation.

We found here that expression of several differentiation-related genes were altered upon suppression of MUC1. For example, ciliary neutrophic factor receptor was down-regulated in S2-013.MTII.C1 and S2-013.MTII.C2. Ciliary neutrophic factor is a member of the interleukin-6 family that is known to promote cell survival and differentiation in neuronal cells, although there are no reports of studies on the role of this gene in differentiation of epithelial cells. In addition, CEACAM6 was reported to inhibit differentiation when overexpressed in colonocytes, and the glycophosphatidylinositol anchor of CEA has been associated with its differentiation blocking activity (43).

In conclusion, suppressing MUC1 expression in pancreatic cancer cells influenced their growth rate and metastatic propensity. This likely resulted from altering the contribution of MUC1 to maintenance of the cell surface environment and effects on morphogenetic signal transduction. Given that MUC1 is also expressed by normal pancreatic cells, it is important to understand the roles of MUC1 in normal epithelial cells and to understand how these are common and different in cancer cells. MUC1 expression may contribute beneficially to cancer cells in some locations or situations and may negatively affect their ability to colonize and grow at other sites. Nonetheless, on balance, we found that suppressing MUC1 in S2-013 cells decreased their growth and malignant potential by direct effects on cell adhesion and growth properties and by influencing expression of other factors that contribute to poor prognosis for pancreatic cancer, including CEACAM6, insulin-like growth factor binding protein-3, or other factors (44). This leads us to conclude that suppression of MUC1 remains a potentially useful strategy to treat pancreatic cancer.