Expression of Core 2 β-1,6-N-Acetylglucosaminyltransferase in a Human Pancreatic Cancer Cell Line Results in Altered Expression of MUC1 Tumor-associated Epitopes*

Many tumor-associated epitopes possess carbohydrate as a key component, and thus changes in the activity of glycosyltransferases could play a role in generating these epitopes. In this report we describe the stable transfection of a human pancreatic adenocarcinoma cell line, Panc1-MUC1, with the cDNA for mucin core 2 GlcNAc-transferase (C2GnT), which creates the core 2 β-1,6 branch in mucin-type glycans. These cells lack endogenous C2GnT activity but express a recombinant human MUC1 cDNA. C2GnT-transfected clones expressing different levels of C2GnT were characterized using monoclonal antibodies CC49, CSLEX-1, and SM-3, which recognize tumor-associated epitopes. Increased C2GnT expression led to greatly diminished expression of the CC49 epitope, which we identified as NeuAcα2,6(Galβ1,3)GalNAcα-Ser/Thr in the Panc1-MUC1 cells. This was accompanied by the emergence of the CSLEX-1 epitope, sialyl Lewis x (NeuAcα2,3Galβ1,4(Fucα1,3)GlcNAc-R), an important selectin ligand. Despite this, however, the C2GnT transfectants could not bind to selectins. Increased C2GnT expression also led to masking of the SM-3 peptide epitope, which persisted after the removal of sialic acid, further suggesting greater complexity of the core 2-associated O-glycans on MUC1. The results of this study suggest that C2GnT could play a regulatory role in the expression of certain tumor-associated epitopes.

ferases, which compete with one another for common acceptor structures (3). When competing glycosyltransferases share the same Golgi localization, the final O-glycan structure is likely to be governed primarily by the relative activities of the enzymes. However, if the enzymes reside in different Golgi compartments, the earlier Golgi enzyme will have an advantage in dictating the oligosaccharide structure.
Glycosyltransferase competition can occur following the creation of the core 1 acceptor structure Gal␤1-3GalNAc␣-Ser/Thr. UDP-GlcNAc:Gal␤1-3GalNAc (GlcNAc to GalNAc) ␤-1,6GlcNAc-transferase (C2GnT; 1  Creation of the core 2 branch enables the O-glycan chain to be extended into complex structures such as polylactosamine chains (reviewed in Ref. 4). The sialyl transferases ST6GalNAc I and II compete with C2GnT for the core 1 acceptor substrate (5,6) and direct the glycosylation pathway toward simpler structures lacking the core 2 branch, such as NeuAc␣2,6-(Gal␤1,3)GalNAc and NeuAc␣2,6GalNAc. The latter structures are recognized by the monoclonal antibody CC49 (7,8), and therefore creation of the CC49 epitope should be influenced by the relative activities as well as the specific Golgi localization of both C2GnT and these sialyl transferases.
C2GnT activity is regulated under certain growth conditions, including maturation of granulocytes (9) and T cells (10) as well as T cell activation (11). Transgenic mice in which C2GnT was overexpressed showed normal T cell development but an * This work was supported by National Institutes of Health Grants RO1 HL48282, RO1 CA57362, RO1 CA69234, and P30 CA36727. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. § Current address: Regional Research Laboratory, Jammu-Tawi-180001, India.
The mucin MUC1 is a type I membrane-bound glycoprotein, which is aberrantly glycosylated in many cancer tissues (reviewed in Ref. 24), displaying several tumor-associated epitopes. For example, monoclonal antibodies SM-3 and HMFG-2 recognize peptide epitopes within the MUC1 tandem repeat region and react preferentially with MUC1 in cancer tissues (25)(26)(27) where aberrant glycosylation results in epitope exposure.
The purpose of this study was to examine the effects of C2GnT on the expression of certain MUC1 tumor-associated epitopes. We stably transfected the human pancreatic adenocarcinoma cell line Panc1-MUC1, which expresses a recombinant human MUC1 cDNA, with a bovine C2GnT cDNA. We found that increased expression of C2GnT resulted in de novo expression of the sLe x epitope. However, this did not render the cells capable of binding selectins. C2GnT expression also led to the elimination of the CC49 epitope. Furthermore, the SM-3 and HMFG-2 tumor-associated MUC1 peptide epitopes were masked by high levels of C2GnT expression. In summary, introduction of C2GnT into a cancer cell line significantly altered the expression of MUC1 tumor-associated epitopes by shifting the mucin-type glycosylation pathway toward more complex O-glycans. This suggests a potential regulatory role for C2GnT in the generation of tumor-associated epitopes.

EXPERIMENTAL PROCEDURES
Materials-The Panc1 cell line was purchased from the ATCC (Manassas, VA). S2-013 is a subline of a human pancreatic tumor cell line derived from a liver metastasis (28). UDP-[ 14 C]GlcNAc, UDP-[ 3 H]Gal, and CMP-[ 3 H]NeuAc were from American Radiolabeled Chemicals (St. Louis, MO). Asialo ovine submaxillary mucin (aOSM) was prepared as described (29). FPDG was isolated as described (30) from the serum of the antarctic fish Dissostichus mawsoni, provided by Dr. Arthur DeVries at the University of Illinois at Champaign-Urbana. E-, P-and L-selectin/IgM chimeras were provided by Dr. John Lowe at the University of Michigan, Ann Arbor. Transferrin (iron-saturated) was from Collaborative Biomedical Products (Bedford, MA). Immobilon P polyvinylidene difluoride membrane was from Millipore (Bedford, MA). Bond Elut C18 cartridges were from Varian (Sunny Vale, CA). M2 anti-FLAG monoclonal antibody was from Kodak IBI. CC49 monoclonal antibody was a gift from Dr. David Colcher at the University of Nebraska Medical Center. CSLEX-1 monoclonal antibody was obtained from the ATCC. HMFG-2 and SM-3 monoclonal antibodies were gifts of Dr. Sandra Gendler at the Mayo Clinic, Scottsdale, AZ. C2GnT monoclonal antibody B5-1 was obtained as described (31). Biotin-conjugated lectins DBA and PNA were obtained from EY Laboratories (San Mateo, CA). Other chemicals were from Sigma unless otherwise noted.
Cell Culture-Panc1 and Panc1-MUC1 cells were grown in minimal essential medium supplemented with 5% fetal bovine serum and antibiotics (50 units/ml penicillin and 50 g/ml streptomycin). Panc1-MUC1 cells stably transfected with C2GnT cDNA were grown in minimal essential medium supplemented with 5% fetal bovine serum and 300 g/ml Zeocin.
Stable Transfection of Panc1-MUC1 Cells with Bovine C2GnT cDNA-A 1.6-kilobase fragment of bovine C2GnT cDNA (containing the complete open reading frame of C2GnT) (31) was subcloned into the mammalian expression vector pcDNA 3.1/Zeoϩ (Invitrogen) using KpnI and NotI. Transfection of pcDNA 3.1-C2GnT into Panc1-MUC1 cells was performed using a transferrin-assisted lipofection protocol as described previously (32), and clones, which stably express C2GnT, were selected based on resistance to the antibiotic Zeocin.
Glycosyltransferase Enzyme Assays-Enzyme assays were carried out on total cell homogenates prepared by washing confluent monolayers of the cells twice with cold PBS, scraping the cells off the flask in 0.25 M sucrose, and disrupting the cells by successive passage of the sucrose suspension through 18-, 20-, and 25-gauge syringe needles. Protein concentration was measured by the Bio-Rad assay (Bio-Rad) using BSA as standard. All assays were conducted at least in duplicate under conditions in which product formation was linear with respect to time and enzyme amount. An additional reaction without exogenous acceptor was performed to measure endogenous enzyme activity. Enzyme activity was calculated by subtracting endogenous activity from total activity and was expressed as nmol of sugar donor transferred/ hour/mg protein. GalNAc TF, which catalyzes the attachment of Gal-NAc in ␣-linkage to serine or threonine of the mucin peptide, was assayed as described (33), using a synthetic 29-amino acid MUC2 peptide as acceptor having the sequence PTTTPITTTTTVTPTPTPTGTQT-PTTTPI. Core 1 GalTF, which catalyzes attachment of galactose in ␤-1,3 linkage to GalNAc␣-Ser/Thr, was assayed as described (29) using aOSM as acceptor. C2GnT activity was assayed as described (34) using Gal␤1-3GalNAc␣-Bzl as acceptor. ST6GalNAc I, which catalyzes attachment of neuraminic acid in ␣-2,6 linkage to GalNAc in GalNAc␣-Ser/Thr and Gal␤1-3GalNAc␣-Ser/Thr (6), was assayed as described (35) using aOSM as acceptor. ST6GalNAc II, which catalyzes attachment of neuraminic acid in ␣-2,6 linkage to GalNAc in Gal␤1-3Gal-NAc␣-Ser/Thr (but not in GalNAc␣-Ser/Thr) (5), was assayed as described (35) using FPDG as acceptor. FPDG also acts as an acceptor for ST6GalNAc I and ST3Gal I, and therefore care must be taken in interpreting assay results using this acceptor. ST3Gal I, which catalyzes attachment of neuraminic acid in ␣-2,3 linkage to Gal of Gal␤1-3GalNAc␣-Ser/Thr (36), was measured in the same manner as ST6GalNAc I and II, except Gal␤1-3GalNAc␣-Bzl was used as acceptor. ST6GalNAc I and II cannot utilize Gal␤1-3GalNAc␣-Bzl as acceptor (5,6), thus preventing their interference with the measurement of ST3Gal I activity.

Preparation of Cell Lysate for Western Blotting and Lectin
Blotting-Confluent cells were washed twice with cold PBS and scraped from culture flasks in lysis buffer (500 l for a T25 flask) containing 10 mM Tris-HCl, pH 8.0, 150 mM NaCl, 1 mM phenylmethylsulfonyl fluoride, and 1% Triton X-100, using a rubber cell scraper. Following a 40-min incubation on ice, lysates were centrifuged at 2000 rpm for 2 min in a microcentrifuge to pellet cell debris. Supernatant was then transferred to a new tube and stored at Ϫ20°C.
Glycosidase Treatment of Cell Lysates-Cell lysates (300 g of total protein) prepared as described above were treated with 100 milliunits of Clostridium perfringens neuraminidase in a total volume of 200 l for 3.5 h at 37°C in 0.05 M sodium acetate, pH 5.5. One-third of the neuraminidase-treated lysates was exposed to 1 unit of peptide Nglycosidase F in 200 mM sodium phosphate, 10 mM EDTA, pH 7.2, at 37°C for 18 h. Another third of the neuraminidase-treated lysates was treated with 10 milliunits of Diplococcus pneumoniae ␤-galactosidase in 200 mM sodium cacodylate, pH 6.0, at 37°C for 44 h. Samples were stored at Ϫ20°C after treatment prior to SDS-polyacrylamide gel electrophoresis analysis and Western blotting or lectin blotting.
Immunoblotting-For Western blot analysis of MUC1 and C2GnT, proteins in cell lysates were resolved by 6% SDS-polyacrylamide gel electrophoresis (with 3% polyacrylamide stacking gels) and 10% SDSpolyacrylamide gel electrophoresis (with 6% polyacrylamide stacking gels), respectively. Protein was electroblotted to Immobilon P polyvinylidene difluoride membrane overnight at 300 mA, then blocked in 5% nonfat milk in TBS (0.9% NaCl, 10 mM Tris, pH 7.5) at room temperature for 1 h. The MUC1 blots were then probed for 1 h at room temperature with various primary antibodies in 5% nonfat milk in TBS, whereas the C2GnT blot was probed with the C2GnT antibody B5-1 diluted 1:2500 in TBS, 1% BSA. Membranes were then washed 15 min in 5% nonfat milk in TBS (for MUC1 blots) or TBS, 1% BSA (for C2GnT blots) followed by two additional 5-min washes with fresh wash mixtures. The membranes were then exposed for 1 h at room temperature to peroxidase-conjugated goat anti-mouse IgG/IgM secondary antibody diluted 1:2000 in 5% nonfat milk in TBS (for MUC1 blots) or TBS, 1% BSA (for C2GnT blots). Washes were repeated as described above, and then ECL reagents (Pierce) were applied per the manufacturer's instructions; the blots were then exposed to ECL-sensitive film (Amersham Pharmacia Biotech).
Lectin Blotting-Proteins in whole cell lysates were resolved by 6% SDS-polyacrylamide gel electrophoresis (with 3% polyacrylamide stacking gels). Protein was electroblotted to Immobilon P polyvinylidene difluoride membrane overnight at 300 mA and blocked in 2% BSA (fraction V) in PBS at room temperature for 1 h. The blots were probed for 1.5 h at room temperature with biotin conjugates of either DBA or PNA 1:500 in TBT (TBS ϩ 0.1% BSA ϩ 0.025% Tween 20). Membranes were washed three times, 5 min each, in TBT. Next, the membranes were exposed for 1 h at room temperature to peroxidase-conjugated streptavidin diluted 1:1000 in TBT. Washes were repeated as described above. ECL reagents were applied as per the manufacturer's instructions, and the blots were exposed to ECL-sensitive film.
Imunoprecipitation of MUC1 Using Antibody Against the MUC1 Tandem Peptide Repeat-500 l of Panc1-MUC1 C2#5 total cell lysate was incubated with 200 l of antibody HMFG-2 at 4°C with mild agitation for 3 h. Protein G-Sepharose (150 l) was then added, and the mixture was incubated for 16 h at 4°C with mild agitation. The immunoprecipitate was washed three times with 1 ml of cold PBS. SDSpolyacrylamide gel electrophoresis sample buffer was added, and the mixture was boiled for 5 min; the supernatant was resolved on 6% SDS-polyacrylamide gel electrophoresis.
Flow Cytometric Analysis of Selectin Binding Ability of Panc1-MUC1 C2#7 Cells-S2-013 cells and Panc1-MUC1 C2#7 cells were grown to approximately 90% confluence and released from the tissue culture flask by incubation for 30 -60 min in PBS containing 0.5 mM EDTA and 0.1% BSA. Aliquots containing 5 ϫ 10 5 cells were pelleted in wells of a 96-well plate and washed with staining medium (Dulbecco's modified Eagle's medium containing 0.1% BSA and 0.1% sodium azide). The cells were then stained for 1 h on ice with IgM-conjugated P-, E-, or Lselectin in staining medium with or without 5 mM EDTA added. Cells were washed twice with staining medium (with or without 5 mM EDTA, as appropriate) and stained 1 h on ice in the dark with 10 g/ml fluorescein isothiocyanate-conjugated goat anti-human IgM in staining medium (with or without 5 mM EDTA, as appropriate). Cells were washed twice with staining medium and fixed for 10 min in 2% formaldehyde. The cells were resuspended in PBS containing 0.1% BSA and 0.1% sodium azide and analyzed on a Becton Dickinson FACScan or FACStar Plus .

Stable Transfection of C2GnT cDNA into Panc1-MUC1 Cells and Selection of Clones for Characterization-MUC1-trans-
fected Panc1 (Panc1-MUC1) cells were chosen as the host for stable transfection of C2GnT because this cell line lacks endogenous C2GnT expression but expresses core 1 Gal TF, which creates the acceptor substrate utilized by C2GnT. These cells also express a recombinant human MUC1 bearing a FLAG epitope (Fig. 1), which provides for ease of detection and purification of the MUC1. Furthermore, MUC1 in the Panc1-MUC1 cells has been previously characterized with a panel of antibodies recognizing specific carbohydrate antigens and was found to express the tumor-associated epitope recognized by CC49 (28), which does not possess a core 2 branch (7,8).
Following transfection of the C2GnT cDNA into Panc1-MUC1 cells, stable clones were isolated and screened for expression of C2GnT. Three clones representing a wide range of C2GnT expression (Fig. 2) were chosen for characterization of the effects of C2GnT on MUC1. The three clones displayed an approximate 40-fold difference in C2GnT activity between the high expressing (C2#7) and low expressing (C2#5) clones ( Fig.  2A). C2GnT expression levels in the clones assayed via Western blotting gave results consistent with those of the C2GnT enzyme assays (Fig. 2B).

Expression of C2GnT
Leads to the Appearance of sLe x but Does Not Confer Selectin Binding Ability on the Cells-The presence of the core 2 branch in mucin-type glycans has been implicated in the generation of the sLe x epitope (13)(14)(15)(16), and transfection of C2GnT cDNA into a cell line, which expresses the P-selectin ligand P-selectin glycoprotein ligand-1, rendered the cells capable of binding to P-selectin (37,38). Therefore, we examined whether expression of C2GnT in Panc1-MUC1 cells could generate the sLe x epitope. Immunoblotting with CSLEX-1, which recognizes sLe x (NeuAc␣2-3Gal␤1-4(Fuc␣1-3)Glc-NAc-R) (17), revealed that the sLe x epitope was not present in Panc1-MUC1 parental cells or in the low C2GnT-expressing clone C2#5. However, the epitope appeared in clones C2#14 and C2#7, which have higher C2GnT expression (Fig. 3A). MUC1 expression levels in the clones varied little (Fig. 3B), showing that the changes in sLe x detection were not because of different amounts of MUC1 in the samples.
Because we detected sLe x on MUC1 in the high C2GnTexpressing transfectants, we tested these clones for their ability to bind to P-, E-, and L-selectins using IgM-selectin fusion proteins followed by staining with a fluorescein isothiocyanateconjugated anti-IgM secondary antibody and subsequent flow cytometry analysis. Despite the presence of the sLe x epitope, no binding of the C2GnT transfectants to the selectins was detectable under conditions in which the positive control cell line, S2-013, was found to bind to the selectins (data not shown).
The CC49 Epitope on MUC1 in Panc1-MUC1 Cells Consists of a Trisaccharide Which Lacks the Core 2 Branch-Because CC49 can recognize both NeuAc␣2,6(Gal␤1,3)GalNAc and NeuAc␣2,6GalNAc (7, 8), we sought to identify which of these was present on MUC1 in the Panc1-MUC1 parental cells, as detected previously by Burdick et al. (28). We performed lectin blotting using DBA and PNA on Panc1-MUC1 cell lysates treated or untreated with neuraminidase. As shown in Fig. 4, A and B, lanes 1 and 2, PNA reacted intensely with MUC1 in neuraminidase-treated lysate, whereas DBA showed no reaction. Because DBA is specific for ␣-linked GalNAc, whereas PNA recognizes Gal␤ 1-3 GalNAc, these results strongly suggest the presence of Gal␤ 1-3 (NeuAc ␣ 2-6) GalNAc, rather than sialyl Tn (NeuAc␣ 2-6GalNAc), on MUC1 in the Panc1-MUC1 parental cells. To rule out the possibility that the struc-ture recognized by PNA is located on asparagine-linked glycans on MUC1 rather than on mucin-type glycans, we treated desialylated lysate with peptide N-glycosidase F to remove Nlinked chains. No change was seen in the intensity of PNA recognition (Fig. 4, A and B, lane 3), confirming that the recognized structure is present on O-glycans. Finally, because PNA can react with terminal ␤1,4-linked galactose in addition to Gal␤ 1-3 GalNAc, we treated desialylated lysate with D. pneumoniae ␤-galactosidase, which cleaves ␤-1,4-linked galactose (but not ␤-1,3-linked galactose). Once again, no change was seen in PNA recognition (Fig. 4, A and B, lane 4), further supporting Gal ␤-1-3GalNAc as the structure recognized by PNA. From these results, we conclude that the structure detected by CC49 in the Panc1-MUC1 parental cells consists mainly of the trisaccharide Gal ␤-1-3 (NeuAc␣ 2-6) GalNAc. This conclusion is supported by our detection of substantial core 1 GalTF enzyme activity in Panc1 cells (Table I), which synthesizes Gal␤-1-3GalNAc.
C2GnT Expression Reduces the Expression of the Tumorassociated Epitope Recognized by CC49 -CC49 was raised against the tumor antigen tumor-associated glycoprotein-72 (39). CC49 is highly tumor-selective, reacting with a high percentage of malignant cells from various carcinomas but with very few normal tissues (39,40). Because the CC49 epitope lacks a core 2 branch, we hypothesized that C2GnT expression could lead to altered expression of this epitope. CC49 immunoblotting showed heavy expression of the CC49 epitope in the Panc1-MUC1 cells (Fig. 5A), confirming the previous report of Burdick et al. (28). Increased C2GnT expression led to decreased expression of the CC49 epitope, and the epitope was abolished in the highest C2GnT-expressing transfectant, C2#7. Again, these results were not because of differences in the  The linkage shown is that catalyzed by ST6GalNAc II using FPDG. However, FPDG is also utilized by ST6GalNAc I and ST3Gal I. The fact that ST6GalNAc I and ST3Gal I activities were the same in Panc1 and Panc1-MUC1 C2#7, as was the total activity measured using FPDG, suggests that ST6GalNAc II activity was also unchanged. c ND, not detectable (Ͻ0.5 nmol/h/mg for C2GnT assay). amount of MUC1 present in the samples, as shown in Fig. 5B. Analysis of MUC1 O-glycans via ␤-elimination and gel filtration chromatography using Bio-Gel P-4 (Bio-Rad) showed that transfection with C2GnT led to a nearly complete disappearance of the lower-MW MUC1 O-glycans (data not shown). This suggests that the reduction in the CC49 epitope was caused by a complete substitution of the O-glycans by C2GnT, instead of by only a partial substitution causing reduced access of the antibody to the unsubstituted O-glycans. These findings support the idea that C2GnT was able to utilize C6 on GalNAc in the core 1 acceptor to a greater extent than ST6GalNAc I and II, shifting the O-glycosylation pathway away from the structure recognized by CC49 (Fig. 6).
C2GnT Expression Masks the Tumor-associated Epitope Recognized by SM-3-SM-3 and HMFG-2 react with the tandem peptide repeat domain of MUC1, and bind more strongly to MUC1 from cancer tissues than to the more heavily glycosylated MUC1 found in normal tissues (25)(26)(27). SM-3 was raised against a chemically deglycosylated form of the mucin recognized by HMFG-2 (25) and possesses a high degree of tumor specificity. SM-3 reacts with 92% of breast carcinoma samples tested, as well as several other carcinomas, and shows virtually no reactivity with normal resting breast, lactating breast, or other normal tissues (25,26).
Because the CSLEX-1 and CC49 immunoblots indicated that C2GnT expression resulted in the synthesis of more complex O-glycans, we examined whether these O-glycans could mask the SM-3 tumor-associated peptide epitope. SM-3 immunoblotting showed strong recognition of MUC1 (ϳ250 kDa) in the lysate of both the Panc1-MUC1 cells and the low C2GnTexpressing clone (C2#5) but virtually no recognition in the lysates of the clones expressing higher levels of C2GnT (Fig. 7). Similar results were seen in immunoblots probed with HMFG-2 (data not shown). This apparent masking effect further supports the hypothesis that introduction of C2GnT caused more complex MUC1 O-glycosylation.
SM-3 also detected a series of lower molecular mass bands between ϳ100 -140 kDa in the cell lysates of both the Panc1-MUC1 and the C2GnT clones (Fig. 7A). These fragments are likely derived from the tandem repeat region of MUC1, because they could be immunoprecipitated by HMFG-2 and were not detected in lysate from the negative control Panc1 cells (Fig.  7A), which do not express the recombinant MUC1. Furthermore, unlike full-length MUC1, the low molecular mass peptides were not masked in the C2GnT transfectants (Fig. 7A) and were not detected by the CC49 or CSLEX-1 antibodies (data not shown), which recognize specific carbohydrate structures. These results suggest that these peptides are poorly glycosylated, lacking even the core 1 acceptor structure on which C2GnT acts. SM-3 also detected full-length MUC1 in both the cell lysate and conditioned medium, whereas the low molecular mass MUC1 fragments were seen only in the cell lysate, suggesting that these peptides remain within the cell, whereas a fraction of full-length MUC1 is released into the medium.
Neuraminidase Treatment Does Not Reverse Masking of the MUC1 Tandem Peptide Repeat Epitope-Removal of sialic acid has been shown to greatly increase access of anti-MUC1 tandem repeat antibodies such as SM-3 and HMFG-2 to the MUC1 peptide epitope on underglycosylated MUC1 in several cancer cell lines (41). Therefore, we treated cell lysates with neuraminidase to determine whether removal of sialic acid could overcome the effects of introducing C2GnT and expose the MUC1 peptide epitope. We performed immunoblots using the HMFG-2 antibody, because heavier glycosylation is required to mask the HMFG-2 epitope than the SM-3 epitope (25,42). As shown in Fig. 8, removal of sialic acid dramatically decreased the SDS-polyacrylamide gel electrophoresis mobility of fulllength MUC1 in the Panc1-MUC1 cells and the clone expressing low C2GnT (C2#5), consistent with heavy sialylation of MUC1. However, desialylation did not enhance MUC1 peptide epitope recognition by HMFG-2 in the clones expressing higher levels of C2GnT. This suggests that the nonsialic acid portion of the MUC1 O-glycans in the high C2GnT-expressing clones is sufficient to mask the peptide epitope, further indicating greater complexity of the MUC1 O-glycans in the C2GnT transfectants. Fig. 8 also shows that the low molecular mass MUC1 fragments were unaffected by neuraminidase treatment, both in their electrophoretic mobility and antibody recognition. This further supports the idea that these fragments are poorly glycosylated, if at all, because mucin-type glycans are often terminated with sialic acid.
Transfection of C2GnT Did Not Significantly Alter the Activity of Other Glycosyltransferases Which Could Affect MUC1 Epitopes-It is conceivable that the changes seen in expression of the tumor-associated epitopes could have been caused by unanticipated changes in the activities of glycosyltransferases other than C2GnT in the C2GnT transfectants. Therefore, we compared the activities of several glycosyltransferases, which could lead to altered MUC1 epitope expression in the original Panc1 cells, with those in the highest C2GnT-expressing transfectant, Panc1-MUC1 C2#7. As shown in Table I, transfection of C2GnT resulted in no apparent changes in activity of any of the enzymes tested. These results suggest that C2GnT activity was solely responsible for the changes in MUC1 epitope detection shown in Figs. 3, 5, 7, and 8.

DISCUSSION
In this report we describe the effects of stable expression of C2GnT on tumor-associated epitopes in a human pancreatic adenocarcinoma cell line, Panc1-MUC1, which does not express endogenous C2GnT. Three C2GnT-transfected clones expressing varying levels of C2GnT, along with the original Panc1 cell line and parental Panc1-MUC1 cells, were compared in their expression of tumor-associated epitopes recognized by antibodies CSLEX-1, CC49, SM-3, and HMFG-2.
The appearance of the sLe x epitope in the high C2GnTexpressing transfectants is consistent with previous studies, which have linked the sLe x structure, NeuAc␣2-3 Gal ␤1-4 (Fuc ␣1-3) GlcNAc ␤1-6, with the presence of the core 2 ␤ 1-6 branch (13)(14)(15)(16). Interestingly, it was recently found that transfection of a cDNA for a specific ␣1-3 fucosyltransferase into COS-1 cells, which express C2GnT activity, 2 also resulted in the creation of sLe x (43). This study, combined with our results, highlights the importance of both C2GnT and the ␣1-3 fucosyltransferase in generating the sLe x structure, as shown in Fig. 6. Our results imply that Panc1 cells express this fucosyltransferase, the ␣2-3 sialyl transferase, and the ␤1-4 Gal 2 P.-W. Cheng, unpublished observations. FIG. 7. C2GnT transfection produces masking of the SM-3 tumor-associated peptide epitope. A, anti-MUC1 tandem repeat immunoprecipitate (from C2GnT transfectant C2#5, using antibody HMFG-2) or total cell lysates (100 g of protein/well) and (B) conditioned medium (100 l/well) from Panc1, Panc1-MUC1, and C2GnT-transfected Panc1-MUC1 clones were separated by 6% SDS-polyacrylamide gel electrophoresis followed by immunoblotting against the MUC1 peptide tandem repeat domain. C2GnT transfectants are displayed in order of increasing C2GnT expression. C, MUC1 expression levels are shown for normalization of blots in A and B. MUC1 was detected by immunoblotting using the M2 anti-FLAG antibody. transferase necessary for generating sLe x and that C2GnT is the missing component in the sLe x synthetic pathway in Panc1 cells.
The inability of our sLe x -bearing C2GnT transfectants to bind to selectins may be explained by the known antiadhesive properties of MUC1, which can hinder cell-cell and cell-matrix interactions (44,45). Such an antiadhesive effect may be sufficient to block the sLe x -selectin interaction. Our results are also consistent with those described in a recent report by McDermott et al. (46), in which introduction of a MUC1 cDNA into the pancreatic cancer cell line S2-013 abolished the selectin binding ability possessed by the parental S2-013 cells, despite the presence of sLe x on the MUC1 expressed by the cells. It is possible that the sLe x structures on MUC1 are not presented in a favorable configuration for high affinity selectin binding. For example, it has recently been suggested that clustering of the epitope may play a key role in selectin binding (21).
The apparent shift in mucin-type glycosylation pathways caused by introducing C2GnT (Fig. 6) may be partially explained by differences in Golgi localization for C2GnT and the sialyl transferases ST6GalNAc I and II. C2GnT has been detected within the cis-to medial-Golgi compartments (47). The specific Golgi localization of ST6GalNAc I and II has not yet been reported. However, a localization beyond the cis-Golgi is suggested by the inability of ST6GalNAc I present in the cells (Table I) to generate NeuAc␣2,6GalNAc, as suggested by our DBA lectin blotting results (Fig. 4). Furthermore, the CC49 epitope was significantly reduced even in the lowest C2GnTexpressing transfectant, C2#5 (Fig. 5), in which the measured C2GnT activity was approximately equal to that of ST6GalNAc I and II (Table I). This result could be explained by earlier Golgi localization for C2GnT than for ST6GalNAc I and II, because this would enable C2GnT to utilize the core 1 acceptor before ST6GalNAc I and II.
C2GnT also competes for the core 1 acceptor structure with ST3Gal I, which attaches neuraminic acid in ␣-2,3 linkage to galactose of the core 1 structure (36). Evidence for such competition includes shortened MUC1 O-glycans seen in certain breast cancer cell lines, which were attributed to increased ST3Gal I activity and reduced C2GnT activity (48). Furthermore, Whitehouse et al. (49) found that transfection of a C2GnT-expressing normal breast cell line with a ST3Gal I cDNA resulted in decreased core 2 branching of MUC1 Oglycans. These investigators localized ST3Gal I to the medialto trans-Golgi, and thus its localization overlaps with that reported for C2GnT (47). This suggests that partial overlap in Golgi localization is sufficient for efficient glycosyltransferase competition for a common acceptor substrate.
Immunoblots using the anti-MUC1 tandem peptide repeat antibodies SM-3 and HMFG-2 showed that increased C2GnT expression significantly obscured these tumor-associated peptide epitopes (Figs. 7 and 8). Interestingly, we found that SM-3 strongly recognized MUC1 in the Panc1-MUC1 parental cell line, which we showed to possess the sialylated core 1 carbohydrate structure (Fig. 4). This suggests that SM-3 is capable of recognizing certain glycosylated forms of MUC1, despite the fact that SM-3 was raised against a chemically deglycosylated form of MUC1 (25). Moreover, this finding is consistent with a recent study by Karsten et al. (50), which showed that glycosylation of MUC1 by GalNAc or Gal␤1,3GalNAc can actually enhance recognition by SM-3. Together with the masking of the SM-3 epitope we observed upon introduction of C2GnT, these results suggest that SM-3 is capable of reacting with MUC1 possessing O-glycan structures below a certain threshold level of complexity.
Masking of the HMFG-2 epitope persisted following removal of sialic acid (Fig. 8). Because HMFG-2 recognition of MUC1 is known to be enhanced by desialylation of underglycosylated MUC1 present in certain cancer cell lines (41), this result strongly suggests significantly increased MUC1 glycosylation in the C2GnT transfectants, beyond a level at which desialylation can expose the epitope. These results are consistent with the more complex glycosylation of MUC1 in the C2GnT transfectants implied by the immunoblotting results using CC49 and CSLEX-1. We conclude that the core 2-associated O-glycans free of sialic acid are sufficient to block antibody recognition, underscoring the importance of the core 2 structure in regulating tumor epitope expression.
In conclusion, we have found that the stable expression of C2GnT in a human pancreatic cancer cell line was able to greatly reduce the expression of tumor-associated epitopes recognized by monoclonal antibodies SM-3 and CC49, although inducing expression of sialyl Lewis x. These findings imply that alterations in C2GnT activity could play a key role in regulating the expression of these epitopes on cancer cells. In particular, our results suggest that C2GnT activity may be suppressed in certain cancer cells, which deserves further exploration.