Human Colon Adenocarcinomas Express a MUC1-associated Novel Carbohydrate Epitope on Core Mucin Glycans Defined by a Monoclonal Antibody (A10) Raised against Murine Ehrlich Tumor Cells

INTRODUCTION

Most mAbs 5 with selectivity for adenocarcinomas have been shown to be directed against epitopes carried out on mucus glycoproteins (mucins), which are present in the glycocalix of epithelial cells in many secretory tissues (1 , 2) . These highly glycosylated molecules often show tumor-associated alterations that have been used as tumor markers. mAbs reacting with cancer mucins have been shown to recognize epitopes on the apomucin backbone and/or their oligosaccharide side chains (1 , 2) . A number of these tumor-associated epitopes are normally cryptic antigens and become exposed after incomplete glycosylation occurring in malignant cells. Others appear as a consequence of abnormal glycosylation events that may lead, for example, to the synthesis of aberrant or incompatible blood group antigens (3) . Interestingly, these epitopes are often associated with a wide variety of adenocarcinomas originating from different tissues (1) . In addition, some of them are considered oncodevelopmental antigens because they are expressed in normal fetal tissues (4) .

A10 is an IgM mAb raised against murine ET cells that inhibits their growth both in vivo (5) and in vitro (6) . It recognizes a yet undefined carbohydrate epitope that is carried by a high molecular weight cell surface glycoprotein (7) . Similar molecules have been shown to bind IgM antibodies produced by ET-immunized mice (8) as well as natural IgM antibodies (5) , both playing a critical role in the host resistance against this tumor. The potent biological activity of A10 mAb has led us to study in greater detail the distribution of the epitope it detects. In the course of this screening, both murine and human tumors cells were included, and all but HT-29 and Caco 2 human colon cancer cell lines were A10 unreactive. 6 These results prompted us to examine the A10 reactivity with frozen sections of normal and malignant human colonic tissues as well as other normal and neoplastic human tissues.

The results presented herein indicate that mAb A10 reacts strongly with most human colon adenocarcinomas but not with normal colon. A10 is reactive with a selected variety of adenocarcinomas as well and shows more restricted reactivity with normal tissues. Moreover, mAb A10 defines a novel tumor-associated carbohydrate epitope carried on MUC1 from colon cancer tissues. This epitope is associated to core mucin glycans and likely contains GlcNAcβ(1–6)GalNAc.

MATERIALS AND METHODS

Reagents.

Trypsin, proteinase K, phenylmethylsulfonyl fluoride, trypsin inhibitor, Tween 20, sodium m-periodate, sodium borohydride, CsCl, TFMSA, phenyl-N-acetyl-α-d-galactosaminide, α-galactosidase (coffee beans), β-galactosidase (bovine testes), DNase II, tunicamycin, and BSA were from Sigma Chemical Co. (St. Louis, MO). Neuraminidases (Vibrio cholerae and Arthrobacter ureafaciens), N-acetyl-β-glucosaminidase (Diplococcus pneumoniae), O-glycosidase (D. pneumoniae), and N-glycosidase (glycopeptidase-F) were from Boehringer-Mannheim (Mannheim, Germany).

Lectins.

Digoxigenin-labeled MAA, SNA, and DSA were purchased from Boehringer-Mannheim. Biotin-labeled PNA, tomato, SBA, BS1, and WGA were from Sigma.

Carbohydrates.

GalNAc1α-O-p-nitrophenyl, GlcNAc1α-O-p-nitrophenyl, Galβ(1–3)GalNAc1α-O- p-nitrophenyl PNP, Galβ(1–3)[GlcNAcβ(1–6)]GalNAc1α-O-p-nitrophenyl, GlcNAcβ(1–3)GalNAc1α-O-p-nitrophenyl, GlcNAcβ(1–6)GalNAc1α-O-p-nitrophenyl from Toronto Research Chemicals (North York, Ontario, Canada).

Tumor Cell Lines.

HT-29 colon cancer cells and an HT-29-derived subpopulation of mucus-secreting cells selected by culture in 10⁻⁶m methotrexate (herein designated HT-29 M6 cells; (9) ) were obtained from Drs. Alain Zweibaum and Thécla Lesuffleur (INSERM U178, Villejuif, France). Caco 2 human colon cancer cell line from the American Type Culture Collection (Rockville, MD).

Monoclonal Antibodies.

A10 and 2A1 are IgMκ derived from fusions from ET-immunized mice. A10 recognizes a cell surface carbohydrate epitope on mucin-like molecules of ET cells (5) , whereas 2A1 is an unreactive control antibody. Both mAbs were purified as described (5) . BC1 (IgG), BC2 (IgG), and CCP58 (IgG) react with peptides encompassing the tandem repeat sequence of MUC1 (BC1 and BC2) and MUC2 (CCP58; (10 , 11) ) and were a generous gift from Drs. P. X. Xing and I. F. C. McKenzie (Austin Research Institute, Victoria, Australia). LDQ10 (IgM) reacts with the MUC2 apomucin tandem repeat (12) .

Tissues.

Normal and malignant frozen tissues were obtained from the Department of Pathology, Hospital Clínico San Carlos (Madrid, Spain). In general, normal tissues were from uninvolved areas of specimens obtained from the very same cancer patients. HT-29 xenografts were obtained by inoculating 5 × 10⁶ HT-29 cells i.m. in the left hind of BALB/c-nu/nu (Charles River España, , Barcelona, Spain) or severe combined immunodeficient (Centro de Biología Molecular, Madrid, Spain) mice. Tumors were excised when they reached ∼1 cm in diameter and frozen until used.

IIF.

Frozen sections (3–5 μm) were incubated (30 min at room temperature) with 3 μg/ml of A10 or 2A1 in PBS-0.3% BSA. After washing with PBS, a 1:30 dilution of fluorescein-conjugated F(ab′)₂ goat anti-mouse IgM (Pel-Freez, Rogers, AR) plus 1% Evans blue was added. The reactions were scored semiquantitatively as 3+, 2+, or 1+ when reactivity was clearly observed at ×100, ×200, and ×400, respectively. IIF was also performed on cytocentrifuge preparations of HT-29 and Caco 2 cells cultured in vitro in the absence or presence of phenyl-N-acetyl-α-d-galactosaminide (2 mm) or tunicamycin (5 μg/ml) to inhibit O- or N-glycosylation, respectively.

Immunoelectron Microscopy.

HT-29, HT-29 M6 cells, normal human colon, and colon cancer tissues were fixed and embedded in Lowicryl K4M as described previously (13) . Gold immunolabeling was performed according to the postembedding methodology. Briefly, thin sections placed in parlodion/carbon-coated nickel grids were floated on a droplet of PBS for 5 min and then incubated with mAb A10 for 60 min at room temperature. Subsequently, sections were rinsed in PBS and further incubated with affinity-purified rabbit anti-mouse IgG+IgM (50 μg/ml in PBS containing 0.1% Tween 20) for 45 min. After two washes in PBS, grids were floated onto a droplet of protein A-gold of 10 or 15 nm (BioCell, Cardiff, United Kingdom) for 45 min. Grids were rinsed in PBS, double distilled water, and allowed to dry. Thin sections were counterstained with 3% uranyl acetate for 6 min and lead citrate for 45 s.

Preparation of Tissue Extracts.

Tissues from HT-29 xenografts and normal and malignant colonic tissues from the same patient were minced in PBS and treated with 0.25 mg/ml of trypsin for 1 h at 37°C. After adding trypsin inhibitor (1 mg/ml) and phenylmethylsulfonyl fluoride (1 mm), tissues were homogenized using a Potter at 1500 rpm. Tissue homogenates were cleared by centrifugation at 20,000 × g for 1 h at 4°C, dialyzed against PBS, and lyophilized.

Gel Filtration.

Tissue extracts at 10 mg/ml (dry weight) were loaded onto Sephacryl S-300 (Pharmacia, Uppsala, Sweden). The column (200 × 2.5 cm) was equilibrated in PBS, and fractions (1 ml) were collected using PBS as eluent at a linear flow rate of 5 cm/h. Protein was monitored at 280 nm, and an aliquot from each fraction at 1:100 dilution was used to assess mAb reactivity by ELISA as described below. The void (Vo) volume of the column was determined using Blue Dextran 2000.

Density Gradient Ultracentrifugation.

The Vo volume fractions from HT-29 xenograft extracts were lyophilized, and 3 mg (dry weight) were dissolved in 7 ml of a CsCl solution containing 4 m guanidine hydrochloride with a density of 1.40 g/ml. The gradient was formed by centrifugation in a Beckmann L8-70 at 100,000 × g for 72 h at 4°C. Fractions (0.5 ml) were weighed to determine their density.

ELISA.

mAb reactivity with tissue extracts was tested by indirect-ELISA exactly as described (5) . In some experiments, tissue extracts were submitted to several treatments (periodate oxidation and enzymatic treatments) once bound to the ELISA plate (see below) before being assayed for A10 reactivity. A similar ELISA was carried out to assess binding of labeled lectins with these tissue extracts after deglycosylation. A sandwich-ELISA was performed in particular experiments to test the relationship between A10 reactivity and MUC1 and/or MUC2, using A10 or 2A1 (control) as catcher mAbs and BC1 or CCP58 as tracer antibodies. Briefly, A10 or 2A1 (5 μg/ml) were allowed to bind to the wells of ELISA plates for 18 h. Tissue extract fractions were then added for 2 h. After washing, a dilution of BC1 or CCP58 (1:1000) was added. Reactions were detected by using Fc-specific peroxidase-coupled goat anti-mouse IgG (Nordick, Tilburg, the Netherlands). ELISA was also used for inhibition studies of A10 reactivity with different carbohydrates. Thus, A10 (1:1000 final dilution in PBS-1% BSA) was incubated with every inhibitor (1–7 mm) overnight at 4°C. Samples were then centrifuged (14,000 × g; 10 min) and used for ELISA without a further dilution. As a positive control, HT-29 xenograft-derived mucins (Vo volume of Sephacryl S-300) were used. Nonspecific inhibition was assessed using an isotypic matched mAb (LDQ 10) recognizing the MUC2 apomucin chain (12) . Results were expressed as the percentage of the initial reactivity (in the absence of inhibitor).

Immunoprecipitation and Western Blotting.

A tissue extract (Vo volume, Sephacryl S-300 column) from HT-29 xenograft was immunoprecipitated with BC-2 (anti-MUC1) by means of protein A-agarose beads (Affi-Gel MAPS II; Bio-Rad, Hercules, CA). Thus, beads were incubated (3 h; 4°C) with BC-2 or none (control beads), using the binding buffer given by the supplier (pH 9.0). After a washing step with PBS, both BC2-coated and control beads were incubated (12 h; 4°C) with HT-29 tissue extract in PBS. After washing, beads were treated with running sample buffer (0.1 m Tris-HCl, 0.1% SDS, 25% glycerol, and 0.01% bromphenol blue, pH 6.8) and boiled. Both BC2 and nonspecific immunoprecipitates were run in 0.1% SDS-5% PAGE and transferred to nitrocellulose sheets in standard conditions (Mini-Protean Cell Electrophoresis; Bio-Rad). Nitrocellulose sheets were neutralized with 1% BSA and incubated (overnight at 4°C) with either PBS (control), BC-2, or A10 (0.1 μg/ml in PBS 1% BSA, 0.05% Tween 20). After a washing step, all runs were incubated (3 h, at room temperature) with a dilution of a goat serum anti-murine immunoglobulins conjugated with alkaline phosphatase (Bio-Rad). After a final washing, the sheets were soaked in a substrate solution (4-nitroblue tetrazolium chloride/5-bromo-4-chloro-3-indolyl-phosphate), and the enzymatic reaction was stopped by rinsing with deionized water.

Enzymatic Treatments.

These treatments were carried out on mucins derived from tissue extracts (Vo volume, Sephacryl S-300 column) bound to plastic after neutralization with PBS-1% BSA. The following treatments were performed. DNase II (5000 units/ml) in 0.02 m phosphate buffer (pH 7.0), 2 mm MgCl₂, and 1 mm CaCl₂ for 1 h at 37°C; neuraminidases from V. cholerae (0.1 unit/ml) and from A. ureafaciens (5 units/ml) in 0.05 m acetate buffer (pH 5.5) and 2 mm CaCl₂, for 18 h at 37°C; α-galactosidase (0.1 unit/well) in PBS (pH 7.0) for 24 h at 37°C; β-galactosidase (0.05 unit/well) in 0.05 m citrate buffer (pH 4.2) for 24 h at 37°C; O-glycanase (5 milliunit/ml) in 0.1 m sodium phosphate (pH 6.0) for 2 h at 37°C; N-glycanase (1 unit/ml) and N-acetyl-β-glucosaminidase (5 milliunits/well) in 20 mm acetate buffer (pH 6.5) for 18 h at 37°C; and trypsin and proteinase K (5 mg/ml) in PBS for 4 h at 37°C. In all cases, plates were washed after treatment and used for ELISA as above.

TFMSA Treatment.

Tissue extracts derived from HT-29 xenografts or human colon were deglycosylated with TFMSA (14) . Briefly, the Vo volume obtained after Sephacryl S-300 fractionation was lyophilized and treated with 0.8 ml of TFMSA for 4 h at room temperature under nitrogen. This was followed by a neutralization step by adding pyridin:water (3:2) at −20°C and dialyzed against 0.1 m carbonate buffer at pH 9.6. The extent of deglycosylation was monitored by ELISA using lectins recognizing peripheral, backbone, and/or core carbohydrates (15) .

Other Treatments.

Tissue extracts (Vo volume, Sephacryl S-300 column) were treated chemically to examine the properties of the A10 epitope: (a) solubility in 0.6 n HClO₄ for 1 h at 4°C followed by centrifugation (14,000 × g for 10 min) to remove insoluble precipitate, neutralization to pH 7.4, and dialysis; (b) phenol extraction with water-saturated phenol as described (8) , followed by dialysis of the aqueous phase; (c) lipid extraction with chloroform:methanol (1:1) as described (8) , followed by extraction of the aqueous phase with water-saturated ether and dialysis against distilled water; (d) alkali treatment (β-elimination), carried out with NaOH (0.05 and 0.5 n) for 48 h at 50°C as described (16) , followed by dialysis against PBS; (e) Smith degradation, carried out on extracts bound to ELISA plates, using a serial periodate oxidation (17) : oxidation with 20 mm sodium m-periodate in 0.05 m sodium acetate buffer at pH 4.4, followed by a reduction step with 0.1 m sodium borohydride, and mild acid hydrolysis with 0.05 n H₂SO₄.

RESULTS

Reactivity of mAb A10 with Human Colonic Tissues.

The reactivity of A10 with normal and pathological colonic tissues was determined using IIF on frozen tissue sections. As shown in Table 1 ⇓ , most samples of normal colonic tissues were unreactive, except for four cases that showed weak punctate cytoplasmic reactivity. In contrast, most carcinomas (36 of 42) were strongly reactive. In cancers, diffuse cytoplasmic and membrane-like staining patterns were observed in virtually all tumor cells (Fig. 1) ⇓ . Remarkably, all three mucinous colon carcinomas tested thus far were scored as negatives. A10 reactivity with other colonic tissues is shown in Table 1 ⇓ . A10 reacted with a highly dysplastic adenomatous villous polyps but not with benign polyps or with the so-called transitional mucosa, i.e., that adjacent to the tumor with well-oriented glands. Interestingly, fetal mucosae was also negative, as well as other nonmalignant colon tissues (Table 1) ⇓ . All samples were unreactive with a control, isotype-matched mAb (2A1; not shown).

Reactivity of mAb A10 with Other Tissues.

As shown in Table 2 ⇓ , most adenocarcinomas tested (lung, pancreas, stomach, prostate, ovary, and endometrium) were reactive with A10. The staining pattern was similar to that observed in colon cancers. Nevertheless, some carcinomas, such as those from the breast, were unreactive with A10; yet benign proliferative lesions from breast (fibroadenoma) were strongly reactive (Table 2) ⇓ . Most normal tissues were unreactive as well (Table 3) ⇓ , although some pancreatic and breast ducts and, to a lesser extent, bronchial mucosae, were A10 reactive. The pattern of A10 reactivity with these tissues was similar to that found in adenocarcinomas, but with staining mainly the the luminal area of glands (not shown). In addition, A10 also stained some basal membranes from skin, alveoli, and blood vessels.

Reactivity of mAb A10 with Human Colon Cancer Cells by Immunoelectron Microscopy.

Fig. 2 ⇓ shows the subcellular distribution of A10 reactivity by immunoelectron microscopy in ultrathin sections of normal colon and colon cancers. In normal colonic goblet cells, mucin droplets were generally unreactive (Fig. 2A) ⇓ . Occasionally, some gold labeling of mucus droplets and Golgi complex was observed (Fig. 2A) ⇓ . Absorptive cells did not show any specific labeling (not shown). In contrast, the content of mucin droplets was labeled in colon adenocarcinomas (Fig. 2B) ⇓ . This was also the case when testing polarized mucin-secreting HT-29 M6 human colon cancer cells, which showed strong labeling of mucin droplets and microvilli while lacking reactivity with the basolateral membrane and the rest of the apical membrane (Fig. 2C) ⇓ . In undifferentiated HT-29 cells, mucins droplets were labeled, but plasma membrane was not (not shown).

Reactivity of mAb A10 with Mucin-enriched Fractions from Colon Adenocarcinomas.

The epithelial distribution and the results of electron microscopy suggested that mAb A10 was recognizing mucin molecules. To examine such hypothesis, extracts from HT-29 xenografts, colon adenocarcinomas, and normal colon were tested. Fig. 3 ⇓ shows the elution profiles of these extracts through a S-300 Sephacryl column and their reactivities with mAbs A10, BC1 (anti-MUC1), and 2A1 (control IgM mAb). As shown, A10 reacted with fractions present in the Vo volume (over 1.5 × 10⁶ Da) of extracts from colon adenocarcinomas but not from normal colon (Fig. 3) ⇓ . Physicochemical treatment of the Vo fractions from HT-29 extracts indicated that heat denaturation (1 h at 100°C), acid treatment (HClO₄, 0.6 n, 1 h), lipid extraction (chloroform:methanol), and phenol precipitation did not reduce A10 reactivity (data not shown). Fig. 4 ⇓ shows the distribution of A10-reactive material in the Vo volume from HT-29 extracts submitted to CsCl isopyknic density gradient. As shown, fractions reactive with A10 had a typical mucin bouyant density (1.42–1.45 g/ml) and were enriched in sialic acid and MUC1 as detected by MAA and BC1, respectively.

Reactivity of mAb A10 with MUC1 Associated with Colon Cancer Mucins.

HT-29 xenografts contain MUC1 and MUC2 as determined by ELISA using antibodies BC1 and CCP58, respectively. To examine the relationship between mucin molecules and A10 reactivity in greater detail, a sandwich-ELISA was performed using A10 bound to ELISA plates as catcher antibody and BC1 (anti-MUC1) or CCP58 (anti-MUC2) as tracer antibodies. In a representative experiment (Table 4) ⇓ , A10 reactive mucins from colon adenocarcinoma or HT-29 xenograft were recognized by BC1 (anti-MUC1) but not by CCP58 (anti-MUC2). When using normal colonic mucins, no reactivity with BC1 or CCP58 was observed using the same assay (Table 4) ⇓ . To confirm that A10-reactive epitopes were borne by MUC1 molecules, A10 reactivity was assessed by immunoblotting on anti-MUC1 immunoprecipitates as described in “Materials and Methods.” As shown in Fig. 5A ⇓ , A10 recognized components of the same mobility as those detected using BC2 (anti-MUC1) antibody (M_r 300,000–400,000), whereas these bands were not identified in control immunoprecipitates. The converse experiment, i.e., immunoblotting on A10 immunoprecipitates, was carried out and yielded similar results, demonstrating that A10 epitopes are carried on MUC1 in these tumors.

Reactivity of A10 with Colon Cancer Mucins after Exoglycosidase, Alkali, and Periodate Treatments.

The fact that A10 recognized a carbohydrate epitope on murine ET cells (5) suggested the recognition of a related epitope on human mucins. To examine this hypothesis, different deglycosylation treatments were carried out (Table 5) ⇓ ; their effectiveness was controlled by means of labeled lectins. As shown, the enzymatic release of sialic acid, whether linked to α(2–3)Gal (V. cholerae neuraminidase) or to α(2–6)Gal (A. ureafaciens neuraminidase), did not modify A10 reactivity significantly. Similarly, no effect was observed by using other exoglycosidases, except for a consistent increase in A10 reactivity after β-galactosidase treatment. Other enzymes such as DNase, trypsin, and proteinase K were also unsuccessful to impair A10 reactivity (data not shown). As shown, no effect was noted after periodate treatment, in striking contrast to the murine epitope (5) , pointing against the involvement of sugar residues in terminal nonreducing positions in the A10 human epitope. Nevertheless, A10 reactivity decreased after several rounds of periodate oxidation in the form of Smith degradation. A similar decrease was also observed after alkali treatment (β-elimination), suggesting the involvement of O-linked carbohydrates. As shown in the same table, none of these treatments decreased the reactivity of mAb BC1 with MUC1 used as control, indicating their carbohydrate specificity.

In addition, HT-29 cells were grown in vitro in the presence of phenyl-N-acetyl-α-d-galactosaminide and tunicamycin, inhibitors of O- or N-glycosylation, respectively. The results (not shown) indicated that A10 reactivity with HT-29 cells was prevented by both phenyl-N-acetyl-α-d-galactosaminide and tunicamycin.

Reactivity of A10 with Colon Cancer Mucins Deglycosylated with TFMSA.

The involvement of peripheral, backbone, or core carbohydrates associated with O-linked mucin glycans in A10 human epitope was further studied. Thus, A10 reactivity was assessed on mucins partially deglycosylated with TFMSA; the extent of chemical deglycosylation was controlled by various lectins. As shown (Table 6) ⇓ , peripheral sugars, detectable with sialic acid-binding lectins (SNA and MAA), were successfully removed by TFMSA treatment from HT-29 xenograft, colon adenocarcinoma, and normal colon-derived mucins. In addition, a decrease in binding of tomato lectin suggested that trimers and tetramers of GlcNAc had been removed. Similarly, a lower binding of DSA suggested this was also the case for type 2 backbone chains. By contrast, binding of lectins recognizing mainly core sugars such as BS1, SBA, and PNA remained unaffected. Under these conditions, A10 reactivity with TFMSA-treated human mucins was fully preserved, even enhanced, in every case (Table 6) ⇓ . These results suggested that A10 human epitope was not dependent on peripheral or backbone carbohydrates but located in or close to core mucin glycans.

TFMSA-treated mucins were further submitted to endoglycopeptidase and protease treatments, taking advantage of a putative higher sensitivity to such enzymatic digestions (18) . As shown in Table 7 ⇓ , O-glycanase released successfully core 1, Galβ(1–3)-GalNAc, as judged by a decrease in binding to PNA. Whereas A10 reactivity still remained, as occurred after N-glycanase treatment. Moreover, reactivity with A10, but not with BC1 used as control, was resistant to protease digestion. This pointed against the possibility of A10 defining a complex carbohydrate-peptide epitope in human colon cancer mucins.

Inhibition Studies of A10 Reactivity with Core Mucin Glycans.

The foregoing results suggested that A10 was reacting with core mucin glycans other than core 1. To assess the involvement of core carbohydrates in A10 human epitope, inhibition studies with different core mucin glycans were carried out. Table 8 ⇓ summarizes the results of these assays in which an isotype-matched mAb (LDQ10) reacting with the MUC2 apomucin chain (12) was always included as control of specificity, as well as a positive control carried out with HT-29 xenograft extracts. As shown, no significant inhibition was achieved with GalNAc or core 2. Core 1 and GlcNAc produced a minor inhibition at the highest concentration tested, likely nonspecific, as was also seen on LDQ10 (not shown). By contrast, core 3 [GlcNAcβ(1–3)GalNAc] and core 6 [GlcNAcβ(1–6)GalNAc] inhibited A10 reactivity specifically. The inhibition achieved by core 6 was greater than by core 3 in all three concentrations tested, reaching to ∼50% with 7 mm but being still remarkable with 1 mm(Table 8) ⇓ . Thus, the results pointed to carbohydrate structures related to core 6 glycans as being involved in A10 reactivity with human colon cancer mucins.

DISCUSSION

mAb A10 was originally raised in a BALB/c mouse immunized with devitalized ET cells and selected because of its strong reactivity with ET cell surface carbohydrates (5) . Here we show that, in addition, A10 reacts strongly with most human colon adenocarcinomas, with the striking exception of mucinous carcinomas. Although this was also the case with a wide variety of other epithelial cancers, few normal tissues were A10 reactive. Of these, the reactivity with breast ducts was remarkable because ductal breast carcinomas, which are thought to be derived from duct cells, were A10 unreactive. In the colon, A10 reactivity was virtually absent in normal epithelium and in the so-called transitional mucosa. Polyps with low malignant potential were also A10 unreactive, suggesting that A10 epitope may be a marker of the malignant transition in colonic epithelium, as described for T (19) and other blood group-related antigens (20) . In this sense, A10 lacked reactivity with A1, A2, B, H, Le^a, Le^b, and i or A-like (Forssman) typed erythrocytes (not shown). Moreover, the lack of A10 reactivity with granulocytes, kidney proximal tubules, normal gastrointestinal epithelia, acinar pancreatic cells, breast carcinomas, neuraminidase treated-lymphocytes/erythrocytes, and acid-treated bovine submaxillary mucins (not shown) made it highly unlikely that A10 was reacting with T, Tn antigens (1 , 19 , 21) , Le^x, Le^y, and/or their extended forms (22, 23, 24) .

The tissue distribution of A10 human epitope prompted us to consider mucins as the target molecules for its reactivity. Several lines of evidence indicated that this was the case, at least regarding colon adenocarcinomas: (a) the subcellular distribution of A10 epitope was restricted to mucin droplets and microvilli in colon adenocarcinoma cells; (b) physico-chemical properties of A10-reactive molecules derived from colon adenocarcinomas, such as their high molecular weight by gel filtration, buoyant density in CsCl gradients, or solubility in perchloric acid, were all them typical of mucins (18 , 25) ; (c) A10 epitope was present on MUC1 molecules from either colon adenocarcinomas or HT-29 xenogeneic grafts, indicating a direct relationship between A10 epitope and this mucin. The fact that A10 epitope could not be demonstrated on MUC2 molecules might suggest that, in colon cancer, expression of A10 epitope is restricted to particular apomucins, e.g., MUC1. Such a restriction could be considered to explain the lack of reactivity with certain colon adenocarcinomas, e.g., mucinous carcinomas, in which the production of A10 unreactive mucins might predominate. It is unlikely, however, that A10 epitope is carried exclusively by MUC1, because A10 and BC1 showed distinct patterns of tissue reactivity in colon adenocarcinomas (data not shown). Indeed, A10 reactivity was detected in both microvilli and mucin droplets of HT-29 M6 cells, whereas MUC1 is present mainly in the microvilli of these cells (26) . Conversely, A10 epitope is likely not be exposed in every MUC1 cancer mucin, because of the low A10 reactivity with breast carcinomas, where MUC1 is known to be overexpressed (27) . Additional work is necessary to elucidate these issues and also to determine whether in normal breast and pancreas the A10 epitope is also carried by MUC1, the major mucin molecule in these tissues (27 , 28) .

Previous data indicated that A10 murine epitope was a carbohydrate in nature, because it is rather sensitive to a mild periodate oxidation (5) . A10 epitope in colon cancer mucins was, in contrast, resistant to such a treatment. This pointed against the involvement of terminal nonreducing sugars in A10 human epitope, which fitted with its resistance to neuraminidase. These features distinguished A10 from a number of mAbs recognizing mucin-like glycoproteins associated with human adenocarcinomas (24 , 28, 29, 30, 31, 32, 33) . Yet, A10 human epitope was sensitive to periodate in the form of Smith degradation, indicating that A10 was reacting with a carbohydrate located in an internal portion of the oligosaccharide side chain. Moreover, it was sensitive to β-elimination, supporting the involvement of O-linked carbohydrates consistent with a mucin-type glycosidic linkage to Ser/Thr (18) . This was also supported indirectly by the fact that HT-29 cells cultured with phenyl-N-acetyl-α-d-galactosaminide, a rather specific O-glycosylation inhibitor (34) , lacked A10 reactivity. Although this was also the case if culturing with tunicamycin, a structural analogue of UDP-GlcNAc used to inhibit N-glycosylation (35) , it should be noted that N-glycosylation in mucins appears to be required before O-glycosylation takes place (36) . Moreover, the possibility of GlcNAc being involved in the A10 epitope (see below) might also be considered to explain the tunicamycin effect. As shown in “Results,” A10 reactivity with human colon cancer mucins was fully preserved, even enhanced, after chemical deglycosylation with TFMSA. The pattern of lectin binding to TFMSA-treated mucins indicated a successful removal of peripheral and backbone sugars but not core glycans. These results were consistent with both the resistance of core mucin glycans to TFMSA (15) and the involvement of these carbohydrates in A10 human epitope. As core mucin glycans are in the innermost portion of the oligosaccharide side chain, the possibility of the apomucin chain being involved in the A10 human epitope was also considered. A10 reactivity, however, was fully resistant to proteolysis, in spite of the previous deglycosylation with TFMSA to favor protease digestions (18) . Therefore, the results pointed against A10 reacting with a complex carbohydrate/peptidic epitope (37) , or a mere cross-reactivity with the apomucin chain, as described for some mucin-reactive mAb interacting with both peptidic and carbohydrate moieties (38) .

A10 reactivity with TFMSA-deglycosylated colon cancer mucins was fully preserved after O-glycanase treatment, in which Galβ (1–3)GalNAc (core 1) was released successfully. This endoglycopeptidase releases GalNAc-O-Ser/Thr when Gal, but not GlcNAc, is attached (15 , 39) , raising the possibility of A10 reacting with core mucin glycans other than core 1 (T antigen), i.e., containing GlcNAc linked to the acetylgalactosaminyl residue. T antigen was, in any case, very unlikely to be recognized by A10 because of the tissue distribution of the A10 epitope, as discussed above, with high expression in locations where T antigen is lacking (e.g., normal breast) and low where it is overexpressed (e.g., breast carcinoma). Moreover, inhibition studies with core 1, as well as with its derivative core 2, Galβ (1–3)[GlcNAcβ(1–6)]GalNAc, pointed against the involvement of these carbohydrates in A10 reactivity. By contrast, this reactivity was significantly inhibited by core 3, GlcNAcβ(1–3)GalNAc, and remarkably, by core 6, GlcNAcβ(1–6)GalNAc. It seems unlikely, however, that core 3 was involved in the A10 human epitope, given its wide expression in normal colonic mucin species (40) . Furthermore, an impairment in core 3 synthesis and also in its derivative core 4 might be expected in colon adenocarcinomas, as occurred in Caco 2 cells, which lack the core 3 β3-GlcNAc transferase activity (41) ; yet these cells were strongly A10 reactive. Interestingly, a similar impairment has been shown for the core 2 β6-GlcNAc transferase activity in mammary carcinoma cells (42) , a fact supported by the structural analyses of their O-linked carbohydrate chains derived from these cells (43, 44, 45) . Although this enzyme is involved in the synthesis of core 2, it has been proposed that it may also be involved in the synthesis of core 6, upon releasing the β3-linked galactose from the former (41) . The striking low reactivity of A10 with breast carcinomas, in contrast to normal breast duct cells, together with the rather strong inhibition of A10 achieved with core 6, points to this mucin glycan as the main candidate to be involved in A10 reactivity with human colon cancer mucins. The fact that colon cancer cells show high β-galactosidase activity (41) , together with the slight but significant increase in A10 reactivity after β-galactosidase treatment (Table 5) ⇓ , also points to the same direction. In this sense, core 6 was described in κ-casein from human milk (46) as well as in mild acid-treated meconium glycoproteins (47) .

Thus far, the only core mucin glycans associated with colon adenocarcinomas are T, Tn, and their sialylated forms. T and Tn antigens are thought to be cryptic carbohydrate determinants normally masked by chain elongation, which are exposed in cancer mucins because of truncated glycosylation of the oligosaccharide side chains (1 , 48) . The A10 epitope seems not to be cryptic in the colon, however, because its differential expression remained in TFMSA-deglycosylated normal and cancer mucins. Nevertheless, overexpression of a given core mucin glycan may also be influenced in cancer cells by factors such as the differential activity of glycosyltransferases competing for the same substrate position (21) . As discussed above, the pattern of reactivity of A10 with normal and cancer cells derived from breast and colon tissues might be explained by the differential activities of β6-GlcNAc versus β3-GlcNAc transferase in breast and colon cancer cells.

Finally, the origin of A10 hybridoma deserves consideration because the murine ET cell line was used as immunogen (5) . This should not be surprising, however, because tumor-associated carbohydrate epitopes, including T antigen, can be shared by human and murine cancer cells (4 , 49) . Alternatively, natural antibodies in normal mouse sera have been described reacting with human oncofetal antigens, including those associated with acid-treated meconium containing core 6 (47) . Although we do not know whether A10 is a monoclonal natural antibody or was induced after the mouse immunization actually, it is clear that either possibility might be. The former can be, however, considered seriously given the similarities between A10 and natural antibodies, with regard to their antibody isotype, antitumor properties, and target molecules recognized on ET cells (5 , 6) .