Differential effect of GalNAcα-O-bn on intracellular trafficking in enterocytic HT-29 and Caco-2 cells: correlation with the glycosyltransferase expression pattern

The specific functions of the apical and basolateral sides of polarized cells are correlated with a marked membrane asymmetry at both the protein and lipid levels. Generally, the apical membrane is characterized by a higher sphingolipid content (glycosphingolipids and sphingomyelin) (Simons and Van Meer, 1988; Van Meer, 1989), and the presence of proteins inserted into the membrane through a glycosylphosphatidylinositol (GPI) anchor (Lisanti et al., 1988). In addition, particular biological functions of apical membranes are mediated through the expression of specific transmembrane proteins, such as digestive hydrolases for enterocytes (Hauri, 1988).

A complex system of sorting machinery is involved to ensure the cell polarity is maintained. Basolateral targeting signals have been identified in the cytoplasmic domain of transmembrane proteins; these are the tyrosine, dileucine and Leu-Val motifs (Matter et al., 1992; Matter and Mellman, 1994; Hunziker and Fumey, 1994; Monlauzeur et al., 1995; Sheikh and Isacke, 1996; Keller and Simons, 1997). For the apical delivery, the association of proteins with glycosphingolipids has been proposed (Simons and Van Meer, 1988; Simons and Wandinger-Ness, 1990). GPI-anchored proteins and glycosphingolipids could be isolated from a Triton X-100 insoluble fraction (Brown and Rose, 1992). The dynamic clustering of glycosphingolipid-enriched microdomains would constitute functional rafts for the apical delivery of proteins (Simons and Ikonen, 1997). In this way, the GPI anchor has been assumed to represent an apical targeting signal; however, several studies suggested that apical sorting information is in fact provided by other determinants (Arreaza and Brown, 1995; Benting et al., 1999).

The potential role of glycosylation upon intracellular trafficking in polarized cells has been highlighted by several recent data. N-glycans can act as apical sorting signals for several secretory, transmembrane or GPI-anchored model proteins in Madin-Darby canine kidney (MDCK) cells (Scheiffele et al., 1995; Gut et al., 1998; Benting et al., 1999). The presence of O-glycosylation sites is also important in the apical delivery of the human neurotrophin receptor in polarized MDCK and Caco-2 cells (Yeaman et al., 1997; Monlauzeur et al., 1998). O-linked glycans were recently shown to be involved in the apical sorting of human intestinal sucrase-isomaltase and dipeptidyl peptidase IV (DPP-IV), but not aminopeptidase N (Alfalah et al., 1999; Naim et al., 1999).

In that field, our previous work showed that the secretion of mucins, which are highly O-glycosylated glycoproteins, is markedly inhibited by the permanent exposure of polarized goblet HT-29 cells to a competitive inhibitor of O-glycosylation, 1-benzyl-2-acetamido-2-deoxy-α-D-galactopyranoside (GalNAcα-O-bn) (Hennebicq-Reig et al., 1998). Inside the cells, this N-acetylgalactosaminide is converted into the benzyldisaccharide Galβ1-3GalNAcα-O-bn, which acts as a potent competitive inhibitor of the sialyltransferase ST3Gal I (Delannoy et al, 1996). Indeed, ST3Gal I is highly involved in the synthesis of the O-linked glycans on the HT-29 MTX mucins, and particularly the sialyl-T antigen Neu5Acα2-3Galβ1-3GalNAc-R, which accounts for 41% of the oligosaccharide side chains of the mucins (Hennebicq-Reig et al., 1998). Besides, a blockade in the apical targeting of membrane brush border glycoproteins, such as DPP-IV, the carcino-embryonic antigen (CEA), and the mucin-like membrane glycoprotein MUC1, occurred beyond the cis-Golgi (Huet et al., 1998) and the trans-Golgi network (Ulloa et al., 2000). These abnormalities were accompanied by morphological changes, in particular the accumulation of numerous intracellular vesicles. GalNAcα-O-bn was also reported to reduce the apical secretion of a soluble form of mouse DPP-IV in MDCK and Caco-2 cells (Ait-Slimane et al., 2000). Surprisingly, our further work showed that enterocyte-like HT-29 cells, but not Caco-2 cells, were able to generate considerable amounts of a series GalNAcα-O-bn derived oligosaccharides (Zanetta et al., 2000). The latter observations suggested that each cell type reacted differently to GalNAcα-O-bn exposure, probably in conjunction with the specific expression pattern of glycosyltransferases. This led us to investigate the alterations induced by GalNAcα-O-bn in intracellular trafficking and glycosylation processes in the two types of cells, HT-29 and Caco-2, differentiated into an enterocytic phenotype. The brush border glycoprotein DPP-IV and the receptor for hyaluronic acid, CD44, both contain potential sites for N- and O-glycosylation (Darmoul et al., 1992; Goldstein et al., 1989), and were therefore chosen as models of transmembrane glycoproteins targeted towards the apical and basolateral membranes, respectively. The basolateral glycoprotein CD44 was furthermore reported to be associated with rafts in polarized mammary epithelial cells (Oliferenko et al., 1999), and the brush border glycoprotein DPP-IV was also described to be associated with rafts (Danielsen et al., 1995).

HT-29 differentiated cells of enterocytic phenotype were derived from the original cell line (Fogh and Trempe, 1975) by culture in a glucose-free medium (Zweibaum et al., 1985). The cells were used after several weekly passages of reversion to standard medium and are referred to as HT-29 G⁻. Cells were grown in DMEM (Eurobio, France) supplemented with 10% inactivated fetal bovine serum (FBS) for 30 minutes at 56°C (Roche Diagnostics, Meylan, France). GalNAcα-O-bn (Sigma Chemical Co., St Louis, MO, USA) was solubilized in DMEM. Cell maintenance and all experiments were done in 25-or 75-cm² T flasks (Corning Glass Works, Corning, NY, USA), and in tissue culture-treated Transwell polycarbonate membranes with a 24.5 mm diameter and 0.4-μm pore size (Costar, Cambridge, MA, USA), at 37°C in a 10% CO₂/90% air atmosphere. Cells were seeded at 2×10⁴ cells/cm². For maintenance purposes, cells were passaged weekly, using 0.025% trypsin, 0.53 mM EDTA in Ca²⁺/Mg²⁺-free PBS (PBS⁻). The medium was changed daily in all culture conditions. Caco-2 cells were cultured as previously reported (Pinto et al., 1983) and analyzed between passages 70 and 80. Cells were studied at a late postconfluent state (day 21). For GalNAcα-O-bn treatment of HT-29 G⁻ or Caco-2 cells, GalNAcα-O-bn was used at the concentration of 2 mM in DMEM with FBS, and cells were treated from day 2 after seeding up to late confluence. In all experiments, GalNAcα-O-bn treatment had no effect on cell viability, as assessed by the absence of cells in suspension and Trypan Blue exclusion.

Rat mAb 4H3 against DPP-IV (Gorvel et al., 1991), mouse mAb 525 against a basolateral glycoprotein of 38-40 kDa (gp525) (Le Bivic et al., 1989), and rabbit antibody against villin, an actin-binding protein of the apical brush border (Arpin and Friederich, 1992) were obtained from D. Massey-Harroche (CNRS Unité de Recherche Associée 1920, Faculté des Sciences de Saint Jérôme, Marseille, France), A. Le Bivic (IBDM, Marseille, France) and S. Robine (UMR 144, Institut Curie, Paris, France), respectively. Mouse mAbs against DPP-IV (M-A261), CD44 (mAb G 44-26) and the sialyl-Lewis^x (SLe^x; Neu5Acα2-3Galβ1-4[Fucα1-3]GlcNAcβ1-R) (mAb 2H5) were from Pharmingen (San Diego, USA). Mouse mAb against CD44 (clone BRIC 222) was obtained from Biogenesis (Kingston, USA). Mouse mAb against the sialyl-Lewis^a (Sle^a; Neu5Acα2-3Galβ1-3[Fucα1-4]GlcNAcβ1-R) (mAb 1H4) was from Seikagaku (Tokyo, Japan). Mouse mAbs against the H type 2 (Fucα1-2Galβ1-4GlcNAc) and A (Fucα1-2[GalNAcα1-3]Galβ1-4GlcNAc) blood group epitopes were from Dako (Trappes, France). Digoxigenin-or alkaline phosphatase-labeled Maackia amurensis agglutinin (MAA) (Wang and Cummings, 1988), Sambucus nigra agglutinin (SNA) (Shibuya et al., 1987), Ulex europeanus agglutinin I (UEA I) (Sughii et al., 1982) and peanut Arachis hypogaea agglutinin (PNA) (Lotan et al., 1975), which recognize the oligosaccharide species Neu5Acα2-3Gal-R, Neu5Acα2-6Gal-R, Fucα1-2-R and Galβ1-3GalNAc-R, respectively, were from Roche Diagnostics. Texas Red-conjugated lectins were from EY Laboratories, Inc. (San Mateo, CA, USA).

Brush border-enriched membrane fractions were prepared according to the procedure of Schmitz et al. (Schmitz et al., 1973). Cells were homogenized by sonication in 2 mM Tris/50 mM mannitol, pH 7.1, containing 40 μg/ml 4-(2-aminoethyl)-benzenesulfonylfluoride (AEBSF) at 4°C, and CaCl₂ was added at a final concentration of 10 mM. After 10 minutes incubation at 4°C, the homogenate was centrifuged for 10 minutes at 950 g. The brush border-enriched membrane fraction was collected by ultracentrifugation of the supernatant for 30 minutes at 33,500 g. Preparations were controlled by measuring the activity of the brush border hydrolase DPP-IV in the brush border-enriched membrane fraction compared to that of the total homogenate. We found an enrichment in DPP-IV activity in the brush border-enriched fractions of Caco-2 and HT-29 G⁻ cells by sevenfold and fivefold, respectively. Indeed, the expression of enterocytic markers is lower in the latter cell type (Zweibaum et al., 1985). Fractions were delipidated in chloroform/methanol mixtures (Zanetta et al., 1980) and the precipitated proteins were washed with methanol, then subjected to acid methanolysis followed by gas chromatography analysis of the heptafluorobutyrate derivatives (Zanetta et al., 1999).

Confocal microscopy was performed on cell monolayers of fully differentiated HT-29 G⁻ and Caco-2 cells grown on filters. The cells were fixed with 4% paraformaldehyde, quenched for 30 minutes with 50 mM NH₄Cl in PBS, and permeabilized with 0.2% saponin in PBS for 20 minutes. The saturation step was performed for 30 minutes in PBS containing 1% BSA and 0.2% saponin. Prior to analysis of the effect of GalNAcα-O-bn upon the cellular localization of the studied glycoproteins and glycan epitopes, double-labeling experiments were carried out in order to visualize simultaneously the basolateral and apical membranes, using gp525 and villin as markers of basolateral and apical membranes, respectively. As we have previously described, GalNAcα-O-bn did not affect the morphological polarity of the cells, as assessed by the apical expression of villin, an actin-binding protein of the apical brush border, and the tight junction protein ZO-1 (Huet et al, 1998). On the other hand, the basolateral localization of the glycoprotein gp525 is not affected by GalNAcα-O-bn treatment, like gp120 (Huet et al., 1998). To detect villin, gp525, DPP-IV, CD44, sLe^x, sLe^a, protein H and protein A antigens respectively, polyclonal rabbit anti-villin antibody (diluted 1/100), mAbs 525 (1/500), M-A261 (1/100), G 44-26 (1/150), 2H5 (1/100), 1H4 (1/150), anti-H (1/100) and anti-A (1/150) in PBS/1% BSA/0.2% saponin, were added overnight. For double labeling with anti-villin and anti-gp525 antibodies, FITC-conjugated goat anti-rabbit antibody and Texas Red-conjugated goat anti-mouse antibody (Jackson) were used. For single labeling, or double labeling with lectins, FITC-conjugated goat anti-mouse antibody (Jackson) was used. Texas Red-conjugated lectins, UEA, SNA, PNA and MAA (50 μg/ml in PBS/1% BSA/0.2% saponin) were incubated for 1.5 hours. One of the tested lectins (MAA for HT-29 G⁻ cells and UEA for Caco-2 cells) was studied in a double experiment with basolateral CD44 and/or apical DPP-IV. Laser confocal microscopy analyses were performed using a Leica-instrument (Model TCS-NT). A high resolution laser confocal analysis was carried out with the anti-DPP-IV and anti-CD44 antibodies.

Classical transmission electron microscopy was performed as previously reported (Lesuffleur et al., 1991) on cells grown in 25-cm² plastic flasks. Samples embedded in Epon (Polysciences Inc., Washington, PA, USA) were re-embedded to make sections perpendicular to the bottom of the flask.

Ultrastructural immunochemistry was performed as previously described (Huet et al., 1998). After rinsing three times in PBS, cells cultured in 25-cm² flasks were fixed in phosphate buffer containing 4% paraformaldehyde and 0.2% glutaraldehyde. The cell layer was scraped with a rubber policeman, the cell pellet was infiltrated with phosphate buffer containing 2.3 M sucrose and 20% polyvinylpyrolidone, and then frozen in liquid nitrogen. Ultrathin cryosections were successively incubated with PBS containing 10% FBS, mouse mAb M-A261 (DPP-IV) or G 44-26 (CD44), rabbit anti-mouse Ig antibody, and 8-nm gold-conjugated protein A. All antibodies and gold-conjugated protein A were diluted in PBS containing 10% FBS. The grids were finally counterstained with methylcellulose uranyl acetate and observed using an electron microscope (model 902; Carl Zeiss, Inc., Thornwood, NY, USA).

For immunoprecipitation of DPP-IV or of CD44, cells were lysed in hot buffer (5 mM Tris-HCl, pH 8.2, containing 0.5% SDS, 5 mM EDTA, 150 mM NaCl, 100 μM AEBSF), then heated at 100°C for 5 minutes, sonicated, and centrifuged at 12,000 g for 10 minutes. The supernatants were adjusted to 2.5% Triton X-100 and incubated with the antibody (mAbs 4H3 and BRIC 222, respectively) overnight at 4°C. Immunocomplexes were collected on protein G-Sepharose 4B (Sigma), eluted in the SDS-sample buffer (0.2 M Tris-HCl buffer, pH 6.8, containing 2% SDS and 30% glycerol) at 60°C for 5 minutes, and then analyzed on 5%-30% polyacrylamide gels (Laemmli, 1970). We checked that the amounts of immunoprecipitated protein were identical in control and GalNAcα-O-bn treated cells by western blotting. The electrophoreses were carried out under reducing conditions, except for the detection of CD44, because the antigenicity of CD44 towards the mAb BRIC 222 is lost after reduction. For detection of DPP-IV and CD44 by western blotting, the proteins were transferred to a hybond C extra membrane (Amersham, Aylesbury, UK). The membrane was then treated with the anti-DPP-IV (mAb 4H3) or anti-CD44 (mAb BRIC 222) antibody, followed by peroxidase-conjugated anti-rat or anti-mouse antibody (Sigma) diluted 4000-fold. Detection was carried out by luminescence using the ECL western-blotting system (Amersham). For detection of sialic acid and fucose on DPP-IV and CD44 immunoprecipitates by lectin blotting, proteins were transferred to a nitrocellulose membrane (BioTrace NT; Gelman Sciences Inc., Ann Arbor, RI, USA) as described (Vaessen et al., 1981). The membranes were then treated for 2 hours with 2% polyvinylpyrrolidone K 30 in 10 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl (TBS). Membranes were incubated with digoxigenin-labeled lectins using MAA, SNA or UEA at a concentration of 5 μg/ml in TBS. Then, the nitrocellulose membranes were incubated for 1 hour with alkaline phosphatase-labeled anti-digoxigenin Fab fragment (1 μg/ml in TBS). For the detection of T antigen (Galβ1-3GalNAc), membranes were directly incubated with alkaline phosphatase-labeled PNA lectin. The labeled glycoproteins were revealed by 4-nitroblue tetrazodium chloride/5-bromo-4-chloro-3-indolyl phosphate staining.

Total RNA was isolated by ultracentrifugation on a cesium chloride gradient (Debailleul et al., 1998). Reverse transcription into cDNA was achieved using the First-Strand cDNA Synthesis kit (Pharmacia Biotech, USA), according to the manufacturer’s protocol, using oligo-d(T) as initiation primer in a final reaction volume of 33 μl. Expression of sialyltransferases was analyzed by the Multiplex PCR method (Recchi et al., 1998), adapted to include the recently cloned ST3Gal VI (Okajima et al., 1999). Briefly, a primer pair (sense primer: 5′-CACATGATTTAAGGTGGCTGTT-3′; anti-sense primer: 5′-TGTTGGGTGTTTAGGTTTCTGA-3′) was designed in the open reading frame of ST3Gal VI to generate a 202 bp specific fragment. The specificity of the amplification was checked, using HepG2 cell cDNA as a positive control, by subcloning the amplified fragment in pCR2 plasmid (InVitrogen, Groningen, The Nederlands) and sequencing. The Multiplex PCR mixture (25 μl) consisted of 1 μl of the retrotranscription reaction, 1× AdvanTaq™ DNA polymerase (Clontech, Palo Alto, USA), 40 mM Tricine-KOH, pH 8.0, 16 mM KCl, 4.5 mM MgCl₂, 3.75 μg/ml BSA, 0.2 mM dNTP and 0.3 μM of each primer. Reactions were run in a PTC-100™ thermal cycler (MJ Research, Watertown, USA) using the following conditions: 1 minute at 94°C, 1 minute at 63°C and 2 minutes at 68°C, 32 cycles. Negative control reactions were performed by replacing cDNA template with sterile water, and positive controls were performed with HepG2 cDNA. Fucosyltransferases were individually analyzed using previously published specific primers (Mas et al., 1998; Kunzendorf et al., 1994; Hanski et al., 1996). For Fuc-TI, Fuc-TII, and for Fuc-TIII and Fuc-TV, annealing was performed at 55°C for 30 seconds and at 60°C for 1 minute, respectively. For Fuc-TIV, Fuc-TVI and Fuc-TVII, annealing and extension were performed at 72°C for 3.5 minutes. 35 cycles of amplification were done for all fucosyltransferases. The amounts of cDNA used for amplification of the glycosyltransferases were checked by PCR amplification of GAPDH.

Cells were lysed in 10 mM cacodylate buffer, pH 6.5, containing 1% Triton CF-54, 20% glycerol and 5 mM DTT. Assays were performed in 0.1 M cacodylate buffer, pH 6.2, 10 mM MnCl₂, 0.2% Triton CF-54, 50 μM CMP-[¹⁴C]-Neu5Ac (N-acetyl neuraminic acid) (1.85 kBq), with 4 mM of acceptor substrate (Galβ1-3GalNAcα-bn, Galβ1-3GlcNAcβ-octyl, or Galβ1-4GlcNAc) and 20 μl of the cell lysate in a final volume of 50 μl. The enzyme reaction was performed at 37°C for 1 hour and reaction products were separated from CMP-[¹⁴C]-Neu5Ac, depending on the acceptor substrate. For arylglycosides, the reaction was stopped by the addition of 450 μl H₂O and products were isolated by hydrophobic chromatography on C₁₈ SepPak cartridges (Millipore Corp., Milford, MA, USA). After washing with 10 ml water, arylglycosides were eluted with 5 ml 30% acetonitrile/H₂O. For N-acetyllactosamine, Neu5Acα2-3 and Neu5Acα2-6-isomers were separated by ion exchange high performance liquid chromatography (HPLC) as previously described (Cacan et al., 1994).

α1,2- and α1,3-fucosyltransferase activities were measured according to the procedure of Goupille et al. (Goupille et al., 2000) with detection by high performance anion exchange chromatography with pulsed amperometric detection (HPAEC/PAD) (Dionex Corp.). Cells were lysed in 50 mM potassium phosphate, pH 6.0, containing 2% (v/v) Triton X-100 and 1 mM AEBSF on ice for 30 minutes. The reaction was carried out with 50 μg of proteins, 20 μM GDP-L-[¹⁴C]-fucose (1.83 kBq), 20 μM N-acetyllactosamine, 10 mM L-fucose, 7.7 mM MgCl₂, 1.9 mM ATP at 37°C. The optimal incubation time was 10 minutes for Caco-2 cells and 1 hour for HT-29 cells. The reaction was stopped by addition of methanol, and the oligosaccharides were injected onto a Carbolac PA-100 Column (4×250 mm). The elution was monitored both by pulsed amperometric detection (PAD2 model, Dionex Corp.) and by radioactivity on-line (high performance liquid chromatography radioactivity detector LB 506 C-1, EG & G, Berthold), using Fucα1-2Galβ1-4GlcNAc, Galβ1-4[Fucα1-3]GlcNAc and Fucα1-2Galβ1-4[Fucα1-3]GlcNAc as standards.

α1,4-fucosyltransferase activity was determined in 125 mM MES buffer, pH 7.0, 12.5 mM ATP, 12.5 mM MnCl₂, with 50 μg of cellular proteins, using 50 μM GDP-[¹⁴C]-Fuc (1.83 kBq) and 1 mM of Galβ1-3GlcNAcβ-octyl as acceptor substrate, at 37°C for 2 hours. The reaction was stopped by dilution with distilled water and products were isolated by hydrophobic chromatography on C₁₈ SepPak cartridges (Millipore Corp, Milford, MA, USA).

Dipeptidyl peptidase IV activity was measured by a kinetic method with the fluorogenic substrate H-Gly-Pro-AMC (Bachem, Voisins-le-Bretonneux, France) according to Gotoh et al. (Gotoh et al., 1989) with some modifications, using a Cobas Fara II centrifugal analyzer (Roche Diagnostics, Meylan, France). Stock and working solutions were prepared as follows: H-Gly-Pro-AMC (20 mmol/l) was dissolved in dimethyl sulfoxide, and the working solution was obtained by diluting 50 μl of stock solution in 450 μl of 1g/l aqueous Brij 35 (polyoxyethylene monolauryl ether). The stock solution of amino-4-methylcoumarin (AMC) (0.25 mmol/l) was diluted 100-fold with 0.15 M NaOH-glycine buffer, pH 8.7, just before use. Briefly, the apparatus distributes successively: (a) 12.5 μl of sample and 250 μl of 0.15 M NaOH-glycine buffer, pH 8.5, (b) 12.5 μl of the AMC working solution as internal standard, and (c) 20 μl of the substrate working solution. The first reading was taken 30 seconds after mixing the sample and the NaOH-glycine buffer, and the second reading after adding AMC. 15 further readings were taken at 1 minute intervals after adding the substrate. The fluorescence was measured with excitation at 340 nm and emission at 450 nm. 1 mU of activity corresponds to the release of 1 nmol amino-4-methylcoumarin per minute.

Proteins were determined by using bicinchoninic acid protein kit (Smith et al., 1985) from Pierce (Rockford, Illinois, USA) with bovine serum albumin as a standard.

The effect of permanent GalNAcα-O-bn treatment upon the intracellular trafficking of DPP-IV and CD44 was studied by confocal microscopy in polarized enterocyte-like HT-29 G⁻ and Caco-2 cells, comparatively, both proteins being potentially N- and O-glycosylated. In order to analyze the effect of GalNAcα-O-bn on polarized differentiated cells, and also regarding the low turnover rate (over 48 hours) of DPP-IV (Darmoul et al., 1992), GalNAcα-O-bn was used as a permanent treatment starting well before the differentiation process and continuing up to the differentiation state. GalNAcα-O-bn was added at a concentration of 2 mM in the culture medium from day 2 after seeding up to late confluence (day 21). As markers of the two poles of these cells, we used the non-glycosylated protein villin, an actin-binding protein of the apical brush border, and the glycoprotein gp525, an N-glycosylated glycoprotein for which the processing was shown to be unaffected in HT-29 MTX cells by GalNAcα-O-bn (Ulloa et al., 2000). Villin and gp525 were used as markers of the apical and basolateral sides of the cells, respectively, by double labeling and confocal microscopy (Fig. 1 for HT-29 G⁻ cells and Fig. 2 for Caco-2 cells). The results of confocal microscopy in HT-29 G⁻ cells (Fig. 1) showed that, under GalNAcα-O-bn exposure, neither DPP-IV nor CD44 were expressed at the apical or basolateral membranes, respectively, but remained strongly accumulated in the cytoplasm. No such changes in the cellular distribution were observed for DPP-IV and CD44 in Caco-2 cells (Fig. 2; the results for CD44 could not been shown on the picture due to low level of fluorescence). These glycoproteins remained localized at the apical and basolateral membrane, respectively. However, a reduction of the apical expression of DPP-IV was observed in GalNAcα-O-bn treated Caco-2 cells.

In order to gain insight into the subcellular localization of DPP-IV and CD44 at the ultrastructural level in GalNAcα-O-bn treated HT-29 cells, high-magnification laser confocal analysis and ultrastructural immunochemistry were carried out with the anti-DPP-IV and anti-CD44 antibodies (Fig. 3). High-magnification laser confocal analysis showed that DPP-IV (Fig. 3d) and CD44 (Fig. 3e) were accumulated as a granular pattern all over the cytoplasm. At the morphological level, we observed that GalNAcα-O-bn induced the formation of numerous vesicles that filled the cytoplasm (Fig. 3b,c). Using ultrastructural immunohistochemistry, both DPP-IV (Fig. 3f) and CD44 (Fig. 3g) were found localized in these vesicles clearly delimited by a membrane.

As we had previously observed that GalNAcα-O-bn dramatically inhibited the incorporation of sialic acid in the apically secreted gel-forming mucins produced by goblet HT-29 cells, we investigated the effect of this compound upon the glycosylation of the brush border membrane glycoproteins. We analyzed the monosaccharide composition of the brush border-enriched membrane fractions isolated from control and GalNAcα-O-bn treated HT-29 G⁻ cells and Caco-2 cells. The results are shown in Table 1. The sialic acid content was higher in HT-29 G⁻ cells than in Caco-2 cells. GalNAcα-O-bn treatment induced a specific decrease in the sialic acid content in HT-29 G⁻ cells, whereas a decrease in both sialic acid and fucose contents were observed in Caco-2 cells.

A panel of lectin- and glycan-specific antibodies was used in order to gain more knowledge about the glycosylation of the apical and basolateral membranes in both cell types cultured on filters in standard conditions or in presence of GalNAcα-O-bn, using confocal microscopy.

MAA stained the apical membrane of HT-29 G⁻ cells intensely, and the basolateral membrane at a much lower level (Fig. 4). These localizations were confirmed using double-labeling experiments, i.e. MAA/DPP-IV and MAA/CD44. PNA staining was observed all over the cell, and in particular in large vesicular structures. No labeling was significantly observed with SNA and UEA (data not shown). After GalNAcα-O-bn treatment, MAA labeling completely disappeared at the apical membrane, and was instead found in the cytoplasm. The double-labeling experiments showed that the changes observed for MAA binding were similar to those observed for DPP-IV and CD44. An increase in the PNA labeling was observed, and the staining was present all over the cells.

All the studied lectins were able to bind to Caco-2 cells. SNA (Fig. 5) and MAA (data not shown) stained the apical membrane intensely, but also the basolateral membrane to a lesser extent. UEA strongly stained both the apical and basolateral membranes. The double-labeling experiment UEA/DPP-IV showed that DPP-IV labeling was colocalized with the UEA labeling at the apical side of the cells. A PNA labeling was visualized in the cytoplasm, but also at the apical membrane of the control cells. GalNAcα-O-bn treatment was unable to induce a change in the distribution of the SNA staining, nor of the MAA staining. In contrast, a strong decrease in the UEA staining, at both apical and basolateral membranes, was observed. On the other hand, the PNA labeling became strongly increased all over the cell, including the apical membrane.

To better characterize the oligosaccharide structures present in both cell types, and to determine which ones were altered after GalNAcα-O-bn treatment, we studied the distribution of the blood group antigens (A for HT-29 G⁻ cells, and H for Caco-2 cells) and of the carbohydrate antigens sLe^a and sLe^x, by confocal microscopy with specific monoclonal antibodies. In HT-29 G⁻ cells (Fig. 6), the epitopes sLe^x, sLe^a, and the blood group epitope A, were highly associated with the apical brush border. After GalNAcα-O-bn treatment, the expression of sLe^x disappeared at the apical brush border and was found in the totality of the cytoplasm. The expression of sLe^a completely disappeared in most GalNAcα-O-bn treated cells, remaining only visualized in few cells, all over the cytoplasm.

But in contrast, no significant change was observed for the blood group epitope A.

In Caco-2 cells (Fig. 7), the blood group H type 2 antigen was strongly expressed at the apical brush border. The epitopes sLe^x and sLe^a were also visualized at the apical brush border. After GalNAcα-O-bn treatment, the apical expression of H type 2 antigen strongly decreased, and only some cells displayed a granular cytoplasmic labeling with the specific antibody. The expression of sLe^x epitope remained apically localized, but that of sLe^a completely disappeared.

Two main conclusions emerge from these experiments, concerning the types of glycan alterations: (1) GalNAcα-O-bn alters the distribution of all α2,3-sialylated epitopes in HT-29 G⁻ cells, but that of non-sialylated α1,2-fucosylated epitopes in Caco-2 cells, and (2) the altered glycan epitopes became intracellularly accumulated in treated HT-29 G⁻ cells, but not in Caco-2 cells.

In addition to the study of the cellular distribution of glycan epitopes, the specific glycosylation of the apical glycoprotein (DPP-IV) and of the basolateral glycoprotein (CD44) was investigated (Fig. 8). We particularly looked for the glycan epitopes modified in the cells by GalNAcα-O-bn. The two glycoproteins were isolated by immunoprecipitation, as described in Materials and Methods, and we checked that identical amounts of these glycoproteins were obtained in control and GalNAcα-O-bn treated cells using western blotting with anti-DPP-IV and anti-CD44 antibodies. Furthermore, we also checked that the amount of DPP-IV activity was similar in control and GalNAcα-O-bn treated cells (data not shown). However, while DPP-IV appeared to be expressed at a similar level in both cell types, CD44 appeared highly expressed in HT-29 G⁻ cells, but detectable only as a trace amount in Caco-2 cells.

In control HT-29 G⁻ cells, DPP-IV reacted only with MAA. In GalNAcα-O-bn treated HT-29 G⁻ cells, DPP-IV lost the MAA labeling, and was revealed by PNA. In control Caco-2 cells, DPP-IV was labeled neither by MAA nor by UEA, but with SNA. In GalNAcα-O-bn treated Caco-2 cells, no PNA labeling appeared, and the SNA staining persisted.

In HT-29 G⁻ cells, as for DPP-IV, CD44 was found labeled by MAA, and GalNAcα-O-bn completely inhibited the MAA labeling and strongly induced the expression of the T-antigen. In Caco-2 cells in contrast, CD44 displayed a glycosylation pattern different from that of DPP-IV in the same cells. CD44 did not appear labeled by SNA, but with UEA. Furthermore, GalNAcα-O-bn induced changes in the glycosylation of CD44, with the disappearance of the UEA labeling, and the appearance of a PNA labeling.

With regard to the differential effects of GalNAcα-O-bn upon the glycosylation in HT-29 G⁻ and Caco-2 cells, we examined the expression of the sialyltransferases and fucosyltransferases in control and GalNAcα-O-bn treated HT-29 G⁻ and Caco-2 cells.

The expression of ST3Gal I, ST3Gal II, ST3Gal III, ST3Gal IV, ST3Gal VI and ST6Gal I, was assessed by multiplex RT-PCR in control and treated HT-29 G⁻ and Caco-2 cells (Fig. 9). Different expression patterns were observed for HT-29 G⁻ and Caco-2 cells. HT-29 G⁻ cells mainly expressed ST3Gal I, while Caco-2 cells mainly expressed ST6Gal I. Furthermore, the number of cycles used for the multiplex RT-PCR was calculated in order to visualize all the expressed sialyltransferases, and in these conditions ST3Gal I is at a saturation step in HT-29 G⁻ cells as well as ST6Gal I in Caco-2 cells. In addition to these respective main sialytransferases, HT-29 G⁻ cells weakly expressed ST3Gal IV and ST6Gal I, and Caco-2 cells ST3Gal II, ST3Gal I and ST3Gal VI. The expression patterns significantly changed after GalNAcα-O-bn treatment, showing an increased expression of different sialyltransferases involved in the transfer of sialic acid in α2,3-linkage in both cell types. We observed a marked increase in ST3Gal IV amplification product, and an induction of the expression of ST3Gal II and ST3Gal III in treated HT-29 G⁻ cells. Treated Caco-2 cells displayed an increase in ST3Gal VI amplification product, and an induction of the expression of ST3Gal III and ST3Gal IV.

The fucosyltransferases Fuc-TI to Fuc-TVII were individually amplified in comparison to positive controls (Fig. 10). Several fucosyltransferases were detected in both cell lines: Fuc-TI, Fuc-TII, Fuc-TIII, Fuc-TVI and Fuc-TVII in HT-29 G⁻ cells, Fuc-TI, Fuc-TII, Fuc-TIII, Fuc-TV, Fuc-TVI and Fuc-TVII in Caco-2 cells. Fuc-TIV was not expressed in either cell line (not shown). The results showed that GalNAcα-O-bn treatment markedly induced the expression of Fuc-TVII in Caco-2 cells.

In order to specify the differences in the glycosyltransferase pattern between the two cell types we undertook further investigations at the enzymatic level.

The results of sialyltransferase assays in control and GalNAcα-O-bn treated HT-29 G⁻ and Caco-2 cells are shown in Table 2. In HT-29 G⁻ cells, the highest transfer of sialic acid was observed onto Galβ1-3GalNAcα-O-bn, which is a specific substrate for ST3Gal I and ST3Gal II. However, as ST3Gal II was not detected in the multiplex RT-PCR assay of control HT-29 G⁻ cells, the activity is very likely relevant to ST3Gal I, the main sialyltransferase expressed in these cells. The transfer of sialic acid in α2,3-linkage to the type 1 Galβ1-3GlcNAc sequence or to the type 2 Galβ1-4GlcNAc sequence was detected at a much lower level. As ST3Gal III and ST3Gal VI were not detected in control HT-29 G⁻ cells by multiplex RT-PCR, these transfers can be attributed to ST3Gal IV. Finally the transfer of sialic acid in α2,6-linkage was also detected at a very low level. In Caco-2 cells, the transfer in α2,3-linkage to Galβ1-3GalNAc was observed at a much lower level in comparison to HT-29 G⁻ cells. The main enzyme corresponded to the addition of sialic acid in α2,6-linkage, in accordance also with the dominant expression of ST6Gal I observed by RT-PCR. A transfer of sialic acid in α2,3-linkage could be detected on the type 2 sequence but not on the type 1 sequence. The latter activity can be attributed to ST3Gal VI, as ST3Gal IV did not appear expressed in control Caco-2 cells. The results of enzymatic activity in control Caco-2 cells are in accordance to those previously published by Dall’Olio et al. (Dall’Olio et al., 1996). All these activities became increased by GalNAcα-O-bn.

The results of fucosyltransferase assays are shown in Table 3. In HT-29 G⁻ cells, a transfer of fucose was visualized in α1,2-linkage to the type 2 sequence, in accordance to the expression of Fuc-TI and Fuc-TII. A transfer in α1,3-linkage was also detected on the type 2 sequence. This activity corresponds to Fuc-TVI, as Fuc-TIV and Fuc-TV were not amplified by RT-PCR in HT-29 cells. Although Fuc-TIII appeared to be expressed at a low level by RT-PCR, the typical activity of this α1,3/1,4-fucosyltransferase was not clearly detectable. In Caco-2 cells, the transfer of fucose in α1,2-linkage was much higher than in HT-29 cells, indicating a higher expression of Fuc-TI and/or Fuc-TII. Likewise, the α1,3-fucosylation was much higher in this cell type, and includes two expressed α3Fuc-T in this cell type, Fuc-TV and Fuc-TVI. The activity of Fuc-TIII was clearly evaluated on the type 1 sequence in this cell type. As for the sialyltransferase activities, all the fucosyltransferase activities became increased by GalNAcα-O-bn.

The results clearly showed strong differences in the glycosyltransferase expression patterns between the two cell types. α2,3 sialyltransferase activity of ST3Gal I is expressed at a very high level in HT-29 G⁻ cells. In contrast, this activity is much lower in Caco-2 cells, for which the main sialyltransferase activity rather occurs in α2,6 linkage by ST6Gal I. On the other hand, Caco-2 cells display much higher fucosyltransferase activity than HT-29 G⁻ cells.

In this study, we investigated whether alteration in the glycosylation process by the competitive inhibitor of O-glycosylation, GalNAcα-O-bn, could be related to disruptions in intracellular trafficking in polarized epithelial cell lines of the enterocytic phenotype. We studied two glycoproteins containing potential sites for both N- and O-glycosylation, of basolateral and apical membrane localization, respectively. Our results obviously revealed that the targeting of these two glycoproteins towards their membrane localization in HT-29 G⁻ cells was specifically blocked by GalNAcα-O-bn treatment. Both glycoproteins remained stored in numerous intracytoplamic vesicles filling the cells. These vesicles also contain GalNAcα-O-bn metabolites which specifically accumulate (1 mg hexose/mg protein) in the HT-29 cell type (Zanetta et al., 2000). In contrast, the basolateral localization of gp525, an N-glycoprotein whose sialylation was not affected by GalNAcα-O-bn (Ulloa et al., 2000), was unchanged after GalNAcα-O-bn treatment, as we have previously shown for another basolateral membrane glycoprotein, gp120 (Huet et al, 1998). These observations led to investigations on the changes induced by GalNAcα-O-bn upon the glycosylation of DPP-IV and CD44 in HT-29 G⁻ cells.

The addition of sialic acid mainly occurs in α2,3-linkage in HT-29 G⁻ cells. The α2,3-transfer is mainly processed through ST3Gal I, which is specific to O-glycans, and through ST3Gal IV acting on both O- and N-linked glycans. Permanent GalNAcα-O-bn exposure of HT-29 G⁻ cells induced a marked alteration in the distribution of α2,3-linked sialic acid. All the oligosaccharide structures substituted by α2,3-linked sialic acid, involving fucose or not, were not expressed at the apical membrane, and appeared instead to be localized inside the cells. The latter observation is at least partially relevant to our previous work demonstrating the biosynthesis of a high amount of several sialylated GalNAcα-O-bn oligosaccharides, including sialyl T, sLe^x, disialyl T and disialyl Le^x motifs (Zanetta et al., 2000). Interestingly, no change was seen for the blood group A epitope, which corresponds to a fucosylated but non-sialylated epitope. Therefore, alteration in the distribution of glycan epitopes was restricted to α2,3 sialylated determinants in GalNAcα-O-bn treated HT-29 G⁻ cells.

Our data clearly showed that DPP-IV is substituted by α2,3-linked sialic acid in HT-29 G⁻ cells, and that the permanent GalNAcα-O-bn treatment completely inhibited this step of sialylation, and generated the expression of the core 1 structure Galβ1-3GalNAc (T-antigen) on the glycoprotein. This antigen is a structure specifically O-linked to threonine or serine residues. DPP-IV thus carries the O-linked oligosaccharide structure Neu5Acα2-3Galβ1-3GalNAc. This structure could be connected with the presence of O-linked glycan(s) in DPP-IV (Matter et al., 1989; Naim et al., 1999) and the potential O-glycosylation sites in the stalk of DPP-IV (Darmoul et al., 1992). These results suggested that the impairment of Neu5Acα2,3-substitution of the T structure might be related to the blockade in the apical delivery of DPP-IV. Indeed, recent work has suggested that O-glycosylation could be involved in the apical sorting of brush border glycoproteins in polarized cells. The potential O-glycosylation sites of the stalk of the human neurotrophin receptor are necessary for the apical delivery of the glycoprotein in MDCK and Caco-2 cells (Yeaman et al., 1997; Monlauzeur et al., 1998). Naim et al. (Naim et al., 1999) recently reported that deoxymanojirimycin, an inhibitor of the processing of N-glycans, was able to induce a reduction of the O-glycosylation of DPP-IV and partially shifted the apical delivery of the glycoprotein towards the basolateral membrane in Caco-2 cells. GalNAcα-O-bn was also reported to inhibit the O-glycosylation of the brush border glycoprotein sucrase-isomaltase and its selective targeting to the apical membrane (Alfalah et al., 1999). Recently, the O-glycosylated domain within the sequence of 12 amino acids Ala³⁷-Pro⁴⁸ of the stalk of sucrase-isomaltase was shown to represent an apical sorting signal in MDCK cells (Jacob et al., 2000). Our data add to information about the nature of O-glycosylation that could be involved in such process, and suggest that the terminal elongation of the O-glycan Galβ1-3GalNAc, i.e. the sialylation by ST3Gal I, might constitute a prerequisite step for apical delivery of DPP-IV in enterocytic HT-29 cells.

As for DPP-IV, GalNAcα-O-bn completely inhibited the α2,3-sialylation of CD44 and greatly induced expression of the T antigen, in agreement with observations made in B16BL6 melanoma cells (Nakano et al., 1996). Altogether, our data suggest that the sialylation by ST3Gal I might also be a key step in intracellular transport of this glycoprotein towards the basolateral membrane, bearing in mind the fact that distribution of the normally sialylated basolateral glycoproteins gp120 and gp525 (Ulloa et al., 2000) was normal. The selective sensitivity of CD44 to GalNAcα-O-bn could be due to an association of this glycoprotein with lipid rafts, which was recently reported in polarized mammary epithelial cells (Oliferenko et al., 1999). Surprisingly, the permanent GalNAcα-O-bn treatment of Caco-2 cells did not result in a similar alteration of the localization of DPP-IV. Indeed, DPP-IV does not become intracellularly stored. Nevertheless, a reduction of DPP-IV activity by GalNAcα-O-bn has been previously reported (Amano and Oshima, 1999), and we also observed a reduction of the apical expression of DPP-IV by confocal microscopy.

Another question was whether DPP-IV is also modified in its glycosylation state. We show here that the glycosylation of DPP-IV is completely different in Caco-2 cells in comparison to HT-29 G⁻ cells. The substitution by sialic acid is only processed in α2,6-linkage and not in α2,3-linkage. The transfer of sialic acid was unaffected by GalNAcα-O-bn, and expression of the O-linked core 1 structure Galβ1-3GalNAc could not be observed. Indeed, the activity of ST3Gal I is very low in Caco-2 cells, and substitution by this enzyme is not predominant. On the other hand, substitution by ST6Gal I cannot be prevented by GalNAcα-O-bn in Caco-2 cells, as no complex GalNAcα-O-bn derivatives, particularly potential competitors of this glycosyltransferase, were revealed in this cell type (Zanetta et al., 2000). The single derivative visualized in high amounts in treated Caco-2 cells was the disaccharide Galβ1-3GalNAcα-O-bn. This derivative acts as a main competitor to ST3Gal I in HT-29 G⁻ cells, but ST3Gal I is weakly expressed in Caco-2 cells. However, Galβ1-3GalNAcα-O-bn is also a potential substrate for the fucosyltransferases α2-Fuc-T (Brockhausen, 1995), which displays a high activity in Caco-2 cells. In this way, the strong inhibition of the expression of H type 2 blood group antigens in GalNAcα-O-bn treated Caco-2 cells, in addition to the inhibition of the expression of H type 1 blood group antigen previously described (Amano and Oshima, 1999), testifies that the substitution by α2-Fuc-T fucosyltransferases is markedly inhibited in this cell type. Furthermore, the increase in expression of the T antigen Galβ1-3GalNAc in treated Caco-2 cells clearly demonstrates that an important inhibition of the elongation of the T antigen also occurs in this cell type. Thus in Caco-2 cells, GalNAcα-O-bn treatment mainly results in an inhibition of the substitution by the fucosyltransferases α2-Fuc-T.

In conclusion, our data show that the permanent GalNAcα-O-bn treatment alters the O-glycosylation of glycoproteins in the two cell types differently. GalNAcα-O-bn was initially designed as a competitive inhibitor of the elongation of O-glycans by core-1 β-3Gal T. In fact, the high affinity of the core-1 β-3Gal T for GalNAcα-O-bn generates a high amount of the disaccharide Galβ1-3GalNAcα-O-bn inside the cells. This disaccharide further constitutes a competitive substrate for sialyltransferases, N-acetylglucosaminyltransferases and/or fucosyltransferases (Huang et al., 1992). The final inhibition of the glycosylation will depend upon the relative expression pattern of these glycosyltransferases. So, the major effects would be an inhibition of the addition of α2,3-linked sialic acid in HT-29 G⁻ cells, but of α1,2-linked fucose in Caco-2 cells. This specificity of GalNAcα-O-bn between the two cell types is very likely relevant to the difference observed in their cellular responses. In this way, GalNAcα-O-bn induces a specific cellular phenotype with enterocyte-like HT-29 cells, morphologically characterized by the formation of numerous cytoplasmic vesicles, which contain GalNAcα-O-bn metabolites and the defectively sialylated apical (DPP-IV) or basolateral (CD44) O-glycosylproteins. Besides, it may be suggested that the sensitivity of both these glycoproteins to GalNAcα-O-bn exposure could be also related to an association with lipid rafts during their intracellular trafficking.

The nomenclature of sialyltransferases is based on Tsuji et al. (Tsuji et al., 1996, Glycobiology6, v-vii) and that of fucosyltransferases on Oriol et al. (Oriol et al., 1999, Glycobiology9, 323-334). Core-1 β3-Gal-T: UDP-Gal: GalNAc-R β1,3-galactosyltransferase, EC 2.4.1.122; α2-Fuc-T: GDP-Fuc: Galβ-R α1,2-fucosyltransferase, including Fuc-TI FUT1-encoded α1,2 fucosyltransferase and Fuc-TII FUT2-encoded α1,2 fucosyltransferase; α3-Fuc-T: GDP-Fuc: Galβ1-4GlcNAc (Fuc to GlcNAc) α1,3-fucosyltransferase, including Fuc-TIV (myeloid enzyme) FUT4-encoded α1,3-fucosyltransferase, Fuc-TV FUT5-encoded α1,3-fucosyltransferase, Fuc-TVI (plasma enzyme, EC 2.4.1.152) FUT6-encoded α1,3-fucosyltransferase, and Fuc-TIX FUT9-encoded α1,3-fucosyltransferase; Fuc-TVII: GDP-Fuc: Neu5Acα2-3Galβ1-4GlcNAc α1,3-fucosyltransferase, EC 2.4.1.-; ST3Gal I: CMP-Neu5Ac: Galβ1-3GalNAc α2,3-sialyltransferase, EC 2.4.99.4; ST3Gal II: CMP-Neu5Ac: Galβ1-3GalNAc α2,3-sialyltransferase, EC 2.4.99.-, ST3Gal III: CMP-Neu5Ac: Galβ1-3(4)GlcNAc α2,3-sialyltransferase, EC 2.4.99.6; ST3Gal IV: CMP-Neu5Ac: Galβ1-4GlcNAc α2,3-sialyltransferase, EC 2.4.99.-; ST3Gal VI: CMP-Neu5Ac: Galβ1-4GlcNAc α2,3-sialyltransferase, EC 2.4.99.-; ST6Gal I: CMP-Neu5Ac: Galβ1-4GlcNAc α2,6-sialyltransferase, EC 2.4.99.1.

We thank M. J. Dejonghe (from the University Lille II), O. Moreau and B. Hemon (both from INSERM U377) for excellent technical assistance. Galβl-3GlcNAcβ-O-octyl was the generous gift from Dr C. Augé (Laboratoire de Chimie Organique Multifonctionnelle, UMR CNRS 8614, Université de Paris-Sud, Orsay, France). We are extremely grateful to Dr J. F. Dubremetz for his help in ultrastructural immunochemistry, Drs D. Massey-Harroche, S. Robine and A. Le Bivic for their gift of antibodies, and to Dr A. Zweibaum and F. X. Real for helpful discussions. Confocal microscopy was carried out in the IFR 22 (Biologie et Pathologie des régulations cellulaires). Wim Steelant has a fellowship from the European Carbohydrate Research Network CARENET 2. This work was supported by a grant from ARC (N°5687) and the European Carbohydrate Research Network CARENET 2.