Differential Distribution of Sialic Acid in α2,3 and α2,6 Linkages in the Apical Membrane of Cultured Epithelial Cells and Tissues

Abstract

We used lectin cytochemistry and confocal microscopy to examine the distribution of sialic acid in epithelial cells. Maackia amurensis lectin and Sambuccus nigra agglutinin were used to detect α2,3 and α2,6 sialic acid, respectively. In Caco-2, HT-29 5M12, and MCF-7 cells, which express sialic acid mainly in one type of linkage, the majority of the signal was observed in the apical membrane. In cells that bound both lectins, α2,3 sialic acid was distributed apically, whereas α2,6 sialic acid showed a broader distribution. In IMIM-PC-1 cultures, α2,3 sialic acid was detected mainly in the apical membrane, whereas α2,6 sialic acid was more abundant in the basoleral domain of polarized cells. In these cells, treatment with GalNAc-O-benzyl led to reduced α2,3 levels and to an increase and redistribution of α2,6 to the apical domain. Similarly, sialic acid was predominantly expressed apically in all epithelial tissues examined. In conclusion, (a) sialic acid is mainly distributed to the apical membrane of epithelial cells, (b) there is a hierarchy in the distribution of sialic acids in polarized epithelial cells, i.e., α2, 3 is preferred to α2, 6 in the apical membrane, and (c) IMIM-PC-1 cells are a good model in which to study the regulation of the levels and distribution of sialic acids.

Sialic acids are a diverse group of monosaccharides present at the outermost end of N-linked and O-linked carbohydrate chains and in lipid-associated glycoconjugates. Sialic acids are found in glycoconjugates of some bacteria and fungi and in animals of the deuterostome lineage (Schauer 1982). In mammalian cells, the most common sialic acids are N-acetylneuraminic acid and N-glycolylneuraminic acid, although only the former is present in human cells (Chou et al. 1998). The structural diversity of sialic acids arises not only from the nature of the monosaccharide but also from its linkage to other sugars, which occurs in three main configurations: α2,3, α2,6, and α2,8. The latter form generally exists as homopolymers in a few glycoproteins expressed in neural tissue, such as N-CAM, and in gangliosides (for a review of the chemistry and biology of sialic acids see Schauer 1982; Reutter et al. 1997).

Despite extensive knowledge about the structural features of sialic acids, relatively little is known about their biological role (Varki 1993). The extended nature of oligosaccharide chains, and possibly their negative charge, plays an important role in cell–cell and cell–matrix interactions (reviewed in Kelm and Schauer 1997). The best physiological evidence supporting such a contention comes from the demonstration that sialyl-Le^x and sialyl-Le^a are ligands for selectins and play a major role in the early steps of leukocyte rolling on endothelial cells (reviewed in Tedder et al. 1995).

Sialyltransferases constitute a family of more than 15 enzymes that transfer sialic acid from CMP–sialic acid to glycoconjugates. They exhibit a notable specificity for both linkage and acceptor substrates. Like other glycosyltransferases, they are Type II glycoproteins residing in trans-Golgi cisternae and the trans-Golgi network. Their expression is highly tissue-specific and is finely regulated during embryonic development and cell differentiation (Tsuji 1996).

In the course of studies on the biosynthesis of mucin glycoproteins by mucous-secreting HT-29 colon cancer cells, we observed that the major effect of short-term treatment with the sugar analogue GalNAc-O-benzyl was not an inhibition of O-glycosylation, as previously reported, but a profound inhibition of sialylation of de novo-synthesized mucin glycoproteins (Huet et al. 1995). This effect resulted from a competitive inhibition of α2,3 sialyltransferases by GalNAc-O-benzyl metabolites produced by the cells (Huet et al. 1995; Delannoy et al. 1996; Zanetta et al. 2000). Indeed, the main sialyltransferase activity detected in HT-29 cells is of the α2, 3 type (Dall'Olio et al. 1993; Huet et al. 1995, 1998). A more thorough analysis of the effects of prolonged treatment with GalNAc-O-benzyl on these cells disclosed several interesting effects. First, in polarized monolayers of mucus-producing HT-29 cells, sialic acid was predominantly distributed to the apical membrane. Second, inhibition of sialylation induced by GalNAc-O-benzyl was associated with an increase in the levels of ST3 Gal I mRNA and activity. Finally, apical glycoproteins were mislocalized in intracellular vesicles accumulating in treated cells (Hennebicq–Reig et al. 1998; Huet et al. 1998). These observations led to the proposal that sialic acid may constitute an apical targeting signal for glycoproteins. We have recently reported that both apical and basolateral glycoproteins are sialy-lated in HT-29 cells. Nevertheless, GalNAc-O-benzyl selectively inhibits the sialylation of apical and lysosomal glycoproteins but it does not affect the two basolateral glycoproteins analyzed (Ulloa et al. 2000). These findings raise important questions about the biosynthesis and intracellular trafficking of membrane sialoglycoproteins as well as the regulation of sialylation.

In this study, we analyzed the distribution of sialic acid in a panel of human cultured cells and examined the effects of treatment with GalNAc-O-benzyl on sialoglycoconjugates in a pancreatic cancer cell line, designated IMIM-PC-1, displaying high levels of ST3Gal I. This work is the basis for a more in-depth analysis of the putative role of sialic acid in glycoprotein targeting in epithelial cells and provides novel cell models in which to analyze the regulation of sialylation.

Materials and Methods

Reagents, Antibodies, and Lectins

Dulbecco's modified minimal essential medium (MEM) and fetal bovine serum (FBS) were purchased from Gibco Life Technologies (Glasgow, UK). GalNAc-O-benzyl, polyvinylpyrrolidone, and Clostridium perfringens neuraminidase were from Sigma Chemical (St Louis, MO). Streptavidin–rhodamine and streptavidin-peroxidase were from Pierce Chemical (Rockford, IL) and Zymed (South San Francisco, CA), respectively. A mouse monoclonal antibody (MAb) TS2/16 against an extracellular epitope of the β1 integrin subunit (Arroyo et al. 1992) was obtained from F. Sánchez-Madrid (Hospital de la Princesa; Madrid, Spain). Mouse MAb A9 detecting an extracellular epitope of the β4 integrin subunit (Kimmel and Carey 1986; Van Waes et al. 1991) was a gift from T. Carey (University of Michigan; Ann Arbor, MI). Mouse MAb 36, detecting the cytoplasmic domain of human and dog E-cadherin, was purchased from Transduction Laboratories (Lexington, KY). Biotinylated Maackia amurensis lectin (MAL) and Sambuccus nigra agglutinin (SNA), recognizing the oligosaccharide species NeucAc α2,3Gal-R and NeucAc α2,6Gal/GalNAc-R, respectively, were from Vector Labs (Burlingame, CA). The specificity of these lectins and their selectivity for sialic acid linkage has been extensively demonstrated in the past (Shibuya et al. 1987; Wang and Cummings 1988).

Cell Culture

A clone derived from HT-29 colon cancer cells and designated HT-29 5M12 was selected because it responds to GalNAc-O-benzyl in a manner similar to that of the unselected parental population, but cells display a polarized absorptive phenotype in postconfluence (Lesuffleur et al. 1998), thus facilitating the analysis of apical membrane glycoproteins without interference from mucus. Caco-2 colon cancer cells display an enterocytic phenotype. Both lines were provided by T. Lesuffleur (INSERM U505; Paris, France). MDCK I cells were provided by B. Alarcón (Centro de Biología Molecular Severo Ochoa; Madrid, Spain), MDCK II cells were provided by K. Matter (University of Geneva; Geneva, Switzerland), MZPC-1 pancreatic cancer cells were obtained from A. Knuth (Nordwestern Krankenhaus; Frankfurt, Germany), MCF-7 breast cancer cells were from the ATCC (Manassas, VA), and IMIM-PC-1 and IMIM-PC-2 pancreatic cancer cells were established in our laboratory (Vilá et al. 1995). Cells were seeded at the following densities: HT-29 5M12, HRT-18, and MCF-7, 2 × 10⁴ cells/cm²; Caco-2, 1 × 10⁴ cells/cm²; MZPC-1, IMIM-PC-1, and IMIM-PC-2, 5 × 10³ cells/cm²; and MDCK I and MDCK II, 1 × 10³ cells/cm². Culture medium (DMEM supplemented with 10% FBS) was changed daily and cells were maintained at 37C in a 5% CO₂ atmosphere for 2 weeks until a polarized monolayer was achieved, except for HT-29 5M12 cells, which were grown for 21 days to obtain a better-polarized monolayer.

Fluorescence Microscopy

Cells grown on coverslips were fixed with 4% paraformaldehyde for 10 min at room temperature (RT), incubated with 50 mM of NH₄Cl for 30 min, and permeabilized with 0.1% saponin in PBS–1% BSA for 30 min. Primary antibody was incubated for 1 hr followed by FITC-conjugated goat anti-mouse Ig. MAL or SNA (20 μg/ml in 0.1% saponin in 50 mM Tris-HCl, pH 7.5, 15 mM KCl, 5 mM MgCl₂) was added for 1 hr, followed by streptavidin–rhodamine (4 μg/ml). Confocal microscopic analysis was performed using a Leica TCS-D instrument. To determine the degree of colocalization of sialic acid-specific lectins and the basolateral marker, >15 random vertical sections of cells were captured at ×630 magnification and co-localization analysis was carried out using Metamorph 4.0 (Universal Imaging; Westchester, PA) software.

Lectin Histochemical Staining

Paraffin blocks from normal and neoplastic human tissues were obtained from the Department of Pathology, Hospital del Mar, Barcelona. Sections were deparaffinized and rehydrated, boiled in 0.1 M sodium citrate, pH 6, for 10 min, and fixed with cold acetone for 20 min. Endogenous peroxidase was blocked with 4% H₂O₂ for 10 min and nonspecific binding sites were blocked with 2% polyvinylpyrrolidone in 50 mM Tris-HCl, pH 7.5, 15 mM KCl, 5 mM MgCl₂. Sections were then incubated with lectins (20 μg/ml in blocking solution) for 1 hr at RT and streptavidin-peroxidase (4 μg/ml) was added for 30 min. Reactions were developed with H₂O₂/diaminobenzidine (DAB). Enzymatic desialylation was performed by incubating tissue sections with neuraminidase from C. perfringens (50 mU/ml in 50 mM citrate, pH 6.0, 0.9% NaCl, 0.1% CaCl₂) for 16 hr at 37C.

Results

Membrane Distribution of Sialic Acid in Cultured Epithelial Cells

We used lectin immunofluorescence to determine the distribution of sialic acid in polarized epithelial cells. The levels and types of sialic acids were analyzed using a panel of cell lines and the lectins MAL and SNA. Representative results obtained with selected lines are shown here. MAbs TS2/16 and A9, detecting β1 and β4 integrins, and clone 36, detecting dog E-cadherin, were used as a control of polarization and accessibility to the basolateral domain. HT-29 and Caco-2 cells were used as reference cells, because it has previously been shown that they express mainly α2,3 and α2,6 sialic acid, respectively, with a predominant apical distribution (Huet et al. 1998; Slimane et al. 2000).

Figure 1 shows vertical sections obtained by confocal microscopy of representative examples of cell lines reactive exclusively, or predominantly, with one of the lectins used. HT-29 5M12, MCF-7, and HRT-18 bound MAL, whereas Caco-2 bound SNA. In all cases, the signal was much stronger in the apical membrane than in the basolateral membrane and there was essentially no co-localization of sialic acid and basolateral integrins. In some cases, especially in MCF-7 cells, cytoplasmic reactivity was also observed. Similar results were obtained when cells were cultured on plastic, glass coverslips, or permeable filters. When cells were fixed with methanol to remove membrane lipids, a predominant apical distribution of sialic acid was also observed, indicating that the above-mentioned findings are not due to the higher levels of glycosphingolipids present in the apical membrane (data not shown). Quantitative analysis of confocal microscopic images showed that, in paraformaldehyde-fixed HT-29 5M12 cells, 79.9% (± 14.2) of the MAL signal was excluded from the signal obtained with the antibody detecting β4 integrin. These results indicate that the selective distribution of sialic acid cannot be accounted for by lack of accessibility of this molecule to the basolateral membrane.

A second category of cells, including the three pancreatic cancer lines examined and MDCK cells, showed labeling with both MAL and SNA. In MZPC-1 cells, MAL bound mainly to the apical membrane, although weak binding to basolateral membrane could also be detected. In these cells, reactivity with SNA was distributed to all membrane domains as well as in the cytoplasm. IMIM-PC-1 cells are a mosaic of polarized and non-polarized cells. The distribution of sialic acid in polarized cells was similar to that observed in MZPC-1 cells, although the SNA signal was fainter (Figure 2). In non-polarized cells, sialic acid showed a uniform distribution along the cell membrane. In MDCK I cells, MAL and SNA signals were located in the apical membrane domain (data not shown). In MDCK II cells, MAL signal was mainly distributed in the apical domain and was not found intracellularly, whereas SNA signal was present in all membrane domains, with basolateral reinforcement, and in the cytoplasm. In a small proportion of MDCK II cells, predominantly apical distribution of SNA reactivity was found. Quantitative analysis of confocal microscopic images revealed that, in these cells, α2, 3 sialic acid and E-cadherin showed an almost mutually exclusive distribution: 15.8% (± 9.6) of MAL signal co-localized with E-cadherin and 18.6% (± 13.4) of E-cadherin signal co-localized with MAL. By contrast, α2, 6 sialic acid and E-cadherin showed much greater overlap in their distribution: 49.4% (± 14.5) of SNA signal co-localized with E-cadherin and 70.2% (± 20.9) of the E-cadherin signal co-labeled with SNA.

In summary, sialic acid was predominantly found in the apical membrane and there appears to be a hierarchy in its distribution. When present in both α2, 3 and α2, 6 linkages, the α2, 3 configuration is preferentially found in the apical domain.

Membrane Distribution of Sialic Acid in Epithelial Tissues

Many studies have previously analyzed the reactivity of sialic acid-reactive lectins with a wide variety of tissues (Roth 1993). We have reassessed such findings using lectin histochemistry in a selected panel of normal tissues, paying particular attention to its subcellular distribution in epithelial cells. Cell types containing mucous droplets were not considered because it was not possible to appropriately assess basolateral membrane reactivity. A remarkable finding was that, in all tissues analyzed, sialic acid showed a predominant apical distribution. MAL labeled the apical membrane of pancreatic and mammary ducts, bile ducts, and Henle's loop epithelium. In hepatocytes, labeling was concentrated in the biliary canaliculi, which correspond to the apical membrane domain in these cells. In kidney proximal and distal tubules, MAL showed a predominant apical reactivity as well as weak cytoplasmic labeling, with only exceptional basolateral reactivity (Figure 3). Similar findings were made with SNA although cytoplasmic reactivity was generally greater than with MAL in all tissues examined (Figure 3).

Figure 1

Sialic acid distributes preferentially to the apical domain of cultured epithelial cells expressing sialic acid predominantly in either α2, 3 or α2, 6 linkage. Cells were grown on coverslips until polarized, fixed with 4% paraformaldehye, and permeabilized with 0.1% saponin in 1% BSA/PBS. Double staining was performed with a sialic acid-specific lectin (rhodamine) and an antibody detecting a basolateral marker (β4 integrin for HT-29 5M12 cells and β1 integrin for MCF-7 and Caco-2 cells) (fluorescein), and was visualized by confocal microscopy. Vertical sections of polarized HT-29 5M12, MCF-7, and Caco-2 cells. The lectin signal is strong in the apical domain and weak in the basolateral membrane. In MCF-7 cells, cytoplasmic signal is also observed. Bars = 5 μm.

Figure 2

Distribution of sialic acid in cultured epithelial cells expressing sialic acid in both α2,3 and α2,6 linkages. Cells were processed as indicated in Figure 1, double stained with sialic acid-specific lectins and an antibody detecting a basolateral marker (β1 integrin for pancreas cancer cells and E-cadherin for MDCK II cells), and visualized by confocal microscopy. Vertical sections of polarized MZPC-1, IMIM-PC-1, and MDCK II cells. Note the preferential apical distribution of MAL signal in the three lines and the overlap between SNA signal and the basolateral marker. In some cases, intracellular binding of lectin is also observed. Bars = 5 μm.

GalNAc-O-benzyl Induces a Redistribution of Sialic Acid in IMIM-PC-1 Cells

As shown above, IMIM-PC-1 cells express both α2,3 and α2,6 sialic acid. GalNAc-O-benzyl has been shown to decrease sialylation in a few cell types (Byrd et al. 1995; Huet et al. 1995, 1998) due to inhibition of α2,3-sialyltransferase activity. IMIM-PC-1 cells express high levels of ST3Gal I mRNA and ST3 Gal activity (data not shown). Cells were treated for 15 days with 2 mM GalNAc-O-benzyl and the distribution of sialic acid was analyzed by confocal microscopy. In control cultures, >90% of cells showed strong apical MAL reactivity; by contrast, <5% of cells were reactive in treated cells. In control cultures, approximately 35% of cells were reactive with SNA and the signal was found in the apical and basolateral membrane as well as intracellularly. By contrast, in treated cells, SNA-binding sites showed a predominant apical distribution in 60% of cells (Figure 4). These findings indicate that inhibition of α2,3 sialylation is accompanied by an increase in the levels of and a redistribution of α2,6 sialic acid.

Discussion

In this study we have analyzed by lectin staining and confocal microscopy the distribution of sialic acid in the membrane of polarized cultures of epithelial cells and in normal epithelial tissues. The three main findings of the study are as follows: (a) sialic acid showed a predominant apical membrane distribution in all cultured cells examined, regardless of their origin; (b) when sialic acid was present both in α2,3 and α2,6 linkages, molecules with an α2,3 linkage were preferentially distributed to the apical membrane; and (c) inhibition of α2,3 sialylation induced by treatment of IMIM-PC-1 pancreatic cancer cells with GalNAc-O-benzyl was associated with an increase in the levels and the relative apical distribution of α2,6 sialic acid.

The predominantly apical distribution of sialic acid supports the idea that this molecule plays a role in the interaction between the host and a wide variety of microorganisms in the lumen of epithelia. Many pathogenic bacteria and viruses produce neuraminidases, and it has been proposed that the genes coding for them have their origin in higher animals (Roggentin et al. 1993; Traving and Schauer 1998). Sialic acid can also be used as a co-receptor for binding and infectivity of some pathogenic viruses (Varki 1993; Barton et al. in press). In addition, a trans-sialidase is released by Trypanosoma cruzi and can induce apoptosis in lymphocytes (Leguizamon et al. 1999). Sialic acid at the luminal surface of epithelial cells may facilitate loss of tissue integrity, invasion, and/or colonization. It has been proposed that glycoconjugate structure may have evolved to elude invasion by pathogens, with sialic acid acquisition and modification (i.e., 9-O-acetylation) being part of such a process (Varki 1993).

Sialic acid has also been proposed to play an important role in cell–cell interactions. It is expressed at the luminal surface of endothelial cells, and binding of selectins to sialyl-Le^a or sialyl-Le^x plays a crucial role in the extravasation of leukocytes at inflammatory sites (Tedder et al. 1995). It has also been proposed that sialylated ligands on tumor cells may contribute to the transendothelial migration occurring in the metastatic process (Kannagi 1997). Increased levels of sialic acid commonly found in tumors, and possibly selected during clonal evolution, and the loss of polarity found in many neoplastic cells may contribute to enhance such interactions.

In the past few years, a growing number of membrane-bound mucins expressed by epithelial (i.e., MUC1, MUC3, MUC4), hematopoietic (i.e., CD43), and/or endothelial cells (i.e., Gly-CAM, MadCAM) have been identified (reviewed in Varki 1997; Fukuda and Tsuboi 1999; Carraway et al. 2000). These molecules contain Ser- and Thr-rich domains that are extensively O-glycosylated and sialylated, and it has been proposed that they play an important role in cell–cell interactions. CD43 is a cell surface glycoprotein expressed by hematopoietic cells. Its sialylation plays an essential role in disminishing the susceptibility of target cells to T-lymphocyte-mediated cytolysis (Manjunath et al. 1995; McFarland et al. 1995). T3AH rat mammary carcinoma cells express high levels of the sialoglycoprotein epiglycanin. On deglycosylation, cell–cell and cell–matrix adhesion is stimulated, and epiglycanin sialylation can overcome E-cadherinmediated homotypic adhesion (Kemperman et al. 1994). Similarly, ectopic MUC1 expression in melanoma cells induced an inhibition of cell aggregation which was due, in part, to sialylation (Ligtenberg et al. 1992), and MUC1 interferes with the binding of non-polarized cells to the matrix mediated by integrins (Wesseling et al. 1995).

Figure 3

Lectin staining of sections of normal human epithelial tissues. Sections of paraffin-embedded tissues were allowed to react with biotinylated MAL or SNA and the reactions were developed with streptavidin–peroxidase and DAB; sections were counterstained with hematoxylin. Note the predominant apical distribution of the lectin signal in pancreatic and mammary gland ducts, biliary epithelium, biliary canaliculi, and renal epithelial cells (arrows).

Figure 4

GalNAc-O-benzyl induces a redistribution of sialic acid in IMIM-PC-1 cells. In control cells, MAL labels the apical membrane, whereas SNA does not show a preferential apical labeling. In GalNAc-O-benzyl-treated cells, MAL reactivity is markedly decreased, whereas SNA shows increased labeling and α2,6 sialic acid is detected in the apical membrane of polarized cells. Arrowheads point to the surface of the coverslip, identified by reflection. Bars = 5 μm.

Regarding the mechanisms leading to the preferential distribution of α2,3 vs α2,6 sialic acid in the apical membrane, several hypotheses can be postulated, although as yet there is no substantial evidence favoring any of them. First, it is possible that the enzymes involved in the synthesis of both types of linkages reside in different compartments of the secretory pathway. Despite the fact that little is known about the specific subcellular distribution of sialyltransferases and that cell-to-cell variability is likely to occur (Velasco et al. 1993), ST3Gal III and ST6Gal I have been shown to co-localize in the Golgi yet to be differentially distributed in a post-Golgi compartment (Burger et al. 1998). Alternatively, a sorting mechanism mediated by a lectin may be responsible for the preferential distribution of α2,3 to the apical domain. Such putative lectins might selectively recruit sialic acid-containing glycoproteins to the apical vs the basolateral domain or might favor the transport of sialylated vs unsialylated cargo. It should be emphasized that the findings described here probably reflect quantitative rather than qualitative differences and that the presence of sialic acid in basolateral glycoproteins has been demonstrated in the past. The existence of saturable glycanindependent mechanisms for apical delivery has recently been proposed (Marmorstein et al. 2000).

With regard to the mechanisms operating in glycoprotein selection and sorting in cultured cells, two points should be made: the fact that we used tumor cells and the possibility that glycoprotein sorting mechanisms might be altered in cultured cells. Against the possibility that our findings are artifactual is the observation that a similar degree of selectivity in sialic acid distribution was present in normal tissues. Although the data reported here on the expression of sialic acids in tissues are not novel, they substantiate available evidence on the distribution of sialic acids in the apical membrane using lectin–ultrastructural techniques (selected references: Sata et al. 1989, 1991; Roth 1993; Kaneko et al. 1995; Babal and Gardner 1996). For example, it has been reported that in human colon, using MAL lectin–ultrastructural cytochemistry, the “luminal surface of absorptive enterocytes was intensely stained along the entire length of the crypts and in the surface epithelium of both right and left colon” (Sata et al. 1991). In rat intestine, sialic acid is present in both apical and basolateral membranes of undifferentiated cells, whereas it is restricted to the apical domain in the surface epithelium (Taatjes and Roth 1988). We have failed to find reports of an exclusive or preferential basolateral distribution of sialic acid (Roth 1993).

We have examined the reactivity of MAL and SNA with a small panel of tumor tissues from prostate, colon, and pancreas. In areas where cells were polarized, predominant apical distribution of sialic acid was also observed (not shown), as has been described by other investigators. Interestingly, in prostate cancers α2,3 sialic acid showed mainly a supranuclear distribution, whereas α2,6 sialic acid was found in the apical membrane, supporting the notion that the differential subcellular distribution of sialic acids is tightly regulated.

In HT-29 cells, we have previously described that inhibition of α2,3 sialylation associated with culture in the presence of GalNAc-O-benzyl is accompanied by concomitant upregulation (threefold) of the levels of ST3 Gal I mRNA and activity (Huet et al. 1998). These findings strongly support the notion that sialylation levels are finely tuned in cells, suggesting the existence of sensor mechanisms that monitor the degree, and perhaps the type, of sialylation of membrane glycoproteins. It is tempting to speculate that lectin-like molecules may participate in this process. In the past few years, several intracellular lectins have been implicated in the secretory pathway (Fiedler and Simons 1994, 1996), although none of them appears to recognize sialic acids. A family of sialic acid-binding cell surface lectins (siglecs) has recently been described, but its members appear to participate mainly in cell–cell interactions (Crocker et al. 1998). If intracellular lectins exist and play a role as sensors, the next question will be how this information translates into changes in steady-state mRNA levels. Two possibilities are the regulation of mRNA stability and transcriptional regulation itself, although neither of these processes is known to be regulated by sialylated proteins. Until now, this concept has not been discussed at any length.

Carbohydrates in glycoproteins and glycolipids display a highly remarkable complexity that does not have a parallel in terms of functional significance (see discussion by Varki 1993). Recent data indicate that Fringe is a fucosyltransferase that modulates the ability of Notch ligands to activate the Notch signaling pathway (Moloney et al. 2000). It is therefore possible that new roles for other sugars, such as sialic acid, will be described in the future. The generation of mice deficient in sialyltransferases should also yield information on their function (Muramatsu 1999; Priatel et al. 2000). The findings described here provide clues regarding the regulation of sialic acid level and distribution in epithelial cells.