Transcriptional Regulation of 1,3-Galactosyltransferase in Embryonal Carcinoma Cells by Retinoic Acid MASKING OF LEWIS X ANTIGENS BY α-GALACTOSYLATION

Treatment of mouse teratocarcinoma F9 cells with all-trans-retinoic acid (RA) causes a 9-fold increase in steady-state levels of mRNA for UDP-Gal:β-D-Gal α1,3-galactosyltransferase (α1,3GT) beginning at 36 h. Enzyme activity rises in a similar fashion, which also parallels the induction of laminin and type IV collagen. Nuclear run-on assays indicate that this increase in α1,3GT in RA-treated F9 cells, like that of type IV collagen, is transcriptionally regulated. Differentiation also results in increased secretion of soluble α1,3GT activity into the growth media. The major α-galactosylated glycoprotein present in the media of RA-treated F9 cells, but not of untreated cells, was identified as laminin. Differentiation of F9 cells is accompanied by an increase in α-galactosylation of membrane glycoproteins and a decrease in expression of the stage-specific embryonic antigen, SSEA-1 (also known as the Lewis X antigen or Le), which has the structure Galβ1-4(Fucα1-3)GlcNAcβ1-R. However, flow cytometric analyses with specific antibodies and lectins, following treatment of cells with α-galactosidase, demonstrate that differentiated cells contain Le antigens that are masked by α-galactosylation. Thus, RA induces α1,3GT at the transcriptional level, resulting in major alterations in the surface phenotype of the cells and masking of Le antigens.


Treatment of mouse teratocarcinoma F9 cells with all-
trans-retinoic acid (RA) causes a 9-fold increase in steady-state levels of mRNA for UDP-Gal:␤-D-Gal ␣1,3galactosyltransferase (␣1,3GT) beginning at 36 h. Enzyme activity rises in a similar fashion, which also parallels the induction of laminin and type IV collagen. Nuclear run-on assays indicate that this increase in ␣1,3GT in RA-treated F9 cells, like that of type IV collagen, is transcriptionally regulated. Differentiation also results in increased secretion of soluble ␣1,3GT activity into the growth media. The major ␣-galactosylated glycoprotein present in the media of RA-treated F9 cells, but not of untreated cells, was identified as laminin. Differentiation of F9 cells is accompanied by an increase in ␣-galactosylation of membrane glycoproteins and a decrease in expression of the stage-specific embryonic antigen, SSEA-1 (also known as the Lewis X antigen or Le X ), which has the structure Gal␤1-4(Fuc␣1-3)Glc-NAc␤1-R. However, flow cytometric analyses with specific antibodies and lectins, following treatment of cells with ␣-galactosidase, demonstrate that differentiated cells contain Le X antigens that are masked by ␣-galactosylation. Thus, RA induces ␣1,3GT at the transcriptional level, resulting in major alterations in the surface phenotype of the cells and masking of Le X antigens.
Retinoids, which are natural and synthetic analogs of vitamin A, alter in vitro growth and differentiation of a variety of normal and neoplastic cell lines (1)(2)(3)(4)(5). Profound changes in gene expression and pattern formation during embryogenesis are regulated by retinoic acid (RA) 1 via its interactions with retinoic acid binding proteins and transcription factors (6 -10). A model system for studying such changes related to cellular differentiation and development has been the mouse teratocarcinoma cell line F9 (11). These cells exhibit characteristics of embryonic cells from the inner cell mass of the blastula stage. When F9 cells are treated with all-trans-retinoic acid, within 3 days they differentiate into primitive endoderm-like cells, which begin to synthesize basement membrane proteins, including type IV collagen and laminin (11)(12)(13)(14)(15)(16).
F9 cells contain many surface carbohydrate antigens that are shared with those in early mouse embryos (17)(18)(19)(20)(21)(22). Surface glycoconjugates in F9 cells contain the repeating disaccharide [3Gal␤1-4GlcNAc␤1] n or poly-N-acetyllactosamine sequence (23)(24)(25) and the stage-specific embryonic antigen SSEA-1 (also known as the Lewis X antigen or Le X ), which has the structure Gal␤1-4(Fuc␣1-3)GlcNAc␤1-R (17,26). Differentiation of F9 cells induced by RA causes changes in expression of surface glycoconjugates, and within 3 days following RA treatment there is a marked decrease in expression of the Le X antigen (14,19,27). However, the mechanism by which RA/F9 cells express less Le X antigen is unclear. Although it has been shown that 5 days following RA treatment of F9 cells there is an 80% decrease in the level of an ␣1,3-fucosyltransferase activity that can synthesize the Le X antigen (28), there is little change in this enzyme activity following 3 days of RA treatment (29). These changes in Le X expression in differentiated F9 cells are relevant to events occurring during early embryogenesis, since there are notable changes in expression of Le X antigens during normal differentiation of mouse embryos, and some studies have suggested that Le X antigens play important roles in cellcell adhesion and development (30 -33).
Cell extracts from both F9 and RA-treated F9 cells (RA/F9 cells) contain a UDP-Gal:␤-D-Gal ␣1,3-galactosyltransferase (␣1,3GT) activity that synthesizes the terminal sequence Gal␣1-3Gal␤-R (29,34). The presence of this enzyme is consistent with studies suggesting that glycoproteins from F9 and RA/F9 cells contain terminal ␣-galactosyl residues (29,35,36). However, the activity of the ␣1,3GT is higher in RA/F9 cells, and there is an apparent increase in ␣-galactosylation of surface glycoconjugates induced by RA (29). Little information is available about the induction of the ␣1,3GT by RA, the glycosylation of proteins accompanying this induction, and whether there is a relationship between changes in the expression of terminal ␣-galactosyl residues on cellular glycoproteins and the apparent decrease in expression of Le X antigens.
We have now examined the steady-state levels of the ␣1,3GT transcript during differentiation by RA and the enzymatic activities in both cell extracts and cell culture media following RA treatment. Our findings demonstrate that (i) ␣1,3GT is transcriptionally regulated by RA treatment; (ii) differentiation of F9 cells results in increased activity of the enzyme and increased expression of terminal ␣-galactosyl residues on selected cellular glycoproteins; and (iii) the apparent loss of the Le X antigens on cell surface glycoconjugates in RA/F9 cells is in part due to masking by terminal ␣-galactosyl residues.
Cell Culture-Mouse teratocarcinoma F9 cells were cultured in Dulbecco's modified Eagle's medium containing 15% fetal calf serum on gelatin-coated tissue culture plates as described (12). For induction of differentiation, cells were grown under identical conditions in media supplemented with either 10 Ϫ7 M RA alone or 10 Ϫ7 M RA plus 10 Ϫ3 M dibutyryl cyclic AMP (dbcAMP) for 3 days. Differentiation was assessed by measuring the amount of laminin secreted in the cell culture media using a solid-phase assay described below. Chinese hamster ovary (CHO) cells were cultured in ␣-minimal essential media containing 10% fetal calf serum.
RNA Blot Analysis-Total cellular RNA was isolated from cultured cells essentially as described (38). For Northern blot analysis, RNA samples were transferred to nylon membranes (Schleicher & Schuell) by capillary action in 20 ϫ SSPE for 16 h and immobilized by UV cross-linking (Autocrosslink on Stratalinker 2400). The filters were prehybridized for at least 4 h in a solution containing 50% formamide, 5 ϫ SSPE (0.18 M NaCl, 10 mM sodium phosphate, pH 7.7, 1 mM EDTA), phycoerythrin (250 mM Tris, pH 7.5, 25 mM EDTA, 0.5% sodium pyrophosphate, 5% SDS, 1% polyvinylpyrrolidone, 1% Ficoll 400, and 1% bovine serum albumin), and 150 g/ml denatured salmon sperm DNA. RNA bound to nylon membrane was hybridized with 32 P-labeled cDNA inserts from murine ␣1,3GT clone PCDM7-␣GT (34), murine ␣-actin cDNA, and rat tubulin cDNA (a gift from Dr. Kelley Moremen, University of Georgia). The RNA blots were hybridized separately with the individual probes. Hybridization was carried out at 42°C overnight in the prehybridization solution using a hybridization incubator (Robbins Scientific, Sunnyvale, CA). The filters were washed twice in 6 ϫ SSPE, 0.5% SDS at room temperature for 15 min, twice in 1 ϫ SSPE, 0.1% SDS at 42°C for 15 min, and three times in 0.1 ϫ SSPE, 0.1% SDS at 65°C for 15 min. Hybridization was visualized by autoradiography, using Kodak XAR-5 film (Kodak). After development, exposed areas of film were quantified by densitometry (Molecular Dynamics model 3008 computing densitometer). The signals were normalized for variations in the amount and quality of RNA using actin mRNA or tubulin mRNA as internal controls.
Nuclear Run-on Assays-Nuclei were isolated from F9 or RA/F9 cells as described (39) and run-on transcription reactions were carried out as described (40) but with the following modifications. Transcription was initiated by incubating 0.2 ml of nuclei (2-3 ϫ 10 7 nuclei) with 12.5 l of [␣-32 P]UTP (800 Ci/mmol, 10 Ci/l) in a 0.4-ml reaction volume containing 126 l of reaction buffer (10 mM Tris-Cl, pH 7.5, containing 35% glycerol, 5 mM MgCl 2 , and 300 mM KCl), 20 l of 10 mM GTP, CTP, ATP, respectively, 1 l of 0.5 M dithiothreitol, and 80 units of RNase inhibitor. After mixing the components, the reactions were incubated at 30°C for 40 min with shaking. RNase-free DNase I (50 units) was added with 22.5 l of 20 mM CaCl 2 . After 30 min incubation at 37°C, 5 l of 16 mg/ml proteinase K and 47.25 l of proteinase K buffer (5% SDS, 50 mM EDTA, 100 mM Tris, pH 7.4) was added along with the 10 l of 10 mg/ml yeast tRNA, and the samples were incubated at 37°C for 30 min. The 32 P-labeled transcripts were resuspended in 0.5% SDS and hybridized onto DNA fragments immobilized on nylon membranes at 65°C for 36 h.
Reaction mixtures for the ␣1,3GT routinely contained 30 mM Nacetyllactosamine, 20 mM MnCl 2 , 0.5 mM UDP-[ 3 H]Gal (35,000 cpm/ nmol), 5 mM ATP, 50 mM D-galactono-1,4-lactone, and about 80 g of cell extracts or 12.5 l of 10-fold concentrated cell culture media in a final volume of 25 l. The reactions were carried out at 37°C for 3 h. The product of the enzyme reaction was analyzed by Dionex high performance anion-exchange chromatography using a Carbopac PA1 column (4 ϫ 250 mm). The separation was carried out in 160 mM sodium hydroxide for 15 min with a flow rate of 1 ml/min. The fractions were collected and monitored by scintillation counting. The radioactivity corresponding to the position of standard Gal␣1-3Gal␤1-4GlcNAc was taken as the product of ␣1,3GT. The values obtained from assays performed in the absence of the acceptor were used as background. Assays for UDP-Gal:GlcNAc ␤1,4-galactosyltransferase (␤1,4GT) were performed as described for ␣1,3GT, except that 20 mM GlcNAc was used as an acceptor, and cell culture media centrifuged at 100,000 ϫ g for 1 h were used as an enzyme source without concentration. The product N-acetyllactosamine was identified by high performance anion-exchange chromatography, as described above.
Detection of Laminin by Enzyme-linked Immunosorbent Assay-The wells of Immulon 4 microtiter plates (Dynatech Laboratories, Chantilly, VA) were coated at 4°C overnight with 100 l of either F9 or RA/F9 cell culture media or 100 l of varying concentrations of mouse EHS laminin in carbonate coating buffer (150 mM Na 2 CO 3 , 348 mM NaHCO 3 , and 0.02% NaN 3 , pH 9.6). After blocking with 5% BSA in PBS/NaN 3 (PBS containing 0.02% NaN 3 ), laminin was detected by incubating the wells with rabbit anti-mouse EHS laminin (100 l of 1:2,000 dilution) followed by alkaline phosphatase-conjugated goat anti-rabbit IgG (100 l of 1:4,000 dilution). Antibody solutions were prepared in dilution buffer (PBS/NaN 3 containing 0.1% BSA and 0.05% Tween 20), and incubation was performed at room temperature for 1 h. The wells were treated with 100 l of a freshly prepared 1 mg/ml alkaline phosphatase substrate p-nitrophenylphosphate in coating buffer containing 1 mM MgCl 2 at 37°C. The optical density at 405 nm of each well was recorded by V-Max kinetic microplate reader (Molecular Devices Corp., Menlo Park, CA). Each assay was performed in triplicate.
Lectin Blotting-GS I-B 4 was biotinylated with Sulfo-NHS biotin according to the manufacturer's instructions. Microsomes were prepared as described (41) in the presence of protease inhibitors, phenylmethylsulfonyl fluoride (1 mM), pepstatin (1 g/ml), aprotinin (10 g/ ml), and leupeptin (10 g/ml). Streptavidin-agarose was added to the microsomes to remove cellular proteins that nonspecifically interact with streptavidin, and the supernatant fraction was recovered following centrifugation at 5,000 ϫ g for 15 s. Total protein was determined by the BCA protein assay. For the preparation of serum-free cell culture media, F9 and RA/F9 cells were grown in Dulbecco's modified Eagle's medium containing 15% fetal calf serum for 3 days, as described above, and the culture media were replaced with serum-free media. Media were collected after 1 day and centrifuged at 100,000 ϫ g for 60 min. The supernatant (soluble proteins) was concentrated about 10-fold using a centriprep-10 concentrator. The microsomal fraction and soluble proteins in culture media were subjected to electrophoresis on 10% SDS/polyacrylamide gels (42). Proteins were transferred electrophoretically onto nitrocellulose membranes (Hybond-ECL Western, Amersham) (43). Blots were blocked with TBS (20 mM Tris, 500 mM NaCl, pH 7.5), 5% BSA at room temperature for 3 h, then incubated with 10 g/ml of biotinylated GS I-B 4 in dilution buffer (TBS, 0.1% BSA, 0.3% Tween 20) for 1 h. Control lanes were incubated with a lectin solution containing 200 mM hapten sugar raffinose. Following three 15-min washes in TTBS (TBS, 0.3% Tween 20), the lectin blots were incubated at room temperature for 1 h with horseradish peroxidase-conjugated streptavidin diluted 1:5,000 in dilution buffer. Blots were washed three times (15 min each) in TTBS followed by one wash in TBS and developed using the enhanced chemiluminescence (ECL) Western blotting kit following the manufacturer's instructions. Molecular weight determinations were made using prestained high molecular mass standards (14.3-200 kDa, Life Technologies, Inc.).
Immunoprecipitation of Laminin in RA/F9 Culture Media-Serumfree RA/F9 cell culture media were prepared, as described above, and 84 l from each of these (corresponding to 400 g of soluble protein) was incubated with 50 l of protein A-Sepharose at 4°C for 1 h to remove components that react nonspecifically with protein A. The supernatant was recovered by centrifugation at 10,000 ϫ g for 15 min and incubated with 5 l of rabbit anti-mouse laminin antibody at 4°C for 2 h. The mixture was subsequently incubated with 50 l of protein A-Sepharose for 1 h and centrifuged to pellet immunoprecipitated complexes. The laminin-antibody complexes immobilized on the protein A-Sepharose were washed five times with washing buffer (1% BSA, 0.05% Tween 20, 1% SDS in PBS) and finally boiled in Laemmli sample buffer (42) for 5 min to dissociate the immune complexes. The samples were subjected to electrophoresis on 6% SDS-PAGE followed by GS I-B 4 lectin blotting, as described above.
Immunoblotting of Laminin in RA/F9 Media-Mouse EHS laminin and RA/F9 serum-free culture media were separated by SDS-PAGE on 6% polyacrylamide gel under reducing conditions and transferred electrophoretically onto nitrocellulose sheets as described above. Immunoblotting was performed as described for the GS I-B 4 blotting, except that rabbit anti-mouse laminin antibody (1:2,000) and anti-rabbit-horseradish peroxidase conjugate (1:5,000) were used as first and second antibodies, respectively.
Flow Cytometry Analysis-Cells were harvested and washed once with Hanks' balanced salt solution. Approximately 5 ϫ 10 5 cells were incubated with 100 milliunits of green coffee bean ␣-galactosidase in HEPES buffer (10 mM HEPES, pH 6.5, 0.15 M NaCl, 5 mM CaCl 2 ) at 37°C for 1 h. Enzymes boiled for 20 min were used as controls. The cells were washed four times with staining buffer (Hanks' balanced salt solution, 1% heat inactivated goat serum) and incubated with 20 l of fluorescein isothiocyanate-conjugated anti-Le X monoclonal antibody without dilution or 50 l of biotinylated GS I-B 4 (22 g/ml), followed by incubation with 8 l of phycoerythrin-streptavidin (1 mg/ml) in 100 l of staining buffer on ice for 30 min, respectively. Flow cytometry was performed with a FACStar-Plus flow cytometer (Becton Dickinson Immunocytometry, San Jose, CA).

␣1,3GT mRNA Is Increased upon RA Treatment of F9
Cells-To address the mechanism(s) underlying the apparent induction of ␣1,3GT in F9 cells, we first examined the effect of 72 h of RA treatment on the induction of ␣1,3GT transcripts. A representative time course of the Northern blot analysis using cells treated with RA is shown in Fig. 1A. The steady-state levels of ␣1,3GT mRNA begin to increase at 36 h following RA treatment and reached a maximum level at 48 h. This result was reproducible in four separate experiments, and the average of all the results obtained from densitometric scans is shown in Fig. 1B. The results demonstrate that RA treatment causes a 9-fold induction of the ␣1,3GT mRNA within 48 h. As a control, we analyzed total RNA from CHO cells and COS7 cells. These cells lack ␣-galactosyltransferase activity (34,45) and lack any detectable transcript for the ␣1,3GT (Fig. 1A).
It has been reported that dbcAMP enhances the transcriptional activation of genes by RA (44). Therefore, we tested the combined effects of RA and dbcAMP on the changes induced in the transcript for the ␣1,3GT. About 15 g of total cellular RNA from untreated and treated cells was analyzed by Northern blot analysis using a 32 P-radiolabeled murine ␣1,3GT cDNA in pCDM7. The results demonstrate that after 3 days of treatment the inclusion of dbcAMP and RA only slightly enhances the level of transcript over that of RA alone (Fig. 2, A and B). Since the levels of ␣1,3GT mRNA message of RA/F9 and that of RA/dbcAMP/F9 are rather similar, we focused on the effect of RA alone in all other experiments.
RA Increases the Transcriptional Rate of ␣1,3GT-Nuclear run-on assays were conducted to address the question of whether the change in ␣1,3GT mRNA induced by RA reflects an effect at the transcriptional level. For these nuclear run-on assays, we also analyzed expression of type IV collagen (␣1), which is known to have a higher rate of transcription in RA/F9 cells compared to F9 cells (46). RNA "run-on" transcripts were synthesized by incubating the nuclei isolated from both F9 and RA/F9 cells in the presence of [␣-32 p]UTP, and similar amounts of radiolabeled RNA (ϳ2 ϫ 10 6 cpm/ml) were then hybridized to DNA fragments immobilized on nylon filters. The transcriptional rate of the type IV collagen (␣1) gene increases during F9 cell differentiation (Fig. 3), which is consistent with earlier reports (39,46). Likewise, the transcriptional rate of the ␣1,3GT gene is much higher in RA/F9 cells than in F9 cells (Fig. 3). As controls, there was no significant change in run-on transcripts for actin, and no transcripts were detected for the control pBluescript vector DNA. These results demonstrate that elevated transcripts for the ␣1,3GT in RA/F9 cells results from an increase in the transcriptional rate for the gene.
Time Course of Induction of ␣1,3GT Enzyme Activity in F9 Cell Extracts and Cell Culture Media Following RA Treatment-To understand the relationship between ␣1,3GT mRNA level and enzymatic activity during differentiation, the ␣1,3GT activity was determined in total cell extracts and concentrated cell culture media at varying times after RA treatment. The specific activity of ␣1,3GT in cell extracts starts to increase at 36 h after RA treatment, and the specific activity in 72-h RA/F9 cell extracts is approximately 5-fold higher than that in F9 extracts (Fig. 4A). This agrees with the previous findings when N-acetyllactosamine was used as an acceptor (29).
In preliminary experiments, we found a considerable amount of activity of the ␣1,3GT in the culture media from RA/F9 cells. We therefore conducted a careful study of the enzymatic activity present in the culture media at various times during differentiation. The total enzyme activity was normalized to the FIG. 1. Kinetics of ␣1,3GT mRNA induction in F9 cells by RA. A, total RNA was isolated from F9 cells (0 h) and F9 cells exposed to RA for increasing periods of time. 40 g of total RNA from each time point was isolated from the cells, electrophoretically separated on a 1% agarose gel containing 7% formaldehyde, and transferred to nylon filters. The immobilized RNAs were then hybridized with a 32 P-labeled ␣1,3GT cDNA probe, followed by stripping and reprobing with a 32 P-labeled mouse actin cDNA probe as a control. As negative controls, RNAs isolated from CHO and COS7 cells were included. B, the magnitude of expression of ␣1,3GT was quantified by densitometry, and the relative signal was normalized to the level of actin message. The levels of expression of the ␣1,3GT transcript in F9 cells were arbitrarily taken as 1 unit. amount of total cellular protein on the dish. The ␣1,3GT activity in the cell culture media begins to rise after 36 h of RA treatment and is approximately 4-fold higher after 72 h of RA treatment (Fig. 4B). It can be seen from the results in Fig. 4B that approximately 2 ⁄3 of the total ␣1,3GT enzyme activity detectable in the cultures of the differentiated F9 cells is present in the culture media, and the remaining 1 ⁄3 is recoverable in cell extracts. These results demonstrate that the ␣1,3GT is efficiently secreted by cells and that a majority of the enzyme activity detected is present in a soluble form. The increase in ␣1,3GT activity in media following RA addition appears to rise with a time course slightly lagging behind the rise observed in the cell-associated enzyme activity (Fig. 4, A and B).
The ␣1,3GT activity in culture media arises by secretion from cells, since the growth media prior to addition to the cultured cells has no detectable activity (data not shown). In addition, the ␣1,3GT activity recoverable in the media of RA/F9 cells cannot be pelleted by prolonged centrifugation at 100,000 ϫ g, although the cell-associated activity is clearly microsomal (data not shown). This indicates that the enzyme in culture media is soluble and is an active and probably a proteolytically cleaved form, as has been observed for some other glycosyltransferases (47)(48)(49)(50)(51)(52).
To assess whether the increase in ␣1,3GT in F9 cells in response to RA treatment is specific for this enzyme, we also measured the activity of ␤1,4GT. It has been reported that ␤1,4GT activity decreases to approximately one-third of control values during the first 3 days of differentiation of F9 cells following RA addition and then begins to rise slowly on day 4 following RA treatment (53). Consistent with this, we found that the activity of the ␤1,4GT in extracts of both F9 and RA/F9 cells is similar during the first 3 days following RA treatment (data not shown). Our results are also consistent with another recent study showing that there is little change in either the ␤1,4GT activity or transcript levels within 3 days of treatment of F9 cells with RA (54). The ␤1,4GT activity is detectable in the cell culture media of both cell types, but RA treatment causes only a small change (Ͻ2-fold increase) in ␤1,4GT activity in media after 72 h (data not shown). These results demonstrate that RA treatment of F9 cells does not cause a general rise in the activities of all galactosyltransferases but results in a pronounced change in the expression and activity of the ␣1,3GT.
Laminin Is Secreted into Culture Media Following RA Treat- F9 cells were exposed to RA for increasing periods of time. A, approximately 80 g of cell extracts was assayed for ␣1,3GT activity by incubation at 37°C for 3 h using N-acetyllactosamine as an acceptor. The product was isolated as described under "Experimental Procedures." B, The culture media were concentrated 10-fold, and 12.5 l portions were assayed for ␣1,3GT activity. ment of F9 Cells-To determine the temporal relationship between induction of the ␣1,3GT compared by RA to other known markers and to assess the efficiency of differentiation of F9 cells in response to 10 Ϫ7 M RA in our system, we determined the amount of laminin secreted into the cell culture media. It has been shown that the steady-state levels of mRNAs for laminin B1 and B2 increase dramatically following differentiation of F9 cells, and laminin has been defined as a differentiation marker, along with other basement membrane components, such as type IV collagen (␣1) (39,(55)(56)(57)(58)(59)(60). Laminin was detected by an enzyme-linked immunosorbent assay using cell culture media immobilized in microtiter wells and anti-laminin antibody as the detecting reagent. Laminin levels in the media of F9 cells treated with RA starts to increase at 36 h after seeding the cells and adding RA and continues to increase during the entire 72-h period of RA treatment (Fig. 5). In contrast, there is only a modest amount of immunoreactive material in media from F9 cells not treated with RA. These studies show that induction of the ␣1,3GT parallels that of laminin in response to 10 Ϫ7 M RA.
Soluble and Membrane-associated Glycoproteins Express Increased Amounts of ␣-Galactosyl Residues After RA Treatment-To determine the effects of induced ␣1,3GT expression on glycosylation of proteins in the cells, we utilized the plant lectin, GS I-B 4 , which binds specifically to terminal ␣-linked galactosyl residues (61). Previous studies indicated that RA/F9 cells have more binding sites for GS I-B 4 than F9 cells (29), but the nature of the bound material was not studied. Microsomes from both F9 and RA/F9 cells were analyzed by SDS-PAGE on 10% polyacrylamide gel and blotted with GS I-B 4 . There is a large increase in the amount of GS I-B 4 -reactive glycoproteins in RA/F9 cells compared to F9 cells, especially in the high molecular mass (Ͼ200 kDa) range and the 68 -200 kDa range (Fig. 6A). The binding of GS I-B 4 to glycoproteins is specific, since inclusion of the hapten sugar raffinose (200 mM) reduces binding (Fig. 6A). In other control experiments, we found no specific binding to microsomal glycoproteins from CHO cells, whereas there was significant specific binding of the lectin to a commercial preparation of EHS laminin, as expected and as discussed below (Fig. 6A). These results confirm that ␣-galactosylation of glycoproteins is dramatically increased upon differentiation of F9 cells. This evidence, coupled with flow cyto- The blots were developed using the ECL chemiluminescence Western blotting kit. Mouse EHS laminin was included as a positive control. B, serum-free culture media from each cell line was also collected, and approximately 40 g of soluble protein was blotted with biotinylated GS I-B 4 . C, RA/F9 media were prepared as described under "Experimental Procedures." Following immunoprecipitation with anti-laminin antibody, bound proteins were subjected to SDS-PAGE on a 6% polyacrylamide gel and blotted with biotinylated GS I-B 4 . D, RA/F9 serum-free culture media and mouse EHS laminin were analyzed by SDS-PAGE and immunoblotted with rabbit anti-mouse laminin antibody followed by goat anti-rabbit IgG-peroxidase conjugate. metric data described below, demonstrate that increased binding of GS I-B 4 may be considered as a marker for F9 cell differentiation induced by RA.
Differentiation of F9 cells induced by RA is also accompanied by a large increase in the terminal ␣-galactosylation of high molecular weight soluble glycoproteins secreted into the cell media (Fig. 6B). A significant band of the high molecular weight soluble glycoprotein in the range of 200 -400 kDa reactive with GS I-B 4 is detectable in RA/F9 serum-free cell culture media, whereas very little material in F9 serum-free cell culture media is reactive with GS I-B 4 . The molecular weight of this GS I-B 4 -reactive material in the growth media of RA/F9 cells suggested that it might be laminin. To confirm the identity of this material, the RA/F9 culture media were immunoprecipitated with anti-mouse laminin antibody, analyzed by SDS-PAGE on 6% acrylamide gel, and then blotted with GS I-B 4 in the presence or absence of 200 mM raffinose. The antimouse laminin antibody immunoprecipitates the major ␣-galactosylated components from RA/F9 cell culture media (Fig. 6C). As a control, the same RA/F9 culture media and commercially purchased murine laminin were subjected to 6% SDS-PAGE and analyzed by immunoblotting directly with rabbit anti-mouse laminin antibody. The results obtained from immunoblotting the media of RA/F9 cells are similar to those obtained with commercial EHS laminin (Fig. 6D). Murine EHS laminin is known to contain terminal ␣-1,3-galactosyl residues and to be reactive with GS I-B 4 (62,63). Taken together, these data demonstrate that the major ␣-galactosylated glycoprotein secreted by RA/F9 cells is laminin.
Flow Cytometric Analysis of Cells for Altered Expression of Le X Antigens in Response to RA Treatment-Treatment of F9 cells with RA reportedly causes a decrease in expression of the Le X antigen within 2-3 days (14), but the cause of this change is not clear. Although this decrease might be due to loss of a GDP-Fuc:␤-D-GlcNAc ␣1,3-fucosyltransferase (␣1,3FT) (28), F9 cells treated with RA for 3 days have approximately the same amount of ␣1,3FT as untreated cells (29). Thus, there is a temporal difference between loss of Le X and decrease in the ␣1,3FT activity. We considered the possibility that the Le X antigen might be apparently lost because it could be "masked" by the addition of terminal ␣-galactosyl residues upon RAinduced differentiation. This might generate the structure Gal␣1-3Gal␤1-4(Fuc␣1-3)GlcNAc-R (termed ␣-galactosylated Le X ). It has been shown that at least one type of ␣1,3FT can utilize the Gal␣1-3Gal␤1-4GlcNAc-R to generate ␣-galactosylated Le X (64), and ␣-galactosylated Le X occurs naturally in mucins of cobra venom (65).
To test for the possibility of masking of Le X determinants, the levels of Le X antigen on F9 and RA/F9 cells were analyzed by flow cytometry using the anti-Le X monoclonal antibody (anti-CD15 monoclonal antibody). As expected, F9 cells express much higher levels of Le X antigen than RA/F9 cells (Fig. 7, A  and B). When F9 cells are treated with ␣-galactosidase, there is no significant change in the expression of the Le X antigen (Fig. 7A). In contrast, ␣-galactosidase treatment of RA/F9 cells causes a clear enhancement in the amount of surface Le X antigen (Fig. 7B). These results demonstrate that terminal ␣-galactosyl residues in differentiated F9 cells mask Le X determinants on RA/F9 cells generated by 3 days of RA treatment.
After treatment of F9 cells with ␣-galactosidase, there is only a slight decrease in GS I-B 4 staining (Fig. 7C). In contrast, treatment of RA/F9 cells with ␣-galactosidase causes a significant decrease in staining by GS I-B 4 (Fig. 7D). These results demonstrate that RA/F9 cells express more surface ␣-galactosyl residues than F9 cells and that many of the terminal ␣-galactosyl residues bound by GS I-B 4 on RA/F9 cells are accessible to ␣-galactosidase. Taken together, these results support the conclusion that RA treatment of F9 cells causes an increase in levels of surface glycoconjugates containing ␣-galac- F9 cells (panels A and C) and RA/F9 cells (panels B and D) were treated with either ␣-galactosidase(ϩ) or heat inactivated ␣-galactosidase(Ϫ) as described under "Experimental Procedures." The washed cells (5 ϫ 10 5 ) were analyzed for surface Le X expression by direct immunostaining with anti-CD15 monoclonal antibody (panels A and B) and for surface expression of the terminal ␣-galactose residues by indirect staining with biotinylated GS I-B 4 and streptavidin-phycoerythrin conjugate (panels C and D). The numbers in each panel refer to the experimental treatments as indicated on the right. tosyl residues and that these residues mask Le X determinants in RA/F9 cells. DISCUSSION To gain a full understanding of the roles of glycoconjugates in development and differentiation and in disease conditions, it is important to define the mechanisms regulating expression of glycosyltransferases and their cognate glycoconjugate structures. In many ways, F9 cells are ideal for studying these changes, since new glycoconjugate expression is inducible by RA within 3 days in the irreversibly differentiated cells. We previously reported that F9 cells have an ␣1,3GT that appears inducible by RA (29,34). Our studies now extend these findings to show that differentiation of F9 cells is accompanied by a 9-fold increase in steady-state levels of ␣1,3GT mRNA and that this increase is transcriptionally regulated. This induction is associated with secretion of the ␣1,3GT and enhanced ␣-galactosylation of surface glycoproteins and secretion of an ␣-galactosylated glycoprotein identified as laminin. Upon induced differentiation, there is a marked decrease in expression of surface Le X determinants, but this decrease is due in part to masking of the Le X determinants by ␣-galactosylation.
Although it is appreciated that expression of glycosyltransferases in some cases is regulated during cellular differentiation and transformation, little is understood about the mechanisms controlling enzyme expression (66 -68). Two of the best studied glycosyltransferases in this respect are the sialyltransferases (69) and the ␤1,4GT. Within the ␣2,6-sialyltransferase (ST6N) gene are at least four promoters, one of which is responsive to liver-restricted transcription factors (70) and another which appears to be B-cell specific and is regulated during B-cell development (71,72). The ␤1,4GT is known to have a regulatory element that controls transcript initiation (73) and may be related to hormone-dependent stimulation of the enzyme in mammary glands (74). Although the complete sequence of the murine ␣1,3GT gene is known (75), the cis-and/or trans-acting elements that regulate the transcription of ␣1,3GT are not yet understood.
There are several cases in vitro where treatments of cells with various agents appears to alter the levels of glycosyltransferases (76 -83). In regard to the ␣1,3GT, it has been observed that enzyme activity is higher in mouse peritoneal macrophages elicited with thioglycollate (84), which may correlate with increased binding to GS I-B 4 observed in stimulated macrophages compared to resident mouse macrophages (85). Other studies have examined the effects of RA on glycosyltransferase activities in F9 cells, but the results obtained are clearly different from those observed here for the ␣1,3GT. Prolonged treatment of F9 cells with RA causes an increase in activity for ␤1,6-N-acetylglucosaminyltransferase V and the core 2 ␤1,6-N-acetylglucosaminyltransferase (86). There is little change of activity in these enzymes, however, after 3 days of differentiation, but an increase in activity begins to occur after 4 days of differentiation.
The activity of some enzymes appears to decline in F9 cells after RA treatment. The N-acetylgalactosaminyltransferase activity that converts globoside to Forssman glycosphingolipid declines by approximately 70% in F9 cells within 3 days following RA treatment (87). Treatment of F9 cells for 5 days with RA and dibutyryl cAMP results in an approximately 80% decline in the activity of an ␣1,3FT (28). However, using a different assay for fucosyltransferase activity another group observed a slight increase in the activity of fucosyltransferase upon RA treatment of F9 cells (88). In the case of ␤1,4GT, it was shown that ␤1,4GT activity in F9 cells declines after 3 days of differentiation and then begins to rise around 5-6 days of differentiation, which correlates with increased levels of tran-script observed after 5-6 days of treatment with RA (53). Recently, it was reported that neither the ␤1,4GT enzyme nor transcript levels change significantly in F9 cells treated with RA alone, but treatment with RA and cAMP causes a 6.5-fold induction of the transcript in 8 days (54). Nuclear run-on experiments demonstrated that this increase in ␤1,4GT was not due to an increase in transcriptional rate, but it was due to post-transcriptional regulation (54). Induction of the ␣1,3GT by RA is clearly different from those other glycosyltransferases cited above, in that elevated transcripts and activity of the ␣1,3GT clearly begin to rise after 24 -36 h and are maximal at 3 days.
The time course of induction of ␣1,3GT transcripts by RA in F9 cells indicates that this gene is induced like other known late or secondary genes in contrast to RA induction of early or primary genes (44). Examples of the early inducible genes, which are transcriptionally activated within hours, are ERA-1/Hox-1.6 (58) and the Hox 1.3 (59). Their activation results from interaction of the RA-RA receptor complex with their promoters (44,89). Examples of the late inducible genes, which exhibit a delayed induction (24 -48 h), are laminin B1 and type IV collagen (␣1) (55,90). Whether the molecular mechanism for induction of the ␣1,3GT is related to those for these other late inducible genes is currently under investigation.
We observed that there is significant ␣1,3GT activity in the cell culture media and that Ϸ 2 ⁄3 of the total activity detectable in the cultured cells is recoverable in the media. The amount of activity in the media increases in response to RA treatment of the cells and is similar to the increase in activity observed in the cell extracts. Soluble forms of some other glycosyltransferases, such as ␣2,6-sialyltransferase and ␤1,4GT, have been found in various secretions and body fluids including milk (91), colostrum (92), and serum (48,52,93,94), and some soluble glycosyltransferases have been purified from these sources (91,92,95). These soluble forms appear to result from the proteolytic cleavage of the membrane-bound forms of the enzymes (47)(48)(49), and the levels of some soluble enzymes are affected by disease status and inflammation (50,51). Although the proteases responsible for the solubilization of glycosyltransferases have not been identified, there are suggestions that cathepsin D-like proteases within the acidic trans-Golgi might be involved (52). The mechanism by which ␣1,3GT is secreted into the culture media is not known, although we are currently attempting to define the N terminus of the secreted ␣1,3GT to identify the cleavage site. It would be expected that the cleavage site(s) occur in the so-called "stem-region," proximal to the C-terminal catalytic domain (67). The biological functions, if any, for secreted glycosyltransferases are not known. It is conceivable that the secretion of ␣1,3GT could be regulated and purposeful and that the soluble enzyme could have a function during differentiation of F9 cells. In preliminary studies, we have detected a soluble and active form of the ␣1,3GT in mouse serum and media from other cultured murine cells, indicating the secretion of soluble forms of this enzyme may be a common occurrence in cells.
Although the full-length transcript for the murine ␣1,3GT is 3.7 kilobases, four different mRNA transcripts that differ in the length of the sequences encoding the putative stem region have been detected by RNA-polymerase chain reaction analysis (75). It is thought that these transcripts arise by alternative splicing of a pre-mRNA according to a cassette model. Interestingly, although all four transcripts are present in both F9 and RA/F9 cells, only a single transcript for bovine ␣1,3GT was observed in bovine thymus and MDBK cells (75). The splice variants generate enzymes predicted to vary particularly in the stem region, which is predicted to contain the putative cleavage site for proteases. Future studies are required to understand the functions of different ␣1,3GT splice variants and whether these isoforms exhibit differential cell and/or tissue expression.
There are many studies demonstrating that expression of cell surface carbohydrates is developmentally regulated, but the mechanisms of regulation are largely unknown. Our finding that Le X antigens may be masked in differentiated F9 cells is consistent with some studies performed on mouse embryos. Both 8-cell mouse embryos and embryonic ectoderm of 5-and 6-day-old embryos express Le X antigens, and ␣-galactosidase treatment enhances expression of the antigen (96). Although there is scant information about antigen masking in vivo, such masking of carbohydrate determinants has been performed in vitro using recombinant glycosyltransferases. For example, transfection of cDNA for ␣2,6-sialyltransferase in Xenopus oocytes inhibits the formation of polysialic acid onto neural cell adhesion molecule (97), and transfection of ␣2,3-sialyltransferase into T lymphoblastoid cells converts the cells from a peanut agglutinin positive (PNA ϩ ) phenotype to PNA Ϫ , possibly caused by masking of the peanut agglutinin positive receptor, Gal␤1-3GalNAc␣1-Ser/Thr, with sialic acid (98). Our study demonstrates, however, that antigen masking by ␣-galactosylation is a biochemical response of the embryonal carcinoma cells to RA.
The observed masking of Le X antigens on RA/F9 cells suggests that differentiated cells synthesize the new sequence Gal␣1-3Gal␤1-4(Fuc␣1-3)GlcNAc␤1-R (␣-galactosylated Le X ). This possibility is supported by recent evidence that Gal␣1-3Gal␤1-4GlcNAc can be fucosylated in vitro using a partially purified ␣1,3-fucosyltransferase from human milk (64). In contrast, fucosylated oligosaccharides, such as Gal␤1-4(Fuc␣1-3)GlcNAc␣1-2Man, do not serve as acceptors for ␣1,3GT (99), indicating that there must be an ordered addition of terminal sugars. Interestingly, ␣-galactosylated Le X structures occur normally in oligosaccharides of mucins from cobra venom (65). It is conceivable that the carbohydrate phenotypes of F9 and RA/F9 cells are reflective of a balance between competing glycosyltransferases for available substrates. Competitive enzyme reactions are known to be important in influencing the structures of newly synthesized glycoconjugates (100 -103). We previously demonstrated that introduction of the murine ␣1,3GT into CHO cells results in synthesis of glycoconjugates containing terminal ␣-galactosyl residues and decreased levels of terminal sialic acid (45). Enzyme competition has been shown to be involved in the biosynthesis of mucin oligosaccharides in porcine and ovine submaxillary glands (103,104) and other mucin chains and blood group antigens (100).
The results of this study demonstrate that differentiation of F9 cells into parietal endoderm-like cells is accompanied by a profound increase in ␣1,3GT transcripts and enzyme activity, resulting in changes in the surface carbohydrate phenotype and masking of antigens. These changes may result in the synthesis of carbohydrate structures containing ␣-galactosyl residues, which could be functionally important for cellular interactions during murine embryogenesis (30 -33). There is particular interest in Le X antigens or masked Le X antigens because of evidence that they may participate in intercellular adhesion events during early embryogenesis via direct interactions with endogenous lectins or via direct carbohydrate-carbohydrate interactions (33). Although no lectins have yet been found that exclusively prefer Le X itself or ␣-galactosylated Le X as a carbohydrate ligand, many different animal lectins bind to oligosaccharides containing the underlying Le X structure, such as sialyl Le X and sulfated sialyl Le X (105). Future studies will be aimed at defining the mechanism by which the ␣1,3GT gene is responsive to RA, the detailed structures of the masked Le X antigens, and the possibility that murine cells express lectins interactive with Le X -containing oligosaccharides.