Suppression of integrin expression and tumorigenicity by sulfation of lactosylceramide in 3LL Lewis lung carcinoma cells.

To investigate the cellular functions of sulfated glycosphingolipids, we introduced the cerebroside sulfotransferase (CST) gene into J5 cells, a subclone of 3LL Lewis lung carcinoma cells. The J5 cells lack acidic glycosphingolipids but accumulate their common biosynthetic precursor, lactosylceramide. We established the stable CST transfectants, J5/CST-1 and J5/CST-2 clones, highly expressing sulfated lactosylceramide (SM3). Both clones exhibited more spherical morphology in comparison to mock transfectant, and their adhesiveness to fibronectin and laminin was significantly lower. The loss of cell-substratum interactions in these SM3-expressing cells could be attributed to decreased expression of integrins (alpha(5), alpha(6), and beta(1)) on the cell surface and their whole cellular levels. However, the levels of H-2K(b) and H-2D(b) antigens remained unchanged. Reverse transcriptase-polymerase chain reaction and Northern blot analyses for these integrins exhibited significant decrease of beta(1) gene expression in J5/CST-1 and 2, but there was no change in the levels of alpha(5) and alpha(6) transcripts. Deglycosylation by endoglycosidase H treatment clearly demonstrated that the precursor form of beta(1) integrin, possessing high mannose oligosaccharide chains, was preferentially decreased in the CST transfectants. These results demonstrate that endogenous SM3 negatively regulates beta(1) integrin expression at the transcriptional level, and the decrease of alpha integrin proteins in the CST transfectants was due to the post-transcriptional modification. We suggest the putative importance of the intracellular pre-beta(1) integrin pool for normal integrin maturation and subsequent function. Although the rates of cell proliferation in vitro for mock and CST transfectants were similar, tumorigenicity of J5/CST-1 and -2 cells inoculated into syngeneic C57/BL6 mice was greatly decreased or even absent. This was probably due to global loss of the efficient cell-matrix interactions, which are essential for the development of malignant tumors in vivo. Thus, we showed the evidence that cellular SM3 negatively regulates the cell-substratum interaction, resulting in the loss of tumorigenicity.

The physiological functions of sulfatides have long been noted for their unique property of having a strong negative electronic charge in the molecule. Sulfatides interact with various biomolecules especially in cell adhesion, differentiation, and signal transduction (reviewed in Refs. 5 and 6). In the case of cell-substratum adhesion, SM4 and SM3 have been found to bind several proteins such as laminin and thrombospondin (7). When exogenous SM4 was incorporated into SMKT-R3 cells (human renal cell carcinoma), their attachment to laminin, but not to fibronectin, was enhanced (8). However, it was uncertain whether the data obtained from these experiments truly reflect the cellular functions of endogenous sulfatides.
The purpose of the present study was to investigate the functional role of endogenous sulfatide in cell adhesion. We employed a subclone of the mouse 3LL Lewis lung carcinoma cell line, 3LL-J5, which has high LacCer content but no galactosylceramide and SM3 (9,10), to introduce the recently cloned CST gene (1). The stable CST transfectants expressing SM3 at a high level exhibited decreased adhesive abilities to both fibronectin and laminin. We demonstrated here an inverse (or negative) relationship between the level of cellular SM3 and cell adhesive ability as well as tumorigenicity.
(Carlsbad, CA), and LipofectAMINE PLUS Reagent was from Life Technologies, Inc. Human fibronectin and mouse laminin were from Biomedical Technologies Inc. (Stoughton, MA). A cell counting kit based on a modified 3-(4,5-dimethyl-2-thiazolyl)-2,5-diphenyl-2H-tetrazolium bromide method was employed (Dojindo, Kumamoto, Japan). For flow cytometry, anti-␣ 5 integrin (clone 5H10-27) was from PharMingen (San Diego, CA); anti-␣ 6 (monoclonal antibody 1378) and -␤ 1 (monoclonal antibody 1997) integrins were from Chemicon International Inc. (Temecula, CA); anti-SM3 (11) was from Seikagaku Corp. (Tokyo, Japan); and fluorescein-conjugated anti-H-2D b and H-2K b were from Meiji Nyugyo Corp. (Tokyo, Japan) as the primary antibodies. Fluoresceinconjugated anti-mouse IgM from Vector Laboratories Inc. (Burlingame, CA) and fluorescein-conjugated anti-rat IgG as the secondary antibodies were from Immunotech (Marseille, France). For Western blotting of integrins, anti-␣ 5 integrin (antibody 1928) and anti-␤ 1 integrin (monoclonal antibody 1997) were from Chemicon. For the chemiluminescence detection of glycolipids, horseradish peroxidase-conjugated anti-mouse IgM from Jackson ImmunoResearch (West Grove, PA) and the enhanced chemiluminescence system (ECL kit) from Amersham Pharmacia Biotech were used. Bicinchoninic acid reagent from Pierce was used for protein determination. All animal experiments were carried out in accordance with NIH Guide for Care and Use of Laboratories Animals and approved by the Animal Care and Use Committee in Hokkaido University.
Semi-quantitative RT-PCR-Total RNA was isolated from cultured cells using the Trizol reagent, and one-step RT-PCR was performed using the mixture of reverse transcriptase and Taq DNA polymerase (Life Technologies, Inc.) according to the manufacturer's instructions. The RNA was reverse-transcribed at 45°C for 30 min, and the cycling conditions were 94°C for 2 min, followed by 25 cycles of 94°C for 0.15 min, 50°C for 0.5 min, and 72°C for 2 min. The following primer pairs were used: glyceraldehyde-3-phosphate dehydrogenase, forward 5ЈACCACAGTCCATGCCATCAC3Ј and reverse 5ЈTCCACCACCCTGT-TGCTGTA3Ј (product size 451 bp); mouse integrin ␣ 5 , forward 5ЈTCG-CCTCGATCTCCTCTCCA3Ј and reverse 5ЈCGCTGCAGATAGATGTA-GAC3Ј (product size 1099 bp); mouse integrin ␣ 6 , forward 5ЈGTGAAC-GTGAGGTGTGTGAAC3Ј and reverse 5ЈCGCATGGTATCGGGGAAT-GC3Ј (product size 377 bp); mouse integrin ␤ 1 , forward 5ЈGACTTCCG-CATTGGCTTTGGC3Ј and reverse 5ЈCAAACACGACACCTGCACAC-G3Ј (product size 1205 bp). The concentration ratio of target cDNA primers to glyceraldehyde-3-phosphate dehydrogenase primers and the PCR cycle number was optimized for each reaction.
Northern Blot Analysis-Total RNA (10 g) from cells was denatured in 50% formamide, 6% formaldehyde, 20 mM MOPS (pH 7.0) at 65°C, electrophoresed on 1% agarose gel containing 6% formaldehyde, blotted onto a nylon membrane (Roche Molecular Biochemicals), and crosslinked by UV irradiation. A digoxigenin-labeled RNA probe for CST mRNA was synthesized from the XhoI fragment of pBS-hCST1 (1) using a DIG RNA labeling kit with T7 RNA polymerase (Roche Molecular Biochemicals) according to the manufacturer's instructions. A digoxigenin-labeled RNA probe for mouse ␤ 1 integrin mRNA was synthesized from the AvaII fragment of pGEM1-mouse ␤ 1 integrin (kindly provided by Dr. R. O. Hynes) using a DIG RNA labeling kit with Sp6 RNA polymerase. Mouse ␣ 5 integrin mRNA was synthesized by RT-PCR using RNA derived from mouse B16 melanoma cells with primers mouse ␣ 5 integrin, forward 5ЈTCGCCTCGATCTCCTCTCCA3Ј and reverse 5ЈCGCTGCAGATAGATGTAGAC3Ј (product size 1099 bp), and cloned into pGEM-T easy (Promega, Madison, WI). The plasmid was linearized with XbaI and transcribed with Sp6 RNA polymerase using a DIG RNA labeling kit as above. The membrane was stained with methylene blue for the detection of 18 S and 28 S rRNA and then hybridized with the RNA probe at 68°C. Detection was with a DIG-Luminescent Detection kit (Roche Molecular Biochemicals).
Sulfotransferase Assay-Cells cultured in a 100-mm dish were scraped off and washed three times with PBS. The cells were resuspended in 300 l of Tris-buffered saline (pH 7.4) containing 0.1% Triton X-100 and homogenized by sonication on ice. The protein concentration was adjusted to 1 mg/ml. CST activity was assayed in 30 l of a mixture containing 25 mM sodium cacodylate (pH 6.4), 50 M GalCer in 5% Triton X-100, 10 mM MnCl 2 , 1% Lubrol PX, 0.25 mM dithiothreitol, 5 mM NaF, 2 mM ATP, 50 mM NaCl, and 40 M [ 35 S]3Јphosphoadenosine 5Ј-phosphosulfate, plus 20 l of the enzyme source. This was incubated for 30 min at 37°C, and then 1 ml of chloroform/methanol/water (30: 60:8) was added. The mixture was applied to a DEAE-Sephadex A-25 column (Amersham Pharmacia Biotech) that was then washed with 2 ml each of chloroform/methanol/water (30:60:8) and methanol. [ 35 S]Sulfatide was eluted with 8 ml of 90 mM AcONH 4 in methanol directly into a scintillation vial. The radioactivity was determined in 10 ml of scintillation fluid by a liquid scintillation spectrometer (Amersham Pharmacia Biotech, type 1211).
Lipid Analysis-The total lipids were extracted from the cells with chloroform/methanol/water (4:4:0.3 and 2:4:0.3), successively. The combined extracts were then applied to a DEAE-Sephadex A-25 column (acetate form, 2.4-ml bed volume), and the neutral lipids were eluted with 12 ml of chloroform/methanol/water (30:60:8). The acidic lipid fraction was then eluted with 12 ml of chloroform/methanol/aqueous 0.8 M sodium acetate (30:60:8). The neutral and acidic lipid fractions were evaporated to dryness, and contaminating esters were methanolyzed with methanolic 0.1 M NaOH for 1 h at 40°C. The solution was neutralized with 1 M acetic acid in methanol, diluted with an equal volume of aqueous 50 mM NaCl, and applied to a Sep-Pak C 18 reverse-phase cartridge. The cartridge was washed with 40 ml of water, and lipids were eluted with 10 ml of methanol and 10 ml of chloroform/methanol (1:1), successively. The eluate was evaporated to dryness, and the lipids were analyzed by HPTLC. The plates were then developed with chloroform/methanol/aqueous 12 mM magnesium chloride (60:25:4, for neutral lipids) or chloroform/methanol/water (65:25:4, for acidic lipids). Glycolipids were visualized by spraying orcinol reagent and heating at 100°C for 10 min and then quantified with a dual-wavelength flying spot scanner (CS9300-PC, Shimadzu, Kyoto, Japan) in the reflectance mode at 500 nm with integrated areas.
Immunological Detection of SM4 and SM3-TLC immunoblotting was performed by the method of Taki et al. (12), slightly modified as follows. Acidic lipids were separated on an HPTLC plate with chloroform/methanol/water (65:25:4) and then immersed in a mixture of isopropyl alcohol/aqueous 0.2% calcium chloride/methanol (40:20:7) for 20 s. The plate was then covered with a polyvinylidene difluoride (PVDF) membrane (Immobilon, Millipore, Bedford, MA) and a glass microfiber filter (Atto Instruments, Tokyo, Japan). This was then pressed (level 8) for 50 s with TLC Thermal Blotter (Atto Instruments) at 180°C, after which the PVDF membrane was separated from the plate and dried. The PVDF membrane was agitated in 5% skim milk/ TBS-T (aqueous 137 mM NaCl, 20 mM Tris-HCl (pH 7.5), 0.05% Tween 20) for 1 h, and then the membrane was shaken in the primary antibody O4 (for galactosyl sulfatide and lactosyl sulfatide) (13) or anti-SM3 (for lactosyl sulfatide) (11) solution in 5% skim milk/TBS-T at 4°C overnight. The membrane was washed with TBS-T and dipped and shaken in the secondary antibody, peroxidase-conjugated goat anti-mouse IgM, solution in 5% skim milk/TBS-T for 1 h. After washing with TBS-T, the membrane was analyzed with an enhanced chemiluminescence system (ECL kit).
Western Blotting of Integrins and Treatment of Endoglycosidase H-Subconfluent mock and CST transfectants cells cultured in 75-cm 2 culture dishes were rinsed with cold PBS and were solubilized in SDS-gel sample buffer containing 0.4% SDS, 1% ␤-mercaptoethanol, and proteinase inhibitors, and the amounts of protein in all samples were determined using BCA reagent. The equal amounts of samples were then boiled for 3 min and clarified by centrifugation for 5 min at 10,000 rpm in a microcentrifuge. When deglycosylation of high mannose oligosaccharide chains was performed prior to SDS-PAGE, each denatured sample was further treated with 1,000 units of endoglycosidase H (EndoH f , BioLabs Inc. Beverly, MA) for 1 h 37°C according to the manufacturer's instructions. Samples were subjected to SDS-PAGE and transferred to PVDF membranes and blocked with 5% skim milk/ TBS-T. The membranes were incubated with primary antibody (anti-␣ 5 or anti-␤ 1 integrin), washed, and then incubated with the appropriate horseradish peroxidase-coupled secondary antibody for 60 min at room temperature. Immunoreactive proteins were visualized by autoradiography using an enhanced chemiluminescence system (ECL kit).
Flow Cytometry-Cells were detached from the culture surface by a 20-min incubation at 37°C in 10 mM EDTA/PBS, pelleted, washed with FACS buffer (0.1% BSA and 0.1% NaN 3 in PBS) twice, and then resuspended in FACS buffer containing 10 g/ml anti-␣ 5 , ␣ 6 , ␤ 1 integrin or anti-SM3 for 1 h at 4°C. For the detection of whole cellular integrins, cells were permeabilized with 0.1% saponin containing FACS buffer for 15 min at 4°C before adding the appropriate primary antibody. Then the cells were washed with FACS buffer three times, treated with fluorescein-conjugated anti-rat IgG or fluorescein-conjugated antimouse IgM for 1 h at 4°C in the dark, and washed with FACS buffer as above. The intensity of cell fluorescence was determined by a FACScan cytometer (Becton Dickinson). For the detection of MHC, cells treated without or with saponin were stained with 100 g/ml fluorescein-conjugated anti H-2D b or H-2K b in FACS buffer for 1 h at 4°C and washed with FACS buffer as above.
Cell Attachment Assay-100 l of fibronectin or laminin (1, 5, 10 and 25 g/ml) in PBS were added to each well of 96-well plates, incubated overnight at room temperature, and removed. The coated wells were further incubated with 100 l of 0.1% BSA in PBS at room temperature for 1 h and washed with PBS three times. Each well was incubated with 50 l of 0.01% BSA in RPMI 1640 medium at 37°C for 1 h. A 50-l suspension of J5/CST-1, J5/CST-2, or mock cells (5 ϫ 10 3 ) in 0.01% BSA/RPMI 1640 was added to the fibronectin-or laminin-coated wells and incubated for 30 min. Non-adherent cells were removed by inverting the plate, and each well was gently washed with 100 l of serumfree RPMI 1640 medium. To each well were added 100 l of the same medium followed by 10 l of cell counting kit. After incubation at 37°C for 2 h, the absorbance (450 nm) of formazan generated in the wells was measured with a dual-wavelength flying spot scanner (CS9300-PC, Shimadzu, Kyoto, Japan). The attachment ability was expressed as the percentage of attached cells (absorbance of attachment cells in the well/absorbance of total cells added to the well ϫ 100).

Establishment of the Stable CST Transfectants
Highly Expressing SM3-Previously, we subcloned the J5 clone, which lacked acidic GSLs and accumulated LacCer, from the wild type of murine 3LL Lewis lung carcinoma cells (9). The CST gene was introduced into the J5 clone to generate sulfatideexpressing clones. A total of 24 transfectant clones were finally obtained by limited dilution, and 2 clones (J5/CST-1 and J5/ CST-2) which expressed relatively high levels of CST gene mRNA (Fig. 1A) and CST activity (Fig. 1B) were chosen for further study. The mRNA and the enzyme activity of CST appeared only in the clones into which the vector constructed with the CST gene had been introduced but not in the mocktransfected clones (Fig. 1, A and B). Because the J5 clone expresses only LacCer but not GalCer (9, 10), introduction of the CST gene into this clone should express only SM3 as a sulfatide. The GSL fractions were prepared from mock and J5/CST-1 and -2 cells, and the SM3 content in proportion to the CST activity was confirmed on HPTLC visualized with orcinol reagent (Fig. 1C). There was no visible staining of SM4 in both mock and the CST transfectants.
Analyses of the neutral GSLs of J5/CST-1 and -2 cells showed that not only the contents of LacCer and its precursor GlcCer were decreased, but also that of the globo type GSL, Gb3 (Fig.  1C). This indicated that the biosynthetic pathway from LacCer (at the LacCer branching point) was greatly shifted to the sulfo type. We further identified only SM3 but not SM4 by immunoblotting on HPTLC using anti-SM3 monoclonal antibody (11), which recognizes only SM3, and with O4 antibody (13), which recognizes both SM4 and SM3 (Fig. 1D). By using the anti-SM3 antibody, strong expression of SM3 on the cell surface was confirmed by flow cytometry (Fig. 1E). Thus, we designated J5/CST-1 and -2 cells as SM3 high expressors, and the following experiments were carried out using these two clones and the mock transfectant.
Comparison of Cell Growth, Morphology and Cell-Substratum Adhesion-When in vitro growth of mock and CST transfectants was compared, there was no difference in the rates of cell proliferation under the normal culture conditions on plastic plates ( Fig. 2A). On the other hand, there was a significant increase of the spherical form in the CST transfectants (Fig.  2B), indicative of weak adhesiveness of the SM3-expressing cells. Both J5/CST-1 and -2 cells exhibited a marked decrease in their adhesive abilities to the plastic coated with laminin and fibronectin (Fig. 3A). Moreover, the cells highly expressing SM3 were unable to spread on the surface coated with fibronectin even 2 h after seeding (Fig. 3B).
Decreased Integrin Expression in CST Transfectants-The expression of integrin molecules on the cell surface involved in the recognition of fibronectin (␣ 5 and ␤ 1 ) and laminin (␣ 6 and ␤ 1 ) was analyzed by flow cytometry ( Table I). All of the integrins examined here exhibited a marked decrease (around 32-43% in both J5/CST-1 and -2 cells). After cell permeabilization with saponin, almost similar decreases in the cellular integrins were observed, demonstrating that the decreases in cell surface integrins in the SM3 highly expressing cells were due to their  (Table I). Therefore, the decrease of integrin expression in the CST transfectants may be a specific cellular event regulated by cellular SM3. As shown in Fig. 4, the significant reduction of ␣ 5 and ␤ 1 proteins in CST transfectants was observed by Western blot analysis, confirming the data of flow cytometry (Table I). It has been reported that cellular integrins, especially ␤ 1 -subunit, can be separated into two bands on SDS-PAGE due to the different oligosaccharide chains on the protein; the lower band has been designated the precursor form of ␤ 1 -subunit (high mannose type) and the upper band is the mature form (complex type) (14,15). We also observed two immunoreactive bands in lysates of both mock and CST transfectants by anti-␤ 1 antibody, and the transition from the pre-␤ 1 to the maturely glycosylated form increases the size from ϳ120 to 130 kDa (Fig. 4A). Removal of immature N-linked glycan chains by digestion with endoglycosidase H (16) reduced the size to ϳ80 kDa, whereas the maturely glycosylated ␤ 1 band of ϳ130 kDa was not cleaved by endoglycosidase H (Fig. 4A). Thus, the decrease of ␤ 1 integrin protein in CST transfectants was mainly due to the dramatic decrease of the precursor form (Fig. 4A). The amount of ␣ 5 integrin protein in CST transfectants was decreased to around 50% in comparison to that of mock transfectant (Fig. 4B). The determination of ␣ 6 integrin by Western blotting using two different anti-␣ 6 antibodies was unsuccessful due to nonspecific staining. By taking the data from Table I and Fig. 4, we concluded that the cell surface expression of ␣ 5 , ␣ 6 , and ␤ 1 integrins as well as their chemical quantities were decreased in CST transfectants.
Selective Decrease of ␤ 1 Integrin mRNA in CST Transfectants-We asked here whether the decreased integrin contents in the CST transfectants originated from the decreased cellular amounts of their mRNAs themselves. Semi-quantitative RT-PCR for ␣ 5 , ␣ 6 , and ␤ 1 integrins showed the selective decrease of ␤ 1 gene expression levels in J5/CST-1 and -2 (Fig. 5). There was no difference between the mRNA levels of ␣ 5 and ␣ 6 integrins in both the mock and CST transfectants at various PCR cycles under non-saturating conditions, and the representative picture was presented in Fig. 5. Northern blot analysis clearly indicated the significantly lowered level of ␤ 1 integrin mRNA in both SM3 high expressing cells, J5/CST-1 and J5/CST-2, but only a slight decrease in the SM3 low expressing J5/CST-3 cells, showing the inverse relationship between the levels of cellular SM3 and ␤ 1 integrin mRNA content (Fig. 6, A versus  C). The levels of ␣ 5 transcript both in mock and CST transfectants were essentially the same (Fig. 6B). These results demonstrate the selective transcriptional down-regulation of ␤ 1 integrin by endogenous SM3.
Loss of Tumorigenicity-Since the tumorigenic and metastatic potentials of tumors are greatly affected by the expression levels of integrins (17)(18)(19)(20), we were interested in the behavior of the SM3-expressing cells in vivo. When the mock and the CST transfectants were inoculated subcutaneously into syngeneic C57/BL6 mice, and then examined for their tumor growth, a remarkable decrease or even no sign of tumor growth was observed (Fig. 7).
Since no difference in in vitro cell growth in culture plastic between the mock and the CST transfectants was observed ( Fig. 2A), the loss of tumorigenicity in vivo could be due to the global loss of integrin functions that are essential for malignant tumor growth. DISCUSSION LacCer is the common precursor of numerous GSLs consisting of six groups classified as the ganglio, globo, isoglobo, neolacto, lacto, and sulfo series. Their expression is primarily determined by cell type-specific expression of enzymes at the lumenal side of Golgi membranes responsible for either glycosylation (21) or sulfation (22,23) of branch point LacCer. Since the composition and distribution of LacCer-derived GSLs are known to be greatly altered during development and oncogenic transformation (reviewed in Ref. 24), the elucidation of the biological significance of LacCer branching in various cellular functions is one of the most important issues of glycobiology. The cDNAs of the branching enzymes have been extensively cloned recently (1,(25)(26)(27)(28)(29)(30), and it is now possible to clarify the functions of GSLs utilizing these synthase genes.
As our first trials to study the biology of LacCer branching, we were able to show the functioning of cellular SM3 in cellsubstratum adhesion using genetically manipulated SM3-expressing cells, containing the CST gene in the J5 clone of 3LL Lewis lung carcinoma cells (9). Significant findings demonstrated here are as follows. 1) The transfectants expressing SM3 at a relatively high level exhibited decreased adhesion to both fibronectin and laminin substrata. 2) The defect in cell adhesion could be attributed to decreased expression of integrin proteins, including ␣ 5 -, ␣ 6 -, and ␤ 1 -subunits, on the cell surface as well as their whole cellular levels. Although there might be a general decrease of cellular integrins in the CST transfectants, MHC expression was, however, not altered significantly, suggesting a selective action of cellular SM3 on integrin expression. 3) Integrin expression is regulated at both transcriptional and post-transcriptional levels, since only ␤ 1 integrin mRNA but not ␣ 5 and ␣ 6 mRNAs decreased in SM3expressing cells. 4) Although the rates of cell proliferation in vitro were similar for the mock and the CST transfectants, tumorigenicity of the SM3-expressing cells in vivo was dramat- ically lower, probably due to the global loss of the efficient cell-matrix interactions in vivo.
The mechanism by which the CST gene causes the selective decrease of ␤ 1 integrin mRNA remains to be elucidated. Many cytokines are known to regulate integrin transcription (reviewed in Ref. 31). The initial signaling cascade of several cytokines may occur at lipid rafts, which are believed to function as signaling domains in the plasma membrane (reviewed in Ref. 32). GSLs including sulfatides are concentrated in the lipid rafts, which can be isolated as the detergent-insoluble microdomains (DIMs) (33). If the distinctive organization and functions of lipid rafts from the mock and CST transfectants can be elucidated, the information might explain the transcriptional suppression of ␤ integrin. We have obtained preliminary evidence regarding differences in the protein composition and patterns of tyrosine-phosphorylated proteins in the two types of DIMs. 2 One could hypothesize the presence of an SM3-binding protein in the DIMs of 3LL Lewis lung carcinoma cell lines, since the sulfatide-binding proteolipid protein, the rat myelin and lymphocyte protein, has been identified in a detergent-insoluble complex obtained from oligodendrocytes and Schwann cells (33). The rat myelin and lymphocyte protein is a member of the tetraspanin membrane protein (TMP) group. Other TMPs, such as CD9 (35)(36)(37)(38), CD53 (39), CD63 (36,39,40), CD81 (37,39), and NAG-2 (41), have been known to associate with inte-   grins to make functional complexes that might modulate integrin signaling (42). Since it has been reported recently that association of integrin and TMP in DIMs is observed under certain conditions (43,44), sulfated GSLs concentrated in DIMs (32) might influence the integrin-TMP complexes. Further studies are needed to examine these hypotheses.
Many cell types have a large intracellular pre-␤ 1 integrin pool in endoplasmic reticulum (45)(46)(47)(48)(49) and form ␣␤ 1 heterodimers when they are still located in this compartment (45). It has been also suggested that the excess of pre-␤ 1 in endoplasmic reticulum might be required for the efficient heterodimer formation (47). As demonstrated in Table I and Figs. 4 -6, we observed a significant decrease in the pre-␤ 1 pool correlated with the reduction of the ␤ 1 -transcript in the SM3expressing cells. Although the protein levels of ␣ 5 and ␣ 6 integrins were considerably diminished similar to that of the pre-␤ 1 , the ␣ 5 -and ␣ 6 -transcripts remained unaltered. An interesting hypothesis is that under conditions in which only a limited number of pre-␤ 1 -chains are available, the ␣-subunits inefficiently form ␣␤ 1 heterodimers, thereby leading the en-hancement of intracellular degradation of ␣-subunits before reaching the plasma membranes. In fact, this idea is supported by our experimental data showing the parallel decreases on the cell surface expression between ␣-subunits and ␤ 1 integrin in the SM3-expressing cells (Table I). We also observed a decrease of spontaneous cell motility in CST transfectants (data not shown). Therefore, we are currently studying the effects of endogenous SM3 on integrin biosynthesis and recycling by the metabolic labeling and pulse-chase experiments.
Since it is known that SM4 and SM3 can directly bind to laminin and other bioactive molecules under cell-free conditions (7), we initially expected enhancement of cell adhesion to laminin in the SM3-expressing J5/CST-1 and -2 cells. However, this expectation was not corroborated in our cell adhesion assay due to the significant down-regulation of integrin expression (␣ 6 and ␤ 1 integrins) which occurred when SM3 was expressed endogenously by CST transfection. In other words, we present the evidence for the first time that endogenous sulfated GSLs, especially SM3, negatively regulate cell-substratum adhesion. If the other anionic GSLs, such as gangliosides, exhibit similar or opposite effect against cellular SM3, then our current approach to express SM3 alone by introducing CST gene into the cells lacking all anionic GSLs would be one of the most practical approaches to reveal and identify the functional role of individual anionic GSL molecule. Interestingly enough, when we transfected GM3 synthase gene (26) constructed with the same vector used in this paper into J5 cells to express GM3, the GM3-expressing cells exhibited the ability to attach and spread on fibronectin more efficiently. 3 In contrast to the SM3expressing cells, ␤ 1 integrin mRNA content was elevated in the GM3-expressing cells. This finding would explain the distinctive roles of sulfation and sialylation on carbohydrates as well as the significance of GSL function at the LacCer branching point. 3 Further detailed study will be performed to verify whether or not the suppressive effect of SM3 on integrin expression is a physiologically and/or pathologically relevant event in cells originally expressing SM3, such as renal cell lines (50,51) and in tissues expressing CST gene, such as stomach, small intestine, brain, kidney, lung, and testis (4). We are currently examining the effects of CST gene or antisense CST gene transfection into various non-transformed and malignant tumor cells in cell adhesion as well as integrin expression. Since many tumor cell lines express ganglioside GM3, which is biosynthe- sized from LacCer by the action of GM3 synthase (26), it is also interesting to compare the degrees of the competing GM3 and SM3 expressions.
Integrins have been implicated in the processes of transformation and differentiation of certain cell types. Transformation of cells to a malignant state is often accompanied by quantitative changes in integrin expression and/or alteration of the types of integrins expressed. For example, Chinese hamster ovary cells deficient in fibronectin receptor (␣ 5 ␤ 1 integrin) exhibited increased tumorigenicity (52). Overexpression of ␣ 5 ␤ 1 integrin into Chinese hamster ovary cells led to loss of tumorigenicity (53). On the other hand, increased expression of ␣ 6 integrin, which involves laminin binding, showed enhanced metastatic potential in Lewis lung carcinoma cells (17). The ␤ 1 -subunit associates with at least 10 ␣-subunits (␣ 1 -␣ 9 and ␣ v ), constituting the largest subfamily of the integrins. Targeted disruption of the ␤ 1 integrin gene (20) or administration of anti-␤ 1 antibody (18) greatly reduced metastatic capacity. Since the CST transfectants obtained here exhibited global loss of ␣ 5 , ␣ 6 , and ␤ 1 integrin expression, we were interested in examining their malignancy in vivo. Surprisingly, tumorigenicity of the SM3-expressing cells inoculated into syngeneic mice was decreased drastically or even lost, although there was no difference in vitro cell growth potential among these clones. Therefore, we assume that the loss of tumorigenicity might be due to the global loss of the efficient cell-matrix interactions, which are essential for the development of malignant tumors in the body. Our findings might open up a new strategy of gene therapy for malignant and metastatic tumors, if it becomes possible to transfect human tumors in vivo with the human CST gene.