Oncogenic Ki-ras but Not Oncogenic Ha-rasBlocks Integrin β1-Chain Maturation in Colon Epithelial Cells*

Human colorectal tumors commonly contain mutations in Ki-ras but rarely, if ever, in Ha-ras. The selectivity for Ki-ras mutations in this tumor was explored using the HD6-4 colon epithelial cell line which contains no ras mutations. After adhesion to an extracellular matrix, HD6-4 cells polarize into columnar goblet cells with distinct apical and basal regions. Stable HD6-4 transfectants were made with mini-gene constructs of the oncogenic cellular Ki-ras4BG12V gene, the oncogenic Ha-ras G12V gene, or mini-gene constructs of wild-type Ki-ras4B as a control. Ki-rasmutations, but not Ha-ras mutations, disrupted colon epithelial cell apicobasal polarity and adhesion to collagen I and laminin. Three Ha-ras transfectants and three Ki-ras transfectants exhibited Ras proteins expressing the Val-12 mutation by Western blotting with pan-ras G12V antibody. Only wild-type Ki-ras transfectant cells and oncogenic Ha-rastransfectant cells synthesized the mature, fully glycosylated forms of β1 integrin. Instead of the mature integrin β1-chain, a faster migrating β1-chain intermediate was detected on the cell surface and in the cytoplasm of the oncogenic Ki-ras transfectants. Expression of the oncogenic Ki-ras gene caused the altered β1 integrin maturation because phosphorothiolated antisense oligonucleotides to Ki-ras reduced expression of both the mutant Ki-Ras protein and the aberrant integrin β1-chain and increased expression of the mature integrin β1-chain. Altered glycosylation generated the new β1 integrin form since integrin core β1-chain proteins of the same molecular weight were yielded in Ki-ras, Ha-ras, and control transfectants after removal of sugar residues with endoglycosidase F or following tunicamycin treatment to inhibit glycosylation. The selective effect of oncogenic Ki-ras on β1 integrin glycosylation was not due to selective activation of mitogen-activated protein kinases because both mutated Ki- and Ha-ras genes activated this pathway and increased cell proliferation. Since blocking the glycosylation of integrin β1-chain inhibited the adherence, polarization, and subsequent differentiation of colon epithelial cells, the selective effects of the oncogenic cellular Ki-ras gene on integrin β1-chain glycosylation may account, at least in part, for the selection of Ki-ras mutations in human colon tumors.

The mammalian ras gene family contains three homologous members, Ki-ras, Ha-ras, and N-ras. Each encodes a 21-kDa protein of either 188 or 189 amino acid residues. The first 85 amino acids of each ras isoform are identical, and the next 80 amino acids exhibit an 85% homology. The remaining C-terminal sequence is highly divergent between the isoforms and so is termed the hyper-variable domain. The last four amino acids in the C terminus constitute the CAAX motif, which is required for post-translational modification and subsequent membrane localization after gene-specific lipid modifications (1). The Cterminal region can substitute for the entire gene in localizing the downstream effector raf to the plasma membrane (2).
Ras proteins switch between an active form that binds GTP and an inactive form that binds GDP (3). Activated ras in turn activates the serine-threonine kinase raf-1 which then activates MAP 1 kinase kinases (MEKs) which in turn activate the p42/44 MAP kinases (Erks). Ras also activates raf-independent signaling pathways leading to activation of JNK kinases. Although there is evidence linking the raf/MEK/MAP kinase pathway to cellular transformation, ras may mediate some aspects of transformation through raf-independent pathways. Non-raf candidate ras effectors include Rho family proteins, two Ras GTPase-activating proteins (p120 and NF1), guanine nucleotide exchange factors for Ral proteins, and phosphatidylinositol-3-OH kinase (reviewed in Ref. 4). The possible contributions of ras activation to transformation include altered transcription and translation, alterations in the cytoskeleton, and altered cell surface carbohydrates (5,6). Oncogenic mutations detected in Ha-, Ki-, and N-ras genes isolated from different human tumors localize in the N-terminal region controlling GTP binding, suggesting the genes become oncogenic when they remain in the GTP-bound state. Mutations in predominantly one of the three mammalian ras genes have been found associated with specific human cancers. For example, Ha-ras mutations have been reported in 18% of transitional cell carcinomas of the human urinary bladder, whereas very few mutations in Ki-ras were observed (7,8). In contrast, the vast majority, over 90%, of pancreatic adenocarcinomas contain mutations in the Ki-ras gene and not in N-ras or Ha-ras (9). Specificity of ras mutations is also seen in colon cancers, in which roughly 40 -50% of cases exhibit activating mutations in the Ki-ras gene. Only a few percent of cases exhibit N-ras mutations, whereas Ha-ras mutations are very uncommon in colon cancers (10,11). N-ras mutations have been found in approximately 20 -25% of cases of acute myeloid leukemia, although mutations in Ki-ras were infrequent (12). The reason for the selectivity for a specific mutated ras isoform in a tumor type is not known but may be due to the importance of that isoform in controlling proliferation within that cell type. Recently it was shown using gene-specific antisense oligonucleotides that Ki-ras, but not Ha-ras, contributes to the proliferation of normal human lung fibroblasts (MCR-5 cells), whereas oncogenic Ha-ras drives proliferation of T24 human bladder carcinoma cells (13).
Both Ha-Ras and Ki-Ras proteins are expressed in colon epithelial cells (14) but may have different functions. We wished to determine whether mutations in the Ki-ras gene which are found in colon carcinoma cells induce another function in addition to cell growth through the MAP kinase cascade. Viral Ki-ras prevents the polarization of MDCK kidney epithelial cells (15,16) suggesting the mutated cellular Ki-ras gene may have a similar function in colon epithelial cells. Studies in tissue culture have shown that cellular polarization may require both E-cadherin-mediated cell to cell contact and integrin-mediated cell-substratum interactions (17). E-cadherin is critical to the formation of the basolateral domain, whereas the orientation of the apicobasal axis depends on integrin-mediated associations with the substratum.
We now compare the effects of two ras genes, the Ki-ras which is often mutated in colon cancer and the Ha-ras which is rarely, if ever, mutated in colon cancer, on colon epithelial cell polarity, integrin-mediated adhesion, and integrin expression and function. The expression plasmids utilized are a mini-gene construct of the large cellular Ki-ras4B gene (18) and an oncogenic cellular Ha-ras gene, both mutated at codon 12 from Gly to Val. The colon epithelial cell used is the HD6-4 line, which binds strongly to collagen I through the integrin heterodimer ␣2␤1 (19). After binding to collagen I, HD6-4 cells then polarize into columnar cells with distinctive basal and apical compartments. These columnar cells differentiate into a specialized epithelial cell type, the colon goblet cell, with a basal nucleus and an apical theca containing mucin granules (20). HD6-4 cells contain inactivating mutations in the tumor suppressor genes p53 and APC but have wild-type ras genes, making them suitable recipients for oncogenic ras genes (21). We now find that constitutive expression of oncogenic Ki-ras, but not oncogenic Ha-ras, prevents the establishment of columnar cell polarity in HD6 cells by blocking glycosylation and maturation of integrin ␤1-chain, thus reducing the capacity of this integrin to mediate binding to extracellular matrix components.

EXPERIMENTAL PROCEDURES
Materials-[␣-32 P]GTP, 125 I, [ 32 P]H 3 PO 4 , and [ 35 S]methionine were obtained from NEN Life Science Products, protein A-Sepharose from Pharmacia Biotech Inc., PVDF transfer paper Immobulin-P from Millipore, and PEI-cellulose-F TLC plates from EM Separations. Pan-ras rat monoclonal antisera Y13-259 which reacts with the p21 translational products of the Ha-, Ki-and N-ras human oncogenes; monoclonal pan-ras Val-12 antibody which reacts with only the forms of the ras oncogenes mutated at codon 12 to valine; c-Ki-ras Ab1 (clone 234-4.2) a mouse monoclonal antibody specific for c-and v-Ki-ras p21 and not recognizing c-Ha-ras or c-Ha-ras p21; c-Ha-ras (Ab-1) clone 235-1.7.1 a mouse monoclonal antibody specific for Ha-Ras and not Ki-Ras or N-Ras p21 proteins, and purified p21 recombinant protein Ki-Ras Val-12 were purchased from Oncogene Science. Phosphorothioate oligodeoxynucleotides were a gift of Dr. Brett Monia, Isis Pharmaceuticals, Carlsbad, CA. Isis-2503 is a 20-mer targeted to the initiation codon (AUG) of c-Ha-ras mRNA (TCC-GTC-ATC-GCT-CCT-CAG-GG), and Isis-13177 is a 20-mer of random sequence; Isis-6957 is a 20-mer targeted to the 5Ј-UTR of Ki-ras (CAG-TGC-CTG-CGC-CGC-GCT-CG). This sequence is found within the promoter region of c-Ki-ras4B gene cloned into pMiK Val-12 (see below), about 60 base pairs upstream of the translational start. Anti-integrin ␤1, a polyclonal mouse IgG-clone 18 raised to a protein fragment corresponding to amino acids 76 -256 of human integrin ␤1-chain, a region of the extracellular domain, was purchased from Transduction Laboratories and was used for immunoblotting. Monoclonal antibody clone P4C10, an IgG1 isotype directed to integrin ␤1-chain was purchased from Life Technologies, Inc. and was used for immunoprecipitation and cell binding studies. Phospho-specific MAP kinase antibody and p44/42 MAP kinase antibody were rabbit polyclonal IgGs from New England BioLabs, and endoglycosidase F (Nglycosidase F-free), also known as endo-␤-N-acetylglucosidase F, was purchased from Boehringer Mannheim. Collagen I was purchased from the Collagen Corp., Palo Alto, CA. Fibronectin and tunicamycin were purchased from Sigma, and laminin was from Life Technologies, Inc.
Cell Culture and Cell Binding Assays-The HD6 human colon carcinoma cell line (20) was subcloned from the HT29 cell line and recloned as the HD6-4 line immediately before transfection. The parental and transfectant lines were maintained in DME medium containing 7% fetal bovine serum or ITS-DME medium, as described (20). Binding to 30 g/ml collagen I-coated, 10 g/ml fibronectin-coated, or 40 g/ml laminin-coated 24-well plates was performed exactly as described (19).
Transfection-pMiK Cys encodes the endogenous cellular Ki-ras promoter, exons 1, 2, 3, and 4B, and 4.2 kilobase pairs of the 3Ј-untranslated sequence. The mini-gene constructs pMiK Gly and pMiK Val-12 plasmids (18) were made from pMiK Cys , which was cut by PstI and selfligated. Into the StuI-NsiI sites of this plasmid was ligated the first exon of Ki-ras amplified by polymerase chain reaction with primers KIUSX and KIDAX from tumors with determined mutations in codon 12 of Ki-ras. A KpnI fragment carrying a lacI-Z gene was ligated into KpnI-SmaI sites of pBluescriptIKsϩ (Stratagene) and designated pML1. A KpnI-BamHI fragment of pML1 and XhoI-BamHI fragment of pMClneopoly(A) (Stratagene) containing the neo gene were ligated into KpnI-XhoI sites of pBluescriptIKsϩ and designated pMLN. pMLN was linearized with BssH2 and then cut by SalI to obtain a fragment carrying the neo gene. This was ligated to BssSalI sites of pMiK Val-12 and pMiK Gly-12 to give plasmids pML Val-12 and pML Gly-12 . These plasmids were transfected into HD6-4 cells by calcium phosphate-mediated transfection followed by osmotic shock and selected in 600 g/ml G418 in DME medium. pUC EJ 6.6 encoding the mutated human Ha-ras gene cloned from the human EJ bladder carcinoma and pSV2neo encoding the neomycin resistance gene under control of the SV40 late promoter were co-transfected at a ratio of 9:1 by calcium phosphate-mediated transfection followed by osmotic shock and selected in 600 g/ml G418 in DME medium. Individual clones of Ha-ras and Ki-ras transfectants were isolated.
Cell Permeabilization-The method was adapted from Buday and Downward (23). Cells were cultured 2 days post-plating at 4 ϫ 10 5 /cm 2 in 10-cm tissue culture dishes to bring cells into log phase. The medium was then changed to serum-free ITS-DME medium, and cells were cultured overnight. After 1 ϫ wash with PBS, cells were placed in 2.4 ml of permeabilization buffer consisting of 150 mM KCl, 37.5 mM NaCl, 6.25 mM MgCl 2 , 0.8 mM EGTA, 1 mM CaCl 2 , 1.25 mM ATP, 12.5 mM PIPES, pH 7.4, and incubated at 37°C for 10 min. 0.6 ml of 2 international units/ml streptolysin O (Sigma) in permeabilization buffer was then added, and the incubation was continued for 5 min. 15 Ci of [␣-32 P]GTP (300 Ci/mmol) were added and incubated for 10 min. The buffer was then removed, the cells lysed, and Ras proteins immunoprecipitated as above. Total specific radioactivity associated with immunoprecipitated Ras proteins was detected by a Beckman 2000 counter. Samples were assayed in triplicate.
Immunodetection-50 g of cell lysate (22) proteins were blotted onto PVDF membranes after separation on 8% SDS-PAGE. The blots were blocked in blocking buffer: 25 mM Tris, pH 8, 125 mM NaCl, 0.1% Tween 20, 4% bovine serum albumin for 1 h at room temperature, incubated for 2 h with a 1:2500 dilution of mouse IgG clone 18 antibody to integrin ␤1-chain, or 5 g/ml of the Ki-ras-specific, Ha-ras-specific, or pan-ras antibodies and detected by enhanced chemiluminescence. After SDS-PAGE and blocking as above, tyrosine-phosphorylated erk1 and erk2 and total erk1 and erk2 were detected, respectively, by 1 g/ml phospho-MAP kinase and 1 g/ml total MAP kinase antibodies and then detected using the Western blotting detection system provided by the manufacturer: a 1-h incubation at room temperature with 1:1000 dilution of alkaline phosphatase-conjugated anti-rabbit secondary antibody and 1:1000 dilution of alkaline phosphatase-conjugated anti-biotin antibody in blocking buffer, followed by a wash, and detection using a 1:500 dilution of CDP-Star for 5 min, followed by autoradiography.
Antisense Oligonucleotide Treatment-Treatment of Ki-ras and Haras transfectants was essentially as described (13). Cells were seeded at 2 ϫ 10 5 per well in 6-well plates. 48 h later the cells were washed with pre-warmed serum-free ITS-DME medium and then incubated in this medium with a fixed ratio of oligonucleotide to Lipofectin (2.4 l of Lipofectin per 40 pmol of oligonucleotides) for 4 h. The oligonucleotidecontaining medium was then replaced with normal growth medium, and growth was continued for 48 h to allow ras turnover plus reduced ras mRNA levels to result in reduced Ras protein level (13).
Differentiation Assay-Cells were cultured on Costar transwells with 2 ml of medium under the layer and 1.5 ml above the layer for 2 weeks post-confluence with 3 times weekly media changes, fixed, and stained for mucin detection by Alcian blue dye with a nuclear fast red counterstain, exactly as described (20).
Kinase Assays in Polyacrylamide Gels Containing MBP-The method is adapted from one used previously (22). Cell lysates in high salt EBC buffer (50 mM Tris-HCl, pH 8.0, 120 mM NaCl, 100 mM NaF, 1% Nonidet P-40, 200 M sodium orthovanadate, 10 g/ml each aprotinin and leupeptin, and 1 mM phenylmethylsulfonyl fluoride) were boiled in Laemmli sample buffer for 5 min and then electrophoresed in a 7.5% SDS-PAGE gel (0.5 mm thick and 5 cm long) containing 0.5 mg/ml MBP (Sigma). After fixing the gel with four changes of 20% 2-propanol in 50 mM Tris-HCl buffer, pH 8.0, for 2 h, SDS was removed by washing the gel twice for 2 h each in several gel volumes of 50 mM Tris-HCl, pH 8.0, containing 5 mM 2-mercaptoethanol. The MBP kinases were then re-denatured with 6 M guanidine HCl for 2 h and then renatured by 10 washes of 20 min each in several gel volumes of 50 mM Tris-HCl, pH 8.0, containing 0.04% Tween 40 and 5 mM 2-mercaptoethanol. After preincubation for 1 h with 5 ml of 40 mM HEPES, pH 8.0, containing 2 mM 2-mercaptoethanol and 10 mM MgCl 2 , phosphorylation of MBP within the gel was carried out by incubating the gel at room temperature for 1 h in 5 ml of 40 mM HEPES, pH 8.0, containing 25 Ci of [␥-32 P]ATP, 40 M ATP, 0.5 mM EGTA, and 10 mM MgCl 2 , and then washing the gel in 5% (w/v) trichloroacetic acid containing 1% sodium pyrophosphate several times until the radioactivity reached background levels.
Digestion with Endoglycosidase F and Tunicamycin Treatment-Cells were treated with 3 g/ml tunicamycin for 24 h before lysis. Cell lysates were denatured by boiling in 2% SDS-Laemmli sample loading buffer (24) for 5 min. For digestion with endoglycosidase F, 0.5 units of endoglycosidase F were added after the lysates were made 1% in Nonidet P-40 and 0.1% in SDS and boiled for 5 min. Denaturation by heating at 100°C in the presence of SDS, but not Nonidet P-40, increases the deglycosylation rate considerably according to the vendor. The lysates were incubated for 36 h before SDS-PAGE and Western blot analysis.
Preparation of Membrane Fraction-After washing with PBS, the cells were swollen for 5 min on ice in hypotonic buffer A (20 mM Tris-HCl, pH 7.5, containing 1 mM NaF, 100 M sodium orthovanadate, 2 mM EDTA, 1 mM EGTA) containing protease inhibitors exactly as described (25). After 20 -25 strokes with a Dounce homogenizer on ice, sucrose was added to a final concentration of 0.25 M, and the nuclei were pelleted at 1000 ϫ g for 5 min. The post-nuclear supernatant was layered on top of a 15% sucrose cushion in buffer A and centrifuged at 150,000 ϫ g for 1 h. The sedimentable membrane pellet was suspended in 1% Nonidet P-40 in buffer A, and any Nonidet P-40-insoluble material was removed at 10,000 ϫ g for 60 min.
Cell Surface Iodination and Immunoprecipitation of ␤1 Integrin-Cells were plated in 6-cm dishes so the majority of cells would be doublets 2 days post-plating, after doubling once. IODO-BEADS (Pierce) were washed twice in freshly prepared PBS, pH 6.5, and dried on filter paper just before using. Cells were rinsed with PBS, pH 6.5, twice, and then 1 ml of PBS, pH 6.5, two washed IODO-BEADS, and 10 l (1 mCi) of 125 I were added per dish. The cells were incubated for 20 min with rotation on an orbital shaker at room temperature. Cells were then washed 3 times with cold PBS and lysates prepared as above. Immunoprecipitations were performed using 5 g of monoclonal antibody clone P4C10 to 400 g of cell lysate exactly as described (22).
Tumorigenicity in Athymic Mice-Cells were collected by trypsinization and adjusted to 10 6 /ml, pelleted, resuspended in PBS at this volume, and 0.1 ml (10 5 cells) was injected in each of 5 male BALB/c nu/nu athymic mice between the shoulder blades. Cells were Ͼ99% viable by trypan blue exclusion. Tumor size was measured in two dimensions every 2-3 days using calipers, and volume was calculated as (width) 2 ϫ length/2. Sections were fixed in formalin, then processed for routine histology, and stained with either hematoxylin and eosin or Alcian blue with a nuclear fast red counterstain.

Transfectant Lines Express Mutated Ras Proteins-Transfec-
tions of HD6-4 cells were performed with a mini-gene construct of the cellular Ki-ras4B gene mutated at codon 12 to valine (Ki-ras G12V ), an expression plasmid encoding a mini-gene construct of the wild-type cellular Ki-ras4B gene, and a cellular Ha-ras G12V expression plasmid. We isolated three independent transfectant clones expressing the Ki-ras2 G12V oncogene: 4V, 4V1 and 4V2; three independent transfectant clones expressing the Ha-ras G12V oncogene: H15, H18, and H25; and several independent clones expressing the transfected wild-type Ki-ras4B gene. One of the latter, 4G1, was arbitrarily selected for a control for expression of excess copies of Ki-Ras protein.
The transfected oncogenes were expressed as active proteins. Proteins in lysates from the transfected lines were size-fractionated by SDS-PAGE and then transferred to PVDF membrane and analyzed for the presence of Ki-Ras proteins by immunoblotting with a Ki-ras-specific antibody (Fig. 1A, top  panel), analyzed for the presence of Ha-Ras proteins by immunoblotting with an Ha-ras specific antibody (Fig. 1A, second  panel), analyzed for the presence of total Ras proteins by immunoblotting with a pan-ras antibody (Fig. 1A, third panel), and analyzed for Ras proteins mutated at Val-12 by immunoblotting with a pan-ras Val-12 antibody (Fig. 1A, bottom panel). Cells transfected with mutant Ki-ras (4V1, 4V2, and 4V) expressed elevated levels of Ki-Ras protein, presumably mutant, compared with the other transfectant and parental lines (Fig.  1A, top panel). Cells transfected with mutant Ha-ras (H15, H18, and H25) expressed elevated levels of Ha-Ras protein, presumably mutant, whereas other transfectant lines and parental cells displayed lower levels of Ha-Ras protein (Fig. 1A,  second panel). All of the transfectants displayed similar levels of total Ras proteins (Fig. 1A, third panel) suggesting endogenous Ras proteins were down-regulated in the transfectant cells (Fig. 1A, third panel). Similar levels of total Ras proteins were also observed by immunoprecipitation experiments (data not shown).
Each of the three lines transfected with the Ki-ras4B G12V oncogene and each of the three lines transfected with the Haras G12V oncogene expressed mutant Ras proteins, which migrated at the expected mobility for proteins of 21 kDa and at the same position as recombinant mutant Ki-ras Val-12 (Fig. 1A). No Ras proteins mutated at valine 12 were detected in the parental line or in the Ki-ras4B wild-type control transfectant (Fig. 1A, bottom panel).
To confirm that the 21-kDa protein bearing the Val-12 mutation detected by immunoblotting was indeed ras, Ras proteins were immunoprecipitated from 35 S-prelabeled HD6-4 parental cells, wild-type c-Ki-ras 4G1 transfectant cells, and 4V transfectant cells expressing the mutated c-Ki-ras gene (Fig.  1B). Similar amounts of total Ras proteins were immunoprecipitated from each cell line (data not shown). Immunoprecipitated wild-type and mutant Ki-Ras proteins were transferred to PVDF membrane and analyzed for the presence of mutated Ras proteins by immunoblotting with the pan-ras Val-12 antibody. Only the 4V line expressed Ras proteins mutated to valine at codon 12 ( Fig. 1B), confirming the results of the Western blot analysis (Fig. 1A).
The mutant Ras proteins bound elevated levels of GTP showing they were functional. The ratio of GTP/GDP bound to immunoprecipitated Ras proteins was determined by thin layer chromatography (22). In duplicate experiments, the 4V line had a ras-bound GTP/GTP ϩ GDP ratio of 15.5%, parental cells a ratio of 8.5%, and wild-type c-Ki-ras transfectants a ratio of 7.0%. The presence of an activated ras was confirmed by a second method. Ras proteins exchange guanine nucleotides at a very low rate that can be measured by addition of [␣-32 P]GTP to permeabilized cells, followed by analysis of nucleotides bound to Ras proteins (23). In 4V cells over three times as much labeled GTP specifically bound to Ras proteins than in control 4G1 transfectant cells or in the parental line (data not shown). Thus mutated Ras proteins were expressed in both oncogenic Ki-ras and Ha-ras transfectant lines, and the immunoprecipitated oncogenic Ki-ras and Ha-Ras proteins exhibited increased GTP binding.
Oncogenic Ki-ras, Not Oncogenic Ha-ras, Prevents Colon Epithelial Cell Polarization into a Columnar Cell Shape-Parental cells, oncogenic Ha-ras transfectants H25, H18, and H15, and oncogenic Ki-ras transfectants 4V, 4V1, and 4V2 were each injected into five athymic mice. The relatively small inoculum of 10 5 cells was used to maximize any differences in tumor cell growth between the three lines. Tumor take was identical in each line, with 4 of 5 of each set of mice showing tumor growth. However, each ras transfectant line grew more rapidly in vivo than the parental line. The Ha-ras transfectants grew 2-4 times as quickly as the parental line, whereas the Ki-ras transfectants grew 4 -7-fold as rapidly (Fig. 2). Thus the presence of oncogenic Ras proteins in each transfectant line was correlated with increased cell growth in athymic mice.
The HD6-4 cell line organized into glands similar to normal colonic crypts seen in cross-section when cells were injected subcutaneously into athymic mice (21). Sections from tumors induced by the transfectant lines were examined after routine histology. Disorganized glands were seen in each tumor induced by each of three oncogenic Ki-ras transfectant cell lines ( Fig. 3D and data not shown), whereas the Ha-ras transfectants and wild-type Ki-ras transfectants displayed normalappearing glands (Fig. 3, B and C). These data suggested that the oncogenic Ki-ras gene disrupted either the tight cell to substratum adherence characteristic of HD6-4 cells or that cell to cell lateral adherence was disrupted.
Cells of each transfectant line were then cultured on transwells to determine whether disorganization seen in the tumors induced by the oncogenic c-Ki-ras gene was due to a loss of colon epithelial cell apicobasal polarity. Two weeks of postconfluent growth were necessary to induce full differentiation of HD6-4 cells so that the extent of apicobasal polarity could be ascertained (20). The HD6-4 parental line, the wild-type c-Kiras transfectant line 4G1, and the oncogenic Ha-ras transfectant H18 grew as monolayers when grown in vitro on a collagen I-coated porous polycarbonate transwell (Fig. 3, E-G, similar data obtained with two other Ha-ras transfectants in duplicate experiments not shown). The cells were fed by media through the basal surface as occurs in vivo. The cells formed lateral attachments and polarized so that their nuclei were found at the basal end of the cell and mucin granule formation was restricted to the apical portion. In contrast, 4V cells expressing mutated c-Ki-ras genes did not form a monolayer but formed a multilayer (Fig. 3H). Nuclei were not found polarized to the basal region of the 4V cell nearest the transwell membrane but

The Differential Effect of Oncogenic Ki-Ras and Ha-Ras Proteins on Colon Epithelial Cell Polarity Is Not Due to Differences in Activation of MAP Kinases-Introduction of either mutated
Ha-ras or Ki-ras into PC12 cells or NIH3T3 cells leads to activation of the MAP kinases erk1 and erk2 through sequential activation of raf and MEK (26). The MAP kinases erk1 and erk2 are activated by phosphorylation on tyrosine and threonine residues in the motif T*EY*. Antibody specific for the tyrosine-phosphorylated forms of erk1 and erk2 was utilized in Western blot analysis of cell lysates to demonstrate the presence of constitutively activated erk1 and erk2 in each of the three oncogenic Ha-ras and each of the three oncogenic Ki-ras transfectant lines, whereas no activated MAP kinases were detected in the parental HD6-4 line (Fig. 4 top). To confirm that the tyrosine-phosphorylated MAP kinases in the Ki-ras and Ha-ras transfectant cells were active kinases, the proteins in cell lysates were size-fractionated by SDS-PAGE in a gel containing immobilized myelin basic protein, a MAP kinase substrate. After removal of the SDS and renaturation of the lysate proteins, radiolabeled ATP was added and an in gel kinase assay was performed. Increased myelin basic kinase activity was detected in each of the Ki-ras and Ha-ras cell lysates at the 42/44-kDa position, the molecular masses of erk1 and erk2 (Fig. 4 bottom). Western blot analysis of the lysates with an antibody that detects both phosphorylated and unphosphorylated forms of erk1 and erk2 showed that the transfectant and parental cell lysates contained equal levels of these MAP kinases (Fig. 4, middle). Thus the presence of tyrosine-phosphorylated MAP kinases and active MBP kinases of 42/44 kDa confirm that the both oncogenic Ki-ras and Ha-ras transfectants expressed activated MAP kinases, with greater activity than those in the parental line (Fig. 4, bottom).
In earlier studies we had determined that colon carcinoma cell lysates that exhibited 42/44-kDa proteins with kinase activity on myelin basic protein also contained tyrosine-phosphorylated erk1 and erk2, as shown by immunoprecipitation of these kinases (22). Thus the increased cell growth rate detected in each transfectant line in vivo (Fig. 2) and in vitro (data not shown) was consistent with an activated ras-MAP kinase path-way. However, differences in Ki-ras and Ha-ras effects on cell polarization were not caused by differences in MAP kinase activation as MAP kinase activation was seen in each type of transfectant.
Oncogenic Ki-ras, Not Oncogenic Ha-ras, Interrupts ␤1 Integrin Maturation-The first step in HD6-4 cell polarization and differentiation is adherence to collagen I (20). HD6-4 cells bind much tighter to collagen I than undifferentiated colon carcinoma cells (19). In these cells collagen binding is mediated by the ␣2␤1 integrin heterodimer (19). Both undifferentiated colon carcinoma cells with poor collagen I binding (19) and the oncogenic Ki-ras transfectants displayed a disorganized monolayer of unpolarized cells when cultured on transwells. These data suggested that the Ki-ras transfectants might have abnormal ␣2␤1 integrin collagen receptors. The establishment of cell polarity is controlled by both cell to cell and cell to substratum interactions (17). Orientation of the apical-basal axis in MDCK cells is known to depend on integrin-mediated interactions with the growth surface. When grown in suspension MDCK cells aggregate to form cysts with the apical domain facing outwards. These cysts reverse their polarity when collagen is added to the culture medium causing apical markers to disappear from the external domain and reappear on the luminal face (27). Collagen interaction with integrins on the basal surface of cells is believed to orient the apical-basal axis, since addition of blocking antibody to ␤1 integrin prevents this polarity reversal (28). Because of these data, our studies concentrated on ␤1 integrin.
Integrin ␤1-chain is present in cells in several forms as follows: partially glycosylated precursors of 105-115 kDa and a more fully glycosylated mature form of lower electrophoretic mobility, about 130 kDa in fibroblasts, both of which do not migrate as sharp bands because of variations in their glycosylation (29). N-Glycosylation of both the ␣ and ␤ subunits of the integrin receptor in the Golgi is essential for the association of the heterodimer and for its optimal function (30). Under reducing conditions the 105-115-kDa integrin ␤1-chain precursors in the HD6-4 colon epithelial cells formed a large, tight band of 115 kDa, and the mature integrin ␤1-chain unexpectedly migrated as a doublet of roughly 140 -145 kDa (Fig. 5A, arrow). HD6-4 cells can differentiate into a colon goblet-like cell (Fig.  3E) that secretes copious amounts of mucins. Possibly posttranslational modifications other than those typical of mesenchymal cells occur in this differentiated epithelial cell, and these modifications result in a mature integrin ␤1-chain with lower electrophoretic mobility.
Each of the three c-Ha-ras transfectant lines (Fig. 5A) and the wild-type Ki-ras transfectant 4G1 line (Fig. 5B, arrow) also exhibited the mature integrin ␤1-chain doublet and the more abundant precursor. In sharp distinction, the mature integrin ␤1-chain doublet was absent in each of the three c-Ki-ras G12V transfectants (Fig. 5, A and B). In its place was an integrin ␤1-chain species with migration intermediate between the 115-kDa precursor and the 145-kDa mature doublet. We postulated that integrin ␤1-chain maturation was impaired in the oncogenic c-Ki-ras transfectants.
Under standard growth conditions immature integrin ␤1chain molecules are not transported to the cell surface but remain in the Golgi. Prolonged treatment of fibroblasts with 1-deoxymannojirimycin, a mannose analogue that specifically inhibits Golgi ␣-mannosidase I, resulted in the appearance of immature integrin ␤1-chains on the cell surface (31). Similar observations have also been been made with keratinocytes (32). Membrane fractions of parental HD6-4 cells and Ha-ras transfectant H18 cells exhibited an enrichment of the mature forms of integrin ␤1-chains, as expected since the mature forms are transported to the cell surface (Fig. 5C). Membrane fractions of the oncogenic Ki-ras transfectant contained the abnormal integrin ␤1-chain form (Fig. 5C), suggesting that the abnormal ␤1-chain was transported to the cell surface where it formed a heterodimer with an ␣-chain and functioned as a matrix receptor.
To confirm that the aberrant integrin ␤1-chain was transported to the cell surface, HD6 and HD6-4V cells were surfacelabeled with 125 I, and integrin ␤1-chain species were immunoprecipitated and then analyzed by SDS-PAGE and autoradiography. The ␤1-chain intermediate form was iodinated, proving its location on the cell surface (Fig. 6, left). A protein that might be an integrin ␣2-chain was co-immunoprecipitated with integrin ␤1-chain in both parental and oncogenic Ki-ras transfectant 4V cells but appeared in slightly less abundance in the 4V cells. To confirm the two large bands immunoprecipitated by ␤1-chain integrin antibody in HD6-4 parental cells and 4V cells, respectively, were the ␤1 mature form and the aberrant form, the same protein blot of the integrin ␤1chain immunoprecipitates containing both surface-iodinated ␤1 integrin and the much larger pool of "cold" cytoplasmic integrin was analyzed by Western blotting with anti-␤1 integrin antibody and chemiluminescence detection. Both iodinated immunoprecipitated bands superimposed upon the total ␤1 integrin forms (Fig. 6, right). Thus the integrin ␤1-chain faster migrating form was transported to the cell surface where it could be iodinated and where it could participate in heterodimer complexes with ␣-chains.
Aberrant ␤1 Integrin Form Seen in Ki-ras Transfectants Due to Impaired N-Glycosylation-To ascertain whether the aberrant ␤1 form was due to blocked N-glycosylation, lysates of HD6-4 parental cells and 4V cells were digested with endoglycosidase F before Western blot analysis for ␤1 integrin, yielding the same pattern, one lower electrophoretic mobility species of approximately 88 kDa (Fig. 7, left). Similar data were obtained with the 4V2 line (data not shown). Endoglycosidase F digests N-linked oligosaccharides of glycoproteins, cleaving high mannose structures, hybrid structures, and biantennary complex structures. In less complete digestions, the mature ␤1 integrin forms in the 4G1 control transfectant, the H18 Ha-ras transfectant, and the parental line were digested to a form roughly the same size as the unusual ␤1 integrin found in 4V2 cells (data not shown) suggesting, but not proving, that the aberrant ␤1 integrin was an immature form. Further studies are in progress to determine the biochemical structure of the aberrant ␤1 integrin form. Abolition of N-glycosylation by tunicamycin pretreatment of cells before lysis and Western analysis also demonstrated that the ␤1 integrin core protein was identical in both parental cells and oncogenic Ki-ras transfectant 4V cells (Fig. 7B). Thus the aberrant ␤1 integrin found in oncogenic Ki-ras transfectant cells was the result of impaired N-glycosylation and thus impaired maturation.
Impaired ␤1 Integrin N-Glycosylation Shown to Be Caused by Oncogenic Ki-ras Gene by Use of Gene-specific Antisense Oligonucleotides-Blocking transcription of the transfected oncogenic Ki-ras gene in 4V2 cells partially reversed accumulation of the ␤1 integrin intermediate. Chen and colleagues (13) used phosphorothiolated antisense oligonucleotides directed to the 5Ј-untranslated region of the Ki-ras gene to demonstrate the dominant role for c-Ki-Ras proteins in controlling the proliferation of diploid human fibroblasts. A 4-h treatment with oligonucleotides (see "Experimental Procedures") was followed by a 48-h chase to allow turn-over of the endogenous Ki-ras mRNA and protein (13). We used the same conditions to treat 4V2 cells. 4V2 cells treated with antisense oligonucleotides integrin form (double arrows). An integrin chain is co-immunoprecipitated (arrow), possibly ␣2 which is expressed in these cells (19). Right panel, same blot probed by Western analysis for total integrin ␤1-chain abundance, indicated by arrows at right. ϩ lanes, immunoprecipitations with anti-␤1 integrin monoclonal antibody P4C10; Ϫ lanes, control immunoprecipitations with isotype control monoclonal MOPC-21.

FIG. 7. Western blot analyses showing the faster migrating ␤1
integrin forms in mutated c-Ki-ras expressing 4V cells are due to aberrant glycosylation. A, the aberrant ␤1 integrin forms and mature ␤1 integrin forms are digested to same mobility species by endoglycosidase F (arrow). B, cells treated with tunicamycin before Western blot analysis contain the same molecular weight integrin ␤1chain core protein (arrow). Ϫ, untreated; ϩ treated. directed to the Ki-ras gene (A) exhibited almost no mutated Ki-Ras protein (Fig. 8A) compared with 4V2 cells treated with random sequence oligonucleotides (R) or untreated 4V2 (C) or parental cells. In each of two experiments performed in duplicate a slower migrating integrin ␤1-chain form was seen in 4V2 cells treated with antisense oligonucleotides directed to the Ki-ras gene (A) compared with 4V2 cells treated with random sequence oligonucleotides (R) (Fig. 8B). The abnormally glycosylated integrin ␤1-chain seen in untreated cells (C lane) was not completely replaced by the mature form, perhaps because the long half-life of the abnormally glycosylated integrin ␤1chain. However, we have shown that specifically inhibiting c-Ki-ras gene transcription by antisense techniques increases the degree of maturation of integrin ␤1-chain. Thus the impairment in ␤1 integrin maturation seen in each of three independently cloned transfectant lines is not due to some extraneous event in the establishment of these lines but is directly due to the functioning of the oncogenic c-Ki-ras gene.
Aberrantly N-Glycosylated Integrin ␤1-Chain Mediates Cell Adherence Poorly-HD6-4 cells tightly adhere to collagen Icoated transwells as a first step in establishing their apicobasal and lateral polarity. HD6-4 cells use integrin ␤1-chain to mediate binding to both collagen I and laminin as inclusion of 1 g/ml anti-␤1-chain monoclonal antibody to a cell adherence assay decreased binding to these substrates 42 and 39%, respectively ( Fig. 9, left). Binding to collagen I films was decreased by one-third in the oncogenic Ki-ras transfectant 4V cells, whereas binding to laminin films was decreased approximately 70% (Fig. 9, right). Both decreases were statistically significant (p Ͻ 0.001 by Student's two-tailed t test). In contrast, expression of oncogenic Ha-Ras proteins or more copies of wild-type Ki-Ras proteins did not alter binding of H18 or 4G1 cells, respectively, to either collagen I or laminin. Oncogenic Ki-Ras proteins block normal N-glycosylation of ␤1 integrin, perhaps at an intermediate stage. The abnormally glycosylated integrin ␤1-chain is transported to the cell surface where it mediates a looser attachment of cells to extracellular matrix components, preventing their normal apicobasal polarization and normal cell to cell adherence. DISCUSSION Integrins are a family of ␣␤ heterodimers that mediate cell adhesion to extracellular matrix proteins, plasma proteins, or to other cells (33). Nine different ␤-chains have been identified. The integrin ␤1-chains form 10 different ␣␤1 heterodimers that mediate binding to collagens, fibronectin, laminin, vitronectin, and VCAM-1, a cell adhesion molecule. Coordinate control between maturation of ␣ and ␤1 integrin chains has been indicated by several studies. In mesenchymal cells iodination of surface proteins followed by immunoprecipitation of either a specific ␣ integrin or ␤1 integrin has demonstrated that the mature forms of ␣2, ␣3, ␣5, or ␣6 integrin chains co-immunoprecipitate with mature ␤1 integrin chains (29,34). Furthermore, an increase in conversion of ␤1 integrin precursor to the mature form led to an increase in the assembly of ␤1 with mature ␣, showing coordinate regulation of maturation. The immature integrin chains ␤1 and ␣6 each transiently associate with the chaperone calnexin prior to assembly of the mature integrin chains to heterodimers in human kidney epithelial 293 cells (35) or human epidermal keratinocytes (32).
In the current study we observed blocked maturation of ␤1 integrin in each of three independently cloned oncogenic Ki-ras transfectants and in none of three independently cloned oncogenic Ha-ras transfectants. Blocking expression of the Ki-ras gene by a transient 4-h transfection of an antisense oligonucleotide partially restored the normal maturation pattern, demonstrating that expression of the oncogenic Ki-ras gene inhibited ␤1 integrin glycosylation. The faster-migrating ␤1 integrin forms seen in oncogenic Ki-ras transfectants and the mature ␤1 integrin species found in control cells yielded products of the same size when digested with endoglycosidase F or when cells were pretreated with the N-glycosylation inhibitor tunicamycin. These results suggested that oncogenic Ki-ras blocked post-translation N-glycosylation of ␤1 integrin at some intermediate point in maturation.
Most of the immature ␤1 integrin molecules remain in the endoplasmic reticulum. After translocation to the cis-Golgi complex, immature forms of integrin ␤1-chains may be processed by multiple glycosyltransferases including branching enzymes, such as GlcNAc-T III, IV, and V, and elongating enzymes, such as ␤1,4-galactosyltransferases and ␤1,3-Glc-NAc-T. Twelve potential N-glucosylation sites (Asn-X-(Ser/Thr) sequences) occur in ␤1 integrin. N-Glycosylation of both the ␣ and ␤ subunits of the integrin receptor in the Golgi is essential for the association of the heterodimer and for its optimal function (30). Removal of GlcNAc residues from purified integrin receptor or from K562 cells bearing such receptors dissociated the heterodimer, preventing co-immunoprecipitation of ␣ and ␤1 integrin chains. The N-glycosylation of precursor ␤1-integrin chains, in particular the addition of GlcNAc residues, is thus an essential component of their maturation. Partial elimination of N-linked structures located at the surface domain of integrin ␤1 abolished adhesive function, mainly by dissociating the heterodimer, suggesting that only a few carbohydrate chains are essential (30). During suspension-induced terminal differentiation of keratinocytes, the maturation of the ␤1 subunit and the associated ␣ subunits was prevented by blocking the N-linked glycosylation of both integrin chains (36). We do not know the biochemical nature of the glycosylation step altered by expression of oncogenic Ki-ras in HD6-4 cells. However, N-acetylglucosaminyltransferase III (GlcNAc-T III) catalyzes the addition of a bisecting GlcNAc structure that inhibits further processing of oligosaccharides by other glycosyltransferases (37). Thus activation of GlcNAc-T III or a similar transferase could block ␤1 integrin processing. Analysis of the biochemical nature of the ␤1 integrin form found in Ki-ras transfectants should allow us to define the pathways altered by expression of this oncogene.
Oncogenic forms of ras genes have been implicated in modulating various Golgi acetylglucosaminyltransferases in different cell types. ␤1-6 N-acetylglucosaminyltransferase V (Glc-NAc-T V) is increased severalfold in activity in rodent fibroblast lines transfected with the oncogenic T24 Ha-ras gene (38). Expression of transfected GlcNAc-T V in MvlLu lung epithelial cells led to an acquisition of tumorigenicity and an increase in the apparent molecular weights of ␣5,␣v, and ␤1 integrins, indicating that the oligosaccharides on these integrins are substrates for GlcNAc-T V (39). NIH3T3 cells expressing the N-ras proto-oncogene exhibited larger cell surface asparagine-linked glycans. Possibly this increase was caused by 5-7-fold increases in the activities of the levels of elongating ␤1-galactosyltransferase and ␤3-N-acetylglucosaminyltransferase, both of which synthesize polylactosaminoglycan chains (40). R-ras is a GTP-binding protein highly homologous to Ha-ras but, unlike Ha-ras, induces neither cell proliferation nor differentiation. Transfection of a constitutively active R-ras into myeloid cells grown in suspension activated integrins by a yet unknown mechanism, increasing cell adhesiveness (41). Thus H, N, and R ras genes either modify the activity of various acetylglucosaminyltransferases which could alter ␤1 integrin maturation in various cell types or, in the case of R ras in myeloid cells, alter integrin ␤1-chain function directly. Cell type specificity in the action of different ras isoforms is also deduced from the following studies. Viral Ki-ras prevents the apical polarization of MDCK kidney epithelial cells (15) by decreasing expression of ␤1 integrin precursor, whereas no decrease in the abundance of mature integrin ␤1-chain was seen (16). No effects on glycosylation of integrin ␤1-chain were observed in contrast to the current study with colon epithelial cells, and an aberrant ␤1 integrin intermediate was not observed in the v-Ki-ras-transformed MDCK cells (16). Transfection of a different ras gene, oncogenic Ha-ras, into another epithelial cell type, CACO2 colon cancer cells, did not lead to loss of growth control or to loss of cell polarization but led instead to induction of multiple markers of normal intestinal brush border differentiation and the capacity for terminal differentiation (42). These studies support our observations in this study that expression of a mutated cellular Ha-ras gene in differentiated colon epithelial cells did not alter cell polarity.
Studies by other investigators have implicated loss of the ␣2 integrin chain in colon cancer progression. Low expression of ␣2␤1 integrin collagen receptors has been found in the more aggressive histological subtypes of colon cancers (43). A second group also reported that reduced expression of the ␣2 integrin subunit was commonly found in both locally invasive and in metastatic colon cancers (44). The decrease in ␣2 integrin observed in these colon cancers may be caused by impaired maturation of ␤1 integrin, since maturation of ␣ integrin is regulated in part by the size of the intracellular ␤1 integrin precursor pool (45). Expression of ␤1-integrin antisense RNA decreased synthesis of the ␤1-integrin, decreasing the size of the precursor pool. The ␤1 integrin precursors that were synthesized exhibited accelerated maturation, whereas the ␣ integrin precursors remained for longer times in the Golgi, yielding fewer mature ␣ integrin molecules. Therefore, the impaired ␤1 integrin maturation we observed in this study could contribute to the decreased ␣2 integrin expression seen by others in colon cancers.
Additional clinical evidence has linked aberrant ␤1 integrin processing to colon cancer progression. Alteration of the ratio of the ␤1 integrin precursor to the mature form of ␤1 integrin is found in invasive and metastatic colorectal cancers (46). Fifteen of 19 colorectal cases (79%) with alteration of ␤1 integrin processing showed lymph node metastases compared with 12 of 32 (38%) without detectable alterations in ␤1 integrin processing (p Ͻ 0.01). Cancer invasion beyond the muscularis propria was observed in all 19 cases with alterations in the ratio of ␤1 integrin precursor to mature form. The mechanism for the alteration in ␤1 integrin protein processing in colorectal cancer is not known. We suggest, from the results of the current study, that the mutation in the cellular Ki-ras gene found in a large subset of colon cancers may prevent the normal processing of ␤1 integrin precursor and lead to characteristic areas of cell multilayering or dysplasia. Expression of the oncogenic Ki-ras gene in benign colon tumors is also seen within areas of dysplasia where cells form multilayers (47). In the current study, the oncogenic Ki-ras HD6-4 cell transfectants also formed multilayers when grown in transwells (Fig. 3). Mutations of both Ki-ras and Ha-ras genes may occur randomly within the colon. However, the specificity of the effects of oncogenic Ki-ras on colon epithelial cell polarization through impairment of ␤1 integrin maturation may lead to selection of cells bearing Kiras mutations. Selection of colon adenoma cells with oncogenic Ki-ras mutations is seen during the growth of benign tumors (47), and this selection is maintained in the transition to malignancy and during further progression of the cancer cells to highly invasive and metastatic states (48,49). Tumor cells freed from tight cell to cell and cell to substratum adherence because of the synthesis of aberrant ␤1 integrin could have a selective advantage in both growth and invasion and become the dominant cell type within a tumor. It has been observed that cells with oncogenic Ki-ras mutations form as foci within benign tumors but rapidly become the sole tumor cell type when benign tumors progress to malignancy (47).