Expression of a Splice Variant of KAI1, a Tumor Metastasis Suppressor Gene, Influences Tumor Invasion and Progression¹

KAI1 (CD82) is a transmembrane glycoprotein that is a member of the TM4SF³ (1). About 20 members of this family have been defined, including MRP-1/CD9, TAPA-1/CD81, ME91/CD63, and KAI1/CD82. TM4SF proteins share 20–30% sequence similarity and contain four highly conserved transmembrane domains, two short cytoplasmic domains at the NH₂ and COOH termini and two relatively divergent extracellular domains, the larger of which contains several conserved amino acid motifs. The precise biochemical function of TM4SF proteins is not yet clear. It has been hypothesized that TM4SF proteins may be transmembrane adaptor proteins that organize the distribution and function of other cell surface molecules and their associated signaling proteins. Members of this superfamily may have a significant role in the regulation of cell proliferation, activation, and motility through modulation of cell adhesion (2, 3, 4). Some TM4SF proteins may be particularly relevant to tumor metastasis. For example, ectopic expression of CD9, CD63, or CD82/KAI1 in tumor cells suppressed their metastatic potential in animal model systems (1). Furthermore, elevated expression of CD9 and CD82/KAI1 were linked to a higher survival rate for various human cancers (5). Other studies have shown that members of the TM4SF are implicated in the assembly of integrin-containing signaling complexes, thus modulating the function of integrin receptors in cell migration (4, 6, 7).

In recent years, it was reported that a dramatically reduced level of KAI1 expression is one of the characteristics of the invasive and metastatic stages of many human cancers (5, 8, 9, 10, 11). In addition, the expression of KAI1 resulted in reduced cell motility and invasiveness in vitro in colon cancer cells and in vivo in melanoma cells (12, 13). These findings, taken together with the fact that the KAI1 protein has an almost ubiquitous tissue distribution (1), suggest that KAI1 functions as a metastasis suppressor gene in many cancers.

In this study, we found a splice variant of KAI1 during a screen for the expression of various genes involved in the invasion and metastasis of gastric cancer. No alternately spliced transcript of the KAI1 gene has been described previously. The spliced-KAI1 was lacking the distal part of the second extracellular loop and part of the fourth transmembrane region. This structural difference conferred increased metastatic potential both in vitro and in vivo. Our findings suggest that expression of spliced-KAI1 may be a useful molecular marker for poor prognosis in cancer patients.

Thirty-one gastric cancer and normal adjacent mucosal tissues were obtained from surgically resected specimens at Chonnam University Hospital (Kwangju, Korea). We also collected 31 normal and metastatic lymph nodes and 5 peritoneal and 6 hepatic metastases. The tumors were histologically examined and pathological stage was also estimated by tumor-node-metastasis score (stage IIIa, 3; stage IIIb, 1; and stage IV, 27). The Ethics Committee of Chonnam University Hospital approved our experimental protocols.

Total RNAs were prepared from normal and tumor tissues or cultured cells. Total RNA (400 ng) was reverse transcribed with random primers and Moloney murine leukemia virus by the RT-PCR kit (Invitrogen, Carlsbad, CA). PCR amplification of cDNA was performed using RT-PCR with two specific KAI1 primer sets. Primer set I was designed to recognize whole coding region (801 bp) of the KAI1 gene (sense, 5′-GGAATTCGATGGGCTCAGCCTGTATCAAAGTCA-3′ and antisense, 5′-GCTCTAGA-TCAGTACTTGGGGACCTTGCTGT-3′). Primer set II was designed to recognize the spliced region (exon 7) of the KAI1 cDNA (sense, 5′-GACAACAGCCTTTCTGTGAGGAAG-3′ and antisense, 5′-GCTCTAGATCAGTACTTGGGGACCTTGCTGTA-3′). PCR consisted of 30 cycles of the following conditions: denaturing for 30 s at 94°C; annealing for 60 s at 60°C (primer set I) or for 30 s at 60°C (primer set II); and extension for 60 s at 72°C. Endogenous spliced-KAI1 mRNA expression levels were investigated by nested RT-PCR using both primer sets. The PCR products were separated by electrophoresis on a 1.5% agarose gel containing ethidium bromide.

To clone the full-length KAI1 and spliced-KAI1 cDNA into the mammalian expression vector pcDNA3 (Invitrogen), each KAI1 cDNA was prepared by RT-PCR using primer set I and gastric cancer mucosa. The resulting 801- and 717-bp PCR products were digested with EcoRI and XbaI and subcloned into the pcDNA vector. pEGFP-wild-KAI1 or pEGFP-spliced-KAI1 plasmid was made by inserting the above KAI1 cDNA into a COOH-terminal enhanced fluorescent protein vector (pEGFP-C1; Clontech, Palo Alto, CA) with EcoRI and SmaI. Each construct was confirmed by sequencing.

The mouse colon adenocarcinoma cell line (CT-26) was grown in DMEM (Invitrogen) supplemented with 10% fetal bovine serum (Sigma, St. Louis, MO) in a humidified atmosphere of 5% CO₂ at 37°C. The cDNA (KAI1/pcDNA) of wild-type KAI1 or spliced-KAI1 was transfected into CT-26 cells. Transfections were performed using the FuGENE 6 transfection reagent (Roche, Indianapolis, IN) as described previously (14). Cells were detached by 0.25% trypsin-EDTA solution (Hyclone, Logan, Utah) and passaged into the selective media containing 500 μg/ml Geneticin (Life Technologies, Inc., Grand Island, NY). Two weeks later, surviving clones were analyzed by Western blot analysis for their expression of KAI1 protein.

RNA samples were quantitated by spectrophotometry at 260 nm. For Northern analysis, total RNA (10 μg) was denatured with glyoxal, separated by size on 1.0% agarose gels, and transferred to Genescreen (DuPont). Probe (KAI1, in-frame full-length cDNA, 801 bp) was radiolabeled by a nick translation method, and hybridization and signal visualizations were performed as described previously (14). The integrity of the RNA sample was established by Northern analysis with a rat glyceraldehyde-3-phosphate dehydrogenase probe.

For Western blot analysis, cell or tissue proteins were solubilized in NP40 lysis buffer containing a protease inhibitor mixture (Roche) and 1 mm phenylmethylsulfonyl fluoride. The resolved proteins (50 μg) were transferred to a nitrocellulose membrane, and blotted with H-173 or C-16 (a rabbit polyclonal antibody against KAI1, Santa Cruz Biotechnology, Santa Cruz, CA) or B-L2 (a monoclonal antibody against KAI1, Serotec, Oxford, United Kingdom), and antirabbit or antimouse immunoglobulin-HRP (Amersham, Arlington Heights, IL) as described (14). The blot was reprobed with antiactin antibody (I-19, Santa Cruz Biotechnology) to control for loading.

Proliferation and viability of cells were measured using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide colorimetric assay (Sigma). Cells were seeded at 1 × 10⁴ cells/well in 24-well plates and grown for 8 days in the presence of serum. 3-(4,5-Dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide reagent was added for 2 h, after which, DMSO was added. The absorbance at 570 nm was determined 16 h later using a microplate reader with SOFTmax PRO software (Molecular Devices, Sunnyvale, CA).

Human plasma fibronectin (Calbiochem, La Jolla, CA) and collagen type IV (Sigma) were diluted with sterile PBS, and 10–50 μg were added to each well in a 96-well plate to coat the bottom. The plates were left for 24 h to allow fibronectin bind to the well surface. The excess liquid was then removed, and the remaining sites were blocked by incubation with 250 μl of 1.5% BSA in PBS for at least 2 h at 37°C. As a control for nonspecific binding, a parallel set of wells without fibronectin was blocked with BSA using identical conditions. Cells were detached from the culture flasks with 5 mm EDTA in PBS, resuspended in culture medium containing 0.02% BSA to 4 × 10⁵ cells/ml, and then 100 μl of cell suspension were added to each well. All cells were assessed in quadruplicate. After incubation for 1 h at 37°C with 5% CO₂, the supernatant from each well was removed, and warm PBS (100 μl) was then used to gently wash the wells three times. Crystal violet (0.04% in water; 100 μl) was then added. After 10 min, the crystal violet was removed, and the stained cells were gently rinsed with PBS. Adherent cells were then counted in three random areas of each well, using the inverted phase contrast microscope to determine the average number of cells/field of view. The results presented are the overall mean ± SE of at least three independent experiments.

The stained cells were counted in six random squares of 0.5 × 0.5 mm of microscope fields for each filter. The results were expressed as the mean ± SE of the number of cells/field.

Cell lysates were prepared using a radioimmunoprecipitation assay buffer containing protease inhibitors. Immune complexes were collected onto protein A/G-agarose beads (Pierce, Rockford, IL) that were prebound with antiKAI1 antibody, H-173 (Santa Cruz Biotechnology), followed by three washes with radioimmunoprecipitation assay buffer. Immune complexes were eluted from beads with 2× SDS sample buffer and resolved by 8% SDS-PAGE under reducing conditions. Proteins were electrophoretically transferred to nitrocellulose membrane (Amersham). After blocking with 5% nonfat milk in Tris-buffered saline with (TBST)-Tween 20 buffer at room temperature for 2 h, nitrocellulose membranes were incubated at room temperature for 1 h with a specific antibody against integrin α₃β₁ (VLA-3; Chemicon, Temecula, CA) and then treated with horseradish peroxidase-conjugated goat antirabbit IgG, washed with TBST-Tween 20 buffer, and developed.

Cell migration was measured using a Transwell migration apparatus (Costar, Inc., Cambridge, United Kingdom) with minor modification. The filters (8-μm pore size) were coated with 1% gelatin solution on both top and bottom surfaces. Cells were harvested, washed once in serum-free DMEM/0.2% BSA, and resuspended at 2 × 10⁶ cells/ml DMEM/0.2% BSA. To start the assay, 120 μl (2.4 × 10⁵ cells) were loaded to the upper chamber of the Transwell, whereas 400 μl DMEM/0.2% BSA containing 10 μg/ml human plasma fibronectin (Calbiochem) were loaded into the lower chamber. The Transwell apparatus was incubated for 24 h at 37°C. At the end of the incubation, the cells attached to the membranes were fixed and stained with Diff-Quick (International Reagents, Kobe, Japan) following the manufacturer’s protocol. The cells on the top surface of the filters were wiped off with cotton balls, and the migrated cells on the bottom surface were counted in six random squares of 0.5 × 0.5 mm of microscope fields for each filter. The results were expressed as the mean ± SE of the number of cells/field.

KM12C, KM1214, and A431 cells were seeded onto an 8-well Lab-Tek Chamber Slide Glass (Nunc, Scotts Valley, CA) and were grown in DMEM supplemented with 10% FBS (Life Technologies, Inc.). The pEGFP-wild-KAI1 and pEGFP-spliced-KAI1 plasmids were transfected into cells using FuGENE 6.

Cells were rinsed with PBS 3 times and fixed with ice-cold 2% buffered paraformaldehyde (pH 7.4) in PBS for 10 min. After washing with PBS and blocking with a buffer containing 0.1% saponin and 0.05% BSA in PBS (pH 7.4) for 30 min, the cells were incubated with anti-E-cadherin antibody (dilution 1:250; BD Biosciences, Franklin Lakes, NJ) for 1 h at room temperature and then washed with the blocking buffer. Tetramethylrhodamine-labeled antimouse IgG antibody (dilution 1:250; BD Biosciences) was added to the cells, and they were incubated for 1 h. After washing with PBS three times, the cells were examined with a Laser Scanning Confocal Microscope (Leica Microsystems TCS NT, Leica, Germany).

Prior approval of the experimental protocol was obtained from the Chonnam National University Medical School Research Institutional Animal Care and Use Committee. Maintenance of animals and all in vivo experiments were performed according to the Guiding Principles in the Care and Use of Animals. Subconfluent CT-26 cells (CT-26/wild-KAI1 or CT-26/spliced-KAI1) were trypsinized and then suspended in DMEM. The cell suspension (5 × 10⁶ cells in 0.1 ml medium/mouse) was injected s.c. into the BALB/c syngeneic mice (n = 10 for each group of CT-26, CT-26/wild-KAI1, and CT-26/spliced-KAI1). Tumor size was measured daily with a vernier caliper from the first week to the fifth week after injection. Tumor volume (V) was calculated using the values of the largest (A) and the smallest (B) diameter according to the formula: V = 0.5 × AB². At the sixth week after injection, the presence of metastasis in the liver and lung tissues in each mouse bearing a tumor mass on the back were evaluated by gross and microscopic examination. After confirming the metastasis, tumor mass and metastatic foci in the lung tissues and liver tissues were excised (n = 5 for each group of CT-26/wild-KAI1 and CT-26/spliced-KAI1), and proteins were prepared for KAI1 analysis in the each tissue.

Experimental differences were tested for statistical significance using ANOVA and Students’ t test. P of <0.05 was considered to be significant.

A splice variant of KAI1 (spliced-KAI1) was identified during the study of the expression of various genes involved in invasion and metastasis. RT-PCR with a KAI1 primer set flanking the whole encoding region of the gene produced two amplification products, 801 and 717 bp (Fig. 1,A). The subcloning and sequencing analyses of 801-bp product showed that it corresponded to the wild-type KAI1 cDNA, whereas the 717-bp product was a KAI1 cDNA sequence that lacked 84 bp in the COOH-terminal region. This variant was named spliced-KAI1. In gastric cancer specimens in which metastasis extended to the peritoneum, expression of spliced-KAI1 mRNA was present in the primary tumor tissue, but absent in the normal adjacent mucosa. The expression of spliced-KAI1 was also present in metastatic lymph nodes and distant peritoneal tissue (Fig. 1,A, no. 1). This expression pattern was the same in the gastric cancer specimens with regional lymph node metastasis (Fig. 1 A, no. 2).

Subsequent genomic DNA analysis revealed that this splice variant displayed selective deletion of exon 7, an 84-bp sequence (Fig. 1,B). Exon 7 consists of 28 amino acids spanning from the distal part of the second extracellular loop to the fourth transmembrane region (Fig. 1 C). Thus, the COOH-terminal region of spliced-KAI1, including the second extracellular loop, is different from that of wild-type KAI1.

The presence of spliced-KAI1 was confirmed by using a different primer set flanking the spliced exon (Fig. 1, D and E). Examination of this splice variant was carried out by RT-PCR in gastric cancer patients with various stages of metastasis (n = 31; Table 1). The 255- and 171-bp bands predicted for wild-type KAI1 and spliced-KAI1, respectively, were amplified, and the identity of the bands was confirmed by subcloning and sequencing. In the gastric cancer specimens from patients with a good prognosis with regional lymph node metastasis (Fig. 1,D, no. 1) or distant metastasis such as peritoneum (Fig. 1,D, no. 3), the expression of spliced-KAI1 in the tumor tissue and metastatic tissues was negligible. However, in the gastric cancer specimens from patients with regional lymph node metastasis and poor prognosis (Fig. 1,D, no. 2) or distant metastasis and poor prognosis (Fig. 1,D, no. 4), the expression of spliced-KAI1 in the tumor tissue and metastatic lymph node was higher than that in the normal adjacent mucosa (Fig. 1 D, nos. 2 and 4). Thus, the expression of spliced-KAI1 mRNA in the extraneoplastic mucosa and primary tumor tissue seemed to correlate with poor prognosis.

We also examined the expression of wild-type and spliced-KAI1 mRNA in normal and various cancer cell lines by RT-PCR (Fig. 1,E). In human colon carcinoma cell lines, spliced-KAI1 mRNA was higher in the KM1214 and KM12SM cells than in KM12C, which has a lesser metastatic potential. Although different levels of spliced-KAI1 mRNA were present in the various human cell lines, ECV304, immortalized human endothelial cells expressed barely detectable endogenous spliced-KAI1 mRNA (Fig. 1,E). In contrast, mouse colon adenocarcinoma cell line CT-26 showed a relatively low level of endogenous wild-type KAI1 mRNA (Fig. 2 A). For subsequent experiments, CT-26 cells were used to establish cell lines through stable transfection of wild-type KAI1 or spliced-KAI1 cDNA.

Patients who had distant metastatic tissues such as liver and peritoneum expressing spliced-KAI1 had significantly reduced survival after surgery compared with those with no spliced-KAI1 expression (7.0 ± 0.5 versus 30.8 ± 5.3 months; Table 1). Similarly, patients with regional metastatic lymph nodes expressing spliced-KAI1 had significantly reduced survival (12.1 ± 3.0 versus 30.7 ± 1.7 months; Table 1). When compared with corresponding stage of pooled patients from the past 10 years, the same pattern was evident in the survival periods of patients who had metastatic tissues expressing spliced-KAI1 (Table 1).

To analyze the role of spliced-KAI1 on the motility and metastatic processes of colon cancer cells, stably expressing cell lines (CT-26/wild-KAI1 and CT-26/spliced-KAI1) were established using CT-26 cells. Wild-type and spliced-KAI1 mRNA expression in resultant transfected clones were then analyzed by RT-PCR and Northern blot analyses. KAI1 expression was always higher in the transfected cells than in the CT-26 parent cells. The 258- and 174-bp bands corresponding to wild-type and spliced-KAI1, respectively, were expressed at various levels (Fig. 2,A, top). In Northern blot analysis, the wild-type KAI1 transfectants expressed KAI1 transcript (∼2.0 kb) more than control parent cells (Fig. 2 A, bottom). Also, the spliced-KAI1 transfectants expressed KAI1 transcript more than parent cells just as the wild-type KAI1 transfectants. However, high level of truncated KAI1 gene (<1.6 kb) was expressed in the spliced-KAI1 transfectants. It shows that transfected wild-type or spliced-KAI1 gene was expressed in the transfectant cells.

Western blot analysis showed that CT-26 parent cells expressed both kinds of KAI1 proteins, wild-type (∼50 kDa) and spliced-KAI1 (∼48 kDa; Fig. 2,B). In contrast to the RT-PCR result, the former was always the predominant form, suggesting the presence of a posttranscriptional modification process. CT-26/wild-KAI1 and CT-26/spliced-KAI1 cells expressed the appropriate transfected protein at a level higher than in the CT-26 parent cells when we used H-173 polyclonal antibody (Fig. 2,B, top). We observed that the similar detection of KAI1 was produced using B-L2 monoclonal antibody compared with the H-173 (Fig. 2,B, middle). However, C-16 polyclonal antibody did not detect the spliced-KAI1 protein, which had different COOH-terminal structure compared with wild-type KAI1 (Fig. 2 B, bottom). The C-16 antibody was raised against a peptide mapping at the COOH terminus of human KAI1, and C-16 antibody could not react with the spliced-KAI1 protein because epitope of C-16 was not present or modified in the spliced-KAI1 protein. ECV304 cells expressed a very low level of spliced-KAI1 protein, consistent with the RT-PCR result.

KAI1 interacts with integrin (6, 15), which plays an important role in the regulation of adhesion between the cell and the extracellular matrix (ECM). Integrins form adhesion contacts during invasive migration. Expression of integrin α₃β₁ in stably expressing cell lines (CT-26/wild- and spliced-KAI1) were analyzed by RT-PCR. Both transfected cell types had similar levels of integrin α₃β₁ expression (Fig. 2,C). The association of wild-type KAI1 and spliced-KAI1 with integrin α₃β₁ was examined in the transfected cells by immunoprecipitation. The association of spliced-KAI1 with integrin α₃β₁ was weaker than that of wild-type KAI1 (Fig. 2 D).

To detect any differences in localization between wild-type and spliced-KAI1 in cell-cell contact sites, we transiently expressed both fluorescent KAI1 proteins into KM12C and A431 cells, both of which express endogenous E-cadherin. We compared the localization of KAI1 with E-cadherin, a cell surface protein that is found at adherens junctions at the edges of cells (Ref. 16; Fig. 3,a). E-cadherin overlapped with fluorescent KAI1 at the edges of cells in the KM12C and A431 cells transiently transfected with wild-type KAI1 (Fig. 3,b, middle; Fig. 3,c, middle). In contrast, cells transiently transfected with spliced-KAI1, the staining patterns of E-cadherin and spliced-KAI1 did not overlap at the edges of cells, indicating that spliced-KAI1 was not present on the cell surface (Fig. 3,b, bottom; Fig. 3 c, bottom). These results indicate that the structural difference in spliced-KAI1 affects its functional characteristics.

CT-26/wild-KAI1 cells had a longer process (Fig. 4,a, middle) than did parental CT-26 cells (Fig. 4,a, top). In contrast, this morphological change was not seen in CT-26/spliced KAI1 cells, which had a short process like the parental cells (Fig. 4 a, bottom). This result indicated that wild-type KAI1 was able to affect the morphological shape of cells, whereas spliced KAI1 did not have this ability.

To investigate whether wild-type KAI1 affected the proliferation rate of CT-26 cells, we examined the cell proliferation of parent CT-26, CT-26/wild-KAI1, and CT-26/spliced-KAI1 cells for 7 days. The proliferation of CT-26/spliced-KAI1 cells was very similar to that of parent CT-26 cells (Fig. 4 b). However, CT-26/wild-KAI1 cells proliferated more slowly, as seen clearly from day 3. CT-26/wild-KAI1 cells exhibited maximal growth at 6 days, whereas CT-26/spliced-KAI1 and parent CT-26 cells exhibited maximal growth at 4 days. Thus, wild-type KAI1 suppressed cell growth, whereas spliced-KAI1 did not have this characteristic.

Tumor cells interact with ECM components and basement membranes, an essential initial event during the process of invasion and metastasis. To analyze the effect of KAI1 expression on the binding of tumor cells to fibronectin and type IV collagen, ECM proteins were examined by a cell attachment assay. In CT-26 cells, wild-type KAI1 expression decreased cell-ECM adhesion, including adhesion to fibronectin and type IV collagen (Fig. 5,a, middle). However, the expression of spliced-KAI1 had different effects on fibronectin and collagen IV adhesion (Fig. 5,a, bottom). Fibronectin exhibited a significantly greater effect on the binding of CT-26/spliced-KAI1 cells than on the CT-26 parental or CT-26/wild-KAI1 cells, whereas collagen exhibited no significant difference in the extent of the changes observed on the three cell types (Fig. 5,b). In contrast, CT-26/wild-KAI1 cells had significantly decreased adhesion to fibronectin compared with parent CT-26 cells (Fig. 5, a and b).

The invasiveness of tumor cells is one of several important properties necessary for metastasis and involves the loss of factors involved in the control of cell-cell and cell-ECM interactions. To analyze the differential effect of wild-type KAI1 and spliced-KAI1 on in vitro cell invasion, a cell invasion assay was carried out using the Transwell migration apparatus. KAI1 interacts with integrin α₃β₁, which is a multiple ligand receptor that binds laminin, fibronectin, and kalinin/epiligrin (6, 15). Thus, cell motility was estimated by using fibronectin and collagen IV as chemotactic factors. CT-26/spliced-KAI1 cells showed significantly increased in vitro motility and invasive potential induced by fibronectin and collagen IV (Fig. 6,a, bottom), compared with CT-26 parental or CT-26/wild-KAI1 cells (Fig. 6,a, top and middle; Fig. 6,b). Furthermore, CT-26/wild-KAI1 cells had lower invasive potential induced by fibronectin, compared with parent CT-26 cells (Fig. 6, a and b). In particular, CT-26/spliced-KAI1 cells showed significantly greater invasive ability when fibronectin was used as a ligand compared with collagen IV (Fig. 6 b).

To demonstrate whether spliced-KAI1 affects colon cancer progression differently from wild-type KAI1, parent CT-26, CT-26/wild-KAI1, and CT-26/spliced-KAI1 cells were injected s.c. in a syngeneic host, BALB/c mice, at a density of 5 × 10⁶ cells. Tumor volumes were measured every day from the first week after injection until necrosis was observed in the primary tumors. All of the CT-26/parent and CT-26/spliced-KAI1 cells formed tumors, whereas 70% (7 of 10) of the mice inoculated with CT-26/wild-KAI1 cells developed tumors. Also, the tumor sizes obtained from CT-26/spliced-KAI1 cells were similar to those from CT-26/parent cells (Fig. 7). These results indicate that wild-type KAI1 suppressed tumor growth, whereas spliced-KAI1 did not.

We observed that spliced-KAI1 was expressed in the tumor tissue and metastatic tissues of gastric cancer specimens from patients with distant metastases and poor prognosis (Fig. 1,d). We examined whether CT-26 cells would metastasize from the back to distant organs such as lung or liver after s.c. injection of CT-26/spliced-KAI1 cells or CT-26/wild-KAI1 cells and whether there would be differences in the KAI1 expression pattern in the primary tumors and metastatic tissues. In mice-bearing tumors formed by CT-26/spliced-KAI1 cells (5 of 5, n = 5), metastatic foci were found in the lung (5 of 5; Fig. 8,d), liver (3 of 5), and peritoneum (2 of 5). In mice-bearing tumors formed by CT-26/wild-KAI1 cells (3 of 5, n = 5), metastatic foci were found only in the lung (1 of 5) and liver (1 of 5). Thus, there were marked differences in the potential for distant metastasis between the tumors formed by CT-26/spliced-KAI1 cells and CT-26/wild-KAI1 cells. In the primary tumors (Fig. 8,a, Lanes 1 and 4), lung (Fig. 8,a, Lanes 2, 5, and 8) and liver tissues (Fig. 8,a, Lanes 6 and 9) with injection of CT-26/wild-KAI1 cells, wild-KAI1 was expressed, whereas expression of spliced-KAI1 was dominant in the metastatic liver tissue (Fig. 8,a, Lane 3). However, wild-KAI1 was not expressed in metastatic lung tissues (Fig. 8,b, Lanes 2, 5, and 8), and expression of spliced-KAI1 was dominant in metastatic liver tissue (Fig. 8 b, Lanes 3, 6, and 9) in the mice injected with CT-26/spliced-KAI1 cells.

We also used immunoprecipitation to examine the association of KAI1 with integrin α₃β₁ in the primary tumor and metastatic tissues of mice injected with CT-26/spliced-KAI1 cells. The association of KAI1 with integrin α₃β₁ was very weak in the metastatic lung and liver tissues compared with that of the primary tumor (Fig. 8 c). Thus, the weak interaction of spliced-KAI1 with integrin α₃β₁ in metastatic tissues may explain why tumor cells transfected with spliced-KAI1 form more metastases.

Cell adhesion plays an important role in many functions such as cell anchorage, proliferation, differentiation, invasion, and signal transduction. Increased cell adhesion is associated with an increase in the tumorigenicity of many tumor cells (17, 18, 19). After discovering an alternatively spliced variant of KAI1, we sought to analyze the influence of this structural difference in the KAI1 protein on the invasiveness of CT-26 cells. Thus, we established cell lines stably expressing wild-type and spliced-KAI1 and examined the adhesive and invasive abilities of these colon cancer cells. CT-26 cells were shown to have a higher potential to attach and respond chemotactically to fibronectin, a ligand of VLA-3, compared with their response to collagens I and IV and laminin (20). Thus, in the present study, cell adhesion and motility were estimated by using fibronectin. Stably transfected wild-type KAI1 clones showed a reduced ability to bind to fibronectin, whereas the inverse was observed for spliced KAI1-transfected clones. In the interaction with other ECMs such as type IV collagen, no significant effect on the binding ability of wild-type KAI1 transfectants was observed, but spliced-KAI1 transfectants showed slightly higher binding activity. Spliced-KAI1 transfectants also showed significantly higher adhesive ability when fibronectin was used as a ligand compared with collagen IV. We observed the increased invasive ability of CT-26/spliced-KAI1 on fibronectin-mediated locomotion, indicating a decrease in the anti-invasive effect of spliced-KAI1 as opposed to wild-type KAI1. Our data are supported by previous studies that wild-type KAI1 has an inhibitory effect on the binding and locomotion of wild-type KAI1-transfectants on fibronectin-coated surfaces, without influencing the level of expression of fibronectin receptors in melanoma and colon cancer cells, suggesting the importance of fibronectin for the expression of the anti-invasive effect of KAI1 (12, 13). The increased in vitro motility and invasive potential in CT-26/spliced-KAI1 cells are consistent with the increased cell-ECM adhesion ability and the weakened association of spliced-KAI1 with integrin α₃β₁. A correlation was found between the expression of spliced-KAI1 and the ability of transfected cells to migrate. This difference in the adhesive ability between wild-type and spliced-KAI1 transfectants can be attributed to their structural difference.

Previous studies showed that fibronectin receptors function not only in attachment and cellular adhesion but also in an intracellular signaling system where the binding of fibronectin to a cell-surface fibronectin receptor conveys signals that control cell locomotion (21, 22). As reported for CD82/KAI1 expressed in hematopoietic cells, KAI1 may play an essential role as a regulating molecule in some signal transduction pathways, and engagement of CD82/KAI1 delivers different signals depending on the cell type. For example, KAI1 can induce a costimulatory signal for full activation of Jurkat cells, a human T-cell line (23). And KAI1 can enhance the signal in FcR-mediated activation of monocyte/macrophage cell lines (24). Some members of the TM4SF family, including KAI1, have been shown to form multimolecular membrane complexes by structurally or functionally associating with other TM4SF proteins and with other molecules, many of which appear to be involved in signal transduction. CD9/MRP-1 associates with VLA integrins in the pre-B-cell and megakaryocytic cell lines and affects aggregation and migration of the two cell lines with this association (25). Consistent with it having a signaling function, CD9/MRP-1 is reported to associate with small GTP-binding proteins (26). This suggests the intriguing possibility that both CD9/MRP-1 and CD82/KAI1, two members of the TM4SF family, are furthermore involved in cell migration, invasion, and metastasis. These findings suggest that KAI1, along with other proteins, might be involved to some extent in modulating the intracellular signaling that regulates cell locomotion.

KAI1 interacts with integrin, playing an important role in the regulation of cell-ECM adhesion during invasive migration (6, 15). The interaction of wild-type KAI1 proteins with integrin α₃β₁ was confirmed by immunoprecipitation, whereas the association of spliced-KAI1 protein with integrin α₃β₁ was lower in CT-26/spliced-KAI1 cells. This suggests that the weaker association of spliced-KAI1 with integrin α₃β₁ is caused by the structural change associated with deletion of part of the second extracellular loop and the second transmembrane region. Integrin-dependent cell adhesion markedly influences cell growth, death, and differentiation through integration of cell signaling pathways and cytoskeletal reorganization (27, 28, 29). Signaling through many different integrins causes similar calcium fluxes, pH changes, and activation of focal adhesion kinase. However, specific integrins may also differ markedly from each other in terms of their effect on cell cycle progression, cell survival, and gene induction (30, 31). Integrins also play a pivotal role in the regulation of rapid turnover of cell-ECM adhesion contacts and actin cytoskeletal dynamics during invasive migration (32), and both integrins and cadherins regulate contact-mediated inhibition of cell migration (33). It was reported that β₁ integrins regulate the polarity and motility of epithelial cells by down-regulating cadherin and catenin function and the activating Rac1 and RhoA (34). In this study, we found that distant metastasis such as that seen in the lung, liver, and peritoneum was observed much more in mice-bearing tumors formed by CT-26/spliced-KAI1 cells than by CT-26/wild-KAI1 cells. The expression of spliced-KAI1 was dominant in the metastatic liver tissue, and wild-KAI1 was not expressed metastatic lung tissues. In addition, the association of KAI1 with integrin α₃β₁ was very weak in the metastatic lung and liver tissues. Thus, the decreased anti-invasive effect of spliced-KAI1 might be partly explained by the weak association of spliced-KAI1 with integrin α₃β₁, and it could result in the loss of the ability of KAI1 to suppress metastasis.

There are several possible mechanisms for fibronectin/integrin α₃β₁-induced cell invasion controlled by the integrin α₃β₁-KAI1 complex (Fig. 9). It seems that integrin α₃β₁-spliced-KAI1 complex formation in CT-26/spliced-KAI1 cells may be hindered by the structural deformity of the second extracellular loop, which contains the binding site for integrin α₃β₁. If the aberrant KAI1 protein resulting from the translation of spliced-KAI1 mRNA is expressed in the cell, it assembles dysfunctional complexes with other regulatory proteins. For example, spliced-KAI1 protein may lose its inhibitory effect on α3β1 function, thus promoting cell motility and invasion. Moreover, spliced-KAI1 protein may counteract the anti-invasive action of wild-type KAI1. Our present results indicate that the metastasis suppressor function was decreased in spliced-KAI1, suggesting that the functional difference of effects on cell motility and growth between wild-type KAI1 and spliced-KAI1 might be partially explained by the structural differences between the two KAI1 proteins. We suggest that the signaling pathways from KAI1 to the cytoskeletal system, as well as the interaction of KAI1 with integrin and/or cadherin, are important in the regulation of cell-cell contacts. Additional studies are currently being undertaken to investigate the effects of the structural and functional difference in the COOH-terminal regions of the two KAI1 proteins.

In this study, we observed that the expression of a shorter spliced form of KAI1 appeared to be linked to increased in vivo tumorigenicity and metastasis compared with wild-type KAI1 expression. This result is consistent with the higher expression of spliced-KAI1 in the metastatic tissues of gastric cancer patients with poor prognosis. KAI1 expression is correlated with the progression of a variety of human cancers such as non-small cell lung, pancreatic, bladder, gastric, and breast cancers (5, 9, 35, 36, 37). A decrease in the levels of KAI1 was associated with poor prognosis in non-small lung cancer (8) and with recurrence in breast cancer (35), suggesting that loss of KAI1 confers an important advantage on tumor cells in general. This notion has been supported by in vitro studies using tumor cell lines, demonstrating that reduced KAI1 protein expression was associated with increased in vitro invasive potential, increased cell-ECM adhesion, reduced cell-cell interactions, and increased in vivo metastatic ability (13, 36, 38, 39). Thus, the expression of spliced-KAI1 and a decrease in wild-type KAI1 level could be used as markers for poor prognosis in a variety of cancers.

We thank Jennifer Macke for assistance preparing the text.