HEMCAM/CD146 downregulates cell surface expression of β1 integrins

Among adhesion receptors a subgroup of the immunoglobulin superfamily with a characteristic V-V-C2-C2-C2 Ig domain structure has recently emerged. Three members of this family have been identified in humans: BCAM (or as a splice variant, the Lutheran blood group antigen); ALCAM, the activated leukocyte-cell adhesion molecule; and MCAM/CD146, a marker for human melanoma tumour progression (Campbell et al., 1994; Rahuel et al., 1996; Parsons et al., 1995; Bowen et al., 1995; Shih, 1999; Johnson et al., 1996; Lehman et al., 1989). In chick, motile c-kit⁺ hemopoietic progenitor cells in the bone marrow express another member of this V-V-C2-C2-C2 subgroup, HEMCAM (Vainio et al., 1996; Dunon et al., 1998; Dunon et al., 1999). Three HEMCAM mRNA splice variants have been identified (Vainio et al., 1996). One has a short cytoplasmic tail (HEMCAM-s), another has a long tail (HEMCAM-l), whereas the third lacks transmembrane and cytoplasmic regions. Both transmembrane forms are found on the cell surface (Taira et al., 1995) but they appear to be expressed differentially in the developing nervous system (Taira et al., 1994; Takaha et al., 1995; Tsukamoto et al., 1996) and in the immune system (Vainio et al., 1996). HEMCAM expression is observed on embryonic endothelial cells, myocytes and epithelial cells of the bursa of Fabricius. HEMCAM is the same molecule as gicerin, a molecule involved in neurite outgrowth during development (Taira et al., 1994). Gicerin participates in the development of the retina, the cerebellum and the ear, and is involved in the regeneration of different organs in the adult organism (Hayashi and Miki, 1985; Hayashi et al., 1987; Kajikawa et al., 1997; Kato et al., 1992; Takaha et al., 1995; Taniura et al., 1991; Tsukamoto et al., 1998; Tsukamoto et al., 1999). Gicerin binds to neurite outgrowth factor (NOF), a 700 kDa extracellular matrix glycoprotein of the laminin family (Taira et al., 1994; Tsukamoto et al., 1996; Tsukamoto et al., 1997), and promotes cell-cell adhesion of transfected cells. These data suggest that HEMCAM exhibits both heterophilic and homophilic adhesion activities (Taira et al., 1995; Taira et al., 1994; Vainio et al., 1996).

Among the three known human V-V-C2-C2-C2 Ig molecules, the highest homology is observed between HEMCAM and MCAM/CD146 (39%) (Vainio et al., 1996). Although this homology is relatively low for the extracellular portion of the molecule, the transmembrane and the cytoplasmic domains show 66% and 69% homology, respectively. MCAM/CD146, also named MUC18, A32, Mel-CAM and S-Endo-1 (Shih, 1999), was first identified as a human melanoma-associated antigen with gradually increasing expression as tumours acquire metastatic potential (Johnson et al., 1996; Luca et al., 1993; Schlagbauer-Wadl et al., 1999; Wang et al., 1996; Xie et al., 1997). In contrast to ubiquitous expression on melanoma, MCAM expression is restricted to some types of carcinomas, such as breast carcinoma, where it has been suggested to act as a tumour suppressor (Shih et al., 1997a). Similar to HEMCAM, MCAM/CD146 mediates cell-cell adhesion both in a homophilic manner and through association with another as yet unidentified partner (Johnson et al., 1997; Shih et al., 1997b). Transfection of MCAM into melanoma cells reduces adhesion to laminin 1 and increases invasive migration through Matrigel-coated filters, although the molecular mechanism for this effect is not known (Xie et al., 1997).

In the present report, we show that HEMCAM is the avian homologue of MCAM/CD146. We investigated the effect of HEMCAM isoforms on cell adhesion, spreading, migration and proliferation on various extracellular matrix molecules including laminin 1 and fibronectin, using normal mouse fibroblasts transfected by HEMCAM isoforms. We demonstrate that the adhesion of transfected cells to laminin 1 is specifically reduced, and by comparing the expression pattern of the laminin receptor integrins, we show that HEMCAM actively downregulates the cell surface expression of the β1 integrins.

Mouse laminin 1 (LN) and bovine fibronectin (FN) were purchased from Sigma. Vitronectin (VN) was purified from bovine plasma as described previously (Yatohgo et al., 1988). c264 monoclonal antibody (mAb) against HEMCAM has been previously described (Vainio et al., 1996) and was used for cytofluorimetric (ELITE ESP, Coulter) and immunoprecipitation assays. Cells were cultured in Dulbecco’s Modified Eagle Medium (DMEM) with 10% fetal calf serum (FCS), 10 IU/ml penicillin and 100 μg/ml streptomycin (Life-Technologies Inc., Gaithersburgh, MD).

Antibodies against integrin subunits were as follows: rat mAb GoH3 and EA-1 anti-mouse α6 (Sonnenberg et al., 1987; Imhof et al., 1991; Ruiz et al., 1993), PS/2 anti-α4 (Miyake et al., 1991), RMV-7 anti-αv (Takahashi et al., 1990), CY8 against α7 (Yao et al., 1996), MFR5 against α5 (Pharmingen), 9EG7 against activated β1 integrins (Lenter et al., 1993), mouse mAb 7A3 anti-human α3 (de Melker et al., 1997) and polyclonal antibodies (pAb) 363E against β1 cytoplasmic region (DeSimone and Hynes, 1988). FITC-conjugated goat-anti mouse IgG (Jackson, West Grove, PA), FITC-conjugated rabbit anti-rat IgG (Biosys, Compiègne, France), and biotinylated donkey anti-rabbit IgG (Amersham) were used as secondary antibodies. Biotin labelling was detected by streptavidine coupled to either FITC (Amersham) or HRP (Pierce Chemical Co.).

Transfection of L cells by pcDNA3 vector coding for each isoform of HEMCAM has been described previously (Vainio et al., 1996). The sequence of the cytoplasmic part of truncated HEMCAM isoform (HEMCAM-t) is KKGKISCGRSSMHLEGPIL (HEMCAM cytoplasmic tail is truncated after the tenth amino acid, underlined sequence). Expression of HEMCAM isoforms was controlled by western blot experiments (data not shown). All results were obtained with cloned cell lines and some of them were confirmed with sorted cell populations.

Microtiter plates (96 flat-bottomed wells) were coated with different purified extracellular matrix proteins, in PBS, for 16 hours at 4°C. Coated plates were washed three times with PBS and the remaining binding sites on the plastic surface were blocked with 1% BSA for 1 hour at 37°C. Cells were trypsinized and resuspended in serum-free DMEM and counted. 50 μl of cell suspension containing about 3000 cells were added to individual substrate-coated wells and incubated at 37°C. Plates were washed twice with PBS to remove the non-adherent cells, and fixed in a 3.7% formaldehyde solution for 15 minutes at room temperature. Attached and spread cells were counted under a Nikon inverted phase-contrast microscope. In all tests, the same number of microscopy fields was analyzed.

Integrin activation assays were performed similar to adhesion assays but Puck’s A saline buffer (Life technologies Inc., Gaithersburgh, MD) was used to dilute laminin 1, Mn²⁺ solution and to wash cells before adhesion (Chen et al., 1999). Divalent cations were mixed with cells just before being added to laminin-coated wells. To determine which integrins were responsible for adhesion to laminin 1, antibodies against different integrins were added to the cell suspension just before adhesion assays.

Kinetics of proliferation were performed in 24-well plates. On each well, 20,000 cells were seeded in 1 ml of complete medium. After different times of culture, cells were trypsinized and counted. Each point of kinetics corresponds to the average of three wells (n=3).

Migratory ability of L cells was analyzed using wound assays. Cells were grown at confluence and the cell layer was wounded with a yellow tip. Adherent cells were washed twice and incubated at 37°C, 5% CO₂ for 12 hours before analysis. Cell migration and cell motility were also analyzed under videomicroscopy.

Alternatively, cell migration was analyzed using lateral migration assays (adapted from Koivisto et al., 1999). Briefly, 24-well plates were coated with extracellular matrix molecules. Each well was seeded with 30,000 L cells into glass cloning cylinders with an opening of 4 mm. Cells were left to attach for 3 hours before cylinders were removed, and afterwards cells were allowed to migrate for 24, 48 or 72 hours. After fixation in 3.7% formaldehyde in PBS, cells were stained with 0.25% Coomassie blue in 50% methanol-10% acetic acid for 15 minutes at room temperature. Cells were washed twice with 50% methanol-10% acetic acid. To measure the average diameter of the cell colony a 20-fold-magnified image was used. Cell migration was calculated by subtracting the area of the original colony to the area of the colony obtained after cell migration.

For surface protein labelling, cells were incubated with 1 mg/ml EZ-Link sulfo-NHS-LC-biotin (Pierce Chemical Co.) for 45 minutes at 4°C. Cells were washed three times with TBS, and lysed in lysis buffer (2% NP-40, 150 mM NaCl, 2 mM PMSF, 2 mM CaCl₂2, 2 mM MgCl₂, 50 mM Tris pH 8.0). For total protein labelling, cells were metabolically labelled, incubating them overnight with 30 μCi/ml [³⁵S]methionine/cysteine in 1% FCS-supplemented DMEM medium without methionine and cysteine. Proteins were extracted for 1 hour at 4°C in 1 ml of lysis buffer. Pulse-chase experiments were carried out to study the maturation rate of β1 chain. Cells were incubated in methionine and cysteine-free DMEM for 1 hour before labelling with [³⁵S]methionine/cysteine for 5 hours. Cells were then washed twice with PBS and incubated in their complete medium. Proteins were extracted after 0, 12, 24 and 36 hours of chase.

For immunoprecipitation, cell extracts were centrifuged for 30 minutes at 13,000 rpm at 4°C. The supernatant was cleared with mouse pre-immune serum coupled to protein G-agarose beads (Boehringer Mannheim) and incubated overnight with the corresponding antibody previously adsorbed to protein G-agarose. For β1 immunoprecipitation, rabbit pre-immune serum and protein A-agarose beads were used to pre-clear extracts. Beads were washed three times with lysis buffer and twice with TBS, boiled in Laemmli buffer, and run on 6% acrylamide SDS-PAGE gels. After the transfer of proteins onto nitrocellulose sheet, analysis of biotinylated proteins was performed with HRP-coupled streptavidine and ECL system (Amersham). For metabolically labelled proteins, gels were autoradiographed following a treatment with an autoradiography enhancer (Dupont).

For western blot experiments, proteins from 75,000 cells were extracted and boiled in Laemmli buffer before separating on a 6% acrylamide SDS-PAGE gel under nonreducing conditions. Proteins were electroblotted onto nitrocellulose. Nitrocellulose membranes were blocked with nonfat milk (5% w/v in TBS, 0.5% Tween-20; TBSTw) and incubated for 1 hour at room temperature with 363E antibody in TBSTw. After washing, membranes were incubated with HRP-coupled anti-rabbit antibody for 1 hour at room temperature. After washing in TBSTw and TBS, anti-rabbit antibodies were detected using the ECL system, following the manufacturer’s instructions (Amersham).

Chick embryos at 2.5 days of embryogenic development were fixed in 4% paraformaldehyde/PBS overnight at 4°C. HEMCAM digoxygenin (DIG)-labelled antisense riboprobe was synthesized by in vitro transcription using T7 RNA polymerase. The coding sequence fragment transcribed was localized between positions 425 and 935 of HEMCAM cDNA. Whole-mount in situ hybridization was performed as described previously (Prince and Lumsden, 1994).

Japanese quail (Corturnix corturnix japonica) embryos were used for immunohistochemistry. Embryos were routinely fixed at room temperature in 3.7% formaldehyde in PBS for 12 hours. After washing in PBS for one hour, embryos were embedded in increasing concentrations of sucrose solution in PBS (5-15% w/v) and subsequently frozen with OCT compound (Tissue-Tek, Life Sciences Inc.) in liquid nitrogen. Using a cryostat, 10 μm sections were obtained (Bright Instrument Co. Ltd, Huntington, UK) and mounted on superfrost slides (CML, France). Nonspecific staining was blocked by incubating the sections on 1% BSA in PBS for 1 hour at room temperature. The sections were incubated with polyclonal anti-HEMCAM serum (1/1000) for 1 hour at room temperature, washed three times in PBS and incubated for 1 hour with biotin-conjugated donkey anti-rabbit antibody (Amersham). PBS-washed sections were incubated for 1 hour with strepavidin-conjugated fluorescein (Amersham), mounted in Immunomount (CML) and observed under a Leitz Orthoplan epifluorescent microscope. Photographs were taken using TMX-400 Kodak film.

Cytoplasmic regions of HEMCAM, murine and human MCAM were amplified by RT-PCR. Total RNA from chicken muscle, mouse muscle and HUVEC was isolated using the guanidium isothiocyanate method (Chomczynski and Sacchi, 1987). 5 μg of total RNA were used as a template for the synthesis of randomly primed single-strand cDNA using mouse mammary leukemia virus reverse transcriptase (Life Technologies Inc.) in a reaction volume of 20 μl, according to the supplier’s instructions. cDNA reaction volume was subsequently made up to 50 μl with water and heated to 94°C for 2 minutes to inactivate the enzyme.

Cytoplasmic regions of HEMCAM, murine MCAM and human MCAM were amplified by PCR using specific primers. The oligonucleotides used for HEMCAM and MCAM were: 15667 (HEMCAM, from nucleotide 1876) 5′TTCCTGCACAAGAAGGGC; 15668 (HEMCAM-l, from nucleotide 2068, antisense) 5′TTAGTTTCTCAGATCGATC; 11674 (human MCAM, from nucleotide 1336) 5′GTGTTGAATCTGTCTTGTGAA; MEMCAM-1 (murine MCAM, from exon 12) 5′AGTGTACAGCCTCCAAC; MEMCAM-2 (murine MCAM from exon 16 and human MCAM from nucleotide 1919, antisense) 5′ATGCCTCAGATCGATG (Lehman et al., 1989; Sers et al., 1993; Vainio et al., 1996). Primers for murine α6 integrin subunit were 5′ATCTCTCGCTCTTCTTTCCG and 5′GACTCTTAACTGTAGCGTGA (Tamura et al., 1991). The 550 bp fragment corresponded to the α6A cytoplasmic domain isoform and the 420 bp fragment to the α6B isoform. Primers for murine β1 integrin subunit were 5′TTGTGGAGACTCCAGACTGTTCTACT and 5′TCATTTTCCCTCATACTTCGGATT (Belkin et al., 1996). The 324 bp fragment corresponded to the β1D cytoplasmic domain isoform and the 243 bp fragment to the β1A isoform. Amplified DNA fragments were gel-purified and cloned into pCRTM2.1TOPO vector (Invitrogen). Sequences were determined from denatured double-stranded recombinant plasmid DNA using SequenaseTM (Amersham) in the chain termination reaction. Sequences were analyzed and assembled with the software package of Infobiogen (Paris, France).

Electrophoresis of RNA samples was carried out in a 1.2% agarose gel in the presence of formaldehyde. The RNA was then transferred to a nylon membrane by capillary blotting and fixed to the filter by exposure to UV light (Stratalinker, Stratagene, La Jolla, CA). RNA was hybridized with an α6 cDNA probe obtained by PCR labelling with [α−³²P]dATP. Hybridization was carried out at 42°C in 50% formamide, 5× SSC, 5× Denhardt’s solution, 0.1% SDS and 300 μg/ml denatured salmon sperm DNA. Filters were washed twice in 1× SSC, 0.1% SDS at room temperature and once at 65°C in 0.1× SSC, 0.1% SDS. Filters were exposed to X-ray film (BIOMAX MR, Kodak) at −80°C with intensifying screens.

Given the homology between HEMCAM and MCAM, we address whether they might in fact be homologous molecules. Because three HEMCAM mRNA splice variants exist in birds, one way of investigating this possibility was to determine whether a similar splicing exists for MCAM in mammalian cells. Indeed, two transcripts were detected by RT-PCR in mouse and human tissues using MCAM-specific oligonucleotides (Fig. 1A). Cloning and sequencing of PCR products demonstrated that two transmembrane isoforms of MCAM were generated by a splicing event and that these two forms differed in the cytoplasmic portion (Fig. 1B). Nucleotide sequence analysis and comparison with the human gene organization (Sers et al., 1993) indicated that exon 15 was excised in the short MCAM form and that this alternative splicing event was identical in birds and mammals. Interestingly, excision of exon 15 led to a frame shift for exon 16 that changed the last six amino acids at the C terminus. Amino acid sequence comparison showed that the short cytoplasmic form of HEMCAM and MCAM contained a putative PDZ domain interaction site (T/S-X-I/V) (Songyang et al., 1997). In the two isoforms of HEMCAM and MCAM, the constant sequence of the cytoplasmic part codes for a putative PKC phosphorylation site (RSGK). An additional PKC phosphorylation site (KSDK) and a YXXL sequence were conserved in the long cytoplasmic form of chicken HEMCAM, and mouse and human MCAM. The YXXL sequence was first described as a half immunoreceptor tyrosine-based activation motif (ITAM) and, more recently, as an endocytosis motif (Deschambeault et al., 1999). The region corresponding to exon 13 of human MCAM was absent in HEMCAM (Vainio et al., 1996). Exhaustive analysis of transcripts was performed around this region in mouse, human and chicken tissues but no other splicing event was detected.

To analyze the expression profile of HEMCAM, we performed immunohistochemistry and whole mount in situ hybridization experiments in chick. At day 2.5 of development, HEMCAM was expressed in migrating neural crest cells, an embryonic cell population known to give arise to melanocytes. HEMCAM immunoreactivity was also detected in epidermis and endoderm, as well as in the notochord.

Whole mount in situ hybridization results clearly show that neural crest cells emigrating from the dorsal part of the neural tube expressed HEMCAM mRNA, which was also observed in dermomyotome cells (Fig. 2B). This expression pattern is confirmed by transverse sections (Fig. 2Bb), which also underline labelling of epidermis. This expression profile was observed along the whole A-P axis.

Because it was shown that human MCAM could alter cell adhesion to laminin (Xie et al., 1997), we analyzed whether its chicken orthologue exerted this function and if its alternative forms exhibited different activities. Stable transfections of each isoform of HEMCAM into L cells were performed. HEMCAM expression was localized at the cell surface and preferentially in the cellular protrusions. This expression pattern was confirmed in cell clones and sorted transfected cells (Fig. 3A). Several properties of L cells were modified by HEMCAM transfection. A first consequence was a delay in proliferation of transfected cells after trypsinization (Fig. 3B). To determine the involvement of HEMCAM in cell migration, wounding and lateral migration assays were performed. HEMCAM expression inhibited cell migration on tissue culture dishes (Fig. 4A,B) and on fibronectin substrate (data not shown). Motile activity was studied by videomicroscopy 24 hours after plating the cells in culture dish plates. Mock- and HEMCAM-transfected cells were motile but the structure of protrusion was different between cell lines. Mock-transfected cells exhibit long and thin filopodia, whereas transfected cells gave rise to short, puffy protrusions. These experiments suggest that these alterations of motility were involved in the inhibition of migration.

To further investigate the role of HEMCAM in this process, adhesion to laminin 1, fibronectin and vitronectin was analyzed at graded concentrations of each extracellular matrix protein but also in a kinetics test (Fig. 4C; Fig. 5). On laminin-1-coated plates, expression of HEMCAM-l, HEMCAM-s, or HEMCAM-t inhibited adhesion of L cells by 85%, compared with that in mock-transfected cells. Cell spreading was completely abolished by transfection of all HEMCAM forms. On fibronectin, HEMCAM-s transfection reduced cell adhesion and spreading, whereas HEMCAM-l and HEMCAM-t had a weak effect on cell spreading (Fig. 5A). These results suggested that the cytoplasmic domain of HEMCAM could be involved in the regulation of cell spreading on fibronectin. In addition, HEMCAM reduced cell adhesion and cell spreading on vitronectin, whichever construct was expressed at the cell surface (Fig. 5B). These results were confirmed using several independent L-cell clones transfected with the different HEMCAM constructs, and sorted populations of recently transfected L cells.

To address whether HEMCAM could affect integrin expression or function in L cells, we first characterized the repertoire of LN and FN integrin receptors in these cells using blocking antibodies (Fig. 5; Fig. 6; Fig. 8). On laminin 1, anti-α6 integrin antibodies EA-1 and GoH3 decreased the adhesion of mock-transfected L cells by 75%. As expected, anti-αv integrin antibody RMV-7 had no effect (Fig. 6A). These results demonstrate that α6 integrins were the main laminin receptors at the L-cell surface. Thus, the inhibition of L-cell adhesion to laminin 1 induced by HEMCAM might be due to inactivation of α6 integrin. Integrin receptors can be activated by Mn²⁺ addition (Shaw and Mercurio, 1993; Shaw et al., 1993). To test whether HEMCAM expression can induce the switch from an active to an inactive conformation of the integrin, we replaced Ca²⁺ by Mn²⁺ in the adhesion assay. Whereas graded concentrations of Mn²⁺ increased adhesion of mock-transfected cells to laminin 1, no effect was observed with HEMCAM-transfected cells (Fig. 6B). This suggests that integrin receptors are not sensitive to Mn²⁺ modulation after overexpression of HEMCAM, either because they cannot be activated or because they are not present at the cell surface.

Flow cytometry results revealed that HEMCAM transfection does indeed induce downregulation of α5, α6 and α7 integrins at the cell surface (Fig. 7; Fig. 8A). By contrast, HEMCAM transfection had no effect on the expression levels of αv integrins (Fig. 7, Fig. 8B). The downregulation of α6 was observed with two different mAbs each recognizing all conformational states of this integrin. This inhibition was reversible, because loss of HEMCAM expression restored α6β1 expression by L cells (data not shown). The disappearance of β1 integrins was observed in different clones of transfected cells whatever the HEMCAM spliced variant was expressed (HEMCAM-l or HEMCAM-s). Using mAb 9EG7, which recognizes activated β1 integrins (Bazzoni et al., 1998; Lenter et al., 1993), we found residual cell surface expression in HEMCAM-transfected cells, suggesting that some β1 integrins are still present (Fig. 7). Immunoprecipitation studies confirmed the FACS analysis (Fig. 8B; Fig. 9A). In L cells, α6 is expressed exclusively associated with the β1 chain, because no band of appropriate molecular size of the β4 chain was coprecipitated with anti-α6 antibodies from surface-biotinylated cells. In addition to the α6β1 integrin, other laminin binding integrins, such as α3β1 and α7β1, were downregulated by HEMCAM (Fig. 8). The low residual β1 integrin detected at the cell surface might correspond to a low accumulation of various α chains associated with the β1 subunit. This residual β1 integrin cannot correspond to αvβ1 because immunoprecipitation experiments indicate that the αv chain is associated with the β3 subunit in all cell lines (Fig. 8B). Together, these results demonstrate that HEMCAM overexpression dramatically alters the β1 integrin repertoire expressed on the surface of L cells.

As α6β1 integrin was the major laminin receptor of L cells, we have analyzed how HEMCAM might regulate its expression. RT-PCR experiments established that HEMCAM expression did not modify the transcription level of α6 and β1 integrin genes (Fig. 10A). These results were confirmed by northern blot experiments (Fig. 10B). These experiments showed that L cells expressed the β1A and α6A integrin splice variants and that no modification of splicing was induced by HEMCAM transfection. Because HEMCAM transfection did not interfere with transcription of β1 and α6 integrin genes (Fig. 10), we tested whether overexpression of HEMCAM induced a redistribution of α6β1 integrin in intracellular compartments. Immunoprecipitation of α6β1 integrin was performed using metabolically labelled L cells. In HEMCAM-transfected cells, a complete disappearance of α6β1 was observed, indicating that α6β1 integrin was not synthesized during labelling (Fig. 9B). Thus this absence at the cell surface was probably not due to α6β1 integrin redistribution. Western blot experiments using anti-cytoplasmic β1-integrin antibody were performed to detect the β1 integrin precursor. Although the precursor form of β1 integrin subunit was clearly present in all cell lines (transfected and mock-transfected), the mature β1 protein was barely detectable in HEMCAM-transfected cells (Fig. 9B), supporting the data obtained by FACS and immunoprecipitation experiments (Fig. 7; Fig. 9A). Pulse-chase experiments confirmed that β1 integrin precursor is synthesized in all cells (transfected and not transfected) and stabilized in this state in HEMCAM-transfected cells (Fig. 9C). The presence of β1 integrin precursor in transfected cells showed that HEMCAM transfection interfere on β1 integrin expression at the post-translational level.

HEMCAM/gicerin was described as an adhesion molecule that exhibits homophilic cell-cell adhesion and heterophilic interaction with NOF, a laminin-like molecule. In this report, we have established that HEMCAM is the avian orthologue of MCAM/CD146. In addition, our studies demonstrate that an additional function of HEMCAM is to regulate cell adhesion by downregulating the expression of mature β1 integrins at the cell surface of fibroblasts.

Based on amino acid comparison, we previously showed that chicken HEMCAM and mammalian MCAM/CD146 were related molecules (Vainio et al., 1996). Here, we further demonstrate that the gene products of these molecules are similarly spliced, coding for proteins with a long and a short cytoplasmic tail. Moreover, human and avian molecules have similar functions: MCAM expression in MCAM-negative melanoma cells and HEMCAM expression in L cells inhibited cell adhesion and spreading to laminin 1. In addition, MCAM, absent from melanocytes, is re-expressed by melanoma cells, and HEMCAM was detected in avian neural crest cells, the precursors of melanocytes. Taken together, our results show that HEMCAM is the avian homologue of MCAM. The main difference between the mammalian and the chicken molecule is the absence of 34 amino acids in chicken HEMCAM located between the last Ig domain and the transmembrane domain. This region is encoded by exon 13 in the mammalian MCAM gene (Sers et al., 1993) and corresponds to a mucin repeat (Gum et al., 1989). Interestingly, such structural differences have previously been encountered in orthologous molecules of different species. For example, in mouse MAdCAM-1, the large mucin-like domain is replaced by an Ig domain in human MAdCAM-1 (Shyjan et al., 1996).

Comparison of amino acid sequences of the long cytoplasmic tail of mammalian MCAM and avian HEMCAM shows two conserved putative PKC phosphorylation sites and the YXXL motif. The latter can be involved in endocytosis and sorting of integral membrane proteins to specific cellular compartments (Deschambeault et al., 1999; Eng et al., 1999; Kamiguchi and Lemmon, 1998). The short cytoplasmic tail of MCAM is generated by excision of exon 15 and conserves the first PCK phosphorylation site. In addition, the short forms of mammalian MCAM and avian HEMCAM both contain a PDZ interaction site generated by a shift in the reading frame due to splicing. Proteins that bind to such sites contain PDZ domains, which are found in diverse membrane-associated proteins including members of the MAGUK family of guanylate kinase homologues, several protein phosphatases and kinases, neuronal nitric oxide synthase, and several dystrophin-associated proteins (Ponting et al., 1997). Most PDZ-domain proteins localize to specific compartments of the plasma membrane and participate in formation of cellular junctions, receptor clustering, channel formation, or intracellular signalling. The functions of different cytoplasmic sites of HEMCAM-s and HEMCAM-l are not yet clear; they might be involved in other functions of HEMCAM such as cellular localization or regulation of protein expression.

Our studies suggest that an important role of HEMCAM in cell adhesion could be the regulation of β1 integrin expression. Regulation of α6β1 integrin expression was already seen for extracellular signals such as 1α,25 dihydroxybutyrate vitamin D3, which reduced α6 integrin expression on melanoma cells, and IL-12, which increased its expression on human type 1 helper T cells (Colantonio et al., 1999; Hansen et al., 1998). Transfection of L cells by HEMCAM-s, HEMCAM-l or HEMCAM-t reduces expression of mature β1 integrins at the surface of fibroblasts. This phenomenon is thus independent of the cytoplasmic part of HEMCAM, like other adhesion characteristics of this molecule (Taira et al., 1999). Consequently, HEMCAM may regulate integrin expression by controlling the translation and/or maturation (cell surface expression) of β1 and/or α integrin subunits. The disappearance of α6 integrin from HEMCAM-transfected cells was not due to downregulation of transcription because the mRNA transcription level remained constant. In addition, no modification of splicing was observed for the α6 or β1 integrin transcripts. Another mechanism for reducing the amount of integrins from the cell surface has been observed in sensory neurones that redistribute α6β1 in different intracellular compartments (Condic and Letourneau, 1997). However, this mechanism is unlikely to play a major role in our system, because immunoprecipitation of metabolically labelled α6 integrin confirmed its absence from any other cellular compartments.

More likely, transfection of HEMCAM either induced the rapid degradation of β1 integrins or downregulated its translation and/or maturation. Regulation of integrin mRNA translation resulting in cell adhesion has been previously reported: integrin occupancy induced formation of focal adhesion complexes and translation of pre-existing messengers (Benecke et al., 1998). Moreover, HUVEC interaction with ECM leads to a relocation of mRNAs and ribosomes to focal adhesion complexes (i.e. at sites of signal reception) (Chicurel et al., 1998). As MCAM can promote phosphorylation of the focal adhesion kinase (FAK) leading to interaction with paxillin (Anfosso et al., 1998; Anfosso et al., 2000), it might interfere with the translation of specific mRNAs possibly at focal adhesion plaques. Because overexpression of HEMCAM led to the disappearance of different β1 integrins, HEMCAM might perturb β1 integrin maturation. Indeed, in HEMCAM-transfected cells, β1 integrin subunit is stabilized in its precursor form. Thus, HEMCAM might interfere with a chaperone molecule, such as calnexin, known to participate in integrin maturation (Lenter and Vestweber, 1994). Finally, HEMCAM might be involved in the regulation of the degradation pathway of α integrin subunits.

The role of HEMCAM/CD146 in cell adhesion is complex. This molecule interacts, on the one hand, with NOF, a laminin-like molecule of the extracellular matrix and, on the other hand, probably with an unknown cell adhesion molecule at the surface of other cells. Moreover, several experiments suggest that HEMCAM interacts homophilically (Taira et al., 1994; Taira et al., 1995; Taira et al., 1999; Vainio et al., 1996). In addition, our data show that, in L cells, HEMCAM regulates cell adhesion by controlling β1 integrin expression at the cell surface. Thus the level of HEMCAM expression might determine the specific laminin that is recognized by fibroblasts: at low levels of HEMCAM the cells can adhere to laminin 1 via laminin-binding integrins, whereas at high levels the loss of these integrins leads to reduced adhesion to laminin 1 and specific interaction with NOF (Taira et al., 1994; Taira et al., 1995; Taira et al., 1999). In addition, CD146 activation by antibodies or a probably extracellular partner establishes that this molecule is involved in outside-in signalling and may contribute to focal adhesion assembly and reorganization of the cytoskeleton (Anfosso et al., 1998). Thus, HEMCAM/CD146 may regulate cell adhesion by several mechanisms.

In chick, neural crest cells that give rise to melanocytes express HEMCAM, and MCAM is re-expressed during melanoma progression in humans. The role of MCAM (CD146) in tumorigenesis is complex: it is a marker of melanoma progression and seems to act as a tumour suppressor in breast carcinoma. These differential effects of HEMCAM (CD146) might be related to the regulation of cell adhesion and cell migration by controlling β1 integrin expression in specific cell types.

The authors thank Pascale Vigneron and Marie-Claude Gendron for excellent technical assistance, and Carolyne Johnson-Léger, Claire Fournier-Thibault and Dominique Alfandari for critical reading and improvement of the manuscript. We also thank Randall Kramer, Douglas De Simone, Arnoud Sonnenberg and Annemieke de Melker for the kind gifts of the CY8, 363E, GoH3 and 7A3 antibodies, respectively. This work was supported by the CNRS contract Biologie cellulaire no. 96108, by the Association pour la Recherche contre le Cancer (ARC) no. 6982 and 9738, by the Fondation pour la Recherche Médicale (FRM), the Ligue Nationale Contre le Cancer (LNCC), the Human Frontier Science Programme Organization (HFSP) no. RG366/96, the Academy of Finland and by the Swiss National Science Foundation. S. A. was supported by a LNCC and a ARC Fellowship.