Migration of highly aggressive melanoma cells on hyaluronic acid is associated with functional changes, increased turnover and shedding of CD44 receptors

CD44 is an ubiquitously expressed cell surface glycoprotein involved in cell activation, cell-cell adhesion and cell-substrate interaction (Underhill, 1992; Haynes et al., 1991; Lesley et al., 1993; Herrlich et al., 1993; Zöller and Kaufmann, 1994; Sherman et al., 1994; Ponta et al., 1994). A variety of CD44 isoforms have been identified with molecular weights between 80 and 200 kDa. This diversity has been attributed to cell typespecific glycosylation (Brown et al., 1991; Jalkanen et al., 1988) as well as to variable usage of 10 exons, encoding parts of the extracellular portion of CD44, which are differentially spliced (Günthert et al., 1991; Hofmann et al., 1991; Screaton et al., 1992). The 85 kDa isoform, also referred to as ‘hematopoietic’ CD44H or ‘standard’ CD44s, does not comprise alternatively spliced exons and is the predominant form expressed by blood and melanoma cells (Stamenkovic et al., 1989). CD44s is the major cellular receptor for the glycosaminoglycan hyaluronic acid (HA) (Aruffo et al., 1990; Stamenkovic et al., 1991; Miyake et al., 1990; Culty et al., 1990). In contrast, CD44E, a 150 kDa isoform primarily expressed by cells of epithelial origin, which contains variant exons v8-10, shows only little affinity for HA (Screaton et al., 1992; Stamenkovic et al., 1991).

Increasing evidence exists that CD44 expression is associated with tumor growth and metastasis. In a rat tumor model, transfection of a non-metastasizing pancreatic adenocarcinoma line with CD44v6 cDNA confers metastatic potential to this tumor cell line (Günthert et al., 1991). In human tumors, however, a strict correlation between variant expression and tumor progression does not appear to exist generally (Ponta et al., 1994). Other studies implicated that expression of the 85 kDa CD44s isoform is sufficient to promote tumor progression. B16 mouse melanoma cells show a more aggressive behaviour when carrying CD44s on their surface (Hart et al., 1991). Transfection of CD44-negative Namalwa Burkitt lymphoma cells with CD44s was associated with a significant enhancement of tumor formation when injected into nude mice, whereas transfection with the 150 kDa CD44E isoform did not reveal such an effect (Sy et al., 1991). The interaction between CD44 and HA has been ascribed a pivotal role for tumor progression since the degree of agressiveness correlated with the capacity to adhere to HA-coated substrates. In a recent study the invasive potential of human breast cancer cell lines could be attributed to their capability to express CD44s as well as to bind and degrade HA (Culty et al., 1994). In a more direct approach it was demonstrated that expression of a HA-binding wild-type CD44 isoform augments tumor formation by melanoma cells in vivo whereas expression of a CD44 mutant incapable of binding HA failed to do so (Bartolazzi et al., 1994). Accordingly, another report describes that the monoclonal antibody (mAb) GKW.A3, which inhibits binding of CD44 to HA in vitro, also led to significant inhibition of melanoma growth and metastasis in vivo (Guo et al., 1994).

However, the 85 kDa CD44s isoform is expressed on the surface of many tumor cells irrespective of their metastatic behaviour (Günthert et al., 1991; Pals et al., 1989). We therefore studied the haptotactic migration on HA-coated membranes of four melanoma cell lines that differ in their metastatic potential in nude mice and compared it with four normal melanocyte cell lines. The migratory behaviour of the cells was found to depend on the formation of the HA-binding Hermes-1 epitope rather than on the overall expression of CD44s, which was virtually identical in all melanoma and melanocyte cell lines studied. Moreover, the enhanced migratory potential of highly aggressive melanoma cells was associated with increased synthesis, turnover and phosphorylation of CD44s. Furthermore, in contrast to melanocytes and less aggressive melanomas, highly aggressive melanoma cells were found to shed CD44 from the cell surface as well as to secrete HA, the ligand of the latter. These observations may be of relevance for understanding melanoma cell invasiveness in HA-rich stroma.

Human melanoma cell lines MV3, BLM, IF6 and 530 (van Muijen et al., 1989, 1991; Versteeg et al., 1988; Danen et al., 1993) were kindly provided by Dr G. van Muijen (Department of Pathology, Academish Ziekenhuis, University of Nijmegen, The Netherlands). These cell lines differ in their capacity to form metastases after subcutaneous inoculation into nude mice (van Muijen et al., 1989, 1991; Danen et al., 1993). Cells were cultured in RPMI 1640 medium supplemented with 10% fetal calf serum, 2 mM glutamine, 1% non-essential amino acids and 100 i.u./ml penicillin/streptomycin, respectively. Melanocyte cultures were established from human foreskin as described earlier and maintained in TIC medium (Halaban et al., 1986). The latter consisted of Ham’s F10 medium (Linaris, Bettingen/Main, Germany) supplemented with 10% Nu-serum (Collaborative Research via Serva, Heidelberg, Germany), 10% fetal calf serum, 85 nM 12-O-tetradecanoyl-phorbol-13-acetate (TPA) (Serva), 0.1 mM 3-isobutyl-1-methylxanthine (IBMX) (Sigma, Deisenhofen, Germany), 2.5 nM cholera toxin as well as 100 U/ml penicillin and streptomycin, respectively.

The mAbs against CD44s employed are listed in Table 1. For control experiments mAbs W6/32 (mouse IgG_2a) against MHC class 1 (Servalab via Biozol, Eching, Germany), 4D10 (rat IgG_2a) against E-selectin (BMA via Dianova, Hamburg, Germany), 84H10 (mouse IgG₁) against ICAM-1 (Immunotech via Dianova), RM3/1 (mouse IgG₁) against a macrophage antigen (Zwadlo et al., 1987) and BT15 (mouse IgG₁) against an 80 kDa cell surface glycoprotein of human keratinocytes (Klein et al., 1987) were used.

For one-colour-immunofluorescence staining cells were detached with 0.02% ethylenediaminetetraacetic acid (EDTA) in phosphatebuffered saline (PBS), washed, suspended in 1% bovine serum albumin (BSA)/PBS to prevent non-specific binding and then incubated with appropriate dilutions of primary antibody (1 μg/ml or 1:20 dilution of supernatants) for 60 minutes at room temperature, as described earlier (Goebeler et al., 1993). Cells were again washed and then incubated with fluorescein-conjugated second-stage rabbit antimouse or rabbit anti-rat F(ab)′₂ antibodies (Dako, Hamburg, Germany). Subsequently, 0.66 mg/ml propidium iodide was added to allow determination of non-viable cells. After additional washing steps cells were immediately analyzed on a FACScan^R using Lysis^R software (Becton Dickinson, Heidelberg, Germany).

Haptotaxis of melanoma cells and melanocytes was determined using modified Boyden chambers (Neuro Probe, via Costar, Bodenheim, Germany) with polycarbonate Nucleopore filters of 8 μM pore size (Costar). Filters were floated overnight at 37°C on a solution of 0.05, 0.5 or 5 mg/ml HA (Sigma) in PBS or on a solution of chondroitin-6-sulfate (Sigma) at similar concentrations. For negative controls, filters were floated on 1% (w/v) BSA/PBS; in additional experiments, membranes coated on a solution of 10 μg/ml fibronectin (Sigma) in PBS were employed. The lower compartment of the chambers was filled with serum-free RPMI 1640 medium containing 0.1% (w/v) BSA. After several washes with PBS, filters were placed into the chambers with the coated membrane side facing the lower compartment. 10⁵ cells, obtained after treatment with 0.02% EDTA/PBS and suspended in 100 μl 0.1% BSA/RPMI 1640, were added to the upper compartment. In a series of experiments, cells were incubated with mAbs against CD44 epitopes and isotype control mAbs at a concentration of 10 μg/ml (i.e. 1 μg/100,000 cells) 30 minutes prior to and during the haptotaxis assay. Chambers were subsequently incubated at 37°C in a 5% CO₂ atmosphere for 4 hours. After removal of filters, cells on the non-coated upper membrane side were gently wiped off. Filters were then washed, fixed, stained and mounted on glass slides. Cells that had migrated to the coated side of the filter were counted in five defined fields at a magnification of 200×.

Subconfluent cultures of melanoma cells or melanocytes were washed, preincubated with methionine-free RPMI1640 medium (Biochrom, Berlin, Germany) for 2 hours and subsequently pulselabelled with 0.125 μCi [³⁵S]methionine (Amersham-Buchler, Braunschweig, Germany) for 30 minutes. After chase periods of 1 hour up to 48 hours cells were lyzed in 0.5% (v/v) NP-40, 0.15 M NaCl, 0.01 M Tris pH 7.5, 2 mM phenylmethylsulfonyl fluoride and 4 μg/ml aprotinin (all obtained from Sigma). Immunoprecipitations were carried out as previously described (Klein et al., 1988). Briefly, lysates were adsorbed to mAb P1G12 directed against CD44s or an isotype control antibody (BT15). After incubation with goat anti-mousecoupled Protein A-Sepharose (Pharmacia, Freiburg, Germany), centrifugation and repeated washes with 0.15 M NaCl, 0.01 M Tris, pH 7.5, containing 0.5% (v/v) NP-40 and 0.1% (w/v) SDS, bound antigen was eluted with SDS-PAGE sample buffer. Samples were separated on 9% SDS polyacrylamide gels under reducing conditions and gels exposed to X-ray films. Autoradiography bands were quantified by densitometrical scanning employing a Fast Scan^R supplied with Image Quant^R software (Molecular Dynamics, Sunny Vale, CA).

Melanoma cells were incubated for 1 hour with phosphate-free RPMI supplemented with 10% dialyzed FCS and then labelled with 500 μCi ortho-³³phosphate (Amersham-Buchler) for 15 hours (experiment A). In a parallel experiment (B) cells were cultured under identical conditions except for the loading with ortho-³³phosphate. Cells of experiment B were lysed after 16 hours and their CD44 content determined by a CD44 sandwich ELISA (sensitivity 0.07 ng/ml) (Bender). Based on the data obtained in experiment B, volumes of the lysates obtained in experiment A were calibrated to a similar CD44 content. Lysates were subsequently processed for immunoprecipitation employing mAb P1G12 against CD44s and an isotype control mAb (BT15) as described above.

To determine secretion of HA into the supernatant, melanoma cells and melanocytes were cultured for 16 hours in defined volumes of serum-free medium. Supernatants were collected, centrifuged at 13,000 g for 10 minutes and analyzed by a HA radiometric assay (sensitivity < 5 ng/ml) (Pharmacia). The test procedure was based on the use of specific ¹²⁵I-labelled HA-binding proteins isolated from bovine cartilage, which reacted with the HA in the supernatant. The unbound [¹²⁵I]HA binding protein was then quantified by incubating with HA covalently coupled to Sepharose particles, which were subsequently separated by centrifugation. Radioactivity bound to the particles, which is inversely proportional to the concentration in the supernatant, was then determined using a gamma counter. The concentration of HA in the samples was finally related to 10⁶ cells.

Firstly, we studied haptotactic migration of melanoma cell lines and melanocytes. Boyden chambers were equipped with polycarbonate membranes, which were coated on the bottom side with either HA or, as a control, chondroitin-6-sulfate. In a series of pilot experiments a coating concentration of 5 mg/ml HA or chondroitin-6-sulfate was found to be appropriate (data not shown). The highly aggressive melanoma cell lines MV3 and BLM, previously described as having a highly aggressive behaviour when injected into nude mice (van Muijen et al., 1989, 1991; Danen et al., 1993), showed a significantly higher transmembrane migration rate (P<0.01, Student’s t-test) through HA-coated membranes than the less aggressive melanoma lines IF6 and 530 (Fig. 1A). Four normal melanocyte cell lines tested in parallel revealed significantly lower migration rates as well (P<0.01, Student’s t-test; Fig. 1B). Almost no migration of either cell line occurred when chondroitin-6-sulfate-coated membranes were employed (Fig. 1).

The different migratory capacities of highly and less aggressive melanoma cells and melanocytes to HA prompted us to study surface expression of CD44, the major cellular HA-receptor. FACS analysis using a panel of mAbs against CD44s was performed to study surface expression of the latter. Interestingly, all melanoma cell lines and melanocytes revealed virtually identical levels of CD44 surface expression, as judged by immunolabelling with mAbs Hermes-3, P3H9, 25–32, F10-44–2, SFF2 (see Fig. 2) and P1G12 and J-173 (data not shown). Only one mAb, Hermes-1, showed differential staining intensities: in the highly aggressive melanoma lines MV3 and BLM the intensity was twoto threefold stronger than in the less aggressive IF6 and 530 cells and about 5-6 times higher than in normal melanocytes (Fig. 2).

To examine whether formation of the CD44s epitope recognized by Hermes-1 is essential for haptotactic migration to HA, blocking experiments using different anti-CD44s mAbs were employed. MV3 melanoma cells were preincubated with CD44s and control mAbs and subsequently tested, in the presence of mAb in the upper chamber, for their migratory capacity through HA-coated membranes. In contrast to the isotype control mAb 4D10 and to mAb W6/32 directed against a monomorphic MHC class 1 epitope, the Hermes-1 antibody led to a strong reduction of the migratory capacity of MV3 cells towards HA (P<0.01, Student’s t-test) (Fig. 3). In contrast, mAbs Hermes-3 (Fig. 3) or P1G12, P3H9, J-173 and OS/37 (data not shown) against CD44s epitopes did not significantly reduce migration rates.

Haptotactic migration of highly aggressive melanoma cells towards HA requires a distinct activation state of CD44, which appears to be reflected by formation of the Hermes-1 epitope. Since variation of activation states is often regulated by phosphorylation events, we studied CD44 phosphorylation in melanoma cells and melanocytes. Lysates of cells labelled with ortho-³³phosphate were calibrated to their CD44 content as described in Materials and Methods and processed for CD44 immunoprecipitation using mAb P1G12. As depicted in Fig. 4, the highly aggressive melanoma cell line MV3 shows constitutive phosphorylation of CD44 whereas 530, a melanoma cell line with less aggressive potential, has no CD44 phosphorylation. Similarly, melanocytes were not found to show phosphorylation of their CD44 molecules (data not shown).

Melanocytes that show almost no migratory capacity through HA-coated membranes reveal low, but still significant, levels of Hermes-1 surface-staining intensity (see Fig. 2). Therefore, CD44-related mechanisms in addition to formation of the Hermes-1 epitope appear to exist, which could contribute to the migratory potential of highly aggressive melanoma cell lines. To further investigate this aspect, CD44 synthesis and turnover was studied in melanoma cells as well as melanocytes by immunoprecipitation and pulse/chase experiments. Bands of identical size, 85 kDa, were precipitated from lysates of MV3 melanoma cells and the normal melanocyte cell line M-SA utilizing mAb P1G12 (Fig. 5). The other melanoma (BLM, IF6, 530) and melanocyte lines showed bands of 85 kDa as well (data not shown). Pulse-labelling for 30 minutes without a subsequent chase period revealed a band of 60 kDa (Fig. 5). Immunoprecipitation after chase periods of 1 hour (data not shown) up to 48 hours (Fig. 5) did not reveal this band any longer, thus indicating that it represents a precursor of the mature 85 kDa CD44 molecule. Comparing M-SA melanocytes with the melanoma cell line MV3, the latter shows a threeto fivefold higher rate of CD44 synthesis as judged by densitometric scanning of autoradiographic bands. Similarly, the CD44 synthesis rate in the highly aggressive MV3 cells was roughly five- and sevenfold higher than that of IF6 and 530 cells, respectively (data not shown). According to quantitative densitometry the half life of CD44s could be calculated to be about 20 hours in MV3 cells and about 48 hours in normal melanocytes. The turnover rate of CD44 thus appears to be higher in aggressive melanoma cells.

Taking into account that both cell lines express similar levels of CD44 on their surface the difference in CD44 synthesis could be caused by shedding from the cell surface in the aggressive melanoma cell line. To elucidate whether CD44 was released into the supernatant, we immunoprecipitated CD44 from the latter using mAb P1G12. A prominent protein band of 65 kDa was detectable in the supernatant of MV3 cells that could not be observed in that of melanocytes (Fig. 6). A much weaker band of 85 kDa, thus indistinguishable by size from that immunoprecipitated from cell lysates, was found in the supernatant of melanoma cells and, but weaker still, of melanocytes. Immunoprecipitations after chase periods of 1, 2, 4, 8, 18, 24 and 48 hours demonstrated that the appearance of this band did not occur before 2 hours of chase (data not shown). The size difference of 20 kDa in comparison to the 85 kDa CD44s molecule detected in cell lysates, as well as time course experiments, suggest that the 65 kDa band released by aggressive melanoma cells represents a released form of CD44. This product was not found in the supernatant of the two melanoma cell lines displaying low aggressiveness (data not shown), nor in the supernatant of normal melanocytes.

The increased haptotactic migration of highly aggressive melanoma cells might not only depend on the CD44 receptor, but also on secretion of its ligand HA, which could contribute to migration in an ‘autocrine motility loop’. To study whether highly aggressive melanoma cells meet the requirement to secrete HA we performed radiometric assays employing ¹²⁵I-labelled HA binding proteins. Interestingly, the highly aggressive cell lines MV3 and BLM were found to secrete substantially higher amounts of HA than less aggressive melanoma cell lines (IF6 and 530) and melanocytes (Fig. 7). These marked differences are not simply due to a generally increased synthesis of molecules in highly aggressive melanoma cells, since the secretion of another extracellular matrix protein, fibronectin, shows no correlation with the aggressive potential (own unpublished observation).

In the present study we analyzed the migratory capacity of melanoma cells and melanocytes towards the extracellular matrix component HA. The highly aggressive human melanoma cell lines MV3 and BLM, which reveal a highly aggressive metastatic behaviour after s.c. inoculation into nude mice (van Muijen et al., 1989, 1991; Danen et al., 1993), were found to show significantly higher rates of haptotactic migration to HA than less aggressive melanoma cells or normal melanocytes. Interestingly, cell surface expression of CD44s, the major cellular receptor for HA, was almost identical in highly and less aggressive melanoma cells and melanocytes and did not correlate with their different migratory capacities. Thus, differences in melanoma cell migration to HA are not necessarily associated with changes in overall levels of CD44s surface expression, which is in contrast to conclusions reported earlier (Thomas et al., 1993). Nevertheless, inhibition experiments utilizing mAb Hermes-1 demonstrated that CD44 rather than another HA receptor is the predominant molecule responsible for migration on HA. RHAMM, another receptor mediating motility on HA (Hardwick et al., 1992), was not found to be expressed by the melanoma cell lines and melanocytes analyzed (own unpublished observation). Studies by Thomas et al. (1992) demonstrated that transfection of melanoma cells with the 85 kDa isoform CD44H enhanced motility on HA-coated substrates whereas transfection with the 150 kDa epithelial isoform CD44E failed to do so. Similar observations have been made in lymphoma cells, where expression of the CD44H but not the CD44E isoform was found to promote tumor growth (Sy et al., 1991). Our immunoprecipitation studies demonstrate that highly and less aggressive melanoma cells as well as normal melanocytes all express the 85 kDa isoform of CD44, thus ruling out CD44E expression as being responsible for reduced motility of the latter. Expression of CD44 variant isoforms, which has been attributed to tumor progression and metastasis (Zöller and Kaufmann, 1994; Sherman et al., 1994; Ponta et al., 1994) was not a striking feature of the melanoma cells analyzed in this study, which only expressed very weak levels of CD44 v3, v5, v6 and v9; normal melanocytes showed no expression of variant peptides at all (own unpublished observation). Detailed analysis of CD44 surface expression with a panel of mAbs to CD44s revealed that only the epitope recognized by mAb Hermes-1, which blocks CD44-dependent HA-binding (Culty et al., 1990), is constituted at much higher levels in those melanoma cell lines, i.e. MV3 and BLM, which exhibit stronger haptotactic migration to HA. Bearing in mind that the different properties of CD44H and CD44E have been attributed to the fact that CD44H but not CD44E binds to HA (Stamenkovic et al., 1991; Sy et al., 1991; Thomas et al., 1992), it appeared reasonable to expect that haptotactic migration of melanoma cells to HA is at least partly dependent on their ability to expose appropriate levels of epitopes necessary for HA recognition. Accordingly, haptotactic migration to HA could be significantly blocked by mAb Hermes-1 but not by other mAbs recognizing CD44s epitopes. Employing site-directed mutagenesis techniques, three HA-binding motifs have recently been identified in the CD44 molecule; two are located in the extracellular domain of CD44, and the third is in the cytoplasmatic region (Peach et al., 1993; Yang et al., 1994). All HA-binding motifs contain two basic amino acids, either arginine or lysine, which are separated by seven amino acids (Yang et al., 1994). Mutation of these ‘B(X7)B’ motifs abolished HA-binding (Peach et al., 1993; Yang et al., 1994) and reduced tumor formation (Bartolazzi et al. 1994).

The Hermes-1 mAb did not appear to directly recognize CD44 B(X7)B epitopes since neither pretreatment of cells with unlabelled HA nor preincubation of Hermes-1 antibody with peptides comprising these motifs inhibited subsequent binding of FITC-labelled Hermes-1 mAb to melanoma cell surface CD44 (own unpublished observation). Similarly, KM114, a rat anti-mouse CD44 mAb that efficiently blocks HA binding to murine CD44 (Miyake et al., 1990), was recently described also to recognize an epitope outside the B(X7)B motifs (Zheng et al., 1995). These observations indicate that multiple CD44 domains appear to be of functional relevance for sufficient HA recognition. Recent reports, furthermore, suggest that glycosylation of CD44 regulates HA recognition (Katoh et al., 1995; Lesley et al., 1995). Since immunoprecipitated CD44 from highly and less aggressive melanoma cells, as well as from melanocytes, show virtually identical relative molecular weights, potential regulation via glycolysation does not appear to be responsibile for differences in haptotactic migration towards HA and Hermes-1 binding.

Differential haptotactic migration towards HA, despite identical CD44 surface expression, indicates that the latter is required but not sufficient for interaction with HA, suggesting that the functional properties of the receptor may be regulated. We now demonstrate that CD44 is phosphorylated in highly aggressive melanoma cells whereas no apparent phosphorylation was detected in less aggressive melanoma cells or melanocytes. Phosphorylation events could thus result in conformational changes of CD44, which may be reflected by different immunoreactivities of the receptor in highly and less aggressive melanoma cells and melanocytes with the mAb Hermes-1. It is therefore conceivable that differences in haptotactic migration may be due to conformational changes of the functional relevant Hermes-1 epitope that are dependent on the phosphorylation state of CD44. Similar data have recently been reported from leukocytes where phosphorylation of serines 325 and 327 within the cytoplasmatic domain of CD44 was found to regulate HA binding (Puré et al., 1995).

Searching for other differences regarding CD44 that could account for the different migratory capacity we found that the highly aggressive melanoma cells MV3 and BLM showed significantly higher rates of CD44 synthesis than less aggressive melanomas and melanocytes. Furthermore, MV3 and BLM cells were found to shed significant amounts of CD44 into the supernatant. Shedded CD44 displayed a molecular weight of about 65 kDa and was thus substantially smaller than the 85-kDa-isoform obtained from cell lysates. The molecular weight of CD44 shedded from melanoma cells thus resembles that found to be shedded from lymphocytes (Bazil et al., 1992). The mechanism of shedding is not yet completely understood. Recent reports suggest that metallo- and serine proteases are involved in the process, suggesting enzymatic cleavage of CD44 (Bazil et al., 1994; Katoh et al., 1994). Preliminary results from our laboratory, however, suggest that inhibitors of serine proteases, N-p-tosyl-L-lysine chloromethyl ketone (TLCK) and 3,4-dichloroisocoumarin, do inhibit cleavage from the surface of granulocytes but not from aggressive melanoma cells (own unpublished observation). The functional significance of shedded CD44 remains speculative: it may reflect some kind of detachment response of cells which are temporarily bound to an immobilized ligand, e.g. HA, during the highly dynamic process of migration. Released CD44 may also have a regulative antiadhesive function by competing with membrane-bound CD44 for its ligand HA. As shown in this study, highly aggressive melanoma cells not only shed CD44 but also secrete HA, the ligand of CD44. HA has previously been associated with metastatic events (Turley, 1992), since it is often enriched in tumor-associated stroma (Turley and Tretiak, 1984) and produced by highly metastatic tumor cells (van Muijen et al., 1995; and Knudson and Knudson, 1993, for a review); its presence has furthermore been demonstrated to facilitate melanoma invasion across chorioallantoic membranes (Turley et al., 1987). These observations implicate a role for HA in cell locomotion via an autocrine motility mechanism, which involves both secreted HA as well as surface-bound and released CD44. In summary, the data presented in our study provide evidence for profound functional differences of CD44 receptors in highly and less aggressive human melanoma cells,with relevance for tumor progression.

The authors thank Charles R. Mackay for providing mAb 25-32, Sirpa Jalkanen for mAbs Hermes-1 and Hermes-3, Shinya Murakami for mAb OS/37, Eva A. Turley for an antiserum against the RHAMM receptor and Clemens Sorg for mAbs 4D10 and RM3/1. We especially thank Heitje Lewrick, Sybille Schmid and Andrea Rivera for excellent technical assistance and Johannes Roth for helpful discussions. This work was supported by a grant from the Deutsche Forschungsgemeinschaft (Kl 510/3-2) and presented in part at the 22nd Annual Meeting of the Arbeitsgemeinschaft Dermatologische Forschung, Würzburg, Germany, November 18–20, 1994.