Melanoma Cell CD44 Interaction with the 1(IV)1263-1277 Region from Basement Membrane Collagen is Modulated by Ligand Glycosylation*

: Invasion of the basement membrane is believed to be a critical step in the metastatic process. Melanoma cells have been shown previously to bind distinct triple-helical regions within basement membrane (type IV) collagen. Additionally, tumor cell binding sites within type IV collagen contain glycosylated hydroxylysine residues. In the present study, we have utilized triple-helical models of the type IV collagen a 1(IV)1263-1277 sequence to (a) determine the melanoma cell receptor for this ligand and (b) analyze the results of single-site glycosylation on melanoma cell recognition. Receptor identification was achieved by a combination of methods, including (a) cell adhesion and spreading assays using triple-helical a 1(IV)1263-1277 and an Asp 1266 Abu variant, (b) inhibition of cell adhesion and spreading assays, and (c) triple-helical a 1(IV)1263-1277 affinity chromatography with whole cell lysates and glycosaminoglycans. Triple-helical a 1(IV)1263-1277 was bound by melanoma cell CD44/chondroitin sulfate proteoglycan (CSPG) receptors, and not by the collagen-binding integrins or MPG. Melanoma cell adhesion to and spreading on the triple-helical a 1(IV)1263-1277 sequence was then compared for glycosylated [replacement of Lys 1265 with Hyl( O - b - D -galactopyranosyl)] versus non-glycosylated ligand. Glycosylation was found to strongly modulate both activities, as adhesion and spreading were dramatically decreased due to the presence of galactose. CD44/CSPG did not bind to glycosylated a 1(IV)1263-1277. Overall, this study (a) is the first demonstration of the prophylactic effects of glycosylation on tumor cell interaction with the basement membrane, (b) provides a rare example of an apparent unfavorable interaction between carbohydrates, and (c) suggests that sugars may chondroitin

Tumor cell invasion, a key step in the metastatic process, involves a complex series of correlated macromolecular interactions. These include interaction with, and movement through, collagen, most often type I and/or basement membrane (type IV) collagen. In general, invasion of the basement membrane is believed to be a critical step in the metastatic process. Human melanoma cells have been shown to bind distinct triple-helical regions within type IV collagen (1)(2)(3)(4)(5). Melanoma receptors for triple-helical collagen fall into one of two categories: members of the integrin heterodimeric protein family (α1β1, α2β1, and α3β1 integrins) or cell surface proteoglycans [such as CD44 and melanomaassociated proteoglycan/melanoma chondroitin sulfate proteoglycan (MPG/MCSP/NG2) 1 ]. Specific ligands from type IV collagen have been described for the α1β1, α2β1, and α3β1 integrins. The α1β1 integrin simultaneously binds Asp 441 from two α1(IV) chains and Arg 458 from the α2(IV) chain (6,7). The Gly-Phe-Hyp-Gly-Glu-Arg motif, in triple-helical conformation, has been shown to bind to the α2β1 integrin (8)(9)(10). This motif is found within type IV collagen at α1(IV)405-410; a triple-helical model of α1(IV)402-413 is bound by melanoma cells. 2 The melanoma cell α3β1 integrin binds to α1(IV)531-543 (2,4,11).
The [IV-H1] sequence contains a glycosylated hydroxylysine (Hyl) residue in position 1265 (28). Hyl is the major glycosylation site within mammalian collagens. The 5-hydroxyl group may be posttranslationally modified by the monosaccharide galactose (β-D-galactopyranosyl) or the disaccharide glucose-galactose [α-D-glucopyranosyl-(1→2)-β-D-galactopyranosyl] (29,30). Interest in glycosylation of collagen stems from by guest on March 23, 2020 http://www.jbc.org/ Downloaded from the recent reports of activation of specific receptor tyrosine kinases by glycosylated type I collagen (31), T-cell recognition of a glycosylated sequence within type II collagen (32), and the identification of melanoma and breast carcinoma binding sites within type IV collagen that contain glycosylated Hyl residues (1,2,4,11,12,14,33). Most secreted and cell surface eukaryotic proteins are found glycosylated in vivo. Glycosylation is believed to have three important biological roles (34)(35)(36)(37)(38). First, glycosylation can serve as a recognition marker for a cell, both in the context of cell-cell and cell-extracellular matrix interactions. A second role concerns the alteration glycosylation has on the physical properties of the protein. Frequently, glycosylation will render a protein resistant to hydrolysis, significantly increase solubility of a protein, or even drastically affect the overall folding and/or physical bulk of a protein. In the case of collagen-like triplehelices, the addition of β-D-galactose to Thr in the Yyy position of either (Gly-Pro-Yyy) 10 or (Gly-Hyp-Yyy) 10 greatly enhances triple-helical stability compared to Thr alone (39,40). A third role for glycosylation is in signal transduction, and may be analogous to or compete with protein phosphorylation (37 Purification of 5-Hydroxy-L-lysine. Hyl was isolated from porcine gelatin by hydrolysis followed by multiple passages over an ion-exchange column (46 1265 ]-α1(IV)1263-1277 THP were prepared as previously described (47,48). Branched THP compositions were confirmed by MALDI-MS analysis of the branch and Edman degradation sequence analysis of the intact THP (47,49). All other peptides were synthesized as C-terminal amides to prevent diketopiperazine formation (50). Peptide-resin assembly was performed by Fmoc solid-phase methodology on a Perkin Elmer/ABD 433A Peptide Synthesizer by methods previously described in our laboratory (18,19). For incorporation of the glycosylated amino acid, the H 2 N-peptidylresin was removed from the instrument. Fmoc-Hyl[ε-Boc,O-(2,3,4,6-tetra-O-acetyl-β-Dgalactopyranosyl)] was coupled manually in an orbital shaker using 3-fold molar excesses of Fmoc-amino acid and HOAt, a 2.7-fold molar excess of HATU, and a 6-fold molar excess of DIEA in 10 ml DMF for 18 h. Subsequent amino acids were coupled on the instrument. Peptide-resins were characterized by Edman degradation sequence analysis as described previously for "embedded" (non-covalent) sequencing (51) on an Applied Biosystems 477A Protein Sequencer/120A Analyzer. Peptide-resins were then either (a) cleaved or (b) acylated with the C 16 alkyl tail (19) and then cleaved. Cleavage and side-chain deprotection of the peptide-resin proceeded for 2 h using ethanedithiol−thioanisole−phenol−water−TFA (2.5:5:5:5:82.5) as described (52). The cleavage solution was extracted with methyl tert-butyl ether prior to purification. The  Cells. SK-Mel2, M14P, M14#5, and M14#11 human melanoma cells were propagated as described previously (2,4,11). Briefly, melanoma cells were cultured in EMEM or RPMI-1640 supplemented with 10% fetal bovine sera, 1 mM sodium pyruvate, 0.1 mg/mL gentamycin (Boehringer Mannheim, Indianapolis, IN), 50 units/mL penicillin, and 0.05 mg/mL streptomycin. Cells were passaged 8 times and then replaced from frozen stocks of early passage cells to minimize phenotypic drift. All cells were maintained at 37 °C in a humidified incubator containing 5% CO 2 . All media reagents were purchased from Fisher Scientific.  The goal was to achieve a 10-15% labeling of amino groups.

Construction and Characterization of Ligands
Design of Potential Ligands. To determine the receptor binding to the α1(IV)1263-1277 region and evaluate the role of glycosylation on melanoma activities, triple-helical models incorporating collagen sequences of interest needed to be constructed. In addition, in order to properly evaluate biological effects, the triple-helices of these "mini-collagens" needed to be stable to assay conditions. We have previously described two methods for assembling THPs of desirable thermal stabilities. One method uses a C-terminal covalent branch (1,47), while the other uses self-assembly driven by pseudo-lipids (18-20). Both approaches were used in the present study to create either the  (Table 1). Thus, glycosylation appeared to slightly increase triple-helical stability.
To assess the biological effects of glycosylation, we prepared the peptideamphiphile models of (Gly-Pro-Hyp) 4  for this peptide-amphiphile (19). However, the peptide-amphiphile concentration for the earlier study was 0.5 mM (19), which causes more extensive aggregation and a correspondingly higher T m value (57). The peptide-amphiphile concentration used for the CD analysis described herein (14 µM) approximates the concentration range required for biological studies (see below). and precleared human melanoma cell lysates were added to the beads. Following application of the cell lysates, the column was washed with 3 volumes of OGS lysis buffer, and then bound materials were eluted with IP lysis buffer. Eluants were incubated with mAbs against either CD44 or the β1 integrin subunit, followed by immunoprecipitation and immunoblotting with the respective mAb. A protein of ~85-90 kDa was immunoprecipitated by the anti-CD44 mAb (Figure 8). This apparent molecular weight corresponded to melanoma CD44s core protein following chondroitinase treatment (60,61). No corresponding proteins were observed using an anti-β1 integrin subunit mAb immunoprecipitation (data not shown; see later discussion).
Immunoprecipitation analysis of whole cell lysates showed the presence of both CD44 and the β1 integrin subunit (Figures 8 and 13), consistent with prior studies (60).
Incubation of the column-bound materials or whole cell lysates with IgG resulted in the detection of only IgG proteins (data not shown).
To further examine the role of CS in the binding of melanoma cells to α1(IV)1263-1277, affinity chromatography was performed using branched α1(IV)1263-chondroitin-4-sulfate and chondroitin-6-sulfate were found to specifically bind to α1(IV)1263-1277 THP, while dermatan sulfate did not (Figure 9). The relative elution profiles of chondroitin-4-sulfate and chondroitin-6-sulfate make it appear that chondroitin-4-sulfate has a greater ability to bind branched α1(IV)1263-1277 THP, but a significant amount (>4000 RFU) of chondroitin-6-sulfate remains bound to the THP and elutes only with successive washes with acetate buffer, pH 4.0, and Tris•HCl buffer, pH 8.0 (data not shown).

Effects of Glycosylation on Melanoma Activities
Human melanoma cell adhesion was examined for C 16 Figure   13). The result for the β1 integrin subunit is identical to that observed when using the non-glycosylated ligand (see earlier discussion). Immunoprecipitation analysis of whole cell lysates showed the presence of both CD44 and the β1 integrin subunit (Figure 13).

DISCUSSION
The development of model triple-helical peptide ligands has led to a better understanding of the role of the triple-helix as a modulator of biological function. In the present study, triple-helical models of α1(IV)1263-1277 have been used to define the roles of both triple-helicity and glycosylation on tumor cell interactions with basement membrane (type IV) collagen. Prior studies had shown that CD44/CSPG from melanoma cells binds directly to single-stranded α1(IV)1263-1277 (15,16), and that CD44/CSPG binds to type IV collagen (60). However, cells may engage different receptors depending upon the conformational state of collagen (21)(22)(23)(24)(25)(26)(27), and thus we needed to determine the receptor for triple-helical α1(IV)1263-1277. A variant of α1(IV)1263-1277 was constructed in which the single Asp residue was replaced with Abu. The three collagenbinding integrins, α1β1, α2β1, and α3β1, require a negatively charged residue (Asp or Glu) for binding (2,(9)(10)(11)58,59). Replacement of the one negatively charged residue in α1(IV)1263-1277 had no effect on melanoma cell binding, indicating that melanoma cell interaction with this triple-helical ligand is not integrin mediated. Inhibition of adhesion assays showed that (a) anti-integrin mAbs had no effect on melanoma cell adhesion, Triple-helical α1(IV)1263-1277 represents the second distinct extracellular matrix ligand described for CD44. CD44 has long been recognized for the ability to bind hyaluronic acid (HA). HA binds to the CD44 amino-terminal globular domain (65). The HA binding motif consists of two basic amino acids separated by seven non-acidic amino acids (B[X7]B) (65). In CD44, HA binding motifs are found within residues 21-45, with Arg 41 of particular importance (65). Several distal residues also contribute to HA binding, such as Lys 158 and Arg 162 (65). Since CS is required for CD44 binding to α1(IV)1263-1277, but interferes with CD44 binding to HA (65), it appears that α1(IV)1263-1277 and HA bind to different regions of CD44.
Position 1265 of the α1(IV) collagen chain can be glycosylated (28). The effects of this glycosylation on either triple-helical structure or CD44 binding are unknown. CD spectroscopic studies have shown that glycosylation at residue 1265 of either the triplehelical peptide or peptide-amphiphile increased the melting temperature by 3.0-3.5 °C compared to the non-glycosylated ligands. Prior work had demonstrated that β-Dgalactose glycosylation of Thr in the Yyy position of (Gly-Hyp-Yyy) 10 enhanced triplehelical stability by 32 °C compared to Thr (40). This corresponds to 3.2 °C per glycosylated residue. Thus, both studies have come to similar conclusions as the role of glycosylation in stabilizing the triple-helix. However, in the case of Hyl glycosylation, this stabilization effect is most likely localized to a specific sequence, and is not a general mechanism by which collagen thermal stability is enhanced. Type II collagen, whether in fully glycosylated (10 residues/1016 total) or lowly glycosylated (2 residues/1016 total) form, has the same T m value (66).
The role of Hyl glycosylation in the CD44 recognition processes was first studied by comparing melanoma cell adhesion to the glycosylated and non-glycosylated ligands.
A dramatic reduction in cell adhesion was observed due to the presence of the single galactose residue, suggesting significant biological consequences of even subtle changes in collagen carbohydrate content. Promotion of melanoma cell spreading had similar, although not identical, trends as seen for cell adhesion. Subsequent affinity chromatography experiments indicated that CD44 no longer bound to the α1(IV)1263-1277 sequence once carbohydrate was present. The exquisite sensitivity of cell interaction with glycosylated ligand has only rarely been observed. T cell hybridoma response to type II collagen fragments has been shown to depend upon contacts from a single glycosylated Hyl with the CD3 loops of the T cell receptor (32).
We have found that glycosylation inhibits CD44 interaction with the α1(IV)1263-1277 region derived from basement membrane collagen. This result is unexpected, as prior studies had shown that melanoma cell binding to α1(IV)1263-1277 is primarily via electrostatic interactions with Lys 1265 and Lys 1268 (67). While it is possible that the glycosylation may mask the side-chain charge of residue 1265, such behavior seems unlikely given the small size of the carbohydrate. It is more likely that we have observed a specific, unfavorable carbohydrate-carbohydrate interaction between the CD44 CS and the α1(IV)1263-1277 galactose residue. Overall, little is known about how carbohydrates interact with cell surface receptors, particularly in the case of unfavorable associations (68,69). More often, such interactions are favorable, as when carcinoma cell surface mucins associate with platelet P-selectin, creating a platelet "cloak" surrounding the tumor cells that aid in the metastatic process (45). While CD44 does bind certain carbohydrates (HA), this interaction requires a minimum of six sugar residues (three repeating disaccharide units), with affinity increasing for longer HA molecules (K d ~ 0.3 nM) (65). The present study suggests that glycosylation can be used for modulating tumor cell behaviors, based on carbohydrate structure and chain length.
The reduced binding of CD44/CSPG due to ligand glycosylation presents a possible "cryptic sites" mechanism by which tumor cells may invade the basement membrane. In the native, glycosylated state, regions within type IV collagen may have minimal interaction with receptors such as CD44/CSPG. After tumor cells bind to type IV collagen (presumably via integrins such as α2β1), cell surface glycosidases could liberate the collagen-bound carbohydrates. This process would expose "cryptic sites" for interaction with CD44/CSPG and/or other cell surface receptors (such as the α3β1 integrin, which also binds to a glycosylated region within type IV collagen (2,11,28)).
Galactosylation has been shown previously to mask Lewis X antigens (70).  [Spiro, 1971]. In addition, a cell surface galactosyltransferase that binds to type IV collagen has been described (71). While a deglycosylation/cryptic sites mechanism provides interesting speculation, it should also be noted that not all Lys residues in type IV collagen are fully hydroxylated and glycosylated (28,72), and thus receptor interaction may just occur with the sub-population of type IV collagen that does not contain carbohydrate.
CD44/CSPG interaction with α1(IV)1263-1277, and subsequent promotion of signaling and spreading activities, is dependent upon triple-helical conformation and level of glycosylation. However, the role of CD44 in tumor cell invasion is just beginning to be unraveled (65,73). CD44 and several isoforms have been characterized on a variety of tumor cell surfaces (74)(75)(76)(77), and have been suggested to be prognostic indicators of malignant melanoma (78,79). Although CD44 binds to types I, IV, VI, and XIV collagen, it is not a primary receptor for cell adhesion to collagen (60,(80)(81)(82). The CD44 cytoplasmic domain binds to ankyrin and members of the ezrin-radixin-moesin (ERM) family of cytoskeletal proteins (83). CD44 is also directly linked to two Tyr kinases, p185 HER2 and c-Src kinase (83). We have found that signaling via the CD44/α1(IV)1263-1277 interaction results in autophosphorylation of p125 FAK (14), while others have shown that CD44 mediates phosphorylation of ZAP-70 and the activation of PLCγ, Ras, PKCζ, and NF-κB binding activity (84). One result of CD44 "outside-in" signaling is up regulation and activation of integrins (85) and the expression of matrix metalloproteinases (61). 2 It is possible that CD44 works in concert with another receptor, such as the α2β1 integrin, to efficiently bind to type IV collagen and subsequently upregulate cell signaling pathways. Amongst the products of these pathways are proteases and growth factors that aid in compromising the basement membrane. Such a mechanism is consistent with our "collagen structural modulation" model previously proposed for tumor cell invasion (4), and will be explored in future studies.

ACKNOWLEDGMENTS
We thank Dr. Barbara Mueller for providing the M14, M14#5, and M14#11 melanoma cell lines.               integrin subunit mAb. Lane 3 contains proteins eluted by EDTA + NaCl from the glycosylated THP column and immunoprecipitated with an anti-β1 integrin subunit mAb. Lane 4 contains lysis buffer immunoprecipitated with an anti-β1 integrin subunit mAb, which serves as a negative control. In lane 2, one protein of ~145 kDa, corresponding to the β1 integrin subunit, was immunoprecipitated. In lane 3, no proteins were immunoprecipitated from the column. In lanes 2-4, the anti-β1 integrin subunit mAb appears at ~50 and ~25 kDa.