Differential Modulation of Human Melanoma Cell Metalloproteinase Expression by α2β1 Integrin and CD44 Triple-helical Ligands Derived from Type IV Collagen*

Tumor cell binding to components of the basement membrane is well known to trigger intracellular signaling pathways. Signaling ultimately results in the modulation of gene expression, facilitating metastasis. Type IV collagen is the major structural component of the basement membrane and is known to be a polyvalent ligand, possessing sequences bound by the α1β1, α2β1, and α3β1 integrins, as well as cell surface proteoglycan receptors, such as CD44/chondroitin sulfate proteoglycan (CSPG). The role of α2β1 integrin and CD44/CSPG receptor binding on human melanoma cell activation has been evaluated herein using triple-helical peptide ligands incorporating the α1(IV)382–393 and α1(IV)1263–1277 sequences, respectively. Gene expression and protein production of matrix metalloproteinases-1 (MMP-1), -2, -3, -13, and -14 were modulated with the α2β1-specific sequence, whereas the CD44-specific sequence yielded significant stimulation of MMP-8 and lower levels of modulation of MMP-1, -2, -13, and -14. Analysis of enzyme activity confirmed different melanoma cell proteolytic potentials based on engagement of either the α2β1 integrin or CD44/CSPG. These results are indicative of specific activation events that tumor cells undergo upon binding to select regions of basement membrane collagen. Based on the present study, triple-helical peptide ligands provide a general approach for monitoring the regulation of proteolysis in cellular systems.

molecule for type IV collagen and is used to contract and remodel type I collagen (17,20,21). The ␣ 2 ␤ 1 integrin also participates in migration of metastatic melanoma on type IV collagen (22). The ␣ 2 ␤ 1 integrin up-regulates the ␣ isoform of p38 MAPK in a three-dimensional collagen matrix, resulting in stimulation of MMP-13 and type I collagen synthesis (23). Engagement of the ␣ 2 ␤ 1 integrin has also been linked to induction of MMP-1 and MMP-14 (24,25) and implicated in MMP-9 regulation (26). Furthermore, MMP-1 binds to the ␣ 2 ␤ 1 integrin via the I domain of the ␣ 2 subunit (27).
The cell surface proteoglycans are a second class of receptors implicated in tumor metastasis. Melanoma cells have been shown to possess at least two distinct cell surface chondroitin sulfate proteoglycans, CD44 and the melanoma-associated proteoglycan/melanoma chondroitin sulfate proteoglycan (4,28). CD44 binds directly to type IV collagen, and this binding is dependent upon chondroitin sulfate (CS) (21). In general, melanoma cell invasion of the basement membrane is inhibited by the selective removal of the CS (21). The nonspliced variant of CD44, CD44s (also known as CD44H), is the sole form of CD44 expressed by human melanoma cells (29).
Several lines of evidence indicate that CD44s plays an important role in melanoma progression. Up-regulation of CD44s mRNA transcription and cell surface expression has been found in highly metastatic melanoma as compared with lowly metastatic melanoma or nontransformed melanocytes (29 -32). Enhanced expression of CD44s is also found in endothelial cells within the vasculature of tumors compared with endothelial cells from normal tissue (33). Transfection of CD44 isolated from metastatic cells into nonmetastatic tumor cells results in conferment of metastatic behavior (34). In addition, tumor cells expressing CD44s display accelerated tumor growth and metastatic spread in immunodeficient mice compared with parental cells (34), whereas a knockout of the CD44 gene virtually eliminates metastasis in mice (35). Approaches that interfere with CD44 binding to its ligand, such as administration of high molecular weight hyaluronic acid (HA), anti-CD44 mAb, or a CD44-receptor globulin, reduce tumor formation in the lung for animal models established from CD44-expressing tumor cell lines (36). Finally, proteolytic removal of CD44 inhibits the growth of primary tumors and curtails metastasis in a mouse B16 melanoma model (37).
Cell Adhesion-Cell adhesion assays were performed to quantify cell binding to ligands. Melanoma cell adhesion to substrate-coated Pro-Bind TM 96-well plates (B & D Biosciences) was performed as described previously (44,55). Lipidated peptides dissolved in PBS were diluted in 70% ethanol, added to the 96-well plate, and allowed to adsorb overnight at room temperature with mixing. Nonspecific binding sites were blocked with 2 mg/ml bovine serum albumin in PBS for 1 h at 37°C. The cells were released with 5 mM EDTA in PBS, washed twice with adhesion medium (20 mM HEPES, 2 mg/ml bovine serum albumin in RPMI 1640) and labeled with 5-or 6-carboxyfluorescein diacetate. Unincorporated fluorophore was removed by repeated washings with adhesion medium. The cells were then resuspended in adhesion medium and added to the plate. The plate was incubated 60 min at 37°C. Nonadherent cells were removed by washing three times with adhesion medium. Adherent cells were lysed with 0.2% SDS and quantified using excitation ϭ 485 nm and emission ϭ 538 nm with a SpectraMAX Gemi-niEM 96-well plate spectrafluorometer and SoftMax Pro 4.3LS software (both from Molecular Devices). Induction of Melanoma Cells-Pro-Bind TM 96-well assay plates were conditioned at room temperature overnight prior to initiation of the induction experiment with the appropriate concentration of ligand (see "Results"). The plates were then blocked by adding 2 mg/ml bovine serum albumin in PBS and incubated overnight at room temperature. M14 cultures used for induction experiments were typically 60 -80% confluent before release from growth flasks with PBS containing 5 mM EDTA, pH 7.3. Subsequent to release, the cells were washed with adhesion media (RPMI 1640 containing 20 mM HEPES) and seeded at ϳ7,500 cells/well. This cell density had been previously shown to produce efficient cell adhesion with minimization of cell-cell contacts (44, 56 -59). Aliquots of the melanoma-conditioned media were harvested at regular intervals over a 24-h period for later determination of metalloproteinase levels. Total RNA (ϳ1-2 g) was isolated at regular intervals over a 24-h period with a S.N.A.P. TM RNA eukaryotic total RNA isolation kit and treated with DNase I to remove genomic DNA, as directed by the manufacturer (Invitrogen). Both sets of samples were stored at Ϫ20°C until analyses could be performed. To evaluate changes in gene expression for cells grown on peptide amphiphiles, RNA was isolated from 90,000 cells/time point up to 24 h. RNA yield was determined by measuring the A 260 of an aliquot of the final preparation. Typical yields, on the order of 1-2 g of total RNA, provided ample mRNA for multiple cDNA syntheses.
Semi-quantitative RT-PCR-All of the reagents were from Invitrogen unless otherwise stated. DNase I-treated, total RNA isolated above was reverse transcribed in a total of 20 l containing 125 ng of RNA following the manufacturer's recommendations for SUPERSCRIPT II. Primer-dropping PCR was performed on 10% of the resulting cDNA in a total of 50 l following the manufacturer's recommendations for Platinum® Taq DNA polymerase. A typical PCR amplification cycle consisted of 30 s of melting at 94°C, annealing at an optimized temperature (approximately equal to T m (primers) Ϫ 10°C) for 30 -45 s, and extension at 72°C for 60 s. Platinum® Taq DNA polymerase was activated for 2 min at 94°C prior to the first amplification cycle. Annealing temperatures, total amplification cycles performed, and amplicon size are listed in Table I for the primers used in this study (see also Refs. 60 and 61). "Primer dropping" PCR was accomplished by providing an internal GAPDH control to co-amplify with the target gene for the last 20 or 21 cycles of the total number of amplification reactions. PCR products were electrophoresed in 3% agarose gels at 90 V for 45 min and stained with SYBR green I (BioWhittaker Molecular Applica tions). Images of stained gels were captured on a Fluor-S™ MutiImager and quantified using Quantity One® v4.3.0 software package. Microsoft Excel 98 was used to perform calculations and graphically represent trends in metalloproteinase expression. Briefly, this consisted of first determining the target to GAPDH ratio followed by normalization to the initial time point. PCR amplicons were gelpurified using an Invitrogen SNAP gel purification kit. The isolated DNA was sent to the Biotechnology Program at the University of Florida for sequencing.
Quantitative RT-PCR-Quantitative PCR was performed on a Roche Applied Sciences LightCycler. All of the reagents were from Invitrogen unless otherwise stated. Custom primers were designed using an Invitrogen LUX TM Designer for MMP-2, -8, and -9. Primer sequences and melting temperatures are located in Table I. Amplifications utilized Platinum® Quantitative PCR SuperMix-UDG following the manufacturer's recommendations. The cycling program was UDG inactivation at 50°C for 2 min, 95°C for 2 min, followed by 45 cycles of (a) 5 s denaturation at 94°C, (b) annealing at 60°C for 10 s, and (c) extension at 72°C for 15 s. Melting curve analysis was performed using continuous acquisition and a slope of 0.1°C/s. The human GAPDH-certified LUX TM primer set was used as a housekeeping control for all amplifications.
Melanoma Cell-specific Protein Expression-Conditioned medium was isolated by withdrawal of the media from growing cells, centrifuging at 1000 ϫ g (to remove any floating cells). For MMP-1 and MMP-3, ELISA was performed as described (62) using appropriate antibodies purchased from Chemicon. MMP-8 and MMP-13 ELISA kits were obtained from R & D Systems (Minneapolis, MN). ELISA was used to measure protein expression in the conditioned media from cells grown on surfaces coated with lipidated peptides.
Cellular MMP Assays-Two assays were utilized to evaluate active MMP production by melanoma cells. The first assay involved MMP immobilization (106). A 96-well plate was incubated with the appropriate MMP mAb for at least 18 h at 4°C. Nonspecific binding sites were blocked by incubating with phosphate-buffered saline containing 0.05% Tween TM 20 and 2 mg/ml bovine serum albumin for at least 4 h at 4°C and washed three times with enzyme assay buffer (50 mM Tricine, 50 mM NaCl, 10 mM CaCl 2 , 0.05% brij-35). Either MMP standards or CCT unknown samples were added to each well, and the plate was mixed for at least 18 h at 4°C. All of the liquid was removed, and 2 mM 4-aminophenylmercuric acetate was added to each well (where applicable). This was followed by a 2-h incubation at 37°C. The wells were washed three times with enzyme assay buffer, and the appropriate fluorogenic substrate was added to each well. The plate was incubated at 37°C in a humidified atmosphere for at least 18 h. Fluorescence readings ( excitation ϭ 325 nm and emission ϭ 393 nm) were taken at appropriate intervals, and a standard curve was created by plotting the percentage of increase in fluorescence versus concentration of active enzyme. Substrate cleavage was found only for the specific mAb and the corresponding MMP; cross-reactivity between MMPs of similar function (i.e. MMP-1 and MMP-13) was not observed.
The second assay was based on hydrolysis of fluorogenic substrates in solution by conditioned media (41,55). One volume of the appropriate fluorogenic substrate was added to each well in a 384-well plate. Where applicable, EDTA was added to each well. The plate was incubated at 30°C in a humidified atmosphere for 30 min, and one volume of conditioned media, adhesion media (as a negative control), or MMP (as a positive control) was added to each well. The plate was incubated at 30°C in a humidified atmosphere for at least 18 h. Fluorescence readings ( excitation ϭ 325 nm and emission ϭ 393 nm) were taken and a standard curve created by plotting the increase in fluorescence versus concentration of MMP standard. This standard curve can be used to calculate the active enzyme concentration in the conditioned media.
Protein Electrophoresis/Gelatin Zymography-Proteins in the cellconditioned medium were resolved by SDS-PAGE with 4 -15% pre-cast Tris⅐HCl gels (Bio-Rad) and silver stained (BioRad). Images of the gels were captured and analyzed on a BioRad Fluor-S TM MutiImager using Quantity One®, version 4.3.0 software package. To prepare the conditioned medium samples, the samples were first concentrated 15 times using spin concentrators (10,000 molecular weight cut-off) at 3000 rpm for 90 min. Fifteen l of sample was mixed with 5 l of 2ϫ sample buffer (0.125 M Tris⅐HCl, pH 6.8, 4% (w/v) SDS, 20% (v/v) glycerol). Fifteen l was loaded and then electrophoresed for 1 h at 180 V on SDS-polyacrylamide gels containing 1 mg/ml gelatin (type A, porcine skin). After electrophoresis, the gels were washed once with deionized water, three times with 2.5% (w/v) Triton X-100 to remove SDS, and three times with the collagenase buffer (50 mM Tris⅐HCl, 200 mM NaCl, 5 mM CaCl 2 , and 0.1% Brij-35, pH 7.6). The gels were then incubated in collagenase buffer at 37°C for 16 h, stained with Coomassie brilliant blue, and destained in 20% methanol and 10% acetic acid until the lysis zones were clear against a blue background. Inhibition studies to determine protease class were performed by incubating gels in collagenase buffer containing 1 mM phenylmethanesulfonyl fluoride, 1 mM N-ethylmaleimide, or 0.1 mM 1,10-phenanthroline.

TABLE II
Relative induction of target genes by the ␣ 2 ␤ 1 integrin and CD44/CSPG triple-helical ligands NM, negligible modulation; ND, not determined; ϩ indicates relative up-regulation; Ϫ indicates relative down regulation; ϩ 3 Ϫ indicates up-regulation followed by down-regulation; Ϫ 3 ϩ indicates down regulation followed by up-regulation.

Target
Receptor engaged mRNA Protein Active enzyme

RESULTS
The present study has examined the changes induced by melanoma ␣ 2 ␤ 1 integrin and CD44/CSPG binding using well defined "peptide amphiphile" substrates. Peptide amphiphiles consist of peptide sequences covalently linked to hydrocarbon chains and are designed to house discrete binding sites (such as those found in various ECM proteins) (41, 44, 47, 56 -59, 63, 64). These small biomolecules assume stable "mini-protein" structures (␣-helical or triple-helical) at physiological temperatures, allowing for the study of cellular behaviors in response to discreet binding sites that maintain the three-dimensional structure of the parent protein. CD spectroscopic analysis of two of the peptide amphiphiles used in the present study, THP-B and THP-D (Fig. 1), indicated T m values of 55.0 and 45.0°C, respectively (data not shown). Thus, both peptide amphiphiles were triple-helical under assay conditions. SSP-A and SSP-C (Fig. 1) did not form triple-helical structures (data not shown). Assays were then performed to determine the ligand concentrations at which cell adhesion was optimal and equivalent. The ligand concentrations that provided high levels (Ͼ80%) of comparable adhesion were 100 M for SSP-A, 10 M for THP-B, 50 M for SSP-C, and 10 M for THP-D (data not shown). These ligand concentrations were used for all subsequent induction experiments.
Expression profiles were initially examined for the M14 melanoma cell line induced by the ␣ 2 ␤ 1 integrin specific THP-B and the linear peptide model SSP-A. RT-PCR analysis was performed for several members of the MMP family (MMP-1, MMP-2, MMP-3, MMP-8, MMP-9, MMP-13, and MMP-14), heparanase A, and several members of the ADAMTS family (ADAMTS-1, ADAMTS-2, ADAMTS-3, and ADAMTS-4). Semiquantitative PCR amplicons were gel-purified and sequenced, whereas real time PCR amplicon sequences were verified by melting curve analysis. DNA sequences matched the respective gene sequences in GenBank TM (data not shown). The expression levels of heparanase A and ADAMTS-1, -2, -3, and -4 gene products remained unchanged relative to the internal GAPDH standard (data not shown). However, melanoma cells displayed differential expression of several MMPs relative to an internal GAPDH control ( Fig. 2A and Table II) and as compared with a linear version of the ligand (data not shown). The expression of MMP-8 ( Fig. 2A) and MMP-9 (data not shown) was unaffected by exposure to the triple-helical ligand, whereas the expression of MMP-3 increased slightly over the initial expression levels (1.8-fold after 6 h; data not shown). In contrast, the relative expression of MMP-1 increased 2-fold by 12 h, whereas MMP-14 expression increased 2-fold after 2 h, decreased by 6 h, and then slowly increased by 12 and 24 h ( Fig. 2A). MMP-2 expression was substantial at 1 h, decreased by 2-fold after 4 h, and then increased back to initial levels after 8 h (data not shown). The relative expression of MMP-13 increased nearly 2.5-fold by 12 h and continued over the induction period until a final value of over 4-fold maximum expression by the 24-h time point was achieved (Fig. 2A). The mounting expression of the MMP-13 gene may be due to either prolonged gene activation or by additional stimulation by type IV collagen normally pro- Expression profiles were subsequently examined for the M14 melanoma cell line induced by the CD44-specific THP-D and the linear peptide model SSP-C. RT-PCR analysis was performed for several members of the MMP family (MMP-1, MMP-2, MMP-3, MMP-8, MMP-9, MMP-13, and MMP-14). Melanoma cells displayed differential expression of several MMPs relative to an internal GAPDH control (Table II) and as compared with a linear version of the ligand. MMP-8 expression, for example, was stimulated 3.5-fold by THP-D after 6 h (Figs. 2B and 3) and was only slightly stimulated by the linear SSP-C (Fig. 3). MMP-1 showed a slight sustained increase in relative expression, whereas MMP-14 was up-regulated at 2 h and then sharply down-regulated at 6 and 8 h (Fig. 2B). MMP-13 expression showed a slight increase at 6 h (Fig. 2B). MMP-2 modulation followed a similar trend as for the ␣ 2 ␤ 1 integrin ligand, in that initial expression levels decreased over time and then increased at 8 h (data not shown).
Protein expression was next examined for MMPs whose mRNA levels were modulated by either the ␣ 2 ␤ 1 integrin or CD44 ligand. Conditioned media from M14 melanoma cells grown on the ␣ 2 ␤ 1 integrin-specific THP-B or CD44/CSPGspecific THP-D were subjected to ELISA analyses ( Fig. 4 and Table II). For MMP-2 and MMP-14, ELISA was found to be unreliable, so fluorogenic substrates selective for MMP-2/ MMP-9 (41) and MMP-14 (66) were used for evaluation of active enzyme (see below). MMP-1 was slightly induced by both ligands, with increases in the range of ϳ1-2 ng/ml over 24 h (Fig. 4, A and B). ELISA for MMP-8 induction showed a slight decrease in protein because of ␣ 2 ␤ 1 integrin binding (ϳ1 ng/ml) and an increase in protein from CD44 engagement (ϳ6 ng/ml) (Fig. 4, A and B), consistent with relative mRNA expression (Fig. 2). ELISA for MMP-13 showed a substantial increase in protein in response to the ␣ 2 ␤ 1 integrin ligand and a lesser increase in response to the CD44 ligand (Fig. 4C). The relative MMP-13 protein levels induced by the ␣ 2 ␤ 1 integrin (ϳ60 pg/ ml) and CD44 (ϳ25 pg/ml) ligands were consistent with mRNA expression (Fig. 2).
The prior analyses have revealed trends in MMP mRNA and protein expression but have not addressed the production of active enzyme. MMPs initially exist as zymogens and may also be rendered inactive by binding to TIMPs or other, more general protease inhibitors. For MMP-1, active enzyme was quantified by solid phase mAb immobilization of the enzyme followed by reaction with the fluorogenic substrate fTHP-4 (49). The solid phase assay showed more activity induced by the ␣ 2 ␤ 1 integrin than by CD44 (Fig. 5). Treatment of samples with an activator of proMMPs (4-aminophenylmercuric acetate) resulted in a further increase in MMP-1 activity (Fig. 5), indicative of the production of both MMP-1 and proMMP-1 by engagement of either the ␣ 2 ␤ 1 integrin or CD44. For other MMPs, calibration of the solid phase assay was problematic and not further pursued.
An additional approach to monitor active enzyme is the use of discriminatory substrates. Unfortunately, such substrates do not yet exist for MMP-1, MMP-8, and MMP-13. General triple-helical peptidase activity can be evaluated using fTHP-4. Engagement of either the ␣ 2 ␤ 1 integrin or CD44 resulted in significant triple-helical peptidase activity detected in melanoma cell conditioned media (Fig. 6A), with greater activity found in response to CD44. This activity was completely inhibited by EDTA (data not shown), suggesting metalloproteinase activity (41). The MMP-14 selective substrate C 10 -(Gly-Pro-Hyp) 5 -Gly-Pro-Lys(Mca)-Gly-Pro-Gln-GlyϳCys(4-methoxybenzyl)-Arg-Gly-Gln-Lys(Dnp)-Gly-Val-Arg-(Gly-Pro-Hyp) 5 -NH 2 was used for comparison with general triple-helical peptidase activity. Soluble MMP-14 activity, which can be generated by nonautocatalytic shedding of MMP-14 (67,68), was significant at early time points and then decreased over 8 h in response to the ␣ 2 ␤ 1 integrin and CD44 ligands (Fig. 6B). The MMP-14 activity profiles correlate well with the mRNA expression profiles in that MMP-14 is induced at early time points. The subsequent decrease in MMP-14 activity may be due to degradation of MMP-14. Other MMP activity profiles do not decrease at later time points (Fig. 6, A and C), and exogenous MMP activity is not significantly effected over a 24-h period by melanoma conditioned media (106). These results suggest that degradation of MMP-14 is specific, which is not surprising considering the multitude of MMP-14 shedding processes (68).

DISCUSSION
Engagement of the ␣ 2 ␤ 1 integrin and CD44 by ECM ligands results in intracellular signaling (25,69,70). However, the precise nature of this signaling within melanoma cells has not been documented. MMP-1, -2, -3, -8, and -13 are expressed in numerous human melanoma cell lines but not melanocytes, with MMP-2 and MMP-8 expression at higher constitutive levels than other MMPs (71,72). Conversely, MMP-14 is expressed in both melanocytes and melanoma (71). The present study has used ␣ 2 ␤ 1 integrin-and CD44-specific ligands to examine the regulation of several members of the MMP family associated with melanoma cell invasion.
Triple-helical ligands for the ␣ 2 ␤ 1 integrin and CD44/CSPG promote expression of relatively low levels MMP-1 mRNA and protein (Table II and Figs. 2 and 4, A and B). However, more active MMP-1 was found in response to ␣ 2 ␤ 1 integrin engagement (Fig. 5). This result may be related to increased MMP-3, which is an activator of proMMP-1 (73), seen in response to the ␣ 2 ␤ 1 integrin ligand. The triple-helical ligand for the ␣ 2 ␤ 1 integrin also promotes expression of MMP-13 and MMP-14 (Table II and Figs. 2A, 4C, and 6B). The profile seen here in response to ␣ 2 ␤ 1 integrin engagement is consistent with high invasion potential, in that the MMPs up-regulated can participate in the dissolution of basement membrane (type IV) and type I collagen.
Examination of gene expression by RT-PCR analysis indicated initial down-regulation, followed by up-regulation, of MMP-2 by either the ␣ 2 ␤ 1 integrin or CD44 (Table II), whereas enzyme activity is clearly increased (Figs. 6C and 7). Prior studies had shown that neither type I collagen nor HA induce MMP-2 expression in melanoma cells (74). The results seen here appear to represent high constitutive levels of MMP-2, as previously reported for human metastatic melanoma (74). Enhanced MMP-2 activity could also be achieved via MMP-14 upregulation as well as TIMP modulation, because MMP-14 and TIMP-2 participate in proMMP-2 activation (73). MMP-14 mRNA was initially up-regulated by both the ␣ 2 ␤ 1 integrin and CD44 ligands (Fig. 2), whereas soluble active enzyme was significantly present up to 4 h (Fig. 6B).
CD44 binding to two different ligands, ␣1(IV)1263-1277 and HA, has been well characterized (34,44,75). HA binds to the amino-terminal globular domain of CD44, using motifs of two basic amino acids separated by seven non-acidic amino acids (B[X7]B) (34). These motifs are found within CD44 residues 21-45 (34). Several distal residues also contribute to HA binding, such as Lys 158 and Arg 162 (34). CD44 binding to type IV collagen and ␣1(IV)1263-1277 is dependent upon CS (21, 44). CS interaction with collagen requires a minimum oligosaccharide chain length and distinct sulfation pattern (76). Conversely, CS interferes with CD44 binding to HA (34). It appears that ␣1(IV)1263-1277 and HA bind to different regions of CD44, and thus these ligands may induce different CD44 signaling pathways. Consistent with this proposal, HA signaling through FAK and extracellular signal-regulated kinase 1/2 results in MMP-9 secretion in glioblastoma cells (77); in turn, extracellular signal-regulated kinase pathway activation results in elevated expression of MMP-3, MMP-9, MMP-14, and CD44 (78). These HA-induced signaling outcomes are clearly different from the results in the present study. However, it should be noted that the identity of the receptor in the prior HA binding study (77), either CD44 or receptor for hyaluronanmediated motility, was not determined. Regardless, it will be of great interest to determine whether different CD44 extracellular environments result in different cellular responses.
The in vitro activation of melanoma cells with selective ligands serves as a means to dissect specific pathways resulting from engagement of a particular receptor. Triple-helical ligands for CD44 and the ␣ 2 ␤ 1 integrin promote expression of MMPs with differing downstream events. The intracellular signaling pathways engaged by these individual receptors have been somewhat defined but do not explain the differing downstream events. For example, FAK induction has long been established as an early signaling event occurring as a result of ␣ 2 ␤ 1 integrin ligation by collagen (79), and we described previously the induction of FAK following binding of CD44/CSPG to triple-helical ␣1(IV)1263-1277 (46). Thus, different signaling outcomes may well be dictated by the levels, timing, and modulation of FAK phosphorylation, as well as the pathways leading to FAK phosphorylation. The overall signaling results may be indicative of specific activation events that the tumor cell undergoes upon binding to select regions of basement membrane collagen. Linear constructs do not promote significant levels of MMP expression, further supporting the notion that triple-helical conformation selectively modulates melanoma cellular activity (46,47,55).
It is possible that CD44 works in concert with the ␣ 2 ␤ 1 integrin to efficiently bind to type IV collagen and subsequently up-regulate cell signaling pathways. Melanoma-associated proteoglycan/melanoma chondroitin sulfate proteoglycan has been shown to mediate ␣ 4 ␤ 1 integrin binding to fibronectin (80). Prior studies showed that removal of cell surface CS by chondroitinase ABC digestion inhibits ␤ 1 -integrin-dependent adhesion to type IV collagen (81). The CD44 ␣1(IV)1263-1277 ligand had the same effect as chondroitinase ABC, implicating CD44/CSPG as the receptor working in concert with the integrin. Adhesion was reestablished by a ␤ 1 -integrin subunit stimulating mAb. Other studies have also shown interactions between membrane-bound carbohydrates and integrins. Tumor gangliosides enhance ␣ 2 ␤ 1 integrin-dependent platelet activation upon binding to collagen (82). Several transmembrane proteins are known to associate with the ␣ 2 ␤ 1 integrin (5,83). A CD44/␣ 2 ␤ 1 integrin interaction may be indirect, because the CD44 cytoplasmic domain binds to ankyrin, which in turn is linked to actin by fodrin/spectrin (84,85). Another possible interaction between the ␣ 2 ␤ 1 integrin and CD44 relates to prior studies showing that CD44 can associate with active MMP-14 at the tumor cell migration front, potentially enhancing invasion of the basement membrane (86). MMP-14 can shed CD44, and the intracellular CD44 fragment subsequently generated by presenilin/␥-secretase action may function as a transcription factor (70,86). In turn, the soluble CD44/CSPG may modulate tumor growth and spreading (87,88) in an analogous fashion to other liberated tumor cell surface proteoglycans (89,90). Because MMP-14 is initially up-regulated by the ␣ 2 ␤ 1 integrin, this integrin may modulate CD44 shedding and subsequent signaling by liberated CD44 domains.
The MMP proteolytic profiles identified here can in turn lead to differential processing of type IV collagen. For example, the sites of hydrolysis within type IV collagen are different for  1 and 4). M14 melanoma cells were allowed to adhere to 10 M THP-B or THP-D for 60 min at 37°C. Three washes of adhesion medium were used to remove nonadherent cells and cells were grown for 12 (lanes 1 and 2) or 24 (lanes 4 and 5) h. Cell supernatants were added either without treatment or following 100 M 1,10-phenanthroline treatment (data not shown). Lane 3 is an activated MMP-2 standard, which has been processed to the 50-kDa form. The majority of gelatinolytic activity corresponded to MMP-2 in the 52-and 50-kDa active forms. A small amount of gelatinolytic activity was seen at 81 kDa, which corresponded to active MMP-9 (standard not shown).
MMP-2/MMP-9 compared with MMP-3 (91,92). Generation of different cleavage products may then influence tumor progression in several ways. First, liberated type IV collagen fragments can modulate tumor angiogenesis (93). Second, proteolysis may reveal formerly "cryptic" binding sites that promote migration (94). Third, decreased triple-helical structure negatively influences binding and signaling through collagen-specific receptors, altering cellular phenotypes (47,55). Fourth, the loss of basement membrane integrity results in a barrier less resistant to invasion (95). Thus, the MMP proteolytic profiles induced by engagement of different cellular receptors (i.e. the ␣ 2 ␤ 1 integrin and CD44/CSPG) may have profound consequences on metastatic potential.
It is noteworthy to consider how the activities induced by the ␣ 2 ␤ 1 integrin and CD44 ligands compare with either constitutive or type IV collagen-induced melanoma behaviors. We have recently examined the constitutive production of active MMPs by M14 melanoma cells and found that levels of active MMP-1, MMP-3, MMP-13, and soluble MMP-14 are low (106). As documented here, the engagement of either the ␣ 2 ␤ 1 integrin or CD44 specifically increases active protease production. Prior studies had shown that binding of the ␣ 2 ␤ 1 integrin to collagen results in increased production of MMP-1, MMP-13, and MMP-14 (23,25,96), analogous to the MMP profile induced by the ␣ 2 ␤ 1 integrin triple-helical ligand. The present study provides the first correlation between CD44 binding and MMP production (see prior discussion). Human melanoma cell binding to type IV collagen induces an inositol 1,4,5-trisphosphateindependent release of intracellular Ca 2ϩ stores (97); inhibition of this release decreases expression of MMP-2 (98). In related studies with human fibrosarcoma cells, calcium ionophores were shown to decrease MMP-9 expression and MMP-2 activation but have no effect on MMP-14 expression (99). The overall effect of melanoma binding to type IV collagen is increased MMP-2 and MMP-9 but no modulation of MMP-14. This clearly differs from the ␣ 2 ␤ 1 integrin and CD44 profiles reported here. However, correlation between melanoma binding to type IV collagen and triple-helical ligands is complicated by the multidomain nature of type IV collagen. The type IV collagen 7S, triple-helical, and NC1 domains are bound by both overlapping and distinct receptors (12,21,93,100). This presumably leads to a multitude of signaling and regulatory pathways. One of the distinct advantages of the peptide amphiphile approach is that, initially, one receptor at a time can be studied, and then a "mini-ECM" can be assembled to examine the cumulative effects of multi-receptor binding (58,59). This will allow for the examination of cooperativity between the ␣ 2 ␤ 1 integrin and CD44/CSPG. In addition, peptide amphiphiles can be used to probe receptor-mediated events in co-culture systems, further defining the stimuli for cross-talk within the melanoma microenvironment (101)(102)(103)(104). Peptide amphiphile ligands ultimately provide a general approach for monitoring receptor-mediated regulation of proteolysis in cellular systems.