Cell surface chondroitin sulphate proteoglycan 4 (CSPG4) binds to the basement membrane heparan sulphate proteoglycan, perlecan, and is involved in cell adhesion

Abstract

Chondroitin sulphate proteoglycan 4 (CSPG4) is a cell surface proteoglycan highly expressed by tumour, perivascular and oligodendrocyte cells and known to be involved cell adhesion and migration. This study showed that CSPG4 was present as a proteoglycan on the cell surface of two melanoma cell lines, MM200 and Me1007, as well as shed into the conditioned medium. CSPG4 from the two melanoma cell lines differed in the amount of chondroitin sulphate (CS) decoration, as well as the way the protein core was fragmented. In contrast, the CSPG4 expressed by a colon carcinoma cell line, WiDr, was predominantly as a protein core on the cell surface lacking glycosaminoglycan (GAG) chains. This study demonstrated that CSPG4 immunopurified from the melanoma cell lines formed a complex with perlecan synthesized by the same cultured cells. Mechanistic studies showed that CSPG4 bound to perlecan via hydrophobic protein–protein interactions involving multiple sites on perlecan including the C-terminal region. Furthermore, this study revealed that CSPG4 interacted with perlecan to support cell adhesion and actin polymerization. Together these data suggest a novel mechanism by which CSPG4 expressing cells might attach to perlecan-rich matrices so as those found in connective tissues and basement membranes.

Proteoglycans are present in intracellular granules, on the cell surface, within the pericellular spaces and in the extracellular matrix (ECM), where they mediate cell–cell and cell–ECM interactions (1). Proteoglycans can bind to various factors, including growth factors and cytokines, which are involved in cell signalling, and thus affecting cell motility, adhesion and proliferation (2). Both the cell surface proteoglycan, chondroitin sulphate proteoglycan 4 (CSPG4), and the basement membrane proteoglycan, perlecan, have been shown to mediate cell–ECM interactions (3, 4).

CSPG4, also known as NG2 and human high molecular weight melanoma associated antigen, is a highly glycosylated integral membrane proteoglycan decorated with a chondroitin sulphate (CS) chain (5). As a transmembrane proteoglycan, CSPG4 has the potential to interact with both extracellular and cytoplasmic binding partners. This suggests that CSPG4 can mediate intracellular signalling downstream of growth factor receptor and integrin interactions, potentiating communication between the extracellular and intracellular compartments of the cell (6, 7). Although CSPG4 is known as a transmembrane proteoglycan, it can also be shed into the ECM (8, 9). CSPG4 potentiates cell adhesion (10), proliferation (5, 11), migration (10, 12, 13) and modulates responses to growth factors (6, 14, 15). CSPG4 can either act directly as an adhesion molecule or indirectly affecting focal contact formation by acting in concert with integrins such as α4β1 and α3β1 (10, 16, 17). CSPG4 can bind to type VI collagen, which is widely distributed in connective tissues (15, 18). CSPG4 can also help to promote the formation of initial adhesive contacts at the leading edge of migrating cells (5) and is involved in cell migration on ECM components including collagen and fibronectin (10, 12, 13).

Perlecan, is a major proteoglycan present in basement membranes and connective tissues (4) and contains binding sites for a range of molecules, including basement membrane components (19), cell adhesion molecules (20) and growth factors (21). Binding with these ligands suggests that perlecan has a broad and diverse range of biological functions including cell adhesion, proliferation and migration (4, 21). Down regulation of perlecan mRNA with antisense RNA reduced the proliferation of melanoma, colon carcinoma and Kaposi’s sarcoma cells (22–24). This reduced proliferation was thought to be linked with reduced angiogenesis, as reduced blood vessel growth was observed in these samples (22). Perlecan expression has been correlated with both increased and decreased tumour cell migration. Downregulation of perlecan in Kaposi sarcoma cells reduced cell migration when coupled with growth factors like VEGF and HGF (24). This reduced migration was thought to be due to the ability of perlecan to bind and inhibit the presentation of the chemokines and growth factors required for cell migration (25–31). On the contrary, suppression of perlecan expression in human fibrosarcoma cells resulted in increased migration toward chemo-attractants, and a heightened ability to invade through the ill-defined ECM model, Matrigel (32). In this ill-defined matrix which can contain ∼1–5% murine perlecan that would confound the results, it is challenging to make any firm conclusions. Even though these studies indicate that perlecan plays an important role in modulating events in tumour progression, more defined and better controlled studies are needed to examine the precise role that perlecan plays in mediating events in cell adhesion to the ECM and subsequent cell migration.

This study demonstrates for the first time that CSPG4 binds to perlecan through protein–protein interactions and that this interaction has a role in cell adhesion and actin polymerization. These studies were performed using melanoma and colon carcinoma cells, which are known to highly express CSPG4 on their cell surfaces, however, we suggest that this interaction will be relevant to other situations where CSPG4-expressing cells are exposed to perlecan-rich matrices such as in the developing blood vessel and brain.

Materials and Methods

Materials

Chondroitinase ABC (C’ase ABC) and heparinase III (Hep III) were purchased from Seikagaku Corporation, Tokyo, Japan. Mouse monoclonal anti-CSPG4 antibody B5 (67) was grown and purified in-house. Mouse monoclonal anti-CSPG4 antibody 9.2.27 was a gift from Prof. Ralph Reisfeld, Scripps Research Institute. Mouse monoclonal antibodies against perlecan clone A74 was purchased from Abcam, Cambridge, MA, USA and clone E-6 was purchased from Santa Cruz Biotechnology, Santa Cruz, CA, USA. The polyclonal rabbit anti-human perlecan (CCN-1) (33, 68) was raised in house against immunopurified HCAEC perlecan. Anti-rabbit secondary antibodies were purchased from Merck-Millipore (Sydney, Australia). Secondary horseradish peroxidase (HRP) conjugated antibodies were purchased from Dako (Sydney, Australia). All other chemicals were purchased from Sigma-Aldrich (Castle Hill, Australia).

Cell culture

Human colon carcinoma cell line, WiDr and human melanoma cell lines, MM200 and Me1007 were cultured in DMEM medium containing 5% and 10% (v/v) foetal bovine serum (FBS), respectively, and 1% (v/v) penicillin/streptomycin at 37°C in 5% CO₂. Conditioned medium was collected every 3 days and stored at −20°C until required.

Endoglycosidase digestion

Samples were digested with 50 mU/ml C’ase ABC in 0.1 M Tris acetate, pH 8.0, at 37°C for 16 h. Samples were digested with 10mU/ml Hep III in DPBS, pH 7.2 for 16 h at 37°C.

Immunocytochemistry

The procedure was performed as previously described (33). Briefly, confluent cell cultures on chamber slides (Thermo Fisher Scientific, Australia) were fixed with 4% (w/v) paraformaldehyde containing 1% (w/v) sucrose for 15 min at 37°C, followed by washing with DPBS twice. Cells were then permeabilized with 300 mM sucrose, 50 mM NaCl, 3 mM MgCl₂, 2 mM HEPES and 0.5% Triton X-100 in deionized water, pH 7.2 for 5 min on ice. The cells were then blocked with 1% (w/v) BSA in DPBS for 1 h at room temperature. Primary antibodies, diluted in blocking solution were added to corresponding compartments and incubated at 4°C overnight. No primary (negative) controls were incubated with blocking solutions only. Cells were then washed twice with DPBS with 1% (v/v) Tween-20 (PBST) for 5 min and then incubated with Alexa Fluor® secondary antibodies in blocking solution (1:500) for 1 h at room temperature in the dark and washed twice with PBST for 5 min. The slides were then counterstained with 4′, 6-diamidino-2-phenylindole, dilactate (DAPI) (Invitrogen, Australia, 1 μg/ml) for 15 min in the dark at room temperature. After four washes with PBST washes, the cells were visualized with a confocal microscope (Olympus Fluoview FV1200).

Western blot analysis

Proteoglycan enriched samples (200 µg/ml based on Coomassie protein assay), with and without endoglycosidase digestion, were electrophoresed in 3–8% (w/v) Tris-Acetate NuPAGE^® SDS-PAGE gels (Life Technologies, Sydney, Australia) under non-reducing conditions using Tris-Tricine running buffer (50 mM Tricine, 50 mM Tris base and 0.1% (w/v) SDS in water at pH 8.3) at 160 V for 60 min. Pre-stained high molecular weight standards (HiMark standard, Invitrogen, Australia) were electrophoresed on each gel. Samples were then transferred to polyvinylidene difluoride (PVDF) membrane using transfer buffer (50 mM Bicine, 50 mM Bis-Tris, 10 mM EDTA, 0.05% (w/v) SDS, 1% (v/v) Methanol at pH 8.3) in a semi-dry blotter at 300 mA and 25 V for 60 min. The membrane was blocked with 1% (w/v) bovine serum albumin (BSA) in TBST (20 mM Tris base, 136 mM NaCl, 0.1% (v/v) Tween-20, pH 7.6) for 2 h at 25°C followed by incubation with primary antibody diluted in 1% (w/v) BSA/TBST overnight at 4°C. Membranes were subsequently rinsed with TBST, incubated with secondary HRP conjugated antibodies (1:50,000) for 60 min at 25°C, rinsed with TBST and TBS before being imaged using chemiluminescence reagent (Thermo Fisher Scientific, Sydney, Australia) and x-ray film.

Proteoglycan enrichment

Anion exchange chromatography, using a diethylaminoethyl (DEAE) column (DEAE Sepharose Fast flow) was used to purify proteoglycans from WiDr, MM200 and Me1007 cell conditioned medium. The column was equilibrated with three column volumes of DEAE running buffer (250 mM NaCl, 20 mM Tris-base, 10 mM EDTA, pH 7.4) at a flow rate of 2 ml/min. Filtered cell conditioned medium was loaded onto the column at 4°C at 2 ml/min followed by extensive washes with DEAE running buffer to remove phenol red and non-bound proteins from the column. The column was then connected to a BioLogic low-pressure chromatography system (BioRad) with a flow rate of 2 ml/min and a baseline absorbance was established with DEAE running buffer. Molecules of interest were eluted with eluting buffer (1 M NaCl, 20 mM Tris-base, 10 mM EDTA, pH 7.5) followed by regenerating buffer (2 M NaCl, 20 mM Tris-base, 10 mM EDTA, pH 7.5). The collected proteoglycan enriched samples were dialyzed against PBS and concentrated.

RT-PCR and quantitative real-time PCR (qPCR)

Total RNA was isolated from MM200, Me1007 and WiDr cells using TRI Reagent (1 ml/10⁷ cells) and then treated with DNase using the RQ1 RNase-free DNase kit (Promega, Madison, WI) to remove contaminating DNA. Subsequently, 1 μg of RNA was transcribed into cDNA using oligo (dT) primer mixer (ProtoScript^® M-MuLV First Strand cDNA synthesis kit, NewEngland Labs, GeneSearch Pty. Ltd., Arundel, Australia) and amplified during 35 cycles by PCR utilizing CSPG4 and HSPG2 primers that amplified regions of the genes that encode for various regions of the proteins. The reactants were cycled at 95°C for 1 min, 60°C for 1 min and 72°C for 1.5 min to enable denaturation, annealing and extension, respectively. PCR products were then separated on 1.5% (w/v) agarose gel at 60 V for 1 h in TBE buffer (89 mM Tris base, 89 mM boric acid and 2 mM EDTA, pH 8). The gels were stained with GelRed (Jomar Diagnostics, Sydney, Australia) for 30 min and then visualized under UV light. The primers used for CSPG4 and HSPG2 are indicated in Table I.

Table I.

Primers used for PCR amplification of CSPG4 and HSPG2 cDNA

Gene	Domain	Primer sequences (5′–3′)	PCR product size (bp)
CSPG4	I	F: CACTCAGGACGAAGGAACCC	964
		R: GTATGTTTGGCCCCTCCGAA
	II	F: AACTACAGGGCACAAGGCTG	997
		R: TGGTAGTGGACCTCATCCCC
	III	F: CCTTGCTGTGGCTGTGTCTT	760
		R: GATGGCGGATGGTAGGATGT
	IV	F: TGTTCAGCGTCATCATCCCC	928
		R: CCCATCCCCAGAGCAACCTA
HSPG2	I	F: ACCTGGGCAGTGGGGACCTG	403
		R: GCCTCCGTGCAGGCTCTTGG
	III	F: ACCCCACCTGTGATGCGTGCTC	796
		R: GTGGCCCGGATCAGGAGCTCAT
	V	F: GTGTCAGTGAATGGCAAACG	578
		R: CTGAAGACAAGGTGCCCG

For qPCR, 50 ng of cDNA was mixed with 100 nM of forward and reverse primers and 10 μl of Power SYBR Green 2× PCR Master Mix (Applied Biosystems, Mulgrave, Australia), and toped up to 20 μl/sample with RNase-free water. Samples were subject to 35 reaction cycles using the ABI StepOne^TM real-time PCR system.

Mass spectrometry

Immunopurified samples (50 μg/ml) were prepared for LC-MS² analysis by in-solution digestion which involved reduction with 10 mM DTT for 10 min at 95°C and alkylation with 25 mM IAA for 20 min at 25°C. Samples were then incubated with 20 µg/ml sequencing grade Trypsin in 50 mM NH₄HCO₃ at 30°C for 16 h and then subjected to peptide analysis by LC-MS². Samples were analysed by LC-MS² using an LTQ mass spectrometer (Thermo Fisher Scientific). The results were analysed with the MASCOT database (www.matrixscience.com).

Immunopurification of perlecan and CSPG4

Perlecan and CSPG4 were isolated from the conditioned medium produced by cultured WiDr, MM200 and Me1007 cells by anion exchange chromatography followed by monoclonal anti-perlecan domain V antibody (clone A74) and monoclonal anti-CSPG4 antibody (clone B5) immunoaffinity chromatography, respectively, as previously described (69, 70). Additionally, some samples were then reloaded onto a 1 ml DEAE column (HiTrap™ DEAE Fast flow, GE Healthcare Life Science) attached to a fast protein liquid chromatography (FPLC) system (ÄKTA purifier, GE Healthcare Life Science) at 1 ml/min and eluted with a linear gradient of 0.1–1.0 M NaCl, 20 mM Tris, 10 mM EDTA, pH 7.5 over 36 column volumes at 0.2 ml/min and collected in 1 ml fractions.

Surface plasmon resonance

The interaction between CSPG4 and perlecan was analysed with a BiaCore 2000 (GE Healthcare) using research grade Au sensor chips. The conditions used were adapted as previously described (14, 71). Briefly, the sensor chip flow channels were first washed with DPBS at 5 μl/min until a stable baseline was achieved. The flow channels were then exposed to immunopurified materials at 20 μg/ml at a flow rate of 5 μl/min until 1500 resonance units (RU) were reached. The flow channels were then blocked with 1% BSA/PBS at 5 μl/min for 4 min then rinsed with DPBS to achieve a stable baseline before injecting either perlecan or CSPG4 at 20 μg/ml over the coated surface at 10 μl/min for 4 min. The stable baseline acted as the negative control and the baseline values were subtracted from the binding RU values. Channel surfaces were regenerated with 1 M NaCl between each sample. Sensorgrams were analysed using the BIAcore 2000 Evaluation software 3.0.2.

Cell adhesion assay

The cell adhesion assays were performed as described previously (55). Briefly, chamber slides were coated with 50 μl of 20 μg/ml HCAEC derived immunopurified perlecan at for 16 h at 4°C. The wells were blocked with 3% (w/v) BSA/PBS for 2 h. Cells were harvested from routine culture with enzyme-free cell dissociation buffer to maintain the cell-surface CSPG4 (Cell dissociation solution non-enzymatic, Sigma Australia). Cells were seeded at a density of 3,30,000 cells/ml in 100 μl serum free DMEM medium. Selected cells were treated with mouse monoclonal anti-CSPG4 antibodies, clones B5 or 9.2.27 (2 μg/ml), anti-perlecan antibody, clone A74 or anti-α2β1 integrin antibody, clone BHA2.1 for 1 h at 37°C prior to seeding. After 1 h of exposure of the cells to the perlecan coating, non-adherent cells were removed by washing twice with DPBS. For live cell imaging, fresh serum free DMEM medium was added to the wells. Images were taken with differential interference contrast (DIC) microscopy (Nikon Eclipse TiE). The number of adherent cells in each condition was counted manually each field of view. For actin staining, adherent cells were fixed as per the immunocytochemistry protocol. Actin filaments were stained with Rhodamine-phalloidin (1:100 dilution in 1% (w/v) BSA in DPBS) at 37°C for 30 min and washed four times with PBST before taking images using a confocal microscope (Olympus Fluoview FV1200). ImageJ was used to analyse the cell perimeter. A minimum of 10 cells were chosen at random and analysed for each condition.

Statistical analysis

A one-way analysis of variance (ANOVA) was performed to compare multiple conditions. Results of P < 0.05 were considered significant. Experiments were performed in triplicate and experiments were repeated three times.

Results

CSPG4 were produced by melanoma and colon carcinoma cell lines

The expression and localization of CSPG4 in melanoma (MM200 and Me1007) and colon carcinoma (WiDr) cell lines were investigated by immunocytochemistry. All three cell lines expressed CSPG4 on their cell surface, but there were differences in CSPG4 staining intensity and distribution, depending on the cell type. MM200 cells, exhibited a punctate granular staining pattern for CSPG4 (Fig. 1A (i)), whereas Me1007 cells demonstrated a more homogeneous pattern of staining, with some punctate staining on the cell membrane (Fig. 1A (ii)). The colon carcinoma cell line, WiDr, showed intense CSPG4 staining focused on the cell membrane, with minimal intracellular CSPG4 staining (Fig. 1A (iii)). In contrast, CSPG4 was found not to be expressed by primary human coronary artery endothelial cells (HCAEC), consistent with previous reports of the absence of CSPG4 on endothelial cells (data not shown).

Fig. 1.

Characterization of CSPG4 produced by melanoma and colon carcinoma cell lines. (A) Immunolocalization of CSPG4 produced by MM200 (i), Me1007 (ii) and WiDr (iii) cell lines. The presence of CSPG4 was detected with an anti-CSPG4 antibody (clone B5) (arrows). Cell nuclei were counterstained with DAPI. Scale bar represents 10 µm. Western blot analysis of CSPG4 (clone B5) in MM200 (i), Me1007 (ii) and WiDr (iii) cell lysates (B) and proteoglycan enriched cell culture medium (C). Samples were treated or untreated with either C’ase ABC or Hep III, or both. Molecular weight standards (in kDa) were electrophoresed on each gel and were indicated on the left.

The biochemical structure of the cell associated CSPG4 produced by the three cell lines was investigated by Western blotting of whole cell lysates (Fig. 1B). MM200 and Me1007 cells expressed multiple forms of CSPG4. MM200 cells expressed a proteoglycan form at a relative molecular mass (Mr) ∼4,60,000 that was found to contain a protein core of Mr ∼2,70,000 and decorated with CS as determined by a reduction of molecular weight after C’ase ABC treatment to remove CS chains (Fig. 1B (i) lanes 1 and 2). Smaller protein forms were found at approximately Mr ∼2,70,000 and ∼1,80,000 (Fig. 1B (i)). Digestion with Hep III had no effect on the size of the immunoreactive bands indicating that the CSPG4 expressed by these cells was not decorated with HS (Fig. 1B (i) lane 3). CSPG4 expressed by Me1007 cells was larger than that produced by MM200 cells, with the proteoglycan form at Mr >4,60,000 (Fig.1B (ii) lane 1). Digestion of Me1007-derived CSPG4 with C’ase ABC produced a protein core of Mr 2,70,000 (Fig. 1B (ii) lane 2). Me1007-dervied CSPG4 was also expressed as smaller protein forms at approximately Mr ∼2,00,000 and ∼1,80,000 (Fig. 1B (i)). Digestion with Hep III had no effect on the size of the immunoreactive bands indicating that the CSPG4 expressed by these cells was not decorated with HS (Fig. 1B (ii) lane 3). The CSPG4 expressed by WiDr cells was present as two forms at Mr ∼2,70,000 and 1,80,000, neither of which were affected by either C’ase ABC or Hep III digestion, indicating that these forms of CSPG4 were not proteoglycans (Fig. 1B (iii) lanes 1–3).

The biochemical structure of the shed form of CSPG4 produced by each of the cell lines was analysed by Western blotting after enrichment of the conditioned medium using anion exchange chromatography. MM200-derived shed CSPG4 was detected in four forms: a proteoglycan form detected as an immunoreactive smear centered at Mr 4,60,000 as well as the full length protein core at Mr 2,70,000 and two smaller protein forms at Mr ∼2,00,000 and ∼1,80,000 (Fig. 1C (i) lane 1). Digestion with C’ase ABC to remove CS indicated that the proteoglycan form was decorated with CS (Fig. 1C (i) lane 2), while HS was not present as confirmed with Hep III digestion (Fig. 1C (i) lanes 3 and 4). The smaller protein forms at Mr ∼2,00,000 and ∼1,80,000 became more intense after C’ase ABC digestion, indicating these forms were present both as protein core and decorated with CS (Fig. 1C (i) lanes 2 and 4). Me1007-derived shed CSPG4 was detected as a proteoglycan form similar in size to the proteoglycan form of MM200-derived CSPG4 (Fig. 1C (ii) lane 1). Digestion with C’ase ABC resulted in the generation of the full length protein core at Mr ∼2,70,000 as well as smaller forms at Mr ∼2,00,000 and ∼1,80,000 indicating these forms were decorated with CS (Fig. 1C (ii) lanes 2 and 4). There was no change in size of immunoreactivity of the Me1007-derived shed CSPG4 after Hep III digestion, indicating that the CSPG4 was not decorated with HS (Fig. 1C (ii) lane 3). Interestingly, the shed form of CSPG4 was not detected in medium conditioned by WiDr cells (data not shown). Together these results suggested that different cell types produce different forms of both cell-associated and shed CSPG4 that are differentially substituted with CS.

The smaller forms of CSPG4 produced by all cell lines may be the result of either proteolytic processing of the full length CSPG4 or alternative splicing to synthesize different forms of CSPG4. To investigate further, PCR was performed using primers to amplify the CSPG4 gene corresponding to each of the domains of the transcribed protein (Table I and Fig. 2A). These analyses confirmed that all the tumour cell lines expressed mRNA for all regions of the CSPG4 gene (Fig. 2B). Although the PCR product for domain IV from WiDr mRNA was very faint, the presence of mRNA encoding this transmembrane/cytoplasmic region was confirmed by qPCR (Fig. 2C). qPCR confirmed that all tumour cell lines tested expressed mRNA for full length CSPG4 (Fig. 2C). Thus, the presence of smaller forms of CSPG4 was most likely the result of proteolytic processing.

Fig. 2.

CSPG4 gene expression in WiDr, MM200 and Me1007 cells. (A) Schematic of CSPG4 denoting the four domains and the location of primers used to amplify regions of the CSPG4 gene corresponding to each of the domains of the transcribed protein. (B) mRNA derived from all three cell lines was isolated and used to generate cDNA, which was amplified using primers to domains I–IV of CSPG4 and electrophoresed on 1.5% (w/v) agarose gels. Products from the GAPDH primer set were electrophoresed on each gel to indicate the amount of cDNA that was loaded on each gel for each cell type. (C) qPCR analysis of CSPG4 using same primer sets as part (B). Data presented as mean threshold cycle (CT) ± standard deviation. The data are representative of three independent experiments.

Perlecan were produced by melanoma and colon carcinoma cell lines

Perlecan produced by MM200 and Me1007 melanoma cell lines was localized intracellularly (Fig. 3A (i) and (ii)). Perlecan produced by WiDr cells was detected on the exposed cell surface of cells at the edge of the dense clusters (Fig. 3A (iii)). There was no evidence of perlecan in the ECM elaborated by any of the tumour cell lines suggesting that they did not lay down an ECM in culture. However, all of the tumour cell lines secreted perlecan into the conditioned medium. Perlecan secreted by MM200 cells was a proteoglycan with a protein core of Mr 4,60,000 and decorated with HS (Fig. 3B (i), lanes 1–4). Perlecan secreted by Me1007 cells was a proteoglycan with protein cores of Mr 4,60,000, 2,00,000 and 1,20,000 (Fig. 3B (ii), lanes 1–4). The protein core of Mr 4,60,000 was decorated with HS (Fig. 3B (ii), lane 3). Perlecan secreted by WiDr cells was a proteoglycan with a protein core Mr 4,60,000 and decorated with CS/HS as well as a fragment at Mr 1,20,000 (Fig. 3B (iii), lanes 1–4). Perlecan produced by Me1007 and WiDr cells at Mr 1,20,000 and 2,00,000 may be proteolytic fragments (Fig. 3B (ii and iii), lanes 1 and 3). Interestingly, these bands disappeared after C’ase ABC treatment. The same bands have been reported for smooth muscle cell-derived perlecan but not endothelial-cell derived form using the same antibody (CCN-1) and they also disappeared after C’ase ABC treatment (33). Analysis of the bands at Mr 2,00,000 and 1,20,000 by mass spectrometry confirmed that they contained perlecan including the N-terminal domain I and the C-terminal domains of IV and V (Fig. 3C). Domains I and V of perlecan have been shown to be decorated with CS. Therefore, it is possible to speculate that the disappearance of these bands was due to their decoration with CS and C’ase ABC digestion resulted in protein core forms too small to be resolved in the 3–8% Tris-acetate gels.

Fig. 3.

Characterization of perlecan produced by melanoma and colon carcinoma cell lines. (A) Immunolocalization of perlecan produced by MM200 (i), Me1007 (ii) and WiDr (iii) cell lines. The presence of perlecan was detected with anti-perlecan antibody, CCN-1 (arrows). Cell nuclei were counterstained with DAPI. Scale bar represents 10 µm. (B) Western blot analysis of perlecan (CCN-1) in proteoglycan enriched MM200 and Me1007 conditioned media. Samples were untreated or treated with either C’ase ABC or Hep III, or both. Molecular weight standards (in kDa) were electrophoresed on each gel and were indicated on the left. (C) Perlecan peptides identified in immunoreactive bands at Mr 2,00,000 and 1,20,000 that were sensitive to C’ase ABC digestion in (B) analysed by mass spectrometry. (D) mRNA expression of HSPG2 in WiDr, MM200 and Me1007 cells. mRNA derived from all three cell lines was isolated and used to generate cDNA, which was amplified using primers to domains I, III and V of perlecan and electrophoresed on 1.5% (w/v) agarose gels. Products from the GAPDH primer set were electrophoresed on each gel to indicate the same amount of cDNA was loaded on each gel for each cell type. The data are representative of three independent experiments.

PCR analysis of the HSPG2 gene indicated the presence of sequences that encode domains I, III and V of the protein core of perlecan at the expected sizes. These data indicated that alternative splicing had not occurred in these cells (Fig. 3D).

CSPG4 and perlecan formed a stable complex via protein core interactions

The shed form of CSPG4 and secreted perlecan were immunopurified from medium conditioned by MM200, Me1007, WiDr and HEK293 cells transfected to produce recombinant NG2 (the rat homologue of CSPG4) via immunoaffinity chromatography. These preparations were analysed by mass spectrometry for their purity. Surprisingly, both CSPG4 and perlecan peptides were identified in all immunopurified samples (Table II). These data indicated that perlecan and CSPG4 co-immunopurified. The MOWSE score for perlecan in the CSPG4 samples immunopurified from the melanoma cell lines was higher than the score for CSPG4 suggesting that there were a larger number of peptides identified for perlecan resulting in a higher coverage of the perlecan protein core. In contrast, NG2 immunopurified from HEK293 cells had very little co-purifying perlecan and immunopurified perlecan from WiDr cells had very little co-purifying CSPG4 using the same methodology (Table II). Immunopurified perlecan from HCAEC cells had no detectable co-purifying CSPG4, since these cells do not express CSPG4 (Table II).

Table II.

The presence of CSPG4 and perlecan in immunopurified CSPG4 and perlecan samples isolated from medium conditioned by different cell lines detected by peptide LC-MS²

Cell line	Sample immunopurified for	MOWSE scorea
		Perlecan	CSPG4
MM200	CSPG4	1265	372
	Perlecan	7527	593
Me1007	CSPG4	2253	443
	Perlecan	2474	125
WiDr	Perlecan	7847	143
HEK-293 cells expressing NG2	NG2	616	3242
HCAEC	Perlecan	2591	—

^aMOWSE score as determined by Mascot query. This is the value (P) which is a measure of the probability that the match is a random event expressed as –10log(P). The higher the score, the more confidence that the match is not due to a random event.

The immunoaffinity chromatography methodology utilized here uses urea to elute the bound fraction from the column. This would indicate that the complex formed between shed CSPG4 and secreted perlecan was stable and involved hydrophobic interactions. To confirm that the complex was not bound via ionic interactions, immunopurified CSPG4 isolated from MM200 cells was subjected to anion exchange chromatography with a linear NaCl gradient. Eluted fractions were analysed for the presence of both CSPG4 and perlecan by Western blotting (Fig. 4A and B). Perlecan was eluted between 0.43 and 0.67 M NaCl and was detected as a heterogeneous smear above Mr 4,60,000 (Fig. 4A). CSPG4 eluted between 0.51 and 0.63 M NaCl and was detected as a heterogeneous smear above Mr 4,60,000 (Fig. 4B). The smaller CSPG4 fragments detected as shed forms from MM200 cells (Fig. 1C (i)) were absent. This is likely due to their low affinity to the CSPG4 immunoaffinity column (mAb B5), resulting in them not being present in this purification step. These data demonstrated that CSPG4 and perlecan were not bound via a relatively weak ionic interaction as all fractions that contained CSPG4 also contained perlecan. These fractions were further analysed by mass spectrometry to confirm the presence of both CSPG4 and perlecan (Fig. 4C).

Fig. 4.

CSPG4 and perlecan co-purify. (A) The presence of perlecan (clone CCN-1) (A) and CSPG4 (clone B5) (B) in immunopurified CSPG4 samples from MM200 cells separated by anion exchange chromatography with a linear elution gradient 0.1–1.0 M NaCl analysed by Western blotting. Fractions containing immunoreactive bands are shown. Molecular weight standards (in kDa) were electrophoresed on each gel and were indicated on the left. Peptide LC-MS2 was performed on the fractions in A and B (C). MOWSE scores were as determined by Mascot query for perlecan and CSPG4 and also represented as the ratio of the perlecan MOWSE score to the CSPG4 MOWSE score.

To investigate whether the interaction between CSPG4 and perlecan was an artefact of the purification process, binding studies were performed using surface plasmon resonance. In these studies, HCAEC-derived perlecan was used as it had no co-purifying CSPG4 and HEK293-derived extracellular domains of NG2 (rat CSPG4) was used as it had negligible amounts of co-purifying perlecan (Table II). Binding was assessed by exposing NG2 in solution to bound perlecan and vice versa. Both formats gave similar binding profiles demonstrating rapid association and dissociation with maximal binding of 500 RU (Fig. 5A).

Fig. 5.

CSPG4 and perlecan bind via protein–protein interactions. The binding between immunopurified NG2 and HCAEC perlecan was analysed by surface plasmon resonance. (A) NG2 was immobilized on gold sensor chips and the binding of HCAEC perlecan was monitored (i) and vice versa (ii). (B) C’ase ABC treated NG2 was immobilized on gold sensor chips and the binding of Hep III treated HCAEC perlecan was monitored (i) and vice versa (ii). (C) Perlecan domain V (i) and C’ase ABC treated perlecan domain V (ii) were immobilized on gold sensor chips, respectively, and the binding of C’ase ABC treated NG2 was monitored. (D) C’ase ABC treated perlecan domain V was immobilized on gold sensor chips and the binding of C’ase ABC treated NG2 in the presence of 0.1% Tween was monitored. All samples were prepared at 20 µg/ml and binding assays were run at 10 µl/min at 25°C. Data shown are representative of three independent experiments.

In previous studies, HCAEC derived perlecan has been shown to be decorated with HS exclusively (33). When NG2 and perlecan were treated with endoglycosidases C’ase ABC and Hep III, respectively, to remove their glycosaminoglycan chains, similar binding profiles were obtained, suggesting that the glycosaminoglycan chains were not involved in the binding (Fig. 5B) and that the mechanism of binding was via protein–protein interactions. To identify the region of perlecan involved, the binding between CSPG4 and perlecan, we employed two recombinant proteins, one NG2 and the other was a C-terminal peptide encompassing 37 amino acids of domain IV and all of domain V of perlecan (perlecan domain V, for the purpose of this paper) (Fig. 5C). The monoclonal antibody against perlecan domain V (clone A74) showed similar reactivity against the recombinantly expressed perlecan domain V compared to the native HCAEC derived perlecan (Supplementary Fig. S1). The monoclonal antibody against CSPG4 (clone B5) also showed a similar phenomenon for MM200 cell derived CSPG4 compared to recombinantly expressed NG2. This suggests that the recombinant proteoglycans were folded in a similar way to their respective regions in the full length molecules when bound on a surface. Binding was assessed by exposing the NG2 recombinant in solution to perlecan domain V adsorbed to the chip surface. NG2 and perlecan domain V interacted with rapid association and dissociation and maximal binding of 500 RU (Fig. 5C (i)). The perlecan domain V recombinant is extensively decorated with CS (34), so we analysed the binding with perlecan domain V treated with C’ase ABC to remove the CS. The binding between immobilized perlecan domain V protein core (without CS) and NG2 in solution exhibited rapid association and dissociation, with maximal binding of 1700 RU (Fig. 5C (ii)). This level of binding was greater than that observed for full length perlecan (Fig. 5A and B), supporting the idea that the extracellular regions of CSPG4 interacted with the C-terminal region of perlecan. In these experiments, the sensor chip was coated with same amount of either perlecan domain V or full length perlecan. Based on the relative size of perlecan domain V and full length perlecan there should be ∼6 times more perlecan domain V immobilized using the recombinant C-terminal region compared to full length perlecan. However, only ∼3 times as much NG2 bound to the perlecan domain V without CS compared to full length perlecan, indicating that NG2 may also interact with regions of perlecan outside of the C-terminal region. Removal of the CS on perlecan domain V enabled more binding to NG2 suggesting that when a glycosaminoglycan chain decorates domain V of perlecan it can modulate the interaction with CSPG4 by potentially masking the protein core binding sites. This phenomenon was not observed using full length HCAEC-derived perlecan as its C-terminal region is not substituted with any glycosaminoglycan chain.

In order to examine whether the interaction between NG2 and perlecan domain V was of a hydrophobic nature, binding was assessed in the presence of Tween-20. Perlecan domain V without CS was immobilized on the sensor surface and exposed to NG2 without CS in solution in the presence of 0.1% Tween-20 (Fig. 5D). Under these conditions rapid association and dissociation were observed with maximal binding of 500 RU. This level of binding was reduced compared to conditions without Tween-20 indicating that the interaction between CSPG4 and perlecan was of a hydrophobic nature.

CSPG4 is involved in cell adhesion to perlecan

As CSPG4 has been shown to be involved in cell adhesion and migration, it was hypothesized that these events involved the interaction between CSPG4 and perlecan. This hypothesis was tested in cell adhesion assays using the melanoma cell lines. Cells were plated on HCAEC-derived perlecan coated surfaces in the presence or absence of anti α2β1 integrin antibody (clone BHA2.1), anti-perlecan antibody (clone A74) or anti-CSPG4 antibodies (clone B5 and 9.2.27). Cells were also plated on fibronectin coated surfaces as a positive control (Fig. 6A (i)). Cells were allowed to adhere for 1 h prior to analysis (Fig. 6A and B). When MM200 cells were plated on perlecan coated surfaces they exhibited a spread morphology with polymerized actin microfilaments and prominent focal adhesions (Fig. 6A (ii) [indicated by arrow] and B (i)). The addition of the anti-α2β1 integrin antibody (clone BHA2.1), anti-perlecan antibody (clone A74) and anti-CSPG4 antibodies (clones B5 and 9.2.27) resulted in the adhered cells becoming rounded and without polymerized actin indicating reduced contact with the perlecan coated surface (Fig. 6A (iii and vi) and B (ii and iii)). In contrast, cells plated on fibronectin exhibited spread morphologies with well-developed actin fibers (Fig. 6A (i)), and in the presence of anti-CSPG4 antibodies, there was no changes in cell morphologies (data not shown). Measurements of cell perimeter for cells plated on perlecan indicated that cells plated on perlecan in the presence of anti α2β1-integrin antibodies and anti-CSPG4 antibodies were significantly (P < 0.05) less spread than cells plated on perlecan in the absence of antibodies (Fig. 6C). However, in the presence of the anti-perlecan antibody, there was no significant (P < 0.05) difference compared to in the absence of antibodies.

Fig. 6.

The CSPG4 and perlecan interaction is involved in cell adhesion and spreading. (A) MM200 cell adhesion on fibronectin coated surface was used as a positive control (i). MM200 cell adhesion on HCAEC perlecan coated surfaces, in the absence of antibody (ii), in the presence of antibodies against α2β1 integrin clone BHA2.1 (iii), against perlecan clone A74 (iv) and against CSPG4 clones B5 (v) and clone 9.2.27 (vi). Cells were stained for actin filaments with Rhodamine-phalloidin (i–vi) and imaged using confocal microscopy. White arrows point out the polymerized actin filaments. Scale bar represents 10 µm. (B) Live cell images using DIC were also taken after MM200 cell adhesion on HCAEC perlecan coated surfaces in the absence of antibody (i), in the presence of antibody against CSPG4 clones B5 (ii) and clone 9.2.27 (iii) for 1 h. Scale bar represents 50 µm. (C) Analysis of the perimeter of cells exposed to the perlecan coating for 1 h in the absence or presence of antibodies from (B). A minimum of 10 cells were analysed from three confocal images taken at random for each condition. Individual data points are indicated as well as mean ± standard deviation. (D) The number of cells adhered to the perlecan coating after 1 h in the absence or presence of anti-CSPG4 antibodies determined from the live cell DIC images in (C). Individual data points were indicated as well as mean ± standard deviation. ‘*’ denoted significant differences compared to cells plated on perlecan in the absence of antibodies (P < 0.05) analysed by one-way ANOVA. Data shown are representative of three independent experiments.

The number of adhered cells in presence of anti-CSPG4 antibodies indicated that there were significantly (P < 0.05) less cells adhered in the presence of 9.2.27 antibody than in the absence of antibodies (Fig. 6D). However, the anti-CSPG4 antibody clone B5 did not significantly reduce the number of cells adhered. These data indicated that both α2β1 integrins and CSPG4 are involved in MM200 cell adhesion and spreading on perlecan, an abundant component of most basement membrane/ECMs.

The adhesion assay was performed in serum-free medium to reduce the confounding effects of serum proteins on cell adhesion. The MM200 cells were able to tolerate these conditions however, while the Me1007 and WiDr cells were able to adhere to perlecan under these conditions, they exhibited a rounded morphology over the 1 h time period of the assay and up to 14 h after the commencement of the assay (data not shown).

Discussion

In this study, the melanoma and colon carcinoma cell lines, MM200, Me1007 and WiDr, expressed CSPG4 on the plasma membrane. These findings are supported by previous reports of CSPG4 being expressed on the plasma membrane of the human uveal melanoma cell line, basal breast carcinoma, squamous carcinoma of the head and neck (SCCHN), mesothelioma, pancreatic carcinoma, some types of renal cell carcinoma, chordoma, soft tissue sarcomas and chondrosarcoma cells (3, 5, 35, 36). CSPG4 is not only associated with tumour cells, as it has been detected on the plasma membrane of smooth muscle cells, pericytes (37), macrophages (38) and oligodendrocyte precursors (39) suggesting a role for CSPG4 in both biology and pathology. This study also found that the melanoma cells produced CS decorated forms of CSPG4 while the colon carcinoma cells produced only the core protein form suggesting that it can be a ‘part-time proteoglycan’ (40) and support different roles and/or interactions for CSPG4 in different cell types.

CSPG4 produced by the melanoma cell lines analysed in this study was shed into the conditioned medium as multiple forms decorated with CS. In contrast, the colon carcinoma cells shed CSPG4 into the medium at levels only detectable by mass spectrometry. The reasons for these cell type differences remain unclear but maybe due to the relative levels of shedding enzymes. Additionally, this study revealed differences in the CS decoration and protein core forms of shed CSPG4 between the two melanoma cell lines suggesting that the shedding process may involve multiple steps and different enzymes produced by the different cell types. Oligodendrocyte precursors generate and shed truncated forms of CSPG4/NG2 (8, 9), while primary human periodontal ligament fibroblast cells also shed CSPG4 (41). Interestingly, several proteinases including TIMPs-2 and -3 sensitive metalloproteinase (8, 9), ADAMTS-10 (42) and MMP-13 (41) have been reported to release CSPG4/NG2 from the plasma membrane and to generate truncated forms of the protein core. Furthermore, it has been suggested that the ectodomain of CSPG4/NG2 was shed by the α-secretase ADAM10 where the C-terminal fragment is subsequently processed by the γ-secretase to release the intracellular domain (42). The importance of this shedding process remains unclear, although it is speculated to be a physiological process as shed and fragmented forms of CSPG4 have been detected in sera from patients with malignant disease as well as patients without any sign of disease (5). It is hypothesized that this process facilitates cell migration by releasing cell membrane associated CSPG4 bound to ECM components.

In this study, perlecan produced by the tumour cell lines was mostly secreted, with some amounts localized intracellularly and on the cell surface. These data align with reports of the intracellular localization of perlecan in ameloblastomas cells (43), mast cells (44) and the cell surface expression on keratocystic odontogenic tumour cells (45). Interestingly, none of the cell lines analysed in this study deposited perlecan in their ECM/pericellular matrix that is characteristic of many cell types including primary endothelial, vascular smooth muscle (33) and chondroprogenitor cells (46). CSPG4 has been reported to interact with a number of ECM components, including laminin, tenascin, fibronectin, as well as collagen types II, V and VI (40, 47, 48). The extracellular central domain II of CSPG4 is reported to bind to collagen types V and VI (49), suggesting a role for cell surface CSPG4 in anchoring cells to ECM components. The present study demonstrated for the first time that CSPG4 interacts with the ubiquitous HSPG, perlecan, which is common to most connective tissues and basement membrane that plays an important role in cancer progression. In a study, perlecan expression and immunolocalization was examined in 27 metastatic melanoma tumours. Twenty-six of the 27 tumour samples showed a significant increase in the perlecan mRNA levels compared to normal tissues, which correlated to an increased deposition of perlecan in the pericellular matrix and vasculature basement membrane (50). This study highlights an additional basement membrane/ECM component, perlecan, is involved in cell adhesion via CSPG4 demonstrating the complexity of cell interactions with ECMs.

This novel binding partner for CSPG4 was identified following purification of CSPG4 by immunoaffinity chromatography that was found to contain both CSPG4 and perlecan. A similar methodology was used to discover the binding between CSPG4 and collagen type VI (51). The interaction between CSPG4 and perlecan was primarily due to hydrophobic protein-protein binding between the protein cores. CSPG4 binds to collagen via similar mechanisms (47). The interaction between CSPG4 and perlecan involved multiple binding sites on the perlecan core protein, including the C-terminal region, while the extracellular region of CSPG4 bound to perlecan as demonstrated by the cell adhesion assays in this study. When only perlecan domain V was exposed to CSPG4, the interaction was modulated by the presence of CS on this form of perlecan, indicating that the glycosaminoglycan chain may have masked the protein core binding sites. This effect was not observed for full length perlecan used in this study as it was derived from endothelial cells and only substituted with HS on its N-terminal domain I, suggesting that this region of perlecan is not involved in binding CSPG4. Interestingly, the CS attached to CSPG4 was not involved in the binding between CSPG4 and perlecan. These data suggest that although different cell types express CSPG4 either as a protein or a proteoglycan with different CS structures, the interaction between CSPG4 and perlecan would not be affected.

Perlecan interacts with a range of ECM molecules including collagen type IV, laminin, nidogen and fibronectin (52, 53). Perlecan also interacts with the cell surface α2β1 integrin via a site located in domain V (31). Interestingly, this integrin was first shown to be involved in cell adhesion to collagens (54). Additionally, the expression of β1 integrins obscures the binding between CSPG4 and collagen types V and VI, which is independent of this integrin (55) and directly mediates cell adhesion, polarization and migration (56, 57). In contrast, the binding between CSPG4 and fibronectin requires α4β1 integrins (10, 58). This study demonstrated that melanoma cells bind to perlecan and that this adhesion involved CSPG4 and α2β1 integrins. The addition of an anti-α2β1 integrin antibody reduced cell adhesion to perlecan indicating α2β1 integrins support MM200 cell adhesion on perlecan. A study using another human melanoma cell line SK-MEL-2, showed that SK-MEL-2 expressed α2β1 integrins were directly involved in invasion of reconstituted basement membrane via a type IV collagen-dependent mechanism (59). Together, the results demonstrated the important role of α2β1 integrins interacting with perlecan in assisting melanoma cell adhesion to the basement membrane/ECM.

The functional role of the interaction between CSPG4 and perlecan was explored in this study using cell adhesion assays; one of the major roles ascribed to CSPG4. The addition of anti-CSPG4 antibodies reduced cell adhesion to perlecan indicating that binding between CSPG4 and perlecan directly supported cell adhesion. Furthermore, the addition of anti-CSPG4 antibodies that bind to the ectodomain of CSPG4 inhibited the formation of polymerized actin fibers. These data are supported by previous findings that the cytoplasmic domain of CSPG4 is involved in anchorage to actin stress fibers (60). Engagement of CSPG4 supports cytoskeletal rearrangement necessary for changes in cell morphology to support cell motility. Tumour cell invasion of basement membranes represents one of the critical steps in the metastatic process. While the assays performed in this study focused on early cell adhesion events, it is possible to speculate that the interaction between perlecan and CSPG4 might support cell migration and metastasis (57).

This study demonstrated for the first time that CSPG4 interacts with perlecan in a protein core dependent fashion to promote cell adhesion and promote actin polymerization. This interaction may be a fundamental process whereby cells attach to perlecan-rich matrices including most connective tissues and basement membranes, to supporting cell adhesion prior to cell migration. Indeed metastatic melanoma cells increase perlecan expression while antisense targeting of perlecan cDNA reduced cell proliferation and invasion (61), supporting the notion that perlecan is involved in the migration of CSPG4-expressing cells. While perlecan promotes actin polymerization supporting cell adhesion, the role of shed CSPG4 bound to perlecan remains to be determined and is speculated to provide a mechanism to enable cell motility.

The binding between CSPG4 and perlecan is likely to have broad relevance to other CSPG4-expressing cells exposed to perlecan-rich matrices such as in the developing vasculature and brain. Indeed, pericytes express high levels of CSPG4 and interact with endothelial cells that produce a perlecan-rich vascular basement membrane. Pericyte-specific NG2 ablation in blood vessels of intracranial B16F10 melanomas decreased pericyte ensheathment of endothelial cells and reduced assembly of the vascular basal lamina (7). Furthermore, NG2 null mice displayed a deficiency in both pericyte and endothelial cell maturation, as well as reduced basal lamina assembly (62). Thus, it is tempting to speculate that the interaction between CSPG4 and perlecan supports pericyte and invasive ‘tip’ endothelial cell interactions and guide the formation of blood vessels in development and angiogenesis (63–66).

Supplementary Data

Supplementary Data are available at JB Online.

Glossary

Abbreviations

C’ase ABC

chondroitinase ABC

CS

chondroitin sulphate

CSPG4

chondroitin sulphate proteoglycan 4

ECM

extracellular matrix

FGF

fibroblast growth factor

GAG

glycosaminoglycan

HCAEC

human coronary artery endothelial cell

Hep

heparinase

HS

heparan sulphate

HSPG

heparan sulphate proteoglycan

SMC

human coronary artery smooth muscle cell

TCPS

tissue culture polystyrene

VEGF

vascular endothelial growth factor