THE NG2 CHONDOITIN SULFATE PROTEOGLYCAN: ROLE IN

MALIGNANT PROGRESSION OF HUMAN BRAIN TUMOURS

M. CHEKENYA,$, % H.K. ROOPRAI,$ D. DAVIES,} J.M. LEVINE,}

A.M. BUTT% and G.J. PILKINGTON$*$Experimental Neuro-oncology Group, Department of Neuropathology, Institute of Psychiatry, King's College, De

Crespigny Park, London SE5 8AF, UK; %KGT School of Biomedical Sciences, Division of Physiology, King's College(St. Thomas's Campus), London SEI 7EH. UK; }FACS Laboratory, ICRF, 61 Lincoln's Inn Fields, London WC2A

3PX. UK; }Department of Neurobiology and Behaviour, SUNY at Stony Brook, Stony Brook, New York, NY 11794,USA

AbstractÐThe expression and function of NG2, a transmembrane chondroitin sulfate proteoglycan wasstudied in human gliomas of various histological types in culture using immunocytochemistry and ¯owcytometry. NG2 was di�erentially expressed in the neoplasms, with higher expression in high comparedto low-grade gliomas. In acutely isolated cells from human biopsies, NG2 +ve and NG2 ÿve popu-lations were morphologically distinct from each other, and NG2 +ve cells were more proliferative thanNG2 ÿve cells. The mitogens platelet derived growth factor (PDGF-AA) and basic ®broblast growthfactor (bFGF) added in combination to serum-free medium (SFM) upregulated NG2 expression onglioblastoma multiforme cells in culture but had little e�ect on NG2 expression on the anaplastic astro-cytoma cells. Furthermore, NG2 was colocalised with the platelet derived growth factor alpha receptor(PDGFaR) and antibody blockade of the PDGF-aR ablated NG2 expression on the glioblastoma mul-tiforme cells, suggesting that increased NG2 expression in the presence of PDGF-AA is mediated viathe PDGF-aR. Assays of migration and invasion indicate that NG2 +ve glioma cells migrated moree�ciently on collagen IV and that NG2 ÿve cells were more invasive than their NG2 +ve counterparts.The results indicate that NG2 may be, respectively, positively and negatively related to the proliferativeand invasive capacity of glioma cells. Thus, expression of the NG2 proteoglycan may have major impli-cations for malignant progression in glial neoplasms and may prove a useful target for future thera-peutic regimens. # 1999 Published by Elsevier Science Ltd on behalf of ISDN. All rights reserved.

Key words: extracellular matrix, growth factors, invasion, migration, NG2, PDGFa receptor, progeni-tor, proliferation.

INTRODUCTION

Intrinsic tumours of the central nervous system (CNS) are characterised by di�use localinvasion of the normal brain parenchyma. This invasion is a multifaceted phenomenoninvolving cell motility, adhesion, enzymatic degradation of the extracellular matrix (ECM)components, and paracrine/autocrine interactions between normal and neoplastic cells.39 Localbrain invasion is mediated by quiescent G0/G1 phase ``guerilla'' cells which are more resistantto radiotherapy and chemotherapy than the proliferative S,G2/M phase cell populations. Whilethe control of tumour growth and invasion are central to providing e�ective therapies forprimary brain tumours, knowledge of the precise time of carcinogenic stimulus and the targetcell populations a�ected would provide the basis for improved diagnostic practice and possiblepreventative medical approaches.

The antibody NG2 binds to a membrane spanning chondroitin sulfate proteoglycan, with acore protein of 260 kDa47 originally identi®ed on rat neural cell lines with characteristics ofboth neurones and glial cells.52 In the developing and adult rat CNS, NG2 identi®esproliferating O-2A progenitor cells22,48, which develop into oligodendrocytes or Type 2

Int. J. Devl Neuroscience, Vol. 17, No. 5±6, pp. 421±435, 1999# 1999 Published by Elsevier Science Ltd on behalf of ISDN.

All rights reserved. Printed in Great Britain0736-5748/99 $20.00+0.00

PII: S0736-5748(99)00019-2

www.elsevier.com/locate/ijdevneu

*To whom correspondence should be addressed. Tel.: +44-171-919-3268; fax: +44-171-708-3895; E-mail: g.pilkington@iop.bpmf.ac.uk

421

astrocytes depending on medium conditions in vitro.42 In the adult CNS (white matter andcerebellum), NG2 labels a population of stellate glial cells, but not mature astrocytes oroligodendrocytes6,23, suggesting that NG2 expression is turned o� before glial terminaldi�erentiation. The existence of NG2 positive cells in the adult rat CNS43 has led to thesuggestion that they may be adult progenitor cells and the in vivo counterpart of the in vitro O-2Aadult progenitor.36 NG2 +ve cells have also been described in adult human brain where theyhave been proposed to be a pool of progenitors which can regenerate oligodendrocytes inmultiple sclerosis.45 NG2 is also expressed by a wide variety of tissues outside the CNS, mainlythose of the mesenchymal origin, such as developing cartilage, skeletal and smooth muscles andblood vessels.32 Signi®cantly, it has been proposed25 that NG2 may play a role in cellularmigration, adhesion and di�erentiation. We suggest that it may serve a similar function in ofsome neoplastic glial cells for the following reasons: (1) NG2 interacts extracellularly via its coreprotein with collagen types II, V, VI49 and with laminin and tenascin5, the latter two also beingproposed as important mediators of invasive behaviour of neoplastic glia.40 (2) NG2 acts as acoreceptor for spreading and focal contact formation, in association with b1 integrin, which isimportant in tumour cell adhesion and migration.16 b1 is the most abundant integrin subunit inhuman gliomas37 and has been shown to positively in¯uence invasiveness.38 (3) The intracellularbinding partner for NG2 is actin, which is involved in cell motility and development ofmorphology.26 (4) NG2 has been shown to potentiate the activity of PDGF-AA and its areceptor signalling pathway in smooth muscle and O-2A cells12,34 and PDGF-AA is a knownmitogen for glioma cells in vitro.17

It is noteworthy that NG2 has been demonstrated on human melanoma cells16 and thehuman homologue of NG2 is expressed by acute myeloid leukemia (MLL) blasts but notnormal haematopoietic cells.2,46 NG2 has been useful for correlating infant MLL patients withtranslocational mutations associated with poor prognosis. NG2 is also a tumour-speci®cantigenic determinant on the chemically induced rat chondrosarcoma HSN, and there isevidence that it is important in the growth and metastasis of the HSN tumour.21

Previously, we have identi®ed cells with an O-2A progenitor phenotype in human19 gliomas,using the ganglioside recognising monoclonal antibody A2B5.9 Gangliosides not only modulateintegrin function31, but also upregulate ECM components such as laminin20 and matrixmetalloproteinases28 which are involved in brain tumour invasion. Furthermore, A2B5expression is indicative of G0/G1 cells which have transiently arrested from the cell cycle andare in the migratory or invasive phase.41 It is possible therefore to envisage a scenario whereintricate interactions between gangliosides, integrins and NG2 modulate the malignantbehaviour of neoplastic glia. In this paper we report that human gliomas express NG2 andprovide evidence that NG2 is important in proliferation rather than migration or invasiveness ofhuman glioma cells in vitro.

EXPERIMENTAL PROCEDURES

Glioma primary cultures and cell lines were propagated in Dulbecco's modi®ed Eagle'smedium (DMEM; without pyruvate, with glucose; Sigma), 10% heat inactivated foetal calfserum (FCS) and 1% antibiotic/antimycotic (at a ®nal concentration of 100 IU Streptomycin,100 mg Penicillin and 0.25 mg/ml Amphotericin: Sigma) until con¯uent at 378C in an humidi®edtissue culture incubator (LEEC), in 5% C02, 95% air atmosphere (see Table 1).

Alternatively, the cells were synchronised sequentially to serum-free medium (DMEM+F12(Ham) with L-Glu+Hepes and 3'3'5'Triiodo-L-thyronine (15 nM), Insulin (5 mg/ml) Transferrin(5 mg/ml), Sodium selenite (5 mg/ml) media supplement, L Glutamine (0.2 M), 1% penicillin/streptomycin and progesterone (20 mg/ml). Medium was changed to 7% FCS, incubated at 378Cfor 24 h, then 3% FCS for 24 h, then serum-free medium for 24 h. Growth factors (humanrecombinant-PDGF AA and human recombinant-bFGF expressed in E. coli; Sigma) at 10 ng/ml were then added to serum-free medium and cells cultured for 24±36 h prior to ICC and ¯owcytometry.

M. Chekenya et al.422

Biopsies

Human tumour specimens from excised gliomas or from stereotactic biopsies were transferredinto a sterile petri dish and rinsed with HBSS (Sigma) to remove blood clots. With a sterilescalpel, blood vessels and necrotic tissue were dissected out and the tissue disaggregated intosmall fragments until a uniform suspension was obtained. The cell suspension was furtherdisaggregated by gently pipetting with a sterile Pasteur pipette. The suspension was thendispersed into 25 cm2 tissue culture ¯asks containing 5 ml DMEM and left undisturbed for3±4 days until settled.

Immunocytochemistry

Living glioma cells were used to detect cell surface NG2, A2B5 and PDGFa receptor byindirect immuno¯uorescence. Cells grown on microscope coverslips were incubated in NG2(1:1000), or PDGFa receptor (1:200) at room temperature for 1 h, followed by swine anti-rabbitbiotinylated antibody (1:100) for 30 min and streptavidin ¯uorescein (1:100), (Amersham Int.)for 15 min in the dark. For detection of the intracellular antigen, GFAP, cells were ®xed in acidalcohol (99 pts ethyl alcohol/l pt HCl) for 2 min then washed in HBSS. All specimen coverslipswere mounted in citi¯uor (Citi¯uor Products) and examined on a Leica epi¯uorescencemicroscope equipped with ¯uorescein and rhodamine optics.

Flow cytometry

Cells were immunolabelled as described above. Analysis was performed on a Facs Calibur(BD, San Jose, CA) equipped with a 15-mW, 488-nm argon laser. Emission ¯uorescence wasmeasured with a DF 530/30 ®lter for FITC and a DF 572/42 ®lter for phycoerythrin. Dataacquisition and analysis were performed with Celiquest software.

DNA staining with propidium iodide

After harvesting and washing, cells were ®xed with icecold 70% ethanol while vortexing. Thesamples were then spun at 2000 rpm for 5 min followed by washing �2 in PBS. Cells were thentreated with 100 ml of ribonuclease 100 mg/ml for 5 min at room temperature. 400 ml ofpropidium iodide 50 mg/ml was then added and samples were analysed on the FACS Calibur.Cell debris and clumps were gated out and PI ¯uorescence at 560±600 nm was collected.

Magnetic cell sorting

Cells were grown in tissue culture ¯asks and harvested by trypsinisation, counted and labelledwith primary mouse antirat NG2 1:10 for 5±10 min at room temperature. Cells were washedwith 500 ml bu�er (PBS pH 7.2, supplemented with 0.5% BSA and 2 mM EDTA), centrifugedat 1000 rpm for 5 min and supernatant removed completely. The pellet was then resuspended in80 ml of bu�er containing 20 ml of Macs Goat Antimouse IgG microbeads (Miltenyi Biotech,

Table 1. Derivation of cell cultures used, passage number, histological type and grade of malignancy. The pre®x ``IP''denotes Institute of Psychiatry, where the cell lines were propagated from biopsies. U denotes cell lines from Uppsala

and NP from the Queen's University of Belfast (Courtesy of Dr. D. McCormick)

Cell culture designation Passage No. Type+grade18

NP785/96 10 glioblastoma multiforme (WHO grade IV)

IPSS 10 glioblastoma multiforme (WHO grade IV)

NP131 12 glioblastoma multiforme (WHO grade IV)

U251-MG >100 glioblastoma multiforme (WHO grade IV)

IPSB-18 13 astrocytoma (WHO grade III)

IPJJ 10 astrocytoma (WHO grade III)

IPCE Primary oligodendroglioma (WHO grade II)

IPDDC-A2 15 astrocytoma (WHO grade II)

IPRW/98 Primary atypical meningioma (WHO grade I)

IPPE/98 Primary pilocytic astrocytoma (WHO grade I)

IPGC/98 Primary meningioma (WHO grade I)

Role of NG2 in human gliomas 423

London), mixed well and incubated for 15 min at 6±128C. Cells were washed carefully andresuspended in 500 ml of bu�er before magnetic separation. Cells were placed onto MS+/RS+

column in the magnetic ®eld of a MiniMacs separator and rinsed with 500 ml of bu�er. Cellsuspension in 1000 ml bu�er was applied and the negative cells allowed to pass through. Cellswere then rinsed with 1500 ml of bu�er. The column was removed from the separator andplaced on a suitable collection tube, 1000 ml of bu�er was pipetted onto the column and positivefraction ®rmly ¯ushed out using the plunger supplied with the column.

Migration assay

Glioma cells were harvested and resuspended at 40� 106 cells/ml in DMEM containing 10%FCS, 1% antibiotics and 1% low melting point agarose (Sigma Chemical Co.), maintained at378C to prevent setting of the agarose. 1.5 ml drops of the cell suspension were applied to thecentre of wells within a 24-well tissue culture dish (Nunc: Gibco Life technologies, Paisley),which was then placed at 48C for 15 min to allow the agarose to solidify. Prior to addition ofdrop, each well was coated with Poly-D-ornithine (10 mg/ml) to mark the centre and facilitateadherence of the drop. Around the agarose drop, 50 ml of ECM components (Sigma) wereadded (murine laminin 5 mg/ml, collagen IV 10 mg/ml, bovine vitronectin 0.5 mg/ml and bovine®bronectin 5 mg/ml) and incubated at 378C for 2 h, after which 400 ml of SFM+10 ng/ml instandard tissue culture PDGFAA chemoattractant was added to cover the cell suspension. Theculture was incubated and cell migration counted from the leading edge of drop on fourquadrants (N±S, E±W) on an Olympus phase contrast optical microscope with a calibratedgraticule in the eyepiece at 24 h, 48 h etc. The width of one grid square represented 0.12 mmactual distance at a magni®cation of 10�. Cells migrated out to form a uniform corona aroundthe drop. Within single experiments each condition was tested in triplicate. The mean migrationwas calculated for each condition This assay is a modi®ed30 Varani agarose drop assay.51

Invasion assay

The Transwell modi®ed Boyden chamber assay was used to monitor the role of NG2 inglioma cell invasion stimulated by PDGF. Polycarbonate micropore ®lters, pore size 8 mm(Marathon Lab. supplies, London) in 12 well transwell units, were coated with 100 ml of ECMcomponents (described above) overnight and then washed with sterile PBS and air dried. On theday of the assay, cells were harvested by trypsinisation. Cells were resuspended in DMEMserum-free medium at 104 cells seeding density and incubated for 8±12 h (depending on thepopulation doubling times), with 10 ng/ml PDGF AA in the bottom chamber as achemoattractant. Invasion assays were done in triplicate for each condition. Cells were ®xedwith methanol and ®lters stained with eosin and methylene blue (Di� Kwik Kit: Merck Ltd,Leicestershire). Cells on the upper surface of the ®lter were counted (®ve ®elds/condition) usingthe phase contrast microscope described above, with a 20� objective and a 10� eyepiece ®ttedwith a 10-mm grid scale. Cells on the top of the ®lter were then wiped o� using a cotton swab,and the procedure repeated for the cells invading through the ®lter. Results were expressed as:

% invasion � No: cells invading� 100

No: noninvading� invading cells

Materials

A polyclonal antiserum against rat NG2 described previously47, was used forimmunocytochernical staining and inhibition experiments and was a kind gift from J.M. Levine.The mouse antirat NG2 is a cocktail of six mouse monoclonals.49 This was used for magneticcell sorting. Fluorescein conjugated swine antibody to rabbit immunoglobulins (DAKO) waspurchased from Amersham International. Rabbit anti-Human PDGFRa neutralising andantibodies (polyclonal) were obtained from R&D systems, London. A2B5 mouse monoclonalantibody is a hybridoma of the original mouse A2B5 IgM9, and was a kind gift from Dr. HelenPilcher (Institute of Psychiatry). Rabbit anticow GFAP was obtained form DAKO.

M. Chekenya et al.424

RESULTS

NG2 is expressed on cultured human glioma cells

The presence of NG2 on human gliomas was con®rmed by immuno¯uorescence staining ofbiopsy derived human glioma cells in culture with antibodies against rat NG2. Glioma cellswere maintained either in SFM or in medium containing 10% FCS to study the changes inNG2 expression. There was di�erential expression of NG2 on the gliomas studied, where highexpression was evident with high grade glioblastoma multiforme (Fig. 1) and reduced expressionwith decreasing malignancy. Approximately 70% of anaplastic astrocytoma cells were NG2positive in FCS supplemented medium; there was a mutual exclusivity between NG2 and GFAPin these conditions with the remaining 30% being GFAP positive (Fig. 3). The glioblastoma-derived cultures were virtually 100% positive for NG2 but were uniformly GFAP negative. Onanaplastic astrocytoma cell lines, small stellate process bearing cells had a punctate NG2distribution on the perikaryon and processes (Fig. 2). This was further quanti®ed by ¯owcytometry which revealed that anaplastic astrocytoma cells had greater NG2 ¯uorescenceintensity in medium containing 10% FCS than in SFM. Grade II oligodendroglioma cellsexpressed lower levels of NG2 whilst grade I pilocytic astrocytoma and meningioma cells didnot express NG2. Under 10% FCS-supplemented conditions 10% of cells in anaplasticastrocytoma cultures were A2B5 positive (Fig. 4).

NG2 and PDGFa receptor are colocalised on human gliomas

When anaplastic astrocytoma cells were labelled after 4 days in medium containing 10% FCSwith a polyclonal antihuman PDGFa receptor antibody, there was colocalisation of the twoantigens. However, whilst NG2 was detected on both cell bodies and processes, the PDGFareceptor was mostly localised to the perikaryon with faint staining on a few processes (Fig. 5).

Fig. 1. NG2 positive human glioblastoma multiforme cell. Glioblastoma multiforme cells form humanbiopsies were labelled with rabbit anti-rat NG2 and swine anti-rabbit secondary antibody. Cells werethen stained with ¯uorescein isothiocyanate and propidium iodide nuclear counterstain (see

Experimental procedures).

Role of NG2 in human gliomas 425

Fig. 2. (a) NG2 positive anaplastic astrocytoma cells. Anaplastic astrocytoma cells from human biopsieswere labelled with rabbit anti-rat NG2 and swine anti-rabbit secondary antibody. Cells were thenstained with ¯uorescein isothiocyanate and propidium iodide nuclear counterstain (see Experimentalprocedures). (b) GFAP and A2B5 positive anaplastic astrocytoma cells. Anaplastic astrocytoma cells

from human biopsies were double labelled with rabbit anti cow GFAP (TRITC) and A2B5 (FITC).

M. Chekenya et al.426

E�ect of PDGF and bFGF on NG2 expression

Cultured anaplastic astrocytoma and glioblastoma multiforme cells were treated for 24 h with10 ng/ml of bFGF, PDGF or in combination in SFM to study the e�ect of the growth factorson intensity of NG2 ¯uorescence, and was quanti®ed by ¯ow cytometry. In anaplasticastrocytoma cells, NG2 ¯uorescence intensity was greater in medium containing 10% FCS thanin SFM, ANOVA (F15.9, df=3, p<0.001 at 95% con®dence interval) with Tukey-HSD multiplecomparison of means at (p<0.05). Addition of bFGF increased NG2 ¯uorescence intensityabove that in SFM alone [Fig. 4(a)], but bFGF in combination with PDGF seemed to havelittle e�ect on NG2 expression.

In contrast, glioblastoma multiforme cells showed greater NG2 immunoreactivity in bothSFM and 10% FCS. Treatment of the cells with bFGF or PDGF in SFM resulted in reducedNG2 ¯uorescence intensity, whereas bFGF and PDGF in combination increased NG2immunoreactivity above that in SFM alone [Fig. 4(b)], ANOVA (F92, df=4, p<0.001 at 95%con®dence interval) with Tukey-HSD multiple comparison of means at (p<0.05).

Cell cycle analysis

Propidium iodide was used to assess cell cycle status of NG2 positive and NG2 negativeastrocytoma cells. Fig. 5 represents the PI DNA histograms showing % cells in proliferativephases (S, G2/M). NG2 positive cells were slightly more proliferative than NG2 negative cells.ANOVA (F15.1, df=4, p=0.004 at 95% con®dence interval) with Tukey-HSD multiplecomparison of means.

E�ects of PDGFa receptor blockade on NG2 expression

The possibility that interference with the normal function of the PDGFa receptor might a�ectNG2 expression was explored by incubating the cells with anti-PDGFa receptor neutralisingantibody in 10% FCS medium, SFM or SFM with either bFGF, PDGF or in combination for24 h prior to ¯ow cytometric analysis. (1) NG2 expression was greater at all times whenPDGFa-R was not blocked. (2) PDGFa-R blockade decreased NG2 expression irrespective ofmedium. (3) PDGF-AA reduced NG2 expression even when PDGFa-R was blocked suggestingthat blockade was incomplete and implying competitive inhibition. (3) bFGF or PDGF reduced

Fig. 3. Anaplastic astrocytoma cells express PDGFa receptor. Anaplastic astrocytoma cells werelabelled with rabbit anti-human PDGFaR and swine anti-rabbit secondary antibody. Cells were then

stained with Texas red and Hoescht blue nuclear counterstain.

Role of NG2 in human gliomas 427

NG2 expression (4) bFGF and PDGF increased NG2 expression and this was only whenPDGFa-R was not blocked [Fig. 4(b)].

Migration

To determine if NG2 positive anaplastic astrocytoma cells will migrate in response to PDGFAA, migration assays were performed in transwell plates coated with various ECM componentsusing a modi®ed migration assay25. 12 h after plating most cells had not extended processes onlaminin, vitronectin or ®bronectin, but on collagen IV at this time point, some cells werealready beginning to migrate beyond the edge of the agarose drop in all quadrants in response

Fig. 4. (a) E�ect of bFGF and PDGF on NG2 expression on anaplastic astrocytoma cells. For each ex-periment (N=3) a control (FITC only) was used to set a background ¯uorescence level. The mean ¯uor-escence intensity (MFI) of the control was used to normalise the other samples, so each sample isexpressed as a % of the control. NG2 MFI was greater in 10% FCS than SFM. ANOVA (F15.9, df=3,p<0.001). (b) E�ect of bFGF and PDGF on NG2 expression on glioblastoma multiforme cells. Foreach experiment (N=3) a control (FITC only) was used to set a background ¯uorescence level. Themean ¯uorescence intensity (MF) of the control was used to normalise the other samples, so each sampleis expressed as a % of the control. bFGF and PDGF in combination increased NG2 MFI above that inSFM. (open bars) ANOVA (F92, df=4, p<0.001). PDGFaR blockade ablated NG2 MFI (®lled bars).

M. Chekenya et al.428

to PDGF AA. Only after 24 h on vitronectin and ®bronectin was there initial movement beyondthe edge, with movement on laminin apparent only at 48 h. There was no di�erence inmigration rate over time of the cells on ®bronectin and vitronectin. Cells migrating on collagenIV moved the greatest distance both at 24 and 48 h (Table 2).

Invasion

In order to establish whether there are di�erences in invasive behaviour between NG2positive cells and NG2 negative cells, the Transwell modi®ed Boyden chamber assay was used.

Fig. 5. NG2 positive cells are more proliferative than NG2 ÿve cells. Immunomagnetically isolatedNG2 +ve and NG2 ÿve anaplastic astrocytoma cells were stained with propidium iodide (50 mg/ml)for DNA content. The upper histogram is a pro®le from NG2 ÿve cells, the lower from NG2 +ve cells(N=3). The region M1 represents cells in the proliferative phases of the cell cycle (S,G2 and M).

ANOVA (F15.1, df=4; p<0.004)

Role of NG2 in human gliomas 429

NG2 ÿve anaplastic astrocytoma cells were more invasive than NG2 +ve cells. ANOVA (F16.03,df=3, p<0.0001 at 95% con®dence levels) with Tukey-HSD multiple comparison of means at(p<0.05). There was no signi®cant di�erence in invasive behaviour of the cells propagatedeither in 10% FCS medium (NG2 +ve), SFM supplemented with bFGF (NG2 +ve). The NG2antibody added to medium containing 10% FCS did not improve invasion (Fig. 6).

Furthermore, NG2 ÿve cells were more invasive through laminin, ANOVA (F3.72, df=4,p=0.019 at 95% con®dence interval) with Tukey-HSD multiple comparison of means at(p<0.05). There was no signi®cant di�erence in invasion on collagen IV, vitronectin, hyaluronicacid or uncoated polycarbonate ®lters (Fig.7).

DISCUSSION

The major aims of our studies were to determine whether human gliomas express NG2 invitro and investigate its role in malignant progression of the neoplasms. NG2 +ve cells wereidenti®ed in human gliomas of various histological types and grades of malignancy in vitro. Ourdata indicated that NG2 expression was related to the proliferative phase and that theproteoglycan could be ``turned on'' by changing culture conditions and driving cells into the

Table 2. Migration of NG2 +ve anaplastic astrocytoma cells on ECM substrates

ECM 24 h 48 h

mean migration (mm) (N=3) S.E.M mean migration (mm) (N=3) S.E.M

Collagen IV 0.24 0.03 0.72 0.06

Fibronectin 0.11 0.03 0.24 0.06

Laminin 0.06 0.034 0.11 0.031

Vitronectin 0.14 0.052 0.24 0.069

Fig. 6. Invasion of NG2 +ve cells through laminin coated polycarbonate ®lters. NG2 ÿve cells propa-gated in SFM were most invasive, ANOVA (F16.03, df=3, p=0.0001) There was no signi®cant di�er-ence in the invasiveness of cells propagated in SFM+bFGF or 10% FCS. These cells were shown by¯ow cytometry and immunocytochemistry to be NG2 +ve. Antibody blockade of NG2 expression in

10% FCS failed to improve their invasive potential.

M. Chekenya et al.430

proliferative phase. NG2 +ve cells were less invasive than their NG2 ÿve counterparts. NG2expression was higher in glioblastoma multiforme cells compared to anaplastic astrocytoma cellsand declined with reduced grades of malignancy. Similar trends of NG2 expression anddi�erentiation have been reported in human melanomas51 where lower levels of theproteoglycan are observed in several well-di�erentiated melanoma sublines.11,15 Furthermore,NG2 downregulation has been achieved in other melanoma sublines when di�erentiation hasbeen induced with exogenous compounds such as cyclic adenosine monophosphate or phorbolesters. We suggest from our in vitro data that, since there is heightened expression on highgrade, less di�erentiated gliomas (indicated by decreased GFAP immunoreactivity), NG2expression may be a marker of malignancy in these neoplasms. In a very recent report4,transfection of B16 melanoma cells with the NG2 proteoglycan reverted their phenotype to thatof a less di�erentiated one with a concurrent increase in proliferative capacity. Indeed, it hasbeen established14 that children with NG2 positive acute myeloid leukemia cells have asigni®cantly worse outcome than NG2 negative cases. In consonance with this idea, it has beenproposed that the concept of dedi�erentiation in oncology27 is associated with an increase in theproliferative capacity of the cells both in vitro and in vivo.1,51 Sometimes this is attributable tochanges in the repertoire of cell surface antigens expressed by transformed cells at varying stagesof di�erentiation.15

Recent ®ndings show that not only is NG2 expressed in human brain sections45, but in ourlaboratory we have found NG2 +ve cells within human glionia sections.7 Although there issome similarity in morphology (bipolarity), this subpopulation of NG2 +ve cells in glialtumours is not coincident with CD68 +ve microglia. Moreover, NG2 +ve cells were generallyseen to be Ki67 +ve on double immunolabelling. As proliferating cell populations are notgenerally invasive and NG2 may play a negative regulatory function in integrin mediatedmigration, the possible role of NG2 in facilitation of proliferation and downregulation ofinvasion is therefore proposed in human gliomas.

Since NG2 is a marker for O-2A progenitors in vitro and precursor cells in vivo24, it istempting to speculate that human gliomas may be derived from the so-called O-2A progenitorlineage. In this case, the transforming oncogenic insults may have occurred either in the

Fig. 7. Invasion of NG2 ÿve cells through ECM coated polycarbonate ®lters. NG2 ÿve cells were moreinvasive through laminin, ANOVA (F3.72, df=4, p=0.0 19). There was no signi®cant di�erence of inva-

siveness through collagen IV, vitronectin or hyaluronic coated polycarbonate ®lters.

Role of NG2 in human gliomas 431

developing brain or in adult life on the O-2Aadult progenitor cells. However, as these cells areneoplastic in nature they may not correspond completely to the phenotypic categories of normalglia seen at di�erent stages of development or di�erent lineage pathways. In addition, sinceprofound changes in the cell surface may accompany malignant transformation, NG2expression inter alia, may di�er in its distribution from those on normal cells.

NG2 is coexpressed with PDGFa-R and is associated with proliferation

Flow cytometric analysis of the DNA-binding dye propidium iodide showed that NG2 +vehuman glionia cells have a higher percentage of cells in the proliferative phases of the cell cyclethan NG2 ÿve glioma cells. This ®nding is in agreement with previous observations4 that lessdi�erentiated NG2 positive melanoma cells exhibited faster rates of proliferation in vitro thantheir NG2-negative counterparts. Previous studies12,34 have suggested that NG2 is involved inactivation of the PDGF-AA/ PDGF-a receptor pathway. Anti-NG2 antibodies block PDGF-AA induced proliferation and migration in smooth muscle cells12, supporting the speci®cinvolvement of NG2 in facilitating the activity of this growth factor pathway. This is onepossible mechanism by which NG2 expression could in¯uence growth of glial neoplasms. ThePDGF-a receptor was colocalised with NG2 indicating that growth factor e�ects may bemediated by the PDGFa-R. This is supported by the demonstration34 of similar subcellulardistribution patterns of NG2 and the PDGFa-R on partially di�erentiated rat O-2A cells. Thisdistribution is also consistent with that observed in the mature rat brain in vivo.33 Our datashows NG2 to be more widely expressed than the PDGFa-R and blockade of the receptor withthe Human PDGFa-neutralising antibody ablated NG2 expression and reduced proliferation.Antibodies to PDGFa receptor block the mitogenic e�ects of PDGF, causing cells to ceasedivision and di�erentiate prematurely. A role for autocrine PDGF stimulation in the growth ofhuman brain tumours is suggested by the coexpression of PDGF and PDGF receptors.35

Furthermore, the ®nding that the PDGFa receptor gene on chromosome 4q11±p12 is ampli®edin some glioblastomas10 and is highly expressed in all astrocytomas lends support for its role intumour growth.

E�ects of bFGF and PDGF on the expression of NG2

The e�ect of the mitogenic growth factors bFGF and PDGF indicated that NG2 expressionis under the in¯uence of these growth factors. Anaplastic astrocytoma cells did not express NG2in SFM but bFGF added to SFM induced NG2 expression. This is consistent with the ®ndingsthat treatment of freshly plated rat O-2A progenitor cells with bFGF in SFM slightly increasedboth NG2 immunoreactivity and the number of A2B5 positive cells33. However, it is not clearwhy in our study PDGF and bFGF in combination did not increase NG2 expression in theanaplastic astrocytoma cells. The e�ects of the growth factors were quite di�erent inglioblastoma multiforme cells. These cells expressed high levels of NG2 regardless of mediumand addition of individual growth factors reduced NG2 ¯uorescence, whereas together theyrestored NG2 expression to original levels. These results suggest a di�erential interaction ofbFGF and PDGF on NG2 expression in astrocytoma and glioblastoma multiforme cells,respectively. It may be indicative of growth factor synergism3 described previously.

NG2 modulates cell-ECM interaction

The NG2 +ve anaplastic astrocytoma cells migrated best on collagen IV at 48 h, comparedto ®bronectin, vitronectin and laminin. Intriguingly, previous studies have shown that the NG2core protein associates with type VI collagen, although the signi®cance of this association forglial cells is uncertain since type VI collagen has not been detected in the brain parenchyma.Type IV collagen is however, found exclusively in the basement membranes.44 Several studieshave suggested that NG2 may serve an anti-adhesive role on the cell surface, where attachmentof mouse melanoma cells to speci®c ligands for CD44 and a4b1 integrin is reduced by theexpression of NG2.4 NG2 has also been found to inhibit attachment and neurite outgrowthfrom cerebellar neurones on laminin substrates.8 Such weakening in cell±cell and cell±ECMinteractions have also been shown to be important in the proliferative properties of neoplastic

M. Chekenya et al.432

cells. Our data would indicate that NG2 interference with integrin mediated attachment toECM substrates would not necessarily arrest migration. The cells probably utilise additional celladhesion molecules to facilitate this migration. The possibility that NG2 expression maymediate glioma cell invasion was not supported by the data obtained in this study, wherebyNG2 ÿve anaplastic astrocytoma cells were more invasive compared to NG2 +ve cells andinvasiveness was not a�ected by growth factors. The use of rabbit antirat polyclonal NG2antibody to block NG2 expression failed to improve invasiveness. We are now exploring thepossibility that NG2 antisense treatment may establish a negative regulatory role of NG2 ininvasive behaviour. In the absence of any ECM coating the NG2 ÿve GBM cells failed to crossthe polycarbonate membrane ®lters in transwell assays. This is not unusual since provision ofECM substrates or treatment with TGFb, which stimulates endogenous production of ECM40,is often necessary for invasion to occur in this experimental system. Laminin and hyaluronicacid were the most permissive substrates for invasion of NG2 ÿve cells. This is particularlyintriguing since NG2 positivity has been demonstrated to interfere with bI integrin subunit andCD444 mediated binding. It is likely that other factors may be mediating invasion in these cells.Certain gangliosides have been shown to interact directly with growth factor receptors tocontrol cell proliferation13,29, di�erentiation and adhesion. Furthermore, ganglioside expressingglial progenitors in the developing mammalian CNS have been shown to be highly migratory. Ithas also been demonstrated that ganglioside expression is positively correlated with motile andinvasive behaviour of growth retarded, noncycling neoplastic glia.41

Thus, we conclude that NG2 expression is associated with proliferative activity in gliomacells, whereas invasiveness is associated with the expression of gangliosides. In this scenario,gangliosides and NG2 would be, respectively, up- and downregulated in the invasive andproliferative phases. The sequence of events may thus involve activation of gangliosideexpression on the cell surface leading to modulation of integrin receptor function31 andincreased binding to ECM components to facilitate invasion. Gangliosides such as GD3 havebeen proposed to modulate the invasive potential of neoplastic cells by upregulating matrixmetalloproteinase28 production and by acting as cellmatrix adhesion molecules within thein®ltrative edge of the tumour as the neoplastic cells migrate and invade the normal brain. NG2expression would inhibit integrin mediated binding to ECM and would accordingly retardinvasiveness but promote cell proliferation within the main tumour mass. Expression of NG2would be expected to be downregulated prior to invasion, implying a ®ne balance between NG2and ganglioside. expression in mediating glioma cell invasion. This would imply that a ``switch''mechanism may exist in gliomas inducing NG2 expression during proliferation and gangliosideexpression during invasion. These possibilities are currently being investigated in histologicalsections and in vitro using three-dimensional confrontation cultures.

AcknowledgementsÐThis work was made possible by a generous grant from the Samantha Dickson Research Trust.

REFERENCES

1. Aubert, C., Rouge, F. and Voulot, C., New variants of the B16 melanoma: tumourigenicity and metastatic propertiesunder di�erent culture conditions. J. Natl. Cancer. Inst., 1990, 82, 952±958.

2. Behm, F., Smith, F., Raimondi, S., Pui, C. and Bernstein, I., Human homologue of the rat chondroitin sulfate pro-teoglycan NG2, detected by monoclonal antibody 7.1, identi®es childhood acute lymphoblastic leukemias witht(4;11) (q2l; q23) or t(11; 19)(q23; p13) and MLL gene rearrangements. Blood, 1996, 87(3), 1134±1139.

3. Bogler, O., Wren, D., Barnett, S., Land, H. and Noble, M., Cooperation between two growth factors promotesextended self renewal and inhibits di�erentiation of oligodendrocytetype2 astrocyte (O-2A) progenitor cells. Proc.Natl. Acad. Sci. USA, 1990, 87, 6368±6372.

4. Burg, M. A., Grako, K. A. and Stallcup, W. B., Expression of the NG2 proteoglycan enhances the growth andmetastatic properties of melanoma cells. Journal of Cellular Physiology, 1998, 177, 299±312.

5. Burg, M., Nishiyama, A. and Stallcup, W., A central segment of the NG2 proteoglycan is critical for the ability ofglioma cells to bind and migrate toward type VI collagen. Experimental Cell Research, 1997, 235, 254±264.

6. Butt, A. M., Duncan, A., Homby, F., Kirvell, S., Hunter, A., Berry, M. and Levine, J., Cells expressing the NG2antigen contact nodes of Ranvier in adult CNS white matter. Glia, 1999, 0, 1±8.

7. Chekenya, M., Pilcher, H., Rooprai, H. K., Davies, D., Butt, A. M., Levine, J. M., Pilkington, G. J. Glia, 19998. Dou Chang-Lin, Levine, J. M., Inhibition of neurite outgrowth by the NG2 chondroitin sulfate proteoglycan.

Journal of Neuroscience, 1994, 14(12), 7616±7628

Role of NG2 in human gliomas 433

9. Eisenbarth, G. S., Walsh, F. S. and Nirenberg, M., Monoclonal antibody to a plasma membrane antigen of neurons.Proc. Natl. Acad. Sci. USA, 1979, 76, 4913±4917.

10. Fleming, T., Matsui, T. and Molloy, M., Demonstration of an activated platelet derived growth factor autocrinepathway and its role in human malignant glioma cell lines. Cancer Research, 1992, 48, 3910±3918.

11. Giu�re, L., Schreyer, M., Mach, J. P. and Cavrel, S., Cyclic AMP induces di�erentiation in vitro of human mela-noma cells. Cancer, 1987, 61, 1132±1141.

12. Grako, K. and Stallcup, W., Participation of the NG2 proteoglycan in rat aortic smooth muscle cell responses toPDGF. Experimental Cell Research, 1995, 221(1), 231±240.

13. Hakomori, S. I., Bifunctional role of glycosphingolipids. Journal of Biological Chemistry, 1990, 265, 3±18716.14. Hilden, J., Smith, F., Frestedt, J., McGlennen, R., Howeels, W. and Arthur, D., MLL gene rearrangement, cytoge-

netic 11q23 abnormalities and expression of the NG2 molecule in infant acute myeloid leukemia. Blood, 1997, 89,3801±3805.

15. Houghton, A. N., Eisinger, M., Albino, A. P., Caimcross, J. G. and Old, L. J., Surface antigens of melanocytes andmelanoma: markers of melanocyte di�erentiation and melanoma subsets. Journal of Experimental Medicine, 1982,156, 1755±1766.

16. Iida, J., Meijine, A., Spiro, R., Ross, E. and McCarthy, J., Spreading and focal contact formation of human mela-noma cells in response to the stimulation of both melanoma-associated proteoglycan (NG2) and a4bl integrin.Cancer Research, 1995, 55(10), 2177±2185.

17. Kirsch, M., Wilson, L. C. and Black, P., Platelet derived growth factor in human brain tumours, . Journal of Neuro-oncology, 1997, 35, 289±301.

18. Kleihues, P., Burger, P. C. and Scheithauer, B. W., The new WHO classi®cation of brain tumours. Brain Pathology,1993, 3, 255±268.

19. Knott, J., Edwards, A., Gullan, R. and Pilkington, G., A human glioma cell line retains expression of GFAP andgangliosides recognised by A2B5 and LB1 antibodies, after prolonged passage. Neuropathology and AppliedNeurobiology, 1990, 16, 489±500.

20. Koochekpour, S., Merzak, A. and Pilkington, W., Extracellular matrix proteins inhibit proliferation, upregulate mi-gration and induce morphological changes in human glioma cell lines. European Journal of Cancer, 1995, 31A, 375±380.

21. Leger, O., Johnson-Leger, C., Jackson, E., Coles, B. and Dean, C., Chondroltin sulfate proteoglycan NG2 is atumour-speci®c antigen on the chemically induced rat chondrosarcoma HSN. Int. J. Cancer, 1994, 58(5), 700±705.

22. Levine, J. and Stallcup, W., Plasticity of developing cerebellar cells in vitro studied with antibodies against the NG2antigen. Journal of Neuroscience, 1987, 7, 2721±2731.

23. Levine, J. and Card, P., Light and electron microscopic localisation of a cell surface antigen (NG2) in the rat cerebel-lum: association with smooth protoplasmic astrocytes. Journal of Neuroscience, 1987, 7, 2711±2720.

24. Levine, J. M., Enquist, L. and Card, P. J., Reactions of oligodendrocyte progenitor cells to alpha herpes virus infec-tion in the CNS. Glia, 1998, 23, 316±328.

25. Levine, J. M. and Nishiyama, A., The NG2 chondroitin sulfate proteoglycan: a multifunctional proteoglycan associ-ated with immature cells. Perspectives on Developmental Neurobiology, 1996, 3, 245±259.

26. Lin, X., Dahlin, H. and Stallcup, W., Interaction of the NG2 proteoglycan with the actin cytoskeleton. J. Cell.Biochem., 1996, 63(4), 463±477.

27. Linskey, M. E., Glial ontogeny and glial neoplasia: the search for closure. Journal of Neurooncology, 1997, 34, 522.28. Maidment, S., Merzak, A., Koochekpour, S., Rooprai, H., Rucklidge, G. and Pilkington, G., The e�ect of exogen-

ous gangliosides on matrix metalloproteinase secretion by human glioma cells in vitro. European Journal of Cancer,1996, 32A(5), 868±871.

29. Merzak, A., Koochekpour, S., McCrea, S. and Pilkington, G., Gangliosides modulate proliferation, migration andinvasiness of human brain tumour cells in vitro. Molecular and Chemical Neuropathology, 1995, 24, 121±135.

30. Milner, R., Anderson, H., Franklin, R. and �renchConstant, C., Contrasting e�ects of mitogenic growth factors onoligodendrocyte precursor cell migration. Glia, 1997, 19, 85±90.

31. Mugnai, G., Barletta, E. and Mannini, A., Modulation of the integrin mediated cell adhesion by complex ganglio-sides. Trends in Glycoscience and Glycotechnology, 1994, 6, 199±214.

32. Nishiyama, A., Dahlin, K., Prince, J. and Stallcup, W., The primary structure of NG2, a novel membrane-spanningproteoglycan. The Journal of Cell Biology, 1991, 14, 359±371.

33. Nishiyama, A., Lin, X. H., Giese, N., Heldin, C. H. and Stallcup, W. B., Colocalisation of NG2 proteoglycan andPDGF alpha receptor on O-2A progenitor cells in the developing rat brain. Journal of Neuroscience Research, 1996,43, 299±314.

34. Nishiyama, A., Lin, X. H., Giese, N., Heldin, C. and Stallcup, W., Interaction between NG2 proteoglycan andPDGF alpha-receptor on O-2A progenitor cells is required for optimal response to PDGF. J. Neurosci. Res., 1996,43(3), 315±330.

35. Nister, M., Libermann, T., Patterson, C., Heldin, C. and Westermark, B., Expression of mRNA for PDGF andTGF-a and their receptors in human malignant glioma cell lines. Cancer Research, 1988, 48, 3910±3918.

36. Noble, M., Wolswijk, G. and Wren, D., The complex relationship between cell division and the control of di�eren-tiation of the oligodendrocyte-type-2-astrocyte progenitor cells isolated from perinatal and adult rat optic nerves.Prog. Growth Factor Res., 1989, 1, 179±194.

37. Panou, C., Rooprai, H., Trillo-Pazos, G., Davies, D. and Pilkington, G., The beta 1 integrin subunit: expression androle in human gliomas. Neuropathology and Applied Neurobiology, 1997, 23, 173.

38. Paulus, D. and Tonn, C., Basement membrane invasion of glioma cells mediated by integrin receptors. J. Neurosurg.,1994, 80(3), 515.

39. Pilkington, G., Tumour cell migration in the CNS. Brain Pathology, 1994, 4, 157±166.40. Pilkington, G. L., The role of the extracellular matrix in neoplastic glial invasion of the nervous system. Brazilian

Journal of Medical and Biological Research, 1996, 26, 1159±1172.41. Pilkington, G. J., Glioma heterogeneity in vitro: the signi®cance of growth factors and gangliosides. Neuropathology

and Applied Neurobiology, 1992, 18, 434±442.

M. Chekenya et al.434

42. Ra�, M., Miller, R. and Noble, M., A glial progenotor cell that develops in vitro into an astrocyte or an oligoden-drocyte depending on culture medium. Nature, 1983, 303, 390±396.

43. Reynolds, R. and Hardy, R., Oligodendroglial progenitors labelled with the o4 antibody persist in the adult rat cer-ebral cortex in vivo. J. Neurosci. Res., 1997, 47, 455±470.

44. Rutka, J. T., Apodaca, G., Stem, R. and Rosenblum, M., The extracellular matrix of the central and peripheral ner-vous systems: structure and function. J. Neurosurg., 1988, 69, 155±170.

45. Scolding, N., Franklin, R., Stevens, S., Heldin, C. H., Compston, A. and Newcombe, J., Oligodendrocyte progeni-tors are present in the normal adult human CNS and in the lesions of multiple sclerosis. Brain, 1998, 121, 2221±2228.

46. Smith, F., Rauch, C., Williams, D., March, C., Arthur, D., Hilden, J. and Bernstein, I., The human homologue ofrat NG2, a chondroitin sulfate proteoglycan, is not expressed on the cell surface of normal hematopoetic cells but isexpressed by acute myeloid leukemia blasts from poorprognosis patients with abnormalities of chromosome band11q13. Blood, 1996, 87, 1123±1133.

47. Stallcup, W. B., Beasley, I. and Levine, J. M., Cell surface molecules that characterise di�erent stages in the develop-ment of cerebellar intemeurons. Cold Spring Harbor Symp. Quant. Biol., 1983, 48, 761±774.

48. Stallcup, W. and Beasley, L., Bipotential glial precursor cells of the optic nerve express the NG2 proteoglycan.Journal of Neuroscience, 1987, 7, 2737±2744.

49. Stallcup, W., Dahlin, K. and Healy, P., Interaction of the NG2 chondroitin sulfate proeoglycan with type VI col-lagen. Journal of Cell Biology, 1990, 111, 3177±3188.

50. Tsukamoto, K., Gersten, D. M., Law and Hearing, V. J., Malignant melanoma: relationship to parameters of di�er-entiation. Melanoma Research, 1991, 1, 223±230.

51. Varani, J., Orr, W. and Ward, P., A comparison of the migration patterns of normal and malignant cells in twoassay systems. Am. J. Pathol., 1978, 90, 159±172.

52. Wilson, A., Baetege, E. and Stallcup, W., Antisera speci®c for cell lines with mixed neuronal and glial properties.Developmental Biology, 1981, 83, 146±153.

Role of NG2 in human gliomas 435