LAB INVESTIGATION-HUMAN/ANIMAL TISSUE

Type I collagen is overexpressed in medulloblastomaas a component of tumor microenvironment

Yu Liang Æ Maximilian Diehn Æ Andrew W. Bollen ÆMark A. Israel Æ Nalin Gupta

Received: 27 May 2007 / Accepted: 25 June 2007 / Published online: 25 July 2007

� Springer Science+Business Media B.V. 2007

Abstract Medulloblastoma is the most common malig-

nant brain tumor of children, and more specific and

effective therapeutic management needs to be developed to

improve upon existing survival rates and to avoid side-

effects from current treatment. Gain of chromosome seven

is the most frequent chromosome copy number aberration

in medulloblastoma, suggesting that overexpression of

genes on chromosome seven might be important for the

pathogenesis of medulloblastoma. We used microarrays to

identify chromosome seven genes overexpressed in

medulloblastoma specimens, and validated using data from

published gene expression datasets. The gene encoding the

alpha 2 subunit of type I collagen, COL1A2, was overex-

pressed in all three datasets. Immunohistochemistry of

tumor tissues revealed type I collagen in the leptomenin-

ges, and in the extracellular matrix surrounding blood

vessels and medulloblastoma cells. Expression of both type

I collagen and the b1 subunit of integrin, a subunit of a

known type I collagen receptor, localized to the same area

of medulloblastoma. Adherence of D283 medulloblastoma

cells to type I collagen matrix in vitro depends on the b1

subunit of integrin. Because medulloblastoma is charac-

teristic of high vascularity, and because inhibition of type I

collagen synthesis has been shown to suppress angiogen-

esis and tumor growth, our data suggest that type I collagen

might be a potential therapeutic target for treating medul-

loblastoma.

Keywords Medulloblastoma � Microarray � Extracellular

matrix � Type I collagen � Adhesion

Introduction

Brain tumors are the primary cause of non-traumatic death

in children and young adults under the age of 20 years.

Primitive neuroectodermal tumors (PNETs) are the most

common malignant brain tumor in children and account for

25% of all pediatric brain tumors [1]. PNETs can arise in a

supratentorial (cerebral) or infratentorial (cerebellar) loca-

tion but most PNETs occurring in children arise within the

cerebellum. These cerebellar tumors are also called

medulloblastoma. The standard treatment for medulloblas-

toma is surgical resection followed by fractionated external

beam radiation and/or chemotherapy [2]. Although medul-

loblastoma is sensitive to both therapeutic modalities,

Electronic supplementary material The online version of thisarticle (doi:10.1007/s11060-007-9457-5) contains supplementarymaterial, which is available to authorized users.

Y. Liang � N. Gupta

Department of Neurological Surgery, Brain Tumor Research

Center, University of California, San Francisco, CA 94143, USA

Y. Liang (&)

Division of Molecular Cell Biology, Applied Biosystems, 850

Lincoln Centre Drive, Foster City, CA 94404, USA

e-mail: liangy01@hotmail.com

M. Diehn

Department of Radiation Oncology, Stanford University School

of Medicine, Stanford, CA 94305, USA

M. Diehn

Department of Biochemistry, Stanford University School

of Medicine, Stanford, CA 94305, USA

A. W. Bollen

Department of Pathology, University of California,

San Francisco, CA 94143, USA

M. A. Israel

Departments of Pediatrics and Genetics, Norris Cotton Cancer

Center, Dartmouth Medical School, Lebanon, NH 03756, USA

123

J Neurooncol (2008) 86:133–141

DOI 10.1007/s11060-007-9457-5

irradiation of the central nervous system (CNS) in young

children often results in serious long-term side effects, such

as deafness, cognitive decline, and neuroendocrine insuffi-

ciency. For this reason, a variety of chemotherapy regimens

are used as initial treatment in the very young patients to

delay the use of radiation therapy. Early diagnosis and

improved treatment have increased the 5 year survival rate

of patients with localized medulloblastoma to greater than

60% [3, 4].

Invasion of the leptomeninges by medulloblastoma cells

and their dissemination via the subarachnoid spaces

throughout the neuraxis to distant sites can occur either at

presentation or at relapse and is associated with poor sur-

vival [5]. Genetic studies and the development of trans-

genic mouse models have provided important information

about the cellular origin and oncogenic pathways under-

lying medulloblastoma [6]. The molecular mechanisms

responsible for invasion into and dissemination within the

neuraxis remain largely unknown. A recent study using

DNA microarrays suggested that expression of platelet-

derived growth factor a and the RAS/mitogen-activated

protein kinase signal transduction pathway might be

up-regulated in disseminated medulloblastoma and could

be targets for more effective treatments [7].

Both medulloblastoma and supratentorial PNET are

malignant and invasive embryonal tumors that have similar

histological features, but the survival for children with

supratentorial PNET is much poorer, only 20–30% at

5 years [8–10]. The degree of surgical resection may play a

role in the improved survival of medulloblastoma, since

gross total resection of supratentorial PNET is usually

more difficult to achieve. PNET arise from both locations

exhibit high microvascular density and express wide range

of angiogenic factors [11, 12]. Despite these similarities,

recent evidence has shown that medulloblastoma and

supratentorial PNET are two genetically distinct tumor

types, primarily based upon gene expression profiling [13]

and the patterns of chromosome copy number aberrations

from comparative genomic hybridization [14–17]. One

important observation is that majority of these studies also

demonstrated that gain of chromosome seven is the most

frequent copy number alteration shared by both medullo-

blastoma and supratentorial PNET [14–18]. For example,

gain of chromosome seven was detected in 44–57% of

medulloblastoma and in 66% of supratentorial PNET [17,

18]. For this reason, we sought to identify genes located on

the chromosome seven which are consistently overex-

pressed in both tumor types might provide the genetic basis

for shared features that characterize these tumors, such as

tumoral vascularity.

In the present study, we used cDNA microarrays to

examine the expression profiles of chromosome seven

genes in three medulloblastoma and three supratentorial

PNET specimens, and in normal brain tissue. We validated

the results with published datasets and confirmed overex-

pression of COL1A2 in both these tumor types. This gene

and COL1A1, each encodes a distinctive subunit that

together heterodimerizes forming type I collagen. We also

detected increased accumulation of type I collagen in

medulloblastoma by immunohistochemistry.

Materials and methods

Cell culture

D283 medulloblastoma cell line was obtained from the

Brain Tumor Research Center Tissue Bank at UCSF. All

cells were maintained in Eagle’s minimal essential medium

with 15% fetal bovine serum and 5% CO2.

Tissue specimens

Medulloblastoma and normal brain specimens were

obtained from the Brain Tumor Research Center Tissue

Bank at UCSF after approval by the Committee on Human

Research.

Microarray and bioinformatic analyses

The detailed microarray methods were published at the

website http://microarray-pubs.stanford.edu/gbm/. Briefly,

total RNA was extracted using Trizol followed by mRNA

purification using FastTrack (Invitrogen). mRNA was

reverse transcribed to cDNA and directly labeled with Cy

dyes (Amersham Biosciences) before hybridization. The

raw data of 22,636 features on the array were extracted

from the Stanford Microarray Database using the ScanA-

lyze-featured extraction software with the following

settings: regression correlation >0.6, channel 1 mean

intensity/median background intensity ‡1.5, and channel

two normalized mean intensity/median background inten-

sity ‡1.5. The DNA sequences of individual clones from

the extracted raw data were examined for chromosomal

locations with the ‘‘Table Browser’’ function at the UCSC

Genome Bioinformatics website (the Human May 2004

assembly at http://genome.ucsc.edu), and those localized

on chromosome seven (1,116 features) were identified.

Among these 1,116 features, only those with available data

in three or more tumor specimens and in one or more

normal brain specimens were selected for final analyses

(757 features).

Gene expression data of three independent published

datasets (four groups of tumor samples in total) for veri-

fication of expression of the COL1A1 and COL1A2 genes

were extracted from the Broad Institute Cancer Program

134 J Neurooncol (2008) 86:133–141

123

Datasets (http://www.broad.mit.edu/cancer/datasets.html)

and Serial Analysis of Gene Expression (SAGE) Anatomic

Viewer (http://cgap.nci.nih.gov/SAGE/AnatomicViewer).

Antibodies

Anti-b1 subunit of integrin (ITGB1), anti-type I collagen,

and anti-integrin linked kinase polyclonal antibodies were

from Santa Cruz Biotechnology (Santa Cruz, CA). Anti-

ITGB1 monoclonal antibodies AIIB2 and P4G11 were

from Developmental Studies Hybridoma Bank (Iowa

City, IA). Anti-CD31 and anti-fibronectin antibodies were

from Novocastro Laboratories (United Kingdom) and

GIBCO-BRL (Gaithersburg MD), respectively. Peroxi-

dase-conjugated and biotinylated secondary antibodies

were from Vector Laboratories (Burlingame, CA). Fluo-

rescine-conjugated and Rhodamine-conjugated secondary

antibodies and normal serum were from Jackson Immu-

noResearch Laboratories (West Grove, PA).

Immunohistochemistry

Frozen tissue sections used for immunohistochemistry

were fixed in 4% formaldehyde, treated with H2O2,

blocked with normal serum, incubated with primary anti-

bodies at 4�C overnight or room temperature (RT) 2 h,

incubated with biotinylated secondary antibody and per-

oxidase-labeled streptavidin at RT for 30 min, and visu-

alized using the DAB Reagent kit (KPL; Gaithersburg,

MA). Some frozen sections were visualized using FITC- or

Texas Red-conjugated secondary antibodies (Vector Lab-

oratories). Staining of paraffin-embedded sections followed

the same protocol, except for prior de-waxing and antigen

retrieval by microwave heating.

Adhesion assay

Wells of 96-well tissue culture plates (Corning, Corning,

NY) were coated with 100 lg/ml of rat-tail type I collagen

(BD Biosciences, Bedford, MA) for 1 hour at room tem-

perature, followed by two PBS washes. 1 · 105 D283 cells

resuspended in serum-free medium were plated into either

uncoated or type I collagen-coated wells for 1 hour. Non-

adhered cells were removed by lightly tapping the plates

and the number of cells remaining adherent was counted.

Monoclonal antibodies AIIB2 and P4G11 were diluted to

1:20 into the serum-free medium during a 1-hour incuba-

tion to inhibit cell adhesion.

Data analysis

All statistical analyses used the Student’s t test (2-tailed

and equal variance not assumed) and Pearson correlation in

SPSS for Windows (Release 11.5.0). A P value £ 0.05 was

considered statistically significant.

Immunoprecipitation and immunoblotting

D283 cells were lysed in RIPA buffer (in 1% NP-40, 0.1%

SDS, 0.5% Na deoxycholate, 50 mM Tris pH 8.0, and

150 mM NaCl) supplemented with 1 mM NaF, 1 mM

Na3VO4, and Complete protease inhibitor cocktail tablets

(Roche, Basel, Switzerland). After high-speed centrifuga-

tion, the lysate was pre-cleared twice with protein A

Sepharose beads (Zymed Laboratories, South San Fran-

cisco) followed by mixing with 1:100 dilution of anti-

bodies at 4�C for 1 h. Immunoprecipitated proteins were

captured by protein G Sepharose beads (Zymed Labora-

tories) and released by sample buffer. Samples were sep-

arated by SDS-PAGE and transferred to nitrocellulose

membranes, followed by 10% skim milk blocking and

antibody incubation, and visualized using the Super Signal

West Pico Chemiluminescent Substrate (Pierce, Rockford,

IL).

Results

Gene expression profiling identifies genes on

chromosome seven overexpressed in primitive

neuroectodermal tumors compared to normal brain

We performed gene expression profiling of three medul-

loblastoma and three supratentorial PNETs using cDNA

microarrays that consisted of ~23,000 features (represent-

ing ~18,000 unique UniGene clusters), and compared the

results to the expression patterns we observed for two

specimens derived from normal cerebrum and one from

normal cerebellum. After data filtering, we selected

approximately 750 genes located on chromosome seven

(see Methods and Additional file 1 for data), and computed

the difference of their expression between all six tumors

and the three normal brain specimens. Six UniGene clones

representing five different genes, ETV1, CDK6, COL1A2,

TAC1, and CAV2, had average expression in tumors at least

4-fold higher when compared to normal tissues (Fig. 1).

Differences in gene expression between tumor and normal

brain specimens reached significance in three genes using

the Student’s t test (ETV1, P = 0.006; CDK6, P = 0.0005;

COL1A2, P = 0.009).

COL1A2 mRNA is overexpressed in independent

groups of primitive neuroectodermal tumors

We verified the increased expression of ETV1, CDK6, and

COL1A2 using other gene expression datasets. In a recent

J Neurooncol (2008) 86:133–141 135

123

study, oligonucleotide microarrays were used to charac-

terize gene expression from embryonal tumors of the CNS,

and found a correlation between gene expression and the

clinical outcome of patients with medulloblastoma [13].

Their study included two groups of medulloblastoma and

supratentorial PNET specimens: the first group had 10

medulloblastoma and eight supratentorial PNET (Fig. 2a),

and the second group consisted of 60 medulloblastoma and

six supratentorial PNET (Fig. 2b). We reviewed the

expression of AEBP1, CDK6, and COL1A2 in these two

groups, and we found that COL1A2 but not the other two

genes demonstrated higher expression in tumors than in

normal cerebellum with a P value of less than 0.05 (see

Additional file 2 for data). We also identified increased

expression of COL1A2 in an additional published dataset

[19] that included 10 medulloblastomas (Fig. 2c).

Finally, we sought to validate this finding using gene

expression data from the SAGE Anatomic Viewer [20] that

displays gene expression in human normal and malignant

tissues based on the number of SAGE tags corresponding

to individual mRNAs [21]. In this database, expression of

COL1A2 in 20 medulloblastoma specimens was signifi-

cantly higher when compared to two normal cerebellum

specimens (Fig. 2d, and see Additional file 3 for data).

Type I collagen is overexpressed in medulloblastoma

COL1A2 gene encodes the alpha 2 subunit of type I col-

lagen. Unlike other types of collagen in which the subunits

of the collagen fibrils heterodimerize in several possible

combinations, type I collagen is always composed of one

alpha 2 chain and two alpha 1 chains (encoded by the

COL1A1 gene). Expression of type I collagen is regulated

primarily at the transcriptional level, and common cis-

acting regulatory elements are present in the promoters of

both COL1A1 and COL1A2 genes [22, 23] This is consis-

tent with the coherent expression patterns between

COL1A1 and COL1A2 in the two groups of normal,

medulloblastoma, and supratentorial PNET specimens used

by Pomeroy et al. [13] in which we verified COL1A2

expression in Figs. 2a,b (group 1, N = 20, P = 1 · 10–8;

group 2, N = 68, P = 2.4 · 10–15; both by Pearson corre-

lation and see Additional file 2 for data). As expected,

COL1A1 mRNA was also significantly higher in these two

sets of medulloblastoma and supratentorial PNET than in

normal cerebellum (Figs. 2ab, and Additional file 2 for

data), so was the number of SAGE tags corresponding to

COL1A1 significantly higher in medulloblastoma than in

normal cerebellum (Fig. 2d).

Overexpression of both the COL1A1 and COL1A2 genes

in both medulloblastoma and supratentorial PNET strongly

suggests increased production of type I collagen in their

tumor microenvironment. We examined the expression of

type I collagen protein in a panel of 17 human medullo-

blastoma specimens using immunohistochemistry. In nor-

mal brain, type I collagen was detected in the

leptomeninges and basement membrane associated with

blood vessels but not in brain parenchyma (Fig. 3a). Two

patterns of type I collagen expression were observed in the

tumor specimens. The first pattern consisted of type I

collagen immunoreactivity surrounding blood vessels,

while the ECM between neoplastic cells was essential

negative (Fig. 3b). The boundaries of the type I collagen-

positive basement membrane were sharply demarcated.

The second pattern of type I collagen immunostaining was

not restricted in the ECM of blood vessels but was dif-

fusely associated with thickened and distended basement

membrane of the tumor vasculature and throughout the

ECM within the tumor (Fig. 3c, e, g). We noticed that both

immunostaining patterns occurred together in some tumors

(data not shown).

b1 integrin subunit is detected in type I collagen-

surrounded medulloblastoma cells and mediates their

in vitro adhesion to type I collagen matrix

Our immunohistochemical data indicate that type I colla-

gen is present in the basement membrane associated with

intra-tumoral vasculature as well as in the ECM sur-

rounding neoplastic cells. This suggests that type I colla-

gen-containing ECM interacts with at least a subset of

neoplastic cells. The b1 subunit of integrin (ITGB1) is a

known constituent of receptors for type I collagen, and

Fig 1 Expression of genes located on chromosome seven in

medulloblastoma and supratentorial PNET specimens compared to

normal brain tissue. The average expression of each gene in three

normal brain specimens was subtracted from its average expression in

all six tumor specimens, and plotted against the base position of this

gene on chromosome seven; the value of gene expression (Y-axis) is

in log2 scale. Five genes with the values greater than two (4-fold

higher expression in tumor than in normal) were selected for further

consideration: from left to right ETV1, CDK6 (two clones), COL1A2,

TAC1, and CAV2. The Student’s t test showed that expression of

ETV1, CDK6, and COL1A2 was significantly greater in tumors than in

normal brains, while expression of TAC1 and CAV2 was not

136 J Neurooncol (2008) 86:133–141

123

could be a candidate for mediating the cellular receptor to

type I collagen. We investigated this possibility by first

examining localization of the ITGB1 in medulloblastoma

tumor tissues. We contrasted the CD31 immunoreactivity

that stains endothelial cells (Fig. 3f) to the type I collagen

immunoreactivity that stains basement membrane of the

vessels and tumoral ECM (Fig. 3g). ITGB1 was detected in

endothelial cells as well as neoplastic cells in regions

where the ECM contained type I collagen (Fig. 3h). Dual

immunofluorescence microscopy also showed that type I

collagen-positive tumoral ECM overlapped with ITGB1

immunoreactivity (Figs. 3j–l).

We then characterized whether type I collagen mediates

medulloblastoma cell adhesion by examining the adhesion

of D283 medulloblastoma cells to type I collagen matrix

in vitro. One hour after plating, D283 cells started to

adhere to a type I collagen-coated surface whereas they

could be easily removed from an uncoated surface by

lightly tapping a culture dish (Fig. 4). Most D283 cells

detached from either coated or uncoated surface after

vigorous shaking (data not shown). AIIB2 and P4G11 are

two monoclonal antibodies that have been previously

characterized for their opposite effects on the ITGB1-

mediated functions, in that AIIB2 blocks binding of b1

integrin subunit to its ligand, [24] whereas P4G11 stimu-

lates the binding [25]. Adherence of D283 cells to type I

collagen is ITGB1-dependent, as AIIB2 abolished the

attachment of cells to type I collagen, while P4G11 had no

such effect (Fig. 4). Our preliminary experiments demon-

strated that the integrin-linked kinase (ILK), a kinase

downstream of ITGB1, was detectable only in neoplastic

cells positive for ITGB1 (Fig. 3i). Anti-type I collagen

polyclonal antibodies co-immunoprecipitated ILK from

D283 cells (Additional file 4) suggesting that medullo-

blastoma cells interacting with type I collagen ECM via

ITGB1 might trigger signaling pathways involving ILK.

Fig 2 Expression of COL1A2 (white boxes) and COL1A1 (grey

boxes) genes is increased in medulloblastoma (Mb) and supratentorial

PNET (PNET) compared to normal brain (Norm) in four independent

groups of specimens (three groups in Pomeroy et al. [13] and

Ramaswamy et al. [19] and one group in SAGE [20] depicted in these

box plots. (a) Two normal brain specimens compared with 10

medulloblastoma (P = 0.007 for COL1A2, P = 0.009 for COL1A1)

and eight supratentorial PNET specimens (P = 0.032 for COL1A2,

P = 0.011 for COL1A1), (b) two normal brain specimens compared

with 60 medulloblastoma (P = 9.8 · 10–5 for COL1A2, P = 2 · 10–

10 for COL1A1) and six supratentorial PNET specimens (P = 0.034

for COL1A2, P = 0.04 for COL1A1), (c) three normal brain

specimens compared with 10 medulloblastoma specimens

(P = 0.001 for COL1A2), (d) SAGE libraries derived from two

normal brain specimens compared with 20 medulloblastoma speci-

mens (P = 0.001 for COL1A2, P = 0.01 for COL1A1). m, extremes;

s, outliers

J Neurooncol (2008) 86:133–141 137

123

Discussion

Microarrays provide a high throughput platform to char-

acterize gene expression profiles of tumor specimens and to

identify clusters of differentially expressed genes that may

mediate specific biological functions or clinical phenotypes

or mark specific cell lineages from which tumors are de-

rived [26]. For example, we have previously used cDNA

microarrays and sequential supervised and unsupervised

algorithmic analyses to identify a group of genes highly

expressed in rapidly progressing glioblastoma tumors. We

validated the prognostic value of the expression of one of

the genes, FABP7, in two independent sets of specimens

[27]. FABP7 was found to be preferentially expressed in

Fig 3 Type I collagen is overexpressed in medulloblastoma and co-

distributed with ITGB1. In normal brain, expression of type I collagen

is associated with the basement membrane of blood vessels (arrow in

a). Two patterns of type I collagen expression were observed in the

medulloblastoma specimens. The first pattern consisted of type I

collagen immunoreactivity in the basement membrane surrounding

blood vessels (arrow in b), but not in the ECM between neoplastic

cells (* in b). The second pattern of type I collagen immunostaining

was diffusely associated with the ECM of the tumor (* in c, e, g) as

well as the vasculature indicated by arrows in from d to i (endothelial

cells were highlighted by CD31 staining in d and f). Tumor cells

surrounded by type I collagen-positive ECM also expressed ITGB1

(h) and ILK (i). Panels d/e and f–i were consecutive sections from the

same specimen. Dual immunofluorescence microscopy showed that

ITGB1 (j) and type I collagen (k) were co-distributed (yellow color in

l) in the ECM surrounding tumor cells counterstained by DAPI (blue

color in l). Bar in a represents 45 lm for from a to i. Bar in jrepresents 15 lm for from j to l

Fig 4 D283 medulloblastoma cells adhered to type I collagen matrix

via the ITGB1 in vitro. The inhibitory monoclonal antibody, AIIB2,

blocked this interaction, whereas the activating monoclonal antibody,

P4G11, did not. The data were plotted from a representative

experiment with quadruplicates, and * indicates a P value of 0.0017

138 J Neurooncol (2008) 86:133–141

123

glioma cells with astrocytic features, [28] and its prog-

nostic value for glioblastoma was associated with EGFR

overexpression of the tumor cells [29]. Because the avail-

ability of only a limited number of medulloblastoma

specimens might easily confound the reliability of clus-

tering analysis, we utilized a different strategy in this

investigation to identify candidate genes of pathological

importance. We built upon recognition of a frequently

occurring chromosomal aberration in medulloblastoma,

gain of chromosome seven, that is shared by supratentorial

PNET, and sought to identify chromosome seven genes

that were overexpressed in both tumor types using cDNA

microarrays. We validated our finding in published data-

bases that used distinctive assay platforms, and repeatedly

found increased expression of the genes encoding the two

subunits of type I collagen. Immunohistochemistry further

confirmed increased deposition of type I collagen protein in

the ECM of medulloblastoma. These findings provide

evidence for roles of type I collagen in the pathogenesis of

medulloblastoma.

ECM components and the cellular machinery that

mediates the interaction of tumor cells with the ECM are of

great importance in efforts to understand the molecular

mechanisms regulating angiogenesis of malignant cells.

We found high levels of expression of type I collagen in

medulloblastoma specimens associated with tumor vascu-

lature. In normal tissues, type I collagen is a major com-

ponent of basement membranes associated with blood

vessels, and it is required for maintaining the integrity of

the vasculature and for regulating angiogenesis, [30, 31]

but type I collagen is not expressed in normal brain

parenchyma. Type I collagen is believed to directly mod-

ulate the behavior of tumor cells mainly based on data from

studies using halofuginone, a low molecular weight qui-

nazolinone alkaloid that inhibits synthesis of type I colla-

gen. While halofuginone inhibits neovascularization by

inhibiting vascular sprouting, tubular formation, and ECM

deposition by endothelial cells both in vitro and in vivo,

[32] it also inhibits tumor invasion and tumor growth in an

animal model for glioma and chemically induced mouse

bladder carcinoma [33–35].

The best-characterized cellular receptors for type I col-

lagen are a family of integrin membrane proteins. Integrins

are heterodimeric glycoproteins composed of one a chain

and one b chain. All type I collagen-binding integrins share

a common b1 integrin subunit [36]. Anti-a2b1 integrin

antibodies inhibit the attachment of endothelial cells to

type I collagen matrix but not to fibronectin or laminin that

are also the components of vasculature ECM [37]. ITGB1

plays central roles in tumor progression, ECM remodeling,

and angiogenesis by transmitting mechanical signals as it

mediates cell adhesion and migration by interacting with

the cytoskeleton proteins, [38] and it also conveys chemical

signals to control cell survival and proliferation by acti-

vating a spectrum of kinases and downstream adaptor

proteins [38]. ILK is a serine-threonine protein kinase that

binds to the cytoplasmic domain of ITGB1 and regulates

ITGB1-mediated signaling following growth factor stimu-

lation [39]. Identification of ILK-binding proteins has

recently demonstrated that the functions of ILK in ITGB1

signaling pathways include anchoring actin filaments and

activating signaling molecules downstream of growth fac-

tor receptors [40].

Type 1 collagen may be of importance in the pathobi-

ology of medulloblastoma because of its role in blood

vessel formation. Medulloblastoma is a highly vascular

tumor and several studies have suggested that anti-angio-

genesis-directed therapy may be a potential strategy for

future approaches to the treatment of medulloblastoma. For

example, an integrin av antagonist peptide successfully

inhibited angiogenesis induced by human DAOY medul-

loblastoma cells growing in chicken chorioallantoic mem-

brane [41] and also reduced growth of DAOY cells

implanted into nude mice [42]. Whether inhibitors of type I

collagen, such as halofuginone, can suppress the spread and

growth of medulloblastoma cells via inhibition of angio-

genesis or by a direct effect on neoplastic cells deserves

further investigation. Type I collagen is also a major

constituent of leptomeninges, [43] and invasion of the

leptomeninges is a frequent route over which medullo-

blastoma cells spread throughout the neuraxis and establish

systemic metastatic disease [44]. However, expression of

type I collagen might not be involved in medulloblastoma

invasion into adjacent brain structures or metastasis through

cerebrospinal fluid, as analyses using the published dataset

did not show correlation between type I collagen expression

and tumor/metastasis stages (data not shown). This is con-

sistent with the findings that the vascularity of medullo-

blastoma is not associated with either metastasis or patient

survival [12, 45].

We demonstrated that type I collagen is also present in

the ECM associated with medulloblastoma cells in vivo.

The type I collagen-positive interstitial matrix may be

essential for the angiogenic behavior of medulloblastoma

cells. Co-distribution of ITGB1 with type I collagen also

suggests the possibility that by binding to its ligand, ITGB1

triggers intracellular signaling pathways that might regu-

late the pathophysiology of medulloblastoma cells, proba-

bly through ILK. In one study, ILK expression was not

seen in normal brain but was detected in all three medul-

loblastoma and four supratentorial PNET specimens

examined [46]. We had preliminary data suggesting that

ILK might be recruited to the type I collagen/ITGB1

complex in D283 cells. These findings support the notion

that the type I collagen/ITGB1 interaction in medullo-

blastoma may activate a signaling cascade via ILK. It

J Neurooncol (2008) 86:133–141 139

123

would be interesting to investigate whether interaction of

type I collagen and ITGB1 relates to angiogenic activities

of medulloblastoma cells.

Our findings suggest the possibility of a role for the

inhibition of type I collagen and its signaling pathways in

the treatment of medulloblastoma. The therapeutic poten-

tial of inhibiting type I collagen synthesis or downstream

signaling events in medulloblastoma might have a two-

prong effect: inhibition of angiogenesis and the tumoral

vasculature in which type I collagen is actively synthe-

sized, and blocking type I collagen-mediated tumor cell

interaction by reducing the activated integrin signaling.

Such therapeutic activities seem very likely to affect the

pathobiology of this tumor in a manner that would have a

significant impact on the clinical behavior of the tumor and

thereby contribute to an improved outcome for patients.

Acknowledgements We thank the Brain Tumor Research Center

Tissue Bank of UCSF for contributing tissue specimens in this study.

This work was supported by funding from the Department of Neu-

rological Surgery at UCSF, and by National Institute of General

Medical Sciences training grant GM07365 (M.D.), and the Theodora

B. Betz Foundation and Kyra Memorial Fund (M.A.I.). UCSF is an

NCI-designated Specialized Program of Research Excellence for

Brain Tumors.

References

1. Jemal A, Murray T, Samuels A (2003) Cancer Statistics. CA

Cancer J Clin 53:5–26

2. Rutkowski S (2006) Current treatment approaches to early child-

hood medulloblastoma. Expert Rev Neurother 6(8):1211–1221

3. Evans AE, Jenkin RD, Sposto R et al (1990) The treatment of

medulloblastoma. Results of a prospective randomized trial of

radiation therapy with and without CCNU, vincristine, and

prednisone. J Neurosurg 72(4):572–582

4. Packer RJ, Sutton LN, Elterman R et al (1994) Outcome for

children with medulloblastoma treated with radiation and cis-

platin, CCNU, and vincristine chemotherapy. J Neurosurg

81(5):690–698

5. Zeltzer PM, Boyett JM, Finlay JL et al (1999) Metastasis stage,

adjuvant treatment, and residual tumor are prognostic factors for

medulloblastoma in children: conclusions from the Children’s

Cancer Group 921 randomized phase III study. J Clin Oncol

17(3):832–845

6. Marino S (2005) Medulloblastoma: developmental mechanisms

out of control. Trends Mol Med 11(1):17–22

7. MacDonald TJ, Brown KM, LaFleur B et al (2001) Expression

profiling of medulloblastoma: PDGFRA and the RAS/MAPK

pathway as therapeutic targets for metastatic disease. Nat Genet

29(2):143–152

8. Dirks PB, Harris L, Hoffman HJ et al (1996) Supratentorial

primitive neuroectodermal tumors in children. J Neurooncol

29(1):75–84

9. Albright AL, Wisoff JH, Zeltzer P et al (1995) Prognostic factors

in children with supratentorial (nonpineal) primitive neuroecto-

dermal tumors. A neurosurgical perspective from the Children’s

Cancer Group. Pediatr Neurosurg 22(1):1–7

10. Cohen BH, Zeltzer PM, Boyett JM et al (1995) Prognostic factors

and treatment results for supratentorial primitive neuroectodermal

tumors in children using radiation and chemotherapy: a Childrens

Cancer Group randomized trial. J Clin Oncol 13(7):1687–1696

11. Huber H, Eggert A, Janss AJ et al (2001) Angiogenic profile of

childhood primitive neuroectodermal brain tumours/medullo-

blastomas. Eur J Cancer 37(16):2064–2072

12. Grotzer MA, Wiewrodt R, Janss AJ et al (2001) High micro-

vessel density in primitive neuroectodermal brain tumors of

childhood. Neuropediatrics 32(2):75–79

13. Pomeroy SL, Tamayo P, Gaasenbeek M et al (2002) Prediction of

central nervous system embryonal tumour outcome based on gene

expression. Nature 415(6870):436–442

14. Russo C, Pellarin M, Tingby O et al (1999) Comparative geno-

mic hybridization in patients with supratentorial and infratento-

rial primitive neuroectodermal tumors. Cancer 86(2):331–339

15. Burnett ME, White EC, Sih S, von Haken MS, Cogen PH (1997)

Chromosome arm 17p deletion analysis reveals molecular genetic

heterogeneity in supratentorial and infratentorial primitive neu-

roectodermal tumors of the central nervous system. Cancer Genet

Cytogenet 97(1):25–31

16. McCabe MG, Ichimura K, Liu L et al (2006) High-resolution

array-based comparative genomic hybridization of medulloblas-

tomas and supratentorial primitive neuroectodermal tumors. J

Neuropathol Exp Neurol 65(6):549–561

17. Bayani J, Zielenska M, Marrano P et al (2000) Molecular cyto-

genetic analysis of medulloblastomas and supratentorial primitive

neuroectodermal tumors by using conventional banding, com-

parative genomic hybridization, and spectral karyotyping. J

Neurosurg 93(3):437–448

18. Reardon DA, Michalkiewicz E, Boyett JM et al (1997) Extensive

genomic abnormalities in childhood medulloblastoma by com-

parative genomic hybridization. Cancer Res 57(18):4042–4047

19. Ramaswamy S, Tamayo P, Rifkin R et al (2001) Multiclass

cancer diagnosis using tumor gene expression signatures. Proc

Natl Acad Sci U S A 98(26):15149–15154

20. Boon K, Osorio EC, Greenhut SF et al (2002) An anatomy of

normal and malignant gene expression. Proc Natl Acad Sci U S A

99(17):11287–11292

21. Velculescu VE, Zhang L, Vogelstein B, Kinzler KW (1995)

Serial analysis of gene expression. Science 270(5235):484–487

22. Karsenty G, Park RW (1995) Regulation of type I collagen genes

expression. Int Rev Immunol 12(2–4):177–185

23. Karsenty G, de Crombrugghe B (1991) Conservation of binding

sites for regulatory factors in the coordinately expressed alpha 1

(I) and alpha 2 (I) collagen promoters. Biochem Biophys Res

Commun 177(1):538–544

24. Werb Z, Tremble PM, Behrendtsen O, Crowley E, Damsky CH

(1989) Signal transduction through the fibronectin receptor in-

duces collagenase and stromelysin gene expression. J Cell Biol

109(2):877–889

25. Wayner EA, Gil SG, Murphy GF, Wilke MS, Carter WG (1993)

Epiligrin, a component of epithelial basement membranes, is an

adhesive ligand for alpha 3 beta 1 positive T lymphocytes. J Cell

Biol 121(5):1141–1152

26. Quackenbush J (2006) Microarray analysis and tumor classifi-

cation. N Engl J Med 354(23):2463–2472

27. Liang Y, Diehn M, Watson N et al (2005) Gene expression

profiling reveals clinically distinct subtypes of glioblastoma

multiforme. Proc Natl Acad Sci U S A 102(16):5814–5819

28. Liang Y, Bollen AW, Nicholas MK, Gupta N (2005) Id4 and

FABP7 are preferentially expressed in cells with astrocytic fea-

tures in oligodendrogliomas and oligoastrocytomas. BMC Clin

Pathol 5:6

29. Liang Y, Bollen AW, Aldape KD, Gupta N (2006) Nuclear

FABP7 immunoreactivity is preferentially expressed in infiltra-

tive glioma and is associated with poor prognosis in EGFR-

overexpressing glioblastoma. BMC Cancer 6:97

140 J Neurooncol (2008) 86:133–141

123

30. Montesano R, Orci L, Vassalli P (1983) In vitro rapid organiza-

tion of endothelial cells into capillary-like networks is promoted

by collagen matrices. J Cell Biol 97(5 Pt 1):1648–1652

31. Jackson CJ, Jenkins KL (1991) Type I collagen fibrils promote

rapid vascular tube formation upon contact with the apical side of

cultured endothelium. Exp Cell Res 192(1):319–323

32. Elkin M, Miao HQ, Nagler A et al (2000) Halofuginone: a potent

inhibitor of critical steps in angiogenesis progression. Faseb J

14(15):2477–2485

33. Abramovitch R, Dafni H, Neeman M, Nagler A, Pines M (1999)

Inhibition of neovascularization and tumor growth, and facilita-

tion of wound repair, by halofuginone, an inhibitor of collagen

type I synthesis. Neoplasia 1(4):321–329

34. Elkin M, Ariel I, Miao HQ et al (1999) Inhibition of bladder

carcinoma angiogenesis, stromal support, and tumor growth by

halofuginone. Cancer Res 59(16):4111–4118

35. Gross DJ, Reibstein I, Weiss L et al (2003) Treatment with

halofuginone results in marked growth inhibition of a von Hippel-

Lindau pheochromocytoma in vivo. Clin Cancer Res 9(10 Pt

1):3788–3793

36. White DJ, Puranen S, Johnson MS, Heino J (2004) The collagen

receptor subfamily of the integrins. Int J Biochem Cell Biol

36(8):1405–1410

37. Gamble JR, Matthias LJ, Meyer G et al (1993) Regulation of

in vitro capillary tube formation by anti-integrin antibodies. J

Cell Biol 121(4):931–943

38. Guo W, Giancotti FG (2004) Integrin signalling during tumour

progression. Nat Rev Mol Cell Biol 5(10):816–826

39. Hannigan GE, Leung-Hagesteijn C, Fitz-Gibbon L et al (1996)

Regulation of cell adhesion and anchorage-dependent growth by

a new beta 1-integrin-linked protein kinase. Nature

379(6560):91–96

40. Wu C, Dedhar S (2001) Integrin-linked kinase (ILK) and its in-

teractors: a new paradigm for the coupling of extracellular matrix

to actin cytoskeleton and signaling complexes. J Cell Biol

155(4):505–510

41. MacDonald TJ, Taga T, Shimada H et al (2001) Preferential

susceptibility of brain tumors to the antiangiogenic effects of an

alpha(v) integrin antagonist. Neurosurgery 48(1):151–157

42. Taga T, Suzuki A, Gonzalez-Gomez I et al (2002) alpha v-Inte-

grin antagonist EMD 121974 induces apoptosis in brain tumor

cells growing on vitronectin and tenascin. Int J Cancer

98(5):690–697

43. Rutka JT, Giblin J, Dougherty DV et al (1986) An ultrastructural

and immunocytochemical analysis of leptomeningeal and

meningioma cultures. J Neuropathol Exp Neurol 45(3):285–303

44. Rutka JT (1997) Medulloblastoma. Clin Neurosurg 44:571–585

45. Ozer E, Sarialioglu F, Cetingoz R et al (2004) Prognostic sig-

nificance of anaplasia and angiogenesis in childhood medullo-

blastoma: a pediatric oncology group study. Pathol Res Pract

200(7–8):501–509

46. Chung DH, Lee JI, Kook MC et al (1998) ILK (beta1-integrin-

linked protein kinase): a novel immunohistochemical marker for

Ewing’s sarcoma and primitive neuroectodermal tumour. Virch-

ows Arch 433(2):113–117

J Neurooncol (2008) 86:133–141 141

123