INFORMATION TO USERS

This manuscript has been reproduœd from the microfilm master. UMI films the text directly from the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, pnnt bleedthrough, substandard margins, and improper alignment can adversely affect reproduction.

In the unlikely event that the author did not send UMI a cornplete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyn0ght material had to be removed, a note will indicate the deletion.

Oversize materials (e.g., rnaps, drawings, charts) are reproduœd by sedioning the original, beginning at the upper left-hand corner and continuing from left to right in equal sections with small overlaps. Each original is also photographed in one exposure and is induded in reduced form at the back of the bwk.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6^ x 9" black and white photographie prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

Bell & Howell Information and Leaming 300 North Zeeb Road, Ann Ahor, MI 48106-1346 USA 8oO-Wl-06W

THE POTENTIAL ROLE OF BASIC FIBROBLAST GROWTH FACTOR IN MALIGNANT TRANSFORMATION, ANGIOGENESIS, INVASION AND PROLIFERATION OF HUMAN GLIOMAS

Stephen T. Gately, B.A.

A thesis submitted to the Faculty of Graduate Studies and Research, McGill

University, in partial fulfillment of the requirementç for the degree of

Doctor of Philosophy

Department ,of Neurology and Neurosurgery

McGill University

Montreal, Canada

Stephen T. Gately O 1997 National Library Bibliothèque nationale 191 of Canada du Canada Acquisitions and Acquisitions et Bibliographie Services services bibliographiques 395 Wellington Street 395, rue Wellington Ottawa ON K1A ON4 Ottawa ON KIA ON4 Canada Canada Your fi& Votre n5férenœ

Our file Notre rëler6nce

The author has granted a non- L'auteur a accordé une Licence non exclusive licence allowing the exclusive pennettant à la National Library of Canada to Bibliothèque nationale du Canada de reproduce, loan, distribute or sel1 reproduire, prêter, distribuer ou copies of this thesis in rnicrofom, vendre des copies de cette thèse sous paper or electronic formats. la fonne de microfiche/film, de reproduction sur papier ou sur format électronique.

The author retains ownership of the L'auteur conserve la propriete du copyright in this thesis. Neither the droit d'auteur qui protège cette thèse. thesis nor substanîial extracts fiom it Ni la thèse ni des extraits substantiels may be printed or otherwise de celle-ci ne doivent être imprimés reproduced without the author's ou autrement reproduits sans son permission. autorisation. ABSTRACT

The role of basic fibroblast growth factor (bFGF) in the malignant phenotype of human gliomas was examined in this dissertation. A bFGF ELISA, demonstrated to detect recombinant human bFGF protein, was used to quantitate bFGF protein in glioma tissue samples. Basic FGF protein was significantly elevated in low and high grade glioma tissues, suggesting increased bFGF expression could be an early event in the progression of gliomas. Because the tumor sections contained non-tumor elements, bFGF gene and protein expression were examined in established human glioma ceIl lines. Human glioma cells demonstrated extracellular release of bFGF and expression of higher molecular weight bFGF isoforms. Because secreted bFGF could be important in the malignant phenotype of gliomas, fibroblasts were stably transfected with an expression vector for a secreted bFGF. These cells were more invasive in vitro and in the brain. To confirm and extend these data, non-neoplastic human astrocytes were stably transfected with the bFGF secretion vector. These cells demonstrated features associated with rnalignant transformation, suggesting increased expression and extracellular release of bFGF could have transforming potential in human astrocytic cells. To determine if bFGF gene and protein expression were required for human glioma ce11 growth, bFGF-specific antisense oligodeoxynucleotides were tested. The antisense treated glioma cells demonstrated deaeased bFGF gene and protein expression and significantly deaeased proliferation. Taken together, the data suggests the potential importance of bFGF in glioma ce11 transformation and growth, and indicates anti-bFGF strategies could be useful in the management of glioma patients. Le but de cette thèse a été d'examiner le rôle du bFGF dans la phénotype malin des gliomes humaines. Le taux élevé de bFGF dans des tissus 5 basse et ii haute concentration gliomateuse, indiquant une expression accrue, est un des premiers énvéments dans le développement des gliomes. Les cellules humaines, contrairement aux astrocytes non-transformés, ont démontré une libération extra-cellulaire de bFGF et une expression d'isoformes bFGF d'un poids moléculaire plus éléve. Les fibroblastes transfectés de facon stable à l'aide d'un vecteur qui permet l'expression et la sécrétion du bFGF se sont avérés invasifs. Ceci semble indiquer que le bFGF joue un rôle déterminant dans le degré d'invasion du cerveau par les fibroblastes. La transfection stable d'astrocytes humains non-néoplastiques avec un de ces vecteurs entraîne leur transformation, ce qui suggère que le bFGF est un élément transformant des cellules gliales. L'inhibition de l'expression de bFGF par des oligodeoxynucléotides a inhibé la prolifération des cellules gliomateuses. On peut en déduire que la prolifération gliomateuse est bFGF dépendante. Ces données démontrent le rôle du bFGF dans la transformation et la prolifération des cellules gliomateuses et indiquent également que les stratégies anti-bFGF pourraient s'avérer très utiles pour le traitement de patients atteints de gliomes. PREFACE

This thesis is written in the form of publications, except for the references that are Uted in the final section (Bibliography) of this dissertation. Chapters 114 are joined by linking pages to give continuity. This option is provided by Section 3 of the Guidelines Concerninp: Thesis Preparation: "Cnndidntes have the option of including, as part of the thesis, the text of one or more papers siibmitted or to be szrbmitted for publication, or the clearly- diiplicnted text of one or more published papers. These texts must be bo und as an integral part of the thesis.

If this option is chosen, connecting texts that provide logical bridges betwee~z the different papers are mandntory. The thesis must be written in

siich LI wny thnt if is more than a collection of manuscripts; in other words, reszilts of n series of papers mut be integrnted.

The thesis rnlist still conform to al1 other reqziirements of the "Guidelines for Thesis Preparntion". The thesis mirst include: A Table of Contents, an abstract in English and French, an introdliction which clenrly states the rationale and objectives of the study, a review of the literntirre, a final conclrision and summary, and a thorolrgh bibliography or reference list.

Additionnl materid must be provided where appropriate (eg. in appendices) and in szlficient detail to allow a clear and precise judgment to be made of the importance and originality of the research reported in the fhesis.

In the case of rnanuscripts coauthored by the candidate and others, f he candidate is reqlrired tu make an explicit statement in the thesis as to w ho contribiited to srrch work and to what extent. Superoisors must attest tu the acciiracy of srich stntements at the doctortzl oral defense. Since the task of the examiners is made more difictilt in these cases, it is in the candidates interest to mnke perfectly clear the responsibilities of al1 authors of the CO-authored papers. /?

Chapter I is a review of the literature on basic fibroblast growth factor and human brain tumors, providhg the fondation for the experiments desdbed in this dissertation. The important points and principal conclusions of this dissertation are summarized in Chapter VI.

Publications:

1.) Gately, S., Tsanaclis, A.M.C., Takano, S., Klagsbnui, M., Brem, S. Cells transfected with the basic fibroblast growth factor (bFGF) gene fused to a signal sequence are invasive in vitro and in situ in the brain. Neurosurgery, 36: 780- 785,1995. (see Appendix 1).

2.) Gately, S., Soff, GA., Brem, S. The potential role of basic fibroblast growth factor in the transformation of cultured prirnary human fetal astrocytes and the proliferation of human gliorna (U-87) cells. Neursurgery, 32723-732,1995. (see Appendix 1). ACKNOWLEDGMENTS

1wish to express my sincere appreciation to my thesis supervisor, Cr. Steven Brem. It was a privilege and honor to train under Dr. Brem, an internationally recognized expert on brain tumor angiogenesis. Under his direction 1 had the opportunity to study the cellular and molecular mechanisms involved in the regdation of tumor angiogenesis and invasion, and to explore potential new and exciting therapeutic agents, that may someday prove to be beneficial in the management of human brain tumors. Outside of the laboratory, Dr. Brem was supportive and welcorned me as a member of his family. 1 am grateful for the professional and persona1 experiences I have shared with Dr. Brem; they have had a significant positive influence on my life, and 1will remain indebted to him for this.

I would like to thank my CO-supervisor Dr. Jack Antel and the members of my comrnittee, Dr. Voon Wee Yong, and Dr. Josephine Nalbantaglu for helpful discussions and support. 1would also like to thank the department chairman, Dr. Daniel Guitton, for his patience and assistance, and Monique Ledermann for al1 her help and support.

The majority of the experimental work was performed in the Brain Tumor Research Laboratory at the Lady Davis Institute for Medical Research. The author wishes to express his gratitude to the then director, Dr. Norman Kalant, and to Mrs. Therese Davila and all her team for their help and kirtdness. I also wish to extend my gratitude to Dr. Ana Maria Tsanacliç, for reviewing the pathological material, and instructing me in neuropathology. Acknowledgments to David Ivancic and Marguerite Wotoczek for expert technical assistance. vii

Dr. Gerald Soff at the Northwestem University School of Medicine has been an invaluable teacher of molecular biological techniques. Dr. Soff has given generously of his thne and expertise, and I consider him one of my closest advisors. Ms. Deborah Cundiff has also been an invaluable colleague, whose expert technical expertise was much appreciated.

The experimental work presented in this dissertation was made possible by persona1 fellowship grants that induded the Mime Fellowship in liver and vascular disease, and the Illinois Neurofibromatosis hcorporated; 1 remain indebted to these groups. In addition the work was supported by the Medical Research Council of Canada, and the National Institutes of Health Grants CA57781 and CA60177.

Finally 1 wish to express appreciation to my family. Without their love, patience, understanding, emotional and financial support none of this work would have been possible. viii

ABBREVIATIONS

bFGF basic fibroblast growth factor

ke base pair

cDNA complementary deoxyribonucleic acid

CRNA complementary ribonucleic acid

DMEM Dulbecco's modified Eagle's medium

DNA deoxyribonucleic acid

ELISA enzyme-linked immunosorbent assa y

FBS fetal bovine serum

GFAP glial fibrillary acidic protein

kilobase

messenger ribonucleic acid

nanogram

nucleotides

PBS phosphate buffered saline

PCR polymerase diain reaction P8 picogram

RT-PCR reverse transcription-polymerase diain reaction

P8 microgram

PI microliter

Pm micrometer Table of Contents

PREFACE +* ~~~~~o~~~~~~~~~*o~~*~.*~***o~~*~~~~~~m~o~o~~m~~~~I~~~~~~~oo~~oooomooo+moIV

ACKNOMÏLEDGMENTS...... e...... *e**....*...m...*.....VI ABBREVIATIONS....~...... e...... VI11

TABLE OF CONTENTS ...... ,...... -...... o...... x LIST OF FIGURES...... XIV

BASIC FIBROBLAST GROWH FACTOR .A REWEW ...... o.,..0.00***....e..*e*t16

INTRODUCTION ~*~*o.**te*.**.*e*.~~********~~~o~~~~*~e*e**~m~*~~mo**~~~**~~*oo~~o~~*~~~~~~~oo*e~o~a~~~~~*e~n~~em**~*~~~*17 RATIONALE ...... J7 GOALS...... 17

Primary Intracranial Gliomas ...... 18 Basic Fibroblast Growth Factor ...... 19 bFGF Gene Structure ...... 20 bFGF and Human Brain Tumors ...... 21 Extracellular Export of bFGF ...... 22 High Molecular Weight bFGF...... 30 - - bFGF Receptors ...... 30

MATERIALS AND METHODS ...... C~~~.m.~~~~m~~~~~m~~~~~~.~,~.~.~~~~~.,~.~~.~e~e~~~~,..J

ELISA Detection of bFGF ...... , ...... 6 Tissue Preparation ...... -.....* ...... -...... -...... 37 Western Blot Analysis ...... b-.*38 Northem Blot Analysis ...... 38 Hybridization to 32~-labeledprobes ...... 39 Statistical Analysis ...... 39 RESULTS ...... o..o.o.040

Basic Fibroblast Growth Factor Gene and Protein Expression in Human Glioma Cells...... 59

ABSTRACT ~~~~~~~*~~~~o~~oo~~~~~~~~~~~~~~~~~~~~~~~~~~~~eoeeooooo~~~~~~~~~~~~~~*~***~~*~~6O

MATERIALS AND METHODS ...... 61 Ce11 Culture ...... 61 Chernoinvasion Assay ...... 62 Quantitation of bFGF by ELISA...... 62 Southern Blot Analysis ...... 63 Northern Blot Analysis ...... 63 Hybridization to 3+labeled9 Probes...... A4 Western Hot Analysis ...... 65 Statistical Analysis ...... 65 RESULTS ...... 65 Quantitation of bFGF by ELISA ...... A6 Southern Blot Analysis ...... 66 Northem Blot Analysis ...... 66 ..,...... bFGF mRNA ...... -...... ,...... 66 bFGF Receptor-2 mRNA ...... 73 Western Blo t Analysis ...... 73

DISCUSSION ...... o.. ~~~~~~~o~~~~~~~**~~~~~~~~~~~~~~~~b~~~~o~~~~o~~~~~ooo~~~~o~~~m~~~~o~o~~oo~o~~~~~~~oo~~~8~

CHAPTER IV ...... 90 Cells Transfected with the Basic Fibroblast Growth Factor Gene Fused to a Signal Sequence are Invasive In Vitro and In Situ in the Brain ...... 90 ABSTRACT ...... o.eo..e...*.91

MATERIALS AND METHODS ~~~OO~~~O~~O~~~OO~~~~O~~1~~oo~~~oo~eb~~b~O~O~oo~o~~~~Oo~~oo~~~~~~~~~~~~O~O~~~~~~~~~~ Basic Fibroblast Growth Factor Quantitation...... 93 Western Blot Analysis ...... 94 In Vitro Invasiveness Assays ...... 95 Chernoinvasion Assay...... 95 xii

Cellular Migration ...... 96 Zymography ...... 96 Comeal Angiogenesis Assay ...... 97 Brain Tumor Model ...... 98 Statistical Analysis ...... 100

Basic FGF Quantitation...... 100 In Vitro Invasiveness Using Matrigel...... 105 Chernoinvasion Assay...... 105 Cellular Migration ...... 105 Zymography ...... 106 Angiogenesis...... 106 Brain Tumor Model of hvasiveness ...... JO6

CHAPTER The Potential Role of Basic Fibroblast Growth Factor in the Transformation of Cultured Primary Human Fetal Astrocytes and the Proliferation of Human Glioma U-87 Cells ...... 128

INTRODUCTION

Ce11 Culture ...... 130 Ce11 Proliferation Assay ...... 131 Basic FGF and Fetal Astrocyte Proliferation...... +...... 131 Reverse Transcription and Polymerase Chain Reaction ...... 132 Transfections...... *...... *...... *...... *...... *...... 133 Western Blo t Analysis ...... 133 Glial Fibrillary Acidic Protein Imrnunocytochemistry...... 136 Oligodeoxynudeotide Synthesis...... J36 Antisense Inhibition of Glioma Ce11 Proliferation ...... 137 Northem Blot Analysis ...... 137 Hybridiza tion to 3%'-labeled probes ...... 138 Quantitation of bFGF by ELISA ...... 138 Statisücal Analysis ...... 139 RESULTS ...... w0.0139 Fetal Astrocyte Proliferation...... 139 Basic FGF Receptor ...... 140 Transfections...... ,...... *...... 140 Western blot Analysis of Transfected Clones ...... *...... 145 Antisense Inhibition of GIioma Growth ...... 145 xiii

Conclusions...... ,...... 162 CONTRIBUTION TO ORIGINAL KNOMILEDGE ...... 00*0******169

Chapter 1...... 171 Chapter II ...... 183 Chapter III ...... ~...... 188 Chapter IV ...... 195 Chapter V ...... 204 Chapter VI ...... 213 xiv

List of Figures CHAPTER I FIGURE 1 Overlap of bFGF Sense and Antisense mRNA ...... -25 FIGURE 2 . Immunolocalization of bFGF in Brain Tumor Tissues ...... 27 FIGURE 3 . Cellular Pathways for bFGF Activities ...... J CHAPTER II FIGURE 1 . Detection of Recombinant Human bFGF by ELISA ...... 43 FIGURE 2 . bFGF Levels in Snap Frozen and Archived Frozen Tissues ...... 45 FIGURE 3 . Quantitation of bFGF in Hurnan Brain Tumors ...... 47 FIGURE 4 bFGF Levels in Human Gliomas ...... 49 FIGURE 5 bFGF Levels in Gliomas and Recurrent Tumors ...... 51 FIGURE 6 Quantitation of bFGF mRNA Expression ...... 53 CHAPfER III FIGURE 1.Glioma Cell Invasion In Vitro ...... 68 FIGURE 2 Quantitation of Glioma CeU-Associated bFGF Protein ...... 70 FIGURE 3 Quantitation of Released bFGF Protein ...... J FIGURE 4 Southem Analysis of bFGF in Human Glioma Cells...... -75 FIGURE 5 Northem Analysis of bFGF Gene Expression ...... 77 FIGURE 6 Cycloheximide Induction of bFGF mRNA Expression ...... 79 FIGURE 7 bFGF Recep tor-2 mRNA Expression ..*...... **...*...... *..*...81 FIGURE 8 High molecular weight bFGF Isoform Expression ...... 83

CHAPTER IV FIGURE 1 Quantitation of Cell-Associated and Released bFGF ...... 32 FIGURE 2 Detection of higher molecular weight bFGF isofoms ...... JO4 FIGURE 3 Cellular morphology on Matrigel...... 108 FIGURE 4 In Vitra Invasiveness ...... 110 FIGURE 5 Quantitation of Cellular Migration ...... 112 FIGURE 6 Gelatui Zymography ...... 117 FIGURE 7 WiId-type-bFGF Cells Do Not Invade the Brain...... Il9 FIGURE 8 Signal-peptide-bFGF Cells Invade the Brain ...... 121 FIGUE 9 Intracerebral Tumor Cell Kinetics ...... 123

CHAPTER V FIGURE 1 Restriction Map of the Signal-Pep tide-bFGF cDNA...... 135 FIGURE 2 Human Fetal Astrocyte Response to bFGF protein ...... 142 FIGURE 3 Detection of bFGF receptor mRNA ...... 144 FIGURE 4 Cellular Morphology ...... 147 FIGURE 5 GFAP Expression ...... 149 FIGURE 6 . Expression of Higher Molecular Weight bFGF ...... 151 FIGURE 7 Suppression of Glioma Ce11 Proliferation ...... 153 FIGURE 8 Suppression of bFGF mRNA Expression ...... J55 FIGURE 9 Suppression of bFGF Protein Expression ...... 153

CHAPTER VI Figure 1 Cellular Model for bFGF Involvement...... 168 CHAPTER I Basic Fibroblast Growth Factor - A Review INTRODUCTION

RATIONALE

"The correlation of gene products and their mechanisms of action with different aspects of glioma, particularlv the develo~ment of malignant potential, must be one of the main goals of brain tumor pathology" (63). The central hypothesis of this dissertation can be stated: Bnsic fibroblnst grozoth factor is a transforming factor, and key positive cellular and molecnlar rnedintor of the malignant phenotype of hzlrnan gliorn~s, inclriding involvement in angiogenesis, invasiveness, and cellular prolqeratio n . This hypothesis predicts that alternative expression, or aberrant forms of bFGF could be involved in the pathogenesis of hurnan gliomas. It also suggests

that inhibitors of bFGF Amotein or Veene exmessionI would have theraoeuticA benefit at multiple sites arresting growth by angiosuppression, blocking the infiltrative spread of malignant tumors, and /or inhibiting glioma ce11 prolifera tion.

GOALS

The major goals of this study are:

1). To quantitate bFGF protein in primary human brain tumors using a sensitive bFGF enzyme-hked immunosorbent assay. Quantitation of bFGF protein will allow clinical correlative studies of bFGF levels with angiogenesis, învasiveness, edema, proliferation, and patient survival, and could be usefd in the identification of a subpopulation of patients that may benefit from anti-growth factor strategies 2). To characterize bFGF gene and protein expression in human glioma ce11 in vitro. Basic FGF has been identified as a key regulator of angiogenesis and invasion in a variety of non-central nervous system tumors, but little is known about its role in glioma cell biology.

3). To develop models in vitro and in situ in the brain to study the effects of hcreased expression of intracellular and extracellular bFGF.

4). To determine if bFGF is a transforming factor for non-transformed human astrocytes that remains necessary for the continued growth of established human glioma cells.

5). To determine the effectiveness of bFGF antisense DNA on the inhibition of human glioma ce11 growth. These studies will provide needed information on the role of the bFGF gene in human glioma cell growth, and could provide the impetus for novel bFGF gene based therapies that may be usehl in the management of patients with rnalignant gliomas.

Primary lntracranial Gliomas

The World Health Organization (WHO) classification system recognizes four histological grades of glioma malignancy including: 1, pilocytic astrocytoma; II, fibrillary, protoplasrnic, gemistocytic, and subependymal gant cell asirocytomas; III, anaplastic astrocytoma; and IV, glioblastoma multiforme. Collectively the gliomas are the most frequent primary brain tumor, that constitute 40-45'' of al1 intracranial tumors. 20.5% of gliomas are of WHO grades 1-III, and 55% are glioblastorna, WHO grade IV (61). Despite recent advances in the molecular biology of human gliomas, and improved neurosurgical techniques, the prognosis for patients with malignant gliomas remains poor: with a median survival of less than one year, and fewer than 10% of patients surviving two years (27). This dismal survival is primarily a function of the invasive spread of malignant gliomas (13), yet little is known about the cellular and molecular mechanisms that regulate this process. Combined with local infiltrative growth, malignant glial tumors stimulate intense angiogenesis (7,9),and have an elevated proliferative rate (72). The identification of a transforming factor that links an~ionenesis,invasion, and elevated cellular ~roliferationin diornas wouid allow for the rational desi~nof novel therapies for the treatment of malignant diornas. t

There is hcreasing scientific evidence that links bFGF with these phenotypes in non central nervous system tumors, however, there have been only a limited number of cellular and molecular mechanisms proposed to explain the role of bFGF in glial tumors, and no reports on the transforming potential of this bFGF in human astrocytic cells.

Basic Fibroblast Growth Factor

The fibroblast growth factors were purified based on their strong affinity for heparin. Once purified, amino terminal sequence analysis identified two forms of acidic FGF, and a single 146 amino acid form of basic FGF (24). Cornparison of the amino atid sequences of acidic FGF and bFGF revealed 55% homology, suggesting the two forms of FGFrs arose by duplication and divergence from a common gene (24). Identification of the arnino acid sequence for the FGF's consolidated a growing list of heparin binding proteins once thought to be unique, to be in fact one or the other FGF. Endothelial ce11 growth factor, eye-derived growth

factor II, heparin binding growth factor a, retinal derived growth factor, astroglial growth factor 1 and prostatropin were identical to aFGF, and eye- derived growth factor 1, heparin binding growth factor 13, astroglial growth factor 2, cartilage-derived growth factor, and chondrosarcoma-derived growth factor were found to be equivalent to bFGF (24).

Basic FGF has been purified from most mesodermal tissues including brain, pituitary, kidney, adrenal glands, corpus luteum, placenta, retina, macrophage, chondrosarcoma, and thymus (24,38). In vitro, bFGF has been shown to be a potent mitogen for cells of ectodermal, mesodermal, endodermal, and neuroectodermal origin. In addition to inducing cellular proliferation, bFGF has been shown to induce cellular motility, differentiation, extension of neurites and enhance neuronal sumival (24,38). In vivo, bFGF has been shown to be involved in embryonic and fetal development, angiogenesis, and wound healing (38). Inappropriate expression of bFGF has been irnplicated in a variety of pathological processes that result in excessive cellular proliferation and angiogenesis (38).

bFGF Gene Structure

The bFGF gene has been estimated to be 38 kB in length, and was cioned and sequenced from bovine and human ïibraries. The gene was mapped to chromosome 4, and was fond to contain 2 introns and thxee exons (1). The bFGF gene gives rise to multiple polyadenylated mRNA's of 7.0, 3.7, 22and 1.5 kB that differ from each other in the length of the 3'- untranslated region; the functional role of the various rnRNA's however, remains unknown (48). Examination of the 5'-untrançlated region of the bFGF mRNA identified five cis-acting RNA elements (two in the 5'- untranslated region and three in an altematively translated region) that have specific effects on the level of synthesis of different bFGF protein isoforms. These cis-acting factors in tum appear to be the targets of tram-acting factors (52). Taken together these data suggest the regdation of bFGF gene expression is cornplex.

Examination of bFGF mRNA trznscnp ts in xenopus oocytes identified that the 1.5 kB bFGF mRNA transcript was present in 20-fold excess over the other bFGF transcripts (30,73). Strand-specific northem blots identified the 1.5 kB bFGF transcript was an "antisense" transcript, that overlapped a portion of exon III of the bFGF gene (Figure 1). Recently the mammalian equivalent of this antisense transcript was identified and cloned (48). The antisense bFGF transcript was undetectable in the human U-87 human glioma ce11 line, known to overexpress the bFGF sense mRNA. By contrast this antisense transaipt was readily detectable in tissues with very low levels of the sense bFGF mRNA (48). These data suggested that the antisense bFGF transcript could regulate bFGF gene and protein expression in mammalian ceus.

bFGF and Human Brain Tumors

Basic FGF is a potent angiogenic mitogen (21), that has been identified at elevated levels in human gliomas (26,42-44,49,67,68,77),and in the microvasculature of glioblastomas, possibly stored in the basement membrane (2,10,65) (Figure 2 A,B). In vitro, bFGF induces cell motility (66) and glioblastoma cdmigration, and invasion (18,36) and stimulates the production of proteases (6,46). Recently, it was found that glioma cell growth was dependent on bFGF (25,43) a function independent of its role in tumor angiogenesis (43). The data also suggested that bFGF release, or secretion was required for the growth promoting propeties (43); one of the unique features of bFGF is the absence of a classical hydrophobic leader sequence that is required for the release of proteins to the extracellular space (38). Despite the largely extracellular activity of these factors only a few mechanisrns have been proposed to explain FGF release (19). Figure 3 describes the possible sites of action for bFGF in human glial tumors. Al1 but one of these pathways requires extracellular bFGF, however, the mechanism of release of bFGF remains unknown (19,40). The identification of a family of seven FGF-related genes (Table 1) that share 30-45% homology to bFGF (3) could provide some insight. Four of these genes, FGF-3 (15,16,50), FGF-4 (14,33,76),FGF-5 (78,79),FGF-6 (37), have been identified as transforming oncogenes, that are associated with secretory signal peptide sequences (8). Despite the importance of these FGF-like oncogenes, there has been only one report on the identification of these genes in human gliomas (25).

Extracellular Export of bFGF

Secreted proteins are first processed in the rough endoplasmic reticulum, then sorted for secretion in the golgi apparatus. The genes for these seaetory proteins code for a 16 to 26 amino aad long signal sequence that are recognized by signal recognition particles, that direct the mRNA to the membrane of the endoplasmic reticulum (54). The role of a signal peptide in association with bFGF and other autocrine growth factors has received increasing attention. Deletion of the signal sequence from the v-sis (4,28), IL3 (17), FGF-4 (69) and FGF-6 (12) oncogenes dramatically reduces the transforming capacity of these genes. Human tumors with known amplification of the signal peptide-bearing bFGF-like oncogenes are reported to be more invasive: coamplification of the FGF-3 and FGF-4 oncogenes in esophageal carcinomas is correlated with metastatic spread and recurrence (31,35); amplification of the FGF-3 oncogene in breast carcinoma is linked to metastatic spread (35); and coarnplification of the FGF-3 and FGF-4 oncogenes is a property specific to invasive ductal carcinomas (71).

Experimental fusion of a code for a signal peptide sequence to the bFGF gene converts a non-transforming growth factor (bFGF) into a transforming protein that is analogous to the bFGF-related oncogenes (75). NIH 3T3 fibroblasts transfected with the bFGF gene fused with a secretory signal sequence (sp-bFGF) are transformed in nitro (22,59) and tumorigenic in syngeneic mice (60). The rnechanism of transformation for the sp-bFGF cells could involve an intemal autocrine loop (11,75) where bFGF can interact with high affinity intracellular receptors, or classical autocrine transformation, where extracellular bFGF interacts with high affinity ce11 surface receptor. The switch to the extracellular export of bFGF has been shown to be an important molecular event in the development of a malignant phenotype in fibrosarcoma (29). This switch is associated with angiogenesis, invasion, increased tumorigenicity and metastasis (23,29). Experimental evidence suggests that glioma cells secrete bFGF (43), however, this has not been demonstrated conclusively. Figure 1. The overlap region between the sense and antisense strands is shown, and is known to be approximately 900 bp. (The diagram is not to scale). Sense bFGF mRNA 3'untranslated

Overlap region

...... ***..**.*..* ......

3'unîransla ted

1.5 kB bFGF antisense mRNA Figure 2. A. bFGF protein was immunolocalized primarily to the microvasculature of a glioblastorna multiforme, with positive immunoreactivity identified by the arrows. B. Using immunogold electron microscopy, bFGF (small black dots) was identified primarily associated with the basement membrane (marked by the anow) of an ependymoma. L - vesse1 lumen.

FIGURE3. Pathway 1 demonstrates that bFGF released by an uncharacterized mechanism, or by ce11 death can stimulate angiogenesis; the release of bFGF can also stimulate surrounding glioma celIs through a paracrine mechanism demonstrated in Pathway 2; Pathway 3 demonstrates classical autocrine stimulation where secreted bFGF interacts with the same ce11 surface receptor, and: Pathway 4 represents an intemal autocrine loop where bFGF exerts its biological activity intracellularly. Endothelial Cell High Molecular Weight bFGF

The human bFGF gene has one ATG codon that initiates the translation of a 155 amino acid protein having a molecular weight of approximately 18 kD (1). Several higher molecular weight forms of the bFGF protein have been detected that could not be explained from initiation at this start codon (32,45,64). The presence of alternative initiation sites upstream (53) that predicted higher molecular mass bFGF proteins corresponding to 22.5, 23.1, and 24.2 kD (20) have now been reported. The significance of the higher molecular weight proteins remains unknown, however, the alternative initiation sites were characteristic of nuclear retention signals (11,51), and the higher molecular weight bFGF proteins have now been reported to be selectively localized to the nucleus (55,57). Despite the importance of these high molecular weight isoforrns acting as nuclear targeting signals (55), and the finding that expression of elevated levels of high molecular weight bFGF induced transformation and tumorigenicity (5,56),there have been only a limited number of reports on the detection of higher molecular mass bFGF isoforms in human brain tumors (34,39,47).

bFGF Receptors

Five distinct receptors for the FGF's, the result of alternate splicing events, have been identified that originate from 4 separate genes (38) (Table

2). The basic structure of the transmembrane receptors includes: (a) three immunoglobulin-like domains; (b) a transmembrane region, and; (c) a tyrosine kinase domain. The speâficity of the receptors for the nine FGF farnily members is variable, with some receptors activated by multiple members, while others are highly speàfic. The switch from FGFR-2 expression to FGFR-1 has been shown to be an important molecular event in the progression of human gliomas (42,74).

Because of the key role of bFGF in a variety of normal and pathological processes, the following experiments were conducted to better understand the role of bFGF in the malignant phenotype of human gliomas. Table 1 Gene Commo~zNme Secreton! Sipnnl Seqrience FGF-1 Acidic FGF No FGF-2 Basic FGF No FGF-3 Yes FGF-4 Yes FGFJ Yes FGF-6 FGF-6 / hst-2 Yes FGF-7 KGF Yes FGF-8 AIGF Yes FGF-9 GAF No -- Table 2 CHAPTER II

Quantitafion of bFGF in Human Brain Tumor Tissues by Enzyme-Linked lmmunosorbent Assay ABSTRACT

Basic FGF is a potent angiogenic mitogen that is linked to the malignant phenotype of solid tumors. Until recently, the amount of bFGF antigen has been determined semi-quantitatively by immunohistochernistry, with no quantitative reports of bFGF in surgical samples of human central nervous system neoplasms. Because the precise quantitation of bFGF could be useful as a prognostic indicator in human brain tumors, we have used a rapid, reproducible and sensitive enzyme-linked immunosorbent assay (ELISA) to quantitate bFGF in 62 samples of neoplastic and control (non- neoplastic) brain tissue samples. The tumor-associated bFGF was significantly elevated in meningiomas (18.5 + 2.8 ng/mg total protein, n=ll, P <0.05), acoustic schwannomas (14.0 I 0.9 ng/mg, n=9, P < 0.05), and gliomas (WHO grades 1-IV) (9.1 f 1.9 ng/mg, n=18, P c 0.05) in comparison to control brain tissue (3.4 f 0.3 ng/mg, n=9). Classification of gliomas into lower grade tumors WHO grades 1-III, (15.2 f 3.1 ng/mg n=4, P < 0.05), and glioblastoma, WHO grade IV, (7.4 k 2.1 ng/mg n=14, P c 0.05) revealed significant elevations of bFGF. Recurrent high grade gliomas (3 WHO grade IV, one WHO grade III) (14.0 t 6.4 ng/mg n=4, P < 0.05) demonstrated significant elevations of bFGF protein in comparison to non-neoplastic brain. Pituitary adenornas (1.8 Ç 0.7 ng/mg, n=3) and metastatic carcinomas to the brain (2.3 k 0.7 ng/mg, n=12) showed no elevations of bFGF. Preliminary examination of total ribonucleic acid revealed elevated levels of bFGF mRNA transcripts in a lower grade glioma, WHO grade II, glioblastoma, WHO grade IV in comparison to non- neoplastic human brain. The direct measurement of bFGF în human brain tumors codd enable clinical correlative studies of bFGF with cellular proliferation, irtvasiveness, angiogenesis, and could be a useful aiterion for anti-bFGF or angiosuppressive therapies.

INTRODUCTION

Basic FGF is a potent angiogenic peptide (7) that is linked to invasion, cellular proliferation and angiogenesis in human brain tumors (5,10,11,15,21,28). Basic FGF has been localized in human brain tumors semiquantitatively by immunohistochemistry (4,21,26,28,32), and nor thern blot (29). Recently biologically active bFGF protein was detected in the cerebrospinal fluid of 62% of children with brain tumors, but not in controls (14). The amount of bFGF was found to positively correlate with the intensity of angiogenesis in these tumors (14). The development of a specific bFGF ELISA (31) has enabled the quantitation of bFGF antigen in serum (9),urine (19,20,27), and cerebrospinal fluid (CSF) (14). Tumor tissues are known to contain numerous growth factors, however, there have been no quantitative studies of bFGF in surgical samples of human brain tumors. Because of the potential importance of bFGF in the biology of adult human brain tumors, and to idenbfy potential subpopulations of patients that may be responsive to anti-growth factor strategies (11,12,30),we have quantitated bFGF antigen in 53 surgical samples of adult human brain tumors and in 9 non-neoplastic control tissue samples.

MATERIALS AND METHODS

ELISA Detection of bFGF

To determine the sensitivity of the bFGF ELISA (Amersham Corporation, Arlîngton Heights, IL), two sources of recombinant human bFGF were serially diluted in the range of 10 pg/d to 640 pg/ml in the tissue extraction buffer (see below), and a standard curve generated for both (Figure 7 1-

Tissue Preparation

Surgical sampIes of the human brain turnors were either snap frozen in liquid nitrogen immediately following surgical removal, or embedded in O.C.T. (Miles Laboratories) compound, then frozen in liquid nitrogen. Al1 samples were stored at -80°C until used. FifSr milligrarns of the frozen tissue was then homogenizecl in chilled extraction buffer, containing 50 mM Tris pH 8.0, 150 mM NaCl, 2% Nonidet P-40 (Sigma Chernical Co., St. Louis, MO), and the protease inhibitors EDTA-Na2, pefabloc, pepstatin and leupeptin (Boehringer Mannheim Corp., Indianapolis, IN). The homogenates were centrifuged at 16,000 x g for 20 minutes at 4OC, to remove insoluble cellular debris, and the supernatants aliquotted and stored at -80°C.

Basic FGF antigen was quantitated from tissue fragments of acoustic schwannomas (n=9), meningiomas (n=ll), WHO grade 1-IV gliomas (n=18), metastatic carcinomas to the brain (n=12), pituitary adenornas (n=3), and non- neoplastic human brain (n=9) utilizing a highly sensitive solid phase ELISA capable of detecting bFGF protein as low as 1.0 pg/d (Amersham Corporation, Arlington Heights, IL). Samples were diluted in the extraction buffer, and determinations were made in duplicate wells on 96-well microtiter plates, and bFGF antigen concentration was calculated as nanograms (ng) of antigen per milligram of total cellular protein using a modified Lowry assay (Bio-Rad, Richmond, CA). Western Blot Analysis

Equal protein amounts of the tumor lysates were loaded on a precast 14% polyaqlamide gel (NOVEX, San Diego, CA), electrophoresed ai 125 constant volts, then electrotransferred to a 0.45 pM nitrocellulose membrane (NOVEX). The membrane was incubated with the DE06 anti-human bFGF monoclonal antibody (26) at a 1:500 dilution, followed by an alkaline phosp ha tase-conjugated goat anti-mouse secondary antibody (Kirkeguaard and Perry Laboratories, Gaithersberg, MD). Basic FGF was visualized with the chromogenic substrate Cnitroblue tetrazolium chloride/5 bromo-4 chloro- indoyl-phosphate (Kirkeguaard and Perry Laboratories). Negative controls for immunoreactivity included: (i) the omission of the primary antibody, to control for the non-specificity of the secondary antibody; and (ii) substitution of the primary antibody with purified mouse IgG (Organon Teknika Corp., West Chester, PA).

Northern Blot Analysis

Total RNA was isolated f-rorn fragments of non-neoplastic human brain, lower grade glioma, glioblastoma multiforme, and meningioma, the RNA pellet resuspended in RNAse free water and stored ai -80°Cuntil used in northern blotting as described previously (23). The RNA was quantitated spectrophotometricallyby its absorbance at 260 nm (1 OD = 40 pg/ml). Fifteen pg of total RNA was denatured by heating to 65°C in 2.2 M formaldehyde and 50% formamide in sample buffer containing 25 ng/@ ethidium bromide. The samples were electrophoretically fractionated on a 1.2% formaldehyde- agarose gel in 0.02M morpholinopropane sulfonate (MOPS) buffer, pH 7.0. Following electrophoresis, gels were photographed under UV light, rinsed in water, and the RNA was transferred over~ghtfrom the gel to a nylon membrane (Boehringer Mannheim Corp., Indianapolis, IN) by capillary transfer using 10X SSC (lx SSC is 0.015 M sodium utrate/0.15 M sodium chloride). The RNA was fixed to the filter by UV crosslinking.

Hybridization to 32 P-la beled probes

Prehybridization was in 30% formamide, and standard buffer (23) for 5 hours at 42OC. A 1.5 kB EcoRl bovine brain bFGF cDNA fragment from the pbFGF plasmid (22) was labeled by randorn priming (Gibco-BRL) incorporating 32P-a-labeled dCTP (Amersham Laboratories, Arlington Heights, IL). The 32P-a-labeled probe was denatured and added to the prehybridization solution, and hybridization to the membrane was performed for 24 hours at 42OC. Nonspecific binding was removed by washing in 0.5X SSC, 0.1% SDS (w/v) at 60°C. Visualization and quantification of the signals were obtained utilizing a scanning densitometer. To ensure equal RNA loading, ethidium bromide stained ribosornal RNA bands, were photographed under UV light. In addition, the membranes were probed secondarily with the "control" cDNA species, glyceraldehyde 3- phosphate dehydrogenase (GAPDH).

Statistical Analysis

Results are expressed as mean f standard error. Statistically significant differences between means were determined using the Statview SE statistical program (Abacus Concepts Inc.) utilizing a one-way analysis of variance. When the overall F-ratio was significant, differences between the means were determined using a Fisher posthoc test with P c 0.05 considered statistically signifiant. RESULTS

As demonstrated in Figure 2, the ELISA was capable of detecting two sources of recombinant human bFGF protein (r2 0.93, and 0.97 respectively). These data demonstrated the bFGF ELISA was specifically detecting bFGF protein. For subsequent analyses, the recombinant bFGF from Amersham Corporation was used for the generation of standard curves for estimating bFGF protein amounts. There were no significant differences in the amount of bFGF antigen between the snap frozen and O.C.T. embedded surgical specimens (P<0.147), values were therefore pooled for subsequent analysis (Figure 2). Basic FGF protein was significantly elevated in meningiomas (18.5 f 2.81 ng/mg protein, P c 0.05), acoustic schwannomas (14.0 t 0.9 ng/rng, P < 0.05), and WHO grade 1-IV gliomas (9.1 t 1.9 ng/mg, P < 0.05) in comparison to non-neoplastic control tissue (3.5 t 0.3 ng/rng) (Figure 3). The gliomas were subdivided for further analysis; lower grade gliomas (15.2 k 3.1 ng/mg, P < 0.05) (WHO grades 1-III), and glioblastomas (7.4 f 2.1 ng/rng P < 0.05) (WHO grade IV) demonstrated significant elevations of bFGF (Figure 4). There were 4 recurrent gliomas (3 glioblastoma and 1 rnalignant oligodendroglioma); bFGF levels were elevated in these tumors (14.0 + 6.4 ng/mg, P < 0.05) in comparison to the non-neoplastic human brain and glioblastoma (Figure 5). Metastatic carcinomas to the brain (2.3 t 0.7 ng/mg), and pituitary adenomas (1.8 t 0.7 ng/mg) did not differ significantly hmthe non-neoplastic control (Figure 3).

Western blot analysis of the tumor cell lysates revealed a single faint immunoreactive band at approximately 14-16 kD (not shown). Northern blot analysis of the human tumor tissues revealed three bFGF mRNA transcripts of approxîmateIy 7.0,4.0, and L.2 kB. The amount of the 7.0 and 4.0 kB bFGF mRNA was significantly elevated in the lower grade glioma in cornparison to the non-neoplastic brain, meningioma and glioblastorna (Figure 6). The levels of the 7.0 and 4.0 kB bFGF mRNA transcripts were barely detectable in the meningioma sample, but a faint band was discemible at 1.5 kB (Figure 6).

Angiogenesis is necessary for the continued growth of solid tumors (8). Basic FGF is a potent angiogenic mitogen reported to be upregulated in brain tumors but there have been no quantitative reports on the amount of tumor- associated bFCF in these neoplasms. We now report the quantitation of bFGF in a variety of human brain tumor tissues using a sensitive ELISA. Two sources of recombinant human bFGF wese serially diluted and added to the bFGF ELISA. The standard curves obtained not only reveal a significant correlation between the absorbance and the amount of bFGF antigen added (Figure 1))they also indicate that the assay is sensitive to small amounts of bFGF protein. Because bFGF is a member of a larger family of structurally homologous growth factors, it is conceivable that the assay could detect one or more of these other bFGF-related factors. Because the other FGF-like proteins were not available, this was not tested in this dissertation. Previous experiments however, have demonstrated the assay to specifically detect bFGF, with no cross-reactivity for other FGF-like proteins (Dr. Peter Riis, Persona1 communication). FIGURE 1. Two sources of recombinant human bFGF were serially diluted in the tumor extraction buffer, and added to the microtitre wells. Both sources of bFGF were detectable in amounts as low as 10 pg/ml. The correlation between absorbance and bFGF antigen amounts was also significant; Arnersham bFGF, ?= 0.943, and DuPont-Merck, ?= 0.979.

FIGURE 2. Comparison of bFGF antigen levels extracted from snap frozen or O.C.T. embedded tumor tissue sarnples. No significant difference in the arnount of bFGF was found. Fresh Frozen O.C.T. Embedded

Meningioma Acoustic Low Grade GBM Metastasis FIGURE3. The amount of bFGF antigen was significantly elevated in meningioma, acoustic schwannoma, and gliomas, in cornparison to non- neoplastic human brain (*, P < 0.05). Metastatic carcinomas and pituitary adenomas showed no significant elevations of bFGF protein. Gliomas Non-neoplastic Metastsns Pituitary Schwannoma Brain Adenoma FIGURE4 Within the gliomas, lower grade tumors (WHO grades 1-III) showed sieficant elevations of bFGF antigen in comparison to non- neoplastic brain (*, P < 0.05), and glioblastoma (WHO grade Iv), (**, P < 0.05). Glioblastoma also demonstrated significant elevations of bFGF protein (*, P < 0.05) in comparison to non-neoplastic brain.

FIGURE 5 There were four recurrent high grade tumors (3 WHO grade IV, and one WHO grade III). The recurrent gliomas demonstrated a significant elevation of bFGF protein in cornparison to non-neoplastic brain (*, P < 0.05) and glioblastoma (**, P < 0.05). 1 Non-neoplastic Low Grade GBM Recurrent Brain Gliomas Gliomas FIGURE 6 Quantitation of bFGF mRNA transcripts reveals elevated Ievels of the 7.0 and 4.0 kB hanscripts in low grade gliomas. GBM - glioblastoma multiforme. Normal Low Grale GBM Meningioma

Normal L O w GBM Meningioma Grade Basic FGF antigen was detected in all of the tumors and non-neoplastic brain exarnined in this report. No significant difference in the amount of bFGF antigen was found between the snap frozen and O.C.T. embedded tumor samples (Figzm 2). This suggestç that archived frozen brain tumor tissue samples could be utilized for the routine measurement of tumor- associated bFGF and/or other growth factors that could complement routine histopathology. The advantages of usîng tumor tissues for the quantitation of bFGF protein in adult patients indude: (a) the amount of bFGF antigen is elevated in brain tumor samples, suggesting that bFGF levels could be obtained from small biopsy specimens; (b) the amount of tissue obtained at surgery, typically exceeds what is required for ELISA measurement, so tissue can be stored for repeat analysis, or used in subsequent analysis and (c) the use of tissue avoids some of the eeneral hazards of obtaining cerebrospinal fluid by lumbar puncture (2), as well as the specific effect of the danger of intracranial hemiation in patients with rnass effect.

The amount of bFGF protein was detected at elevated levels in meningiomas, acoustic schwannomas, consistent with the reported elevation of bFGF messager RNA in these turnors (17,29). Angiogenesis is necessary for the development of benign and malignant tumors; the high level of bFGF protein in the benign, yet vascular meningioma and acoustic schwannoma indicates that bFGF protein could be a mediator of the intense neovascularization associated with these tumors. In addition, these data suggest that elevated levels of bFGF alone may not be sufficient for the development of a malignant phenotype.

The gliomas demonstrated elevated levels of bFGF in cornparison to non-neoplastic brain. The significantly elevated levels of bFGF protein in lower grade gliomas (WHO grades 1-Ill) (Figure 5) suggests that elevated bFGF expression is an early event in the multi-step development of central nervous system gliornas. The preliminary finding of a significant increase in bFGF protein in recurrent high grade gliomas (3 grade IV, 1 grade m) could suggest recurrent lesions are dependent upon bFGF, and would be consistent with the recently reported finding of undeteciable levels of bFGF in children diagnosed with medulloblastoma, that became detectable when the tumor recurred (14). The lowered amounts of bFGF antigen in gliomas in comparison to the meningiomas and acoustic schwamomas has not previously been reported. One potential mechanism that could explain this finduig is the extracellular export of bFGF by the gliomas. The switch to the extracellular export of bFGF was correlated with tumorigeneicty, angiogenesis and metastasis in fibrosarcoma (13), and increased cellular invasion in the brain (10). The exported bFGF could be bound to a high affinity ce11 surface receptors, known to be upregulated in malignant gliornas (16), and may not be released in our extraction protocol.

The amount of bFGF antigen in metastatic tumors to the brain did not differ significantly from the non-neoplastic control brain (Figure 4). The lowered levels of bFGF protein suggests that the metastatic tumors express other angiogenic factors e.g. vascular endothelid cell growth factor, that may be important in angiogenesis associated with these tumors. Alternatively, the lowered levels of bFGF in the metastatic tumors could be a function of the organ microenvironment; for example, the expression of bFGF rnessenger RNA and protein in human renal ceU carcinoma cells was enhanced when these cells were inoculated in the kidney, but bFGF levels deaeased significantly when these cells were injected at sites other than the kidney (25).

Analysis of the cellular lysates by western blot revealed only a single weak immunoreactive band at approximately 1416 kD for both the normal and tumor tissue homogenates. Prototypical bFGF migrates as an 18 kD protein, however in the presence of proteinases, bFGF is degraded to a 1416 kD protein. Because of the delay following surgical it is likely that tumor- associated proteinases are degrading the bFGF protein. This makes analysis of bFGF isoforms larger than 16 kD difficult in tissue samples. Because the ELISA recognizes all molecular weight bFGF isoforms the amount of bFGF protein can still be quantitated, even in the presence of proteinases. Western analysis of bFGF protein in tissue sections has been reportedly difficult because the amount of bFGF antigen is typically below detectable levels, and sam~lesoften reauire ~urificationand concentration of bFGF protein using heparin-sepharose duomatography. The application of this protocol, would be difficult with over 62 tissue samples, and the final result would not permit the quantitative comparison between tumor types.

One problem with the ELISA and the western blot analysis is that they camot distinguish whether bFGF is expressed in tumor or non-tumor tissues. We have exarnined a series of tumor tissue sections for bFGF localization by immunocytochemistry (4) (see Appendix 1). Basic FGF was detectable in 97% of malignant brain tumors, confirming that one source of bFGF is from the tumor cell cornpartment. In addition, bFGF was prominently deteded in the endothelial ceIl compartment, suggesting levels of bFGF detected in tumor tissue section by ELISA, reflect bFGF from endothelid ceils. To more definitively quantitate bFGF levels in human gliomas, bFGF levels should be measured in vitro from human glioma cell lines.

The prelhinary finding of elevated levels of bFGF mRNA in low grade gliomas and glioblastoma is consistent with the increased amount of bFGF protein reported in this study, and suggests increased bFGF gene transcription may be responsible, in part, for the elevated bFGF protein detected in these tumors. The amount of bFGF protein was detected at elevated levels in meningiomas; the lowered levels of the 7.0 and 4.0 kB bFGF mRNA transcripts in the meningioma specimen, in cornparison to the glioma and non-neoplastic brain specimens, could be the result of an increase in bFGF mRNA stability (17). Alternatively, the decreased levels of the bFGF mRNA transcripts could be the result of the increased ce11 density in the tumor; bFGF mRNA levels are reported to be elevated in cells sparsely plated, but rapidly declined when the cells reached confluence (18).

The direct measurement of bFGF antigen in brain tumor tissues at the time of surgery may complement histopathology, and provide a molecular phenotype of the tumor. The finding that O.C.T. embedded tissue samples provide measurable bFGF levels confirms that archived tissue samples cm be used retrospectively in clinical correlative studies of bFGF with microvascular and cellular proliferation, invasiveness, patient survival, and may be useful as a criterion for anti-bFGF, or angiosuppressive therapy . Chapter II - Limitations The bFGF ELISA. Basic FGF has been immunolocalized and quantitated semiquantitatively in human brain tumors (4,21,26,28,32). To conhand extend these obse~ations,a comrnercially available ELISA for bFGF was used to quantitate bFGF protein from a variety of brain turnor tissues. In Fig. 1 the sensitivity of the ELISA for increasing concentrations of recombinant human bFGF protein from two different commercial sources was tested, however, the ELISA was never tested for selectivity. To dernonstrate selectivity the ELISA would have to be tested with the other eight FGF protein farnily members to confirm the assay was specifically rneasuring bFGF protein. In a persona1 communication, Dr. Peter Riis at Amersham Corporation provided data that demonstrated the antibodies used in this ELISA did not recognize the other FGF family members. Despite the persona1 communication, the inclusion of selectivity control experimen ts would have strengthened the data obtained using this bFGF ELISA. Another way to confirm the results obtained by this bFGF ELISA would have been to confirm the results with a second bFGF ELISA. In addition, the use of westem blots to determine if similar levels of bFGF protein are recognized would be useful. Although a western blot was performed on a small number of hurnan brain tumor samples, the monoclonal bFGF antibody was different than that used in the bFGF ELISA. The results in this chapter would have been strengthened by performing quantitative westem blots with the monoclonal antibody used in the bFGF ELSA, to corifhm the antibody was indeed recognizing bFGF protein speaficdy, and quantitativeIy. There have been numerous studies on the immunolocalization of bFGF protein in human gliomas. In these reports, the amount of bFGF protein was subjectively quantitated based upon the intensity of the immunochemical reactivity. In Chapter II we demonstrated the validity of a bFGF ELISA, and its usefulness in quantitating objective differences in bFGF protein levels between histologically distinct himors. In Chapter II we report a significant elevation of bFGF protein in low and high grade gliomas in cornparison to the non-neoplastic brain tissue. Because we have previously demonstrated that bFGF is immunologically detectable in the endothelial cell, and vesse1 wall (See Appendix l),it is not clear whether differences in bFGF protein levels are due to variations in the tumor or endothelial ce11 compartments. To confirm and extend these findings, we have examined bFGF gene and protein expression in established human glioma ce11 lines, and non-neoplastic human astrocytes in vitro. In Chapter III the results of these experiments are demonstrated. CHAPTER III

Basic Fibroblast Growth Factor Gene and Protein Expression in Human Glioma Celk Malignant gliomas are characterized by heightened cellular proliferation, local invasion and angiogenesis, processes known to be stimulated by basic fibroblast growth factor @FGF). Because of the important role of bFGF in the biology of human gliomas we examined bFGF gene and protein expression in four established human glioma cell lines, U-373,U-118, A-172 and U-87, and non-transformed human fetal astrocytes. Quantitation of cell-associated bFGF protein by ELISA, indicated that al1 cells expressed detectable bFGF protein. Basic FGF antigen was detected in the serum-free conditioned medium of the glioma cells, not the non-transformed human astrocytes, suggesting a switch to the extracellular export of bFGF in malignancy. Southern blot analysis of genomic DNA from the four human glioma ce11 lines confirmed no amplification or rearrangernent of the bFGF gene. Northern blot analysis demonstrated two bFGF messenger RNA (mRNA) trançcripts of approximately 7.0 and 4.0 kB in the human glioma and fetal astrocyte cells. Cyclohexirnide caused a rapid and significant elevation of the bFGF mRNA transcripts in the human glioma cells, but a decrease in these levels in the fetal human astrocytes. Western blots of the glioma ce11 lysates revealed four bFGF protein isoforms of approximately 18, 21, 22 and 24 kD. As a control, two non-transformed bovine vascular endothelial, and human fetal astrocyte cells were examined and found to express only the 18 kD bFGF isoform. These findings demonstrate a differential regulation of bFGF gene and protein expression between non- transformed and malignant human glial celis, and confinn the important and complex role of bFGF in the malignant phenotype of human gliomas. INTRODUCTION

Basic fibroblast growth factor (bFGF) is a key positive regulator of angiogenesis (16,42), that is linked to invasion, cellular proliferation, and angiogenesis in human gliomas (12/23,34). Recently, the amount of bFGF antigen was quantitated in human brain tumor tissues using a highly sensitive enzyme-linked immunosorbent assay, and the arnount of bFGF protein was found to be significantly elevated in gliomas (WHO grade 1-IV) in cornparison to non-neoplastic brain (Gately et RI, unpublished data). This study was undertaken to further investigate bFGF gene and pro tein expression in vitro in four established human gliorna ce11 lines, and in non- transformed human fetal astrocytes.

MATERIALS AND METHODS

Cell Culture

Established human glioma cells U-373, U-118, A-172 and U-87, (American Type Culture Collection (ATCC),Rockville, MD) were cultured in Dulbecco's Modified Eagles Medium (Dh4EM) supplemented with 10% calf serum, 100 IU/ml penicillin and 100 pg/d streptomycin (Gibco BRL, Grand Island, New York). Primary human fetal astrocytes W585, W586 and W766 (51), bovine pulmonary artery endothelial cells (BEC,ATCC), and fetal aortic endothelial cells AG 7680 (AGO,Institute for Medical Research, Camden, NJ) were grown in DMEM supplemented with 10% fetal bovine serum, 100 N/ml penicillin and 100 pg/ml streptomyûn. Ce& were maintained in a humidified incubator at 37OC in an atmosphere of 5% COand 95% air, with the medium replaced twice per week. Chernoinvasion Assay

The chernoinvasion assay method was similar to that reported previously (1). Cell culture inserts containing an 8 Pm pore membrane, pre- coated with Matrigel, 100 pg/cm2 (Collaborative Biomedical, Inc.), were placed in a humidified incubator at 37OC for 60 minutes to polymerize the Matrigel. The human glioma cells resuspended in serum free DMEM supplemented with 1.0% bovine serurn albumin (BSA) (Sigma Chernical Co., St. Louis, MO), were then carefully added to the upper chamber at a concentration of 4.0 x 104cells/well. Self-conditioned serum-free DMEM (24 hours) from each ce11 line was used as a chemoattractant in the lower chamber. Cells were allowed to invade for 24 hours, fixed in chilled 100% methanol, then Matrigel and cells remaining in the upper chamber were mechanically removed with a cotton swab. The filters were then stained with hematoxylin and mounted on glass slides. Cellular invasion was then quantitated by computerized image analysis (Quantimet 570, Leica, Deerfield, IL); the number of cells that invaded the Matrigel and migrated to the lower surface of the membrane were counted from triplicate wells.

Quantitation of bFGF by ELISA

The human glioma and fetal astrocyte cells were plated in duplicate at 1.0 x105 cells/6O mm plate, and allowed to attach ovemight. The following day (day O) cells were fed with fresh medium. Cells isolated for bFGF quantitation on day 2 (subconfluence), day 4 (confluence) and day 6 (post confluence). One plate of ceus were tsrpsinized, and the number of cds were counted with a hemacytometer. The cells fiom the second plate were lysed in 1 ml of chilled extraction buffer, containing 50 mM Tris pH 8.0,150 rnM NaCI, 2% Nonidet P-40, and the protease inhibitors EDTA-Na2, pefabloc, pepstatin and leupeptin (Boehringer Mannheim Corp.). The lysates were then centnfuged at 16,000 x g for 20 minutes at 4OC to remove insoluble cellular debris, and the supernatants aliquotted and stored at -80°C. The amount of bFGF was determined utilizing a solid phase bFGF ELISA capable of detecting bFGF in amounts as low as 1.0 pg/d (Amersham, Arlington Heights, L). Determinations were made in duplicate wells on 96-well microtiter plates, and bFGF antigen concentrations calculated as nanograms of antigen per 1.0 x 106 cells.

Southern Blot Analysis

Confluent monolayers of the glioma cells were harvested by trypsinization, and genornic DNA prepared using the PureGene kit (Gentra Systems, Inc., Research Triangle Park, NC). The DNA was quantitated spectrophotometrically by its absorbance at 260 nm (1 0.D.260= 50 pg/ml), then l5pg of genomic DNA was digested with EcoRl and Pst1 for two hours at 37"C, then samples were size fractionated by electrophoresis on a 0.8% agarose gel containing ethidium bromide. The samples were then transferred to a nylon filter (Boehringer Mannheim Corp., Indianapolis, IN) by capillary transfer using 10X SSC (lx SSC is 0.015 M sodium citrate/O.l5 M sodium chloride). The DNA was fixed to the filter by UV crosslinking. Southern hybridization was performed as described below for northem hybridization.

Northern Blot Analysis

Total cellular RNA was isolated from cultured cells by acid guanidinium isothiocyanate-phenoldoro form extraction (9). The RNA pellet was resuspended in RNase free water and stored at -80°C until used in northem blotting as described previously (44). The RNA was quantitated spectrophotometrically by its absorbance ai 260 nm (1 0.D.260 = 40 pg/ml), then 15pg of total RNA waç denatured by heating to 65OC in 2.2 M formaldehyde and 50% formamide in sample buffer containhg 25 ng/p1 ethidium bromide. The samples were then size fractionated by electrophoresis on a 1.2% formaldehyde-agarose gel in 1X morpholinopropane sulfonate (MOPS) buffer. Following electrophoresis, gels were photographed under W light, rinsed in water, and the RNA was transferred overnight from the gel to a nylon membrane (Boehringer Mannheim Corp.) by capillary transfer using 10X SSC (lx SSC is 0.015 M sodium citrate/O.lS M sodium chloride). The RNA was fixed to the filter by UV crosslinking.

Hybridization to 32P-labeled Probes

Prehybridization was in 30% formamide, and standard buffer (44) for 5 hours at 4Z°C. For bFGF rnRNA identification, a 1.5 kB EcoRl bovine brain bFGF cDNA fragment from the pbFGF plasmid (43) was utilized. For detection of the bFGF receptor-2 (bFGFR-2) mRNA, a bFGFR-2 cDNA isolated from non-transformed human fetal astrocytes (17) was blunt-end ligated into a plasmid Bluescript vector (Stratagene) and used for probing. The cDNAs were labeled by random priming (Gibco-BRL, Grand Island, NY) incorporating 32P-a-labeled dCTP (Amersham Laboratories, Arlington Heights, IL). The 32P-labeled probe was denatured and added to the prehybridization solution, and hybridization to the membrane was performed for 24 hours at U0C. Nonspecific binding was removed by washing in 0.5X SSC, 0.1% SDÇ (w/v) ai 60°C. Visualization and quantification of the signals were obtauied utilizing a scanning densitometer. To ensure equal RNA loading, the ribosomal RNA bands, stained with ethidium bromide, were photographed under UV light. In addition, the membranes were probed secondarily with the "control" cDNA species, human p-actin.

Western Blot Analysis

Equal protein amounts of the cellular lysates of the human glioma, fetai astrocyte and bovine endothelial cells were loaded on a precast 14% polyacrylamide gel (NOVEX, San Diego, CA), electrophoresed at 125 constant volts, then electrotransferred to a 0.45 pM nitrocellulose membrane (NOVEX). The membrane was incubated with the DE4 anti-human bFGF monoclonal antibody (40) at a 1:500 dilution, followed by an alkaline phosphatase-conjugated goat anti-mouse secondary antibody (Kirkeguaard and Perry Laboratories, Gaithersberg, MD). Basic FGF was visualized with the chromogenic substrate 4-nitroblue tetrazolium chloride/5 bromo-4 chloro- indoyl-phosp hate (Kirkeguaard and Perry Laboratories). Negative controls for immunoreactivity included: (i) the omission of the primary antibody, to control for the non-specificity of the secondary antibody; and (ii) substitution of the primary antibody with purified mouse IgG (Organon Teknika Corp., West Chester, PA).

Statistical Anaiysis

Results are expressed as mean k standard deviation. Statistically signiCicant difierences were determined by one-way analysis of variance. Differences with p < 0.05 were considered statistically significant.

RESULTS

The four glioma cell lines were heterogeneous with respect to cellular invasiveness, The mean number of invaded U-373, 695 + 155, and U-118,430 k 220, human glioma cells was signihcantly higher than A-17525 + 20, and U- 87,6 t 7, (P < 0.04) (Figure 1).

Quantitation of bFGF &y ELISA

Quantitation of cell-associated bFGF in the non-neoplastic fetal astrocytes, and human glioma cells revealed detectable amounts of bFGF antigen in al1 ce11 lines (Figure 2). The amount of bFGF protein, except for the A472 ce11 line, did not differ significantly as cells grew to confluence (Figure 2). Basic FGF was detected in the senim-free conditioned medium of confluent human glioma cells (Figure 3). By cornparison, non-transformed human feetal astrocytes released barely detectable levels of bFGF antigen (Figttre 3).

Southern Biot Anaiysls

Equal amounts of genomic DNA isolated from the four human glioma ce11 lines, was cut with EcoRl(1anes 1-4) and Pst 1, (lanes 5-8) then probed with a bovine bFGF cDNA. (Figure 4) The hybridizing genomic fragments correlated with the restriction map of the bFGF gene suggesting no rearrangement. The intensity of the hybridizing material was equal between samples consistent with no amplification of the bFGF gene. (Figure 4).

Northern BIot Analysis

Two prominent bFGF mRNA transcripts of 7.0 and 4.0 kB were detected in the established human glioma cell Lines and fetal astrocytes. Only FIGURE 1. - The mean number of cells that invaded the Matrigel was significantly higher for the U-373 and U-118 human glioma cells (* P c 0.04) in cornparison to A472 and U-87. Number of Cells Invaded FIGURE 2- The amount of ce11 associated bFGF protein in human glioma and non-neoplastic human astrocytes at subconfluence (Day 2), confluence (Day 4), and post-confluence (Day 6).

FIGURE 3- Basic FGF protein was detected in the semm-free conditioned medium of human glioma ce11 lines. In contrast bFGF was barely detected in the non-transformed W585 human fetal astrocytes conditioned medium.

the A-172 glioma ce11 line demonstrated a signihcant elevation of the 4.0 kB bFGF mRNA transaipt in comparison to all other cells (Figure 5A,B) The addition of cycloheximide caused a rapid and significant elevation of these transcripts following two hours exposure (Figure 6). The U-118 glioma ce11 line showed maximal elevation of both the 7.0 (465 % of control), and 4.0 (327%) kB mRNA transcripts, in comparison to the U-87, (220% and 251%) and U-373, (153% and 157%) glioma ce11 lines. By contrast, cycloheximide suppressed the two bFGF mRNA transcripts in the W585 human fetal astrocyte cell strain (Figure 6). The 7.0 kB transcript was inhibited to 47% of control, and the 4.0 kB transcript to 62% of control.

bFGF Receptor-2 mRNA

Northern blots revealed an mRNA transcript that hybridized to the bFGFR-2 cDNA probe in the human glioma and fetal astrocytes cells (Figure 7). Quantitation of this banscript demonstrated the heterogeneous expression of receptor mRNA that was highest in the non-transformed human fetal astrocyte, W585, and lowest in the U-373 human glioma ce11 line.

Western Blot Analysis

Cellular extracts of the human glioma ceIl lines revealed three prominent immunoreactive bFGF bands at approximately 18, 22, and 24 kD (Figure 8). Examination of the two non-transformed fetal astrocyte ce11 strains, and bovine endothelial ce11 lines demonstrated the prototypical 18 kD bFGF isoform only (Figure 8). The saline and mouse IgG controls were negative confirming the specificity of the antibody-antigen interaction. FIGURE 4- Southern blot of genomic DNA for bFGF. The fragments correspond to the previous restriction digest reports of the human bFGF gene. Lanes 1 and 5, U-373;Lanes 2 and 6,U-118; Lanes 3 and 7, A-172; and Lanes 4 and 8, U-87 ceIl lines. Eco RI Pst 1

1 2 3 4 5 6 7 8 FIGURE 5- Northem blot of bFGF mRNA in confluent human glioma cells. After standardizing for loading, the amount of 4.0 kB bFGF mRNA transcript was found to be significantly elevated in the A472 human glioma cell line (* P < 0.05).

FIGURE 6- Cycloheximide caused a rapid and significant elevation of the 7.0 and 4.0 kB bFGF mRNA transcripts. By contrast, it caused a decrease in these transcripts in the non-transformed human fetal astrocytes, W585. UntRated - - - Conti01 Level FIGURE 7- The amount of the bFGFR-2 mRNA transcripts were significantly higher in the W585 human fetal astrocytes (* P < 0.05) in cornparison to the human glioma ce11 lines.

FIGURE 8- Western blot detection of bFGF in the cellular lysates of human glioma cells, and control non-tranzformed human astrocytes W585, W586, and bovine endothelial cells AG07680 and GM7373.

DISCUSSION

Basic FGF has been identified as a key factor in the biology of malignant gliomas (13,19,20,24,26). The present study was designed to hrther examine bFGF gene and protein expression in human glioma ce11 lines. The findings presented are important because: (i) they dernonstrate the quantitation of bFGF protein by ELISA fiom human glioma cells and non-transformed human astrocytes; (ii) report the extracellular export of bFGF by the glioma cells, not the non-transformed astrocytes; (iii) demonstrate no alteration of the bFGF gene in the human glioma cells; (iv) the cycloheximide data demonstrates differences in the regdation of bFGF transcription between non-transformed fetal astrocytes and human gliomas; and (v) the presence of high molecular weight bFGF in malignant glioma cells, and the absence of these proteins in non-transforrned fetal astrocvtes.

This is the first report on the quantitation of glioma-associated bFGF using a sensitive, and specific bFGF ELISA. Previously, investigators utilized slot-blot hybridization (26),and the growth stimulatory properties of heparin- affinity purified cell extracts on quiescent 3T3 fibroblasts to estimate the amount of bFGF antigen (26). Basic FGF ELISA have been utilized as non- invasive diagnostic tools for bladder and other cancers (32,33). In this study, quantitation of total bFGF in the non-transformed fetal astrocytes and human glioma ceU lines supports the finding that both normal and tumor cells produce bFGF (28). Because the level of bFGF in the non-transformed fetal astrocytes was comparable to the gliomas, expression of specific bFGF isoforms may be more important than the overall level of bFGF in determinhg the malignant p henotype of cells. Alternatively, differences in the extracellular export of bFGF may be involved in the malignant phenotype of gliomas. In this study, examination of senun-free conditioned medium of the four established human gliomas demonstrated detectable levels of bFGF. By contrast bFGF was undetectable in the conditioned medium of one of the non-transformed human fetal astrocytes (Figure 3). This data would be strengthened by the analysis of the conditioned medium of the W586 and W766 ce11 strains, however this was not performed in this study. The lack of detectable bFGF in the conditioned medium of non-transformed human astrocytes however is consistent with previous studies that demonshated bFGF is a cell-associated protein; the switch to the extracellular export of bFGF was associated with angiogenesis and tumorigenicity of fibrosarcoma cells (21), and cells transfected with bFGF expression vectors modified for secretion of bFGF are reported to be transformed (17,43), tumorigenic (43), and invasive (18). Because native bFGF protein has no signal peptide necessary for directing the extracellular export of proteins, the rnechanisms responsible for the release of bFGF in gliomas is unknown, but could involve novel release mechanisms (15).

Northern analysis demonstrated the human gliomas and non- transformed fetal astrocytes expressed primarily two bFGF mRNA transcripts of approximately 7.0 and 4.0 kB, similar to that reported previously (30,31,45). Cydohewmide, an inhibitor of protein synthesis, caused a rapid and significant increase in bFGF mRNA transcripts in the glioma cells, but decreased mRNA transcripts in the fetal astrocytes. These data suggest that bFGF transcription in the non-transformed fetal ashocytes is dependent upon the function of a labile protein intermediary. The rapid upregulation of bFGF rnRNA in the malignant glioma cell lines is similar to that desmibed in C6 glioma (35), human fibroblasts (47), and bovine endothelial cells (49). The superinduction of bFGF mRNA could reflect an increase in mRNA stability (IO), or could be due to the loss of transcriptional repressors by bFGF specific, or nonspecific protein factors, that would normally shut off transcription (522). The data suggest the non-transformed and malignant ceus may differ in the mechanisms of regdation of bFGF gene expression.

The glioma and non-transfomed human astrocytes expressed an mRNA transcript for the bFGFR-2. Taken together with the detection of extracellular bFGF in the glioma cells, it indicates that bFGF could be an autocrine growth factor for malignant human gliomas. In this study, quantitation of the bFGFR-2 receptor mRNA revealed differential levels of expression between the non-transformed and malignant glial cells. The aciount of the bFGFR-2 transcript was highest in the non-transformed human astrocytes, consistent with the reported loss of bFGFR-2 in the malignant progression of human gliomas (25,50). The amount of bFGFR-2 mRNA has been reported to be in abundance in normal white matter and in low grade gliomas, but was undetectable in malignant gliomas (50). The data presented supports the hypothesis that loss of bFGFR-2 mRNA expression is associated with increasing malignancy in human gliomas.

Western blot analysis of the human glioma and two fetal astrocyte ce11 lines in this study revealed the glioma cells expressed at Ieast three bFGF isoforrns (Figure 8). The complete absence of high molecular weight bFGF proteins in the two non-transformed astrocyte cell strains suggests the high molecular weight bFGF proteins may play a role in the malignant phenotype of human gliomas. Previous studies have identified high molecular weight bFGF proteins to be primarily associated with the nucleus (838,41). Murphy et al showed the selective suppression of 23 and 25 kD bFGF isoforms, utilizing antisense bFGF oligonucleotides, correlated with the inhibition of anchorage-dependent and anchorage-independent glioma ce11 growth (29). In addition, the higher molecular weight bFGF isoforms were show to represent a significantly higher proportion of bFGF protein than the prototypical 18 kD bFGF isoform (29). This finding is supported in the present study; bFGF levels measured by ELISA, reveal higher levels of bFGF protein in the U-373 human glioma cell line in comparison to the W585 ce11 line (Figure 2). By contrast the western Hot, though not quantitative, suggests the W585 ce11 line has more bFGF protein. When taken together these data demonstrate the higher bFGF protein level as measured by ELISA is associated with the expression of the higher molecular mass bFGF isoforms.

Basic FGF protein exists as several isoforms, the predominant form is an 18 kD protein (42), with several high molecular weight bFGF isoforms detected that result fiom alternative CUG translational start sites (14,27,36,37), and/or by alternative bFGF mRNA splicing (53). The relative hction of different molecular weight bFGF proteins is unclear, however, transfection of 3T3 fibroblasts with a bFGF cDNA that codes for the selective expression of high molecular weight bFGF results in transformation of the 3T3 fibroblasts in vitro and tumorigenicity in v iao (4,39). Additional experiments demonstrated that the oncogenic properties of bFGF required the CO- expression of 18 kD and high molecular weight bFGF isoforms (11). Transfection and constitutive expression of the 18 kD bFGF protein in endothelid cells conferred the ability to grow in soft agar (11). Those cds that expressed only the high molecular weight bFGF were immortalized, but the coexpression of the 18 kD and high molecular weight bFGF proteins induced tumorigenicity (11). Therefore it is likely that the different isofoms confer different properties to the target cells.

The findings presented in this study identify bFGF is an important factor in the pathobiology of human gliomas. In this paper we have shown that human glioma cell lines differ from non-transformed hurnan astrocytes in several respects: (i) the extracellular release of bFGF; (ii) cydoheximide superinduction of bFGF rnRNA; (iii) decreased expression of FGFR-2 mRNA; and (iv) expression of higher molecular weight bFGF isofonns. When taken together these data suggest that the differential regulation of bFGF gene and protein expression may be responsible in part for the malignant phenotype of human gliomas. Chapter III - Limitations The experiments included in this chapter suggest differential regdation of bFGF gene and protein expression for human glioma cells in comparison to the non-transformed human fe ta1 astrocytes. An important finding was shown in Fig. 3, where bFGF was detectable in the serum-free conditioned medium of human glioma cells, but not detectable in the non-transformed huma. fetal astrocyte conditioned medium. These data would have been strengthened by measuring the amount of bFGF detectable in the medium from the other fetal astrocyte ce11 strains. Unfortunately, these ce11 strains were not available ai the time these experiments were perfomed. The preluninary findings of secreted bFGF in the transformed gliomas would however, be consistent with the reported export of bFGF in the multi-step development of fibrosarcoma (Kandel, J., et al Ce11 66:1095, 1991), and provides the hpehis to examine the role of secreted bFGF as reported in Chapters IV and V. In Chapter III we demonstrated that the amount of bFGF protein may not be as important as its cellular localization or molecular size. In preliminary studies human gliorna cells were found to release bFGF into the serum-free conditioned medium. By contrast the bFGF in non-neoplastic human astrocytes was cell-associated. This data suggested the switch to the extracellular export of bFGF codd be an important event in the malignant progression of human gliomas. To explore this hypothesis NIH 3T3 fibroblasts were stably hansfected with a bFGF cDNA that would code for a cell-associated bFGF protein, or a modified bFGF cDNA that would code for a secreted form of bFGF protein. In Chapter N we demonstrate that the cells transfected with the bFGF cDNA modified for secretion express higher molecular weight bFGF isoforms, are highly invasive in vitro and in situ in the brain in comparison to the cells expressing increased amounts of ce& associated bFGF. CHAPTER IV

Cells Transfected wifh the Basic Fibroblast Growth Factor Gene Çused to a Signal Sequence are lnvasive In Vitro and In Situ in the Brain ABSTRACT

Invasiveness is a critical event in the development of malignancy in brain tumors. A potential molecular mediator is basic fibroblast growth factor @FGF). NM-3T3 cells transfected with the bFGF gene fused with a signal peptide sequence (signal peptide-bFGF) acquire an invasive phenotype as measured by in uitro assays of invasion including: (A) the formation of branching networks on Matrigel; (8)invasiveness in a chernoinvasion assay; (C) migration in a cell spreading assay; (D) detection of an M, 92,000 gelatinase; and (E) local invasion into the surrounding neuropil after injection in the atktymic mouse brain. By contrast, cells transfected with only the native bFGF gene (wild-type bFGF): (A) formed discrete cell clusters on Matrigel; (B) were less invasive and migratory in vitro; (C) released minimal M, 92,000 gelatinase; and (D) in vivo fonned a pseudocapsule that separated the tumor cells from the neuropil. Quantitation of bFGF in the serum-free conditioned medium of the ce11 lines by enzyme-linked immunosorbent assay demonstrated that the signal-peptide-bFGF cell clone seaeted bFGF. These hdings suggest a role for bFGF-mediated pathways and collagenase as molecular determinants of invasiveness in the brain.

INTRODUCTION

A pathological hallmark of malignant brain tumors is local invasion into the surrounding normal brain (18, 47). Despite the importance of invasiveness to the dinical outcome for patients with brain tumors (9), little is known about the cellular and molecular mechanisms that regulate neoplastic infiltration. Basic fibrablast growth factor (bFGF) is a potent angiogenic mitogen (13), that has been identified in elevated levels in human gliomas (15, 16, 2426, 31, 41, 42, 50) and in the microvasculature of glioblastomas (3, 5,39). In vitro, exogenous bFGF induces ce11 motility (40), glioblastoma ceU migration, and invasion (10, 22) and stimulates the production of proteases (4,27). Recently, glioma cell growth was shown to be dependent on bFGF (25), and the release, or secretion, of bFGF was required for prornoting glioma ce11 growth (25). The native gene for bFGF does not normally code for a secretory signal peptide (1) therefore, bFGF rernains primarily a cell-associated protein (33), and its mode of action is undefined.

The overexpression of the bFGF gene was detected in the human glioblastoma cell line A-172, in cornparison with other human glioma cell lines and normal brain (11). When this ceil line was transplanted into the athymic rat brain, bFGF messenger ribonucleic acid (RNA) was detected by in situ hybridization primarily at the tumor-normal brain interface (Il), suggesting a role for bFGF in the heightened angiogenesis, proliferation, and invasion at the turnor edge. Because of the obvious relevance of bFGF to the biology of brain tumors, we hypothesized that ce11 transfected with the bFGF complementary deoxyribonucleic acid @NA) fused with the code for a secretory signal sequence, signal peptide-bFGF (34), would be both angiogenic and invasive in contrast to cells transfected with the native, wild-type-bFGF gene. We report that NIH-3T3 cells transfected with the signal peptide-bFGF expression vector are highly invasive in nitro and in situ in the brain.

MATERIALS AND METHODS

NZH-3T3 fibroblasts that have receptors for, and respond mitogenically to bFGF (46), were CO-transfectedwith a vector for neomycin resistance, and with a previously described vector (34) directhg the expression of normal bovine brain bFGF. The signal peptide-bFGF cells were transfected with the bFGF complementary DNA modified by the fusion of the code for a mouse irnmunoglobulin heavy-diain signal peptide (34). Ce11 dones were isolated by selection with G418, and multiple ce11 clones analyzed for growth and bFGF expression (34). After extensive characterization of multiple ce11 clones, two clones demonstrated to be representative of the transfection conditions (34) were generously provided by Dr. Michael Klagsbrun. Cells were cultured in 10 ml of Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 100 N/ml penicillin, 100 pg/ml streptomycin, and 0.5 mg/ml geneticin (Life Technologies, Inc., Bethesda, MD). Cells were maintained at 37OC in an atrnosphere of 5% carbon dioxide and 95% air, in a humidified incubator, with medium replaced twice per week, and cells were harvested for experimental procedures at confluence. After removal of the medium, the cells were washed in a phosphate-buffered saline (PBS)solution and detached from the culture vessels afier a five minute incubation with 0.05% trypsin-EDTA (Life Technologies, Inc.). The =sin was inactivated by reexposure of the cells to the culture medium. After centrifugation at 1000 rpm for 10 minutes, the ce11 pellet was resuspended in PBS and the number of cells counted with a hemacytometer.

Basic Fibroblast Gro wth Factor Quantitation

The signal pep tide-bFGF and wild-type-bFGF cells were grown to confluence, then were lysed on ice in 1 ml of chilled extraction buffer (50 mM Tris pH 8.0,150 mM NaCl, 2% NP-40,and the protease inhibitors EDTA-Na, pefabloc SC, pepstatin and leupeptin (Boehringer Mannheim Corp., Indianapolis, IN). The cell lysates were then centrifuged at 16,000 x g for 20 minutes at 4OC to remove insoluble cellular debris, and the supernatants aliquotted and stored at -80°C. For conditioned medium, cells were grown to confluence, then the medium replaced with sem-free DMEM for 48 hours. The medium was then collected, cenhifuged at 16,000 x g for 20 minutes at 4OC. and the supematants aliquotted and stored at -80°C. Protein concentrations were determined utilizing a modified Lowry assay (Bio-Rad, Richmond, CA). The amount of bFGF was determined utilizing a sensitive solid phase bFGF ELISA capable of detecting bFGF in amounts as low as 2.0 pg/ml (Amersham Corporation, Arlington Heights, IL,). Determinations were made from duplicate wells and bFGF antigen concentrations calculated as nanograms of antigen per milligram of total protein.

Western Biot Analysis

Equal protein arnounts of the cellular Iysates of the transfected cells were loaded on a precast 14% polyacrylamide gel (NOVEX, San Diego, CA), electrophoresed at 125 constant volts, then electrotransferred to a 0.45 pM nitrocellulose membrane (NOVEX). The membrane was incubated with the DE-6 anti-human bFGF monoclonal antibody (26) at a 1:500 dilution, followed by an alkaline phosphatase-conjugated goat anti-mouse secondary antibody (Kirkeguaard and Perry Laboratories, Gaithersberg, MD). Basic FGF was visualized with the chromogenic substrate 4nitroblue tetrazolium chloride/5 bromo-4 chloro-indoyl-phosphate (Kirkeguaard and Perry Laboratories). Negative controls for immunoreactivity induded: (i) the omission of the primary antibody, to control for the non-specificiv of the secondary antibody; and (ii) substitution of the prirnary antibody with purified mouse IgG (Organon Teknika Corp., West Chester, PA). In Vitro lnvaslveness Assays

For the qualitative assay of invasiveness, the basement membrane Matrigel (Collaborative Biomedical, hc., Bedford, MA) was used with protein concentrations ranging from 9.5 to 10.1 mg/ml. Six-well culture plates (Becton Dickinson, Lincoln Park, NJ) were coated with 0.5 ml/well Matrigel and placed in a humidified incubator at 37OC for 60 minutes. Following polymerization of the Matrigel, the signal peptide-bFGF and wild-type-bFGF cells in culture medium were carefully added to the wells at a concentration

of 5.0 x 104 cells/well, and examined after 24 and 48 hours in culture. At the conclusion of each time interval, tissue culture medium was removed and 2 ml of Dispase (Collaborative Biomedical, Inc.) was added to each culture well for dissolution of the Matrigel and the formation of a single ce11 suspension; the number of cells were counted by hemacytometer.

Chernoinvasion Assay

The chernoinvasion assay method was sirnilar to that reported previously (2). Ce11 culture inserts containing an 8 pm pore membrane, pre- coated with Matrigel, 100 pg/cm2 (Collaborative Biomedical, hc.),were placed in a hurnidified incubator at 37OC for 60 minutes to polymerize the Matrigel. The wild-type-bFGF and signal peptide-bFGF cells resuspended in serum free DMEM supplemented with 1.0% bovine serum albumin (BSA) (Sigma Chemical Co., St. Louis, MO), were then carefully added to the upper chamber at a concentration of 4.0 x 104cells/well. Self-conditioned serum-free DMEM (24 hours) from each ce11 line was used as a chemoattractant in the lower Chamber, Non-transfected NI.-3T3 fibroblasts were used as a control. Cells were allowed to invade for 24 hours, fixed in chilled 100% methanol, then Matrigel and ceUs remaining in the upper chamber were mechanically removed with a cotton swab. The filters were then stained with hematoxylin and mounted on glass slides.

Cellular Migration

The tumor cell migration assay was developed from that of Sato and Rifkin previously used to quantitate migration of endothelial cells (37,42). Confluent monolayers of the signal peptide-bFGF and wüd-type-bFGF cells were grown in 100 mm dishes. Cells were cut, and the plate marked by pressing a razor blade down ont0 the plate, then the blade was gently moved to one side to remove part of the ce11 sheet. After scraping, the adherent cells were washed twice with PBS, incubated for 16 hours in DMEM containing 1% fetal bovine serum, then fixed with absolute methanol, and stained with alcohof eosh.

Cellular invasion and migration in vitro were quantitated by computerized image analysis (Quantimet 570, Leica, Deerfield, IL). For invasion, the number of cells that invaded the Matrigel and migrated to the lower surface of the membrane were counted from triplicate wells. For ce11 migration, the distance under 10x mapification, between the leading edge of the cells from ten random fields was measured using the original cut as the origin (42).

Zymograph y

The signal peptidebFGF and wild-type-bFGF cells were grown to 80% coduence, then serum containing medium was replaced with serum-free DMEM supphmented with 1%BSA. Conditioned medium was removed at 24 and 48 hours, centrifuged for 10 minutes to sediment cellular debris, the supernatant collected and concentrated 10-fold with a centriprep concentrator (Amicon hc. Beverly, MA). Protein concentrations were determined using a modified Lowry assay (Bio-Rad Laboratories) and samples stored at -80°C. To detect gelatinases, 25 pl sarnples of the conditioned medium, standardized by the protein concentration, were applied to precast 10% acrylamide, 0.1% gelatin gels (NOVEX, San Diego, CA), and electrophoresed for approximately 90 minutes at 125 volts. Gels were renatured in 2.5% (v/v) Triton X-100 in water, then developed overnight at 37OC in developing buffer (NOVEX). Gels were stained with Coomassie R-250 (Gibco Laboratories), and gelatinolytic activity detected as a clear band of lysis against the blue background of stained gelatin (43).

Corneal Angiogenesis Assay

For quantitation of angiogenesis, the signal pep tide-bFGF and wild- type-bFGF tumors, grown in athymic mice, were harves ted for implantation into corneal micropockets (14,42). Twenty male New Zealand white rabbits were anesthetized with xylazine hydrochloride 5 mg/kg, and ketamine hydrodoride 35 mg/kg by intramuscular injection. The rabbit comea was flushed with sterile 0.5% tetracaine hydrochloride, and the globe proptosed. Using a stereomicroscope, a comeal micropocket was formed to within 2.0 mm of the lirnbus, and 1 mm fragments of the tumors were placed into the

micro~ockets.A Comeas were examined daily after implantation of tumor, until the tirne of sacrifice, 7 days post-implantation. A total of 120 comeal micropockets were made, divided equally into three groups (n=40), containing: a) signal peptide-bFGF; b) wild-type-bFGF tumor fragments; or c) blank micropockets, as a control for injury. The angiogenic response was quantitated using an ocular microscale, measuring the vesse1 length from the limbus to the maximal edge of the vascular sprout. The vascular density was scored utilizing the following scale (44): O = no blood vessels; 1 = 1-10 blood vessels; 2 = >IO vessels, loosely arranged, details of the iris could be viewed between gaps in the growing vessels; 3 = >IO vessels tightly packed, the iris could not be viewed; an angiogenesis index was calculated multiplying vessel length by density.

Twenty male athymic mice (Nu/Nu homozygous) between 25-30 grams were anesthetized with sodium pentobarbital, 65 mg/kg, by intraperitoneal injection supplemented with a subcutaneous injection of atropine sulfate 0.4 mg/kg. Animals were divided into two equal groups, receivuig either the wild-type-bFGF or the signal peptide-bFGF ce11 line. A 50 p1 Hamilton syringe with a 27s-gauge needle, was used to inject 10 pl of the ce11 suspension (1.O x 106 cells) into the right frontoparietal lobe of the mice to a depth of approximately 2.8 mm. Bone wax was applied at the injection site to prevent the reflux of cells, the skull surface flushed with sterile saline, and the scalp closed with nylon suture. Upon awakening, each animal was examined to exclude neurologie injury or other untoward effects.

Animals were sacnficed 10 days following the implantation of cells. On the day of sacrifice, skulls were opened, brains removed and the maximal diameters of the tumor measured in three planes: (dl) coronal; (d2) sagittal; (d3) transverse, and tumor volume calculated with the formula dl x d2 x d3 x rc /6 (48). Coronal sections were prepared for light rniaoscopy by fixation for 24 hours in 70% ethanol. The tumor and surroundhg brain were then embedded in paraffin and sectioned at 5pthidmess. For electron microscopy, 1 mm cubes of both tumor and normal brain adjacent to the tumor were immersed for 2 hours in chil1ed 3% gluteraldehyde in 0.1 mM cacodylate buffer, then postfixed in osmium tetroxide, dehydrated, and embedded in epon. Thin sections were stained with 3% uranylacetate and lead citrate, and examined with a Phillips 301 transmission electron microscope.

To determine the rate of tumor ce11 proliferation in situ in the brain, a monoclonal antibody to bromodeoxyuridine (BUdR)was utilized (8). Three hours prior to sacrifice, the athymic mice received 60 mg/kg by intraperitoneal injection of BUdR and 6 mg/kg fluorodeoxyuridine (Sigma), a cornpetitive inhibitor of thymidine uptake. Sections were incubated with 2N HC1 for 30 minutes, followed by borax buffer, pH 8.5 for 5 minutes and then washed three times in PBS. Tissues were then incubated with 30% hydrogen peroxide in absolute methanol (1:9) for 10 minutes, followed by 10% normal goat semm for 20 minutes, and then the anti-BUdR monoclonal antibody at a 1:20 dilution for 30 minutes at 37'C. After washing in PBS, tissues were incubated with a biotinylated secondary antibody (Zymed Streptavidin-Biotin System, Dimension Laboratories, Mississauga, ON) for 10 minutes at room temperature, washed again in PBS, and then incubated for 5 minutes with concentrated streptavidin-peroxidase conjugate. The was localized by final incubation of 5 to 10 minutes with diaminobenzidine 1 mg/ml in PBS prepared with an equal volume of 0.03% hydrogen peroxide. The slides were washed in water, counterstained with nuclear fast red, dehydrated, deared in xylene, and mounted. Cells that incorporated BUdR showed a bladc nuclear pigment. The BUdR tumor cell labeling index was determined as the percentage of IabeIed cells in relation to the total number of cells scored. Statistical Analysis

Results are expressed as mean + standard deviation. Statistically significant differences between means were determined utilizing a two-tailed t-test for unpaired samples, or a one-way analysis of variance. Differences with P < 0.05 were considered statistically significant.

RESULTS

b FG F Qumtita tion

The amount of ce11 associated bFGF was slightly higher in the signal peptide-bFGF ce11 clone (29.6 ng/rng protein), compared to the wild-type-bFGF ce11 clone (24.9 ng/mg protein) (Figure 1A). Basic FGF was almost undetectable in the conditioned serum-free medium of the wild-type-bFGF ce11 clone (5.9 pg/mg protein). By contrast, it was detected in the conditioned medium of the signal peptide-bFGF ce11 clone (72.6 pg/mg protein) (Figure 123).

Analysis of the ce11 clones by western blot revealed the signal peptide- bFGF expressing cell clone expressed in addition to the prototypical 18 kD bFGF isoform, several higher molecular weight bFGF isoforms (Figure 2). The mechanism responsible for the expression of these higher molecular weight isoforms is not known. The detection of higher molecular weight isoforms was similar to that described for the human glioma cells (Chapter III), and supports the hypothesis that expression of higher molecular mass bFGF isoforms is associated with a malignant phenotype. FIGURE 1 A, the amount of cell-associated bFGF was quantitated by enzyme-linked immunosorbent assay, demonstrating that the wild-type-bFGF (wt-bFGF) and signal peptide-bFGF (sp-bFGF) ce11 clones express comparable levels of bFGF. By contrast, B, demonstrates that the signal peptide-bFGF ce11 clone released bFGF into the serum-free conditioned medium.

FIGURE 2. CeU-associated bFGF expression was examined by western blot. The non-transfected and wild-type-bFGF cells expressed only the prototypical 18 kD bFGF isoform. By contrast the signal peptide-bFGF ceIl clone expressed several higher molecular weight bFGF isoforms.

In Vitro ln vasiveness Using Matrigel

Growth of the wild-type-bFGF and signal peptide-bFGF cells when sparsely plated on plastic showed no morphological differences after 24 hours. By contrast, there were significant qualitative differences, following twenty- four hours growth on Matrigel-coated plates: the signal peptide-bFGF cells formed brandiing colonies whereas the wild-type-bFGF cells formed individual aggregate colonies. At 48 hourç, the signal peptide-bFGF cells continued to form extensive branching colonies (Figure 3A) ; in sharp contrast, the wild-type-bFGF cells remained as individual ce11 cluçters (Figure 3B). There was no difference in the number of signal peptide-bFGF or wild- type-bFGF cells growing on Matrigel at 24 or 48 hours.

Chernoinvasion Assay

When Matrigel was excluded in the chernoinvasion assay, equal numbers of the wild-type-bFGF and signal peptide-bFGF cells migra ted through the 8 PM pores to the lower surface of the membrane. The control, non-transfected NM-3T3 fibroblasts did dot invade the Matrigel. The wild- type-bFGF cells were more invasive than non-transfected cells with 132 k 114 cells invading. The signal peptide-bFGF cells were the most invasive with 418 t 144 cells invading the Matrigel and crossing the membrane (P = 0.05). (Figure 4).

Cellular Migration

Using the migration assay, the wild-type-bFGF cells spread a distance of only 244.5 f 8.3 p into the denuded area (Figure 5A), in contrast the signal peptide-bFGF ceUs migrated significantly farther, to a distance of 871.21 9.0 pm (P = 0.001) (Figure 5B).

Zymography

Zones of gelatin gel lysis corresponded to gelatinases of approximately 68 to 72 kD were detected in the serum-free conditioned medium of the wild- type-bFGF and signal peptide-bFGF cells. Zones of lysis were also seen at approximately 42 kD, that correspond to activated interstitial collagenase. There was a slight increase in these bands in the signal peptide-bFGF cell line conditioned medium at 48 hours (Figure 6). A higher molecular weight zone of lysis corresponding to approximately 92 kD was dearly evident in the 48 hour signal peptide-bFGF ce11 conditioned medium (Figure 6) but far less prominent in the conditioned medium of the wild-type cells.

Angiogenesis

The angiogenic activity of the two ce11 clones implanted in the rabbit comea assay showed an increase in vascular density produced by the signal peptide-bFGF cells (Table l),but no sipificant difference in the vesse1 lengths or in the overall angiogenesis index (44). As a control, the forty control rnicropockets that contained no tumor implants, showed no angiogenic response.

Brain Tumor Model of lnvasiveness

Electron rnicroscopy revealed that the wild-type bFGF ce& grew intracerebrally as well-circumsaibed tumors, with a sharp border that separated the tumor from the adjacent normal brain (Figure 7A). A moderately electron dense material was deteded, that constituted a "pseudocapsule" that separated the tumor ce& from the neuropil (Figure FIGURE 3 Appearance of signal-peptide-bFGF (A) and wild-type-bFGF cells (B) at 48 hours after plating on basement membrane Matrigel. The non- transfected fibroblasts adopt the phenotype of the wild type-bFGF cell clones (not shown). Following digestion of the Matrigel with dispase, the number of cells was determined by counting on hemacytometer. There was no significant difference in the number of cells between wild type and signal peptide-bFGF ce11 clones.

FIGURE 4 The number of signal peptide-bFGF (sp-bFGF) cells that invaded the Matrigel and crossed the porous membrane is significantly higher than for the wild-type-bFGF (wt-bFGF) cells. *, P < 0.05, Student's t-test. None of the control, nontransfected NM-3T3 fibroblasts were able to invade the Matrigel and cross the membrane. Mean Number of Invading Cells

NIH 3T3 wt-bFGF sp-bFGF FIGURE 5 After the razor blade cut (arrow), the wild-type-bFGF cells were less rnigratory (A); however, the signal peptide-bFGF cells were highly rnigratory, with significantly farther spreading distance into the denuded area (B)(P < 0.0001, Student's t-test).

Table 1. Fragments of the wild-type-bFGF, or signal-peptide bFGF tumors were impianted in the rabbit comea. The angiogenic response to the tumor fragments demonstrated a significantly increased vascular density for the signal-peptide tumors in cornparison to the wild-type bFGF tumors. Control corneal pockets that contained no tumor did not induce an angiogenic response. Table 1 Wild-Type Signal Peptide P 7B). In contrast, the signal peptide-bFGF cells formed tumors with an ill- defined border with infiltration of the tumor cells into the neuropil (Figure BA). There were frequent distortions of the neuropil, and a distinct absence of the "pseudocapsule" separating the -or mass from normal neighboring structures (Figure 8B).

Both ce11 lines were tumorigenic, forming tumors in al1 mice following subcutaneous and intracerebral injection. No significant difference in the intracerebral tumor volume was found between wild-type-bFGF, 26.7 + 11.6 mm3, and the signal peptide-bFGF tumors, 39.1+ 20.3 mm3. The BUdR labeling index of the wild-type-bFGF tumors, 35.8 + 5.376, was slightly higher in cornparison to the signal peptide-bFGF tumors, 27.8 + 5.4% (P = 0.05) (Figure 9).

DISCUSSION

The mechanisms of invasion is central to the biology of malignant brain tumors. The rnicroscopic spread of malignant gliomas into the surrounding normal brain results in their local recurrence (28,46), and is a factor that accounts for the failure of traditional therapies. To study the role of bFGF in neoplastic invasion in the brain, we utilized well characierized dones transfected with bFGF expression vectors, with and without a seuetory signal sequence. We now report that ceUs transfected with the bFGF cDNA fused to a seuetory signal peptide sequence are invasive in vitro and in situ in the brain. These findings are important because they: (i) provide insight into possible mechanisrns of neoplastic invasion in the brain pointing to bFGF and the 92 kD gelatinase as potential molecular determinants; (ii) underscore the dose relation between angiogenesis and neoplastic invasion; FIGURE 6 Zymogram gel illustrating the digestion of gelatin, with zones of enzymatic activity characterized by negative staining. A, wild-type-bFGF cell conditioned medium at 24 and 48 hours; B, signal peptide-bFGF ce11 conditioned medium at 24 and 48 hours. The signal peptide-bFGF cell clone at 48 hours released significant amounts of M, 92,000 gelatinase and M, 42,000 activated interstitial collagenase (arrows). 24 Hours 48 Hours

A B A B FIGURE 7 Ten athymic mice were implanted (5 with wild-type, and 5 with signal peptide bFGF cells). Tumors formed in al1 animals, and sections were examined by electron microscopy. Representative sections are displayed here: A, tumor cells (T) lacking the signal peptide do not invade the neuropil (N) (original magnification, x 400); B, the himor border iç weli defined, with tumor cells separated from the neuropil by a pseudocapsule (arrow) (original magnification, x 2500).

FIGURE 8 A, neoplastic cells transfected with signal peptide-bFGF display an invasive phenotype and infiltrate the neuropil at the tumor edge, spreading along perivascular pathways (original magnification, x 400); B, the infiltrative spread of signal peptide-bFGF cells occurs without a defined border, and signal peptide-bFGF cells are found invading the neuropil (original magnification, x 4500). N, neuropil; T, tumor cells.

FIGURE 9 Wild-type-bFGF and signal peptide-bFGF (sp-bFGF) intracerebral tumor cell BUdR labeling index. The wild-type bFGF tumor ce11 labeling index in situ in the brain is significantly higher in comparison with the signal peptide-bFGF hunor cells. +,P < 0.05, Student's t-test. Tumor Cell BUdR Labeling Index (%) and (iii) provide a reproducible mode1 to study the molecular mechanisms of bFGF-related neoplastic invasion in vitro and within the brain.

Invasion is an active process involving a cascade of cellular events that include adhesion, migration, and proteolytic degradation of the extracellular matrix (20). In this study, the signal peptide-bFGF transfected cell clone was highly migratory (Figure SB) and invasive. Zyrnographic analysis of the conditioned medium from the wild-type-bFGF and signal peptide-bFGF ce11 clones demonstrated elevated 92 kD gelatinase only in the signal peptide- iFGF clone (Figure 6). This finding identifies an important role for proteinases, particularly the 92 kD gelatinase, as a molecular mediator of invasiveness in the brain, similar to that described in tissue extracts of human glioblastoma (32).

The molecular mechanisms responsible for the elevated gelatinolytic activity of the signal peptide-bFGF cells remains unknown, but could be linked indirectly to the extracellular release of bFGF (23). Basic FGF was detected in the serum-free conditioned medium of the signal peptide-bFGF ce11 clone (Figure ZB). Because the signal peptide is responsible for directing the extracellular release of proteins, it is not surprising that bFGF was detected in the medium. However, previous reports (34,3547) did not find bFGF in the medium of the signal peptide-bFGF cells, perhaps because the ELISA method of bFGF detection is highly sensitive. Extracellular bFGF is a potent stimulator of urokinase-type plasminogen activator (12) an enzyme identified at elevated levels in glioblastoma multiforme (19). Urokinase is responsible for the conversion of inactive plasminogen to activated plasmin (17), capable of converting the inactive matrix metalioproteinases, including the gelatinases, to the active enzyme (17). Gelatinases have been linked to angiogenesis (27), and neoplastic invasion in several tumor systems (20), including human glioblastoma (32).

The signal peptide-bFGF and wild-type-bFGF ce11 dones formed tumors in all animals both subcutaneously and intracerebrally, with no significant difference in the tumor volumes, and only a slight difference in BUdR labeling index (Figure 9). These findings demonstrate that the invasive phenotype of the signal peptide-bFGF cells is not related to turnor size or the proliferative rate. Invasiveneçs, however, was linked to angiogenesis. In the corneal assay, the signal peptide-bFGF tumor fragments induced an angiogenic response with increased vascular density. In situ in the brain, the signal peptide-bFGF cells were found invading dong the preexisting blood vessels within the brain parenchyma (Fig 8B), similar to that described for rat C6 glioma (29,50) VX2 carcinoma (48), and a known route of spread for malignant human glial tumors cells (36). These data suggest that invasion in this mode1 may not only be functionally reIated to angiogenesis (6, 21,23,27), but may be angiogenesis-dependent (30), consistent with the finding that agents that suppress angiogenesis also inhibit invasion (6,7), or the production of proteolytic enzymes (4243).

In surnmary we have shown that MH 3T3 fibroblasts transfected with a signal peptide-bFGF expression vector are invasive in aitm and in situ in the brain in cornparison to celis transfected with the wild-type-bFGF gene. Continuing studies will focus on the role of bFGF in the invasiveness of human glial lumors. Recently, non-transformed human fetal astrocytes transfected with the signal peptide-bFGF expression vector were found to be transformed in vitra. Taken together these findings suggest thai a fibroblast cell line engineered to secrete bFGF are more invasive and migratory than a ce11 line secreting smaller arnounts of bFGF. Furthemore, secreted bFGF may have the potential to mediate tumor ce11 invasion and migration in the brain. Chapter IV - Limitations Transfections The transfection experiments reported in this chapter would have been strengthened by the analysis of multiple clones from the wild-type and signal peptide-bFGF transfectants. Although the cell clones utiüzed in these studies were considered "representative" of the phenotype obtained following transfection, the data would have been more conclusive if multiple ce11 clones exhibited the same invasive phenotype in vitro and in the brain. The data would have also been strengthened by the inclusion of ceLl clones that expressed bFGF at levels greater than the non-transfected 3T3 fibrobtasts. Without this control it is difficult to detemine if the observed phenotypes are the result of bFGF overexpression, rather than a switch to extracellular export of bFGF. The original description of the bFGF transfected ce11 clones included a clone that expressed 100 times excess of bFGF in cornparison to the non-transfected control cells (Rogelj, S., et al. Nature 331:173, 1988). Despite the overexpression of the cell-associated bFGF, these ceIl clones did not adopt the aggressive phenotype of the signal peptide-bFGF cells clones, suggesting the invasive phenotype of the sp-bFGF cell line may not be due to the overexpression of bFGF protein. In Chapter IV we demonstrated that cells expressing a secreted fom of bFGF were highly invasive in vitro and in the brain, suggesting bFGF may have the potential to mediate tumor ce11 invasion and migration in the brain. Taken together with the data fiom Chapter III, we have asked at what point in the progression of gliornas is there a switch to the extracellular export of bFGF? To address this question in Chapter V we report on the stable transfection of non-transformed human fetal astrocytes with the modified bFGF cDNA that codes for a secreted fonn of bFGF.

In addition, in Chapter V, we have asked if modulation of bFGF gene expression in an established glioma cell line would suppress growth. W e now report on the application of bFGF antisense oligodeoxynucleotides in the regdation of glioma ce11 proliferation. Chapter V

The Potential Role of Basic Fibroblast Growth Factor in the Transformation of Cultured Primary Human Fetal Astrocytes and the Proliferation of Human Glioma U-87 Cells ABSTRACT

Basic fibroblast growth factor @FGF) is a potent stimulator of angiogenesis, proliferation, and invasion in human gliomas. To test the hypothesis that bFGF is important in the development of the rnalignant phenotype of human gliomas, bFGF expression was prospectively modulated in primary human fetal astrocytes and in an established human glioma ce11 line. Fetal astrocytes were transfected with a vector expressing bFGF modified by the addition of a secretory signal peptide sequence (sp-bFGF). Two sp-bFGF astrocyte clones examined in vitro demonstrated anchorage independent growth, loss of contact inhibition, and decreased glial fibrillary audic protein immunoreactivity, changes consistent with cellular transformation. To analyze the inhibition of bFGF expression, phosphorothioated bFGF antisense oligodeoxynucleotides were added to cultures of the U-87 human glioma ce11 line. The U-87 cell proliferation was inhibited to 70.6 I 0.4% of control at 10 PM, and 53.2 I 5.6% of control at 20 pM (P c 0.05). Both the 7.0 and 4.0 kB bFGF mRNA transcripts were reduced following antisense oligodeoxynucleotide exposure, and cell-associated bFGF protein was reduced by 44%. The sense oligodeoxynucleotide, a negative control, failed to inhibit U-87 proliferation. These data support the concept that bFGF expression could be a key event in glial tumorigenesis, that may be necessary for the sustained growth of human gliomas.

INTRODUCTION

The malignant phenotype of human gliomas is diaracterized by local infiltrative grow th (Il),angio genesis (6,26,50) and an elevated p rolifer ative rate (53,54). Basic FGF is a potent angiogenic mitogen (14), that stimulates the production of proteinases (14,27, 28, 29, 52), and proliferation in glioma cells (17, 18, 32, 33). The addition of bFGF to human glioma cells in vitro was shown to induce cellular motility and invasiveness (12). By contrast, depletion of copper, an ionic cofactor of bFGF (48) inhibited glioma ce11 invasiveness (8). These observations have focused our laboratory on the expression of bFGF in human malignant brain tumors (7) and on the role of bFGF in the growth of human glioma cells (17). The following studies were designed to test the hypothesis that bFGF regulates glioma ce11 growth and may be used as a target to suppress glioma ceIl proliferation.

To determine the transforming potential of bFGF, human fetal astrocytes were stably transfected with a bFGF expression vector, and examined for indications of cellular transformation in vitro. Phosphorothioated bFGF antisense oligodeoxynucleotides were also utilized to identify a role for bFGF in the proliferation of an established human glioma cell Iine. The results suggest that bFGF could be important in glial tumorigenesis and glioma ce11 proliferation.

MATERIALS AND METHODS

Cell Culture

Primary human fetal astrocytes, W585 (55), were grown in Dulbecco's Modified Eagles Medium @MEM) supplemented with 10% fetal bovine serum (FBS), 100 IU/ml penicillin and 100 pgfml streptomycin. The established human glioma cell luie U-87,(American Type Culture Collection, Rockville, MD) was cultured in DMEM supplemented with 10% calf serum, 100 IU/d penicillin and 100 pg/ml streptomycin (Gibco BRL, Grand Island, NY). Cds were maintained in a humidified hcubator at 37OC in an atmosphere of 5% CCP and 95% air, with the medium replaced twice per week.

Ce11 Proliferation Assay

Ce11 proliferation was determined utilizing the Ce11 Titer 96TM Aqueous Non-radioactive Ce11 Proliferation Assay (Promega Corp., Madison, WI). This system rneasures the bioreduction of (3-(4,s-dimethylthiazol-2-y1)- 5-(3-carboxyme thoxypheny1)-2-(4-sdfophey) tetrazolium (MTS) to an aqueous soluble formazan in the Dresence of the electron coudine; rea~ent, A A U V . methosulfate (PMS) by proliferating cells. An absorbance reading the quantity of formazan produced was obtained using an automated microplate reader (Bio-Tek Instruments, Winooski, VT) . Background absorbance from wells containing culture medium only was subtracted from the sample absorbance values to obtain the corrected absorbance.

Basic FGF and Fetal Astrocyte Proliferation

The W585 fetal astrocyte cells were plated in triplicate in a 96-well tissue culture plate (Becton Dickinson, Lincoln Park, NJ) ai a concentration of 2.5 x 103 cells/well. The following day fresh medium was added with increasing concentrations (0-75 ng/well) of recombinant human bFGF (provided by the DuPont Merck Pharmaceutical Corporation, Wilmington, DE). The cells were hcubated for 48 hours, then 20 pl of the cornbined MTS/PMS reagent was added to each well, and plates were incubated an additional 2 hours at 37OC. Reverse Transcription and Polymerase Chain Reaction

Total RNA was isolated from the W585 human fetal astrocytes and the U-87 malignant glioma ce11 line (10). Basic FGF receptor primers were chosen based on the sequence of the transmembrane domain of the FGF receptor-2 (FGFR-2), a region conserved in the FGF receptor family (2136); (Oligonucleotide primer 1) 5' GCC AGC ACT CCC GCA TCA TCA 3' complementary to nucleotides 2180-2200, and (Oligonucleotide primer 2) 5 ' GAC GCA ACA GAG AAA GAC ?TG 3' corresponding to nucleotides 1543- 1563. Reverse transcription was carried out for 60 minutes at 37OC in a 33 pl reaction consisting of 5 pg of total RNA template, 3 pl of 20pM oligonucleotide primers 1 and 2, 1X reverse transcription buffer (50 mM Tris- HCI, (pH 8.3); 40 mM KCl; 6 mM MgCl, 1mM dithiotreitol (DTT)), 2 p1 each dNTP (25 mM), 6 pl DTT (0.1 M), and 600 units of Superscript reverse transcriptase (Gibco BRL). Polymerase chah reaction (PCR) was perforrned in 60 pl volumes consisting of 10 pl of the reverse transcription reaction, 2 pl of oligonucleotide primers 1 and 2 (20 PM), 1X PCR buffer ( 100 rnM Tris-HCl, (pH 8.3); 500 mM KCl), 2.5 mM MgCl,;, 0.2 mM each dNTP, and 2.5 U Taq DNA polymerase (Gibco, BRL). Forty cycles were performed (1 minute at 94OC, 1 minute at 50°C and 2 minutes at 72OC). The predicted 658 base-pair cDNA product was detected. To confirm that the obtained cDNA was correct, it was re-amplified using nested bFGF receptor primers; (Oligonucleotide 3) 5 ' CCG CAT CAT CAT GTA CAG CTC GTT GG 3' complementary to nucleotides 2165-2190, and (Oligonucleotide 4) 5' GAC T'TG TCA GAC CTG ATC TCA GAA ATG G 3' corresponding to nucleotides 1558-1585. The reaaion mix was then examined following size fractionation by electrophoresis on a 1.2% agarose gel containing ethidium bromide. Transfections

Human fetal astrocyte cells were transfected utilizing Lipofectin reagent according to the manufacturers instructions (Gibco BRL). Approxirnately 2.0 x IO5 cells were plated in six-well tissue culture dishes, 24 hours prior to the transfection. One microgram of pSV2a 3.6, an expression vector that induces the expression of the alpha-chain of murine Na+/K+-ATPase, that confers resistance to ouabain (24), was transfected alone or in combination with 10 pg of the sp-bFGF expression vector (42) (Figtire 1). The DNA-Lipofectin mixture was incubated at room temperature for 10 minutes then added dong with serum-free DMEM in a final volume of 2.0 ml to the cells. After 5 hours of incubation at 37OC, the medium was removed, and fresh DMEM + 10% FBS added to quench the reaction, and the cells hcubated overnight. The following day, cells were split into three 24well tissue culture plates/condition, and ce11 clones allowed to grow.

Western Blot Analysis

Equal protein amounts of the cellular lysates of the transfected cells were loaded on a precast 14% polyacrylamide gel (NOVEX, San Diego, CA), electrophoresed at 125 constant volts, then electrotransfened to a 0.45 pM nitrocellulose membrane (NOVEX). The membrane was incubated with the DE6 anti-human bFGF monoclonal antibody at a 1:500 dilution, followed by an alkaline phosphatase-conjugated goat anti-mouse seconda7 antibody (Kirkeguaard and Perry Laboratories, Gaithersberg, MD). Basic FGF was visualized with the chromogenic substrate Cnitrobhe tetrazolium chloride/5 bromo-4 chloro-indoyl-phosphate (Kukeguaard and Perry Laboratories). Negative controls for immunoreactivity Iliduded: (i) the omission of the primary antibody, to control for the non-specificity of the secondary antibody; FIGURE 1 Restriction map of the sp-bFGF cDNA. The bFGF is under the transcriptional control of the Moloney murine leukemia virus long terminal repeat (Mo-MuLV LTR) (41). After translation, the mouse immunoglobulin heavy diain signal peptide (SP) is found fused to the second amino acid of the bovine brain bFGF. NS, intervening sequence; UT, untranslated; SV, sarcoma virus. Xba 1 Xba 1 EcoRl Sa1 1 I 1 1 I Coding bFGF 3' UT SV40 poly (A) Mo *MuLV 1 , SP IVS Sequenee 140 250 500 462 and (ii) substitution of the primary antibody with purified mouse IgG (Organon Teknika Corp., West Chester, PA).

Glial Fibriliary A cidic Protein lmmunocytochemistry

Glial fibrillary acidic protein (GFAP) was detected by immunocytochemistry in cells growing on chambered slides. The cells were plated at a concentration of 1.0 x 10' cells/ml, and allowed to attach overnight to the chambered slide. The following day, the medium was removed, and the cells fixed for 20 minutes in acetone/methanol (1:l) cooled in liquid nitrogen. The cells were then air dried, and incubated with 10% normal goat serum for 20 minutes, followed by an anti-human GFAP monoclonal antibody (Sigma Chernical Co., St. Louis, M.O.) at a 1:200 dilution for 60 minutes at room temperature, or with normal mouse serurn diluted in phosphate buffered saline (PBS) as a negative control. After washing in PBS, the cells were incubated with a biotinylated secondary antibody (Zymed Streptavidin-Biotin System, Zymed Laboratories, San Francisco, CA) for 10 minutes at room temperature, washed again in PBS, and then incubated for 5 minutes with concentrated streptavidin-fluorescein isoihiocyanate conjugate (Zymed Laboratories). The slides were mounted with an aqueous mounting medium (Aqua Mount, Lemer Laboratories) and GFAP was visualized with a Leitz Aristoplan microscope equipped with an epi-illuminator, interference and barriers filters for florescence.

Oligodeoxynucleotide Synthesis

Fifteen-base oligodeoxyiucleotides were synthesized at the Biotechnology Facility of Northwestern University, with a phosphorothioate substitution at each base. The oligodeoxynucleotides were desalted, lyophilized, then resuspended in sterile water and stored at -20°C. The oligodeoxynucleotides were targeted against the first splice donor-acceptor site at codon 60 of the human bFGF gene (2), and the sequence confirmed to be specific for bFGF using the BLAST algorithm (3). The antisense oligodeoxynucleotide was 5'-TAG-CTT-GAT-GTG-AGG-3';the corresponding sense oligodeoxynucleotide 5'-CCT-CAC-ATC-AAG-CTA-3'was used as a negative control.

Antisense Inhibition of Glioma Cell Proliferation

The U-87 human glioma cells were plated in triplicate into a 96-well tissue culture plate (Becton Dickinson, Lincoln Park, NJ) ai a concentration of 2.5 x 10) cells/well. The following day serum-free DMEM supplemented with Ham's F-12 and sense or antisense oligodeoxynucleotides at the concentration of 0,10 and 20 were added. Cells were incubated for 5 days, then 20 pl of the combined MTS/PMS reagent was added to each well, and plates were incubated an additional 2 hours at 37OC.

Northern 81ot Analysis

The U-87 human glioma cells were plated in 60 mm culture plates (Becton Dickinson) at a concentration of 1.0 x IO5 ceIls/well. The following day serum-free DMEM supplemented with Ham's F-12 and sense or antisense oIigodeoxyiucleotides at the concentration of O or 20 were added. Cells were grown for five days, then total cellular RNA was isolated (10). The RNA pellet was resuspended in RNAçe free water and stored at -80°C until used in northern blotting as described previously (46). The RNA was quantitated spebrophotometrically by its absorbance at 260 nm (1 OD = 40 pg/ml). Fifteen pg of total RNA was denatured by heating to 65OC in 2.2 M formaldehyde and 50% formamide in sample buffer containing 25 ng/pl ethidium bromide. The samples were eledrophoretically fractionated on a 1.2% formaldehyde- agarose gel in 0.02M morpholinopropane sulfonate (MOPS)buffer, pH 7.0. Following electrophoresis, gels were photographed under UV light, rinsed in water, and the RNA was transfened overnight hom the gel to a nylon membrane (Boehringer Mannheim Corp., Indianapolis, IN) by capillary transfer using 10X SSC (1X SSC is 0.015 M sodium citrate/O.lS M sodium chloride). The RNA was fixed to the filter by UV crosslinking.

Hybridization to 32P-labeled probes

Prehybridization was in 30% formamide, and standard buffer (46) for 5 hours at 42OC. A 1.5 kB EcoRl bovine brain bFGF cDNA fragment from the pbFGF plasmid (41) was labeled by random priming (Gibco-BRL) incorporating 32P-a-labeleddCTP (Amersham Laboratories, Arlington Heights, IL). The "P-labeled probe was denatured and added to the prehybridization solution, and hybridization to the membrane was performed for 24 hours at 42°C. Nonspecific binding was removed by washing in 0.5X SSC, 0.1% SDS (w/v) at 60°C. Visualization and quantification of the signals were obtained utilizing a Fujix BAS 2000 Phosphorimager, (Fuji Photo Film Co. Ltd., Japan). To ensure equal RNA loading, ethidium bromide stained ribosomal RNA bands, were photographed under W Iight. In addition, the membranes were probed secondarily with the "control" cDNA, p-actui.

Quantitation of bFGF &y ELISA

The U-87 human glioma cells were plated in 60 mm culture plates (Becton Dickinson) at a concentration of 1.0 x 105 cells/weU. Mer 24 hours, the medium was changed to serum-free DMEM supplemented with Hm's F- 12 and sense or antisense oligodeoxynuc~eotidesat the concentration of O and 20 PM. Cells were grown for five days, then lysed in 1 ml of chilled extraction buffer, containing 50 mM Tris pH 8.0, 150 mM NaCl, 2% NP40 (Sigma Chernical Co.), and the protease inhibitors EDTA-Na, pefabloc, pepstatin and leupep tin (Boehringer Mannheim Corp .) . The lysates were then centrifuged at 16,000 x g for 20 minutes at 4OC to remove insoluble cellular debris, and the supernatants aliquotted and stored at -80°C. Protein concentrations were detemined utilizing a modified Lowry assay (Bio-Rad, Richmond, CA). The amount of bFGF antigen was detennined utilizing a highly sensitive solid phase bFGF ELISA capable of detecting bFGF in amountç as low as 1.0 pghl (Amersham Corporation) (38). Basic FGF antigen concentrations were calculated as nanograms of antigen per milligram of total protein.

Statistical Analysis

Results are expressed as mean f standard deviation. Statistically significant differences between means were determined utilizing a two-tailed Student t-test for unpaired sarnples, or a one-way analysis of variance. Differences with P < 0.05 were considered statistically significant.

RESULTS

Fetal Asfrocyte Proliferation

The W585 fetal astrocytes were responsive to inaeasing concentrations of exogenous recombinant human bFGF (Figure 2). Cellular proliferation was significantly elevated ai 25 ng/well (P c 0.04), and 75 ng bFGF/well (P < 0.0001) in cornparison to non-bFGF exposed fetal astrocytes (Figure 2).

Basic FGF Receptor

Reverse transcription-PCR confirmed the presence of mRNA for the transmembrane domain of the bFGF receptor in both the W585 and U-87 ce11 lines. The bFGF receptor-specific primers 1 and 2 produced the predicted 658 nucleotide cDNA fragment. The nested primers 3 and 4 permitted the specific amplification of the bFGF receptor cDNA, and produced the predicted 633 nucleotide cDNA fragment (Figure 3, Lanes 1 and 3). The identity of the PCR products was confirmed by restriction digestion; the region of amplification of the bFGF receptor cDNA contains a single BglII restriction site (36). The BglII digestion produced the predicted 505 nucleotide fragment (Figure 3, Lanes 2 and 4) and a 128 nucleotide fragment (not visible).

Transfections

The fetal astrocytes in control - sham transfected and SV2a3.6 transfected groups failed to survive the split into the 24-well tissue culture dishes. Because cells in the control groups did not grow, sp-bFGF ce11 clones were selected on the basis of survival and proliferation alone. Of eleven sp- bFGF clones, two (84 and D3) were randornly selected for further analysis. The non-transfected astrocytes grew uniformly, and were contact inhibited (Figure 4A). By contrast, the B4 and D3 clones grew with an altered morphology (Figure 4B), and at confluence developed visible foci. The nontransfected fetal astrocytes were 100% positive for GFAP (Figure SA). By contrast, many of the sp-bFGF transfected cells had lost GFAP immunoreactivity (Figure SB). Northern analysis demonstrated two bFGF transcripts of approxîmately 7.0 and 4.0 kB in the nontransfected fetal FIGURE 2 The W585 human fetal astrocyte ce11 strain was responsive to increasing concentrations of human recombinant bFGF protein. Cellular proliferation was significantly stîmulated at 25 and 75 ng/well, cornpared with control. *, P < 0.04; **, P < 0.0001. 0.8 1 i rvmil r ..mmam I I . IV I~I Control 1 10 100 bFGF Concentration (ng I well) FIGURE 3 Detection of bFGF receptor mRNA in W585 and U-87 cells. Lanes 1 and 3 represent the predicted 633-base pair bFGF receptor cDNA product. Digestion of this fragment with the BglII restriction endonuclease demonstrates the predicted 505- (arrow) and 128 - (not visible) base pair fragments. Standard astrocytes. The sp-bFGF cells alço expressed these transcripts, as well as an additional transcript of approximately 988 bp that corresponded to the transfected bFGF cDNA.

Western Mot Analysis of Transfected Clones

To confirm that the B4 and D3 ce11 clones were expressing the transfected bFGF cDNA, cellular lysates were analyzed by western blot. The non-transfected W585 fetal astrocytes expressed only the prototypical 18 kD bFGF isoform. By contrast the B4 and D3 cell clones expressed several higher molecular weight bFGF isoforms (Fisire 6).

Antisense inhibition of Glioma Growth

The antisense oligodeoxynucleotides caused a concentration-dependen inhibition of U-87 ceIl proliferation to 70.6 f 0.4 % of control at 10 PM, and 53.2 f 5.6 % of control at 20 (P c 0.05) . The sense oligodeoxynucleotide did not significantly decrease ce11 proliferation (Fignre 7).

Northern blo t analysis demons trated two specific 'ÙFGF transcrip ts corresponding to approximately 7.0 and 4.0 kB in the U-87 ce11 line. The addition of 20 PM of the antisense bFGF oligodeoxynucleotide resulted in a 53% reduction in the amount of the 7.0 kB transcript, and a 42% reduction in the level of the 4.0 kB bFGF mRNA transcript (Figure 8). Quantitation of bFGF in the cellular lysates of the antisense exposed U-87 ce11 line demonstrated a 44% reduction in bFGF antigen levels in comparison to control, non-oligonucleotide exposed U-87 glioma cells (Fipire 9). The sense oligonucleotide did not cause a significant reduction in bFGF protein, in comparison to control (Fi're 9). FIGURE4 A, the nontransfected human fetal astrocyte ce11 strain was contact-inhibited at confluence. B, in contrast, the signal peptide-bFGF transfected astrocytes revealed an altered morphology, with the Ioss of contact inhibition.

FIGURE 5. A, the W585, nontransfected human fetal astrocytes were uniformly positive when immunostained for GFAP. B, the B4 and D3 signal peptide-bFGF transfected astrocytes demonstrated a significant loss of GFAP immunoreactivity and developed loci at confluence (Clone B4 is shown).

Figure 6. Examination of the cellular lysates of the transfected human fetal astrocytes demonstrated higher molecular weight bFGF isoforms of 22 and 24 kD. By contrast, the cellular lysates of the non-transfected fetal astrocytes demonstrated only the prototypical 18 kD bFGF isoform (see Chapter III).

FIGURE 7. The bFGF antisense oligodeoxynucleotide caused a concentration-dependent inhibition of the U-87 human glioma ce11 proliferation. *, P c 0.05. The sense oligodeoxynucleotide caused only a slight decrease in cellular proliferation. Sense Antisense

10 20

Oligodeoxynucleotide Concentration (PM) FIGURE S. Reduction in the U-87 human glioma ce11 bFGF mRNA. Exposure to 20 pol/L of the bFGF antisense oligodeoxynucleotide resulted in a 53% reduction in the 7.0-kB and 42% reduction in the 4.0 kB bFGF mRNA transcripts. The sense oligodeoxynucleotide caused only a slight decrease in mRNA transcrip ts. Control Sense Antisense FIGURE 9 Quantitation of the bFGF protein in the U-87 glioma ce11 line by ELISA demonstrated that the antisense oligodeoxynucleotide caused a 44% reduction in the amount of cell-associated bFGF. The sense oligodeoxynucleotide caused only a slight decrease in the bFGF antigen. Control Sense Antisense DISCUSSION

In the present study, we have analyzed the contribution of bFGF to the malignant phenotype of human glial-derived tumors. Induction of bFGF protein expression in primary human astrocytes resulted in the acquisition of a transformed phenotype. Conversely, antisense oligodeoxynucleotide inhibition of bFGF mRNA and protein expression in the U-87 established hurnan glioma cell line resulted in the significant inhibition of cellular proliferation. The findings presented are important because they: (i) link bFGF to the in vitro transformation of human astrocytes; (ii) suggest that aberrant expression of bFGF could be an early event in the multi-step development of human gliomas; and (iii) demonstrate that a decrease in bFGF mRNA transcripts and bFGF protein using antisense oligodeoxynucleotides results in a significant decrease in glioma ce11 proliferation.

The transforming potential of bFGF has not previously been investigated in human astrocytes. These data demonstrate that transfection of primary human fetal astrocyte cells with the sp-bFGF expression vector resulted in a transformed phenotype in vitro demonstrated by the loss of contact inhibition, GFAP irnmunoreactivity and anchorage independent growth. Loss of GFAP immunoreactivity supports the in vitro transformation because it is a pathological change consistent with increasing grade in human glial tumors (45). The mechanism of transformation of the sp-bFGF astrocytes remains unknown, but codd be linked to the extracellular release of bFGF and/or the expression of higher molecular weight bFGF isoforrns. Because the seaetory signal sequence is responsible for the extracellular export of prote&, the sp-bFGF transfected astrocytes rnay secrete bFGF in the culture medium, as shown previously (5, 15). The importance of this shift in the localization of bFGF fiom cell-associated to extracellular release, has previously been shown to correlate with increased tumorigenecity and angiogenesis in the development of fibrosarcoma (22). In addition, bFGF and a novel member of the FGF family, FGF-9, have been detected in the conditioned medium of human and rat glioma cells (31, 37, 47). Because mRNA for the bFGF receptor was detected in the W585 ce11 strain (Figure 3), and these cells were found to be responsive to increasing concentrations of exogenous bFGF (Figure 2) this suggests that the sp-bFGF transfected astrocytes could be transfomed by classical autocrine stimulation (49). Four structurally related genes that encode five distinct FGF receptors have been described (20, 21, 23, 30, 39, 44, some soluble (19, 25). The PCR primers utilized in this study were specific for the cytoplasmic domain of the FGFR-2 receptor (36) a region that is highly conserved in the FGF receptor gene family (21, 36). Although we have not identified the speufic FGF receptor present on the W585 cells, Our results indicate that one or more of these receptors is present, and functional. These results support previous work that bFGF signal transduction pathways may play a critical role in the phenotype of astrocytic tumors (34,40).

Basic FGF has been identified as a key factor in the biology of malignant gliomas (1,13,17,18,26,40,50,51). Based on the important role of bFGF in glioma cell biology, there have been several reports on the inhibition bFGF expression in gliomas (4, 16, 17, 35, 36, 40, 51). We now report a concentration-dependent decrease in glioma cell proliferation following the addition of phosphorothioated bFGF antisense oligodeoxynucleotides. The specïficity of the bFGF antisense oligodeoxynucleotide was indicated by: (a) inhibition of U-87 glioma ceil proliferation by antisense, not the sense oligodeoxynucleotide; (b) the amount of the 7.0 and 4.0 kB bFGF mRNA transcripts was reduced following exposure to the antisense oligodeoxynucleotide, not the sense oligodeoxynucleotide; and (c) bFGF protein quantitated by ELISA, was decreased following exposure to the antisense not the sense oligodeoxynucleotide.

Antisense oligodeoxynucleotides can downregulate gene expression by three mechanisms: (a) blocking the function of mRNA through steric hindrance; @) irreversible binding of the oligonucleotide to the target mRNA sequence blocking translation; or (c) activation of RNAse H due to the presence of RNA/DNA duplexes (9'43); our results suggest that one or al1 of these mechanisms may be responsible for the activiiy of the bFGF antisense oligodeoxynucleotides~ The decreased expression of the 7.0 and 4.0 kB bFGF mRNA transcripts (Figure 8), following a single administration of bFGF antisense oligodeoxynucleotides is translated into a 44% reduction of gIioma cell-associated bFGF protein (Figure 9). The suppression of U-87 glioma ce11 proliferation (Figure 7) following antisense exposure could be the result of this reduction in bFGF protein; excess, or aberrant forms of bFGF may otherwise be available for extracellular release, stimulating proliferation by a classical autocrine mechanism (47, 49) or for the activation of a putative intracellular bFGF receptor, stimuIating proliferation b y an intemal autocrine mechanism (5,41). This study provides direct evidence that U-87 glioma ceIl proliferation is dependent in part, on bFGF.

The findings presented support bFGF as an important factor in the biology of human gliomas. The transforming potential of aberrant bFGF expression, including higher molecular weight bFGF isoforms, and/or the extracellular release of bFGF in human astrocytes in vitro, suggests that bFGF may be involved in glial turnorigenesis. The suppression of glioma ce11 proliferation by antisense oligodeoxynucleotides suggests that bFGF remains important in established gliomas. These results suggest anti-bFGF strategies could be useful in the treatment of malignant human gliomas, and provide the impetus for the design of pharmacological and gene based therapies that target bFGF expression. Chapter V - Limitations Transfections To understand the roie of secreted bFGF isofoms on the biology of non-transformed human astrocytes, astrocytes were stably transfected with the signal-peptide-bFGF cDNA. After splitting the cells for selection, only the sp-bFGF transfected ce11 clones survived; cells were therefore selected on the basis of proliferation alone. Because the transfected cells also had a selectable marker, resistance to oubain, CO-transfected,the selection of clones would have been strengthened if they were chosen following exposure to oubain, demonsbating the cells were indeed stably transfected. Additionally, the clones were never characterized for secreted bFGF levels by ELISA or western blot. These experiments would have provided additional support for the central hypothesis of this dissertation, that aberrant expression or extracellular export of bFGF was an important event in the malignant phenotype of human gliornas. As in Chapter IV, it would have been informative to obtain stably transfected wild type-bFGF expressing astrocytes, to rule out that bFGF overexpression, rather than a switch to secreted bFGF was responsible for the in vitro transformation. Antisense Experiments Following exposure to bFGF antisense oligonucleotides, the established human glioma cell U-87 revealed decreased expression of bFGF mRNA, protein and proliferation, in cornparison to sense oligonudeotide exposure. Because antisense oligonucleotides can non-specifically inhibit the expression of proteins, the data would have been strengthened by the inclusion of additional experiments: (a) demonstrate that the bFGF antisense oligonudeotide does not alter the expression Ievels of other FGF famiIy members; or @) the addition of bFGF protein to antisense treated glioma ceus to demonstrate reversal of the inhi'bitory effects. CHAPTER VI

Conclusions The studies reported in this dissertation represent an integrated examination of the role of bFGF gene and protein expression in the malignant phenotype of human gliomas. The new findine3 presented in the preceding chapters have demonstrated that:

(1) The amount of bFGF protein is signhcantly elevated in tissue fragments of primary human gliomas, WHO grades 1-IV, in cornparison to non-neoplastic human brain tissue. This significant elevation is an early event in the progression of gliomas because it is detected in lower grade gliomas.

In contrast to non-neoplastic human astrocytes, human glioma cells: (i) release bFGF into serum-free conditioned medium; (ii) express higher molecular weight bFGF isoforms; and (iii) express decreased amounts of bFGFR-2 mRNA.

Cells transfected with a bFGF expression vector for a secreted form of bFGF are transformed, express higher molecular weight bFGF isoforms, and demonstrate a highly invasive phenotype in vitro, and in situ in the brain.

Transfection of non-transformed human astrocytes with a bFGF expression vector modified to produce a secreted form of bFGF resulted in the detection of higher molecular weight bFGF isoforms, and evidence of in vitro transformation of the astrocytes. This data suggests that higher molecular weight, or secreted isofoms of bFGF could be a transfonning factor for human astrocytes. Antisense bFGF oligodeoxynucleotides inhibit established human glioma cell proliferation, related to the decreased expression of bFGF protein and bFGF mRNA transcripts. This data suggests that bFGF expression is required for the continued growth of human glioma cells .

Higher molecular weight bFGF isoforms were detected in a series of established human glioma ceus, not non-trans formed fetal astrocytes (Chapter III). Examination of the cellular lysates for bFGF by western blot from a varieiy of experimental cells demonstrated a relationship between the expression of higher molecular mass bFGF and a more malignant phenotype. In Chapter IV, the more invasive bFGF transfected ce11 clone not only secreted bFGF, but was shown to express higher molecular weight bFGF isoforms. In Chapter V, transfected fetal astrocytes that exhibited a transformed phenotype, expressed higher molecular weight bFGF isoforms, in comparison to non- trasnfected astrocytes that expressed only the 18 kD bFGF protein. The important contribution of higher molecular weight bFGF isoforms to glioma cell growth was demonstrated in a previous study. Basic FGF antisense treatrnent of an established human glioma cell line caused a significant inhibition of glioma ce11 growth, and a selective decrease in the higher molecular weight bFGF isoforms (17). The precise role of the higher molecular weight bFGF isoforms remains unknown, but could be related to the extracellular release of bFGF. Recently, high molecular weight bFGF isoforms were shown to be released from cells, in comparison to the prototypical bFGF that remained cell associated (18). In addition, the higher molecular weight bFGF isoforms were show to have deaeased affinity for hepa~(18). The mechanisms that regulate the expression of higher molecular weight bFGF isoforms, and the novel mechanism for their release are unknown. The data presented in this dissertation suggest that this is an important topic for future studies examining the contribution of bFGF to glioma ce11 pathobiology.

The experiments conducted in this dissertation were focused on bFGF protein, however it is dear that bFGF can act indirectly by effectïng other growth factors known to be upregulated in human gliomas. Vascular endothelial ce11 growth factor (VEGF) has been shown to be an important regulator of angiogenesis (7,8,9,16)and vasogenic edema (12) assocîated with human gliomas. Recently several studies indicated that bFGF and VEGF are co-localized immunohistochemically in human gliomas (1) and that these factors act synergistically in the induction of angiogenesis (3,6). Indirect angiogenic proteins like platelet derived growth factor or transforming growth factor-1 are known to upregulate the expression of both VEGF and bFGF (2). Other studies have indicated that bFGF upregulates VEGF, suggesting VEGF expression is bFGF dependent (11J3). The experiments conducted in this dissertation support that bFGF gene and protein expression may exert direct effects on human gliomas, and there is accumulating evidence that bFGF may also exert biological effects indirectlyf regulating the expression of proteins known to be important in angiogenesis and glioma ce11 biology.

The data presented in this dissertation suggest that the upregdation of bFGF gene expression, and/or extracellular release of bFGF have transforming properties for human astrocytes. This data indicaies that the upregulation of bFGF gene and protein expression may represent an independent pathway for astrocytic transformation. Altematively glial cell transformation could result from upstream genetic alterations, that could exert their effeds through the upregulation of bFGF gene expression. One of the earliest genetic alterations of glioma progression is the mutation of the p53. tumor suppresser gene (10,15), a finding conelated with the detection of angiogenesis (15). Examination of the bFGF gene promoter revealed it was responsive to p53 protein (14); in the presence of the mutant-type p53 protein bFGF gene expression is stimulated (14), by contrast the wild-type p53 protein has been shown to inhibit bFGF gene expression (14). It is therefore possible that upstream genetic events could exert their effects through the regulation of bFGF gene and protein expression.

When taken together, the experiments reported in this dissertation provide the impetus for the continued studies on the mechanisms that regulate bFGF gene and protein expression in normal and malignant glial cells. The results reported in Chapters II-V provide evidence that identifies bFGF as a contribuüng factor to the malignant phenotype of human gliomas. These findings are sumrnarized diagramatically in Figure 1. The results of Chapter V confirm the important role of bFGF in a human glial model, and they suggest inhibition of bFGF gene and protein expression could represent a unique therapeutic objective for the management of human gliomas. Figure 1 Cellular mode1 for the involvement of bFGF in the malignant progression of human gliomas. Normal Lower Grade Glioblastorna

lncreasing Malignancy

LEGEND

18 kD bFGF protein High molecular weight bFGF isofoms e 92 kD gelatinase Contribution to Original Knowledge

We have quantitated the amount of bFGF protein in tissue fragments of primary human brain tumors. We have identified that bFGF protein levels are stable in both the snap frozen and archived frozen tissue samples. These data will allow for clinical correlative studies of bFGF with cellular proliferation, angiogenesis, invasion, and could be useful in identifying sub- populations of patients that may benefit from anti-growth factor strategies.

We have demonstrated the differential regdation of bFGF gene and protein expression between non-neoplastic human astrocytes and established human glioma ce11 lines. In contrast to the non-transformed human astrocytes, the glioma cells released bFGF extracellularly, and expressed high molecular weight bFGF isoforms.

We developed in vitro and in aioo models to study the result of transfection-mediated increased bFGF gene expression. Using these models, we identified that cells stably transfected with a bFGF expression vector for a secreted form of bFGF expressed high molecular weight bFGF isoforms, and were highly invasive and angiogenic. When implanted in the brain, these cells grew with an invasive phenotype. By contrast cells transfected with a wild-type bFGF expression vector were non-invasive in vitro and in sitri in the brain. These studies established that aberrant expression of bFGF, and/or the extracellular release of bFGF could be a molecular determinant of neoplastic invasion in the brain.

We identified bFGF as a transforming factor for human astrocytes. Cds transfected with a bFGF expression vector for a secreted form of bFGF expressed higher molecular weight bFGF isoforms, were transformed in vitro, demonstrating loss of contact inhibition, anchorage independent growth, and loss of glial fibrillary acidic protein immunoreactivity. These studies revealed that elevated bFGF levels, higher molecular weight bFGF isoforms, and/or the extracellular release of bFGF could be important early evenis in the transformation of human astrocytes.

We developed a rapid 96-well plate assay for the quantitation of the growth inhibitory properties of bFGF antisense oligodeoxynudeotides. Using this assay, we dernonstrated the significant concentration dependent inhibition of established human gliorna ceIl proliferation. This decrease in glioma cell proliferation was associated with the specific decrease in bFGF mRNA transcripts, and bFGF protein.

1 This section is a mandatory requirement of Ph.D. Theses submitted to the Faculsr of Graduate Studies and Research, McGill University, Montreai, Quebec, Canada. REFERENCES

Chapter I

Abraham, J.A., Whang, J.L., Tumolo, A., Friedman, J., Hjerrild, KA., Gospodarowiu, D., Fiddes, J.C. Human basic fibroblast growth factor: nudeotide sequence and genomic organization. EMBO Journal 5: 2523- 2528,1986.

Alvarez, J.A., Baird, A., Tatum, A., Daucher, J., Chorsky, R., Gonzalez, A.M., Stopa, E.G. Localization of basic fibroblast growth factor and vascular endothelial ce11 growth factor in human glial neoplasms. Modem Pathology 5: 303307,1992.

Baird, A., Klagsbrun, M. The fibroblast growth factor famüy. Cancer Celis 3: 239-243,1991.

Bejcek, B.E., Li, DY., Deuel, T.F. Transformation by v-sis occurs by an interna1 autoactivation mechanism. Science 245: 1496-1499, 1989.

Bikfalvi, A., Klein, S., Pintucci, G., Quarto, M., Mignatti, P., Rifkin, D.B. Differential modulation of ceIl phenotype by different molecular weight forms of basic fibroblast growth factor: possible intracellular signaling by the high rnolecular weight forms. Journal of Ce11 Biology, 129: 233-243, 1995.

Blei, F., Wilson, EL., Mignatti, P., Rifkin, D.B. Mechanism of action of angiostatic steroids: suppression of plasminogen activator activity via stimulation of plasminogen activator inhibitor synthesis. Journal of Cd Physiology 155: 568-578,1993. Brem S The role of vascular proliferation in the growth of brain tumors. Clinical Neurosurgery 23: 440-453,1976.

Brem, H., Klagsbrun, M. The role of fibroblast growth factors and related oncogenes in turnor growth. Cancer Treatrnent Research 63: 211-231,1992.

Brem, S., Coiran, R., Folkman, J. Tumor angiogenesis: A quantitative method for histologie grading. Journal of the National Cancer Institute 48: 347-356,1972.

10. Brem, S., Tsanaclis, A.M.C., Gately, S., Gross, J.L., Herblin, W.F. Immunolocalization of basic fibroblast growth factor to the microvasculature of human brain tumors. Cancer 70: 2673-2680,1992.

11. Bugler, B., Amalric, F., Prats, H. Alternative initiation of translation determines cytoplasmic of nuclear localization of basic fibroblast growth factor. Molecular and Cellular Biology 11: 573-577,1991.

12. Coulier, F., Batoz, M., Marks, L, Putative structure of the FGF6 gene product and role of the signal peptide. Oncogene 6: 1437-1444,1991.

13. de Ridder, L, Calliauw, L. Invasion of human brain tumors in vitro: relationship to dinical evolution. Journal of Neurosurgery 72: 589-593, 1990,

14. Delli Bovi, P., Curatola, A.M., Kern, F.G., An oncogene isolated by transfection of Kaposi's sarcoma DNA encodes a growth factor that is a member of the FGF family. CeIi 50: 729-737,1987. 15. Dickson, C., Acland, P., Smith, R., Dixon, M., Deed, R., MacAllan, D., Walther, W., Fuller-Pace, F., Kiefer, P., Peters, G. Characterization of in t- 2: a member of the fibroblast growth factor family. Journal of Cell Science (Suppl) 13: 87-96,1990.

16. Dickson, C., Smith, R., Brookes, S., Peters, G. Tumorigenesis by mouse mammary tumor virus: Proviral activation of a cellular gene in the common integraiion region int-2. CeIl 37: 529-536, 1984.

17.Dunbar, C.E., Browder, T.M., Abrams, J.S., Nienhuis, A.W. COOH- terminal-modified interleukin-3 is retained intracellularly and stimulates autocrine growth. Science 245: 1493-1496,1989 .

18. Engebraaten, O., Bjerkvig, R., Pedersen, P.H., Laem, O.D. Effects of EGF, bFGF, NGF, and PDGF (bb) on ce11 proliferative, migratory and invasive capacities of human brain-tumor biopsies in vitro. International Journal of Cancer 53: 209-214,1993.

19.Florkiewiu, R.Z., Majack, R.A., Buechler, R.D., Florkiewicz, E. Quantitative export of FGF-2 occurs through an alternative, energy- dependent, non-ER/golgi pathway. Journal of Cellular Physiology, 162: 388399,1995.

20. Florkiewiu, R.Z., Sommer, A. Hurnan basic fibroblast growth factor gene encodes four polypeptides: Three initiate translation from non-AUG codons. Proceedings of the National Academy of Sciences USA 86: 3978- 3981,1989. 21. Folkrnan, J., Klagsbrun, M. Angiogenic factors. Science 235: 442-447, 1987.

22. Gately, S., Soff, GA., Brem, S. The potential role of basic fibroblast growth factor in the transformation of cultured ~rimarvhurnan fetd astrocvtes L d 4 and the proliferation of human glioma (U-87) cells. Neurosurgery, 37: 723-

23. Gately, S., Tsanaclis, A.M.C., Takano, S., Klagsbrun, M., Brem, S. Cells transfected with the basic fibroblast growth factor gene hsed to a signal sequence are invasive in vitro and in situ in the brain. Neurosurgery, 36: 780-788,1995.

24. Gospodarowicz, D., Neufeld, G., Schweigerer, L. Fibroblast growth factor. Molecular and Cellular Endocrinology, 46: 187-204,1986.

25.Gross, J.L., Herblin, W.F., Eidsvoog, K., Horlick, R., Brem, S. Tumor growth regulation by modulation of basic fibroblast growth factor, in Steiner R, Weisz PB, Langer R. (eds): Angiogenesis: Key Principles - Science - Technology - Medicine. Switzerland: Birkhauser Verlag, 1992, pp 421-427.

26.Gross, J.L., Morrison, R.S., Eidsvoog, K., Herblin, W.F., Komblith, P.L., Dexter, D.L. Basic fibroblast growth factor: a potential autocrine regulator of human glioma ce11 growth. Journal of Neuroscience Research 27: 689- 696,1990.

27. Gruber, M.L., Hochberg, F. Systematic evaluation of ptimary brain tumors. Journal of Nuclear Medicine 31: 969-971,1990. 28. Hanninck, M., Donoghue, D.J. Requisement for a signal sequence in biologicd expression of the v-sis oncogene. Science 226: 1197-1199, 1984.

29. Kandel, J., Bossy-Wetzel, E., Radvanyi, F., Klagsbm, M., Folkman, J., Hanahan, D. Neovascularization is assouated with a switch to the export of bFGF in the multi-step development of fibrosarcoma. Cell, 66: 1095- 1104,1991.

30. Kimelman, D., Kirsduier, M.W. An antisense messenger RNA directs the covalent modification of the transcript encoding fibroblast growth factor in Xenopus oocytes. Cell, 59: 687-696,1989.

31. Kitagawa, Y., Ueda, M., Ando, N., Shinozawa, Y., Shimizu, N., Abe, O. Significance of int-2/hst-l coamplification as a prognostic factor in patients with esophageal squamous carcinoma. Cancer Research 51: 1504-1508,

32. Klagsbrun, M., Smith, S., Sullivan, R., Shing, Y., Davidson, S., Smith, J.A., Sasse, J. Multiple forms of basic fibroblast growth factor: Amino-terminal clervages by tumor cell- and brain cell-derived acid proteinases. Proceedings of the National Academy of Sciences USA û4: 1839-1843,1987.

33. Koda, T., Sasaki, A., Matsushima, S., Kakinuma, M. A transforming gene, hst, found in NM 3T3 cells transformed with DNA from three stomach cancers and a colon cancer. Japanese Journal of Cancer Research 78: 325- 328,1987.

34. Li, Y., Koga, M., Kasayama, S., Matsumoto, K., Arita, N., Hayakawa, T., Sato, B. Identification and characterization of high molecular weight foms of basic fibroblast growth factor in human pituitary adenornas. Journal of Clinical Endocrinology and Metabolism 75: 1436-1441,1992.

35. Lidereau, R, Callahan, R., Dickson, C., Peters, G., Escot, C., Ali, LU. Amplification of the int-2 gene in primary human breast tumors. Oncogene Research 2: 21-27/1987.

36. Lund-Johansen, M., Forsberg, K., Bjerkvig, R., Laerum, O.D. Effects of growth factors on a human gliorna ce11 Iine during invasion into rat brain aggregates in culture. Acta Neuropathalogica 84: 190497,1992.

37. Marics, I., Adelaide, J., Raybaud, F., Mattei, MG., Coulier, F., Planche, J., de Lapeyriere, O., Birnbaum, D. Characterization of the HST-related FGF-6 gene, a new member of the fibroblast growth gene family. Oncogene 4: 335-340,1989.

38. Mason, I.J. The ins and outs of fibroblast growth factors. Ce11,78: 547-552, 1994.

39. Megyesi, J.F., Klagsbrun, M., Folkrnan, J. Analysis of fibroblast growth factors produced by a human glioblastoma cd1 line: Evidence for multiple bFGF proteins. Journal of Cellula Biochemistry 15F: 224,1991 (Abstract #

40. Mignatti, P., Morimoto, T., Rifkin, D.B. Basic fibroblast growth factor, a protein devoid of secretory signal sequence, is released by ce& via a pathway independent of the endoplasmic reticulum-Golgi cornplex. Journal of Cellular Physiology 151: 81-93,1992.

41. Morrison, R.S. Suppression of basic fibroblast growth factor expression by antisense oligonucleotides inhibits the growth of transfonned astrocytes. Joumal of Biological Chemistry 266: 728-734,1991.

42. Morrison, R.S., Yamaguchi, F., Saya, H., Bruner, J.M., Yahanda, A.M. Donehower, L.A., Berger, M. Basic fibroblast growth factor and fibroblast growth factor receptor I are implicated in the growth of human astrocytomas. Journal of Neuro-Oncology 18:207-16,1994.

43. Morrison, R.S., Giordano, S., Yamaguchi, F., Hendrickson, S., Berger, M.S., Palczewski, K. Basic fibroblast growth factor expression is required for clonogenic growth of human gliorna cells. Journal of Neuroscience Research 34: 502-509,1993.

44. Morrison, R.S., Gross, J.L., Herblin, W.F. Reilly, T.M., LaSala, P.A., Alterman, R.L., Moskal, J.R., Kornblith, P.L., Dexter, D.L. Basic fibroblast growth factor-like activity and receptors are expressed in a human glioma cell line. Cancer Research 50: 2524-2529,1990.

45. Moscatelli, D., Joseph-Silverstein, J., Manejias, R., Rifkin, D.B. M, 25,000 heparin-binding protein from guinea pig brain is a high rnolecular weight form of basic fibroblast growth factor. Proceedings of the National Academy of Sciences USA û4: 57785782,1987.

46. Moscatelli, D., Rifiun, D.B. Membrane and matrix localization of proteinases: a common theme in turnor cell invasion and angiogenesis. Biochemica et Biophysica Acta 948: 67-85,1988.

47. Murphy, P.R. Sato, Y., Knee, R.S. Phosphorothioate antisense oligonucleotides against basic fibroblast growth factor inhibit anchorage- dependent and anchorage-independent growth of a malignant glioblastoma ce11 line. Molecular Endocrinology 6: 877-884,1992.

48. Murphy, P.R., Knee, R.S. Identification and characterization of an antisense RNA transcript (gfg) fiom the human basic fibroblast growth factor gene. Molecular Endocrinology, 8: 852-859,1994.

49. Paulus, W., Grothe, C., Sensenbrenner, M., Janet, T., Baur, L, Graf, M., Roggendorf, W. Localization of basic fibroblast growth factor, a mitogen and angiogenic factor, in human brain tumors. Acta Neuropathologica 79: 418423,1990.

50. Peters, G., Brookes, S., Smith, R., Placzek, M., Dickson, C. The mouse homolog of the hst/kFGF gene is adjacent to int-2 and is activated by proviral insertion in some virally induced mammary tumors. Proceedings of the National Academy of Sciences USA 86: 5678-5682,1989.

51. Powell, P., Klagsbrun, M. Three forms of rat basic fibroblast growth factor are made frorn a single mRNA and localize to the nucleus. Journal of Cellular Physiology 148: 202-210,1991.

52. Prats, AC., Vagner, S., Prats, H., Amalric, F. us-acting elements involved in the alternative translation initiation process of human basic fibroblast growth factor mRNA. Molecular and Cellular Biology, 12:47964805,1992. 53. Prats, H., Kaghad, M., Prats, A.C., Klagsbrun, M., Lelias, J.M., Liauzun, P., Chalon, P., Tauber, J.P. Amalric, F. Smith, J.A. High molecular mass foms of basic fibroblast growth factor are initiated by alternative CUG codons. Proceedings of the National Academy of Sciences USA 86: 1836- 1840,1989.

54. Prescott, D.M. Function of Genes. In, Cells - Principles of Molecular Structure and Function. Boston: Jones and Bartlett Publishers, 1988, pp. 250-293.

55. Quarto, N., Finger, FR,Rifkin, D.B. The NH2-terminal extension of high molecular weight bFGF is a nuclear targeting signal. Journal of Cellular Physiology 147: 311-318,1991.

56. Quarto, N., Talarico, D, Florkiewicz, R., Rifkin, D.B. Selective expression of high molecular weight basic fibroblast growth factor confers a unique phenotype to NM 3T3 cells. Cell Regulation 2: 699-708,1991.

57.Renk0, M., Quarto, N., Morimoto, T., Rifkin, D.B. Nuclear and cytoplasmic localization of different basic fibroblast growth factor species. Joumal of Cellular Physiology 144: 108-114,3990.

58. Rifkin, D.B., Moscatelli, D. Recent developments in the cell biology of basic fibroblast growth factor. Journal of CeII Biology 109: 106,1989.

59. Rogelj S, Weinberg RA, Fanning PA, Klagsbrun, M. Basic h'broblast growth factor fused to a signal peptide transforms cells. Nature 331: 173- 176,1988. 60. Rogelj, S., Weinberg, RA., Fanning, P., Klagsbrun, M. Characterization of tumors produced by signal peptide-basic fibroblast growth factor- transfonned cells. Journal of Cellular Biochemistry 39: 13-23,1989.

61. Kleihues, P., Burger P.C., Scheithauer, B.W. The new WHO classification of braui tumours. Brain Pathology 3:255-68,1993.

62. Sato, R., Rifkin, D.B. Autocrine activities of basic fibroblast growth factor: regulation of endothelial ce11 movement, plasminogen activator synthesis, and DNA synthesis. Journal of Ce11 Biology 107: 1199-1205,1988.

63. Schiffer, D. Patterns of Tumor Growth, in Salcman M (ed): Neurobiology of Brain Tumors. Baltimore: Williams and Wilkins, 1991, pp 85-135.

64. Sommer, A., Moscatelli, D., Rifkin, D.B. An amino-terminally extended and post-translationally modified form of 25 kD basic fibroblast growth factor. Biochemical and Biophysical Research Communication 160: 1267- 1274,1989.

65. Stefanik, D.F.,Rizkalla, L.R., Soi, A.S., Goldblatt, A., Rizkalla, W.M. Acidic and basic fibroblast growth factors are present in glioblastoma multiforme. Cancer Research 51: 5760-5765,1991.

66. Stoker, M., Gherardi, E. Regdation of cell movement: the rnotogenic cytokines. Biochemica et Biophysica Acta 1072: 81-1021991.

67.Takahashi, J.A., Fukumoto, M., Igarashi, K., Oda, Y., Kikuchi, H., Hatanaka, M. Correlation of basic fibroblast growth factor expression levels with the degree of malignancy and vascularity in human gliomas. Journal of Neurosurgery 76: 792-798,1992.

68. Takahashi, J.A., Mon, H., Fukumoto, M., Igarashi, K., Jaye, M., Oda, Y., Kikuchi, H., Hatanaka, M. Gene expression of fibroblast growth fador in human gliomas and meningiomas: Demonstration of cellular source of basic fibroblast growth fador mRNA and peptide in tumor tissues. Proceedings of the National Academy of Science USA 87: 5710-5714,1990.

69. Talarico, D., Basilico, C. The K-fgfFst oncogene induces transformation through an autocrine mechanism that requises extracellular s tirnula tion of the mitogenic pathway. Molecular and Cellular Biology 11: 1138-1145, 1991.

70. Taylor, W.R., Greenburg, A.H., Turley, E.A., Wright, J.A. Ce11 motility, invasion, and malignancy induced by overexpression of K-FGF or bFGF. Experimental Ce11 Research 204: 295-301,1993.

71. Theillet C, Le Roy, X., De Lapeyriere, O., Grosgeorges, J., Adnane, J., Raynaud, S.D.,Simony-Lafontaine, J., Goldfarb, M., Escot, C. Birnbaum D Amplification of FGF-related genes in human tumors: possible involvement of hst in breast carcinomas. Oncogene 4: 915-922,1989.

72. Tsanaclis, A.M.C., Robert, F., Michaud, J., Brem, S. The cycling pool of cells within human brain tumors: In situ cytokinetics using the monoclonal antibody Ki-67. Canadian Journal of Neurologic Sciences 18: 12-17,1991. 73. Volk, R., Koster, M., Poihg, A., Hartmann, L., bochel, W. An antisense transcript from the Xenopus laevis bFGF gene coding for an evolutionarily conserved 24 kD protein. EMBO Journal, 8: 2983-2988,1989.

74. Yamaguchi, F., Saya, H., Bruner, J.M., Morrison, R.Ç. Differential expression of two fibroblast growth factor-receptor genes is associated with malignant progression in human astrocytomas. Proceedings of the National Academy of Sciences USA. 91:4848,1994.

75. Yayon, A., Klagsbm, M. Autocrine regdation of ce11 growth and transformation by basic fibroblast growth factor. Cancer Metastasis Review 9: 191-202,1990,

76. Yoshida, T., Miyagawa, K., Odagiri, H., Sakamoto, H., Little, P.F., Terada, M., Sugimura, T. Genomic sequence of hst, a transforming gene encoding a protein homologous to fibroblast growth factors and the int-2 encoded protein. Proceedings of the National Academy of Sciences USA 84: 7305- 7309,1987.

77.Zagzag, D., Miller, D.C., Sato, Y., Rifkin, D.B., Burstein, D.E. Imrnunohistochemical localization of basic fibroblast growth factor in astrocytomas. Cancer Reseaxch 50: 7393-7398,1990.

78. Zhan, X., Bates, B., Hu, X., Goldfarb, M. The human FGF-5 oncogene encodes a novel protein related to fibroblast growth factors. MolecuIar and Cellular Biology 8: 3487-3495,1988.

79.Zhan, X., Culpepper, A., Reddy, M., Loveless, J. Goldfarb M. Human oncogenes detected by a defined medium culture assay. Oncogene 1: 369- 376,1987.

CHAPTER II

1. Brem, S., Cotran, R., Follanan, J. Tumor angiogenesis: A quantitative method for histologie grading. Journal of the National Cancer Institute 48: 347-356,1972.

2. Brem, S.S., Hafler, D.A., Van Uitert, R.L., Ruff, R.L., Reichert, W.H. Spinal subarachnoid hematoma: a hazard of lumbar puncture resulting in reversible paraplegia. New England Journal of Medicine. 304:1020-1021, 1981.

3. Brem, S.S., Zagzag, D., Tsanaclis, A.M.C., Gately, S., Elkouby, M.P., Brien, S.E. Inhibition of angiogenesis and tumor growth in the brain. Suppression of endothelial ce11 turnover by penicillamine and the depletion of copper, an angiogenic cofactor. American Journal of Pathoiogy 137 1121-1142,1990.

4. Brem, S., Tsanaclis, A.M.C., Gately, S., Gross, J.L., Herblin, W.F. Immunolocalization of basic fibroblast growth factor to the microvasculature of human brain tumors. Cancer 70: 2673-2680,1992.

5. Engebraaten, O., Bjerkvig, R., Pedersen, P.H., Laerum, O.D. Effects of EGF, bFGF, NGF, and PDGF (bb) on cell proliferative, migratory and invasive capacities of human brain tumor biopsies in vitro. International Journal of Cancer 53: 209-214,1993. 6. Finkelstein, S.D., Black, P., Nowak, T.P., Hand, CM., Christensen, S., Finch, P.W. Histological characteristics and expression of acidic and basic fibroblast gowth factor genes in intracerebral xenogenic transplants of human glioma cells. Neurosurgery, 34: 136-143,1994.

7. Folkman, J., magsbmn, M. Angiogenic factors. Science 235: 44247,1987.

8. Folkman, J. What is the evidence that tumors are angiogenesis dependent? Journal of the National Cancer Institute 82: 44,1990.

9. Fujimoto, K., Ichimori, Y., Kakizoe, T., Okajima, E., Sakamoto, H., Sugirnura, T., Terada, M. Increased semm levels of basic fibroblast growth factor in patients with renal cell carcinoma. Biochemical and Biophysical Research Communication 180: 386-392,1991.

10. Gately, S., Tsanaclis, A.M.C., Takano, S., Klagsbrun, M., Brem, S. Cells transfected with the basic fibroblast growth factor gene fused to a signal sequence are invasive in vitro and in situ in the brain. Neurosurgery 36:780-788,1995.

11. Gately, S., Soff, GA., Brem, S. The potential role of basic fibroblast growth factor in the transformation of cultured primary human fetal astrocytes and the proliferation of human glioma (U-87) ceus. Neurosurgery 37:723- 732,1995.

12. Gross, J.L., Herblin, W.F., Dusak, B.A., Czemiak, P., Diamond, M.D., Sun, T., Eidsvoog, K., Dexter, D.L., Yayon, A. Effectç of modulation of basic fibroblast growth factor on tumor growth in vivo. Journal of the National Cancer Institute 85: 121-131,1993.

13. Kandel, J., Bossy-Wetzel, E., Radvanyi, F., Klagsbrun, M., Folkman, J. and Hanahan, D., Neovascularization is associated with a switch to the export of bFGF in the multistep development of fibrosarcoma. Cell 66:1095-1104, 1991.

14. Li, V.W., Folkerth, R.D., Watanabe, H., Yu, C., Rupnick, M., Bames, P., Scott, R.M., Black, P.M., Sallan, S.E., Folkman, J. Microvessel count and cerebrospinal fluid basic fibroblast growth factor in children with brain tumors. Lancet 344: 82-86,1994.

15. Lund-Johansen, M., Forsberg, K., Bjerkvig, R., Laerum, O.D. Effects of growth factors on a hurnan glioma ce11 line during invasion into rat brain

aggregates in culture. Acta Neuropathologica 84: 190-197,1992.

16. Morrison, R.S., Yamaguchi, F., Bniner, J.M., Tang, M., McKeehan, W., Berger, M.S. Fibroblast growth factor receptor gene expression and immunoreactivity are elevated in human glioblastoma multiforme. Cancer Research 54:2794-2799,1994.

17.Murphyf P.R., Myal, Y., Sato, Y., Sato, R., West, M., Friesen, G. Elevated expression of basic fibroblast growth factor messenger ribonudeic acid in acoustic neuromas. Molecular Endocrinology 3: 225-231,1989.

18. Murphy, PR, Sato, R., Sato, Y., Friesen, H.C. Fibroblast growth factor messenger ribonucleic aad expression in a human astrocytoma cell line: regdation by senun and cell density. Molecular Endocrinology 2: 591-598, Nguyen, M., Watanabe, H., Budson, A.E., Richie, J.P., Hayes, D.F., Folkman J. Elevated levels of an angiogenic peptide, basic fibroblast growth factor, in the urine of patients with a wide spectrurn of cancers. Joumal of the National Cancer Institute 86: 356-361,1994.

Nguyen, M., Watanabe, H., Budson, A.E., Richie, J.P., Folkman, J. Elevated levels of the angiogenic peptide basic fibroblast growth factor in urine of bladder cancer patients. Journal of the National Cancer Institute 85: 241- 242,1993.

21. Paulus, W., Grothe, C., sensenbre~er,M., Janet, T., Baur, L, Graf, M., Roggendorf, W. Localization of basic fibroblast growth factor, a mitogen and angiogenic factor, in human brain tumors. Acta Neuropathologica. 79: 418423,1990.

22. Rogelj S, Weinberg RA, Fanning PA, Klagsbrun M: Basic fibroblast growth factor fused to a signal peptide transforms cells. Nature 331: 173476,1988.

23. Sambrook, J., Fritsch, R.F., Maniatis, R. Malecular Cloning. Cold Spring Harbor, N.Y. 1989.

24. Silverlight, JJ., Prysor-Jones, R.A., Jenkins, J.S. Basic fibroblast growth factor in human pituitary tumors. Clidcal Endocrinology 32: 669676, 1990.

25. Singh, R.K., Bucanna, CD., Gutman, M., Fan, D., Wilson, M.R., Fidler, IJ. Organ site-dependent expression of basic fibroblast growth factor in human renal ce11 cascinoma cells. American Journal of Pathology 145:365- 374,1994

26. Stefanik, D.F.,Rizkalla, L.R., Soi, A., Goldblatt, S.A., Rizkalla, W.M. Acidic and basic fibroblast growth factors are present in glioblastoma multiforme. Cancer Research 51: 5760-5765,1991.

27. Soutier, A.D., Nguyen, M., Watanabe, H. and Folkman, J. Basic fibroblast growth factor secreted by an animal tumor is detectable in urine. Cancer Research 53:5297-5299,1993.

28. Takahashi, J.A., Fukumoto, M., Igarashi, K., Oda, Y., Kikuchi, H., Hatanaka M. Correlation of basic fibroblast growth factor expression levels with the degree of malignancy and vascularity in human gliomas. Journal of Neurosurgery 76: 792-798,1992.

29. Takahashi, J.A., Mori, H., Fukumoto, M., Igarashi, K., Jaye, M., Oda, Y., Kikuchi, H., Hatanaka, M. Gene expression of fibroblast growth factors in human gliomas and rneningiomas: Demonstration of cellular source of basic fibroblast growth factor mRNA and peptide in tumor tissues. Proceedings of the National Academy of Science USA 87: 5710-5714,1990.

30. Takano, S., Gately, S., Neville, M.E., Herblin, W.F., Gross, J.L., Engelhard, H., Pemcone, M., Eidsvoog, K., Brern, S. Suramin, an anticancer and angiosuppressive agent, inhibits endothelial cell binding of basic fibroblast growth factor, migration, proliferation, and induction of urokinase-type plasminogen activator. Cancer Research 54: 26542660,1994. 31. Watanabe, H., Hori, A., Seno, M., Kozai, Y., Igarashi, K., Idumori, Y., Kondo, K. A sensitive enzyme immunoassay for human basic fibroblast growth factor. Biochemical and Biophysical Researdi Communication 175: 229-235,1991.

32.Zagzag, D., Miller, D.C., Sato, Y., Rifkin, D.B., Burstein, D.E. Irnmunohistochemical localization of basic fibroblast growth factor in astrocytomas. Cancer Research 50: 7393-7398,1990.

CHAPTER Ill

1. Albini, A., Iwamoto, Y., Kleinman, H.K., Martin, G.R., Aaronson, S.A., Kozlowski, J.M., McEwan, R.N. A rapid in vitro assay for quantitating the invasive potential of tumor cells. Cancer Research 47: 32393245,1987.

2. Alrnendral, J.M., Sommer, D., Macdonald-Bravo, H., Burchardt, J., Perera, J., Bravo, R. Complexity of the early genetic response to growth factors in mouse fibroblasts. Molecular and Cellular Biology 8: 2140-2148 1988.

3. Alvarez, J.A., Baird A., Tatum, A., Daucher, J., Chorsky, R., Gonzdez, AM., Stopa, E.G. Localization of basic fibroblast growth factor and vascular endothelial ce11 growth factor in human glial neoplasms. Modem Pathology 5: 303-307,1992.

4. Bikfalvi, A., Klein, S., Pintucci, G., Quarto, M., Mignatti, P., Rifkin, D.B. Differential modulation of cell phenotype by different molecular weight forms of basic fibroblast growth factor: possible intracellular signahg by the high molecular weight forms. Journal of Cell Biology, 129: 233-243, 1995. 5. Brem, S., Tsanaclis, A.M.C., Gately, S., Gross, J.L., Herblin, W.F. Immunolocalization of basic fibroblast growth factor to the microvasculature of human brain tumors. Cancer 70: 2673-2680 1992.

6. Brigstock, D.R., Klagsbrun, M., Sasse, J., Farber, P.A., Iberg, N. Species- spetific high molecular weight foms of basic fibroblast growth factor. Growth Factors 4: 45-52,1990.

7. Brigstock, D.R.,Sasse, J., Klagsbrun, M., Subcellular distribution of basic fibroblast growth factor in human hepatoma cells. Growth Factors 4: 189- 196,1991.

8. Bugler B, Amalric F, Prats H. Alternative initiation of translation determines cytoplasmic of nuclear localization of basic fibroblast growth factor. Molecular and Cellular Biology 11: 573-577 1991.

9. Chomczynski, P., Sacchi, N. Single-step method of RNA isolation by acid guanidinium thiocyanate-phenol-chloroform extraction. Analytical Biochemistry 162: 156-159 1987.

10. Cleveland, D.W., Yen, T.I. Multiple determinants of eukaryotic m RN A stability* New Biology 1: 121-126 1989.

11. Couderc, B., Prats, H., Bayard, F., Amaliic, F., Potential oncogenic effects of basic fibroblast growth factor requires cooperation between CUG and AUG- initiated fonns. Cell Regulation 2: 709-718 1991.

12. Engebraaten, O., Bjerkvig, R., Pedersen, P.H., Laerum, O.D. Effects of EGF, bFGF, NGF, and PDGF @b) on cell proliferative, migratory and invasive capacities of human brain-tumor biopsies in vitro. International Journal of Cancer 53,209-214 (1993).

13. Finkelstein, S.D., Black, P., Nowak, T.P., Hand, C.M., Christensen, S., Finch, P.W. Histological characteristics and expression of acidic and basic fibroblast growth factor genes in intracerebral xenogenic transplants of human glioma cells. Neurosurgery 34: 136-143 1994.

14. Florkiewicz, R., Z., Sommer, A. Human basic fibroblast growth factor gene encodes four polypeptides: Three initiate translation from non-AUG codons. Proceedings of the National Academy of Science USA 86: 397s 3981 1989,

15.Florkiewicz, R.Z., Majack, R.A., Buechler, R.D., Florkiewicz, E. Quantitative export of FGF-2 occurs through an alternative, energy- dependent, non-ER/golgi pathway. Journal of Cellular Physiology, 162: 388-399,1995.

16. Folkman, J., Klagsbnui, M., Angiogenic factors. Science 235: 442-447 1987.

17. Gately, S., Soff, GA., Brem, S. The potential role of basic fibroblast growth factor in the transformation of cultured primary human fetal astrocytes and the proliferation of human glioma (U-87) cells. Neurosurgery, 37: 723- 732,1995.

18. Gately, S., Tsanaclis, A.M.C., Takano, S., Klagsbm, M., Brem, S. Cells transfected with the basic fibroblast growth factor gene fused to a signal sequence are invasive in vitro and in situ in the brain. Neurosurgery 36:780-788,1995

19. Gross, J.L., Herblin, W.F.,Dusak, B.A., Czemiak, P., Diamond, M.D., Sun, T., Eidsvoog, K., Dexter, D.L., Yayon, A. Effects of modulation of basic fibroblast growth factor on tumor growth in vivo. Journal of the National Cancer Institute 85: 121-131 1993.

20. Gross, J.L., Morrison, R.S., Eidsvoog, K., Herblin, W.F., KombIith, P.L., Dexter, D.L. Basic fibroblast growth factor, a potential autocrine regulator of human glioma ce11 growth. Journal of Neuroscience Research 27: 689- 696 1990.

21.Kande1, J., Bossy-Wetzel, E., Radvanyi, F., Klagsbrun, M., Folkman, J., Hanahan, D. Neovascularization is associated with a switch to the export of bFGF in the multi-step development of fibrosarcoma. Cell, 66: 1095- 1104,1991.

22. Lau, L.F., Nathans, De Expression of a set of growth-related immediate early genes in BALB/c3T3 cells: Coordinate regulation with c-fos or c-myc. Proceedings of the National Academy of Science USA 84: 11824186 1987.

23. Lund-Johansen, M., Forsberg, K., Bjerkvig, R., Laenim, O.D. Effects of growth factors on a human glioma cell line during invasion into rat brain aggregates in culture. Acta Neuropathologica 84: 190-197 1992.

24. Morrison, R.S. Suppression of basic fibroblast growth factor expression by antisense oligo deoxyiudeotides inhibits the growth of transformed astrocytes. Journal of Biological Chemistry 15: 728-734 1991.

25. Morrison, R.S., Yamaguchi, F., Saya, H., Bruner, J.M., Yahanda, A.M. Donehower, L.A., Berger, M. Basic fibroblast growth factor and fibroblast growth factor receptor I are implicated in the growth of human astrocytomas. Journal of Neuro-Oncology 18:207-16,1994.

26. Morrison, R.S., Gross, J.L., Herblin, W.F., Reilly, T.M., Lasala, P.A., Alterman, R.L., Moskal, J.R., Kornblith, P.L., Dexter, D.L. Basic fibroblast growth factor-like activity and receptors are expressed in a human glioma ce11 line. Cancer Research 50: 2524-2529 1990.

27. Moscatelli, D., Joseph-Silverstein, J., Manejias, R., Rifkin, D.B. M, 25,000 heparin-binding protein from guinea pig brain is a high molecular weight form of basic fibroblast growth factor. Proceedings of the National Academy of Science USA 84: 5778-5782 1987.

28. Moscatelli, D., Presta, M., Joseph-Silverstein, J., Rifkin, D.B., Both normal and tumor cells produce basic fibroblast growth factor. Journal of Cellular Physiology, 129: 273-276 (1986).

29. Murphy, P.R., Sato, Y., Knee, R.S., Phosphorothioate antisense oligonucIeotides against basic fibroblast growth factor inhibit anchorage- dependent and anchorage-independent growth of a malignant glioblastoma cell Iine. Molecular Endocrinology 6: 877-8û41992.

30. Murphy, P.R., Sato, Y., Sato, R., Friesen, HG. Fibroblast growth factor messenger ribonucleic acid expression in a human astrocytoma cell line: regdation by serum and ce11 density. Molecular Endocrinology 2: 591-598

31. Murphy, P.R., Sato, Y., Sato, R., Friesen, HG. Regulation of multiple basic fibroblast growth factor messenger ribonucleic acid transcripts by protein kinase C activators. Molecular Endocrinology 2: 1196-1201 1988.

32. Nguyen, M., Watanabe, H., Budson, A.E., Richie, J.P., Folkman, J. Elevated levels of the angiogenic peptide basic fibroblast growth factor in urine of bladder cancer patients. Journal of the National Cancer Institute 85: 241- 242,1993.

33. Nguyen, M., Watanabe, H., Budson, A.E., Richie, J.P., Hayes, D.F., Folkman, J. Elevated levels of an angiogenic peptide, basic fibroblast growth factor, in the urine of patients with a wide spectrum of cancers, Journal of the National Cancer Institute 86: 356-361 1994.

34. Paulus, W., Grothe, C., Sensenbrenner, M., ane et, T., Baur, I., Graf, M., Roggendorf, W. Localization of basic fibroblast growth factor, a mitogen and angiogenic factor, in human brain tumors. Acta Neuropathologica 79: 418-423 (1990).

35. Powell, P.P., Klagsbnin, M. Regulation of basic fibroblast growth factor mRNA expression in rat C6 glioma cells. Experimental CelI Research 209: 224-230 1993.

36. Powell, PP., Klagsbrun, M. Three forms of rat basic fibroblast growth factor are made from a single mRNA and iocalize to the nucleus. Journal of Cellular Physiology 148: 202-210 1991.

37. Prats, H., Kaghad, M., Prats, A.C., Klagsbrun, M., Lelias, J.M., Liauzun, P., Chalon, P., Tauber, J.P., Amalric, F., Smith, J.A., Caput, D. High molecular mass forms of basic fibroblast growth factor are initiated by alternative CUG codons. Proceedings of the National Academy of Science USA 86: 1836-1840 1989.

38. Quarto, N., Finger, F.P., Rifkin, D.B. The NH2-terminal extension of high molecular weight bFGF is a nuclear targeting signal. Journal of Cellular Physiology 147: 311-318 (1991).

39. Quarto, N., Talarico, D., Florkiewicz, R., Rifkin, D.B. Selective expression of higher molecular weight basic fibroblast growth factor confers a unique phenotype to NM 3T3 cells. Ce11 Regulation 2: 699-708 1991.

40. Reilly, T.M., Taylor, D.S., Herblin, W .Fe, Thoolen, M.J., Chiu, A.T., Watson, D.W., Timmermans, P.B.M.W.M. Monoclonal antibodies directed against basic fibroblast growth factor which inhibit its biological activity in vitro and in vivo. Biochemical and Biophysical Research Communications 164: 736-743 1989.

41. Renko, M., Quarto, N., Morimoto, T., Rifkin, D.B. Nuclear and cytoplasmic localization of different basic fibroblast growth factor speaes. Journal of Ce11 Physiology 144: 108-114 1990.

42. Rifkin, D.B., Moscatelli, D. Recent developments in the ceU biology of basic fibroblast growth factor. Jounial of Cell Biology 109: 1-6 1989. 43. Rogelj, S., Weinberg, R.A., Fanning, P., Klagsbmn, M. Basic fibroblast growth factor hsed to a signal peptide transforms cells. Nature 331: 173- 175 1988.

44. Sambrook, J., Fritsch, R.F., Maniatis, R. Molecular Cloning. Cold Spring Harbor, N.Y. 1989.

45. Çato, Y., Murphy, FR., Sato, R., Friesen, H.G. Fibroblast growth factor release by bovine endothelial cells and human astrocytoma cells in culture is density dependent. Molecular Endocrinology 3: 744-748 1989.

46.Stefanik, D.F., Rizkalla, L.R., Soi, A.S., Goldblatî, S.A., Rizkalla, W.M. Acidic and basic fïbroblast growth factors are present in glioblastoma multiforme. Cancer Research 51: 5760-5765 1991.

47. Sternfeld, M.D., Hendrickson, J.E., Keeble, W.W., Rosenbaum, J.T., Robertson, J.E., Pittelkow, M.R., Shipley, D. Differential expression of mRNA coding for heparin-binding growth factor type 2 in hurnan cells. Journal of Ce11 Physiology 136: 297-304 1988.

48. Takahashi, J.A., Fukumo to, M., Igarashi K., Oda, Y., Kikuchi, H., Hatanaka, M. Correlation of basic fibroblast growth factor expression levels with the degree of malignancy and vascularity in human gliomas. Joumd of Neurosuigery 76: 792-798 1992.

49. Weich, H.A., Iberg, N., Klagsbrun, M., Folkman, J. Transcriptional regdation of basic fibroblast growth factor gene expression in capiilary endothelial cells. Journal of Cellular Biochemistry 47: 158-164 1991. 50. Yamaguchi, F., Saya, K., Bruner, J.M., Morrison, R.S. Differential expression of two fibroblast growth factor-receptor genes is associated with malignant progression in human astrocytomas. Proceedings of the National Academy of Sciences USA. 91:484-8,1994.

51. Yong, V.W., Kim, S.U., Pleasure, D. Growth factors for fetal and adult human astrocytes in culture. Brain Research 444: 59-66 1988.

52. Zagzag, D., Miller, D.C., Sato, Y., Rifkin, D.B., Burstein, D.E. Immunohistochemical localization of basic fibroblast growth factor in astrocytomas. Cancer Research 50: 7393-7398 1990.

Zuniga, A., Bo ja, M., Meijers, C., Zeller, R. Expression of alternatively spliced bFGF first coding exons and antisense mRNAs during chicken embryogenesis. Developmental Biology 157: 110-118 1993.

Chapter IV

Abraham, J.A., Whang, J.L., Tumolo, A., Friedman, J., Hjerrild, Gospodarowicz, D., Fiddes, J.C. Human basic fibroblast growth factor: nucleotide sequence and genornic organization. EMBO Journal 5: 2523- 2528,1986.

2. Albini, A., Iwamoto, Y., Kleinman, H.K., Martin, G.R., Aaronson, S.A., Kozlowski, J.M., McEwan, RN. A rapid in vitro assay for quantitating the invasive potential of tumor cells. Cancer Research 47: 3239-3245,1987.

3. Alvarez, J*Av Baird, A., Tatum, A., Daucher, J., Chorsky, R., Gonzaiez, AM., Stopa, E.G. Localization of basic fibroblast growth factor and vascular endothelial ce11 growth factor in human glial neoplasms . Modem Pathology 5: 303-307,1992.

4. Blei, F., Wilson, E.L., Mignatti, P., Rifkin, D.B. Mechanism of action of angiostatic steroids: suppression of plasminogen activator activity via stimulation of plasminogen activator inhibitor synthesis. Journal of Cell Physiology 155: 568-578,1993.

5. Brem, S., Tsanaclis, A.M.C., Gately, S., Gross, J.L., Herblin, W.F. Immunolocalization of basic fibroblast growth factor to the microvasculature of hurnan brain tumors. Cancer 70: 2673-2680, 1992.

6. Brem, S., Tsanaclis, A.M.C., Zagzag, D. Anticopper treatment inhibits pseudopodial protrusion and the invasive spread of 9L gliosarcoma cells in the rat brain. Neurosurgery 26: 391-396,1990.

7. Brem, S.S., Zagzag, D., Tsanaclis, A.M.C., Gately, S., Elkouby, M-P., Brien, S.E. Inhibition of angiogenesis and tumor growth in the brain. Suppression of endothelial ce11 turnover by penicillamine and the depletion of copper, an angiogenic cofactor. American Journal of Pathology 137: 1121-1142,1990.

8. Brien, S., Zagzag, D., Brem, S. Rapid in situ cellular kuietics of intracerebral tumor angiogenesis using a monoclonal antibody to bromodeoxyuridine. Neurosurgery 25: 715-719,1989.

9. de Ridder, L., Cailiauw, L. Invasion of human brain tumors in vitro:

relationship to clinical evolution. Joumal of Neurosurgery 72: 589-593, 10. Engebraaten, O., Bjerkvig, R., Pedersen, P.H., Laerum, O.D. Effeds of EGF, bFGF, NGF, and PDGF (bb) on ce11 proliferative, migratory and

invasive camcitiesI of human brain-tumor bio~siesl. in vitro. International Joumal of Cancer 53: 209-214, 1993.

Finkelstein, S.D., Black, P., Nowak, T.P., Hand, C.M., Christensen, Finch, P.W. Histological characteristics and expression of acidic and basic fibroblast growth factor genes in intracerebral xenogenic transplants of human glioma cells. Neurosurgery 34: 136443,1994.

12.Flaumenhaft, R., Moscatelli, D., Saksela, O., Rifkin, D.B. Role of extracellular mahk in the action of basic fibroblast growth factor: matrix as a source of growth factor for long-term stimulation of plasminogen activator production and DNA synthesis. Journal of Cellular Physiology 140: 75-81,1989.

13. Folkman, J., Klagsbrun, M. Angiogenic factors. Science 235: 442447,1987.

14.Gimbrone, M.A., Cotran, R.S., Leapman, S., Folkman, J. Tumor growth and neovascularization: an experimental model using the rabbit cornea. Joumal of the National Cancer Institute 52: 413-427,1974.

EGross, J.L., Herblui, W.F., Eidsvoog, K., Horlick, R., Brem, S. Tumor growth regulation by modulation of basic fibroblast growth factor, in Steiner R, Weisz PB, Langer R. (eds): Angiogenesis: Key Principles - Science - Technology - Medicine. Switzerland: Birkhauser Verlag, 1992, pp 16. Gross, J.L., Morrison, R.S., Eidsvoog, K., Herblin, W.F., Komblith, P.L., Dexter, D.L. Basic fibroblast growth factor: a potential autocrine regdator of human glioma ce11 growth. Journal of Neuroscience Research 27: 689- 696,1990.

17. Kwaan, H.C. The plasminogen-plasmin system in malignancy. Cancer Metastasis Review 11: 291-311,1992.

18. Laerum, O.D., Bjerkvig, R., Steinsvag, S.K., de Ridder, L. Invasiveness of primary brain tumors. Cancer Metastasis Review 3: 223-236,1984.

19. Landau, B.J., Kwaan, H.C., Verrusio, E., Brem, S.S. Elevated levels of urokinase-type plasminogen activator and plasminogen activator inhibitor type1 in malignant human brain tumors. Cancer Research 54: 1105-1108,1994.

20. Liotta. L.A. Mechanisms of cancer invasion and metastases, in Kaiser HE, Liotta LA. (eds): Cancer Growth and Progression: Influence of Tu mor Development on the Host. Dordrecht: Kluwer Acadenic Publishers, 1989,

21. Liotta, L.A., Steeg, P.S., Stetler-Stevenson, W.G. Cancer metastasis and angiogenesis: an imbalance of positive an(i negative regdation. Ce11 64: 327-336,199 1.

22. Lund-Johansen, M., Forsberg, K., Bjerkvig, R., Laenun, O.D. Effects of growth factors on a human glioma cell line during invasion into rat brain aggregates in culture. Acta Neuropathologica 84: 190-197,1992.

23. Mignatti, P.P., Rifkin, D.B. Biology and biochemistry of proteinases in tumor invasion. Physiological Review 73: 161495,1993

24. Morrison, R.S., Giordano, S., Yamaguchi, F., Hendrickson, S., Berger, M.S., Palczewski, K. Basic fibroblast growth factor expression is required for clonogenic growth of human gliorna cells. Journal of Neuroscience Research 34: 502-509,1993.

25. Morrison, R.S., Gross, J.L., Herblin, W.F., Reilly, T.M., LaÇala, P.A., Alterman, R.L., Moskal, J.R., Kornblith, P.L., Dexter, D.L. Basic fibroblast growth factor-like aciivity and receptors are expressed in a human glioma ce11 line. Cancer Research 50: 2524-2529,1990.

26. Morrison, R.S. Suppression of basic fibroblast growth factor expression by antisense oligonucleotides inhibits the growth of transformed astrocytes. Journal of Biological Chemistry 266: 728-734,1991.

27. Moscatelli, D, Rifkin, D.B. Membrane and matrix localization of proteinases: a common theme in tumor ce11 invasion and angiogenesis. Biochemica Biophysica Acta 948: 67-85,1988.

28. Moser, R.P. Surgery for gliorna relapse: Factors that influence a favorable outcorne. Cancer 62: 381-390,1988.

29.Nagan0, N., Sasaki, H., Aoyagi, M., Hirakawa, K. Invasion of experimental rat brain turnor: early morphological changes following microinjection of C6 glioma cells. Acta Neuropathologica 86: 117-25,1993.

30. Nicosia. R.F., Tchao, R., Leighton, J. Angiogenesis dependent tumor spread in reinforced fibrin clot culture. Cancer Research 43: 2159-2166, 1983.

31. Paulus, W., Grothe, C., Sensenbrenner, M., Janet, T., Baur, L, Graf, M., Roggendorf, W. Localization of basic fibroblast growth factor, a mitogen and angiogenic factor, in human brain turnors. Acta Neuropathologica 79: 41&423,1990.

32. Rao. J.S., Steck, P.A., Mohanam, S., Stetler-Stevenson, W.G., Liotta, L.A., Sawaya, R. Elevated levels of Mr 92,000 type IV collagenase in human

brain tumors. Cancer Research 53: 2208-2211,1993.

33. Rifkin, D.B., Moscatelli, D. Recent developments in the ce11 biology of basic fibroblast growth factor. Journal of Cell Biology 109: 1-6,1989.

34. Rogelj, S., Weinberg, R.A., Faming, P., Klagsbnin, M. Characterization of tumors produced by signal peptidebasic fibroblast growth factor- transfomed cells. Journal of CeII Biochemistry 39: 13-23,1989.

35. Rogelj, S., Weinberg, R.A., Fanning, P.A., Klagsbnin, M. Basic filroblast growth factor fused to a signal peptide transforms cells. Nature 331: 173- 176,1988.

36. Russell. D.S., Rubenstein, L.J. Pathology of Tumors of the Nervous System. Baltimore, Williams and Wilkins, 1989, ed 5, pp 83-289.

37. Sato, R., Rif',D.B. Autocrine activities of basic fibroblast growth factor: regulation of endothelial cell movement, plasminogen activator synthesis, and DNA synthesis. Jounial of Ce11 Biology 107: 1199-1205,1988.

38. Stefanik, D.F., Rizkalla, L.R., Soi, A.S., Goldblatt, A., Rizkalla, W.M. Acidic and basic fibroblast growth factors are present in glioblastoma multiforme. Cancer Research 51: 5760-5765,1991.

39. Stoker, M., Gherardi, E. Regdation of cell movement: the motogenic cytokines. Biochemica Biophysica Acta 1072: 81-102,1991.

40.Takahashi, J.A., Fukumoto, M., Igarashi, K., Oda, Y., Kikuchi, H., Hatanaka, M. Correlation of basic fibroblast growth factor expression levels with the degree of malignancy and vascularity in human gliomas. Journal of Neurosurgery 76: 792-798,1992.

41. Takahashi, J.A., Mori, K., Fukumoto, M., Igarashi, K., Jaye, M., Oda, Y., Kikuchi, H., Hatanaka, M. Gene expression of fibroblast growth factor in human gliomas and meningiomas: Demonstration of cellular source of basic fibroblast growth factor mRNA and peptide in tumor tissues. Proceedings of the National Academy of Science USA 87: 5710-5714,1990.

42. Takano, S., Gately, S., Neville, M.E., Herblin, W.F., Gross, J.L., Engelhard, H., Perricone, M., Eidsvoog, K., Brem, S. Suramin, an anticancer and angiosuppressive agent, inhibits endothelial ceIl binding of basic fibroblast growth factor, migration, proliferation, and induction of urokinase-type plasminogen activaior. Cancer Research 54: 26542660,1994.

43. Takano, S., Gately, S., Jiang, J.B., Brem, S. A diaminoanthraquinone inhibitor of angiogenesis. Journal of Pharmacology and Experimental Therapeutics 271: 10274033,1994.

44. Tamargo, R.J., Leong, K.W., Brem, H. Growth inhibition of the 9L glioma using polymers to release heparin and cortisone acetate. Journal of Neurooncology 9: 131-138,1990.

45. Vlodavsky, L, Folkman, J., Sullivan, R., Friedman, R., Ishai-Michaeli, R., Sasse, J., Klagsbrun, M. Endothelial cell-derived basic fibroblast growth factor; synthesis and deposition into subendothelial extracellular matrix. Proceedings of the National Academy of Science USA 84: 2292-2296,1987.

46. Wilkinson, H.A., Litofsky, N.S. Therapy of brain tumors. Contemporary Neurosurgery 16: 1-8,1994.

47. Yayon, A., Klagsbrun, M. Autocrine regulation of ce11 growth and transformation by basic fibroblast growth factor. Cancer Metastasis Review 9: 191-202,1990.

48. Zagzag, D., Brem, S., Robert, F. Neovascularization and tumor growth in the rabbit brain. A mode1 for experimental studies of angiogenesis and the blood-brain barrier. American Journal of Pathology 131: 361-371,1988.

49. Zagzag, D., Miller, D.C., Sato, Y., Rifkin, D.B., Burstein, D.E. Immunohistochemical localkation of basic fibroblast growth factor in astrocytornas. Cancer Research 50: 7393-7398,1990.

50. Zama, A., Tamura, M., houe, H.K. Three-dimensional obsemations on microvascular growth in rat glioma using a vascular casting method. Journal of Cancer Research and Clinical Oncology 117: 396-402,1991.

Chapter V

1. Abe, T., Okamura, K., Ono, M., Kohno, K., Mori, T., Hori, S., Kuwano, M. Induction of vascular endothelial tubular morphogenesis by human glioma cells. A mode1 system for tumor angiogenesis. Journal of Clinical Investigation 92: 54-61,1993.

2. Abraham, J.A., Whang, J.L., Tumolo, A., Friedman, J., Hjerrild, KA., Gospodarowicz, D., Fiddes, J.C. Human basic fibroblast growth factor: nucleotide sequence and genomic organization. EMBO Journal 5: 2523- 2528,1986.

Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J. Basic local alignrnent search tool. Journal of Molecular BioIogy 215: 403-410,1990.

Behl, C., Winkler, J., Bogdahn, U., Meüxensberger, J., Schligensiepen, K.H., Brysch, W. Autocrine growth regdation in neuroectodemal tumors as detected with oligodeoxyiucleotide antisense molecules. Neurosurgery 33: 679-684,1993.

Blam, S.B., Mitchell, R., Tischer, E., Rubin, J.S., Silva, M., Silver, S., Fiddes, J.C., Abraham, J.A., Aaronson, S.A. Addition of growth hormone secretion simal to basic fibrobIast mowth factor results in cell transformation and secretion of aberrant forms of the protein. Oncogene 3: 129-136,1988.

Brem, S., Cotran, R., Folkman, J. Tumor angiogenesis: A quantitative method for histologie grading. Journal of the National Cancer Institute 48: 347-356,1972.

Brem, S., Tsanaclis, A.M.C., Gross, J.L., Herblin, W.F. Immunolocalization of basic fïbroblast growth factor to the microvasculature of human brain tumors. Cancer 70:2673-2680, 1992.

Brem, S., Tsanaclis, A.M.C., Zagzag, D. Anticopper treatment inhibits pseudopodial protrusion and the invasive spread of 9L gliosarcoma cells in the rat brain. Neurosurgery 26: 391-396,1990.

Carter, G., Lemoine, N.R. Antisense technology for cancer therapy: does it make sense? British Journal of Cancer 67: 8694376,1993.

10. Chomczynski, P., Sacchi, N. Single-step method of RNA isolation by acid guanidi~um thiocyanate-phenol-chloroform extraction. Analytical Biochemistry 162: 156-159,1987.

11. de Ridder, L., Calliauw, L. Invasion of human brain tumors in vitro: relationship to chical evolution. Journal of Neurosurgery 72: 589593, 1990.

12. Engebraaten, O., Bjerlcvig, R., Pedersen, P.H., Laerum, O.D. Effects of EGF, bFGF, NGF, and PDGF (bb) on cell proliferative, migratory and invasive capacities of human brain tumor biopsies in vitro. International Journal of Cancer 53:209-214,1993.

13.Finkelstein, S.D., Black, P., Nowak, T.P., Hand, C.M., Christensen, S., Finch, P.W. Histological characteristics and expression of acidic and basic fibroblast growth factor genes in intracerebral xenogenic transplants of human glioma cells. Neurosurgery 34: 136-143 1994.

14. Folkman, J., Klagsbrun, M. Angiogenic factors. Science 235: 442-447,1987.

15.Gately, S., Tsanaclis, A.M.C., Takano, S., Klagsbm, M., Brem, S. Cells transfected with the basic fibroblast growth factor (bFGF) gene fused to a signal sequence are invasive in vitro and in sitzl in the brain. Neurosurgery 36: 7&788,1995.

16. Gross, J.L., Herblin, W.F., Dusak, B.A., Czerniak, P., Diamond, MD., Sun, T., Eidsvoog, K., Dexter, D.L., Yayon, A. Effects of modulation of basic fibroblast growth factor on tumor growth in vivo. Journal of the National Cancer Institute 85: 121-131 1993.

17. Gross, J.L., Herblin, W.F., Eidsvoog, K., Horiick, R., Brem, S.S. Tumor growth regulation by modulation of basic fibroblast growth factor. In, Angiogenesis, Stein R, Weisz PB, Haudensdiild C, Langer R (eds); Birkhauser Verlag, Basel, pp. 42147,1992-

18. Gross, J.L., Morrison, R.S., Eidsvoog, K, Herblin, W.F., Komblith, P.L., Dexter, D.L. Basic fibroblast growth factor: a potential autocrine regulator of human glioma cell growth. Journal of Neuroscience Research 27: 689- 19. Ha~eken,A., Ying, W., Ling, N., Baird, A. Identification of soluble forms of the fibroblast growth factor receptor in blood. Proceedings of the National Academy of Science USA: 91: 9170-9174,1994.

20. Houssaint, E., Blanquet, P.R., Champion-Amaue, P., Gesnel, M.C., Torriglia, A., Courtois, Y., Breathnach, R. Related fibroblast growth factor receptor genes exist in the human genome. Proceedings of the National Academy of Science USA 87: 8180-8184,1990.

21. Johnson, D.E., Lu, J., Chen, H., Werner, S., Williams, L.T. The human fïbroblast growth factor receptor genes - a common structural arrangement underlies the mechanisms for generating receptor forms that differ in their third immunoglobulin dornain. Molecular and Cellular Biology 11: 4627-2634,1991.

22. Kandel, J., Bossy-Wetzel, E., Radvanyi, F., Klagsbrun, M., Folkman, J.,

Hanahan, D. Neovascularization is assoaated with a switch to the export of bFGF in the multistep development of fibrosarcoma. Ce11 66: 10954204, 1991.

23. Keegan, K., Johnson, D.E., Williams, L.T., Hayman, M.J. Isolation of an additional member of the fibroblast growth factor receptor family, FGFR-3. Proceedings of the National Academy of Science USA 88: 1095-1099,1991.

24. Kent, R.B., Emanuel, J.R., Neriah, Y.B., Levenson, R., Housman, D.E. Ouabain resistance conferred by expression of the cDNA for a murine Na+, K+-ATPase a subunit. Science 237:901-903,1987.

25. Kishi, T., Yoshida, T., Terada, M. A soluble form of K-çam/FGFR2 protein in the culture medium of human gastnc cancer cells. Biochemical and Biophysical Research Communication 202: 1387-1394,1994.

26. Li, V.W., Folkerth, R.D., Watanabe, H., Yu, C., Rupnick, M., Bames, P., Scott, R.M., McL Black, P., Sallan, S.E., Folkman, J. Microvessel count and cerebrospinal fluid basic fibroblast growth factor in children with brain tumours. Lancet 344: 82-86,1994.

27. Lund-Johansen, M., Forsberg, K., Bjerkvig, R., Laerum, O.D. Effects of growth factors on a human glioma ce11 line during invasion into rat brain aggregates in culture. Acta Neuropathologica 84: 190497,1992.

28. Mignatti, P., Mazzieri, R., Rifkin, D.B. Expression of the urokinase receptor in vascular endothelial cells is stimulated by basic fibroblast gcowth factor. Journal of Ce11 Biology 113:1193-1201,1991.

29. Mignatti, P., Rifkin, D.B. Biology and biochemistry of proteinases in tumor invasion. Physiol Rev 73: 161-195, 1993.

30. Miki, T., Bottaro, D.P., Fleming, T.P., Smith, CL., Burgess, W.H., Chan, A.M.L., Aaronson, S.A. Determination of ligand-binding specificity by alternative splicing: two distinct growth factor receptors encoded by a single gene. Proceedings of the National Academy of Science USA 89: 246- 250,1992. 31. Miyarnoto, M., Namo, K., Seko, C., Matsumoto, S., Kondo, T., Kurokawa, T. Molecular doning of a novel cytokine cDNA encoding the ninth member of the fibroblast growth factor family, which has a unique secretion property. Molecuiar and Cellular Biology 13: 42514259,1993.

32. Morrison, R.S., Giordano, S., Yamaguchi, F., Hendrikson, S., Berger, M.S., Palczewski, K. Basic fibroblast growth factor expression is required for clonogenic growth of human glioma cells. Journal of Neuroscience Research 34: 502-509,1993.

33. Morrison, R.S., Gross, J.L., Herblin, W.F., Reilly, T.M., LaSala, P.A., Alteman, R.L., Moskal, J.R., Komblith, P.L., Dexter, D.L. Basic fibroblast growth factor-like activity and receptors are expressed in a human gliorna ce11 line. Cancer Research 50: 2524-2529,1990.

34. Morrison, R.S., Yamaguchi, F., Bruner, J.M., Tang, M., McKeehan, W., Berger, M.S. Fibroblast growth factor receptor gene expression and immunoreactivity are elevated in human glioblastoma mu1tiforme. Cancer Research 54: 27942799,1994.

35. Morrison, R.S. Suppression of basic fibroblast growth factor expression by antisense oligodeoxynucleotides inhibits the growth of transformed astrocytes. Journal of Biological Chemistry 15: 728-734,1991.

36. Murphy, P.R., Sato, Y., Knee, R.S. Phosphorothioate antisense oligonucleotides against basic fibroblast growth factor inhibit anchorage- dependent and anchorage-independent growth of a malignant glioblastoma ceU lhe. Molecular Endocnnology 6: 877-884,1992. 37. Naruo, K., Seko, C., Kuroshima, K., Matsutani, E., Sasada, R., Kondo, T., Kurokawa, T. Novel secretory heparin-binding factors from hum an glioma cells (glia-activating factors) involved in glial ceIi growth. Journal of Biological Chemistxy 268: 2857-2864,1993.

38. Nguyen, M., Watanabe, H., Budson, A.E., Richie, J.P., Hayes, D.F., Folkman J. Elevated levels of an angiogenic peptide, basic fibroblast growth factor, in the urine of patients with a wide spectnun of cancers. Journal of the National Cancer Institute 86: 3560361,1994.

39. Partanen, J., Makela, T.P., Eerola, E., Korhonen, J., Hirvonen, H., Claesson- Welsh, L., Alitalo, K. FGFR-4, a novel acidic fibroblast growth factor receptor with a distinct expression pattern. EMBO Journal 10: 1347-1354, 1991.

40. Redekop, G.J., Naus, C.C.G. Transfection with bFGF sense and antisense cDNA resulting in modification of malignant glioma growth. Journal of Neurosurgery 82: 83-90,1995.

41. Rogelj, S., Stem, D., Klagsbm, M. Construction and expression of transforming gene resulting from fusion of basic fibroblast growth factor gene with signal peptide sequence. Methods in Enzymology 198: 117-124, 1991.

Rogelj, S., Weinberg, R.A., Fannuig, P.A., Klagsbm, M. Basic fibroblast growth factor fused to a signal peptide trançforms cds. Nature 331: 173- 176,1988. 43. Rothenburg, M., Johnson, G., Laughlin, G., Green, L, Cradock, J., Sarver, N., Cohen, J.S. Oligonucleotides as antisense inhibitors of gene expression: therapeutic implications. Journal of the National Cancer Institute 81: 1539-1544,1989

44. Ruta, M., Burgess, W., Givol, D., Epstein, J., Neiger, N., Kaplow, J., Cmmley, G., Dionne, C., Jaye, M., Schlessinger, J. Receptor for acidic fibroblast growth factor is related to the tyrosine kinase encoded by the fms-like gene (KG). Proceedings of the National Academy of Science USA 86: 8722-88726,1989.

45. Rutka, J.T., Hubbard, S.L., Fukuyama, K., Matsuzawa, K., Dirs, P.B., Becker, L.E. Effects of antisense glial fibriliary acidic protein complementary DNA on the growth, invasion, and adhesion of human astrocytoma cells. Cancer Research 54: 326703272,1994.

46. Sambrook, J., Fritsch, R.F., Maniatis, R. Moleculnr Cloning. Cold Spring Harbor, N.Y. 1989.

47. Sato. Y., Murphy, P.R., Sato, R., Friesen, H.G. Basic fibroblast growth factor release from bovine comeal endothelial cells and human astrocytoma cells is ce11 density dependent. Molecular Endocrinology 3: 744748,1989.

48. Shing, Y. Biaffinity chromatography of fibroblast growth factors. Methods in Enzymology 198: 91-95,1991.

49. Spom, M.B., Roberts, A.B. Autocrine growth factors and cancer. Nature 313: 745-747,1985. 50. Takahashi, J.A., Fukumoto, M., Igarshi, K., Oda, Y., Kikuchi, H., Hatanaka, M. Correlation of basic fibroblast growth factor expression levels with the degree of malignancy and vascularity in human gliomas. Journal of Neurosurgery 76: 792-798,1992.

51. Takahashi, J.A., Fukumoto, M., Kozai, Y., Ito, N., Oda, Y., Kikudii, H., Hatanaka, M. Inhibition of ce11 growth and turnorigenesis of human glioblastoma cells by a neutralizing antibody against hurnan basic fibroblast growth factor. FEBS Letters 288: 65-71,1991.

52. Takano, S., Gately, S., Neville, M.E., Herblin, W.F., Gross, J.L., Engelhard, H., Perricone, M., Eidsvoog, K., Brem, S. Suramin, an anticancer and angiosuppressive agent, inhibits endothelial ce11 binding of basic fibroblast growth factor, migration, proliferation, and induction of urokinase-type plasminogen adivator. Cancer Research 54: 2654-2660,1994.

53. Tsanaclis, A.M.C., Brem, S., Gately, S., Skipper, H.M., Wang, E. Statin imrnunolocalization in human brain tumors. Cancer 68: 786-792, 1991.

54. Tsanaclis, A.M.C., Robert, F., Michaud, J., Brem, S. The cycling pool of cells within human brain tumors: In situ cytokinetics using the monoclonal antibody Ki-67. Canadian Journal of Neurological Sciences 18: 12-17,1991.

55. Yong, V.W., Kim, S.U.,Pleasure, D. Growth factors for fetal and adult human astrocytes in culture. Brain Research 444: 59066,1988. 1. Alvarez, J.A., Baird, A., Tatum, A., Daucher, Jv Chorsky, R., Gonzalez, A.M., Stopa, E.G. Localization of basic fibroblast growth factor and vascular endothelial growth factor in human glial neoplasms. Modern Pathology. 5:303-7,1992.

2. Brogi, E., Wu, T., Namiki A., Isner, J.M. Indirect angiogenic cytokines upregulate VEGF and bFGF gene expression in vascular smooth muscle cells, whereas hypoxia upregulates VEGF expression only. Circulation. 90549-52,1994.

3. Goto, F., Goto, K., Weindel, K., Folkman J. Synergistic effects of vascular endothelial growth factor and basic fibroblast growth factor on the proliferation and cord formation of bovine capillary endotheiial cells within collagen gels. Laboratory Investigation. 69:508-17,1993.

4. Louis, D.N. The p53 gene and protein in human brain tumors. Journal of Neuropathology & Experimental Neurology. 53:ll-21,1994.

5. Mikkelsen, T., Cairncross, J.G., Cavenee, W.K. Genetics of the malignant progression of astrocytoma. Journal of Cellular Biochemistry. 46:3-8,1991.

6. Pepper, M.S., Fenara, N., Orci, L., Montesano, R. Potent synergism between vascular endothelia1 growth factor and basic fibroblast growth factor in the induction of angiogenesis in vitro. Biochemical & Biophysical Research Communications. 189:82431,1992.

7. Plate, K.H., Breier, G., Weich, H.A., Risau, W. Vascular endothelial growth factor is a potential tumour angiogenesis factor in human gliomas in vivo. Nature. 359:845-8,1992.

8. Samoto, K., Ikezaki, K., Ono, M., Shono, T., Kohno, K., Kuwano, M., Fukui, M. Expression of vascular endothelial growth factor and its possible relation with neovascularization in human brain tumors. Cancer Research. 55:1189-93,1995.

9. Samoto, K., Ikezaki, K. Ono, M., Shono, T., Kohno, K., Kuwano, M., Fukui M. Expression of vascular endothelial growth factor and its possible relation with neovascularization in human brain tumors. Cancer Research. 55:1189-93,1995.

10. Sidransky, D., Mikkelsen, T., Schwechheimer, K., Rosenblum, M.L., Cavanee, W., Vogelstein, B. Clonal expansion of p53 mutant cells is associated with brain tumor progression. Nature. 3552346-847,1992.

11. Stavri, G.T., Zachary, LC., Baskerville, P.A., Martin, J.F., Erusalimsky, J.D. Basic fibroblast growth factor upregulates the expression of vascular endothelial growth factor in vascular smooth muscle celis. Synergistic interaction with hypoxia. Circulation. 92:ll-4,1995.

12.Strugar, J.G., Criscuolo, GR., Rothbart, D., Harrington, W.N. Vascular endothelial growth/permeability factor expression in human glioma specimens: correlation with vasogenic brain edema and tumor-associated cysts. Journal of Neurosurgery. 83:682-9,1995.

13. Tsai, J.C. Goldman, CX. Gillespie, GY. Vascular endothelial growth factor in human glioma cell lines: induced secretion by EGF, PDGF-BB, and bFGF. Journal of Neurosurgery. 82:864-73,1995.

14.Ueba, T. Nosaka, T., Takahashi, J.A., Shibata, F., FIorkiewicz, R.Z., Vogelstein, B., Oda, Y., Kikuchi, H., Hatanaka, M. Transcriptional regulation of basic fibroblast growth factor gene by p53 in human glioblastoma and hepatocellular carcinoma cells. Proceedings of the National Academy of Sciences USA. 91:9009-13,1994.

15. Van Meir, E.G., Polverini, P.J. Chazin, V.R., Su Huang, H.J., de Tribolet, N., Cavenee, W.K. Release of an inhibitor of angiogenesis upon induction of wild type expression in glioblastoma cells. Nature Genetics. 8:171-6, 1994.

16. Weindel, K., Moringlane, J.R., Manne, D., Weich, H.A. Detection and quantification of vascular endothelial growth factor/vascular permeability factor in brain tumor tissue and cyst fluid: the key to angiogenesis?. Neurosurgery. 35:439-448,1994.

17. Murphy, P.R., Sato, Y., Knee, R.S. Phosphorothioate antisense oligonucleotides against basic fibroblast growth factor inhibit anchorage- dependent and anchorage-independen growth of a malignan t glioblastoma ce11 line. Molecular Endocrinology 6: 877-884, 1992.

18. Hanahan, D., Christofori, G., Naik, P., Arbeit, J. Transgenic mouse models of tumor angiogenesis: the angiogenic switch, its molecular controls, and prospects for preclinical therapeutic models. European Journal of Cance~ 32A: 2386-2393,1996.

Reprintcd hmCANCER, VOL 70, No, LI, Dccrnik 1, 1997 Copyn'gbt @ 1992. by the Amerka Cancer Society, Inc I. B. tippincott Company. Rintcd in USA,

lmmunolocalization of Basic Fibroblast Growth Factor to the Microvasculature of Human Brain Tumors Steoen Brem, M.D.," Ana Maria C. Tsanaclis, MD., Ph.D.,*t Stephen Gately, BA.,* Janet L. Gross, Ph.D.,$ and William F. Hrrblin, Ph.D.$

Background. Microvascular proliferation, a promi- tumars (1.2 2 1.1%) than the 31 malignant turnors (10.3 nent feature of tumors of the central nervous system, is a 10.5Y0; P < 0.001). The bFGF was immunolocalized in prime brget for anti-cancer therapy. the tumor cell nuclei in 23 of 52 tumors (4446)and in the Methods. Because basic fibroblast growth factor cytoplasm of 8 of 52 (35%) tumors. fmmunostaining to (bFGF) plays a key role in the regulation of angiogenesis, bFGF was prominent in the microvascular endotherial surgical specimens from 52 human brain tumors were campartment in 84% of the malignant tumors and only examined by immunocytochernical s tudies wi th a mu- 52% of benign tumors (P < 0.01). Immunostaining was rine monoclonal antibody to bFGF* Sections from these not present after preabsorption of the antibody with pure tumors also were incubated with Ki-67 monoclonal anti- human recombinant bFGF* body to rneasure the growth fraction. Conclusions, The presence of bFGF predominantly Results. Imrnunostaining for bFGF was observed in within the tumor microvasculature indicates a cellular 45 of 32 (87%) neoplasrns, reacting with 97% of the malig- depot for this potent growth factor that mediates angio- nant brain tumors and 67% of benign tumors [P < 0.01). genesis and tumorigenesis. These data support a role for The nonreactive tumors were a medulloblastoma and 7 bFGF in the transition From the benign to the malignant of 21 (33%) benign, noninvasive, slow-growing neo- phenotype. Cancer 1992; ï0:2673-80. plasms (1 acoustic schwannoma, 3 meningiornas, 2 pitu- itary adenomas, and 1 cholesteatoma). The indices of pro- Key words: angiogenesis, brain tumors, endothelium, fi- liferation [Ki-67labeling) were lower for the 21 benign broblast growth factor, immunocytochemistry, monoclo- nal antibody.

From the 'Lady Davis fnstitute for Medicd Research. Sir Mor- timer B. Davis-Jewish Genenl Hospitat, and the Department of Neu- LLiaovascular proliferation is fundamen ta1 to tumor rology and Neurosurgery, McGill University School of Mediane, patho&ysiology and is linked to tumor growthtib2inva- Pulontréal, Québec; the Division of Neucosurgery, Northwestern Me- sivenes~,~vascular permeability,2' and rdemao5An- morial Hospital. Northwestem University of School of Medicine, Chicago, Diinois; and the $Medical Products Department, the DuPont giogenesis is assoâated with maiignant progression6*' Mexk Pharmaceuti~;~ICompany, Wilmington, Delaware. and is requinte for the sustained growth of a solid tu- Supported by grants hmthe Medical Research Coundof Can- mor? Amortg human neoplasrns, mafigrtant himors of ada and the Cdncer Research Society. tnc (to S.B., the recipient of a the central nervous system fom the highest degree of Chercheur Bounier Clinicien Award of the Fonds de Ia liecherthe du angiogene~is?~~~Neovascularization guides the dassin- Santé du Québec); the Mme Feliowship in tiver and Vascular Dis- ease (to S.G.); and an institutional gant fmm EL du Pont de Ne- cation," radiologie detecti~n,'~"and establishment of mours and Company, Inc progno~is~.~~of many central nervous system turnors. t Dr. Tsanaciis is a visiting saentist fmm the Department of Biochemical mediaton of tumor angiogenesis are Pathofogy. SZo Paulo University School of Medicine, SJo Paulo, rnembers of a family of potent, polypeptide, heparin- BraPI. binding, endothebal cell mitogens.' A growth factor re- The authors thank David Ivanac and Marguerite Wotoczek for th& skilied technial assistance and Francine Dotson for typing the ceivuig major interest is basic fibrobIast growth factor manuscript: the juvede brain tumors were suppiied by Dr- J. Mi- (bFGF), a povverhrl shulator of endothdial growth in chaud of the University of Montréal; and the bFGF antibodies were vitro and in vivo>'f'" This gmwth factor is capabIe of puided by Drs. T. Ray, J, Duke, and A. Nahnpetlan (DuPont stimulating the replication of multiple cdtypes, in- M& Pharmaceutical Company)- duding astrocyticH and gli~rna'~"' ceIIs. Cells tram- Address for reprints: Steven Bcem, MD., NeuronuginugicalOncoi- ogy, Northwestern Memonal HospitaL 233 Eâst Erie Street, Suite fected with bFGF, fised to a signal peptide, stimdate 500, Chicago. iL 606tI-2906. neoplastic transformation in vitroz and produce vascu- Accepted for pubkation May 5,1993 larized tumors in vivo? 2874 CANCER Decornber 2,1993, Volume 70, No. 11

Initial immunohistochemicd midies, using polydo- to be specüic for human recombinant bFCF? Both nd antisera, demonstrated restricted tissue distribution DG2 and DE6 showed high Wtyto bFGF and neu- of bFGF in vivo in basement membranes'43; develop- Wedthe mitogenic activity of bFGF in vitro." Both in? and embry0nic2~capillaries, but not in aduit capü- antibodies recognize the bFGF epitope in a varie. of lary endotheliumua; arteriai endothelium and suben- species and diverse cell types, induding rat and human dothelial matrÏx2B;and the rnargins of subacute brain glioma,LLand both antibodies are equally sensitive in wounds assoaated with capillari~ation~~-cowistent detecting bFGF wilh Western blot anaiysis."" The with the role of bFGF as a physiologie mediator of an- conhol antibody consisted of DG2 absorbed with an giogenesis. Despite the importance to tumor biology, excess concentration of human recombinant bFGF anti- there have been relatively few reports of bFGF in hu- gen (provided by S.mergen, Boulder, CO). man rnalignant tissues. Most of the fa& known about After they were ruised in phosphate-buffered sa- the possible role of bFGF in malignancy corne fkom line, sections were exposed to 0.3?40 hydrogen peroxide nonglioma models or tissues outside the central ner- in phosphate-butfered saline for 30 minutes to blodc vous system?* endogenous peroxidase. After a thorough rime in phos- The doning and production of purified recombi- phate-buffered saline, the avidin-biotin complex nant bFGF have led to the developrnent of high-afnnity me thod was used (Vectastain ABC Kit, MK4002, Vec- monocIona1 antibodies, DG2 and DE6, two neutraiking tor Laboatories, Burlingame, CA), consisting of prein- antibodies directed against bFGF31; the use of monodo- cubation with diluted horse çerum for 20 minutes, fol- nai antibodies allows one to avoid Limitations inherent lowed by incubation in a humidified chamber tvith the to polyclonal antisera, induding heterogeneity and vari- anti-bFGF antibody at 20 ~g/rnlfor 2 hours at 37°C. ability? The sections thrn were exposed to biotinytated horse The current study had the follotving objectives: (1) secondary antibody against mouse immunogIobulin G to demonstrate the in situ expression of the bFGF epi- and subsequentiy to avidin-bioan-peroxidase complex tope in human brain tumors; (1)to determine the distri- for 20 minutes. The sections were stained with EreshIy bution of bFGF within the respective neoplastic and prepared 3,3'-diaminobenziduie-cobalt diloride solu- microvascular ce11 populations; and (3) to correlate tion (Sigma Chemicak, St. Louis, MO) and counter- bFGF expression and the biologic aggressiveness of the stained with Kernechtrot stain. Sections of these tumon himor based on the histopathologic dassification and were incubated with the Ki-67 monodonal antibody, as measurement of the growth Fraction. desaibed previously, to determine the growth fraction of the tumop and were evaluated on an Aristopian Materials and Methods microscope (Leica, Inc., ~eirfield,iL) adapted for pho- tomicrography. Tissue Samp les Immunostaining for bFGF was scored with a semi- quantitative scale: 0.0, absent bFGF immunostaining; Fi-two human brain tumors were studied: 46 addt 0.3, Iporadic bFGF immunostaining; 1.O, faint and tumon prospectively and 6 juvenile tumors (medullo- patchy bFGF imrnunostaining; 1.0, moderate bFGF im- blastomas) retrospeaively from tissue froren at -BO0 C. munostaining; and 3.0, prominent bFGF immunostain- The spearnens were sectioned imrnediately after re- hg. The values were grouped according to histopatho- moval; one portion was fked in 10% buffered formalin logic diagnosis. for routine histopathologic examination, and the other was embedded in OTC compound (Mies SaenSc, Ce11 Culture Napendle, IL) and fiozen in isopentane suspended in üquid nîûogen at -150°C for 1 minute. Routine sec- Human Shi-19, rat RT3, and C6 cells were main- tions for histopathologic examination were shedwith tained, and preparations of cytosol and Western blot hematoxylin and eosin. Frozen tissue blocks were anaiysis of these samples were perfomçd as previously stored at -WC. Cryostat sections, sediy cut at a de~cn'bed?~ thickness of 7 Fm, were air-dried at nom temperature for 1 hour, then fked in methanol-acetone (LI) for 20 minutes at -2U°C- Resuits are expressed as the mean r standard dcviation. bFGF ImmzrnocytochemicaI Shdi~ The chi-square test was used to determine si@cant differences between malignant and berügn tumors in Two murine monoclonal immunoglobuün G1 antibod- tnmunostaining of bFGF and the fiequenq of micro- ies, designated DG2 and DE6, were show previously vascular staining. S tudent's t test was used to evaluate bFGF and Human Brain Tumors/Brm et al.

TabIe 1. Basic Fibmblast Growth Factor and Ki-67 Imrnunostaining in Human Centrai Nervous System Tumors Vd ~topatfio1ogiccikgnasis walI Nuderw CytopIa~m KibT(%) MaLignantt Malipant soma (n = 8) 1.88 r 0.35 0.63 c 0.88 0.06 2 0.18 7.87 IC: 4.61 GliobIastoma (n = 12) 1.42 2 0.90 0.63 r 0.56 0.25 e 0.38 6.52 .i- 2.71 Medulioblastoma (n = 6) 2.17 r 1.17 0.33 2 0.82 0.00 11.09 2 6.90 Ependymoma (n = 3) 1.33 2 0.58 0.00 0.00 1.76 + 1.07 Metastatic carchtama (n = 2) LOO 2 1.41 0.00 1.00 1 41.30 s 3.68 Penign Schwannoma (n = 7) 1.14 k 0.90 0.79 2 0.86 0.14 r O24 1.13 2 1.53 Meningioma (n = If) 0.82 2 0.87 0.32 a 0.25 0.05 c, 0.15 1.34 2 0.91 Pihritary adenoma (n = 2) 0.00 0.00 0.00 0.09 ChoIesteatoma (n = 1) 0.00 0.00 0.00 NA NA: not mila able. 5xntKI47 tabehg indices are fmm derence 32. t Chdication of araiignancy Y fmm ceference 8 1.

the Merences in proüferative rates between benign samples trom malignant glioma and glioblastoma, re- and malignant tumon. Differences with P values less spectively. nuclear immunostaining was hetemgeneous than 0.05 were considered statistically signincant. and relative. rare. In contrast to the generally unifom and cowpicuous staining of the miaovascular netivork, the expression of bFGF within the himor cells was less prominent and more heierogeneous (Fie. 1and 2 corn- The results of the bFGF immunost~gare reported in pared with Figs. 3 and 4). Staining of the tumor cell Table 1. Positive immunostaining for bFGF was ob- cytoplasm rvas observed in 8 of 52 (15%) bain tumors served in al1 but one maUgnant human brain tumor, (1 malignant glioma, 3 gIiobIastomas, 1metastatic carci- with the use of DG2. The imrnunoreactivit)t was more noma, 2 schwannomas, and 1 meningiorna). In 6 of 52 prominent in the maiignant tumors, with the probabü- tumon (12%), bFGF reactivity was observed in both the ity of bFGF immunostaining significantly higher for the tumor ceii nuclear and cytoplasmic compartments. maiignant tumon in cornparison with the benign tu- There was no sipificant difference between benign or mors (P < 0.01). The tumors that were nonreactive were malignant brain tumors with respect to cytoplasmic or a medulloblastoma with a low index of proliferation nuclear localiza tion, (2.5%) and benign, noninvasive, slow-growing neo- A,high growth fraction as detexminrd by Ki-67 la- plasms (one acoustic schwannoma, three meningiomas, belingi a marker of the cyding pool ol ~ells."~'is one hvo pituitary adenornas, and one cholesteatoma). measure of malignant potential. Growth fractions for In 84% (26 of 31) of the malignant brain tumors, malignant brain tumors ranged from 1.846 to 11.3%, immunoreactivity was localued predominantly to the and, as reported above, all but one (97%) were immuno- endotheiîôl ceU nuclei and the microvascular waU, as reactive for bFGF. In conûast, the Ki-67Iabeling indices shown by immunostaining of a meddoblastoma (Fie. of the ben@ brain tumors ranged hom 0.1% to 1.3%, 1, top). This was significantly greater than the immuno- and 7 of 21 (3396) of the tumors were nonreactive for reactivity of the ben@ neopIasms (52%, 11 of 21 tu- bFGF. lmmunostaining of vesse1 wdswas most promi- mors; P < 0.01). This immunoreactivity was speâfic for nent in the medu.iIoblastomas, foUowed by metastatic bFGF because, in the presence of excess human recom- carcinomas, tumors with the highest proMerative rate. bhant bFGF, immimmunoreactivity was not observed (Fig. The monoclonal antibody to human recombinant 1, bottom). In tissue derived from a metastatic carci- bFGF (DE6) recognized several fomof immunoreac- noma, the nonimmunoreactive neoplastic ce& of a tu- tive bFGF that diffexed in their apparent moledar morcord nurounded a microvesse1 that was immunore- weights, as demonstrated by Western blot analysis of active for bFGF (Tg. 2). cytosoüc samples prepared from cultured human and In the neoplastic ceUuIar cornpartment the tumor rat güomas (Fig. 5). Aithough the relative intensities of celi nudei were reactive for bFGF in 23 of 52 (44%) each band differed among the glioma exûacts. the ap- tumors. As shown in Figures 3 and 4, which represent parent molecular weights of the predominant immune- 2676 CANCER Deconber 1,1992, Volume 70, No. 11

Figure 1. immunostaining of a meduiioblsstom~with the mondonai antibody DG2 to human recombinant bFGF. (Top left) The microvesseis are outiined by the darWy stained immunoreactive materiai (original magrtifkation, XIOO). (Top right) Higher magnifintion shows that the miaovascuIar wall is highfy reactive for bFGF antibody (original mogniîktion. X4OO). (Bottom) Contrd siide showing disappeannce of inununoreactivity when the monodonal antibody is treated with u

Figure Zbletastatic cminorna, The normative neoplastic ceils of a Figure 3. htak'gnant giiorna. Heterogeneous nuciear tumar cord surround a aiicrovessei that is immunoreactive to bFCF. immunostaining to bfGF antiiody (otiguiat magrUâcation. X630). bFGF and Human Brain Tumors/Brem et al.

These studies show that the microvessels in situ in rna- lipant brain tumors contain bFGF, a potent cellular mitogen capable of stimulating both angiogenesis and tumorigenesis. These observations stem from the use of two well-characterized monodonal angbodies that rec- ognize the bFGF epitope?' The application O f monodo- na1 antibodies permits reproduabüity and standardka- tion not possible with polydonal antisera," comrnoniy used to study bFGF.2G2g The original paradip of a biochemical "factor" pmduced by a soüd neoplasm (himor angiogenesis fac- to$' has been rehed in recent year~'~~~to take into account new data gieaned mainly from ceU-ceU inter- actions in vitr~.'~*~'~-"The principal heparin-binding, tumor rnitogenic, angiogenic factor'9 is adcnowledged widely to be synonymous with ~FGFP*~"O Previous cy- tokinetic studies underscore that the nomaily quies- cent endotheliurn of the brain becomes rnitoticalIy ac- tive to keep Pace with the proliferation of the tumor C~US.~*~' A iink between tissue bFGF content and indices of ceii proliferation has not been estabiished pre~iously~~ but it is noteworthy that there was less bFGF expression arnong the ben@ tumors. The ciinical course of a pa- tient with a brain tumor is a function of the relative proportion of tumor ceils Lhat aie mitotically active." For example, astrocytomas with a labeling index of less than 1%manifest a slow growth a te and carry a favor- able prognosis in contrast to tumon with a higher prolif- erative rate? More significantly, bFGF expression ap- FÏgure S. Western btot analysis of intracetlulac ùnmunoreactive peaa to be a consistent feature of the rnalignant tumors, bFGF proteins from rat (C6, RT2) and human (SM-19) glioma indicating a link between bFGF and maiignant progres- iines, devdoped with the DE6 anti-bFCF immunog1obulin sion, A Iarge prospective series is necessary to assess the antibody, demonstrates several distinct forms of the bFGF protein propostic signincance of bFGF expression in patients among Fr varioru ceU iines. with brain tumors. 3 The localllation of bFGF within the tumors is of interest. Immunolocalization to the vasculature sup- ports the work of other~.~~~-~~Increased angiogenic capacity6 and vasdarizatiod in human breast tumors are iinked to dinicd aggressiveness. Immunoreadivity to bFGF also was observed in nudear and cytoplasmic compartments of himors cells. Distinct moIecdar weight forms of bFGF are derived from initiation of message at different codonsMRecent data suaest that the diverse molecular weight [omis of bFGF have sepa- rate intracduiar localizations (either nudear or cyto- p~a~xnic).'~ Because DG2 and DE6 detect multiple forms of immunoreactive bFGF Pig. S), the observation of himorcells in situ shoiving cytosolicor nuclearimmuno- reactivity niggests that @orna celis are expresshg one Figure 4, Giioblastoma. Rare bFGF-immunoreactive ceiis (arrows) or more of these forms. Paradoxicdy, onIy 4816 of tu- with staining OFnudei (original magnification, X630). mor ceIIs were found to express immunoreadive bFGF, 2678 CANCER December 2,1992, Volume 70, No. 11

althaugh previous rtudier rhowed high leveis of intra- human cancer, also may contribute to the complex an- cellular bFGF in cultured @orna cells." The variable giogenesis code. expression of bFGF by subpopuIatiow of himor cells After t\iis work was completed, a role of bFGF in (Figs. 3 and 4) may represent another facet of heteroge- brain -or angiogenesis was corroborated by other in- neity in brain tumor karyotype, antigeniaty, angiogen- vestigator~.~~~A direct cornparison of the data is not es&, and histologic dia~act&tics.'~*~~~~ possible, however, because of the differences in the an- The bFGF e.xists as either a stable, inactive molecule tibodies used. It is of interest that Zagzag et al. found or a rapidly degradable. bioactive rnolec~le.'~The ac- bFGF expression to be characteristic of maiignant but tive fonn of bFGF stimulates endothelid cell mitosis, not benign astrocytomas; the capülary endothelium and remodeling of basement membranes, was immunoreative withui and at the margin of malig- three major steps in the process of angiogenesis." Endo- nant astrocytomas but not in the bain distant to the theiiai tek synthesue and store bFGF within the ceU tumor* Takahashi et alamdemonstrated in situ expres- and its extracellular matri.., which in vivo includes the sion of messenger RNA for bFGF in human gliomas and basement membrane of the microvascuiar wakSOLa&- meningiomas. High levels of immunoreactive bFGF ing a signal peptide, bFGF is not normally a secreted have been demowtated in glioma cdlines. in contrat mitogen, but rather remains sequestered In a function- to other solid turnors." aily inactive ~tate,f'~~explainhg the observation that, Evidence of in situ storage of a speafic angiogenic even in highly vasdar bain tumors, the endotheiium molecule in human brain himors provides additional is rnainiy quiescent and n~n~roliferative?~~Local stim- impetus to the developmrnt of biologic response modi- uli to initiate angiogenesis lead to the production by fies that target rnicrovascular proliferaticin for anti- endothelial cells of proteases," which are abundant in cancer therapy. Antagonists of positive autocrine human brak turn~rs.'*~and capable of degrading the growth factors can inhibit the growth of cancer cells in endothelial celi membrane, releasing active bFGP3 experimrntal animal^.^' Antagonists of bFGF are and thereby stimulating the production of more pro- emetging that can interfere with the receptorfn inacti- tease." The enzyme-growth factor interaction could be vate the genç,'* prevent the mitogenic effect," or bind higgered by a genetir or epigenetic event (e.g., local directly to the protein?*"*7 What would the effect of injury resulting korn hypoxia, ischemia, necrosis, or ex- anti-bFGF therapy be in situ? An uisight cornes €rom powe to radiation)?' recent animal experiments in which the host tissues The ma& cornponents in the basement membrane were depleted of copper, a cofactor of angiogenesis that of the endothelium that bind bFGF also can release the has remarkable affinity to bFGF." Treated animais rnitogen locaiiy for presentation to the adjacent tumor showed an inhibition of angiogenesis, tumor growth, containing receptor to bFGF," thereby stimu- and neopla~tics~read."The current article, demonstnt- lating tumorigenesis in a paracrine mannrr. Surround- ing bFGF ivithin the tumor microvasculature, provides ing a tumor capillary, the h-ction of proiiferating tumor a molecular targe t for adjuvant therap y by angioscp- ceUs decreases with increasing distance hmthe capil- pression for maügnant tumors of the cenûai nervous lary,"' possibly represenüng a diffusion gradient of 5ystèm.".~' bFGF. e~plainingthe topographie organization of exper- brain imental and human tumor celis around prolifeat- References hg capillaries. Irnmunoreactivity within the tumor cell nudei or cytoplasm of 31 of 52 bain himors is consis- tent with a positive autocrine rnechani~m~~in certain Brem S. ïhiz rote of vascufar proliferation in the growth of brain turnors. Uitr Yrturosug 1976; U:440-53. brain tumor dones. For tumors unable to produce Zagzag D, Brern 5, Robert F. Neovdxuliuization and hmor bFGF, local "host" bFGF could be mobilited by the growth in the brain: a mode1 for etperimental studie of angio- invading tumor tek." genesis and btood-bnin barries. Am / Parltof 1988; ISI:361-72. Neither bFGF nor any other trophic factor acts Rifkin DB, Moscatelli D, Cross J, Jaffe E. Proteases, angiogene- sis, and invasion, ln: Nicoison CL, Mias t, editon. Cancer inva- alone in isolation fiom O rher cytokinesb2;ulhatel y, the sion and metastasis: bioiogic and thenpeutic aspects. New York in vivo response of a re3p.iatory peptide such as bFGF is Uaven Press, 198.2:tSï-200. the combined rdtof multiple regdatory grotvth lac- Hauser tS, Miller FN. Differential rn~cromotecu1i~Iraicage from tors working in concert? Addidonal angiogenic mole- the vascuiature of tumors. Cancer 1986; Si:4614. cules (e.g., epidermal growth factor, transfomring Del Maestro RF, hfegyesi JF, FmeU CL. Mechanisms of tumor- Ssi 1990; LZ177-83. growth factor-alpha, pIa teiet-denved growth factor, associateci edema: a Ieview. Can 1 Xeumf Brem SS, Jensen HM, Guiiino Pb1 bgiogenesis as s market of and tenin) each have been desaibed as mediators of preneaptsstic lesions of the human break Ctacer t978;41239- neovdarization in human brain tumor~.-~ Newly 44. dkcovered angiogenic-- rnolec~les~'~'asyet untes ted in Weidner N, Sempfe jP, Wdch WR, FoIkman JI 'Tumucangiogert- bFGF and Human Brain Tumors/Brem et al. 2679

esis and metastasis: correIation in invasive breast caninorna. rV sic fibroblast growth factor in bovine adrenal giand, ovary, and Engf f Med 199t; 324A-8. pituitary. / Histocftem Cytochern 1989; 3731877433, Folkman J, Kiagsbrun U Angiogenic factors. Scimce 1987; CasxeUs W, Speir E. Sasse J, Klagsbrun M, Men P, Lee M. et al, 235:M2-7. isolation, characterization. and lociiiization of heparin-bliding Feigin 1, Men LB, Lipkin L Gross SW. The endothelid hyper- growth factors in the heart. J Clin lnvest 1990; 8543341. piasia of the cerebral blood vesseh with braui hunOE3, and its FdteinS, Apostoiides PJ, Caday CG, Prosset J. Phiiiips MF, saccornatous transformation. Cancer 1958; 1 l:26+T. Klagsbrun M. inaeased basic fibroblast growth factor (bFGF) Brexrt S, Cotran R. Folkman J. Tutor angiogenesis: a quantita- immunoreactivity at the site of focai wounds. Brain Res 1988; tive method for histologieal grading. / Nad Cancer Inst 1972; 460:253-9, 48:347-56. WestpM M. Herrmann H-D. Gmwth factor biology and onco- Daumas-Duport C, Scheithauer 0, O'Failon J, KeUy P. Grading gene activation in human @ornas and theh implications for spe- of asbwytomas: a simple and reproduâble rnethod Gncer afic therapeutic concepts. ?@umsorgery 1989; 25681-94. 1988; 62.215245. RdyTM, Taylor DS, Herbün WF, ThooIen MJ,Chiu AT, Wat- Kurnar PP, Cood M. Jona EO, Skultety FM, Leibrodc LG, son DW, et al. Monodona1 antibodies directed agsinst basic fi- McCornb RD. Conûast-enhanchg compukd tomopphy ring broblsst grawth factor which inhibit its bioIagiagidativity in zri- in güoblastoma multiforme after intraoperzttive endocuriethet- fm and in oiuo. Biochern Biophys Res Commun 1989; 16k736-43. apy. Cancer 1988; 61:1759-65. Tsana& AMC, Robert F, Michaud J, Brem S. The cyciing pool Zagzag D, Goldenberg M, Brem S. Angiogenesk and blood- of ceIIs within human brain tumots: in situ cytokinetics using the brain barrier breakdown modulate CT contnst enhancement: monudonal antibody Ki-67. Can 1Neural Sci 1991; 18:12-7. an experimental study in a rabbit brain tunior model. Am Gros JL, Behrens DL, MuhDE, Kornblith PL, Dater DL. 1989: 10529-34. Piasminogen activator and inhibitor ativity in human giioma Germano bt, Ito M, Cho KG, Hoshino T, Davis RL, Wion CB. celis and modulation by sodium butpte. Cmcer Rzs 1988: Correlation of histopathologicai features and proüferative po- 48:29 1-6. tential of gliomas. J Nrurosurg 1989; 70:701-6. Tsanadi .WC. Brem SS, Gately S, Schipper HM, Wang E, Sta- Schweigerer L. Neufeld G, Friedman J, Abraham JA, Fiddes JC, h'n immunolocaiiwtion in human bnin tuutors: detection of GuspodarowiczD. Capiflasr endothebat celk express basic fibro- noncyciing ce[& using a novel marker of ce11 quiescence. Cancer biast growth factor, a mitogen that promotes thdr own growth. 1991; 68:786-92. Nature 1987; 325257-9. Fuchs HE, Zalutsky MR. Bigner SH, Bigner DO. Monoclonal Cospodarowia D, Ferran N, Schweigerer L, Neufeld G. Stmc- antibodies: thair applications in the dirignosis and management tun1 chancterizrition and biologicai functions of Cibroblast of bnin tumors. In: Wilkins RH, Rengachary SS, editors. Xeuro- growth factor. Emfocr Rm f 982 8:gS- 114. nugery update. vol, 1. New York: McGnw-Hiil, 1990:152-9. Rifkin DB,MoscatelIi D. Recent developments in the ce11 biology Folkman J. Tumor ongiogenesis thecapeutic implications. N of basic fibroblast growth factor. / Ce12 Bi01 1989; 109:l-6. Engf 1Med 1971; 1851182-6. Baird A, Walicke PA. Fibroblast gmwth factors. Br rMrd Bull Kiagsbrun M, Vlodavsky 1. Biosynthesis and storage of basic 1989; 453.238-52. Gbroblast growth factor (bFGF) by endothelia1 cells: implication Kniss DA, Burry RW. Semm and fibroblast growth factor stimu- for the mechanism OC action of angiogenesis. Pmg Ciirt Biol Rrs late quiescent astrocytes to ce-enter the ce11 cycle. Bmin Rcts 1988; 266:53-61. 1988; 439:lS 1-8. Vlodavsky 1, Fotkman j, Sullivan R, Freidman R, [shai-Xlichaeü Morrison RS. Cross JL, Herblin WF, Rdly TM. LaSara PA. Al- R, Sasse J. et aI. Endothelia1 cell-derived basic fibroblast growth tman PL. et JI. Basic Bbroblast growth hctor-üke activity and factor, synthesis and deposition into subendotheliai irxmcelIu- receptors are expressed in a human giioma ceIl üne. Cancer Res Iar matrix. Froc Xatl Acad Sci Li SA 1987; 84:2292-6. 1990; 302524-9. Shing Y, Folkmiin 1, Sullivan R, Butterfield C Murray J, KIags- Gross JL. bIonisan RS, Eidsvoog K, Herbiin W. Komblith PL, bqM. Heparin affinity: purikation of a tumor-derived capil- Dexter DL. Basic 6bmblast growth factot: a potential autoaine r;iryCendotheLiaiceIl growth factor. Science 1984; 2231296-8. reguIator of human @orna ce11 growth. J Nzurosci Rts 1990 Nagashima T, Hoshino T,Cho KG. Pmliferjtive potentiaI of 27:689-96. vascular componenk in human giiabiastoma multiforme- Acta Rogeij S, Weinberg RA, Fanning P, Klagsbmn hl. Basic 6bm- Nmrupathol @cri) 1987; 73:3O 1-5. bbt pwth factor fused to a signal peptide traruforms cells. Bnen SE, Zagzag D, Brem S. Rapid in situ cellular kinetics of Naturc 1988; 331:173-5. intracerebrai tumor angiogenesis using a monodonal antibody Rogelj S, Weinberg RA. Fanning P, Klagsbmn M. Characterira- to bmmodeoxyuridine- Nrurosurgirry 1989; t5:715-9. tion of tumors produced by signal peptide-basic fibroblast Rosenblum ML Berens ME. Rutka JT.Recent perspectives in growth factor-transformed celis / Cdt Biochm 1989; 39:13-23. brain turnor bioIogy and treatrnent. Clil1 lVcurosrrrg 1989; Folkman J, Kiagsbrun hl, Sasse 1, Wadzinski hf, fngber D, Vio- 35 :3 14-35. dnvsky 1. A heparin-binding angiogenic protein-basic fibro- Hoshino T, Rodriguez LA, Cho KG, Lee Kç, Wion CB, Ed- bIast growth factor-is stored within basemen t membrane. Am J wards MSB, et ai. Propostic impücations of the prolifera tive Pathof 1988; 130:393-400. potential of Iow-grade astrocytomas. Nrurosug 1988; 69:839- Hanneken A, Lutty GA, McLeod DS, Robey F, Harvey AK, Hjd- 42 meland LM. Localkation of basic fibtoblast growth factor to the Dickon C. Fuller-Pace F, Kiefer P, Adand P, Mac-Ulen D, deveioping capillaries of the bovine retùia- / GIf Physiol 1989; Peters G. Expression, pmcessing and properties of int-7 Attn LW 138:115-20, Acad Sci 1991; 638:18-26, Ganr;tIez A-hl, Buscaglia LMI Ong M, Baird A- Dismiution of Bugfer B, Am;iInc F, Prats H, Alternative initiation of translation basic fibrobtast pwth factor in the 18-&y rat fetus: Iouliza- determines cytopIasmic or nudear lacaikation of basic fibro- tion in the basement membrane of diverse tissues. Cell Biol biast growth kctor. Mol Ceil Biol 199 1; 11573-7. 1990; 110:753-65. Bignst DD. Biohgy of gliomas: po tentiai dkUcaI impiiations of Grothe C, tfnsidcer KI fmmun~cytoche~caflocaiiition of ba- glioma cellular heterogeneity*Nmmsurgery 1981; 9520-6- .-C.. -. -. 1. 2680 CANCER December 1,1992, Volume 70, No. 11

47. Pauius W. ~~erJ. ïntranimoni hiltoiogic he

Stephen Cately, B.A., Gerald A. Soff, M.D., Steven hem, M.D. Division oi Xriiroloçic.iI Srirçcrv iSG, SB) .incl Hern,itologvÎC)nc~ologviC.ASi. Northwestern University School oi hkdieine, Chicago, Illinois, and the Department of Neurology and Neurosurgery rSGI. ~LtcCilIUniversity School oi Medicine. ,\\ontre,il. Qrirf)t>c. C~n.1cl.1

BASIC FlBROBLAST CROWTH factor (bFCF) is a potent stimulator of angiogenesis, proliferation, and invasion in human gliomas. To test the hypothesis that bFGF is important in the development of the malignant phenotype of human gliomas, bFGF expression was prospectively modulated in primary human fetal astrocytes and in an established human glioma cell line. Fetal astrocytes were transfected with a vector expressing bFGF modified by the addition of a secretory signal peptide sequence. Two of these bFCF astrocyte clones examined in vitro demonstrated anchorage-independent growth, loss of contact inhibition, and decreased glial fibrillary acidic protein immunore- activity, changes consistent with cellular transformation. To analyze the inhibition of bFGF expression, phosphoro- thioated bFGF antisense oligodeoxynucleotideswere added to cultures of the U-87 human glioma cell line. The U-87 cell proliferation was inhibited to 70.6 I0.4% of control at 10 pmol/L and to 53.2 t 5.65 of control at 20 pnol/L (P< 0.05). 60th the 7.0- and 4.0-kilobase bFGF messenger ribonucleic acid transcripts were reduced after exposure to the antisense oligodeoxynucleotide, and cell-associated bFCF protein was reduced by 44%. The sense oligode- oxynucleotide, a negative control, failed to inhibit U-87 proliferation. These data support the concept ihat bFCF expression could be a key event in glial tumorigenesis that mav be necessary for the sustained growth of humùn gtiomas. iNeurosurgery 3 1:7?3-732, 1 995)

he malignant phenotype ot: hurnan gliornas is chancter- To &termine the transfwrning potential ot tiFGF, hrirnm izrd bu 10cd intiltrative growth ( Il1, cingiugrnt5is (6,X fetaI ,istrucyteil \vert? stably h;inskcted wi th a bFGF rz\prtl+ T501, and an elewted pr01ifer;itir.r rate 63, 3).Basic .;ion wctor and euamineci for indication^ ci t cellular transfor- fibroblast grorvth iaïtor (trFGF) is CI potent m~tugtrnicmito- mation in vitru. Pl'iciiphoruthiotiterf tiFGF .intist.n+e 14ir;or.e- gen (14 th~tdrnulates the prridirction ot protrinitst5 (14. ~\ynucleutidt's~-ttre ~II~L, u;td tc) identity a ride ttr CFCF in 27-19,521 and proliieration in glioma crlls (17,1S, 33,341- The the pro1irer;ition of an established human glioma id line. T1iè rlddition of bFGF to human ghma cefls in vitro has bren resufts suggest that bFGF ma! be important in glial tumciri- shown to induce ceIlular motility and invasiveness (13. By genesis and ylioma ceIl proliferation. contmst, the depietion ot copper, 'tn ionic mhctor of bFGF (481,inhibited giiomcl cell invasiveness (8). Tfiese observations have focuseci our Iaboratory on the expression of bFGF in MATERIALS AND METHODS human malignant brain tumors (7l and on the rofe of bFGF in the growth of human giioma ceHs (11)- The shtdies desaibed Cell culture in this article wre designeci to test the hypcithrsis that bFGF Prirn,in; human retal ~lstroqtes,tV385 (3,wcre grotvn in regulates gfioma ce11 gowth and ma! be used as a target to Dulbecco's mod ifieti Eagle's medium ( DMEW supplemen td suppress giioma ceII prolifent'< ion- with 10% ietal bovine serum, 100 W/ml penicillin, and 100

Neurosurgery, I'd. 3. 724 Gately et al. pg/nil streptomyciii. Tlic e~t~iblisliedIi~iiti~in glioma cc11 linc U-ûï(Xmerican Tvpe Culture Collection. R(ickvi1le. MD) was iu1turcd in DMEhl supplt.mentrd rvitli IOr, calt scrum, II111 IU / mi penicillin, and 100 pg/rnl streptomycin (Life Technol- ogies. t~ic.,Bethtn;cla, MD). Cells were maintained in a humid- ifid incubator at 370C in an atmosphere of -3CO2 anci 95i air, rvitli the medium replaceci trvice prr \\.eek-

Cell proliferation assay Ce11 prolifention was determined bv using the Ce!l Titer% .Arl~ec'u~on-r~~diciactiveCeIl Prolifemtim Assy (Promepi Corp., blariison, WH. This systern rneasures the bioredurtion of 3-~4,S-dimtitliylthiazol-2-yl)-5-(3-~nyl)- 2-(-L-sulfophenyl)-2H tetnzolium tX1TS) to an aqueuus du- Transfections ble hrniazan in the presenet! ut the electrcin coupling reagrnt, phenazine methosukitt! ( f%!S), by prriiifen ting ct?l!s. An ab- sorbance reading rctlecting the quantitu ~iihmnran pro- duced wsobtaind by using an au ton~~tdmicro pl,^ te rdèr (Bio-Tek fnstrumrnts, C\'iric~c~sbi.LTl BxAgruund abricirb,ince from wells containing only culture medium was subtrrictd froni the sample absr~rt?;incevdiies tci ohtain the correcteci at7s~irb~iiice. bFCF and fetùl astrocyte proliferation TIie LV5Si tetd astrocyte cells tvere ~idtèdin tripliate in .I 96-rvdl tissue culture pI;ittt (Becton Dickinson, Lincoln Park, NJ) ot .i concentration nt 2.3 x le' ceIls/ well. The nrxt dav, frrsh medium rms addeci with incre'ising concentrations (il-75 ng,/rvt'll) of recombinant human bFGF (provideci by the Dupont hierck Pharrnace~iticrtl Curpmtion. IVilmingtuii, DE 1. The c'elIl; t\we incuhted ie~r4s liotirs. Tlwn 211 pl at the comtiine~iLITS, PMS rmgent tuas aci~ledtu each tvt.11, and the phtes wre incub,ited ,in additioncil 2 horirs at 3TC. Glial fibrillary acidic protein immunocytochemistry

Reverse transcription and polymerase chain reaction

Total ribcinucleic xiJ ( RSX) rvss isola ted irom the hf3S_5 huniaii tetal ,istrucyttts ,ind the L-$7 rnLiIignant glioma cd1 linr i 101.The bFGF receptor primrrs wre ihosen on the basi5 of the qut'nct? of the transrnembriint! Jomain of the Fibro- bl~stGrotvtii Fxtur Recep tor 2 ( FGFR-3. a region conserveci in the FGFR fctmily (21,361:Oligonuclwticie Primer 1 (5'-GCC ACC AGT CCC CCX TCX TCX-3'1, complementaq- to nucle- titides 21SO-120(l: and Oligonucler?ti~iePrimer 1 (3'-G.K GC.4 -AC.-! GAG AAX GAC TG-3'). correspnriing to nucitrotides 1343-13S. Re\-erçe t~inxriptionrviis; ~ttrtrirtnedhr 60 min- utes dt 37C in LI 33-pl reaitiun irmmting {tk 5 ps ut tutal kYA tt'mphte, 3 ot' 2d pmd:' L Oligunucleotide Primers 1 and 2, 1 x reverse transcription butier (50 mmol/L Tris-HC1, pH Y .3);40 mrnW L Ki: b mmol! L MgCl,; 1 mrnoli L dithiotrei- tol; 2 pI of each deo~yibonucleotidtriphosphate (75 mm01 / LI; 6 pI of dithiotreitol (0.2 maI/L); and 600 U of Superscript FIGURE 1, Restriction map of the sp-bfGF cDNA. The bFGF reverse transcriptase ( Life TechnoIogies, Inc.1. PoIymerase gene is under the transcriptional control of the ~oioney chain reaction (PCRj ruas pert'onned in 60-pl volumes con- murine leukemia virus long terminal repeat (Mo-MutV LTR) sisting oi 11) pi of the reverse transcription reaction, 2 pl of (41). After translation, the mouse immunoglobulin heavy O~igonucltwtirit'Prirners 1 mi 2 (20 pmoVL), 1 x PCR buffer chain signal peptide (SP)is found fused to the second amino (100 mmoVL Tris-HC1, pH 33);500 mmol/L KCI; 3r acid of the bovine brain LFGF. IVS, iniervening sequence; $ mmol/L LM@,; 0.1 mrnoI/L each deoxyribonudeotilfe UT, untranslated; SV, sarcoma virus. bFGF and Human Gliomas 725

phospli,itc-buffered saline, the cells wre incubatcd rvith CI liiotinyla ttd 'stlcondary aiitihciriv (Zvrnt.4 5trcptbiviJin-Biiititi Systeni; Zymd Labontories, San Fr,inciscu, CA) for 10 min- utes at roclm temperature, rvashtxi dgiin in phosphatt~biiff- ered siliiie, and then intubateci for 5 minutes with concen- tntd strep tavid in-fluorescein isotiiiocyana te con juga te Hybridization to 32~-labeledprobes (ZymeJ Laborat~iries). nie slides were niountcd with an Prehybriciization was in 30% Corm,imide and standard aqueoiis rnuunting medium (Aqua Mount: Lttrner Lcibunto- hiiffer (46) for 5 hours &it4232 X 1.5-kilobase tkb) EcoRI ries, Pittsburg, PA), and the CFAP rvns r-isualized with a Leitz bovine brain bFGF cDNA fragment irom the pbFGF plasmici Aristoplan microscope (Deeriield, ILI. (411 rvas Iabrled by ran~iciniprirning (Litc TeclinciIogies, Inc.), incorpomting a-"P-labeld ~itwvrih~tidinr5'-tnphus- Oligodeoxynucleotide synthesis phate (Amersham Laboratories, Xrlingtc.rn Heights, Il). The Fiiteen-base oligodeo~nuclt.~tides tvere syn tlwsized st the "P-laheled probe rvas denahird snd addeci to the prehybrid- Biutechnolugv Faality ot ilorthrvestern L niversity iChiwgo, iz,iti~msolution, and hytiriciization to the membrane! rvas Fer- IL), with a pliusphorot hiobite substitu tir)it a t each bdse. The icirmtrd for 24 hours ;tt 42C. Nonspecifï hinding was re- iiligodwuynucltwtiiies wrrr ~ies~i1tt.d.Iyii~hilized, and thrn niiived by wclshing in 0.5u ÇSC 0.1 (; 'icidium dodecyl ~i~lhtr resiispen~iedin sterile watrr ~nLf st~irtliat ->WC. The di- tweight/vc.ilume) at hOcC. Visualization and yuantiticaticm ot godeo~ynuclei)ti~iesrwre targeted 11gainst the first sptice do- the signais rwre obtaineci by using a Fuji\ BAS 2tN10 Phos- nor-dcceptur site at Codon 00 ot the liuman hFGF gene (2). phorimager (Fuji Photo Film Co. Ltd., Tokyo, Japan).To en- and the sequence was confirmed to be specific for the bFGF by sure eyuat RNA loading, èthidium bromide-stained rib- iising the basic local alignmcnt search twl algcirithm (3). Tlie .;~rni,ilRNA band.; rverr ~hotogrnphedunder ultraviokt light. mtisense ~iligucitwynucle~tiiicN'GIS ,T'-TAC CTT GAT GTC ln ,~dclition,the membranes were probd secondarily with the AGG-3'; the correspondhg sense oligodeoxynucfeotide 5'- "i~ititri~l"cDN.4 c;pccie, g1ycer;ilcieliycie 3-phciçph~tc~iehy- CC7 C.K ATC AXG CT.4-3' w.is iisd LI$.i ntlgbltlvecuntrcil. ~iri~çen~ise. Antisense inhibition of gliorna cell proliferation The Lr-ij7 human gliomci ceIll; wrrr phtrd in triplic~teintv Quantitalion of bFGF by enzyme-linked a Ya-rvelf tissue culture plate (Becton Dickinson) .iit a concen- immunosorbent assay (ELISA) tration ot ?.? d. liit cellr.'udI. The ne\t ~hy,seriim-irer DMEXI supplemtmtrd with Ham's; F-12 .mi wnse or antisense oligoci~~~;ynucItloti~irsat ccmcrntratic~ns d O, II), ~iid21) rzmol/ L were addeci. The ceils were inctlbated for 5 clays, 'Y nd 21) pl oi the combineci XITS!PMS reagent tuas biddtsl to exh rveil. the addition cii the MTS/PILIS migent, the plates rvere incubateci an ~dditionalZIiours at 3TC.

Northern blot anaiysis The t'-S7 humm gliomd cells were plciteci in ho-mm ~mlture plates t Becton Dickinson) at a concentration ot 1.i) Ili' ïeIls/r\.ell. The nelt &y, ~enirn-frrrD'rlEhI supptenienttrd rvith Hbmi's F- 12 mci sense or an tisemr oligc)~Ieo.ryniicIei~- ticies nt a concentration ut U or 2tl pmriI/ L rvert. il~idd.The cells were gruttn fur 3 days, ,inri the total cellular RV.4 w,tr i~ol~itedi IO). The RS.4 pellet r\.as resti~;~enddin ribonide- asr-t'rer r\;iter .inJ $toreci at -W'C untii iiseti in Sorthcrn tilotting as cles~ribedpreviuusly (4bi. The RS.4 tvas quanti- tateri spectrophotometrically by its atisorbance at 260 nm ([ optical density = 40 pg/ml). Fiiteen miuognms of total RVA \vas Jenaturd bu heating it to 63°C in 3.2 mdjL tomciide- hyde and SOC; formamide in sarnple buiier containing 23 ng/~lethidium bromide. The sarnples tvrre eIectrophoreti- Statistical analysis dlyhctionaterl on a t .ZS formaIdehyk-aprose geI in 0.02 mol/L molinopr~p~inrsdfonate buffer (pH, 7.0). Afer elec- Results are es~ressedas mean = stcindard deviation. Sta- trophoresis, the gels were photographeci under ultraviolet tistically significant differences between means were deter- light ancl rinsed in s\.;iter, ,mi the RLiA \vas tramferred wer- mined tiy wing either a two-tai1ec.f 5tu~ittnt'st test for trn- night €rom the gel to a nylon membrane (Bcwhringer Mann- paird samples, or a one-way anaiysis of variance- Differences heim Corp., indianapoIis, [NI by capitlary transkr, using 10 x with P c 0.05 were considered statistically signifiant.

Neurosurgery, LfoL 3.-. .Vu. 4, October 1?95 726 Gately et al.

RESULTS tut wsm

FetaI astrocyte proliferation The W585 fetal ash-ocytes wre reçpmsiw to increasing concentrations of exogenous recombinaint human bFGF (Fis. 11. Celiular proliferation was significantiy eievatd at 25 ng of bFGF/well (P< 0.04) md at 73 ng of bFGF,'welI (P C 0.0001), compared with the fetal astrocytes that rwre not exposed to bFGF (Fi$. 2). FIGURE 3. Detection of bFGF receptor mRNA in W585 and b FGF receptor U-87 cells. Lanes 1 and 3 represent the predicted 633-base Reverse transcription-PCR confirmeLi the presence of the pair bFGF receptor cDNA product, Digestion of this frag- messenger RNA (mRNA) for the trsinsmembnne domain of ment with the BghI restriction endonuctease demonstrates the bFGF receptor in both the \C'jSS L-87id1 lin*. The the predicted 505- (arrowl and 128- (not visible) base pair bFGF receptor-specific Primers 1 'Inci 2 pduced the pre- fragments. dictd 658-nucleotide cDNA fragment. The nested Primers 3 ,id 4 pérmitted the specific amp~ificationdi the bFGF recep- tor cDNA anci prociuced the ~redkted63.3-nucleotide cDNA fragment (Fis. 3. hrrtes I and 3). The identity ut' the PCR products was confirmd by restriction digestion; the region of iimpIification of the bFGF receptor cD5.A contains a single BglH restriction site (36). The 6~111digestion produced the predicted 305-nucleotide fragment (Fig. 3. hrws 2 and 4) and ,i 12s-nuïle~tifngrnent tnut visible).

Transfections The fetal astrocytes in control (rh'im-transfected) and pSV2cr3.6-tmnskected groups tail& to suri-ive the spht into the Zt\\-el1 tissue culture dishes. Brc.~u.;e celis in the control grouys diJ not grow, sp-bFGF ce11 clone5 wre selected on the h-is t 1i1and prdiirritticm .ilcmr. Of 1 1 sp-bFGF clones, two !EN and D3) were nndomly selectd tor hrther ,in,ilysis. The nontransiected ,istrocytes grrw uniiorrnly md rvere contact inhibiteci (Fig. -CA). In contnst, the EH md D3 iiones grew with an aitered morpholosy r Fi.9. 46) anci at ï~~ntluencedeveloprd visible foci. The nontranstectrd tetal

FIGURE 4. A, the nontnnsfected human fetal astrocyte cell stnin was contact-inhibited at confluence. fi, in contrast, the sp-bFCF transfeded astrocytes revealed an altered morphol- ogy, with the loss of contact inhibition. ;i>trucyttti;tvere Li113' ; pi;itivtt ter GF-4P (Fis.5-41. In cmtr;i.;t, man! ot the sp-bFGF tritnstected crlh had lost GFAP :mmu- noreactivity (Fis. 3B). Sorthem analysis demonstratd two bFGF transcripts of a~proximateiy7.0 and 4.0 kb in the non- 044transkted ietal astroc? tes. The 2ip-b FGF ce1Is aIso ex pressed Control 1 10 100 these h-rinscrfpts, as rveI1 as an additional tnnscrîpt of approx- bFQF Concentration (ng / wdl) imately 9S8 kh tliat correspondeci to the trrinsfected bFGF FIGURE 2. The W58S human fetal astrocyte ceil strain was ïDN.4. responsive to increasing concentrations of human recombi- nant bFGF protein- Ceflular proliferation was significantly Antisense inhibition of gliorna growth stimulated at 25 and 75 ng/well, compared with control. The antisense oligodeoxynudeotides mused a concentra- *, P c 0.04; ", P c 0.0001. tion-dependent inhibition of Li47 ceIl proIiferation to 70.6 r bFGF and Human Gliomas

0llgodwxynucl.oUds Concontratton (PMI FIGURE 6, The bFCF antisense oligodeoxynucleotide caused a concentration-dependent inhibition of the U-87 human gli-

orna cell proliferation. O, P < 0.05. The sense oligode- oxynucleotide caused only a slight decrease in cellular prolif- eration.

FIGURE 5, A, the W585 nontransfected human fetal astro- cytes were uniformly positive when irnmunostained for CFAP. 8, the 64 and D3 sp-bFGF-transfected astrocytes dem- onstrated a significant loss of CFAP immunoreactivity and FIGURE 7. Redudion in the CI-87 human glioma celf bFCF developed foci at confluence (Clone B-4 is shown). mUNA. Exposure to 20 pmol/L of the bFCF antisense di- godeoxynucleotide resotted in a 53% reduction in the :.O-kb and 42% reduction in the 4.0-kb bFCF mRNA tnnscripts. The sense otigodeoxynucleotide caused onlv a slight decrease 0.4% 'coi control at IO pmol/L and 33.2 r j.+?.ot controt at 20 in the mRNA transcripts. prnol/L (P < 0.03. The sense olirgc7iIeoxynircleotide did not signifimntIy detrease ceII prolifera tirin (Fis. 6). Northern blot mdysis drmonstratd two specific bFGF DISCUSSION transcripts conesponcfing to appro.rimately 7-11 anci 4.0 kb in the U-Sï ce11 linr, ïhe addition of 20 pmol/ L ot the antisense In the present study, ive have an;iIyzd the contribution ot bFGF oiigodeoxynucIeotide resuited in a 53% reduction in the bFGF to the maIignant phenotype of human $ai-derived amount of the 7-û-kb hanscript, and a 42'3 reduction in the tumors. Induction of bFGF protein expression in prirnary Ievel of the 4.0-kb bFGF &VA tnm~Tipt(fi$. 3. Quantita- human astrocyts resulted in the mpisition of a tr;insformrd tion of bFGF in the cellular lyscttes of the U-Si ce11 tine phenotype. Conversely, antisense oligodeoxynucleotide inhi- esposeci to the antismse oligdeoxynucltrotide dernonstrateci bition of the bFCF rnRX.4 and protein expression in the U-87 a 44% reduction in the bFGF antigen Ieveh, compard with eshblished human giiorna ceII Iine resdteci in a significant control LM7 glioma ceils that had not been exposeci to the inhibition of cellular proliferation. The hdings presented are oligonuclmtide tFis. s'). The seme 01igonucIwtide did not important, bernuse they link bFGF to the in t-itrc~transforma- cause a significant reduction in the bFGF protein, compareci tion of human astroqtes, suggest that the aberrant expression with the controt group (Fig. 8). of bFGF could be an end? event in the multistep developrnent

Neurosurgery, Vol. 3,-, No. 4. Octaber 1995 Gately et al.

(31, ,i region that is 1it~~IiIycr~tistwd in the FCF reccptcir gent. f4imilv (71, 3h). Alth~~it~l~IYC h'ir~a nclt idcntifitd ttfitb 'rpt.iltic FGF rcccptc)r prcwit rbti tlw \\. -35 ic*lls, rur rcwltb indicate that one or more of these receptors is present and tunctional, These resuIts support prwious work that bFGF signal transduction pathways may plav a critical role in the phencltype oi astrocytic tumors (33, -40). bFGF has been identifieci as a key factor in the biology tif malignant giiomas (1,13,17,18,26,40,50,5 1). On the basis of the important role of hFGF in glioma ceIl tiialogy. there have been several reports on the inhibition of bFGF expression in #ornas (4, 16, 17, 32, 3b. 40, 51). Wt? notv report a concentra- tion-dependent de~xasein giioma ceil prolifention aiter the Liil~iitiunof phu~phrirrithioatt bFGF miisenw digide- citynucleotides. The speciiicity of the bFGF an tisense oligocie- mynucleotide was indicated by the t'crllowing results: 11 U-87 FIGURE 8, Quantitation of the bFGF protein in the U-87 glirirnn ce11 proliiemtion luas in hibiteci t?y the sntisensr but glioma ceIl line by ELISA demonstrated that the antisense not by the sense oligod~~uynucleotide;3 the amount ot the oligodeoxynucleotide caused a -!Ji% reduction in the amount 7.U- and 4.O-kb bFGF mR\X tnnscripts was reduceti dtter of ceIl-associated bFCF. The sense oligodeoxynucfeotide exposure to the antisense oligodeoxynucleotide but not after caused only a slight dectease in the bFGF antigen. cqwure to the sense ~jli~cideo\vniicltiotide:and 3) the bFGF prutein, GISqumtitrlteci by ELISA, ciecreaseci aiter exposure to the antisense but not ,if er ttxposure to the sense oligode- or human gliom&~s.anci Jemonstrate tliat J dt'creast! in bFGF o~yniicleotide. mRSA triinscripts and bFGF protein using mtiscnse digode- Antisense oIigo~le~.r~ynuslt'oti~iescm ~i~iwn-regiil~ite grnc ouynudeotides results in a significant decrease in glioma ceII expression by three mechanisms: 1) blocking the hnction of proliferit'ion. the mRXA through iiterii hindnnce, 2) irreversible bincling ui The transiorming potentiai ot bFGF hns not been rnvesti- the oligonucleo tide to the target rnRY.A seyuence blocking ~atedpret'iouslv in humm astn~cytes.These db~ta&mm- tr'~nsla tion. or 3) ,ictiv,i tion of riboni1ciebic;rH tiecitiic;e. tif tlic strate that transiection of prirnary humm ietal iistroeyte cellc; presc'nce oi RKAiDN.4 duplexes (9.-13,. Ottr resulrs suggest with the sr-bFGF expression vectcir rtlsultd in a tmnsicmned thbitone or ai1 ut thesr niecti,inism~m~iy tw rrtlsponsiblt, kir the phenotype in vitro that was demonstnted by the loss of xtivity oi the bFGF &intisenc;e ~iligorieoxynitclt!~~tides. The contact inhibition, GFAP irnmunoreattivi;y. and cinch~iraye- dtxreascrd expression ot the 7.0- mci 4.0-kb bFGF mR\A independent gowth. The loss oi GFAP immunclreiictivity transcripts (FiLf.3, aiter ,i sint;le administration of the bFGF supports the in vitro h-ansl'ormation, brcause it is 'i patholog- antisense oligoiIeoxynucleotide5, iç trans1,ited intu '1 44'; re- icd change consistent rvith increasing grade in human glial cf uctiun ot glioma ceil-,i~s~~ci~iteibFCF protein (Fi:;. Si. Tlir tumors 443. The rnechanism ot the tmnsiormntion d the suppression oi L'S7 gli~rn,i id1 pro1iler;ttion (Fis.a) 'iitrr sp-bFGF astrocytes remains unknown, but the tnnsfomation antisense expure cuuld bt. the result ~it'this rtlciuction in could be linked to the extracellular relrase of bFGF. Because hFGF protein; excess, or bibrrrant. iurms cit bFGF may c~ther- the jecm tory signd 5ey urnce is responsible tor the e\ trmd- ;vise bt! biv~iiL~bltrt~r elther t'\t~ielillklrreiedse. 5timui;iting Mar export of proteins. the sp-bFGF-transiected astrocytes prditerati~mby '1 dbis~dciutruine mechanim 147, 491, or may wcrete bFGF in the cxlture medium, as sho5k.n previ- tc~rthe bictiration of LI putative intr;icellulc~rbFGF rticeptur, ously (5, 13. The importance of this shiit in the localization ot iitimulating proliieration by an interna1 autoirine rneclianism bFCF, Erom cell-associcited to eutr;icellular release, has previ- i 5, 4 1). This striif!. pn-r~lec;direct cvi~ittncethat C-s7 gIiorn.~ otisly been shuwn to correl~te ivi tli inc~trasedturnorigenicity ceil ~roliteratim ~ie~t'n~lentin part on t9FGF. ,inci mgiogenesis in the deveIopment of tibrus.ircoma t'7). [n addition, bFGF and a novel member of the FGF tamilu, FCF-9, The iindings presentrd support bFGF ~isan impurtant hc- have ben Jetected in the conditioned medium of human and tor in the biology ot hurnan gliomas. The tmnsf'urrning poten- rat gIiomcl celIs (31, 37,47'). Beause the mRYA for the bFGF tia1 ot' aberrant bFGF espression in human astrocutes in vitro receptor was detecteit in the W5S5 ce11 strain (Fixa3) and thesr suggests that bFCF ma' be involveci in glial turnorigenesis. dis were found to be responsive to increasing concentnt-IO~S The suppresion of glioma ceIl prdifrmtion by mtisense oli- ofexogenous bFGF (fig.3, the sp-bFGF transfected asmetes godeo.i~ynucleoti~iessuggests that the bFGF remains irnpor- couid be transformed by classicztl auto^-rine stimulation (49). tant in established glioma. Thse resuits suggest that anti- Four structuraIly related gens that encode five distinct FGF bFGF shategies couId be usehl in the treatment of malignant receptors have been ~iescribed420. 21. 23, 30. 39. W),wme human giiomas. anci the? proride the impetus for the design soluble (19, 3).The PCR pcirners used in this study were of pharmacologicztl and gene-baseci thtirapies that target bFGF . specific for the cytoplasmic domain of' the FGFR-2 receptor expression-

Neurosurgery, Vol. 3.Y ACKNOWLEDGMENTS 13. Finkelstein ÇD. Blaïh P. Nt~tv,~kTI), Hmd CM,Christtlnwn S. Fi~ich PL\ 1IÏ~t~dt I+-.I: cii.~r.ictm-t~~.*.inLi ~b\pr~-+itwtbt .IL-IL~C \Ve V.W. Yong tor th thank Dr. genmiusly pruviding and basic iibroblast grutvth tactor gents in intracerebnl seno- human fetal astrocytes and Dr. M. Klagsbrun for providing genic transplants ot hunian glioma cells. Neurosurgery 34: 136- the psp-bFGF and pbFGF plasmids. CVe ais0 thank Debornh 143, 1994 Cundiff, Holly Duncan, David Ivancic, and Marguerite 14. Folkman J, Klagsbrun hl: Angiogrnic tactors. Science 235:U2- Wotoczek-ûbadia for expert technical assistance. This work 4-47, 1987. was supported in part by gants from the Riinois NF Inc. (SC;), 15. Gately S, Tsanaclis .L\IC.Takano S, Kiagsbmn M, Brem S: Cells a grant-in-aid from the American Heart Association of LM&- tnnsfected with the basic fibroblast growth factor (bFGF) gene ropolitan Chicago (GASI, the Feinberg Cardiovascular Re- fuscd to a signal squence are invasive in vitro and in situ in the search Institute (GAS), and National Institutes of Health bnin. Neumsurgery 36:fS0-7S5, 1995. Grants CAS7781 and CA60177 (SB). 16. Gross IL, Herblin WF. Dusik BA, Czerniak P. Didmond MD, Sun T, Eidsvoog K. Dexter DL. Ynyon A: Effects of modulation of' basic Received. Sptember 14, 1994. iihrublast growth factor on tumor growth in vivo. j Natl Cancer Accepted, April 14. 1995. Inst 85121-131, 1993. Reprint requests: Steven Brem, MD.. Director ot Xeurosurgical On- 17. Cross JL. Herblin WF, Eidsvoog K. Horiick R, Brem SS: Tumor cology, Division of Neurological Surgery. Sorthwestem MemoriaI growth regulation by modulation or basic fibroblast growth iac- Hospitai. 233 East Erie Street, Suite 500, Chiago. IL a061 1-2906. tur. in Stein R, Weisz PB, Haudenschild C, Langer R tds): Aqb ,ftntiwis.Basel. Birkhauwr i'erlag, 1992. pp 421427. 18. Gros JL, Momson E, Eidsvoog K. Herbtin \VF, Kornbfith PL. Dexter DL: Basic fibrublat growth factor: A potentiai dutocrine REFERENCES replator ot human glioin,i ceII groueth. J Neuroscirnce Res 27 t&q-69b, 1C)'jU. L. Abe T, thmura K. Ono SI. Kohno K. Mon T. Hori S. Kuwano bI: induction of vascular endothelial tubular morphogenesis bv hu- 19. Hanneken A. Ying W. Ling N, Baird A: Identification OS soluble man glioma celb: h mode1 system tor tumcr ~nçiogeneis.J Clin iorms or the iibrob1,ir;t ptvth factor reccptvr in blrid. Proc Natl Invest Q2:51-61. 1993. Xcad Sci U S A ~l:ql7-Q174.1W-i. -1 Abraham JA, Whang JL, Tumolo .A, Friedman J, Hjerrild LA, 2U. Houssaint E, Blanquet PR. Champion-Arnaue P, Genel MC. Tor- Cospodiirowicz D, Fiddes JC: Human basic iibroblast growth rigiis A. Courtois Y. Breathnach R: Relatd iibroblast growth tactor: NucIeotide sequence and genomic organization. EMBO j iactor receptor genes exist in the human gnome. Proc Natl Xcad S:LSZbiiX, 1986. Sci U S A Sï:St8û-S~S4,1WtI. 3. Altschul SF. Gish \V. Miller W. Mvers RI'. Lipmùn O]: Basic lacal 21. lohnson DE. Lu J. Chzn H, lVerner S. CVilli,ims LT The humm alignment search tool. J Mol Biol ZfS:4i1~410,l9W. fibroblast grorvth iactor rtïeptor genes: h iommuri structunI 4. Behl C, LVinkler J, Bogdahn C. Meiixensbrrger J. Schligensirprn .vriingement unrier!ie.; the mtxhmisrns ior senentins rece~tor KH. Bysch W: Autocrine growth regulacion in neurotrctdermd iorms that Jit'ter in thetr rhirci immunor;lobulin domain. .Mol Ce11 tumors as detrcted with oligdeoxynucleutide antisense mole Bi01 Il:-Ib27--!62, 1991. - cules. Neurosurgery 33:6M,1993. 2. Kandel J, Bussy-CVetzel E. Radvanui F, Klagsbrun M. Folkman J, -.'I Blam SB, Mitchell R. Tischer E, Rubin JS. Silva hl, Silver S. Ficides Hanahan D: Ntwwsclilariwtion is associateci with a switch to the JC,Abraham JA, Aaronson Sh: Addition of growth hormone evport ot bFCF in the multistep ifevdopment of iilirosan-orna. secretion signal to basic tibroblast ~rorvthfactor reults in ce11 Ceil oe:1095-1lU-1,1QJI. tr~nstormationand secretion oi ~bernntrorms oi the protein. 3. Keegan K. Johnson DE. \VilIiarns LT, Hdynan MJ:I.wli~tion ot an Oncogene 3:119-136, 1988. itdditional member or the tibroblast gotvth fxtw receptur tamily, b. Brem 5, Cohan R. Folkman J: Tumor angioyenesis: A quantitative FGFR-3. Proc Nat1 Ac~dSci U S X SY:1619%1WJ, 1991. methoü tor histvlogic gading. J Nat1 Cancer Inst W:j-li-356. 24. Kent RB. Emanuel JR. Stiriih YB, Lrrenson R, Housman DE: 192 - Ouabain resistance iuniened by exprtsion of the cDKX for a 1. Brem S. T~anaclisAMC. Gross IL. Herblin \L'E ImmunoIualiza- murinr Na ',K'-ATPase u subunit. Science 3F:WI-W. 1987. tion ot basic fibroblast growth factor to the rnicrov~isculatureot 5.&hi T, Foshida T, Tend&Si: X soluble tom of K-mn;'FGFRl hurnan bnin tumos. Cancer 70:26;3-1&I;O, 1991 protein in the ~xlturemedium ot human g.i?;tnc c,incer i&. 5. Brem S, Tsanaclis .WC. Zigzag D: Xntiimw matment inhibits Biochem Biophys Res Commun ZL12l.t,47-1W,1994. pseudopociial promion and the invasvr ?preiid ot QL gliosar- ln. Li LIV. Folkerth RD. Li',ttan,tbtl H, Yu C. Rupnich M. B~rnesP. coma cells in the rat bnui. Neumsurgery 1tx3u1-396, 1Wl. 9. Carter G,Lemoine SR: Antisense tdnoiogy for cancer therapy: kott RXI, Black PILlcL. SaUan SE, Folkman J: .l!iuvveswi count Does it make wnse? Br J Cancer 6ï:86%K6. 1993. and ïerebrospinal tluid basic fibroblsst growth factor in children 1O. Chomcynski P, Satchi N. Singktep methoci of &VA isolation \tlth brain tumours. Lancet 3.W:S2-S6, 1994. by aad panidinium thiocyanate-phenol-crhioroh extraction. 27. Lund-Johansen ,LI, Forsberg K. Bjerktig R, Laentm OD.Etftxts of Anal Biochem 162156-159, 1987. growth factors on a human glioma ce11 line during invasion into Il. Je Ridder L. Calliauw L: Invasion of hurnan bnin tumors in rat brain dggregates in cuIture. Acta Neuropathol Sk19û-197, vitro: ReIationship to cIinicaI evolution. J Neurosurg 2589- 1993 . 593, 1990. 28. SLignatti P, Rifkin DB: BioIogy and biochemistry of proteinases in 12. Engebraaten O, Bjerkvig R, Pedersen PH, hemm OD: Effects of tumor invasion. Physiol Rev ;3:161-195, 1993. ECF, bFGF. NGF, and PDCF (bb)on ceff proMerative, rnigntory 19. htignatti P, Mrinieri R, Riikin DB: Expression oi the urokinase and invasive capacities of human brain tumor biopsies in vitro. receptor in v;iscuIar endothelia[ cek is stimuiated by basic fibm- Int j Cancer 3~214,1993. blast pwth factor. J Cell Biol 1l3:I I93-l2Ot, 1991.

Neurosurgery, Vol. 37, Xo. 4. October 1995 no Gately et al.

30. Miki T, 8ottaro DP, Fleming TP, Smith CL. Burgess WH, Chan AML. &. %mbrook J, Fritsch RF, Maniatis R: Molvrular C1unin.q. Cold Aamnson SA: Determination iit lipnd-binding spxifi~ityby dtrr- Spring HLirbor,Colif S~ringHarbw Prts, 1989. native spiicing Two distinct growth factor receptors encoded by a 47. Sato Y, Murphy PR, Sato R, Friesen HG: Basic fibroblast grorvth - single gene, PmNat1 Acad Sci U S A S:2&250, 19% factor release frorn h~vinecorneal endothelial celk .md human 3 t. Mipnoto M, Namo K. Seko C. Matsumoto 5, Kondo T, Kuroksw astrocytoma celis is ce11 density dependent. Mol Endocrinol T Molmlar clonùig of a nvve! cytokine sDNh encociing the ninth 3:74+748, 1989. member of the fibroblast growth factor Ssmily, which has a unique 4s. Shing Y: Biaifinity chromatognphy oi tibroblast growth tactors. seaetion property. Mol CeU Bi01 I3:4251-1259,1993. Methods Enzymol 198:91-95, 1991. 32. Morrison RS: Suppression of basic fibmblast growth factor ex- 49. Sporn MB, Roberts AB: Autocrine growth hctors and cancer. pression by antisense otigodeoxynucleoticies inhibits the growth Nature 31 3:745747, t985 of transformeci astrocytes. J Bioi Chem 13:28-734, 1991. 30. T~kahashiJA, Fukumoto M. Igarshi K, Oda Y, Kikuchi H, Ha- !kt: 33. Morrison Rç, Giordano S, Yamaguchi F, Hendrikson S, Berger tanaka Correlation ot basic fibroblast gmwth factor expression MS, Palaewski K: Basic fibroblast gowth factor expression is leveis with the degriir of malignancy and vascularity in human 1992 required for clonogrnic growth of human glioma cells. J Neurosci gliomas. J Neurosurg 76:792-798, 51. Takahashi J.4,Fukumoto M, Kozai Y, Ito hi, ma Y, Kikuchi H. Res 3J:JU-5W, 1993. Hatanaka hi: Inhibition or ce11 growth and tumorigenfiis of hu- M. Momson RS, Cross JL, Herblin WF, Reilly TM, LaSah PA, Alter- man glioblastoma cells by a neutralkïng antibodv against human man RL, Mosbl IR. Komblith PL, Delter DL: Basic t'ibroblast basic fibroblast growth factor. FEBS Lett 288:65-71, 1991. growth factor-like activip and receptorc 'ire expressecf in a hu- =2. fakano S, CdteIy S, Sevik ME, Herblin WF, Gross JL, Engelhard man gliama ce11 line. Cancer Res ?0:324-2523, 1990. H, Perricone Itt. Eidswos K, Brem S: Suramin, an anticancer and RS, 1V. 35. .lIonison Yamaguchr F, Bruner Jht. Tmg hl, blcKwhan ~ngiosuppressiveagent, inhibits endothelial celi binding of basic Mç. gene Berger Fibroblast growth factor receptor expression and fibroblast growth factor. migntion, prolifention, and induction immunoreactivity are eleva ted in humlin glioblastoma mu1 ti- di urokinaschtype phminogen activator. Cancer Res 326% forme. Cancer Res EI:Z"4l-ZYo. 19%. 2660, 1993. 36. hlurphy PR, Sato Y, Knee RS: Phosphorothioate antisense oligo- 3. Tsanaclis AM, Brem S. Gately S. Schipper HM, Wang E: Statin nucieotides against basic tibroblast growth factor inhibit anchor- imrnunolocalizatian in human brain tumors. Cancer bS:3GQ2. age-ciependent dnd anchoragrindependent growth ot (1 mdig- 1091. nant glioblastoma ce11 linr. Mol Endocrinol 6:8r4W, 1992. 3- Tsanaclis .AMC Robert F. hIichaud J, Brem 5: The qciing ~ooiof 37. Namo K, SkoC, Kuroshim K, Matsutani E, Sasada R, Kondo T, ceUs rvithiri human bnin tumors: In situ cytokinetics using the Kurokawa T: Novel secretory heparin-binding factors from hu- monoclonal antibody K-67. Can J Neurol Sci t Y:12-17. 1991. man glioma cells (gfia-activating factors) invoIved in glial ceIl 55. Yong VW, Kim SU, Pleasure D: Growth tactors tor fetal and ndult pwth. J Bi01 Chem 268:255ï-286.4. 19Q3. human astrocytes in ~ulture.Bnin Res W:S~a,1JSS. 38. Nguyen hl, Watanabe H. Budson AE. Richie JP. Hayes DF, Folk- man J: Elrvriteci levels of ,in dngioo;enic peptide basic tibroblast growth factor in the urine of patients with a wide qxctrum di COMMENTS cancers. J Nat1 Cancer Inst 3b:%-361. 1994. 39. Partanen J, blakela TP, Eeroia E, Korhonen J. Hirvonen H. The controt of the grorvth of human glioma ceils is an area Ciaesson-Webh L, AIitalo K: FGFR-4: 4 novrl acidic fibrobl.xt ot extreme scientific interest. It now appears that there is a growth factor receptor with a distinct zkpression pattern. series of peptide growth factors that influence the growth EMBO J 10:I347-1354, 1991. parameters of malignant gliomas and may play a rolr in their W. Rdekop CJ, Naus CCC: Tmnstection rvith bFGF sense ~ndanti- evolution and progression. One of the most relevant peptides sense cDNA resulting in rnodificahon ot malignant gIioma is basic fibroblast growth factor (bFGD, which appears to growth. J Neumsurg S283-90, 1995. have certain characteristics of an autocrine grorvth factor for 41. RogeIi S. Stem D. KIagsbrun hl: Conshuction and expression of gtiornas. transtorming gene resulting trom hlsrun or basic fibroblast In this artide, the preriousiy reported findings that anti- growth factor gene with signal peptide qurnce. Methods En- sense 01igonucIeotids against bFGF can decxease glioma zymol 198:117-124, 1991. growth are confirmeci and extended. In particuIar, the de- 42 Rogelj S. Weinberg RA, Fanning PX. KI.igshrun .LI: Basic fibro- mase in the bFGF mRY.4 h-anscripts is notable. blat gowth bctor Neci to a srgn~lpeptide trmsionns tek. In addition, the authors have obsen-ed that human kta1 Nature 331:13176, 19M. astrocytes cm bFGF 43. Rothenberg LI. Johnson C. LaughIin C. Green 1. Crridock J, Sain-er be traniiectd with a expression vector N, Cohen JÇ: Oligonucieotides as antisense inhibitors of gene rnodiiied by addition oi a secretory signa1 peptide squence. expression: Therapeutic implications. j Nat1 Cancer Cwt S1:1339- This transfection resuited in evidence of transformation by the 1544,1989. vector. 44. Ruta M, Burgess W, Givol D, Epstein J, Neiger N, bplow J, 00th sets of experiments lend hrther support to the poten- Cntmley G,Dionne C,Jaye M, Sdilessinger 1: Receptor for aadic tial roIe of bFGF in matignant glioma bioiogy. The issue is fibroblast growth factor is related to the tyrosine kinase enrded ma Je even more intriguing by the relationship of bFGF to the by the fitu-üke gene (KG).Proc Nat1 Acad Sci U S A 36:SZ- process of angiogenesis in gliornas. mus, bFGF has a potentiaI 8726,1989. rote in a series of parameters of glial malignancy, including 45. Rutka JT.Hubbard SL, Fukuyama K. Matsuzarva K, Dirs PB. proIiferation, transformation, and angiogenesis. Becker LE: Effects of antisense glid tibnllarr; aadic protein com- t piementa- DNA on the growth, invasion, and adhirion of hu- Paul L. Komblith ' man astmcytoma c&. Cancer Res 33267-32r7,,199.1. Pittsb tir& Pennsy Irnnin bFGF and Human Gliomas ni

Previous work has established the importance of bFGF in hvpothesized, could increase proliferation bv an intemal or the biolugy 05 bnin tumors. Such work is tvdl rttferencd in t.\ttlrnLilutucrine mechmisrn, this article. In the current study, the authors have successhilly The Iindings reported in the current article confirm those of translected a hurnan fetal astrocyte ceLi line with a bFGF Morrison's group (3,4), in addition to icientifying a dmease expression vector, resutting in the loss of glial fibrillary aadic in the bFGF mRNA transcriptç. The conclusion that the tram- protein expression and the Ioss of contact inhibition, changes fection of fetaI astrocytes causes in vitro hnsfomsition re- that are consistent with rnaiignant transformation. The article mains questionable despite the failure of the authors to detect would have been strengthened had the authors ben abIe to growth in any of the control fetal astrocytic lines. Although demonstrate that these changes could be reversed with the the decrease in glioma ce11 proliferation uçing 70 pmoI/L use of antisense oligodeox~ynucleotides. antisense oligodeoxynudeotide was modest (47%), it is com- The mechanisms by which an increased fibroblast growth parable with that reported in the literature and was Linked to factor expression ieads to malignant progression remain to be decreases in the bFGF mWA &inscripts and protein levels. elucidated. The article presents indirect evidence regarding I found tha t the authors' logic in conducting the studies was the possibility of the presence of an autocrine loop; however, sound. The methodologies and conclusions, for the most part, there are undoubtedly other molmlar mechanisms at play were also appropriate, Takahashi et al. (6)abo suggested, in and kirther studies are warranted. 1990, that tumor-derived bFGF is involved in the progression oi gliomas and meningiornas in vivo. This group has reported Raymond Sawaya thst expression of the bFGF receptor gene was not ijignifi- Houstorr, Tri.ws cantly different between tumor and normal tissues (7)and The authors report that bFGF is important in the cleveiop- that neutralizing the bFGF with a mouse monocIonal anti- ment ot the malignant phenotype in human gliomas in ~ltro. body in hibits bo th anchorage-dependent and -independent Transfection of a human fetal astrocyte ce11 line with the growth of U-87MG and T98G human glioblastoma cells, a11 of signal peptide bFGF expression vector resulted in changes rvhich express the bFGF receptor genes. The subseyuent .id- that were consistent with a more aggressive or malignant rninistntion of this monoclonal antiboctv suppressd tumor phenotype. These included anchorage-independent growth, deveiopment in nude niice. They toncluded that the neutnl- Ioss of contact inhibition, and decreased glial fibdlary acidic ization of bFGF may control the growth OC himors that bar protein immunoreactivity. Blocking bFGF expression with a the bFGF teceptors (5). phosphorothioatd antisense ~Iigodeo~ynucleotideresulted In transgenic tumor models of fibrosarcoma, it has been in the inhibition of cell proliferation. The authors suggest that reported that there is a change in the localization of the bFGF this inhibition rsults irom ,ln inhibition of the translation oi from its normal cell-associated state to the extr;icellular re- the protein, and they conciude that the bFGF expression is a lecise that correlates with tumorigenicity. The authors suggest "key event in glial tumorigenesis" and thst it rnay be neces- that antisense ~Iigonucleotirfesinhibit the bFGF protein trans- sary for the sustained growth of human gliomas. lation. Tumorigenesis may result from 3 disc~eteswitch to ,in bFGF is known to be neurotrophic and gliotrophic, with angiogenic phenorne that causes the export of the bFCF (2). sequence homologies thst are similar to interieukin-1. It has Lntortunately, it remains difficult to explain how a transla- been hund to be mitogenic for capillary endotheiial cells and tional inhibition results in the inhibition ot prolii'ttration. es- to s tirnula te angiogenesis. This a bility to stimula te angiogen- pecially in light of the idea that unlike most growth tactors, wis and enhance neovascularization is a likely contributor to the prima. translation product for the bFGF appea- to Iack tumor progression. A number ot ce11 lines Jerivecl irom ma- a senetory signal peptide. Given that the secretion d bFCF Iignant gliomas express the gene t'or atidic fibrobiast grorvth exhibits Little mitogenic activity, it has been conchdecl that the factor, which shares a 50% homoIogy with bFGF. The bFGF interaction of the bFGF with its receptor probably occurs messenger ribonudeic aad (mR!A) has also been isolated in while the fusion protein is being processeci along the senetory large smounts in some, but not all, ceil Iines deriveci from pathtcav (1). malignant gliomas, tvith at Ieast four distinct hybridizing mRWAs. It is now wel established that the unrestricted growth of tumors depends on angiogenesis. It rernains un- clear as to when or how the transition to an angiogenic state occurs Juring early neoplastic development. ïhe authors suggest that the bFGF transiected cefi clones 1. BIam SB. Mitchel1 R. TierE, Rubin JS.Silva ,Li, Silver S. ricides couid be transformeci by cIassical autocrine stimulation. The! JC. Abraham JA. Aamnson Sk Addition of pwth hormone secretion signal to basic fibrobIast growth factor resdts in ceU have identified that the bFGF antisense oiigodmxynucleo- transformation and wcretion of aberrant forms ot' the protein. tides seiectively bIo& the 7.0- and 40-kilobase bFGF tran- Oncogene 3:179-136,1988. scripts and dmase the amount of the bFGF protein. They 2. KAndcl J, ksy-Wetzel E. Radvanyi F, Klagiibrun ,CI, Folkman J, also suggest that the decrease in proliieration could be the Hanahan D: Neovascularization is associateci with a nvitch to the resdt of this decrease in the bFGF protein, whereas prolifer- export of bFGF in the mdtistep development of fibtosa~oma. atîon is the resdt of excess production of bFGF, which, it is Ce11 66:2095-21W, 2992. 732 Gately et al.

Momson RS, Giordano S. Yamaguchi F. Hendrikson S. Berger 6. Takahashi JA. Mori H. Fukurnoto M. [garshi K. ],ive 'LI. Oda Y. 'LIS, P>i7Ic~ewskiK: Basic fibroblast grcwth f,~cturexpression is hikiichi H. HatanùL.1 U: GtliiC c\prtlsswn iit ribruhl.ist gruwth rquired for clonogenicgrowth of hurnan giioma cek. J Neurosci factors in human giiornas and meningiomas: Demonstration of - Re5 34502-509, 1993. ccliufar source of basic tibrciblast gmwth factor mRNA and pep- RS, IL, WF, Motrison Gross Herblin ReiIIy TM. LaÇala PA. tide in tumor tissues. Proc Nat1 Acad Sci U S A S75flû-5ïi4, Alterman RL, Moskal IR. Kornblith PL. Dexter DL: Basic fibro- blast growth factor-like activity and receptors are expressed in 1990. a human gliorna cell line. Cancer Res 50:2523-2529, 1990. 7. Takahashi JA, Suzui H, Yasuda Y. Ito N. Ohta M, jaye M, Takahashi JA, Fukumoto M, Kozai Y, Ito N, Oda Y, Kikuchi H, Fukumoto M, Oda Y, Kikuchi H, Hatanaka LM:Cene expression Hatanaka M: Inhibition of ce11 growth and tumorigenesis of of fibroblast growth factor receptors in the tissues of human human gliobtastoma ceils by a neutrslizing antiboriy against hu- gliomas and meningiomas. Biochem Biophys Res Commun man basic fibmblast growth factor. FEBS Lett B8:bSX, 1991. I77:l-7, 1991.

Neurosurgery, Vol. 37, No- 4, October 1 995 Cells Transfected with the Basic Fibroblast Growth Factor Gene Fused to a Signal Sequence Are Invasive In Vitro and In Situ in the Brain

Stephen Gately, B.A., Ana Maria C. Tsanaclis, M.D., Ph.D., Shingo Takano, M.D., Ph.D., Michael Klagsbrun, Ph.D., Steven Brem, M.D. Division of Neurosurgery, Northwestern University School oi Medicine, Chicago. Illinois (SG,ST, SB). Department of Neurology and Neurosurgery, McGili Universitv School ui Medicine, Montréal. Québec. Canada iSCI, Department of Paihology, University of Sao Paulo, Sa0 Paulo, Brazil (AMCT), and Oepartments oi Biological Chemisty and Surgery, Children's Hospital. Harvard Medical School. Boston, ~\iassachusetts(MK)

tNVASlVENESS IS A critical event in the developrnent of malignancy in bain tumors. A potential molecular mediator is basic fibroblast growth factor (bFGF). NlH-3T3 cells transfected with the bFGF gene fused with a signal peptide sequence (signal peptide bFCF) acquire an invasive phenotype as rneasured by in vitro assays of invasion including: 1) the formation of branching netwotks on Matrigel; 2) invasiveness in a chemoinvasion assay; 3) migration in a cell spreading assay; 4) detection of an M, 92,000 gelatirtase: and 5) local invasion into the surrounding neuropil after injection in the athymic mouse bain. By contrast, cells transfeded with only the native bFCF gene (wild-type bFGF): 1) formed discrete cell clusters on Matrigel; 2) were less invasive and migratory in vitro; 3) released minimal M, 92,000 collagenase; and 4) in vivo formed a pseudocapsule that separated the tumor cells from the neuropil. Quantitation of bFGF in the conditioned senirn-free medium of the cell lines by enzyme-linked immunosorbent assay demonstrated that the signal peptide-bFGF cell clone secreted bFGF. These findings suggest a role for bFCF-médiated pathways and collagenase as molecular determinants of invasiveness in the brain. (Neurosurgery 36:780-788. 1995) Key rvords: .Angiogenesis, Basic iibroblast grawth hctor, Coll~genase,~eoplasm imasiveness

pathoIogica1 hallmark of malignant brain tumors is primarily a cell-associated protein (331, and its mode of action local invasion into the surrounding normal brain (18, is undefineci. A;?- Des)pite the importance of invasiveness to the clin- The overexpression of the bFGF gene was detected in the idoutcome for patients with bnin tumon 19), Iittle is known hurnan gIioblastoma ceII üne A-172,in cornparison with other about the ceIlular and molecular mechanisms thar regdate hurnan güoma lines and normal brain (11). When this cell Iine neoplastic infiltration. Basic fibroblast growth factor (bFGD is was transpIanted into the athymic rat brain, bFGF messenger a potent angiogenic mitogen (131, which has been identifid in RiiA \vas detected by in situ hybridization primady at the elevated Ievels in human gbomas (15,26,24-26,37,41,42,50) tumor-normal brain interface (Il), suggesting a role for bFGF and in the microvasdature of gliobIastomas (3, 3, 39). In in the heightened angiogenesis, proliferation, and invasion at vitro, exogenous bFGF induces ceIl motility (401, glioblastoma the tumor edge, Because of the obvious rdevance of bFGF to ceii migration, and invasion (10,221 and stirndates the pro- the biology of brain tumors, we hypothesized that cells &ans- duction of protaçes (4,227).Recentiy, giiorna ceU growth was fected with the bFGF complernentary deoxyribonudeic add shown to be dependent on bFGF (251, and the release, or (DNA) fused with the code for a seaetory signal sequence, secretion, of bFGF was required for promoting glioma ceU signal peptide-bFGF (34, wouid be both angiogenic and in- growth (25). The native gene for bFGF does not normaily code vasive in contrast to celIs transf&ed with the native, wiid- for a seaetory signal peptide (1); therefore, bFGF remains type-bFGF gene. We report that NM-3T3 cells transfected

780 Neurosurgery, Vol. 36, No. 4, April 1995 bFGF and Neoplastic invasion i

with the signal peptidebFGF expression vctctor are highly In vitro invasiveness assays invasive in vitro and in situ in the brain. For the qualitative assay of invasiveness, the basement membrane Matrigel (Collaborative Biomedical, Inc., Bedford, MA) tuas used with protein concentrations ranging from 95 MATERIALS AND METHOOS to 20.1 mg/ml. Six-well culture plates (Becton Dickinson, Cell Iines Lincoln Park, NJ) were coated with 0.5 rnl/well Matrigel and placed in a humidified incubator at 3°C for 60 minutes. After MW-3T3 Cibroblasts, which have receptors for and respond polymerization of the Matrigel, the signal peptide-bFGF and mitogenically to bFGF (461, were transfected with a previ- tvild-type-bFGFcells in dturemedium tvere carelully added ously describeci vector (34) directing the expression of normal to the wells at a concentration of 5.0 x 1@ceils per well and bovine brain bFGF. The signal peptide-bFGF cells were trans- examined after 24 and 43 hours in culture. At the conciusion fected with the bFGF complementary DNA modified by the of each time interval, tissue ~ulhiremedium was removed, fusion of the code for a mouse imrnunoglobulin heavy-chain and 2 ml of ûispase (Collaborative BiornedicaI, hc.) were signal peptide (34). CeHs rvere cuttured in 10 ml of Dulbecco's added to each culture well for dissolution of the Matrigel and modified Eagle's medium (DMEM) supplernenteci with 10% the formation of a single ce11 suspension; the number of celis fetd bovine serum, 100 IU/d peniàllin, 100 wg/ml strepto- was counted by a hemacytorneter. mycin, and 0.5 mg/ml geneticin (Life Technologies, Inc., Be- thesda, LMD).CelIs were maintained at 3iC in an atmosphere Chernoinvasion assay of 5% carbon dioxide and 95% air, in a humidified incubator, re with medium replaced twice per week. and ce& were har- The chernoinvasion assav method was sirnitar to that ported previously (2.CeU ~xltureinserts containing an 8-pm vested for experimental procedures cl t confluence. Mter re- 100 rg/cm2 moval of the medium, the ceüs were washrd in a phospha tr pore membrane, preco.~ted with Matrigel, tvere buffered saline (PBS)soIution and detached from the culture piaceci in a humidifiecl incubator at 37OC for 60 minutes to polymerize the Matrigel. The wild-type-bFCF and signal pep vessels after a 5-minute incubation with 0.05% trypsin-ethyl- DMEM mediamine tetraacetic acid (Life Technologies, Inc.1. The tntp- tide-bFGF ceUs resuspended in serurn-free supple i sin mented with 1.0'7c bovine serurn albumin (Sigma Chemical was inactivnted by reevposure of the cells to the culture MO) medium. After centrifugation at 1000 rprn for IO minutes, the Cu., St. Louis, were then carefully added to the upper ce11 pellet was resuspended in PBS and the number of cells chamber at ri concentration of -1.0 I04ceIls per rvell. Self- counted with a hemacytometer. ccinditioned sentm-free DMEM (24 hours) lrom each cell line tas used as a chemoattractant in the lotver chamber. Non- transfected tMH-3T3 fibroblasts were used as a control. The bFCF quantitation cells were allowed to invade ior 24 hours and were fked in The signal peptide-bFGF and wild-type-bFG F cells were ïhiI1ed 100% methanol. and Matrigel and ceils remaining in grown to confluence and were lysed on icr in 1 ml of chilted the upper chamber were mechanically removed with a Cotton extraction buffer (50 mmol/ L Tris, pH S.0, 150 mrnol/ L so- swb. The filters were then stained with hematoxylin and dium chloride, 2%hionidet P4O [Sigma, St. Louis, MO]) and mounted on glass siides. the protease inhibitors ethylenediamine tetnacetic acid-so- dium, pefabloc SC. pepstatin, and Ieupeptin (Boehringer Cellular migration Mannheim Corp., Indianapolis, IN). The ce11 lysates were then The tumor ceIl migration assny was developed from that of centrifuged at 16,000 x g for 20 minutes at -Lac to remove Sato and Rifkin (38) pretlouslv used to quantitate migration insoluble cellular debris, and the nipernatants tvere aliquoted of endothelia1 cells (43, 4-41. Confluent monolayers of the and stored at -YO°C. For conditioned medium, ceils were signal peptide-bFGF and ruild-type-bFGF cells were gmwn in gram to confluence, and the medium rvas replaced with 100-mm dishes. The cells rvere cut and the plate mark& by sem-free DPvlElLf for 48 hours. The medium was then col- pressing a razor blade dorvn onto the plate; then the blade lected and cenaifuged at 16,000 x g for ZO minutes at 4"C, and was gently moved to one side to remove part of the ceLi sheet. the supematants were aliquoted and stored at -80°C-Protein Mer scraping, the adherent celis were washed ~jce14th concentrations were detennined using a modifieci Lowry as- PBS, incubated for 16 hours in DMEM containing 1% fetal say (Bio-Rad, Ridunond, CA). The amount of bFGF rvas de- bovine serum, ttved with absolute methanol, and stained with termined using a sensitive soiid-phase bFGF enzyme-Wed aleoh01 eosin. immunosorbent assay capable of detecting bFGF in amounts Ceiiular invasion and migration in vitro were quantitateci as low as 30 pg/d (hiinhamCorp., hrhgton Heights, IL). by computerized image anaiysis (Quantimet 570, Leica, Deer- Determinations were made hm dupiicate weIls, and bFGF field, IL). For invasion, the number of cells that invaded the antigen conctintrations were caiculated as nanograms oi anti- Mahigel and migrateci to the lower nirfiice of the membrane t - gen per milligrarn of total protein- tvas counted from hipiicate wek. For ceU migration, the

Neurosurgery, Vol. 36, No. 4, April 1995 481 782 Gately et al. distance under x 10 magnifica tion, bétween the Ieading edge neal injection, supplementd with a subcutaneous injection of of the celIs from 10 random fieIds, was measured using the 0.4 mg/kg atropine sulfate. .himals were divided into two originaI crut as the origin (44. qua1groups, receiving either the wild-type-bFGF or the sig- nal peptide-bFGF celi line h 50-4 Hamilton syringe with a Zymography 27s-gauge needIe (Fisher ScientSc, Itasca, IL) was used to The signal peptide-bFGF and wild-tyebFGF cells were inject 10 4 of the ceII susptmsion (1 .O X IOb cells) into the nght grown to 80% confluence, then serum-containing medium frontoparietal lobe of the mice to a depth of approximately 2.8 mm. Bone wax waç appiiai at the injection site to prevent the was replaced with serum-free DMEM supplemented with 1% bovine serum albumin. The conditioned medium was re- reflux of cells, the cranial surface was flushed with sterile moved at 24 and 48 hours and cenûifüged for 10 minutes to saline, and the scalp was dos& with a nylon suture. On sediment celluIar debns, and the supernatant was collected awakening, each animal rvas examined to exclude neuroIog- and concentrated IO-fold with a Centriprep concentrator ical injury or other untoivard effects. (Amicon inc. Beverly, MA). Protein concentrations were de- The anirnals were killed 10 days after the implantation of tennined using a modifieci LOT assay and samples stored at cells. On the day of killing, the craniums were opened, the -60°C. To detect gelahases, &pl sarnples of the conditioned brains removed, and the maimal diameters of the tumors 62, 63, medium, standardized by the protein concentration, were measured in three planes: dl, coronal; sagittal; and applied to precast 10% acrylarnide, 0.2% gelatin gels (NOVU(, transverse. Tumor volume was calculated wîth the formula: (49). San Diego, CA) and electrophoresed for approximately 90 dl x 62 x 63 x 7r/6 Coronal sections were prepared for minutes at 225 volts. The gels were renatured in 23% (vol- light microscopy by fixation for 24 hours in 70% ethanol. The and urne/volume) Triton X-100 in water and developed ovemight tumor surrounding brain were then ernbedded in paraf- at 37°C in developing buffer (NOEX). The gels were stained fin and sectioned at 5-grn thickness. rvith Coomassie Blue R-250 (Life Technologies, [nc.) and ge- For electron microscopy, 1-mm cubes of both tumor and latinolytic activity detected as a clear band ot lysis against the normal brain adjacent to the tumor were irnmersed for 2 blue background of stained gelatin (43). hours in chilied 3% gluteraldehyde in 0.1 rnrnol/L cacodyiate buffer. They were then poctfixed in osmium tetroxide, dehy- Corneal angiogenesis assay drated, and ernbedded in epon. Thin sections were staind with 3% uranylacetate and lead citrate and examined with a For quantitation of angiogenesis, the signal peptide-bFGF Phillips (Mawah, NJ) 301 transmission electron microscope. and wild-type-bFGF turnors, grown in athymic mice, were To determine the rate ot' tumor ceU proliferation in situ in hanrested for implantation into comeal micropockets (14,441. the brain, a monoclonal antibody to bromodeoxyuridine Twnty male New Zealand White rabbits [Harian Sprague- (BLjdR) was useri (81. Thrrr houn before iulling, the athymic Dawley, Indianapoiis, IN) were anesthetizd with 5 mg/kg mice received 60 mg/kg BL'dR and 6 mg/kg fluorodeos-vuri- uylazine hydrochloride and 35 mg/ kg ketamine hydrochlo- dine (Sigma), a corn petitive inhibitor ot thymidine upta ke, by ride by intramuscular injection. The rabbit cornea was flushed intraperitonea1 injection. Sections were incubated with 2 N with steriie 05% tetracaine hydrochloride and the globe prop- hydrochloric acid for 30 minutes and with borax buffer, pH tosed. Using a stereomiaoscope, a comeal micropocket was 8.5, for 5 minutes and were then washed three tirnes in PBS. fonned to within 2.0 mm of the limbus, and 1-mm fragments Tissues were incubated with 309 hydrogen peroxide in abso- of the tumors were placed into the micropockets. The corneas lute methanol (1:9) ior 10 minutes, with 10% normal goat were examined daily after the implantation of tumor, until the semrn for 20 minutes, and rvith the anti-BLidR monoclonal time of kiIling, 7 days after implantation. A total of 120 antibody at a 1%) dilution for 30 minutes at 3FC. After corneal micropockets were made, divided equally into three washing in PBS, tissues twre incuba ted wit h a biotiny iated groups (n = 40), containing: 1) signai peptide bFGF; 2) wild- secondary antibody (2-vmd Streptavidin-Biotin System, Di- type-bFGF tumor Fragments; or 3) blank micropockets as a mension Laboratories, hiississauga, ON) for IO minutes at control for injury. The angiogenic response was quantitated room temperature, washed again in PBS, and incubated for 5 ushg an ocular miaoscale, measurïng the vesse1 length from minutes with concentra ted streptavidin-peroxidase conjugate. the limbus to the maximal edge of the t-ascular sprout The This rvas loalùed by a ha1 incubation of 5 to 10 minutes vascular density was score3 using the idlot-ùig sale (45): 0, rvith 1 rng/ml diaminobenridine in PBS prepared with an no biood vessets; 2, 1 to 10 blood vessels: 2, more than 10 qua1 volume of 0.039 h~drogenperoxide. The slides tvere vessels, Ioosely arranged (details of the iris could be viewed washed in water, counterstained with nudear fast red, dehy- between gaps in the growing vesseis); and 3, more than IO drated, deared in xylene, and mounted. Ceüs that incorpo- vessels tightiy packed (the iris coutd not be viewed, and an rated BUdR showed a black nudear pigment- The BUdR angiogenesis index was calculated multipl-ying vesse1 length tumor ceii-hbehg index was determined as the percentage of by density). Iabeled c& in reiation to the total number of cek scored. %mintumor model Twenty male athyinic mice (Harlan SpragueDawley) Nu/Xu homozygous) between 25 and 30 g were anesthe- Resuits are expressed as r *ean standard deviation. Sta- tîzed with sodium pentobarbitd, 65 mg/kg, by intraperito- tisticaiiy significant ciifferen 3 between means were deter-

Neurosurgery, Vol. 36, tto. 4, April 1995 bFGF and Neoplastic Invasion 783 mineci using a ho-tailed Student's t-test for unpaireci Sam- ical riifferences after 24 hours. 8y contrast, there were signif- ples or a one-way analysis of variance. Differences with P < iï.int yuaIit,itive ditterences after 2-5 hours' growth on blatn- C 0.05 were considered statisticallv significant. gel-conted plates; the signal peptidebFGF cells formed trrmching donit)s,i\*liert..is the wild-type-bFGF cclls tormed RESULTS individual aggregate colonies. At 48 hours, the signal peptide- bFCF crlls continued to form rxtensiw branching colonies bFGF quantitation (Fis.?A); in sharp conhst, the wild-type-bFGFceils remaineci The amount of cd-associateci bFGF was slightly higher in as individual ceIl clusters (Fig. ZB). There was no difftmnce in the signal peptide-bFGF cell done (29.6 ng/rng protein) than the number of signal peptide-bFGF or wild-Srpe-bFGF cells in the wild-type-bFGF cddone (24.9 ng/mg protein) (Fig. growing on Matrigel at 24 or 423 hours. ZA). bFGF was aImost undetettable in the conditioned semm- (5.9 kee medium of the wild-type-bFGF ceIl clone pg/mg Chernoinvasion assay protein). By contrast, it was detected in the conditioned me- dium of the signal peptide-bFGF ceIl clone (72.6 pg/mg pro- When Matrigel was excluded in the chernoinvasion may, tein) @. ZB). qua1 numberç of the wild-type-bFGF and signal peptide- bFGF cells migrateci through the 8-~rnpores to the hwer In vitro invasiveness using Matrigel surface of the membrane. The control, nontransfected NIH- 3T3 fibroblasts, did not invade the Matrigel. The wild-type- Growth of the wild-type-bFGF and signal peptide-bFCF bFGF celIs were more invasive than nontransfected celis, with cells when sparsely plated on plastic showd no morpholog- 232 z II4 cells invading. The signal peptide-bFGF cells were the most invasive, with 418 s 144 celis invading the Mahigel and crossing the membrane (P = 0.05) (Fis. 3).

Cellular migration Lsing the migration dssay, the wild-type-bFCF cells spread a distance of only 244.5 r 5.3 gm into the denuded area (Fis. 4A); in contrasr, the signal peptide-bFGF cells rnigrated sig- nificantly iarther, to a distance of 871.2 = 9.0 pm (P = t1.001) (Fi$ 45).

w-bFaF **== B WI cm FIGURE 1. A, amount of ceIl-associated bFGF was quanti- bted by enzyme-linked immunosorbent assay, demonstrating Chat the wild-type-bFGF (wt-bFGR and signal peptide-bFGF (sp-bFC8 cell clones express comparable levels of bFCF. By i . . contnst, 8 demonstrates that the signal peptidebFCF celi FIGURE 2. Appearance of signal peptide-bFCF (A) and wild- clone released bFGF into the serum-free conditioned me- type-bFGF cells (B) at 18 houn after plating on basement 1 dium. membrane Matrigel,

Neurosurgery, Vol, 36, No- 4, April 1995 784 Gately et al.

terstitial cdlagenase. There t\-.~s a slight increase in these bands in the signal peptide-bFGF ce11 line-conditioned me- dium ,ILAS Iiours (Fig.3. .-\ 1iiglit.r mr~lrcularwi~ht zone of lysis corresponding to apprusimately M, 92,000 was clearly evident in the 48-hour signal pep tide-bFGF cell-conditioned medium (Fis. 3, but far les prominent in the conditioned medium of the wild-type celk.

The angiogenic activity ot the two cdclones implanteci in the rabbit comea assay showd an innease in vascular den- sity producd by the signal peptide-bFGF cells (Krble 1)but no rl-DFW *vw significant difference in the vesse1 lengths or in the overd 40 FIGURE 3. The number of signal peptide-bFCF (sp-bFGfi angiogenesis index (45). AS a control, the control micro- cells that invaded the Matrigel and crossed the porous pockets that contained no tumor implants showed no angio- membrane is significantly higher than for the wiid-type-bFGF genic response. (wt-bFGn cells. ', P < 0.05, Student's [-test. None of the control, nontransfected NlH-3T3 fibroblasts were able to in- Brain tumor mode1 of invasiveness vade the Matrigel and cross the membrane, Electron microçcopy revealed that the wild-type-bFGF ceIls grew intracerebrally as tuell-circumscritred tumors, with sharp borders that separated the tumors from the adjacent normal brains (Fis. 6.4. A moderatelv electron- dense material was detected, which constihted a pseudo- capsule that seprirated the tumor cells hmthe neuropil (Fig. 6B).In contrast, the signal peptide-bFGF cells formed tumors with ill-defined borders with infiltration of the tumor cells into the neuropil (Fis. 78).There were trequent distortions of the neuroptI .ind a distinct nbsenct! of the pseudocapsule separating the tumor rnass trorn normal neighboring structures IFy. .-BI. Both ce11 lines were tumorigenic, forming tumors in all mice a tter subcutaneous and intracerebral injection. No significant difference in the intracerebral turnor volume was found be- h4.een the wild-typtrbFGF, 26.7 = 11.6 mm?,and the signal

FIGURE 4. After the razor blade cut (arrow), the wild-type- bFGF ceIls were less migratory (A); however, the signa! pep- tide-bFCF ceIls were highly rnigratory, with a significantly Farther spreading distance into the denuded area (BI (P < 0.0001, Student's t-test). FIGURE 5. Zymograrn gel illustrating the digestion of gela- th, with zones of enzymatic activity characterized by nega- tive staining. A, wild-type-bFGF cell-conditioned medium at Zones of gelatin gel Iysis corresponded to gelatincases of 24 and 18 houn; 6, signal peptide-bFCF cell-conditioned approximately M, @,O00 to 72,000 were detected in the se- medium at 24 and 48 hours, The signal peptide-bFCF ceIl rum-free conditioned medium of the wild-typebFGF and done at 48 hours reieased significant amounts of M, 92,000 signal peptide-bFGF cells. Zones of Iysis were ais0 seen at gelatinase and M, 42,000 activated interstitial collagenase :' appmrimately h.Ir12,ûûû, which correspond to activated in- (arrctws). bFGF and Neoplastic Invasion 785

TABLE 1. Density of the Angiogenic Response to Wild-type- and Signal Peptide-Basic Fibroblast Cro~thFactor Tumor Fragments"

' Mean * standard dcviation, wt, wild-tvpe; bFGF, basic fibroblast growth factor; sp, signal peptide. b~=0.01.

FIGURE 7. A, neoplastic cells transfected with signal peptide bFCF display an invasive phenotype and infiltrate the neuro- pi1 at the tumor edge, spreading along perivascular pathways (original magnification, x 400). B, the infiltrative spread of signal peptide-bFGF cells occurs without a deiined border, and signal pep tide-bFCF cells are-found invading the neuro- pi1 (original magnification, ~4500).N, neuropil; T, tumor cells.

DISCUSSION The mechanisms ot invasion are central to the biology oi mali~nantbnin tumors. The microscopic spread of mtilignant giiomas into the surrounding namal bnin results tn th& local remmence (3.47 ,lnd is a factor th~taciuunti t'or the FIGURE 6, A, tumor cells (T) lacking the signal peptide do failure ot traditional :hrrrzpies. Tu study the role or bFGf in not invade the neuropil (hl) (original rnagnification, ~400); neoplastic invasion in the brain, we usai cells tmnstecteci 8, the tumor border is well defined, with tumor cells sepa- with bFGF expression vwtors, with and without a secreton; rated from the neuropil by a pseudocapsule (arrow)(original signa1 squence. CVe now report tha t cttlls transfected r\*ith the magnification, x 2500). bFGF comptementary DNX fus& to a secretory signal pep- tide sequence are invasive in vitro and in situ in the brain. These iindings are important, because the!: 1) provide insight peptidebfGF tumors, 39.1 = 203 mmJ.The BWR-laboling into possible mechanisms of neoplastic invasion in the brain index of the wild-type-bFGF tumors, 35.5 s 53%. was slightiy pointing to bFGF and the iLI, 92,000 gelatinase as potential higher than the signal peptide-bFGF tumors, 27.5 r 5.4% molmlar deterrninants; 3) underscore the close relation be- (P = 0.03 (F*. 8). tween angiogenesis and neoplastic invasion; and 3) pro~idea

Neurosurgery, VOL 336, No, 4..-\pril 1995 786 Gatety et al.

Sa.. Lin-lp. rvhich art. known to induce the expression of the M, 93,WO gelatinase (36). Gelatinases have been linked to angio- gt'ne.iir; (27 and nwplastic invasion in several turnor systcms 44 (20L including human glioblastorna (32). The signal peptide-bFGF and wild-type-bFGF ce11 clones formd tumors in al1 animais both subcutaneously and intn- 3a cerebrriIly. with no significant difference in the tumor vol- umes and only a slight difference in the BUdR-labeling index 2G (Fis.8). These findings demonstrate that the invasive pheno- type of the signa1 peptide-bFGF cells is not related to tumor sire or the proliferative nte. Invasiveness, however, was 1 e linked to angiogenesis. In the comeal assay, the signal pep- < tide-bFGF tumor fragments induced an angiogenic response with increased vascular demil. In situ in the brain, the signal peptiJebFGF ceUs were iound invading aiong the preexisting FIGURE 8, Native-bFCF and signal peptide-bFGF (sp-bFCR hlood vessels withiri the bnin parenchyma (FisIS), simifar to intracerebral tumor cell 8UdR-labeling index. The wild-type- that described ter rat C6 glioma (29, 521, VX2 carcinoma (491, bFGF (wt-bFGF) tomor cell-labeling index in situ in the brain and ,I knorvn route of spread for malignant human glial is significantly higher in comparison with the signa1 peptide- tumors cells (371. These ricita suggest that invasion in this (6, bFGF tumor cells, *, P < 0.05, Student's t-test- mode1 may not oniy be hnctionaily related to angiogenesis 71,3,27). but may be angiogenesis dependent (30),consistent r\.ith the finding that agents :bat suppress c~ngiogxwsisaIso reproducibIe mode1 to study the mole~ularmechanisms oi inhibit invasion (6,7) or the production ut' proteolytic en- bFGF-related neoplastic invasion in vitro and within the zymes (43, 44). brain, ln summ~lv,rve have shuwn that NIH-3T3 fbrobiasts Invasion is an active process involving a cascade of cellular transfecteci rvith a signal peptide-bFGF expresion rector .ire events that include adhesion, migration, and proteolytic deg- invasive in titro and in situ in the brain in comparison with mdation of the extr;icellular matrix (20). Ln this studv, the cells transtected with the r-ild-type-bFGF gene. Cclntinuing signal peptide-bFGF-tnnsfected cell cione was highly migra- studies rvill ioas on the role oi bFCF in the invasivenesç ot tory (Fix. 4B) anci invasive. fyrnogrnphic analvsis d the con- human glial tumors. RwentIy, nontmnsiormed human ktal ditioned medium hmthe r\.ilri-type-bFGF ~ndsignd pep- .~stroqtestransfectrd rvith the signal peptide-bFGF rxprrs- tide-bFGF ce11 ciones dernonstrated elevated A[, Q2.UW sion vtxtcir were iound tu be transformeci in vitro (G~telyS. gelatincise only in the signal peptidebFCF clone (Fis.5). fhis Sort GA, Brem S; rnanusmpt submitted). Taben together, finding identifies an important role for proteinases, pirticu- these iinciings demonstrate bFGF as an important moIeculx lady the M, 92,000 gelritinclse, as a molecular rnediator of ïne~iiatorot' the malignant phenotype of human glial tumors. invasivenijss in the bnin, simiIar to that describai in tissue extracts r3f human gliublastoma (32). ACKNOWLEDGMENTS The molectrlar meçhanisms responsible tor the elevated gelatinolytic activity tif the signal peptide-bFCF celli remain This ruork was supported in part by gmnts irom the unknown but couId be linkecl to the extracellular relrase of ical Research Couna1 of Cmxia, L.S. Public Hralth Senice bFGF (23).bFGF tvas detected in the senim-t'ree conditiond Gnnts CX6C)IN and CA5781 (to SB), L.S. Public Health medium of the sign.11 peptide-bFGF ceII clone (Fis. 16). Be- Service Gnnt C.43392 itr, LN,the! Mime FrIlwv..;hip (tuSG), cause the signal peptide is responsible tor directing the extra- and a gmnt from Minois Seurotibrornatosis [nc. (tu SG and cellular relerise of proteins, it is not surprising that bFGF was STL Ctre thank Dr. Patricra Dimond and Mark h~I~C~Itcighof detecttd in the medium. However. previous reports (34. 35, Coihbontive Biomdic;il for generousIy ~rovidingthe Maki- 4) did not fincl bFGF in the medium ot the signal peptide- del invasion rhcimbers. Ll'e dso acknowledge the e\pert t~h- bFGF cells, perhaps because the enzyrnelinked immtinosor- ntd r~s.ilstanceot blarperi te Et'uto~z~eh-ObaJ LI, Dry rid '\.an- bent assay method ot bFGF detection is highly sensitive. cic, \Valter G togowshi, Seeta Drsai, 'mi SanJn Lavoie. EutracelIular bFGF is a potent stimulator of urokinase-type plasminogen activator ( 121, an enzyme identified at elevated Received. %ptember 14, 1994. IeveIs in giiobtastoma multiforme 1 19). Urokinase is respon- Xccepted, Norember 7, 1994. sibIe for the conversion of inactive phsminogen to activated Reprint requests: Dr. Steven Brem. Director, Neumurgitzil Oncol- qp-,Division of Nrurologiml Surgery, Northwestrrn %ternoria1Hos- plasmin (ln, capable of converting the inactive matri. met- pitd, 233 East Erie Street. Suite 500, Chica~o,fL 6061 1-2906. alloproteinases, including the gelatinases, to the active en- zyme (17). Aternatively, transformation by the signal pep REFERENCES tide-bFGF constnict, or the eutracelular export of bFGF from the signal peptide-bFGF ceils, may induce the production of 1. Abraham J.4, bt'hcing JL. Tumolo A, Friedman J, Hjerriid ka, other factors, cytokines, tumor neCros% factor-a, or interfeu- Cospodarowicz D, Fiddes IC: Human basic fibrobiast grorvth bFGF and Neoplastic lnvasion 787

factor: Nuclturtidt. sequene and pnamic i~r!~.iniz;itiiin.EMBO J W. L.indliu BI. Krvaan HC. brrusio E, Brmi SS: EIevated levels of S:3r-L52.., 1'4%. undm.ise-tyfc. pIL~smino~r.n.~iti\.ator ancf plùsminogen dctivator 3 Albint A. Irvamato Y. Kleinman HK. Mmtin CR. A,inrn?;crn SA. inhibitor type1 in m.ilign.int human brain tumors. Cancer Res KcizlrirvsLi fXI, 5lsEr\*,in KN: A rbipiJ in vttnp .i>+.iy tor qu.iiittt.~t- 3:i It&l ItM, IYW ing the invasive potential of tumor cells. Cancer Res 47323% 3. Liotts LA: blechanisms oi mncer invasion and metastases. in 3245. lys?. K.iiz;er HE, Lima LA itkisk L;itrrt~rGrorcth ,uid Pn~p?;sîort:hfltr- 3. Alvarez JX, Baird A, Tatum A, Daucher 1. Chorsky R. Gonza1t.z twa* of Ttrttrrw Dr'~~hj~>~r'l~toit t/w Hmt. Dordrecht, Kluwcr .\LX- AM, Stop* EG: L~ualizationrif bLwictifrt+ia~t growth tattor an~i deniic Publishers, I9HQ FF 3-3. vclscuhrendothelid ceil growth factor in human gli.11 ntwpl~s~.II. Li»tta LA, Steeg PS, Stetier-Stevenson WC: Cdncer metastasis and Mod Pathol530>307, 1993 angiogenesis: An imhalance of positive and negative regula tion. 4. Blei F, CViIson EL, Mignatti P, Rilkin DB: Mechankm of action of CeIl M:327-336, 1991. angirwt.1 tic stcroiris: Supprtvsion of plasniincitp iirtiv.itor xtiv- 22 Lund-lohansen M, Fonbeq, K. Bierkvig R. LemmOD: Eff~tsof ity via stimulation of plasminugcn activat~irinhibitlit synthesis. J gnwth ractorç on a hunun glioma ce11 line during invasion intrt Ce11 Physiol IS?:iô&378, 1993. rat bnin aggrc'gates in ~ulture.Acta Neuropathol (Beril S4:IY0- S. Brem S. Tsinaclis AMC. Gately '5, Cross IL, Hrrblin WF: Imrnii- lQ7, IW2. noioc~itilationot basic fibroblast grinvth t,ictor to the microv.is- 3. Mignatti Pi'. Ritlin DB: Biolvgv and biwhemistry of proteinases culature of hurnan brain tumors. Cancer -ii.2nlXhSt). 1W3 in tumor invrision. Physiol Rev 3:tbI-195, 19% o. Brem S. T~~inaclis.-MC, Ligzal; D: .4ntiï~t~prtrcitnient inhibits 24. \lcwrkun RS: Supprez;si~~nkir basic fibroblast growth ractor CS- pseudopodial protruston and the invasive ?preail or 4L cJios.ir- prtM»n hv antisenrie ciligonuclwtidtts inhibits the growth of coma tells in the rat bnin. Neurosurgecy h-3I-Yn. !"W. tr.insiormril ~istrcicvtes.J Bi01 Chem ZO~:~S-I'~,I Wl. 7. Brem SS. Zigzaç D. TscinticlisAXIC, G,iteIy 5. Elkci~b~il-P. Brien 3. Jlorri_son RS. Giordano 5. i,imahwchi F, Hendrickson S. Berger SE: Inhibitirin angiur;enti;i.s md tuni~jr~rot\.th in tlit. br.11~: 11s. P.lli~et~*skiK: BJ-t~ ribrcibiast gruivth txtor ekpressi~miti Suppression of endothelia1 ceil turnover bv penicillamine and the required for donogenic grorvth of human glioma cells. J Netuosci depletion of copper. m ,~ngiopnicci~t.ictor. Am / Pathol 137 Res 3:32-iO9. 1943 1121-L 142. IW). lb. Jtiini-irn =.Cr~)';s IL Hdrblin LVF. Rrill!. TU. LaS.ilci P.A. Alter- 8. Brien S. &gag D, Brem S: hpid in situ cellular Linetics or man RL, hlosk~ljR. Kcirnblith PL, Dater DL: Basic iibroblast intrricerebnl tumor angiogentrjis iising ,i mi~niiclonalmtibtdv tu crcwth f,ictor-lihc ,icti\.itv ,incl receptors are etpr~sdin .i hu- bromocieosyuriciine. Seurosurgery I?:lt t'lu. 1Wq. nian &)ma iell linr. Cancer Res ?0:5l-C3Y.IWt). Q. de Ridder L. C,itliIiutv L: Inv~si~intir hunun br,iin tumirs tn 27. \tLh;c.~tt'iI~D. RiAin @6; Jlemtiranr and rnb1trix loc.ilizatii~n d vitro: Relationship to clinical evolution. j Seurosurg 239-593. proteinases: A tommon therne in tumor ceil invasion and angio- 1990. pwis. Biochim Biophys Acta W:b745,IW. 10. Engebnaten O. Bjerkvig R. P~derwnPH. L.ierum OD: Eitects ot 2s. ILloser RP: 5urget-y tcir $iorna relapse: Factors that intlurncr ,i EGF. hFGF, NGF. and PDGF tbb) on ceIl proliteratiw. rn:cntrip i.ivcinbIe outairne. Cancer oh351-3W. Io#. and invam.t, capacitie ot hurnan bmn-tunrcir btip+" in vttrca. 3. \.I+I~LI \ , 5,t~LiU. ,4L>\..1931, i-iir.~h\-.iK: In~.mmoi e\pbr- Int J Cancer 53:209-$14. lQQ5 irnr.nt.11 nt bnin tttmtir- Exiy morpholoçical ihm';rs idlorvin~ 11. Finhelstein SD. Bisi~hP. Sowaf Ti', H,inci CM. C1imtrnc;rn 5. miirriinjtirr.on Sn ;i~~rrn.iie1l.j. Acta Neuropathol (Berl) 50: Finch PW: Hi5tdogrcal characteristics .ind e\preszron Lit mdic 11,-15, t'w. and baic iibroblLwt grow th tactor gen~in tntnzertlbral teno- 3). Siiu&i RF. Tchao R. Lr:~it~-. R~yeij5. \l*etntrer$ KA.. F.inning P. KI,i.gsbrun IL I: Clwncteriza- Prfircipf~5cit'11~-t'-Tt'ci1t~ulo~~y-~~lt~dic~~ ttvLBad Birkhm3er C'dro, riim or tumurs producd t-y 'ignal peptidebasic fibrobl~stgrowth 19Z. pp -1214Z. r,iïtor-tr.~no;rc)rmtldc&. J Ce11 Biochem ?9:I>13, 1QS9. ID. Gtiiss lt. .'clorrison RS, Eidsvoog K. Herbiin LVF, KimtkIith PL. th. Ro-ienbeq C.A. Dencutt JE, Mc€uire PG, Lotta LA. Stetfer- Dater DL Basic ribroblast bmwth iactot: .A potenti~l.iutomne 5tevrnsor. LYG: Injury-inJucd 92-kilrdalton gettinae umkinase regdator of hurnan ghma ce11 grotvtfi. J Yeurosci Res 27:oSY- expression in nt brsin. Lab Invest 71:417-&12, 1994 696, 1990. 31. Rusdl ES. Rubenstem LJ: Tumors of the centra1 nruroepitheiid 17. Kwaan HC: The pIasminogen-piasmin liystem in rnalignancy. origin. in P~ltliulo~yL+ T:,trwn of th< rUer;.uits 5-wt~.Baltimore, Cancer Metastasis Rev 1 1379 1-32 1, 1993 CVilIiams iG Wiikins. 1989. ed S. pp 8S89. IS. Lemm OD. Bierkvig R. Steinsrag SK, de Ridder L- Inta~iivrnt.4- ;S. %to R. RiAin DB: .i\uto~lrneactivitit! cii basic tibmblast ptvtft of primary brain tumors. Cancer Metastasis Rev 3:33b,19s. iactorr Replation ot rndothelial cell movrment, plasminogen 788 Gately et al.

activator synthesis, and DNA 5~nthtu;is.J CeII Biol 1021 199- The tinding that the signal pq?tidt'-bFGF cc11 cIcme waç cap- 1205, 1988. ble ot secreting bFGF iij evidence of the potential autocrine Stefanik DF, Rizk;itla LR, %i AS. Gol~ibl.itt .A. Ri~k~1l1.1W\I: rdt. ot bFGF. hcidic mci basic tibroblast growth factors dre prwnt in glioblas- These iindings in an experinientai mode1 system, whicli toma multiforme. Cancer Res 51:~6U-576, 1991. does not involve human astro~yttw;,cleariy needs to be? eval- Stoker hi, Ghendi E: R~ylationor cell muvement: The me uated in other more Jirectly relateci mociels. Nonetheles, the togenic cytokines. Biochim Biophys Acta 1U2Y 1-102. 1991. data presented serve ta Surther support the potential signifi- Takahashi fA. Fukumoto M. Igsnshi K. Ori.1 Y. Kikuchi H. Ha- tanaka M:Correlation of basic fibroblast grotvth factor expression cance of bFGF as a relevant peptide in the evolution ot ma- bels with the degree of malignancy and wscularity in human Iignant gliomas. gliomas. j Newsurg 76392-798, 1992. Tahhashi IA, Mori H, Fukumott, M. Iganshi K. laye M, Oda Y. Kikuchi H, HaLinaka M: Gent! expression ot tibroblast growth factor in human gliomas and meningiomiis: Demonstmtion ai One tvay to determine the function of a given growth factor cellular source of basic tibroblast gmwth factor mRNA and in ce11 biology iç to up-regulate the expression of the growth peptide in Nmor tissues. Proc Nat1 Acad Sci USA 87571tl-3714, tactor by the celI using gene-tramter techniques. In this manu- 1990. Takano S, C~telyS, Jiang JB, Brem 5 h diaminoanthnquinont!~uhone script. Gately et JI. have taken the bFCF gene iused to 'in inhibitor of angiogenesi';. J Phannacol Exp Ther 271:1077-1033. important signa1 peptide sequence and tnnsfectrd the NIH- 1994. Yi7 41line. Their results shoiid that the transfected NIH- Takano S. Cateiy S. Neville ME. Herbtin CVF. Gros5 IL. Engelhard 3T3 cetls acquired a mure inv,ii;ive phenotype in vitro thm H. Prnicone hl, Eiclsvoog K. Brrm 5: Çunmjn. .in antic~ncrrand controls and a tencienc?; ior locd inv'ision in vivo in nrhymic: angiosuppressive agent, inhibits endothelia1 cell binding of bdsic mice. fibroblast growth i~ctor.misration. proliteration, ~ndinduction One of the mechanisms 5~ which the bFGF transiecticm of 3:ZbI+ urukinLilir-type plC~srnincipn.~ittv,it~r. Cancer Res iwrAed is bu incrrasing the expression ot the .I[,9Z,UOr) gelci- 2660,1994. tinases (ppe [V caliagenase). [ncreased expression of this T'~margo RI. Leong KCV. Brem H: Crriwth inhibition 05 the QL gtioma using pulymttrs to releasz htlparin .mi cc~rtismeacetate. J enzyme tuouid iacilitate ~enetr~itiontlirough .\.ic~trigel.Ex- te': ,i Neurooncol 3: 13 1-135, IWil ~ctlyho\u bFGF up-re&i the M, 32.il110 ~t.l.itinast~tter Vtodavsky 1, Folkman j, Sullivan R. Frittriman R. Ishai-Michaeli R. tnnstection rvould be an interesting question to answer. Sasse J, Klagsbrun .LI: Endothelid cel14erivd basic fibroblast This is an excellent report demonstrating the power ui gene growth factor Synthesis and deposition into wbendothrlial es- transter experiments. The results suygest that bFGF phys a ~~celluLrrnatnx. Proc Nat1 Acad Sd USA S:EYZE%,lW7. rrile in ce11 invasion. 1 lcwh i'c~n\.itrdto kiture eupriment5 by \Uilkinson HA, Litotsky NS: Thr.r.ipy ci! bran tumors. Conternp the mthors ln tvhich either astrocytes clr cistrucpmaceIl !ines Neurosurg 16: 1-8. 1994. ,ire used in similx e\periments tcb Jetermine hoiv bFGF gene Yciyon A. K1apbn.m M:Xutrxrintr regul.rtton oi tell pwth md transrer ~ikt~irills ot the centrd nrrvous systern. transformation by basic tibrotrlnst growth ixtor. Cancer ,Metas- tasis Rev 4: Ir3 1-202. 1490. L~gzagD, Brem 5, Robert F: Swvsiscu1.~nz,itionand tumor growth in the nbbit brain: h mode1 for e~prnrnent~~fstudies or mgiogenesis and the blod-bnin barriet. Am J Pathot I-i1:3hl- This c~rticletiy &tel? and colleiigues provides acfdition,~l 311, 19S8. evtdencr for the importance fit cfetermining the rnol~uiar Zigzag D. Stiller DC. Satu, Y, Rttlin DB. Burstein DE: :nimunu associateci wth migratm and invmon of rnalipant histochemical Iocalization or basic fibmbi,it grotvth fxtor in giroma cells, In this study, the mthors zihoii. that .I clone ot asnocytomas. Cancer Res 5039>ij'JS, IWO. nontumorigenii tells trLin.;ieited :i-ith the signal peptide A. HK: Zsmù T~murnhi, [noue Thrtir-dxmewionai obrewiitions bFGF is h~ghiyrn~ograto? and int.;i'iive ,inci is as';oci.itrd ri-ith on rnicrovascwlar grotvth in nt g1iom.i usin5 a v~scui~irastins &vateci Ieveis oi gelatinase. pruperties that h'ive ben as- method. J Cancer Res Clin Oncol 1 II:39M02. I9Y 1. cribed to human gliobfastorna. This wems to suggirst a pos- sible mi>lecuI,ir thera~eurictclrget. Hor\xwr, it is dl! not dear COMMENTS that trrain tumor cdl- in 5itu zornrn~mIyiontain bFGF ivitit a sicput sequence. that the! e\~ressth&. ,inci thdt it has an The peptide growth tactor basic fibrriblalit growth factor impact on those cdls under normal cunditions. lonetheless, it (bFGD has been shown to have rnarked etiects on the profif- is redistic to sa? that angiopesis and invasion are important eration of astrocytoma celis, and it is cIearly a relevant factor eIements in gIioma ceii growth and destructivtrnes, and a in the grorvth of malignant astroqtomas. It is a peptide with rrasonable case is made that bFGF cmplay a significant role documented mgiogenic effects, another key ieature of human in this açtivity. Thererore, this wrkadds important informa- giioblastoma. tion to our undershnding ot this aspect of malignant gliomris, in this work, the bFGF gene tuseci with a signal sequrnce and the authors' erfo* are noteworthy. has been shown to enhance the invasive properties of NIH- 3T3 ce&. nie Fuidingof Ioal invasion in the brains of ath-pic Robert L. Martuza mice &ter transfection of the cek 3 of particular relevance. !t:ldtitzprr,District ùf' Calrtitdrirr

Neurosurgery, Vol. 36. ,No. 4. April 1995