Tumor suppressor genes and medulloblastoma

Journal of Neuro-Oncology 29: 103-112, 1996. © 1996 Kluwer Academic Publ&hers. Printed in the Netherlands.

Tumor suppressor genes and meduiloblastoma

Philip H. Cogen 1and Jeffrey D. McDonald 2 t Pediatric Neurosurgery, Children's National Medical Center, Washington, DC, USA; 2 Neurosurgery, The University of Utah, Salt Lake City, UT, USA

Key words: medulloblastoma, chromosome arm 17p, tumor suppressor gene, ABR gene, prognostic significance Summary Although primary intracranial neoplasms are the most common type of solid cancer in children, little is known about their etiology at the molecular genetic level. Recently, studies have shown that a class of genes known as tumor suppressors play an important role in the origin of several different types of human tumors, including those located in the central nervous system (CNS). Using a variety of techniques, selective loss of DNA sequences has been identified in tissue specimens from children with medulloblastoma, one of the most common pediatric brain tumors. The most consistent losses to date have been shown for probes located on distal chromosome arm 17p. Although the known tumor suppressor p53 is located on this chromosome, and deletion and mutation of the p53 gene are the most common genetic events in human cancers of many types, such alterations have been infrequently detected in medulloblastoma specimens. These results suggest that inactivation of another tumor suppressor gene or genes located on 17p is important in medulloblastoma tumorigenesis. Deletion of 17p has also been shown to have implications for clinical management, as the loss of DNA sequences located on this chromosome arm is strongly associated with a negative prognosis for these patients. The identification and cloning of this tumor suppressor gene or genes will aid in understanding of the pathogenesis of medulloblastoma, as well as guiding the development of novel and more effective strategies for a cure.

Introduction Brain tumors are the most frequent type of childhood cancer second only to leukemia, yet there is a marked contrast between the significant improvement in understanding the pathogenesis and improving the treatment for leukemia, and the paucity of knowledge and often dismal outlook for brain tumors. Most children who suffer from a brain tumor will die as a result of disease, this despite the advances in surgery, radiation, and chemotherapy made in the past several years [1]. Significantly improving the therapeutic outcome for these children must involve not only the identification of

better treatment modalities, but as well an understanding of the pathogenetic basis for the tumors themselves. Studies over the past several years have identified at least two groups of genes that play an important role in the genesis of neoplasia [2]. The oncogenes are dominantly-acting growth enhancing genes whose transformation by mutation, deletion, or rearrangement leads to uncontrolled cellular proliferation. The tumor suppressors are recessively-acting growth control genes whose inactivation by mutation, deletion, or rearrangement also leads to neoplastic transformation. Using the analogy of the uncontrolled cellular growth leading to the for-

104 mation of a cancer as a rapidly moving vehicle heading down a hill, the oncogenes are the accelerators, whereas the tumor suppressors are the brakes. Thus, the activation of an oncogene that causes the cascade of events leading to neoplasia has the same end result as the inactivation of a tumor suppressor gene. Several different lines of evidence have suggested that tumor suppressor genes are involved in the etiology of medulloblastoma, one of the most common childhood intracranial neoplasms [3-8]. Cytogenetic and molecular genetic studies have shown the consistent loss and rearrangement of several chromosomal regions in tumor specimens, findings that often signal tumor suppressor activity. Recent studies have provided additional data that further localize the sites of these potentially novel genes, aiding in their identification and cloning [9].

The tumor suppressor hypothesis Initial evidence for the role of tumor suppressor genes in neoplasia came from the elegant epidemiologic studies of Knudsen on the sporadic and hereditary forms of retinoblastoma, the most common ocular malignancy in children [10]. Although the many experiments performed since Knudsen's seminal work have resulted in the identification and cloning of other tumor suppressors, the first of these genes to be identified, the retinoblastoma gene (Rb), remains the most useful paradigm for explaining the tumor suppressor mechanism. Knudsen observed that two types of retinoblastoma were clearly distinguishable by their distinct clinical patterns [10]. The hereditary type, an autosomal dominant with high penetrance, presented in early infancy, often with binocular lesions. These patients also showed an up to 1000 fold increase in second malignancies, particularly osteogenic sarcomas, occuring both within and outside of the site of the radiation used to treat the retinal tumors. In contrast, the sporadic type of retinoblastoma was more often seen in older children, was typically uniocular, and was not associated with an increased incidence in secondary malignancy. Knudsen reasoned that this pattern of clinical

presentation of retinoblastoma could result from as few as two genetic events, if each one of these lead to the inactivation of one copy of a gene whose function would be to control cellular proliferation. The loss of both copies of the gene, which Knudsen called a 'tumor suppressor', would result in neoplasia. According to this hypothesis, the first inactivating genetic event, or 'hit', would occur in the germline in those patients with hereditary retinoblastoma, and thus be carried in all of the cells of this individual. The second hit would be to the somatic cells, in this instance the retinoblasts, resulting in the formation of a particular tumor. Patients with the sporadic type of retinoblastoma would also develop tumors after the two 'hits', but both of these events would be found only in the somatic tissues. The germ-line location of the first hit for patients with the herediatary type of retinoblastoma would account for the high degree of penetration of the disease, the more extensive (bilateral) involvement, and the increased frequency of secondary malignancies, as all of the cells in the body would be predisposed to involvement in tumorigenesis from a second somatic inactivating event. The initial evidence for Knudsen's hypothesis was cytogenetic: translocations and deletions of chromosome 13 were detected in the somatic cells (blood) of patients with hereditary retinoblastoma, and in the tumor cells from patients with both types of the disease [11]. Examination and comparison of several different cases resulted in the identification of a minimal affected region, located at 13q14. Subsequently, the hypothesis was verified by restriction fragment length polymorphism (RFLP) technology, allowing for tumor deletion mapping and genetic linkage studies [12]. The power of RFLP analysis lies in the ability of the investigator to identify whether both copies (alleles) of a particular DNA sequence are present. To perform these experiments, DNA is extracted from a tissue specimen and subject to digestion (cutting into smaller fragments) using bacterial enzymes called restriction endonucleases that cleave the strands only at sites of specific base-pair sequences. There are many stretches of the DNA of all individuals that contain naturally occuring variations in the base-pair sequence which are called polymorphisms [13]. These

105 variations are most often found in the non-translated part of the DNA (introns), and therefore carry no phenotypic consequence. If a polymorphism changes the sequence of a naturally occuring restriction endonuclease cleavage site, digestion of the DNA will result in a larger fragment, as a cut cannot be made. If a polymorphism results in the establishment of a new cleavage site, digestion of the DNA will result in an additional smaller fragment. The net effect of a polymorphism on one of the two alleles is to generate fragments of varying sizes when the DNA is digested with the appropriate enzyme, and probed using a short stretch of DNA complimentary to the site of the polymorphism. By radioactively labeling this DNA probe, a hybridization signal can be detected on x-ray film, and the presence of more than one signal (band) shows that both copies of the DNA sequence are contained in the tissue specimen. The presence of DNA fragments of a particular size after restriction enzyme digestion was also determined to be inherited in a mendelian fashion. Thus, the origin of an allele could be determined by studying DNA obtained from other family members. The first RFLP studies performed on retinoblastoma patients used a technique called Southern blotting [12]. DNA was extracted from blood (control) and tumor specimens, digested with restriction enzymes, and size-selected by electrophoresis on agarose gels. The DNA fragments were immobilized in the gel, transferred to a membrane (Southern blot), and probed using a radioactively labeled DNA marker of interest. An individual without a polymorphism showed only one band on the film, as the DNA fragments from the digestion of both strands were of the same size. This person would be called homozygous, and it would not be possible to detect whether only one or both copies of the DNA probe were present (non-informative). An individual with a polymorphism showed two bands, as the fragments from each alleles were of a different size. Thus, it was possible to detect whether or not both copies of the DNA probe were present in the tissue (informative). DNA from blood and tumor specimens from retinoblastoma patients was examined side by side using RFLP technology using probes derived from

chromosome 13. A DNA probe sequence that was absent in tumor tissue but present in control tissue from the same patient was evidence for loss of an allele. Using these probes, a common region of deletion at 13q14 was confirmed in these patients. Simultaneously, analysis was performed using DNA specimens from the blood of family members affected with the hereditary type of retinoblastoma, and these same probes were noted to be associated with affected individuals (genetic linkage). By these and other techniques, finer mapping and ultimately identification and cloning of the gene for retinoblastoma (Rb) was achieved [14]. The Rb gene, a DNA-binding protein, has been shown to have an important role in a number of cellular interactions [15]. Rb has also been shown to be deleted in a number of different tumor types, including osteogenic sarcoma [14] and small-cell lung carcinoma [16]. These and other studies have shown that a multiplicity of roles in many different tumor types is typical for a single tumor suppressor gene, most likely reflecting the natural function of the control of cellular proliferation. Since the cloning of the retinoblastoma gene, other tumor suppressor loci have also been identified. Experimental studies have resulted in the cloning of one of the genes for Wilm's tumor, a common pediatric renal malignancy (Wtl, located on chromosome 11) [17], several of the genes that function in the neoplastic cascade resulting in colorectal cancer, including the gene for familial adenomatous polyposis (FAR located on chromosome 5) [18], the genes for the neurofibromatoses type 1 (NF1, located on chromosome 17) [19] and type 2 (NF2, located on chromosome 22) [20], one of the genes for familial breast cancer (BRCA1, located on chromosome 17) [21], and at least one gene that appears to function as a multiple tumor suppressor for several different neoplasms, including melanoma and certain astrocytomas (MTS1, located on chromosome 9) [22]. To date, the most frequent genetic event in all types of human cancer involves mutation and/or deletion of a tumor suppressor gene: the ubiquitous p53, located on chromosome arm 17p [23].

106

The p53 gene This most common of tumor suppressors was initially detected as the 53 kD binding product of the large T antigen of the SV40 tumor virus. The p53 gene was originally classified as an oncogene based on the results of an NIH 3T3 cell transformation assay. Subsequently, it was determined that the p53 gene used for these studies had a mutation, and that the wild-type p53, rather than being oncogenic, was in fact growth suppressive [24]. The initial evidence showing that p53 had an important role in human cancer came from studies from the Vogelstein laboratory on the etiology of colorectal cancer [25]. These investigators determined by RFLP analysis that loss of chromosome arm 17p was a common event in colorectal tumorigensis. This loss included 17p13.1, the site of the p53 gene. Sequence data was obtained from several tumor specimens in which one copy of the p53 gene was determined to be absent, and revealed mutations in the remaining allele. Together, these alterations inactivated both copies of the gene, documenting its role as a tumor suppressor [25]. Subsequent experiments also revealed that replacement of wild-type p53 in colorectal cell lines carrying a mutated form of the gene resulted in the cessation of cellular growth. Since these observations were made, experiments by these and many other investigators have shown mutations of the p53 gene in all types of cancers, including those of the breast, lung, and brain [23, 27]. The acquisition of p53 mutations in astrocytomas has been documented as part of the transition from the more indolent type to the more aggressive type [28]. The p53 gene has also been shown to be involved in a multiplicity of cellular growth control functions, including progression through the cell cycle, and the repair of radiationinduced DNA damage [29].

Cytogenetic analysis of medulloblastoma The histologic similarity of medulloblastomas to other primitive small round blue cell tumors including retinoblastoma and small-cell lung carcinoma suggested that tumor suppressor genes might be in-

volved in their etiology. The initial clues to the location of such a gene or genes again turned out to be cytogenetic. Karyotype analysis of medulloblastoma specimens revealed isochromosome 17q [i(17q)] as the most common cytogenetic abnormality [3, 4, 30]. This chromosomal anomaly is the result of loss of the short (p) arm of the chromosome, and subsequent 'head to head' aposition of the long (q) arms. The formation of i(17q) is the most common chromosomal rearrangement of this type, and is frequently detected in leukemias, lymphomas, and solid tumors of the stomach and colon [30]. In medulloblastomas, i(17q) has been observed in approximately 2/3 of the karyotypes examined; in 1/2 of these cases it is the sole chromosomal anomaly [3, 4]. Other chromosomal losses that have been shown in medulloblastomas with a lesser frequency include those of chromosomes 1, 6, 16 and 22 [3, 4].

Molecular genetic analysis of medulloblastoma The frequent identification of i(17q) in tumor specimens prompted initial evaluation of chromosome arm 17p as a potential site for a tumor suppressor gene or genes important in medulloblastoma. RFLP studies from our laboratory as well as those from several other investigators subsequently revealed loss of DNA sequences derived from the short (p) arm of chromosome 17 in 35-45% of tumor specimens [5-8]. Although losses at other chromosomal loci were also detected in these specimens, they were far less frequently observed than those from 17p. The initial composite 17p deletion map suggested that the most common site of loss was centered at 17p13.1, the location of the p53 gene [6]. These results, as well as the growing number of p53 gene mutations detected in almost all types of human cancers [23], prompted an analysis of this gene in medulloblastoma. Although RFLP analysis will reveal deletion of a gene, the presence of mutations in the gene cannot be detected by this methadology. Thus, the search for p53 gene mutations in medulloblastoma specimens necessitated the use of other techniques. Although direct or indirect DNA sequencing techniques remain the standard for the identification of

107 gene mutations, they are often cumbersome, timeconsuming, and expensive, particularly if the mutations are infrequent. Thus, we chose to use a different technique for our initial screening of the p53 gene in medulloblastoma specimens: a combined polymerase chain reaction (PCR)-denaturing gradient gel (DGGE) technique. For this study, we employed DNA sequencing not only to verify mutations detected by this technique in our specimens, but to examine all of the DNA we initially screened to validate the PCR-DGGE technique for the detection of p53 gene mutations [31, 32]. The PCR-DGGE technique relies upon the ability of PCR to generate large numbers of faithfully replicated DNA sequences, and the ability of the DGGE technique to detect single base-pair changes in these DNA fragments. PCR is a process that utilizes a DNA polymerase that is temperature-sensitive, and thus can be alternately activated or inactivated by changing the temperature of the reaction. After determining the sequence of the DNA fragment to be amplified, primers are then produced that span the sequence, and the reaction performed in a DNA thermal cycler, resulting in up to a million-fold amplification of the identical DNA fragment. These fragments are then analyzed for mutations using DGGE, a technique that depends upon strand-separation (melting) of double-stranded DNA [32]. This process will occur at a specific temperature or concentration of alkaline denaturing agents that depends on the individual basepair sequence of the DNA strands. To perform DGGE, PCR-amplified DNA fragments are run out on a gel prepared with a gradient of alkaline denaturing agents. The DNA strands will separate at a particular point in the gel that is particular to their unique base-pair sequence, causing the DNA to become immobilized at that location, and appear as a single band characteristic for the fragment. As small a change as mutation of a single base-pair in these DNA fragments will result in a band at a different location in the gel than that of wild-type DNA. For our p53 gene analysis, we prepared primers for each part of the gene to be studied, as subdivision was necessary to perform PCR amplification with maximal efficiency [31]. We chose to evaluate exons 4-9 of the gene in our initial

studies because primers for this region were able to cover the sites of greater than 95 % of previously described p53 mutations, including several mutational 'hot spots' in human cancers. The results of our studies revealed that mutation of the p53 gene is an infrequent event in medulloblastoma: only 2 of 20 tumors showed such mutations [31]. The relative paucity of p53 mutations in medulloblastoma was also documented by several investigators using other types of mutational analysis [33-35], suggesting that the conserved region of the p53 gene was not the target of the 17p deletions seen in this tumor. Further experiments were then performed using newly identified 17p probes that were of particularly informative types: variable number of tandem repeat (VNTRs) markers and microsatellite or CArepeat markers. VNTR probes are based on polymorphisms generated by variation in the number of copies of a repetitive DNA sequence that is present on each of the chromosomal strands. VNTR sequences are frequently detected in the human genome, and the detection of their polymorphism by multiple restriction enzymes renders most individuals informative with many of these types of probes. Microsatellite or CA-repeat markers are based on stretches of repetitive highly CA-rich DNA sequences whose size variation results in a polymorphic allelic pattern. The ability to use PCR-based technology combined with the frequency of CA-repeats in the genome together generates useful information with these markers for most individuals. Recent studies with new markers of these types have shown that in medulloblastoma specimens there is preferential loss of DNA sequences at a site distal to the p53 gene on chromosome 17 [9, 35]. Loss of this region of 17p has also been detected in several other types of tumors including sporadic breast cancer [36]. Our current investigations suggest that the common deleted region is located at 17p13.3, bracketted distally by the 144-D6 probe, the most telomerically located 17p marker as yet identified, and proximally by the ABR gene (Fig. 1). The ABR gene was originally identified by hybridization of a human genomic library with BCR, a gene located on chromosome 22 at the common breakpoint site whose translocation results in the Philadelphia (Ph') chromosome, a common cytoge-

108 suppressor. We are currently testing the hypothesis that the gene is located within the A B R locus by mutational analysis using PCR-DGGE with primers that cover the A B R exons with the highest homology to active BCR sites (mBCR region). Other sites of DNA sequence loss in medulloblastoma that have been identified to date include those on chromosome arms 6q, 16q, and 22q [3, 4]. Molecular genetic analysis of larger numbers of tumor specimens will help to define the importance of these sites.

IDistal Chromosome 17p Loci : Medulloblastoma I

)

Telomere

1oeun:

co~osFr~

D17S34 D17S926 D17S695 D17S849 D17S643 ABR VNTR-B ABRVNTR-A D17S28 D17S5 D17S30 D17S66

13.2 13.1

Preserved

•

Lost

0

Noninformative

Clinical correlation Fig. 1. Composite deletion map of distal chromosome arm 17p for medulloblastoma tumor specimens.

Although a tumor suppressor gene for medulloblastoma has yet to be definitively identified and cloned, the frequency of 17p deletions in tumor specimens allows for clinical correlative analysis to be performed. The use of molecular genetic markers to guide the course of treatment and predict the outcome of intervention for cancers of many different types has been steadily increasing. For example, previous studies have shown that for neuroblastoma, another small round blue cell tumor of children, the copy number of the N-myc oncogene is a better predictor of treatment outcome than any of the clinical parameters evaluated to date [38]. We have performed a correlative analysis of our first 31 medulloblastoma patients treated at the University of California, San Francisco (UCSF) Medical Center, by dividing them into so-called

netic abnormality in chronic myelogenous leukemia [37]. The gene was named for its homology to BCR and its transcription in several types of tissues (Active BCR-Related). Our laboratory has previously determined that A B R is located 240-260 kb proximal to the 144-D6 probe, with a 5' to 3' transcriptional orientation [9]. Using probes for two previously identified VNTR regions in the gene, we have subsequently determined that in 3 medulloblastoma tumor specimens there is differential loss of A B R (Fig. 2). The preservation of the more distally located VNTR-B and loss of the proximally located VNTR-A suggest that a breakpoint exists within the A B R gene that may be located at or near the site of a novel tumor

ABR (toward cantromere)

f

Mbcr homologousregion axons

~r

pYNH37.3

VNTR-A m

(toward telomere)

1

2

3

4 5

•

•

•

mm

_Ij

m

m

m

mm

, ' , "

VNTR-B

p'144-D6

Fig. 2. Map of the ABR gene highlighting the region between the variable number of tandem repeat (VNTR) loci A and B. The arrow represents the location of the potential breakpoint identified in 3 of our tumor specimens.

109 Kaolan.Meler Curves for Medulloblasloma Patients: Survival tOO

P

150

i

,

I

~f~.

pO0, lt~k. (-, dete'~on$ {n=l @)

I,! I_ . . . . .

I

I

(~-~}

i

[

. . . . . . . . . . . J

I

10

I

I

I

t

20

30

40

EO

I

60

I

I

I

70

:BO

50

"1 I

II 1100

I 110

month~

Fig. 3. Survival curves obtained for 3i medulloblastoma patients from the UCSF series grouped by clinical ('good-risk' and 'poor-risk') and molecular genetic [with deletions (+) and without deletions (-)] data.

'good risk' and 'poor risk' groups based on conventional clinical staging [39]. Patients were considered as 'good-risk' if they met the following criteria: age at diagnosis greater than 3 years, total or near-total tumor resection, absence of metastatic deposits in the brain or spinal cord as determined by contrastenhanced magnetic resonance imaging (MRI), and cerebrospinal fluid (CSF) cytology that was negative for malignant cells. All of the patients who did not meet one or more of these criteria were placed in the 'poor-risk' category. 'Good-risk' patients received post-operative radiation, with a total dosage of 7200 rads to the posterior fossa delivered by a hyperfractionated technique, and 3000 rads to the craniospinal axis. Patients in the 'poor-risk' group below the age of 3 years at the time of diagnosis received a multi-drug chemotherapy regimen as the first line of post-operative treatment, followed by the previously outlined radiation protocol. Those 'poor-risk' patients above the age of 3 years at diagnosis received radiation as previously outlined as their first post-operative intervention, followed by a multi-drug chemotherapy regimen. The patients

in each group were then further subdivided into those with and without demonstrable 17p deletions in their tumor specimens. The results of these analyses were of great interest. Those patients in the 'poor-risk' group fared poorly whether or not there were detectable 17p deTable 1. Risk Group

Endpoint

P-value sided)

Good (N = 15) Good Poor (N = 16) Poor

Progression Survival Progression Survival

0.0727 0.0218 0.6541 0.1336

(two-

Statistical analysis of progression and survival data for 31 medulloblastoma patients from the UCSF series with and without deletion of chromosome arm 17p. A n exact logrank test was used to obtain these results. There is a statistical significance between the time to tumor progression and time to death for the 'goodrisk' patients with and without chromosome arm 17p deletions. While the 'poor-risk' patients did not show a similar statistically significant difference at the 0.1 level, the small sample size makes it difficult to definitively draw this conclusion.

110 Table 2.

Group

Median Timeto Progression

Median Survival

Good withoutdeletions Good withdeletions Poor withoutdeletions Poor with deletions

Not reached 18 months 52 months 12 months

Not reached 26 months 82 months 16 months

Kaplan-Meiermedianscalculatedfor progressionand survivalforeachpatientgroup.In our study,the 'good-risk'patientswithdeletions shared the same prognosisnegativeas the 'poor-risk'patients despite their differencein clinicalpresentation.

letions (Fig. 3). This result was anticipated, based on the assignment of patients to this category, and the inhomogeneity of the patient population, which included patients with no evidence of residual disease after surgery yet were under the age of 3 years at diagnosis, to those patients who showed wide-spread disseminated disease at the time of presentation. The results of the analysis for the 'good-risk' patients, however, were quite revealing-(Fig. 3). Those 'good-risk' patients without 17p deletions did well after treatment: all of the patients alive and free of tumor at the time the study was completed. Subsequently, one of the patients developed a recurrence located where the ports selected for radiation failed to overlap. In contrast, all but one of the 'good-risk' patients with 17p deletions recurred by the completion of the study, and all of these patients have since expired. These results were determined to be statistically significant (Tables 1 and 2). Thus, the results of this study showed that 'goodrisk' patients with 17p deletions had essentially the same prognosis negative as 'poor-risk' patients with or without such deletions. This result has potential clinical relevance, as the 'good-risk' patients received post-operative radiation alone, since their 17p deletions status was not available at the initiation of treatment. Those 'good-risk' patients with such deletions might have benefitted from the addition of the chemotherapy that was withheld owing to their favorable clinical status. Although the results of this study are striking, they will require replication on larger series of patients, as analyses of other medulloblastoma series have not yielded the identical results, perhaps owing to the smaller sample size of other populations studied. It is likely that

a national cooperative study such as that performed for children with neuroblastoma will be required to absolutely define the clinical implications of the molecular genetic data [40].

Acknowledgements The authors would like to thank the neurosurgeons from the University of Chicago (U of C) Medical Center and UCSF Medical Center for their assistance in obtaining tumor specimens, and the members of the Division of Neuro-Oncology, Department of Neurological Surgery, UCSF and the Department of Radiaton Oncology, UCSE for patient treatment data. We would also like to thank Ms. Kathleen Lamborn and Ms. Janet Wynne, UCSF Medical Center, for the statistical analysis of the clinical correlative data. Particular gratitude is also expressed to the research staff of the Brain Tumor Research Center, UCSF, and the Maggie McNamara/Barrett 'Bear' Krupa Memorial Laboratory, U of C, including Natalia Abrikosova, Margaret Burnett, Eugene Choi, Vivian Dai, Ruby Kalra, Ellen Mack, Andrew Metzger, Sharon Sih, Markus van Haken, Jennifer Willert, Shan-Mei Xu, and most especially, Eileen White and Laleh Daneshvar. Funding for these experiments was obtained from the American Cancer Society (PDT 429 and EDT 75341), the Brain Reserach Institute, U of C, the Cancer Research Coordinating Committee and School of Medicine, the University of California, the Maggie McNamara Memorial Fund and the Bear Necessities Pediatric Cancer Foundation.

iii References 1. Belza MG, Donaldson SS, Steinberg GK: Medulloblastoma: freedom from relapse longer than 8 years - a therapeutic cure? JNS 75: 575-582, 1991 2. Friend SH, Dryja TR Weinberg RA: Oncogenes and tumorsuppressing genes. NEJM 318: 618-622, 1988 3. Griffen CA, Hawkins AL, Packer RJ: Chromosome abnormalities in pediatric brain tumors. Cancer Res 48: 175-180, 1988 4. Bigner SH, Friedman HS, Biegel JA: Structural chromosomal abnormalities in human medulloblastoma. Cancer Genet Cytogenet 30: 91-101, 1988 5. Thomas GA, Raffel C: Loss of heterozygosity on 6q, 16q, and 17p in human central nervous system primitive neuroectodermal tumors. Cancer Res 51: 639-643, 1991 6. Cogen PH, Daneshvar L, Metzger AK: Deletion mapping of the medulloblastoma locus on chromosome 17p. Genomics 8: 279-285, 1990 7. Biegel JA, Burk CD, Barr FG: Evidence for a 17p tumor related locus distinct from p53 in pediatric primitive neuroectodermal tumors. Cancer Res 52: 3391-3395, 1992 8. James CD, He J, Carlbom E: Loss of genetic information in central nervous system tumors common to children and young adults. Genes, Chrom, Cancer 2:1990 9. McDonald J, Daneshvar L, Willert JR: Physical mapping of chromosome 17p13.3 in the region of a putative tumor suppressor gene important in medulloblastoma. Genomics 23: 229-232, 1994 10. Knudsen AJ Jr: Mutation and cancer: Statistical study of retinoblastoma. PNAS 68: 820-823, 1971 11. Sparkes RS, Sparkes MC, Wilson MG: Regional assignment of genes for esterase D and retinoblastoma to chromosome band 13q14. Sci 208: 1042-1044, 1980 12. Cavenee WK, Dryja TR Phillips NA: Expression of recessive alleles by chromosomal mechanisms in retinoblastoma. Nature 305: 779-784, 1983 13. Sambrook J, Fritsch EF, Maniatis T: Molecular cloning: A laboratory manual. Cold Spring Harbor Press, Cold Spring Harbor, NY, 1989 14. Friend SH, Bernards S, Rogelj S: A human DNA segment with properties of the gene that predisposes to retinoblastoma and osteosarcoma. Nature 313: 101-106, 1985 15. Lee W-H, Shew J-Y, Hong FD: Human retinoblastoma susceptibility gene: Cloning, identification, and sequence. Nature 329: 642-645, 1987 16. Harbour JW, Lai S-L, Whang-Peng J: Abnormalities in structure and expression of the human retinoblastoma gene in SCLC. Sci 241: 353-357, 1988 17. Call KM, Glaser T, Ito C: Isolation and characterization of a zinc finger polypeptide gene at the human chromosome 11 Wilms' tumor locus. Cell 60: 509-520, 1990 18. Kinzler KW, Nilbert MC, Vogelstein B: Identification of a gene located at chromosome 5q21 that is mutated in colorectal cancers. Sci 251: 1366-1371, 1991 19. Marchuk DA, Saulino AM, Tavvakol R: cDNA cloning of

20.

21.

22.

23.

23. 25.

26.

27.

28.

29. 30. 31.

32.

33. 34.

35. 36.

37.

38.

the type I neurofibromatosis gene: complete sequence of the NF gene product. Genomics 11: 931-940,1991 Rouleau GA, Merel R Lutchman M: Alteration in a new gene encoding a putative membrane-organizing protein causes neuro-fibromatosis type 2. Nature 363: 515-522,1993 Miki Y, Swensen J, Shattuck-Eidens D: A strong candicate for the breast and ovarian cancer susceptibility gene BRCA1. Sci 266: 66-72, 1994 Kamb A, Gruis NA, Weaver-Feldhaus J: A cell cycle regulator potentially involved in genesis of many tumor types. Sci 264: 436-440, 1994 Soussi T, de Fromentel CC, May P: Structural aspects of the p53 protein in relation to gene evolution. Oncogenes 5: 945952, 1990 Lamb R Crawford L: Characterization of the human p53 gene. Mol Cell Biol 6: 1379-1385, 1986 Baker SJ, Fearon ER, Nigro JM: Chromosome 17 deletions and p53 gene mutations in colorectal carcinomas. Sci 244: 217-221, 1989 Nigro JM, Baker SJ, Preisinger AC: Mutations in the p53 gene occur in diverse human tumour types. Nature 342: 705708, 1989 Metzger AK, Sheffield VC, Duyk G: Identification of a germ-line mutation in the p53 gene in a patient with an intracranial ependymoma. PNAS 88: 7825-7829, 1991 Sidransky D, Mikkelsen I", Schwechheimer K: Clonal expansion of p53 mutant cells is associated with brain tumor progression. Nature 355: 846-847, 1992 Hollstein M, Sidransky D, Vogelstein B: p53 mutations in human cancers. Sci 253: 49-54, 1991 Mertens F, Johansson B, Mitelman F: Isochromosomes in neoplasia. Genes, Chrom, Cancer 10: 221-230, 1994 Cogen PH, Daneshvar L, Metzger AK: Involvement of multiple chromosome 17p loci in medulloblastoma tumorigenesis. Am J Hum Genet 50: 584-589, 1992 Sheffield VC, Cox DR, Lerman LS: Attachment of a 40base-pair G + C-rich sequence (GC clamp) to genomic DNA fragments by the polymerase chain reaction results in improved detection of single base-pair changes. PNAS 86: 232-236, 1989 Saylors RL, Sidransky D, Friedman HS: Infrequent p53 mutations in medulloblastoma. Cancer Res 51: 4721-4723,1991 Adesina A, Nalbantoglu J, Cavenee WK: p53 mutation and MDM2 gene amplification are uncommon in medulloblastoma. Cancer Res 51: 5649-5651, 1994 Ohgaki H, Eibl RH, Wiestler D: p53 mutations in nonastrocytic human brain tumors. Cancer Res 51: 6202-6205, 1991 Coles C, Thompson AM, Eider PA: Evidence implicating at least two genes on chromosome 17p in breast carcinogenesis. Lancet 336: 761-763, 1990 DiSilvestre D, Morris C, Heisterkamp N: A second hypervariable RFLP within the ABR gene. Nucl Acids Res 18: 5925, 1990 Seeger RC: Association of multiple copies of the N-myc oncogene with rapid progression of neuroblastoma. NEJM 313: 1111-1116, 1985

112 39. Kopelson G, Linggold RM, Kleinman GM: Medulloblastoma: The identification of prognostic subgroups and implications of multimodality management. Cancer 51: 312-320, 1983 40. Cogen PH: Prognostic signifcance of molecular genetic

markers in childhood brain tumors. Ped Neurosurg 17: 245250, 1991-1992

Address for offprints: Philip H. Cogen, Department of Neurosurgery, Children's National Medical Center, III Michigan Avenue, NW Washington, DC 20010, USA