Molecular Cytogenetic Analysis in the Study of Brain Tumors

Jane Bayani, M.H.Sc.; Ajay Pandita, D.V.M., Ph.D.; Jeremy A. Squire, Ph.D.

Classic cytogenetics has evolved from black and white to technicolor images of chromosomes as a result of advances in fluorescence in situ hybridization (FISH) techniques, and is now called molecular cytogenetics. Improvements in the quality and diversity of probes suitable for FISH, coupled with advances in computerized image analysis, now permit the genome or tissue of interest to be analyzed in detail on a glass slide. It is evident that the growing list of options for cytogenetic analysis has improved the understanding of chromosomal changes in disease initiation, progression, and response to treatment. The contributions of classic and molecular cytogenetics to the study of brain tumors have provided scientists and clinicians alike with new avenues for investigation. In this review the authors summarize the contributions of molecular cytogenetics to the study of brain tumors, encompassing the findings of classic cytogenetics, interphase- and metaphase-based FISH studies, spectral karyotyping, and metaphase- and array-based comparative genomic hybridization. In addition, this review also details the role of molecular cytogenetic techniques in other aspects of understanding the pathogenesis of brain tumors, including xenograft, cancer stem cell, and telomere length studies.

A Brief History of Human Cytogenetics

The science of human cytogenetics (see review by Smeets255) is attributed to the Austrian cytologist Walther Flemming, who published the first illustration of the human chromosome in 1882. Six years later, in 1888, Waldeyer introduced the term “chromosome.” Sutton later combined the disciplines of cytology and genetics to coin the term cytogenetics: the study of chromosomes. The classic work of Theodor Boveri in the 1880s provided the foundation for understanding chromosomes as the units of inheritance, their involvement in embryonic development, and later, their role in disease. He postulated that chromosomal changes could lead to the development of cancer. In 1959, the first human karyotypes prepared from peripheral lymphocytes were visualized by Hungerford and col- leagues.[109] The ability to visualize numerical and structural chromosomal abnormalities helped reveal the genetics of Down syndrome (trisomy 21), Turner syndrome (45,X), and Klinefelter syndrome (47,XXY).[255]

Cancer cytogenetics took a major leap in the late 1960s with studies of hematological malignancies, which finally led to the discovery of the Philadelphia chromosome(Ph),[180] which was later found to be a consistent chromosomal change among chronic myelogenous leukemias.[230] These findings provided the impetus to identify consistent/ recurrent/nonrandom chromosomal changes in various disease conditions, yielding a plethora of simple and complex structural and numerical cytogenetic aberrations.

The development of reliable cloning strategies in the 1980s facilitated the genomic analysis and sequencing of specific DNA fragments. In addition, improvements in fluorescence microscopy permitted the visualization of these cloned DNA fragments to the chromosomal target. The emergence of FISH in the late 1980s and early 1990s paved the way for an effective and direct means of mapping specific DNA fragments to their chromosomal locations.[275] Besides being used as an important tool for gene mapping, FISH was also applied to ascertain the presence, absence, copy number, or location(s) of a particular chromosomal locus/gene in cancer cells. The FISH analysis could be applied not only to chromosomes (metaphase-based FISH), but to the interphase nuclei (interphase-based FISH) of cultured specimens, as well as to cells from tissues embedded in paraffin, touch preparations, or smears.

The complexities and heterogeneity of karyotypes in some cancers forced investigators to find means of determining the overall genomic changes in a given tissue. The difficulty in obtaining good cytogenetic preparations from the majority of solid tumors led to the development of the two types of CGH assays: metaphase- and (micro)arraybased CGH. Comparative genomic hybridization is a two-color FISH-based[125] or array-based[3] method used to identify the net gains and losses of genomic material in a given DNA sample. Equal amounts of tumor and normal DNA are differentially labeled, denatured, and hybridized to a normal metaphase spread or cloned DNA arrayed on glass slides. Any deviation in the ratio from 1 denotes gains or losses of those regions in the tumor DNA. This technique, which has enabled researchers to identify common regions of gain, loss, or high-level amplification without the need for actively dividing cells to provide metaphase spreads, is usefully applied to DNAs retrieved from archived material.

Although these methods proved to be useful in revealing patterns of genomic alterations among different tumors, the information regarding the way in which these genomic changes were exhibited (that is, simple deletions/balanced translocations compared with complex rearrangements/unbalanced translocations) in the karyotype was lost. The structural configurations in which amplifications, gains, and deletions were occurring could provide clues to the mechanisms influencing or causing these chromosomal alterations. In the past, the numerical and structural complexities of certain cancers made G-banding descriptions often incomplete and prone to errors. In the late 1990s SKY, a multicolor FISH assay, was developed;[247] this technique permitted the visualization of the entire genome in one experiment (for review see Bayani and Squire[12,14]). It was now possible to identify the chromosomes involved in complex structural aberrations and to reveal subtle chromosomal translocations that otherwise would have been missed or incorrectly annotated.

Due to the large body of literature and space constraints, the citation of all findings will not be possible, and we apologize in advance to those authors for the omission of their contributions. In recent years, a number of online resources have become useful tools for cataloging molecular cytogenetic findings. In this review we refer mostly to the data accumulated in the CGAP website, which is properly known as the Mitelman Database of Chromosome Aberrations in Cancer (2005). Mitelman F, Johansson B and Mertens F (eds.) (http://cgap.nci.nih.gov/Chromosomes/Mitelman).

We also refer to the NCI and NCBI’s SKY/M-FISH and CGH Database (2001) (http://www.ncbi.nlm.nih.gov/sky/skyweb.cgi); the Progenetix CGH online database (http://www.progenetix.net/); and PubMed/Medline (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi). Readers are encouraged to visit these websites regularly for updates.
Cytogenetic Findings in Brain Neoplasms
Neuroepithelial Tumors of the CNS: Glial Tumors

Astrocytic Tumors. Astrocytic tumors comprise the largest and most common group of brain tumors. The subcategories of astrocytic tumors included in this review are as follows: astrocytomas, anaplastic astrocytomas, GBM, pilocytic astrocytomas, subependymal giant cell astrocytomas, and pleomorphic xanthoastrocytomas.

Astrocytomas, Anaplastic Astrocytomas, and GBMs. This group of glial tumors illustrates the potential for low-grade astrocytomas to progress to a more malignant phenotype, and corresponds to the WHO grading of CNS tumors based on their histological features. Astrocytomas (WHO Grade II) are also known as low-grade diffuse astrocytomas and are characterized by slow growth and infiltration of neighboring brain structures. Anaplastic astrocytomas (WHO Grade III), also known as malignant and high-grade astrocytomas, may arise from a diffuse astrocytoma or de novo with no indication of a less malignant precursor. Glioblastomas or GBM (WHO Grade IV) may develop from a diffuse or an anaplastic astrocytoma (termed secondary GBM), but more commonly present de novo with no evidence of a less malignant precursor (termed primary GBM). The major genetic determinants that distinguish these two types of GBMs are EGFR amplification and TP53 mutation,[110] with the first being predominantly associated with the spontaneous variant, and the latter being primarily associated with GBMs arising from malignant progression. Because progression toward malignancy typically arises from a precursor lesion, low- and high-grade astrocytomas share similar changes. Classic cytogenetic analyses have revealed that karyotypes range from being karyotypically normal to grossly abnormal in structure and chromosome number. A general observation has been that the progression in malignancy is concomitant with an increase in complexity, both in structure and ploidy.[25]

A survey of the CGAP site (http://cgap.nci.nih.gov/Chromosomes/Mitelman) yields 102 low-grade astrocytomas[1,26,50,51,58,62,83,92,94,98,126,154,175,182,209,214,232,236,249,270,271,278,280,295,296,298] (astrocytoma not otherwise specified/astrocytoma I and II) that possess normal or abnormal karyotypes with near-diploid chromosomal counts (45–53), most commonly with the loss of one of the sex chromosomes. The most common whole chromosomal gains and losses are as follows: +7, -9, -10, +19, and -22. Tetraploid karyotypes represent approximately 25 to 30% of the karyotypes and recapitulate the findings in their diploid counterparts of whole chromosomal gains and losses. Structural chromosomal abnormalities present as partial deletions and translocations with the presence of structural changes including ring chromosomes. No specific, recurrent chromosomal translocation has been reported.

In 410 cases of astrocytomas Grades II and IV found at the CGAP site (large studies[1,22,26,27,29,30,50,62,117,141,149,154,159,175,201,207,211,249,270,295,298]), karyotypes similar to those among the low-grade astrocytomas persist, namely normal karyotypes or those only missing a sex chromosome, as well as karyotypes with copy number changes from chromosomes 7, 9, 10, 19, and 22, as described earlier. As with the low-grade tumors, these karyotypes typically possess gains and losses of the entire chromosome. Karyotypes with tetraploid and triploid chromosomal counts occur more frequently, suggesting increasing genomic instability and errors in the mitotic machinery.[176] Complex structural aberrations, including unbalanced translocations, insertions, the presence of double minute chromosomes, ring chromosomes, and unidentifiable “marker chromosomes” present more frequently, and contribute to the amplification of chromosomal regions that are believed to harbor oncogenes.

These gross findings from the conventional cytogenetic studies have been confirmed by CGH studies. In a survey of published results in 509 gliomas analyzed using CGH before 2001,[134] common changes between the three astrocytic tumor types include partial or whole gains of chromosome 7, loss of chromosome 10, loss of chromosome 22, and loss of 9p and 13q, confirming the cytogenetic findings from the previous two decades. The CGH assay also has been used to identify novel regions of change, including gains at 1p34-p36, 12p13, and 20q13. When present, amplifications were characteristically found at 1p36.2, 3q26.3q27, 7p12, 7q21-q31, 8q24.1, 12p13, 12q13-q15, 17q24, 19q13.2, and 20q13.1, and net genomic losses at 1p22, 4q33-q35, 6q16, 6q23-q27, 9p21, 10q25-q26, 13q21.1, and 22q13. Examination of the Progenetix database reveals CGH profiles from 78 astrocytomas not otherwise specified, 60 anaplastic astrocytomas, and 108 glioblastomas. A majority of the cases referenced at this site have been reviewed by Koschny, et al.,[134] and the reader may also refer to the website for a listing of these cases and their references.

The composite genomic profiles show that the 78 astrocytomas not otherwise specified have overall net losses at chromosomes 1p (16.7%), 2 (20%), 3p (3.9%), 4 (7.7%), 9p (7.7%), 10q (3%), 13q21 (6.4%), 18 (3.9%), 19q (16.7%), and X (10%). Gains are primarily identified on chromosomes 5 (7.7%), 7 (16.7%) or 7q (15%), 8q (5.1%), 9 (6%), 10p (7.7%), 12p (6.4%), and 19p (7.7%). Amplifications are restricted to the region spanning 8q21-8qter (2.6%). Among the 60 anaplastic astrocytomas, similar trends can be seen and the effects of changes in ploidy are apparent in the increase of whole chromosomal gains and losses across the genome. For anaplastic astrocytomas, losses were identified at 1p (20%), 3 (10%), 8p (8%), 9p (21.7%), 10 (26.7%), 12q21-qter (16.6%), 13q11-q32 (20%), 14 (13.3%), 17p (13.3%), 19q (22%), and 22 (23.3%). Gains were identified at 1q (15%), 2q (13.3%), 5 (5%) and 5q11-q23 (13.8%), 7 (35%), 8q (10%), 12q11-q21 (11.7%), 17q (8.3%), and 20 (11.7%). Amplifications are present at 1p31 and 1p32 (10% each), 7p11.2 (8.3%), 7q21 (1.7%), 7q22-q33 (3.3%), 8q13q23.3 (3.3%), 12q13-q21 (1.7%), 15q26 (1.7%), and 20p12 (1.7%). Among 108 glioblastomas, the affected chromosomes and overall frequency of gains and losses were found to be similar to those in the anaplastic group. In addition, the frequency of amplifications at the same chromosomal loci revealed some increases. Based on CGH studies, regions of amplification, gains, and loss have allowed identification of new candidate tumor suppressor genes and oncogenes as well as confirmation of the status of other genes previously identified using other molecular techniques. The roles of these and other putative tumor suppressor and oncogenes is reviewed in Ichimura, et al.[110]

Identification of the chromosomal changes associated with tumor progression was the subject of investigation in many early CGH studies. Weber, et al.,[291] identified alterations in primary astrocytomas (Grade II) to include losses on Xp and 5p, gains on 8q and 19p, and gain/amplification on 12p. Common progression-associated changes found in anaplastic astrocytoma (Grade III) or GBM included losses on 4q, 9p, 10q, 11p, and 13q, and gains on 1q, 6p, and 20q. The most frequent amplification site in all tumors was located on 12p13.[291]

In a similar study by Nishizaki and colleagues[179] low-grade astrocytomas were characterized by gains at 8q, 9q, 12q, 15q, and 20q; anaplastic astrocytomas were characterized by loss of 10q, 9p, and 13q, and gains of 1q, 7, 11q, and Xq; and GBMs were characterized primarily by losses of 9p, loss of all or part of chromosome 10, and loss of 13q, 22q, and Xq. More recently, Wiltshire, et al.,[297] examined 102 astrocytomas by using CGH to identify the genomic changes associated with each histological subtype and its clinical findings. Low-grade astrocytomas (Grade I) showed losses of chromosome 19p. In Grades II and III, losses of 9p and 10q with gains of 19p and 19q were identified. Grade IV tumors obtained in patients younger than 45 years of age showed changes including the loss of 9p and/or 9q, 10p and/or 10q, and chromosome 22, and gains of 7p and/or 7q and 19p. Tumors resected in patients older than 45 years of age had changes including the loss of 9p and 10p and/or 10q, and gains of 7p and/or 7q, 19p and/or 19q, and 20p and/or 20q. Cox proportional hazards statistical modeling showed that the presence of +7q and 10q CGH alterations significantly increased a patient’s risk of dying, independent of histological grade.

The information provided by these CGH studies has enabled investigators to validate these findings, both retrospectively and prospectively, by using conventional molecular assays; however, many have preferred to “FISH” the gene/chromosomal locus of interest directly to a cytogenetic specimen or tissue section.[150,248] The “FISH-ing” of specific probes directly to tissue or cytogenetic specimens has since revealed the heterogeneity of the tumor genome. The FISH examinations of EGFR amplification have revealed that cells within a given tumor specimen possess different levels of gene amplification,[185] information that is lost during the bulk DNA extraction of the specimen.

The technological improvements in microdissection methods in recent years have permitted investigators to select specific cells for extraction and study. This has helped researchers overcome the shortcomings of early CGH studies that were conducted using DNA contaminated with surrounding normal and abnormal cells. Proper and careful microdissection reduced or eliminated the “diluting” effects of contaminating cells with normal or questionable histological features. In a study by Hirose and associates,[105] microdissection was used to extract small regions of pure tumor from the paraffin-embedded sections of Grade II astrocytomas for CGH analysis. Thirty cases of Grade II astrocytoma were analyzed, and copy number changes were detected in 83% of cases. The most frequent aberrations were gains on 7q, 5p, 9, and 19p. Losses were detected at 19q, 1p, and Xp. As a result, two subgroups of Grade II astrocytomas were identified: those with a gain on 7q and those with losses on 1p/19q. Because only the microdissected cells were shown to be purely astrocytic, not oligodendritic, the authors of this study suggest that genetic differences exist within the grade, and may be influenced by the patient’s age and tumor location.

In addition to advancing researchers’ ability to select specific cells for analysis, the increased resolution of array-based CGH (see Albertson and Pinkel[3]) has resulted in refinements of genomic signatures to a 1-Mb level, a significant improvement over the 5- to 10-Mb resolution for metaphase-based CGH. Recently, Misra and coworkers[171] used array-based CGH to identify subgroups among 50 primary Grade IV astrocytomas (GBMs). A 2246 BAC array with a mean 1.5-Mb resolution was used. Thirty-three candidate sites for amplification and homozygous deletion were detected, and three major genetic subgroups within the GBM tumors were identified, including those with chromosome 7 gain and chromosome 10 loss; tumors with only chromosome 10 loss in the absence of chromosome 7 gain; and tumors without a copy number change in chromosomes 7 or 10. Correlation to clinical data suggested that there was no overall difference in survival between the groups; however, the group showing the loss of chromosome 10 and gain of chromosome 7 showed characteristics typical of GBM survivors, whereas the group without chromosome 10 loss or chromosome 7 gain showed characteristics of typical and long-term survivors.

The benefit of BAC array-based CGH is the ability to identify genes contained within the BAC contig, making validation possible by using FISH on cytogenetic or tissue specimens. Amplification of EGFR appeared to occur primarily in the group with chromosome 10 loss and chromosome 7 gain, and the authors postulated that this was associated with the primary form of GBM rather than the secondary form, which is associated with the group lacking chromosome 10 loss or chromosome 7 gain. The benefit from the increased resolution of array-based CGH is clear, and has led to much finer analysis beyond the level of individual chromosomes. De Stahl, et al.,[61] used a tiling-path array for chromosome 22 in a CGH experiment in 50 patients with GBM to identify germ-line and tumor-specific aberrations. Hemizygous deletions were detected in 28% of the tumors, with a predominant pattern of monosomy 22 in 20% of cases. The tiling nature of the array revealed the distribution of overlapping hemizygous deletions to delineate two putative tumor suppressor loci across 22q. Two distinct loci were affected by regional gains; both were of germ-line origin and were identified as TOP3B and TAFA, whose gene functions show promise for further investigation.

The advantage of CGH-based assays is the requirement of only small amounts of tumor DNA; thus, the need for mitotically active cells is not required, as it is in metaphase preparations of tumors. Nevertheless, invaluable information regarding the mechanism of genomic change/instability at the chromosomal level has been lost. The advent of whole-genome FISH assays provided the means for revealing the information classified as unidentifiable in classic cytogenetic studies (see Bayani and Squire[14]) and revealed the true complexity of the karyotypes. Furthermore, these studies provided clues to the nature of epigenetic changes. Several SKY studies revealed complex chromosomal changes.[31,55,135,139,140,257,307]

A study by our group[257] examined glial tumors derived from short-term cultures by conventional cytogenetics, CGH, and SKY (http://www.ncbi.nlm.nih.gov/sky/). The combination of different molecular cytogenetic techniques allowed us to identify the frequent involvement of chromosomes 1 and 10, which were affected by translocations, in addition to chromosomes 3, 5, 7, and 11. No specific recurrent chromosomal translocation was identified; however, the resulting breakpoint analysis and the identification of chromosomal origins in complex aberrations and “marker” chromosomes, together with net genomic changes, provided a more comprehensive cytogenetic description of the tumors. An example of the SKY analysis of the glioma cell line SF549 is shown in Fig. 1. Breakpoint analysis also provided a chromosomal basis for disruption of gene function and expression, because breakpoints were found to occur near regions of gains/amplification and deletion detected using CGH analysis.

A larger study by Krupp and colleagues[139] investigated 23 diffuse astrocytomas by combinations of either SKY, metaphase-based CGH, or metaphase-based FISH. According to their findings, most of the identified structural rearrangements were localized on chromosome arms 2p and 7q, with numerical changes most frequently involving chromosomes 7, Y, X, 10, and 17. A review of interphase-based FISH data indicated that cells with polysomy 7 were found in 75% of Grade II astrocytomas as well as in 100% of Grade III astrocytomas and GBM cases. Monosomy 10 was found in 75% of Grades II and III astrocytomas as well as in 100% of GBM cases. More recently, Cowell, et al.,[55] examined four GBM cell lines derived from primary tumors by SKY and array-based CGH, with FISH and PCR validation experiments. Their findings confirmed previous molecular cytogenetic observations of GBMs: karyotypes with chromosomal counts near normal or in the triploid, tetraploid, or hexaploid range as well as complex structural changes. Using a 6000 BAC array, CGH enabled the identification of deletions at the 9p13~p21 region harboring the CDKN2A gene (seen in all four tumors), which were confirmed with FISH assays. Amplifications of EGFR (7p12.3) were also identified and confirmed using PCR analysis.

Pilocytic Astrocytomas. Pilocytic astrocytomas, tumors classified as Grade I by WHO, typically occur in children and have a relatively good prognosis.[110] These tumors can maintain their Grade I status over a long period of time and rarely become more malignant in phenotype. Results of classic cytogenetic analysis of pilocytic astrocytomas are characterized by normal karyotypes or abnormal karyotypes in the near-diploid range. As in the Grade II astrocytomas, whole chromosomal gains and losses characterized the tumors (particularly the loss of sex chromosomes) as well as the gains of chromosome 7, loss of 10, and loss of 22. The CGAP database describes 31 cases of pilocytic/juvenile astrocytomas.[22,30,62,117,207,225,295,306]

The Progenetix site currently reports nine cases[250] of pilocytic (juvenile) astrocytomas, showing predominately losses for all of 1p (11%) or at 1p31-p36 (22%), 3 (11%), 14q11q24 (11%), 15 (11%), and 19, 20, 21, and 22 (each at 11%). Gains occur predominately for chromosomes 4 (22%) or 4q21-q32 (33%), 5q14-q31 (11%), 6q14-q23 (33%), 7 (11%) or 7q31(33%), 10 (10%), 11 (10%) or 11q14-qter (22%), and 13q21-q31 (22%). Other CGH studies[233,297] confirm the low frequency of net genomic changes stemming from either the normal or near-diploid karyotypic changes. Wiltshire and coworkers[297] however, identified the loss of 19p in a subset of pilocytic astrocytomas as the only significant numerical alteration. To date, no SKY or multicolor FISH analysis of pilocytic astrocytomas has been reported in the literature.

Subependymal Giant Cell Astrocytoma. Subependymal giant cell astrocytoma (WHO Grade I) is a benign, slow-growing tumor. This lesion occurs almost exclusively in patients with tuberous sclerosis complex.[251] Genetic linkage studies have tied tuberous sclerosis complex to two different loci, one on chromosome 9q34 (TSC1) and another on chromosome 16q13.3 (TSC2).[198] Only one study of the cytogenetic basis for subependymal giant cell astrocytoma has been identified. Debiec-Rychter, et al.,[65] published two cases of this tumor that were identified using conventional cytogenetics. One of the tumors had a complex, near-diploid karyotype with a translocation involving chromosome 22 at band q12, whereas the second one showed chromosome 1 loss and chromosome 22 deletion at band q12. Both lesions also had normal karyotypes.

Pleomorphic Xanthoastrocytoma. Pleomorphic xanthoastrocytoma is a low-grade glioma corresponding to WHO Grade II. It is uncommon and accounts for less than 1% of all astrocytic tumors. Most patients have a relatively favorable prognosis; however, tumors that have undergone progressive anaplastic transformation to high-grade gliomas or GBM have also been reported.[110,299] Few cytogenetic studies have been conducted on pleomorphic xanthoastrocytoma. In a case report by Sawyer and colleagues[237] the hyperdiploid karyotype showed a gain of chromosomes 3 and 5 and the loss of chromosomes 20 and 22 as well as the addition of two unbalanced translocations; one involving chromosome 7 and an unknown chromosomal partner and another involving telomeric fusions of chromosomes 15 and 20. Later in 1992, this same group reported the recurrent tumor.[239] The apparent telomeric fusion between chromosomes 15pter and 20qter, and between an extra copy of the long arm of chromosome 1 and chromosome 22qter, evolved in a stepwise fashion to ring chromosomes 20 and 22. A report by Li and associates[148] described the karyotype from a recurrent pleomorphic xanthoastrocytoma following treatment, showing a near-diploid chromosome count with complex structural abnormalities. Lai, et al.,[141] reported two pleomorphic xanthoastrocytoma karyotypes, of which only one showed an abnormal diploid karyotype with translocation involving chromosomes 1, 12, 16, and 19. More recently, Yin and colleagues[299] conducted CGH experiments on three pleomorphic xanthoastrocytoma tumors. These findings revealed gains on 2p (one of three), 4pter (one of three), 7 (two of three), 11qter, 12, 15q, and 19 (each locus with one of three); and losses on 8p (two of three), 9p, 10p, and 13 (one of three in each of these loci).

Oligodendrogliomas, Anaplastic Oligodendrogliomas, and Mixed Oligoastrocytoma. According to their histopathological appearances, diffusely infiltrative gliomas can be divided into astrocytic, pure oligodendroglial, and mixed oligoastrocytic tumors. Based on similarities in clinical features and genetic aberrations, oligodendrogliomas and oligoastrocytic tumors are often grouped together as oligodendroglial tumors. The WHO classifications identify oligodendrogliomas as Grade II, and these lesions behave much like diffuse astrocytomas, whereas anaplastic oligodendrogliomas are characterized as Grade III. An accurate distinction between the two is important, however, because it has prognostic and therapeutic implications (see Jeuken, et al.,[121]). The CGAP website references 45 cases of oligodendrogliomas.[94,117,154,201,207,209,270,298] Among these cases, approximately 25% involve normal karyotypes with only the loss of a sex chromosome. The gain of chromosome 7 and the loss of chromosomes 21 or 22 appear as the sole change in another 25% of karyotypes. The remaining karyotypes show combinations of these changes as well as the addition of translocations and unidentified marker chromosomes as different clones within the same tumor, indicating heterogeneity in the cell population.

The CGAP website also references one case of an oligoastrocytoma,[53] showing heterogeneity in the karyotypes present in the tumor. In each of the 14 karyotypes identified for this tumor, the gain of chromosome 7 occurred in all clones, and the authors suggest that the largest and most widely distributed clonal population (47,XY,+7) underwent further evolution to give rise to seven additional sidelines. Two karyotypes displayed a tetraploid content, with two more showing 50 chromosomes, and the remaining karyotypes were described as diploid. No apparent translocations or structural changes were identified on G-banding analysis, with all aberrations existing as whole chromosomal gains or losses.

Analysis of the CGH studies reveals more wide-ranging genomic changes. A current survey of the Progenetix web-site yields 40 cases of oligodendrogliomas (not otherwise specified),[119,127,137,190,191,250] with the primary changes including the loss of all or part of 1p (42.5% of cases), gain or amplification of 7p (15%), gain of part or all of 7q (27.5%), gain or amplification of 8q (~ 12%), loss of part or all of 9p (20%), loss of chromosome 10 (10%), gain of chromosome 11 (~ 12%), loss of chromosome 13 (15%), gain of 17q (20%), the loss of all of chromosome 19 (10%) or the loss of 19q (27%), and the loss of chromosome 22 (~ 17%). Twelve cases of anaplastic oligodendrogliomas[127,137,190] were associated with more striking genomic changes, including the loss of all or part of 1p (41–50%), loss of part or all of chromosome 2 (16–25%), loss of part or all of chromosome 4 (~ 40%), loss of all or part of chromosome 6 (~ 16%), loss of all of chromosome 9 (16%) or the loss of 9p (9%), loss of all of chromosome 10 (25%), loss of chromosome 13 (16%), loss of chromosomes 14, 15, and 16 (25% each), loss of 19q (25%), and loss of 21 (16%).

Also summarized on the Progenetix site are the findings in 16 anaplastic oligoastrocytomas[127,137,160] showing similar oligodendroglioma-like changes as well as others, including the predominant loss of 1p (43%), 2q (12.5%), 4q (12.5%), 9p (18.8%), 11p (18%), 12q (6.3%), 13 (43.8%), 14q (18%), 18q (6.3%), and 19q (18.8%), and gains of 7/7q (25%), 8q (12%), and 10p (6.3%). These findings are consistent with many LOH studies (see Jeuken, et al.,[121] for review) in which losses of 1p and 19q are identified as the hallmark changes characteristic of oligodendrogliomas. Array-based CGH has been used to refine the deletion of 1p and 19q in both tumors and cell lines. Law and colleagues[145] used homozygosity mapping, FISH, and CGH to arrayed BACs to screen 17 glioma cell lines for chromosome 1 and 19 deletions. Array-based CGH and homozygosity mapping of these cell lines defined a 700-kb common deletion region encompassed by a larger deletion region previously known in sporadic gliomas. The common deletion region was localized to 1p36.31 and included CHD5, a putative tumor suppressor gene.

Other novel changes have been refined and identified with the increased resolution of array-based CGH, including findings by Rossi and associates[229] showing an approximately 550-kb region in 11q13 and an approximately 300kb region in 13q12 displaying hemizygous deletion invirtually all the tumors analyzed regardless of their 1p/19q status. These findings were confirmed by interphase-based FISH analyses of nuclei from the same tumors used for array-based CGH, making this specific change a diagnostic marker for this subgroup of low-grade tumors.

In another array-based CGH study, Kitange, et al.,[127] examined 31 oligodendrogliomas of different grades and histological features and identified the most frequent aberrations, including the loss of 1p (49%) and 19q (43%), and the combined loss of 1p/19q (37%) as well as the deletions of 4q, 5p, 9p, 10q, 11p, and 13q, and gains of 7p, 8q, 10p, and 11q. Whole-chromosome losses of 4, 9, and 13 were also detected, with whole-chromosome gains of 7 and 11. The minimally altered regions were identified at chromosomal bands 1p36.32, 4q33, 5p15, 8q24, 11p15, and 19q13.3. A subsequent univariate analysis of these cases suggested that combined deletion of 1p and 19q was associated with better survival (p = 0.03), whereas an 8q gain in oligodendrogliomas was strongly associated with poor outcome (p = 0.002). Also associated with poor disease outcome were alterations that had a low prevalence in the pure oligodendrogliomas, including loss of 3q, 9q, and 12q and gain of 1p, 8p, and 10q. The common changes shared by low-grade oligodendroglioma and its high-grade counterparts suggest that the loss of 1p and 19q are early events in oncogenesis, with varying reports on whether the initial change occurs on 1p or 19q.[121] Furthermore, the presence of astrocytic components in the mixed subtypes suggested similar clonal origins, with the tumor microenvironment imposing differentiating influences.

The search for tumor suppressor genes on both 1p and 19q has led to the identification of minimal regions of interest by an assortment of molecular analyses, including array-based CGH and FISH methods, and has identified candidate genes, including TP73 (1p36.32), CDKN2C and RAD54 (both 1p32), GLTSCR1 (19q13.3), EDH2 (19q13.3), and GLTSCR2 (19q13.3).

Ependymal Tumors. Ependymomas are well-delineated, moderately cellular gliomas and are the third most common brain tumors in children.[267] The WHO classification differentiates four major types: ependymoma (WHO Grade II), anaplastic ependymoma (Grade III), myxopapillary ependymoma (Grade I), and subependymoma (Grade I) as well as the ependymal variants (cellular, papillary, epithelial, clear cell, and mixed). Whereas ependymomas occur in both children and adults, subependymomas and myxopapillary ependymomas are more common in adults.[267]

The CGAP website identifies 106 ependymoma karyotypes[1,22,50,62,64,83,92,117,161,175,207,209,225,227,257,270,278,281,293,294] (also reviewed by Mazewski, et al.[161]). The karyotypes described are predominantly normal, and when abnormal are near-diploid, and are characterized by gains and losses of entire chromosomes. Normal karyotypes have been estimated to occur in approximately 34% of published cases.[161] The most commonly gained chromosomes include 4, 5, 7, 8, and 9, either as the sole change or in combination. Loss of chromosome 10, 17, and 22 is also a frequent occurrence. Structural chromosomal aberrations often involve chromosomes 2, 6, 7, 12, 13, 16, 17, and 22 and are frequently simple in nature. No specific translocation has been identified and the only significant breakpoint appears at 22q11-13.

The LOH studies have identified LOH of 22q as the most frequent change in approximately 30% of ependymomas,[267] contributed partially by the loss or structural abnormalities of chromosome 22 observed in the karyotypes. Clinically, adult ependymomas and the myxopapillary subtype are most likely to have chromosome 22 changes. The 22q region contains the NF-2 tumor suppressor gene, making this a candidate gene for ependymomas (see discussion in a later section). A number of other tumors seen in NF2, including vestibular schwannomas and meningiomas, have also shown chromosomal aberrations involving chromosome 22q.[267] A SKY analysis has been performed in one case of ependymoma reported by our group,[257] and showed no additional structural aberrations from the original G-banded karyotype (http://www.ncbi.nlm.nih.gov/sky/) (Fig. 2). Cytogenetic analysis of subependymomas[57,260] revealed normal karyotypes and nonclonal changes involving chromosome 17. When multicolor FISH analysis was conducted on an intracranial ependymoma,[91] the tumor was found to possess chromosomal aberrations including i(1q) as well as aberrations involving chromosomes 6p and 17p.

The Progenetix Database summarizes the CGH findings of 165 ependymomas (not otherwise specified) and 29 cases of anaplastic ependymomas (see for references http://www.progenetix.net/). The most common genomic changes among the 165 ependymomas summarized included the gain of 1q (17%), 4 (14%), 5 (15%), 7 (17%), 9 (16%), and 12 (10%), and the loss of chromosomes 3 (11.5%), 6p (14.6%), 6q (20%), 10 (14%), 13 (12%), 16 (18.2%), 17 (12%), 19 (10.9%), 20q (12.7%), and 22 (29%). In the 29 anaplastic cases,[118,241,288] gains of 1q (17.2%), 7 (10.3%), 9q (13.8%), and 15 (10%) were detected as well as amplification at 2p24.[241] The amplification at 2p24 was confirmed by FISH to be amplification of MYCN in a spinal ependymoma. Losses were detected on chromosomes 3 (6%), 9p (6.9%), 10p (10.3%), 10q (17.2%), 13q21 (10.3%), and 22 (10%). The LOH studies have confirmed the presence of deletions in these regions,[106] specifically at chromosomes 6 and 9 as well as at loci 3p14, 10q23, and 11q. Moreover, investigation of chromosome 22 using tiling-path arrays by Ammerlaan, et al.,[6] revealed the presence of overlapping interstitial deletions of 2.2 Mb and approximately 510 kb in two patients. The deletions were also found to be present in the constitutional DNA of these two patients and in some of their unaffected relatives. Microsatellite analysis of these families further refined the commonly deleted segment to a region of 320 kb between markers RH13801 and D22S419, suggesting the presence of a low-penetrance ependymoma susceptibility locus at 22q11.

The CGH assay has also been used to correlate clinical parameters and outcome. Analysis of 42 primary and 11 recurrent pediatric ependymomas by metaphase-based CGH was correlated to clinical outcome in a study by Dyer, et al.[72] Hierarchical clustering of the findings identified three distinct genetic patterns. The first group showed few and mainly partial imbalances, which the authors suggested were a result of structural changes. The second “numerical group” showed 13 or more chromosome imbalances with a nonrandom pattern of gains and losses of entire chromosomes. The remaining tumors showed a balanced genetic profile that was significantly associated with a younger age at diagnosis (p < 0.0001), suggesting that ependymomas arising in infants were biologically distinct from those occurring in older children. Multivariate analysis showed that the structural group had a significantly worse outcome compared with tumors in which a numerical (p = 0.05) or balanced profile (p = 0.02) was found.

For the myxopapillary ependymomas, a molecular cytogenetic study by Mahler-Araujo and colleagues[156] attempted to identify common aberrations within this group of tumors. Seventeen myxopapillary ependymomas were studied by combinations of CGH, microsatellite analysis, and interphase FISH. Of seven tumors analyzed using CGH, a concurrent gain on chromosomes 9 and 18 was the most common finding. Microsatellite and interphase-based FISH analysis revealed results consistent with CGH findings; these results included the gains of both chromosomes 9 and 18 in 11 of 17 cases, the gain of either chromosomes 9 or 18 and imbalance of the other chromosome in three of 17 tumors, and allelic imbalances of chromosomes 9 or 18 in three and one of 17 tumors, respectively.

The FISH assay has been performed using 1p/1q, 19p/ 19q, centromere 18/DAL1, and bcr/NF2 probe pairs in the analysis of 10 clear cell ependymomas.[81] No deletions involving 1p, 19q, or NF2 were detected. Furthermore, the tumors in five of seven patients, all showing anaplasia, had losses of both centromere 18 and DAL-1.
Neuroepithelial Tumors of Uncertain Origin
Spongioblastoma

Spongioblastomas are classified by WHO as Grade IV, and are tumors containing spongioblast cells. The CGAP site reports one case of spongioblastoma that had a diploid count with a stemline showing the gain of chromosome 2 and the loss of both chromosomes 7 and 9. Double minute chromosomes, which are indicative of gene amplification, were detected in a cell, along with the presence of marker chromosomes in a hypotetraploid cell.[83] Gliomatosis

Gliomatosis cerebri (WHO Grade III/IV) is a rare, diffuse glial tumor with extensive brain infiltration that involves more than two lobes, frequently occurs bilaterally, and often extends to the infratentorial structures and spinal cord. The peak incidence appears to occur in patients between 40 and 50 years of age. Unfortunately, the prognosis is typically poor.

The CGAP website reports two cases of gliomatosis,[30,99] in which the karyotypes are described as near-diploid. Bigner and associates[30] reported the presence of unidentifiable marker chromosomes and double minutes against an otherwise normal karyotype. The karyotype reported by Hecht, et al.,[99] revealed structural aberrations involving chromosomes 6q, 14q, 15q, 18q, 19p, 20p, and 21q. A tetraploid version of the diploid karyotype was also reported, indicating failure of the cell to undergo cytokinesis.

In a recent CGH analysis, Kros and colleagues[138] examined the idea of “field cancerization.” Because gliomatosis cerebri is a rare condition in which the brain is infiltrated by a diffusely growing glial cell population involving at least two lobes, and sometimes even affecting infratentorial regions, the neoplastic proliferation may have a monoclonal origin, or alternatively, it may reflect progressive neoplastic change of an entire tissue field, which is known as field cancerization. Thus, the presence of an identical set of genetic aberrations throughout the lesion would point to monoclonality of proliferation, whereas the presence of nonidentical genetic changes in widely separated regions within the neoplasm would support the concept of field cancerization. The CGH analysis revealed losses on 2q11-q31 in 13 of 24 samples and losses on 19q13-qter in 10 of 24 samples from both left and right hemispheres. Other widespread chromosomal aberrations included losses on 3q13qter and 16q22-qter and gains on 7q22-qter, supporting the concept of monoclonal tumor proliferation.
Astroblastoma

Astroblastomas are a rare glial tumor occurring preferentially in young adults. Lesions are characterized by a perivascular pattern of glial fibrillary acidic protein–positive astrocytic cells with broad, nontapering processes radiating toward a central blood vessel. Low-grade astroblastomas appear to have a better prognosis than those with high-grade histological features.

Four cases of astroblastoma that have been reported at the CGAP site[141,232,257,269] show karyotypes in the diploid range. Structural aberrations such as translocations are frequent as well as whole chromosomal gains and losses. Chromosomes 7, 10, 12, 21, and 22 are frequently involved in both numerical and structural changes. The SKY analysis of an astroblastoma case,[257] previously reported by Jay, et al.,[113] refined the original karyotype, describing the loss of chromosomes 10, 21, and 22 and the presence of two marker chromosomes; the identification of the markers as an unbalanced translocation between chromosomes 10 and 21; and the presence of a chromosome classified as some duplication of chromosome 22 (http://www.ncbi.nlm.nih.gov/sky/). Cytogenetic analysis of a high-grade astroblastoma[169] revealed a hypodiploid clone showing deletion of 1p36 and 11p13 and an unbalanced translocation between chromosomes 14 and 15. Chromosome 22 was found to be rearranged, which was confirmed on FISH analysis, whereby it was determined that a rearrangement of 22q resulted in a complex translocation with chromosome 11. Results of the FISH assay also confirmed the loss of distal 1p.

Brat and coworkers[37] conducted CGH studies on seven astroblastomas, identifying genomic changes including gains of chromosome 20q (four of seven) and 19 (three of seven). The combination of these gains occurred in three tumors, including two well-differentiated and one malignant astroblastoma. Other alterations, noted in two tumors each, were losses on 9q, 10, and X.
Ganglioglioma

Gangliogliomas are rare tumors of the CNS that account for approximately 1% of all brain tumors and are classified as WHO Grade I. Histologically, gangliogliomas are composed of intimately admixed glial and neuronal components, with pathological origins have are not yet been identified.

The CGAP site lists 10 ganglioglioma cases.[22,62,175,257] Of the 10, seven were described as diploid, two were tetraploid, and the remaining case was hyperdiploid. Whole chromosomal gains and losses characterized the cytogenetic descriptions; however, these also included structural aberrations including deletions, additions, and translocations. Gains and structural aberrations involving chromosome 7 occurred frequently. There were also instances of copy number and structural changes involving chromosome 13. Other case reports,[116,133,286] have described cytogenetic findings consistent with those presented at the CGAP site. A SKY analysis of an adult ganglioglioma by our group[257] showed no additional chromosomal alterations from those detected on conventional G-banding studies (http://www.ncbi.nlm.nih.gov/sky/). One anaplastic ganglioglioma was reported by Jay, et al.,[115] who described the malignant transformation of a ganglioglioma that showed a complex abnormal karyotype with three sublines containing several structural chromosomal abnormalities.

There are few CGH studies of gangliogliomas. The Progenetix website displays two cases,[218] one normal and the other whose sole abnormality was a loss of the region of 4q13-q31. Squire, et al.,[257] used CGH to analyze a pediatric ganglioglioma, and the lesion was shown to have no net changes, despite the identification of structural changes involving chromosomes 1, 2, 3, 13, 17, and 22 as well as the detection of some cells with a tetraploid count. The disparity between the CGH results and cytogenetic findings lies in the heterogeneity of abnormal cells, the presence of normal karyotypes, and the possibility of contaminating normal tissue. Two publications by Yin and colleagues[300,301] described the CGH findings in five gangliogliomas, including the loss of material on 9p in three of five cases. Gains of parts or all of chromosome 7 were also detected and confirmed on FISH analysis. Genomic losses were detected at 2q33-q34, 8q12-q22, 14q21-qter, and 15q26-qter.
Desmoplastic Infantile Ganglioglioma and Astrocytoma

Desmoplastic infantile gangliogliomas and astrocytomas are classified as WHO Grade I, and are rare tumors. Desmoplastic infantile gangliogliomas are described as having both neuronal and astroglial elements, whereas desmoplastic infantile astrocytomas have primarily astrocytic elements. Although both may have aggressive cellular features, these tumors are often benign and the patient’s prognosis is good when the lesion is completely resected.

Cytogenetic analysis by Park and associates[189] described a case of desmoplastic infantile ganglioglioma with no consistent clonal abnormalities. The majority of cells, however (25 of 40), showed structural rearrangements, specifically telomere associations, resulting in dicentric and other derivative chromosomes. The breakpoints most often observed included 17q25, 19p13.3, 17p13, 14q32, 11q25, 9p24, 5q35, and 22q13. Bhattacharjee, et al.,[22] reported a case in which the lesion showed a hypotetraploid karyotype with both structural and numerical changes. As in other glial tumors, changes involving chromosomes 1, 7, 9, and 10 were identified. Kros, et al.,[136] conducted molecular analysis including CGH in three typical cases of desmoplastic infantile astrocytoma and ganglioglioma, and revealed loss of 8p22-pter in one case, whereas in another a gain of 13q21 was detected. Their findings led them to suggest that the genetic aberrations found in desmoplastic infantile ganglioglioma differ from those encountered in common astrocytomas.
Central Neurocytoma

Central neurocytomas are rare, benign, slow-growing neoplasms that have a favorable prognosis. They compromise 0.25 to 0.5% of brain tumors.[246] The mixed histological features of these tumors has prompted studies to determine their cellular origins.

The CGAP website reports one central neurocytoma, showing the sole abnormality as a loss of chromosome 17.[49] Jay and colleagues[112] reported a diploid case in which three copies of 1q were involved with rearrangements of chromosomes 4 and 7. A FISH analysis investigating the copy number status of chromosome 7 has been conducted by Taruscio, et al.,[264] and the chromosome was found to be gained in 33% of the tumors studied (nine), and in one case it was the sole abnormality. The CGH studies performed in 10 central neurocytomas by Yin, et al.,[302] are summarized at the Progenetix website and show identified gains at 2p (40%), 10q (40%), and 18q (30%).

The mixed cellular features of these tumors has led to several studies to identify their similarity with other more defined tumors, including oligodendrogliomas[84] and neuroblastoma.[273] Using FISH analysis for markers on 1p and 19q, which are characteristic of oligodendrogliomas; and probes for 1p26 and MYCN (2p24), which are characteristic of neuroblastoma, these studies have shown that central neurocytomas do not share the characteristic changes associated with oligodendrogliomas or neuroblastomas.
Dysembryoplastic Neuroepithelial Tumor

The DNETs (WHO Grade I) are a benign, usually supratentorial, neuronal–glial neoplasm occurring primarily in children and young adults with a long-standing history of partial seizures. These tumors may occasionally occur in patients with NF1, and they carry a good prognosis. Like central neurocytomas, the mixed cellular features of DNETs have prompted investigators to determine whether these lesions possess the changes known to appear in other tumors.[84,192,199] These FISH investigations have also shown that DNETs do not have loss of 1p or 19q, as is the case in oligodendrogliomas; nor do they show amplification of MYCN or EGFR, which is typical of neuroblastomas and astrocytomas, respectively.
Olfactory Neuroblastoma (Esthesioneuroblastoma)

There are few cytogenetic studies of esthesioneuroblastomas; however, these lesions have been shown to range from diploid to polyploidy, with relatively simple to complex changes.[88,123] The CGAP website reports a karyotype[123] showing a diploid tumor with numerous structural aberrations involving chromosomes 1, 3, 7, 8, 10, and 13. The Progenetix site reports three cases analyzed using CGH,[215] showing gains of 1p32pter (two of three), 8q23-qter (all three), 9q31-qter (two of three), and 15q25-qter, 19, and 22q (all three in each instance). Losses were detected for 4q and 13q in all cases.

More recently, Bockmuhl, et al.,[32] examined 22 esthesioneuroblastomas. They found deletions on chromosomes 3p and overrepresentations on 17q in up to 100% of cases. In more than 80% of cases, deletions were detected on 1p, 3p/q, 9p, and 10p/q, along with gains on 17p13, 20p, and 22q. The most consistent finding was a pattern for involvement of chromosomes 3, 10, 17q, and 20 occurring almost exclusively by deletions or overrepresentations, respectively. High copy gains/amplifications were seen on 1p34, 1q23-q31, 7p21, 7q31, 9p23-p24, 17q11-q22, 17q24-q25, 19, 20p, 20q13, and 22q13. The analysis of metastatic/recurrent lesions indicated a higher percentage of pronounced alterations, such as the high-copy DNA gains at 1q34-qter, 7q11, 9p23-p24, 9q34, 13q33-q34, 16p13.3, 16p11, 16q23q24, and 17p13. The authors suggested that deletions of chromosome 11 and gains of 1p may be associated with metastasis formation and/or worse prognosis. These recent findings add to the ongoing debate whether these tumors are similar to the primitive peripheral neuroectodermal tumor (Ewing group)[256] or are a distinct group.[32,168] Nonglial Tumors
Tumors of the Choroid Plexus

Choroid plexus papilloma is a rare, benign tumor most common in children younger than 2 years of age. The choroid plexus carcinoma is the malignant form of this tumor. The CGAP database describes 15 cases[1,22,50,69,71,147,165,175,196,202,203,228] of choroid plexus papilloma or carcinoma. The karyotypes are predominantly near-diploid, hypodiploid, or hypotriploid, and are characterized by whole chromosomal gains and losses; these include gains of chromosomes 5, 6, 7, 8, 9, 12, 15, 18, and 20 and losses of chromosomes 1, 3, 10, 16, 17, 21, and 22. In two cases,[50,69] diploid karyotypes were characterized by structural changes, including translocations, deletions, inversions, and the presence of markers, which were present as subclones. All other karyotypes showed no structural aberrations.

The Progenetix website summarizes 41 cases of choroid plexus tumors (not otherwise specified)[95,222] and 15 WHO Grade III carcinomas.[222] Of the 41 choroid plexus tumors, gains were identified on chromosomes 5 (up to 53% from loci 5p14-q13), 7 (53%), 8 (24%), 9 (34%), 12p (36%), 12q (17–31%), 15 (24.4%), 18 (19.5%), and 20 (19.5%). Losses were restricted to chromosomes 1 (9.8%), 2 (7.3%), 3 (14.6%), 10 (43%), 16 (9.8%), 17 (12.2%), 21 (19.5%), and 22 (36.6%). Grade II carcinomas showed a different pattern of change among the 15 cases. All chromosomes showed almost equal instances of gains and losses, with the following exceptions: gains of chromosome 1 in 40% of cases; loss of chromosome 3 in 26.7% of cases; loss of chromosome 6 in 26% of cases; gain of chromosome 12 in 60%; gain of chromosome 14 in 40%; loss of chromosome 15 in 26.7%; loss of chromosome 16 in 20%; loss of chromosome 18 in 33%; gain of chromosome 20 in 53%; gain of 21 in 33%; and loss of most or all of chromosome 22 in up to 72% of cases. Other regions of prominent gain were identified at 7q11-q31 (33%), 8q11-q23 (46%), and chromosome 4q (40%).

In the CGH study conducted by Rickert and colleagues[222] which investigated the genomic changes between choroid plexus papillomas and choroid plexus carcinomas, chromosomal imbalance differences characteristic of a tumor entity or age group were identified. In choroid plexus papillomas, +5q, +6q, +7q, +9q, +15q, +18q, and -21q were found to be significantly more common, whereas choroid plexus carcinomas were characterized by +1, +4q, +10, +14q, +20q, +21q, -5q, -9p, -11, -15q, and -18q. Among choroid plexus papillomas, the gains +8q, +14q, +12, and +20q occurred more often in children, whereas adults mainly presented with +5q, +6q, +15q, +18q, and -22q. On their own, the number of overall aberrations as well as gains and losses had no significant effect on survival among patients with choroid plexus tumors; however, a significantly longer survival duration among patients with choroid plexus carcinomas was associated with +9p and -10q.
Pineal Parenchymal Tumors
Pineocytomas, Pineoblastomas, and Mixed Pineocytoma/Pineoblastoma

Pineal parenchymal tumors arise from pineocytes or their precursors, and they are distinct from other pineal gland neoplasms such as astrocytic and germ cell tumors. Pineocytomas (WHO Grade II) are a slow-growing pineal parenchymal neoplasm that primarily occurs in young adults, accounting for less than 1% of all brain tumors, and they comprise approximately 45% of all pineal parenchymal tumors. Adults 25 to 35 years of age are most frequently affected, with a 5-year survival rate of greater than 80%. Pineoblastomas (WHO Grade IV) are a generally rare but highly malignant primitive embryonal tumor of the pineal gland manifesting primarily in children. Tumors similar in appearance to pineoblastomas have been observed in patients with familial (bilateral) retinoblastoma. Outcomes are generally favorable with appropriate treatment. Pineal parenchymal tumors of intermediate differentiation are monomorphous lesions exhibiting moderately high cellularity, mild nuclear atypia, occasional mitosis, and the absence of large pineocytomatous rosettes. They comprise approximately 10% of all pineal parenchymal tumors and occur in all age groups, with varying clinical outcomes. A comprehensive discussion of the pathogenesis and cytogenetic aspects of pineal region neoplasms is reviewed by Taylor and associates.[266]

The CGAP website reports three cases of pineocytomas,[18,60,205] with one case in the diploid range, one hypodiploid, and the other hyperdiploid. No specific pattern of chromosomal change is evident, although whole gains, losses, and structural abnormalities can be found. For pineoblastomas, the CGAP site reports four cases,[30,225,258] which are characterized by near-diploid karyotypes and whole chromosomal gains and losses.

A CGH study conducted by Rickert, et al.,[221] consisted of nine pineal parenchymal tumors, including three pineocytomas (WHO Grade II), three pineal parenchymal tumors of intermediate differentiation (WHO Grade III), and three pineoblastomas (WHO Grade IV). On average, 0 chromosomal changes were detected per pineocytoma, 5.3 per pineal parenchymal tumor of intermediate differentiation (3.3 gains compared with 2.0 losses), and 5.6 per pineoblastoma (2.3 gains compared with 3.3 losses). The most frequent DNA copy number changes among pineal parenchymal tumors of intermediate differentiation and pineoblastomas were gains of 12q (three of six cases) and 4q, 5p, and 5q (two of six each), as well as losses of 22 (four of six), 9q, and 16q (two of six each). Among pineal parenchymal tumors of intermediate differentiation, the most common chromosomal imbalances were +4q, +12q, and -22 (two of three cases each), and in pineoblastomas they were –22 (two of three). Five high-level gains were identified, all of them in pineoblastomas; these were found on 1q12-qter, 5p13.2-14, 5q21-qter, 6p12-pter, and 14q21-qter.

Regarding clinical outcome, all patients with pineocytomas and pineal parenchymal tumors of intermediate differentiation were alive after a mean observation time of 142 and 55 months, respectively, whereas all patients with pineoblastomas had died after a mean of 17 months, indicating that pineal parenchymal tumors of intermediate differentiation are cytogenetically more similar to pineoblastomas and prognostically more similar to pineocytomas. Imbalances in higher-grade pineal parenchymal tumors were mainly affected by gains of 12q and losses of chromosome 22.

An interesting case study published by Sawyer and coworkers[238] reported on a 6-month-old girl with a PNET of the pineal region. The tumor exhibited a constitutional reciprocal translocation t(16;22)(p13.3;q11.2~2), suggesting that the presence of this translocation, specifically the breakpoint at 22q11.2~2, may have predisposed the patient to the development of the tumor.
Tumors With Neuroblastic or Glioblastic Elements (Embryonal Tumors)
Medulloepithelioma

Medulloepitheliomas are a rare type of neuroepithelial tumor affecting young children. These tumors are usually found in the brain or retina, and are composed of primitive neuroepithelial cells lining the tubular spaces. Because of their classification as embryonal tumors and the presence of neuroblastic and glioblastic elements, specific reports are rare; however, the CGAP website reports one case of an ocular medulloepithelioma[21] with a diploid karyotype showing the loss of chromosome 15 and a balanced translocation between chromosomes 1 and 16. A second line within this tumor identified the same aberrations, in addition to the partial deletion of one of the chromosome 6 homologs.
Multipotent Differentiating PNETs: Medulloblastoma, Supratentorial PNETs, Medullomyoblastomas, Melanocytic Medulloblastoma, Desmoplastic Medulloblastoma, and Large-Cell Medulloblastoma

The major groups of PNETs are medulloblastomas and supratentorial PNETs, with variants including medullomyoblastomas, melanocytic medulloblastomas, and desmoplastic medulloblastomas. Med-ulloblastomas (WHO Grade IV) are a malignant, invasive embryonal tumor of the cerebellum that occurs primarily in children, has a predominantly neuronal differentiation, and has a tendency to metastasize through cerebrospinal fluid pathways. In adulthood, 80% of medulloblastomas occur in people 21 to 40 years of age. Medulloblastomas have been diagnosed in several familial cancer syndromes, including TP53 germ-line mutations, the nevoid basal cell carcinoma syndrome, and Turcot syndrome Type 2.[204] Supratentorial PNETs (WHO Grade IV) are embryonal tumors in the cerebrum or suprasellar region that are composed of undifferentiated or poorly differentiated neuroepithelial cells, which have the capacity for differentiation along neuronal, astrocytic, ependymal, muscular, or melanocytic lines. These tumors are also known as cerebral medulloblastoma, cerebral neuroblastoma, cerebral ganglioneuroblastoma, “blue tumor,” and PNET. These lesions are generally rare and occur in children.
Tumors of Cranial and Spinal Nerves
Schwannoma (Neurinoma, Neurilemoma)

The CGAP database reports 79 schwannomas.[15,54,80,90,124,151,164,175,208,212,236,259,268,290] The karyotypes can be generally characterized as diploid, the partial or complete loss of chromosome 22 being the most common chromosomal aberration, and occurring in approximately half of the cases. Other recurrent abnormalities include the loss of a sex chromosome as well as the losses of chromosomes 17 and 15. Gains of chromosome 7 occur, albeit in low frequencies.

There are few CGH studies of schwannomas,[7,128,289] and in all three of these, comparisons between sporadic schwannomas and those associated with NF2 were discussed. Warren and colleagues[289] screened 76 vestibular schwannomas obtained in 76 patients (66 sporadic and 10 NF2-related tumors), and found the most common change was the loss of chromosome 22, which was more frequent in sporadic than in NF2-related tumors. In 10% of cases, gains of 9q34 were detected. Other gains were detected on regions of chromosomes 10, 11, 13, 16, 19, 20, X, and Y. Koga, et al.,[128] examined 50 cases of PNSTs; among them were 14 schwannomas, which revealed similar findings. Nevertheless, the authors noted that the schwannomas had losses of chromosomes 17 and 19 in less than 50% of cases, compared with other PNSTs. Finally, Antinheimo and coworkers[7] analyzed 25 schwannomas (12 NF2 and 13 sporadic schwannomas), with some chromosomal regions further studied using LOH or FISH analysis. The CGH analysis detected genomic abnormalities in 60% of schwannomas, with the most common alteration being the loss on 22q, which was detected in 32% of cases. No consistent changes were detected in other chromosomal regions between the sporadic and NF2-associated cases.

Due to the high frequency of chromosome 22 loss in both sporadic and NF2-associated tumors, array-based studies across the genome as well as chromosome 22 have been conducted. Mantripragada, et al.,[158] constructed an array spanning 11 million bp of 22q encompassing the NF2 gene, with 100% coverage and a mean resolution of 58 kb. Moreover, the 220-kb genomic sequence encompassing the NF2 gene was covered by 13 cosmids to further enhance the resolution of analysis. Using CGH methods, the array was used to map and size chromosome 22q deletions, around NF2, in 47 sporadic schwannomas. Heterozygous deletions were detected in 45% of the lesions. Of these 21 tumors with deletions, the authors identified three profiles to classify the lesions. The predominant profile (in 12 of 21) was a continuous deletion of the 11-Mb segment, consistent with monosomy 22. The second profile (five of 21), was also in agreement with a continuous 11-Mb heterozygous deletion, but displayed a distinctly different level of deletion when compared with the first profile. The third profile was composed of four cases displaying interstitial deletions of various sizes. Two did not encompass the NF2 locus, emphasizing the importance of other loci in schwannoma pathogenesis. Similarly, to determine the frequency and extent of deletions, Bruder, et al.,[41] examined a 7-Mb interval in the vicinity of the NF2 gene by using high-resolution array-based CGH in 116 patients with NF2.At least 90% of this region of 22q, around the NF2 locus, was included on the array. Deletions were detected in eight severe, 10 moderate, and six mild cases. The authors concluded that these findings do not support the correlation between the type of mutation affecting the NF2 gene and the disease phenotype occurring in constitutional or tumor-derived DNA.

Cellular schwannomas appear to possess chromosomal[124,151,208,259] changes similar to classic schwannomas, namely the near-diploid karyotypes and involvement of chromosomal aberrations involving the sex chromosomes and chromosomes 22 and 7. In two published cases by Rao, et al.,[208] one patient displayed whole chromosome losses of one homolog of chromosomes 13 and 22 as well as the presence of a marker chromosome. The other patient, interestingly, was found to have an apparently balanced translocation between chromosomes 1 and 17 (t(1;17)(p12;q11.2)), with the 17q breakpoint occurring in the region of the NF1 gene. Cytogenetic analysis of a rare variant, the plexiform cellular schwannoma, was recently published by Joste, et al.,[124] and the authors reported the presence of clonal numerical changes consisting of an additional chromosome 17 in one clone and the addition of chromosomes 17 and 18 in the second one.
Perineuriomas

The literature shows few publications on the cytogenetics of perineuriomas; however, the CGAP website reports two perineuriomas[75,164] showing abnormal diploid karyotypes, with aberrations involving chromo-some 22 appearing in both cases. Mott and associates[173] reported a soft-tissue perineurioma with the sole abnormality of the loss of chromosome 13. A FISH analysis for chromosome 22 showed no apparent deletion or structural abnormalities.
Neurofibroma

The CGAP reports eight cases of neurofibromas[50,172,174,177,213,216] with diploid karyotypes. Structural aberrations are present, including balanced and unbalanced translocations, additions, and deletions, as well as marker chromosomes. No consistent chromosomal aberration appears among these cases.

The Progenetix website reports 40 cases of neurofibromas, both sporadic and NF1 associated,[128,188,243] showing fairly well-defined patterns of chromosomal gains and losses, including gains from 1p31-p11, 1q, 2q21-q34, 3, 4, 5p15-qter, 5q14-q21, 6q11-q25, 8p, 8q11-q23, 12q11-q21, 13, 13q21-q32, and 18. Losses were restricted primarily to chromosomes 9, 14q31-qter, 16, 17p, 17q, 19, 20, 21, and 22. Within these studies, the differences between sporadic and NF1-associated neurofibromas were assessed. Koga, et al.,[128] examined 27 neurofibromas. Both sporadic and NF1associated neurofibromas exhibited losses atchromosome 22q in more than 50% of cases. In NF1-associated neurofibromas, the most frequent losses were found in chromosomes 17p11.2-p13 (60% of cases), 17q24-25 (40%), 19p13.2 (53%), and 19q13.2-qter (53%). Gains were more frequently detected in plexiform neurofibromas (two of three cases) than in other benign tumors included in this study.

These findings were recapitulated in another publication by this group,[129] in which 12 NF1-associated tumors and 12 sporadic cases were analyzed using CGH. In this study, chromosomal imbalances were more common in NF1-associated tumors than in sporadic neurofibromas, with both groups showing predominantly more losses than gains, and it revealed novel chromosomal imbalances including chromosomes 17, 19, and chromosome arm 22q, which may be related to oncogenes or tumor suppressor genes in neurofibromas. For NF1-associated neurofibromas, the most frequent losses (minimal common regions) were found in chromosomes 17p11.2-p13, 17q24-q25, and 19p13. In addition, both NF1-associated and sporadic neurofibromas often exhibited losses at chromosome arms 19q and 22q, with 19q13.2-qter in NF1 tumors.
Malignant PNSTs

The CGAP website reports 93 malignant PNSTs and triton tumors (with references to large studies[80,122,164,166,197,208,244]). The majority of cases show complex karyotypes, with a near-triploid or tetraploid chromosomal complement. Whole chromosomal gains and losses affecting all chromosomes are apparent, with no specific pattern of genomic change. Plaat and colleagues[197] used computer analysis to determine recurrent cytogenetic alterations in 51 malignant PNSTs, including 44 from the literature and seven new cases. In addition, direct cytogenetic comparison between NF1-associated and sporadic malignant PNSTs was also performed. Significant losses (p < 0.05) were observed at chromosomal regions 9p2, 11p1, 11q2, and 18p1. Also, losses at 1p3, 9p1, 11q1, 12q2, 17p1, 18q1-q2, 19p1, 22q1, X, and Y were detected. Chromosome 7 was frequently gained, especially at 7q1 (p < 0.05). The most frequently involved breakpoints were identified at 1p13, 1q21, 7p22, 9p11, 17p11, 17q11, and 22q11.

The cytogenetic differences between NF1-associated and sporadic malignant PNSTs included a relative loss of chromosomal material in NF1-associated malignant PNSTs at 1p3, 4p1, and 21p1-q2 as well as the relative gain in 15p1q1. Differences in breakpoints between NF1-associated and the sporadic malignant PNST group were observed at 1p2122, comprising 28% of NF1 compared with 0% of sporadic malignant PNSTs, 1p32-34 (17 compared with 0%), 8p1112 (7 compared with 27%), and 17q10-12 (24 compared with 7%). Thus, the authors theorized that losses in 9p2 and gains in 7q1 could be of oncogenetic importance in malignant PNSTs; and the uncommon loss of 17q1, the NF1 region (17q11.2), in NF-associated malignant PNSTs may reflect a difference in the oncogenic pathway between NF1associated and sporadic malignant PNSTs.

Recently, Bridge and colleagues[39] used conventional cytogenetics and FISH and SKY analysis to identify recurrent chromosomal aberrations and breakpoints in malignant PNSTs and triton tumors. Twenty-one malignant PNSTs and malignant triton tumors obtained in 17 patients (nine with peripheral neurofibromatosis, also called NF1) were analyzed using standard G-banding and SKY studies for a subset of cases. These findings revealed structural aberrations most frequently occurring at 1p31-36, 4q28-35, 7p22, 11q22-23, 19q13, 20q13, and 22q11-13. Overall, loss of chromosomal material was much more common than gain. Loss of chromosomes or chromosomal regions 1p36 (48%), 3p21-pter (52%), 9p23-pter (57%), 10 (48%), 11q23-qter (48%), 16/16q24 (62%), 17 (43%), and 22/22q (48%), and gains of 7/7q (29%) and 8/8q (29%) were most prominent, with gains and losses distributed equally between malignant PNST and malignant triton tumor.

The Progenetix website reports 70[128,163,244,245] malignant PNSTs showing the frequent gain of 1q (7.1 to 33%), 2 (11.4%), 2p16-2p23 (11.4 to 20%), 3 (17%), 4 (10%), 5p (35%), 5q (10%), 6p (20 to 25%), 6q (20%), 7p (37%), 7q (25%), 8p (14%), 8q (48%), 12p (21%), 12q (17%), 16p (14%), 17q (48%), 18q (18%), 20q (24%), and 21 (20%). Losses are present on all chromosomes at a low frequency; however, predominant losses occur on 1p (8.6%) and 17p (15.7%). Amplifications are also present at 5p, 7p, 8q (8q21-8qter), 12q21, and 17q24.

Identification of the genomic changes between the malignant and other PNSTs based on CGH was examined by Koga and coworkers[128] and included 50 cases, consisting of nine malignant PNSTs, 27 neurofibromas (with three plexiform neurofibromas), and 14 schwannomas. These authors’ findings were consistent with other CGH studies of the PNSTs, showing gains and amplifications occurring more prominently in the malignant forms, reflecting tumor progression along oncogenic pathways, and losses and deletions were more common in the benign forms, suggesting a pathway involving tumor suppressor genes. In addition, Perry, et al.,[195] used interphase FISH for NF1 (17q), NF2 (22q), p16 (9p), and EGFR (7p) to examine malignant PNST and morphologically similar neoplasms, to determine whether these alterations are involved in malignant PNST tumorigenesis. Twenty-two malignant PNSTs (nine NF1-associated, 13 sporadic), 13 plexiform neurofibromas, five cellular schwannomas, eight synovial sarcomas, six fibrosarcomas, and 13 hemangiopericytomas were analyzed using two-color FISH. The NF1 deletions, often in the form of monosomy 17, were found in malignant PNSTs (76%), neurofibromas (31%), hemangiopericytomas (17%), and fibrosarcomas (17%), but not in synovial sarcomas or cellular schwannomas. The NF1 losses were encountered more frequently in malignant PNSTs compared with other sarcomas (p < 0.001), as were p16 homozygous deletions (45 compared with 0%; p < 0.001), EGFR amplifications (26 compared with 0%; p = 0.006), and polysomies for either chromosomes 7 (53 compared with 12%; p = 0.003) or 22 (50 compared with 4%; p < 0.001). Hemizygous or homozygous p16 deletions were detected in 75% of malignant PNSTs, but not in benign ones (p < 0.001).

Triton tumors are a rare variant of the malignant PNST group characterized as possessing a mixture of cells with nerve sheath and skeletal muscle differentiation. Karyotypic studies[103,162,277,283] and three SKY studies[39,96,155] have revealed that these tumors possess karyotypic characteristics similar to malignant schwannomas. In most cases, structural aberrations involved chromosomes 7, 8, 17, and 22. Amplification of MYCC was also identified. Interestingly, both Magrini, et al.,[155] and Hennig, et al.,[103] report the detection of an isochromosome 8q as one of two aberrations in a diploid karyotype (four copies of isochromosomes 8q were detected by Magrini, et al., suggesting the formation of the isochromosome 8q (containing MYCC) as an early event of malignancy).
Summary

The reviews of cytogenetic and molecular cytogenetic findings presented here have shown both common and unique chromosomal changes to the various brain tumor types and are summarized in Table 1 and Fig. 6. The distribution of genomic alterations shows a general trend of aggressive tumor types possessing changes including -1p, +1q, 2, -3p, +3q, -4, +6, +7, +8q, -9, -10, +12, -17p, and +17q as well as the presence of gene amplification; whereas more benign tumor types are characterized by fewer cytogenetic changes, the prominent alteration being loss of chromosome 22. The integration of genomics and proteomics will facilitate a more comprehensive understanding of the relationship between DNA copy number and gene and protein expression levels in carcinogenesis, and will lead to better diagnosis, novel treatments, and better quality of life.
Other Molecular Cytogenetic Contributions to the Study of Brain Tumors

Although the greatest contribution of molecular cytogenetic analysis to the study of brain tumors has primarily been the identification of chromosomal changes and imbalances, molecular cytogenetic techniques have also been implemented in other areas of study.
Studying Amplicon/Gene Structure in Double Minute Chromosomes or Homogeneously Staining Regions

Understanding and revealing patterns of gene amplification within the cell yield important information on gene expression. Several investigations of gene amplification have used molecular cytogenetic analysis. We have examined EGFR amplification in vivo and in vitro,[185] and have found that although EGFR amplification is observed in one third of GBMs, only one GBM cell line has maintained this amplification in vitro to date. It is interesting that of all the common genetic alterations observed in GBMs, EGFR amplification is the only aberration with a proportionally low incidence in vitro, in comparison with its presence in patient samples. Also, because EGFR mutations in GBM manifest primarily, if not exclusively, in amplified form, few cell lines with mutation of endogenous EGFR are available, suggesting that in vivo environments select for EGFR amplification and mutation, whereas in vitro environments (cell culture), select against this gene alteration. To contrast directly the fates of EGFR amplification in vivo and in vitro, as well as to examine potential relationships between EGFR amplification and mutation, we established and maintained GBM explants as xenografts by serial passaging in nude mice. Analysis of EGFR copy number and EGFR mutation status in 11 patient tumors and their corresponding xenografts, as well as the monitoring of EGFR copy number during the establishment of a GBM cell line from a xenograft with amplified EGFR, indicate that selection for EGFR amplification is an in vivo phenomenon. Furthermore, during tumor propagation as a xenograft the EGFR mutation occurs only in tumors with EGFR amplification, and selection of amplified mutant EGFR over amplified wild-type EGFR is rapid and complete.

The genomic structure of EGFR amplification units has also been investigated in detail in brain neoplasms. Double minute chromosomes are extrachromosomal circular DNA that comprise amplicon structures ranging in both size and complexity. Double minute chromosomes have been reported in several brain neoplasms, including GBMs and neuroblastomas. Vogt and colleagues[284] investigated the genetic content and organization of the repeat elements in the double minute chromosomes of seven gliomas. A FISH analysis showed that the EGFR locus was never deleted from its expected location on chromosome 7 in the cells of all analyzed tumors; furthermore, the chromosome regions corresponding to each breakpoint were sequenced after PCR amplification, compared with the normal sequence, and shown to be normal. A FISH study with a probe spanning a deletion region in a subset of cases revealed that the deletion was specific to the double minute chromosomes and not the resident site on chromosome 7. These molecular and FISH findings led to the conclusion that the observed double minute chromosomes originated from a single founding extrachromosomal body. In each of the gliomas, the founding molecule was generated by a simple event that circularized a chromosome fragment overlapping EGFR. The authors hypothesized that this likely occurs from the postreplicative excision of a chromosomal fragment, whereby the two ends are ligated by microhomologybased nonhomologous end-joining.

The FISH assay has also been used in identifying other genes that may be coamplified within double minute chromosomes as well as homogeneously staining regions. Methods have been introduced for releasing chromatin fibers from interphase nuclei in preparation for FISH analysis.[100,303] With these approaches, the map order, orientation, and distance between closely linked genes can readily be determined by analysis of the linear arrangement of probes on released free chromatin fibers. Chromatin FISH mapping techniques conveniently bridge the resolution gap between pulsed-field gel electrophoresis and FISH metaphase mapping, so that molecular contigs generated by conventional molecular hybridization methodology can be oriented with respect to chromosomal landmarks and established genetic markers for each chromosomal region. We have used free chromatin to study the complex structures of amplified genes within the amplicons in homogeneously staining regions and double minute chromosomes in neuroblastoma (Fig. 7).[186] No regular reiterated amplicon repeat unit was present in the homogeneously staining regions, but detailed analysis of the configurations of the amplified genes within each array indicated that multiple rearrangements generated a complex homogeneously staining region amplicon structure.
Mouse Models and Xenograft Fidelity

Animal models for studying brain tumors have important implications for understanding tumor initiation, pathogenesis, progression, and treatment strategies. Molecular cytogenetic techniques have enabled investigators to probe the murine genome with considerable ease as a result of advancements in cloning techniques and the sequencing of the mouse genome. The most common use of cytogenetic analysis is to identify chromosomal aberrations in the mouse genome and to determine whether the chromosomal abnormalities detected in mouse tumors are comparable to those in humans.

Ding and associates[68] generated a transgenic mouse astrocytoma model by using the glial fibrillary acidic protein promoter to express oncogenic V(12)Ha-ras, specifically in astrocytes. The chimeras expressing high levels of V(12)Ha-ras in astrocytes died of multifocal malignant astrocytomas within 2 weeks, with 95% of the mice dying of solitary or multifocal low- and high-grade astrocytomas within 2 to 6 months. The astrocytomas derived from these mice were pathologically similar to human astrocytomas. Banding, FISH, and SKY analyses revealed consistent clonal aneuploidies of chromosomal regions that were syntenic with comparable loci altered in human astrocytomas. Extra copies of mouse chromosome 10 were identified as a major clonal abnormality on SKY as well as FISH analyses. Mouse chromosome 10 harbors regions that are syntenic with a large portion of human chromosome 12q; 12q13-q14 contains CDK4 and MDM2, which are amplified and overexpressed in approximately 10% of human GBMs. This was reflected by the overexpression of both CDK4 and MDM2 by both mouse astrocytoma cells compared with normal mouse astrocytes.

The occurrence of a trisomy for mouse chromosome 8 and an extra copy of chromosome 3 translocated to chromosome 18 were also detected at relatively high clonal frequency, with no obvious syntenic relationship to any human astrocytoma-specific genetic alterations reported to date. A mouse model for oligodendroglioma described by Weiss and associates[292] used the S100 beta promoter to generate transgenic mice expressing v-erbB, a transforming allele of EGFR, resulting in the development of low-grade oligodendrogliomas. Transgenic animals that were heterozygous for ink4a/arf or p53 developed high-grade tumors. The CGH assay was used and revealed loss of distal mouse chromosome 4, which is syntenic to the human chromosome 1p region that is commonly lost in oligodendroglioma.

Finally, animal models have also been used to test treatment strategies. Branle and coworkers[36] used CGH to monitor tumor profiles, particularly the loss of 1p and 19q, during the implantation and tumor development of glioma cell lines injected into mouse and rat brains. Moreover, Bradford and colleagues[35] used six characterized clonal cell lines derived from the VM spontaneous murine astrocytoma for drug sensitivity testing. Various drugs, including the Vinca alkaloids, nitrosoureas, adriamycin, and cis-platinum were used, and it was found that there was a relationship between chromosome number and sensitivity of a wide variety of cytotoxic drugs, including the aforementioned nitrosoureas, Vinca alkaloids, and cis-platinum as well as procarbazine and bleomycin, but not adriamycin or fluorouracil. Clones with small numbers of chromosomes were more resistant than were clones with gross polyploidy.

Xenografts have been an invaluable method for regenerating precious fresh tissue specimens and understanding tumor biology in an in vivo setting; however, a xenograft’s value lies in its ability to reflect the original tumor. Many of the studies referenced in this review have resulted from cell lines or primary tissues engrafted into animals. Monitoring by FISH, SKY, or CGH techniques has shown that most engraftments maintain the integrity of the original genome.[120] Notably, the available model systems, whether at the cellular, tissue, or animal level, do not accurately represent the biology of human brain tumors. This is particularly true for EGFR amplification/mutation in glioblastoma.

There have been multiple attempts at developing a model to investigate the effects of EGFR mutation in GBMs. These include the introduction of altered EGFR complementary DNAs into cultured brain tumor cells, developing EGFR mutant transgenic mice, and transducing mutant EGFR in vivo by using viral constructs. The latter two approaches reflect a growing interest in the use of genetically engineered mice to study brain tumors. Nevertheless, such models also have limitations, and it seems that none of the approaches developed to date provide as accurate a representation of human GBMs with EGFR amplification/mutation as direct xenograft transplantation and propagation of material resected in patients. Alternative model systems cannot accurately represent the overexpression of EGFR that occurs in primary tumors with EGFR amplification, and it is quite likely that high-level expression is key for obtaining model biological properties that are consistent with those of human glioblastoma. Therefore, the xenograft model represents the most valid system with which to study the molecular biology associated with and to test therapeutic strategies directed against this gene alteration.

The FISH method monitors the amplification of EGFR in this model system very well (Figs. 9 and 10). Also, in the majority of the cell culture model systems developed to describe EGFR mutations, the expression and retention of the exogenous mutant EGFR is imposed on tumor cells. These transfected GBM lines lose their mutant EGFR expression unless they are maintained in media containing a high concentration of selective drugs, which clearly is not the case in vivo. The models currently available have not facilitated the development of a GBM model that is readily accessible for studying the effects of high-level EGFR signaling that results from amplification of EGFR. The lack of such models has hindered a large number of neurooncology researchers in their efforts to answer straightforward questions regarding the effects and treatment of EGFR amplification in these tumors.