J Neurosurg 77:432-437, 1992
Clonal composition of glioblastoma multiforme RICHARD A. BERKMAN, M.D., W. CRAIG CLARK, M.D., PH.D., ABHA SAXENA, PH.D., JAMES T. ROBERTSON, M.D., EDWARD H. OLDFIELD, M.D., AND IQBAL U. ALl, PH.D. Surgical Neurology Branch, National Institute of Neurologic Disorders and Stroke, National Institutes of Health, Bethesda, Maryland, and Department of Neurosurgery, University of Tennessee, Memphis, Tennessee v, Glioblastoma multiforme, the most common and most lethal primary central nervous system neoplasm, is noted for its phenotypic and biological heterogeneity. This heterogeneity may result from genetic alterations accumulated by a single transformed astrocyte as it evolves into a monodonal tumor. Alternatively, it may be attributed to the presence of multiple biologically and genetically distinct astrocytic populations within a polyclonal tumor. To address the issue of clonal composition of glioblastoma multiforme the authors used two independent approaches: analysis of X-chromosome inactivation and analysis of chromosomes 10 and 17 for tumor-specific somatic deletions. The analysis included 10 tumors from nine female patients with glioblastoma multiforme (eight primary and two recurrent tumors), who were heterozygous at either of two X-chromosome genes (hypoxanthine phosphoribosyl-transferase or phosphoglycerate kinase). Nine glioblastomas multiforme demonstrated a monoclonal pattern on X-chromosome analysis; contamination with normal tissue obscured the analysis in one tumor. Somatic deletions on chromosomes 10 and/or 17 occurred in nine tumors, supporting a monoclonal composition for these tumors. These data suggest that glioblastoma multiforme is a monoclonal neoplasm, derived from the clonal expansion of a single transformed astrocyte that has, as a fundamental step in tumorigenesis, sustained a critical genetic alteration on chromosome 10 and/or 17. KEY WORDS 9 clonal composition X - c h r o m o s o m e inactivation analysis
LIOBLASTOMA multiforme is composed of distinct subsets of neoplastic astrocytes that differ in histological morphology, biological activity, metastatic potential, and radiation sensitivity. 5'~6'2s Such heterogeneity within a single tumor may arise from the many genetic and epigenefic alterations that occur as a single transformed cell evolves into a monoclonal tumor, or it may arise by the participation of multiple genetically distinct cell populations within a polyclonal tumor. While rare, a polyclonal neoplasm might be predicted to occur when a field of cells is exposed to an endogenous or an exogenous carcinogenic stimulus. Ionizing irradiation and intrathecal chemotherapy, for example, induce solitary and multifocal glioblastomas multiforme in humans ~3'24and oncogenic viruses induce multifocal gliomas in animals, 34 supporting the possibility of a polyclonal composition of glioblastomas multiforme. However, a polyclonal composition would be inconsistent with the recent findings of clonal alterations on chromosomes 10 and 17 in some glioblastomas multiforme. 8't4'2~ Therefore, to address the issue of clonality in this type of tumor, we applied two independent techniques: 1) X-chromosome inacti-
G
432
9
glioblastoma multiforme
9
somatic deletion
9
vation analysis, described originally by Vogelstein, et aL; 32'33 and 2) restriction fragment length polymorphism analysis of matched lymphoeytes and tumors to detect tumor-associated alMic deletions. Analysis of X-chromosome inactivation is based on the Lyon hypothesis, which states that, in each female embryonic cell, one of the parental X chromosomes is randomly inactivated and is then faithfully maintained inactive in the progeny of that embryonic cell. 22 Deoxyribonucleic acid extracted from tumor cells can be analyzed to determine whether all of the cells have the same parental X chromosome inactivated, which would indicate that the tumor evolved from a single transformed progenitor cell (monoclonal neoplasm), or whether there is a mix of cells with both parental X chromosomes inactivated, which would indicate that the tumor evolved from multiple transformed progenitor cells (polyclonal neoplasm). This is accomplished by using restriction fragment length polymorphisms at defined genes to differentiate the maternal and paternal X chromosomes and a methylation-sensitive enzyme to identify which parental X chromosome has been inactivated. 3t
J. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma multiforme Analysis of tumor-specific chromosomal alterations provides another approach to assess the clonal composition of tumors. Some human tumors, of which retinoblastoma is the prototype, result from an inherited or acquired specific mutation at a locus on one copy of a particular chromosome. 25 This is followed by another mutational event, frequently a deletion, on the homologous chromosome. A monoclonal tumor, derived from a single progenitor cell that has sustained a critical somatic deletion, will be populated with cells that all share a deletion at the same locus. In contrast, a polyclonal neoplasm, derived from multiple transformed progenitor cells, will be populated with some cells that have maintained heterozygosity at the informative locus, thereby obscuring the unequivocal demonstration of a somatic deletion. We screened the peripheral blood lymphocytes (PBL's) of 21 female patients with glioblastoma multiforme for polymorphisms at the two X-chromosomelinked genes, hypoxanthine phosphoribosyl-transferase (HPRT) and phosphoglycerate kinase (PGK). Ten tumors (eight primary and two recurrent tumors) from nine patients, who were constitutionally heterozygous at the HPRT or PGK locus, were used for X-chromosome inactivation analysis. The results were verified by examining the tumors for somatic deletions at several loci which span both arms of chromosomes 10 and 17. Materials and Methods
Isolation of DNA Normal human brain tissue was obtained at temporal lobectomy for intractable epilepsy. Glioblastomas multiforme were removed at surgery with meticulous efforts to isolate tumor from surrounding brain. The specimens were immediately frozen in liquid nitrogen and stored at -70~ The tissue diagnoses were confirmed by microscopic pathology according to standard cri.teria3 Genomic DNA was obtained from a human glioblastoma multiforme cell line (SNB 110), PBL, normal human brain tissue, and 10 glioblastomas multiforrne by sodium dodecyl sulfate (SDS)-proteinase K digestion and phenol/chloroform extraction according to established methods] 6 X-Chromosome Inactivation Analysis Analysis for X-chromosome inactivation was performed according to the method described by Vogelstein, et al., 3~with the following modifications. Locus for HPRT. Genomic DNA (10 ug) from a glioblastoma multiforme cell line, PBL's (Case 8), and glioblastoma multiforme tissue (tumors from Cases 8 and 43) was digested to completion with BamHl at 37"C according to conditions recommended by the manufacturer. Half of the samples were then digested at 37"C with 9 U/ug of the methylation-sensitive enzyme HpalI. Digested DNA was size-fractionated through a 0.8% agarose gel and then transferred to Nytran paper in 10 x standard saline citrate (SSC). Filters were baked under vacuum for 2 hours at 80"C. J. Neurosurg. / Volume 77/September, 1992
Locus)or PGK. Genomic DNA (6 ug) from PBL's (Case 20), normal human brain tissue, and glioblastoma multiforme tissue (tumors from Cases 20, 28, 36 first and second recurrence, 37, 39, 63, and 88) was digested with EcoRI and BglI in buffers according to the manufacturer's recommendation, then digested with BglII in 120 ul of 10 mM Tris, 1 mM ethylenediamine tetraacetic acid (EDTA, pH 8.0) with the addition of 25 ul of glycine buffer (100 mM Tris, 100 mM glycine, 100 mM MgCI2, 1 M NaC1, 10 mM 1,4-dithiothreitol, pH 9.5). Samples were ethanol-precipitated and divided into equal aliquots. Digestion with HpaII, gel electrophoresis, and DNA transfer were performed as described above. Somatic Deletion Analysis Genomic DNA (10 izg) from matched PBL and glioblastoma multiforrne tissue (tumors from Cases 8, 20, 28, 36 first and second recurrence, 37, 39, 43, 63, and 88) was digested with the appropriate restriction enzyme: DIOSI (Bg/lI), D10S4 (TaqI), D10S28 (TaqI), and D 17S34 (TaqI) at 37"C. Electrophoretic separation of DNA and transfer to Nytran membrane were performed as described above. Hybridization Insert probes HPRT-800, PGK, D10S1, D10S4, D10S28, and D17S34 were radiolabeled by random priming to a specific activity of more than 108 cpm/ug of DNA. 1~ The filters were prehybridized in a buffer containing 50% formamide, 5 x standard sodium phosphate EDTA, 1 x Denhart's medium, 0.1% SDS, and 5 mg denatured single-stranded salmon sperm DNA. Hybridization of the radiolabeled probes (5 to l0 • 10~ cpm/ml) to the appropriate filters was performed for 24 to 36 hours at 37*C in the same buffer with the addition of 7% dextran sulfate. Filters were then washed once in 2 x SCC and 0.5% SDS at room temperature and three times at 65"C in 0.1 x SSC and 0.5% SDS. Phosphoglycerate kinase blots were washed a fourth time at 68*C in 0.1 x SSC and 0.5% SDS. This stringent washing condition was necessary to eliminate cross-hybridization to a 1.7-kb PGK-related fragmentY The blots were exposed to Kodak XAR 2 film for appropriate intervals at -70~ with intensifying screens. Both methods of clonal analysis (X-chromosome inactivation and somatic deletion) were repeated at least three times for each tumor to minimize the effects of potential variabilities in experimental conditions. Results
X-Chromosome Inactivation Analysis Lymphocyte DNA from 21 female glioblastoma multiforme patients was screened for polymorphisms at the HPRT and PGK loci. Constitutional heterozygositywas detected in two of these patients (Cases 8 and 43) at the HPRT locus and in seven patients (Cases 20, 28, 36, 37, 39, 63, and 88) at the PGK locus. First and second tumor recurrences were obtained from Case 36; prior 4313
R. A. Berkrnan, et al. TABLE 1
Clonal composition of GBM:X-chromosome inactivationand tumor-specific allelicdeletionanalyses* Case No.
X-Chromosome Inactivation
Allelic Deletions: Chromosome:~
HPRT
PGK
10
17
8 20
monoclonal N.I.
N.I. monoclonal
-
+ +
28
N.I.
f
+t
-1
36R N.I. monoclonal + 36RR N.I. monoclonal + 37 N.I. monoclonal 39 N.I. monoclonal + 43 monoclonal N.I. + 63 N.I. monoclonal + 88 N.I. monoclonal + + * N.I. = not informative; GBM = glioblastoma multiforme; HPRT= hypoxanthinephosphoribosyl-transferase;PGK = phosphoglycerate kinase; R = first recurrence; RR = secondrecurrence. - = loss of heterozygosity;+ = maintenanceof heterozygosity. 1"See text for explanation. This represents a summaryof the results from 13 loci examined on chromosomes10 and 17. Only the resultsfrom four of the loci are presented in Fig. 3.
to both recurrences the patient had received irradiation and chemotherapy. The clonal status of these l0 tumors was determined by X-chromosome inactivation analysis at the HPRT and PGK loci. Peripheral blood lymphocytes have often been used as polyclonal tissue controls in previous X-chromosome inactivation experiments. However, DNA samples from PBL's sometimes fail to demonstrate balanced mosaicism.~9'27 This could be due to the clonal nature in which lymphocytes are generated and released into the circulation. Therefore, we have included normal human brain tissue, obtained at temporal lobectomy for intractable epilepsy, along with PBL's from two patients with glioblastoma multiforme, as polyclonal tissue controls. A glioblastoma multiforme cell line served as a control for monoclonal composition.
Locus for HPRT. Genomic DNA from a glioblastoma multiforme cell line, PBL's (Case 8), and two glioblastomas multiforrne (tumors from Cases 8 and 43) was digested with BamHI to reveal the HPRT polymorphic allele and to frame the HpalI methylation sites that are unmethylated in the allele on the inactive X chrom o s o m e ) : Equal aliquots were then treated with the methyl-sensitive enzyme HpaII in order to digest the inactive HPRT allele. Following treatment of PBL DNA with HpalI, a near-equal reduction in intensity of both the 12- and 24-kb HPRT alleles was observed. This is the expected result for polyclonal tissue in which there is a near-equal mix of cells with random inactivation of both paternal and maternal X chromosomes. In comparison, treatment of DNA from a control glioblastoma multiforme cell line and from two glioblastomas multiforme (from Cases 8 and 43) with HpaII resulted in near-complete digestion of only one of the alleles (the 12-kb inactive allele in the control and the 434
FIG. 1. X-chromosome inactivation analysis at the hypoxanthine phosphoribesyl-transferase (HPRT) locus. The DNA from a glioblastoma multiforme cell line (GCL), peripheral blood lymphocytes (PBL), and two glioblastomas multiforme (GBMs) was digested with BamHl to reveal the allele polymorphism and to frame the HpaII methylation sites. Lanes marked with a plus were additionally digested with the methylsensitive enzyme HpaII. The numbers to the left represent the size of the HPRT alleles. Treatment of control PBL DNA with HpaII results in equal digestion of both HPRT alleles, consistent with polyelonal tissue. Similar treatment of DNA from the control GCL and the GBM cases shows an unequal digestion of one of the two alleles, the expected pattern for monoclonal tissue.
24-kb inactive allele in tumors from Cases 8 and 43), suggesting that all of the cells within each tumor sample have the same parental X chromosome inactivated, the expected pattern for monoclonal tissue (Fig. 1).
Locus for PGK. Genomic DNA from PBL's (Case 20), normal human brain tissue, and eight glioblastomas multiforme (tumors from Cases 20, 28, 36 first and second recurrence, 37, 39, 63, and 88) was digested with Bgll to reveal the PGK allele polymorphism, and with BgIII and EcoRI to frame the nine HpaII sites that are unmethylated in the PGK allele on the active X chromosome. Eight of these sites are methylated in the inactive PGK allele; the ninth site is methylated 60% to 80% of the time. 32 Digestion of the PBL and normal human brain DNA with HpaII resulted in nearequal reduction in the intensity of both PGK alleles. Therefore, both tissues contained an equal distribution of cells with either parental X chromosome inactive, as would be expected for polyclonal tissue. Following treatment of the DNA from eight glioblastomas multiforme with HpaII, complete or near-complete digestion of only one of the two alleles in seven of the tumors was observed, consistent with a monoclonal composition of these tumors; these were the 1.7-kb active allele in tumors from Cases 20, 37, and 63, and the 1.3-kb active allele in tumors from Cases 36 (first and second recurrence), 39, and 88 (Fig. 2). Tumor DNA from Case 28 showed only a slight preferential reduction in the intensity of the 1.3-kb allele; the remaining 1.3-kb signal may be due to contamination from normal tissue or clonally distinct subpopulations of tumor cells. Somatic Deletion Analysis Deoxyribonucleic acid from multiforme was also screened lelic deletions on chromosomes both analyses are summarized
all 10 glioblastomas for tumor-specific al10 and 17. Data from in Table 1. Multiple
J. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma rnultiforme
FIG. 2. X-chromosome inactivation analysis at the phosphogylcerate kinase (PGK) locus. The DNA from peripheral blood lymphocytes (PBL's), normal human brain tissue (NHB), and eight glioblastomas multiforme (GBMs) was digested with BglI, BgllI, and EcoRI to reveal the allele polymorphism and to frame the HpalI methylation sites. Lanes marked with a plus were additionally digested with the methyl-sensitive enzyme HpalI. The numbers to the left represent the size of the PGK alleles. Control DNA from NHB and PBL demonstrates a polyclonal pattern following treatment with HpalI, while DNA from seven GBM's reveals a monoclonal pattern. The pattern from the tumor in Case 28 is consistent with contamination from normal tissue (see Fig. 4 right).
polymorphic DNA probes (13 total) on both arms of chromosomes 10 and 17 were used in the analysis. Representative blots of matched sets of lymphocyte and tumor DNA, which were heterozygous at the D10SI, D10S4, D10S28, or D17S34 locus are displayed in Fig. 3. A decrease in the intensity of one of the two alleles occurred for nine of the 10 tumor DNA samples.
Tumor Histopathology Figure 4 shows representative pathological findings in two tumors. Tumor tissue from Case 37, in which the typical morphological cellular heterogeneity of glioblastoma multiforme is apparent is demonstrated in Fig. 4 left. A portion of the tumor from Case 28 is shown in Fig. 4 right; most of the sample consists of normal brain tissue with scattered infiltrating anaplastic astrocytes.
Discussion
Clonal Composition of Human Tumors Early studies of clonality of human tumors used Xchromosome-linked isoenzymes; in particular, electrophoretic variants of glucose-6-phosphate dehydrogenase were employed. Most human tumors analyzed by this technique were reported to be monoclonal, although there were a few exceptions reported as polyclonal, including some parathyroid adenomas, lj neurofibromas, ~2 and trichoepitheliomas. 15 More recently, variable methylation patterns of specific genes of the X chromosome have been used to assess the clonal composition of tumors. 32'33With this technique, many human tumors have been shown to be monoclonal, including colorectal carcinomas, 9 parathyroid adenomas, ~ meningiomas and schwannomas, ~9 and various pituitary adenomas. ~'~7'~8One exception has been the finding of monoclonal and polyclonal corticotroph adenomas of the pituitary, suggesting that mechanisms other than somatic mutation play a role in the genesis of corticotroph adenomas. 27
Glioblastoma Multiforme Clonal Analyses We used X-chromosome-linked inactivation of spe-
J. Neurosurg. / Volume 77/September, 1992
FIG. 3. Somatic deletion analysis showing loss of heterozygosity at loci on chromosomes 10 and I7 in nine cases of glioblastoma multiforme (GBM). The DNA from peripheral blood lymphocytes and GBM's was analyzed with polymorphic markers as described in the text and representative blots presented. Numbers 1 and 2 refer to the larger and smaller alleles, respectively, for the multiallele markers D10S1, D10S4, DIOS28, and D17S34. L = peripheral blood lymphocyte DNA; T = GBM tumor DNA; T~ = first tumor recurrence from Case 36; 1"2 = second tumor recurrence from Case 36. A somatic deletion, as demonstrated by a decrease in the intensity of one of the alleles at the four loci presented, occurred in nine of the 10 tumor DNA samples. cific genes in conjunction with somatic loss of heterozygosity on chromosomes 10 and 17 to assess the clonal composition of glioblastoma multiforme. Nine of the 10 tumors analyzed yielded a monoclonal pattern of X-chromosome inactivation (Figs. I and 2). Both recurrent tumors in Case 36 demonstrated inactivation of the same parental X chromosome and sustained an allelic deletion at the same locus on chromosome 10, consistent with a genetically identical second tumor recurrence. The one tumor that failed to demonstrate clearly a monoclonal composition (Case 28) was contaminated with normal tissue (Fig. 4 right). Uniform loss of DNA sequences on a specific chro435
R. A. Berkman, et al.
FIG. 4. Photomicrographs showing tumor histopathology. H & E, x 30. Left: Tumor from Case 37. Several subtypes of neoplastic astrocvtes are apparent, including small hyperchromic astrocytes, fibrillary astrocytes with dispersed chromatin, and several multinucleated giant astrocytes. Right. Tumor from Case 28 in which normal tissue predominates with some infiltrating anaplastic astrocytes.
mosome in a tumor represents strong evidence for its clonal origin. It is possible, however, for monoclonal tumors to show retention of heterozygosity at specific involved loci. This has recently been reported for meningiomas and schwannomas, which were monoclonal by X-chromosome inactivation analysis but did not consistently have allelic deletions on chromosome 22.19 Considering the heterogeneity of genetic alterations in human solid tumors, and/or the difficulty in detecting small localized deletions, it is not surprising to find monoclonal tumors that do not show clonal chromosomal alterations at involved loci. In human glial tumors, chromosomes 10 and 17 have been reported to be the targets of such mutational events. 8''4'2~ In our study, nine of 10 glioblastomas examined by somatic deletion analysis suffered allele losses on chromosomes 10 and/or 17 (Fig. 3). This high incidence of somatic deletions on chromosomes 10 and 17 suggests a fundamental role for these loci in glioblastoma multiforme tumorigenesis. In spite of the careful selection of tumor tissue for analysis of clonal composition, contamination with normal cells in surgical specimens of tamor sometimes cannot be avoided. The two recurrent tumors from Case 36 and the primary tumors from Cases 28 and 63 probably represent examples of contamination by normal tissue. Our data indicate that somatic deletion analysis is more sensitive to influence by contaminating normal tissue than is X-chromosome inactivation analysis. There are two likely explanations for this observation. First, while 100% of the contaminating normal cells will show maintenance of heterozygosity at the polymorphic loci, only about 50% of these cells will have a particular inactive X-chromosome PGK allele. In fact, this portion is probably less than 50% because of the differential methylation of one of the HpalI sites 436
in the PGK gene. n Second, a particular somatic deletion may be a late event in tumor evolution and, therefore, would not be shared by the entire tumor cell population which, nonetheless, contains the same inactive parental X chromosome. Either or both of these possibilities may explain the apparent discrepancy between the results of the X-chromosome inactivation and somatic deletion analyses. A monoclonal composition for glioblastoma multiforme is consistent with the somatic mutation theory of single-cell carcinogenesis and suggests that the phenotypic and biological heterogeneity of this tumor resuits from the accumulation of genetic alterations by a clonally evolving transformed astrocyte. However, it is possible that distinct polyclonal cells may be present in glioblastoma multiforme but are too sparse to alter the results of clonal analyses significantly. Multifocal gliomas occur sporadically,TM with inherited genetic disorders4"29'3~and in patients previously exposed to ionizing irradiation and intrathecal chemotherapy for other diseases, is'2' If multifocal glioblastomas multiforme arise by a mechanism of carcinogenesis similar to that of solitary glioblastomas multiforme, then their multifocal presentation is likely attributed to the direct extension or intraparenchymal metastasis of clonal cells from a primary focus of monoclonal origin. A true multifocal glioma, on the other hand, requires that multiple cells undergo independent neoplastic transformation events. Gliomatosis cerebri, a rare feature of glioma neoplasms in which malignant astrocytes are spread diffusely throughout the central nervous systemy s may also represent an exception to the singlecell mutation origin of monoclonal glioblastoma multiforme. Alternatively, it too may result from the diffuse spread of clonal cells from a monoclonal site. The clonai status of gliomatosis cerebri and multifocal glioJ. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma multiforme blastoma multiforme can be determined by analyzing tumor DNA obtained from separate tumor foci. Conclusions
Previous reports of loss of heterozygosity on chromosomes 10 and/or 17 in malignant glial tumors, ~:4'2~ and the results of the present study of X-chromosome inactivation in conjunction with tumor-associated allelic deletions, suggest that solitary glioblastomas are monoclonal tumors that are derived from the clonal evolution of a single transformed cell that has lost critical loci on chromosomes 10 and/or 17. Acknowledgments
We acknowledge the contributions of William Reinhold, B.S., and Barbara lkejiri, B.S., for their preparation of lymphocyte DNA samples. We are also grateful to Conrad Kufla, M.D., for supplying normal brain tissue obtained during seizure surgery. We are indebted to Burt Vogelstein, M.D., for generously supplying the probes HPRT and PGK. References
1. Alexander JM, Biller BMK, Bikkal H, et al: Clinically nonfunctioning pituitary tumors are monoclonal in origin. J Clin Invest 86:336-340, 1990 2. Arnold A, Stauton CE, Kim HG, et al: Monoclonality and abnormal parathyroid hormone genes in parathyroid adenomas. N Engl J Med 318:658-662, 1988 3. Barnard RO, Geddes JF: The incidence of multifocal cerebral gliomas. A histologic study of large hemisphere sections. Cancer 60:1519-1531, 1987 4. Balzdorf U, Malamud N: The problem of multicentric gliomas. J Neurosurg 20:122-136, 1963 5. Burger PC, Green SB: Patient age, histologic features, and length of survival in patients with glioblastoma multiforme. Cancer 59:1617-1625, 1987 6. Burger PC, Vogel FS: Surgical Pathology of the Nervous System and Its Coverings, ed 2. New York: John Wiley & Sons, 1982 7. Couch JR, Weiss SA: Gliomatosis cerebri. Report of four cases and review of the literature. Neurology 24: 504-511, 1974 8. EI-Azouzi M, Chung RY, Farmer GE, et al: Loss of distinct regions on the short arm of chromosome 17 associated with tumorigenesis of human astrocytomas. Proc Natl Acad Sei USA 86:7186-7190, 1989 9. Fearon ER, Hamilton SR, Vogelstein B: Clonal analysis of human colorectal tumors. Science 238:192-196, 1987 10. Feinberg AP, Vogelstein B: A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal Biochem 137:266-267, 1984 (Addendum) 11. Fialkow PJ, Jackson CE, Block MA, el al: Multicellular origin of parathyroid "adenomas." N Engl J Med 297: 696-698, 1977 12. Fialkow PJ, Sagebiel RW, Gartler SM, et al: Multiple cell origin of hereditary neurofibroma. N Engl J Med 284: 298-300, 1971 13. Fontana M, Stanton C, Pompili A, et al: Late multifocal gliomas in adolescents previously treated for acute lymphoblastic leukemia. Cancer 60:1510-1518, 1987 14. Fults D, Pedone C, Thomas GA, et al: Allelotype of human malignant astrocytoma. Cancer Res 50:5784-5789, 1990 15. Gartler SM, Zipkrewski L, Krakowski A, et al: Glucose6-phosphate dehydrogenase mosaicism as a tracer in the J. Neurosurg. / Volume 77 / September, 1992
study of hereditary multiple trichoepithelioma. Am J Hum Genet 18:282-287, 1966 16. Giangaspero F, Burger PC: Correlations between cytologic composition and biologic behavior in the glioblastoma multiforme. A postmortem study of 50 cases. Cancer 52: 2320-2333, 1983 17. Herman V, Fagin J, Gonsky R, et al: Clonal origin of pituitary adenomas. J Clin Endocrinnl Metab 71:1427-1433, 1990 18. Jacoby LB, Hedley-Whyte ET, Pulaski K, et al: Clonal origin of pituitary adenomas. 3 Neurosurg 73:731-735, 1990 19. Jacoby LB, Pulaski K, Rouleau GA, et al: Clonal analysis of human meningiomas and schwannomas. Cancer Res 511:6783-6786, 1990 20. James CD, Carlbom E, Dumanski JP, et al: Clonal genomic alterations in glioma malignancy stages. Cancer Res 48:5546-5551, 1988 21. James CD, Carlbom E, Nordenskjold M, et al: Mitotic recombination of chromosome 17 in astrocytomas. Proc Natl Acad Sei USA 86:2858-2862, 1989 22. Lyon MF: X-chromosome inactivation and developmental patterns in mammals. Biol Rev 47:1-35, 1972 23. Okazald H, Scheithauer BW: Atlas of Nenropathology. New York: Gower, 1988, p 72 24. Piatt JH Jr, Blue JM, Schold SC Jr, et al: Glioblastoma multiforme after radiotherapy for acromegaly. Neurosurgory 13:85-89, 1983 25. Ponder B: Gone losses in human tumours. Nature 335: 400-402, 1988 26. Sambrook J, Fritsch EF, Maniatis T: Molecular Cloning: A Laboratory Manual, ed 2. Cold Spring Harbor, NY: Cold Spring Harbor Laboratories, 1989 27. Schulte HM, Oldfield EH, Allolio B, et al: Clonal composition of pituitary adenomas in patients with Cushing's disease: determination by X-chromosome inactivation analysis. J Clin Endocrinol Metab 73:1302-1308, 1991 28. Shapiro JR, Yung WKA, Shapiro WR: Isolation, karyotype, and clonal growth of heterogeneous subpopulations of human malignant gliomas. Cancer Res 41:2349-2359, 1981 29. Sherman JL: Neurofibromatosis 1 (Recklinghausen disease) and neurofibromatosis 2 (bilateral acoustic neurofibromatosis). An update. Ann Intern Med 113:39-52, 1990 (see pp 42-43) 30. Todd DW, Christoferson LA, Leech RW, et al: A family affected with intestinal polyposis and gliomas. Ann Neurol 10:390-392, 1981 31. van der Ploeg LHT, Flavell RA: DNA methylation in human GDB-giobin locus in erythroid and nonerythroid tissues. Cell 19:947-958, 1980 32. Vogelstein B, Fearon ER, Hamilton SR, et al: Clonal analysis using recombinant DNA probes from the Xchromosome. Cancer Res 47:4806-4813, 1987 33. Vogelstein B, Fearon ER, Hamilton SR, et al: Use of restriction fragment length polymorphisms to determine the clonal origin of human tumors. Science 227:642-645, 1985 34. Walsh JW, Zimmer SG, Perdue ML: Role of viruses in the induction of primary intracranial tumors. Neurosnrgory 10:643-662, 1982. Manuscript received November 25, 1991. Accepted in final form February 11, 1992. Address reprint requests to: Iqbal U. All, Ph.D., Surgical Neurology Branch, NINDS, Building 10/Room 5D37, National Institutes of Health, 9000 Rockville Pike, Bethesda, Maryland 20892. 437
Clonal composition of glioblastoma multiforme RICHARD A. BERKMAN, M.D., W. CRAIG CLARK, M.D., PH.D., ABHA SAXENA, PH.D., JAMES T. ROBERTSON, M.D., EDWARD H. OLDFIELD, M.D., AND IQBAL U. ALl, PH.D. Surgical Neurology Branch, National Institute of Neurologic Disorders and Stroke, National Institutes of Health, Bethesda, Maryland, and Department of Neurosurgery, University of Tennessee, Memphis, Tennessee v, Glioblastoma multiforme, the most common and most lethal primary central nervous system neoplasm, is noted for its phenotypic and biological heterogeneity. This heterogeneity may result from genetic alterations accumulated by a single transformed astrocyte as it evolves into a monodonal tumor. Alternatively, it may be attributed to the presence of multiple biologically and genetically distinct astrocytic populations within a polyclonal tumor. To address the issue of clonal composition of glioblastoma multiforme the authors used two independent approaches: analysis of X-chromosome inactivation and analysis of chromosomes 10 and 17 for tumor-specific somatic deletions. The analysis included 10 tumors from nine female patients with glioblastoma multiforme (eight primary and two recurrent tumors), who were heterozygous at either of two X-chromosome genes (hypoxanthine phosphoribosyl-transferase or phosphoglycerate kinase). Nine glioblastomas multiforme demonstrated a monoclonal pattern on X-chromosome analysis; contamination with normal tissue obscured the analysis in one tumor. Somatic deletions on chromosomes 10 and/or 17 occurred in nine tumors, supporting a monoclonal composition for these tumors. These data suggest that glioblastoma multiforme is a monoclonal neoplasm, derived from the clonal expansion of a single transformed astrocyte that has, as a fundamental step in tumorigenesis, sustained a critical genetic alteration on chromosome 10 and/or 17. KEY WORDS 9 clonal composition X - c h r o m o s o m e inactivation analysis
LIOBLASTOMA multiforme is composed of distinct subsets of neoplastic astrocytes that differ in histological morphology, biological activity, metastatic potential, and radiation sensitivity. 5'~6'2s Such heterogeneity within a single tumor may arise from the many genetic and epigenefic alterations that occur as a single transformed cell evolves into a monoclonal tumor, or it may arise by the participation of multiple genetically distinct cell populations within a polyclonal tumor. While rare, a polyclonal neoplasm might be predicted to occur when a field of cells is exposed to an endogenous or an exogenous carcinogenic stimulus. Ionizing irradiation and intrathecal chemotherapy, for example, induce solitary and multifocal glioblastomas multiforme in humans ~3'24and oncogenic viruses induce multifocal gliomas in animals, 34 supporting the possibility of a polyclonal composition of glioblastomas multiforme. However, a polyclonal composition would be inconsistent with the recent findings of clonal alterations on chromosomes 10 and 17 in some glioblastomas multiforme. 8't4'2~ Therefore, to address the issue of clonality in this type of tumor, we applied two independent techniques: 1) X-chromosome inacti-
G
432
9
glioblastoma multiforme
9
somatic deletion
9
vation analysis, described originally by Vogelstein, et aL; 32'33 and 2) restriction fragment length polymorphism analysis of matched lymphoeytes and tumors to detect tumor-associated alMic deletions. Analysis of X-chromosome inactivation is based on the Lyon hypothesis, which states that, in each female embryonic cell, one of the parental X chromosomes is randomly inactivated and is then faithfully maintained inactive in the progeny of that embryonic cell. 22 Deoxyribonucleic acid extracted from tumor cells can be analyzed to determine whether all of the cells have the same parental X chromosome inactivated, which would indicate that the tumor evolved from a single transformed progenitor cell (monoclonal neoplasm), or whether there is a mix of cells with both parental X chromosomes inactivated, which would indicate that the tumor evolved from multiple transformed progenitor cells (polyclonal neoplasm). This is accomplished by using restriction fragment length polymorphisms at defined genes to differentiate the maternal and paternal X chromosomes and a methylation-sensitive enzyme to identify which parental X chromosome has been inactivated. 3t
J. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma multiforme Analysis of tumor-specific chromosomal alterations provides another approach to assess the clonal composition of tumors. Some human tumors, of which retinoblastoma is the prototype, result from an inherited or acquired specific mutation at a locus on one copy of a particular chromosome. 25 This is followed by another mutational event, frequently a deletion, on the homologous chromosome. A monoclonal tumor, derived from a single progenitor cell that has sustained a critical somatic deletion, will be populated with cells that all share a deletion at the same locus. In contrast, a polyclonal neoplasm, derived from multiple transformed progenitor cells, will be populated with some cells that have maintained heterozygosity at the informative locus, thereby obscuring the unequivocal demonstration of a somatic deletion. We screened the peripheral blood lymphocytes (PBL's) of 21 female patients with glioblastoma multiforme for polymorphisms at the two X-chromosomelinked genes, hypoxanthine phosphoribosyl-transferase (HPRT) and phosphoglycerate kinase (PGK). Ten tumors (eight primary and two recurrent tumors) from nine patients, who were constitutionally heterozygous at the HPRT or PGK locus, were used for X-chromosome inactivation analysis. The results were verified by examining the tumors for somatic deletions at several loci which span both arms of chromosomes 10 and 17. Materials and Methods
Isolation of DNA Normal human brain tissue was obtained at temporal lobectomy for intractable epilepsy. Glioblastomas multiforme were removed at surgery with meticulous efforts to isolate tumor from surrounding brain. The specimens were immediately frozen in liquid nitrogen and stored at -70~ The tissue diagnoses were confirmed by microscopic pathology according to standard cri.teria3 Genomic DNA was obtained from a human glioblastoma multiforme cell line (SNB 110), PBL, normal human brain tissue, and 10 glioblastomas multiforrne by sodium dodecyl sulfate (SDS)-proteinase K digestion and phenol/chloroform extraction according to established methods] 6 X-Chromosome Inactivation Analysis Analysis for X-chromosome inactivation was performed according to the method described by Vogelstein, et al., 3~with the following modifications. Locus for HPRT. Genomic DNA (10 ug) from a glioblastoma multiforme cell line, PBL's (Case 8), and glioblastoma multiforme tissue (tumors from Cases 8 and 43) was digested to completion with BamHl at 37"C according to conditions recommended by the manufacturer. Half of the samples were then digested at 37"C with 9 U/ug of the methylation-sensitive enzyme HpalI. Digested DNA was size-fractionated through a 0.8% agarose gel and then transferred to Nytran paper in 10 x standard saline citrate (SSC). Filters were baked under vacuum for 2 hours at 80"C. J. Neurosurg. / Volume 77/September, 1992
Locus)or PGK. Genomic DNA (6 ug) from PBL's (Case 20), normal human brain tissue, and glioblastoma multiforme tissue (tumors from Cases 20, 28, 36 first and second recurrence, 37, 39, 63, and 88) was digested with EcoRI and BglI in buffers according to the manufacturer's recommendation, then digested with BglII in 120 ul of 10 mM Tris, 1 mM ethylenediamine tetraacetic acid (EDTA, pH 8.0) with the addition of 25 ul of glycine buffer (100 mM Tris, 100 mM glycine, 100 mM MgCI2, 1 M NaC1, 10 mM 1,4-dithiothreitol, pH 9.5). Samples were ethanol-precipitated and divided into equal aliquots. Digestion with HpaII, gel electrophoresis, and DNA transfer were performed as described above. Somatic Deletion Analysis Genomic DNA (10 izg) from matched PBL and glioblastoma multiforrne tissue (tumors from Cases 8, 20, 28, 36 first and second recurrence, 37, 39, 43, 63, and 88) was digested with the appropriate restriction enzyme: DIOSI (Bg/lI), D10S4 (TaqI), D10S28 (TaqI), and D 17S34 (TaqI) at 37"C. Electrophoretic separation of DNA and transfer to Nytran membrane were performed as described above. Hybridization Insert probes HPRT-800, PGK, D10S1, D10S4, D10S28, and D17S34 were radiolabeled by random priming to a specific activity of more than 108 cpm/ug of DNA. 1~ The filters were prehybridized in a buffer containing 50% formamide, 5 x standard sodium phosphate EDTA, 1 x Denhart's medium, 0.1% SDS, and 5 mg denatured single-stranded salmon sperm DNA. Hybridization of the radiolabeled probes (5 to l0 • 10~ cpm/ml) to the appropriate filters was performed for 24 to 36 hours at 37*C in the same buffer with the addition of 7% dextran sulfate. Filters were then washed once in 2 x SCC and 0.5% SDS at room temperature and three times at 65"C in 0.1 x SSC and 0.5% SDS. Phosphoglycerate kinase blots were washed a fourth time at 68*C in 0.1 x SSC and 0.5% SDS. This stringent washing condition was necessary to eliminate cross-hybridization to a 1.7-kb PGK-related fragmentY The blots were exposed to Kodak XAR 2 film for appropriate intervals at -70~ with intensifying screens. Both methods of clonal analysis (X-chromosome inactivation and somatic deletion) were repeated at least three times for each tumor to minimize the effects of potential variabilities in experimental conditions. Results
X-Chromosome Inactivation Analysis Lymphocyte DNA from 21 female glioblastoma multiforme patients was screened for polymorphisms at the HPRT and PGK loci. Constitutional heterozygositywas detected in two of these patients (Cases 8 and 43) at the HPRT locus and in seven patients (Cases 20, 28, 36, 37, 39, 63, and 88) at the PGK locus. First and second tumor recurrences were obtained from Case 36; prior 4313
R. A. Berkrnan, et al. TABLE 1
Clonal composition of GBM:X-chromosome inactivationand tumor-specific allelicdeletionanalyses* Case No.
X-Chromosome Inactivation
Allelic Deletions: Chromosome:~
HPRT
PGK
10
17
8 20
monoclonal N.I.
N.I. monoclonal
-
+ +
28
N.I.
f
+t
-1
36R N.I. monoclonal + 36RR N.I. monoclonal + 37 N.I. monoclonal 39 N.I. monoclonal + 43 monoclonal N.I. + 63 N.I. monoclonal + 88 N.I. monoclonal + + * N.I. = not informative; GBM = glioblastoma multiforme; HPRT= hypoxanthinephosphoribosyl-transferase;PGK = phosphoglycerate kinase; R = first recurrence; RR = secondrecurrence. - = loss of heterozygosity;+ = maintenanceof heterozygosity. 1"See text for explanation. This represents a summaryof the results from 13 loci examined on chromosomes10 and 17. Only the resultsfrom four of the loci are presented in Fig. 3.
to both recurrences the patient had received irradiation and chemotherapy. The clonal status of these l0 tumors was determined by X-chromosome inactivation analysis at the HPRT and PGK loci. Peripheral blood lymphocytes have often been used as polyclonal tissue controls in previous X-chromosome inactivation experiments. However, DNA samples from PBL's sometimes fail to demonstrate balanced mosaicism.~9'27 This could be due to the clonal nature in which lymphocytes are generated and released into the circulation. Therefore, we have included normal human brain tissue, obtained at temporal lobectomy for intractable epilepsy, along with PBL's from two patients with glioblastoma multiforme, as polyclonal tissue controls. A glioblastoma multiforme cell line served as a control for monoclonal composition.
Locus for HPRT. Genomic DNA from a glioblastoma multiforme cell line, PBL's (Case 8), and two glioblastomas multiforrne (tumors from Cases 8 and 43) was digested with BamHI to reveal the HPRT polymorphic allele and to frame the HpalI methylation sites that are unmethylated in the allele on the inactive X chrom o s o m e ) : Equal aliquots were then treated with the methyl-sensitive enzyme HpaII in order to digest the inactive HPRT allele. Following treatment of PBL DNA with HpalI, a near-equal reduction in intensity of both the 12- and 24-kb HPRT alleles was observed. This is the expected result for polyclonal tissue in which there is a near-equal mix of cells with random inactivation of both paternal and maternal X chromosomes. In comparison, treatment of DNA from a control glioblastoma multiforme cell line and from two glioblastomas multiforme (from Cases 8 and 43) with HpaII resulted in near-complete digestion of only one of the alleles (the 12-kb inactive allele in the control and the 434
FIG. 1. X-chromosome inactivation analysis at the hypoxanthine phosphoribesyl-transferase (HPRT) locus. The DNA from a glioblastoma multiforme cell line (GCL), peripheral blood lymphocytes (PBL), and two glioblastomas multiforme (GBMs) was digested with BamHl to reveal the allele polymorphism and to frame the HpaII methylation sites. Lanes marked with a plus were additionally digested with the methylsensitive enzyme HpaII. The numbers to the left represent the size of the HPRT alleles. Treatment of control PBL DNA with HpaII results in equal digestion of both HPRT alleles, consistent with polyelonal tissue. Similar treatment of DNA from the control GCL and the GBM cases shows an unequal digestion of one of the two alleles, the expected pattern for monoclonal tissue.
24-kb inactive allele in tumors from Cases 8 and 43), suggesting that all of the cells within each tumor sample have the same parental X chromosome inactivated, the expected pattern for monoclonal tissue (Fig. 1).
Locus for PGK. Genomic DNA from PBL's (Case 20), normal human brain tissue, and eight glioblastomas multiforme (tumors from Cases 20, 28, 36 first and second recurrence, 37, 39, 63, and 88) was digested with Bgll to reveal the PGK allele polymorphism, and with BgIII and EcoRI to frame the nine HpaII sites that are unmethylated in the PGK allele on the active X chromosome. Eight of these sites are methylated in the inactive PGK allele; the ninth site is methylated 60% to 80% of the time. 32 Digestion of the PBL and normal human brain DNA with HpaII resulted in nearequal reduction in the intensity of both PGK alleles. Therefore, both tissues contained an equal distribution of cells with either parental X chromosome inactive, as would be expected for polyclonal tissue. Following treatment of the DNA from eight glioblastomas multiforme with HpaII, complete or near-complete digestion of only one of the two alleles in seven of the tumors was observed, consistent with a monoclonal composition of these tumors; these were the 1.7-kb active allele in tumors from Cases 20, 37, and 63, and the 1.3-kb active allele in tumors from Cases 36 (first and second recurrence), 39, and 88 (Fig. 2). Tumor DNA from Case 28 showed only a slight preferential reduction in the intensity of the 1.3-kb allele; the remaining 1.3-kb signal may be due to contamination from normal tissue or clonally distinct subpopulations of tumor cells. Somatic Deletion Analysis Deoxyribonucleic acid from multiforme was also screened lelic deletions on chromosomes both analyses are summarized
all 10 glioblastomas for tumor-specific al10 and 17. Data from in Table 1. Multiple
J. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma rnultiforme
FIG. 2. X-chromosome inactivation analysis at the phosphogylcerate kinase (PGK) locus. The DNA from peripheral blood lymphocytes (PBL's), normal human brain tissue (NHB), and eight glioblastomas multiforme (GBMs) was digested with BglI, BgllI, and EcoRI to reveal the allele polymorphism and to frame the HpalI methylation sites. Lanes marked with a plus were additionally digested with the methyl-sensitive enzyme HpalI. The numbers to the left represent the size of the PGK alleles. Control DNA from NHB and PBL demonstrates a polyclonal pattern following treatment with HpalI, while DNA from seven GBM's reveals a monoclonal pattern. The pattern from the tumor in Case 28 is consistent with contamination from normal tissue (see Fig. 4 right).
polymorphic DNA probes (13 total) on both arms of chromosomes 10 and 17 were used in the analysis. Representative blots of matched sets of lymphocyte and tumor DNA, which were heterozygous at the D10SI, D10S4, D10S28, or D17S34 locus are displayed in Fig. 3. A decrease in the intensity of one of the two alleles occurred for nine of the 10 tumor DNA samples.
Tumor Histopathology Figure 4 shows representative pathological findings in two tumors. Tumor tissue from Case 37, in which the typical morphological cellular heterogeneity of glioblastoma multiforme is apparent is demonstrated in Fig. 4 left. A portion of the tumor from Case 28 is shown in Fig. 4 right; most of the sample consists of normal brain tissue with scattered infiltrating anaplastic astrocytes.
Discussion
Clonal Composition of Human Tumors Early studies of clonality of human tumors used Xchromosome-linked isoenzymes; in particular, electrophoretic variants of glucose-6-phosphate dehydrogenase were employed. Most human tumors analyzed by this technique were reported to be monoclonal, although there were a few exceptions reported as polyclonal, including some parathyroid adenomas, lj neurofibromas, ~2 and trichoepitheliomas. 15 More recently, variable methylation patterns of specific genes of the X chromosome have been used to assess the clonal composition of tumors. 32'33With this technique, many human tumors have been shown to be monoclonal, including colorectal carcinomas, 9 parathyroid adenomas, ~ meningiomas and schwannomas, ~9 and various pituitary adenomas. ~'~7'~8One exception has been the finding of monoclonal and polyclonal corticotroph adenomas of the pituitary, suggesting that mechanisms other than somatic mutation play a role in the genesis of corticotroph adenomas. 27
Glioblastoma Multiforme Clonal Analyses We used X-chromosome-linked inactivation of spe-
J. Neurosurg. / Volume 77/September, 1992
FIG. 3. Somatic deletion analysis showing loss of heterozygosity at loci on chromosomes 10 and I7 in nine cases of glioblastoma multiforme (GBM). The DNA from peripheral blood lymphocytes and GBM's was analyzed with polymorphic markers as described in the text and representative blots presented. Numbers 1 and 2 refer to the larger and smaller alleles, respectively, for the multiallele markers D10S1, D10S4, DIOS28, and D17S34. L = peripheral blood lymphocyte DNA; T = GBM tumor DNA; T~ = first tumor recurrence from Case 36; 1"2 = second tumor recurrence from Case 36. A somatic deletion, as demonstrated by a decrease in the intensity of one of the alleles at the four loci presented, occurred in nine of the 10 tumor DNA samples. cific genes in conjunction with somatic loss of heterozygosity on chromosomes 10 and 17 to assess the clonal composition of glioblastoma multiforme. Nine of the 10 tumors analyzed yielded a monoclonal pattern of X-chromosome inactivation (Figs. I and 2). Both recurrent tumors in Case 36 demonstrated inactivation of the same parental X chromosome and sustained an allelic deletion at the same locus on chromosome 10, consistent with a genetically identical second tumor recurrence. The one tumor that failed to demonstrate clearly a monoclonal composition (Case 28) was contaminated with normal tissue (Fig. 4 right). Uniform loss of DNA sequences on a specific chro435
R. A. Berkman, et al.
FIG. 4. Photomicrographs showing tumor histopathology. H & E, x 30. Left: Tumor from Case 37. Several subtypes of neoplastic astrocvtes are apparent, including small hyperchromic astrocytes, fibrillary astrocytes with dispersed chromatin, and several multinucleated giant astrocytes. Right. Tumor from Case 28 in which normal tissue predominates with some infiltrating anaplastic astrocytes.
mosome in a tumor represents strong evidence for its clonal origin. It is possible, however, for monoclonal tumors to show retention of heterozygosity at specific involved loci. This has recently been reported for meningiomas and schwannomas, which were monoclonal by X-chromosome inactivation analysis but did not consistently have allelic deletions on chromosome 22.19 Considering the heterogeneity of genetic alterations in human solid tumors, and/or the difficulty in detecting small localized deletions, it is not surprising to find monoclonal tumors that do not show clonal chromosomal alterations at involved loci. In human glial tumors, chromosomes 10 and 17 have been reported to be the targets of such mutational events. 8''4'2~ In our study, nine of 10 glioblastomas examined by somatic deletion analysis suffered allele losses on chromosomes 10 and/or 17 (Fig. 3). This high incidence of somatic deletions on chromosomes 10 and 17 suggests a fundamental role for these loci in glioblastoma multiforme tumorigenesis. In spite of the careful selection of tumor tissue for analysis of clonal composition, contamination with normal cells in surgical specimens of tamor sometimes cannot be avoided. The two recurrent tumors from Case 36 and the primary tumors from Cases 28 and 63 probably represent examples of contamination by normal tissue. Our data indicate that somatic deletion analysis is more sensitive to influence by contaminating normal tissue than is X-chromosome inactivation analysis. There are two likely explanations for this observation. First, while 100% of the contaminating normal cells will show maintenance of heterozygosity at the polymorphic loci, only about 50% of these cells will have a particular inactive X-chromosome PGK allele. In fact, this portion is probably less than 50% because of the differential methylation of one of the HpalI sites 436
in the PGK gene. n Second, a particular somatic deletion may be a late event in tumor evolution and, therefore, would not be shared by the entire tumor cell population which, nonetheless, contains the same inactive parental X chromosome. Either or both of these possibilities may explain the apparent discrepancy between the results of the X-chromosome inactivation and somatic deletion analyses. A monoclonal composition for glioblastoma multiforme is consistent with the somatic mutation theory of single-cell carcinogenesis and suggests that the phenotypic and biological heterogeneity of this tumor resuits from the accumulation of genetic alterations by a clonally evolving transformed astrocyte. However, it is possible that distinct polyclonal cells may be present in glioblastoma multiforme but are too sparse to alter the results of clonal analyses significantly. Multifocal gliomas occur sporadically,TM with inherited genetic disorders4"29'3~and in patients previously exposed to ionizing irradiation and intrathecal chemotherapy for other diseases, is'2' If multifocal glioblastomas multiforme arise by a mechanism of carcinogenesis similar to that of solitary glioblastomas multiforme, then their multifocal presentation is likely attributed to the direct extension or intraparenchymal metastasis of clonal cells from a primary focus of monoclonal origin. A true multifocal glioma, on the other hand, requires that multiple cells undergo independent neoplastic transformation events. Gliomatosis cerebri, a rare feature of glioma neoplasms in which malignant astrocytes are spread diffusely throughout the central nervous systemy s may also represent an exception to the singlecell mutation origin of monoclonal glioblastoma multiforme. Alternatively, it too may result from the diffuse spread of clonal cells from a monoclonal site. The clonai status of gliomatosis cerebri and multifocal glioJ. Neurosurg. / Volume 77/September, 1992
Clonal composition of glioblastoma multiforme blastoma multiforme can be determined by analyzing tumor DNA obtained from separate tumor foci. Conclusions
Previous reports of loss of heterozygosity on chromosomes 10 and/or 17 in malignant glial tumors, ~:4'2~ and the results of the present study of X-chromosome inactivation in conjunction with tumor-associated allelic deletions, suggest that solitary glioblastomas are monoclonal tumors that are derived from the clonal evolution of a single transformed cell that has lost critical loci on chromosomes 10 and/or 17. Acknowledgments
We acknowledge the contributions of William Reinhold, B.S., and Barbara lkejiri, B.S., for their preparation of lymphocyte DNA samples. We are also grateful to Conrad Kufla, M.D., for supplying normal brain tissue obtained during seizure surgery. We are indebted to Burt Vogelstein, M.D., for generously supplying the probes HPRT and PGK. References
1. Alexander JM, Biller BMK, Bikkal H, et al: Clinically nonfunctioning pituitary tumors are monoclonal in origin. J Clin Invest 86:336-340, 1990 2. Arnold A, Stauton CE, Kim HG, et al: Monoclonality and abnormal parathyroid hormone genes in parathyroid adenomas. N Engl J Med 318:658-662, 1988 3. Barnard RO, Geddes JF: The incidence of multifocal cerebral gliomas. A histologic study of large hemisphere sections. Cancer 60:1519-1531, 1987 4. Balzdorf U, Malamud N: The problem of multicentric gliomas. J Neurosurg 20:122-136, 1963 5. Burger PC, Green SB: Patient age, histologic features, and length of survival in patients with glioblastoma multiforme. Cancer 59:1617-1625, 1987 6. Burger PC, Vogel FS: Surgical Pathology of the Nervous System and Its Coverings, ed 2. New York: John Wiley & Sons, 1982 7. Couch JR, Weiss SA: Gliomatosis cerebri. Report of four cases and review of the literature. Neurology 24: 504-511, 1974 8. EI-Azouzi M, Chung RY, Farmer GE, et al: Loss of distinct regions on the short arm of chromosome 17 associated with tumorigenesis of human astrocytomas. Proc Natl Acad Sei USA 86:7186-7190, 1989 9. Fearon ER, Hamilton SR, Vogelstein B: Clonal analysis of human colorectal tumors. Science 238:192-196, 1987 10. Feinberg AP, Vogelstein B: A technique for radiolabeling DNA restriction endonuclease fragments to high specific activity. Anal Biochem 137:266-267, 1984 (Addendum) 11. Fialkow PJ, Jackson CE, Block MA, el al: Multicellular origin of parathyroid "adenomas." N Engl J Med 297: 696-698, 1977 12. Fialkow PJ, Sagebiel RW, Gartler SM, et al: Multiple cell origin of hereditary neurofibroma. N Engl J Med 284: 298-300, 1971 13. Fontana M, Stanton C, Pompili A, et al: Late multifocal gliomas in adolescents previously treated for acute lymphoblastic leukemia. Cancer 60:1510-1518, 1987 14. Fults D, Pedone C, Thomas GA, et al: Allelotype of human malignant astrocytoma. Cancer Res 50:5784-5789, 1990 15. Gartler SM, Zipkrewski L, Krakowski A, et al: Glucose6-phosphate dehydrogenase mosaicism as a tracer in the J. Neurosurg. / Volume 77 / September, 1992
study of hereditary multiple trichoepithelioma. Am J Hum Genet 18:282-287, 1966 16. Giangaspero F, Burger PC: Correlations between cytologic composition and biologic behavior in the glioblastoma multiforme. A postmortem study of 50 cases. Cancer 52: 2320-2333, 1983 17. Herman V, Fagin J, Gonsky R, et al: Clonal origin of pituitary adenomas. J Clin Endocrinnl Metab 71:1427-1433, 1990 18. Jacoby LB, Hedley-Whyte ET, Pulaski K, et al: Clonal origin of pituitary adenomas. 3 Neurosurg 73:731-735, 1990 19. Jacoby LB, Pulaski K, Rouleau GA, et al: Clonal analysis of human meningiomas and schwannomas. Cancer Res 511:6783-6786, 1990 20. James CD, Carlbom E, Dumanski JP, et al: Clonal genomic alterations in glioma malignancy stages. Cancer Res 48:5546-5551, 1988 21. James CD, Carlbom E, Nordenskjold M, et al: Mitotic recombination of chromosome 17 in astrocytomas. Proc Natl Acad Sei USA 86:2858-2862, 1989 22. Lyon MF: X-chromosome inactivation and developmental patterns in mammals. Biol Rev 47:1-35, 1972 23. Okazald H, Scheithauer BW: Atlas of Nenropathology. New York: Gower, 1988, p 72 24. Piatt JH Jr, Blue JM, Schold SC Jr, et al: Glioblastoma multiforme after radiotherapy for acromegaly. Neurosurgory 13:85-89, 1983 25. Ponder B: Gone losses in human tumours. Nature 335: 400-402, 1988 26. Sambrook J, Fritsch EF, Maniatis T: Molecular Cloning: A Laboratory Manual, ed 2. Cold Spring Harbor, NY: Cold Spring Harbor Laboratories, 1989 27. Schulte HM, Oldfield EH, Allolio B, et al: Clonal composition of pituitary adenomas in patients with Cushing's disease: determination by X-chromosome inactivation analysis. J Clin Endocrinol Metab 73:1302-1308, 1991 28. Shapiro JR, Yung WKA, Shapiro WR: Isolation, karyotype, and clonal growth of heterogeneous subpopulations of human malignant gliomas. Cancer Res 41:2349-2359, 1981 29. Sherman JL: Neurofibromatosis 1 (Recklinghausen disease) and neurofibromatosis 2 (bilateral acoustic neurofibromatosis). An update. Ann Intern Med 113:39-52, 1990 (see pp 42-43) 30. Todd DW, Christoferson LA, Leech RW, et al: A family affected with intestinal polyposis and gliomas. Ann Neurol 10:390-392, 1981 31. van der Ploeg LHT, Flavell RA: DNA methylation in human GDB-giobin locus in erythroid and nonerythroid tissues. Cell 19:947-958, 1980 32. Vogelstein B, Fearon ER, Hamilton SR, et al: Clonal analysis using recombinant DNA probes from the Xchromosome. Cancer Res 47:4806-4813, 1987 33. Vogelstein B, Fearon ER, Hamilton SR, et al: Use of restriction fragment length polymorphisms to determine the clonal origin of human tumors. Science 227:642-645, 1985 34. Walsh JW, Zimmer SG, Perdue ML: Role of viruses in the induction of primary intracranial tumors. Neurosnrgory 10:643-662, 1982. Manuscript received November 25, 1991. Accepted in final form February 11, 1992. Address reprint requests to: Iqbal U. All, Ph.D., Surgical Neurology Branch, NINDS, Building 10/Room 5D37, National Institutes of Health, 9000 Rockville Pike, Bethesda, Maryland 20892. 437
