The distribution of nuclear DNA from human brain-tumor cells

J Neurosurg 49:13-21, 1978 The distribution of nuclear DNA from human brain-tumor cells Flow cytometrie studies TAKAO HOSHINO, M.D., D.M.Sc., KAZUHIRO NOMURA, M.D., D.M.Sc., CHARLES B. WILSON, M.D., KATHY D. KNEBEL, B.S., AND JOE W. GRAY, PH.D. Brain Tumor Research Center, Department of Neurological Surgery, University of Ca!ifornia School of Medicine, San Francisco and Livermore, California Flow cytometry (FCM) is a technique that measures the quantity of DNA contained in individual nuclei and records a frequency distribution of the DNA content per nucleus in the sampled cell population. Nuclei from a variety of human braintumor types were isolated by means of tissue grinding, purified by centrifugation through 40% sucrose (15 minutes at 4000 rpm), fixed with 10% formalin, stained with acriflavin-feulgen, and analyzed by FCM. Profiles of DNA distribution in histologically benign tumors, such as meningiomas, pituitary adenomas, neuroblastomas, and low-grade astrocytomas, revealed a large diploid population (2C) with a few nuclei in DNA synthesis, as well as a small premitotic population (G2 cells) that contains a 4C DNA complement. In contrast, malignant gliomas, including glioblastomas, consist of more cells in DNA synthesis; these tumor cells show a highly variable distribution of ploidy consisting not only of diploid, and/or aneuploid, but also of triploid, tetraploid, and possibly octaploid populations. Also, a large variability between different regions of each tumor was always observed. In contrast, metastatic brain tumors, despite the fact that they contain a considerable number of cells undergoing DNA synthesis, demonstrate little variability within each individual tumor. The ability to rapidly characterize the cell populations of human brain tumors with FCM may enhance the effectiveness of their clinical management. KEY WORDS 9 flow cytometer 9 brain tumor 9 glioma 9 DNA measurement 9 nuclei isolation 9 karyotype P'll'~HE genetic information of a cell is I carried in its nuclear DNA, which controls its cellular functions through the production of mrna. The amount of DNA in the nucleus of a non-dividing cell is usually constant throughout a cell's lifespan. In most dividing cells the DNA content immediately after division represents the diploid chromosomal complement (2C); this increases during the DNA synthetic phase to double that amount (4C), and then returns to the diploid state again after mitosis. However, some nondividing cells have been found to contain a 4C, or tetraploid, amount of DNA in both normal and abnormal tissues. For example, by cytophotometric analysis, J. Neurosurg. / Volume 49 / July, 1978 13

T. Hoshino, et al. FIG. 1. A flow cytometry profile of 9L rat brain-tumor cells in vitro. The relative fluorescence intensity values of 1.0 and 2.0 correspond to the position of nuclei with a 2C and 4C DNA content, respectively. For details refer to the text. S = DNA synthesizing cells. the liver contains many tetraploid cells, and some cells in the central nervous system also have a 4C DNA complement),l~-x4,18,22'2s By the cytophotometric method, one can determine the DNA content of each cell type in a particular tissue, ~5,16 but the method is time-consuming and the results are somewhat unsatisfactory in terms of resolution and statistical validity. Usually the coefficient of variation in each peak is so great that slight differences in amounts of DNA cannot be quantitated. Recent developments in biophysics have resulted in a more efficient and accurate method of analyzing the amount of DNA in individual cells, namely by flow cytometer (FCM) analysis 24,25 (also called flow microfluorometer (FMF) analysis). The FCM technique measures the amount of DNA per cell by quantitating the intensity of the fluorescence emitted by a DNA-bound dye as nuclei flow rapidly past a high intensity laser beam. From these data, a histogram can be constructed showing the frequency with which these DNA contents appear in the cell population. The histogram (number of cells versus intensity of fluorescence) has a single narrow peak when the entire population consists of diploid cells. Multiple peaks are found at positions relative to the intensity of their fluorescence when cells with, for example, tetraploid (4C), or octaploid (8C) DNA contents exist in the population. If there are cells synthesizing DNA before cell division, their relative fluorescence values fall between the 2C and 4C peaks. Figure 1 illustrates a profile obtained from 9L rat brain-tumor cells growing exponentially in vitro. Under such conditions, all the cells analyzed were in the proliferating pool. The first peak from the left represents cells containing a 2C DNA complement; these are cells in the G1 (postmitotic, pre-dna synthetic) phase. The second peak represents a population of 4C DNA cells; these are cells in the post-dna synthetic and/or mitotic phases. Cells that register either between these peaks, or partially overlap with these peaks, are considered to be cells in the process of DNA synthesis (shaded portion, S). If many cells become nonproliferating (non-cycling) cells (Go cells), which are also in the category of 2C DNA cells, the proportion of cells in the first peak increases relative to the rest of the population. The FCM technique does not differentiate between Go and G1 cells, since both contain a 2C amount of DNA. This paper describes the preliminary results of an FCM analysis of human brain-tumor nuclei isolated from surgical specimens. Materials and Methods Fresh pieces of brain tumor were obtained at the time of surgery, and selected portions (two to six pieces) from these specimens were processed as follows: A representative piece of tumor was removed for histological study, and a 50- to 100-mg sample of tissue was homogenized in a Dounce tissue grinder (pestle b) in 4 ml of ice-cold PIPES buffer (1.0 mm CaCI2, 0.05 mm piperizine-n,n'-bis(2- ethanesulfonic acid) and 0.5 M 2 methyl-2,4- pentanediol at ph 6.5), and then filtered through a layer of 20 ~tm Nitex.* The homogenate was centrifuged at 4 ~ C for 5 minutes at 221 G in a refrigerated IEC-PR6 centrifuge, the supernatant was removed, and the pellet was resuspended in 2 ml of ice-cold PIPES buffer. This solution was then layered with a pipette on top of 30 ml of 40% sucrose containing 1.5 mm CaCI2 in a 50-ml conical centrifuge tube and, finally, recentrifuged for 15 minutes at 3536 G. The supernatant was removed with a pipette, the nuclei were *PIPES buffer available from Calbiochem, La Jolla, California. Nitex manufactured by Tetko Inc., Elmsford, New York. ]4 J. Neurosurg. / Volume 49 / July, 1978

Nuclear DNA in brain tumor cells resuspended in 5 ml of saline G (8 gm NaC1, 0.2 gm KC1, 1.15 gm Na2HPO4, 0.2 KH~PO4 in 1000 ml distilled water (DW)) and then fixed with 5 ml of 20% formalin. These sam- a ples could then be stored at 4 ~ C for several weeks before they were stained. To stain the,,,. 2 isolated nuclei, the DNA was hydrolyzed for 20 minutes in 4N HCI at room temperature and then exposed to acriflavin (0.02 gm ~, 1 acriflavin, 0.5 gm K2S~O5 in 10 ml of 0.5 N HCI and 90 ml of DW) in the dark at room ~ o temperature for 20 minutes? ~ The stained ua specimens were analyzed on the Lawrence > _ Livermore Laboratory flow cytometer.t To,~ 4 check the presence of multiplet formation, each peak population was sorted on several a occasions and examined with fluorescein microscopy. 4 -- (A) 1.0 2.0 9,_ L m_ (B) Results We have analyzed over 50 brain tumors and obtained results from 18 malignant gliomas (including glioblastoma multiforme and anaplastic astrocytoma), 10 well differentiated gliomas (astrocytoma, oligodendroglioma, and ependymoma), eight meningiomas, seven metastatic brain tumors, and others (including pituitary adenoma, neuroblastoma, choroid plexus papilloma, and medulloblastoma). Benign Brain Tumors Figure 2 illustrates two profiles obtained from two different pieces of a single meningioma (meningotheliomatous). The resulting profiles from each piece were nearly identical in appearance. There were two peaks, a large 2C peak and a small 4C peak, and few nuclei contained an intermediate DNA complement indicative of cells undergoing DNA synthesis. These profiles suggest that this meningioma has very few proliferating cells or a very long cell cycle time composed primarily of a long G1 phase (pre-dna synthetic phase). In addition, the profiles indicate that this tumor is relatively homogeneous with respect to its cell population. We have analyzed eight meningiomas to date, and profiles obtained from them are basically tflow cytometer, prototype of the FACS-II, manufactured by Becton-Dickinson, Mountain View, California. 1.0 2.0 FIG. 2. Flow cytometry analysis of nuclei isolated from a human meningioma. A and B are profiles obtained from two different regions of the same tumor. The relative fluorescence intensity value of 1.0 and 2.0 have the same meaning as in Fig. 1. similar to this profile. These findings correspond well to the benign and slow-growing nature of this tumor. When a case of pituitary adenoma was studied in a similar manner, the basic features of profiles obtained were almost identical to those of meningiomas: namely, a large G1 or Go population with very few nuclei synthesizing DNA. Again, two different pieces from the same tumor yielded identical profiles, regardless of depth of biopsy site. Since normal diploid cells, such as lymphocytes, macrophages, and cells from vessel walls are included in tumor pieces, a profile should present double or multiple peaks if the amount of nuclear DNA in those tumors differs from that in the nuclei of normal cells. Therefore, the presence of only one peak indicates that these tumor cells possess an essentially normal nuclear DNA complement. The majority of low-grade astrocytoma cases J. Neurosurg. / Volume 49 / July, 1978 15

f T. Hoshino, et al. n~ z.d u.l U m I-. 4 3 2 I 4 3 2 I- (A) 1.'0 2'.O (c) i 1.0 2.0 i (B) (D) FIG. 3. Flow cytometry profiles obtained from a fibrillary astrocytoma. Note the double peak at 1.0 position in profiles A and C. Details are given in the text. showed similar profiles, so their biological homogeneity and minimal proliferative rates are obvious. Figure 3 illustrates a fibrillary astrocytoma. General findings were similar, except that profiles A and C demonstrated a double peak that corresponded to the arbitrary 1.0 position of relative fluorescence intensity value. Microscopic examination of the tissue counterpart of specimens A and C revealed that these pieces contained normal brain tissue, whereas B and D contained only tumor tissue. Therefore, if nuclei in the right-hand side of the double peak from A and C represented normal diploid nuclei derived from the normal brain tissue, nuclei in this tumor contained a hypodiploid amount of DNA in the G1 or Go state. Malignant Gliomas In marked contrast, when four pieces of a glioblastoma were analyzed, very different profiles were obtained (Fig. 4). The FCM profiles revealed three main peaks: an initial sharp peak at the 2C (diploid) position, and broad peaks at both the 4C and 8C positions. Since there was a reasonable number of nuclei with DNA contents between 2C and 4C and between 4C and 8C, it is unlikely that these latter peaks are due simply to groups of 2 and 4 nuclei sticking together. This was confirmed by an examination of the sorted nuclei located in each 4C and 8C peak, which revealed that doublet and quadruplet formation in each peak constituted less than 5% of the peak population. In all probability these peaks represent the fractions of the population that contain a tetraploid or octaploid amount of nuclear DNA. The ratio of these peaks varied considerably among the four samples analyzed, which indicates that this glioblastoma was much more heterogeneous with respect to its population than was the meningioma or adenoma. Not all glioblastomas examined contained such prominent tetraploid and octaploid populations, but the variability was always observed when different regions of each glioblastoma were compared. Figure 5 is another example of a glioblastoma, in which the profiles also showed three major peaks with heights varying in different pieces from the same patient. In this case, the peak positions were located at 2C, 3C (triploid amount of DNA), and 6C, if we assume the initial peak represents a 2C population. It is quite possible that the 6C peak represents a G2 population of 3C cells. When four regions of the same anaplastic astrocytoma were analyzed (Fig. 6), two samples (Fig. 6 B and D) had profiles similar to those obtained from well differentiated astrocytomas; this indicates a small proliferating population, because few cells are seen between the 2C and 4C peaks. In contrast, the other two specimens (Fig. 6 A and C) indicated a larger proliferating population, as well as the existence of a 4C population suggested by an abnormally high second peak for G2 population alone. A double peak seen in Fig. 6 C suggests a considerable aneuploid population m this piece of tumor. Histological examination of the tumor analyzed in Fig. 6 revealed that sections corresponding to profiles B and D consisted primarily of fibrillary astrocytes with moderate pleomorphism, microcysts, and gemistocytes, but no mitoses. Sections corresponding to Fig. 6 A and C contained pleomorphic astrocytes with many giant cells and mitoses. 16 J. Neurosurg. / Volume 49 / July, 1978

84 Nuclear DNA in brain tumor cells 5 (A) (e) m D I J 1.0 2.0 4.0 c 5 - (D) 4-3- 2 t 1.02.0 4.0 0 I,r... - - 1.02.0 4.0 FIG. 4. Flow cytometry analysis of nuclei isolated from a glioblastoma multiforme; profiles obtained from four regions of the same tumor. The relative fluorescence intensity values of 1.0, 2.0, and 4.0 correspond to the position of nuclei with a 2C, 4C, and 8C DNA content, respectively. Metastatic Brain Tumors Seven metastatic brain tumors studied demonstrated similar characteristics. There was little variability within each tumor, although a number of cells were in DNA synthesis. This observation indicates that these tumors consist of a relatively homogeneous, actively proliferating cell population. Using the FCM profiles obtained for 50 brain tumors, we estimated the percentage of cells in S phase in each tumor. Most well differentiated gliomas and meningiomas contained very few cells undergoing DNA synthesis, as shown in the representative profiles. However, malignant gliomas and metastatic brain tumors contained a significant number of cells synthesizing DNA. The position of each peak relative to the first peak was plotted in Fig. 7. If a tumor cell population consists of a single karyotype, it usually shows one large peak that contains Go and G1 cells of the population, with a small second peak representing the G2 population. This was true for most well differentiated gliomas. Peaks in the profiles from malignant gliomas showed a broad distribution as well as a variable number of peaks. The multiple peaks in the profiles from malignant gliomas indicate that each of these tumors consisted of several different types of cells that had unusual DNA contents. In addition, the variable distribution of peaks in these profiles suggests that malignant gliomas may vary in the nature of their polyploid or hyperdiploid cell population from one tumor to the other. We J. Neurosurg. / Volume 49 / July, 1978 17

w t~ 4 Y 1.0 1.5 3.0 (A) (c) ~ B) I i i 1.'o 11s i.o (D) FIG. 5. Four profiles obtained from another glioblastoma multiforme. Relative fluorescence intensities of 1.0, 1.5, and 3.0 indicate 2C, 3C, and 6C DNA populations, assuming the first peak population from the left represents diploid cells. do not know whether all of the nuclei in each peak were derived from clonogenic tumor components. One of these peaks may represent only diploid populations of normal cells intermingled within the tumor, as was the case for the profile of several transplantable experimental tumors. Meningiomas and metastatic brain tumors gave a second peak which was in a position similar to that for 311 A 3FI B 10 20,I<2 ~ (c) Io 20 (D) I0 20 FIG. 6. Flow cytometry analysis of nuclei isolated from an anaplastic astrocytoma; profiles from different regions of the same tumor. Details are given in the text. T. Hoshino, et al. benign gliomas; only a few examples showed a third or fourth peak also. It appears that most of these tumors may contain a homogeneous nuclear DNA complement that is close to the amount found in the nuclei of normal cells, even though chromosomal aberrations have been revealed by karyotype analysis of these tumors? ~ Discussion The FCM and cell-sorter system we have used routinely gives a coefficient of variation (CV) of 1.5% or less in CHO 2C DNA peak. Capable of analyzing 5000 cells per second, this system permits the detection of various cell populations with different amounts of DNA more rapidly and accurately 4 than any other existing method, such as microspectrophotometry, karyotype analysis of cells in metaphase, or biochemical measurement. However, we needed to overcome one major obstacle to make the FCM technique a useful analytical tool. Suspensions of either single cells or reasonably intact isolated nuclei are required for FCM analysis. The usual mincing procedure with or without trypsinization was not very successful in dissociating tissue with many entangled fibrils, such as long axons, dendrites, and astrocytic processes. In an effort to overcome this difficulty we modified a nuclei isolation procedure reported by Kato and Kurokawa, 9 and prepared suspensions of nuclei from various parts of brain tissue and tumors in rat, dog, and man. 7 Recovery of intact nuclei with this tissue grinding method is more than 70% 9 and we found no selective destruction correlated with any specific nuclear DNA content (paper in preparation). However, it should be pointed out that the DNA from mitotic dividing cells could not be recovered because the nuclear membrane disappears. Microscopic observation indicated that nuclei in the homogenate immediately after tissue grinding were singlets. Nevertheless, formation of doublets or triplets by the end of the whole procedure occurred, as shown in most of the profiles analyzed. However, the frequency of such multiplet formation was low (0.25%-0.5% of total nuclei analyzed). We made several efforts to separate these multiplet nuclei before FCM analysis by sonicating them after fixation or by homogenation and centrifugation without Ca ++ in the 18 J. Neurosurg. / Volume 49 / July, 1978

Nuclear DNA in brain tumor cells presence of detergent and EDTA, but none of these attempts was successful. Although solving this problem might be critical if this method is to be used for cell kinetic studies, it does not introduce any serious error into our current study. Another problem to be solved is elimination of the background, which consisted of cytoplasmic DNA, debris, and nuclei of dead cells. Background was minimal when the specimen contained no necrotic foci, as seen in the profiles obtained from tumors such as meningioma and pituitary adenoma. Since dead cells degrade and release their nuclear DNA fairly rapidly, the higher the proportion of necrotic area, as seen in most glioblastomas, the more prominent the background. There are two ways to solve this problem: 1) to remove necrotic cells from the specimen before the analysis; and 2) to develop mathematical models to adjust the profile obtained. Both methods are currently under investigation, and some of these mathematical models can be applied to an experimental tumor system. However, no definitive method has been developed for analysis of solid tumors in humans. Elimination of this artifact is an important factor in estimating the number of cells in DNA synthesis. The staining method itself may account for apparent differences in DNA content. Other authors s,l~ have confirmed the DNA-specific dye binding ability of the acriflavin-feulgen method, and we will not present the arguments in this paper. According to our observations, benign brain tumors, such as meningiomas, pituitary adenomas, and well differentiated gliomas, all of which grow slowly, have common characteristics: 1) there are very few cells in DNA synthesis; 2) a majority of the tumor is composed of a single karyotype (if the term "karyotype" is used here to denote cells that have identical amounts of DNA in their nuclei rather than the same number of chromosomes); and 3) there is little variability within a tumor. In contrast to these findings, malignant gliomas, including glioblastoma multiforme, exhibited the following characteristics: 1) there was a substantial population of cells in DNA synthesis; 2) there was a wide distribution of nuclear DNA complement, ranging from 2C to 8C (assuming that the initial peak represented ;o _- r. ~ -~ T: = Malignant Gliomas z- _-- -- * z -" (18 cases); e - -- ='--- x l z _ 2 = ] 9 2 Benign Gliomas ~ x (10 cases) == x -- ~_- x x - - I- Meningiomas I ~. (8 cases) ~ = Medulloblastomas I : : (2 cases) i Metastatic Brain : ~ _- Turnors(7 cases)~ : -2 : Pituitary Adenoma I (1 case) I Neurinoma = (1 case) I Neuroblastoma (1 case) l 2 1 2 :3 4 Relative Peak Position FIG. 7. Comparison of relative peak positions on FCM profiles of human brain tumors. Ratio of the distance of subsequent peaks to the first peak from the ordinate is indicated on the horizontal line. It can be seen that DNA from 15 out of 18 of the malignant gliomas was distributed in three or more peaks. This heterogeneity in DNA content is characteristic of the more malignant tumor types. diploid nuclei); and 3) different regions of the same tumor varied greatly in the distribution of the predominant ploidy; for example, in some areas the nuclear population seemed to be primarily diploid, while in other regions the nuclei were mainly tetraploid (4C).:]: Thus, malignant gliomas are heterogeneous not only in the sense of ploidy, but 1:Some authors use the term "multiploidy" not to denote a multiploid number of chromosomes as the term was originally used, but in respect to the amount of DNA. There is evidence that multiploidy in karyotype does not involve proportionate amounts of DNA; for instance, 9L rat brain tumors contained a population with a tetraploid amount of DNA, even though the number of chromosomes is 62 (triploid) compared to the normal rat chromosome number of 42 (unpublished data). J. Neurosurg. / Volume 49 / July, 1978 19

T. Hoshino, et al. also in respect to the populations within different areas of the same tumor. Heterogeneity and a higher number of cells undergoing DNA synthesis in malignant gliomas have been shown by several autoradiographic studies of human brain tumors, e,a Also, chromosomal aberrations and the existence of multiploid populations in a glial tumor series have been well documented by Bicknell, 2 Lubs and Salmon, 17 Mark, TM and Wilson, et al. 28 The most extensive survey of karyotypes in a malignant glioma series was carried out by Mark, 19 who found that only three out of 50 gliomas were characterized by the normal chromosome number of 46, while 15 had hypodiploid, and 32 had hyperdiploid, triploid, or tetraploid populations. He also found that 20 out of 50 malignant gliomas consisted of more than one karyotype. Flow cytometry profiles on prostatic tumors indicated the existence of multiple or multiploid populations, in terms of the amount of nuclear DNA, 1 although one of these populations might represent normal diploid cells included in the tumor, such as leucocytes, macrophages, and/or cells derived from vessel walls. Metastatic brain tumors demonstrated little variability within each tumor, although a number of cells were in DNA synthesis. This finding indicates that these tumors contain a relatively homogeneous tumor cell population with a significant proliferating component, and suggests that one metastatic lesion may arise from a single clonogenic cell. Cell kinetic analysis of human brain tumors, based on alterations in the pattern of DNA distribution, has not been established because cell dispersion techniques are not yet adequate. Some specimens contain considerable debris and show too large a coefficient of variation in each peak. Our sample of human brain tumors is limited, and we cannot offer an unequivocal interpretation of our data. The information obtained by FCM analysis may provide clues to the degree of malignancy, and indicate probable responses to any clinical management. For example, if a tumor consisted of a heterogeneous cell population, each cell type may have unique responses to various chemotherapeutic drugs. Therefore, a combination of several drugs may be as necessary as the use of multiple antibiotics to treat mixed bacterial infections. Also, the amount of DNA contained in tumor nuclei may correlate with the effectiveness of certain drugs. Of particular importance is the possibility that because FCM analysis is rapid, a patient's postsurgical treatment might be based in a more sophisticated and precise manner on information obtained from his own tumor. To date, conventional light and electron microscopic analysis of tumors has contributed limited information relative to the selection of specific therapies. Acknowledgments The authors thank Kenneth T. Wheeler, Ph.D., for invaluable advice in developing the nuclei isolation technique, Barbara Riddle, Ph.D., for her editorial assistance, and Benny Usog for the preparation of graphics. References 1. Bichel P, Frederiksen P, Kjaer T, et al: Flow microfluorometry and transrectal fine needle biopsy in the classification of human prostatic carcinoma. Cancer 40:1206-1211, 1977 2. Bicknell JM: Chromosome studies of human brain tumors. Neurology 17:485-490, 1967 3. Crissman HA: Cell preparation and staining for fow system in Richmond CR, Petersen DF, Mullaney PF, et al (eds): Mammalian Cells: Probes and Problems. Oakridge, Tenn: Energy Research Development Agency, Symposium Series 731007, 1975, pp 94-106 4. Gray JW, Carrano AV, Moore DH II, et al: High-speed quantitative karyotyping by flow microfluorometry. Ciin Chem 21:1258-1262, 1975 5. Herman C J, Lapham LW: DNA content of neurons in the cat hippocampus. Science 160:537, 1968 6. Hoshino T, Barker M, Wilson CB, et al: Cell kinetics of human gliomas. J Neurosurg 37:15-26, 1972 7. Hoshino T, Nomura K, Gray JW, et al: Flow microfluorometric studies of brain and brain tumors. J Neuropathoi Exp Neurul 35:363, 1976 (Abstract 128) 8. Hoshino T, Wilson CB, Rosenblum ML, et al: Chemotherapeutic implications of growth fraction and cell cycle time in glioblastomas. J Neurosurg 43:127-135, 1975 9. Kato T, Kurokawa M: Isolation of cell nuclei from the mammalian cerebral cortex and their assortment on a morphological basis. J Cell Biol 32:649-662, 1967 10. Kraemer PM, Deaven LL, Crissman HA, et al: DNA constancy despite variability in chromosome number, in DuPraw EF (ed): Ad- 20 J. Neurosurg. / Volume 49 / July, 1978

Nuclear DNA in brain tumor cells vances in Cell and Molecular Biology. New York: Academic Press, 1972, Vol. 2, pp 47-108 11. Lapham LW: The tetraploid DNA content of normal human Purkinje cells and its development during the perinatal period. A quantitative cytochemical study. International Congress of Nenropathology. Amsterdam: Excerpta Medica, International Congress Series No. 100, 1965, pp 445-449 12. Lapham LW: Tetraploid DNA content of Purkinje neurons of human cerebellar cortex. Science 159:310-312, 1968 13. Lapham LW, Johnstone MA: Cytologic and cytochemical studies of neuroglia. II. The occurrence of two DNA classes among glial nuclei in the Purkinje cell layer of normal adult human cerebellar cortex. Arch Neurol 9:194-202, 1963 14. Lapham LW, Johnstone MA: Cytologic and cytochemical studies of neuroglia. III. The DNA content of giant fibrous astrocytes, with implications concerning the nature of these cells. J Neuropathol Exp Neurol 23:419-430, 1964 15. Leuchtenberger C: Quantitative determination of DNA in cells by Feulgen microspectrophotometry, in Danielli JF (ed): General Cytochemical Methods. New York: Academic Press, 1958, Vol. 1, pp 219-278 16. Leuchtenberger C, Leuchtenberger R, Davis AM: A microspectrophotometric study of the desoxyribose nucleic acid (DNA) content in cells of normal and malignant human tissues. Am J Pathol 30:65-85, 1954 17. Lubs HA Jr, Salmon JH: The chromosomal complement of human solid tumors. II. Karyotypes of glial tumors. J Neurosurg 22:160-168, 1965 18. Mann DMA, Yates PO: Polyploidy in the human nervous system. Part 2. Studies of the glial cell populations of the Purkinje cell layer of the human cerebellum. J Neurol Sci 18:197-205, 1973 19. Mark J" Chromosomal characteristics of neurogenic tumours in adults. Hereditas 68:61-100, 1971 20. Mark J: Chromosomal patterns in human meningiomas. Eur J Cancer 6:489-498, 1970 21. Mark J: Two benign intracranial human tumours with an abnormal chromosomal picture. Acta Neuropathol 14:174-184, 1969 22. Sandritter W, Nov~ikov~i V, Pilny J, et al: Cytophotometrische Messungen des Nukleinsiiure und Proteingehaltes von Ganglienzellen der Ratte w~ihrend der postnata!en Entwicklung und im Alter. Z Zellforseh 80:145-152, 1967 23. Swartz F J: The development in the human liver of multiple desoxyribose nucleic acid (DNA) classes and their relationship to the age of the individual. Chromosoma 8:53-72, 1956 24. Van Dilla MA, Steinmetz LL, David DT, et al: High-speed cell analysis and sorting with flow systems: biological applications and new approaches. Nucl Sci 21:714-720, 1974 25. Van Dilla MA, Trujillo TT, Mullaney PE, et al: Cell microfluorometry: a method for rapid fluorescence measurement. Science 163: 1213-1214, 1969 26. Wilson CB, Kaufmann L, Barker M: Chromosome analysis of glioblastoma multiforme. Neurology 20:821-828, 1970 This work was supported in part by NIH Grants CA-13525 and CA-19992 from the National Cancer Institute, and the Association for Brain Tumor Research. Address for Dr. Gray: Lawrence Livermore Laboratory, University of California, Livermore, California. Address reprint requests to: Takao Hoshino, M.D., Department of Neurological Surgery, University of California, San Francisco, California 94143. J. Neurosurg. / Volume 49 / July, 1978 21