Comparison of bromodeoxyuridine labeling indices obtained from tissue sections and flow cytometry of brain tumors

J Neurosurg 68:388-392, 1988 Comparison of bromodeoxyuridine labeling indices obtained from tissue sections and flow cytometry of brain tumors TADASHI NAGASHIMA, M.D., TAKAO HOSHINO, M.D., KYUNG G. CHO, M.D., MORRIS SENEGOR, M.D., FREDERICK WALDMAN, M.D., AND KAZUHIRO NOMURA, M.D. Brain Tumor Research Center and the Department of Laboratory Medicine, School of Medicine, University of California, San Francisco, California," and Department of Neurosurgery, National Cancer Center, Tokyo, Japan Sixteen patients with brain tumors were given a 30- to 60-minute intravenous infusion of bromodeoxyuridine (BUdR), 200 mg/sq m. Grossly viable fragments were taken from the biopsied tumor specimens and divided into two portions. One portion was dissociated into single cells, stained both with fluorescein isothiocyanate (FITC) using anti-budr monoclonal antibody as the first antibody and with propidium iodide (for deoxyribonucleic acid), and analyzed by flow cytometry (FCM). The labeling index (LI) was calculated as the number of FITC-labeled cells expressed as a percentage of the total number of cells analyzed. The other portion was fixed in 70% ethanol, embedded in paraffin, sectioned, and stained with immunoperoxidase using anti-budr monoclonal antibody as the first antibody. The LI of these tissue sections was calculated in two ways: from selected areas in which the labeled cells were evenly distributed and from the entire tissue section. The LI's obtained by FCM correlated closely with those from the entire tissue sections (r = 0.99, p < 0.000001) and were usually lower than LI's from selected areas of tissue sections. The LI's determined by FCM also correlated well with the LI's from selected areas of tissue sections (r = 0.82, p < 0.00012), despite the difference in values between them. Thus, the FCM-derived LI and the tissue LI can both provide useful information for predicting the biological malignancy of individual tumors and for designing treatment regimens for individual patients with brain tumors; however, different standards should be used to interpret the LI's obtained by these two methods. KEY WORDS " brain neoplasm 9 cell kinetics 9 bromodeoxyuridine 9 monoclonal antibody 9 flow cytometry p REVIOUS studies have demonstrated the advantage of using bromodeoxyuridine (BUdR) and anti-budr monoclonal antibody 8 to perform cell kinetics studies on tumors of the central nervous system. 4,~~176 Two methods may be used to obtain the BUdR labeling index (LI), which indicates the fraction of cells undergoing deoxyribonucleic acid (DNA) synthesis: nuclei containing BUdR can be identified in tissue sections by immunoperoxidase staining; ~92 ~ alternatively, BUdR-labeled single cells derived from biopsy specimens can be stained with immunofluorescent antibody and detected by flow cytometry (FCM). 6'22 In the present study, the LI's of various human brain tumors were determined by those two methods and the correlation between the results was analyzed in individual tumors. Materials and Methods Permission to administer BUdR was received from the Human Experimentation Committee at the University of California, San Francisco, and from the National Cancer Institute. Informed consent was obtained from each patient or from a responsible relative. The glial neoplasms were classified according to criteria currently in use at the University of California, San Francisco? Astrocytomas were graded as nonanaplastic or as mildly, moderately, or highly anaplastic based on the degree of cellularity, nuclear and cytoplasmic pleomorphism, and vascular proliferation. Sixteen patients with various brain tumors were entered in this study. All patients received a 30- to 60- minute intravenous infusion of BUdR (200 mg/sq m) 388 J. Neurosurg. / Volume 68/March, 1988

Bromodeoxyuridine labeling indices in brain tumors at the start of surgery. Each excised tumor specimen was divided into two portions. One portion was immediately placed in 70% ethanol, fixed for at least 12 hours, and embedded in paraffin. From the other portion, small tumor pieces that appeared grossly homogeneous and non-necrotic were selected, minced, and placed in an enzyme cocktail consisting of 0.02% collagenase II (139 U/mg), 0.02% deoxyribonuclease I (7 10 dornase U/mg, B grade), and 0.05% pronase (45 PKU/mg, B grade) and stirred with a magnetic stirrer at 37~ for 30 minutes to produce a clean single-cell suspension. The cells were filtered through a 37-urn Nitex mesh filter* and fixed in 70% ethanol for at least 30 minutes. Immunoperoxidase Histochemistry The paraffin-embedded tumor specimens were cut into sections 6 um thick, deparaffinized, and immersed in methanol containing 0.3% H202 for 30 minutes to block endogenous peroxidase activity. After denaturation of DNA with 2 N HC1, the sections were reacted for 45 minutes with a 1:5000 dilution of IU-4,t a monoclonal antibody against iododeoxyuridine that cross-reacts with BUdR. The slides were then immersed for 45 minutes in a 1:50 dilution of peroxidase-conjugated anti-mouse rabbit immunoglobulin G antibody,$ developed with diaminobenzidine tetrahydrochloride and H202 in Tris buffer, and lightly counterstained with Gill No. 1 hematoxylin. The LI was calculated as the number of BUdRlabeled nuclei expressed as a percentage of the total number of nuclei analyzed, excluding the nuclei of vascular components and hematogenous cells. Several viable areas that had a fairly even distribution of labeled cells were selected for analysis. Two hundred to 600 tumor cells from each area and at least 1000 cells (usually approximately 5000 cells) in each specimen were evaluated. The LI was also calculated from the entire tissue section without selecting the well-labeled areas; all of the tumor cells in each section were counted (approximately 15,000 cells), excluding erythrocytes. Immunofluorescence Staining and FCM Analysis Ethanol-fixed single cells were incubated with ribonuclease (1 mg/ml) in phosphate-buffered saline for 15 minutes. Histones were then extracted using 0.1 N HC1 and 0.7% Triton X-100. After denaturation of DNA in water at 90~ for 10 minutes, the cells were incubated with a 1:5000 dilution of IU-4 monoclonal antibodies for 30 minutes and then with a 1:100 dilution of FITC- * Nitex mesh filter manufactured by Tetco, Inc., Elmsford, New York. t Anti-iododeoxyuridine monoclonal antibody 1U-4 supplied by Martin Vanderlaan, Ph.D., Biomedical Science Division, Lawrence Livermore Laboratory, University of California, Livermore, California. Peroxidase-conjugated anti-mouse rabbit immunoglobulin G antibody manufactured by Zymed, South San Francisco, California. - / /... o-/ ~,0/,_v o ~. / Relative DNA Content o o o o Fie. 1. Relative deoxyribonucleic acid (DNA) content (red fluorescence from propidium iodide) and relative bromodeoxyuridine (BUdR) content (green fluorescence from fluorescein isothiocyanate) were measured concomitantly in each cell by flow cytometry and displayed three-dimensionally. Height represents relative numbers of cells. conjugated anti-mouse rabbit immunoglobulin G antibodiesw for 30 minutes. 3~ The cells were then stained for DNA with propidium iodide in phosphate-buffered saline. Approximately 1 x 106 cells were resuspended in 1 to 2 ml of distilled water and filtered through a 37-#m Nitex mesh. Approximately 1 x 105 cells were analyzed with a 5-W argon laser adjusted to emit a 488-nm wavelength beam at 200 mw. Green fluorescence from FITC bound to BUdR-tagged DNA was collected through a 535-nm bandpass filter to obtain a BUdR distribution histogram; red fluorescence from propidium iodide bound to DNA was collected through a 600-nm long wavelength pass filter. The LI was calculated as the number of green fluorescent cells (BUdR-positive cells) expressed as a percentage of the total number of cells gated by appropriate DNA content. The number of green fluorescent cells and total cells were computed from the bivariate DNA/BUdR distribution histograms. Results The LFs obtained by FCM analysis and by analysis of peroxidase-stained tissue sections in 16 patients are shown in Table 1. A bivariate DNA/BUdR distribution histogram of Case 3 is shown in Fig. 1. From the DNA distribution of the histogram, aggregated cells and nonneoplastic components (such as red blood cells and fragments of cells) were excluded by computing the number of cells between the diploid and tetraploid cohorts. In most cases, the FCM-derived LI was considerably lower than the LI of selected areas of tissue section; however, there was a reasonable correlation between these two values (r = 0.82, p < 0.00012). The LI's of entire sections showed very high standard deviations, w F1TC-conjugated anti-mouse rabbit immunoglobulin G antibodies manufactured by DACO, Santa Barbara, California. J. Neurosurg. / Volume 68/March, 1988 389

T. Nagashima, et al. TABLE 1 Labeling indices measured by immunoperoxidase staining ~f tissue sections and by flow cytometry* BUdR Dose Tissue LI (%)t Case No. Age (yrs), Type of Tumor FCM LI (%) Sex (mg/sq m) Entire Section Selected Areas 1 71, M glioblastoma multiforme 200 1.7 1.3 + 1.4 I 1.9 + 3.4 2 40, M glioblastoma multiforme 200 0.2 1.6 + 3.1 8.0 + 1.0 3 40, M glioblastoma multiforme 200 3.2 3.6 _+ 2.8 6.9 + 1.3 4 54, M glioblastoma multiforme 200 3.5 3.3 + 2.0 6.2 + 1.2 5 43, F highly anaplastic astrocytoma 200 4.5 3.0 + 2.6 10.8 _+ 3.4 6 39, F highly anaplastic astrocytoma 200 0.8 0.9 -+ 0.7 2.2 -+ 0.7 7 26, F highly anaplastic astrocytoma 200 0.4 0.2 -+ 0.3 0.7 -+ 0.3 8 23, M moderately anaplastic astrocytoma 200 1.5 0.4 _+ 0.3 0.8 -+ 0.3 9 33, M moderately anaplastic astrocytoma 200 0.6 0.2 _+ 0.7 0.8 -+ 0.3 10 21, F moderately anaplastic astrocytoma 200 1.7 0.04 + 0.2 0.9 + 0.4 11 2, F ependymoma 200 1.0 0.3 _+ 0.4 0.9 _+ 0.4 12 43, E medulloblastoma 200 1.3 0.2 -+ 0.7 4.0 -+ 2.0 13 59, F malignant meningioma 200 1.3 0.8 _+ 0.7 2.3 +_ 0.6 14 55, F meningioma 200 1.5 1.2 + 0.9 2.3 _+ 0.3 15 60, F meningioma 200 0.2 0.2 -+ 0.2 0.4 -+ 0.1 16 66, M metastatic 200 22.9 17.7 + 1.2 20.9 + 2.6 * BUdR = bromodeoxyuridine; FCM = flow cytometry; LI = labeling index. t Mean _+ standard deviation. which implies heterogeneous labeling patterns, but correlated more closely with the FCM LI's (r = 0.99, p < 0.000001) than the LI's computed from selected areas of tissue sections. Discussion Studies of BUdR labeling have provided a considerable amount of information on the growth and biology of various brain tumors. 4'1~ ~3.2o.24 These data supplement the histopathological findings and thereby allow more accurate prediction of prognosis and aid in the design of treatment regimens for individual patients. For more than 20 years, BUdR has been used as a radiosensitizing agent without causing serious side effects.2142629 Although teratogenic and mutagenic effects could be induced by high doses or prolonged administration of BUdR, v-~s the doses required for in vivo labeling studies ( 150 to 200 mg/sq m) were small compared with the doses used therapeutically 1625 and resulted in a low serum BUdR concentration that did not induce mutation experimentally ~'9273~ and did not increase the sister chromatid exchange rate. 32 Thus, BUdR labeling studies can be applied with fewer restrictions than in therapeutic protocols to elucidate the complex proliferation kinetics of human tumors in situ and to better understand their malignant behavior. In calculating the LI from tissue sections, viable areas in which the labeled cells are evenly distributed were usually selected for analysis, as these areas represent the greatest proliferative potential of each tumor. The areas in which labeling is minimal or absent, possibly as a result of necrosis, bionecrosis, or restricted delivery of BUdR, j72~ are excluded. In this study, however, the LI of entire sections was also calculated without consid- eration of the labeling pattern. As expected, this LI value was considerably lower than the LI of selected areas of tissue sections and had a very high standard deviation, indicating very heterogeneous labeling. The single-cell suspension for FCM analysis was obtained from the whole tumor specimen and therefore contained cells from the areas in which the labeling was limited or absent as well as an unpredictable number of normal cells (such as hematogeneous cells and fibroblasts). In most cases, the presence of such cells appeared to make the FCM-derived LI much lower than the LI of selected areas of tissue sections, from which obviously non-neoplastic cells were excluded. Nevertheless, these two values correlated closely. In previous studies, the tissue LI of selected areas was fairly consistent with the histopathological diagnosis and the clinical behavior of individual tumors ~~ ~ 3~20,24 and had a prognostic value in individual patients. 4'~ Because the FCM-derived LI correlates with the LI of selected areas of tissue sections, it can serve as an index to predict the biological and clinical behavior of each tumor. In this study, five of six tumors that had high tissue LI's in selected areas (6.2% to 20.9%) had FCMderived LI's of more than 2%, and all 10 tumors that had tissue LI's of less than 5% in selected areas had LI's of less than 2% based on FCM. Thus, the FCMderived LI may also predict the proliferative potential of individual tumors, but different standards must be used to interpret the FCM and tissue LI's. Bookwalter, et al., 3 investigated the correlation between the patient's survival time and the LI obtained by analysis of single-cell suspensions of 38 glioblastomas multiforme labeled in vitro with tritiated thymidine. The tumors were classified into two groups based on whether the LI's were above or below 5%. Their 390 J. Neurosurg. / Volume 68/March, 1988

Bromodeoxyuridine labeling indices in brain tumors results did not confirm the findings of an early study from our laboratory'" in which LI's obtained by analysis of selected areas of tissue sections of tumors labeled with tritiated thymidine in vivo and in which an LI of 5 % was critical in predicting the prognosis. The discrepancies between these two studies can be explained by the fact that in this current study the LI's obtained from a single-cell suspension of an entire tumor specimen were much lower than those from selected areas of tissue sections. Therefore, it is not surprising that Bookwalter, et al., found LI's below 5% in more than 50% of the glioblastomas in their study, whereas in our previous studies ~~ most such tumors had LI's greater than 5%. We believe that the Ll's measured by different methods cannot be compared directly using the same criteria for interpretation. The conclusion of Bookwalter, et al., that cell kinetics studies are useless for predicting prognosis is, at the very least, premature. The good correlation between survival time and L! obtained from tissue sections by in situ labeling studies has been confirmed in other kinds of solid tumors.~s2~ The results of the current study indicate that the LI obtained by FCM analysis of single-cell suspensions from biopsied tumor specimens may also have a prognostic value in individual patients. The importance of this line of research is to elucidate the significance of LI's obtained by various methods in relation to the actual rate of tumor growth, response to treatment, and patients' survival time so that the benefit of cell kinetics studies are validated for routine clinical practice. Acknowledgments The authors thank Dr. Martin Vanderlaan for supplying the IU-4 monoclonal antibody against BUdR. They are grateful to Frances James for manuscript preparation and Stephen Ordway for editorial assistance. References 1. Aebersold PM: Mutagenic mechanism of 5-bromodeoxyuridine in Chinese hamster cells. Murat Res 36:357-362, 1976 2. Bagshaw MA, Doggett RLS, Smith KC, et al: Intra-arterial 5-bromodeoxyuridine and x-ray therapy. A JR 99: 886-894, 1967 3. Bookwalter JW IIl, Selker RG, Schiffer L, et al: Braintumor cell kinetics correlated with survival. J Neurosurg 65:795-798, 1986 4. Cho KG, Hoshino T, Nagashima T, et al: Prediction of tumor doubling time in recurrent meningiomas. Cell kinetics studies with bromodeoxyuridine labeling. J Neurosurg 65:790-794, 1986 5. Davis RL, Liu HC, Vestnys P, et al: Correlation of survival and diagnosis in supratentorial malignant gliomas. J Neurooncol 2:267, 1984 (Abstract) 6. Dolbeare R, Gratzner H, Pallavicini MG. et al: Flow cytometric measurement of total DNA content and incorporated bromodeoxyuridine. Proc Natl Acad Sci USA 80:5573-5577, 1983 7. Goz B: The effects of incorporation of 5-halogenated deoxyuridines into the DNA of eukaryotic cells. Pharmacol Rev 29:249-272, 1978 8. Gratzner HG: Monoclonal antibody of 5-bromo- and 5- iododeoxyuridine: a new reagent for detection of DNA replication. Science 218:474-475, 1982 9. Heartlein MW, O'Neill JP, Pal BC, et al: The induction of specific-locus mutations and sister-chromatid exchanges by 5-bromo- and 5-chloro-deoxyuridine. Murat Res 92:411-416, 1982 10. Hoshino T, Nagashima T, Cho KG, et al: S-phase fraction of human brain tumors in situ measured by uptake of bromodeoxyuridine. Int J Cancer 38:369-374, 1986 11. Hoshino T, Nagashima T, Murovic J, et al: Cell kinetic studies of in situ human brain tumors with bromodeoxyuridine. Cytometry 6:627-632, 1985 12. Hoshino T, Nagashima T, Murovic JA, et al: In situ cell kinetics studies on human neuroectodermal tumors with bromodeoxyuridine labeling. J Neurosurg 64:453-459, 1986 13. Hoshino T, Nagashima T, Murovic JA, et al: Proliferative potential of human meningiomas of the brain. A cell kinetics study with bromodeoxyuridine. Cancer 58: 1466-1472, 1986 14. Hoshino T, Sano K: Radiosensitization of malignant brain tumors with bromouridine (thymidine analogue). Acta Radiol Ther Phys Biol 8:15-26, 1969 15. Hoshino T, Wilson CB: Cell kinetic analyses of human malignant brain tumors (gliomas). Cancer 44:956-962, 1979 16. Kinsella T J, Russo A, Mitchell JB, et al: A Phase I study of intermittent intravenous bromodeoxyuridine (BUdR) with conventional fractionated irradiation. Int J Radiat Oncol Biol Phys 10:69-76, 1984 17. Molnar P. Groothuis D, Blasberg R, et al: Regional thymidine transport and incorporation in experimental brain and subcutaneous tumors. J Neurochem 43:421-432, 1984 18. Moran RE, Straus M J: Labeling indices of human lung cancer. Correlation with histologic type and survival. Anal Qnant Cytol 5:250-254, 1983 19. Morstyn G, Hsu SM, Kinsella T, et al: Bromodeoxyuridine in tumors and chromosomes detected with a monoclonal antibody. J Clin Invest 72:1844-1850, 1983 20. Murovic J, Nagashima T, Hoshino T, et al: Pediatric central nervous system tumors: a cell kinetic study with bromodeoxyuridine. Neurosurgery 19:900-904, 1986 21. Nagashima T, DeArmond SJ, Murovic J, et al: Immunocytochemical demonstration of S-phase cells by antibromodeoxyuridine monoclonal antibody in human brain tumor tissue. Aeta Neuropathol 67:155-159, 1985 22. Nagashima T, Hoshino T: Rapid detection of S-phase cells by anti-bromodeoxyuridine monoclonal antibody in 9L brain tumor cells in vitro and in situ. Acta Neuropathol 66:12-17, 1985 23. Nagashima T, Hoshino T, Cho KG: Proliferative potential of vascular components in human glioblastoma multiforme. Acta Neuropathol 73:301-305, 1987 24. Nagashima T, Murovic JA, Hoshino T, et al: The proliferative potential of human pituitary tumors in situ. J Neurosurg 64:588-593, 1986 25. Russo A. Gianni L, Kinsella T J, et al: Pharmacological evaluation of intravenous delivery of 5-bromodeoxyuridine to patients with brain tumors. Cancer Res 44: 1702-1705, 1984 26. Sano K, Hoshino T, Nagai M: Radiosensitization of brain tumor cells with a thymidine analogue (bromouridine). J Neurosurg 28:530-538, 1968,L Neurosurg. / Volume 68/March, 1988 39"1

T. Nagashima, et al. 27. San Sebastian JR, O'Neill JP, Hsie AW: Induction of chromosome aberrations, sister chromatid exchanges, and specific locus mutations in Chinese hamster ovary cells by 5-bromodeoxyuridine. Cytogenet Cell Genet 28: 47-54, 1980 28. Straus M J, Moran RE: The cell kinetics of human breast cancer. Cancer 46:2634-2639, 1980 29. Szybalski W: X-ray sensitization by halopyrimidines. Cancer Chemother Rep 58:539-557, 1974 30. Tsuji H, Shiomi T, Tsuji S, et al: Aphidicolin-resistant mutants of mouse lymphoma L5178Y cells with a high incidence of spontaneous sister chromatid exchanges. Genetics 113:433-447, 1986 31. Vanderlaan M, Watkins B, Thomas C, et al: Improved high-affinity monoclonal antibody to iododeoxyuridine. Cytometry 7:499-507, 1986 32. Zhang RX, Nagashima T, Hoshino T: Cytotoxic effect and induction of sister chromatid exchange in exponential growing rat 9L gliosarcoma cells after brief exposure to bromodeoxyuridine. Cell Tissue Kinet 20:357-362, 1987 Manuscript received June 17, 1987. This work was supported in part by Grant PDT-159 from the American Cancer Society, Grant CA-13525 from the National Cancer Institute in Japan, and a Grant-in-Aid from the Ministry of Health and Welfare for Comprehensive 10- Year Strategy for Cancer Control. Address reprint requests to: Takao Hoshino, M.D., c/o The Editorial Office, Department of Neurological Surgery, 1360 Ninth Avenue, Suite 210, San Francisco, California 94122. 392 J. Neurosurg. / Volume 68/March, 1988