Binding of Acridine Orange to DNA in Situ of Cells from Patients with Acute Leukemia1

(CANCER RESEARCH 49. .1692-3695. July 1. 1989] Binding of Acridine Orange to DNA in Situ of Cells from Patients with Acute Leukemia1

Alexander J. Walle2 and George Y. Wong Cornell Ã•'nirersily Medical (allege /A. J. H'./and Memorial Sloan-Kettering Cancer Ccaler Â¡G.Y. W.], New York, New York 10021

ABSTRACT cytes bound AO in a manner significantly different from that of human leukemic blood lymphoblasts [L3 variety of the Fluorescence flow cytometry was used to generate DNA histograms of French-American-British classification (6)]. We quantitatively at nilim- orange stained leukemic cell populations in G(I-G| phase of the described the interaction of AO with DNA of leukemic cell cell cycle. Complexes of the intercalating agent, acridine orange, with populations in G0-G| phase of the cell cycle by measuring the double-stranded DNA in situ, emit green fluorescence upon excitation with blue laser light. The histograms were evaluated by tirsi determining standard deviation of green fluorescence intensity pulses emit ted from AO-DNA complexes in single cells around the mean the standard deviation of the fluorescence intensity relative to the mean channel of fluorescence, i.e., the coefficient of variation, and then dividing channel of the fluorescence histograms. This represents the CV the coefficient of variation of a patient's sample by that of a control of the fluorescence intensity histograms (Fig. 1). To standardize sample (rCV). The mean rCV of cell populations of acute lymphoblastic and calculate the CVs of patient cell histograms, and subse leukemia (31 patients) differed significantly from that of nonlymphoblas- quently to analyze the quantitative differences between the tic leukemia (21 patients). When cells were treated with a solution of mean CVs of different disease groups, we divided the CV of citric acid and magnesium sulfate prior to their staining with acridine each patient by that of a control, i.e., the CV of a lymphocyte orange, the mean rCV of cell populations of acute lymphoblastic leukemia population obtained from a healthy blood donor, measured on increased while that of acute nonlymphoblastic leukemia decreased com the same day and using identical machine settings and staining pared to their respective pretreatment values. The mean difference of rCVs between untreated and treated cells (rCVD) within each disease solutions. category was statistically significant. A logistic regression model, based In this study of ALL and ANLL cell populations by using on rCVD, confirmed the conventional classification of acute lymphoblas AO flow cytometry, we showed that the rCVs of the histograms tic leukemia and acute nonlymphoblastic leukemia cells in 90% of the of leukemic cells before and after treatment with citric acid and cases. magnesium were significantly associated with the type of dis ease. In particular, it was the difference between the rCV of cells before and after treatment with acid that carried a high INTRODUCTION power of differentiating ALL from ANLL. Acute leukemias are mainly differentiated by the use of light microscopic criteria outlined in the French-American-British MATERIALS AND METHODS classification of acute leukemia (1). One group of criteria is based on cytoplasmic features, and another group on nuclear Patients. Between 1981 and 1983 we selected 52 previously untreated features. Diagnostic uncertainty arises in a sizeable number of patients with the diagnoses of ALL and ANLL made by these criteria: cases because microscopic findings are subject to variations microscopic evaluation of tetrachrome (Sigma Chemical Co., St. Louis, MO)-stained smears of bone marrow aspiration specimens, sudan black stemming from both technical factors (2) and subjective inter and myeloperoxidase staining reactions, and biochemical quantitation pretation by different observers (1, 3). of the enzyme deoxynucleotidyl transferase (7). No cases equivocal by With respect to the group of nuclear criteria, differences between ALL' and ANLL in the appearance of nuclear chro- these criteria were included here. The monoclonal antibodies now in use to subclassify acute leukemias into immunological categories were matin have long been noted in Romanowsky-stained smears, not available at the time. such as the discernibly clumped, irregularly distributed areas of In a pediatrie group of 22 patients, 14 [mean age of 6.1 Â±3.3 (SD) eliminatili in ALL, whereas in ANLL the nuclear chromatin years] had the diagnosis of ALL, and 8 patients (mean age of 7.6 Â±3.7 may appear distinct but "lacy," or it may look very fine, non- years) had the diagnosis of ANLL. In an adult group of 30 patients, 17 clumped, or finely stippled (4). Moreover, the chromosomes (mean age of 35.3 Â±18.6 years) had the diagnosis of ALL, whereas 13 are known to have a "fuzzy" appearance in metaphase spreads (mean age of 45.5 Â±17.1 years) had the diagnosis of ANLL. more often in ALL than in ANLL, even if the karyotype is The lymphocytes of 36 healthy blood donors served as controls. Blood lymphocytes of healthy individuals are quiescent cells residing normal (5). in the Go compartment of the cell cycle. We measured at least one In line with our hypothesis that microscopically different control specimen on each day when one or more patient specimens chromatin structures may exhibit different binding patterns of were measured. the intercalating agent, AO, to DNA, we found that the densely Preparation of Cell Specimens. We evaluated 52 leukemic cell sus packed chromatin of normal human peripheral blood lympho- pensions. Heparinized bone marrow aspirates (0.9-4.0 ml) were diluted with 20 ml of PBS, and heparinized blood samples (10-20 ml) were Received 8/12/88; revised 12/7/88. 3/8/89: accepted 3/29/89. diluted with double their volumes of PBS prior to the Ficoll-Hypaque The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in (Lymphoprep, Nyegaard; Oslo, Norway) gradient centrifugation. The accordance with 18 U.S.C. Section 1734 solely to indicate this fact. resulting mononuclear cell preparations were adjusted to 1-2 x IO6 1Supported in part by grants \Va460/l-2 from the Deutsche Forschungsge cells per ml of PBS, and stored at 2-4Â°Cuntil used. The cell suspensions meinschaft, and C'A-08748 and CA-05826 from the National Cancer Institute. J To whom requests for reprints should be addressed, at The New York contained between 6 and 100% (median, 69%; mean, 59%) blast cells Hospital and The Rogosin Institute. 430 East 71 St.. New York, NY 10021. on tetrachrome stained cytocentrifuge (Shandon-Elliot, Broomall, PA) 'The abbreviations used are: ALL. acute lymphoblastic leukemia: ANLL. preparations. Two of 21 ANLL and 3 of 31 ALL had <40% blast cells. acute nonlymphoblastic leukemia; AO, acridine orange: CV. coefficient of varia The percentage of blast counts did not differ statistically between ALL tion; MCF, mean channel of fluorescence intensity; PBS, phosphate-buffered saline: rCV. coefficient of variation of a patient's sample divided by that of a and ANLL. There was no significant correlation of the percentages of control sample: rCVC, rCVN, rCVD. relative CV of cells, nuclei, and the blasts in the cell suspensions with the measured CVs. This observation difference in rC'V between cells and nuclei, respectively. suggests that the morphological blast cell count may be inaccurate or 3692

ANLL ALL or nuclei, populations. This relationship was depicted in Fig. 2, where each of 3 specimens was measured at different MCFs as indicated on the ordinate. During the study, the machine settings were adjusted such cells that the G0-G, cell subpopulations of interest were always measured ir between channels 30 and 40 of the F-.Mphotomultiplier tube. Moreover, Z the flow rate of cells was kept at about 200/s. 13 50 100 O 50 100 Evaluation of Fluorescence Measurements. /â€¢'..â€ž,measurementswere CV3.4ICV'3.511 ui used to identify the ploidy levels, to calculate the CVs, and to analyze o nuclei the cell cycle phases of the malignant cell populations. The cell cycle evaluation with respect to the delineation of the G0-G, population was performed according to previously described software programs (10- 50 100 0 50 100 12). Aneuploid populations were detected and defined as previously MEAN CHANNEL OF GREEN FLUORESCENCE described (6). The incidence of aneuploidy in the current study was Fig. 1. Histograms of green fluorescence emitted by AO-DNA complexes in 26% with 3 of 21 ANLL and 10 of 31 ALL specimens displaying cells from ANLL and ALL cell populations prior to (cells) and after (nuclei) aneuploid stem lines. The present number of samples was too small for treatment with citric acid and magnesium. The peaks represent the G0-G, sub- a separate statistical analysis of the influence of aneuploidy on AO- populations. Note that treatment with citric acid and magnesium decreases the DNA binding patterns. The CV was calculated for each G0-G, DNA CV in ANLL but increases it in ALL. peak of control and patients' cell populations. The relative CVs of nuclei and cells of G0-G, cell populations were calculated by determin that some morphologically nonmalignant-appearing cells may be en ing the ratio of the CV of patient cells, or nuclei, to that of control dowed with abnormal AO-DNA binding features, thus obfuscating a cells. Moreover, the difference between the rCVs of cells and nuclei of statistically prominent correlation of blast cell counts with AO-DNA the same specimen was calculated: rCVD = rCVC - rCVN. The binding patterns of leukemic G0-G, cell populations. replicate measurements of control samples on the same day under Isolation of Nuclei. This was done as described by Walle (8). In brief, 0.2-0.4 ml of a concentrated mononuclear cell suspension (1-1.5 x 106 identical staining and machine conditions were within Â±3%as far as cells) was added to 3 ml of a freshly prepared solution of 1.5% citric the CV was concerned. Similar data on replicate measurements of acid and 0.5 mM MgSO4 in distilled water. The nuclei were isolated by patient samples are currently not available. moving the piston of a 10 ml plastic syringe, attached to a 26-gauge Statistical Evaluation. Each patient contributed one set of measure hypodermic needle, 8-12 times up and down. The isolated nuclei, ments to the evaluation. The one-sample t test was used for comparisons washed and resuspended in Hanks' balanced salt solution without of means of parameters among disease and control groups. The two- calcium, were kept on ice until used (5-60 min). sample t test was applied to comparisons of means of rCVCs and Staining of Cells and Nuclei with Acridine Orange. The procedure rCVNs within disease groups. The method of maximum likelihood was described by Darzynkiewicz et al. (9, 10) and TrÃ¡ganoset al. (11). logistic regression was used to determine the joint prognostic signifi In brief, permeable cells were stained with AO at a dye/DNA-P molar cance of the rCVs in predicting the probability of a patient having ALL ratio of >2, in the presence of 1 mM sodium EDTA. Under these rather than ANLL. Likelihood ratio tests were used to assess signifi circumstances the intercalation of the dye into DNA results in green cance levels. The estimated logit equation was combined with an fluorescence with a maximum emission at 530 nm (F5jo), whereas the appropriate log odds of having ALL rather than ANLL to arrive at a dye stacking to DNA emits RNase-insensitive red luminescence with a decision rule to discriminate ALL patients from ANLL patients. Such maximum emission at 640 nm (Ftm), as described (12,13). The intensity a logistic discriminant procedure can be shown to minimize the prob of green fluorescence was shown to be proportional to the content of ability of misclassification (for example, see Ref. 17). DNA per cell (14, 15). The specificity of staining in the leukemic cells was estimated by preincubating permeabilized cells with DNase I or RNase A, as described (8, 11). Two hundred Â¿ilofcells in suspension 100-, were treated with 0.4 ml of a solution containing 0.08 N HC1, 0.146 N Nad, and 0.1% Triton X-100 at 0Â°C.Thirty s later an aliquot of 1.2 v u a> ml of AO solution (6 Mg/ml; 20 /IM) in 0.2 M phosphate-citric acid e 80- buffer (pH 6.0) containing 1 mM EDTA and 0.146 M NaCl was added. o Aliquots of 0.2 ml of nuclei in suspension were mixed with 1.2 ml of 3 AO solution (20 MMAO) in 0.2 M phosphate-citric acid buffer, at pH o 60 6.0, and incubated for 30 s. In the staining procedure the number of 'S nuclei did not exceed 2 x 10"/ml in the final staining solution contain C. â€¢ C ing approximately 13-20 MMAO, so that the AO/DNA-P ratio was o 4O .C always >2 (12). O Fluorescence Measurements. The fluorescence of individual cells or nuclei was measured in an FC200 flow cytometer (Ortho Diagnostics, g 20- Westwood, MA) equipped with a 50-mW argon ion laser, at 15 mW o laser power output. The green (filter, 515-575 nm) fluorescence emis IO K> 0-1 sion from each particle was measured, and the data were stored in a 1.0 0.8 0.6 0.4 Nova minicomputer (Data General Corp., Westboro, MA). The fluo rescence pulse-width values of nuclei were recorded to determine nu Coefficient of Variation clear diameter and to eliminate cell doublets from the evaluations (16). Fig. 2. The green DNA-derived fluorescence intensity, indicated as mean Five to 10 x IO6 cells or nuclei were measured per sample. Each channel of fluorescence is plotted against the coefficient of variation of the green patient's sample was measured with and without addition of 50% fluorescence histogram. Each of 3 control cell specimens from different days was measured at the indicated MCFs. To account for day-to-day variations, the CVs normal peripheral blood lymphocytes from donors both at low and were normalized for each day. The standard deviations for the measurements high photomultiplier tube settings. This enabled us to detect deviations after normalization did not exceed Â±5%ofthe mean, indicating that the displayed from the diploid DNA content of the G0-G, cell populations of >5%. relationship between MCF and CV is rather similar from day to day. At a low G0-G] cell populations of both control and patient samples were meas MCF the CV is high, whereas the CV decreases with increasing MCF. The observed effect may be due to bit precision error in the assignment of cells to ured at the same green MCF, including whole cells as well as nuclei. channels. As such, this effect is most pronounced at low MCFs. The extent of Because of the influence of the MCF on the CV, this proved to be of this effect may differ among different flow cytometers. Each of 3 control cell crucial importance, in particular with respect to aneuploid G0-G, cell, specimens was measured at the indicated MCFs. 3693

RESULTS parameter logistic regression model suggests that a one-param eter model based on rCVD can fit the data adequately. This The summary statistics of the rCVCs, rCVNs and rCVDs are was indeed the case and the simpler one-parameter model was listed separately for the two age as well as disease groups of chosen instead. The fitted equation was patients, and the controls (Table 1). There was no significant difference between the two age groups in any of the measured /Â»(ALL)= parameters within a given disease category; moreover, separate 1 + exp(jc), logistic regression analyses on the two groups did not show any difference in the relationship of any of the parameters with the where jc= 1.136- 15.01 rCVD. The usefulness of this equation probability of having ALL rather than ANLL. For these rea is illustrated in Fig. 3. The curve is characterized by a steep sons, the data from all the 52 patients were pooled for subse slope between rCVDs of 0.00 and 0.15, with the point of quent statistical analyses. inflection being at rCVD of 0.0757. Obviously, small changes The mean rCVCs of both ALL and ANLL were significantly in values for rCVD within this range result in large changes of larger (P < 0.00001) than the mean rCVC of control cells, probability values. which had the assigned value of 1.0. The mean rCVN of control The logistic discriminant rule was given by cells was significantly smaller than that of ALL (P < 0.00001) and that of ANLL (P = 0.014) (Table 1). D = 1.136 - 15.01 rCVD + In 1 -/Â»' Striking differences between the two disease categories were observed (Table 1). First, the rCVC of ALL was significantly where p is the prevalence rate of ALL over ANLL in the general smaller than that of ANLL (P < 0.00001), while the mean pediatric/adult patient population from which the pediatrie/ rCVN of ALL was significantly larger than that of ANLL (P< adult patient sample was drawn. A patient was classified as 0.00001 ). Second, the rC VD values of ALL were predominantly ALL if D > 0, and as ANLL if D < 0. It is an epidemiological negative, and the mean rCVD was significantly smaller than experience that ALL is the prevalent type of acute leukemia in zero (P < 0.00001); the reverse, however, was true of ANLL. the pediatrie population, whereas ANLL is the prevalent type This characteristic pattern was violated by 1 of 21 cases of in the adult population. For Memorial Hospital, we estimated ANLL and by 4 of 31 cases of ALL. As will be seen below, this p to be between 0.7 and 0.8 for pediatrie patients and between difference between rCVC and rCVN, i.e., rCVD, led to a very 0.2 and 0.3 for adult patients. The discriminant rule, when sharp discrimination between ALL and ANLL. applied to the present pediatrie patient sample confirmed the In the logistic regression analysis relating the probability of a patient having ALL rather than ANLL [/Â»(ALL)]to the morphological diagnosis in 86% of cases, and in 90% of cases when applied to the adult patient population. Our results sug measured parameters, rCVC was found to be a much more gest that rCVD potentially carries a very high power of accom important prognostic indicator (P < 0.00001) than rCVN (P = plishing the therapeutically and prognostic-ally critical differ 0.0043). With rCVC in the model, however, rCVN still added entiation of ALL from ANLL in pediatrie as well as in adult significantly to the prediction (P = 0.00001). The fitted two- patients. TableI Summary-statisticsofmeasuredparametersfarpediatrieandadult patientswithALLandANLL,andthecontrolgroup DISCUSSION

Pediatrie22)ALL(n= patients (n = This report on a pilot study describes how flow cytometrically 14)rCVCrCVNrCVDANLL measured histograms of green fluorescence derived from AO- stained DNA of cells and their isolated nuclei can be used to discriminate lymphoblastic from nonlymphoblastic leukemic 8)rCVCrCVNrCVDAdult(n = cell populations with up to 90% probability of confirming the conventional classification. The values of rCVC as well as those of rCVN discriminated significantly between ALL and ANLL. The rCVD values, how- 30)ALLpatients (n = (n =17)rCVCrCVNrCVDANLL(n= 1.0 0.9 13)rCVCrCVNrCVDPediatrie Q8 t Â°7 i Â°-6 patientbinedand adult com-1.1631.393-0.2301.4321.1630.2701.0001.006-0.006s 2 0.5 52)ALL(n (n = o =31)rCVCrCVNrCVDANLL(n Â£ 0.4 0.3 0.2 21)rCVCrCVNrCVDControl= 0.1 0.0 -0.50 -025 0.00 0.25 0.50 (n =36)rCVC"rCVNrCVD"rCVC rCVO Fig. 3. The probability of a given sample of acute leukemia cells of being lymphoblastic rather than nonlymphoblastic is plotted as a function of the measured rCVD. Note that the point of inflection is at rCVD = 0.0757. This of control cells iMean1.2251.436-0.2111.4011.1560.2451.1121.357-0.2451.4521.1670.285s1 by definition.Median1.1651.370-0.1401.4001.2300.2351.1101.300-0.1701.3401.1900.2501.1301.340-0.1601.3601.1900.2401.0001.0131.025SD0.2380.3090.2890.2310.2470.1430.1540.3310.2750.2010.2430.2290.2010.3180.2770.2090.2380.1980.0000.1890.189indicates that only 4 of 51 cases had a probability of correct diagnosis of <51 ' < 3694

Downloaded from cancerres.aacrjournals.org on September 23, 2021. © 1989 American Association for Cancer Research. DIFFERENTIATION OF ALL FROM ANLL BY DNA FLOW CYTOMETRY ever, discriminated between the two diseases even more signif ACKNOWLEDGMENTS icantly. We are indebted to Dr. M. Andreeff, whose flow cytometry facility Moreover, this study revealed that the binding of AO to DNA was used to perform part of this work, and to the attending physicians of ALL differs from that of ANLL cell populations by virtue of of the Hematology and Pediatrie Services of Memorial Sloan-Kettering their responses to brief treatment with citric acid and magne Cancer Center, who had cared for the patients we studied. This work sium. After treatment of cell populations, the mean rCV of the is dedicated to Annemarie. isolated nuclei populations increased significantly in ALL but decreased significantly in ANLL. There were only 5 of 52 cases of acute leukemias which violated this characteristic pattern. REFERENCES Whether some of these 5 cases were misclassified by the con 1. Bennett. J. M.. Catovsky, D.. Daniel. M. T., Flandrin. G.. Galton. D. A. G., ventional laboratory parameters in the absence of monoclonal Gralnick. H. R., and Sultan. C. Proposal for the classification of the acute antibody analyses such as T-cell or myeloid cell antibodies, leukemias. Br. J. Haematol.. 33:451-458. 1976. 2. Baumann. I.. Gunzer. U.. and Aus, H. M. Influence of pH. buffer system remains unanswered. However, a number of reasons suggest and Di-K-EDTA on standard Romanowsky-Giemsa staining and the effect that rCVD represents a parameter of high prognostic value, in both cellular architecture and color measured in peripheral blood cells. particularly since not a perfect single one method of classifica Cytometry. A. Suppl. I, 35. 1987. 3. Lampen. F., Henzc, G., Langermann, H. J., Schellong. G., Gadner. H., and tion is currently available. These reasons include that rCVD is Rifinii. H. J. Acute lymphoblastic leukemia: current status of therapy in a new marker describing AO-binding patterns to DNA of leu- children. In: E. Thiel and S. Thierfelder (eds.). Recent Results in Cancer Research, pp. 159-181. Berlin: Springer-Verlag. 1984. kemic cell populations; it is a parameter independent of current 4. Winirobe. M. M., Lee. G. R.. Boggs. D. R.. Bithell. T. C., Foerster, J., ones used to classify leukemias; it is devoid of subjective bias Athens. J. W., and Lukes, J. N. Clinical Hematology, Ed. 8, pp. 1493-1506. which is a drawback of light microscopic evaluations of bone Philadelphia: Lea & Febiger. 1981. 5. Sandberg, A. A. The chromosomes in Human Cancer and Leukemia, pp. marrow smears; it does not require special expertise on the part 262-347. New York: Elsevier Science Publishing Co.. 1980. 6. Walle. A. J. Identification of L3 leukemia and Burkitt's lymphoma cells by of the investigator as is necessary for microscopic evaluations; flow cytometric quantitation of nuclear DNA and RNA content. Leuk. Res., and it correlates significantly with the disease category. 10: 303-312, 1986. The mechanism(s) for the characteristic response patterns of 7. Mertelsmann, R., Koziner, B.. Filippa, D. A.. Grossbard. E., Icefy, G., Moore, M. A. S., and Clarkson, B. D. Clinical significance of Tdt, cell surface ALL and ANLL cell populations to treatment with citric acid markers and CFU-C in 279 patients with hÃ©matologieneoplasia, in: R. Nelh, and magnesium is presently unclear. We have observed some R. C. Gallo. P. H. Hofschneider, and K. Mannweiler (eds.). Modern Trends cases of lymphoid and myeloid blast crisis of chronic myeloid in Human Leukemia. Vol. 3, p. 131. 8. Walle. A. J. Distribution and content of nuclear and cellular RNA among leukemias that exhibited patterns of responses similar to the Cancercell populations Res.. 45: 5193-5195'. of acute lymphoblastic 1985. and nonlymphoblastic leukemia. ones characteristic of ALL and ANLL, respectively. 9. Darzynkiewicz, Z., TrÃ¡ganos, F., Sharpless, T. K., and Melamed. M. R. Because of the highly significant association of the reaction Conformation of RNA in situ as studied by acridine orange staining and patterns with the disease types, we developed a statistical pro automated cytofluorometry. Exp. Cell Res.. 95: 143-153, 1975. 10. Darzynkiewicz, Z., TrÃ¡ganos, F., Sharpless. T. K., and Melamed, M. R. cedure using logistic discriminant analysis to differentiate ALL Lymphocyte stimulation: a rapid multiparameter analysis. Proc. Nati. Acad. from ANLL. Assuming a prevalence rate of 0.2 of ALL over Sci. USA. 73: 2881-2884. 1976. ANLL in our adult patient population, the equation confirmed 11. TrÃ¡ganos, F.. Darzynkiewicz. Z.. Sharpless, T. K.. and Melamed, M. R. Simultaneous staining of ribonucleic and deoxyribonuclcic acids in unfixed the morphological diagnosis with a probability of 90%, based cells using acridine orange in a flow cytometric system. J. Histochem. solely on rCVD. Fig. 3 displays the relationship of rCVD and Cytochem., 25: 46-56. 1977. 12. Darzynkiewicz. Z. Acridine orange as a molecular probe in studies of nucleic the estimated probability of correct diagnosis in the form of a acids in situ. In: M. R. Melamed. P. Mullaney. and M. Mendelsohn (eds.). nomogram. If based on a much larger number of cases, such a Flow Cytometry and Sorting, pp. 283-316. New York: John Wiley & Sons, 1979. nomogram can be used for the practical purpose of obtaining 13. Kapuscinsky. J.. Darzynkiewicz, Z., and Melamed. M. R. Luminescence of the probability of having ALL rather than ANLL for the solid complexes of acridine orange with RNA. Cytometry, 2: 201-211, 1982. 14. Lerman, S. The structure of the DNA-acridine orange complex. Proc. Nati. measured variable. Acad. Sci. USA. 49: 94-102, 1963. The present pilot study indicates that a single parameter of 15. Coulson, P. B.. Bishop, A. O.. and Lenarduzzi. R. Quantitation of cellular certain malignant cell populations carries an enormous amount deoxyribonucleic acid by flow microfluorometry. J. Histochem. Cytochem.. of disease-specific information. The described system of auto 25: 1147-1153. 1977. 16. Sharpless, T. K., TrÃ¡ganos.F., Darzynkiewicz, Z.. and Melamed. M. R. Flow mated specimen analysis and evaluation may help in defining cytometry: discrimination between single cells and cell aggregates by direct size measurements. Acta Cytol., 19: 577-581, 1975. in more detail the binding pattern of AO to DNA in situ of 17. Dillon, W. R.. and Goldstein, M. Multivariate Analysis. New York: John undifferentiated or mixed phenotype acute leukemias. Wiley & Sons. Inc., 1984.

3695

Alexander J. Walle and George Y. Wong

Cancer Res 1989;49:3692-3695.

Updated version Access the most recent version of this article at: http://cancerres.aacrjournals.org/content/49/13/3692

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/49/13/3692. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.