- Email: [email protected]
Multiparameter analysis and sorting of mammalian cells
Printed in Sweden Copyright Q 1974 by Academic Press, Inc. All rights of reproduction in any form reserued
Experimental Cell Research84 (1974) 15-23
MULTIPARAMETER
ANALYSIS AND SORTING OF MAMMALIAN
J. A. STEINKAMP,
CELLS
A. ROMERO, P. K. HORAN, and H. A. CRISSMAN
Biophysics and Instrumentation Group, Los AIamos Scientific Laboratory, University of California, Los Alamos, N. Mex. 87544, USA
SUMMARY An improved multisensor cell sorting instrument for quantitative analysis and sorting of cells has been developed. Cells stained with fluorescent dyes enter a flow chamber where cell volume, fluorescence, and light scatter sensors simultaneously measure multiple cellular properties. Cells then emerge in a liquid jet that is broken into uniform liquid droplets. Sensor signals are electronically processed in one of several ways for optimum cell discrimination and are displayed as pulse-amplitude distributions using a multichannel pulse-height analyzer. Processed signals activate cell sorting according to preselected parametric criteria by electrically charging droplets containing cells and electrostatically deflecting them into collection vessels. Illustrative examples of multiparameter cell analysis and sorting experiments using a model mouse tumor cell system, human and animal leukocytes, and cultured mammalian cells are presented.
The development of high-speed, flow-system, cell-sorting methods based on physical and biochemical measurements of single cells has provided a new dimension to biological investigation by being able to physically isolate cells with particular properties from a heterogeneous mixture [l, 2, 31.Such systems have been used to differentiate and sort human leukocytes based on volume measurement [4], to isolate antigen-binding cells which are precursors to antibody-producing cells by use of rmmunofluorescence [5], and to separate abnormal human cells from uterine cervical carcinoma specimens based on UV absorption and visible light scatter [2]. New and improved instrumentation for separating cells based on simultaneous measurement of cell volume, total or two-color fluorescence, and light scatter recently has been developed [6]. This system incorporates several existing analytical techniques, permitting multiple measurements to be per2-741811
formed on the same cell. Coupled with cell sorting, multiparameter analysis methods provide unusual versatility in experimental cell research by measuring and processing several cell characteristics simultaneously and separating cells according to selected combinations of parameters. Cell sorting permits the correlation of machine-measured cell properties with cell morphology and cell sample enrichment (i.e., concentration of cell subclasses). Presented in this report is a description of the cell analysis and sorting techniques made possible with this instrument. Validation of the methodology is discussed with illustrative examples of cell analysis and sorting on the basis of single parameters, ratios, and two parameter combinations of cell volume and fluorescence. Analysis and sorting of (a) SH-thymidine pulse-labeled CHO cells and mouse tumor cells on the basis of DNA content; (b) dog leukocytes Exptl Cell Res 84 (1974)
16 Steinkamp et al.
Coulter Volume
Current
Source
Sensor
I
8 Preamp
Multiparameter Signal Processing Electronics - 1
‘3 &gigI
/
Sample Collection
-
Fig. 1. Multisensor cell separator system diagram illustrating the flow chamber, laser illumination, cell sensors, signal processing and cell separation electronics, and droplet charging and deflection scheme. The piezoelectric transducer used for droplet generation is not shown.
into subclassesof lymphocytes, neutrophils, and eosinophils according to cytoplasmic granulation; (c) human lymphocytes into two subclassesbased on the degree of cytoplasmic granulation density; and (d) HeLa cells based on DNA-cell volume relationships are presented. MATERIALS AND METHODS Instrumentation A comprehensive description of the instrumentation has been described elsewhere [6]. Cells stained with fluorescent dyes that label specific biochemical components are suspended in normal saline and introduced into the flow chamber (fig. 1) at approx. 500 cells/set. Cells first pass centrally through a 75 ,um diameter volume-sensing orifice. (Coulter principle) and then across a fluid-filled viewing region intersecting an argon laser beam (488 nm wavelength), causing fluorescence and light scatter. Both are Exptl Cell Res 84 (1974)
electro-optically measured, the fluorescence sensor being a dual photomultiplier tube array which measures total (above 520 MY wavelength) or two-color fluorescence of selectable color separation regions. Light scatter can also be measured by optically focusing the forward scattered light onto a photodiode. After optical measurement, the liquid stream carrying suspended cells emerges into air as a jet and is broken into uniform droplets (45 OOO/sec).Thus, cells are isolated into droplets [l] with approx. 1 % of the droplets containing a cell. Parametric signals from the cell sensor amplifiers (fig. 1) are processed on a cell-by-cell basis as single parameters, ratios (analog pulse divider), and gated single parameters. A multichannel pulse-height analyzer accumulates and displays processed signals as pulse-amplitude frequency distribution histograms. Gated single parameter analysis permits the examination of particular subclasses of cells within selectable ranges on different cellular property values. For example, the volume distribution for Gl cells can be obtained by analysing only those processed volume signals from cells yielding Gl DNA content as determined by fluorescence measurement. That is, if the DNA content signal falls within the Gl DNA content range as determined by adjustable signal amplitude gates, only then will the associated volume
Multiparameter signal be accumulated and displayed. Gated single parameter techniques also permit distributions to be obtained from weakly fluorescing signals (e.g., fluorescent antibody) by requiring electronic coincidence with cell volume or light scatter signals. In this manner, fluorescence signals from debris is eliminated. Complete two parameter analysis is available by selecting two processed signals as inputs to a dual parameter pulse-height analyzer and displaying subsequent three-dimensional pulse-height distributions and two-dimensional contour views. Processed signals activate cell sorting if the amplitudes fall within preselected ranges of single-channel, pulse-height analyzers (SCAs) located within the cell separation logic. An electronic time delay is activated which then triggers a positive or negative droplet charging pulse, causing a group of droplets containing a-cell-to be charged and deflected by a static electric field into a collection vessel. Those cells not satisfying the sorting criteria pass undefleeted into a separate collection vessel. The sorted suspension is introduced into a cytocentrifuge (Shandon Scientific Co., Sewickley, Pa) and deposited onto microscope slides for counterstaining and microscopic examination. Some experiments require sorted cells to be reanalyzed and separated again.
Cell staining and dispersal Chinese hamster cells (line CHO) grown in spinner culture in F-10 medium supplemented with 10 % calf and 5 % fetal calf sera were pulse-labeled with 5 ,&i/ ml of aH-thymidine for 15 min, dispersed with trypsin, fixed in 70% ethanol for 0.5 h, treated with RNase, and stained with propidium iodide [7]. Stained cells were then suspended in normal saline for fluorescence analysis of DNA content and sorting. Sorted samples were subjected to autoradiography. Mouse spleen and a methylcholanthrene-induced squamous cell mouse tumor (MCA-1) grown in C3H/Hej mice were minced, trypsinized [8], and then dispersed. Cells were filtered by passing through a nylon mesh filter (80 pm pore size), fixed in 10% formalin for 12-18 h, stained with acriflavine using the fluorescent-Feulgen DNA staining procedure [9], and suspended in normal saline for fluorescence analysis of DNA content and subsequent sorting. The sorted cells were then counterstained using the standard Pananicolaou method. Fresh whole blood from normal humans and dogs (beagles) containing 0.01 ml of 10% EDTA/ml of blood was prepared for fluorescence analysis and sorting by diluting one part blood with 25 oarts of acrid&e -orange golution [lo]. Stained leukocytes exhibit a uniform green nucleus with distinct red cytoplasmic granules as viewed with the fluorescence microscope. Erythrocytes do not fluoresce at this dye concentration. In these measurements, the twocolor fluorescence sensor was set to measure zreen 520-580 nm) and red (590-800 nm) fluoresc&ce. Sorted leukocytes were fixed in 100 % methanol and counter-stained using Wright-Giemsa. HeLa cells (line S3), grown in double-strength BME supplemented with 20% fetal calf serum on monolayer, were dispersed with trypsin, fixed in
cell sorting
1I
formalin, and stained with acriflavine for DNA as described above. Cells were resuspended in normal saline for analysis of cell volume and DNA content and for demonstration of two parameter cell sorting capability.
RESULTS AND DISCUSSION for CHO The DNA content distribution cells pulse-labeled for 15 min with 3Hthymidine is shown in fig. 2. The first peak represents Gl cells and the second peak G2 +M cells, having twice the DNA content. Cells synthesizing DNA (S phase) lie in the region between the peaks. Equal numbers of cells sorted on the basis of Gl, S, and G2 +M DNA content, designated by sort regions 1, 2, and 3, respectively, were examined autoradiographically. The percentage of labeled cells in each fraction is shown in fig. 2. The cell population had a total labeling index of 33 %. The small percentage of labeled cells in the sorted Gl fraction is due to early S cells, whereas those labeled cells in the G2+M sorted fraction are late S
)r
10,00c
Fig. 2. Abscissa: channel no. (proportional to cellular DNA content); ordinate: no. of cells. DNA content freauencv distribution histomam of CHO cells pulse-labeled for 15 mm with $H-th-&dine and stained with propidium iodide. The dotted line is an approximation of a second-degree polynomial computer curve-fit of cells in S phase [8, 111. Exptl Cell Res 84 (1974)
18 Steinkamp et al.
10%
(
,
,
,
T
,
[
,
,
’ 60
’
’ SO
’
I (4
100
104-
OO
20 -q Sort1
tel-"OOSort
2
100 L
Fig. 3. Abscissa: channel no. (proportional to cellular DNA content)+;ordinate: no. of cells. Frequency distribution histograms of acriflavine-Feulgen DNA stamed mouse (C3H/Hej) spleen and MCA-1 tumor cells: (A) DNA content distribution of normal spleen cells; (I?) DNA content distribution of MCA-1 tumor cells; (C) photomicrographs of dispersed tumor cells prior to sorting; (D) sorted mouse leukocytes of sort region 1; (E) sorted tumorigenic cells of sort region 2. Both DNA content distributions of (A) and (I?) were measured at the same fluorescence channel amplifier gain setting.
cells. These techniques currently are being used to study the nature of cell-cycle kinetics in populations of human diploid cells maintained in low serum-containing medium [12]. Fig. 3 shows the DNA content distributions for normal mouse spleen and methylcholanthrene-induced mouse tumor (MCA-1) cells. The mouse spleen cell DNA distribution (fig. 3A) provides an indication for normal 2C diploid DNA content. The DNA distribution of MCA-1 tumor cells (fig. 3B) shows three peaks, the first peak being cells having a normal 2C diploid DNA content and the second and third peaks representing cells having 4C and 8C DNA contents, respectively. The tumor cell population was sorted on the basis of DNA content into Exptl Cell Res 84 (1974)
two groups, designated by sort regions 1 and 2 (fig. 3 B). Fig. 3 C is a photomicrograph of dispersed and counterstained tumor cells prior to sorting. Cells having normal DNA content from sort region 1 are leukocytes (fig. 3 D), whereas cells containing an elevated or abnormal DNA content from sort region 2 are tumorigenic cells (fig. 39, the two subpopulations being morphologically distinct. The second and third peaks of fig. 3 B are Gl and G2+M phase tumor cells, with S phase cells contained between the peaks. These results have been verified for MCA-1 tumor cells cultured in vivo and in vitro [13]. Other tumor cells with non-elevated DNA content most certainly will not be identifiable on the basis of DNA alone. Other para-
Multiparameter
cell sorting
19
Fig. 4. Abscissa: channel no. (proportional to cellular fluorescence); ordinate: no. of cells. Frequency distribution histograms of acridine orange-stained normal dog (beagle) leukocytes: (A) green fluorescence distribution; (B) red fluorescence distribution; (C) photomicrographs of blood smear prior to sorting; (0) sorted lymphocytes; (I?) sorted neutrophils; (F) sorted eosinophils. Erythrocytes contained in the photomicrographs were separated along with selected leukocytes.
meters which characterize cell abnormalcy are under investigation and include DNA, protein and cell volume relationships, nuclear-to-cytoplasmic ratios by laser beam scanning, and others such as antigenic properties of tumorigenic cells.
The red and green fluorescence pulseamplitude distributions of normal dog (beagle) leukocytes vitally stained with acridine orange are shown in fig. 4. The green fluorescence distribution (fig. 4A) is unimodal, illustrating uniform nuclear staining, whereas Exptl Cell Res 84 (1974)
20 Steinkamp et al.
Fig. 5. Abscissa: channel no. (proportional to cellular fluorescence); ordinate: no. of cells. Frequency distribution histograms of acridine orange-stained normal human leukocytes: (A) 0, green; l , red fluorescence distributions; (B) green-to-red fluorescence ratio distribution; (C) photomicrographs of blood smear prior to sorting; (0) sorted large lymphocytes; (E) sorted small lymphocytes. Erythrocytes contained in the photomicrographs were separated along with selected leukocytes.
the red fluorescence distribution (fig. 4B) shows three distinct peaks, characterizing dog leukocytes primarily into 3 subpopulations on the basis of cytoplasmic granules which exhibit red fluorescence. Leukocytes having red fluorescence signal amplitudes corresponding to sort regions 1, 2, and 3 (fig. 49 were separated, counterstained, and identified by microscopic examination as principally consisting of lymphocytes, neutrophils, and eosinophils, respectively. Fig. 4C-F shows photomicrographs of the blood smear and sorted leukocyte fractions. Eosinophils are characterized by coarse cytoplasmic granules, compared with finer (dust-like) granulation in lymphocytes and neutrophils, as exemplified in the red fluorescence distribution. Exptl Cell Res 84 (1974)
The green and red fluorescence distributions of normal human leukocytes vitally stained with acridine orange are shown in fig. 5A. The green fluorescence distribution again is unimodal, with the red fluorescence distribution indicating three distinct leukocyte subpopulations (peaks l-3), having been identified as lymphocytes, monocytes, and granulocytes, respectively [lo, 141. The green-to-red fluorescence ratio distribution (fig. 5B), which shows four distinct subpopulations (peaks l-4), permits leukocyte characterization as a function of nuclear-tocytoplasmic staining properties. Leukocytes having green-to-red fluorescence ratio signal amplitudes corresponding to sort regions 1, 2, 3, and 4 (fig. 5B) were separated, counterstained, and identified as granulocytes,
Multiparameter
cell sorting
21
6,000
Fig. 6. Abscissa: channel no.; ordinate: no. of cells. Frequency distribution histograms of acriflavine-Feulgen DNA stained HeLa cells: (A) l , DNA content; 0, cell volume distributions. (II) Cell volume distributions of l , Gl; a, G2 +M cells as obtained by gated single parameter analysis. (C) Two parameter DNA-cell volume plots illustrating three-dimensional isometric and two-dimensional contour views of cells prior to sorting. (D) Cells reanalyzed after sorting. The contour views were recorded near the zero cell number level.
monocytes, and two subclasses of lymphocytes, respectively. Fig. 5 C-E shows photomicrographs of the blood smear and sorted lymphocyte fractions 3 and 4. Preliminary investigations on sorted lymphocytes subjected to volume measurement indicate that fractions 3 and 4 are classesof large and small lymphocytes, respectively, with some overlap of the volume distributions. Since nuclear staining is approximately uniform, the greento-red fluorescence ratio measurement allows lymphocytes to be classified as to the degree of cytoplasmic granulation density. This technique provides possible new methodology for study of lymphocyte kinetics with respect to immune response. Cell analysis and sorting on the basis of two parameters are demonstrated using acriflavine-Feulgen stained HeLa cells. Single parameter DNA content and cell volume distributions were first recorded (fig. 6A) as was the DNA-cell volume two parameter
plot (fig. 6C). The DNA content distribution for exponentially grown HeLa cells is similar to the CHO cell DNA distribution (fig. 2). The volume distribution of fixed and stained HeLa cells is broad, unimodal, and typical of cells growing exponentially [15]. Fig. 6 C shows a three-dimensional DNA-cell volume distribution (isometric view) and a twodimensional contour view which is a baseplane section of the isometric display. The contour display illustrates the correlation of DNA content and cell volume around the cell cycle, with additional fine structure capable of being observed with this type of analysis. Two parameter analysis also provides a method to quantitate the number of cells within a given range of two cellular properties. Recent two-color fluorescence analysis of DNA and protein content relationships in doubly stained HeLa and CHO cells [6, 71 compare favorably with DNA-cell volume measurements. Exptl Cell Res 84 (1974)
22
Steinkamp et al.
Using gated single parameter analysis methods, the individual Gl and G2+M cell volume distributions were next recorded (fig. 6B) by analyzing the volume of cells having Gl and G2+M DNA content, respectively (channels 25-33 and 58-70 of the DNA distribution). Equal numbers of cells were then sorted into two groups, designated by sort regions 1 and 2 (fig. 6B), resulting in small volume Gl and large volume G2-l-M cells being collected. The two subpopulations were mixed together and rerun through the cell sorter, recording the DNA-cell volume two parameter distribution (fig. 60). The two parameter DNA-cell volume distribution of the combined sorted populations vividly illustrates this method of cell sorting and provides additional capability to measure quantitatively and characterize cell properties on the basis of two or more parameters. We have presented several examples of analysis and sorting of mammalian cells on the basis of single and two parameter combinations of cell volume and total or twocolor fluorescence. Multiparameter signal processing coupled with cell sorting provides a versatile and extremely useful method for cell biology research with the range of applications expanding as new methods of obtaining information on cells become available. Many other combinations of measurements are possible with the present instrumentation. The briefly described flow chamber contains a Coulter cell volume sensor coupled with fluorescence and light scatter optical sensors. Both Coulter and light scatter sensors yield cell size information. We have determined experimentally that the Coulter channel output is directly proportional to cell volume over allarger range of cell sizes than small angle light scatter [6], laser beam shape, wavelength, and cell diameter having important relationships in cell Exptl Cell Res 84 (1974)
size determinations [16]. Although cell fixation reduces the Coulter volume signal, we find only minimal cell volume distribution distortion as measured on cultured cells [7]. An additional electronic circuit has been constructed which converts cell volume signals to the two-thirds power, providing methodology for cell surface area determinations [17]. In principle, other cellsensing and analysis methods may give meaningful information on physical and biochemical cellular characteristics as both instrumental development and biological applications are in their initial stages of growth. This work was performed at the Los Alamos Scientific Laboratory, Los Alamos, New Mexico, under joint sponsorship of the US AEC and the NCI. We thank Mrs Julia Grilly for photomicrography of sorted cells, L. M. Holland, D. V. M., and Mrs J. E. London for obtaining blood samples, and Mrs P. C. Sanders for autoradiography. MC&-l tumor was obtained from the Laboratory of Pathology, NCI, NlH, Bethesda, Md, USA.
REFERENCES 1. Fulwyler, M J, Science 150 (1965) 910. 2. Kamentsky, L A & Melamed, M R, Science 156 (1967) 1364. 3. Hulett, H R, Bonner, W A, Barrett, J & Herzenberg, L A, Science 166 (1969) 747. 4. Van Dilla, M A, Fulwyler, M J & Boone, I U, Proc sot exptl biol med 125 (1967) 367. 5. Julius, M H, Masuda, T & Herzenberg, L A, Proc natl acad sci US 69 (1972) 1934. 6. Steinkamp, J A, Fulwyler, M J, Coulter, J R, Hiebert, R D, Homey, J L & Mullaney, P F, Rev sci instr 44 (1973) 1301. 7. Crissman, H A & Steinkamp, J A, J cell biol 59 (1973) 766. 8. Kraemer, P M, Deaven, L L, Crissman, H A & Van Dilla. M A. Advances in cell and molecular biology (ed E’ DuPraw) vol. 2, pp 47-108. Academic Press. New York and London (1972). 9. Tobey, R A, Crissman, H A & Kraemer,‘P M, J cell biol 54 (1972) 638. 10. Adams, L R & Kamentsky, L A, Acta cytol 15 (1971) 289. 11. Dean, P N & Jett, J H, J cell biol(1974). In press. 12. Dell’Orco, R, Crissman, H A, Steinkamp, J A & Kraemer,.P M. In preparation. 13. Horan, P K, Romero, A, Steinkamp, J A 8~
Multiparameter cell sorting 23 Petersen, D F, J natl cancer inst US (1974). In press. 14. Steinkamp, J A, Romero, A & Van Dilla, M A, Acta cytol 17 (1973) 113. 15. Anderson, E C, Bell, G I, Petersen, D F & Tobey, R A, Biophys j 9 (1969) 246.
16. Mullaney, P F & Dean, P N, Biophys j 10 (1970) 764. 17. Steinkamp, J A, Larkins, J H, Carr, L J & Kraemer, P M. In preparation. Received September 3, 1973
Exptl Cell Res 84 (1974)
Experimental Cell Research84 (1974) 15-23
MULTIPARAMETER
ANALYSIS AND SORTING OF MAMMALIAN
J. A. STEINKAMP,
CELLS
A. ROMERO, P. K. HORAN, and H. A. CRISSMAN
Biophysics and Instrumentation Group, Los AIamos Scientific Laboratory, University of California, Los Alamos, N. Mex. 87544, USA
SUMMARY An improved multisensor cell sorting instrument for quantitative analysis and sorting of cells has been developed. Cells stained with fluorescent dyes enter a flow chamber where cell volume, fluorescence, and light scatter sensors simultaneously measure multiple cellular properties. Cells then emerge in a liquid jet that is broken into uniform liquid droplets. Sensor signals are electronically processed in one of several ways for optimum cell discrimination and are displayed as pulse-amplitude distributions using a multichannel pulse-height analyzer. Processed signals activate cell sorting according to preselected parametric criteria by electrically charging droplets containing cells and electrostatically deflecting them into collection vessels. Illustrative examples of multiparameter cell analysis and sorting experiments using a model mouse tumor cell system, human and animal leukocytes, and cultured mammalian cells are presented.
The development of high-speed, flow-system, cell-sorting methods based on physical and biochemical measurements of single cells has provided a new dimension to biological investigation by being able to physically isolate cells with particular properties from a heterogeneous mixture [l, 2, 31.Such systems have been used to differentiate and sort human leukocytes based on volume measurement [4], to isolate antigen-binding cells which are precursors to antibody-producing cells by use of rmmunofluorescence [5], and to separate abnormal human cells from uterine cervical carcinoma specimens based on UV absorption and visible light scatter [2]. New and improved instrumentation for separating cells based on simultaneous measurement of cell volume, total or two-color fluorescence, and light scatter recently has been developed [6]. This system incorporates several existing analytical techniques, permitting multiple measurements to be per2-741811
formed on the same cell. Coupled with cell sorting, multiparameter analysis methods provide unusual versatility in experimental cell research by measuring and processing several cell characteristics simultaneously and separating cells according to selected combinations of parameters. Cell sorting permits the correlation of machine-measured cell properties with cell morphology and cell sample enrichment (i.e., concentration of cell subclasses). Presented in this report is a description of the cell analysis and sorting techniques made possible with this instrument. Validation of the methodology is discussed with illustrative examples of cell analysis and sorting on the basis of single parameters, ratios, and two parameter combinations of cell volume and fluorescence. Analysis and sorting of (a) SH-thymidine pulse-labeled CHO cells and mouse tumor cells on the basis of DNA content; (b) dog leukocytes Exptl Cell Res 84 (1974)
16 Steinkamp et al.
Coulter Volume
Current
Source
Sensor
I
8 Preamp
Multiparameter Signal Processing Electronics - 1
‘3 &gigI
/
Sample Collection
-
Fig. 1. Multisensor cell separator system diagram illustrating the flow chamber, laser illumination, cell sensors, signal processing and cell separation electronics, and droplet charging and deflection scheme. The piezoelectric transducer used for droplet generation is not shown.
into subclassesof lymphocytes, neutrophils, and eosinophils according to cytoplasmic granulation; (c) human lymphocytes into two subclassesbased on the degree of cytoplasmic granulation density; and (d) HeLa cells based on DNA-cell volume relationships are presented. MATERIALS AND METHODS Instrumentation A comprehensive description of the instrumentation has been described elsewhere [6]. Cells stained with fluorescent dyes that label specific biochemical components are suspended in normal saline and introduced into the flow chamber (fig. 1) at approx. 500 cells/set. Cells first pass centrally through a 75 ,um diameter volume-sensing orifice. (Coulter principle) and then across a fluid-filled viewing region intersecting an argon laser beam (488 nm wavelength), causing fluorescence and light scatter. Both are Exptl Cell Res 84 (1974)
electro-optically measured, the fluorescence sensor being a dual photomultiplier tube array which measures total (above 520 MY wavelength) or two-color fluorescence of selectable color separation regions. Light scatter can also be measured by optically focusing the forward scattered light onto a photodiode. After optical measurement, the liquid stream carrying suspended cells emerges into air as a jet and is broken into uniform droplets (45 OOO/sec).Thus, cells are isolated into droplets [l] with approx. 1 % of the droplets containing a cell. Parametric signals from the cell sensor amplifiers (fig. 1) are processed on a cell-by-cell basis as single parameters, ratios (analog pulse divider), and gated single parameters. A multichannel pulse-height analyzer accumulates and displays processed signals as pulse-amplitude frequency distribution histograms. Gated single parameter analysis permits the examination of particular subclasses of cells within selectable ranges on different cellular property values. For example, the volume distribution for Gl cells can be obtained by analysing only those processed volume signals from cells yielding Gl DNA content as determined by fluorescence measurement. That is, if the DNA content signal falls within the Gl DNA content range as determined by adjustable signal amplitude gates, only then will the associated volume
Multiparameter signal be accumulated and displayed. Gated single parameter techniques also permit distributions to be obtained from weakly fluorescing signals (e.g., fluorescent antibody) by requiring electronic coincidence with cell volume or light scatter signals. In this manner, fluorescence signals from debris is eliminated. Complete two parameter analysis is available by selecting two processed signals as inputs to a dual parameter pulse-height analyzer and displaying subsequent three-dimensional pulse-height distributions and two-dimensional contour views. Processed signals activate cell sorting if the amplitudes fall within preselected ranges of single-channel, pulse-height analyzers (SCAs) located within the cell separation logic. An electronic time delay is activated which then triggers a positive or negative droplet charging pulse, causing a group of droplets containing a-cell-to be charged and deflected by a static electric field into a collection vessel. Those cells not satisfying the sorting criteria pass undefleeted into a separate collection vessel. The sorted suspension is introduced into a cytocentrifuge (Shandon Scientific Co., Sewickley, Pa) and deposited onto microscope slides for counterstaining and microscopic examination. Some experiments require sorted cells to be reanalyzed and separated again.
Cell staining and dispersal Chinese hamster cells (line CHO) grown in spinner culture in F-10 medium supplemented with 10 % calf and 5 % fetal calf sera were pulse-labeled with 5 ,&i/ ml of aH-thymidine for 15 min, dispersed with trypsin, fixed in 70% ethanol for 0.5 h, treated with RNase, and stained with propidium iodide [7]. Stained cells were then suspended in normal saline for fluorescence analysis of DNA content and sorting. Sorted samples were subjected to autoradiography. Mouse spleen and a methylcholanthrene-induced squamous cell mouse tumor (MCA-1) grown in C3H/Hej mice were minced, trypsinized [8], and then dispersed. Cells were filtered by passing through a nylon mesh filter (80 pm pore size), fixed in 10% formalin for 12-18 h, stained with acriflavine using the fluorescent-Feulgen DNA staining procedure [9], and suspended in normal saline for fluorescence analysis of DNA content and subsequent sorting. The sorted cells were then counterstained using the standard Pananicolaou method. Fresh whole blood from normal humans and dogs (beagles) containing 0.01 ml of 10% EDTA/ml of blood was prepared for fluorescence analysis and sorting by diluting one part blood with 25 oarts of acrid&e -orange golution [lo]. Stained leukocytes exhibit a uniform green nucleus with distinct red cytoplasmic granules as viewed with the fluorescence microscope. Erythrocytes do not fluoresce at this dye concentration. In these measurements, the twocolor fluorescence sensor was set to measure zreen 520-580 nm) and red (590-800 nm) fluoresc&ce. Sorted leukocytes were fixed in 100 % methanol and counter-stained using Wright-Giemsa. HeLa cells (line S3), grown in double-strength BME supplemented with 20% fetal calf serum on monolayer, were dispersed with trypsin, fixed in
cell sorting
1I
formalin, and stained with acriflavine for DNA as described above. Cells were resuspended in normal saline for analysis of cell volume and DNA content and for demonstration of two parameter cell sorting capability.
RESULTS AND DISCUSSION for CHO The DNA content distribution cells pulse-labeled for 15 min with 3Hthymidine is shown in fig. 2. The first peak represents Gl cells and the second peak G2 +M cells, having twice the DNA content. Cells synthesizing DNA (S phase) lie in the region between the peaks. Equal numbers of cells sorted on the basis of Gl, S, and G2 +M DNA content, designated by sort regions 1, 2, and 3, respectively, were examined autoradiographically. The percentage of labeled cells in each fraction is shown in fig. 2. The cell population had a total labeling index of 33 %. The small percentage of labeled cells in the sorted Gl fraction is due to early S cells, whereas those labeled cells in the G2+M sorted fraction are late S
)r
10,00c
Fig. 2. Abscissa: channel no. (proportional to cellular DNA content); ordinate: no. of cells. DNA content freauencv distribution histomam of CHO cells pulse-labeled for 15 mm with $H-th-&dine and stained with propidium iodide. The dotted line is an approximation of a second-degree polynomial computer curve-fit of cells in S phase [8, 111. Exptl Cell Res 84 (1974)
18 Steinkamp et al.
10%
(
,
,
,
T
,
[
,
,
’ 60
’
’ SO
’
I (4
100
104-
OO
20 -q Sort1
tel-"OOSort
2
100 L
Fig. 3. Abscissa: channel no. (proportional to cellular DNA content)+;ordinate: no. of cells. Frequency distribution histograms of acriflavine-Feulgen DNA stamed mouse (C3H/Hej) spleen and MCA-1 tumor cells: (A) DNA content distribution of normal spleen cells; (I?) DNA content distribution of MCA-1 tumor cells; (C) photomicrographs of dispersed tumor cells prior to sorting; (D) sorted mouse leukocytes of sort region 1; (E) sorted tumorigenic cells of sort region 2. Both DNA content distributions of (A) and (I?) were measured at the same fluorescence channel amplifier gain setting.
cells. These techniques currently are being used to study the nature of cell-cycle kinetics in populations of human diploid cells maintained in low serum-containing medium [12]. Fig. 3 shows the DNA content distributions for normal mouse spleen and methylcholanthrene-induced mouse tumor (MCA-1) cells. The mouse spleen cell DNA distribution (fig. 3A) provides an indication for normal 2C diploid DNA content. The DNA distribution of MCA-1 tumor cells (fig. 3B) shows three peaks, the first peak being cells having a normal 2C diploid DNA content and the second and third peaks representing cells having 4C and 8C DNA contents, respectively. The tumor cell population was sorted on the basis of DNA content into Exptl Cell Res 84 (1974)
two groups, designated by sort regions 1 and 2 (fig. 3 B). Fig. 3 C is a photomicrograph of dispersed and counterstained tumor cells prior to sorting. Cells having normal DNA content from sort region 1 are leukocytes (fig. 3 D), whereas cells containing an elevated or abnormal DNA content from sort region 2 are tumorigenic cells (fig. 39, the two subpopulations being morphologically distinct. The second and third peaks of fig. 3 B are Gl and G2+M phase tumor cells, with S phase cells contained between the peaks. These results have been verified for MCA-1 tumor cells cultured in vivo and in vitro [13]. Other tumor cells with non-elevated DNA content most certainly will not be identifiable on the basis of DNA alone. Other para-
Multiparameter
cell sorting
19
Fig. 4. Abscissa: channel no. (proportional to cellular fluorescence); ordinate: no. of cells. Frequency distribution histograms of acridine orange-stained normal dog (beagle) leukocytes: (A) green fluorescence distribution; (B) red fluorescence distribution; (C) photomicrographs of blood smear prior to sorting; (0) sorted lymphocytes; (I?) sorted neutrophils; (F) sorted eosinophils. Erythrocytes contained in the photomicrographs were separated along with selected leukocytes.
meters which characterize cell abnormalcy are under investigation and include DNA, protein and cell volume relationships, nuclear-to-cytoplasmic ratios by laser beam scanning, and others such as antigenic properties of tumorigenic cells.
The red and green fluorescence pulseamplitude distributions of normal dog (beagle) leukocytes vitally stained with acridine orange are shown in fig. 4. The green fluorescence distribution (fig. 4A) is unimodal, illustrating uniform nuclear staining, whereas Exptl Cell Res 84 (1974)
20 Steinkamp et al.
Fig. 5. Abscissa: channel no. (proportional to cellular fluorescence); ordinate: no. of cells. Frequency distribution histograms of acridine orange-stained normal human leukocytes: (A) 0, green; l , red fluorescence distributions; (B) green-to-red fluorescence ratio distribution; (C) photomicrographs of blood smear prior to sorting; (0) sorted large lymphocytes; (E) sorted small lymphocytes. Erythrocytes contained in the photomicrographs were separated along with selected leukocytes.
the red fluorescence distribution (fig. 4B) shows three distinct peaks, characterizing dog leukocytes primarily into 3 subpopulations on the basis of cytoplasmic granules which exhibit red fluorescence. Leukocytes having red fluorescence signal amplitudes corresponding to sort regions 1, 2, and 3 (fig. 49 were separated, counterstained, and identified by microscopic examination as principally consisting of lymphocytes, neutrophils, and eosinophils, respectively. Fig. 4C-F shows photomicrographs of the blood smear and sorted leukocyte fractions. Eosinophils are characterized by coarse cytoplasmic granules, compared with finer (dust-like) granulation in lymphocytes and neutrophils, as exemplified in the red fluorescence distribution. Exptl Cell Res 84 (1974)
The green and red fluorescence distributions of normal human leukocytes vitally stained with acridine orange are shown in fig. 5A. The green fluorescence distribution again is unimodal, with the red fluorescence distribution indicating three distinct leukocyte subpopulations (peaks l-3), having been identified as lymphocytes, monocytes, and granulocytes, respectively [lo, 141. The green-to-red fluorescence ratio distribution (fig. 5B), which shows four distinct subpopulations (peaks l-4), permits leukocyte characterization as a function of nuclear-tocytoplasmic staining properties. Leukocytes having green-to-red fluorescence ratio signal amplitudes corresponding to sort regions 1, 2, 3, and 4 (fig. 5B) were separated, counterstained, and identified as granulocytes,
Multiparameter
cell sorting
21
6,000
Fig. 6. Abscissa: channel no.; ordinate: no. of cells. Frequency distribution histograms of acriflavine-Feulgen DNA stained HeLa cells: (A) l , DNA content; 0, cell volume distributions. (II) Cell volume distributions of l , Gl; a, G2 +M cells as obtained by gated single parameter analysis. (C) Two parameter DNA-cell volume plots illustrating three-dimensional isometric and two-dimensional contour views of cells prior to sorting. (D) Cells reanalyzed after sorting. The contour views were recorded near the zero cell number level.
monocytes, and two subclasses of lymphocytes, respectively. Fig. 5 C-E shows photomicrographs of the blood smear and sorted lymphocyte fractions 3 and 4. Preliminary investigations on sorted lymphocytes subjected to volume measurement indicate that fractions 3 and 4 are classesof large and small lymphocytes, respectively, with some overlap of the volume distributions. Since nuclear staining is approximately uniform, the greento-red fluorescence ratio measurement allows lymphocytes to be classified as to the degree of cytoplasmic granulation density. This technique provides possible new methodology for study of lymphocyte kinetics with respect to immune response. Cell analysis and sorting on the basis of two parameters are demonstrated using acriflavine-Feulgen stained HeLa cells. Single parameter DNA content and cell volume distributions were first recorded (fig. 6A) as was the DNA-cell volume two parameter
plot (fig. 6C). The DNA content distribution for exponentially grown HeLa cells is similar to the CHO cell DNA distribution (fig. 2). The volume distribution of fixed and stained HeLa cells is broad, unimodal, and typical of cells growing exponentially [15]. Fig. 6 C shows a three-dimensional DNA-cell volume distribution (isometric view) and a twodimensional contour view which is a baseplane section of the isometric display. The contour display illustrates the correlation of DNA content and cell volume around the cell cycle, with additional fine structure capable of being observed with this type of analysis. Two parameter analysis also provides a method to quantitate the number of cells within a given range of two cellular properties. Recent two-color fluorescence analysis of DNA and protein content relationships in doubly stained HeLa and CHO cells [6, 71 compare favorably with DNA-cell volume measurements. Exptl Cell Res 84 (1974)
22
Steinkamp et al.
Using gated single parameter analysis methods, the individual Gl and G2+M cell volume distributions were next recorded (fig. 6B) by analyzing the volume of cells having Gl and G2+M DNA content, respectively (channels 25-33 and 58-70 of the DNA distribution). Equal numbers of cells were then sorted into two groups, designated by sort regions 1 and 2 (fig. 6B), resulting in small volume Gl and large volume G2-l-M cells being collected. The two subpopulations were mixed together and rerun through the cell sorter, recording the DNA-cell volume two parameter distribution (fig. 60). The two parameter DNA-cell volume distribution of the combined sorted populations vividly illustrates this method of cell sorting and provides additional capability to measure quantitatively and characterize cell properties on the basis of two or more parameters. We have presented several examples of analysis and sorting of mammalian cells on the basis of single and two parameter combinations of cell volume and total or twocolor fluorescence. Multiparameter signal processing coupled with cell sorting provides a versatile and extremely useful method for cell biology research with the range of applications expanding as new methods of obtaining information on cells become available. Many other combinations of measurements are possible with the present instrumentation. The briefly described flow chamber contains a Coulter cell volume sensor coupled with fluorescence and light scatter optical sensors. Both Coulter and light scatter sensors yield cell size information. We have determined experimentally that the Coulter channel output is directly proportional to cell volume over allarger range of cell sizes than small angle light scatter [6], laser beam shape, wavelength, and cell diameter having important relationships in cell Exptl Cell Res 84 (1974)
size determinations [16]. Although cell fixation reduces the Coulter volume signal, we find only minimal cell volume distribution distortion as measured on cultured cells [7]. An additional electronic circuit has been constructed which converts cell volume signals to the two-thirds power, providing methodology for cell surface area determinations [17]. In principle, other cellsensing and analysis methods may give meaningful information on physical and biochemical cellular characteristics as both instrumental development and biological applications are in their initial stages of growth. This work was performed at the Los Alamos Scientific Laboratory, Los Alamos, New Mexico, under joint sponsorship of the US AEC and the NCI. We thank Mrs Julia Grilly for photomicrography of sorted cells, L. M. Holland, D. V. M., and Mrs J. E. London for obtaining blood samples, and Mrs P. C. Sanders for autoradiography. MC&-l tumor was obtained from the Laboratory of Pathology, NCI, NlH, Bethesda, Md, USA.
REFERENCES 1. Fulwyler, M J, Science 150 (1965) 910. 2. Kamentsky, L A & Melamed, M R, Science 156 (1967) 1364. 3. Hulett, H R, Bonner, W A, Barrett, J & Herzenberg, L A, Science 166 (1969) 747. 4. Van Dilla, M A, Fulwyler, M J & Boone, I U, Proc sot exptl biol med 125 (1967) 367. 5. Julius, M H, Masuda, T & Herzenberg, L A, Proc natl acad sci US 69 (1972) 1934. 6. Steinkamp, J A, Fulwyler, M J, Coulter, J R, Hiebert, R D, Homey, J L & Mullaney, P F, Rev sci instr 44 (1973) 1301. 7. Crissman, H A & Steinkamp, J A, J cell biol 59 (1973) 766. 8. Kraemer, P M, Deaven, L L, Crissman, H A & Van Dilla. M A. Advances in cell and molecular biology (ed E’ DuPraw) vol. 2, pp 47-108. Academic Press. New York and London (1972). 9. Tobey, R A, Crissman, H A & Kraemer,‘P M, J cell biol 54 (1972) 638. 10. Adams, L R & Kamentsky, L A, Acta cytol 15 (1971) 289. 11. Dean, P N & Jett, J H, J cell biol(1974). In press. 12. Dell’Orco, R, Crissman, H A, Steinkamp, J A & Kraemer,.P M. In preparation. 13. Horan, P K, Romero, A, Steinkamp, J A 8~
Multiparameter cell sorting 23 Petersen, D F, J natl cancer inst US (1974). In press. 14. Steinkamp, J A, Romero, A & Van Dilla, M A, Acta cytol 17 (1973) 113. 15. Anderson, E C, Bell, G I, Petersen, D F & Tobey, R A, Biophys j 9 (1969) 246.
16. Mullaney, P F & Dean, P N, Biophys j 10 (1970) 764. 17. Steinkamp, J A, Larkins, J H, Carr, L J & Kraemer, P M. In preparation. Received September 3, 1973
Exptl Cell Res 84 (1974)