- Email: [email protected]
An improved flow microfluorometer for rapid measurement of cell fluorescence
Copyright All rights
6 1973 by Academic Press, Inc. of reproduction in any form reserwd
Experimental Cell Research 80 (1973) 105-110
AN
IMPROVED
FLOW
MEASUREMENT
MICROFLUOROMETER OF CELL
FOR
RAPID
FLUORESCENCE
D. M. HOLM and L. S. CRAM Biomedical Research Group, Los Alamos Scientific Laboratory, University of California, Los Alamos, N. Mex. 87544, USA
SUMMARY An improved instrument for the making of high-speed fluorescent measurements on single cells has been constructed and characterized. Instrumental optimization was accomplished with Chinese hamser cells (line CHO) stained by the fluorescent Feulgen procedure using either auramine- or acriflavine. Contributions of instrumental resolution and artifacts to cellular DNA distributions have been determined. The instrument introduces a total coefficient of variation of less than 2 % for the case of CHO cells stained with bright dyes such as acriflavine, where photon-statistical effects are minimal. The design of the instrument is such that any of a number of cellular constituents or properties can be studied and quantitated on a single-cell basis at a rate of 50000 cells/min.
The advent of flow systems for the making of high-speed fluorescent measurements on single cells [l-5] has elicited new biological information on the relationship between DNA content per cell and chromosome number [6, 71 and in the area of immunofluorescence [8, 91. As a result of such biological experiments, many researchers are now using the flow-system technique in a wide variety of biological studies. Since the instruments of this type are in their early stages of evolution, it is important to understand performance characteristics, limitations, and possible artifacts in the data that would lead to erroneous conclusions. The purpose of this paper is to describe our flow microfluorometer (FMF) for high-speedfluorescence measurements and to show what factors influence the quality of data derived from it. Armed with this information, other experimenters can better evaluate their own instruments and data.
We estimate that, with bright dyes such as acriflavine, this instrument introduces a total coefficient of variation (defined as the S.D. divided by the mean; assuming a normal distribution, it is calculated by dividing the full width at half maximum by the modal channel times l/2.35) of less than 2%; therewith, details of small variations in fluorescence can be observed. It is possible to quantitate a large number of physical/cellular parameters for which there is a known relationship between dye binding and quantity being measured. We have found FMF useful for basic research in biology and medicine and potentially useful for disease diagnosis by immunofluorescence methods. The instrument considered here is an improvement of that described by Van Dilla et al. [l] and has been used for nearly 2 years in routine biological experiments. Single-cell suspensionsare stained with a fluorescent dye specific for a cellular component. The cells Exptl Cell Res 80 (1973)
106 D. M. Holm & L. S. Cram
Fig. 1. Schematic diagram of the flow microfluorometer (FMF). The argon-ion laser beam (lower righthand corner) is focused by 20 cm (placed at 28 cm from the cell stream) and 2 cm focal length cylind&al lenses to an ellipt~al cross section 7.5 x 100 ym (height and width of the la&r beam-at half the maximum intensity) at intersection with the sample stream.
are caused to pass through a laser beam which excites the dye and causes a flash of fluorescent light proportional to the amount of dye present in the cell. This flash of light is viewed by a photomultiplier tube and converted into a voltage pulse whose amplitude is proportional to the amount of light emitted by cell and, hence, to the amount of dye per cell. A histogram showing the number of cells as a function of cell fluorescence is accumulated on a pulse-height analyzer. Analysis of the pulse-height distribution enables the experimenter to determine the quantitative distribution of dye within the cell population. From the distribution and relationship between the dye and cellular material binding the dye, the experimenter can deduce the quantities of interest. MATERIALS AND METHODS A schematic diagram of flow microfluorometer Model II (FMF II) is shown in fig. 1. The cell suspension is introduced through a hypodermic tube of 0.5 mm inside diameter at the bottom of the flow cell. A distilled water sheath flow surrounds the sample tube, at the end of which the two flows merge under laminar flow conditions. The combined flow at this point is 3.2 mm in diameter. While maintaining laminar flow conditions, the combined flow is reduced in diameter to 250 ,um by a nozzle, from which it exits into a quiescent water volume where it Exptl Cefl Res 80 (1973)
intersects the laser beam. The combined flow leaves the chamber through a 250 pm orifice and is collected in a water traD on a 360-Torr vacuum system. An argon-ion laser ICoherent Radiation Model 54) is used as the source of blue exciting light. The exciting wavelength is 457 or 476 nm for cells stained by the auramine-O/Feulgen procedure and 488 nm for cells stained with the acriflavine/Feulgen procedure or with fluorescein. The laser beam is focused to an elliptical cross section at the cell stream by two cylindrical lenses of focal length 2 and 20 cm. Laser beam power is monitored with a Jodon Model 450A. Optimum laser power at the front of the laser is 50 mW at 457 nm, 135 mW at 476 nm, and 355 mW at 488 nm. The fluorescent signal is collected with a projection lens having f/1.6 optics. Following the first projector lens, where the light is parallel, is a Corning glass filter (No. 3-70 for 457 or 476 nm excitation, No. 3-69 for 488 nm excitation) to block scattered light. A similar second projection lens focuses the fluorescent light on a 200 pm pinhole. After passing through the pinhole, the light is made parallel by a small collimating lens. A prism is used to reflect the light on the photocathode of the RCA 4526 photomultiplier. The optical system utilizes darkfield illumination and has a magnification of 1. Thus, only light generated in the region of the intersection of the laser beam and cell stream is efficiently collected by the photomultiplier. A viewing window with a blocking filter to cut out the blue laser light is located opposite the photmultiplier and is used for alignment of the system. For normal alignment the only adjustment necessary is sweeping the laser beam horizontally across a fluorescent dye stream. This is done by positioning the 20 cm focal length cylindrical lens such that photomultiplier anode current is maximum. The result is that the cell stream intersects the central maximum of the laser beam. Since the laser beam normally has an approximately Gaussian intensity distribution, it is necessary to make certain that the cell stream intersects the peak intensity of the laser beam. The laser beam was shaped into an ellipse so that
Flow microfluorometry each cell would be illuminated in the same way and the total fluorescent light pulse would be short. For FMF II, the major axis of the ellipse has a full width at half the maximum light intensity of about 100 pm, while the minor axis of the ellipse has a full width at half the maximum intensity of about 7.5 ,um. These dimensions are obtained for the major and minor axes, respectively, with a 20 cm focal length cylindrical lens 28 cm from the cell stream and a 2 cm focal length cylindrical lens focused on the cell stream. The 20 cm lens is defocused to increase the major axis dimension. It is possible to discriminate between two cells stuck together and one cell having twice as much DNA. This is important in making accurate measurements of the DNA distribution in cell populations for life cycle analysis or percent tetraploidy. Discrimination is accomnlished with a beam minor axis smaller than the nuclear diameter and pulse-shape discrimination. Alignment of two cells takes place in the flow system; therefore, two peaks separated by a saddle are observed when two attached cells cross the laser beam. For cells at time intervals greater than 4 psec (about 15 pm apart), a complete discrimination is made. Since the minor axis of the laser beam is of the order of one nuclear diameter, it is necessary to integrate the electrical pulse produced by the passage of a cell through the laser beam. Pulse integration is accomplished either by RC integration mainly in the preamplifier (Los Alamos Model 225V) or by an active pulse integrator (Los Alamos Model 8082); both yield equivalent results. Integration in the preamplifier uses an integrating time constant of about 6 psec, which is long compared to the transit time of cells through the laser beam. Fidelity of the integrated pulse is maintained through the amplifier (Camberra Model 1415). The amplifier time constant is set at 1.25 ,usec, the first differentiator is set at 5 ,usec, and the second differentiator is turned off. When active pulse integration is used, the preamplifier time constant is adjusted to 0.2 ,ksec, and electrical pulse shape follows the light pulse shape until integration. Good fidelity is also necessary if pulse-shape analysis is to be performed. The integrated pulses are analyzed by a pulse-height analyzer (Victoreen SCIPP Model 1600). For bright cells stained by the acriflavine/Feulgen procedure, typical amplifier gain is about 100 and photomultiplier gain 300.
RESULTS Considerable progress has been made in understanding and minimizing errors or artifacts introduced by the instrumentation. Fig. 2 shows a pulse-height distribution from exponentially growing Chinese hamster (CHO) cells. These cells were stained with acriflavine using the Feulgen procedure for DNA [lo]. The abscissa is proportional to cellular
107
Fig. 2. Abscissa: channel no. (proportional to cell fluorescence); ordinate: no. of cells. Fluorescence distribution of acriflavine/Feulgen/ DNA-stained Chinese hamster ovary cells growing asynchronously in suspension culture.
fluorescence, and the ordinate is the number of cells having that amount of fluorescence. The large peak at the left corresponds to cells in the GI state of their life cycle. The coefficient of variation of this peak is 3.3 %. As will be shown below, approximately half of the coefficient of variation of this peak is caused by a number of factors including variation introduced by the staining procedure, optical absorption effects, intrinsic variation in DNA per cell, and variation in laser beam intensity as a function of position and as a function of time. A series of experiments were performed to identify and measure the contributions Exptl Cell Res 80 (1973)
10% D. M. Holm & L. S. Cram
,LI
m/(,aA,,
, IO
,~ ~_~I ,-;\,
,,I 100
Fig. 3. Abscissa:
laser beam power (mW); ordinate: (left) coefficient of variation ( 06); (v&+iht)cell brightness.(arbitrary scale). (A) Brightness of cells as a function of laser beam power at I = 457 nm, showing no dye saturation effects; (B) Coefficient of variation of the Gl peak of CHO cells stained by the auramine-O/Feulgen procedure as a function of laser beam power, indicating that statistical effects for these cells are significant at lower power and that laser beam instability is significant at higher power. Curves A and B were obtained using a higher power laser and different laser beam dimensions than used on FMF II. (C) Coefficient of variation of a photodiode light flasher of brightness equal to stained cells at the specified laser beam power. The straight line is that expected from statistical effects and indicates that these effects are important with this stain.
to the coefficient of variation of the Gl peak. The coefficient of variation is a good measure of the quality of the instrument, if there is reason to believe that cells do not have significant variation in DNA content. This is the case for CHO cells. For this reason, they have been used along with uniform fluorescent plastic microspheres [ 1I] to monitor instrumentation effects. One of the first considerations for optimizing the instrument was to measure the fluorescent emission from cells as a function of laser power. The purpose of this experiment was to see if there was any evidence of dye saturation by the intense blue laser light. Curve A of fig. 3 shows the fluorescent emission from CHO cells stained by the auramine-O/Feulgen procedure as a function Exptl Cell Res 80 (1973)
of laser beam power. These experiments were done with an experimental FMF system using a more powerful laser (Coherent Radiation Model 52G) which has several times the power of the Model 54, and different optics were used for beam shaping. Since the data points fall on a 45” line on loglog paper, dye saturation is not indicated. To determine the effect of photon statistics on the coefficient of variation of the Gl peak, fluorescent intensity distributions were measured as a function of laser power. CHO cells stained by the auramine-O/Feulgen procedure were used because these effects are larger with less bright dyes. The coefficient of variation of the Gl peak was calculated and is plotted in fig. 3 (curve B) as a function of laser power. A light flasher was introduced in the optical path so that pulses of an amplitude equal to that of Gl cells could be generated. The coefficient of variation of the flasher peak is also shown as a function of laser power in fig. 3 (curve C). The most
I 300
400
500
600
Fig. 4. Abscissa: wavelength (nM); ordinate:
rel. intensity. Excitation and emission spectra of auramine-O/ Feulgen stained Chinese hamster cells (line CHO) as measured with an Aminco-Bowman spectrophotofluorometer. Extrapolated points on the excitation curve (U-U) were obtained from experiments using available argon-ion laser excitation wavelengths.
Flow microfluorometry significant feature of this plot is that of the flasher slope. The straight line drawn through those points has the slope expected for purely statistical fluctuations in N, the number of photoelectrons emitted at the photocathode of the photomultiplier tube. The coefficient of variation from statistical fluctuations is 1%/N= l/l%. Thus, for auraminestained cells, the light signal reaching the photomultiplier is weak enough (about 1 200 photoelectrons per cel1 with FMF II, as calculated from the flasher coefficient of variation of 2.9~4) to make statistical fluctuation a significant factor in the coefficient of variation. It will be noted that cells follow the same slope at lower power but that, at higher power, they deviate significantly from this slope. This deviation is probably due to increased laser beam power fluctuations. The absolute difference between the flasher curve and cell curve is due to other factors referred to above. Electronic effects seem to be be insignificant. During these experiments the laser beam was not monitored for fast intensity fluctuations; however, subsequent experiments showed that instabilities do occur, and corrective measures are being taken to minimize this effect. Statistical effects can also be reduced by using optics with greater light-collecting power. The flow chamber can accommodate f/l optics which, if used, would improve light collection by approximately a factor of 2; the current f/1.6 optics have a solid angle of about 2.2%. The RCA 4526 photomultiplier was chosen because it had a high (25 “/o) quantum efficiency; photodiodes have higher quantum efficiency but are noisier. Recent use of the acriflavine/Feulgen procedure yielded much brighter cells and reduced the statistical effects considerably so that the total coefficient of variation of the Gl peak is between 3 and 476.
[ :
INDIVIOUAL POPULAT ,ONS cvA=44% cvB=3 6 ‘6
1000
A
109
MIXTURE A 50% B 50% CV= 6.0%
B
500
0 y:
45
50
55
+
Fig. 5. Abscissa: channel no. (proportional to microsphere fluorescence); ordinate: no. of microspheres. Fluorescence distributions of uniform fluorescent plastic microsphere populations: two individual populations (A) and (B); a 50:50 mixture of (A) and (II); and a 80:20 mixture of (A) and (B).
Fig. 4 shows the excitation and emission curves of auraminebound to DNA of CHO cells. Since the absorption maximum is at about 440 nm, it is obvious that the rather weak line at 457 nm is not the optimum excitation wavelength. However, in spite of the fact that the 457 nm line is lower in power than the 488 or 476 nm lines, it gives a signal that is equal to or greater than that obtained with the longer wavelength lines of higher intensity. The signal can also be increased by using a blocking filter which transmits maximum fluorescence. Since it is necessary to block out scattered blue light, use of dyes whose absorption and emission bands are widely separated is desirable. Investigations were made of the effect of cell stream diameter and non-uniformity of laser beam illumination. For these experiments cells were examined under normal conditions (i.e., 21 ml/min flow rate for the sheath water that surrounds the sample and 0.15 ml/min sample flow volume). These rates correspond to a cell stream diameter of 15 pm at the laser beam intersection. Since this dimension is very nearly equal to a cell diameter, there is very little uncertainty as to cell position in the laser beam. Exptl Cell Res 80 (1973)
110 D. M. Helm & L. S. Cram Keeping the sheath flow constant, sample flow was increased up to 10 times normal, and the coefficient of variation of cells in the G 1 peak was measured. A 4% coefficient of variation was obtained for normal sample flow conditions (cell stream diameter 15 ,um) and only 4.6% coefficient of variation for 10 times that flow (cell stream diameter 45 pm). To eliminate accidental coincidences at the high flow rate, a low cell concentration was maintained. Cells with a small coefficient of variation (4 %) were used so that changes due to non-uniform illumination would be observed easily. Thus, under normal operating conditions, laser beam illumination of the cell stream is very uniform. Counting rates of the order of 1 000 cells/ set are feasible with FMF II. The total transit time of a cell across the laser beam is 2-3 psec, and electronic base-line recovery is within 20 psec. Instrumental quality is further demonstrated by measuring the fluorescence distribution of very uniform 12 pm fluorescent plastic microspheres. Fig. 5 illustrates the capabilities of the instrument to resolve two fluorescence distributions resulting from mixing two populations of microspheres differing in intensity by 10%. The two Gaussian distributions add mathematically, as expected. These results demonstrate the importance of minimizing instruto the width of mental contribution the distribution. Similar experiments indicate that two subpopulations can be resolved if the modal fluorescence intensities differ by at least 5 “/u and if the fraction of one population is at least 10% of the total.
Exptl Cell Res 80 (1973)
This instrument has been in routine use for biological experiments for about two years in a wide variety of applications. It can be operated by a technician and is reliable. It appears that resolution can be improved still further by compensating for laser beam fluctuations as a function of time, and efforts in this direction are underway. Once this compensation has been accomplished, it appears that instrumental effects can be reduced to the neighborhood of 2 % or less for brighter dyes such as acriflavine/Feulgen. We wish to thank Dr M. A. Van Dilla, Mr T. T. Trujillo, and Mr R. D. Hiebert for their assistance. This work was performed under the auspices of the US AEC.
REFERENCES 1. Van Dilla, M A, Trujillo, T T, Mullaney, P F & Coulter, J R, Science 163 (1969) 1213. 2. Kamentsky, L A, Cytology automation (ed D M D Evans) pp 177-185. Livingstone, Edinburgh, London (1970). 3. Schumann, J, Ehring, F, GGhde, W & D&rich, W, Arch klin exptl Derm 239 (1971) 377. 4. Springer, E, Bohn, N & Sandritter, W, Histochemie 26 (1971) 238. 5. Holly, F E, Ph.D. thesis, University of California, Los Angeles (1970). 6. Kraemer, P M, Petersen, D F & Van Dilla, M A, Science 174 (1971) 714. 7. Kraemer, P M, Deaven, L L, Crissman, H A & Van Dilla, M A, Advances in cell biology (ed E Du Praw) vol. 2, pp 47-108. Academic Press, New York and London (1972). 8. Hulett, H R, Bonner, W A, Barrett, J & Herzenberg, L A, Science 166 (1969) 747. 9. Cram, L S & Hensley, J C, Biophysical society abstr 12 (1972) 144~. Abstract SaPM-A12. 10. Trujillo, T T & Van Dilla, M A, Acta cytol 16 (1972) 26.
II. Cram, L S, Fulwyler, M J & Perrings, J D, Biophysical society abstr (1971) 155~. Abstract WPM-G2. Received November 27, 1972
6 1973 by Academic Press, Inc. of reproduction in any form reserwd
Experimental Cell Research 80 (1973) 105-110
AN
IMPROVED
FLOW
MEASUREMENT
MICROFLUOROMETER OF CELL
FOR
RAPID
FLUORESCENCE
D. M. HOLM and L. S. CRAM Biomedical Research Group, Los Alamos Scientific Laboratory, University of California, Los Alamos, N. Mex. 87544, USA
SUMMARY An improved instrument for the making of high-speed fluorescent measurements on single cells has been constructed and characterized. Instrumental optimization was accomplished with Chinese hamser cells (line CHO) stained by the fluorescent Feulgen procedure using either auramine- or acriflavine. Contributions of instrumental resolution and artifacts to cellular DNA distributions have been determined. The instrument introduces a total coefficient of variation of less than 2 % for the case of CHO cells stained with bright dyes such as acriflavine, where photon-statistical effects are minimal. The design of the instrument is such that any of a number of cellular constituents or properties can be studied and quantitated on a single-cell basis at a rate of 50000 cells/min.
The advent of flow systems for the making of high-speed fluorescent measurements on single cells [l-5] has elicited new biological information on the relationship between DNA content per cell and chromosome number [6, 71 and in the area of immunofluorescence [8, 91. As a result of such biological experiments, many researchers are now using the flow-system technique in a wide variety of biological studies. Since the instruments of this type are in their early stages of evolution, it is important to understand performance characteristics, limitations, and possible artifacts in the data that would lead to erroneous conclusions. The purpose of this paper is to describe our flow microfluorometer (FMF) for high-speedfluorescence measurements and to show what factors influence the quality of data derived from it. Armed with this information, other experimenters can better evaluate their own instruments and data.
We estimate that, with bright dyes such as acriflavine, this instrument introduces a total coefficient of variation (defined as the S.D. divided by the mean; assuming a normal distribution, it is calculated by dividing the full width at half maximum by the modal channel times l/2.35) of less than 2%; therewith, details of small variations in fluorescence can be observed. It is possible to quantitate a large number of physical/cellular parameters for which there is a known relationship between dye binding and quantity being measured. We have found FMF useful for basic research in biology and medicine and potentially useful for disease diagnosis by immunofluorescence methods. The instrument considered here is an improvement of that described by Van Dilla et al. [l] and has been used for nearly 2 years in routine biological experiments. Single-cell suspensionsare stained with a fluorescent dye specific for a cellular component. The cells Exptl Cell Res 80 (1973)
106 D. M. Holm & L. S. Cram
Fig. 1. Schematic diagram of the flow microfluorometer (FMF). The argon-ion laser beam (lower righthand corner) is focused by 20 cm (placed at 28 cm from the cell stream) and 2 cm focal length cylind&al lenses to an ellipt~al cross section 7.5 x 100 ym (height and width of the la&r beam-at half the maximum intensity) at intersection with the sample stream.
are caused to pass through a laser beam which excites the dye and causes a flash of fluorescent light proportional to the amount of dye present in the cell. This flash of light is viewed by a photomultiplier tube and converted into a voltage pulse whose amplitude is proportional to the amount of light emitted by cell and, hence, to the amount of dye per cell. A histogram showing the number of cells as a function of cell fluorescence is accumulated on a pulse-height analyzer. Analysis of the pulse-height distribution enables the experimenter to determine the quantitative distribution of dye within the cell population. From the distribution and relationship between the dye and cellular material binding the dye, the experimenter can deduce the quantities of interest. MATERIALS AND METHODS A schematic diagram of flow microfluorometer Model II (FMF II) is shown in fig. 1. The cell suspension is introduced through a hypodermic tube of 0.5 mm inside diameter at the bottom of the flow cell. A distilled water sheath flow surrounds the sample tube, at the end of which the two flows merge under laminar flow conditions. The combined flow at this point is 3.2 mm in diameter. While maintaining laminar flow conditions, the combined flow is reduced in diameter to 250 ,um by a nozzle, from which it exits into a quiescent water volume where it Exptl Cefl Res 80 (1973)
intersects the laser beam. The combined flow leaves the chamber through a 250 pm orifice and is collected in a water traD on a 360-Torr vacuum system. An argon-ion laser ICoherent Radiation Model 54) is used as the source of blue exciting light. The exciting wavelength is 457 or 476 nm for cells stained by the auramine-O/Feulgen procedure and 488 nm for cells stained with the acriflavine/Feulgen procedure or with fluorescein. The laser beam is focused to an elliptical cross section at the cell stream by two cylindrical lenses of focal length 2 and 20 cm. Laser beam power is monitored with a Jodon Model 450A. Optimum laser power at the front of the laser is 50 mW at 457 nm, 135 mW at 476 nm, and 355 mW at 488 nm. The fluorescent signal is collected with a projection lens having f/1.6 optics. Following the first projector lens, where the light is parallel, is a Corning glass filter (No. 3-70 for 457 or 476 nm excitation, No. 3-69 for 488 nm excitation) to block scattered light. A similar second projection lens focuses the fluorescent light on a 200 pm pinhole. After passing through the pinhole, the light is made parallel by a small collimating lens. A prism is used to reflect the light on the photocathode of the RCA 4526 photomultiplier. The optical system utilizes darkfield illumination and has a magnification of 1. Thus, only light generated in the region of the intersection of the laser beam and cell stream is efficiently collected by the photomultiplier. A viewing window with a blocking filter to cut out the blue laser light is located opposite the photmultiplier and is used for alignment of the system. For normal alignment the only adjustment necessary is sweeping the laser beam horizontally across a fluorescent dye stream. This is done by positioning the 20 cm focal length cylindrical lens such that photomultiplier anode current is maximum. The result is that the cell stream intersects the central maximum of the laser beam. Since the laser beam normally has an approximately Gaussian intensity distribution, it is necessary to make certain that the cell stream intersects the peak intensity of the laser beam. The laser beam was shaped into an ellipse so that
Flow microfluorometry each cell would be illuminated in the same way and the total fluorescent light pulse would be short. For FMF II, the major axis of the ellipse has a full width at half the maximum light intensity of about 100 pm, while the minor axis of the ellipse has a full width at half the maximum intensity of about 7.5 ,um. These dimensions are obtained for the major and minor axes, respectively, with a 20 cm focal length cylindrical lens 28 cm from the cell stream and a 2 cm focal length cylindrical lens focused on the cell stream. The 20 cm lens is defocused to increase the major axis dimension. It is possible to discriminate between two cells stuck together and one cell having twice as much DNA. This is important in making accurate measurements of the DNA distribution in cell populations for life cycle analysis or percent tetraploidy. Discrimination is accomnlished with a beam minor axis smaller than the nuclear diameter and pulse-shape discrimination. Alignment of two cells takes place in the flow system; therefore, two peaks separated by a saddle are observed when two attached cells cross the laser beam. For cells at time intervals greater than 4 psec (about 15 pm apart), a complete discrimination is made. Since the minor axis of the laser beam is of the order of one nuclear diameter, it is necessary to integrate the electrical pulse produced by the passage of a cell through the laser beam. Pulse integration is accomplished either by RC integration mainly in the preamplifier (Los Alamos Model 225V) or by an active pulse integrator (Los Alamos Model 8082); both yield equivalent results. Integration in the preamplifier uses an integrating time constant of about 6 psec, which is long compared to the transit time of cells through the laser beam. Fidelity of the integrated pulse is maintained through the amplifier (Camberra Model 1415). The amplifier time constant is set at 1.25 ,usec, the first differentiator is set at 5 ,usec, and the second differentiator is turned off. When active pulse integration is used, the preamplifier time constant is adjusted to 0.2 ,ksec, and electrical pulse shape follows the light pulse shape until integration. Good fidelity is also necessary if pulse-shape analysis is to be performed. The integrated pulses are analyzed by a pulse-height analyzer (Victoreen SCIPP Model 1600). For bright cells stained by the acriflavine/Feulgen procedure, typical amplifier gain is about 100 and photomultiplier gain 300.
RESULTS Considerable progress has been made in understanding and minimizing errors or artifacts introduced by the instrumentation. Fig. 2 shows a pulse-height distribution from exponentially growing Chinese hamster (CHO) cells. These cells were stained with acriflavine using the Feulgen procedure for DNA [lo]. The abscissa is proportional to cellular
107
Fig. 2. Abscissa: channel no. (proportional to cell fluorescence); ordinate: no. of cells. Fluorescence distribution of acriflavine/Feulgen/ DNA-stained Chinese hamster ovary cells growing asynchronously in suspension culture.
fluorescence, and the ordinate is the number of cells having that amount of fluorescence. The large peak at the left corresponds to cells in the GI state of their life cycle. The coefficient of variation of this peak is 3.3 %. As will be shown below, approximately half of the coefficient of variation of this peak is caused by a number of factors including variation introduced by the staining procedure, optical absorption effects, intrinsic variation in DNA per cell, and variation in laser beam intensity as a function of position and as a function of time. A series of experiments were performed to identify and measure the contributions Exptl Cell Res 80 (1973)
10% D. M. Holm & L. S. Cram
,LI
m/(,aA,,
, IO
,~ ~_~I ,-;\,
,,I 100
Fig. 3. Abscissa:
laser beam power (mW); ordinate: (left) coefficient of variation ( 06); (v&+iht)cell brightness.(arbitrary scale). (A) Brightness of cells as a function of laser beam power at I = 457 nm, showing no dye saturation effects; (B) Coefficient of variation of the Gl peak of CHO cells stained by the auramine-O/Feulgen procedure as a function of laser beam power, indicating that statistical effects for these cells are significant at lower power and that laser beam instability is significant at higher power. Curves A and B were obtained using a higher power laser and different laser beam dimensions than used on FMF II. (C) Coefficient of variation of a photodiode light flasher of brightness equal to stained cells at the specified laser beam power. The straight line is that expected from statistical effects and indicates that these effects are important with this stain.
to the coefficient of variation of the Gl peak. The coefficient of variation is a good measure of the quality of the instrument, if there is reason to believe that cells do not have significant variation in DNA content. This is the case for CHO cells. For this reason, they have been used along with uniform fluorescent plastic microspheres [ 1I] to monitor instrumentation effects. One of the first considerations for optimizing the instrument was to measure the fluorescent emission from cells as a function of laser power. The purpose of this experiment was to see if there was any evidence of dye saturation by the intense blue laser light. Curve A of fig. 3 shows the fluorescent emission from CHO cells stained by the auramine-O/Feulgen procedure as a function Exptl Cell Res 80 (1973)
of laser beam power. These experiments were done with an experimental FMF system using a more powerful laser (Coherent Radiation Model 52G) which has several times the power of the Model 54, and different optics were used for beam shaping. Since the data points fall on a 45” line on loglog paper, dye saturation is not indicated. To determine the effect of photon statistics on the coefficient of variation of the Gl peak, fluorescent intensity distributions were measured as a function of laser power. CHO cells stained by the auramine-O/Feulgen procedure were used because these effects are larger with less bright dyes. The coefficient of variation of the Gl peak was calculated and is plotted in fig. 3 (curve B) as a function of laser power. A light flasher was introduced in the optical path so that pulses of an amplitude equal to that of Gl cells could be generated. The coefficient of variation of the flasher peak is also shown as a function of laser power in fig. 3 (curve C). The most
I 300
400
500
600
Fig. 4. Abscissa: wavelength (nM); ordinate:
rel. intensity. Excitation and emission spectra of auramine-O/ Feulgen stained Chinese hamster cells (line CHO) as measured with an Aminco-Bowman spectrophotofluorometer. Extrapolated points on the excitation curve (U-U) were obtained from experiments using available argon-ion laser excitation wavelengths.
Flow microfluorometry significant feature of this plot is that of the flasher slope. The straight line drawn through those points has the slope expected for purely statistical fluctuations in N, the number of photoelectrons emitted at the photocathode of the photomultiplier tube. The coefficient of variation from statistical fluctuations is 1%/N= l/l%. Thus, for auraminestained cells, the light signal reaching the photomultiplier is weak enough (about 1 200 photoelectrons per cel1 with FMF II, as calculated from the flasher coefficient of variation of 2.9~4) to make statistical fluctuation a significant factor in the coefficient of variation. It will be noted that cells follow the same slope at lower power but that, at higher power, they deviate significantly from this slope. This deviation is probably due to increased laser beam power fluctuations. The absolute difference between the flasher curve and cell curve is due to other factors referred to above. Electronic effects seem to be be insignificant. During these experiments the laser beam was not monitored for fast intensity fluctuations; however, subsequent experiments showed that instabilities do occur, and corrective measures are being taken to minimize this effect. Statistical effects can also be reduced by using optics with greater light-collecting power. The flow chamber can accommodate f/l optics which, if used, would improve light collection by approximately a factor of 2; the current f/1.6 optics have a solid angle of about 2.2%. The RCA 4526 photomultiplier was chosen because it had a high (25 “/o) quantum efficiency; photodiodes have higher quantum efficiency but are noisier. Recent use of the acriflavine/Feulgen procedure yielded much brighter cells and reduced the statistical effects considerably so that the total coefficient of variation of the Gl peak is between 3 and 476.
[ :
INDIVIOUAL POPULAT ,ONS cvA=44% cvB=3 6 ‘6
1000
A
109
MIXTURE A 50% B 50% CV= 6.0%
B
500
0 y:
45
50
55
+
Fig. 5. Abscissa: channel no. (proportional to microsphere fluorescence); ordinate: no. of microspheres. Fluorescence distributions of uniform fluorescent plastic microsphere populations: two individual populations (A) and (B); a 50:50 mixture of (A) and (II); and a 80:20 mixture of (A) and (B).
Fig. 4 shows the excitation and emission curves of auraminebound to DNA of CHO cells. Since the absorption maximum is at about 440 nm, it is obvious that the rather weak line at 457 nm is not the optimum excitation wavelength. However, in spite of the fact that the 457 nm line is lower in power than the 488 or 476 nm lines, it gives a signal that is equal to or greater than that obtained with the longer wavelength lines of higher intensity. The signal can also be increased by using a blocking filter which transmits maximum fluorescence. Since it is necessary to block out scattered blue light, use of dyes whose absorption and emission bands are widely separated is desirable. Investigations were made of the effect of cell stream diameter and non-uniformity of laser beam illumination. For these experiments cells were examined under normal conditions (i.e., 21 ml/min flow rate for the sheath water that surrounds the sample and 0.15 ml/min sample flow volume). These rates correspond to a cell stream diameter of 15 pm at the laser beam intersection. Since this dimension is very nearly equal to a cell diameter, there is very little uncertainty as to cell position in the laser beam. Exptl Cell Res 80 (1973)
110 D. M. Helm & L. S. Cram Keeping the sheath flow constant, sample flow was increased up to 10 times normal, and the coefficient of variation of cells in the G 1 peak was measured. A 4% coefficient of variation was obtained for normal sample flow conditions (cell stream diameter 15 ,um) and only 4.6% coefficient of variation for 10 times that flow (cell stream diameter 45 pm). To eliminate accidental coincidences at the high flow rate, a low cell concentration was maintained. Cells with a small coefficient of variation (4 %) were used so that changes due to non-uniform illumination would be observed easily. Thus, under normal operating conditions, laser beam illumination of the cell stream is very uniform. Counting rates of the order of 1 000 cells/ set are feasible with FMF II. The total transit time of a cell across the laser beam is 2-3 psec, and electronic base-line recovery is within 20 psec. Instrumental quality is further demonstrated by measuring the fluorescence distribution of very uniform 12 pm fluorescent plastic microspheres. Fig. 5 illustrates the capabilities of the instrument to resolve two fluorescence distributions resulting from mixing two populations of microspheres differing in intensity by 10%. The two Gaussian distributions add mathematically, as expected. These results demonstrate the importance of minimizing instruto the width of mental contribution the distribution. Similar experiments indicate that two subpopulations can be resolved if the modal fluorescence intensities differ by at least 5 “/u and if the fraction of one population is at least 10% of the total.
Exptl Cell Res 80 (1973)
This instrument has been in routine use for biological experiments for about two years in a wide variety of applications. It can be operated by a technician and is reliable. It appears that resolution can be improved still further by compensating for laser beam fluctuations as a function of time, and efforts in this direction are underway. Once this compensation has been accomplished, it appears that instrumental effects can be reduced to the neighborhood of 2 % or less for brighter dyes such as acriflavine/Feulgen. We wish to thank Dr M. A. Van Dilla, Mr T. T. Trujillo, and Mr R. D. Hiebert for their assistance. This work was performed under the auspices of the US AEC.
REFERENCES 1. Van Dilla, M A, Trujillo, T T, Mullaney, P F & Coulter, J R, Science 163 (1969) 1213. 2. Kamentsky, L A, Cytology automation (ed D M D Evans) pp 177-185. Livingstone, Edinburgh, London (1970). 3. Schumann, J, Ehring, F, GGhde, W & D&rich, W, Arch klin exptl Derm 239 (1971) 377. 4. Springer, E, Bohn, N & Sandritter, W, Histochemie 26 (1971) 238. 5. Holly, F E, Ph.D. thesis, University of California, Los Angeles (1970). 6. Kraemer, P M, Petersen, D F & Van Dilla, M A, Science 174 (1971) 714. 7. Kraemer, P M, Deaven, L L, Crissman, H A & Van Dilla, M A, Advances in cell biology (ed E Du Praw) vol. 2, pp 47-108. Academic Press, New York and London (1972). 8. Hulett, H R, Bonner, W A, Barrett, J & Herzenberg, L A, Science 166 (1969) 747. 9. Cram, L S & Hensley, J C, Biophysical society abstr 12 (1972) 144~. Abstract SaPM-A12. 10. Trujillo, T T & Van Dilla, M A, Acta cytol 16 (1972) 26.
II. Cram, L S, Fulwyler, M J & Perrings, J D, Biophysical society abstr (1971) 155~. Abstract WPM-G2. Received November 27, 1972