Direct and indirect immunofluorescence analysis of bacterial populations by flow cytometry

Journal of Immunological Methods, 101 (1987) 219-228 Elsevier

219

JIM 04412

Direct and indirect immunofluorescence analysis of bacterial populations by flow cytometry A.P. Phillips, K.L. Martin and A.J. Capey Chemical Defence Establishment, Porton Down, Salisbury, Wiltshire SP4 0JQ, and Department of Microbiology, The Medical School, University of Bristol, Bristol BS8 1TD, U.K. (Received 18 November 1986, revised received 23 February 1987, accepted 23 March 1987)

Bacillus anthracis spores and Escherichia coli were stained with fluorescein-conjugated antibody using direct and indirect methods, then analyzed by means of a commercial flow cytometer. To reduce the cytometer's fluorescence component resulting from unreacted conjugate, reaction mixtures were either diluted or were centrifuged through a sucrose solution using a moving zone technique. Evidence is produced that the fluorescence statistics for centrifuged samples closely represent the fluorescence distribution of stained single bacteria in the reaction mixture at the end of incubation; in particular, centrifugation did not cause aggregation of bacteria. Centrifugation is proposed as more effective than mere dilution for use with a wide range of bacterial concentrations, and the moving zone technique is to be preferred to conventional centrifugation in which bacteria tend to aggregate in the pellet. In indirect assays, it was shown that the washing step after reaction with antibacterial antibody may be omitted. The performance of direct and indirect staining methods was compared, including the use of either Staphylococcus aureus protein A or polyclonal sheep anti-rabbit antibody as the indirect reagent. When the bacterial concentration in reaction mixtures was increased the median fluorescence intensity fell, indicating that specific antibody had become limiting at low concentrations of the polyclonal antibody preparations. The implications of this for the design of flow cytometry assays of bacteria are discussed. Key words: Immunofluorescence; Flow cytometry; Antibacterial antibody; Bacterial; Antigen distribution

Introduction

The technique of flow cytometry, well established in studies of eukaryotic cells, has only recently been applied to the analysis of bacterial populations by immunofluorescence (IF) staining of surface antigens. Published reports for bacteria have essentially been limited to the detection of particles assumed from their light scatter and fluorescence characteristics to be individual bacteria

Correspondence to: A.P. Phillips, Chemical Defence Establishment, Porton Down, Salisbury, Wiltshire SP4 0JQ, U.K.

specifically stained with antibody (Ingram et al., 1982; Steen et al., 1982; Tyndall et al., 1985). In the flow immunofluorescence (FIF) methods used, the excess unreacted antibacterial antibody was removed by centrifugation washing. However, in our hands, many bacterial preparations that can readily be resuspended as single bacteria after centrifugation from saline buffer, are recovered with low efficiency when centrifuged in the presence of specific antibacterial antibody; this is presumably due to the formation of immune complexes (Phillips and Martin, 1985a). Even when a high proportion of bacteria can be recovered when antibody-containing reaction mixtures are washed,

0022-1759/87/$03.50 © 1987 Elsevier Science Publishers B.V. (Biomedical Division)

220

centrifugation is likely to encourage the formation of small aggregates of bacterial doublets, triplets, etc. Tyndall et al. (1985) suspected that the broad, tailing fluorescence distributions for FIF analysis of Legionella were partly the result of bacterial aggregation. A further disadvantage of centrifugation washing is the risk of loss of antigen and/or bound antibody during the traumatic resuspension process. These consequences of centrifugation would reduce the utility of FIF data in applying cytometry to population studies of surface antigen expression in bacteria. In order to avoid the problems inherent in the centrifugation of bacteria in the presence of antibody, we developed an FIF procedure that involves direct cytometry of reaction mixtures. The technique relies on using low concentrations of fluorescein-conjugated antibacterial antibody in reaction mixtures, and flowing the sample slowly through the cytometer so as to achieve a narrow sample stream. Both factors help to minimize triggering of the fluorescence photodetector other than for a bacterial-size particle (Phillips and Martin, 1985a, b). Cell sorting of FIF reaction mixtures and subsequent fluorescence microscopy was used to establish the proportion of single bacteria in the cytometer sample. Thus, cell sorting of Bacillus anthracis Vollum spores gated on the central region of the cytometer fluorescence peak, which included 93% of fluorescence events, indicated that 98% of the bacteria in this region existed as single particles (Phillips and Martin, 1985a). The potential advantage of our bacterial FIF method over techniques where stained bacteria are washed by centrifugation, for generating fluorescence data that accurately reflect the staining of individual bacteria without confusion by data from aggregates, is somewhat degraded by the fact that in our method even such small amounts of free fluorescent conjugate in the sample stream result in falsely high fluorescence values for stained bacteria (Phillips and Martin, 1983). The problem of reducing the fluorescence contribution of free antibody is addressed in the present communication. Centrifugation of FIF bacterial reaction mixtures through a sucrose solution in a moving zone technique is evaluated as a simple means of removing unreacted antibody, thus reducing the an-

tibody component of the fluorescence signal without encouraging aggregation of bacteria. Sucrose centrifugation has permitted the previous direct IF technique to be modified to an indirect method. We also examine the implication that polyclonal antibacterial antibodies used at low concentrations may become the limiting reagent in determining fluorescence intensity when the bacterial concentration increases.

Materials and methods

Bacteria, antibodies and other reagents Escherichia coli MRE 162 (O8 : K9) and spores of Bacillus anthracis Vollum were grown, inactivated in formaldehyde, and washed in saline phosphate buffer (SPB) as previously described (Phillips and Martin, 1982b, 1984). Working dilutions of bacteria were prepared freshly each day by dilution in SPB that had been filtered three-fold through 0.22 #m membrane filters (3 x SPB) as earlier described (Phillips and Martin, 1983); in the case of spores, suspensions containing 108 organisms/ml were sonicated before dilution (Phillips and Martin, 1982a, b). The immunoglobulin G (IgG) fraction of sheep anti-rabbit globulin (SAR) and the IgG of rabbit anti-B, anthracis spore (aBa) were prepared by salt precipitation and ion exchange chromatography, as earlier described (Phillips and Martin, 1982a; Phillips et al., 1984). Rabbit IgG against E, coli (aEc) was prepared from antiserum by two-fold salt precipitation. Fluorescein conjugates of aBa (F-aBa), of aEc (F-aEc), and of SAR (F-SAR) were prepared having molar F / P ratios of 7.0, 7.9, and 6.6, respectively (Phillips and Martin, 1982a). Fhiorescein-conjugated protein A (F-PA) of Staphylococcus aureus was purchased from Pharmacia, and had a molar F / P ratio of 6.0. Unlabelled and conjugated IgG, and F-PA, were diluted in bovine serum albumin buffer (BSAB : 1% BSA (Sigma), 0.1% Brij 35 (BDH), in SPB) to a protein concentration of 200 or 1000 /,g/ml, filtered through 0.22 /Lm membrane filters, and kept for up to 5 days at 4 °C. Immunofluorescence (IF) reaction Direct IF. To 106 or 107 organisms of B.

221

anthracis or E. coli suspended in 990 #1 of 3 x SPB were added 2 or 10 #g of the homologous conjugate F-aBa or F-aEc, contained in 10 #1 of BSAB. After incubating for 1 h at room temperature (19-21°C), reaction mixtures were diluted 1/10 or 1/100 in 3 x SPB, then analyzed using the flow cytometer. Indirect IF." simultaneous addition assay. To 106 or 107 bacteria in 970 /~1 of 3 x SPB were added the amounts shown of aBa or aEc in 10/~1 of BSAB, followed immediately by either F-SAR or F-PA, also in 10 #1 of BSAB. After incubating for 1 h, the reaction mixture was diluted 1/10 or 1/100 in 3 x SPB and analyzed. Indirect IF." deferred addition assay. This procedure was as for the simultaneous addition assay, except that bacteria were incubated for 30 rain with the aBa or aEc before addition of the F-SAR or F-PA and incubation for the final 30 min. Indirect IF." sequential addition assay, with intermediate sucrose centrifugation. 107 bacteria were incubated with aBa or aEc for 30 min, as for the deferred addition assay. Then, 0.5 ml of 40% (w/v) sucrose solution from a motorized syringe was introduced over about 1 rain below 0.5 ml of the reaction mixture, taking care not to disturb the visible interface of the two phases. This mixture was centrifuged in a Quickfit model 320 microcentrifuge (angle rotor, gmax 14000 ×g), power being applied for 10 s in the case of B. anthracis tests and for 15 s in the case of E. coli tests. Using a syringe and needle, 0.25 ml of the lower, sucrose layer were removed, keeping the needle tip away from the tube walls and from the phase interface. The recovered sample was diluted 1/5 in 3 x SPB, incubated for 30 min with F-SAR or F-PA, then diluted 1/10 in 3 × SPB and analyzed. Sucrose centrifugation immediately before cytometry. As an alternative to diluting direct and indirect reaction mixtures after incubation to reduce the conjugate levels for cytometry, incubated reaction mixtures were subjected to sucrose centrifugation as described above. After centrifugation, the recovered sample was diluted 1/5 or 1/20 in 3 x SPB for analysis. Operation of the flow cytometer The flow cytometer was a Cytofluorograf 50H (Ortho Instruments) fitted with a 2 W argon-ion

laser (Coherent). The laser was operated at 488 nm, at a power of 150 mW. Sheath fluid was membrane filtered deionized water (Milliq, Millipore Corp.). Combined red and green fluorescence peak photomultiplier signals gated on narrow forward angle (NFA) light scatter were inputted into a PDP 11/34 computer (Digital). Optical alignment was maximized by the use of 1.78 #m diameter fluorescent monodisperse carboxylated microspheres (Poly Sciences) at a standard gain of the fluorescence photomultiplier. Periodically during FIF experiments, any drift in optical alignment (usually less than 2%) was corrected. In FIF analyses, the movement of the meniscus in the graduated pipette in the sample entry line was timed to allow a flow rate of 20 t*l in about 30 s (Phillips and Martin, 1983). Data were acquired for 5120 cytometer events, or less (down to 1536 events) in the case of samples with the lowest bacterial concentrations, when acquisition was stopped after about 70 s as the end of the sample stream approached the flow cell. Fluorescence and NFA scatter histograms and cytograms of fluorescence versus NFA scatter were prepared off-line. To calculate mean, median, standard deviation (SD), and coefficient of variation (CV) for the bell-shaped region of fluorescence histograms believed to be due to stained bacteria (Phillips and Martin, 1983, 1985a), the left hand limit for statistical analysis was set just above the sharp peak due to unstained particles. The fight hand limit was at computer channel 256. Fluorescence was measured at one of four photomultiplier gains, sufficiently close to allow some reaction mixtures to be assayed at two gain settings and the ratio of the two medians calculated. Thus, a set of ratios was established encompassing all four levels of photomultiplier performance. Fluorescence median values for duplicate IF reaction mixtures were averaged, then normalized by means of these photomultiplier ratios. As previously reported (Phillips and Martin, 1985a), the coefficient of variation for mean and median fluorescence within replicate sets was generally less than 3%. Pairs of cumulative fluorescence distributions constructed from individual assays or sets of replicates were compared by the Kolmogorov-Smirnov two-sample test, using a computer program written by one of us (AJC). The origin of cumulative

222

distributions was offset above the unstained-particle peak, so that variation in the size of this peak did not affect the comparison of the fluorescence distribution of stained bacteria. Since the maximum percentage difference D' between two distributions could reflect several sources of difference - differences in location (central tendency), in dispersion, skewness, etc. - the comparison was repeated ( D " ) after normalizing for location by multiplying the data of one distribution by the ratio of the two medians.

150

A

100 0 ~J

50

isoiI

B

Results

Effect of centrifugation of direct IF reaction mixtures on fluorescence distribution In a preliminary study of B. anthracis spores stained with F-aBa at a concentration of 2/~g/ml, IF reaction mixtures were analyzed directly and also after a series of doubling dilutions in 3 × SPB. The estimate of fluorescence median fell to 90% after 1/2 dilution and continued to fall to 79% at the highest dilution, 1/64. Sucrose samples recovered after centrifugation of reaction mixtures, and diluted at least 1/4 before cytometry, gave NFA scatter histograms indistinguishable from those for bacteria that had not been centrifuged. Lower dilutions of sucrose samples gave rise to displaced NFA scatter peaks presumably because of refractive index changes. Integration of NFA scatter histograms indicated that the bacterial concentration in sucrose samples recovered after centrifugation was about 50% of the concentration in the bacterial suspension applied to the centrifuge tube. Thus, to investigate the effect of sucrose centrifugation of bacteria on cytometry fluorescence data the standard conditions of 1/10 dilution of IF reaction mixtures and 1/5 dilution of sucrose samples were formulated so as to provide similar bacterial concentrations in the two types of samples and hence similar acquisition rates during cytometry. The behaviour of free antibody conjugate during sucrose centrifugation was modeled by centrifuging solutions of fluorescein-isothiocyanate through sucrose and measuring fluorescence by means of a spectrofluorometer. In this way, it was estimated that the concentration of free conjugate

1

50 100 150 200 Ftuorescence (channet no )

250

Fig. 1. Fluorescence histograms for B. anthracis spores stained with F-aBa conjugate. B. anthracis at a concentration at 107 organisms/ml were incubated for 1 h with F-aBa at 2 ~ g / m l . A: The reaction mixture was diluted 1/10 for analysis. B: The reaction mixture was centrifuged on sucrose, and the recovered sample was diluted 1/5 for analysis.

would be reduced about 20-fold by sucrose centrifugation. Typical fluorescence histograms for stained B. anthracis analyzed before and after the standard sucrose treatment are shown in Fig. 1. Statistical estimates for these fluorescence distributions are compared in Table I with estimates for samples of the same IF reaction mixture centrifuged on sucrose for shorter times, 3 s and 5 s. For a narrow fluorescence range between channels 28-112 that included the histogram peak, the SD and CV altered little after centrifugation. However, when the fluorescence range was extended to channels 17-256 to allow routine analysis of bacterial samples widely differing in fluorescence intensity, centrifugation was accompanied by a larger reduction in CV, increasing with centrifugation time. This trend was apparently due to the reduction in fluorescence count between channels 113-256. For analysis in either the narrow or broad fluorescence range, the fluorescence median

223 TABLE I F L U O R E S C E N C E STATISTICS F R O M CYTOMETRY OF A DIRECT I M M U N O F L U O R E S C E N C E REACTION M I X T U R E OF B. A N T H R A C I S Data refer to the reaction mixture described in Fig. 1. Statistical analysis between channels no.

Percentage of fluorescence count between channels no.

Treatment of IF reaction mixture Centrifugation

Dilution

2-12

13-27

28-112

113-256

time (s) 3 5 10

1/10 1/5 1/5 1/5

4.5 9.8 10.3 12.5

3.5 1.7 0.8 1.0

84.0 82.5 83.2 82.0

8.1 6.0 5.6 4.5

28-112

17-256

Median

CV%

Median

CV%

71 66 67 67

26.9 26.9 26.2 25.8

72 67 68 68

38.2 36.9 34.9 33.6

(71) a (68) (68) (67)

(35.5) a (34.4) (31.8) (31.7)

a Fluorescence statistics in parentheses were calculated after gating on the low scatter region in Fig. 2, as described in the text.

was reduced by 4-5 channels after centrifugation, and the mean fell by between 2.6-4.4 channels. The cytograms for the reaction mixtures referred to in Fig. 1 are shown in Fig. 2. In the NFA scatter region that accounted for the great major480

"'I

I

I

I

I

I

I

I

I

_

384

i A

-,- 288 t.J

<

192

LI_

:7

;

96

'....,.....:..

:..:.

0 0

480

I

i

I

120

I

•

'

:~i'.'.,k~,.,'....:.

I

I

I

240

I

I

1

:

..

I

I

I

I

I

360 4B0 600 Ftuorescence

I

I

I

I

I

I

720

I

I

I

I

I

840

1

I

I B

38t~

960

-

'~ 2 8 8

.,ero

U-

z

192

--

96

"~

..

'

..

.-....

-

.'..

.

.

-4"

I

0

I

120

I

I

240

~ I I 1 1 1 1 1 1 1 360 480 600 720 B40 Fluorescence I

960

Fig. 2. Cytograms for B. anthracis spores stained with F-aBa conjugate. Data refer to the reaction mixture described in Fig. 1. Conditions for A and B as for Fig. 1, but the intensity scales were not equivalent. The dotted line intercept at 150 U on the y-axis was applied after the data were acquired (see text).

ity of the bacteria, i.e., below the dotted line set by inspection at 150 on the y-axis, the number of highly fluorescent particles was reduced after centrifugation. (Scales on the histograms and cytograms are not equivalent.) The proportion of fluorescent particles with high NFA scatter characteristics; i.e., above the dotted line, also fell after centrifugation. Thus, in the assays of Table I for fluorescence analysis between channels 17-256, the proportion of highly scattering particles was 4.6% for the 1/10 diluted reaction mixture and 2.6% after centrifugation for 10 s. In both cases the fluorescence statistics for this high scatter population had higher values than for low scatter events, with the fluorescence median in channel 119. However, recalculation of the fluorescence statistics of Table I after gating on the low scatter region (estimates in parentheses in the table) had little effect, reducing median values by one channel number and mean values by up to two channels. As an alternative to specifying the NFA scatter range for gating, a statistical analysis computer program was written in which fluorescence channels containing a low count were ignored; this threshold count could be preset between 1-5 events. In the data of Table I, derived from 5120 events, ignoring channels containing two or less events gave statistical estimates close to the low NFA scatter-gated statistics shown in parentheses in the table.as Kolmogorov-Smirnov comparisons of cumulative fluorescence distributions for replicate IF reaction mixtures generally gave difference values

224

D' and D " between 3-5%. Normalized D " values were also in this range when 1/10 diluted reaction mixtures were compared with sucrose centrifugation derivatives. Essentially similar results were obtained when the effect of sucrose centrifugation on direct IF reaction mixtures of E. coli was investigated.

Effect of bacterial and conjugate concentration on direct FIF assay performance Fluorescence estimates for direct IF reaction mixtures based on differing concentrations of bacteria and fluorescent conjugates may be seen in Table II. Antibody binding was apparently near equilibrium after incubating for 1 h, since extending the reaction time did not greatly increase the degree of staining. It is clear that use of the antibacterial conjugates at a concentration of 2 /~g/ml did not provide sufficient specific antibody to saturate the antigenic sites on 10 6 bacteria under these reaction conditions, since fluorescence staining was considerably enhanced by increasing the conjugate concentration to 10 #g/ml. At either concentration of conjugate, increasing the bacterial concentration to 10 7 organisms/ml greatly reduced the degree of staining. Bacterial reaction mixtures based on 10 6 organisms/ml were not normally diluted more than 1/10, or 1/5 after sucrose centrifugation, TABLE I1 FLUORESCENCE STATISTICS FROM CYTOMETRY OF DIRECT IF REACTION MIXTURES OF B. A N T H R A C I S AND E. COLI Bacterial concentration (organisms/ml)

F-aEc or F-aBa concentration

(~g/~)

Fluorescence median for IF reaction mixtures Diluted Centrifuged and diluted 1/10

E. coh 106 107 106 107

2 2 10 10

168 36 400 189

B. anthrac~ 106 107 106 107

2 2 10 10

201 80 321 252

1/100

31 167

75 245

1/5 144 30 359 160 193 77 309 249

1/20

29 152

76 247

because with the increased cytometer analysis time needed for higher dilutions, the bacterial fluorescence histograms became increasingly noisy in the sharp left hand peak, and in the region between this and the bacterial peak, increasing values of SD and CV. The greater number of bacteria in reaction mixtures based on 107 organisms/ml allowed these to be diluted further without apparently affecting SD values; as shown in Table II, fluorescence median estimates were reduced by such additional dilution of the 1/10 reaction mixture or of the 1/5 sucrose sample. Use of the Kolmogorov-Smirnov method to compare reaction mixtures of E. coli processed by sucrose centrifugation (1/5 dilution) in which either the bacteria or conjugate differed in concentration as in Table II, gave substantial values of the normalized difference parameter D", from 7 to 25%. This suggests a change in fluorescence dispersion or skewness, although no statistical significance tests could be performed. Smaller D " values were obtained for the same comparisons of B. anthracis assays, ranging from 4 to 10%.

Indirect FIF assays Preliminary studies of indirect staining with F-SAR established that detectable levels of fluorescence staining could only be achieved when the concentration of F-SAR was at least 10 /xg/ml and was equal to or greater than the concentration of the unlabelled direct antibodies aBa and aEc. Typical results for indirect assays of B. anthracis in which various concentrations of aBa and F-SAR were used may be seen in Table III. The fluorescence levels achieved in simultaneous assays were roughly comparable with levels in sequential assays performed using similar antibody concentrations, and were always higher than in the equivalent deferred addition assays. Increasing the bacterial concentration from 10 6 to 10 7 organisms/ml affected fluorescence signals far less than in the case of direct assays, even at aBa concentrations of 2 or 3/~g/ml. However, the high fluorescence median values obtained in direct assay reaction mixtures based on high concentration ratios of conjugate:bacteria could not be achieved in F-SAR indirect assays; after allowing for the differing F / P molar ratio of the direct conjugate and the F-SAR, the number of fluorescent anti-

225 TABLE III FLUORESCENCE STATISTICS FROM CYTOMETRY OF INDIRECT IF REACTION MIXTURES OF B. A N T H R A C I S USING F-SAR AS THE INDIRECT REAGENT Assay type

B. anthracis

concentration (organisms/m1)

aBa concentration ( # g/ml)

F-SAR concentration (/~g/ml)

Fluorescence median for IF reaction mixtures Diluted 1/10

Simultaneous addition

106 107 106 107

3 3 10 10

20 20 10 10

87 85 56 51

Deferred addition

106 107

10 10

10 10

41 39

Sequential addition

107 107

3 3

10 20

Centrifuged and diluted 1/100 78

1/5 79 78 53 44

1/20 76

38 31 83 103

88 109

85 105

TABLE IV FLUORESCENCE STATISTICS FROM CYTOMETRY OF INDIRECT IF REACTION MIXTURES OF B. A N T H R A C I S AND E. C O L I USING F-PA AS THE INDIRECT REAGENT Assay type

Bacterial concentration (organisms/rnl)

Simultaneous addition

B. anthracis 106

Deferred addition

B. anthracis 106

Sequential addition

B. anthracis 107

Simultaneous addition

E. coli

Sequential addition

F-PA concentration Otg/ml)

Fluorescence median for IF reaction mixtures Diluted Centrifuged and diluted 1/10

1/100

1/5

1/20

10 10

1 1

117 111

105

112 105

107

10 10

1 1

98 91

90

90 86

87

]07

10 10

1 2

116 119

111 111

110 108

106 107

10 10

1 1

142 85

126 82

81

107

10

l

75

67

65

107 107

E. coli

aBa or aEc concentration (/~g/ml)

body molecules bound on average was always lower in indirect assays. In view of the undesirability of the high fluorescence contribution expected from unreacted FSAR used at 10 ~tg/ml or more, even after centrifugation on sucrose, use of F-PA as the indirect reagent was investigated. As is evident from Table IV, fluorescence levels for B. anthra¢is similar to those in the F-SAR assays could be obtained using as little as 1 ~tg/ml of F-PA. Even at this low concentration of F-PA, median fluorescence values were reduced somewhat by further dilution

109

81

of the 1/10 reaction mixtures; however, further dilution of 1/5 sucrose samples did not alter the median by more than three channel numbers, and in some cases an increased median was recorded. Simultaneous addition assays compared well with sequential addition assays. F-SAR indirect assays of E. coli were not performed, but the F-PA assay results of Table IV indicate comparable simultaneous addition and sequential addition performance, and independence of bacterial concentration similar to that found for indirect B. anthracis assays.

226

Kolmogorov-Smirnov comparisons of fluorescence distributions for indirect assays performed using identical concentrations of reagents but by the different assay methods, simultaneous, deferred and sequential, gave normalized D " values of 2-5%.

Discussion The high reproducibility of immunofluorescence measurements by cytometry has permitted the effect of excess unreacted antibody conjugate on fluorescence statistics to be stated with some precision in the present study. It is clear that dilution of bacterial IF reaction mixtures before analysis provides a simple means of effecting a limited reduction in the false fluorescence effect, but the disadvantages of analyzing the bacteria at very low data acquisition rates, including the increased noise, impose a practical limitation on the utility of dilution. The centrifugation of bacterial IF reaction mixtures through sucrose solution in a simple and rapid moving zone technique provided an estimated 20-fold reduction of free conjugate concentration whilst only reducing the bacterial concentration by a factor of 2, thus leading to ten-fold improvement in the concentration ratio of bacteria:free conjugate. Consistent with this, for the wide range of bacterial IF reaction mixtures tested fluorescence median values for recovered sucrose samples diluted 1/5 before analysis were in most cases indistinguishable from data for the original reaction mixture diluted 1/100, in view of the 2-3% CV of fluorescence measurement. There was no indication that further dilution of sucrose samples to 1/20 significantly reduced the fluorescence median. Examination of the small changes in cytometry data that accompany sucrose centrifugation suggest that three minority types of particles in the reaction mixtures were depleted by centrifugation. These are particles with low fluorescence values, particles with high fluorescence values, and particles with high NFA light scatter values. It is not unreasonable to suppose that these minority particles are variously stained debris or bacterial aggregates, and this leads to the conclusion that fluorescence statistics obtained after centrifuga-

tion of the bacteria through sucrose reflect the antigenic distribution on single bacteria even more closely than data obtained by analyzing the original reaction mixture. It is only by clearly demonstrating that a flow cytometry technique can provide accurate fluorescence measurement of single bacteria in a population, without confusion from aggregates, that the potential of flow cytometry for studying antigen distribution in bacteria is likely to be realized. At the low concentrations of antibacterial antibody that must be used to provide a small fluorescence component for unreacted antibody during cytometry, even given the advantage of sucrose centrifugation, the activity of specific antibody was clearly insufficient to saturate antigenic sites when the bacterial concentration was raised. This should not be a problem for population studies of surface antigens, when the relative staining of bacteria in a population sample will be more important than the absolute degree of staining. Moreover, bacteria being studied in this way are likely to be obtained from pure culture and it should be a simple matter to adjust bacterial concentration to a standard value. The evidence from Kolmogorov-Smirnov comparison of cumulative distributions of reaction mixtures differing in the concentration ratio of bacteria:conjugate, that even after normalizing the median fluorescence value differences in dispersion and/or skewness remain, indicates that it would be important to standardize reagent concentrations in order to study small changes in antigen distribution. Different criteria apply to an FIF method intended for detecting specific bacteria in unknown sampies; specific immunofluorescence ideally should occur in an intensity range as narrow as possible but for bacterial concentrations as diverse as possible. Increasing the concentration of antibacterial antibody so as to increase the dynamic range for bacterial concentration would be inadvisable not so much because of the loss of accuracy of fluorescence measurement, which would still compare favourably with other methods, but because the increase in unwanted triggering of the fluorescence detector and thus the increased dead time in detector circuitry would prejudice the detection of stained bacteria at low concentrations. In dealing with unknown samples where a high bacterial

227

concentration is a possibility, we would prefer to use a low concentration of antibacterial antibody but to react test samples at two concentrations in FIF reaction mixtures, say 100-fold different. The choice between direct and indirect versions of the bacterial assays studied is complex. The performance of direct assays was more dependent on the concentration ratio of bacteria:conjugate than were analogous indirect assays. Assuming that the amount of bacterial debris contaminating these bacterial preparations after 6-fold washing would contribute little to the antibody-binding capacity of the preparation, then if on average each bacterium can bind even as much as 1/10 of its weight of antibody, 10 6 bacteria of this size would only bind 10 ng of antibody. Since extending the IF reaction times had little effect on the degree of fluorescence staining, it may be inferred from the direct assay data that the content of high avidity specific antibody in these polyclonal preparations is likely to be no more than 1%. The relative independence of indirect assays to bacterial concentration suggests that the fluorescein conjugation procedure has denatured a proportion of the specific antibacterial antibodies a n d / o r reduced their avidity. In view of the accuracy of fluorescence measurement by flow cytometry, it seems an ideal method to use in investigating alternative methods of conjugating fluorochromes to antibody. Comparison of fluorescence data for indirect assays of the simultaneous and deferred addition types with data for the classical sequential addition design demonstrates that the washing step between application of the 1st and 2nd antibody reagents may be omitted. We previously reached a similar conclusion for solid-phase IF assays of B. anthracis spores (Phillips and Martin, 1982c). In the solid-phase assay it was calculated that about twice as many F-PA molecules as F-SAR molecules are bound at saturation. Under the reaction conditions used in the present study, believed to be near equilibrium and as may be inferred from the data of Tables III and IV for different concentrations of the indirect reagent, not far from saturation, the molecular ratio was near 4 : 1. The average number of F-PA molecules bound per bacterium in the FIF analysis was approximately equal to the number of antibacterial antibody

molecules bound in direct assays, but the F-SAR binding level was about half the direct level. When the binding of these preparations of SAR and direct anti-B, anthracis IgG were compared previously in solid-phase assays (Phillips and Martin, 1982b), a direct:indirect molecular binding ratio close to unity was calculated. However, in the solid-phase assays, antibody reagents were used at much higher concentrations, 100 /~g/ml, and the FIF data may merely reflect a low concentration of active antibody in the present preparation. It is also possible that the proximity of the glass surface of the solid-phase support in some way affected the ultimate ratio of 1st to 2nd antibody molecules. The present results for F-PA confirm the expectation that a high proportion of active molecules with a high avidity for rabbit IgG will be encountered in a protein A preparation (Goding, 1978). The corollary that in FIF assays a low concentration of F-PA can be used hence reducing the non-specific fluorescence contribution to fluorescence signals, is likely to confer a considerable advantage on F-PA over a majority of anti-species antibody reagents.

Acknowledgements We are grateful to Mr. N.L. Cross for advice in statistical analyses.

References Goding, J.W. (1978) Use of staphylococcal protein A as an immunological reagent. J. Immunol. Methods 20, 241. lngram, M., Cleary, T.J., Price, B.J., Price, R.L. and Castro, A. (1982) Rapid detection of Legionella pneumophila by flow cytometry. Cytometry 3, 134. Phillips, A.P. and Martin, K.L. (1982a) Evaluation of a microfluorometer in immunofluorescence assays of individual spores of Bacillus anthracis and Bacillus cereus. J. Immunol. Methods 49, 271. Phillips, A.P. and Martin, K.L. (1982b) Assessment of immunofluorescence measurements of individual bacteria in direct and indirect assays for Bacillus anthracis and Bacillus cereus spores. J. Appl. Bacteriol. 53, 223. Phillips, A.P. and Martin, K.L. (1982c) Variations on the staining method in quantitative indirect immunofluorescence assays for Bacillus spores, and the use of fluoresceinprotein A. J. Immunol. Methods 54, 361.

228 Phillips, A.P. and Martin, K.L (1983) Immunofluorescence analysis of Bacillus spores and vegetative cells by flow cytometry. Cytometry 4, 123. Phillips, A.P. and Martin, K.L (1984) The effect of the environment on the immunoflnorescence staining of bacteria. J. Immunol. Methods 69, 85. Phillips, A.P. and Martin, K.L. (1985a) Dual-parameter scatter-flow immunofluorescence analysis of Bacillus spores. Cytometry 6, 124. Phillips, A.P. and Martin, K.L (1985b) Identification of bacteria by flow immunofluorescence. In: K.O. Habermehl (Ed.), Rapid Methods and Automation in Microbiology and Immunology (Springer-Verlag, Berlin), p. 408.

Phillips, A.P., Martin, K.L. and Horton, W.A. (1984) The choice of methods for immunoglobulin IgG binding: yield and purity of antibody activity. J. Irnmunol. Methods 74, 385. Steen, H.B., Boye, E., Skarstad, K., Bloom, B., Godal, T. and Mustafa, S. (1982) Applications of flow cytometry on bacteria: Cell cycle kinetics, drug effects and quantitation of antibody binding. Cytometry 2, 249. Tyndall, R.L., Hand, R.E., Mann, R.C., Evans, C. and Jernigan, R. (1985) Application of flow cytometry to detection and characterization of Legionella spp. Appl. Environ. Microbiol. 49, 852.