This material was originally published in the Purdue Cytometry CD-ROM Series

Flow Cytometry and Microbiology

Flow cytometric analysis of gram-negative bacterial cells.

Dave Duncan, Peggy Ooi, and Robert Zagursky
Wyeth-Lederle Vaccines and Pediatrics
211 Bailey Road
West Henrietta, NY 14586 USA


Since the mid-to-late 1980 s, flow cytometry has begun to be used as an important tool by microbiologists. At Wyeth-Lederle Vaccines and Pediatrics we wished to take advantage of the multiparametric analysis of potentially heterogeneous populations, afforded by flow cytometry, in our work with pathogenic bacteria. There has not been as widespread use of flow cytometry in analyses of bacteria as there has been for eukaryotic cells, and, based on our experiences, we suspect that in part, this may be due to difficulty in discriminating bacteria from debris. However, we have found conditions allowing analysis of a variety of bacterial species differing in size and shape. We also have found that methods in common use for analysis of eukaryotic cells apply surprisingly well to analysis of bacteria. We have explored the use of the Becton Dickinson FACSort unit in bacterial analysis, and we find this instrument well suited to this task. What follows is a description of some of our experiences that may be of general interest to other users. We describe 1) staining conditions and instrument settings on our FACSort that are applicable to a variety of bacteria differing in size and shape, 2) the limits of light scatter resolution on our flow unit, and 3) a strategy for FACSort analysis of bacterial cells less than 1 micron in size.

Staining Conditions and FACSort Settings
Neisseria meningitidis cells were stained either in tubes or in 96-well U bottom plates. A 100 ul volume of a cell suspension (OD620 approximately 0.05 when brought to 1.0 ml PBS) was stained with 0.1-0.5 ug purified antibody in PBS/0.05% NaN3 for 20 minutes at room temperature. Cells were washed twice and stained with secondary, if required. After the final wash, cells were suspended to 1.0 ml PBS/NaN3 and analyzed.

Nontypeable Haemophilus influenzae (NTHi) cells were stained in 1.5 ml microcentrifuge tubes. A 200 ul volume of a cell suspension (OD600 approximately 0.04 when brought to 1.0 ml of PCM [PCM buffer: 10mM NaPO4, 150mM NaCl, 0.5 mM MgCl2, 0.15mM CaCl2]) was stained with primary antibody (1:200) for one hour at 37 C. Cells were washed and stained with secondary biotinylated antibody (1:20) for 15 minutes at 37 C, and then washed and stained with Quantum Red-conjugated streptavidin (1:40, Sigma) for 15 minutes at 37 C. One ml of PCM was added to the labelled cells, washed and resuspended in a final volume of 1 ml PCM.

We have analyzed dead as well as viable cells using the viability stain Syto 9 (Molecular Probes) added before analysis according to the manufacturer's specifications. In our particular applications, we find no evidence of nonspecific staining of dead cells. This can be advantageous when analyzing biohazardous organisms, in that a bulk suspension of dead cells can be prepared and analyzed at one's leisure, as opposed to staining a viable cell suspension, fixing, and assessing survival before analysis.

We collected events triggered on side scatter (SSC) with a threshold of 0. As discussed below, the SSC PMT is more sensitive than the forward scatter (FSC) photodiode, and so we used this parameter to trigger. Typically we do not set a higher threshold or gate on either light scatter parameter; when it is necessary to discriminate cells from debris we use an additional parameter, as described below. For all analyses, FSC and SSC were set in log mode; the FSC gain was EO1 and the SSC gain and voltage were typically 1 and 300. FL1, FL2, and FL3 were also in log mode with gains of 1-2 and voltages set to 400-600, as warranted by comparisons of control-stained and specifically-stained cells. Superior results were obtained when the suspension was sampled on a low setting (12 Šl/min), giving a data acquisition rate of no more than 300-600 events/second. With higher flow rates we observed increased light scatter and fluorescence, but with a greater CV. We concluded that under these conditions there is a greater chance for coincidence, where multiple cells may be analyzed as a single event.

Data Examples
Examples of the serological specificity of bacterial stains are shown in Figures 1 and 2. In Figure 1, suspensions of two heat-killed strains of Neisseria meningitidis (strain 1 and strain 2) were stained with either of two serotype-specific monoclonal antibodies followed by a FITC-conjugated secondary. In each panel is shown the reactivity of one of the bacterial strains with both monoclonal antibodies. In the left panel, strain 1 stained with anti-strain 2 had low fluorescence, with the histogram filling the first decade; when stained with anti-strain 1, fluorescence shifted to decades 3-4. Analogous results were observed with strain 2 in the right panel: labelling was observed with anti-strain 2 antibody, but not anti-strain 1 antibody. In Figure 2, strain 1 was stained with pooled pre- or post-vaccination serum from a group of mice that had been vaccinated with a subunit formulation derived from this strain. The panels show the staining profiles for cells stained with 1/100 dilutions of pre- or post-vaccination sera followed by a secondary antibody specific for different mouse isotype: IgM, IgG1, IgG2a, and IgG2b. As can be seen, the immune sera shifted the staining profile relative to the pre-immune sera to various degrees for all the isotypes tested. These results indicate that there is serological specificity under these conditions of staining and analysis.

Figure 1. Detection of serotype-specific determinants on different N. meningitidis strains. Heat-killed suspensions of strain 1 (A) and strain 2 (B) were stained with monoclonal antibodies specific to one strain or the other, and then analyzed. The fluorescent histograms for both stains are shown in each panel. For both strain 1 and strain 2, the monoclonal with the inappropriate reactivity stained at background levels whereas the reactive monoclonal gave a shift in fluorescence.

Figure 2. Detection of antigen-specific isotypes in sera from vaccinated mice. In each panel is shown the fluorescence of strain 1 cells stained with a 1/100 dilution of pre- or post-vaccine serum followed by the indicated secondary antibody.

The asymmetry of the right-shifted one-parameter histograms in Figures 1 and 2 is notable. In fact, with specifically-stained cells we observed increases not only in the fluorescent signal, but light scatter as well. We have concluded that these changes are derived from antigen-antibody complexes, for several reasons. First, aggregates were evident by fluorescent microscopy; second, the more brightly fluorescent cells had higher FSC and SSC as well, with all three parameters showing a direct relationship; third, these characteristics were seen with specific stain but not control stain; and fourth, a mild bath sonication step reduced FSC and SSC signals to the level of control-stained cells, and caused a reduction in the fluorescent CV, being more symmetrical, but maintaining similar peak fluorescence. Some of these results are shown in Figures 3 and 4.

Figure 3. Light scatter of stained and unstained strain 1 cells. A). FSC and SSC are shown for cells that were stained with antibody. B). The same sample as in A was bath sonicated for 5 seconds, and then immediately analyzed. C). FSC and SSC of unstained cells.

Figure 4. Effect of sonication on the fluorescence profile of stained cells. A). Strain 1 was stained, washed and analyzed. B). The same sample as in A was sonicated and then immediately analyzed.

In Figure 3A it can be seen that light scatter increased in specifically stained cells compared to the unstained cells in Figure 3C; when the same tube was bath sonicated for 5 seconds, light scatter was equivalent to that of unstained cells (Figure 3B). Figure 4 shows the effects of sonication on the fluorescent signal of specifically-labeled cells. Figure 4A shows the fluorescence of cells stained with a reactive primary antibody, followed by a FITC-conjugated secondary antibody. As also seen in Figures 1 and 2, the fluorescent histogram was asymmetric; the most brightly fluorescent cells also had the greatest FSC and SSC (not shown). After sonication, light scatter was reduced, as in Figure 3; fluorescence was reduced as shown, resulting in a more symmetric peak. From these data we conclude that fluorescence and light scatter signals from a portion of the events in these samples is the collective signal of cells trapped in an antigen-antibody complex. Thus some events in these samples consist of complexes of cells recorded as a single event, with light scatter and fluorescence signals not the result of a single-cell event, but rather the sum of the individual signals of all the cells trapped in the complex.

These results demonstrate the serological specificity of bacterial strains. Although there can be aggregation in stained samples, this effect is dependent on a specific reaction between an antibody and its antigen. Thus in situations where a qualitative question is being asked, as in strain serotyping or identification of bacterial species in environmental or biological samples, aggregation in itself would not invalidate the result. Obviously, though, aggregation does present a problem in studies of expression levels of a determinant. For this reason we wished to define the parameters critical in examining single-cell events on the FACSort.

Light Scatter Resolution on the BD FACSort
We began by assessing the limits of light scatter resolution on our instrument using 0.2, 0.5, and 1.0 micron beads (Polysciences, Inc.). Figure 5 shows the FSC and SSC profiles obtained for each of these beads.

Figure 5. Light scatter profiles of 0.2, 0.5 and 1.0 micron beads. A). FSC histogram. B). SSC histogram.

As can be seen in Figure 5A, the 0.2 and 0.5 micron beads could not be distinguished with FSC. However, the 1.0 micron beads showed a shift in the FSC parameter. In contrast, each of the three bead sizes was clearly resolved on the basis of SSC (Figure 5B). Thus in our hands, the FACSort has a limited capacity to resolve particles less than 1 micron in size on the basis of FSC; but SSC is much more sensitive and can resolve particles at least as small as 0.2 microns. These results are consistent with the differences in sensitivities of the photodiode and photomultiplier tube used in the detection of FSC and SSC, respectively. There are two consequences that follow from these observations. First, we prefer to trigger on SSC, since FSC is less sensitive and may not trigger for some bona fide events. Second, we do not rely exclusively on FSC and SSC thresholds or gates to resolve bacterial cell populations less than 1 micron in size, but use an additional parameter as discussed below.

Bacterial Cell Resolution on the BD FACSort
In our studies of aggregation we noticed that the sample dilution can have a significant effect on the fluorescence histograms. This is illustrated in Figure 6. Heat-killed N. meningitidis was stained, washed, resuspended to 1.0 ml, and briefly sonicated as described above. Serial dilutions of the sonicated suspension were made and the samples were immediately analyzed. As can be seen in Figure 6, the frequency of positively staining cells apparently decreased as the sample was diluted; intriguingly the fluorescence intensity of the positive population remained constant. As described below, what we found was that debris is often present in our samples and, when the cells were diluted, the cells become less frequent relative to this debris. Because N. meningitidis fell below the limits of FSC resolution on our instrument, the light scatter profiles gave no indication of this effect.

Figure 6. The effect of sample density on the fluorescence profile. Strain 1 was stained, washed, sonicated and analyzed. Shown are the fluorescent profiles of A) undiluted cells, B) cells diluted 1/4, and C) cells diluted 1/16.

In order to distinguish cells from debris, we fluorescently labelled the bacteria using the intracellular viability stain, Syto 9, and detected viable bacteria using one of the fluorescent PMTs. Suspensions of viable NTHi cells were stained with antibody plus secondary, washed, resuspended in 1.0 ml, and then counter-stained with Syto 9.

Figure 7. Identification of antibody labelled and Syto 9 stained NTHi cells: analysis gate. A). Light scatter profile of the cells. B). FL1 (Syto 9) fluorescent histogram of the cells. C). FL3 (Quantum Red) fluorescent histogram of the cells. D). Resultant FL3 fluorescent histogram when gated on Syto 9 positive cells.

Figure 7A shows the light scatter of these cells. In this particular run there was less debris than is often seen, and a distinct population was suggested by light scatter. However, debris was evident and in fact the apparent bacterial cell population, as defined by light scatter, represented only 15% of the total events. Figure 7B shows the profile obtained with the viability stain, with the positive population also representing about 15% of the total events. Figure 7C shows the staining profile obtained with a NTHi-specific antibody: again, 15% of the total events were positive. Without the means to definitively resolve the bacterial cells, we could not make any conclusions about the percentage of positive cells. In Figure 7D, a post data analysis gate was used to analyze only Syto 9 positive events and it can be seen that essentially all of these events stained with the NTHi-specific antibody. Dual-parameter analysis of Syto 9 staining versus antibody staining indicated that all Syto 9 positive events were antibody positive, and all antibody positive events were Syto 9 positive (not shown).

Figure 8. Identification of antibody labelled and Syto 9 stained NTHi cells: acquisition gate. A). Light scatter of the gated cells. B). FL3 fluorescent histogram of gated cells. C). FL3 fluorescence histogram of only Syto 9 labelled cells.

Analogous results (Figure 8) were obtained when a data acquisition gate was used to collect only Syto 9 positive events. Figure 8A shows a scatter plot obtained using this data acquisition gate. Analysis of these cells now shows that 100% of the cells are labelled with NTHi specific antibody (Figure 8B). In an identical experiment using cells NOT labelled with NTHi specific antibody, only the expected background staining is observed (Figure 8C). This supports our conclusion that Syto 9 does label cells and is a useful parameter to use for gating.

General Discussion
There have been two basic challenges that have shaped our approach to the flow cytometric analysis of bacteria: resolution of the cells from debris, and aggregation of specifically-stained cells. These challenges may or may not be critical depending on the question being addressed in the analysis. As discussed above these issues are less relevant for qualitative analyses where one is asking a yes/no question, and more relevant for quantitative analyses where one would like to determine expression levels of a determinant. With regard to these two challenges, we have defined conditions of staining and analysis that allow us to discriminate bacterial cells from debris using the FACSort; and we also have defined conditions that minimize antibody-induced aggregation. However, we have not yet established if we are analyzing single-cell events. We are exploring the use of pulse analysis to see if we can discriminate single cells from small, 2-3 cell aggregates using DNA stains of fixed cells. Once we have defined conditions under which we can consistently restrict our analysis to single-cell events, we will determine the optimal staining conditions for expression level analysis. Specifically, we will compare the effects of indirect staining, direct staining, and staining with Fab fragments on aggregate formation.

Back to Flow Cytometry and Microbiology Introductory Page

| Facility Infomation | Email Archive | Web Sites | Organization and Meetings |
| Home | CD-ROMs | Research | Education | Email Us |