Real-time flow cytometry to assess qualitative and quantitative responses of oral pathobionts during exposure to antiseptics

ABSTRACT

Antiseptics are widely used in oral healthcare to prevent or treat oral diseases, such as gingivitis and periodontitis. However, the incidence of bacteria being tolerant to standard antiseptics has sharply increased over the last few years. This stresses the urgency for surveillance against tolerant organisms, as well as the discovery of novel antimicrobials. Traditionally, susceptibility to antimicrobials is assessed by broth micro-dilution or disk diffusion assays, both of which are time-consuming, labor-intensive, and provide limited information on the mode of action of the antimicrobials. The abovementioned limitations highlight the need for the development of new methods to monitor and further understand antimicrobial susceptibility. In this study, we used real-time flow cytometry, combined with membrane permeability staining, as a quick and sensitive technology to study the quantitative and qualitative responses of two oral pathobionts to different concentrations of chlorhexidine (CHX), cetylpyridinium chloride (CPC), or triclosan. Apart from the real-time monitoring of cell damage, we further applied a phenotypic fingerprinting method to differentiate between the bacterial subpopulations that arose due to treatment. We quantified the pathobiont damage rate of different antiseptics at different concentrations within 15 minutes of exposure and identified the conditions under which the bacteria were most susceptible. Moreover, we detected species-specific and treatment-specific phenotypic subpopulations. This proves that real-time flow cytometry can provide information on the susceptibility of different microorganisms in a short time frame while differentiating between antiseptics and thus could be a valuable tool in the discovery of novel antimicrobial compound, while at the same time deciphering their mode of action.

IMPORTANCE

With increasing evidence that microorganisms are becoming more tolerant to standard antimicrobials, faster and more accessible antimicrobial susceptibility testing methods are needed. However, traditional susceptibility assays are laborious and time-consuming. To overcome the abovementioned limitations, we introduce a novel approach to define antimicrobial susceptibility in a much shorter time frame with the use of real-time flow cytometry. Furthermore, phenotypic fingerprinting analysis can be applied on the data to study the way antiseptics affect the bacterial cell morphology over time and, thus, gain information on the mode of action of a certain compound.

INTRODUCTION

Over the last decades, the administration of antiseptics in oral healthcare for the treatment and prevention of oral diseases, such as gingivitis and periodontitis, has been intensified. However, there is increasing evidence that microorganisms become more tolerant to antiseptics, and this phenomenon is often combined with increased resistance toward antibiotics (1, 2). This stresses the need for, on the one hand, better surveillance for the antiseptic tolerance (3, 4) and, on the other hand, the acceleration of novel antimicrobial discovery.

Traditionally, antimicrobial susceptibility testing includes broth micro-dilution assays or disk diffusion assays (5, 6) with which the minimum inhibitory concentration (MIC) is determined. These methods are confronted with limitations. They are based on the active growth of a bacterial strain under specific antiseptic concentrations and are therefore time-consuming as they need at least 24 hours of incubation from the antiseptic application until endpoint measurement. Furthermore, they cannot differentiate between bacteriostatic and bactericidal conditions. To identify bactericidal conditions (determining the minimum bactericidal concentration—MBC), another 24 hours are required. Additionally, these bulk-based methods do not permit the detection of differences between bacterial subpopulations. Nonetheless, it is reported that even isogenic bacterial populations can harbor phenotypically variable subpopulations with different levels of tolerance toward antimicrobials (7, 8), and such a differential response can affect the treatment outcomes. Furthermore, the abovementioned methods are not informative about the mode of action of tested compounds. Consequently, we need rapid antimicrobial susceptibility testing methods that can further provide information on the subpopulation level or the compounds’ mode of action. These methods can facilitate better infection management, antimicrobial tolerance surveillance, and research for novel compounds.

Flow cytometry is a method used to detect antimicrobial susceptibility (9, 10). One of the main advantages of this method is that it provides a large amount of quantitative data in a short time frame as it can measure hundreds to thousands of cells in a few seconds. Besides the rapid measurement, flow cytometry protocols can further accelerate testing because they do not require long incubation times in contrast to the conventional methods, allowing to move from an incubation window of 24 to 48 hours to 1–2 hours (10, 11). Moreover, with the use of appropriate dyes, it can directly detect cell damage and lysis, thus enabling the study of bactericidal or bacteriolytic activity (12). In addition, the application of real-time flow cytometry could increase the time resolution as it enables the immediate observation of physiological changes of a given microbial subpopulation within a few seconds (13).

In this study, we evaluated real-time flow cytometry as an accurate and fast method to study the response of two oral pathobionts when exposed to antiseptics commonly used in oral care. Aggregatibacter actinomycetemcomitans was exposed to chlorhexidine (CHX) and cetylpyridinium chloride (CPC), while Streptococcus mutans was tested under the abovementioned antiseptics and triclosan. By exposing the bacteria to the antiseptics for 15 minutes, we evaluated the quantitative and qualitative responses in real-time under different concentrations of the antiseptics by detecting cell damage and observing the dynamics of the phenotypical subpopulations that arose during the exposure to the antimicrobials.

RESULTS

To demonstrate the use of real-time flow cytometry in assessing the susceptibility of oral bacteria to antiseptics, two species with different physiological properties that are linked to oral diseases were incorporated in the study. More specifically, Streptococcus mutans, a Gram⁺ coccus linked to dental caries (14), and Aggregatibacter actinomycetemcomitans, Gram^- coccobacillus that has been associated with more aggressive forms of periodontitis (15), were used. The bacteria were subjected to different antiseptics, commonly found in oral care products, at different concentrations. Α cell membrane permeability staining protocol was employed to distinguish between intact and damaged cells. SYBR Green I was used to stain all cells, while propidium iodide, which only penetrates cells with a disrupted cell membrane, was used to stain damaged cells. After staining, the cells were exposed to the respective antiseptics and continuously measured by flow cytometry for 15 minutes. Due to the antiseptic-induced cell damage, the cells were gradually stained with propidium iodide, thus allowing the real-time observation of cell damage.

Real-time determination of the cell damage rates

Our first objective was to measure the cell damage rate for each treatment, i.e., antiseptic type and concentration. We quantified the number of intact and damaged cells over time and segmented the 15-minute continuous data in smaller time frames of 30 seconds. Two gates were drawn manually in the green (FL1) vs red fluorescence intensity (FL3) density plots corresponding to the intact and damaged cell populations for each species (Fig. S1). To define the gates, we used a nontreated sample as a control for the intact population and a heat-killed sample as a control for the damaged population. Subsequently, these gates were used to count the number of intact and damaged cells in each 30-second frame.

The number of intact cells was used to further calculate the cell damage rate. More specifically, we expressed the data as the percentage of surviving cells, which corresponds to the ratio of intact cells at each time point over the intact cells from the nontreated control samples (Fig. 1 and 2). To define the rate of cell damage of each treatment, log-logistic models were fitted on the surviving cell percentage data. The Hill coefficient (slope) and the effective time 50 (ET50), calculated based on the model, were used to evaluate the effect of each treatment on the survival of the bacterial cells. The Hill coefficient describes the steepness of the curve, while ET50 indicates the time after which 50% of the cells were damaged. A higher Hill coefficient means that the slope is steeper, which corresponds to a faster response. A smaller ET50 value means that less time was needed to obtain half of the bacterial population in a damaged state.

Fig 1.

The percentage of surviving cells (intact cells) of A. actinomycetemcomitans over the time of 15 minutes for two different antiseptics, chlorhexidine (CHX) and cetylpyridinium chloride (CPC) in three different concentrations (2 mg/mL, 10 mg/mL, and 20 mg/mL). The points represent the average of three replicates, while the ribbon represents the sample standard deviation. The vertical lines represent the ET50 time point.

Fig 2.

The percentage of surviving cells (intact cells) of S. mutans over the time of 15 minutes for two different antiseptics, chlorhexidine (CHX), cetylpyridinium chloride (CPC), and triclosan in two different concentrations (2 mg/mL and 20 mg/mL). The points represent the average of three replicates, while the ribbon represents the sample standard deviation. The vertical lines represent the ET50 time point.

Although CHX caused faster cell damage on A. actinomycetemcomitans cells, CPC could cause more damage over time when 2 and 10 mg/mL concentrations were applied. After choosing a three-parameter log-logistic model as the best-fitted model (Fig. S3A) and based on the slope steepness, we observed that the CHX treatment, independent of concentration, exhibited a lower Hill coefficient and less steep killing curve than the CPC treatment. The Hill coefficient for CHX was increasing with increasing concentrations, while the opposite trend was true for CPC (Fig. 3). On the other hand, all CHX treatments exhibited a lower ET50 than equal concentrations of CPC (Fig. 3B). Besides calculating the time–response curves for different treatments, the dose–response curve was calculated for samples that were exposed to treatment for 10 minutes. This time was chosen to ensure that sufficient time from the exposure had passed. A three-parameter log-logistic model was used to calculate the Hill coefficient and the ED50, the dose for which 50% of cell damage should be observed (Fig. S3B). CPC had an ED50 of 1.5 mg/mL, while CHX had an ED50 of 1.9 mg/mL (Table 1), indicating that a slightly lower concentration of CPC is needed to reach 50% of cell damage after 10 minutes of treatment.

Fig 3.

A. The Hill coefficient (slope of the curve) and B. median effective time (ET50) as they were calculated based on time-dependent log-logistic models that were fitted in the cell damage curves of A. actinomycetemcomitans and S. mutans for each treatment. The values for S. mutans exposed to CPC could not be calculated as all cells were damaged in the beginning of the measurement. The values for the same species exposed to 2 mg/mL triclosan are not depicted as no killing was observed. The points represent the mean value of three biological replicates, while the error bars represent the standard deviation.

TABLE 1.

The mean and standard deviation of the Hill coefficient (slope of the curve) and median effective dose (ED50) as they were calculated based on dose-dependent log-logistic models that were fitted in the cell damage curves of A. actinomycetemcomitans

Species	Antiseptic	Time (min)	Hill coefficient	ED50 (mg/mL)
A. actinomycetemcomitans	CHX	10	1.29 (0.71)	1.9 (0.73)
A. actinomycetemcomitans	CPC	10	0.83 (0.5)	1.52 (1.06)

Different patterns were observed when S. mutans was exposed to the same antiseptics. Besides, this strain was also exposed to triclosan, which has a different mode of action than CHX and CPC. Exposure to CPC, either 2 or 20 mg/mL, led to immediate damage of most of the cells (< 99%) (Fig. 2). Consequently, we could not fit a time–response curve on these data, and these conditions were not used in the next steps of the analysis. A four-parameter log-logistic model was used to calculate the rate of cell damage under treatment with the other two antiseptics (Fig. S3C). A higher Hill coefficient was observed for 20 mg/mL CHX compared to 2 mg/mL (Fig. 3A). The Hill coefficient of 2 mg/mL triclosan was almost 0, (0.02 ± 0.12) because no cell damage occurred during this treatment. Concerning the dose-dependent data, a good model that would properly describe the phenomenon could not be fitted, probably because of the lack of a range of concentrations that will capture different phases of killing.

Another gate representing the total bacterial cells, thus separating both intact and damaged populations from the background, was used to evaluate whether the tested conditions lead only to cell damage or also to cell lysis. For most of the tested conditions, minimal cell lysis was observed (< 10%) (Fig. S4 and S5). However, 20 mg/mL triclosan rapidly caused cell lysis of the S. mutans cells (Fig. S5).

Flow-cytometric phenotypic fingerprinting

In the previous section, real-time flow cytometry data were used to calculate the effectiveness and rate of permeabilization of different antiseptic treatments by dividing the cell population in intact and damaged subpopulations. Nevertheless, we noticed that the response to the different antiseptics was much more dynamic than could be captured by this binary classification and that intermediate physiological phenotypes appeared through time. To consider this information in our analysis, we employed an alternative approach that is based on the phenotypic fingerprint of the samples. Instead of the binary split in the manually designed gates, a Gaussian mixture model was used to identify different physiological subpopulations (phenotypes).

After denoising the data based on the “total bacteria” gate (Fig. S2), a representative subset of samples was chosen to estimate the parameters for the Gaussian mixture model. The information of four parameters, (i) forward scatter (FSC), (ii) side scatter (SSC), (iii) green fluorescence intensity (FL1), and (iv) red fluorescence intensity (FL3), was used, and the model was set to identify 20 phenotypes (i.e., Gaussian mixtures). Including the forward and side scatter measurements on the model provided an extra layer of information about the morphological differences between the subpopulations. By observing the mean value of each phenotype in the four channels, we found that separation was based on both fluorescence intensity and scatter information (Fig. 4). Subsequently, the model was used to calculate the abundance of the different phenotypes in all time points.

Fig 4.

Heat map representing the mean values (a.u.) for each parameter (FL1, FL3, FSC, and SSC) for the 20 phenotypes as they have been predicted by the Gaussian mixture model, when allowed to identify a maximum of 20 phenotypes using 1,000 cells per sample as a training set. Phenotypes with higher red fluorescence (FL3-H) represent damaged cells, while phenotypes with high green (FL1-H), but low red fluorescence, represent intact phenotypes. The rest of the phenotypes represent intermediate states, most of which appeared during treatment. The color panel indicates whether the phenotypes were more abundant in A. actinomycemcomitants samples (purple) or in S. mutans samples (orange) or in both (green), and the phenotypes have been named accordingly.

The phenotypes clustered in two main groups according to the mean values of the four parameters (Fig. 4), and the members of one group (named Aa phenotypes 1–8) were more abundant in the A. actinomycetemcomitans samples, while the members of the other group were more abundant in the S. mutans samples (named Smu phenotype 1–9), except from three phenotypes that were abundant in samples of both species (named Aa/Smu phenotypes 1–3). Aa phenotypes 3, 5, and 6 describe the intact A. actinomycetemcomitans under no stress conditions. We observed a transient shift from these phenotypes with higher green fluorescence to the newly appearing phenotypes, when A. actinomycetemcomitans was exposed to the antiseptics. In the samples treated with CHX, these phenotypes had similar values of green and red fluorescence (Aa/Smu phenotypes 1 and 2). The shift to these newly appearing phenotypes was slower or faster depending on the antiseptic concentrations (Fig. 5). We also observed that Aa phenotypes 7 and 8 appeared only under 10 and 20 mg/mL CPC and in a very low abundance under 2 mg/mL but were not present under treatment with CHX. Finally, it is noteworthy that the heat-killed cells clustered in two completely different phenotypes (Aa phenotype 1 and Aa/Smu phenotype 3) and not together with the cells damaged by antiseptics.

Fig 5.

The relative abundances of phenotypic subpopulations in the different treatments and concentrations of A. actinomycetemcomitans as they have been estimated based on the Gaussian mixture model, which allowed to identify a maximum of 20 phenotypes and trained in a subsection of the data using 1,000 cells per sample. The bars represent the average of three biological replicates. The information about the mean value of the two fluorescence parameters was used to accordingly color the phenotypes. Samples with high red fluorescence are in different shades of red. Samples with high green and low red fluorescence are in shades of green. Samples with low green and red fluorescence are in shades of yellow, while samples of median green and red fluorescence are in shades of orange.

Intact nonstressed S. mutans cells were clustered in different phenotypes, with the most abundant being Smu phenotypes 1, 2, and 3. When S. mutans was exposed to CHX, different phenotypes appeared over time, moving from ones characterized from higher green fluorescence to ones with higher red fluorescence (Fig. 6), while their abundances were dependent on the antiseptic concentration. Interestingly, a range of different phenotypes was also observed in the conditions that all cells were clustered as damaged according to the previous method (Fig. S7). This was clearer, for the CPC-treated S. mutans cells were all damaged from the first time point. However, differences between the damaged cells were observed with this method (Fig. 6), with, for example, Smu phenotype 5 being the most abundant under treatment with 20 mg/mL CPC and mainly found in this treatment. In addition, we observed that Smu phenotypes 5 and 6 that mostly characterized 20 mg/mL CPC treatment clustered together with Smu phenotype 4, more abundantly in 20 mg/mL CHX, based on their mean values (Fig. 3). The latter probably indicates that high concentrations of the two antiseptics induce similar but distinct cell subpopulations.

Fig 6.

The relative abundances of phenotypic subpopulations in the different treatments and concentrations of S. mutans as they have been estimated based on the Gaussian mixture model, which was trained in a subsection of the data using 1,000 cells per sample. The bars represent the average of three biological replicates. The information about the mean value of the two fluorescence parameters was used to accordingly color the phenotypes. Samples with high red fluorescence are in different shades of red. Samples with high green and low red fluorescence are in shades of green. Samples with low green and red fluorescence are in shades of yellow, while samples of median green and red fluorescence are in shades of orange.

Hence, with cytometric fingerprinting, we could observe the dynamic shift from phenotypes that were describing intact cells to a range of damaged cell phenotypes. Most importantly, we detected phenotypes that were either compound-specific or concentration-specific.

DISCUSSION

Flow cytometry is a quick and reliable method for determining the microbial susceptibility of clinically important bacterial isolates (9, 12, 16), for detecting bactericidal conditions (12) and for accelerating prognosis in a clinical setting from 1 to 2 days to a couple of hours (10, 11). The combination of viability dyes (Syto9/PI) (12, 17) or membrane polarization dyes [DiOC_n(3), DiBAC₄(3)] (9) allows corroborating membrane permeabilization and pore formation under certain stress conditions. Different methods are in place to quantify the antimicrobial-induced changes, such as manual gating (12) or averaging ratios of the green and red fluorescence intensity to calculate the percentage of damaged cells (17). Yet, our current approach of combining dynamic flow cytometry analysis with SGPI staining allowed us to achieve unprecedented real-time monitoring of membrane permeabilization of the bacterial population of two oral species during exposure to three different antiseptics.

Initially, the killing dynamics upon exposure to a certain antiseptic were calculated based on the number of intact cells through time. A. actinomycetemcomitans was exposed to two different antiseptics, CHX and CPC, in three concentrations that were ~1,000 x above the MIC value for the two compounds (Table S1). It was observed that CHX was faster in damaging the bacterial cells, as indicated by the ET50 values (Fig. 3). However, the rate of damage was decreasing over time, as this was calculated by the Hill coefficient (Fig. 3). Contrastingly, when the two antiseptics were evaluated based on the concentration needed to decrease the cell population by half, it was found that slightly less CPC was needed to achieve the same effect (Table 1). Our findings suggest that, although CHX treatment acted more rapidly than CPC in the beginning, CPC was slightly more potent than CHX.

S. mutans cells exhibited three different patterns of response depending on the supplemented antiseptic. Immediate cell permeabilization was observed with CPC, gradual permeabilization over 15 minutes with CHX, and a biphasic response with triclosan, for which one concentration, namely, 20 mg/mL, led to immediate permeabilization, while 10 x lower concentration, 2 mg/mL, caused no cell damage after 15 minutes of exposure. Triclosan at low concentrations acts by blocking lipid synthesis via inhibition of enoyl-acyl carrier protein (ACP) reductase (FabI) (18) but at high concentrations seems to act simultaneously on multiple targets, such as inhibition of lipid, RNA, protein synthesis, and membrane perturbations (19). A different mode of action between lower and higher concentrations could be the reason for the biphasic response, which would require a longer time for cell permeabilization under lower concentrations as both tested concentrations are well above the MIC value for S. mutans (Table S1). It is also important to note that 20 mg/mL triclosan was the only treatment that induced cell lysis. Overall, the flow cytometric approach could detect bacterial damage already after 15 minutes of exposure for all conditions, except 2 mg/mL triclosan tested in S. mutans. The abovementioned finding corroborates the results of the traditional MIC test in a much shorter time frame (Table S1). Moreover, real-time flow cytometry provided additional information on the different responses between CPC and CHX and revealed that the highest concentration of triclosan caused lysis of S. mutans cells, which would not be possible to be determined with a traditional microdilution method.

Despite the substantial information that was acquired by this approach, some limitations must be considered. Short inactivation times of less than 1 minute prevented us from fitting a mathematical model to the antimicrobial effects, thus making the approach non-applicable for high concentrations of antiseptics. In addition, the subpopulations of intact and damaged cells were based on the two-dimensional space of green vs red fluorescence intensity, and the gating strategy was designed manually. Nonetheless, manual gating is subjected to bias (20), and antimicrobial stress can affect not only cell permeability but also cell morphology (21). Moreover, by splitting the cells into two subpopulations, intact or damaged, we enforced a discretization of the data, which in reality represented a continuous spectrum of phenotypes with intermediate levels of cell permeability. Hence, meaningful biological information was not considered with this approach.

To overcome the abovementioned limitations, i.e., avoid the bias of manual assigning populations and make use of more dimensions, a cytometric fingerprinting approach was subsequently applied to the data. Fingerprinting techniques have been successfully used in the past to discriminate between different bacterial strains, the same strain in different physiological conditions, or upon exposure to different antimicrobials (22–24). They are superior to manual gating because they are not subjected to user bias (20), they allow the observation of phenotypical changes that might not be captured by the use of a binary system for intact/damaged cell classification (20), and they can take advantage of the multiparametric measurements using both fluorescence and scatter information. In the past, very few efforts have been made to combine more than two dimensions in susceptibility studies by flow cytometry, such as the study by Huang et al. (16), which use the three-dimensional data to calculate the change between control and treated conditions as a one-dimensional distance (16).

In this study, we used a Gaussian mixture model approach to partition our data into different phenotypic subpopulations, which allows for the use of a smaller number of phenotypes as compared to the previous approaches (25) and thus can better reflect the complexity of the physiological changes and cell damage caused by the treatment. Four parameters were considered for identifying different phenotypes (forward and side scatter and green and red fluorescence intensity), which provided information on the morphology, nucleic acid content, and the membrane permeability of every single cell. Both fluorescence intensity and scatter played a role in the separation of the phenotypes. This is in accordance with previous studies that have shown that stress, which results in cell wall damage, can alter the morphology of a bacterial cell (21, 26).

Applying the Gaussian mixture model flow cytometric fingerprinting method increased the resolution of our data analysis and revealed further information on how each treatment affected the phenotypic fingerprint of the two bacterial populations. Our results indicate that the relative abundances of the different phenotypes were dependent on the (i) bacterial species, (ii) antiseptic and concentration of antiseptics used, and (iii) the time after the application of the treatment. This method also revealed how the cells pass through different phenotypical stages with different values of cell permeabilization and cell morphology, revealing the dynamic process of antimicrobial action (Fig. S5 and S6). We noticed that the two main groups, in which the phenotypes were clustered (Fig. 4), could be linked to one or the other species, which suggests that phenotypic differences of the two species are even larger than their state after antiseptic treatment. This could be explained by the lower forward and scatter values of the phenotypes mainly abundant on A. actinomycetemcomitans samples, which suggests a smaller size of this species. Additionally, the distinct phenotypes between antiseptic-treated A. actinomycetemcomitans and S. mutans could be explained by the differential action of the antiseptics against Gram⁺ and Gram^- species, such as the location where the cell wall is disrupted (27). These findings indicate that the method could be applied for species identification together with susceptibility testing, which might be an advantage in the case of clinical isolates. It must be noted that this study includes only two oral strains, when the oral cavity is inhabited by a wide range of both commensal microorganisms and pathobionts. When evaluated with the traditional methods, it has been observed that strains exhibit variable tolerance on oral antiseptics independently of their relevance to human health (28). This comes as no surprise as the response to an antiseptic depends on cell morphology, cell size, the number of porins, and cell membrane lipid composition (29). Thus, to acquire a more complete view of the arising phenotypical changes in different members of the oral cavity, follow-up studies should include a broader range of microorganisms from across the taxonomical spectrum.

Some phenotypes only appeared under one of the antiseptic treatments and not another, such as Aa phenotypes 7 and 8 that appeared only under CPC treatment of A. actinomycetemcomitans and Smu phenotype 5 under CPC treatment of S. mutans. We hypothesize that these distinct phenotypes could be linked to differences in the mode of action of the compounds. CHX is a bisbiguanide compound, and CPC is a quaternary ammonium compound. Both of the compounds are positively charged, and their mode of action is linked to their ability to bind in the negatively charged cell membrane, destabilizing the membrane and creating pores (30). The only difference between the two compounds is that the hydrophobic region of CPC becomes solubilized within the hydrophobic core of the cell membrane, while the hydrophobic region of chlorhexidine does not (31). Therefore, no major differences in the damaged phenotype would be expected from the use of the two compounds. However, for both species, it was apparent that CHX treatment resulted in phenotypes with higher red fluorescence than the ones that appeared under CPC treatment (Fig. 3). Different pore sizes induced by each antiseptic could lead to the incorporation of different amounts of propidium iodide in the cells and could potentially explain the distinct phenotypes. Concerning triclosan-treated S. mutans, when damage was observed, 20 mg/mL triclosan, the most abundant phenotype was Aa/Smu phenotype 3, which was not so abundant under treatment with CPC and CHX. Nevertheless, the mode of action of triclosan in high concentrations is not clear. Generally, information on the exact mode of action is lacking for many antiseptics, and the above-described pipeline could help in understanding the physiological changes that antiseptics induce in microbial cells. Furthermore, flow cytometric fingerprinting can be used to detect persister subpopulations, by distinguishing phenotypes whose cell concentrations do not change over time. Identifying antimicrobial compounds that will specifically target persister subpopulations can be of major importance as the increased tolerance of those subpopulations to antimicrobials hinders the efficacy of antimicrobial chemotherapy (32).

In general, in the present study, we demonstrated the applicability of real-time flow cytometry for the study of antiseptic susceptibility. By using two different approaches, our analysis provided information on the rate of cell damage under different concentrations of antiseptics and allowed the comparison between treatments, while at the same time permitted the observation of the physiological changes induced by the different compounds over 15 minutes of exposure. We propose that the potential of this method can be further strengthened with the use of alternative staining dyes, e.g., membrane polarity stains, that can detect different physiological changes. Additionally, this method can be applied for other environmental stressors with clinical or biotechnological relevance that induce phenotypic changes, e.g., osmotic stress, UV, and temperature. The temporal resolution can prove extremely beneficial to observe and quantify the dynamics at high concentrations of the stressor where the cell response is instantaneous, and thus distinct time points are insufficient to capture the intermediate phenotypic states. We believe that our study may trigger the broader use of real-time flow cytometry in the study of antimicrobial susceptibility as it could be applied for the detection of resistant strains or in the quest for novel antimicrobial compounds.

MATERIALS AND METHODS

Strains and growth conditions

Aggregatibacter actinomycetemcomitans ATCC 43718 and Streptococcus mutans ATCC 25175 were used in all described experiments. The strains were maintained on blood agar No2 (Oxoid, Hampshire, UK) supplemented with hemin (5 mg/mL) (Sigma Aldrich, Belgium), menadione (1 mg/mL) (Sigma Aldrich, Belgium), and 5% sterile horse blood (Oxoid, Hampshire, UK) or cultured in liquid medium in brain heart infusion (BHI) (Carl Roth, Belgium) broth under aerobic conditions at 37°C.

Flow cytometric measurements

Sample preparation and treatment

After overnight culture in the conditions that were previously described, the bacterial cultures were measured by flow cytometry and SYBR Green/ propidium iodide (SGPI) staining to determine the intact cell concentration. In more detail, the samples were diluted in PBS, sterile and filtered through 0.2-um pore filter, (PBS tablet, Merck, Belgium) and stained with the nucleic acid stain SYBR Green I and propidium iodide that stains permeabilized cells (30). SYBR Green I (10,000X concentrate in DMSO, Invitrogen) was diluted 100 times in 0.22-µm-filtered DMSO (IC Millex, Merck, USA), and propidium iodide (20 mM in dimethyl sulfoxide (DMSO), Invitrogen, USA) was diluted 50 times. Samples were stained with 10 µL/mL staining solution. Next, they were incubated at 37°C for 20 minutes and measured with a benchtop Accuri C6 +cytometer (BD Biosciences, Belgium).

Overnight cultures were subsequently diluted in sterile dH₂O (for S. mutans) or sterile Evian bottled water (Evian, France) (for A. actinomycetemcomitans) to a final concentration ~1×10⁶ cells/mL according to the previous measurement and further stained with SGPI with the abovementioned approach. Just before measurement, the corresponding concentrations of chlorhexidine digluconate (Merck, Belgium), cetylpyridinium chloride (Carl Roth, Germany), or triclosan (Merck, Belgium) were added to the diluted culture and mixed well. A. actinomycetemcomitans was subjected to CHX or CPC in 2 mg/mL, 10 mg/mL, or 20 mg/mL. S. mutans was subjected to CHX, CPC, and triclosan either in 2 mg/mL or in 20 mg/mL.

Sample measurement/ instrument settings

All samples were measured with a benchtop Accuri C6 +cytometer (BD Biosciences, Belgium). The stability of the instrument was verified daily using CS&T RUO beads (BD Biosciences, Belgium). The blue laser (488 nm) was used for the excitation of the stains. The filters for the (fixed gain) photomultiplier detectors used during the measurements were 533 nm with a bandpass filter of 30 nm for the green fluorescence (FL-1) and 670 nm longpass filter for the red fluorescence (FL-3). The threshold was set on the 533/30 nm (FL-1) detector at the arbitrary unit of 1,200 and on the 670 nm(FL-3) at the arbitrary unit of 1,200. The threshold was decided based on the measurements of control samples (growing culture, heat-killed culture, medium, and dH₂O/diluent) in order to avoid background noise and allow for maximum measurement of total events in the same acquisition. Sample acquisition took place at a flow rate of 66 µL/minute continuously at a time of 15 minutes.

Data analysis

Data filtering and cleaning

Flow cytometric data were analyzed in R (version 4.0.3). The Phenoflow’s (v1.1.2) (33) function “time_discretization” was used to concatenate the files into smaller files of fixed time frames of 30 seconds.

Two different gating strategies were used for each species to gate a. intact cells, b. damaged cells, and c. total cells (Fig. S1 and S2). Nontreated and heat-killed samples were used as control samples to define the intact and damaged cell gates.

Time– and dose–response analysis

Intact cell numbers, based on the “intact cells” gate, were used for calculating the killing rate under the different antimicrobials. First, the percentage of surviving cells was calculated as the ratio of intact cells at each time point over the average intact cells of nontreated control samples. Subsequently, the “drm” function from the drc package (v3.0.1) (34) was used to fit different log-logistic models to the time-effect or dose–response data. For all log logistic models, the maximum asymptote was constraint to 100. The fit of the different models was evaluated with the “mselect” function of the drc package to identify the model that best fitted the data based on the Akaike’s information criterion (AIC) and the lack-of-fit test (against a one-way ANOVA model). The model with the lowest AIC and highest P-value in the lack-of-fit test was chosen. The Hill coefficient and the effective time 50 or effective dose 50 were calculated from the model. Effective time/dose 50 is the time or dose for which 50 per cent of killing is reached.

Phenotypic subpopulations PhenoGMM

The “PhenoGMM” function (24) of the PhenoFlow package was used to determine phenotypes to which cells can be assigned. Initially, the background was removed by applying the “total cells” gate (Fig. S2). Subsequently, the “PhenoGMM” function was applied on a representative subset of the data, using the FSC-H, SSC-H, FL1-H, and FL3-H parameters as the input. The fcs samples were resampled with replacement to 1,000 events. The best number of mixtures/phenotypes to describe the data set was chosen from 1:20 based on the Bayesian information criterion (BIC). Increasing the number of allowed phenotypes was leading to a higher number by the BIC, but eventually the 20 phenotype limit was chosen as a number that allowed to capture the data variation without increasing the complexity and result interpretation. After the Gaussian mixture model was fitted, it was used to calculate the abundance of each mixture in the total data set.

Minimum inhibitory concentration (MIC) determination

Minimum inhibitory concentration of the different antiseptics for the two bacteria was determined by the broth microdilution assay. In more detail, the three antiseptics were diluted via serial dilutions in fresh BHI in a flat-bottom 96-well plate (Greiner, Belgium). S. mutans and A. actinomycetemcomitans were added in a final concentration of 10⁶ cells/mL or 10⁷ cells/mL in each well. 10⁷ cells/mL was chosen for A. actinomycetemcomitans because it was found that the minimal inoculation concentration allows for growth under control (no antiseptic) conditions. Cell concentrations were previously determined by flow cytometry following the procedure that was explained previously.

The cultures were incubated at 37°C for 24 hours, with A. actinomycetemcomitans incubated under anerobic conditions. After incubation, growth was defined by measuring OD₆₀₀ with a plate reader (Tecan Infinite M200 PRO, Tecan, Belgium). The MIC was defined as the lowest concentration where no growth was observed.

DATA AVAILABILITY

Flow cytometry data (.fcs format) are available on the FlowRepository archive under repository ID FR-FCM-Z4VR.

SUPPLEMENTAL MATERIAL

The following material is available online at https://doi.org/10.1128/spectrum.00955-24.

Supplemental material. spectrum.00955-24-s0001.pdf.

Fig. S1 to S7; Table S1.

ASM does not own the copyrights to Supplemental Material that may be linked to, or accessed through, an article. The authors have granted ASM a non-exclusive, world-wide license to publish the Supplemental Material files. Please contact the corresponding author directly for reuse.