A Nutrient-Tunable Bistable Switch Controls Motility in Salmonella enterica Serovar Typhimurium

ABSTRACT Many bacteria are motile only when nutrients are scarce. In contrast, Salmonella enterica serovar Typhimurium is motile only when nutrients are plentiful, suggesting that this bacterium uses motility for purposes other than foraging, most likely for host colonization. In this study, we investigated how nutrients affect motility in S. enterica and found that they tune the fraction of motile cells. In particular, we observed coexisting populations of motile and nonmotile cells, with the distribution being determined by the concentration of nutrients in the growth medium. Interestingly, S. enterica responds not to a single nutrient but apparently to a complex mixture of them. Using a combination of experimentation and mathematical modeling, we investigated the mechanism governing this behavior and found that it results from two antagonizing regulatory proteins, FliZ and YdiV. We also found that a positive feedback loop involving the alternate sigma factor FliA is required, although its role appears solely to amplify FliZ expression. We further demonstrate that the response is bistable: that is, genetically identical cells can exhibit different phenotypes under identical growth conditions. Together, these results uncover a new facet of the regulation of the flagellar genes in S. enterica and further demonstrate how bacteria employ phenotypic diversity as a general mechanism for adapting to change in their environment.

genes. Conversely, when the nutrient level is low, YdiV expression is enhanced and expression of the motility genes is repressed. The cAMP-CRP complex also regulates the transcription of the flhDC operon in S. enterica (9), although YdiV apparently masks the effect, at least under the conditions in which these experiments were performed.
YdiV participates in a double-negative feedback loop involving the flagellar regulator FliZ. FliZ directly represses ydiV transcription, and YdiV indirectly represses fliZ transcription through FlhD 4 C 2 ( Fig. 1) (10). FliZ and YdiV have also been shown to influence the population dynamics of flagellar gene expression. In particular, multiple studies have observed coexisting populations of motile and nonmotile cells (11)(12)(13)(14); however, these coexisting populations are not observed in ⌬fliZ (12) or ⌬ydiV (13) mutants. Based on these results, FliZ and YdiV have been hypothesized to function in a genetic on-off switch, causing some cells to be motile and others not (15).
Although this mechanism is appealing, it has yet to be proven. Moreover, we previously found that these coexisting populations are transient, so that the entire population eventually becomes motile (12). However, these experiments were performed in rich media. One aspect that has yet to be explored is the role that nutrients play in shaping this dynamic response.
In this study, we investigated how nutrients tune flagellar gene expression dynamics in S. enterica. Our results demonstrate that nutrients tune the fraction of motile cells. While coexisting populations are observed at all nutrient concentrations, they persist only at intermediate nutrient concentrations. Using a combination of experimentation and mathematical modeling, we further investigated the mechanism that governs this tunable response. In support of previous models, we found that FliZ and YdiV are necessary for the response. In addition, we found that the positive feedback loop involving FliA is required, although its role appears to be solely to enhance FliZ expression. Together, these results reveal a new facet of motility and flagellar gene regulation in S. enterica.

Nutrients tune the fraction of motile cells. Previous experiments
investigating the dynamics of flagellar gene expression in S. enterica were performed in rich Luria-Bertani (LB) medium. Based on the recent discovery that nutrients tune YdiV expression (6), we hypothesized that nutrients may also tune the fraction of motile cells. To test this hypothesis, we grew cells in Vogel-Bonner medium E (16) supplemented with 0.2% (wt/wt) glucose and various concentrations of yeast extract. The cells were harvested during late exponential phase, and their swimming behavior was analyzed by video microscopy. Consistent with our hypothesis, we found that nutrients, specifically yeast extract, tune the fraction of motile cells ( Fig. 2A). We also performed growth experiments. Except in its complete absence, we found that the concentration of yeast extract does not strongly affect the growth rate (see Fig. S1 in the supplemental material). These results show that the response to yeast extract at the concentrations tested is not determined by the growth rate but is instead regulated by nutrient availability.
Nutrients tune the fraction of cells expressing flagellar genes. We next investigated whether regulation occurs at the level of gene expression. Flagellar genes can be divided into a transcriptional hierarchy comprising three classes (Fig. 1). We used flow cytometry to measure the expression from a representative promoter from each hierarchical class, using single-copy, chromosomally integrated transcriptional fusions to the fluorescent protein Venus (17). The class 1 P flhDC promoter was active in all cells irrespective of yeast extract concentrations (Fig. 2B). However, we found that the class 2 P flhB and class 3 P fliC promoters were active in only a subpopulation of cells at intermediate yeast extract concentrations, giving rise to a bimodal distribution ( Fig. 2C and D). The active fraction increased with yeast extract concentrations in a manner consistent with the video microscopy experiments ( Fig. 2E and F). We note that nutrients increased the relative expression of the P flhB and P fliC promoters in cells where the promoters were active, though the effect is minor. As the P flhB and P fliC promoters are nearly identical in their response to yeast extract, our remaining investigations focused on the class 2 P flhB promoter.
In dynamic gene expression experiments, coexisting populations of cells with active and inactive promoters were transiently observed at all yeast extract concentrations (Fig. 3). However, the population with inactive promoters persisted only at low yeast extract concentrations. At the higher concentrations, the promoters in all cells eventually became active. These results are consis-  (56,57). The sole class 1 operon encodes the FlhD 4 C 2 master regulator. FlhD 4 C 2 activates the expression of class 2 operons, which encode the hook-basal-body (HBB) proteins. In addition, FlhD 4 C 2 activates the expression of the alternate sigma factor, FliA (also known as 28 ), and FliZ. FliA, in turn, activates the expression of the class 3 operons, which encode the motor proteins (MotAB), flagellar filament (FliC), and chemotaxis pathway (Che). YdiV binds FlhD 4 C 2 and prevents it from activating class 2 promoters (6). In addition, YdiV promotes the degradation of FlhD 4 C 2 via ClpXP (7). Both nutrients and FliZ repress the expression of YdiV (10). tent with our previous findings, in which we observed transient heterogeneity in nutrient-rich media (12). At the lower concentration of yeast extract, the coexisting populations persist for many hours with no significant change in their distribution. These results suggest that the observed response is bistable: that is, genetically identical cells can exhibit different phenotypes that persist under identical growth conditions. Multiple nutrients activate flagellar gene expression. Yeast extract is a complex mixture containing many nutrients. We tested a number of different compounds for their ability to activate the P flhB promoter (Table 1). Among the compounds tested, we found that amino acids were able to activate the P flhB promoter, although not to the same degree that yeast extract was able to. Analysis of individual amino acids suggested that most were able  to activate the P flhB promoter weakly, whereas a combination of all twenty was able to activate the P flhB promoter to roughly half the level achieved with yeast extract. These results indicate that the activating signal is not a simple compound but rather a mixture of them, of which amino acids are a subset. We also note that previous studies have shown that that the RNA-binding protein CsrA, which is involved in carbon storage, regulates YdiV translation, though YdiV is still subject to nutritional regulation in a ⌬csrA mutant (6). CsrA is regulated by CsrB and CsrC, two noncoding RNAs that are transcriptionally regulated by the BarA/SirA twocomponent signal transduction system (18,19). One study found that formate and acetate regulate CsrB transcription through BarA and SirA, respectively (20). We also tested the ability of formate and acetate to activate the P flhB promoter and found that neither was able to do so ( Table 1). As we were unable to isolate a single activating compound, we employed yeast extract in the remainder of our studies.
YdiV and FliZ are necessary for bimodal flagellar gene expression. Both YdiV and FliZ have previously been shown to affect single-cell gene expression dynamics (12,13). To test how these two proteins contribute to the nutritional response to yeast extract, we measured flagellar gene expression in ⌬ydiV and ⌬fliZ mutants. In a ⌬ydiV mutant, the P flhB promoter was strongly active in most cells, irrespective of yeast exact concentration ( We note that there is a tail in the distribution, indicating that the P flhB promoter is weakly active in a small population of cells. This population was observed at all yeast extract concentrations and was also present in the wild type, even at the highest concentration of yeast extract employed ( Fig. 2C and D). In contrast, the ⌬fliZ mutant exhibited a homogeneous nutrient response consisting of a single population ( Fig. 4B; see also Fig. S2B in the supplemental material). We also found that ydiV was dominant, so that a ⌬ydiV ⌬fliZ double mutant was indistinguishable from the ⌬ydiV single mutant (see Fig. S2C in the supplemental material).
YdiV and FliZ are known to repress each other, with FliZ directly repressing ydiV transcription and YdiV indirectly repressing fliZ transcription via FlhD 4 C 2 ( Fig. 1) (10). To observe this competitive interaction at single-cell resolution, we employed twocolor flow cytometry to measure simultaneous expression from the P ydiV and P flhB promoters. Expression from the P ydiV promoter was measured using a transcriptional fusion to the cyan fluorescent protein (CFP). Note that the P ydiV -cfp fusion was active only when expressed from a low-copy-number (pSC101* origin of replication) plasmid; single-copy transcriptional fusions, like those employed for the P flhB and P fliC promoters, were not sufficiently active to measure a response for the P ydiV promoter. Figure 4C shows a comparison of P flhB and P ydiV promoter activities as a function of yeast extract concentrations. In the absence of yeast extract (0%), a single population was observed in  which the P ydiV promoter is active and the P flhB promoter is inactive. At intermediate concentrations of yeast extract (0.2 to 1%), two populations were observed: one in which the P ydiV promoter was active and the P flhB promoter was inactive and one in which the reciprocal pattern occurred. As yeast extract concentrations were increased, the relative number of cells occupying the population where P flhB promoter was active increased. At a high yeast extract concentration (Ͼ1%), only the P flhB active population was observed, although transcription from the P ydiV promoter was still detectable. These results suggest that the heterogeneous nutrient response arises from the mutual repression of YdiV and FliZ. We note that a previous study did not report any significant changes in ydiV transcription in response to nutrients. Instead, only changes in the level of YdiV protein were observed, suggesting that nutrients principally regulate YdiV via a posttranscriptional mechanism (6). We, on the other hand, found that yeast extract decreased expression from the P ydiV promoter, indicating that the mechanism involves a significant transcriptional component. One possible explanation for this discrepancy is that the decrease in ydiV transcription is due to FliZ, which is more strongly expressed at high yeast extract concentrations and which is known to bind the P ydiV promoter. We found, however, that yeast extract also decreased expression from the P ydiV promoter in a ⌬fliZ mutant (see Fig. S3A in the supplemental material). In both the wild type and the ⌬fliZ mutant, we observed more than a 2-fold decrease in promoter activity at high concentrations of yeast extract. When YdiV expression was measured using a translational fusion to superfolder green fluorescent protein (SGFP), we observed similar decreases in its expression (see Fig. S3B), indicating that YdiV is regulated at the transcriptional level.
Flagellar gene expression is bistable and exhibits hysteresis. Our data suggest that flagellar gene expression is bistable. As bistable systems exhibit hysteresis, which reflects history dependence (21), we wished to determine how cells transition between different states of flagellar gene expression. To test whether flagellar gene expression exhibits hysteresis, we first replaced the native P flhDC promoter with a tetracycline-inducible one, as described previously (22). We then grew cells in the presence and in the absence of 10 ng/ml anhydrotetracycline (aTc)-referred to here as the on and off states, respectively-prior to subculturing into fresh medium containing different aTc concentrations. We fixed the yeast extract concentration at 0.2%, as bistability is most pronounced at this concentration. If flagellar gene expression is bistable, then the response should be different for these two cases.
Consistent with a bistable response, we observed that cells exhibited different patterns of P flhB promoter activity depending on whether they were initially in the on or off state (Fig. 5A). In general, P flhB promoter activity was lower when cells were initially in the off state than when cells were initially in the on state. Bimodality was not observed when cells transitioned from an on to an off state ( Fig. 5B and C). We performed similar experiments using the ⌬ydiV and ⌬fliZ mutants. For both mutants, no hysteresis was observed: the response was the same irrespective of whether they had previously been induced with aTc ( Fig. 5A; see also Fig. S4 in the supplemental material). These results demonstrate that YdiV and FliZ are essential for the hysteresis response. Similar experiments were performed at different concentrations of yeast extract, and hysteresis was again observed (see Fig. S5A to C in the supplemental material). However, we cannot directly compare these experiments with one another, as the responses to yeast extract and aTc are not orthogonal: as yeast extract concentrations increase, expression from the tetracyclineinducible P tetA promoter (23) decreases, for unknown reasons (see Fig. S5D in the supplemental material). Despite this cross talk, the conclusions that flagellar gene expression is bistable and exhibits hysteresis do not change.
The FliA positive feedback loop is necessary for bistability. The flagellar network possesses two feedback loops in addition to the ones involving FliZ and YdiV. One is a negative feedback loop involving FliT. Expressed from a hybrid class 2/3 promoter (24), FliT binds to FlhD 4 C 2 and prevents it from activating its cognate class 2 promoters (25,26). The second is a positive feedback loop involving FliA and FlgM. The fliAZ operon is under the control of both class 2 and class 3 promoters (27). The class 2 promoter functions in a double-negative feedback loop involving FliZ and YdiV; the class 3 promoter functions in an autogenous loop involving the alternate sigma factor FliA. Note that the latter loop is not directly autocatalytic, as FliA is inefficiently translated from the class 3 transcript (28). Rather, positive feedback is indirect, such that FliA activates FliZ expression and FliZ indirectly activates FliA expression by repressing the expression of YdiV.
An additional facet of the regulation involves FlgM, which regulates FliA activity by binding to it and preventing it from activating its cognate class 3 promoters (29). FliA and FlgM function in a developmental checkpoint involving protein secretion (30). In addition, they are believed to function in a regulatory circuit involved in controlling the number of flagella produced per cell (31)(32)(33).
To determine whether these regulatory loops contribute to bistability, we examined the activity of the P flhB promoter in ⌬fliT, ⌬fliA, and ⌬flgM mutants. In the case of the ⌬fliT and ⌬flgM mutants, the response to yeast extract was similar to that of wildtype cells (Fig. 6A and B), demonstrating that neither gene product is required for bistability. The ⌬fliA mutant, on the other hand, exhibited a homogenous response to yeast extract (Fig. 6C) in a manner equivalent to a ⌬fliZ mutant. We also tested a P fliA :: P flhB promoter mutant, where the hybrid class 2/3 P fliA promoter was replaced with a pure class 2 promoter (33), and found that it also exhibited a homogeneous response to yeast extract (Fig. 6D). These results demonstrate that both FliA and the class 3 component of the P fliA promoter are essential for bistability.
Mathematical model reproduces bistability and provides insight regarding FliA. We compared two simple mathematical models of the flagellar gene network to gain insight regarding the mechanism for bistability (details and equations are provided in Text S1 in the supplemental material). The first model ignores the contribution of FliA and accounts only for the interaction between YdiV and FliZ. The second model includes FliA. Both models are semiquantitative, in the sense that they were formulated solely to capture the general trends in our data rather than to reproduce them exactly.
The first model is capable of generating bistability (Fig. 7). In our simulations, we considered two scenarios. First, we varied the level of YdiV expression in the model as a proxy for nutrient availability (Fig. 7A). For intermediate levels of YdiV expression, the model admits three steady-state solutions that represent possible states of flagellar gene expression (denoted by the shaded region in Fig. 7A). The upper and lower branches are stable and correspond to the on and off states, respectively. The middle branch is unstable and not physiological. In contrast, the model yields a single steady state when it is simulated for a ⌬fliZ mutant (gray line in the plot in Fig. 7A) or a ⌬ydiV mutant. In the latter, there is obviously no expression of YdiV. Simulations for different parameter values are provided in Fig. S6 in the supplemental material. Next, we performed simulations in which we varied the FlhD 4 C 2 expression rate while keeping the YdiV expression rate fixed (Fig. 7B). These simulations were used to mimic the experiments whose results are shown in Fig. 5. The model again reproduced the general features of our experimental data, including those for the ⌬fliZ and ⌬ydiV mutants (Fig. 5A). These results show that positive feedback via FliA is not necessary for bistability: the regulatory interactions between FliZ and YdiV suffice.
We next considered a model that includes FliA. This second model was also capable of generating bistability (see Fig. S7 in the supplemental material). The parameters were chosen to match the response of the first model. When we compared the two models and associated parameters, the key difference was that the first model required that FlhD 4 C 2 strongly activate the expression of FliZ to generate bistability, whereas the second model did not. Introducing FliA positive feedback relaxes the requirement for FliZ because the positive feedback functions to amplify FliZ expression by FlhD 4 C 2 (see equation 24 in Text S1 in the supplemental material). These results are consistent with the notion that the contribution of FliA is indirect and functions to enhance FliZ expression.

DISCUSSION
The findings presented here show that the fraction of motile cells in a population of S. enterica is determined, at least in part, by nutrient availability. YdiV and FliZ control two feedback loops with opposing activities that govern this response, as previously proposed (15). Although these feedback loops are necessary for bistability, they are not sufficient; a positive feedback loop involving FliA is also required. This third loop is not strictly autoregulatory, in the sense that FliA does not directly enhance its own expression.
To further understand the mechanism of bistability, we constructed a simplified mathematical model of the flagellar gene network. Analysis of this model demonstrated that, in principle, the two feedback loops controlled by FliZ and YdiV are sufficient for bistability. The positive feedback loop involving FliA, however, does not provide a critical feature for bistability and nutrient control. Rather, the FliA loop contributes to bistability by ensuring that expression of FliZ is strongly enhanced by FlhD 4 C 2 due to fliZ being transcribed from a hybrid class 2/3 promoter.
These features of nutrient regulation raise the question of why fliZ is transcribed from a hybrid class 2/3 promoter when expression from a strong class 2 promoter would suffice to generate bistability. One potential explanation is that the FliA feedback loop couples the switch to the completion of assembly of the hook-basal body complex. A developmental checkpoint involving FliA and FlgM couples class 3 flagellar gene expression to flagellar assembly (30). Prior to completion of the hook-basal body, FlgM binds to FliA and prevents it from activating class 3 promoters. Upon completion of the hook-basal body complex, FlgM is secreted from the cell by the flagellar export apparatus, freeing FliA to transcribe from class 3 promoters. Mutations that inhibit formation of the hook-basal body complex are deficient in FlgM secretion and thus prevent transcription from class 3 promoters. This checkpoint likely ensures that cells do not switch to the on state for class 3 transcription until they are able to make functional hook-basal body complexes. In support of this prediction, we discovered that gene expression is homogenous in hook-basal body mutants (12) and in a ⌬fliA mutant (Fig. 6C). Moreover, fliA and fliZ reside in the same operon in most bacteria that possess fliZ (34). Thus, coupling of fliZ and fliA may be necessary for the full function of FliZ to be realized.
In our work, we used yeast extract as the nutrient to tune flagellar gene expression. How yeast extract regulates the P ydiV promoter is not known. Our work demonstrates that amino acids are able to activate the P flhB promoter, but they are unlikely to be the only compounds that do so (Table 1). One additional possibility is that the P ydiV promoter responds to the growth rate of the cell.
Although our experiments provide no support for this possibility (see Fig. S1 in the supplemental material), a caveat is that we used relatively high concentrations of nutrients that supported approximately equal growth rates. If flagellar gene regulation also responds to the growth rate of cells, as one might expect, then our experiments would not have detected this phenomenon, as our conditions were chosen to keep the growth rate nearly constant.
The present study raises the question of why nutrients control bistable flagellar gene expression in S. enterica. One possible explanation derives from the fact that motility is intimately coupled to virulence and host colonization in S. enterica. Numerous studies have shown that flagellar gene expression is coupled to the expression of the invasion genes associated with the type III secretion system encoded within Salmonella pathogenicity island 1 (SPI-1) (35)(36)(37)(38)(39)(40). Of note, FliZ positively regulates the expression of the SPI-1 invasion genes. In addition, flagellin activates the innate immune response (41). The nutrient response likely ensures that flagellin is expressed at specific sites within the host, as previously shown (11). Bistability, as argued by Stewart and Cookson (15), may enable a division of labor and a degree of bet hedging, where motile cells are invasive and nonmotile cells noninvasive. Nonmotile cells could thus avoid the inflammatory environment of the intestinal epithelium and serve as a reservoir for the next phases of colonization. Our results extend this model by showing that nutrients control these fractions of motile and nonmotile cells.
Nutrients repress motility in E. coli, a form of regulation that is consistent with E. coli employing motility as a foraging mechanism. Only when starved of nutrients are these bacteria motile. However, recent results suggest that motility in E. coli is also employed for host colonization. For example, the bacterial quorum sensing signal AI-2 and interkingdom signaling molecule norepinephrine control motility in E. coli (42)(43)(44)(45). These results indicate that both S. enterica and E. coli employ motility for multiple purposes.
We note that the ydiV gene is also present in E. coli (46). Although the gene is transcriptionally active, it is poorly translated (28). However, the E. coli ydiV gene is efficiently translated in S. enterica, suggesting that some factor represses its translation in its native host. These results suggest that the flagellar gene networks in E. coli and S. enterica are quite plastic, in the sense that small changes in the expression of individual genes can result in significantly different responses to nutrients and perhaps to other environmental cues. Such plasticity in the regulatory pattern of the flagellar gene network possibly may reflect the disparate roles motility plays in these two closely related organisms and enable them to adapt readily to new environments in which these roles differ.
We conclude by noting that flagellar gene expression is also bistable in Bacillus subtilis (47), although the mechanism governing bistability is quite different than the one in S. enterica (48). In B. subtilis, the flagellum-specific sigma factor SigD resides at the end of a large operon containing thirty-one flagellar genes. Presumably, SigD expression is weak or nonexistent in many cells because of incomplete transcription of the full operon. In cells in which the entire operon is transcribed and expression of SigD exceeds some threshold, it can further enhance its own expression through a SigD-dependent promoter that resides in the middle of the operon. This positive feedback mechanism, which involves the stochastic triggering of the loop, generates the observed bistability in B. subtilis motility. Indeed, moving the sigD gene upstream in the operon increases the fraction of motile cells (48). Whether the bistable response in B. subtilis is tuned by external factors to the same degree as in S. enterica is not known. Nonetheless, the bistable response in these two distantly related bacteria suggests that heterogeneous expression of flagellar genes is a general phenomenon and may reflect a widespread strategy for deploying motility as an adaptational response to the environment.

MATERIALS AND METHODS
Media and growth conditions. All experiments were performed at 37°C in Vogel-Bonner minimal E medium (16) supplemented with 0.2% (wt/ wt) glucose and yeast extract at specified concentrations (MG medium). Luria-Bertani (LB) medium was used for strain and plasmid construction. Strains harboring plasmids pKD46, pCP20, and pINT-ts were grown at 30°C. Antibiotics were used for gene deletion and plasmid maintenance at the following concentrations: 40 g/ml kanamycin, 20 g/ml chloramphenicol, and 100 g/ml ampicillin.
Bacterial strains and plasmid construction. All strains are derivatives of S. enterica serovar Typhimurium 14028 (American Type Culture Collection). The ⌬fliA (region from 2056152 to 2056871), ⌬flgM (region from 1215209 to 1215502), and ⌬ydiV (region from 1432774 to 1433487) mutants were constructed using the method of Datsenko and Wanner (49). Prior to removal of the antibiotic marker, the integrated cassette was moved to a clean wild-type background by P22 transduction. The ydiV=cfp transcriptional fusion was made by introducing the ydiV promoter (region from 1431646 to 1432784) and CFP from pMUTIN-CFP (50) into the plasmid pZS*31 (51) using the XhoI, EcoRI, and XbaI restriction sites. The YdiV-SGFP translational fusion was made by amplifying ydiV with its native promoter (region from 1431646 to 1433484) and the SGFP gene (52) with the Gly-Gly-Ser-Gly linker using overlap extension PCR (53). The PCR product was then cloned into the pVenus plasmid (54) using the SalI and NheI restriction sites and finally integrated into the chromosome using the method of Haldimann and Wanner (55). The class 1 P flhDC promoter (region from 2034484 to 2033584), class 2 P flhB promoter (region from 2023494 to 2022815), and class 3 P fliC promoter (region from 2061043 to 2060527) were used as representatives for each hierarchical class of flagellar promoters. Single-copy transcriptional fusions in the chromosome were made by first cloning respective promoters into the plasmid pVenus (54) using KpnI and EcoRI restriction sites and integrating the plasmids into the chromosome using the CRIM method (55). The integrated plasmids were then moved into the wild type and the different mutants by P22 transduction. All strains and plasmids used in this work are listed in Table 2. Strain numbers are provided in the figure captions to match individual experiments to specific strain genotypes. Flow cytometry. Cells were grown overnight at 37°C in MG medium supplemented with 0.2% yeast extract. Cells were subcultured to an optical density at 600 nm (OD 600 ) of 0.05 in fresh MG medium supplemented with 0%, 0.2%, 0.5%, 1%, or 2% yeast extract. After subculture, the cells were then allowed to grow at 37°C for 4 h (unless noted otherwise) before harvesting. For dynamic gene expression experiments, samples were collected every hour for 7 h, centrifuged at 3,200 ϫ g for 10 min, resuspended in phosphate-buffered saline (PBS) containing 50 g/ml chloramphenicol, and kept on ice. The cells were pelleted by centrifugation at 3,200 ϫ g for 10 min and resuspended in DAPI (4=,6-diamidino-2-phenylindole) staining buffer with 14.3 M DAPI and 50 g/ml chloramphenicol. The cells were then incubated at room temperature for half an hour. The cells were then analyzed using a BD LSR II flow cytometer. Fluorescence values for approximately 100,000 cells were recorded using the Pacific Blue channel (excitation, 405 nm; emission, 450/50 nm) for DAPI and the fluorescein isothiocyanate (FITC) channel (excitation, 488 nm; emission, 530/ 30 nm) for Venus. The cells were distinguished from other debris by gating only the population stained with DAPI. For cells expressing both CFP and Venus, the cells were not stained with DAPI; instead, the population was gated according to side scatter (SSC) and forward scatter (FSC) channels. CFP fluorescence was recorded using the Alexa Fluor 430 channel (excitation, 405 nm; emission, 525/50 nm). Data extraction and analysis for the FACS experiments were done using FCS Express version 4 (De Novo Software). The data were exported to Microsoft Excel (2010) and further processed to obtain the data for fluorescence and relative density distributions.
The hysteresis experiments were performed by first growing the cells at an initial OD 600 of 0.05 for 3 h in 5 ml of MG medium supplemented with 0.2% of yeast extract in either the presence or absence of 10 ng/ml anhydrotetracycline (aTc). The cells were then harvested by centrifuging at 3,200 ϫ g for 10 min, washed once with PBS, and finally resuspended in 50 ml MG medium supplemented with 0.2% yeast extract. Both cultures were then supplemented with various concentrations (0.01, 0.1, 0.2, 0.5, 0.8, 1, 5, and 10 ng/ml) of aTc and grown for 4 h before harvesting. The samples were stained with DAPI and analyzed with a flow cytometer as described above.
Cell tracking. Cells were grown overnight at 37°C in MG medium supplemented with 0.2% final concentration of yeast extract. Cells were subcultured to an OD 600 of 0.05 in fresh MG medium supplemented with 0%, 0.2%, 0.5%, 1%, and 2% yeast extract. After subculture, the cells were then allowed to grow at 37°C for 5 h before harvesting. Glass slides and coverslips were soaked in 1 M KOH for 15 min and washed with deionized water prior to use. A 5-l volume of appropriately diluted sample, such that there would be roughly 50 cells in the view frame, was put on a glass slide, covered, and sealed with epoxy. Cells were tracked by phase-contrast using a Zeiss standard RA microscope equipped with a Hyper HAD blackand-white video camera. The movie was then analyzed using custom Mat-Lab software. The algorithm ignores all the cells that are stuck on the glass slides and analyzes only the cells that are swimming (motile) or drifting in the liquid (sessile).