Heterogeneous Flagellar Expression in Single Salmonella Cells Promotes Diversity in Antibiotic Tolerance

ABSTRACT Phenotypic heterogeneity among single cells in a genetically identical population leads to diverse environmental adaptation. The human and animal pathogen Salmonella enterica serovar Typhimurium exhibits heterogeneous expression of virulence genes, including flagellar and Salmonella pathogenicity island (SPI) genes. Little is known about how the differential expression of flagellar genes among single cells affects bacterial adaptation to stresses. Here, we have developed a triple-fluorescence reporter to simultaneously monitor the expression of flagellar and SPI-1 pathways. We show that the two pathways cross talk at the single-cell level. Intriguingly, cells expressing flagella (fliC-ON) exhibit decreased tolerance to antibiotics compared to fliC-OFF cells. Such variation depends on TolC-dependent efflux pumps. We further show that fliC-ON cells contain higher intracellular proton concentrations. This suggests that the assembly and rotation of flagella consume the proton motive force and decrease the efflux activity, resulting in antibiotic sensitivity. Such a trade-off between motility and efflux highlights a novel mechanism of antibiotic tolerance.

tolerance has been mostly associated with slow growth or low metabolism (10)(11)(12), although the underlying mechanisms remain elusive.
Salmonella enterica serovar Typhimurium is a bacterial pathogen in animals and humans and causes tens of millions of gastrointestinal infections in humans worldwide each year (13)(14)(15). In addition to gastroenteritis, S. Typhimurium and other nontyphoidal Salmonella strains also cause invasive human disease that often leads to death (14). Among the major virulence genes of S. Typhimurium are Salmonella pathogenicity island (SPI) and flagellar genes (16,17). SPI-1 and SPI-2 genes are important for invasion of host cells and intracellular survival, respectively (18). Flagella are critical for Salmonella to move toward nutrients as well as within the host (19). In addition, the flagellin protein FliC stimulates the host immune response by activating caspase-1 in macrophages (15). The assembly and rotation of flagella both require proton motive force (PMF) and are energetically costly (19,20). It has been shown that shutting off flagellar expression helps Salmonella cells evade the host immune response (21) and improve growth in culture (22). Interestingly, flagellar and SPI-1 expression is heterogeneous in Salmonella (23)(24)(25)(26)(27). The expressions of flagellar and SPI-1 pathways are interconnected at the population level (28). In this study, we constructed a triple-fluorescence reporter to detect SPI-1 and flagellar expression simultaneously and found that the expression of these two pathways was positively correlated among single S. Typhimurium cells. We next applied this reporter to study the functional impact resulting from the heterogeneous expression of flagellar and SPI-1 genes. To our surprise, we found that fliC-OFF cells displayed improved tolerance to bactericidal antibiotics compared with fliC-ON cells, in a manner independent of SPI-1 genes. We further show that the heterogeneous responses of fliC-ON and fliC-OFF cells to antibiotics depend on the efflux activity driven by PMF. Our results suggest that the expression and rotation of flagella compete with the efflux process for PMF. This trade-off leads to lower efflux activity and drug tolerance in cells expressing flagellar genes.

RESULTS
Cross talk between SPI-1 and flagellar pathways in single Salmonella cells. Our recent study of the role of protein synthesis in Salmonella virulence suggests that perturbing translational fidelity results in the downregulation of both SPI-1 and flagellar genes (29), leading us to further investigate the interplay between these two pathways. Previous studies have shown that the expressions of SPI-1 and flagellar genes in Salmonella are correlated at the population level (28). To monitor these two pathways concurrently in single cells, we constructed a low-copy-number plasmid carrying three fluorescence proteins. The mCherry gene (encoding a red fluorescent protein) is under the control of a constitutive Ptet promoter, serving as a reference for the overall gene expression activity in cells; the YFP gene (encoding a yellow fluorescent protein) is controlled by the promoter of prgH, which is a downstream SPI-1 gene; and the eCFP gene (encoding an enhanced cyan fluorescent protein) is fused to the promoter of fliC, a class 3 flagellar gene encoding the flagellin, as a reporter for the expression of the flagellar pathway. Using fluorescence microscopy, we observed the bimodal expression of both the prgH and fliC promoters in wild-type (WT) S. Typhimurium ATCC 14028s cells grown in high-salt Luria broth (LB Miller) (Fig. 1A). This was consistent with previous results from single reporters of SPI-1 and flagellar pathways, respectively (24,25). The master regulator of SPI-1 genes is HilD, which controls the expression of HilA (28,30,31). HilA in turn activates the expression of downstream SPI-1 genes, including prgH ( Fig. 2A). As expected, deleting hilD or hilA abolished the expression of the prgH promoter, whereas deleting the master flagellar regulator flhDC abolished the expression of PfliC ( Fig. 2B and C). These results confirm that our triple-fluorescence reporter accurately and sensitively measures the expression of both the SPI-1 and flagellar pathways in single cells.
We also found that deleting flhDC or fliZ decreased the fraction of prgH-ON cells (Fig. 2), supporting the cross talk between flagellar and SPI-1 pathways at the singlecell level.
Previous studies suggest that nonmotile Salmonella mutants are defective in host invasion (32). We confirmed that deleting flhDC indeed abolished the attachment of Salmonella to macrophage cells (Fig. 3). Interestingly, in the WT strain, prgH-ON/fliC-ON cells were substantially enriched in the subpopulation that attached to macrophages (Fig. 3), suggesting that both SPI-1 and flagellar pathways are required for effective host cell interactions.
Flagellar expression decreases the tolerance of single Salmonella cells to antibiotics. We next investigated the phenotypic variations of fliC-ON and fliC-OFF cells. We recorded the motility of single Salmonella cells using fluorescence and phase-contrast microscopy. While ;50% of fliC-ON cells were highly motile, only ,10% of fliC-OFF cells showed motility ( Fig. S2 and Video S1). In addition to FliC, some Salmonella cells also encode another flagellin, FljB, in phase variation (33). The small percentage of motile cells among the fliC-OFF cells could be due to FljB-dependent motility. Next, we tested the regrowth of early-stationary-phase cells and did not observe a significant difference between the fliC-ON and fliC-OFF subpopulations (Fig. S3). However, following a brief treatment (15 min) of early-stationaryphase Salmonella cells with the bactericidal antibiotic ciprofloxacin (Cipro) or streptomycin (Strep), a significantly higher percentage of fliC-OFF cells resumed multiple divisions than fliC-ON cells (Fig. 4, Fig. S4, and Video S2). The expression of flagellar genes is multilayered, and class 3 genes (such as fliC) are under the control of the sigma factor FliA. We further validated that similar to fliC-OFF cells, fliA-OFF cells were also more tolerant to antibiotic killing than fliA-ON cells (Fig. 5). Antibiotic Tolerance in Salmonella ® Early-stationary-phase prgH-OFF cells also appeared to survive Cipro treatment better than prgH-ON cells. Given the cross talk between flagellar and SPI-1 genes, we next tested which pathway was primarily responsible for the variation in antibiotic tolerance. We found that deleting flhDC abolished the difference in Cipro tolerance between prgH-ON and prgH-OFF cells, whereas fliC-OFF cells remained more tolerant than fliC-ON cells in the DhilA strain (Fig. 4C). Deleting flhDC also increased the fraction of Cipro-tolerant cells compared with the WT (Fig. S5A). Collectively, these results suggest that heterogeneous flagellar expression, instead of the SPI-1 pathway, promotes variation in transient antibiotic tolerance in Salmonella under the tested conditions.
Flagella compete with efflux for PMF to decrease antibiotic tolerance. Many Gram-negative bacteria, including Salmonella, exhibit robust efflux activities to remove toxic small molecules (e.g., antibiotics) from cells (34,35). TolC is a key component for many resistance-nodulation-cell division (RND) efflux pumps, which use PMF to move antibiotics (including Cipro and Strep) across the double membrane from the cytoplasm to the exterior (34)(35)(36)(37). We showed that deleting tolC or adding an efflux pump inhibitor, Phe-Arg-b-naphthylamide (PAbN) (38), abolished the difference between fliC-ON and fliC-OFF cells in Cipro tolerance (Fig. 6A). Using an efflux reporter dye, Nile red (39), we found that the DflhDC mutant exhibited significantly higher efflux activity than the WT (Fig. 6B). In contrast, deleting hilA or hilD only modestly affected efflux (Fig. S6). This could not be explained by the expression of efflux genes, as deleting flhDC appeared to decrease, rather than increase, the percentage of cells with high expression levels of tolC (tolC-HIGH) or acrAB (acrAB-HIGH) (Fig. 7). In the WT, the fliC-ON subpopulation also displayed higher percentages of tolC-HIGH and acrAB-HIGH cells (Fig. 7). As both efflux and flagellar motility are driven by PMF, we hypothesized that there is a trade-off between efflux and flagellar expression due to the competition for PMF. Indeed, adding the PMF uncoupler carbonyl cyanide m-chlorophenylhydrazone (CCCP) or carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) ablated Antibiotic Tolerance in Salmonella ® the difference observed between fliC-ON and fliC-OFF cells in Cipro tolerance (Fig. 8).
Deleting the motAB genes, which encode flagellar motor proteins that drive the rotation of flagella using PMF, also abolished the difference in antibiotic tolerance between fliC-ON and fliC-OFF cells (Fig. S5B). Furthermore, we used an intracellular pH indicator, BCECF-AM [29,79-bis-(2-carboxyethyl)-5 (and -6)-carboxyfluorescein acetoxymethyl ester], and found that in fliC-OFF cells, a larger percentage exhibited lower intracellular proton concentrations (higher BCECF signals and pH) than in fliC-ON cells (Fig. 8). The DflhDC mutant also had a larger percentage of cells with higher BCECF signals than the WT. As the extracellular pH remained the same for all cells, this implies that fliC-OFF cells have higher DpH (and likely higher PMF) than fliC-ON cells overall. Furthermore, the addition of glucose to the media abolished the difference in Cipro tolerance between fliC-ON and fliC-OFF cells (Fig. S7), presumably due to an increase in PMF. Together, our data support that the expression of flagella costs PMF and reduces efflux efficiency, thereby decreasing tolerance to antibiotics (Fig. 9).

DISCUSSION
Flagellar motility is a mechanism commonly used by bacteria to search for nutrients and hosts. However, the expression and assembly of flagella are also energetically costly processes (40,41). It is remarkable that some bacteria, such as Salmonella, use a bet-hedging mechanism to express flagellar genes heterogeneously within a population (23,24,42). Our work here reveals a previously unknown benefit of the differential expression of flagellar genes. We show that fliC-OFF Salmonella cells that do not express flagella are more tolerant to antibiotics ( Fig. 4; see also Fig. S4 in the supplemental material). Due to the cost of proton motive force to drive motility, flagellum-ON cells are less capable of removing intracellular antibiotics through efflux than flagellum-OFF cells (Fig. 9). It is also interesting to note that the expression of flagellar and SPI-1 genes appears to be positively correlated (Fig. 1). SPI-1 genes are essential for Salmonella to invade host cells (15), and nonmotile Salmonella mutants have been shown to exhibit a substantial decrease in host cell invasion in vitro and in vivo (32). Consistently, we show that DflhDC mutant cells are defective in attachment to macrophage cells (Fig. 3). At the single-cell level, macrophage-attached Salmonella cells are significantly enriched in prgH-ON and fliC-ON cells (Fig. 3). Coupling the expression of the SPI-1 secretion system with flagella would therefore enable prgH-ON/fliC-ON cells to quickly move toward and invade the host cells, whereas prgH-OFF/fliC-OFF cells remain in the extracellular environment, such as the intestinal lumen of mammalian hosts. The intestinal lumen is enriched in antimicrobial molecules, such as antimicrobials secreted by competing microbes as well as bile acids from the host (43,44). Shutting off both SPI-1 and flagellar pathways would enable these Salmonella cells to maximize their efflux activity and remove toxic molecules, allowing survival and growth.
The cross talk between SPI-1 and flagellar pathways is complex and involves multiple factors, including FlhDC, FliZ, RtsAB, and HilD. HilD, RtsA, and HilC form a positive Antibiotic Tolerance in Salmonella ® feedforward loop to control the expression of SPI-1 genes (45). A previous study shows that overexpressing HilD activates the transcription of flhDC (46). Our reporter assay reveals that deleting hilD does not substantially decrease the fraction of fliC-ON cells (Fig. 2), suggesting that the native expression level of HilD under our growth conditions may not be high enough to dominate the cross talk between the two pathways. We also show that deleting flhDC or fliZ negatively impacts both the SPI-1 and flagellar pathways (Fig. 2), which is consistent with previous studies (47,48).
The hierarchical expression of flagellar genes is regulated at multiple levels (49). The class 1 master regulator FlhDC controls the expression of class 2 genes such as fliA, and FliA controls the expression of class 3 genes. Salmonella encodes two class 3 flagellins, FliC and FljB, and undergoes phase variation (33). Our motility assay shows that the fraction of motile cells is 7-fold higher in the fliC-ON group than in the fliC-OFF group (Fig. S2), suggesting that FliC is the prevalent form of flagellin under our tested conditions. The small percentage of motile cells among the fliC-OFF cells could be due to FljB-dependent motility. To test the effect of phase variation on the heterogeneity  Fig. 3. Cells in the right panels were treated with the efflux inhibitor PAbN (20 mg/ml) together with Cipro. These results are representative (top) and averages (bottom) of data from three independent experiments. The P values were determined using the x 2 test. (B) S. Typhimurium cells grown to the early stationary phase were treated with the efflux reporter dye Nile red (48 mg/ml) and monitored for fluorescence using a plate reader. As expected, due to the efflux defect, the DtolC mutant accumulated more Nile red than the WT. The DflhDC mutant displayed a significantly lower Nile red signal than the WT, indicating higher efflux activity. Error bars represent 1 SD from the mean (n = 4). The P value was determined using the unpaired t test. ***, P , 0.001. of antibiotic tolerance, we constructed a reporter using YFP under the control of PfliA, which is upstream of phase variation. We show that like fliC-ON cells, fliA-ON cells are also more sensitive to antibiotic killing than fliA-OFF cells (Fig. 5), further supporting our model of a flagellum-efflux trade-off (Fig. 9).
Antibiotic tolerance and resistance pose a severe and urgent threat to human health (6)(7)(8). A significant cause of antibiotic failure is the efflux activity in many bacterial pathogens, especially the Gram-negative bacteria (34). Salmonella expresses multiple efflux pumps. One major family is the TolC-dependent RND efflux pumps, which use PMF to remove many clinically important antibiotics, such as Cipro and Strep used in this study (34)(35)(36)(37). Cipro and Strep are representatives of fluoroquinolone and aminoglycoside antibiotics that target DNA replication and protein synthesis, respectively, and Cipro is frequently used as a frontline antibiotic to treat Salmonella infections. We show that the differential tolerance to Cipro between fliC-ON and fliC-OFF cells depends on the presence of TolC and PMF (Fig. 5 to 7), revealing a novel mechanism of antibiotic tolerance resulting from the trade-off between flagellar motility and efflux. It is therefore advantageous for the Salmonella population to maintain the heterogeneous expression of flagellar genes among individual cells, which provides a bet-hedging mechanism for optimal adaptation to ever-changing environments. In addition to Salmonella, heterogeneous expression of flagella has been observed in some pathogenic Escherichia coli strains (50). Given that PMF is used to drive flagellar motility and efflux in many bacteria (35,41), it is tempting to speculate that the trade-off between the two also occurs in other bacterial species, which needs to be investigated in future studies. Previous studies have also suggested that the heterogeneous expression of Antibiotic Tolerance in Salmonella ® SPI-1 cells contributes to persistence in the Salmonella population (51,52). Our work here shows that the flagellar pathway, instead of SPI-1, contributes to the transient tolerance to antibiotics (Fig. 3), thus highlighting distinct mechanisms underlying antibiotic tolerance and persistence.  (54). Therefore, all three ionophores disrupt the PMF. These results are representative (top) and averages (bottom) of data from three independent experiments. The P values were determined using the x 2 test.

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. All Salmonella strains (Table 1) used in this study are derived from S. Typhimurium ATCC 14028s. Salmonella serovar Typhimurium gene deletion mutants were constructed essentially as previously described (53). Briefly, the Flp recombination target (FRT)-flanked chloramphenicol resistance gene (cat) was amplified by PCR from plasmid pKD3 using primers shown in Table S1 in the supplemental material. The resulting PCR products were purified and electroporated into S. Typhimurium ATCC 14028s cells harboring plasmid pKD46 expressing the Red recombinase. The recombinants were selected on Luria broth (LB) plates containing chloramphenicol at 37°C and verified by PCR. All strains used in this study were cultured in LB Lennox (containing 10 g/liter of tryptone, 5 g/liter of yeast extract, and 5 g/liter of NaCl) or LB Miller (containing 10 g/liter of tryptone, 5 g/liter of yeast extract, and 10 g/liter of NaCl). The following antibiotics were used: ciprofloxacin at 0.1 mg/ml, ampicillin at 100 mg/ml, chloramphenicol at 25 mg/ml, and streptomycin at 60 mg/ml.
For the construction of the PprgH-YFP, PfliA-YFP, and PfliC-eCFP fusions, the promoter regions of prgH, fliA, and fliC containing sequences of 500 bp upstream of the start codons were amplified from strain ATCC 14028s genomic DNA by PCR. The DNA fragments were fused to plasmid pZS Ptet-mCherry Ptet-YFP or pZS Ptet-eCFP using the In-Fusion HD cloning kit according to the manufacturer's instructions.
Time-lapse microscopy. Cultures grown overnight in LB Lennox were diluted 1:200 in LB Miller and grown aerobically for 5 h at 37°C. All cultures were normalized to an optical density at 600 nm (OD 600 ) of ;0.1, and ciprofloxacin was added to a final concentration of 0.1 mg/ml. Following 15 min of incubation at 37°C with agitation, cultures were harvested by centrifugation and resuspended in 100 ml LB. Two microliters of the resulting cultures was placed on a 1.5% agarose LB pad within a Gene Frame (Thermo Fisher Scientific). Fluorescence images were taken at the initial time point for quantitation. Cells were then imaged for 6 to 10 h at room temperature with a 60Â phase-contrast lens at 20-min intervals. Image analysis was performed using a BZ-X800 analyzer (Keyence).
Efflux activity assay. Cultures of bacteria grown overnight were diluted 1:200 into fresh LB Miller and incubated for 5 h at 37°C. All cultures were normalized to the same OD 600 , and aliquots (100 ml) were transferred to 96-well plates. Nile red was added to the cells at a final concentration of 48 mg/ml, and the cells were incubated with agitation for 3 h at 37°C. The fluorescence intensity was recorded in a Synergy H1 microplate reader (BioTek) using an excitation wavelength of 549 nm and an emission wavelength of 628 nm.
Flow cytometry analysis. Cultures grown overnight in LB Lennox were diluted 1:200 in LB Miller and grown aerobically for 5 h at 37°C. BCECF acetoxymethyl ester (BCECF-AM) was added to a final concentration of 40 mM, and the cultures were incubated for 30 min at 37°C. Cells were diluted in phosphate-buffered saline (PBS) and directly measured on a BD FACSCanto II flow cytometer at a low flow rate. In all, 10,000 to 30,000 gated events were acquired for each sample. The density plots show the distribution of promoter activities in individual cells as determined based on YFP and eCFP fluorescence. Data were further analyzed using FlowJo software.
Susceptibility to ciprofloxacin. The MIC of ciprofloxacin was determined in 96-well microtiter plates. Mid-log-phase cultures were added to LB medium containing 2-fold serial dilutions of ciprofloxacin with a final inoculum size of 10 5 CFU/ml. Plates were incubated at 37°C with agitation for 18 h. The MIC was defined as the lowest concentration that completely inhibited visible growth.
We declare no competing interests.