Each Salmonella enterica serovar Typhimurium cell produces a discrete number of complete flagella. Flagellar assembly responds to changes in growth rates through FlhD4C2 activity. FlhD4C2 activity is negatively regulated by the type 3 secretion chaperone FliT. FliT is known to interact with the flagellar filament cap protein FliD as well as components of the flagellar type 3 secretion apparatus. FliD is proposed to act as an anti-regulator, in a manner similar to FlgM inhibition of σ28 activity. We have found that efficient growth-dependent regulation of FlhD4C2 requires FliT regulation. In turn, FliD regulation of FliT modulates the response. We also show that, unlike other flagellar-specific regulatory circuits, deletion of fliT or fliD did not lead to an all-or-nothing response in FlhD4C2 activity. To investigate why, we characterized the biochemical interactions in the FliT : FliD : FlhD4C2 circuit. When FlhD4C2 was not bound to DNA, FliT disrupted the FlhD4C2 complex. Interestingly, when FlhD4C2 was bound to DNA it was insensitive to FliT regulation. This suggests that the FliT circuit regulates FlhD4C2 activity by preventing the formation of the FlhD4C2:DNA complex. Our data would suggest that this level of endogenous regulation of FlhD4C2 activity allows the flagellar system to efficiently respond to external signals.
Salmonella enterica serovar Typhimurium utilizes flagella, self-assembled nanomachines anchored in the cell envelope, to swim through liquids and swarm over surfaces. Approximately 60 flagellar genes, organized into 13 operons, are involved in flagellar assembly, motility and chemotaxis. The expression of the flagellar genes follows a strict hierarchy resulting in the subdivision of the flagellar operons into three classes based on the promoters that drive their transcription: Pclass1, Pclass2 and Pclass3 (Chilcott and Hughes, 2000). There is only one Pclass1 promoter that drives transcription of the genes flhD and flhC. FlhD and FlhC form a hetero-hexameric complex, FlhD4C2, which is the master regulator of the enteric flagellar systems (Wang et al., 2006). FlhD4C2 activates transcription itself through a direct interaction with the alpha subunit of RNA polymerase, allowing σ70 to activate transcription from FlhD4C2 specific promoters (Liu et al., 1995). Previous studies of the FlhD4C2 have shown that the DNA binding component is primarily directed through FlhC. FlhD has been proposed to stabilize the FlhD4C2:DNA complex (Claret and Hughes, 2000). There are eight FlhD4C2-dependent flagellar-specific Pclass2 promoters. One of these promoters drives the transcription of the fliA gene, encoding an alternative sigma factor, σ28, required for transcription from the eight Pclass3 promoters (Ohnishi et al., 1990). The gene products of flhD, flhC and fliA are essential for the expression of the flagellar genes.
Along with fliA, a further regulatory component transcribed from both Pclass2 and Pclass3 promoters is flgM. The gene product of flgM is an anti-sigma factor that directly interacts with σ28 in the cell, preventing σ28 from activating transcription from the Pclass3 promoters during HBB assembly (Gillen and Hughes, 1991a,b). Upon HBB completion Pclass3 activation is coupled to flagellar assembly through the physical secretion of FlgM via the flagellum, thus freeing σ28 to activate transcription from Pclass3 promoters (Hughes et al., 1993; Kutsukake and Iino, 1994).
The coupling of assembly and gene expression suggests that some gene products are only expressed at a given stage of the assembly pathway. However, one limitation of this model is that it only considers a single flagellum. S. enterica does not produce a single flagellum per cell. Rather, it produces many, approximately four to eight per cell (Iino, 1974). Thus, in a growing cell with more than one flagellum, all components required to assemble a complete flagellum are in theory present. Therefore, the flagellar system likely requires additional levels control in the presence of growing or assembled flagella that allow it to co-ordinate and adjust flagellar gene expression accordingly. Specifically, the presence of an upper limit in flagella number suggests that a negative feedback loop, sensitive to the number of flagella present, exists to modulate the activity of the regulators FlhD4C2 and σ28.
For σ28:FlgM, the physical secretion of FlgM itself through completed HBB structures provides a given level of control for σ28. We have recently shown that changing the rate of FlgM secretion modulates σ28 activity (Brown et al., 2008). Ideally, a level of negative feedback similar to the σ28:FlgM circuit would regulate the activity of the master regulator FlhD4C2 in response to flagellar assembly. As FlhD4C2 is at the top of the transcriptional hierarchy, any changes in its activity will have the net effect of reducing both Pclass2 activity and σ28-dependent activation of Pclass3. Other than FlgM and σ28, there are two regulatory proteins that modulate flagellar gene transcription: FliT and FliZ (Kutsukake et al., 1999; Yamamoto and Kutsukake, 2006; Saini et al., 2008). The necessary negative feedback required to regulate FlhD4C2 activity to secretion is the anti-FlhD4C2 factor, FliT (Yamamoto and Kutsukake, 2006). FliT is known to inhibit FlhD4C2-binding to DNA through its direct interaction with FlhC (Yamamoto and Kutsukake, 2006), while it also interacts with the filament-capping protein FliD and two soluble proteins, FliI and FliJ, of the flagellar-specific type 3 secretion (T3S) apparatus (Fraser et al., 1999; Evans et al., 2006; Imada et al., 2010). FliJ enhances FliT/FliD interactions with the C-terminal cytoplasmic domain of FlhA, a key protein of the flagellar-specific T3S apparatus (Bange et al., 2010; Saijo-Hamano et al., 2010). The interactions of FliT with FliD, FlhA and FliJ are thought to facilitate the secretion of FliD upon completion of the HBB. With respect to its interaction with FlhC, FliT has been defined as an anti-FlhD4C2 factor and FliD an anti-anti-FlhD4C2 factor (Yamamoto and Kutsukake, 2006). In this regulatory circuit, FlhD4C2 would regulate flagellar assembly in response to the rate of FliD secretion.
FliT consists of four α-helices (α1, α2, α3 and α4) (Imada et al., 2010). A truncation of the C-terminal α4-helix, leading to FliT94, increases the binding affinity for the FlhD4C2 complex, suggesting that α4 plays an autoregulatory role in modulating FliT activity. A surface exposed residue Lys-79 in α3 is critical for the interaction with FliD. The α4 helix binds to the hydrophobic cleft formed by α2 and α3. As a truncation of α4 does not affect the interaction with FliD or FlhD4C2per se, it is proposed that the interaction between α4 and the hydrophobic cleft modulates the binding affinity for the FlhD4C2 complex (Imada et al., 2010).
In this work, we investigate the role of FliT regulation of FlhD4C2 during continual flagellar assembly. We show that FliT regulation of FlhD4C2 activity allows the system to efficiently respond to bacterial growth rates. In addition, unlike the all-or-nothing response observed with respect to the deletion of fliA (σ28) or flgM, genetic analysis of the FliT circuit suggests FliT tunes the activity of the FlhD4C2, but never switches it off. To investigate the mechanism behind this behaviour, we performed a biochemical characterization of the FliT:FliD:FlhD4C2 circuit. Using this approach, we show that FliT regulates FlhD4C2-dependent activation of Pclass2 by disrupting the FlhD4C2 complex. However, FliT-dependent inhibition is unable to disrupt the FlhD4C2–DNA complex. Based on our results, we propose the model that FliT regulation of FlhD4C2 allows the flagellar system to respond efficiently to external regulatory inputs directed through changes in FlhD4C2 activity.
The assembly of flagella in S. enterica responds to changes in growth rates
Recently, a twofold change in flhDC gene expression was shown to increase flagellar numbers by twofold (Erhardt and Hughes, 2010). This study used, in part, a chromosomally encoded, functional FliM–GFP fusion to numerate flagella by the number of HBB per cell (Aldridge et al., 2006a). A significant number of studies over the past five decades monitor S. enterica flagellar assembly in Luria–Bertani (LB). We asked if the growth conditions were changed, would a similar flagella distribution or a growth-dependent phenotype be observed. Using the FliM–GFP construct we counted the number of HBB per cell of S. enterica grown in four conditions. We used a minimal medium base (Minimal E plus 0.2% glucose) containing increasing amounts of yeast extract (YE) (0.04 g l−1, 0.2 g l−1, 1 g l−1 and 3 g l−1). Yeast extract was chosen, as it could increase growth rates to that equivalent of LB (3 g l−1 YE, data not shown). Growth was performed at 30°C for consistency with previous experiments performed (Brown et al., 2008). The rationale of adding rich nutrients to a base medium was similar to seminal growth experiments performed by Schaechter et al., to observe changes in growth mass in S. enterica at different growth rates (Schaechter et al., 1958).
Flagellar numbers were counted at late exponential phase of standard shaken cultures prior to the transition into stationary phase (Fig. 1A, Fig. S1 and Table 1). This time point is suggested to coincide with maximal flagellar gene expression (Pruss and Matsumura, 1996; 1997). We observed an unexpected growth-dependent phenotype that went against our perception of flagella utilization. In our nutrient-limiting conditions (0.04 g l−1 YE) we observed a significant reduction in flagellar numbers compared with 3 g l−1 YE. Changing carbon source or using a defined amino acid mix did not alter this response (data not shown). On average, growth in 0.04 g l−1 YE produced 0.3 flagella per cell, while 3 g l−1 YE produce 3.74 flagella per cell (Table 1). No cells were observed for 0.04 g l−1 YE with more than two foci (Fig. 1A). In contrast to growth in 0.04 g l−1 that exhibited an exponential distribution of flagellar numbers, 3 g l−1 YE growth produced a Gaussian (or normal) distribution of flagellar numbers. For the 3 g l−1 YE population, 90% of cells observed possessed one or more foci (Table 1). This suggests that only when nutrients are perceived will flagellar numbers increase, dependent on nutrient availability, to achieve optimal swimming behaviour and therefore chemotaxis. These observations are consistent with previous reports on the influence of media composition on the flagellation of E. coli (Adler and Templeton, 1967).
Table 1. Statistics of the FliM–GFP foci distribution in wild-type. ΔfliT and ΔfliD.
Calculated from three independent growth cultures.
Standard deviations can be found in Table S1.
0.30 ± 0.58
1.28 ± 1.59
0.22 ± 0.55
0.50 ± 0.81
2.88 ± 3.88
0.30 ± 0.54
1.91 ± 2.26
5.46 ± 2.72
0.22 ± 0.51
3.74 ± 2.41
5.32 ± 3.27
1.40 ± 1.74
FliT regulation is required to allow the system to respond accurately to external signals
The response of the S. enterica flagellar system to growth conditions suggests the system is flexible enough to respond to external cues, even though motility is defined as a robust phenotype (Barkai and Leibler, 1997). The majority of external signals are transduced into the flagellar system through changes in flhDC expression or FlhD4C2 stability, observable as a change in FlhD4C2 activity (Soutourina and Bertin, 2003). Yamamoto and Kutsukake (2006) have shown that FlhD4C2 activity is negatively regulated via a direct interaction with FliT. This suggests that FlhD4C2 activity is regulated by internal signals from the flagellar system. In their work, it was proposed that FliD regulates the ability of FliT to inhibit FlhD4C2 activity (Yamamoto and Kutsukake, 2006). Therefore, FliT regulation of FlhD4C2 activity sensitizes the system to flagellar assembly via FliD secretion rates (Brown et al., 2008). As a growth-dependent phenotype for flagella numbers was observed, we asked which level of regulation was dominant, the external regulation of FlhD4C2 or the endogenous regulation through FliT. We therefore compared FliM–GFP foci numbers in our growth conditions for ΔfliT and ΔfliD mutants to wild-type (Fig. 1B and C).
We observed a significant increase in flagellar numbers at all growth rates for the ΔfliT mutant compared with wild-type (Fig. 1B and Fig. S1). The percentage of cells with foci also significantly increased for the ΔfliT mutant compared with wild-type. Grown in 0.04 g l−1 YE 60% of cells from the ΔfliT mutant population contained FliM–GFP foci as compared with 24% in wild-type. The average number of FliM–GFP foci in the ΔfliT mutant also increased in all conditions (Table 1). Consistently, a reciprocal drop in flagellar numbers was observed for the ΔfliD mutant (Fig. 1C, Fig. S1 and Table 1). Even for 3 g l−1 YE the 54% of foci containing cells of the ΔfliD mutant had only one or two foci (16/50 cells) with only a small proportion having 3 foci (4 cells), 4 foci (3 cells), 5 foci (2 cells) or 6 foci (2 cells) (Fig. 1C). Importantly, the ΔfliT and ΔfliD phenotypes were observed at all growth rates, suggesting that the level of regulation exerted through FliT on FlhD4C2 is independent of growth. Furthermore, the data suggest that for the system to respond to external cues efficiently, endogenous regulation of FlhD4C2 by FliT is required.
Overexpression of fliT does not halt flagellar gene expression completely
Consistent with the flagellar numbers of the ΔfliT and ΔfliD mutants, plasmid-based expression of fliT or fliD resulted in the expected change. An increase of full-length fliT concentration by approximately fourfold (data not shown) was unable to completely inhibit movement as assayed using motility agar plates (Fig. 2A). This level of increased fliT expression resulted in only 2% of cells possessing one FliM–GFP foci in comparison to 90% in wild-type. The recent description of the FliT structure suggested that a C-terminal α-helix modulates FliT activity (Imada et al., 2010). Deletion of this α-helix, producing the variant fliT94, resulted in a further reduction in motility on motility agar after 8 h incubation when fliT94 was expressed from a high copy number plasmid (pSE-fliT94) or as a GST fusion protein (pGST-fliT94 (pMMT002) –Fig. 2A) (Imada et al., 2010). In comparison, overexpression of either flgM or fliA from pSE380 did not reduce motility in soft agar plates significantly (Fig. 2A). This suggests that the σ28:FlgM and FliT regulatory circuits play distinct roles in regulating flagellar assembly, albeit that they are interlinked by σ28-dependent transcription of the fliDST operon.
We have recently established a flagellar gene expression assay that monitors flagellar gene expression through a Fla- to Fla+ transition (Brown et al., 2008). The motility phenotypes of plasmid-based expression of fliT and fliT94 suggest that a basal level of flagellar gene expression is still present. To confirm this, flagellar gene expression dynamics were measured for plasmid-based fliT or fliT94 expression from pSE380 (Fig. 2B). As compared with wild-type, plasmid-based expression of fliT and fliT94 resulted in a significant delay in activation of Pclass2 and Pclass3 promoters. However, the flagellar transcriptional hierarchy was still evident in all strains (Fig. 2B). When expressing fliT, PflgA activity peaked at 1672 RLU at 91 min, a 4.9-fold reduction compared with its vector control that peaked at 8219 at 119 min. For fliT94 PflgA activity reached a maximum activity of 1067 RLU at 98 min, a 7.7-fold reduction in activity. For the Pclass3 promoter PfliC, fliT and fliT94 expression resulted in maximum activity of 130 and 118 RLU, respectively, compared with 15 339 RLU for the vector control. These observations are consistent with the proposed function of α4 modulating FliT activity by binding into a cleft formed by helices α2 and α3. Importantly, both constructs were unable to completely inhibit flagellar gene expression. Instead, flagellar gene expression is significantly reduced to a basal level that can still sustain some motility. Therefore, this suggests that for the observed phenotypes, some aspect of the FliT:FlhD4C2 regulatory circuit prevents FliT completely inhibiting FlhD4C2 from activating flagellar gene expression.
FliT disrupts the free FlhD4C2 complex
The in vivo data on FliT regulation suggest that even in excess, FliT is unable to completely inhibit FlhD4C2. Instead, the data suggest that FliT acts to significantly reduce FlhD4C2 activity. However, the in vivo data does not offer an explanation to how this level of regulation by FliT is achieved or why it exists. We therefore investigated the biochemical characteristics of the FliT regulatory circuit in vitro to determine if the nature of the interactions between FliT, FliD, FlhD4C2 and the FlhD4C2 DNA binding site could aid our understanding of FliT regulation in vivo.
Yamamoto and Kutsukake (2006) were able to show that FliT interacts with FlhC of the FlhD4C2 complex. All in vitro analysis in this work focussed on FliT regulation of the FlhD4C2 complex rather than FlhC itself. The rationale for this approach argued that to understand the in vivo data, how the FlhD4C2 complex behaves in the presence of FliT rather than FlhC alone was important. We began by asking whether we could isolate an FlhD4C2:FliT complex. FlhD4C2 and FliT were purified separately, mixed at different molar ratios and subjected to gel filtration. No high-molecular-weight species characteristic of a stable FliT:FlhD4C2 complex was observed (data not shown). The analyses of the FliT:FlhD4C2 interaction using surface plasma resonance (SPR) and isothermal calorimetry were unable to add any value to understanding the gel filtration data. We therefore utilized analytical ultra-centrifugation (AUC) to characterize the FliT:FlhD4C2 interaction further (Fig. 3).
FliT exists in equilibrium between its monomeric and dimeric form in solution, as judged by sedimentation equilibrium AUC measurements (Imada et al., 2010). Analysis of FlhD4C2, by sedimentation velocity (SV) AUC, produced a single protein species with a mass of 94.9 kDa. This is in good agreement with the theoretical mass of the FlhD4C2 complex (96.7 kDa) (Wang et al., 2006). Addition of FliT to FlhD4C2 at a 1:1 molar ratio using 8 µM FlhD4C2 did not produce a high-molecular-weight species larger than 94.9 kDa. Instead, the FlhD4C2 peak was reduced in size and a long shoulder that tapered towards the left of the FlhD4C2 peak appeared. Higher ratios of FliT:FlhD4C2 were then analysed and resulted in the shoulder becoming a defined peak with a predicted mass of 61–63 kDa. According to the size of this new protein species we assume that it could either be the dissociated FlhD4 tetramer or a FliT:FlhC2 complex. These data suggest that on interaction of FliT with FlhD4C2, the net result is the disassociation of the FlhD4C2 complex through the interaction of FliT with FlhC. This is in agreement with a previous report that the amount of FlhD co-eluted with GST-FliT94 from a GST column is much lower than that of FlhC (Imada et al., 2010).
Inhibition of FlhD4C2:DNA complex formation requires excess FliT
Within the FliT regulatory circuit, the key interaction is FlhD4C2-binding to its DNA binding site. Based on our AUC results, we speculated that excess amounts of FliT would provide increased inhibition of FlhD4C2:DNA binding. To test our hypothesis, we chose an SPR approach. We immobilized 350–400 resonance units (RUs) of double stranded 80-mers, biotinylated on one strand, to a streptavidin (SA) sensor chip containing the FlhD4C2 DNA binding site from flgB (PflgB) and flhB (PflhB) and an intergenic region of fliC coding sequence as a control DNA. Consecutive injections of FlhD4C2 in a dilution series to PflgB DNA exhibited a concentration-dependent response (Fig. 4A). Addition of FliT in a 1:1 ratio with FlhD4C2 prior to injection resulted in a significant reduction in FlhD4C2:DNA interaction but did not completely inhibit it (Fig. 4A). Analysis of the data showed that Vmax for FlhD4C2 alone to be 3933 compared with 560 for FlhD4C2 and FliT (Table S2). Binding to PflhB and fliC DNA (data not shown) showed consistency with previous studies on the FlhD4C2:DNA interaction. Stafford et al. showed that FlhD4C2 could bind to DNA without a defined FlhD4C2 binding site (Stafford et al., 2005; Wang et al., 2006). Therefore, the observation that FlhD4C2 bound to a coding region of fliC is not that surprising.
Our AUC data suggested that increasing molar ratios of FliT:FlhD4C2 would significantly disrupt the FlhD4C2 complex. Using our SPR assay the ability of increasing FlhD4C2:FliT ratios to inhibit FlhD4C2 to bind its DNA binding site was tested. Increasing amounts of FliT were added to 100 nM FlhD4C2 giving the ratios of 1:1, 1:1.5, 1:2.5 and 1:4 prior to injection over PflgB (Fig. 4B). Increasing amounts of FliT significantly reduced the ability of FlhD4C2 to bind to its binding site (Table S2). Molar ratios higher than 1:4 gave similar results (data not shown), suggesting that the maximal level of inhibition in our assays of the FlhD4C2 to DNA interaction is a ratio of approximately 1:4.
A FlhD4C2:DNA complex is resistant to FliT regulation
The above observations from our AUC and initial SPR analysis are sufficient to explain the observed phenotypes of the ΔfliT and ΔfliD mutants and fliT overexpression with respect to FliM–GFP foci. Our data suggest that to completely inhibit FlhD4C2 a high ratio of FliT to FlhD4C2 is required. Previous studies characterizing the FlhD4C2:DNA interaction have established that the FlhD4C2:DNA complex is very stable with a half-life of approximately 40 min (Claret and Hughes, 2000). Thus, irrespective of the level of free FliT, its regulation of FlhD4C2 will potentially encounter FlhD4C2 bound to DNA. Therefore, we asked what effect would FliT have upon an FlhD4C2:DNA complex using our SPR assay (Fig. 5). The injection of PBS running buffer did not show any effect, as expected. Unexpectantly, FliT would not disrupt the already existing FlhD4C2:DNA complex (Fig. 5A– addition of a 100 nM solution of FliT to bound FlhD4C2 complexes is shown). In these assays a constant concentration of 100 nM FlhD4C2 was first bound to the immobilized DNA. Even a 100-fold excess of FliT (10 µM) could not disrupt the FlhD4C2:DNA complex (data not shown). In contrast and as a control to show that FlhD4C2 was bound and could be disassociated, injecting 100 nM free PflgB DNA resulted in disassociation of the FlhD4C2:DNA complex from all three DNA species tested (Fig. 5B). Importantly, this control also showed that the interaction dynamics of FlhD4C2 with the three DNA species behaved in agreement with a previous work (Stafford et al., 2005). These results therefore suggest that FliT can prevent FlhD4C2 from binding DNA but if FlhD4C2 is already bound to DNA, FlhD4C2 is resistant to FliT regulation.
FliD acts as an anti-anti-FlhD4C2 factor
The phenotype of a ΔfliD mutant with respect to changes in flagellar gene expression and that FliD directly interacts with FliT led Yamamoto and Kutsukake (2006) to propose that FliD acts as an anti-anti-FlhD4C2 factor. However, this hypothesis has not been experimentally explored. We therefore used the SPR-based assay of FlhD4C2 binding DNA to observe what effect FliD would have upon FliT inhibition of FlhD4C2.
As FliD:FliT interact in a 1:1 molar ratio, a 1:1:1 ratio of FliD:FliT:FlhD4C2 was assayed. Addition of FliD to FlhD4C2, in the absence of FliT, did not alter the ability of FlhD4C2 to bind DNA (Fig. 6). This was in agreement with gel filtration and bacterial two-hybrid analysis that did not identify a FliD:FlhD4C2 interaction (data not shown). In contrast, a mix of FliD, FliT and FlhD4C2 did alter the ability of FliT to inhibit FlhD4C2 from binding DNA (Fig. 6). A similar response to the addition of FliD to a 1:1 mix of FliT94:FlhD4C2 was also observed (Fig. S2). Therefore, we can confirm Yamamoto and Kutsukake's proposal that FliD does act in an anti-anti-FlhD4C2 manner. Interestingly, a complete inhibition of FliT regulation was not observed.
The FliD/FlhD4C2 binding sites on FliT overlap
The data above suggest that FliD interaction with FliT when all proteins are at a 1:1:1 ratio is not sufficient to completely inhibit the regulation of FlhD4C2 by FliT. One explanation for this is that the binding sites for FliD and FlhD4C2 are on distinct surfaces of FliT. As a result even in the presence of FliD, FliT could disrupt the FlhD4C2 complex. An alternative explanation is that the sites of interaction overlap and the observed response is due to FliD/FlhD4C2 competition for FliT. To distinguish between these two models, a genetic analysis of fliT was performed. Our intent was to determine if fliT point mutations would exist that could bind FliD but not interact with FlhD4C2 and vice versa. The Pclass1flhDC promoter is negatively autoregulated by FlhD4C2 (Fig. 7). Deletion of fliT although observed to have increased Pclass2 activity exhibits reduced Pclass1 activity (Fig. 7). This suggests that FliT regulation effects not only Pclass2 activity but also FlhD4C2 levels through regulation of the autoregulatory control of the Pclass1 promoter region by FlhD4C2. Our screen used the drop in Pclass1 activity associated with loss of FliT regulation to isolate fliT mutations exhibiting reduced activity against FlhD4C2. Mutations in fliT were isolated using random PCR mutagenesis (Aldridge et al., 2006b). From 2000 colonies, 17 colonies with reduced Pclass1 activity were isolated randomly and sequenced. From these nine had single point mutations, three had two point mutations, two had triple mutations, one had a C-terminal extension of 25 amino acids due to the mutation of the stop codon and one did not contain any mutation in the coding sequence of fliT (data not shown). Using the recently described structure of FliT the single point mutations were found to all clusters in α1 or α3. Interestingly, one mutation N74D was at the same residue as had been identified to show weaker FliD interaction (Imada et al., 2010). This suggests that the FliD and FlhD4C2 binding sites on FliT may in fact overlap.
To further investigate whether the FliD/FlhD4C2 binding sites on FliT overlap, we assayed alanine substitution mutants in the C-terminal portion of α3 in motility assays, an FlhC pull-down assay (Imada et al., 2010) and the SPR-based assay. Motility assays of these mutations introduced into fliT94 showed that five exhibited a strong reduction in the ability of fliT94 to inhibit motility: I68A, N74A, E75A, K79A and L82A (Fig. 8A). Consistently, in a pull-down assay one of these mutations (E75A) showed a significant drop in its ability to interact with FlhC compared with those that showed a similar phenotype to wild-type fliT94 (Fig. 8B).
Four substitutions showing motility were also assayed for their ability to inhibit the FlhD4C2:DNA interaction. In agreement with the motility and pull-down assays FliT94(E75A) exhibited the weakest inhibition of the DNA interaction (Fig. 8C and Table S2). In contrast, the substitutions K79A and L82A behaved in a similar fashion showing inhibition at ratios greater than 1:2 of FlhD4C2:FliT94, while N74A exhibited a slight drop in inhibition at all concentrations. Importantly, all alanine substitutions still retained the ability to bind to FlhC to some extent and therefore inhibited wild-type motility (Fig. 8). Therefore, these mutants strongly suggest that the FliD/FlhD4C2 interface on FliT somehow overlap and the observed response in the SPR analysis is the result of competition for FliT.
A key factor often overlooked with respect to flagellar assembly is how the cell co-ordinates the synthesis of multiple flagella during bacterial growth. Here we have presented evidence suggesting that enteric bacteria are able to sense the extent of their flagellation via an endogenous regulatory circuit that modulates the activity of FlhD4C2. Furthermore, this internal regulation is key to allow the system to respond accurately to extracellular environmental, internal metabolic or growth-dependent signals.
Two key observations from this study showed that (i) loss of FlhD4C2 regulation by FliT disrupted an efficient response to changes in growth conditions and (ii) if bound to DNA, FlhD4C2 was insensitive to FliT regulation. Consistent with both observations was the finding that deleting fliD or expressing full-length or a truncated fliT variant from a plasmid was unable to completely inhibit FlhD4C2 activity in any growth condition. That when bound to DNA FlhD4C2 was insensitive to FliT regulation is an intriguing twist to an already complex regulatory network that couples flagellar gene expression to the assembly pathway.
Hughes and co-workers predicted that FlhC was the major DNA binding component of the FlhD4C2 complex (Claret and Hughes, 2000). Yamamoto and Kutsukake (2006) when describing the FliT:FlhD4C2 interaction were able to show that FliT was interacting with FlhC. Our data are in agreement with these observations. The resistance of the FlhD4C2:DNA complex to FliT suggests that the FliT binding site on FlhC overlaps the FlhC DNA binding domain. This is different from the model presented by Wang et al. (2006) when FlhD4C2 was crystallized. Wang et al. (2006) proposed a configuration that shows FlhD being the major DNA binding component. However, when discussing this model, they conceded that the data available were insufficient to rule out other plausibility's, such as FlhC making the majority of the DNA:protein contacts (Wang et al., 2006). Our data, with that of the co-workers of Hughes and Kutsukake suggests that the proposed alternative is more feasible but still requires direct biochemical or structural evidence to confirm the correct orientation.
Our genetic analysis of the FliT:FlhD4C2 interaction suggests that it is not only the interaction sites of FliT and DNA on FlhC that overlap. Mutations in FliT itself, previously shown to reduce FliD interaction, were also found to have reduced activity in inhibiting FlhD4C2 (Imada et al., 2010). A conclusion based on the analysis of alanine substitutions in the truncated variant FliT94. The genetic analysis, in conjunction with our biochemical analysis, suggests that a key factor in the way that FliT regulates FlhD4C2 is the extent of competition between all components. As a result this circuit is sensitive to the availability of FliT to bind FliD, FliT to bind FlhC and the availability of FlhD4C2 to bind DNA. A swing in the balance between these three reactions will alter the level of FlhD4C2 output in terms of transcription activation.
In agreement with previous data we have shown that FlhD4C2 can bind DNA un-specifically (Stafford et al., 2005). The mode of binding may be different, as seen during the competition assays using unbiotinylated PflgB DNA to disassociate the FlhD4C2:DNA complexes (Fig. 5B). The ability of FlhD4C2 to bind DNA even without a defined binding site could potentially have an impact of FliT regulation as it means FlhD4C2 does not only have to bind to its specific sites to become resistant to FliT.
What is the advantage of FlhD4C2 being regulated by FliT only when not bound to DNA? A recent study using SPR, which our assays were based on, showed that the RelBE system behaves somewhat different from FlhD4C2 regulation by FliT. The RelBE system is a toxin–antitoxin locus in which the toxin component can cleave mRNA associated with ribosomes but also autoregulates the transcription of the relBE locus (Overgaard et al., 2008). If RelE interacts with a RelB:DNA complex the result is disassociation of the proteins from the DNA (Overgaard et al., 2008). The flagellar and RelBE systems are very different with respect to their regulation. For the flagellar system there is an advantage for a given basal level of activity to exist even when flhDC expression is under significant negative influence from the long list of regulators that can influence its expression (Soutourina and Bertin, 2003). Both a Pclass2 and a Pclass3 promoter drive the transcription of fliT. Thus, even at low levels of flhDC expression some FliT can be made, even if no complete HBBs are present. The apparent competition between all components that our data suggest predicts that even during HBB assembly, when FliD is not being secreted, a given level of FliT would be available to interact with free FlhD4C2. Note also that activation of FlgM secretion allows σ28 to increase transcription of fliT. Therefore, once one structure is complete the level of free FliT will increase due to changes in FliD and FlgM levels as a result of secretion. We propose the working model that under growth conditions that exhibit significant inhibition of flhDC expression, FliT regulation will control the extent of flagellation by further inhibiting any free FlhD4C2 found in the cell. However, if the FlhD4C2:DNA interaction occurs first then bound FlhD4C2 will be insensitive to FliT regulation.
Our data suggest that under certain environmental conditions a basal level of FlhD4C2 activity is maintained via a balance between external regulation of flhDC expression and internal regulation through, for example, FliT inhibition of FlhD4C2. The result is that a small proportion of cells among the population will possess a low number of flagella. It is then the subpopulation of flagellated cells that are able to begin to move out of the current location to seek more preferential growth conditions. Adler and Templeton (1967) drew similar conclusions from their work to gain maximum motility of E. coli in a defined chemical medium. Interestingly, how efficient chemotaxis would be in E. coli or S. enterica cells possessing low flagella numbers has not been explored in detail and in the light of this work would be an interesting line of investigation (Turner et al., 2000; Darnton et al., 2007).
Thus, in conclusion we have shown that the enteric flagellar system requires a given level of endogenous negative feedback regulation to respond to external signals. For enteric bacteria this regulation is via the negative regulation of FlhD4C2 activity by FliT. We further conclude that FliT is unable to completely inhibit FlhD4C2 from activating flagellar gene expression, thus preparing the population to respond, with respect to the flagellar system, accordingly to external signals. What is now of interest is to begin to question how non-FlhD4C2 dependent systems control the same response.
Bacterial strains, plasmids and cultivation
Strains and plasmids used during this study are described in Table 2. Media used in this study included LB broth and Minimal E (Vogel and Bonner, 1956; Davis et al., 1980). All plates contained 1.5% agar. Motility assays were performed using semi-solid motility agar containing 0.3% difco agar. Cultures were grown at either 37°C or 30°C with constant shaking at 180 r.p.m. Antibiotic concentrations have been described previously (Bonifield et al., 2000). For the growth condition experiment Minimal E base salts were supplemented with 0.2% glucose and the stated concentration of yeast extract from a stock solution of 25 g l−1. For purification of FliT, FliD and FlhD4C2 LB cultures were grown to an OD600 between 0.4 and 0.6 at 37°C. Induction of expression for 3–5 h was achieved using a final concentration of 1 mM IPTG with incubation at 30°C. Cell pellets were harvested and stored at −80°C until needed.
Throughout this study standard cloning and DNA manipulation techniques have been used. All plasmids were sequenced prior to use. Primer sequences used for constructs are available on request. A modified version of pET28a-mod was used in this study that reduced the size of the his-tag fused to FliT or FliD with a region of the multiple cloning site deleted (Table 2). For the purification of FlhD4C2 the entire flhDC operon was subcloned into pET28a-mod. On induction this produced a complex that was his-FlhD4C2. All protein identities were also confirmed by MALDI-TOF analysis using an in-house facility at Newcastle (PINNACLE).
Random mutagenesis of fliT was performed using the natural error rate of Taq DNA polymerase. A PCR reaction generated using Taq DNA polymerase was recombined onto the chromosome of strain TPA993 selecting on Tetracycline-sensitive (TetS) plates (Aldridge et al., 2006b). After selection, crossing once more on to TetS plates was used to isolate 2000 colonies. A pool of the 2000 colonies was then transduced with plasmid pRG38 (PflhDC::luxABCDE) to screen for potential fliT mutants. To isolate colonies exhibiting reduced light production, images of the transduction plates were made using the chemiluminescent settings of an intelligent light box. Seventeen colonies were taken at random and purified to isolate phage-free colonies before light production was confirmed using liquid cultures assayed in a BMG Fluostar microplate reader. The fliT-coding region was then sequenced from PCR products. Deletion of fliD in TPA2421 used previously described recombineering protocols (Datsenko and Wanner, 2000; Aldridge et al., 2003).
To visualize FliM–GFP, cultures were grown as described above. At appropriate time points or OD600 values, one microlitre of culture was spotted onto a 1% agarose pad containing 1× E-salts but no glucose in a Gene frame (AB gene). The agarose pad was sealed with a coverslip, a procedure sufficient to immobilize the motile bacteria. Images were acquired on a Zeiss Axiovert 200M system (Murray and Errington, 2008). An exposure time for all GFP images of 1.5 s was used and 100 ms for a phase-contrast image. Images were captured through METAMORPH (version V.6.2r6) and then processed using ImageJ (NIH).
To count FliM–GFP foci all cells were counted using the dot function of ImageJ. From the counted cells a random number generator ( http://www.randomizer.org) was used to identify 50 cells from each time point and strain. Using a stacked RGB composite image of the GFP and phase-contrast images, the FliM–GFP foci were then counted manually in the identified cells. Histograms of the number distributions and statistical analysis were performed using Excel (Microsoft).
All protein isolations were done using an Akta Prime Plus (GE Healthcare). Purification of FliT and FliD was performed using standard purification protocols. Briefly, cell pellets of induced cultures were resuspended in His-loading buffer: 50 mM HEPES, 150 mM NaCl, 20 mM imidazole, pH 7.5 and sonicated for 5 min on ice. To remove all cell debris the sample was centrifuged at 31 500 g for 30 min at 4°C prior to injection over a pre-equilibrated 5 ml his-trap column (GE Healthcare). Proteins were eluted using an 80% imidazole gradient over 50 ml of His–elute buffer: 50 mM HEPES, 150 mM NaCl, 1 M imidazole, pH 7.5.
FlhD4C2 purification was performed using a modified method described by Wang et al. (2006). Plasmid pPA158 was freshly transformed into BL21 and induced cell pellets isolated from up to 6 l of culture and resuspended in Buffer A: 20 mM Tris, pH 7.9. The cell suspension was sonicated for 5 min, centrifuged for 30 min at 31 500 g at 4°C and injected over a 5 ml Heparin column (GE Healthcare). After injection, the column was washed with a 20% solution of Buffer B: 20 mM Tris, 1 M NaCl, pH 7.9. After the baseline reached a pre-loading level, elution of the FlhD4C2 complex was achieved using a gradient of Buffer B up to 100% applied over 60 min. All protein fractions were checked by SDS page on 16.5% Tricine gels.
After purification FliT, FliD and FlhD4C2 proteins were concentrated using VIVAspin columns (10 kDa cut-off) and subjected to gel filtration using a S200 column (GE Healthcare). The gel filtration buffer used was 20 mM Tris, 300 mM NaCl, 1 mM DTT, pH 7.9. Pull-down assays by GST affinity chromatography have been previously described (Imada et al., 2010).
Surface plasma resonance
The majority of SPR experiments were performed using the BIAcore 2000. Experiments using FliT94 were performed on a Bio-Rad Proteon XPR36. DNA–protein interaction experiments were done using an SA or neutravidin coated sensor chips (GE Healthcare and Bio-Rad). The running buffer for all experiments was phosphate-buffered saline (PBS). For the DNA–protein experiments using the BIAcore 2000, 365 RUs of DNA were immobilized on the SA sensor chip, while for the XPR36 100 RUs were immobilized. Prior to immobilization, three consecutive injections over all flow cells for 60 s at a flow rate of 10 µl min−1 with 1 M NaCl then 0.05 M NaOH were performed to ensure no contamination of unspecific bound compounds. An empty flowcell was always used as a negative control for background subtraction and therefore had no DNA or protein bound to it. The DNA used was an 80 bp double-stranded oligonucleotide, with a 5′ biotinylated forward strand of the flgB and flhB binding sites of FlhD4C2 (Table 3) (Stafford et al., 2005). Also as a negative control we used an intergenic region of fliC (Table 3). Annealing of the DNA was performed using a Biometra T3000 Thermocycler: 94°C for 2 min then ramp cooled to 25°C over 45 min then kept at 4°C until use. Annealing of samples was confirmed by native gel electrophoresis on 10% native gels. For all experiments FlhD4C2 protein was applied for 60 s over all four flow cells. Regeneration of the flow cells was done using a 10 s injection of 6 M guanidine-HCL. As a control for chip quality FlhD4C2 alone was always run at the beginning, middle and end of an experimental set. For protein–protein interactions standard BIAcore protocols were followed using 300 RUs of FliT immobilized on a CM5 sensor chip as ligand and either FliD or FlhD4C2 as analyte.
The SV experiments were carried out in a ProteomeLab XL-I analytical ultracentrifuge (Beckman Coulter, USA) using both absorbance (at 280 nm) and interference optics. All AUC runs were carried out at a rotation speed of 48 000 r.p.m. and an experimental temperature of 4°C using an 8-hole rotor and double-sector aluminium-epon centrepieces. The sample volume was 400 µl for the SV experiments with sample concentrations between 0.2 and 1.2 mg ml−1. The partial specific volumes () for the proteins were calculated from the protein amino acid sequence, using the program sednterp and extrapolated to the experimental temperature. The density and viscosity of the buffer (10 mM Tris pH 7.5, 1 mM EDTA, 200 mM NaCl) at the experimental temperature was also calculated using sednterp.
The distributions of sedimenting material were modelled as a distribution of Lamm equation solutions where the measured sedimentation boundary was modelled as an integral over differential concentration distributions c(s) as implemented in the program SEDFIT ( http://www.analyticalultracentrifugation.com) (Schuck, 1998). The SV profiles were fitted using a maximum entropy regularization parameter of P = 0.95. The weight average sedimentation coefficient was calculated by integrating the differential sedimentation coefficient distribution (Schuck, 2003). Sedimentation coefficients (s) were extrapolated to zero concentration and converted to standard conditions, i.e. those that would be measured at 20°C in water. The diffusion coefficient (D) corresponding to each sedimentation coefficient value was estimated from a weight-average frictional ratio (f/f0)w (Schuck, 1998; 2003). Conversion to a molecular mass distribution from s and D was made using the Svedberg equation
where R is gas constant, T is temperature, is a partial specific volume of solute and ρ is density of the buffer. The integration of the mass distribution c(M) was made similarly to the integration of c(s) distribution in order to determine the weight average molecular mass of each species.
We would like to thank all lab members and colleagues for critical reading of this manuscript. Special thanks go to W. Vollmer for his input. We would also like to thank R. Daniel, H. Murray and J. Errington for use of their microscope suite and M. Kinoshita-Minamino for technical assistance. Thanks also go to Bob Lightowlers for access and advice on the use of the BIAcore 2000. BBSRC Grant No. BB/D015855/1 awarded to P.A. funded the majority of this work. K.P. was awarded a PhD scholarship from Nareasuan University (Thailand). C.R. was partially supported by the National Science Foundation Grant CBET 0644744 and by Public Health Service Grants GM083601 and GM054365. Funding for K.I. and T.M. is through the Targeted Proteins Research Program from the Ministry of Education, Science and Culture of Japan (K.I. and T.M.). An International Project Grant awarded jointly by the Diawa Foundation and Royal Society (UK) funded T.M. and P.A.