Introduction

The flagellum in the enteric bacteria, Escherichia coli and Salmonella enterica, has been studied extensively for over fifty years and provides the canonical example for bacterial motility. These studies have revealed not only the complex structure of the enteric flagellum but also its role in host colonization, pathogenesis, and cellular physiology1,2,3,4. In addition, these studies have identified many of the complex regulatory processes that coordinate the assembly and control of this exquisitely complex biological machine3,4,5.

The flagellum in E. coli and S. enterica are structurally very similar and are often tacitly assumed to be effectively identical aside from differences in the filament structure. However, in the case of regulation, these assumptions are based more on sequence similarity rather than on actual experimental data5,6. Indeed, a number of studies have shown that these two systems are regulated in entirely different manners in response to environmental signals despite strong gene synteny. For example, many common E. coli strains are motile only during growth in nutrient-poor conditions whereas many common S. enterica strains are motile only during growth in nutrient-rich conditions7. In addition, E. coli is more motile at 30 °C than at 37 °C whereas motility S. enterica is generally insensitive to these temperature differences8. E. coli flhDC are transcribed from a single transcriptional start site that is responsive to OmpR, RcsB and CRP regulation, to name only a few regulatory inputs8. In contrast S. enterica flhDC transcription is significantly more complex with up to 5 transcriptional start sites, albeit with only a subset being responsible for the majority of flhDC transcription9.

Part of the problem is that different questions have been asked when studying the regulation of motility in these two bacterial species. Most studies in E. coli have focused on the environmental signals and associated regulatory process that induce bacterial motility. In particular, they have focused on the processes that regulate the expression of the master flagellar regulator, FlhD4C28. Most studies in S. enterica, on the other hand, have focused on the regulatory processes that coordinate the assembly process following induction4. In particular, they have focused on the downstream regulatory processes induced by FlhD4C23.

Despite differences in regulation, the protein subunits of master flagellar regulators, FlhC and FlhD, exhibit high sequence similarity sharing 94 and 92% identity, respectively (Figure S1), between E. coli and S. enterica. For both proteins the most significant amino acid changes are within the last 8 amino acids. Other substitutions are scattered across each protein and do not provide a consistent mutational pattern that provide a clear phenotypic explanation. Given that modifications to transcription factors and/or promoter structure can lead to divergence in regulatory circuits10, we were interested in how FlhD4C2 functions in different genetic backgrounds. Previously, it was shown that E. coli flhDC can complement a ∆flhDC mutant in S. enterica, suggesting that these proteins are functionally identical in the two bacterial species11. However, it is not clear whether they are regulated in the same manner. We, therefore, investigated the impact of replacing the native master regulator in S. enterica with the one from E. coli. Defining the impact of known FlhD4C2 regulators such as ClpP, RflP (previously known as YdiV), FliT and FliZ on the two complexes suggest that these two species have adapted in how they perceive FlhD4C2. We argue that these phenotypic differences arise from adaptations E. coli and S. enterica have made during evolution to expand or modify cellular function with respect to movement within specific environmental niches.

Results

Orthologous flhDC from E. coli can functionally complement flhDC in S. enterica

Given the similarities between the flagellar systems in S. enterica and E. coli, we sought to determine whether the FlhD4C2 master regulator is functionally equivalent in these two species of bacteria. To test this hypothesis, we replaced the flhDC genes in S. enterica (flhDCSE) with the flhDC genes from E. coli (flhDCEC). The reason that we performed these experiments in S. enterica rather than E. coli was that the flagellar system is better characterized in the former, particularly with regards to transcriptional regulation. To avoid plasmid associated artefacts associated with the ectopic expression of flhDC, we replaced the entire S. enterica flhDC operon with the flhDC operon from E. coli at the native chromosomal locus (Figure S2).

We first tested whether flhDCEC was motile as determined using soft-agar motility plates. As shown in Fig. 1A and B, these strains formed rings similar to the wild type. These results demonstrate that flhDCEC is functional in S. enterica. However, motility plates measure both motility and chemotaxis and do not provide any insights regarding possibly changes in the number of flagella per cell. To determine the impact flhDCEC had upon flagellar numbers we used a FliM-GFP fusion as a proxy for flagellar numbers (Fig. 1C). When this fluorescent protein fusion is expressed in cells, it forms spots associated with nascent C-rings that loosely correlate with the number of flagella12,13,14. By counting the number of spots per cell, we can determine the number of flagella made per cell. As shown in Fig. 1C, flhDCEC did not change flagellar numbers as compared to the wild type. These results demonstrate flhDCEC induces flagellar gene expression at similar levels as the wild type.

Figure 1
figure 1

(A) Motility of flhDCST and flhDCEC driven by PflhDC. (B) Quantification of swarms produced in motility agar after 6 to 8 hours incubation. Error bars indicate calculated standard deviations. (C) Percentage frequency of FliM-GFP foci for flhDCEC compared to S. enterica with flhDC under the control of PflhDC. Colors of bars in the graph correspond to the source of flhDC as shown in (B).

Full size image

flhDC requires a specific transcription rate to maintain optimal flagellar numbers

The flagellar network in S. enterica contains a number of feedback loops to ensure that the cells regulate the number of flagella produced4. One possibility is that these feedback loops mask any differences in FlhD4C2EC activity. To test this hypothesis, we replaced the native PflhD promoter with the tetracycline-inducible PtetA/tetR promoters. We then measured flagellar gene expression using a luciferase reporter system15. In this case, a consistent and significant change (e.g at 10 ng for PflgA ANOVA P = 0.0008) in flagellar gene expression was observed when comparing activity across all strains tested (Fig. 2A and B). Maximal expression of PflgA and PfliC, chosen to reflect flagellar gene expression at different stages of flagellar assembly5, for both complexes was observed between 10 and 25 ng/ml of anhydrotetracycline, when flhDC transcription was from PtetA (Fig. 2A and B). In contrast, PtetR, the weaker of the two tetracycline inducible promoters, reached a maximal output between 50 to 100 ng/ml anhydrotetracycline. When comparing PtetA and PtetR activity around the transition points in each experiment, for example 10 ng anhydrotetracycline for PflgA, the difference between PtetA and PtetR expression was significant (see Fig. 2 legend for P-values). However, the observed differences between FlhD4C2EC to FlhD4C2SE for either PtetA or PtetR expression were not significant (e.g. at 10 ng for PflgA via PtetA expression ANOVA P = 0.186).

Figure 2
figure 2

Titration of PtetA::flhDCST/EC and PtetR::flhDCST/EC activity suggests a given rate of transcription drives optimal flagellar assembly. (A) Activity of PflgA in response to PtetA or PtetR transcription of flhDC from S. enterica (S.e.) or E. coli (E.c.). Data sets that exhibit statistical significance at P < 0.03 are shown with ‘*’. Using 10 ng anhydrotetracycline as an example, due to this being where FlhD4C2SE reaches maximal activity via PtetA expression, the following comparisons are significant: S.e. PtetA v PtetR (P = 0.008) E.c. PtetA v PtetR (P = 0.009), while S.e. PtetA v E.c. PtetA is not (P = 0.186). Error bars show the standard error of the mean. (B) Activity of PfliC in response to PtetA or PtetR transcription of flhDC. As in (A) the ‘*’ identifies data sets that exhibit ANOVA statistical significance at P < 0.005. In agreement with PflgA activity PtetA v PtetR ANOVA comparisions were significant for 5 to 25 ng anhydrotetracycline (e.g. at 10 ng S.e.: P = 0.012; E.c.: P = 0.002) while S.e. PtetA v E.c. PtetA comparisons were not. Error bars show the standard error of the mean. C. flagellar numbers as defined by FliM-foci in response to PtetA or PtetR transcription of flhDC. In agreement with the statistical analysis FliM-Foci profiles reflect the statistical significance associated with the expression data shown in (A) and (B). All data represents the analysis of gene expression or FliM-Foci from 3 independent repeats. FliM-Foci data is based on n > 400 cells for each data point. The colours of lines reflect the strains used in Figs 4, 5 and 6, for example in these figures, when graphs are used, S.e. flhDC is represented as gray and its E.c. flhDC replacement as light blue.

Full size image

We also measured the number of FliM-GFP foci at different anhydrotetracycline concentrations. PtetR::flhDC expression generated on average of approximately two FliM-foci per cell at 25 ng/ml of anhydrotetracycline for both FlhD4C2 complexes (Fig. 2C). In contrast, 5 ng/ml induction of the PtetA::flhDCEC strain was sufficient to generate typical FliM-foci numbers (approx. 8 flagellar foci per cell). These data reflect the statistical significance of the expression data where a marked difference between PtetA and PtetR expression was observed (Fig. 2A and B). Even with the strong decrease in average foci per cell at these levels of induction for PtetR, the number of basal bodies observed is sufficient to allow motility at comparable levels in the motility agar assay (Figure S3).

Replacement of flhC but not flhD in S. enterica with the E. coli orthologs affects motility

The hetero-oligomeric regulator FlhD4C2 is unusual in bacteria as the majority of transcriptional regulators are believed to be homo-oligomeric complexes. To determine the relative contributions of the two subunits, we individually replaced the flhC or flhD genes from S. enterica with their ortholog from E. coli (Figure S2). When we tested the two strains using motility plates, we found that motility was inhibited in the strain where flhCEC replaced the native S. enterica flhC (Fig. 3A; blue bars), with an 88% reduction in swarm diameter when compared to WT S. enterica. The introduction of flhDEC compared to flhDCEC or flhDCSE produced swarms of a comparable size (Fig. 3A; blue bars).

Figure 3
figure 3

Motility phenotypes and gene expression of flhDCST, flhDCEC, flhDEC and flhCEC strains in the absence of known FlhD4C2 regulators. (A) Quantification of n = 3 swarms per strain produced in motility agar after 6 to 8 hours incubation at 37 °C. Error bars indicate calculated standard deviations. (B) Relative activity of PfliC in all strains as a percent of the maximal activity observed in flhDEC ∆rflP.

Full size image

Using the dose-dependent inducible PtetA promoter16 we observed that PtetA expression of flhCEC led to reduced PflgA transcription and strongly reduced PfliC transcription (Fig. 4). Strains expressing flhDEC in S. enterica showed a mild increase in PflgA gene expression and a similar response for PfliC, although these changes were not significant (see Fig. 4 for P values). These data suggest that the combination of FlhDSE and FlhCEC generates an inefficient FlhD4C2 complex, resulting in reduced motility.

Figure 4
figure 4

Titration of PtetA::flhDC for S. enterica, flhDCEC, flhDEC and flhCEC suggests that flhCEC exhibits low motility due reduced PflgA activity and a strong reduction in PfliC activity. Note that the legend indicates which gene has been replaced compared to S.e. flhDC. The colours of lines reflect the strains used in Figs 2, 5 and 6, for example in all figures S.e. flhDC is represented as gray and the E.c. flhDC replacement as light blue. Inducible expression was driven from the PtetA promoter within the TetRA cassette of Tn10 as in Fig. 2. Data represents n = 3 independent repeats of the expression assays. Data sets exhibiting ANOVA statistical significance of P < 0.03 are indicated with a ‘*’. Error bars show the standard error of the mean. The PflgA variation observed between S.e. flhDC, E.c. flhDC and E.c. flhD at 10 and 25 ng was not significant (ANOVA P = 0.64 and 0.33 respectively). In agreement for PfliC data S.e. flhDC, E.c. flhDC and E.c. flhD exhibits ANOVA P-values of 10 ng: 0.16 and 25 ng: 0.07. All ANOVA statistical comparisons to E.c. flhC were significant P < 0.04.

Full size image

Orthologous FlhC and FlhD interaction is species specific and a key determinant of promoter recognition by the FlhD4C2 complex

The results above demonstrate that flhCEC is not functionally identical to flhCST. One possibility is that that FlhCEC is impaired in FlhD4C2 for DNA-binding. Alternatively, the stability of the FlhD4C2 complex is reduced in the flhCEC strain, leading to reduced FlhD4C2 activity. To test these hypotheses, we purified all combinations of the FlhD4C2 complex using affinity (Ni+ and heparin) chromatography (Fig. 5A). In each complex, FlhD was tagged with a carboxy-terminal hexa-histidine to facilitate affinity purification. Such expression constructs have previously been used successfully to purify the FlhD4C2 complex17,18. Using either Ni+ affinity or heparin purification, we observed complete complex retrieval for three combinations (Fig. 5A). FlhC recovery was less efficient in the FlhDSE/FlhCEC complex. In contrast, no FlhDSE/FlhCEC complex was recovered via Heparin purification, used to mimic DNA during protein purification of DNA-binding proteins (Fig. 5A). This suggests that the FlhDSE/FlhCEC complex is less stable, resulting on a lower yield of complex retrieval.

Figure 5
figure 5

The FlhDSTFlhCEC complex is an active but unstable complex. (A) Protein gel showing purified complexes with either HIS6 or Heparin based purification protocols. The nature of the FlhDC complex allows isolation of both proteins in these assays. Arrows indicate the FlhC and FlhD bands. The image shown is the complete gel down to the leading edge of the loading buffer. The unprocessed raw image is shown in Figure S4. (B) Quantification of the unbound DNA during EMSA to define the binding ability of the complex combinations compared to S. enterica FlhD4C2. The protein complexes used in these assays were isolated via the HIS6 protocol as indicated in A by the corresponding coloured symbols that the act as the key for (B). All colours reflect the same complex associated with data shown in Figs 2, 4 and 6 for continuity, for example FlhDCSE is gray. Error bars show the standard error of the mean. See text for values of the calculated slopes using the excel built-in function SLOPE to highlight the impact of FlhCEC in each isolated complex.

Full size image

We next used the EMSA assays to test all four protein complexes for their ability to bind the S. enterica PflgAB promoter region. Quantification of the DNA shifts showed that complexes containing the orthologous FlhCEC reduced the PflgAB promoter binding profile, compared to FlhCSE complexes (Fig. 5B). This difference is exemplified when calculating the SLOPE (an excel function) of each data set. For FlhDCSE and FlhDECFlhCSE the slopes were −906 and −784 respectively. In comparison FlhDCEC and FlhDSEFlhCEC were much shallower at −1570 and −1116 respectively. This is consistent with FlhC being the DNA binding subunit of the complex and the variation in FlhD4C2 activated promoter-binding sites between S. enterica and E. coli19. Therefore, these results suggest that FlhC is a key determinant of DNA binding ability. Furthermore, the reduction in FlhCEC motility and flagellar gene expression in S. enterica is a result of the FlhDSE/FlhCEC complex being unstable, ultimately reducing the cellular concentration of the FlhD4C2 complex.

FlhD4C2EC responds to proteolytic regulation

S. enterica and E. coli both regulate the FlhD4C2 complex through ClpXP-mediated proteolytic degradation. Proteolytic degradation of FlhD4C2 plays a fundamental role in facilitating rapid responses to environmental changes that require motility20,21. The FlhD4C2 complex has a very short half-life of approximately 2–3 minutes22. Proteolytic degradation of FlhD and FlhC is regulated in E. coli and S. enterica by RflP (previously known as YdiV)23. However, rflP is not expressed under standard laboratory conditions in model E. coli strains, suggesting that ClpXP activity is modulated in a species-specific manner7.

Previous work has shown that RflP delivers FlhD4C2 complexes to ClpXP for degradation24. We have assessed the impact on motility for ∆clpP and ∆rflP mutations (Fig. 3). The ∆clpP and ∆rflP mutants exhibited improved motility and flagellar gene expression, including the FlhDSE/FlhCEC strain (Fig. 3A and B). These results suggest that proteolytic degradation mechanism of FlhD and FlhC, and its regulation, is common to E. coli and S. enterica.

To complement the motility assays, we investigated how ∆clpP and ∆rflP mutations impact the number of FliM-foci in cell. Both ∆clpP and ∆rflP mutants showed an increased number of FliM-foci compared to the wild type (Fig. 6A–C). For flhCEC strain, FliM-foci were observed in 13% of the population where individual cells exhibited just one or two foci. However, the ∆clpP or ∆rflP mutants increased the flagellated population of the flhCEC strains to 51 and 46% respectively, albeit with the majority still possessing only a single FliM focus (Fig. 6 B and C).

Figure 6
figure 6

Impact of protein stability regulators of FlhD4C2 on flagellar numbers as defined by FliM-foci. Quantification of FliM-foci was performed using the semi-automatic protocols defined with in Microbetracker. (A) Wild Type foci distribution; (B) ∆clpP; (C) ∆rflP. All line and symbol colours reflect the same complex associated with data shown in Figs 2, 4 and 5 for continuity, for example S. enterica (FlhDCSE) is gray.

Full size image

FliT and FliZ regulation of FlhD4C2 complexes

FlhD4C2 activity has an additional level of regulation in S. enterica via the flagellar-specific regulators FliT and FliZ. FliT functions as an export chaperone for the filament cap protein, FliD, and is a regulator of FlhD4C2 activity17,25. FliT disrupts the FlhD4C2 complex but is unable to disrupt a FlhD4C2:DNA complex. Therefore, FliT modulates availability of FlhD4C2 complexes for promoter binding17. In contrast, FliZ is a negative regulator of rflP expression26,27 and modulates the activity of HilD28,29 and thus increases the number of FlhD4C2 complexes in S. enterica.

In motility assays of ∆fliT mutants, we observed a difference between the flhDC strains. Motility is increased in a ∆fliT mutant background in S. enterica30 (and Fig. 3A). However, when flhDCEC and flhDEC replaced the native genes, a reduced swarm size was observed (Fig. 3A). Furthermore, quantification of PfliC activity agreed with the motility profile for ∆fliT mutants, where flhDCEC and flhDEC containing strains had reduced promoter activity compared to wild type (Fig. 3B). This suggests that the FlhD4C2 complexes are being perceived differently by FliT in S. enterica. The results for ∆clpP and ∆rflP mutants suggest that this is not due to protein stability, as all complex combinations reacted in a comparable fashion (Figs 3 and 6).

In contrast, the loss of fliZ resulted in a consistent reduction in motility, except for the flhCEC strain. However, as the flhCEC strain was already impaired in motility, it is possible that the resolution of the motility assay was unable to identify differences in the ∆fliZ mutant. Flagellar gene expression activity did, however, suggest a 2-fold drop in PfliC expression in the flhCEC ∆fliZ strain as compared to the otherwise wild-type (Fig. 3B).

Analysis of FliM-foci distribution in ∆fliT mutant reinforced the observed discrimination of flhDCEC and flhDEC gene replacements. Calculating the average foci per cell, S. enterica ∆fliT mutants showed an increased average number of foci per cell from 2.9 to 6.3, while the flhDEC (fliT+: 3.4 versus ∆fliT: 4.2) and flhDCEC replacements (fliT+: 3.6 versus ∆fliT: 2.7) exhibited no significant changes (Fig. 7A). Interestingly, in a ∆fliZ mutant background, the FliM-foci analysis was able to differentiate flhDCEC and flhDEC from the native S. enterica flhDC strain. Both replacements exhibited an increase in the average foci compared to S. enterica ∆fliZ (Fig. 7A).

Figure 7
figure 7

FliT and FliZ regulation reflects when FlhCEC or FlhDEC are present. (A) FliM-Foci quantification is consistent with the observed motility phenotype of ∆fliT mutants. For ∆fliZ FliM-foci numbers discriminate between the source of FlhD, FlhDSE exhibits a consistnet drop in foci while FlhDEC containing strains show comparable foci averages. (B) Testing the hypothesis that ∆fliT mutants respond differently in E. coli compared to S. enterica. Note: this experiment in (B) uses the species E. coli and S. enterica not engineered replacements.

Full size image

These data suggest that there is a fundamental difference in how the FlhD4C2 complexes in E. coli and S. enterica respond to, at least, FliT regulation. There are two explanations for this: a) the E. coli combinations are being regulated via an unidentified mechanism in S. enterica or b) that they are insensitive to FliT regulation. Both arguments predict that in the species E. coli, FlhD4C2 may respond differently to FliT regulation. Comparing the species, not gene replacement strains, S. enterica and E. coli does indeed identify a difference in the response to a ∆fliT mutant. While a ∆fliT mutant in S. enterica leads to a consistent increase in FliM-foci, no significant difference is noted for an E. coli ∆fliT mutant compared to E. coli wild type (Fig. 7B). This suggests that the regulatory impact of FliT is very different in these two flagellar systems and the role FliT plays in S. enterica is potentially adaptive and species specific.

Discussion

Two model flagellar systems that form the foundation of the flagellar field are those from the enteric species E. coli and S. enterica. These two systems have led to key discoveries in relation to many aspects of flagellar structure, type 3 secretion, flagellar cell biology and the regulation of flagellar assembly. Textbook explanations suggest that most flagellar systems are being activated, regulated and built according to the models for E. coli and S. enterica. Modifications of transcriptional regulatory circuits contribute to the phenotypic diversity we see in closely related gene sets and we are only now able to investigate this in depth due to the tools available. Here we have taken a simple step and asked how do orthologous FlhD4C2 complexes function in the closely related species E. coli and S. enterica?

At the onset of our work it was known that FlhD4C2 from E. coli could sustain motility in S. enterica11. Our work was focussed on understanding and defining the species-specific differences in the regulon of two orthologous genes. Here we took advantage of the well-defined flagellar assembly tools to measure outputs such as, motility, flagellar assembly per cell and flagellar gene expression. Bioinformatic analysis identifies only an 8 and 6% identity difference between FlhD and FlhC in E. coli and S. enterica respectively, suggesting that these proteins function in an analogous fashion. It is well established that related taxa usually rely on orthologous regulators to coordinate response to a given signal10.

The fine detail of the differences in the FlhD4C2 complexes only became apparent when we began to focus on their effect on flagellar gene expression and flagellar assembly. Biochemical analysis of isolated complexes showed that FlhCEC had weaker DNA binding ability to the PflgAB promoter region from S. enterica, consistent with previous investigations into FlhD4C2 DNA binding activity19. The isolation of FlhD4C2 complexes from our strains suggested that a key aspect of the phenotypes we observed, was the stability of the complexes formed.

With respect to flhDC transcription we show a discrepancy in flagellar numbers defined by FliM-foci when using PtetA/PtetR::flhDC expression. This was somewhat surprising as all constructs exhibited good swarming ability on motility agar plates (Figure S3). Original studies on the regulation of PtetA/PtetR from Tn10 have shown that these two promoters have differing activities but both respond to TetR regulation. We show that even though maximal activity of PflgA and PfliC can reach 40–50% of PtetA::flhDC expression for PtetR strains, this results in an average of 2 flagella per cell. This suggests that even though the majority of the literature states that E. coli and S. enterica produce between 4 and 8 flagella per cell, only 1 or 2 per cell is needed for an optimal output of the system with respect to motility agar assays. This conclusion correlates with the observation that swimming speed does not depend on flagella numbers in E. coli31.

It has been shown that FliT interacts with FlhC and that in S. enterica the output of this circuit is to destabilize FlhD4C2 complexes that are not bound to DNA. Our data suggests that this level of regulation does not impact E. coli FlhC. The nature of the adaptability needed by the favourable conditions to drive motility in E. coli may have led to the FliT regulatory input becoming less critical. Indeed, the specific amino acid substitutions between FlhCEC and FlhCST merits further investigation, outside the focus of this study, to determine whether this can be defined by a single substitution or requires the combination of the changes observed between these two proteins (Figure S1). Similarly, the impact of FliZ regulation becomes apparent for FlhDEC containing complexes when we assess flagellar numbers. FliZ regulates the transcription of rflP in S. enterica27. It is plausible that the impact in changing rflP regulation is the source of this differentiation, especially as RflP is proposed to interact with FlhDSE. Furthermore, we know that rflP is not expressed in model E. coli strains, strengthening the argument that FlhDEC has adapted to the absence of RflP or vice versa FlhDSE to RflP. However, regulation of flagellar gene expression in S. enterica via FliZ must take in to consideration other regulators such as HilD and its impact on flhDC gene expression9,28,29.

Importantly our analysis shows that even though these two systems are genetically similar, investigation of FlhD4C2 activity identifies subtle but key differences into how the FlhD4C2 complex is modulated in two closely related species. We argue that this is a valid example of the caution needed in the age of synthetic biology to exploit heterologous systems in alternative species or chassis’. Our data shows that even systems showing significant synteny may not behave in exactly the same manner and due diligence is required in making assumptions based on heterologous expression.

Materials and Methods

Bacterial Strains and Growth conditions

S. enterica and E. coli strains used in this study have been previously described elsewhere12,15,17,30. This study used S. enterica serovar Typhimurium strain LT2 as the chassis for all experiments. E. coli genetic material was derived from MG1655. All strains were grown at either 30 °C or 37 °C in Luria Bertani Broth (LB) either on 1.5% agar plates or shaken in liquid cultures at 160 rpm17. Antibiotics used in this study have been described elsewhere32. Motility assays used motility agar17 incubated at 37 °C for 6 to 8 hours. Motility swarms were quantified using images captured on a standard gel doc system with a ruler in the field of view and quantified using ImageJ to measure the vertical and horizontal diameter using the average as the swarm size. All motility assays were performed in triplicate using single batches of motility agar.

Genetic Manipulations

For the replacement of flhDC coding sequences the modified lambda red recombination system described by Blank et al. (2011) was used33. Deletion of clpP, rflP, fliT and fliZ was performed using the pKD system described by Datsenko and Wanner (2000)34. PtetA/PtetR replacements of the PflhDC region was also performed using Datsenko and Wanner (2000) with the template being Tn10dTc35. For Blank et al. (2011) replacement experiments we used autoclaved chlortetracycline instead of anhydrotetracycline as described for the preparation of Tetracycline sensitive plates36. All other gene replacements were performed as previously described17. All primers used for these genetic manipulations are available on request.

Quantification of flagellar gene expression

Flagellar gene expression assays were performed using the plasmids pRG39::cat (PfliC) and pRG52::cat (PflgA)15. Both plasmids were transformed into strains using electroporation. Gene expression was quantified as described previously and analysis was based on a minimum of n = 3 repeats for each strain tested15.

Quantification of FliM-GFP foci

FliM-GFP foci were quantified using Microbetracker on images captured using a Nikon Ti inverted microscope using filters and exposure times described previously14. Strains were grown to an OD600 of 0.5 to 0.6 and cells immobilised using a 1% agarose pad containing 10% LB14,17. For each strain a minimum of 5 fields of view were captured from 3 independent repeats. This allowed analysis of approximately 400–1000 cells per strain. For the comparison of FliM foci in E. coli ∆fliT to S. enterica ∆fliT shown in Fig. 7B the chemostat growth system described by Sim et al. (2017) was used. For this experiment the growth rate of both strains was similar to batch culture in LB at 37 °C where the media used was a Minimal E base salts, a minimal media previously described14,17, supplemented with 0.1% Yeast extract and 0.2% glucose.

Purification of FlhD4C2 complexes

Purification of proteins complexes was based on previously described methods17. Wild type FlhD4C2SE was purified using pPA158. The other 3 complexes were purified from plasmids generated using the New England Biolabs NEBuilder DNA Assembly kit on the backbone of pPA158. The E. coli strain BL21 was used for all protein induction experiments prior to protein purification using either a pre-equilibrated 5 ml His-trap column or a 5 ml heparin column (GE Healthcare). Proteins were visualised using Tricine-based SDS polyacrylamide gel electrophoresis and standard commassie blue staining17.

Electrophoretic mobility shift assay (EMSA)

All EMSA assays were performed using Ni++ (his-trap) purified proteins as this allowed analysis of all four complexes (Fig. 5A). Buffer exchange from elution buffer to a 100 mM Tris-HCl, 300 mM NaCl 1 mM DTT (pH 7.9) buffer was performed through 10 cycles of protein concentration in VivaSpin columns with 20 ml buffer reduced to 5 ml per round of centrifugation at 4500 rpm. A protein concentration range of 100 to 700 nM was used with 80 ng/ml of a PCR product containing PflgAB from S. enterica. After incubation bound and unbound DNA were resolved using 5% acrylamide gels made with 1x TBE buffer. Quantification of gel images was performed using ImageJ.