The temporal and spatial transcription of late flagellar genes in Caulobacter crescentus is regulated by the σ54 transcriptional activator, FlbD. One requirement for FlbD activity is the assembly of a structure encoded by early, class II flagellar genes. In this report, we show that the trans-acting factor FliX predominantly functions as a negative regulator of FlbD activity in the absence of the class II-encoded flagellar structure. In contrast, a mutant FliX that bypasses the transcriptional requirement for early flagellar assembly is incapable of repressing FlbD in a class II flagellar mutant. Expression of this mutant allele, fliX1, does not alter the temporal pattern of FlbD-dependent transcription. Remarkably, this mutation confers the correct cell cycle timing of hook operon transcription in a strain that cannot assemble the flagellum, indicating that the progression of flagellar assembly is a minor influence on temporal gene expression. Using a two-hybrid assay, we present evidence that FliX regulates FlbD through a direct interaction, a novel mechanism for this class of σ54 transcriptional activator. Furthermore, increasing the cellular levels of FliX results in an increase in the concentration of FlbD, and a corresponding increase in FlbD-activated transcription, suggesting that FliX and FlbD form a stable complex in Caulobacter. FliX and FlbD homologues are present in several polar-flagellated bacteria, indicating that these proteins constitute an evolutionarily conserved regulatory pair in organisms where flagellar biogenesis is likely to be under control of the cell division cycle.
The late flagellar genes encode structures that are assembled externally relative to the cytoplasmic membrane, including the basal body rods, the outer membrane rings, the hook and filament (Fig. 1) (reviewed in Gober and England, 2000; England and Gober, 2001). Flagellum synthesis initiates in the stalked cell type with the activation of the global transcription factor, CtrA (Quon et al., 1996). CtrA-phosphate activates the transcription of the early, class II flagellar genes encoding the cytoplasmic membrane MS-ring (FliF), the flagellar switch proteins and the flagellum-specific type III secretory apparatus (Quon et al., 1996; Domian et al., 1997; Reisenauer et al., 1999). The subsequent expression of the remaining flagellar genes (totalling nearly 30) is subject to a trans-acting regulatory hierarchy, such that the assembly of the class II-encoded polypeptides into a basal body substructure is required for the transcription of the late, class III/IV flagellar genes (Fig. 1) (Newton et al., 1989; Xu et al., 1989; Ramakrishnan et al., 1994; Mangan et al., 1995). In a subsequent assembly checkpoint, the assembly of the rods, outer basal body rings and then the hook structure is required for the translation of flagellins (Mangan et al., 1995; Anderson and Newton, 1997). Mutations in the regulatory gene flbT bypass this checkpoint, permitting flagellin translation in the absence of a hook structure (Mangan et al., 1999; Anderson and Gober, 2000). Additionally, flbT mutant strains exhibit an aberration in the cell cycle timing of flagellin translation, indicating that assembly checkpoints significantly influence temporal gene expression (Mangan et al., 1999).
The trans-acting factor encoded by the class II fliX gene has been shown to be a key regulator in coupling early flagellar assembly events to the activation of late flagellar gene transcription (Muir and Gober, 2001; 2002; Muir et al., 2001). Strains bearing a deletion in fliX are non-motile, and exhibit a cell division defect characteristic of class II flagellar mutants (Mohr et al., 1998; Muir et al., 2001). These strains fail to transcribe class III and class IV genes (Mohr et al., 1998; Muir et al., 2001), a characteristic that can be bypassed by a bfa mutation in flbD (Muir et al., 2001; Muir and Gober, 2002). Introduction of a bfa mutation into fliX mutant strains also restores motility to a level that is indistinguishable from wild-type cells, suggesting that FliX is a positive regulator of FlbD activity (Muir et al., 2001; Muir and Gober, 2002). Interestingly, a mutant allele of fliX (fliX1) was isolated that bypassed the assembly requirement for late flagellar gene transcription, indicating that FliX also functions as a negative regulator of class III/IV transcription (Muir et al., 2001). In one proposed model, FliX ‘senses’ the completion of assembly of the MS-ring-basal body complex and transduces this information to FlbD by switching from a negative to a positive regulator of FlbD activity (Muir et al., 2001).
In this study, we investigate the mechanism of FliX-mediated regulation of late flagellar gene expression. Through epistasis experiments, we show that FliX negative regulation of FlbD activity is accentuated in strains containing mutations in class II flagellar genes. In contrast, FliX1, the allele of FliX that bypasses the transcriptional requirement for a early flagellar structure, is only capable of activating FlbD and cannot repress gene expression. Unlike gain-of-function FlbD mutants, the presence of FliX1 did not significantly alter the temporal pattern of transcription of the class III, fliK gene. Both wild-type and mutant versions of FliX influenced the cellular concentration of FlbD, with the levels of FlbD being directly proportional to those of FliX, suggesting that FliX and FlbD may form a complex inside the cell. In support of this idea, we show that FliX and FlbD directly interact with each other in a bacterial two-hybrid assay.
Previous experiments have shown that fliX is required for motility as well as the expression of FlbD-dependent, class III and IV flagellar genes (Mohr et al., 1998; Muir et al., 2001; Muir and Gober, 2002). FliX is hypothesized to couple the assembly of a class II-encoded flagellar structure to the activation of the transcription factor FlbD, because a mutant allele of FliX, fliX1, was shown to bypass the transcriptional requirement for an early flagellar structure (Muir et al., 2001). The fliX1 mutant was isolated by screening for fliX alleles that would permit expression of a class III/IV transcriptional reporter, fljL-lacZ, in a class II flagellar mutant strain. Here, we first wanted to test the mutant fliX1 gene for the ability to restore motility to a strain bearing a deletion in the chromosomal copy of fliX (ΔfliX). When inoculated into semi-solid motility agar and permitted to incubate for 4 days, wild-type cells exhibited a swarm of growth that radiated away from the point of inoculation, whereas non-motile, mutant cells, such as a strain with a Tn5 insertion in flbD, grew as a relatively compact colony (Fig. 2A). A strain containing a deletion in fliX characteristically produced a compact colony, with cells containing motile suppressor mutation(s) (usually gain-of-function mutations in flbD; Muir et al., 2001; Muir and Gober, 2002) exhibiting a flare of growth on the periphery of the colony (Fig. 2A). Both fliX and fliX1 introduced into this strain on a plasmid resulted in the restoration of motility to a level comparable to that of wild-type cells (Fig. 2A).
Sequence analysis revealed that the fliX1 coding region possessed a −1 frameshift at codon number 141 (out of 144) predicted to generate a larger polypeptide of ≈22 kDa (wild-type FliX is 14.6 kDa) (Muir et al., 2001). We compared the electrophoretic mobility of FliX1 with wild-type FliX using immunoblot with anti-FliX antibody after SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (Fig. 2B). The fliX1-encoded polypeptide was ≈6 kDa larger than that encoded by wild-type fliX as determined by relative mobility in the gel (Fig. 2B). Previously, we demonstrated that a fraction (≈10–20%) of FliX was peripherally associated with the cytoplasmic membrane, with the bulk of the protein in the cytoplasm (Muir et al., 2001). One plausible hypothesis is that FliX activity may change upon either association with the membrane or the nascent flagellar structure. Because FliX1 is apparently active in the absence of flagellar assembly, we wanted to determine whether it possessed a subcellular localization pattern that differed from wild-type FliX. The mutant, however, localized similarly to its wild-type counterpart, with the majority of the protein in the cytoplasmic fraction (data not shown). This result suggests that FliX1 does not have different regulatory properties because of an altered association with the cytoplasmic membrane.
The FliX1 mutation abolishes negative regulation of FlbD activity
Next we wanted to compare the performance of the mutant fliX1 with that of wild-type fliX in regulating FlbD activity. In this first experiment, we compared the ability of these two FliX alleles to repress FlbD activity when expressed from a multicopy plasmid (pBBR1-MCS). Specifically, we assayed the expression of a fliF-lacZ transcriptional reporter fusion in both wild-type and class II flagellar mutant strains when either fliX or fliX1 was overexpressed. Previous experiments have shown that FlbD represses the early, class II fliF promoter by binding to a DNA sequence that overlaps the −35/−10 promoter sequences (Mullin et al., 1994; Wingrove and Gober, 1994). For example, in wild-type cells, the fliF-lacZ fusion generated 2904 units of β-galactosidase activity, and in a strain with a Tn5 insertion in flbD, expression increased almost fivefold (14 263 units) (Fig. 3). Likewise, a strain bearing a deletion in fliX (ΔfliX) exhibited an approximately twofold increase in fliF-lacZ expression relative to wild-type cells, which is attributable to a decrease in FlbD activity (Muir et al., 2001; Muir and Gober, 2002) (see also Fig. 4A). Introduction of wild-type fliX on a multicopy plasmid (pX10) into the ΔfliX strain had no effect on fliF-lacZ expression (Fig. 2). In contrast, the introduction of the mutant fliX1 on a multicopy plasmid (pfliX1) into the ΔfliX resulted in a decrease in β-galactosidase activity to levels resembling wild-type cells. This result indicates that the mutant FliX1 can function to enhance FlbD activity.
We next tested how flagellum assembly influenced fliF-lacZ promoter activity in cells expressing either wild-type fliX or the mutant fliX1. We performed the same basic experiment described above, but instead used a strain containing a Tn5 insertion in the class II, fliP gene, which encodes a component of the flagellum-specific secretory apparatus (Gober et al., 1995). In this strain, the fliF-lacZ fusion was expressed at levels three times greater than wild-type cells, indicative of a decrease in FlbD activity (Fig. 3). Introduction of wild-type fliX on a plasmid into the fliP::Tn5 mutant caused an even greater reduction in FlbD activity, increasing reporter gene expression almost sixfold over wild-type cells. Importantly, this strain (fliP::Tn5/pX10) exhibited a marked reduction of FlbD activity compared with the ΔfliX/pX10 strain (15 881 versus 5791 units). Thus, the repressive effect of overexpressing fliX is accentuated in a strain that cannot assemble a flagellum (i.e. fliP::Tn5). In contrast, FlbD activity was significantly higher (7248 versus 15 881 units) in the fliP::Tn5 strain expressing the mutant fliX1 (Fig. 3), indicating that this allele of fliX could not repress FlbD activity as effectively as wild-type fliX. However, the fliP::Tn5/pfliX1 strain still had more fliF-lacZ promoter activity than a strain that was competent for flagellum assembly (ΔfliX/pfliX1), suggesting that there may be some negative regulation of FlbD activity. One possibility was that the copy of wild-type fliX on the chromosome of the fliP::Tn5/pfliX1 strain was responsible for negatively regulating FlbD. In order to test this, we performed an additional set of experiments assaying fliF-lacZ expression in a strain that contained a deletion in fliX and a Tn5 insertion in fliP (ΔfliX fliP::Tn5). Therefore, the sole copy of fliX in these cells was on the multicopy plasmid. When wild-type fliX was overexpressed in this strain, the activity of the fliF-lacZ reporter increased relative to the strain without fliX on a plasmid (9087 versus 4862 units). A distinctly opposite effect was observed when fliF-lacZ expression was assayed in the ΔfliX fliP::Tn5 mutant strain with fliX1 on a plasmid. The expression of the mutant fliX1 resulted in an activation of FlbD (i.e. a reduction in fliF-lacZ expression) to a level that approached wild-type cells (3837 versus 2904 units) (Fig. 3). These results suggest that wild-type fliX, when overexpressed from a plasmid, functions predominantly as a negative regulator of FlbD activity in the absence of an early flagellar structure. The mutant fliX1 was incapable of repressing FlbD, and only exhibited the ability to activate FlbD, regardless of whether or not the strain was competent for flagellum assembly. These results are consistent with the idea that FliX switches between two states: functioning as a negative regulator in the absence of an early assembled flagellar structure, and as a positive regulator of FlbD activity, after assembly.
Effect of fliX1 on the temporal transcriptional activation of late flagellar genes
We next compared the mutant FliX1 with wild-type FliX in regulating FlbD-activated transcription. We performed the same basic experiment as described above, but used a class III reporter gene, fliK-lacZ (Fig. 4). This fusion generated ≈10-fold less β-galactosidase activity in the ΔfliX strain compared with wild-type cells (Fig. 4A). Introduction of wild-type fliX on plasmid completely restored promoter activity to wild-type levels. However, even though the mutant fliX1 functions solely as a positive regulator, its presence on a plasmid in the ΔfliX strain resulted in about one-half the promoter activity than when wild-type fliX was supplied in trans (1236 versus 2510 units) (Fig. 4A). This may be a consequence of the fact that FliX1 protein concentration is lower than that of wild-type FliX when both are expressed from this plasmid (see below) or may simply indicate that FliX1 does not activate FlbD as well as wild-type FliX. We next compared the ability of wild-type FliX and mutant FliX1 to activate transcription in the class II flagellar mutant fliP::Tn5 (Fig. 4A). Similar to results that were reported previously, only FliX1, and not wild-type FliX, was capable of stimulating class III transcription, although to modest levels, in the class II flagellar mutant (Muir and Gober, 2002).
Mutations of flbD (bfa) that bypass the transcriptional requirement for early flagellar assembly result in a partial loss of temporally regulated transcription (Mangan et al., 1995; Muir and Gober, 2002). Cell cycle experiments with strains containing these mutant alleles show strong expression of class III reporter genes in swarmer cells, whereas in wild-type cells, class III transcription ceases before cell division. Because the fliX1 mutation bypasses the same flagellar assembly checkpoint, we wanted to determine whether or not the presence of this mutant gene would also result in an alteration in the temporal pattern of class III transcription. In order to accomplish this, mutant strains carrying either wild-type fliX or mutant fliX1 on a multicopy plasmid and a class III, fliK-lacZ transcription reporter gene were synchronized, the swarmer cells were placed in fresh medium and permitted to proceed through the cell cycle. At various time points, a sample was removed, protein was pulse-labelled and β-galactosidase was immunoprecipitated. A ΔfliX strain with a wild-type copy of fliX supplied in trans exhibited a wild-type pattern of fliK-lacZ expression, with promoter activity peaking in early to late predivisional cells (Fig. 4B, top). Interestingly, the fliX1 allele, when present in multicopy on a plasmid, had no detrimental effect on the timed transcription of the fliK-lacZ reporter, with the peak in expression at the same predivisional cell stage (Fig. 4B, middle). Additionally, in both cases, there was little to no transcription in the newly isolated swarmer cells. Thus, unlike the flbD bfa strain, cells containing fliX1 do not have a grossly altered cell cycle pattern of late flagellar gene expression.
An implication of the above observation is that the progression of flagellar assembly may have a minor role in the temporal activation of FlbD. In order to explore this idea, we introduced pfliX1 into a ΔfliX fliP::Tn5 strain, which does not assemble a class II flagellar structure, and assayed the temporal transcription of the fliK-lacZ transcriptional reporter (Fig. 4B). This strain exhibited a low level expression of fliK-lacZ throughout the swarmer and early stalked cell phase; however, a strong peak in expression occurred in the predivisional cell stage (Fig. 4B). Thus, remarkably, even in the absence of a flagellar structure, the presence of fliX1 could confer a fliK-lacZ cell cycle expression pattern similar to that of wild-type cells (Fig, 4B). This observation is consistent with the conclusion that flagellar assembly is not required for the temporal pattern of class III and IV transcription, but is, however, necessary for maximal FlbD activity.
These experiments also show that the constitutive fliX1 allele has a distinctly different effect on temporal class III gene transcription from the bfa alleles of FlbD. We performed a direct comparison of the effect of an flbD bfa mutation (flbD-1204) on the relative levels of expression of the fliK-lacZ fusion early in the cell cycle, with wild-type cells, and with ΔfliX cells expressing either fliX or fliX1 from a multicopy plasmid (Fig. 5A). (Note that, data presented for both wild-type and flbD-1204 in Fig. 5A were quantified from phosphorimages of previously published immunoprecipitation experiments (Muir and Gober, 2002). In freshly isolated, wild-type swarmer cells, fliK promoter activity is low, and gradually increases upon differentiation into stalked cells (Fig. 5A). The same basic pattern of expression occurs in ΔfliX cells containing either wild-type fliX or fliX1 on a plasmid (Fig. 5A). In contrast, cells containing a gain-of-function mutation in flbD (flbD-1204) exhibited relatively high expression in swarmer cells that gradually decreased as the cells become stalked cells (Fig. 5A).
We wanted to determine whether the high levels of fliK-lacZ expression in flbD-1204 swarmer cells could be attributable to active transcription or, as is the case with flagellin mRNA, from transcripts that were synthesized in predivisional cells and became trapped in the nascent swarmer cell upon division (Gober et al., 1991; Wingrove et al., 1993). In order to distinguish between these two possibilities, we isolated swarmer cells bearing the flbD-1204 mutation and the fliK-lacZ reporter (Fig. 5B). After the addition of rifampicin to inhibit transcription, protein was pulse-labelled over time, and labelled β-galactosidase was immunoprecipitated and subjected to SDS-PAGE. Rifampicin treatment resulted in a relatively sharp decrease in the rate of β-galactosidase synthesis over the course of the experiment indicating that the mutant FlbD-1204 protein was activating fliK transcription in swarmer cells. The ability to activate transcription in swarmer cells indicates that FlbD-1204, in contrast to FliX1, not only evades regulation by flagellar assembly, but also regulation by a cell cycle-related cue.
Bacterial two-hybrid analysis of FliX–FlbD interaction
There exist at least two distinct ways which FliX could modulate FlbD activity. In one scenario, FliX could regulate the activity of a sensor histidine kinase that phosphorylates FlbD. FlbD has been shown to undergo a temporally regulated phosphorylation, the peak of which coincides with the cell cycle peak of class III and IV flagellar gene transcription (Wingrove et al., 1993). Because the mutant FliX1 had no effect on the temporal pattern of class III transcription (see Fig. 4), we propose that FliX probably does not influence the phosphorylation state of FlbD through interaction with a sensor kinase. Alternatively, FliX may directly interact with FlbD to regulate its activity. In support of this idea is the observation that the presence of FliX is required for maintaining wild-type levels of FlbD (Muir and Gober, 2002). A plausible mechanism is that FliX and FlbD form a complex which is required for the stability of each partner. In order to test this possibility, we assayed for FliX–FlbD interaction using a LexA-based bacterial two-hybrid system (Dmitrova et al., 1998). This particular experimental system for assaying protein–protein interaction has the flexibility of permitting the analysis of both homodimerization or the formation of heterodimers (Porte et al., 1995; Dmitrova et al., 1998; Figge et al., 2003). In order to assay homodimer formation by FliX, we constructed a protein fusion with a carboxyl-terminal deleted derivative of LexA that cannot dimerize. The inability to dimerize results in expression of a lacZ reporter (Escherichia coli SU101) that contains a LexA binding site as an operator sequence. For example, previous experiments have shown that the introduction of a LexA(1-87)::Fos zipper fusion (pMS604) into E. coli strain SU101 results in the repression of lacZ expression (Dmitrova et al., 1998) (Fig. 6). Likewise, replacement of the Fos fusion on this plasmid with the entire fliX coding region fused in frame to LexA(1-87) also resulted in a significant repression of lacZ expression (Fig. 6), indicating that FliX is capable of forming dimers, and/or perhaps higher-order oligomers.
In order to test for FliX–FlbD interaction, we used a different strain (SU202) that contains the lacZ reporter with a mutant LexA operator. This mutant operator (op+/op408) possesses a wild-type LexA half binding site (op+) and a mutant LexA half-site (op408) that can only be recognized by a LexA derivative with a suppressor mutation in the DNA-binding domain (LexA408). Thus, full repression of lacZ expression in strain SU202 requires dimerization between wild-type LexA and LexA408. For example, as was shown previously (Dmitrova et al., 1998), expression of the LexA(1-87)::Fos zipper fusion (pMS604) and a LexA408(1-87)::Jun zipper fusion (pDP804) results in repression of lacZ expression (Fig. 7) (131 versus 1608 units in strain SU202 without plasmids). [Note that the presence of either fusion-containing plasmid, pMS604 or pDP804, by itself resulted in levels of β-galactosidase activity that were comparable to those assayed in SU202 without plasmid (data not shown)]. We fused the entire flbD coding region in frame to LexA408(1-87) in pDP804 replacing the Jun fusion (Dmitrova et al., 1998). Introduction of the LexA408(1-87)::FlbD (pREM88) fusion along with pMS604 into SU202 resulted in an over twofold reduction in β-galactosidase activity suggesting that the LexA408(1-87)::FlbD fusion could weakly interact with the LexA408(1-87)::Fos zipper fusion on pMS604. Similarly, when the LexA(1-87)::FliX fusion (pREM87) was introduced into SU202 with the pDP804 plasmid, there was also a weak repression of lacZ expression (1090 versus 1608 units). In contrast to both of these cases, when the LexA408(1-87)::FlbD and LexA(1-87)::FliX fusions (pREM88/pREM87) were both expressed in SU202 the reduction in lacZ expression was comparable to that of the positive control [i.e. a strain containing pDP804 (LexA408(1-87)::Jun) and pMS604 (LexA(1-87)::Fos)] (188 versus 131 units), indicative of a direct interaction between these fusions (Fig. 7A). As an additional control, we assayed the relative levels of the LexA408(1-87)::FlbD and LexA(1-87)::FliX fusions when induced with different concentrations of IPTG (Fig. 7B). The addition of increasing amounts of IPTG to the growth medium resulted in an increase in the steady state concentrations of the LexA408(1-87)::FlbD and LexA(1-87)::FliX fusions up to an IPTG concentration of 0.1 mM (Fig. 7B). Unfortunately, this analysis revealed that a LexA(1-87)::FliX1 fusion was degraded and/or poorly expressed so that we were unable to determine whether this mutant protein interacted with FlbD (not shown). The repression of β-galactosidase synthesis was directly proportional to the concentration of IPTG in the medium (up to 1 mM) (data not shown). Increasing the IPTG concentration in the medium beyond this, resulted in no further induction of the fusions (Fig. 7B) or significant decrease in β-galactosidase activity (data not shown). The strong repression of lacZ expression when the LexA408(1-87)::FlbD and LexA(1-87)::FliX fusions are induced at this high concentration of IPTG indicates that FliX and FlbD directly interact with each other in a relatively stable complex.
Evidence for a direct relationship between FliX–FlbD levels and the magnitude of flagellar promoter activity
Although both FlbD and FliX are present throughout the cell cycle, their levels increase in late predivisional cells at the time when class III and IV flagellar genes are maximally expressed (Wingrove et al., 1993; Mohr et al., 1998). It is possible that the concentration increase in these regulators contributes to the temporal pattern of class III/IV transcription. In previous experiments, we had shown that the levels of FlbD were decreased in a ΔfliX mutant strain (Muir and Gober, 2002). We wanted to test whether increasing the cellular concentration of FliX or FliX1 would result in an increase both in the levels of FlbD and in the transcription of FlbD-dependent genes. In order to accomplish this, we assayed the expression level of a class IV flagellar transcriptional reporter, fljL-lacZ, integrated into the chromosome of a ΔfliX mutant strain (Fig. 8). We introduced fliX into this strain, either on a high-copy-number plasmid (pX10; 15–20 copies per cell) or on a lower-copy-number plasmid (pX09P; three to six copies per cell) and assayed the effect on FlbD levels and fljL-lacZ expression (Fig. 8). The presence of fliX on the lower-copy-number plasmid, pX09P, resulted in a twofold increase in FliX concentration and 4.2-fold increase in FlbD levels over those of wild-type cells (Fig. 8A) as well as a slight increase in fljL-lacZ expression (Fig. 8B). Expression of fliX from pX10 caused an even greater increase in fliX levels (2.5-fold) as well as the cellular concentration of FlbD (6.5-fold over wild-type cells), and a still greater increase in fljL-lacZ expression (Fig. 8B). This result indicates that a rise in the cellular concentration of FliX results in a parallel increase in FlbD levels and consequently, an increase in fljL promoter activity. We also tested whether the mutant FliX1 could have a similar effect on cellular levels of FlbD. When fliX1 was expressed from the lower-copy-number plasmid, pfliX1P, it was barely detectable by immunoblot (Fig. 8A). These low levels are most probably attributable to instability of the mutant FliX1 protein (see proteolysed fragments in pfliX1 lane). This low level of FliX1 consequently resulted in very low levels of FlbD (32% of wild-type levels), similar to those observed in ΔfliX cells (Fig. 8A). However, the level of fljL-lacZ expression almost approached that of wild-type cells (1273 versus 1621 units), in spite of slightly reduced levels of FlbD, strengthening the notion that FliX1 functions solely as a positive regulator. In results similar to those observed with wild-type cells, the presence of fliX1 on the higher-copy-number plasmid resulted in still higher cellular levels of FlbD (2.5-fold greater than wild-type cells) (Fig. 8A), and a resulting increase in fljL-lacZ expression (Fig. 8B). In summary, these results indicate that the level of class III and IV flagellar promoter activity responds to an increase in the cellular concentrations of FliX and FlbD and suggests that the temporally regulated increase in both these proteins may be a contributing factor in the cell cycle transcriptional activation of these genes.
Regulation of flagellar gene transcription in Caulobacter requires the integration of multiple temporal and spatial cues. Among these are the progression of the cell cycle and flagellar assembly. The focal point of these multiple inputs is the modulation of the activity of FlbD, the global transcriptional regulator of both early and late flagellar genes. In this report, we have investigated how the trans-acting factor FliX influences FlbD activity in response to early flagellar assembly. The evidence presented here suggests that FliX can exist either as a positive or as a negative regulator of FlbD activity, depending on the state of flagellar assembly. For example, we show that wild-type FliX functions predominantly as a negative regulator of FlbD-dependent transcription in the absence of a class II-encoded flagellar structure. A mutant form of FliX with a −1 frameshift in the carboxyl-terminal coding region (fliX1) abolishes the ability of the protein to respond to the absence of a completed class II structure, and repress FlbD activity. Unlike the gain-of-function mutations in FlbD that also bypass the transcriptional requirement for early flagellar assembly, the presence of the fliX1 allele had no effect on the temporal transcription pattern of a class III flagellar gene. This result suggests the progression of flagellar assembly is probably not the sole influence on FlbD-regulated cell cycle transcription.
FlbD is a sigma 54 transcriptional activator that consists of three distinct domains, a carboxyl-terminal DNA-binding domain, a conserved, central ATPase domain of the AAA+ family and an amino-terminal, response-regulator domain. The amino-terminal domain typically is phosphorylated in this class of transcriptional activator by a cognate sensor histidine kinase. Phosphorylation, in turn, activates ATPase activity in the conserved AAA+ domain, which ultimately results in stimulating open complex formation catalysed by sigma 54-containing RNA polymerase (Weiss et al., 1991; North et al., 1993; Porter et al., 1993; 1995). The sensor kinase(s) that catalyses the phosphorylation of FlbD has yet to be identified (Muir and Gober, 2001); however, phosphorylation more than likely plays a major role in regulating FlbD activity. In vivo experiments have shown that FlbD is phosphorylated in predivisional cells at a time coinciding with the peak in class III and IV flagellar gene expression (Wingrove et al., 1993). Furthermore, in vitro transcription experiments have shown that low-molecular-weight phosphodonors, such as phosphoamidate, can enhance FlbD-mediated transcriptional activation (Benson et al., 1994b). Thus, FlbD is similar in this regard to other sigma 54 transcriptional activators that possess response regulator domains. However, the primary amino acid sequence of the amino-terminal receiver domain of FlbD differs significantly from the vast majority of response regulator domains in other signal transduction proteins (see Volz, 1993; Muir and Gober, 2002). For example, FlbD is missing the critical conserved aspartate residues located at positions 12 and 13 in CheY (10 and 11 in NtrC) (Volz, 1993); these residues have been shown to have a role in the coordination of a magnesium ion that has an essential function in catalysis (Stock et al., 1993; Volkman et al., 1995; Cho et al., 2001). Likewise, FlbD is also lacking a catalytically essential lysine residue (K109 in CheY) (Cho et al., 2001). We propose that these critical differences in the amino-terminal receiver domain reflect a mode of regulation that is unique to FlbD, and very likely to involve FliX.
An analysis of the recently completed genome sequences within the α-proteobacterial subdivision reveals the existence of several highly conserved FlbD homologues (Fig. 9A). Comparison of the Caulobacter FlbD sequence with the predicted sequences of these homologues reveals a significant degree of sequence conservation throughout the entire polypeptide, including the unusual amino-terminal response-regulator domain (Fig. 9A). To date, FlbD homologues have been identified in five different members of the α-proteobacteria group including Magnetospirillum magnetotacticum, Rhodop-seudomonas palustrus, Rhodospirillum rubrum, Bradyrhizobium japonicum and Azospirillum brasiliense. Interestingly, in each case where the complete genome has been sequenced (all except Azospirillum), these organisms also possess fliX homologues. Each of these organisms possess polar flagella, sometimes in a bipolar arrangement, occurring either singly or in tufts. Both flbD and fliX homologues are located within clusters of flagellar genes, and in most cases (Magnetospirillum, Rhodop-seudomonas palustrus, Rhodospirillum rubrum) are adjacent to the same homologous genes in Caulobacter. This distribution of flbD and fliX infers that these organisms share a common mechanism in regulating late flagellar gene expression. We speculate that this mode of regulation may be related to cell cycle-regulated synthesis of the flagellum. The restriction of flagella to the cell poles, whether in an organism that is singly or bipolarly flagellated, requires that flagellum synthesis and/or assembly be coordinated with the cell division cycle, such that each progeny cell inherits a functional flagellum. Therefore, it is likely that the mechanisms of temporal and spatial regulation of flagellar biogenesis mediated by FliX/FlbD are evolutionarily conserved within this group of organisms.
The idea that FliX and FlbD constitute a conserved regulatory pair is strengthened by the experimental evidence indicating that these proteins physically interact with each other. We used two different experimental approaches to investigate this issue. First, the bacterial two-hybrid experiments indicate that FliX forms a stable, relatively strong interaction with FlbD. Second, immunoblot analysis showed that each protein is required for maintaining the cellular levels of the other, suggesting the existence of a stable complex in Caulobacter cells. In support of this idea, recent experiments indicate that FliX has a reduced stability in the absence of FlbD, and likewise, FlbD is unstable in the absence of FliX (data not shown). Here we show that increasing the cellular concentration of FliX resulted in a concomitant increase in FlbD levels that were, in turn, accompanied by an increase in class III/IV promoter activity. These data also suggest that direct interaction with FliX is required both for repression and activation of FlbD activity. As a consequence of instability of the LexA-FliX1 fusion, we were unable to assay for an interaction of this mutant with FlbD in the bacterial two-hybrid experiment. However, overexpression of fliX1 did result in an increase in FlbD levels suggesting that this positively acting form of FliX also interacts with FlbD, although with apparently less avidity than wild-type FliX.
Regulation of a sigma 54 transcriptional activator through direct contact with another protein, and in the absence of apparent covalent modification, is not common. NifA and PspF are two well-known examples of sigma 54 transcriptional activators that are repressed through interaction with regulatory proteins. NifA activates the transcription of genes encoding nitrogenase in a diverse group of organisms, and possesses a distinct amino-terminal domain that, along with the central AAA+ domain, interacts with the negative regulatory protein, NifL (Dixon, 1998; Money et al., 1999; Barrett et al., 2001). NifL represses NifA-dependent transcription in response to high oxygen tension, high ADP levels and nitrogen sufficiency (reviewed in Martinez-Argudo et al., 2004). The phage-shock protein, PspF, lacks an amino-terminal domain, containing only the conserved central ATPase and carboxyl-terminal DNA-binding domain. Thus, this protein is constitutively active (Jovanovic et al., 1999). Transcriptional activation is held in check by the binding of the regulatory PspA protein to the central AAA+ domain (Elderkin et al., 2002). The interaction of FliX with FlbD seems unusual when compared with NifA and PspF because FlbD, unlike these proteins, possesses a response regulator domain that is capable of receiving a phosphate. We hypothesize that the regulation of FlbD activity via FliX interaction incorporates flexibility into the control of gene expression, permitting FlbD to receive information regarding both the progression of the cell cycle and flagellar assembly. Previous genetic experiments have indicated that FliX does not repress FlbD activity by influencing its phosphorylation state (Muir and Gober, 2002). The formation of a PspA–PspF or NifL–NifA complex results in an inhibition of ATPase activity, and an accompanying inability to activate transcription (Elderkin et al., 2002; Martinez-Argudo et al., 2004). We speculate that FliX, with analogy to PspA and NifL, also influences FlbD-catalysed ATPase activity. Because, in contrast to these two proteins, FliX also activates FlbD, it will be interesting to determine whether this property will also be attributed to a alteration in ATPase activity.
How does FliX sense the status of flagellar assembly and transduce this information to FlbD? In enteric bacteria, a well-studied flagellar assembly checkpoint couples the completion of a hook-basal body structure to the transcription of genes encoding flagellins and the chemotaxis apparatus. In this case, an alternative sigma factor, σ28, that is required for flagellin gene transcription, is inhibited by the FlgM anti-sigma factor. Completion of the hook-basal body structure results in the secretion of FlgM out of the cell, thus relieving the inhibition of late flagellar gene transcription (Hughes et al., 1993; Kutsukake, 1994). A distinctly different mechanism operates in Caulobacter because FliX, unlike FlgM, is not exported outside the cell and, additionally, has both a positive and a negative regulatory role. We imagine two possible ways by which flagellar assembly could influence FliX activity. First, FliX could be covalently modified when the assembly of a class II-encoded structure is completed, and thus be switched to an active form. In a second scenario, FliX may interact with a secreted substrate of the flagellum-specific type III secretory system (TTSS), or with a TTSS chaperone, as a way of monitoring the status of flagellum assembly. While there is currently no precedent for the first mechanism, there is evidence that TTSS chaperones in conjunction with their bound, secreted substrates have a role in regulating flagellar gene expression. In Salmonella typhimurium, FlgN, the chaperone for the hook-associated proteins (FlgK and FlgL) (Fraser et al., 1999), is a positive regulator of flgM translation (Karlinsey et al., 2000). The positive effect of FlgN on FlgM translation can be reversed by an increase in the intracellular concentration of FlgK and FlgL (Aldridge et al., 2003). This observation suggests that the binding of TTSS substrates to their cognate chaperones can be used to monitor the state of flagellar assembly and thereby influence gene expression (Aldridge et al., 2003). Analogously, but probably utilizing a different mechanism, the filament cap chaperone (Fraser et al., 1999), FliT, can influence transcription of middle (class II) flagellar promoters in Salmonella (Kutsukake et al., 1999). Interestingly, Caulobacter does not contain homologues of these enteric bacterial chaperones. It is also not apparent whether or not FliX can function as a chaperone for TTSS substrates because members of this class of protein share little sequence homology, although certain attributes are generally conserved. These include low molecular weight (less than 15 kDa), a relatively acidic isoelectric point and a high degree of coiled-coil regions, which are presumed to be involved with substrate interactions (reviewed in Parsot et al., 2003). While FliX does not have an acidic isoelectric point, it does have two highly conserved regions predicted to form coiled-coils (data not shown). Intriguingly, the fliX1 frameshift mutation is located at the end of the second of these highly conserved, coiled-coil sequences in the carboxyl-terminus (see Fig. 9B). Future studies will be required to determine whether this region is important for interaction with FlbD, or possibly with flagellar structural proteins.
Bacterial strains, plasmids and growth conditions
The bacterial strains and plasmids used in this work are listed in Table 1. In order to construct a fusion of FliX with the wild-type LexA(1-87) DNA-binding domain, a 448 bp AgeI–XhoI [polymerase chain reaction (PCR) generated]fliX fragment was cloned in frame behind the valine residue following the IPTG-inducible (i.e. lacUV5) LexA(1-87) DNA-binding domain of pMS604 (Dimitrova et al., 1998) in which the Fos zipper coding sequence has been removed (this work; see Porte et al., 1995 for sequence of fusion junction). In order to construct a fusion of FlbD with the mutant LexA(1-87)408 DNA-binding domain, a 1.38 kb XhoI–BamHI (PCR generated) flbD fragment was cloned in frame behind the six residues (SIARLE) following the IPTG-inducible (REG9 mutant lacUV5 promoter) LexA(1-87)408 DNA-binding domain of pDP804 in which all but the first five residues (IARLE) of the Jun zipper coding sequence has been removed (this work; see Porte et al., 1995 for sequence of fusion junction). The Caulobacter transcriptional reporters used, pfliF/lacZ/290 (Wingrove and Gober, 1994), pfliK/lacZ/290 (formerly pflbG/lacZ/290; Gober and Shapiro, 1992), and pfljL/lacZ/NPT228 (Muir et al., 2001), have been described previously.
Table 1. Bacterial strains and plasmids used in this study.
448 bp AgeI–XhoI (PCR generated) fliX fragment cloned in frame with LexA(1-87)D NA-binding domain of pMS604
1.38 kb XhoI–BamHI (PCR generated) flbD fragment cloned in frame with LexA(1-87)408 DNA-binding domain of pDP804
DNA manipulations were performed essentially as described (Ausubel et al., 1989). Plasmids were introduced into Caulobacter by bacterial conjugation using helper E. coli strain S17-1 (Simon et al., 1983). Caulobacter strains were grown at 31°C in peptone-yeast extract (PYE) (Poindexter, 1964) or M2 minimal media (Johnson and Ely, 1977) either alone or supplemented with one or more of the following: chloramphenicol (2.5 µg ml−1), tetracycline (2.0 µg ml−1), naladixic acid (20 µg ml−1) and rifampicin (1.0 µg ml−1). PYE motility plates contained 0.3% agar. The E. coli strains were grown at 37°C in Luria–Bertani (LB) medium (Miller, 1972) either alone or supplemented with one or more of the following: chloramphenicol (30 µg ml−1), kanamycin (50 µg ml−1), tetracycline (12.5 µg ml−1), ampicillin (100 µg ml−1), and IPTG (0.001–1.0 mM).
Flagellar gene expression
Synchronizable Caulobacter strains harbouring the fliK-lacZ transcriptional reporter fusion in addition to either pX10 or pfliX1 were used for the cell cycle gene expression experiments that were performed essentially as described by Wingrove and Gober (1994) and with the modifications detailed by Muir and Gober (2002). The cell cycle experiment carried out in the presence and absence of rifampicin with the synchronizable JG1181 (flbD-1204) strain harbouring the fliK-lacZ reporter was adapted from Gober et al. (1991) and was performed essentially as described by Muir and Gober (2002), varying only by the addition of 1.0 µg ml−1 rifampicin at time equalling zero minutes. Expression of the flagellar gene transcriptional reporters in unsynchronized cultures was measured as previously described (Mangan et al., 1995). All quantitative measurements of β-galactosidase activity exhibited by unsynchronized cultures were determined in triplicate, on three separately grown cultures. Immunoblots were performed essentially as described by Towbin et al. (1979) and were analysed using anti-FliX (Muir and Gober, 2002), anti-FlbD (Muir and Gober, 2002), anti-flagellin (Anderson and Gober, 2000) and anti-MreB antibodies (Figge et al., 2004) anti-sera.
FliX homodimerization and FliX–FlbD interaction
Construction of the LexA(1-87) DNA-binding domain-FliX fusion and the LexA(1-87)408 mutant DNA-binding domain-FlbD fusion proteins used in the interaction experiments is described above. Homodimerization of FliX and direct interaction of FliX with FlbD were assayed using the LexA-based bacterial two-hybrid system described by Dimitrova et al. (1998). Briefly, homodimerization experiments were performed in the SU101 reporter strain harbouring the lacZ reporter gene controlled by the wild-type LexA operator sequence (op+/op+) encompassing two wild-type half-sites (CTGT). Homodimerization was assayed by the transcriptional repression of the reporter gene as a result from the binding of a LexA dimer. In this system, dimerization of the wild-type LexA(1-87) DNA-binding domains is directly dependent on the intermolecular interactions of the carboxyl-terminally positioned ‘bait’ protein (FliX), as the LexA(1-87) DNA-binding domain alone can not form dimers. Heterodimerization experiments were performed in the SU202 reporter strain harbouring the lacZ reporter gene controlled by the hybrid LexA operator sequence (op408/op+) encompassing one wild-type half-site (CTGT) and one mutated half-site (CCGT). Heterodimerization was measured by the transcriptional repression of the lacZ reporter gene and resulted from the binding of a LexA heterodimer consisting of one wild-type LexA(1-87) DNA-binding domain that recognizes the op+ half-site and one mutant LexA(1-87) 408 DNA-binding domain (triple mutation LexA408 corresponding to P40A, N41S, A42S) that recognizes the op408 half-site. Homodimers of the wild-type and mutant LexA(1-87) DNA-binding domains formed through homo-association of the carboxyl-terminally positioned ‘bait’ proteins do not bind the hybrid operator with much affinity and are unable to significantly repress transcription of the lacZ reporter gene (Dimitrova et al., 1998). In this system, as in that mentioned previously, neither the wild-type LexA(1-87) nor the mutant LexA(1-87)408 DNA-binding domains alone have the capacity to form stable complexes with each other or themselves.
β-Galactosidase reporter activity was measured for the plasmid-containing SU101 and SU202 strains essentially as described by Dimitrova et al. (1998). Cells were grown over night in the presence or absence of 0.001, 0.01, 0.1 and 1.0 mM IPTG. The overnight cultures were diluted 1:100 into fresh media containing an identical IPTG concentration to that with which they were grown in overnight and let to grow at 37°C to an ODλ600nm = 0.6–0.8. Cells were then measured for β-galactosidase activity in triplicate, on three separately grown cultures.
We thank R. Dutton, J. England, M. Llewellyn and Z. Xu for helpful discussions and critical reading of this manuscript. R.E.M. was supported by a USPHS Predoctoral Fellowship GM07104. This work was supported by Public Health Service Grant GM48417 from the National Institutes of Health.