FlbT, the post-transcriptional regulator of flagellin synthesis in Caulobacter crescentus, interacts with the 5′ untranslated region of flagellin mRNA

Authors


Abstract

Flagellar gene expression‘ in Caulobacter crescentus is regulated by a complex trans-acting hierarchy, in which the assembly of early structural proteins is required for the expression of later structural proteins. The flagellins that comprise the filament are regulated at both the transcriptional and the post-transcriptional levels. Post-transcriptional regulation is sensitive to the assembly of the flagellar basal body and hook structures. In mutant strains lacking these structures, flagellin genes are transcribed, but not translated. Mutations in the flagellar regulatory gene, flbT, restore flagellin translation in the absence of flagellar assembly. In this report, we investigate the mechanism of FlbT-mediated post-transcriptional regulation. We show that FlbT is associated with the 5′ untranslated region (UTR) of fljK (25 kDa flagellin) mRNA and that this association requires a predicted loop structure in the transcript. Mutations within this loop abolished FlbT association and resulted in increased mRNA stability, indicating that FlbT promotes the degradation of flagellin mRNA by associating with the 5′ UTR. We also assayed the effects on gene expression using mutant transcripts fused to lacZ. Interestingly, the mutant transcript that failed to associate with FlbT in vitro was still repressed in mutants defective in flagellum assembly, suggesting that other factors in addition to FlbT couple assembly to translation.

Introduction

The Gram-negative bacterium Caulobacter crescentus undergoes an asymmetric cell division, resulting in the formation of a non-motile stalked cell and a motile swarmer cell possessing a single polar flagellum. Flagellar biogenesis, the most extensively studied aspect of this simple programme of differentiation, occurs in the predivisional cell and results in the assembly of a complex flagellum structure at one pole of the cell (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000). Genetic analysis has shown that flagellar biogenesis requires over 50 genes and is regulated by a trans-acting hierarchy that couples early stages of flagellar assembly to gene expression (Fig. 1).

Figure 1.

The flagellar regulatory hierarchy in Caulobacter crescentus (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000). A schematic diagram of the assembled structures is depicted above the gene designations. In response to a cell cycle cue, the sensor histidine kinase encoded by cckA phosphorylates the response regulator transcription factor (Jacobs et al., 1999), CtrA, which, in turn, activates the transcription of class II genes (Quon et al., 1996). Class II genes encode components of the type III flagellar secretory system, structural proteins of the MS-ring and switch and transcriptional activators of late flagellar genes. The expression and successful assembly of the class II flagellar structure is required for the expression of genes that encode the class III structure, which consists of the rods and rings of the basal body and the hook. Class III gene expression is activated by the transcription factor, FlbD, and requires the σ54 subunit of RNA polymerase, integration host factor (IHF) (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000) and the trans-acting factor, FlbE (Wingrove and Gober, 1997). A mutation in an unidentified gene, bfa, permits the expression of class III genes in the absence of early flagellar structures and is thus considered to encode a negative regulator that couples the assembly of class II gene products to the expression of class III genes (Mangan et al., 1995). Similarly, the expression of class IV genes (flagellins) is post-transcriptionally repressed by the flbT gene product in the absence of a completed hook structure.

The earliest flagellar structural components expressed encode the MS-ring of the basal body, the flagellar switch and the flagellum-specific, type III secretion system (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000) (Fig. 1). Epistasis experiments have shown that a mutation in any of these class II genes prevents the transcription of class III genes that encode the outer rings of the basal body, the basal body rod structure and the flagellar hook and hook-associated proteins. From this type of experiment, it has been inferred that assembly of a class II-encoded flagellar structure is required for the transcription of class III genes.

The promoters of class III flagellar genes share conserved sequences, each containing binding sites for the general transcription factors integration host factor (IHF) and σ54-containing RNA polymerase (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000). Control of transcriptional activation is accomplished by regulating the phosphorylation state of the transcriptional activator, FlbD (Ramakrishnan and Newton, 1990; Wingrove et al., 1993; Wingrove and Gober, 1994; 1996; Benson et al., 1994a,b; Mullin et al., 1994; Wu et al., 1995). Mutations, termed bfa for bypass of flagellar assembly, were isolated that no longer require class II flagellar gene products for the transcription of middle or late flagellar genes (Mangan et al., 1995). The function of the bfa gene product is not currently known, but is hypothesized to encode a negative regulator of class III transcription that responds to the assembly of a class II-encoded flagellar structure. Interestingly, although the 27 kDa flagellin gene, fljL, is transcribed in the class II/bfa double mutants, flagellin protein is not produced. In addition, it has been demonstrated that both fljL::lacZ and fljK::lacZ protein fusions (25 kDa flagellin) are not expressed in either class II or class III flagellar mutants (Mangan et al., 1995; Anderson and Newton, 1997). These experiments indicate that the expression of flagellin is regulated by a post-transcriptional mechanism in response to the assembly of a class III flagellar structure.

Recently, we have shown that the stability of flagellin mRNA is greatly reduced in strains containing mutations in class III flagellar genes (Mangan et al., 1999). Furthermore, we have found that mutations in flbT can restore flagellin expression and mRNA stability in class III flagellar mutants. Therefore, FlbT regulates a checkpoint that couples the assembly of class III flagellar structures to flagellin expression. As fljK is expressed almost throughout the cell cycle in an flbT mutant strain, this checkpoint is essential for the correct temporal regulation of flagellin gene expression (Mangan et al., 1999). In this report, we examine the mechanism of post-transcriptional regulation of fljK by FlbT. We show that FlbT is associated with the 5′ untranslated region (UTR) of fljK using cell extracts and in vitro-transcribed fljK in electromobility shift assays. This binding is abolished by mutations in the 5′ UTR at positions 27–29 nucleotides relative to the start of transcription. Additionally, we demonstrate that a mutant transcript that is not associated with FlbT has an increased half-life, indicating that FlbT regulates flagellin expression through binding to the 5′ UTR. We also test the effect of site-directed mutations in the 5′ UTR of the fljK transcript and on the regulation of fljK as measured by reporter fusions to lacZ.

Results

FlbT is associated with the 5′ UTR of fljK

Translation of fljK is restored in an flgE mutant background when combined with a mutation in flbT (Mangan et al., 1999). The same mutation in flbT restores translation of fljK in all class II and class III flagellar mutant backgrounds tested, including mutations in genes encoding components of the flagellar export machinery and structural proteins of the basal body and hook (Mangan et al., 1999). Deletion of nucleotides between +24 and +38 relative to the start of transcription within the 5′ UTR were shown to be required for post-transcriptional regulation (Anderson and Newton, 1997). As these data indicate that FlbT functions as a translational repressor, it is likely that FlbT exerts its regulatory effect on gene expression by interacting, either directly or indirectly, with the 5′ UTR. In an effort to demonstrate an interaction between the fljK transcript and FlbT, we performed gel shift assays using in vitro-transcribed message and either purified protein or cell extracts (Fig. 2). To preserve the 5′ end, and thus any secondary structure of the RNA, a T7 promoter was introduced by site-directed mutagenesis just upstream of the previously determined transcription start site (Minnich and Newton, 1987). This configuration changes the 5′ end by several nucleotides from CCA to GAGA; however, these bases are predicted to be unpaired (see Fig. 3) and therefore should have no effect on the secondary structure. Plasmids carrying the T7 promoter were used to make labelled run-off transcripts of 132 nucleotides for use as substrate in RNA-binding assays.

Figure 2.

RNA-binding assays using fljK transcript and purified FlbT or cell extracts. Transcripts (132 nucleotides) were generated in vitro and were labelled internally with [α-32P]-UTP using T7 RNA polymerase. Lane 1, wild-type fljK transcript and C. crescentus NA1000 S30 extract; lane 2, wild-type fljK transcript and C. crescentus SC276 (flbT650) S30 extract; lane 3, wild-type fljK transcript and C. crescentus NA1000 S30 extract supplemented with anti-FlbT antibody; lane 4, wild-type fljK transcript and C. crescentus SC276 (flbT650) S30 extract supplemented with anti-FlbT antibody; lane 5, mutant fljK201 transcript and C. crescentus NA1000 S30 extract; lane 6, mutant fljK201 transcript and C. crescentus SC276 (flbT650) S30 extract; lane 7, mutant fljK201 transcript and C. crescentus NA1000 S30 extract supplemented with anti-FlbT antibody; lane 8, mutant fljK201 transcript and C. crescentus SC276 (flbT650) S30 extract supplemented with anti-FlbT antibody. Unbound probe and protein–RNA complexes are indicated by arrows.

Figure 3.

Two secondary structures for the fljK transcript predicted using the mfold program (Zuker, 1989). Each structure shows 113 nucleotides (nt) from the start of transcription, which includes 63 nt of 5′ untranslated sequence and 17 codons. The position of site-directed mutants and sequence changes described in this paper are shown for each structure. The structure depicted in (A) is predicted to have a free energy of folding equalling −38.3 kcal mol−1. This structure is predicted to support translation as a consequence of little basepairing near the ribosome-binding sequence (RBS) (AGGA; in bold) allowing for efficient basepairing with 16S ribosomal RNA. The structure depicted in (B) is predicted to have a free energy of folding equalling −36.6 kcal mol−1. Owing to extensive secondary structure in the area of the RBS and start codon (AUG; in bold), translation of structure B is predicted to be inefficient.

When purified FlbT was used in a gel shift assay with in vitro-transcribed RNA, no shift in mobility was observed at any of the RNA–protein concentration ratios tested (data not shown). In contrast, when labelled probe was used in binding reactions using S30 extracts (Zubay, 1973) prepared from either wild-type or flbT strains, a mobility shift was observed (Fig. 2, lanes 1 and 2). The mobility of the RNA was shifted using extracts of both strains tested, although the extract lacking FlbT produced a different band pattern from extracts containing FlbT. Samples containing wild-type extract produced one predominant shifted species as well as two minor species (labelled bands 2, 3 and 4 in Fig. 2). When extracts were prepared from the flbT mutant strain, the abundance of the major species (labelled band 2 in Fig. 2) was remarkably reduced. Instead, there was an increase in the abundance of free probe as well as in the two slower migrating species (labelled bands 3 and 4 in Fig. 2). This suggests that FlbT is associated with the RNA, although the interaction may be indirect or require an additional protein to form a stable complex. The addition of purified histidine-tagged FlbT (FlbTH6) to extracts did not alter the pattern of shifted bands in either wild-type or flbT extracts (data not shown). To confirm that FlbT was present in the shifted complex, and thus responsible for the shift patterns in extracts prepared from different strains, RNA-binding assays were repeated in extracts that were supplemented with anti-FlbT antibody (Fig. 2, lanes 3 and 4). The addition of anti-FlbT antibody produced an additional shifted band (Fig. 2, lane 3) that was not present in extracts prepared from an flbT mutant strain (SC276) (Fig. 2, lane 4). These data suggest that FlbT binds to the fljK transcript and that binding is probably dependent on another protein present in cell extracts.

In order to identify the RNA sequences responsible for the association of FlbT with the fljK transcript, we performed site-directed mutagenesis on the 5′ UTR. Analysis of the fljK transcript with the mfold program (Zuker, 1989) revealed two distinct predicted secondary structures of nearly equal free energies (Fig. 3). Although both structures share a common stem–loop at the 5′ end as well as a common stem–loop structure beginning 26 nucleotides after the start codon, the structures surrounding the ribosome binding site (RBS) and start codon are different. In one predicted structure (Fig. 3A), the RBS and start codon exist in two different stem–loops (ΔG = −1.9 and −10.6 kcal mol−1 respectively) and may be more accessible to the ribosome than in the other predicted structural conformation (Fig. 3B), in which they lie within a larger stem–loop (ΔG = −22.9 kcal mol−1). One possible effect of FlbT could be to shift the equilibrium between the formation of these two structures.

Mutations were created in sequences predicted to form loops or bulges [fljK201 (AAA27–29CCC), fljK203 (A22U), fljK205 (AG22–23GA) and fljK206 (AG22–23GU)] and are therefore designed to interfere with the binding of potential proteins without significantly altering the free energy of either structure. The exception being mutant fljK202 (ΔAAA27–29), which contains a deletion in one of the loops and is predicted to have a direct effect on the free energy of the structures, favouring the predicted untranslated structure in Fig. 3B. To test the effect of these mutations in the 5′ UTR on interaction with FlbT, the RNA-binding reaction was performed using the mutant transcripts as substrates. As with wild-type transcript, binding reactions were performed in extracts and extracts supplemented with anti-FlbT antibody. Mutant transcripts fljK203, fljK205 and fljK206 exhibited the same shifted species as the wild-type transcript and possessed a supershifted species when anti-FlbT antibody was included in the extracts (data not shown). The fljK201 mutant transcript, however, differed from the wild type when used in the RNA-binding assay (Fig. 2, lanes 5–8). Only one new shifted band was demonstrable for this mutant transcript (labelled band 1 in Fig. 2, lanes 5–8). Interestingly, the fljK201 transcript lacked all the bands that were present when wild-type transcript was used as a substrate. In addition, supplementing the binding reactions with anti-FlbT antibody had no effect on the mobility of the shifted bands using cell extracts prepared from either wild-type or an flbT mutant strain (SC276) (Fig. 2, lanes 7 and 8). These experiments demonstrate that the mutant fljK201 transcript is either weakly associated or not associated at all with FlbT in cell extracts. Furthermore, as these binding assays also lacked other shifted species that were present using wild-type transcript, it is likely that the mutations in fljK201 also abolish the binding of another protein(s) in addition to disrupting FlbT binding. One possibility is that the sequences within the loop structure that are mutated in fljK201 are required for the binding of FlbT and these other proteins. Interestingly, in this regard, mutant transcript fljK202, which possesses a deletion of this sequence, was also not associated with FlbT or these other proteins (data not shown).

Assay of fljK mRNA stability

Previous experiments have shown that fljK mRNA exhibits a dramatically increased stability in flbT mutant strains (Mangan et al., 1999). This is consistent with the proposal that flbT is a negative regulator of fljK expression. We therefore wished to determine whether the mutant fljK transcript that was not associated with FlbT in vitro also possessed an increased stability. To accomplish this, we performed mRNA half-life experiments using wild-type, fljK201 (AAA27–29CCC) or fljK206 (AG22,23GU) message fused as a protein fusion to lacZ RNA. Half-lives were measured in both a wild-type and a flgE/flbT double mutant background (Fig. 4). As observed previously, wild-type fljK mRNA showed an increase in stability in an flbT mutant strain. In the flgE/flbT mutant, the half-life for a wild-type fusion increased from approximately 1 to 5 min. To determine whether the FlbT-dependent destabilization of message occurred as a result of interaction with FlbT, we measured the half-life of the fljK201 transcript fused to lacZ. In wild-type cells, this transcript exhibited an ≈ 2.5-fold increase in stability (t1/2 = 2.5 min) (Fig. 4), consistent with the idea that association with FlbT is, in part, responsible for regulating the stability of fljK mRNA. Although this mutant transcript either cannot be or is weakly bound by FlbT, like wild-type transcript, it possesses an increased half-life in the flgE/flbT double mutant (t1/2 > 10 min). fljK206, which is associated with FlbT in cell extracts, had the greatest increase in stability, showing no visible decay in either strain background even after 10 min (Fig. 4). As mutations at positions 22 and 23 are predicted to have little effect on structure, and the RNA-binding assays for the wild-type fusion and fljK206 show no visible difference, these base changes may interfere with an mRNA processing pathway, resulting in an unusually stable message independent of strain background.

Figure 4.

Effects of mutations in the 5′ UTR of fljK on mRNA half-lives in both wild-type and flgE/flbT double mutant cells. Rifampicin was added to a final concentration of 200 µg ml−1 to exponentially growing cultures containing fljK, fljK201 or fljK206 fused to lacZ. At times 0.5, 1, 2.5 and 10 min, 10 ml aliquots were removed and frozen rapidly. Total RNA was isolated, and quantitative primer extension was performed to measure levels of fljK–lacZ fusion mRNA. The 32P-labelled primer-extended product was subjected to electrophoresis in an 8% denaturing gel. Radioactive bands were quantified using a phosphorimager and imagequant.

Effect of mutations in the fljK 5′ UTR on post-transcriptional regulation

We next tested the effect of mutations in the 5′ UTR of the fljK transcript on post-transcriptional regulation by flagellar assembly. To accomplish this, we constructed both transcriptional and translational fusions to lacZ. Each fusion contained the 505 bp restriction fragment that includes 376 bases upstream of the start of transcription, the 63 nucleotide 5′ UTR and 66 nucleotides of coding sequence. This fragment contains all the necessary elements for cell cycle and transcriptional control by the flagellar regulatory hierarchy (Wingrove et al., 1993; Mangan et al., 1999). β-Galactosidase assays were then performed on both transcriptional and translational fusions in wild-type, flgE (hook), flbT or flgE/flbT double mutant backgrounds (Table 1). As expected, protein levels decreased by 89% in an flgE mutant (a Tn5 in the class III structural gene that codes for the hook protein). When a mutation in flbT was introduced either alone or in combination with an flgE mutation, protein expression increased by 143% and 461% relative to wild type even though transcription levels fell to just 6% and 14% of wild type respectively.

Table 1. β-Galactosidase assays of transcriptional and translational fusions.
FusionStrain
background
Protein
fusion
Operon
fusion
Translationa
index
  • a

    . Translation index represents the units of β-galactosidase generated from a protein fusion divided by the units generated from an operon fusion (see text).

fljK::lacZ LS10797744090.21
flgE 10039180.03
flgE/flbT 52006228.4
flbT 22552907.8
fljK201 LS107249941940.60
AAA27–29CCC flgE 10538040.03
flgE/flbT 30011300.27
flbT 14808391.8
fljK202 LS1072523670.01
ΔAAA27–29 flgE 1018210.01
flgE/flbT 2417630.01
flbT 4840890.01
fljK203 LS10776518010.42
A22U flgE 3528380.01
flgE/flbT 14095782.4
flbT 16 50023171.4
fljK205 LS10792538750.23
AG22,23GA flgE 6513620.05
flgE/flbT 2845440.52
flbT 9821895.2
fljK206 LS10768045290.15
AG22,23GU flgE 7033110.02
flgE/flbT 3145770.54
flbT 10602514.2

For each fusion, the level of β-galactosidase activity for the translational fusion is divided by the level of transcriptional activity to provide an index of the efficiency of translation for a given transcript in each of the strain backgrounds (Table 1). With the wild-type fusion, the efficiency of translation decreases when measured in a class III mutant background (Table 1), as would be expected if the gene is regulated post-transcriptionally. The efficiency of translation is greatly increased when the mutant flbT650 allele is introduced either alone or in combination with a class III mutation (Table 1). These data are consistent with the idea that FlbT functions as a post-transcriptional regulator.

Surprisingly, all the mutant transcripts were regulated by flagellar assembly, exhibiting a marked decrease in protein expression in the flgE mutant strain (compare LS107 with flgE for all mutant transcripts) (Table 1). Interestingly, these mutant transcripts varied in their response to an absence of flbT. The most notable effect in this regard was mutation fljK201, which changes three unpaired adenines to cytosines in a predicted loop structure. In this case, the effect of a loss of flbT is greatly reduced (Table 1). The presence of the flbT650 allele causes a three- to ninefold increase in the translation index in a fusion containing fljk201 compared with a 37-fold and 335-fold effect on the wild-type fusion. Mutation fljK202, which deletes the same three adenines and would be predicted to favour strongly the structure shown in Fig. 2B, shows a severe decrease in translation in all strain backgrounds, supporting the idea that the critical determinant of translation competence is accessibility of the ribosome to the RBS. Mutation fljK203, which changes an unpaired adenine to a uracil, is still subject to post-transcriptional regulation by FlbT. Mutations fljK205 and fljK206, which are both designed to move the bulge up by 1 bp while either maintaining the unpaired adenine or changing it to a uracil, had little effect on repression. These data suggest that FlbT may exert its regulatory effect by interacting with the 5′ UTR of the fljK transcript. However, as all mutant transcripts were still repressed in the absence of a flagellar hook structure, it is likely that FlbT is not the sole regulator of translation in response to flagellar assembly (see Discussion).

Regulation of FlbT expression

Flagellin expression in C. crescentus is under cell cycle control. The transcription of fljK peaks in predivisional cells at a time when flagellin monomers are required (Wingrove et al., 1993). Cell cycle transcription is accomplished through the cell cycle-dependent phosphorylation of the transcription factor FlbD. FlbT also influences the temporal expression of flagellin genes. In flbT mutants, flagellin–lacZ protein fusions continue to be translated after cell division (Mangan et al., 1999). This effect is presumably a consequence of the remarkable stability of fljK mRNA in flbT mutant cells. This experimental observation has prompted the hypothesis that FlbT functions to repress the translation of flagellin mRNA late in the cell cycle. How does FlbT regulate translation in a cell cycle-dependent manner? We envisage two possible mechanisms. First, the cellular levels of FlbT may be temporally regulated or, alternatively, the activity of FlbT may be regulated in a cell cycle-dependent fashion. To distinguish between these two possibilities, we performed immunoblots with anti-FlbT antibody on extracts prepared from cells at different stages of the cell cycle. The levels of FlbT did not change during the course of the cell cycle, indicating that FlbT activity must be temporally regulated (data not shown). To test this idea, we also assayed the temporal levels of flbT in an flgE mutant. In this case, also, the level of FlbT remained constant throughout the cell cycle. In addition, the flgE mutant strain contained levels of FlbT that were similar to amounts found in wild-type cells (data not shown). This result indicates that flagellin assembly regulates FlbT activity.

Discussion

In this paper, we have investigated the mechanism of FlbT-mediated post-transcriptional regulation of flagellin genes in C. crescentus. The temporal regulation of flagellar gene expression in C. crescentus is influenced not only by progression through the cell cycle, but also by flagellum assembly (reviewed by Brun et al., 1994; Gober and Marques, 1995; Wu and Newton, 1997; Gober and England, 2000). The assembly of early flagellar structures (MS-ring, secretory system, switch) is required for the transcription of genes encoding middle structures (basal body rods, outer rings and hook), which, in turn, are required for the expression of late flagellar structures (filament). The coupling of middle flagellar gene expression to the synthesis of flagellins was previously found to be regulated by a post-transcriptional mechanism. In class III flagellar mutants, the genes encoding flagellins are transcribed, but flagellin protein is not expressed (Mangan et al., 1995; 1999; Anderson and Newton, 1997). Mutations in flbT restore flagellin expression in class III flagellar mutants, thus indicating that FlbT has a critical role in coupling the assembly of class III structures to flagellin expression.

Previous experiments have shown that a deletion of the 5′ UTR of the fljK transcript abolished post-transcriptional regulation (Anderson and Newton, 1997). We have demonstrated that FlbT in cell extracts is associated with the 5′ UTR, a result consistent with its proposed role as a post-transcriptional regulator of gene expression. Association was not observed using purified FlbT and transcript, indicating that the binding of FlbT to the fljK transcript is probably dependent on another protein(s). Using the mfold program (Zuker, 1989) for RNA secondary structure prediction revealed two possible, equally energetically favourable conformations for the fljK transcript. Here, we have analysed the effect of mutations in predicted loop or bulge structures in the 5′ UTR on the association of FlbT and other proteins present in cell extracts. Two mutant transcripts, fljK201 (AAA22–24CCC) and fljK202 (ΔAAA22–24) were not associated with FlbT in cell extracts.

We also tested the effect of these mutations on post-transcriptional regulation. Reporter protein fusions containing a wild-type fljK 5′ UTR are typically not expressed in strains bearing class III flagellar mutations. The introduction of a mutant flbT into these strains restores protein expression. One mutant fusion, fljK201 (AAA22–24CCC), was not subject to this regulation. This fusion was expressed in wild-type cells at levels greater than wild-type fusions and was repressed in a hook mutant. In contrast to a lacZ fusion containing a wild-type 5′ UTR, the expression of this mutant was not restored by introducing a mutant flbT allele into the hook mutant strain. As the RNA-binding experiments showed that FlbT did not associate with this mutant transcript in cell extracts, this provides a basis for the lack of an increase in expression in flbT mutants. As previous experiments, as well as those presented here, have shown that a mutation in flbT is sufficient completely to restore flagellin protein expression in class III flagellar mutants, this result represents an apparent paradox. If FlbT is the sole negative regulator of flagellin gene expression as postulated, how can a mutant transcript that is neither associated with nor responds to a lack of FlbT be repressed in a hook mutant? A possible explanation is that an additional factor must be bound to the fljK transcript, which promotes stability and/or translation in order for flagellin protein to be expressed, and the mutation in the fljK201 transcript disrupts the binding of that factor. The RNA binding experiments presented here support this idea. The mutation in fljK201 not only resulted in a loss of FlbT bound to the transcript, but also at least two other protein–RNA complexes (Fig. 2, bands 3 and 4). Interestingly, these same protein–RNA complexes were increased in abundance in flbT mutant extracts when binding was assayed using a wild-type transcript. This raises the possibility that an additional protein(s) requires some of the same sequences as FlbT to bind to the fljK 5′ UTR. Based on these results, we propose a model in which post-transcriptional regulation of the fljK message is accomplished through the interaction of at least two proteins with the message. One possibility is that FlbT competes with a positive factor for the same binding site within the fljK 5′ UTR (Fig. 5). We hypothesize that FlbT is bound to flagellin message when the nascent flagellum structure is not competent to assemble flagellin monomers. When hook assembly is completed, repression is relieved by the binding of this proposed positive factor to the fljK transcript.

Figure 5.

Model depicting the interaction of regulatory molecules with the fljK transcript. The top of the figure shows a schematic representing the stages of flagellar assembly. We propose that FlbT binds to the 5′ UTR of the fljK transcript and promotes degradation, possibly by stabilizing a structure that is susceptible to ribonuclease action. Based on the experiments presented here, we hypothesize that a positive factor, labelled X, which promotes translation, binds to some of the same RNA sequences as FlbT. We envisage that the progression of flagellar assembly dictates whether FlbT binds and promotes message degradation or the putative positive factor binds and promotes translation.

Although regulation of bacterial gene expression at the level of transcription has been studied exhaustively, many post-transcriptional regulatory mechanisms are not thoroughly understood. There are several possible points at which the bacterial cell can regulate protein synthesis. Rates of elongation have been shown to be affected through codon usage (Grosjean and Friers, 1982); RNA secondary structures (Sorensen et al., 1989) and bacterial polyadenylation have been implicated in regulating translation (reviewed by Sarkar, 1997). Experimental data have shown the major point of regulation in bacterial translation is at the step of initiation, particularly in ternary complex formation, in which either protein repressors or RNA secondary structures physically block association of the mRNA, initiator tRNA and 30S ribosomal subunit (reviewed by Gold, 1988; McCarthy and Gualerzi, 1990; Kozak, 1999). What is the mechanism of FlbT-mediated repression? Previous experiments have demonstrated that the stability of fljK mRNA is reduced in strains bearing a class III flagellar mutation (Anderson and Newton, 1997; Mangan et al., 1999). Introduction of a flbT mutation into these strains resulted in increased stability of message. These results indicated that FlbT either promoted the degradation of message directly or functioned as an inhibitor of translation, and this, in turn, led to a decreased message stability. The experimental results presented here favour the former mechanism. Although FlbT contains none of the sequence motifs commonly associated with RNA-binding proteins (reviewed by Burd and Dreyfuss, 1994; Arnez and Cavarelli, 1997), it is clearly associated with the 5′ UTR of fljK. The mutant fljK201 transcript that is not associated with FlbT has an increased stability in wild-type cells, as would be predicted if FlbT binding promoted message degradation. In an flgE/flbT double mutant, the half-life of the mutant transcript was also greater than that of a wild-type transcript in the same strain. In this case, however, the message was for the most part not translated. Therefore, for the fljK message, degradation and translation are separable processes. The mutation in fljK201 resulted in both an increased stability and a lack of translation in a hook mutant. This, again, is consistent with the idea that the fljK201 mutation disrupts an association with two proteins: FlbT, which promotes mRNA degradation; and an antagonizing positive factor, which promotes translation. We propose that these two opposing regulatory events are influenced by flagellum assembly (Fig. 5). The sequences required for FlbT as well as this other unknown protein(s) to interact with the fljK message centre 24 bases upstream of the Shine–Dalgarno sequence. We hypothesize that this positive factor functions to stabilize a secondary structure that permits translation of the 25 kDa flagellin.

Post-transcriptional regulation of flagellin synthesis in response to flagellum assembly in C. crescentus contrasts with regulatory mechanisms operating in enteric bacteria. In Salmonella typhimurium, flagellins are transcribed by a σ28-containing RNA polymerase (Komeda, 1986; Kutsukake et al., 1990). Co-ordination of class III flagellar expression (flagellin) with assembly of the basal body–hook structure is accomplished through the use of an anti-sigma factor encoded by flgM (Gillen and Hughes, 1991a,b; Onishi et al., 1992). Intracellular levels of FlgM remain high in the absence of flagellar assembly, preventing σ28 activity. Upon successful completion of assembly of the hook structure, FlgM is exported allowing transcription of flagellin (Hughes et al., 1993). By analogy, one possible mechanism for relieving FlbT-mediated negative regulation of flagellin expression could involve the export of FlbT when the basal body–hook structure has completed assembly. Alternatively, the activity of FlbT or perhaps an activator of translation could be modulated by either assembly or events during the cell cycle. We favour the second model for several reasons. First, intracellular levels of FlbT are constant throughout the cell cycle. Immunoblots using an FlbT antibody and synchronized cultures of both wild-type and hook mutant strains showed no variation in the level of FlbT. In addition, the level of FlbT is not increased in a hook mutant compared with a wild-type background. Attempts to identify FlbT in the supernatant of cultures by precipitation and immunoblotting have also failed to detect any secreted protein (data not shown).

What is the regulatory logic in controlling flagellin expression at the post-transcriptional level? The temporal transcription of flagellin genes is activated by the transcription factor, FlbD, whose activity is regulated by cell cycle-dependent phosphorylation that spans a period from the early predivisional cell to the time of cell division (Wingrove et al., 1993). In the late predivisional cell, FlbD activity and the transcription of flagellin genes is restricted to the swarmer compartment (Gober et al., 1991; Wingrove et al., 1993; Wingrove and Gober, 1997). The swarmer pole-specific transcription of flagellin genes serves to supply the nascent swarmer cell with flagellin mRNA, which persists and continues to be translated after cell division (Milhausen and Agabian, 1983). This supports continued flagellar filament assembly in the nascent swarmer cell well after the transcription of flagellar genes has ceased. The level of flagellin mRNA declines rapidly upon differentiation into a stalked cell. FlbT-mediated post-transcriptional regulation is responsible for this decline in flagellin mRNA, as flbT mutant cells continue to express flagellin in the stalked cell phase (Mangan et al., 1999). Thus, co-ordinately regulated transcription and mRNA degradation fine tunes the temporal and spatial expression of flagellin genes.

Experimental procedures

Bacterial strains and growth conditions

The bacterial strains and plasmids used in these experiments are listed in Table 2. C. crescentus LS107, a synchronizable, ampicillin-sensitive derivative of CB15 (Stephens et al., 1997), was used as a wild-type strain. AE8006 is a mutant strain with a Tn5-VB32 (Bellofatto et al., 1984) insertion in flgE (Champer et al., 1987). Translational and transcriptional fusions were made with flgE by transduction from AE8006 and SC1039, respectively, using φCR30 phage (Ely and Johnson, 1977). All C. crescentus strains were grown with shaking at 31°C in PYE medium (Poindexter, 1964). Escherichia coli strains were grown at 37°C in Luria–Bertani (LB) medium (Miller, 1972). Antibiotics were added to a final concentration of 2 µg ml−1 tetracycline, 50 µg ml−1 kanamycin and 20 µg ml−1 naladixic acid for C. crescentus and 12.5 µg ml−1 tetracycline, 100 µg ml−1 ampicillin, 50 µg ml−1 kanamycin and 30 µg ml−1 chloramphenicol for E. coli.

Table 2. Strains and plasmids used in this study.
Strain or plasmidGenotypeSource
Plasmids
 pJBZ282 lacZ protein fusion vectorM. R. K. Alley
 placZ/290 lacZ transcription fusion vector Wingrove et al. (1993)
 pBluescript KS+ Stratagene
 pET-15bN-terminal His-tag cloning/expression vectorNovagen
 pPA200pET15b-FlbTH6 
 pPA201pBluescript KS+ with 505 bp PstI–EcoRI fragment containing
 the fljK promoter and 23 codons
This work
E. coli
 S17-1Rp4-2, Tc::Mu Km::Tn7 Simon et al. (1983)
 BL21(DE3)FompT hsdSB (rB mB; an E. coli B strain)with
 λ prophage carrying T7 RNA polymerase gene
Studier et al. (1990)
C. crescentus
 LS107 syn-100, amps derivative of NA1000 Stephens et al. (1997)
 AE8006 flgE806::Tn5-VB32 Champer et al. (1987)
 SC276 flbT650 Johnson and Ely (1979)
 SC1039 flgE::Tn5B. Ely
 JG866LS107 fljK203-JBZ282This work
 JG868LS107 fljK205-JBZ282This work
 JG869LS107 fljK206-JBZ282This work
 JG898SC276 fljK201-JBZ282This work
 JG900SC276 fljK206-JBZ282This work
 JG901SC276 fljK205-JBZ282This work
 JG904LS107 fljK201-lacZ/290This work
 JG906LS107 fljK202-lacZ/290This work
 JG908LS107 fljK203-lacZ/290This work
 JG910LS107 fljK205-lacZ/290This work
 JG911LS107 fljK206-lacZ/290This work
 JG912LS107 fljK-lacZ/290This work
 JG913SC276 fljK201-lacZ/290This work
 JG914SC276 fljK202-lacZ/290This work
 JG915SC276 fljK203-lacZ/290This work
 JG918SC276 fljK205-lacZ/290This work
 JG919SC276 fljK206-lacZ/290This work
 JG920SC276 fljK-lacZ/290This work
 JG939LS107 fljK-JBZ282This work
 JG940SC276 fljK-JBZ282This work
 JG952JG940 flgE806::Tn5-VB32This work
 JG976JG901 flgE806::Tn5-VB32This work
 JG977JG898 flgE806::Tn5-VB32This work
 JG978JG902 flgE806::Tn5-VB32This work
 JG979JG901 flgE806::Tn5-VB32This work
 JG982JG867 flgE806::Tn5-VB32This work
 JG983JG866 flgE806::Tn5-VB32This work
 JG984JG869 flgE806::Tn5-VB32This work
 JG985JG868 flgE806::Tn5-VB32This work
 JG986JG987 flgE806::Tn5-VB32This work
 JG987CB15NfljK201-JBZ282This work
 JG988SC276 fljK203-JBZ282This work
 JG989JG988 flgE806::Tn5-VB32This work
 JG990SC276 fljK202-JBZ282This work
 JG991JG990 flgE806::Tn5-VB32This work
 JG992CB15N fljK202-JBZ282This work
 JG993JG992 flgE806::Tn5-VB32This work
 JG1055SC276 flgE::Tn5This work
 JG1057SC1039 fljK-lacZ/290This work
 JG1058SC1039 fljK201-lacZ/290This work
 JG1059SC1039 fljK202-lacZ/290This work
 JG1060SC1039 fljK203-lacZ/290This work
 JG1062SC1039 fljK205-lacZ/290This work
 JG1063SC1039 fljK206-lacZ/290This work
 JG1064JG1055 fljK-lacZ/290This work
 JG1065JG1055 fljK201-lacZ/290This work
 JG1066JG1055 fljK202-lacZ/290This work
 JG1067JG1055 fljK203-lacZ/290This work
 JG1069JG1055 fljK205-lacZ/290This work
 JG1070JG1055 fljK206-lacZ/290.This work

Construction of lacZ reporter fusions

Site-directed mutagenesis was performed on the fljK 5′ UTR using mutagenic oligonucleotides (Kunkel and Roberts, 1987). All mutations were confirmed by DNA sequencing. A 505 bp PstI–EcoRI fragment of wild-type and mutant versions of fljK in Bluescript was subcloned as a 529 bp BamHIHindIII fragment into either pJBZ282 to create lacZ reporter protein fusions or placZ/290 to create lacZ reporter operon fusions. For protein fusion vectors, plasmids were first transformed into S17-1 carrying the conjugation helper plasmid pLVC9 (Simon et al., 1983), then mated into LS107 and SC276 and plated on selective media. Transconjugants were transduced to tetracycline resistance with φCR30 lysates grown on AE8006. Transcriptional fusions were first transformed into S17-1 and then mated into LS107 and SC276. C. crescentus cells harbouring the plasmid were plated on selective media. Transconjugants were transduced to kanamycin resistance with φCR30 lysates grown on SC1039. Cultures were grown overnight in PYE supplemented with the appropriate antibiotics and assayed as described previously (Miller, 1972).

Protein purification and antibody production

To overexpress histidine-tagged FlbT (FlbTH6), flbT was amplified from C. crescentus DNA by polymerase chain reaction (PCR) and cloned as a BamHI fragment into pET-15b. A cell extract prepared from BL21(DE3) carrying this plasmid was applied to a nickel agarose (Qiagen) column, washed with NSLB (0.5 M NaCl, 20 mM Tris-HCl, pH 7.5, 5 mM imidazole) and eluted in an imidazole gradient from 0 mM to 1000 mM in NSLB. Protein was then dialysed against 10 mM HEPES, pH 7.9, 250 mM NaCl and 1 mM EDTA. Dialysed protein was further purified to apparent homogeneity using a Bio-Rad preparative electrophoresis cell according to the manufacturer's instructions. Polyclonal antibodies were prepared by a commercial source (Cocalico Biologicals).

RNA-binding assays

S30 extracts were made by harvesting 1 l of C. crescentus cells grown in PYE supplemented with the appropriate antibiotic to an OD600 between 0.6 and 0.8, washing once in 10 mM Tris-acetate, pH 8.0, 14 mM Mg-acetate, 60 mM KCl and 6 mM 2-mercaptoethanol and suspending in 10 ml of 10 mM Tris-acetate, pH 8.0, 14 mM Mg-acetate, 60 M KCl and 1 mM dithiothreitol (DTT). Cells were lysed using a French pressure cell, and the extract was centrifuged twice at 21 000 r.p.m. in a Beckman SA-600 rotor. Supernatants were then supplemented to 100 mM Tris-acetate, pH 8.0, 0.5 mM DTT, 1 mM Mg-acetate, 0.3 mM ATP, 3.2 mM phosphoenolpyruvate, 0.2 mM amino acids and 20 µg ml−1 pyruvate kinase and incubated for 80 min at 37°C to remove endogenous RNA. After dialysis in 10 mM Tris-acetate, pH 8.0, 14 mM Mg-acetate, 60 M potassium acetate and 1 mM DTT, extracts were stored at −80°C.

In order to prepare labelled RNA, plasmids containing either wild-type or mutant fljK DNA (2 µg) were digested with EcoRI and PstI and gel purified. The DNA was suspended in 1 × transcription buffer (Promega), 10 mM DTT, 0.4 mM NTP mix, 125 µCi α[32P]-UTP (3000 Ci mmol−1), RNasin and T7 RNA polymerase (1 unit each) and incubated at 37°C for 60 min, after which the reaction mixture was phenol extracted at pH 5.2, chloroform extracted, ethanol precipitated and suspended in DEPC-treated water. The labelled RNA was incubated with a 200-fold molar excess of tRNA in RNA binding buffer (20 mM HEPES, pH 7.9, 100 mM KCl, 1 mM MgCl2, 1 mM DTT, 5% glycerol) for 3 min at 95°C and cooled on ice. Labelled RNA (25 fmol) was then added to appropriate S30 cell extract or pure protein in RNA binding buffer and RNasin at 2 U µl−1 in a total volume of 15 µl. After 15 min on ice, glycerol (10% final concentration) was added, and the entire reaction was loaded on a prerun 6% TBE PAGE gel and electrophoresed at 150 V for 180 min.

Synchronization of cell cultures

C. crescentus was grown in M2 medium (Contreras et al., 1987) and centrifuged through colloidal silica (Percoll, Sigma) as described previously (Evinger and Agabian, 1977). The proteins were subjected electrophoresis in a 12% SDS–polyacrylamide gel and subjected to immunoblotting (Towbin et al., 1979) using anti-FlbT antibodies.

Assay of mRNA stability

At time 0 min, 10 ml of culture was removed from exponentially growing cultures at an OD600 of 1.0, and rifampicin was added to a final concentration of 200 µg ml−1. At 30 s, 1 min, 2.5 min and 10 min, 10 ml portions were removed and frozen rapidly in dry ice–ethanol. RNA was extracted by suspending cell pellets in 0.5 ml of 30 mM sodium acetate, pH 5.2, extracting twice with phenol (equilibrated in the same buffer) at 65°C, extracting once with chloroform and precipitating with ethanol. An oligonucleotide complementary to lacZ (100 pmol) was end labelled using [γ-32P]-ATP (> 7000 Ci mmol−1) and polynucleotide kinase. Extension products were generated by incubation for 30 min at 42°C using AMV reverse transcriptase and stopped by the addition of four volumes of 1 mM EDTA and 0.2% SDS in 300 mM sodium acetate, pH 5.2. Samples were subjected to electrophoresis in a denaturing sequencing gel and quantified using a phosphorimager.

Acknowledgements

We would like to thank G. C. Draper, R. E. Muir and C. H. Boyd for assistance with this manuscript. This work was funded by grant GM48417 from the National Institutes of Health to J.W.G.

Ancillary