Extensive alanine scanning reveals protein–protein and protein–DNA interaction surfaces in the global regulator FlhD from Escherichia coli

Authors


Abstract

FlhD and FlhC are the transcriptional activators of the flagellar regulon. The heterotetrameric complex formed by these two proteins activates the transcription of the class II flagellar genes. The flagellar regulon consists not only of flagellar genes, but also of the chemotactic genes and some receptor proteins. Recently, a connection between the flagellar regulon and some virulence genes has been found in some species. Furthermore, FlhD, but not FlhC, regulates another non-flagellar target. As a first attempt to understand the mechanism of the flagellar transcriptional activation by FlhD and FlhC, the structure of FlhD has been solved. In order to understand the mechanism of the action of FlhD when it regulates the flagellar genes, we conducted site-directed mutagenesis based on its three-dimensional structure. Six interaction surfaces in the FlhD dimer were mapped by alanine scanning mutagenesis. Two of them are surface clusters formed by residues His-2, Asp-28, Arg-35, Phe-34 and Asn-61 located at each side of the dimer core. The other four are located in the flexible arms of the dimer. The residues Ser-82, Arg-83, Val-84, His-91, Thr-92, Ile-94 and Leu-96 are located at this region. All these residues are involved in the FlhD/FlhC interaction with the exception of Ser-82, Arg-83 and Val-84. These three residues affect the DNA-binding ability of the complex. The three-dimensional topology of FlhD and the site-directed mutagenesis results support the hypothesis of FlhC as an allosteric effector that activates FlhD for the recognition of the DNA.

Introduction

FlhD and FlhC proteins form the heterotetrameric complex (D2C2) which is the transcriptional activator of the flagellar regulon in Eubacteria (Liu and Matsumura, 1994; Macnab, 1996; Yanagihara et al., 1999). Their genes (flhD and flhC) belong to the class I flagellar operon flhD (the master operon). The co-ordinated flagellar regulon expression is dependent on the FlhD/FlhC complex and it occurs hierarchically (Komeda, 1982, 1986; Kutsukake et al., 1990; Kutsukake and Iino, 1994). The promoters of the class II flagellar genes are the targets of the FlhD/FlhC complex. fliA, which belongs to the class II operons and is under the direct control of the master operon, is an alternative sigma σ28 specific for the expression of the class III operons (Ohnishi et al., 1990; Liu and Matsumura, 1995). In turn, FlgM (class III), which is an anti-sigma factor, negatively regulates the class III operons (Ohnishi et al., 1992; Hughes et al., 1993; Kutsukake et al., 1994). The hierarchy of expression of the flagellar operons parallels their roles in the flagellar assembly pathway (Aizawa, 1996; Macnab, 1996).

This genetic system consists of genes that not only code for the flagellum apparatus but also for the proteins that comprise the chemotaxis system. Also, the other four genes, which belong to the flagellar regulon and are under the control of the flagellar master operon, code for receptor proteins: Tap, a dipeptide receptor (Manson et al., 1986); Tar, aspartate/maltose/Co2+/Ni2+ receptor (Clarke and Koshland, 1979; Hedblom and Adler, 1980); Trg, ribose receptor (Ordal and Adler, 1974a, b; Kondoh et al., 1979; Harayana et al., 1982); and Tsr, serine receptor (Clarke and Koshland, 1979; Hedblom and Adler, 1980; Wang and Koshland, 1980).

There are other species besides Escherichia coli and Salmonella typhimurium in which the flagellar regulon and its products have been studied. A connection between motility and virulence has been found, and there is a growing interest in the study of the control of this system in pathogenic bacteria (Ottemann and Miller, 1997). Some indications of this connection were obtained when an analogy between some flagellar genes and genes from the system for the secretion of virulence proteins of mammalian and plant pathogens (type III secretion system) was observed (Dreyfus et al., 1993; Van Gijsegem et al., 1995). Even the position of certain genes suggested a relationship between these two systems. For example, the inv gene from Yersinia enterocolitica is located between flagellar operons (Fauconnier et al., 1997). Furthermore, a co-ordinate expression of the flagellar genes and virulence genes has been described in different bacteria species (Allison et al., 1992; Gygi et al., 1995; Kapatral et al., 1996; Badger and Miller, 1998; Eichelberg and Galán, 2000). Later, the induction of phospholipase under the control of the flhD operon was demonstrated in Serratia liquefaciens (Givskov and Molin, 1992, 1993; Givskov et al., 1995) and in Y. enterocolitica (yplA gene) (Young et al., 1999a; Schmiel et al., 2000). Recently, it has been shown that certain virulence genes can be under the control of some flagellar proteins and, consequently, under FlhD/FlhC complex transcriptional regulation. FliZ, a protein whose gene belongs to the class II operon, fliAZY in Salmonella enterica, is required for the activation of the hilA invasion gene in this species (Lucas et al., 2000).

FlhD, but not FlhC, negatively regulates the transcription of non-flagellar genes when the cells enter the stationary phase (Prüß and Matsumura, 1996; Prüßet al., 1997). FlhD is also involved in several cellular processes at the transcriptional regulation level (B. M. Prüßet al., submitted).

In order to improve our knowledge about this system, we have crystallized the FlhD protein, and its three-dimensional structure has been solved to 1.8 Å resolution (Campos et al., 1998; 2001). The structure provided several insights about the possible function and mechanism of action of this global regulator. In the accompanying paper (Campos et al., 2001), the crystal structure of this protein demonstrates that FlhD is a dimer that contains two flexible arms formed by the C-terminus of each molecule and located outside the main dimer core. A putative helix–turn–helix (HTH) DNA-binding motif was found at the beginning of each flexible domain. It was also proposed that FlhC is an allosteric protein, able to activate FlhD by forming the FlhD/FlhC complex (Campos et al., 2001). Extensive alanine scanning mutagenesis directed to several surface residues has been carried out. This collection of alanine mutants develops some insights into the protein–protein interaction between FlhD and FlhC and also into the protein–DNA interaction between FlhD (when forming the FlhD/FlhC complex) and the DNA of one promoter belonging to the class II flagellar genes, fliA. Our results showed that the FlhD dimer possesses at least four FlhC-interacting surfaces and two DNA-interacting surfaces. Some of the residues that are important for the FlhD/FlhC interaction and all the residues involved in the FlhD–DNA interaction are located in the flexible arms of the dimer.

Results

FlhD is conserved among bacteria species

Given the importance of FlhD as a global regulator in the cell, we were interested in determining the relationship between homologous sequences from other bacteria. The addition of new FlhD sequences from different species allowed us to carry out an alignment in order to determine the conservation of this protein. As expected for a global regulator, FlhD is a conserved protein among members of the Eubacteria (Fig. 1). FlhD from E. coli shows a 93.1% similarity (92.24% identity) to its homologues from S. typhimurium; 82% similarity (72.4–74.1% identity) to FlhD from S. liquefaciens, Serratia marcescens, Y. enterocolitica, and Xenorhabdus nematophilus; 78.4% similarity (69% identity) to FlhD from Proteus mirabilis; 63% similarity (56% identity) to FlhD from Erwinia carotovora; and 62% similarity (49% identity) to FlhD from Bordetella bronchiseptica. The first three residues of FlhD from S. typhimurium, S. liquefaciens, S. marcescens, Y. enterocolitica, E. carotovora and B. bronchiseptica were not included in the calculations, as it is believed that they are not part of the protein in E. coli (Soutourina et al., 1999). The lowest similarity value was obtained when E. coli FlhD was compared with E. carotovora and B. bronchiseptica FlhDs. B. bronchiseptica is the only species that belongs to a different subdivision (beta subdivision; Alcaligenaceae) from all the other species (gamma subdivision; Enterobacteriaceae).

Figure 1.

Deduced amino acid sequence alignment of FlhD protein from different species. Sequence numbering is according to the E. coli FlhD residues (Bartlett et al., 1988; Soutourina et al., 1999). The secondary structure elements observed in the X-ray crystal structure of FlhD from E. coli are colour coded as follows: red, α-helix; cyan, β-sheet; black, unclassified structure. Dotted lines in the secondary structure representation and elements enclosed in the green box do not display traceable electron density in the Se-Met-FlhD crystal (Campos et al., 2001). Cys-65 mutation data were taken from the accompanying manuscript (Campos et al., 2001). Colours in FlhD from E. coli residues represent the position of every residue that was changed to alanine. The phenotype of each mutation is colour coded as follows: green, swarming; red, partial swarming or non-swarming (see text). FlhD amino acid sequences are indicated as follows: Ec, E. coli (Bartlett et al., 1988; Soutourina et al., 1999); St, S. typhimurium (Yanagihara et al., 1999; GenBank accession no. D43640); Sl, S. liquefaciens (Givskov et al., 1995; GenBank accession no. 2126176); Sm, S. marcescens (GenBank accession no. AF077334); Ye, Y. enterocolitica (Young et al., 1999b; GenBank accession no. AF081587); Xn, X. nematophilus (Givaudan and Lanois, 2000; GenBank accession no. AJ012828); Pm, P. mirabilis (Furness et al., 1997; GenBank accession no. U96964); Erwc, E. carotovora (GenBank accession no. AF130387); and Bb, B. bronchiseptica (Akerley et al., 1995; GenBank accession no. U17998). Identical residues to the FlhD E. coli sequence are represented as (*); conserved residues are underlined, and gaps in the sequences are represented as (–). The alignment was performed manually.

Given the high similarity value obtained when comparing the different sequences from these species, it seems that the secondary and tertiary structure of FlhD may be conserved among them. Certain important elements found in the crystal structure of FlhD from E. coli are conserved among the different species: Cys-65, responsible for the disulphide bond between the FlhD monomers (Campos et al., 2001); Gly-93, highly conserved among the HTH DNA-binding motifs of different regulators.

Alanine scanning mutagenesis

To date, nothing is known about the specific contacts between FlhD and FlhC and between the FlhD/FlhC complex and DNA. Alanine scanning mutagenesis was performed in combination with a genetic screening to identify critical residues involved in the FlhD/FlhC protein interaction and in the FlhD–DNA interaction. Site-directed mutagenesis was addressed to all charged residues on this protein located on the surface (Table 1A). Because the C-terminal structure was disordered and because of the presence of a putative HTH DNA-binding motif (Campos et al., 2001), every residue in the C-terminus was also changed to alanine. We excluded from these assays residues 110 and 116, as they are already alanines. During our studies, two surface clusters that map at each side of the core dimer were found. As the initial alanine scanning was directed only to the charged residues of the FlhD N-terminal, we also carried out alanine scanning in some of the residues located at the periphery of these surface clusters (see below and Table 1B).

Table 1A. DNA sequence changes of flhD′ alanine scanning mutants (first screening).
PlasmidBase substitution(s)aAmino acid change(s)bPhenotype in E. coli YK4131
  • All plasmids in this table were obtained by site-directed mutagenesis carried out on pXL27 (Table 3).

  • a . The number indicates the position of the first base of the codon containing the base substitution. Nucleotide numbering is as for the flhD + sequence, with number 1 corresponding to the first nucleotide of the flhD+ coding sequence, when the first ATG codon is considered as the start codon (base 274 in Bartlett et al., 1988).

  • b

    . Amino acids in FlhD′ have the same position number that they have in the intact FlhD protein.

  • pACC65A data were taken from the accompanying manuscript (Campos et al., 2001).

pACH2A 4CAT–GCTH2APartial swarming
pACE5A 13GAG–GCGE5ASwarming
pACK8A 22AAA–GCAK8ASwarming
pACH9A 25CAC–GCCH9ASwarming
pACD12A 34GAC–GCCD12ASwarming
pACR23A 67CGT–GCTR23ASwarming
pACD28A 82CGT–GCCD28APartial swarming
pACK29A 85AAA–GCAK29ASwarming
pACR35A 103CGT–GCTR35APartial swarming
pACE40A 118GAA–GCAE40ASwarming
pACE41A 121GAA–GCAE41ASwarming
pACK56A 166AAG–GCGK56ASwarming
pACE59A 175GAA–GCAE59ASwarming
pACC65A 193TGT–GCTC65ASwarming
pACH66A 196CAC–GCCH66ASwarming
pACR68A 202CGT–GCTR68ASwarming
pACD70A 208GAC–GCCD70ASwarming
pACH72A 214CAC–GCCH72ASwarming
pACD81A 241GAT–GCTD81ASwarming
pACS82A 244TCC–GCCS82APartial swarming
pACR83A 247CGC–GCCR83APartial swarming
pACV84A 250GTT–GCTV84APartial swarming
pACD85A 253GAC–GCCD85ASwarming
pACD86A 256GAT–GCTD86ASwarming
pACL87A 259CTC–GCCL87ASwarming
pACQ88A 262CAG–GCGQ88ASwarming
pACQ89A 265CAG–GCGQ89ASwarming
pACI90A 268ATT–GCTI90ASwarming
pACH 91A 271CAT–GCTH91APartial swarming
pACT92A 274ACC–GCCT92ANon-swarming
pACG93A 276GGC–GCCG93ASwarming
pACI94A 280ATC–GCCI94ANon-swarming
pACM95A 283ATG–GCGM95ASwarming
pACL96A 286CTC–GCCL96APartial swarming
pACS97A 289TCA–GCAS97ASwarming
pACT98A 292ACA–GCAT98ASwarming
pACR99A 295CGC–GCCR99ASwarming
pACL100A 298TTG–GCGL100ASwarming
pACL101A 301CTG–GCGL101ASwarming
pACN102A 304AAT–GCTN102ASwarming
pACD103A 307GAT–GCTD103ASwarming
pACV104A 310GTT–GCTV104ASwarming
pACN105A 313AAT–GCTN105ASwarming
pACQ106A 316CAG–GCGQ106ASwarming
pACP107A 319CCT–GCTP107ASwarming
pACE108A 322GAA–GCAE108ASwarming
pACE109A 325GAA–GCAE109ASwarming
pACL111A 331CTG–GCGL111ASwarming
pACR112A 334CGC–GCCR112ASwarming
pACK113A 337AAG–GCGK113ASwarming
pACK114A 340AAA–GCAK114ASwarming
pACR115A 343AGG–GCGR115ASwarming
Table 1B. DNA sequence changes of flhD′ alanine scanning mutants around the interface located at H2, D28 and R35.
PlasmidBase substitution(s)aAmino acid change(s)bPhenotype in E. coli YK4131
  • All plasmids in this table were obtained by site-directed mutagenesis carried out on pXL27 (Table 3).

  • a . The number indicates the position of the first base of the codon containing the base substitution. Nucleotide numbering is as for the flhD + sequence, with number 1 corresponding to the first nucleotide of the flhD+ coding sequence, when the first ATG codon is considered as the start codon (base 274 in Bartlett et al., 1988).

  • b

    . Amino acids in FlhD′ have the same position number that they have in the intact FlhD protein.

pACT3A 7ACC–GCCT3ASwarming
pACL7A 19CTG–GCGL7ASwarming
pACI25A 73ATT–GCTI25ASwarming
pACQ27A 79CAG–GCGQ27ASwarming
pACA30S 88GCG–TCGA30SSwarming
pACS31A 91TCC–GCCS31ASwarming
pACM33A 97ATG–GCGM33ASwarming
pACF34A 100TTT–GCTF34APartial swarming
pACG37A 109 GGC–GCCG37ASwarming
pACN39A 115AAT–GCTN39ASwarming
pACT60A 178ACC–GCCT60ASwarming
pACN61A 181AAT–GCTN61APartial swarming
pACQ62A 184CAA–GCAQ62ASwarming

As mentioned in Experimental procedures, mutagenesis of flhD was carried out using the plasmid pXL27, which carries both flhD and flhC. After site-directed mutagenesis, a collection of flhD mutants was obtained (Tables 1A and B). In order to confirm that the desired changes in the DNA sequence were present in each candidate, we sequenced each flhD mutant gene in all the plasmids. Also, we sequenced the upstream region from the pT7 promoter of the expression plasmid to the start codon of flhD in order to confirm the integrity of this region.

The effect of each mutation on FlhD flagellar function was studied by transforming each mutant plasmid onto strain YK4131 (flhD). As FlhD, together with FlhC, is required for the transcription of the class II flagellar genes, the functionality of FlhD was evaluated by its ability to complement this flhD strain. The bacteria harbouring each mutant plasmid were grown on swarming plates in order to confirm their individual phenotype (data not shown). By testing each single mutant, we identified 10 mutations (out of 52) that are somehow compromised in the function of FlhD. We found three different phenotypic displays: swarming cells in which the mutation does not affect the flagellar function of FlhD; partial swarming cells in which the FlhD flagellar activity is reduced but not abolished; and non-swarming cells in which flagellar function could not be observed (Table 1A and Fig. 1). Some differences were observed in the mutants with partial swarming. The halo diameter was not the same in all cases, and the time of incubation required to observe partial swarming was different. Thus, the phenotypic analyses were carried out for at least 11 h.

Our data show that most of the charged surface residues in this protein can be changed to alanine with no observable effect on function. As mentioned before, only 10 out of the 52 tested residues have an effect on the FlhD function. The transcriptional activation of flagellar genes by FlhD is decreased or abolished when His-2, Asp-28, Arg-35, Ser-82, Arg-83, Val-84, His-91, Thr-92, Ile-94 and Leu-96 are changed to alanine (Figs 1 and 2, and Table 1A). Hence, these are critical residues that affect the function of FlhD in two possible ways: by altering the binding of FlhD with FlhC or by altering the FlhD DNA binding with the promoter sequences of the class II flagellar genes. However, we cannot exclude the possibility of a structural change in the FlhD dimer structure.

Figure 2.

Stereo schematic diagram of the FlhD dimer from E. coli shown as a CPK model. Van der Waals' dots were plotted by increasing the radius of each atom by 1.4 Å. Residues in green do not affect the flagellar function of FlhD when they are changed to alanine. Residues that affect the flagellar transcription of FlhD by altering the protein–protein interaction with FlhC (yellow) or by affecting the FlhD–DNA interaction (red) are shown. Orientation of the FlhD dimer is the same as in Fig. 4 of the accompanying manuscript (Campos et al., 2001).

A. Bottom view.

B. Front view.

C. Top view.

Residue M1A was not displayed as it was disordered in the crystal structure (Campos et al., 2001). T92A is not visible in any of the views because it is covered by other residues in the C-terminus. The figure was generated with Swiss-pdb viewer v3.6b3 (Guex and Peitsch, 1997; software available from URL: http://www.expasy.ch/spdbv/) and rendered with POV-Ray 3.1 (software available from URL: http://www.povray.org/).

We have found that some of these residues that affect the flagellar function of FlhD are located in the N-terminus domain. As FlhD is a dimer, when these residues are mapped in the crystal structure of FlhD, they defined two cluster surfaces in the main core dimer. These surface clusters were defined by at least three residues (His-2, Asp-28 and Arg-35) that map together at each side of the core dimer (Fig. 2). Each monomer of FlhD is twisted together in the three-dimensional structure (Campos et al., 2001), and these surface clusters are not formed with residues from the same chain. For example, one of the cluster surfaces is formed by His-2A, Asp-28B and Arg-35B, were A and B indicate the monomer of the FlhD dimer to which the residue belongs (Fig. 2). The alanine scanning directed to the charged amino acids did not include all the residues around these surface cluster. As there was the possibility that other non-charged residues were part of these surface clusters, additional alanine scanning mutagenesis was directed to the residues located at the edges of the surface clusters in the dimer core (Table 1B). Of 13 residues from this second alanine screening, only two alanine changes were defective in FlhD activity when tested on swarming plates. Phe-34 and Asn-61 displayed a partial swarming phenotype in the YK4131 (flhD) strain (Table 1B). These additional results showed that the surface clusters located at both sides of the FlhD dimer core are formed by the residues His-2, Asp-28, Phe-34, Arg-35 and Asn-61. The following is the combination at one of the cluster surfaces: His-2A, Asp-28B, Phe-34B, Arg-35B and Asn-61A; and the following is the combination of the other cluster surface: His-2B, Asp-28A, Phe-34A, Arg-35A and Asn-61B (Fig. 2).

The rest of the interaction surfaces found in the first alanine scanning experiment were located at the C-terminal flexible domain of each FlhD monomer. Of 10 critical residues identified in the first alanine scanning experiments (Table 1A and Fig. 1), seven (Ser-82, Arg-83, Val-84, His-91, Thr-92, Ile-94 and Leu-96) mapped at these domains (Fig. 2). Furthermore, all these seven residues are located in the region in which the putative HTH motif was found (Fig. 1) (Campos et al., 2001).

The FlhD dimer core holds two FlhD/FlhC interaction surfaces, and each C-terminus domain contains interaction surfaces for both FlhC and DNA.

As the flexible domain displays characteristics of a DNA-binding motif, it was possible that the two interacting surfaces formed by His-2, Asp-28, Arg-35, Phe-34 and Asn-61 at the dimer core are needed for the interaction with FlhC. The current structure of FlhD agrees with the mutation data, which suggest that the C-terminal flexible domain (from Ser-82 to Leu-96) is critical for the interaction between FlhD and FlhC or FlhD and DNA or maybe both. Furthermore, given the fact that the C-terminal domain is highly flexible, holds a putative HTH motif and the distance between one domain and the corresponding domain in the other chain fits with the distance between two adjacent major grooves of the DNA, it was likely to be involved in binding to the promoters of the flagellar class II operons (Campos et al., 2001).

In order to understand how the function of FlhD is affected in each of the alanine mutants obtained, a FlhD/FlhC complex formation assay was carried out. It is known that the FlhD/FlhC complex, but not FlhD by itself, has binding affinity to heparin (Liu and Matsumura, 1994) (Fig. 3). This interaction can be disrupted by the addition of a high salt concentration (0.5 M NaCl). In fact, this characteristic of the FlhD/FlhC complex has been exploited to purify it to homogeneity from a cell lysate. Thus, this property was used to test the ability of each FlhD mutant to interact with FlhC and form the heterotetrameric complex that is responsible for the transcriptional activation of the flagellar genes. The binding to heparin is suggestive evidence that FlhD and FlhC form the complex. As all the pAC plasmid series also carry the flhC gene, each flhD mutant gene was overexpressed in conjunction with the flhC gene in strain MC1000flhD::kan harbouring the plasmid pGp1-2cml. In fact, the co-expression of both genes in the same cell is required to obtain the FlhD/FlhC complex. Each cell lysate was loaded directly onto a HiTrap heparin column cartridge and eluted with a linear gradient of NaCl as described in Experimental procedures. The fractions containing 0.5 M NaCl were checked by electrophoresis on an SDS–PAGE gel in order to ensure the presence or absence of the FlhD/FlhC complex. Figure 3 shows representative gel analyses of the FlhD/FlhC complex formation. When present, the complex proteins were eluted when 0.5 M NaCl was applied to the column (Fig. 3A and C). As mention before, the FlhD protein does not bind to heparin if FlhC is absent (Fig. 3B). FlhC alone was not tested, as it precipitates when FlhD is not present. Thus, the presence of both proteins in the 0.5 M NaCl fraction was interpreted as an indication that the mutation does not affect FlhD/FlhC complex formation. On the other hand, the absence of both proteins in that fraction indicated that the mutation affects the protein–protein interaction between FlhD and FlhC. Figure 3C and D show two representative complex formation analyses. FlhDS82A was able to form the FlhDS82A/FlhC complex (Fig. 3C). FlhDH91A did not form the FlhD/FlhC complex (Fig. 3D). The same procedure was applied to each of the other mutants, and the results are summarized in Table 2.

Figure 3.

FlhD/FlhC complex formation assay. Protein fractions eluted from a heparin column, analysed on SDS−20% polyacrylamide gels and visualized by Coomassie brilliant blue staining. Fractions were obtained by applying an NaCl gradient through the column.

A. Cell lysate of overexpressed native FlhD/FlhC complex (from pXL27) was used as a positive control.

B. Cell lysate of overexpressed FlhD alone (from pXL25) was used as a negative control.

C and D. Overexpression of two representative FlhD mutants together with native FlhC, FlhDS82A and FlhDH91A respectively.

Lines: 1, Mr protein markers (BSA, 68 kDa; ovalbumin, 45 kDa; carbonic anhydrase, 29 kDa; trypsin inhibitor, 20 kDa; and lysosyme, 14 kDa); 2, clear lysate; 3–9, equal volumes of fractions 17–23 taken during elution with a gradient of NaCl. When the FlhD/FlhC complex is present, both proteins are co-eluted with 0.5 M NaCl (fraction 20, line 6). The fraction containing 0.5 M NaCl is indicated.

Table 2. FlhD/FlhC complex formation assay for the mutants that affect the flagellar function of FlhD.
MutantPhenotypeFlhD/FlhC complex formation
  • a

    . NA, not applicable. Overexpression of the FlhD/FlhC complex was not obtained.

  • b

    . Non-swarming phenotype in this mutant was observed in the first hours. Longer incubations (more than 6 h) resulted in the partial swarming phenotype.

FlhDH2APartial swarmingNAa
FlhDD28APartial swarmingNo
FlhDF34APartial swarmingNo
FlhDR35APartial swarmingNo
FlhDN61APartial swarmingNo
FlhDS82APartial swarmingYes
FlhDR83APartial swarmingYes
FlhDV84APartial swarmingbYes
FlhDH91APartial swarmingNo
FlhDT92ANon-swarmingNo
FlhDI94ANon-swarmingNo
FlhDL96APartial swarmingbNo

Of the 11 mutants tested (for some unknown reason, it was not possible to overexpress the FlhD/FlhC complex from the plasmid pACH2A), only three of them were able to form the FlhD/FlhC complex (Table 2), showing that the C-terminal flexible domain contains residues important for the interaction with FlhD and FlhC and also probably with FlhD and DNA. Mutations to alanine in residues Asp-28, Arg-35, Phe-34, Asn-61, His-91, Thr-92, Ile-94 and Leu-96 affected the flagellar function of FlhD at the level of complex formation with FlhC (protein–protein interaction). The partial swarming that the mutants D28A, F34A, R35A, N61A, H91 and L96 exhibited in our experiments (Table 2) suggests that the interaction of FlhD and FlhC is affected (diminished) but not abolished. If this is true, the FlhD/FlhC interactions become weak when these residues are changed to alanine, and the protein–protein interaction becomes unstable. The quantity of complex formed under these conditions may be lower than the quantity that our system can detect (heparin column), but high enough to complement the flhD mutant strain partially. As it was not possible to obtain the complex from the heparin column in some of the mutants that display partial complementation, their DNA-binding ability was not tested.

The FlhD/FlhC complex formation for FlhDS82A, FlhDR83A and FlhDV84A was further confirmed by passing each heparin eluate through a Superdex 200 HR column in a fast protein liquid chromatography (FPLC) system. In these three cases, the proteins (FlhD and FlhC) were co-eluted from the column at around 68 kDa, which is the approximate size of the oligomeric complex FlhD/FlhC (Liu and Matsumura, 1994). The presence of the FlhD/FlhC complex when FlhD was mutated to alanine in these residues was confirmed. As changes of Ser-82, Arg-83 and Val-84 to alanine do not affect FlhD/FlhC complex formation, the other possibility was that these changes affect the DNA binding between FlhD and the DNA. In order to confirm this possibility, DNA mobility shift analyses were carried out using the promoter region of fliA, a gene of a class II flagellar operon, as a template.

As the FlhDS82A, FlhDR83A and FlhDV84A proteins have the ability to interact with FlhC and form the FlhD/FlhC heterotetrameric complex, the DNA mobility shift analyses were carried out with these purified complexes. fliA belongs to the class II flagellar genes, and its product, FliA, is an alternative sigma factor specific for the class III flagellar genes. Thus, the template used in the experiments was the promoter region of the fliA operon from E. coli (Ohnishi et al., 1990; Liu and Matsumura, 1994; 1995). FlhD alone (from pXL25) does not bind to the DNA and was used as a negative control, and the native FlhD/FlhC complex (from pXL27) was used as a positive control. The results indicated that the complex formed with these three mutants showed lower affinity than the native FlhD/FlhC complex. Figure 4 shows a representative mobility shift analysis. Complexes FlhD/FlhC, FlhDS82A/FlhC, FlhDR83A/FlhC and FlhDV84A/FlhC (500 ng of each) were tested for their ability to bind to the DNA promoter region of the fliA operon (−185 to +26). The native complex completely shifted the probe (1 ng) at that concentration, whereas only a partial shift was observed when any of the three FlhD mutants was tested at the same concentration (Fig. 4). In other experiments, even a high concentration of FlhD mutant/FlhC complex (1.75 µg), the DNA template (1 ng) was not completely shifted, whereas 400 ng of native complex was enough to bind completely to the same quantity of DNA template (data not shown). This result was expected, as these mutants displayed a partial phenotype when complemented with the YK4131 (flhD) strain (Table 2). What is relevant here is that the C-terminal flexible domain of FlhD is not only important for the protein–protein interaction between FlhD and FlhC, but also for the DNA interaction between the FlhD/FlhC complex and the DNA fliA promoter sequence.

Figure 4.

Mobility shift analysis of purified protein (500 ng of each) performed on the fliA promoter (−185 to +26). Lines: 1, free probe; 2, FlhD native protein from MC1000 flhD::kan/pGp1-2cml/pXL25; 3, FlhD–FlhC native protein from MC1000 flhD::kan/pGp1-2cml/pXL27; 4, FlhDS82A/FlhC complex from MC1000 flhD::kan/pGp1-2cml/pACS82A; 5, FlhDR83A/FlhC complex from MC1000 flhD::kan/pGp1-2cml/pACR83A; and 6, FlhDV84A/FlhC complex from MC1000 flhD::kan/pGp1-2cml/pACV84A.

Our results indicate that the FlhD dimer holds two interface surfaces important for DNA binding located at the putative HTH motif and four interface surfaces for FlhD/FlhC interaction: two of them located at the putative HTH motif region and the other two in surface clusters at both sides of the dimer core.

Discussion

The expanding list of DNA sequences of bacteria motility genes, including those from the flagellar master operon, reflects an increasing interest in the understanding of this system. As expected from a global regulator, FlhD is a conserved protein among Eubacteria. In our mutagenesis results indicating that changing any one of the last residues of FlhD (from 97 to 116) to alanine did not result in a loss of function, the C-terminal is the less conserved region of this protein (Fig. 1 and Table 1A). Furthermore, if this region is excluded from the calculations, the identity values when compared with E. coli are increased by 4% in S. typhimurium and by 8.1–8.4% among the other species, excluding E. carotovora (1.7%) and B. bronchiseptica (which, in fact, decreases by 1.6%). FlhD from E. carotovora and B. bronchiseptica are the most divergent sequences of the group. In fact, FlhD from B. bronchiseptica does not present the last residues found in the other species, again suggesting that this region is not essential. Sequence conservation reflects the functional and structural conservation of FlhD. The high degree of conservation of FlhD suggests that this protein should play similar roles in other bacteria. Furthermore, flhDC from different species is able to complement mutations in E. coli: P. mirabilis (Furnerss et al., 1997), S. liquefaciens (Givskov et al., 1995) and, more recently, Y. enterocolitica (Young et al., 1999b). Thus, we expect that FlhD proteins from other species could also be involved in non-flagellar regulation as has been found in E. coli, i.e. the cadBA operon (Prüßet al., 1997).

Previous results have shown that the Cys-65A:Cys-65B disulphide bond is not required for the flagellar function of FlhD (Campos et al., 2001). However, Cys-65 is conserved among the different species, suggesting that the disulphide bridge could be present in all the FlhD dimer proteins. This fact reinforces the idea that the disulphide bridge may play another role in an unknown function of FlhD (Campos et al., 2001). One of the most highly conserved residues in the turn of the HTH motif is a small amino acid, either glycine (most common) or alanine (Pabo and Sauer, 1984; 1992; Branden and Tooze, 1991). Gly-93 is also present in almost all the FlhD proteins from different species. However, its alanine substitution does not seem to affect the conformation of the putative HTH.

In order to study the relationship between protein structure and function on FlhD, we performed an alanine scanning based on its three-dimensional topology. Clues to the FlhD/FlhC interface surface were found in the crystal structure (Campos et al., 2001) and combined with the alanine scanning and a genetic screen provided insight into the biological function of FlhD. We have identified critical residues for the function of this protein. The structural data are consistent with our genetic analysis that identifies three domains in the FlhD dimer. The domain 1 is formed by the N-terminus from each FlhD monomer. This domain consists of residues from Met-1 to Asp-81 chain A and from Thr-3 to Asp-81 chain B, which forms a compact core dimer. This domain contains the α-helices 1–5 of both chains, the β-strand 1 of both chains, the Cys-65:Cys-65′ disulphide bond and two of the four interaction surfaces for FlhC (Fig. 2). Domains 2 and 3 are formed by flexible arms that extend from domain 1. Domain 2 is composed of residues Ser-82 to Ala-116 of chain A, and domain 3 is composed of residues Ser-82 to Ala-116 of chain B. Domains 2 and 3 contain a putative HTH motif (α-helices 6 and 7) that interacts with both FlhC and the DNA.

We identified at least six interaction surfaces on FlhD. Two surface clusters were located at opposite sides of the main core, and the other four mapped at the C-terminal flexible domain (Fig. 2). The putative HTH motifs at the flexible arms suggested that they are functional domains and that FlhD is part of the complex with DNA-binding properties (Campos et al., 2001). This premise was further supported by the genetic analysis, which showed that Ser-82, Arg-83, Val-84, His-91, Thr-92, Ile-94 and Leu-96 are critical residues for FlhD function. Our data suggest that the surface clusters and part of the C-terminus in the dimer are important for the FlhD/FlhC interaction. On the other hand and as expected from the crystallographic studies, the FlhD C-terminus is also important for the FlhD/FlhC–DNA interaction, as shown by the mobility shift analyses (see below).

Residues His-2A, Asp-28B, Phe-34B, Arg-35B, Asn-61A and His-2B, Asp-28A, Phe-34A, Arg-35A and Asn-61B form two surface clusters in the core dimer (Fig. 2), and their topology is in agreement with the observation that the heterocomplexes present interfaces that are more planar than the homodimers (Jones and Thornton, 1996). Complex formation assay showed that these cluster surfaces form part of the FlhD–FhC interface. In fact, when the electrostatic potential is calculated on the surface, the position of these residues involved in the FlhD/FlhC interaction correlates with the position of the neutral and positive patches located at both sides of the core dimer. Compare Fig. 2B of this work with Fig. 5E of the accompanying paper (Campos et al., 2001). However, we cannot exclude the possibility that some of the mutations can affect FlhD function by changing the molecular structure, i.e. Arg-35 is not completely exposed in the surface.

As His-2 is part of these interface clusters, it is possible that Met-1 is also included in the interface. Unfortunately, only the Met-1A was observed with poor resolution in the crystal structure, and its relative position is not known (Fig. 1 of Campos et al., 2001). As it is not possible to obtain a Met-1 mutant, further studies directed at solving the crystal structure of the FlhD/FlhC complex should answer this concern. The fact that the surface clusters in the main core are formed by residues from both chains supports the hypothesis that FlhD is an obligate dimer (Campos et al., 2001). Even if FlhD can be separated in monomers, these surface clusters will not be present in the single polypeptide. As shown by the complex formation analyses, His-91, Thr-92, Ile-94 and Leu-96, located at the flexible domain, are also involved in the FlhD/FlhC interaction.

It was surprising that these residues are located in the putative HTH. This is an unusual function for an HTH motif. As FlhD is a dimer, all the FlhD/FlhC interface surfaces we found are represented twice (Fig. 2).

Before the crystal structure of FlhD was obtained, nothing was known about the FlhD/FlhC–DNA interaction. The mere presence of the putative HTH motifs in the FlhD dimer suggested that FlhD is the factor in the FlhD/FlhC complex that has the ability to bind to the DNA. Our mutagenesis and the mobility shift DNA analyses are consistent with this idea. Mutants FlhDS82A, FlhDR83A and FlhDV84A are able to form the FlhD/FlhC complex, but the affinity for complex–DNA binding was reduced as shown by the mobility shift experiments. This is in agreement with our phenotypic analyses, which showed a reduced activity in the mutants (Tables 1A and 2). Ser-82, Arg-83 and Val-84, located in the first helix of each putative HTH motif, form a DNA interaction surface. The fact that only three residues in the putative HTH motif affect DNA binding was not expected. Alanine mutations in this region may be tolerated because alanine is a small amino acid with no chain beyond the β-carbon to interact sterically with the DNA. The FlhD component of the FlhD/FlhC complex appears to interact with the DNA, but it is not known whether FlhC also interacts with the DNA.

FlhD does not have DNA-binding activity unless it is complexed with FlhC (Liu and Matsumura, 1994). Specific DNA binding only occurs when the FlhD/FlhC complex is formed. Based on the FlhD crystal structure and on the results of genetic analyses, we believe that the DNA specificity of FlhD is given by FlhC. FlhC is an allosteric activator able to change the conformation of FlhD by holding the two C-terminal flexible domains in a suitable position, which permits interaction with the DNA (Fig. 5). In our working model, each molecule of FlhC interacts with one of the surface clusters in the main core, while another region of the same molecule holds one of the flexible arms of FlhD dimer by interacting with the residues His-91, Thr-92, Ile-94 and Leu-96 of one of the flexible arms (Fig. 5). We do not know whether FlhC is also a dimer, but two molecules of FlhC are needed to maintain both flexible arms rigid in order to confer the DNA-binding specificity. This allosteric interaction could stabilize the flexible arms of the FlhD dimer, so that the residues Ser-82, Arg-83 and Val-84 can face the DNA (Fig. 5).

Figure 5.

Hypothetical allosteric activation of FlhD by FlhC.

A. In order to visualize the partial C-terminus at both sides of the FlhD dimer, the current crystal structure of the FlhD dimer was modelled with two chain As from the crystal structure (Campos et al., 2001).

B. Hypothetical representation of the FlhD/FlhC and FlhD–DNA interaction. Our results support the hypothesis that each molecule of FlhC interacts with two different regions of the FlhD dimer. As the structure of FlhC is not known, only the putative interfaces with FlhD are displayed as a blob in transparent models. Each blob represents part of each molecule of FlhC: one located at the front and the other behind the FlhD dimer model. Each molecule of FlhC may interact with the FlhD core dimer while it holds one of the C-terminal flexible domains. This allosteric interaction should generate a rotation on the flexible arms (red arrows) of the FlhD dimer, so that the residues Ser-82, Arg-83, and Val-84 can face the DNA. See text for further details. Notice that the rotation of the flexible domains can be opposite to that shown in the figure. Residues involved in the FlhD/FlhC interaction (yellow) and FlhD–DNA interaction (red) are displayed as a CPK model. The FlhD dimer and DNA models were generated with Swiss-pdb viewer (Guex and Peitsch, 1997; software available from URL: http://www.expasy.ch/spdbv/). Additional elements (FlhC blob representation) were added and rendered with POV-Ray 3.1 (software available from URL: http://www.povray.org/).

In this work, we have shown that most of the charged residues in FlhD can be changed to alanine with no observable effect on flagellar function. Excluding the C-terminus (from Ser-82 to Ala-116), there are 16 charged residues that do not affect the flagellar function of FlhD when they are changed to alanine. It is possible that some of them are involved in some other function(s) of FlhD. As a global regulator, FlhD may change its DNA-binding specificity depending on the protein with which it interacts. Maybe FlhD has the ability to form different complexes inside the cell with other partners to regulate different targets. Some associations will need a strong binding of FlhD, whereas other circumstances may dictate weaker binding to another protein.

The combinatorial specificity makes FlhD a versatile and appropriate protein not only as a model for the study of protein–protein interactions and protein–DNA interactions studies, but also as a model for the study of transcriptional activation by complex formation.

Experimental procedures

Bacterial strains, plasmids and media

The genotypes of E. coli strains and plasmids used in this study are outlined in Table 3. E. coli strain Epicuran Coli XL1-Blue (Stratagene) was used to propagate the plasmids generated by in vitro site-directed mutagenesis. Strain YK4131 was used to examine the phenotype of the alanine scanning mutants in swarming plates. E. coli strain MC1000 flhD::kan harbouring the plasmid pGp1-2cml was used to overexpress the FlhD/FlhC complex. Luria–Bertani broth (LB) (Miller, 1992) was used for all purposes other than swarm phenotypic analysis. Antibiotics (Sigma) were added when required at the following concentrations: penicillin G, 100 µg ml−1; kanamycin, 30 µg ml−1; chloramphenicol, 25 µg ml−1. Phenotype assays were performed on tryptone soft agar plates (1% tryptone, 0.5% NaCl and 0.3% bacto-agar). Plasmid pGp1-2cml was used for the expression of the T7 DNA polymerase, which in turn induces expression in the pT7-7 derivative plasmids of the cloned genes. Plasmids pXL25 and pXL27 (and derivatives) were used to overexpress the FlhD E. coli protein and the FlhD/FlhC complex respectively.

Table 3. Escherichia coli strains and plasmid genotypes.
E. coli strainsGenotypeReference
YK4131 flhD derivative of YK410 [F, araD139, Δlac(U169), rpsL, thi, pyrC46, nalA, thyA, his]Komeda et al. (1980)
MC1000 flhD::kan flhD flhC derivative of MC1000 [F, araD139, Δ(araAB, leu)7697, Δ(lacX74), galK, strA]Malakooti et al. (1989)
Epicuran Coli XL1-Blue recA1 endA1 gyrA96 thi-1 hsdR17 supE44 relA1 lac[F′proAB lacIqZΔM15 Tn10 (Tetr)]Stratagene
Plasmids
 pGp1-2cmlT7 RNA polymerase, CmlRRendone (1992)
 pXL21pAN518-fliA (−182 to +26)Unpublished data
 pXL25pT7-7-flhD Liu and Matsumura (1994)
 pXL27pT7-7-flhD/C Liu and Matsumura (1994)
 pAC seriespXL27 derivatives (see Tables 1A and B)This study

Alanine scanning by site-directed mutagenesis

In vitro site-directed mutagenesis of flhD was performed in accordance with the QuikChange site-directed mutagenesis kit (Stratagene) protocol. Plasmid pXL27, which carries the complete flhD and flhC genes, was used for each flhD mutagenesis experiment according to the instructions of the manufacturer. Changes were introduced through individual and internal mutagenic oligonucleotide primers, complementary to flhD except at the position of the desired mutation (Table 1A and B). The changes were located in the middle of the primer with ≈ 8–16 bases of correct sequences on both sides. In vitro site-directed mutagenesis reactions were performed using 50 ng of pXL27 as the template for each reaction. Twenty cycles were completed with the following parameters: 95°C for 30 s, 55°C for 1 min and 70°C for 8 min. Plasmids carrying each flhD mutant were transformed into Epicuran Coli XL1-Blue competent cells after 1 h of digestion with DpnI. Each flhD mutant plasmid was then purified using the QIAprep spin miniprep kit (Qiagen), sequenced (Sanger et al., 1977) using the CircumVent thermal cycle dideoxy DNA sequencing kit (New England Biolabs) and transformed into the flhD strain YK4131. For each plasmid generated (Table 1A and B), the whole sequence of the flhD mutant gene was confirmed. The upstream region from the PT7 to the start codon of flhD was also sequenced to exclude any random change that might possibly affect the expression level of the flhD gene. Phenotypic analyses were carried out in swarming plates at 30°C in a humid box for 5–6 h.

Genetic screening of FlhD alanine mutants in swarming plates

Phenotypic analyses were carried out by growing the cells (E. coli YK4131) harbouring the desired plasmids on LB at 37°C overnight. Two microlitres of an overnight culture were used to inoculate each swarming plate supplied with the appropriate antibiotic when required. Incubation of the swarming plates was carried out at 30°C into a humid camera to avoid evaporation of the media. The swarm diameter of each strain was monitored every 30 min for 11 h.

FlhD and FlhD/FlhC complex purification and complex formation assay

Overexpression and purification of FlhD native protein was carried out using the two pT7 systems (Tabor, 1990) in E. coli MC1000 flhD::kan harbouring plasmids pGpP1-2cml and pXL25, as described previously (Campos et al., 1998; 2001).

Overexpression of E. coli FlhD/FlhC complex was carried out using the two-plasmid pT7 system (Tabor, 1990) in E. coli MC1000 flhD::kan harbouring plasmids pGpP1-2cml and pXL27 (or pXL27 flhD mutant derivatives; Table 3) in 1 l of LB broth as described previously (Liu and Matsumura, 1994). pXL27 and derivatives contain the complete coding sequence of flhD and flhC. Overinduced culture cells were disrupted by sonication in 20 mM Tris (pH 7.9) and centrifuged at 31 000 g for 30 min. The FlhD/FlhD complex was then purified to homogeneity by loading the supernatant onto a HiTrap heparin column cartridge (Amersham Pharmacia Biotech) in a Bio-Rad Econo system. The heterotetramer complex was eluted with a linear gradient of NaCl from 0 to 1.4 M in 20 mM Tris (pH 7.9). FlhD/FlhC complex unbinds the heparin at a 0.5 M NaCl concentration. Protein fractions from the heparin column were analysed on SDS−20% polyacrylamide gels and visualized by Coomassie brilliant blue staining. Soluble FlhD/FlhC protein was stored at −20°C until use for DNA mobility shift assay when applicable.

FlhD/FlhC complex size was confirmed by loading each heparin-purified complex onto a Superdex 200 HR 10/30 filtration column (Pharmacia Biotech) in an ÄKTAfplc system (Amersham Pharmacia Biotech). The column was equilibrated using BSA (65 kDa), trypsin inhibitor (20 kDa) and lysosyme (14 kDa). Chromatography of each FlhD/FlhC complex was carried out using 20 mM Tris, pH 7.9, and 0.5 M NaCl as a buffer at a flow rate of 0.5 ml min−1.

Labelling of fliA promoter region

The fliA promoter sequence template used in this study for mobility shift assays was prepared as follows. Amplification of the fliA promoter region from E. coli was carried out by polymerase chain reaction (PCR) (Saiki et al., 1988) using primers fliA_5′B, 5′-GCGCATCCGGCAACATAAAG-3′ and fliA_3′B, 5′-CCTTCAGCGGTATAGAGTG-3′. The primers were used to amplify a 211 bp fragment whose 5′ and 3′ termini extended to –185 and +26, respectively, relative to the transcriptional +1 position of the fliA gene (Liu and Matsumura, 1995). The plasmid pXL21 was used as the template (Table 3).

The PCR reaction was performed in 100 µl volumes of Tris buffer (75 mM, pH 8.8) containing 200 mM ammonium sulphate, 0.01% Tween 20, 2 mM magnesium chloride, 0.4 mM each dNTP, 50 pmol of each primer, 50 ng of template DNA and Taq DNA polymerase (5 U; MBI Fermentas). Thirty cycles were completed with the following parameters: 95°C for 30 s, 50°C for 1 min and 72°C for 1 min. The PCR product was then 5′ labelled with T4 polynucleotide kinase (Promega) with [γ-32P]-ATP (Amersham Pharmacia Biotech). One microlitre of DNA was incubated with 10 U of T4 polynucleotide kinase in 10 µl of 1× buffer (distributed by the supplier) and incubated for 30 min at 37°C. Further purification of the labelled fragment was carried out using the QIAquick nucleotide removal kit (Qiagen) according to the manufacturer's instructions and by elution of the probe with 100 µl of 10 mM Tris-Cl, pH 8.5, supplied with the kit. The probe was stored at −20°C until use in the DNA mobility shift assays.

DNA mobility shift assay

To test the ability of each FlhD/FlhC complex for DNA binding, the heparin-eluted complexes were used. Mobility shift assay was carried out according to the method of Fried and Crothers (1981). The binding reactions were carried out in a 20 µl volume of 5 mM Tris (pH 7.9), 50 mM KCl, 5 mM EDTA, 1 mM dithiothreitol, 5% glycerol, 10 ng of poly-(dI–dC)·(dI–dC) (Amersham Pharmacia Biotech), 1 ng of 32P-labelled DNA and 50–1750 ng of purified protein. After incubation at 30°C for 30 min, the samples were loaded onto 5% polyacrylamide gels, and the bands were visualized by autoradiography.

Acknowledgements

The authors thank Azucena Rosas for her help in the DNA characterization of the alanine scanning mutants, and Bryan Shimkos and Rhonda T. Fleming for their help in the second alanine scanning. We also acknowledge Peggy O'Neill for critical reading of the manuscript. This work was supported by National Institutes of Health grant GM59484 to P.M.

Ancillary

Advertisement