FliZ, a flagellar regulator, is at the crossroads between motility, haemolysin expression and virulence in the insect pathogenic bacterium Xenorhabdus


  • Anne Lanois,

    1. INRA and
    2. Université Montpellier 2, UMR 1133 Laboratoire EMIP, F-34000 Montpellier, France.
    Search for more papers by this author
  • Grégory Jubelin,

    1. INRA and
    2. Université Montpellier 2, UMR 1133 Laboratoire EMIP, F-34000 Montpellier, France.
    Search for more papers by this author
    • Present address: UMR 1225 (INRA-ENVT), Ecole Nationale Vétérinaire de Toulouse, 23 chemin des Capelles BP87614, 31076 Toulouse Cedex 3, France.

  • Alain Givaudan

    Corresponding author
    1. INRA and
    2. Université Montpellier 2, UMR 1133 Laboratoire EMIP, F-34000 Montpellier, France.
    Search for more papers by this author

E-mail givaudan@univ-montp2.fr; Tel. (+33) 467 14 48 12; Fax (+33) 467 14 46 79.


There is a complex interplay between the regulation of flagellar motility and the expression of virulence factors in many bacterial pathogens. We investigated the role of FliZ in the regulation of flagellar and virulence genes in Xenorhabdus nematophila, an insect pathogen. The fliZ gene is the second gene in the fliAZ operon in X. nematophila. In vivo transcription analysis revealed a positive feedback loop of fliAZ transcription in which FliZ activates flhDC, the master operon of flagellar regulon in X. nematophila, leading to an increased transcription of the FlhDC-dependent promoter of fliAZ. We also showed that fliAZ and flhDC mutants lacked motility, had no haemolysin or Tween lipase activity and displayed an attenuated virulence phenotype in insects. Lipase activity is controlled by FliA, whereas haemolysin production and full virulence phenotype have been reported to be FliZ-dependent. Transcriptional analysis revealed that FliZ directly controlled expression of the xhlBA and xaxAB operons, which encode haemolysins from the two-partner secretion system and the binary XaxAB toxin family respectively. We suggest that this regulatory pathway may also occur in other pathogenic enterobacteria with genes encoding members of these two growing families of haemolysins.


Flagella are complex surface structures that serve as the primary means of locomotion for many bacterial species and allow many bacterial pathogens to adhere to and invade cells and to secrete virulence factors (Ramos et al., 2004). More than 50 genes are involved in the biogenesis and function of a flagellum of Escherichia coli or Salmonella enterica serovar Typhimurium (Macnab, 1992). These genes are transcriptionally regulated as a cascade and are co-ordinated with the flagellar hierarchy (Macnab, 1992). At the top of the hierarchy is the class I operon, flhDC, whose products, FlhD4C2 heterohexamers, are required for the expression of all other flagellar genes (Bartlett et al., 1988; Kutsukake, 1994; Wang et al., 2006). The E. coli FlhD4C2 complex activates class II operons, including most of the structural genes for flagellar hook-basal body components [a type III secretion system (TTSS)] and the alternative sigma factor fliA (Liu and Matsumura, 1994). The fliA gene is the first gene of the fliAZY operon in E. coli and S. enterica and its product, σ28, directs transcription of the class III genes encoding the filament protein, hook-associated proteins, motor proteins and various chemotaxis proteins (Ohnishi et al., 1990).

Several genes encoding regulators within the flagellar regulon are known to affect the expression of class II genes. The central channel of the flagellar apparatus is thought to serve as a passage for both flagellar component proteins and for flagellar regulatory protein, FlgM, an anti-sigma factor (Hughes et al., 1993; Kutsukake, 1994). Thus, the accumulation of FlgM in the cell due to the prevention of its export blocks the transcription of class III genes, including that encoding flagellin. In addition to flgM, two other genes fliT and fliZ have been shown to regulate class II gene transcription in S. enterica (Kutsukake et al., 1999). Disruption of the fliT gene increases class II gene transcription, whereas disruption of the fliZ gene decreases class II gene transcription, suggesting that FliT and FliZ are negative and positive regulators respectively. FliT has been shown to act as an anti-FlhD4C2 factor, binding to FlhD4C2 by interacting with the FlhC subunits and inhibiting binding to the class II promoter (Yamamoto and Kutsukake, 2006). However, the precise molecular mechanisms underlying the upregulation of class II operons by FliZ remain unknown.

Flagellar regulators have recently been shown to be involved in processes other than flagellar gene expression. E. coli flhD mutants cannot sense the serine depletion of the medium that induces wild-type cells to reduce their rate of cell division (Pruss and Matsumura, 1997). The FlhDC complex controls the expression of metabolic genes in E. coli (Pruss et al., 2003) and Yersinia enterocolitica (Kapatral et al., 2004). Many studies have shown that flagellar motility is essential for some stages in the life cycle of many bacterial pathogens and that the expression of virulence genes is often coupled to that of flagellar genes via regulatory networks (for a review, Josenhans and Suerbaum, 2002). In Y. enterocolitica, FlhDC and FliA have been shown to have negative effect on plasmid-encoded virulence gene expression (Bleves et al., 2002; Horne and Pruss, 2006). Another overlap between the regulatory mechanisms controlling the flagellar regulon and the expression of TTSS-dependent invasion genes has also been described in S. enterica (Eichelberg and Galan, 2000). Moreover, it has been suggested that type III protein secretion by the flagellar apparatus may be a general mechanism for transporting proteins involved in virulence (Young et al., 1999). The secretion of virulence proteins, such as exoenzymes, requires a functional flagellar type III export system in Y. enterocolitica (Young et al., 1999), Serratia liquefaciens (Givskov et al., 1995), Campylobacter jejuni (Konkel et al., 2004) and two bacteria pathogenic to insects: Bacillus thuringiensis (Ghelardi et al., 2002) and Xenorhabdus nematophila (Park and Forst, 2006).

Other studies revealed the existence of common regulatory mechanisms for controlling flagellar gene expression and the virulence-associated TTSS (Eichelberg and Galan, 2000; Bleves et al., 2002). In S. enterica serovar Typhi, fliA disruption has been reported to reduce significantly the expression of invasion genes clustered in the SPI-1-encoded TTSS (Eichelberg and Galan, 2000). Other studies attributed this effect to FliZ, which regulates hilA, a positive regulator of SPI-1 genes (Lucas et al., 2000; Iyoda et al., 2001).

Xenorhabdus nematophila, a motile Gram-negative bacterium belonging to the Enterobacteriaceae family, kills various insects. These bacteria also form a species-specific mutualistic association with the entomopathogenic nematode, Steinernema carpocapsae (Thomas and Poinar, 1979). X. nematophila is transported and released by nematode vectors into the haemocoel (body cavity) of insect hosts. Xenorhabdus initially colonizes the connective tissue surrounding the anterior midgut and haemolymph (bloodstream) of lepidopteran insects (Sicard et al., 2004), leading to the death of the insect, probably due to a combination of the effects of toxins and septicaemia. X. nematophila can independently kill insects and grow within their bodies if introduced into the haemolymph by direct injection in the laboratory (Sicard et al., 2004). X. nematophila proliferates in the haemolymph before the insect dies and must therefore be able to escape the insect immune response. Cell-mediated immunity comes into play immediately after the insect haemocoel is penetrated by a foreign body, resulting in the clearance of non-pathogenic bacteria from the haemolymph within hours. Haemocytes are the immunocytes of insects, and cellular immune responses to bacteria involve cells of several different lineages (Ribeiro and Brehelin, 2006). Cytotoxic factors targeting immunocytes are good candidates for the mediators of immunosuppression. The culture supernatants of most Xenorhabdus species, including X. nematophila, are cytotoxic (Brillard et al., 2001). C1, a cytotoxin targeting phagocytes, is one of the most active cytotoxic molecules in X. nematophila growing in broth cultures. C1 also lyses sheep red blood cells (SRBC) but not rabbit red blood cells (RRBC). Another cytotoxin, C2, lyses insect plasmatocytes and erythrocytes from rabbit, but not sheep. A peptide cytotoxin purified from a culture medium with C1 activity has been shown to lyse cells via a mechanism involving pore formation (Ribeiro et al., 2003) and to trigger apoptosis (Vigneux et al., 2007). Two genes, xaxAB, have been shown to encode this toxin in X. nematophila and their inactivation fully abolishes C1 cytotoxic activity. The xax gene family is also present in other entomopathogenic bacteria of the genus Photorhabdus and in Pseudomonas entomophila, in the human pathogen Y. enterocolitica and in the plant pathogen Pseudomonas syringae (Vigneux et al., 2007). A cell surface-associated haemolysin, XhlA, also lyses the two most common types of insect immune cell (granulocytes and plasmatocytes), as well as rabbit and horse erythrocytes (Cowles and Goodrich-Blair, 2005). XhlA, like its Photorhabdus homologue PhlA (Brillard et al., 2002), belongs to a family characterized by a two-partner secretion system (TPSS).

We have shown that FlhDC positively regulates haemolysin and lipase production and is required for full virulence in X. nematophila (Givaudan and Lanois, 2000). More recently, FliA-regulated lipase (xlpA) and protease (xrtA) genes were identified (Park and Forst, 2006). In this study, we carried out a precise transcriptional analysis of the flagellar cascade in Xenorhabdus. We showed that expression of the fliAZ operon displayed positive autoregulation of the FlhDC-dependent promoter via FliZ but not via FliA. We also found that the class II regulators, FliA and FliZ, were at a crossroads in Xenorhabdus, mediating the co-ordinate regulation of lipase activity and motility, and haemolysis and virulence. We also found that FliZ bound to the promoter specific to both haemolysin-encoding genes, xaxAB and xhlAB, in vivo, inducing the transcriptional activation of these haemolysin genes even in the absence of the FlhD4C2 complex. Finally, we clearly demonstrated that only FliZ-dependent genes in the flagellar cascade are involved in the virulence of Xenorhabdus in insects.


Sequence analysis of the fliAZ operon in X. nematophila

Flagellar genes and operons are tightly clustered on the chromosome of enterobacteria (Soutourina and Bertin, 2003). Region III of the E. coli and S. enterica serovar Typhimurium chromosomes contains genes encoding export apparatus proteins, flagellar structural proteins and the flagellar-specific sigma factor, FliA. The gene encoding flagellin, fliC, and the hook-associated protein 2 gene, fliD, have been cloned from a genomic library constructed from the X. nematophila chromosome (Givaudan et al., 1996). We initially cloned a 15 kb Sau3A insert from plasmids harbouring the fliC gene. We showed that the plasmid containing this insert, pAL3, restored the motility of E. coli YK4104, a strain that is non-motile due to a mutation in the fliA gene. Subcloning steps yielded pAL3011, with a 2.8 kb HindIII insert, which complemented an E. coli fliA mutant, conferring full motility. Analysis of the DNA sequence of this insert revealed two complete ORFs in the same orientation (Fig. 1). These two ORFs were downstream from a partial gene encoding FliC (Givaudan et al., 1996) and had predicted primary amino acid sequences 80% and 52% identical to E. coli FliA and FliZ respectively. X. nematophila FliA and FliZ were most similar to their respective homologues from Photorhabdus luminescens (93–87% identity) and Proteus mirabilis (87–85% identity). The low level of identity between the FliZ proteins from X. nematophila and E. coli (52% identity) can be accounted for by divergence in the C-terminal region. Stem–loop structures were found downstream from each gene (fliC, fliA and fliZ), but these structures were not followed by the series of uridine residues characteristic of E. coli rho-independent terminators (d'Aubenton Carafa et al., 1990).

Figure 1.

Identification of fliAZ operon by complementation experiments.
A. Physical and genetic map of the fliCAZ chromosome region of Xenorhabdus nematophila. Grey arrows represent ORFs. Solid bars represent the DNA fragments cloned in pUC19.
B. Complementation analysis. Restoration of the motility of the E. coli fliA mutant (YK4104) is shown. +, large spreading area in mot agar.

A search for conserved domains (CDD) revealed that the C-terminal sequence of FliZ from X. nematophila contained a domain found in the phage integrase family: the N-terminal SAM-like domain. Multiple sequence alignments of C-terminal sequences from the FliZ proteins of X. nematophila, E. coli, Yersinia pestis and P. mirabilis generated with Tito software (Labesse and Mornon, 1998) showed that all FliZ sequences displayed folding patterns consistent with the three-dimensional structure of the N-terminal domain of the site-specific recombinase XerD (Subramanya et al., 1997). The XerD integrase belongs to the tyrosine recombinase family, some of the members of which are involved in the site-specific integration and excision of lysogenic bacteriophage genomes, the transposition of conjugative transposons, the termination of chromosomal replication and stable plasmid inheritance. The larger C-terminal domain containing the RHR triad and the conserved tyrosine nucleophile binds, cleaves and religates DNA strands at the core sites, whereas the N-terminal domain is largely responsible for high-affinity binding to DNA sequences (Subramanya et al., 1997). Such structural compatibility between FliZ and XerD suggests that the C-terminal FliZ domain may be involved in DNA binding.

fliAZ transcription in X. nematophila

We investigated whether the fliAZ operon was transcribed as a polycistronic mRNA by carrying out Northern blots with probes specific for fliA and fliZ. The detection of a single band of about 1400 nt with both probes demonstrated the production of large amounts of polycistronic mRNA in X. nematophila (data not shown). Reverse transcription polymerase chain reaction (RT-PCR) analysis with primers amplifying overlapping fragment of the polycistronic mRNA confirmed the presence of polycistronic mRNA molecules encompassing both fliA and fliZ (data not shown). A second faint band was detected at 900 nt with the fliZ probe only, suggesting that this band corresponded to a transcript containing fliZ alone present at low copy number.

The fliAZY operon in E. coli and Salmonella is under the control of two promoters: a FliA-dependent class III promoter and an FlhDC-dependent class II promoter. Alignments of the upstream sequences flanking the fliA genes from X. nematophila and other enterobacteria confirmed the presence of a consensus sequence containing a potential −10 region and displaying imperfect dyad symmetry [AA(C/T)G(C/G)N2/3AAATA(A/G)CG], known to be involved in FlhD4C2 binding (Claret and Hughes, 2002), in all sequences (see Fig. S1). Surprisingly, the FliA-dependent class III promoter consensus was found only in sequences upstream from fliA in Salmonella and E. coli. This suggests that the fliAZ operon from Xenorhabdus, Photorhabdus, Yersinia, Serratia and Proteus is essentially under the control of the flagellar class II promoter.

Primer extension was carried out to validate these putative promoters of fliAZ and three extension products were obtained (Fig. 2). Two faint signals mapping to an adenine nucleotide and a guanidine nucleotide far upstream from the initiation codon of the fliA gene were designated P1. A second promoter, P2, was detected by the mapping of a strong positive signal to an adenosine residue 28 bases upstream from the initiation codon of the fliA gene. The regulatory elements of the P2 promoter correspond to the −10 region typical of σ70 promoters and the −40 to −80 dyad-containing region detected in silico (see Fig. 2 and Fig. S1). These findings are consistent with footprinting data showing that a region between nucleotides −30 and −80 with respect to the transcriptional start site for E. coli class II promoters is protected in vitro by the FlhD4C2 complex (Claret and Hughes, 2002).

Figure 2.

Localization of the start sites of transcription upstream from the fliA gene.
A. The fliApext primer (Table S1) was hybridized with total RNA extracted from the X. nematophila F1 wild-type strain (WT) grown in mot broth to an OD540 of 0.5, which was then extended with reverse transcriptase. Sequencing reactions (A, C, G and T) were performed using the same primer on pAL3011 as a template (Fig. 1).
B. DNA nucleotide sequence upstream from the fliA gene and deduced amino acid sequence corresponding to the 5′fliA gene region. Boldface uppercase letters indicated the +1 residues of both putative promoters P1 and P2. A putative FlhDC complex binding the consensus sequence (Claret and Hughes, 2002) and the −10 region of P2 are double underlined in the sequence upstream from P2 initiation. Single underlined nucleotides indicate RBS (ribosome binding sites). The sequence complementary to the fliApext primer is indicated by a broken line.

As a minor transcript containing only fliZ was identified, we carried out primer extension analysis with a primer binding to the 5′ end of fliZ to assess whether transcription was initiated in the fliA–Z intergenic region. Numerous faint signals corresponding to transcript termini (intergenic region 3, IR3) were detected in the 5′ region of fliZ. However, none of these signals was compatible with initiation of transcription from a putative σ70 promoter or from a flagellar regulator-dependent promoter.

FliAZ expression in X. nematophila is essentially controlled by an FlhDC-dependent promoter

We first investigated expression from the various fliAZ promoters in wild-type and FlhD-defective backgrounds, using plasmids containing transcriptional fusions with the uidA reporter gene. The P1fliA–uidA and IR3fliZ–uidA fusions were expressed only weakly in both chromosomal backgrounds, revealing P1 and IR3 to be FlhD-independent promoters with low levels of activity. The P2fliA–uidA fusion displayed higher levels of activity in the wild-type strain than in the flhD mutant (Fig. 3A). Consistent with the presence of the consensus sequence for FlhDC binding, β-glucuronidase assays clearly showed that P2 is an FlhDC-dependent promoter.

Figure 3.

Transcriptional analysis and protein–DNA interaction studies on fliA promoters.
A. Gus activities of the fliAZ promoter regions in wild-type and flhD mutant strains. We constructed uidA fusions in plasmids with the P1, P2 and IR3 regions of fliAZ and these plasmids were transferred into the wild type and flhD mutant. pBB-UIDK was used as a negative control. GUS activity was assayed on bacteria grown in mot broth to an OD540 of 0.5. Specific GUS activity was calculated as GUS activity per unit optical cell density. The mean and SD were calculated from the data for three independent cultures.
B. Activation of the P2 promoter of the fliA gene by FliZ but not by FliA. The P2–uidA fusion was transferred into ΩfliA (pRK404), ΩfliA (pKA7), ΩfliA (pKZ4), ΩflhD (pRK404) and ΩflhD (pKZ4). GUS activity was assayed on bacteria grown in mot broth at an OD540 of 0.5. Specific GUS activity was calculated as GUS activity per unit optical density. The mean and SD were calculated from the data for three independent cultures.
C. In vivo promoter binding to the FlhD protein. ChIP assays were performed with FlhD–HA in the presence of FlhC (WT and ΩflhDpKC2) or in the absence of FlhC (ΩflhDpRK404). Enrichment with the fliA promoter region in the different backgrounds was analysed and compared with that of the mreB internal region (non-target DNA). The relative values of enrichment for the indicated promoter were normalized with respect to a control region in the gyrB gene. The flagellar genotype of each strain is described below the graph. The data represent the means of three independent cultures at an OD540 of 0.4.

In order to confirm the in silico analysis suggesting that the fliAZ operon from Xenorhabdus is not under the control of a flagellar class III promoter, we investigated the expression of P1fliA and P2fliA fusions with uidA in a fliA mutant, with and without complementation mediated by a plasmid overexpressing fliA. As expected, FliA alone had no effect on expression from the P2 promoter (Fig. 3B). Similar results were obtained with the P1 promoter region (data not shown). These data suggest that the FlhD4C2 complex is the main activator of fliAZ operon in X. nematophila.

In order to show that the FlhD4C2 complex bound to the fliA promoter in vivo, we carried out chromatin immunoprecipitation (ChIP) assays. A polar flhD mutant expressing the flhC gene and a fusion encoding an FlhD–HA construct in trans was used to immunoprecipitate DNA bound to FlhD–HA protein. Quantitative PCR showed that the immunoprecipitated fraction displayed a 75-fold specific enrichment of the fliA promoter DNA region, whereas no such enrichment was observed if the flhD strain did not carry the flhC-expressing plasmid (Fig. 3C). Thus, FlhD and FlhC bind as a complex to the fliA promoter region in X. nematophila, thereby promoting fliAZ transcription.

FliZ is an autogenous activator controlling the transcription of fliAZ (class II genes) through the activation of the flhDC (class I genes)

To further examine the function of FliZ in the regulatory cascade, we conducted experiments to evaluate the effect of FliZ on fliAZ expression. The introduction of a plasmid containing fliZ under the control of Plac into a polar fliA mutant increased the expression of P2fliA–uidA fusion genes by a factor of 13 (Fig. 3B), while no increase of the activity of P1fliA alone was detected (data not shown). As expected, very low levels of expression of the P2fliA–uidA fusion were detected in the presence or absence of FliZ in the flhD background (Fig. 3B). Thus, expression of the fliAZ operon involved strong positive autoregulation by FliZ, via its FlhDC-dependent promoter. In order to test the hypothesis that FliZ plays a role in flhDC expression, we constructed an flhD::uidA transcriptional fusion and introduced it, in the presence or absence of FliZ, in the flhD strain (Fig. 4A). We observed that the flhD strains containing FliZ (pKZ4) have a 2.7-fold increase in flhDC transcription, as compared with the strains containing the control plasmid or pKA7 (fliA gene alone). Taken together, these data indicate that FliZ acts as a positive regulator of flhDC (class I genes) expression and that, in turn, flhDC products control fliAZ transcription in X. nematophila.

Figure 4.

fliZ is an autogenous activator.
A. Activation of the promoter region of the flhD gene by FliZ. The PflhD–uidA fusion was transferred into ΩflhD (pRK404), ΩflhD (pKA7) and ΩflhD (pKZ4). GUS activity was assayed on bacteria grown in mot broth at an OD540 of 0.5. Specific GUS activity was calculated as GUS activity per unit optical density. The mean and SD were calculated from the data for three independent cultures.
B. In vivo promoter binding to the FliZ protein. ChIP assays with FliZ–HA were performed. Enrichment with the flhD and fliA promoter regions was analysed in the wild-type and flhD backgrounds and compared with that of the mreB internal region (non-target DNA). The relative values of enrichment for the indicated promoter were normalized with respect to a control region in the gyrB gene. The data represent the means of three independent cultures at an OD540 of 0.4.

Further investigations were conducted to know whether the FliZ protein binds directly to the flhD promoter in X. nematophila by carrying out ChIP assays. A FliZ recombinant protein fused to an HA tag at the C-terminus was placed in trans into the flhD mutant. Quantitative PCR showed that the immunoprecipitated fraction displayed a fivefold specific enrichment of the flhD promoter DNA region with respect to the non-target DNA regions upstream from mreB (Fig. 4B). The ChlP experiments also revealed that no fliA promoter fragment enrichment was observed in the HA-immunoprecipitated DNA fraction, as shown by comparison with the total DNA fraction (Fig. 4B). Thus, the FliZ protein of X. nematophila binds directly to the flhDC promoter, but does not belong to the FlhD4C2/DNA complex activating the fliAZ operon transcription.

FliZ is the checkpoint between haemolysin production and motility in X. nematophila

We have previously shown that FlhDC is required for haemolysin and lipase production, in addition to motility, in X. nematophila (Givaudan and Lanois, 2000). It is therefore possible that the production of lipase and haemolysin depends on FliA or FliZ. We investigated this possibility by comparing the phenotypic characteristics of the wild type and flhD and fliA mutants. Like the flhD strain, the fliA mutant differed from the parental strain in three main phenotypic aspects: (i) it lacked swimming motility in soft agar, (ii) no haemolysis reaction was observed on sheep blood agar and (iii) no lipolytic activity was observed on Tween agar (Table 1). We investigated whether FliA or FliZ alone, or both proteins were involved in these phenotypic defects, by carrying out a phenotypic analysis of the polar omega-insertion fliA mutant complemented with a low-copy-number mobilizable plasmid containing fliA or fliZ. As expected, the motility of the polar fliA mutant strain in soft mot agar was completely restored by introducing pAL3512, which carried the fliAZ genes, but only partially when complemented with pAL3511 (fliA gene alone) (Table 1). It is likely that the difference of motility between the two complemented strains is due to the feedback control of FliZ on the flhDC operon expression. In contrast, motility was not restored with pAL3513 (fliZ gene alone) (Table 1). The Tween lipase activity of the fliA strain was fully complemented by plasmids carrying fliA, but the mutant displayed a haemolytic phenotype after introduction of the fliZ-containing plasmid in trans (Table 1). FliZ overproduction also led to the recovery of a haemolytic phenotype in the flhD background, indicating that FliZ does not act in concert with FlhDC. Plate assays of haemolysis are not sufficient to determine the nature of the active haemolysin. We therefore used liquid haemolytic assays for this purpose. We have previously shown that XaxAB haemolysin is fully responsible for the early extracellular cytolytic activity (C1 activity; Brillard et al., 2001) observed against immunocompetent cells and SRBC (Vigneux et al., 2007). Determination of the XaxAB titre in early extracellular supernatants revealed that the fliA strain expressing fliZ had a wild-type phenotype for XaxAB activity (Table 1). These findings indicate that swimming motility and Tween lipase activity are controlled by the σ28 factor, FliA, whereas the haemolysin activity of XaxAB is FliZ-dependent and does not require σ28 factor activity.

Table 1.  Phenotypic characteristics of fliA and flhD mutants with and without complementation with fliA, fliZ and fliAZ expressing plasmids. a
StrainDescriptionChemotaxisbSheep blood haemolysiscHU titredLipolysis of Tween 20 and 80eBtb adsorptionfAntibiotic productiong
  • a. 

    All plates incubated for 2 days at 28°C before assays were interpreted unless otherwise indicated.

  • b. 

    Chemotaxis halo: ++, large spreading area (> 20 mm); +, 10 mm;

  • −, no spreading area.

  • c. 

    T, total haemolysis; P, partial haemolysis; v (variable); partial or annular haemolysis;

  • −, no haemolysis.

  • d. 

    XaxAB haemolytic unit (see Experimental procedures for titre calculation). ND, not done. Supernatants from 20 h cultures were collected and assayed for haemolysis, using SRBC.

  • e. 

    Halo of precipitation: ++, diameter > 6 mm; +, 4 mm;

  • −, no halo.

  • f. 

    Btb, bromothymol blue; B, dark blue colonies; N, no adsorption of Btb.

  • g. 

    +, Zone of growth inhibition of Micrococcus luteus (laboratory collection).

F1Wild type++T0.043+B+
ΩflhDflhD mutantV< 6 × 10−3B+
ΩfliAfliAZ mutant< 6 × 10−3B+
ΩfliA (pAL3512)pAL3512 carries fliAZ++P0.07++N+
ΩfliA (pAL3511)pAL3511 carries fliA only+< 6 × 10−3+B+
ΩfliA (pKA7)pKA7 carries Plac-fliA+ND+B+
ΩfliA (pAL3513)pAL3513 carries fliZ onlyP0.044B+
ΩfliA (pKZ4)pKZ4 carries Plac-fliZTNDB+
ΩflhD (pKZ4)pKZ4 carries Plac-fliZTNDB+
ΩflhD (pBBZ5)pBBZ5 carries Plac-fliZTNDB+

FliZ is directly involved in the expression of two haemolysin-encoding operons, xaxAB and xhlBA

As the haemolysin activity of XaxAB was found to be FliZ-dependent in this study, we checked the involvement of FliZ in xaxAB transcriptional expression. We first confirmed that xaxA and xaxB were co-transcribed as an operon, by RT-PCR analysis (data not shown). We investigated the participation of FliZ in xaxAB transcription, by assessing the expression of a PxaxA–uidA construct in different chromosomal backgrounds. β-Glucuronidase activity levels in both the polar flhD and fliA mutants were less than half those in the wild type (Fig. S2), indicating that expression of the haemolysin operon is controlled by regulators of the flagellar regulon. As the presence of FliZ alone is sufficient for the haemolysis of sheep blood cells (see above), we assessed the levels of xax transcripts in both mutants in the presence of the fliA and fliZ genes, placed under the control of Plac in the fliA and flhD backgrounds. The presence of FliZ alone resulted in wild-type levels of xax transcript in both mutants with impairment of the major flagellar regulators, whereas the presence of FliA had no significant effect on xax expression (Fig. S2). The ratio between the amount of xax transcripts in flagellar mutants harbouring pKZ4 and that in mutants harbouring pRK404 (plasmid control) was 4.4 for fliA mutants and 7 for flhD mutants. This indicates that activation of the xaxAB gene by FliZ occurs even in the absence of FlhD/FlhC. Quantitative RT-PCR (qRT-PCR) analysis showed that the ratio between the amount of xax transcripts in flhD mutants harbouring pKZ4 and pRK404 (plasmid control) was 62, providing further evidence that xaxAB may be activated by FliZ even in the absence of FlhD/FlhC (Fig. 5A). We also carried out the ChIP assay with the wild-type strain producing the FlhD–HA translational fusion. As expected, quantitative PCR with a primer specific for the region upstream from xaxAB showed that the immunoprecipitated DNA fraction was not enriched in this region, demonstrating that FlhD is not involved in the activation of xaxAB operon transcription (data not shown).

Figure 5.

Role of FliZ in the expression of haemolysin genes.
A. Effect of FliZ on xaxAB and xhlBA transcript levels. Real-time RT-PCR was carried out for xaxAB and xhlBA, using total RNA from stationary-phase cultures (OD540 of 4.5) of wild-type, ΩflhD (pRK404) and ΩflhD (pKZ4) strains and specific internal primers for each gene. Quantitative PCR was performed in triplicate and data are presented as a ratio, with gyrB used as the control gene (95% confidence limits).
B. FliZ binding to xaxAB and xhlBA promoters. ChIP assays were performed on cultures (OD540 of 1.2) of the F1 (FliZ–HA) and ΩflhD (FliZ–HA) strains, using anti-HA antibody. The regions upstream of xaxA and xhlB were immunoprecipitated with FliZ–HA. The upstream fliA and internal mreB regions were used as non-target DNA. The relative values of enrichment for the indicated promoter were normalized with respect to a control region in the gyrB gene. The mean and SD were calculated from the data for three independent cultures.

A cell surface-associated haemolysin, XhlA, that lyses the two most prevalent types of insect immune cells and rabbit and horse erythrocytes was recently described (Cowles and Goodrich-Blair, 2005). As the xhlBA operon was found to be FlhDC-dependent in X. nematophila (Cowles and Goodrich-Blair, 2005), we carried out qRT-PCR analysis with the same RNA extracts and primers specific for xhlB and xhlA. Figure 5A shows that expression of FliZ in the flhD background resulted in a 10-fold increase in the levels of xhlA and xhlB transcripts as compared with the flhD strain containing the control plasmid. In contrast, FliA alone had no effect on xhlBA transcript levels (data not shown). Thus, FliZ controls the expression of both haemolysin operons xaxAB and xhlBA at the transcriptional level.

We carried out ChIP experiments to assess the binding of FliZ protein to haemolysin gene promoter regions, using wild-type and flhD strains producing the recombinant FliZ–HA protein. Quantitative PCR on DNA immunoprecipitated with the anti-HA antibody, using a primer specific for the promoter region of xaxA and xhlB, showed substantial enrichment of these DNA regions linked to FliZ–HA (about eightfold) with respect to the non-target DNA regions upstream from fliA and mreB (Fig. 5B). Thus, FliZ belongs to the transcriptional complex linked to the promoter of both haemolytic operons in X. nematophila.

FliZ-dependent genes are involved in insect virulence

We investigated the contribution of various components of the X. nematophila flagellar regulon to virulence during insect infection, by comparing the pathogenicity of different bacterial mutants. We did this by monitoring larval mortality over a 3-day period, after injection into our insect model, the lepidopteran Spodoptera littoralis. At least four independent experiments with one dose [300–500 colony-forming units (cfu)] of bacterial cells in the exponential growth phase were analysed using the Statistical Package for Social Science version 11.0.1 (SPSS) to compare individual survival times for each strain. Living cells of wild-type X. nematophila are highly virulent when injected into insect haemolymph. The calculated LT50 for S. littoralis was about 30 h. As previously shown, the flhD mutant displayed attenuated virulence in this insect (LT50 = 41 h) (data not shown).

Statistical analysis also demonstrated that the fliA mutant had an ability (P > 0.05) to kill insect larvae similar to that of the flhD mutant, with a higher LT50 at 11 h than the wild-type strain (P < 0.001) (Fig. 6). In control experiments, we first demonstrated that the presence of the control plasmid (pBBR1K) had no significant effect on the virulence of the fliA mutant and the wild-type strain (P > 0.05). We investigated whether FliA-dependent or FliZ-dependent genes were involved in the virulence defects, by comparing the LT50 of fliA mutant strains complemented with plasmids containing fliA (pAL3511) or fliZ (pAL3513) with the fliA mutant strain (Fig. 6). The pathogenicity of the polar fliA mutant carrying the plasmid encoding FliA was not significantly different from that of the fliA mutant (P > 0.05). Thus, the FliA-dependent phenotypes of motility and lipase activity are not essential for the full virulence phenotype of X. nematophila. Unlike FliA, which seems to have no effect on Xenorhabdus virulence, FliZ was found to be required for full virulence in insects (Fig. 6). Indeed, statistical analysis revealed that fliA strains harbouring a fliZ-expressing plasmid were as able to kill insects as the wild-type strain (P > 0.05). To strengthen the view of FliZ, rather than FlhDC, as the main regulator of virulence in X. nematophila, we examined whether the presence of the fliZ gene placed under the control of Plac in the flhD background could restore the full virulence in S. littoralis. To perform in vivo infection assays, the use of a stable plasmid is required. As pRK404 derivatives are not stable without antibiotic pressure, we placed fliZ under the Plac control of pBBR1 derivatives to yield pBBZ5. As expected, the flhD mutant harbouring pBBZ5 displayed the same phenotypes as flhD (pKZ4) strain (Table 1). Pathology assays showed that flhD strains harbouring a fliZ-expressing plasmid displayed an even stronger ability to kill insects, with a lower LT50 value (about 3 h early), in comparison with the wild-type strain (P < 0.05) (data not shown). Taken together, the pathology assays showed that one or several virulence genes from the FliZ regulon contribute to the virulence of X. nematophila against S. littoralis larvae.

Figure 6.

fliZ is necessary for full virulence in S. littoralis. Insect mortality was monitored over time, after the injection of wild-type X. nematophila (black diamonds), fliA (open triangles), fliA (pAL3511): a fliA-complemented strain (black squares); fliA (pAL3513): a fliZ-complemented strain (open circles) and xaxAB9: a xaxAB null mutant (open squares). The mortality values are based on data obtained after the injection of 300–500 cfu into 20 insect larvae. The mortality curves shown are from a representative experiment (n = 4).

Having demonstrated that the xaxAB operon is directly regulated by FliZ, we investigated whether the XaxAB toxin was required for full virulence. We compared insect larval mortality after infection with xaxAB9 (a xaxAB mutant; Vigneux et al., 2007), flhD (pBBxaxAB) mutant or a fliA (pBBxaxAB) mutant constitutively producing the XaxAB haemolysin with that for the wild-type strain. The xaxAB9 mutant displayed full virulence (Fig. 6), whereas both the fliA and flhD mutants expressing xaxAB displayed the same attenuated virulence phenotype as the polar fliA mutant containing the control plasmid (P > 0.05) (data not shown). These data suggest that the XaxAB toxin is not essential for the full virulence of X. nematophila in S. littoralis.


We show here that FliZ is the key regulator coupling motility and virulence in the insect pathogenic bacterium, X. nematophila. The fliZ gene is the second of the two genes of the fliA operon, both of which are class II flagellar genes. Like other flagellar genes, fliZ was identified not based on its effects on motility, but based on its transcription from flagellar promoters and its location in an operon known to contain flagellar genes.

Similarities in the flagellar hierarchy between E. coli and X. nematophila include the identification of FlhD/FlhC as the master regulator (Bartlett et al., 1988; Givaudan and Lanois, 2000), FliA as the sigma factor (Ohnishi et al., 1990; Park and Forst, 2006) and structural flagellar genes, such as fliC and fliD (Givaudan et al., 1996). However, there are a number of differences between the flagellar systems of E. coli and X. nematophila. In E. coli and S. enterica serovar Typhimurium, the fliA, fliZ and fliY genes are organized into a single operon (Mytelka and Chamberlin, 1996; Ikebe et al., 1999). A different gene organization, in which fliA and fliZ are the sole members of the corresponding operon, is found in X. nematophila, but also in Y. enterocolitica (Kapatral et al., 2004), P. mirabilis, Serratia marcescens (http://www.sanger.ac.uk/Projects/Microbes/) and P. luminescens (Duchaud et al., 2003). A second major difference concerns the regulatory elements in the fliA promoter region. E. coli and Salmonella species have two promoters in this region: a class II promoter, which is recognized by σ70 RNA polymerase in the presence of the FlhD and FlhC activator proteins, and a class III promoter, which is recognized by the FliA RNA polymerase (Liu and Matsumura, 1994; Ikebe et al., 1999). In addition, the presence of multicopy fliZ genes increases fliA transcription from its class II promoter in S. enterica serovar Typhimurium, and these effects are FlhDC-dependent (Kutsukake et al., 1999). We show here that the fliA operon in Xenorhabdus is mainly controlled by FlhDC. However, unlike the corresponding proteins in E. coli and Salmonella, X. nematophila FliA has no effect on transcription of its own gene. Consequently, the absence of FliZ reduced the class II promoter of fliA activity by a factor of more than 10 (Fig. 3B) in X. nematophila, whereas it decreased that in S. enterica serovar Typhimurium by a factor of only three (Kutsukake et al., 1999). These data reveal the significant role of FliZ in the Xenorhabdus flagellar regulon. As in X. nematophila, the putative promoter sequences of fliA from P. luminescens, Y. enterocolitica, S. marcescens and P. mirabilis did not contain any consensus for a FliA-dependent class III promoter (Fig. S1), suggesting that other enterobacteria share a regulatory network similar to that proposed for X. nematophila.

As yet, the precise molecular mechanisms by which FliZ modulates class II promoter remain unknown in enterobacteria. We have demonstrated, for the first time, that FliZ cannot bind in vivo to the fliA promoter in X. nematophila (Fig. 4B). Other regulatory proteins are known to regulate class II expression in Salmonella through specific protein–protein interactions. It has been reported that the ClpXP protease and FliT, which belongs to the flagellar regulon, function as negative regulators, modifying FlhD4C2 turnover and inhibiting FlhD4C2 binding to the class II promoter respectively (Kutsukake et al., 1999; Tomoyasu et al., 2003; Yamamoto and Kutsukake, 2006). Like FliZ, the DnaK chaperone acts as a positive regulator for class II operons. DnaK is thought to be involved in the conversion of FlhD4C2 into a stable form able to bind to class II promoters to form an FlhD4C2/DNA complex (Takaya et al., 2006). Our ChIP experiments, conducted in vivo with Xenorhabdus cells grown to mid-exponential growth phase, showed that the FlhD subunit belongs to the transcriptional complex, whereas FliZ does not immunoprecipitate the protein complex associated with the upstream region of fliA. As ChIP provides snapshots of protein–protein and protein–DNA interactions at a particular point time, complementary ChIP experiments were performed, using bacterial cells in different growth phases. Whatever the growth phase studied, no binding of FliZ to the upstream region of fliA was detected (A. Lanois, unpubl. data). Alternatively, it has been reported that both in S. enterica and in enterohaemorrhagic E. coli (EHEC), FliA was able to upregulate flhDC transcription, giving rise to a feedback regulatory loop (Kutsukake, 1997; Clarke and Sperandio, 2005). We therefore investigated the option that FliZ directly controls the expression of flhDC in X. nematophila. Our ChIP experiments and reporter gene assays revealed for the first time that FliZ is involved in the positive feedback regulation of fliAZ transcription by a direct transcriptional activation of the master operon flhDC. While FliA is involved in the feedback loop between the class II genes (fliAZ) and the class I genes (flhDC) in EHEC (Clarke and Sperandio, 2005), we demonstrated here that FliZ is the key protein of the feedback regulation in the flagellar regulon in X. nematophila (Fig. 7).

Figure 7.

Model depicting the role of X. nematophila FliZ in the co-ordinate regulation of motility, virulence and haemolysin gene expression. The master activator of the flagellar cascade, FlhD4C2, directly controls the expression of the fliAZ operon. FliA (the σ28 factor) controls motility and lipase production. The FliA-regulated lipase gene, xlpA, has previously been identified (Park and Forst, 2006). FliZ is involved in autogenous control of fliAZ expression through a retroactive control of flhDC transcription. FliZ directly controls expression of the xhlBA and xaxAB operons which encode haemolysins and is also required for full virulence towards S. littoralis insects. See text for more details.

In addition to its role in swimming motility, FlhDC has been shown to regulate a number of non-flagellar promoters in Escherichia coli and other enteric bacteria (Pruss et al., 2003; Kapatral et al., 2004; Stafford et al., 2005; Horne and Pruss, 2006). However, it remains unclear whether FlhDC-dependent regulation is determined directly by FlhD4C2 binding to non-flagellar gene promoters or whether it reflects indirect pathways acting through FliA, FliZ or other regulators. We previously showed that FlhDC positively regulates swarming behaviour, haemolysin and lipase production, and is required for full virulence in X. nematophila (Givaudan and Lanois, 2000). The transcription of flhDC is negatively influenced by the two-component regulator EnvZ/OmpR (Park and Forst, 2006) and is positively regulated by the global regulator Lrp (A. Lanois and A. Givaudan, unpubl. data; Goodrich-Blair and Clarke, 2007). Mutations in the genes encoding these regulators have been shown to affect swimming and swarming motilities, lipase and protease production and haemolysis. The regulatory pathway controlling these phenotypes has been identified, but the hierarchy of these factors remains unclear. In this study, we showed that FliA co-ordinates the expression of lipase activity and flagellar motility, whereas FliZ controls xaxAB and xhlAB haemolysin expression and virulence in insects (Fig. 7). Two non-flagellar FliA-regulated genes, xlpA and xrtA, encoding a lipase and a protease, were recently characterized in X. nematophila (Park and Forst, 2006). The lack of lipase activity in our fliA strain therefore probably results from the lack of expression of the xlpA gene (Fig. 7). However, the disruption of fliA affected the lipase activity observed with Tween 20 to Tween 80 as substrates (Table 1), whereas xlpA mutant strains retained Tween 80 lipase activity (Park and Forst, 2006), suggesting that other FliA-regulated genes encoding lipolytic enzymes are present in the X. nematophila genome.

Our study in Xenorhabdus reveals for the first time the existence of a FliZ regulon, which consists of genes directly regulated by FliZ such as flhDC, but also xaxAB and xhlBA. MEME motif discovery tool (http://meme.sdsc.edu/) was used to search for FliZ binding sites upstream of FliZ-dependent genes of X. nematophila. No consensus sequences were found, but these regions are characterized by numerous AT-rich sequences which suggest that FliZ may be a transcriptional activator binding to AT-rich regions rather than binding to a defined DNA sequence motif. Structural compatibility between the C-terminal domain of FliZ and the recombinase XerD suggests that this domain may mediate the DNA-binding activity of FliZ. Moreover, as both direct targets of FliZ encode haemolysins, the FliZ regulon may contain a specific functional class of genes. Our transcriptional analysis also revealed that the haemolytic genes of the FliZ regulon could be expressed in an FlhDC-independent manner. However, no significant transcription of fliZ was detected in an flhD background in classic bacterial growth conditions (Fig. 3A). That may suggest that the regulation of genes encoding FliA-dependent and FliZ-dependent phenotypes (motility, lipase activity and haemolysin production respectively) in response to environmental cues is integrated through pathways that affect flhDC expression. Additional studies are required to determine if the coupling of motility and haemolysin expression in X. nematophila also occurs during insect infection.

The two directly FliZ-dependent genes, xaxAB and xhlBA, encode haemolysins from two different families. XaxAB is the prototype of a new extensive family of haemolysins with apoptotic and pore-forming activities in mammalian and invertebrate cells. xaxAB homologues are found in various enterobacteria and pseudomonads, such the insect pathogenic bacteria, P. luminescens and P. entomophila, the plant pathogen P. syringae and the human pathogens Y. enterocolitica, Photorhabdus asymbiotica and P. mirabilis (Vigneux et al., 2007). The function of Xax homologues has been established only in X. nematophila. However, a recent microarray analysis in Y. enterocolitica identified xaxA and xaxB homologues (YE1984 and YE1985 respectively) as a RovA-regulated locus (Cathelyn et al., 2007). RovA is a member of the MarR/SlyA family and it has been defined as a major transcriptional activator of Yersinia invasin. It has been proposed that RovA acts as an anti-H-NS gene activator (Cathelyn et al., 2007). The second FliZ-dependent haemolysin, XhlA, belongs to a family characterized by a two-partner secretion system (TPSS). The first member of the TPSS family to be characterized was the ShlB protein of S. marcescens, which exports the ShlA haemolysin from the periplasm of the Gram-negative bacterial envelope into the external medium (Poole et al., 1988). Several ShlBA homologues have been functionally characterized in pathogenic bacteria, such as P. luminescens, P. mirabilis, Haemophilus ducreyi, and the fish pathogens Edwardsiella tarda (for synthesis, Hertle, 2005) and Yersinia ruckeri (Fernandez et al., 2007). These pore-forming toxins of the TPSS family are generally induced under iron-starvation conditions (Hertle, 2005). Direct binding of the Lrp regulatory protein to the xhlBA promoter region has been shown in X. nematophila (Cowles et al., 2007). FliZ may be one of the main positive regulators of the Xax and TPSS haemolysins in all of these pathogenic bacteria. For instance, the HpmA haemolysin toxin (TPSS family) of P. mirabilis, encoded by the hpmBA locus, is upregulated co-ordinately with the synthesis and assembly of flagella during differentiation into hyperflagellated swarm cells (Fraser et al., 2002). Fraser et al. (2002) also showed that Lrp directly regulates hpmBA transcription, whereas FlhD4C2 acts indirectly and the hpmBA operon is transcribed from a σ70 promoter, rather than a putative σ28 promoter. FliZ therefore remains the main candidate linking HpmA haemolysin to swarming motility in P. mirabilis. Moreover, it has been reported that FliZ positively regulates the invasion genes encoded on the SPI-1 locus in S. enterica. Indeed, FliZ controls the expression of hilA, a positive regulator of SPI-1 genes (Lucas et al., 2000; Iyoda et al., 2001). As yet, the molecular mechanism by which FliZ activates Salmonella invasion genes remains unknown. As observed for FliZ-dependent haemolysin genes in Xenorhabdus, the hilA gene is activated by FliZ even in the absence of FlhC in Salmonella (Lucas et al., 2000). Because we have shown that FliZ is able to bind directly to promoter regions in the absence of FlhDC, it is likely that FliZ directly activates regulatory gene(s) of SpI1 expression, linking the flagellar and invasion systems in Salmonella.

Xenorhabdus is unusual in that it is a pathogenic bacterium without a TTSS (Brugirard-Ricaud et al., 2004). It therefore displays no interplay between flagellar genes and TTSS-encoded genes. This suggests that virulence factors may be secreted through the flagellar export apparatus in Xenorhabdus. The lipase gene of X. nematophila, xlpA, has recently been identified (Park and Forst, 2006) as a homologue of yplA, the FliA-dependent phospholipase gene of Y. enterocolitica (Young et al., 1999). As previously shown for the virulence-associated phospholipase YplA, the secretion of lipase XlpA in Xenorhabdus is dependent on a functional flagellar export apparatus (Park and Forst, 2006). However, another FliA-dependent enzyme, the protease XrtA, is secreted by a dedicated ABC transport system (Park and Forst, 2006). It has also been shown that an flhA mutant of X. nematophila impaired in flagellar export systematically displays a haemolytic phenotype on sheep blood agar plates (Park and Forst, 2006). We used the flhD strain as a flagellar export mutant, to confirm that active XaxAB haemolysin is not secreted by the flagellar export system. The expression of the xaxAB genes under the control of Plac in the flhD genomic context resulted in larger amounts of active haemolysin, with the characteristics of XaxAB, in the culture supernatant (A. Givaudan, unpubl. data). Thus, XaxAB secretion is not dependent on the flagellar export system. We are currently investigating the precise molecular mechanisms underlying the secretion of both components of this new binary toxin. However, XaxAB is clearly not a member of the TPSS family and the xaxAB operon is not flanked by genes encoding a dedicated ABC transport system (Vigneux et al., 2007). The FliZ-regulated XhlA haemolysin is probably secreted by a type V secretion system, the TPSS system (Cowles and Goodrich-Blair, 2005), even if this has not been functionally demonstrated in Xenorhabdus. These data demonstrate that the non-flagellar FliA and FliZ-dependent proteins of Xenorhabdus are secreted via different pathways.

Complex interplay between flagellar motility and bacterial virulence has been described in many bacteria pathogenic in mammals (for synthesis, Josenhans and Suerbaum, 2002). Insects are thought to be a useful model system for the testing of pathogenicity in vivo. Wild-type X. nematophila has been shown to be highly virulent when injected directly in the haemolymph of various insect larvae. We have shown that flhDC and fliAZ mutant strains of X. nematophila display a delayed virulence phenotype in the lepidopteran S. littoralis and Galleria mellonella and the locust, Locusta migratoria (this study; A. Givaudan unpubl. data; Givaudan and Lanois, 2000). We have also shown that FliA-dependent properties, such as motility and lipase production, are not involved in the pathogenicity of X. nematophila towards insects (Fig. 7). Non-motile fliC mutants (which lack the flagellin subunit) have been shown to display wild-type virulence (Herbert and Goodrich-Blair, 2007). Thus, motility per se is not required for the virulence of X. nematophila following injection into insect larvae. Finally, we show here that only the FliZ regulon contributes to virulence during insect infection by X. nematophila.

When introduced by injection or released by their nematode vector, Xenorhabdus cells grow within the body of the insect. In vivo studies of the process of bacterial infection in insect hosts showed that X. nematophila is extracellular in the haemolymph (Sicard et al., 2004). As cellular immunity is induced by the penetration of a foreign body, such as a bacterium, into the insect haemocoel, Xenorhabdus must therefore be able to escape the immune response. Haemocytes are the immune cells of insects, and cellular immune reactions against bacteria involve haemocytes of several different lineages. In Lepidoptera, plasmatocytes (Pl) build nodules that isolate clumps of bacteria and necrotic insect tissues, and granular haemocytes 1 (GH1) are the professional phagocytes (Ribeiro and Brehelin, 2006). Both the FliZ-dependent haemolysins targeting these immune cells, XhlA and XaxAB, are good candidates for involvement in immunosuppression. The FliZ-dependent haemolysin XaxAB strongly induced necrosis and apoptosis in insect immunocompetent cells in vitro (Vigneux et al., 2007), but the xaxAB mutant was shown to be as virulent as wild-type X. nematophila (this study). Furthermore, expression of the xaxAB genes under the control of the Plac promoter in a fliAZ or an flhD background did not restore full virulence to these strains. Thus, XaxAB is not required for the full virulence of X. nematophila. It is noteworthy that the xaxAB9 does not necessarily have early extracellular cytolytic activity (called C1 activity) (Brillard et al., 2001) against immunocompetent cells, but it systematically induces haemocyte apoptosis in insects (Vigneux et al., 2007). Haemocytes apoptosis ability may therefore explain the full virulence of xaxAB mutant. Regarding the cell surface-associated haemolysin XhlA, it has been shown to be required for the full virulence of X. nematophila in another lepidopteran, Manduca sexta (Cowles and Goodrich-Blair, 2005). As xhlBA transcript levels were low in the absence of FliZ, at 10% wild-type levels, the attenuated virulence of fliAZ and flhDC mutant strains in insects may result from lower levels of XhlA production during S. littoralis infection. To strengthen this hypothesis, we evaluated the virulence of the xhlA mutant using our insect model, S. littoralis. As compared with the wild-type strain, an attenuated virulence phenotype for xhlA mutant was also found after injection into S. littoralis (A. Givaudan, unpubl. data). However, xhlA mutant has been reported to have cytolytic activity against insect haemocytes and mammalian blood erythrocytes similar to that of wild-type X. nematophila (Cowles and Goodrich-Blair, 2005). Thus, the cellular targets of XhlA haemolysin during insect infection remain unclear in X. nematophila. In addition, it may seem surprising that both haemolysins, XaxAB and XlhA, belong to the same regulatory network and have redundant cellular target specificity. Future studies will therefore focus on host cell cytotoxicity and timing of the expression of FliZ-dependent genes encoding haemolysins during insect infection with X. nematophila, making it possible to determine the role of each haemolytic factor during insect infection. Moreover, it is likely that other unknown genes of the X. nematophila FliZ regulon are also involved in full virulence in insects.

In conclusion, our data show that not all enteric bacteria have a flagellar gene network similar to that described in E. coli and S. enterica. Both FliZ and FliA display positive autoregulation of expression of the fliAZY operon in E. coli, whereas only FliZ can activate the FlhDC-dependent promoter of fliAZ in Xenorhabdus. We also demonstrated that FliZ directly regulated the expression of two haemolysin operons (Fig. 7). FliZ is, thus, a key factor for the co-ordination of flagellar motility and the expression of several virulence factors in X. nematophila. The FliZ regulatory pathway probably occurs in other enteric bacteria, such as P. luminescens, Y. enterocolitica, S. marcescens and P. mirabilis, also harbouring loci encoding XaxAB and TPSS haemolysin homologues in their genomes.

Experimental procedures

Bacterial strains, plasmids and growth conditions

The strains and plasmids used in this study and their sources are listed in Table 2. Bacteria were routinely grown in Luria–Bertani (LB) medium or mot broth (1% tryptone, 0.5% NaCl, 10 mM MgSO4) at 28°C for X. nematophila and at 37°C for E. coli. Swimming motility was observed in mot agar (mot broth with 0.35% agar) at 28°C. When required, the final concentrations of the antibiotics used for selection were as follows: ampicillin, 100 mg l−1 for E. coli strains and 50 mg l−1 for X. nematophila; kanamycin, 20 mg l−1; gentamicin, 30 mg l−1; chloramphenicol, 20 mg l−1 for E. coli strains and 15 mg l−1 for X. nematophila and tetracycline, 10 mg l−1 for E. coli strains and 7.5 mg l−1 for X. nematophila.

Table 2.  Bacterial strains and plasmids used in this study:
Strain or plasmidGenotype or relevant characteristicsSource or reference
E. coli strain
 DH5αMCRF-mcrA Δ(mrr-hsdRMS-mcrBC) φ80ΔlacZΔM15 Δ(lacZYA-argF)U169 endA1 recA1 deoR thi-1 supE44 lambda- gyrA96 relA1Laboratory collection
 YK4104YK 410 fliA- (spontaneous mutant)Matsumara
YK410 = F-araD139 ΔlacU169 rpsL thi nalA thyA pyrC46 hi 
 S17.1pro r- n- TpRSmR RP4-2-Tc::Mu::Tn7 recA thiSimon et al. (1983)
X. nematophila strain
 F1Wild type isolated from Steinernema carpocapsae nematode, Plougastel (Brittany)Laboratory collection
 ΩflhDF1 flhD::ΩCmGivaudan and Lanois (2000)
 ΩfliAF1 fliA::ΩCmThis work
 xaxAB9F1 xaxAB::ΩCmVigneux et al. (2007)
 pUC19ApR cloning vehicleBiolabs
 pAL315 kb Sau3A insert from F1 containing fliD-C-A-Z-putA cloned into pUC19/BamHIThis work
 pAL30112.8 kb HindIII fragment from pAL3 containing ext 3′fliC-fliA-Z-ext 3′putA cloned into pUC19/HindIIIThis work
 pAL30122 kb BstEII–ApaI blunt end fragment containing fliA-Z-ext 3′putA into pUC19 HincII siteThis work
 pJQ200KSGmrsacRB mob oriV (p15A replicon)S. Forst
 pHP45-ΩCmApr Cmr interposon ΩCmFellay et al. (1987)
 pAL3012-ΩCm3.5 kb BamHI blunt ΩCm fragment from pHP45-ΩCm into pAL3012 HincII siteThis work
 pJQ-fliA-ΩCm5.5 kb BamHI–PstI fragment from pAL3012-ΩCm into pJQ200KS/BamHI–PstIthis work
 pBBR1KKmrmobS. Kohler
 pAL3511PCR-amplified fragment using oli3 and fliA/Z primers cloned into pBBR1KThis work
(contains fliA gene only) 
 pAL35122 kb fliAZ fragment from pAL3012 cloned into pBBR1K/Asp718I–BamHIThis work
 pAL3513600 bp AccI–SacI deletion from pAL3512 (contains fliZ gene only)This work
 pUIDK1uidA Kmr AprBardonnet and Blanco (1992)
 pBBR1-MCS1broad host range vector CmrmobKovach et al. (1995)
 pBBR1-MCS2broad host range vector KmrmobKovach et al. (1995)
 pBBR1-MCS5broad host range vector GmrmobKovach et al. (1995)
 pBBZ5SacI–ApaI fliZ fragment into pBBR1-MCS5 (under Plac promoter control)This work
 pBB-UIDKuidA Kmr cassette (HindIII–SmaI) from pUIDK1 cloned into pBBR1-MCS1 lacking the Cm resistance gene and T3, T7 and β-galactosidase promotersThis work
 pAL72200 bp PCR fragment using amt5′ and amt3′ primers cloned into pBB-UIDK (contains P1fliA promoter)This work
 pAL73200 bp PCR fliZ2-fliApext blunt end fragment cloned into pBB-UIDK/SmaI (contains P2fliA promoter)This work
 pAL74300 bp HincII–HaeIII fragment cloned into pBB-UIDK/SmaI (contains IR3fliZ region)This work
 pAL75580 bp PCR fragment cloned into pBB-UIDK/SacII–XbaI (contains PflhD promoter)This work
 pXABU350 bp PCR fragment using xax-uidA-f and xax-uidA-r primers cloned into SacII–XbaI sites of pBB-UIDK (contains PxaxAB region)This work
 pRK404lacZ mob TcR (mobilizable vehicle)Ditta et al. (1985)
 pKA7BstEII–HaeIII fliA fragment into pRK404 (under Plac promoter control)This work
 pKZ4SacI–ApaI fliZ fragment into pRK404 (under Plac promoter control)This work
 pKC2HpaIScaI flhC fragment into pRK404 (under Plac promoter control)This work
 pBBxaxAB2558 bp region overlapping xaxAB cloned into pBBR1MCS5 (GmR)Vigneux et al. (2007)
 pBB-FlhDFliZ–HApBBR1-MCS2 (KmR) containing FlhD–HAThis work
 pBB-FliZFliZ–HApBBR1-MCS2 (KmR) containing FliZ–HAThis work

Molecular genetic techniques, RNA preparation and Northern analysis

DNA manipulations were carried out as previously described (Ausubel et al., 1993). DNA plasmids were transferred into E. coli and X. nematophila by transformation and mating experiments respectively (Givaudan and Lanois, 2000). All constructs were sequenced by Millegen (Labège, France). The primers used in this study (Eurogentec) are described in Table S1. Total RNA was extracted with TRIzol reagent, according to the manufacturer's instructions (Invitrogen), and purified with the High Pure RNA Isolation kit (ROCHE), including a DNase I incubation. Concentration was assessed by determining absorbance at 260 nm. For each RNA preparation, DNA contamination was assessed by performing a control PCR reaction before RT-PCR analysis. Northern blot analysis was performed as previously described (Givaudan and Lanois, 2000). PCR fragments generated with R1 and fliA/Z primers were used as a probe for fliA, and primers oliZ5′ and oliZ3′ were used as a probe for fliZ.

Construction of the fliA mutant and of complemented strains

An omega cassette conferring resistance to chloramphenicol and with transcriptional and translational terminators was inserted into the HincII restriction site within the X. nematophila fliA gene of pAL3012 (2 kb BstEII–ApaI fliAZ fragment cloned into pUC19) to generate pAL3012-ΩCam. The 5.5 kb BamHI–PstI fragment harbouring fliA-ΩCm was then inserted into pJQ200KS to yield pJQ-fliA-ΩCm. The pJQ200KS plasmid is a derivative of pACYC184 carrying the sacB gene and the mob site from RP4. The pJQ-fliA-ΩCm plasmid (a sacB-negative selection plasmid) was used to transform the E. coli strain S17.1 and was introduced into X. nematophila F1 by mating. Cmr and Sacr exconjugants were selected on 4% sucrose and chloramphenicol LB agar. Omega insertion was confirmed by PCR analysis and the resulting clone was designated ΩfliA.

For ΩfliA complementation, a PCR fragment generated with the oli3 and fliA/Z primers and containing the fliA gene alone or the 2 kb Asp718–BamHI fliAZ fragment from pAL3012 was cloned into pBBRIK to yield pAL3511 and pAL3512 respectively. We constructed pAL3513, containing the fliZ gene alone, by deleting the 600 bp AccI–SacI fragment from pAL3512. For ΩflhD complementation, a 1.2 kb SacI–ApaI fliZ fragment from pAL3011 was cloned into pBBRI-MCS5 to yield pBBZ5.

Phenotypic characterization of the fliA mutant and complemented strains

Antibiotic production, bromothymol blue adsorption and extracellular lipase activity were assessed as previously described (Givaudan and Lanois, 2000). Swimming motility was observed in mot agar (mot broth complemented with 0.35% agar). Haemolytic activity was determined using blood agar plates and liquid haemolytic assay, as previously described (Brillard et al., 2001). Bacteria were grown on trypticase soy agar (bioMérieux) supplemented with 5% (v/v) defibrinated sheep blood (bioMérieux). Haemolysis was scored positive if a clearing halo was observed surrounding bacteria grown on standard sheep blood agar plates. The extracellular haemolysin was titrated in a liquid haemolytic assay. Briefly, SRBC and RRBC were obtained from BioMérieux (France) as a 50% suspension. They were thoroughly washed in phosphate-buffered saline (PBS) pH 7.2, and were diluted in this buffer to give a 5% suspension (v/v). Growing bacterial cells were harvested at various time points, for 5 days. Cells were centrifuged and subjected to ultrafiltration (Millipore, 0.2 μm) and extracts were then mixed with a suspension of erythrocytes washed in PBS (bioMérieux). The mixture was incubated at 37°C for 2 h. Intact cells and cell membranes were removed by centrifugation and the amount of haemoglobin released and present in the samples was determined by measuring A540. The titre of a XaxAB haemolysin solution can be calculated in haemolytic units (HU) from the absorbance value obtained, according to the following formula, deduced from numerous absorbance determinations with serial dilutions of haemolysin: Titre = 210(OD−0.72) (Vigneux et al., 2007).

Primer extension analysis

The fliApext primer was used for fliA primer extension and the fliZpext primer was used for fliZ primer extension. We carried out 5′ end labelling of 10 pmol of primer, with [γ-32P]-ATP and T4 polynucleotide kinase (New England Biolabs), at 37°C for 30 min. Labelled primer (1 pmol) and 30 μg of X. nematophila total RNA were added to AMV reverse transcriptase (Roche) buffer. The reaction mixture was incubated at 94°C for 3 min, cooled below −80°C and thawed on ice for 1 h. We then added 2 mM of each dNTP and 24 units of AMV reverse transcriptase (Roche) to the reaction mixture, which was then incubated at 45°C for 30 min. We then added 20 μg of RNase A and incubated at 37°C for 10 min. The reaction was stopped by adding 5 μl of sequence loading buffer. We subjected 8 μl of the extension reaction to electrophoresis in a 6% urea-polyacrylamide gel, which was then dried and autoradiographed.

RT-PCR and qRT-PCR analysis

RT-PCR was performed with the RT-PCR Access kit (Promega). After a second treatment with 2 U of DNase I (Promega), 100–500 ng of total RNA was mixed with 0.5 mM of each primer, denatured at 94°C for 2 min and then added to the reaction mixture (0.2 mM each dNTP, 1 mM MgCl2, 5 U AMVRT, 5 U Tfl polymerase in AMVRT/Tfl reaction buffer). The mixture was incubated at 49°C for 45 min and then subjected to 30 amplification cycles as follows: denaturation at 94°C for 30 s, at the annealing temperature determined by the primers Tm for 1 min and elongation at 68°C for 2.5 (min), followed by a final extension step at 68°C for 7 min. The RT-PCR products were then analysed by agarose electrophoresis.

We carried out qRT-PCR in two steps. First, cDNA was synthesized from 1 μg of total RNA from Xenorhabdus wild type, mutants and complemented strains, using the Super Script II Reverse Transcriptase from Invitrogen and random hexamers (100 ng μl−1) from Roche Diagnostics. We carried out quantitative PCR in triplicate, using the LightCycler FastStart DNA MasterPLUS SYBR Green I kit from Roche Diagnostics with 1 μl of cDNA synthesis mixture and 1 μM specific gene primers for xaxA, xaxB, xhlA and xhlB. The enzyme was activated by incubation for 10 min at 95°C. Reactions were performed in triplicate, using 45 cycles of 95°C for 5 s, 60°C for 5 s and 72°C for 10 s, with monitoring in a Light Cycler (Roche). Melting curves were analysed for each reaction and each curve contained a single peak.

The amount of PCR product was calculated from standard curves obtained by PCR with serially diluted genomic DNA from X. nematophila F1. All data are presented as a ratio, with gyrB used as the control gene (95% confidence limits).

Chromatin immunoprecipitation assay

For the production of FlhD–HA protein, the AspLflhD–HA primer (harbouring the ribosome binding site region of fliC and the first 12 codons of flhD sequence) and the BamRflhD–HA primer (designed to encode the HA sequence immediately upstream from the stop codon of the flhD gene) were used to amplify an flhD–HA PCR fragment from F1 genomic DNA.

The same strategy was used to generate a fliZ–HA PCR fragment, using the AspLfliZ–HA (harbouring the ribosome binding site region of fliC and the first 12 codons of fliZ sequence) and BamRfliZ–HA primers. These PCR fragments were inserted into pBBR1-MCS2 digested with KpnI and BamHI, to generate the pBB-FlhD–HA and pBB-FliZ–HA plasmids.

The pKC2 plasmid, containing the flhC gene under Plac control, was also constructed by inserting the flhC fragment into pRK404 to complement the flhD polar mutant for FlhC. We then transferred pBB-FlhD–HA into the F1, ΩflhD (pRK404) and ΩflhD (pKC2) strains. We transferred pBB-FliZ–HA into the F1 and ΩflhD strains and then we assessed the functionality of recombinant proteins. Restoration for motility and haemolytic phenotypes of ΩflhD (pKC2) (pBB-FlhD–HA) and ΩflhD (pBB-FliZ–HA), respectively, were confirmed before ChIP assays. ChIP assays were performed using a modified version of a published technique (Shin and Groisman, 2005).

Following the growth in mot broth of X. nematophila strains containing FlhD–HA and FliZ–HA tagged proteins, formaldehyde (1%) was added to the cultures (10 ml), which were placed at room temperature for 30 min before quenching the reaction by incubation with glycine (125 mM) for 5 min. Cells were collected and washed twice with cold PBS. Cells were lysed by incubation in 1 ml of lysis solution (10 mM Tris pH 8, 10 mM EDTA, 50 mM NaCl, 20% sucrose and 10 mg ml−1 lysozyme) for 30 min at 37°C and in 1 ml of 2× IP buffer (100 mM Tris pH 8, 300 mM NaCl, 2% Igepal CA-630, 1% sodium deoxycholate, 0.2% SDS and 1 mM PMSF) at 37°C for 15 min. The cell extract was sonicated to obtain an average DNA fragment size of 600 bp and then centrifuged. For the immunoprecipitation of FliZ-cross-linked DNA, 700 μl of the extract was removed and incubated with 7 μl of rabbit anti-HA Tag (Sigma) at room temperature for 90 min. The mixture was incubated with Protein G Sepharose 4 Fast Flow (Amersham) for 1 h at room temperature and centrifuged for 1 min at 10 000 g; 50 μl of supernatant was removed for total DNA preparation. The beads were washed twice with 1× IP buffer then twice in LiCl detergent solution (10 mM Tris pH 8, 250 mM LiCl, 1 mM EDTA, 0.5% Igepal CA-630, 0.5% sodium deoxycholate), and finally with TE buffer. The immunoprecipitated material was eluted with 100 μl of elution buffer (50 mM Tris pH 8, 10 mM EDTA and 1% SDS). It was treated with proteinase K for 2 h at 37°C; the cross-linking of immunoprecipitated and total DNA was reversed by incubation at 67°C for 8 h. Immunoprecipitated and total DNA were then purified with the QIAquick PCR Purification kit (Qiagen). The enrichment of the immunoprecipitated material with specific DNA fragments was determined by quantitative PCR on a LightCycler (Roche), using the LightCycler FastStart DNA MasterPLUS SYBR Green I kit and promoter-specific primers.

Construction of reporter fusion plasmids and β-glucuronidase (GUS) assay

We created a uidA fusion with the fliAZ and flhDC regions, by constructing pBB-UIDK, by inserting the 3.6 kb uidA-Km cassette (from pUIDK1, Bardonnet and Blanco, 1992) into the pBBR1MCS broad-host-range cloning vector (Kovach et al., 1995) from which the Cm resistance gene and the T3, T7 and β-galactosidase promoters had been deleted. The blunt-ended fragments containing the potential promoter regions of the fliAZ locus identified by primer extension were inserted into the SmaI restriction site of pBB-UIDK, resulting in plasmids pAL72 (P1fliA promoter), pAL73 (P2fliA promoter), pAL74 (IR3fliZ region). To construct the PflhD–uidA fusion, a PCR fragment generated with the L-flhDuid and R-flhDuid primers was cloned into pBB-UIDK to yield pAL75 (Table 2). We studied the effect of the σ28 and FliZ proteins on the fliAZ and flhD fusions, by also inserting fragments containing the promoterless fliA and fliZ genes separately into the pRK404 mobilizable vector under transcriptional control of the Plac promoter, giving rise to pKA7 and pKZ4 respectively.

We generated a PxaxA–uidA fusion, by amplifying a 350 bp DNA fragment corresponding to the xaxAB promoter region by PCR from F1 chromosomal DNA for use as a template with primers xax-uidA_f and xax-uidA_r, which contain SacII and XbaI restriction sites respectively. The PCR fragment digested with SacII and XbaI was inserted into the corresponding sites of pBB-UIDK to generate pXABU. Finally, these plasmids were transferred into X. nematophila strains by mating experiments.

β-Glucuronidase activity was measured by spectrophotometrically monitoring the hydrolysis of p-nitrophenyl-β-d-glucuronide into p-nitrophenol (Bardonnet and Blanco, 1992). Bacterial cells were cultured in LB for various times (OD540 of 0.5 for fliAZ fusions and ∼3 for xaxAB fusion) and frozen at −20°C to lyse the cells. We then added 100 μl of each defrosted sample to 900 μl of GUS buffer containing 50 mM NaPi, pH 7.0, 10 mM β-mercaptoethanol, 0.1% Triton X-100 and 1 mM GUS substrate. This mixture was incubated at 37°C until a yellow colour developed. Each reaction was then stopped by adding 0.7 M Na2CO3 and the degradation product was detected at 415 nm. Specific activity was expressed in units such that 1 U corresponded to 1 nmol of p-nitrophenol liberated per minute and per unit optical density at 540 nm.

In vivo pathogenicity assay

The common cutworm, S. littoralis, were reared on an artificial diet at 23°C with a photoperiod of 12 h. Fifth-instar larvae were used for virulence assays, as previously described (Sicard et al., 2004). Briefly, the surfaces of insect larvae were sterilized with 70% (v/v) ethanol. A Hamilton syringe was then used to inject 20 μl each of the appropriate dilution of exponentially growing bacteria in PBS (300–500 cfu) into groups of 20 larvae. Treated larvae were incubated individually for up to 96 h, and the times of death of the insects were recorded. At least four independent experiments were performed for each strain. Statistical analysis was carried out with the Statistical Package for Social Science version 11.0.1 (SPSS, Chicago, IL), comparing individual survival times in each group with individual survival times in the other groups.

Nucleotide sequence accession number

The sequence of the fliAZ operon from X. nematophila F1 strain has been assigned EMBL Accession No. AJ131736.


We are grateful to Didier Lereclus for his help concerning the extension primer, Sylvie Pagès and Nadège Ginibre for pathology assays, Christine Laroui for plasmid preparations and M.C. Guérin and J. Martin for expert technical assistance with real-time PCR. We also thank H. Goodrich-Blair and A. Andersen for the generous gift of the AN6 and xhlA strains. This work was supported by Grant No. 2004 1133 02 from INRA.