Erwinia chrysanthemi requires a second iron transport route dependent of the siderophore achromobactin for extracellular growth and plant infection

Authors


E-mail expert@inapg.fr; Tel. (+33) 1 44 08 17 06; Fax (+33) 1 44 08 16 31.

Summary

Full virulence of the pectinolytic enterobacterium Erwinia chrysanthemi strain 3937 depends on the production in planta of the catechol-type siderophore chrysobactin. Under iron-limited conditions, E. chrysanthemi synthesizes a second siderophore called achromobactin belonging to the hydroxy/carboxylate class of siderophore. In this study, we cloned and functionally characterized a 13 kb long operon comprising seven genes required for the biosynthesis (acs) and extracellular release (yhcA) of achromobactin, as well as the gene encoding the specific outer membrane receptor for its ferric complex (acr). The promoter of this operon was negatively regulated by iron. In a fur null mutant, transcriptional fusions to the acsD and acsA genes were constitutively expressed. Band shift assays showed that the purified E. chrysanthemi Fur repressor protein specifically binds in vitro to the promoter region of the acsF gene confirming that the metalloregulation of the achromobactin operon is achieved directly by Fur. The temporal production of achromobactin in iron-depleted bacterial cultures was determined: achromobactin is produced before chrysobactin and its production decreases as that of chrysobactin increases. Pathogenicity tests performed on African violets showed that achromobactin production contributes to the virulence of E. chrysanthemi. Thus, during infection, synthesis of these two different siderophores allows E. chrysanthemi cells to cope with the fluctuations of iron availability encountered within plant tissues. Interestingly, iron transport mediated by achromobactin or a closely related siderophore probably exists in other phytopathogenic bacterial species such as Pseudomonas syringae.

Introduction

Essential to the biological activity of many proteins mediating electron transfer and redox reactions, iron is absolutely necessary for most forms of life. However, in aerobic conditions, its bioavailability is severely compromised by its poor solubility. At pH 7, ferric ion is available at 10−17 M, while microorganisms require micromolar concentrations of this metal to multiply (Braun and Killmann, 1999). Iron catalyses the formation of reactive free radicals in the presence of oxygen and can trigger chain reactions that are harmful to macromolecules (Luo et al., 1994). This high reactivity means that iron is generally not present in biological tissues as a free element and not readily accessible to invading microorganisms. A wide variety of microorganisms solubilize iron and control their intracellular iron levels by excreting siderophores. Siderophores are high-affinity Fe(III)-scavenging/solubilizing molecules that, once loaded with iron, are specifically imported into the cell ( Neilands, 1995). In Gram-negative bacteria, production of a siderophore and proteins involved in uptake of its ferric complex is accurately controlled by the sensory and regulatory protein Fur. Fur protein acts as a dimer, each monomer containing a non-haem ferrous iron site (Hantke, 2001). If the cellular iron level becomes too low, the active Fur repressor losses Fe2+, its co-repressor, and is no longer able to bind to its operator sites (Escolar et al., 1999; Gonzalez de Peredo et al., 2001). The production of siderophores by pathogenic bacteria can greatly contribute to their virulence, because these molecules can remove iron from a wide variety of organic substrates (Ratledge and Dover, 2000). The role of siderophores in microbial pathogenesis has been extensively studied in infectious diseases of mammals, which express an array of defences aimed at withholding nutritional iron from the pathogen (Weinberg, 1993; 2000).

Interestingly, the virulence of the plant pathogenic enterobacterium Erwinia chrysanthemi strain 3937 on African violets (Saintpaulia ionantha) depends on the production of the siderophore chrysobactin (Enard et al., 1988; Neema et al., 1993; Masclaux and Expert, 1995). Chrysobactin is a bidentate ligand consisting of a monomer of 2,3-dihydroxybenzoyle-d-lysyl-l-serine (Persmark et al., 1989) and thus is a less powerful ferric ion ligand than hexadentate siderophores like the tris-catecholate enterobactin. Siderophores are compared by calculating their pFe values at a concentration of ligand of 10−5 M and pH of 7.4 (pFe is - log[Fe3+]): a higher pFe value corresponds to lower free Fe3+ concentration and stronger binding of the iron by the siderophore. The pFe values of chysobactin and enterobactin are 17.1 and 35.5 respectively (A.-M. Albrecht-Gary, pers. comm.). Chrysobactin is essential for E. chrysanthemi cells to cause systemic infection (Expert, 1999). The soft rot symptom produced by E. chrysanthemi consists of a progressive disorganization of parenchymatous tissues called maceration and results from several bacterial enzymatic activities, including pectinases and endoglucanases which work in concert to degrade plant cell walls (Pérombelon, 2002). E. chrysanthemi cells colonize intercellular spaces of leaf tissues and progress intercellularly. Deconstruction of the cell walls causes cell lysis and allows bacteria free access to cellular nutrients (Murdoch et al., 1999).

Most of the proteins involved in chrysobactin-mediated iron transport are encoded by a 25 kb contiguous region of the E. chrysanthemi chromosome. The presence of four operons divergently transcribed from two bidirectional promoters is reminiscent of the genetic organization of the enterobactin region in Escherichia coli K12 (Expert et al., 2004). In particular, the fct-cbsCEBA operon encodes the receptor Fct and the enzymes leading to the catechol moiety in chysobactin biosynthesis (Franza and Expert, 1991). The promoter region for this operon is controlled by iron via a direct interaction with the ferric uptake regulatory protein Fur, which is highly similar to the E. coli Fur repressor (Sauvage et al., 1996; Franza et al., 1999; 2002). In addition, the E. chrysanthemi Fur protein negatively controls the transcription of several pectinase encoding genes, according to an anti-activation mechanism. Indeed, the Fur binding sites are located in the close vicinity of the sequences recognized by the cAMP receptor protein CRP that plays a crucial role in expression of the pectinolysis genes (Nasser et al., 1997). Fur could thus inhibit the binding of CRP and the subsequent gene transcription activation (Franza et al., 2002). Fur behaves as a global regulator coupling two functions important for pathogenicity.

However, in response to iron limitation, E. chrysanthemi synthesizes a second siderophore, achomobactin. Achromobactin belongs to a new class of siderophores derived from citrate (Münzinger et al., 2000), based on carboxylate and hydroxy donor groups rather than the commonly encountered hydroxamates and catecholates (Fig. 1). Achromobactin was uncovered on the basis of the phenotype of chrysobactin biosynthetic mutants, which are still able to form a halo of discoloration on Chrome Azurol S (CAS) agar medium, used to detect siderophore production. In a previous work, we isolated mutants accumulating achromobactin in the medium that are derepressed for the production of chrysobactin. The mutations were designated as cbr (for chrysobactin regulation), and were shown unexpectedly to map in an operon, cbrABCD, encoding the ABC permease for ferric achromobactin (Mahéet al., 1995).

Figure 1.

Structure of achromobactin and its cyclized form that prevails in neutral aqueous solution (Münzinger et al., 2000). Proteins AcsD, AcsC and AcsA belong to a family of siderophore synthase components catalysing amide bond formation reactions (Martinez et al., 1994). Further details are in the text.

In this study, we sequenced a region of 15 kb extending 5′ upstream to the cbrA gene. We uncovered a cluster of eight genes involved in the biosynthesis and transport functions of achromobactin. We reported a functional analysis of this second iron transport pathway and investigated its role in pathogenicity of E. chrysanthemi.

Results

Cloning and analysis of the nucleotide sequence of the 5′ upstream region of the ferric achromobactin permease operon

A previous study indicated that one mutation (acsA::lacZ) located upstream of the cbrABCD operon abolished achromobactin production (Fig. 2). We used a wild-type gene library and looked for homology with the left-hand 1.9 kb EcoRI–HindIII fragment of plasmid pDE4 that contains the cbrABCD operon (Fig. 2). We identified one recombinant cosmid, pL9G1, that harboured an insert of ≈30 kb and restored the wild-type phenotype to a cbrA mutant. We then isolated a second recombinant cosmid, pL2G12, that shared homology with the 4.2 kb EcoRI fragment present in the genomic insert of pL9G1 (Fig. 2). The nucleotide sequence of a contiguous region of 15 kb was determined.

Figure 2.

Physical and genetic analysis of the E. chrysanthemi achromobactin genomic region. The transcriptional direction for each ORF/gene (see the text and Table 1) is shown by an arrow. The hatched box corresponds to the promoter of the cbrABCD operon (Mahéet al., 1995). The polar mutations and gene fusions described in Table 2 and in the text are indicated by their number and reporter gene respectively. Their position on the nucleotide sequence is numbered relative to the StuI restriction site (position 0) present on cosmid pL2G12. Physical maps for recombinant cosmids and plasmids used in this work are shown. Restrictions sites are abbreviated as follows: E, EcoRI; H, HindIII; Bg, BglII; P, PstI; N, NotI; S, StuI. The underlined HindIII site comes from the pLA2917 vector.

Scanning of the sequence allowed the detection of eight open reading frames (ORFs), displaying the same transcription polarity as that observed for the cbrABCD operon (Table 1). ORF 8 ends 354 bp upstream of the translational start of the cbrA gene. A blastp search of the sequence databases with the predicted amino acid sequences of these eight ORFs revealed a number of relevant informations (Table 1) with respect to the achromobactin biosynthesis and transport. On the basis of these predictions, we designated ORFs 1–8 as acsF, acr, acsD, acsE, yhcA, acsC, acsB, acsA. The presence of short intergenic regions and, for acsE, acsB and acsA, of overlapping stop and start codons (Fig. 3) suggested that this gene cluster was functionally organized as an operon. As for the ferric achromobactin permease encoding genes, the G+C content of this region is 62 mol%, i.e. higher than the average G+C content of E. chrysanthemi (56 mol%).

Table 1.  Predicted functions of the proteins encoded by the different ORFs from the achromobactin genomic region, based on sequence similarities (blastp).
Protein
(ORF)
Amino
acid
Closest demonstrated function
and/or homology
BacteriumIdentity
(%)
Similarity
(%)
AcsF (ORF 1)448DABA AT (diaminopropane production)Acinetobacter baumannii6679
Gamma-aminobutyrate aminotransferase
4-Aminobutyrate aminotransferase
Pseudomonas syringae B728a7586
Acr (ORF 2)704IutA (aerobactin receptor)Escherichia coli2560
AcsD (ORF 3)620PvsD (vibrioferrin biosynthesis)Vibrio parahaemolyticus3046
IucC/IucA (aerobactin synthetase units)E. coli25/2535/34
RhbF (rhizobactin 1021 biosynthesis)Sinorhizobium meliloti2431
AlcC (alcaligin biosynthesis)Bordetella species2333
SbnE (siderophore biosynthesis)Staphylococcus aureus2543
Siderophore synthetase componentP. syringae B728a6577
AcsE (ORF 4)450LysA (diaminopimelate decarboxylase)E. coli3240
PvsE (vibrioferrin biosynthesis)V. parahaemolyticus4360
Diaminopimelate decarboxylaseP. syringae B728a7280
YhcA (ORF 5)482Major Facilitator Family efflux pump   
VceB (multidrug exporter)Vibrio cholerae2736
Multidrug resistance efflux proteinBacillus halodurans3858
Permease of the major facilitator familyP. syringae B728a6473
AcsC (ORF 6)636IuC/IucA (aerobactin synthetase units)E. coli30/2541/32
RhbF (rhizobactin 1021 biosynthesis)S. meliloti2940
AlcC (alcaligin biosynthesis)Bordetella species2738
AcsD (achromobactin biosynthesis)Erwinia chrysanthemi2432
SbnF (siderophore biosynthesis)S. aureus4563
Siderophore synthetase componentP. syringae B728a7080
AcsB (ORF 7)258GarL 2-dehydro-3-deoxyglucarate aldolaseE. coli3149
SbnG (siderophore biosynthesis)S. aureus4864
4-Dihydroxyhept-2-ene-1,7-dioic acid aldolaseP. syringae B728a7483
AcsA (ORF 8)647PvsB (vibrioferrin biosynthesis)V. parahaemolyticus2443
IucC/IucA (aerobactin synthetase units)E. coli23/2432/31
RhbF (rhizobactin 1021 biosynthesis)S. meliloti2532
AlcC (alcaligin biosynthesis)Bordetella species2232
SbnC (siderophore synthesis)S. aureus3249
AcsC (achromobactin synthesis)E. chrysanthemi2432
Siderophore synthetase componentP. syringae B728a6473
Figure 3.

Intergenic regions of the acsF–acsA operon and mapping of the acsF promoter by primer extension. Short intercistronic regions are indicated by their nucleotide sequences, longer ones are indicated by the termination and initiation codons, and the number of nucleotides contained within the region is given. Initiation codons are referred by a number indicating their position on the nucleotide sequence as presented in Fig. 2. The nucleotide sequence of the 5′acsF gene upstream region is shown. The −35 and −10 elements for the P1 predicted promoter are underlined, the putative Fur binding site is boxed and the two arrows indicate the acsF transcription starts determined from primer extension reactions performed under iron-replete (+) and -depleted conditions (–) as shown on the autoradiogram. Lanes T, G, C, A are sequencing ladders.

Characterization of acsF, acr, acsD, acsE, acsC, acsB, acsA polar mutants

To investigate the role of proteins encoded by this gene cluster, we first used the aphA-3 cassette to generate non-polar mutations in each gene. Our attempts remained unsuccessful indicating that accumulation of some achromobactin biosynthesis precursors might be toxic for the cell. Thus, we constructed mutants using insertion elements that generate polar mutations. The mutations acsA-37, acsC-33, acsD-14, acsD::uidA, acsE-1, acsF-29, acsF-1 and acr-1 were introduced into the chromosome of strain 3937 by marker exchange recombination. We constructed several double mutants also carrying the chrysobactin or receptor/chrysobactin-negative mutation cbsE-1 or fct34, according to the compatibility of antibiotic resistance markers. Unlike chrysobactin defective mutants, all double mutants failed to grow on L agar medium containing 2,2′-dipyridyl. They did not produce halo of discoloration on CAS agar medium and their low-iron culture supernatant fluids were negative in the CAS assay (Schwyn and Neilands, 1987). These phenotypes indicate that these mutants were impaired in achromobactin production. The mutants harbouring the mutations acr-1, acsF-29 or acsF-1 failed to use ferric achromobactin as an iron source. SDS-PAGE analysis of the outer membrane of such mutants grown in Tris medium showed that a polypeptide migrating in the 78 kDa range induced in the wild type was missing (Fig. 4). These results indicate that the acr gene encodes the ferric achromobactin receptor, Acr. The apparent molecular mass of this receptor protein fits with the size calculated for the predicted product, i.e. 78.747 kDa. The amino acid sequence revealed the presence of an N-terminal signal peptide (MKVRALYATPLFLFAQA). In the N-terminal part of the matured protein, we identified the VVTA signature that characterizes TonB-dependent receptors (Mey and Payne, 2003).

Figure 4.

SDS-PAGE analysis of outer membrane proteins from iron-depleted E. chrysanthemi cells. The wild-type (wt) strain 3937 and the mutants harbouring the mutations acr-1 or acsF-29, as indicated on each lane, were grown in Tris medium. Protein samples were run on 11% polyacrylamide gels. The three low-iron regulated polypeptides are indicated. The apparent molecular sizes of standard proteins in kDa are indicated on the left of the photograph.

The possibility that the mutations isolated had a polar effect on the downstream genes was examined. We checked whether the presence of cosmid pL9G1 or pL2G12 in the mutants was able to restore the wild-type phenotypes (Fig. 2). None of the mutations including acsA-37, acsC-33, acsD-14, acsE-1, acsF-1, acsF-29 and acr-1 was rescued, suggesting the lack of internal promoters within this gene cluster. We transduced the acsE-1, acr-1 and acsF-1 insertions in L2 cells containing the transcriptional fusion acsA::lacZ. The chromosomal DNA boundaries of transduced insertions were amplified by polymerase chain reaction (PCR), cloned in the pGEM-T easy vector and the DNA sequence was carefully checked as indicated in Experimental procedures. The acsA::lacZ fusion is derepressed in bacterial cells grown under iron limitation (Fig. 5). The presence of either insertion totally abolished the expression of this fusion. For these mutants, the addition in the medium of achromobactin or its biosynthetic precursors did not result in the expression of this fusion (data not shown). Thus, we concluded that the lack of expression of the acsA::lacZ fusion resulted from a polar effect of these insertions on transcription of acsA, the last gene of the cluster. Hence, the acsF–acsA gene cluster must be transcribed as a polycistronic operon.

Figure 5.

Fur-mediated regulation of the acsF–acsA operon. Graphs show the expression of the acsA and acsD gene fusions in L2 and 3937 cells, respectively (circles), or their fur derivatives (triangles) in iron-replete (solid symbols) and -depleted (open symbols) conditions. Experiments were carried out in triplicate and standard deviations are shown. The autoradiogram shows the interaction between Fur and the acsF promoter region analysed by electrophoretic mobility shift assay. Restriction fragment containing the acsF promoter region was incubating with increasing concentrations of purified Fur (monomer) as indicated. E. ch, Erwinia chrysanthemi.

Transcription start for the acsF, acr, acsD, acsE, yhcA, acsC, acsB, acsA genes

Sequence analysis predicted two potential promoters located upsteam of the acsF gene. A putative Fur binding site matching the consensus sequence of the Fur operator, at 15 out of 19 positions (Escolar et al., 1999; Lavrrar and McIntosh, 2003), overlaps this promoter region (Fig. 3). To determine the transcriptional start site of the acsF gene, we performed a primer extension analysis. Templates were mRNA transcripts from 3937 cells grown in Tris medium supplemented with FeCl3 or without supplementation. An iron-regulated cDNA that co-migrated with the positions 86 and 89 of the ladder sequence (see Experimental procedures) is consistent with the occurrence of a transcriptional start at T nucleotides (positions 683 or 680) and with the predicted P1 promoter (Fig. 3).

The acsF gene promoter is directly controlled by Fur

In a previous work, we used a fur null mutant to show that the chromosomal acsA::lacZ fusion is regulated by iron in a Fur-dependent manner (Franza et al., 1999). For comparison, we constructed the chromosomal fusion acsD::uidA. This fusion was derepressed in Tris medium and its expression level in the presence of FeCl3 was relatively high, as observed before for the acsA::lacZ fusion (Fig. 5). In the fur null mutant the acsD::uidA fusion like acsA::lacZ was fully derepressed regardless the iron concentration (Fig. 5). To investigate whether the putative Fur operator site overlapping the P1 promoter of the acsF gene was functional, we analysed the binding of the E. chrysanthemi Fur protein to this promoter region in mobility shift assays. The promoter region of acsF was incubated with increasing concentration of Fur protein in the presence of a 250-fold molar excess of pUC19 DNA. The migration of the acsF promoter DNA was retarded by Fur in a concentration-dependent manner (Fig. 5). When 1 mM EDTA was added to the incubation mixture, Fur did not bind to the acsF regulatory region (data not shown), indicating that only the metalled form of Fur displays affinity for this promoter.

Are achromobactin and chrysobactin systems two complementary iron transport routes?

As the achromobactin system was directly controlled by the Fur repressor like the chrysobactin system, we investigated the expression of these two iron transport routes in iron-depleted wild-type cells. We analysed the time-course production of achromobactin and chrysobactin in cultures grown in Tris medium, using bioassays (Fig. 6B). We detected the presence of achromobactin before that of chrysobactin, and its production declined as that of chrysobactin was increasing. In contrast, in the chrysobactin-negative mutant PPV11, the production of achromobactin increased gradually along the growth (Fig. 6D) and in the achromobactin-negative mutants including PPV27, the production of chrysobactin was slightly advanced and higher compared with wild-type strain (Fig. 6C). In Tris medium, the growth of the two mutants PPV11 and PPV27 was not impaired compared with wild-type strain. Rather, the growth of the achromobactin-negative mutant was improved (Fig. 6A). These results are consistent with the facts that the two systems are homeostatically controlled by the Fur protein and are not together required if the iron environmental conditions are not drastically reduced. On the other hand, the addition of the ferrous ion chelator 2,2′-dipyridyl to the medium limited the growth of both mutants and had an additive effect on the growth inhibition of the double mutant PPV32 (Fig. 6E). In Tris medium containing the ferric ion chelator EDDHA, the growth of the chrysobactin-negative mutant was severely reduced compared with that of the achromobactin non-producer but was similar to that of the double mutant (Fig. 6F). Thus, if ferric ions are not complexed to strong ligands, achromobactin is efficient enough to supply the bacteria with iron. If iron is strongly chelated, only chrysobactin can remove ferric ions from the complex.

Figure 6.

Achromobactin and/or chrysobactin production in iron-depleted E. chrysanthemi cells. In (A), the wild-type strain 3937 (solid circles) and the mutants harbouring the cbsE-1 (open circles) or the acsD-14 (open triangles) mutation were grown in Tris medium. For each strain, achromobactin (black bars) and/or chrysobactin (grey bars) were/was determined (B, wild type; C, acsD-14; D, cbsE-1) using bioassays (see Experimental procedures). The data were expressed as the mean of two independent experiments. The growth in the presence of 2,2′-dipyridyl (E) or EDDHA (F) of the wild-type strain (solid circles) and of the mutants (acsD-14, open triangles; cbsE-1, open circles; acsD-14 cbsE-1, open squares) was recorded. Two independent experiments were carried out and the results of one typical experiment are shown.

Is the yhcA gene involved in achromobactin excretion?

The yhcA gene was predicted to encode a membrane protein, YhcA, comprising 12 helical domains that represent transmembrane segments (TMS) (Fig. 7). A blastp search revealed that the YhcA protein falls into the class of bacterial sugar efflux transporters, including the arabinose efflux permease AraJ that defines a cluster of solute and drug transporters of the MFS. Typical of these transporters is a higher level of conserved amino acids in the N-terminal half of the proteins compared with the C-terminal half. The YhcA protein revealed two conserved motifs, motif A in the cytoplasmic loop between TMS2 and TMS3 (GYLGhRlGnKPLy) and motif C in TM5 (ApalGPlcGGiL) (Fig. 7). Motif A [consensus is GxLaDrxGrkxx(x)l] which is predicted to contain a β-turn structure may be involved in the transport function per se, while motif C (consensus is gxxxGPxxGGxl), found only in multidrug transporters and specific efflux pumps, may dictate the transport direction (Putman et al., 2000). Motif C is not present in MFS symport proteins. Motif B in TMS4, which contains a basic arginine residue and is predicted to be involved in proton transfer, is not conserved in YhcA. Thus, we investigated whether the YhcA protein was involved in achromobactin excretion and/or extrusion.

Figure 7.

Amino acid sequence of the E. chrysanthemi YhcA protein and phenotype analysis of the yhcA-negative mutant. The 12 transmembrane regions (TM) predicted with TMbase (a database of membrane spanning protein regions: http://www.ch.embnet.org/software/TMPRED) are underlined. Bold letters depict the conserved motifs in transporters of the MFS. Further details are in the text [capital letters correspond to amino acids occurring in > 70% of the examined sequences and lowercase letters amino acids occurring in > 40%. x represents any amino acid and [x] not always being present (Putman et al., 2000)]. In (A), shown is the growth of the two yhcA-positive (solid circles) and -negative (open circles) strains in Tris medium. In (B), the time-course production of achromobactin in the culture medium was reported (black bars, yhcA-positive construct; grey bars, yhcA-negative construct). The data were expressed as the mean of two independent experiments. DFO, desferrioxamine.

As indicated before, our attempts to construct a non-polar yhcA-negative mutant were unsuccessful. We thus introduced a polar mutation, yhcA-2, in a chrysobactin-defective background. As expected, the double mutant PPV25 was impaired in achromobactin biosynthesis. We introduced the plasmid pDE10 that expresses the acsCBA genes under the control of vector Plac promoter in this mutant as well as in the double mutant PPV20. For these two constructs, we compared the levels of achromobactin released in bacterial cultures grown in Tris medium (Fig. 7B). Bacterial cells from the yhcA-positive strain produced achromobactin in the culture medium on the onset of the exponential phase of growth and the production increased gradually during the growth. The growth of yhcA-negative bacterial cells was slightly delayed and the production of achromobactin in the medium was severely reduced (Fig. 7A and B). However, in late exponentially grown cultures, the levels of achromobactin present in the medium were almost equivalent for both strains. These data indicate that the YhcA protein enhances the release of achromobactin in the medium, which suggests the existence of a coupling between achromobactin biosynthesis and excretion. The significant levels of achromobactin present in late exponentially grown cultures of the yhcA-negative mutant may reflect cell lysis or an alternative export process to release achromobactin into the medium after prolonged growth.

To investigate whether the YhcA protein could promote the efflux of iron internalized via achromobactin, we analysed the levels of incorporation of 59Fe-labelled achromobactin in bacterial cells from strains PPV20 and PPV25 (Fig. 8A). For both strains, the ferric achromobactin transport rate seemed to be equivalent during the first 10 min of the experiment. However, after 25 min, the yhcA-negative mutant had incorporated twice more iron than the yhcA-positive strain. In a parallel experiment, we checked that a cbrA-negative mutant impaired in the production of ferric achromobactin permease failed to incorporate iron (Fig. 8A). To investigate the physiological effect of increased ferric achromobactin incorporation in the yhcA-negative mutant, we compared the growth of the yhcA-negative mutant in Tris medium supplemented with ferric achromobactin with that of the yhcA-positive strain. The growth of the yhcA mutant was stimulated by ferric achromobactin at a higher level than that observed for the yhcA-positive strain (Fig. 8B). This result shows that the increased incorporation of ferric achromobactin observed in the yhcA mutant was not toxic. We thus compared the stimulation of growth of yhcA-positive and -negative strains by ferric achromobactin in a dose–response bioassay (Fig. 8C). For ferric achromobactin amounts lower than 4 picomoles, the growth zones of the yhcA-negative mutant were larger than those for the yhcA-positive strain. With higher amounts of ferric achromobactin, this difference was no longer visible. These data suggest that the YhcA protein could pump the iron internalized via achromobactin out, if it is not immediately processed.

Figure 8.

Ferric achromobactin acquisition in the yhcA-negative mutant.
A. 59Fe-labelled achromobactin transport was analysed in the yhcA-positive strain PPV20 (solid circles), in the yhcA-negative strain PPV25 (open circles) and in the cbr-negative mutant PPV12 as a negative control (solid triangles).
B. The growth curves of the strains PPV20 (circles) and PPV25 (triangles) in Tris medium amended with ferric achromobactin to the final concentration of 7 µM at the time indicated (open symbols), or without supplementation (solid symbols).
C. Increasing amounts of ferric achromobactin were used in a growth promotion assay for strains PPV20 (black bars) and PPV25 (open bars). Transport and growth promotion assays were carried out in triplicate and standard deviations are shown.

The contribution of achromobactin in pathogenicity

We examined the pathogenic behaviour of achromobactin or/and chrysobactin simple (PPV19 and PPV11) and double mutants (PPV20) on potted African violets by comparison with the wild-type strain. Appearance and progression of the symptoms were scored daily (Fig. 9A). The achromobactin-deficient mutant was less virulent than the wild-type strain but was more virulent than the chrysobactin-deficient strain. The double mutant failed to initiate symptoms in one-third of the inoculated plants. However, it was still able to produce a systemic infection in one-half of the inoculated plants and in this respect, it behaved like the chrysobactin-deficient strain. Interestingly, the absence of symptom production caused by the double mutant correlated with a progressive decline in the bacterial population (Fig. 9B). In contrast, symptom initiation resulted in a drastic increase in the bacterial populations, but the multiplication of the double mutant was 10-fold lower than that of the wild-type strain. We thus conclude that both siderophores contribute to successful infection. Initiation of the symptomatic phase of the disease appears to be critical if there is no siderophore to be produced. It seems that the role of chrysobactin prevails once the symptoms have started, thus indicating that iron is not readily available, despite the degradation of tissues by pectinases.

Figure 9.

Pathogenicity and growth on S. ionantha plants of the wild-type strain and its siderophore-negative mutants.
A. The significant phases of the disease induced by the various strains tested (genotypes are indicated on each graph) was scored as follows: white columns, no symptom; grey columns, maceration localized at the inoculated area or covering the inoculated leaf; black columns, maceration invading the petiole and spreading towards the other aerial parts of the plant.
B. Enumeration of the bacterial populations (cfu) present in intercellular fluids from leaves inoculated with the wild-type (circles) or the siderophore double mutant used in (A) (triangles) was determined on M9 medium, during the symptomless phase (open symbols) or during the beginning of the maceration (solid symbols). Experiments were performed in duplicate and the results of one typical experiment are shown.

Discussion

In a first report, Münzinger et al. (2000) elucidated the structure of achromobactin, a new siderophore of the citrate/carboxylate family produced by E. chrysanthemi 3937 in addition to chrysobactin. In this study, we presented a genetic and functional analysis of this second iron transport route, and compared it with that mediated by chrysobactin in order to estimate the individual role of these two siderophores in pathogenesis. The different questions emerging from this work are successively considered.

A eight-gene iron-regulated operon involved in achromobactin biosynthesis, uptake and release

Nucleotide sequencing, phenotypic analysis of insertion mutants, transcriptional gene expression studies and marker rescue experiments led to the identification of an eight-gene cluster, acsF acr acsD acsE yhcA acsC acsB acsA, encoding proteins required for the biosynthesis and export of achromobactin and for uptake of its ferric complex (Fig. 2). Genetic analysis confirmed that these eight genes are functionally organized as an operon. Primer extension experiment showed that they are transcribed from a single promoter, corresponding to P1 (Fig. 3).

This analysis also confirmed that this operon is regulated by the iron availability. Under iron limitation, both acsD::uidA and acsA::lacZ fusions are derepressed by a 12-fold factor, but in the presence of FeCl3, the basal expression level of these fusions is 10- to 20-fold higher than that observed for the chrysobactin gene fusions. Moreover, the derepression of the achromobactin operon requires less severe conditions of iron limitation than those required for the derepression of the chrysobactin system. The presence of a functional Fur binding site overlapping the P1 promoter elements of the acsF gene is in agreement with the fact that the achromobactin gene fusions are no longer iron regulated in a fur mutant (Fig. 5). We showed that the Fur regulation is direct and could occur via a direct competition between RNA polymerase and the repressor. Thus, the differential expression between the achromobactin and chrysobactin gene systems may result from differences in the promoter strength and the affinity of the Fur repressor to the operator sites.

The occurrence of long operons encoding siderophore systems is rather uncommon. Interestingly, the structure of vibrioferrin, the siderophore produced by Vibrio parahaemolyticus is related to that of achromobactin, as it also contains a chiral citrate moiety and 2-oxo-glutarate in the ring form of pyrrolidone (Tanabe et al., 2003). The genes required for biosynthesis and transport of vibrioferrin are organized in at least four transcriptional units (Tanabe et al., 2003), while the two systems are functionally related (see next paragraph). On the other hand, in Staphyloccocus aureus, the nine genes encoding a siderophore system that seems to be related to that of achromobactin are clustered in a single operon (Dale et al., 2004). The structure of the siderophore produced is still unknown. In addition, we found that the genome of Pseudomonas syringae pv. syringae 728a (Deng et al., 2003) but not that of P. syringae pv. tomato DC3000 (Buell et al., 2003) contains a locus encoding proteins which share between 64% and 75% of identity to those involved in the achromobactin system, except for the outer membrane receptor protein (Table 1). The siderophore produced by P. syringae 728a is probably achromobactin or is structurally closely related to it, as it can be used specifically by E. chrysanthemi 3937 cells as an iron source, via the ferric achromobactin outer membrane receptor Acr (data not shown).

Biosynthesis of achromobactin

Achromobactin consists of citrate substituted on the distal carboxyls with residues of ethanolamine and diaminobutyric acid, respectively, both condensed to a residue of 2-oxo-glutaric acid (Fig. 1). Computer analysis enabled us to assign enzymatic fonctions to AcsF, AcsD, AcsE, AcsC and AcsA proteins encoded by the acsF, acsD, acsE, acsC and acsA genes respectively. A blastp search for the AcsF protein gave a significant alignment with amino acid sequences of pyridoxal-5′-phosphate-dependent enzymes of class III, that catalyse reactions of transamination. Given the high level of homology of AcsF with the diaminobutyrate:2-ketoglutarate 4-aminotransferase of Acinetobacter baumanii (Table 1) (Ikai and Yamamoto, 1997), we assume that AcsF is an enzyme that fulfils a similar reaction, involving glutamate as the amino group donor: aspartate β-semialdehyde + glutamate = diaminobutyrate + 2-oxo-glutaric acid. The AcsE protein falls into the family of pyridoxal-dependent decarboxylases acting on ornithine, lysine, arginine and related substrates. This protein may catalyse the decarboxylation of a precursor amino acid, possibly containing a serine moiety, to produce a residue of ethanolamine. Moreover, the AcsE amino acid sequence has a significant level of homology with that of the PvsE protein, required for biosynthesis of vibrioferrin that also contains a residue of ethanolamine (Tanabe et al., 2003). A blastp analysis for AcsD, AcsC and AcsA revealed that these proteins belong to a family of synthases, including the IucA and IucC components of aerobactin synthase, which catalyses the consecutive condensation of two molecules of N6-acetyl-N6-hydroxylysine to each of the terminal carboxyl groups of citric acid, generating amide bonds (Martinez et al., 1994). As achromobactin contains three amide bonds, we assume that AcsD, AcsC and AcsA catalyse similar reactions and thus constitute the three components of a synthase. A number of IucA and IucC homologues has been identified in the biosynthesis of other citrate-derived siderophores such as rhizobactin 1021, alcaligin and vibrioferrin ( Brickman and Armstrong, 1999; Lynch et al., 2001; Tanabe et al., 2003), but globally the levels of homology between these components are relatively weak (Table 1). On the other hand, the potential role of the AcsB protein in achromobactin biosynthesis is less clear. A blastp analysis revealed that the AcsB belongs to the HpaH/HpaI aldolase family that includes two aldolases using 2,4-dihydroxyhept-2-ene-1,7-dioic acid and 4-hydroxy-2-oxovalerate as substrates respectively (Hubbard et al., 1998). A multiple alignment of amino acid sequences of AcsB, of these two proteins and of the E. coli DDG aldolase GarL, showed that besides several conserved motifs distributed along the sequences, the amino acid residues participating in the active site of GarL including Ser-124, Arg-75 and His-50 are present (Izard and Blackwell, 2000). As GarL is involved in the (d)-glucarate and galactarate catabolic pathway and galactarate is used as a carbon source in E. chrysanthemi, we checked whether the acsB mutant still catabolized galactarate (data not shown). There was no difference in assimilation of this substrate compared with the parental strain, indicating that AcsB is not an enzyme of the galactarate catabolism in E. chrysanthemi. We thus suggest that AcsB is an enzyme that catalyses the conversion of some carbon intermediate into pyruvate and aldehyde. Such an enzymatic activity would have the advantage to recycle the Krebs cycle intermediates or to maintain the pool of central carbohydrates necessary for the formation of achromobactin. In this respect, it is noteworthy that achromobactin biosynthesis occurs on the beginning of the exponential phase of growth, when cell metabolic activities are increasing while the iron availability decreases. Finally, we did not find a gene in this operon that would encode an enzyme catalysing the formation of an ester bond between citrate and ethanolamine or its derivative. Such a gene must be present on another part of the chromosome.

Transport functions

We identified the yhcA gene, encoding a protein that displays the amino acid sequence characteristics of efflux pumps of the MFS (Putman et al., 2000). We showed that this transporter facilitated the release of achromobactin in the medium. A likely interpretation of these data is that the YhcA protein would be involved in transport of achromobactin out the cell. Similar export pumps for catecholate siderophores (Furrer et al., 2002; Page et al., 2003), including the EntS enterobactin exporter, have already been described. We thus propose that the primary role of the YhcA protein would be to excrete the newly synthetized siderophore. We have evidence that the AcsD, AcsC and AcsA components of achromobactin synthase interact physically with proteins located in the inner membrane (B. Py, pers. comm.). A transmembrane efflux pump coupled with a siderophore synthase would prevent the bacterium from accumulating cumbersome levels of the neosynthetized siderophore. However, we cannot exclude that the YhcA transporter is involved in the recycling of achromobactin outside the cell or/and in the extrusion of ferric achromobactin if the iron is not immediately removed from its ligand. We found that in a chrysobactin-deficient mutant, increased intracellular levels of ferric achromobactin resulting from the yhcA mutation were well tolerated. This situation may be different in the wild-type strain, if the two siderophores are simultaneously produced. Extrusion of overflowing ferric achromobactin might be required if ferric chrysobactin is also internalized.

We found that the ferric achromobactin receptor, Acr, is one of the three low-iron-inducible outer membrane proteins, previously identified in E. chrysanthemi 3937. Although the homology of the Acr protein with other aerobactin receptors is relatively weak (Table 1), we must note that it is the best score found, compared with other TonB-dependent receptor specific for citrate-derived siderophores.

The contribution of achromobactin in pathogenicity

We found that E. chrysanthemi requires both achromobactin and chrysobactin during the infection process. This indicates that these two ligands act in a complementary way to satisfy the bacterial iron needs, suggesting that the bacteria are confronted to fluctuations in the iron status during their journey within the plant. In this regard, the fact that achromobactin and chrysobactin have different affinity for iron and would be produced in planta in a sequential manner could be considered as an advantage. In planta bacterial growth curves (Fig. 9B) and microscopic observations (Murdoch et al., 1999) have shown that wild-type cells can survive for several days in intercellular spaces of parenchymatous tissues without multiplying substantially, a situation that strongly contrasts with the phase of maceration during which the number of viable counts increases dramatically. Therefore, it appears that E. chrysanthemi is very well suited to acquire iron in conditions where the environment is continually changing. Ferric citrate is present in the plant apoplasm but it is noteworthy that E. chrysanthemi does not use ferric citrate as an iron source. Achromobactin, the pFe of which is higher than that of citrate (pFe of citrate is 14.8), can readily compete with this compound and other plant iron ligands such as nicotianamine (pFe of nicotianamine is 16). In conclusion, achromobactin resembles a preventive system that helps bacterial cells to cope transiently with a loss of iron until chrysobactin is fully induced. Investigations have been initiated to further elucidate the role of these two siderophores in planta.

Experimental procedures

Bacterial strains, phages, plasmids and media.

The bacterial strains, except some double acs cbsE-1 mutants, and bacteriophages used in this work are described in Table 2. Plasmids are described in Table 2 and Fig. 2. The rich media used were L broth and L agar (Sambrook et al., 1989). L agar was iron-depleted by adding ethylenediamine-N,N′-bis(2-hydroxy-phenylacetic acid) (EDDHA, Sigma Chemical), or 2,2′-dipyridyl (Sigma Chemical) to give a final concentration of 80 µM or 200 µM respectively. Tris medium was used as the low-iron minimal medium (Franza et al., 1999). It was iron-depleted by adding EDDHA or 2,2′-dipyridyl to give a final concentration of 40 µM or 100 µM respectively. For iron-rich conditions, it was supplemented with 20 µM FeCl3. Glucose (2 g l−1) was used as the carbon source. The anti-bacterial agents and chemicals used were as reported previously (Franza et al., 1991; 1999).

Table 2.  Bacterial strains, bacteriophages and plasmids used in this study.
Strain/plasmidRelevant characteristicsSource/reference
Strain
Erwinia chrysanthemi
 3937Wild type isolated from African violetOur collection
 L2Lac derivative of 3937Hugouvieux-Cotte-Pattat and Robert-Baudouy (1985)
 PPV13Lac, acsA1::lacZ, Acs, KmRFranza et al. (1999)
 PPV15Lac, fur::Ω, acsA1::lacZ, Fur, Acs, SpecR, KmRFranza et al. (1999)
 PPV22acsD::uidA, Acs, KmRThis work
 PPV23fur::Ω, acsD::uidA, Fur, Acs, SpecR, KmRThis work
 PPV18Lac, fct34::lacZ, Fct, Cbs, KmRFranza et al. (2002)
 PPV11cbsE-1, Cbs, SpecRMahéet al. (1995)
 PPV12Lac, cbrA-21, acsA1::lacZ, cbsE-1 Cbr, Acs, Cbs, CmR, KmR, SpecRMahéet al. (1995)
 PPV19acsA-37::MudIIpR13, Acs, CmRThis work
 PPV20acsA-37, cbsE-1, Acs, Cbs, CmR, SpecRThis work
 PPV21acsC-33::MudIIpR13, Acs, CmRThis work
 PPV24yhcA-2::MudII1734, YhcA, Acs, KmRThis work
 PPV25yhcA-2::MudII1734 cbsE-1, YhcA, Acs, Cbs KmR, SpecRThis work
 PPV26acsE-1::Ω, Acs, SpecRThis work
 PPV27acsD-14::MudII1734, Acs, KmRThis work
 PPV28acr-1::Ω, Acr, Acs, SpecRThis work
 PPV29acr-1::Ω, fct34::lacZ, Acr Acs Fct , Cbs SpecR, KmRThis work
 PPV30acsF-29::MudII1734, Acs, Acr, KmRThis work
 PPV31acsF-1::Ω, Acs, Acr, SpecRThis work
 PPV32acsD-14::MudII1734, cbsE-1, KmR, SpecRThis work
 PPV17cbsE-1, tonB60, Cbs, TonB, SpecR, KmREnard and Expert (2000)
Escherichia coli K-12
 M8820F LacaraD 139Δ(ara-leu)7697Δ(proAB-argFlacIPOZYA)XIII rpsL, SmRCastilho et al. (1984)
 POII1734FaraD139 ara::(Mucts)3 MudII1734 (lac, KmR) Δ(lac)X74 galU galK rpsLCastilho et al. (1984)
 MC4100 (Mucts)araD139 Δ(argF-lac)U169 rpsL relA1 flbB5301 ptsF25 deoC1 (Mucts) MudIIPR13, SmR CmRRatet et al. (1988)
 ED8767supE44 supF58 hsdS3 (rBmB) recA56 galK2 galT22 metB1Sambrook et al. (1989)
 DH5αsupE44 ΔlacU169 (f80lacZΔM15) hsdR17 recA1 endA1 gyrA96 thi relA1Sambrook et al. (1989)
 C600Fthi thr-1 leuB lacY fhuA supESambrook et al. (1989)
Phage
 øEC2Generalized transducing phage from E. chrysanthemi strain 3690Résibois et al. (1984)
Plasmids/cosmids
pUC182.7 kb vector, AmpRSambrook et al. (1989)
pWSK295.4 kb vector, pSC101 derivative, AmpRWang and Kushner (1991)
pBC3.4 kb vector, pUC19 derivative, CmRStratagene
pGEM-T Easy3.015 kb vector, pGEM-5Zf derivative, AmpRPromega
pUIDK1pNB4 vector carrying the uidA-Km cassette KmR and AmpRBardonnet and Blanco (1992)
pRK2013Mobilization helper plasmid, KmRFigurski and Helinski (1979)
pLA291721 kb mobilizable cosmid, TcR and KmRAllen and Hanson (1985)
pL9G1Sau3A fragment of E. chrysanthemi genomic DNA cloned in the BglII site of the KmR gene of pLA2917, Cbr+, TcRThis work
pL2G12Sau3A fragment of E. chrysanthemi genomic DNA cloned in the BglII site of the KmR gene of pLA2917, TcRThis work
pDE410 kb HindIII fragment of E. chrysanthemi genomic DNA cloned in pUC18, Cbr+, AmpRExpert et al. (1992)
pDE105.7 kb NotI–BglII fragment from pL9G1 cloned in NotI–BamHI linearized pWSK29This work
pDE114.2 kb EcoRI fragment from pL9G1 cloned in pBCThis work
pTF309 kb HindIII fragment from pL9G1 cloned in pBCThis work
pDE121284 bp amplified fragment of yhcA gene region cloned in pGEM-T EasyThis work
pTF311260 bp amplified fragment of acr gene region cloned in pGEM-T EasyThis work
pTF321320 bp amplified fragment of acsF gene region cloned in pGEM-T EasyThis work
pDE13645 bp amplified fragment containing the acsF promoter regions cloned in pGEM-T EasyThis work

General microbiological techniques

Insertional mutagenesis with MudII1734 (Castilho et al., 1984) of plasmids pDE11, pDE12 and pTF32 and with MudIIpR13 (Ratet et al., 1988) of plasmid pDE4 was carried out as described previously (Enard et al., 1988). The interposon Ω coding for spectinomycin resistance was inserted in the PstI site of the acsF gene cloned in pTF32, in the HpaI site of the acr gene from pTF31 and in the SmaI site of the acsE gene from pDE11. The uidA-Km cassette from pUIDK1 was inserted into the HpaI site of the acsD gene from pDE11 to generate a transcriptional fusion acsD::uidA. Plasmid transformation, marker exchange recombination into the chromosome of strain 3937 and phage phiEC2 transductions were performed as described previously (Franza et al., 1999). When necessary, the chromosomal DNA regions flanking the transduced mutations acsF-1, acr-1 and acsE-1 were amplified by PCR by using an internal primer present on both ends of the interposon Ω (5′-agcataaagcttgctca-3′) and the primers described below. These amplicons were cloned into the pGEM-T Easy vector and their DNA sequence was checked as indicated in General DNA methods. Tri-parental matings with the helper plasmid pRK2013 were performed as described previously (Franza et al., 1991).

Determination of siderophore and phenotype analysis of mutants

Siderophore activity was detected as CAS-reacting material using CAS agar plates or in the culture supernatant (Schwyn and Neilands, 1987), using desferrioxamine BDFO (Desferal, Novartis Pharma SA) as the standard. The biological activity of achromobactin or chrysobactin was determined in bioassays under low-iron conditions. Plates were poured with 15 ml of EDDHA-L agar medium seeded with 10 µl of an overnight L broth culture of indicator strains, PPV20 and PPV29 or of other strains to be tested. Sterile disks of 6 mm diameter were placed on the agar surface and 15 µl of filter-sterilized culture supernatants of strain to be tested grown in Tris medium were added. The source of achromobactin was a filter-sterilized supernatant of a culture of strain PPV17 grown overnight in Tris medium. Formation of the ferric complex of achromobactin was estimated using the CAS assay. The diameters of zones of growth of the indicator strains were measured after 24 h. When necessary, analysis of the outer membrane proteins from low-iron cultures was performed as described by Enard et al. (1988).

General DNA methods

DNA manipulations (plasmid and chromosomal DNA isolation, cloning and electrophoresis) and the E. chrysanthemi genomic library constructed in cosmid pLA2917 were described previously (Franza et al., 1991; 1999). All cloning experiments were performed in the DH5α strain of E. coli. DNA/DNA hybridization analysis was performed by using the Denhardt's method described by Sambrook et al. (1989). DNA sequencing was performed on double-strand plasmid (pDE4 and pTF30) by the dideoxynucleotide chain termination method with the Sequenase version 2.0 and [α-35S]-dATP or [α-33P]-dATP (Amersham Biosciences). When necessary, DMSO (Sigma) was added in the annealing mixture (5% final concentration). Nucleotide sequence determined on cosmid pL2G12 was obtained from Genome Express. Data were analysed with the software package UW GCG provided by Bisance (Dessen et al., 1990). The nucleotidic sequence of the achromobactin gene cluster has been submitted to GenBank under the Accessions No. AF416739 and AF416740. For PCR amplification of E. chrysanthemi genomic fragments, we used the following primers: 5′-tcactactgaagacg gcat-3′ and 5′-gcgacgccaccacgatcg-3′ for the yhcA gene region (pDE12), 5′-gggcaagatgaactact-3′ and 5′-atgtgatg tactccttgt-3′ for the acr gene region (pTF31), 5′-gatcatgtgca gagaacc-3′ and 5′-cacggcttcaatcaacgc-3′ for the acsF gene region (pTF32), 5′-tcacccacctcaacgtag-3′ and 5′-gcgatcacct gcgagatg-3′ for the acsE gene region. PCR was performed in a DNA thermocycler with denaturation at 94°C for 60 s, annealing at 52°C for 75 s, and an extension at 72°C for 75 s, which was followed by an extension reaction at 72°C for 10 min. PCR products were cloned into the plasmid pGEM-T Easy, according to the manufacturer's instructions. Nucleotide sequence of PCR products was obtained from Genome Express.

RNA isolation

An overnight culture in L broth was diluted 60-fold in Tris medium-glucose supplemented with FeCl3 or without supplementation. The culture was grown under shaking until an absorbance of 0.5 at 600 nm was reached. Culture (7.5 ml) was harvested by centrifugation for 10 min at 4°C (8000 g). The cell pellet was then resuspended in 600 µl of buffer A (20 mM sodium acetate, pH 5.5, 1 mM EDTA) at 0°C. After addition of 33 µl of 10% SDS and 600 µl of hot acidic phenol (65°C) equilibrated with buffer A, the sample was vigorously mixed for 30 s and incubated for 10 min at 65°C. The aqueous phase was re-extracted with phenol/chloroforme (1:1) equilibrated with Tris 10 mM, pH 7. RNA was precipitated overnight with 30 µl of 3 M sodium acetate and 800 µl of ethanol. The RNA pellet was washed with ethanol 70% and resuspended in 35 µl of water treated with diethyl pyrocarbonate.

Primer extension

The following oligonucleotide, complementary to the sequence between positions 790 and 770 relative to the left StuI site (position 0 in Fig. 2), was used for primer extension experiments: 5′-ctcaaaccatccctgtcgctcc-3′. This primer was labelled with [γ-32P]-ATP with the T4 polynucleotide kinase enzyme. Two to five micrograms of total RNA were mixed with 0.1 pmole of [γ-32P]-end-labelled specific primer complementary to the acsF gene in 12 µl of water. The mixture was incubated at 70°C for 5 min and chilled on ice. Synthesis of first strand cDNA was then performed with the MBI Fermentas RevertAidTM H Minus First Strand cDNA Synthesis Kit according to the manufacturer's instructions. Samples were concentrated under vacuum and 4 µl of loading stop solution was then added. Extension products were separated by electrophoresis on a 5% polyacrylamide sequencing gel. A DNA sequence ladder was obtained by sequencing pUC19 with the 22-mer (-46) M13 forward sequencing primer.

Preparation of operator fragment and gel retardation assays

The promoting region (645 bp) from the acsF gene was amplified by PCR, using the primers 5′-aaataatatcgagaaggg-3′ and 5′-gcccgacgccagaaat-3′, and cloned into the pGEM-T Easy vector (pDE13) according to the procedure described above. Nucleotide sequenced was checked before use. Promoter DNA was hydrolysed with EcoRI and purified after electrophoresis on low-melting agarose gels by using the QIAquick extraction kit from Qiagen. The DNA fragment was end-labelled with [α-33P]-dATP by using the Klenow enzyme. The acsF promoter DNA was diluted to a final concentration of 0.4 nM and incubated with increasing amount of purified Fur protein in the presence of an excess of pUC19 DNA (100 nM), as described by Franza et al. (2002). Samples were separated by electrophoresis on a non-denaturing 4% polyacrylamide gel at 140 V in a buffer containing 40 mM BisTris Borate (pH 7) and 250 µM MnCl2.

Ferric achromobactin transport experiments

An overnight culture in L broth supplemented with ampicillin of the strains PPV25 and PPV20 harbouring pDE10 was diluted 1:40 in Tris medium supplemented with glucose and ampicillin and the culture was grown with shaking until an absorbance of 0.5 at 600 nm was reached. A culture of strain L37 cbrA-21 acs1 cbsE-1 was used as a control. Bacterial cells were harvested by centrifugation, washed and suspended in phosphate-free Tris medium with glucose and kept on ice until use. The transport medium was phosphate-free Tris medium containing 8 µM achromobactin and supplemented with 1 µM 59FeCl3 (8.92 mCi ml−1 Fe(III) chloride in 0.5 M HCl; Amersham Biosciences). For transport experiments, the bacterial suspension was diluted in transport medium to give an absorbance of 0.4 at 600 nm in a total volume of 5 ml and placed in a 50 ml Erlenmeyer flask. At intervals of 5–30 min, 200 µl were withdrawn and immediately filtered through a filter with 0.45 µm pores that had been soaked for at least 12 h in Tris medium supplemented with 20 µM unlabelled FeCl3. Filters were immediately washed with 20 ml of phosphate-free Tris medium. The filters were placed in scintillation vials and air dried, and radioactivity was measured by liquid scintillation counting. Twenty microlitres of samples of each bacterial culture were counted to check the total amount of radioactivity. For each strain, experiments were performed in duplicate.

Determination of β-galactosidase and β-glucuronidase activities in bacterial culture

An inoculum of an overnight LB culture was 60-fold diluted in Tris medium containing glucose and appropiate antibiotics. The inoculated culture was divided in two subcultures supplemented or not with 20 µM FeCl3. Cultures were grown aerobically at 30°C. Samples were collected and immediately frozen. Enzymatic activities were assayed as reported previously (Sauvage and Expert, 1994; Masclaux et al., 1996). β-Galactosidase activity is expressed in Miller units whereas β-glucuronidase activity is expressed in nanomoles of paranitrophenol liberated per minute, per unit of optical density at 600 nm.

Pathogenicity assays and determination of in planta bacterial growth

Pathogenicity test was performed on potted African violets (S. ionantha) cv. Blue Rhapsody as described previously (Sauvage and Expert, 1994; Franza et al., 1999). Two independent series of inoculations were carried out, using two different batches of plants. Twenty-four plants were inoculated per strain. For in planta bacterial growth determination, a set of 70 plants were inoculated (one leaf per plant) with the wild-type strain or the siderophore-negative mutant PPV20. Intercellular fluids from five inoculated leaves were taken up at different times after infection as described by Neema et al. (1993) and Masclaux and Expert (1995). Inoculated leaves displaying no symptom were also collected. The bacterial population present in the intercellular fluids was determined on agar minimal medium containing the appropriate antibiotics for the mutant strain.

Acknowledgements

This work was supported by the Institut National de la Recherche Agronomique (INRA). B.M. was a fellow of the Ministère de la Recherche et de l’Enseignement Supérieur and D.E. is a researcher from the Centre National de la Recherche Scientifique (CNRS). We are grateful to the Molecular Microbiology reviewers for their suggestions to improve the manuscript.

Ancillary