Upon iron limitation, Bacillus subtilis secretes the catecholic trilactone (2,3-dihydroxybenzoate-glycine-threonine)3 siderophore bacillibactin (BB) for ferric iron scavenging. Here, we show that ferri-BB uptake is mediated by the FeuABC transporter and that YuiI, a novel trilactone hydrolase, catalyses ferri-BB hydrolysis leading to cytosolic iron release. Among several Fur-regulated ABC transport mutants, only ΔfeuABC exhibited impaired growth during iron starvation. Quantification of intra- and extracellular (ferri)-BB in iron-depleted ΔfeuABC cultures revealed a fourfold increase of the extracellular siderophore concentration, confirming a blocked ferri-BB uptake in the absence of FeuABC. Ferri-BB was found to bind selectively to the periplasmic binding protein FeuA (Kd = 57 ± 1 nM), proving high-affinity transport of the iron-charged siderophore. During iron starvation, a ΔyuiI mutant displayed impaired growth and strong intracellular (30-fold) and extracellular (6.5-fold) (ferri)-BB accumulation. Kinetic studies in vitro revealed that YuiI hydrolyses both BB and ferri-BB. While BB hydrolysis led to strong accumulation of the tri- and dimeric reaction intermediates, ferri-BB hydrolysis yielded exclusively the monomeric reaction product and occurred with a 25-fold higher catalytic efficiency than BB single hydrolysis. Thus, ferri-BB was the preferred substrate of the YuiI esterase whose gene locus was designated besA.
In contrast, remarkably less is known about further siderophore-dependent iron acquisition pathways. While siderophore biosynthesis genes and their products are described in many bacteria, the components of transport and iron release are largely unknown. However, there seems to be a change as several recent reports deal with siderophore transport mechanisms in important pathogens such as Mycobacterium smegmatis (Zhu et al., 1998), M. tuberculosis (Rodriguez and Smith, 2006), Yersinia pestis (Fetherston et al., 1999) or Listeria monocytogenes (Jin et al., 2006). The catecholic siderophore bacillibactin (BB) is produced by the Gram-positive model organism Bacillus subtilis and related species such as human and/or animal pathogenic Bacillus anthracis, B. cereus and B. thuringiensis. Synthesis of BB and its precursor 2,3-dihydroxybenzoate (DHB) depends on the dhbACEBF operon (Rowland and Taber, 1996; Rowland et al., 1996; May et al., 2001). The trimodular NRPS system DhbEBF assembles DHB, glycine and threonine and cyclotrimerizes the tripeptidic intermediates to yield the BB trilactone scaffold that is structurally related to Ent (May et al., 2001). The crystal structure of the DHB adenylation domain DhbE was solved and was used for inhibition studies (May et al., 2002; Miethke et al., 2006). Transcriptome analyses showed the Fur-dependent regulation of the BB biosynthesis genes and several iron transporters (Baichoo et al., 2002). While the ferric hydroxamate uptake system FhuBCG-FhuD was previously characterized (Schneider and Hantke, 1993), a recent study reported the substrate preferences of further Fur-dependent ABC transporters and indicated the FeuABC-YusV-dependent utilization of BB and Ent (Ollinger et al., 2006). However, the intracellular iron release from ferri-BB remained obscure.
In its first part, this study characterizes BB biosynthesis by showing both the superiority of BB among the endogenous B. subtilis ferric iron chelators and the physiological relevance of BB-dependent iron scavenging. The suitability of the BB NRPS for a mutasynthesis approach with various aryl acid analogues is demonstrated. In the second part, two cellular components acting downstream of the BB biosynthesis are introduced by in vivo and in vitro characterization. Evidence is provided that ferri-BB uptake depends exclusively on the ABC transporter FeuABC and that its periplasmic binding protein FeuA senses ferri-BB in the nanomolar range. For the following step of iron release in the cytosol we identified and characterized the esterase YuiI as the major catalyst of ferri-BB hydrolysis.
Bacillibactin is the dominant ferric iron chelator of B. subtilis
In order to monitor siderophore production by a quick and efficient screening, we intended to use the chrome azurol sulphonate (CAS) agar plate assay (Schwyn and Neilands, 1987) as the method of choice. Due to reasons published earlier (May et al., 2001), we had to establish this assay for B. subtilis using another growth medium that allowed siderophore-dependent halo formation in the presence of surfactin (see Experimental procedures). Colonies of the wild-type (WT) strain ATCC 21332 grown for 3 days on the CAS agar formed bright orange halos. To confirm that the halo formation was the result of BB secretion, we used the strain JJM405 lacking the entire second part of the dhbF gene that was described to be BB-deficient (May et al., 2001). Halo formation of JJM405 colonies on CAS agar plates was drastically reduced compared with the WT, although very small halos were observed after several days of growth (Fig. 1A). Because the dhb operon encodes not only for the BB NRPS DhbEBF but also for the DHB synthesis enzymes DhbCBA, strain JJM405 should still be able to produce DHB acting as a potential iron chelator of lower efficacy. To prove this, the isochorismate synthase gene dhbC was replaced in frame by an erythromycin resistance cassette. The phenotype of the resulting strain BMM100 was analysed on CAS agar plates. BMM100 colonies showed no halo formation after 3 days of incubation, confirming that production of DHB was mainly responsible for the slight halo formation of JJM405 (Fig. 1A). Subsequently, growth of WT and the BB biosynthesis mutant strains was monitored in iron-depleted minimal medium (Fig. 1B). Both mutants displayed a diminished growth rate in the exponential phase and a decreased cell density in the stationary phase. In contrast, no growth differences were observed when the cultures were supplemented with 50 μM FeSO4 (data not shown). The mutational analyses showed that BB is the most efficient compound for extracellular ferric iron scavenging produced by B. subtilis during iron limitation and that alternative endogenous iron chelators such as DHB are not sufficient to sustain WT-like cell growth in the absence of BB.
Generation of BB derivatives by mutasynthesis
When the dhbC mutant strain BMM100 was grown excessively long (more than 3 days) on CAS agar plates, BB production was restored to a certain extent. We concluded that this mutation suppression was probably due to the availability of a DhbC-independent DHB source. Possibly, the missing DhbC activity was partially compensated at later stages of incubation by its paralogue MenF that functions as isochorismate synthase in the menaquinone synthesis pathway (however, during culture incubation for 24 h, MenF was reported not to interfere with the DhbC-dependent DHB synthesis; Rowland and Taber, 1996). To investigate whether DHB could be added in trans to the BB NRPS system, we supplied iron-depleted liquid cultures of BMM100 with 2.5 mM DHB. After 10 h of growth, culture supernatants were extracted with ethyl acetate (see Experimental procedures) and BB with [M+H]+ at m/z 883 was found in extracts of DHB-supplemented, but not in non-supplemented (control) cultures by mass spectrometry (MS) analysis. As studies on several NRPS systems have shown that mutations in the genes for the biosynthesis of non-essential precursors facilitated derivative formation by adding precursor analogues to the cultures (Hojati et al., 2002; Weist et al., 2004; Gregory et al., 2005), we tested the flexibility of the BB NRPS for the incorporation of DHB analogues into the BB scaffold. When cross-feeding the cultures with 2.5 mM salicylate instead of DHB, a substance mass peak with [M+H]+ at m/z 835 was detected that differed in 48 Da (= 3[O]) from the BB mass peak, revealing that salicylate fully replaced DHB as aryl cap during the in vivo assembly leading to the secreted BB derivative [salicylate (SAL)-glycine (Gly)-threonine (Thr)]3, termed SAL-BB (Figs S1 and S2). Analysis of the compounds by analytical reversed-phase high-performance liquid chromatography (RP-HPLC) revealed retention times of 14.2 min and 17.3 min for BB and SAL-BB respectively. SAL-BB was purified from fermentation broth and the substance was used as internal standard for subsequent BB quantifications by mass analysis (see Table 1). Among further DHB analogues that were supplemented to BMM100 cultures, benzoic acid, 3-hydroxybenzoic acid and picolinic acid were also assembled as aryl caps into the macrocyclic peptide scaffold yielding products with [M+H]+ at m/z 787, 835 and 790 respectively (Fig. S2).
Table 1. Intra- and extracellular bacillibactin (BB) quantification.
n.o., not observed.
BB intracellular (μg g−1)
1.25 ± 1.08
1.49 ± 0.65
36.13 ± 7.00
BB extracellular (μg ml−1 OD−1)
2.20 ± 0.97
8.17 ± 3.43
14.17 ± 4.90
Identification and characterization of the ferri-BB uptake system FeuABC
In order to identify the specific ferri-BB uptake system, we deleted the Fur-regulated ABC transporters FeuABC, YfiYZ-YfhA, YclNOPQ and YfmCDEF yielding mutant strains BMM110 (ΔfeuABC), JJM410 (ΔyfiYΔyfiZΔyfhA), JJM420 (ΔyclNOPQ) and JJM430 (ΔyfmCDEF), respectively, and analysed their phenotypes on CAS agar plates (Fig. 2A). The colonies of JJM410, JJM420 and JJM430 did not show significant differences in growth and halo formation compared with the WT. In contrast, strain BMM110 displayed severely impaired growth and an enhanced halo formation after 3 days of incubation. Subsequent cultivation in iron-depleted minimal medium showed no differences between JJM410, JJM420, JJM430 and WT. In contrast, growth of BMM110 was drastically reduced (Fig. 2B). Supplementation with an excess of free available iron (50 μM FeSO4) led to a homogenous growth of all mutants and WT (Fig. S3), confirming the iron-dependent growth defect of BMM110. Because we initially showed that BB was the major compound produced by B. subtilis leading to halo formation on CAS agar plates, the phenotype of BMM110 strongly indicated extracellular accumulation of (ferri)-BB. To prove this, strain BMM110 was used for intra- and extracellular (ferri)-BB quantification (Table 1). The intracellular fractions of WT and BMM110 contained comparable concentrations of (ferri)-BB in the range of 1–2 μg per gram fresh cell weight. However, the (ferri)-BB amounts in the supernatant of the BMM110 culture were fourfold increased compared with the WT culture. This extracellular accumulation correlated with the expected phenotype of a specific ferri-BB importer knockout. As DHB was also shown to be capable of halo formation on CAS agar, we furthermore addressed the question whether FeuABC might additionally play a role in uptake of DHB in monomeric form or as ferri-(DHB)3 complex. To investigate this, feuABC was deleted in the backgrounds of strains BMM100 (DHB– BB–) and JJM405 (DHB+ BB–), resulting in strains BMM111 and BMM121 respectively. Growth of BMM100, JJM405 and BMM121 was monitored during iron limitation. If uptake of ferri-(DHB)3 was dependent on FeuABC, growth of the DHB-producing strain BMM121 should correspond to the DHB-deficient strain BMM100. However, BMM121 and JJM405 grew very similar to each other, whereas BMM100 displayed a diminished growth rate. This indicated that DHB-mediated iron acquisition in the absence of BB was not strictly dependent on FeuABC. Furthermore, strain BMM111 was cultured in iron-depleted minimal medium with and without supplementation of 2.5 mM DHB. BB was detected in supernatant extracts of DHB-supplied cultures at a concentration of 25 μM, confirming that also uptake of monomeric DHB occurred by FeuABC-independent transport activities.
To prove the specific sensing of ferri-BB by FeuABC, the periplasmic substrate-binding protein FeuA of the transporter was further characterized. FeuA is assumed to be a lipoprotein anchored to the outer side of the cytoplasmic membrane by a diacylglycerol moiety attached to the N-terminus of the mature protein. FeuA was recombinantly produced in E. coli as a 35 kDa C-terminal His6-tag fusion and purified by Ni-NTA affinity chromatography and gel filtration with a final yield of 8 mg l−1 pure protein (Fig. S4). For matters of solubility, the recombinant FeuA variant was untethered and unacylated as it lacked the first 20 N-terminal amino acids including the signal sequence (amino acids 1–19) and the cystein at position 20 which is the predicted residue for lipid modification. After purification, separated batches of 50 μM FeuA-His6 and 50 μM bovine serum albumin (BSA) (as a control protein) were incubated with 50 μM ferri-BB (yielding solutions of a pale red colour) for 15 min at room temperature. After short centrifugation of the samples in Centrifugal Filter Units (Millipore), a dark red fraction accumulated at the bottom of the filter containing FeuA-His6, but not in the BSA control, indicating a specific interaction of ferri-BB with FeuA. For detailed binding studies, 50 nM pure FeuA-His6 was incubated with varying concentrations of non-loaded BB and ferri-BB. Quenching of the FeuA-His6 emission maximum at 329 nm (after excitation at 280 nm) upon ligand binding was measured by fluorescence spectroscopy yielding concentration-dependent saturation curves (Fig. 3). The dissociation constants were determined with Kd = 191 ± 12 nM for BB and Kd = 57 ± 1 nM for ferri-BB, revealing an approximately threefold higher binding affinity for the iron-charged than for the non-loaded siderophore. To estimate the FeuA binding specificity for further relevant ligands, additional binding studies with various ferric iron chelators were carried out. The exogenous ferri-siderophore ferri-Ent and the endogenously derived iron source ferri-(DHB)3 revealed affinities to FeuA that were 80% and 0.8% compared with ferri-BB binding respectively. The addition of ferric dicitrate as a further endogenous iron source, however, resulted in unspecific emission quenching below 15% without saturation up to a ligand concentration of 12.5 μM, indicating a binding incompatibility of FeuA and ferric dicitrate.
Identification and characterization of the ferri-BB trilactone hydrolase YuiI
After entering the cell, it is essential that the iron is released from the ferri-siderophore complex to become available for cellular processes. BB contains a trilactone backbone and therefore we suspected that it might be susceptible for a trilactone hydrolase activity. Searching for such an activity in B. subtilis unveiled the gene yuiI upstream of the dhb cluster that was reported to be Fur-regulated and found to be an IroE homologue (Baichoo et al., 2002; Zhu et al., 2005). YuiI shows approximately 20% identity to the IroE esterase and contains the typical GxSxG serine esterase motif. A database search with genomes of related Bacillus species capable of BB biosynthesis revealed that five B. subtilis relatives possess a yuiI orthologue downstream of their feuABC(feuD) transport genes (Fig. S5). Bacillus licheniformis harbours an undivided feuABC-yuiI-dhbACEBF gene cluster and seems to represent an intermediate between the branching arrangements of BB biosynthesis, ferri-BB uptake and esterase genes among these Bacillus species. The YuiI orthologues were found to be highly conserved among the considered species (35–62% identity compared with B. subtilis YuiI) (Fig. S6A). The cladogram based on the YuiI primary sequence similarities (Fig. S6B) coincidently shows the same pattern of species clustering as observed for the genomic association of yuiI with BB synthesis and/or transport genes. To investigate the phenotype of a yuiI mutant during iron limitation, the strain BMM140 carrying a pMUTIN insertion in the yuiI gene was used for a CAS agar analysis (Fig. 2A). After 3 days of incubation, the BMM140 colonies showed both a reduced growth and an increased halo formation in comparison with WT colonies. The cultivation of the yuiI mutant in iron-depleted minimal medium confirmed the impaired growth (Fig. 2C) which was strictly iron-dependent as the corresponding control cultures of mutant and WT (repleted with 50 μM FeSO4) displayed no growth differences (Fig. S3). If natively occurring breakdown of ferri-BB in the cytosol was initially dependent on a trilactone hydrolase, the elimination of this activity should lead to a preferentially intracellular accumulation of the siderophore. Because the CAS assay only visualizes extracellular siderophore accumulation, the intra- and extracellular (ferri)-BB concentrations in iron-depleted BMM140 cultures were quantified by RP-HPLC/MS (Table 1). Strikingly, we found an approximately 30-fold increased (ferri)-BB concentration in the cytosol of the yuiI mutant, confirming that (ferri)-BB breakdown is severely blocked in the absence of YuiI. Furthermore, a 6.5-fold extracellular siderophore accumulation was observed and was likely to be the consequence of the intracellular (ferri)-BB overflow.
To confirm that YuiI indeed hydrolyses (ferri)-BB, the protein was recombinantly produced in two variants (both about 34 kDa) containing either an N- or a C-terminal His6-tag fusion and purified by Ni-NTA affinity chromatography with a final yield of 30 mg l−1 pure protein (Fig. S4). In an initial activity study, both YuiI variants (1 μM each) were incubated with both BB and ferri-BB (50 μM each) for 10 min at 25°C. RP-HPLC/MS analysis of the samples revealed a complete hydrolysis of the substrates into the DHB-Gly-Thr monomeric product in all reactions containing YuiI, while the substrates were completely stable in the controls without enzyme. Thus, YuiI was confirmed to act as an esterase hydrolysing the BB trilactone cycle. As both His6-tag fusions of YuiI proved to be functional, the N-terminal variant was used for all further analyses that were performed to study the reaction characteristics in more detail. Incubation of the substrates with a reduced enzyme concentration of 50 nM due to the rapid substrate turnover unveiled different patterns of product formation during hydrolysis of non-loaded and iron-charged BB (Fig. 4). While in the case of non-loaded BB hydrolysis all three theoretical hydrolysis fragments (with [M+H]+ at m/z 901 for the trimeric, [M+H]+ at m/z 607 for the dimeric and [M+H]+ at m/z 313 for the monomeric product) were present in high abundance, only the monomeric product was detected in the case of ferri-BB hydrolysis. As subsequent analyses of ferri-BB hydrolysis with shorter incubation times and further reduction of enzyme concentration did not lead to the detection of the trimeric and dimeric reaction intermediates, this suggested distinct reaction mechanism for BB and ferri-BB hydrolysis. Reaction kinetics that should reflect the substrate specificity for the cyclic substrates were therefore examined considering formation of the trimeric hydrolysis fragment as the first detectable reaction product of BB hydrolysis and formation of the monomeric product in the case of ferri-BB hydrolysis. Because an incubation time of 20 s was short enough to allow only the first hydrolysis step to occur in the reaction with non-loaded BB, varying concentrations of BB and ferri-BB were accordingly incubated with YuiI and the kinetic paramaters were determined (Table 2). It turned out that the catalytic efficiency for the overall hydrolysis of ferri-BB was approximately 25-fold higher than for the trimeric product formation during hydrolysis of non-loaded BB. To examine the YuiI activity with a structurally related catecholic siderophore containing a trilactone cycle, Ent was used as YuiI substrate for a comparative analysis. As the detection of linear Ent upon hydrolysis of non-loaded Ent was possible, the Km and kcat/Km for this reaction were determined with 221 ± 3.6 μM and 5.0 min−1 μM−1, respectively, indicating a ninefold lower substrate specificity and a threefold lower catalytic efficiency for hydrolysis of Ent compared with BB hydrolysis. Thus, YuiI was capable to hydrolyse the trilactone cycle of Ent; however, BB was clearly preferred as substrate, indicating that the overall size of the siderophore scaffold is decisive for the efficiency of this reaction.
Table 2. Kinetic paramaters for YuiI-dependent BB and ferri-BB hydrolysis.
kcat/Km//kobs/Kobs (min−1 μM−1)
Cyclic trimer → linear trimer
24.4 ± 0.7
348 ± 31.4
Cyclic trimer →→→ monomer
0.54 ± 0.09
205 ± 27.8
The BB-dependent iron acquisition pathway in B. subtilis is not nearly as well understood as the Ent-dependent pathway in E. coli (Fischbach et al., 2006). Here, several aspects of the BB biosynthesis were examined and the ferri-BB uptake and iron release systems were characterized.
The analyses of BB synthesis mutants on plate and in liquid culture during iron limitation demonstrated that B. subtilis utilizes BB as the major compound for high-affinity iron scavenging. Alternative endogenous iron chelators such as DHB or citrate were not efficient enough to compensate for a BB deficiency. Furthermore, the BB biosynthesis machinery was found to be a flexible system regarding the aryl capping of its macrocyclic product. This additionally confirms the relaxed substrate specificity of DhbE in vivo. In previous work, DhbE was shown to activate non-specific salicylate in vitro with a catalytic efficiency of 5.6 × 104 s−1 M−1 that was sixfold lower than observed for specific DHB activation (May et al., 2001). Taking advantage of the structural similarities between BB and the mutasynthesis product SAL-BB, we used this derivative successfully as an internal standard for MS-based quantification of BB in iron-depleted cultures. The high sensitivity of this method allowed the BB quantification from small culture volumes with product concentrations in the nanomolar range for reproducible results.
The investigation of further steps in the BB pathway led us to study the ferri-BB uptake system mediating the transport of the iron-charged siderophore into the cell and the ferri-BB trilactone hydrolase catalysing the intracellular iron release. The feuABC locus coding for an ABC transporter that comprises the periplasmic binding protein FeuA and the membrane permeases FeuB and FeuC was crucial for the maintenance of cell growth during iron limitation. Only the addition of high ferrous iron amounts (50 μM) led to comparable growth of the feuABC deletion strain BMM110 and the WT. The extracellular accumulation of BB in iron-depleted BMM110 cultures strongly indicated that this transport system is specifically involved in the uptake of ferri-BB. In vitro studies with purified FeuA showed the highest binding affinity to ferri-BB with a Kd of 57 ± 1 nM. FeuABC is the closest homologue to the E. coli FepBDG(C) system for ferri-Ent transport across the cytoplasmic membrane. The dissociation constants for Ent and ferri-Ent to the periplasmic binding protein FepB of E. coli are Kd = 60 nM and Kd = 30 nM respectively (Sprencel et al., 2000). In the case of FeuA, we observed a threefold preference for the binding of ferri-BB compared with the binding of non-loaded BB. In contrast to FepB, FeuA is assumed to be tethered to the extracellular surface of the cytoplasmic membrane. Furthermore, while B. subtilis possesses FeuA as a single layer of ferri-BB sensing, E. coli FepB is actually the second sensor of ferri-Ent and FepA in the outer membrane the ‘first gate of recognition’ with a Kd for ferri-Ent of 0.1 nM (Newton et al., 1999). Altogether, this indicates that the ferri-Ent uptake system of E. coli is at least two orders of magnitude more sensitive for ferri-Ent than the B. subtilis uptake system for ferri-BB. However, the recombinant FeuA variant that was used in this study must not necessarily display binding activities as high as engendered by the native form. In accordance to recent studies suggesting one shared transport system for ferri-BB, ferri-Ent and ferric itoic acid (DHB-Gly dipeptide) (Dertz et al., 2006) and showing FeuABC-dependent growth stimulation with both BB and Ent (Ollinger et al., 2006), we now identified FeuA as sensor for ferri-BB, ferri-Ent and, to a low extent, also for ferri-(DHB)3. Thus, although the ferri-BB and ferri-Ent complexes were reported to have opposite chirality (Bluhm et al., 2002), the binding affinity to the FeuA receptor was only of minor difference. However, as the ferri-Ent outer membrane transporter FepA greatly discriminates ferri-BB binding (Annamalai et al., 2004), there seem to be different modes of recognition for iron-charged catecholic siderophores at translocation across the outer and the cytoplasmic membrane. As the cellular uptake of both DHB and ferri-(DHB)3 was not abolished in a DHB-deficient feuABC mutant (strain BMM111), this confirmed the presence of further DHB-dependent transport activities. The growth experiments with BMM110 revealed furthermore that the uptake systems for alternative endogenous ferric iron chelators (such as DHB or citrate) were insufficient to compensate for a blocked BB-dependent iron uptake, most likely because the secreted BB competes successfully with these chelators leading to a severe iron self-depletion of the cells. The ATP-binding protein as the cytosolic component of ABC transporters is missing in the B. subtilis feu operon, tempting to speculate whether FeuABC interacts with a specific or with various trans-located ATP-binding protein(s). The observation of YusV-dependent growth stimulation with BB and Ent indicated the involvement of this ATP-binding protein in FeuABC-mediated transport (Ollinger et al., 2006). In contrast, the feu operon of B. anthracis, B. cereus and B. thuringiensis comprises a gene locus for an ATP-binding component that we termed feuD (see Fig. S5). Sequence similarities suggest that FeuD represents a FhuC paralogue in these Bacillus species.
Upon import, the iron must be released from the ferri-siderophore by cellular mechanisms. In many organsims, this process is mediated by ferri-siderophore reductases (Schröder et al., 2003). It was assumed for a long time that iron release in B. subtilis was mainly a reductive process. Indeed, former studies showed that non-hydrolysable Ent analogues were sources of iron for cell growth (Lodge et al., 1980) and that release of iron from ferri-(DHB)3 was mediated by a ferri-siderophore reductase activity that preferred NADPH as reductant in B. subtilis and B. megaterium (Arceneaux and Byers, 1980; Gaines et al., 1981). As B. subtilis is capable to take up ferric hydroxamates, ferri-(DHB)3 or ferric dicitrate, a reductive iron release for utilization of these alternative iron sources is conceivable. However, the BB scaffold contains three ester bonds which implies the possibility of hydrolytic iron release from the siderophore by breaking down its trilactone cycle. So far, the presence of a ferri-BB esterase activity in B. subtilis was obscure. This study presents evidence that YuiI is an esterase that plays a key role in both iron release and siderophore degradation by catalysing the trilactone hydrolysis of ferri-BB. The yuiI mutant (BMM140) showed, in contrast to the feuABC transport mutant, a strong accumulation of intracellular BB indicating that the YuiI esterase acts in the cytosol as a (ferri)-BB breaker. The impaired growth of the esterase mutant during iron limitation furthermore implied a decisive role of YuiI for hydrolysing the iron-charged form of the siderophore. The activity of YuiI in vitro confirmed its capability to hydrolyse the trilactone ring of BB with great preference for the iron–siderophore complex rather than for the non-loaded siderophore.
On account of these data, we suggest to designate the gene locus of B. subtilis yuiI as ferri-BB esterase besA. To our knowledge, BesA, belonging to the family of α,β-hydrolases, is the first trilactone hydrolase described in a Gram-positive bacterium. Although the protein sequences of B. subtilis BesA and E. coli IroE are quite similar, the enzymes differ in cellular localization, substrate specifity and product formation. Premature IroE possesses an N-terminal signal sequence for membrane translocation and the mature protein is predicted to be tethered to the cytoplasmic membrane at the periplasmic side (Hantke et al., 2003; Lin et al., 2005). It tends to prefer the non-loaded forms of Ent and salmochelins as substrates and the main product observed was the linearized trimer due to a single hydrolysis event (Lin et al., 2005; Zhu et al., 2005). In contrast, there is no signal sequence predicted in the N-terminal part of BesA in all considered Bacillus species. Thus, and in consistency with the experimental observations, BesA can be considered as a cytosolic enzyme. BesA accepts both BB and ferri-BB as substrate and hydrolyses the trilactone cycle in three steps leading to the monomeric tripeptide as final product. However, the catalytic efficiency for ferri-BB hydrolysis is much higher than for the hydrolysis of non-loaded BB. Furthermore, the different patterns of product formation suggest two distinct reaction modes for BB and ferri-BB hydrolysis in a way that in the first case the intermediates dissociate from the enzyme whereas in the second case they stay enzyme-bound. Presumably, the shape of ferri-BB is more suitable for a complete hydrolysis without intermediate accumulation than the non-loaded scaffold conformation. Moreover, Ent was not as efficiently hydrolysed by BesA as BB. Thus, although the central trilactone cycle of the siderophore is the target of the catalytic action, the efficacy of the hydrolysis is constraint by both the shape and the overall structure and size of the macrocyclic scaffold. In the B. subtilis genome, the Fur-regulated gene ybbA is located downstream of feuABC and was suggested to code for an esterase (Baichoo et al., 2002; Zhu et al., 2005). As YbbA has protein sequence identities of about 15% to both E. coli IroD, Fes and IroE, B. subtilis probably harbours a second trilactone hydrolysing activity which might serve as a kind of backup system (that, however, does not compensate for a besA mutation) or display different substrate preferences. In Fig. 5, a current model of the BB-dependent iron acquisition pathway in B. subtilis is presented. Further studies are currently initiated to identify the BB exporter and to elucidate the roles of additional putative pathway components.
In comparison, the BB pathway and all other known siderophore-mediated iron acquisition pathways are non-cyclic. Therefore, they exhibit a strict phenotypic hierarchy showing dominant phenotypes of the upstream components. As both the FeuABC transporter and the BesA esterase were found to be crucial for maintenance of cell growth while BB was produced and secreted to overcome the iron starvation, their potential to serve as new downstream model targets for inhibition of ferri-siderophore utilization with respect to iron-dependent pathogen control is under investigation.
Strains and general methods
The bacterial strains and plasmids used in this study are listed in Table 3. Antibiotics for selection of B. subtilis strains were used at the following concentrations: chloramphenicol (5 μg ml−1), erythromycin (1 μg ml−1), lincomycin (25 μg ml−1), kanamycin (10 μg ml−1), spectinomycin (200 μg ml−1). Antibiotics for selection of plasmid containing E. coli TOP10 strains were used at the following concentrations: ampicillin (100 μg ml−1), kanamycin (50 μg ml−1). DNA manipulations and transformations were carried out as described previously (Sambrook et al., 1989; Hoch, 1991; Klein et al., 1992).
pCB28a+ containing an N-terminal His6-tag fusion of yuiI lacking the first 25 codons of the annotated gene sequence (SubtiList Database)
pET28a+ containing yuiI as C-terminal His6-tag fusion lacking the first 25 codons of the annotated gene sequence (SubtiList Database)
pCB28a+ containing a C-terminal His6-tag fusion of feuA lacking the first 20 codons of the annotated sequence (SubtiList Database)
Deletion mutants were constructed following the PCR synthesis method of marker cassettes with long flanking homology regions (Wach, 1996). Hereby, the target gene was replaced by a resistance marker to facilitate mutant selection. The primers used to generate PCR fusions are listed in Table S1. All PCRs were carried out with a Platinum®Pfx DNA Polymerase (Invitrogen). In a first round of PCR, chromosomal DNA of B. subtilis ATCC 21332 was used as template to amplify the homologous regions up- and downstream of the target gene. The resulting PCR products carrying complementary 3′-ends to a resistance cassette were used in a second round of PCR generating a fusion construct of the homologous genomic regions and the resistance marker. The PCR fusions for construction of strains BMM100, BMM110 and BMM120 were directly used for transformation of B. subtilis. The PCR fusions for construction of strains JJM410, JJM420 and JJM430 were cloned into pΔsrf prior to the transformation. Transformants were selected on antibiotic-containing LB plates. Chromosomal DNA of the mutants was isolated and the recombinations were confirmed by PCR. The resistance determinants used for the different constructs were obtained from several plasmids. The erythromycin resistance cassette from pMUTIN was used to construct strain BMM100 by in-frame replacement of the dhbC gene. The chloramphenicol resistance cassette of pΔsrf was used to construct the strains JJM410, JJM420 and JJM430 by replacing the genes yfiYyfiZyfhA, yclNOPQ and yfmCDEF respectively. From pDG783, the kanamycin resistance cassette was used to construct strain BMM110 by replacing the genes feuABC. Strain BMM120 was constructed using the spectinomycin resistance cassette from pUS19 replacing the genes feuABCybbA. Strain BMM140 was generated by transforming B. subtilis ATCC 21332 with chromosomal DNA of strain BFA1447. Because this strain contained a gene disruption based on Campbell-type integration, it was always grown under selecting conditions except during the CAS agar analysis.
General liquid culture conditions
Strains in liquid culture were grown under agitation (250 r.p.m.) at 37°C. For growth during iron depletion, standard minimal medium (0.5% glucose) without citrate and iron salt was used (Stülke et al., 1993). Stock solutions were stored in polyethylene tubes and cultures were grown in polycarbonate flasks to avoid iron contamination during growth. The iron content of this medium was 0.13 ± 0.04 μM which was determined by inductively coupled plasma mass spectrometry (ICPMS) using an ICPMS Agilent 7500 with a hydrogen cell. For iron repletion controls, 50 μM FeSO4 was added. For cross-feeding experiments with strain BMM100, 2.5 mM of either DHB, salicylate, benzoate, 3-hydroxybenzoate, picolinate or 3-hydroxypicolinate was added to the cultures.
CAS agar plate assay
A chrome azurol sulphonate-hexadecyltrimethylammonium bromide (CAS-HDTMA) stock solution was prepared according to published protocols (Schwyn and Neilands, 1987). The CAS-HDTMA solution was mixed 1:10 with standard minimal medium containing 1.2% agar-agar and, optionally, 160 μg ml−1 tryptophane and phenylalanine. To test the specificity of halo formation, 2,3-dihydroxybenzoic acid (10 mM; pH 4.0), DHB (10 mM; pH 8.0) as well as purified BB and surfactin (3 mM each) were applied to the CAS agar. Only DHB at pH 8.0 (deprotonated form) and BB led to an orange halo formation that was very faint in the first and very large in the second case. To test halo formation of B. subtilis WT and mutants, strains were first grown on LB plates over night at 37°C. From these plates, cells were spotted on the CAS agar plates that were then incubated for 20 h at 30°C. Further incubation took place at room temperature. The plates were scanned once a day to monitor the proceeding halo formation.
Preparation of BB and mutasynthesis products
Bacillibactin was extracted from B. subtilis culture broth using ethyl acetate as described previously (May et al., 2001). Dried pellets of the ethyl acetate extracts were solved in 50% methanol and purified by RP-HPLC using a column Macherey-Nagel 250/21 Nucleodur 100-5 C18 and a gradient from 5% to 60% water/acetonitrile/0.1% trifluoroacetic acid (TFA) over 30 min. The acidic conditions led to ligand protonation of BB and therefore iron was removed from the siderophore during the purification process. After pure BB fractions were identified by electrospray ionization mass spectrometry (ESI-MS) using an ESI-quadrupol mass spectrometer (Agilent 1100 MSD, Series A), they were lyophilized and the non-loaded BB was solved in dimethyl sulphoxide (DMSO) to a stock concentration of 20 mM that was stored at −20°C. To load BB quantitatively with ferric iron, equal volumes of 20 mM BB and a freshly prepared 20 mM FeCl3 solution were mixed, resulting immediately in complex formation monitored by an absorption change from colourless to dark violet. The mixture was incubated for 5 min at room temperature. To remove traces of free iron, the mixture was subsequently incubated with a cation exchange powder [Amberlite CG-50-II (Fluka)] for 20 min at 22°C under gentle shaking. The solid phase was then separated from the supernatant by centrifugation (13 200 g at 22°C for 1 h). In contrast to non-loaded BB, iron-charged BB did not lead to any halo formation when applied to CAS plates, confirming the complete saturation of BB with ferric iron in the stock solution that was stored at −20°C.
Bacillibactin mutasynthesis products were extracted according to the BB extraction procedure. Samples of all extractions were screened for product contents by high resolution Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) using a Finnigan LTQ-FT. For RP-HPLC purification of SAL-BB, the acetonitrile gradient used for BB preparation was from 15% to 55% due to the increased hydrophobicity of the compound. Analytical RP-HPLC/MS with BB and SAL-BB was performed with a column Macherey-Nagel 125/2 Nucleodur 100-3 C18 (using a flow rate of 0.3 ml min−1, column temperature 45°C, and a linear gradient from 10% to 95% water/methanol/0.05% formic acid over 20 min) coupled with an ESI-quadrupol mass spectrometer (Agilent 1100 MSD, Series A).
Bacillus subtilis WT and strains BMM100, JJM405, BMM110, BMM130 and BMM140 were grown in standard minimal medium without iron for 10 h. Twenty millilitres from each culture were harvested and cells and culture supernatant were separated by centrifugation at 4000 g for 20 min at 4°C. Sample preparation for intracellular quantification was performed as follows: after the wet weight of the cell pellets was determined, the cells were washed twice with 1 ml of TE buffer, finally resuspended in 400 μl of TE and 0.2 mg ml−1 lysozyme was added. Complete cell disruption was performed by subsequent sonication (5 × 1 min at 54 W). Cell debris were separated by centrifugation (20 000 g for 1 h at 4°C) and the cytosolic fractions were extracted three times with equal volumes of ethyl acetate. The organic fractions were pooled and the solvent was evaporated by vacuum centrifugation at 45°C. The obtained pellets were solved in 100 μl of 50% methanol. Sample preparation for extracellular quantification was performed by extracting the culture supernatants according to the BB extraction method (see above). After ethyl acetate evaporation, dried pellets were solved in 400 μl of 50% methanol. Samples were analysed by RP-HPLC and ESI-MS (Agilent 1100 MSD, Series A) using SAL-BB as internal standard for signal normalization. For that, 10 μl of a SAL-BB stock solution (1 mM) was added to the 100 μl of samples. With the RP-HPLC method applied (column Macherey-Nagel 250/2 Nucleodur C18 Pyramid, flow rate 0.2 ml min−1, column temperature 45°C, linear gradient from 10% to 95% water/methanol/0.05% formic acid over 55 min), BB and SAL-BB showed retention times of 29.3 and 35.4 min respectively. The calibration curve was obtained by using 18 BB containing samples with defined concentrations ranging from 10 μM to 500 μM. The intracellular BB quantities were calculated in μg g−1 of wet cell weight. The extracellular BB quantities were calculated in μg ml−1 OD600–1. Quantifications were performed three times from independent culture extracts, and mean values and standard deviations were calculated from the total data set. As controls, cytosolic fractions and culture supernatants were extracted a fourth time and no BB was detectable in these extracts, confirming a quantitative extraction in the first three steps.
Cloning, overexpression and purification of FeuA and YuiI
Intrinsic fluorescence measurements of His6-FeuA without or with several ligands [non-loaded and iron-charged forms of BB and Ent (in varying concentrations from 0.5 nM to 4 μM) as well as ferri-(DHB)3, ferric dicitrate and FeCl3 (in varying concentrations from 0.1 to 1000 μM)] were performed with a FP-6500 spectrofluorometer (Jasco) at 22.5°C (constant temperature) with an excitation and emission band width of 5 nm each, a response of 0.5 s and with medium sensitivity. The excitation and emission maxima determined for His6-FeuA were 280 and 329 nm respectively. The measuring range was from 295 to 400 nm with a data pitch of 0.5 nm. Measurement started with a Solution A (50 mM HEPES, 100 mM NaCl; pH 7.0) containing 50 nM His6-FeuA without any ligands and was pursued by adding defined quantities of a Solution B containing 50 nM His6-FeuA and the ligand at high concentration in the same buffer. The solution was then mixed by magnetic stirring in the cuvette for 5 min. To avoid stirring effects, the mixture was allowed to come to rest for 5 min before the measurement started. Negative controls with 50 nM BSA under the same conditions revealed no emission quenching with the non-loaded and iron-charged forms of BB and Ent. In contrast, FeCl3 at concentrations above 1 μM induced emission quenching of both His6-FeuA and BSA, indicating unspecific quenching effects for non-chelated iron. After substraction of background fluorescence, the ligand-dependent quenching of the His6-FeuA emission maximum was plotted appropriately and the dissociation constants were determined as described previously (Sprencel et al., 2000) as means of three independent measurements for each ligand.
Kinetics with YuiI
All kinetic studies were carried out with the N-terminal His6-tag fusion of YuiI at a concentration of 50 nM. Non-loaded and iron-charged BB and Ent (concentrations ranging from 0.3 to 1000 μM) were pre-incubated in 75 mM HEPES buffer pH 7.5 at 25°C for 5 min before the reactions were started by adding the enzyme. The samples were mixed and further incubated at 25°C and after 20 s the reactions were stopped with 1/2 volume of 6% (v/v) TFA in methanol. The samples were analysed by RP-HPLC and ESI-MS using a column Macherey-Nagel 125/2 Nucleodur 100-3 C18 with a flow rate of 0.3 ml min−1, column temperature 45°C and a linear gradient from 10% to 95% water/methanol/0.05% formic acid. The UV absorption of substrate and products was monitored at 215 nm. For an optimal peak resolution, the gradient for separation of the hydrolysis products of non-loaded BB and Ent was set over 38 min. For analysis of the ferri-BB reactions, the gradient was set over 20 min. The peak areas of the products (the trimeric hydrolysis products of non-loaded BB and Ent with [M+H]+ at m/z 901 and 688, respectively, and the monomeric hydrolysis product of ferri-BB with [M+H]+ at m/z 313) were integrated and the kinetic reaction parameters were determined as means of three independent measurements. No spontaneous hydrolysis of the substrates in the control samples without enzyme was observed during the time of the analysis.
We are indebted to Michael Fischbach and Dr Henning Lin for the kind gift of Ent and many helpful discussions. We are grateful to Dr Oscar P. Kuipers and Dr Helga Westers for providing the B. subtilis strain BFA1447. Christian Renner is acknowledged for helping with the fluorescence spectroscopy as are Oliver Happel and Jürgen Knöll for ICPMS measurement. We thank Christiane Bomm, Antje Schäfer and Sven Siebler for excellent technical assistance. This work was supported by the EC-Grant LSHG-CT-2004-503468, by grants from the DFG, the Fonds der Chemischen Industrie and the Graduiertenkolleg ‘Protein function at the atomic level’.