The Legionella pneumophila Siderophore Legiobactin Is a Polycarboxylate That Is Identical in Structure to Rhizoferrin

Abstract

Legionella pneumophila, the agent of Legionnaires' disease, secretes a siderophore (legiobactin) that promotes bacterial infection of the lung. In past work, we determined that cytoplasmic LbtA (from Legiobactin gene A) promotes synthesis of legiobactin, inner membrane LbtB aids in export of the siderophore, and outer membrane LbtU and inner membrane LbtC help mediate ferrilegiobactin uptake and assimilation. However, the past studies examined legiobactin contained within bacterial culture supernatants. By utilizing high-pressure liquid chromatography that incorporates hydrophilic interaction-based chemistry, we have now purified legiobactin from supernatants of virulent strain 130b that is suitable for detailed chemical analysis. High-resolution mass spectrometry (MS) revealed that the molecular mass of (protonated) legiobactin is 437.140 Da. On the basis of the results obtained from both MS analysis and various forms of nuclear magnetic resonance, we found that legiobactin is composed of two citric acid residues linked by a putrescine bridge and thus is identical in structure to rhizoferrin, a polycarboxylate-type siderophore made by many fungi and several unrelated bacteria. Both purified legiobactin and rhizoferrin obtained from the fungus Cunninghamella elegans were able to promote Fe³⁺ uptake by wild-type L. pneumophila as well as enhance growth of iron-starved bacteria. These results did not occur with 130b mutants lacking lbtU or lbtC, indicating that both endogenously made legiobactin and exogenously derived rhizoferrin are assimilated by L. pneumophila in an LbtU- and LbtC-dependent manner.

INTRODUCTION

The Gram-negative bacterium Legionella pneumophila is both an inhabitant of natural and man-made water systems and the primary etiologic agent of Legionnaires' disease, an increasingly common and serious form of pneumonia (1–3). In its aquatic habitats, L. pneumophila survives in biofilms and as an intracellular parasite of amoebae, and in the lung, it grows primarily in macrophages (4–7). Iron has long been recognized as a major aspect of L. pneumophila replication, intracellular infection, and virulence (8, 9). For ferrous iron assimilation, L. pneumophila utilizes a secreted pyomelanin that has ferric reductase activity and the inner membrane Fe²⁺ transporter FeoB (10, 11). For ferric iron uptake, the bacterium uses the siderophore legiobactin (12, 13). Both of these iron acquisition pathways are required for infection of the lungs by L. pneumophila (10, 14).

The investigation of legiobactin has followed a rather circuitous path. Indeed, it was originally thought that L. pneumophila does not produce siderophores; this conclusion was based on negative data obtained from both the Arnow and Csáky assays, which detect catecholate and hydroxymate structures, respectively, and the Chrome Azurol S (CAS) assay, which detects iron chelators independently of their structure (15–17). However, in 2000, we demonstrated that L. pneumophila strains can secrete a siderophore activity that is detected by the CAS assay when the bacterium is cultured in a deferrated, chemically defined medium (12). The failure of previous efforts by us and others at finding siderophore activity was likely due to the fact that the CAS reactivity is obtained only in cultures that are inoculated with bacteria that had been grown to either the log or early stationary phase. Curiously, inocula obtained from the late-stationary phase, despite being capable of initiating growth in the deferrated media, do not induce CAS reactivity. In a follow-up study, we determined that the CAS-reactive material released into the supernatants could stimulate the growth of iron-starved L. pneumophila, including wild-type strains and a mutant that lacks FeoB (18). Based on mutant analysis, we also identified, at that time, two linked genes, lbtA and lbtB, that are required for expression of the siderophore (18). Cytoplasmic LbtA has similarity to bacterial siderophore synthetases, and LbtB, a member of the major facilitator superfamily, is analogous to inner membrane exporters from a variety of bacteria. Thus, we defined legiobactin as representing a CAS-positive, bioassay-positive activity whose synthesis is lbtA dependent. Based on the combined results obtained from Southern blot analyses using an lbtA probe and CAS assays testing deferrated culture supernatants, many but not all Legionella species produce legiobactin (18–20). Recent studies have identified both LbtU as a novel, TonB-independent receptor for legiobactin and LbtC as an inner membrane, ABC-type permease that helps mediate legiobactin assimilation (13, 21). We also demonstrated that cytochrome c₄ is required for optimal siderophore production, although the means by which this cytochrome promotes legiobactin expression are unclear (22). Despite these various advances in our understanding of the genetics and membrane transport of legiobactin, elucidation of the structure of the L. pneumophila siderophore has remained elusive.

Here, we report the determination of the structure of legiobactin based on mass spectrometry (MS) and nuclear magnetic resonance (NMR) analyses and prove the ability of purified legiobactin to mediate iron uptake and assimilation. Interestingly, legiobactin was identical in structure to rhizoferrin, a siderophore produced by several fungi and unrelated bacteria.

MATERIALS AND METHODS

Bacterial strains and growth media.

L. pneumophila 130b (American Type Culture Collection strain BAA-74) was used as both our wild-type strain and the parental control for mutant strains (18). Mutants tested in this study were lbtA mutant NU302, lbtC mutant NU305, and lbtU mutant NU383 (18, 21). Complemented derivatives of the lbtC mutant and the lbtU mutant were also previously described (13, 18, 21). Legionellae were routinely cultured at 37°C on buffered charcoal yeast extract (BCYE) agar or in buffered yeast extract (BYE) broth (23). When appropriate, kanamycin (25 μg/ml), gentamicin (2.5 μg/ml), and chloramphenicol (3 μg/ml) were added to the BCYE agar at the indicated concentrations. The BCYE agar and BYE broth typically contain 0.25 g of ferric pyrophosphate per liter (24), but, as indicated below, this iron supplement was sometimes omitted in order to achieve iron-limiting growth conditions. In order to collect supernatants that contained siderophore, strains were grown in deferrated chemically defined medium (CDM) as previously described (18). To monitor the growth of L. pneumophila in BYE or CDM broth, the optical density at 660 nm (OD₆₆₀) of the cultures was determined using a DU720 spectrophotometer (Beckman, Fullerton, CA).

Siderophore assays.

To test for the presence of iron-chelating activity in culture supernatant fractions, the CAS assay was performed as previously described, with the exception that the CAS reagent used included 5-sulfosalicylic acid in order to speed up the reaction (18, 25). Fractions were also tested for siderophore biological activity by assessing their capacity to stimulate the growth of the NU269 feoB mutant on iron-limited agar (10, 18). For this bioassay, 50 μl of the sample was placed into a well cut into the center of a non-iron-supplemented BCYE agar plate onto which had been spread 10⁴ CFU of the mutant and then growth was judged after 4 days of incubation at 37°C.

Purification of legiobactin by HILIC.

In order to obtain bacterial supernatants to be used for the purification of legiobactin, strain 130b was first cultured in BYE broth that lacked the usual iron supplement. To that end, bacteria that had been grown for 1 day on BCYE agar were inoculated into 1 liter of the non-iron-supplemented BYE broth (in a 2-liter flask) to an OD₆₆₀ of 0.3 and then incubated at 37°C with shaking at 225 rpm. Upon reaching the log phase, i.e., an OD₆₆₀ of 1.0, 50-ml aliquots of the culture were subjected to centrifugation at 5,000 × g for 5 min, and then the bacterial pellets were washed twice with 35 ml of deferrated CDM base buffer (50 mM morpholinepropanesulfonic acid [MOPS], 2 mM KH₂PO₄, 50 mM NaCl, pH 6.5). Bacteria were next resuspended in 1 liter of deferrated CDM (in a 2-liter flask) to an OD₆₆₀ of 0.3 and incubated for 18 h at 37°C with shaking at 225 rpm. After centrifugation of the resulting culture, supernatants were sterilized by passage through a 0.22-μm-pore-size polyethersulfone filter (EMD Millipore, Billerica, MA) and then acidified with HCl at a concentration of 0.04 M. The legiobactin-containing supernatants were concentrated using disposable cartridges containing a mixed-mode, reversed-phase/weak-anion-exchange (WAX) resin (Waters Corporation, Milford, MA). We specifically used 6-ml WAX cartridges (catalog no. 186004647) that contained 500 mg of sorbent with a 60-μm-diameter particle size, and 25 ml of supernatant was passed through each cartridge. No CAS-reactive material was present in the flowthrough, indicating that legiobactin was completely retained by the resin. Material bound to the WAX resin was washed with four 5-ml increments of 50 mM MOPS in increasing order of pH levels, i.e., 2, 5, 10, and 12. Each wash was collected and monitored for both CAS reactivity and the ability to rescue the growth of the feoB mutant. CAS-reactive, bioassay-positive material was detected only in the second and third elutions at pH 10, and the material was then lyophilized and stored at −80°C until further purification. After removal from −80°C storage, lyophilized supernatant samples, corresponding to 40 ml of elution material from the WAX column, were resuspended in 5 ml of 100% acetonitrile and then subjected to centrifugation at 5,000 × g for 5 min. No CAS-reactive or bioassay-positive material was present in the 100% acetonitrile supernatant. The insoluble material remaining after the extraction was resuspended in 10 to 12 ml of 65% acetonitrile–10 mM ammonium acetate (pH 9)–0.04% ammonium hydroxide. Aliquots (250 μl) of this material were then subjected to further purification using high-pressure liquid chromatography (HPLC). To that end, we used a Waters Breeze HPLC system (Waters Corp.) that was equipped with a 1525 model binary pump, model 2489 UV light/visible light (UV-Vis) detector, and a 10-by-150-mm Xbridge BEH Amide ODB Prep column with a 5-μm-diameter particle size (catalog no. 186006601). The mobile phase pumped by pump A (A) consisted of 50% acetonitrile, 10 mM ammonium acetate (pH 9), and 0.04% ammonium hydroxide. The mobile phase pumped by pump B (B) consisted of 65% acetonitrile, 10 mM ammonium acetate (pH 9), and 0.04% ammonium hydroxide. Each 25-min HPLC run began with a 2-min hold at 99% B and 1% A at a flow rate of 2.4 ml/min. Over the next 10 min, a linear gradient was run from 99% B/1% A to 1% B/99% A. At 12.01 min, an immediate gradient change back to 99% B/1% A was implemented and the reaction mixture was then held for the remainder of the run. Absorbance at 210 nm was monitored over the entire length of the 25-min run, and peaks that reached a threshold of 5% and had a slope of 0.035 V/min were collected. Samples were dried in a 60-Hz Savant SpeedVac DNA 100 concentrator (Thermo Fisher Scientific, Waltham, MA) set at 65°C for 3 to 4 h. The resulting pellets were resuspended in 100 μl of HPLC-grade water (Sigma-Aldrich, St. Louis, MO). Each fraction was tested in both the CAS assay and feoB bioassay. CAS-positive, bioassay-positive fractions were pooled, desalted using PD MidiTrap G-10 columns (catalog no. 28-9180-11; GE Healthcare Life Sciences, Piscataway, NJ), and dried in the SpeedVac apparatus prior to MS and NMR analyses.

MS analysis.

In preparation for mass spectrometry measurements, the desalted, HPLC-purified samples were reconstituted in 100 μl of a 50:50 solution of HPLC-grade acetonitrile (Sigma-Aldrich) and water, and then liquid chromatography separation was performed on an Agilent 1200 series HPLC system (Agilent, Santa Clara, CA) with a binary pump, where solvent A consisted of 0.1% formic acid–water and solvent B consisted of 0.1% formic acid–acetonitrile. An Agilent Poroshell 120 EC C₁₈ column (50 mm length by 2.1 mm internal diameter; 2.7 μm pore size) was utilized for all runs. A 20-μl volume of sample was injected per run, and each run began with a 2-min hold at 100% A at a flow rate of 0.25 ml/min. Over the next 3 min (i.e., min 2 to min 5), a linear gradient was applied until 100% B was reached, and this was held for the next 3 min (i.e., min 5 to min 8). At 8 min, a linear gradient was applied until 100% A was reached by min 8.1 and this was held for the next 1.7 min until the run completed. More than 80% of the legiobactin came off at the 2-min point, with some residual material appearing at the 5-min point at the start of the organic phase of the gradient. MS measurements were performed using an Agilent 6210A LC-time of flight (ToF) instrument with a dual-spray electrospray ionization (ESI) source and a high-resolution ToF mass analyzer. The instrument conditions were as follows: drying gas flow rate, 8 liters/min; drying gas temperature, 330°C; nebulizer gas level, 35 lb/in²/g; scan range, 100 to 3,200 atomic mass units (amu); fragmentor voltage, 230 V; skimmer voltage, 64 V; reference lock mass, 922.0098. Positive-ion spectra were collected in 2-GHz detector mode, and all data acquisition and analyses were done using Agilent MassHunter software.

NMR analysis.

A sample for NMR analysis was prepared by dissolving 0.1 mg of HPLC-purified legiobactin in 0.5 ml deuterium oxide (D₂O), and the resulting clear solution was transferred into a 535-pp NMR tube (Wilmad, Vineland, NJ) (outer diameter, 5 mm). All NMR experiments, except the one-dimensional (1D) carbon experiment, were carried out on an Agilent DD2 600 MHz NMR spectrometer equipped with a Varian HCN triple-resonance cold probe (5 mm). All data were collected at 25°C and without sample spinning. The regular 1D proton data were acquired by using a delay time of 1 s, an acquisition time of 2.7 s, a spectral width of 6,000 Hz, and 8 scans. The proton-carbon 2D heteronuclear single-quantum coherence (HSQC) experiment was applied using gradient and adiabatic pulses and 1 s of delay, 0.15 s of acquisition time, and 16 scans with 128 increments. The proton-carbon 2D heteronuclear multiple-bond correlation (HMBC) data were recorded under conditions similar to those used for the HSQC, except that there were 32 scans and 200 increments. The 1D carbon spectrum was obtained on an Avance III 500 MHz spectrometer (Bruker, Billerica, MA) equipped with a 5-mm-outer-diameter DCH Cryoprobe, which has a much better signal-to-noise ratio, 1,400:1, for 13C NMR. Due to the diluted concentration of the NMR sample, the quaternary carbons were barely seen with 8,192 scans. However, those quaternary 13C peaks could be detected by 2D HMBC, even when those peaks were very weak from the 1D 13C spectrum. All the NMR data were processed and analyzed by MNova software version 9.0 (Mestrelab Research, Santiago de Compostela, Spain).

Siderophore utilization and iron uptake assays.

In order to confirm the biological activity of purified legiobactin, we assayed the molecule's ability to promote the uptake of radiolabeled iron by L. pneumophila. To that end, purified legiobactin was collected from 11 HPLC runs, pooled, desalted, and dried. In the next step of iron loading, purified siderophore was resuspended in 35 μl of a 0.192 mCi/ml stock of ⁵⁵FeCl₃ (PerkinElmer, Boston, MA) that had been diluted in 10 mM HCl and then allowed to incubate at room temperature for 14 h. Next, excess unbound ⁵⁵Fe was precipitated by incubation with 0.5 M dibasic sodium phosphate buffer for 30 min at room temperature (26). The insoluble material was pelleted by centrifugation at 21,000 × g for 10 min at room temperature, leaving soluble, iron-loaded legiobactin in the supernatant. As a negative control, an identical amount of ⁵⁵FeCl₃ in the absence of added siderophore was precipitated with 0.5 M dibasic sodium phosphate and processed as described above. Finally, 50 μl of either ⁵⁵Fe-loaded legiobactin or the negative control was added to 9.5 ml of iron-starved L. pneumophila resuspended in CDM base buffer, and the amount of bacterium-associated radiolabel was quantified at 0 and 120 min as previously described (11). For this assay, wild-type strain 130b had been grown in deferrated CDM for 14 h, washed, and resuspended to an OD₆₆₀ of 1.0. At each time point, 1 ml of the bacterial suspension was removed and filtered through a 0.45-μm-pore-size nitrocellulose membrane (Millipore, Billerica, MA), and then the membranes were washed with 0.5% thioglycolic acid to remove free iron. Cell-associated radiolabel was quantified using a Beckman LS6500 scintillation counter (Beckman, Fullerton, CA) with data collected as the mean counts per minute over a 5-min period. To examine the ability of rhizoferrin to deliver iron to L. pneumophila, an identical procedure was followed, with 58 μg of purified, iron-free rhizoferrin (EMC Microcollections, Tübingen, Germany) being iron loaded upon overnight incubation with 35 μl of the 0.192 mCi/ml stock of ⁵⁵FeCl₃.

To assess the ability of the wild-type and mutant L. pneumophila strains to utilize purified siderophore for growth stimulation, 10⁴ CFU of each strain was spread onto non-iron-supplemented BCYE agar that contained 10 μM deferoxamine mesylate (DFX) to further decrease the iron content of the media, and then either 50 μl of HPLC-purified legiobactin or 50 μl of a 1 mg/ml solution of purified iron-free rhizoferrin was placed into the center well. Bacterial growth around the wells was indicative of an ability to use the supplied siderophore (13). Positive-control wells contained 5 mM ferric pyrophosphate, and negative-control wells contained HPLC-grade water, which was the solvent for purified legiobactin and rhizoferrin. To test the complemented lbtC mutant, 1 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) was added into the agar in order to induce expression of plasmid-encoded lbtC (13).

RESULTS

Purification of legiobactin.

Culture supernatants obtained from wild-type strain 130b grown in deferrated CDM were first concentrated using a mixed-mode, reversed-phase/weak-anion-exchange cartridge. CAS-reactive, bioassay-positive fractions eluted at pH 10 were then subjected to HPLC using hydrophilic interaction liquid chromatography (HILIC)-based-column chemistry and a detection wavelength of 210 nm. HPLC analysis demonstrated the presence of a strong CAS-reactive peak that eluted at 7.9 min (Fig. 1A). This peak also contained bioactive material that rescued the growth of iron-starved, feoB mutant bacteria (inset, Fig. 1A). In order to determine if the CAS-reactive, bioassay-positive material was legiobactin, we obtained concentrated culture supernatants from an lbtA mutant and then analyzed them by HPLC in a manner identical to that employed for the wild-type supernatants. The peak at 7.9 min that had been observed in the 130b supernatants was lacking in the lbtA mutant samples (Fig. 1B). Importantly, when the mutant's HPLC fraction that encompassed the region where legiobactin elutes from wild-type samples was collected and analyzed, no CAS activity or bioactivity was detected (inset, Fig. 1B). These results indicated that the CAS-reactive, bioassay-positive peak at 7.9 min represented purified legiobactin. Indeed, after the residual ammonium acetate carried over from the mobile phase of the HPLC was subtracted, the sample proved to be at least 98% legiobactin, based on the peak integration values in the proton NMR spectrum (see below).

FIG 1.

HPLC purification of legiobactin. CDM culture supernatants collected from wild-type strain 130b (WT) and lbtA mutant NU302 (lbtA-) were concentrated by a mixed-mode reversed-phase/weak-anion-exchange column, lyophilized, and then injected into an HPLC system equipped with an Xbridge BEH Amide ODB Prep column. Samples taken were then tested for the presence of siderophore activity, as measured by both the CAS assay and bioassay. The inset in panel A depicts the CAS-positive and bioassay-positive peak that eluted from WT supernatants at 7.9 min, and the inset in panel B denotes the corresponding region from mutant supernatants that was devoid of activity. AU, absorbance units.

MS analysis indicates structural similarity between legiobactin and rhizoferrin.

As a next step toward determining the structure of legiobactin, we used high-resolution MS to discern the molecular mass of the HPLC-purified legiobactin. The MS analysis demonstrated two major peaks with masses of 437.140 Da and 459.122 Da (Fig. 2A). These masses were consistent with protonated and sodiated forms. To further confirm that the production of these species is linked to legiobactin, the HPLC fraction at 7 to 8.5 min (Fig. 1B) was collected from the lbtA mutant supernatants and analyzed by MS. Overall, very low counts, representative of near-background noise levels, were obtained upon analysis of the mutant samples (Fig. 2B). Based on the high sensitivity of our instrumentation, the 437.126-Da and 459.128-Da species that appeared to be present for the mutant were not conclusively matched to the two species seen for the wild-type, especially given their very low signal levels and background noise interference. A small amount of sample carryover during the analysis might have also contributed to the appearance of species that were seemingly close in mass to 437.140 Da and 459.122 Da. Together, these data indicated that the 437.140-Da and 459.122-Da species obtained from wild-type samples are the protonated and sodiated forms of legiobactin.

FIG 2.

MS analysis of purified legiobactin. HPLC-purified legiobactin obtained from WT strain 130b supernatants (A) and the corresponding HPLC elution from the lbtA mutant NU302 supernatants (B) were analyzed by high-resolution MS in positive-ion mode, and the spectra from the peak that eluted from min 1.9 to min 2 are presented. The data for the WT sample in panel A include the major species with m/z 437.140 and m/z 459.122, which correspond to the masses of the protonated and sodiated forms, respectively, as well as fragmentation products of the major species. Although there are various peaks labeled in the mutant sample (B) for the sake of completeness, they are at noise background levels.

In considering the MS data that were obtained for the wild-type sample, we were struck by the fact that masses of 437.140 Da and 459.122 Da are consistent with the protonated and sodiated forms of rhizoferrin, respectively (27). With a molecular formula of C₁₆H₂₄N₂O₁₂, rhizoferrin is composed of two citric acid residues linked by a putrescine bridge (Fig. 3). Support for the idea of legiobactin being structurally similar to rhizoferrin was gained from the fact that known fragmentation products of rhizoferrin (27, 28) were observed in the MS analysis. These fragmentation products included the protonated mass minus a H₂O molecule (m/z 419), minus H₂O-CO (m/z 391), minus 2H₂O-CO (m/z 373), and minus a citryl group (m/z 263) (Fig. 2A). Comparing the predicted mass to the observed mass for each species, the protonated, sodiated, and fragmentation products all had errors of less than 3 mDa, the standard for an acceptable difference between predicted and observed masses (see Table S1 in the supplemental material). Legiobactin purified from an independent set of wild-type supernatants yielded the same MS results as those presented in Fig. 2A (data not shown).

FIG 3.

Structure of rhizoferrin and its L. pneumophila counterpart, legiobactin. The siderophore, whose molecular formula is C₁₆H₂₄N₂O₁₂, consists of two citrate moieties linked by a putrescine bridge by amide bonds. To facilitate the discussion of NMR data (below), the atoms in the structure have been given numerical designations ranging from 1 to 30.

NMR analysis confirms the structural identity of legiobactin and rhizoferrin.

Next, multiple types of NMR were done in order to confirm whether legiobactin had a structure identical to that of rhizoferrin. Due to the mirror symmetry of rhizoferrin (Fig. 3), only half of the protons and carbons were required to be assigned based on the NMR analysis. From the proton 1D NMR spectrum (Fig. 4A), the methylene (CH₂) protons 3′, 3″, 6′, 6″, 13, and 14 were determined based on their chemical shifts and scalar coupling patterns. While protons at position 14 had a lower frequency signal in the range of 1.5 to 1.6 ppm, methylene protons 3 and 6 had chemical shifts in the 2.5-to-2.7-ppm region due to their proximity to an oxygen atom. The doublets observed for protons 3 and 6 were due to germinal coupling at approximately 15 Hz. The signal observed at approximately 3.2 ppm corresponded to protons at position 13, and the shift downfield was due to their close proximity to a nitrogen atom. In addition, the integration values for protons 3, 6, 13, and 14 indicated two protons at each position. The signals observed from 1.8 to 2.1 ppm were due to the presence of residual ammonium acetate that was carried over from the mobile phase utilized during the HPLC purification of legiobactin. Since purified legiobactin was dissolved in D₂O, the protons bound to either oxygen (positions 1, 5, and 10) or nitrogen (position 12) were not seen due to hydrogen-deuterium exchange.

FIG 4.

1D NMR analysis of purified legiobactin. HPLC-purified legiobactin obtained from WT strain 130b supernatants was analyzed by 1D proton NMR (A) and 1D 13C NMR (B). The numbering above the peaks corresponds to the atoms designated in the legiobactin structure in Fig. 3. The asterisks indicate peaks from the ammonium acetate that was carried over during the purification of legiobactin.

Next, purified legiobactin was analyzed by 13C NMR. Similarly to the proton NMR analysis, the symmetry of legiobactin resulted in only 8 of the 16 total carbons being observed during analysis. Although the low signal-to-noise ratio for the 1D 13C NMR made it difficult to detect all eight of the carbons, those at positions 3, 6, 13, and 14 were identified from these spectra (Fig. 4B). In order to detect the chemical shifts for the remaining four carbons, we utilized two types of 2D NMR analyses, i.e., HMBC (Fig. 5A) and HSQC (Fig. 5B). From these analyses, the other four carbons were identified, i.e., three carbonyl carbons (numbered 4, 7, and 9) and the carbon at position 2 (Fig. 3). Based on these NMR results, combined with the MS data, legiobactin was determined to have a structure identical to that of rhizoferrin (27, 29).

FIG 5.

2D NMR analysis of purified legiobactin. HPLC-purified legiobactin was analyzed by proton-carbon 2D HMBC NMR (A) and by proton-carbon 2D HSQC NMR (B). The numbering within the plots corresponds to the atomic designations in Fig. 3. To facilitate further interpretation, the cross-peaks corresponding to carbons 4, 7, and 9 in the HMBC analysis have also been magnified (see Fig. S1 in the supplemental material).

Purified legiobactin and rhizoferrin facilitate iron uptake and utilization by L. pneumophila.

In order to confirm the biological activity of purified legiobactin, we next examined its ability to promote the uptake of radiolabeled iron by wild-type strain 130b. With this aim, purified legiobactin was iron loaded by incubation with ⁵⁵FeCl₃. For the purpose of having a negative control, ⁵⁵FeCl₃ was incubated without siderophore and then processed in an identical fashion. Samples containing either ⁵⁵Fe-labeled legiobactin or the negative control were added to the iron-starved L. pneumophila, and the amount of cell-associated radiolabel was determined at 0 and 120 min. Because nonprecipitated ⁵⁵Fe can be taken up by bacteria in the absence of siderophore, there was some radioactivity in the negative control. Importantly, however, the bacteria that had been incubated with the iron-loaded legiobactin incorporated significantly more ⁵⁵Fe than did those bacteria that had been mixed with the control (Fig. 6A, left panel), indicating that legiobactin can directly facilitate iron uptake by L. pneumophila. That the difference between the negative control and the sample exposed to legiobactin was not greater was most likely due to the limited amount of purified siderophore that we added into the mixture. Since legiobactin had a structure identical to that of rhizoferrin, we hypothesized that purified rhizoferrin isolated from another organism would also be capable of delivering iron to L. pneumophila. To test this, we obtained rhizoferrin that had been purified from the fungus Cunninghamella elegans (30) and repeated the radiolabeled-iron-uptake assay. Bacteria that had been exposed to iron-loaded rhizoferrin incorporated more ⁵⁵Fe than did those incubated with the negative control (Fig. 6A, right panel), indicating that rhizoferrin from an alternative source is also capable of delivering iron to L. pneumophila. Next, we determined if wild-type strain 130b could use both purified legiobactin and fungal rhizoferrin to support its growth under iron-limiting conditions. Comparably to the growth stimulation observed following exposure to ferric pyrophosphate, the ability of L. pneumophila to grow on iron-depleted BCYE agar was rescued by addition of purified legiobactin or rhizoferrin (Fig. 6B). Together, these data indicate that L. pneumophila can directly utilize both endogenously produced legiobactin and exogenously derived rhizoferrin for the purpose of Fe³⁺ uptake and assimilation and growth under iron-limiting conditions.

FIG 6.

Legiobactin and rhizoferrin uptake and utilization by L. pneumophila wild type and mutants. (A) WT strain 130b that had been grown in deferrated CDM was incubated with ⁵⁵Fe-loaded legiobactin (horizontally striped bars, left panel), ⁵⁵Fe-loaded rhizoferrin (black bars, right panel), or a negative control lacking added siderophore (gray bars), and then the amount of cell-associated radiolabel was determined at 0 and 120 min. The asterisks indicate levels of label uptake that were significantly higher than that of the negative control (*, P = 0.013; P**, = 0.002 [two-tailed, Student's t test, unequal variances]). (B and C) WT strain 130b (B) and lbtU mutant NU383 (lbtU-), the complemented lbtU mutant (lbtU-/lbtU+), lbtC mutant NU305 (lbtC-), and the complemented lbtC mutant (lbtC-/lbtC+) (C) were spread onto non-iron-supplemented BCYE agar containing DFX, and wells cut in the agar were, as indicated, filled with water, HPLC-purified legiobactin (lbt), fungal rhizoferrin (rhizo), or ferric pyrophosphate (Fe pyro). After 4 days of incubation at 37°C, bacterial growth around the well was recorded. The results depicted in panels A, B, and C are representative of three independent trials.

LbtU and LbtC are required for the ability of L. pneumophila to use legiobactin and rhizoferrin.

Previously, we determined that the outer membrane protein LbtU and the inner membrane protein LbtC are required for the ability of L. pneumophila to utilize legiobactin, based on the inability of legiobactin-containing supernatants (obtained from wild-type strain 130b) to stimulate the growth of lbtU and lbtC mutants on iron-depleted agar media (13, 21). To confirm the role of the LbtU receptor and LbtC transporter in bacterial utilization of legiobactin, we repeated the bioassay using the lbtU and lbtC mutants and purified legiobactin. In contrast to parental strain 130b (Fig. 6B), the two mutants were unable to be rescued for growth on low-iron BCYE agar following exposure to purified legiobactin (Fig. 6C). The mutants were also unable to use fungal rhizoferrin (Fig. 6C). However, mutants that were complemented by reintroduction of an intact lbtU or lbtC gene were rescued for growth by the addition of either siderophore (Fig. 7). These data indicate that endogenously made legiobactin and exogenously derived rhizoferrin are assimilated by L. pneumophila in an LbtU- and LbtC-dependent manner.

FIG 7.

Proposed biosynthetic pathway for legiobactin. The biosynthetic pathway is divided into three steps—the production of citrate (A), the production of putrescine (B), and the conjoining of citrate and putrescine to form legiobactin (C). For each enzymatic reaction, the enzyme and gene (open reading frame [ORF]) involved are indicated to the left and right of the vertical arrows, respectively.

DISCUSSION

In determining that legiobactin is identical in structure to rhizoferrin, we confirmed that the L. pneumophila siderophore belongs to the polycarboxylate family of siderophores, which derive their iron-binding capacity from carbonyl oxygen atoms (31). We also confirmed the presence of 3 carbonyl carbons within legiobactin. A total of 8 carbons were assigned during NMR analysis, which, due to the mirror symmetry of legiobactin, translated to 16 carbons total. Finally, we determined that legiobactin is composed of methylene and amine protons. These conclusions are compatible with results that were previously obtained using the Arnow, Csáky, and Rioux assays and that had indicated the absence of any catecholate or hydroxymate structure within legiobactin (12). In the one previous effort aimed at purifying legiobactin and discerning its structure, we had also concluded, based on limited proton NMR and proton-decoupled 13C NMR, that the molecule was a polycarboxylate (14). However, unacceptably high levels of NaCl contamination from the HPLC run prevented our obtaining the MS data needed to complete the structural determination. We also now believe that a contaminant(s) copurified with legiobactin during that prior purification, causing us to both miscalculate the number of carbons as 13 and suggest the presence of methine hydrogens in addition to methylene and amine protons. Fortunately, the method of HPLC adopted in the current study eliminated the contaminants, permitted analysis by both MS and NMR methods, and yielded a complete picture of legiobactin's structure.

Knowing now that legiobactin consists of two citrate molecules linked by a putrescine bridge, we can propose a biosynthetic pathway for the siderophore (Fig. 7). To begin, citrate is produced by L. pneumophila during the Kreb's cycle, when citrate synthase, encoded by the gltA gene, condenses acetyl-coenzyme A (CoA) with oxaloacetate (Fig. 7A) (32). The production of putrescine by L. pneumophila is believed to result from the conversion of l-arginine to putrescine via a three-step process, i.e., (i) conversion of l-arginine to agmatine by an arginine decarboxylase, (ii) conversion of agmatine to N-carbamoylputrescine by an agmatine deiminase, and (iii) conversion of N-carbamoylputrescine to putrescine by an N-carbamoylputrescine amidohydrolase (Fig. 7B) (33). In addition to having the genes necessary to synthesize putrescine, L. pneumophila appears to have the ability to import exogenous putrescine-utilizing proteins analogous to the PotABCD putrescine/spermidine transporter of Escherichia coli (34). Finally, we propose that LbtA helps catalyze the final step of legiobactin synthesis, namely, the joining of two citrates via a putrescine link (Fig. 7C). That LbtA should catalyze such a reaction is compatible with the fact that LbtA has sequence similarity to various amide bond-forming enzymes that are involved in siderophore production in other bacteria (14, 18). This final, LbtA-mediated step in legiobactin synthesis could theoretically occur in one trimolecular reaction or in two reactions, with one citrate added at a time. Alternately, it is also possible that there is a yet-be-defined enzyme that works with LbtA to complete the legiobactin synthesis pathway. Although our studies were based on strain 130b, we strongly suspect that all L. pneumophila strains produce legiobactin; i.e., an examination of sequenced genomes (35–39) reveals conservation of the genes implicated in the legiobactin biosynthetic pathway (Table 1).

TABLE 1.

Presence of proteins and genes that are proposed to be part of the biosynthetic pathway of legiobactin within sequenced strains of L. pneumophila

Strain	Gene encoding indicated proteina
	Citrate synthase	Arginine decarboxylase	Agmatine deiminase	N-Carbamoylputrescine amidohydrolase	LbtA
130b	lpw14291 (100, 100)	lpw00051 (100, 100)	lpw00041 (100, 100)	lpw00061 (100, 100)	lpw13341 (100, 100)
Paris	lpp1370 (99, 100)	lpp0006 (99, 99)	lpp0005 (97, 98)	lpp0007 (98, 99)	lpp1280 (98, 99)
Phil-1	lpg1415 (100, 100)	lpg0006 (99, 99)	lpg0005 (99, 99)	lpg0007 (98, 99)	lpg1325 (98, 99)
Alcoy	lpa02078 (100, 100)	lpa00006 (99, 99)	lpa00005 (98, 99)	lpa00008 (99, 99)	lpa01955 (98, 98)
Lens	lpl1366 (100, 100)	lpl0006 (99, 100)	lpl0005 (99, 99)	lpl0007 (100, 100)	lpl1279 (99, 98)
Corby	lpc0831 (100, 100)	lpc0006 (99, 99)	lpc0005 (98, 99)	lpc0007 (99, 99)	lpc0740 (98, 98)

^aNumbers in parentheses indicate percent amino acid identity and percent similarity to the L. pneumophila 130b sequence.

With the discovery that legiobactin is identical to rhizoferrin, we can also revisit the issue of why legiobactin is critical for L. pneumophila infection of the murine lung (14). Because lbtA mutants grow normally in macrophages and lung epithelial cells in vitro, it was suggested that legiobactin might perform a key role outside intracellular replication in resident lung cells (14). To do that, legiobactin would presumably need to resist the action of host lipocalin-2, a siderophore-binding, antimicrobial protein that is present in the lung as well as on other mucosal surfaces (40–42). Importantly, it has been shown that rhizoferrin purified from fungi is not bound by lipocalin-2 (41). Given the structural identity that exists between legiobactin and fungal rhizoferrin, we hypothesize that legiobactin from L. pneumophila is also not bound by lipocalin-2; however, this remains to be determined experimentally. Therefore, it is possible that legiobactin production during lung infection may be advantageous as a consequence of allowing L. pneumophila to scavenge iron despite the presence of lipocalin-2.

Finally, we now add L. pneumophila to the growing list of microbes that make and use rhizoferrin. First discovered as a fungal siderophore, rhizoferrin is produced by numerous members of the Zygomycetes, including Rhizopus microsporus var. rhizopodiformis, Mucor mucedo, Phycomyces nitens, Chaetostylum fresenii, Cokeromyces recurvatus, Cunninghamella elegans, Basidiobolus microsporus, Mycotypha africana, and Mortierella vinacea (27, 30). More recently, rhizoferrin has been shown to be produced by the bacteria Francisella tularensis and Ralstonia pickettii (28, 29). Interestingly, LbtA shares 33% amino acid identity and 52% amino acid similarity with the FslA/FigA protein that is required for siderophore production by F. tularensis (28, 43). LbtA also shares sequence similarity with a putative protein (DP23_3679) encoded in the genome of R. pickettii strain ATCC 27511 (44). In the rhizoferrin produced by R. microsporus, the two chiral carbons in the citryl residues adopt an R,R configuration (27). However, in the case of the rhizoferrin made by R. pickettii, the citryl residues adopt an S,S configuration (29). It is believed that the difference in chirality between the fungal rhizoferrin and bacterial rhizoferrin is due to differences in the biosynthetic pathways (45). This would suggest that the chiral centers in legiobactin adopt an S,S configuration. In thinking about what other microbes might also produce rhizoferrin, it is notable that Coxiella burnetii (46) encodes a putative protein that is 32% identical and 51% similar to LbtA. It is not readily apparent why these many various organisms, spanning the bacterial and fungal kingdoms, should produce the same siderophore. Indeed, whereas L. pneumophila, F. tularensis, and C. burnetii are best known as intracellular parasites of amoebae and macrophages (5, 28, 43, 47, 48), R. pickettii and the above-mentioned fungi are typically free-living inhabitants of soil and water environments (30, 49). However, it is possible that these organisms share an ecological niche or inhabit similar niches in the environment in which rhizoferrin serves as an effective chelator of iron and perhaps other metals (50). Were they coexisting, it is possible that L. pneumophila would be able to utilize the rhizoferrin made by other bacteria or fungi. Although there are many examples of a single type of siderophore being produced by multiple bacterial species or multiple fungal species, there are relatively few situations analogous to that of rhizoferrin, in which a single siderophore is produced by such a wide range of microorganisms, including both fungi and bacteria (31, 51). Moving forward, we will refer to legiobactin as L. pneumophila rhizoferrin.