Listeria monocytogenes is a Gram-positive bacterium that causes severe opportunistic infections in humans and animals. We biochemically characterized, for the first time, the iron uptake processes of this facultative intracellular pathogen, and identified the genetic loci encoding two of its membrane iron transporters. Strain EGD-e used iron complexes of hydroxamates (ferrichrome and ferrichrome A, ferrioxamine B), catecholates (ferric enterobactin, ferric corynebactin) and eukaryotic binding proteins (transferrin, lactoferrin, ferritin, haemoglobin). Quantitative determinations showed 10–100-fold lower affinity for ferric siderophores (Km ≈ 1–10 nM) than Gram-negative bacteria, and generally lower uptake rates. Vmax for [59Fe]-enterobactin (0.15 pMol per 109 cells per minute) was 400-fold lower than that of Escherichia coli. For [59Fe]-corynebactin, Vmax was also low (1.2 pMol per 109 cells per minute), but EGD-e transported [59Fe]-apoferrichrome similarly to E. coli (Vmax = 24 pMol per 109 cells per minute). L. monocytogenes encodes potential Fur-regulated iron transporters at 2.031 Mb (the fur-fhu region), 2.184 Mb (the feo region), 2.27 Mb (the srtB region) and 2.499 Mb (designated hupDGC region). Chromosomal deletions in the fur-fhu and hupDGC regions diminished iron uptake from ferric hydroxamates and haemin/haemoglobin respectively. In the former locus, deletion of fhuD (lmo1959) or fhuC (lmo1960) strongly reduced [59Fe]-apoferrichrome uptake. Deletion of hupC (lmo2429) eliminated the uptake of haemin and haemoglobin, and decreased the virulence of L. monocytogenes 50-fold in mice. Elimination of srtB region genes (Δlmo2185, Δlmo2186, Δlmo2183), both sortase structural genes (ΔsrtB, ΔsrtA, ΔsrtAB), fur and feoB did not impair iron transport. However, deletion of bacterioferritin (Δfri, lmo943; 0.97 Mb) decreased growth and altered iron uptake: Vmax of [59Fe]-corynebactin transport tripled in this strain, whereas that of [59Fe]-apoferrichrome decreased 20-fold.
Listeria monocytogenes is a ubiquitous Gram-positive bacterium that lives saprophytically in nature. This species is not a usual constituent of the human flora, but its ability to grow at 4°C allows it to contaminate foodstuffs, sometimes resulting in digestive infections that may become systemic in immunocompromised individuals. The severity of listeriosis (an overall mortality of 25–30% in spite of antibiotic therapy) mainly results from a high frequency of neurological damage in such infections. L. monocytogenes survives and multiplies in many cell types and tissues, including macrophages and hepatocytes (Portnoy et al., 2002). This intracellular survival is fundamental to its pathogenicity, and each step of its intracellular parasitism depends on the production of virulence factors (Cossart, 2002; Kreft et al., 2002). The major virulence genes so far identified cluster in two loci on the chromosome, and are controlled by a single pleiotropic regulatory activator, PrfA, which is also required for virulence (Sheehan et al., 1995; Wong and Freitag, 2004). One locus contains the hly gene that encodes the major virulence factor listeriolysin O (LLO). This protein is a pore-forming cytolysin that promotes bacterial escape from the phagosomal compartment into the cytoplasm (Portnoy et al., 2002). Inactivation of hly prevents access to the cytosol (where the bacterium usually multiplies), resulting in complete avirulence in the mouse infection model. LLO also lyses erythrocytes (Parrisius et al., 1986), and thus can provide an abundant source of iron-loaded Hb.
The proliferation of L. monocytogenes within cells, and its spread from cell to cell, suggest that the organism readily obtains iron in the intracellular environment. But, L. monocytogenes does not secrete iron chelators. Rather, it utilizes the ferric siderophores of other organisms or other iron-containing compounds, including catecholamines (Simon et al., 1995; Coulanges et al., 1996), citrate (Adams et al., 1990) and transferrin (Hartford et al., 1993). The genetic systems, membrane permeases, or surface receptors for such ferric complexes are not yet known. In the related organism S. aureus, a ferrichrome (Fc) uptake system exists with a cytoplasmic membrane-anchored binding proteins (Sebulsky and Heinrichs, 2001) and an ABC-type membrane permease (Sebulsky et al., 2000). A particularly informative report on the binding protein of this uptake system, FhuD2 (Sebulsky et al., 2003), showed that it has higher affinity for ferric hydroxamates than the homologous periplasmic binding protein of Escherichia coli, FhuD.
Like Staphylococcus (Mazmanian et al., 2003), Streptococcus (Brown et al., 2001) and Corynebacteria (Drazek et al., 2000), L. monocytogenes utilizes haemin (Hn) and Hb (Newton et al., 2005). S. aureus contains Isd-encoded, sortase-anchored surface proteins that were proposed to function in Hn/Hb uptake (Mazmanian et al., 2002; 2003; Skaar et al., 2004). L. monocytogenes contains homologous, sortase-dependent proteins in its Fur-regulated srtB locus, but these do not transport Hn/Hb (Newton et al., 2005). A cell surface-localized reductase (Deneer et al., 1995; Cowart, 2002) was postulated to transport iron in L. monocytogenes, by broadly recognizing the iron-centres of different ferric complexes, reducing the metal, and releasing Fe++ into the cell (Coulanges et al., 1997; 1998). The experiments that we report defined the biochemical parameters of ferric siderophore uptake by L. monocytogenes strain EGD-e (Glaser et al., 2001), demonstrated its ability to acquire iron from several eukaryotic iron binding proteins, identified the structural genes for transporters of ferric hydroxamates and Hn/Hb, and evaluated the virulence of EGD-e mutants with deletions in the potential iron uptake loci. Only the elimination of Hn/Hb transport (ΔhupC) increased the 50% lethal dose (LD50) of L. monocytogenes (50-fold in mice).
Iron uptake by L. monocytogenes: siderophore nutrition tests
To initially assess the ability of strain EGD-e to utilize different iron sources, we adapted the siderophore nutrition test (Wayne et al., 1976; Newton et al., 1999) to the Gram-positive organism. ApoFcA (Neilands, 1976) deprives Gram-negative bacteria of iron (Wayne et al., 1976), but it was ineffective for L. monocytogenes, which utilizes it (Table 1). α,α-bipyridyl (BP), on the other hand, restricted growth of L. monocytogenes in brain–heart infusion (BHI), and nutrition assays on BHI agar containing BP (0.1 mM) showed the ability of L. monocytogenes to obtain iron from siderophores and mammalian proteins (Table 1). Contrary to previous reports (Coulanges et al., 1996; 1997), EGD-e utilized the hydroxamates Fc and ferrichrome A (FcA). No bacteria of which we are aware are able to transport FcA. EGD-e also utilized Hn and Hb in nutrition assays, but the test did not reveal uptake of elemental iron, ferric catecholates, or from holotransferrin (HTf) or Ftn (data not shown). Regarding the eukaryotic iron binding proteins, BP has sufficient affinity to remove their iron, thereby preventing them from supplying it to L. monocytogenes. Therefore, we also conducted nutrition tests with KRMT agar, where iron is limited by alkaline pH (see Experimental procedures). These assays confirmed the stimulation of EGD-e growth by HTf, Ftn, ferric citrate (FeCit) and FeSO4 (Table 1). Subsequent experiments found transport of ferric enterobactin (FeEnt) and ferric corynebactin (FeCrn), but at very low rates that the nutrition test did not detect (see following).
Table 1. Siderophore nutrition tests and mouse infection experiments with EGD-e and its mutant derivatives.
For nutrition tests, the tabulated values represent the diameter (in mm) of the halo of growth surrounding a paper disc embedded with 10 µl aliquots of the test compound. Fc and FcA, FxB, Hb and Hn were tested on BHI agar containing 0.1 mM BP; HTf, Ftn, FeCit and FeSO4 were tested on KRMT agar plates. Fc and FcA were evaluated at 50 and 0.5 µM; the two hydroxamate siderophores always gave similar growth halos with all the strains, and the results of three or four experiments with each compound were averaged and tabulated. The concentrations of all other iron compounds are also micromolar; each tabulated value represents the mean of three or four tests, which had minimal variation. NA, not applicable; ND, no data.
Iron acquisition and growth in iron-restricted minimal media: KRM
In our hands, L. monocytogenes grew poorly in existing minimal media (Pine et al., 1989; Premaratne et al., 1991; Phan-Thanh and Gormon, 1997), and we devised Klebba's RPMI modification (KRM) to study its iron acquisition processes. RPMI contains no extrinsic iron, and its components are highly purified for cell culture applications. We added other supplements (vitamins, trace metals, casamino acids, glucose and adenine; Newton et al., 2005) to increase growth rate and maximum cell density. The concentration of adventitious iron in KRM was < 10−6 M (Newton et al., 2005), and growth in this medium rendered L. monocytogenes iron-deficient, as seen by increased doubling time and decreased final cell density. These effects were reversed by addition of elemental iron or ferric siderophores.
Tests of iron acquisition in KRM recapitulated the nutrition test results, and further showed that elemental iron (FeSO4, FeCl3), catecholate ferric siderophores (FeEnt, FeCrn) and mammalian proteins (Hb, HTf, Ftn) were effective iron sources (Fig. 1). When any of these compounds was added to KRM, EGD-e had a doubling time of about 1 h at 37°C, and reached a final density of 4–5 × 108 cells ml−1.
Binding and transport of 59Fe metal complexes
Using 59Fe (Newton et al., 1999), we determined the affinity and velocity with which EGD-e transports ferric siderophores. The adsorption of [59Fe]-complexes of apoFc, corynebactin and enterobactin to listerial cells was a saturable process with roughly the same affinity: Kd ≈ 10 nM (Fig. 2). The affinity of these binding reactions was about 100-fold lower than that of E. coli FepA and FhuA for FeEnt and Fc respectively; E. coli does not adsorb FeCrn (Annamalai et al., 2004). The capacities of EGD-e for both FeCrn and Fc was 50–60 pMol per 109 cells; the capacity for FeEnt was lower, 20 pMol per 109 cells (Fig. 2). Above 200 nM, the amounts of all three 59Fe-complexes bound to the cells began to linearly increase with concentration (data not shown), indicating a non-specific binding process in this higher range. However, with Km values of about 10 nM, this non-specific binding did not significantly distort the transport affinity and rate parameters that we observed.
Despite their negative siderophore nutrition tests, EGD-e bound and transported 59FeEnt and 59FeCrn (Table 2; Fig. 2). The listerial uptake reactions for ferric catecholates were significantly less efficient than those of Gram-negative bacteria. The overall uptake affinity of EGD-e for FeEnt (Km = 2–4 nM; Fig. 2) was 20-fold lower than the affinity of E. coli FepA for the same compound (Km = 0.1–0.2 nM; Annamalai et al., 2004; Klebba, 2004), and its uptake affinity for FeCrn was another log lower (Km = 10 nM). Second, the rate of ferric catecholate transport by L. monocytogenes was 100-fold (FeCrn; Vmax = 0.4 pMol per 109 cells per minute) to 400-fold (FeEnt; Vmax = 0.1 pMol per 109 cells per minute) slower than that of FeEnt uptake by E. coli FepA [Vmax = 40 pMol per minute per 109 cells, for the chromosomal system (Newton et al., 1999; Klebba, 2004)]. Next, the presence of FeEnt in the growth media enhanced the rate of 59FeEnt uptake (Fig. 2), suggesting that the Gram-negative ferric catecholate positively regulates the synthesis of its own transport system. This apparent positive regulation of FeEnt uptake differentiates L. monocytogenes from most Gram-negative species, which (with some exceptions, e.g. Pseudomonas) negatively regulate the FeEnt uptake (Klebba et al., 1982). Conversely, the addition of FeCrn to KRM did not stimulate 59FeCrn uptake (data not shown), intimating that FeEnt and FeCrn are acquired by separate transport systems.
Table 2. Transport properties of EGD-e and its mutants.
Bacteria were cultured in BHI broth overnight, subcultured in KRM media to stationary phase, and subcultured a second time in KRM to mid-log. Uptake of 59Fe complexes of corynebactin and apoFc was determined and plotted by Grafit 5.09 (Erithacus, London). Km (nanomolar) and Vmax (pMol per 109 cells per minute) were calculated using the ‘Enzyme Kinetics’ algorithm. The mean standard deviations of Km and Vmax determinations for FeCrn and Fc were 42% and 14%, and 41% and 8.5% respectively.
EGD + +
The overall uptake affinity of L. monocytogenes for Fc was also 100-fold lower (Km = 10 nM) than that of Gram-negative transport systems for this compound [Km of E. coli FhuA for Fc is 0.2 nM (Scott et al., 2001)]. On the other hand, the rate of Fc uptake by L. monocytogenes was about the same (Vmax = 24 pMol per 109 cells per minute) as that of E. coli (Scott et al., 2001). So, although the Fc uptake process had lower affinity in L. monocytogenes than in E. coli, the maximal rates of 59Fc transport by the two bacteria were equivalent. EGD-e transported the ferric hydroxamate roughly 20- and 200-fold faster than FeCrn and FeEnt respectively. Finally, unlike FeEnt, the presence of Fc in culture media decreased the rate of 59Fc uptake by the same cells (Fig. 2). Thus, synthesis of the Fc transport system was negatively regulated by iron availability, just as it is in Gram-negative bacteria, and these data concur with its Fur-mediated regulatory system (see following).
Iron transport by chromosomal deletion mutant derivatives of EGD-e
On the basis of their sequence similarities to iron metabolism or acquisition systems of other organisms, we chose target genes in five regions (Fig. 3) of the L. monocytogenes genome (Glaser et al., 2001), and engineered individual, precise, in frame chromosomal deletions. Some of the site-directed mutants were impaired in iron uptake, others had deficiencies in iron metabolism, and some had no phenotypes.
The fur-fhu region (2.031 Mb) encodes lmo1956, a 150-residue protein with 76% identity to fur of B. subtilis. fur precedes lmo1957–lmo1960, a potential ABC-transporter for ferric hydroxamate siderophores, and lmo1961, a putative oxidoreductase. Although we ultimately generated deletions in three different putative ferric hydroxamate ABC-transporters, only deletions in the fur-fhu region affected ferric hydroxamate transport. Δlmo1959 and Δlmo1960, which eliminated a putative binding protein and membrane ATP binding permease component respectively, both prevented the uptake of Fc, FcA and ferrioxamine B (FxB) in siderophore nutrition tests (Table 1), without affecting the uptake of Hn, Hb, transferrin or Ftn. These data demonstrate the specificity of the transport system encoded by the fur-fhu locus for ferric hydroxamates. Both Δlmo1959 and Δlmo1960 severely reduced 59Fe uptake from the iron hydroxamates. (Fig. 2, Table 2). Although Δlmo1959 and Δlmo1960 grew like wild type in BHI broth and KRM, the growth of Δlmo1960 became restricted when it was repeatedly subcultured in KRM (doubling time > 6 h; final density of 2 × 108 cells ml−1). Deletion of the putative oxidoreductase (Δlmo1961) did not impair either growth or iron uptake.
The srtB region (2.274 Mb) contains genes for a secreted, peptidoglycan-associated surface protein (lmo2185), the sortase (srtB) that anchors it to the cell wall (Bierne et al., 2004; Newton et al., 2005), and a putative ABC-transporter, lmo2182–2184. A precise deletion of lmo2183, the putative cytoplasmic permease of the srtB region, did not affect the uptake of any of the iron-containing compounds we tested. Besides siderophore nutrition tests on Fc, FcA, FxB, Hn, Hb, HTf and Ftn, we performed 59FeEnt and 59FeCrn uptake studies on the Δlmo2183 derivative of EGD-e, but we did not observe a phenotype in any of these assays. Similarly, deletions of the upstream genes lmo2186 and lmo2185[formerly called svpA (Newton et al., 2005)], and the downstream gene, srtB, did not affect iron uptake from Fc, Hb or Hn (Fig. 3; Table 2). We also tested a deletion of the unlinked sortase A structural gene, srtA (lmo0929; Bierne and Cossart, 2002; Garandeau et al., 2002) and the double mutant ΔsrtAB: neither mutation, nor the double mutation, impaired iron uptake from the compounds we surveyed.
The hup region (2.499 Mb) encodes a third ABC transporter containing an ATP binding component (lmo2429) that was annotated for potential participation in ferric hydroxamate uptake. Yet, the deletion of lmo2429 did not impair the transport of iron hydroxamates. Instead, Δlmo2429 eliminated the uptake of Hn and Hb (Fig. 3; Table 1). Because of its poor solubility, the halos around Hn in nutrition tests are smaller than those seen for Hb, but the effect of Δlmo2429 was distinct: neither Hb nor Hn was utilized by this strain. Conversely, the deletion did not affect uptake of HTf or Ftn, demonstrating the selectivity of the permease system for Hn/Hb. Thus we named the locus hupDGC for Hn/Hb uptake, and designated lmo2429 as hupC.
The feoAB region (2.184 Mb) contains a Fur box followed by contiguous feoA (lmo2104) and feoB (lmo2105) genes that encode 75 and 664 residue proteins respectively, identical in length to those of E. coli. Listerial FeoB is 34% identical and 51% similar to E. coli FeoB (Kammler et al., 1993). In aerobic (Table 1) and anerobic (data not shown) nutrition tests, and in aerobic 59FeCrn uptake experiments (Table 2), EGD-e ΔfeoB behaved like wild-type strain EGD-e.
Δfri (0.979 Mb). In our hands also, the elimination of bacterioferritin retarded bacterial growth in defined media (Olsen et al., 2005). In early log phase in KRM, EGD-e Δfri had a doubling time of about 1.25 h at 37°C, and only reached a density of 3.5 × 108 cells ml−1 in stationary phase. Δfri also decreased cytoplasmic iron availability (see following), and doubled the Vmax of FeCrn uptake. This increased efficiency of FeCrn uptake likely derived from the depletion of intracellular iron stores, resulting in overexpression of the FeCrn transporter. Unexpectedly, Δfri impaired the uptake of ferric hydroxamates in nutrition tests, and [59Fe]-apoFc in transport studies (Fig. 3).
To better understand these data, we determined the effect of Δfri on intracellular iron availability, relative to wild-type EGD-e and Δfur, by transforming the three strains with pATgfp7(Newton et al., 2005). The plasmid carries the green fluorescent protein under the control of the Fur-regulated srtB locus promoter (Fig. 4). In BHI broth at stationary phase, EGD-e/pATgfp7 was predominantly non-fluorescent (mean fluorescence < 1), indicating that intracellular iron was sufficient to repress the srtB promoter-gfp construct. A fraction of the cells (about 5%) was significantly fluorescent (mean fluorescence between 10 and 50). At the other extreme, most of the Δfur cells were fluorescent (mean fluorescence of 3), and a fraction of these bacteria (10%) was highly fluorescent (mean fluorescence between 20 and 100). The main Δfri population was intermediate between EGD-e and Δfur (mean fluorescence of 1.6), and it also contained a fraction (20%) that was highly fluorescent (mean fluorescence between 20 and 100). These data show that the Δfri mutation impaired the ability of EGD to regulate intracellular iron availability, which in some cells resulted in complete derepression of Fur-regulated promoters.
Pathogenicity of mutants in the mouse model system
We performed animal pathogenesis studies in the mouse infection model to evaluate the effects of the site-directed chromosomal mutations on bacterial virulence, by measuring LD50 values. Groups of outbred Swiss mice were inoculated by intravenous injection with different doses, in 10-fold dilutions. Mortality was scored for the following 7–10 days, and LD50 determined by the graphic probit method (Roth, 1961). Three mutations reduced the pathogenesis of L. monocytogenes in mice: Δfur, ΔhupC (Table 1) and Δfri (Dussurget et al., 2005; Olsen et al., 2005). Among loci involved in ferric hydroxamate uptake, ΔfhuD and Δlmo1961 had no effect on LD50; the decrease in growth rate conferred by ΔfhuC prevented us from fairly evaluating its impact on virulence.
Molecular analyses of L. monocytogenes FhuD, FhuC and HupC
blast analyses indicated similarity of L. monocytogenes FhuD to the E. coli periplasmic binding proteins FhuD, BtuF and MalE. The listerial protein was most similar to FhuD2 of S. aureus, a membrane anchored binding protein for ferric hydroxamates (Sebulsky et al., 2003). Like FhuD2, the listerial binding protein contains an unpaired Cys residue (C21) that may be lipidated to form a membrane anchor. Sequence analyses indicated that FhuC and HupC of L. monocytogenes, which are themselves very closely related in primary structure, are similar to know ATP binding subunits of ABC-type membrane permeases, including the crystallized proteins BtuD, HisP and MsbA of E. coli. Using their most obvious identities and similarities as a basis, we correlated the sequences of the L. monocytogenes proteins to the crystal structures of FhuD and BtuD (Fig. 5). In the case of the FhuD family, the most similar regions of primary structure distribute throughout both lobes of the binding protein, in sites that give rise to ligand-contact residues. In the ATP binding membrane permease component family, we found a higher degree of homology, which was more uniformly distributed throughout tertiary structure.
Like many infective bacteria, L. monocytogenes acquired iron from ferric siderophores and eukaryotic proteins. The experiments reported herein provide the first thermodynamic and kinetic descriptions of Gram-positive bacterial iron transport in vivo, which are of additional interest because L. monocytogenes is an intracellular pathogen. We furthermore identified, also for the first time, listerial cytoplasmic membrane transporters for particular iron-containing compounds. In light of other data (Sebulsky et al., 2003), the results suggest that these ABC-type permeases are the principal components of Gram-positive bacterial iron uptake systems.
Listeria monocytogenes utilized FcA, a hydroxamate iron complex that most other bacteria do not transport. The unsuitability of apoFcA for sequestration of iron from L. monocytogenes led us to seek another iron-deprivation agent, and BP effectively rendered the cells iron deficient. Unlike apoFcA, which is triply negatively charged, BP may permeate through membrane bilayers. We found, however, that addition of iron or iron complexes reversed growth inhibition by BP (Fig. 3, data not shown), indicating that the chelator is not generally cytotoxic to L. monocytogenes and that its effects are related to iron availability. The consistency of BP-based nutrition tests and [59Fe]-apoFc uptake studies supports this conclusion. In addition, BP was previously used to study iron metabolism or transport in both Gram (+) [Bacillus (Baichoo et al., 2002), Staphylococcus (Lim et al., 1998)] and Gram (–) [Salmomella (Reissbrodt et al., 1997; Zaharik et al., 2002; Ho et al., 2004), Vibrio (Holmstrom and Gram, 2003)] bacteria, often at 10-fold higher concentrations (1 mM) than we used (0.1 mM). The single discrepancy we observed with BP, the inability of FeEnt and FeCrn in nutrition assays, was not likely attributable to use of BP as an iron sequestration agent, because in plate tests with E. coli, BP acted just like apoFcA (data not shown). We found instead that L. monocytogenes transports both catecholate 59Fe-complexes extremely slowly, at low rates that may preclude the formation of growth halos on plates.
Site-directed chromosomal deletions identified the structural loci for two iron transport systems and elucidated other aspects of listerial iron acquisition. (i) Elimination of lmo1960 or lmo1959 reduced Fc transport more than 90%. These genes, in the fur-fhu region, encode the primary ferric hydroxamate transporter of L. monocytogenes. Nevertheless, both the ΔfhuC and ΔfhuD strains had low-level, residual 59Fc uptake (Vmax = 2 pMol per 109 cells per minute; Km = 0.1–0.2 nM), suggesting that EGD-e contains a second, slower Fc uptake system, which becomes visible when the fhuCDBG transporter is inactivated by deletion. Such transport redundancy is known in other systems. A single membrane protein may recognize multiple substrates [the Fhu permease system of E. coli transports Fc, FxB, ferric aerobactin, and ferric rhodotorulate (Rohrbach et al., 1995)], or, multiple membrane transporters may exist for a single compound [in E. coli, FepA and FecA both transport FeEnt (Annamalai et al., 2004; Zhou et al., 1995)]. In L. monocytogenes, a second Fc uptake system exists with higher affinity but lower rate, whose activity is most relevant when ferric hydroxamates are present at low concentrations. (ii) Among the 11 deletion mutants in this study, only the elimination of lmo2429 prevented the uptake of iron from Hn/Hb. This gene encodes a putative ATP binding membrane permease component with homology to other known Hn/Hb transporters, as for example those that were genetically identified in Corynebacterium[hmuV (Drazek et al., 2000)] and Streptococcus[shp (Liu and Lei, 2005)] HupC has identity/similarity to these proteins at levels of 37%/57% and 40%/61% respectively. The acquisition of iron from mammalian proteins is a common attribute of pathogens (Cornelissen and Sparling, 1994), and EGD-e ΔhupC was attenuated in mice, consistent with other findings on strains that are defective in Hn/Hb uptake (Tai et al., 1993; Stojiljkovic et al., 1995; Stevens et al., 1996; Torres and Payne, 1997). In Yersinia pestis, though, Hn/Hb utilization is not connected to virulence (Simpson et al., 2000). (iii) The elimination of bacterioferritin decreased intracellular iron availability, as shown by enhanced expression of a Fur-regulated GFP reporter construct, and by an increase in the rate of FeCrn uptake. Fc uptake in the Δfri mutant was paradoxical. The fhu locus is negatively regulated by Fur, and Δfri decreased intracellular iron availability. Therefore, we expected Δfri to enhance FhuBCDG biosynthesis, and consequently, the Fc uptake rate. Instead, the precipitous drop in Fc uptake in the Δfri strain intimates a functional relationship between bacterioferritin and the Fc transport system. These data suggest a previously unrecognized link between the iron storage system and ferric hydroxamate transport. (iv) The feoAB locus encodes a membrane uptake system for ferrous iron (Kammler et al., 1993). FeoAB homologues contribute to the virulence of Helicobacter pylori (Velayudhan et al., 2000) and Legionella pneumoniae (Robey and Cianciotto, 2002); conflicting findings exist on the relationship of FeoAB to the pathogenesis of Salmonella typhimurium (Tsolis et al., 1996; Boyer et al., 2002). EGD-e ΔfeoB was neither defective in iron transport, nor attenuated in mice.
Data that we report address a fundamental question about Gram-positive transport mechanisms: the participation of sortase-anchored proteins in iron acquisition from mammalian iron binding proteins. Whereas experiments on S. aureus indicated that proteins bound to peptidoglycan by srtA or srtB act in the uptake of iron from Hb (Mazmanian et al., 2002; 2003; Skaar et al., 2004), our analysis of ΔsrtA,ΔsrtB and ΔsrtAB mutants found that sortase-anchored proteins play no discernible role in iron (haem/Hb or ferric siderophore) utilization by L. monocytogenes. The elimination of sortase genes (ΔsrtA,ΔsrtB or ΔsrtAB) did not impair listerial iron acquisition from any of the ferric siderophores or iron binding proteins we tested. These results conflict with the idea (Mazmanian et al., 2003; Skaar et al., 2004) that sortase-anchored proteins play a major role in iron acquisition. It is conceivable, although perhaps unlikely, that these two closely related organisms acquire iron from Hb by different mechanisms. On the other hand, even if sortase-anchored iron binding proteins exist in L. monocytogenes and/or S. aureus, it remains unclear how adsorption to the PG layer may facilitate iron uptake into the cell. Once bound, no obvious driving force exists to transfer iron, ferric siderophores Hn, Hb or Fe-transferrin from the peptidoglycan polymer to CM permeases.
Listeria monocytogenes EGD-e used all of the iron sources we tested, including several (FcA, FeCrn, FxB, Hb, Hn HTf, Ftn) that are not utilized by E. coli. Relevant to this broad specificity, others postulated a surface-localized membrane reductase that catalyses iron uptake (Adams et al., 1990; Deneer et al., 1995; Barchini and Cowart, 1996; Cowart, 2002). Yet, the identification of two ABC-transporters with differential selectivity for ferric hydroxamates and Hn/Hb argues against this idea. Reduction of iron from these compounds on the cell surface will eliminate the specificity of their cytoplasmic membrane transport reactions. Furthermore, thermodynamic and kinetic barriers exist to the removal of iron from haem by biological reduction. Rather, Hn/Hb and ferric hydroxamates enter through two different ABC-transport systems. Uptake of Fc has moderately high affinity (Km = 10 nM) and a maximum velocity (24 pMol per 109 cells per minute) comparable with that of the equivalent E. coli transporter. The listerial permease components also equate with those of hydroxamate transporters in S. aureus (Sebulsky et al., 2000; 2003) and B. subtilis (Schneider and Hantke, 1993).
Bacterial strains, plasmids and media
We used L. monocytogenes strain EGD-e (Glaser et al., 2001) for mutant constructions, and the thermosensitive shuttle vector pKSV7 (Smith and Youngman, 1992) for construction of deletions and their recombination into the chromosome. E. coli DH5α was the host for plasmids prior to their introduction in EGD-e. E. coli was grown in Luria–Bertani (LB) (Miller, 1972) or trypticase soy broth (Difco); L. monocytogenes was grown in BHI broth or on BHI plates (Difco). We subcultured (1%) EGD-e strains from BHI broth into KRM (Newton et al., 2005), an iron-deficient synthetic medium based on RPMI 1640. Because insoluble Fe(OH)n forms at alkaline pH, growth at pH 7.4 renders Gram-negative bacteria iron-deficient (McIntosh and Earhart, 1976). So, in some experiments we buffered KRM with 0.1 M Tris-Cl, pH 7.4 (KRMT), to study listerial iron uptake.
Siderophores and iron binding proteins
Ferrichrome and FcA were purified from cultures of Ustilago sphaerogena (Emery, 1971). Enterobactin and corynebactin, the native siderophores of Gram-negative and Gram-positive bacteria, were purified from E. coli and B. subtilis respectively, and their iron complexes (FeEnt and FeCrn) were purified by passage over Sephadex LH20 (Wayne and Neilands, 1975; Annamalai et al., 2004). FxB was a gift from J. B. Neilands. We purchased purified Hn, bovine Hb, bovine HTf (approximately 80% saturated with Fe+++) equine Ftn, FeCit and ferrous sulphate (FeSO4) from Sigma-Aldrich (St. Louis).
For PCR we used Taq polymerase (New England BioLabs) or AmpliTaqGold (Applied Biosystems) according to the manufacturer's instructions, and oligonucleotide primers from Eurogentec, France, IDT Biotechnologies, or Invitrogen. We performed PCR-colony tests of L. monocytogenes for the presence of plasmids, or their integration into or excision from the chromosome, by resuspension of a small portion of a colony into a 50 µl PCR reaction containing appropriate primers. The annealing temperature was 50°C; the extension time depended on the size of the expected fragments (1 min per kb).
Growth in iron-limiting media
Previous experiments showed that L. monocytogenes did not utilize iron complexed by BP, so we used the chelator to sequester iron in BHI (Newton et al., 2005). After growth in BHI to OD600 = 0.1, BP was added to 0.1 mM, and the cells were shaken at 37°C until the cultures reached an OD600 of 0.9. For experiments in defined media, EGD-e and its mutants were grown overnight in BHI, subcultured into KRM, grown to stationary phase (OD600∼1.2), and then subcultured again into KRM (1%) and grown to mid-log phase.
Bacteria were grown in BHI and exposed to BP as described above, and 2 × 107 cells were plated in BHI agar containing 0.1 mM BP. Paper discs were applied to the agar, 10 µl aliquots of sterile ferric siderophores or iron binding proteins were applied to the discs, and the plates were incubated overnight at 37°C. The diameters of the growth halos were measured.
59Fe binding and uptake experiments
For binding and transport studies we prepared and chromatographically purified 59Fe complexes of corynebactin, enterobactin and apoFc (specific activity 150–1000 cpm pMol−1). We measured their binding and transport (Newton et al., 1999) over a range of concentrations, by adding appropriate amounts of 59Fe-complexes to two aliquots of 2 × 107 cells of EGD-e or its mutants, and incubating them for 15 s, or 1 h and 15 s respectively, before collecting and washing the cells on 0.2 micron filters. The 15 s aliquot measured the amount initially bound to the cells, which we subtracted from the second time-point to obtain the amount transported during a 1 h period. The long duration of the transport reactions was necessary to measure the low uptake rates we observed for some of the ferric siderophores. At each concentration, data were collected in triplicate and averaged. The Kd and capacity of 59Fe-siderophore binding were determined by using the ‘Bound-vs.-Total’ equation of Grafit 5.09 (Erithacus, Middlesex, UK), and Km and Vmax of transport were calculated using the ‘Enzyme Kinetics’ equation.
In uptake experiments on bacteria that were cultured in the presence of ferric siderophores, when the cells reached mid-log phase they were pelleted by centrifugation at 5000 g for 10 min at 4°C, washed three times with fresh KRM media to remove adsorbed iron complexes, and resuspended in fresh media before measurement of the uptake of [59Fe]-siderophore complexes.
Assessment of intracellular iron availability in EGD-e Δfri
We transformed EGD-e and its Δfur and Δfri derivatives with the Fur-regulated-srtB promoter–gfp construct pATgfp7 (Newton et al., 2005), and cytofluorimetrically measured the extent of expression of the reporter green fluorescent protein, with a Beckman Coulter Epics Elite flow cytometer.
Determination of virulence in the mouse model
Bacteria were grown in BHI medium overnight at 37°C with agitation. Pathogen free Swiss female mice (Janvier), 6–8 weeks old, were intravenously inoculated (0.5 ml per mouse) in the lateral tail vein with bacterial suspensions diluted in 0.15 M NaCl. Groups of five mice were challenged with various doses of bacteria, and the mortality was followed for 10 days. The virulence of the strains was estimated by the LD50 using the graphic probit method (Roth, 1961).
Molecular analyses of genes and proteins
We obtained chromosomal DNA sequences from the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov/) or from Institut Pasteur [http://www.pasteur.fr/externe], and subjected the translated sequences to blastp analysis (http://www.ncbi.nlm.nih.gov/blast/). Among homologues of the L. monocytogenes proteins of interest (FhuD, FhuC and HupC), we selected related binding proteins and membrane permease ATP binding subunits respectively, including several in each class whose structures are known from crystallography. We subjected these families of proteins to clustalw (http://www.ebi.ac.uk/clustalw/) and protein domain search analyses. We ultimately mapped the sequences of L. monocytogenes binding protein FhuD, and ATP binding permease components FhuC and HupC, according to the known structures of E. coli FhuD and BtuD respectively. X-ray structure co-ordinates were obtained from the Protein Data Bank (http://pdbbeta.rcsb.org/pdb).
This work was supported by OCAST Grant 6074, NIH Grant GM53836, an INSERM Post Rouge, and contributions of the US Department of State, Phillipe Foundation and J. William Fulbright Foundation to Phillip E. Klebba, and an INSERM Post Orange to Salete M. C. Newton. This work was also supported by CNRS, INSERM and Université Paris V.