Bacterial iron homeostasis

Abstract

Iron is essential to virtually all organisms, but poses problems of toxicity and poor solubility. Bacteria have evolved various mechanisms to counter the problems imposed by their iron dependence, allowing them to achieve effective iron homeostasis under a range of iron regimes. Highly efficient iron acquisition systems are used to scavenge iron from the environment under iron-restricted conditions. In many cases, this involves the secretion and internalisation of extracellular ferric chelators called siderophores. Ferrous iron can also be directly imported by the G protein-like transporter, FeoB. For pathogens, host–iron complexes (transferrin, lactoferrin, haem, haemoglobin) are directly used as iron sources. Bacterial iron storage proteins (ferritin, bacterioferritin) provide intracellular iron reserves for use when external supplies are restricted, and iron detoxification proteins (Dps) are employed to protect the chromosome from iron-induced free radical damage. There is evidence that bacteria control their iron requirements in response to iron availability by down-regulating the expression of iron proteins during iron-restricted growth. And finally, the expression of the iron homeostatic machinery is subject to iron-dependent global control ensuring that iron acquisition, storage and consumption are geared to iron availability and that intracellular levels of free iron do not reach toxic levels.

Iron, Fur, Siderophore, Ferritin, DtxR, Feo

1 Introduction

1.1 Role and properties of iron

Iron is a first row transition metal. Under physiological conditions, it mainly exists in one of two readily inter-convertible redox states: the reduced Fe²⁺ ferrous form and the oxidised Fe³⁺ ferric form. It can also adopt different spin states (high or low) in both the ferric and ferrous form, depending on its ligand environment. These properties make iron an extremely versatile prosthetic component for incorporation into proteins as a biocatalyst or electron carrier. Iron is also an abundant metal, being the fourth most plentiful element in the Earth's crust. For these reasons, iron is considered to have been the ideal choice for incorporation into proteins during the evolution of early life and was probably the key constituent in the first prosthetic moieties [1]. Iron is now absolutely required for life of all forms (although there are a few exceptions to this rule; Section 7). It participates in many major biological processes, such as photosynthesis, N₂ fixation, methanogenesis, H₂ production and consumption, respiration, the trichloroacetic acid (TCA) cycle, oxygen transport, gene regulation and DNA biosynthesis. Its biological functionality is almost entirely dependent upon its incorporation into proteins, either as a mono- or binuclear species, or in a more complex form as part of iron–sulfur clusters or haem groups. Insertion of iron into proteins allows its local environment to be ‘controlled’ such that iron can adopt the necessary redox potential (ranges from −300 to +700 mV), geometry and spin state required for it to fulfil its designated biological function.

Unfortunately, the dependence of life on iron has come at a price, for the properties of iron are not entirely satisfactory for the majority of organisms inhabiting oxic environments. Although, during the early years of life, iron is considered to have been a biologically ‘friendly’ metal, once oxygenic photosynthesis began to pollute the atmosphere with molecular oxygen (from 2.2 to 2.7 billion years ago), the prevailing iron chemistry altered significantly. The predominant form of iron switched from the relatively soluble (0.1 M at pH 7.0) ferrous state to the extremely insoluble (10⁻¹⁸ M at pH 7.0) ferric form. Thus, this valuable minor nutrient upon which life is now so dependent became scarce and growth-limiting within many ecological niches. This is particularly true for some of the oceans of the world, such as the South Pacific and Southern Ocean [2]. In addition, iron can be extremely toxic under aerobic conditions. Oxygen and reduced oxygen species interact with iron in a fashion that can be devastating for living systems [3]. So, in the presence of oxygen, iron is both poorly available and potentially toxic. Thus, in order to achieve effective iron homeostasis, organisms must balance their need to efficiently scavenge iron from their surroundings to ensure adequate supplies are maintained, with the careful management of cellular free iron levels to guard against iron-induced toxicity. The way in which bacteria meet the conflicting demands imposed by iron is the focus of this review.

1.2 Basic principles of iron homeostasis

There are essentially five strategies used by bacteria in the management of iron.

1. High-affinity iron transport enabling iron to be scavenged, in various forms, from the surroundings.

2. Deposition of intracellular iron stores to provide a source of iron that can be drawn upon when external supplies are limited.

3. Employment of redox stress resistance systems (e.g. degradation of iron-induced reactive oxygen species and repair of redox stress-induced damage).

4. Control of iron consumption by down-regulating the expression of iron-containing proteins under iron-restricted conditions.

5. An over-arching iron-responsive regulatory system that co-ordinates the expression of the above iron homeostatic machinery according to iron availability.

Although these five strategies form the general basis for iron homeostasis in bacteria, the manner in which these strategies are fulfilled varies considerably depending on prevailing environmental conditions, the adopted ecological niche and phylogeny. Since bacterial iron metabolism is best understood in Escherichia coli, this organism will be the focus of this review but some different approaches used by other bacteria in their attempts to achieve iron homeostasis will also be highlighted.

2 Iron acquisition

2.1 Introduction

The iron content of E. coli ranges from ∼10⁵ to 10⁶ atoms per cell, depending on growth conditions (derived from [4]). Similar iron levels are found in other bacteria (although they can be as high as 1.8% of dry weight) [5]. Using these values it can be calculated that, at high cell densities (10⁹ cells ml⁻¹), each generation would be expected to consume up to 10¹⁸ iron atoms per litre [6]. Unfortunately, ferric ion has a solubility of just 10⁻¹⁷ M at pH 7 providing a mere 6×10⁶ iron atoms per litre and indeed bacteria generally require iron at around 10⁻⁷ to 10⁻⁵ M to achieve optimal growth. Bacteria can close the ‘concentration gap’ between ferric iron solubility and iron requirement by solubilising ferric oxides. This can be achieved in three way: (i) lowering the external pH to render ferric iron more soluble; (ii) reducing ferric iron to the relatively soluble ferrous form; or (iii) employing ferric ion chelators as solubilising agents [7]. Chelation and reduction are the main approaches that have been adopted by bacteria.

2.2 Siderophore-based iron acquisition

2.2.1 Overview

Bacteria elaborate and secrete high-affinity extracellular ferric chelators, called siderophores (iron carriers), to solubilise iron prior to transport [8, 9]. Gram-negative bacteria take up ferri-siderophore complexes via specific outer membrane (OM) receptors in a process that is driven by the cytosolic membrane (CM) potential and mediated by the energy-transducing TonB-ExbB-ExbD system. Periplasmic binding proteins shuttle ferri-siderophores from the OM receptors to CM ATP-binding cassette (ABC) transporters that, in turn, deliver the ferri-siderophores to the cytosol (Fig. 1A) where the complexes are probably dissociated by reduction. Bacteria, such as Gram-positives, that lack an OM require neither OM receptors nor TonB-ExbB-ExbD systems (Fig. 1B). To allow ferri-siderophores to traverse their CM, they use binding-protein-dependent ABC permeases that are analogous to those employed by Gram-negatives except that the binding protein is generally a lipoprotein tethered to the external surface of the CM [8].

Figure 1

Schematic representation of siderophore-mediated iron uptake in Gram-negative (A) and Gram-positive (B) bacteria.

2.2.2 Siderophores

Siderophores are of low molecular mass (<1000 Da) and are characterised by their high specificity and affinity (K_aff10³⁰[10]) towards ferric iron. They are generally synthesised and secreted by bacteria (and fungi and monocotyledonous plants) in response to iron restriction although some, such as the mycobactins of mycobacteria, remain associated with the cell envelope [11]. Siderophores can reach extremely high concentrations, up to 200 mg l⁻¹ can be achieved for aerobactin produced by some E. coli strains [12]. Siderophores usually form hexadentate octahedral complexes with ferric iron and typically employ hydroxamates, α-hydroxycarboxylates and catechols as extremely effective Fe³⁺ ligands [13, 14]. Approximately 500 siderophores have been characterised [15] which can be classified according to the functional groups they use as iron ligands. The well known siderophores, enterobactin (a catecholate) and ferrichrome (a hydroxamate), are shown in Fig. 2. Common precursors for siderophore biosynthesis include citrate, amino acids, dihydroxybenzoate and N⁵-acyl-N⁵-hydroxyornithine [14]. Many siderophores are composed of peptides, often cyclic, assembled via non-ribosomal peptide synthetases similar to those used for the biosynthesis of peptide antibiotics. Following synthesis, siderophores are secreted across the CM. Since siderophores are generally polar, this process is likely to require specific transport proteins. Indeed, secretion of enterobactin in E. coli has recently been shown to be mediated by the EntS protein [16].

Figure 2

Structures of the siderophores enterobactin (or enterochelin) and ferrichrome.

2.2.3 Fe³⁺–siderophore receptors

Fe³⁺–siderophore complexes are too large to pass through the porins (size limit of ∼600 Da) that allow smaller solutes to permeate the OMs of Gram-negative bacteria. For this reason, the first step in the internalisation process (for Gram-negatives) requires OM receptor proteins that bind cognate ferri-siderophores with high specificity. The OM receptors are also thought to enhance the rate of ferri-siderophore uptake, allowing bacteria to more efficiency scavenge ferri-siderophores from their surroundings. This is enabled by the high affinities (K_d 0.1–100 nM) of receptors for their corresponding siderophore complexes [6, 17]. All known OM siderophore receptors are related [8] and the crystal structures of at least three (FepA, FecA and FhuA) have been determined [18–20]. They are β-barrel proteins consisting of a 22-β-stranded tube that traverses the OM allowing the siderophore complex to access the periplasm (Fig. 3). Unlike the structurally similar porins, the central channel of the siderophore receptors is ‘plugged’. The ‘plug’ or ‘cork’ consists of a globular domain derived from the first ∼160 residues at the N-terminus and is positioned in the pore towards the periplasmic end of the barrel. The plug gates the channel (external diameter of ∼4 nm) limiting access to the periplasm. Co-crystal structures of FecA with ferric citrate, and FhuA with iron-containing ferrichrome, have provided an amazing insight into the siderophore transport process [19, 20]. Binding of ferri-siderophore results in conformational changes in the external loops and periplasmic pocket which are believed to prime the liganded receptors for the next stage in the transport process. FhuA derivatives lacking the cork domain still possess the ability to act as TonB-dependent ferri-siderophore transporters [21] and it appears that the plug domain (of FepA) has only weak affinity for ferri-siderophore [22].

Figure 3

Structures of FepA and FhuA: the OM receptors for ferri-enterobactin and ferrichrome. Helices are shown as rods and β-strands as arrows. Lipids and the FhuA-dependent antibiotic, δ-2-albomycin, that co-crystallise with FhuA are also shown as space-filled molecules. Structures are from Buchanan et al. [18] and Ferguson et al. [19, 20, 174], and pictures are from the Protein Data Bank [23].

OM siderophore receptors are generally induced by iron starvation and characteristically are not present under iron-sufficient conditions. One important reason for this may be the targeting of OM receptors by bacteriophages, colicins and antibiotics as an entry point into bacterial cells. Bacteria often possess multiple OM receptors, each providing the bacterium with specificity for different siderophores. E. coli K-12 possesses at least six OM receptors that enable acquisition of eight iron–chelate complexes (Fig. 4). Of these, four (coprogen, ferrichrome, ferrioxamine and rhodotorulic acid) are produced exogenously (i.e. produced by other organisms). Only enterobactin (and its precursor and breakdown products, dihydroxybenzoate and dihydroxybenzoyl serine) is produced endogenously. It is quite typical for bacteria to utilise exogenous siderophores since, in this way, they can piratise the siderophores of their competitors and escape any bacteriostatic effect caused by exogenous siderophores. E. coli also possesses an OM receptor for ferric dicitrate and can utilise this as an iron source [24]. It should be noted that citrate is not thought to be produced by E. coli in sufficient quantities to drive iron uptake [6].

Figure 4

Schematic representation of siderophore-mediated iron uptake systems in E. coli K-12. Note that the TonB-ExbB-ExbD complex energises and interacts with all the OM receptors shown (not just FepA).

2.2.4 The TonB-ExbB-ExbD complex

Although binding of ferri-siderophores to OM receptors is independent of other factors, transport of ferri-siderophores through OM receptors requires energy. This energy is provided by the electrochemical charge gradient of the CM and is delivered by the energy-transducing TonB-ExbB-ExbD protein complex [25, 26]. Although TonB-ExbB-ExbD proteins are found in many Gram-negative bacteria, most is known about the E. coli system. In E. coli, the energy transduction process requires direct contact between TonB and OM receptors. A conserved hydrophobic seven-amino acid segment, the ‘TonB box’, at the N-terminus of TonB-dependent OM receptors has been found to be required for TonB-mediated uptake of ferri-siderophores [27]. Substitutions in this region inactivate the transport capability of the corresponding receptor, but such effects can be reversed by complementary substitutions in TonB suggesting that the TonB box region, thought to be located in the periplasm, physically interacts with the TonB protein. However, it should be noted that other regions of OM receptors are also involved in TonB interaction [28–30]. ExbB and ExbD are integral CM proteins, whereas TonB (239 amino acid residues) is periplasmic and anchored to the CM by its hydrophobic N-terminal domain. Cross-linking studies show that these three proteins form a complex in vivo, and quantification of the levels of the corresponding subunits suggests a ratio of 1:7:2 for the TonB:ExbB:ExbD subunits [31]. TonB contains a Pro-rich central domain that is thought to form an extended rigid structure that allows TonB to span the periplasmic space enabling the C-terminal domain to contact TonB-dependent receptors in the OM. It is believed that ExbB and ExbD use the membrane electro-chemical charge gradient to produce an ‘energised’ form of TonB that mediates a conformational change in the liganded OM receptor. This in turn leads to translocation of the associated ferri-siderophore to the periplasm and de-energisation of TonB [32, 33]. The de-energised form of TonB then ‘recycles’ back to the CM [34, 35]. The recently determined crystal structure of the C-terminal domain of TonB shows that it forms a cylinder composed of two intertwined subunits [36]. However, the molecular details of TonB action still remain unresolved.

In some bacteria, there is more than one TonB-ExbB-ExbD system (E. coli possesses just one). For instance, Vibrio cholerae contains two TonB proteins, TonB1 and TonB2, which appear to have specificities for different OM receptors [37]. The reason for this dichotomisation is unclear.

2.2.5 Transport across the periplasm and cytoplasmic membrane

Transport of ferri-siderophore complexes across the periplasmic space and cytoplasmic membrane is mediated by periplasmic binding proteins and associated CM transporters. Each binding protein accepts one ferri-siderophore complex at a time but it is unclear whether the binding protein collects the ferri-siderophore directly from the OM receptor, or whether the binding protein simply picks up the free ferri-siderophore from the periplasm [8]. The binding protein acts as a shuttle, collecting the ferri-siderophore released from the OM receptor and delivering it to a cognate permease in the inner membrane. Ferri-siderophore- (as well as haem- and vitamin B₁₂-) binding proteins are evolutionarily related and so are expected to possess similar structural and functional properties. The best characterised ferri-siderophore-binding protein is FhuD from E. coli. It interacts with a range of ferri-hydroxamates with dissociation constants ranging from 0.3 to 79 μM [8]. The crystal structure of FhuD shows that, like other binding proteins, it possesses a bilobal structure somewhat resembling a Venus flytrap [38, 39] (Fig. 5). The ligand-binding site is located within a shallow pocket located between the two lobes and ligand specificity is achieved through interaction of the iron-hydroxamate centre with residues in the binding pocket. The backbone of the siderophore does not directly interact with FhuD which explains how the protein is able to recognise different types of hydroxamate. FhuD is folded differently from the classical periplasmic binding protein involved in sugar and amino acid transport. It lacks the flexible hinge region and does not appear to undergo the major opening and closing movements seen for other types of binding protein upon binding and release of ligand [8].

Figure 5

Structures of the periplasmic binding proteins, FhuD of E. coli and Fbp (ferric-binding protein) of Haemophilus influenzae, complexed with their corresponding binding substrates (ferrichrome and ferric iron, respectively). Structures were determined by Clarke et al. [38, 39] and Bruns et al. [40], and the pictures were obtained from the Protein Data Bank [23].

The ABC permease complexes consist of four ‘modules’: two identical or homologous integral membrane permease modules and two ABC modules located on the inner surface of the inner membrane. The two permease modules can be supplied by two independent subunits (e.g. FepD and FepG), two copies of the same subunit (e.g. haem uptake) or by one large two-module subunit (e.g. FhuB), whereas the two ABC modules are generally composed of two identical subunits. There is good evidence that the binding protein interacts with its cognate CM permease: FhuD protects FhuB from proteolysis; FhuD and FhuB can apparently be cross-linked when combined together in a spheroplast system; and various FhuB peptides bind to purified FhuD [41, 42].

Although E. coli K-12 possesses six iron-transporting OM receptors, it only contains three associated binding-protein-dependent ABC systems: one each for the transport of ferric citrate, ferric hydroxamates and ferric catechols. Thus, the OM receptors display a greater specificity than the binding-protein-dependent ABC systems. In Pseudomonas aeruginosa, the situation appears even more extreme since the genome sequence suggests 35 TonB-dependent OM receptors but just four potentially associated ABC transporters [8].

2.2.6 Fate of internalised ferri-siderophores

Once internalised, the ferri-siderophore complex must be dissociated to liberate the complexed iron for use in cellular metabolism. This process is thought to involve reduction of the siderophore-associated iron resulting in dissociation due to the relatively low affinity of siderophores for ferrous iron. Several intracellular enzymes with ferric reductase activity have been identified in bacteria (e.g. flavin reductase enzyme, sulfite reductase, flavo-haemoglobin) but there is no strong evidence to indicate that these enzymes have any major physiological role in iron metabolism [43]. In E. coli, the utilisation of ferri-enterobactin requires the esterase protein which is encoded by the fes gene located within a cluster of genes (ent-fec) involved in enterobactin synthesis, export and uptake [44]. The esterase hydrolyses the ester bonds of internalised ferri-enterobactin producing dihydroxybenzoyl serine which, as mentioned above, functions as a weak siderophore [44]. Importantly, the esterase also seems to function as a ferri-enterobactin-specific reductase reducing the ferric iron carried by enterobactin leading to release of iron from the siderophore [44–46].

The mechanism employed to release iron from exogenous siderophore complexes is less clear. The fhuF gene of E. coli was originally identified as being required for ferrioxamine B utilisation and was assumed to be involved in transport. However, it is now known that the fhuF gene encodes a cytoplasmic ferredoxin-like protein thought to act as a ferrioxamine B reductase in the release of iron from the internalised ferri-siderophore [47, 48]. The activity of FhuF appears to be dependent on the Suf system (Section 6.2).

2.3 Ferrous iron transport and ferric reduction

Many bacteria possess a type of iron transport system, called Feo, which is quite different from the siderophore-dependent systems. Instead of ferric iron, ferrous iron is transported. The first bacterial ferrous iron transporter discovered was in E. coli and is encoded by the anaerobically induced, and iron-repressed, feoAB genes. These genes are conserved in many bacteria (although feoB is often found without feoA) and specify: FeoA, a 75-amino acid residue protein that possesses clear homology (unpublished observations) with the ‘flexible’ C-terminal domain of the iron repressor protein, DtxR (see Section 5.6); and FeoB, a membrane-bound 773-amino acid protein (FeoB) which was initially thought to be a transport ATPase and was thus presumed to use ATP hydrolysis to energise ferrous iron uptake [181, 51]. However, a recent report shows that FeoB has G protein functionality [182]. FeoB proteins consist of two main regions: a hydrophilic N-terminal domain (exhibiting GTPase activity in the E. coli protein) containing a well conserved guanine nucleotide-binding site required for Fe²⁺ uptake, and a hydrophobic C-terminal region with 7–12 predicted membrane-spanning α-helices, of which seven are highly conserved [182]. Unsurprisingly, the iron transport function of Feo appears to be of particular importance during low oxygen conditions when ferrous iron remains stable and predominates over ferric iron. Indeed, E. coli and Salmonella feoB mutants are attenuated in their ability to colonise the mouse intestine presumably due to their inability to transport ferrous iron within the anaerobic environment of the mouse gut [49, 50]. The feoB mutation does not affect Salmonella virulence indicating that ferrous uptake by FeoB is not an important route for iron acquisition during infection. In the microaerophile, Helicobacter pylori, the affinity of the FeoB system has been measured as 0.54 μM and inhibition studies indicate that FeoB is ATP-dependent [51]. This suggests that FeoB may possess ATPase, as well as GTPase, activity. FeoB is required for H. pylori colonisation of mouse gastric mucosa, as well as for normal growth and iron uptake under iron-restricted conditions [51].

Clearly, the transport of ferrous iron by bacteria would be facilitated by the ability to convert extracellular ferric iron into ferrous iron, and indeed extracellular ferric reductase activity has been identified in many bacteria, including E. coli and H. pylori [52, 53]. Such activity is also observed in eukaryotes and the reductases responsible have been identified in yeast, plants, and mammals [54–57]. However, the equivalent bacterial systems have not yet been defined.

2.4 Metal-type ABC transporters

Another type of bacterial iron transporter is the metal-transporting binding-protein-dependent ABC systems that have specificity for iron but do not necessarily require OM receptors or siderophores [8]. Such systems include the SfuABC, SitABCD, YfeABCD, FbpABC and FutABC transporters of Serratia marcescens, Salmonella typhimurium, Yersinia pestis, Neisseria gonorrhoeae and Synechocystis PCC 6803 [58–60, 177–179]. The Sit system is encoded by genes within the Salmonella pathogenicity island 1 and is required for mouse infection. However, it now appears that this system has much greater affinity for Mn²⁺ than Fe²⁺ (and cannot transport Fe³⁺), and so may not function primarily in iron transport as was originally suggested [175, 176]. The Yfe system of Y. pestis is repressed by iron and manganese, and is required for full virulence in mice and for normal growth under iron-restricted conditions. The Fut system of Synechocystis is also iron-repressed and involved in ferric iron uptake. The Sfu and Fbp systems seem to be involved in transporting iron, delivered by transferrin and lactoferrin, across the CM.

2.5 Low-affinity iron transport

Although much is known about the iron transporters used by bacteria under iron-restricted conditions, the system(s) utilised under iron sufficiency (∼10 μM iron) have not been identified. Such ‘low-affinity’ iron transport has been measured in E. coli and H. pylori but the pathways used are obscure [44, 51]. For E. coli, it has been suggested that low-affinity iron uptake is mediated adventitiously via transporters with specificities for other metals or cations [61]. Low-affinity iron uptake in H. pylori appears to be energy-independent [51]. It is suggested that this may be non-biological (i.e. iron binding to the cell surface) since in the absence of the high-affinity Feo system, low-affinity iron uptake does not enable normal growth even in the presence of high concentrations of iron [51].

2.6 Iron acquisition by pathogens

The problems that bacteria face in acquiring sufficient iron from their surroundings are particularly acute for pathogens. The host specifically limits iron availability as part of its innate defence against invading cellular micro-organisms. Mammals employ iron-binding proteins (transferrin, lactoferrin) to reduce the levels of free extracellular iron to around 10⁻¹⁸ M [62], levels insufficient to allow bacterial growth [63]. In addition, the host produces proteins that bind haem and haemoglobin (e.g. haemopexin and haptoglobin) and consequently limit the availability of haem as an iron source for pathogenic bacteria. Pathogens often use low environmental iron levels as a signal for the induction of virulence genes [63]. For instance, the Shiga-like toxin I of enterohaemorrhagic E. coli is induced by iron starvation [64]. Pathogens are able to counter the iron restriction imposed by their hosts through the use of siderophores. Siderophores can compete with host iron-binding proteins and indeed some siderophore-based transport systems (such as the plasmid pColV-K30-encoded aerobactin system found in E. coli ColV strains [65, 66]) are known to be required for effective host colonisation. However, it is common for pathogens to acquire iron directly from host iron-binding proteins by using receptor-mediated transport systems specific for host–iron complexes.

2.6.1 Transferrin and lactoferrin receptors

Independent transferrin (Tf) and lactoferrin (Lf) receptors (TfR and LfR) have been identified in pathogenic bacteria such as Neisseria species [67]. These receptors are located in the OM and are induced by iron starvation. Iron is stripped from Tf and Lf at the bacterial cell surface and the iron-free proteins are released extracellularly rather than being internalised and accumulated. The receptors appear to be able to partially discriminate between the apo and holo forms of Tf and Lf [68–71]. The receptors are composed of two subunits: Tbp1 and Tbp2 for TfR, and Lbp1 and Lbp2 for LfR. Tbp1 and Lbp1 are homologous and are related to the siderophore OM receptors [72]. These subunits possess affinity for Tf or Lf, respectively, and are likely to provide the route for Fe³⁺ translocation across the OM. Unsurprisingly, uptake of Tf and Lf iron is TonB-ExbB-ExbD- and pmf-dependent. Tbp2 and Lbp2 are also homologous but are not related to OM siderophore receptors. As for Tbp1 and Lbp1, they have affinity for Tf and Lf respectively. Tbp2 and Lbp2 appear to be lipoproteins anchored by their N-terminal lipid groups to the outer surface of the OM [73]. It is suggested that Tbp2 and Lbp2 act to increase the affinity of the receptors for Tf and Lf and to provide the ability to discriminate between the iron-loaded and iron-free proteins [67].

Transport of the iron released from Tf and Lf across the periplasm and CM is dependent upon a periplasmic binding protein ABC permease system. The periplasmic binding protein (known as the ferric-binding protein or Fbp) accepts ferric iron released from Tf or Lf and shuttles it across the periplasm to the ABC permease in the CM which in turn translocates the ferric iron across the CM into the cytosol. The structures of Fbp from N. gonorrhoeae and Haemophilus influenzae have been determined (Fig. 5). They are similar to other periplasmic binding proteins (such as phosphate-binding proteins) and consist of two similar domains arranged to form a Venus flytrap-like structure.

2.6.2 Haem uptake

The most abundant source of iron in the body is haem and so it is not surprising to find that pathogenic bacteria can use haem as an iron source. Indeed, some bacteria (e.g. Bacteroides fragilis) have an absolute requirement for exogenous supplies of haem (or its precursor, protoporphyrin IX) since they have dispensed with the biosynthetic machinery required for haem elaboration [74]. For extracellular pathogens, access to host haem requires the liberation of haem and haemoglobin from red blood cells by haemolysins and proteases. Once released, haem is rapidly bound by host proteins (haemopexin, albumin) but may also be directly transported by bacteria. Bacteria can use haem, haemoglobin or the haemopexin–haem complex as direct sources of iron. For Gram-negative bacteria, this involves binding of haem or haem complexes to OM receptors followed by transport of the isolated haem group across the OM in a TonB-ExbB-ExbD-dependent fashion. Thus the OM receptors are TonB-dependent and are related to those involved in siderophore and Tf/Lf uptake. Some bacteria secrete ‘haemophores’ that bind extracellular haemoglobin and haemopexin, and mediate their delivery to the OM receptors. Transport of haem across the CM appears to require an ABC permease (no periplasmic binding protein seems to be involved) and once in the cytoplasm the haem is degraded by haem oxygenase releasing the complexed iron. For a recent review of this subject, see Genco and Dixon [75].

3 Iron storage

3.1 Structural and functional properties

Extracellular iron is not the only source of iron available to bacteria. Many bacteria deposit intracellular reserves of iron within iron storage proteins [76]. These iron stores can then be used to enhance growth when external iron supplies are restricted. Three types of iron storage protein are recognised in bacteria: the archetypal ferritins which are also found in eukaryotes, the haem-containing bacterioferritins found only in eubacteria and the smaller Dps proteins present only in prokaryotes. Interestingly, all three types can exist in the same bacterium and multiple ferritin or bacterioferritin genes are common. Although the three types of iron storage protein form evolutionarily distinct families, they are distantly related to each other and have thus retained many structural and functional similarities. The key feature of these proteins is their molecular architecture, which provides them with their iron-storing capability. They are composed of either 24 (ferritins and bacterioferritins) or 12 (Dps proteins) identical (or similar) subunits that assemble to form an approximately spherical protein shell surrounding a central cavity that acts as an iron storage reservoir (Fig. 6). Each subunit is folded to form a four-α-helix bundle. The relatively large (∼500 kDa) ferritins and bacterioferritins can accommodate at least 2000–3000 iron atoms per 24-mer, whereas the smaller (∼250 kDa) Dps proteins have a lower storage capacity of just ∼500 iron atoms per 12-mer.

Figure 6

Structures of the 24-meric and 12-meric iron storage proteins. Structures shown are of bacterioferritin [77] and Dps [78] from E. coli, and are approximately to scale. The haem groups and Mn atoms associated with Bfr are shown as black/red and blue space-filled molecules; sodium ions in the Dps structure are in green. Pictures were obtained from the Protein Data Bank [23].

Iron storage proteins take up iron in the soluble ferrous form, but iron is deposited in the central cavity in the oxidised ferric form (Fig. 7B shows the hollow centre of a ferritin molecule). The iron storage process thus requires a ferroxidation step which is catalysed by specific sites within the iron storage proteins themselves. For ferritins and bacterioferritins, this site is the so-called ferroxidase centre located in the central region of individual subunits. The residues at this site are highly conserved and act as ligands for the binding of two ferrous ions, the first step in the iron uptake process. The bound ferrous ion pair is subsequently oxidised by O₂ which results in the formation of an oxo-bridged di-ferric intermediate. The ferric iron thus formed then migrates to the central cavity where either a ferrihydrite core is formed, or an amorphous ferric phosphate core builds up, depending on whether phosphate is present (as is the case in vivo). The ferroxidase residues are not conserved in Dps proteins. These proteins bind and oxidise ferrous iron at a completely different site located at the two-fold interface between subunits [79]. The residues involved in iron binding at this site are well conserved in the Dps proteins, but not the ferritins and bacterioferritins. Thus, the 12-meric and 24-meric iron storage proteins oxidise iron quite differently.

Figure 7

The outer (A) and inner (B) surfaces of a ferritin molecule, showing the hollow centre where iron is sequestered. Image obtained from the Research Collaboratory for Structural Bioinformatics and Protein Data Bank [23].

3.2 Dps proteins

The first Dps protein discovered was from E. coli. This protein was found to be induced in stationary phase by the sigma factor, σ^S, and to be a non-specific, DNA-binding protein with a role in protecting DNA from redox stress [117]. Subsequently, redox- and iron-induced homologues of Dps were found in other bacteria, and an iron-storing Dps-like protein was discovered in Listeria monocytogenes[80] and elsewhere. Recent work shows that the Dps of E. coli can also store iron, but it prefers H₂O₂ as the oxidant, with O₂ being rather a poor alternative [81]. This finding indicates that the primary role of Dps in E. coli is in protecting DNA against the combined action of ferrous iron and H₂O₂ in the production of the hydroxy free radical [81]. Thus, Dps probably does not have a strict function in iron storage. Whether the Dps-like proteins from other bacteria also function mainly as DNA-protecting anti-redox agents remains to be established.

3.3 Ferritins

Inactivation of the ferritin A (FtnA) gene (ftnA) of E. coli resulted in two phenotypes: a ∼50% reduction in stationary-phase cellular iron content following growth under iron-sufficient conditions and a reduced rate of growth under iron-restricted conditions [4]. The growth phenotype was only observed when strains were precultured with sufficient iron to allow iron stores to be accumulated by FtnA. This suggests that the function of FtnA is to accumulate iron during post-exponential growth in the presence of excess iron for use as an intracellular iron source during subsequent growth under iron-deficient conditions. Thus FtnA fulfils the classical iron storage role that was originally suggested for the ferritins of eukaryotes [82]. The expression pattern for ftnA matches the physiological function of FtnA –ftnA is induced by iron and by post-exponential growth. No role could be discovered for FtnA in iron detoxification or redox stress resistance, although amplification of the ftnA gene reduces the sensitivity of fur (ferric uptake regulation) mutants to redox stress [83].

Ferritins from other bacteria (Campylobacter jejuni, H. pylori) have also been found to have a role in enhancing iron-restricted growth [84, 85] and so would appear to also act primarily as iron storage proteins. However, the ferritin mutants of C. jejuni and H. pylori were more sensitive to redox stress or metals suggesting that ferritins can detoxify, as well as store, iron.

3.4 The bacterioferritins

Although bacterioferritins (Bfr) are more common in bacteria than ferritins, and were discovered in bacteria some 10 years before the ferritins, the physiological purpose of these proteins is less clear. All known bacterioferritins are haem-containing. The haem is normally in the form of protoporphyrin IX, although the Bfr of Desulfovibrio desulfuricans uses a novel type of haem, iron-coproporphyrin III [86]. There are normally 12 haem groups per 24-mer located at each of the 12 two-fold interfaces between subunits. The haem is positioned within a pocket towards the inner surface of the protein shell, with the haem being exposed to the inner cavity. The haem iron atom is coaxially ligated by a pair of methionine residues, a haem co-ordination scheme that is unique to the bacterioferritins. The haem iron is low-spin, with a relatively low redox potential (−225 mV for Azotobacter vinelandii[87]). The purpose of the haem group is uncertain, but it is likely that the presence of haem is central in distinguishing the function of the haem-free ferritins from that of the bacterioferritins. Haem-free Bfr variants from E. coli take up iron in vitro at rates indistinguishable from those of the wild-type suggesting that haem is not required for iron uptake. However, the haem-free variants accumulated approx. four-fold more iron in vivo than the wild-type protein which indicates that the haem group is involved in mediating release of iron from Bfr by facilitating reduction of the iron core [88].

Many bfr genes are associated with a gene (bfd) encoding a [2Fe–2S] ferredoxin known as Bfd (the fr-associated erre oxin). This protein is somewhat similar to FhuF which is thought to be involved in intracellular reduction of ferrichrome (Section 2.2.6). The Bfd of E. coli is induced by iron starvation and has a low redox potential (−254 mV) [89]. There is evidence that Bfd and Bfr specifically interact, which, together with the low redox potential of Bfd, suggests that Bfd may act as an iron-starvation-induced Bfr reductase mediating the release of iron from Bfr under iron restriction [89, 90]. The bfr gene of the anaerobe D. desulfuricans is associated with a gene (rd2) encoding rubredoxin-2 (a mononuclear Fe-Cys₄ protein likely to function as an electron carrier) [91]. The bacterioferritin and rubredoxin-2 form a complex in vitro and the rubredoxin-2 is able to reduce the haem group of bacterioferritin. Indeed, the redox potential of rubredoxin-2 is 115 mV lower than that of the haem of the Bfr which supports the notion that rubredoxin-2 functions in vivo to release iron from Bfr by reduction. Thus, at least two different types of electron carrier protein may act as Bfr-specific reductases in the mobilisation of iron from Bfr. Nothing is known about the in vivo release of iron from bacterial ferritins or Dps proteins, but it is presumed that this process would be mediated by reduction of the ferric iron cores.

So far, no phenotype has been identified for the bfr mutant of E. coli and it appears that Bfr does not play a major role in iron storage in this bacterium. It was also not possible to identify a phenotype for the bfr mutant of Brucella militensis[92]. However, the bfrA mutant of P. aeruginosa is sensitive to peroxides suggesting a role for Bfr in redox stress resistance. It was speculated that this phenotype arises from the use of Bfr as an iron source for the manufacture of the haem groups incorporated into the catalase encoded by the kat gene adjacent to the bfrA and bfrB genes [93]. Thus, Bfr may not be directly involved in redox stress resistance.

4 Iron and redox stress

Although it is clearly important for bacteria to secure the iron supplies required for growth, it is equally important to ensure that their intracellular iron is maintained in a safe, non-toxic form. This requires that cellular iron is not allowed to interact with reactive oxygen species in an unrestricted manner. Reactive oxygen species are partially reduced derivatives of molecular oxygen that are produced as a natural consequence of aerobic metabolism [94]. The one- and two-electron-reduction products of oxygen, namely superoxide and hydrogen peroxide, are only mildly reactive physiologically. However, iron interacts with these species to generate the highly reactive and extremely damaging hydroxyl radical. The key reactions are shown below:

1

2

3

In vivo superoxide concentrations are considered to be too low (at ∼10⁻¹⁰ M) to cause iron reduction (Eq. 1) but can be sufficiently high to damage the exposed [4Fe–4S] clusters of dehydratase-lyase family members (e.g. aconitase and α,β-dihydroxyacid dehydratase) which in turn leads to release of free iron. The relevant in vivo mediator of iron reduction during iron-induced redox stress appears to be flavins [95]. Because of this, redox stress is enhanced by factors that favour the accumulation of reduced flavins (such as inhibition of respiration [95]).

The role of iron as a major protagonist in redox stress is indicated by the increased sensitivity of bacteria to redox stress agents following growth under iron-rich conditions [4, 96]. In addition, inactivation of the fur gene (resulting in deregulation of iron metabolism) increases sensitivity to redox stress, an effect that can be reversed by iron chelation, a tonB mutation (blocking iron uptake) or by increasing iron storage capacity through overexpression of the ferritin (ftnA) gene [83]. Also, the impairment of E. coli superoxide dismutase mutants can be suppressed by further mutations that raise the intracellular concentrations of the iron chelator, dipicolinate [97]. It is probable that the enhanced sensitivity of E. coli fur mutants to redox stress is a direct consequence of increased cellular free iron. This notion is supported by the observation that fur mutants, which express iron transport systems constitutively (with respect to iron), have low levels of iron storage proteins and have increased free iron levels (although total cellular iron levels are actually reduced) [4, 98]. These findings suggest that Fur regulates the concentration of intracellular free iron through modulation of iron acquisition and iron consumption (e.g. iron storage), and that, in the absence of Fur, iron uptake and consumption are improperly balanced such that free iron levels become excessive.

5 Control of gene expression by iron

5.1 Fur – the global iron-dependent regulator of E. coli

Bacteria typically regulate their iron metabolism in response to iron availability (Fig. 8). In E. coli (and many other bacteria [99]) this regulation is mediated by the erric- ptake egulator protein (Fur) that controls the iron-dependent expression of more than 90 genes in E. coli strains [47, 100]. Fur is a homodimer composed of 17-kDa subunits [101]. It acts as a positive repressor, i.e. it represses transcription upon interaction with its co-repressor, Fe²⁺, and causes de-repression in the absence of Fe²⁺. It binds one ferrous ion per subunit, but can also bind chemically related metals (e.g. Co²⁺ and Mn²⁺) in vitro, although these metals are present at insufficient levels in vivo to be physiologically relevant in Fur interaction [102]. Metal binding increases the affinity of Fur for its DNA-binding site by ∼1000-fold. It has also been suggested that Fur binds haem [103], although the physiological significance of this is unclear. The E. coli Fur protein also contains at least one ‘structural’ (non-regulatory) zinc ion per dimer, probably at a site that includes cysteines 92 and 95 in the C-terminal half of the subunit [104, 105]. This site is not apparently conserved, and zinc is not present, in Fur proteins from all bacteria [106]. The Fur subunit can be divided into two domains, an N-terminal DNA-binding domain and a C-terminal domain, rich in His residues, thought to bind the Fe²⁺ co-effector and to mediate dimerisation [101, 107].

Figure 8

Schematic representation of Fur-mediated gene repression.

The Fe²⁺–Fur complex normally binds between the −35 and −10 sites at the promoters of Fur-repressed genes. Fur-binding sites were originally found to conform to a 19-bp palindromic consensus sequence known as the ‘iron box’ or ‘Fur box’:

It should be noted that this exact sequence is not found anywhere in the E. coli genome and Fur-binding sites may only match 11/19 bp of the consensus sequence (e.g. for tonB[108]). Also, DNase I footprinting studies show that Fur protects a region somewhat larger than the Fur box (∼31 bp) at Fur-binding sites indicating that Fur interacts with DNA outside the Fur box region. Often, Fur-binding sites consist of two or more adjacent or overlapping iron boxes suggesting the binding of several Fur dimers. Indeed, at some Fur-controlled promoters, several Fur dimers can bind to an extended region of up to ∼100 bp [109]. At the extended Fur interaction site of the aerobactin biosynthesis operon promoter (Paer), Fur dimers initially bind at a ‘primary’ (high-affinity) site which then stimulates further Fur binding at adjacent, weaker, secondary sites such that Fur appears to ‘polymerise’ along the DNA duplex [110]. Footprinting and methylation-protection studies, as well as atomic force and electron microscopy studies, indicate that multiple Fur dimers wrap around the double helix in a screw-like manner extending into regions that do not appear to match the Fur box consensus [110, 111].

Such non-consensus binding regions have led to a reinterpreted Fur box sequence as three repeats of a 6-bp motif [112]:

Multiple (more than three) repeats of the hexanucleotide motif lead to increased Fur binding, and many Fur sites can be interpreted in this fashion. Indeed, 18 tandem hexanucleotide motifs were identified in the extended Fur-binding site of Paer supporting the reinterpretation of the Fur box [109]. Thus, the reinterpretation of the Fur-binding site would appear to explain the ability of Fur binding to extend into adjacent sites lacking Fur boxes and to apparently polymerise along the DNA duplex. However, the reinterpretation by Escolar et al. [109] does not clarify exactly how individual Fur dimers are arrayed along the double helix at Fur-binding sites and precisely how many hexanucleotide repeats are required to bind a single Fur dimer. A further reinterpretation of the Fur box does provide an adequate explanation for how Fur dimers interact with Fur-binding sites. Lavrrar et al. [113] suggest that Fur-binding sites are overlapping 13-bp ‘6-1-6’ motifs:

This arrangement would allow two Fur dimers to bind at each Fur box, on opposite faces of the double helix at sites displaced by approximately half a helical turn. This arrangement would thus explain the corkscrew manner in which Fur appears to wrap around the DNA duplex. According to this model, each additional consecutive hexameric motif would allow a further Fur dimer to bind, extending the Fur-binding site accordingly and explaining the extended DNase I footprint observed for Paer. A similar reinterpretation of the Fur box was made by Baichoo and Helmann [114] for Bacillus subtilis, although they identified a binding site slightly larger (15 bp) than that of Lavrrar et al. [113].

The Fur protein is surprisingly abundant at 5000 copies per E. coli cell during exponential growth, rising to 10.000 copies in stationary phase [115]. Similarly high levels (2500) of Fur are found in V. cholerae[116]. The reason for such high levels is uncertain. This may be related to the tendency of Fur to polymerise along the DNA duplex and the large number (>90) of genes controlled by Fur in E. coli, or it might be that Fur has an additional function as a ferrous ion ‘buffer’, binding free ferrous iron in the cell. The affinity of Fur for Fe²⁺ is ∼10 μM [102] (80 μM for Mn²⁺[103]) which matches the estimated levels of free ferrous iron in the cell [98]. Thus, the affinity of Fur for Fe²⁺ is ideally set to respond to physiologically relevant fluctuations in free Fe²⁺ levels. Note that ferrous iron (as opposed to ferric iron) is considered to be the physiologically relevant form of free iron since this is the form that is utilised by proteins and the cytosolic environment is relatively reducing.

5.1.1 Regulation of the fur gene of E. coli

The E. coli fur gene is the downstream gene in the bi-cistronic fldA-fur operon. This arrangement is found in other bacteria such as Klebsiella pneumoniae, H. influenzae and Y. pestis[115]. The fldA gene is essential and encodes flavodoxin, a flavin-containing protein involved in redox chemistry. It is possible that the flavodoxin has a role in maintaining cytosolic free iron in a reduced state so it could be important in providing Fe²⁺ for the Fur protein. The fldA-fur operon is induced (10-fold) by the SoxRS system in response to superoxide-mediated redox stress [115]. In addition, the fur gene possesses its own promoter that is induced 10-fold by OxyR in response to H₂O₂-mediated redox stress. However, it appears that redox stress increases Fur levels by only two-fold (from 5000 to 10.000 copies per cell). This regulation reveals an interesting link between redox stress management and iron homeostasis. Increased Fur levels during redox stress would be expected to increase the Fe²⁺-binding capacity of the cytosol, repress iron transport and induce iron storage systems. These effects should reduce free iron levels in the cytosol during oxidative stress which should in turn help counter iron-induced toxicity.

The fur gene is also weakly autoregulated via a Fur box located in the fldA-fur intergenic region [118]. In addition, it has been reported that fur expression is under the control of the cAMP receptor protein (Crp) [118]. However, it is now clear that the fur-lacZ fusion used in the Crp regulation study did not include the true fur promoter region [115], so the Crp effect on fur expression requires re-analysis.

5.1.2 The Fur modulon

E. coli genes known to be regulated by Fur are listed in Table 1. The main physiological function of Fur is to repress the iron acquisition genes (35 known in total) under iron sufficiency, thus ensuring that the iron transport systems are induced by iron restriction (below 5–10 μM external iron concentration). However, there is evidence that genes (cyoA, flbB, fumC, gpmA, metH, nohB, nrdH, purR and sodA) with ‘non-iron’ functions (respiration, flagella chemotaxis, the TCA cycle, glycolysis, methionine biosynthesis, phage-DNA packaging, DNA synthesis, purine metabolism and redox stress resistance) are also repressed by Fur [119–122]. Thus Fur is a truly ‘global’ regulator and the genes under Fur control can be considered to form a ‘modulon’. Inactivation of the fur gene in E. coli leads to the inability to grow on non-fermentable carbon sources [123] which suggests a defect in respiration. In Salmonella, fur inactivation increases sensitivity to acid and it is suggested that Fur plays an important role in the acid stress response [124]. In addition, E. coli fur mutants are more sensitive to iron-induced redox stress [83] presumably due to increased free cytosolic iron levels (Section 4).

Table 1

Fur- and iron-regulated genes in E. coli K-12

Gene	Function	+/−	Reference(s)
acnA	Aconitase, [Fe–S] protein	+	[126]
bfd	Release of iron from Bfr?	−	[89]
bfr	Iron storage	+	[127]
cir	Ferric dihydroxybenzoate uptake	−	[63]
cyoA	Terminal respiratory oxidase subunit	−	[119]
entABCDEF	Enterobactin biosynthesis	−	[63]
entS	Export of enterobactin	−	[16]
exbBC	Siderophore and vitamin B12 transport	−	[63]
fecABCDE	Ferric dicitrate transport	−	[63]
fepA	Ferri-enterobactin transport	−	[63]
fepBCDEG	Ferri-enterobactin transport	−	[63]
fes	Ferri-enterobactin utilisation	−	[63]
fhuABCD	Ferric hydroxamate uptake	−	[63]
fhuE	Ferric coprogen rhodotorulate uptake	−	[63]
fhuF	Ferrioxamine utilisation	−	[125]
fiu	Dihydroxybenzoyl serine uptake	−	[63]
flbB	Motility	−	[119]
ftnA	Iron storage	+	[4]
fumA	Aerobic fumarase, [Fe–S] protein	+	[120]
fumB	Anaerobic fumarase, [Fe–S] protein	+	[128]
fumC	Non-[Fe–S] fumarase	+	[120]
fur	Ferric uptake regulation	−	[118]
nohA	Phage recombinase	−	[121]
nrdHIEF	Deoxyribonucleotide reductase 2	−	[121]
orf78	Unknown	−	[121]
gpmA	Glycolysis	−	[121]
metH	Methionine biosynthesis	−	[119]
nohB	Phage function	−	[119]
purR	Purine regulon regulation	−	[119]
ryhB	Small regulatory RNA	−	[127]
sdhCDAB	TCA cycle	+	[129]
sodA	Mn-superoxide dismutase	−	[130]
sodB	Fe-superoxide dismutase	+	[130]
tonB	Siderophore and vitamin B12 transport	−	[63]
ygaC	Unknown	−	[121]
yhhY	Unknown	+	[121]

‘+/−’ indicates induction (+) or repression (−) by the Fe²⁺–Fur complex.

5.1.3 Positive regulation by Fur

Several E. coli genes (acnA, bfr, ftnA, fumA, fumB, sdhCDAB, sodB; Table 1) are known to be induced by iron in a Fur-dependent manner [120, 127, 128]. However, such genes do not apparently possess a Fur box and thus seem to be indirectly regulated by Fur. The only gene known to be induced by iron through direct interaction with Fur is the ferritin (pfr) gene of H. pylori[131]. The expression of pfr is repressed by the iron-free form of Fur through direct interaction of apo-Fur with the pfr promoter at a novel Fur-binding site. Formation of an Fe²⁺–Fur complex appears to result in de-repression of pfr.

In E. coli, the most thoroughly studied Fe²⁺–Fur-induced gene is sodB. Fur induction of sodB was found to require a cis element consisting of a palindrome at the +1 site and an adjacent downstream AT-rich sequence [132]. The half-life of the sodB transcript was three-fold enhanced by Fur suggesting that Fur induction of sodB is at least partly at the post-transcriptional level. The histone-like protein, H-NS, was found to repress sodB transcription through interaction with the same cis element required for Fur induction suggesting that H-NS and Fur may compete for the same binding site [132]. However, it appears that neither Fur nor H-NS directly affects sodB transcription and Fur does not bind to the sodB message, and so it is unclear from this work precisely how Fur induces sodB at the transcriptional and post-transcriptional level [133].

The recent discovery of an Fe²⁺–Fur-repressed gene (designated ryhB) specifying a small non-coding RNA (RyhB) would appear to provide an explanation for the mechanism of Fe²⁺–Fur activation of transcription for sodB and other Fur-activated genes in E. coli[127]. RyhB appears to act as an Fe²⁺–Fur-repressed negative regulator of genes that are induced (indirectly) by the Fur complex. The Fur induction of the acnA, bfr, ftnA, sdh and sodB genes was found to be eliminated by inactivation of ryhB suggesting that their Fur dependence is mediated by RyhB (Fig. 9). It is currently unclear at what level (transcription or transcript stability) RyhB affects gene expression and what mechanism is employed. No common cis element was found to be associated with RyhB-controlled genes. Also, it is not yet clear whether all Fur-induced expression in E. coli is RyhB-mediated. However, RyhB homologues (two in Salmonella) are present in other bacteria suggesting that RyhB-mediated induction by Fur could be a common mechanism [127].

Figure 9

Roles of Fur and RyhB in mediating Fe-dependent gene regulation in E. coli (adapted from Masse and Gottsman [127]).

5.1.4 Fur controls iron consumption in E. coli

The observation that Fur induces the expression of genes encoding iron-containing proteins suggests that the biosynthesis of some iron-requiring proteins may be geared to iron availability [4, 127]. Such a mechanism would avoid the wasteful generation of iron-requiring proteins when there is insufficient iron to combine with them. In addition, the down-regulation of non-essential iron-requiring processes would increase the supply of iron for other more important processes (such as DNA biosynthesis) thus allowing available iron to be utilised more efficiently. Evidence supporting such a mechanism comes from the observation that the iron contents of E. coli fur mutants are reduced by ∼70%[4]. Part (50%) of this reduction is due to lack of iron storage proteins, but the residual 20% reduction is speculated to arise from lower levels of cellular iron proteins [4]. The ability of E. coli to modulate the expression of iron-requiring proteins in response to iron availability would explain how it can achieve and tolerate a 10-fold reduction in cellular iron content during iron-limited growth [4].

5.1.5 Functional replacement of Fe proteins by non-Fe-requiring isoenzymes

In addition to down-regulating the levels of some iron proteins during iron restriction, bacteria can also replace certain of these Fe proteins with non-iron-requiring alternatives. Under iron sufficiency, the sodB gene encoding the Fe-superoxide dismutase (SOD) enzyme is induced by Fe²⁺–Fur, whereas the sodA gene encoding the Mn-containing SOD is repressed by Fe²⁺–Fur [134]. Thus, sodA and sodB are reciprocally regulated in response to iron by Fur indicating that the cell replaces Fe-SOD with Mn-SOD when iron levels are reduced. A similar Fe²⁺–Fur-dependent regulatory relationship exists between the fumarase isoenzymes. The [4Fe–4S] fumarases (A and B) are apparently Fe²⁺–Fur-induced whereas the non-iron-dependent fumarase C is Fe²⁺–Fur-repressed [120, 128]. In some cyanobacteria, a reciprocal iron-dependent regulatory relationship exist between ferredoxin (an iron–sulfur-containing electron carrier) and flavodoxin (a flavin-containing electron carrier) levels [135, 136].

5.2 Fur in other bacteria

Following the initial discovery of Fur in E. coli, Fur homologues have subsequently been found to be responsible for iron-responsive expression in large numbers of Gram-negative bacteria and some Gram-positives [47, 137]. In many bacteria (e.g. Rhizobium leguminosarum, Synechococcus), the fur gene appears essential since it has not been possible to generate fur mutants [138, 139]. It should be noted that some Fur homologues (Zur and PerR) do not function as iron-dependent repressors: the Zur protein of E. coli is a zinc-dependent repressor and PerR of B. subtilis is a peroxide- and manganese-responsive repressor [140, 141].

5.3 Global iron-dependent regulation in P. aeruginosa and B. subtilis

A recent global analysis of iron-dependent gene expression in P. aeruginosa has revealed 205 iron-regulated genes (118 were induced by iron starvation, and 87 were repressed) [142], many of which are likely to be Fur-dependent. This number represents a high proportion (4%) of P. aeruginosa genes and confirms that iron availability can have a major influence on gene expression in bacteria. Genes induced by low iron included those encoding: iron acquisition systems (∼30), proteases, exotoxin A, fumarase C, SodA, a ferredoxin and ferredoxin reductase, and several oxidoreductases and dehydrogenases. In B. subtilis, the effects of a fur mutation and an iron chelator on global gene expression were measured [143]. Forty -six genes were found to be repressed by both iron and Fur, all of which had defined or predicted Fur-binding sites. Twenty-seven of these were known or putative iron uptake genes, and the remainder included genes encoding flavodoxins, flagella subunits, a cytochrome P450 monooxygenase homologue as well as genes involved in amino acid metabolism and nitrite reduction. No genes were found to be induced by both iron and Fur, although several cytochrome systems (e.g. cydABCD) and aconitase (citB) were reported to be weakly depressed by iron limitation. The above studies illustrate that, as in E. coli, a variety of cellular functions are affected by iron availability.

5.4 FecA-FecR-FecI and the ‘iron starvation sigmas’

The genes (fecIR fecABCDE) specifying the diferric citrate transport system of E. coli are repressed by Fe²⁺–Fur and induced by the fec-specific FecI-FecR system in response to ferric citrate [144]. FecI is a cytosolic ‘extracytoplasmic function’ (ECF) sigma factor [145] that acts as the direct regulator of the fecABCDE operon. FecR spans the cytoplasmic membrane with an N-terminal cytosolic domain (residues 1–85) and a C-terminal periplasmic domain (residues 101–317). Its periplasmic domain apparently interacts with an ‘extension peptide’ at the C-terminus of the OM receptor, FecA. In this way, FecR senses whether FecA is bound to its transport substrate, ferric dicitrate, and in response to interaction of FecA with ferric dicitrate activates FecI. Activated FecI then induces transcription of fecABCDE. The surprising feature of this system is that FecI seems to be absolutely specific for the fec operon. Thus, FecA, FecR and FecI constitute a novel three-component regulatory system for the fecABCDE operon. How FecR activates FecI is unclear – there is no evidence that FecI is covalently modified by FecR. It is suggested that, in the absence of ferric citrate, the cytoplasmic region of FecR binds FecI such that FecI is unable to function as a transcriptional activator [146]. Ferric citrate binding by FecA would then induce conformational changes leading to release of FecI and subsequent transcriptional activation.

Related ECF systems are found in other bacteria. Of these, the PvdS protein of P. aeruginosa has been particularly well studied [147]. PvdS induces the genes required for biosynthesis of the siderophore pyoverdine. FecI and PvdS are now known to be part of a subfamily of ECF sigma factors involved in iron metabolism – these have been named the ‘iron starvation sigmas’ [147]. In P. aeruginosa, there are 14 fecIR-like gene clusters, 10 of which are adjacent to known or putative siderophore receptors [147].

5.5 Other iron-dependent regulators in E. coli and Salmonella

There is evidence that the modification of some tRNA species is dependent on iron availability [148, 149], although the physiological significance of this effect is not clear. The PmrA-PmrB two-component sensor–regulator system of Salmonella enterica senses and responds to extracellular iron. It induces genes involved in lipopolysaccharide modification that are required for resistance to polymyxin B and neutrophil antimicrobial peptides in response to extracellular iron [150]. Thus Salmonella modifies its OM composition in response to extracellular iron levels.

In mammals, iron regulation is achieved at the post-translational level by iron-regulatory proteins 1 and 2 (IRP1 and 2). Under iron-replete conditions, IRP1 functions as a cytosolic [4Fe–4S]-containing aconitase (a TCA cycle enzyme), but when iron levels are low it loses its [Fe–S] cluster and so no longer functions as an aconitase. Instead, the apo-aconitase acts as an mRNA-binding protein. It specifically recognises an ‘iron regulator element’ (IRE) in the mRNA of relevant genes (e.g. ferritin, transferrin receptor) and its binding can either inhibit translation or stabilise the message, depending on the position of the IRE within the transcript. Interestingly, there is evidence that some bacterial aconitases may function similarly. In an in vitro system, the [4Fe–4S]-free aconitase A of E. coli binds to the aconitase A transcript and stabilises it, resulting in an increase in aconitase A activity [151]. There is also evidence of regulatory activity for the aconitases of B. subtilis and Xanthomonas campestris [152, 153].

5.6 DtxR

Although Fur functions as the major iron-responsive regulator in many bacteria, in high-GC-content Gram-positive bacteria (e.g. Corynebacterium, Mycobacterium, Streptomyces) the DtxR (diphtheria toxin regulator) protein (called IdeR and SirR in mycobacteria and staphylococci) mediates global iron regulation [154–156]. DtxR displays little or no sequence similarity to Fur, but as appears to be the case for Fur, it acts as an Fe²⁺-dependent repressor and binds to a hyphenated palindromic sequence as two dimers on opposite faces of the duplex. DtxR regulates iron acquisition, and expression of the diphtheria toxin, in response to iron [157]. The crystal structure of the Corynebacterium diphtheriae DtxR and that of IdeR from Mycobacterium tuberculosis have been determined [158, 159] (Fig. 10). Like Fur, DtxR has an N-terminal DNA-binding domain [159, 160] followed by a dimerisation domain likely to provide iron-sensing capability since it contains two metal-binding sites, but unlike Fur it has a third highly flexible C-terminal domain of uncertain function. The IdeR crystal structure is less clear than that of DtxR since only the N-terminal domain is sufficiently ordered to be resolved [158]. The similarities between DtxR and Fur, as described above, are striking, yet surprisingly, no clear homology exists between them. Any evolutionary relationship between these two proteins should become clear once a three-dimensional structure is provided for any Fur-like protein. Note that the DtxR proteins are homologous with the Mn²⁺-dependent repressor, MntR [183].

Figure 10

Structure of DtxR bound to its DNA-binding site. Two DtxR dimers are bound on opposite faces of the DNA duplex. Image obtained from The Nucleic Acid Database Project, Rutgers, The State University of New Jersey, USA.

6 Insertion into proteins

6.1 Regulation of haem biosynthesis in Bradyrhizobium japonicum

The generation of iron-containing prosthetic groups requires specific assembly pathways. Haem production involves the initial synthesis of an iron-free protoporphyrin, via a well-conserved pathway ron esponse egulatory protein, Irr. Irr is a member of the Fur family of bacterial transcriptional regulators (Section 5.1). In addition to Irr, B. japonicum contains a Fur protein that appears to regulate iron transport. Irr regulates the biosynthesis of δ-aminolevulinic acid dehydratase (hemB gene product) which catalyses the second step in haem biosynthesis. Irr-mediated regulation ensures that protoporphyrin production does not exceed iron availability. Irr directly interacts with ferrochelatase and seems to be able to sense whether ferrochelatase possesses haem or protoporphyrin at its active site. High iron levels allow haem formation by ferrochelatase which results in Irr binding to ferrochelatase and consequent inactivation of Irr. When iron levels are low, protoporphyrin accumulates which seems to enable Irr dissociation from ferrochelatase resulting in repression of the haem biosynthetic genes [162–164]. Thus, protoporphyrin production is controlled to ensure that it does not exceed iron availability.

6.2 [Fe–S] cluster assembly

Although iron–sulfur clusters can be produced by chemical synthesis from Fe^2+/3+- and S²⁻-releasing agents ron– ulfur luster (Isc) biosynthesis proteins [167] that utilise Fe²⁺ and l-cysteine as substrates. In E. coli, these proteins are encoded by the iscSUA-hscBA-fdx gene cluster (Fig. 11) and are required for the full activity of many [Fe–S] proteins [167]. It appears that [Fe–S] cluster generation in E. coli (as well as other bacteria) is regulated according to cellular requirements, i.e. is under homeostatic control. This regulation is mediated by the IscR protein, a SoxS homologue encoded by a gene directly upstream of, and co-operonic with, the iscSUA genes [169]. The anaerobically purified IscR protein possesses a [2Fe–2S] cluster which is thought to be required for IscR-mediated repression of iscRSUA expression. A negative feedback regulation model is proposed in which [Fe–S] proteins and IscR are in competition for the [Fe–S] clusters generated by the proposed Isc assembly complex (Fig. 11). It is presumed that IscR competes poorly with other [Fe–S] proteins for [Fe–S] clusters and so only receives clusters when [Fe–S] assembly rates exceed cellular requirements and other [Fe–S] proteins are therefore fully [Fe–S]-replete. Under such conditions, the iscRSUA operon would be repressed, but would become de-repressed should [Fe–S] assembly rates drop below cellular requirements. We are aware of no evidence that the E. coli iscR-fdx genes, or their orthologues in other bacteria, are regulated by Fur or iron. However, the sufABCDSE operon of E. coli is Fe²⁺–Fur-repressed (and induced by redox stress) [48, 170]. This operon contains homologues of iscA and iscS (namely, sufA and sufS) suggesting that is involved in [Fe–S] cluster formation. In addition, it seems to be required for [Fe–S] cluster formation in FhuF (Section 2.2.6). The sufBCD genes are highly conserved and encode the SufABC proteins that are ‘orphan’ (not obviously partnered with a membrane-integral permease) ABC proteins [180]. The proteins interact and probably form a cytosolic complex in vivo. The Suf system is required for formation and/or repair of oxygen-labile [Fe–S] clusters under oxidative stress conditions [180], and thus forms an auxiliary [Fe–S] assembly pathway.

Figure 11

Regulation of iron–sulfur cluster formation in E. coli. It is speculated that the products of the iscS-fdx gene form a complex that synthesises [Fe–S] clusters. The entire iscR-fdx operon is subject to IscR-mediated negative feedback regulation in response to [Fe–S] cluster formation. Adapted from Frazzon and Dean [168].

7 Life without iron

It is interesting to consider how life would have evolved had the Earth's pre-biotic atmosphere been aerobic. It is probable that life's dependence on iron would have been far less extreme. As it is, there are only a few organisms thought not to require iron. The Lactobacilli were the first iron-independent organisms identified. They were found to contain just one to two iron atoms per cell [171], levels considered too low to provide iron with any conceivable biological function. This lack of iron requirement presumably explains their ability to grow in milk (a highly iron-restricted medium, due to the high concentrations of the iron-binding protein, lactoferrin) and to predominate in the natural gut flora of breast-fed infants [172]. More recently, the genome sequence of the Lyme disease pathogen, Borrelia burgdorferi, was found to contain no identifiable iron-protein-encoding genes [173]. In addition, severe iron limitation had no effect on its growth and its iron content was less than 10 atoms per cell. It is speculated that this pathogen has dispensed with the need for iron in order to circumvent its host's innate iron-withdrawal defence mechanism. The syphilis pathogen, Treponema pallidum, likewise appears to lack a requirement for iron [173]. It should be noted that both B. burgdorferi and T. pallidum are obligate intracellular parasites with minimal genomes and although formally they may not require iron, they rely upon the iron-dependent metabolic processes of their host for energy and biosynthesis – so strictly speaking they still depend upon iron, albeit indirectly.

8 Note added in proof

Since this manuscript was accepted for publication, a paper describing the crystal structure of the Fur protein of P. aeruginosa has been published [184]. The structure shows that each subunit consists of an N-terminal, winger helix, DNA-binding domain (residues 1–83) that is thus structurally similar to that of DtxR and a novel C-terminal alpha/beta dimerisation domain (residues 84–135).

Author notes

1

Present address: The Microbiology Department, University College, National University of Ireland, Cork, Ireland.