Free Fe²⁺ is rarely present in the host, except under conditions where the redox potential or the pH are disturbed, such as ischemia caused by a trauma or following reduction of the environment by proliferating bacteria (Bullen et al., 2005). In bacteria, Fe²⁺ enters the periplasm through non-specific porins and is delivered to the cytoplasm through different transporters.

Under anaerobic-microaerophilic conditions, bacteria use the FeoB pathway (Cao et al., 2007). Members of the FeoB-family mediate transport of free Fe²⁺ across the inner membrane to the cytoplasm in a GTP-dependent manner. FeoB is located in the periplasmic membrane. It is expressed with FeoA and FeoC of recently characterized functions (Braun, 2001; Kim et al., 2013; Lau et al., 2013). FeoB is produced by the majority of Enterobacteriaceae Table S1; Fetherston et al., 2012; Grim et al., 2012; Hood and Skaar, 2012). In some E. coli strains, EfeUOB (YcdNOB) is also involved in Fe²⁺ uptake under aerobic, low-pH, low-iron conditions (Cao et al., 2007). As EfeO is able to oxidize ferrous iron, it has also been proposed that EfeU could transport Fe³⁺ (Rajasekaran et al., 2010). In E. coli K-12, EfeB and its paralog YfeX, widespread in Enterobacteria, may also promote iron extraction from periplasmic haem (Letoffe et al., 2009; Dailey et al., 2011). EfeUOB is also present in Y. pestis (Fetherston et al., 2012). An uncharacterized system called FetMP has also been identified in E. coli and Y. pestis as involved in ferrous iron uptake (Forman et al., 2010).

Inner membrane ATP-Binding Cassette (ABC) transporters can also import Fe²⁺. The ABC superfamily of transporters consists in several protein complexes which together are capable of transporting various solutes across membranes. They are composed of three or four different subunits, usually encoded by different genes grouped in an operon (Ma et al., 2009; Klein and Lewinson, 2011). The Mn²⁺/Fe²⁺ ABC-transporters of Enterobacteria include the Yersinia YfeABCD system and its Sit homolog identified in S. Typhimurium, ExPEC, K. pneumoniae, P. mirabilis and Shigella spp. Table S1; Angerer et al., 1992; Bearden and Perry, 1999; Kehres et al., 2002b; Sabri et al., 2008; Fisher et al., 2009; Himpsl et al., 2010).

Siderophores are small secreted molecules that display higher affinity for iron than the host iron-binding proteins such as transferrin and lactoferrin. They are synthesized in the cytoplasm and require specific export systems to reach the extracellular space. The export systems usually involve an inner membrane protein from the major facilitator family and an outer membrane channel protein such as TolC (Garénaux et al., 2011). In the extracellular space, iron-bound siderophores are recognized by specific receptors (Braun, 2001; Chakraborty et al., 2007). The TonB-ExbB-ExbD energy transduction system is required for transport of the ligand into the periplasmic space. While the inner-membrane-embedded ExbD-ExbB complex tranduces energy from the proton motive force of the cytoplasmic membrane, the TonB protein spans the periplasm to transfer this energy to the outer membrane receptor (Higgs et al., 2002). Then, a specific ABC-transporter mediates entry of the iron-bound siderophores through the inner membrane. A chaperone may be involved in the transfer of the iron-bound siderophore from the outer membrane receptor to an inner membrane ABC-transporter. It is the case for ferric enterobactin and aerobactin siderophores, requiring the FepB and FhuD chaperones, respectively (Koster and Braun, 1990; Chenault and Earhart, 1991). Once in the periplasm, ferric iron can be directly reduced to its ferrous state by reductases, so as to be released from the siderophores and transferred to iron-dependent proteins. As such, in some cases, apo-siderophores can be recycled to the extracellular space without the need for new siderophore biosynthesis. In other cases where the redox potential of the ferri-siderophore is too high, degradation by specific esterases is required before the reductases can release iron (Miethke et al., 2011).

Nearly all E. coli and K. pneumoniae strains produce enterobactin (Bachman et al., 2011; Garénaux et al., 2011). Enterobactin (or enterochelin), is internalized through the FepA receptor and FepCDG ABC-transporter thanks to the FepB chaperone (Chenault and Earhart, 1991). Some strains also produce a set of pathogen-specific siderophores potentially comprising salmochelins, aerobactin or yersiniabactin (Bachman et al., 2011; Garénaux et al., 2011). Salmochelins, glycosylated forms of enterobactin, are internalized through the IroN receptor and the same ABC-transporter as enterobactin, FepCDG. Aerobactin uptake involves the IutA receptor as well as FhuBC ABC-transporter, while yersiniabactin uptake occurs via the Psn/FyuA receptor and the YbtPQ ABC-transporter (Koster and Braun, 1990; Perry and Fetherston, 2011). The types of siderophores produced are dependent on the pathotypes of the strains. While ExPEC strains are able to synthesize up to four different siderophores, IPEC such as O157:H7 strains only produce enterobactin. Some non-O157 IPEC strains and commensals may also produce aerobactin (Kresse et al., 2007), and IPEC enteroaggregative strains also contain yersiniabactin- and aerobactin-encoding genes (Okeke et al., 2004). Likewise, all Shigella isolates produce siderophores, namely enterobactin, salmochelins or aerobactin, but the types of siderophores produced vary from one species to another (Payne et al., 2006). Salmochelins were first identified in Salmonella, which also produces enterobactin. Some Salmonella strains also produce yersiniabactin or aerobactin (Carniel, 2001; Muller et al., 2009; Izumiya et al., 2011). Siderophores produced by the different pathogenic Enterobacteria are summarized in Table S1.

Other siderophores have been identified in Enterobacteria that do not produce enterobactin. The genome of Cronobacter species contains a non-functional enterobactin gene cluster and a plasmid-encoded aerobactin cluster renamed cronobactin (Grim et al., 2012). Virulent Yersinia species are known to produce yersiniabactin as well as the alternative yersiniachelin and pseudochelin siderophores (Rakin et al., 2012). P. mirabilis produces proteobactin and a yersiniabactin-like siderophore (Himpsl et al., 2010).

In addition to the different siderophores produced by E. coli strains, it is not unusual to find more than 10 genes encoding distinct siderophore receptors in their genomes. In non-pathogenic E. coli, enterobactin can be internalized through 3 different TonB-dependent receptors: FepA, Cir, and Fiu (Andrews et al., 2003). The Iha TonB-dependent receptor mediates enterobactin uptake in UPEC (Leveille et al., 2006). Another TonB-dependent receptor identified in UPEC, IreA, is an additional potential siderophore transporter (Russo et al., 2001, 2002). Many Enterobacteria are able to internalize exogenous siderophores, such as the fungal siderophore ferrichrome, internalized through FhuABCD (Andrews et al., 2003). Low molecular weight iron chelators such as citrate can also be used as a source of iron. The typical ferric citrate transport system involves the FecA TonB-dependent receptor, the FecB chaperone and the FecCDE ABC-transporter at the inner membrane (Braun et al., 2003). These transport genes form the fecABCDE operon (Mahren et al., 2005). The S. marcescens SfuABC, the K. pneumoniae KfuABC, the E. coli and Cronobacter sp EitABC and the Y. pestis YfuABC and YiuABC systems transport ferric iron bound to small chelators across the cytoplasmic membrane (Angerer et al., 1990; Gong et al., 2001; Ma et al., 2005; Johnson et al., 2006; Kirillina et al., 2006; Grim et al., 2012).

This redundancy in siderophore-mediated iron acquisition systems suggests that their fundamental role lies in adaptation to different iron-limited niches in which bacteria are competing with other microorganisms in the environment or with host proteins for iron acquisition (Valdebenito et al., 2006).

Haem represents one of the most abundant iron sources inside the host. Organization of iron in haem allows iron solubilization, but also enhances its catalytic activity by 5 to 10 times, making it an even more efficient cofactor but increasing its toxicity (Anzaldi and Skaar, 2010). Pathogens have evolved direct haem uptake systems, as well as haemophore systems.

As haem is not freely available, bacteria secrete exotoxins such as haemolysins, proteases or cytolysins to release haem for direct uptake. Certain proteins of SPATE family (serine proteinase autotransporters of Enterobacteriacae), degrade haemoglobin to allow pathogenic bacteria to gain access to haem (Krewulak and Vogel, 2008). Haem and haemoproteins bind to specific TonB-dependent cell surface receptors, where haem contained in haemoproteins is extracted and transported to the periplasm. It is further imported by ABC transporters from the periplasm to the cytoplasm. In the cytoplasm, iron is released from haem through degradation by haem oxygenases (Hood and Skaar, 2012). S. dysenteriae encodes a haem transport system composed of shuA (coding for the haem receptor), shuT, U and V (coding for a periplasmic chaperone and an ABC-transporter) as well as shuS (which protects against haem toxicity by allowing its intracellular processing) and shuWXY coding for proteins of unknown functions (Wyckoff et al., 2005). Although the shu locus is only found in S. dysenteriae, other Shigella use haem as the sole iron source, suggesting another haem transport system (Payne et al., 2006). The Shu system is found in some E. coli strains including IPEC O157:H7 or UPEC CFT073 (Chu), Y. pestis (Hmu), P. mirabilis (Hmu) and S. marcescens (Hem) (Table S1), suggesting that horizontal transfer might have occurred (Thompson et al., 1999; Payne et al., 2006; Hagan and Mobley, 2009; Benevides-Matos and Biville, 2010; Himpsl et al., 2010). In UPEC, an additional haem receptor, Hma, has also been identified (Hagan and Mobley, 2009).

HasA-type haemophores were first identified in Serratia marcescens but are also conserved in other Gram-negative bacteria such as Yersinia pestis and Y. pseudotuberculosis (Table S1; Ghigo et al., 1997; Rossi et al., 2001). Haemophores are proteins with higher affinity for haem than haem-containing proteins such as haemopexin or myoglobin. The S. marcescens Haem Acquisition System (=Has) obtains haem through different steps. First, HasA haemophores are secreted through a Type I Secretion System (TISS). The SecB general chaperone is required to maintain HasA in a secretion-competent state and to facilitate its secretion. Once in the extracellular space, HasA acquires haem from the host haem carrier proteins due to its higher affinity, regardless of its redox state (Fe²⁺ or Fe³⁺). Then, haem-containing HasA is recognized by the HasR specific TonB-dependent receptor. In the presence of the HasB-ExbB-ExbD energy transduction system, interaction of haem-bound HasA with HasR leads to a conformational change allowing haem transfer to HasR and apo-HasA release in the extracellular space for recycling (Cescau et al., 2007). HasB is a TonB homolog dedicated to HasR (De Amorim et al., 2013). In S. marcescens, the Hem haem uptake system is active at low haem concentrations. Even lower iron concentrations are required for activation of the Has haemophore system, suggesting that these two systems might be involved in haem acquisition under different conditions (Benevides-Matos and Biville, 2010).

If studies on iron homeostasis in enterobacteria mainly focus on acquisition systems, the potential role of iron efflux in virulence should also be considered. Indeed, in S. Typhimurium, oxidative stress caused by an iron overload is encountered after macrophage invasion. Enterobactin production and iron-citrate efflux have been shown to confer oxidative stress resistance in those conditions (Frawley et al., 2013).

It is still unclear how copper enters the bacterial cytoplasm. Uncharacterized energy-independent channels such as OmpC porins probably allow passage of Cu²⁺ and Cu¹⁺ through the outer membrane (Rensing and Grass, 2003). In E. coli, permeability to copper is reduced by the ComC outer membrane protein through an unknown mechanism. ComC expression decreases under low copper conditions to allow copper uptake. Homologs of ComC are present in many Gram-negative bacteria, suggesting that it might play an important role in copper homeostasis (Rademacher and Masepohl, 2012). Only Cu¹⁺ ions enter the cytoplasm by an unknown mechanism, potentially involving Zn²⁺ uptake systems or some ATPases (Ma et al., 2009; Hood and Skaar, 2012; Nies and Herzberg, 2013). However, in the case of pathogenic Enterobacteria, copper efflux is more crucial than copper uptake. To prevent copper from competing with natural ligands of metalloproteins, unbound copper is excluded from the cytoplasm. In bacteria, all copper-dependent proteins are located in the periplasm or within the cytoplasmic membrane with binding sites in the periplasm (Ma et al., 2009; Dupont et al., 2011; Hodgkinson and Petris, 2012; Nies and Herzberg, 2013). The copper tolerance response in Gram-negative bacteria involves three different mechanisms: (1) exporting cytoplasmic copper to the periplasm using P_1B-type ATPases, (2) detoxifying copper using the CueO multicopper oxidase, or (3) eliminating unbound periplasmic copper by exporting it or sequestering it (Dupont et al., 2011; Hodgkinson and Petris, 2012).

In Gram-negative bacteria, inner membrane heavy metal pumps (P_1B-type ATPases) export cytoplasmic copper to the periplasm. These pumps are monomers and their structure includes a group of three cytoplasmic domains responsible for the ATPase activity. The phosphorylation cycle responsible for ATPase activity induces conformational changes resulting in substrate translocation through the inner membrane (Gourdon et al., 2011; Klein and Lewinson, 2011).

In E. coli, the CopA ATPase is involved in Cu¹⁺ export from the cytoplasm to the periplasm (Figure 2; Arguello et al., 2011). In Salmonella, copper ATPases are not only required for copper resistance, but also for efficient copper availability to cupro-enzymes such as Cu/Zn SodCII under copper-limited conditions (Osman et al., 2013). They might play a role in delivering cytoplasmic copper contained in unidentified storage compounds (Nies and Herzberg, 2013). Once in the periplasm, highly reactive Cu¹⁺ is detoxified through re-oxidation by the CueO multicopper oxidase to form less harmful Cu²⁺ (Dupont et al., 2011; Nies and Herzberg, 2013).

In some enterobacteria such as E. coli, the CusCBA heteromultimeric transport complex can transfer periplasmic copper to the extracellular milieu (Outten et al., 2001; Gudipaty et al., 2012). Using the proton-motive force, the CusCBA complex shuttles between three conformational states to act as a peristaltic pump and excrete copper (Dupont et al., 2011). The periplasmic transfer of Cu¹⁺ from the P_1B-ATPases to the CusB adaptor protein is achieved by the CusF chaperone to limit its potential toxic effects (Hodgkinson and Petris, 2012; Nies and Herzberg, 2013). This Cus system is also present in K. pneumoniae (Zulfiqar and Shakoori, 2012). Other chaperones have been described in some E. coli strains, such as plasmidic PcoC or PcoE (Espirito Santo et al., 2008).

Salmonella and Yersinia lack the CusCBA system to export Cu¹⁺ from the periplasm to the extracellular milieu. Alternatively, another periplasmic protein, CueP, sequesters copper to neutralize toxicity (Pontel and Soncini, 2009; Osman et al., 2010; Dupont et al., 2011). This chaperone also helps supply copper to SodCII, (Nies and Herzberg, 2013; Osman et al., 2013). S. Typhimurium uses two different P_1B-type ATPases, GolT, and CopA (Dupont et al., 2011; Nies and Herzberg, 2013). Initially described as a gold efflux system, GolT is predominantly involved in copper efflux (Osman et al., 2013). S. Typhimurium expresses a CueO multi-copper oxidase, also called CuiD, which buffers copper toxicity in the periplasm (Arguello et al., 2011; Hodgkinson and Petris, 2012). Periplasmic Cu²⁺ is efficiently reduced to Cu¹⁺ by NADH dehydrogenase 2 and components of the respiratory chain. Re-oxidation by multicopper oxidases ensures Cu²⁺ availability for copper-dependent enzymes (Nies and Herzberg, 2013).

Iron and copper homeostasis are linked, as some siderophores interact with copper. In vitro, catecholate siderophores increase copper sensitivity in E.coli, reducing Cu²⁺ to generate toxic Cu¹⁺. The CueO multicopper oxidase is essential, as it allows oxidation of catecholates to create molecules able to sequester Cu²⁺ instead of reacting with it (Grass et al., 2004). However, in S. Typhimurium, a cueO mutant does not display higher sensitivity to siderophores (Achard et al., 2010). Conversely, yersiniabactin production increases copper resistance, as it sequesters Cu²⁺ before it interacts with catecholate siderophores to produce highly toxic Cu¹⁺ (Chaturvedi et al., 2012).

The transport of Zn²⁺ across the outer membrane is not defined in Enterobacteria. Zinc uptake across the cytoplasmic membrane is mediated by two major types of transporters: ZnuACB, which belongs to the cluster C9 family of (TroA-like) ABC transporters, and ZupT, which is a member of the ZIP (ZRT/IRT-like protein) family of transporters that are also present in eukaryotes (Hantke, 2005). ZnuACB is a high-affinity transporter whereas ZupT is a low-affinity uptake system (Hantke, 2005). Under conditions of moderate zinc availability, zinc uptake is carried out by ZupT, whereas it is carried out by ZnuACB in environments characterized by very low zinc availability.

The gene znuA encodes for the periplasmic, zinc-binding component of the transporter, znuB encodes for the transmembrane component and znuC encodes for the ATPase subunit. Zinc uptake mediated by the Znu system requires ATP hydrolysis by dimeric ZnuC to transport Zn²⁺ captured by ZnuA through the pore formed by a ZnuB dimer in the cytoplasmic membrane (Patzer and Hantke, 1998, 2000). In Enterobacteria, numerous studies in different strains of E. coli, in S. Typhimurium, S. Enteritidis, Proteus mirabilis or in different species of Yersinia have demonstrated that ΔznuA, ΔznuC, ΔznuB or ΔznuACB mutants had decreased Zn²⁺ uptake compared to the wild-type strains (Table S1; Patzer and Hantke, 1998; Campoy et al., 2002; Ammendola et al., 2007; Gunasekera et al., 2009; Sabri et al., 2009; Desrosiers et al., 2010; Nielubowicz et al., 2010; Gabbianelli et al., 2011).

In some bacterial species such as IPEC strain O157:H7 and S. Typhimurium, zinc uptake involves another protein, ZinT (formerly known as YodA). This protein is involved in periplasmic zinc binding under zinc-limiting conditions, and studies carried out in S. Typhimurium have suggested that ZinT participates in zinc uptake through ZnuACB, by a mechanism involving direct interaction with ZnuA (Petrarca et al., 2010; Gabbianelli et al., 2011).

Proteins of the ZIP family were initially identified as iron or zinc transporters in eukaryotes, but some members were subsequently shown to also transport other metals, such as manganese or cadmium. ZupT is the first characterized bacterial member of this family and was shown to be responsible for zinc uptake in E. coli (Grass et al., 2002). Studies on the role of ZupT on Zn²⁺ uptake have been performed in non-pathogenic E. coli and in the UPEC strain CFT073 (Table S1; Grass et al., 2002; Sabri et al., 2009). The zupT gene is constitutively expressed and ZupT can transport iron and cobalt in addition to zinc and possibly manganese. When overexpressed, ZupT can also transport copper (Grass et al., 2002, 2005).

Zinc detoxification is primarily achieved by the P_1B-type ATPase ZntA and the cation diffusion facilitators (CDF) ZitB and YiiP (Table S1; Rensing et al., 1997; Grass et al., 2001; Wei and Fu, 2006). P_1B-type ATPases and CDF transporters catalyze metal translocation across the inner membrane, and the substrate is transported from the cytoplasm to the periplasm (Klein and Lewinson, 2011). CDF is a ubiquitous family of metal transporters found in prokaryotes and eukaryotes. Functional analysis of YiiP and ZipT indicated that these two proteins are proton-linked antiporters that utilize the free energy derived from H⁺ influx to pump cytosolic Zn²⁺ out of the cells (Chao and Fu, 2004; Wei and Fu, 2006).

In addition to Zn²⁺, ZntA is able to transport Cd²⁺ and Pb²⁺ (Binet and Poole, 2000). It has been proposed in S. Typhimurium that the periplasmic C-terminal domain of ZraP, a periplasmic protein with two zinc-binding domains, facilitates modulation of transporters such as ZntA (Appia-Ayme et al., 2012). It has been suggested that ZntA is critical for the survival of E. coli in the presence of high zinc concentrations, while ZitB maintains zinc homeostasis under normal growth conditions, i.e., low environmental zinc stress (Rensing et al., 1997; Wang et al., 2012). Overexpression of zitB in E. coli resulted in a significant increase in zinc tolerance and reduced uptake of zinc, while overexpression of yiiP did not confer additional zinc resistance, and deletion of yiiP did not alter zinc resistance (Grass et al., 2001).

A recent study identified new zinc exporters in E. coli, MdtABC, a RND-type efflux pump, and MdtD, a MFS (Major Facilitator Superfamilly) transporter, as well as a periplasmic protein, Spy, involved in zinc detoxification (Table S1). It has been proposed that the role of Spy in relieving zinc stress may be to facilitate the folding and to protect the integrity of transmembrane and periplasmic transporters that function in zinc export (Wang and Fierke, 2013).

As for Zn²⁺, the mechanisms of Mn²⁺ transport across the outer membrane are not yet defined in Enterobacteria. For import across the cytoplasmic membrane, two major manganese transporters have been identified: a proton-dependent Nramp-related transport system typified by MntH and an ABC transporter typified by SitABCD and YfeABCD (Figure 2; Goswami et al., 2001; Forbes and Gros, 2003). The Nramp (Natural resistance-associated macrophage protein) transporter family was first described in plants, animals, and yeasts (Cellier et al., 1996). Mammalian Nramp1 and Nramp2 are H⁺-dependent transition metal divalent cation transporters with physiologically relevant affinities for at least Mn²⁺, Fe²⁺, and Zn²⁺ (Papp-Wallace and Maguire, 2006).

MntH (for Proton-dependent manganese transporter) has been characterized in many enterobacterial species (Table S1; Makui et al., 2000; Boyer et al., 2002; Zaharik et al., 2004; Runyen-Janecky et al., 2006; Champion et al., 2011; Perry et al., 2012). Affinity studies for Mn²⁺ have shown that the apparent K_0.5 for Mn²⁺ uptake was 0.1 μM in S. Typhimurium and 0.5 to 1 μM in E. coli. The K_0.5 for Fe²⁺ was about 100 μM, far higher than physiological concentrations of Fe²⁺ (Kehres et al., 2000).

High-affinity Mn²⁺ acquisition can also be mediated by ABC transporters. The SitABCD transporter, present in some pathogenic E. coli strains, was first described in S. Typhimurium (Table S1; Zhou et al., 1999). In S. Typhimurium, SitABCD is primarily a Mn²⁺ transporter rather than a Fe²⁺ transporter. It mediates influx of Mn²⁺ with a K_0.5 of 0.1 μM, while it mediates influx of Zn²⁺, Cd²⁺, and Fe²⁺ with affinities of 3, 3 and 30 μM, respectively. Moreover, it operates optimally in slightly alkaline medium, whereas MntH seems to be more effective in acid medium, which suggests distinct physiological roles (Kehres et al., 2002b). In the APEC strain χ7122, the SitABCD transporter has affinities for Mn²⁺ and Fe²⁺ of about 4 and 1 μM, respectively. However, the affinities changed in accordance with the genetic background of the strain. In a mntH mutant, manganese was better transported than iron; in an aroB feoB mutant iron was better transported than manganese. SitABCD is thus a highly versatile and adaptable transporter (Sabri et al., 2006). The SitABCD transporter in S. flexneri or YfeABCD system in Y. pestis is also able to transport Mn²⁺ and Fe²⁺ (Bearden and Perry, 1999; Runyen-Janecky et al., 2006).

As manganese is essential for enzymatic catalysis and protection against oxidative stress, molecular mechanisms of manganese toxicity are not yet clear. A manganese efflux pump, MntP, has been recently described in E. coli K-12 (Waters et al., 2011), and this transporter seems to be present in other pathogenic Enterobacteria (Veyrier et al., 2011).

The ComC outer membrane protein is under the control of the TetR-like ComR repressor. In the presence of high copper concentrations, Cu²⁺ binds to ComR and releases it from the comC promoter region, leading to its activation (Mermod et al., 2012).

The copper ATPase CopA, the multicopper oxidase CueO and the periplasmic copper-binding protein CueP are regulated by CueR, belonging to the MerR family of regulators (Outten et al., 2000; Stoyanov et al., 2001; Osman et al., 2013). The MerR family contains nearly exclusively transcriptional activators of the expression of genes required for metal efflux or detoxification, or defense against oxidative stress and drug resistance. Both the apo- and effector-bound forms are capable of binding to their operator DNA sequences with similar affinities. However, binding of the metal ion provokes an allosteric change at the DNA-binding domain of the protein, so that only the effector-bound form can significantly optimized RNA polymerase binding and transcription initiation (O'halloran et al., 1989; Brown et al., 2003). CueR has a copper affinity of 10⁻²¹ M (Changela et al., 2003), and is able to induce the expression of its target genes in response to Cu⁺, Ag⁺, and Au⁺ (Ma et al., 2009; Perez Audero et al., 2010). GolS is a CueR-like sensor and the regulator of the P-type ATPase GolT (Osman et al., 2013). GolS and CueR have similar affinities for Cu⁺, but in vivo GolS distinguishes Au⁺ from Cu⁺ or Ag⁺ to activate its target genes (Ibanez et al., 2013; Osman et al., 2013).

The copper efflux genes cusCFBA are induced by the CusSR TCRS which responds to extracellular copper concentrations (Gudipaty et al., 2012). The importance of the Cus system is also dependent on the presence of oxygen. In E. coli, the Cus system is not required for copper resistance under aerobic conditions, in which the primary line of defense, the CueR regulon, is sufficient (Outten et al., 2001). However, in K. pneumonia, the Cus system is activated under both aerobic and anaerobic conditions and the activation level is higher under aerobic conditions (Zulfiqar and Shakoori, 2012). These differences might be explained by the lower capacity of E. coli to store copper under anaerobic conditions (Zulfiqar and Shakoori, 2012).

The zinc uptake system ZnuACB is regulated by Zur, belonging to the Fur family, and by SoxR, belonging to the MerR family of regulators. This system is repressed by Zn-Zur in zinc-rich environment and, under zinc depletion conditions, Zur becomes inactive, leading to the activation of znuACB (Li et al., 2009). Zur is able to sense zinc at concentrations as low as 10⁻¹⁵ M in vitro (Waldron and Robinson, 2009). SoxR responds to oxidative stress and is part of the SoxRS regulon. SoxS, activated by SoxR, activates the expression of this regulon containing numerous genes such as znuACB (Brown et al., 2003; Warner and Levy, 2012). A network biology approach has predicted that RybA sRNA might regulate ZupT uptake system in E. coli (Modi et al., 2011).

The ZntA export system is regulated by the MerR-like ZntR regulator. The apo-ZntR dimer binds to the promoter of zntA and weakly represses transcription; Zn-ZntR is a transcriptional activator (Wang et al., 2012). ZntR can also bind Pb²⁺ and Cd²⁺ to activate zntA expression (Binet and Poole, 2000). As Zur, ZntR has zinc affinity of 10⁻¹⁵ M in vitro but, in vivo, ZntR up-regulates zntA transcription in response to nanomolar intracellular concentrations of free zinc (Outten and O'halloran, 2001; Wang et al., 2012). Finally, ZntA is up-regulated under high zinc stress by ZntR to effectively lower the intracellular zinc concentration, while the other export system, ZitB is constitutively expressed to function as a first-line defense against zinc influx (Wang et al., 2012).

In Enterobacteria, responses to zinc are mediated by BasSR, BaeSR, PmrAB, and ZraSR TCRS (Yamamoto and Ishihama, 2005; Appia-Ayme et al., 2012; Chen and Groisman, 2013; Wang and Fierke, 2013). MdtABC and MdtD zinc exporters are up-regulated by BaeSR upon exposure to high concentrations of zinc (Wang and Fierke, 2013). Copper and zinc exposure lead to overexpression of Spy through regulation by CpxAR and BaeSR, respectively (Wang and Fierke, 2013), and ZraP is upregulated by ZraSR (Appia-Ayme et al., 2012).

In Enterobacteria, except for Yersinia, the DtxR-like MntR regulator is the primary sensor of manganese abundance. When bound to manganese, it represses the transcription of the uptake systems MntH and SitABCD (Patzer and Hantke, 2001; Ikeda et al., 2005; Papp-Wallace and Maguire, 2006; Runyen-Janecky et al., 2006). Of the other metal cations, only Cd²⁺ can compete efficiently for binding to MntR (Lieser et al., 2003). MntR has a 10⁻⁵ M affinity for manganese, value that matches the estimated 10⁻⁵ M intracellular concentration of this metal (Waldron and Robinson, 2009). Mn²⁺ is the most potent and most effective cation in E. coli and S. Typhimurium. In S. Typhimurium, Fe²⁺ repressed transcription through interaction with MntR in the absence of Fur (but this requires high extracellular concentrations of Fe²⁺) (Patzer and Hantke, 2001; Kehres et al., 2002a; Ikeda et al., 2005). In Y. pestis, in which MntR is absent, Fur represses YfeABCD and MntH in response to iron and manganese (Perry et al., 2012). Control of Fur repression by manganese is not specific to Y. pestis, as E. coli aerobactin and fhuF genes also respond to manganese in a Fur-dependent manner. However, in both cases, Fur repression through manganese binding is just specific to some promoters (Privalle and Fridovich, 1993).

MntH is activated by the LysR-type transcriptional regulator OxyR in the presence of H₂O₂in S. Typhimurium and Shigella (Kehres et al., 2002a; Runyen-Janecky et al., 2006). In the presence of oxidative stressors, OxyR activates genes by directly interacting with RNA polymerase to enhance initiation of transcription (Ma et al., 2009). MntH is also part of the peroxide stress response in E. coli (Anjem et al., 2009). It has been predicted that RybA and RyhB sRNAs might coordinately regulate mntH (Modi et al., 2011).

In addition to repressing gene expression, Mn-MntR activates the manganese efflux pump MntP (Waters et al., 2011).

In Shigella, single inactivation of either the Shu haem transport system, the Sit and Feo ferrous iron transporters or enterobactin biosynthesis does not affect invasion or intracellular growth (Reeves et al., 2000; Runyen-Janecky et al., 2003). The IutA aerobactin receptor is not expressed in Shigella grown in HeLa cells. However, combined deletion of siderophore (IucD) and ferrous iron acquisition systems (Feo and Sit) leads to decreased growth and spreading in epithelial cells, suggesting a synergistic role of these different systems (Runyen-Janecky et al., 2003). Similarly, in S. Typhimurium, a single FeoB mutation does not affect intramacrophage replication, which is impaired in simultaneous absence of SitABCD, MntH, and FeoB, suggesting that ferrous iron and manganese acquisition might play a critical role in virulence (Boyer et al., 2002).

The CopA ATPase, involved in export of cytoplasmic Cu¹⁺ to the cytoplasm, is necessary for intramacrophage survival of E. coli (White et al., 2009). In S. Typhimurium, it is overexpressed upon phagocytosis by macrophages cells, which supports the theory that copper is accumulated in phagosomes. Deletion of copA or golT in S. Typhimurium has no effect on survival in cultured macrophages, whereas deletion of both results in significantly reduced survival (Achard et al., 2012; Hodgkinson and Petris, 2012). On the contrary, a cueO mutant shows no defect in survival in macrophages (Achard et al., 2010).

Concerning zinc, znuA was strongly induced in an E. coli O157:H7 strain adhering to Caco-2 cultured epithelial cells and a znuA mutant was significantly less able to adhere to Caco-2 cells in competition with the wild-type strain (Gabbianelli et al., 2011). In S. Typhimurium and S. Enteritidis, znuA mutants were impaired for growth in Caco-2 epithelial cells and bacteria starved for zinc displayed reduced multiplication in phagocytes (Ammendola et al., 2007).

In Y. pseudotuberculosis, a mntH mutant was defective in survival and growth in macrophages expressing functional Nramp1, but survived and replicated in macrophages deficient in Nramp. This mutant was also susceptible to killing by H₂O₂ when grown under manganese-limited conditions (Champion et al., 2011). In S. flexneri, a mntH sitA mutant was more sensitive to hydrogen peroxide, but not to superoxide generators. Moreover, the mutant had impaired survival in activated macrophage lines, but was able to form the same number and size plaques on Henle cell monolayers, suggesting Sit and MntH are not required for survival in this epithelial cell line. Expression of sitA and mntH was higher when Shigella was in Henle cells compared to LB medium (Runyen-Janecky et al., 2006). Upregulation of sitABCD in Shigella was also observed by microarray analysis in HeLa cells and human macrophage-like U937 cells (Lucchini et al., 2005). A most recent study in S. flexneri demonstrated that when cultured Henle cells were infected with a mixture of wild-type and sitA mutant strains, the sitA mutant was recovered in lower numbers than the wild-type strain, indicating that Sit provides an intracellular growth advantage (Fisher et al., 2009). In S. Typhimurium, sitABCD was induced in vivo after invasion of the intestinal mucosa (Janakiraman and Slauch, 2000). The mntH sit mutant of S. Typhimurium Keller strain was defective for replication in Nramp1^−/− RAW 264.7 macrophages. Overexpression of mntH in a mntH sit mutant improved the intracellular survival of the strain in macrophages (Boyer et al., 2002). In S. Typhimurium strain SL1344, a mntH mutant showed no defect in invasion of or survival in cultured HeLa or RAW 264.7 macrophages but was more susceptible to killing by H₂O₂. However, expression of mntH was induced several fold after 3h within macrophages (Kehres et al., 2000). Using the same strain, Zaharik et al. demonstrated that sitA and mntH were upregulated when strains were internalized by Nramp1-expressing macrophages (Zaharik et al., 2004). As a rule, mutants for manganese transporters impaired in their ability to replicate in macrophages were also impaired in their ability to resist oxidative stress.

The FeoB ferrous iron transporter is not required for systemic infections caused by APEC or S. enterica infections (Tsolis et al., 1996; Sabri et al., 2008). A decrease in infection of intravenously inoculated mice was observed for S. Typhimurium, but the mice used in this study were Nramp^−/−, thus, more susceptible to iron-induced oxidative stress (Boyer et al., 2002). Y. pestis is significantly less virulent after deletion of the yfe, feo, and mntH genes in a bubonic plague model, suggesting that iron and manganese transport are important in a subcutaneous model of infection (Fetherston et al., 2012). Pathogen-specific siderophore production is crucial for the establishment of systemic infections. Enterobactin mutants of S. Typhimurium are not attenuated in a systemic model of infection (Benjamin et al., 1985). Enterobactin can efficiently sequester iron from transferrin due to its particularly high affinity for ferric iron, as demonstrated in K. pneumoniae, in which enterobactin promotes survival in the perivascular space. However, lipocalin-2, also called NGAL or siderocalin, is an innate immune defense protein that can sequester enterobactin (Bachman et al., 2012). Chicken and quail also produce avian homologs of lipocalin-2 (Garenaux et al., 2013). Thus, production of additional siderophores that are not captured by siderocalin, are required for systemic infection by Enterobacterial pathogens (Fischbach et al., 2006). Salmochelins are involved in systemic infection caused by S. Typhimurium (Crouch et al., 2008). Salmochelins and aerobactin are both important for chicken systemic infection caused by APEC (Dozois et al., 2003). If yersiniabactin plays a role in bubonic plague caused by Y. pestis, it is not required for septicemic plague following intraveinous injection (Fetherston et al., 2010; Sebbane et al., 2010). In addition, neither the Has haemophore nor Hmu were involved in virulence in both bubonic plague or systemic infection models (Thompson et al., 1999; Rossi et al., 2001).

Inside the host, infection leads to increased copper levels in the serum. This could be due to ceruloplasmin secretion by the liver during the acute-phase response. Ceruloplasmin is a serum copper-containing protein that is associated with 85% of the copper circulating inside the host (Hodgkinson and Petris, 2012). In S. Typhimurium, CopA and GolT ATPases are not involved in systemic infection (Achard et al., 2012; Hodgkinson and Petris, 2012). By contrast, deletion of cueO, resulted in decreased virulence during systemic infection (Achard et al., 2010). This is consistent with the observation that host copper deficiency increases susceptibility to infection by S. Typhimurium (Hodgkinson and Petris, 2012). Surprisingly, a UPEC cueO mutant displays a hypervirulent phenotype. Indeed, in the absence of CueO, copper-stressed cells displayed a mucoid phenotype and an aggregative behavior. This could lead to an increase in capsule production and virulence by evading the host immune response (Tree et al., 2007).

In Y. ruckeri, a znuACB mutant was unable to compete with the wild-type strain and survived poorly in rainbow trout kidney (Dahiya and Stevenson, 2010). By contrast, ZnuACB was not important for high-level infectivity and virulence of Y. pestis in either subcutaneous or intranasal infection models (Desrosiers et al., 2010). In S. Typhimurium, virulence of a znuC mutant was attenuated compared to the wild type strain (Campoy et al., 2002). Studies conducted with S. Typhimurium and S. Enteritidis showed similar results with znuA mutants (Ammendola et al., 2007).

An yfeABCD mutant strain of Y. pestis was attenuated in Nramp1^+/+ mice following intravenous infection (Bearden and Perry, 1999). Similarly, yfeAB, feoB yfeAB, yfe mntH mutant strains were attenuated in a bubonic plague model (Fetherston et al., 2012; Perry et al., 2012). Galleria mellonella larval survival following inoculation with a Y. pseudotuberculosis mntH strain was significantly greater than survival following challenge with the wild-type strain (Champion et al., 2011). During intra-peritoneal competition experiments with S. Typhimurium, the sitA mutant was consistently out-competed by the wild-type strain in the spleens and livers of mice (Janakiraman and Slauch, 2000). Moreover, intravenous inoculation of sit, mntH and feoB mutant strains in Nramp1^−/− mice showed that the mntH mutant was fully virulent and the sitABCD mutant was markedly attenuated. The mntH feoB, sit feoB and mntH sit feoB mutants were completely avirulent (Boyer et al., 2002). Another study showed in Nramp1^+/+ mice that the mntH and sitABCD mutants were significantly attenuated and the mntH sit mutant was completely avirulent (Zaharik et al., 2004). Depending on the mouse model used, MntH and SitABCD systems in S. Typhimurium seem to be more or less important alone but play an important combined role during infection. In S. flexneri, a sitA mutant was attenuated in a mouse lung model of virulence (Fisher et al., 2009). An APEC sitA mutant demonstrated reduced colonization of the lungs, liver and spleen compared to the wild-type strain. The mntH sit mutant demonstrated reduced persistence in blood and reduced colonization in the lungs, liver, and spleen. The mntH mutant was as virulent as the wild-type strain (Sabri et al., 2008). The mntH sit mutant strain was more sensitive to H₂O₂ compared to the wild-type strain (Sabri et al., 2006).

Role of iron transporters. The FeoB ferrous iron transporter promoted colonization of the intestine by S. enterica in mice (Tsolis et al., 1996). Iron acquisition by siderophores also plays a particularly important role in gut colonization. Enterobactin (and other catecholate siderophores) is involved in gut colonization of the mouse by Gram-negative bacteria (Pi et al., 2012). Enterobactin is produced in large quantities by commensals or pathogens producing no other pathogen-specific siderophores, such as E. coli O157:H7. The Iha siderophore receptor is involved in colonization of the intestine by this enterohaemorragic E. coli. However, in this case, its virulence potential is related to its adhesin properties (Yin et al., 2009). Other pathogenic strains preferentially produce pathogen-specific siderophores rather than enterobactin to promote colonization (Henderson et al., 2009). As such, aerobactin is involved in mouse intestinal colonization by E. coli O104:H4 (Torres et al., 2012). SitA also contributes to S. Typhimurium colonization of the small intestine (Janakiraman and Slauch, 2000).

Role of zinc transporters. In S. Typhimurium, ZnuACB contributes to resistance against host calprotectin-mediated Zn²⁺ chelation. This transporter promotes resistance of extracellular S. Typhimurium to calprotectin accumulated in the host intestine following infection. Moreover, S. Typhimurium exploits calprotectin-mediated Zn²⁺ chelation in order to out-compete host microbiota, which is less well adapted to the zinc-limited environment in the infected intestine (Liu et al., 2012).

Role of iron transporters. The Iha siderophore receptor is involved in bladder and kidney colonization by UPEC (Leveille et al., 2006). FyuA, the yersiniabactin receptor, promotes biofilm formation in the bladder (Hancock et al., 2008; Brumbaugh et al., 2013). In P. mirabilis, yersiniabactin, unlike proteobactin, also allows better fitness in the bladder and the kidneys in a coinfection model (Himpsl et al., 2010). However, as demonstrated in UPEC, yersiniabactin might be involved in copper sequestration rather than iron acquisition (Chaturvedi et al., 2012). The IutA and IroN receptors also promote bladder colonization (Garcia et al., 2011; Watts et al., 2012). On the contrary, enterobactin production is not involved in kidney colonization (Torres et al., 2001).

Haem is an important iron source in tissues. Competition assays showed that the ChuA haem receptor contributes to iron acquisition in kidneys following urinary tract infection by E. coli (Hagan and Mobley, 2009; Garcia et al., 2011). The Hma haem receptor is also involved in kidney colonization by UPEC (Hagan and Mobley, 2009; Garcia et al., 2011). By contrast, a chuT deletion mutant in mono-infections in both APEC and UPEC models had no change in virulence (Gao et al., 2012).

Role of zinc transporters. In competitive infections, a UPEC zupT mutant was not outcompeted by the wild-type strain. In contrast, the znuA and znuA zupT mutants demonstrated significantly reduced numbers in the bladders and kidneys. In single-strain infections, znuA and znuA zupT mutants were reduced in the kidneys. Moreover, the double mutant demonstrated decreased motility and less resistance to hydrogen peroxyde (Sabri et al., 2009). A UPEC znuB mutant exhibited a defect in biofilm formation under static conditions and in motility (Gunasekera et al., 2009). In Proteus mirabilis, a znuC mutant displayed reduced swimming and swarming motility. This mutant was outcompeted by the wild-type strain during competitive infections in urine, bladder and kidneys of mice but colonized mice as well as the wild-type during independent infections (Nielubowicz et al., 2010).

Role of manganese transporters. In UPEC, mntH and sitABCD mutants colonize bladder and kidneys as well as the wild type strain. However, the mntH sit double mutant displayed lower colonization rates in kidneys. Moreover, the mntH and mntH sit mutants were more sensitive to H₂O₂ and plumbagin compared to the wild-type strain (unpublished data from our laboratory).

Role of iron transporters. Salmochelins and aerobactin play important roles in colonization of the lungs during APEC infections (Dozois et al., 2003), whereas haem uptake through the Chu system does not play a significant role (Gao et al., 2012). Salmochelins and yersiniabactin are both required for K. pneumoniae infections (Bachman et al., 2011, 2012). In Y. pestis, yersiniabactin is important for pneumonic plague but may not be critical for iron acquisition. Indeed, receptor mutants are less attenuated than synthesis mutants and it has been suggested that yersiniabactin may damage pulmonary epithelial cells or affect immune cells (Fetherston et al., 2010; Perry and Fetherston, 2011).

Role of copper transporters. Copper accumulation at the sites of infection, such as infected lungs by M. tuberculosis has been reported (Hodgkinson and Petris, 2012). As a consequence, copper tolerance might play an important role in the establishment of some pulmonary infections. However for Enterobacteria, thus far no data to support this have been reported.

Role of manganese transporters. Yfe and Feo systems are not essential for pneumonic plague, even though enhanced transcription of yfe genes was measured in vivo in a pneumonic plague model. However, the environment encountered by bacteria in the lungs is limited in manganese, and manganese transporters have been shown to play significant roles in other types of lung infections. Manganese requirements of Y. pestis during lung infection might be lower than that of other pathogens (Fetherston et al., 2012).

The Yfu and Yiu ABC transporters are not involved in virulence in a bubonic plague model (Gong et al., 2001; Kirillina et al., 2006). Deletion of the yersiniabactin siderophore-mediated iron acquisition system resulted in complete loss of Y. pestis virulence in a bubonic plague model following subcutaneous inoculation (Fetherston et al., 2010). Using a flea-to mouse infection model, it has also been established that yersiniabactin played a critical role during the early stages of the infection (Sebbane et al., 2010).