Canonical and ECF-type ATP-binding cassette importers in prokaryotes: diversity in modular organization and cellular functions

Authors


  • Present address: Mathias Grote, ESRC Centre for Genomics in Society, University of Exeter, Byrne House, St. German's Road, Exeter EX4 4PJ, UK.

  • Editor: Mecky Pohlschröder

Correspondence: Erwin Schneider, Institut für Biologie/Physiologie der Mikroorganismen, Humboldt-Universität zu Berlin, Chausseestraße 117, 10115 Berlin, Germany. Tel.: +49 30 2093 8121; fax: +49 30 2093 8126; e-mail: erwin.schneider@rz.hu-berlin.de

Abstract

Since their discovery in the 1960s as ‘osmotic shock-sensitive’ transporters, a plethora of so-called binding protein-dependent (canonical) ATP-binding cassette (ABC) importers has been identified in bacteria and archaea. Their cellular functions go far beyond the uptake of nutrients. Canonical ABC importers play important roles in the maintenance of cell integrity, responses to environmental stresses, cell-to-cell communication and cell differentiation and in pathogenicity. A new class of abundant micronutrient importers, the ‘energy-coupling factor’ (ECF) transporters, was originally identified by functional genomics. ABC ATPases are an integral part of both canonical ABC and ECF importers. Fundamental differences include the modular architecture and the independence of ECF systems of extracytoplasmic solute-binding proteins. This review describes the roles of both types of transporters in diverse physiological processes including pathogenesis, points to the differences in modular assembly and depicts their common traits.

Introduction and historical background

Transport across biological membranes is fundamental to any form of life. Referring to the source of energy used by transport proteins, one can distinguish channels, primary and secondary transporters (Saier, 2000). Whereas secondary transporters couple transport to electrochemical gradients across the membrane, for example a proton or a sodium potential, primary transporters such as proton-pumping ATPases or ATP-binding cassette (ABC)-transporters harness the free energy of ATP hydrolysis.

What is known today as binding protein-dependent ABC transporters was historically distinguished from other classes of transport systems by their susceptibility to cold osmotic shock (Neu & Heppel, 1965). The sensitivity of these transport systems against osmotic shock was soon found to be due to the release of an essential protein component located in the periplasm of Escherichia coli and Salmonella enterica subsp. enterica serovar Typhimurium (S. Typhimurium), between the inner and the outer membrane – the periplasmic solute-binding protein (SBP) (Pardee, 1968; Berger, 1973; Berger & Heppel, 1974). Because E. coli and S. Typhimurium were the preferred bacterial model organisms at that time, it was assumed that this type of transporter must be confined to gram-negative bacteria. This (first) dogma was overthrown in the late 1980s when similar transport systems were discovered in gram-positive bacteria encompassing a solute-binding lipoprotein that is anchored to the extracellular side of the cytoplasmic membrane by fatty acids covalently linked to the N-terminal cysteine residue of these proteins (reviewed in Sutcliffe & Russell, 1995).

With the first nucleotide sequences of the histidine and maltose transporters of S. Typhimurium and E. coli, respectively, at hand in the early 1980s, it became obvious that despite their chemically different substrates, both systems contained a homologous component (Gilson et al., 1982). Moreover, in 1986, a mammalian protein (P-glycoprotein or multidrug resistance protein) involved in resistance of certain cancer cells to chemotherapeutic agents was recognized to share significant similarity to the bacterial transport proteins (reviewed in Ames, 1986). These findings gave rise to the notion that a common transport principle might be present in all organisms from bacteria to humans. Sequence motifs were identified that are required for nucleotide binding and this led to the designation of ‘ABC’ transport proteins by Higgins and colleagues (Hyde et al., 1990). Using this information as a template, subsequent genome sequence projects led to the discovery of a plethora of ABC transporters in organisms belonging to all three kingdoms of life that made them one of the largest protein families, including both ABC importers (the originally described ‘osmotic shock-sensitive’ systems) and export systems. The occurrence of a solute-binding component was taken as an earmark of ABC-type importers.

This (second) dogma was subsequently proven wrong by two pioneering studies. First, in 1993, a new class of secondary transporters – the TRAP family (tripartite ATP-independent periplasmic transporter) – was discovered, which includes a solute receptor for the capture and delivery of their substrates (Forward et al., 1993). Second, in 2006, bioinformatic analysis of whole-genome sequences from bacteria and archaea led to the discovery of the new and abundant subclass of energy-coupling factor (ECF)-type ABC import systems that lack an extracytoplasmic SBP (Rodionov et al., 2006, 2009; Hebbeln et al., 2007). The designation ‘ECF transporters’ was chosen to honor the work of Henderson et al. (1979), who published already in 1979 that the folate, thiamine and biotin transport systems of Lactobacillus casei each function via a specific ‘binding protein’ (later designated as ‘S components’, i.e. transmembrane proteins that are unrelated to SBPs of canonical ABC importers), and that they require, in addition, a common, but unknown component present in limiting amounts per cell. The latter was hypothesized to couple energy to these transport processes and named the ‘ECF’. More than 25 years later, ‘ECFs’ were identified as protein modules consisting of two ABC ATPases and a moderately conserved transmembrane protein. These ECF modules form stable complexes even in the absence of an S unit (Rodionov et al., 2006, 2009; Hebbeln et al., 2007).

Numerous excellent review articles and two books focusing in great detail on the crystal structures and the mechanistical aspects of ABC transporters have been published in recent years (e.g. Schmitt & Tampé, 2002; van der Heide & Poolman, 2002; Holland et al., 2003; Albers et al., 2004; Biemans-Oldehinkel et al., 2006a; Dawson et al., 2007; Linton, 2007; Davidson et al., 2008; Moussatova et al., 2008; Locher, 2009; Ponte-Sucre, 2009; Rees et al., 2009). In contrast, the genetics, molecular biology, biochemistry and physiology of ABC transporters mediating the uptake of diverse solutes in prokaryotes were last summarized in 2001 (Dassa & Schneider, 2001). Thus, it is the aim of this article to provide the reader with an up-to-date and comparative overview on canonical (binding protein-dependent) and ECF-type ABC importers in the context of their physiological significance. While in the case of canonical systems, we have focused on summarizing available experimental data, bioinformatic and experimental results are reviewed for ECF-type transporters since bioinformatics provided major clues to these systems.

ABC transporters

General description and classification

ABC systems are found in all three domains of life and form one of the largest protein superfamilies of paralogous sequences (Davidson et al., 2008). The human genome, for example, includes 48 genes for ABC transporters, while in the genome of E. coli serotype K-12, between 52 and 80 genes were reported to encode discrete ABC transport systems (reviewed in Moussatova et al., 2008). Plant genomes code for a high number of ABC systems, for example >120 have been identified in Arabidopsis thaliana and Oryza sativa (rice) (Rea, 2007). Yet, in relation to genome size, the highest number of ABC systems is found in bacteria (Davidson et al., 2008).

Canonical ABC transporters share a common structural organization comprising two transmembrane domains (TMDs) that form the translocation pore and two nucleotide-binding domains (NBDs) that hydrolyze ATP. The ABC superfamily includes exporters and importers, the latter being, until recently, confined to prokaryotes (but see ‘ABC and ECF importers in plants?’), and nontransport ABC systems that do not possess TMDs. A new mode of classification for both eukaryotic and prokaryotic ABCs has been proposed by Davidson et al. (2008). On the basis of sequence comparison, the authors postulate three classes of ABC systems that were already present in the last common ancestor of archaea, bacteria and eukarya. Class I comprises transporters with fused TMDs and NBDs, class II includes nontransport ABCs lacking TMDs and class III includes mainly transporters with NBDs and TMDs formed by separate polypeptide chains. Interestingly, class III is absent in eukaryotes.

Substrates transported by ABC transporters are diverse, such as sugars, amino acids, peptides, vitamins, ions, xenobiotics and even polypeptides, linking these proteins to various cellular functions that range from energy supply to osmoregulation, detoxification and virulence.

Prokaryotic canonical ABC importers that are dependent on high-affinity substrate-binding proteins, mediate the uptake of nutrients, osmoprotectants, various growth factors or trace elements (Holland et al., 2003). Canonical ABC importers, such as the enterobacterial maltose transporter MalFGK2-E (Ehrmann et al., 1998), the histidine transporter HisQMP2-J (Ames et al., 2001) or the glycine betaine (OpuA) transporter of Lactococcus lactis (Patzlaff et al., 2003), are model systems for the study of ABC transporters' structure and function. Crystal structures of two archaeal and four bacterial ABC importers have been resolved.

ECF transporters comprise an abundant class of importers for micronutrients in bacteria and archaea. ECF transporters consist of pairs of ABC ATPase domains (A components), a conserved transmembrane protein (T component) and a transmembrane substrate-capture protein (S component) in an unknown stoichiometry, but lack an SBP (Rodionov et al., 2009).

The transporter classification (TC) system currently includes a total of 33 subfamilies of ABC import systems within family 3.A.1.1 (http://www.tcdb.org). It is analogous to the Enzyme Commission system for classification of enzymes, except that it incorporates both functional and phylogenetic information. In an alternative classification system, Dassa & Bouige (2001) subdivided the ABC transporters into three uptake and six export classes. For a most recent description, see Dassa (2007). Here, the TC system will be used for the classification of canonical ABC importers (22 subfamilies) (Table 1), but will not be applied to the ECF transporters. Currently, the TC system does not contain the category of ECF transporters, but rather classifies certain members or components of ECF systems as ABC systems or as secondary active transporters.

Table 1.   Subfamilies of canonical ABC importers*
  • *

    The second column refers to the TC code.

  • Please note that the abbreviation ThiT is also used for the S units of thiamine-specific ECF transporters.

Carbohydrate uptake transporter-1 (CUT1)3.A.1.1.1–3.A.1.1.33Mono-, di- and oligosaccharides, glycerolphosphates, polyolsMaltose/maltodextrin transporter MalE FGK2 (E. coli/Salmonella)
Carbohydrate uptake transporter-2 (CUT2)3.A.1.2.1–3.A.1.2.13Monosaccharides (tetraoses, pentoses, hexoses, purine nucleosides), autoinducer AI-2Ribose porter RbsABCD, arabinose porter AraFGH, galactose/glucose porter MglABC (all from E. coli)
Polar amino acid uptake transporter (PAAT)3.A.1.3.1–3.A.1.3.20Polar amino acids, octopine, nopaline, diaminopimelate, cystine, cystathionine, ectoine/hydroxyectoineHistidine, lysine, arginine, ornithine porter HisJ/LAO-HisQMP2 (S. Typhimurium); Glutamine porter GlnHPQ, arginine porter ArtI/ArtJ-ArtMQP (both from E. coli); Arginine, lysine, histidine, ornithine porter ArtJ-MP (Geobacillus stearothermophilus)
Hydrophobic amino acid uptake transporter (HAAT)3.A.1.4.1–3.A.1.4.6Hydrophobic amino acids, general amino acids, urea/thiourea/hydroxyl-ureaLeucine, isoleucine, valine porter LivJ/LivK-LivHMGF (E. coli)
Peptide/opine/nickel uptake transporter (PepT)3.A.1.5.1–3.A.1.5.24Oligopeptides, mono-, di- and oligosaccharides, β-glucosides, glutathione, proline/betaine, microcins, EDTAOligopeptide porters OppABCDF (S. Typhimurium, Lactobacillus lactis); Ni porter NikABCDE (E. coli)
Sulfate/tungstate uptake transporter (SulT)3.A.1.6.1–3.A.1.6.6Sulfate, thiosulfate, tungstate, vanadate, molybdateSulfate/thiosulfate porter CysPTWA (E. coli)
Phosphate uptake transporter (PhoT)3.A.1.7.1–3.A.1.7.2PhosphatePstSACB (E. coli)
Molybdate uptake transporter (MolT)3.A.1.8.1–3.A.1.8.2Molybdate, tungstateMolybdate porter ModABC (E. coli, Methanosarcina acetivorans)
Phosphonate uptake transporter (PhnT)3.A.1.9.1–3.A.1.9.2Phosphonate, organophosphate ester, phosphatePhosphonate/phosphate porter PhnCDE (Mycobacterium smegmatis)
Ferric iron uptake transporter (FeT)3.A.1.10.1–3.A.1.10.3Fe3+ (Ga3+, Al3+)Ferric iron porter SfuABC (Serratia marcescens)
Polyamine/opine/phosphonate uptake porter (POPT)3.A.1.11.1–3.A.1.11.6Polyamines, mannopine, chrysopine, aminoethylphosphonate, γ-aminobutyratePolyamine porter PotABCD (E. coli)
Quarternary amine uptake transporter (QAT)3.A.1.12.1–3.A.1.12.11Glycine betaine, proline, choline, carnitine, ectoineGlycine betaine/proline porter ProVWX (E. coli), OpuAA-OpuAB porter (L. lactis)
Vitamin B12 uptake transporter (B12T)3.A.1.13.1Vitamin B12BtuCDF (E. coli)
Iron chelate uptake transporter (FeCT)3.A.1.14.1–3.A.1.14.20Fe3+, Fe3+-citrate, Fe3+-enterobactin, Fe3+-hydroxamate, iron-chrysobactin, heme, iron-vibriobactin, iron-vibrioferrin, corrinoid, desferrioxamine, coelichelin, bacillibactinFe3+, Fe3+-citrate porter FecBCDE (E. coli); Fe3+-enterobactin porter FepBCDG (E. coli); Corrinoid porter BtuCDE (Halobacterium sp.) heme porter Shp/HtsABC (Streptococcus pyogenes)
Manganese/zinc/iron chelate uptake transporter (MZT)3.A.1.15.1–3.A.1.15.11Mn2+, Zn2+, Fe2+Mn2+/Zn2+ porter MntABC (Neisseria meningitides)
Nitrate/nitrite/cyanate uptake transporter3.A.1.16.1–3.A.1.16.3Nitrate, nitrite, cyanate, bicarbonateNitrate/nitrite porter NrtABCD (Synechococcus sp.)
Taurine uptake transporter (TauT)3.A.1.17.1–3.A.1.17.5Taurine (2-aminoethane sulfonate), aromatic sulfonate, hydroxymethylpyrimidine, phthalateTaurine porter TauABC (E. coli), phthalate porter OphFGH (Burkholderia capacia)
Thiamin uptake transporter (ThiT)3.A.1.19.1Thiamin, thiamin monophosphate, thiamine pyrophosphateThiBPQ (S. Typhimurium)
Brachyspira iron uptake transporter (BIT)3.A.1.20.1IronBitABCDEF (B. hyodysenteriae)
Siderophore-Fe3+ uptake transporter (SIUT)3.A.1.21.1–3.A.1.21.2Fe3+-yersiniabactin, Fe3+-carboxymycobactinFe3+-yersiniabactin porter YbtP (Yersinia pestis)
Methionine uptake transporter (MUT)3.A.1.24.1–3.A.1.24.4l- and d-methionine, formyl-l-methionineMethionine porter MetNIQ (E. coli)
γ-Hexachlorocyclohexane (HCH) family3.A.1.27.1–3.A.1.27.2γ-HexachlorocyclohexaneLinKLMN porter (Sphingobium japonicum)

Canonical ABC importers

Modular organization

The architecture of canonical ABC importers reflects their universal mechanism. All members consist of two TMDs that form the translocation pore and an NBD dimer at the cytoplasmic face of the membrane. The NBDs bind and hydrolyze ATP, thereby generating conformational changes that are coupled to the TMDs and ultimately lead to substrate translocation. These domains can be arranged as a single fused polypeptide chain, as half-transporters or – in prokaryotic ABC importers only – as separate proteins. Importers require an additional component, the SBP, for function (Fig. 1).

Figure 1.

 Diversity in the domain architecture of canonical ABC importers. Examples: (a) vitamin B12 transporter, Btu(CD)2; (b) maltose/maltodextrin transporter, MalFGK2; (c) Fe3+-hydroxamate transporter, FhuBC2; (d) oligopeptide transporter, OppBCDF; (e) arabinose transporter, AraH2G (ATPase activity of the C-terminal NBD is unclear as denoted by a question mark); (f) nitrate/nitrite transporter, NrtM2CD (the second NBD-NrtC- is a C-terminal fusion with a solute-binding domain). Examples (a)–(e) are from Escherichia coli, (f) is from Synechococcus. SBP, (extracytoplasmic) solute-binding protein; TMD, transmembrane domain; NBD, nucleotide-binding domain; CM, cytoplasmic membrane. See text for details.

NBDs

The NBDs can be considered as the ‘motor domains’ of ABC transporters. The tertiary structure and catalytic cycle of these domains are relatively well understood (Schmitt & Tampé, 2002; Davidson & Chen, 2004; Dawson et al., 2007; Davidson et al., 2008). NBDs share a conserved architecture among im- and exporters. They consist of a RecA-like and an α-helical subdomain, interconnected by two flexible loop regions, one of which contains a conserved glutamine residue (Q-loop). The RecA-like subdomain is also found in other ATPases, whereas the helical subdomain is specific for the NBDs of ABC systems (Davidson et al., 2008).

In the assembled transporter, NBDs are present as a dimer, with the nucleotide-binding sites of each monomer facing the other. Binding and hydrolysis of ATP are catalyzed by residues from several conserved motifs (see Davidson & Chen, 2004 as a general reference). Both of the so-called ‘Walker’ motifs map to the RecA-like subdomain. The ‘Walker A’ motif forms a loop structure that binds to the β- and γ-phosphate groups of ADP or ATP, while the ‘Walker B’ motif is arranged as a β-strand, whose terminal aspartate coordinates a Mg2+ ion through water needed for ATP hydrolysis. The ‘Q-loop’ is part of the loop region connecting RecA-like and helical subdomains. It is characterized by a conserved glutamine that binds to the Mg2+ cofactor. Located at the dimer interface as well as at the interface to the TMDs, this region is probably involved in interdomain signaling (Dawson et al., 2007). Mobility of the Q-loop is essential to complete the catalytic cycle of the transporter (Daus et al., 2007b). Moreover, NBDs contain the ‘H-loop’ or ‘switch’ with a conserved histidine residue that contacts the γ-phosphate of ATP and is required for hydrolysis, and finally the ‘signature’ or the ‘LSGGQ’ motif (C-loop) (Schneider & Hunke, 1998).

The dimeric arrangement of the NBDs corresponds well with the finding that ATPase activity is highly cooperative (Davidson et al., 1996; Liu et al., 1997; Moody et al., 2002). According to several crystal structures of dimeric NBDs (Hopfner et al., 2000; Smith et al., 2002; Chen et al., 2003) and a full transporter (Oldham et al., 2007), the NBD monomers form a ‘nucleotide sandwich’ (Fig. 2). In this arrangement, ATP is bound along the dimer interface by Walker A residues of one monomer and residues from the LSGGQ motif of the other monomer. Upon binding of two ATP molecules, the NBD dimer closes in a ‘tweezer-like’ fashion (Chen et al., 2003). Hydrolysis of ATP occurs in the tightly dimerized state, which then reopens to release phosphate. The ADP-bound conformer of the NBD dimer resides in an intermediary, semi-open state, which presumably returns to the open apo-state after the dissociation of ADP (Lu et al., 2005). The observed positive cooperativity of ATP hydrolysis makes it highly likely that two molecules of ATP are needed for NBD dimer closure (Chen et al., 2003; Davidson et al., 2008).

Figure 2.

 Structures of the maltose/maltodextrin transporter of Escherichia coli. The transporter consists of MalE (SBP, yellow), MalF (TMD, blue), MalG (TMD, magenta) and two copies of the NBD MalK (shown in green and red for clarity). MalFGK2 represents the inward-facing (apo-) state and MalE-FGK2 represents the outward-facing (intermediate) state of the transporter. In the latter, MalE, in its open conformation, is tightly associated with MalFG, maltose resides within the pore by interaction with residues from MalF only and two ATP molecules are sandwiched between both MalK monomers. The large periplasmic P2 loop of MalF, which is not resolved in the apo-state, is in close contact to MalE. See text for details. The figure was drawn with discovery studio visualizer 2.5 (Accelrys, Cambridge, UK) using the coordinates from entry 3FH6 and 2R6G, respectively, in the Brookhaven Protein Data Bank.

TMDs

The TMDs mainly consist of four to 10 membrane-spanning α-helices. As compared with the NBDs, the sequence of the TMDs is less conserved. A substrate-binding site has thus far been clearly identified only in the MalF subunit of the E. coli maltose importer MalFGK2 (Oldham et al., 2007) (Fig. 2). One maltose molecule is interacting with 10 surrounding residues through H-bonds, van der Waals interactions and aromatic ring stacking. Six of these residues were already known to reduce transporter activity when mutated in vivo (Oldham et al., 2007). Intra- and extracellular loops connecting the transmembrane segments are of varying lengths and possess functional significance for contacting the NBDs or extracellular interaction partners. The ‘coupling helices’ are architecturally conserved elements forming the NBD–TMD interface. In the crystal structures, they are located in grooves on the NBDs' surfaces (Locher, 2009).

Nucleotide-induced conformational changes are thought to be transmitted via noncovalent interactions from the Q-loop region of the NBDs to these short helices. In prokaryotic importers, the coupling helices are characterized by a conserved sequence of amino acids [consensus EAA-X3-G-X9-I-X-LP (Saurin et al., 1994)]. These ‘EAA’ or ‘L’ loops map to the last cytoplasmic loop of the respective TMDs (Mourez et al., 1997). Certain loops on the extracellular side are possibly involved in the recognition or the stabilization of interacting proteins or subsequent signal transduction. Large loops are found, for example in MalFG of the enterobacterial maltose transporter (comprising 180 amino acids) (Oldham et al., 2007).

SBPs

Extracellular SBPs (also called receptors) are essential components of canonical ABC importers. In gram-negative bacteria, the binding proteins diffuse freely in the periplasm between the inner and the outer membrane. Gram-positive bacteria and archaea anchor the binding proteins to the outer surface of the cell membrane via an N-terminal lipid moiety (Sutcliffe & Russell, 1995) or, as observed for some archaea, an N-terminal hydrophobic helix. Moreover, solute-binding domains can also be fused to the TMDs, resulting in two or four substrate-binding sites (van der Heide & Poolman, 2002; Biemans-Oldehinkel et al., 2006a) (Fig. 3).

Figure 3.

 Structural organization of SBPs. In gram-negative bacteria, SBPs reside in the periplasm (a) while in gram-positive bacteria and archaea, lacking an outer membrane, they are attached to the extracellular side of the cytoplasmic membrane via an N-terminal lipid anchor (solute-binding lipoproteins are also found in cyanobacteria) (b). In some archaea, the SBP is associated with the membrane by an N-terminal peptide (c). In some bacteria and archaea, including Lactococcus lactis, Helicobacter pylori and Sulfolobus solfataricus, transporters exist that comprise two or four SBPs fused to the TMDs (d). NBDs and TMDS [except in (d)] are omitted for clarity. CM, cytoplasmic membrane.

Substrates bound by SBPs with a high specificity are highly diverse, ranging from inorganic or organic ions and sugars to peptides or vitamins (Wilkinson & Verschueren, 2003). Because of their high substrate affinities (Kd values typically range from 0.01 to 1 μM), capture and accumulation of substrate in proximity to the transporter can be regarded as the main raison d'être of SBPs. However, as these proteins are also indispensible at very high substrate concentrations, they probably also play an important functional role in the catalytic cycle of the transporter (Shilton, 2008) (discussed below).

Generally, SBPs consist of two symmetrical lobes or domains (termed N- and C-lobe according to the protein's termini), both of which display an α-/β-fold, connected by a hinge region (Quiocho & Ledvina, 1996; Davidson et al., 2008). The amino acid sequences of SBPs were originally grouped into eight clusters or subfamilies based on their similarities (Tam & Saier, 1993). Alternatively, SBPs have been classified by the topology of their globular domains (Fukami-Kobayashi et al., 1999; Wilkinson & Verschueren, 2003; Dwyer & Hellinga, 2004) or by the number of interdomain connections (Quiocho & Ledvina, 1996). In all three families or groups identified so far, the substrate-binding site is located in a cleft between the two lobes. Substrate specificity is achieved mostly by differential H-bonding. In type I and II SBPs, the lobes differ in the number and order of β-strands and are connected by a flexible hinge region. They are represented, for example by the E. coli galactose/glucose (MglB) – and maltose-binding proteins (MalE), respectively. Substrate binding causes a rotation of these lobes toward each other, rendering the central cleft inaccessible from the aqueous surrounding (‘Venus flytrap’ model, Quiocho & Ledvina, 1996).

In type III SBPs (constituting the new cluster 9 based on sequence similarities), as represented, for example by the E. coli vitamin B12-binding protein, BtuF, the two ligand-binding domains are connected by a rigid α-helix, suggesting a rather small conformational change upon ligand binding (Lawrence et al., 1998; Lee et al., 1999; Karpowich et al., 2003). However, a molecular dynamics simulation indicated that BtuF appears to be more flexible than implied by the X-ray structures, displaying clear opening and closing motions that are compatible with the Venus fly trap model (Kandt et al., 2006).

Interaction patterns of SBPs with the transmembrane domains of the transporter could be system specific, as indicated by the different affinities of SBPs for their cognate membrane components (Chen et al., 2001; Hollenstein et al., 2007; Oldham et al., 2007). Positive cooperativity between two substrate-binding domains fused to the TMDs exists in the glycine betaine transporter OpuA from L. lactis (Biemans-Oldehinkel et al., 2006b).

Binding protein-independent (BPI) mutants have been described for the maltose and the histidine transporters of E. coli and S. Typhimurium, respectively. In case of the maltose transporter, the mutations map to transmembrane regions of the TMDs, MalF and MalG. The phenotype of these mutants is characterized by a spontaneous ATPase activity and a decline in affinity for maltose by three orders of magnitude. Moreover, maltodextrins are no longer recognized as substrates (Covitz et al., 1994). Maltose-binding protein actually inhibits the activity of the mutants, likely by preventing the access of free maltose to the substrate-binding site within the pore (Dean et al., 1992). BPI mutants are thought to reside in a transition state-like conformation (Mannering et al., 2001; Daus et al., 2006, 2007b, 2009). As inferred from the known crystal structures of the transporter, all mutations lie at sites of interaction that are altered in the inward-to-outward transition (Khare et al., 2009).

In contrast, mutations causing a BPI phenotype of the histidine transporter all map to a region in HisP, the NBD of the system, immediately preceeding the Walker B motif and in between the D and the H loops (Speiser & Ames, 1991). These mutants exhibit high spontaneous ATPase activity, but rather low transport rates in vitro that can be stimulated substantially by the binding protein, HisJ (Liu et al., 1999). The latter finding is consistent with the notion that the phenotype is caused by an altered nucleotide-binding site (Speiser & Ames, 1991).

The ‘alternating access’ model of transport

X-ray structures of isolated NBDs and complete transporters as well as biophysical evidence support an alternating access mechanism for substrate transfer (Borbat et al., 2007; Hollenstein et al., 2007; Oldham et al., 2007; Khare et al., 2009): conformational changes generated in the NBDs by ATP binding and hydrolysis cause an alternation between an outward-facing (open to the exterior) and an inward-facing (open to the cytoplasm) structure of the pore (Figs 2 and 4). In the following, we will present a short transport scenario that is largely based on structural, biochemical and biophysical data accumulated for the maltose transporter of E. coli/Salmonella, by far the best-studied ABC importer to date, but also includes evidence from other systems (Boos & Shuman, 1998; Davidson & Chen, 2004; Daus et al., 2006, 2007a, b; Oldham et al., 2007; Grote et al., 2008, 2009; Orelle et al., 2008; Khare et al., 2009). The model might not be applicable in all aspects to larger ABC importers, such as the vitamin B12 transporter BtuCD-F from E. coli (Lewinson et al., 2010). For a more detailed discussion of current transport models, the reader is referred to (Davidson et al., 2008; Locher, 2009; Rees et al., 2009).

Figure 4.

 Simplified model of the translocation mechanism of an SBP-dependent ABC transporter. In the resting state (I), SBP and NBDs are in their open conformations while the TMDs expose the substrate-binding site within the pore to the cytoplasm (inward-facing). Under physiological conditions, ATP is likely to bind to the NBDs. When a substrate becomes available, the SBP is shifted to the closed state (II), thereby triggering a conformational change that results in closing of the NBD dimer, opening of the TMDs to the extracellular side (outward-facing) and release of the substrate into the pore (III). ATP hydrolysis enables the TMDs to revert to the inward-facing conformation, thereby completing translocation of the substrate to the cytoplasm. Subsequent dissociation of phosphate and ADP and reloading of the NBDs with ATP restore the resting state. See text for further details.

The maltose transporter consists of the membrane-integral subunits, MalF and MalG, a homodimer of the ATPase, MalK and the periplasmic maltose-binding protein, MalE. In the absence of substrate, the resting state is thought to be represented by the ATP-bound form of the transporter (Daus et al., 2007a; Oldham et al., 2008) (Fig. 4, state I). Although it is usually considered that MalE docks onto the MalFGK2 complex in the substrate-loaded, closed conformation, several lines of experimental evidence suggest a permanent association of a copy of MalE with the transport complex (Daus et al., 2007b, 2009; Grote et al., 2009) (see Daus et al., 2007a, for a detailed discussion) (Fig. 4, state II).

As a result of substrate binding, MalE is shifted to its closed conformation, which in turn promotes a tight association of the MalK monomers, thereby resulting in opening of the transport channel to the periplasmic side. Concomitantly, MalE is pushed toward its open conformation (Orelle et al., 2008), the substrate becomes released and enters the pore. At this stage, MalE forms a stable complex with the transporter (Oldham et al., 2007) (Fig. 4, state III). The unusually large periplasmic P2 loop of the MalF subunit plays a pivotal role in communicating substrate availability to the NBDs (Grote et al., 2009). The chemical nature of the substrate appears not to be important for initiating the transport cycle as demonstrated using a MalE mutant that binds either maltose or sucrose (Gould & Shilton, 2010). The scenario is complicated by the observation that unliganded binding proteins can also have a stimulatory effect on the ATPase activity of the transporter. In case of the maltose transporter, this effect was contributed to open, unliganded MalE, which might interact with a yet to be identified small population of transporter molecules in the required conformation (Gould et al., 2009).

The following steps are still highly speculative. Hydrolysis of one ATP, as can be experimentally achieved by preventing dissociation of ADP with vanadate, results in no or little opening of the MalK dimer (Daus et al., 2007b; Grote et al., 2008), and thus might push the substrate only further into the pore. However, release of the substrate to the cytoplasm at this stage cannot yet be fully excluded (Chen et al., 2001). Hydrolysis of the second ATP would then switch the transporter back to the inward-facing conformation, thereby lowering the affinity of the binding protein for the transporter and allowing release of the substrate in the case of the first scenario. Dissociation of phosphate and ADP and subsequent rebinding of ATP would restore the resting state.

The model implies that two molecules of ATP are consumed per substrate molecule transported, which is consistent with a study on the glycine-betaine importer, OpuA, of L. lactis (Patzlaff et al., 2003). However, the numbers of ATP molecules hydrolyzed and substrate molecules translocated per catalytic cycle as well as the mechanistic role of each hydrolysis step are still a matter of debate. For a detailed discussion, see for example Davidson et al. (2008).

ECF transporters

Classification and modular organization

ECF transporters were shown to be an abundant class of importers for micronutrients in bacteria and archaea. The evolving list of substrates contains a set of water-soluble vitamins, cofactors and their metabolic precursors, the transition-metal ions Ni2+ and Co2+, the amino acid tryptophan and queuosine and its metabolic precursors (Table 2). Many bacteria with restricted cofactor biosynthetic capacities rely on vitamin uptake by ECF import systems that are encoded by essential genes in several gram-positive pathogens including Streptococcus pneumoniae, Staphylococcus aureus and Mycoplasma genitalium (Rodionov et al., 2009).

Table 2.   Known and predicted substrate specificities of S components of ECF transporters
Vitamins, cofactors and precursors
 BioYBiotin
 RibURiboflavin, FMN
 FolTFolate, 5-formyltetrahydrofolate
 ThiTThiamine
 PanTPantothenate
 YkoEHydroxymethylpyrimidine (HMP)
 ThiWThiazole
 PdxUPyridoxine
 NiaXNiacin
 LipTLipoate
 CbrTCobalamin
 CblTDimethylbenzimidazole (DMB)
 MtsTS-adenosylmethionine (SAM)
 MtaTMethylthioadenosine
Transition metal ions
 NikMNickel (Ni2+)
 CbiMCobalt (Co2+)
Amino acids
 TrpPTryptophan
Queuosine and its precursors
 QrtTQueuosine
 QueT7-Aminomethyl-7-deazaguanine (preQ1)

Like the aforementioned canonical ABC importers, ECF systems contain ABC ATPase subunits and are thus considered to couple ATP hydrolysis to substrate uptake. Fundamental differences between classical ABC importers and ECF transporters lie in the modular architecture and in the absence of extracytoplasmic substrate-binding proteins among the ECF systems.

ECF transporters consist of pairs of ABC ATPase domains (A components), a conserved transmembrane protein (T component) and a transmembrane substrate-capture protein (S component, ≈21 kDa in many cases) in an unknown stoichiometry. As depicted in Fig. 5, the systems contain two different A components (A1, A2), two copies of the same A component (A1, A1) or two ABC ATPase domains fused to yield a single polypeptide (A1, A2, indicated by encircling in the very left panel in Fig. 5). Based on the utilization of a dedicated (subclass I) or a shared (subclass II) AAT module, ECF systems fall into two groups. Subclass I ECF systems are encoded in operons containing one or two genes for A units, a T-unit and an S-unit gene. About half of the ECF transporters – subclass II, found in most Firmicutes, Thermotogales and some archaeal species – share the same EcfA1A2T module, i.e. various and highly diverse S components in a cell compete for the same module in order to form the holotransporters. These EcfA1A2T modules are encoded by operons, whereas the S-unit genes are scattered around the genome. In Firmicutes, ecfA1A2T genes are part of larger operons encoding essential cellular functions (ribosomal proteins, RNA polymerase subunit, pseudouridine synthase, etc.), suggestive of their robust and constitutive expression.

Figure 5.

 Modular architecture of ECF transporters. ECF transporters consist of (1) two different ABC ATPases (A1, A2) that are fused to a single protein in certain subclass I importers or copies of the same ABC protein (A1, A1), (2) a conserved transmembrane protein (T) and (3) a substrate-specific integral membrane protein (S). Cobalt- and nickel-transporting ECF systems contain bipartite (or rarely tripartite, not shown) S units (green contours). Shared use of the same A1A2T module by many highly diverse S proteins is a hallmark of subclass II systems. Subclass I ECF transporters have dedicated A and T units.

Recent bioinformatic analyses identified >20 highly diverse families of S components, each with a distinct predicted or an experimentally established substrate specificity (Rodionov et al., 2009). Most of these S-unit genes are individually expressed under control of corresponding metabolite-responsive riboswitches or transcription factors (Rodionov et al., 2009). The presence of ecfA1A2T operons in Methanosarcina sp. that do not contain any of the previously identified S components points to the existence of S components of ECF transporters with as yet unrecognized specificities.

Properties

Analyses of bacterial Co2+ and biotin transporters provided an initial indication of the role of individual components of ECF transporters. Heterologous production of the CbiMNQO system from S. Typhimurium and of the α-purple bacterium Rhodobacter capsulatus in E. coli resulted in recombinants with Co2+-uptake activity (Rodionov et al., 2006). Whereas the role of CbiO (A unit) as an NBD that needs to dimerize for function was immediately obvious from its amino-acid sequence, clues to the roles of the three transmembrane proteins CbiM, CbiN and CbiQ were not provided by sequence analysis. Mature CbiM proteins have seven predicted TMDs and an invariant His residue at position 2. CbiNs are unique small integral membrane proteins with two TMDs and a large extracellular loop in between. CbiQ proteins belong to the family of T proteins. Sequential deletion of genes from the 3′-end of cbiMNQO operons and assays of Co2+-transport activity in recombinant E. coli yielded the surprising result that significant basal activity was detectable for CbiMNQ and CbiMN. This finding suggests a secondary active mode of transport in the absence of the NBD (CbiO) and identified CbiMN as the minimal module required for basal uptake activity (Rodionov et al., 2006; P. Hebbeln & T. Eitinger, unpublished data). The Co2+ (and Ni2+) transporters are an exception among the ECF systems inasmuch as their S components (e.g. CbiMN) are heterooligomeric.

More detailed analyses were reported for the R. capsulatus biotin transporter BioMNY. The sequence similarities between CbiO and BioM (A components) and between CbiQ and BioN (T components) in conjunction with the strong diverseness between CbiMN and BioY (S components) suggested that BioMN and BioY represent the energy-coupling module and the S unit, respectively (Hebbeln et al., 2007). Biochemical analyses of the BioMNY system confirmed its modular organization and the function of BioY as the S unit. Notably, this protein can function as a high-capacity transporter in its solitary state. High-affinity transport reflecting the requirements in the natural environment, however, depends on the presence of a functional BioM ATPase. Copurification studies revealed that BioM and BioN form stable bipartite complexes and tripartite complexes together with BioY. Because the expression of bioMY and bioMNY resulted in comparable high-affinity biotin-uptake rates of recombinant E. coli, compared with low-affinity uptake caused by the solitary bioY, the existence of BioMY complexes in the absence of BioN (the T component) is conceivable. Such complexes, however, could not be purified, suggesting that BioN is required as a stabilizing factor. ATPase assays with purified components in a detergent solution, as well as experiments with proteoliposomes, found that BioM is capable of hydrolyzing ATP only in the presence of a transmembrane partner. (Hebbeln et al., 2007; A. Alfandega & T. Eitinger, unpublished data).

Shared use of the same EcfA1A2T module by many diverse S units is an amazing feature of subclass II ECF transporters and was experimentally demonstrated for lactobacterial folate, pantothenate, riboflavin and thiamine importers (Neubauer et al., 2009; Rodionov et al., 2009). The molecular patterns that allow efficient recognition between the energy-coupling module and various highly diverse S components remain elusive. The T units among the various ECF transporters differ in the number of TMDs and membrane topology (Fig. 6). A function of T units beyond that of a molecular adhesive that holds together the S and A components was recently shown for the BioMNY transporter, and likewise, for the promiscuous vitamin (folate, pantothenate and riboflavin) transporters of the lactobacterium Leuconostoc mesenteroides (Neubauer et al., 2009). Two short motifs with an arginine residue in the center, frequently occurring as Ala-Arg-Gly and located on the cytoplasmic face in the C-terminal region, are the most conserved feature of T units (Fig. 6). Individual replacements of the Arg residue by Glu in motif II introduced into BioN and EcfT led to diminished activity of the biotin transporter and inactivated the L. mesenteroides vitamin transporters, respectively. The double mutations strongly destabilized the transporter complexes. Single replacements of the Arg residue in motif I affected the activity of both the biotin and the subclass II vitamin transporters, but did not destabilize the complexes. The latter finding suggests that T units are essential for intramolecular signaling and effective coupling of ATP hydrolysis to substrate translocation.

Figure 6.

 Topology of T components. XRX, conserved three-amino-acid motifs, mainly found as ARG, SRG, LRG, VRG, MRG or ARS. cTP, predicted plastid-specific transit peptide in a couple of plant T components (see ‘ABC and ECF importers in plants?’ for a description of ‘Cyano_T’ and T components in plants). Classification of T units: EcfT, promiscuous systems; BioN, biotin transporter; CbrV, transporter for cobalamin precursor; QrtU, transporter for queuosine precursor; YkoC, transporter for thiamine precursor.

Substrate recognition

The substrate specificity of the S components CbiMN (Co2+), NikMN (Ni2+) (Rodionov et al., 2006), BioY (biotin) (Hebbeln et al., 2007), FolT (folates) (Eudes et al., 2008; Rodionov et al., 2009), PanT (pantothenate) (Neubauer et al., 2009), RibU (riboflavin) (Burgess et al., 2006; Duurkens et al., 2007; Vogl et al., 2007) and ThiT (thiamine) (Eudes et al., 2008; Rodionov et al., 2009; Schauer et al., 2009) has been analyzed experimentally. The metal specificity of CbiMNQO and NikMNQO systems was addressed by uptake assays with recombinant E. coli cells expressing the cbiMNQO operons of R. capsulatus or S. Typhimurium, and the nikMNQO operons of R. capsulatus or Methanothermobacter thermoautotrophicus (Rodionov et al., 2006; Hebbeln, 2009). These studies showed a strong preference for Co2+ over Ni2+ for the Cbi systems and the opposite for the Nik systems. PanT of L. mesenteroides was shown to transport pantothenate, provided that the EcfA1A2T components were coproduced (Neubauer et al., 2009). BioMNY mediates high-affinity biotin transport into recombinant E. coli cells; the biotin concentration resulting in half-maximal transport velocity was estimated to 5 nM (Hebbeln et al., 2007). High-affinity, but substoichiometric binding of biotin was shown for the purified BioMNY complex and the solitary BioY in detergent solution by equilibrium dialysis (A. Alfandega & T. Eitinger, unpublished data). Analysis of riboflavin binding to the L. lactis S unit RibU, purified from cells grown in the absence of riboflavin, by isothermal titration calorimetry, flow dialysis and fluorescence titration, revealed a 1 : 1 stoichiometry (riboflavin : RibU) and a dissociation constant (Kd) of 0.6 nM. FMN and the toxic flavin analog roseoflavin also bind to RibU with a high affinity, but FAD does not (Duurkens et al., 2007). Growth experiments and riboflavin-uptake experiments with Bacillus subtilis suggested that the RibU-containing transporter accepts riboflavin and its analogs in the order riboflavin≈roseoflavin>FMN>>FAD (Vogl et al., 2007). Dissociation constants in the nanomolar range were also determined via quenching of intrinsic tryptophan fluorescence upon addition of vitamins to the purified L. casei FolT and ThiT. The reported Kd values are 9 nM folate and 5 nM 5-formyltetrahydrofolate for FolT, and 0.5 nM thiamine for ThiT. Binding stoichiometries are as low as about 1 : 10 and this finding was explained by copurification of vitamins that are tightly bound to the S units and occupy binding sites in the ‘as-isolated’ state (Eudes et al., 2008). Thus, the available data point to extremely high affinity as a characteristic trait of ECF transporters that renders them well suited for their role in micronutrient uptake.

Genomic and phylogenetic analysis

The comparative analysis of ∼400 microbial genomes using the SEED platform (Overbeek et al., 2005) provided the basis for the identification of a variety of families of ECF transporters and the genomics-based prediction of their specificities (Rodionov et al., 2009). Similarity searches with the cbiQO and bioMN genes identified multiple homologs of A- and T-component genes unevenly distributed in prokaryotic genomes. About one half of these genes were found in potential operons with genes encoding unrelated hypothetical transmembrane proteins (S components). The S components were classified into ∼20 families. A combination of genome context analysis (gene clustering, coregulation, co-occurrence) and metabolic reconstruction was successfully utilized for the prediction of substrate specificities for the majority of S-component families; many of these were later validated in experiments (Table 2 and ECF-type transporters).

The distribution of ECF-transporter components across prokaryotic genomes was captured in the ‘ECF class transporters’ subsystem in the SEED comparative genomics resource (http://theseed.uchicago.edu/FIG/index.cgi). This subsystem provides an easy access to information about the sequences, genomic locations, functional annotations and distribution of ECF component proteins in a periodically updated genome collection. Table 3 presents an overview of the distribution of transport systems per ECF family in major taxonomic classes of prokaryotes. The tabulated subclass I and subclass II ECF systems include ∼700 S components and ∼400 AT (or AAT) modules in 35 archaeal and 168 bacterial genomes. Of these, the majority of these S components were detected in Firmicutes (68%), Actinobacteria (8%), Proteobacteria (8%), Archaea (8%), and Thermotogae (3%). Overall, nearly one half of the analyzed prokaryotic species have at least one ECF transporter.

Table 3.   Occurrence of subclass I and subclass II ECF transporters in prokaryotic genomes
 # genomesArchaeaBacteriaTotal
CrenarchaeotaEuryarchaeotaKorarchaeotaThermotogaeFirmicutes/BacillalesFirmicutes/ClostridiaFirmicutes/
Lactobacillales
Firmicutes/
Mollicutes
ActinobacteriaProteobacteriaChloroflexiDeinococcus-ThermusFusobacteriaCyanobacteriaSpirochaetesChlamydiaeBacteroidetesChlorobiAquificaePlanctomycetes
12291715172517241804311955163384
  1. The number of genomes possessing a particular type of ECF transporter is indicated. Families of S components that occur in both subclasses are indicated by an asterisk. ‘BioY alone’ shows the distribution of BioY transporters that are neither accompanied by a dedicated nor a shared ‘AAT’ module.

Subclass I*BioMNY1121121721248
NikMNQO21919241661161
CbiMNQO11613151293152
YkoCDE3111131442
MtsTUV1141173633
*HtsTUV11565119
*QrtTUVW1135616
*CbrTUV1352617
MtaTUV1214
*RibU +AAT123
*PanT +AAT112
*TrpP +AAT729
*QueT +AT22
*PdxU +AAT2114
*ThiW +AAT347
*CblT +AAT11
Subclass IIEcfA1–A2–EcfT171715162517190
*BioY461516221165
*RibU16131125561
*PanT1823436
ThiT81111333
FolT18141437
*TrpP5914
*QueT111711181141
*PdxU143111938
*ThiW1559121
NiaX71118
*CblT3811
*HtsT22
*QrtT257
*CbrT11112
LipT33
BioY alone3422119554

The relative abundance of various S components of subclass I and subclass II ECF transporters in the analyzed prokaryotic genomes is presented in Fig. 7. BioY is the most abundant S unit and belongs to either subclass I (48 cases) or subclass II (65 cases) ECF systems. Another 54 bioY genes occur in organisms that do not contain any AAT module genes. Five S component families (NikM, CbiM, YkoE, MtsT and MtaT) always belong to subclass I ECF systems. Three families (CbrT, QrtT and HtsT) are mostly represented by subclass I systems with exceptions when some of their homologs (found in Firmicutes) appear to belong to subclass II. The remaining 11 S component families mainly belong to the subclass II ECF systems and were predominantly found in Firmicutes and Thermotogae. Among seven such families, a small proportion of S-component genes (26 cases, found mainly in Archaea and Actinobacteria) are accompanied by AAT module genes and were classified into subclass II systems.

Figure 7.

 Relative distribution of ECF transporters in prokaryotic genomes. The occurrence of 20 families of S components was analyzed in 341 bacterial and 42 archaeal genomes. The total numbers of organisms that encode members of different families classified to either subclass I (black bars) or subclass II (white bars) of ECF transporters are presented. The gray bar for the BioY family indicates the number of organisms that encode a solitary BioY protein, but neither a dedicated (BioMN) nor a shared (EcfA1A2T) module.

Major families of S components as well as all A and T components were analyzed by generating phylogenetic trees [illustrated with dendroscope (Huson et al., 2007) in the supplemental figures] using protein sequences collected from the ‘ECF class transporters’ subsystem in the SEED database. Preliminary similarity searches using the position-specific iterated (psi-) blast program did not reveal any significant cross-identification between different S-component families, with the only exception being the homologous CbiM and NikM families. These results agree with the apparently weak (if any) similarity detected between S proteins from different families (average 11%, maximum 21% identity) (Rodionov et al., 2009). Therefore, with the exception of the CbiM/NikM proteins, separate phylogenetic trees of other S components were generated (Supporting Information, Fig. S1). The CbiM/NikM tree is subdivided into two large branches corresponding to nickel and cobalt transporters (Rodionov et al., 2006). BioY proteins are most diverse; their phylogenetic tree contains multiple separate branches corresponding to the BioMN-associated homologs (subclass I transporters), the EcfA1A2T-dependent homologs (subclass II transporters) and the solitary BioY proteins (independent of any AAT module). For instance, the BioY proteins from various Alphaproteobacteria are not clustered together, but dispersed between two very diverged branches of the tree corresponding to the BioMNY systems and the solitary BioY proteins, respectively. This tree suggests the complex evolutionary history of the BioY family including multiple duplication and horizontal transfer events, and different rates of sequence divergence in the different organisms caused by different (if any) AAT module partners. In addition to BioY, some other S-component families (e.g. FolT, PdxU and QueT) also include multiple diverse paralogs in the same genome, suggesting a likely divergence in their substrate specificities.

The amino acid sequence identity between individual T components of subclass I and subclass II transporters and among T units of different subclass I systems varies considerably and lies between ∼15% and 60%. The phylogenetic tree of T components (Fig. S2a) was constructed using the representative subset of ∼320 prokaryotic proteins and 10 additional homologs encoded in genomes of the various plants. The tree shows an overall consistent clustering of the major functional groups of ECF transporters with several exceptions including multiple distinct branches representing the BioN and EcfT proteins from Archaea, and a separate branch of QrtU proteins from enterobacteria. Interestingly, EcfT homologs from plants form a separate branch together with homologs from Cyanobacteria (those that are not encoded by operons for subclass I systems), suggesting their common evolutionary origin. These cyanobacterial and plant EcfT homologs are not accompanied by any recognizable EcfA homolog and their interaction partners and function remain elusive (see ‘ABC and ECF importers in plants?’).

A phylogenetic tree of A components of ECF systems (Fig. S2b) was constructed using a joint collection of ∼260 single-domain ATPase proteins and ∼230 domains of duplicated ATPases (MtsU, YkoD, HtsV, CbrU and MtaU) from prokaryotes. The tree shows a clear separation of duplicated A domains into diverse branches, suggesting that the duplication of ATPase domains occurred early in their evolution. The only exception from this observation is a subgroup of YkoD proteins from Lactobacillales, whose N- and C-terminal ATPase domains are most closely related to each other, suggesting that they probably originated from a relatively recent duplication.

The NBDs from canonical ABC and ECF importers form two distinct branches in a phylogenetic tree. A more detailed comparison is presented in Common features of canonical ABC and ECF importers, and perspectives.

Cellular functions of ABC import systems

Binding protein-dependent ABC importers

Nutrient supply and cell growth

This chapter summarizes the current knowledge on canonical ABC import systems involved in the uptake of macronutrients, trace elements, vitamins and polyamines.

Carbon and energy sources

SBP-dependent ABC transporters mediating the uptake of a diverse range of carbohydrates for carbon and energy metabolism cluster primarily in the CUT1 and CUT2 families of the TC classification system (Saier, 2000; Schneider, 2001). The ‘earmark’ of CUT2 family members is the primary structure of their NBDs, which appear to be a ‘natural’ fusion of two ABC domains (Schneider, 2001). Some carbohydrate transporters of hyperthermophilic Archaea, for example Pyrococcus furiosus (Koning et al., 2001) and Sulfolobus solfataricus (Elferink et al., 2001), Thermotoga maritima (a hyperthermophilic bacterium) (Conners et al., 2005; Nanavati et al., 2006) and of the root nodule-forming bacterium Rhizobium meliloti are also found within the PepT subfamily, otherwise comprising transporters for peptides, opines and nickel ions.

While the per se existence of mono- and disaccharides in nature is rather limited (e.g. fruits, milk, sugar-rich diet of animals, including humans), the major sources of sugars that can be utilized by prokaryotes are plant- and fungi-derived polysaccharides, including starch, cellulose, xylan, pectin and chitin. Thus, most prokaryotes that thrive in soil, aquatic environments, sewage digesters or the digestive tracts of animals secrete hydrolases for the degradation of biopolymers. The resulting mono-, di- and oligosaccharides are then internalized primarily due to the action of specific ABC transporters.

Maltose, maltodextrins, cyclodextrins. The maltose/maltodextrin transporter MalEFGK2 of E. coli/S. Typhimurium, by far the best-characterized member of the CUT1 family that also serves as a model for ABC transporters in general (see The ‘alternating access’-model of transport), enables the facultative anaerobes at the entry of the large intestine to feed on sugars formed by the cleavage of starch from the animals' diet in the stomach and that escaped adsorption in the small intestine (Flint et al., 2008). Quantitatively predominant bacteria in the large intestine such as Bifidobacterium have multiple gene clusters encoding putative oligosaccharide ABC transporters in their genomes (Schell et al., 2002).

The receptor, MalE, binds maltose and maltooligosaccharides up to seven glucose units (Boos & Shuman, 1998), which are delivered to the MalFGK2 complex for translocation to the cytoplasm. In contrast, reduced or oxidized maltodextrins or β-cyclodextrin bind to MalE, but the resulting liganded receptor fails to initiate the transport cycle, due to a different binding mode (Hall et al., 1997a, b). The NBD MalK and MalFGK2 were purified and extensively characterized in liposomes (reviewed in Schneider, 2003; Davidson & Chen, 2004). Crystal structures were resolved of MalE complexed with various substrates (Quiocho et al., 1997), of the MalK dimer (reviewed in Davidson & Chen, 2004) and the complete transporter in the apo form in the absence of MalE (Khare et al., 2009) and in an ATP-bound form with tightly associated MalE (Oldham et al., 2007). The C-terminal extension of the MalK, the NBD of the maltose/maltodextrin transporter from E. coli/Salmonella, binds MalT, the positive transcriptional regulator of the mal regulon, thereby preventing its activation through binding of ATP and maltotriose and subsequent oligomerization (Joly et al., 2004; Richet et al., 2005). It is proposed that substrate availability is sensed through the transporter, which, in the idling mode, binds MalT and thereby represses mal gene transcription. In the presence of a substrate, however, transport activity is switched on, i.e. ATP is hydrolyzed at the MalK subunits, thus causing the release of MalT and subsequent induction of maltose-regulated gene expression (Boos & Böhm, 2000). The maltose transporter is also involved in a second regulatory process called ‘inducer exclusion’, which is part of the global carbon regulation in enteric bacteria. Here, in the presence of the preferred carbon source, glucose, the transport of inducer molecules for alternative metabolic pathways is prevented. This is achieved by inhibition of the respective transport systems via a component of the glucose transporter, the dephosphorylated enzyme IIAGlc of the phosphoenolpyruvate phosphotransferase system (PTS) (Postma et al., 1996). In the case of the maltose transporter, enzyme IIAGlc binds predominantly to the C-terminal extension of the MalK subunits, thereby inhibiting ATP hydrolysis (reviewed in Schneider, 2003). Although the precise mechanism of inhibition is still unknown, the structure of the MalK dimer suggests binding sites for enzyme IIAGlc on the C-terminal domain of one monomer and the N-terminal domain of the opposing monomer (Samanta et al., 2003). Although an additional C-terminal domain of their NBDs is shared by all members of the CUT1 family, in the majority of cases, nothing is known on the possible regulatory processes involved.

Other related transporters discriminate between maltose and maltooligosaccharides by specific properties of their cognate receptors. Thus, the maltotriose porter of Thermus thermophilus does not recognize the disaccharide maltose, while the MalEFGK2 transport system of Thermococcus litoralis accepts maltose and trehalose, but no longer dextrins. On the other hand, trehalose is not a substrate of the E. coli maltose/maltodextrin porter (Cuneo et al., 2009b). Comparison of crystal structures of these proteins in complex with the cognate substrates revealed the presence of four subsites that bind individual glucose rings (Samanta et al., 2003; Cuneo et al., 2009b).

An ABC transporter homologous to maltose/maltodextrin porters of the CUT1 family also plays a key role in the proposed ‘carbophor’ function of the pseudo-maltotetraose acarbose produced by bacteria from the genera Actinoplanes and Streptomyces (reviewed in (Wehmeier & Piepersberg, 2004). Acarbose, which is also a substrate of the E. coli maltose transporter (Brunkhorst et al., 1999), consists of an unsaturated C7 cyclitol bound via an imino bridge to 4-amino-4,6-dideoxyglucose (together named acarviosine) to which a maltose moiety is attached. According to this hypothesis, acarbose is synthesized and secreted into the environment, where, due to the action of a transferase, the acarviosyl group is hooked to maltotriose or other oligosaccharides. These are taken up by the GacHFG-MsiK ABC transporter, thereby resulting in a net gain in carbon and energy. Crystal structures and biochemical characterization of the binding protein GacH have demonstrated that longer acarbose homologs are in fact ligands of the protein, thereby corroborating this notion (Vahedi-Faridi et al., 2010).

Two CUT1 family members with strong sequence similarities to the E. coli maltose transport proteins, but distinctive substrate specificities, have been described in Streptococcus mutans, which is central to dental caries in humans. While the MsmEFGK system transports raffinose, melibiose and stachyose, the MalXFGK transporter is specific for maltodextrins. Maltose is a poor substrate for both systems. Mutations affecting the binding proteins, MsmE and MalX, respectively, were shown to cause the respective defects in sugar utilization (Webb et al., 2008). These results somewhat contradict a study by Kilic et al. (2007), who reported overlapping substrate specificities for both systems. However, mutating either one of the ATPase subunits, MsmK and MalK, had no phenotype, suggesting that the remaining NBD is shared by both systems (Webb et al., 2008).

Similar conclusions were drawn previously in the case of the MsiK protein, which assists in the uptake of several oligosaccharides through distinct transporters in Streptomyces (Hurtubise et al., 1995).

Bacteria such as Klebsiella oxytoca secrete cyclodextrin glycosyltransferases, which can cyclize linear maltodextrins first to α-cyclodextrins that are converted mainly to β-cyclodextrins. Utilization of cyclodextrins as carbon and energy source involves a cyclodextrin-ABC transport system of the CUT1 family (Fiedler et al., 1996). Crystal structures of a cyclodextrin-binding protein from Thermoactinomyces vulgaris (TvuCMBP) have been reported in complex with α-, β- and γ-cyclodextrins as well as with maltotetraose (Matsumoto et al., 2009). The structures revealed that TvuCMBP, in contrast to E. coli MalE, adopts the closed conformation with α-and β-cyclodextrins, but the open conformation with maltotetraose.

Cellooligodextrins. Extracellular degradation of cellulose results mainly in the production of cellotriose and cellobiose, which are internalized by ABC transport systems. In species of the genus Streptomyces that thrive in soil, an operon (cebEFG) was identified, encoding a solute-binding lipoprotein, CebE, and two membrane proteins: CebF and CebG (Schlösser et al., 1999). Transcription of the operon is induced by cellobiose. The gene for an NBD, as often observed for sugar ABC transporters in gram-positive bacteria, is not linked, but the MsiK protein was demonstrated to assist in cellobiose/cellotriose and maltose transport (Schlösser et al., 1997).

While the Ceb system of Streptomyces is a member of the CUT1 family, the hyperthermophilic archaea P. furiosus and S. solfataricus contain high-affinity cellobiose-uptake systems, homologous to oligopeptide transporters. This holds for quite a few sugar transporters of archaea and of the bacterium T. maritima, whereas the majority of systems belong to the CUT family (Albers et al., 2004). The P. furiosus Cbt transporter exhibits a Km for uptake of 175 nM and a Kd for solute binding of 45 nM. The binding protein, CbtA, which is anchored to the cytoplasmic membrane via an N-terminal transmembrane helix, has a broad substrate specificity, accepting cellodextrins up to five glucose units and laminaribiose, a degradation product of laminarin (1,3-β-d-glucan) (Koning et al., 2001). The crystal structure of a cellobiose-binding protein homologous to oligopeptide-binding proteins from T. maritima revealed a semi-specific recognition of the substrates. While the disaccharide cellobiose binds specifically at its nonreducing end, additional rings up to five (cellopentaose) are located in a solvent-filled groove. Interactions of the reducing end with the protein define the acceptable length of the substrate (Cuneo et al., 2009a).

The anaerobic thermophilic bacterium Clostridium thermocellum grows very efficiently on cellulose due to a multienzyme complex, the cellulosome. Five sugar ABC transporters were identified in the organism, which, by analysis of the binding properties of their cognate solute-binding lipoproteins, were demonstrated to display specificities for cellodextrins (CbpB-D), cellotriose (CbpA) and laminaribiose (Lbp), respectively (Nataf et al., 2009).

Hemicelluloses. Because of the tight association of hemicelluloses with cellulose fibrils in the plant cell wall, cellulose degraders also require hemicellulolytic activity. Hemicelluloses are a mixture of branched and linear polysaccharides. Endo-1,4-β-xylanases release short, modified oligoxylose units of two or more sugars from the polymer's backbone.

In Streptomyces thermoviolaceus, an ABC transporter, BxlEFG (a gene for an ATPase subunit is not linked), operates that takes up xylodextrins. The recombinant binding protein, BxlE, shows the highest affinity for xylobiose (Kd∼10−8 M) and xylotriose (Kd∼10−7 M). Repression of transcription by the regulator protein BxlR is relieved in the presence of xylobiose (Tsujibo et al., 2004). A similar system, XynEFG, was found in Geobacillus stearothermophilus. The purified binding protein, XynE, binds the substrates up to six xylose units with Kd values in the low μM range. Transcription of the operon is repressed by a specific regulator and activated in the presence of xylose by the regulator component of a two-component system (Shulami et al., 2007).

A putative ABC transporter proposed to mediate the uptake of methyl-α-d-glucuronosyl-xylotriose, a degradation product of side groups of xylan, was identified in Bacillus stearothermophilus T-6, but not characterized further (Shulami et al., 1999).

Pectin. The plant cell wall component pectin, the major matrix polysaccharide, consists of α-1,4-linked galacturonate residues forming the backbone of the molecule. Attached to it are neutral sugars such as rhamnose, arabinose, galactose or xylose. Pectin can be depolymerized by pectinases, leading to maceration.

In the plant pathogen Erwinia chrysanthemi, which causes soft rot disease on various plants, an ABC transporter, TogMNAB, besides others, internalizes oligogalacturonides and is involved in the chemotactic response toward these compounds (Hugouvieux-Cotte-Pattat et al., 2001). The structure of the binding protein component, TogB, of the homologous transporter from Yersinia enterocolitica was resolved and revealed selectivity for digalacturonic acid, especially for the 4,5-unsaturated form of the sugar (Abbott & Boraston, 2007).

Ramified regions of pectin contain galactan (β-1,4- or β-1,3-linked d-galactopyranose residues), which can also be utilized by E. chrysanthemi. The ganEFGK operon, encoding a transporter displaying a high similarity to the maltose transporter of E. coli, is required for growth on galactan as shown by mutational analysis (Delangle et al., 2007).

Rhamnose, a methyl-pentose sugar that is also found in the mucilage of legume plants, is taken up by the root–nodule symbiont Rhizobium leguminosarum bv. trifolii via the activity of the RhaSTPQ system, representing a binding protein, an ATPase and two membrane proteins, respectively (Richardson et al., 2004). Transport experiments revealed strict specificity for rhamnose as other pentoses, such as l-arabinose or l-fucose, failed to inhibit the uptake of radiolabeled rhamnose. The capability to internalize rhamnose is required for the bacteria to successfully compete for nodule occupancy. As an interesting yet to be unraveled feature, the transport activity is dependent on an active rhamnose kinase (Richardson & Oresnik, 2007). The rhamnose ABC transporter, like those for other pentoses found in numerous bacteria, for example arabinose or xylose, is a member of the CUT2 family (Schneider, 2001).

Chitin. Chitin, the major cell wall component of fungi, is a polymer of N-acetyl-d-glucosamine. Soil bacteria of the genus Streptomyces secrete chitinases and internalize the chitooligomers produced, chitobiose (N, N′-diacetylchitobiose) and N-acetylglucosamine (NAG), which can be utilized as carbon and energy sources.

The Ngc transporter, comprising the solute-binding lipoprotein, NgcE, and two membrane-integral subunits, NgcF and NgcG, of Streptomyces olivaceoviridis, is the only known uptake system for NAG besides a phosphotransferase system. NgcE specifically binds NAG and chitobiose (Saito & Schrempf, 2004). Analogous ABC transporters were found by bioinformatics in Silicibacter and Rhizobiales (Yang et al., 2006). In addition, Streptomyces contain a gene cluster, dasABC, that encodes a transporter for the uptake of chitobiose, which also plays a role in cell differentiation (Saito et al., 2007). The binding protein, DasA, exhibits the highest affinity for chitobiose (32 nM), but also accepts chitooligomers up to five NAG units with a reasonable affinity. Low-affinity binding was found for NAG (25 μM). The transporter is completed by the MsiK-ATPase as demonstrated by mutational analysis (Saito et al., 2008).

Chitin is also produced in massive amounts in marine environments by crustaceans. Gene clusters for chitin utilization including genes that encode an ABC importer were found in Vibrio chlolerae and Vibrio furnissii (Li & Roseman, 2004).

Alginate. Sphingomonas sp. A1 can grow on alginate, a linear polymer composed of α-l-guluronate β-d-mannuronate (polymerization grade >100), and produced by brown seaweed and certain bacteria. Strikingly, and unlike other biopolymers, alginate is not degraded by secreted hydrolases before internalization. This is achieved by the formation of a large pit on the cell surface, which facilitates the transport of alginate to the periplasm, where it is captured by two binding proteins, AlgQ1 and AlgQ2, and delivered to an ABC transporter, consisting of two membrane-integral subunits, AlgM1 and AlgM2, and an ATPase, AlgS. Sequence analysis of the latter groups the transporter in the CUT1 family (Schneider, 2001). Both receptors bind alginate with Kd values around 0.1 μM. X-ray structures of AlgQ1 and AlgQ2, in their open form and in complex with an alginate tetrasaccharide, have been resolved and revealed a larger cleft in between the N- and the C-terminal lobes as observed for other solute receptors. Many positively charged residues in the cleft enable both proteins to bind preferentially to alginate (reviewed in Murata et al., 2008, original references therein).

Other sugars. In bacteria, glucose is often internalized as glucose-6-phosphate via the PTS. In the hyperthermoacidophile S. solfataricus, belonging to the phylum Archaea that apparently lack PTS, a glucose ABC transporter was identified. The SBP binds glucose with a Kd of 0.4 μM at low pH (Albers et al., 2004). Crystal structures of the NBD, GlcV, with bound nucleotide and in the ligand-free state, were solved (Verdon et al., 2003).

A transporter specific for maltose and trehalose (disaccharide of α-1,1-linked glucose residues, originating from plants and fungi) that also accepts sucrose and palatinose (Silva et al., 2005) was characterized in the archaeon T. litoralis (Xavier et al., 1996). The structure of the cognate-binding protein complexed with trehalose is known (Diez et al., 2001) and the biochemical properties of the complete transporter (Greller et al., 2001) and the isolated NBD (MalK) (Greller et al., 1999) were investigated. The crystal structure of MalK was resolved, but the conformation of the dimer is likely to be an artifact because it differs from almost all other solved NBD dimers (Diederichs et al., 2000).

A transporter, AglEFGK, allowing growth on a variety of α-glucosides, including sucrose, maltose and trehalose, was identified in the root–nodule symbiont Sinorhizobium meliloti (Willis & Walker, 1999). The sucrose isomer palatinose (6-O-α-d-glucopyranosyl-d-fructofuranose) is produced from sucrose by some bacteria as a means of carbon storage. The plant tumorigenic bacterium Agrobacterium tumefaciens utilizes palatinose as a carbon source, which is internalized by the PalEFGK transporter displaying homology to the putative maltose/trehalose transporter of S. meliloti. Trehalose, which is an osmoprotectant in S. meliloti, is not transported by the Pal system as demonstrated by analysis of a palK mutant. In addition, the mutation only affected growth on palatinose, but not on glucose, maltose, sucrose or galactose (de Costa et al., 2003).

Besides ABC transporters for the uptake of glucose, galactose and xylose, the gram-negative bacterium Agrobacterium radiobacter also possesses a binding protein-dependent transport system for lactose, LacEFGK, which is usually transported by pmf-driven porters or PTS (Greenwood et al., 1990). The system displays considerable sequence similarity to the MalEFGK transporter of E. coli, which is underscored by the observation that the ATPase subunit LacK can replace MalK in maltose transport (Wilken et al., 1997).

Ribose, a degradation product of nucleosides, is transported in E. coli by the RbsBCA system, the best-characterized member of the CUT2 family (Schneider, 2001).

The hyperthermophilic bacterium T. maritima has multiple archaeal homologs of ABC transporters from the PepT subfamily characterized as sugar transporters of various specificities (Nanavati et al., 2006). Among 15 SBP components of these ABC transporters tested for sugar binding using fluorescence spectroscopy, 11 were characterized as sugar transporters with their own specific profiles of substrate specificities. Among the substrates bound by these T. maritima SBP proteins were several monosaccharides (xylose, ribose), disaccharides (cellobiose, laminaribiose, xylobiose, mannobiose, maltose, trehalose), various oligosaccharides, as well as myo-inositol and α-1,4-digalacturonic acid.

Polyols. Sugar alcohols (polyols) are utilized by a variety of bacteria and often internalized by PTS. The existence of an ABC transporter with a specificity for d-mannitol and d-glucitol (sorbitol) and homology to the maltose ABC transporter was deduced from sequence analysis in the purple, nonsulfur bacterium Rhodobacter sphaeroides Si4 (Stein et al., 1997). An operon, mtlEFGK, encoding genes for a transporter with a similar specificity (mannitol, araitol, glucitol) was identified in Pseudomonas fluorescens (Brünker et al., 1998).

Erythritol is a likely carbon source of root–nodule bacteria from the species R. leguminosarum. An operon encoding an ABC transporter, EryABCD, was identified on a plasmid of R. leguminosarum bv. viciae, which is induced by erythritol. A mutant lacking an intact eryA gene that encodes the putative binding protein was impaired in its capability to compete for nodulation against the wild type (Yost et al., 2006). Orthologs were found in the genomes of Brucella sp..

γ-Hexachlorocyclohexane (γ-HCH). Sphingomonas japonicum strain UT26 can grow aerobically on γ-HCH, an insecticide, as the sole source of carbon and energy. The compound is degraded to β-ketoadipate (Lal et al., 2010). Analysis of mutants identified a gene cluster, linKLMN, encoding a putative ABC transporter to be involved in γ-HCH metabolism (Endo et al., 2007). While LinL is a typical nucleotide-binding protein, LinK has features of a membrane-integral subunit, but lacks the EAA motif of canonical ABC importers. LinM has an N-terminal periplasmic signal peptide and is similar to Mce proteins of Mycobacteria (see Cholesterol) and might thus be the SBP of the putative transporter. LinN has the features of a lipoprotein. Mutational analysis suggested to the authors that the LinKLMN transporter might contribute to a controlled access of the hydrophobic and thus a potentially toxic substrate to the membrane by slowing down γ-HCH diffusion into the cell (Endo et al., 2007).

Phthalate. The capability of bacteria from the genus Burkholderia to utilize phthalate as the sole source of carbon and energy by degradation via protocatechuate is well documented (Chang & Zylstra, 1998 and references therein). Phthalates, or phthalate esters, are mainly used in the chemical industry as plasticizers to soften polyvinyl chloride. Besides a secondary transporter, uptake is achieved by the action of an ABC transporter, OphFGH. The encoding genes are part of an operon, together with the gene for a specific outer membrane porin, OphP (Chang et al., 2009). The putative SBP, OphF, does not show high levels of similarity to other SBPs in the database. Phthalate uptake is strongly inhibited by 4-chlorophthalate and salicylate, suggesting that both compounds are accepted as substrates.

Cholesterol. In actinobacteria, including Mycobacterium tuberculosis and Rhodococcus jostii, evidence for a novel type of ABC transporter for the uptake of cholesterol that is used as a carbon source was provided. Mycobacterium tuberculosis, which is the causative agent of tuberculosis, requires host-derived cholesterol for persistence (Pandey & Sassetti, 2008). Genes encoding two ABC-type membrane-integral proteins are part of an operon (mce) containing between nine and 13 genes of an otherwise unknown function. The Mce1A protein is considered a candidate for a solute-binding lipoprotein of the system (Sutcliffe & Harrington, 2004). In fact, M. tuberculosis contains four mce operons and it was proposed (Pandey & Sassetti, 2008) that each putative ABC transporter is energized by an interaction with a genetically unlinked common nucleotide-binding protein of the Mkl family (Davidson et al., 2008). For the Mce4 system of R. jostii, cholesterol uptake was verified by transport assays (Mohn et al., 2008). Related operons were also identified by bioinformatics in gram-negative bacteria. Because mutations were found to affect cell envelope integrity, it was proposed that the substrates might be organic acid precursors of cell envelope biogenesis (Casali & Riley, 2007). How this proposal correlates with the experimental findings for cholesterol transport as cited above remains to be elucidated.

Bicarbonate. In the cyanobacterium Synechococcus sp., a gene cluster encoding a bicarbonate ABC transporter was identified by mutational analysis and uptake experiments that is activated under CO2-limiting growth conditions (Omata et al., 1999). The transporter comprises a solute-binding lipoprotein, CmpA, a membrane-integral subunit, CmpB, and two ATPase subunits, CmpC and CmpD. Binding experiments with recombinant CmpA demonstrated that HCO3 (Kd=0.5 μM) rather than CO2 is the substrate (Maeda et al., 2000). The proteins are homologous to the nitrate ABC transporter (NrtABCD) of Synechocystis sp. (Koropatkin et al., 2006; see Nitrogen sources). This is underscored in the crystal structure of CmpA, which was resolved in complex with bicarbonate and carbonic acid and in the absence of ligands. Bicarbonate was found to bind in a nearly identical position as nitrate in NrtA. Bicarbonate binding is accompanied by a Ca2+ ion that might act as a cofactor or as a cosubstrate in bicarbonate transport (Koropatkin et al., 2007b). Moreover, the ATPase subunit CmpC, like NrtC of the nitrate transporter, contains a C-terminal solute-binding domain involved in the regulation of the transporter's activity (see Nitrogen sources). Interestingly, the C-terminal domain is about 50% similar to NtrA, the SBP of the nitrate transporter, and it was proposed that nitrate and not bicarbonate is the likely substrate (Koropatkin et al., 2006).

Carbon and nitrogen sources

Peptides. Peptides can serve as sources of nitrogen or amino acids in auxotrophs such as lactic acid bacteria. Thus, ABC transporters mediating the uptake of peptides play an important role in the nutrition of these organisms. In addition, peptide transporters play crucial roles in signaling processes and in virulence (reviewed in Detmers et al., 2001; Doeven et al., 2005), which will be discussed in Oligopeptide transporters. Peptide transporters specific for di- and tripeptides (Dpp) or oligopeptides containing five and more residues) (Opp) are grouped within the PepT family. Generally, like the well-characterized systems of S. Typhimurium and L. lactis, oligopeptide transporters consist of an SBP, OppA, and a heterodimer each of the membrane-integral subunits, OppB and OppC, and the ATP-binding subunits, OppD and OppF (Detmers et al., 2001). The oligopeptide-binding proteins determine the selectivity of the system as was shown in an in vitro study using purified and membrane-reconstituted Opp of L. lactis (Doeven et al., 2004). Structures of DppA from E. coli and three OppA proteins from different organisms revealed that the specificity for peptides is determined by hydrogen bonds with the ligand backbone. The side chains are located in pockets that can accommodate any side chain, which is in line with the observed lack of sequence specificity for the ligands (reviewed in Doeven et al., 2005). Most OppA proteins bind peptides with two to seven residues. By contrast, the OppA protein of L. lactis MG 1363 handles peptides up to 35 residues (Doeven et al., 2005). Crystal structures of OppA in the open and closed (liganded) conformations provided a clue for this unusual property (Berntsson et al., 2009). The protein has an enlarged substrate-binding cavity due to the movement of two loops to the surface, which, in other OppA orthologs, confine the binding cleft. Analysis of peptides bound to OppA revealed a preference for peptides between nine and 17 residues, enriched in prolines. Crystal structures in complex with peptides revealed a hydrophobic pocket that accommodates one of the peptide's side chains, which was often an isoleucine. These findings coincide with the organism's requirement for branched-chain amino acids and its preference for proline-rich casein.

The diversity of oligopeptide ABC transporters in L. lactis and other lactic acid bacteria is reflected by the observation that most strains contain more than one copy of an Opp system and in some organisms more than one peptide-binding protein is associated with a given transporter (Lamarque et al., 2004; Doeven et al., 2005). Likewise, in the genome of Borrelia burgdorferi, the causative agent of Lyme disease, which is deficient in genes for the biosynthesis of amino acids, a single Opp transporter is encoded that might interact with five different peptide-binding proteins, OppA1–OppA5 (Wang et al., 2004). The expression of the encoding genes seems to be controlled by different transcription factors (Medrano et al., 2007). Similarly, four Opp systems were identified by in silico analysis in S. aureus and shown to be expressed differently (Hiron et al., 2007). However, only one system (Opp3) was required for the growth of S. aureus in a medium deficient in glutamate/glutamine, but supplemented with glutamate/glutamine-containing peptides (4-mers to 8-mers).

The OppBCDE transporter of gram-negative bacteria is involved in recycling of cell wall components that are delivered by the specific peptide-binding protein, MppA. The regular receptor, OppA, is not involved. Amidases release murein peptides, such as l-alanyl-γ-d-glutamyl-meso-diaminopimelate, into the periplasm, from which they diffuse out of the cell, or may enter the cytoplasm via the MppA-OppBCDF transporter (Park et al., 1998). MppA, which exhibits sequence similarity to OppA, was also reported to transport heme into E. coli cells when combined with the dipeptide ABC transporter, DppBCDF. However, the cognate-binding protein, DppA, could replace MppA (Létofféet al., 2006).

Amino acids. Prokaryotes can utilize amino acids as carbon and/or nitrogen sources or they are required as precursors under auxotrophic growth conditions. Transport systems with overlapping specificities belonging to the pmf-driven major facilitator superfamily and to the ABC transporter superfamily coexist in many organisms. ABC transporters mediating the uptake of amino acids are grouped into three families: the polar amino acid transporters (PAAT), the hydrophobic amino acid transporters (HAAT) and the methionine porters (MUT; discussed under Sources of sulfur) (Hosie & Poole, 2001; Zhang et al., 2003). The histidine-lysine/arginine/ornithine transporter of S. Typhimurium is the prototype of the PAAT family (reviewed in Hosie & Poole, 2001; Schneider, 2003). Genetic, molecular biological and biochemical studies on the system by G. Ames and colleagues contributed substantially to the current knowledge on ABC transporters. The transporter consists of two SBPs, HisJ and LAO (product of the argT gene), two transmembrane subunits, HisQ and HisM, each spanning the membrane only five times, and a homodimer of the ATPase subunit, HisP. HisJ and LAO exhibit high affinities in the nanomolar range for histidine and lysine/arginine/ornithine, respectively, but each binds the preferred substrate(s) of the other with about 10-fold lower affinities. The proteins are 70% identical and may have thus evolved by gene duplication. The crystal structures of HisJ and LAO in complex with their substrates revealed that the ligand preference of both proteins might be the result of a single amino acid exchange (Oh et al., 1994a, b). HisP was the first NBD of any ABC transporter whose structure was resolved, showing the typical nucleotide-binding (RecA-like) and α-helical subdomains (Hung et al., 1998). The reported HisP dimer structure, however, has the NBDs in a different orientation than the vast majority of all other NBD structures and is most likely an artifact. The complete transporter was purified and intensively characterized in proteoliposomes (Ames et al., 2001).

Other PAAT family members closely related to the histidine transporter that have been partially characterized include the E. coli glutamine (GlnH-PQ) and arginine-specific (ArtIJ-QMP) transporters (reviewed in Hosie & Poole, 2001) and the arginine/lysine/ornithine/histidine transporter (ArtJ-MP) of the thermophile G. stearothermophilus DSMZ 13240 (Fleischer et al., 2005; Vahedi-Faridi et al., 2008). The latter is homologous to the YqiXYZ porter identified in B. subtilis (Sekowska et al., 2001). Crystal structures of ArtP of G. stearothermophilus, which is 52% identical to HisP, in complex with nucleotides are known and show a dimer organization consistent with that of most other NBDs (PDB code 2OUK, 2OLK, 2OLJ, 2QOH, 3C4J and 3C41).

Two other close relatives of the histidine transporter, the OccT-QMP and NocT-QMP systems, are worth mentioning as they mediate the uptake of the arginine-derived modified amino acids octopine and nopaline as nutritional sources in the plant pathogen A. tumefaciens. Agrobacterium tumefaciens causes the formation of a tumorous growth on a wide variety of dicotyledonous plants and elicits the synthesis of a number of modified amino acids called opines by the plant. Both transporters were studied by uptake experiments with intact cells only (Zanker et al., 1992). Transporters for other opines are members of the PepT and POPT families, respectively (Table 1).

ABC transporters mediating the uptake of acidic amino acids were characterized from several bacteria. In Corynebacterium glutamicum, a high-affinity glutamate transporter, GluBCDA, was identified by mutational analysis and transport assays with intact cells. Whether the transporter is specific for glutamate or also accepts aspartate was not investigated (Kronemeyer et al., 1995). The AatJMQP transporter of Pseudomonas putida was shown to transport glutamate and aspartate. The purified SBP, AatJ, binds glutamate and asparagine with an equally high affinity (Kd values of 0.34 and 1.3 μM, respectively), while glutamine and asparagine are recognized with a much lower affinity (Singh & Rohm, 2008). In contrast, transport competition assays revealed that glutamate, aspartate, glutamine and asparagine are substrates of the BztABCD transporter from R. capsulatus (Zheng & Haselkorn, 1996).

An interesting ensemble of amino acid transporters was found in the filamentous cyanobacterium Anabaena sp. that can fix molecular nitrogen in differentiated cells termed heterocysts. Basically, only three ABC transporters are responsible for amino acid uptake of the organism. While the NatFGH transporter recognizes acidic and neutral amino acids, the BgtAB system takes up basic amino acids. In addition, the NatABCDE porter belonging to the HAAT family prefers proline and hydrophobic amino acids. The BgtAB system is composed of the ATPase subunit, BgtA, and BgtB, a fusion of an N-terminal solute-binding domain and a C-terminal TMD. The bgtA gene is not linked to bgtB, but included in the gene cluster encoding the NatFGH proteins and shared by both transporters as revealed from uptake assays. Both Nat systems appear to contribute to nitrogen fixation (Picossi et al., 2005; Pernil et al., 2008).

While most of the transporters are rather specific for chemically closely related amino acids, a few systems accepting a wide range of amino acids are known that seem to be widespread. The best-characterized general amino acid transporter is the AapJQMP porter of the root symbiont R. leguminosarum, which prefers basic and acidic amino acids, but also transports aliphatic amino acids (reviewed in Hosie & Poole, 2001). Transport experiments with intact cells revealed that the Aap system, the branched-chain amino acid transporter, Bra, of R. leguminosarum, and the histidine transporter of S. Typhimurium mediate not only the uptake but also the export of amino acids, which might be a means to cope with the accumulation of amino acids as a consequence of metabolism. This finding challenges the concept of the unidirectional transport of ABC import systems depending on an extracellular binding protein (reviewed in Hosie & Poole, 2001). The authors assumed that substrate molecules must gain access to a (low affinity) binding site within a transporter from the cytoplasmic site that does not necessarily have to be identical to the one used for uptake. How this could be achieved was not yet further explored under in vitro conditions and whether it is only an intrinsic property of amino acid transporters is not known. The currently available crystal structures do not provide a clue in favor of this notion.

Within the HAAT family, transporters for branched-chain amino acids have been characterized from E. coli, S. Typhimurium and Pseudomonas aeruginosa (reviewed in Hosie & Poole, 2001). The LIV-1 system of E. coli consists of two SBPs, LivJ and LivK, and a membrane-bound complex comprising two membrane-integral subunits, LivH and LivM, and two ATPase subunits, LivG and LivF. Crystal structures of the LivG homolog from Methanococcus janaschii in its monomeric form have been resolved (Karpowich et al., 2001). LivK was originally reported to be specific for leucine (Kd∼1 μM), while LivJ binds leucine, isoleucine and valine with similar affinities (Kd∼0.1–1 μM) and threonine, serine and alanine with a lower affinity. In a more recent study, using an E. coli strain deficient of all known transporters for aromatic amino acids, it was demonstrated that the LIV-1 system also transports phenylalanine. Uptake studies further revealed that the transporter, when equipped with LivK, accepts isoleucine and valine similar to leucine (Koyanagi et al., 2004).

A similar modular organization is found for the Salmonella LIV-1 transporter, whereas the homologous system from P. aeruginosa (BraC-DEFG) contains only one SBP (BraC). The Bra system of P. aeruginosa is one of the few that have been characterized in proteoliposomes, thereby demonstrating that alanine and threonine are true substrates (Hoshino et al., 1992). Other members of the HAAT family display a much broader range of substrate specificity (Hosie & Poole, 2001).

Both broad-specificity amino acid transporters of R. leguminosarum (Aap and Bra) are required for effective nitrogen fixation in pea nodules (Lodwig et al., 2003). It was shown that bacteroids (differentiated forms of the bacteria within root nodules) become symbiotic auxotrophs for branched-chain amino acids, and depend on the plant for supply (Prell et al., 2009).

γ-Aminobutyric acid (GABA) was suggested as a candidate amino acid (beside glutamate) that, in root–nodule symbiosis, is donated by the plant to the bacteroid in exchange for ammonia as a result of nitrogen fixation (Lodwig et al., 2003). In mutants of R. leguminosarum bv. viciae 3841 that grow faster than the wild type on GABA as the sole source of carbon and nitrogen, an ABC transport system, GtsABCD, exhibiting specificity for GABA and related compounds was identified (White et al., 2009). The transporter, belonging to the POPT family (Table 1), consists of an SBP, GtsA, two transmembrane subunits, GtsBC, and an ATPase, GtsD. The Gts system was not expressed in pea bacteroids, thereby questioning a role in amino acid cycling between host and symbiont. Rather, the authors consider it likely that this function is achieved by the Bra system, which was also shown to transport GABA (Hosie et al., 2002).

Nucleosides. Uptake or scavenging of ribonucleosides from the environment is a means to provide cells with precursors for nucleic acid synthesis and sugars as a source of carbon and energy. In S. mutans, the predominant causative agent of dental caries, an ABC transporter for the uptake of ribonucleosides, was identified as the only member of the carbohydrate uptake family CUT2 present in this organism (Webb & Hosie, 2006). The transporter, RnsBACD, consists of an ATPase subunit, RnsA, two membrane-integral subunits, RnsCD, and a solute-binding lipoprotein, RnsB. RnsA, like RbsA and other ATPases of the CUT2 family, is a fused heterodimeric protein, containing two putative NBDs. Uptake experiments with intact cells using radiolabeled cytidine in combination with competing solutes revealed the acceptance of most ribonucleosides, whereas ribose and nucleobases were not recognized.

An ortholog of the Rns system was found in the genome of Treponema pallidum, the causative agent of syphilis (Deka et al., 2006). The putative transporter, PnrABCDE, which could not be studied in intact cells due to the inability to cultivate the organism in vitro, was characterized via its cognate-binding protein, PnrA. The protein, which is a lipoprotein despite the dual membrane system of the cell envelope, was demonstrated by isothermal titration calorimetry to bind purine nucleosides with a Kd of ∼0.1 μM. The crystal structure of PnrA complexed with inosine (likely procured from the E. coli host) was resolved and revealed a polypeptide with structural similarities to family 1-binding proteins, such as ribose- and glucose/galactose-binding proteins of E. coli. Because T. pallidum lacks the capacity for de novo synthesis of purines, the transporter might enable the organism to take up nucleic acid precursors from the human host.

Choline, glycine betaine, proline betaine. Quaternary ammonium compounds are commonly used as osmoprotectants in many bacteria and archaea (Welsh, 2000). In a few bacteria, including the root–nodule symbiont S. meliloti, osmoprotectants can be utilized at low osmolarity exclusively as carbon and nitrogen sources or in addition to osmoprotection. Choline (trimethylammonium) is a main constituent of eukaryotic plasma membranes in the form of phosphatidylcholine and is thus readily available in different environments including the soil and the rhizosphere. For its use as a nutrient, it has to be enzymatically converted to glycine betaine, from which pyruvate is subsequently formed (Smith et al., 1988). Uptake of choline in S. meliloti is achieved by an ABC transporter composed of the SBP, ChoX, the membrane-integral subunit, ChoW, and the ATP-binding subunit, ChoV. The proteins share the highest similarities to the corresponding subunits from ProU-like transporters involved in glycine betaine and proline uptake for osmoprotection (Dupont et al., 2004). However, the transporter is induced by choline, but not by high salt. The purified ChoX protein binds choline with a high affinity (Kd, 2.7 μM) and acetylcholine with a low affinity, but not glycine betaine or proline betaine. The crystal structures of ChoX, in complex with its ligands and in an unliganded, closed form, were determined (Oswald et al., 2008). Most recently, a crystal structure of ChoX in a semi-open substrate-free form was reported that was attributed to the movement of a subdomain in the N-lobe that might be present in other SBPs as well (Oswald et al., 2009).

In P. aeruginosa and Pseudomonas syringiae, the CbcXWV system mediates the uptake of choline, betaine and carnitine. The choline-binding protein, CbcX, is an ortholog of ChoX of S. meliloti. Two other binding proteins, BetX and CaiX, with specificities for betaine and carnitine, respectively, that are genetically unlinked, deliver their substrates to the CbcWV transporter. Interestingly, transport experiments revealed that CheX and BetX compete with CaiX-betaine for docking to the core transporter in their liganded forms only, thereby suggesting that productive receptor–transporter interactions are dependent on the presence of a substrate (Chen et al., 2009).

Two ABC transporters have been implicated in the transport of proline betaine (N,N-dimethylproline) in S. meliloti. The Hut transporter displays a high affinity for histidine, proline and proline betaine and is controlled at the transcriptional level by histidine, but not by salt stress. This is different from the ProU system of E. coli with which Hut shares homology, but is consistent with a role of the transporter in catabolism (see also Uptake of compatible solutes). In contrast, the Prb system, with a periplasmic-binding protein, PrbA, two membrane-integral subunits, PrbB and PrbC, and an ATP-binding subunit, PrbD, is a member of the oligopeptide transporter family (Alloing et al., 2006). prb gene expression is induced by both proline betaine and sodium chloride, thereby differing from the cho genes. Moreover, uptake experiments with intact cells revealed that besides proline betaine, glycine betaine and choline are substrates.

EDTA. Bacterium BNC1, closely related to Mesorhizobium and Agrobacterium sp., was demonstrated to degrade the metal-chelating agent EDTA for use as a nitrogen source. Uptake of EDTA is probably mediated by the EppABCD transporter, whose encoding genes are adjacent to the gene for EDTA monooxygenase. Purified SBP, EppA, binds free EDTA with a Kd of 25 nM, but not stable metal–EDTA complexes (Zhang et al., 2007).

Nitrogen sources

Nitrate, nitrite, cyanate, urea. Besides amino acids, prokaryotes can utilize a wide range of nitrogen sources that are transported by specific ABC uptake systems belonging to the NitT family.

Cyanobacteria require nitrate for growth, which is usually severely limited in aquatic environments. Organisms such as Synechococcus sp. and others are equipped with a high-affinity ABC transporter for the uptake of nitrate and nitrite. The Nrt system consists of a solute-binding lipoprotein, NrtA, localized to the outer leaflet of the inner membrane, which is an unusual feature of binding proteins from gram-negative bacteria, but typical for cyanobacteria, a membrane integral subunit, NrtB, a canonical ATPase subunit, NrtD, and a unique NBD-SBP fusion protein, NrtC (Omata, 1995; Maeda & Omata, 1997). As already mentioned (Uptake of compatible solutes), the Nrt system shows significant similarities to the CmpA-BCD transporter, mediating the uptake of bicarbonate in cyanobacteria (Omata et al., 1999). NrtA has a signal sequence containing a double arginine motif at its N-terminus, which marks it as a substrate for the Tat export system. The crystal structure of NrtA was resolved in complex with nitrate (Koropatkin et al., 2006) and revealed an α/β protein composed of two domains that encompass the binding cleft that belongs to family 2 of SBPs (Wilkinson & Verschueren, 2003). The nitrate flux through the transporter is inhibited by ammonium ions that bind to the C-terminal solute-binding domain of NrtC (Kobayashi et al., 1997).

The N-terminal domain of NtrC from Phormidium laminosum, when expressed separately, was found to bind ATP, but no ATPase activity could be detected (Nagore et al., 2003).

A transporter displaying sequence similarity to the Nrt (and Cmp) system, NasFED, was reported for the enterobacterium K. oxytoca, which can utilize nitrate and nitrite as nitrogen sources (Wu & Stewart, 1998). Here, the periplasmic-binding protein, NasF, lacks an N-terminal cysteine residue required for modification with fatty acids and thus, as expected for gram-negative bacteria, is not a lipoprotein. NasE is homologous to NrtB and NasD is about 47% identical to NrtD. No NrtC-like component was found, although nitrate uptake by the NasFED transporter was inhibited by ammonium.

In the cyanobacterium Synechococcus elongatus, an ABC transporter, previously identified as a cyanate porter, CynABD (Espie et al., 2007), was also shown to transport nitrite, in addition to the Nrt system (Maeda & Omata, 2009). CynA displays a high degree of sequence similarity to the periplasmic SBPs NrtA and CmpA. The transporter is thought to play a role in utilizing low concentrations of cyanate under nitrogen-limiting conditions and to allow those cyanobacteria that lack an Nrt system to assimilate nitrite.

Urea is a readily available nitrogen source due to its excretion by a variety of organisms in the environment. It can be metabolized by many microorganisms. An ABC transporter mediating the uptake of short-chain amides and urea (FmdDEF) was found in Methylophylus methylotrophus (Mills et al., 1998) and characterized in more detail in cyanobacteria (Valladares et al., 2002). The UrtABCDE porter of Anabaena sp. PCC 7120 consists of the periplasmic-binding protein, UrtA, two transmembrane subunits, UrtB and UrtC, and two ATP-binding subunits, UrtD and UrtE. The proteins share the highest sequence identities with the components of the branched chain amino acid transporter from P. aeruginosa. Uptake experiments with intact cells revealed a Km of about 0.11 μM. Expression of the genes was induced under nitrogen-limiting conditions. A similar system was identified in C. glutamicum, whose expression is under the control of the global nitrogen regulator of the organism, AmtR (Beckers et al., 2004).

Sources of phosphorus

Phosphate, phosphite, phosphonate, glycerophosphates. ABC transporters mediating the uptake of phosphorus-containing compounds are grouped within the PhoT and PhnT families. The high-affinity phosphate ABC transporter Pst – a member of the PhoT family – is present in many bacteria. Current knowledge on the prototype E. coli PstSCAB porter will be summarized here. The role of Pst in pathogenesis is discussed in Uptake of nutrients from the host.

The Pst system is part of the Pho regulon in E. coli, which is considered as a global regulatory network involved in phosphate metabolism (Lamarche et al., 2008). It consists of the binding protein, PstS, whose structure is known (Wang et al., 1994), the inner membrane proteins, PstA and PstC, and the ATPase subunit, PstB, for which enzymatic activity was demonstrated (Chan & Torriani, 1996) as was for the PstB homolog of M. tuberculosis (Sarin et al., 2001). The transporter was studied by mutational analysis (Webb et al., 1992), but biochemical characterization was never reported. The Pst system is involved in the regulation of the Pho regulon. This notion is based on the observation that mutations in pst genes result in the constitutive expression of the Pho regulon in many bacteria (Wanner, 1996). The current model implies that excess phosphate (>4 μM) causes the Pst transporter to form an inhibition complex with the sensor kinase PhoR, thereby preventing activation (phosphorylation) of the cognate transcriptional regulator, PhoB. The complex dissociates at limiting concentrations of phosphate, resulting in autophosphorylation of PhoR and subsequent activation of PhoB by a phosphotransfer reaction. Dephosphorylation of PhoB-P, which is return to repression under phosphate-rich conditions, is thought to occur through the phosphatase activity of PhoR in concert with PstB and PhoU. The latter is part of the pst-phoU operon (Lamarche et al., 2008). The molecular basis of the proposed protein–protein interactions including the Pst complex remains elusive. Genetic evidence for the modulation of Pst transporter activity by PhoU was recently provided (Rice et al., 2009).

ABC transporters are also involved in the utilization of alternative sources of phosphorus such as phosphite and phosphonates (reviewed in Kononova & Nesmeyanova, 2002). They are members of the PhnT family. A transport system with homology to the E. coli PhnCDE proteins was identified in Mycobacterium smegmatis as a second phosphate transporter besides several copies of the Pst system that does not recognize phosphite or phosphonates (Gebhard et al., 2006). Expression of the system is controlled by a specific repressor, PhnF, but also by the global two-component system SenX3-RegX3 (Gebhard & Cook, 2008).

The UgpBAEC transporter of E. coli allows the uptake of glycerol-3-phosphate (Schweizer & Boos, 1983) and glycerol-2-phosphate (Yang et al., 2009a) as sources of phosphate. The transporter displays strong sequence similarity to the maltose transporter and thus belongs to the CUT1 family. In fact, UgpC, the NBD of the system, was demonstrated to substitute for MalK in maltose transport (Hekstra & Tommassen, 1993).

Sources of sulfur

Prokaryotes obtain sulfur either from inorganic sulfate or from organosulfur compounds including sulfonates, sulfate esters, sulfur-containing amino acids and glutathione (Kertesz, 2001).

Sulfate/thiosulfate/taurine. Sulfate/thiosulfate ABC transporters are members of the SulT family. In E. coli, the transporter consists of two binding proteins with preferred specificities for sulfate (SuBP) and thiosulfate (CysP), respectively, two membrane components, CysT and CysW, and the ATPase, CysA. The crystal structures of SuBP of S. Typhimurium and CysA from Alicyclobacillus acidocaldarius have been resolved. In SuBP, sulfate is bound only by hydrogen bonds to neighboring amino acids and entirely dehydrated (Pflugrath & Quiocho, 1985, 1988). CysA revealed a C-terminal extension similar to those observed in the ABC proteins from CUT1 family members, but with an as yet unknown function (Scheffel et al., 2005). The cysPTWA genes, together with other cys genes required for the synthesis of cysteine, are repressed in the presence of cysteine. Expression is activated by action of the positive regulator, CysB, and the coinducer, N-acetylserine (van der Ploeg et al., 2001).

The utilization of taurine (2-aminoethanesulfonic acid) and alkanesulfates in E. coli and other bacteria requires the ABC transporters TauABC and SsuABC, belonging to the TauT family (Kertesz, 2001). Expression of the encoding genes is derepressed under conditions of sulfate or cysteine starvation (van der Ploeg et al., 2001). Analysis of deletion mutants revealed that components of both transport systems are not functionally interchangeable (Eichhorn et al., 2000).

Cysteine/cystine/methionine. In E. coli and S. Typhimurium, experiments with intact cells revealed the participation of two binding protein-dependent transport systems in the uptake of cystine. The FliY/YecS/YecC system, one of the E. coli porters, has a rather broad substrate specificity also recognizing diaminopimelic acid. None of the systems has been characterized further, which also holds for cysteine transporters (Hosie & Poole, 2001).

A cysteine-binding protein (CjaA) was purified and crystallized from Campylobacter jejuni, a food-borne pathogen and a leading cause of acute human gastroenteritis. CjaA, which is a conserved immunodominant protein in C. jejuni, was shown to specifically bind l-cysteine with a Kd of ∼0.1 μM, while the affinity for l-serine was >200-fold lower. Binding of other amino acids could not be detected (Müller et al., 2005).

In B. subtilis, two ABC transporters were identified by mutational analysis and uptake experiments with intact cells to mediate l-cystine transport (Burguière et al., 2004): TcyJKLMN and TcyABC. The former includes two SBPs, TcyJ and TcyK, exhibiting 57% identity with unknown specificities. However, besides cystine, the system accepts cystathionine, djenkolic acid and S-methyl-cysteine, while the TcyABC porter seems to be more specific for cystine. Expression of the genes encoding the TcyJKLMN system was high in the presence of methionine, but reduced in the presence of sulfate or cystine (Even et al., 2006). Expression of the TcyABC-encoding genes was low under all the conditions tested. TcyABC-like transporters are found in many gram-positive bacteria. The putative YxeMNO transporter that is most similar to the E. coli FliY/YecS/YecC is not involved in cystine uptake (Burguière et al., 2004). The first crystal structure of a cystine-binding protein, NGO372 from Neisseria gonorrhoeae, which displays 50% and 41% sequence identity, respectively, to TcyA of B. subtilis (Burguière et al., 2004) and BspA of Lactobacillus fermentum (Turner et al., 1999), was recently resolved (H. Bulut, F. Scheffel, S. Moniot, W. Saenger & E. Schneider, unpublished data).

ABC transporters that transport l- and d-methionine, formyl-l-methionine and likely organic sulfur compounds in gram-positive bacteria are clustered in the MUT family (Zhang et al., 2003). In E. coli, the long-known metD locus was shown to encode the MetNIQ transporter (Gál et al., 2002; Merlin et al., 2002) with the ATPase subunit, MetN, the membrane subunit, MetI, and the receptor, MetQ. These proteins were formerly known as ABC, YaeE and YaeC, respectively. Interestingly, the receptor, MetQ, is proposed to be a lipoprotein. The crystal structure of the Met(NI)2 complex at a 3.7 Å resolution was resolved and revealed an explanation for an allosteric regulatory mechanism that operates at the level of transport activity (Kadaba et al., 2008). Increased cellular concentrations of the transported ligand stabilize an inward-facing conformation of the transporter by binding to a site located in between the C-terminal extensions of the respective NBD dimers. As a consequence, the transporter is locked in an inactive state, thereby preventing further ligand uptake (Kadaba et al., 2008). A similar mechanism was found for the molybdate/tungstate transporter of Methanosarcina acetivorans, Mod(AB)2 (see Trace elements).

The MetNPQ transporter of B. subtilis, which is also distributed among other gram-positive bacteria, was shown to also transport methionine sulfoxide (Hullo et al., 2004). In S. mutans, the homologous AtmBDE system is involved in the uptake of l- and d-methionine, selenomethionine and homocysteine (Sperandio et al., 2007).

A methionine-binding lipoprotein (TP32) was identified from its crystal structure in T. pallidum. The protein shares significant similarity to the putative methionine-binding protein MetQ (YaeC) of E. coli and is thus suggested to be a constituent of a methionine ABC transporter (Deka et al., 2004). The crystal structure of a related protein, GNA1946, from Neisseria meningitidis, which is assumed to be a lipoprotein, also revealed l-methionine as the most likely substrate (Yang et al., 2009b).

Glutathione. Mutational analysis and transport assays revealed a novel type of ABC transporter in E. coli, mediating the uptake of glutathione that is used as the sole source of sulfur (Suzuki et al., 2005). The transporter is a member of the PepT family and consists of the solute receptor, YliB, two inner membrane components, YliC and YliD, and an ATP-binding subunit, YliA. Homologs are found in other enterobacteria.

Iron

Iron has a vital function for almost all prokaryotes due to its role in electron transfer proteins, for example cytochromes, and in some proteins, such as ribonucleotide reductase and soluble methane monooxygenase. Although iron is one of the most abundant elements on earth, its requisition is a major challenge for many prokaryotes because it is not readily available under most growth conditions. In the presence of oxygen and at neutral pH, iron forms insoluble hydroxides and most bacteria and certain fungi are forced to synthesize and secrete low-molecular-weight chelators (siderophores) that sequester Fe3+-ions with a high affinity. Only under anaerobic conditions and at pH values below 3 can ferric ions be transported without chelators (see Crosa et al., 2004 for review). Sufficient iron supply was also shown to contribute to the virulence of many pathogens. Consequently, iron transporters of pathogenic bacteria are virulence factors and are discussed in Uptake of nutrients from the host. Here, we will summarize the current knowledge on iron ABC transporters of E. coli K-12 and other nonpathogens.

Escherichia coli is equipped with three ABC transporters mediating the uptake of Fe3+-siderophores, namely, Fe3+-hydroxamate (FhuBCD), Fe3+-enterobactin (FepBCDG) and Fe3+-dicitrate (FecBCDE) (Köster, 2001). They are grouped within the FeCT family. While mutational and biochemical analyses have been performed for individual components of the transporters, an in vitro characterization of the assembled transport complexes has not been reported as yet. All three systems have in common the dependence of a specific outer membrane receptor, which, powered by the proton-motive force, translocates the substrate into the periplasm. Removal of a so-called plug, a domain of the polypeptide blocking the channel formed of β-strands (β-barrels), was suggested to be energy-consuming (Braun et al., 2004; Ferguson & Deisenhofer, 2004). The molecular basis of this process, which requires the presence of a cytoplasmic membrane protein complex comprising TonB, ExbB, and ExbD remains elusive (Postle & Larsen, 2007).

The purified Fe3+-hydroxamate-binding protein, FhuD, binds its substrates (ferric coprogen, ferric aerobactin, ferrichrome and albomycin) with dissociation constants in the low μM range (Braun et al., 2004). The crystal structures of FhuD that have been resolved in complex with several substrates place the protein into family 3 of solute receptors (Krewulak et al., 2004). The transmembrane subunit, FhuB, has about double the molecular mass of a typical TMD, each half spanning the membrane 10 times, and is proposed to represent a fusion of two TMDs with internal sequence homology. It was demonstrated that separate expression of the N- and C-terminal halves resulted in an active protein, thereby corroborating the above notion. The FhuC protein is a typical ABC ATPase (reviewed in Köster, 2001).

Fe3+-dicitrate is passing the outer membrane by the action of the FecA receptor. It is assumed that in the periplasm only Fe3+ is delivered to the binding protein, FecB, based on the finding that 10 times more radiolabeled Fe3+ is transported into the cytoplasm than citrate (Braun et al., 2004). The membrane-bound transport complex is composed of the transmembrane subunits FecB and FecC and the ATPase, FecE. The interaction of FecB with FecCD, as suggested for other ABC importers to initiate the transport cycle, was proposed based on the BtuF-CD structure to involve the formation of salt bridges. Evidence for this notion was presented recently by mutational analysis (Braun & Herrmann, 2007). The Fec system is induced by ferric citrate in the periplasm through a signaling pathway comprising the transmembrane protein FecR (Härle et al., 1995).

The first crystal structure of a binding protein in complex with a native catecholate siderophore was reported recently. The FeuA protein of B. subtilis binds its substrate, ferribacillibactin, by electrostatic interactions through two lysine and one arginine residue (basic triad) from the N- and C-terminal lobes, respectively, and via hydrogen bonds from two glutamine residues (Peuckert et al., 2009). A basic triad was also found in the structure of the enterobactin-binding protein CeuE from C. jejuni, which was crystallized in complex with an artificial substrate (Müller et al., 2006). FeuA belongs to family 3 solute receptors and is a component of the FeuABC-YusV transporter, which also imports ferrienterobactin (Ollinger et al., 2006).

Bacillus subtilis can also utilize the exogenous siderophore petrobactin, which, among others, is produced by members of the Bacillus cereus group, including the pathogen B. anthracis. Fe-petrobactin is internalized by the YclNOPQ transporter, comprising two transmembrane subunits, YclON, an ATPase, YclP and a solute-binding protein, YclQ. The crystal structure of YclQ revealed CeuE from C. jejuni as its closest relative, but also a similarity in shape to FeuA. The importance of three basic residues for interaction with the substrate was confirmed (Zawadzka et al., 2009).

Cyanobacteria have a high demand for iron to sustain photosynthesis. In Synechocystis PCC 6803, an iron transport system operates consisting of two SBPs, FutA1 and FutA2, which are 52% identical, a membrane-integral subunit, FutB, and an ATPase, FutC (Katoh et al., 2001). FutA1 and FutA2 bind iron directly without siderophores or anions, respectively, as revealed from their crystal structures. Iron is coordinated by four tyrosine and one histidine residue, as also observed in the crystal structure of the homologous FbpA protein of C. jejuni (Tom-Yew et al., 2005). Whether ferrous or ferric irons are the preferred ligands under physiological conditions is controversially discussed (Koropatkin et al., 2007a; Badarau et al., 2008). Moreover, the function of FutA1 as a periplasmic iron-binding protein has been questioned. Instead, it was suggested to act intracellularly under conditions of iron limitations. Accordingly, only FutA2, which has an export signal targeting it to the Tat secretion machinery, is proposed to deliver iron to the FutBC transporter (Badarau et al., 2008).

Vitamins

Vitamin B12. Coenzyme B12 or adenosylcobalamin (vitamin B12 is the cyanocobalamin analogue), containing a corrin ring, serves as a cofactor for enzymatic radical reactions. The uptake of vitamin B12 in E. coli is mediated by the Btu system, which is closely related to the FeCT family comprising iron-chelate ABC transporters (Köster, 2001), and for which by far the most experimental data have been accumulated. Both families share the necessity of a specific substrate receptor in the outer membrane that requires the proton-motive force established at the cytoplasmic membrane for translocation of the substrate to the periplasm (see Iron). The Btu(CD)2 transporter was the first ABC transporter whose crystal structure could be resolved (Borths et al., 2002). It consists of two copies of the transmembrane subunit, BtuC, each spanning the membrane 10 times, and of two ATPase subunits, BtuD. Characterization of the purified transporter in proteoliposomes revealed that, like in other cases, the Btu(CD)2 complex exhibits a basal ATPase activity that is significantly stimulated by the cognate-binding protein, BtuF, and is sensitive to the inhibitor vanadate (Borths et al., 2005). BtuF was found to bind tightly to the transporter under all conditions tested. BtuF belongs to family 3 of SBPs (Borths et al., 2002; Karpowich et al., 2003) (see SBPs).

Thiamine pyrophosphate (TPP) (Vitamin B1). TPP is indispensable in central metabolism as a cofactor of enzymes such as pyruvate dehydrogenase and oxo-glutarate dehydrogenase. An ABC transporter, consisting of the SBP, ThiB, a membrane-integral subunit, ThiP, and an ATPase subunit, ThiQ, mediating the uptake of thiamine and TPP was first identified in S. Typhimurium mutants defective in de novo synthesis of TPP (Webb et al., 1998). The E. coli homolog was originally designated SfuABC to reflect its similarity to the ferric ion transporter of Serratia marcescens. Uptake experiments with intact cells demonstrated that both thiamine and TPP are accepted as substrates. Expression of the encoding operon is repressed by excess thiamin. The E. coli thiamine-binding protein designated TbpA (identical to ThiB of S. Typhimurium) was shown to exhibit similarly high affinities for thiamin (Kd, 3.8 nM), thiamin monophosphate (Kd, 2.3 nM) and TPP (Kd, 7.4 nM) (Soriano et al., 2008). Its crystal structure was resolved in complex with thiamin monophosphate. Both lobes of TbpA are connected by two crossovers (Soriano et al., 2008). The protein is structurally most similar to the Fe3+-binding protein from Haemophilus influenzae, but similarity was also found to thiaminase-I, a thiamine-degrading enzyme, suggesting a common evolutionary origin. Additional aspects of transport and salvage of thiamine and thiamine-related compounds are discussed in ‘Transport of water-soluble vitamins and vitamin-derived compounds’ of this review.

Trace elements

Manganese, zinc. Manganese plays a crucial role in the water-splitting enzyme associated with photosystem II of oxygen-evolving phototrophic bacteria and is an important metal for an oxidative stress response (Horsburgh et al., 2002). Zinc is not involved in electron transfer reactions due to its single oxidation state in solution. Rather, it is a cofactor in a variety of enzymes with diverse functions such as alkaline phosphatase, RNA polymerase, zinc finger proteins and some ribosomal proteins. Both transition metals (like iron) are toxic at higher intracellular concentrations and thus metal homeostasis pathways must operate. Among others, high-affinity ABC transporters for the uptake of these metals are part of such pathways. They cluster in the MCT family of the TC system and, based on sequence alignments, their cognate extracellular SBPs constitute cluster 9 of solute receptors (Claverys, 2001). Some have been recognized as important virulence factors of pathogens and are addressed in Uptake of nutrients from the host.

In the cyanobacterium Synechocystis sp. PCC 6803, a manganese ABC transporter, MntCAB, operates with a high affinity (Km, 1–3 μM), which is induced under manganese-starvation conditions (Bartsevich & Pakrasi, 1996). The crystal structure of its SBP, MntC, was resolved with bound Mn2+ and revealed the presence of an unusual, but functionally crucial disulfide (Rukhman et al., 2005).

Manganese ABC transporters related to an oxidative stress response were found in numerous bacteria (Claverys, 2001; Horsburgh et al., 2002). The putative iron/manganese transporter of the root–nodule symbiont S. meliloti was shown to be important for the organism's response to oxidative stress. The SitABCD system for which a homolog exists in S. Typhimurium (Kehres et al., 2002) (Uptake of nutrients from the host) favors manganese over iron as a substrate. A mutant lacking the solute-binding subunit, SitA, is symbiotically defective and displays elevated sensitivity to superoxide due to decreased levels of superoxide dismutase B. The S. meliloti SodB (also called SodA in older references) can operate with iron or manganese, therefore termed ‘cambialistic’, but exhibits a higher activity with the latter. The mutant was rescued by exogenous addition of 10 μM MnSO4, but not FeSO4, although the latter was not applied under reducing conditions to maintain the Fe(II) redox state. The authors concluded that the SitABCD transporter plays a crucial role in manganese uptake (Davies & Walker, 2007), which is consistent with data obtained for the Salmonella Sit system.

Escherichia coli and others have a ZnuABC system, consisting of the extracellular binding protein, ZnuA, a membrane-integral subunit, ZnuB, and an ATPase, ZnuC. Expression of the znuABC gene cluster is regulated by zinc and a specific repressor, Zur (Hantke, 2005). Crystal structures of Ec-ZnuA are available and show that both β-/α-domains are connected by an α-helix as it is characteristic of family 3 receptors (Wilkinson & Verschueren, 2003). Three conserved histidine residues and a glutamate were found to coordinate the zinc ion (Chandra et al., 2007; Li & Jogl, 2007; Yatsunyk et al., 2008). The glutamate was replaced by water in a structure of the homologous ZnuA protein from Synechocystis (Banerjee et al., 2003). Similar to MntC, the Ec-ZnuA structures show an unusual disulfide bond in the C-terminal domain, which might be important for structural integrity or regulation of Zn2+ binding. In one structure, a second metal-binding site with an unclear function was observed (Yatsunyk et al., 2008). Ec-ZnuA binds Zn2+ with an estimated Kd of <20 nM and other divalent cations, but not manganese. Only Zn2+, Cd2+ and Cu2+ caused conformational changes in the protein thought to be required for metal delivery to the cognate transport complex (Yatsunyk et al., 2008). A highly charged and mobile loop observed in the vicinity of the binding cleft is proposed to act as a zinc chaperone to facilitate acquisition (Banerjee et al., 2003).

The crystal structures of two other proteins, PsaA from S. pneumoniae and TroA from T. pallidum, were resolved in complex with Zn2+, but the nature of the physiological substrate (zinc or manganese) is still controversially discussed (Hantke, 2005). A binding study with TroA purified from E. coli revealed that Zn2+, Mn2+ and possibly iron might be substrates, but that the transcriptional regulator of the tro operon, TroR, is likely to bind Zn2+ (Hazlett et al., 2003).

Nickel. In E. coli, Ni2+ is an essential cofactor for NiFe-hydrogenases that operate only under anaerobic conditions. To fulfill the demand for Ni2+ ions, the nikABCDE operon encoding an ABC transporter is expressed in close correlation with the hydrogenase expression levels. Positive regulation of nikABCDE is achieved by the redox sensor FNR while transcription is repressed in response to excess Ni2+ by NikR (reviewed in Eitinger & Mandrand-Berthelot, 2000). Furthermore, nikABCDE expression is also under the control of the nitrate-regulatory system (Rowe et al., 2005). The transporter consists of two heterodimers representing the transmembrane (NikBC) and nucleotide-binding subunits (NikDE), and of a SBP, NikA. Because of sequence similarities to oligopeptide transporters, it is grouped within the PepT family of the TC system. NikA was purified and a Kd for Ni2+ was estimated to be around 0.1 μM based on equilibrium dialysis and quenching of intrinsic tryptophan fluorescence, while that for Co2+ was 10-fold higher (de Pina et al., 1995). In a subsequent study, using isothermal titration calorimetry, unphysiologically high Kd values of ∼10 and 250 μM were reported for Ni2+ and Co2+, respectively (Heddle et al., 2003). The lower affinity for Ni2+ found by this group corroborates with a study using fluorescently labeled NikA (Salins et al., 2002). Although several crystal structures of NikA have thus far been reported, a clear-cut answer on how Ni2+ is coordinated in the protein remains elusive. Heddle et al. (2003) presented structures obtained in the absence and presence of Ni2+. In the latter, the ligand is not occluded in the binding cleft, but is rather accessible to the solvent, which is in contrast to other binding proteins. Also, the rotational motion between both lobes upon binding Ni2+ was less than that observed for related binding proteins. In 2005, Cherrier et al. (2005) reported on a refined structure (1.8 Å resolution) of NikA in complex with FeEDTA(H2O). These authors, on close inspection of the data deposited by Heddle et al. (2003), arrived at the conclusion that the nature of the ion in the ‘Heddle’ structure remains unclear and could very well be an iron. Among other considerations, this was based on the fact that both groups used EDTA-containing buffers for the purification of NikA. Moreover, it was speculated that Ni2+ binding by NikA might require a chelating cofactor. This notion was further corroborated by the same group by resolving the crystal structure of NikA prepared in the absence of EDTA (Cherrier et al., 2008). Here, an unidentified chelator that might be butane-1,2,4-tricarboxylate was found to contribute to the binding of a transition metal ion of unknown identity. To make the picture even more complex, NikA was also shown to bind heme under anaerobic growth conditions, although the binding site is predicted to be remote from the nickel-binding cleft (Shepherd et al., 2007).

Homologs of the NikABCDE system with an experimentally confirmed function in nickel uptake were reported for Brucella suis (Jubier-Maurin et al., 2001) and Yersinia pseudotuberculosis (Sebbane et al., 2002). Because of the similarity among the SBPs within the PepT family, it is difficult to divide its members by amino-acid sequence comparisons into subclasses with substrate specificity for either metal ions or peptides. Based on genomic colocalization with genes for nickel-dependent enzymes or the presence of upstream NikR-binding sites, a subset of the PepT family was identified as metal transporters. A function in Ni2+ uptake has been ascribed to the vast majority of the members of this subset (Rodionov et al., 2006; Zhang et al., 2009). (see Uptake of Ni2+ and Co2+ ions, additional aspects of Ni2+ and Co2+ uptake).

Copper. Methanobactin, a siderophore-like molecule with a peptidic nature, is implicated in copper uptake in methanotrophic methylotrophs. Cu2+ is an essential cofactor of particulate methane monooxygenase. The acquisition of copper ions is achieved by methanobactin. Although nothing is currently known on the nature of the transporters involved, a TonB-dependent outer-membrane transporter and an ABC transporter in the cytoplasmic membrane were predicted for internalization of the Cu–methanobactin complex (Balasubramanian & Rosenzweig, 2008).

Molybdate/tungstate/vanadate. Molybdenum serves as a cofactor in a variety of enzymes, including (1) nitrogenase of nitrogen-fixing bacteria such as cyanobacteria or root–nodule symbionts and (2) a large group of pterin-based molybdenum enzymes in all kingdoms of life that are grouped into three families represented by sulfite oxidase, xanthine oxidase and dimethyl sulfoxide reductase (DMSOR). The DMSOR family includes many more enzymes involved in electron-transport chains under anoxic conditions, such as nitrate reductase and formate dehydrogenase (see Zhang & Gladyshev, 2008; Schwarz et al., 2009 for recent reviews). Molybdenum is never directly bound by an enzyme, but always via a cofactor, the ‘FeMoco’ (nitrogenase only) or a molybdenum cofactor (Moco), whose structure is derived from a metal-binding pterin (MPT). The same holds true for tungsten, which is found as a relative of Moco (called Wco) in tungsten-containing formate dehydrogenase (a member of the DMSOR family) and Wco-containing aldehyde : ferredoxin oxidoreductase (see Andreesen & Makdessi, 2008 for a review).

The E. coli ModABC transporter is the best-characterized uptake system for the naturally predominant form of Mo, the anoxyion molybdate (Self et al., 2001). It consists of the binding protein, ModA, the transmembrane subunit, ModB, and the ATPase, ModC. ModA binds molybdate with a very high affinity in the nanomolar range and, at equimolar quantities, also tungstate. Crystal structures of ModA in complex with molybdate and tungstate are available (Hu et al., 1997). Expression of the modABC genes is tightly regulated by the repressor, ModE, and requires molybdate starvation. Interestingly, the E. coli sulfate ABC transporter mediates molybdate uptake with a low affinity while, likewise with a low affinity, the ModABC system accepts sulfate as a substrate (Kertesz, 2001; Self et al., 2001).

The crystal structure of a molybdate/tungstate transporter, ModBC, from the sulfate-reducing archaeon Archaeoglobus fulgidus in complex with its cognate-binding protein, ModA/WtpA, has been resolved (Hollenstein et al., 2007). In contrast to ModA proteins from E. coli and Azotobacter vinelandii, which bind molybdate and tungstate as a tetrahedral complex, the crystal structures of ModA/WtpA of A. fulgidus and four other archaebacterial molybdate/tungstate-binding proteins revealed an octahedrally coordinated central metal ion (Hollenstein et al., 2009).

The crystal structure of a homolog from M. acetivorans, ModBC, revealed a C-terminal extension of the ATPase subunit, ModC, which provides the basis for the mechanism of trans-inhibition of transport similar to the methionine transporter of E. coli (Gerber et al., 2008) (Sources of sulfur). These regulatory domains provide two anoxy-binding pockets for molybdate/tungstate. When occupied by a substrate, the transporter is locked in the ‘inward-facing’ conformation, resulting in an inactive ATPase activity. Transport activity i.e. energy costs, can thus be controlled by cellular levels of the substrate.

A clear preference for tungstate over molybdate is exhibited by the receptor TupA of the TupABC transporter from the amino acid-degrading bacterium Eubacterium acidaminophilum (Makdessi et al., 2001). Unpublished data obtained by isothermal titration calorimetry revealed a Kd for tungstate of 0.2–1 nM, at least 1000-fold lower than that for molybdate. The crystal structure of TupA in the unliganded form was resolved (reported in Andreesen & Makdessi, 2008).

The genome of C. jejuni, which causes gastrointestinal infections, contains two distinct ABC transport systems (Cj0300–0303 and Cj1538–1540) with similarity to molybdate transporters and the tungstate transporter of E. acidaminophilum, respectively (Smart et al., 2009). Binding studies with the purified SBPs using isothermal titration calorimetry and intrinsic tryptophan fluorescence revealed equal affinities of Cj0303 (ModA) for molybdate and tungstate (Kd∼4–8 nM), but a clear preference of Cj1540 (TupA) for tungstate over molybdate (5 × 104-fold). Tungstate is bound with an unusually low Kd of 1 pM. By mutational analysis, it was demonstrated that a tupA strain displays a significant reduction in formate dehydrogenase activity, suggesting a role of tungstate as a cofactor of this enzyme (Smart et al., 2009). These data were corroborated by a study of Taveirne et al. (2009), who monitored the activities of different molydate- and tungstate-dependent enzymes in strains carrying mutations in the mod and/or tup genes.

Another tungstate and molybdate uptake system, WtpABC, was first identified in P. furiosus (Bevers et al., 2006). In contrast to the ModABC transporters, which are present in the vast majority of bacteria, the WtpABC transporter appears to be the predominant ABC transporter for these substrates in archaea (Zhang & Gladyshev, 2008). WtpA, the solute receptor, binds tungstate in the low picomolar range, while the dissociation constant for molybdate is about 103-fold higher (Bevers et al., 2006).

Bioinformatics identified a new subclass of putative molybdate transporters (ModA-like) in species of the hyperthermophilic archaeon Pyrobaculum. No experimental data on these systems are currently available (Zhang & Gladyshev, 2008).

A few nitrogen-fixing organisms produce an alternative nitrogenase containing FeVco (a vanadium-containing variant of the aforementioned FeMoco) when molybdenum is unavailable. The natural vanadium source vanadate is not transported by the known molybdate transport systems. A high-affinity vanadate ABC transporter, VupABC, was identified in the genome of the cyanobacterium Anabaena variabilis in the region near the genes encoding V-nitrogenase. It is not found in completely sequenced genomes from other prokaryotes in which a V-nitrogenase is known to operate. Mutational analysis revealed that a vupB mutant cannot produce V-nitrogenase activity at low concentrations of vanadate (Pratte & Thiel, 2006).

Polyamines

Polyamines, especially spermidine (NH2-(CH2)4–NH-(CH2)3-NH2) and putrescine (NH2-(CH2)4–NH2), both derived from arginine, are the major polycations in the cell and have long been associated with nucleic acid and protein biosynthesis and structure, cell growth and differentiation (Tabor & Tabor, 1985). Together with Mg2+, polyamines can bind to intracellular polyanions such as nucleic acids and ATP. Other processes involving polyamines include incorporation into the cell wall, biosynthesis of siderophores or acid resistance (Wortham et al., 2007). New results with an E. coli mutant carrying deletions of all genes involved in polyamine biosynthesis demonstrated a requirement of polyamines under strictly anaerobic conditions while no effect was observed under aerobic conditions (Chattopadhyay et al., 2009).

Two ABC transport systems mediating the uptake of spermidine and putrescine in E. coli have been studied extensively (reviewed in Igarashi et al., 2001). The PotABCD porter preferentially transports spermidine with a Km value of 0.1 μM, but also accepts putrescine with a 15 times lower affinity. In contrast, the PotFGHI system is specific for putrescine (Km=0.5 μM). The crystal structures of the respective SBPs, PotD and PotF, have been resolved in complex with spermidine and putrescine, respectively. There is evidence that spermidine uptake is inhibited by excess of intracellular spermidine through binding to the C-terminal domain of the NBD, PotA (trans-inhibition). Expression of the potABCD genes is unusually repressed by the PotD precursor through binding to two regions close to the transcription initiation site of the operon.

A polyamine-binding lipoprotein with a preference for putrescine over spermidine was identified by structural analysis and binding studies in T. pallidum. Based on these data, together with the X-ray structures of the E. coli PotD and PotF proteins, the authors proposed several key residues that determine specificity for spermidine or putrescine (Machius et al., 2007).

Maintenance of outer membrane integrity

Asymmetrical distribution of lipids in the outer membrane of gram-negative bacteria, with lipopolysaccharides in the outer leaflet and phospholipids in the inner leaflet, is thought to be crucial for the barrier function of the outer membrane (Nikaido, 2005). An ABC importer that is conserved in most gram-negative bacteria was proposed to play an important role in a newly discovered pathway for maintaining the lipid asymmetry of the E. coli outer membrane (Malinverni & Silhavy, 2009). Based on mutational analysis of the mlaFEDCB gene cluster, it was shown that deletion of mlaC or mlaB caused sensitivity of the cells against SDS/EDTA. Moreover, deletion of mlaC resulted in increased levels of palmitoylated lipid A, which reflects increased phospholipids in the outer leaflet of the outer membrane. The putative transporter consists of an SBP located at the periplasmic side of the cytoplasmic membrane via its uncleaved signal sequence, MlaD, a transmembrane subunit, MlaE, and an NBD, MlaF. A cytoplasmic protein of still unknown function, MlaB, is hypothesized to be a part of the transport complex. MlaC is a second SBP that is likely to bind phosphatidylethanolamine, as demonstrated by the crystal structure of an MlaC ortholog (PDB ID: 2QGU). However, the structure bears no resemblance to canonical SBPs. A genetically unrelated gene, mlaA, encodes an outer-membrane lipoprotein of unknown function. The authors hypothesize that the transporter mediates retrograde phospholipid transport from the outer leaflet of the outer membrane to the cytoplasmic membrane (Malinverni & Silhavy, 2009). The evidence for the transporter as a virulence factor is summarized in ‘Maintenance of cell integrity’.

Chemotaxis

Escherichia coli, S. Typhimurium and other motile bacteria respond to chemical gradients by moving toward higher concentrations of attractants and lower concentrations of repellents, a phenomenon known as chemotaxis. The signaling cascade mediating this behavior comprises chemoreceptor proteins localized to the cytoplasmic membrane that sense and transmit the presence of chemotactically active compounds. Sensing is achieved either by direct binding of the chemicals or indirectly via SBP components of ABC import systems. As a result, the chemoreceptors that are stable homodimers undergo a conformational change that is transmitted to a series of kinases, eventually leading to a change of the rotational state of the flagellar rotary motor (reviewed in Falke & Hazelbauer, 2001).

SBPs including those for maltose, ribose, glucose/galactose and dipeptides, respectively, are known to interact with the chemoreceptors Tar, Trg and Tap in case of attractant chemotaxis (Falke & Hazelbauer, 2001). The binding protein of the nickel ABC transporter of E. coli, NikA, was shown to be involved in the repellent reaction mediated by Tar (de Pina et al., 1995). Maltose-binding protein (MalE) interacts with Tar, which also senses free aspartate, by two well-defined regions located on each domain of the protein. These regions partially overlap with those involved in contacting the MalFG transmembrane subunits. Mutational analysis and modeling of the interaction between MalE and the periplasmic domain of Tar based on available crystal structures suggested a mechanism by which this interaction could trigger a conformational change similar to aspartate (Zhang et al., 1999). In case of ribose-binding protein, four distinct regions in the N- and C-terminal domains are involved in interaction with the transporter, the chemoreceptor, Trg, or both (Binnie et al., 1992; Eym et al., 1996).

Cell-to-cell communication and cell differentiation

Quorum sensing

Bacteria communicate with one another using secreted chemical signal molecules termed autoinducers in a process called ‘quorum sensing’ (Waters & Bassler, 2005; Williams et al., 2007). While a large number of gram-negative bacteria produce acyl-homoserine lactones, gram-positive bacteria synthesize oligopeptide autoinducers. Furthermore, parallel quorum-sensing circuits are known that rely on derivatives of 4,5-dihydroxy-2,3-pentadione, some of which contain boron (AI-2), as signals. AI-2 plays a universal role as a signal used for interspecies communication in many gram-negative and -positive bacteria. In S. Typhimurium and other enteric bacteria, an AI-2-specific ABC transporter, LsrABCD, operates, which displays homology to the ribose ABC transporter, RbsABC, of E. coli (Taga et al., 2001). Expression of the encoding genes that are part of the lsr operon is repressed by a specific transcriptional regulator, LsrK, which becomes inactivated by binding of phosphorylated AI-2 due to the action of a cytoplasmic kinase, LsrK (Taga et al., 2003; Xavier & Bassler, 2005). It is proposed that internalization and subsequent enzymatic inactivation of AI-2 causes a decline in the external concentration, which is indicative of a monospecies environment (Waters & Bassler, 2005).

The SBP LsrB of the gram-negative bacterium Aggregatibacter actinomycetemcomitans, a causative agent of periodontitis, was shown to differ from ribose-binding protein, RbsB, in the interaction with AI-2 from different organisms (James et al., 2006; Shao et al., 2007).

Competence and sporulation

Competence, which is the capability of cells to import DNA fragments from the environment, is a phenomenon largely distributed among gram-positive bacteria. Competence is not a constitutive trait, but is regulated on a community-wide scale by quorum sensing. In gram-positive bacteria, this process includes the secretion of short peptides (pheromones) acting either from the outside or after internalization by oligopeptide ABC transporters (Sturme et al., 2002).

In B. subtilis, the Opp transporter encoded by spo0K takes up the competence-stimulating factor (CSF), one of two peptides known to be involved in competence. CSF is a pentameric peptide (ERGMT) and has different functions depending on its intracellular concentration. At low nanomolar concentrations, it stimulates the expression of genes involved in competence whereas at higher concentrations (up to 1 μM), it inhibits genes required for competence and stimulates the expression of genes required for sporulation (reviewed in Detmers et al., 2001; Waters & Bassler, 2005). Another pentameric peptide, PhrA (ARNQT), is internalized by the Opp system in B. subtilis and acts as an effector of a transcriptional regulator in the phosphorelay signal transduction pathway that controls competence/sporulation.

In a recent study, the oligopeptide transporter called Ami in Streptococcus thermophilus was shown to be involved in the development of natural competence in a peptide-free medium (Gardan et al., 2009). The authors demonstrated that the Ami transporter controlled the expression of comX, encoding a peptide that regulated the transcription of competence genes. One out of two oligopeptide-binding proteins, AmiA3, is predominantly involved in this process.

Bacillus subtilis produces acetoin as an external carbon storage compound that is reused during the stationary phase and sporulation. Mutational analysis suggested that the products of the ytrABCDEF operon constitute an ABC transporter consisting of a solute-binding lipoprotein, YtrF, two transmembrane subunits, YtrCD, and two NBDs, YtrBE, which is required for the reuptake of acetoin (Yoshida et al., 2000).

Conjugation

Conjugation is a mechanism of cell-to-cell transfer of certain plasmids, which, in Enterococcus faecalis and related bacteria, is initiated by the secretion of hydrophobic peptides by the recipient cells and internalization of these peptides by the donor cells. The peptides, also called sex pheromones, contain six to eight amino acids and are transported into the donor cells by components of the Opp system. However, a particular binding protein, PrgZ in case of plasmid pCF10, is involved, which, in contrast to OppA, displays a high specificity for the respective peptide (Leonard et al., 1996). Another pheromone-binding protein, TraC, which shares a high sequence identity with PrgZ, delivers its peptide to an unknown ABC transporter because a strain lacking the oppD gene that encodes a nucleotide-binding subunit exhibited normal conjugative activity (Nakayama et al., 1998).

Swarming

The process of bacterial swarming is distinct from swimming as it is a multicellular event that takes place on solid supports and requires the differentiation of vegetative cells into specialized swarmer cells (Rather, 2005). For E. coli, it was shown that swarming depends on polyamines such as spermidine (Kurihara et al., 2009). A polyamine-deficient mutant carrying a deletion of the potABCD genes encoding the polyamine ABC transporter (see Polyamines) was shown to display a swarming-negative phenotype in the presence of exogenous spermidine. The result suggests a role of the transporter in the induction of swarming.

Multicellular differentiation in Streptomyces

Strains of the gram-positive, filamentous soil bacteria from the genus Streptomyces undergo complex morphological differentiation, resulting in the formation of sporulating aerial hyphae. Two basic classes of developmental mutants of Streptomyces coelicolor A3(2) are known: bald mutants (bld), which fail to produce fuzzy aerial hyphae, and white mutants (whi), which form aerial hyphae, but no gray-pigmented spores (Ohnishi et al., 2002).

The bldK locus encodes an ABC transporter (BdlK-ABCDE) with sequence similarity to oligopeptide porters Opp from S. Typhimurium and SpoOK of B. subtilis. Mutations in the bldK locus caused a bald phenotype and resistance of cells to the toxic peptide bialaphos, known to be internalized by oligopeptide transporters. It was suggested to mediate the uptake of an extracellular signal required for the formation of aerial hyphae (Nodwell et al., 1996). BdlK and other putative oligopeptide ABC transporters are upregulated by exogenous S-adenosylmethionine (SAM), and documented to activate secondary metabolism in Streptomyces (Shin et al., 2007).

An ABC transporter was demonstrated in Streptomyces griseus to be involved in the regulation of morphological differentiation in response to glucose (Seo et al., 2002). It consists of a solute-binding lipoprotein, DasA, and two membrane-spanning subunits, DasBC, which are encoded by the dasRABC cluster. A dasA mutant failed to produce aerial hyphae, thus displaying a bald phenotype. Overexpression of dasA in a wild-type strain caused a severe defect in normal development in the presence of glucose, resulting in an early ectopic septation of young substrate hyphae and subsequent sporulation. The observation that overproduction of DasA was sufficient to cause the phenotype led to the proposal that DasA, besides a proposed role in substrate transport, might interact with a two-component sensor–regulator system under certain growth conditions (Seo et al., 2002).

DasA was later identified as a component of the DasBC-MsiK ABC transporter mediating the uptake of chitobiose in S. coelicolor (Colson et al., 2008; Saito et al., 2008). MsiK is a nucleotide-binding protein that is shared with other sugar ABC transporters in Streptomyces (Schlösser et al., 1997; Saito et al., 2008) (see Carbon and energy sources). DasR, a transcriptional regulator also included in the das gene cluster, represses the expression of dasA. Colson et al. (2008) found induction of dasA expression by chitin, a readily available biopolymer, for soil-dwelling bacteria. A dasA mutant showed increased chitinolytic activity. It was proposed that DasA, in the absence of chitin, binds to the chitin sensor, ChiS, thereby preventing activation of the cognate regulator, ChiR, resulting in the chitinolytic system of the organism to be shut off. When the chitin-derived disaccharide chitobiose is available, DasA is engaged in its transport, and thus ChiS is relieved and the chitinolytic system will subsequently be turned on as a consequence of ChiR-P action. Furthermore, a role of DasABC in the control of development was suggested based on the observation that extracellular NAG arrests the cells in the vegetative growth phase due to a complex regulatory control involving the sugar-transporting PTS (Rigali et al., 2006). In contrast, development was normal in the presence of chitin, a polymer of NAG. The authors propose that DasABC-mediated uptake of chitobiose, internal hydrolysis to NAG and subsequent phosphorylation by a kinase bypass the PTS, and thus enable the latter to target developmental proteins for phosphotransfer (Colson et al., 2008).

Response to environmental stress

Uptake of compatible solutes

Prokaryotes have developed mechanisms to maintain their cell volume and turgor pressure when facing sudden osmotic changes in the water content of their environment, termed ‘osmoadaptation’, ‘osmoregulation’ or ‘osmoprotection’. Because of the lack of active transport systems for water, turgor can only be controlled by altering the concentration of osmotically active compounds in the cytoplasm, such as potassium ions and so-called ‘compatible solutes’ (Wood, 1999). The latter include certain sugars, polyols, amino acids and their derivatives, quartenary amines and others (Welsh, 2000). The uptake of K+ ions is often an immediate first response to an increase in environmental osmolarity, which is followed by synthesis or, energetically favored, internalization of compatible solutes mediated in many prokaryotes at least in part by specific ABC transporters belonging to the QAT family of the TC system (Wood, 1999).

The prototype of an ATP-driven glycine betaine uptake system is the ProVWX transporter, encoded by the proU operon of E. coli. While proline betaine is likewise a major substrate, proline and ectoine are much poorer substrates as is also reflected by the binding properties of the periplasmic receptor, ProX. The transporter is indirectly induced by high osmolality (reviewed in Csonka & Epstein, 1996). Crystal structures of ProX in complex with glycine betaine and proline betaine, respectively, have been reported (reviewed in Horn et al., 2005).

The best-characterized ABC transporter mediating the uptake of the compatible solutes glycine betaine and proline, the latter with a low affinity, as a response to high-osmolarity growth conditions is the OpuA system of L. lactis (reviewed in Poolman et al., 2004). The transporter, which was intensively studied at the level of purified proteins in proteoliposomes, is composed of an unusual transmembrane subunit to which a solute-binding domain is fused, OpuAB, and an ATP-binding subunit, OpuAA (van der Heide & Poolman, 2002). Moreover, the transporter is known to sense intracytoplasmic ionic strength through tandem cystathionine-β-synthase (CBS) domains fused to the C-terminus of OpuAA (Biemans-Oldehinkel et al., 2006b). An anionic surface of the membrane is also required. Mutational analysis identified residues from both a surface-exposed cationic region of the CBS domain as well as of its C-terminal anionic tail to be crucial for an electrostatic switching mechanism that was proposed to underlie the activation/inactivation cycle of the transporter (Mahmood et al., 2006, 2009). According to this proposal, the transporter is locked in an inactive state by interaction of the CBS domains with the membrane under ‘normal’ physiological conditions. Upon an increase in ionic strength as a consequence of an osmotic upshift, the CBS domain is repulsed from the membrane, thereby turning on the transporter's activity (Biemans-Oldehinkel et al., 2006b). It was also speculated that the CBS domains might contribute to OpuAA dimer formation (Biemans-Oldehinkel et al., 2006b) as it was shown for the C-terminal domain of MalK of the maltose transporter (Chen et al., 2003).

The gram-positive soil bacterium B. subtilis is equipped with five active transport systems involved in the uptake of compatible solutes. Among these are two ABC transporters with specificities for choline (OpuB), and proline betaine/glycine betaine (OpuA) and a third ABC transporter exhibiting a rather broad specificity range (OpuC). The OpuA system displays homology to the ProU ABC transporter of E. coli and the OpuA transporter of L. lactis, and transcription of the encoding genes that are organized in an operon is induced under high-osmolarity growth conditions. OpuA is composed of an SBP, OpuAC, a transmembrane subunit, OpuAB, and a nucleotide-binding subunit, OpuAA (reviewed in Horn et al., 2005). OpuAC binds glycine betaine with a Kd of 19 μM while the dissociation constant for proline betaine is about 10 times higher. Crystal structures of OpuAC in complex with both substrates are available, revealing a missing hydrogen bond in the structure with proline betaine as a possible explanation for the differences in affinity. Coordination of the quaternary trimethylammonium group of glycine betaine is achieved by cation-stacking involving conserved Trp residues, slightly different from the arrangement found in the E. coli homolog, ProX (Horn et al., 2006). The transporter was intensively characterized biochemically. Results on monomer/dimer formation of the purified OpuAA protein (reviewed in Horn et al., 2005) contributed to the alternating access model of ABC importers (see The ‘alternating access’-model of transport). Because OpuAA of B. subtilis also contains the C-terminal CBS domains as does OpuAA from L. lactis, their contribution to the dimerization of OppAA cannot be excluded (Horn et al., 2005; Mahmood et al., 2006).

In the gram-negative bacterium P. syringiae, a plant pathogen, an OpuC transporter was identified by uptake experiments with salt-stressed intact cells as the sole or the predominant porter involved in osmoregulation (Chen & Beattie, 2007). OpuC is more closely related to the Bacillus system than to osmoprotectant transporters from other gram-negative bacteria. It displays a high affinity for glycine betaine and a broad substrate specificity, accepting choline, acetylcholine, carnitine and proline betaine. Like nucleotide-binding subunits from other Opu systems, the OpuCA subunit contains tandem CBS domains and a C-terminal tail. However, unlike in OpuAA of L. lactis, the C-terminal tail of the P. syringae protein and those from other Pseudomonas sp. is positively charged. Thus, the mechanism by which ionic strength is sensed by the CBS domains might differ between organisms (Chen & Beattie, 2007).

Two rather than one transporter, Ota and Otb, with homology to the OpuA system of bacteria were identified in the methanogenic archaeon Methanosarcina mazei Gö1. The Ota system is composed of a substrate-binding protein, OtaC, a transmembrane subunit, OtaB, and the ATPase, OtaA (Spanheimer et al., 2008). In addition to genes encoding the A and C components, the otb operon contains two genes coding for a transmembrane subunit (Saum et al., 2009). When overproduced in an E. coli strain lacking a glycine betaine transport system, the Ota transporter exhibited high-affinity glycine betaine uptake with a Km around 10 μM and was activated only by salinity gradients (Schmidt et al., 2007). This finding is in line with the observation that M. mazei accumulates glycine betaine in response to hypersalinity. The Ota system was found by genetic analysis to be the only glycine betaine transporter of M. mazei. The function of Otb remains elusive (Saum et al., 2009).

In S. meliloti, an ABC transporter, EhuABCD, was identified that is induced by ectoine (a tetrahydropyrimidine) and hydroxyectoine, but not by glycine betaine or high osmolality (Jebbar et al., 2005). The transporter displays similarity to amino acid porters of the PAAT family. The SBP EhuB binds ectoine and hydroxyectoine with similarly high affinity (Kd values 1.6 and 0.5 μM, respectively; Hanekop et al., 2007). The crystal structures of EhuB in complex with both substrates were determined and revealed a ligand coordination by cation-stacking involving aromatic amino acid residues as has been found for ProX from E. coli and OpuAC from B. subtilis with bound glycine betaine (Hanekop et al., 2007). Mutational analysis of key residues from the binding site confirmed their importance in ligand interaction, although most residues were more tolerant to changes than observed in ProX. This result was attributed to an additional stabilization of the delocalized positive charge of the ligands by salt bridges and hydrogen bonds, which is missing in ProX.

ABC transporter involved in response to pH stress

A regulatory role of a putative sugar ABC transporter (CUT2 family), encoded by the supABCD operon, in potassium transporters of S. meliloti was suggested in the context of alkaline stress (Lin et al., 2009). A supA mutant lacking the predicted SBP is impaired in survival at pH 10, which is enhanced in the presence of high levels of potassium ions. Expression of potassium transport systems and putative cation/proton antiporters, thought to be involved in alkaline resistance, was reduced in the mutant. How the SupABCD system might regulate the potassium transporters remains elusive.

A glutamate transporter was found by blast search in the oral lactic acid bacterium S. mutans using glutamine transporter genes as a query. Accordingly, the transporter genes were termed glnQHMP, encoding two transmembrane subunits, GlnMP, an ATPase, GlnQ, and a solute-binding protein, GlnH. Competition experiments with intact cells revealed that, besides glutamate, glutamine, aspartate and cysteine might also be transported (Krastel et al., 2010). In the presence of glutamate, wild-type strain UA159 showed increased production of lactate and displayed a decreased acid-tolerance response (ATR). The latter finding was interpreted to mean that the cells might have difficulties in coping with higher ‘acid load’ due to glutamate degradation (via phosphoenol pyruvate) by the glycolytic pathway. The observation that a glnQHMP deletion mutant produced less lactate in the presence of glutamate and exhibited increased ATR was taken as evidence for this notion. Furthermore, transcriptional downregulation of the transporter genes at acidic pH is also consistent with a link between glutamate transport and ATR (Krastel et al., 2010).

ABC importers as virulence factors in pathogens

ABC importers play various roles in the survival of pathogens within mammalian hosts. Moreover, components of ABC importers are candidates for the development of vaccines and might be targets for novel antimicrobials (Garmony & Titball, 2004).

Uptake of nutrients from the host

Amino acids. The uptake of certain amino acids from the host has been implicated in bacterial virulence. Neisseria meningitidis is a colonizer of nasopharyngeal mucosa in about 10% of healthy humans, which, in a few cases, may cause meningitis and sepsis. Genome analysis identified a gene cluster encoding a putative four-component ABC transporter for the uptake of glutamate, termed GltT. A mutant strain (serogroup B) deficient in the integral membrane subunit (NMB 1965) displayed reduced glutamate transport at a low sodium concentration and was essential for growth in infected epithelial cells (Monaco et al., 2006). In a mouse model, a mutation inactivating the homologous protein (NMC1937) of a serogroup C strain caused an attenuated phenotype in mixed infections, indicating that the wild-type bacteria outcompeted the GltT-deficient cells (Colicchio et al., 2009). Similarly, a mutant with a defective ATP-binding protein (NMB1966) was also demonstrated to be crucial for virulence in vivo (Li et al., 2009). Glutamate may be required as a precursor for the synthesis of the antioxidant glutathione.

Besides glutamate, glutamine is a primary nitrogen source of S. Typhimurium. Together with glutamine synthetase, the glutamine ABC transporter GlnHPQ (>95 sequence identity with the E. coli system; see Carbon and nitrogen sources) was shown by mutational analysis to be essential for virulence of the organism. While a single mutation in glnH, encoding the periplasmic-binding protein, allowed full virulence in mice, a combination with a mutation in glnA, encoding glutamine synthetase, caused an attenuated phenotype. Moreover, the double mutant protected mice against a subsequent challenge with wild-type bacteria (Klose & Mekalanos, 1997).

In N. gonorrhoeae, two putative glutamine ABC transporters displaying homology to the Gln system were shown to be transcriptionally upregulated twofold in a pilT mutant deficient in a protein required for retraction of type IV pili that triggers responses of epithelial cells during infection. Glutamine reversed downregulation of the transporter genes by PilT. The physiological meaning of this observation is not yet understood (Friedrich et al., 2007).

In group B streptococci, which cause neonatal sepsis and meningitis, transposon mutagenesis inactivating the glnQ gene of a glutamine transporter, GlnPQ, resulted in decreased adherence to and invasion of respiratory epithelial cells in vitro and in decreased virulence in vivo (Tamura et al., 2002). GlnP is an unusual fusion protein composed of two solute-binding domains, followed by a TMD (van der Heide & Poolman, 2002), while GlnQ is the ATPase subunit. Glutamine transport was demonstrated to be strongly impaired in the glnQ mutant. The role of GlnQ in adhesion remains elusive (Tamura et al., 2002).

Disruption of genes encoding nine ABC transporters in S. pneumoniae resulted in identification of the LivJHMGF transporter as a virulence factor. The SBP was demonstrated by binding assays to accept leucine, isoleucine and valine as substrates, which is in agreement with its sequence similarity to the E. coli branched-chain amino acid transporter. A deletion strain showed reduced virulence in pneumonia and septicemia models when tested by competitive infection (Basavanna et al., 2009).

In the gram-positive bacterium Listeria monocytogenes, a food-borne pathogen, the product of the imo0135 gene was shown to contribute to virulence in a murine model of intravenous infection. The protein, termed CtaP, displays homology to oligopeptide-binding proteins, but was identified by mutational analysis to be required for high-affinity cysteine uptake. CtaP was also shown to contribute to host cell adhesion, colonization of the gastrointestinal tract of mice and acid resistance (Xayarath et al., 2009).

Phosphate. The high-affinity phosphate transporter, PstSCAB, has been implicated in the virulence of both extracellular and invasive pathogenic bacteria, such as avian pathogenic E. coli (APEC) (Lamarche et al., 2005), Proteus mirabilis (Jacobsen et al., 2008; O'May et al., 2009), Shigella flexneri (Runyen-Janecky et al., 2005) and M. tuberculosis (Peirs et al., 2005). Strains carrying mutations in the periplasmic-binding protein, PstS, but also in other transporter components are generally attenuated in virulence. The suggested roles of Pst include regulation of invasion, resistance to cationic antimicrobial peptides, colonization and biofilm formation (recently summarized in Lamarche et al., 2005).

Iron. Pathogenic bacteria have developed three mechanisms for acquiring iron from the environment involving ABC import systems: (1) secretion of siderophores to chelate iron and their subsequent uptake as it is widely found in nonpathogens, (2) capturing iron directly from storage molecules of the host and (3) uptake of iron from free heme or heme-containing proteins (Tong & Guo, 2009). Usually, the genetic equipment of pathogenic bacteria allows more than one mechanism to operate.

In gram-negative pathogens, the synthesis and secretion of several siderophores are generally observed, which require uptake into the periplasm by a specific outer-membrane receptor by a TonB-dependent mechanism. The subsequently operating ABC transporters share sequence similarities to Fhu, Fec and Fep transport systems described for E. coli (see Iron). For example, the ViuPDGC system, identified in Vibrio cholerae, the causative agent of cholera, is required for the internalization of catechol siderophores such as vibriobactin and enterobactin. The transporter consists of the periplasmic-binding protein, ViuP, two membrane-integral subunits, ViuDG, and ATPase, ViuC, which display the highest sequence identity to the Fep system. Interestingly, ViuP was demonstrated to be a lipoprotein (Wyckoff et al., 1999). In pathogenic E. coli strains derived from human isolates, a putative iron uptake system, FitABCDE, was identified by an in vivo expression technology protocol. Although expressions of the encoding genes were found to be induced by iron limitation, its precise role including the transported substrate remained unclear (Ouyang & Isaacson, 2006).

Siderophores are also produced and secreted by gram-positive pathogens. Staphylococcus aureus, an opportunistic human pathogen of surgical wounds and implanted medical devices, produces two polycarboxylate-type siderophores termed staphyloferrin A and B [the latter might be identical to staphylobactin (see Beasley et al., 2009)]. Staphyloferrin A is internalized by the HtsABC transporter as demonstrated by uptake experiments. Although the HtsABC transporter had previously been implicated in heme acquisition, a deletion strain showed no growth defect in the serum (Beasley et al., 2009). The crystal structure of the solute-binding component HtsA identified it as a member of class III receptors with similarities to other siderophore- and heme-binding proteins. The htsABC operon lacks a gene encoding an ATPase component, but results from mutational analysis suggested that FhuC of the iron-hydroxamate transporter FhuCBG also assists in staphyloferrin A uptake (Beasley et al., 2009). Staphyloferrin B (staphylobactin) was shown to require the SirABC transporter, consisting of a binding protein, SirA, and two membrane-integral subunits, SirBC, for internalization (Dale et al., 2004). Again, experimental evidence indicated that FhuC is also shared by the Sir system. Moreover, an fhu deletion strain had decreased virulence in a mouse model (Speziali et al., 2006). A third putative siderophore transporter, SstABCD, has been found in S. aureus, but the chemical nature of its substrate(s) remained elusive (Morrissey et al., 2000).

In M. tuberculosis, the causative agent of tuberculosis, the irtAB system, when mutated, decreased the ability of M. tuberculosis to survive under iron-deficient conditions in vitro as well as in human macrophages or mouse lungs (Rodriguez & Smith, 2006). The authors postulated a role of IrtAB as an importer for siderophores. This view was subsequently challenged by (Farhana et al., 2008), who identified IrtA as an exporter for carboxymycobactin, while IrtB was shown to mediate the transport of iron-loaded carboxymycobactin both in vivo and in proteoliposomes. Interestingly, IrtB is composed of a TMD that is C-terminally fused to an NBD. The protein Rv2895c was demonstrated by a pulldown assay to interact with IrtB as a ferrated siderophore-binding protein (Farhana et al., 2008). Most recent in vivo experiments by the Rodriguez lab, which also indicated an FAD-binding domain in the N-terminus of IrtA, were taken as further evidence in favor of their original proposal of IrtAB as a heterodimeric importer (Ryndak et al., 2010). Thus, the precise role of IrtA is still a matter of debate.

Some gram-negative pathogens, including H. influenzae, Neisseria sp., Mannheimia haemolytica and others, deprive iron ions from host-storage proteins such as transferrin and lactoferrin by a siderophore-independent mechanism. The system involves a specific outer-membrane receptor that binds and releases iron from proteins directly. Uptake of free iron into the cytoplasm requires transport to the periplasm by a TonB-dependent process and subsequent translocation across the cytoplasmic membrane via an ABC transporter. The latter is usually composed of a ferric iron-binding protein, FbpA, a transmembrane subunit, FbpB, and an ATPase, FbpC. When assayed in an E. coli strain deficient in siderophore synthesis, the H. influenzae FbpABC transporter displayed an apparent Km for Fe3+ uptake of 0.9 μM. The transporter prefers trivalent (Fe3+, Ga3+, Al3+) over divalent cations (Anderson et al., 2004). The crystal structure of hFbp(A), the first of its kind, revealed coordination of Fe3+ by two conserved tyrosines: one histidine and one glutamate residue. In addition, a water molecule and an anion (phosphate) are part of the binding site. An anion (carbonate) involved in ligand binding is present in human transferrin as well, suggesting a common evolutionary origin of both protein families, which also include ferric ion-binding proteins from nonpathogens, such as cyanobacteria (see Iron) (Bruns et al., 2001). The synergistic anion, together with the arrangement of the other ligand-binding residues, is thought to be responsible for the extremely high affinity for ferric ions (Bruns et al., 2001). For recombinant Fbp(A) from Neisseria, the effective binding constant for ferric ion was determined to be in the order of 1018 per mol, similar to that of transferrin (Taboy et al., 2001). A carbonate ion was also found in Mannheimia haemophila FbpA, but otherwise, the ligand is bound only by three tyrosine residues. Comparison with the unliganded open form suggested a role of carbonate in modulating the overall conformation of the protein (Shouldice et al., 2004).

Less knowledge exists on transport systems for free iron in gram-positive pathogens. Most recently, the crystal structure of a binding lipoprotein, MtsA, a component of an ABC transporter, MtsABC, of Streptococcus pyogenes, the causative agent of scarlet fever, was reported (Sun et al., 2009). The structure was refined at a 1.8 Å resolution and the bound metal was assumed to be Fe2+. The ligand-binding site includes two histidines, a glutamate and an aspartate, as it is characteristic for Fe–Mn-specific receptors (Claverys, 2001). Furthermore, a bicarbonate ion participates in the coordination of the metal. Binding studies confirmed a 10-fold preference for Fe2+ over Mn2+. Mutational analysis revealed that the glutamate residue within the binding site is dispensable (Sun et al., 2009) (see Trace elements).

Utilization of heme, a tetrapyrrole complexing a ferrous ion, and its oxidized form, hemin, as iron sources has mostly been studied in gram-negative bacteria. Hemolysins, secreted by the bacteria, lyse erythrocytes thereby releasing hemoglobin, which, then in a complex with serum haptoglobin, is delivered to the liver for degradation. Heme, after being released from hemoglobin, is sequestered by hemopexin or serum albumin. At least two mechanisms of heme acquisition that differ in their initial step are known: (1) Heme-containing proteins or free heme are bound directly to a specific outer membrane receptor or (2) extracellular hemophores bind free heme or extract heme from carrier proteins and shuttle it to hemophore-specific outer-membrane receptors (reviewed in Cescau et al., 2007). Subsequently, transfer to the periplasm is achieved by a TonB-dependent process, followed by internalization via an ABC transporter. Once in the cytoplasm, heme is bound by a specific binding protein and delivered to an oxygenase, which opens the porphyrin ring, yielding free iron (Tong & Guo, 2009).

In P. aeruginosa, the Phu system operates, which represents a well-studied model for direct heme uptake. It involves the outer-membrane heme receptor, PhuR, a periplasmic-binding protein, PhuT, and an ABC transporter, composed of an integral membrane protein, PhuU and an ATPase, PhuV. In addition, a cytoplasmic heme-binding protein, PhuS, is required for heme storage (Ochsner et al., 2000; Tong & Guo, 2009). Crystal structures of PhuT and of its homolog, ShuT, from Shigella dysenteriae, have been resolved and identified the proteins as members of class III of periplasmic-binding proteins (Ho et al., 2007). In both proteins, heme is coordinated by a tyrosine residue, although the architecture of the binding site is somewhat different between PhuT and ShuT. Structurally, PhuT/ShuT resemble BtuF more than other class III members, which might reflect ligand similarity. Apo-PhuT binds heme with a Kd of ∼1.2 nM (Tong & Guo, 2009). The ShuUV transport system was purified and reconstituted into proteoliposomes (Burkhard & Wilks, 2008). Basal ATPase activity was observed with ShuT enclosed in the lumen of the vesicles, which was about fourfold increased in the presence of heme. Moreover, transport of heme as monitored by binding to the (cytoplasmic) heme-binding protein ShuS occurred concomitantly with ATP hydrolysis, providing clear evidence for a functional heme transport system.

A (direct) heme transport system (HugBCD) from Pleisiomonas shigelloides, a gram-negative bacterium that causes diarrheal disease in humans, was recently shown to enhance recombinant hemoglobin production when overexpressed in E. coli. Such studies are aimed to establish an alternative to human blood for patients requiring blood transfusions (Villarreal et al., 2008).

Heme utilization was studied only in a few gram-positive pathogens, with the Isd system of S. aureus and the Hts (Sia) system of S. pyogenes being the best understood. In S. aureus, hemoglobin and haptoglobin–hemoglobin are first bound to specific cell surface receptors, from which heme is extracted and passed to a cell wall protein, subsequently delivering it to an ABC transporter. The latter is composed of a binding lipoprotein, IsdE, an integral membrane protein, IsdF, and an ATPase, IsdD (Tong & Guo, 2009). The crystal structure of soluble IsdE in complex with heme revealed a class III receptor, with the heme iron being uniquely coordinated by a methionine and a histidine residue. The histidine residue was shown by mutational analysis to be essential for the IsdE-mediated growth of S. aureus on heme as the sole iron source. Unlike PhuT/ShuT of gram-negative bacteria, IsdE is structurally rather distantly related to BtuF (Grigg et al., 2007).

In S. pyogenes, only two surface proteins with little sequence similarity to the Isd proteins, Shr and Shp, deliver heme to the solute-binding lipoprotein component HtsA (SiaA) of the ABC transporter HtsBC (SiaBC) (Lei et al., 2003; Nygaard et al., 2006). Characterization of HtsA by biophysical means suggested that heme is six-coordinate with methionine and histidine as axial ligands, which is similar to IsdE, but different from PhuT (Sook et al., 2008).

Manganese, zinc. It is well established that ABC transporters mediating the uptake of manganese and zinc are of importance for pathogenic bacteria during infection of host tissues in which the concentrations of these metals are generally low (Claverys, 2001).

The first systems that have been described in this respect were the AdcCBA and PsaABC of S. pneumoniae that transport Zn2+ and Mn2+, respectively (Dintilhac et al., 1997; Novak et al., 1998) (but see discussion in Trace elements). The crystal structure of the solute-binding lipoprotein PsaA revealed a novel structural type of SBPs, termed class III receptors (reviewed in Claverys, 2001) (see Solute-binding proteins). Orthologs of the Psa system were subsequently identified in other Streptococci. In S. mutans, which is associated with dental caries and human endocarditis, a putative Mn/Fe transporter, composed of a solute-binding lipoprotein, SloC, a membrane-integral subunit, SloB, and an ATPase, SloA, was shown to be required for endocarditis virulence in a rat model. Growth experiments and transport assays using intact cells revealed a preference for Mn2+ over Fe2+ and repression of the system by Mn2+ only, suggesting that the latter is the physiological substrate (Paik et al., 2003).

A similar system, MtsABC, was found in S. pyogenes (Janulczyk et al., 1999) and characterized at the level of intact cells and mutants. An mtsABC mutant displayed increased susceptibility to reactive oxygen species due to the reduced activity of manganese-dependent superoxide dismutase (Janulczyk et al., 1999).

In S. Typhimurium, the SitABCD transporter, composed of the periplasmic-binding protein, SitA, two membrane-integral subunits, SitCD, and an ATPase, SitB, is encoded on a pathogenicity island and its mutation attenuates the virulence of the bacterium. Transport assays revealed an apparent affinity constant of ∼0.1 μM and an optimal activity at a slightly alkaline pH (Kehres et al., 2002). The Sit system also transports Fe2+, but the physiological relevance of this activity is controversially discussed (Boyer et al., 2002; Kehres et al., 2002). Together with the pmf-driven manganese transporter MntH, SitABCD also contributes to resistance to oxidative stress (Boyer et al., 2002). Similarly, the Sit transporter, which is encoded on a conjugative plasmid in the APEC O78 strain χ7122, was shown to contribute to virulence in a chicken infection model using isogenic mutants (Sabri et al., 2008). In N. gonorrhoeae, the etiological agent of the sexually transmitted disease gonorrhea, the MntABC transporter was shown by mutational analysis to be required for the intracellular survival of the bacteria in a human cervical epithelial cell model and also showed reduced ability to form a biofilm. The solute-binding lipoprotein, MntC, binds Mn2+ and Zn2+ with equal dissociation constants (∼0.1 μM), similar to those described for the TroA protein of T. pallidum. The specificity for both metal ions is in line with the apparent lack of other class III receptors in the genome of N. gonorrhoeae (Lim et al., 2008).

In S. pneumoniae, a second Zn-binding protein, besides the AdcA of the AdcCBA transporter, was identified by whole-genome analysis (Loisel et al., 2008). AdcAII is not genetically included in a classical ABC operon structure and displays only ∼30% sequence identity to AdcA and PsaA. The crystal structure in complex with one Zn2+ was resolved and revealed a tetrahedral coordination geometry composed of three histidine and one glutamate residue. Together with results from binding experiments using biophysical means, a role of AdcAII as a Zn2+-binding protein was suggested. adcAII is colocalized with two pht genes encoding proteins of an as yet unknown function, which, however, induce immune protection in animal models. Pht proteins (also termed Htp) only exist in Streptococci and contain several His-triad motifs that might form potential binding sites for Zn2+ delivered by AdcAII. Thus, a role of AdcAII in Zn2+ homeostasis was postulated (Loisel et al., 2008; Linke et al., 2009).

In S. pyogenes, a homolog of AdcAII, Lbp (or Lsp) was identified as a zinc receptor that also binds to laminin, a highly glycosylated multidomain protein found in all human tissues (Linke et al., 2009). Adhesion to laminin by S. pyogenes is an initial step in infection. The crystal structure of Lbp shows one Zn2+ coordinated as in AdcAII. These results are corroborated by a study demonstrating that mutations in Lsp affecting the Zn-binding pocket cause attenuated virulence in a murine soft tissue infection model (Weston et al., 2009). Although genetically not included in an operon encoding an ABC transporter, a role of Lbp/Lsp as the binding protein component of a Zn transporter is suggested.

The zinc transporter, ZnuABC, of S. Typhimurium is required for virulence as has been demonstrated by an increase in the 50% lethal dose of a znuC mutant, defective in the ATPase subunit, when inoculated orally or intraperitoneally in mice (Campoy et al., 2002). Furthermore, it was shown that a znuA mutant, lacking the SBP, is impaired in its ability to grow in epithelial cells (Ammendola et al., 2007). Similar results were reported in a mutational study on the role of the ZnuABC transporter in a uropathogenic E. coli strain (Sabri et al., 2009). Most recently, ZinT (formerly known as YodA), which is induced under zinc starvation conditions, was shown to assist ZnuA in zinc recruitment in S. Typhimurium (Petrarca et al., 2010). ZinT is homologous to the N-terminus of ZnuA and other Zn-binding proteins, including AdcA from streptococci, and forms a binary complex with ZnuA in vitro. Together, the data suggest a role of the ZnuABC transporter in environments poor in zinc or with limited zinc availability such as the intracellular compartment of host cells.

Determinants of resistance against antimicrobial peptides and bile

Small cationic peptides with antimicrobial properties are produced by a variety of animal species, including humans. These peptides are responsible for a potent, nonspecific humoral immune response against invading microorganisms. Although differing in length and amino acid sequence, many of these peptides attack pathogens by the same postulated mode of action, that is, by destabilizing the cytoplasmic membrane and, in case of gram-negative species, also the outer membrane. Successful pathogens have evolved mechanisms to withstand the antibiotic activity of these molecules (Otto, 2009).

In S. Typhimurium, the sap operon (sensitivity to antimicrobial peptides) is required for virulence and resistance to antimicrobial peptides such as protamine and melittin (Groisman et al., 1992). The gene products, including an SBP, SapA, two membrane-integral subunits, SapBC, and two nucleotide-binding subunits, SapDF, display homology to oligopeptide ABC transporters (Parra-Lopez et al., 1993). It was proposed that host defense peptides are internalized by the transporter, followed by degradation in the cytoplasm (Groisman, 1994). In a subsequent study, Parra-Lopez et al. (1993) identified the sapG locus of S. Typhimurium, also found to be required for protamine resistance, as being identical to the E. coli TrkA protein, implicated in potassium transport.

The role of the sap operon in E. coli remains unknown. However, it was shown that the ATP-binding subunit SapD is recruited by the potassium transport systems TrkH and TrkG, thereby explaining the dependence of both systems on the proton-motive force and ATP. Mutational analysis suggested that ATP binding rather than ATP hydrolysis is sufficient to promote potassium transport activity (Harms et al., 2001). Moreover, Stumpe & Bakker (1997) demonstrated that in E. coli K-12, this Trk-related function is responsible for protamine resistance, thereby challenging the above proposal by Groisman and colleagues. According to their hypothesis, protamine causes the leakage of potassium from the cytoplasm by forming a channel, but the cells can be rescued by high potassium influx until the peptide is detoxified.

The sap operon was also studied in nontypeable H. influenzae and shown to be upregulated upon exposure of the cells to antimicrobial peptides (Mason et al., 2006). The organism is a nasopharyngeal commensal, but also an opportunistic invader of the middle ear space, where it is likely to become exposed to antimicrobial peptides (AMPs). It was further demonstrated that purified recombinant SapA binds to AMPs and that SapD is required for both resistance and potassium transport. The authors conclude that SapABCDE is likely an importer for AMPs to be degraded internally, but also emphasize a role of SapD in potassium uptake as hypothesized by Stumpe & Bakker (1997) for E. coli.

A transporter of unknown cellular function, YejABEF, was found in E. coli that is responsible for uptake and thus sensitivity against the translation inhibitor microcin C, a peptide–nucleotide antibiotic (Novikova et al., 2007). Mutations in any of the four components resulted in a microcin-resistant phenotype. The transporter with YejA being an SBP, YejBE, the transmembrane subunits and YejF, an ATPase is related to oligopeptide porters. It is also the target of regulation by the small RNA RydC at the mRNA level (Antal et al., 2005). RydC degrades the message that was found to be associated with growth defects on certain sugars.

In S. Typhimurium, the yejABEF operon was shown to be upregulated in host cells and to confer resistance to AMPs (Eswarappa et al., 2008). A yejF deletion strain exhibited higher sensitivity to protamine, melittin and polymyxin B, a phenotype that was not observed with a ÄyejA strain. The capability of the ÄyejF strain to proliferate in activated macrophages and epithelial cells was compromised and it was attenuated for survival in a murine model of typhoid fever. The authors suggest that the transporter might play a role in resistance by mediating the uptake of AMPs for subsequent internal degradation, similar to the model proposed for the Sap complex. However, the missing phenotype of the ÄyejA mutant points to a specific function of the ATPase subunit YejF, reminiscent of that of SapD for the potassium transporter TrkH in E. coli (Stumpe & Bakker, 1997).

A transporter exhibiting sequence similarity to betaine ABC transporters was found in L. monocytogenes (Sleator et al., 2005). The BilE proteins also share a modular organization with the Opu system of L. lactis, that is, besides a putative ATPase, BilA, it consists of a fusion of the transmembrane and solute-binding domains, BilEB. Although transcriptional upregulation was observed in response to salt and low-temperature stress, the authors found no evidence for osmolyte transport. Rather, the transporter seems to confer resistance to bovine bile (oxgall) as suggested from a bile-sensitive phenotype, a bilE mutant. Furthermore, wild-type L. monocytogenes cells accumulated lower levels of radiolabeled bile than the mutant, leading the authors to propose a role in bile exclusion by an unknown mechanism.

Oligopeptide transporters

An oligopeptide-binding protein (Hly-OppA) was identified in V. furnissii, thought to cause acute gastroenteritis in humans, which exhibits hemolytic activity (Wu et al., 2007). The protein was purified and found to share high sequence identity with orthologs from other Vibrio sp. and to contain the signature sequence of cluster 5 SBPs (Tam & Saier, 1993). Binding of a nonapeptide by the purified protein and Hly-OppA-dependent accumulation of fluorescent substrates in intact cells were taken as evidence for a functional oligopeptide-binding protein. Deletion of the encoding gene caused less mortality of BALB/c mice and also effected biofilm formation.

In Vibrio fluvialis, a halophilic human pathogen, an opp gene cluster was identified that appears to be involved in biofilm formation. A mutation in the oppA gene encoding the oligopeptide-binding component, resulted in the formation of much more biofilm compared with the wild type for currently unknown reasons (Lee et al., 2004).

Osmoregulation

In M. tuberculosis, mRNA levels of proXVWZ, encoding a glycine betaine transporter, are elevated in response to host interactions, such as phagocytosis. A deletion strain is impaired in the accumulation of betaine under conditions of osmotic stress, resulting in a reduced growth within human macrophages. It was further shown that the ProXVWZ transporter is the means by which M. tuberculosis acquires betaine from the host cells (Price et al., 2008).

Maintenance of cell integrity

The recently discovered Mla system, encompassing the ABC importer MlaDEF, was suggested to be involved in maintaining the lipid asymmetry in E. coli (Malinverni & Silhavy, 2009) (see Maintenance of outer membrane integrity). Orthologs are present in pathogenic bacteria, including enteroinvasive E. coli, S. flexneri and Burkholderia pseudomallei. In S. flexneri, VacJ, the ortholog of the outer-membrane protein MlaA, was found to be essential for intercellular spreading of the bacteria within the host (Suzuki et al., 1994), leading to the speculation that an intact outer membrane is a prerequisite for escape from the host double membrane (Malinverni & Silhavy, 2009).

In Haemophilus ducreyi, the causative agent of a sexually transmitted disease termed chancroid, which is most prevalent in developing countries, an ABC transporter mediating the uptake of sialic acid (N-acetyl-neuraminic acid) was identified by transposon mutagenesis (Post et al., 2005). In bacteria of the genera Haemophilus and Neisseria, sialic acid is added to the terminal galactose residue of lipooligosaccharides in their outer membrane, thereby promoting resistance to the antimicrobial activity of human serum (Severi et al., 2007). The transporter consists of an SBP, SatA, two membrane-integral subunits, SatBC, and an ATPase, SatD. Interestingly, the SatC protein is a fusion of an N-terminal TMD and a C-terminal ATPase domain, suggesting that the latter forms a heterodimer with SatD. Transport of radiolabeled sialic acid was verified for the wild type, while mutants carrying insertions in any of the four genes displayed a transport-negative phenotype (Post et al., 2005).

SBPs

Agrobacterium tumefaciens is the causative agent of crown gall tumors in a wide range of dicotyledonous plants. The bacteria transfer and integrate T-DNA, originating from their tumor-inducing (Ti) plasmid into the plant genome in response to environmental (chemical) signals produced at the wound site of a plant (Brencic & Winans, 2005). The Ti plasmid carries virulence (vir) operons, whose induction by plant-released signals requires the combined action of the VirA/VirG sensor-regulator system and the periplasmic sugar-binding protein, ChvE. It is hypothesized that sugar-loaded ChvE interacts with periplasmic regions of VirA, thereby increasing its sensitivity for plant-derived phenolic compounds. ChvE displays homology to type I binding proteins, such as the glucose/galactose- and ribose-binding protein (Peng et al., 1998) and is suggested to be associated with the putative sugar ABC transporter, GguAB (Kemner et al., 1997). GguAB displays homology to CUT2 family members, which is consistent with ChvE being a monosaccharide-binding protein, but its role in sugar-dependent growth is still a matter of debate (Kemner et al., 1997; He et al., 2009). However, the homologous system of B. suis was demonstrated by mutational analysis to be required for growth on monosaccharides (Alvarez-Martinez et al., 2001). Mutational analysis and fluorescence-based sugar-binding studies with purified wild type and mutant ChvE variants identified two overlapping sites on ChvE for interaction with VirA and the cognate transporter (He et al., 2009). ChvE is also involved in chemotaxis to the sugar it binds.

Mycoplasma hyorhinis is a swine pathogen, but has also been proposed to be a source of oncogenic transformation. Mycoplasmas lack a cell wall and survive attached to or within eukaryotic host cells. Protein p37 was found to induce invasiveness in cancer cells and identified by X-ray crystallography as a thiamine-binding protein related to TbpA of S. Typhimurium p37 (renamed CypI) is encoded in an operon together with genes for components of an ABC transporter (Dudler et al., 1988). It has been hypothesized that the CypI might sequester thiamine from the surrounding of host cells, thereby increasing the pathogen's chances for invasion (Sippel et al., 2009).

Substrate-binding lipoproteins from gram-positive pathogenic bacteria were shown to function as native ligands of Toll-like receptors (TLR) of mammals, which play a crucial role in sensing microorganisms. These include SitC, a triacylated substrate-binding protein from an iron transporter in S. aureus (Kurokawa et al., 2009). The purified protein was capable of stimulating specifically TLR2, responsible for recognizing lipopolysaccharides and lipoproteins. Two other solute-binding lipoproteins from B. subtilis and Microccocus luteus displayed the same properties, suggesting to the authors a general role as TLR ligands for the binding proteins of ABC transporters from gram-positive bacteria.

SBPs as targets for immunotherapy and for the development of vaccines

Immunogenic ABC transporter proteins, mostly SBPs and a few NBDs, have been identified as immunodominant proteins through screening with immune sera raised against whole bacterial cells (summarized in Garmony & Titball, 2004). SBPs were also studied as potential antigens for the development of vaccines. Two binding lipoprotein components (PiuA, PiaA) of iron transport systems from S. pneumoniae that are essential for virulence were shown after immunization to protect mice against a systemic challenge similar to an established control vaccine (Brown et al., 2001). In a subsequent report, the same group provided evidence for the notion that the antibodies raised against PiuA and PiaA enhanced bacterial opsonophagocytosis as analyzed with a human neutrophil cell line rather than inhibiting iron transport (Jomaa et al., 2005). Similarly, mice were protected against systemic infection with S. pneumoniae by immunization with the polyamine-binding protein PotD, which contributes to virulence as shown in two model systems (Shah & Swiatlo, 2006).

An immunochemical analysis of sera from rabbits immunized with inactivated cells of Yersinia pestis, the causative agent of plague, identified four SBPs as immunoreactive proteins: OppA, PstS, YrbD and PiuA. Only OppA was found to increase the mean time to death of mice when challenged with Y. pestis in a correlation with the level of IgGs (Tanabe et al., 2006).

ECF-type transporters

Uptake of Ni2+ and Co2+ ions

Nickel ions are essential components of various enzymes involved in nitrogen and energy metabolism and in detoxification processes. This list includes urease, [NiFe] hydrogenase, anaerobic carbon monoxide dehydrogenase, acetyl-CoA synthase/decarbonylase, [Ni] superoxide dismutase, methyl-coenzyme M reductase and glyoxalase I (see Hausinger & Zamble, 2007 for a review). Uptake of Ni2+ ions into microbial cells is mediated by primary (ABC-type) and secondary active transporters (reviewed by Eitinger & Mandrand-Berthelot, 2000; Eitinger, 2001; Eitinger et al., 2005). In many cases, synthesis of the transporters is controlled by the Ni2+-responsive transcriptional repressor NikR (Rodionov et al., 2006; Zhang et al., 2009). The NikABCDE system of E. coli, a member of the peptide/opine/nickel (PepT, TC 3.A.1.5) transporter family, is the prototype of canonical ABC-type Ni2+ transporters (Navarro et al., 1993) and has been discussed in Trace elements. It is composed of the periplasmic SBP NikA, the transmembrane subunits NikB and NikC and the NBDs NikD and NikE (Fig. 8a).

Figure 8.

 ECF, canonical ABC (and some secondary) transport systems in prokaryotes mediating the uptake of Co2+ and Ni2+ (a), biotin (b), riboflavin (c), thiamine and its metabolic precursors (d), cobalamin (e) and tryptophan (f). Components from different protein families are shown by different colors. Examples of genomic clusters of genes encoding the thiamine, cobalamin and tryptophan transport systems are shown by stretches of arrows with colors corresponding to the protein families. Candidate regulatory elements for these genes – either transcription factor-binding sites or riboswitches – are shown by colored dots.

In 1993, sequence analysis of the cobalamin synthesis gene cluster of S. Typhimurium identified the cbiMNQO genes encoding an ABC ATPase (CbiO) and three membrane proteins, but no periplasmic substrate-binding protein (Roth et al., 1993). Because of the similarity of CbiO to the NBDs of canonical ABC importers and the fact that mutations in the cbiMNQO region could be corrected by increasing the concentration of Co2+ ions in the medium, the authors suggested that the cbiMNQO genes encode a transporter for Co2+. This hypothesis was confirmed >10 years later by experimental analysis, and the CbiMNQO system may be considered the founding member of the emerging class of ECF transporters (Rodionov et al., 2006). Because cbiQ encodes the T component of the transporter, homologs are widespread in prokaryotic genomes and flanked by one or two ABC ATPase genes in most cases. This situation led to substantial confusion during annotation of genome sequences. Independent of the presence or the nature of an adjacent S-unit gene, a multitude of T components are misannotated as cobalt transporters or as components of cobalt transporters.

Homologs of CbiMNQO with a function in nickel uptake were described for Streptococcus salivarius (Chen & Burne, 2003) and Actinobacillus pleuropneumoniae (Bosse et al., 2001). In both organisms, the metal transporter is encoded adjacent to genes for the synthesis of the nickel-dependent urease and was shown to be essential for urease activity under nickel limitation. Systematic analysis of colocalization and coregulation of ‘MNQO’ genes in prokaryotic genomes and of the phylogeny of the ‘M’ proteins revealed that these systems fall into two major branches. The members of one branch are colocalized with genes for cobalamin synthesis and/or are regulated by coenzyme B12-responsive riboswitch elements, and thus are clearly implicated in Co2+ uptake. The second branch contains operons that are colocalized with genes for nickel-dependent enzymes or under control of a NikR repressor. As a consequence, those systems were predicted to function as Ni2+ transporters and were renamed ‘Nik’ (Rodionov et al., 2006). Experimental analyses of the metal specificity of the CbiMNQO systems of R. capsulatus and S. Typhimurium and the NikMNQO systems of R. capsulatus and Methanothermobacter thermoautotrophicus are in full agreement with the bioinformatic predictions (Rodionov et al., 2006; Hebbeln, 2009).

CbiM and NikM are homologous membrane proteins with seven predicted membrane-spanning domains in their mature form and a strongly conserved extracytoplasmic N-terminus containing a His residue at position 2 (Fig. 9). This His residue is essential for the activity of both CbiMNQO and NikMNQO systems (P. Hebbeln, D.A. Rodionov & T. Eitinger, unpublished data). CbiN and NikN proteins have a similar topology, with two TMDs separated by an extracytoplasmic loop, but they are unrelated at the sequence level. NikN is fused to NikM in many cases and is replaced by NikL or NikK and NikL, two other small transmembrane proteins, in a number of prokaryotes (Rodionov et al., 2006). A heterooligomeric structure of the S unit is restricted to the metal transporters among the ECF systems (Fig. 8), but the distinct functions of the ‘M’, ‘N’, ‘K’ and ‘L’ proteins are currently unknown.

Figure 9.

 Schematic view of CbiM and NikM sequences. Roughly half of the bacterial and archaeal CbiM proteins contain a predicted N-terminal signal peptide cleavage of which between Ala and Met releases the strongly conserved N-terminus with MHIME as the consensus sequence. NikM proteins have a similar N-terminus, but do not contain signal peptides. TMDs, transmembrane domains.

A number of NikO ATPases from sulfate reducers, methanogens and some other organisms contain a histidine-rich C-terminal tail that may function in metal binding. Assuming a metal-sensing and regulatory role of these tails is an attractive hypothesis. Indeed, initial characterization of the NikO from M. thermoautotrophicus pointed to such a function (M.M.P. Murphy, University of British Columbia, Vancouver, Canada, pers. commun.).

Transport of water-soluble vitamins and vitamin-derived compounds

Substrate specificities of most ECF transporters were originally predicted by bioinformatic techniques including studies of genomic colocalization, coregulation by proteinaceous transcription factors or substrate-sensing riboswitch elements, and co-occurrence. These analyses suggested prevalence among the ECF systems of transporters for water-soluble vitamins, vitamin-derived cofactors and cosubstrates, and intermediates of vitamin-salvage pathways. In many gram-positive bacteria including pathogenic Streptococcus and clostridial species, ECF systems provide a means of transport for vitamins and vitamin precursors that are essential for growth because these bacteria do not contain all of the pathways for de novo biosynthesis (Rodionov et al., 2009). For a number of transporters (biotin, riboflavin, thiamine, folate and pantothenate transporters), the predicted substrate specificity was subsequently verified by physiological and biochemical analyses. These systems as well as the predicted transporters for thiamine precursors, niacin, vitamin B6, vitamin B12-related compounds and lipoate will be discussed in the ensuing paragraphs.

Biotin

Biotin (also known as vitamin B7, vitamin B8 or vitamin H) is an essential cofactor in carboxylation and decarboxylation reactions and is thus necessary, for example for fatty acid biosynthesis. It is produced by many bacteria, certain archaea, fungi and plants but can also be taken up from environmental sources. Contrasting the situation in eukaryotes, the mechanisms behind biotin transport into prokaryotic cells have, until recently, not been understood. Active transport of biotin into E. coli K-12 cells was reported about 35 years ago (Prakash & Eisenberg, 1974), but the carrier responsible for uptake has not been identified until now. Evidence pointing to a role of BioM (A unit) and BioN (T unit), which are homologs of CbiO and CbiQ, respectively, in biotin uptake, was reported for the root-nodulating Alphaproteobacteria S. meliloti (Entcheva et al., 2002) and Rhizobium etli (Guillén-Navarro et al., 2005). Comparative genomic analyses of biotin-responsive transcriptional repression by the widespread regulator BirA and the regulators BioR in Alphaproteobacteria and BioQ in actinobacteria predicted a role for the transmembrane protein BioY in biotin uptake (Rodionov et al., 2002a; Rodionov & Gelfand, 2006; Rodionov, 2007). A six-transmembrane-domain architecture with the N-terminus inside is predicted for the majority of BioY proteins (Fig. 8b). bioMNY operons encode subclass I ECF transporters in many proteobacteria, actinobacteria and Euryarchaeota (Table 3). In many Firmicutes, Thermotogales and in archaea from the Thermococcus genus, BioY is predicted to interact with the endogenous EcfA1A2T module to form subclass II ECF biotin-transport systems (Rodionov et al., 2009). Many Alphaproteobacteria, some actinobacteria and chlamydia contain BioY homologs in the absence of any recognizable A and T unit of ECF transporters, suggesting that those BioY proteins function as transporters in a solitary state and independent of ATP hydrolysis.

As described above, the best-investigated prokaryotic biotin transporter is the BioMNY system of R. capsulatus. Analysis of the contributions of the individual subunits showed that BioY functions as the biotin-capture protein and acts as a transporter in its solitary state at high (i.e. above 1 nM) substrate concentrations. BioMN represent the ECF. A functional BioM ATPase is required for high-affinity transport at biotin concentrations in the pM range (Hebbeln et al., 2007).

bioY is clustered with biotin synthesis genes in many prokaryotes. Positional clusters of bioY with bioB are found in Alphaproteobacteria and Clostridia. BioB encodes biotin synthase, which inserts a sulfur atom into dethiobiotin in the last step of the biosynthetic path. This genomic organization points to a role of BioY in the uptake of dethiobiotin in addition to biotin.

Riboflavin

Riboflavin (vitamin B2) is the precursor of the redox cofactors FMN and FAD, which are indispensable for all organisms and involved in many metabolic reactions (Fraaije & Mattevi, 2000). The vitamin is synthesized by many microorganisms and plants, but is also taken up from the environment. The S unit RibU (termed YpaA in some cases and found mostly in Firmicutes) and the secondary active transporter PnuX (found in Actinobacteria) were predicted to mediate riboflavin uptake because the respective genes encode integral membrane proteins and are regulated by the FMN-responsive riboswitch (RFN element) in many bacteria (Gelfand et al., 1999; Kreneva et al., 2000; Vitreschak et al., 2002). With only a few exceptions, for example in Bifidobacterium sp. and in the archaeon Caldivirga maquilingensis in which ribU is located adjacent to an ecfA1A2T cluster, RibU is the S unit of subclass II ECF transporters. In silico topological analyses of individual RibU proteins predict between four and six TMDs. Experimental investigation using fusions of GFP, PhoA and LacZ to the B. subtilis RibU suggests that the protein has five TMDs and that the C-terminus faces the cytoplasm (Vogl et al., 2007). Comparing riboflavin uptake of wild-type and RibUB. subtilis strains confirmed the function of RibU as the central component of the riboflavin transporter (Vogl et al., 2007; Rodionov et al., 2009). In contrast to RibU (Km for riboflavin in the lower nM range), PnuX has a much lower affinity for its substrate (Km in the lower μM range) and does not seem to depend on an energy source (Fig. 8c) (Vogl et al., 2007).

Two lines of evidence support the conclusion that RibU proteins interact with EcfA1A2T to form subclass II systems. Disruption of ecfA1 in B. subtilis abolishes riboflavin uptake as does disruption of ribU (Rodionov et al., 2009). Coproduction of RibU and EcfA1A2T from L. mesenteroides in E. coli allows the isolation of stable complexes containing all four subunits from membranes of the recombinants (Neubauer et al., 2009; Rodionov et al., 2009).

Transport and salvage of thiamine and derived compounds

TPP, the active form of vitamin B1, is required for central metabolic reactions including the decarboxylation of 2-oxo-acids in sugar catabolism and in the citric acid cycle. TPP is produced by prokaryotes, fungi and plants in two separate branches, yielding the intermediates hydroxyethylthiazole phosphate (HET-P) and hydroxymethylpyrimidine pyrophosphate (HMP-PP), which are combined to yield thiamine monophosphate (TMP) that, in the final step, is phosphorylated to the active cofactor (Jurgenson et al., 2009). Bacteria can salvage thiamine and a number of components of TPP. In the absence of central parts of the TPP biosynthetic pathway, for example in several pathogens among the Firmicutes, salvage is essential. A canonical ABC transporter for thiamine, TMP and TPP, consisting of the NBD ThiQ, the transmembrane protein ThiP and the SBP ThiB, was identified in S. Typhimurium (Webb et al., 1998). Crystal structure analysis characterized the E. coli ThiB (also known as TbpA) as a group II periplasmic-binding protein that exhibits similarity to thiaminase I, an enzyme involved in thiamine degradation (Soriano et al., 2008).

Reconstruction of the regulons controlled by the TPP-responsive riboswitch (THI element) identified various candidate transporters involved in thiamine salvage pathways in bacteria, in addition to the ThiBPQ system, which is widespread among gram-negative bacteria (Fig. 8d) (Rodionov et al., 2002b). A predicted secondary transporter for thiamine (PnuT) was found in several lineages of proteobacteria that lack the ThiBPQ system. The subclass II ECF transporter for thiamine, ThiT (also known as YuaJ), was identified in Firmicutes that lack both ThiBPQ and PnuT, at that in some thiamine-auxotrophic species (e.g. Streptococcus pyogenes), it complements the absence of thiamine biosynthesis genes. The prediction that ThiT proteins bind thiamine and represent the S units of subclass II ECF transporters was confirmed in a series of biochemical assays with the L. casei and L. monocytogenes homologs (Eudes et al., 2008; Rodionov et al., 2009; Schauer et al., 2009).

At least two types of transporters, the canonical ABC system ThiXYZ and the subclass I ECF system YkoEDC, are implicated in recycling pyrimidine moieties of TPP that arise through base degradation of the cofactor (Fig. 8d) (Rodionov et al., 2002b). In a recently discovered pathway, base-degraded hydrolysis products of thiamine in soil such as formylaminopyrimidine are transported into the cells, deformylated by an amidohydrolase and subsequently hydrolyzed by thiaminase II (TenA) to HMP, which, as the pyrophosphate, is a building block of TPP (Jenkins et al., 2007). A canonical ABC system encoded adjacent to tenA in B. cereus and Bacillus halodurans consisting of the SBP ThiY, the integral membrane protein ThiX and the NBD ThiZ serves as the transporter for formylaminopyrimidine (Jenkins et al., 2007). The subclass I ECF system YkoEDC may play this role in other bacteria. It is composed of the S unit YkoE, the A unit YkoD containing duplicated NBDs and the T component YkoC. In E. faecalis, Streptococcus sp. and Geobacillus kaustophilus, the ykoEDC genes are colocalized with tenA. In B. subtilis and Oceanobacillus iheyensis, the ykoFEDC operon contains an additional gene. YkoF of B. subtilis is a thiamine-binding protein that is unrelated to substrate-binding proteins of canonical ABC transporters. It binds thiamine at two sites with dissociation constants of 10 and 250 μM (Devedjiev et al., 2004). Because these values are by far above physiological requirements, it is unlikely that YkoF contributes to the transport of thiamine precursors.

ThiW (S unit)-containing ECF transporters are predicted to function in the salvage of the thiazole moiety of TPP. The assumption is based on positional clustering of thiW with thiM encoding HET kinase (Rodionov et al., 2009). ThiW proteins from numerous Firmicutes and a single archaeon (Korarchaeum cryptofilum) belong to subclass II ECF systems, whereas their orthologs from Crenarchaeota and Chloroflexi are encoded by operons including genes for A and T components and are thus classified as subclass I ECF transporters.

Folate

Tetrahydrofolate is a carrier of one-carbon units and involved in many central metabolic reactions including the formylation of methionyl-tRNA, the last step of methionine synthesis, thymidylate synthesis and synthesis of purines. Tetrahydrofolate is produced in plants and microorganisms in two branches: the pterin branch starting from GTP and the 4-aminobenzoate branch starting from chorismate. The resulting pterin and 4-aminobenzoate moieties are condensed to a pteroate, which is glutamylated and reduced twice to yield tetrahydrofolate. Tetrahydrofolate may become polyglutamylated in a subsequent step. Salvage pathways include the uptake and utilization of folates and of the degradation product 4-aminobenzoyl glutamate (Carter et al., 2007; de Crecy-Lagard et al., 2007). Folate transporters similar to related systems in plants were found in cyanobacteria and may also occur in proteobacteria (Klaus et al., 2005). Uptake of folates by naturally auxotrophic organisms such as Lactobacillus sp. is known for >30 years ago (Henderson et al., 1977; Kumar et al., 1987), but only recently subclass II ECF transporters were shown to be the systems involved (Eudes et al., 2008). FolT, which is widespread among the Firmicutes, was predicted to be the S unit based on three lines of evidence: the sequence similarity of folT to sequences predicted by reverse genetics of peptides derived from the L. casei‘folate-binding protein’ (=FolT), the colocalization of folT with folC, encoding folylpolyglutamate synthase, in some genomes, and the strict co-occurrence of folT only in those genomes that encode an EcfA1A2T module. Lactobacillus casei FolT was shown to bind folate and folinate with a high affinity and to transport folate into L. lactis cells in the presence of a cognate EcfA1A2T module. Using a folate-auxotrophic (and naturally folate-transport deficient) E. coli strain as the host, it was unequivocally shown that heterologously produced FolT and EcfA1A2T from L. mesenteroides form a stable and functional folate transporter (Eudes et al., 2008; Neubauer et al., 2009; Rodionov et al., 2009).

Pantothenate

Pantothenate (vitamin B5) is a building block of coenzyme A and is thus essential for central metabolic reactions such as fatty acid synthesis. During de novo synthesis in a number of prokaryotes, in fungi and in plants, pantothenate arises by condensation of β-alanine and pantoate (see Spry et al., 2008 for a recent review). In prokaryotes, β-alanine is either produced by decarboxylation of aspartate or taken up from the environment. In E. coli, β-alanine uptake is mainly mediated by the CycA amino-acid transporter (Schneider et al., 2004). The pantoate moiety is derived from 2-oxo-isovalerate. Pantothenate is converted into coenzyme A in a five-step path via 4-phosphopantothenate, 4-phosphopantothenoylcysteine, 4-phosphopantetheine and dephospho-coenzyme A. Pantothenate and probably pantetheine can also be used in salvage pathways. Pantothenate is taken up either by a widespread Na+/pantothenate symporter (PanF, TC 2.A.21.1.1), or as recently predicted, by ECF transporters (mostly subclass II) using PanT as the S component (Rodionov et al., 2009). PanT is found in many pathogenic (e.g. Streptococcus sp.) and nonpathogenic lactobacteria and in Clostridia including Clostridium difficile, Clostridium perfringens and Clostridium tetani. These organisms are known to depend on exogenous pantothenate for growth. Heterologous production of PanT and EcfA1A2T from L. mesenteroides in E. coli resulted in stable quadripartite complexes in membranes of the recombinant host (Rodionov et al., 2009) and conferred pantothenate-transport activity on a PanF-deficient strain (Neubauer et al., 2009). These results clearly demonstrate the role of PanT-EcfA1A2T as a high-affinity pantothenate transporter.

Niacin

NAD is an important redox coenzyme, but also a substrate for certain enzymes such as bacterial DNA ligase and ADP-ribosyl transferases that decompose the molecule. NAD is synthesized de novo from aspartate via quinolinate and nicotinate mononucleotide. A variety of pathways for recycling of endogenous or exogenous NAD degradation products including nicotinate and nicotinamide (named niacin or vitamin B3 independent of the acid or amide function) are known, but many of the transporters for the exogenous compounds are not characterized. A recent functional genomic analysis of the NAD synthesis and salvage regulons governed by the niacin-responsive transcriptional repressor NiaR in Firmicutes identified the ribosyl nicotinamide transporter gene pnuC and the genes for three putative transporters (niaP, niaY and niaX) as regulated targets (Rodionov et al., 2008). NiaP of B. subtilis, a member of the major facilitator transporter family, complemented an NAD-auxotrophic E. coli mutant in the presence of nicotinamide and is thus implicated in niacin transport. Experimental data for such a role of NiaY and the more widespread NiaX are not available. NiaX was found only in Firmicutes containing an EcfA1A2T module, and it was tentatively predicted to function as the S unit of a subclass II ECF transporter for niacin (Rodionov et al., 2009).

Vitamin B6

Vitamin B6 exists in the three forms: pyridoxine, pyridoxal and pyridoxamine. The active form, pyridoxal-5-phosphate (PLP), is a cofactor of a large number of enzymes mainly involved in amino-acid metabolism. PLP is synthesized by microorganisms and plants by one of two ways: most bacteria including E. coli combine 4-phosphohydroxy-l-threonine (derived from erythrose-4-phosphate) and deoxyxylulose-5-phosphate, a metabolite also used for isoprenoid and thiamine synthesis, to produce pyridoxine-5-phosphate, which is oxidized to PLP. Some other bacteria including B. subtilis, archaea and eukarya synthesize PLP directly from ribulose-5-phosphate, dihydroxyacetone phosphate and glutamine (see Mooney et al., 2009) for a recent review on the metabolic roles and biosynthesis of vitamin B6). The various forms of vitamin B6 and their 5-phosphates are interconvertible, which allows recycling of the endogenous compounds released for instance during protein degradation. Salvage of exogenous vitamin B6 variants depends on the uptake of the metabolites. Surprisingly little is known on vitamin B6 transport into prokaryotic cells including lactobacterial species that are natural auxotrophs. Although vitamin B6 uptake was demonstrated for various lactobacteria (Mulligan & Snell, 1977) and S. enterica serovar Typhimurum (Mulligan & Snell, 1976), the transporters behind this process were not known. Homologous S units of ECF transporters, PdxT and HmpT, are predicted to function as vitamin B6 transporters in Firmicutes, Thermotogales and some archaea (Rodionov et al., 2009). In many Lactobacillales, pdxT/hmpT genes are colocalized with genes encoding pyridoxine-related kinases of two different types (similar to PdxK from E. coli and to ThiD/PdxK from B. subtilis), both of which have specificity for pyridoxine, pyridoxal and pyridoxamine (Park et al., 2004; Safo et al., 2006). In our previous study, based on the colocalization with thiD kinase genes, we proposed that HmpT transporters are involved in the thiamine salvage pathways. Similarity searches using psi-blast identified HmpT and PdxT as homologous proteins from the same family (COG4720). Thus, we revise our initial annotations and combine the HmpT and PdxT families into a single family of the pyridoxine-related transporters PdxU that may be specific to PLP precursors, pyridoxine, pyridoxal and pyridoxamine.

Vitamin B12-related compounds

Corrinoid-containing cofactors are synthesized in prokaryotes, and utilized for essential reactions by prokaryotes, animals and algae, but not by higher plants. Reactions that rely on B12 cofactors include 5′-deoxyadenosyl cobalamin-dependent rearrangements, methyl-cobalamin- or methyl-cobamide-dependent methyl-group transfer and reductive dehalogenation (see Banerjee & Ragsdale, 2003; Matthews et al., 2008 for a review). B12 cofactors are produced de novo from uroporphyrinogen III, an intermediate of tetrapyrrole synthesis, by two complex metabolic routes, resulting in the formation of an adenosylated cobyric acid derivative as another intermediate. The two routes, known as the aerobic and the anaerobic pathway, differ in the cobalt-insertion reaction that occurs early in the anaerobic path and late in the aerobic path, and in the elimination of the C20 carbon atom that is present in tetrapyrroles, but absent from the corrin skeleton. An aminopropanol moiety derived from threonine is attached to the adenosylated cobyrate to yield adenosyl cobinamide, which is converted to 5′-deoxyadenosyl cobalamin by the addition of α-ribazole. The latter is derived from dimethylbenzimidazole and nicotinic acid mononucleotide (reviewed by Scott & Roessner, 2002). Prokaryotes, for example those that do not contain the complete pathway for de novo B12 synthesis (e.g. E. coli K-12), can take up corrinoids from the environment and many prokaryotes use a canonical ABC transporter for this purpose. In E. coli, corrinoids enter the periplasm through the TonB-dependent outer-membrane transporter BtuB and are transported into the cytoplasm via the BtuCD-BtuF ABC transporter (Fig. 8e). These proteins have been analyzed in great detail and the crystal structures of BtuB (Chimento et al., 2003), the SBP BtuF (Borths et al., 2002), the complex of BtuC (TMD) and BtuD (NBD) (Locher et al., 2002) and the BtuCD-BtuF (Hvorup et al., 2007) holotransporter have been determined. Related ABC transporters are widespread among bacteria and archaea, and the respective genes are regulated by adenosylcobalamin-dependent riboswitch (B12 element) in most cases (Rodionov et al., 2003). The E. coli BtuCD-BtuF system was shown to transport adenosylcobalamin, the non-natural cyanocobalamin (vitamin B12) and, although at a lower rate, cyanocobinamide (Bradbeer et al., 1978).

CbrT is the predicted S component of subclass II ECF transporters for cobalamin in Lactobacillales (Rodionov et al., 2009). Several Lactobacillus sp. contain an nrdJcbrS–cbrTpduO gene cluster encoding adenosylcobalamin-dependent ribonucleotide reductase (NrdJ), a hypothetical lipoprotein (named CbrS), the putative S unit CbrT and a cobalamin adenosyltransferase (PduO) (Fig. 8e). The subclass I transport system CbrTUV was identified in other Firmicutes, in some Actinobacteria and in the archaeon Methanosphaera stadtmanae. It is accompanied by one or two distinct hypothetical lipoprotein components (e.g. CbrY-Z in Moorella thermoacetica and CbrX in B. cereus). The absence of btuFCD in the genomes that possess the cbrT or cbrTUV genes suggests the involvement of both transport systems in the uptake of corrinoids. The absence of CobU and CobS (enzymes converting adenosylcobinamide to adenosylcobalamin) and the presence of PduO [an enzyme that adenosylates cobalamin and cobinamide, for example in Lactobacillus reuteri (Park et al., 2008)] in Lactobacillus sp. suggest that CbrT has specificity to cobalamin.

In a number of Firmicutes, CblT is predicted to be the S unit of subclass II ECF systems and in a Desulfitobacterium sp. it belongs to a subclass I system. CblT may be involved in the salvage of dimethylbenzimidazole (Rodionov et al., 2009). A couple of these bacteria have adjacent cblT and cblS genes, but lack CobT that is required for the synthesis of α-ribazole-5′-phosphate from dimethylbenzimidazole. Thus, it was predicted that the cblT and cblS genes encode a salvage path for dimethylbenzimidazole consisting of a transporter and a novel type of α-ribazole-5′-phosphate synthase (Rodionov et al., 2003).

Lipoate

Lipoic acid is a sulfur-containing cofactor that resembles biotin in structure and is essential for the function of a number of enzymes including pyruvate dehydrogenase and 2-oxoglutarate dehydrogenase. Lipoic acid is covalently bound via its carboxyl group to a lysine residue in the lipoyl domain of lipoate-dependent enzymes. This reaction is catalyzed by lipoate-protein ligase (LplA) if free lipoate is available. De novo synthesis of lipoate has been extensively investigated in E. coli. It is produced from n-octanoic acid by two routes. In the minor path, octanoate is bound by LplA to the lipoyl domain of the target protein. Subsequently, lipoate synthase (LipA) inserts two sulfur atoms into the octanoyl moiety in a complex reaction resembling the biotin-synthase reaction. The major route depends on LipB that transfers the octanoyl moiety from octanoyl-acyl carrier protein, derived from fatty acid synthesis, to the target with the subsequent transfer of sulfur by LipA (see Cronan, 2008, for an overview). Transporters that may enhance the salvage of exogenous lipoate are unknown, but ECF systems may be involved in certain organisms. Several Phytoplasma sp. that contain neither lipA nor lipB contain an lplAlipT cluster encoding lipoate-protein ligase and the predicted S unit of a subclass II ECF transporter (LipT) that has a weak similarity to ThiT from Phytoplasma (Rodionov et al., 2009).

Other salvage pathways

Methionine

Besides its role as a proteinogenic amino acid, methionine fulfills many other functions in the cell. The adenosylated form, SAM, is the major donor of methyl groups, but is also used to produce adenosyl radicals for reactions catalyzed by enzymes of the radical/SAM family, as a donor of the ribosyl group for the queuosine modification in tRNA, and as a donor of its aminobutyryl moiety used for the synthesis of spermidine, biotin and N-acyl-homoserine-lactones, and for tRNA modification (see Hondorp & Matthews, 2006, for a review). Many organisms contain salvage pathways for recycling the degradation products of SAM. Splitting off the aminobutyryl group of SAM produces methylthioadenosine (MTA), which is a starting point of the salvage cycle. In the first step, adenine is cleaved off to yield methylthioribose-1-phosphate (MTR-1-P), either in a one-step reaction catalyzed by MtnP or in a two-step path via methylthioribose (MTR) by the serial action of MtnN and MtnK. Escherichia coli does not further recycle MTR, but excretes this compound. In many other organisms, MtnA isomerizes MTR-1-P to methylribulose-1-phosphate, which is used to produce methionine in a series of four or five additional reactions (reviewed by Sekowska et al., 2004).

Methionine is taken up by members of the MUT family (TC 3.A.1.24) of canonical ABC transporters encoded in E. coli by the metNIQ operon and in B. subtilis by the metQPN operon (see Sources of sulfur). In addition, secondary active transporters for methionine [MetPS (Trötschel et al., 2008)], S-methylmethionine [in E. coli (Thanbichler et al., 1999)] and SAM [in Rickettsia prowazekii (Tucker et al., 2003)] have been described. Comparative genomic analyses predicted the existence of subclass I ECF transporters for methionine precursors that arise by the aforementioned reactions (Rodionov et al., 2004, 2009). In a few prokaryotes, the genes mtaU (encoding a duplicated ABC ATPase), mtaV (encoding a T component) and mtaT (encoding a putative S component) are clustered with mtnA and mtnP, suggesting that MtaTUV transporters are involved in the uptake of MTA, MTR and/or MTR-1-P. MtsTUV is another transporter under the control of the methionine regulons (SAM riboswitch or methionine T-box regulatory elements in Firmicutes; MetJ repressor in Vibrio). Comparative genomic analysis (gene co-occurrence, colocalization and coregulation) of the ‘methionine metabolism’ subsystem in the SEED database allowed us to tentatively predict that MtsTUV systems are transporters for SAM. The reduced genomes of Phytoplasma lack any methionine biosynthesis enzyme including the SAM synthetase MetK. The absence of these essential enzymes in Phytoplasma is complemented by two transport systems: the methionine transporter MetNPQ and the predicted SAM transporter MtsTUV.

Queuosine and archaeosine

Queuosine (Q) and archaeosine (G+) are hypermodified 7-deaza-purine nucleosides present in the anticodon wobble position of bacterial and eukaryotic tRNAAsp, tRNAAsn, tRNAHis and tRNATyr, and in the D-loop of archaeal tRNAs, respectively. Whereas mammals take up the free base queuine with the diet or from the intestinal flora and use it to replace a guanine base by means of a tRNA guanine transglycosylase (TGT) enzyme, the modified nucleosides are produced de novo from GTP in prokaryotes. The PreQ0 precursor (a 7-cyano-7-deaza purine) is synthesized in a four-step pathway that is common to bacteria and archaea. In bacteria, PreQ0 is converted to PreQ1 (7-aminomethyl-7-deazaguanine), which is then transferred to tRNAs by TGT and subsequently converted to Q by the QueA enzyme. In archaea, PreQ0 is incorporated into tRNA and converted to G+ (Roth et al., 2007; Phillips et al., 2008; McCarty et al., 2009). Nucleolytic cleavage of Q-containing tRNAs could result in PreQ1, but not in PreQ0 or other precursors. Degradation of G+-containing tRNAs would hardly result in the free PreQ0 base. Because bacterial, archaeal and eukaryotic TGT enzymes are specific for PreQ1, PreQ0 and queuine, respectively (Stengl et al., 2005), variations of the predicted salvage pathway containing individual enzymes that produce the appropriate precursors are conceivable.

Two different families of ECF transporters operating with the S units QrtT and QueT were predicted to be involved in a hypothetical Q salvage pathway in prokaryotes (Rodionov et al., 2009). QrtT occurs in many prokaryotic groups and is the predicted S unit of subclass II ECF transporters and of subclass I ECF systems in combination with QrtU (T unit), QrtV and QrtW (A units). Likewise, QueT is found in diverse bacteria and archaea and considered as an S component of subclass II ECF systems (Rodionov et al., 2009). Predictions on the substrates of QrtT- and QueT-containing ECF systems are based on the genome-context evidences captured in the SEED database (Overbeek et al., 2005). In the genomes of Firmicutes, the qrtT and queT genes are coregulated with the Q-biosynthetic operon by representatives of two classes of PreQ1-responsive riboswitches (Roth et al., 2007; Meyer et al., 2008), and co-occur with the tgt and queA genes, but not with the de novo PreQ1 biosynthesis genes, suggesting the involvement of the respective transporters in the PreQ1 salvage. Colocalization of qrtT with a hypothetical homolog of the inosine-uridine nucleoside N-ribohydrolase gene iunH in many Firmicutes suggests that these organisms utilize queuosine salvage. Projection of the predicted specificities of QueT and QrtT in bacteria to archaeal homologs is challenging because the G+ biosynthesis in Archaea does not utilize PreQ1.

Tryptophan

Tryptophan is one of the rarest amino acids in most proteins and can be synthesized by most microorganisms, except some pathogenic (e.g. Streptococcus) and symbiotic species. Both auxotrophic and prototrophic species often possess Trp transporters, for instance E. coli and related organisms possess the Mtr permease, whose production is under the control of the Trp repressor (Heatwole & Somerville, 1991; Sarsero et al., 1991). The tryptophan-specific S unit TrpP (also known as YhaG) was identified as a member of the tryptophan regulon operated by the TRAP attenuation protein in B. subtilis (Sarsero et al., 2000; Yakhnin et al., 2004). Involvement of TrpP in tryptophan transport was confirmed by sensitivity tests of the trpP mutant to growth-inhibiting levels of tryptophan analogs (Sarsero et al., 2000). trpP orthologs in Clostridia are regulated by the Trp-specific T-box attenuator (Vitreschak et al., 2008). TrpP proteins in Firmicutes are a part of subclass II ECF transporters for tryptophan (Fig. 8f), whereas its homologs in methanogenic archaea belong to the subclass I systems. A canonical ABC transporter for tryptophan (termed TrpXYZ) regulated by the Trp-specific T-box attenuator was predicted in some gram-positive bacteria that lack TrpP (Vitreschak et al., 2008).

ABC and ECF importers in plants?

As noted at the outset, ABC transporters in eukaryotes were until recently considered to mediate substrate movement only in an inside-to-outside direction from the cytoplasm across the plasma membrane or across the membranes of vacuoles and peroxisomes. Based on the genome sequences of A. thaliana and O. sativa, plants are considered to contain a total of >120 ABC transporters, the majority of which are involved in the defense against biotic and abiotic stresses (see Rea, 2007; Yazaki et al., 2009, for a review). In recent times, however, experimental evidence accumulated, suggesting that distinct ABC systems in plants transport substrates across the plasma membrane into cells. The list of ABC importers in the plasma membrane includes the AtABCB4 and AtABCB14 systems in A. thaliana and the CjMDR1 transporter of the medicinal plant Coptis japonica. CjMDR1 is implicated in the transport of berberine, a benzylisoquinoline alkaloid with antibacterial and antimalarial activity, from xylem vessels into rhizome cells. The function of CjMDR1 as an uptake system was confirmed upon production in Xenopus oozytes (Shitan et al., 2003). Heterologous production of the CjMDR1 homolog AtABCB4 resulted in indole-3-acetic acid (IAA)-uptake activity of HeLa cells (Terasaka et al., 2005) and in the hypersensitivity of yeast cells to IAA and the toxic auxin analog 5-fluoroindole (Santelia et al., 2005). These results suggest a function of AtABCB4 in auxin uptake. AtABCB14 mediates the uptake of malate from the apoplast into guard cells, a process involved in the CO2-responsive regulation of closure of stomata. Recombinant production of AtABCB14 conferred malate-uptake activity on E. coli cells lacking the endogenous dicarboxylate transporter and enhanced malate uptake of HeLa cells (Lee et al., 2008). These findings suggest that at least distinct full-size ABC transporters such as AtABCB4, AtABCB14 and CjMDR1 have the potential to couple intracellular ATP hydrolysis to the uptake of substrates across the plasma membrane.

The genomes of various plant lineages ranging from deeply branching algae (e.g. Ostreococcus) via mosses (e.g. Physcomitrella) to crops encode homologs of prokaryotic T components and thus it is tempting to speculate about the existence of ECF transporters in plants. With very few exceptions (e.g. a BioY homolog encoded in the chloroplast genome of Mesostigma viride, a freshwater flagellate and very deeply branching member of the green plants), homologs of prokaryotic S components with known or predicted substrate specificity have not been identified in plants. Thus, questions on potential substrates remain elusive. In silico analyses of the available plant T protein sequences predict plastid-specific transit peptides in several cases, suggesting that those T units may be located in the chloroplast envelope (Fig. 6). Plant organelles play important roles in vitamin synthesis. The entire de novo synthesis of folate, a central step in pantothenate synthesis and the final step of biotin synthesis, occur in the mitochondria (reviewed by Smith et al., 2007). Because biotin is required for fatty acid synthesis in chloroplasts, the vitamin must be exported from the mitochondria and imported into chloroplasts. Central steps of thiamine synthesis are catalyzed in chloroplasts and in the mitochondria (Kong et al., 2008; Huang et al., 2009). These facts indicate that vitamins and vitamin precursors must be transported across the organellar envelopes. Given the substrate preference of prokaryotic ECF transporters and the existence of T components in plants, the assumption that ECF-related systems may be involved in vitamin transport across organellar membranes is an attractive speculation.

A group of T components of unknown function in cyanobacteria is most closely related to the aforementioned plant homologs (Fig. S2a). In a number of cyanobacteria, these T proteins are encoded in a conserved cluster containing engA, the T-unit gene and pipX. PipX is a small protein that interacts with PII and NtcA and is thus involved in nitrogen regulation. EngA is a widespread GTPase that plays an important role in ribosome assembly. At present, the genome context of this special type of cyanobacterial T components does not allow to draw conclusions on potential physiological roles.

Common features of canonical ABC and ECF importers, and perspectives

The structure and mechanism of many canonical ABC importers have been analyzed in great detail. On the other hand, and due to the fact that ECF systems as an abundant new class of transporters were described only recently, experimental analysis of the latter is still in its infancy. Nevertheless, the available data allow a comparative view of the two types of ABC importers.

Utilization of ABC ATPases is the immediately obvious common feature of the two types of ATP-dependent transporters. Thus, questions of evolutionary origins are of interest. Phylogenetic analyses of sets of ATPases of canonical ABC importers and of ECF transporters clearly show that the NBDs of the two transporter groups form distinct branches (Fig. S3), indicating that they have diverged early in evolution. Nevertheless, sequence alignments show that the NBDs of ECF-type transporters contain all the signatures that are known as being essential for ABC proteins. Structural data are available for a multitude of NBDs of canonical ABC importers, but for only a few NBDs of ECF systems. Structural alignments of CbiO, the NDB of the ECF-type Co2+ transporter of C. perfringens (Ramagopal UA, Morano C, Toro R, Dickey M, Do J, Burley SK & Almo SC, unpublished data; PDB code 3GFO), with its canonical-type relatives, do not reveal significant differences in the catalytically important domains.

At first glance, the property of at least certain S components to function as transporters in the solitary state seems to be completely different from canonical ABC importers that strictly depend on an extracytoplasmic SBP and the NBDs in addition to the transmembrane components. However, distinct mutations in the genes for TMDs (Covitz et al., 1994) or NBDs (Speiser & Ames, 1991) of two canonical ABC importers are known, which render substrate uptake independent of the SBP. Substrate transport by the TMDs in the absence of both the SBP and the NBDs has not been reported, but preliminary results indicate that such functional variants of the maltose transporter might exist (E. Schneider, unpublished data). In this respect, it is noteworthy that a pmf-driven activity was demonstrated for LmrA, a multidrug ABC exporter from L. lactis, when lacking its NBDs (Venter et al., 2003).

Intramolecular signaling between the Q loops of the NBDs and the TMDs may follow similar principles. In canonical ABC systems, the Q loop interacts with the EAA loop, a conserved helical region in the TMDs. The conserved Ala-Arg-Gly signatures (with variations as regards the Ala and Gly residue, indicated in the legend to Fig. 6) in a helical segment of T units may play a similar role in ECF systems.

Questions of whether or not the transport cycles of canonical ABC and ECF importers follow similar principles need future experimentation. According to the standard model for the former systems, the substrate-charged SBP plays an important role in initiating the cycle. Obviously, the mechanism must be different in the case of ECF systems because soluble SBPs are not involved. The scarcely available data on a biotin importer suggest that ATP hydrolysis of the BioMNY holotransporter is independent of the addition of a substrate. Because of the extremely high affinity of ECF systems for their substrates, it cannot be excluded that biotin remains bound to the BioMNY complex during purification. Thus, questions of whether the substrate induces hydrolysis of ATP remain enigmatic.

Other questions concern the supramolecular assembly of ECF systems in comparison with canonical ABC importers. The latter consist of two TMDs, two NBDs and (except for the hitherto only few importers in plants) an SBP and thus have a five-domain structure. The subunit assembly of ECF systems is unclear. With the exception of the heterooligomeric S units of the metal transporters, core transporters have six (or fewer) predicted transmembrane helices. Certain S units can function as stand-alone transporters, i.e. in the absence of a T unit and the NBDs (A units), and it is not likely that they act as monomers. Data on the subunit assembly of the holotransporter complexes have not been reported and thus, additional experimentation on this intriguing class of membrane transporters is needed.

Finally, it is worth comparing the substrate spectra of canonical ABC and ECF importers. Canonical systems in prokaryotes transport a wide variety of substrates ranging from carbon, energy and nitrogen sources, which are required in large amounts by the cells, to transition metal ions that are required and tolerated only in trace amounts. The majority of ECF systems are transporters of micronutrients, frequently vitamins, and nickel and cobalt ions. This specificity correlates with an extremely high affinity for the substrate in the low nanomolar or picomolar range of the S components. The specific contributions of the A and T components in flux control remain to be elucidated.

Acknowledgements

Work by the authors' groups was supported by grants SCHN 274/9-3 and SFB 449 (TP B14) from the Deutsche Forschungsgemeinschaft to E.S., the program ‘Molecular and Cellular Biology’ of the Russian Academy of Sciences and a Russian President's grant (MK-422.2009.4) to D.A.R. and grants EI 374/3-1 and EI 374/4-1 from the Deutsche Forschungsgemeinschaft to T.E.

Note added in proof

In a recent overview, Berntsson et al. (2010) proposed a new classification system for SBPs of canonical ABC importers and other transport systems based on structural features. The key role of Cu in methanotrophy and the function of methanobactin as a chalkophor has been summarized by Semrau et al. (2010). A recent report by Englert et al. (2010) challenges the previous suggestion that the SBP NikA participates in repellent chemotaxis. Picomolar substrate affinity was identified for another S unit (Erkens & Slotboom, 2010).