What makes a heavy metal resistant bacterium heavy metal resistant? The mechanisms of action, physiological functions, and distribution of metal-exporting proteins are outlined, namely: CBA efflux pumps driven by proteins of the resistance–nodulation–cell division superfamily, P-type ATPases, cation diffusion facilitator and chromate proteins, NreB- and CnrT-like resistance factors. The complement of efflux systems of 63 sequenced prokaryotes was compared with that of the heavy metal resistant bacterium Ralstonia metallidurans. This comparison shows that heavy metal resistance is the result of multiple layers of resistance systems with overlapping substrate specificities, but unique functions. Some of these systems are widespread and serve in the basic defense of the cell against superfluous heavy metals, but some are highly specialized and occur only in a few bacteria. Possession of the latter systems makes a bacterium heavy metal resistant.
1 Introduction: what kind of heavy metal resistance will be described
Metal resistance determinants were initially found on bacterial plasmids [1–13]. However, after the genome sequences of several bacteria became available, homologues to metal resistance determinants popped up in many bacteria (see below). If the possession of a heavy metal resistance determinant makes a bacterium metal resistant, which bacteria remain sensitive? Since Escherichia coli serves as model bacterium for many purposes, metal resistance of a given bacterium could be compared to metal resistance of E. coli.
This comparison was done with the well-characterized heavy metal resistant bacterium Ralstonia metallidurans strain CH34 (previously Alcaligenes eutrophus or Ralstonia sp. ). It originates from a zinc decantation tank and related bacteria have been found in many metal-contaminated environments around the world [16, 17]. The minimal inhibitory concentrations (MICs) of the biologically important heavy metals  for R. metallidurans strain CH34 and for E. coli strain K38 were compared in Tris-buffered mineral salts medium . Growth of neither bacterium was inhibited by up to 50 mM MoO42−, WO42− and, due to its low solubility, Fe3+. Both species had a MIC of 2 mM UO22−. E. coli was more resistant to VO3−, AsO43− and SbO2− than R. metallidurans; however, R. metallidurans was 2.5-fold more resistant to CrO42− than E. coli. Interaction of microbes with arsenic has recently been reviewed  and will not be elucidated here. Thus, of these seven biologically important oxyanion-forming heavy metals , only chromate resistance will be discussed. The first part of this review will concentrate on the heavy metal cations.
The MIC values of both bacteria to biologically important heavy metal cations (with the exception of iron) were plotted against the solubility constants of the respective metal sulfide complexes (, Fig. 1). This was done because heavy metal cations that escape metal efflux systems may cause toxicity by interaction with cellular thiol compounds such as glutathione. The solubility of the sulfide complexes may therefore serve as a measure of the affinity of a metal to thiol compounds and for the severity of damage resulting from this interaction. The different degrees of metal resistance roughly followed the solubility constants (regression coefficients >80%); so, this interaction may indeed be part of metal toxicity.
Heavy metal toxicity depends on the affinity of a cation to sulfur. The logarithm of the MIC of heavy metal cations for E. coli (open circles) and R. metallidurans (closed circles) was plotted against the logarithm of the solubility for the respective metal sulfide as a measure of the affinity to sulfur . The positions of the cations are indicated with arrows at the top. An exponential curve fit was performed for both data sets that yielded regression coefficients for E. coli (thin line) of 81.1% and for R. metallidurans (thick line) of 86.7%.
Metal resistance of R. metallidurans was in general higher than metal resistance of E. coli (Fig. 1). Especially resistance of R. metallidurans to the cations of cobalt, zinc, nickel, and copper was very high, cadmium resistance was similar to that of E. coli. This review will focus on the description of all resistance systems that are responsible for this remarkable heavy metal resistance of R. metallidurans and it will give an overview of the distribution of these heavy metal resistance determinants in other organisms.
2 The first layer of heavy metal resistance: members of the resistance–nodulation–cell division (RND) protein family export superfluous cations
The RND protein family was first described as a related group of bacterial transport proteins involved in heavy metal resistance (R. metallidurans), nodulation (Mesorhizobium loti) and cell division (E. coli) . This family has grown into a huge superfamily that includes seven protein families that can be found in all major kingdoms of life . In bacteria and archaea members of this superfamily are involved in transport of heavy metals, hydrophobic compounds, amphiphiles, nodulation factors, and, with SecDF, in protein export. The members of the RND superfamily in eukaryotes transport sterols or serve as receptors [24–45]. This section deals with the heavy metal efflux family (HME-RND ) of the RND superfamily that has received the number TC 2.A.6.1.1 in Milton Saier's functional-phylogenetic classification system for transmembrane solute transporters . RND proteins involved in export of hydrophobic and amphiphilic compounds (HAE-RND) will be included, since data obtained for these systems allow the development of models for the function of RND proteins and RND-driven efflux complexes. Therefore, the expression ‘RND protein’ will be used exclusively hereafter for HME-RND and HAE-RND proteins, but not for eukaryotic proteins or bacterial proteins involved in protein export.
Bacterial RND proteins like company. Adjacent to a gene that encodes a RND protein is – in most cases – a second gene that encodes a member of the membrane fusion protein family MFP . This family has also been designated periplasmic efflux protein family  or, more recently, periplasmic adapter protein family . This review will use the oldest scientific name, that is MFP. Besides MFPs, many RND proteins cooperate with yet a third protein that belongs to the family of outer membrane factors (OMF) [47, 49]. These three proteins form an efflux protein complex that may export its substrate from the cytoplasm, the cytoplasmic membrane or the periplasm across the outer membrane directly to the outside [50–53]. To distinguish this type of transport system from ATP-driven ABC transport systems, the RND-driven export systems will be referred to as CBA efflux systems or CBA transporters.
2.1 CzcCBA-mediated efflux is the outer shell of Co2+, Zn2+, Cd2+ and Ni2+ resistance in R. metallidurans
The first identified member of the RND family was the CzcA protein from R. metallidurans followed by the CnrA protein from the same bacterium . This β-proteobacterium contains two megaplasmids involved in heavy metal resistance [17, 19]. The presence of the larger of both plasmids, pMOL30, increased the MICs to Zn2+ 50-fold, to Co2+ 33-fold and to Cd2+ seven-fold , while the second plasmid, pMOL28, mediated resistance to Ni2+ and a 16-fold increase in cobalt resistance.
The plasmid pMOL30-mediated resistance to Co2+, Zn2+ and Cd2+ was designated Czc and the structural gene region of the czc resistance determinant of pMOL30 was cloned  and sequenced [55, 56]. This structural gene region contains the genes for the OMF CzcC, the MFP CzcB and the CzcA protein of the RND family. The three genes form an operon czcCBA that is transcribed tricistronically  and is flanked by a multitude of genes involved in metal-dependent regulation of czcCBA expression (C. Große, A. Anton, T. Hoffmann, S. Franke, G. Schleuder and D.H. Nies, unpublished) [58–60].
The czcCBA structural gene region was expressed in a plasmid-free R. metallidurans strain under the control of a constitutive promoter on a low copy number vector. The presence of the czcCBA genes in this bacterial strain mediated a metal resistance level similar to that reached by the complete megaplasmid pMOL30 [54, 55]. Czc-mediated resistance to Co2+, Zn2+ and Cd2+ was based on a diminished cellular accumulation of the three cations resulting from cation efflux  driven by the proton motive force .
The cobalt–nickel resistance determinant cnr on plasmid pMOL28 is also based on cation efflux [61, 63]. The resistance determinant is composed of a cnrCBA structural region  that is preceded by the regulatory gene region cnrYXH [65, 66]. Similar to czcCBA, cnrCBA encodes the OMF CnrC, the MFP CnrB and the RND protein CnrA. Another cobalt–nickel resistance determinant, ncc, was cloned from R. metallidurans strain 31A (previously Achromobacter xylosoxidans subsp. xylosoxidans) and further characterized . Similar to cnr, ncc is composed of a regulatory gene region nccYXH followed by the structural region nccCBA, again encoding a putative outer membrane protein (NccC), a MFP (NccB), and an RND protein (NccA). In contrast to Cnr, Ncc mediates resistance to cadmium in addition to cobalt and nickel resistance .
Czc alone seemed to be responsible for the increase in Co2+, Zn2+ and Cd2+ resistance that was assigned to plasmid pMOL30. However, the genomic sequence of R. metallidurans (preliminary sequence obtained from DOE/JGI, http://www.jgi.doe.gov/JGI_microbial/html/ralstonia/ralston_homepage.html) indicated the presence of a ncc determinant very similar to ncc from R. metallidurans strain 31A in the genome of strain CH34. The ncc determinant from R. metallidurans was assigned to plasmid pMOL30 by a polymerase chain reaction and homologous recombination experiment (S. Franke and D.H. Nies, unpublished). Thus, Ncc from R. metallidurans strain CH34 may contribute to the cobalt, zinc and cadmium resistance mediated by plasmid pMOL30. The exclusive assignment of this resistance to Czc could be an experimental artifact, e.g. resulting from the higher copy number of the vector plasmid compared to the native plasmid pMOL30. However, this possibility was excluded when the czcA gene was deleted from pMOL30. The resulting pMOL30(ΔczcA)-containing mutant strain was even more sensitive to cobalt, zinc and cadmium than the plasmid pMOL30-free strain . Since polar effects were carefully avoided when the ΔczcA deletion was constructed, this indicates clearly the essential function of Czc as the outermost shell of cobalt, zinc, and cadmium resistance in R. metallidurans. In agreement with this fact, no regulatory genes were found upstream of nccCBA on plasmid pMOL30 and no expression of this operon could be measured (G. Rehbein, C. Grosse and D.H. Nies, unpublished). Thus, ncc seems to be a silent determinant in R. metallidurans CH34.
2.2 Localization and speciation of the cations effluxed by CzcCBA
Why does the CzcCBA efflux system mediate such a huge increase in metal resistance? Maybe, this system diminishes not only the cytoplasmic concentration of heavy metal cations but additionally (or even exclusively?) the periplasmic concentration. In this case, the cations could be removed before they had the opportunity to enter the cell. Moreover, the CzcCBA efflux system could mediate further export of the cation that had been removed from the cytoplasm by other efflux systems.
Such export of heavy metal cations from the periplasm may happen indirectly, that is by the combined action of an uptake and a CBA transenvelope efflux system, but periplasmic cations could also be directly funneled into a CBA transenvelope efflux system. Yet, both mechanisms may contribute to the physiological function of CzcCBA and other CBA transport systems. As shown with purified protein that has been reconstituted into proteoliposomes, CzcA should be able to accept heavy metal cations from the cytoplasm . Since transport kinetics of the reconstituted CzcA protein and the CzcCBA complex as characterized in inside-out vesicles are similar , the complete CzcCBA complex should also be able to export cations from the cytoplasm. Efflux of cadmium by CzcCBA fits into a quantitative physiological model describing cadmium homeostasis . This favors the possibility of export from the cytoplasm over export directly from the periplasm.
On the other hand, the K50 values of CzcA and the CzcCBA complex for Co2+, Zn2+ and Cd2+ are in the lower mM range [62, 69]. The trace elements Co2+ and Zn2+ are transported with a sigmoidal substrate saturation function (with a Hill coefficient of 2). This keeps the efflux velocity of the CzcCBA complex slow when cytoplasmic concentrations are below the K50 value, but allows rapid export at higher concentrations. A cytoplasmic flow equilibrium model predicts the apparent cytoplasmic concentration of Co2+ to be 8.4 mM and of Zn2+ to be below 6.6 mM . These values are in contrast to the situation described for E. coli by Outten and O'Halloran ; in this bacterium, the apparent concentration of Zn2+ is kept at 0.2 mM. Moreover, the authors predict by analysis of the interaction of Zn2+ with regulatory proteins a free cellular zinc concentration in the fM range. If R. metallidurans is not completely different from E. coli, the concentration of free zinc in the cell will not allow efficient export of zinc from the cytoplasm by CzcCBA in the cellular model of Outten and O'Halloran.
As shown in Fig. 1, heavy metal cations differ in their affinity for sulfur. This affinity relates to the toxicity of the individual heavy metal cation. Gram-negative bacteria contain an intracellular glutathione pool of about 10 mM [71, 72], other bacteria harbor different thiol compounds . Glutathione complexes of extremely sulfur-loving heavy metal cations have complex binding constants of 1042 (Hg2+–bis-glutathionate complex ) and 1039 (Cu+–glutathionate complex ). Copper should never occur ‘free’ in a cell , but the complex binding constants for Cd2+, Zn2+, Co2+, and Ni2+ should be below the copper value. To explain the fM concentrations of ‘free’ zinc observed by the regulatory proteins in E. coli at an apparent cytoplasmic concentration of 0.2 mM Zn2+ and 10 mM glutathione, a binding constant of at least 1013 is required, which is a moderate assumption of this value. Thus, in the cells of Gram-negative bacteria cellular zinc that is not needed in the active centers of enzymes may be present exclusively in glutathionate complexes once the thermodynamic equilibrium has been reached.
However, CzcA was not able to transport 65Zn2+ into proteoliposomes when glutathione was present . Formation of a zinc–mono-glutathionate complex in vitro has a rate constant of 3.9×103 M−1 s−1. This would lead to cytoplasmic concentrations of ‘free’ Zn2+ in the μM range in a flow equilibrium between uptake and a theoretical flux into a glutathione-bound pool (calculation not shown). This concentration is 1010-fold higher than the fM concentration of ‘free’ zinc predicted for E. coli, but still 1000-fold lower than the K50 value of CzcA for zinc . If the in vivo rate of the formation of zinc–mono-glutathionate complexes is two or three orders of magnitude lower than the rate determined in vitro , CzcCBA will be able to detoxify cytoplasmic Zn2+. Otherwise, this transport system relies on other transporters that export Zn2+ into the periplasm, meaning that in this case CzcCBA-mediated efflux is a direct periplasmic detoxification process. But, is there any evidence for a direct export of metal cations out of the periplasm? To investigate this question, a copper/silver-transporting system from E. coli will be discussed.
2.3 The Cus system from E. coli: is there direct export from the periplasm?
Copper and silver have a much higher affinity towards sulfur than does zinc. The half-life of Cu+ and Ag+ as ‘free’ ions in the cytoplasm is thus probably much shorter than the half-life of ‘free’ Zn2+, and no ‘free’ copper should occur in a bacterial cell . Thus, evidence for direct periplasmic detoxification may be more easily obtained for silver and copper than for zinc or cobalt.
A plasmid-encoded copper resistance determinant for E. coli has been known for a long time [9, 59, 78–86]. However, copper resistance factors encoded by the chromosome of this bacterium were described only recently. A copper-exporting P-type ATPase of E. coli, CopA [87, 88], transports the heavy metal cation from the cytoplasm into the periplasm like other soft metal-transporting P-type or CPx-type ATPases , without any evidence for further transport from the periplasm to the outside. The only HME-RND of E. coli is the CusA protein. Its gene is embedded in the structural gene operon cusCFBA on the bacterial chromosome that encodes CusA, the MFP CusB, the OMF CusC and a small additional protein, CusF . Upstream of cusCFBA and divergently transcribed on the other DNA strand are two regulatory genes that encode the factors of a two-component regulatory system CusS (sensor/histidine kinase) and CusR (response regulator). Both are involved in control of cellular copper homeostasis  and of cusCFBA expression .
The proteins of the Cus system mediate resistance to silver [90, 92]. Resistance to copper could not be assigned to Cus in the first study , not even in a mutant background with the copA gene for the copper-detoxifying P-type ATPase disrupted. However, copper was a better inducer of cus expression than silver , indicating involvement of Cus in copper resistance. Outten et al. found a clear contribution of Cus to copper resistance under anaerobic conditions  while copper detoxification under aerobic conditions was performed by CopA and the periplasmic protein CueO. Under anaerobic conditions periplasmic copper ions control copper homeostasis via CusRS regulation, while under aerobic conditions cytoplasmic copper exerts control via the regulator CueR .
Finally, Grass and Rensing  demonstrated that an essential function of Cus for copper resistance is easily visible in mutant strains devoid of CueO. This protein is a multi-copper oxidase located in the periplasm  explaining why CopA and CueO work together in detoxifying copper, but only under aerobic conditions . CopA exports Cu+ from the cytoplasm into the periplasm . If not prevented by further transport, Cu+ may react with periplasmic components to form Cu2+ and would therefore reduce and impair these components. CueO could re-oxidize and repair these periplasmic components again. Electrons can be transferred by the copper atoms of CueO  to molecular oxygen leaving a repaired periplasmic component and water. Since Ag+ cannot be oxidized to Ag2+ so easily, Ag+ should not be able to damage periplasmic components by oxidation (but it may damage them by binding). Thus, CueO cannot repair silver damage and silver resistance mediated by Cus was visible in the presence of CueO 
In the absence of CueO or in the absence of molecular oxygen [93, 94], however, the Cus system is required to maintain copper homeostasis. Cus complements a missing repair or protection process in the periplasm. Recent work on the plasmid-encoded copper resistance system Pco discusses the importance of oxidizing periplasmic Cu+ to Cu2+ to protect the cell [83, 96]. Periplasmic proteins, the MFP CusB and the OMF CusC, were required for full copper resistance . Substitutions of methionine residues in the periplasmic portion of the CusA protein resulted in loss of copper resistance demonstrating their functional importance .
The major difference in the organization of the czcCBA and the cusCFBA structural gene regions is the presence of cusF in the cus determinant. Similar genes only occur in putative or identified silver/copper resistance determinants. The small 10-kDa protein CusF is a periplasmic protein that binds one copper per polypeptide and may function as a copper usher that facilitates access of periplasmic copper towards the CusCBA efflux pump . Therefore, the main function of Cus may be the direct export of Cu+ and Ag+ from the periplasm to the outside [93, 94]. This is necessary because both monovalent cations should never occur ‘free’ in the cytoplasm owing to their high affinity to sulfur. More evidence for possible direct periplasmic access of substrates into CBA efflux systems comes from the HAE-RND proteins that transport organic substances [98, 99].
2.4 Export of organic substrates by HAE-RNDs
Because of their eminent medical importance , many HAE-RND proteins have recently been characterized. These RND proteins are involved in the detoxification of organic substances, including antibiotics, in pathogenic or non-pathogenic bacteria. Resistance to antibiotics mediated by RND proteins was responsible for half of the multiple drug resistant Pseudomonas aeruginosa strains that emerged in vivo in a hospital , showing the importance of RND proteins for the medical problem of antibiotic resistance. This pathogenic bacterium contains at least 13 RND transport systems (see below). One is a HME-RND protein and exports heavy metals , the other 12 are probably HAE-RND proteins. Best characterized of these proteins are MexB, MexD and MexF. Together with their respective MFPs (MexA, MexC, MexE) and OMFs (OprM, OprJ, OprN), these systems detoxify a variety of organic substances. Unfortunately, nomenclature of the HAE-RNDs did not follow the older nomenclature of the HME-RNDs, which may lead to confusion. MexB is the CzcA-related RND protein while MexA is the CzcB-related MFP.
The specificity of the MexB-driven system mainly includes β-lactam antibiotics, β-lactamase inhibitors and organic solvents [102–104], while the MexD-driven system expels cepheme compounds . Chloramphenicol, fluoroquinolones and imipenem are best exported by the MexF-driven exporter . Recently, the substrate specificities of two more Mex systems were investigated, MexJK and MexGHI [107, 108], and the functions of some systems were characterized in multiple deletion mutant strains .
E. coli contains six HAE-RND proteins and the Cus system as the sole HME-RND protein. The different substrate specificities of all six HAE-RNDs have been characterized [110, 111], but most data are available for the AcrAB-TolC system [98, 112–115] that protects the bacterium against a variety of toxic substances especially during slow growth . The AcrD-driven system from E. coli is an aminoglycoside efflux pump .
Examples of characterized HAE-RND-driven systems from other bacteria are AcrAB from Haemophilus influenzae, MtrCDE from Neisseria gonorrhoeae [119, 120], SrpABC and TtgDEF from Pseudomonas putida [121, 122], CmeABC from Campylobacter jejuni [123, 124], AdeABC from Acinetobacter baumannii, SmeDEF from Stenotrophomonas maltophilia, IfeAB and AmeABC from Agrobacterium tumefaciens [127, 128], NorMI from Brucella melitense, XepCAB from Porphyromonas gingivalis, and a CBA efflux system from Serratia marcescens. The substrate specificity of an RND-driven export system can include many organic substances with different chemical structures. Common to all these components is only a hydrophobic ability or at least a hydrophobic side chain. Examples are various β-lactam antibiotics [132–134], β-lactamase inhibitors , other hydrophobic antibiotics and toxic substances [117, 129, 130, 136, 137], organic solvents [104, 112, 114, 115, 121, 122, 138], bile salts , indole , and even homoserine lactone compounds involved in quorum sensing [127, 140, 141]. A title announcing resistance to vanadate mediated by a RND-driven system , however, has to be considered with caution, since the effect on vanadium resistance is via the well-known action of an RND-driven system on quorum sensing.
2.5 Biochemistry of RND proteins: towards a model of function for these proton-driven antiporters
Contribution to resistance to a multitude of toxic substances has been assigned to RND proteins, but biochemical and genetic analysis has only reached a few members. The RND protein CzcA that expels Co2+, Zn2+, and Cd2+ was purified and reconstituted actively into ammonium chloride-charged proteoliposomes . The result demonstrated that CzcA is a proton–cation antiporter depending solely on the Z·ΔpH portion of the proton motive force and not on its ΔΨ portion . This was in agreement with previous data obtained with the CzcCBA complex that was heterologously expressed in E. coli and characterized in inside-out vesicles .
Not many transport data are available for other RND proteins. Only AcrB has also been purified and reconstituted into proteoliposomes . Depending on the presence of ΔpH, the protein catalyzed the extrusion of fluorescent phospholipids. Known substrates of AcrAB, such as bile acids, erythromycin, and cloxacillin, inhibited this activity. Action of various drugs in the presence of an inverted pH gradient (inside acidic) suggested that AcrB is a proton antiporter  like CzcA. Thus, members of both RND families, HME and HAE, are proton–substrate antiporters.
With reporter gene fusions, the topology of several RND proteins was investigated. These proteins include CzcA , MexB , and MexD . RND proteins are huge proteins of more than 1000 amino acid residues (aa) in size. They are composed of two related halves that are probably the result of an ancient gene duplication and fusion , each containing a large periplasmic loop of about 300 aa in size that is flanked by 12 transmembrane α-helices, TMH I to TMH XII.
TMH IV contains amino acid residues that are conserved in most RND proteins . HME-RND proteins involved in the export of divalent heavy metal cations (CzcA, CnrA) exhibit the consensus sequence DFGX3DGAX3VEN in this TMH (Fig. 2). Exporters for monovalent cations like CusA and the silver transporter SilA  have consensus sequences similar to AVGX3DAAX3IEN (Fig. 2) and do not contain the ‘DFG’ motif. The HAE-RND proteins such as AcrB and MexB that export organic substances do not have the first aspartate residue in the DFG region, but a double DD in the middle (characteristic DD motif, Fig. 2). These conserved sequences allow the prediction for the specificity of many RND proteins with merely a glance.
Signature sequences of selected RND proteins. Representative RND proteins of the HAE family (AcrB from E. coli), HME group 1 (CzcA from R. metallidurans), HME group 2 (CnrA from R. metallidurans), HME group 3 (two putative proteins from R. metallidurans denoted as ‘p'contig number/gene number, and two proteins from Helicobacter pylori and Aquifex aeolicus), HME group 4 (SilA from Salmonella enterica pv. typhimurium), and HME group 5 (all7631 from Nostoc) were analyzed to demonstrate the relation of the groups to each other. A short sequence alignment of the transmembrane helix IV is shown, which contains the signature motifs of these proteins. The red box indicates the ‘DFG’ occurring in HME1, HME2 and a few HME3 proteins. The green box above the red one is the ‘AI/VG’ at the same position in the other protein groups. The blue box frames the ‘DD’ motif of HAE and HME5 proteins, the gray one below the ‘DSS’ at the same position that is a characteristic of HME group 3. Finally, the purple arrows mark unique and indicative polar and acidic residues that occur at this position only in HME5 proteins.
Both aspartate residues and the glutamate residue in TMH IV of CzcA were essential for function . The same was true for the single aspartate and the glutamate residue of CusA . For MexF, one of the two aspartate residues in the DD motif and the glutamate residue were also essential . Thus, the presence of charged amino acid residues in TMH IV of RND proteins is required for the proton–substrate antiport. Both aspartate residues of CzcA are probably involved in proton translocation . As in other proteins [147–150], the aspartate residues may therefore be part of a charge relay system that forms the path of the protons across the cytoplasmic membrane.
These data led to a preliminary model of how CzcA and other RND proteins might function  (Fig. 3). This simple model for the efflux of cytoplasmic cations by an RND export system can easily be modified to fit other postulated functions for RND proteins (Fig. 3). It is not relevant whether the substrates access the periplasmic substrate binding sites from the cytoplasm (as in the case of Zn2+?), or from the periplasm (as in the case of Cu+?), or from inside the cytoplasmic membrane (maybe in the case of hydrophobic substrates for HAE-RND proteins). In all cases, a protonation/deprotonation cycle couples the process of substrate binding and release to the exergonic proton import reaction. Moreover, protonation may easily change the substrate binding site between negatively charged and neutral (HME-RND proteins) or hydrophilic and hydrophobic (HAE-RND proteins). The substrates bound initially to some substrate binding site of the RND protein are consecutively forwarded to the other two components of the CBA efflux system for further transport.
Proposed cellular function and mechanism of RND proteins. This model is based on the structure of AcrB, TolC and published biochemical data (see text for details). The RND protein (red) is shown as a trimer with a periplasmic substrate binding site. The MFP protein is shown in green, the OMF in yellow. Action by RND proteins is driven by protons (black dots) that were previously expelled by the respiratory chain (1). In the case of HME1 and HME2 proteins, divalent heavy metal cations (blue, purple and gray circles, M2+) migrate through the cation channel of the protein and are bound to a periplasmic substrate binding site (orange box, 2a). Alternatively (blue arrows), this site may also be accessible for periplasmic cations (2b), especially in the case of Cu+ and HME4 proteins. HAE transporter of organic substrates may bind these substrates from the hydrophobic environment of the cytoplasmic membrane (2c). Affinity of the hypothetical periplasmic binding site is decreased by protonation with periplasmic protons (3) and this energy is used to accumulate the substrates into the OMF tube or a microenvironment formed by the RND protein (4). The high concentration in this tube allows easy diffusion of the substrates from the tube to the outside (5). For energetic coupling, the hypothetical periplasmic substrate binding site is deprotonated by import of the protons to the cytoplasm (6). This complex reaction allows transport of substrates across the complete envelope of a Gram-negative bacterium that is composed of outer membrane (OM), cell wall (CW), periplasmic space (PP) and cytoplasmic membrane (CPM).
A lot of insight into the possible mechanisms of action has come from the recently available three-dimensional structure of the AcrB protein obtained after crystallization [151, 152]. Three AcrB protomers are organized as a homotrimer in the shape of a jellyfish. Each protomer is composed of a transmembrane region 5 nm thick, composed of the previously identified 12 transmembrane spans, and a 7-nm protruding headpiece containing a central cavity at the bottom. The cavity has three vestibules at the side of the headpiece that may lead into the periplasm. This structure implies that substrates may be translocated from the cell interior through the transmembrane region and/or from the periplasm through the vestibules and are collected in the central cavity , before they are forwarded to the other two components of the CBA efflux complex.
2.6 MFPs and OMFs enhance the function of RND proteins
Most RND proteins are encoded in operons that additionally contain the gene for a MFP. Many of these operons harbor a third gene, encoding an OMF . The position of this OMF-encoding gene varies from operon to operon. It may be located at the 5′ end of operons like czc, at the 3′ end of the operon as in most mex operons, it could be not even part of the operon as in the case of tolC.
Cross-linking experiments with the CzcCBA efflux complex displayed a CzcB multimer . This protein migrates in polyacrylamide gels at a much lower apparent molecular mass than predicted from the derived amino acid sequence. This unusual migration behavior may be based on the extensive coil-coil structure proposed for MFPs . Depending on which size for the CzcB monomer is used in the calculation, the CzcB multimer is a trimer or a tetramer of the MFP . For the MFP AcrA, a trimeric form has also been determined in cross-linking experiments .
In many CBA protein complexes, loss of the MFP or the RND protein abolishes the respective resistance [55, 90, 94, 97, 154, 155]. In contrast, loss of the OMF usually has a relatively moderate influence [55, 97]. Since many Gram-negative bacteria contain a variety of OMF-encoding genes, another OMF may take over and complement for the loss of the native OMF. TolC is known to interact with a variety of transenvelope efflux systems that are driven by RND proteins or even ABC export systems . The CnrC protein could be substituted partially in artificially mixed operons with CzcC or NccC (G. Grass and D.H. Nies, unpublished). Chimeric MexCBA efflux pumps were also functional [156–158]. Thus, the OMF part of a CBA efflux complex is not involved in the substrate specificity, although its overproduction might lead to increased resistance to antibiotics .
TolC is the first OMF for which the structure is known at high resolution . A TolC homotrimer forms a tube-like channel with a length of 14 nm and an outer diameter of 3.5 nm. This tube is hollow, allowing substrates to pass. The tube is sufficient in length to span the entire periplasmic space and the outer membrane, connecting the periplasmic domains of the RND protein with the outside. Insertion in the outer membrane and the ability to form a channel has been determined for OprM and TolC [48, 161, 162]. OprM formed a channel for small molecules, but was only able to allow transport of antibiotics after protease treatment .
This may indicate that the iris-like structure at the periplasmic end of the OMF tube  could serve as a gate, controlling access of the substrates from the RND protein. The top of the headpiece of AcrB opens like a funnel, where TolC might directly dock into AcrB . This funnel was connected to the central cavity of AcrB by a pore that may allow substrate transport from the central cavity to TolC. All these pores and cavities in RND and OMF proteins may serve as microenvironments explaining a common mechanism of action of CBA efflux systems: an active accumulation of a substrate into a tube or cavity may yield a high substrate concentration inside this microenvironment allowing a rapid discharge of the substrate to the outside by diffusion (Fig. 3).
MFPs interact with their RND proteins and are attached to the cytoplasmic membrane by a hydrophobic residue, some by a hydrophobic amino-terminus [53, 164]. MFPs contain a flexible coil-coil structure that may allow a flexible contact between the RND pump and the OMF tube . A recent structure proposal based on liquid layer crystallization  describes AcrA as a doughnut-like structure with an inner diameter of 3 nm. Andersen et al. have suggested that MFPs may serve as adapter proteins between RND or other export pumps and OMF tubes . MFPs probably do not fuse the cytoplasmic and the outer membrane directly, but they attach the OMFs to the pump proteins (Fig. 3).
Thus, OMF and MFP proteins have a rather static function during CBA-mediated transenvelope efflux. Recent findings indicate that the large periplasmic loops (LPL) of HAE-RND proteins have a function in substrate recognition. When the LPLs of AcrD were exchanged with the corresponding LPLs from AcrB, the chimeric protein displayed the substrate specificity of AcrB rather than AcrD . Similar results were obtained when chimeric AcrB–MexB proteins were constructed . Unfortunately, chimeric CzcA–CnrA proteins were not stable (K. Helbig, A. Legatzki and D.H. Nies, unpublished) and similar experiments could not be performed for these HME-RNDs.
Methionine residues in the LPL of CusA were essential for CusA function as mentioned above . Similarly, essential amino acids probably involved in substrate recognition were identified in both LPLs of MexD . Moreover, the LPLs of MexF were also involved in correct assembly of this CBA efflux pump , assigning a second essential function to the headpiece of RND proteins.
Taken together, the RND proteins are indeed the central and most important component of CBA efflux systems: they mediate the active part of the transport process, determine the substrate specificity and are involved in the assembly of the transenvelope protein complex. Thus, distribution of RND proteins in the kingdoms of life may give some clues about the distribution and frequency of RND-mediated metal resistance.
2.7 Distribution of RND proteins in the kingdoms of life
The genomes of 63 prokaryotes from both kingdoms (Archaea and Bacteria, http://tigrblast.tigr.org/cmr-blast, ) were compared to the genome of R. metallidurans (http://www.jgi.doe.gov/JGI_microbial/html/ralstonia/ralston_homepage.html, Table 1). All RND proteins with sufficient homology to CzcA (E value <0.01 in the E value of the BLAST score) were selected and compared to CzcA and the 50 protein sequences in the databases (1.034.995 sequences) most similar to CzcA. Multiple alignments yielded homology trees (data not shown) that were used to group and subgroup the RND proteins (Table 2). Moreover, the occurrence of the Czc-, Cus-, and Mex-specific consensus motifs DFG/D/EN, AIA/D/EN and DD, respectively (see above), was used to confirm the group assignments (Table 2).
Distribution of members of heavy metal-exporting protein families in the genomes of 64 prokaryotesa
Other transport systemsd
2: Cu1, Me0
2: Cu1, Cu16
2: 2a, 3aii
2: 3f, 3f
3: Cu9, Cu12, ff
4: Cu2, Cu14, Zn1, ff
2: 3f, 3f
6: Cu9, Cu12, Zn1, ff
2: MgtA, ff
2: 2c, 3f
3: Cu1, KdpB, ff
2: Cu1, KdpB
4: M3b, M4, A5ff
2: 2g, 3ai
3: Cu1, Cu12, ff
2: 2c, 3e
2: 2h, 3c
3: Cu3, Zn2, KdpB
6: M5, A4, A5ff
9: Cu5, Cu10, Zn7, Zn9, ff
8: M5, M5, A1, A4, A5ff
3: 1b, 2e, 3ai
13: Cu2, Cu5, Cu5, Cu10, Zn4, Zn9, Me1, Me1, ff
4: 2g, 3e, 3e, 3f
4: Cu6, Zn7, Zn10, ff
2: 3f, 3f
4: Cu6, Zn5, Zn10, ff
3: 2g, 3c, 3e
6: Cu6, Zn10, KdpG, ff
3: 2g, 3c, 3e
7: Cu6, Zn7, Zn10, KdpG, ff
3: 2f, 2h, 3e
2: Cu6, KdpG
5: Cu9, Zn7, ff
13: Cu9, Cu12, Zn7, Zn10, Zn12, ff
2: 2c, 3e
4: Cu2, Me1, ff
2: 2c, 3e
3: Cu9, Zn10, ff
2: 2c, 3e
9: Cu9, Cu12, Zn8, Zn12, ff
12: Cu2, Cu2, Cu2, Zn9, Zn9, Zn12, Zn12, ff
4: Cu2, Cu2, Zn12, ff
7: M1, M2, A3, A5ff
2: 2e, 3b
3: Cu14, Zn9, ff
2: 2h, 3c
5: Cu3, Cu14, Zn2, Zn9, ff
2: 2g, 3c
7: Cu4, Cu14, Zn2, Zn12, ff
10: A3, A5ff
2: 1a, 3c
5: Cu3, Cu3, Cu14, Zn2, ff
11: A3, A5ff
2: 1a, 3c
9: Cu3, Cu3, Cu4, Cu14, Cu14, Zn2, Zn12, KdpG, ff
14: M4, A1, A4, A5ff
4: 1a, 3f, 3f, 3f
2: Cu7, Cu14
2: Cu8, Cu14
15: M1, M3a, M3b, M4, A3, A5ff
2: 1a, 3f
4: Cu7, Cu14, Zn6, ff
CHR2, CHR5, CnrT
20: 2×M1, 2×M2, 3×M3a, 3×M3b, 2×M4, A5ff
3: 1a, 2h, 3f
10: Zn6, Zn6, Zn6, Cu4, Cu7, KdpG, ff
CHR1, 2×CHR2, CHR5, CnrT
2: 2c, 3b
3: Cu15, Cu15, KdpG
3: M3b, M4, A5
3: Cu8, Cu14, Zn7
7: M4, A5ff
2: 2g, 3b
4: Cu8, Zn3, KdpG, MgtA
2: 2g, 3b
6: Cu3, Cu8, Zn3, KdpG, ff
2: 2g, 3b
5: Cu8, Zn3, ff
7: A1, A5ff
2: 2g, 3b
5: Cu8, Zn3, ff
2: 2g, 3b
13: M1, A5ff
3: 1a, 2h, 3b
7: Cu7, Cu14, Zn6, Zn12, ff
7: A1, A1, A5ff
3: Cu8, Cu14, Zn3
5: A1, A5ff
2: M5, A4
2: 2b, 3f
Green sulfur bacteria
3: 2b, 3ai, 3aii
aThe comprehensive microbial resource at http://www.tigr.org and the R. metallidurans sequencing homepage of DOE/JGI at http://www.jgi.doe.gov/JGI_microbial/html/ralstonia/ralston_homepage.html was used. A BLAST search [280–282] was performed against the ‘hook’ proteins CzcA from R. metallidurans (RND family), CzcD from R. metallidurans (CDF family), CadA from S. aureus (CPx type ATPases), ChrA from R. metallidurans (CHR family), NreB from R. metallidurans strain 31A and CnrT from R. metallidurans. Sequences with an E value smaller than 0.01 were collected and analyzed in two ways. Firstly, the occurrence of specific signatures was used to pre-group the proteins into putative functional clusters. Secondly, assignment of a sequence to a group of proteins was cross-checked by multiple alignments performed with BioTechnix3d (http://www.biotechnix3d.com). For RND proteins, sequences with low similarity to CzcA and a signature of a RND protein (‘DD motif’) were not further analyzed. Likewise, analysis of a single genome was not further pursued when a magnesium- or potassium-transporting P-type ATPase was encountered. ‘ff’ indicates the presence of such non-characterized RND- and P-type proteins that were non-RND and non-CPx-type proteins, respectively. For RND-, CDF-, and P-type proteins, the total number of the respective proteins encoded by the particular genome is listed. After a colon, the association of the single proteins to a group within the protein family is given (see below). For CHR proteins, the number of these proteins is given and the assignment to the five groups of these proteins (similarity tree not shown). CnrT and NreB proteins were only mentioned. A zero (0) indicates that the organism does not contain any protein from the respective protein group or family.
bFor RND proteins, ‘M’ refers to the respective cluster of the HME-RND proteins, ‘A’ to the respective cluster of the HAE-RND proteins (Table 2).
cLikewise, the cluster assignments of the specific CDF proteins (Table 3) is indicated.
dThe data for the P-type ATPases (Cu, Cu-CPx-type ATPase; Zn, Zn-CPx-type ATPase. MeO, MgtA and KdpG are other P-type proteins, the number after ‘Cu’ or ‘Zn’ indicates the assignment to a group within the respective subfamily and a possible fine differentiation of function) and the other protein groups are not shown.
aFor better orientation, the amino acid residues corresponding to G404, D408, E415, and N416 of CzcA are shown in bold. The dash corresponds to three mostly hydrophobic amino acid residues. Amino acids shown in parentheses are alternatives for the same position. Amino acid names in lower-case letters indicate rare choices, X is any amino acid.
bRepresentative sequences are from R. metallidurans strains CH34 (CzcA, CnrA) and 31A (NccA), Ralstonia solanacearum (RSA1040), Pseudomonas fluorescens (CztA), Legionella pneumophila (HelA), Helicobacter pylori (HP0969), Salmonella enterica pv. typhimurium (SilA), E. coli (CusA, AcrB), Nostoc sp. (All7631, Alr1656, Alr5294), Bacillus subtilis (YerP), and S. meliloti (NolG).
HME- or HAE-RND proteins were found in low frequency in early-branching bacteria (Table 1), but Aquifex aeolicus contains members of both RND protein families. RND proteins are very rare in flexible bacteria. Since spirochetes contain members of the unusual HAE3 family , this observation may indicate a limitation of the method used. Gram-positive bacteria are the third bacterial group with a low frequency of RND proteins (Table 1). Given the current hypothesis about the function of RND proteins as transenvelope efflux pumps, a CBA efflux complex should not function in a Gram-positive cell wall. However, CzcA alone mediates a small degree of resistance , thus, RND proteins in Gram-positive bacteria might function as single-subunit efflux systems. All RND proteins found in Gram-positive bacteria grouped in the HAE2 cluster (identical with the HAE2 protein family ) that appeared only in these bacteria and in Borrelia burgdorferi. This indicates that the mechanism of action of the RND proteins in Gram-positives might be different from that in most Gram-negative bacteria.
Cyanobacteria and proteobacteria contained many RND proteins from both families (Table 1). Exceptions are the symbiotic bacterium Buchnera with no RND protein and three pathogenic bacteria with only one RND (Rickettsia prowazekii, H. influenzae, Pasteurella multocida). Most RND proteins were found in R. metallidurans strain CH34 (20 putative or candidate genes, some of them not complete), followed by its phytopathogenic relative Ralstonia solanacearum (15 genes), Magnetococcus (14 genes) and P. aeruginosa (13 genes). Thus, R. metallidurans contains more RND proteins than every other bacterium in the analyzed group.
Most of the many RND proteins in cyanobacteria and proteobacteria belong to the HAE family. When bacteria contain only a few RND proteins, these are HAE-RND proteins in most cases. Examples of bacteria that exclusively harbored HAE-RND proteins are H. influenzae (with one RND protein that has already been characterized ), R. prowazekii, C. jejuni, P. multocida, Thermotoga maritima, Deinococcus radiodurans, B. burgdorferi, and Gram-positive bacteria.
When bacteria contain putative HME-RND proteins at all, most of these are putative copper/silver exporters of the HME4 group (Tables 1 and 2). Examples are A. aeolicus (two genes, HME3b and HME4), Magnetococcus (HME4), Helicobacter pylori (HME3b and HME4) and E. coli (HME4=CusA). A second group of bacteria harbored members of the highly unusual HME5 cluster, the cyanobacteria Synechocystis (one HME5 gene), Nostoc (two HME5 genes), Legionella pneumophila, and P. gingivalis (one HME5 gene). The HME5 cluster contains members of the HME-RND protein family by homology. These proteins exhibit the ‘DD’ motif in the TMH IV that is characteristic of HAE proteins, but contain some conserved polar residues in their TMH IV unique to this group (Fig. 2).
There is only indirect evidence that these HME5 proteins might transport heavy metals, especially Ni2+. A putative HME5 protein (sequence slr0794, http://www.kazusa.or.jp/cyano/Synechocystis) is part of an operon designated nrsBACD. Expression of this operon is controlled by nickel via a two-component regulatory system NrsRS (=RppAB) encoded upstream of nrsBACD on the other DNA strand. The genes in the nrsBACD operon refer to the genomic designations slr0793 (MFP), slr0794 (HME5), slr0795 (hypothetical protein), and slr0796 (putative NreB protein). It has been shown for NreB-like proteins that they are able to exert nickel resistance alone (, see below), so the nickel resistance assigned to the nrsBACD may be mediated by nrsD. It is not clear if nrsC is an OMF and if a complete CBA efflux complex can be encoded by this operon or parts of it. Therefore, confirmation of the HME5 proteins as nickel-exporting RND proteins has to await further experimental evidence.
A third group of bacteria that contain putative HME-RND proteins may harbor proteins of the HME1 cluster highly related to CzcA. In the analyzed genomes and the current databases, 10 HME1 sequences were found. These were proteins from R. metallidurans (CzcA from different strains [55, 56]), P. aeruginosa (CzrA ), Pseudomonas fluorescens (CztA), a fragment from R. metallidurans (gene 848 on contig 456), R. solanacearum (Rsp 0493), Caulobacter (CC2390), two proteins from either pathovar of Xanthomonas campestris and HelA from L. pneumophila. Only these eight bacteria can be expected to exhibit Co2+/Zn2+/Cd2+ resistance based on RND-driven transenvelope efflux.
The Cnr-related HME2 group is much smaller and contains only four members, two of them from R. metallidurans strain CH34 (CnrA and NccA), one from strain 31A (NccA ) and one from Caulobacter (CC2724). These three bacteria should exhibit RND-based nickel/cobalt/cadmium resistance.
Group HME4 contains 10 sequences of putative copper/silver transport systems. These sequences came from E. coli (CusA [90, 94, 97]), Salmonella enterica pv. typhimurium (SilA ), two sequences from R. metallidurans and one sequence from R. solanacearum, Pseudomonas syringae, H. pylori, L. pneumophila, A. aeolicus and Magnetococcus sp. CusA has been introduced above and by Grass and Rensing , SilA by Silver . The H. pylori copper export system has also been characterized . Because of the similarity of the other eight sequences to CusA and SilA, these bacteria can be expected to detoxify copper and/or silver in a way that has been shown for CusA and SilA.
The remaining HME3 cluster contains all sequences not closely associated with HME1, HME2 or HME4, but with a unique DSS motif in TMH IV (Fig. 2). There are two subgroups in this cluster (Table 2). Subgroup HME3a exhibits the DFG consensus motif and the proteins in this subgroup may transport divalent heavy metal cations. The five members of HME3a come from R. solanacearum, Xanthomonas axonopolis pv. citri, and three sequences from R. metallidurans strain CH34. The other subgroup, HME3b, shows mostly the AIG motif characteristic of transporters of monovalent cations. The five proteins of this cluster come from R. solanacearum, H. pylori, A. aeolicus and again three sequences from R. metallidurans. The sequence from A. aeolicus is unusual since it contains a GIG motif instead of AIG, which gives this putative protein (aq_469) an ancient touch.
The relationship of RND proteins leads to a scenario of how heavy metal-transporting HME proteins may have evolved: the earliest branch of HME proteins are the HME5 proteins that still contain the DD motif of RND proteins, and which are much more diverse and widely distributed than HME proteins. Have HME proteins originated from HAE proteins? The other four groups, HME1 to HME4, are currently found only in proteobacteria and A. aeolicus. This could indicate that the divalent heavy metal transport systems originated from an ancient monovalent transport system such as aq_469 from A. aeolicus during evolution of the proteobacteria.
3 The second layer of metal resistance: cation diffusion facilitators (CDF family) serve as secondary cation filters in bacteria
CDF proteins form a protein family of metal transporters (TC 2.A.4.1.1–2 ) occurring in all three domains of life . Primary substrates of CDF proteins are Zn2+, but also Co2+, Ni2+, Cd2+, and Fe2+. CDF-mediated transport is driven by a concentration gradient, a chemiosmotic gradient, ΔΨ, ΔpH or a potassium gradient [176–178] (A. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished). First, the physiological function, biochemistry and substrate specificities of the CDF protein family will be analyzed, before the distribution of CDF proteins in sequenced bacterial genomes will be compared to that of R. metallidurans strain CH34.
3.1 Physiological function and substrate specificity of CDF proteins
All CDF proteins described so far in bacteria are involved in resistance to Zn2+ and other heavy metal cations. The archetype of the family, CzcD from R. metallidurans strain CH34, was first described as a regulator of expression of the CzcCBA high resistance system , but CzcD is also able to mediate a small degree of Zn2+/Co2+/Cd2+ resistance in the absence of the high resistance CzcCBA system . The presence of CzcD decreases the cytoplasmic concentration of these metals  which serve as inducers for czc expression (C. Große, A. Anton, T. Hoffmann, S. Franke, G. Schleuder and D.H. Nies, unpublished). Heterologously expressed in E. coli, CzcD was driven by the proton motive force and effluxes Zn2+ and Cd2+, but not Ni2+ and Co2+ (. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished). Cross-linking experiments indicated that CzcD is probably a dimer in vivo.
Although CDF-encoding genes are found on the chromosomes of many sequenced organisms, the function of the proteins has only been characterized in a few bacteria. In Bacillus subtilis, the czcD gene is in an operon together with the gene for a TrkA dehydrogenase  and mutant studies show some involvement in cobalt resistance . CzcD from B. subtilis was driven by the proton motive force as well as by a potassium gradient, and the exchange seemed to be electroneutral. Substrates for this CzcD protein were Zn2+, Co2+ and Cd2+.
Two groups cloned from Staphylococcus aureus a small determinant containing a CDF-encoding gene and a regulator. The CDF protein was named ZntA (now RzcB ) or CzrB . Both proteins are nearly the same size (325 or 326 aa) and differ only in nine positions; they are essentially identical. CzcD from S. aureus is involved in zinc and cobalt resistance. A CDF protein from Thermus thermophilus was shown to mediate zinc and cadmium, but not cobalt resistance .
Recently, one of the two CDF proteins from E. coli was characterized, ZitB (the ybgR gene product), which is also a zinc resistance system . This protein diminishes the cellular accumulation of Zn2+ and is driven by a potassium gradient in addition to the proton motive force . E. coli contains another putative CDF-encoding gene, the enigmatic yiiP. Expression of this protein is induced by zinc, deletion of the gene does not lead to an increase in zinc resistance, but overexpression has no effect . The function of this protein is therefore unknown.
The bacterial CDF proteins mentioned so far are mainly Zn2+/Co2+/Cd2+-exporting cytoplasmic membrane proteins. Magnetotactic bacteria are able to sense the magnetic field of the Earth with the help of specific intracellular organelles, the magnetosomes, which contain the mineral magnetite (Fe3O4) as field-sensing material. MamB was described as a CDF protein located in the magnetosome membrane . Its function is probably to transport iron into this organelle. Magnetosome-like magnetite was discussed as a marker for remnants of ancient life forms from the planet Mars [187, 188]. That would increase the distribution of CDF proteins to another planet! However, a magnetic field is the prerequisite for the occurrence of magnetotactic life forms. Since the magnetic field of Mars was only 1/30 as strong as the field of our Earth when the putative Martian fossils were liberated by a giant impact  and this is rarely sufficient to drive any magnetotactic behavior, we probably have to wait for better evidence for life forms and CDF proteins from other planets.
Usually, bacteria contain one or a few CDF-encoding genes. In eukaryotes, the number of CDF-encoding genes increases in parallel with evolution from yeasts to animals and plants. Saccharomyces cerevisiae contains five CDF-encoding genes . ZRC1 was initially described in the same year as CzcD and this protein mediates resistance to zinc and cadmium . Another yeast CDF, COT1, mediates cobalt resistance , but ZRC1 and COT1 can complement each other . Both can even substitute for CzcD in its transport and regulatory function in R. metallidurans. ZRC1 and COT1 are located in the vacuolar membrane and transport the cations Zn2+, Cd2+ and Co2+ into this compartment for detoxification and possible future re-use [194–196]. This explains why both proteins are also needed to survive zinc starvation [194, 197, 198]. Moreover, since the cytoplasm is not detoxified in ZRC1 knock-out strains, the cell reacts with an increased glutathione level . Similar to the function of CzcD from R. metallidurans, the postulated regulatory function of ZRC1 is therefore an indirect consequence of its transport function.
A third S. cerevisiae protein, MSC2 (YDR205W), is also involved in zinc transport . The respective deletion strain, however, accumulated excess zinc in the nucleus-enriched membrane fractions. Therefore, a function of MSC2 in zinc homeostasis of the nucleus was suggested . The other two CDF proteins from S. cerevisiae are not involved in Zn2+/Cd2+/Co2+ transport, but these two proteins, MFT1 and MFT2, are probably Fe2+ transport systems located in the mitochondrial membrane . That adds to the evidence that CDF proteins may transport Fe2+.
The ZRC1-related CDF protein Zhf from the fission yeast Schizosaccharomyces pombe is located in the endoplasmic reticulum and the nuclear envelope like MSC2 . Deletion of zhf leads to increased zinc and cobalt sensitivity, but to increased cadmium and nickel resistance. The cytoplasmic content of zinc, cobalt and cadmium was diminished and detoxification of cadmium by phytochelatin synthesis was not increased . Thus, this protein may be responsible for supply of the endoplasmic reticulum and the nucleus with zinc and/or cobalt. Involuntary transport of cadmium and nickel into the same compartment by Zhf could mediate increased toxicity of these cations.
The currently known CDF proteins in higher animals are all zinc transport systems. ZnT1 is located in the cytoplasmic membrane and seems to be a zinc efflux system . In contrast, ZnT2 transports Zn2+ into vesicles and is therefore comparable in function to ZRC1 and COT1 . ZnT3 also transports Zn2+ into vesicles, but exclusively into synaptic vesicles . The protein is most abundant in the zinc-enriched mossy fibers that project from the dentate granule cells to hilar and CA3 pyramidal neurons in the brain. Thus, ZnT3 seems to be responsible for the ability of these neurons to release zinc upon excitation . Age-dependent low expression of ZnT3 leads to deficiencies of Zn2+ in synaptic vesicles of the mossy fiber pathway and induces glutamatergic excitotoxicity in the hippocampal neurons and the deterioration of learning and memory .
The fourth known CDF protein from higher animals, ZnT4, is similar to ZnT3 with respect to zinc transport and abundant expression in the brain, but this protein is also expressed in the mammary epithelia and responsible for excretion of zinc into mother's milk . The CDF family in mammals kept growing with ZnT5 that transports zinc into secretory granules in pancreatic β cells  and also plays an important role in maturation of osteoblasts and in maintenance of the cells involved in the cardiac conduction system . ZnT5 proteins from mouse and man are twice the size of other CDF proteins. Since dimerization has been indicated for CzcD (A. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished) and ZnT4 , gene duplication followed by gene fusion could have been responsible for the occurrence of ZnT5 and other large CDF proteins.
Another large CDF protein from mammals was named ZTL1 (ZnT-like protein). It is located in the enterocyte apical membrane, e.g. in kidney . Comparison of the 765-aa ZnT5 protein with the 523-aa ZTL1 proteins displays a 495-aa middle part that differs in only two amino acid residues (99.6% identity). This middle part starts at M172 of ZnT5 that equals M1 of ZTL1, and extends to K666 of ZnT5. The 100-aa ZnT5 and 28-aa ZTL1 carboxy-termini are completely different. The genomic sequence encoding the ZTL1 protein consists of 11 exons separated by large introns with the first ones encoding possible alternative splicing . The ZnT5 protein has a predicted size of 84 kDa, but its size in vivo was determined by Western blot analysis to be 55 kDa , which is similar to the predicted size of ZTL1 of 57.3 kDa. Thus, the original gene may encode at least two proteins by splicing of the RNA and post-translational processing of a pre-protein, and the two proteins may differ in function and localization.
Mammals contain at least two more CDF proteins, ZnT6 and ZnT7. The protein ZnT7 is currently only putative (AY094606) awaiting publication of more data. ZnT6 may transport cytoplasmic zinc into the Golgi apparatus and its expression is regulated by zinc in kidney cells . All these results suggest that intracellular zinc homeostasis in mammals is mediated by a variety of ZnT-like proteins that differ in intracellular localization and organ-specific expression. These proteins transport zinc from the cytoplasm into another compartment, to the outside (ZnT1, ZnT4) like bacterial transport systems, or into vesicles (ZnT2, ZnT3, ZnT5) like ZRC1 and COT1, or into the Golgi apparatus (ZnT6) like Zhf. However, experiments with oocytes expressing ZTL1 indicate that CDF proteins may also be involved in uptake into the cell .
The first CDF protein from plants was ZAT . Arabidopsis thaliana plants overexpressing ZAT (AtCDF1) exhibited enhanced Zn2+ resistance and strongly increased Zn2+ content in the roots under high Zn2+ exposure. Antisense mRNA-producing plants were viable, with a wild-type level of Zn2+ resistance and content . The ZAT protein was purified and functioned as a zinc uptake system in several microbial model systems . ZAT did not transport other heavy metal cations with the exception of a slow transport of cadmium. A. thaliana contains at least seven more CDF proteins . As in mammals, a variety of CDF proteins may also be responsible for zinc homeostasis in plants.
A ZAT-like gene was also cloned from the heavy metal hyperaccumulator plant Thlaspi caerulescens. From the related plant T. goesingense, a CDF-encoding gene with interesting properties was cloned. Transcription of the MTP1 gene from this plant leads to a spliced and an unspliced mRNA. Both mRNAs encode proteins which differ within a histidine-rich putative metal binding domain and in their substrate specificity . The protein translated from the unspliced mRNA confers high level tolerance to Cd2+, Co2+ and Zn2+, whereas the product of the spliced mRNA confers nickel resistance . This is reminiscent of the ZnT5/ZTL1 situation in animals and of nickel as a possible substrate of Zhf.
3.2 Biochemistry of CDF proteins
With transport of Ni2+ by the T. goesingense protein, the substrate spectrum of CDF proteins includes Fe2+, Co2+, Ni2+, Zn2+ and Cd2+. The ionic radii of these cations are 76 pm, 74 pm, 72 pm, 74 pm and 97 pm, respectively . Of the other divalent heavy metal cations of biological importance, Mn2+ (80 pm) may also fall in the size range of possible CDF substrates. In contrast, Mg2+ (65 pm) and Cu2+ (69 pm) may be too small and Ca2+ (99 pm) and Hg2+ (110 pm) too big to serve as CDF substrates in general. Additionally, affinity for a substrate binding site should also differ between substrate cations. The cations of the first transition period usually have a higher affinity for free electron pairs of oxygen and nitrogen atoms, whereas the cations of the second and third transition periods prefer the ‘soft’ electrons of sulfur atoms. The majority of the CDF substrates are divalent metal cations of the first transition period with sizes of 74±2 pm diameter with Cd2+ being a piggy-back rider due to its chemical similarity to Zn2+.
Most CDF proteins have a size of 300–400 aa. CzcD from R. metallidurans exhibits six transmembrane spans  and this can also be assumed as the rule for most other CDF proteins. CDF proteins contain many histidine residues that are located at the amino-terminus, the carboxy-terminus, or between transmembrane helices IV and V, or at all three locations. Histidine residues of CzcD bound two to three Zn2+ per CzcD polypeptide (A. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished) and were also essential for transport of zinc by ZitB from E. coli. Although the histidine-rich carboxy-termini of CzcD and ZitB were not essential for transport (A. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished) , they could have a function as regulators of transport activity of these proteins.
The CzcDs from B. subtilis and R. metallidurans and ZitB from E. coli are efflux pumps driven by the proton motive force or a potassium gradient (A. Anton, G. Grass, C. Rensing and D.H. Nies, unpublished) [176, 177, 180, 185] and function of ZRC1 from S. cerevisiae also required a proton gradient . In eukaryotic cells, however, not all membranes harbor a proton, sodium or potassium gradient. Therefore, some CDF proteins may function as simple diffusion facilitators in these organisms. Cation transport by the ZAT protein from A. thaliana was not stimulated in proteoliposomes by a proton gradient. Moreover, ZAT functioned as an uptake system in several microbial cells . Therefore, cation transport by CDF proteins could be either an antiport reaction against a chemiosmotic gradient or a facilitated diffusion driven by the concentration gradient of the respective cation.
3.3 Distribution of CDF proteins in the living world
In a similar manner to the RND proteins, the distribution of CDF proteins in sequenced bacterial genomes was analyzed (Table 1). The proteins in these genomes and in current databases were compared in multiple alignments, and used to group and subgroup these proteins (Table 3). Three groups of CDF proteins could be established. Group 1 contained two subgroups, 1a and 1b. Cluster 1b is composed of the large proteins ZnT5, ZTL1 and MSC2, together with three other sequences. This assignment may be an artefact of their size and the multiple cluster algorithm used. Cluster 1a, on the other hand, consists of six sequences of putative proteins from proteobacteria significantly related to each other. This might indicate a common mechanism of function.
aThe number of question marks after the predicted substrates indicates the probability of this prediction; the more question marks, the lower the probability for the putative substrate.
bRepresentative sequences are from P. aeruginosa (PA1297), S. cerevisiae (MSC2), Pyrobaculum aerophilum (PAE2715), Streptococcus pneumoniae (Spr1672), Pyrococcus horikoshii (PH0896), A. thaliana (ZAT), mammals (ZnT2, ZnT3, ZnT5, ZnT7, ZTL1), Nostoc (All7610), S. aureus (SA0155), E. coli (ZitB, YiiP), B. subtilis (CzcD_Bacsu, YeaB, YbdO), R. metallidurans (CzcD_Ralme), A. aeolicus (Aq_2073), D. radiodurans (DR1236), Enterococcus faecalis (EF0859), Magnetococcus gryphiswaldensis (MamB).
The remaining CDF proteins fall into two groups, each containing several subgroups. Group 2 contains most of the known zinc-transporting CDF proteins in bacteria (CzcDs, ZitB), eukaryotes and archaea (Table 3). Cluster 2d occurs exclusively in eukaryotes and also contains the nickel-transporting plant protein that is highly related to ZAT from A. thaliana. Group 2a, which branches off first, is found in Crenarchaeota (Table 3). This might indicate that the mainly zinc-transporting CDF proteins may have originated very early in evolution.
Group 3 is assembled around YiiP from E. coli. No data are available for proteins in this group with the exception of MamB, the probable Fe2+-transporting CDF protein from magnetosome membranes of magnetotactic bacteria . In aerobic environments, iron occurs mainly as the insoluble Fe3+, however, in anaerobic or micro-aerobic environments, a considerable concentration of Fe2+ may be present. Fe2+ can also be a substrate of CorA-like magnesium uptake systems . Similar to Zn2+, a second-step filtering of Fe2+ against Mg2+ could be required, which yields a speculative initial idea for the function of the group 3 CDF proteins.
When the distribution of CDF proteins is analyzed (Table 1), most prokaryotes contain two putative CDF proteins, one from group 2 and one from group 3. Of the 64 genomes analyzed, 24 contained two putative CDF proteins, 15 one, 16 none, seven three and only two contained four. Prokaryotes with two putative CDF proteins always harbored one member of group 3 plus a putative CDF from group 2 (75%), group 1 (12.5%), or a second protein from group 3 (also 12.5%). Thus, the general outfit of a bacterium is one group 3 and one group 2 CDF protein. The prokaryotes with three or four CDF proteins always contained a group 3 protein, exclusively or in addition to a group 2 or a group 1 protein. This could indicate that the gene for a group 3 CDF protein is less likely to have been sacrificed during evolution than that of a group 2 zinc-transporting CDF protein. If group 3 proteins are indeed iron transporters, this may mirror the higher biological importance of iron in comparison to zinc.
4 The third layer of heavy metal resistance: P-type ATPases are the basic defense against heavy metal cations
P-type ATPases constitute a superfamily of transport proteins that are driven by ATP hydrolysis . Members of this family occur in all three kingdoms of life. Substrates are inorganic cations such as H+, Na+, K+, Mg2+, Ca2+, Cu+, Ag+, Zn2+ and Cd2+. An individual P-type ATPase can either import its substrate from the outside or from the periplasm to the cytoplasm (importing P-type ATPase) or export it from the cytoplasm to the outside/periplasm (exporting P-type ATPase). With respect to heavy metal homeostasis, P-type ATPases may be important for two reasons. Firstly, import systems for macroelements such as Mg2+ may also import heavy metals . Secondly, exporting P-type ATPases may detoxify heavy metal cations by efflux.
Members of the huge P-type ATPase superfamily that transport heavy metal cations carry a conserved proline residue, preceded and/or followed by a cysteine residue. This motif is essential for the function of at least one member of this family . Work of the last few years on these ‘CPx-type’ or soft metal-transporting ATPases has been reviewed before  and will be partially covered in other reviews in this issue. A survey of heavy metal cation-transporting CPx-type ATPases in this section will therefore concentrate on data that are important in understanding the interconnections between CPx-type ATPases and members of the other two metal efflux protein families, CDF and RND.
4.1 Families of CPx-type ATPases and their substrates
One family of CPx-type ATPases is involved in export of Cu+ (but not Cu2+) and Ag+ (Cu-CPx-type ATPases [88, 89]. Members of this group were characterized in a variety of organisms, in the archaeon Archaeoglobus fulgidus, the cyanobacterium Synechocystis, the Gram-positive bacteria B. subtilis, Enterococcus faecium, Enterococcus hirae [225–227], Streptococcus mutans, Lactobacillus sakei, the proteobacteria Rhizobium leguminosarum, Sinorhizobium meliloti, P. putida and E. coli. Cu-CPx-type ATPases also occur in lower eukaryotes as Dictyostelium, Cryptosporidium parvum and Candida albicans. Seven of the 45 genes from A. thaliana that may encode P-type ATPases are putative CPx-type ATPases . One of these proteins may transport copper and is required for ethylene signaling in the plant . In mammals including human, miss-expression or mutant proteins of Cu-CPx-type ATPases were assigned to the copper homeostasis defects Menkes’ and Wilson's diseases [237, 238]. This demonstrates a broad distribution of Cu-CPx-type ATPases in the living world.
A second family transports Zn2+, Cd2+ and Pb2+ (Zn-CPx-type ATPases), again, mainly from the cytoplasm to the outside or to the periplasm. The very first member of the CPx-type ATPases described was CadA, a cadmium resistance-mediating protein encoded on a plasmid from S. aureus [13, 239]. Examples from other bacteria such as B. subtilis, Listeria monocytogenes, P. putida, S. maltophilia, E. coli [244–248], and the yeast S. cerevisiae followed, but the distribution of Zn-CPx-ATPases seems to be not as broad as the distribution of the Cu-CPx-type ATPases.
The substrate specificity of CPx-type ATPases is Cu+/Ag+ or Zn2+/Cd2+/Pb2+. A CPx-type ATPase from the cyanobacterium Oscillatoria brevis uses heavy metal cations from either group as substrates . Overexpression of CadA from H. pylori has some effect on cobalt resistance, so Co2+ may also be a substrate of Zn-CPx-type ATPases , as has been suggested before for a P-type ATPase from Synechocystis. There is, however, no clear evidence for the transport of Ni2+ by a CPx-type ATPase. A publication claiming the cloning of a P-type ATPase  involved in nickel resistance deals with a possible nickel uptake protein from a completely different protein family (NiCoT) not known to be involved in metal resistance, but in metal uptake [254–256].
The direction of transport of CPx-type ATPases in bacteria is mostly export to the outside or the periplasm, but the other direction of transport also seems to occur. The first examples of Cu-CPx-ATPases came from the Gram-positive bacterium E. hirae. Two Cu-CPx-type ATPases were found, CopA involved in copper uptake and CopB in copper resistance by efflux [257–261]. These are reviewed in . Physiological evidence indicates that two CPx-type ATPases from the cyanobacterium Synechocystis, CtaA and PacS, may be involved in copper uptake . They facilitate the synthesis of the copper-dependent electron transfer factor plastocyanin , although disruption of pacS decreased copper tolerance. Disruption of ctaA diminished the amount of cell-bound copper , clearly indicating a function of CtaA in copper uptake. There is also evidence from B. subtilis that the Zn-CPx-type ATPase ZosA is a zinc uptake system . It seems to be easy for evolution to change an efflux system into an uptake system when this is required by the specific environment the cell struggles to survive in.
4.2 Structure and function
The membrane topology of CPx-type ATPases was analyzed exemplarily for the first described member, CadA from S. aureus. The protein contains eight transmembrane segments with the conserved, eponymous CPC motif located in the sixth of these segments. The ATP binding domain is situated in a large cytoplasmic domain that follows this transmembrane domain. The amino-terminal 109 aa carry a CXXC motif that is also present in most members of the CPx-ATPases . Although heavy metal cations are able to bind to the amino-terminal domain , this domain is not essential for transport  and a regulatory function has been discussed .
When the HME-RND proteins were discussed above, the problems in detoxifying thiol-bound heavy metal cations were mentioned. The complex binding constant for Cu+–glutathionate complexes is 1039. The energy to remove the Cu+ cation from this complex is ΔG′o=−RT ln1039=−226 kJ mol−1 at 30°C. Thus, to detoxify heavy metal cations bound to thiols, paired cysteine residues in the substrate binding or substrate funneling site are required and additionally a lot of energy. CPx-type ATPases exhibit both features and it has already been demonstrated that the substrates in vivo are likely metal–thiolate complexes rather than the ‘free’ metals .
4.3 Distribution of CPx-type ATPases in prokaryotes
If CPx-type ATPases clean out the thiol-bound heavy metal pool, they should occur even more frequently in the living world than CDF or RND proteins. In a similar manner to the approach described above for the other two protein families, the distribution of putative CPx-type ATPase in prokaryotic genomes was analyzed. A number of 16 groups of Cu-CPx-type ATPases and 12 groups of Zn-CPx-type ATPases were found by multiple alignments (data not shown). Members of the other families of P-type ATPases such as MgtA-like magnesium transporters and KdpG-like potassium transporters were only considered as examples. Generally, the KdpG-like proteins were more related to the CPx-type ATPases than the MgtA-like proteins (data not shown).
Starting from the 64 prokaryotes analyzed, only eight species did not contain a gene for a putative P-type ATPase at all (Table 1). Five species contained at least one P-type ATPase, but none of the CPx-type. Most (29 species) had CPx-type ATPases of both families, the Cu-CPx-type and the Zn-CPx-type ATPases. If prokaryotes harbored only a member of one of these two families, this was mostly a member of the Cu-CPx-type ATPases (18 species) and in just four cases only a member of the Zn-CPx-type ATPases. Thus, nearly half of the species analyzed contained members of both CPx families, which indicates the importance of the resistance mediated by these proteins. As with the RND proteins, export of copper/silver was more important than efflux of zinc/cadmium/lead, mirroring the different toxicities of these cations.
R. metallidurans contains genes or candidate genes for two Cu-CPx-type ATPases, three Zn-CPx-like ATPases and five other P-type ATPases giving a total number of 10 (Table 1). One of the Zn-CPx-type ATPases, PbrA, is part of a plasmid-mediated lead resistance determinant in R. metallidurans, the other two, ZntA and CadA, are involved in resistance to Zn2+ and Cd2+. This complement of CPx-type ATPases is not unusually high for a bacterium; cyanobacteria, mycobacteria or Rhizobium species exhibited similar numbers (Table 1).
5 Other heavy metal export systems: CHR protein family, NreB, CnrT
Before the data about RND, CDF, and CPx proteins are assembled into a picture that describes heavy metal detoxification in prokaryotic cells, three small groups of transporters will be introduced. All three are linked to resistance determinants that may allow a bacterium to survive in the specific nickel-, cobalt- and chromate-rich serpentine environment . ChrA-like proteins of the CHR protein family  detoxify the oxyanion chromate, while NreB and CnrT are nickel efflux systems.
5.1 CHR proteins detoxify chromate
The CHR family contains proteins that are bound to the cytoplasmic membrane by 10 transmembrane spans . The protein probably originated by gene duplication followed by gene fusion. Several bacteria contain operons that comprise two successive genes encoding the amino- and the carboxy-terminal parts, respectively, of a full-length CHR protein. This led to the hypothesis that both halves of the full-length proteins may differ in the mode of their action . CHR proteins are probably chromate efflux pumps [268, 269] driven by the chemiosmotic gradient. Since the export of an anion is in the direction of the electric field of the proton motive force, ΔΨ alone is sufficient to drive chromate efflux.
The CHR proteins of 64 prokaryotic genomes were compared by multiple alignments and grouped into five categories that may mirror differences in the physiological or biochemical modes of action. Only a quarter of the bacterial genomes contained at least one putative CHR member (Table 1). Two CHR-encoding genes were found in R. solanacearum and Bacillus halodurans, four in R. metallidurans. Again, R. metallidurans contained the highest number of members of a protein family involved in heavy metal export.
Of the 21 putative CHR proteins analyzed, involvement in chromate efflux has only been shown for two of the four proteins from R. metallidurans [267, 270–273], the ChrA protein from P. aeruginosa [1, 268, 269, 274], and Synechococcus, a cyanobacterium roughly related to Synechocystis.
5.2 NreB-like proteins
In three bacterial species (Synechocystis sp., Nostoc sp., S. meliloti), the presence of a CHR protein was accompanied by the presence of a NreB-like protein in the genome (Table 1), in Magnetococcus it appeared without a CHR protein. The NreB protein has been characterized in R. metallidurans strain 31A . This bacterium contains two distinct nickel resistance determinants on its native megaplasmids pTOM8 and pTOM9: ncc, which was mentioned above, and nreB. Expression of the nreB gene was specifically induced by nickel and conferred nickel resistance in both strain 31A and E. coli. Cells expressing nreB showed decreased accumulation of Ni2+, suggesting that NreB mediates nickel efflux.
The NreB-like proteins have a size of about 400 aa and contain probably 12 transmembrane spans. One putative protein from the enterobacterium Hafnia alvei is much smaller and may be the result of a sequencing error. The other six NreB-like proteins found in current databases all exhibit a characteristic, histidine-rich carboxy-terminus that resembles a natural His tag.
5.3 CnrT-like proteins
Only three proteins in the analyzed genomes and the current databases were similar to CnrT from R. metallidurans. These were putative proteins from Yersinia pestis, R. solanacearum and M. loti (Table 1). The cnrT gene is located directly downstream of the cnr cobalt–nickel resistance determinant in plasmid pMOL28 of R. metallidurans and encodes a small degree of nickel resistance by itself (G. Grass, S. Kühnemund, B. Fricke, K. Sutter, G. Krauß and D.H. Nies, unpublished) . CnrT-like proteins exhibit 12 or 14 transmembrane spans and, like NreB proteins, a histidine-rich carboxy-terminus.
CnrT- and NreB-like proteins do not exhibit any potential ATP binding sites and are thus probably driven by the proton motive force, like CDF proteins. Moreover, all three groups of proteins contain cytoplasmic loops of the polypeptide chain rich in putative metal binding regions and export their substrates across the cytoplasmic membrane. Therefore, CnrT- and NreB-like proteins may act as secondary filter proteins for Ni2+ similar to CDF proteins that may act as secondary filter proteins for Zn2+, Co2+ and Cd2+.
6 Heavy metal resistance in prokaryotes revisited
Which modes of heavy metal efflux do most bacteria have? Export of zinc and cadmium is accomplished at least by CDF2, Zn-CPx and HME1-RND proteins. Of the 64 prokaryotic genomes analyzed, there are approximately equal numbers of species harboring no system (17 species), only one CDF2 protein (13 species), only one Zn-CPx (15 species), or a CDF2 plus a Zn-CPx protein (13 species). Only a minority of prokaryotes analyzed have all three proteins (three species), an ATPase plus a HME1-RND protein (two species), a CDF2 plus a HME1-RND protein (one species) or exclusively an RND protein (no species). Thus, the presence of a HME1-RND protein is the exception and marks high level resistance to zinc, cobalt and cadmium, while possession of a CDF2 and/or a Zn-CPx-type ATPase is the general case and therefore part of the cellular homeostasis system. Moreover, since no CBA transport system occurs without the presence of a helper protein that imports the cations into the periplasm, this shows that periplasmic access to the CBA pump may also be important for resistance to this group of heavy metal cations. Bacteria relying exclusively on a CDF2 protein for their zinc homeostasis are more adapted to an anaerobic or micro-aerobic environment than to higher oxygen concentrations. The need to remove metals from the thiol pool by a CPx-type ATPase may be smaller when the additional oxidation stress by molecular oxygen is absent.
As far as we know today, CPx-type ATPases or CDF proteins detoxify nickel only in rare cases. RND proteins known to detoxify nickel and cobalt (HME2 group) or candidates for this process (HME5 group) are rare and mark species as adapted to nickel-rich environments. Moreover, NreB- and CnrT-like nickel transport systems occur in only a few species. Nickel is used seldom as a cofactor  and if needed, is imported into the cell by chemiosmotically driven slow and specific uptake systems of the NiCoT family [46, 254–256, 276–278] or ABC transport systems . The concentration of nickel in seawater and its toxicity to E. coli is comparable to that of zinc, but this comparison has to be viewed in the light of two known zinc-exporting systems in E. coli (Table 1) and no known nickel efflux system. The toxicity of nickel seems to be a smaller problem than the toxicity of zinc or, especially, copper.
Copper is mainly detoxified by a Cu-CPx-type ATPase. Of the analyzed 64 genomes, 41 contained the gene for a putative Cu-CPx-ATPase, 17 no such protein, and six an additional HME4-RND protein. The presence of such a RND protein may mark a demand for a supplementary export of the periplasmic copper, maybe under anaerobic conditions . The presence of an RND-driven CBA pump is again an exception, indicating increased heavy metal resistance. However, the highest degree of copper resistance is mediated by plasmid-bound copper resistance determinants [9, 79–81, 83, 84] that are mentioned in the article by Grass and Rensing .
Seventeen species did not contain any known zinc export system and the same number no known copper efflux system (Table 1). The two groups overlap, but not completely; nine species contained neither zinc- nor copper-exporting systems. Four of them, the pathogenic spirochete B. burgdorferi, two species of Rickettsia and the methanogenic archaeon M. jannaschii, contained at least a CDF3 and/or a CHR protein indicating some skills in export of heavy metals: the unknown CDF3 substrate and/or chromate. The remaining five species without known competence to detoxify heavy metals are the cell-wall-free pathogenic Mycoplasma and Ureaplasma species and the aphid symbiont Buchnera. In their pathogenic or symbiotic way of life, these bacteria are probably not confronted with heavy metal stress, rendering metal homeostasis genes superfluous.
7 Concluding remarks: heavy metal resistance is the product of the interaction of many systems
We may now be able to answer the question we started with: what makes a heavy metal resistant bacterium like R. metallidurans heavy metal resistant? Firstly, there is a multitude of resistance systems. When R. metallidurans is compared to R. solanacearum, an increase in the numbers of the CPx-type, CDF and RND proteins is observed (Table 1). These additional genes are the product either of horizontal gene transfer or of a duplication of genes already present in the common Ralstonia ancestor. If the duplicated genes in R. metallidurans were closely related to a respective gene in R. solanacearum, this would indicate a possible gene duplication that had happened during the evolution of the genus R. metallidurans. If this is not the case and if the genes are additionally harbored by a plasmid, this would point towards horizontal gene transfer.
The number of CPx-type ATPases increases from R. solanacearum to R. metallidurans from one to three Zn-CPx proteins, while the number of Cu-CPx-ATPases is constant at two. The four Zn-CPx proteins from both Ralstonia species belong to the same group of proteins, Zn6, which indicates an origin by gene multiplication. CHR proteins duplicated from two to four. R. metallidurans contains one CDF protein more than R. solanacearum. Both species harbor a CDF1 and a CDF3 protein, the additional protein in R. metallidurans is the CDF2 protein CzcD that is encoded by plasmid pMOL30. The number of HME-RND proteins increased from four in R. solanacearum to an outstanding number of 12 in R. metallidurans. When the various clusters of HME proteins are compared, the number of HME1 and HME4 proteins is duplicated, the number of two HME3 proteins in R. solanacearum is even tripled to six proteins in R. metallidurans. In all cases, predicted proteins from R. metallidurans and R. solanacearum group closely with each other indicating an origin from a common ancestral gene. The remaining two HME2 proteins, CnrA and NccA, which occur only in R. metallidurans and not in R. solanacearum, are both encoded on plasmids that may have been acquired by horizontal gene transfer.
Secondly, there is differentiation of the function of metal resistance systems after the duplication event has occurred. The three Zn-CPx-type ATPases of R. metallidurans became lead-, cadmium- and zinc-specific efflux pumps. Thirdly, there is combination of the specialized proteins with others to form resistance systems with increased efficiency. Of the two determinants that encode a HME2 pump, ncc on plasmid pMOL30 of strain CH34 became inactivated while cnr combined with the cnrT gene to form an efficient nickel/cobalt pump. The HME3 proteins that may detoxify copper and silver combined with other copper resistance systems. The additional CDF protein in R. metallidurans, CzcD, became attached to the Czc system and is involved in regulation of expression and provides additionally a minor degree of metal resistance. Other genes also became part of the Czc system that is intriguing and unique in its complexity. The three steps, (1) gene multiplication by duplication and horizontal transfer, (2) differentiation of function and (3) combination of genes into highly efficient operons, created the outstanding heavy metal resistance of R. metallidurans.
This work was funded by the Deutsche Forschungsgemeinschaft, the state Sachsen-Anhalt and by the Fonds der Chemischen Industrie. I am grateful to Chris Rensing, Gregor Grass and Cornelia Große for their comments on this review. The R. metallidurans CH34 web site of DOE/JGI (http://www.jgi.doe.gov/JGI_microbial/html/ralstonia/ralston_homepage.html) and the comprehensive microbial resource http://www.tigr.org/tigr-scripts/CMR2/CMRHomePage.spl of The Institute of Genomic Research (TIGR) were used intensively to compile Table 1. I would also like to thank all the colleagues, mentors and students in my lab who in the last 20 years have contributed to the progress of the CH34 story.
(1986) The nucleotide sequence of the mercuric resistance operons of plasmid R100 and transposon Tn501: further evidence for mer genes which enhance the activity of the mercuric ion detoxification system. Mol. Gen. Genet. 202, 143–151.
Silver, S., Lee, B.T.O., Brown, N.L. and Cooksey, D.A. (1993) Bacterial plasmid resistances to copper, cadmium, and zinc. In: Chemistry of Copper and Zinc Triads (Welch, A.J. and Chapman, S.K., Eds.), pp. 38–53. The Royal Society of Chemistry, London.
(1994) Newer systems for bacterial resistances to toxic heavy metals. Environ. Health Perspect. 102, 107–113.
Silver, S. and Walderhaug, M. (1995) Bacterial plasmid-mediated resistances to mercury, cadmium and copper. In: Toxicology of Metals. Biochemical Aspects (Goyer, R.A. and Cherian, M.G., Eds.), pp. 435–458. Springer Verlag, Berlin.
(1972) Mercury resistance in a plasmid-bearing strain of Escherichia coli. J. Bacteriol. 112, 1228–1236.
(1998) Related membrane domains in proteins of sterol sensing and cell signaling provide a glimpse of treasures still buried within the dynamic realm of intracellular metabolic regulation. Curr. Opin. Lipidol. 9, 137–140.
(1996) Co-cultivation of Niemann-Pick disease type C fibroblasts belonging to complementation groups alpha and beta stimulates LDL-derived cholesterol esterification. J. Inherit. Metab. Dis. 19, 769–774.
(1997) A family of Gram-negative bacterial outer membrane factors that function in the export of proteins, carbohydrates, drugs and heavy metals from Gram-negative bacteria. FEMS Microbiol. Lett. 156, 1–8.
Nies, D.H. and Brown, N.L. (1998) Two-component systems in the regulation of heavy metal resistance. In: Metal Ions in Gene Regulation (Silver, S. and Walden, W., Eds.), pp. 77–103. Chapman and Hall, London.
Legatzki, A., Franke, S., Lucke, S., Hoffmann, T., Anton, A., Neumann, D. and Nies, D.H. (2003) First step towards a quantitative model describing Czc-mediated heavy metal resistance in Ralstonia metallidurans. Biodegradation (in press).
Brown, N.L., Lee, B.T.O. and Silver, S. (1993) Bacterial transport of and resistance to copper. In: Metal Ions in Biological Systems, Vol. 30 (Sigel, H. and Sigel, A., Eds.), pp. 405–430. Marcel Dekker, New York.
Bryson, J.W., O'Halloran, T.V., Rouch, D.A., Brown, N.L., Camakaris, J. and Lee, B.T.O. (1993) Chemical and genetic studies of copper resistance in E. coli. In: Bioinorganic Chemistry of Copper (Karlin, K.D. and Tylekar, Z., Eds.), pp. 101–109. Chapman and Hall, New York.
Lee, B.T.O., Brown, N.L., Rogers, S., Bergemann, A., Camakaris, J. and Rouch, D.A. (1990) Bacterial response to copper in the environment: Escherichia coli as a model system. In: Metal Speciation in the Environment (Gucer, S. and Broeckaert, J.A.C., Eds.). Springer-Verlag, New York.
(1997) Copper-inducible transcriptional regulation at two promoters in the Escherichia coli copper resistance determinant pco. Microbiology 141, 1191–1202.
(2001) Construction of a series of mutants lacking all of the four major mex operons for multidrug efflux pumps or possessing each one of the operons from Pseudomonas aeruginosa PAO1: MexCD-OprJ is an inducible pump. FEMS Microbiol. Lett. 202, 139–143.
(1998) Involvement of outer membrane protein TolC, a possible member of the mar-sox regulon, in maintenance and improvement of organic solvent tolerance of Escherichia coli K-12. J. Bacteriol. 180, 938–944.
(2000) Ineffectiveness of topoisomerase mutations in mediating clinically significant fluoroquinolone resistance in Escherichia coli in the absence of the AcrAB efflux pump. Antimicrob. Agents Chemother. 44, 10–13.
(2002) Inactivation of the acrA gene is partially responsible for chloramphenicol sensitivity of Escherichia coli CM2555 strain expressing the chloramphenicol acetyltransferase gene. Microb. Drug Resist.-Mech. Epidemiol. Dis. 8, 179–185.
(2002) On the mechanism of substrate specificity by resistance nodulation division (RND)-type multidrug resistance pumps: the large periplasmic loops of MexD from Pseudomonas aeruginosa are involved in substrate recognition. Mol. Microbiol. 46, 889–901.
(2002) Identification by RNA profiling and mutational analysis of the novel copper resistance determinants CrdA (HP1326), CrdB (HP1327), and CzcB (HP1328) in Helicobacter pylori. J. Bacteriol. 184, 6700–6708.
(2000) Two types of Bacillus subtilis tetA(L) deletion strains reveal the physiological importance of TetA(L) in K+ acquisition as well as in Na+, alkali, and tetracycline resistance. J. Bacteriol. 182, 2088–2095.
Luhmann, J.G. and Russell, C.T. (1997) Mars: magnetic field and magnetosphere. In: Encyclopedia of Planetary Sciences (Shirley, J.H. and Fainbridge, R.W., Eds.), pp. 454–456. Chapman and Hall, New York.
(2000) Expression of ZRC1 coding for suppressor of zinc toxicity is induced by zinc-starvation stress in Zap1-dependent fashion in Saccharomyces cerevisiae. Biochem. Biophys. Res. Commun. 276, 879–884.
(1997) Ultrastructral localization of zinc transporter-3 (ZnT-3) to synaptic vesicle membranes within mossy fiber boutons in the hippocampus of mouse and monkey. Proc. Natl. Acad. Sci. USA 94, 12676–12681.
(2002) tcrB, a gene conferring transferable copper resistance in Enterococcus faecium: occurrence, transferability, and linkage to macrolide and glycopeptide resistance. Antimicrob. Agents Chemother. 46, 1410–1416.
(2000) Development of genetic tools for Lactobacillus sakei: Disruption of the beta-galactosidase gene and use of lacZ as a reporter gene to study regulation of the putative copper ATPase, AtkB. Appl. Environ. Microbiol. 66, 4272–4278.
(2002) Molecular characterization of an operon, cueAR, encoding a putative P1-type ATPase and a MerR-type regulatory protein involved in copper homeostasis in Pseudomonas putida. Microbiology (UK) 148, 2857–2867.
(2002) Zinc binding to the NH2-terminal domain of the Wilson disease copper-transporting ATPase – Implications for in vivo metal ion-mediated regulation of ATPase activity. J. Biol. Chem. 277, 13409–13414.