Newer systems for bacterial resistances to toxic heavy metals.

Bacterial plasmids contain specific genes for resistances to toxic heavy metal ions including Ag+, AsO2-, AsO4(3-), Cd2+, Co2+, CrO4(2-), Cu2+, Hg2+, Ni2+, Pb2+, Sb3+, and Zn2+. Recent progress with plasmid copper-resistance systems in Escherichia coli and Pseudomonas syringae show a system of four gene products, an inner membrane protein (PcoD), an outer membrane protein (PcoB), and two periplasmic Cu(2+)-binding proteins (PcoA and PcoC). Synthesis of this system is governed by two regulatory proteins (the membrane sensor PcoS and the soluble responder PcoR, probably a DNA-binding protein), homologous to other bacterial two-component regulatory systems. Chromosomally encoded Cu2+ P-type ATPases have recently been recognized in Enterococcus hirae and these are closely homologous to the bacterial cadmium efflux ATPase and the human copper-deficiency disease Menkes gene product. The Cd(2+)-efflux ATPase of gram-positive bacteria is a large P-type ATPase, homologous to the muscle Ca2+ ATPase and the Na+/K+ ATPases of animals. The arsenic-resistance system of gram-negative bacteria functions as an oxyanion efflux ATPase for arsenite and presumably antimonite. However, the structure of the arsenic ATPase is fundamentally different from that of P-type ATPases. The absence of the arsA gene (for the ATPase subunit) in gram-positive bacteria raises questions of energy-coupling for arsenite efflux. The ArsC protein product of the arsenic-resistance operons of both gram-positive and gram-negative bacteria is an intracellular enzyme that reduces arsenate [As(V)] to arsenite [As(III)], the substrate for the transport pump. Newly studied cation efflux systems for Cd2+, Zn2+, and Co2+ (Czc) or Co2+ and Ni2+ resistance (Cnr) lack ATPase motifs in their predicted polypeptide sequences. Therefore, not all plasmid-resistance systems that function through toxic ion efflux are ATPases. The first well-defined bacterial metallothionein was found in the cyanobacterium Synechococcus. Bacterial metallothionein is encoded by the smtA gene and contains 56 amino acids, including nine cysteine residues (fewer than animal metallothioneins). The synthesis of Synechococcus metallothionein is regulated by a repressor protein, the product of the adjacent but separately transcribed smtB gene. Regulation of metallothionein synthesis occurs at different levels; quickly by derepression of repressor activity, or over a longer time by deletion of the repressor gene at fixed positions and by amplification of the metallothionein DNA region leading to multiple copies of the gene.


Introduction
Bacterial mechanisms for resistances to toxic heavy metals have become a mature scientific subject over the last 20 years. Highly specific systems exist for most toxic cations and oxyanions ( Figure 1). The genes determining these resistances are generally (but not always) found on bacterial plasmids. Heavy-metal resistance mechanisms were recently reviewed in Plasmid with articles on mercury resistance (1), the most thoroughly understood of bacterial This paper was presented at the Second International Meeting on Molecular Mechanisms of Metal Toxicity and Carcinogenicity held 10-17 January 1993 in Madonna di Campiglio, Italy. resistance systems (2,3) and on resistances to arsenic (4), copper (5), cadmium (6), chromate, silver, and tellurite. Here, we have reviewed five less familiar bacterial heavy-metal resistance systems.
Bacterial Plasmid Copper-Resistance Systems Bacterial plasmid copper-resistance systems have been found on plasmids in Escherichia coli (5,7) and Pseudomonas syringae (8) and on the chromosome of Xanthomonas campestris (8). It was thought that the basic mechanism of resistance might be different in E. coli and in Pseudomonas (3,5).
However, new DNA sequence analysis of the E. coli plasmid determinant has shown that the systems are basically equivalent and consist of four structural genes, now called pcoA, pcoB, pcoC, and pcoD in E. coli, and copA, copB, cop C, and copD in Pseudomonas (7). The Pseudomonas and E. coli systems both have paired regulatory genes, with an apparently membranebound Cu2+ sensor (the pcoS or copS gene product) coupled with the pcoR or copR gene products that may be DNA-binding repressor proteins (5,7). The genes and their protein products are summarized in Figure 2, along with the percentages for Plasmid Heavy Metal Resistance Systems and Mechanisms. 1. mer. Mercury and organomercurials are enzymatically detoxified. 2. ars. Arsenate is reduced to arsenite enzymatically. Arsenite is "pumped" from cells by a membrane protein or a two-component ATPase. Sb(lll) is also a transport substrate. 3. cadA. Cadmium (and zinc) are "pumped" from the cells by a P-type ATPase. 4. cop. Copper resistance in Pseudomonas and pco. Copper resistance in E coli results from bioaccumulation of copper from the medium and binding by surface proteins. 5. czc. Cadmium, zinc, and cobalt are "pumped" from the cells by a three-componenet membrance complex that appears NOT to be an ATPase. cnr. Cobalt and nickel are "pumped" from the cells by a related membrane complex that may (after mutation) also pump zinc. 6. chr. Chromate resistance results from reduced cellular uptake but an efflux pump has not been demonstrated. Additional resistance systems await understanding: bismuth, boron, lead, silver, thalium, and tellurium.   Figure 2. Comparison of the genes for copper resistance from an E coli plasmid, an P. syringae pv tomato plasmid and the chromosomal determinant of Xanthomonas campestris pv juglandis. The copABCDRS gene products (9,12) and the pcoR gene product (5) have been published. The remainder of this analysis is based on preliminary sequences by NL Brown (personal communication) for the Pco sequence from E coli and by Y-A Lee, M Hendson, and MN Schroth (personal communication) for the Xanthomonas sequence. The precise numbers may be subject to minor corrections and the "missing" genes and the nature of pcoEremain for additional work.
amino acid identities between corresponding protein products. Southern blotting DNA/DNA hybridization fails to detect homologies between the E. coli and Pseudomonas genes (8)(9)(10). However, amino acid identities ranging from 38 to more than 75% allow little question but that the proteins will fnmction in comparable roles. For the E. coli system, an additional reading frame called pcoE was identified in the DNA sequence; it appears to be transcribed in a Cu2-inducible manner. The function of this gene product is not known, and the comparable region for the Pseudomonas determinant has not been sequenced. Figure 2 also includes information from a chromosomal Xanthomonas copper-resistance determinant for which only the four structural genes have been sequenced (Y-A Lee, M Hendson, and MN Schroth, personal communication). It was earlier shown that Xanthomonas and Pseudomonas copper-resistance determinants do not detectably cross-hybridize (10). However, it is surprising that the copper-resistance systems from P. syringae and E. coli are more similar (Figure 2) than are the Pseudomonas and Xanthomonas determinants. The bacteria from which these copper-resistance systems have been isolated are all pathogens of agricultural interest, but are otherwise quite different in their sources, with the E. coli from pigs with diarrhea, and the Pseudomonas and Xanthomonas pathogens of tomato plants and walnut trees, respectively. The Pseudomonas CopA, CopB and CopC proteins have been isolated (CopD remains to be identified). The CopA and CopC proteins are blue periplasmic proteins, containing 11 and 1 Cu2+ cation respectively by direct analysis (11). The CopB protein is an outer membrane protein and CopD is probably an inner membrane protein ( Figure 3). How these four proteins function together to allow periplasmic "bioaccumulation" of copper by resistant cells is still unknown. However, colonies of the copper-resistant Plasmid genes C. 2+ Pseudomonas turn bright blue, while the supporting agar becomes colorless, as the copper salts are taken up by the cells. The colonies of copper-resistant E. coli become brown, however, when grown on coppercontaining agar. The nature of the brown pigment is not known.
Regulation of the copper resistance determinants is also under study. Thus far, the pcoR and pcoS [or copR and copS (12)] gene products are known only from DNA sequence analysis and the proteins have not be isolated. The sequences are typical of the 100 or so "two-component" regulatory systems that have been identified in bacteria (13), so there is little doubt of the overall mechanism ( Figure 4). PcoS appears to be a membrane protein. It is homologous in sequence to other "sensor" proteins (kinases) (13) that respond to external stimuli by autophosphorylating (from ATP) a specific histidine residue (for PcoS, His257) that is always conserved in this class of membrane sensors. The phosphorylated sensor protein then transfers its phosphate to a specific aspartate residue (Asp52 for PcoR) of the secondary transducer protein (12,13). In many cases, the transducer is activated by phosphorylation and binds to the appropriate DNA operator region, to initiate transcription of the operon, for example, here pcoABCD. However, there is a question as to whether PcoR might be a repressor protein that is inactivated (and released from the DNA) by phosphorylation. Furthermore, in the Pseudomonas cop system, there is evidence for a third regula-Chromosomal genes  The current (and tentative) picture for the cut (for Cu2+ transport) chromosomal gene products involved in uptakei intracellular movement, and efflux of Cu + (3,5,7) are also shown in Figure 3. In E. coli, these genes were initially tagged by mutations leading to copper sensitivity (14). Only one of the cut genes (cutE) has been cloned and sequenced to date (15) and its sequence does not immediately lead to an explanation of its intracellular role ( Figure 3). Comparable cut chromosomal mutations to copper sensitivity have been obtained with Pseudomonas (DA Cooksey, personal communication). The recent identification of a previously sequenced gene for a P-type ATPase in Enterococcus as a cut gene (16) opens the search for comparable Cu2+ efflux ATPases in other bacteria. Furthermore, the human X-chromosome gene responsible for the lethal disease Menkes syndrome appears to determine a P-type ATPase (17). The discussion of the Menkes gene and bacterial P-type ATPases in the section on Cd2+ resistance below provides further information.
Metallothionein from the Cyanobacterium Synechococcus A bacterial metallothionein from the cyanobacterium Synechococcus was purified and sequenced (18). The gene encoding metallothionein is called smtA (19). Bacterial metallothionein probably represents the independent evolutionary development of the same solution to heavy-metal binding obtained with animal and plant metallothioneins, since the sequence is shorter (only 56 amino acids) and contains fewer cysteine residues (9, compared with 20 out of 60 amino acids for animal metallothioneins, or 12 out of 75 amino acids for plant metallothionein). Unlike animal metallothioneins, the Synechococcus metallothionein contains two aromatic tyrosine residues in the intracysteine-rich region (19). Synechococcus metallothionein synthesis is regulated by heavy metals such as cadmium and zinc (20) by means of a repressor protein synthesized from a constitutively expressed divergently transcribed gene smtB (20), ( Figure 5). In addition to rapid response to toxic metals by derepression (20), two long-term mechanisms of increasing smtA metallothionein gene expression have been found. Upon growth on high cadmium concentrations, amplification of the metallothionein gene region occurs (21) leading  (22), whereas the HIP1 frequency is lower in bacteria other than cyanobacteria and is virtually unknown in E. coli. In summary of Figure 5, three levels of gene regulation occur: rapidly by derepression of the SmtB repressor, and more slowly in response to continued toxic metal stress by amplification of the gene copy number and by deletion of the repres-Metallothionein 56 amIno acids 9 cyateines Metal-binding la SmtA I t t sor gene between fixed chromosomal points ( Figure 5).

Bacterial Plasmid Arsenic Resistance
Bacterial plasmid arsenic resistance results from the rapid energy-dependent efflux of oxyanions added as either arsenate [As(V)], arsenite [As(III)], or antimonite [Sb(III)]. The same basic mechanism (and genes) exists in both gram-negative and gram-positive bacteria ( Figure 6). The mechanism of bacterial arsenic resistance has been repeatedly reviewed (3,4,23,24), but there are several recent and surprising findings. After an operator/promoter regulatory region, both the gram-positive and the gram-negative versions of the ars operon start with the arsR regulatory gene (Figure 6) that encodes a transacting repressor protein  (25,26). ArsR of the gram-negative plasmid R773 and the gram-positive version from two plasmids (26,27) are identical in 30% of their amino acid positions. The ArsR proteins are not very similar to other known bacterial regulatory proteins, but ArsR is weakly similar to the SmtB regulatory protein for cyanobacterial metallothionein regulation (above) and to the small CadC protein in the cadmium resistance system (below). The ArsR structure and means of binding to the DNA operator region (25,27) are not known. Since the regulatory protein is made as part of the same transcriptional unit as the other arsenic resistance genes, the system cannot be turned "tighdy off': low levels of ArsR are always needed. The arsenic resistance operons of grampositive and negative bacteria diverge after arsR. The plasmid R773 system has two genes, arsD and arsA, that are missing from the staphylococcal arsenic resistance systems (26,27). ArsD appears to be a minor regulatory protein that "puts a cap" on the amount of transcription that occurs in derepressed cells. ArsA is the ATPase subunit, binding to the ArsB membrane protein and forming the arsenite efflux ATPase (3,4,23,24). ArsA from plasmid R773 has been studied extensively in vitro. It is an oxyanion-stimulated soluble ATPase protein (4,23). How can one have an ATPase efflux pump in gram-positive bacteria without the ATP-binding protein? This major question is not resolved, but there is preliminary evidence that ArsB alone functions in Staphylococcus as a chemiosmotic secondary transport system (24,28). It was hypothesized (23) that an ancestral electrogenic oxyanion system (ArsB) may have acquired through evolution a second ATPase protein component, converting it into a more effective ATPase primary pump. The basic similarity of membrane regions from differing primary and secondary membrane transporters (29,30) is consistent with this idea. In summary, there are still major unresolved questions concerning the mechanism of arsenic resistance.
The next gene common to all three sequenced arsenic resistance determinants is arsB (Figure 6), the determinant of the membrane protein that is needed for arsenic efflux. The ArsB proteins from gram-negative and gram-positive bacteria are 58% identical in amino acids. They are so similar in overall sequence that chimeric ArsB proteins with the first third of ArsB from E. coli and two-thirds of ArsB from Staphylococcus (and the reverse) have been made by swapping gene segments (S Dey, D Dou, BP Rosen, personal communication). These chimeric ArsB proteins give partial resistances to arsenate and arsenite, but they are not equivalent to the native gene products. Higher levels of resistances occur with the ArsA protein present in addition to ArsB. The details of how the regions of the chimeric ArsB proteins function remain to be deduced.
The final gene common to all three arsenic resistance operons is arsC, but the protein products are only 18% identical in amino acid sequences. Nevertheless, the function of arsC is the same: to provide arsenate resistance to a system limited to arsenite and antimonite resistances in the absence of arsC. Recent experiments have shown that the arsC protein i__an enzyme, a reductase that converts AsO4 [As(V)] to AsO2 [As(III)] (31). The gram-positive version of this small, soluble protein couples in vitro to thioredoxin (31), which is a general intracellular redox protein. In vitro ArsC reductase activity can be driven by thioredoxin and an artificial reduced thiol compound, dithiothreitol (DTT: Figure  7), or by the larger protein thioredoxin reductase and NADPH (Figure 7). Since thioredoxin reductase catalyzes an oxidized to reduced dithiol cycle in thioredoxin, it is expected that thioredoxin similarly carries out reduction of an oxidized dithiol [two of the four cysteine residues in staphylococcal ArsC (26,27); however, there are only two cysteines in the E. coli ArsC] and that ArsC protein is oxidized as it reduces arsenate to arsenite (Figure 7). The picture of the arsenic resistance system given here shows major changes from that reported in 1992 (3,4

CadA Cd2 Efflux ATPase
The CadA Cd2+ (and perhaps Zn2+) efflux ATPase is a P-type ATPase (32)(33)(34) found in gram-positive bacteria. Three versions of this genetic determinant have now been sequenced, including one from S. aureus plasmid p1258 (32). A closely related system was found in a alkalophilic soil bacterium Bacillusfirmus (35) and still another has been sequenced from the chromosomal cadmium-resistance determinant of a methicillin-resistant S. aureus isolate (DT Dubin, personal communication). These are three of the ten currently sequenced Ptype ATPases from bacteria, half of which are yet to be published. There are more than 30 P-type ATPase enzymes known from animal sources (3,33,34, and references therein). Whereas the P-type ATPases were once thought to be exclusively animal in origin (and they include the Ca + ATPase of the muscle sarcoplasmic reticulum and the numerous known Na+/K+ ATPases), additional P-class ATPases have been identified in plants and lower eukaryotes as well (34). The basic properties of P-type ATPases are demonstrated in the model of the CadA membrane polypeptide in Figure 8. First, the protein is embedded in the membrane by six hydrophobic sections that presumedly form transmembrane a-helices. These regions are thought to form the transmembrane cation channel. All P-type ATPases are cation-translocating (34); some function for uptake (of K+ and Mg2+) whereas others function for efflux (Cd2+ and Ca2+). In each case, the P-type ATPases are thought to be about 75% intracellular, about 20% in the membrane, and with very little extracellular sequence (8). Positively charged amino acids reside more frequently on the inner membrane surface, whereas negatively charged amino acids are found more often on the outer membrane surface.
There are a number of key conserved positions in P-type ATPases, and in particular in the CadA sequence and candidate human Menkes sequence in Figure 8. Starting from the amino terminus, the CadA protein has a hypothesized Cd2+binding motif of perhaps 30 amino acids.
This motif occurs also in the Cd2+ ATPase of Bacillus (35), twice (back to back) in an unpublished chromosomal Cd2+ resistance determinant in a methicillin-resistant S. aureus (DT Dubin, personal communication), and six times over a 600 amino acid region at the beginning of the gene product for human Menkes disease (17,36). This is a striking example ofwhere studies on bacterial model systems have facilitated understanding of human disease (17,36) (Figure 8).
Menkes syndrome is a human disease of copper deficiency (37) and the characteristics of the defective cells are consistent with the new hypothesis that Menkes disease patients will be defective in synthesis of a P-type ATPase that is a Cu2+ efflux enzyme, more similar to the bacterial Cd2+ ATPases in sequence than it is to known eukaryotic P-type ATPases (17).
After the first membrane hairpin, all Ptype ATPases have a region of 150 to 300 amino acids that includes a "phosphatase motif" Thr-Gly-Glu-Ser in CadA or Thr-Gly-Glu-Ala in the Menkes sequence ( Figure 8). This region has been assigned the role of "transducing" the intracellular protein-bound cation to the membrane channel region. The second membrane hairpin appears to be involved in transmembrane cation movement and includes a conserved proline residue in all P-type ATPases. For CadA and Menkes, this appears in a conserved Cys-Pro-Cys sequence and along with the metal-binding motif cysteines. In addition to these two cysteines in the membrane region, only the metal-binding motif cysteines occur in the entire 727 amino acid CadA polypeptide or 1500 amino acid Menkes sequence (17). The Cys-Pro-Cys sequence is found in this position in six of the ten known prokaryote P-type ATPases (all except in KdpB, MgtA, MgtB, and the Enterococcus possibly-Cu2+ ATPase (16). Following the conserved proline is a second proline residue, eight positions further along in CadA, in the Menkes ATPase, and in the six closest prokaryotic P-type ATPase sequences. Following the second hairpin is an ATP-kinase domain of approximately 300 amino acids in P-type ATPases. This domain includes the aspartate-415 in CadA, which is in the Asp-Lys-Thr-Gly-Thr-Leu (or Ile)-Thr sequence conserved in all P-type ATPases. The aspartate residue is phosphorylated by ATP during the transport cycle. There follows a series of conserved residues involved in ATP binding, and the "kinase" region ends with a highly conserved "hinge motif" that is thought to be involved in allosteric movement of the kinase region so as to modify the cation binding affinity. After a third membrane hairpin, the CadA structure in Figure 8 ends at position Lys727. The Menkes ATPase sequence continues further and ends at Leu1500.
After the sequence of CadA was available (32,35), it was found that the second gene transcribed with cadA (called cadC) was necessary for full resistance (38) and that the cadA system is tightly repressed but inducible by toxic cations (39,40). It is still unclear whether the cadC gene is directly involved in the resistance mechanism (38) or is involved in regulation (39). The CadC amino acid sequence is somewhat related to the ArsR arsenic regulatory protein sequence (see above). Direct mea-surements of uptake of radioactive Cd2+ by inside-out membrane vesicles from cells containing CadA (38) added direct biochemical support for the deductions from seq2uence homologies. CadA-dependent Cd + uptake requires ATP (38), and the CadA protein can be labeled with 32p from ATP, with the labeled protein showing exchange and alkali-sensitivity characteristics similar to other P-type ATPases (41). The CadA P-type ATPase, for which work is only beginning, may become increasingly valuable as a model P-type ATPase for detailed molecular genetic and biochemical studies.

Gram Negative Soil Bacterium Alcaligenes Eutrophus
The gram-negative soil bacterium Alcaligenes eutrophus harbors two divalent cation efflux systems on its large megaplasmids (each more than 150 kilobases in 2+ 2+ 2+ length): the Cd , Zn , and Co resistance system called Czc and the Co2+ and Ni2+ resistance system called Cnr. Cnr has only recently been shown to be related to Czc at the level of genes and their protein products (42,43). For both systems, energy-dependent efflux of divalent cations has been demonstrated to be the resistance mechanism (6,42). The predicted gene products from DNA sequences show no sign of ATPase motifs (42,43). There are preliminary suggestions that the mechanism of energy-coupling may be a chemiosmotic divalent cation/proton antiporter (6,44). The primary differences between the two systems appeared to be cation specificity and the nature of gene regulation. New mutations of the cnr system add zinc resistance (and presumedly efflux) (43,45) to cobalt and nickel resistances; and cnr and czc may be less different than initially thought. As cloned and sequenced, the czc system was found to consist of four genes ( Figure   9). The first is czcA, the determinant of a large membrane protein that is thought to be the principal component of the transport system (6), since deletions involving czcA lost all resistances (42). The smaller membrane (CzcB) and soluble (CzcC) proteins appear to play secondary roles in determining specificity (6,42). The fourth gene product, CzcD, is thought to affect regulation of the system, but not resistance directly (6,42,44). More recent sequencing of the "upstream" region of czc and czc-Pgalactosidase fusion experiments (44) indicate that the divergently transcribed czcR gene and czcD cooperate in gene regulation. CzcR may be an activator protein rather than a repressor (44). Although the paired regulatory proteins CzcD and CzcR are thought to be respectively membrane sensor and DNA-binding responder (6,44), these proteins do not fit into the familiar two-component class of kinase-"sensor" plus transphosphorylated "responder" as do PcoR and PcoS of the copper resistance system (above).
The new sequence of the cnr system (43) has yielded a few surprises. First, although the two genetic systems are sufficiently different that their DNAs do not hybridize on Southern blots, the gene products of CzcA (and CnrA), CzcB (and CnrB), and CzcC (and CnrC) are significantly homologous (Figure 9). Therefore, it is likely that the two systems will function in a fundamentally similar manner. However, the regulatory regions appear at the moment to be unrelated ( Figure 9). The cnr lacks both cnrR and cnrD genes, but has instead three unrelated upstream genes cnr}i cnrR, and cnrH ( Figure 9) that appear to be involved in gene regulation (43,45). It is the cnrYregion that contains the mutations to constitutive function that also extend the system to zinc resistance (43,45). We can look forward to more understanding and possibilities with regard to toxic divalent cation efflux.