The Sensory Histidine Kinase CusS of Escherichia coli Senses Periplasmic Copper Ions

Homeostasis of essential-but-toxic transition metal cations such as Zn(II) and Cu(II)/Cu(I) is an important contributor to the fitness of environmental bacteria and pathogenic bacteria during their confrontation with an infected host. Highly efficient removal of threatening concentrations of these metals can be achieved by the combined actions of an inner membrane with a transenvelope efflux system, which removes periplasmic ions after their export from the cytoplasm to this compartment. ABSTRACT Two-component regulatory systems composed of a membrane-bound sensor/sensory histidine kinase (HK) and a cytoplasmic, DNA-binding response regulator (RR) are often associated with transenvelope efflux systems, which export transition metal cations from the periplasm directly out of the cell. Although much work has been done in this field, more evidence is needed for the hypothesis that the respective two-component regulatory systems are indeed sensing periplasmic ions. If so, a regulatory circuit between the concentration of periplasmic metal cations, sensing of these metals, and control of expression of the genes for transenvelope efflux systems that remove periplasmic cations can be assumed. Escherichia coli possesses only one transenvelope efflux system for metal cations, the Cus system for export of Cu(I) and Ag(I). It is composed of the transenvelope efflux system CusCBA, the periplasmic copper chaperone CusF, and the two-component regulatory system CusS (HK) and CusR (RR). Using phoA- and lacZ-reporter gene fusions, it was verified that an assumed periplasmic part of CusS is located in the periplasm. CusS was more important for copper resistance in E. coli under anaerobic conditions than under aerobic conditions and in complex medium more than in mineral salts medium. Predicted copper-binding sites in the periplasmic part of CusS were identified that, individually, were not essential for copper resistance but were in combination. In summary, evidence was obtained that the two-component regulatory system CusSR that controls expression of cusF and cusCBA does indeed sense periplasmic copper ions. IMPORTANCE Homeostasis of essential-but-toxic transition metal cations such as Zn(II) and Cu(II)/Cu(I) is an important contributor to the fitness of environmental bacteria and pathogenic bacteria during their confrontation with an infected host. Highly efficient removal of threatening concentrations of these metals can be achieved by the combined actions of an inner membrane with a transenvelope efflux system, which removes periplasmic ions after their export from the cytoplasm to this compartment. To understand the resulting metal cation homeostasis in the periplasm, it is important to know if a regulatory circuit exists between periplasmic metal cations, their sensing, and the subsequent control of the expression of the transenvelope efflux system. This publication adds evidence to the hypothesis that two-component regulatory systems in control of the expression of genes for transenvelope efflux systems do indeed sense metal cations in the periplasm.

cytoplasmic location of the amino-and large carboxy-terminal part of CusS. In contrast, the measured high PhoA and low LacZ enzyme activities provided evidence for a periplasmic location of CusS when fusions were made at positions 38,112, and 185 within the protein. High activity of both PhoA and LacZ was measured when the fusion was made at position 207, following the second predicted transmembrane alpha helix of CusS (Fig. 1, gray arrow).
Western blotting of the fusion proteins produced in E. coli cells after separation of polypeptides in crude extracts by SDS-PAGE ( Fig. 1) yielded signals for LacZ fusion proteins at positions 15 and 17 (blue arrows) but not PhoA fusion proteins, which agreed with the low PhoA-specific activity detected (Table 1). At positions 38, 112, and 185, PhoA fusion proteins, but also some LacZ fusion proteins, could be visualized (magenta arrows). At position 207 (gray arrows), the LacZ fusion protein was clearly visible, while the corresponding PhoA activity showed degradation. Finally, at position 480, no PhoA fusion protein was detectable, but a LacZ fusion protein and many of its degradation products could be detected. The specific activities of the corresponding enzymes (Table 1) in combination with the presence or absence of the fusion proteins in E. coli ( Fig. 1) clearly verified the predicted topology of CusS with a short cytoplasmic aminoterminal region and a large carboxy-terminal domain of CusS, two transmembrane alpha-helices, and a periplasmic domain between positions 35 and 187.
The structure of the periplasmic domain between positions 38 and 185 has been solved (24), and amino acid residues involved in binding of the Cu(I) proxy Ag(I) have been identified, namely, F43, H42, and H176 between alpha-helices leading down to the cytoplasmic membrane as well as S84, M133, M135, and H145 in a disordered loop of the periplasmic domain. The next question was whether the predicted metalbinding sites in the periplasmic domain of CusS (Fig. 2) were required for activation of cus via the HK CusS and the RR CusR.
Sensing of copper by the periplasmic part of CusS is required for Cus-mediated copper resistance under anaerobic conditions. To investigate the contribution of the periplasmic part of CusS to copper resistance in E. coli, growth conditions had to be identified that provided a difference between a DcusS strain and its respective parent. In a DcopA DcueO double mutant, the Cus system was the major remaining copper resistance system in E. coli. In Tris-buffered mineral salts medium (TMM) (Fig. 3), deletion of cusS had a clear effect under aerobic and anaerobic conditions. Under aerobic conditions, the 50% inhibitory concentration (IC 50 ) of the triple mutant was about a Translational fusions of the cusS gene with the phoA and the lacZ genes were created at the indicated codon/amino acid residue positions as shown in Fig. 1. The specific activities of the respective fusion constructs were determined. The ratio Q between the LacZ and the PhoA activities indicated a cytoplasmic location of the PhoA and LacZ domains at the amino and carboxy termini (blue arrows in Fig. 1), a periplasmic location for those in light gray rows (magenta arrows in Fig. 1), and an ambiguous value for position 207 following the second transmembrane alphahelix (dark gray row and gray arrow in Fig. 1). Copper resistance under anaerobic conditions was significantly lower (Fig. 3B), with a difference between DcusS and its parent at copper concentrations above 2 mM. However, the differences between the triple mutant and the double mutant with an intact cusS gene were small in the DcopA DcueO background, even in mineral salts medium, probably due to the strong decrease in copper resistance as result of the DcopA deletion. Nevertheless, these data confirmed that Cus contributed to copper resistance, even under aerobic conditions when CueO is absent (5,21). In the complex medium LB, the IC 50 of the wild-type strain under aerobic conditions was greater than 2 mM (Fig. 4). Under anaerobic conditions, the IC 50 of the wild type was above 100 mM, while the IC 50 of the DcusS single mutant was about 25 mM, revealing a sufficiently large difference between the presence and absence of cusS for the subsequent experiments; it also confirmed the important role of Cus under anaerobic conditions (5,21).
Three HKs from E. coli were reported to cross-talk with Cus: BarA, UhpB, and YedV (27). The respective genes for these histidine kinases were deleted in E. coli and additionally cusS in the resulting triple mutant. Under aerobic conditions in LB, the copper resistance of a DbarA DuhpB DyedV triple mutant, a DbarA DuhpB DyedV DcusS quadruple mutant, a DcusS single mutant, and the parental strain was not significantly different (Fig. 4A). Under anaerobic conditions, copper resistance of the DbarA DuhpB DyedV triple mutant was similar to that of the parental strain, while that of the DbarA DuhpB DyedV DcusS quadruple mutant was similar to that of the DcusS single mutant (Fig. 4B). Deletion of cusS resulted in loss of copper resistance under anaerobic conditions, and the HKs BarA, UhpB, and YedV did not interfere with Cus-mediated copper resistance. Therefore, cross-talk between nonmetal-sensing HKs (27) was not evident in the control of Cus-mediated metal resistance.
CusS and mutant derivatives thereof were all produced in E. coli as Strep-tagged proteins from a plasmid carrying the cognate gene, so that the presence of the proteins in the cells and their location in the cytoplasmic membrane could be verified by western blotting. A CusS-specific signal was observed in the Western blot but was absent in the vector control (Fig. 5A). Even without the inducer anhydrotetracycline (AHT) being added, CusS could be detected in the membrane fraction (Fig. 5A, lane 9). After induction with AHT, CusS was found in the membrane fraction and additionally in the cellular debris, indicating that some CusS may have ended up in inclusion bodies under these conditions. CusS was produced in E. coli in trans from a plasmid as a Strep-tagged protein and was located in the membrane. Expression of cusS from this plasmid resulted in a complete restoration of copper resistance in the DcusS-mutant strain under anaerobic conditions (Fig. 6), while the copper resistance of the DcusS mutant with only the vector without gene insert remained at a low level. It can be concluded, therefore, that the Strep-tagged version of CusS produced from a plasmid was fully functional.
The periplasmic region of CusS is required for copper resistance. Because Cu(I) may bind to several sites in the periplasmic region of CusS (24), mutations in singleamino-acid residues might possibly have no effect on CusS function. Therefore, a larger portion of the periplasmic domain of CusS was removed by introducing a corresponding deletion in the coding region of the cusS gene (Fig. 2). The coding sequence of a disordered loop in the periplasmic portion of CusS was removed. Deletions did not affect the sequences encoding the two alpha-helices, which continue into the transmembrane alpha-helices and contain the Cu(I)-binding site H176, F43, and H42 (Fig. 2,  red). Such a deletion would most likely disturb the overall structure of CusS. Instead, two amino acid sequences with metal-binding motifs were removed, specifically those in the histidine-rich stretch at positions 136 to 148 with H138, H149, H142, M143, and H145 (Fig. 2, blue and green) and at positions 129 to 138 with the methionine triad M133, M134, and M135 (magenta and blue). Moreover, a peptide spanning both regions (positions 129 to 153) was also removed. An even larger deletion (positions 67 to 153; Fig. 2, in yellow) should disturb the structure of CusS and served as a negative control.

CusS of E. coli Senses Periplasmic Copper Ions
Microbiology Spectrum proteins CusSD(129 to 152), CusSD(136 to 148), and CusSD(129 to 138) were identified in the crude extract, were enriched in the debris, and were more abundant in the membrane fraction, but they were not identified in the soluble fraction ( Fig. 5C and D). Thus, all four deletion derivatives of CusS were produced in E. coli and were located in the membrane fraction. The CusSD (129 to 138) derivative, which no longer contained the histidine-rich stretch, mediated copper resistance on the level of the full-length protein (Fig. 6, open inverted triangles). Resistance mediated by CusSD(136 to 148) without the methionine triad was lower (open triangles) than copper resistance in the strain complemented with full-length CusS but was much higher than that of the strain with the vector control. When both regions were deleted, the mutant protein could no longer mediate upregulation of cus (open squares). Copper resistance was similar to that of the vector control (open circles) and to that mediated by the CusS derivative with the largest deletion CusSD(67 to 153) (open diamonds).
Together, these data indicate that CusS has a periplasmic domain that is required for the copper-mediated function of the histidine kinase.

DISCUSSION
The CusRS system from E. coli was used to verify that the signals sensed by the histidine kinase CusS originate from the periplasm. PhoA and LacZ protein fusions defined the topology of membrane-bound CusS and clearly demonstrated that a domain of CusS that was predicted to be in the periplasm indeed was located in this compartment. In agreement with published data (3), a copper-dependent activation of CusS was required for cusCFBA expression and consequently Cus-mediated copper resistance under anoxic conditions (Fig. 3 and 4). Cus was required under these conditions (5) because the periplasmic Cu(I) oxidase CueO could not function due to the absence of its electron acceptor, molecular oxygen.
The residues T17, F18, and V202 within the two transmembrane alpha-helices ( Fig. 1) are in close proximity to each other in the active CzcS dimer. Distance changes from intramolecular in the signaling-inactive CusS dimer form to intermolecular in its active form (28). Both transmembrane alpha-helices continue into alpha-helices that define the beginning and the end of the metal-sensing periplasmic domain of CusS. The alpha-helices downstream of M38 and upstream of N185 (Fig. 2), including the Cu (I)-binding site H42, F43, and H176 (24), delimit the beginning and the end of the periplasmic part of CusS (Table 1). Disturbance of the conformation downstream of F43 and upstream of H176 affected the functionality of CusS. In particular, the flexible loop from S129 to N152 was important for function of the protein. Loss of the histidine-rich loop (Fig. 2, blue and green) could be fully substituted by binding of Cu(I) to other sites, loss of the methionine triad (magenta) to a large degree (Fig. 6).
Binding of substrate ions to the residues F43, H42, and H176 (Fig. 2), which are located in the periplasmic part of both alpha-helices close to the outer face of the cytoplasmic membrane, triggers change of the conformation of the CusS dimer (8,24). Studies conducted with the H34, H38, and H171 amino acid residues in CorS of Myxococcus xanthus confirm this hypothesis for CusS (29). Additional metal-binding sites in a loosely organized loop of the periplasmic domain are involved in activation of CusS (Fig. 6). Methionine-rich loops, such as those formed by M133, M134, and M135, are also involved in binding of copper to blue copper-dependent oxidases (30). Together, these data demonstrate that periplasmic copper ions are the signals that are sensed by CusS.
CusS and related transition metal-sensing HKs plus their associated response regulators use periplasmic cations as signals to control the expression of their genes. This provides further evidence that transenvelope efflux systems, which are encoded by determinants together with the genes for a two-component regulatory system, function to remove periplasmic metal cations from the periplasm before these cations can be imported into the cytoplasm. This ultimately results in a decrease of the cytoplasmic concentration of the respective metal. In this way, the transenvelope system also interacts with efflux systems of the inner membrane, which antagonize the action of the uptake systems. The cell thus achieves optimal control over the cytoplasmic metal ion concentration by indirectly restricting the import/export by the inner membrane system. Moreover, the cations are exported in a two-step mechanism, first by the inner membrane efflux system to the periplasm and later by the transenvelope efflux system. Demonstration of the dependence of CzcCBA-mediated Co(II) resistance in C. metallidurans on the presence of DmeF clearly supports this hypothesis (31).
In M. xanthus, the CusRS homologs CorRS also control expression of a gene for a copper-exporting P-type ATPase (29). Consequently, the genome of M. xanthus contains genes for orthologs of CusA, for instance WP_011551109. In C. metallidurans and Pseudomonas syringae, CusRS orthologs control the expression of genes for a CueO-like periplasmic Cu(I) oxidase plus helper genes (32)(33)(34)(35)(36). In contrast, cueO expression in E. coli is not under the control of a two-component regulatory system but instead is controlled by the MerR-type regulator CueR, which also controls copA expression (18)(19)(20)(37)(38)(39). Under aerobic conditions, the number of copA and cueO mRNAs reach a peak level of about 350 and 180 copies per cell, respectively, when E. coli is challenged by 1 mM Cu(II). Subsequently, the number of mRNA copies of either gene decreases to about 20 copies per cell within 20 min (40). Under anaerobic conditions, these numbers reach 1,200 copies per cell for copA and about 350 copies per cell for cueO within 20 min and remain at this level subsequently if the anaerobic conditions persist. Because CueR senses Cu(I) in the cytoplasm (38), this indicates a rapid increase in cytoplasmic Cu(I) after the cells are confronted with 1 mM Cu(II). Accumulation of cytoplasmic Cu(I) could be decreased again by the action of CueO and CopA under aerobic but not anaerobic conditions, indicating that oxidation of periplasmic Cu(I) by CueO results in decreased cytoplasmic Cu(I) accumulation; Cu(I) is a better substrate for import than Cu(II).
Evidence presented in this publication strengthens the hypothesis that the CusRS system senses periplasmic Cu(I) to control the expression of cus. E. coli as a facultative anaerobic bacterium can be confronted with a higher ratio of Cu(I) to Cu(II) than an aerobic bacterium such as C. metallidurans. E. coli suffers from a high probability that periplasmic Cu(I) oxidation by CueO may not function, making CueO a facilitator for CopA, as is evident from their common regulation by the MerR-type one-component regulator CueO.

MATERIALS AND METHODS
Bacterial strains and growth conditions. E. coli strains and primers are provided in Tables 2 and 3. Tris-buffered mineral salts medium (41) containing 2 g/L sodium gluconate (TMM) was used. E. coli was cultivated under aerobic conditions at 37°C with 2 g/L glycerol and under anaerobic conditions in Hungate tubes with 2 g/L glucose instead of gluconate in a TMM adapted for E. coli. LB medium was also used (Becton, Dickinson, Heidelberg, Germany) (42,43).
Dose-response growth curves were conducted in TMM or LB. A preculture was incubated for 16 h at 37°C and 200 rpm up to early stationary phase, diluted 1:400 into fresh medium with glycerol as a carbon source in the case of TMM, incubated for 24 h at 37°C and 200 rpm, diluted again 400-fold into fresh medium with glucose as a carbon source, and incubated for 2 h with shaking at 37°C. CuCl 2 was added, the incubation was continued for 6 h at 37°C, and the turbidity was determined at 600 nm using a Spectronic 201 (Milton Roy, Ivyland, PA). Under anaerobic conditions, TMM contained 2 g/L glucose, and the incubation time with copper ions was 16 h; LB contained 0.5 g/L glucose, and the incubation time was 6 h before the turbidity was measured.
Genetic techniques. Standard molecular genetics techniques were used (44,45). Genes were deleted by insertion of resistance cassettes using the l Red-recombinase system (46), as previously published (21). Initial deletions were performed in E. coli strain BW25113, in which the target genes were exchanged for a chloramphenicol (cat) resistance cassette, and subsequently transferred by general transduction with phage P1 into E. coli strain W3110 or its derivatives. In the resulting mutant strains, the genes were disrupted by insertion of the cat resistance cassette through homologous recombination. Multiple deletions were constructed by FLP recombination target (FRT)-dependent elimination of the respective resistance cassette assisted by flippase from plasmid pCP20 (46) and subsequent general phage P1 transduction. All mutations were verified by PCR.