Interplay between Two-Component Regulatory Systems Is Involved in Control of Cupriavidus metallidurans Metal Resistance Genes

ABSTRACT Metal resistance of Cupriavidus metallidurans is based on determinants that were acquired in the past by horizontal gene transfer during evolution. Some of these determinants encode transmembrane metal efflux systems. Expression of most of the respective genes is controlled by two-component regulatory systems composed of a membrane-bound sensor/sensory histidine kinase (HK) and a cytoplasmic, DNA-binding response regulator (RR). Here, we investigated the interplay between the three closely related two-component regulatory systems CzcRS, CzcR2S2, and AgrRS. All three systems regulate the response regulator CzcR, while the RRs AgrR and CzcR2 were not involved in czc regulation. Target promoters were czcNp and czcPp for genes upstream and downstream of the central czc gene region. The two systems together repressed CzcRS-dependent upregulation of czcP-lacZ at low zinc concentrations in the presence of CzcS but activated this signal transmission at higher zinc concentrations. AgrRS and CzcR2S2 interacted to quench CzcRS-mediated expression of czcNp-lacZ and czcPp-lacZ. Together, cross talk between the three two-component regulatory systems enhanced the capabilities of the Czc systems by controlling expression of the additional genes czcN and czcP. IMPORTANCE Bacteria are able to acquire genes encoding resistance to metals and antibiotics by horizontal gene transfer. To bestow an evolutionary advantage on their host cell, new genes must be expressed, and their expression should be regulated so that resistance-mediating proteins are produced only when needed. Newly acquired regulators may interfere with those already present in a host cell. Such an event was studied here in the metal-resistant bacterium Cupriavidus metallidurans. The results demonstrate how regulation by the acquired genes interacts with the host’s extant regulatory network. This leads to emergence of a new system level of complexity that optimizes the response of the cell to periplasmic signals.

cycle enzymes. Surprisingly, most recessive metal resistance determinants are maintained, although their central components had been inactivated by mutation (15). This indicates that some of their genes may be required in the natural environment and under laboratory conditions. This appears to be the case for genes encoding membrane fusion proteins, outer membrane factors, and two-component regulatory systems on these recessive determinants, which were expressed under nonchallenging conditions or upregulated under challenging conditions. Could it be that metal resistance in C. metallidurans is a mosaic phenotype mediated by the dominant metal resistance determinants that interact with a few genes from the otherwise inactivated recessive determinants (15)? In this way, heterologous transenvelope efflux systems may exist or indeed cross talk between two-component regulatory systems could be involved, as has been described before in Escherichia coli (23). The aim of this study is to test this hypothesis.
Two-component regulatory systems are usually composed of a membrane-bound sensor(y) histidine kinase (HK) and cytoplasmic DNA-binding response regulator (RR), which is phosphorylated upon a signal being sensed by the HK (24). The RR dimerizes within seconds after the HK has received the stimulus and activates transcription (25). In C. metallidurans, the dominant czc resistance determinant ( Fig. 1 and see Fig. S1 in the supplemental material) encodes the HK CzcS and the RR CzcR. The genes czcRS are part of the predicted operon Op1819f_1 czcCBADRSE on plasmid pMOL30, which is transcribed together with the czcI gene upstream of czcC from three transcriptional start sites (TSSs), one reliant on the housekeeping sigma factor RpoD, the other two not; 4 weak TSSs upstream of czcI and within the operon also exist. CzcD is a member of the CDF (cation diffusion facilitator) family of inner membrane secondary metal efflux systems and influences together with the periplasmic zinc-and copper-binding protein CzcE via CzcS czc expression (26)(27)(28). CzcI quenches the activity of the CzcCBA pump to prevent "overpumping" of essential metal cations such as Co(II) (22).
Many czc transcripts have been identified (Fig. S1), for instance, a czcI monocistronic message, czcICBA, and czcDRS (27,29). The czcCBADRSE central czc region ( Fig. 1; Fig. S1) is also expressed under nonchallenging conditions, and most genes are additionally upregulated under metal stress. The gene czcN upstream and czcP, coding for a zinc-exporting P IB4 -type ATPase, downstream of the central czc determinant and separated from it by a transposon insertion were only weakly expressed under nonchallenging conditions (nucleotide activities per kilobase of exon model per million mapped reads [NPKM] value of 8 with 10 usually serving as a threshold; NPKM is a measure of the RNA abundance). Deletion of czcP decreases the zinc resistance of C. metallidurans in Tris-buffered mineral salts medium by one-third from an 50% inhibitory concentration (IC 50 ) of 3.4 mM to 2.3 mM (30), but deletion of czcS or of czcR decreases resistance not at all (27). While czcP was upregulated under metal stress, czcN was not (15). Transcription of czcN and czcP depends on CzcR, and a binding site of CzcR had been identified at czcNp (27). Although the full picture of the regulation of czc by (i) RpoD, (ii) one or more non-RpoD-type sigma factors, and (iii) CzcRS is not known at this stage, expression of czcN and czcP cannot be upregulated without CzcR.
Intriguingly, genes of partially inactivated, recessive metal resistance determinants that were retained by C. metallidurans in the natural environment and laboratory are membrane fusion proteins, outer membrane factors, and two-component regulatory systems (15). This suggests fine-tuning of the activity of the dominant determinants by periplasmic signals or a function of proteins located in this compartment. We selected CzcR as a target to ask if CzcR activity as a regulator of two czc promoters might be under the control of cross-talking histidine kinases, CzcS and another HK encoded by recessive determinants. To that end, two-component regulatory systems from C. metallidurans were selected to investigate a possible cross talk with CzcRS. The results demonstrated that a cross talk indeed exists. It influences CzcR activity, especially at low zinc concentrations, to quench expression of czcP for an additional zinc-exporting P IB4type ATPase and of czcN upstream of the main czc determinant. This might be useful to decrease loss of the essential trace element zinc at low external zinc concentrations and explains why the recessive determinants were retained under environmental and laboratory conditions.

RESULTS
Selection of possible cross-talking two-component regulatory systems in C. metallidurans. The genome of C. metallidurans contains 160 genes for response regulators or sensor proteins, forming 42 gene regions with at least one gene for a response regulator and one for a sensory histidine kinase in close proximity. Among the predicted proteins for 29 histidine kinases, clustering of these proteins from C. metallidurans with CusS from E. coli (red) and an adjacent triple group containing CzcS (orange) (see Fig. S2 in the supplemental material) suggested an involvement in upregulation of copper or zinc resistance genes, respectively (31). The CusS and the CzcS clusters were related to YedV from E. coli and two more HKs from C. metallidurans. Another group of 10 HKs around QseC from E. coli (green) contained ZneS from C. metallidurans, which is encoded as part of the znezni metal resistance determinant (15,29). While some of these HKs and RRs were associated with the genes for active (CzcRS, ZneRS, ZneR 2 S 2 , ZniRS, CopRS, and CopR 2 S 2 ; Table  S1, genes on a light green field) resistance determinants, other were affiliated with inactive (CzcR 2 S 2 , HmzRS; light red field) genes for transmembrane metal efflux systems (1,18,29). Except for ZneR 2 S 2 , the genes for these eight systems responded to metal stress (response of $5; Table S1). CzcRS, CopR 2 S 2 , HmzRS, Rmet_5797/98, and ZniRS were also expressed under nonchallenging conditions with an NPKM value of $10 for both genes (encoding the HK and RR), indicating a housekeeping function. ZneR 2 S 2 , not responding to metal stress, was also expressed in nonchallenged cells. Other HKs encoded in metal resistance determinants such as ZneS 2 and ZniS were not closely related to CzcS from C. metallidurans (Fig. S2).
A multiple alignment of the RRs of metal resistance determinants in C. metallidurans displayed a related pair, CopR 1 and CopR 2 (red), while CzcR, CzcR 2 , and Rmet_1751 were less related to the CopR pair (Fig. S3, orange). Other RRs clustered around ZneR (green). The two-component regulatory systems in C. metallidurans with the highest probability for a metal-dependent cross talk were the systems CzcRS, CzcR 2 S 2 , and Rmet_1751/Rmet_1752. The respective three HKs form a cluster of 3 proteins (Fig. S2, orange) distant from HKs involved in copper-dependent regulation (Fig. S2, red). The two-component regulatory system Rmet_1751/Rmet_1752 is involved in a spontaneous development of silver resistance in C. metallidurans (32) and was renamed AgrRS. Consequently, possible cross-talking twocomponent regulatory systems in C. metallidurans involved in fine-tuning of zinc homeostasis could be CzcRS, CzcR 2 S 2 , and AgrRS. These three systems were selected for further investigation.
Transcription of the czc region. Transcriptional organization of the czc region on plasmid pMOL30 has been extensively studied (15,26,27,29,33,34), and the data are summarized in Fig. 1 and Fig. S1. Several promoters upstream of czcI are responsible for transcription of czcICABDRS. The determinant is expressed even in nonchallenged cells and responds strongly to changes in metal availability.
Transcription of czcN in comparison to the other czc genes was analyzed in detail using reverse transcriptase quantitative PCR (qRT-PCR) in the parent strain AE128, a derivative of CH34 with only plasmid pMOL30 (2), and its isogenic DczcR, DczcS, and DczcP mutants ( Table 1). The czcN gene did not respond to metal stress in gene array experiments (Fig. S1); however, in these experiments, C. metallidurans cells were challenged with a metal mixture containing 30 mM (each) Co(II), Ni(II), Cu(II), Zn(II), and Cd(II) (33). This may have been too low a zinc concentration to induce upregulation of czcN. Instead of the metal mix, the cells were confronted with 300 mM Zn(II), and a qRT-PCR experiment was performed. At this concentration, czcN and the other tested czc genes were clearly upregulated in the presence of Zn(II) in C. metallidurans AE128 (Table 1).
In contrast to the parental strain AE128, czcN was not upregulated by treatment for 10 min with 300 mM Zn(II) in the DczcR and the DczcS mutants (Table 1), while all other czc genes were still upregulated in both mutants. In the DczcP mutant, which does not produce the zinc-exporting P IB4 -type ATPase CzcP (30), czcN was again upregulated. Expression of czcN was clearly CzcRS dependent, but the expression of other czc genes in operon Op1819f_1 from czcI to czcE was not, because these genes rely on the czcI promoters ( Fig. 1). Expression of czcN occurred only at high zinc concentrations.
Control of expression of czcPp-lacZ on plasmid pMOL30. The lacZ gene was inserted downstream of the czcP gene on plasmid pMOL30, which has a copy number of 1 compared to the chromosome (35). Time-and concentration-dependent upregulation of czcP-lacZ expression was determined ( Table 2, Fig. 2, and Fig. S4). As previously observed (30), upregulation of czcP was strictly zinc and czcR dependent (Fig. 2). First, expression of czcP essentially required the zinc-dependent activation of CzcR, probably by phosphorylation (24). Second, none of the other RRs in C. metallidurans related to CzcR (Fig. S3) were able to substitute for CzcR. These RRs were not relevant for further discussion.
In the DczcS mutant, expression of czcP-lacZ was twice as strong as that in the parent. In the DczcS, DagrRS, and DczcR 2 S 2 strains, the maximum reporter activity was reached at 100 mM Zn(II), and in the parent, it was reached at 150 mM ( Fig. 2A). The time-dependent increase in reporter activity in the DczcS mutant was 5.2 times higher than in the parent, but the regulatory response displayed a lag of 34 min ( Table 2). Other HKs were able to phosphorylate CzcR, which took about 30 min before activation of transcription began. This lag also explains why for the other CzcR-regulated gene, czcN, no upregulation was observed in the DczcS mutant when analyzed by qRT-PCR after only 10 min (Table 1).
In the two double mutant DczcS DagrRS and DczcS DczcR 2 S 2 strains, the zinc-dependent expression level of czcP-lacZ was between those of the DczcS mutant and the parental strain AE128 (Fig. 2B). The time-dependent increase in reporter activity was 3.9-and 4.7-fold higher in these double mutants than in the parent ( Table 2) but with  a The cells were incubated with or without 200 mM Zn(II), and the specific beta-galactosidase activity was measured after the indicated time (see Fig. S4 in the supplemental material). Compared to Table 1, the inducing zinc concentration was lowered from 300 mM to 200 mM because the response of the czcPp-lacZ fusion was already past the maximum at 300 mM zinc (Fig. 2).  a lag of 1 h. Expression of czcP-lacZ in the DczcS DagrRS DczcR 2 S 2 triple mutant was not higher but was on a similar level as the parent at zinc concentrations up to 75 mM and on a lower level at higher zinc concentrations (Fig. 2C). The increase in reporter activity in the triple mutant was 3.3-fold higher than in the parental strain, and the lag phase was again about 1 h (Table 2). This means that at least one other two-component system was able to activate CzcR. Second, AgrRS and CzcR 2 S 2 were responsible for the high activation level of czcP-lacZ in the absence of CzcS. Third, activation of CzcR by the other systems was delayed while activation in the presence of CzcS was a rapid process (Table 2). Finally, CzcS 2 and AgrS cooperated to reduce the lag time of activation by approximately 50%.
In the presence of CzcS, expression of czcP-lacZ in the DczcR 2 S 2 mutant was similar to that in the parental strain up to concentrations of 75 mM Zn(II) and on a slightly lower level at higher zinc concentrations, while that of the DagrRS mutant was lower at all concentrations than the parental strain ( Fig. 2A). The increase in reporter activity was lower than in the parent, and no lag could be observed ( Table 2). The increased expression observed in the DagrRS DczcR 2 S 2 double mutant was similar to that in the parental strain (Table 2) but started at a higher level in the zinc-dependent measurement when no zinc was added. This paralleled the expression profile measured for the parental strain up to 75 mM Zn(II) but reached a lower overall expression level at higher zinc concentrations (Fig. 2C). AgrRS and CzcR 2 S 2 together repressed CzcRS-dependent upregulation of czcP-lacZ at low zinc concentrations in the presence of CzcS but activated this signal transmission at higher zinc concentrations. Both needed to interact for this purpose. In the absence of the one of the systems, and at low zinc concentrations up to 75 mM, CzcR 2 S 2 alone had a small activating effect, which was observed only in the time-dependent experiments, while this effect was evident for AgrRS in the time-and the concentration-dependent experiments. AgrRS and CzcR 2 S 2 thus interacted with CzcRS to control expression of czcP via CzcR.
Expression of czcNp-lacZ cloned on plasmid pVDZ92. The DNA region between the 39 end of czcN and the beginning of czcI includes several transcriptional start sites and a possible open reading frame, Rmet_6485, encoding a protein with unknown function (Fig. S1). Insertion of lacZ downstream of czcN might disturb expression of czc by causing deleterious effects. Consequently, czcNp was cloned upstream of a lacZ gene on plasmid pVDZ92, a derivative of plasmid RP4/RK2, which should have a 3-to 5-fold-higher copy number than plasmid pMOL30 (36). Due to this copy number, titration of the regulators can occur, as has been shown to occur in the Fur titration assay (FURTA) (37). Thus, the beta-galactosidase activity should be higher than the fusion on plasmid pMOL30. The negative controls indeed showed a high expression level, even in the promoterless control. Again, no zinc-dependent upregulation occurred in the absence of CzcR (Fig. S5), as previously reported (26).
The time-dependent expression of czcNp-lacZ in plasmid pVDZ92 was strictly zinc dependent (Fig. 3). In the presence of CzcS (Fig. 3A), zinc-mediated upregulation of czcNp-lacZ expression in the DagrRS and the DczcR 2 S 2 single mutants paralleled the profile of the parental strain on a slightly higher expression level so that these systems alone activated CzcRS-mediated expression of czcNp-lacZ to a small degree. The DagrRS DczcR 2 S 2 double mutant started from a higher basic expression level but reached the expression level of the parental strain after 1 h. In the absence of zinc, AgrRS and CzcR 2 S 2 interacted to quench CzcRS-mediated expression of czcNp-lacZ, similarly to czcPp-lacZ.
While in all single, double, and triple mutants bearing a DczcS mutation expression of czcP-lacZ on plasmid pMOL30 started after a lag and increased subsequently following a linear function ( Table 2; Fig. S4), the zinc-dependent expression profiles of czcNp-lacZ on plasmid pVDZ92 followed a polynomial function "y = a 1 bt 1 ct 2 " (Fig. 3B). The a and b coefficients of the four mutants were not different (Table S2) whereas the c values increased, compared to the triple mutant, 1.8-fold in the DczcS DczcR 2 S 2 mutant, 2.4-fold in the DczcS DagrRS mutant, and 3-fold in the DczcS single mutant. With this result in mind, the data from the czcP-lacZ fusion in the DczcS mutants were also fitted to a polynomial function (Table S2). Again, the a values of the four mutants were similar. The b values had large deviations, indicating that a linear function with a lag (Table 2) delivered a better modeling of the data than a polynomial function. The b value decreased from the DczcS mutant to the triple mutant in both models. The c values of the double and triple mutants were similar, the c value for the DczcS mutant being half of this value, reminiscent of the doubling of the lag from the single to the double and triple mutants. Since the first derivative of a second-grade polynomial function at time still depends on the time, both a polynomial fitting and a linear fitting with a lag described the data by different modeling approaches. These two sides of the same coin demonstrated that activation of CzcR by the other histidine kinases was a delayed process. Possible explanations could be a lower phosphorylation rate of CzcR by CzcS 2 and AgrS than by CzcS or an upregulation of czcR 2 S 2 and agrRS expression that had to occur before CzcR could be efficiently activated.
The time-dependent czcNp-lacZ experiments indicated also a higher contribution of CzcR 2 S 2 to expression in the absence of CzcS than of AgrRS, while that of the third, unknown system was minor (Fig. 3B). This was also demonstrated by the concentration-dependent expression profiles of czcNp-lacZ on plasmid pVDZ92 (Fig. 4). In the DczcS single mutant (Fig. 4A), the expression level was higher than that of the parental strain at zinc concentrations up to 200 mM. No upregulation occurred in the DczcS DagrRS DczcR 2 S 2 triple mutant. The level of the DczcS DczcR 2 S 2 mutant was barely above the level of the triple mutant while that of the DczcS agrRS mutant was between those of the triple mutant and the parental strain. AgrRS and CzcR 2 S 2 , together, were able to activate czcNp-lacZ expression, and the contribution of CzcR 2 S 2 was higher than that of AgrRS (Fig. 4A). In the presence of CzcS, the expression profile of the DczcR 2 S 2 mutant was similar to that of the parent and that of the DagrRS mutant was similar to that of the DagrRS DczcR 2 S 2 double mutant. In both DagrRS-containing strains, the czcNp-lacZ expression level was higher than that of the parent (Fig. 4B). In the presence of CzcS, AgrRS quenched zinc-dependent czcNp-lacZ expression while CzcR 2 S 2 did not interfere. In the absence of CzcS, AgrRS and CzcR 2 S 2 interacted to mediate expression of czcNp-lacZ after a lag period with a stronger contribution coming from CzcR 2 S 2 than from AgrRS.
Expression of czcIp-lacZ on plasmid pVDZ92 was on an even higher level of reporter activity than that of czcNp-lacZ (Fig. S6). The differences between the mutants and their parent were small with a slightly higher expression level of the DczcS mutant at low zinc concentration than of the parent and a lower expression level of the two double mutants and the triple mutant carrying a DczcS mutation. This finding was in agreement with an observed upregulation of the genes from czcI to czcE (Table 1) and all czc structural genes (27) in the DczcS mutant. It also indicated that AgrRS and CzcR 2 S 2 interacted to influence regulation of expression of the czcI promoter region with its minimum of 4 transcriptional start sites ( Fig. 1; Fig. S1).
Identification of the czcN promoter. Previously identified possible transcriptional start sites for czcN are located 643 bp, 834 bp, 844 bp, and 1,395 bp upstream of czcN (29). This is further upstream of czcN than the beginning of the 246-bp fragment that was cloned in plasmid pVDZ92, which displayed a clear metal-dependent upregulation of the lacZ reporter ( Fig. 3 and 4) (29) was annotated to the incorrect open reading frame due to repeated changes in the annotation of the C. metallidurans genome. This TSS was not assigned to the RpoD sigma factor. Its activity score under nonchallenging conditions is 37.3 6 9.3 (29) and represents a weak activity. In a new RNA sequencing and TSS determination experiment, C. metallidurans CH34 was treated with a modified metal mixture (33) and either with 50 mM EDTA or with no addition. RNA was isolated and used to determine the transcriptional start sites. In this experiment, the score of TSS_7414114 in RNA from metal-challenged cells was 60,069 6 10,166, in nonchallenged cells it was 232 6 42, and in EDTA-treated cells it was 236 6 43. There was no upregulation of czcNp in the presence of the metal chelator EDTA but a 259-fold upregulation under conditions of metal stress. The 246-bp fragment thus carries the czcNp promoter, which had previously not been identified.

DISCUSSION
Accumulation of genes by the czc metal resistance determinant on plasmid pMOL30 mediated an increase in complexity. Plasmid pMOL30 of C. metallidurans strain CH34 is a horizontally acquired replicon that provides sophisticated copper cop 1 and cobalt-zinc-cadmium resistance czc determinants to its host, in addition to another mercury (mer), a lead (pbr), and an inactivated nickel-cobalt-cadmium (ncc) resistance determinant (1,(38)(39)(40). The cop 1 and czc determinants contain more genes than paralogous determinants on the chromosome and chromid. Plasmid pMOL30 will be maintained by C. metallidurans only if the plasmid-carried determinants provide a better function than the chromosomal or chromid paralogs, so that the plasmid-carried determinants become dominant over the others.
The paralogous chromid-carried czc 2 determinant of C. metallidurans displayed clear signals of a recessive determinant (15). Originally, a zntA,.czcI 2 C 2 B 2 A 2 , czcR 2 S 2 determinant as found in related bacterial strains (11,16) was interrupted within the czcB 2 genes. Subsequently, the two parts of the determinant were separated from each other on the chromid by a rearrangement of this replicon. ZntA is a P IB2 -type ATPase and the major zinc-exporting inner membrane efflux system of C. metallidurans (30,41). The zntA gene is under the control of an RpoD-dependent promoter and the MerR-type regulator ZntR (22). This fate of czc 2 as a result of acquisition of plasmid pMOL30 with its czc determinant by C. metallidurans indicates that czc provides a function to C. metallidurans that was superior to that mediated by czc 2 . The additional benefit should be encoded by those genes of czc that czc 2 does not contain, namely, czcD, czcE, czcJ, czcP, and czcN. Among these beneficial genes, czcP and czcN are expressed under the control of CzcRS and its cross-talking partners, while czcD and czcE flank czcRS and influence the activity of CzcRS (26).
The central czcICBADRSE region of czc ( Fig. 1) is upregulated when the cells are challenged by high metal concentrations, with zinc being the best inducer, followed by cobalt and cadmium (26,34). Depending on the method used, czc allows an IC 50 of zinc of about 3.4 mM in liquid culture (30) or a MIC of 12 mM on solid medium (2). On the other hand, czc is expressed even under nonchallenging conditions in Tris-buffered mineral salts medium (see Fig. S1 in the supplemental material). The zinc content of this medium is 200 nM (42). Despite this low zinc concentration and expression of czc in these cells, they are able to obtain sufficient zinc to grow. This changes when the central zinc importer ZupT is deleted. In this case, pMOL30 is rapidly cured from the cells (42). A forced expression of czcCBA from a plasmid in trans results in a disappearance of the central CzcA protein from the cells despite the presence of its mRNA. This suggests either that translation of the mRNA is impeded or that degradation of CzcA occurs. This indicates that the pMOL30-carried czc determinant has the ability to interact indirectly with ZupT and the other components of the Zur regulon (43)(44)(45) so that zinc homeostasis is maintained from 200 nM to the lower-millimolar range of zinc concentrations, covering a 10,000-fold difference when considering only the zinc concentration. Additionally, czc mediates homeostasis of the minor bioelements cobalt and the toxic-only cadmium. All the additional genes of czc could be involved in this process, bringing a greater advantage to the cells than czc 2 . Since czc mediates resistance to three metal cations, one toxic only and two essential but toxic, expression of czc also needs a high level in sophistication of its regulation.
Under nonchallenging conditions, four transcriptional start sites were identified 284 bp, 97 bp, 53 bp, and 31 bp upstream of czcI (29). The highest-abundance mRNAs initiated transcription at the 53-bp position, and this start site was also identified previously by primer extension (27). The associated promoter displayed a medium-strong consensus sequence for the housekeeping sigma factor RpoD. The abundance of mRNAs with 59 ends at the 97-bp and 31-bp upstream positions was about 10% of that of the 53-bp 59 mRNA, and both appear not to be RpoD promoters. Other sigma factors contribute to czcI expression. The remaining start site most distal to czcI had a 2% abundance but displayed a strong consensus motif for RpoD-dependent promoters.
Three additional start sites within the czcICBADRSE region contributed to a low abundance (2% to 3% of that from the 53-bp position) of mRNAs with 59 ends at these sites, which were located 32 bp upstream of czcA, 1,003 bp upstream of czcD, and 37 bp upstream of czcE, respectively. None was an obvious RpoD-dependent promoter (29), so that other sigma factors might initiate transcription of parts of czc from these promoters (Fig. S1, white arrowheads). A 59 mRNA end 223 bp upstream of czcC was identified by primer extension but not by transcriptome sequencing (RNA-Seq) under nonchallenging conditions (27,29). Provided such a start site cannot be found under challenging conditions, the appearance of czcI and czcICBA mRNAs in Northern blots (27) and the decrease in mRNA abundance from czcI to czcC (Fig. S1) would indicate an mRNA cleavage site 223 bp upstream of czcC, which is within the czcI gene. No transcript continued between czcA and czcD, or between czcS and czcE in zinc-treated cells, in agreement with stem-loop structures downstream of czcA and czcS that may act as transcriptional terminators (27). The abundance of mRNAs starting 1,003 bp upstream of czcD and 37 bp upstream of czcE was 36 6 4 and 26 6 5 in nonchallenged cells, respectively, which fits the NPKM values for czcD of 26.3 6 0.6 and for czcE of 26.3 6 3.1. The promoter sequences associated with both start sites were clearly not RpoD-dependent promoters (29), so that czcDRS and czcE were expressed as a tricistronic and monocistronic message, respectively, under the control of one or two non-RpoD-dependent RNA polymerase holoenzymes. This indicates that the predicted operon Op1819f_1 (Fig. S1) contains in fact three transcriptional units, czcCBA, czcDRS, and czcE.
Upstream of the lacZ reporter gene on plasmid pVDZ92, the czcI promoter region mediated a high beta-galactosidase activity, which is upregulated about 30% by increasing zinc concentrations. CzcS and CzcR are not needed for upregulation (26), and the influence of CzcS and the other two histidine kinases was small (Fig. S6). Consequently, CzcR and CzcS do not control expression of the central czcICBA gene region. Transcription of this core part of czc is guaranteed by the housekeeping sigma factor RpoD and additionally by at least one non-RpoD sigma factor. Instead, CzcR and CzcS influence expression of flanking genes czcN and czcP. On the next level, expression of czcR and czcS, as part of the tricistronic czcDRS mRNA, is under the control of a non-RpoD sigma factor (Fig. S7, fields with a red or blue surrounding for exclusively non-RpoD control or control by both, non-RpoD and RpoD, respectively), for instance, the RpoH heat shock factor, the RpoS stationary-phase factor, or one of the 11 sigma factors of the extracytoplasmic function (ECF) family (33,46,47). Since the consensus sequences for all these sigma factors have not yet been identified, anything between one non-RpoD sigma factor controlling both non-RpoD-dependent czcI promoters, czcDp and czcEp, and four sigma factors controlling each one of these promoters can be assumed. This third pillar of metal homeostasis (33) regulates expression of czcICBA, of czcDRS, and of czcE, and subsequently, CzcD, CzcR, CzcS, and CzcE control transcription initiation of czcN and czcP. Finally, the cross talk between the two-component regulatory systems was now added to the model (Fig. S7).
Function of the products of the czcICBA gene region. The large transmembranespanning CzcCBA protein complex is at the core of the Czc system. There is clear evidence that the RND protein CzcA and the related copper transporter CusA transport their substrates in vitro across a membrane that would correspond to the inner or cytoplasmic membrane (48)(49)(50). But there is also accumulating evidence that CzcCBA and CusCBA export their substrates in vivo mainly from the periplasm through the outer membrane to the outside (30,(51)(52)(53)(54). Metal-binding sites of CzcA in the cytoplasm and CzcB in the periplasm might be involved in flux control that prevents export of zinc by CzcCBA under conditions of low zinc availability (55,56). The periplasmic CzcI protein quenches CzcCBA activity with respect to the essential minor bioelements zinc and cobalt, but not for the toxic-only cadmium (22). CzcI thus prevents an overactive efflux of essential periplasmic zinc and cobalt ions at low concentrations of these ions, allowing the presence of the CzcCBA transmembrane efflux complex under these conditions. CzcCBA has not to be degraded and stands ready in case of a sudden increase in the cobalt, zinc, and cadmium concentrations. However, the zinc importer ZupT is needed to compete with CzcCBA for periplasmic zinc ions. That way, ZupT guarantees zinc import despite the presence of CzcCBA, which explains why cells cannot keep CzcCBA when ZupT is absent.
The products of the czcDRSE region control expression of czcN and of czcP. The czcP gene encodes a P IB4 -type zinc-exporting ATPase (30,57,58) that exports loosely bound cytoplasmic zinc ions with a high transport rate, while ZntA effluxes firmly bound ions with a lower transport rate. CzcP thus cannot provide zinc resistance without ZntA or one of the other two P IB2 -type ATPases CadA and PbrA in C. metallidurans but is able to enhance resistance mediated by ZntA. The czcP gene is expressed only on a low level in nonchallenged cells but upregulated under metal stress ( Table 2, Fig. 2, and Fig. S4) (30). A transcriptional start site is located 41 bp upstream of czcP and is not RpoD dependent (29).
Expression from czcNp and czcPp strictly and exclusively depends on the response regulator CzcR and on Zn(II) as inducer (Tables 1 and 2, Fig. 2, and Fig. S4 and S5) (26,30). Missing upregulation in a DczcR mutant can be complemented with czcR in trans on a plasmid. CzcR binds to the respective promoter regions. While czcNp contains one binding site, czcPp possesses two (27,30). This agrees with a maximum level of expression from czcPp at 150 mM Zn(II) (Fig. 2) compared to a maximum level reached for czcNp at 500 mM (Fig. 4). This indicates that a cooperative effect of two CzcR dimers bound to czcPp may activate czcP expression at lower zinc concentrations than one CzcR dimer bound to czcNp. Rapid export of loosely bound cytoplasmic Zn(II) by CzcP is needed at lower zinc concentrations than the function provided by CzcN.
Cross-talk between two-component regulators embeds the Czc system into metal homeostasis of its host. A cross talk between the response regulators YedW and CusR in E. coli (59), CopRs and CzcR in Pseudomonas stutzeri (60), and other response regulators (61) has been shown. In C. metallidurans, the response regulators closely related to CzcR (Fig. S3) are not able to substitute for CzcR. Instead, CzcS 2 and AgrS converge on CzcR, which is reminiscent of the histidine kinases RocS 1 and RocS 2 that act on the response regulator RocA 1 in Pseudomonas aeruginosa to control expression of the cupC gene involved in copper resistance. Both sensors also act on the response regulator RocA 2 to repress expression of the mexAB-oprM genes, which encode a transenvelope efflux system for organic substances instead of metal ions as the substrate (62). CzcR 2 S 2 might control expression of the recessive czc 2 determinant, more precisely the genes czcI 2 C 2 . These genes and czcR 2 S 2 are expressed under RpoD control, czcI 2 C 2 from a 134-bp common promoter region but in the opposite direction of transcription from zntA (29). Expression of zntA is activated by the MerR-type regulator ZntR (22), and the RpoD-dependent transcriptional start site is located 12 bp upstream of zntA (29). The transcriptional start site of czcI 2 , which is also RpoD dependent, is directly at the 59 end of czcI 2 , indicating lmRNA-specific translation initiation (leaderless mRNA, [63]). This leaves sufficient distance between the promoters to allow simultaneous binding of ZntR and CzcR 2 so that the RpoD-dependent RNA polymerase holoenzymes attracted to both promoters by both activators may not automatically interfere with each other. On the other hand, this also cannot be excluded. CzcI 2 interferes with CzcCBA, similar to CzcI, but in a slightly different manner (22). Expression of czcC 2 is strongly upregulated by metals (15). When czcBA fragments with DczcC deletions were expressed in trans in the plasmid-free strain AE104, this strain maintained zinc resistance and the ability to efflux zinc compared to a strain expressing czcCBA (64). Since an outer membrane factor such as CzcC is an essential part of the transenvelope efflux system (65-68), CzcC 2 was probably able to substitute for CzcC in these experiments. Because cadmium and cobalt resistance was not maintained, CzcC 2 may have a higher selectivity for zinc over cobalt and cadmium than CzcC so that an upregulation of czcI 2 C 2 may be an advantage under conditions of high zinc availability.
A spontaneous mutation in agrRS results in increased silver resistance of C. metallidurans (32). The agrRS genes are transcribed in the opposite direction from the agrABC genes for a transenvelope efflux system from a common promoter region. Two promoters 39 bp and 76 bp upstream of agrRS are non-RpoD dependent, but the promoter for agrABC depends on the housekeeping sigma factor (29). It is located 45 bp within agrA so that the 59 end of the gene might have been misannotated. Other promoters are far upstream of agrABC within agrRS and might cause an antisense effect, as has been observed for many transcriptional events in C. metallidurans (29). The RND protein AgrA is related neither to the metal-transporting RND proteins of the HME-RND protein family such as CzcA or CusS nor to an AcrB-like transporter for organic substances (Fig. S8). It clearly shows the conserved EN motif at the end of a transmembrane alpha-helix that is important for proton transport. RND proteins involved in transport of divalent transition metal cations possess within this alpha-helix a DFG motif followed by a conserved aspartate 3 positions downstream, which is essential for proton transport (48). Transporters for monovalent cations exhibit an AVG instead of the DFG. In RND proteins for organic substance, an AIG is followed 3 positions downstream by a double aspartate instead of just one (54,69). AgrA also has the DD signature of organic substrates plus an AVG upstream and a potential metal-binding site, HHRE, downstream of the EN motif in a cytoplasmic part of the RND protein, reminiscent of metal-binding sites in CusAs, CzcA, and SilA from C. metallidurans (Fig. S8). AgrA shows hybrid features of an organic and a metal cation transporter. It may transport a metal complex, but an upregulation of agr genes under different conditions of metal availability has not yet been observed (15). The most important contribution of agr is thus the quenching of CzcR activation under low-zinc conditions.
Signaling by two-component regulatory systems accelerates with HK expression but decelerates with RR expression (25). Expression of czcS and of czcS 2 is upregulated when metal availability changes but not that of czcR (Table S1). Since transcription of czcDRS and of czcE is under the control of one or two non-RpoD sigma factors and CzcD and CzcE interact with CzcS, a complicated network is in control of the expression of czcN and czcP (Fig. S7).
Conclusion. Together, these data demonstrated a cross talk between AgrRS, CzcR 2 S 2 , and CzcRS involved in control of the czc promoters czcNp and czcPp via CzcR. This cross talk probably used the periplasmic zinc concentration as a signal to regulate expression of czcP and czcN. In the presence of CzcS, AgrRS and CzcR 2 S 2 quenched activation of CzcR at low zinc concentrations, with AgrRS being more important than CzcR 2 S 2 . In the absence of CzcS, both cross-talking systems mediated activation of CzcR, albeit after a lag phase. For czcP-lacZ in a single-copy environment, the two cross-talking systems contributed equally with an maximum of activation at 100 mM Zn(II) compared to 150 mM Zn(II) in the parent. For czcNp-lacZ on a plasmid with a higher copy number, CzcR 2 S 2 contributed more to activation of CzcR in the absence of CzcS than AgrRS; the maximum was at 200 mM Zn(II) compared to 500 mM in the parent.

MATERIALS AND METHODS
Bacterial strains and growth conditions. Plasmids and C. metallidurans strains are provided in Table S3 in the supplemental material. Tris-buffered mineral salts medium (2) containing 2 g sodium gluconate/L (TMM) was used for C. metallidurans under aerobic conditions at 30°C.
Genetic techniques. Standard molecular genetic techniques were used (70,71). For conjugative gene transfer, overnight cultures of donor strain E. coli S17/1 (72) and of the C. metallidurans recipient strains grown at 30°C in Tris-buffered medium were mixed (1:1) and plated onto nutrient broth agar. After 2 days, the bacteria were suspended in TMM, diluted, and plated onto selective medium as previously described (70).
Plasmid pECD1002, a derivate of plasmid pCM184 (73), was used to construct deletion mutants in C. metallidurans. These plasmids harbor a kanamycin resistance cassette flanked by loxP recognition sites. Plasmid pECD1002 additionally carries alterations of 5 bp at each loxP site. Using these mutant lox sequences, multiple gene deletions within the same genome are possible without interference by secondary recombination events (74,75). Fragments of 300 bp upstream and downstream of the target gene were amplified by PCR, cloned into vector pGEM-T Easy (Promega), sequenced, and further cloned into plasmid pECD1002. The resulting plasmids were used in a double-crossover recombination in C. metallidurans strains to replace the respective target gene with the kanamycin resistance cassette, which was subsequently also deleted by transient introduction of cre expression plasmid pCM157 (73). Cre recombinase is a site-specific recombinase from the phage P1 that catalyzes the in vivo excision of the kanamycin resistance cassette at the loxP recognition sites. The correct deletions of the respective transporter genes were verified by Southern DNA-DNA hybridization. For construction of multiple deletion strains, these steps were repeated. The resulting mutants carried a small open reading frame instead of the wild-type gene to prevent polar effects.
b-Galactosidase assay and lacZ reporter constructions in C. metallidurans. To construct reporter operon fusions, a respective promoter region was cloned together with the lacZ gene in plasmid pVDZ92 as described before (26). Alternatively, the lacZ reporter gene was inserted downstream of czcP. This was done by single crossover recombination in C. metallidurans strains. A 300-to 400-bp PCR product of the 39-end region of the respective target gene was amplified from total DNA of strain CH34, and the resulting fragments were cloned into plasmid pECD794 (pLO2-lacZ) (30). The respective operon fusion cassettes were inserted into the open reading frame of the target gene by conjugation and single crossover recombination. C. metallidurans cells with a lacZ reporter gene fusion were cultivated as a preculture in TMM containing 1.5 g L 21 kanamycin at 30°C and 250 rpm for 18 h, diluted to a turbidity of 30 Klett units into fresh medium, and incubated with shaking at 30°C for 3 to 4 h until a cell density of 60 Klett units was reached. This culture was distributed into sterile 96-well plates (Greiner Bio-One, Frickenhausen, Germany). After addition of metal salts, incubation in the 96well plates was continued for 3 h at 30°C in a neoLab DTS-2 shaker (neoLab Migge Laborbedarf, Heidelberg, Germany). The turbidity at 600 nm was determined in a Tecan Infinite 200 Pro reader (Tecan, Männersdorf, Switzerland), and the cells were sedimented by centrifugation at 4°C for 30 min at 4,500 Â g. The supernatant was discarded, and the cell pellets were frozen at 220°C. For the enzyme assay, the pellet was suspended in 190 mL Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 50 mM beta-mercaptoethanol), and 10 mL permeabilization buffer was added (6.9 mM cetyltrimethylammonium bromide [CTAB], 12 mM sodium deoxycholate). The suspension was incubated with shaking at 30°C, and 20 mL ONPG solution (13.3 mM ortho-nitrophenyl-beta-D-galactopyranoside in Z buffer without beta-mercaptoethanol) was added. Incubation was continued with shaking in a neoLab DTS-2 shaker at 30°C until the yellow color of o-nitrophenol was clearly visible and stopped by addition of 50 mL 1 M Na 2 CO 3 . The extinction at 420 nm and 550 nm was measured in a Tecan Infinite 200 Pro reader. The activity was determined as published previously (76)  Specific activity was activity divided by the cellular dry mass as published previously (76). For the time-dependent beta-galactosidase assay, at a cell density of 60 Klett units, metal salts were added up to various final concentrations into tubes and the cells were incubated with shaking for a further 3 h. The specific beta-galactosidase activity was acquired in permeabilized cells as published previously with 1 U defined as the activity forming 1 nmol of o-nitrophenol per min at 30°C: activity U = 355.6 Â E 420 / reaction time (76)(77)(78).
RNA isolation and qRT-PCR. Total RNA was isolated, and the RT reaction was performed as previously described (46). To exclude experimental artifacts resulting from DNA contaminations, only RNA preparations that did not generate products in a PCR with chromosomal primers without a previous RT reaction were used. As an endogenous control, rpoZ was used. A no-template control was performed under identical conditions as for the target genes. An average for two different cDNAs as well as an average for two independent biological examples was calculated. For normalization, transcript levels of rpoZ were used.
TSS determination. RNA was prepared from C. metallidurans CH34 cells cultivated in TMM (Tris-buffered mineral salts medium with 2 g/L gluconate as the carbon source) for three independent biological repeats in the presence and absence of a metal ion mix (33), respectively. The composition of this metal cation mix was modified from the published version for a better representation of the individual toxicity of the respective cation. The total metal ion concentration of the metal ion mix used for C. metallidurans RNA-Seq was performed by Vertis Biotechnology AG (Freising, Germany) using a Cappable-seq protocol for TSS determination (79). The TSSs were trimmed and mapped to the reference genomes CP000352 (chromosome), CP000353 (chromid, also named "megaplasmid"), CP000354 (plasmid pMOL30), and CP000355 (pMOL28), and potential TSSs were annotated as peaks using program tools made available by Laurence Ettwiller (New England Biolabs; https://github.com/Ettwiller/TSS). Since the number of control reads was small compared to the TSS reads, TSSs were calculated without using the control reads. The n io value was the number of reads at position i in orientation o, and N was the total number of mapped reads. The RRS io value for each position and orientation was the reads per million and was defined as RRS io = (n io /N) Â 10 6 for TSS determination and control. For each TSS, the score was RRS io _TSS/RRS io _control. For the TSS determination, a cutoff value of RRS io of 5 and a cluster value of 5 were used, the latter defining the size in base pairs of the upstream and downstream region used for clustering conditions. Only TSSs that appeared in all three biological repeats and had a score of 10 were further considered. Promoter sequences per TSS were extracted as sequence regions 290 to 110 bp around the TSS position on the respective replicon (GenBank accession numbers CP000352.1, CP000353/NC_007974.2, CP000354/NC_007971.2, and CP000355/NC_007972.2) and according to the strand orientation of the TSS. From the resulting database (unpublished data), the mean scores for TSS_7414114 upstream of czcN in RNA from metal-challenged cells, nonchallenged cells, and cells treated with 50 mM EDTA were selected.
Statistics. Student's t test was used, but in most cases the distance (D) value has been used several times previously for such analyses (80)(81)(82). It is a simple, more useful value than Student's t test because nonintersecting deviation bars of two values (D . 1) for three repeats always mean a statistically relevant ($95%) difference provided the deviations are within a similar range. At n = 4, significance is $97.5%, at n = 5, it is $99% (significant), and at n = 8, it is $99.9% (highly significant).

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 2.3 MB.