Copper homeostasis networks in the bacterium Pseudomonas aeruginosa

Bacterial copper (Cu+) homeostasis enables both precise metallation of diverse cuproproteins and control of variable metal levels. To this end, protein networks mobilize Cu+ to cellular targets with remarkable specificity. However, the understanding of these processes is rather fragmented. Here, we use genome-wide transcriptomic analysis by RNA-Seq to characterize the response of Pseudomonas aeruginosa to external 0.5 mm CuSO4, a condition that did not generate pleiotropic effects. Pre-steady-state (5-min) and steady-state (2-h) Cu+ fluxes resulted in distinct transcriptome landscapes. Cells quickly responded to Cu2+ stress by slowing down metabolism. This was restored once steady state was reached. Specific Cu+ homeostasis genes were strongly regulated in both conditions. Our system-wide analysis revealed induction of three Cu+ efflux systems (a P1B-ATPase, a porin, and a resistance-nodulation-division (RND) system) and of a putative Cu+-binding periplasmic chaperone and the unusual presence of two cytoplasmic CopZ proteins. Both CopZ chaperones could bind Cu+ with high affinity. Importantly, novel transmembrane transporters probably mediating Cu+ influx were among those largely repressed upon Cu+ stress. Compartmental Cu+ levels appear independently controlled; the cytoplasmic Cu+ sensor CueR controls cytoplasmic chaperones and plasma membrane transporters, whereas CopR/S responds to periplasmic Cu+. Analysis of ΔcopR and ΔcueR mutant strains revealed a CopR regulon composed of genes involved in periplasmic Cu+ homeostasis and its putative DNA recognition sequence. In conclusion, our study establishes a system-wide model of a network of sensors/regulators, soluble chaperones, and influx/efflux transporters that control the Cu+ levels in P. aeruginosa compartments.

Copper is a prosthetic group of redox enzymes (1)(2)(3). However, Cu ϩ is also toxic to cells, disrupting Fe-S centers and participating in harmful Fenton reactions (4,5). Thus, cells must mobilize the metal ion to specific protein targets while preventing the presence of potentially toxic free Cu ϩ . Our goal is to understand how bacterial cells accomplish Cu ϩ homeostasis, shifting the attention from understanding how cells tolerate this redox active metal to how physiological needs for this ion are satisfied.
The toxicity of free Cu ϩ has been employed in many studies to pinpoint genes responsible for conferring tolerance to external Cu 2ϩ (6 -8). Various cytoplasmic Cu ϩ -sensing transcriptional regulators (CueR, CsoR, and CopY) (9 -11) as well as periplasmic Cu ϩ -sensing two-component systems (CopR/S, CusR/S, PcoR/S, and CinR/S) (7,(12)(13)(14) have been identified. Biochemical analyses have shown that cytoplasmic Cu ϩ -sensing transcriptional regulators bind Cu ϩ with K D in the 10 Ϫ17 to 10 Ϫ21 M range (10,15), suggesting the virtual absence of free Cu ϩ inside cells (16). In most bacteria, these might only regulate the expression of a single Cu ϩ efflux ATPase and a cytoplasmic chaperone (9,17,18). On the other hand, the periplasmic two-component sensor CusR/S regulates expression of the multicomponent trans-periplasmic Cu ϩ transporter CusCFBA, a resistance-nodulation-division (RND) 2 -type transport system (12,19). In any case, studies have not yet provided a clear description of the coordinated work of the independent periplasmic and cytoplasmic metal sensors.
Transmembrane P 1B -type ATPases (CopA) responsible for cytoplasmic Cu ϩ efflux are extensively characterized elements of Cu ϩ tolerance systems. They receive Cu ϩ from the cytoplasmic chaperone (CopZ) and transfer it to periplasmic chaperones (CusF) via protein-protein interactions (20,21). In some bacterial species, an outer membrane porin (PcoB) and the RND-type CusCBA system appear to contribute to periplasmic Cu ϩ efflux (22,23). Overall, these molecular studies have showed that high binding affinities and metal transfer via protein-protein interactions are key mechanisms to confer directional specificity while maintaining the absence of free Cu ϩ . However, the integrated interplay among various transporters present in a given system is not understood. The role of Cu ϩ chaperone proteins mobilizing the metal within compartments has been characterized (24,25). In some organisms, a periplasmic chaperone (CusF) associated with the RND-type efflux system is also present (26,27). On the other hand, organisms apparently lacking cytoplasmic or periplasmic chaperones have been identified (27). Consequently, identification of the complete set of chaperones (Cu ϩ -bind-ing soluble proteins) participating in compartmental metal levels homeostasis would significantly contribute to establish Cu ϩ distribution mechanisms.
Beyond Cu ϩ tolerance, metallation of cuproenzymes seems similarly mediated by Cu ϩ transport and chaperone proteins mobilizing and targeting Cu ϩ via protein-protein interactions (28,29). For instance, SenC, a membrane protein that binds Cu ϩ in its periplasmic domain, acts as a chaperone transferring the metal to cytochrome c oxidase in the plasma membrane (28). However, this process requires the routing of Cu ϩ across the cytoplasm (i.e. influx into the cytoplasm via the CcoA transporter (30) and efflux to the periplasm by a dedicated P-type ATPase (CopA2) (31)). Thus, there is a Cu ϩ distribution network devoted to the process and cytoplasmic control of the events as the targeted Cu ϩ ion is mobilized through this compartment.
As a result of Cu ϩ tolerance studies, a preponderant model composed of a metal-sensing transcriptional regulator, a chaperone, and a plasma membrane efflux ATPase has emerged to explain the fate of Cu ϩ in bacterial cells. Microarray analyses of transcriptional responses to high, or even lethal in some cases, Cu 2ϩ levels have further contributed to this rudimentary model (13,(32)(33)(34)(35)(36)(37). Consequently, there is still a lack of an integrated vision of a system-wide Cu ϩ homeostatic network in a given organism. We have chosen Pseudomonas aeruginosa to search for, identify, and describe novel elements involved in this role. This opportunistic pathogen is a major cause of infection and mortality among immunosuppressed patients and it is among the microorganisms for which there is an urgent need for novel antibiotics (38). BLAST searches in the P. aeruginosa genome using known homologues reveal the presence of Cu ϩ -sensing transcriptional regulators, cuproenzymes, putative Cu ϩ -binding chaperone/storage proteins, and likely Cu ϩ transmembrane transporters (supplemental Table S1). However, the connections among these proteins and their participation in alternative Cu ϩ pathways (metallation of cuproenzymes or tolerance) are not self-evident.
Previous studies have shown numerous genes activated during the P. aeruginosa transcriptomic response to lethal copper concentrations (exposure to 10 mM Cu 2ϩ for 45 min or 6 h) (13). A complex response was described, including the overexpression of various porins, several P-type ATPases, ABC ATPases, RND transporters, and other genes that are probably part of a general response to cell damage. These observations clearly include pleiotropic changes associated with the lethal conditions employed in these studies. More recently, it was reported that P. aeruginosa CueR was activated by the LasR/ LasI quorum-sensing system (17). The transcriptomic response of the ⌬cueR mutant strain to 10 M Cu 2ϩ stress during 3 h under anaerobic conditions revealed a regulon that included copA1, an RND transporter, putative chaperones PA3520 and PA3574.1, and genes PA3515 to PA3519 coding for lyases, methyl transferases, and hypothetical proteins. The functional role or integrated interplay of the coding proteins was not considered.
Here we show the genome-wide transcriptomic response of P. aeruginosa to Cu 2ϩ stress under selected treatment conditions designed to prevent nonspecific pleiotropic effects. RNA-Seq analysis was performed at short treatment times and after Cu ϩ fluxes reached steady-state conditions. Our studies showed the specific up-regulation of Cu ϩ transporters, the uncommon occurrence of two cytoplasmic Cu ϩ chaperones, soluble periplasmic proteins that probably mediate the trafficking to and from Cu ϩ efflux transporters, and the previously unobserved repression of putative Cu ϩ influx transporters. Comparison of the response in ⌬cueR and ⌬copR mutant strains reveals two independent regulons, controlling the distribution networks in the cytoplasmic and periplasmic compartments, respectively. Departing from copper tolerance studies, our transcriptomic and functional data allowed us to generate a model of compartmentally defined and independently regulated Cu ϩ distribution networks. The framework presented below as a summary of these studies suggests that Cu ϩ fluxes and pool sizes should be described considering multicomponent functional parameters beyond the sensor affinities or plasma membrane transport rates.

Cells maintain normal growth rates under a high-intracellular copper steady-state condition
Our goal was to identify specific mechanisms of Cu ϩ homeostasis in P. aeruginosa. We hypothesized that exposure to nontoxic external Cu 2ϩ stress would selectively activate homeostatic mechanisms rather than pleiotropic responses to metal toxicity, such as redox, permeability, osmotic, and metabolic imbalances. Fig. 1A shows the effect of 2-h exposure to high external CuSO 4 concentrations on P. aeruginosa viability. As expected, 1-4 mM CuSO 4 led to significant cell death, and no survival was observed after a 2-h exposure to 4 -6 mM CuSO 4 . Alternatively, 0.5 mM CuSO 4 had no effect on cell viability (Fig.  1A). Importantly, P. aeruginosa survived in the presence of 0.5 mM CuSO 4 , although the levels of intracellular copper increased Ͼ5-fold compared with untreated samples (Fig. 1B). Thus, the not only does the cell achieves homeostasis by extruding intracellular Cu ϩ via efflux transporters, but other mechanisms are also in play.
Based on these observations, it is reasonable to hypothesize a multifaceted cellular response to external copper, with the expression of alternative sets of genes immediately following exposure and upon stabilization of intracellular metal levels (i.e. pre-steady-state and steady-state conditions). Surprisingly, no information could be found in the literature on the kinetics of whole-cell Cu ϩ fluxes leading to a steady state in bacterial cells. Defining these metabolic states, Fig. 1C shows the rate of Cu ϩ uptake into P. aeruginosa cells upon exposure to 0.5 mM CuSO 4 . The influx of metal was relatively fast, and a maximum level of intracellular Cu ϩ was reached after 10 min. Importantly, in the presence of 0.5 mM CuSO 4 , the intracellular Cu ϩ level remained constant for at least 6 h (Fig. 1D). This implies that the system stabilizes with equal Cu ϩ influx and efflux rates. Moreover, this remarkable result indicates that regulation of Cu ϩ -responsive genes occurs before the maximum Cu ϩ level is reached (i.e. within 10 min after copper exposure).

Copper homeostasis networks Copper stress triggers a metabolic shift
Toward capturing this initial response, as well as the distinct landscape under steady-state conditions, genome-wide transcriptomic analysis at 5 min (pre-steady state) and 2 h (steady state) of exposure to 0.5 mM CuSO 4 was performed using RNA-Seq. A first examination of the genome-wide expression profile clearly showed distinct responses under each condition ( Fig. 2 and supplemental Data Set S1). Cluster analysis of the differentially expressed genes (DEGs) (Ͼ2-fold change) was employed to define groups of genes with similar expression patterns. As an immediate response after 5 min of treatment (pre-steady state), a relatively small number of DEGs were induced, whereas a much larger number were repressed (Figs. 2 (A and C) and 3A (clusters 2, 4, and 7)). Examination of the functions of repressed genes points out the slowdown of cell metabolism. Genes related to fatty acid and amino acid degradation; pyruvate, nitrogen, and purine metabolism; and chemotaxis and biofilm formation pathways were all significantly down-regulated (supplemental Fig. S1) (i.e. the cell slows its metabolism, diverting resources toward the immediate response to metal stress). It was noticeable that the cluster of genes up-regulated only at 5 min included genes coding for factors associated with iron homeostasis (femR, femI, fiuR, fiuI, foxR, and foxI), the Fe 2ϩ transporter FeoB, and PA3530, a bacterioferritin-associated ferredoxin. These findings are not surprising, considering the likely disruption of Fe-S centers by the high copper levels in these cells (4). Alternatively, after 2 h of exposure to external Cu 2ϩ , many DEGs were up-regulated, including some that were repressed at 5 min (Figs. 2 (B and C) and 3A (clusters 3, 4, and 8)). The functional analysis of up-regulated DEGs at 2 h showed the activation of various metabolic pathways, predominantly genes involved in the metabolism and transport of sulfur and amino acids (supplemental Fig. S2). These data indicate that a new metabolic status is established once a steady state for Cu ϩ levels has been achieved.

Compartmentally defined protein pools maintain copper levels
A relatively small subset of genes was strongly up-regulated (Ͼ6-fold change) by the exposure to Cu 2ϩ , indicating the presence of a core adaptive regulon (Fig. 3A, cluster 1). Considering first those largely up-regulated genes (Fig. 3B), five code for membrane transporters: pcoB (PA2064), an outer membrane porin; copA1 (PA3920), the Cu ϩ -ATPase; and PA3521 (cusC), PA3522 (cusA), and PA3523 (cusB), an RND system similar to the Cu ϩ efflux Cus system (referred to as CusCAB in this study) (12,39). We have previously shown Cu ϩ efflux by P. aeruginosa CopA1 (31). Although establishing the role of the RND-type system is beyond the scope of the present report, the abundance of Met residues in CusA and its induction by Cu ϩ suggest a role

Copper homeostasis networks
in extrusion of Cu ϩ to the extracellular milieu. Genes coding for proteins probably targeted to the periplasm, PA2807 (putative Cu ϩ -binding protein with cupredoxin folding), PcoA (PA2065), and PtrA (PA2808), were also significantly up-regulated. copR (PA2809), encoding part of the periplasmic Cu ϩ two-component sensor CopR/S, belonged to this cluster of highly induced genes, whereas copS expression increased 4.8fold at 5 min and 3.7-fold at 2 h. Moreover, an array of genes coding for cytoplasmic proteins fall into this cluster (i.e. queF (PA2806), PA3520, and PA3574.1, which encode the putative cytoplasmic Cu ϩ chaperones CopZ1 and CopZ2, and PA3516, PA3517, PA3518, and PA3519, which encode adenylsuccinate lyases and hypothetical proteins, respectively). Intriguingly, nalD (PA3574), a member of the TetR family of repressors, and the transcriptional regulator brlR (PA4878) were among the highly expressed genes. Some of these largely overexpressed genes (copA1, cusCAB, copZ1, copZ2, PA3516, PA3517, PA3518, and PA3519) make up the CueR regulon (17). However, cueR (PA4778) appeared in a different cluster, with a -fold change of 2.1 at 5 min and non-induction at 2 h. pcoB, pcoA, queF, PA2807 (cupredoxin-like), ptrA, and the transcriptional regulators nalD and brlR do not respond to CueR (see below). Interestingly, except for the transcriptional regulators and QueF, these remaining proteins reside in the outer membrane or in the periplasm. Moreover, soluble proteins PcoA, PA2807 (cupredoxin-like), and PtrA are probably cuproproteins with potential redox/chaperone roles. PtrA has been proposed to play a role in Cu 2ϩ tolerance (40). A more detailed analysis of how these highly up-regulated genes might work together is presented under "Discussion."

Copper influx transporters participate in copper homeostasis
The occurrence of cell influx transporters is a prerequisite for homeostasis of a given cellular constituent. Considering that Cu ϩ efflux proteins are up-regulated under Cu ϩ stress, genes coding for Cu ϩ influx proteins should be repressed. We focused our attention on strongly down-regulated DEGs (Ͻ0.3-fold change) to screen for putative participants in Cu ϩ homeostasis (Fig. 3A, cluster 2). Three interesting proteins were identified among these: OprC (PA3790), an outer membrane porin with a proposed role in Cu ϩ influx; PA3789, an inner membrane pro-tein; and PA5030, a member of the major facilitator superfamily (MFS) of transporters with an abundance of Met and His in its primary sequence. qPCR analysis confirmed these observations (Fig. 5A). Alternatively, genes encoding previously proposed Cu ϩ importers ccoA and hmtA (30, 48) did not change their expression levels (Fig. 5).

Metallation of cuproenzymes is regulated independently of copper levels
Contrary to the highly regulated genes involved in the control of Cu ϩ levels, genes associated with cuproproteins and their metallation (senC, copA2, ccoA, ccoN, nozZ, cotA, and azu; see supplemental Table S1) were not regulated. This indicates that independent Cu ϩ distribution networks are involved in these functions. These observations were further confirmed by qPCR of selected genes (csp1, copA2, and azu) (data not shown). The induction of genes that might participate in producing Cu ϩ storage pools was also explored. No changes in the expression of PA2140 (metallothionein-like) or glutathione biosynthesis genes were observed. Recently, the high Cu ϩ -binding capability of Methylosinus trichosporium OB3b and Bacillus subtilis Csp1s has led to a suggested role for these molecules in Cu ϩ homeostasis (49). However, P. aeruginosa csp1 was not

Copper homeostasis networks
induced in any condition, suggesting perhaps an alternative role for Csp1.

Compartmental copper levels are independently controlled by the CopR/S and CueR regulons
Studies have shown the role of the cytoplasmic regulator CueR and the two-component system CusR/S in sensing periplasmic copper and regulating the Cus RND-type transport system (9,12,17). The RNA-Seq analysis showed that a number of P. aeruginosa genes strongly up-regulated during Cu 2ϩ exposure were not part of the described CueR regulon (17). RNA-Seq analysis of ⌬cueR and ⌬copR mutant strains after 5-min and 2-h exposure to 0.5 mM CuSO 4 was performed to

Copper homeostasis networks
seek an integrated view of the different roles of each transcriptional regulator (supplemental Data Set S1). Both strains were fully viable in medium supplemented with this level of Cu 2ϩ (Fig. 6A). As expected, cluster analysis of the DEGs from wild type, ⌬copR, and ⌬cueR revealed two distinct subsets of genes whose Cu ϩ -dependent induction was abolished in each one of the mutant strains (Fig. 6B, clusters 2 and 6). Importantly, expression of copR in the ⌬cueR mutant background and expression of cueR in the ⌬copR mutant background were not affected (Fig. 7A). Mutation of cueR abolished expression of copA1, cusCAB, copZ1, copZ2, PA3516, PA3517, PA3518, and PA3519, confirming the described CueR regulon (17). This observation was further verified by qPCR analysis of selected genes (Fig. 7B). Alternatively, expression of pcoB, pcoA, queF, PA2807 (cupredoxin-like), and ptrA was suppressed in the ⌬copR mutant strain (Fig. 7C). These genes are organized in three operons (pcoAB, PA2808, and PA2807/06) and appear to constitute the full CopR/S regulon. Analysis of their 5Ј-untranslated regions revealed that the CopR putative DNA-bind-ing region sequence is homologous to the reported Escherichia coli CusR-binding site (32) (Fig. 7D). A BLAST search of the P. aeruginosa genome did not show the additional presence of the CopR DNA-binding motif (TGACXXXTGTAAT). These results are consistent with the idea that CueR and CopR exert different regulatory roles (i.e. CueR regulates the expression of genes implicated in cytoplasmic Cu ϩ homeostasis, whereas control of the periplasmic metal level relies on CopR).

Discussion
The goal of this work was to obtain a global description of the bacterial mechanisms that control cellular Cu ϩ levels. Transcriptomic analysis, under carefully selected stress conditions minimizing pleotropic responses to Cu ϩ toxic effects, revealed key components of Cu ϩ homeostasis. The data unveil the overall cellular physiology under pre-steady-state and steady-state conditions, discriminate between elements involved in controlling compartmental Cu ϩ levels from others related to cuproprotein metallation, and uncover novel putative periplasmic

Copper homeostasis networks
chaperones, metal influx transporters, and the presence of two cytoplasmic chaperones. In addition, the identified regulons of both periplasmic and cytoplasmic Cu ϩ sensors demonstrate that Cu ϩ levels are controlled separately according to subcellular compartments. These findings led us to propose a testable model for Cu ϩ compartmental distribution in the whole organism (Fig. 8).

Copper levels reach steady state in P. aeruginosa
Limited information is available on the kinetics of Cu ϩ distribution among bacterial cellular compartments and the sizes of the various metal pools. This is relevant in transcriptomic response studies to identify Cu ϩ distribution proteins while neglecting those involved in the general cellular response to a homeostatic imbalance. Nevertheless, substantial progress can be made by analyzing the rate of cellular Cu ϩ change upon homeostasis disruption and how fast an eventual steady state is reached. In bacterial systems, steady state implies not only equal influx/efflux rates but, more importantly, cell integrity (i.e. the absence of toxic effects leading to cell death). We observed that in the presence of sublethal external Cu 2ϩ levels (Ͻ0.5 mM Cu 2ϩ ), P. aeruginosa was in a steady state, because intracellular Cu ϩ level and cell viability remained constant for as long as the experiments were extended. In addition, our determinations showed that under aerobiosis, P. aeruginosa reaches steady state between 5 and 10 min following the change in external Cu 2ϩ concentration. Notably, we did not observe an excess Cu ϩ load that was later decreased to basal or lower steady-state levels upon expression of homeostatic proteins. This implies that transcriptional regulation and expression/repression of relevant proteins is achieved before Cu ϩ equilibria. This regulation is continued, albeit slightly attenuated, maintaining a steady-state Cu ϩ level higher than the basal (in the absence of external Cu ϩ ). Importantly, the normal growth with high intracellular Cu ϩ levels implies that an unidentified intracellular Cu ϩ -binding molecular pool should be present.

Metabolic changes occur in response to copper stress
Our approach led us to observe that the most strongly upregulated or repressed (6 -70-fold) genes were those responsible for Cu ϩ distribution, whereas differential expression of all others genes was significantly lower (2-3-fold). Among the latter, the pre-steady-state repression of genes participating in general metabolism was observed, suggesting the need for energy economy or reassignment to essential tasks specific to the Cu ϩ stress response. As cells entered steady state, there was an increase in the number of genes with a low degree of up-regulation, with the consequent restoration of primary metabolic functions.
Subsets of genes involved in Cu ϩ metallation of cytochrome c oxidases and denitrifying enzymes, NosZ and NirS, as well as genes coding for possible Cu ϩ storage pools were also distinctly unresponsive to Cu ϩ stress. It has been proposed that CcoA, SenC, and CopA2 are involved in metallation of cytochrome c oxidases (30,31,50). No induction of their coding genes was observed, suggesting separate regulatory networks for metallation of cytochrome c oxidases. However, this might be quite complex. Consider, for instance, that there are three cytochrome c oxidase systems in P. aeruginosa, the cbb3-1 oxidase (Cbb3-1), the cbb3-2 oxidase (Cbb3-2), and the aa3 oxidase (Aa3) (51), but only one set of CcoA/SenC/CopA2 proteins. Is this protein set participating in the metallation of all cytochrome c oxidases? Or are alternative unidentified transporters part of this network? Among missing elements, a cytoplasmic molecule carrying Cu ϩ from CcoA to CopA2 should be taken into account, because both chaperone CopZs are up-regulated by Cu ϩ independently of CopA2. Assembly and copper metallation of the periplasmic NosZ and NirS oxidoreductases are poorly understood, preventing further consideration. A single report indicated that the NosF and NosY, subunits of an ABC ATPase, and the periplasmic NosD, but not OprC, are required for function (52).
Among possible mechanisms contributing to Cu ϩ homeostasis, the presence of cytoplasmic molecules buffering a Cu ϩ pool has been proposed (53,54). We observed that there is a large increase in the intracellular Cu ϩ under stress conditions. Early studies showed that glutathione influences the metal tolerance of E. coli cells lacking metal efflux systems; however, wild-type cells do not use glutathione to cope with metal stress (53). In agreement with these reports, we did not observe changes in the expression of enzymes involved in glutathione

Copper homeostasis networks
metabolism. Recently, a small tetrameric protein, Csp1, with high Cu ϩ -binding capability was described in M. trichosporium OB3b and B. subtilis (49). However, no changes in expression levels in the csp1 gene were observed. Thus, whereas a Cu ϩbuffering cellular function appears necessary, no evidence of likely responsible molecules emerged from our studies.

Defined transport mechanisms control compartmental copper levels
P. aeruginosa-specific response to Cu ϩ stress could be described in the form of three components: genes up-regulated by CueR that handle cytoplasmic Cu ϩ levels; genes under control of the two-component CopR/S system affecting Cu ϩ periplasmic distribution; and genes largely down-regulated, probably involved in copper influx (Fig. 8). Previous studies suggested that P. aeruginosa OprC functions as an outer membrane porin involved in copper uptake into the periplasm (55). Our transcriptomic data support this idea as oprC is strongly downregulated in the presence of external Cu 2ϩ . oprC is part of an operon with PA3789, coding for a protein probably located in the plasma membrane. However, the structure of PA3789 does not suggest a transport function but rather a peptidase with a PepSy domain (56). Alternatively, the gene PA5030, coding for an MFS transporter localized in the plasma membrane, was also strongly down-regulated. The abundance of Met and His in its sequence suggests metal-binding/transport capabilities. More importantly, deletion of either oprC or PA5030 conferred significant tolerance to external Cu 2ϩ . It has been suggested that CcoA, also a MFS transporter, and the P 1Btype ATPase HtmA might function as plasma membrane cop-per uptake systems (30,48). Rhodobacter capsulatus CcoA is required for metallation of cytochrome c oxidase, and deletion of ccoA does not confer Cu ϩ tolerance (30). In agreement with a role in protein metallation, external Cu 2ϩ did not affect the expression of P. aeruginosa ccoA. Similarly, P. aeruginosa htmA does not appear to respond to Cu ϩ stress. HmtA increases Cu 2ϩ and Zn 2ϩ levels but does not alter Ag ϩ and Cd 2ϩ when heterologously expressed in E. coli (48). This surprising result appears contrary to the transport mechanism of P-type ATPases, where the transported ion is required in the phosphorylation step, when metal sites face the cell cytoplasm (57). Moreover, the putative metal-coordinating residues present in HtmA are distinct from Cu ϩ/2ϩ -or Zn 2ϩ -ATPases (58), and similar selectivity of biomolecules for elements in the same periodic group (i.e. electronic structure (group 10 Cu ϩ /Ag ϩ /Au ϩ and group 11 Zn 2ϩ /Cd 2ϩ )) has been described (59). Thus, considering the data shown in this report, OprC and PA5030 are likely to work in tandem providing Cu ϩ to the cytoplasm. In this direction, we believe that an unidentified periplasmic chaperone is likely to mediate the metal movement between both transporters.
CueR-controlled genes are primarily involved in maintaining lower cytoplasmic Cu ϩ levels. The key players of this system are the well-characterized plasma membrane CopA1 (31) and a CusCAB transport system spanning the outer and inner membranes (39). Previous studies indicated that several P. aeruginosa RND systems might participate in the response to Cu ϩ stress (13). However, we observed that P. aeruginosa CusCAB is the sole RND system probably exerting this function. This, Figure 8. Copper homeostasis networks in P. aeruginosa. Model depicting key components of P. aeruginosa Cu ϩ distribution mechanisms. The candidates for Cu ϩ influx transporters (green, down-regulated genes), the CopR and CueR regulon (red, up-regulated genes), and the proteins responsible for metallation of cuproproteins (non-induced) are shown. Blue, transmembrane transporters; orange, transcriptional regulators; green, cuproproteins; yellow, chaperones; gray, proteins of unknown role. Hexagonal shapes highlight redox activity.

Copper homeostasis networks
however, has a distinctive characteristic, the regulation by CueR the cytoplasmic sensor, rather than a periplasmic sensing two-component system as occurs in E. coli (17). It is interesting that the periplasmic chaperone CusF, observed in E. coli for instance, is not present in P. aeruginosa. These two observations might suggest that CusCAB acquires Cu ϩ from the cytoplasm. Consider that alternative transport mechanisms have been proposed for this system, acquiring Cu ϩ either from the cytoplasm (60) or the periplasm (61). Moreover, considering the proximity of the copZ1 (PA3520) in the genome and the presence of an extra chaperone, CopZ2, it is tempting to speculate that CopZ1 delivers Cu ϩ not to an ATPase but to CusCAB instead. The data suggest that both CopZ1 and CopZ2, albeit with higher affinity, are cytoplasmic Cu ϩ chaperones. The presence of two Cu ϩ -ATPases in P. aeruginosa and the demonstrated role of chaperones delivering Cu ϩ to these transporters might suggest that each chaperone delivers metal to a distinct ATPase. However, the independent regulation and participation of CopA2 in the metallation of cytochrome c oxidase make this possibility less appealing.
We have shown that CopA delivers Cu ϩ to the periplasmic chaperone CusF in E. coli (21). Because CusF is not present in P. aeruginosa, it appears unlikely that CopA1 works in tandem with CusCAB. Alternatively, CopA1 might operate in conjunction with PcoB, a porin in the outer membrane under the control of CopR. However, the need for a periplasmic chaperone is still pertinent (see below). In any case, the possibility of cytoplasmic Cu ϩ extrusion via CopA1/PcoB and CusCAB presents the question of when the cell uses one or the other system.
The relevance of periplasmic Cu ϩ -sensing two-component systems has been shown previously (62)(63)(64). However, a complete description of a CopR regulon was lacking. The comparative analysis of the wild-type, ⌬cueR, and ⌬copR strain transcriptomes in response to external Cu 2ϩ uncovered a small set of genes that were exclusively induced in wild-type and ⌬cueR and thus likely to constitute the CopR regulon. All members of the regulon shared a conserved operator sequence that matches the E. coli CusR DNA-binding site (CusR box). Curiously, among the genes controlled, only ptrA appears to affect copper tolerance, as evaluated by bacterial survival in high metal levels (40). PcoA and PcoB were initially described as part of a copper resistance Cop operon in E. coli and Pseudomonas syringae (7,65). As mentioned, pcoB codes for a copper-binding outer membrane porin (65). E. coli PcoA has multicopper oxidase activity and requires interaction with the periplasmic chaperone PcoC for full activity (66,67). In P. aeruginosa, this role might be exerted by the cupredoxin-like protein (PA2807), also part of the CopR regulon. Alternatively, we hypothesize that PA2807 might participate in the copper trafficking from CopA1 to PcoB as well as in the delivery of copper to periplasmic cuproproteins. PA2807 along with ptrA and queF form the gene cluster next to the copRS operon. P. aeruginosa PtrA, a periplasmic protein of unknown function, has been associated with Cu ϩ tolerance (40). PA2807 and QueF are induced by Cu ϩ , but deletion of these genes in P. putida does not lead to an increase in Cu 2ϩ sensitivity (40). QueF is a nitrile reductase that participates in the synthesis of the tRNA-modified nucleoside queosine, which in turn improves translational accuracy (68). Con-sequently, QueF might be necessary to fine-tune the activation of the Cu ϩ homeostatic machinery.
In summary, selective copper stress conditions revealed specific homeostasis mechanisms enabling compartmental metal sensing and distribution coupled to transmembrane influx and efflux (Fig. 8). This model, although it does not include components required for cuproenzyme metallation, provides a testable system to model copper fluxes and pools under various environmental conditions, such as those encountered during bacterial virulence in eukaryotic hosts.

Copper uptake assays
Copper uptake was measured using cells in mid-log phase in antibiotic-free LB medium. Upon the addition of 0.5 mM CuSO 4 , aliquots were taken at the indicated times, treated with 1 mM DTT and 1 mM bathocuproinedisulfonic acid (BCS), and harvested by centrifugation. Pellets were washed twice with 150 mM NaCl. Cells were mineralized with 15.6 M HNO 3 (trace metal grade) for 1 h at 80°C and overnight at 20°C. Digestions were completed with 2 M H 2 O 2 . Copper levels were measured by atomic absorption spectroscopy (AAS). When required, samples were diluted with 300 mM HNO 3 before analysis by AAS.

RNA isolation and sequencing
P. aeruginosa PAO1, ⌬copR, and ⌬cueR mutant strains were grown to mid-log phase in antibiotic-free LB medium and treated with 0.5 mM CuSO 4 for 0 min, 5 min, and 2 h. At each point, 7.5 ϫ 10 8 cells were collected for RNA extraction, and 2ϫ RNAprotect Bacteria Reagent (Qiagen) was added to stabilize RNA. Total RNA extraction was performed using an RNeasy Plus kit (Qiagen), following the manufacturer's instructions. To minimize biological variation within samples, RNA from three biological replicates was pooled to constitute single RNA sequencing samples. Furthermore, two independent RNA samples were sequenced for each strain and condition (71).
Library construction and sequencing were performed at the Biopolymers Facility, Harvard Medical School (Boston, MA). Briefly, mRNA was enriched using the Ribo-Zero rRNA removal kit (Illumina). Strand-specific cDNA libraries were synthesized with the Apollo324 TM system (WaferGen). Transcriptome sequencing was carried out to generate 150-bp pair end reads on a Next Seq 500 Mid Output platform. The number of raw reads ranged 5-10 million/sample (average ϭ 8,429,273), ensuring optimal sequencing depth (72).

Gene expression analysis and functional categorization
Workflows were built with the Galaxy platform (73). Clean reads (average quality Ն25) were mapped against the P. aeruginosa PAO1 genome using TopHat (mean inner distance between mate pairs ϭ 50; S.D. for distance between mate pairs ϭ 20) (74). Count files were generated from aligned reads with HTSeq (mode to handle reads that overlap more than one feature ϭ intersection-strict; minimum alignment quality ϭ 20) (75). Screening of DEGs in pairwise comparisons (treated versus untreated) was performed using DEseq2 (76). We defined differential expression as the ratio between expression values Ն Ϯ2 (p value Յ 0.01 and corrected p value Յ 0.05) at 5 min and 2 h versus time 0. -Fold change values and statistical significance for each gene are provided in supplemental Data Set S1. Putative function of DEGs was manually assigned based on PseudoCap annotation (77) and literature search. Pathway annotation of genes down-regulated at 5 min and up-regulated at 2 h was performed and visualized using ClueGO App within Cytoscape (78). DEGs were clustered using Cluster 3.0, and the corresponding heat maps were generated with Java Tree View (79).
Quantitative PCR analysis was performed on samples obtained under the same conditions as those used for RNA-Seq. Total RNA was isolated as described above. Following DNase I treatment (New England Biolabs), RNA was purified with phenol/chloroform extraction and ethanol precipitation. cDNA was synthesized from 1 g of total RNA using the ProtoScript II first strand DNA synthesis kit (New England Biolabs) in accordance with the manufacturer's protocol. All reactions were prepared in a 20-l final volume with 10 l of 2ϫ Fast Essential DNA Green Master (Roche Applied Science), 0.4 M each primer, and 25 ng of cDNA. Amplification, after a 10-min preincubation at 95°C, consisted of 40 cycles of 95°C for 10 s and 60°C for 30 s, followed by a melting stage. To compare the transcript levels between different samples, we used the comparative Ct (2 Ϫ⌬⌬Ct ) method (80).

Cu ؉ loading to chaperone proteins
Prior to Cu ϩ -binding studies, apo-PA3520 (CopZ1) and apo-PA3574.1 (CopZ2) were fully reduced by incubation overnight with 5 mM TCEP. Reducing agent was removed with 3-kDa Centricon filters immediately before the assay. Cu ϩ loading to apo-PA3520 (CopZ1) and apo-PA3574.1 (CopZ2) was performed by incubating each protein with a 2-fold molar excess of CuSO 4 in 25 mM Hepes, pH 8, 150 mM NaCl, 5 mM DTT for 5 min at room temperature with gentle agitation. DTT and unbound Cu ϩ were removed by passage through a Sephadex G-10 column (GE Healthcare). The amount of Cu ϩ bound to chaperones was determined by the bicinchoninic acid method (83), using BCS as color reagent. CuSO 4 was used as a standard.

Copper-binding affinity
Protein-Cu ϩ dissociation constants (K D ) were determined by using a competition assay with BCS following the formation of the CuBCS 2 complex at 483 nm (␤ 2 formation constant for CuBCS 2 ϭ 10 19.8 M Ϫ2 , ⑀ ϭ 13,000 M Ϫ1 cm Ϫ1 (45)). Briefly, 10 M Cu ϩ , 25 M BCS in 25 mM Hepes, pH 8, 150 mM NaCl, 10 mM ascorbic acid was titrated with 1-20 M protein and incubated for 5 min at room temperature, and the 300 -800 nm absorption spectra were recorded. K D values were calculated by curve fitting of the experimental data to the equilibrium equation (45).
Reported errors are asymptotic S.E. values provided by the fitting software (KaleidaGraph, Synergy).

Molecular modeling
Homology modeling of P. aeruginosa PA3520 (CopZ1) and PA3574.1 (CopZ2) was performed with the Phyre2 server (87), using Salmonella enterica GolT (Protein Data Bank entry 4Y2I) as a template. Quality of the models was assessed with the ProQ2 algorithm (88). The electrostatic surface potentials were calculated using the Adaptive Poisson-Boltzmann Solver (APBS) with the PyMOL APBS tool (89).