Widespread Distribution and Functional Specificity of the Copper Importer CcoA: Distinct Cu Uptake Routes for Bacterial Cytochrome c Oxidases

ABSTRACT Cytochrome c oxidases are members of the heme-copper oxidase superfamily. These enzymes have different subunits, cofactors, and primary electron acceptors, yet they all contain identical heme-copper (CuB) binuclear centers within their catalytic subunits. The uptake and delivery pathways of the CuB atom incorporated into this active site, where oxygen is reduced to water, are not well understood. Our previous work with the facultative phototrophic bacterium Rhodobacter capsulatus indicated that the copper atom needed for the CuB site of cbb3-type cytochrome c oxidase (cbb3-Cox) is imported to the cytoplasm by a major facilitator superfamily-type transporter, CcoA. In this study, a comparative genomic analysis of CcoA orthologs in alphaproteobacterial genomes showed that CcoA is widespread among organisms and frequently co-occurs with cytochrome c oxidases. To define the specificity of CcoA activity, we investigated its function in Rhodobacter sphaeroides, a close relative of R. capsulatus that contains both cbb3- and aa3-Cox. Phenotypic, genetic, and biochemical characterization of mutants lacking CcoA showed that in its absence, or even in the presence of its bypass suppressors, only the production of cbb3-Cox and not that of aa3-Cox was affected. We therefore concluded that CcoA is dedicated solely to cbb3-Cox biogenesis, establishing that distinct copper uptake systems provide the CuB atoms to the catalytic sites of these two similar cytochrome c oxidases. These findings illustrate the large variety of strategies that organisms employ to ensure homeostasis and fine control of copper trafficking and delivery to the target cuproproteins under different physiological conditions.

MFS-type transporters (http://www.tcdb.org/) (36). Recent studies showed that conserved Met and His residues of CcoA are important for its function, possibly acting as metal ligands (37). It is noteworthy that both the Cu importer CcoA and the Cu exporter CcoI are required to incorporate the Cu B center into CcoN, implying trafficking of Cu across the cytoplasmic membrane during cbb 3 -Cox biogenesis (32).
In this study, we conducted comparative genomic analyses of CcoA orthologs in alphaproteobacterial genomes. This search revealed a higher degree of co-occurrence of CcoA with cbb 3 -Cox than with aa 3 -Cox, suggesting that CcoA activity is specific to class C HCOs. To test this hypothesis, we investigated the function of CcoA in R. sphaeroides, a close relative of R. capsulatus, which contains both functional cbb 3 -and aa 3 -Cox with identical heme-Cu B binuclear centers, belonging to different HCO families. Upon the identification of R. sphaeroides ccoA (RSP_2726), appropriate mutants were constructed and their physiological and biochemical properties were characterized. We also identified bypass suppressors of ΔccoA mutants in R. sphaeroides copA (RSP_2890) that restored cbb 3 -Cox activity at the expense of increased Cu 2ϩ sensitivity. This study showed that CcoA is specific to cbb 3 -Cox and is not involved in the biogenesis of aa 3 -Cox. Therefore, we concluded that the Cu atoms needed for the formation of the heme-Cu B binuclear centers in these two similar enzymes must be provided by distinct Cu uptake pathways.

RESULTS
Distribution of CcoA homologues in alphaproteobacteria. Using CcoA from R. capsulatus as a query, we identified 144 CcoA-like MFS proteins in 125 of the 327 alphaproteobacterial species interrogated, with several genomes containing up to three distinct copies of ccoA. We also compiled a phylogenetic profile of ccoA together with the presence of the cbb 3 -Cox and aa 3 -Cox structural genes (Fig. 1A). Most of the CcoA homologues were predicted to contain 12 transmembrane helices with conservation of the motifs MXXXM in helix 7 and HXXXM in helix 8, both of which are required for Cu uptake and for cbb 3 -Cox activity in R. capsulatus (37), with the exception of a group of CcoA-like proteins in Rhizobiales (Fig. 1B). These putative transporters also contained the two conserved motifs in helices 7 and 8 but were truncated at the C terminus, lacking predicted helix 12. Intriguingly, the genes encoding these truncated CcoA homologues were found right downstream of the ccoNOQP and ccoGHIS clusters, encoding the structural and assembly genes of cbb 3 -Cox, respectively (Fig. 1B), suggesting that the Rhizobiales CcoA-like proteins might play a role similar to that of R. capsulatus CcoA. For the complete set of data pertinent to Fig. 1, see Table S1 in the supplemental material.
We found that of the 327 alphaproteobacterial genomes analyzed, only 44 had no CcoA or Cox enzyme ( Fig. 2; Table S2). Among the remaining genomes, 118 (62%) of 192 coding for cbb 3 -Cox (CcoNOQP) and 122 (46%) of 274 coding for aa 3 -Cox (CoxI, CoxII, and CoxIII) also contained CcoA. In contrast, 74 (23%) of these genomes had cbb 3 -Cox but not CcoA (mainly from Caulobacterales, Brucella, Rhizobiaceae, Hyphomonadaceae, Rhodospirillaceae, and Sphingomonadales), while 152 (46%) had aa 3 -Cox without CcoA ( Fig. 2; Table S2). Thus, the data suggested a higher degree of co-occurrence of ccoA and cbb 3 -Cox than of ccoA and aa 3 -Cox. This co-occurrence was particularly evident in Methylocystaceae and Methylobacteriaceae, where species of the same genus would have both CcoA and cbb 3 -Cox or neither (Fig. 1C). In addition, we also observed strain differences; e.g., both Paracoccus denitrificans strains SD1 and PD1222 had cbb 3 -Cox and aa 3 -Cox but only strain PD1222 contained CcoA. Similarly, all of the Rhizobium leguminosarum strains analyzed had cbb 3 -Cox and aa 3 -Cox but individual biovars differed in the presence of CcoA. Finally, we found six species containing cbb 3 -Cox without CcoA and seven species containing CcoA but not cbb 3 -Cox, suggesting that CcoA-independent provision of Cu to cbb 3 -Cox and an additional unknown function(s) of CcoA, that is unrelated to Cu provision to the Cu B center of this enzyme might exist in some species (Table S2).
Phenotypes of ccoA mutants of R. sphaeroides. To understand the role of CcoA in the biogenesis of the Cu B site of HCO and to test whether CcoA is involved in the provision of Cu to aa 3 -Cox, as it is in the provision of Cu to cbb 3 -Cox, we investigated the function of a CcoA ortholog in an organism that contains multiple HCOs. R. sphaeroides has a CcoA homologue (RSP_2726, previously annotated as a multidrug/ metabolite efflux pump) containing the conserved Met motifs in transmembrane helices 7 and 8 (Fig. 3A). Unlike R. capsulatus, which is rare among the alphaproteobacterial species in having only one HCO (cbb 3 -Cox), R. sphaeroides also contains the canonical type A aa 3 -Cox. To assess the effect of lacking CcoA on both cbb 3 -and aa 3 -Cox activities, a ccoA deletion allele was introduced into appropriate R. sphaeroides strains. The wild-type (Ga) strain and the Δaa 3 (JS100 [38]) and Δcbb 3 (MT001 [39,40]) mutant strains yielded the ΔccoA (HW3) single mutant and the ΔccoA Δaa 3 (HW2) and ΔccoA Δcbb 3 (HW4) double mutants, respectively (Table S3). The Δcbb 3 Δaa 3 double mutant (ME127 [40]), lacking both cbb 3 -and aa 3 -Cox activities, served as a negative control (Table S3). The Cox activities of these strains were visualized qualitatively by ␣-naphthol and N=N=-dimethyl-p-phenylenediamine (NADI; blue) staining of colonies grown aerobically on enriched medium (Fig. 3B). The wild-type and Δaa 3 mutant strains were NADI ϩ (i.e., stained dark blue in seconds), whereas the Δcbb 3 mutant was  Table S1. (B) Schematic representation of the cbb 3 -Cox structural (ccoNOQP) and assembly (ccoGHIS) gene clusters together with the ccoA homologue in the Rhizobiales genomes indicated. (C) Co-occurrence plot with circles indicating the presence or absence of ccoA (red or white, respectively), the cbb 3 -Cox structural and assembly gene clusters (black or white, respectively), and the aa 3 -Cox-related genes (gray or white, respectively). Not all species are shown because of space limitations, but for a complete profile and a summary, see Tables S1 and S2, respectively. NADI slow (i.e., stained blue in a few minutes), indicating that cbb 3 -Cox provides most of the Cox activity under these growth conditions. The Δaa 3 Δcbb 3 double mutant was NADI Ϫ (i.e., no blue staining after 15 min), consistent with the absence of both Cox enzymes (39). Both the ΔccoA single mutant and the ΔccoA Δcbb 3 double mutant had a NADI slow phenotype, similar to that observed when only aa 3 -Cox activity (Δcbb 3 ) was present (Fig. 3B). Importantly, the double mutant lacking both CcoA and aa 3 -Cox (ΔccoA Δaa 3 ) but containing the intact structural genes of cbb 3 -Cox was NADI Ϫ like the double mutant (Δcbb 3 Δaa 3 ) lacking both Cox activities (Fig. 3B). Upon complementation with a plasmid carrying a wild-type allele of R. sphaeroides ccoA, both the single (ΔccoA) and double (ΔccoA Δaa 3 ) mutants lacking CcoA became NADI ϩ (Fig. 3B). Thus, the data indicated that in R. sphaeroides, the absence of ccoA affected cbb 3 -Cox, but not aa 3 -Cox, activity.
The NADIphenotypes of R. sphaeroides ΔccoA mutants were restored upon the addition of 5 M Cu 2ϩ to the growth medium, similar to that seen in the R. capsulatus ΔccoA mutant (34). In contrast, the ΔccoA Δcbb 3 double mutant, which has only a functional aa 3 -Cox, remained NADI slow upon Cu 2ϩ supplementation, suggesting that Cu 2ϩ addition had no effect on aa 3 -Cox activity (data not shown).
Absence of CcoA affects heme and subunit compositions of cbb 3 -Cox but not aa 3 -Cox. To assess how the absence of CcoA affects R. sphaeroides HCO biogenesis, the c-, b-, and a-type heme contents of membrane fractions derived from appropriate mutants were analyzed by using optical difference (dithionite-reduced minus ferricyanide-oxidized) spectra. In membranes of a wild-type R. sphaeroides strain, prominent peaks around 605, 560, and 551 nm, corresponding to the a-, b-, and c-type hemes, respectively, were readily detectable (39) (Fig. 3C). As expected, a significant decrease in the 605-nm peak and in the 560-and 551-nm peaks was observed in the Δaa 3 and Δcbb 3 mutants, respectively (41). Note that in R. sphaeroides membranes, only aa 3 -Cox has a-type heme but other proteins besides cbb 3 -Cox contain b-and c-type hemes (e.g., cyt bc 1 ) under the growth conditions tested. Accordingly, in the double mutant (Δcbb 3 Δaa 3 ) lacking both Cox enzymes, all three peaks decreased substantially compared with the wild-type strain (Fig. 3C), as reported earlier (39). Remarkably, in the ΔccoA single mutant only the content of b-and c-type hemes decreased, as in the mutant lacking only cbb 3 -Cox (Δcbb 3 ) or the ΔccoA Δcbb 3 double mutant. Moreover, in the ΔccoA Δaa 3 double mutant, all three peaks, corresponding to the a-, b-, and c-type hemes, decreased drastically, similar to what was seen in the double mutant (Δcbb 3 Δaa 3 ) (Fig. 3C). In summary, the data showed that in the absence of CcoA, the content of b-and c-type hemes in the membrane fraction (corresponding partly to cbb 3 -Cox) decreased significantly, whereas the a-type heme content (corresponding to aa 3 -Cox) remained unchanged, consistent with CcoA being involved in cbb 3 -Cox, but not aa 3 -Cox, production. We emphasize that these data are merely semiquantitative be- cause of the presence of other b-and c-type cyts (in addition to cbb 3 -Cox) whose content may vary in the presence or absence of different HCOs.
Next, the steady-state amounts of cbb 3 -Cox subunits present in membranes from appropriate mutants were examined by SDS-PAGE and 3,3=,5,5=-tetramethylbenzidine (TMBZ) staining, which specifically reveals membrane-bound c-type cyts (42). In wildtype R. sphaeroides membranes, four distinct c-type cyts, including the CcoO (cyt c o ) and CcoP (cyt c p ) subunits of cbb 3 -Cox, can be detected (Fig. 3D, top). As expected, in Heme Staining were grown aerobically at 35°C on LB medium, and the presence of Cox activity was visualized by NADI staining (see Materials and Methods). Colonies that contain wild-type levels of Cox activity turn dark blue within a few seconds (NADI ϩ ), while those that have low or no Cox activity show lighter blue (NADI slow ) or no blue staining (NADI Ϫ ) upon longer exposure, respectively. Note that the ΔccoA Δaa 3 mutant is NADIlike the Δaa 3 Δcbb 3 mutant, unless it is complemented with a plasmid carrying a wild-type allele of ccoA (ΔccoA Δaa 3 CcoA ϩ ). (C) Absorption difference spectra of membrane fractions of R. sphaeroides mutants recorded between 500 and 625 nm by using oxidized membrane preparations as the baseline and reducing the sample with an excess of sodium dithionite. The intensity of the peaks centered at 551, 560, and 605 nm indicates the contents of c-, b-and a-type hemes, respectively. (D) Steady-state levels of structural subunits of cbb 3 -and aa 3  The presence of the Cox1 subunit of R. sphaeroides aa 3 -Cox was identified with P. denitrificans Cox1 polyclonal antibodies that cross-react with it. The white lines seen on the blot next to some lanes are scanning artifacts and do not reflect spliced gels. (E) cyt c activity of membrane fractions of R. sphaeroides ΔccoA mutants. Total Cox (cbb 3 -Cox plus aa 3 -Cox) activities were determined using membrane preparations of various R. sphaeroides strains by monitoring the rate of oxidation of reduced horse heart cyt c. R. sphaeroides wild-type strain Ga exhibited an activity of~1.33 mol of cyt c oxidized/min/mg of total membrane proteins, which was referred to as 100%. Three independent assays were carried out for each strain. The ΔccoA Δaa 3 mutant has no activity, like the Δaa 3 Δcbb 3 mutant that lacks both Cox enzymes. the absence of aa 3 -Cox, the c-type cyt profile remained unchanged, but in mutants lacking cbb 3 -Cox (Δcbb 3 and Δcbb 3 Δaa 3 ), cyt c o and cyt c p were not present, leaving only the cyt c 1 subunit of cyt bc 1 and the membrane-anchored electron carrier cyt c y . Remarkably, in strains lacking CcoA, such as the ΔccoA and ΔccoA Δaa 3 mutants, the amounts of cyt c o and cyt c p decreased at different levels, even though these strains contained an intact copy of the cbb 3 -Cox structural genes. These data, together with the spectral data showing that the amount of b-type heme, and hence that of CcoN, also decreased, indicated that production of cbb 3 -Cox was defective in the absence of CcoA. Finally, the presence of the Cox1 subunit of aa 3 -Cox was monitored by using polyclonal antibodies raised against Cox1 of Paraccocus denitrificans aa 3 -Cox (43) (Fig. 3D, bottom). As expected, Cox1 was absent from mutants lacking aa 3 -Cox, like the Δaa 3 , Δcbb 3 Δaa 3 , and ΔccoA Δaa 3 mutant strains. However, it was readily detected in strains lacking CcoA (ΔccoA mutant), cbb 3 -Cox (Δcbb 3 mutant), or both proteins (ΔccoA Δcbb 3 mutant) at levels comparable to those of the wild type, in agreement with the Cu-containing Cox1 subunit of aa 3 -Cox being unaffected by the absence of CcoA in R. sphaeroides.
Cox activities of mutants lacking CcoA. The total cyt c oxidation activity (accounting for both aa 3 -Cox and cbb 3 -Cox activities) present in membranes of different R. sphaeroides strains was measured by using reduced horse heart cyt c. R. sphaeroides wild-type strain Ga exhibited an activity level of 1.33 mol of cyt c oxidized/min/mg of total membrane proteins (referred to as 100%) (Fig. 3E). Addition of 200 M KCN, a specific inhibitor of the HCO catalytic binuclear center, abolished this activity almost completely (96% inhibition). The mutants lacking aa 3 -Cox (Δaa 3 mutant) and cbb 3 -Cox (Δcbb 3 mutant) showed Cox activities corresponding to 73 and 62% of the wild-type level, respectively, whereas the Δcbb 3 Δaa 3 double mutant had no activity. A strain lacking only CcoA (ΔccoA mutant) or both CcoA and cbb 3 -Cox (ΔccoA Δcbb 3 mutant) showed similar amounts of Cox activity, 59 and 60% of the wild-type level, respectively. In contrast, a strain lacking both CcoA and aa 3 -Cox (ΔccoA Δaa 3 mutant), although it contained intact cbb 3 -Cox structural genes, had no Cox activity, similar to a Δcbb 3 Δaa 3 double mutant (Fig. 3E). Therefore, the absence of CcoA affected only cbb 3 -Cox, and not aa 3 -Cox, in R. sphaeroides.
Suppressors of ⌬ccoA restore cbb 3 -Cox activity at the expense of Cu 2؉ hypersensitivity. During the phenotypic characterization of ΔccoA mutants, we observed that the NADIdouble mutant lacking both CcoA and aa 3 -Cox (ΔccoA Δaa 3 mutant) readily yielded wild-type-like NADI ϩ revertants (Fig. 4A). Similar revertants had previously been obtained with R. capsulatus ΔccoA mutants, and their characterization showed that these suppressor mutations restored cbb 3 -Cox deficiency and conferred Cu 2ϩ sensitivity (32). Using whole-genome sequencing, we determined that these mutations were single base-pair indels in a rare stretch of 10 conserved cytosine base pairs located in copA, which encoded the P 1B -type ATP-dependent Cu exporter (CopA) (32). These indels caused translational frameshifts that inactivated copA and increased cellular Cu content and Cu 2ϩ sensitivity (32).
Intrigued by the occurrence of similar revertants of R. sphaeroides, we retained four independent NADI ϩ derivatives (HW2R 1 to HW2R 4 ) of the ΔccoA Δaa 3 double mutant (Table S3) and tested their Cu 2ϩ tolerance in enriched medium. Indeed, they were hypersensitive to Cu 2ϩ (above~200 M) compared with their wild-type and ΔccoA Δaa 3 mutant parents (tolerant to~1 mM) (Fig. 4B). Thus, similar to R. capsulatus, these R. sphaeroides revertants regained cbb 3 -Cox activity at the expense of becoming hypersensitive to Cu 2ϩ . DNA sequencing of the genomic copies of R. sphaeroides copA (RSP_2890) (44) from these revertants showed that they all contained two base-pair (CG) deletions in copA (Fig. 4C). Remarkably, these deletions were located in a region of copA containing five consecutive CG repeats, presumably causing translational frameshifts that inactivated copA and increased the Cu 2ϩ sensitivity of cells. The data indicated that in R. sphaeroides, as in R. capsulatus, suppression of the CcoA defect occurred via mutations (two base-pair CG deletions and single base-pair C indels, respectively), inducing translational frameshifts that inactivated CopA.
Distribution of the CcoA family among Bacteria and Eukarya. Given the functional specificity of CcoA for cbb 3 -Cox in Rhodobacter species, we widened our bioinformatic search for CcoA-like MFS transporters beyond the alphaproteobacteria and queried their co-occurrence with CcoN in the SEED database (45). We found CcoA homologues in all major classes of Proteobacteria, Bacteroidetes, and Spirochaetia and in all major divisions of the Terrabacteria group, including Chloroflexi and Deinococcus (Fig. 5A). Moreover, we found that most of the genomes that contained CcoA also encoded CcoN (i.e., cbb 3 -Cox), except Actinobacteria and Firmicutes (Fig. 5B). The CcoA-like proteins were also present in the nuclear genomes of eukaryotic algae, with both primary and secondary plastids, in two fungal genomes from Chytridiomycota (Fig. 5A), in addition to the group of Actinobacteria and Firmicutes (Fig. 5B), which are known to lack cbb 3 -Cox (26). Remarkably, these "orphan" CcoA-like transporters encountered in organisms lacking cbb 3 -Cox still contained the conserved MXXXM and HXXXM motifs in helices 7 and 8, suggesting that they might also transport Cu to other protein targets. Finally, we note that, similar to alphaproteobacterial genomes, about one-third of ccoN-containing organisms also contain ccoA and exhibited species level  (Table S4). For the set of data pertinent to Fig. 5A and B, see Tables S4 and S5, respectively.

DISCUSSION
The R. capsulatus MFS-type transporter CcoA is the prototypical bacterial Cu importer and the key Cu provider to cbb 3 -Cox under limited Cu availability (32,34,37). Earlier, we observed that R. capsulatus mutants lacking either cbb 3 -Cox (ΔccoNOQP) or CcoA (ΔccoA) contained similar smaller amounts of total cellular Cu (~80% of the wild-type amount) (46), suggesting that the Cu imported by CcoA is allocated primarily to cbb 3 -Cox biogenesis. To assess the functional specificity of CcoA toward other cuproenzymes, we initiated a broad-based comparative genomic study to examine the presence of CcoA homologues and their co-occurrence with cyt c oxidases in organisms of known genome sequences. We found that the CcoA-like transporters are widespread in bacteria and some microbial eukaryotes. They are present in all major classes of Proteobacteria, Bacteroidetes, Spirochaetia, and Terrabacteria, as well as in the nuclear genomes of eukaryotic algae and fungi (Table S1). Interestingly, our all-inclusive bioinformatic analyses showed that not all CcoA family members are involved in cbb 3 -Cox biogenesis. Numerous species that have no cbb 3 -Cox, such as Actinobacteria and Firmicutes species, still contained CcoA-like transporters that possibly perform other functions that have yet to be uncovered. A closer look to the group of alpha-  Table S5. The positions of CcoA of R. sphaeroides (Rs) and R. capsulatus (Rc) in the tree are also indicated. Note that ccoA is present in Actinobacteria and Firmicutes (clusters 8 to 10) that are devoid of ccoN (i.e., cbb 3 -Cox).
Cu Importer CcoA Is Specific to cbb 3 -Cox ® proteobacteria showed that about one-third of these species contained at least one CcoA homologue together with the genes encoding cbb 3 -Cox or aa 3 -Cox, showing a high degree of co-occurrence of HCO with CcoA. Finally, similar to some genera of alphaproteobacteria, we observed species level variations in the presence of ccoA, which were particularly evident in Vibrio (Table S4), which may reflect that CcoA provides a selective advantage in some environmental niches.
Using R. sphaeroides, which contains both cbb 3 -and aa 3 -Cox (from C and A HCO families, respectively) and an ortholog of R. capsulatus CcoA (RSP_2726), we tested experimentally whether CcoA could also provide Cu to the canonical aa 3 -Cox, whose biogenesis has been studied (43,47). Physiological, genetic, and biochemical data gathered by using appropriate ΔccoA mutants lacking either cbb 3 -or aa 3 -Cox established unequivocally that the absence of CcoA affected cbb 3 -Cox, but not aa 3 -Cox, production in this organism. Earlier work had shown that the absence of the Cu chaperone Cox11, which is required for Cu B insertion into aa 3 -Cox, had no effect on cbb 3 -Cox production in R. sphaeroides (25) or Pseudomonas pseudoalcaligenes KF707 (48). Therefore, we concluded that the Cu atoms inserted into the binuclear centers of the cbb 3 -Cox and aa 3 -Cox enzymes are not only delivered by distinct pathways but also provided by different uptake systems (Fig. 6).
CcoA being an exclusive Cu importer for cbb 3 -Cox was rather unexpected, especially because the catalytic subunits and the heme-Cu binuclear centers of all HCOs are very similar (11,49). The existence of specialized Cu trafficking pathways for different cuproproteins has been documented in different organisms (50), and their specificity is generally conferred by target-specific chaperones rather than transporters (51). Thus, the independent Cu uptake systems operating during the biogenesis of different HCOs and the specificity of CcoA for cbb 3 -Cox are intriguing. Since the Sco-like and PCu A C-like Cu chaperones are involved in the biogenesis of both cbb 3 -Cox (23,27,33) and aa 3 -Cox (19,22,24,52), they are less likely to confer specificity. Thus, a possibility is that CcoA may do so by conveying Cu either directly or via an unknown partner, to CcoI, which is the P 1B -type ATPase required for cbb 3 -Cox production (29) (Fig. 6). Interestingly, the physical clustering of ccoA with the cbb 3 -Cox assembly genes ccoGHIS in members of the order Rhizobiales (Fig. 1B) suggests that these proteins function together and possibly interact during cbb 3 -Cox production. Undoubtedly, under low Cu availability, the occurrence of a membrane-integral complex containing both CcoI and CcoA would be advantageous for efficient biosynthesis of cbb 3 -Cox.
In both R. sphaeroides and R. capsulatus mutants lacking CcoA, the defect in cbb 3 -Cox production can be restored by providing a high concentration of exogenous Cu 2ϩ , which leads to an increase in cellular Cu content (34). The components of this putative CcoA-independent low-affinity Cu uptake pathway remain unknown. However, this pathway still relies on CcoI, whose absence cannot be palliated by Cu 2ϩ supplementation, to provide Cu to cbb 3 -Cox. Alternatively, the defect in cbb 3 -Cox biogenesis can be bypassed via frameshift mutations in copA, which encodes the P 1B -type ATPase involved in Cu export and detoxification, resulting in inactivation of CopA and consequent greater cellular Cu content and hypersensitivity to Cu 2ϩ (32). Elucidation of how the Cu imported by CcoA is conveyed to CcoI is needed to understand how the increase in cellular Cu content bypasses the role of CcoA in cbb 3 -Cox biogenesis. The molecular natures and locations of the suppressor mutations that inactivate CopA are different between the two Rhodobacter species. In R. capsulatus, these mutations are single base-pair C indels in a region of copA where 10 consecutive C base-pairs are located (bp 230 to 239) (32), whereas in R. sphaeroides, they are two base-pair CG deletions in a region of copA where five consecutive CG repeats are present (bp 863 to 872). Hypermutable nucleotide tandem repeats (NTRs), which are prone to DNA slippage during replication and increased recombination, are widespread in genomes of different organisms (53,54), and they can reversibly inactivate or regulate the expression of specific coding sequences (55). Computational analyses suggested that in prokaryotes, the monomeric NTRs of G/C (e.g., C repeats of R. capsulatus copA) are more mutagenic than dimeric (e.g., CG repeats of R. sphaeroides copA) or trimeric NTRs (56). The different types of mutagenic NTRs located in copA may reflect different strategies used for Cu homeostasis governing Cu availability to cbb 3 -Cox via CcoA-independent pathways. P 1B -type ATPases such as CopA and CcoI contain conserved domains for ATP binding and for phosphorylation, in addition to their N-terminal metal-binding domains (MBDs), harboring a Cu-binding CXXC motif and a membrane-embedded Cu binding site (CPX) (57) (Fig. 4C). The frameshift mutations that inactivate CopA still conserve the genetic ability to produce truncated N-terminal CopA derivatives with intact N-terminal MBDs that, if produced and stable, could hypothetically facilitate Cu delivery to cbb 3 -Cox. An R. capsulatus CopA derivative would become soluble with a single MBD, whereas an R. sphaeroides CopA derivative would remain membrane attached and conserve its two MBDs, reminiscent of the Cu chaperone CupA in Streptococcus pneumoniae. The membrane-anchored CupA protein enhances Cu sequestration and mediates its binding to the MBD of CopA as an adaptation to Cu toxicity (58). Arabidopsis thaliana chaperone PCH1 is produced by alternative splicing of the P 1B -type Cu ϩ ATPase PAA1 pre-mRNA and acts as its specific Cu chaperone (59). In Escherichia coli, a fragment of CopA containing the N-terminal MBD, resulting from programed ribosomal frameshifting during the translation of copA mRNA, is able to bind Cu and increase tolerance of Cu toxicity (60,61). The molecular mechanisms underlying these cases are distinct from the NTR mutations in copA, yet they reflect similar responses that organisms have evolved to maintain Cu homeostasis and avoid its toxicity.
The isolation of mutations in copA of both R. capsulatus and R. sphaeroides may suggest that CcoA is not required for cbb 3 -Cox metalation, depending on the mechanisms of Cu homeostasis used by the organisms. Indeed, our comparative genomic analyses indicated that CcoA-like MFS proteins are absent from about one-third of cbb 3 -Cox-containing alphaproteobacterial species. That cbb 3 -Cox metallation in these species does not require CcoA while it does so in R. capsulatus and R. sphaeroides under low Cu availability is intriguing. These species may have other Cu acquisition pathways, similar to R. capsulatus ccoA mutants at high Cu 2ϩ concentrations. As an example, P. denitrificans PD1222 has an MFS-type CcoA Cu importer and a typical P 1B -type ATPase CopA ortholog with an N-terminal heavy-metal-associated (HMA) domain acting as its MBD. In contrast, P. denitrificans SD1 does not have CcoA but has a CopA homologue with a different MBD, an N-terminal TRASH domain (62). These differences are in agreement with the proposal that the copA NTR mutations occurring after the N-terminal MBD of CopA in both Rhodobacter species may result in HMA-containing derivatives acting as chaperones. Further investigation of these species and characterization of different strains with respect to their CcoA-independent Cu trafficking pathways will be informative.
In summary, this work established that Cu incorporation into the catalytic site of different HCOs, in particular cbb 3 -Cox and aa 3 -Cox, occurs not only via distinct delivery pathways but also via distinct uptake pathways (Fig. 6). While the MFS-type transporter CcoA is required for Cu incorporation into cbb 3 -Cox, it is not involved in the metallation of aa 3 -Cox. The occurrence of dedicated Cu uptake pathways, critical for the maintenance of intracellular Cu homeostasis, might be an evolutionary example of different strategies to improve fitness encountered in many organisms.

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. The E. coli and R. sphaeroides strains and plasmids used in this study are listed in Table S3. The standard molecular biology techniques used are described in reference 63, and all plasmid and strain constructions are described below Table S3. E. coli strains were grown at 37°C in Luria-Bertani (LB) broth supplemented with 100, 50, 50, 12.5, and 12 g/ml (final concentrations) ampicillin, kanamycin (Kan), spectinomycin (Spe), tetracycline (Tet), and gentamicin (Gen), respectively (46). R. sphaeroides strains were grown in either minimal (64) or LB medium supplemented with 10, 10, 2.5, and 1 g/ml (final concentrations) Kan, Spe, Tet, and Gen, respectively (39).
Biochemical and spectroscopic techniques. R. sphaeroides cells grown under semiaerobic conditions on LB medium were harvested and resuspended in 50 mM Tris-HCl (pH 7.2), 1 mM KCl assay buffer. Intracytoplasmic membrane vesicles (chromatophores) were prepared as previously described (65). The protein concentration of membrane fractions was determined with the bicinchoninic acid assay (Sigma, Inc.). Visualization of c-type cyts was done by TMBZ staining following the separation of~200 g of total membrane proteins by 15% SDS-PAGE as done earlier (42). Immunoblot analysis to identify R. sphaeroides Cox1 was done with~40 g of total membrane proteins separated by 12% SDS-PAGE. Proteins were transferred onto polyvinylidene difluoride membranes and incubated with P. denitrificans anti-Cox1 specific polyclonal antibodies cross-reacting with R. sphaeroides protein (47). Alkaline phosphataseconjugated secondary antibodies and 5-bromo-4-chloro-3-indolyl phosphate (BCIP)-nitroblue tetrazolium were used for visualization of Cox1 polypeptide.
Visible spectra were taken with 50 g of total membrane proteins in 1 ml of assay buffer containing 0.2% n-dodecyl-␤-D-maltoside (DDM). Samples were oxidized by the addition of a few grains of potassium ferricyanide, and the absorption spectra taken between 480 and 660 nm were saved as a baseline. After reduction of the samples by the addition of a small amount of sodium dithionite, the spectra were rerecorded in the same wavelength range (39).
Determination of Cox activities. The cbb 3 -Cox activity of colonies was visualized by using the NADI reaction (␣-naphthol ϩ N=N=-dimethyl-p-phenylenediamine ¡ indophenol blue ϩ H 2 O) by staining the plates with a 1:1 (vol/vol) mixture of 35 mM ␣-naphthol and 30 mM N=,N=-dimethyl-p-phenylenediamine (66). Colonies with cbb 3 -Cox activity exhibited dark blue staining (NADI ϩ ) within 30 s to 1 min, while those with low activity or lacking it showed light blue (NADI slow ) or no staining (NADI Ϫ ) up to 15 min, respectively. Total aa 3 -Cox and cbb 3 -Cox activity levels were determined with reduced cyt c as a substrate as done previously (39). Chromatophore membranes were solubilized at room temperature by the addition of 1 mg of DDM/mg of total proteins. Activity assays were initiated by the addition of~10 g of solubilized membranes to 1 ml of assay buffer containing 25 M reduced cyt c. Rates of cyt c oxidation were determined by monitoring the time-dependent decrease in absorbance at 550 nm and expressed in micromoles of cyt c oxidized/min/mg of total membrane proteins by using the extinction coefficient at 550 nm for cyt c ( 550 ϭ 20.0 mM Ϫ1 cm Ϫ1 ). The specificity of Cox activity was confirmed by inhibition with 200 M KCN, a specific inhibitor of HCO enzymes, which stopped cyt c oxidation almost completely. Any residual cyanide-insensitive cyt c oxidase activity (air oxidation was negligible) was subtracted from the final rates.
Bioinformatic analysis. Genes encoding CcoA-like, CcoN, and Cox1 proteins were identified in the SEED database (45). In addition to amino acid sequence similarity, annotation of a protein as being CcoA-like required conservation of the MXXXM and HXXXM motifs of transmembrane helices 7 and 8. Patterns of co-occurrence and genomic colocalization were detected with the set of tools for comparative genome analysis available in SEED. For the phylogenetic trees of CcoA-like proteins, full-length amino acid sequences (Tables S1, S4, and S5) were aligned through the CIPRES web portal (67) with MAFFT on XSEDE (v. 7.305) (68) and an approximate maximum-likelihood estimation was performed with FastTreeMP on XSEDE (v. 2.1.9) (69). The resulting phylogenetic trees were visualized and annotated with the Interactive Tree of Life (iTOL) tool (70). A comprehensive identification of CcoA homologues in sequenced genomes was performed with a protein similarity network as implemented with the EFI-EST tool (http://efi.igb.illinois.edu/efi-est/) with R. capsulatus CcoA as the seed sequence, an E value of 1E-4 for the blast search, and an alignment score of 80. EFI retrieved 2,490 proteins (see Table S4), which were incorporated into the network and visualized with the yFiles organic layout provided with the Cytoscape software (http://www.cytoscape.org).

ACKNOWLEDGMENTS
This work was supported mainly by the Division of Chemical Sciences, Geosciences and Biosciences, Office of Basic Energy Sciences of the Department of Energy (DOE DE-FG02-91ER20052 to F.D.), and partly by the National Institutes of Health (NIH GM38237 to F.D.). We acknowledge the partial support provided by the Office of Biological and Environmental Research of the Department of Energy to C.E.B.-H. Support of H.-G.K. by the Deutsche Forschungsgemeinschaft (IRTG 1478 and RTG 2202) is greatly appreciated. The funding agencies had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Bernd Ludwig for providing the polyclonal antibodies specific for the Cox1 subunit of P. denitrificans aa 3 -Cox, Shalini Paliwal for help constructing plasmid pSP1, and Stefan Steimle for critical reading of the manuscript.