The CopC Family: Structural and Bioinformatic Insights into a Diverse Group of Periplasmic Copper Binding Proteins

Abstract

The CopC proteins are periplasmic copper binding proteins believed to play a role in bacterial copper homeostasis. Previous studies have focused on CopCs that are part of seven-protein Cop or Pco systems involved in copper resistance. These canonical CopCs contain distinct Cu(I) and Cu(II) binding sites. Mounting evidence suggests that CopCs are more widely distributed, often present only with the CopD inner membrane protein, frequently as a fusion protein, and that the CopC and CopD proteins together function in the uptake of copper to the cytoplasm. In the methanotroph Methylosinus trichosporium OB3b, genes encoding a CopCD pair are located adjacent to the particulate methane monooxygenase (pMMO) operon. The CopC from this organism (Mst-CopC) was expressed, purified, and structurally characterized. The 1.46 Å resolution crystal structure of Mst-CopC reveals a single Cu(II) binding site with coordination somewhat different from that in canonical CopCs, and the absence of a Cu(I) binding site. Extensive bioinformatic analyses indicate that the majority of CopCs in fact contain only a Cu(II) site, with just 10% of sequences corresponding to the canonical two-site CopC. Accordingly, a new classification scheme for CopCs was developed, and detailed analyses of the sequences and their genomic neighborhoods reveal new proteins potentially involved in copper homeostasis, providing a framework for expanded models of CopCD function.

Graphical abstract

Copper is an essential but toxic metal in biology.^1,2 It is a critical cofactor in proteins such as cytochrome c oxidase, superoxide dismutase, and numerous oxygenases, but excess copper causes oxidative damage and may disrupt iron–sulfur clusters.^1–4 To satisfy intracellular copper requirements and avoid toxicity, bacteria have evolved multiple copper homeostasis pathways. Strikingly, more than 44% of the copper proteome is dedicated to copper homeostasis.⁵ In prokaryotes, there are at least three systems that impart copper resistance: the Cue system (Cu efflux), the Cus system (Cu sensing), and the Cop/Pco system (copper resistance, or plasmid-borne copper resistance).^6–9

The Cue and Cus systems are generally believed to function in copper export and detoxification.^6,7 The Cue system typically comprises three proteins: a P_1B-type ATPase (often annotated as CopA and herein termed CopA*) that pumps Cu(I) from the cytoplasm to the periplasm, the periplasmic multicopper oxidase CueO, which couples the oxidation of Cu(I) to the reduction of molecular oxygen, and the cytoplasmic copper binding regulatory protein CueR.^10–12 In contrast to the Cue system, the Cus system does not rely on the reduction of oxygen and is therefore capable of functioning under anaerobic conditions. This system includes six proteins. The CusABC complex spans the entire periplasmic envelope and effluxes Cu(I) ions, driven by a proton gradient.¹³ CusF delivers Cu(I) ions to the CusABC complex,¹⁴ and expression of the Cus proteins is regulated by the CusR/CusS two-component system.¹⁵

While the Cue and Cus systems are strongly implicated in copper efflux, it is not yet clear how the Cop/Pco system imparts copper resistance to the cell. It has been suggested that the Cop proteins export copper,⁹ import copper to the cytoplasm,^2,16 and/or sequester copper in the periplasm.^17,18 The possibility of import has been relatively underinvestigated, likely due to the fact that there is no known function for copper in the bacterial cytoplasm beyond its involvement in copper homeostasis and gene regulation.² The traditional Cop/Pco system is composed of seven proteins, CopABCDERS. CopD and CopB are inner and outer membrane proteins, respectively (CopB should not be confused with the identically named copper-transporting P_1B-type ATPase, not traditionally part of the cop operon).⁸ CopA is a periplasmic multicopper oxidase that can substitute for CueO.¹⁹ CopE is a soluble protein that binds Cu(I) and is hypothesized to sequester copper in the periplasm.²⁰ CopC is believed to be a periplasmic copper chaperone.²¹ The PcoABCD proteins are homologous to the CopABCD proteins but were originally identified on plasmids; annotations may show either set of names. Like the cus operon, the canonical cop operon is regulated by a two-component system, the CopRS proteins.¹⁹

The proposed physiological functions of the Cop/Pco systems are based largely on phenotypic characterization of strains expressing different combinations of the cop genes and on biophysical characterization of CopC/PcoC.^8,22,23 Knockouts of the individual cop/pco genes result in increased copper sensitivity, suggesting that all four proteins work in concert to impart copper resistance.^23,24 Counterintuitively, expression of copCD without copAB leads to copper hypersensitivity, suggesting that CopD imports copper and that CopD-mediated copper import is facilitated by CopC.²⁵ A definitive model for the functions and interactions of the Pco/Cop proteins remains elusive, however. Interpretation of phenotypic data has been further complicated by biophysical studies of traditional CopC/ PcoC proteins. Until recently, CopCs were generally thought to contain both a Cu(I) binding site and a Cu(II) binding site.^21,26–28 Structural studies show that the Cu(I) ion binds in a methionine rich loop, and that the Cu(II) ion is coordinated by the amino terminus, two histidines, and a water molecule. One of the coordinating histidines is the N-terminal residue, which is generated by cleavage of a signal peptide, while the other histidine is part of a DXH motif in which the aspartic acid hydrogen bonds to the coordinating water molecule.^21,26,29,30 Biochemical studies with PcoC indicate that the Cu(I) and Cu(II) sites are capable of intermolecular copper transfer and that PcoA and CueO are capable of oxidizing the Cu(I) site.^21,28,31

Over the years, evidence that Cop/Pco systems are significantly more diverse and less similar to the traditional cue and cus copper efflux systems than was believed originally has accumulated. First, a CopC lacking the Cu(I) binding site, but with an unusually high affinity for Cu(II), has been characterized.³² The higher affinity is believed to derive from an additional coordinating histidine, which is present in the third position of the mature peptide. Second, proteins that appear to be fusions of CopC and CopD have been reported in the genomes of numerous organisms, and there is strong biochemical evidence that these Cop-like systems are involved in the import of copper into the cytoplasm.^16,33,34 Finally, operons containing copC and copD without copA, copB, or copE have been observed in multiple organisms.^18,33,35 For example, in methanotrophic bacteria, copC and copD genes are located adjacent to the operon encoding particulate methane monooxygenase (pMMO), a copper-containing enzyme that catalyzes the first step of methane metabolism.³⁶ Notably, the copD gene in a mutant strain of Methylosinus trichosporium OB3b that is impaired in cytoplasmic copper uptake and regulation contains a frameshift deletion, consistent with CopC and CopD functioning in copper import in this organism.³⁷

To further understand the methanotroph CopCD system, a CopC from M. trichosporium OB3b (Mst-CopC) was structurally characterized, and an extensive bioinformatics study was undertaken to understand Mst-CopC in the broader context of the CopC protein family. A total of 8636 CopC homologues were examined, providing an overview of the diversity of CopC sequences and operons. We find that CopCs containing two copper binding sites (named Type A by Wijekoon et al.³²) are relatively rare, accounting for approximately 10% of CopCs. The vast majority of CopCs contain only a Cu(II) site similar to that in either Mst-CopC or the recently characterized higher-affinity Pseudomonas fluorescens SBW25 CopC (Psf-CopC, named Type B by Wijekoon et al.).³² These data, in conjunction with the crystal structure of the newly identified CopC type, warrant the creation of a new system for classifying CopC-like proteins. Moreover, when combined with a growing body of evidence of CopCD-mediated copper uptake, these data necessitate a re-examination of the existing models of CopCD function.

MATERIALS AND METHODS

Materials

Unless otherwise stated, all chemicals were purchased from VWR, Research Products International, or Sigma-Aldrich and were not purified further before being used.

Construction of the Expression Plasmid

The Mst-copC gene (IMG gene ID 2507406428, locus tag Mettr-DRAFT_0380) was codon-optimized using Escherichia coli codon frequencies, synthesized by GenScript, and subcloned into pE-SUMO (Life Sensors) using the BsaI restriction site. This sequence contains a C-terminal tag encoding a TEV restriction site, a sequence corresponding to β-strand 11 of green fluorescent protein,³⁸ and a Strep-Tactin (IBA) affinity tag in that order. In place of the native N-terminal signal peptide sequence, an N-terminally encoded His-SUMO tag was introduced, which allows for generation of the authentic N-terminus (His 23) after the protein is incubated with SUMO protease.³⁹

Protein Expression and Purification

The resultant vector was transformed into chemically competent BL21 Star (DE3) (Life Technologies) cells, and one colony was used to inoculate a 1 L culture of autoinduction medium.⁴⁰ Cultures were shaken at 37 °C until they reached an OD₆₀₀ of ~0.5, at which point the temperature of the incubator was lowered to 20 °C. Following overnight growth, cells were harvested by centrifugation for 15 min at 7100g and flash-frozen for later use.

Cells were suspended in Tris buffer [50 mM Tris and 500 mM NaCl (pH 8.0)] and lysed by sonication on ice for 8 min using 1 s pulses with 3 s rests. Lysates were centrifuged for 1 h at 186000g and applied to a 15 mL Strep-Tactin affinity column (IBA) that had been pre-equilibrated with equilibration buffer [50 mM Tris and 500 mM NaCl (pH 8.0)]. Following extensive washing with equilibration buffer, Mst-CopC was eluted from the Strep-Tactin column with elution buffer [50 mM Tris, 500 mM NaCl, and 2 mM desthiobiotin (pH 8.0)].

To cleave the SUMO tag, fractions containing Mst-CopC were incubated in TCEP-containing Tris buffer [50 mM Tris, 500 mM NaCl, and 2 mM TCEP (pH 8.0)] at 37 °C overnight with excess TEV and SUMO proteases that were prepared using previously published protocols.^39,41 Cleaved Mst-CopC containing the native amino-terminal histidine residue was then applied to a Cu-loaded 5 mL HiTrap Chelating HP column (GE) to remove the SUMO protein and the His-tagged proteases. Mst-CopC was then exchanged into 100 mM NaPO₄ and 500 mM NaCl (pH 7.0) and purified using a 24 mL Superdex 200 column (GE). Protein purity was assessed by sodium dodecyl sulfate–polyacrylamide gel electrophoresis using a 15% gel, and protein concentrations were measured using the calculated molar extinction coefficient at 280 nm (5960 M⁻¹ cm⁻¹).⁴² Protein yields of 10–50 mg/L of culture were obtained.

Copper Loading and Metal Analysis

Purified Mst-CopC in 100 mM NaPO₄ and 500 mM NaCl (pH 7.0) was incubated with 10 equiv of CuSO₄ at room temperature for several minutes. Samples were then centrifuged to remove precipitated copper(II) phosphate, and unbound copper was removed by buffer exchange [into 100 mM NaPO₄ and 500 mM NaCl (pH 7.0)] on a 24 mL Superdex 200 column (GE). Metal content was measured by inductively coupled plasma mass spectrometry (ICP-MS, Thermo iCAP Q). The concentrations of iron, cobalt, nickel, copper, and zinc were measured using a multielement standard (Inorganic Ventures). ICP-MS samples were prepared in 5% nitric acid with 5 ppb indium, lithium, scandium, and yttrium internal standards (Inorganic Ventures).

Crystallization and Structure Determination

Mst-CopC was crystallized by sitting drop vapor diffusion at room temperature by mixing 1 µL of 10 mg/mL protein with 1 µL of a well solution containing 0.1 M HEPES sodium salt, 0.8 M sodium phosphate, and 0.8 M potassium phosphate (pH 7.5). Football-shaped crystals grew within 1 week. All crystals were cryoprotected in a well solution containing 50% glycerol and flash-frozen in liquid nitrogen. Diffraction data were collected at LS-CAT Sector 21 and GM/CA Sector 23 of the Advanced Photon Source at Argonne National Laboratory (Argonne, IL) at an energy of 12 keV. HKL-2000 was used to index, integrate, and scale all data sets.⁴³ Molecular replacement using E. coli PcoC [Esc-PcoC, Protein Data Bank (PDB) entry 1LYQ]²⁶ as a starting model was used to determine phases. Molecular replacement solutions were calculated using Phaser, and structures were refined using REFMAC as implemented in the CCP4 program suite (Table 1).^44–46 During the final stages of refinement, TLS and anisotropic refinement parameters for the copper ion were added to the model.

Table 1.

Data Collection and Refinement Statistics of Cu- Loaded Mst-CopC

space group	P212121
resolution (Å)	1.46 (1.50–1.46)
Rsym	0.048 (0.067)
I/σI	41.9 (22.1)
completeness (%)	99.9 (99.9)
redundancy	8.6 (8.2)
no. of reflections	17434
data cutoff	F > 0
no. of atoms	876
Rwork/Rfree (%)	16.6/19.1
B factor (Å2)
average	8.5
protein	7.7
water	20.3
ligands/ions	18.4
root-mean-square deviation
bond lengths (Å)	0.0141
bond angles (deg)	1.661

Identification of CopC Homologues in Bacterial Genomes

All protein sequences matching the CopC PFAM profile hidden Markov model (HMM) (PF04234) and associated metadata were obtained from the JGI-IMG database (http://img.jgi.doe.gov/w/, September 17, 2014) (Supplemental Files 1 and 2). To avoid the loss of potentially relevant minor differences at copper binding sites, the data set was analyzed with and without pruning on the basis of sequence similarity. In the pruned data set (Supplemental Files 3 and 4), sequences were clustered according to similarity using UClust,⁴⁷ and in groups with >95% similarity, a centroid sequence was chosen to represent the cluster, resulting in a smaller data set of 2550 sequences. Unless otherwise specified, this is the data set discussed in the body of this paper, and discussion of the auxiliary unpruned data is relegated to the Supporting Information. All sequences were aligned against the HMM via hmmalign (Supplemental Files 5–9),⁴⁸ and additional domains not matching the HMM were trimmed for further sequence-based analyses.

Sequences were classified into one of several categories depending on the conservation of their Cu(I) and Cu(II) sites (Table 2). The Cu(I) site was considered to be present if, after alignment with the CopC HMM, residues corresponding to at least four of the five residues, M63, M66, M69, H72, and M75 (using Esc-PcoC residue numbering, IMG entry 640940805), were conserved. In the case of H69 and M75, the presence of an M or H within one residue of that position was considered sufficient for conservation of the residue. The Cu(II) site was considered to be present if, after alignment with the CopC HMM, residues corresponding to H24, D113, and H115 were conserved. Sequences containing a Cu(II) site with an additional conserved histidine at H26 were assigned to a separate category, higher-affinity Cu(II) binding sequences (using Esc-PcoC residue numbering, IMG entry 640940805). Systemic misalignments appear to be minimal or due to the presence of significant insertions, deletions, or multiple-residue differences in sequence with the potential to alter the structure or function of the resulting proteins. For the generation of consensus sequences in Jalview,⁴⁹ only residues fitting the HMM were used [along with residues or gaps in the alignment corresponding to Esc-PcoC M51, which is part of the Cu(I) binding site].

Table 2.

Prevalence of CopC Types and Selected Supersets^a

	Type	Cartoon of copper ions	Cu(I) ligand set (M40, M43, M46, H48, M51)	Cu(II) ligand set (H1, H3, D79, H81)	Number of Sequences (Unpruned)	Number of Sequences (Pruned)	Typical protein name	Classification in Wijekoon et al.32
CopC types	C0−1		<4	3(H3 excluded)	3,807 (44.1 %)	1,832 (71.8%)	CopC, YcnJ	-
	C0−2		<4	4(H3 included)	3,471 (40.2 %)	332 (13.0%)	YobA,CopC	Type B
	C1−1		≥4	3(113 excluded)	930 (10.8%)	137 (5.4 %)	PcoC, CopC	Type A
	C0−0		<4	<3(H3 excluded)	411 (4.8 %)	238 (9.3 %)	-	-
	C1−0		≥4	<3(H3 excluded)	17 (0.2 %)	11 (0.4 %)	-	-
CopC supersets	No Cu(I)		<4	-	7,689 (89.0 %)	2,402 (94.2 %)	-	-
	Cu(I)		≥4	-	947 (11.0%)	148 (5.8 %)	-	-
	No Cu(II)		-	<3(H3 excluded)	428 (5.0 %)	249 (9.8 %)	-	-
	Cu(II) only		<4	3(H3 excluded)	7,278 (84.3 %)	2,164 (84.9 %)	-	-

Cu(I) Cu(II)2 His, 1Asp Cu(II) 3 His

^aThe “Type” column shows the CopC types defined herein with pictorial representations of each type shown in the adjacent column. The shaded columns indicate the ligand sets used to define each CopC type. The “Number of Sequences (Unpruned)” column shows the number of all known CopC sequences (8636 total) with the specified ligand set. The “Number of Sequences (Pruned)” column shows the number of CopC sequences with the specified ligand set after highly similar sequences (>95%) have been removed (2550 total). Supersets include all CopCs with the listed sequence feature; e.g., “Cu(I)” includes any sequence containing a Cu(I) site, regardless of Cu(II) site residues.

Network Analysis of CopC Sequences

Sequence similarity networks were generated as described previously.⁵⁰ Nodes represent the trimmed CopC amino acid sequences, and edge values were generated via an all-by-all BLAST+ (NCBI) against a database consisting of either the 8636 raw or the 2500 pruned CopC sequences (Supplemental Files 2–6), trimmed but unaligned. Initial networks were generated using the Blast2Sim plug-in⁵¹ in Cytoscape 2.8.3⁵² with an initial BLAST threshold of 100, a coverage factor of 15, and a “sum of all hits” similarity function. Edges corresponding to low similarity were removed according to a numeric filter; filter values were set below a level that would result in >1% of the nodes dissociating from the main cluster. Transitivity clustering was applied via the TransClust plugin, using a threshold of 35 with the Blast2SimGraph_sim score as an edge value and similarity as an edge weight determiner.⁵¹ In the final cluster analysis, edges with a value below 35 were deleted and the network was organized according to the edge-weighted, force-directed Cytoscape layout.

A tab-delimited text file identifying characteristics of interest for each node was loaded and overlaid onto the network via the Cytoscape VizMapper plug-in. Copper site assignment was calculated as described above. The genus and gene length were obtained from exported JGI metadata for each gene, while the phylum was determined using the NCBI Taxonomy Database on the basis of genus (http://www.ncbi.nlm.nih.gov/taxonomy). Operon categories were assigned as described below.

Hierarchical Clustering of CopC Genome Neighborhoods

In the JGI-IMG database, each gene has a gene object id (gene_oids). On a DNA contig, scaffold, or chromosome, neighboring ORFs are assigned contiguous numeric gene oids. On the basis of this information, metadata were obtained for all genes within five ORFs of copC (i.e., all genes with gene_oids ±5 from the gene_oid of the relevant copC). Neighboring genes were classified by annotated Pfam⁵³ and TIGRFAM⁵⁴ profile HMMs present in the gene, and for genes observed near >1% of copCs in a given position, the distance in number of genes from copC was recorded, from 0 (indicating fusion with copC) to 5. A binary value was assigned to indicate the presence or absence of a given trait at a given distance from the copC gene; in a separate set of analyses, a binary value was assigned to the presence or absence of those traits in the genomic neighborhood of copC regardless of their relative positions. Several other traits were also assigned binary numeric values for each gene: presence of a given type of copper site, membership in one of the 15 most common genera in this data set, and membership in the most common phyla. Only gene presence or presence/distance values and copper site values were used for hierarchical clustering.

The gene IDs and their associated traits were subjected to hierarchical clustering in R.⁵⁵ For both genes and traits, the distance matrix was calculated using the Manhattan method (absolute distance) in the dist function, and the clustering was determined using the Ward.D2 method⁵⁶ in the hclust function (Supplemental File 10). A heatmap and dendrograms were generated from the hierarchical clustering data via the heatmap.2 function in gplots.⁵⁷ Co-clustered sets of traits were presumed to represent potential operons, and five randomly selected copCs from each of the 20 largest operon families were checked to visually confirm the clustering results. Similar sets of analyses were performed using the unpruned CopC data set (Figure S1, SI Supplemental File 11). Only genes with unidentified roles or with roles plausibly related to copper homeostasis were considered to be potential operon members, although all gene neighbors are represented in the raw data in the Supporting Information (Supplemental File 5). In some clusters with more morphological variation, only the core operon structure observed in the significant majority of the cluster is depicted.

RESULTS AND DISCUSSION

CopC from M. trichosporium OB3b

The cloned copC gene used in this work is located downstream of the pmoCAB operon, which encodes the three subunits of pMMO, between a gene termed pmoD and the copD gene.⁵⁸ The copD gene is followed by a gene encoding a putative member of the DUF461 family, which includes the proteins known as PCu_AC, PcuC, and EcuC.^35,59–61 This entire gene cluster forms an operon that is coregulated in response to copper.³⁷ The predicted Mst-CopC protein includes an N-terminal signal peptide for export to the periplasm. Previous studies of CopC proteins indicate that cleavage of the signal peptide results in an N-terminal histidine, which coordinates a Cu(II) ion with both the N-terminal amino group and the imidazole side chain.²¹ Accordingly, the N-terminal histidine in Mst-CopC was generated by genetically inserting a SUMO tag before the copC gene. Cleavage of this tag yields the native N-terminus, and ICP-MS analysis of the resulting protein indicates that 1.0 ± 0.3 (n = 4) copper ions bind per monomer after metal loading.

The crystal structure of Mst-CopC was determined to 1.46 Å resolution in space group P2₁2₁2₁ with a single molecule in the asymmetric unit (Table 1 and Figure 1). There is no evidence of biologically relevant oligomerization in the crystal packing or for interference of crystal lattice contacts with the metal center. All regions of the protein are well-ordered, and all of the residues following the signal peptide (residues 23–120) in addition to a C-terminal TEV cleavage site were modeled. Similar to previously characterized CopCs, Mst-CopC forms a seven-stranded β-barrel, albeit with a slightly different arrangement of β-strands. The crystal structure reveals a single copper ion, modeled at an occupancy of 1.0.

Figure 1.

Structural alignment of Mst-CopC (light pink) and Pss-CopC (purple, PDB entry 2C9Q). Copper ions are shown as spheres.

Two CopCs have previously been structurally characterized: PcoC⁶² encoded on plasmid PRJ1004^23,63 found in an E. coli strain isolated from a piggery (Esc-PcoC)²⁶ and CopC⁶⁴ from Pseudomonas syringae pv. tomato (Pss-CopC).^21,24,29,30 These CopCs contain Cu(I) and Cu(II) binding sites, located at opposite ends of the molecule. In Pss-CopC, the Cu(I) ion is coordinated by methionine(s) and a histidine (Figures 1 and 2), with variable coordination observed in crystal structures obtained at different pH values and in different crystallographic space groups.²¹ The Cu(II) ion is coordinated in square planar geometry by two histidines, the amino terminus, and a water molecule, with an aspartic acid within hydrogen bonding distance of the ligands (Figure 2B). At low pH values in a different space group, an altered coordination involving a histidine from a second molecule is also observed. Copper coordination was not observed in the Esc-PcoC structure because the recombinantly produced protein used for crystallography did not include authentic N-terminal histidine, which is found in the native protein.

Figure 2.

Topology diagrams of (A) Mst-CopC and (B) Pss-CopC and Esc-PcoC. Cu(II) and Cu(I) ions are shown as dark and light blue spheres, respectively. Water is shown as small red spheres.

Mst-CopC contains only the Cu(II) binding site and completely lacks the conserved residues necessary for Cu(I) binding (Figure 1). The residues near the Cu(II) site, H23, D105, H107, and the amino terminus, are similar to those observed in Pss-CopC²¹ but coordinate Cu(II) differently (Figures 2 and 3A). In Mst-CopC, Cu(II) is bound in a distorted square pyramidal geometry with H23, H107, D105, and the amino terminus comprising the planar ligands and a water molecule serving as the axial ligand (Figure 3B). The D105 coordinating distance is 2.30 Å, whereas the histidine nitrogen to copper distances range from 1.96 to 2.06 Å. The 2.30 Å distance is likely due to anisotropy along the D105 H23 imidazole axis, as evidenced by egg-shaped electron density in the copper omit map (Figure 3B). The distance between the copper ion and water molecule is 2.37 Å. This elongation is likely a result of Jahn-Teller distortion. Similar to Met-CopC, P. fluorescens SBW25 CopC (Psf-CopC) also contains only a Cu(II) binding site, which was characterized biochemically and spectroscopically. The Psf-CopC Cu(II) site has a Cu(II) affinity 2 orders of magnitude higher than that of Pss-CopC, which was attributed to the presence of an additional histidine ligand in the third sequence position of the mature peptide (Figure 4).³² Notably, this residue is absent in Mst-CopC. Thus, Mst-CopC appears to be a Cu(II)-only chaperone distinct from the two types of CopCs previously characterized: those that bind both Cu(I) and Cu(II) (e.g., Pss-CopC and Esc-PcoC) and those that bind only Cu(II) with an additional histidine ligand (e.g., Psf-CopC).

Figure 3.

Coordination environment of the Cu(II) site. (A) Structural alignment of the Cu(II) site in Mst-CopC (light pink) and Pss-CopC (purple, PDB entry 2C9Q). Copper ions are shown as large spheres, and water molecules are shown as small spheres. (B) Electron density map (2F_o − F_c map contoured at 1.6σ, colored gray) and difference density map (F_o − F_c map contoured at 4.5σ with positive density colored green, negative density not observed at this σ, and copper and water omitted) of the Mst-CopC Cu(II) site.

Figure 4.

CopC sequence logos and alignment. Residues that do not match a position in the profile HMM have been removed for the sake of clarity except for M72, which is a key Cu(I) binding residue; it should be noted that M72 and H75 are located in a somewhat variable loop region. Copper binding residues are highlighted in blue (histidine), red (aspartate), or green (methionine). Representative sequences are from the following organisms: C₁₋₁, Pseudomonas sp. GM17 (IMG ID 2511266057); C₀₋₁, M. trichosporium OB3b (IMG ID 2507406428); C₀₋₂, P. fluorescens R124 (IMG ID 2503652063); C₀₋₀, Mycobacterium tusciae JS617 (IMG ID 2508745664); and C₁₋₀, Sphingobium yanoikuyae XLDN2-5 (IMG ID 2549030586).

The fold of Mst-CopC is the same as that of Pss-CopC and Esc-PcoC, but the β-strands adopt a slightly different arrangement (Figures 1 and 2). In particular, a “dog leg”⁶⁵ is introduced into β-strand 4, resulting in secondary structure contacts with β-strands 3 and 5 and the formation of a tighter loop between β-strands 3 and 4 (Figure 1). The contacts between β-strand 3 and the “dog leg” consist of backbone hydrogen bonds from a single pair of amino acids (labeled β-strand 4B in Figures 1 and 2), whereas four pairs of amino acids mediate the interaction between the “dog leg” and β-strand 5 (labeled β-strand 4A in Figures 1 and 2). By contrast, β-strand 4 of Pss-CopC contacts only β-strand 5, resulting in a more extensive loop that also includes the Cu(I) binding residues. It has been suggested that the Cu(I) bound in this loop can be oxidized by periplasmic oxidases,²⁸ and the extended loop may play a role in protein-protein interactions between Cu(I) binding CopCs and cognate oxidases or other partner proteins. Although the overall structure of all three CopCs is similar, the differences in metal binding site arrangement and ligands are consistent with Mst-CopC representing a new type of CopC.

Bioinformatic Classification of CopC Subfamilies

To determine if the Mst-CopC type is unique to methanotrophs or more prevalent, we performed an extensive bioinformatics study. At the time of analysis, the JGI-IMG database contained 8636 genes predicted to include a CopC-like domain. After alignment against the CopC profile HMM (PF04234), these genes were categorized on the basis of the composition of the copper sites, considering each of the two sites independently. This classification scheme is more comprehensive than previous analyses, in which CopCs were identified on the basis of the presence or absence of the Cu(I) binding site.³² We have assigned each sequence according to whether the Cu(I) site is present in the canonical form represented by the Pss-CopC and Esc-PcoC ligand set (1) or absent (0) and defined three categories for the Cu(II) site: absent (0), present in the canonical/Mst-CopC form (1), or present in the higher-affinity form with an additional histidine ligand (2) (Table 2 and Figure 4). Using this nomenclature, we have given each CopC class a designation consisting of “C” followed by two subscripted digits, of which the first identifies the Cu(I) site type and the second identifies the Cu(II) site type, resulting in five categories: C₀₋₁, C₀₋₂, C₁₋₁, C₀₋₀, and C₁₋₀, in order of prevalence (no C₁₋₂ CopCs were identified). The sequences were also grouped into CopC supersets, including combinations of multiple types (Table 2). Surprisingly, this analysis indicates that C₀₋₁ CopCs, of which Mst-CopC is the founding member, are the most abundant (71.8%), followed closely by members of the C₀₋₂ class (13.0%). Thus, the overwhelming majority of CopCs contain only the Cu(II) site, and only 5.4% of the sequences belong to the C₁₋₁ class, upon which most previous studies and functional hypotheses have been based. A total of 9.3% of the sequences lack one or more ligands to either copper site (C₀₋₀), and a very few (0.4%) appear to have only the Cu(I) site (C₁₋₀).

To improve our understanding of this newly discovered CopC diversity, similarity network analysis was performed using nodes composed of trimmed CopC sequences and edge values generated by an all-by-all BLAST. The resulting clusters were examined for the following traits: CopC type, source organism phylum, and length as a proxy for detecting fusions to other proteins (Figure 5). The results indicate that type C₀₋₁ is the most widely distributed, appearing in the actinobacteria, proteobacteria, and firmicutes; it is clearly not limited to methanotrophs. By contrast, C₀₋₂ and C₁₋₁ CopCs are mostly confined to three clusters within the Proteobacteria phylum (Figure 5A, B). The C₀₋₂ CopCs are found primarily in gamma-proteobacterial genera like Escherichia and Pseudomonas.³²

Figure 5.

CopC sequence similarity networks colored by (A) copper site, (B) phylum, (C) full sequence length in amino acids, and (D) CopD presence, absence, or fusion. Labeled variants correspond to CopC sequences highlighted in Figure 4 [PsG, Ps. sp. GM17; Mst, M. trichosporium OB3b (IMG ID 2507406428); Psf R, P. fluorescens R124 (IMG ID 649637371); Myt, My. tusciae JS617; and C₁₋₀Spy., Sp. yanoikuyae XLDN2-5].

In terms of predicted final polypeptide length, the C₁₋₁ and C₀₋₂ CopCs are typically 90–100 amino acids and do not contain additional domains (Figure 5C). Although some C₀₋₁ CopC sequences, including the Mst-CopC sequence, are similar in length and composition to the C₁₋₁ and C₀₋₂ CopCs, a significant subset of these sequences are fused to CopD homologues. These CopCD fusions are especially common in Gram-positive bacteria. The YcnJ CopCD fusion protein from Bacillus subtilis has been investigated by gene disruption and copper regulation studies as well as whole cell copper uptake experiments.^16,33 Some of the fused genes encode additional C-terminal domains, in particular the functionally uncharacterized YktA domain (PF05256); both the fused CopC and the C-terminal domains are predicted to be periplasmic in Gram-negative bacteria and external to the cell in Gram-positive bacteria. In Gram-positive bacteria, C₀₋₁ CopC sequences not fused to CopD are frequently longer than CopCs in other classes likely due to the need for an N-terminal transmembrane region that tethers the protein to the external surface of the Gram-positive cell membrane (Figure 5C).

Genomic Neighborhood Analysis

To gain insight into CopC functional diversity, we analyzed the genomic neighborhoods of all the copC genes. For each copC gene, predicted genes within a five-gene radius were identified via profile hidden Markov models (TIGRFAMs and PFAMs), as were copper site types and the presence of any fusion proteins. These traits and the CopCs associated with them were submitted to hierarchical clustering to identify conserved gene groups and sets of CopCs with similar genomic neighborhoods (Figure 6). In addition, investigation of particularly common clusters allowed us to eliminate genes likely to be irrelevant to copper homeostasis on the basis of predicted function, relative distance and direction, and membership in an overrepresented genus. Examination of the 21 remaining common operon types highlights possible functional differences between CopC subfamilies as well as genes encoding several previously uninvestigated proteins that may be involved in copper homeostasis (Figures 5D and 6).

Figure 6.

Hierarchical clustering of copC genes and traits [including genome neighborhood, copper site identity, and the presence of additional domains within the copC gene (e.g., copD, ytkA)]. copC genes are clustered by which traits are observed in and around them (left), while traits are clustered by the copC genes to which they are connected (top); correlation between a given copC gene and trait is indicated by black (central heatmap). The 21 most widespread sets of neighboring genes are depicted (bottom). In some cases, multiple potential operon architectures are seen; the most common one is depicted. Neighboring genes to either side with no predicted relationship to metal homeostasis have been omitted. The copC genes in operons 2–8 are often adjacent to a range of other copper-related genes and operons beyond the five-gene neighborhood monitored in this data set; these genes include cusABCF, other cop genes (copE, copK, and copZ), regulatory proteins (cueR, additional cusRS/copRS), and disulfide isomerases.

The C₁₋₁ copC genes are almost invariably found preceding copD and are frequently found in operons containing other cop machinery (Figures 5 and 6). Common examples include copABCDRS, copBARSA*CD, and copAB(DUF411) (DUF2933)A*CD. Proximity to czc and cus metal homeostasis operons is common, as is the existence on a plasmid rather than a chromosomal operon. Of special note in these putative cop operons are DUF411 and DUF2933 gene sequences. DUF411 has occasionally been conflated with several different regulatory proteins (including the CopY family repressor CopR⁶⁶ and the characterized repressor CopG)^67,68 and as a potential exporter;⁶⁹ the sequences of DUF411 proteins support a third role, namely a periplasmic copper binding protein.^70,71 It has been hypothesized that DUF411 is involved in copper resistance,^70–72 whereas DUF2933 (occasionally annotated as CopO) is completely uncharacterized. Both hypothetical proteins also appear in the proximity of cus operons. The 11 type C₁₋₀ copC genes are found in operons that are similar to C₁₋₁ copC-containing operons. It is not clear that C₁₋₀ CopCs are a significant and distinct group of CopCs; rather, they seem to be the result of mutations to the Cu(II) site in otherwise C₁₋₁ CopCs.

The C₀₋₂ copC genes, which are frequently (although not invariably) annotated as yobA,⁶ occur almost exclusively before separate copD genes (often annotated as yebZ). Sometimes these two genes comprise the entire operon, but in many cases, the copD homologue is followed by an uncharacterized DUF2511 gene sequence (sometimes annotated as yebY) (Figure S1 and Figure 6). DUF2511 is predicted to be a small periplasmic protein, and the proximity of its gene to those encoding CopC and CopD homologues suggests that it too may be involved in copper homeostasis. However, unlike the C₁₋₁ copC genes, C₀₋₂ copC genes do not appear in the genomic proximity of other cop genes, and they are broadly chromosomal rather than plasmid-based.

The genome neighborhoods of the C₀₋₁ copC genes, which include Mst-CopC, are quite varied. This diversity is consistent with the range of species containing these genes (Figure 6). DUF1775 sequences (PF07987, sometimes annotated ycnI) are frequently adjacent to copC or copD genes, and this protein is produced in response to copper in some species, though the function remains unknown.⁷³ Two other members of the ycn operon are sometimes observed, although primarily in Bacillus species and related genera; examples include the DeoR family regulator ycnK (PF08220) and a reductase or disulfide isomerase sometimes annotated as ycnL. Other frequent neighbors across species and genera include genes encoding DyP-type peroxidases (PF04261), periplasmic copper binding proteins belonging to the DUF461 family (PF04314, also known as PCu_AC, PcuC, and EcuC),⁶¹ and genes encoding other periplasmic copper chaperones resembling Sco1 and SenC (PF02630).⁷⁴ Perhaps because of their wide distribution across many species, likely aided by horizontal gene transfer, C₀₋₁ copC genes occur without a neighboring copD gene at a frequency higher than those of other copC types (Figure 5D), although even these isolated copC genes frequently occur in the genomic vicinity of genes encoding DyP peroxidases, DUF461 proteins, Sco1 homologues, and YcnI homologues.

Unlike the type C₁₋₀ CopCs, many type C₀₋₀ CopCs appear to be distinct CopC variants. They lack one or more metal binding residues at the copper binding sites or have regions of divergent length or sequence composition that render comparison to the standard HMM difficult to interpret. Several clusters are seen within groups of related species (Figure 5), suggesting that these sequences have been maintained despite likely alterations in copper binding. The genes encoding C₀₋₀ CopCs are found next to or fused with copD a little less than half the time. Given that many of CopCs encoded by these genes are unlikely to bind copper or to have Cu(I) or Cu(II) sites resembling those of established CopC families, their function remains ambiguous.

Functional Implications

The crystal structure of Mst-CopC and accompanying bioinformatic analysis reveal a new CopC subfamily, the C₀₋₁ CopCs, characterized by the presence of a single Cu(II) binding site similar to those found in the previously characterized C₁₋₁ CopCs.^21,26,29,30 This subfamily is broadly distributed throughout the microbial world and is in fact the most common form of CopC (Table 2). Combined with the higher-affinity Cu(II) C₀₋₂ subfamily, CopCs that bind only Cu(II) comprise approximately 85% of all CopC family proteins. Thus, previous functional roles suggested for CopC, including delivery of Cu(I) to outer membrane components or to the multicopper oxidase CopA as a source of reducing equivalents,⁷⁵ as well as models involving intra- or intermolecular transfer of copper ions between the two sites in CopC or to other partners such as CopA, CopB, and CopD,^21,30 may not be applicable to the broader CopC family. Whereas the C₁₋₁ copC genes are typically found in operons with these previously proposed partner proteins, the genomic neighborhoods of C₀₋₁ copC genes are variable. The copD gene is usually present, however, and in many instances, the encoded CopCD pairs or CopCD fusion proteins appear to be functional units by themselves. This observation is particularly notable given the increasingly strong evidence of the role of CopD as a copper importer, despite the lack of a well-understood nonregulatory role for copper in the cytoplasm of microorganisms.^{16,18,25,33,37}

It is likely that the C₀₋₁ and C₀₋₂ CopCs are periplasmic Cu(II) chaperones for a Cu(II) import system that utilizes CopD for transport across the inner membrane followed by reduction to Cu(I) in the cytoplasm.^2,16 These CopCs could also donate Cu(II) to other proteins in the periplasm, including Cop proteins. In this scenario, the C₁₋₁ CopCs would function similarly and may also participate in Cu(I) detoxification. As noted previously, the presence of the two distinct sites for Cu(I) and Cu(II) is unusual,³⁰ and Cu(I) binding could serve a specialized function in some organisms. Alternatively, the methionine rich Cu(I) binding site could function in protein-protein interactions,²⁶ not necessarily always binding Cu(I) in vivo. Methionine residues are often found in patches at protein interaction interfaces.^76,77

The full range of functional roles for CopC and CopD remains unclear. In some systems, copper imported by CopD clearly interacts with cytoplasmic regulatory proteins. For example, copper imported by B. subtilis YcnJ, a CopCD fusion protein, binds the copper-dependent regulator CsoR, resulting in expression of the copZA genes encoding a cytoplasmic copper efflux system, which pumps copper to the extracellular matrix.³³ In P. fluorescens SBW25, the Cop and Cue systems work in concert to regulate copper levels.¹⁸ In support of a similar model in methanotrophs, a frameshift/deletion in the copD gene adjacent to the gene encoding Mst-CopC results in a copper-starved phenotype, suggesting that copper is not reaching cytoplasmic regulators.³⁷ This type of model likely applies to other systems and diverse cytoplasmic regulators. For example, in E. coli, copper imported by CopC and CopD could activate the Cue system as well as the outer membrane protein ComC, which reduces the permeability of the outer membrane to copper. These systems are regulated by the copper binding transcription factors CueR⁷⁸ and ComR,⁷⁹ respectively.

Beyond import for regulation, it is also possible that CopC and CopD import copper into the cytoplasm for protein assembly. In Rhodobacter capsulatus, assembly of active cbb₃-type cytochrome c oxidase (COX) depends on CcoA, a member of the major facilitator superfamily involved in the uptake of copper to the cytoplasm.^80,81 The frequent appearance of periplasmic copper proteins implicated in COX assembly, including DUF461 and Sco1 homologues,^{35,59–61,82–84} in the genomic neighborhood of C₀₋₁ copC and copD genes suggests that CopD could function analogously to CcoA. In support of this notion, at least 15% of copD gene sequences contain a C-terminally fused domain annotated as ctaG (not to be confused with the COX11/CtaG protein family), which is a transmembrane protein implicated in cytochrome c biogenesis in Bacillus species.^37,85,86 The presence of multiple copC genes in 37.6% of CopC-producing species may also support functional roles of CopC and CopD other than regulation (SI File 4). Copper imported by the CopCD system might also be used to metalate fully folded copper-containing proteins that are exported from the cytoplasm via the twin arginine translocation (TAT) pathway. Disruption of the TAT secretion pathway affects metal homeostasis in E. coli,⁸⁷ and numerous copper proteins are exported by the TAT pathway, including multicopper oxidases, nitric oxide reductases, and nitrite reductases.⁸ This mechanism of metal loading has been previously proposed for CueO⁶ and PcoA.²⁸ Finally, in the case of the relatively abundant copCD mini-operons, it seems equally probable they could import copper for regulation, for loading, or for both functions.

In this study, structural biology and bioinformatics have been used to significantly expand and thoroughly categorize the CopC family of periplasmic copper binding proteins. Contrary to the established model, the CopC family is dominated by proteins that bind only Cu(II), namely, the recently discovered C₀₋₂ CopCs and the newly identified C₀₋₁ CopCs. The vast majority of these genes are not found in canonical cop operons, and the composition of their genomic neighborhoods depends in part on the identity of their copper sites. Several conserved and uncharacterized proteins potentially associated with copper homeostasis can be identified on the basis of their genomic association with copCD pairs. Taken together, these data indicate that truly representative cop operons and CopC copper binding properties cannot be defined. Exploring the range of potential CopC functions in the context of this previously unappreciated diversity will form the basis for further studies of the widespread yet poorly understood CopCD copper homeostasis system.

ABBREVIATIONS

COX

cytochrome c oxidase

Esc-PcoC

E. coli PcoC

HMM

hidden Markov model

Mst-CopC

M. trichosporium OB3b CopC

Psf-CopC

P. fluorescens SBW25 CopC

Pss-CopC

P. syringae pv tomato

TAT

twin arginine translocation