A hint for the function of human Sco1 from different structures

Abstract

The solution structures of apo, Cu(I), and Ni(II) human Sco1 have been determined. The protein passes from an open and conformationally mobile state to a closed and rigid conformation upon metal binding as shown by electrospray ionization MS and NMR data. The metal ligands of Cu(I) are two Cys residues of the CPXXCP motif and a His residue. The latter is suitably located to coordinate the metal anchored by the two Cys residues. The coordination sphere of Ni(II) in solution is completed by another ligand, possibly Asp. Crystals of the Ni(II) derivative were also obtained with the Ni(II) ion bound to the same His residue and to the two oxidized Cys residues of the CPXXCP motif. We propose that the various structures solved here represent the various states of the protein in its functional cycle and that the metal can be bound to the oxidized protein at a certain stage. Although it now seems reasonable that Sco1, which is characterized by a thioredoxin fold, has evolved to bind a metal atom via the di-Cys motif to act as a copper chaperone, the oxidized form of the nickel-bound protein suggests that it may also maintain the thioredoxin function.

Sco is a family of proteins ubiquitous to all kingdoms of life. Ortholog and paralog genome browsing has shown that one or more representative of this class are present in most bacterial and eukaryotic genomes (1, 2). In the bacterial operons, Sco proteins often are associated with copper enzymes, suggesting that they are involved in the maturation or functioning of such enzymes (1). Eukaryotic genomes contain two paralogs, Sco1 and Sco2 (3, 4), both involved in copper-dependent assembly of cytochrome c oxidase (CcO) (5). CcO contains two functional copper ions located in the binuclear Cu_A site and one located in the binuclear Cu_B-heme a₃ site (6). CcO is a multimeric enzyme complex embedded in the inner mitochondrial membrane of all eukaryotes and in the plasma membrane of prokaryotes, and it functions as terminal enzyme of the respiratory chain (7). Sco1 was first suggested to be involved in copper ion delivery to the CcO complex based on the observation that, in the presence of high copper concentrations, overexpression of either Sco1 or the homologous Sco2 can restore the CcO activity of Saccharomyces cerevisiae strains lacking the gene of the mitochondrial copper chaperone Cox17 (8). Yeast strains lacking Sco1 also are respiratory deficient, and an excess of copper and/or overexpression of either Cox17 or Sco2 cannot compensate for the Sco1-associated CcO deficiency (8). The absolute requirement of Sco1 in the activation of CcO indicates that Sco1 functions downstream from Cox17 in the delivery of copper to CcO. Indeed, in vitro Sco1 can receive copper from the copper chaperone Cox17 (9).

Human Sco1 (HSco1) is a 301-residue polypeptide anchored through a single helix to the inner mitochondrial membrane of eukaryotes (10). The functional part of the Sco1 protein is composed of a single soluble domain, located in C-terminal region, whereas the N terminus contains a mitochondrial-targeting sequence followed by a transmembrane helix (11). The structure of the soluble domain was first resolved by NMR for apoSco1 from Bacillus subtilis (12). The structure revealed a potential metal binding site constituted by two Cys residues present in a conserved motif CPXXCP and a fully conserved His residue, in agreement with earlier extended x-ray absorption fine structure investigations of yeast Cu(I)Sco1 (13). Besides extended x-ray absorption fine structure data many other spectroscopic data on the human, yeast, and B. subtilis Sco proteins have confirmed that the protein binds Cu(I), Cu(II), and other metal ions (14–17). The fold of apoSco1 protein, which contains four α-helices and seven β-strands organized in two β-sheets (12), is atypical for a metal chaperone given that it resembles the fold of thioredoxins, which are enzymes specialized for the reduction of protein disulfides (18). Also, the x-ray structures of the apo forms of the B. subtilis Sco1 and HSco1 proteins have been determined (16, 19). Some crystals of B. subtilis apoSco1 contained Sco1 with S—S bonds, thus supporting a role of the protein in redox processes. In addition, it has been recently suggested on the basis of the extreme sensitivity of the yeast sco1-null mutant to hydrogen peroxide that HSco1 might function as a mitochondrial redox signaling molecule (19). Thus, the specific role of Sco1 in maturation of CcO, either as a copper chaperone or connected with redox processes, is still elusive.

To date there are no structures available for any metallated forms of Sco1 proteins. The structure of the metal adducts is, however, crucial for understanding the mechanism of Sco-mediated copper insertion into CcO. We have succeeded in preparing a human Cu(I)Sco1 derivative [hereafter referred to as Cu(I)HSco1] in vitro and have determined its solution structure through NMR. We also have investigated the structure of the Ni(II) derivative of HSco1 as a model for the binding of bipositive metal ions like Cu(II). We would like to stress that NMR solution structure determination of metalloproteins is a difficult challenge as far as the protein metal-binding mode is concerned because NMR does not provide direct information on protein–metal interaction. In contrast, it is also common that the crystallization procedures might provide derivatives different than the physiological ones. Still, these derivatives may be significant as models of transient species and for the mechanism of action. Indeed, the x-ray structure of human Ni(II)Sco1 [hereafter referred to as Ni(II)HSco1], also reported here, displays a completely different metal binding with respect to the solution structure, being the metal bound to the S—S bond of the oxidized HSco1. Structural information for different conformers of the same protein is extremely valuable because individual structures often mimic transient species, which enable elucidation of the mechanism of protein action.

Here, we propose that the Cu(I)HSco1 species act as chaperones for copper ion delivery and that the specific oxidation of metal-ligating thiolates of HSco1 could have a role in the transfer of copper ions to CcO.

Results

Two truncated forms of the HSco1 gene have been engineered, both lacking the N-terminal mitochondrial targeting sequence and the single-transmembrane helix (Fig. 1). The two truncated forms differ by a 14-aa segment at the N-terminal side. This 14-aa segment is rich in positively and negatively charged residues and is predicted to be unstructured. Size-exclusion chromatography equipped with multiangle light scattering shows that the shorter apoHSco1 construct eluted in fractions corresponding to a monomeric state for the protein, whereas the longer construct eluted as a dimer (see Fig. 6, which is published as supporting information on the PNAS web site). This observation suggests that the residues separating the transmembrane helix from the folded domain are essential to promote dimerization. In vivo studies also have shown that this N-terminal region is crucial for yeast Sco1 function and cannot be replaced, even by its Sco2 counterpart (20).

Fig. 1.

Diagram of the protein sequence and cloned constructs of HSco1. The positions of the mitochondrial target sequence (MTS), transmembrane helix (TM), and the soluble fragment of HSco1 are orange, blue, and yellow, respectively. The N terminus of HSco1 protruding into the mitochondrial matrix is green. The soluble fragments used in the present study are named long and short truncated constructs. The truncates lack only the mitochondrial target sequence and transmembrane helix. The essential CPXXCP motif, the tobacco etch virus protease recognition site (GSFT), and the 14-aa segment (red dotted box) at the N terminus side are indicated in bold. The positively and negatively charged amino acids are represented in blue and red, respectively.

For the structural and electrospray ionization (ESI)-MS characterization, the short construct of 170 aa (corresponding to residue segment 132–301) of the HSco1 gene was used. Additionally, this construct contains an additional 4-aa segment (GSFT), corresponding to the tobacco etch virus protease recognition site at the N terminus (Fig. 1), thus producing a final construct of 174 aa whose identity was confirmed by ESI-MS analysis. The ESI-MS spectrum of the oxidized apoWT-HSco1 showed two main peaks corresponding to +9 and +10 ions, and the deconvolution of the spectrum gave the expected molecular mass of 19,741.3 Da. Incubation of oxidized HSco1 with 1 mM DTT increased molecular mass by 2.3 Da (molecular mass = 19,743.9 Da), which indicates that the disulfide bridge in HSco1 could be reduced easily by 1 mM DTT.

Reconstitution of Fully Reduced HSco1 with Metal Ions.

We are able to prepare HSco1 containing Cu(I) ions at 1:1, a ratio suitable for structural studies. Mass spectra of fully reduced HSco1 reconstituted at pH 7.5 with increasing concentrations of Cu(I) ions, are presented in Fig. 7, which is published as supporting information on the PNAS web site. Addition of one equivalent of Cu(I) ions to reduced HSco1 generates a new peak in MS spectrum, which corresponds to Cu₁HSco1 (Fig. 7). Addition of further equivalents of Cu(I) ions leads to a slight additional increase of the Cu₁HSco1 peak but does not induce metalloforms with higher metal stoichiometry (Fig. 7). We also showed that oxidized HSco1 does not form complex with Cu(II) ions in an ESI-MS experiment (Fig. 7), which indicates that such a complex, even if present in solution, is weak and dissociates during the ESI process. Accordingly, addition of Ni(II)Cl₂ to oxidized ¹⁵N HSco1 does not affect the ¹H–¹⁵N heteronuclear single quantum correlation (HSQC) spectrum. However, we have prepared the Cu(II) and Ni(II)HSco1 derivatives of reduced HSco1 in a 1:1 metal/protein ratio. The ultraviolet/visible (UV/VIS) and EPR spectra of Cu(II)HSco1 are identical to those recently reported (15). Similar UV/VIS spectra also were reported for bacterial Sco homologues (12, 14, 17) and Cu(II) nitrosocyanin (21, 22). In addition, the UV/VIS spectrum of Ni(II)HSco1 (Fig. 8, which is published as supporting information on the PNAS web site) is similar to that reported for the B. subtilis and Rhodobacter sphaeroides homologues (14, 17), with two weak bands at 380 and 540 nm, respectively, correlating with the two low-energy bands of the Cu(II)HSco1 spectrum although shifted to lower energy (Fig. 8). The intense thiolate-Ni(II) charge transfer band (23) is also red-shifted at 304 nm with respect to the Cu(II)HSco1 spectrum (Fig. 8).

The aggregation state and the conformational properties of the Cu(I) and Ni(II)HSco1 forms have been investigated by multiple techniques, including size-exclusion chromatography, ESI-MS, and NMR. In NMR, ¹⁵N relaxation rates are modulated by the correlation time for the protein tumbling (τ_m), which is directly related to the molecular weight of the protein, thus monitoring its aggregation state (24). The correlation times of Cu(I)HSco1 and Ni(II)HSco1 proteins (at millimolar concentrations) are 14.5 ± 1.1 ns and 15.6 ± 1.2, respectively, as expected for a protein of this size in a monomeric state. These values also are similar to the τ_m of apoHSco1 (13.8 ± 1.6 ns), which reorients in solution as a monomeric protein as shown by size-exclusion chromatography and multiangle light scattering experiments. ESI-MS experiments conducted at a 1.8 μM concentration of protein did not detect any higher aggregates for apoHSco1 or Cu₁HSco1. ESI-MS spectra and, especially, the charge-state distribution of ions also can yield information about the conformational states of proteins under a variety of conditions (25). Upon the binding of Cu(I), the charge state distribution of Cu₁HSco1 species shifted toward ions with lower charges (+9 and +8), which indicates that binding of metal induces a conformation change of the protein to a more compact state (Fig. 7).

Cu(I), Ni(II), and apoHSco1 Solution Structures.

The solution structures of Cu(I)HSco1 and Ni(II)HSco1 (Fig. 2) were determined by using distance and angle restraints as obtained from 2D and 3D heteronuclear NMR experiments (Table 1, which is published as supporting information on the PNAS web site). The overall fold of Cu(I)HSco1 and Ni(II)HSco1 structures is the same as that of the crystal structure of apoHSco1 (19) and contains four α-helices and nine β-strands organized into the thioredoxin fold. The solution structure of apoHSco1 also displays the same global thioredoxin fold. However, the β-hairpin present in the extended, solvent-exposed loop-8 region does not form anymore (Fig. 2).

Fig. 2.

Solution structures of human Cu(I), Ni(II), and apoSco1. (Left) The superimposition of 20 structures of Cu(I), Ni(II) and apoHSco1 are shown. α-helices and β-strands are colored in red and cyan, respectively. (Right) The average structures of the lowest energy ensemble are shown. The metal-binding residues Cys-169, Cys-173, and His-260 are shown in yellow and blue, respectively. Cu(I) and Ni(II) ions are depicted as orange and pink spheres, respectively.

Cu(I) is coordinated by the two Cys residues of the CPXXCP conserved motif, shared by the third loop and helix α1, and by the conserved His-260 (Fig. 2), located in the seventh β-strand, as shown by ²J_NH coupling-based ¹H–¹⁵N HSQC experiments (Fig. 9, which is published as supporting information on the PNAS web site). From these experiments, it appears that, in Cu(I)HSco1, His-260 acquires a preferential conformation where N^δ1 is protonated and N^ε2 is coordinated to the metal ion. The involvement in the metal binding of residues from two different regions of the protein produces a compact state of protein with respect to the apo form, in agreement with observations in ESI-MS experiments. Backbone NH resonances of three regions comprising residues 166–180, 202–204, and 244–264 are indeed not detected in the ¹H–¹⁵N HSQC spectrum of apoHSco1, although they are present in both Cu(I) and Ni(II)HSco1 spectra. These three regions comprise the CPXXCP metal-binding motif and the surrounding loops 5 and 8, the latter containing the third ligand, His-260. The inability to detect the backbone NH signals listed above is because of their fast exchange with the bulk solvent or because of the presence of multiple backbone conformations in the metal-binding area of apoHSco1, whereas the metal binding is able to “freeze” these regions in a more rigid conformation. In particular, the large conformational variability of the long loop 8 observed in the apoHSco1 solution structure (no long-range nuclear Overhauser effects are detected in loop 8) indicates that backbone structural changes are necessary to locate the metal ligand His-260 in the vicinity of the other two ligands, Cys-169 and Cys-173 (Fig. 2). Analysis of backbone dynamics (Fig. 10, which is published as supporting information on the PNAS web site) also is in agreement with the latter picture and demonstrates that both metallated forms of HSco1 do not display extensive motions on both milli- to microsecond and/or nano- to picosecond time scales, with the exception of the C and N termini.

When reduced, HSco1 binds a divalent cation, as Ni(II), and the metal is still bound by the two Cys residues of the CPXXCP motif and by His-260 through the N^ε2 atom (Fig. 2), as confirmed by ²J ¹H–¹⁵N HSQC NMR experiments. Because Ni(II) is expected to be at least four-coordinated, it is feasible that a forth ligand is completing its coordination sphere. This ligand could be a water molecule or a residue donated by the protein. The solution structure of Ni(II)HSco1 shows that two acidic groups (Asp-171 and Asp-259) could complete the Ni(II) coordination sphere, but our data do not allow discrimination between these two possibilities.

Ni(II)HSco1 Crystal Structure.

Crystals of the Ni(II)HSco1 derivative also were obtained in aerobic conditions. The overall structure of the Ni(II)HSco1 complex is essentially superimposable with that of apoHSco1 [Protein Data Bank (PDB) ID code 1WP0] (19), which was used as the model in the molecular replacement. The main exception is the solvent exposed a region involving residues 240–260 (loop 8 and the β-hairpin), ending with the metal-binding His-260. This loop indeed acquires a more ordered conformation as a consequence of the metal binding, according to the behavior in solution; this greater stability is confirmed by the good quality of the electron density map in that region for both molecules in the asymmetric unit, which is indeed better than that observed in the apoHSco1 crystal structure (19). A further confirmation is the significantly lower temperature factors of the atoms belonging to the above-mentioned loop in the structure of Ni(II)HSco1 with respect to those of the crystal structure of apoHSco1 (19).

The coordination sphere of Ni(II) in the crystal structure of Ni(II)HSco1 is quite odd and unexpected. In this structure, the two metal-binding Cys residues are oxidized and form a disulfide bond (Fig. 3; see also Fig. 11, which is published as supporting information on the PNAS web site) and therefore not capable of binding the Ni(II) ion as thiolates. Still, the metal ion remains in contact with the S—S bridge with a Ni—S distance of 2.0–2.2 Å, suggesting the formation of bonds with the available lone pairs of sulfur atoms (Fig. 3). The coordination sphere of Ni(II) is completed by His-260 (N^ε2—Ni, 2.03–2.45 Å), in agreement with the solution structure of Ni(II)HSco1, and a water molecule or more likely an anion such as Cl⁻ arranged in a distorted square planar geometry.

Fig. 3.

The distorted square planar coordination of nickel as present in the x-ray structure of oxidized Ni(II)HSco1. Bonding distances between nickel and the coordinating moieties are shown in green along with the distance between the two Cys residues. The distance between the nickel ion and the OH oxygen of Tyr-163 is also shown in magenta, and it is too large to be considered a bonding interaction.

We have been able to trace one case only in the PDB of a nickel ion coordinated to Cys residues (PDB ID code 1FRF) that are at bonding distance; in this case, the metal-binding site is dinuclear (iron and nickel) and is made of four Cys residues pointing toward two metal ions that are at a distance of 3.2 Å from each other. Cys-75 and Cys-546 are at interaction distance (2.4 Å), whereas Cys-72 and Cys-543 are farther apart (3.0 Å) (26). The distances between nickel and the four sulfur atoms are as follows: Cys-72–Ni, 2.15 Å; Cys-543–Ni, 2.11 Å; Cys-75–Ni, 1.61 Å; and Cys-546–Ni, 2.44 Å.

Discussion

The solution and crystal structures of the metal derivatives of HSco1 are completely superimposable along the entire amino acid sequence (Fig. 4) (backbone rms deviation to the new structure within 0.8 Å). These structures also are very similar to the solution structure of apoHSco1 with the exception of loop 8, which displays a different backbone conformation in apoHSco1, positioning the imidazole ring of His-260 at ≈10 Å from the sulfur atoms of the metal binding Cys residues (Fig. 4). In addition, helix α2, which encompasses the CPXXCP metal-binding site at its N terminus, is tilted in the apo solution structure with respect to the structure of the metallated HSco1 (Fig. 4). From the NMR structures, it appears that the apo form is highly disordered around the metal binding site, sampling more open conformations than in the metallated forms (Fig. 2). This observation also agrees with the results of conformation analysis by ESI-MS. In the x-ray structure of apoHSco1 (19), however, the protein is frozen in a specific conformation that does not reflect the real condition in solution. Therefore, metal binding is accompanied by a relatively large, albeit localized, effect on the protein structure, mainly involving loop 8: From an open conformation with local disorder, the structure converts into a well defined, compact state in which a metal ion is bound. In particular, the presence of the His ligand, suitably located in loop 8 to coordinate both divalent and monovalent metal ions, is important to modulate the order and disorder state of loop 8 observed in the metallated and apo forms, respectively. Also taking into account that disordered regions in protein structure often are engaged in protein–protein interactions (27), one may speculate that loop 8 modulates association–dissociation of HSco1 with its partner, the Cu(I) chaperone Cox17. For example, it is possible that, once Cu(I)Cox17 interacts transiently with apoHSco1 and donates its copper cargo to HSco1, loop 8 structurally rearranges and allows His binding and concomitant formation of the compact Cu(I)HSco1 structure, which might not exchange copper with Cox17. The formation of the stable, compact Cu(I)HSco1 state could thus constitute the important driving force of the copper transfer from Cox17 to HSco1.

Fig. 4.

Overlay of the backbone of apo (blue), Cu(I) (green), and Ni(II) (red) HSco1 solution structures and of Ni(II) (black) HSco1 x-ray structure.

Biological Context.

The debate on whether HSco1 is a metalloprotein or a thioredoxin can significantly be advanced in light of the structural results presented here. HSco1 forms 1:1 complexes with the Cu(I) and Ni(II) ions by exploiting the same metal binding ligands, which confirms that HSco1 is suitable for binding both monovalent and divalent metal ions (15). Reduced HSco1 also can bind one equivalent of Cu(II); however, the reconstituted Cu(II)HSco1 complex shows two different coordination environments with different populations (15). Similar results on Cu(II)Sco1 complexes also were obtained for the B. subtilis and yeast proteins (12, 15). It is therefore reasonable to assign a copper chaperone role to HSco1, the metal ion being coordinated by two Cys residues and one His residue. A similar metal-binding site also is found in another copper chaperone, i.e., the ATX1 from Synecocystis (28). Similarly to the latter system, the metal-donating and the metal-receiving coordination sites are different, thus overcoming the condition that “donor” and “recipient” protein partners in metal transfer processes need metal-binding sites similar in structure, as recently suggested (29). The presence of three ligands, one of each being a “flexible” His residue, also makes the site suitable for the binding of divalent metal ions. Indeed, Cu(II), Ni(II), and, presumably, Zn(II) (14) can bind at the same site of HSco1. In the case of a divalent metals, the coordination can be completed by an additional exogenous ligand, e.g., H₂O, or by a protein carboxylate. The latter hypothesis is supported by the observation that the affinity of HSco1 for the Cu(II) ion is reduced if Asp-259 is mutated (15).

The fold of HSco1 is similar to that of redox-active proteins like thioredoxins and peroxiredoxins, with the metal-binding Cys residues located at the same positions as the conserved catalytic Cys residues in thioredoxins. This feature became apparent when the first structure of a Sco1 homolog was solved (12). Therefore, a thioredoxin fold has evolved as a metal chaperone to bind the metal atom via the di-Cys motif, and one may speculate that the thioredoxin function is still maintained. Indeed, apoHSco1 can be easily oxidized to form S—S bonds. In this research, we also have shown that, in the oxidized form, the protein has low affinity for metal ions because their addition did not affect the ¹H–¹⁵N HSQC spectrum of oxidized apoHSco1, and no metal adduct has been detected by ESI-MS experiments. In the PDB, only one example of a metal ion bound to an oxidized S—S bond is reported (26). Therefore, it is feasible that the present crystal structure of the nickel derivative of oxidized HSco1 represents the transient copper-delivery complex, which might exist just before the copper is transferred to the Cu_A site of the COXII subunit. Indeed, biochemical and genetic studies on yeast Sco1 demonstrated its ability for direct interaction with the COXII subunit (30, 31). Because numerous proteins in the mitochondrial intermembrane space have disulfide bonds (32–34), the two Cys residues of the Cu_A site of COXII may also require reduction before metal transfer. The same requirement applies to the bacteria in which the COXII subunit is exposed to oxidizing periplasmatic or extracelluar environments. Therefore, it may be argued that Sco1 participates, along with metal transfer, in the reduction of the Cu_A site of COXII and its metal-transfer mechanism might include the following steps: (i) CuHSco1 interacts with the oxidized COXII and reduces the Cys residues of the Cu_A site as a thioredoxin, and (ii) the oxidized HSco1 protein transfers Cu to the reduced Cu_A site (Fig. 5). The oxidized apoHSco1 could be reduced before the next metal transfer and its reduction can be performed by another thioredoxin-like protein, such as HSco2, which is known to play a key role in Cu_A formation (35), or by cytochrome c, an electron-transfer protein that is found in the same operon of Sco1 homologs in some bacteria or is fused to some Sco1 paralog proteins (1). The copper delivery process therefore seems to involve several proteins in a complex molecular mechanism that requires further investigation. Interestingly, a dual functional role in the assembly mechanism of the copper enzyme superoxide dismutase has been similarly proposed for its copper chaperone CCS, which is, along with the copper donation, involved in the formation of an intramolecular disulfide essential for superoxide dismutase activity (36).

Fig. 5.

Proposed mechanism for copper transfer from HSco1 to the COXII subunit of CcO. This model implies that HSco1 may form a transient species characterized by an oxidized S—S moiety still able to interact with a copper ion. This state might exist just before the copper is transferred to the Cu_A site of CcO. In the metal transfer mechanism, HSco1 also can work as a thioredoxin in the reduction of the Cys residues in the Cu_A site of COXII. At this stage, the oxidization state of copper during the metal transfer cannot be assessed. The Cu_A site of the COXII subunit can indeed accept both Cu(I) and Cu(II) ions from HSco1. Broken lines indicate the metal coordination bonds.

Concluding Remarks.

We have succeeded in preparing Cu(I) and Ni(II) derivatives of HSco1 and determined their solution structures. The solution structure of apoHSco1 also has been determined. The structures confirm that the metal ions are bound by two Cys residues and one His residue, and they show the transition from a locally disordered apo protein to a compact metallated form, as confirmed by ESI-MS studies. We also obtained crystals of Ni(II)HSco1, which suggest the binding of the metal ion to the oxidized form of HSco1. This species may represent a transition state of the copper transfer from HSco1 to the Cu_A site of COXII. This transient species may present a missing link that integrates the metal transfer and thioredoxin functions already proposed for this fascinating protein. A similar mechanism of copper transfer also has been previously suggested for the bacterial Sco1 homologue PrrC from Rhodobacter sphaeroides (17).

Methods

Protein Preparation and Characterization.

Long and short soluble domains of HSco1 (lacking the first 351 and 393 bp, respectively, corresponding to residues 1–117 and 1–131) were amplified by PCR, cloned into the Gateway Entry vector pENTR/tobacco etch virus/D-TOPO (Invitrogen), and subcloned into pETG-30A (European Molecular Biology Laboratory Protein Expression and Purification Facility) by Gateway LR reaction to generate N-terminal, His-GST fused proteins. The proteins were expressed in Escherichia coli BL21-Gold(DE3) cells (Stratagene), which were grown in LB and minimal medium [(¹⁵NH₄)₂SO₄ and/or [¹³C]glucose] for the production of labeled samples. Purification was performed by using a HiTrap chelating HP column (Amersham Pharmacia Biosciences) charged with Zn(II). His-GST tag was cleaved with AcTEV, and separated from the C-terminal domain with a second purification step. After this purification, the protein preparations showed a single component by SDS/PAGE with <5% of copper bound to the protein, as checked through ESI-MS spectrometry. DTT was added to the apoprotein in a 10 mM concentration to reduce the Cys residues of the CPXXCP motif before metal reconstitution. The Cu(I), Cu(II), Ni(II) metallated forms were obtained by addition of stoichiometric amounts of the metal ions {as [Cu(I)(CH₃CN)₄]PF₆, CuSO₄, and NiCl₂} to diluted protein solutions in 50 mM phosphate buffer at pH 7.2, followed by protein concentration under nitrogen atmosphere. The metal content was finally determined by inductively coupled plasma MS.

Electronic spectra on the metal derivatives were recorded on a Cary 50 spectrophotometer (Varian). EPR spectra on Cu(II)HSco1 were recorded at 180 K on an Elexsys E500 spectrometer (Bruker) equipped with a X-band microwave bridge (microwave frequency, 9.45 GHz) and an ER 4131 VT unit for temperature control. To investigate the aggregation state of HSco1, 0.5–1 mM protein samples were run on a Superdex75 HR-10/30 size-exclusion column on an AKTA-FPLC system (Amersham Pharmacia Biosciences) connected with a multiangle light scattering (DAWN-EOS, Wyatt Technologies, Santa Barbara, CA) coupled with quasielastic light-scattering detectors.

Before ESI-MS experiments, purified apoHSco1 protein was brought into 50 mM ammonium acetate buffer (pH 7.5) by using a HiPrep26/10 desalting column (Amersham Pharmacia Biosciences). In MS experiments, 1.8 μM protein samples were infused by a syringe pump at 15 μl/min into an Ettan API ESI-TOF mass spectrometer (Amersham Pharmacia Biosciences). Mass spectra were recorded during 2–3 min at a capillary exit voltage of 150 V. apoHSco1 was reduced by addition of 0.5 or 1.0 mM DTT at 25°C. Reconstitution of apoHSco1 with copper was conducted as follows. First, Cu(II) acetate was dissolved at 150 μM concentration in argon-saturated 50 mM ammonium acetate, pH 7.5, and Cu(II) was reduced to Cu(I) by addition of 0.5 mM DTT. Different equivalents of freshly prepared Cu(I)DTT complex were added to the apoHSco1 samples (protein concentration, 1.8 μM), the mixture was incubated for 1 min at 25°C, and ESI-MS spectra were recorded as described above.

Solution Structures Determination.

NMR spectral assignment and structure determination were obtained through the experiments listed in Table 1, which also indicates the magnetic fields at which they were collected. Overall, the resonances of 95% of carbon atoms, 97% of nitrogen atoms, and 90% of protons were assigned in Cu(I)HSco1 and Ni(II)HSco1 (deposited in the Protein Data Bank database). In the case of apo form, the resonances of 85% of carbon atoms, 80% of nitrogen atoms, and 80% of proton atoms were assigned. The ¹H, ¹³C, and ¹⁵N resonance assignments of the apo, Cu(I), and Ni(II)HSco1 forms are reported, respectively, in Tables 2–4, which are published as supporting information on the PNAS web site. The His ring protons were assigned through a ¹H–¹⁵N HSQC experiment tailored to the detection of ²J ¹H–¹⁵N couplings and from the analysis of the ¹³C-NOESY-HSQC spectra. For all His residues, all of the nonexchangeable protons were assigned in Cu(I)HSco1 and Ni(II)HSco1. The exchangeable proton of the metal-binding ligand, His-260, also was detected in both metallated forms. R₁ and R₂ ¹⁵N relaxation rates and ¹H–¹⁵N nuclear Overhauser effects (with and without ¹H saturation) (Table 1) were measured at 298K on Avance 500 and 600 MHz Bruker spectrometers and then analyzed by using a standard procedure (37). After conversion of the NMR data in structural constraints [3,035, 2,776 and 2,066 meaningful proton–proton distances, together with 85 ψ and 83 φ angle constraints for Cu(I)HSco1, Ni(II)HSco1, and apoHSco1, respectively], the structures were calculated using the program dyana (38). The best 30 structures of the dyana family were then subjected to restrained energy minimization with amber 8.0 (39). The force-field parameters for the metal ions were adapted from similar systems (40, 41). The statistical analysis of the restrained energy minimization family of apoHSco1, Cu(I)HSco1, and Ni(II)HSco1 structures are reported, respectively, in Tables 5, 6, and 7, which are published as supporting information on the PNAS web site. The programs procheck and procheck-nmr (42, 43) were used in the evaluation of the quality of the structures. More than 90% of residues were located in the allowed regions of the Ramachandran plot.

Crystallization, Data Collection, and Structure Solution.

Crystals of Ni(II)HSco1 grew at 20°C from a 0.1 M Tris·HCl/30% polyethylene glycol 6000 solution at pH 8.5 by the vapor diffusion technique. The final protein concentration was ≈10 mg/ml. The data set was collected by using synchrotron radiation at beamline ID-29 (European Synchrotron Radiation Facility, Grenoble, France) at 100 K, with the crystal cryocooled, in the presence of 10–15% of ethylen glycol. The Ni(II)HSco1 crystal diffracted up to 2.5-Å resolution and belongs to space group P2₁2₁2₁ (a = 51.46 Å, b = 52.44 Å, c = 136.41 Å) with two molecules in the asymmetric unit, a solvent content of 47.1%, and a mosaicity of 0.7°. The structure was solved by using the molecular replacement technique, with the structure of the apoHSco1 (PDB ID code 1WP0) as starting model.

The Ramachandran plot of the refined model shows that 97.5% of residues are in allowed regions of the plot, 2.5% of residues are in generously allowed, and no residues are in disallowed regions. Table 8, which is published as supporting information on the PNAS web site, reports the data collection and refinement statistics.

Abbreviations

CcO

cytochrome c oxidase

ESI

electrospray ionization

HSco1

human Sco1

HSQC

heteronuclear single quantum correlation.