X-ray absorption spectroscopy of the copper chaperone HAH1 reveals a linear two-coordinate Cu(I) center capable of adduct formation with exogenous thiols and phosphines.

The human copper chaperone HAH1 transports copper to the Menkes and Wilson proteins, which are copper-translocating P-type ATPases located in the trans-Golgi apparatus and believed to provide copper for important enzymes such as ceruloplasmin, tyrosinase, and peptidylglycine monooxygenase. Although a substantial amount of structural data exist for HAH1 and its yeast and bacterial homologues, details of the copper coordination remain unclear and suggest the presence of two protein-derived cysteine ligands and a third exogenous thiol ligand. Here we report the preparation and reconstitution of HAH1 with Cu(I) using a protocol that minimizes the use of thiol reagents believed to be the source of the third ligand. We show by x-ray absorption spectroscopy that this reconstitution protocol generates an occupied Cu(I) binding site with linear biscysteinate coordination geometry, as evidenced by (i) an intense edge absorption centered at 8982.5 eV, with energy and intensity identical to the rigorously linear two-coordinate model complex bis-2,3,5,6-tetramethylbenzene thiolate Cu(I) and (ii) an EXAFS spectrum that could be fit to two Cu-S interactions at 2.16 A, a distance typical of digonal Cu(I) coordination. Binding of exogenous ligands (GSH, dithiothreitol, and tris-(2-carboxyethyl)-phosphine) to the Cu(I) was investigated. When GSH or dithiothreitol was added to the chaperone during the reconstitution procedure, the resulting Cu(I)- HAH1 remained two-coordinate, whereas the addition of the phosphine during reconstitution elicited a three-coordinate species. When the exogenous ligands were titrated into the Cu(I)-HAH1, all formed three-coordinate adducts but with differing affinities. Thus, GSH and dithiothreitol showed weaker binding, with estimated KD values in the range 10-25 mm, whereas tris-(2-carboxyethyl)-phosphine showed stronger affinity, with a KD value of <5 mm. The implications of these findings for mechanisms of copper transport are discussed.

Copper is an essential metal for all organisms. Important mammalian enzymes such as cytochrome c oxidase (respiration), Cu,Zn-superoxide dismutase (antioxidation), lysyl oxidase (collagen cross-linking), peptidylglycine monooxygenase (activation of neuropeptides), dopamine-␤-monooxygenase (catecholamine biosynthesis), and tyrosinase (pigmentation) all contain copper. Plants and bacteria utilize copper in photosynthetic and denitrification pathways, and new roles for copper in enzyme catalysis are being discovered (1). However, the oxidation potential of the Cu(I)/Cu(II) pair in an uncomplexed aqueous environment is sufficient to induce the formation of damaging radicals, and consequently free cellular copper concentrations are maintained at extremely low levels (2). Copper homeostasis is therefore highly regulated within the cell by both transcriptional control and selective transport mechanisms.
Copper transport to membrane-bound proteins such as copper-transporting ATPases or delivery to copper-dependent enzymes is carried out by small cytosolic proteins called chaperones that are specific to each protein. HAH1 (3,4) is the chaperone (5) for Menkes (MNK) 1 and Wilson's disease proteins (WND) (6 -11). These copper-transporting ATPases are located in the trans-Golgi apparatus and are believed to provide copper for ceruloplasmin (12), peptidylglycine monooxygenase (13), and tyrosinase (14). They share ϳ50% sequence homology and differ mainly in their location in the body. Whereas WND is predominantly found in the liver and brain tissues, MNK is found in all other tissues but liver and brain. HAH1's counterpart in yeast, Atx1 (15,16), delivers copper to Ccc2, the yeast homologue for MNK and WND (17). COX17, COX11, and SCO1 are involved in copper shuttling to the mitochondrial membrane and incorporation of copper into cytochrome c oxidase (18 -21). Copper chaperone for superoxide dismutase shuttles copper to Cu,Zn-superoxide dismutase (22)(23)(24)(25)(26)(27).
Solution structures for several of the copper-bound and apo forms of chaperones as well as isolated metal-binding repeats of some ATPases have been published (28 -36). All structures adopt a ␤␣␤␤␣␤ fold. The conserved metal-binding motif, MX-CXXC, which is anticipated to bind Cu(I), is located in the connecting loop between the first ␤-sheet and the first ␣-helix.
of the bound Cu(I), more indicative of a three-coordinate structure (28,29), with S-Cu-S angles of 120 Ϯ 40°and 115 Ϯ 26°, respectively. Comparison with the NMR-derived apo structures (28,29) indicated that one of the ligated cysteines (the one closest to the N terminus) occupies a different conformation in the apo proteins, where it points away from the metal-binding pocket toward solvent. This residue appears to flip inwards as the result of Cu(I) binding becoming much less conformationally mobile in the process. Similar conclusions have been obtained from the NMR structures of apo and Cu(I)-loaded forms of the chaperone ATPase partners S. cerevisiae Ccc2a (35) and B. subtilis CopAb (36) (where "a" and "b" denote the first and second subdomains, respectively). These proteins likewise show a large conformational change in the position of the N-terminal cysteine residue as it flips from a solvent-exposed conformationally mobile location in the apo form to its Cu(I)coordinated position. Again, the S-Cu-S angles are estimated to be 115°and 132°, respectively, much less than the 180°value expected for linear two-coordination. Three-coordination involving an exogenous thiol has been proposed as the origin of the third ligand, since all potential protein-derived third ligands are at too great a distance to influence the Cu(I) coordination directly.
NMR spectroscopy does not directly observe the bound copper ion or the putative exogenous ligand (37). Assumptions about the copper coordination are made on the basis of the known van der Waals radius of the Cu(I) ion and the distance of the coordinating atoms (i.e. two cysteine-sulfurs in these chaperones). X-ray absorption spectroscopy (XAS), on the other hand, is unique in its ability to define the local coordinate structure of the otherwise spectroscopically silent Cu(I) ion in proteins and reveals detailed information about the immediate coordination environment within a radius of Ͻ5 Å. Metalligand distances can be determined to high accuracies (0.02 Å), whereas information on coordination number and oxidation state of the metal can also be determined. XAS spectroscopy at the copper K-edge has been used to confirm the presence of two-coordinate Cu(I) in MNK (38) and WND (39), where the Cu(I) center is ligated by the two cysteine residues of the MXCXXC motif. However, XAS studies by Penner-Hahn and co-workers (40) on the copper-bound form of Atx1 in S. cerevisiae found a three-coordinate Cu(I) center bound by two "short" (2.24-Å) and one "long" (2.4-Å) Cu-S interaction. As with the NMR studies, it was suggested that the third ligand was derived from the reductant, dithiothreitol (DTT), that had been added to the protein prior to metal reconstitution. George and co-workers (41) have performed EXAFS spectroscopy on Enterococcus hirae CopZ and the full-length and second metalbinding motif of MNK (42). Again, rather than the expected digonal coordination, mixtures of two-and three-coordinate species were found. Most recently, an XAS study of B. subtilis Cu(I)-CopZ by Banci et al. (37) has reported a three-coordinate environment (three Cu-S at 2.25 Å) in both DTT-and dithionite-reduced samples and in the presence of carboxylate anions as potential exogenous ligands. Ascorbate reduction, on the other hand, was found to elicit a perturbed spectrum with evidence for a low Z scatterer (Cu-O at 1.95 Å) in addition to the two S scatterers (two Cu-S at 2.25 Å).
These studies pose two important questions. First, can the two-coordinate Cu(I)-bound chaperone be prepared free from ligation by an exogenous third ligand, and if so, what is the structure? Second, what is the role (if any) of binding of a third ligand in chaperone function? In the present study, we report the results of biochemical and XAS studies of the HAH1 metallochaperone, which for the first time reveal the presence of a linear two-coordinate Cu(I) center, free from ligation by any exogenous ligand. We have also investigated the affinity of this two-coordinate site for the exogenous thiol ligands, GSH and DTT, as well as for the phosphine ligand tris-(2-carboxyethyl)phosphine (TCEP), which is often used as a reductant for disulfides in proteins. Our results establish that exogenous ligands can bind to the two-coordinate Cu(I) center but with rather low affinity.

EXPERIMENTAL PROCEDURES
Cloning and Isolation of HAH1-The clone for HAH1 was obtained from the expressed sequence tag data bank (accession number AA703181). The cDNA was amplified using PCR and oligonucleotides designed to introduce a NdeI restriction site at the 5Ј-end and an EcoRI restriction site at the 3Ј-end. The primers used were as follows: 5Ј-AGC GCT CAC ATA TGC CGA AGC ACG AG-3Ј (HAH1 forward primer) and 5Ј-AGC TCG AAT TCT ACT CAA GGC CAA GGT A-3Ј (HAH1 reverse primer). Cloning was performed using the TOPA TA cloning kit (Invitrogen) and Taq polymerase. The PCR product was digested with NdeI and EcoRI endonucleases and cloned into the pET24a(ϩ) vector (Novagen). The resulting plasmid was transformed into the Escherichia coli strain BL21 (DE5). In a typical expression experiment, cultures were grown in 1 liter of Luria-Bertani medium at 37°C to a final A 600 between 0.5 and 0.9. The cells were harvested by centrifugation in a Sorvall GS-3 rotor for 20 min at 15,000 ϫ g; resuspended in buffer containing 50 mM Tris-HCl, 100 mM NaCl, and 1 mM EDTA at pH 8.0; and subsequently lysed in a French press (1000 p.s.i.). To minimize protease activity, one tablet containing a protease inhibitor mixture (Complete; Merck) was added per 50 ml of lysis buffer. The soluble portion was concentrated in a Centriprep (Amicon; molecular mass cut-off, 3000 Da) to ϳ12 ml and loaded on an Amersham Biosciences HiLoad 26/60 Superdex column. Eluted fractions were assayed by SDS-PAGE on an Amersham Biosciences PHAST system (20% homogeneous gel). Fractions that contained HAH1 were pooled and concentrated in a Centriprep concentrator (molecular mass cut-off, 3000 Da) to ϳ7 ml. The concentrate was then ultrafiltered in a Centriprep (molecular mass cut-off, 50,000 Da). The ultrafiltrate was checked for purity on SDS-PAGE. No impurities were detected. The protein concentration was determined by the Bradford assay; typical yields were 15-20 mg of pure HAH1/liter of culture. The Bradford assay was calibrated by comparing a sample analyzed for its amino acid content with the result obtained from Bradford analysis. The protein concentration by Bradford assay was 1.27 mg/ml compared with 1.57 mg/ml determined by amino acid analysis, and a correction factor of 1.24 was applied to all Bradford protein assays.
Protein masses were determined by electrospray ionization mass spectrometry. Proteins were injected onto a 1.0 ϫ 250-mm C4 column (214 MS C4; Vydac), and masses were determined on-line using a model LCQ ion trap (ThermoFinnigan). The flow rate was 20 ml/min with a linear gradient of 18 -50% acetonitrile over 40 min in a mobile phase containing 0.2% acetic acid and 0.005% heptafluorobutyric acid. Samples were autoinjected and concentrated/purified using a microprotein trap cartridge (Michrom Bioresources, Inc.). Mass spectra of proteins eluted from the C4 column were deconvoluted using BioWorks Browser software (ThermoFinnigan). Mass accuracy of better than 0.01% was confirmed using horse myoglobin.
Metal Reconstitution-All procedures described below were performed in an inert atmosphere in an anaerobic chamber to prevent oxidation of the cysteine residues. Copper was added as the tetraacetonitrile Cu(I) complex to minimize disproportionation of free Cu(I) to Cu(II) and metallic copper. DTT (Sigma) was used to reduce the cysteine residues in HAH1 prior to metal reconstitution. DTT was added as a 10-fold excess to the protein, and the mixture was incubated for 10 min on ice. The protein was then dialyzed overnight under argon into a buffer containing 50 mM HEPES (Sigma) and 10% acetonitrile (Sigma) at pH 7.5. Copper was added in a 1:1 ratio by slow infusion of a 100-fold stock solution of [(CH 3 CN) 4 Cu(I)]PF 6 over a 1-h period utilizing a syringe pump at a rate of 1 l/min. Successive dialyses against HEPES plus 10% acetonitrile, HEPES plus 5% acetonitrile, and HEPES without acetonitrile for 12 h per buffer in a Slide-a-lyzer cassette (Pierce) yielded reconstituted Cu-HAH1. In all samples, copper/ protein ratios were between 0.4 and 1.0 as determined by flame atomic absorption spectroscopy (Varian AA-5). The copper content of the sample was monitored throughout the various dialysis steps. In a typical experiment, 8 ml of ϳ100 M protein was reconstituted to a final copper concentration of ϳ50 -80 M. The sample was sealed and stored at Ϫ80°C. For XAS experiments, the protein was concentrated using an Ultrafree centrifugation system (Millipore Corp.; molecular mass cut-off, 3500 Da) modified to maintain an anaerobic atmosphere. In this protocol, a concentrator was filled with dilute sample and sealed with a rubber septum in the anaerobic chamber and then transferred to a tabletop centrifuge for concentration outside the chamber. The oxidation state of copper in the final sample was checked by EPR (Bruker Elexsys). The Cu(I)/Cu(II) ratio was determined by comparison of the integral of a cavity/base line-corrected EPR spectrum of Cu(I)-HAH1 to a 100 M Cu(II) EDTA sample. Concentrations of Cu(II) never exceeded 10% and were typically around 2-4%.
Titrations with Exogenous Ligands-Binding of three commonly used reductants (GSH (Sigma), DTT (Sigma), and TCEP (Sigma)) to Cu-HAH1 was probed in two sets of experiments. All three substances provide a potential ligand to Cu-HAH1, either a sulfhydryl group (DTT and GSH) or a phosphine group (TCEP). In the first set of experiments, GSH, DTT, and TCEP were added in a 10-fold excess (typically 1000 -1500 M) to the apoprotein and incubated for 1 h before adding copper. The sample was then reconstituted with copper and dialyzed as described above.
In the second set of experiments, the three ligands were each titrated into the protein after metal reconstitution, with ratios of 1:1, 5:1, and 25:1 reductant/protein. After the addition of the ligand, the samples were concentrated to ϳ100 M [Cu] for XAS measurements. For a complete list and description of all XAS samples, see Table I.
XAS Data Collection and Analysis-XAS data were collected in March, May, and December of 2002 on beam line 9-3 at the Stanford Synchrotron Radiation Laboratory. Beam line 9-3 was operated at 3.0 GeV with beam currents between 50 and 100 mA. The samples were measured in fluorescence mode using a 30-element array detector (Canberra). A rhodium-coated mirror positioned upstream of the Si220 monochromator was used to cut off all energy above 12 keV. A second mirror located downstream from the monochromator was used to focus the beam (5 ϫ 1 mm). Soller slits with a 6-nickel filter were placed in front of the detector to selectively attenuate the elastic scatter peak with respect to the copper K ␣ fluorescence. Under these conditions, no dead time corrections were necessary. The samples were measured at 10 -15 K using a liquid helium flow cryostat. Data for each channel were inspected for background noise and glitches. The nickel K ␤ fluorescence, which is generated by the nickel filter and is always present in the single channel analyzer window, was removed by collecting data on a blank sample under identical conditions of detector/Soller slit geometry and subtracting this as a background file from each data set. The energy was calibrated by assigning the first inflection point of the copper edge from a copper metal foil that was inserted between the second and third ionization chamber to 8980.3 eV. The number of scans collected for each sample varied from 2 to 15, depending on the copper concentration of the samples. The scans were collected to k ϭ 12.8 Å Ϫ1 because of interference by traces of zinc in the sample.
Data reduction and analysis were performed using the EXAFSPAK computer suite (43). Theoretical phase and amplitude functions were calculated using FEFF 8.0 (44). The inspected raw data were averaged, background-subtracted, and normalized. The energy corresponding to the start of the EXAFS (k ϭ 0) was set at 9000 eV. A k 4 -weighed spline fit was used to extract the EXAFS oscillations from the raw data. The EXAFS data were simulated by curve-fitting in the OPT module of EXAFSPAK using a nonlinear Marquadt algorithm. The following parameters were refined: ⌬E 0 (a small energy correction at k ϭ 0), R i (the distance between the central absorber and atom i), and 2 (the Debye-Waller factor, defining the mean square deviation of R i ). A goodness of fit (F) parameter was used to evaluate the merit of the fit. F is defined as follows.

RESULTS
Recombinant HAH1 was overexpressed in E. coli, and purified to homogeneity with high yield. Mass spectrometry of the purified protein revealed two major peaks at 7399.0 Ϯ 3 and 7269.0 Ϯ 3 Da (Fig. 1a), and the protein was homogeneous on SDS-PAGE (Fig. 1b). The peak at 7399.0 Da in the mass spectrum corresponds to the full-length HAH1 (m/z calculated from FIG. 1. a, electrospray ionization mass spectrum of HAH1. The peak at 7399.0 Da in the mass spectrum corresponds to the full-length HAH1 (m/z ϭ 7401.63), whereas the peak at 7269.0 Da corresponds to the full-length protein lacking its NH 2 -terminal methionine (m/z ϭ 7370.63). Data were collected after separation on a 1.0 ϫ 250-mm C4 column, and masses were determined on-line using a model LCQ iontrap. The flow rate was 20 ml/min with a linear gradient of 18 -50% acetonitrile over 40 min in a mobile phase containing 0.2% acetic acid and 0.005% heptafluorobutyric acid. b, purification of HAH1 on SDS-PAGE (20%, homogeneous). Lanes from the left correspond to (i) protein standards, (ii) before induction, (iii) after induction, (iv) soluble fraction, and (v) after purification. c, structure of Hg-Atx1 showing the mercury atom linearly coordinated to two cysteine residues, Cys 15 (exposed to solvent) and Cys 18 (part of helix 1). The coordinates were taken from Protein Data Bank entry 1CC8, and the structure was displayed by Weblab Viewerpro 3.5 (Molecular Simulations). sequence ϭ 7401.63), whereas the peak at 7269.0 Da corresponds to the full-length protein lacking its NH 2 -terminal methionine (m/z calculated ϭ 7370.63).
Metal reconstitution of HAH1 with Cu(I) poses two problems. The metal binding cysteines in the CXXC motif of HAH1 are exposed to solvent and therefore prone to oxidation and formation of a disulfide bridge (28,29,45) (Fig. 1c). To prevent oxidation within the unoccupied metal binding site, experiments have to either be performed anaerobically or in the presence of a reductant. In addition, Cu(I) as the free ion is unstable in aqueous solution and is sensitive to oxidation or disproportionation. Because of the high hydration energy of Cu 2ϩ versus Cu ϩ (Cu ϩ ϭ 0.139, Cu 2ϩ ϭ 0.522 kcal/mol), the latter will disproportionate in water into Cu 2ϩ and Cu 0 . However, complexes like [Cu(I)CN 4 ] 3Ϫ or [Cu(I)(CH 3 CN) 4 ] ϩ stabilize Cu(I), whereas maintaining dilute Cu(I) concentrations will shift the disproportionation equilibrium toward the Cu(I) species.
To address both problems, we performed all experiments in an anaerobic chamber and used [Cu(I)(CH 3 CN) 4 ]PF 6 with 10% acetonitrile (the complexing ligand) present in all buffers during reconstitution. The copper was added through a syringe pump over a period of 1 h. During the addition, the sample was gently agitated to prevent the accumulation of high local Cu(I) concentrations. After the addition, the amount of acetonitrile in the buffer was reduced stepwise by 5% to 0% by dialysis. Storage of concentrated samples at Ϫ80°C instead of Ϫ22°C turned out to be crucial to the integrity of the samples. No precipitation or oxidation was observed after storage times of over 2 months. Earlier experiments had shown that even after insertion of copper into the protein, exposure to oxygen leads to at least partial oxidation and loss of Cu(I) from its binding site. EPR spectra of samples that were dialyzed and concentrated under aerobic conditions showed that after 48 h, ϳ25-30% of the copper had been oxidized to Cu(II) (data not shown). Table I lists the identity of the samples investigated by XAS, their copper concentrations, and their copper/protein ratio. Three series of experiments were carried out: (i) HAH1 was reduced with DTT and dialyzed to remove the thiol prior to addition of copper; (ii) a 10-fold excess of reductant (GSH, DTT, and TCEP) was added immediately prior to the addition of copper, followed by successive dialysis steps to remove acetonitrile and unbound metal; and (iii) reductant (GSH, DTT, and TCEP) was titrated into the copper-reconstituted sample. The second and third experiments differ in that excess thiol is removed by dialysis in the second, whereas it remains available to complex the protein-bound copper in the third. The EXAFS spectra for all samples were dominated by a strong Cu-S backscattering. For Cu-HAH1 in the absence of any reductant, two samples from two independent preparations were analyzed to ensure reproducibility of the results (HAH1-1 and HAH1-2); parameters used to fit the data are shown in Table II. The EXAFS data exhibited one main frequency (Fig. 2, inset), which was also reflected in the Fourier transform (FT) with one main peak centered at ϳ2 Å and some minor peaks between 3 and 6 Å (Fig. 2). An excellent fit could be obtained with two Cu-S at 2.16 Ϯ 1 Å, a distance typical of linear S-Cu-S coordination (46,47). No significant improvements to the fit were achieved by including a second shell in the model.
More information about the copper coordination in a complex can be extracted from the x-ray absorption near edge structure (XANES) region, which spans the range 10 eV below to 10 eV above the absorption edge, and provides information about electron transitions and the oxidation state of the absorbing atom. XANES spectra of Cu(I) complexes show a pre-edge feature at ϳ8983 eV, which is due to a 1s 3 4p x,y electronic transition, the shape and intensity of which depends on the local copper coordination geometry (46,48,49). The transition shows the highest intensity and resolution for linear two-coordinate complexes and becomes less intense and broadened as the coordination increases and/or the site symmetry is lowered. Thus, it can be used as an indicator of two-versus threecoordination. As expected for a linear coordinated Cu(I) ion, the XANES spectrum for HAH1 showed an intense pre-edge feature (Fig. 3). The intensity of the pre-edge feature of Cu(I)-HAH1 was almost identical to that of the linear two-coordinate model compound bis-2,3,5,6-tetramethylbenzenethiolate Cu(I) (47), suggesting a linear S-Cu-S bond angle close to 180°in the Cu(I) chaperone. We conclude that the copper in HAH1 is in a digonal bis-cysteinate environment. In contrast to the NMR studies and the earlier XAS studies on Atx1, the data showed no indication of the presence of a third ligand.
The previous proposal, that commonly used reductants such as DTT or GSH could act as a potential source of additional ligation, led us to the experiments in which we reconstituted HAH1 in the presence of a 10-fold excess of either DTT (HAH1-DTT), GSH (HAH1-GSH), or TCEP (HAH1-TCEP) ( Table I). Fig. 4, a and b, shows a comparison of the FTs and the EXAFS for the three reductants, respectively. All three samples were comparable in copper/protein ratios (0.5-0.6) and in total copper concentrations (250 -400 M). The EXAFS data of HAH1-DTT and HAH1-GSH showed the same digonal, almost linear S-Cu-S environment. Cu-S distances for both samples were refined to 2.16 Ϯ 1 Å (HAH1-GSH) and 2.17 Ϯ 1 Å (HAH1-DTT), respectively, with an F parameter of 0.30 (HAH1-GSH) and 0.29 (HAH1-DTT). No indication of a third ligand was found. Accordingly, the FTs of HAH1-GSH and HAH1-DTT looked almost identical to HAH1-1 and HAH1-2, with the former spectra exhibiting a better signal/noise ratio. Comparison of the intensity of the 8983 eV edge feature for the samples is shown in Fig. 4c. It is obvious that adding a 10-fold excess of DTT or GSH to HAH1 prior to metal reconstitution has no effect on the copper coordination in the protein. On the other hand, the 8983-eV edge feature for HAH1-TCEP is much less intense and slightly shifted toward lower energy and is characteristic of a three-coordinate Cu(I) complex (46). As in HAH1-2, the EXAFS data of HAH1-TCEP is dominated by strong Cu-S/P backscattering, but the intensity of the oscilla-tions and of the main peak in the FT is greater than for the former. The EXAFS data were best fit with three Cu-S at 2.24 Ϯ 1 Å. We assume that the actual coordination is two Cu-S plus one Cu-P as the third ligand and is most likely provided by a phosphine group from TCEP. EXAFS spectroscopy cannot distinguish between elements that are next to each other in the periodic table.
We conclude that metal reconstitution of HAH1 in the presence of a 10-fold excess of DTT or GSH to prevent local oxidation of the cysteinates has no impact on the copper coordination. However, use of TCEP as reductant results in binding of the phosphine as an exogenous ligand to Cu(I). This implies a lower affinity of Cu(I)-HAH1 for SH groups of GSH and DTT than for the PR 3 groups of TCEP.
To further probe the binding of a third ligand, we titrated reconstituted Cu(I)-HAH1 with specific ratios of DTT, GSH, and TCEP (1:1, 5:1, and 25:1, ligand to copper, respectively). The results of the best fits to the EXAFS are listed in Table II. Adding equimolar amounts of DTT or GSH to Cu-HAH1 has no impact on the resulting copper coordination (Fig. 5, a and b). Increasing the amount of reductants to 5-fold does not change the coordination (data not shown). However, the spectra for a 25-fold excess of thiol ligand show an altered copper coordination. In HAH1-DTT25, the Cu-S distance is elongated to 2.20 Ϯ 1 Å with a coordination number of 2. For HAH1-GSH25, the , phase-corrected) and EXAFS (inset) for HAH1 reconstituted with Cu(I) using the Cu(I)-acetonitrile protocol with no added ligand. The reconstitution protocol was as follows. DTT was added as a 10-fold excess to the apoprotein, and the mixture was incubated for 10 min on ice to reduce any oxidized disulfide at the metal binding site. The reduced apoprotein was then dialyzed overnight under argon into a buffer containing 50 mM HEPES and 10% acetonitrile at pH 7.5. Copper was added in a 1:1 ratio as the [(CH 3 CN) 4 Cu(I)]PF 6 salt dissolved in the same buffer over a 1-h period, utilizing a syringe pump at a rate of 1 l/min. The sample was dialyzed successively against HEPES plus 10% acetonitrile, HEPES plus 5% acetonitrile, and HEPES without acetonitrile for 12 h per buffer change and then concentrated anaerobically. The data correspond to sample HAH1-2 (Tables I and II). Thick lines, experimental data; thin lines, simulated data. The fit corresponds to two Cu-S at 2.16 Å with no contribution from any third ligand. data best fit 2.5 Cu-S at 2.22 Ϯ 1 Å. Fig. 6, a and b, summarizes the results for the TCEP titration of Cu-HAH1. When added as equimolar amount to Cu-HAH1, a linear Cu-S coordination was observed with two Cu-S at 2.14 Ϯ 1 Å. However, adding a 5-fold excess of TCEP to Cu-HAH1 shows an already completed conversion to three-coordination, which confirms the higher affinity of Cu-HAH1 for phosphine over thiol. Fitting the EXAFS for HAH1-TCEP5 and HAH1-TCEP25 gave identical results of three Cu-S at 2.24 Ϯ 1 Å. The changes in coordination with higher ligand/copper ratios for this experimental series are also reflected in a lowered intensity of the 8983-eV edge feature, which is depicted in Fig. 7.
Coordination numbers in EXAFS spectroscopy are not well defined because of their high correlation with the Debye-Waller factor. Thus, errors of Ϯ30% are not uncommon. However, distances derived from EXAFS data are defined to high accuracy (Ϯ0.01 Å or better) and can be used to quantitatively determine the ratios of two-and three-coordinate Cu-S species within a sample (46). Using the equation of Pickering et al. (46) with pure two-and three-coordinate Cu-S distances of 2.16 and 2.25 Å, respectively, the fractions of two-and three-coordinate copper species in our samples can be estimated. For DTT, equimolar and 5-fold excess of ligand show no increase in Cu-S above the two-coordinate value of 2.16 Å, whereas a 25-fold excess gives an average Cu-S distance of 2.20 Å, corresponding to 65 Ϯ 10% of the copper remaining in a two-coordinate site and 35 Ϯ 15% of the copper in a three-coordinate site. For GSH, we calculate trigonal fractions of 25 Ϯ 10, 16 Ϯ 9, and 43 Ϯ 12%, for 1:1, 5:1, and 25:1 ratios, respectively. For TCEP, the trigonal fractions are 0 Ϯ 8, 84 Ϯ 16, and 84 Ϯ 16%, respectively. The error of these ratios is calculated, considering an inaccuracy of the refined distance of 0.01 Å. It should be noted that the calculated ratios for TCEP are likely to be less precise, because the copper coordination consists of two S and one P ligand, which alters the base distances used in our previous calculation.

DISCUSSION
Previous studies of copper coordination in HAH1 and its homologues reported the presence of at least three ligands to the Cu(I) center. NMR structures indicated S-Cu-S angles close to 120°, suggestive of binding of an exogenous ligand, whereas XAS data were interpreted by two short and one long Cu-S distance (Atx1) (40), a mixture of two and three-coordinate Cu-S (E. hirae CopZ) (41), or three-coordinate Cu-S (B. subtilis CopZ) (37). It was suggested that the third ligand was derived from the reductant (DTT) that had been added to the protein prior to metal reconstitution. These previous studies posed the following questions. (i) Can the two-coordinate  Table I. a, Fourier transforms (k ϭ 2.2-12.5 Å Ϫ1 , phase-corrected) of (bottom to top) unligated HAH1 (HAH1-2; Tables I and II) and HAH1 reconstituted in the presence of GSH, DTT, and TCEP as described above. b, experimental (thick lines) and simulated (thin lines) EXAFS corresponding to FTs in a using the parameters listed in Table II. c, absorption edges for unligated HAH1 and HAH1 reconstituted in the presence of GSH, DTT, and TCEP. The absorption edge of bis-2,3,5,6tetramethylbenzenethiolate Cu(I) is included for reference. The solid vertical line corresponds to 8982.5 eV.  Tables I and II. a (bottom to top), Fourier transforms (k ϭ 2.2Ϫ12.5 Å Ϫ1 , phase-corrected) for unligated HAH1 (HAH1-2), the adduct with GSH (HAH1-GSH) at 1:1 and 25:1 ratios, and the adduct with DTT (HAH1-DTT) at 1:1 and 25:1 ratios. b, experimental (thick lines) and simulated (thin lines) EXAFS corresponding to FTs in a using the parameters listed in Table II.  Table II. Cu(I)-bound chaperone ever be prepared free from ligation by an exogenous third ligand, and if so, what is the structure? (ii) What is the role (if any) of binding of a third ligand in chaperone function? In the present study, we used biochemical and XAS studies to show that a linear two-coordinate Cu(I) center free from ligation by any exogenous ligand can be prepared in HAH1. We also investigated the affinity of this two-coordinate site for the exogenous thiol ligands, GSH and DTT, as well as for the phosphine ligand TCEP and were able to establish that exogenous ligands do indeed bind to the two-coordinate Cu(I) center but with rather low affinity.
The method of reconstitution was critical to the preparation of the unligated two-coordinated Cu(I)-chaperone species. Although we used DTT and/or GSH as a reductant of disulfides prior to copper reconstitution, we were successful in preparing Cu(I)-chaperone complexes free from exogenous third ligand coordination. The key to success was to ensure the presence of a high concentration of a weakly coordinating ligand, which can compete effectively with thiols during the reconstitution process. Thus, use of [(CH 3 )CN) 4 Cu] ϩ as the source of Cu(I), with a large excess of acetonitrile (10%) maintained in the buffer during initial copper loading, resulted in good levels of reconstitution (0.5-1.0 molar equivalents). It is likely that the excess acetonitrile served to completely sequester the Cu(I), rendering it soluble and stable to disproportionation during subsequent dialysis. It is also possible that thiols were more effectively removed during dialysis as mixed Cu(I)-acetonitrile-thiol complexes. Whatever the mechanism, XAS measurements showed that the Cu(I) centers are linear two-coordinate with Cu-S bond lengths of 2.16 Å, a value typical of digonal copper (38,46,47). Additionally, the x-ray absorption edge exhibited an intense feature at 8983 eV of the same magnitude as that of the rigorously linear two-coordinate bis-2,3,5,6-tetramethylbenzenethiolate Cu(I) complex, which has Cu-S distances of 2.137 Å and a S-Cu-S angle of 178.6° (47). Taken together, these parameters unequivocally establish the linear two-coordinate geometry of the unligated form of the Cu(I)-HAH1 chaperone.
With a genuine two-coordinate Cu(I)-chaperone species in hand, we next investigated whether exogenous ligands could bind to generate the kind of three-coordinate species postulated from NMR. The addition of the thiol-containing ligands GSH and DTT had little or no effect on the XAS spectra at 1:1 ratios, but increasing concentrations of the thiol led to a decrease in intensity of the 8983 eV peak and a gradual increase in the average Cu-S bond length (Table II) Titrations with the phosphine TCEP showed evidence of a higher affinity for third ligand binding. Although a 1:1 molar ratio of phosphine to copper had little effect on the spectroscopy, the titration appeared complete at a 5:1 molar ratio (5 mM TCEP), implying a K D between 1 and 5 mM. Furthermore, samples reconstituted in the presence of phosphine showed three-coordination even after exhaustive dialysis. TCEP thus appears to bind more strongly to the Cu(I) centers in chaperones to produce a complex that is resistant to removal by dialysis and hence must be rather inert kinetically.
It is unclear whether the estimated magnitude of K D for thiol ligation to HAH1 is small enough to explain the presence of a third ligand in the NMR and EXAFS studies from other laboratories. Cobine et al. (41) rigorously excluded thiols from the preparation of their sample for XAS yet still observed a threecoordinate site. The authors suggest that the chloride ion, which has phase and amplitude parameters similar to sulfur, might be the origin of the third ligand. The NMR studies of Banci and co-workers (28,29) on CopZ and Atx1 were prepared by methods similar to the initial EXAFS measurements of Pufahl et al. (40) on Atx1, utilizing a 20-fold excess of DTT for reduction but following this by an ultrafiltration or dialysis step. This experiment was similar to that carried out in the present study, which yielded a two-coordinate Cu(I) site with no bound third ligand.
Another potential source of a third ligand is dimerization. A crystal structure of HAH1 has shown the feasibility of dimerization mediated by metal binding, with Cu(I) strongly coordinated by two cysteines of one monomer and one cysteine of the other. The fourth cysteine appears weakly bound in pseudotetrahedral geometry (31). Kihlken et al. (50) have recently reported that Cu(I) binding induces dimerization of B. subtilis CopZ but only in the absence of DTT. Presumably, binding of DTT to the bis-cysteinate Cu(I) center prevents dimerization (28,29). The recent XAS and NMR studies of the B. subtilis CopZ suggest that nonspecific protein aggregation may occur at high protein concentrations (37). If dimerization or aggregation is a contributing factor to the observation by other laboratories of three-coordinate structures, it is possible that the conditions used to reconstitute HAH1 in this study favored the monomeric state. Notwithstanding these uncertainties, we have firmly established the existence of the two-coordinate form of the HAH1 chaperone and its ability to form three-coordinate ad- ducts with thiol ligands, albeit with low affinity. These results provide evidence in support of proposed mechanisms (31,34,36) for copper transfer in which the solvent-exposed unligated N-terminal cysteine of the apo-partner flips inward to interact with the two-coordinate Cu(I) center of the chaperone and form a three-coordinate intermediate within a chaperone-partner heterodimer. The relatively low affinity of the exogenous thiol interaction with the two-coordinate Cu(I) center suggests that it would be insufficient to drive the heterodimer formation on its own so that electrostatic complementarity (predicted from NMR structures) between chaperone and ATPase domain surfaces would provide additional driving force for metal transfer (34,36). The interplay between these two factors most likely contributes to the selectivity of the chaperone for individual subdomains of the ATPase partner (5) and to the observed reversibility of copper transfer from ATPase to chaperone, which has been proposed as an important regulatory mechanism of copper homeostasis (51). With respect to the latter, we note that the reverse mechanism would require the two-coordinate center on the ATPase (38,39) to also be able to form adducts with exogenous thiol ligands to generate the common intermediate. Studies aimed at establishing this chemistry are under way.