High Affinity Binding between Copper and Full-length Prion Protein Identified by Two Different Techniques*

The cellular prion protein is known to be a copper-binding protein. Despite the wide range of studies on the copper binding of PrP, there have been no studies to determine the affinity of the protein on both full-length prion protein and under physiological conditions. We have used two techniques, isothermal titration calorimetry and competitive metal capture analysis, to determine the affinity of copper for wild type mouse PrP and a series of mutants. High affinity copper binding by wild type PrP has been confirmed by the independent techniques indicating the presence of specific tight copper binding sites up to femtomolar affinity. Altogether, four high affinity binding sites of between femto- and nanomolar affinities are located within the octameric repeat region of the protein at physiological pH. A fifth copper binding site of lower affinity than those of the octameric repeat region has been detected in full-length protein. Binding to this site is modulated by the histidine at residue 111. Removal of the octameric repeats leads to the enhancement of affinity of this fifth site and a second binding site outside of the repeat region undetected in the wild type protein. High affinity copper binding allows PrP to compete effectively for copper in the extracellular milieu. The copper binding affinities of PrP have been compared with those of proteins of known function and are of magnitudes compatible with an extracellular copper buffer or an enzymatic function such as superoxide dismutase like activity.

The cellular prion protein is known to be a copper-binding protein. Despite the wide range of studies on the copper binding of PrP, there have been no studies to determine the affinity of the protein on both full-length prion protein and under physiological conditions. We have used two techniques, isothermal titration calorimetry and competitive metal capture analysis, to determine the affinity of copper for wild type mouse PrP and a series of mutants. High affinity copper binding by wild type PrP has been confirmed by the independent techniques indicating the presence of specific tight copper binding sites up to femtomolar affinity. Altogether, four high affinity binding sites of between femto-and nanomolar affinities are located within the octameric repeat region of the protein at physiological pH. A fifth copper binding site of lower affinity than those of the octameric repeat region has been detected in full-length protein. Binding to this site is modulated by the histidine at residue 111. Removal of the octameric repeats leads to the enhancement of affinity of this fifth site and a second binding site outside of the repeat region undetected in the wild type protein. High affinity copper binding allows PrP to compete effectively for copper in the extracellular milieu. The copper binding affinities of PrP have been compared with those of proteins of known function and are of magnitudes compatible with an extracellular copper buffer or an enzymatic function such as superoxide dismutase like activity.
The cellular prion protein PrP c is a cell surface glycoprotein with large ␣-helical content that is attached to the plasma membrane by a glycosylphosphatidylinositol anchor. This cellular form of the protein binds copper, and it is argued that this copper binding to PrP c is intimately linked to the normal cellular function of the protein either in copper transport, sequestration, or antioxidant activity (1)(2)(3). The native cellular protein may undergo a conformational transformation to a proteaseresistant species known as a prion (PrP Sc ) with high ␤-sheet content and a tendency to aggregate. This transformed species is the putative proteinaceous infectious agent of transmissible spongiform encephalopathies or prion diseases (4). Prion diseases are characterized by a metal imbalance in the brain that occurs simultaneously to conversion of PrP c to PrP Sc . The metal imbalance is associated with the loss of copper binding from PrP c on conversion to PrP Sc (5,6). This raises important questions concerning the role of copper and other metals in prion diseases. Loss of protein function as a result of impaired copper binding in diseased PrP Sc could have serious implications for disease progression.
As a result, knowledge of copper binding to PrP is important in considering both the normal function of PrP c and its influence in prion diseases.
Although it is accepted that PrP c is a Cu(II) binding protein, there are many facets of the interaction of Cu(II) and PrP c that are disputed. These include the exact number of binding sites and their relative affinity for Cu(II). Initial studies focused on a fragment of the protein termed the octameric repeat region because it contains four repeats of an octamer. This region contains histidine and so was a likely candidate for the binding site. These early studies (7,8) suggested Cu(II) binds to this region with a micromolar affinity that was little more specific than histidine on its own. Further studies with larger fragments also suggested that Cu(II) binds PrP c , but again the affinity seen was similarly a low micromolar association (9). A study of the effect of PrP c expression on Cu(II) uptake into cells suggested that the affinity must at least be in the nanomolar range (1). Nanomolar affinities for the octameric repeat region have rarely been suggested (10). Only one publication has suggested that the affinity of Cu(II) for PrP is in the femtomolar range (11). Although such a high affinity implies that PrP is a very specific Cu(II)binding protein, this finding has never been reproduced. Therefore, there is currently no consent about the true affinity of Cu(II) for PrP c .
Initial studies focusing on the octa-repeat region suggested that there might be as many as four Cu(II) binding sites on PrP c . However, study of a peptide with amino acid residues 23-98 of PrP c suggested that this fragment could bind 5 Cu(II) atoms (9). Although four of these sites were designated as belonging to the octameric repeat region, the location of the so-called fifth site has been difficult to identify. There have been three suggestions, each associated with a histidine at either amino acid residue 96 (9), 111 (12), or 187 (13,14). However, the majority of reports suggest that, if there is a fifth site, it is also in the N terminus. A recent study suggests that the fifth site involves both histidines at 96 and 111 (15). Also, there is no evidence that five Cu(II) ions bind to PrP c in vivo. The only study to quantify Cu(II) bound to PrP c purified from mouse brain indicated that an average of three Cu(II) ions is bound to each molecule (3). It has also been suggested that the alternative Cu(II) binding site has higher affinity for Cu(II) than the octameric repeat region (11,15). None of these studies has estimated the relative affinity of these sites within full-length PrP and has relied on peptides and fragments. Therefore, resolution of these issues is still waiting for further study.
To address these issues, two independent techniques have been used to examine high affinity binding within wild type PrP. First, the thermodynamics of Cu(II) binding has been examined by isothermal titration calorimetry (ITC) 2 to measure affinity. Second, competitive metal capture analysis (CMCA) has been used to compare Cu(II) binding in PrP to other Cu(II) chelators of know affinity, enabling direct comparisons of Cu(II) affinities to be made. Where possible full-length protein PrP has been utilized to avoid the discrepancies associated with peptides, and Cu(II) has been added in a chelated form to mimic physiological conditions.

MATERIALS AND METHODS
Recombinant Protein Production-Recombinant mouse PrP protein was produced as described previously (16). Briefly, PCR-amplified product was cloned in the expression vector pET-23 (Novagen) and transformed into Escherichia coli ad494(DE3) or BL21(DE3) strains. Expressed proteins were solubilized by sonication in 8 M urea and recovered by immobilized metal ion affinity chromatography. Columns were charged with nickel in the case of His-tagged proteins and copper for untagged proteins. The eluted material was treated with 0.5 mM EDTA to ensure the protein was free from leached metal ions. All subsequent steps used double-deionized water treated with Chelex resin (Sigma) to remove residual metal ions. The denatured protein was refolded by a 10-fold dilution of the urea in deionized water, concentration by ultrafiltration, and two rounds of dialysis to remove residual urea, imidazole, and EDTA. Protein concentrations were measured using theoretical extinction coefficients at 280 nm (us.expasy.org/tools/protparam.html) and confirmed by BCA assay (Sigma). The protein was adjusted to 20 M in Chelex-treated water, pH 7, as 3-ml aliquots for immediate use in ITC experiments. Protein purity was checked using polyacrylamide gel electrophoresis under denaturing conditions and stained with Coomassie Brilliant Blue.
Mutagenesis-Constructs expressing mouse PrP amino acids 23-231 with and without C-terminal His 6 and mouse PrP-(23-231) with the octameric repeats deleted (PrP-(⌬51-89)) have previously been described (16). Mutations of the histidine residues in the proposed 5th Cu(II) binding site of mouse PrP were prepared using a PCR-based mutagenesis procedure involving paired oligonucleotides to create the single H96A (PrPH96A), H111A (PrPH111A), and double H96A,H111A (PrPH96A,H111A) mutants (residue numbering is based on the human PrP sequence). Mutagenesis was confirmed by DNA sequencing.
Isothermal Titration Calorimetry Measurements-All measurements were made on a Microcal VP-Isothermal Titration Calorimeter instrument. Briefly, a time course of injections of ligand to macromolecule or vice versa is made in an enclosed reaction cell maintained at a constant temperature. The instrument measures the heat generated or absorbed as a ligand-macromolecule reaction occurs. A binding isotherm is fitted to the data, expressed in terms of heat change per mole of ligand against the ligand to macromolecule ratio. From the binding isotherm values for the reaction stoichiometry, association constants K a , the change in enthalpies ⌬H°, and change in entropies ⌬S are obtained.
All solutions were filtered through a 0.22-filter and degassed before use. Weak metal-buffer interactions can cause discrepancies in ITC data (17). Titration of Cu(Gly) 2 into water adjusted to pH 7 resulted in less heat change than use of MES buffer, presumably due to a weak Cu(II)-MES interaction, whereas water only measures a heat of dilution.
Direct titration of protein solutions with aqueous Cu(II) salts was avoided because of the tendency of PrP to aggregate under such conditions (18,19) and to prevent the formation of insoluble Cu(II) hydroxide species. These may be avoided by the use of a Cu(II) chelate. The amino acids glycine and histidine have both previously been used as copper chelators in biochemical experiments (1,9). However titration of histidine in the absence of copper into PrP gave an isotherm indicating that a significant histidine-PrP interaction occurs (data not shown). Copper(II) forms a bis-glycine complex, Cu(Gly) 2 , in the presence of excess glycine. Therefore, a copper:glycine ratio of 1:4 was used by dissolving 3.0 mM copper(II)chloride and 12 mM glycine in Chelex-treated water. The excess glycine ensured that titrated Cu(II) was either chelated to glycine or incorporated into the protein, avoiding aqueous Cu(II) in the reaction cell. It also acted as a competitor to nonspecific protein Cu(II) interactions.
In our hands as others (17), the most reproducible data were obtained from injections of the metal into a solution of protein. Typically an initial injection of 2 l of Cu(II) chelate solution was followed by a further 29 injections of 4 l of Cu(II) into the protein in the sample cell stirred at 300 rpm. Injections were separated by 120 s to allow equilibration, and sample temperature was maintained at 25°C. All experiments were repeated at least three times. Data were analyzed with the Origin 5.0 software package from MicroCal. A base-line correction was applied to each experiment by subtraction of data from a series of injections of Cu(II) chelate solution into a buffer blank correlating to the heat of dilution of the Cu(II) complex. After subtraction of the blank data, a nonlinear least squares method was used to minimize 2 values and obtain best fit parameters for the association constants, K a , and the change in enthalpies, ⌬H°. In all cases best-fit parameters were obtained from the sequential binding sites model, whereby the user defines the number of binding sites to be fitted in a sequential manner. All sequential binding site models between 1 and 10 were assessed to ensure that the optimal data fit was obtained. Attempts to fit data to one or two identical site models gave unsatisfactory results. The affinity of Cu(II) to glycine was measured in water, pH 7, identical conditions to the protein experiments. The best fit model for the coordination of two glycine molecules to Cu(II), corresponding to the Cu(Gly) 2 complex, results in K 1 ϭ 4.0 ϫ 10 5 M Ϫ1 and K 2 ϭ 1.7 ϫ 10 4 M Ϫ1 , with ␤ 2 ϭ 6.8 ϫ 10 9 M Ϫ1 (data not shown), in agreement with the pH-adjusted literature values of K 1 ϭ 3.0 ϫ 10 5 M Ϫ1 and ␤ 2 ϭ 8.7 ϫ 10 9 M Ϫ1 (20). Affinities recorded for PrP were obtained in competition with Cu 2ϩ chelated by glycine; therefore, this must be factored into the reported affinities. This process is complicated from a lack of knowledge of the mechanism of copper binding. There is precedent for the use of the ␤ 2 value as the correct factor (11). However, an important prerequisite is that no mixed ligand complexes form (21) and generally assumes K 2 approximates or is larger than K 1 , not the case for Cu(Gly) 2 . Although it cannot be discounted that the initial interaction produces a PrP-Cu-Gly complex, requiring only the K 2 value to be factored in, this awaits further investigation. In the case of K 1 Ͼ K 2 , there is precedent for the K 1 value to be used as the correction factor assuming the dissociation of ligand (L) from ML 1 to be the competing event. Such an assumption is made in the subsequent CMCA assay for chelators able to form bis adducts (22) and has, therefore, been applied to the ITC data. Such a prudent approach also avoids the risk of overstating the affinity. Therefore, data reported in TABLE ONE are the product of the measured Cu-PrP interaction and 4.0 ϫ 10 5 M Ϫ1 (K 1 of Cu(Gly) 2 ); however, some mention is made in the text to values obtained from a correction using ␤ 2 ϭ 6.8 ϫ 10 9 M Ϫ1 were a concerted mechanism to be invoked.
CMCA-This technique was performed as previously described (22). It involved mixing one of a series of chelators (TABLE TWO) with Cu(II) at a ratio of 1:2 (metal:chelator). The Cu(II) binding affinities of the chelators have been published previously (20,22). The metal-chelator mixture was then mixed with the recombinant protein for 24 h at 37°C before being separated from the Cu(II) chelator solution by filtration. Filtration was carried out using concentration filters with a cut off of 3 kDa. The use of a copper-glycine blank and negative control protein, lysozyme, with Cu(II)-glycine has confirmed that three washes are sufficient to remove all of the Cu-ligand complex. The Cu(II) present in the protein and filtrate was assessed using a spectrophotometric assay (19). We have confirmed that this assay provides accurate measurements of bound Cu(II) by comparing results from this assay to results from inductively coupled plasma mass spectroscopy. Cu(II) used in these experiments was prepared by mixing pure copper at 10 mg/ml (National Institute of Standards and Technology, no. 3114) in 10% HNO 3 . All solutions were prepared in double-deionized water treated with the Chelex reagent to remove any trace of residual metals. Solutions were also filtered using a 0.22-m filter before the experiments. The reaction was carried out in a buffer containing 20 mM Tris and 150 mM NaCl. The pH of the buffer was adjusted to either pH 6.6 or 7.4. The reaction volume (1 ml) contained Cu(II) at 50 M, chelator at 100 M, and protein at 2 M. The proteins tested were wild type mouse PrP (PrP-(23-231)), PrP with the octameric repeat region deleted (PrP-(⌬51-89)) and a truncated PrP (PrP-(90 -231)). These proteins were prepared in an untagged form. Samples of the proteins reacted only in buffer were tested for Cu(II) content in parallel to ensure that no Cu(II) bound the protein as a result of the buffer conditions. Determination of the affinity constants (K app ) for each potential binding site was determined as previously described (22) and by using the online analysis program Equilibrate (www.equilibrate.homestead.com).

RESULTS
Analysis Using Isothermal Titration Calorimetry-The thermodynamic properties of ligand-macromolecule interactions can be accurately determined by ITC. The addition of Cu(Gly) 2 to all of the PrP isoforms used in this study resulted in an overall evolution of heat, indicated by negative peaks in the plots of raw ITC data (Fig. 1a). In contrast, titration of glycine alone into PrP showed no evidence of an amino acid-protein interaction (Fig. 1b) nor did Cu(II) chelate addition to protein where the Cu(II) binding sites had been saturated by refolding the protein in the presence of 5 mM Cu(II) (Fig. 1c). Binding affinities of Cu(II) to PrP species without His 6 tags studied are summarized in TABLE ONE. In all cases data have been analyzed by sequential binding sites models.
ITC data from the wild type protein PrP-(23-231) measured at pH 7 can be fitted to either 4 or 5 binding site model isotherms. By comparison, the minimized 2 values for fits of 1, 2, 3, and 6 binding sites were 52, 47, 2.7, and 9.2 times greater than the optimal fit, respectively. The highest binding affinity is observed for the first ion of Cu(II) to coordinate to PrP regardless of the stoichiometry of fit. When a 4-site fit is adopted, an affinity of 10 12.7 M Ϫ1 is measured, whereas a 5-site fit suggests an affinity of 10 11.8 M Ϫ1 for the first binding atom. These are affinities in the high femtomolar to low picomolar range. These values rise to 10 16.9 and 10 16.1 , respectively, if competition from both the K 1 and K 2 of the Cu(Gly) 2 complex are factored in (see "Materials and Methods"). This would place the affinity at the sub-femtomolar level. Three further Cu(II) ions proceed to bind in the 10s to 100s pM range (10 10 to10 10.8 M Ϫ1 ). In contrast, the fifth site has a lower affinity 10 8.8 M Ϫ1 . Interestingly, the PrP-(23-171) fragment largely mirrors the wild type protein with four or five binding sites measured; Fig. 2. This protein lacks the final 60 amino acids of the C terminus, including one of the candidate histidines for the fifth Cu(II) binding site. The affinities of this fragment to Cu(II) are approximately an order of magnitude lower than those measured for the wild type protein.
Two exceptions are the fourth sequential binding site, which has the largest affinity in this fragment, and the fifth site, which has a similar order of magnitude to that in the wild type protein.
Removal of the octa-repeat region (PrP(⌬51-89)) causes a significant reduction in the heat evolved on Cu(II) binding, suggesting a major disruption to the pattern of Cu(II) binding; Fig. 3. The data fit a binding isotherm of two sequentially filled sites with affinities of 10 10.5 to 10 8.7 M Ϫ1 . The associated errors of data fitted to 1, 3, and 4 binding site models were 13, 7, and 2.5 times greater than the 2-site model obtained. Significantly, Cu(II) binding capacity is lost on removal of the octarepeats. It is also apparent that the higher affinity binding sites require the presence of the octa-repeat region. The measured affinities are com-parable with the two lowest affinity sites of the five-site wild type protein.
The contribution to Cu(II) binding of the putative fifth site, based around histidine residues at positions 96 and 111, were also investigated. Mutants with histidine to alanine substitutions were produced. Alanine is considered innocent in metal binding. Titration of (PrPH96A,H111A) in which both histidines are substituted results in best fit isotherms for a four-binding-site model; Fig. 3. Surprisingly the 2 value for a five-site fit to the data was Ͼ30 times that of the four-site model, the next closest fit being three sites with a four times greater 2 . Again, the first Cu(II) binds with the highest affinity to 10 12.2 M Ϫ1 (fM) and the subsequent three sites from 10 10.7 to 10 10.1 M Ϫ1 . To probe the contributions of the individual histidine residues, the individual mutants PrPH96A and PrPH111A were produced. ITC data show that Cu(II) binding to the two histidine residues is not equivalent. The mutant PrPH96A fits a binding isotherm largely similar to the wild type protein; however, a five-site model always produces the best fit (lowest 2 ). Comparison of the log ␤ 5 values shows the combined affinities of the wild type and PrPH96A species to be identical. ⌻he PrPH111A mutant, however, consistently fits a four-site model and in this respect more closely resembles the double mutant species. Curiously, however, this single mutant species shows a trend to lower affinity Cu(II) binding than even the double mutant PrPH96A,H111A, with the first Cu(II) binding at 10 11.1 M Ϫ1 and a log ␤ 4 value of 40.5 (versus PrPH96A,H111A log ␤ 4 ϭ 43.5) Verification Using Competitive Metal Capture Analysis-The results from ITC analysis present a range of affinities for the wild type PrP protein suggestive of high affinity. Previous studies of Cu(II) affinity for PrP have been repeatedly unable to provide values that could be reproduced using different methods. It was, therefore, essential that we demonstrate the values obtained using ITC are not simply a result of the chosen technique. We used a second technique to assess the affinity of Cu(II) for PrP. This technique involved competing PrP with a range of chelators of known affinity listed in TABLE TWO and assessing the amount of Cu(II) bound to the protein. This technique differs markedly from ITC in that it assesses the ability of PrP to compete with a range of other molecules but maintains Cu(II) in a chelated form. The assessment of Cu(II) binding is retention of Cu(II) by the protein and not energy release. In this way PrP could be ranked with other Cu(II) binding molecules by direct observation. For this study three non-tagged forms of PrP were used: the wild type protein (PrP-(23-231)), the octameric repeat deletion (PrP-(⌬51-89)), and a truncated protein (PrP-(90 -231)). The experiment was carried out at two different pH values to assess the effect of pH on the observed affinities (Fig. 4). Pro-

DISCUSSION
The cellular prion protein PrP c is a cell surface glycoprotein attached to the plasma membrane by a glycosylphosphatidylinositol anchor. It is also widely accepted to be a Cu(II)-binding protein. However, there has been a lack of consensus in the literature concerning the stoichiometry and affinity of Cu-PrP interactions. These are pertinent to understanding the physiological relevance of Cu(II) binding. To address these matters we have undertaken a characterization of Cu(II) binding to recombinant PrP by the use of two separate analytical techniques. This has allowed a controlled and systematic study of Cu(II) binding in which high affinity binding to wild type PrP is measured to be in the femtomolar range for the first of five Cu(II) binding sites detected.
Recently, ITC has been used to examine the interaction of the prion protein with plasminogen (23). Additionally ITC has been used to obtain thermodynamic data of metal protein interactions (24,25) including copper in both its ϩ1 and ϩ2 oxidation states (17, 26 -28). To confirm the utility of ITC in our hands, we first showed that we could reproduce data for the binding affinities of the Cu(Gly) 2 species. Furthermore, use of a second method, competitive metal capture analysis, confirms the high affinity of Cu(II) binding measured is not purely a result of the technique employed.
High Affinity Cu-PrP Interactions-High affinity Cu(II) binding to wild type PrP is confirmed by both methods of analysis. The high (femtomolar) affinity measured for the first Cu(II) binding site is consistent with the observation that PrP c isolated from cerebellar cells grown in low Cu(II) media contain a minimum of one bound Cu(II) (3). The range of affinities measured by CMCA is between nano-to femtomolar for all but the fifth site, which is only detected at pH 7.4. ITC measurement of affinities of the second to fourth Cu(II) binding sites are all in the picomolar range separated by less than 1 order of magnitude, whereas the fifth site is an order of magnitude lower again. Notably, fitting of ITC data from sequential binding sites can result in an averaging of affinities when there is no change in phase of thermodynamics, as in this case, leading to slight grouping of the values (29). The product of all five binding sites ␤ 5 is in agreement between the two methods (log ␤ 5 Ϸ 50). The spread of affinities observed by CMCA is consistent with the observation that PrP c extracted from mouse brain has three ions of Cu(II) per protein molecule (3). The superoxide dismutase-like activity of PrP is proportional to Cu(II) occupancy of the octameric repeat regions (3,16). Our in vitro data support copper to protein stoichiometries of greater than one normally being accessible to PrP at the cell surface.
The observed decrease in binding affinities seen in subsequent copper binding events in the octameric repeats may be explained by the nature of the binding site. Recent peptide studies have suggested that a single Cu(II) occupant in the octameric repeat region is coordinated by three or more imidazole side chains of the histidine residues in the octameric repeat region (30). Such a coordination environment is likely to be of higher affinity than the proposed binding sites as copper to octameric repeat stoichiometry increases. These are thought to contain a single imidazole ligand and additional contributions from deprotonated atoms of the amide backbone (30,31).
Cu(II) binding to wild type PrP is sensitive to pH. The CMCA analysis shows a reduction of binding affinities of between 1.4 to 3.85 orders of magnitude on reduction of pH from 7.4 to 6.6. This is in accordance with the proposed octa-repeat binding motif composed of coordination through a histidine imidazole ring and the backbone amides of the following glycine residues (31). Notably, the fifth site in the wild type protein is not detected at pH 6.6 but emerges at pH 7.4. The ITC experi- Three recombinant mouse PrPs were subject to analysis of metal binding in the presence of each of the following chelators: Tris, glycine, arginine, methionine, asparagine, histamine, Bicine, ethylenediamine, histidine, bipyridyl, N-(2-hydroxyethyl)iminodiacetic acid, ethylenediamine-N-NЈ-diacetic acid, EDTA, diethylenetriaminepentaacetic acid, and CDTA, listed in order of increasing affinity for copper (see TABLE TWO). The proteins tested were wild type PrP (PrP-(23-231) E), protein without the octameric repeat region (PrP-(⌬51-89) •), and a PrP with the amino acid from the N terminus to position 89 deleted (PrP-(90 -231) ‚). After competition with the chelators the amount of copper bound to the protein was determined as was the unbound copper fraction. Experiments were carried out either at pH 7.4 or 6.6. The ratio of copper bound to the protein was plotted against the log K app for each of the chelators tested. Shown are the mean and S.E. of four experiments. PrP Copper Binding Affinity DECEMBER 30, 2005 • VOLUME 280 • NUMBER 52 ments were measured at pH 7 to maximize protein solubility at concentrations used, and this gives rise to the first measured affinity splitting those of the CMCA assays at pH 6.6 and 7.4. The pH sensitivity of binding probably accounts for some experiments with the wild type protein fitting a four as well as a five-site model. Functional Implications-Several functions have been ascribed to PrP as a result of its Cu(II) binding, including signaling (31), Cu(II) sequestration or uptake (1), and antioxidant function (2,3). Cu(II) has been shown to cause PrP c uptake into cells (32,33), but PrP c is not necessary for copper uptake. There is evidence that PrP c is involved in mechanisms against oxidative stress both in the transport of copper to Cu,Zn-superoxide dismutase (34) and as a superoxide dismutase like enzyme itself (2,3,35). The affinities of Cu(II) binding to PrP measured in this study are consistent with an ability to bind Cu(II) at the plasma membrane. Indeed cellular uptake has implied that the affinity must be at least in the nanomolar range. Extracellular Cu(II) is largely bound to ceruloplasmin and serum albumin in addition to lower molecular weight chelators such as amino acids. The high affinity Cu(II) site of serum albumin has been measured in the range 10 11.1 -10 13.2 M Ϫ1 (36 -38). Human ceruloplasmin, with 6 high affinity Cu(II) sites providing ferroxidase activity, also binds additional Cu(II) in lower affinity sites with K A ϭ 10 5-7 M Ϫ1 (38) as a mediator of extracellular copper homeostasis. PrP c could certainly compete for available Cu(II) with a femtomolar site and may suggest a role in Cu(II) buffering at the cell surface. Any such potential buffering capacity, however, appears not to greatly alter the brain copper content of wild type to Prnp knock out mice (39,40). The affinity of only one specific Cu(II) chaperone is available, that of the bacterial protein CopC, K A ϭ 10 4.1 M Ϫ1 (41), whereas the Cu(I)specific chaperones Atox1 and Cox17 have K A values of 10 5.4 and 10 6.8 M Ϫ1 , respectively (28,26). In contrast, the Cu(II) binding affinity of PrP is higher than these copper chaperones, which probably achieve specific delivery of the metal by kinetic control through a slow k off rather than a high affinity constant. Interestingly, the affinities of copper binding to PrP are also in the range of those previously obtained for Cu,Zn-superoxide dismutase, K A ϭ 10 12.6 -10 15.6 M Ϫ1 (42,43). The affinities of Cu(II) binding measured here strongly support PrP c being able to bind Cu(II) in the extracellular milieu and are also in the range expected of an enzyme such as superoxide dismutase.
Copper Binding and Pathology of Prion Diseases-It has been suggested that the presence of the copper cofactor stabilizes the N terminus of PrP by inducing secondary structure in a region unstructured in the metal free state (44). Indeed comparison of the affinity data of the wild type and 23-171 PrP fragment suggests that copper binding in the octarepeat region is influenced by the structured C terminus. Combined with the observation that copper-bound protein is less prone to aggregation suggests that copper binding either blocks a potential site for aggregation or stabilizes the structure, disfavoring misfolding and subsequent aggregation (45,46). Considering such a scenario suggests high affinity copper binding to be important in ensuring copper incorporation into the protein and, hence, protection from aggregation and prevention of a mechanism associated with prion disease.
It is known that metal imbalance occurs in prion disease (5,6). Furthermore, diseased PrP Sc isolated from infected animal brains contains no copper (6) nor does it have affinity for immobilized Cu(II), unlike PrP c (47). Therefore, the high affinity copper binding measured in this study is lost in diseased protein, presumably as a result of conformational change. The resultant accumulation of PrP Sc at the expense of PrP c explains the observed metal imbalance in terms of breakdown in copper regulation. Unregulated copper causes oxidative stress as a of Fenton-like chemistry. Oxidative stress is intimately linked to the pathology of prion diseases (48,49). This holds true whatever the function of PrP if copper is released as PrP Sc accumulates. However, the situation is exacerbated if as previously suggested PrP mediates copper homeostasis or acts as an antioxidant. In the first case serious loss of copper regulation leads to greater oxidative stress as an element of the copper management system breaks down. Meanwhile in the second case loss of antioxidant function renders neurons unable to counter the increased stress caused by the released copper. Therefore, we can postulate that high affinity copper binding acts as a barrier to prion disease; however, its loss severely contributes to disease pathology through oxidative stress.
Physiological Conditions of Copper Binding-Both of the methodologies employed in this study add Cu(II) to PrP in a chelated form as the addition of Cu(II) as an aqueous salt is not physiological and, therefore, not representative of an in vivo environment. Within the cell copper is extremely tightly regulated as it is in plasma (50,51). There is no evidence to suggest that extracellular copper management is different in the brain. Indeed a copper-histidine chelate is the most effective treatment of the copper deficiency associated with Menkes disease, suggesting that amino acid complexes are physiological copper carriers (52). Furthermore, the tendency of recombinant PrP to aggregate in the presence of Cu(II) salts is well documented as is the formation of insoluble Cu(II) hydroxide complexes at neutral to alkali pH (18,19). It is apparent from tissue culture studies that Cu(II) must be chelated for high affinity uptake by PrP on the cell surface (1). These are compelling arguments for Cu(II) to be added as a chelate and would account for the lower binding affinities at the micromolar level recorded by some studies. These have commonly added Cu(II) as an aqueous salt (53,54), which we have carefully avoided in our choice of technique and experimental conditions. Alternatively, the competition of the chelating ligand has not been accounted for in the affinity calculation (9). This initial study of Cu(II) binding to PrP did, however, equate the affinity of the Cu(II) PrP interaction to that of serum albumin. The apparent requirement of copper chelation to observe high affinity binding combined with the undoubted physiological relevance of chelated Cu(II) raises interesting points about the mechanism of uptake. In the absence of a definitive mechanism we have assumed that Cu(II) is incorporated into the PrP molecule through a monoglycine complex (based on K 1 Ͼ K 2 for the Cu(II)-Gly complex). However, we have already noted that should Cu(II) be transferred to PrP through a concerted mechanism involving loss of both glycines, the affinity of the first binding site would be subfemtomolar. Alternatively, it is also conceivable that PrP does not account for all of the ligands around the Cu(II). For example, an amino acid could fill one or more coordination sites to the exclusion of the deprotonated amide backbone. This would account for the high affinity uptake of copper by cells in the presence of a chelator and the higher affinities measured when copper is complexed (Refs. 1 and 11; this study).
This study has also used full-length protein rather than peptides. Previously two femtomolar affinity Cu(II) sites within PrP were identified by the use of peptides (11). This study, however, suggested that the remaining sites in the protein were unable to compete with glycine to bind Cu(II). In contrast, we have shown higher affinities in all of the first four binding sites, all of which are able to compete with glycine for Cu(II) binding. Furthermore, comparison with the PrP-(23-171) fragment suggests that truncation of the protein causes a lowering of binding affinities. This implies that the full-length protein is required for optimum Cu(II) binding. Accordingly, Cu(II) binding affinities suggested with peptides based on the octameric repeat region have been significantly lower, the majority measuring affinity in the micromolar range (7,8,55), although values up to nanomolar have been obtained (10,42). The PrP-(23-171) fragment does, however, have five binding sites that would appear to exclude a role for the histidine at residue 187 as the fifth Cu(II) site.
The Fifth Copper Binding Site-High affinity binding to the wild type protein occurs through the octa-repeats; both techniques show a loss of Cu(II) binding in mutants lacking the octameric repeats. Interestingly, in these mutants the affinity of the fifth site is enhanced, and a second lower affinity site, hidden in the wild type protein, emerges. Truncation of the protein, therefore, increases the Cu(II) binding affinity of two sites outside of the octameric repeat region. This supports the high affinity Cu(II) site measured by Jackson et al. (11) in the PrP-(90 -231) peptide, although we do not reproduce an affinity as high as the femtomolar range. Recently it was observed that the PrP-(90 -115) peptide binds two Cu(II) ions independently at histidine 96 and 111 (56). Substitution of both histidines by alanine only results in the loss of one Cu(II) binding site in the full-length protein. Significantly the lost site is that with the lowest affinity; therefore, in the wild type protein the octameric repeats provide the significant Cu(II) binding sites. Interestingly, it has recently been proposed that dipeptides of the avian hexa-repeat form more stable copper chelates than their mammalian octameric repeat analogs. This is attributed to the structural effects of four rather than two proline residues (57). Unlike the octa-repeats, the fifth-site histidines have no adjacent prolines that may stabilize the structure, leading to a higher binding affinity. Substitution of the histidine at residue 96 alone (PrPH96A) reveals no deviation from the behavior of wild type protein.
In contrast, substitution of histidine 111 results in the loss of the fifth site and a binding isotherm similar to that of the PrPH96A,H111A double mutant. This indicates that Cu(II) binding to the fifth site is modulated by the histidine at position 111. This is the dominant residue in the two Cu(II) binding PrP-(90 -115) peptide (56). In contrast, x-ray absorption fine structure indicated two histidines in the coordination sphere of this site (58), perhaps detecting both sites. His-96 has been suggested to be the dominant Cu(II) binding histidine of the fifth site; however, this study used a large excess of aqueous Cu(II) rather than a chelate (59).
Biological Implications-High affinity Cu(II) binding by wild type PrP has been confirmed by two independent techniques indicating the presence of specific tight copper binding sites. This binding is located within the octameric repeat region of the protein and is independent of the fifth-site histidine residues. Copper binding within the fifth binding site is of lower affinity in comparison to the octameric repeats and is modulated by histidine 111. Removal of the octameric repeats leads to the enhancement of affinity of two binding sites outside of this region. High affinity copper binding leading to a structure resistant to protein aggregation could be important in preventing prion diseases. However, once high affinity copper binding is lost with conversion to PrP Sc , the metal imbalance and oxidative stress associated with prion diseases occurs. High affinity copper binding allows PrP to compete effectively for Cu(II) in the extracellular environment and suggests a copper buffer in the style of serum albumin but could preclude a specific copper chaperone function. The binding affinities measured are also of magnitudes compatible with an enzymatic function, such as the proposed superoxide dismutase-like activity of the protein.