Prion Protein Binds Copper within the Physiological Concentration Range

: The prion protein is known to be a copper-binding protein, but aﬀinity and stoichiometry data for the full-length protein at a physiological pH of 7 were lacking. Furthermore, it was unknown whether only the highly flexible N-terminal segment with its octarepeat region is involved in copper binding or whether the structured C-terminal domain is also involved. Therefore we systematically investigated the stoichiometry and aﬀinity of copper binding to full-length prion protein PrP23–231 and to different N-and C-terminal fragments using electrospray ionization mass spectrometry and fluorescence spectroscopy. Our data indicate that the unstructured N-terminal segment is the cooperative copper-binding domain of the prion protein. The prion protein binds up to five copper(II) ions with half-maximal binding at ฀2 ฀m. This argues strongly for a direct role of the prion protein in copper metabolism, since it is almost saturated at about 5 ฀m,and the exchangeable copper pool concentration in blood is about 8 ฀m. The prion protein is known to be a copper-binding protein, but affinity and stoichiometry data for the full-length protein at a physiological pH of 7 were lacking. Furthermore, it was unknown whether only the highly flexible N-terminal segment with its octarepeat region is involved in copper binding or whether the structured C-terminal domain is also involved. Therefore we systematically investigated the stoichiometry and affinity of copper binding to full-length prion protein PrP 23–231 and to different N- and C-terminal fragments using electrospray ionization mass spectrometry and fluorescence spectroscopy. Our data indicate that the unstructured N-terminal segment is the cooperative copper-binding domain of the prion protein. The prion protein binds up to five copper(II) ions with half-maximal binding at ; 2 m M . This argues strongly for a direct role of the prion protein in copper metabolism, since it is

Prion diseases are fatal neurodegenerative diseases thought to be caused by conformational transition of the native and predominantly ␣-helical prion protein (PrP C ) 1 to the significantly more ␤-sheet-containing pathogenic isoform (PrP Sc ) (1). This conformational transition apparently induces the formation of PrP Sc aggregates (2), which are, in contrast to PrP C , highly protease-resistant (3). During infection, PrP Sc appears to serve as a template for the conversion of the native prion protein, because host PrP C expression is required for infection with prions (4). There is evidence that copper plays a role in the formation of PrP Sc (5).
PrP C is nearly ubiquitously distributed to all tissues. The highest levels are found in the brain (6). Native PrP C is an asparagine-linked sialoglycoprotein that is attached to the surface of the plasma membrane via a C-terminal glycophosphatidylinositol anchor (7). The structured C-terminal half of the full-length recombinant PrP  (amino acids 23-231) from amino acids 126 -231 is made up of two strands of one small antiparallel ␤-sheet and three ␣-helices (8). The N-terminal half from amino acids  showed no structure in NMR analysis and is therefore postulated to be highly flexible (9,10).
Although many efforts have been made the physiological function of PrP C has not yet been identified (11,12). In 1995 it was suggested that the octarepeat region ([PHGGGWGQ] 4 ) from amino acids 60 -91 within the flexible N-terminal half of the prion protein may play a role in binding copper (13). We have shown that PrP C has indeed a function in synaptic copper binding (14). Heavy metal binding studies indicated that PrP C seems to specifically bind copper (15). Half-maximal binding of two copper ions at 14 M was reported for Syrian hamster ShamPrP 29 -231 at pH 6.0 (15). In contrast, binding of 4 and 5.6 copper ions to the octarepeat region peptide (16,17) and humPrP   (18) at pH 7.4 with half-maximal binding at about 6 M were found, respectively. The copper binding to the octarepeat region was shown to be strongly pH-dependent in the range from 5 to 7 (14) with almost no binding at pH 5.
The lack of affinity and stoichiometry data for the full-length prion protein at a physiological pH of 7, as well as the question of which part of the prion protein is responsible for the copper binding, prompted us to systematically investigate the copper binding to the full-length prion protein as well as to different N-and C-terminal fragments using fluorescence spectroscopy and electrospray ionization mass spectrometry.
In this paper we present for the first time copper binding data on the full-length prion protein under physiological conditions at pH 7. Our data show that the prion protein binds up to five copper ions and is almost saturated at 5 M copper(II). Furthermore, our data suggest that the highly flexible N-terminal half is the cooperative copper-binding domain of the prion protein.
Expression and Purification of Full-length murPrP  and N-terminal Fragments humPrP  and humPrP  -The N-terminal fragment humPrP  was expressed in E. coli according to Brown et al. (18). After expression and thrombin cleavage of the GST-humPrP 23-98 fusion protein, the reaction mixture was applied to an EMD-COOH column and humPrP  was eluted with a gradient from 0 M up to 1 M ammonium acetate pH 6.0.
The plasmid pEThuPrP  was constructed for the expression of humPrP  following the cloning of the human PRNP ORF from genomic DNA (20). The coding region of the N-terminal segment of PrP (aa 23 to 112) was amplified via PCR under standard conditions using primers huPrP23-112up (5Ј-GGCCGGTCATGAAGAAGCGCCCGAAG-CCT-3Ј) and huPrP23-112do (5ЈGCCGGGAATTCTTATCACATGTGC-TTCATGTTGGT-3Ј). Primer huPrP23-112up contains a BspHI cleavage site and a start codon. Primer huPrP23-112do contains two stop codons and an EcoRI cleavage site. The construct was cloned into the NcoI and EcoRI sites of the inducible expression vector pET-21d (Novagen) and the sequence of the insert was verified by DNA sequence analysis The expression of the humPrP  in E. coli was induced with IPTG. Harvested cells were lysed with 1% Triton X-100 and centrifuged at 18,500 g for 15 min. The supernatant was applied to an EMD-COOH column, washed with 10 mM ammonium acetate pH 6.0. humPrP  and eluted with a linear gradient up to 1 M ammonium acetate pH 6.0 from the column.
Finally, both N-terminal PrP fragments were purified by reversed phase HPLC on a Vydac C 4 column (4 mm x 250 mm) using a linear water acetonitrile gradient of 10% to 50% acetonitrile including 0.1% trifluoroacetic acid. The full-length murPrP 23-231 was expressed and purified according to Liemann et al. (21). Before lyophilization, protein was dialyzed with Chelite P to remove traces of bound copper. The C-terminal fragment murPrP 121-231 was expressed and purified as described earlier (22).
Fluorescence Titration of PrP  and Its Fragments-Fluorescence spectroscopy was carried out on an LS 50B from PerkinElmer Life Sciences (Ü berlingen, Germany). Measurements were performed in 20 mM MOPS, pH 7.2, 100 mM NaCl, and 1 mM DTAC. MOPS buffer was treated with Chelite P to remove traces of heavy metal ions. To excite selectively tryptophan fluorescence, a wavelength of 295 nm was chosen, and the fluorescence signal was detected on the emission side at 355 nm. Excitation slit width varied from 3 to 5 nm, and emission slit width was in the range of 10 -15 nm. For reasons of sensitivity the excitation wavelength for murPrP 121-231 was set to 285 nm. Copper sulfate was added carefully to prevent possible precipitation (monitored by systematic excitation peak broadening in the emission spectra). Each point in the fluorescence titration curves represents the average of at least three measurements. Fluorescence titrations were performed at 20°C. Due to absorption of copper in the UV range, fluorescence titration curves were corrected for inner filter effect (23). All curves were corrected for background fluorescence.
ESI Mass Spectrometry-Mass spectrometry of N-terminal PrP fragments with and without copper were performed on a TSQ7000 (Finnigan) in the nano spray mode. The buffer was 1 mM NEMO/formic acid at pH 7.4. The applied spray voltage was 0.8 -1.0 kV. Capillary temperature was set to 150°C.
Peptide and Protein Quantification-Peptide and protein concentrations were determined in 6 M guanidinium chloride according to Gill and von Hippel (24). Peptide quantification by amino acid analysis for humPrP 60 -91 resulted in identical values.

Electrospray Ionization Mass Spectrometry of Full-length
PrP  and Its Fragments with Copper-ESI mass spectrometry was used to elucidate the stoichiometry of copper binding to the different N-terminal fragments humPrP 60 -91 , humPrP 60 -109 , humPrP  , and humPrP 23-112 as well as for the C-terminal fragment murPrP 121-231 and for the full-length  prion protein murPrP  . The experimental conditions for metal interaction studies at pH 7.4 with respect to suitable buffers and additives were intensively investigated. Finally, we introduced N-ethylmorpholine at pH 7.4 as a new buffer system for metal interaction studies with ESI MS, because it did not interfere with copper binding to peptides and protein compared with ammonium salt-containing buffers. Furthermore, addition of solvents like acetonitrile or methanol significantly affected the copper population in the ESI mass spectra (data not shown) and were therefore omitted. To avoid unspecific binding the copper concentration as well as the peptide concentration were kept as low as possible.
For a negative control experiment we used hen egg lysozyme, since it is similar to the prion protein with respect to its size and isoelectric point. Besides the potential unspecific copper binding sites at the N-terminal amino group as well as Cterminal and side chain carboxyl groups, it has one histidine at amino acid position 15 just after the first helix, which is exposed to the solvent. As clearly shown in the spectra of lysozyme with (Fig. 1, B and C) and without copper (Fig. 1A), even at a 10-fold excess of free copper(II) no significant binding occurred. For the first and second copper adduct mass differences of 61.7 and 61.3 average mass units were obtained, respectively. The low unspecific copper binding to lysozyme as a negative control demonstrated that NEMO is an excellent new buffer system for copper interaction experiments with proteins at a physiological pH of 7.
ESI mass spectrometry of the N-terminal fragments ( resulted in the expected average masses for the apopeptides of humPrP 60 -91 , humPrP 60 -109 , and humPrP   (Table I).Inthe spectrum of apo humPrP 60 -109 , a by-product (Fig. 2C, marked with triangles) with a corresponding mass of 4934.3 amu is observed. The mass difference of 127.6 suggests a missing lysine or glutamine residue in the by-product. The mass calculated from the spectrum of humPrP   (Fig. 2G) was 1 amu higher than the theoretically expected 9354.2 amu. This is probably due to the hydrolysis of Asn 108 to Asp 108 (25). The same holds for the by-product (marked with triangles) for which a mass of 8958.6 amu was obtained. The mass difference indicated that the three C-terminal amino acids KHM after Met 109 have been cleaved off from humPrP  .
Adding a 5-10-fold excess of copper(II) to the apopeptides resulted in mass spectra (Fig. 2, B, D, F, and H) with several copper adduct peaks. On addition of 100 M copper sulfate to 20 M humPrP 60 -91 , new m/z peaks appeared (Fig. 2B) corresponding to up to four bound copper(II) per peptide molecule. The two small peaks between the two following m/z copper peaks are always due to unspecific sodium and potassium adducts commonly observed in mass spectrometry of peptides and proteins at neutral pH. Adding 100 M copper(II) to 10 M humPrP 60 -109 resulted in the appearance of additional m/z peaks due to the complex formation of up to five bound copper per humPrP 60 -109 (Fig. 2D). Note that the peaks apparently corresponding to zero (marked by diamonds) and one bound copper for humPrP 60 -109 are also the main peaks for the byproduct with two and three bound copper, because their masses differ only by 4 amu and thus appeared as poorly resolved double peaks. Addition of 70 M copper(II) to 10 M humPrP   (Fig. 2F) led to the appearance of m/z peaks for up to six bound copper ions. It should be noted that adding a 10-fold excess (data not shown) above all significantly raised the intensity of the m/z peaks for six bound copper ions, sug-gesting a less specific copper binding site. In contrast, only adduct peaks for up to five bound copper were observed on addition of 100 M copper sulfate to 10 M of the 14 amino acids longer humPrP   (Fig. 2H). The significant shift of m/z peaks to lower charge values z and thus higher m/z values on addition of copper suggests conformational changes associated with copper binding, resulting in a more folded or structured conformation compared with the apopeptides.
The structured C-terminal domain of the prion protein murPrP 121-231 was analyzed by ESI mass spectrometry (Fig.  2I) and yielded a mass of 13335.5, which corresponds to the expected average molecular weight of 13334.8. On addition of up to 100 M copper to 13 M protein, only one bound copper was observed (Fig. 2J). This could be seen easier from the deconvoluted spectra (Fig. 3A). This site appears to be of lower affinity, because at 20 M copper(II), which is the 1.5-fold excess in copper, only a peak of little intensity is observed. At 40 M copper (a 3-fold excess of copper) the apoprotein peak almost disappeared, and the molecular weight peak for one bound copper became prominent. This remained almost unchanged, increasing the copper concentration up to 100 M and indicating that there is probably one copper-binding site with lower affinity in murPrP 121-231 . In contrast to the N-terminal fragments, there is no shift of the m/z charge envelope to lower values. Thus, presumably no conformational change occurs on binding of the copper ion.
From the mass spectrum of the apoprion protein murPrP 23-231 (Fig. 2K) an average mass of 23110.6 with a S.D. of 3.1 is obtained, which is about 6 amu higher than the expected molecular mass of 23104.4. The reason for this difference was unclear. On addition of up to 100 M copper, several not very well resolved copper adduct peaks (Fig. 2L) appeared. Due to the broad half-peak width and the high charges, the copper adduct peaks could be observed better from the deconvoluted The average molecular masses for all used peptides and proteins are summarized in Table I. From this table the average mass differences between the copper sites almost corresponds to 61.5 amu, which is typically observed in copper(II) binding ESI mass spectrometry experiments. Because the average mass of copper is 63.5 amu, the difference of 2 amu is explained as displacement of two hydrogens on binding of each copper ion (M ϩ n⅐Cu Ϫ 2⅐n⅐H ϩ ).
Comparing the results of humPrP 60 -109 , humPrP  , and humPrP  only to humPrP  , an additional sixth copper ion appears to be bound. For humPrP  , which is almost identical to humPrP 23-98 but C-terminally extended, only five bound copper were observed. Also humPrP 60 -109 , which shares an almost identical C-terminal extension to the octarepeat region, binds only five copper ions. Therefore the binding of a sixth copper ion to humPrP 23-98 appears to be an unspecific copper-binding site. Thus up to five copper ions appear to be bound specifically to humPrP 60 -109 , humPrP  , and humPrP  , whereas four copper ions are bound to the octarepeat region, suggesting an additional copper-binding site within the N-terminal segment. Because the additional copperbinding site occurs already in humPrP 60 -109 , which is compared with humPrP 60 -91 C-terminally extended, this additional binding site has to be located C-terminally of the octarepeat region.
Fluorescence Titration of PrP C and Its N-terminal Fragments with Copper-Fluorescence measurement and quantification of the copper binding to PrP C and its fragments were difficult with respect to some items. First, to prevent unspecific binding of copper to PrP C and its fragments, 100 mM NaCl was included. Second, to prevent unspecific aggregation and adsorption to surfaces 1 mM DTAC was added at a concentration clearly below critical micelle concentration (26). Third, for the quantification of affinities the peptide or PrP C concentrations were chosen to be significantly lower than the dissociation constants, to assume that the free copper concentration is approximately equal to the added copper concentration. Fourth, in contrast to, e.g. calcium, the fluorescence of peptides and proteins is not only quenched by conformational changes induced by copper(II) binding but also by diffusion-controlled collision of copper(II) with tryptophan and tyrosine side chains (27)(28)(29). Thus fluorescence titration curves have to be recorded at copper concentrations far beyond the saturation point to get quantitative data from the fluorescence titration curves.
The fluorescence titration curves with copper(II) for the Nterminal fragments humPrP 60 -91 , humPrP 60 -109 , humPrP  , and humPrP  ; the structured C-terminal fragment mur-PrP 121-231 ; as well as for the full-length recombinant murPrP  are shown in Fig. 4, A-F. All curves, except that for murPrP 121-231 , show more or less clearly visible sigmoidal changes in fluorescence intensity upon addition of copper, suggesting a basic cooperative binding mechanism. It appears that the peptides/protein cannot be saturated with copper at high concentrations. This is due to collisional quenching, which can be shown by plots according to the Stern-Volmer equation (28), fluorescence without quencher, and K SV is the quenching constant also called the Stern-Volmer constant. Above the saturation point of copper binding to the peptides or protein, the changes of F o /F should be linear in Stern-Volmer plots if this is only due to collisional quenching. Indeed, with the exception of murPrP 121-231 (Fig. 4F, inset) the linearity in the Stern-Volmer plots of PrP C and its N-terminal fragments (Fig. 4, A-E, insets) showed that the fluorescence changes above ϳ30 -50 M are the result of collisional quenching. The nonlinearity of the Stern-Volmer plot for murPrP 121-231 might be explained by the different excitation wavelength of 285 nm instead of 295 nm used for reasons of sensitivity. The fluorescence of murPrP 121-231 , which has 11 tyrosines and only 1 tryptophan, probably reflects not only the Trp fluorescence but also the fluorescence of tyrosines, which are expected to have different dynamic copperquenching properties.
From these Stern-Volmer plots, the Stern-Volmer constants as well as the fraction of conformational fluorescence change caused by binding of copper f conf are calculated by linear regression analysis at high copper concentrations. For this purpose Equation 1 has to be modified, because it is only valid where exclusively collisional quenching occurs. In the presence of conformational fluorescence changes true values for K SV can be calculated, introducing a correction factor according to Equation 2.
where f conf is the fraction of conformational fluorescence changes. Thus to obtain the true value for K SV the slopes from the Stern-Volmer plots have to be divided by the value of the ordinate 1/(1 Ϫ f conf ). The Stern-Volmer plots yield values for f conf and K SV (Table II). The Stern-Volmer constant is used later to fit fluorescence data by nonlinear regression to obtain copper binding data. The values for K SV represent a measure for the accessibility of tryptophan residues. For humPrP 60 -109 , humPrP  , and humPrP  , K SV decreases from almost 4000 M Ϫ1 to about 2500 M Ϫ1 , suggesting that the tryptophans become more buried in the holopeptides with longer fragment size. The reason for the at least 5-10-fold higher value for murPrP  is not known but has to be an effect of the C-terminal domain. Although the Stern-Volmer plot of murPrP 121-231 (Fig. 4F, inset) could not be fitted very well with Equation 2, a similar Stern-Volmer constant of 22,000 M Ϫ1 could be estimated considering only data from 30 to 80 M copper. From the fluorescence titration curves (Fig. 4, A-E)i ti s apparent that there is a significant increase in affinity for copper(II) going from humPrP 60 -91 (Fig. 4A)t ohumPrP 60 -109 (Fig. 4B). Again a small increase in affinity can be observed going further to humPrP   (Fig. 4C), whereas the affinities of humPrP  and humPrP   (Fig. 4D), as well as for the full-length recombinant murPrP   (Fig. 4E) seem to be identical.
To get quantitative copper binding data the titration curves were fitted by nonlinear regression analysis. ESI mass spectrometry has confirmed that four copper ions bind to the octarepeat region (humPrP 60 -91 ). The mass spectrometry data suggest that one additional copper binding site exists for humPrP 60 -109 , humPrP  , humPrP  , and murPrP 23-231. This site was assumed to be independent, because particularly the titration curves for humPrP  , humPrP  , and murPrP 23-231 could only be fitted with a mathematical term for an independent binding site. Assuming almost equal fluorescence changes per cooperative binding site, data of titration curves were fitted according to Equation 3, where F rel is the relative fluorescence; f ind and K ind are the fraction of fluorescence change and the dissociation constant for the independent binding site, respectively; f coop , K coop , and n are the fractional fluorescence change, the dissociation constant, and the Hill coefficient for the four cooperative binding sites, respectively; K SV is the Stern-Volmer constant; and [Cu 2ϩ ] is the free copper concentration. Because the peptide/ protein concentration was chosen to be significantly lower than the dissociation constants, the added or total copper concentration was assumed to be equal to the free copper concentration. For the fitting procedure the Stern-Volmer constants obtained from Stern-Volmer plots were inserted for K SV in Equation 3. Equation 3 was applied to all fluorescence titration curves except for humPrP 60 -91 , because only cooperative binding sites were assumed, and thus the term for the independent site was omitted. The same held for murPrP 121-231 , but n was set to 1, because mass spectrometry suggested binding of only one copper ion, and thus no cooperative binding can occur.
The quantitative data (Table II) show that C-and N-terminal extensions in the sequence of the octarepeat region humPrP 60 -91 have a significant influence on both dissociation constant K coop and Hill coefficient n. The dissociation constants for the cooperative binding of four copper ions to the octarepeat region decrease significantly from 5.5 to 2.5 M for humPrP 60 -91 and finally only slightly to 2.2 M for humPrP  , humPrP  , and murPrP  . Also Hill coefficients significantly increase from 2.4 to values ranging from 3.6 to 4.2 for humPrP 60 -109 , humPrP  , humPrP  , and murPrP    For the one putative copper-binding site in murPrP 121-231 ,a dissociation constant of about 8 M was estimated considering that no precise mathematical fitting was obtained (see Fig. 4F).

DISCUSSION
Here we present for the first time affinity and stoichiometry data on the copper binding to humPrP 60 -109 , murPrP 121-231 , and the full-length prion protein murPrP 23-231 under physiological conditions as well as for humPrP  representing a natural cleavage product of the prion protein (30,31). Our results clearly demonstrate that the full-length prion protein binds copper within the physiological concentration range with binding of five copper ions per molecule. Furthermore, we proved that the positive cooperativity of copper binding is not restricted to N-terminal fragments but is a property of the full-length protein. The systematical comparison of the data for full-length PrP with C-and N-terminal fragments indicates that the N-terminal segment represents the cooperative copper-binding domain of the prion protein.
Mass spectrometry has evolved as a method in the assessment and elucidation of copper and heavy metal binding stoichiometry of proteins and peptides (32,33). In contrast to Whittal et al. (34) we used ESI MS only for the copper binding stoichiometry studies for several reasons. First, our fluorescence affinity data indicated the copper binding sites of the prion protein and its fragments to be almost saturated with copper under the chosen concentrations in ESI MS. Therefore, it is expected to obtain mass spectra where predominantly the four or five copper adduct peaks are observed. This is not the case as readily visible from the spectra (see Fig. 2). It might be explained by the high flexibility of the N-terminal copper binding domain and thus its higher sensitivity to mechanical stress during the spraying procedure than more classical copper binding domains in which the copper binding sites are surrounded by several stabilizing secondary structure elements (35). Thus the ESI mass spectra appear not to reflect quantitatively the situation in solution but rather qualitatively for the prion protein and its peptides. Second, further investigations showed that ESI spectra are quite sensitive to changes of buffer and additives. The effect of ammonium salts on the ESI mass spectra of humPrP 60 -91 compared with NEMO as buffer is shown in Fig. 5. In contrast to the ammonium buffer system the third and fourth copper-binding sites are significantly populated using the N-ethylmorpholine instead. Additionally, the use of Tris with a functional primary amine prevented copper binding to humPrP 58 -91 in CD spectroscopy experiments (17). Summarized, ESI mass spectrometry of copper binding to prion protein and peptides appears to be quite sensitive to the chosen buffer and additive conditions. Therefore we have introduced N-ethylmorpholine as new buffer system in ESI mass spectrometry for metal interaction studies at a physiological pH of 7.
Our finding that the octarepeat region (humPrP 60 -91 ) binds four copper ions confirmed earlier mass spectrometry (13,34) and circular dichroism spectroscopy data (17). The mass spectra of humPrP 60 -109 , humPrP  , and humPrP 23-112 revealed one additional copper-binding site. This is in good agreement with the binding of 5.6 copper(II) ions to humPrP 23-98 from our earlier equilibrium dialysis experiments (18). From these experiments, as well as from the mass spectrum of humPrP   (Fig. 2C), a sixth binding site, which was also observed earlier (34), might be assumed. But because only five copper ions were observed for humPrP 60 -109 and humPrP  , this site has to be unspecific. The involvement of the N-terminal amino group of humPrP  for the unspecific site was proved by N-terminal acetylation (34).
Mass spectrometry data showed binding of up to five copper ions to the full-length protein, confirming one additional copper-binding site compared with the four of the octarepeat region. The data for the structured C-terminal domain murPrP 121-231 containing three histidine residues revealed one possible copper-binding site with an estimated dissociation constant of 8 M. This affinity is probably too low to be of physiological significance. For copper dissociation constants in the micromolar range, at least two side chains are essential from which one has to be a histidine or cysteine residue (36). Looking for this site in the three-dimensional NMR structure of the C-terminal fragment, we found a site made up of His 140 , Asp 147 , and Met 138 . These residues are on the surface and close enough to each other (distance: ϳ5 Å).
In contrast to the putative C-terminal copper(II)-binding site for the fifth binding site in humPrP  , humPrP  , and full-length prion protein, we obtained a dissociation constant of about 2 M. Since the additional fifth copper is already bound by humPrP 60 -109 , it has to occur C-terminal of the octarepeat region. This binding site might be located at His 96 , because it is known that besides the cysteine residue, the histidyl side chain is the major copper-binding site in peptides and proteins (36). In structured proteins at least two ligands, e.g. two histidine side chains, are necessary for copper binding in the micromolar concentration range (37). In flexible peptides without secondary structure only one histidine is essential for the copper binding. In such peptides the peptide backbone can adopt conformations in which peptide bond nitrogen is involved as a ligand in copper binding after deprotonation (38). Histidine hinders deprotonation of peptide bonds in the C-terminal position but promotes the deprotonation of N-terminally located peptide bonds as for N-acetyl-Gly-Gly-Gly-His (36). In this tetrapeptide the abstraction of the three adjacent peptide-bond hydrogens, together with the imidazole nitrogen, results in a 4N coordinated copper complex. The sequence at His 96 ... -Gly-Gly-Gly-Thr-His-...isstrikingly similar to N-acetyl-Gly-Gly-Gly-His, but also His 112 might be involved in copper binding. Further investigations are required to clarify the role of His 96 , His 112 ,o rH i s 140 in the copper binding of prion protein.
The sigmoidal form of fluorescence titration curves for the prion protein and its N-terminal fragments clearly indicated an underlying positive cooperative binding mechanism. This agreed well with CD spectroscopy data for humPrP 58 -91 , which is almost identical to our octarepeat fragment humPrP 60 -91 (17). In contrast, Whittal et al. negative cooperativity. This is probably explained by the limitations of ESI mass spectrometry in quantitative analysis and the buffer conditions as discussed above. Our results demonstrate that the positive cooperative copper binding is not restricted to the octarepeat fragment but is a general feature of the full-length prion protein.
The fluorescence data for humPrP 60 -91 yielded a cooperative dissociation constant of 5.5 M, which is almost identical to the value of 6 M for humPrP 58 -91 obtained by CD spectroscopy (17). Compared with equilibrium dialysis experiments for humPrP  with a dissociation constant of 5.9 M (18), we have calculated significantly lower dissociation constants of 2.2 and 1.4 M for the cooperative and the assumed independent site, respectively. This might be explained by the different experimental conditions. Our affinity data for humPrP 60 -91 and humPrP  are difficult to compare with those obtained by quantitative ESI MS (34) for the already mentioned reasons and because we have obtained one cooperative constant instead of four individual constants from which only two were calculable in the case of humPrP   (34).
The similarity of the fluorescence titration curves of humPrP  and humPrP  with the full-length prion protein murPrP 23-231 as well as the obtained data (Table II) indicated that the N-terminal segment is a copper-binding domain. The putative low affinity copper-binding site in the C-terminal domain (murPrP 121-231 ) apparently had no influence on the copper-binding mode and affinity of the N-terminal segment. Although the total N-terminal segment up to amino acid 126 is highly flexible, C-and N-terminal extensions of the octarepeat region (humPrP 60 -91 ) have a significant influence on the affinity and cooperativity of the independent and cooperative binding sites, respectively. They clearly increase if the octarepeat region is C-terminally extended to humPrP 60 -109 (Table II). The affinity of the assumed independent site becomes significantly higher on N-terminal sequence extension from humPrP 60 -109 to humPrP  or humPrP  . Thus the fluorescence data of the four cooperative sites as well as for the assumed independent site indicated that there are additional interactions stabilizing the holocopper-binding domain.
Fluorescence titration curve data show that the full-length murPrP  is almost saturated at about 5 M copper(II). This is within the physiological concentration range of the total copper concentration in blood plasma of 18.6 M (39). But this concentration does not represent the physiologically available copper pool, because about 10.2 M copper (65%) is bound to ceruloplasmin in a nonexchangeable way. In contrast about 2.8 M is bound to serum albumin, 1.9 M to transcuprein, and 3.6 M is bound to low molecular weight components like amino acids in an exchangeable way. Considering only the copper pool of 3.6 M bound by low molecular weight components and the copper affinities of the prion protein, there is strong argument for a direct role of the prion protein in copper metabolism. Additionally, copper concentrations ranging from 0.5 to 2.5 M are found in the cerebrospinal fluid, whereas 15 M is present in the synaptic cleft (40). In contrast to blood plasma the conditions at the synaptic cleft and in the cerebrospinal fluid are probably different in that no ceruloplasmin, serum albumin, or transcuprein is present.
The cooperative copper-binding mode of the prion protein within the physiological concentration range suggests two possible functions. First, similar to the cooperative oxygen binding of hemoglobin, it might play a role in copper transport. Second, the prion protein might be a copper sensor similar to the intracellular calcium-binding protein calmodulin, which also binds calcium in a cooperative way.
In mammals, relatively little is currently known about the precise components involved in copper transport and the mechanism by which copper is transported across the plasma membrane into cells (41). Cellular copper uptake by the prion protein might be similar to transferrin, which is mediated by endocytosis. Both transferrin and prion protein release their bound iron and copper at an acidic pH of 5.0 (15). A possible transport function for PrP C is supported by the stimulation of endocytosis with copper (42), the pH-dependent copper binding (17), and the lower copper content of brain in PrP C -deficient mice (18). This was recently further supported by our results showing that PrP C is apparently located in presynaptic membranes and that the loss of PrP C in Prnp 0/0 mice strongly affected the copper content of synaptosomes (14). This suggests that PrP C is involved in the synaptic copper homeostasis. The in vivo binding mode of copper to the prion protein is not yet known. Even if PrP C does not bind copper under normal conditions, it might serve as an extracellular copper sensor. It is known that cancer and infections increase plasma copper concentrations in blood (39). Thus it might be that PrP C binds copper and regulates cells on increasing copper concentration.
Copper binding to PrP appears to play a role in the prion disease as well (43). Copper apparently stabilized prions, because it enhances the regeneration of partially denatured PrP Sc (5). Our finding of a low affinity copper-binding site in the C-terminal domain of PrP C might be interesting for the conformational change to PrP Sc .
Further investigations are necessary to explore the molecular nature of the unique, cooperative, copper binding motif in the octarepeat region as well as to elucidate the role of copper binding for the prion protein in vivo.