Zinc modulates copper coordination mode in prion protein octa-repeat subdomains

Abstract

In this work we present and analyse XAS measurements carried out on various portions of Prion-protein tetra-octa-repeat peptides in complexes with Cu(II) ions, both in the presence and in the absence of Zn(II). Because of the ability of the XAS technique to provide detailed local structural information, we are able to demonstrate that Zn acts by directly interacting with the peptide, in this way competing with Cu for binding with histidine. This finding suggests that metal binding competition can be important in the more general context of metal homeostasis.

Introduction

Transmissible spongiform encephalopathies (TSE), also known as Prion diseases, belong to a broad class of diseases including pathologies that can affect different species. Among these diseases we find Creutzfeldt–Jakob disease in humans, scrapie in sheep (from which the generic name for the misfolded form of the Prion protein is taken), and bovine spongiform encephalopathy (or BSE) in cattle.

There is increasing experimental evidence that in all TSEs the occurrence of an abnormal form of a protein called a Prion (shorthand for proteinaceous infectious particle; Prusiner 1998) protein (PrP) is of crucial importance. PrP is a membrane protein that is anchored to the membrane surface via a glycosyl phosphatidyl inositol group. It occurs in two alternative conformers: the cellular native harmless conformer, PrP^C, rich in α-helix, and the aberrant pathogenic conformer, PrP^Sc (or scrapie PrP) that has self-replicating properties and is characterized by larger β-sheet content (Gasset et al. 1993; Pan et al. 1993; Morante 2008).

The structure of the C-terminal domain of PrP^C has been solved by both NMR (Dominikus et al. 2005) and X-ray diffraction (Burns et al. 2002) techniques, whereas the only available structural information about PrP^Sc is on its secondary structure (Eghiaian et al. 2004). The unstructured N-terminal region of mammalian PrP is characterized by a highly conserved domain consisting of a variable number (from four to six depending on the species) of tandem repeats of eight amino acid residues (PHGGGWGQ) termed octa-repeat.

Although the structural changes characteristic of the switch between the two conformers occur in the C-terminal domain, the N-terminal region seems to have a physiological function in subcellular trafficking (Nunziante et al. 2003). Among the different functions already tentatively attributed to PrP^C, Cu²⁺ binding is the only one that has been correlated with the physiological impairments linked to TSE (Stockel et al. 1998; Rachidi et al. 2003). The main Cu²⁺ binding site has been located within the octa-repeat region by use of several different experimental techniques (Hornshaw et al. 1995; Whittal et al. 2000; Burns et al. 2002; Qin et al. 2002; Wells et al. 2006). Other Cu²⁺ binding sites have been identified in the C-terminal globular domain (Hasnain et al. 2001; Cereghetti et al. 2001; Burns et al. 2003). Studies with PrP^C models in the form of synthetic and recombinant pep-tides have shown binding stoichiometries of up to one Cu²⁺ per octa-repeat group.

In previous work (Morante et al. 2004) X-ray absorption spectroscopy (XAS) was used to explore the Cu²⁺ site geometry in PrP peptides containing one, two, and four octa-repeats and in the whole recombinant form of bovine PrP (BoPrP). In that paper a model was proposed in which the Cu²⁺ coordination mode depends on the [Cu²⁺]:[octa-repeat] concentration ratio. The model is in good agreement with the EPR results of Chattopadhyay et al. (2005), in which it is shown that Cu²⁺ can be found in three different coordination modes, called “components”, whose relative abundance also depends on the [Cu²⁺]:[octa-repeat] concentration ratio.

In particular, the geometry of the component 1 metal site is seen to be characterized by the same structure visible in the crystal (Burns et al. 2002), where the metal binding site appears to be located within the octa-repeat subsegment HGGGW. Component 2 binding is observed only in PrP structures containing two or more octa-repeat segments and EPR data have shown that there His acts as a bidentate ligand, thus giving rise to a 2N₂O Cu²⁺ coordination. Component 3 coordination is, instead, observed only in tri and tetra-octa-repeat structures at low Cu²⁺ concentration (< 1.0 equiv.) suggesting that it only occurs if three or four sequential HGGGW segments (or in other words three or four (neutral) imidazole rings) are available.

More recently, new EPR measurements carried out by the same group (Walter et al. 2007) have shown that the presence of Zn²⁺ modulates mode of the Cu²⁺ binding to the synthetic PrP tetra-octa-repeat peptide. In particular it has been suggested that, even if Zn²⁺ is not able to completely remove Cu²⁺, increasing the Zn²⁺ concentration can progressively change the way Cu²⁺ is bound to the tetra-octa-repeat. This modification has been interpreted in terms of variation of the relative amount of each of the three components. The question that cannot be answered by EPR experiments is by which mechanism Zn²⁺ affects Cu²⁺ binding.

In the work discussed in this paper we exploited the possibility offered by XAS spectroscopy of directly and separately looking at the structure around both Cu²⁺ and Zn²⁺ ions when they are simultaneously present in the sample, thus overcoming the intrinsic limitation of Zn²⁺ being EPR silent.

Experimental procedures

XAS experiments

XAS experiments on the Cu and Zn K-edges of PrP-octa-repeat samples were performed at the BM30B beamline of the European Synchrotron Radiation Facility (Grenoble, France) (Proux et al. 2005). The storage ring was operating in 16 bunches mode at 6 GeV with a ~90 mA current. The beam energy was selected using an Si(220) double-crystal monochromator with an experimental resolution close to that theoretically predicted (namely ~ 0.5 eV) (Proux et al. 2006). The beam spot on the sample was approximately 300 × 200 µm² (H × V, FWHM). Because of the low Cu²⁺ and Zn²⁺ concentrations, spectra were recorded in fluorescence mode using a 30-element solid-state Ge detector. To avoid photo-degradation and spectra evolution during XAS measurements, all the samples were cooled to 13 K (−260°C) by use of a liquid helium cryostat. To prevent Cu photo-reduction samples are moved at each scan in order for the radiation not to hit the same portion of the sample.

The energy range of Zn K-edge spectra is shortened by the need to avoid effects on Zn K_α fluorescence of the Cu K_β fluorescence line contribution.

XAS data analysis

For each sample the several collected spectra were first averaged and then normalized using ATHENA software (Ravel and Newville 2005). This normalization procedure is necessary to enable quantitative comparison of the X-ray absorption near edge structure (XANES) regions among different samples. ATHENA is also used to extract the χ(k) signal¹ in the extended X-ray absorption fine structure (EXAFS) region. The χ(k) data are fitted by use of the EXCURV98 package (Binsted 1998) and structural information is extracted. The theoretical expression of χ(k) is calculated by taking into account single and multiple scattering contributions (Fonda 1992). The so-called constrained refinement strategy, which consists in treating each amino acid residue as a rigid body, is used. This enables reduction of the number of free variables characterizing the structural models taken into consideration in the analysis (Binsted et al. 1992).²

Sample preparation

Synthetic PrP octa-repeat peptides were prepared using the fluorenylmethoxycarbonyl (Fmoc) method, as previously described by Burns et al. (2002) and Aronoff-Spencer et al. (2000). The amino acid sequence of the synthetic peptide was: Ac-KKRPKPWGQ(PHGGGWGQ)₄-NH₂.

The initial nine amino acid residues, preceding the four consecutive octa-repeat groups present in the N-terminal of natural PrP, were inserted to increase peptide solubility. The peptide was, as usual, acetylated at the N-terminus and amidated at the C-terminus to mimic the presence of the missing portion of the natural protein. All samples were prepared by dissolving the peptide in degassed solvent containing 25 mM N-ethylmorpholine (NEM) buffer and 20% (v/v) glycerol at pH 7.4. The peptide concentration used for XAS samples was 0.2 mM. Cu²⁺ and Zn²⁺ were added as the salts CuSO₄ and ZnCl₂, respectively. As shown in Table 1, two classes of sample were prepared and measured: samples S_i (i = 1, 2, 3) prepared in the absence of Zn²⁺, and samples Si_Zn (i = 1, 2, 3), prepared in the presence of Zn²⁺. The corresponding two buffers were also measured. In Table 1 the Cu²⁺ concentrations of the measured samples, expressed in mM (column 2) or in equivalents (column 4), are reported. In column 5 we specify the metal at the K-edge of which measurements had been performed. We performed 11 (i.e. 9 plus 2 for buffers) sets of measurements. In the last two columns we also list the approximate Cu²⁺ binding component percentage, extracted from EPR measurements (Walter et al. 2007). As already suggested by Walter et al. (2007), we group components 1 and 2, which for short we call component α. For uniformity of notation we then denote component 3 by β.

Table 1.

List of measured samples

Sample	[Cu2+] (mM)	[Zn2+] (mM)	Cu2+ bound (eq.)	Edge	Cα (%)	Cβ (%)
S1	0.16	0	0.8	Cu	0	100
S2	0.4	0	2	Cu	80	20
S3	0.6	0	3	Cu	100	0
S1_Zn	0.16	0.6	0.8	Cu, Zn	55	45
S2_Zn	0.4	0.6	2	Cu, Zn	85	15
S3_Zn	0.6	0.6	3	Cu, Zn	100	0
B_Cu	2.0	0		Cu
B_Zn	0	2.0		Zn

For each sample (column 1) the concentration of Cu²⁺ in mM and equivalents (columns 2 and 4) and of Zn²⁺ in mM (column 3) is given. In column 5 the edge at which each spectrum has been collected is reported. In the last two columns we give the percentage of α and β components (see text) as obtained from Fig. 1

Results

Figure 1, where we summarize the EPR results already presented by Walter et al. (2007), clearly shows that the mode of Cu²⁺ coordination is indeed modified by the presence of Zn²⁺. In particular it is seen that the effect of Zn²⁺ is stronger at moderate Cu²⁺ concentrations and tends to disappear at high Cu²⁺ concentration. At moderate Cu²⁺ concentrations, where Zn²⁺ is more effective, the presence of Zn²⁺ induces an increase of the α component, thus pushing the system toward Cu²⁺ binding modes in which the number of bound His units is lower.

Fig. 1.

Plot of α (red) and β (black) component concentrations versus Cu²⁺ concentration. Full lines are in the absence and dotted lines in the presence of Zn²⁺. The three blue vertical lines are drawn to indicate the three Cu²⁺ concentrations at which XAS data were collected (Table 1). Concentrations are given both in mM (right and top axes) and in equivalents (left and bottom axes)

XAS technique has the potential to (dis)prove this mechanism because, at variance with EPR, it enables direct investigation of the atomic environment of Zn²⁺. In particular we want to discriminate between the situation in which Zn²⁺ affects the Cu²⁺ binding mode by directly binding to the peptide and the situation in which, by some more complicated and indirect mechanism, Zn²⁺ induces structural peptide modifications that in turn affect the geometry of the Cu²⁺-binding site. With the purpose of investigating these features we acquired XAS spectra of samples with different Cu²⁺ concentrations in the absence and in the presence of Zn²⁺ at the Cu and Zn K-edges.

Starting with a 0.2 mM tetra-octa-repeat solution which yields a histidine concentration [His] = 0.8 mM, we prepared three samples containing Cu²⁺ at 0.16, 0.4 and 0.6 mM, corresponding to 0.8, 2 and 3 equivalents (= [Cu²⁺]/[pept]), respectively (Table 1; Fig. 1).

Each solution was further subdivided into two sub-samples and 0.6 mM Zn²⁺ was added to one of these. The two reference buffer solutions were prepared as described above.

To gain an initial understanding of the structural properties of the metal (Zn²⁺ and Cu²⁺) coordination modes, we started by examining gross similarities and differences among the many spectra we measured. For this purpose we first compared the XANES and/or EXAFS regions of the spectra qualitatively but systematically.

Qualitative XAS data analysis

From Fig. 2 it is immediately evident from the shape of the feature at the top of the sharply rising part of the spectrum (called “white line”) that the spectrum of Cu²⁺ in buffer is significantly different from that of Cu²⁺ bound to the peptide at all the Cu²⁺ concentrations at which data were acquired.

Fig. 2.

Comparison among XANES spectra taken at the Cu K-edge of the three samples prepared at the three different Cu²⁺ concentrations of Table 1, in the absence of Zn²⁺, namely S₁ (black), S₂ (blue), S₃ (green). The spectrum of the B_Cu buffer (red) is also shown. Here µ(E) is the normalized absorption coefficient²

In Fig. 3, the same type of comparison as in Fig. 2, but referred to the EXAFS region, is displayed. As usual, in order to partly compensate for the rapid signal decrease with increasing k, the experimental data have been multiplied by k³. Because of the very low Cu²⁺ concentration in sample S₁, the corresponding spectrum (black line) appears to be rather noisy. Nevertheless, fixing our attention on the low-k part of the EXAFS spectrum, where noise is under control, we can observe that again the Cu-buffer (B_Cu) spectrum is markedly different from that of all the other samples, whereas the S₂ and S₃ spectra are very similar but substantially different from that of S₁. By focussing on the appearance of a double peak in the wave number region between k ≈ 3.5 Å⁻¹ and k ≈ 4.5 Å⁻¹, which is a fingerprint feature for the presence of bound histidine residues (Ferreira et al. 2002), we can conclude that their number decreases going from S₁ to S₃, and it is obviously zero in B_Cu.

Fig. 3.

The same comparison as in Fig. 2 for the EXAFS region. Colours are as in Fig. 2

This qualitative behaviour is the same as that observed in EPR experiments (Walter et al. 2007). In fact, at low Cu²⁺ concentration (sample S₁) only the β component is present, and its fraction decreases with increasing Cu²⁺ concentration. Indeed, the fraction of β component is already very low at 2 Cu²⁺ equivalents (sample S₂) and completely disappears at 3 Cu²⁺ equivalents (sample S₃), as seen in Fig. 1.

The same type of analysis was then performed by comparing the spectra taken at the Zn K-edge at the three different Cu²⁺ concentrations (samples S_i_Zn in Table 1). Now the reference spectrum is that of the Zn–buffer sample (B_Zn). Looking at the XANES region (Fig. 4), one can conclude that, at least for sample S₁_Zn, a significant amount of Zn²⁺ must be bound to the peptide, because its spectrum shows general features substantially different from that of the B_Zn sample.

Fig. 4.

Comparison among XANES spectra taken at the Zn K-edge of the three samples prepared at the three different Cu²⁺ concentrations of Table 1, in the presence of Zn²⁺, namely S₁_Zn (black), S₂_Zn (blue), S₃_Zn (green). The spectrum of the B_Zn buffer (red) is also shown

The same qualitative conclusions can be drawn by studying the EXAFS part of these spectra, shown in Fig. 5. Indeed the general spectral features of S₁_Zn (which is the sample at the lowest Cu²⁺ concentration) are substantially different from those of Zn²⁺ in buffer, whereas the spectra of the samples in which the Cu²⁺ concentration is higher (S₂_Zn and S₃_Zn) are almost identical to each other and very similar to the spectrum of Zn²⁺ in buffer. In other words, the local Zn²⁺ environment in the S₁_Zn sample is significantly different from that of Zn²⁺ in solution, thus proving that at least some of the Zn²⁺ is bound to the peptide.

Fig. 5.

Same comparison as in Fig. 4 for the EXAFS region

Finally, in order to better display the effect of Zn²⁺ on the Cu²⁺ coordination geometry, the EXAFS spectra of S_i and S_i_Zn samples are compared pair-wise in Fig. 6, focussing attention on the low-k region of the spectrum in which the data are statistically more accurate. The comparison confirms that only in sample S₁ is the mode of Cu²⁺ coordination substantially affected by the presence of Zn²⁺.

Fig. 6.

Pair-wise comparison between samples in the absence, S_i, and in the presence, S_i_Zn, of Zn, in the low-k region of the EXAFS spectrum at the Cu K-edge

Quantitative XAS data analysis

The qualitative analysis presented in the previous sections has led to two main conclusions:

1. the presence of Zn²⁺ does affect the mode of Cu²⁺ coordination, and

2. Zn²⁺ is able to bind to the PrP-octa-repeat domain by partially displacing Cu²⁺.

In this section we want to make the second of these statements more quantitative and precise by trying to structurally characterize the Zn²⁺ binding site environment. This is done by fitting the EXAFS data at the Zn K-edge to appropriate atomic model structures.

Measurements at the Cu K-edge are not really useful for this discussion because the only sets of data in which the EXAFS signal-to-noise ratio is sufficiently good (i.e. those corresponding to samples S₃ and S₃_Zn) do not appear to be in any appreciable way affected by the presence of Zn²⁺. This conclusion also follows from the EPR data of Fig. 1, in which it is clearly apparent that at this high Cu²⁺ concentration the solid lines (EPR data in the presence of Zn²⁺) tend to coincide with the broken lines (EPR data in the absence of Zn²⁺).³

We shall, then, concentrate on data at the Zn K-edge, and in particular for sample S₁_Zn, for which the effect of the presence of Zn²⁺ on the Cu²⁺ coordination mode looks more pronounced (compare solid and broken curves in Fig. 1). Our objective is to extract from the XAS data the important Zn²⁺-related structural information that is not directly available from EPR.

To proceed to quantitative analysis, we make the assumption that the Zn²⁺ can be present in our samples in two different structural configurations, one corresponding to Zn²⁺ in solution (complexed as in the buffer sample B_Zn) and the second corresponding to Zn²⁺ directly bound to the peptide. At this point, because the geometrical information that one can extract from the XAS data is the result of an average of the signals from all the structural Zn²⁺ environments, we need an estimate of the relative weight of these two Zn²⁺ coordination modes.

At a given the total His concentration, [His]_tot, the His fraction available for Zn²⁺ coordination, [His]_av, is:

$${[\text{His}]}_{\text{av}}={[\text{His}]}_{\text{tot}}-{[\text{His}]}_{\text{Cu}}$$ (1)

where [His]_Cu is the His fraction that is already engaged in Cu²⁺ binding. To estimate [His]_Cu in Eq. (1), we can exploit the EPR results of Chattopadhyay et al. (2005). EPR data show, in fact, that, in the presence of Zn²⁺ and at 0.8 Cu²⁺ equivalents (sample S1_Zn), approximately 45% of Cu²⁺ is found in component β⁴ and the remaining 55% in component α (Fig. 1), and no Cu²⁺ is found free in solution.⁵ In other words, all Cu²⁺ ions are involved in some kind of peptide binding because none can be fully substituted by Zn²⁺. This implies that the amount of His bound to Cu²⁺ can be estimated by use of the formula:

$${[\text{His}]}_{\text{Cu}}=(0.45\times 4+0.55\times 1)\times 0.16\text{ mM}=0.38\text{ mM},$$ (2)

where the numbers 0.45 and 0.55 are the fractions of Cu²⁺ in the β and α components, respectively, and 0.16 mM is the Cu²⁺ concentration in sample S₁_Zn; the integers 4 and 1 are the numbers of His units bound to the Cu²⁺ in the β and α components, respectively.

Because, according to our assumption, the total amount of Zn²⁺ in the sample, [Zn²⁺]_tot, is distributed between Zn²⁺ bound to the peptide, [Zn²⁺]_His, and Zn²⁺ free in solution, [Zn²⁺]_buffer,⁶ in the context of our two-component model, we obtain:

$${[{\text{Zn}}^{2+}]}_{\text{tot}}={[{\text{Zn}}^{2+}]}_{\text{His}}+{[{\text{Zn}}^{2+}]}_{\text{buffer}}$$ (3)

We also assume for simplicity that for each sample the number of His residues bound to Zn²⁺ is homogenous. We are thus led to consider four possible models, M_n, according to the number n (=1, 2, 3, and 4) of His residues bound to Zn²⁺. With the above simplifying assumptions we obtain:

$${[{\text{Zn}}^{2+}]}_{\text{His}}^n=\frac{{[\text{His}]}_{\text{av}}}{n}$$ (4)

where n is the number of His units supposed to be bound to Zn²⁺. At this point, by use of Eq. (1), we finally obtain:

$${[\text{His}]}_{\text{av}}={[\text{His}]}_{\text{tot}}-{[\text{His}]}_{\text{Cu}}=(0.8-0.38)\text{ mM}=0.42\text{ mM}$$ (5)

In Table 2 we have reported the concentrations of Zn²⁺ bound to His (column 3) together with the corresponding concentration of Zn²⁺ in the buffer (column 4) for each of the four models considered (column 1), computed using Eqs. (1)–(4).

Table 2.

List of proposed Zn²⁺ coordination models (column 1) with n the number of Zn²⁺ bound His’s (column 2)

Two-component model	n	[Zn2+]His (in %)	[Zn2+]buffer (in %)
M1	1	0.42 mM (70)	0.18 mM (30)
M2	2	0.21 mM (35)	0.39 mM (65)
M3	3	0.14 mM (23)	0.46 mM (77)
M4	4	0.11 mM (18)	0.49 mM (82)

Columns 3 and 4 report the Zn²⁺ concentrations (in %) bound to His units, [Zn²⁺]_His, and free in solution, [Zn²⁺]_buffer, respectively

To check the validity of this two-component model, we must first obtain reliable structural information about the atomic environment of Zn²⁺ in buffer. The best fit of the EXAFS B_Zn data is shown in Fig. 7 and is obtained by assuming Zn²⁺ is coordinated to six oxygen atoms located at a distance of 2.07 Å in octahedral geometry, as expected on the basis of, for instance, the results of D’Angelo et al. (2002a, b).

Fig. 7.

Best fit of the EXAFS Zn²⁺ buffer spectrum (upper panel). Experimental data are in black and the best-fit curve is in grey. The amplitude of the corresponding FT spectra is shown in the lower panel

As for the S₁_Zn EXAFS spectrum, four different fits (Fig. 8) have been performed according to the number of His units assumed to be bound to Zn²⁺. The fraction of Zn²⁺ that remains free in solution in each case can be found in Table 2.

Fig. 8.

Comparison between experimental XAS spectra of sample S₁_Zn at the Zn K-edge (black) and the simulated spectra (grey) corresponding to the four models of Table 2. EXAFS data are shown in the left column. The amplitudes of the corresponding FT spectra are displayed in the right column

It is rather clear from Fig. 8 that by increasing the number of His units supposedly bound to a single Zn²⁺ ion (i.e. moving from model M₁ to model M₄) and thus correspondingly increasing the fraction of Zn²⁺ in solution (Table 2), the quality of the fit (quantified in terms of the standard quality factor, R⁷) significantly deteriorates.⁸ In particular, R seems to be unacceptably high for models M₃ and M₄ (49 and 60%, respectively). In contrast, it decreases to rather good values, 34 and 35%, for models M₁ and M₂.

The net outcome of the analysis is that there are qualitative and quantitative reasons to discard models M₃ and M₄ as unphysical, and, although we do not have enough “resolution” to choose between models M₁ and M₂, they seem to yield a quite accurate description of the Zn²⁺ structural environment.⁹

We end by noting that in the Fourier transform (FT) of the simulated spectra the peak at approximately 3 Å (which is a fingerprint of the presence of histidine and is expected to become increasingly intense with the number of bound histidines (Strange et al. 1990)) is not visible in our case, because, as we explained above, in the situation in which more histidines are bound to Zn²⁺, more Zn²⁺ is found in solution (Table 2). Indeed, because the simulated XAS spectrum is the result of the sum of the features coming from all possible Zn²⁺ coordination modes, the histidine peak is either too small (when only one histidine is bound) or obscured by the superimposed contribution coming from Zn²⁺ in buffer (when a larger number of histidines is bound).

Fourier transform and BVS analysis

Further information about the structure of the Zn²⁺ binding site, confirming our previous results, can be extracted by using the FT of the XAS spectra obtained from sample Si_Zn (i = 1, 2, 3) to set up a bond valence sum (BVS) analysis (Brown and Altermatt 1985).

Assuming the validity of the two-component model described above, we expect the location, $\langle r_i^{\text{FT}}\rangle$, of the main peak in the |FT| (i.e. the mean distance of the first shell atoms from the metal), to be given by the formula:

$$\langle r_i^{\text{FT}}\rangle =\frac{x_{\mathit{\text{in}}}\times 6\times \langle r_{\text{buffer}}\rangle +(1-x_{\mathit{\text{in}}})\times 4\langle r_{\mathit{\text{in}}}\rangle }{x_{\mathit{\text{in}}}\times 6+(1-x_{\mathit{\text{in}}})\times 4}$$ (6)

in which the ratio:

$$x_{\mathit{\text{in}}}=\frac{{[{\text{Zn}}^{2+}]}_{\text{buffer}}^{\mathit{\text{in}}}}{{[{\text{Zn}}^{2+}]}_{\text{total}}}$$ (7)

represents the fraction of Zn²⁺ free in solution in the three Si_Zn (i = 1, 2, 3) samples expected for the four models M_n (n = 1, 2, 3, 4). The quantity ${[{\text{Zn}}^{2+}]}_{\text{buffer}}^{\mathit{\text{in}}}$ is computed by use of Eqs. (3) and (4) and hence depends on the sample and the model. In Eq. (6) 〈r_buffer〉 is the Zn²⁺ first shell mean distance of the (x_in fraction of) Zn²⁺ that is free in solution. From the fit to the XAS data we find 〈r_buffer〉 = 2.07. Finally 〈r_in〉 is the Zn²⁺ first shell mean distance of the remaining (1 – x_in) Zn²⁺ fraction bound to the peptide and represents the unknown quantity we ought to determine in the different situations under study. The coordination number is taken to be 6 for Zn²⁺ in solution and 4 for Zn²⁺ bound to the peptide.

In Table 3 the quantities appearing in Eq. (6) for all the samples (column 1), i = 1, 2, 3, and all the models (column 2), n = 1, 2, 3, 4, are reported. The first shell mean distance, $\langle r_i^{\text{FT}}\rangle$ (column 3), is obtained by first shell Fourier filtered data fitting in which four oxygen Zn²⁺ coordination is assumed. The values of 〈r_in〉, obtained solving Eq. (6), are reported in column 5.

Table 3.

Summary of the results from analysis of the four proposed Zn²⁺ coordination models, M_n

Sample	Model	〈ri〉(Å)	xin	〈rin〉(Å)	(BVS)in
S1_Zn	M1	2.026 ± 0.009	0.30	1.99 ± 0.02	1.88 ± 0.08
	M2		0.65	1.89 ± 0.03	2.5 ± 0.3
	M3		0.77	1.78 ± 0.05	3.3 ± 0.5
	M4		0.82	1.67 ± 0.07	4.6 ± 0.9
S2_Zn	M1	2.051 ± 0.007	0.58	2.00 ± 0.02	1.8 ± 0.1
	M2		0.79	1.92 ± 0.04	2.3 ± 0.4
	M3		0.86	1.83 ± 0.07	2.9 ± 0.5
	M4		0.90	1.7 ± 0.1	4 ± 1
S3_Zn	M1	2.075 ± 0.007	0.67	2.08 ± 0.03	1.5 ± 0.1
	M2		0.83	2.08 ± 0.05	1.5 ± 0.2
	M3		0.89	2.08 ± 0.09	1.5 ± 0.4
	M4		0.92	2.1 ± 0.1	1.5 ± 0.5
Buffer		2.075 ± 0.005			2.2 ± 0.2

Column 1 shows the measured samples, column 2 the proposed model, column 3 the fraction of Zn²⁺ in buffer (from Eq. (7)), column 4 the mean first shell distance (computed from Eq. (6)), column 5 the (BVS) values computed from Eq. (9)

With this information available a possible means of choosing among the different metal coordination modes is computation of the so-called bond valence sum (BVS) (Brown and Altermatt 1985). The value of BVS is, in fact, expected to be approximately the formal oxidation number of the metal ion, i.e. 2 in our case. (BVS) is computed by use of the formula:

$$(\text{BVS})=\sum _{\ell }\text{exp}\frac{r_{\ell }^{(0)}-\langle r_{\ell }\rangle }{B_{\ell }},$$ (8)

where the sum over ℓ extends over the contributions from all the metal ligands. The quantities $r_{\ell }^{(0)}$ and B_ℓ are phenomenological variables depending on the nature of metal and ligands, and can be found in Thorp (1992). In our case Zn²⁺ is bound to four ligands, all at the same distance, which can be either O or N, according to the model we consider (column 2 of Table 3). Because r⁽⁰⁾ and B for O and N are very similar, with $r_{\text{Zn}-N}^{(0)}=1.72\AA$ (Brown 2010), $r_{\text{Zn}-O}^{(0)}=1.70\AA$, and B_Zn−O≈B_Zn−N (34), Eq. (8) can be usefully simplified to the form:

$${(\text{BVS})}_{\mathit{\text{in}}}=4\times \text{exp}\frac{{\overline{r}}^{(0)}-\langle r_{\mathit{\text{in}}}\rangle }{B}$$ (9)

with r̄(0) the average value between $r_{\text{Zn}-N}^{(0)}$ and $r_{\text{Zn}-O}^{(0)}.$ Indices i and n (with the meaning specified above) which have been appended to BVS as the actual Zn–ligand distances, 〈r_in〉, depend both on the model and the sample. The BVS values computed according to Eq. (9) are given in the last column of Table 3.

Looking at the whole set of numbers reported in Table 3, the first striking observation is that in sample S₃_Zn, the mean distance obtained from the |FT| data (column 3) is equal, within experimental error, to the first shell distance of Zn²⁺ in buffer (last row). This equality leads in turn to values of 〈r_3n〉 that are perfectly consistent with the first shell distance of Zn²⁺ in buffer. We then conclude that in sample S₃_Zn, Zn²⁺ is free in solution, and Eq. (6) loses its meaning, because in this case Zn²⁺ is expected to be coordinated by six ligands.

Information contained in BVS can be used to identify the range of “acceptable” values for the fraction, x_in, of Zn²⁺ free in solution in each sample. For this purpose we exploit Eq. (9) to obtain the 〈r_in〉 values corresponding to an acceptable BVS range of values for the Zn²⁺ oxidation number. A 10% error is normally attributed to a BVS calculation, thus leading to the acceptable interval (BVS_min = 1.8, BVS_max = 2.2). Inverting Eq. (9) leads to r_max = 2.01 Å and r_min = 1.93 Å . In Fig. 9 〈r_in〉 is plotted against x_in, with the two horizontal lines drawn in correspondence to r_max and r_min.

Fig. 9.

The three curves are obtained by plotting 〈r_in〉 against x_in as computed from Eq. (6) for S₁_Zn (red), S₂_Zn (blue), and S₃_Zn (green). Full dots correspond to the values given in Table 3. Horizontal black lines are drawn in correspondence to the r_max and r_min values computed from Eq. (9). Vertical, red and blue lines are drawn from the point where each curve intersects the horizontal black lines, thus limiting the interval of acceptable x_in values

Continuous lines are obtained from Eq. (6), whereas full dots indicate the x_in values corresponding to the four M_n models. The vertical lines identify the region of “acceptable” x_in values for sample S₁_Zn (red) and S₂_Zn (blue) according to our criteria. In particular x_in is found to lie between 0.18 and 0.56 for sample S₁_Zn and between 0.54 and 0.78 for sample S₂_Zn. These findings are in agreement with the values (0.30 and 0.58, respectively) one obtains for x_in by assuming that in both samples Zn is bound as in model M₁ (Table 3). It should be noted, however, that for sample S₂_Zn, the binding mode suggested in model M₂ cannot be completely excluded.¹⁰

With regard to sample S₃_Zn (green line), the BVS values are not informative, because, as we said above, the hypothesis of a fourfold coordination in Eq. (9) cannot be true if all Zn²⁺ is supposed to be in solution. On the other hand, assuming that in the buffer Zn is coordinated to six ligands, the very good value BVS_buffer = 2.2 ± 0.2 is obtained (last line of Table 3).

Discussion

In this paper we have shown that XAS measurements can be used to clarify the mechanism of Cu²⁺ affinity modulation induced by Zn²⁺ that was left open in Walter et al. (2007). We have found that Zn²⁺ modifies the Cu²⁺ coordination mode by directly interacting with the PrP peptide. This conclusion can be directly inferred from the fact that the XAS spectrum of sample S₁_Zn is definitely different from that of Zn²⁺ in buffer (Figs. 4 and 5).

The most natural hypothesis of the way Zn²⁺ interacts with the peptide is to assume that it directly binds His units. In fact, EPR results have shown that the presence of Zn²⁺ promotes the partial switch of Cu²⁺ coordination mode from component β, in which the Cu²⁺ is bound to four His units, to component α, in which only one His is bound to Cu²⁺ (see Walter et al. 2007 and Fig. 1) suggesting that Zn²⁺ competes for His binding.

Under the assumption that this picture is correct, one can compute the number of His units involved in Zn²⁺ binding. Careful analysis based on XAS data and (BVS) calculation leads to the conclusion that the number of Zn²⁺ bound His units is most probably one and certainly no larger than two (Fig. 9 and Table 3).

The analysis presented in the paper confirms the important result of Walter et al. (2007) according to which Cu²⁺ is never completely removed by Zn²⁺. This finding suggests that the Cu-His binding strength depends on the number of His units already bound to Cu²⁺. Results in this direction are obtained by the numerical computation carried out by Guerrieri et al. (2009), in which the effect of Gly deprotonation is shown to be crucial in determining the metal coordination mode.

In Fig. 10 we show a sketch of the structural modifications induced by adding Zn. In the first reaction (upper line) that occurs at low Cu²⁺ concentration in which only the β component is present, Zn²⁺ binds one or two His units, displacing Cu²⁺. The second reaction (lower line) shows that when Cu²⁺ concentration is sufficiently high to have only the α component, Zn²⁺ is no longer able to remove Cu²⁺ and no longer interacts with the peptide, thus having only very limited effect on the Cu²⁺ coordination mode.

Fig. 10.

Sketch of Zn²⁺ action at low (upper reaction) and high (lower reaction) Cu²⁺ concentration. The tetra-octa-repeat is drawn as a blue string. His units are shown explicitly. Green dot is Cu²⁺, red dot is Zn²⁺, and grey dots are water oxygens

We close with a comment concerning the interesting result reported by Walter et al. (2007), in which DEPC modification followed by mass spectroscopy was used to measure Zn²⁺ affinity for PrP peptides, spanning both the non-octa-repeat Cu²⁺ binding sites, PrP(90–114), and the octa-repeat region, PrP(60–91). Experiments have shown that, in the absence of Cu²⁺, Zn²⁺ only binds to the octa-repeat region with all four octa-repeats engaged in a stable Zn²⁺ to four-His units bonded configuration. The lack of Zn²⁺ binding to PrP peptides with fewer than four octa-repeats emerging in DEPC modification experiments may seem to contradict the result found here according to which, in the presence of Cu²⁺, a Zn²⁺ ion can be found to bind one His by partially displacing Cu²⁺. To reconcile simultaneous Cu²⁺ and Zn²⁺ uptake, Walter et al. (2007) proposed the coexistence in solution of two metal-bound PrP species: one with a single Zn²⁺ coordinated by four His residues and another in which each of the four available His units is bound to a Cu²⁺ ion. This picture, however, does not clarify the reason for the effect of Zn on the Cu coordination mode.

The results in this paper provide a rather different and simpler picture in which a single octa-repeat domain takes up both Cu²⁺ and Zn²⁺. Moreover, our findings suggest that Cu²⁺ may introduce a kink or local conformation that assists Zn²⁺ binding in a somewhat unusual 2-His coordination mode, as shown in Fig. 10. This kind of binding leads to a mechanism of cross-regulation between Cu²⁺ and Zn²⁺ that is worth further investigation in other biological systems (for example the Aβ-peptides involved in Alzheimer’s disease). This mechanism may, in fact, suggest the existence of an additional general strategy for fine regulation of metal binding to avoid cell damage.