|Home | About | Journals | Submit | Contact Us | Français|
Recent evidence suggests that the prion protein (PrP) is a copper binding protein. The N-terminal region of human PrP contains four sequential copies of the highly conserved octarepeat sequence PHGGGWGQ spanning residues 60–91. This region selectively binds Cu2+ in vivo. In a previous study using peptide design, EPR, and CD spectroscopy, we showed that the HGGGW segment within each octarepeat comprises the fundamental Cu2+ binding unit [Aronoff-Spencer et al. (2000) Biochemistry 40, 13760–13771]. Here we present the first atomic resolution view of the copper binding site within an octarepeat. The crystal structure of HGGGW in a complex with Cu2+ reveals equatorial coordination by the histidine imidazole, two deprotonated glycine amides, and a glycine carbonyl, along with an axial water bridging to the Trp indole. Companion S-band EPR, X-band ESEEM, and HYSCORE experiments performed on a library of 15N-labeled peptides indicate that the structure of the copper binding site in HGGGW and PHGGGWGQ in solution is consistent with that of the crystal structure. Moreover, EPR performed on PrP(23–28, 57–91) and an 15N-labeled analogue demonstrates that the identified structure is maintained in the full PrP octarepeat domain. It has been shown that copper stimulates PrP endocytosis. The identified Gly–Cu linkage is unstable below pH ≈6.5 and thus suggests a pH-dependent molecular mechanism by which PrP detects Cu2+ in the extracellular matrix or releases PrP-bound Cu2+ within the endosome. The structure also reveals an unusual complementary interaction between copper-structured HGGGW units that may facilitate molecular recognition between prion proteins, thereby suggesting a mechanism for transmembrane signaling and perhaps conversion to the pathogenic form.
A misfolded form of the prion protein (PrP)1 is responsible for the transmissible spongiform encephalopathies (TSEs), including mad cow disease (BSE) and Creutzfeldt–Jakob disease (CJD) in humans (1). Despite nearly 20 years of PrP research, the physiological function of this remarkable protein is only now becoming clear. Recent work suggests that the membrane-bound prion protein binds copper in its N-terminal domain (2–11). It has been proposed that PrP participates in the regulation of copper (2, 12, 13) and possibly modulates the concentrations of reactive oxygen species (14–16). A hallmark of the PrP N-terminal domain is the octarepeat region, residues 60–91 in human and Syrian hamster PrP, which is composed of four or more repeats of the fundamental eight residue sequence PHGGGWGQ (Figure 1). Most current work suggests that copper binding takes place within this region and also in an adjacent region between the octarepeats and PrP's globular C-terminal domain. The octarepeat domain and copper have been implicated in neurological disease. Humans with PrPs containing extra octarepeats are predisposed to CJD (17). Elimination of the octarepeat domain slows disease progression (18). Addition of copper to wild-type PrP confers protease resistance, a property also characteristic of the pathogenic form (19).
There are ongoing efforts to determine how the octarepeat domain binds copper. The cellular, GPI-anchored, prion protein (PrPC, Figure 1) exists predominantly on the extracellular membrane surface (20) where exchangeable copper is in the divalent oxidation state (21, 22). Mass spectrometry studies performed at an extracellular pH of approximately 7.4 have demonstrated that peptides corresponding to the full octarepeat domain bind four Cu2+ ions with micromolar affinity (2, 7), with significant cooperativity (2) and select for this species over other divalent metal ions (2). These findings have been confirmed in full-length PrP (23). Raman studies have suggested that each PHGGGWGQ octarepeat binds a single Cu2+ by a nitrogen from the His imidazole side chain and deprotonated amide nitrogens from the second and third glycines following the His (8). Alternatively, circular dichroism (CD) and electron paramagnetic resonance (EPR) studies on the peptides PHGGGWGQ and HGGG led to the conclusion that coordination within an octarepeat arises from the His imidazole and three amide nitrogens all within the HGGG unit (5).
In a study combining peptide design, CD, multifrequency EPR, and electron spin–echo envelope modulation (ESEEM) spectroscopy, we recently demonstrated that the peptide HGGGW bound to Cu2+ gave spectroscopic signatures that were indistinguishable from those of Cu2+-bound multioctarepeat sequences (4). EPR and CD titration experiments demonstrated a rigorous 1:1 Cu2+/octarepeat binding stoichiometry regardless of the number of octarepeats within a sequence. These findings led us to conclude that HGGGW comprises the fundamental Cu2+ binding unit. By comparing the EPR of HGGGW with that of HGG, we further concluded that nitrogen coordination was from the His imidazole and deprotonated amides from the next two glycines. Although the Trp in HGGGW was found to be essential in providing the proper coordination environment, it was not clear how this residue interacted with the Cu2+.
Determining the precise Cu2+ coordination environment within the octarepeat domain is essential for advancing the understanding of PrP's role as a copper binding protein. Here we report the X-ray crystal structure of HGGGW in complex with Cu2+. In addition, we determine whether the coordination environment identified in the crystal structure holds in solution for HGGGW, PHGGGWGQ, and the PrP octarepeat domain PrP(23–28, 57–91) using ESEEM, HYSCORE, and low-frequency S-band EPR. A library of 15N-labeled PrP peptides was used in these EPR experiments to assign specific Cu–N couplings. These combined crystallographic and EPR studies provide the first atomic resolution view of the copper binding site within an octarepeat. In addition, the structural features allow us to hypothesize how PrP functions as a copper sensor or transporter, how octarepeats participate in transmembrane signaling, and how copper-mediated PrP recognition contributes to the formation of PrPSc.
HGGGW, the octarepeat (PHGGGWGQ), and PrP(23–28, 57–91) were prepared with an acetyl group at the N-terminus and amidated at the C-terminus. Methods for synthesis, purification, and characterization have been described previously (4). N-Fmocglycine(15N, 98+%) was obtained from Cambridge Isotope Laboratories, Inc.
HGGGW in complex with Cu2+ was crystallized by the hanging-drop vapor diffusion method by the addition of a 2 μL sample solution (15 mg/mL, pH 7.4) to 2 μL of a reservoir solution containing 100 mM HEPES at pH 7.4 and 10–30% Jeffamine M-600. Crystals having dimensions 0.1 mm by 0.2 mm by 0.03 mm grew within 24 h. For cryoprotection, crystals were soaked in paratone and flash-cooled in a stream of liquid nitrogen. Crystals were mounted in the 92 K nitrogen cold stream provided by a CRYO Industries low-temperature apparatus on the goniometer head of a Bruker SMART 1000 diffractometer. Diffraction data were collected with graphite-monochromated Mo Kα radiation employing a 0.3° ω scan and approximately a full sphere of data to a resolution of 0.68 Å. An empirical correction for absorption was applied. A total of 10 657 reflections were collected, of which 8394 were unique [R(int) = 0.045] and 7581 were observed [I > 2σ(I)]. The structure was solved by direct methods and refined by full-matrix least-squares on F2. All non-hydrogen atoms were refined with anisotropic thermal parameters. Hydrogen atoms were located on a difference map. The maximum and minimum peaks in the final difference Fourier map corresponded to 0.64 and −0.53 eÅ−3, respectively. Crystal Data: C25H41CuN9O12, MM = 723.21, triclinic, P1, a = 8.7641(6) Å, b = 9.4145(7) Å, c = 10.8450(8) Å, α = 82.840(7)°, β = 79.744(7)°, γ = 62.342(7)°, Z = 1. The refinement converged with a wR2 value of 0.082 using all data and an R1 value of 0.037 for the observed data using 431 parameters. Programs used were the SHELXTL 5.10 crystallographic software suite (Sheldrick, G. M.; Bruker Analytical X-ray Instruments, Inc.).
All samples were prepared with degassed buffer containing 25 mM N-ethylmorpholine (NEM), 150 mM potassium chloride (KCl), and 20% glycerol (v/v) where the glycerol served as a cryoprotectant. 63Cu (99.62%, Cambridge Isotope Laboratories) was used to avoid inhomogeneous broadening of the S-band EPR lines that would otherwise be present with the mixture of naturally occurring isotopes. S-band spectra (3.5 GHz) were acquired in D2O solution at 133 K using a loop gap resonator as part of a specially designed spectrometer housed at the Biomedical ESR Center at the Medical College of Wisconsin. The mI = −1/2 lines of these spectra were simulated using stick diagrams with Gaussian line shapes. Three pulse ESEEM measurements were obtained at 4.2 K on an X-band pulsed-EPR spectrometer. The instrument, cavity, and resonator were constructed in-house and have been previously described (24). Data were obtained at g, the point of greatest spectral intensity (3280 G at 9.47 GHz). Data processing to attain frequency domain spectra for 3-pulse ESEEM was carried out using software described in previous work (25). The HYSCORE pulse sequence was π/2-τ-π/2-t1-π-t2-π/2 (26). Data were processed using the software package WINEPR (Bruker). Experimental parameters for all pulsed experiments are provided in the figure legends.
It has been demonstrated that the suboctarepeat segment HGGGW possesses all the functional groups that directly coordinate Cu2+ and thus constitutes the fundamental copper ion binding site (4). The X-ray crystal structure (0.7 Å resolution) of HGGGW–Cu2+ was determined from crystals grown in pH 7.4 aqueous solution. Each unit cell contains a single HGGGW–Cu2+ complex along with six ordered water molecules. Figure 2a shows a stereo representation of the HGGGW–Cu2+ complex crystal structure with the electron density (blue) contoured at 2 σ. Additional density found by difference maps contoured at 3.5 σ reveals ordered hydrogens on the axially bound water (white). (Density arising from other ordered water molecules was omitted for clarity.) The detailed structure, shown in Figure 2b, reveals that the Cu2+ possesses a pentacoordinate environment with equatorial ligation from the δ1 nitrogen of the His imidazole and deprotonated amide nitrogens from the next two Gly residues. The second Gly also contributes its amide carbonyl oxygen. With the exception of the His backbone nitrogen and alpha carbon, all atoms from the His through to the nitrogen of the third Gly lie approximately in the equatorial plane and the copper is just above this plane, as consistent with a pentacoordinate complex.
The Trp indole also participates in the coordination environment, but in a rather unusual fashion. There is a water molecule axially bound to the copper. The indole nitrogen from the Trp side chain is 3.0 Å from the oxygen of this bound water, suggesting a hydrogen bond as indicated by a dashed line. This arrangement places the plane of the indole ring above the copper such that it is nearly parallel with the equatorial plane. Two additional water molecules hydrogen bond to the axial water forming a network extending from the backbone carbonyl preceding the His to the carbonyl of the third Gly.
The crystal structure also reveals hydrogen bond contacts between copper binding units as shown in Figure 2c for four HGGGW–Cu2+ units. The NH groups from the imidazole and Trp backbone position to form complementary hydrogen bonds to the backbone amide carbonyls of the His and the first Gly of the neighboring HGGGW segment, creating stable contacts between structured HGGGW–Cu2+ units.
The constructs used for EPR studies are listed in Table 1. HGGGW is the fundamental Cu2+ binding unit, PHGGGWGQ is the consensus octarepeat, and PrP(23–28, 57–91) contains the full octarepeat domain (Figure 1). The string of basic residues KKRPKP, corresponding to PrP(23–28), is included in this latter construct to improve solubility. S-band EPR along with 15N scanning was used to evaluate whether the equatorial coordination environment identified in Figure 2 is preserved in low-temperature aqueous solution for HGGGW and the PHGGGWGQ octarepeat. The benefit of S-band EPR arises from a partial cancellation of g-strain- and A-strain-induced inhomogeneous broadening specifically for the 63Cu mI = −1/2 hyperfine line (27). The mI = −1/2 line of the low-frequency S-band (3.4 GHz) EPR spectrum reveals superhyperfine couplings from nitrogens directly coordinated to Cu2+ centers as seen in Figure 3 (27, 28). We previously demonstrated that the couplings seen here in the top spectra are consistent with three 14N couplings, in agreement with the equatorial coordination geometry of Figure 2 (4). Because15N possesses a different nuclear spin (I = 1/2) than 14N (I = 1), site-specific 15N labeling can be used to identify the directly coordinated nitrogens by an observed change in multiplet structure.
A series of HGGGW peptides was prepared with each member of the series 15N-labeled at a unique glycine amide (Table 1). S-band EPR were obtained from samples in D2O solution at measured pH = 7.4. Figure 3 shows the family of spectra down each column where the 15N substitution is at the underlined glycine. The vertical lines drawn from the most prominent features of the spectra obtained from the unlabeled peptides are included to guide the eye and emphasize changes in multiplet patterns. For HGGGW, a change in multiplet structure relative to the unlabeled species is seen only for the first and second glycines, demonstrating that the nitrogens of these residues directly coordinate to the copper center, which is consistent with the crystal structure. Interestingly, comparison of the spectra of HGGGW and HGGGW reveals slightly different coupling patterns, with the latter giving a wider separation among the most prominent lines. The origin of this difference is unclear but may reflect unequal Cu–N bond lengths from the first and second glycines to the copper center. Nevertheless, the qualitative change in multiplet patterns among the 15N-labeled constructs demonstrates conclusively that only the first and second glycines directly coordinate to the copper center.
Simulations were employed using a previously described model (4) to determine whether the observed changes in multiplet structure with 15N labeling are consistent with the hyperfine coupling constants determined previously. Although the 15N scanning data here may suggest a slight inequivalence for the ligated nitrogen couplings, we previously showed that a simple model of three equivalent nitrogens (aN = 13 G) and a single hydrogen (aH = 10 G) is sufficient to reproduce the S-band EPR data (4) as shown in Figure 3. To incorporate the effect of 15N labeling, one nitrogen spin was changed to I = 1/2 along with rescaling of the nuclear g value. The resulting simulation is shown in Figure 3 and indeed predicts the hyperfine structure observed for HGGGW and HGGGW.
All four glycines were individually labeled in a series of octarepeat peptides (Table 1), and the results are shown in Figure 3. Again, a change in multiplet structure is observed only for the first two glycines. The spectra shown in Figure 3 confirm that the nitrogen contacts to the copper identified for HGGGW are preserved in the octarepeat segment.
Finally, several attempts were made to obtain superhyperfine couplings from S-band spectra of PrP(23–28, 57–91). In all cases, no couplings were resolved. The mI = −1/2 line was slightly broader than that obtained from the octarepeat and HGGGW, suggesting that inhomogeneous broadening was masking the desired multiplet pattern. Fully loaded PrP(23–28, 57–91) contains four Cu2+ ions (see Discussion), and, hence, weak dipolar interactions between the paramagnetic centers or structural inhomogeneities may contribute an additional source of inhomogeneous broadening that is not canceled in the mI = −1/2 line at S-band.
ESEEM spectroscopy is a pulsed EPR technique that is sensitive to spin-active nuclei within approximately 10 Å of the paramagnetic copper center (29). For example, the method readily identifies imidazole coordination by detection of the remote nitrogen (30). In its application here, ESEEM serves as a complement to S-band EPR which detects directly coordinated nitrogens. ESEEM spectra were obtained for HGGGW and PHGGGWGQ, each with a single bound Cu2+, and for PrP(23–28, 57–91) with four bound Cu2+ at pH = 7.4 and are shown in Figure 4a. These spectra reveal typical features that have been previously assigned to imidazole coordination (4, 30), consistent with the crystal structure reported here. There is an additional set of peaks with prominent features, in particular, at 2.0 and 2.8 MHz. To assign these peaks, we used the library of 15N-labeled peptides (Table 1). The ESEEM spectrum of HGGGW is shown in Figure 4b and reveals only those peaks assigned to the imidazole. [With the experimental conditions used here, 3-pulse ESEEM amplitudes of the quadrupolar 14N nucleus are much greater than those of the dipolar 15N nucleus (31) and thus peaks associated with this latter species are not observed in Figure 4.] All other 15N-labeled HGGGW peptides gave spectra equivalent to that observed from the unlabeled species (data not shown). These observations demonstrate that the 14N of the third glycine [Gly(3)] is responsible for the additional set of ESEEM peaks in HGGGW. To test this assignment in the full PrP octarepeat domain, a labeled analogue of PrP(23–28, 57–91) was prepared with 15N-glycine at the Gly(3) position of all four octarepeats (specifically, PrP sequence positions 64, 72, 80, and 88; Table 1). ESEEM of PrP(23–28, 57–91, 15N-64, 72, 80, 88) with four bound Cu2+ is shown in Figure 4b and is found to be equivalent to that obtained from HGGGW and reveals only those peaks assigned to the imidazole. In turn, this demonstrates that the nitrogen of Gly(3) is not coordinated to the copper center either in HGGGW or in the corresponding positions in PrP(23–28, 57–91), in agreement with the crystal structure (Figure 2).
The low-frequency lines at 0.57, 0.90, and 1.47 MHz, which were previously assigned to the remote nitrogen of the imidazole (4), are characteristic of a weakly coupled 14N nucleus close to the exact cancellation limit. This limit results from the cancellation of the vector sum of the 14N nuclear Zeeman and hyperfine fields in one of the electron spin manifolds, and gives rise to 14N quadrupolar transitions (denoted NQI for nuclear quadrupole interaction) (30). The ESEEM frequencies for these transitions are given by the following equations:
where q is the z-component of the electric field gradient across the nucleus, Q is the 14N quadrupole moment, η is the asymmetry parameter describing the x- and y-components of the electric field gradient across the nucleus, h is Planck's constant, and e is the electron charge.
In addition, ESEEM peaks are evident at ~0.8, 2.0, and 2.8 MHz. From 15N scanning experiments, it is clear that these result from coupling to the noncoordinated amide nitrogen of Gly(3). Use of eq 1 to determine NQI parameters (e2qQ/h and η) is valid only if these peaks can be considered predominantly quadrupolar in nature, e.g., the nucleus is at or near the condition of exact cancellation: vI = Aiso/2, in which vI is the nuclear Larmor frequency and Aiso is the isotropic component of the hyperfine interaction (30). A property of quadrupole transitions near exact cancellation is that their frequency is largely insensitive to applied magnetic field strength (and corresponding EPR excitation frequency) (32, 33). We investigated the field/EPR frequency dependence of these Gly(3) peaks by registering ESEEM patterns at 8.9 and 10.5 GHz (data not shown) in addition to the 9.47 GHz spectra shown in Figure 4. The low-frequency peaks assigned to Gly(3) exhibited little or no field dependence, indicating that their frequencies are primarily determined by the (field-independent) nuclear quadrupolar interaction. Use of eq 1 to determine e2qQ/h and η yields values of 3.20 MHz and 0.50, respectively. Correlations observed in HYSCORE experiments (see below) provide further confirmation of this assignment and lead to an Aiso value of 1.41 MHz for this 14N nucleus. Although this means that the condition of exact cancellation has not been precisely achieved at this magnetic field value (3280 G), it has been shown previously that ESEEM spectra of I = 1 nuclei display properties indicative of exact cancellation provided deviations from exact cancellation (defined by Δ = |2vI − Aiso|) are less than e2qQ/3h (32). For the spectra of the amide 14N considered here, Δ = 0.61 MHz < e2qQ/3h = 1.06 MHz. This confirms the assignment of the spectra as at approximate exact cancellation and, together with the multifrequency and HYSCORE experiments, justifies the assignment of the ~0.8, 2.0, and 2.8 MHz peaks as quadrupolar and the use of eq 1 in determining e2qQ/h and η.
The ESEEM spectrum of HGGGW (Figure 4b) reveals those transitions due solely to the remote nitrogen of the imidazole and thus allows for a more accurate determination of its quadrupolar parameters. The transitions and corresponding e2qQ/h and η values are given in Table 2 and represent only a slight adjustment over that reported previously (4). The quadrupolar parameters are fully consistent with a remote nitrogen in the protonated state (30).
The quadrupolar parameters for the amide nitrogen of the third glycine in HGGGW were determined from the well-resolved v− and v+ peaks and are reported in Table 2. The values e2qQ/h = 3.2 MHz and η = 0.5 are very close to those observed for amide nitrogens of other O-coordinated glycine–metal ion complexes as determined by NQR spectroscopy (34). For example, the O-coordinated amide in the chloride and bromide salts of the CdII(Gly-Gly)-(imidazole) complex give quadrupolar parameters of e2qQ/h = 3.22 MHz, η = 0.474 and e2qQ/h = 3.20, η = 0.483, respectively. They are also in agreement with quadrupole parameters assigned to an amide nitrogen in azurin using ENDOR and ESEEM at 95 GHz (35) and HYSCORE at 9 GHz (36).
The two nitrogens giving rise to the 14N ESEEM spectra are rigidly fixed in place by the coordination geometry shown in Figure 2. According to the crystal structure, the remote nitrogen of the imidazole and the amide nitrogen of the third glycine are 4.12 and 4.09 Å from the Cu2+, respectively. The equivalence of the ESEEM spectra from HGGGW, the octarepeat, and PrP(23–28, 57–91) and their 15N-labeled analogues indicates that the electronic and molecular structure of the environment of the two 14N nuclei are similar for all three peptides. This finding, in turn, suggests that the equatorial coordination environment identified by the crystal structure is preserved in HGGGW, in the octarepeat, and in PrP(23–28, 57–91).
HYSCORE experiments allow for enhanced resolution of spectra complicated by the presence of multiple nuclei or significant anisotropic line broadening (26). The HYSCORE experiment represents an important orthogonal technique for confirming the relevance of the X-ray crystal structure to copper-bound HGGGW in solution. As was stated earlier, ESEEM suggests a common equatorial ligand sphere about the central copper. As an extension of the spin–echo experiment, HYSCORE provides a means to directly measure anisotropic hyperfine interactions and more definitively assign transitions to their parent nuclei (37).
In the HYSCORE sequence, the addition of a π pulse between the second and third π/2 pulses introduces correlations between nuclear spin coherences of opposing manifolds which, for the I = 1 exact cancellation case, links the NQI transitions of one electron spin manifold with the higher frequency double quantum transition in the opposite manifold (36). In addition, the appearance of an echo-modulation echo allows detection of anisotropically broadened ESEEM lines as observed for dipolar-coupled I = 1/2 species (26, 37, 38).
The HYSCORE spectra in Figure 5 were obtained with the same HGGGW samples studied in Figure 4. Figure 5a shows the spectra of the natural isotope HGGGW. Here, cross-peaks are observed between the NQI and DQ transitions for both the remote nitrogen of imidazole and the 14N of Gly(3) previously assigned in 3-pulse ESEEM spectra. Those peaks assigned to the Gly(3) 14N-amide in the ESEEM experiment are maintained in HYSCORE, and correlations between NQI and DQ lines in this experiment are consistent with the exact cancellation assignment (Figure 5a). Further, in the 15N-substituted peptide HGGGW (Figure 5b), a new set of cross-peaks at 0.47 and 2.43 MHz are observed and replace those at 2.85 and 4.50 MHz. We assign these peaks, which are not resolved in 3-pulse ESEEM, to the I = 1/2 15N nucleus of Gly(3). The theoretical treatment of I = 1/2, S = 1/2 HYSCORE spectra appropriate for this case has been presented by Dikanov et al. (39). Assuming that the anisotropic component of the hyperfine interaction between an I = 1/2 nucleus and an S = 1/2 metal electron may be adequately described by the point dipole approximation, the hyperfine tensor for this system is axial with principal values of (−T, −T, 2T), and the shape of the cross-peaks is described by the equations:
Aiso and T are, respectively, the isotropic and anisotropic components of the hyperfine coupling constant, vI is the nuclear Larmor frequency, vα and vβ are nuclear frequencies of α and β spin manifolds, respectively, determined from the HYSCORE spectra, and Q and G are derived from the cross-peak contour line shapes. Analysis of the vα, vβ cross-peaks of Figure 5b yields Q = −0.205 and G = 1.39. Solving eqs 2–4 with vI = 1.42 MHz yields |T| = 0.18 MHz and Aiso(15N) = 1.96 MHz (corresponding to Aiso = 1.41 MHz for 14N). These values together with the quadrupole parameters obtained from the 14N data (see above) are consistent with the assignment to a nonligated amide nitrogen as found in the crystal structure (40).
The spectral region near the proton Larmor frequency for the sample HGGGW–Cu2+ is displayed in Figure 5c. The peak observed arises from a proton(s) experiencing a predominantly dipolar interaction with Aiso ~ 0 and T ~ 2 MHz. Peaks assigned to protons of equatorially coordinated water have been clearly observed in HYSCORE spectra of copper complexes (38, 41). These peaks are typically quite intense and span a 6–8 MHz range perpendicular to the diagonal. These HYSCORE studies, together with previous ENDOR studies (42), measure T = 4.8–5.2 MHz and Aiso ≈ 0.7–2.2 MHz for protons of equatorially coordinated water. We observe no peaks in Figure 5c or any other data which can be assigned to equatorially ligated water. This is consistent with the crystal structure of copper-bound HGGGW in which the equatorial ligands consist of three nitrogen atoms and one carbonyl oxygen. The protons of axially ligated water have been studied by ENDOR and assigned coupling values of T = 3.7 MHz and Aiso < 0.2 MHz (42). The peak in Figure 5c is eliminated when the sample is prepared in D2O, indicating it is solvent-exchangeable. It is possible that it may arise from an axially coordinated water, as observed in the crystal structure, but a definite assignment is difficult to make because other weakly coupled solvent-exchangeable protons may contribute to this peak. The main conclusion drawn from Figure 5c is that no equatorial water ligation is observed in the HGGGW–Cu2+ solution structure, in agreement with the crystal structure.
We have obtained the crystal structure of the copper binding site within the HGGGW segment of the PrP octarepeat domain. The copper is coordinated by the His imidazole and deprotonated amides from the next two glycines within the HGGGW segment. The Trp indole also participates through a hydrogen bond to the axially coordinated water molecule. Low-frequency S-band EPR along with a library of 15N-labeled peptides demonstrates that in HGGGW and PHGGGWGQ the first two glycines directly coordinate to the Cu2+, consistent with the crystal structure. ESEEM and HYSCORE spectroscopy reveal the expected pattern for His imidazole coordination as well as an additional coupled 14N which, using the 15N-labeled peptides, is assigned to the noncoordinated amide nitrogen of the third glycine in HGGGW and the corresponding third glycine of each octarepeat in PrP(23–28, 57–91). These site-specific couplings revealed by the ensemble of EPR experiments on HGGGW, PHGGGWGQ, and PrP(23–28, 57–91) are fully consistent with the crystal structure of Figure 2. Within the copper binding segment HGGGW, our studies show that only Gly(3) does not directly interact with the Cu2+ or its axial water. Interestingly, examination of the octarepeat sequences (e.g., Figure 1) shows that this is the only position that exhibits sequence variability (43).
In previous work, we found that Cu2+-bound complexes of HGGGW and PrP(23–28, 57–91) X-band EPR spectra were almost indistinguishable from each other. In addition, titrations showed that HGGGW bound a single Cu2+ whereas PrP(23–28, 57–91), with four HGGGW repeats, bound approximately four Cu2+ ions. Analysis of the X-band EPR signal integrals further showed that there was no detectable magnetic coupling among the four Cu2+ in the fully bound PrP(23–28, 57–91). These findings lead to the proposal that each individual octarepeat binds a single Cu2+ within the HGGGW unit. Indeed, the crystal structure reported here demonstrates that HGGGW provides a five ligand coordination environment as often found for Cu2+ complexes. When considered along with the identical site-specific couplings for HGGGW, the octarepeat, and PrP(23–28, 57–91) as revealed by EPR here, these findings make a compelling argument for a structure where the fully copper-loaded octarepeat domain binds each Cu2+ in an HGGGW segment in a fashion equivalent to that of Figure 2.
Recent findings have led to speculation that PrP is a metal transporter. Pauli et al. found that copper stimulates the endocytosis of PrP (12). Deletion of the octarepeat domain eliminates this function (44). Brown has demonstrated that wild-type cells import more copper than do PrP knockouts (45). While we present no direct evidence to support the copper transport function for PrP, the steep pH-dependent binding of copper in the range of 6.0–7.4 (2, 4), which represents the milieu of PrP in the endosome and on the cell surface, respectively, could indicate that the protein functions as a structural or biochemical switch, transducing information based on the level of bound copper and, in turn, on the concentration in the extracellular matrix. Interestingly the relatively weak affinity for copper (micromolar) may be akin to the constitutive glucose transporter whose Kd is on the order of the blood glucose concentration and is thus exceptionally sensitive in that range. That is, PrP may act as a cuprostat, activating cellular machinery in response to changes in ambient copper levels.
The specific PrP sequence HGGGW is highly conserved across species (43) and is found primarily within the prion family of proteins. The three-dimensional structure reported here represents a newly identified copper binding site. Previous structural work has demonstrated that the N-terminal domain of PrPC, corresponding approximately to PrP(23–90), is unstructured in the absence of Cu2+ (46). The studies presented here suggest that in the physiological milieu, however, this region provides a novel Cu2+ binding domain that functions to bind the metal ion in a pH-dependent manner, so that the affinity is switched between the extracellular and endosomal pH range. The structure of the prion copper binding site provides a structural basis for the putative pH-dependent molecular switch that governs the uptake and release of Cu2+. The glycine amide–Cu linkages are highly pH-sensitive and release the metal ion below pH 6.5–7.0 (47). Mass spectrometry studies reveal that indeed PrP(57–91), which contains four octarepeats, nevertheless binds only two copper ions with high affinity (Kd < 5 μM) at pH 6.0 (2). Pulsed EPR studies of PrP(23–231) at pH = 5.6 identify only imidazole coordination and no evidence of amide backbone coordination as observed here at extracellular pH 7.4 (6). The combined physiological and structural studies suggest a mechanism for PrP's function as a copper sensor or transporting protein that functions through endocytosis as shown schematically in Figure 6a (2, 12). This mechanism is similar to that for iron transport by transferrin and the transferrin receptor (48). In the prion protein, however, the metal ion binding domain (the octarepeats) and the membrane-associated domain (the globular C-terminus with GPI anchor) are integrated into a single protein. Currently, there is no direct structural information as to how the octarepeat domain organizes at low endosomal pH or at low copper concentration. However, Viles et al. have suggested that at low [Cu2+], copper ions are bound by two His imidazoles, thereby linking separate octarepeat segments (11).
Examination of the intermolecular contacts found in the crystal structure reveals a potentially important docking interaction between HGGGW–Cu2+ units that may explain PrP's cooperative copper binding. Figure 2c (see Results) shows four adjacent copper binding units as they are found in the crystal and illustrates how the fully copper-loaded octarepeat domain might be organized in the protein. To accommodate this type of octarepeat domain structure, the intervening -Gly-Gln-Pro- sequence must reach from the C-terminus of one HGGGW unit to the N-terminus of the next. Preliminary modeling studies suggest that this copper-induced stabilizing interaction can be achieved with sterically allowed backbone conformations. While we find that such a model is reasonable based on steric arguments, spectroscopic data suggest caution. The low-frequency components (<2.0 MHz) of imidazole ESEEM spectra are sensitive to the hydrogen bonding environment of the remote nitrogen (49). The similarity of these low-frequency features among the PrP segments suggests that the remote nitrogen experiences a hydrogen bonding environment that is equivalent in HGGGW, the octarepeat, and PrP(23–28, 57–91). Consequently, if PrP(23–28, 57–91) orders as suggested by Figure 2c, then buffer may act as the imidazole NH hydrogen bond acceptor for HGGGW and the octarepeat. Alternatively, HGGGW–Cu2+ units within PrP(23–28, 57–91) may organize in a yet unknown fashion that is distinct from that observed in the crystal.
It is noteworthy that within the octarepeat domain, glutamine is the only side chain possessing a functional group that does not participate in copper binding. Yet, this residue is highly conserved among mammalian PrPs (Figure 1). This observation may contribute to the understanding of PrP function. Recent work suggests that proteins with Gln/Asn-rich regions participate in regulatory processes through protein–protein contacts (50). Participation of octarepeat glutamine suggests an intermolecular recognition mechanism for PrPC analogous to the glutamine zipper-mediated aggregation in Huntington's disease (51, 52). When the octarepeat domain of PrPC orders in the presence of copper, perhaps as suggested by the contacts within the crystal shown in Figure 2c, the glutamines are brought into close proximity to one another within the PrP N-terminal domain between HGGGW–Cu2+ units. These glutamine side chains may facilitate recognition between membrane-bound prion proteins through noncovalent interactions that stimulate endocytosis as shown in Figure 6b. Indeed, antibody cross-linked PrPs on cultured cells stimulate signal transduction (53), and it has been proposed that copper-induced oligomerization is a feasible mechanism for endocytotic trafficking of PrP (19). Glutamine cross-linked PrPs might occasionally allow interaction between the adjacent sequence PrP(90–120) on nearby PrPs as implicated in the formation of pathogenic PrPSc (54). Such a templating mechanism is consistent with that previously proposed in a study where additional octarepeats in some human family lines were found to confer predisposition to CJD (17). This concept may also help explain how copper converts PrPC to a protease-resistant form (19).
The studies here have focused on copper binding in the octarepeat domain. However, recent work suggests that PrP possesses additional Cu2+ binding sites. Recombinant murine PrP with four octarepeats, nevertheless, binds five copper ions (23). Fluorescence and NMR experiments suggest an additional binding site involving histidines 96 and 111 in human PrP (55), and XAFS of PrP(91–231) supports the assignment of copper coordination by two histidines as well as a sulfur which is assigned to Met109 (56). In addition, aggregation of the neurotoxic peptide PrP(106–126) is highly metal ion dependent (57). How additional copper binding sites influence PrP function and copper binding in the octarepeat domain is yet to be determined.
We are grateful to Professors H. Ball, F. E. Cohen, and S. B. Prusiner (UCSF) for helpful discussion. In addition, the peptide PrP(23–28, 57–91), originally designed by H. Ball, was instrumental for obtaining high-quality EPR spectra of the octarepeat domain.
†This work was supported by NIH Grants GM65790 (G.L.M.), GM60609 (G.J.G.), and GM40168 (J.P.) and by NSF Grant MCB-0090994 (W.G.S.).
1Abbreviations: PrP, prion protein; PrPC, cellular isoform of PrP; PrPSc, scrapie isoform of PrP; PrP(57–91), residues 57 through 91 of PrP; EPR, electron paramagnetic resonance; ESEEM, electron spin–echo envelope modulation; HYSCORE, hyperfine sublevel correlation spectroscopy; e2qQ/h, maximum quadrupole coupling; η, quadrupole asymmetry parameter; NQI, nuclear quadrupole interaction; DQ, double quantum; Aiso, isotropic component of the hyperfine coupling constant; T, anisotropic component of the hyperfine coupling constant; CD, circular dichroism; NMR, nuclear magnetic resonance; XAFS, X-ray absorption fine structure; GPI, glycosylphosphatidylinositol; NEM, N-ethylmorpholine; Kd, dissociation constant.