Skip to main content
Proceedings of the National Academy of Sciences of the United States of America logoLink to Proceedings of the National Academy of Sciences of the United States of America
. 2007 Aug 8;104(33):13313–13318. doi: 10.1073/pnas.0702669104

Engineering metal ion coordination to regulate amyloid fibril assembly and toxicity

Jijun Dong *, Jeffrey M Canfield , Anil K Mehta *, Jacob E Shokes , Bo Tian §, W Seth Childers *, James A Simmons *, Zixu Mao §, Robert A Scott , Kurt Warncke *,, David G Lynn *,
PMCID: PMC1948904  PMID: 17686982

Abstract

Protein and peptide assembly into amyloid has been implicated in functions that range from beneficial epigenetic controls to pathological etiologies. However, the exact structures of the assemblies that regulate biological activity remain poorly defined. We have previously used Zn2+ to modulate the assembly kinetics and morphology of congeners of the amyloid β peptide (Aβ) associated with Alzheimer's disease. We now reveal a correlation among Aβ-Cu2+ coordination, peptide self-assembly, and neuronal viability. By using the central segment of Aβ, HHQKLVFFA or Aβ(13–21), which contains residues H13 and H14 implicated in Aβ-metal ion binding, we show that Cu2+ forms complexes with Aβ(13–21) and its K16A mutant and that the complexes, which do not self-assemble into fibrils, have structures similar to those found for the human prion protein, PrP. N-terminal acetylation and H14A substitution, Ac-Aβ(13–21)H14A, alters metal coordination, allowing Cu2+ to accelerate assembly into neurotoxic fibrils. These results establish that the N-terminal region of Aβ can access different metal-ion-coordination environments and that different complexes can lead to profound changes in Aβ self-assembly kinetics, morphology, and toxicity. Related metal-ion coordination may be critical to the etiology of other neurodegenerative diseases.

Keywords: copper-binding, neurotoxicity, self-assembly


Protein intermolecular assembly, especially formation of amyloid fibrillar structures, is correlated with a variety of human neurodegenerative diseases, including Alzheimer's, Parkinson's, Huntington's, and Creutzfeldt–Jakob diseases (1). More recently, amyloid has been tied to many nonpathological functional roles. For example, formation and self-perpetuation of amyloids in Saccharomyces cerevisiae regulate diverse yeast phenotypic expression as a positive response to environmental fluctuations (2), and amyloid may be involved in long-term memory and synapse maintenance in the marine snail, Aplysia (3, 4). Many proteins, including archetypical globular proteins such as myoglobin, can also form amyloid fibrils, suggesting that amyloidogenesis may be an intrinsic property of any α-amino acid polymer (5). Accordingly, these highly ordered paracrystalline protein self-assemblies have now been recognized as useful for nanostructure fabrication and biotechnology (68). Fully capturing these technological opportunities and understanding the biological roles of amyloid will depend on further definition of the organized structure and assembly pathway.

Increasing evidence now implicates transition metal ions, including Zn2+, Cu2+, and Fe3+, as contributors both to amyloid β (Aβ) assembly in vitro and to the neuropathology of Alzheimer's disease, AD (9). The obligatory region of metal ion (Zn2+/Cu2+) binding of Aβ has been mapped to the N terminus, amino acids 1–28 (1016). In its soluble nonamyloid conformation, the peptide contains multiple intramolecular binding sites for Zn2+ and Cu2+ (9, 17), and intermolecular Zn2+ binding can promote Aβ aggregation (14, 18, 19). We have previously examined the role of these intermolecular Zn2+-binding sites in the truncated peptide HHQALVFFA-NH2, Aβ(13–21)K16A. This peptide contains the His-13/His-14 dyad previously implicated in metal binding (14, 15, 19) and the core hydrophobic sequence, LVFFA (20, 21) that is crucial for Aβ assembly. This short peptide self-assembles in vitro into typical amyloid fibrils that are morphologically similar to the full-length Aβ peptide (22). Zn2+ ions accelerate the assembly by means of coordination with two imidazole side chains from different peptide molecules. These two peptide molecules could be within a single β-sheet (intrasheet coordination), or come from adjacent sheets (intersheet coordination). Different metal coordination structures result in distinct self-assembled morphologies, ranging from typical amyloid fibrils to twisted ribbons and homogeneous nanotubes (22).

Complex coordination environments for Cu2+ appear to exist in the longer Aβ(1–40) peptide. An intermolecular His residue-bridging binding site of Cu2+ in the amyloid fibril (14), similar to Zn2+-bridged His coordination (14, 18, 19, 22) has been indicated, whereas other results support an intramolecular Cu2+–Aβ complex existing in both soluble and fibrillar Aβ(1–40) (23). As both inhibitory and fibril-inducing activities have been reported for Cu2+ (12, 24, 25), it is very likely that the observed differences in Cu2+ coordination might be directly responsible for the differences in kinetics and morphologies and, further, that these different Cu2+–Aβ coordination structures can be accessed under slightly different experimental conditions. Indeed, Aβ(1–40) is capable of aggregating into a range of structures under slightly different assembly conditions (26). To simplify the study of Cu2+-induced Aβ self-assembly, we prepared a series of homogeneous Aβ(13–21) complexes with both Zn2+ (22) and Cu2+ incorporation. The structural models that emerged from visible, infrared, x-ray absorption and electron paramagnetic resonance (EPR) spectroscopies reveal a molecular basis for the self-assembly characteristics of different Aβ(13–21) peptides. These models and initial neurotoxicity results implicate a critical role of metal ions and common structural features of amyloid assemblies in a range of neurological diseases.

Results and Discussion

Zn2+/Cu2+ and the Assembly of Aβ(13–21)K16A.

We considered Aβ(13–21), HHQKLVFFA, a minimum sequence for investigating the role of metal-ion His binding in amyloid fibril formation because it contains the His-13/His-14 dyad previously implicated in metal binding (14, 15, 19) and a core hydrophobic sequence, LVFFA (20, 21), crucial for Aβ assembly. To isolate His-13/14 as the sole metal-binding site, Lys-16 was replaced with Ala, giving Aβ(13–21)K16A, which assembles into typical amyloid fibrils, following the characteristic conversion from random coil into β-strand (22). Zn2+ greatly accelerates the self-assembly rate and induces either typical amyloid fibrils or twisted ribbons and nanotubes depending on metal/peptide stoichiometry (22). These different morphologies arise from either intra- or intersheet His–Zn2+–His complexation for the fiber and ribbons/nanotubes, respectively (22). In contrast, Cu2+ induces no CD β-signature in this peptide over extended time periods [see supporting information (SI) Fig. 6A], even in the presence of excess Zn2+ (data not shown), and no aggregates were pelleted when centrifuged at 16,110 × g for 30 min (SI Fig. 6B), supporting the lack of observable self-assembly. More strikingly, Cu2+ rapidly disaggregates amyloid fibrils preformed by Aβ(13–21)K16A (SI Fig. 6C) but was not able to dissociate peptide assemblies preformed in the presence of equimolar Zn2+ (data not shown).

Aβ(13–21)K16A Cu2+ Coordination.

Inhibition of self-assembly in the presence of Cu2+ suggests that Cu2+ binds to the peptide. Isothermal titration calorimetry (ITC) measured the heat change upon Cu2+ binding and fit best to a one-site binding model, with each Cu2+ binding two Aβ(13–21)K16A molecules with an affinity constant (Ka) of 108 M (SI Fig. 7). Displacement of water from the Cu2+(H2O)6 complex by a deprotonated amide nitrogen, amino group, imidazole nitrogen, carboxylate, and/or a water donor, causes a significant blue shift in λmax (27) (see SI Materials and Methods). The Cu2+ transitions in the presence of one equivalent of Aβ(13–21)K16A in pH 5.6 buffer shifted to 600 nm (ε ≈ 40 M−1 cm−1), which is characteristic of three- or four-nitrogen coordination in a type (II) square-planar or distorted tetragonal complex (Fig. 1A). The four bands present in the CD spectrum (Fig. 1B), are assigned as a 264-nm transition for a strong NH2 and/or π2 imidazole-to-Cu2+ charge transfer, 300 nm as charge transfer from N to Cu2+, a weak 340-nm transition as charge transfer from π1 imidazole to Cu2+, and a 596-nm Cotton effect for a Cu2+ d–d transition (28). Virtually identical Cu2+ ITC, UV, and CD data were obtained for wild-type Aβ(13–21) (data not shown), suggesting that Cu2+ adopts the same coordination with both peptides.

Fig. 1.

Fig. 1.

Characterization of Cu2+–Aβ(13–21) (HHQKLVFFA) congener complexes. (A and B) UV-Vis absorbance (A) and CD spectra (B) of 0.4 mM Cu2+ in the presence of 0.4 mM Aβ peptide. Black, Aβ(13–21)K16A; green, Aβ(Q15A); red, Aβ(H14A); and blue, Aβ(Ac-N). (C) CW-EPR spectra of the Cu2+–Aβ(13–21)K16A complex with different Cu2+-to-peptide ratios. The concentration of Aβ(13–21)K16A peptide is 0.5 mM. (D) Fourier transform (solid line) and simulation (dashed line) of three-pulse ESEEM for the soluble Cu2+–Aβ(13–21)K16A complex prepared from 0.4 mM Cu2+ in the presence of 0.5 mM Aβ(13–21)K16A.

The assignments were further tested with three fiber-forming single-point substitutions in Aβ(13–21)K16A: H14A, Q15A, and acetylation of the N-terminal amino group, Ac-N. Cu2+ bound and inhibited self-assembly of all three peptides (data not shown). Q15A retained the UV-Vis λmax absorption at 600 nm (ε ≈ 40 M−1 cm−1) (Fig. 1A) and a CD spectrum very similar to Aβ(13–21)K16A (Fig. 1B). These results argue against involvement of Gln-15 in Cu2+ binding. In contrast, both H14A and Ac-N exhibited red-shifted UV maxima and decreased absorption intensity (Fig. 1A) as well as dramatic changes in the CD spectra (Fig. 1B). These results (consistent with Eq. 5 in SI Materials and Methods and SI Table 1) implicate both His-14 and the N-terminal amine as coordinating ligands.

Cu2+ was interrogated directly with continuous wave—EPR (CW-EPR) after titration of Aβ(13–21)K16A with CuCl2 (Fig. 1C). Identical line shapes were observed for samples with 0.2 and 0.4 molar equivalent (eq) of Cu2+. At 0.8 eq, weak features characteristic of the aqueous copper complex, Cu2+(H2O)6, emerged that broadened and distorted the line shape. At 1.0 eq Cu2+, the mI = −3/2 copper hyperfine feature of Cu2+(H2O)6 at 2,605 G and the narrow derivative feature of Cu2+(H2O)6 (peak, 3,200 G; trough, 3,250 G) appeared. The amplitude of the Cu2+(H2O)6 spectrum becomes dominant over the Cu2+–Aβ(13–21)K16A spectrum with 2 and 3 eq of Cu2+. These results are consistent with ITC confirming a stoichiometry of 0.5 eq of Cu2+ per peptide. The EPR spectrum with 0.4 eq of Cu2+ (Fig. 1C) best represents the line shape of the Cu2+–peptide complex and is characterized by g = 2.236, and a copper hyperfine splitting at a of 196 G, consistent with coordination by 2N2O or 3N1O or 4N equatorial ligands (29) (summarized in SI Table 1).

The three-pulse electron spin echo envelope modulation (ESEEM) spectrum of the Aβ(13–21)K16A complex with Cu2+ (Fig. 1D) is characteristic of superhyperfine coupling (shfc) between the electron on Cu2+ and a 14N nucleus, where the shfc and nuclear Zeeman contributions approximately cancel for one electron spin manifold (A/2 ≈ νN, where A is the shfc constant and νN is the 14N free nuclear frequency) (30, 31). This “exact cancellation” condition gives rise to the sharp lines at 0.6, 0.8, and 1.4 MHz in the ESEEM spectrum, which are assigned to the ν0, ν and ν+ 14N nuclear quadrupole transitions, respectively (32). The broad feature at 4 MHz arises from the ΔmI = ±2 splitting in the electron spin manifold where the shfc and nuclear Zeeman contributions are additive. The pattern in the ESEEM spectrum (Fig. 1D) and the simulation values (SI Table 2) are characteristic of imidazole coordination to Cu2+, and arise from the uncoordinated, remote nitrogen of imidazole (30, 31, 33). These results establish that one histidine imidazole coordinates Cu2+ in an equatorial position.

Therefore, ITC established that Cu2+ binds two Aβ(13–21)K16A peptides. The CD blue shifted d–d transition and ellipticity at 596 nm strongly support Cu2+ coordination to a deprotonated amide nitrogen (11, 3436). Cu2+ is known to chelate backbone amides, particularly if available histidine residues or an α amino group provide an anchor point for the metal ion (37). In these cases, Cu2+ can promote peptide nitrogen ionization at pH values as low as 5–6, even though the pKa is 12–13 in solution. The value of Δε/n for the charge transfer between N and Cu2+ at 300 nm is fairly constant, in the range of 0.2–0.3, where n is the number of negatively charged peptide nitrogens involved in coordination (28). Therefore, Δε (300 nm) ≈ 0.23 (Fig. 1B) indicates Cu2+ coordination with one deprotonated backbone amide. Coordination to the N-terminal amine, the amide backbone, and one equatorial histidine residue requires His-13 and His-14 to form a tridentate metal complex, making structure 1 (SI Fig. 8) the most energetically accessible structure. The asymmetric α-carbon of His-14, held in the chelating ring between a main-chain amide nitrogen and a histidine imidazole ring, provides a vicinal effect (38, 39), consistent with the intense CD signal at 596 nm for the Cu2+ d–d transition. Finally, the conclusion from ITC that Cu2+ binds with two Aβ(13–21)K16A peptides indicates that the open coordination site is occupied by a ligand from the second peptide.

Design of Ac-Aβ(13–21)H14A for Cu2+-Induced Self-Assembly.

In contrast to Zn2+ (22), Cu2+ inhibits assembly of Aβ(13–21)K16A by deprotonating a backbone amide nitrogen and rearranging the peptide backbone to create a chelated metal complex. Even when the amino group is acetylated, as in Aβ(Ac-N), which results in loss of the amide nitrogen as a ligand, as suggested by the disappearance of the CD absorption at 600 nm (Fig. 1B and SI Table 1), amyloid fibril formation remains inhibited. In this case, Cu2+ appears to bind with both His-13 and His-14 intramolecularly, preventing extended β-strand formation and self-assembly (SI Table 1). Therefore, we designed a peptide Ac-Aβ(13–21)H14A, CH3CO-HAQKLVFFA-NH2, where removal of both His-14 and the free amino group would limit the ligands accessible to Cu2+ in each peptide molecule to a single His-13.

During incubation at pH 7.0, Ac-Aβ(1–40). Although the absent or present metal-ion fibrils displayed similar structure, their morphology differed. In the absence of metal ions, typical amyloid fibrils and tightly twisted fibers were apparent by atomic force microscopy (SI Fig. 9C) and transmission electron microscopy (TEM, data not shown) with diameters of 8 nm. The fibrils formed in the presence of Zn2+ or Cu2+ were both nontwisted and smooth, with diameters of 8–9 nm (SI Fig. 9 D and E).

Structural Characterization of Metal Ion-Induced Amyloid Fibrils.

Isotope-edited FT-IR (Fig. 2) was used to probe the local structures of the metal-free and metal-induced amyloid fibrils (4145). In Ac-Aβ(13–21)H14A, 13CInline graphicO labels were placed in the middle of the sequence, [1-13C]L17, or close to the C terminus, [1-13C]F20. In metal-free amyloid fibrils, the [1-13C]F20 amide stretch has an extremely weak shoulder at 1,605 cm−1. The [1-13C]L17 peptide also has a band at 1,605 cm−1, but with increased intensity, and an absorption maximum shift from 1,630 to 1,637 cm−1 for 12CInline graphicO. This shift results from disturbance of the 12C hydrogen-bonding network by 13C (46) and is not observed for [1-13C]F20-labeled fibrils. The stronger intensity for the [1-13C]L17 carbonyl carbon is consistent with Leu-17 being buried in the fibril hydrophobic core with more ordered hydrogen bonding. The isotope-edited IR spectra of fibrils formed in the absence and presence of Zn2+ or Cu2+ ions are identical in both labeling schemes. This result further confirms a very similar, if not identical, peptide conformation and packing order within the fibril β-sheets.

Fig. 2.

Fig. 2.

Isotope edited FT-IR. Absorption of amide I bands of Ac-Aβ(13–21)H14A (Ac-HAQKLVFFA) fibrils formed with free peptide (P) and in the presence of Zn2+ or Cu2+. UL, unlabeled peptide; F20, 13CInline graphicO was introduced into the peptide at Phe-20; L17, 13CInline graphicO was introduced into the peptide at Leu-17.

Extent of Metalation of the Ac-Aβ(13–21)H14A Fibrils.

ITC cannot deconvolute metal binding from peptide–peptide association during Ac-Aβ(13–21)H14A amyloid assembly. Therefore, amino acid analysis (AAA) and inductively coupled plasma mass spectrometry (ICP-MS) were performed to determine the metal-to-peptide ratio in the metalated Ac-Aβ(13–21)H14A fibrils (SI Table 3). Initial peptide concentration was 2 mM, with differing initial metal concentrations ranging from 0.5 to 2 mM. After maturation, fibers were pelleted, washed, and subjected to AAA and ICP-MS measurements. The Zn2+-to-peptide ratio varied from 0.48 to 0.64, and the Cu2+-to-peptide ratio varied from 0.55 to 0.73. The imprecision in incorporation ratios probably arises from nonspecifically bound metal ions. The results establish a peptide-to-metal ratio close to 2:1 in both the Zn2+ and Cu2+ assemblies.

Coordination Environment of Metal Ions in Ac-Aβ(13–21)H14A Fibrils.

To specifically characterize the Cu2+ coordination responsible for induction of amyloid fibrils, Cu2+–fibrillar complexes were pelleted, rinsed, and resuspended in fresh buffer. The three-pulse ESEEM spectrum of the Cu2+–fibrillar complex (Fig. 3A) shows ν0, ν, and ν+ nuclear quadrupole features at the same frequency positions, 0.6, 0.8, and 1.4 MHz, as observed for the Cu2+–Aβ(13–21)K16A complex (Fig. 1D). In addition, weaker peaks are present that correspond to combinations of the fundamental frequencies, including bands centered at 2.1 MHz ([ν0+] and [ν + ν+]) and at 2.8 MHz ([ν+ + ν+]). Combination lines in ESEEM spectra indicate the presence of multiple nuclear couplings, and the relative amplitudes of the fundamental and combination features (47) are consistent with two coupled 14N nuclei, with e2qQ/h = 1.55 MHz and η = 0.751 (SI Table 2). The simulation parameters and spectral pattern, including the enhanced amplitude of the ΔmI = ±2 feature at 4 MHz, demonstrate coupling of Cu2+ to two remote 14N atoms of distinct histidine imidazoles, establishing that two histidine imidazoles are coordinated equatorially to Cu2+ in the fibrillar complex.

Fig. 3.

Fig. 3.

Copper ion analysis in the Ac-Aβ(13–21)H14A (Ac-HAQKLVFFA) fibrillar complex. (A) Fourier transform (solid line) and simulation (dashed line) of three-pulse ESEEM for Cu2+-Ac-Aβ(13–21)H14A fibrils. Simulation values are summarized in SI Table 2. (B) The Fourier transform of extended x-ray absorption fine-structure (EXAFS) (Inset) of Cu2+-Ac-Aβ(13–21)H14A fibrils. Solid line, experiment; dashed line, fit 8 (values listed in SI Table 4).

Following previously developed procedures for probing Zn2+ coordination environments (22), Ac-Aβ(13–21)H14A fibrils were characterized by extended x-ray absorption fine structure (EXAFS) spectroscopy. Curve fitting indicates 3N/1O or 2N/2O or 1N/3O atoms in the first shell (SI Table 4). The coincident appearance and intensities of the 3-Å and 4-Å peaks in the Fourier transform EXAFS spectra [Fig. 3B for Cu2+ and ref. 22 for Zn2+ Ac-Aβ(13–21)H14A fibrillar complexes] are diagnostic of two imidazole ligands. The best fit (dashed line in Fig. 3B) was generated by using two His imidazoles and two additional first-shell light atoms (N or O) (SI Table 4, Fit 8). Therefore, both Zn2+ and Cu2+ induce Ac-Aβ(13–21)H14A fibril formation by bridging two His-13 residues between parallel hydrogen-bonded β-strands in a His–metal2+–His chelated structure.

Toxicity of Cu2+–Peptide Complexes.

Initial neurotoxicity assays in cultured neurons suggested that the two Cu2+–peptide coordination environments [Cu2+–Aβ(13–21)K16A complex and Cu2+–Ac-Aβ(13–21)H14A fibrils] were biologically distinct (data not shown). To more specifically characterize the activity, a myocyte enhancer factor 2 (MEF2) assay was developed in SN4741, a dopaminergic neuron cell line derived from the mouse midbrain. The MEF2 family of transcription factors play critical roles in diverse cellular processes including neuronal survival (48). MEF2s are an endpoint for several neurotoxic signaling pathways that control the molecular machinery of cellular apoptosis (49, 50). As shown in Fig. 4Left, Aβ(1–42) regulates MEF2. In addition, inhibition of MEF2 activity by mature peptide assemblies is dose-dependent and peptide-specific. The Cu2+–Ac-Aβ(13–21)H14A fibrils are equally as inhibitory as Aβ(1–42), whereas the Cu2+-free Ac-Aβ(13–21)H14A fibrils and Cu2+ ions alone are ineffective. The soluble Cu2+–Aβ(13–21)K16A complex is also nontoxic, suggesting that different Cu2+ chelation environments, as mediated by different Aβ conformations, compromise neuron viability.

Fig. 4.

Fig. 4.

MEF2 luciferase reporter gene assay. (Left) Aβ(1–42) dose-dependent response. (Right) Cu2+–Ac-Aβ(13–21)H14A fibrils are as toxic as Aβ(1–42) and Aβ(10–35) fibrils and Cu2+–Aβ(13–21)K16A complex is nontoxic. Peptide concentration is 10 μM.

Summary.

Metal ion association with Aβ has frequently been correlated with Alzheimer's disease (11, 35, 51, 52), and the link among metal-coordination structure, the impact on assembly kinetics, and overall aggregation remains critical to understanding disease etiology. Our attempts to identify individual metal sites within Aβ segments now reveal that Cu2+ coordination can radically alter assembly kinetics and morphology. In the short Aβ fragment Aβ(13–21)K16A, backbone deprotonation occurs even at low pH to generate a high affinity (Ka = 108 M) Cu2+-binding site. The five- {NH2, Namide} and six-membered {Namide, NIm} metal chelate in structure 1 (SI Fig. 8) coordinates with a second peptide, probably through a histidine residue. A virtually identical {NH2, Namide, NIm} complex was previously observed for several short peptides that contain His as the second residue from the peptide N terminus, such as X-His (X = Ala, Gly, His) or X-His-Y (X = Ala, Gly, and Y = Ala, Leu, Lys) (5356). In Aβ(13–21)K16A, this X-His (X = His) motif retards β-strand and β-sheet assembly, but this inhibition may be unique to Cu2+ because Zn2+ readily promotes amyloid self-assembly (22). Even with acetylation of the N terminus as in Ac-Aβ(13–21)K16A, Cu2+ coordinates with both His-13 and His-14 intramolecularly to retard peptide assembly. This inhibition of assembly was previously observed for Ac-Aβ(11-X) (X = 16, 20, 28) (57, 58), probably through a similar complex.

Removing one histidine, as in Ac-Aβ(13–21)H14A, generates a peptide in which both Zn2+ and Cu2+ accelerate amyloid fibril formation by capturing intermolecular His–metal–His coordination (Fig. 5C). The isotope-edited FT-IR results predict a parallel, in-register β-sheet organization, an assignment further supported by preliminary solid-state NMR results (data not shown). The same conformation was previously determined for the well characterized Aβ(10–35) peptide assembled in the absence of metal ions (21, 59). Here, both Zn2+ and Cu2+ associate with two His-13 residues from adjacent peptides arranged along the growing β-sheet surface (22), increasing the growth rate and stability of the growing fibril. Stoichiometry analyses are consistent with the organized linear array of metal ions along the fibril surface (Fig. 5C). Therefore, the ability of Cu2+ to access the intermolecular coordination environment depends on subtle changes in the intramolecular site for Cu2+.

Fig. 5.

Fig. 5.

Structural models for the Cu2+–peptide complexes. (A) Proposed model for Cu2+ coordination with Aβ(13–21)K16A showing only the first four residues (HHQA). (B) Crystal structure for Cu2+–HGGGW complex, the N-terminal repeat sequence of human prion protein (63). Only equatorial ligands are illustrated in both models. Purple, Cu ion; green, carbon; red, oxygen; blue, nitrogen; gray, unidentified ligand. (C) Structural models for Cu2+ arrangement in Ac-Aβ(13–21)H14A fibrils. (Left) View of three stacked β-sheets. (Right) View down peptide backbone. H-bonds are parallel to fibril long axis, and fibrils consist of parallel in-register β-sheets in both the absence and presence of metal ions. Sheets stack perpendicular to the H-bond direction. Both Zn2+ and Cu2+ coordinate two histidines along the H-bonding dimension within the same β-sheet. Red, His-13; blue, remaining Ac-Aβ(13–21)H14A residues; magenta, Zn2+, Cu2+; and green, H-bonds between backbone carbonyl and amide. Amino acid side chains, except His-13, are removed for clarity.

In this regard, Aβ(1–40) are not included in the core β-sheet of the fibrils but, rather, are structurally disordered and exposed along the fibril surface (60). Intramolecular metal complexation close to this structurally flexible region may slow assembly (24), but, as observed in these short peptides, does not block assembly.

“Switching” to achieve different Cu2+-binding sites and peptide conformations has also been considered for the mammalian prion protein (PrP). Cu2+ binds primarily to the N terminus of PrP, a region that is structurally flexible and outside of the core assembly region (61). Therefore, metal-binding regions in both Aβ and PrP are in the structurally flexible N-terminal domain and lie outside the amyloid-determining core (21, 62). The N terminus of PrP contains four copies of the Cu2+-binding octarepeat sequence, PHGGGWGQ. In the presence of two or more molar equivalents of Cu2+, the metal ions bind in an intrarepeat manner, with ligands supplied by one histidine imidazole, two deprotonated amides from the next two glycines, and the amide carbonyl of the second glycine (63) (Fig. 5B). At low Cu2+ concentrations or under acidic pH conditions, multiple His imidazoles from different octarepeats coordinate with Cu2+, forming interrepeat binding sites (64, 65). These interrepeat sites could be intramolecular, intermolecular, or a combination (65).

Moreover, several proteins associated with neurodegenerative diseases also bind metal ions, and metal binding modulates the aggregation behavior of these proteins. α-Synuclein (α-Syn), a protein associated with Parkinson's disease (PD), is intrinsically unfolded and Cu2+ binding results in rapid aggregation (66). Superoxide dismutase 1 (SOD1), a CuZn metalloprotein that catalyzes the dismutation of the superoxide anion, is associated with the progressive neurodegenerative disease amyotrophic lateral sclerosis (ALS). Cu2+ and Zn2+ bind and organize a loop region of SOD1, and mutations around the binding site that result in alterations of the coordination environment of Cu2+ and Zn2+ lead to misfolding and aggregation of SOD1 (67, 68). The oxidative modification and aggregation of SOD1 has been correlated with both AD and PD (69). Thus, it seems reasonable to speculate that AD, PD, and ALS share common mechanisms of metal ion regulation and/or metal-ion induction of protein aggregation, which might lead to a common, or at least overlapping, pathogenic mechanism(s).

Our studies now establish that the N-terminal region of Aβ can access a range of metal-coordination structures. These results suggest that the effect of metal ions in neurodegenerative diseases is not simply a result of metal ion on/off binding, but is due to switching between different metal-binding modes. Clearly, the short peptides allow these structures to be constructed and evaluated directly, whereas in the larger Aβ peptides, the accessible conformations and subtle environmental conditions can alter ligand accessibility and modulate both amyloid assembly and cellular toxicity. The striking quasilinear array of Cu2+ ions along the surface of the biologically active Ac-Aβ(13–21)H14A fibril suggests that assembly within a complex cellular milieu will certainly produce novel protein architectures, architectures that can now be constructed and evaluated directly for functional neuron toxicity and prion function.

Materials and Methods

Fibril Formation.

Each Aβ(13–21) congener peptide was dissolved completely in distilled deionized H2O, sonicated for 10 min, and centrifuged at 16,110 × g for 10 min. The supernatant was used as stock solution for each peptide. Samples with desired peptide and metal concentration in 25 mM Mes buffer (pH 5.6) or Hepes buffer (pH 7.2) and 10 mM NaCl were prepared from stock solutions.

Spectroscopy.

UV-Vis absorption spectra were obtained on a V-530 UV/VIS spectrometer (Jasco, Easton, MD) with a 1-cm path-length quartz cuvette. CD spectra were recorded on a Jasco 810 CD spectropolarimeter. Ellipticity (θ, m degrees) was converted to mean residue molar ellipticity (MRME) [θ] = θ/(10 × n × C × l) or to molar ellipticity Δε = θ/(32980 × C × l), where n is number of amide bonds per peptide, C is molar concentration (mol/liter), and l is path length (cm). FT-IR spectra were collected on a Magna 560 IR spectrometer (Nicolet, Madison, WI). Mature fibrils were spun down, lyophilized, mixed, and ground with KBr and pressed into transparent disks. Typically, 100 scans were averaged with 4-cm−1 resolution. X-ray absorption spectra were collected at the Stanford Synchrotron Radiation Laboratory on beamline 9-3. Mature fibrils were spun down and washed with fresh buffer solution three times. The hydrated pellet was transferred to a 25-well sample holder, frozen, and stored in liquid nitrogen until data collection. See SI Materials and Methods for instrumental setup and data analysis.

EPR.

Samples for both EPR and ESEEM were mixed with an equal volume of ethylene glycol, transferred to 4 mm o.d. quartz EPR tubes, frozen in liquid nitrogen-chilled 2-methylbutane (≈150 K), and stored in liquid nitrogen. Continuous-wave EPR spectra were collected on a ER200D EPR spectrometer (Bruker, Billerica, MA). ESEEM spectra were collected on a home-built pulsed-EPR spectrometer by using the three-pulse stimulated echo pulse sequence. Details of EPR and ESEEM experiments and simulations (70, 71) are described in SI Materials and Methods.

Microscopy.

AFM samples were placed on silicon chips and imaged under dry conditions in tapping mode on a JSPM-4210 AFM (JEOL, Tokyo, Japan) by using ultrasharp noncontact silicon cantilevers with typical frequencies from 240 to 350 kHz.

MEF2 Assay.

SN4741 cells were transiently transfected with a DNA construct containing a luciferase gene under control of MEF2 enhancer by using the Lipofectamine 2000 transfection system (Invitrogen, Carlsbad, CA). Total transfected DNA was kept constant. After 24 h of transfection, cells were treated with Aβ peptides that had been assembled as described above. Twenty-four hours after Aβ treatment, cell lysates were analyzed for luciferase activity by using Luciferase reporter gene assay kit (Roche, Mannheim, Germany).

Supplementary Material

Supporting Information

Acknowledgments

We thank James Lah and Craig Heilman for initial neurotoxicity assays, Yi Xu and Rong Ni for assistance with peptide synthesis and purifications, and C. L. Emerson for AFM instrumentation. This work was supported by Department of Energy (DOE) Grant ER15377 (to D.G.L.), National Institutes of Health (NIH) National Institute of General Medical Sciences Grant GM42025 (to R.A.S.), NIH Grants AG 023695 and NS 048254 and the R. W. Woodruff Health Sciences Center Fund (to Z.M.), Emory Alzheimer's Disease Center Grant P50 AG025688 (to D.G.L. and Z.M.), and National Science Foundation Grant CHE-0131013 (for CD instrumentation). The Stanford Synchrotron Radiation Laboratory is a national user facility operated by Stanford University (Stanford, CA) on behalf of the U.S. DOE, Office of Basic Energy Sciences. The SSRL Structural Molecular Biology Program is supported by DOE, Office of Biological and Environmental Research and by the NIH National Center for Research Resources and Biomedical Technology Program. AAA analysis was performed by the Keck Biotechnology Resource Laboratory at Yale University (New Haven, CT), and ICP-MS was performed at the Chemical Analysis Laboratory at the University of Georgia.

Abbreviations

Aβ(13–21)K16A

HHQALVFFA-NH2

Ac-Aβ(13–21)H14A

CH3CO-HAQKLVFFA-NH2

ESEEM

electron spin echo envelope modulation

shfc

superhyperfine coupling.

Footnotes

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

This article contains supporting information online at www.pnas.org/cgi/content/full/0702669104/DC1.

References

  • 1.Carrell RW, Lomas DA. Lancet. 1997;350:134–138. doi: 10.1016/S0140-6736(97)02073-4. [DOI] [PubMed] [Google Scholar]
  • 2.Patino MM, Liu JJ, Glover JR, Lindquist S. Science. 1996;273:622–626. doi: 10.1126/science.273.5275.622. [DOI] [PubMed] [Google Scholar]
  • 3.Si K, Giustetto M, Etkin A, Hsu R, Janisiewicz AM, Miniaci MC, Kim JH, Zhu H, Kandel ER. Cell. 2003;115:893–904. doi: 10.1016/s0092-8674(03)01021-3. [DOI] [PubMed] [Google Scholar]
  • 4.Si K, Lindquist S, Kandel ER. Cell. 2003;115:879–891. doi: 10.1016/s0092-8674(03)01020-1. [DOI] [PubMed] [Google Scholar]
  • 5.Dobson CM. Nature. 2003;426:884–890. doi: 10.1038/nature02261. [DOI] [PubMed] [Google Scholar]
  • 6.Scheibel T, Parthasarathy R, Sawicki G, Lin XM, Jaeger H, Lindquist SL. Proc Natl Acad Sci USA. 2003;100:4527–4532. doi: 10.1073/pnas.0431081100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Reches M, Gazit E. Science. 2003;300:625–627. doi: 10.1126/science.1082387. [DOI] [PubMed] [Google Scholar]
  • 8.Silva GA, Czeisler C, Niece KL, Beniash E, Harrington DA, Kessler JA, Stupp SI. Science. 2004;303:1352–1355. doi: 10.1126/science.1093783. [DOI] [PubMed] [Google Scholar]
  • 9.Bush AI, Pettingell WH, Multhaup G, Paradis MD, Vonsattel JP, Gusella JF, Beyreuther K, Masters CL, Tanzi RE. Science. 1994;265:1464–1467. doi: 10.1126/science.8073293. [DOI] [PubMed] [Google Scholar]
  • 10.Bush AI, Multhaup G, Moir RD, Williamson TG, Small DH, Rumble B, Pollwein P, Beyreuther K, Masters CL. J Biol Chem. 1993;268:16109–16112. [PubMed] [Google Scholar]
  • 11.Syme CD, Nadal RC, Rigby SE, Viles JH. J Biol Chem. 2004;279:18169–18177. doi: 10.1074/jbc.M313572200. [DOI] [PubMed] [Google Scholar]
  • 12.Atwood CS, Moir RD, Huang XD, Scarpa RC, Bacarra NME, Romano DM, Hartshorn MK, Tanzi RE, Bush AI. J Biol Chem. 1998;273:12817–12826. doi: 10.1074/jbc.273.21.12817. [DOI] [PubMed] [Google Scholar]
  • 13.Curtain CC, Ali F, Volitakis I, Cherny RA, Norton RS, Beyreuther K, Barrow CJ, Masters CL, Bush AI, Barnham KJ. J Biol Chem. 2001;276:20466–20473. doi: 10.1074/jbc.M100175200. [DOI] [PubMed] [Google Scholar]
  • 14.Miura T, Suzuki K, Kohata N, Takeuchi H. Biochemistry. 2000;39:7024–7031. doi: 10.1021/bi0002479. [DOI] [PubMed] [Google Scholar]
  • 15.Liu ST, Howlett G, Barrow CJ. Biochemistry. 1999;38:9373–9378. doi: 10.1021/bi990205o. [DOI] [PubMed] [Google Scholar]
  • 16.Yang DS, McLaurin J, Qin K, Westaway D, Fraser PE. Eur J Biochem. 2000;267:6692–6698. doi: 10.1046/j.1432-1327.2000.01767.x. [DOI] [PubMed] [Google Scholar]
  • 17.Atwood CS, Scarpa RC, Huang XD, Moir RD, Jones WD, Fairlie DP, Tanzi RE, Bush AI. J Neurochem. 2000;75:1219–1233. doi: 10.1046/j.1471-4159.2000.0751219.x. [DOI] [PubMed] [Google Scholar]
  • 18.Stellato F, Menestrina G, Dalla Serra M, Potrich C, Tomazzolli R, Meyer-Klaucke W, Morante S. Eur Biophys J. 2006;35:340–351. doi: 10.1007/s00249-005-0041-7. [DOI] [PubMed] [Google Scholar]
  • 19.Morgan DM, Dong J, Jacob J, Lu K, Apkarian RP, Thiyagarajan P, Lynn DG. J Am Chem Soc. 2002;124:12644–12645. doi: 10.1021/ja0273086. [DOI] [PubMed] [Google Scholar]
  • 20.Lu K, Jacob J, Thiyagarajan P, Conticello VP, Lynn DG. J Am Chem Soc. 2003;125:6391–6393. doi: 10.1021/ja0341642. [DOI] [PubMed] [Google Scholar]
  • 21.Benzinger TLS, Gregory DM, Burkoth TS, Miller-Auer H, Lynn DG, Botto RE, Meredith SC. Proc Natl Acad Sci USA. 1998;95:13407–13412. doi: 10.1073/pnas.95.23.13407. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Dong J, Shokes JE, Scott RA, Lynn DG. J Am Chem Soc. 2006;128:3540–3542. doi: 10.1021/ja055973j. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Karr JW, Kaupp LJ, Szalai VA. J Am Chem Soc. 2004;126:13534–13538. doi: 10.1021/ja0488028. [DOI] [PubMed] [Google Scholar]
  • 24.Zou J, Kajita K, Sugimoto N. Angew Chem Int Ed Engl. 2001;40:2274–2277. [PubMed] [Google Scholar]
  • 25.Suzuki K, Miura T, Takeuchi H. Biochem Biophys Res Commun. 2001;285:991–996. doi: 10.1006/bbrc.2001.5263. [DOI] [PubMed] [Google Scholar]
  • 26.Petkova AT, Leapman RD, Guo Z, Yau WM, Mattson MP, Tycko R. Science. 2005;307:262–265. doi: 10.1126/science.1105850. [DOI] [PubMed] [Google Scholar]
  • 27.Sigel H, Martin RB. Chem Rev. 1982;82:385–426. [Google Scholar]
  • 28.Daniele PG, Prenesti E, Ostacoli G. J Chem Soc Dalton. 1996:3269–3275. [Google Scholar]
  • 29.Peisach J, Blumberg WE. Arch Biochem Biophys. 1974;165:691–708. doi: 10.1016/0003-9861(74)90298-7. [DOI] [PubMed] [Google Scholar]
  • 30.Mims WB, Peisach J. J Chem Phys. 1978;69:4921–4930. [Google Scholar]
  • 31.Flanagan HL, Singel DJ. J Chem Phys. 1987;87:5606–5616. [Google Scholar]
  • 32.Lucken EAC. Nuclear Quadrupole Coupling Constants. London: Academic; 1969. [Google Scholar]
  • 33.Jiang F, McCracken J, Peisach J. J Am Chem Soc. 1990;112:9035–9044. [Google Scholar]
  • 34.Stanczak P, Valensin D, Juszczyk P, Grzonka Z, Migliorini C, Molteni E, Valensin G, Gaggelli E, Kozlowski H. Biochemistry. 2005;44:12940–12954. doi: 10.1021/bi051177e. [DOI] [PubMed] [Google Scholar]
  • 35.Kowalik-Jankowska T, Ruta M, Wisniewska K, Lankiewicz L. J Inorg Biochem. 2003;95:270–282. doi: 10.1016/s0162-0134(03)00128-4. [DOI] [PubMed] [Google Scholar]
  • 36.Brown DR, Guantieri V, Grasso G, Impellizzeri G, Pappalardo G, Rizzarelli E. J Inorg Biochem. 2004;98:133–143. doi: 10.1016/j.jinorgbio.2003.09.006. [DOI] [PubMed] [Google Scholar]
  • 37.Sundberg RJ, Martin RB. Chem Rev. 1974;74:471–517. [Google Scholar]
  • 38.Nakanishi K, Nina E, Woody RW. Circular Dichroism: Principles and Applications. New York: Wiley–VCH; 1994. p. 223. [Google Scholar]
  • 39.Tsangaris JM, Martin RB. J Am Chem Soc. 1970;92:4255–4260. doi: 10.1021/ja00717a020. [DOI] [PubMed] [Google Scholar]
  • 40.Fabian H, Szendrei GI, Mantsch HH, Otvos L., Jr Biochem Biophys Res Commun. 1993;191:232–239. doi: 10.1006/bbrc.1993.1207. [DOI] [PubMed] [Google Scholar]
  • 41.Lu K. Atlanta: Emory Univ; 2005. PhD dissertation. [Google Scholar]
  • 42.Paul C, Axelsen PH. J Am Chem Soc. 2005;127:5754–5755. doi: 10.1021/ja042569w. [DOI] [PubMed] [Google Scholar]
  • 43.Hiramatsu H, Goto Y, Naiki H, Kitagawa T. J Am Chem Soc. 2005;127:7988–7989. doi: 10.1021/ja050844o. [DOI] [PubMed] [Google Scholar]
  • 44.Lansbury PT, Jr, Costa PR, Griffiths JM, Simon EJ, Auger M, Halverson KJ, Kocisko DA, Hendsch ZS, Ashburn TT, Spencer RG, et al. Nat Struct Biol. 1995;2:990–998. doi: 10.1038/nsb1195-990. [DOI] [PubMed] [Google Scholar]
  • 45.Silva RA, Barber-Armstrong W, Decatur SM. J Am Chem Soc. 2003;125:13674–13675. doi: 10.1021/ja036725v. [DOI] [PubMed] [Google Scholar]
  • 46.Brauner JW, Dugan C, Mendelsohn R. J Am Chem Soc. 2000;122:677–683. [Google Scholar]
  • 47.McCracken J, Pember S, Benkovic SJ, Villafranca JJ, Miller RJ, Peisach J. J Am Chem Soc. 1988;110:1069–1074. [Google Scholar]
  • 48.Mao Z, Bonni A, Xia F, Nadal-Vicens M, Greenberg ME. Science. 1999;286:785–790. doi: 10.1126/science.286.5440.785. [DOI] [PubMed] [Google Scholar]
  • 49.Gong X, Tang X, Wiedmann M, Wang X, Peng J, Zheng D, Blair LA, Marshall J, Mao Z. Neuron. 2003;38:33–46. doi: 10.1016/s0896-6273(03)00191-0. [DOI] [PubMed] [Google Scholar]
  • 50.Tang X, Wang X, Gong X, Tong M, Park D, Xia Z, Mao Z. J Neurosci. 2005;25:4823–4834. doi: 10.1523/JNEUROSCI.1331-05.2005. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Karr JW, Akintoye H, Kaupp LJ, Szalai VA. Biochemistry. 2005;44:5478–5487. doi: 10.1021/bi047611e. [DOI] [PubMed] [Google Scholar]
  • 52.Kozin SA, Zirah S, Rebuffat S, Hoa GH, Debey P. Biochem Biophys Res Commun. 2001;285:959–964. doi: 10.1006/bbrc.2001.5284. [DOI] [PubMed] [Google Scholar]
  • 53.Kozlowski H, Bal W, Dyba M, Kowalik-Jankowska T. 1999;184:319–346. [Google Scholar]
  • 54.Bruni S, Cariati F, Daniele PG, Prenesti E. Spectrochim Acta A. 2000;56:815–827. doi: 10.1016/s1386-1425(99)00283-8. [DOI] [PubMed] [Google Scholar]
  • 55.Conato C, Gavioli R, Guerrini R, Kozlowski H, Mlynarz P, Pasti C, Pulidori F, Remelli M. Biochim Biophys Acta. 2001;1526:199–210. doi: 10.1016/s0304-4165(01)00127-1. [DOI] [PubMed] [Google Scholar]
  • 56.Livera CE, Pettit LD, Bataille M, Perly B, Kozlowski H, Radomska B. J Chem Soc Dalton. 1987:661–666. [Google Scholar]
  • 57.Kowalik-Jankowska T, Ruta-Dolejsz M, Wisniewska K, Lankiewicz L, Kozlowski H. J Chem Soc Dalton. 2000:4511–4519. [Google Scholar]
  • 58.Kowalik-Jankowska T, Ruta-Dolejsz M, Wisniewska K, Lankiewicz L. J Inorg Biochem. 2002;92:1–10. doi: 10.1016/s0162-0134(02)00495-6. [DOI] [PubMed] [Google Scholar]
  • 59.Burkoth TS, Benzinger TLS, Urban V, Morgan DM, Gregory DM, Thiyagarajan P, Botto RE, Meredith SC, Lynn DG. J Am Chem Soc. 2000;122:7883–7889. [Google Scholar]
  • 60.Petkova AT, Ishii Y, Balbach JJ, Antzutkin ON, Leapman RD, Delaglio F, Tycko R. Proc Natl Acad Sci USA. 2002;99:16742–16747. doi: 10.1073/pnas.262663499. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Donne DG, Viles JH, Groth D, Mehlhorn I, James TL, Cohen FE, Prusiner SB, Wright PE, Dyson HJ. Proc Natl Acad Sci USA. 1997;94:13452–13457. doi: 10.1073/pnas.94.25.13452. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.James TL, Liu H, Ulyanov NB, Farr-Jones S, Zhang H, Donne DG, Kaneko K, Groth D, Mehlhorn I, Prusiner SB, et al. Proc Natl Acad Sci USA. 1997;94:10086–10091. doi: 10.1073/pnas.94.19.10086. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Millhauser GL. Acc Chem Res. 2004;37:79–85. doi: 10.1021/ar0301678. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 64.Chattopadhyay M, Walter ED, Newell DJ, Jackson PJ, Aronoff-Spencer E, Peisach J, Gerfen GJ, Bennett B, Antholine WE, Millhauser GL. J Am Chem Soc. 2005;127:12647–12656. doi: 10.1021/ja053254z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 65.Morante S, Gonzalez-Iglesias R, Potrich C, Meneghini C, Meyer-Klaucke W, Menestrina G, Gasset M. J Biol Chem. 2004;279:11753–11759. doi: 10.1074/jbc.M312860200. [DOI] [PubMed] [Google Scholar]
  • 66.Rasia RM, Bertoncini CW, Marsh D, Hoyer W, Cherny D, Zweckstetter M, Griesinger C, Jovin TM, Fernandez CO. Proc Natl Acad Sci USA. 2005;102:4294–4299. doi: 10.1073/pnas.0407881102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 67.Antonyuk S, Elam JS, Hough MA, Strange RW, Doucette PA, Rodriguez JA, Hayward LJ, Valentine JS, Hart PJ, Hasnain SS. Protein Sci. 2005;14:1201–1213. doi: 10.1110/ps.041256705. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 68.Valentine JS, Hart PJ. Proc Natl Acad Sci USA. 2003;100:3617–3622. doi: 10.1073/pnas.0730423100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 69.Choi J, Rees HD, Weintraub ST, Levey AI, Chin LS, Li L. J Biol Chem. 2005;280:11648–11655. doi: 10.1074/jbc.M414327200. [DOI] [PubMed] [Google Scholar]
  • 70.Canfield JM, Warncke K. J Phys Chem B. 2002;106:8831–8841. [Google Scholar]
  • 71.Canfield JM, Warncke K. J Phys Chem B. 2005;109:3053–3064. doi: 10.1021/jp046167m. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supporting Information
pnas_0702669104_9.pdf (100.8KB, pdf)
pnas_0702669104_5.pdf (71.9KB, pdf)
pnas_0702669104_6.pdf (17.5KB, pdf)
pnas_0702669104_7.pdf (16.2KB, pdf)
pnas_0702669104_8.pdf (78.3KB, pdf)
pnas_0702669104_1.pdf (42.8KB, pdf)
pnas_0702669104_2.pdf (15.4KB, pdf)
pnas_0702669104_3.pdf (16.5KB, pdf)
pnas_0702669104_4.pdf (1.3MB, pdf)

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences

RESOURCES