Structural Changes of Region 1-16 of the Alzheimer Disease Amyloid β-Peptide upon Zinc Binding and in Vitro Aging*

Amyloid deposits within the cerebral tissue constitute a characteristic lesion associated with Alzheimer disease. They mainly consist of the amyloid peptide Aβ and display an abnormal content in Zn2+ ions, together with many truncated, isomerized, and racemized forms of Aβ. The region 1-16 of Aβ can be considered the minimal zinc-binding domain and contains two aspartates subject to protein aging. The influence of zinc binding and protein aging related modifications on the conformation of this region of Aβ is of importance given the potentiality of this domain to constitute a therapeutic target, especially for immunization approaches. In this study, we determined from NMR data the solution structure of the Aβ-(1-16)-Zn2+ complex in aqueous solution at pH 6.5. The residues His6, His13, and His14 and the Glu11 carboxylate were identified as ligands that tetrahedrally coordinate the Zn(II) cation. In vitro aging experiments on Aβ-(1-16) led to the formation of truncated and isomerized species. The major isomer generated, Aβ-(1-16)-l-iso-Asp7, displayed a local conformational change in the His6-Ser8 region but kept a zinc binding propensity via a coordination mode involving l-iso-Asp7. These results are discussed here with regard to Aβ fibrillogenesis and the potentiality of the region 1-16 of Aβ to be used as a therapeutic target.

damages. These chemical modifications occur through a common pathway involving a neutral cyclic succinimide intermediate (37,38). The potential contribution of such aspartyl modifications to A␤ amyloidosis has been addressed, because unusually high contents of racemized and isomerized Asp residues were found in A␤ isolated from amyloid deposits (39). In addition, these modifications were shown to be related to an increase in ␤-sheet content and to in vitro fibrillation (40,41), leading to the protein aging hypothesis of AD (42). On the other hand, the assessment of plaque age by using antibodies targeting specifically a particular isomer of A␤ suggests that these modifications would occur rather after the amyloid deposition (43).
The N-terminal region of A␤ appears as an attractive therapeutic target, especially for active or passive immunization approaches. Indeed, only the antibodies raised against the N-terminal part of A␤ are able to reduce the plaque burden and restore cognitive deficits in the mice model of AD (16,44,45). In addition, targeting this region should enable us to exert this effect without eliciting an inflammatory response, which had been critical in the first clinical tests of active immunization on humans (46). We have shown previously that A␤-(1-16) zinc binding induces an agonist effect on the 4 -10 epitope recognition by different monoclonal antibodies, suggesting a folding of the peptide that would render the epitope more accessible (47). Because clinical testing for passive immunization has started, it is of major importance to characterize the structural changes of the N-terminal region of A␤ upon zinc binding and protein aging-induced modifications.
NMR experiments were carried out on AVANCE 400 and DMX 600 spectrometers (Bruker Biospin, Wissembourg, France) both equipped with shielded gradients z and set up with 1 H broad band reverse gradient and triple resonance 1 H-13 C-15 N gradient probe heads, respectively. Temperature was controlled with a BCU-05 refrigeration unit and a BVT 3000 control unit on both spectrometers. 1 H and 13 C chemical shifts were externally referenced to sodium 2,2-dimethyl-2-silapentane-5-sulfonate. Conventional one-and two-dimensional experiments, one-dimensional 1 H, 1 H-1 H DQF-COSY, TOCSY (using a spin-lock field produced by an MLEV-17 spin-locking sequence for a spin-lock time of 120 ms), NOESY (with mixing times of 100, 200, 300, and 400 ms), as well as natural abundance 1 H-13 C HSQC and HMBC (with longrange coupling evolution delays of 70 ms and 90 ms) were performed. Water suppression was achieved by means of either selective low power irradiation for the DQF-COSY experiment or pulsed field gradients in a water suppression by a gradient tailored excitation scheme included in the pulse sequences for both TOCSY and NOESY experiments. Data were processed on Silicon Graphics Indigo 2 XL or O 2 workstations, using XWINNMR 3.1 and AURELIA software (Bruker Biospin, Wissembourg, France).
The determination of temperature coefficients (⌬␦/⌬T HN ) was achieved for each sample from five series of 1 H one-dimensional and TOCSY spectra recorded between 288 and 313 K. 3 J HN-H␣ coupling constants were determined from 1 H one-dimensional and DQF-COSY spectra. The chemical shift deviations (CSD) were calculated for H-␣ and C-␣ atoms considering the reference chemical shifts proposed by Wishart et al. (48,49) for each amino acid in random structure.
The pK a values of the histidines were measured in the absence of Zn 2ϩ ions by using 3.5 mM A␤-(1-16) peptide solutions in 10 mM H 2 O/ D 2 O (9:1) sodium phosphate buffer. The effects of pH on the proton chemical shifts of A␤-(1-16) were determined from a series of 1 H onedimensional, TOCSY, and NOESY spectra recorded at different pH values ranging from 3.0 to 9.5 at 278 K. The pH values were adjusted before each NMR experiment, and the absence of pH variation over the period of acquisition was checked immediately after. The pK a values were determined by analyzing the pH titration curves by nonlinear least square fit to the equation ␦ ϭ (␦ 1 ϩ ␦ 2 ϫ 10 (pH Ϫ pKa) )/(1 ϩ 10 (pH Ϫ pKa) ), where ␦ is the chemical shift of a resonance measured as a function of pH, and ␦ 1 and ␦ 2 are its chemical shifts at the lowest and highest pH values, respectively. This procedure was carried out using the software Curve Expert 1.3. The equation used derives from the Henderson-Hasselbalch equation, assuming a rapid equilibrium between protonated and unprotonated forms (50) and considering a noninteracting model (51).
Distance and Dihedral Angle Constraints-Distance constraints resulting from integrated NOESY spectra and dihedral angles derived from the 3 J HN-H␣ coupling constants using the Pardi relation (52) were used for structure calculation. The NOESY experiment with 200 ms of mixing time was selected for distance calculation to get rid of spin diffusion associated with T 1 relaxation. A tolerance range of Ϯ25% of the NMR-derived distances was used to define the upper and lower values of the constraints.
Structure Calculation and Analysis-The three-dimensional structures of A␤-(1-16) at pH 6.5 in the absence and in the presence of zinc were calculated using simulated annealing and energy minimization protocols in X-PLOR 3.851. Alternatively, the conformational calculations were performed with a general purpose internal coordinate molecular dynamics program ICMD (53) by using the variable target function approach (54) adopted for dynamics in the torsion angle space. The program ICMD was further used to calculate the three-dimensional structure of the A␤-(1-16)-Zn 2ϩ complex. The structures obtained from ICMD calculations are only presented here as final structures.
The molecular dynamics calculations in X-PLOR 3.851 were performed with a target function similar to that used by Nilges et al. (55) and a force field adapted for NMR structure determination (parallhdg-.pro and topallhdg.pro). When no stereospecific assignments could be made for methyl and methylene protons, the constraints were considered with an appropriate treatment in X-PLOR. Several rounds of structure calculation and assignment were performed to resolve ambiguities. Starting from an extended template structure, a set of 100 structures was calculated. A first phase of 400-ps dynamics (time step ϭ 2 fs) at 1000 K was followed by 80-ps slow cooling step to 100 K (time step ϭ 2 fs; temperature step ϭ 20 K). A weak weight of the van der Waals repulsive term was used at high temperature in order to allow a large conformational sampling. Refinement of the structures was achieved by using the conjugate gradient Powell algorithm with 7000 cycles of energy minimization, using the CHARMM 22 force field (files topallh22x.pro and parallh22x.pro) (56).
The ICMD program was used with the standard geometry of amino acids and peptide bonds and involved multiple cycles of simulated annealing starting from an arbitrary extended peptide conformation. The AMBER99 all-atom force field parameters (57) were applied, with nonbonded interactions truncated at 6 Å by a force-shift method to maintain reasonable atom-atom distances and avoid any bias from longer range interactions. The A␤-(1-16) peptide was modeled with all torsion degrees of freedom. According to the variable target function principle (54), all NOE-based distance constraints were first checked for the number of free torsions that separate every particular proton pair. The corresponding number is further referred to as torsion separation. The simulated annealing started from an arbitrary extended conformation obtained by an unrestrained MD simulation under high temperature (7000 K). At the beginning, only constraints of torsion separation one were applied. The following separation levels were added one by one in the course of the protocol when reasonable convergence was achieved for all previous torsion separation levels. As the ICMD protocol does not use pseudoatoms, instead all candidate proton pairs corresponding to a given resonance were analyzed from time to time, and the corresponding constraint was reassigned to the pair with the shortest distance in the current conformation. 4 The structure calculation of the A␤-(1-16)-Zn 2ϩ 1:1 complex was carried out using a strategy including two independent steps. In the first step, the structure of the polypeptide chain only was refined in calculations performed without explicit Zn 2ϩ ion and without any assumption on zinc coordination. The second calculation step was used in order to obtain a specific peptide-zinc complex structure with a metal coordination geometry satisfying chemical requirements. The experimental set of constraints used in the first step was updated with additional distance and angle constraints that enforced a tetrahedral ligand coordination of the Zn 2ϩ ion. To this end, we added four ambiguous constraints that linked the Zn 2ϩ ion with all possible partners. These constraints were arbitrarily assigned the torsion separation one, and they were treated along with other constraints, as explained above. Their geometry was derived from zinc-binding sites of relevant high resolution x-ray structures available in the Protein Data Bank. The following atoms and groups present in the A␤-(1-16) peptide were considered as potential zinc chelators: N-␦ and/or N-⑀ atoms of His 6 , His 13 , and His 14 , O-␦1-O-␦2 atoms of Asp 1 and Asp 7 , and O-⑀1-O-⑀2 atoms of Glu 3 and Glu 11 . These Zn 2ϩ ligand hypotheses were in agreement with those authorized by the NMR data obtained. The 2.0 Å harmonic distance constraints were applied for the distance between the Zn 2ϩ ion and each of its four chelators. If one of the histidine nitrogens took part in the complex, the Zn 2ϩ ion was kept in the plane of the imidazole ring and in the bissector plane of the corresponding nitrogen atom by using additional angle constraints. In this case the other nitrogen of the same imidazole ring was temporarily excluded from the list of possible chelators. In a similar way, an appropriate symmetrical orientation of Zn 2ϩ with respect to the Asp and Glu carboxyl groups was ensured. The calculated conformers were analyzed using MOLMOL The modified peptide mixtures resulting from the incubation were analyzed at different incubation times by RP-HPLC, as described above, using a linear gradient of 13-18% ACN in 0.1% aqueous trifluoroacetic acid over 40 min at a flow rate of 1 ml/min.

Isolation of A␤-(1-16) Isomers Produced from in Vitro
Aging-The modified peptides resulting from incubation of A␤-(1-16) at 70°C for 14 days were purified by semi-preparative RP-HPLC on a C-18 Uptisphere, 5 m, 7.8 ϫ 250-mm column (Interchim, Montluçon, France) with a linear gradient of 27-35% MeOH in 0.1% aqueous trifluoroacetic acid over 40 min at a flow-rate of 2 ml/min. Absorbance was monitored at 226 nm. Purity of the isolated species was checked by RP-HPLC and ESI-MS (Q-STAR Pulsar, Applied Biosystems, Courtaboeuf, France).

Identification of A␤-(1-16) Isomers Produced from in Vitro
Aging-Truncated peptides were identified by ESI-MS in positive mode (Q-STAR Pulsar, Applied Biosystems, Courtaboeuf, France). The peptide sequences were checked by collision-induced dissociation experiments. Isoaspartic acid residues within A␤-(1-16) isomers were quantified by enzymatic methylation catalyzed by the L-isoaspartyl methyltransferase, using the Isoquant protein deamidation kit (Promega, Charbonnieres, France), as described previously (61). Briefly, 10 l of a 7.5 M peptide solution in water were added to the reaction buffer (NaH 2 PO 4 /Na 2 HPO 4 125 mM, pH 6.8, 1.25 mM EGTA, 0.005% NaN 3 , 0.2% Triton X-100) containing L-isoaspartyl methyltransferase and 0.1 mM S-adenosyl-L-methionine. Enzymatic reactions were performed in duplicate at 30°C for 30 min and stopped by adding 10 l of 0.3 M phosphoric acid. Solutions were centrifuged at 10,000 ϫ g for 8 4 T. Malliavin et al., manuscript in preparation. min at 4°C. Supernatants were kept at 4°C until RP-HPLC analyses, which were performed within the day using a C-18 Uptisphere, 5 m, 4.6 ϫ 250-mm column (Interchim, Montluçon, France) eluted with a linear gradient of 10 -30% MeOH in NaH 2 PO 4 /Na 2 HPO 4 10 mM, pH 6.2, over 20 min at a flow-rate of 1 ml/min. Absorbance was monitored at 260 nm. Iso-Asp residues within the incubation products were quantified by plotting the S-adenosylhomocysteine peak intensity on a standard curve drawn up from RP-HPLC profiles of different S-adenosylhomocysteine dilutions.
Before N-terminal sequencing, partial hydrolysis of the A␤-(1-16) isomers was performed in trifluoroacetic acid at 48°C (62). The peptides cleaved between residues 1 and 2 were purified by RP-HPLC (Uptisphere, Interchim, 5 m, 4.6 ϫ 250 mm) with a linear gradient of 10 -25% ACN in 0.1% aqueous trifluoroacetic acid over 15 min, at a flow rate of 1 ml/min. They were then resuspended in MilliQ TM water (Millipore, Saint-Quentin-en-Yvelines, France), and each sample was loaded and argon-dried on a Biobrene-coated filter before being subjected to Edman degradation on a Procise 492 automatic protein sequencer (PerkinElmer Life Sciences).
Absolute configuration of the amino acids of the A␤-(1-16) isomers was determined by gas chromatography (GC) analysis of the N-triflu-oroacetylisopropyl ester derivatives of the amino acids present in the total hydrolysates, as described previously (63). GC analyses were performed with a 5890 series II chromatograph (Hewlett-Packard, Les Ulis, France) equipped with a flame ionization detector, on a 25-m length, 0.2-mm internal diameter, 0.12-m film thickness Chirasil-L-Val (N-propionyl-L-valine tert-butylamide polysiloxane) quartz capillary column (Varian, Les Ulis, France), using helium as carrier gas at a flow rate of 1.25 ml/min. Injector and detector temperatures were maintained at 300°C, and the oven temperature was programmed from 50 to 310°C at a rate of 3°C/min and then kept at 310°C for 8 min. The L-or D-configuration of the amino acids was determined by comparing the GC profiles obtained with those of standard L-or D-amino acids derivatized in the same conditions.  tion of the peptide at such concentrations in agreement with our previous CD data (34).

Structures of A␤-(1-16) and A␤-(1-16)-Zn 2ϩ in Aqueous Solution
Spin system identification and complete NMR 1 H assignments (supplemental Table A) were generally obtained from DQF-COSY, TOCSY, and NOESY experiments, using the sequential assignment protocol (64). Spectra recorded at 278 K were primarily used for assignments, and spectra measured at other temperatures, as well as HSQC and HMBC data, were used to resolve ambiguities (see 13 C assignments in supplemental Table B).
The conformational parameters were analyzed for A␤-(1-16) at pH 7.4. They were not indicative of any regular secondary structure elements in the peptide conformation (Fig. 1). However, the dNN(i,i ϩ 1) and d␣ N(i,i ϩ 2) NOEs displayed in the regions 2-5 and 7-13, and in particular the intense dNN between Gly 9 -Tyr 10 , Tyr 10 -Glu 11 , and Glu 11 -Val 12 (data not shown), suggested that these regions folded into a noncanonical structure.
The preliminary analysis of A␤-(1-16) by CD showed that it formed a soluble and stable complex with Zn 2ϩ ions in the pH 6.0 -8.0 range, leading to a conformational change, whereas the homologous peptide with unprotected N and C termini precipitated upon zinc addition (34). Therefore, A␤-(1-16) was chosen for a complete NMR analysis in the absence and in the presence of Zn 2ϩ , in order to characterize the structural change of the N-terminal region 1-16 of A␤ upon zinc binding. The NMR conformational parameters were acquired at 278 K to ensure observation of the amide protons of His 6 and His 14 , which were not observable at higher temperatures, due to amide proton exchange in the considered pH range. Given the potential contribution of the three histidines to the zinc coordination sphere, their pK a values were first measured in order to determine whether these residues were in ionized or neutral form under physiological pH values. The pK a values were calculated from nonlinear least square fit of H-␦2 and H-⑀1 proton titration curves to the Henderson-Hasselbalch equation, and the values 7.0, 6.9 and 6.8 Ϯ 0.1 were assigned to His 6 , His 13 , and His 14 , respectively. An equilibrium between protonated and deprotonated imidazole rings was observed in the 6.0 -8.0 pH range (supplemental Fig. S1) that corresponded to the pH domain where the zinc-induced conformational change and the soluble A␤-(1-16)-Zn 2ϩ complex had been previously characterized by CD (34). Addition of Zn 2ϩ to a 2 mM A␤-(1-16) sample, either in phosphate buffer at pH Ͼ6 or in Tris-d 11 /HCl buffer at pH 5.8 led to the formation of a low amount of precipitate. From our previous analysis of the complex by CD and ESI-MS, this precipitate could be identified as the insoluble fraction of the A␤-(1-16)-Zn 2ϩ 1:1 complex formed at the higher concentration used for NMR. The supernatant was thus submitted to the NMR experiments. One-dimensional 1 H NMR recordings over the 0.3-3.5 mM range of concentrations discarded the presence of a change in the aggregation state and was in favor of a monomeric state of the 1:1 complex at the 2 mM peptide concentration used for the structure analysis. The 1 H one-dimensional NMR spectra of A␤-(1-16) recorded either in the absence or in the presence of 1.5 Zn 2ϩ equivalents in sodium phosphate buffer at pH 6.5 and 7.5 did not illustrate any significant Zn 2ϩ -induced change in the chemical shifts (Fig. 2, A--D). However, a strong and selective peak broadening was observed at pH 7.5 (Fig. 2, D and F). This mainly concerned region 5-15, whereas the N-terminal region 1-4 remained unaffected. The three histidines were particularly affected, with a complete disappearance of the signals involving the amide protons and H-␦2 and H-⑀1. In addition, the spin systems of Tyr 10 , Glu 11 , and Val 12 nearly disappeared.
Such a peak broadening, similar to that already observed by Curtain et al. (28) upon adding Zn 2ϩ to a solution of A␤-(1-28) could be indicative of exchange between zinc-complexed and uncomplexed A␤-(1-16). The absence of any change in the rest of the spectrum was in agreement with the monomeric state of the complex analyzed and indicated that the presence of soluble oligomers that could have resulted in broadened resonances was negligible. Finally, pH 6.5, which allowed good quality spectra and was still included in the protonation/deprotonation pH range of the three histidines (supplemental Fig. S1), was selected for investigation of the three-dimensional structure of the apoA␤-(1-16) and the A␤-(1-16)-Zn 2ϩ complex. Although most of the conformational parameters of A␤-(1-16), i.e. H-␣-and C-␣-CSD, 3 J HN-H␣ and ⌬␦/⌬T HN , were unchanged upon Zn 2ϩ addition, a number of additional specific NOEs were observed in the presence of Zn 2ϩ (Fig. 3), which suggested a structural change.
The three-dimensional structures of the apoA␤-(1-16) and the A␤-(1-16)-Zn 2ϩ complex were calculated by NMR-restrained molecular modeling, using the constraints summarized in Table 1. The apoA␤-(1-16) appeared poorly structured in the N-terminal 1-6 region (Fig. 4A), whereas an irregular structure was characterized in the 7-15 part, including a turn centered at residues 7-8. The 20 selected structures displayed a good fit of the side chains in the C-terminal region. Their orientation suggested the presence of stabilizing intramolecular hydrogen bonds involving the pairs Asp 7 /Lys 16 and Glu 11 /His 13 , as well as the backbone carbonyl group of His 14 and the Tyr 10 side chain (Fig. 4B). In the presence of Zn 2ϩ , the modeling of the A␤-(1-16) structure using X-PLOR or ICMD programs without introducing any constraints related to the cation binding resulted in a more compact structure (Fig. 4C). Compared with the apoA␤-(1-16) folding, the structure of the N-terminal region was better defined, and a global reorientation of most side chains toward the inside of the structure was observed, which particularly concerned those of all three histidines (Fig. 4D). The ICMD protocol (53) was used to assign the peptide/Zn 2ϩ -binding sites and obtain the three-dimensional structure of the A␤-(1-16)-Zn 2ϩ complex (Fig. 4, E and F, and Fig. 5). In most of the selected structures, the Asp 7 -Gly 9 region adopted an irregular 3 10 helical structure (Fig. 4F). The Zn(II) cation was tetrahedrally coordinated to His 6 , His 13 , and His 14 through their N-␦1, N-⑀2, and N-␦1 atoms, respectively, and to Glu 11 through its carboxylate. It is worth noting the coordination sphere proposed from our calculation conveniently fits the 111 Cd NMR data obtained by Mekmouche et al. (36). In all the selected structures, the Phe 4 residue appeared to be located in the inner core of the structure, and Tyr 10 was systematically located on the face opposed to the zinc atom. The binding sites thus obtained corresponded to the more broadened spin systems displayed on the TOCSY spectrum registered in the presence of Zn 2ϩ (Fig. 2, D and F). An analysis of the surface features of the complex (Fig. 5, C and D) showed that most of the surface is hydrophobic or neutral, with a negatively charged patch located near the N terminus (Asp 1 and Glu 3 ) and a single restricted positively charged region due to Lys 16 . This is in agreement with the fact that all positively charged residues present in A␤-(1-16) reoriented toward the inner core of the complex.
In Vitro Aging of A␤- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) in the Absence and in the Presence of Zn 2ϩ Ions and Subsequent Isomerizations and Racemizations-In vitro aging of A␤-(1-16) was performed by incubation at pH 7.4 and 37°C, either in the absence or in the presence of Zn 2ϩ . In order to accelerate the rate of succinimide formation (38) and to facilitate the identification of a maximal range of modified species, another set of experiments was conducted at 70°C. The aging experiments at 37°C were conducted over an incubation period of 70 and 130 days in the absence and in the presence of Zn 2ϩ , respectively (Fig. 6, A and B), whereas this incubation period could be shortened to 14 days at 70°C. The chromatograms obtained in the absence and in the presence of EDTA were similar (data not shown), ruling out the contribution to the chemical modifications observed of trace metal contaminants, such as copper-induced redox processes.
The Asp 1 -and Asp 7 -isomerized and/or racemized species formed at 70°C (Fig. 6C) were identified from a panel of experiments, including (i) L-isoaspartyl methyltransferase-assisted quantification of isoaspartate residues, (ii) Edman sequencing, and (iii) GC analysis on a chiral capil-  (1-16)-Zn 2؉ complex at pH 6.5 calculated with the ICMD protocol. A and B, views showing the location of the zinc ion, which is coordinated through the imidazole nitrogens of the three histidine, His 6 , His 13 , and His 14 , side chains and the carboxylate of the Glu 11 side chain. C and D, corresponding electrostatic molecular surface of the complex; C, view is rotated 180°about the x axis relative to the A view. Blue and red correspond to negatively and positively charged areas, respectively. lary column of the derivatized amino acids obtained in the total acid hydrolysate ( Table 2). The A, D, and E species, which are the three major isomers produced, were assigned to A␤-(1-16)-L-iso-Asp 1,7 , A␤-(1-16)-L-Iso-Asp 7 , and A␤-(1-16)-L-iso-Asp 1 , respectively. Furthermore, liquid chromatography-MS analysis revealed the presence of N-or C-truncated peptides (Fig. 6C, circled peak), which were identified from both their m/z ratios and MS/MS data as A␤-(2-16), A␤-(1-7), and A␤-(1-6) (data not shown).
At 37°C, the truncated peptides appeared favored as compared with the isomerized species, A␤(5-16) being the major modified species from 20 days of incubation ( Fig. 6A and 7A). In addition, the racemizations of L-Asp or L-iso-Asp residues to their D-isomers were totally absent. The presence of Zn 2ϩ ions in the incubation medium significantly altered the kinetic profile of the in vitro aging at 37°C by disfavoring truncations at the advantage of isoaspartate formation at positions 1 and 7 in the sequence (Fig. 6B and Fig. 7B). By contrast, the presence of Zn 2ϩ ions in the incubation medium was without effect on the nature and the relative abundances of the species formed from A␤-(1-16) at 70°C (data not shown).
The conformational parameters of A␤-(1-16)-L-iso-Asp 7 , one of the major species formed in the presence of Zn 2ϩ , were quite similar to those of A␤-(1-16) (Fig. 1). Local changes in the His 6 -Ser 8 region were observed because of the presence of L-iso-Asp 7 , as illustrated by the CSD profiles (Fig. 1, B and C). Furthermore, as already observed for L-iso-Asp-containing proteins (65), this residue triggered a change in the intensity of sequential NOEs between residues 7 and 8, with a strengthened d␤ N connectivity and a weakened d␣ N connectivity (data not shown). However, the observed changes in the NMR parameters appeared to be local and not to express a perturbation of the global peptide fold. One-dimensional 1 H and TOCSY spectra of A␤-(1-16)-Liso-Asp 7 were performed in the presence of Zn 2ϩ ions in order to assess the effect of isoaspartate on the A␤-(1-16)-Zn 2ϩ interaction. Because a precipitate formed at pH values above 6, the spectra were recorded at pH 5.8. They showed a particularly strong broadening of the L-iso-Asp 7 signals, which almost disappeared from the spectra; His 6 and residues 10 -14 were also affected (data not shown). These results indicated that the modified L-iso-Asp 7 residue did not prevent the interaction and even was contributing to the zinc coordination.

DISCUSSION
The N-terminal 1-16 part of A␤ is a flexible region that remains accessible in amyloid deposits (14,15) and is involved in the zinc binding capacity and protein aging propensity of A␤. In this study, we characterized for the first time the three-dimensional structure of the soluble A␤-(1-16)-Zn 2ϩ complex as well as the chemical modifications of this region of A␤ upon protein aging.
Structuration of A␤- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16) Upon Zinc Binding-In phosphate buffer at pH 6.5, the apoA␤-(1-16) shows a rather well defined structure, particularly in region 7-15, but without any canonical element. The addition of Zn 2ϩ to the solution mainly affects the Tyr 10 -Gln 15 region and particularly the H-␦2 and H-⑀1 resonances of all three histidines, in agreement with previous data on A␤-(1-28) (28). The A␤-(1-16) struc-  Table 2), with peak G being unmodified A␤- (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13)(14)(15)(16). The MeOH percentages in the gradients used for separations are represented as dotted lines. ture becomes more compact upon zinc binding. The three-dimensional structure of the A␤-(1-16)-Zn 2ϩ soluble 1:1 complex calculated from the NMR constraints reveals His 6 , Glu 11 , His 13 , and His 14 as the four ligands involved in the zinc coordination sphere. Such a motif of zinc attachment is reminiscent of a motif identified previously in the case of short peptides designed to bind zinc ions (66). The identification of the three histidines as the zinc ligands does not contradict previous studies using various spectroscopic techniques, which proposed that one to all three histidines in this region of A␤ acted as metal-binding sites (31)(32)(33)36). The identification of the fourth zinc chelator is much more debated. The present calculation of the complex identifies Glu 11 as the fourth partner, acting through its carboxylate side chain. In our previous MS study on the A␤-(1-16)-Zn 2ϩ complex (35), Arg 5 was proposed as the additional ligand. This discrepancy with the present work is most likely attributable to the different acidic/basic properties of the amino acids in the gas phase and in solution (67). Tyr 10 has been also suggested previously as a possible ligand participating in Cu 2ϩ or Zn 2ϩ chelation by A␤ (28,32). In the structure obtained here for the A␤-(1-16)-Zn 2ϩ soluble 1:1 complex, Tyr 10 is excluded from the coordination sphere and is located on the opposite face from the Zn 2ϩ ion. This is in agreement with Mekmouche et al. (36), who also excluded the involvement of Tyr 10 in the coordination sphere from both NMR and absorption spectroscopy data. In this last study, Asp 1 was proposed as the fourth coordination site of the zinc ion in A␤-(1-16) at pH 8.7 (36). This is not consistent with our NMR data at pH 6.5 and 7.5. Those exclude Asp 1 as metal chelator, because (i) the corresponding NMR signals were not broadened upon zinc addition, and (ii) this residue was not selected as contributing to the zinc coordination sphere in our structure calculations, although it was considered as a potential ligand. The lower pH and the N-terminal acetylation of A␤-(1-16) in our study can contribute to understanding this difference.
In the absence of zinc, A␤-(1-16) displays a bend in its central region, as illustrated by the intense dNN(i,i ϩ 1) NOEs between Gly 9 and Val 12 . This bending has also been described for the full-length A␤ and has been proposed to contribute to A␤ fibrillation by promoting the formation of salt bridges involving the histidine and aspartic/glutamic acid side chains (69). From our study, zinc binding appears to stabilize this bended structure, which could reinforce the hypothesis of the involvement of zinc in A␤ aggregation (19 -21, 36). However, the nonfibrillar nature of A␤ aggregates formed upon zinc addition (26,27) contradicts this interpretation. Furthermore, the involvement of the histidines in zinc binding would rather compete with the formation of salt bridges with aspartic/glutamic acid side chains. Therefore, we propose that zinc binding would not be implicated in A␤ fibrillogenesis but would rather occur after amyloid deposition as a result of the accessibility of the A␤ N-terminal region within fibrils. In this context, zinc binding-induced aggregation of A␤ would result from another mechanism, such as the formation of intermolecular histidine-metal-histidine bridges (18,28).
Our CD (34) and NMR data have shown that A␤-(1-16) is monomeric in the 5 M to 1 mM concentration range. The absence of oligomerization upon zinc binding has been shown previously by size exclusion chromatography (36). This trend is also emphasized by the absence of variation in A␤-(1-16) NMR chemical shifts and of unspecific NMR signal broadening upon zinc addition. Similarly, Curtain et al. (28) have characterized 1:1 monomeric complexes with copper for A␤-(1-28) and A␤- . In the conditions of our study, zinc binding to A␤-(1-16) appears to lead to the soluble monomeric complex accompanied by a small amount of insoluble aggregates, as proposed by Miura et al. (32).
Protein Aging-induced Modifications of A␤-(1-16)-Because age-related peptide modifications are suggested to participate in the AD pathogenesis, in vitro aging of A␤-(1-16) has been investigated here. Numerous modifications of the synthetic peptide are observed, i.e. truncations, isomerizations, and racemizations. Truncated peptides, including an N-terminal pyroglutamate at positions 3 and 11 of A␤, are species that have been detected within amyloid fibrils (71). Such species are not formed here during the in vitro aging process of A␤-(1-16) or A␤-(1-16) hemi at 37°C. Thus, the truncated molecular forms detected within fibrils presumably represent by-products of metabolic intermediates toward degradation and are not produced spontaneously upon protein aging.
The incorporation of iso-Asp residues is suggested to increase the tendency to form ␤-sheet in amyloid peptides (41,70). Here we observe that the isomerization of L-Asp 7 to L-iso-Asp leads to a local conformational change in the region His 6 -Ser 8 but does not affect the overall structure of the peptide. More interestingly, the presence of Zn 2ϩ ions significantly changes the chemical modification profile of A␤-(1-16) upon the in vitro aging process at 37°C, by favoring isomerizations at the expense of truncations. This trend suggests that the association of

Identification of the isomers of A␤-(1-16) produced upon in vitro aging at pH 7.4 and 70°C
The outcome of the L-isoaspartyl methyltransferase (PIMT)-catalyzed methylation, Edman sequencing, and chiral GC results is shown. The asterisk indicates a yield decrease in Edman sequencing.

Isomer
No the modified peptides A␤-(1-16)-L-iso-Asp 1 , A␤-(1-16)-L-iso-Asp 7 , and A␤-(1-16)-L-iso-Asp 1,7 with Zn 2ϩ ions displaces the equilibrium of the A␤-(1-16) isomerization reactions toward the formation of L-iso-Asp species. Furthermore, A␤-(1-16)-L-iso-Asp 7 , one of the two major modified species generated upon protein aging, appears to bind Zn 2ϩ with a different coordination mode, involving the L-iso-Asp 7 residue itself. Finally, the presence of a peptide bond before Asp 1 does not affect the propensity of this residue to undergo isomerizations/racemizations, because the nonacetylated A␤-(1-16) hemi also exhibits aging-related modifications at Asp 1 and Asp 7 . The impact of the aging-related modifications of the aspartate residues on both the conformation of A␤-(1-16) and its zinc binding capacity has been subsequently assessed here. Altogether, our results reflect the high potential of chemical and conformational modifications of region 1-16 of A␤ upon both aging and zinc binding, which could partly account for its important structural heterogeneity within amyloid fibrils.
Implications for Therapeutic Approaches with Region 1-16 of A␤ as Target-The occurrence of truncated, isomerized, and racemized species of A␤ and of zinc-bound species within amyloid deposits must be considered in the frame of a therapeutic approach targeting the N-terminal region of A␤. On the one hand, the modified species of A␤ could be used as immunological targets, while avoiding autoimmune response because of the recognition of the soluble A␤. The established potentiality to target specifically a particular isomer of A␤ present within amyloid deposits (43) suggests the applicability of such an approach to isomerized species of A␤. On the other hand, the high zinc content within amyloid plaques might not be adverse but even favorable to the interaction of A␤ with anti-A␤ antibodies, which constitutes a positive trend in the frame of an epitope-based vaccination approach (16,47).