Báo cáo khoa học: High-resolution NMR studies of the zinc-binding site of the Alzheimer’s amyloid b-peptide pdf

Interactions of copper and zinc with Ab induce peptide aggregation if the metal concentration is high enough, i.e.. On the other hand, binding of low concentrations of copper or zinc ion

of the Alzheimer’s amyloid b-peptide

Jens Danielsson1, Roberta Pierattelli2, Lucia Banci2 and Astrid Gra¨slund1

1 Department of Biochemistry and Biophysics, Stockholm University, Sweden

2 Department of Chemistry and Magnetic Resonance Center, Universita` di Firenze, Sesto Fiorentino, Italy

The amyloid b-peptide (Ab) is the major component

of senile plaques and soluble oligomers, amyloid

b-derived diffusible ligands, which are considered to

play an important role in Alzheimer’s disease (AD)

pathology Ab is the product of cleavage of amyloid

precursor protein (APP) [1] at position(s) 39–42, which

creates a soluble monomer

There is evidence that APP is involved in copper

homeostasis APP has a selective metal-binding site

and is able to reduce bound Cu(II) to Cu(I) It also

participates in the regulation of copper levels, and its

expression is affected by copper concentration Also

other metal ions, such as Zn2+ and Fe2+, are known

to interact with APP [2–6]

The Ab peptide mainly appears as a random coil

in aqueous solution, but contains some secondary structure elements: A poly(proline II) helix (PII) in the N-terminus, and two b-strands in the central part and in the C-terminus [7–9] The soluble mono-meric peptide has a high tendency to form Ab oligo-mers, which eventually produce Ab fibrils Although the monomeric peptide is not neurotoxic and neither are the fibrils, Ab oligomers have been shown to induce cognitive loss due to neurodegeneration, via

Keywords

aggregation; Alzheimer amyloid b-peptide;

copper binding; NMR; zinc binding

Correspondence

A Gra¨slund, Department of Biochemistry

and Biophysics, Stockholm University,

S-106 91 Stockholm, Sweden

Fax: +46 8 155597

Tel: +46 8 162450

E-mail: astrid@dbb.su.se

(Received 22 June 2006, revised 20 October

2006, accepted 27 October 2006)

doi:10.1111/j.1742-4658.2006.05563.x

Metal binding to the amyloid b-peptide is suggested to be involved in the pathogenesis of Alzheimer’s disease We used high-resolution NMR to study zinc binding to amyloid b-peptide 1–40 at physiologic pH Metal binding induces a structural change in the peptide, which is in chemical exchange on an intermediate rate, between the apo-form and the holo-form, with respect to the NMR timescale This causes loss of NMR signals in the resonances affected by the binding Heteronuclear correlation experiments,

15N-relaxation and amide proton exchange experiments on amyloid b-pep-tide 1–40 revealed that zinc binding involves the three histidines (residues 6,

13 and 14) and the N-terminus, similar to a previously proposed copper-binding site [Syme CD, Nadal RC, Rigby SE, Viles JH (2004) J Biol Chem

279, 18169–18177] Fluorescence experiments show that zinc shares a com-mon binding site with copper and that the metals have similar affinities for amyloid b-peptide The dissociation constant Kdof zinc for the fragment a-myloid b-peptide 1–28 was measured by fluorescence, using competitive binding studies, and that for amyloid b-peptide 1–40 was measured by NMR Both methods gave Kd values in the micromolar range at pH 7.2 and 286 K Zinc also has a second, weaker binding site involving residues between 23 and 28 At high metal ion concentrations, the metal-induced aggregation should mainly have an electrostatic origin from decreased repulsion between peptides At low metal ion concentrations, on the other hand, the metal-induced structure of the peptide counteracts aggregation

Abbreviations

Ab, amyloid b-peptide; AD, Alzheimer’s disease; APP, Alzheimer precursor protein; HSQC, heteronuclear single quantum coherence; LMCT, ligand to metal charge transfer.

pathophysiologic mechanisms that are not completely

understood The Ab oligomers are thus more

neuro-toxic than the Ab fibrils [10–15] The detailed

struc-ture of these Ab oligomers has yet to be resolved,

although they seem to be similar to other

amyloid-forming peptide oligomers [16]

The aggregation process is accompanied by a

signifi-cant change in structure, whereby monomeric, mostly

unstructured Ab folds to form oligomeric b-sheet-rich

forms The detailed mechanism of this transition is not

known

Increased metal concentrations (mainly copper, iron

and zinc) have been found in the brains of AD

patients, both in the amyloid plaques (copper and zinc)

and in the cortical tissue (zinc) [17,18] Interactions of

copper and zinc with Ab induce peptide aggregation if

the metal concentration is high enough, i.e > 1 : 1

metal⁄ peptide ratio [19–21] The formed aggregate is

suggested to be amorphous, and thus high

concentra-tions of copper and zinc prevent fibril formation by

promoting the formation of nonfibril aggregates

[20,22–24] The N-terminal fragment Ab(1–28) has,

however, been shown to undergo increased fibril

formation upon interaction with zinc [25] Another

N-terminal fragment of Ab, Ab(1–16), forms stable

oligomers in the presence of both copper and zinc [26]

On the other hand, binding of low concentrations of

copper or zinc ions to the full-length Ab(1–40) at less

than 1 : 5 metal⁄ peptide ratio reduces the Ab

oligo-meric stability and prevents aggregation, whether

amorphous or fibril forming [27,28] The effect of

metals on Ab is thus dependent on the experimental

conditions, such as pH, salt concentration and, most

important, metal concentrations [21,29] In addition,

Ab-bound Cu(II) may be reduced to Cu(I), and the

complex may produce hydrogen peroxide, which has

been suggested to be neurotoxic in AD [30]

Soluble monomeric Ab has a high-affinity

copper-binding site in the N-terminus, and the metal ion is

suggested to coordinate in a planar configuration with

the three histidines (6, 13 and 14) and the N-terminus

or Tyr10 N-terminal deletions alter the binding

affin-ity, suggesting that the N-terminal amide participates

in copper coordination and that Tyr10 does not [31–

34] The metal-binding site, involving the three

histi-dines, is also able to bind a zinc ion The fourth ligand

for zinc binding has been suggested to be either Tyr10,

Glu11, Arg5 or the N-terminus [35–37]

Both copper and zinc have been suggested to have a

second binding site The ligands of this hypothetical

weaker binding site are unknown [31,38,39]

The binding affinity of the metal ions to Ab is

pH-dependent: copper has a higher affinity under mild

acidic conditions, whereas the affinity for zinc is less pH-dependent over a range of pH values between 6.5 and 7.5 [29] At pH below 6, when the histidines are mainly protonated, no zinc binding occurs [21] Under physiologic conditions, Ab has a higher propensity to bind zinc, whereas under mildly acidic conditions, as

in physiologic acidosis following an inflammatory pro-cess, copper is preferentially bound [40] This differ-ence between copper and zinc affinity at different pH values has been suggested to be a pathogenic mechan-ism of Ab: under normal conditions, zinc protects against copper-induced Ab toxicity, which is induced

by physiologic acidosis [22,29,40]

The present study shows for the first time the detailed molecular effects on the full-length Ab of zinc binding Thanks to the combined use of heteronuclear NMR, fluorescence and CD spectroscopy, a molecular model for the zinc interaction in solution is obtained

We show that the high-affinity zinc-binding site is formed by three histidines and the N-terminus, whereas a second, low-affinity binding site comprises residues 23–28 The results of these studies are dis-cussed within the context of the general mechanism of the onset of pathologic states upon Ab aggregation Results

NMR spectroscopy Interaction with zinc) heterocorrelation experiments Ab(1–40) in aqueous solution has been extensively characterized by NMR spectroscopy, mainly through two-dimensional 1H–1H NMR and 1H–15 N-heteronuc-lear single quantum correlation (HSQC) experiments

In 1H–15N correlation experiments, some signals are affected by an exchange process with the solvent and are missing in the spectra Thus, despite the good reso-lution, not all resonances are visible in the 1H–15N HSQC spectra In contrast, 13C-bound protons are not affected by an exchange process and show a suitable signal dispersion, which allows high-resolution charac-terization Therefore, the combination of these two sets

of data (1H–15N and 1H–13C) provides a detailed char-acterization of metal binding to Ab As the metal ions could be sequestered from solution by complexation with the buffer, we can only obtain an estimate of the order of magnitude of the dissociation constants On the other hand, from NMR we obtain an unambigu-ous description of the binding

When zinc ions are added to a solution of Ab(1–40), all the signals in 1H–15N and 1H–13C HSQC experi-ments are reduced in intensity At one equimolar zinc concentration, the remaining intensity fraction for the

CHasignals of the residues in the C-terminal region is

0.69 ± 0.04, whereas His13 has a relative intensity of

0.15 ± 0.02, and His6 and His14 have relative

intensi-ties of 0.3 ± 0.01 This observation confirms the

involvement of the histidines as metal ligands Asp1

shows a 0.60 residual signal intensity, whereas Val12

has a residual signal intensity of 0.52 ± 0.06 The

sig-nal reduction of Asp1 is equal to the reduction seen

for other residues located in the N-terminal region,

such as Ser8 and Arg5 Asp1, His6, Val12, His13 and

His14 also show small but significant chemical shift

changes (0.010–0.015 p.p.m in the proton dimension)

upon zinc binding (Fig 1) The chemical shift

varia-tions are probably the result of conformational chan-ges of the backbone upon metal binding For other residues, no chemical shift changes are observed The

Cbregion shows the same signal intensity reduction pattern but with no or very small induced chemical shift changes The selective spectral changes observed upon zinc addition for Asp1, His6, His13 and His14 clearly indicate the involvement of these residues in metal coordination In particular, the specific changes

in the Asp1 chemical shift provide a direct indication

of the involvement of the N-terminus in zinc binding The chemical shift change observed for Val12 is prob-ably due to its proximity to the histidine ligands

V12

D1

Y10

H6/H14 H13

Fig 1 1 H– 13 C HSQC spectra of the C a region of 50 l M Ab(1–40) in 10 m M phosphate buffer at pH 7.4 and 286 K without (black) and with (red) 30 l M zinc as chloride All resonance peaks show reduced intensity, but some specific reduction is present for the histidines, Asp1, Tyr10 and Val12 The resonances of His6, His13, His14, Asp1 and Val12 exhibit induced chemical shift changes upon zinc binding The inset shows the broadening and chemical shift changes of Asp1 upon zinc titration with 0, 20 and 50 l M zinc, indicating the induced chemical shift changes.

The aromatic region of the1H–13C HSQC spectrum

is particularly informative about the effects of Zn2+

(Fig 2) The Tyr10 aromatic resonances are unchanged

upon zinc addition, indicating that the aromatic ring of

Tyr10 is not involved in binding of the metal Instead,

the e and d resonances of the histidines show significant

intensity losses, of > 80% The d-proton signals of His6

and His13 show some induced chemical shift changes,

whereas His14 only exhibits intensity loss for the

d-pro-ton signal (Fig 2A) Furthermore, there is a total loss of

the e-protons of the His6 and His14 signals and a

weak-ening (> 80% reduction) of His13 resonance (Fig 2B)

The zinc binding can thus be envisaged as coordination

of the metal ion by the nitrogens of the imidazole rings

of the histidines and the amino group in the N-terminus Use of the signal intensity reduction of the 1H–13C HSQC crosspeaks to estimate the dissociation constant

of zinc yields Kd 1 lm at pH 7.2, in phosphate buf-fer, at 286 K

Also, the 1H–15N HSQC spectrum shows that the reduction in intensity does not affect all the signals to the same extent This signal reduction, where the signal

is lost due to addition of zinc, could as a first approxi-mation be interpreted in terms of a first-order binding process An apparent dissociation constant appKd* can

be calculated; the star indicates that the dissociation

Y10

Y10

A

B

H14

H6

Fig 2 1 H– 13 C HSQC spectra of the

aroma-tic regions of 50 l M Ab(1–40) in 10 m M

phosphate buffer at pH 7.4 and 286 K

with-out (black) and with (red) 30 l M zinc as

chloride (red) The d and e resonances of

the histidines shown in (A) and (B),

respect-ively, show significant signal intensity

los-ses All three histidines are affected In (A),

it is shown that H6 and H13 show some

induced shift but H14 only exhibits signal

intensity loss for the d-protons The

e-pro-tons shown in (B) exhibit a total loss of H6

and H14 resonances, and a weak H13

res-onance remains at the same chemical shift.

The tyrosine aromatic crosspeaks are

unaf-fected upon zinc addition.

constant calculated here is not a quantitative measure

of binding of zinc by the molecule, but rather a

meas-ure of which residues are most involved in the binding

Such an apparent dissociation constant was calculated

for all individual resonances (Fig 3A) The apparent

dissociation constant varies along the peptide chain,

indicating that this is a generalized description of

binding, which in turn reflects the induced changes in

the peptide upon binding The C-terminus residues 29–40,

which are not directly affected by the zinc interaction,

show a fairly constant apparent dissociation constant,

but four signals, originating from residues 23, 24, 26

and 29, are significantly more affected In the

N-termi-nus, the regions corresponding to the binding site show

a very high apparent affinity constant With a simpler approach, the decrease in intensity of the crosspeaks can be studied, and using this approach a similar pat-tern is observed (data not shown) It can be seen that,

in addition to a reduction over the entire polypeptide sequence, a more dramatic effect is observed for the first 15 residues This behavior is similar to that observed with 13C HSQC, and can be interpreted in the same terms as specific binding of the metal in the N-terminal part of the peptide

These observations confirm that the zinc-binding site

is in the N-terminus, and are consistent with three his-tidines participating in the binding, possibly with the formation of a bend or a turn in region 7–12 and one

Kd

R2

A

B

C

10

0

0

-4

-8

0.4

0.2

0

0.4

0.2

0

residue

D

** * **

** * **

* * * **

* * * * *

Fig 3 (A) 1 H– 15 N HSQC signal intensity of

50 l M Ab(1–40) in 10 m M phosphate buffer (pH 7.2) at 286 K The calculated apparent dissociation constant, app Kd*, of zinc for Ab.

A second interaction site is suggested in residues 23, 24, 26 and 28 The residues indicated with an asterisk (*) were not observed, due to rapid exchange with sol-vent water (B) Change in transverse relaxa-tion rate, R2, upon addition of zinc under the same conditions as in (A) The increase in

R2upon addition of zinc (negative differ-ence) suggests greater rigidity at the binding site (C) The amide proton exchange vari-ation with temperature measured as des-cribed in the text The amide proton stability

is increased along the whole peptide How-ever, the stability increase is more promin-ent close to the binding site This approach shows a good correlation with relaxation data (D) Signal intensity decrease upon cop-per binding The fractional remaining inten-sity after addition of 20 l M copper to 50 l M

Ab Copper binding shows the same pattern

as zinc binding.

in region 2–5 The 1H–13C HSQC and 1H–15N HSQC

results show that Tyr10 does not participate in zinc

binding, supporting previous proposals for copper

binding [25,31,32] The presence of a second binding

site for zinc is suggested by the selective effect for

some residues in the C-terminal region

The general signal intensity reduction in the1H–15N

HSQC spectra due to zinc binding could have multiple

origins, such as aggregation, the presence of chemical

exchange rate between conformations in an

intermedi-ate regime with respect to the NMR timescale,

increased amide proton exchange rate in the binding

region, and⁄ or increased relaxation rates due to

decreased dynamics in the binding region We favor

aggregation as the main explanation for the general

intensity loss, as this should give signal reduction due

to the line-broadening of the large aggregates

In the 1H–13C HSQC spectrum, there is some

cross-peak intensity left at high zinc concentrations, whereas

there is no residual crosspeak intensity in the amide

pro-ton region of the1H–15N HSQC spectra This different

feature suggests that in the 1H–15N HSQC case, there

are further mechanisms (in addition to aggregation and

metal-binding effects) that lead to specific reductions in

signal intensity, e.g relaxation effects and proton

exchange effects In order to vary the chemical exchange

rate and to be able to detect additional resonances,

a 1H–15N HSQC spectrum was recorded at a lower

temperature (278 K) As shown in Fig 4, four new

crosspeaks appeared, suggesting that some additional residues exhibit slower exchange at lower temperatures The intensity of the new peaks was too low to attempt the assignment of the newly identified resonances

Relaxation and diffusion measurements The local mobility changes upon zinc binding to Ab were studied by NMR relaxation measurements Amide HN R1 and R2 were measured The average R1 value for the peptide with no zinc bound is 2.16 ± 0.40 s)1 and is unchanged upon zinc binding This suggests that the peptide remains monomeric upon zinc binding This was also confirmed using pulsed field gradient-NMR diffusion studies Ab(1–40)’s diffusion coefficient is 1.04 ± 0.02· 10)10m2Æs)1 at

286 K, which is in fairly good agreement with the expected value of 1.09· 10)10 m2Æs)1, with the visco-sity changes accounted for by an empirical function [8,41] Upon zinc titration, no significant change of dif-fusion coefficient was measured (data not shown) All changes were within the range of experimental error, suggesting that no stable small soluble oligomers of the peptide were induced by zinc binding The zinc-induced peptide aggregates immediately grow to sizes where the linewidth in the NMR spectra broadens beyond detectable limits, so the relaxation and diffu-sion properties of the aggregate cannot be measured with NMR

G38 G29

G9 G25

S8

S26

N27

I31 V24

V39 V12 V18 L34

Y10

A30 R5

F19 L17 F20 V36 F4 D7 K28 M35

I32 A21

V40

Fig 4.1H–15N HSQC spectra of 75 l M

Ab(1–40) in 10 m M phosphate buffer at

pH 7.4 (278 K) and 50 l M ZnCl2 The

assign-ment is shown Four new crosspeaks

appear when the temperature is lowered

and the exchange rate is therefore slowed.

The new crosspeaks are highlighted in

circles.

The mean R2value of the peptide changes upon zinc

addition from 4.85 ± 1.40 s)1 to 7.12 ± 6.44 s)1

Figure 3B shows the changes in R2 upon zinc binding

for all the residues The C-terminal residues show no

or small changes in R2, whereas the residues close to

the binding site show increased R2 upon zinc binding,

as expected because of the presence of a zinc-induced

structure in this region To confirm that the R2

chan-ges are due to decreased local mobility, amide proton

exchange stability was measured As shown in Fig 3C,

the amide protons show slightly increased stability,

which is most prominent close to the binding site We

conclude that zinc binding induces increased order in

the N-terminus of the peptide This is reflected by the

increased R2 in this region, and also by the increased

amide proton stability, which is most likely due to

increased protection of the amide protons from the

solvent water

Interaction with copper

To obtain some complementary information, the

inter-action of copper was investigated using1H–15N HSQC

experiments and the paramagnetic broadening effect

expected from bound copper Figure 3D shows that

copper addition produced a significant decrease in the

crosspeak intensities for residues 3, 5, 8, 9, 10, 11, 12,

16 and 17, and a total loss of the Phe4 signal when

40% of the peptide is bound to copper The signals of

residues 7 and 15 also show a signal intensity decrease,

but not as prominent as for the others Crosspeaks for

residues 1, 6, 13 and 14 are not visible even with no

copper added As well as the selective effects described

above, there is a general copper-induced reduction of

the crosspeak intensities, which could be ascribed to

aggregation of the peptide At high copper

concentra-tions, the crosspeak signal of Ala21 is also lost This

confirms the presence of a second, weaker binding site

in the central part of the peptide All the signals

affec-ted by the addition of copper reappeared, even if not

completely, upon EDTA addition (EDTA

concentra-tion exceeded metal concentraconcentra-tion about 20-fold) (data

not shown)

Fluorescence spectroscopy

The peptide has a tyrosine residue in the N-terminal

part of the peptide, Tyr10, and the copper-binding site

is close to this site, as shown above Addition of

cop-per quenches the tyrosine fluorescence signal [31,38]

Fluorescence quenching can be used to directly

meas-ure the dissociation constant of copper ions, and

indi-rectly to estimate the dissociation constant of zinc

Full-length Ab has a tendency to adhere to the wall

of the quartz cuvette, and this slow process imparts a time dependence to the fluorescence spectra Treatment with ethylenimine [42] increased the stability, but adhe-sion was still detectable Use of the full-length peptide yielded an approximate affinity for copper, Kd  0.5 lm, but with large deviations of the data points from the fitted line (data not shown) The shorter frag-ment Ab(1–28) showed less tendency to adhere, and the sample was stable for more than 3 days The fluor-escence binding studies were therefore performed using the shorter fragment The dissociation constant for copper was determined as Kd¼ 0.36 ± 0.1 lm for this fragment at pH 7.2, assuming a 1 : 1 stoichiometry (Table 1) This Kd is in good agreement with the pre-liminary results obtained using the full-length peptide and the dissociation constant determined earlier [31], but it is significantly higher than that earlier reported for the full-length peptide at physiologic pH [43] There is a fractional fluorescence signal left (approxi-mately 50%) at the end of titration, suggesting that the quenching copper ions are not in direct contact with the fluorescent side chain of tyrosine, but close enough to cause partial quenching There are some systematic deviations of the residuals from the fit, probably arising from induced aggregation Including

a term accounting for the induced aggregation in the binding equation essentially removes all systematic res-iduals (this modified model is described in supplement-ary Doc S1) The dissociation constant is unchanged upon extending the model with this term

Zinc has no fluorescence-quenching abilities, but its binding affinity can be estimated by competitive titra-tions with copper, provided that the binding site is the same Ab(1–28) was incubated with various amounts

of zinc Copper was added to these mixtures, and pro-duced signal quenching similar to what was observed after the addition of copper alone (Fig 5)

Table 1 Dissociation constants of copper and zinc for Ab The dis-sociation constants were calculated from tyrosine fluorescence.

1.2 ± 0.03 b

a Dissociation constant for Ab(1–28) calculated using fluorescence;

10 m M sodium phosphate buffer, 20 C b Dissociation constant for Ab(1–40) calculated using 1H–13C HSQC crosspeak intensity;

10 m M sodium phosphate buffer, 20 C c Dissociation constant for Ab(1–28) calculated using fluorescence; 10 m M Hepes buffer,

20 C.

Addition of EDTA to release the copper from the

peptide by competition recovered most but not all of

the tyrosine fluorescent signal Approximately 30%

was still missing, suggesting that this fraction is

aggre-gated and⁄ or precipitated Assuming that zinc and

copper compete for the same binding site, the dissoci-ation constant for zinc can be calculated using an expression for the bound fraction as a function of the dissociation constants of the two metal ions The dis-sociation constant for zinc was estimated to be 1.08 ± 0.08 lm, indicating only slightly lower affinity than for copper at pH 7.2, close to physiologic condi-tions (Table 1) This dissociation constant for zinc is in excellent agreement with that obtained from NMR data, which was measured on the full-length peptide and is also included in Table 1 Thus, the affinities of zinc for the shorter fragment Ab(1–28) and for Ab(1– 40) are the same

Under mildly acidic conditions (pH 6.5) and in phosphate buffer, the dissociation of copper is slightly increased (Kd¼ 1.16 ± 0.08 lm), whereas the dissoci-ation constant for zinc increases from 1.08 lm to 3.19 ± 0.08 lm (Table 1) The changes in affinity due

to changes in pH close to pH 7 are not large

To investigate the effect of the buffer used in these experiments, the fluorescence measurements were repeated in 10 and 50 mm Hepes buffer at pH 7.2 and with 0, 10 and 20 lm Zn2+ Under these conditions, the dissociation constant for the Ab(1–28)–copper interaction was determined to be 2.5 ± 0.2 lm, and that for the Ab(1–28)–zinc interaction was 6.6 ± 0.1 lm, similar to what was estimated in the experiments using phosphate buffer

CD spectroscopy

To monitor the structural changes of Ab induced by metal ion interactions, CD spectra were recorded for

Ab with increasing amounts of added copper and zinc (supplementary Fig S1) The interaction of Ab with copper reduces the amount of PII helix present in apo-Ab, as demonstrated by examining the difference spectra The N-terminal part of the Ab peptide has a relatively high propensity to adopt a PII helix [8] Thus, a reduction of PII helix secondary structure is consistent with a copper interaction in the N-terminal part of the peptide At low copper concentrations, the structural transition is between two states, and there-fore the CD spectra show an isoelliptic point

Zinc binding gives similar, but not identical, results

to those obtained with copper The zinc interaction mainly reduces the signal intensity but does not clearly reduce the amount of PII helix This may be due to induced aggregation and subsequent precipitation At neutral pH, as in this present study, zinc has a higher propensity to induce aggregation than copper [40] This aggregation can mask the structural change of the peptide

1

0.8

0.6

0.4

30 20

10 0

0.02

0.01

0

-0.01

-0.02

Copper Concentration / [µM]

Fig 5 The tyrosine fluorescence intensity of 10 l M Ab(1–28) in

25 m M phosphate buffer (pH 7.2) at 298 K and 305 nm as a

func-tion of copper concentrafunc-tion The three datasets correspond to 0

(s), 60 (n) and 90 (h) l M zinc acetate added The solid lines are

the fitted curves of the one-to-one binding equation The plateau

values differ between the attenuating intensities This could be due

to different peptide aggregation propensities at different total metal

ionic strengths In the bottom panel, the residuals from the fit are

shown.

Zinc is suggested to have a major effect on aggregation

of Ab [19–21,44], either increasing the aggregation at

high zinc concentrations or reducing the aggregation

at low concentrations [27,28] Here we have studied

the zinc-binding site in soluble monomeric Ab Both

zinc and copper induce specific NMR changes,

affect-ing the same residues in the peptide This is supported

by the fluorescence data, which also show that zinc

and copper compete for the same high-affinity

bind-ing-site (Fig 5) This in agreement with the findings

for the shorter fragments Ab(1–16) and Ab(1–28) [45]

However, both copper and zinc have a putative second

weaker binding site, as shown by NMR This is in

agreement with the finding of two binding sites for

copper in earlier studies [31,33]

The details of the high-affinity binding site for zinc

were studied by NMR 1H–13C HSQC and 1H–15N

HSQC experiments showed a selective zinc-binding site

with His6, His13 and His14 and the N-terminal Asp1 as

ligands (Fig 1) Direct study of the1H–13C HSQC

cross-peaks of the aromatic amino acid side chains shows that

Tyr10 is not directly involved in the binding, but is

located close to the bound metal (Fig 2) The quenching

effect of copper on the tyrosine fluorescence signal

con-firms this view From the present data, a second binding

site for zinc can be proposed, which involves residues

23, 24, 26 and 28 For Cu2+, a similar central region is

involved, manifested by a loss of signal intensity of

resi-due 21 A more detailed study of the second binding site

of copper is not possible, due to the general

paramag-netic line-broadening exhibited by copper

Different ligands for metal coordination by Ab were

suggested in earlier studies, but they were mainly

per-formed on truncated fragments of Ab with or without

acetylated N-terminals All studies, however, showed

the histidines to be necessary ligands [40,46,47] The

truncated fragments show varying binding modes with

respect to the fourth ligand Acetylation of the

N-ter-minus does not inhibit zinc binding to the N-terminal

fragment Ab(1–16), but the fourth ligand is proposed

to be Glu11 in this variant [36] In the same fragment

but without an acetylated N-terminus, Asp1 was

sug-gested to be the fourth ligand [35] Recently Syme

et al published a study on Ab(1–16) and Ab(1–28) in

which they also suggested the fourth ligand to be the

N-terminus, and indeed an N-terminal-blocked variant

of Ab(1–28) showed less effects when zinc was added

[45]

In a recent paper by Hou et al., the interaction of

copper and zinc with full-length Ab(1–40) was studied

using 1H–15N HSQC They suggested that, after

anchoring of the copper ion by the histidine side chains, a less precise binding mode of metal prevails for full-length Ab, compared to the shorter fragments They also observed a reduction on signal intensity that they interpreted as being due either to deprotonation

of amides or line-broadening due to an intermediate chemical exchange rate between the apo-form and holo-form [48] In the present study, we used the full-length Ab(1–40) and combined the use of 1H–13C HSQC and 1H –15N HSQC Under these conditions, zinc binding occurs with His6, His13, His14 and Asp1

as ligands The binding seems to be specific and affects mainly the ligands and the neighboring residues The detected residue-specific signal loss upon metal binding arises from line-broadening due to chemical exchange between conformations in an intermediate rate regime with respect to the NMR timescale (Fig 1)

The N-terminal Asp1 may bind zinc either with the amine group or with the carboxylate groups on the side chain Our data give no direct evidence for which

of these is responsible for binding However, the

1H–13C HSQC findings shows that the Cacrosspeak of Asp1 is more affected by zinc binding (both intensity and chemical shift) than are the Cb crosspeaks (data not shown) The reason could be that Cais closer than

Cb to the binding site, suggesting zinc binding to the amine group of Asp1, in agreement with the recent findings of Mekmouche et al [35] However, our data cannot rule out the possibility of zinc binding to the Asp1 side chain, and this would be in agreement with EPR studies that have reported copper coordination

by three nitrogens and one oxygen, 3N1O, suggesting involvement of the carboxylate oxygen in divalent metal binding [32,47]

The dissociation constant Kdfor copper and Ab has been reported earlier to be approximately 1–5 lm [29,38,48] For zinc, a Kdof 3–300 lm has been repor-ted for full-length Ab(1–40) [38,49,50] Our present results (Table 1) show micromolar dissociation con-stants for both copper and zinc, with a somewhat higher affinity for copper, in good agreement with the earlier reports This holds for experiments in phos-phate as well as in Hepes buffer NMR measurements were similar to fluorescence measurements made under the same conditions Use of the induced chemical shift changes to estimate the dissociation constant yields

Kd 2.6 lm (supplementary Fig S2) This is close to the values obtained with other techniques, and hence provides further evidence for N-terminal involvement

in the metal binding of Ab As previously mentioned, these quantitative results may be somewhat biased, due

to metal–phosphate complex formation and peptide aggregation, which in turn may depend on precise

experimental conditions such as choice of buffer and

temperature Binding of zinc to Ab does not induce

any such well-defined structure of the peptide that can

be determined by NMR methods This is in agreement

with previous reports [45,48] However, the relaxation

data show that the N-terminus becomes more

struc-tured upon zinc binding The results indicate that the

N-terminal region folds around the ion, similar to

ear-lier suggested structures induced by copper [31,33,40]

The CD data show that copper and zinc have only

minor and somewhat different effects on the spectra

The reason may be that copper binding to the

histi-dines induces a charge transfer from the ligand

imidaz-ole to the metal (so called ligand to metal charge

transfer, LMCT), and thus changes the chiral

proper-ties of the peptide The copper-induced changes may

therefore have this origin, and need not necessarily be

due to a change in secondary structure Zinc does not

have this effect on the histidines, and the zinc-induced

changes in CD are very small We conclude that CD

under the present conditions does not provide much

information on the potential metal-induced changes in

the secondary structure of Ab

From the relaxation and diffusion data, we conclude

that no stable, soluble, metal-induced dimers⁄

oligo-mers are present This is in contrast to what has been

reported for Ab(1–16) [26] The full-length peptide

dif-fers from the shorter fragments also in this respect

Our results give rise to a model of the induced

struc-ture of the peptide when bound to zinc (Fig 6) The

binding involves the histidines and the N-terminus

There is a turn at Glu3, bending the N-terminus

towards His6 We propose a second turn at Gly9, to

put His13 and His14 close to the metal ion The model

is similar to the model of Ab(1–28) bound to copper,

proposed by Syme et al [31] We also propose that

zinc has a second, possibly cooperative, binding site

involving the middle segment Asp23, Val24, Asn26

and Lys28 with an induced turn at Gly25

The N-terminal region of free Ab in aqueous

solu-tion has an extended conformasolu-tion rich in PII helix

that is proposed to help to keep the peptide soluble

and protected from amorphous aggregation [8,51–53]

When the N-terminus binds zinc (or copper), it folds

around the metal, forming another relatively

well-defined structure The previous reports on differential

metal-binding effects [20,27] on Ab aggregation at low

and high metal ion concentrations may now be

under-stood in the following terms Metal ions at high

concentrations saturate the binding site(s) of Ab and

lower the electrostatic repulsion between the overall

negatively charged Abs (net charge nominally ) 3 at

pH 7) This effect predominates at high metal ion

con-centrations, and explains the higher aggregation pro-pensity under these conditions The structure induction brought about by the metal-induced fold of the N-ter-minus counteracts aggregation This effect, masked at high metal ion concentrations, should dominate at low ion concentrations, thereby explaining the decreased aggregation propensity under these conditions

Experimental procedures

The peptides, unlabeled Ab(1–40), as well as 15N-labeled and the 13C–15-N-labeled Ab(1–40), were purchased from rPeptide (Athens, GA, USA) and were used without further purification Ab(1–28) was purchased from Neosystems (Strasbourg, France) All peptides were nonmodified in the termini Solvation of the peptide was performed using the protocol suggested by Zagorski et al [54] This protocol prescribes that the peptide is dissolved in a base, here

10 mm NaOH, at high concentration (up to 2 mgÆmL)1), and sonicated in a water⁄ ice bath for 2 min The stock solution was diluted first with water, and then with buffer

to the desired concentration and pH In the present study, the NMR peptide concentration was 50–80 lm and the pH was 7.0–7.3 The peptide was in 10 mm phosphate buffer, and NMR samples contained 10% D2O The peptide and

N

CH2

N H

N

CH2

N H

CH2

N N

H

N-terminus

C-terminus

His6

His13

His14

Fig 6 A schematic representation of the structural model of Ab binding zinc or copper The structure was constructed using a com-bination of signal intensity changes, relaxation data and induced amide proton stability.