Amino acid position-specific contributions to amyloid beta-protein oligomerization.

Understanding the structural and assembly dynamics of the amyloid (cid:1) -protein (A (cid:1) ) has direct relevance to the development of therapeutic agents for Alzheimer disease. To elucidate these dynamics,wecombinedscanningaminoacidsubstitutionwitha method for quantitative determination of the A (cid:1) oligomer frequency distribution, photo-induced cross-linking of unmodified proteins (PICUP), to perform “scanning PICUP.” Tyr, a reactive group in PICUP, was substituted at position 1, 10, 20, 30, or 40 (for A (cid:1) 40) or 42 (for A (cid:1) 42). The effects of these substitutions were probed using circular dichroism spectroscopy, thioflavin T binding, electron microscopy, PICUP, and mass spectrometry. All peptides displayed a random coil 3 (cid:2) / (cid:1) 3 (cid:1) transition, but substitution-dependent alterations in assembly kinetics and conformer complexity were observed. Tyr 1 -substi-tuted homologues of A (cid:1) 40 and A (cid:1) 42 assembled the slowest and yielded unusual patterns of oligomer bands in gel electrophoresis

Recent studies of A␤ oligomers have sought to correlate oligomer size and biological activity. Oligomers in the supernatants of fibril preparations centrifuged at 100,000 ϫ g caused sustained calcium influx in rat hippocampal neurons, leading to calpain activation and dynamin 1 degradation (27). A␤-derived diffusible ligand-like A␤42 oligomers induced inflammatory responses in cultured rat astrocytes (28). A 90-kDa A␤42 oligomer (29) has been shown to activate ERK1/2 in rat hippocampal slices (30) and bind avidly to human cortical neurons (31), in both cases causing apoptotic cell death. A comparison of the time dependence of the toxic effects of the 90-kDa assembly with that of A␤-derived diffusible ligands revealed a 5-fold difference, A␤-derived diffusible ligands requiring more time for equivalent effects (31). A 56-kDa oligomer, "A␤*56," was reported to cause memory impairment in middle-aged transgenic mice expressing human amyloid precursor protein (32). A nonamer also had adverse effects. Impaired long term potentiation in rat brain slices has been attributed to A␤ trimers identified in media from cultured cells expressing human amyloid precursor protein (33). Dimers and trimers from this medium also have been found to cause progressive loss of synapses in organotypic rat hippocampal slices (10). In mice deficient in neprilysin, an enzyme that has been shown to degrade A␤ in vivo (34), impairment in neuronal plasticity and cognitive function correlated with significant increases in A␤ dimer levels and synapse-associated A␤ oligomers (35).
The potent pathologic effects of A␤ oligomers provide a compelling reason for elucidating the mechanism(s) of their formation. This has been a difficult task because of the metastability and polydispersity of A␤ assemblies (36). To obviate these problems, we introduced the use of the method of photoinduced cross-linking of unmodified proteins (PICUP) to rapidly (Ͻ1 s) and covalently stabilize oligomer mixtures (for reviews see Refs. 37,38). Oligomers thus stabilized no longer exist in equilibrium with monomers or each other, allowing determination of oligomer frequency distributions by simple techniques such as SDS-PAGE (37). Recently, to obtain population-average information on contributions to fibril formation of amino acid residues at specific sites in A␤, we employed a scanning intrinsic fluorescence approach (39). Tyr was used because it is a relatively small fluorophore, exists natively in A␤, and possesses the side chain most reactive in the PICUP chemistry (40). Using this approach, we found that the central hydrophobic cluster region (Leu 17 -Ala 21 ) was particularly important in controlling fibril formation of A␤40, whereas the C terminus was the predominant structural element controlling A␤42 assembly (39). Here we present results of studies in which key strategic features of the two methods have been combined to enable execution of "scanning PICUP" and the consequent revelation of site-specific effects on A␤ oligomerization.

EXPERIMENTAL PROCEDURES
Chemicals and Reagents-Chemicals were obtained from Sigma and were of the highest purity available. Water was double-distilled and deionized using a Milli-Q system (Millipore Corp., Bedford, MA).
Peptide Design and Synthesis-In addition to studying native A␤40 and A␤42, each of which contains Tyr 10 , single Tyr substitutions were made in each A␤ alloform at Asp 1 , Phe 20 , Ala 30 , and the C terminus (Val 40 or Ala 42 ) (Fig. 1). In each non-native peptide, Tyr 10 was replaced by Phe, which is largely unreactive in PICUP, so that data interpretation would not be complicated by multiple potential cross-linking sites. A similar substitution strategy proved effective in Tyr intrinsic fluorescence studies (39) in which Phe was fluorometrically, as opposed to chemically, insignificant. A␤ synthesis, purification, and characterization were done as described (41). Briefly, A␤40, A␤42, and their Tyr-substituted peptides (Fig. 1) were made on an automated peptide synthesizer (model 433A, Applied Biosystems, Foster City, CA) using Fmoc (N-(9fluorenyl)methoxycarbonyl)-based methods. Peptides were purified using reversed phase high performance liquid chromatography (RP-HPLC). Quantitative amino acid analysis and mass spectrometry yielded the expected compositions and molecular weights, respectively, for each peptide. Purified peptides were stored as lyophilizates at Ϫ20°C. When possible, to maximize chemical homogeneity among related peptides, multiple peptides were synthesized from the same starting resin by resin splitting at sites of sequence variation. A␤40 and its Tyr-substituted analogues were synthesized using preloaded [ Sample Preparation-All peptides were pretreated with dilute NaOH to increase their solubility and decrease de novo peptide aggregation (42). Briefly, peptides were dissolved initially in 2 mM NaOH (1 mg/ml), sonicated for 3 min in an ultrasonic water bath (model B1200-R, Branson Ultrasonics Corp., Danbury, CT), and then lyophilized. This treatment and other treatments designed to produce unaggregated "starting" peptide preparations have not been found to affect the primary structure of the peptide or its subsequent folding and self-assembly (21,(42)(43)(44)(45)(46).
For CD studies, lyophilizates of pretreated peptides were dissolved in 1 volume of water, after which an equal volume of 20 mM phosphate buffer, pH 7.4, containing 0.02% (w/v) sodium azide was added. Samples were sonicated for 1 min at 22°C, transferred into centrifugal filters (10,000 molecular weight cutoff, Centricon YM-10, Millipore Corp.), and centrifuged at 16,000 ϫ g using a bench top microcentrifuge (Eppendorf model 5415C, Brinkmann Instruments) for 30 min. The filtrate, containing low molecular weight (LMW) A␤, was incubated at 22°C without agitation to allow peptide assembly. By definition (47), LMW A␤ contains monomeric A␤ in equilibrium with low order, unstructured oligomers (47)(48)(49). The concentration of A␤ in the filtrates was determined by quantitative amino acid analysis, as described (50).
For cross-linking experiments of A␤40, LMW A␤ was isolated using size exclusion chromatography (SEC), as described (41). Briefly, A␤ was dissolved at a concentration of 2 mg/ml in dimethyl sulfoxide and sonicated for 1 min, after which 170 l of this solution was injected onto the SEC column. The column was eluted with 10 mM sodium phosphate buffer, pH 7.4, at a flow rate of 0.5 ml/min. Peptides were detected by UV absorbance at 254 nm, and fractions of 350-l volume were collected during elution of the LMW A␤ peak. The concentrations of Tyr-substituted A␤42 peptides isolated by SEC were lower than those obtained using the centrifugation method (51); therefore, the latter method was used to prepare A␤42 peptides for study. Both peptide preparation methods yielded A␤ solutions that produced CD spectra indicative of predominantly disordered secondary structure. Previous studies have shown that LMW A␤42 prepared using filtration or SEC produces similar oligomer distributions in the monomer to octamer region (48). An advantage of the filtration method for preparing A␤42 is that larger oligomers (e.g. dodecamers and octadecamers) are not present (48).
Cross-linking and SDS-PAGE Analysis-Peptides were covalently cross-linked using PICUP immediately after preparation (for a review see Ref. 38). Briefly, 1 l of 1 mM Tris(2,2Јbipyridyl)dichlororuthenium(II) and 1 l of 20 mM ammonium persulfate in 10 mM sodium phosphate, pH 7.4, were added to 18 l of a 20-30 M solution of A␤ or its analogues immediately after preparation. The mixture was irradiated for 1 s with visible light, and the reaction was quenched immediately with 10 l of Tricine sample buffer (Invitrogen) containing 5% (v/v) ␤-mercaptoethanol. The cross-linked oligomer mixtures were fractionated by SDS-PAGE using 10 -20% Tricine gels (1.0 mm ϫ 10 well) (Invitrogen) and silver-stained using a Silver-Xpress silver staining kit (Invitrogen), and then the band intensities were quantified by densitometry, as described (49). The amounts taken for SDS-PAGE analyses were adjusted according to the peptide concentration, determined by amino acid analysis, so that equal amounts of protein were loaded in each lane. Gels were dried and scanned, and the intensities and gel mobilities (R f ) of the resulting monomer and oligomers bands were quantified by densitometry using the program One-Dscan (Scanalytics, Fairfax, VA). The relative amount of each band in a lane as a percentage of all bands in the same lane (I r i ) was determined according to Equation 1, where I i is the intensity of the band i. Circular Dichroism Spectroscopy-A␤, at a concentration of 30 -35 M in 10 mM phosphate buffer, pH 7.4, was prepared by filtration, and the spectra were acquired daily during incubation of peptides at 22°C without agitation. Samples were prepared for analysis by gently drawing up and then expelling the peptide solution in a 200-l pipette tip. After three such cycles, the peptide solution was placed into a 0.1-cm path length quartz cell (Hellma, Forest Hills, NY). Spectra were acquired using an Aviv model 62A DS spectropolarimeter (Aviv Associates, Lakewood, NJ). Following measurements, samples were returned to the original sample tubes. All measurements were done at 22°C. Spectra were generally recorded over the wavelength range of 198 -260 nm. Three independent experiments were performed with each peptide. Raw data were manipulated by smoothing and subtraction of buffer spectra, according to the manufacturer's instructions.
Spectral deconvolution was performed using the CDPro software package (52), which contains the deconvolution programs SELCON3, CDSSTR, and CONTIN. In this package, reference sets of proteins from different sources are combined to create a large reference set of CD spectra. Depending on the spectrum wavelength range, the number of proteins in the reference set (IBasis) can be as large as 48 (IBasis 7). Deconvolutions were done with each of the three programs. If the data thus obtained were similar, they were averaged to obtain the percentage of each secondary structure element. In some cases, the results obtained from one program were highly divergent from those of the other two. In this case, averaging was done only with data from the two consistent programs. Deconvolutions were performed on data acquired at the initiation of assembly, when the presence of significant ␣-helix was observed (by visual inspection of the spectra), and when assembly was complete (spectra remained identical during repeated monitoring). We note that the kinetics of assembly in these studies differ quantitatively, but not qualitatively, from those acquired in previous work (53). Here, the assembly reactions were performed using 20 mM phosphate buffer, pH 7.4. The earlier studies were done in 10 mM Gly-NaOH, pH 7.5. The latter buffer slows the kinetics and facilitates occupation of regions of conformational space containing ␣-helix.
Thioflavin T (ThT) Binding-A 100-l aliquot of each A␤ sample in 10 mM phosphate buffer, pH 7.4, containing 0.01% (w/v) sodium azide, was mixed with 5 l of 100 M ThT prepared in the same buffer. Immediately after addition of ThT, fluorescence was measured. The measurements were made using a Hitachi F4500 spectrofluorometer (Hitachi Instruments Inc., Rye, NH) with excitation at 450 nm and emission at 480 nm. A rectangular 10-mm quartz microcuvette was used. All fluorescence measurements were carried out at 22°C with a scan rate of 240 nm/min. Slit widths used for excitation and emission were 5 and 10 nm, respectively. Three independent experiments were performed for each peptide.
Electron Microscopy-For studies of fibrillar A␤, 5 l of each sample were spotted on a glow-discharged, carbon-coated Formvar grid (Electron Microscopy Sciences, Fort Washington, PA), incubated for 5 min, washed with distilled water, and then stained with 1% (w/v) aqueous uranyl formate. Uranyl formate (Pfaltz & Bauer, Waterbury, CT) solutions were filtered through 0.2-m sterile syringe filters (Corning Glass) before use. EM analysis was performed using a JEOL 1200 transmission electron microscope. Four independent experiments were carried out for each peptide.
For studies of cross-linked and noncross-linked LMW A␤ peptides, cross-linking reactions were quenched with 1 M dithiothreitol (FisherBiotech, Fair Lawn, NJ) in water instead of 5% (v/v) ␤-mercaptoethanol in Tricine sample buffer (Invitrogen). The same LMW preparation also was cross-linked and quenched immediately with 10 l of 5% (v/v) ␤-mercaptoethanol in Tricine sample buffer and then used for SDS-PAGE to verify that the expected oligomer distribution was obtained. Ten l of each sample were incubated for ϳ20 min on the grid. The solution was gently removed using Whatman grade 2 qualitative filter paper, and then the grid was incubated with 5 l of 2.5% (v/v) glutaraldehyde for 4 min, after which fluid again was removed using filter paper. The peptide then was stained with 5 l of 1% (w/v) uranyl acetate (Pfaltz & Bauer) for 3 min. This solution was wicked off, and the grid was air-dried. Samples were examined using a JEOL CX100 electron microscope.
Quantitative analysis of oligomer geometry in cross-linked and noncross-linked LMW A␤ samples was performed by manual determination of oligomer dimensions by inspection of EM images. A representative sample was obtained with particle number n ϭ 40. Particle diameter and length statistics were calculated using Mathematica 6.0 (Wolfram Research, Inc., Champaign, IL).
Reverse Staining and Isolation of Individual Oligomers-Cross-linked A␤ oligomers separated by SDS-PAGE were detected by imidazole-zinc staining, essentially as described (54). Briefly, after SDS-PAGE, the gel was rinsed in distilled water for 30 s and then incubated in 0.2 M imidazole (Sigma) solution containing 0.1% (w/v) electrophoresis grade SDS (Fisher) for 15 min. The solution then was discarded, and the gel was incubated in 0.2 M zinc sulfate in distilled water for ϳ0.5-1 min, until the gel background became white and the A␤ oligomers bands were transparent and colorless. Further staining was prevented by rinsing the gel in distilled water. The staining process was monitored by placing the gel in a transparent tray over a piece of black paper. Immediately after staining, the oligomer bands were excised using a scalpel blade (Fisher) and placed into 1.5-ml conical microcentrifuge tubes. The gel slices were incubated (two times for 5 min each) in 1 ml of 25 mM ammonium acetate buffer, pH 7.4, containing 100 mM EDTA, during which time the gels become completely colorless. The gel slices then were washed twice (two times for 5 min each) using 25 mM ammonium acetate buffer, pH 7.4, and frozen quickly on dry ice (to make the gels pieces brittle and fragile). The pieces were crushed using a 1.5-ml microcentrifuge tube as a mortar and a geometrically matched pestle (Fisher). After crushing the gel pieces, the pestle was held over the tube and washed with 25 mM ammonium acetate buffer, pH 7.4, to ensure that all the gel pieces were collected. Additional buffer was added to the tube to make the volume of the resulting suspension twice that of the original gel pieces. The microcentrifuge tube containing the gel suspension then was agitated by moderate vortexing for 10 min using a Multi-Tube Vortexer (VWR International, Bristol, CT). The gel mixture was centrifuged for ϳ1 min at 14,000 ϫ g, and the supernatant was then collected and placed in a glass tube. A volume of buffer twice that of the crushed gel was added to the pellet, which then was vortexed for 10 min. After another centrifugation, done as above, the supernatant was collected and combined with the first supernatant. A volume of 25 mM ammonium acetate, 50% (v/v) acetonitrile, 0.1% (v/v) trifluoroacetic acid, equal to that of the pellet, was then added to the pellet, and the tube was vortexed for 10 min. A third supernatant then was obtained by centrifugation and was combined with the first two supernatants. Because A␤ oligomers are hydrophobic, a fourth extraction was done for 10 min using 25 mM ammonium acetate buffer, 25% (v/v) acetonitrile, 25% (v/v) isopropyl alcohol, and 0.1% (v/v) trifluoroacetic acid. A final extraction was done for 2 min using 80% (v/v) acetonitrile. The pooled supernatants were concentrated by centrifugal evaporation (Savant SpeedVac Concentrator, Thermo Scientific, Waltham, MA).

RESULTS
Secondary Structure Dynamics-To probe amino acid sitespecific contributions to peptide conformation and assembly, we used CD spectroscopy and ThT binding to monitor temporal changes in secondary structure. In the CD analyses, A␤40 and its substituted alloforms were predominantly unstructured immediately after preparation, as indicated by prominent negative molar ellipticities at ϳ198 nm ( Fig. 2A). Conformational changes in A␤40 were noticeable after ϳ6 days. At day 7, mixed ␣/␤ character was indicated qualitatively by double inflections in the 205-225 nm region and quantitatively by spectral deconvolution. Visual inspection suggested that ␣-helix content was maximal at ϳ8 days. Quantitative analysis of the spectra was done following deconvolution (see "Experimental Procedures"). At 0 day, ␣-helix, ␤-strand, ␤-turn, and random coil (RC) 5 secondary structure elements were present at levels of ϳ8, ϳ9, ϳ6, and 77%, respectively. In contrast, at 8 days, ␣-helix, ␤-strand, ␤-turn, and RC levels were ϳ26, ϳ24, ϳ19, and 33%, respectively. The mixed ␣/␤ conformer population at 8 days includes in part a previously described ␣-helix-rich intermediate (53). Between days 7 and 14, an ␣/␤ 3 ␤ transition was observed that produced a classical ␤-sheet-type spectrum with a negative ellipticity of significant magnitude centered at ϳ215-218 nm. The spectral deconvolution at day 14 showed that the ␣-helix, ␤-strand, ␤-turn, and RC levels were ϳ10, ϳ66, ϳ5, and 20%, respectively.
[Phe 10 ]A␤40, the "scaffold" upon which the Tyr-substituted peptides was built, displayed time-dependent conformational changes qualitatively similar to that of A␤40, as did the four other Tyr-substituted alloforms. As reported previously in studies of A␤40 assembly (53), absolute convergence of the spectra onto an isodichroic point was not seen, suggesting a multistate transition process. We do note, however, that some convergence of the spectra to a single point occurred in the [Tyr 40 ]A␤40 experiment.
To obtain quantitative insight into the kinetics of the RC 3 ␣/␤ 3 ␤ conformational conversions, we studied the time dependence of 222 (supplemental Fig. 1A). The half-time for the development of ␣-helix structure and the time at which maximal ␣-helix structure is observed are later for [Tyr 1 ]A␤40 (8.5 and 11 days, respectively) than they are for A␤40 (7 and 8 days, respectively). These two times also are longer than those for [ Consistent with the CD studies, no ThT binding was observed by any peptides immediately after their preparation (Table 1). However, substantial binding was detected when characteristic ␤-sheet spectra were seen by CD. Binding levels among the different peptides were equivalent, within experimental error (FU avg ϭ 9475 Ϯ 598), where FU is fluorescence units.
CD analysis revealed that A␤42, [Phe 10 ]A␤42, and the Tyrsubstituted analogues displayed RC 3 ␣/␤ 3 ␤ transitions (Fig.  2B) qualitatively similar to those observed with A␤40, but with accelerated kinetics. Consistent with this acceleration, the observed lifetimes of the ␣-helix-containing conformers were relatively short (1 day instead of 3-5 days in A␤40). The ␣-rich conformer appeared for all peptides at day 3, except for [Tyr 1 ]A␤42, in which the ␣-helix-rich conformer appeared at day 4. The rapidity with which the RC 3 ␣/␤ 3 ␤ transition occurs in the A␤42 peptide family is indicated by a monotonic decrease in 222 that begins earlier and does not reach the magnitude of those seen in the A␤40 samples (supplemental Fig.  1B). This increased A␤42 assembly rate is consistent with results of earlier comparative studies of A␤40 and A␤42 peptides linked to familial forms of AD and cerebral amyloid angiopathy (53).
The spectral deconvolution at day 8, where all A␤42 spectra were mostly ␤-sheet (by visual inspection), confirmed that all  It should be noted that the reproducibility of studies of A␤ assembly kinetics depends on careful peptide preparation and manipulation (for a recent review see Ref. 36). The experiments presented here all were performed using the same peptide lot that was prepared and incubated in precisely the same manner for all samples. Solutions were not agitated and any study of the reactions was done with minimal perturbation of the tubes. Under these conditions, the kinetics was reproducible within experiments. Absolute changes in kinetics can be observed between experiments, but the rank order of rates of conformational change remains constant among experiments. Thus, peptides that form the ␤-sheet structure the fastest in any one experiment always form the ␤-sheet structure the fastest. Similarly, the "slowest" peptides always are the slowest. The system variability thus is so small that the trends in rank order of ␤-sheet formation are always the same.
Morphologic Analysis of Assemblies-To determine the morphologies of the assemblies present at the completion of the CD and ThT studies, EM was done. All 12 peptides formed long, unbranched fibrils with smooth margins (Fig. 3, A and B). A␤40   30 ]A␤40 produced a more structurally diverse population of fibrils that displayed diameters ranging from 6 to 12 nm and were composed of 2-5 filaments. Some fibrils had no observable twist, whereas others displayed a helical pitch of ϳ100 -150 nm.
In contrast to the variation in fibril morphologies observed in the A␤40 samples (Fig. 3A), A␤42 peptides formed fibrils that were morphologically similar (Fig. 3B). Most fibrils were ϳ4 -6 nm in diameter and were composed of two filaments. Little discernible substructure was apparent in many fibrils, whereas others appeared with irregular twists or helical twists with pitches of ϳ40 -80 nm. The only departure from these shared morphologic features was observed with [Phe 10 ]A␤42, which produced fibrils with a diameter range, 5-7.5 nm, that overlapped with but was slightly larger than that of the other peptides. The largest fibril included three filaments. The approximate 2-fold difference in diameter between fibrils formed by A␤40 and A␤42 is consistent with and provides a plausible explanation for the magnitude of the difference in ThT binding observed between the two peptide families (see above). However, this explanation assumes a linear relationship between ThT binding and fluorescence intensity. It also is possible that variations in average order or self-quenching could account for the observed differences.
Determination of Oligomer Size Distributions-To probe A␤ oligomerization, LMW fractions of A␤ and its Tyr-substituted alloforms were isolated and immediately analyzed by PICUP and SDS-PAGE ( Fig. 4 and supplemental Table 1). Distinct oligomer size distributions were observed. A␤40 ([Tyr 10 ]A␤40) produced a mixture consisting of predominantly monomer (ϳ20%), dimer (ϳ25%), trimer (ϳ25%), and tetramer (ϳ17%), along with small amounts of pentamer (ϳ6%) and hexamer (ϳ3%). The oligomer distribution of [Tyr 1 ]A␤40 differed from that of A␤40. Presumptive [Tyr 1 ]A␤40 monomers through trimers electrophoresed slightly faster than wild type monomer, dimers, and trimers. In the gel region corresponding to the native trimer, two bands existed (Fig. 4, closed arrowhead). Above this region, three additional prominent bands were seen.  30 ]A␤40 resembled that of [Tyr 1 ]A␤40 in that bands 3 and 4 migrated close to each other, like a doublet. In addition, the M r values for the higher order oligomers were higher than those of the corresponding A␤40 oligomers, and at least one or two higher order oligomer bands were observed (band 7 and above). Cross-linking of [Tyr 40 ]A␤40 produced the greatest number of bands (10 bands). The [Tyr 40 ]A␤40 distribution was similar to that of A␤40 in the monomer-trimer region, but a doublet was formed by bands 4 and 5 (see Fig. 4, arrowhead).
All the substitutions caused increases in the number of oligomer bands and the M r of the highest order band. These effects were most apparent in the peptides in which substitutions were made at the C terminus (positions 30 and 40/42). We note also that the migration differences of dimers and trimers in the substituted peptides relative to A␤40 were greatest in [Tyr 20 ]A␤40 and progressively smaller in [Tyr 30 ]A␤40 and [Tyr 40 ]A␤40.
Cross-linking of A␤42 produced a characteristic (48) distribution with nodes at monomer and pentamer. Heptamers were visible clearly. In the [Tyr 1 ]A␤42 distribution, the predominant oligomer was the tetramer (Fig. 4, closed arrowhead). Interest- ingly, the oligomer distribution in the [Tyr 1 ]A␤42 sample was similar to that of the [Tyr 1 ]A␤40 sample. The M r values of the bands were lower than those of wild type A␤42 and bands 3 and 4 migrated as a doublet. In addition, a larger number of bands were seen (9 versus 7), and the largest M r exceeded heptamer. The oligomer distributions of the Tyr 20 , Tyr 30 , and Tyr 42 alloforms showed M r values of tetramers and higher order oligomers higher than their A␤42 homologue. In addition, as with A␤40, all the substitutions caused increases in the number of oligomer bands and the M r of the highest order band, and these effects were most apparent in the peptides in which substitutions were made at the C terminus (positions 30 and 42).
We note that, at the peptide concentrations used in these experiment, small variations in experimental conditions (peptide and reactant concentrations, temperature, irradiation time, etc.) do not alter significantly the oligomer frequency distributions. These distributions were reproducible among experiments. Importantly, based on careful study of the crosslinking system itself (49) and on results of extensive structureactivity studies (38,48,49), the differences observed among samples in the oligomer frequency distribution and M r values of individual bands are meaningful. The data are not due to random collisional events "captured" by cross-linking.
Determination of Oligomer Morphology-To determine whether increased oligomerization propensity, as indicated by expansion of the oligomer distribution range to higher molecular weight, was reflected in increased oligomer size, EM studies were performed. Among all the peptides studied, A␤40 and [Tyr 40 ]A␤40 displayed the largest difference in oligomerization propensity, and thus we determined their morphologies immediately following isolation by SEC and cross-linking by PICUP (Fig. 5).
Both noncross-linked and cross-linked A␤40 displayed clusters of relatively amorphous structures with low aspect ratios. The average diameter of the globular structures seen in noncross-linked A␤40 was 16.0 Ϯ 0.54 nm (Fig. 5a, inset, white  arrow), and 93% of these globules had diameters between 12 and 20 nm (supplemental Fig. 2A). Aggregates often were observed. Cross-linked A␤40 produced aggregates that were more thread-like in appearance (Fig. 5b, black arrow) but had similar average diameters (15.8 Ϯ 0.61 nm; Fig. 5b, inset, white arrow, and supplemental Fig. 2B). Short protofibril-like structures were also observed, and these were more frequent in the cross-linked sample. The lengths of these structures varied significantly (ϳ11-111 nm; supplemental Fig. 2C), but the distribution of average diameter (d ) was narrow (d ϭ 6.2 Ϯ 0.2 nm; supplemental Fig. 2D).
Non cross-linked [Tyr 40 ]A␤40 also produced relatively amorphous globules and thread-like structures (Fig. 5c, black  arrows). The average diameter of these particles was 9.7 Ϯ 1 nm (supplemental Fig. 2E), smaller than that of A␤40. However, the range of lengths of protofibril-like assemblies produced by [Tyr 40 ]A␤40 was similar to that of A␤40 (15-100 nm; supplemental Fig. 2F). The average diameter of the amorphous structures formed in the cross-linked [Tyr 40 ]A␤40 sample (Fig. 5d, white arrows), 19.6 Ϯ 0.9, was approximately twice that of the noncross-linked peptide and significantly (p Ͻ 0.0001) larger than that of cross-linked A␤40 (supplemental Fig. 2G). This increase is consistent with the increased order observed in the SDS gels of the cross-linked peptides (Fig. 4).
We note, in principle, that a direct correspondence between PICUP and EM may not be observed because the former method uses the denaturing and dissociative characteristics of SDS-PAGE to reveal covalent association among monomers, whereas the latter method requires only adherence of aggregates to a solid support and not their pre facto covalent association. Here, however, the data produced by each method are consistent.
Determination of Oligomer Order-Our analyses of the oligomer size distributions of the Tyr-substituted peptides (Fig. 4) revealed that certain substitutions, e.g. [Tyr 1 ]A␤40, produced oligomer distributions distinct from those of wild type A␤. To determine the oligomer order within the various gel bands, their component peptides were isolated and subjected to MALDI-TOF mass spectrometry. We first determined the mass of all oligomer bands from wild type A␤40 (band numbers are shown in Fig. 4 with white letters). Band 1 from A␤40 produced a major peak of 4332 atomic mass units with an additional small broad peak (Ͻ5%) of 8574 atomic mass units. We assigned the 4332 atomic mass unit peak to the singly protonated A␤40 monomer (theoretical m/z ϭ 4331; Fig. 6A). Bands 2, 3, and 4 produced major peaks of 8656, 13,010, and 17,383 atomic mass units, corresponding to dimer, trimer, and tetramer, respectively (Fig. 6A). Masses from the higher order bands 5 and 6 could not be obtained, likely because the high molecular weight oligomers were present in small quantities, or they may not have been desorbed from the MALDI matrices as readily as the smaller oligomers. Mass spectrometric analysis thus confirmed that A␤40 produced a simple oligomer "ladder." We next analyzed [Tyr 1 ]A␤40 (Fig. 6B). Band 1 displayed a mass of 4363 atomic mass units, which is the average mass of a singly protonated monomer. In linear mode, band 2 produced two almost equally intense peaks, one at 8741 atomic mass units and one at 4380 atomic mass units. The ion of 4380 atomic mass units is consistent with oxidized [Tyr 1 ]A␤40 (expected mass of 4379 atomic mass units), whereas 8741 atomic mass units is consistent with a dimer in which one monomer is oxidized (expected mass of 8739 atomic mass units). A spectral component of mass 4363 atomic mass units also was observed, consistent with the presence of un-oxidized monomer. Measurements done in reflector mode revealed 1 atomic mass unit isotope spacings in the 4400 atomic mass units region, establishing that doubly charged ϳ8800-Da species were not solely responsible for the 4400 atomic mass units ions. Band 3 produced major ions of 4388 and 13,188 atomic mass units, consistent with monomer mass and trimer, respectively. Although the 13,188 atomic mass unit ion is ϳ100 Da higher in mass than expected for the trimer (13,090), it is reasonable to assign it to the trimer as other oligomers clearly could not produce such a mass. Band 4 yielded four significant peaks, the masses of which were 4384, 8767, 13,109, and 17,475 atomic mass units. We assign these to peptide monomer, dimer, trimer, and tetramer, respectively. Band 5 displayed five peaks that were consistent in M r with oxidized monomer (4383), oxidized dimer (8757), oxidized trimer (13,196), oxidized tetramer (17,617), and oxidized pentamer (21,943). The data show, as with A␤40, that an oligomer ladder was produced in the SDS-PAGE experiment that comprised neighboring oligomers differing in order by one. Importantly, the combined electrophoretic/spectroscopic analyses reveal that the Tyr 1 substitution significantly alters the M r of the peptide tetramer and pentamer, but it has much less effect on the monomer, dimer, and trimer. As with A␤40 itself, sufficient signal could not be obtained to analyze the presumptive hexamer and heptamer bands of this peptide (Fig. 4, bands 6  and 7).
Despite extensive efforts using a broad range of matrices and solvents, spectra of [Tyr 1 ]A␤42 oligomers were not obtained. Insolubility did not appear to explain the phenomenon, as the dried analytes dissolved completely in the solvents employed. Rather, the result may be due to the follow- ing: 1) the inability of these oligomers to be incorporated within crystals of the MALDI matrices, perhaps because of their exceptional hydrophobicity; or 2) the disruption of labile covalent bonds or weak noncovalent interactions by the desorption/ionization process.

DISCUSSION
Evidence from in vitro and in vivo studies suggests that A␤ oligomers are potent neurotoxins and may be the proximal effectors of the neuronal dysfunction and death occurring in AD (4,(55)(56)(57). This strong association of A␤ oligomers with AD pathogenesis provides a rationale for the performance of studies to elucidate the structural dynamics of oligomerization, in particular how specific residues within A␤ control the process. To achieve this goal, here we have executed a strategy we term scanning PICUP that employs Tyr in a classical scanning amino acid substitution paradigm.
The use of Tyr, which is highly reactive in the photochemical cross-linking reaction on which PICUP is based, allowed us to simultaneously probe the effects of alteration of specific amino acid side chains on A␤ conformational dynamics and to determine how side-chain modifications affected peptide oligomerization. We note that the results thus obtained depend on the local environment of the Tyr residue within the A␤ monomer (e.g. its solvent accessibility) and on the oligomerization state of this monomer. The data thus reflect the sum of these phenomena. His and Met also may function in the PICUP chemistry, but their reactivity is substantially lower than that of Tyr (58), and thus the data discussed here reflect the activity of the substituted Tyr.
Qualitative analysis of the results of the CD studies of the A␤40 and A␤42 families of peptides comprising wild type and Tyr-substituted homologues showed that each family underwent an RC 3 ␣/␤ 3 ␤ transition, as has been observed in prior studies (39,53). However, detailed analysis of the results revealed substitution-dependent alterations in folding kinetics and conformer complexity. For each peptide family, the Tyr 1substituted homologue folded slowest, as assessed by determination of the midpoint of the secondary structure transition from RC 3 ␤-sheet. This type of divergence from the dynamics of the wild type peptide also was observed in PICUP studies of oligomerization and EM studies of fibril formation. The oligomer frequency distribution of the substituted peptide was distinct, with an unusual doublet band apparent in SDS gels of the cross-linked population. It is significant that the Tyr 1 substitution in A␤42 also produced an unusual oligomer distribution, one qualitatively similar to that seen in A␤40, because this datum suggests that the effects of this substitution on the conformational dynamics of the peptide dominate those linked to the presence of Ile 41 and Ala 42 .
In analyzing SDS-PAGE data such as those discussed above, one generally infers from the M r of a particular band that it includes n-order oligomers, where n ϭ M r /MW monomer . However, this inference is valid only if the electrophoretic behavior of the analyte is ideal. We did not assume ideal behavior and therefore sought to formally determine the M r of a number of the bands. In doing so, we found that putative [Tyr 1 ]A␤40 monomers, dimers, and trimers yielded mass spectra consistent with our order calculation. However, the putative tetramers and pentamers did not. Mass spectrometrically determined molecular weights for bands 4 and 5 were consistent with the masses of tetramers and pentamers, respectively, yet calculation of their M r values determined electrophoretically suggested they contained trimers and tetramers, respectively. These oligomers thus migrated anomalously. First principles suggest that the anomalously low M r was due to oligomer compaction. Direct experimental evidence for A␤ oligomer compaction has been obtained previously in studies of A␤40, A␤42, and homologues thereof (59,60).
A␤ assembly involves intramolecular (within monomer) and intermolecular (within oligomers) interactions. Prior experimental and computational studies have revealed that important interaction sites controlling A␤ fibril formation exist within and adjacent to the central hydrophobic cluster (CHC; Leu 17 -Ala 21 ) and the C terminus (48,(61)(62)(63)(64). This knowledge has led naturally to the design of potential therapeutic agents targeting these sites, e.g. KLVFF-like peptide inhibitors (for a review, see Ref. 65) and inhibitors targeting the A␤ C terminus (66,67). The ability of a single amino acid substitution at the N terminus (Tyr 1 ) to alter the oligomer frequency distribution and the assembly kinetics suggests that the N terminus is not a benign peptide sub-region uninvolved in peptide assembly, as might be inferred from studies of fibril structure (68). It may be argued instead that the contribution of this region in the wild type peptide to the overall assembly energetics is significant but differs in magnitude from the contributions of the CHC and C terminus. Does this mean that therapeutic attention to this region is unwarranted? We would argue the contrary, namely that if the inhibitory effects of the Tyr 1 substitution can be amplified through selection of an appropriate small molecule inhibitor, then new and potentially efficacious assembly inhibitors may be discovered.
In addition to the CD experiments revealing effects of substitutions on folding kinetics, insights into the complexity of the conformational space of A␤ were obtained. The lack of a precise isodichroic point in CD spectra from the A␤40 family shows that peptide assembly is not a simple two-state process (69,70). However, we note that the spectra produced by the Tyr 40 [A␤40] peptide did converge, albeit imprecisely, near a specific point. This convergence does suggest a decrease in the conformational complexity of the system, as is clearly observed in the spectra of A␤42, which display an isodichroic point. The Tyr 40 substitution in A␤40 thus produces a secondary structure dynamics more akin to that seen in A␤42. Consistent with this conclusion is the fact that the Tyr 40 substitution in A␤40 substantially extended the highest order of oligomers observed in the PICUP experiments. For A␤42, its conformational dynamics certainly must correlate with the increased hydrophobic surface of the peptide C terminus, which in turn may affect the stability of a C-terminal hinge (71) or turn (72) and the interactions of the C terminus with other regions of the peptide monomer, including the CHC. Because A␤40 lacks two C-terminal residues and has not been found to form a turn in this region (72), our data suggest that the Tyr substitution in A␤40 facilitates structural organization of the peptide monomer through interactions of the C terminus with the CHC. These interac-tions may stabilize the monomer, restricting its exploration of conformational space and accounting for the quasi-isosbestic point in the CD spectra, an observation suggestive of the absence of intermediates in the conformational conversion process. We note, however, that the structural stabilization imparted on the A␤40 peptide by the substitution of Tyr 40 does not result in an A␤42-like oligomer distribution. An A␤40-like oligomer distribution is maintained.
Our results elucidate the surprisingly complex structural dynamics of the relatively small A␤ peptide and how peptide segment-specific interactions may control the dynamics. A key determinant of how these interactions occur is the location of hinge or turn regions in the peptide. Evidence exists for turns in the Val 24 -Lys 28 region of A␤ in fibrils (for a recent review see Ref. 68) and in the A␤ monomer (72,73), at the C terminus of A␤42 (71,72), at Glu 22 -Asp 23 in A␤42 (74), and at other sites (62,75,76). Such turns bring regions relatively distant in the primary structure into proximity. For example, the turn at Gly 25 -Asn 27 would induce contacts between the CHC and C terminus, as well as among residues immediately adjacent to the turn itself. In the PICUP experiments reported here, we observed increased frequencies of higher order oligomers formed by the peptides in which the Tyr was substituted at Ala 30 , Val 40 , or Ala 42 . A reasonable explanation for these data is the effect of increased hydrophobicity of the substituted Tyr residue relative to the Ala and Val residues replaced. This would increase the stability of conformers in which hydrophobic side-chain packing occurred, as for example among residues forming the C-terminal turn or residues at the interface between the CHC and C-terminal peptide regions. This explanation also provides a mechanistic rationale for the increased oligomerization potential of the N-terminally substituted peptides, in which Tyr replaced Asp. This substitution produces a more hydrophobic N terminus, the increased solvophobic nature of which would favor intramolecular interactions with the apolar CHC (62). It is interesting in this regard that the N-terminal dipeptide substitution Glu-Val, which also increases the hydrophobicity of this peptide segment, facilitates protofibril formation (77).
In conclusion, one working hypothesis supported by our data is that A␤ conformational dynamics and assembly is a competition among interacting regions. For example, the ability of the extreme C terminus of A␤42 to form a stable turn-or hinge-like structure creates a hydrophobic surface that can interact with the CHC to stabilize assembly-competent conformers. Because it lacks Ile 41 and Ala 42 , A␤40 cannot do so, yet the simple substitution of the A␤40 C-terminal Val with Tyr does create a much more "A␤42-like" peptide. The altered C terminus now "wins" the competition with the N terminus, an outcome observed in silico (62). The effect of Tyr substitution of Asp 1 is mechanistically similar. The substitution increases the stability of N terminus-CHC interactions, a result consistent with the observed compaction of the Tyr 1 -substituted peptides observed in SDS-PAGE. In this case, the N terminus competes more effectively with the C terminus. This mechanism suggests that rational targeting of specific A␤ subregions could be an effective therapeutic strategy. For example, agents could be designed to block C terminus-CHC interactions by binding to either or both subregions. Alternatively, agents that enhanced this interaction might facilitate fibril formation, which increasing evidence suggests may be protective (78). In fact, recent work has shown that C-terminal fragments of A␤, when mixed with the full-length peptide, coaggregate and block the neurotoxic activity of the free peptide (78). Agents that bound to the N terminus and facilitated its binding to the CHC also could block formation of toxic assemblies.