The Ratio of Monomeric to Aggregated Forms of Aβ40 and Aβ42 Is an Important Determinant of Amyloid-β Aggregation, Fibrillogenesis, and Toxicity*

Aggregation and fibril formation of amyloid-β (Aβ) peptides Aβ40 and Aβ42 are central events in the pathogenesis of Alzheimer disease. Previous studies have established the ratio of Aβ40 to Aβ42 as an important factor in determining the fibrillogenesis, toxicity, and pathological distribution of Aβ. To better understand the molecular basis underlying the pathologic consequences associated with alterations in the ratio of Aβ40 to Aβ42, we probed the concentration- and ratio-dependent interactions between well defined states of the two peptides at different stages of aggregation along the amyloid formation pathway. We report that monomeric Aβ40 alters the kinetic stability, solubility, and morphological properties of Aβ42 aggregates and prevents their conversion into mature fibrils. Aβ40, at approximately equimolar ratios (Aβ40/Aβ42 ∼ 0.5–1), inhibits (>50%) fibril formation by monomeric Aβ42, whereas inhibition of protofibrillar Aβ42 fibrillogenesis is achieved at lower, substoichiometric ratios (Aβ40/Aβ42 ∼ 0.1). The inhibitory effect of Aβ40 on Aβ42 fibrillogenesis is reversed by the introduction of excess Aβ42 monomer. Additionally, monomeric Aβ42 and Aβ40 are constantly recycled and compete for binding to the ends of protofibrillar and fibrillar Aβ aggregates. Whereas the fibrillogenesis of both monomeric species can be seeded by fibrils composed of either peptide, Aβ42 protofibrils selectively seed the fibrillogenesis of monomeric Aβ42 but not monomeric Aβ40. Finally, we also show that the amyloidogenic propensities of different individual and mixed Aβ species correlates with their relative neuronal toxicities. These findings, which highlight specific points in the amyloid peptide equilibrium that are highly sensitive to the ratio of Aβ40 to Aβ42, carry important implications for the pathogenesis and current therapeutic strategies of Alzheimer disease.

(A␤) 2 proteins in the form of diffuse and neuritic plaques in regions of the brain that are affected by the disease (1)(2)(3)(4). The discovery of A␤ fibrils as principal constituents of amyloid plaques led to the emergence of the amyloid hypothesis, which implicates the aggregation of A␤ as the primary trigger for a cascade of pathogenic events culminating in neurodegeneration and development of AD (1,(5)(6)(7). A␤ proteins are produced in neuronal and non-neuronal cells as a result of sequential proteolytic cleavage of the type I transmembrane amyloid precursor protein (APP) by ␤and ␥-secretases (8 -12). Depending on the site of APP cleavage by ␥-secretase, A␤ proteins of various chain lengths are generated (13)(14)(15)(16). The predominant A␤ species in human plasma and CSF, as well as in conditioned media of APP-expressing cells, is A␤40 (ϳ90%) followed by A␤42 (ϳ10%). Despite the preponderance of A␤40, in vivo studies reveal that A␤42 is a major constituent of amyloid plaques and suggest that A␤42 aggregation plays a critical role in the initiation of plaque formation and AD pathogenesis (17)(18)(19)(20). In vitro, A␤42 exhibits lower solubility and has the propensity to form protofibrils and fibrillar aggregates at lower concentrations and higher rates than A␤40 or other A␤ variants (21)(22)(23). A␤42 aggregates (protofibrils and fibrils) have also been reported to be more toxic to cultured neurons than A␤40 aggregates (24,25).
Although the majority of late-onset AD cases occur sporadically, genetic mutations in APP or subunits of ␥-secretase (presenilins PS1 or PS2) account for a significant proportion of early-onset familial AD (FAD) cases (26,27). Animal models and in vitro cell culture studies have shown that, in most instances, FAD mutations enhance total A␤ production, promote its aggregation and brain deposition, and/or alter the A␤40/A␤42 ratio in favor of A␤42 production (28 -30). Recent studies in human subjects also highlight the importance of A␤40/A␤42 ratio, rather than the total concentration of A␤, as an important biomarker for AD progression and disease severity (31)(32)(33). To evaluate the consequences of altering the ratio A␤40/A␤42, several groups have investigated the effect of co-expressing the two A␤ variants (A␤40 and A␤42) or altering the expression level of one or the other variant in cellular and animal models of AD. These studies and other studies in human patients demon-strate that the ratio of A␤40 to A␤42 is an important determinant of the distribution of amyloid pathology (i.e. parenchymal or vascular amyloid deposition) in the brains of patients with AD and transgenic AD mouse models (18, 34 -37).
The molecular mechanisms by which changes in the A␤40/ A␤42 ratio modulate the aggregation and toxicity of A␤ and influence the amyloid pathology distribution in the AD brain remain the subjects of considerable debate. A␤40 inhibits fibril formation by A␤42 (22,38), and co-incubation of the two A␤ variants leads to formation of mixed prefibrillar aggregates in vitro (39). A␤40 prevents A␤42-induced neurotoxicity in cultured cells and in vivo (40), underscoring the regulatory effects of A␤40/A␤42 ratio on important events associated with A␤ aggregation and toxicity. More recently, Yan and Wang (41) used differential NMR isotope labeling to demonstrate that A␤40 prevents aggregation of monomeric A␤42 and is capable of being exchanged for A␤42 monomer in A␤42 aggregates.
In the present study, we determined the preferential effect of A␤40 on the kinetic stability, solubility, and fibrillogenesis rate of specific aggregation states of A␤42, including monomers, protofibrils, and fibrils. Additionally, we explored the dynamics of exchange between monomeric A␤40 and A␤42 at the end of protofibrils and fibrils formed by each peptide and determined the effect of these interactions on the aggregate growth and morphology in vitro. Finally, we examined the ability of A␤42 fibrils and protofibrils to seed the aggregation of A␤40 and vice versa by monitoring the seeding effects of homologous and heterologous sequence on the lag phase, elongation phase, and steady-state phase of fibril formation of monomeric A␤40 and A␤42. The specificity of interaction between A␤40 and different A␤42 aggregation states was validated by using A␤40-1 (reverse) as a control peptide. The present work provides novel mechanistic and structural insights into the molecular mechanisms underlying the consequences associated with altered A␤40/A␤42 ratio. The implications of these findings for intervention strategies for AD are also discussed.

EXPERIMENTAL PROCEDURES
Chemicals and reagents of analytical grade were purchased from Sigma-Aldrich unless indicated otherwise. Best quality distilled water was used for preparation of buffers and solutions, which were filtered through 0.65-m DVPP membranes (Millipore) before use.

Preparation of Protofibrillar and Monomeric A␤
A␤ peptides 1-40, 1-42, and 40-1 were synthesized and purified by Dr. James I. Elliott at Yale University (New Haven, CT). Protofibrillar (PF) and monomeric (M) forms of A␤ were prepared according to the protocols described previously (42). Briefly, A␤ peptides were dissolved in 100% DMSO and adjusted to 1 mg/ml by adding distilled H 2 O. The pH of the resultant solutions was adjusted with 2 M Tris base, pH 7.4. After centrifugation (8000 ϫ g at 4°C for 10 min) the supernatant was injected into a size exclusion chromatography column Superdex 75 HR 10/30 (GE Healthcare) that had been equilibrated previously with 10 mM Tris-HCl, pH 7.4. Peptides were fractionated at a flow rate of 0.5 ml/min and eluted in 1.5-col-umn volumes. A␤ elution was monitored by UVabsorbance at three different wavelengths: 210, 254, and 280 nm. Under these conditions, A␤42 eluted as two well separated peaks; one corresponding to the void volume of Superdex 75 containing A␤ protofibrils (A␤42 PF ) and the second peak corresponding to monomeric A␤ (A␤42 M ) (supplemental Fig. 1). A␤40 and A␤40-1 elute predominantly as single peaks corresponding to monomeric species (data not shown). The protein concentrations of the various A␤ fractions was estimated by UV absorbance at 280 nm in 10-mm path-length cuvettes using the theoretical molar extinction coefficient at 280 nm (1490 M Ϫ1 cm Ϫ1 ) (43) and/or using the Micro BCA protein assay (Pierce) when needed.

Fibrillogenesis Studies
Co-incubation of Monomeric A␤40 with Protofibrillar or Monomeric A␤42-Monomeric and protofibrillar preparations of A␤42 and A␤40 obtained by size exclusion chromatography (SEC) were adjusted to a final concentration of 10 -20 M in 10 mM Tris-HCl, pH 7.4, and placed in a 37°C incubator without agitation. For co-incubation studies, protofibrillar and monomeric A␤42 fractions were mixed with monomeric A␤40 at molar ratios (A␤42 M :A␤40 M ) of 10:1, 10:5, and 10:10 (all concentrations in M) and allowed to aggregate at 37°C without agitation. Fibril formation and protein solubility were monitored by thioflavin T (ThT) binding assay, negative staining transmission electron microscopy (TEM), and analytical SEC as described below. In parallel, the specificity of A␤40 interactions with A␤42 was validated by co-incubation with the control peptide A␤40-1.
Fibril Elongation Studies-To probe the effect of monomeric A␤ on the elongation and reassociation of A␤42 fibrils, we first generated A␤42 fibrils by incubating 10 M monomeric A␤42 (A␤42 M ) for 96 h. The fibrils were mechanically fragmented into smaller fibrillar structures (100 -300 nm long) by ultrasonication on ice using a Vibra Cell TM instrument (Sonics Inc.) equipped with a 2-mm diameter microtip (20 ϫ 5 s pulses; amplitude, 40%; output, 6 watts). The sonicated fibrils (A␤42 SF ) were incubated: (i) in isolation; (ii) with monomeric A␤42 (10 M); (iii) with monomeric A␤40 (10 M); or (iv) with a 1:1 mixture of monomeric A␤42:40 (20 M final A␤ concentration). Fibril elongation and reformation of mature A␤ fibrils was monitored by ThT fluorescence and TEM.

Seeding Polymerization Studies
Seeding the Fibrillogenesis of Monomeric A␤ (A␤40 and A␤42) with Fibrils Derived from A␤40 and A␤42-To probe the ability of A␤42 fibrils to accelerate the fibrillogenesis of monomeric A␤40 and vice versa, 20 M monomeric A␤ (A␤42 or A␤40) was incubated with the following: (i) 10 g/ml sonicated A␤42 fibrils (A␤42 SF , once); (ii) 20 g/ml sonicated A␤42 fibrils (A␤42 SF , twice); (iii) 10 g/ml sonicated A␤40 fibrils (A␤40 SF , once); and (iv) 20 g/ml sonicated A␤40 fibrils (A␤40 SF , twice). The extent of fibril formation was determined by ThT fluorescence and TEM. ThT fluorescence data were normalized by subtracting the contribution from sonicated fibrils.
Seeding the Fibrillogenesis of Monomeric A␤ (A␤40 and A␤42) with A␤42 Protofibrils-20 M monomeric A␤ (A␤40 or A␤42) was co-incubated with different molar ratios of protofibrillar A␤42 (monomeric A␤:A␤42 PF ; 20:1, 20:2 and 20:4; all final concentrations in M) and fibril formation was monitored by ThT fluorescence after subtracting the contribution from protofibrillar A␤42. To probe the specificity of interactions between the monomeric and aggregated forms of A␤42 and A␤40, we also evaluated the capacity of sonicated fibrils and protofibrillar species of both peptides (A␤40 and A␤42) to seed the fibrillogenesis of A␤40-1 (20 M) in a similar fashion.

Thioflavin T Binding Assay
ThT binding assay was performed by mixing aliquots of 10 -20 M A␤ with 10 -20 M ThT dye (A␤:ThT 1:1) and 50 mM glycine-NaOH, pH 8.5, in Nunc 384-well fluorescence plates (Fisher Scientific). ThT fluorescence of each sample was measured in an Analyst AD fluorometer (Molecular Devices) at excitation and emission wavelengths of 450 and 485 nm, respectively. The samples were analyzed in duplicates at selected time points.

Transmission Electron Microscopy
5 l of sample was applied to carbon-coated Formvar 200 mesh grids (Electron Microscopy Sciences) and incubated at room temperature for 60 s. The grids were then washed sequentially by depositing 10-l droplets of double distilled sterile water (2 times) followed by a 10-l droplet of fresh 2% (w/v) uranyl acetate, which remained on the grid for 30 s. After each step, the excess solution was blotted with Whatman filter paper, and the grids were vacuum-dried from the edges. The samples were analyzed using a Philips CM-10 TEM microscope operated at 100 kV acceleration voltage.

Analytical Size Exclusion Chromatography
Analytical SEC was performed to quantify the relative amount of soluble (monomeric and protofibrillar) A␤ in solution at selected time points during the aggregation experiments. For this purpose a SEC column Superdex 75 PC 3.2/30 (GE Healthcare) was connected to a Waters Separation Module 2795 equipped with a photo diode array detector (Waters Corp.). Aliquots (150 l) of the samples were centrifuged (8500 ϫ g at 4°C for 10 min), and 50 l of supernatant was injected into the column. Samples were individually analyzed by UV absorbance (wavelengths 210, 254, and 280 nm) at a flow rate of 0.05 ml/min.

Cell Culture Toxicity Studies
Primary Cell Cultures-Rat embryonic (E16) cortical cultures were established using a previously described procedure (44). Briefly, neurons were plated at a density of 30,000 cells/ well in 96-well dishes (Costar TM , Corning) previously coated with poly-L-lysine (Mr 30Ј000 -70Ј000). On in vitro day 4, half of the medium was replaced with freshly prepared Neurobasal TM medium supplemented with 2% B27 (Invitrogen), 1ϫ penicillin-streptomycin (Invitrogen), 0.5 mM L-glutamine, and 15 mM KCl. Subsequently, half of the medium was changed weekly. On in vitro day 23, half of the primary culture medium was replaced with complete Neurobasal medium containing one-fifth or one-tenth volume of amyloid peptide species with varying concentrations of A␤40 M , A␤42 M , A␤42 PF , and A␤42 F and 1:1 molar mixtures thereof. Cells were subsequently incubated with A␤ species for 7 days. All amyloid peptide species were delivered in 140 mM NaCl, 10 mM Tris, pH 7.4.
Immunostaining-Cell cultures were washed with PBS and fixed with 4% paraformaldehyde (Fluka) for 15 min at 4°C. Cultures were subsequently washed with PBS and then incubated for 1 h in a blocking solution of PBS supplemented with 10% normal goat serum (DakoCytomation) and 0.1% Triton X-100 in PBS. The cells were then incubated overnight at 4°C in blocking solution containing mouse monoclonal anti-NeuN antibody (1/400, Chemicon Inc.). The next day, cells were washed and incubated for 2 h with a fluorescent secondary antibody (Cy3-conjugated goat anti-mouse; 1:1,000; Jackson ImmunoResearch Laboratories) followed by PBS washes. Immunostained cells were then analyzed with a BD Pathway 855 Bioimager (BD Biosciences). Images for quantitative analyses were acquired under nonsaturating exposure conditions, and image analysis was performed with NIH ImageJ software.

RESULTS
Isolation and Characterization of Protofibrillar and Monomeric A␤-TEM images of protofibrillar A␤42 (A␤42 PF ) fractions revealed predominantly curvilinear structures with an average length and diameter of 60 -100 and 4 -6 nm, respectively (supplemental Fig. 1B). In addition, spherical aggregates of different diameters (6 -10 nm) were also observed occasionally. The second elution peak corresponding to monomeric A␤42 (A␤42 M ) did not show the presence of any recognizable aggregates by TEM (supplemental Fig. 1D). 10 -20 M concentrations of protofibrillar and monomeric A␤42 species converted readily into fibrils upon incubation at 37°C (supplemental Fig. 1, C and E). After 96 h of incubation, almost all A␤42 M had aggregated into a network of long intertwining fibrils (supplemental Fig. 1E), whereas significant amounts of A␤42 protofibrils persisted and coexisted with fibrils under identical conditions (supplemental Fig. 1C). Unlike A␤42, A␤40 and A␤40-1 (reverse) eluted predominantly as a single monomeric peak under the same solubilization conditions (data not shown) and did not form fibrils under the conditions used for A␤42 fibrillogenesis studies (supplemental Fig. 2, A, C, and E). Fibril formation by monomeric A␤40 (A␤40 M ) required gentle agitation (300 rpm) and higher concentrations (supplemental Fig. 2, B, D, and F).
Monomeric A␤40 Inhibits the Fibrillogenesis of Monomeric A␤42 in a Concentration-dependent Manner-To probe the effects of monomeric A␤40 interactions on the self-assembly and fibril formation of monomeric A␤42, we co-incubated SEC-isolated A␤42 M with increasing concentrations of A␤40 M (see "Experimental Procedures"). We found that A␤40 M inhibited the fibrillogenesis of A␤42 M in a concentration-dependent manner. After co-incubation for 96 h, strong (Ͼ50%) inhibition of monomeric A␤42 M fibrillogenesis was observed in samples containing both peptides at A␤40 M /A␤42 M ratios Ϸ 0.5-1 (Fig.  1A). However, the presence of A␤40 M , even at lower concentrations (A␤40 M :A␤42 M ϳ 0.1) led to a transient population of protofibrillar species (Fig. 1B), which disappeared subsequently. In the absence of A␤40, A␤42 M aggregated rapidly and formed long intertwining fibrillar networks without the accumulation of protofibrils (supplemental Fig. 1E). In contrast, coincubation with A␤40 M favored the formation of short, flexible protofibrillar structures (Fig. 1, C and D, and Scheme 1), which did not appear to convert to mature elongated fibrils. Our data suggest that A␤40 M interferes with the ability of A␤42 M to form mature fibrils but does not interfere with its ability to form higher order prefibrillar aggregates. Under no conditions did we observe that mixtures of monomeric A␤40 and A␤42 remained as stable monomers or formed stable heterodimers on the time scale of our experiments (24 -96 h).
Kinetic Stabilization of A␤42 Protofibrils by Monomeric A␤40-Next, we probed the interactions of monomeric A␤ (A␤40 and A␤42) with protofibrillar A␤42 and assessed the consequences of such interactions on the elongation A␤42 protofibrils into mature fibrils. For this purpose, we co-incubated A␤42 PF with monomeric A␤ (A␤40 M or A␤42 M ) at different molar ratios (A␤ PF :A␤ M , 10:1, 10:5, and 10:10; all final concentrations in M) (Scheme 2). The addition of A␤42 M to A␤42 PF at increasing molar ratios resulted in enhanced ThT binding ( Fig. 2A), consistent with the accelerated conversion of A␤42 PF   OCTOBER 17, 2008 • VOLUME 283 • NUMBER 42 into mature fibrillar structures (Fig. 2E). SEC analysis after 48 h of co-incubation revealed that the majority of monomeric and protofibrillar A␤42 were converted into insoluble aggregates and were no longer detectable in solution supernatant after centrifugation (Fig. 2C). Interestingly, the effect of A␤40 M coincubation on inhibition of A␤42 PF fibrillar conversion was independent of ratio. In the presence of A␤40 M , ThT fluorescence of A␤42 PF remained virtually unchanged even after coincubation for 96 h (Fig. 2B). To determine whether the lack of a rise in ThT fluorescence was due to stabilization of A␤42 PF or formation of low ThT-binding A␤ aggregates, samples containing mixtures of A␤42 PF and A␤40 M were subjected to further analyses by SEC and TEM. When purified A␤42 PF were reinjected into an analytical Superdex 75 PC 3.2/30 SEC column, we consistently observed two peaks corresponding to A␤42 PF and A␤42 M . This may have happened because A␤42 PF are in rapid equilibrium with monomers and/or a population of the A␤42 PF is more susceptible to dissociation upon interaction with the column matrix. After 24 h of co-incubation in the presence of Ն1 M A␤40 M , we observed a disappearance of the monomeric peak (Fig. 2D, 25 min), whereas the intensity and area of the protofibril peak increased (Fig. 2D,  18 min). Upon further incubation, the intensity and area of the protofibril peak was markedly reduced and exhibited a slight shift to higher molecular weight elution (16 min). These results suggest that the presence of A␤40 M blocks the formation of mature A␤42 fibrils but does not interfere with the growth of A␤42 protofibrils and/or possibly leads to the formation of mixed A␤40/A␤42 protofibrillar aggregates. This hypothesis was confirmed by TEM studies that revealed protofibrillar clusters with a predominantly curvilinear morphology for all samples containing mixtures of A␤42 PF and A␤40 M regardless of the peptide ratios (Fig. 2F). The average length of these aggregates was variable and significantly larger than that of SECisolated A␤42 PF (supplemental Lack of Effect of A␤40-1 on Monomeric or Protofibrillar A␤42 Fibrillogenesis-To validate the specificity of the inhibitory effect of A␤40 on the fibrillogenesis of monomeric and protofibrillar A␤42, we co-incubated A␤42 PF or A␤42 M with A␤40-1 at varying molar ratios (A␤42:A␤40-1, 10:1, 10:5, and 10:10; all final concentrations in M). The results of these studies show that A␤40-1 does not inhibit fibril formation by protofibrillar A␤42 at any of the concentrations tested (supplemental Fig. 3A). However, A␤40-1 had a small inhibitory effect on A␤42 M fibrillogenesis at a 1:1 ratio (supplemental Fig. 3B), but this activity was much less pronounced than the inhibitory affect of A␤40 M (Fig. 1A). These results underscore the specificity of the interactions between the two endogenous A␤ variants (A␤40 and A␤42) during amyloid fibril formation.

Effect of A␤40 on Fibrillization of A␤42
Monomeric A␤40 Interferes with Growth and Reassembly of Preformed A␤42 Fibrils-Whereas the above findings indicate that A␤40 inhibits de novo fibrillogenesis of monomeric and protofibrillar A␤42, we next sought to explore the possible effect of A␤40 on the growth and assembly of preexisting A␤42 fibrils. A␤42 fibrils were mechanically disrupted by ultra-sonication to generate a narrow distribution of small, 100 -300-nmlong fibrillar structures with an average diameter of 10 to 12 nm (Fig. 3, B and C). The sonicated A␤42 fibrils (A␤42 SF ) were allowed to reassociate at 37°C as follows: (i) in isolation; (ii) in the presence of A␤42 M (10 M); (iii) in the presence of A␤40 M (10 M); or (iv) in the presence of a 1:1 mixture of A␤42 M and A␤40 M (20 M final A␤ concentration) (Scheme 3). Upon incubation for 72 h at 37°C, the fragmented A␤42 fibrils reassociated to form elongated fibrils with morphological features similar to the parent fibrils (data not shown). The addition of A␤42 M to fragmented fibrils accelerated fibril growth and led to the emergence of mature fibrillar structures (Fig. 3, A and D), whereas the addition of A␤40 M retarded their growth and blocked fibril reassembly as evidenced by significantly reduced ThT fluorescence (Fig. 3A). TEM images of the fragmented fibrils incubated in the presence of A␤40 M also revealed predominantly truncated fibrillar aggregates and the absence of mature fibrils (Fig. 3E). Co-incubation of sonicated A␤42 fibrils with a 1:1 mixture of A␤42 M :A␤40 M resulted in fibril growth and reassembly but to a lesser extent than seen after A␤42 M addition (data not shown). These observations suggest that A␤40, at near equimolar ratios, retards the elongation and maturation of fragmented A␤42 fibrils into classical ThT-binding amyloid fibrils (as shown in Figs. 2E and 3B and supplemental Fig. 1E).
Monomeric A␤42 and A␤40 Are Constantly Recycled and Compete for Binding to the Ends of Protofibrillar and Fibrillar A␤ Aggregates-The above observations highlight a significant degree of molecular cross-talk between A␤40 and different aggregation states of A␤42 (monomer, protofibrils, and fibrils) during amyloid fibril formation. Hence, we ventured to further explore this phenomenon and probe the dynamics of exchange between monomeric A␤42 and A␤40 at the end of protofibrillar and fibrillar forms of A␤42. For this purpose, A␤42 protofibrils (10 M) were incubated in presence of A␤42 M (10 M), A␤40 M (10 M), or a 1:1 mixture of both peptides (20 M final A␤ concentration) for 72 h (Fig. 9). After this, a second 10 M addition of freshly prepared monomeric A␤ species (A␤42 M , A␤40 M , or A␤42 M plus A␤40 M ) was carried out, and the resultant mixtures were incubated for an additional 78 h (150 h total) (Scheme 4). At each step of the process, the presence of fibrils was monitored by ThT binding and negative staining TEM. As expected, the addition of A␤42 M to A␤42 PF resulted in enhanced ThT fluorescence (Fig. 4A, II, 72 h) and the formation   of dense fibrillar networks (Fig. 4B). Interestingly, when A␤40 M was added to this fibrillized A␤42 and incubated for an additional 78 h, ThT fluorescence was substantially reduced (Fig.  4A, II, 150 h), and in addition to fibrils, a substantial amount of curvilinear, non-fibrillar structures were also observed (Fig.  4C). In contrast, when A␤42 PF were co-incubated with A␤40 M for 72 h, there was a negligible rise in ThT fluorescence (Fig. 4A, III, 72 h), and non-fibrillar structures resembling protofibrils were observed as the predominant species (Fig. 4D). However, the addition of A␤42 M overcame the stabilizing effect of A␤40 on protofibrils, resulted in enhanced ThT fluorescence (Fig. 4A,  III, 150 h), and led to emergence of long mature fibrils (Fig. 4E). Finally, serial additions of a 1:1 mixture of monomeric A␤42 M : A␤40 M to A␤42 protofibrils resulted in a slight rise in ThT fluorescence relative to that obtained upon the addition of A␤40 M alone (Fig. 4A, IV). An examination of this sample by TEM revealed mainly curvilinear non-fibrillar structures and short protofibrils (Fig. 4, F and G). The observation that A␤40 M can alter the structure of preformed A␤42 F supports the notion that A␤40 and A␤42 monomers are exchangeable at the growing ends of amyloid aggregates (protofibrils and fibrils).
Seeding Polymerization of Monomeric A␤ (A␤42 and A␤40) with Protofibrillar and Fibrillar A␤ (A␤42 and A␤40)-Amyloid fibril formation occurs via a nucleation-dependent polymerization process akin to protein crystallization (45,46). Conventionally, a nucleated polymerization process is divisible into three closely related phases: (i) a slow nucleation "lag" phase comprising thermodynamically unfavorable interactions of protein molecules leading to the formation of nuclei (or seeds); (ii) an elongation phase leading to the emergence of higher order aggregates (fibrils); and lastly, (iii) a steady-state phase in which there is a dynamic equilibrium between the fibrils and monomers in solution (46). The duration of lag phase is significantly shortened in the presence of preformed "nuclei" or "seeds," which promote accelerated self-assembly and amyloid formation by monomeric protein (47). This seeding effect has also been demonstrated to be a sequence-specific process (48 -50).
A␤42 Fibrils Nucleate the Fibrillogenesis of Monomeric A␤40 and Vice Versa-Previous studies have established the ability of A␤ fibrils to induce fibril formation by monomeric A␤ through the process of nucleated polymerization (22,38). Thus, we sought to explore the ability of both well characterized protofibrillar and fibrillar forms of A␤42 to seed the fibrillogenesis of monomeric A␤40 and A␤42. For these experiments, we isolated aggregate-free monomeric A␤40 and A␤42 by SEC and characterized its fibrillogenesis in the presence of different amounts of fibrillar seeds of either A␤40 or A␤42 (sonicated fibrils of A␤42 SF at 10 and 20 g/ml and sonicated fibrils of A␤40 SF at 10 and 20 g/ml) (Scheme 5). A␤42 M (20 M) alone (with no seeds) exhibited an appreciable rise in ThT binding after 15 h and reached a plateau after 50 h of incubation at 37°C. The addition of sonicated fibrils of A␤42 SF abolished the lag . Monomeric A␤40 and A␤42 exchange along different stages of amyloid formation imparts distinct structural features on amyloid aggregates. A␤42 protofibrils were exposed to sequential addition of either monomeric A␤42 or monomeric A␤40 for an extended period of time (A) (see "Experimental Procedures" and Fig. 8 for details). When added to A␤42 fibrils, monomeric A␤40 alters the fibrillar structure (B) and leads to the emergence of curvilinear morphologies (C). Moreover, the addition of monomeric A␤42 to A␤40-stabilized protofibrils (D) results in conversion of protofibrillar structures into elongated fibrils (E). When added to A␤42 protofibrils, a 1:1 mixture of monomeric A␤42 and A␤40, delays the emergence of mature fibrils (72 h (F) and 150 h (G)). Scale bar ϭ 200 nm. The error bars in A represent the mean Ϯ S.D. in duplicate samples. M, monomeric A␤; PF, protofibrillar A␤42; a.u., arbitrary units. SCHEME 5. A␤42 fibrils nucleate the fibrillogenesis of monomeric A␤40 and vice versa.
phase and reduced the time required to reach near equilibrium levels of fibril formation (ϳ9 -15 h) (Fig. 5A). Interestingly, addition of sonicated fibrils of A␤40 SF also decreased lag phase time but exhibited a longer time to reach near equilibrium (ϳ25-30 h) (Fig. 5A). Regardless of the seeding species, A␤42 M forms similar fibrils that are Ն1 m long and have diameters of 10 -14 nm. These fibrils appear to be composed of 2-3 proto-filaments and exhibit a helical twist spaced at 60 -180 nm (Fig.  5, E and F). For the purpose of discussion, we will refer to these fibrils as type I A␤ fibrils.
In parallel, we investigated the ability of A␤40 SF and A␤42 SF to seed the fibrillogenesis of monomeric A␤40. For this purpose, A␤40 M (20 M) was co-incubated with A␤42 SF (10 and 20 g/ml) and A␤40 SF (10 and 20 g/ml). It is noteworthy that . The addition of fibrillar A␤ (A␤40 or A␤42) to monomeric A␤42 abolishes the lag phase of fibril formation, and similar equilibrium levels are achieved regardless of the nature and concentration of fibrillar seeds added (A). In contrast, when monomeric A␤40 aggregation is seeded with fibrillar A␤ (A␤40 or A␤42), although the lag phase is shortened, steady-state levels depend on the amount of fibrillar seeds added (B). Electron microscopic images of monomeric A␤42 aggregation seeded with fibrillar A␤42 (E) and fibrillar A␤40 (F), along with monomeric A␤40 aggregation seeded with fibrillar A␤42 (G) and fibrillar A␤40 (H), reveal that the structure of the resultant amyloid fibrils is determined by the monomeric A␤ regardless of the nature and concentration of fibrillar A␤ added as seeding effectors. Scale bar ϭ 200 nm. ThT data were normalized by subtracting the ThT fluorescence values obtained from incubation of sonicated fibrils (10 and 20 g/ml) in isolation. M, monomeric A␤; SF, sonicated A␤ fibrils; 1ϫ ϭ 10 g/ml; 2ϫ ϭ 20 g/ml; a.u., arbitrary units. A␤40 M (20 M) alone (with no seeds), under stagnant conditions, did not form amyloid fibrils during the time scale of these experiments ( Fig. 5B and supplemental Fig. 2, A and E) but formed fibrils readily at higher concentrations or under agitating conditions (supplemental Fig. 2, B and F). When A␤40 SF or A␤42 SF was added to the solution of monomeric A␤40, we observed two distinct effects on A␤40 assembly: (i) the elongation phase was sensitive to the concentration of seeding fibrils, and (ii) at all time points, the level of ThT binding near equilibrium was relatively higher in the presence of A␤40 SF than of A␤42 SF , suggesting preferential seeding in the presence of A␤40 SF (Fig. 5B). Moreover, A␤40 M generated similar fibrils regardless of the nature and amount of seeding species (Fig. 5, G  and H We found that even the lowest concentration of A␤42 PF (1 M) was sufficient to accelerate the fibrillogenesis of monomeric A␤42 (Fig. 6A), whereas no such effect was observed (for any molar ratio) when A␤42 PF were added to monomeric A␤40 under identical conditions (Fig. 6B). TEM analysis of these coincubation samples confirmed the validity of the above mentioned observations such that when protofibrillar A␤42 was added to monomeric A␤42, extended networks of fibrillar structures were observed at all molar ratios (A␤42 M :A␤42 PF , 20:4 M (Fig. 6C)). In contrast, the addition of A␤42 PF to A␤40 M resulted in the formation of mainly short flexible nonfibrillar structures (A␤40 M :A␤42 PF , 20:4 M (Fig. 6D)). The absence of mature A␤40 fibrils indicates that A␤42 protofibrils are less efficient as seeding nuclei for A␤40 fibrillogenesis as compared with fibrillar A␤42.
Treatment of Cultured Neurons with Specific Aggregationprone A␤ Species or Mixtures Is Associated with Decreased Neuronal Viability-To assess the differential effects of monomeric A␤ (A␤40 or A␤42) on the toxicity of A␤42 aggregates to pri- mary cortical neurons, we applied these species to cultured E16 rat cortical neurons. Cell survival was measured by the numbers of surviving NeuN-positive cells after 7 days of treatment (see "Experimental Procedures"). A␤40 M showed no significant effect on neuronal viability at concentrations between 0.1 and 10 M (Fig. 7, A and F, and supplemental Fig. 6), consistent with its reduced propensity to aggregate and fibrillize in vitro. At concentrations from 0.1 to 5 M, neither A␤42 M nor A␤42 PF had an effect on neuronal viability ( Fig. 7A and supplemental  Fig. 6). Interestingly, when 5 M monomeric A␤ (A␤40 or A␤42) was mixed with A␤42 PF , only the mixtures containing A␤42 M resulted in significant loss of neuron viability (n ϭ 5 p Ͻ 0.002; Fig. 7, A, C, and D). To determine whether the decrease in neuronal viability was due to an increase in the total concentration of A␤42, we probed the effect of monomeric, protofibrillar, and fibrillar forms of A␤42 on neuronal viability at 10 M con-centration. Both monomeric and protofibrillar A␤42, but not A␤42 fibrils, caused significant reduction in neuronal viability, with monomeric A␤42 consistently showing a pronounced effect (A␤42 M : n ϭ 5 p Ͻ 10 Ϫ5 (Fig. 7, A and E); A␤42 PF : n ϭ 5 p Ͻ 0.01 (Fig. 7, A and G)). Exposure to a 1:1 mixture of A␤42 M and A␤40 M (5 M each) had no effect on neuronal viability (Fig. 7A). These results support a central role of A␤42 in neurotoxicity as well as the notion that A␤-mediated cell death requires ongoing nucleationdependent polymerization of A␤42 (51).

DISCUSSION
Increasing evidence from genetics, clinical, and cell culture studies suggests that the ratio of A␤40 to A␤42, rather than the total amount of A␤, is an important determinant of A␤ aggregation, fibrillogenesis, and toxicity (52). However, the molecular mechanisms by which changes in this ratio alter the aggregation, clearance, and distribution of A␤ in vivo remain uncertain. Similar studies reported previously have examined the effects of single A␤ species or total A␤ concentrations (A␤40 ϩ A␤42) but have not explored the interactions of distinct A␤ aggregate states (e.g. monomers, oligomers, protofibrils, and fibrils). Given that both A␤42 and A␤40 peptides exist in equilibrium between these different states in vivo, it is crucial to determine how the interactions between distinct aggregation states of each peptide influence A␤ aggregation, fibril formation, and cellular toxicity. In the present work, we have uncovered new aspects of A␤ aggregation equilibria by investigating the ratio-dependent effects of monomeric A␤40 on the fibrillogenesis of monomeric and protofibrillar forms of A␤42 and the dynamic reassembly of short A␤42 fibrils. By studying well defined aggregation states of A␤, our findings provide critical new mechanistic insights about how A␤40 and A␤42 interact at different stages along the amyloid formation pathway. These findings have allowed us to develop a new working model of A␤42 and A␤40 interactions, which could serve as mechanistic explanation for in vivo amyloid formation (Fig. 9).

The Ratio of A␤42 Aggregates to Monomeric A␤40 Is an Important Determinant of A␤ Aggregation and Fibrillogenesis-
The in vitro studies presented here demonstrate that maintain- FIGURE 7. Conditions favoring accelerated aggregation of A␤42 in vitro are associated with enhanced neurotoxicity. Cultured rat E16 primary cortical neurons were treated for 7 days with A␤40 (monomeric), A␤42 (monomeric, protofibrillar, and fibrillar) or 1:1 mixtures thereof, as indicated under "Experimental Procedures." Cell viability was quantified after immunostaining for the neuronal marker, neuronal nuclei (NeuN). Graphs represent the means Ϯ S.E. of at least five replicates. Treatment with 10 M A␤42 protofibrils (A and G) caused significant (t test; *, p Ͻ 0.01), but milder toxicity than 10 M monomeric A␤42 (A and E) or a mixture of monomeric and protofibrillar A␤42 (A and C)(5M each) (t test; *, p Ͻ 0.002). In contrast, cells exposed to 10 M monomeric A␤40 (A and F), 10 M A␤42 fibrils (A), or a mixture of monomeric A␤40 and protofibrillar A␤42 (5 M each; A and D) did not show significant neurotoxicity. Veh, vehicle. OCTOBER 17, 2008 • VOLUME 283 • NUMBER 42 ing a delicate balance not only between the monomeric forms of A␤42 and A␤40 but also between the monomeric and aggregated forms of both peptides is essential for maintaining A␤ solubility and preventing A␤ fibrillogenesis and toxicity. First, in the presence of excess A␤40 M (as is expected under normal physiological conditions), A␤42 M aggregation is discouraged because of its low concentration and the inhibitory effect of excess A␤40 M . As the concentration of A␤42 M increases (due to FAD-associated mutations in APP or other local factors), the formation of high molecular weight protofibrillar aggregates or low ThT-binding fibrillar structures is favored (Fig. 8). The molecular size, kinetic stability, and potential toxicity of such species depend on the local concentration of A␤40 at the site of their formation or localization. The stabilization of protofibrillar and/or low ThT-binding A␤ aggregates could explain the lack of clear correlation between amyloid plaque pathology and cognitive decline in patients with AD (53)(54)(55). Figs. 8 and 9 summarize the effect of the ratio of monomeric (A␤40 and A␤42) to aggregated forms of A␤40 and A␤42 in the structural and ThT binding properties of the resultant A␤ aggregates. At higher A␤42 M / A␤40 M ratios, amyloid fibril formation is favored and can be accelerated by seeding with fibrils of either A␤40 or A␤42. At A␤40 M /A␤42 PF and A␤40 M /A␤42 F ratios of 0.5 to 1, the formation of mature amyloid fibrils is inhibited, and predominantly protofibrillar and/or low ThT binding fibrils are formed. Interestingly, A␤42 fibrils seed the fibrillogenesis of both peptides (A␤40 and A␤42); however, A␤42 protofibrils were found to seed the fibrillogenesis of monomeric A␤42 but not of monomeric A␤40. These observations support the notion that A␤42 is more highly fibrillogenic and plays a crucial role in the initiation and progression of A␤ aggregation.

A␤42 Fibril Growth and Reassembly Is Modulated by Constant Recycling and Competition between
Monomeric A␤40 and A␤42-Previous studies have shown that amyloid fibril growth occurs by monomer addition to the growing fibril ends (56 -59). In addition, NMR and mass spectrometry studies have also shown that monomeric subunits constantly recycle in and out of the fibrillar backbone (60,61). Given our findings that A␤40 inhibits the fibrillogenesis of monomeric and protofibrillar A␤42, we hypothesized that A␤40 also intervenes in the growth and reassembly of A␤42 fibrils. We envisioned that his could occur by several mechanisms, two of them being; (i) by competing at the free ends of A␤42 fibrils and preventing addition of monomeric A␤42; or (ii) by interacting and sequestering free monomeric A␤42 in solution and preventing its docking back into A␤42 fibrils. To test these two possibilities, we examined the effect of A␤40 on the growth and reassembly of fragmented A␤42 fibrils (see "Experimental Procedures"). In the presence of excess monomeric A␤42, the fragmented A␤42 fibrils grew and reassembled rapidly to form fibrillar structures identical to that of the parent fibrils (Figs. 3B and 9). In contrast, the addition of monomeric A␤40 retarded A␤42 fibril reassembly, led to the accumulation of truncated fibrillar structures that exhibited low ThT binding and did not convert to mature fibrils in our experimental time scale (up to 150 h). Extending this approach further, we carried out sequential additions of monomeric A␤42 and monomeric A␤40 to different A␤ aggregation states (Fig. 9). The addition of monomeric A␤40 to A␤42 protofibrils resulted in the formation of curvilinear, non-fibrillar aggregates (Figs. 4D and 9). However, upon reintroduction of monomeric A␤42, these structures disappeared, and mature amyloid fibrils of A␤42 emerged as the predominant species (Fig. 4E). These observations provide experimental support for the recycling and exchange of monomeric A␤42 and monomeric A␤40 at the growing ends of A␤ aggregates (protofibrillar and fibrillar). This is consistent with recent NMR-based reports that A␤40, when added to the solution of aggregated A␤42, is capable of displacing A␤42 monomers with the latter being recycled into solution (41). Studies are currently under way to determine quantitatively the rate of monomer exchange or preferential incorporation of A␤40 or A␤42 monomers into A␤42 aggregates (protofibrils and fibrils).
The fact that A␤42 fibrils continue to grow, albeit slowly, when co-incubated with monomeric A␤40 indicates that A␤40 can dock and self-assemble on the free ends of A␤42 fibrils, thus resulting in the formation of mixed fibrils (Fig. 3, A and E) (56). This latter observation also suggests that A␤40 inhibits the formation of mature amyloid fibrils by A␤42 but does not block self-assembly into non-fibrillar aggregates (Fig. 9). Furthermore, at A␤40 M /A␤42 M ratios of 0.5 to 1, we observed nearly complete conversion of monomeric A␤40 and A␤42 into high molecular aggregates over time (Fig. 1B). This latter finding suggests that A␤40-mediated inhibition of A␤42 fibrillogenesis is due to its interaction with some oligomeric/aggregated form(s) of A␤42 rather than to the formation of stable A␤ heterodimers.
Kinetic Stabilization of A␤42 Aggregates by A␤40 May Underlie Its Effect on A␤ Aggregation, Toxicity, and Pathology Distribution in Vivo-The transient formation of protofibrils during the fibrillogenesis of almost all amyloid forming proteins supports the notion that protofibrils are obligate intermediates that are likely to exist, at least transiently, in vivo. A␤ protofibrils and oligomers (dimers, trimers, up to dodecamers) have been prepared in vitro from synthetic peptides or purified from conditioned media of APP-overexpressing cells (62,63), are biologically active entities (64), and bind, albeit weakly, to the amyloid-specific dyes such as ThT and Congo red (65). Several independent studies have implicated A␤ protofibrils as the predominant toxic species through effects that include inhibition of long term potentiation (64,66), disruption of signal transduction (67), and formation of channels or pores in cellular membranes (68 -70). These findings have led to the hypothesis that conversion of small toxic A␤ aggregates into fibrils, and their sequestration within inclusions or plaques, is neuroprotective and may represent an active natural detoxification mechanism (71). Hence, it can be argued that factors that alter the distribution and/or kinetic stability of A␤ protofibrils should influence their aggregation and toxicity in vivo. Therefore, stabilizing a toxic population of protofibrils should enhance toxicity and accelerate disease progression. In contrast, stabilization of a nontoxic protofibrillar or fibrillar forms of A␤ may have beneficial effects. In fact, a recent report suggests that reducing the lifetime of protofibrils by accelerating their conversion into amyloid fibrils protects against A␤ induced toxicity (72).
Our study demonstrates that A␤40 enhances the kinetic stability of A␤42 protofibrils and, at high concentrations, promotes the formation of low ThT-binding aggregates. This could provide possible mechanistic explanations for the effect of the ratio of A␤40 to A␤42 on the distribution of amyloid pathology (parenchymal versus vascular). In the presence of high A␤40 M /A␤42 F or A␤42 M /A␤40 F ratios (Ն10), A␤ fibril formation is accelerated and leads to local accumulation of fibrillar aggregates in brain parenchyma. In contrast, subphysiologic A␤40 M /A␤42 (monomeric or protofibrillar) ratios (0.5-2) favor the accumulation and persistence of soluble A␤ aggregates (protofibrils and other non-fibrillar structures). Being relatively soluble in nature, such aggregates can be transported across different brain compartments, including the vasculature, where they can fibrillize if encountered by supraphysiologic concentrations of monomeric A␤42 and A␤40. This model supports a critical role for A␤42 role in initiating and accelerating A␤40 fibrillization in the vasculature, as has also been reported (19,73).
Fibrillogenesis of A␤ Contributes to A␤ Neurotoxicity-To establish the neurotoxic activities of individual and mixed A␤ FIGURE 9. A␤42 fibril growth and reassembly is modulated by constant recycling and competition between monomeric A␤40 and A␤42. This is a schematic depiction of the dynamic exchange between A␤42 and A␤42 monomers at the ends of A␤42 aggregates and the consequences of such an exchange on the A␤42 fibrillar morphology and ThT dye binding.
(A␤40 and A␤42) monomers and aggregates, we treated cultured neurons with defined aggregate states of A␤42 (protofibrils and fibrils) with or without the addition of monomeric A␤ (A␤40 or A␤42). Comparative analysis of NeuN positivity revealed that neither monomeric A␤ (A␤40 or A␤42) nor aggregated A␤42 (protofibrillar or fibrillar) was sufficient to impair neuronal viability at lower (0.1 and 5 M) concentrations ( Fig. 7A and supplemental Fig. 6). Furthermore, adding monomeric A␤40 to A␤42 (monomeric, protofibrillar, or fibrillar; 5 M each) did not enhance A␤ toxicity (Fig. 7, A and D). However, the addition of monomeric A␤42 to protofibrillar A␤42 (5 M each) or 10 M monomeric A␤42 significantly impaired neuronal viability (Fig. 7, A, C, and E), suggesting that perpetuating the aggregation of A␤42 increases A␤ toxicity. Interestingly, a higher concentration (10 M) of protofibrillar A␤42 or preformed A␤42 fibrils did not show comparable toxicity (Fig.  7, A and G). These data are consistent with a dominant role for A␤42 in impairing neuron viability (19,25) and further implicate the ongoing A␤ self-assembly process (rather than any specific A␤ aggregation state) in A␤-associated toxicity (51,74). These results also add credence to therapeutic strategies aimed at disrupting the nucleation or elongation of amyloid fibrils to inhibit A␤ neurotoxicity in AD.
Conclusions-Potential interactions between A␤40 and A␤42 have been investigated previously by various research groups. These studies have shown that A␤40 can inhibit the aggregation and fibril formation of A␤42 (38,41) and protect cultured neurons from A␤-induced neurotoxicity (40). In addition to confirming the findings of these studies, the present study serves as a critical contribution to the current understanding of A␤40 and A␤42 interactions by examining the specific interplay of biophysically defined A␤ species (and mixtures thereof). Unlike previous studies suggesting that A␤40 blocks the aggregation of monomeric A␤42 (41), our results show that A␤40 does not prevent further aggregation of the various A␤42 quaternary structures (monomeric, protofibrillar, and fibrillar). Instead, we have demonstrated that A␤40 alters the kinetic stability, solubility, and morphological properties of A␤42 aggregates and prevents the formation of mature fibrils. Accordingly, we found that A␤40 inhibits fibril formation of monomeric A␤42 in a concentration-dependent manner (Fig. 1), whereas the effect on protofibril and fibril elongation is concentration-independent (Fig. 2). We have also shown that changes induced by A␤40 are reversible upon reintroduction of high concentrations of A␤42. Our results also confirm previous findings on the seeding capacity of A␤40 and A␤42 fibrils (Fig. 5), but to our knowledge this is the first report showing that A␤42 protofibrils have the capacity to enhance the fibrillogenesis of A␤42 but not of A␤40 (Fig. 6D). Importantly, we have also begun to establish the relevance of the interactions of specific A␤ species in vivo. These results suggest that the continued pursuit of a quantitative and structural understanding of how A␤40 and A␤42 interact may eventually provide better biomarkers and therapies for the treatment of AD.