BRICHOS Domains Efficiently Delay Fibrillation of Amyloid β-Peptide*

Background: Alzheimer disease (AD) is associated with Aβ protein misfolding and aggregation into fibrils rich in β-sheet structure. Results: BRICHOS domains prevent fibril formation of Aβ far below the stoichiometric ratio. Conclusion: Aβ is maintained as an unstructured monomer in the presence of BRICHOS. Significance: BRICHOS domain can have a natural protective role against Aβ aggregation, which may open new routes toward AD therapy. Amyloid diseases such as Alzheimer, Parkinson, and prion diseases are associated with a specific form of protein misfolding and aggregation into oligomers and fibrils rich in β-sheet structure. The BRICHOS domain consisting of ∼100 residues is found in membrane proteins associated with degenerative and proliferative disease, including lung fibrosis (surfactant protein C precursor; pro-SP-C) and familial dementia (Bri2). We find that recombinant BRICHOS domains from Bri2 and pro-SP-C prevent fibril formation of amyloid β-peptides (Aβ40 and Aβ42) far below the stoichiometric ratio. Kinetic experiments show that a main effect of BRICHOS is to prolong the lag time in a concentration-dependent, quantitative, and reproducible manner. An ongoing aggregation process is retarded if BRICHOS is added at any time during the lag phase, but it is too late to interfere at the end of the process. Results from circular dichroism and NMR spectroscopy, as well as analytical size exclusion chromatography, imply that Aβ is maintained as an unstructured monomer during the extended lag phase in the presence of BRICHOS. Electron microscopy shows that although the process is delayed, typical amyloid fibrils are eventually formed also when BRICHOS is present. Structural BRICHOS models display a conserved array of tyrosine rings on a five-stranded β-sheet, with inter-hydroxyl distances suited for hydrogen-bonding peptides in an extended β-conformation. Our data imply that the inhibitory mechanism is reliant on BRICHOS interfering with molecular events during the lag phase.

Misfolding of protein is the underlying cause of at least 27 amyloid diseases, e.g. Alzheimer (AD), 4 Parkinson, and prion diseases. These diseases are associated with a specific form of misfolding, where the protein aggregates into amyloid fibrils with ␤-strands running perpendicular to the fibril axis (1)(2)(3). AD is a progressive neurodegenerative disorder (4), and the senile plaques found in the brains of AD patients contain amyloid fibrils of amyloid ␤-peptide (A␤) (5,6). A␤ is derived from the transmembrane region of amyloid precursor protein (APP), which is processed in one amyloidogenic and one nonamyloidogenic pathway. Amyloidogenic A␤, mainly consisting of 42 residues, is released by ␤and ␥-secretase (7). In a pure system, A␤ aggregation has the appearances of a nucleated polymerization reaction and a phase transition (8). Fibrillar aggregates have low solubility but remain in dynamic equilibrium with free protein (8). In vivo, the aggregation process occurs in the presence of other proteins, membranes, metabolites, etc., each of which may affect both its rate and equilibrium. Many studies indicate that small soluble oligomers are more potent than monomers or fibrils in causing neuronal dysfunction (9 -11), suggesting that either oligomeric intermediates or the aggregation process are toxic.
In search for effective AD prevention and therapy, many studies have identified high or low molecular weight compounds that interfere with secretase activities or A␤ aggregation or derived antibodies against A␤ monomer and oligomers (for review see Ref. 12). A complementary approach is to identify and quantify the effects of intrinsic human components on A␤ aggregation to allow future therapies to be directed toward activation of inbuilt defense routes. Molecular chaperones have evolved to counteract misfolding in the cell. Examples are the heat-shock proteins (Hsp) 70, Hsp90, and Hsp104 (13)(14)(15)(16) and * This work was supported by the Swedish Research Council. the small heat shock protein ␣A-crystallin, which affects fibril formation of A␤ in vitro (17). Hsp70 and Hsp104 also affect the aggregation of ␣-synuclein (18) and Sup35 (19). sHsp, Hsp20, and HspB8 co-localize with amyloid plaques in AD and other neurodegenerative disorders (20). This suggests that the cellular quality control machinery is activated in an attempt to prevent the accumulation of misfolded species, although overload of chaperone capacity may impede the prevention (21); however, co-localization alone is not an indication of disease-related interactions. An extracellular quality control system includes clusterin (apolipoprotein J), haptoglobulin, and ␣ 2 -macroglobulin (22)(23)(24), which are up-regulated during stress, interact with pre-fibrillar species, inhibit amyloid formation, occur in amyloid deposits (25)(26)(27)(28)(29), and promote uptake of misfolded or aggregated proteins by cells for further degradation (30).
The BRICHOS domain of ϳ100 amino acids ( Fig. 1) was identified in the protein Bri2 related to familial British dementia, chondromodulin associated with chondrosarcoma, and lung surfactant protein C (SP-C/pro-SP-C) (31). The BRICHOS domain, found in 12 distantly related protein families with very similar predicted secondary structure, contains two strictly conserved cysteine residues, which form a disulfide bridge in pro-SP-C (32). Pro-SP-C BRICHOS exhibits a chaperone function during SP-C biosynthesis and prevents its transmembrane domain from misfolding (33,34). Mutations in the BRICHOS domain of pro-SP-C lead to pro-SP-C aggregation, amyloid formation, and interstitial lung disease (35,36). Pro-SP-C BRICHOS has broad substrate specificity and binds unfolded or extended stretches of nonpolar residues, but it lacks general chaperone activity (37,38). Nerelius et al. (37) found that pro-SP-C BRICHOS prevents fibril formation of A␤ 40 and forms a complex with A␤ oligomers in vitro. Bri2 is expressed in neuronal tissue and has been shown to suppress A␤ deposition and influence APP processing in transgenic mice, suggesting an in vivo role for Bri2 in AD (39 -41) The C-terminal region of Bri2 has high ␤-sheet propensity. In familial British and Danish dementia, mutants, longer and more aggregation-prone peptides, derived from this region form amyloid fibrils (referred to as ABri or ADan) (42). The Bri2 BRICHOS domain binds the C-terminal region of Bri2 and A␤ 40 in vitro. In the latter case, aggregation and fibril formation are inhibited. Bri2 BRICHOS binds segments of hydrophobic residues flanked by charged residues (43), but its exact substrate specificity is unknown.
A general function of BRICHOS domains may be to limit misfolding of aggregation-prone polypeptide regions. This function may be harnessed in future amyloid disease therapy requiring extensive investigations of the substrate specificity and concentration dependence of the inhibitory effect. Therefore, we investigate here in detail the inhibition of aggregation of A␤ 40 and the more disease-relevant A␤ 42 by recombinant forms of human pro-SP-C and Bri2 BRICHOS. Aggregation kinetics are studied using an approach that generates highly reproducible kinetic data (8), combined with structural investigations by circular dichroism (CD) and NMR spectroscopy, interaction studies by size exclusion chromatography, and morphological analyses by electron microscopy (EM). Our data reveal a very high potency of BRICHOS domains to delay A␤ amyloid formation and provide insights into the mechanism behind the anti-amyloidogenic function of this novel chaperone.

EXPERIMENTAL PROCEDURES
A␤ Peptides-A␤(M1-40) and A␤(M1-42) were expressed in Escherichia coli BL21 from synthetic genes and purified in batch format using ion exchange and size exclusion steps as described (44), which results in highly pure monomeric peptide (amino acid sequences presented in Fig. 1). Purified peptide was divided into 20 -30 identical aliquots and stored at Ϫ20°C. Monomer was then isolated by gel filtration of an aliquot of purified peptide just prior to setting up each of the experiments to remove traces of aggregate formed during freezing and thawing and to exchange buffer to the one used in the respective experiments. The latter part of the monomer peak was collected in low-bind Eppendorf tubes (Axygene) on ice, and the concentration was determined by absorbance and amino acid analysis after acid hydrolysis. The monomer was used as is or was diluted to the desired concentration for the respective experiment.
Bri2 BRICHOS-The expression and purification of the Bri2 BRICHOS domain have been described previously (43). Briefly, the Bri2 BRICHOS construct was expressed in E. coli as a fusion protein with thioredoxin-His 6 and S tags. The protein was then purified using two rounds of nickel-nitrilotriacetic acid-agarose column chromatography. Thrombin was used to remove the thioredoxin and His 6 tag. The eluted protein was analyzed with SDS-PAGE and nondenaturing PAGE. The concentration was determined by amino acid analysis after acid hydrolysis. The amino acid sequence is presented in Fig. 1.
Pro-SP-C BRICHOS-Cloning, expression, and purification were performed as described previously (45), with the following changes. The cells were lysed by lysozyme (1 mg/ml) for 30 min The alignment of pro-SP-C and Bri2 BRICHOS domains is made with ClustalW and corresponds to the pro-SP-C domain as derived from the x-ray structure, Protein Data Bank code 2yad (36). Asterisks and double dots mark identical residues and conservative replacements, respectively. and incubated with DNase and 2 mM MgCl 2 for 30 min on ice. The cell lysate was centrifuged at 6000 ϫ g for 20 min, and the pellet was suspended in 2 M urea in 20 mM Tris, 0.1 M NaCl, pH 8, and sonicated for 5 min. After centrifugation at 6000 ϫ g for 30 min at 4°C, the supernatant was filtered through a 5-m filter and poured on a 2.5-ml nickel-agarose column (Qiagen, Ltd., West Sussex, UK). The column was washed with 100 ml of 2 M urea in 20 mM Tris, 0.1 M NaCl, pH 8, then with 100 ml of 1 M urea in 20 mM Tris, 0.1 M NaCl, pH 8, and finally with 100 ml 20 mM Tris, 0.1 M NaCl, 20 mM imidazole, pH 8. The protein was eluted with 200 mM imidazole in 20 mM Tris, 0.1 M NaCl, pH 8, dialyzed against 20 mM Tris, 0.05 M NaCl, pH 8, cleaved by thrombin for 16 h at 4°C (enzyme/substrate weight ratio of 0.002) to remove the thioredoxin and His 6 tag, and then reapplied to a Ni 2ϩ column to remove the released tag. The protein was then further purified using ion exchange chromatography as described previously (45). The concentration was determined by amino acid analysis after acid hydrolysis. The amino acid sequence is presented in Fig. 1.
Cystatin C-Chicken cystatin C was purified from egg white, as described by Lindahl et al. (46).
Monellin-Single-chain monellin with net charge Ϫ2 (scMN-2; obtained through mutagenesis to incorporate the five substitutions C41S, Q13E, N14D, Q28E, and N50D) was expressed in E. coli from a synthetic gene and purified using ion exchange and size exclusion chromatography as described (47).
Size Exclusion Chromatography-Size exclusion chromatography on a Superdex75 column (GE Healthcare) was performed using a BioLogic HR FPLC system (Bio-Rad). The column was equilibrated and operated in degassed buffer (20 mM sodium phosphate, 200 M EDTA, 0.02% NaN 3 at pH 7.4 or pH 8.0 to prepare samples for aggregation studies, and 10 mM sodium phosphate, 40 mM NaF, pH 7.4, to prepare samples for CD studies). Samples were injected from a 1-ml loop and chromatograms recorded by monitoring the absorbance at 280 nm. To monitor protein interactions, mixtures of A␤ and BRICHOS domains were injected directly after mixing, or after 2 and 20 h of incubation at 37°C in 20 mM sodium phosphate, 200 M EDTA, 0.02% NaN 3 at pH 7.4 or pH 8.0. Fractions (0.3-0.7 ml) were collected during the chromatogram, lyophilized, and analyzed by SDS-PAGE in a 10 -20% gradient gel.
Aggregation Kinetics-Aggregation kinetics were studied by recording the ThT fluorescence intensity as a function of time in a plate reader (FLUOStar Omega from BMG Labtech, Offenberg, Germany). The fluorescence was recorded using bottom optics in half-area 96-well polyethylene glycol-coated black polystyrene plates with clear bottom (Corning Glass, 3881) using a 440-nm excitation filter and a 480-nm emission filter. To each well was then added 90 l of the ice-cold A␤ monomer solution, and the plate was immediately placed in the plate reader at 37°C, with fluorescence read every 6 min with continuous shaking at 100 rpm between readings. A␤(M1-40) was studied alone or with pro-SP-C BRICHOS at concentrations ranging from 17 pM to 17 M or Bri2 BRICHOS at concentration ranging from 60 pM to 6 M. A␤(M1-42) was studied alone or with pro-SP-C BRICHOS at concentrations ranging from 60 pM to 17 M or Bri2 BRICHOS at concentration ranging from 12 nM to 6 M. The concentrations of A␤ and BRICHOS proteins were determined by amino acid analysis after acid hydrolysis.
The half-time t1 ⁄ 2 was obtained by fitting a sigmoidal function to each kinetic trace as shown in Equation 1, and the lag time, t lag , was defined as shown in Equation 2, CD Spectroscopy-CD spectra were recorded in a 10-mm quartz cuvette using a JASCO J-815 spectropolarimeter. Far-UV spectra were recorded at 1-nm intervals between 185 and 250 nm using a scan rate of 20 nm/min, with response time of 8 s and a band pass of 1 nm. A␤(M1-40) monomer was isolated by gel filtration in 5 mM sodium phosphate buffer, pH 7.4, with 40 mM NaF and 200 M EDTA, collected on ice, and divided into three samples that were supplemented with buffer, pro-SP-C or Bri2 BRICHOS to final concentrations of 8 M A␤(M1-40) alone or with 0.8 M pro-SP-C BRICHOS or 0.8 M Bri2 BRICHOS. The samples were heated to 37°C and studied directly or after different times of incubation at 37°C with 100 rpm shaking, up to 18 h. A spectrum of the buffer was recorded separately in the same cuvette and subtracted from the A␤ spectra. Spectra of 0.8 M pro-SP-C or Bri2 BRICHOS were recorded separately and subtracted from the spectra of A␤ and BRICHOS.
NMR Spectroscopy-NMR data were collected at 4°C using a Bruker Avance 400-MHz spectrometer. The protein concentrations were 110 M 15 N-labeled A␤ 40 and 140 M pro-SP-C or Bri2 BRICHOS in 20 mM sodium phosphate buffer at pH 6.8, 10 mM EDTA, and 0.02% NaN 3 . All samples contained 90% H 2 O and 10% (v/v) D 2 O. 15 N, 1 H HSQC experiments were performed with 1024 points ϫ 128 increments and referenced according to the water resonance frequency. NMR data were processed using NMRPipe (48) and analyzed using SPARKY (T. D. Goddard and D. G. Kneller, SPARKY 3, University of California, San Francisco).
Electron Microscopy (EM)-Aliquots of 2 l were taken at different time points, during the aggregation kinetic experiments, for A␤ 40 or A␤ 42 in the absence or presence of pro-SP-C or Bri2 BRICHOS protein. A low concentration of ThT, 0.5 M, was used to enable samples to be picked at time points representing specific levels of reaction progression. The samples were adsorbed on copper grids, stained with 2.5% uranyl acetate in 50% ethanol, and examined and photographed with a Hitachi H7100 microscope operated at 75 kV.
Homology Modeling-A homology model for Bri2 BRICHOS was constructed using coordinates for the human pro-SP-C BRICHOS domain (Protein Data Bank code 2yad (36).). The sequences of Bri2 and pro-SP-C BRICHOS were aligned using ClustalW ( Fig. 1) (49). The insertion in the Bri2 sequence was manually adjusted to align the strictly conserved cysteine residues at positions 164/121 and 223/189 in Bri2 and pro-SP-C, respectively. A model for Bri2 BRICHOS was generated using the software SOD (50) and the molecular graphics program O (51). Side chain rotamers were selected to correspond as closely as possible to those in the experimental model except when a different rotamer had to be chosen to avoid steric clashes.

RESULTS
Aggregation Kinetics-Thioflavin T (ThT) was used as a reporter on fibril formation in kinetic experiments for A␤(M1-40) or A␤(M1-42), herein referred to as A␤ 40 and A␤ 42 , respectively, alone or with different concentrations of the BRI-CHOS proteins ranging from 0.00001 to 0.6 molar eq. Examples of aggregation kinetics for A␤ 40 alone and with 0.018 to 0.18 molar eq of pro-SP-C BRICHOS are shown in Fig. 2A, and with 0.006 to 0.061 molar eq of Bri2 BRICHOS in Fig. 2B. The mid-point of the aggregation process, t1 ⁄ 2 , and the lag time were obtained by fitting a sigmoidal function to each kinetic trace and is plotted versus molar ratio of BRICHOS/A␤ 40 in Fig. 2C. Clearly, the lag time for A␤ 40 aggregation has increased extensively in the presence of pro-SP-C or Bri2 BRICHOS, whereas the elongation rate is largely unaffected. Very large effects on the lag time are observed far below equimolar concentration of pro-SP-C or Bri2 BRICHOS relative to A␤ 40 . A doubling of the lag time for A␤ 40 aggregation requires ϳ0.01 molar eq of pro-SP-C BRICHOS, and a 10-fold increase in lag time is seen around 0.01 eq of Bri2 BRICHOS or 0.03 eq of pro-SP-C BRICHOS. The retarding effect increases with increasing BRICHOS concentration, and the lag time exceeds 1 week and becomes practically difficult to quantify above 0.025 (1 Bri2 BRICHOS per 40 A␤ 40 ) or 0.06 molar ratio (1 pro-SP-C BRICHOS per 16 A␤ molecules). Thus, both BRICHOS proteins are very potent inhibitors of A␤ 40 aggregation, with the strongest effects observed for Bri2 BRICHOS.
Both pro-SP-C and Bri2 BRICHOS retard also the aggregation of A␤ 42 , and only substoichiometric amounts of the BRICHOS proteins are required (Fig. 2, D-F). At 0.06 -0.1 molar ratio (1 BRICHOS protein per 10 -16 A␤ 42 ), both the lag time and half-time are doubled compared with the uninhibited case, and the elongation rate is not affected. A 10-fold increase in lag time is seen at ϳ0.6 molar eq of the BRICHOS proteins, under which conditions the elongation rate is found to be significantly reduced. Although strong effects are seen on A␤ 42 aggregation kinetics, it is clear that higher concentrations of the BRICHOS proteins are needed to exert the same effect as on A␤ 40 aggregation, and the retarding effects of pro-SP-C and Bri2 BRI-CHOS are quantitatively more similar in the case of A␤ 42 .
The BRICHOS domain contains two strictly conserved Cys residues, which have been shown to form a disulfide bond in pro-SP-C BRICHOS (32). We performed fibrillation kinetics experiments for A␤ 40 and A␤ 42 in the absence and presence of Bri2 or pro-SP-C BRICHOS that had been reduced with 1 mM DTT prior to the ThT measurements. Both BRICHOS proteins were found to be significantly less efficient in hindering A␤ from forming fibrils in their reduced forms (supplemental Fig.  1). A Bri2 BRICHOS construct lacking the N-terminal S tag shows virtually identical effects on A␤ 42 aggregation kinetics as the S-tagged counterpart, showing that the presence of the S tag does not affect BRICHOS activity (data not shown).
Additional control experiments were set up to study the aggregation kinetics of A␤ 40 and A␤ 42 in the presence of the following three proteins: human anti-thrombin (HAT), egg white cystatin C, and a single chain monellin variant (scMN-2). HAT belongs to the serpin family, other members of which have been reported to affect A␤ aggregation (52,53). Egg white cystatin C has about the same molecular mass as the BRICHOS domain, and scMN-2 was chosen because it has the same net charge (Ϫ2) as pro-SP-C BRICHOS to mimic any nonspecific protein effect. Each control protein was added at 0.01 and 0.1 molar eq to A␤ 40 , or to A␤ 42 , and aggregation followed by the ThT assay (supplemental Fig. 2). HAT and scMN-2 were found to inhibit aggregation of A␤ 40 but required 0.01 and 0.1 molar eq, respectively, to produce the same effects as 0.006 molar eq of pro-SP-C BRICHOS or 0.0006 molar eq of Bri2 BRICHOS. Thus, pro-SP-C BRICHOS was found to be a 10 -100-fold more effective inhibitor of A␤ 40 aggregation than HAT and scMN-2, and Bri2 BRICHOS 100 -1000-fold was more effective than these control proteins. HAT also showed effects against A␤ 42 when added at 0.1. molar eq. No inhibiting effect was observed for cystatin C (supplemental Fig. 2).
CD and NMR Spectroscopy-Structural transitions during the aggregation process were studied using CD spectroscopy (Fig. 3A). The data for A␤ 40 alone agrees with other reports (54), and it shows a continuous progression from a spectrum typical for a random coil peptide toward a spectrum indicative of ␤-sheet structure. The structural transition starts to develop, although the aggregation process as observed by ThT fluores- cence is still in the lag phase and thus reports on the appearance of intermediates with ␤-sheet structure. In the presence of 0.1 molar eq of pro-SP-C BRICHOS or Bri2 BRICHOS, the structural transition appears to be delayed; spectra at 300 min and 15 h report on mainly random coil structure. Thus, the presence of BRICHOS reduces the concentration of intermediates with ␤-sheet structure and keeps A␤ in a mainly unstructured state during the extended lag phase.
To further investigate structural conversions of A␤ 40 when interacting with BRICHOS, 1 H-15 N HSQC experiments of 15 Nlabeled A␤ 40 with and without pro-SP-C BRICHOS were performed. The 1 H-15 N HSQC spectra for A␤ 40 show typical random coil resonances as reported previously (Fig. 3B, red spectrum) (55). With a 30% excess of pro-SP-C BRICHOS added, 1 H-15 N HSQC spectra show no difference in resonances compared with A␤ 40 alone (Fig. 3B, blue spectrum). This argues that the interaction between A␤ and pro-SP-C BRICHOS does not induce structural conversion of the NMR visible A␤ 40 peptide. After 48 h at room temperature, a new spectrum was collected without any change in resonances (data not shown). After 2 weeks at room temperature, the spectrum of A␤ 40 in the presence of BRICHOS still has the same resonances as the spectrum of newly dissolved A␤ 40 alone (Fig. 3B, black spectrum), although a decrease in the intensity of the resonance signals suggests that some aggregation had occurred over time. The same results were obtained with Bri2 BRICHOS (supplemental Fig. 3). The sample with only A␤ 40 had, after 2 weeks at room temperature, aggregated, and no resonances were visible.
Size Exclusion Chromatography-The interaction between the BRICHOS proteins and A␤ was studied using gel filtration. Samples of 8 M A␤ 40 and 0.8 M pro-SP-C BRICHOS or 8 M A␤ 40 and 0.8 M Bri2 BRICHOS were incubated for 20 h at 37°C followed by gel filtration on a Superdex 75 column and SDS-PAGE of collected fractions (Fig. 4 and supplemental Fig.  4). At 20 h, A␤ alone has fibrillated and reached the equilibrium plateau, whereas samples containing 0.1 molar eq of pro-SP-C or Bri2 BRICHOS are still in the lag phase (Fig. 2). The data in Fig. 4 show clearly that in the mixed samples almost all of A␤ 40 is monomeric after 20 h, whereas a minor fraction (Ͻ1%) elutes at 10 -12 ml together with pro-SP-C BRICHOS (supplemental Fig. 4). Similar results are observed for 8 M A␤ 40  Stopping Experiments-To monitor the effect of BRICHOS addition during an ongoing aggregation process, samples with 8 M A␤ 40 were monitored by recording the ThT fluorescence intensity as a function of time. 800 nM Bri2 BRICHOS was added from a concentrated stock before start or at time points ranging from 0.3 to 11.2 h. A similar experiment was performed for 3 M A␤ 42 with 1.8 M Bri2 BRICHOS added before start or at time points ranging from 6 to 109 min. As shown in Fig. 5 and supplemental Fig. 5, the aggregation process is delayed if BRICHOS was added anywhere during the lag time. If BRI-CHOS protein was added during the early part of the elongation phase, the process appears to halt with no further growth of ThT-positive aggregates. When added close to the mid-point of  the transition, the BRICHOS protein seemed to halt the process from further progression or cause the process to reduce its speed and progress at a lower rate. When added at the end of the transition, no effect was seen.
EM Analyses-Samples were taken for EM analyses at different time points of the fibril formation process, which was followed by ThT fluorescence (Fig. 6). At the end of the lag phase, A␤ 42 on its own has formed typical fibrils (Fig. 6d), whereas such structures are rare in the presence of pro-SP-C and absent in the presence of Bri2 BRICHOS at the equivalent time point (Fig. 6, e and f). When A␤ 42 alone has reached equilibrium, fibrils are clearly present in this sample (Fig. 6g). At the same time samples with BRICHOS are in the elongation phase and fibrils have started to appear (Fig. 6, h and i). At later time points, when the ThT fluorescence has reached the plateau level also for the samples with BRICHOS, EM analyses show the presence of larger amounts of amyloid fibrils (Fig. 6, k and j). Similar results were observed for A␤ 40 with fibrils detected at the end of the lag phase in the absence and presence of BRI-CHOS and larger amounts of fibrils when the ThT plateau was reached (supplemental Fig. 6).

DISCUSSION
Stimulation of chaperone activity may be explored as a route toward prevention or treatment of amyloid diseases (30,56) requiring detailed molecular understanding. Broad substrate specificity makes it difficult to target chaperones to specific proteins, and some chaperones seem to promote amyloid formation (57). Therefore, it will be essential to identify chaperones with some level of specificity for A␤ and/or natural co-localization with APP/A␤ and to know at what stage of APP processing and/or A␤ aggregation a specific chaperone might interfere.
Here, we investigate the mechanism by which the BRICHOS domains affect A␤ fibrillation kinetics, structure, and complex formation between A␤ and BRICHOS. We find that BRICHOS domains from two human proteins, Bri2 and pro-SP-C, delay A␤ fibril formation in a concentration-dependent manner. The aggregation of the more disease-relevant A␤ 42 is retarded at substoichiometric BRICHOS/A␤ 42 ratios with a doubling of the aggregation lag time at 1 BRICHOS domain per 10 A␤ 42 . This is an important result, which may be harnessed in the design of future AD therapy. Even lower BRICHOS concentration is needed to inhibit aggregation of A␤ 40 . As little as one Bri2 BRICHOS per 400 A␤ 40 (or 1 pro-SP-C BRICHOS per 160 A␤ 40 molecules) is needed for doubling of the lag time. Above one BRICHOS per 40 or 10 A␤ 40 , the aggregation process is so much retarded that it does not occur within 1 week as compared within a few hours for A␤ 40 alone.
For A␤ 40 , the lag phase is prolonged in the presence of BRICHOS, although the elongation rate is essentially unaffected. The results of the stopping experiments show that an ongoing aggregation process can be strongly delayed as long as BRICHOS domains are added during the lag phase. The process is only temporarily halted if BRICHOS is added at t1 ⁄ 2 , the mid-point of the process, after which it is too late to interfere. This is well in line with the previous observation that addition of pro-SP-C BRICHOS to already formed fibrils resulted in no apparent effect (37). The present results imply that BRI-CHOS domains interfere with molecular events that occur during the lag phase. For A␤ 42 in the presence of BRICHOS, the FIGURE 6. EM of A␤ assemblies at different time points during fibril formation. Aliquots from aggregation kinetics experiments with A␤ 42 and BRICHOS proteins were adsorbed on copper grids, stained with uranyl acetate, and analyzed by EM. Each sample is taken at a specific time point during the aggregation process as followed by ThT fluorescence, which is shown schematically for A␤ 42 alone (black line) or A␤ 42 co-incubated with BRICHOS protein (dotted line). Electron micrographs of only pro-SP-C or Bri2 BRICHOS incubated for the same time periods showed no assemblies (supplemental Fig. 6). effect on the lag phase is less pronounced, and the apparent elongation rate is affected. This suggests that the BRICHOS mechanism of action may differ for A␤ 40 and A␤ 42 . Further detailed kinetic analyses are required to allow a description of the mechanisms of action for BRICHOS on A␤ 40 and A␤ 42 .
Another potent inhibitor, the small engineered protein Z(A␤3), was recently reported to inhibit aggregation of A␤ by monomer binding and monomer depletion from solution (58,59). The extensive delay of fibrillation by very low concentrations of BRICHOS proteins, far below equimolar amounts, implies that its mechanism of action is not to bind and reduce the concentration of free monomer. The lag time for aggregation of pure A␤ 42 displays a strong concentration dependence (8), from which we can estimate that the removal of 0.1 molar eq by monomer capture would cause only a 17% increase in lag time, compared with the 100% increase seen with 0.1 molar eq of pro-SP-C or Bri2 BRICHOS. At the same molar ratio, A␤ 40 aggregation is retarded beyond detection.
The mechanism of action of the BRICHOS domain must therefore clearly be something else than monomer binding to BRICHOS. CD and NMR spectroscopy as well as gel filtration experiments were employed to study structural transitions and protein-protein interactions during the lag phase. The CD and NMR analysis implies that A␤ 40 is kept in a mainly unstructured state during the extended lag phase in the presence of BRICHOS. In addition, size exclusion chromatography identifies only a minor fraction of A␤ eluting together with the BRICHOS proteins, although the majority of the peptide elutes as monomer. The results from the EM studies indicate that the presence of BRICHOS delays the formation of fibrillar assemblies, which start to appear at the end of the lag phase for A␤ alone, and at the end of the extended lag phase for samples containing BRICHOS. Eventually, similar amounts of amyloid fibrils are formed both in the absence and presence of BRICHOS. This suggests that BRICHOS interferes with nucleation processes. Because no binding of A␤ monomer to BRICHOS was detected, it is likely that BRICHOS instead binds to assemblies formed early in the aggregation processes. The chaperones Hsp104, Hsp70, and Hsp40 are active at substoichiometric amounts and preferentially interact with intermediates formed in the fibril formation process (13,14). In overexpressing cell lines, Hsp70 protects against A␤ 42 -induced neuronal death (60), whereas ␣B-crystallin, Hsp27, Hsp20, and HspB2/B3 reduce A␤ cell toxicity (61). The extracellular chaperones clusterin, ␣ 2 -microglobulin, and haptoglobin inhibit aggregation by binding to prefibrillar A␤ 42 species, and clusterin decreases A␤ 42 toxicity (26,28,29).
The present results indicate that the pro-SP-C and Bri2 BRICHOS domains interfere with A␤ fibril formation in mechanistically similar ways, although they are evolutionarily distant, with sequence identity less than 25% (Fig. 1). We have speculated that the BRICHOS domains bind to regions within their respective precursor proteins that show high propensity to form ␤-sheet structures, preventing them from aggregation during biosynthesis (43,62,63). The A␤ central hydrophobic segment flanked by charged residues ( 14 HHKLVFFAED 23 ) is suggested as a target for both BRICHOS proteins (34,38,43).
To gain further insight into potential binding determinants in BRICHOS, a homology model for Bri2 BRICHOS was built using the x-ray structure of pro-SP-C BRICHOS (36) as a template (Fig. 7A). The structure consists of a central five-stranded ␤-sheet with one helix on each side. The ␤-sheet is highly conserved throughout BRICHOS sequences; in particular, a number of highly conserved tyrosine residues are found centrally located on face A of the sheet (Fig. 7B). This exposed array of phenol rings is a striking structural feature with resemblance to polyphenolic compounds found to retard A␤ aggregation (64). Common to other amyloid retarding compounds, for example N-methylated peptides (65), and polymeric nanoparticles with amide units in the backbone (66,67) is the hydrogen bonding capacity, which may allow these compounds to interact with edge strands of growing aggregates to prevent further propagation. The hydroxyl groups of the conserved BRICHOS tyrosine array (maybe together with adjacent His/Asp/Glu/Lys) may be able to form hydrogen bonds with peptide in extended ␤-structure, thereby preventing further growth of aggregated species. Indeed, in our structural model, the distances between the hydroxyl groups on the conserved array of Tyr-113, Tyr-122, FIGURE 7. Bri2 BRICHOS structure model. A, ribbon diagram of the molecular model of Bri2 BRICHOS, based on the recent crystal structure of pro-SP-C BRICHOS, Protein Data Bank code 2yad (36). The secondary structure elements are labeled from the N-terminal strand, ␤1 to the last and fifth strand ␤5. The ␤-sheet is enclosed by two helices ␣1 and ␣2. All residues that are shown as stick representation are situated on face A, which is the putative binding site. B, sequence comparison between Bri2/pro-SP-C in the ␤-sheet face A. Identical residues in pro-SP-C and Bri2 BRICHOS are boxed in green, and conservative replacements are in open boxes. and Tyr-195, match the distances between the carbonyls of residues i, i ϩ 2, and i ϩ 4 of a peptide in extended ␤-conformation. Mapping sequence differences to the structure reveals a number of changes from hydrophobic residues in pro-SP-C BRICHOS to charged residues in Bri2 BRICHOS on face A of ␤-sheet (Fig. 7B), suggesting that this surface may be important for substrate specificity, as hypothesized previously (37,38,43).
There are clear pathological and clinical similarities between familial British or Danish dementia on the one hand and AD on the other hand, and it has been shown that Bri2 interacts with A␤ and regulates APP processing (68). Bri2 suppresses A␤ deposition in APP transgenic mice by influencing APP processing (39). Together with our results, it is tempting to speculate that the Bri2 BRICHOS domain indeed has a natural protective role against A␤ aggregation and AD. Our results may open new routes toward AD therapy based on this natural human chaperone.
In conclusion, both A␤ 40 and A␤ 42 aggregate significantly slower in the presence of BRICHOS domains from two human proteins, pro-SP-C and Bri2. The main effect is on the lag time for aggregation. The inhibitory mechanism involves interference with nucleation processes, such that the main pool of the peptide is kept monomeric and unstructured for extended time.