Recombinant Bri3 BRICHOS domain is a molecular chaperone with effect against amyloid formation and non-fibrillar protein aggregation

Molecular chaperones assist proteins in achieving a functional structure and prevent them from misfolding into aggregates, including disease-associated deposits. The BRICHOS domain from familial dementia associated protein Bri2 (or ITM2B) probably chaperones its specific proprotein region with high β-sheet propensity during biosynthesis. Recently, Bri2 BRICHOS activity was found to extend to other amyloidogenic, fibril forming peptides, in particular, Alzheimer’s disease associated amyloid-β peptide, as well as to amorphous aggregate forming proteins. However, the biological functions of the central nervous system specific homologue Bri3 BRICHOS are still to be elucidated. Here we give a detailed characterisation of the recombinant human (rh) Bri3 BRICHOS domain and compare its structural and functional properties with rh Bri2 BRICHOS. The results show that rh Bri3 BRICHOS forms more and larger oligomers, somewhat more efficiently prevents non-fibrillar protein aggregation, and less efficiently reduces Aβ42 fibril formation compared to rh Bri2 BRICHOS. This suggests that Bri2 and Bri3 BRICHOS have overlapping molecular mechanisms and that their apparently different tissue expression and processing may result in different physiological functions.

To maintain cellular homeostasis, organisms have evolved ways to prevent misfolded and aggregated proteins from causing cellular damage 1 . Molecular chaperones are part of this defence arsenal and can interfere with virtually all steps from co-translational protein folding to degradation, which promote the formation of correctly folded and functional proteins over-accumulation of self-assembled aggregates that may cause toxicity 1,2 .
Amyloid fibrils are the end-result of one type of misfolding and are composed of β-sheets in which the β-strands run perpendicular to the fibril axis 3 . Many proteins can form amyloid-like fibrils in vitro 4,5 but tissue deposits of specific proteins in the amyloid state are coupled to only about 40 diseases, including many severe neurodegenerative disorders 6 . Alzheimer's disease (AD) is the most common cause of dementia, and there is no disease-modifying treatment available. AD is histopathologically characterised by the presence of extracellular amyloid plaques mainly in the neocortex and hippocampus, and these plaques contain fibrils of the amyloid β-peptide 1-40 (Aβ40 and 1-42 (Aβ42) 7 . AD is also defined by the formation of another type of protein depositintracellular neurofibrillary tangles composed of hyperphosphorylated tau protein 8 . The formation of Aβ and tau deposits is thought to result in the loss of synapse function, eventually leading to cell death and brain atrophy. The exact mechanisms, by which protein misfolding and aggregation result in neuronal loss and dysfunction in AD are unresolved, but it is believed that the main toxic agents are soluble oligomeric forms of Aβ rather than the mature fibrils 9,10 .
The BRICHOS domain was initially defined by multiple amino acid sequence alignments of the proteins Bri2, chondromodulin-1 and prosurfactant protein C (proSP-C) 11 . The BRICHOS domain has been found in 10 protein families in humans, and BRICHOS containing proproteins share a common architecture 12 . These proproteins contain specific regions with high β-sheet propensity, which BRICHOS is believed to chaperone during biosynthesis [13][14][15] . Mutations in the BRICHOS domain of proSP-C result in lethal disease already in infancy, associated with lung amyloid composed of the proSP-C β-prone region, which supports the hypothesis that BRICHOS is a chaperone that prevents amyloid toxicity 16 . Bri2 is expressed in the CNS and in peripheral tissues. Mutations that extend its β-prone region result in amyloid deposits and familial British and Danish dementia 17,18 . Recombinant human (rh) BRICHOS domains from Bri2 and proSP-C reduce amyloid fibril formation and neurotoxicity of Aβ40 and Aβ42 [19][20][21][22][23] , and transgenic overexpression of BRICHOS reduces Aβ42 related pathology in fruit flies 21,24 . These findings suggest that BRICHOS domains are able to prevent fibril formation and toxicity of other clients than the β-prone regions of their respective proprotein. In support of this, rh proSP-C BRICHOS prevents fibril formation of medin, a fragment of lactadherin that forms amyloid in the human aortic wall 25 , and we have recently found that Bri2 BRICHOS potently inhibits amyloid formation and toxicity of type 2 diabetes associated islet amyloid polypeptide (IAPP) 26 .
The conservation of amino acid sequences between BRICHOS domains from different proproteins is low, but their predicted secondary structures are similar 12,14 , and three residues are strictly conserved -two Cys and one Asp residue 11,12 .
Bri3 is a 267-residue protein, and as other BRICHOS proteins, it contains an N-terminal part, a transmembrane segment, a linker region, the BRICHOS domain and a C-terminal segment. Bri3 is structurally similar to Bri2 with an overall 44% sequence identity 12,27 . However, Bri3 differs from Bri2 in several ways -it is exclusively expressed in the CNS, the C-terminal segment is cleaved by furin but the BRICHOS domain is apparently not released 28 , and it interacts with amyloid plaques and Aβ in different manners than Bri2 29 .
BRICHOS domains from different proproteins have been found to interfere with distinct steps in the reaction pathway from monomeric Aβ42 to fibrils. ProSP-C BRICHOS exclusively blocks the secondary nucleation step 23,30 . Bri2 BRICHOS, however, has shown a more complex inhibition effect by blocking both a surface catalysed secondary nucleation and elongation of fibrils 21,30 .
Moreover, Bri2 BRICHOS, but not the proSP-C BRICHOS domain also prevents non-fibrillar aggregation of partly denatured proteins 21 . These observations raise the question of how different types of chaperone activities of BRICHOS are mediated. Isolated rh Bri2 BRICHOS monomers potently prevent hippocampal neuronal network toxicity of Aβ42, while dimers strongly suppress Aβ42 fibril formation. The dimers assemble into high-molecular-weight oligomers with an apparent two-fold symmetry, which are efficient inhibitors of non-fibrillar protein aggregation. Thus Bri2 BRICHOS affects qualitatively different aspects of protein misfolding and toxicity via different quaternary structures 31,32 . Rh Bri3 BRICHOS also reduces Aβ42 fibril formation but less efficiently than Bri2 BRICHOS 29 but the detailed effects of Bri3 BRICHOS on Aβ42 fibril formation have not been studied. It is not known whether it similarly to Bri2 BRICHOS also prevents non-fibrillar protein aggregation, or whether it like proSP-C BRICHOS exclusively affects fibril formation.
In this study, we sought a more comprehensive understanding of Bri3 BRICHOS and its ability to reduce Aβ42 amyloid fibril formation and non-fibrillar aggregation.

Results and Discussion
Homology model of Bri3 BRICHOS and comparison to Bri2 BRICHOS. Prediction using the I-TASSER web server 33 gave similar overall tertiary structures for Bri3 and Bri2 BRICHOS ( Fig. 1) with confidence scores (C-scores) of −1.37 for Bri2 BRICHOS and −1.42 for Bri3 BRICHOS. The only determined structure of a BRICHOS domain, from proSP-C, showed a central five-stranded β-sheet structure with one α-helix on each side 34 . Though the sequence identity between different BRICHOS domains is variable, for example between proSP-C and Bri2 BRICHOS it is less than 25% 20 , the secondary structure elements and tertiary structures are predicted to be conserved 12 . Many hydrophobic residues of β-sheet face A (see Fig. 1) of pro-SP-C BRICHOS are changed to charged residues in Bri2 BRICHOS, which correlates well with the physico-chemical properties of their supposed physiological target peptides (the hydrophobic transmembrane segment in case of proSP-C and the highly charged C-terminal Bri23 region in the case of Bri2), suggesting that face A is important for the substrate specificity 20 .
Sequence alignment of Bri2 and Bri3 BRICHOS (Fig. 1) shows that they share 58% identical residues and reveals that there are several residue replacements between the two domains where the physicochemical properties are largely unchanged, making 79% of residues in the two BRICHOS domains similar or identical. In total, 20 out of 51 non-identical residues include charge variations. Judging from the predicted structures most of the residue substitutions that result in different physicochemical properties in Bri3 BRICHOS compared to Bri2 BRICHOS are located in unstructured loops or in β-sheet with residue side chains orientated towards face B (Fig. 1c,f). More specifically, five residues with charge variations are in the central β-sheet and 15 are located in the loop regions, except for one residue that is in the helix 2 on the face B side. Side chains are clearly facing towards face B for 11 charge-changed residues, five of them are in β-sheet and six are around helix 2. From the x-ray structure of proSP-C BRICHOS 34 , it was suggested that face A of the β-sheet interacts with the client peptide regions during biosynthesis 14,20,34 . Face A harbours none of the now predicted non-conservative replacements between Bri2 and Bri3 BRICHOS but instead has four out of five conserved charged residues in the central β-sheet (Fig. 1b,c,e). The 23-residue proposed client peptide regions of Bri2 and Bri3 contain 18 identical residues. These observations suggest that Bri2 and Bri3 BRICHOS domains may have similar functions in terms of interacting with their respective β-prone region during proprotein biosynthesis, and that they have diverged during evolution primarily by varying the loop regions and the face B of the β-sheet. Further experiments, in particular changing the physicochemical properties of either side of the central β-sheet in different BRICHOS domains by site-directed mutagenesis will be needed to test this hypothesis.

Characterisation of recombinant human (rh) Bri3 BRICHOS domain.
Size exclusion chromatography (SEC) (Fig. 2a, Supplementary Fig. S1) and native PAGE (Fig. 2b) show that rh Bri3 BRICHOS migrates mainly as high molecular weight oligomers. SDS-PAGE under non-reducing conditions shows that disulfide bonds to a significant extent hold together the larger aggregates. Also, monomers, dimers and smaller oligomers are detected by SDS-PAGE under non-reducing conditions (Fig. 2b), which indicates that non-covalent inter-subunit interactions contribute to the formation of the high molecular weight oligomers. Rh Bri3 BRICHOS The disulfide bond between the two strictly conserved cysteine residues (distance between the sulfur atoms of the two Cys is 2 Å) is highlighted in gold for Bri3 BRICHOS Residue side-chains that are highlighted by space-filling show either conserved (b and e) or changed (c and f, the residues shown are those present in Bri3 BRICHOS) charged residues between Bri2 and Bri3 BRICHOS, dark blue: positive charge, red: negative charge, grey: loss of charge to a neutral side-chain. (g) Sequence alignment of Bri2 and Bri3 BRICHOS. Shading represents identical residues (green) and similar residues (yellow), based on a Blosum62 score matrix with a threshold of 1. Residues indicated by small boxes above the sequences highlight the changed residues in panels c and f using the same colour-coding. www.nature.com/scientificreports www.nature.com/scientificreports/ monomer has a calculated mass of 17.3 kDa, and it migrates as expected on SDS-PAGE (Fig. 2b). Comparison of the SEC elution volume to standards with known hydrodynamic Stokes radii gives an apparent diameter of 17 nm for the Bri3 oligomers. This result is in a good agreement with diameters obtained from transmission electron micrographs (average maximum diameter ± SD: 15 ± 1 nm) and from dynamic light scattering data (26 ± 11 nm, polydispersity index 0.163) (Fig. 2e, Supplementary Fig. S1c). Rh Bri2 BRICHOS has an oligomer diameter of 14 ± 2 nm. Comparison of minimum and maximum Feret's diameters, indicates that rh Bri3 BRICHOS forms larger oligomers compared to rh Bri2 BRICHOS (Fig. 2f). The same tendency is also observed when comparing rh Bri3 BRICHOS and rh Bri2 BRICHOS by SEC; rh Bri3 BRICHOS shows a slight shift towards larger species and relatively more of it elutes in the void volume compared to rh Bri2 BRICHOS ( Supplementary Fig. S1b).
Circular dichroism (CD) spectroscopy of rh Bri3 BRICHOS shows that its overall secondary structure content is similar to that of rh Bri2 BRICHOS (Fig. 2c) and to isolated rh Bri2 BRICHOS oligomers 19,31 , but with somewhat increased residual molar ellipticity in the 205-225 nm regions. The CD spectra suggest that rh Bri3 BRICHOS is folded into a structure with a mixture of α-helices and β-sheet, which is compatible with the model structure ( Fig. 1).
To probe for the presence of exposed hydrophobic surfaces, the hydrophobic fluorescent dye bis-ANS 35 was used. Incubation of rh Bri3 BRICHOS with bis-ANS shows a considerable increase in emission intensity and a marked blue shift in emission maximum to 472 nm, compared to bis-ANS in buffer; very similar to the behaviour of rh Bri2 BRICHOS (Fig. 2d). Bis-ANS binding to proteins is dominated by hydrophobic interactions, and the spectral changes thus indicate that rh Bri3 BRICHOS exposes hydrophobic regions.
Effect of rh Bri3 BRICHOS on aggregation of partly denatured proteins. BRICHOS domains from different proproteins have different chaperone activities; rh proSP-C BRICHOS for example, is essentially unable to prevent non-fibrillar protein aggregation while rh Bri2 BRICHOS efficiently suppresses aggregation of partly denatured proteins including citrate synthase (CS) 21 . For Bri2 BRICHOS, the activity against nonfibrillar protein aggregation is almost entirely mediated by oligomers while dimers and monomers are essentially inactive 31 . Rh Bri3 BRICHOS efficiently suppress aggregation of partly heat-denatured CS in a concentration-dependent manner, with clear effects already at sub-stoichiometric ratios (Fig. 3a). Rh Bri3 BRICHOS mostly forms high molecular weight disulphide-linked oligomers (Fig. 2, Supplementary Fig. S1) and no difference in chaperone activity is seen between the unresolved crude form of Bri3 BRICHOS and the high molecular weight fractions isolated by SEC (data not shown). Bri3 BRICHOS is able to almost completely prevent the aggregation of CS at 1:1 molar ratio, while for Bri2 BRICHOS two-fold molar excess is needed for the same effect (Fig. 3b, Supplementary Fig. S2a). Likewise, rh Bri3 BRICHOS prevents aggregation of thermo-destabilized rhodanese already at substoichiometric amounts and is somewhat more efficient against non-fibrillar aggregation of also www.nature.com/scientificreports www.nature.com/scientificreports/ this model protein (Fig. 3c,d, Supplementary Fig. S2b). Rh BRICHOS concentrations were also measured with BCA and Bradford assays and the results were compared to spectrophotometric concentration determinations ( Supplementary Fig. S3). There results show, that independent of method used for protein concentration determination, rh Bri3 BRICHOS is somewhat more effective than rh Bri2 BRICHOS in preventing thermal aggregation of CS ( Supplementary Fig. S4).
In order to study potential complex formation between rh Bri3 BRICHOS domain and CS after thermal aggregation, samples were collected after completion of aggregation assay of a CS:BRICHOS mixture at 1:2 molar ratio, spun down and the supernatant as well as pellet fractions were analysed by SDS-PAGE and the supernatant was analysed by SEC. The results show that rh Bri3 BRICHOS forms a complex with thermo-destabilized CS that elutes in the void volume (Fig. 4b), which can explain how rh Bri3 BRICHOS keeps CS in a soluble state. In contrast, after heat treatment of CS alone essentially all protein is in the pellet and no protein is found in the soluble fraction (Fig. 4a). Freshly mixed samples gave separate peaks for CS and rh Bri3 BRICHOS upon SEC and non-heat-treated CS elutes on SEC and migrates on SDS-PAGE as an about 45 kDa protein, compatible with a monomer (Fig. 4c). Most CS is detected in complex with rh Bri3 BRICHOS (Fig. 4), while as shown in the earlier study, only a fraction of CS co-migrated with rh Bri2 BRICHOS despite that a larger excess of CS was used 21 .
The rh Bri2 and Bri3 BRICHOS domains seem to preserve their folded states with only minor changes occurring at the elevated temperatures used in the aggregation assays, as judged from their CD spectra and native-PAGE ( Supplementary Fig. S5). Overall, rh Bri3 BRICHOS is an efficient molecular chaperone that is able to suppress thermally induced amorphous aggregation of different substrates and can form a stable complex with thermally destabilized CS.
Effects of rh Bri3 BRICHOS on kinetics of Aβ42 fibril formation. Amyloid fibril formation can be suppressed by molecular chaperones via different mechanisms. Different BRICHOS domains influence distinct steps in the fibrillization pathway from monomeric Aβ42 to amyloid fibrils 21,30 . ProSP-C BRICHOS for example exclusively inhibits the secondary nucleation, thereby blocking the main source of toxic Aβ42 oligomers 30 , while Bri2 BRICHOS blocks both the secondary nucleation and the elongation of fibrils 21,30 .
To find out how rh Bri3 BRICHOS affects Aβ42 fibrillization in vitro, we first incubated 3 µM Aβ42 in the absence and in the presence of rh Bri3 BRICHOS from 1:0.1 to 1:1 Aβ42:BRICHOS molar ratios and the formation of fibrils www.nature.com/scientificreports www.nature.com/scientificreports/ was monitored by thioflavin T (ThT) fluorescence. Rh Bri3 BRICHOS shows a dose-dependent reduction of Aβ42 fibril formation already at substoichiometric concentrations (Fig. 5a) as also seen for rh Bri2 BRICHOS (Fig. 5,  Supplementary Fig. S6). The fibrillization half time, τ 1/2 , was increased linearly against the concentration of rh Bri3 BRICHOS (Fig. 5b), while the maximum rate of aggregation, r max , shows a mono-exponential dependence (Fig. 5c). The effects of rh Bri3 BRICHOS on τ 1/2 are smaller than what is seen for the rh Bri2 BRICHOS (Fig. 5b); both are smaller compared to an unresolved mixture of rh Bri2 BRICHOS assembly states 21 , and they are also smaller than the effects of isolated rh Bri2 BRICHOS oligomers on Aβ42 τ 1/2 31 . The γ-value, which describes the aggregation  Supplementary Fig. S6). (d) Aβ42 in the absence (black) and presence (red) of 3 µM rh Bri3 BRICHOS exhibits a similar dependence of the aggregation half time described by the γ-exponent. (e) Global analysis from the data sets in Fig. 6a,b revealed a dominant effect rh Bri3 BRICHOS on + k k 2 , related to secondary nucleation. (f) The dependencies of the relative combined rate constants obtained from the fits of rh Bri3 BRICHOS in panel (a) reveal a strong effect of rh Bri3 BRICHOS on secondary nucleation (k 2 k + ) but not on primary (k n k + ) pathways.

Scientific RepoRtS |
(2020) 10:9817 | https://doi.org/10.1038/s41598-020-66718-y www.nature.com/scientificreports www.nature.com/scientificreports/ mechanism of Aβ42 36 , is similar in the absence and presence of rh Bri3 BRICHOS (Fig. 5d), suggesting that aggregation follows mainly secondary pathways in both cases. Electron microscopy shows that the morphology of Aβ42 fibrils formed in the absence or presence of rh Bri3 BRICHOS (Supplementary Fig. S7) is similar. www.nature.com/scientificreports www.nature.com/scientificreports/ To identify microscopic processes in the Aβ42 fibrillization, we performed kinetic experiments at different Aβ42 concentrations with 3 µM rh Bri3 BRICHOS and fitted the results with a nucleation model, where + k k n (for primary pathways) and + k k 2 (for secondary pathway) are constrained globally (Fig. 6a,b). The kinetic fitting showed that the parameter + k k n is not much affected, whereas + k k 2 is markedly reduced by rh Bri3 BRICHOS (Fig. 5e). We also performed the kinetic analysis with the dataset for constant Aβ42 concentration and different rh Bri3 BRICHOS concentrations, which allows to elucidate the quantitative effects, by leaving both + k k n and + k k 2 free. This likewise showed that the + k k 2 was changed significantly while k k n + did not (Fig. 5f), suggesting again that rh Bri3 BRICHOS mainly affects the secondary pathway during Aβ42 fibrillization.
To study how the individual microscopic rate constants (k n , k + or k 2 ) for Aβ42 fibrillization are affected by rh Bri3 BRICHOS, global kinetic analysis was performed on the dataset for constant Aβ42 concentration and different rh Bri3 BRICHOS concentrations, leaving only one fitting rate, k 2 , k + , or k n , free (Fig. 6c-e). Both k 2 and k + but not k n describe well the time evolution of Aβ42 fibril formation in the presence of rh Bri3 BRICHOS, with 2 of 0.68, 2.79 and 13.69, respectively. This suggests that rh Bri3 BRICHOS mainly inhibits secondary nucleation as well as the elongation process during Aβ42 fibrillization, which is similar as for rh Bri2 BRICHOS 21,30 .

Functions of Bri3 and Bri2 and their BRICHOS domains. Bri2 and its BRICHOS domain have been
proposed to play crucial roles in various aspects of Alzheimer disease 37 . Bri3 is homologous to Bri2 with 44% overall sequence identities 27 , which suggest that the two proteins have similar functions. Both Bri2 and Bri3 are transported through the secretory pathway and are present in the plasma membrane, but they appear to be processed partly different 28 . Bri2 goes through a C-terminal cleavage by furin or furin-like convertases in the Golgi apparatus, resulting in a small secreted peptide and mature membrane-bound Bri2 38 , which then undergoes shedding of its ectodomain by ADAM10 39 , thereby releasing a soluble extracellular Bri2 BRICHOS domain. Bri3 is cleaved by furin 38 but fails to undergo shedding of the BRICHOS domain in transfected HEK293 cells 28 . It appears form the present study that the Bri2 and Bri3 BRICHOS domains are quite similar in structures and functions, while the entire Bri2 and Bri3 proteins, and particularly their expression and processing seem more likely to generate potentially different physiological functions.
Both Bri2 and Bri3 interact with the Aβ precursor protein (APP) and regulate its processing and thereby have an effect on Aβ production [40][41][42][43] . Furthermore, it has been shown that the furin cleaved mature Bri2 is localised in the plasma membrane and in endocytic vesicles and there interacts with APP 41,43 . Bri3 also binds to APP 40 and similar to Bri2, interferes with APP processing by specifically binding to mature APP rather than having a direct effect on secretase activities. For both Bri3 42 and Bri2 44 , the region that binds APP is located in the juxtamembrane linker region and APP binding is thus not directly dependent on the presence of the BRICHOS domain.
A recent study demonstrated that different assembly states of rh Bri2 BRICHOS domain have distinct activities: monomers most efficiently inhibit Aβ42 neurotoxicity, dimers are best in reducing fibril formation, and oligomers are by far most efficient against nonfibrillar aggregation 31 . The present study shows that also rh Bri3 BRICHOS has potent molecular chaperone activities. Bri3 BRICHOS forms mainly high molecular weight disulphide linked assemblies and only minor amounts of smaller species like dimers and monomers are observed (Fig. 2). The observed differences between rh Bri3 and Bri2 BRICHOS chaperone profiles shown here (Figs. 3 and 5) and references 21,31 can thus be rationalized by the higher tendency of Bri3 BRICHOS to form larger assemblies compared to Bri2 BRICHOS (Fig. 2f, Supplementary Fig. S1). For rh proSP-C BRICHOS, it has been shown that trimer is a relatively inactive storage form while the monomer is more efficient in delaying Aβ42 aggregation 45 . The lower potency of rh Bri3 BRICHOS against Aβ42 fibril formation can be rationalized along the same line, i.e. that rh Bri3 BRICHOS forms less monomers and dimers than rh Bri2 BRICHOS.
The two BRICHOS domains share about 60% identical residues. The face A of Bri2 and Bri3 BRICHOS is well conserved, it harbours no residue exchanges that result in significant changes in physicochemical properties, while for the face B there are abundant residue replacements that change the properties of the side-chains (Fig. 1). The overall net charge of face B however is essentially unchanged. We speculate that the divergent charge distribution of face B of Bri2 versus Bri3 BRICHOS might influence their different tendencies to form larger oligomeric species, which in turn affects activity against non-fibrillar as well as fibrillar protein aggregation. Our present data suggests that rh Bri3 BRICHOS distribution towards larger oligomers makes it a more efficient chaperone against non-fibrillar aggregation. However, additional factors, like e.g. hydrophobicity of exposed residues and their distribution in the Bri2 and Bri3 structures likely also influence their capacities as molecular chaperones. Further studies, e.g. using site-and motif-specific mutants are warranted to understand the mechanisms that underlie the regulation of assembly states and how they affect BRICHOS chaperone activity.

Material and Methods
Bioinformatic analysis. Tertiary structure predictions of Bri2 BRICHOS (113-231) and Bri3 BRICHOS (112-230) were performed using the I-TASSER web server 33 . Structures with the highest C-score were visualized and aligned using the software UCSF Chimera 46 . The confidence scores (C-score) of the predicted models were: −1.37 for Bri2 BRICHOS and −1.42 for Bri3 BRICHOS. The I-Tasser database does not predict the formation of disulphide bonds. Therefore, we measured the distance between the sulphur groups of the two only cysteine residues and assumed that a disulphide bond is formed when the distance is below 2.3 Å. The measured distance between the sulphur groups of the Bri2 BRICHOS model was 2.019 Å, and that for the Bri3 BRICHOS model was 2.016 Å.
Purification of rh Bri3 and Bri2 BRICHOS was performed essentially as described by Peng et al. 19 for Bri2 BRICHOS. Briefly, cells were lysed by sonication with Sonopuls HD2070 homogeniser (Bandelin electronic, Germany) for 3 min (1 sec on, 1 sec off for 1 min, then left to stand for 30 sec and repeated twice) and centrifuged at 6000 × g for 20 min at 4 °C. Pellets were suspended in 2 M urea in 0.15 M NaCl, 20 mM PB, pH 7.4, and sonicated as described above for 5 min and centrifuged at 15557 × g for 30 min at 4 °C. The collected supernatant was filtered through a 5 μm filter and frozen at −20 °C o/n. Next day, the thawed supernatant was poured onto an IMAC Ni-Sepharose Absorbance spectra from 260 to 180 nm with a wavelength increment of 1 nm were recorded in 1 mm path length quartz cuvettes. The bandwidth was set to 1 nm. Spectra shown are averages of 5 consecutive scans. As controls, rh Bri3 and rh Bri2 BRICHOS CD spectra were also recorded at 25 °C, after 24 h of incubation at 37 °C and after 2 h at 45 °C using a J-1500 CD spectrometer (JASCO) with a PTC-517 Peltier thermostat cell holder. CD spectra were recorded in 1 mm path length quartz cuvettes using wavelength from 260 to 175 nm with an increment of 0.5 nm. The bandwidth was set to 1 nm.
Bis-ANS fluorescence. Bis-ANS fluorescence emission spectra were recorded in quartz glass cuvettes using a Fluorolog-3 spectrofluorometer (FL3-22, Horiba Jobin Yvon). Fluorescence was excited at 395 nm and emission spectra recorded from 420 nm to 600 nm. The sample temperature was adjusted to 25 °C using a Peltier sample cooler. 1 µM rh Bri3 BRICHOS or rh Bri2 BRICHOS together with 2 µM bis-ANS were prepared in 20 mM phosphate buffer, 0.2 mM EDTA, pH 8 and measured after preincubation for 10 min at 25 °C. www.nature.com/scientificreports www.nature.com/scientificreports/ molar ratio of BRICHOS in 160 µl for 18 h. The samples were centrifuged at 22000 × g for 1 h at 4 °C. The pellet was resuspended in 20 µl 1x Tris-buffered saline. After sample preparation, 2 μL aliquot was adsorbed onto glow-discharged continuous carbon-coated copper grids (400 mesh, Analytical Standards) for 1 min. The grids were subsequently blotted with a filter paper, washed with two drops of Milli-Q water, and negatively stained with one drop of uranyl acetate [2% (w/v)] for 45 sec before final blotting and air-drying. The samples were imaged using a JEOL JEM2100F field emission gun transmission electron microscope (JEOL, Japan) operating at 200 kV. Single micrographs of the sample were recorded on a Tietz 4k × 4k CCD camera, TVIPS (Tietz Video and Image Processing Systems, GmbH, Gauting, Germany) at magnification of 85,000 (pixel size 1.76) and 1.5-3 μm defocus was used and a total of 12 and 8 images were recorded for rh Bri2 and Bri3 BRICHOS, respectively.
All rh Bri2 and Bri3 BRICHOS oligomer images were imported to EMAN2 (version 2.3) for further processing. After importing and estimating defocus with e2evalimage.py single particles in different orientations were selected from the images using e2boxer.py in manual mode, resulting in 2385 particles for rh Bri2 BRICHOS and 2394 particles for rh Bri3 BRICHOS. For each image, the contrast transfer function (CTF) parameters were estimated on boxed out regions containing particles (288 × 288 pixels for Bri2 and 240 × 240 pixels for Bri3) using the e2ctf.auto.py program. A reference-free 2D classification was performed afterwards. The Feret's and minimal Feret's diameters of 30 obtained 2D oligomer classes for each rh Bri2 and Bri3 BRICHOS were measured by bordering the oligomers and manual thresholding using Fiji ImageJ (National Institutes of Health) software.
Analysis of chaperone activity using citrate synthase and rhodanese. Thermal aggregation of citrate synthase (CS) from porcine heart (Sigma-Aldrich, Germany) was monitored by following the light scattering at 360 nm 48 . After buffer exchange to 40 mM HEPES/KOH pH 7.5 by PD-10 (GE Healthcare Life Sciences) chromatography CS was diluted to an end concentration of 600 nM. Rhodanese from bovine liver was purchased from Sigma (Merck; Darmstadt, Germany), dissolved in PBS, pH 7.4 and concentration was determined by measuring the absorbance at 280 nm and using an extinction coefficient of 60640 M −1 cm −1 . The final concentration of 3 µM rhodanese was used in experiments.
Thermal aggregation of substrate protein CS or rhodanese alone or in the presence of different concentrations rh Bri3 BRICHOS (determined with UV-Vis spectroscopy at 280 nm) was followed by measuring the change in the absorbance at 360 nm during 45 °C incubation for 2 h. Aggregation kinetics were measured in black, clear-bottom half-area polystyrene 96-well plates with non-binding surface (Corning Glass 3881, USA) in triplicates and using FLUOStar Galaxy plate reader (BMG Labtech, Offenberg, Germany). Samples were prepared in triplicates and all measurements were repeated 3 times. The same procedure was followed for the rh Bri2 BRICHOS. Rh Bri3 BRICHOS and rh Bri2 BRICHOS protein concentrations were also quantified by BCA protein assay (PierceTM, Thermo Scientific, USA) and Bradford based assay (Bio-Rad Protein Assay) following the manufacturer instructions. Different dilutions of rh BRICHOS (stock concentrations determined by UV-Vis spectrophotometric measurement) were compared ( Supplementary Fig. S3). BCA and Bradford assay corrected BRICHOS protein concentrations were then plotted against the normalised final intensities in both non-fibrillar protein aggregation assays ( Supplementary Fig. S4).
To study potential complex formation between rh Bri3 BRICHOS domain and CS after thermal aggregation, samples of CS:BRICHOS, 1:2 molar ratio, and also CS and rh Bri3 BRICHOS alone were collected after 2 h incubation at 45 °C, centrifuged for 5 min at 20,000 x g; supernatant and pellet fractions were analysed by SDS-PAGE under reducing conditions. Soluble fractions of incubated and comparable fresh samples were analysed by SEC and individual collected fractions were analysed by reducing SDS-PAGE.
Aβ42 aggregation kinetics by thioflavin T assay and kinetic analysis. Aggregation kinetics under quiescent conditions were studied by recording the ThT fluorescence intensity as a function of time in a FLUOStar Galaxy plate reader (BMG Labtech, Germany). The fluorescence was measured in black, clear-bottom half-area polystyrene 384-well microplates with non-binding surface (Corning Lifesciences, USA), at 37 °C, using bottom optics with 440 nm excitation filter and 480 nm emission filter. For amyloid fibril formation kinetics of monomeric Aβ42 ThT analysis was carried out. Samples were prepared and quickly mixed on ice and pipetted into 384-well microplates in 4 replicates. Measurements were carried out in volume of 20 µl with each well containing 3 μM Aβ42 monomer, rh Bri3 BRICHOS or rh Bri2 BRICHOS at different concentrations relative to Aβ42, 10 μM ThT in 0.02% NaN 3 , 0.2 mM EDTA and 20 mM PB pH 8.0. Also, the aggregation kinetics of Aβ42 monomers with different concentrations from 1.5, 2.0, 2.5, 3.0, 4.0, 5.0, 7.0 to 9 µM in the presence of a constant concentration (3 µM) of rh Bri3 BRICHOS species were measured in the same manner.
First, the time evolution of fibril formation of 3 µM Aβ42 with different concentrations of Bri3 BRICHOS between 0 to 100% were fitted to a sigmoidal equation , from which half time t 1/2 and maximal growth rate r max were extracted 31 . Second, the fibrillization kinetics was analysed by the nucleation model that is determined by primary (k n ) and secondary nucleation (k 2 ) events as well as for elongation (k + ) 36 . Kinetic profiles at different Aβ42 monomer concentration were fitted globally, where the fitting parameters λ (for primary nucleation) and κ (for secondary nucleation) are dependent on nucleation rates by λ = ⋅ ⋅ + k k m 2 ( 0) The parameters n C and n 2 are the reaction orders for primary and secondary nucleation, respectively. In this model, the time dependence of the fibril mass M(t) is given by Eq. (1), where the additional coefficients are functions of λ and κ 36 . We also performed a global fit of the experimental data set at constant Aβ42 concentration and different Bri3 BRICHOS concentration, in which the fit was constrained such that one fitting parameter was held to a constant value across all Bri3 BRICHOS concentration, while the second parameter was the only free parameter. This fitting constraint results in a sole fitting rate constant i.e. k n , k + or k 2 31 . Size-exclusion chromatography of Bri3 BRICHOS. Size-exclusion chromatography (SEC) was conducted using a Superose 6 10/300 GL column with ÄKTA pure 25 FPLC system (GE Healthcare Life Sciences, UK). Rh Bri3 BRICHOS samples were run at a flow rate of 0.5 ml·min −1 using 20 mM PB, 0.2 mM EDTA at pH 8.0 or 20 mM ammonium acetate buffer, pH 8.0 and the eluates were monitored at 280 nm. Calibration curves from LMW and HMW Gel Filtration Calibration Kits (GE Healthcare, UK) were used for the determination of the hydrodynamic radius. Bri3 hydrodynamic diameter was derived from the Stokes radius obtained by comparison to a standard curve derived from elution volumes of standard proteins with known Stokes radii. For analyses of CS and rh Bri3 BRICHOS complex formation, 500 μL of each sample were eluted in 20 mM ammonium acetate buffer, pH 8.0 using flow rate of 0.7 ml·min −1 .

Data availability
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.