Molecular chaperone ability to inhibit amyloid-derived neurotoxicity, but not amorphous protein aggregation, depends on a conserved pH-sensitive Asp residue

Proteins can self-assemble into amyloid fibrils or amorphous aggregates and thereby cause disease. Molecular chaperones can prevent both these types of protein aggregation, but the respective mechanisms are not fully understood. The BRICHOS domain constitutes a disease-associated small heat shock protein-like chaperone family, with activities against both amyloid toxicity and amorphous protein aggregation. Here, we show that the activity of two BRICHOS domain families against Alzheimer’s disease associated amyloid-β neurotoxicity to mouse hippocampi in vitro depends on a conserved aspartate residue, while the ability to suppress amorphous protein aggregation is unchanged by Asp to Asn mutations. The conserved Asp in its ionized state promotes structural flexibility of the BRICHOS domain and has a pKa value between pH 6.0–7.0, suggesting that chaperone effects against amyloid toxicity can be affected by physiological pH variations. Finally, the Asp is evolutionarily highly conserved in >3000 analysed BRICHOS domains but is replaced by Asn in some BRICHOS families and animal species, indicating independent evolution of molecular chaperone activities against amyloid fibril formation and non-fibrillar amorphous protein aggregation.


Introduction
This suggests that at pH between 7.0 and 6.0 Asp148 can be protonated with subsequent shift 221 of the dimer-monomer equilibrium towards dimers. To corroborate the supposition that 222 Asp148 titrates between pH 6.0 and 7.0 we turned to a rh Bri2 BRICHOS variant with 223 Thr206 replaced by Trp (T206W) that displayed an identical oligomerization profile as the rh 224 wildtype Bri2 BRICHOS (Supplementary Fig. 9a), but a different Trp fluorescence profile 225 in the monomeric state compared to the dimeric state (Supplementary Fig. 9b). In 226 agreement with SEC data, titration of the rh Bri2 BRICHOS Trp mutant showed a 227 fluorescence evolution (monomer/dimer transition) with a pKa of 6.7 (Fig. 2i, 228 Supplementary Fig. 9c). 229 230 These observations indicate that the D148N mutation in rh Bri2 BRICHOS promotes 231 monomer to dimer conversion, and mimics pH-induced dimerization of the wildtype 232 monomers with an apparent pKa of 6.7, but it does not significantly change the assembly of 233 dimers into larger oligomers. Analogously, the mutation D105N in rh proSP-C BRICHOS 234 results in a more compact conformation and possibly more complete assembly into trimers, 235 and the wildtype rh proSP-C BRICHOS conformational shift by CD titrates with an apparent 236 pKa of 6.1. 237

238
The conserved Asp of BRICHOS is essential for the capacity to prevent Aβ42-induced 239 neurotoxicity 240 To investigate the importance for the ability to alleviate amyloid-induced neurotoxicity, we 241 tested the efficacies of Asp to Asn mutated vs. wildtype BRICHOS in preventing A42-242 induced reduction of  oscillations in mouse hippocampal slices.  oscillations correlate with 243 learning, memory, cognition and other higher processes in the brain(41, 42) and progressive 244 cognitive decline observed in AD goes hand-in-hand with a progressive decrease of  245 oscillations (43)(44)(45). The BRICHOS domains from human proSP-C and Bri2 can efficiently 246 prevent Aβ42-induced decline in hippocampal  oscillations (27,30,32,33,46). In addition, 247 rh Bri2 BRICHOS rescues  oscillations and neuronal network dynamics from A42-induced 248 impairment at hippocampal CA3 area ex vivo (47). We recorded  oscillations in hippocampal 249 slices from wildtype C57BL/6 mice preincubated for 15 min either with 50 nmol L -1 Aβ42 250 alone, or co-incubated with 100 nmol L -1 rh wildtype or D105N proSP-C BRICHOS (Fig. 251 3a).  oscillations were elicited by application of 100 nmol L -1 kainic acid (KA) and allowed 252 to stabilize for 30 min prior to any recordings. As previously observed (27,30,46), 100 nmol 253 L -1 wildtype proSP-C BRICHOS prevented Aβ42-induced degradation of  oscillations, 254 which were remained at control levels (Fig. 3b and c, control: 1.7 ± 0.16 × 10 -8 V 2 Hz -1 , n= 255 20; Aβ42: 0.34 ± 0.06 × 10 -8 V 2 Hz -1 , n= 14, p< 0.0001 vs. control; + rh wildtype proSP-C 256 BRICHOS: 1.41 ± 0.24 × 10 -8 V 2 Hz -1 , n= 9) (Supplementary Table 2, Figure 3-source 257 data). By contrast, rh proSP-C BRICHOS D105N showed a complete loss of the preventative 258 efficacy at the same concentration (Fig. 3b and c, rh proSP-C BRICHOS D105N: 0.36 ± 259 0.09 × 10 -08 V 2 Hz -1 , n= 9, p= 0.0202 vs. rh wildtype proSP-C BRICHOS, p= 0.0002 vs. 260 control, p> 0.9999 vs. Aβ42) (Supplementary Table 2, Figure 3-source data). We have 261 previously observed that the rh Bri2 BRICHOS monomer is most efficient at preventing 262 Aβ42 induced degradation of  oscillations(32, 33). We therefore tested whether the D148N 263 mutation could affect the efficacy of rh Bri2 BRICHOS monomers. Rh Bri2 BRICHOS 264 D148N monomers (50 nmol L -1 ) showed reduced tendency of potency to prevent  induced neurotoxicity compared to wildtype monomers (50 nmol L -1 ), but did not completely 266 lose preventative efficacy, as observed for rh proSP-C BRICHOS D105N (Fig. 3d and e, rh 267 wildtype Bri2 BRICHOS monomers: 1.35 ± 0.29 × 10 -8 V 2 Hz -1 , n= 8, p> 0.9999 vs. control; 268 rh Bri2-BRICHOS D148N monomers: 0.7 ± 0.14 × 10 -8 V 2 Hz -1 , n= 11, p= 0. The rh wildtype Bri2 BRICHOS oligomers are efficient molecular chaperones against non-277 fibrillar protein aggregation, evaluated by thermo-induced citrate synthase aggregation as a 278 model(32). This was found to be true also for the rh Bri2 BRICHOS D148N oligomers (Fig. 279 5a and b). Similarly, the D105N mutation did not show any effects on the capacity of rh 280 proSP-C BRICHOS trimer against amorphous protein aggregation, i.e., both the wildtype and 281 D105N mutant are essentially inactive as regards canonical chaperone activity 282 ( Supplementary Fig. 10a). This shows a strong difference in the importance of the 283 conserved Asp in BRICHOS for activity against amyloid toxicity compared to activity 284 against amorphous protein aggregation. 285 286 Although the D148N mutation changed the balance between monomers and dimers of rh Bri2 287 BRICHOS, the mutant still formed large oligomeric species with a similar secondary 288 structure as wildtype oligomers (Supplementary Fig. 5e, g and i). We isolated rh Bri2 289 BRICHOS D148N oligomers by SEC, recorded transmission electron microscopy (TEM) 290 micrographs and calculated 3D reconstructions ( Fig. 4c and d, Supplementary Fig. 10b and 291 c, Figure 4-source data). The micrographs and 2D class averages revealed mostly 292 homogenous large assemblies with two-fold symmetry (Fig. 4c). As a result, an EM map was 293 reconstructed with D2 symmetry from 10 223 manually extracted single particles (Fig. 4d, 294 To further explore whether the diminished BRICHOS capacity against amyloid neurotoxicity 303 caused by Asp to Asn mutation correlates with the activity of suppressing macroscopic 304 amyloid fibril formation, we used thioflavin T (ThT)(48) fluorescence to monitor the kinetics 305 of Aβ42 fibril formation in the absence and presence of different concentrations of rh Bri2 or 306 proSP-C BRICHOS (Fig. 5). Both the rh wildtype proSP-C BRICHOS and the D105N 307 mutant showed dose-dependent progressive reduction of Aβ42 fibril formation at 308 substoichiometric concentrations, and the aggregation kinetics follows a typical sigmoidal 309 behavior ( Fig. 5a and d-f, Supplementary Fig. 11a-c). The Aβ42 fibrillization half time, 310 τ 1/2 , increases with increasing rh proSP-C BRICHOS concentration (Fig. 5b), while the 311 maximum rate of Aβ42 aggregation, r max , shows a mono-exponential decline (Fig. 5c). 312 Interestingly, rh proSP-C BRICHOS D105N showed remarkably improved inhibition on 313 Aβ42 fibril formation compared to the wildtype, manifested with both τ 1/2 and r max (Fig. 5a-314 c), which is opposite compared to the effects on the capacity to prevent Aβ42-induced 315 neurotoxicity (Fig. 3). On the other hand, similarly, the rh wildtype Bri2 BRICHOS 316 monomers showed similar effects on τ 1/2 and r max of A42 fibrillization as reported  i)(32). The rh Bri2 BRICHOS D148N oligomers, dimers and monomers also presented dose-318 dependent inhibition on A42 fibril formation with typical sigmoidal behavior ( Fig. 5g-i, 319 Supplementary Fig. 12). Along the same lines, compared to wildtype monomers, the D148N 320 monomers were significantly more efficient in inhibiting A42 fibril formation (Fig. 5g-i), 321 suggesting the Asp to Asn mutation pronouncedly enhances BRICHOS activity against 322 overall fibril formation. The D148N monomers and dimers showed very similar inhibitory 323 effects on A42 fibrillization (Supplementary Fig. 12a and b), which likely can be 324 explained by that D148N monomers and dimers are in equilibrium (Supplementary Fig. 6b  325   and 8). 326 327 To find out the molecular mechanisms underlying how the Asp to Asn mutation decreases 328 BRICHOS capacity against Aβ42 neurotoxicity but enhances its inhibitory activity against 329 overall Aβ42 fibril formation, we investigated the effects of both rh wildtype BRICHOS and 330 the corresponding Asp to Asn mutants on Aβ42 amyloid fibril formation and related 331 microscopic events. Aβ42 fibrillization kinetics are described by a set of microscopic rate 332 constants, i.e., primary (k n ) and surface-catalyzed secondary nucleation (k 2 ) as well as 333 elongation (k + )(49), and perturbations of the individual microscopic rates are most relevant 334 since their relative contributions determine the generation of nucleation units, which might be 335 linked to the neurotoxic Aβ42 oligomeric species (27,50). To evaluate the effects on 336 individual microscopic processes, we performed global fits of the kinetics data sets at 337 constant Aβ42 and different BRICHOS concentrations for both rh wildtype proSP-C and the 338 D105N mutant, where the fits were constrained such that only one single rate constant, i.e., k n , 339 k 2 or k + , is the sole fitting parameter ( Fig. 5d-f, Supplementary Fig. 11a-c). As previously 340 described(27), the rh wildtype proSP-C BRICHOS mainly interfered with the secondary 341 nucleation, indicated by the perfect fits when k 2 was the sole global fitting rate constant 342 ( Supplementary Fig. 11a-c). Also, the secondary nucleation rate k 2 as the sole fitting 343 parameter gave the best fits for the Aβ42 fibrillization kinetics with rh proSP-C BRICHOS 344 D105N (Fig. 5d-e), but with worse quality compared to the wildtype, especially for the start 345 of the aggregation traces. This suggests that a complex microscopic mechanism is present, 346 which might also include fibril-end elongation (k + ) and/or primary nucleation (k n ) in addition 347 to secondary nucleation (k 2 ). To further study whether the fibril-end elongation process or the 348 primary nucleation are affected in the presence of rh proSP-C BRICHOS D105N and 349 wildtype forms, we determined aggregation kinetics in the presence of a high initial fibril 350 seed concentration(49) and surface plasmon resonance (SPR) analysis. With seeding, the 351 fibrillization traces typically follow a concave aggregation behavior (Supplementary Fig.  352 11d and e), where the initial slope is directly proportional to the elongation rate k + . These 353 seeding experiments revealed that rh proSP-C BRICHOS D105N decreases the elongation in 354 a dose-dependent manner, and already at low concentrations fibril-end elongation is 355 noticeably retarded (Fig. 5j, Supplementary Fig. 11e). The rh wildtype proSP-C BRICHOS, 356 in contrast, showed only slight effects on fibril-end elongation (Fig. 5j, Supplementary Fig.  357 11d), which is further supported by immuno-transmission electron microscopy (immuno-EM) 358 images where both rh proSP-C BRICHOS D105N and the wildtype attach along the Aβ42 359 fibril surface, however, the fibril ends are apparently mainly blocked by the D105N mutant 360 (Fig. 5k). Additionally, SPR analyses of Aβ42 monomers immobilized on a sensor chip 361 indicated that rh wildtype proSP-C BRICHOS showed weak binding to the immobilized 362 Aβ42 monomers in line with previous reports (27,28). The D105N mutation significantly 363 enhanced the BRICHOS-Aβ42 monomer interactions ( Supplementary Fig. 11f) with an 364 apparent affinity value K D around 5 µmol L -1 (Fig. 5l) from global kinetics fits, and under 365 steady-state conditions a K D value of around 25 µmol L -1 was obtained ( Supplementary Fig.  366 11g). Interfering with primary nucleation (k n ) delays Aβ42 fibril formation without changing 367 the total number of oligomers generated, suppressing the secondary nucleation (k 2 ) efficiently 368 prevents the generation of oligomers, while blocking elongation (k + ) significantly increases 369 the number of oligomers formed(27). Of relevance for the results on A42 neurotoxicity of 370 wildtype vs D105N proSP-C BRICHOS (Fig. 3), we found that in the presence of equimolar 371 ratio of rh proSP-C BRICHOS D105N, the Aβ42 fibrillization reaction generated 372 approximately 5.2 times more oligomers than from Aβ42 alone, by significantly suppressing 373 the fibril end elongation process, while the wildtype only showed slight effects (Fig. 5j  To further study the mechanism underlying the reshaped interference of BRICHOS with 382 Aβ42 monomer and fibrils, we used the non-polar fluorescent dye bis-ANS to probe the 383 exposure of hydrophobic areas. Bis-ANS shows a blue shift of the emission maximum and 384 increased emission intensity upon binding to exposed hydrophobic protein surfaces, which 385 has been applied to rh wildtype Bri2 and proSP-C BRICHOS(32, 51, 52). When incubated 386 with bis-ANS, rh wildtype proSP-C BRICHOS gave an increase of emission intensity 387 compared to bis-ANS in buffer, and a blue shift of the emission maximum from about 520 388 nm to 495 nm ( Supplementary Fig. 11h). Notably, the blue shift and the intensity increase 389 for the D105N mutant at pH 8.0 are more pronounced compared to the wildtype protein 390 ( Supplementary Fig. 11h), showing that D105N results in more exposed hydrophobic areas, 391 which may be important for Aβ42 fibril-end binding(32, 33) and monomer binding. 392 Interestingly, when incubating bisANS with rh wildtype proSP-C BRICHOS at pH 6.0 an 393 identical fluorescence spectrum as for the D105N mutant at pH 8.0 was observed 394 ( Supplementary Fig. 11h), indicating that the D105N mutation gives similar effects as 395 protonating Asp105. For Bri2 BRICHOS, dimers were found previously to be most efficient 396 in preventing A42 overall fibril formation compared to the monomers and oligomers, while 397 the monomers are most potent in preventing A42 induced disruption of neuronal network 398 activity(32, 33). The microscopic mechanisms of rh Bri2 BRICHOS D148N species were 399 similar to the wildtype species, both the secondary nucleation and elongation of A42 were 400 affected ( Supplementary Fig. 12c-k). However, the equilibrium between monomer and 401 dimer was modified by the Asp to Asn mutant towards the dimer (Supplementary In this work, we show that the capacity of sHSP-like chaperone domain BRICHOS to inhibit 408 amyloid-associated neurotoxicity is dependent on a phylogenetically conserved Asp, whereas 409 the capacity to suppress non-fibrillar, amorphous protein aggregation is not affected by 410 mutating this Asp to Asn. Moreover, conformational changes occur as a result of Asp to Asn 411 mutations in both proSP-C and Bri2 BRICHOS, those changes can partly be mimicked by 412 lowered pH and the conserved Asp titrates with an apparent pKa of about 6.5 in both 413 BRICHOS domains studied. 414

415
The Asp residue studied herein is the only conserved non-Cys residue in all known 416 BRICHOS domains and two mutations of this residue in human proSP-C BRICHOS (D105) 417 are linked to ILD(16,20,36). Based on molecular dynamic simulation(16), monomeric 418 wildtype proSP-C BRICHOS and the D105N mutant behaved differently: only minor 419 conformational changes were seen in the mutant, but several large-scale changes occurred in 420 the wildtype at moderately elevated temperatures, which resulted in a more loosely folded 421 structure. We can rationalize our results against this background ( Fig. 6a and b). D105N 422 mutation in rh proSP-C BRICHOS indeed results in a more ordered conformation and 423 apparently more efficient trimer formation, while D148N mutation of rh Bri2 BRICHOS 424 results in more readily transition of monomers to more compact dimers ( Supplementary Fig.  425 5a and f). Similar effects as observed in the mutants were seen for the wildtype proteins 426 when pH was lowered to 6-7 ( Fig. 2h and i). This suggests that a negatively charged Asp 427 side-chain is necessary for maintaining a "loose" flexible state of the BRICHOS subunit and 428 that protonation, or replacement with a neutral Asn, results in a more "compact" state that is 429 prone to form oligomers. In the open conformation, BRICHOS is efficient in alleviating 430 Aβ42 amyloid neurotoxicity, while the compact conformation is more potent against overall 431 amyloid fibril formation but inefficient against amyloid induced neurotoxicity (Fig. 6b). 432 Interestingly, in brains of AD patients the pH is lower than that in the brains of healthy 433 individuals(53), and at low pH Aβ can form fibrils more efficiently than at neutral pH(54). BRICHOS domains, with <20% sequence identities, makes it likely that the common effects 444 now observed between them also apply to other BRICHOS domains that exhibit similar 445 evolutionary distances. Therefore, an elevated pKa value of the only strictly conserved non-446 Cys residue, as now found for proSP-C and Bri2 BRICHOS, is probably relevant for a 447 common function of all BRICHOS domains, which thus likely requires a maintained "loose" 448 conformation. Our results indicate that such a hypothetical common function is likely to be 449 related to prevention of amyloid toxicity rather than prevention of amorphous protein 450 aggregation. So far, molecular chaperones or chaperone-like domains have not been much 451 studied extensively as regards sensitivity to pH. One exception is clusterin, whose activity is 452 enhanced at mildly acidic pH, which appears to result from an increase in regions with 453 solvent-exposed hydrophobicity, but independent of any major changes in secondary or 454 tertiary structure(55). The phylogenetic analyses showed the presence of the BRICHOS 455 domain in a broad array of proproteins and species. In some cases where multiple BRICHOS 456 domains are found, the conserved Asp is preferentially replaced with Asn ( Fig. 1 and 2). 457 Considering the experimental results herein this may suggest that BRICHOS has adopted 458 different types of molecular chaperone activities during evolution. conserved Asp in response to pH changes. In the human brain, pH decreases with aging (56),  464 and also pH is significantly lower in patients than in heathy controls in different human brain 465 disorders (53, 57 Incomplete sequences were filtered out, resulting in total of 3 190 sequences. Identical 475 sequences were filtered by means of CD-HIT(59), with a threshold of 100% of sequence 476 identity, which gave 3 093 amino acid sequences. The hidden markov model profile (HMM) 477 of BRICHOS (PF04089) was extracted from PFAM database(60). The CD-HIT filtered 478 BRICHOS protein sequences were then scanned against the HMM profile using HMMER 479 software v3.3.2(61) with an E-value cut-off less than 1.0×10 -5 . An in-house python script was 480 written to filter out the significant BRICHOS proteins from the HMM result file and to 481 extract the BRICHOS domain sequences (each BRICHOS amino acid sequence was extended 482 six residues upstream from the BRICHOS domain starting position defined by PFAM). 483 Further, BRICHOS domain sequences less than 68 aa were removed as one rodent BRICHOS 484 domain is just 69 residues and still functional against amyloid fibril formation (62) and 508 verified by sequencing (Eurofins Genomics). As described(32, 33) , the Bri2 BRICHOS 509 variants were expressed in SHuffle T7 E. coli cells. Briefly, the cells were incubated at 30C 510 in LB medium with 15 µg mL -1 kanamycin until an OD 600 nm around 0.9. For overnight 511 expression, the temperature was lowered to 20C, and 0.5 mmol L -1 (final concentration) 512 isopropyl -D-1-thiogalactopyranoside (IPTG) was added. The induced cells were harvested 513 by centrifugation (4C, 7 000 ×g) and the cell pellets were resuspended in 20 mmol L -1 Tris 514 pH 8.0. After 5 min sonication (2 s on, 2 s off, 65% power) on ice, the lysate was centrifuged 515 (4C, 24 000 ×g) for 30 min and the protein of interest was isolated with a Ni-NTA column. 516 To remove the His 6 -NT* part, the target proteins were cleaved with thrombin (1:1 000, w/w) 517 at 4C  were grown at 37°C in LB medium containing 100 μg mL -1 ampicillin until an OD 600 nm 531 around 0.9. The temperature was lowered to 25°C and 0.5 mmol L -1 (final concentration) 532 IPTG was added for overnight expression. The cells were harvested by centrifugation at 7 533 000 × g for 20 min, and the cell pellets were resuspended in 20 mmol L -1 Tris pH 8.0. The 534 protein was purified using Ni-NTA column and ion exchange chromatography (QFF, GE 535 Healthcare). Thrombin (1:1 000, w/w) was used to remove the thioredoxin tag. The purified 536 rh proSP-C BRICHOS variants were analysed by Superdex 200 GL columns (GE Healthcare, 537 UK) using an ÄKTA system (GE Healthcare, UK). The BRICHOS mutants in this study 538 were expressed and purified in parallel with their wildtype counterparts. 539

NMR spectroscopy 541
For the NMR experiments, gene fragment encoding human proSP-C BRICHOS was 542 transformed into SHuffle T7 E. coli cells and was grown in LB with gradually increasing 543 D 2 O content (25%, 50%,75% and 100%). At 100% D 2 O, 1 mL LB was used to inoculate 100 544 mL M9 in 100% D 2 O enriched with 15 N H 4 Cl and 13 C glucose, and was grown over night at 545 31°C. After overnight incubation, the 100 mL was added to 900 mL M9 in D 2 O enriched 546 with 15 N and 13 C and grown at 30°C until OD 600 nm was around 0.8. The temperature was 547 lowered to 20°C and 0.5 mmol L -1 (final concentration) IPTG was added for overnight 548 expression. The purification was performed as described above. Rh Bri2 BRICHOS D148N oligomers after SEC isolation were immediately stored on ice 587 followed by grid preparation. Aliquots (4 µL) were adsorbed onto glow-discharged 588 continuous carbon-coated copper grids (400 mesh, Analytical Standards) for one min. The 589 grids were subsequently blotted with filter paper, washed with two drops of milli-Q water, 590 and negatively stained with one drop of 2% (w/v) uranyl acetate for 45 s before final blotting 591 and air-drying. The sample was imaged using a Jeol JEM2100F field emission gun 592 transmission electron microscope (Jeol, Japan) operating at 200 kV. Single micrographs for 593 evaluating the quality of the sample were recorded on a Tietz 4k × 4k CCD camera, TVIPS 594 (Tietz Video and Image Processing Systems, GmbH, Gauting, Germany) at magnification 595 of× 85 200 (1.76Å per pixel) and 1.0-2.8 µm defocus. A total of 16 micrographs were 596 recorded for single particle analysis. All 16 micrographs were imported to EMAN2 (version 597 2.3) for further processing(70). After importing and estimating defocus with e2evalimage.py, 598 single particles in different orientations were selected from the images using e2boxer.py in 599 manual mode (11 094 particles, after one more manual selection, 10 223 particles were left). 600 For each image, the contrast transfer function (CTF) parameters were estimated on boxed out 601 regions (containing particles, 168×168) using e2ctf.auto.py program. A reference-free 2D 602 classification based on the selected 10 223 phase-flipped particles (low-pass filtered to 20 Å) 603 was performed using e2refine2d.py. The 2D classes show an approximate 2-fold symmetry, 604 which is consistent with both the results of rh wildtype Bri2 BRICHOS oligomer and the 605 biochemical data. Generated 2D classes were used as the input for building the 3D initial 606 model using e2initialmodel.py. 3D refinement was performed in several rounds using 607 e2refine_easy.py applying D2 symmetry aiming at a final resolution of 15 Å. The first two 608 rounds of 3D refinements were performed with pixel size of 3.52 Å after binning the data by 609 a linear factor 2. In the last round of refinement, the data was resampled to 2.464 Å per pixel. spikes were excluded for global fits to reflect the binding affinity. Since the response signals 639 of the two lowest protein concentrations (i.e., 1.56 and 3.13 μmol L -1 ) used in kinetic analysis 640 were too weak, only sensorgrams obtained from rh proSP-C BRICHOS D105N mutant 641 ranging from 6.25 μmol L -1 to 100 μmol L -1 were included in the global fits. The dissociation 642 was globally fitted to a biexponential model as described by Eq. (1)(28, 73): 643 Eq. (1)  644 where R 1 is the response signal at the starting time for the dissociation phase t 1 , and x is 645 between 0 and 1. k d1 and k d2 are the dissociation rate constants for the fast and slow 646 dissociation phases, respectively. The global fitted value k d2 was used to calculate the 647 apparent K D value. 648 The association phase was fitted to Eq. (2)(28): 649 (2)  650 where R 0 and R f are the initial and final response signal of the association phase, respectively. 651 k obs is the observed rate constant described by Eq. (3)(28, 73): 652 where c is the protein concentration, k a is the association rate constant, and k d from this 654 analysis is related to secondary binding artifacts corresponding to k d1 . Linear regression was 655 used to determine k a . The apparent K D value was calculated as ratio of the dissociation rate 656 constant and association rate constant. Eq. (5) where M(t) is the total fibril mass concentration, and the intermediate coefficients are 695 functions of λ and κ, and n C and n 2 are the reaction orders for primary and secondary 696 nucleation, respectively: 697 ± = ± 2 /2/ 2 ∞ = √2 2 /( 2 ( 2 + 1)) + 2 2 / ̃∞ = √ ∞ 2 -4 + − 2 ± = ( ∞ ± ̃∞)/2/ = √2 ⋅ + ⋅ (0) The microscopic rate constants k n , k + , and k 2 are for primary nucleation, elongation, and 698 secondary nucleation, respectively. The kinetic data were globally fitted to Eq. (5), where the 699 fits were partially constrained with one fitting parameter held to a constant value, resulting in 700 that only one rate constant (k n , k + or k 2 ) is the sole fitting parameter(32, 33). To investigate 701 the generation of nucleation units, according to the nucleation rate r n (t)(27): 702 r n (t) = k n m(t) n c + k 2 M(t)m(t) n 2 , 703 the total number of nucleation units was calculated by integrating the nucleation rate r n (t) 704 over the reaction. 705 706

Immunogold staining of A42 fibrils and transmission electron microscopy analysis 707
Five µmol L -1 A42 monomer was incubated at 37°C with 100% rh wildtype proSP-C 708 BRICHOS and the D105N mutant overnight, and the fibrils were collected at 4°C by 709 centrifugation for 1 h at 22 000×g. The fibrils were gently resuspended in 20 µL 1×TBS, of 710 which 2 µL were applied to carbon coated copper grids, and incubated for about 5 min. Before sacrificed, all the mice were anaesthetized deeply using isoflurane. 726 727 The brain was dissected out and placed in modified ice-cold ACSF (artificial cerebrospinal 728 fluid). The ACSF contained 80 mmol L -1 NaCl, 24 mmol L -1 NaHCO 3 , 25 mmol L -1 glucose, 729 1.25 mmol L -1 NaH 2 PO 4 , 1 mmol L -1 ascorbic acid, 3 mmol L -1 NaPyruvate, 2.5 mmol L -1 730 KCl, 4 mmol L -1 MgCl 2 , 0.5 mmol L -1 CaCl 2 and 75 mmol L -1 sucrose. Horizontal sections 731 (350 µm thick) of the ventral hippocampi from both hemispheres were sliced with a Leica 732 VT1200S vibratome (Microsystems, Sweden). The sections were immediately transferred to 733 a submerged incubation chamber containing standard ACSF: 124 mmol L -1 NaCl, 30 mmol 734 L -1 NaHCO 3 , 10 mmol L -1 glucose, 1.25 mmol L -1 NaH 2 PO 4 , 3.5 mmol L -1 KCl, 1.5 mmol L -1 735 MgCl 2 and 1.5 mmol L -1 CaCl 2 . The chamber was held at 34°C for at least 20 min after 736 dissection and it was subsequently cooled to room temperature (~22°C) for a minimum of 40 737 min. Proteins (A42 and rh BRICHOS) were first added to the incubation solution for 15 min, 738 and then the slices were transferred to the interface-style recording chamber for extracellular 739 recordings. During the incubation, slices were supplied continuously with carbogen gas (5% 740 CO 2 , 95% O 2 ) bubbled into the ACSF. The electrophysiological data is presented as means ± standard errors of the means. Prior 767 statistical analysis all the data was subjected to outlier determination and removal with the 768 ROUT method (Q = 1%) followed by D'Agostino & Pearson omnibus normality test. Based 769 on the previous experience with the outliers and overall sample behaviour, sample size was 770 determined based on previous studies performed in an interface-type chamber (27, 32, 33, 47, 771 74, 75). Each experimental round was performed with parallel controls (Control KA and 772 Aβ42) from the same animal and randomized preparations (slices incubated with Aβ42 + rh 773 proSP-C BRICHOS D105N, + wildtype rh proSP-C BRICHOS, + rh Bri2-BRICHOS D148N 774 monomers or + wildtype rh Bri2-BRICHOS). For comparison purposes data from Control 775 KA and Aβ42 was pooled from interleaved slices recorded in these conditions. The number 776 of recorded slices per condition (at least from 3-5 mice) are shown in the corresponding 777   precursors account for 94% and 5%, and adopt "loose" and "compact" conformation, 1146 respectively. At neutral pH, the conserved Asp is ionized (Asp -) and BRICHOS presents an 1147 open structure (from ref (16) ). At pH 6-7, the Asp is protonated and BRICHOS is more 1148 compact (PDB accession number 2YAD, from crystals prepared at pH 6.5 (ref(16) ) and prone to 1149 the 2 019 BRICHOS domains, which was grouped into thirteen families (Fig. 1c)  1284 aggregation traces (cross) were constrained such that only one single rate constant is the free 1323 fitting parameter, indicated in each panel. 2 values describing the quality of the fits: 43 for 1324 k n free, 3.9 for k 2 free and 7.8 for k + free. (i-k) Aggregation kinetics of 3 µmol L -1 A42 in 1325 the presence of rh Bri2 BRICHOS D148N oligomers at different concentrations. The global 1326 fits (solid lines) of the aggregation traces (cross) were constrained such that only one single 1327 rate constant is the free fitting parameter. 2 values describing the quality of the fits: 8.6 for 1328 k n free, 1.4 for k 2 free and 0.4 for k + free. For the different D148N species, both k 2 and k + as 1329 sole free fitting rate, the fibrillization traces were described with similar 2 values, 1330 suggesting both k 2 and k + might be affected, like the wildtype species.