Protofibril–Fibril Interactions Inhibit Amyloid Fibril Assembly by Obstructing Secondary Nucleation

Abstract Amyloid‐β peptides (Aβ) assemble into both rigid amyloid fibrils and metastable oligomers termed AβO or protofibrils. In Alzheimer's disease, Aβ fibrils constitute the core of senile plaques, but Aβ protofibrils may represent the main toxic species. Aβ protofibrils accumulate at the exterior of senile plaques, yet the protofibril–fibril interplay is not well understood. Applying chemical kinetics and atomic force microscopy to the assembly of Aβ and lysozyme, protofibrils are observed to bind to the lateral surfaces of amyloid fibrils. When utilizing Aβ variants with different critical oligomer concentrations, the interaction inhibits the autocatalytic proliferation of amyloid fibrils by secondary nucleation on the fibril surface. Thus, metastable oligomers antagonize their replacement by amyloid fibrils both by competing for monomers and blocking secondary nucleation sites. The protofibril—fibril interaction governs their temporal evolution and potential to exert specific toxic activities.


Introduction
Amyloid fibrils are cross-b structured protein assemblies that represent the hallmark of many protein aggregation disorders. [1] Fors everal disease-related proteins,a myloid fibrils correspond to the thermodynamic minimum of the free energy landscape for folding and aggregation. [2] For example,A b amyloid fibrils are the core components of the senile plaques found in Alzheimersd isease (AD)-affected brains. [3] Ab fibrils are polymorphic,v ariably constructed from in-register parallel b-sheets. [4][5][6] They form by nucleated polymerization, where initial fibril nuclei grow by monomer addition to the fibril ends. [7] Af requent contributor to the typical sigmoidal growth profile of amyloid fibrils is fibrilmediated secondary nucleation. In this process,t he fibril surface acts as the preferential site for new fibril nucleation, leading to the autocatalytic proliferation of amyloid fibrils. [7] Asecond type of assemblies that Ab is prone to form are metastable globular oligomers with am olecular weight > 50 kD,a nd their associated curvilinear fibrils with typical lengths up to 200 nm. [8][9][10][11][12][13][14] These oligomers are collectively referred to as AbOo rp rotofibrils. [8,12,15] As these oligomers are formed in ar eaction distinct from fibril formation (i.e., off-pathway), [8,11,13,16] the term protofibril can be misleading. Similarly,t he term AbOi su sed interchangeably for onpathway oligomers.B elow we use the designations globular oligomer (gO) and curvilinear fibril (CF) to refer specifically to the off-pathway,m etastable assemblies.G O/CFs form in al ag-free oligomerization reaction with am uch higher reaction order than that observed for fibril formation. [11] Like amyloid fibrils,gO/CFs are rich in b-sheets,but their structure has not been resolved to the same level of detail yet. [17] GO/ CFs have been reported for several amyloidogenic proteins, suggesting that they are ageneral alternative assembly type of this class of proteins. [16,[18][19][20] Ab gO/CFs may represent the main toxic species in AD, as they are more effective than amyloid fibrils at inducing synaptic dysfunction, inhibiting long-term potentiation, triggering inflammation, and disrupting membranes. [8,13] Several receptors that mediate toxic signaling of extracellular Ab gO/ CFs have been identified. [21] In addition, intracellular Ab gO/ CFs show cytotoxic effects. [8] Ab gO/CFs are enriched in ahalo surrounding senile plaques,pointing to apotential role of gO/CF-fibril interactions. [22,23] Fore xample,f ibril plaques have been suggested to serve as ar eservoir,o rb uffer,o fA b oligomers. [22,23] However,g O/CF-fibril interactions have not been characterized in detail.
We have recently reported that the high concentration dependence of gO/CF formation results in at hreshold monomer concentration required for gO/CF formation, denoted critical oligomer concentration (COC), which is significantly higher than the threshold for fibril formation. [11,20] Above the COC,the assembly kinetics are biphasic, with an initial lag-free gO/CF formation phase,f ollowed by asigmoidal phase representing the nucleation and growth of fibrils which slowly replace the metastable gO/CFs.S urprisingly,w eo bserved that gO/CF formation above the COC progressively increased the lag period for subsequent fibril nucleation and growth, revealing that gO/CFs inhibit fibril formation not only by competing for monomers,but also in an active fashion. These observations were made with two distinct amyloid proteins,adimeric variant of Ab40 (dimAb) and hen egg-white lysozyme (hewL). [11] Here,w ei nvestigate how gO/CFs actively inhibit fibril formation. We first show that the inhibitory effects of offpathway gO/CF formation on subsequent fibril nucleation and growth are similarly present in the two dominant AD peptides Ab40 and Ab42. We then demonstrate for Ab as well as for hewL that gO/CFs bind to fibril surfaces.G O/CF binding also promotes fibril bundling, thereby further reducing fibril surface area. We finally take advantage of the Ab-dimAb system to show that the gO/CF-fibril interaction interferes with secondary nucleation and blocks the proliferation of amyloid fibrils.

Results and Discussion
To investigate gO/CF formation of Ab,wehave generated dimAb,adimeric Ab variant in which two Ab40 units are linked in one polypeptide chain through af lexible glycerinserine-rich linker. [11] Thec onformational properties of the Ab40 units in dimAb are the same as those of unlinked Ab40. [11] However,d ue to the increased local Ab concentration, gO/CF formation of dimAb is strongly promoted, which is reflected in the comparatively low COC of % 1.5 mM at neutral pH. [11] Above the COC,T hioflavin T( ThT) fluorescence indicates biphasic assembly kinetics of dimAb ( Figure 1A). During the first phase,gO/CFs form ( Figure 1C) in an oligomerization reaction with ah igh reaction order of % 3. [11] After alag-time,amyloid fibril formation is observed, in agreement with anucleation-polymerization reaction (Fig-ure 1A,C). [11] Upon prolonged incubation, the metastable gO/ CFs are slowly replaced by amyloid fibrils. [11] Above the COC, the lag-time of amyloid fibril formation develops an inverse dependence on protein concentration, i.e., the lag-time increases with protein concentration ( Figure 1B), indicating that gO/CFs actively interfere with amyloid fibril formation. [11] We tested if these observations,p reviously made for dimAb and hewL, are reproduced for Ab40 and Ab42. A logarithmic plot of the ThTtime course of Ab40 assembly at ac oncentration of 20 mMo rb elow shows as igmoidal curve with al ag-time of several hours.T his is in agreement with amyloid formation by an ucleation-polymerization reaction with prominent contributions from secondary nucleation ( Figure 1D). In contrast, for Ab40 concentrations of 40 mM or above,anadditional, lag-free kinetic phase occurred during which gO/CFs assembled ( Figure 1D,F). These gO/CFs were replaced by amyloid fibrils during as econd kinetic phase ( Figure 1D,F). Ab40 assembly thus follows the same pattern as dimAb assembly,a lbeit with an approximately 20-fold higher COC ( % 30 mM), which is expected considering the lack of ac ovalent connection between Ab monomers in unlinked Ab40. ThTk inetics recorded with Ab40 by the deGrado and Prusiner lab,f or concentrations at or above those used here,g enerated similar biphasic kinetics and produced long-lived Ab gOs. [24] As with dimAb and hewL, the lag-time of amyloid fibril formation of Ab40 started to increase above the COC ( Figure 1E). This indicates that Ab40 gO/CFs share the ability to interfere actively with fibril formation. ForA b42, the ThTt ime courses indicated at ransition to biphasic kinetics at ac oncentration between 10 and 30 mM( Figure 1G), in line with previous observations. [25] Theshort lag times of Ab42 amyloid fibril formation undermined our efforts of correlating biphasic ThTk inetics with the onset of gO/CF formation in that system. Never- theless,t he data for Ab40 and Ab42 show that the observations made for dimAb and hewL extend to the two prevalent Ab variants,w ith higher COCs of the unlinked peptides.
One possible mechanism by which gO/CFs might actively inhibit amyloid formation would be by interfering with secondary nucleation. GO/CFs might bind to amyloid fibril surfaces,where they could block the sites capable of catalyzing fibril nucleation. To test this hypothesis,w ef irst investigated if gO/CFs bind to amyloid fibril surfaces.F ibrils were formed from Ab40 at ac oncentration of 10 mM. Since this concentration is below the COC of Ab40, only fibrils but no gO/CFs were formed. Upon centrifugation, the fibrils were found in the pellet (Figure 2A,left). GO/CFs were formed by quiescently incubating dimAb at aconcentration of 10 mMfor 24 hours.Under these conditions dimAb assembled into gO/ CFs whereas amyloid fibrils were still absent. Theg O/CFs were collected from the supernatant after centrifugation (Figure 2A,m iddle). When Ab40 fibrils and dimAb gO/CFs were mixed and subsequently centrifuged, the pellet contained amyloid fibrils decorated with gO/CFs (Figure 2A, right). This indicates that the fibril surfaces have an affinity for gO/CFs,leading to co-precipitation of the two species.The experiment was repeated for hewL. HewL amyloid fibrils grown under sigmoidal (sub-COC) conditions ( Figure 2B, left) and hewL gO/CFs formed during the early phases of biphasic growth ( Figure 2B,middle) were mixed, resulting in binding of gO/CFs to the lateral surfaces of the fibrils ( Figure 2B,r ight). In addition, mixing of hewL gO/CFs with fibrils at growth temperatures dramatically increased lateral bundling and precipitation of fibrils,w hile isolated fibrils remained unchanged ( Figure 2C). Both binding and bundling reduce the fibril surface area available for secondary nucleation.
In order to isolate the consequences of this gO/CF and fibril interaction on fibril growth mechanisms we performed seeded fibril growth experiments with increasing gO/CF admixtures.T od os o, we took advantage of the different COCs for dimAb vs.Ab40:atlow mMconcentrations dimAb assembles into gO/CFs,w hereas Ab40 continues to exhibit the sigmoidal kinetics of nucleated-polymerizationw ith secondary nucleation. Furthermore,d imAb gO/CFs possess high kinetic stability and persist even for several hours after dilution to sub-COC concentrations,t hereby allowing to investigate effects of gO/CFs down to sub-mMc oncentrations. [26] Amyloid fibril formation is am ultistep reaction ( Figure 3G). [27] To test the effects of gO/CFs specifically on fibril elongation and secondary nucleation, we seeded Ab40 monomers with different concentrations of sonicated Ab40 fibrils in the presence of increasing concentrations of dimAb gO/CFs ( Figure 3A). When 10 %A b40 seeds were added to 2.5 mMA b40 monomers,f ibril elongation was the dominant reaction as evident from the immediate linear increase in ThT signal ( Figure 3B). Addition of 1.25 mMd imAb gO/CFs (corresponding to an Ab40 subunit concentration of 2.5 mM) did not have asubstantial effect, showing that gO/CFs do not actively interfere with amyloid fibril elongation ( Figure 3B). When al ower amount, that is,0 .1 %, of Ab40 seeds was applied, sigmoidal time traces were obtained, indicating the importance of autocatalytic amplification of amyloid fibrils by secondary nucleation (Figure 3C). In this case,a ddition of dimAb gO/CFs led to aconcentration-dependent increase in lag-time ( Figure 3C). Since primary nucleation does not contribute to the ThTs ignal on this time scale at this Ab40 monomer concentration ( Figure 1D)a nd fibril elongation is not affected by gO/CFs ( Figure 3B), we conclude that gO/ CFs inhibit secondary nucleation. Thei nhibitory effect was already discernible at ac oncentration of 60 nM gO/CFs, which corresponds to ag O/CF:monomer ratio of 1:20 in numbers of Ab40 units.S uch as ubstoichiometric effect is compatible with inhibition of an autocatalytic process.T o confirm that inhibition of Ab40 fibril formation is in fact caused by gO/CFs and not due to any other activity of dimAb on Ab40, we compared the effects of i) dimAb gO/CFs Figure 2. GO/CFsb ind to amyloid fibril surfaces. AFM images of assemblies of A) dimAb and Ab40 or B),C) hewL. A) Amyloid fibrils formed from 10 mMA b40 were found in the pellet upon centrifugation at 14 000 g(left);g O/CFsformed from 10 mMd imAb remained in the supernatant (middle).U pon mixing equimolar amounts, dimAb gO/ CFsc o-precipitated with Ab40 fibrils and decorated fibril surfaces (right). B) Amyloid fibrils and gO/CFsformed from 1.75 mM hewL were grown below (50 mM NaCl) or above (250 mM NaCl) the COC, respectively.After isolation and adjusting NaCl for both to 450 mM,100 mMo ffibrils were mixed with 1mMofgO/CFsatroom temperature and in 450 mM NaCl. C) Mixing hewL gO/CFsand fibrils at growth temperature (52 8 8C), instead, induced rapid fibril bundling and precipitation while, under the same conditions, fibrils themselves remained unchanged. prepared above the COC and diluted to as ub-COC concentration of 0.3 mMwith those of ii)dimAb monomers that were freshly eluted from size exclusion chromatography and kept at as ub-COC concentration of 0.3 mM. Thed imAb preparation that contained gO/CFs due to incubation above the COC exhibited amuch stronger effect on fibril formation than the one kept below the COC ( Figure 3D). Theinhibition is not an unspecific effect of any polypeptide assembly in the size range of gO/CFs,asit is not observed for ferritin, a24-mer of helical bundles with am olecular weight of 440 kD ( Figure S1).
To further confirm that the kinetics data are in agreement with inhibition of secondary nucleation, we computed global fits to the gO/CF concentration-dependent data for two different models of fibril formation using the software package Amylofit. [27] First, we applied an ucleation-elongation model and performed global fits that attributed the effects of gO/CFs to an altered rate constant of either primary nucleation or fibril elongation (all parameters were shared among the data sets apart from the rate constants of primary nucleation or fibril elongation, respectively). These fits showed clear deviations from the experimental data (Figur- es 3E and S2A,B). Second, we applied as econdary nucleation-elongation model and performed global fits that attributed the effects of GO/CFs to altered rate constants of either primary nucleation, secondary nucleation, or fibril elongation (again, keeping all other fitting parameters the same among the data sets). Theglobal fit to this model using avariable rate constant of primary nucleation did not reproduce the Figure 3. GO/CFsi nhibit secondaryn ucleation of amyloid fibrils. A) Scheme of the kinetics assays. The effects of dimAb gO/CFsons econdary nucleation and elongation of Ab40 amyloid fibrils were probed. B) Elongation of Ab40 fibril seeds by Ab40 monomers in the absence and presence of dimAb gO/CFs. C) Secondary nucleation-elongation of Ab40 fibril seeds by Ab40 monomers in the absence and presence of dimAb gO/CFs. D) Secondary nucleation-elongation of Ab40 fibril seeds by Ab40 monomers in the absence (grey) and presence of dimAb gO/CFs formed above the COC and diluted below the COC (orange) or dimAb monomers below the COC (blue). E) Global fits to the data using an ucleation-elongation model. All parameters were shared apart from the elongation rate constants. F) Global fits to the data using asecondary nucleation-elongation model. All parameters were shared apart from the secondary nucleation rate constant. G) Nucleation-growth model including binding of gO/CFstoamyloid fibril surfaces, which inhibits secondary nucleation. P, fibril particle concentration;M,f ibril mass concentration;m ,monomer concentration; n c ,n ucleus size; k n ,p rimary nucleation rate constant; k 2 ,s econdary nucleation rate constant; k + , elongation rate constant; K D ,affinity of gO/CF for the fibril surface. H), I) Numericals imulationsa pplying the model outlined in G), using the rate constantsobtained for the nucleation-elongation model in F) (uninhibited trace) and a K D of 160 nM. Duplicate or triplicate measurements per condition are shown in panels (C), (E), (F), (H), (I).
decreasing slope during the exponential growth phase with increasing gO/CF concentration ( Figure S2C). In contrast, when the rate constants of secondary nucleation or fibril elongation were variable,g ood agreement with the data was obtained ( Figures 3F and S2D,E). These fits do not differentiate between effects on secondary nucleation and fibril elongation, as the rate constant of secondary nucleation occurs in the regression equation only in the form of its product with the rate constant of fibril elongation. [28] However,a sw ec an exclude any substantial effect of gO/CF on fibril elongation (Figure 3B), the global fits further strengthen the case for gO/CFs inhibiting amyloid fibril formation through an effect on secondary nucleation. As gO/CFs bind to amyloid fibril surfaces,t hey likely inhibit secondary nucleation by blocking the sites capable of catalyzing secondary nucleation ( Figure 3G). This mode of inhibition of Ab fibril formation has previously been described for the BRICHOS chaperone. [29] Ther eduction in the number of active sites effectively corresponds to ar eduction in the fibril surface available for autocatalytic amplification rather than to ad ecrease in the secondary nucleation rate constant. We extended the nucleation-polymerizationm odel by including an equilibrium of gO/CF binding to fibrils that reduces the fibril mass engaged in secondary nucleation ( Figure 3G). Numerical simulations with the modified model were performed, using the rate constants obtained by Amylofit for the uninhibited case of nucleation-polymerization with variable secondary nucleation (black fit in Figure 3F). In particular, the same secondary nucleation rate constant was used for all gO/CF concentrations,a ttributing the gO/CF concentration dependence of the kinetics solely to changes in the fibril mass available for secondary nucleation according to the gO/ CF:fibril interaction equilibrium. Theg O/CF:fibril interaction was treated as a1 :1 interaction in the number of Ab subunits.W hena pplying ad issociation constant of K D = 160 nM the numerical simulations yielded good agreement with data obtained both at 2.5 mMand 5 mMAb40 monomer concentration ( Figure 3H,I).

Conclusion
We previously observed ar emarkable inversion of the scaling relation between increasing protein concentration and decreasing lag-times for dimAb and hewL amyloid fibril formation upon crossing the COC. [11] Here,w er eproduced the surprising increase in lag-time with increasing protein concentration for Ab40, which indicates that gO/CFs actively inhibit fibril formation ( Figure 1E). Collectively,t he AFM data ( Figure 2) and chemical kinetics data ( Figure 3) provide strong evidence that gO/CFs inhibit Ab amyloid fibril formation by binding to amyloid fibril surfaces,b locking the sites that would otherwise promote secondary nucleation. The same mode of inhibition was observed for the BRICHOS chaperone,b ut not for as et of control proteins. [29] This suggests that this inhibitory activity is rather specific.Itisalso in line with the relatively high affinity of the gO/CF:fibril interaction, as indicated by the observed inhibition at low nM gO/CF concentration.
Our observations provide insight into the structure specificity of secondary nucleation. Decoration of amyloid fibril surfaces with gO/CFs formed from the same protein results in less efficient secondary nucleation. This demonstrates that gO/CF surfaces do not possess the same capacity as amyloid fibril surfaces to catalyze fibril nucleation, suggesting that the cross-b structure of amyloid fibrils is essential for efficient secondary nucleation. This is consistent with the distinct structural signatures of gO/CFs vs.f ibrils seen in the amide-I bands of their respective infrared spectra that we have shown for hewL and that have been reported for Ab,aswell. [20,30] Figure 4s hows an updated Scheme of oligomer and amyloid fibril formation. GO/CFs are an alternative (offpathway), metastable assembly type and form rapidly and extensively above the COC.G O/CFs inhibit amyloid formation by competing for the monomers that are required for amyloid fibril nucleation and elongation. [11] In addition, as we show here,GO/CFs actively inhibit the autocatalytic amplification of fibrils by blocking secondary nucleation sites on amyloid fibrils.
Recently,p rotofibril-fibril interactions were observed under conditions of biphasic Ab42 assembly,a nd the protofibrils were interpreted to represent nuclei formed by secondary nucleation. [31] This interpretation is in conflict with the off-pathway nature of protofibrils. [11,13] Ther esults reported here show that protofibril-fibril interactions do not represent, but rather interfere with secondary nucleation.
Thei nterplay between gO/CFs and amyloid fibrils has ah igh relevance for AD pathogenesis:G O/CFs,w hich are thought to represent the main toxic Ab species, [8,13,21,32] were shown to associate with amyloid fibril plaques in vivo,w ith potential consequences for the neurotoxic activities of both assembly types. [22,23] Fore xample,a myloid fibril plaques might serve as reservoir of toxic gO/CFs. [22,23] Our results demonstrate that the interaction of gO/CFs with amyloid fibrils affects the kinetics of formation and depletion of the . Scheme of oligomer and amyloid fibril formation.G O/CFs constitute an alternative (off-pathway)a ssembly type that competes with amyloid fibrils for monomers and that inhibits the autocatalytic amplification of amyloid fibrils by secondary nucleation. GO/CFs interfere with secondary nucleation by binding to amyloid fibrils surfaces and blocking the sites that catalyze nucleation. two species.B yb inding to amyloid fibrils,g O/CFs inhibit formation of new fibrils and thereby delay their own replacement by amyloid fibrils.T he dimAb-Ab40 system may serve as av aluable tool for further elucidation of the interplay between gO/CFs and amyloid fibrils.