Direct Observation of Oligomeric Species formed in the Early Stages of Amyloid Fibril Formation using Electrospray Ionisation Mass Spectrometry

Abstract

Numerous debilitating human disorders result from protein misfolding and amyloid formation. Despite the grave nature of these maladies, our understanding of the structural mechanism of fibril assembly is limited. Of paramount importance is the need to identify and characterize oligomeric species formed early during fibril assembly, so that the nature of the initiating assembly mechanism can be revealed and species that may be toxic to cells identified. However, the transient nature of early oligomeric species, combined with their heterogeneity and instability, has precluded detailed analysis to date. Here, we have used electrospray ionisation mass spectrometry (ESI-MS), complemented by analytical ultracentrifugation (AUC) and measurements of thioflavin-T fluorescence, to monitor the early stages of assembly of amyloid-like fibrils formed from human beta-2-microglobulin (β₂m) in vitro. We show that worm-like fibrils that form with nucleation-independent kinetics assemble by a mechanism consistent with monomer addition, with species ranging from monomer to ≥13-mer being identified directly and uniquely as transient assembly intermediates. By contrast, only monomers, dimers, trimers and tetramers are observed during nucleated growth, which leads to the formation of long straight fibrils. The results highlight the unique power of non-covalent ESI-MS to identify protein assembly intermediates in complex heterogeneous systems and demonstrate its great potential to identify and characterise individual species formed early during amyloid assembly.

Amyloidosis is a pathological disorder associated with the self-assembly of proteins into highly ordered amyloid fibrils,¹ causing debilitating diseases such as Alzheimer’s, type II diabetes and the transmissible spongiform encephalopathies.² Despite the importance of these disorders, our knowledge of the structural and molecular mechanism of amyloid formation remains limited, primarily because of the complexity and heterogeneity of these self-assembling systems. Several models for amyloid fibril formation have been suggested, based on monitoring fibril assembly via microscopy, light-scattering and/or the binding of amyloid-specific dyes.^3–7 Some types of amyloid fibrils have been shown to assemble via nucleated growth involving the formation of a critical, high-energy nucleus in a lag phase, after which fibril formation occurs rapidly.⁸ By contrast, the assembly of spherical oligomers and other fibrillar forms, including curved or worm-like fibrils, occurs with nucleation-independent kinetics.^6,9 In some cases, these species have been purported to be direct precursors of long straight amyloid fibrils,^10,11 whilst in other cases an off-pathway role has been proposed.^7,9,12 However, the inherent heterogeneity and complexity of the assembly mechanism, especially in the early stages when oligomeric species are in rapid equilibrium, has precluded direct and unique identification of species formed in the lag phase of nucleated growth or in the initial stages of non-nucleated assembly. As a consequence, only the average properties of the ensemble can be measured,^6,13 and the role of individual oligomeric species in amyloid formation remains an open question.

By contrast with other methods of protein separation and analysis, electrospray ionisation mass spectrometry (ESI-MS) is unique in its ability to transfer proteins into the gas phase. Under carefully selected conditions, protein–protein interactions can be preserved, whereupon co-existing macromolecular protein complexes can be separated and identified according to their mass-to-charge ratios and fragmentation patterns.^14–16 Advanced methods such as nano-electrospray ionisation with time-of-flight analysis have allowed studies on several large complexes, including ribosomes,¹⁷ chaperones¹⁸ and intact viruses,¹⁹ whilst ion mobility techniques can reveal differences in the conformational properties of different ions, even those resulting from species of identical mass.^20–23 In the context of amyloid fibril formation, ESI-MS has been used to monitor monomer depletion during fibril assembly of the amyloidogenic peptides, insulin and amylin, ^24,25and to identify oligomeric species formed from prions, insulin and a fragment of Aβ,^12,24,26 but the technique has not been used to date for a detailed analysis of the process of amyloid assembly under different kinetic regimes. Here, we describe a series of experiments using non-covalent nanoESI-MS to directly monitor and identify co-existing species formed in real time during the early stages of assembly of amyloid-like fibrils from the 100 residue protein (including the initiating methionine) human β₂-microglobulin (β₂m), under different growth conditions. Specifically, oligomeric species formed during nucleated and non-nucleated assembly are compared, and their presence and kinetics of assembly validated directly using analytical ultracentrifugation (AUC). The results highlight the immense power of modern MS methods to resolve and identify intermediates in these complex self-assembling systems, providing new insights into the assembly landscape of this protein.

Monitoring monomer depletion during fibril formation by ESI-MS

Previous studies have shown that fibril formation from β₂m can be initiated in vitro by partial or more extensive denaturation of the native protein in acid.^27–31 Under these conditions, β₂m assembles readily into amyloid-like fibrils of different morphological type. At pH 2.5 and low ionic strength, long straight fibrils with a defined left-handed twist are formed in a nucleation-dependent manner.³² By contrast, small rod-like particles and worm-like fibrils form rapidly at pH 3.6 in the absence of a lag phase.^9,33 Both fibrillar species are amyloid-like, in that they bind the dyes thioflavin-T (Thio-T) and Congo red, the latter resulting in red-green bire-fringence when viewed under polarised light, and give rise to X-ray fibre diffraction patterns consistent with a cross-β architecture,³³ the classical hallmarks of amyloid-like material. A previous detailed and systematic analysis of the effect of protein concentration, agitation, pH and ionic strength on β₂m fibril assembly using atomic force microscopy (AFM) has shown that the path-ways for formation of worm-like and long straight fibrils compete.⁹ Thus, whilst assembly at pH 2.5 results exclusively in the formation of long straight fibrils in the absence of amorphous aggregates or fibrils of other morphological types, incubation at pH 3.6 at high ionic strength results in the formation of only worm-like fibrils that are kinetically trapped as the end-product of assembly. At pH or ionic strength values between these extremes, both kinetic mechanisms occur concomitantly, resulting in the formation of both long straight and worm-like fibrils under the same incubation conditions. In order to examine the early stages of amyloid fibril formation by nucleated and non-nucleated mechanisms, therefore, we chose to study the assembly mechanisms of β₂m into amyloid-like fibrils at pH 2.5 and at pH 3.6, conditions that result in fibril growth with uniquely defined kinetic mechanisms. Our aim was to reveal the identity of very early oligomeric species, and to follow their formation and depletion during fibril assembly, in order to cast more light on the molecular species involved in fibril assembly under different kinetic regimes.

ESI-MS has been used to determine the concentration of protein monomers during oligomerisation, employing either a non-amyloidogenic isoform of the target protein^24,25 or a small peptide³⁴ as an internal calibrant of protein concentration. Here, we chose bradykinin as internal standard, as this peptide does not form adducts with β₂m under the conditions of these experiments and is detected predominantly as doubly charged ions (M+2H)²⁺ at m/z 530.8, which do not overlap with the multiply charged ion series of β₂m (n = 12+ to 6+; m/z 989–1978; Figure 1(a) and (b)). Fibril formation was initiated either at pH 3.6 without agitation in 150 mM ammonium formate buffer, which results in the formation of worm-like fibrils without a lag phase (monitored using the fluorescence of Thio-T; Figure 1(c)), or at pH 2.5 with agitation in 100 mM ammonium formate, resulting in nucleated growth of long straight fibrils (Figure 1(d)), consistent with the results obtained under each condition in non-volatile buffers.^9,33 Initially, a concentration calibration curve ranging from 0.01–0.4 mg/ml monomeric β₂m was determined in the same buffer but at lower ionic strength (50 mM), conditions under which no aggregation occurs over the time-scale of the experiment (data not shown). Under these conditions, accurate measurement of ion concentration could be made by calibration of the ion intensities of β₂m and bradykinin, as shown by the close agreement between the protein concentration determined by nanoESI-MS compared with UV spectroscopy. The concentration calibration curve (ion count versus protein concentration) was linear over the range analysed (data not shown) and, therefore, was utilised directly to convert the experimental ion intensities into protein concentration. Fibril formation from 0.2 mg/ml β₂m was then initiated at pH 3.6 or pH 2.5 by diluting the protein into the appropriate buffer solution and determining, in real time, the concentration of monomer remaining at different times during assembly using nanoESI-MS. In parallel, assembly-incompetent samples (at the same pH and protein concentration, but in 50 mM buffer) showed no decrease in the concentration of monomeric protein over the timescale analysed (data not shown), confirming that any loss in monomeric protein can be attributed to amyloid fibril formation and demonstrating the accuracy and stability of the ion calibration over the entire time-course of the experiment. Experiments were conducted on an LCT Premier time-of-flight mass spectrometer (Waters Corp., Manchester, UK) with collisional cooling capabilities to preserve non-covalently bound species and an extended m/z range of 60,000, enabling detection of as many oligomeric species as possible. Sample injection and ionisation was performed using a temperature-controlled NanoMate (Advion, Ithaca, USA) nanoESI autosampler and ionisation interface for optimum reproducibility.

Figure 1. Assembly time-course of β₂m fibrils analysed by nanoESI-MS.

(a) and (b) Mass spectra acquired during amyloid fibril formation from 17 μM β₂m at 37 °C in (a) 150 mM ammonium formate at pH 3.6 and (b) in 100 mM ammonium formate pH 2.5 (the latter shaken at 200 rpm). The spectra are normalised to the (M+2H)²⁺ ions of bradykinin (1.4 μM, added immediately before analysis) at m/z 530.8 and shown over m/z 500–2200 to highlight the charge state distribution arising from monomeric β₂m. The 6+ to 10+ charge states of β₂m are labelled. (c) and (d) Fibril growth kinetics at (c) pH 3.6 and (d) pH 2.5 monitored using Thio-T fluorescence (continuous line in (c) and open squares in (d)).³³ The concentration of monomeric β₂m measured by nanoESI-MS is shown in both (c) and (d) as filled circles (error bars represent one standard deviation from the mean). (c) NanoESI-MS data at pH 3.6 can be described by an exponential function (continuous line). (d) Data at pH 2.5 show that significant depletion of monomer occurs during the lag phase but that this is not related directly to the detection of fibrils. Grey line: theoretical curve of the concentration of monomer computed assuming that the Thio-T fluorescence signal is inversely proportional to the monomer concentration, highlighting the decline in monomer concentration observed in the lag phase. (e) and (f) Negative stain EM images of fibrils formed at (e) pH 3.6 and (f) pH 2.5 (the scale bar represents 200 nm).

Examples of spectra acquired during fibril assembly at pH 3.6 are shown in Figure 1(a). The results demonstrate that the concentration of monomeric β₂m (11.8 kDa) decreases immediately upon dilution of the protein into the assembly buffer, and its concentration declines rapidly throughout assembly; quantification of the peak intensities show that only ~20% of monomeric material remains after 5 h (Figure 1(c)). Interestingly, the decline in monomer concentration mirrors precisely the rate of formation of amyloid-like material capable of binding Thio-T (Figure 1(c) and (e)). Control experiments performed under identical conditions, but at a buffer concentration of 50 mM, showed the absence of fibrillar material and no significant change (< 10%) in monomer concentration over the same time-course, consistent with the known ionic strength-dependence of the rate of fibril growth at this pH (data not shown).^9,33

By contrast with the results obtained at pH 3.6, fibril assembly at pH 2.5 follows nucleation-dependent kinetics.^32,33 Under the conditions used here, assembly at pH 2.5 involves a lag phase of ~8 h monitored by the fluorescence of Thio-T, after which time amyloid-like, long straight fibrils form rapidly (Figure 1(d) and (f)). Analysis of the sample during fibril growth using nanoESI-MS demonstrated that the concentration of monomer decreases significantly, but only gradually, during the lag phase, such that ~50% of the protein remains monomeric at the end of this phase (Figure 1(b) and (d)). Over the same time-scale, no change in Thio-T fluorescence is noted (Figure 1(d)) and no high molecular mass oligomers or fibrillar material are observed using negative stain EM (data not shown). The data suggest, therefore, that the loss of monomeric protein during the lag phase results from the formation of low molecular mass oligomers that are too small to detect using EM and are either not capable of binding Thio-T or bind Thio-T but result in a fluorescence intensity too low to be detected. Control experiments performed under identical conditions, but in the absence of agitation, showed no fibril formation over 24 h, and no change in monomer concentration over the entire time-course (data not shown). During the elongation phase (10–15 h in the experiment shown in Figure 1(b) and (d)) the nanoESI-MS data show that the concentration of monomer decreases rapidly until it is reduced to less than 0.01 mg/ml, at which point fibril elongation halts. The data obtained at pH 3.6 and pH 2.5 indicate, therefore, that the rate of depletion of monomer is markedly dependent on the growth conditions (compare Figure 1(c) and (d)), consistent with the known differences in the kinetic mechanism of fibril formation at each pH value.

Identification by ESI-MS of oligomeric species formed early during fibril assembly

To identify oligomeric species formed during both nucleated and non-nucleated assembly of β₂m fibrils, samples after 1 h of fibril assembly at pH 3.6 and pH 2.5 were analysed by infusion into the LCT Premier mass spectrometer and examined over an extended m/z range of 500–20,000. The results obtained were striking, revealing clear differences in the number and type of oligomeric species formed in each assembly mechanism (Figure 2). Thus, whilst a wide range of oligomers is detected during fibril formation at pH 3.6, resulting in complex spectra (Figure 2(a)), few higher molecular mass species are detected during fibril growth at pH 2.5 (Figure 2(b)). To assign peaks to different oligomeric species, the m/z values of the different charge states of all possible oligomers (up to a maximum of 13-mers) were calculated (see Supplementary Data). Subsequently, each set of ions in the spectrum was assumed to arise from the oligomers that could result in ions at this particular m/z value. Next, a minimum of two consecutive charge states was required to identify any particular oligomer, and the charge states assigned to a particular species had to have consecutive charges (i.e. “missing” charges were not allowed in oligomer identification; see Supplementary Data for further details). Whilst many peaks are degenerate (Figure 2; and Supplementary Data Table S1) several ions, particularly those carrying a prime number of charges, represent unique oligomeric species. The presence of such ions, therefore, is of great importance in ascertaining that an oligomer of a particular molecular mass is populated during assembly. Using this approach, all species from the monomer to the 13-mer (154.2 kDa) could be identified unambiguously during fibril assembly at pH 3.6, despite the complexity of the spectrum.

Figure 2. Oligomeric species formed during β₂m amyloid fibril formation.

ESI mass spectra of the region m/z 3200–8000 of 17 μM β₂m after 1 h at (a) pH 3.6 or (b) pH 2.5. The peaks are numbered with the oligomers that have been identified, showing the spread and range of detected oligomers at pH 3.6 compared with that at pH 2.5. X denotes peaks that could contain contributions from all detected oligomers (see Supplementary Data Table S1). For these spectra, an LCT Premier with a standard ESI source was operated in V mode with a source temperature of 60 °C, capillary voltage of 2800 V, cone voltage of 40 V, ion guide 1 of 140 V, aperture 1 of 20 Vand an MCP of 2700 V. Samples were infused at a rate of 3 μl/min. Acquisitions were recorded for a total of 10 min using a 2.4 s scan time in continuum data format over a range of m/z 500–20,000 to optimise the signal-to-noise ratio. Spectra were calibrated using a separate introduction of cesium iodide.

Next, a time-course of assembly was monitored using nanoESI-MS under the conditions described above. Example spectra, expanded in the m/z range 3200–5500 for clarity, obtained at five specific time-points during fibril growth at pH 3.6 are shown in Figure 3(a)–(d). The spectra indicate that before the initiation of fibril formation (in 50 mM buffer) β₂m is predominantly monomeric, only low concentrations of dimers are found in this m/z range (Figure 3(a)). As soon as fibril formation is initiated by dilution of the sample into 150 mM ammonium formate (t ≥ 2 min), higher molecular mass species are observed (Figure 3(b)). Detailed analysis revealed that oligomers ranging from dimers (23.7 kDa) to hexamers (71.2 kDa) are formed immediately (t=2 min) upon initiation of fibrillogenesis at this pH (Figure 3(b)). These peaks decrease in intensity over the next 2 h (Figure 3(c)) and are replaced by peaks corresponding to larger oligomers. Figure 3(c) highlights the range m/z 3200–5500 in which oligomers ≤11-mer appear. Most importantly, the unique and unambiguous identification of every species ranging from monomer to at least the 11-mer appearing in this m/z range during assembly (Figure 4(a)), combined with the apparently exponential decrease in monomer concentration with time (Figure 1(c)), suggest strongly that at least the initial stages in the formation of worm-like fibrils occurs by a mechanism involving monomer addition, in agreement with proposals for other amyloid-forming proteins that form flexible fibrillar assemblies.³⁵ Although a detailed quantitative analysis of the concentration of individual oligomers versus time is not possible (due, in part, to peak degeneracy and because the ability of individual species to enter the gas phase cannot be calibrated), the successive formation of oligomers of increasing molecular mass as a function of assembly time is clearly illustrated (Figures 3(b) and (c) and 4(a)). At longer times (8 h and 24 h), these higher molecular mass species are no longer observed (Figure 3(d)), presumably because the species formed are too large or at a concentration too low to be detected (Figure 4(a)).

Figure 3. The β₂m oligomers detected during fibril assembly using nanoESI-MS (m/z 3200–5500).

Fibril assembly was performed at 37 °C at (a)–(d) pH 3.6 and (e)–(h) pH 2.5. The LCT Premier was operated in V mode with a source temperature of 60 °C, cone voltage of 80 V, ion guide 1 of 120 V, aperture 1 of 30 V and the MCP at 2700 V. A capillary voltage of 1800 V was used on the NanoMate ionisation device. Acquisitions were recorded for a total of 3 min using a 2.4 s scan time in continuum data format over a range of m/z 500–20,000. Triplicate spectra were processed using MassLynx software by summation of spectra over a 3 min period. Spectra are all normalised to the intensity of bradykinin and shown here relative to the most intense peak in each series (a)–(d) and (e)–(h)). The peaks are labelled with all oligomers that could contribute to that m/z value, which have at least one adjacent charge state, and all charge states relating to that species are consecutive (see Supplementary Data). The insets in (d) and (h) show the absence of detectable oligomeric species at the completion of assembly (t = 24 h).

Figure 4. Oligomer distributions observed during β₂m assembly measured by nanoESI-MS over a range of m/z 3200–5500 and AUC.

(a)–(b) Summary of species detected during fibril assembly at 37 °C at (a) pH 3.6 and (b) pH 2.5 using nanoESI-MS. Note that as the concentration of each oligomer cannot be quantified by nanoESI-MS, the population of individual species is denoted as either present or absent in this chart. (c) and (d) Sedimentation velocity derived distributions of sedimentation coefficients of samples taken during the assembly of 42 μM β₂m at (c) pH 3.6 and (d) pH 2.5. Samples were analysed using a Beckman Optima XL-1 ultracentrifuge (Beckman Coulter) at 20 °C in standard double-sector cells in an An60Ti four-cell rotor at rotor speeds ranging from 3000–50,000 rpm. The absorbance at 280 nm was measured versus time and the data were analysed using SEDFIT.⁴⁷ Time-points from 0 h to 8 h (black to brown) in steps of 1 h are shown. The size distribution of the starting solution (obtained in aqueous solution in the absence of buffer) is shown as a broken black line. Insets indicate the Thio-T fluorescence (filled circles) and the total absorbance of soluble material measured in the first scan (open circles) of individual samples.

Analysis of oligomeric species formed during nucleated growth of β₂m fibrils at pH 2.5 resulted in spectra that differ profoundly from those obtained at pH 3.6 (Figures 3(e)–(h) and 4(b)). No species larger than a tetramer was detected during the entire time-course of fibril growth at this pH. Instead, rapid formation of dimer (23.7 kDa), trimer (35.6 kDa) and tetramer (47.4 kDa) peaks occurred immediately upon dilution of the protein into the assembly buffer, and these species remained significantly populated throughout the lag phase. These oligomeric species persist for ~8 h, consistent with the observed decrease in concentration of the monomer during the lag phase and the lack of fibrillar material assayed using Thio-T fluorescence (Figure 1(d)) and EM (not shown). During the elongation phase, the concentration of these small oligomers decreases rapidly as fibrillar species capable of binding Thio-T develop, without the population of further detectable oligomeric species.

The lack of higher molecular mass oligomers, despite the sensitivity of ESI-MS and the previously demonstrated ability to resolve such species by this technique, suggests that oligomers are either not populated or are formed at a concentration too low to be detected by this approach.

Sedimentation velocity analytical ultracentrifugation (AUC)

In order to validate the results of MS obtained above, a parallel series of experiments were performed in which the average molecular mass of species formed during fibril assembly at pH 3.6 or pH 2.5 was monitored using sedimentation velocity AUC^36,37 under buffer conditions identical to those used for the MS experiments, but at a higher concentration of protein than that used for the ESI-MS experiments (0.5 mg/ml, see the legend to Figure 4). In parallel, the development of species capable of binding Thio-T was monitored and the total concentration of soluble material was determined by measuring the absorbance (A_280nm) of samples immediately upon application of the centrifugal force (3000 rpm). The resulting data (Figure 4(c) and (d)) show a close qualitative agreement with the results of the ESI-MS experiments (Figure 4(a) and (b)). Thus, in the absence of buffer salts at pH 3.6 and pH 2.5, β₂m is predominantly monomeric with a sedimentation coefficient of 1.5 S and a molecular mass of ~12 kDa (Figure 4(c) and (d) dotted line). Upon commencement of fibril growth at pH 3.6, the concentration of monomeric β₂m decreases rapidly and a continuous shift in the sedimentation coefficient distributions, c(s), towards higher molecular mass species (up to a maximum of ~28 S) is observed (Figure 4(c)). The breadth of the peaks observed in the c(s) profiles indicates that the sedimenting species interconvert on a time-scale that is rapid compared with the rate of sedimentation. As a consequence, the observed peaks cannot be assigned to discrete oligomeric species, but instead reflect the mass-average sedimentation coefficient of the sedimenting particles. Concomitant with the increase in molecular mass, an increase in Thio-T fluorescence occurs, consistent with the early oligomers possessing the ability to bind this amyloidophilic dye, as has been observed for oligomeric species of β₂m at neutral pH.³⁸ The approximately constant concentration of soluble material throughout fibril growth (at a rotor speed of 3000 rpm) highlights the fact that these early oligomers remain soluble (Figure 4(c) inset). The final oligomer size distribution at pH 3.6 was determined to be of the order of 500–600 kDa by this analysis (40–50-mers), assuming a single frictional ratio for all species in the sample. Such species are not detected by ESI-MS, presumably because of the complexity of the mixture and the fact that peaks for the large number of individual species within the heterogeneous mixture would be expected to be of very low intensity. In addition, large oligomers may ionise inefficiently relative to lower molecular mass species under the conditions used, adding further challenges to their detection.

In the case of fibril formation at pH 2.5, the AUC results are again strikingly different from those obtained at the higher pH value. Thus, whilst monomeric β₂m predominates before fibrillogenesis is initiated, an equilibrium between species of 1.5 S and up to ~6 S is rapidly established early during assembly. Two distinct species (~2 S and ~4.3 S) are clearly visible and remain in solution throughout the lag phase (Figure 4(d) inset), consistent with the dimers, trimers and tetramers visualised by ESI-MS. At times longer than 6 h, the sample sediments rapidly, even at the lowest rotor speed used (3000 rpm), precluding analysis of the molecular mass distribution of material at this time, consistent with rapid elongation into large fibrils, leaving a low concentration of soluble species (Figure 4(d) inset). The data confirm that β₂m assembles into long straight fibrils at pH 2.5 after a lag phase without substantial population of species larger than tetramers. By contrast, a continuous distribution of oligomers ranging from monomer to ~50-mers is observed during assembly at pH 3.6, validating the ESI-MS approach for the analysis of non-covalent species on the aggregation pathway, and underlining its power in the identification of discrete species in complex mixtures.

Mechanistic insights into the formation of oligomeric species

Assembly of β₂m into amyloid-like fibrils at pH 2.5 involves significant conformational rearrangement from the initially highly dynamic, unfolded monomer^30,31 to the compact, stable cross-β structure of amyloid.^33,39–41 Using non-covalent ESI-MS and AUC experiments, we demonstrate that in the lag phase of assembly observed by Thio-T fluorescence, monomeric to tetrameric species are populated, suggesting either that nucleation-dependent assembly occurs via an unstable (high-energy) oligomer greater than tetramer in size or that conformational rearrangement of one or more of these species constitutes nucleation (Figure 5). The population of soluble oligomeric species in the lag phase has been described for other amyloid-like systems.²⁴ The experimental approach described here in which the concentration of monomer is quantified with concomitant identification of the higher-order oligomeric species within the same sample throughout the entire fibril assembly process demonstrates the power of real-time nanoESI-MS to resolve and identify individual species, even when formed transiently and in complex mixtures during fibril assembly.

Figure 5. A diagram of the different mechanisms of β₂m amyloid formation under acidic conditions.

Fibril formation at pH 3.6 is initiated by the sequential addition of monomers into large molecular mass oligomers (up to 40–50mers) in a nucleation-independent reaction. Whether monomer addition continues throughout assembly or involves the association of existing oligomers into worm-like fibrils at later stages is not known. By contrast, the formation of long straight fibrils at pH 2.5 proceeds by a rate-limiting nucleus formation, followed by rapid fibril extension. NanoESI-MS and AUC data reveal that each pathway proceeds via a mechanistically distinct development of oligomeric species.

By contrast with the classical nucleated growth mechanism at pH 2.5, spontaneous self-association of β₂m at pH 3.6 at high ionic strength proceeds from an initially compact, partially folded monomer,²⁹ without a lag phase, and results in the generation of flexible worm-like fibrils via an extensive series of oligomeric forms (Figure 5). Here, our results demonstrate that the concentration of monomeric β₂m declines in an exponential manner during the initial stages of non-nucleated assembly, as has been proposed for other amyloid systems.^25,42 While previous studies of monomer depletion using ESI-MS have also analysed the elongation phase in seeded reactions,²⁵ we show here the mechanism by which monomer is depleted during de novo non-nucleated fibril formation at pH 3.6. Whether the 100–500 nm worm-like fibrils assemble exclusively by monomer addition, or whether coalescence of different oligomeric forms occurs in the later stages of assembly, is unknown. The equilibrium between partially folded and more highly unfolded species is highly dependent on the solution conditions.⁹ Thus, at pH 3.6 the accumulation of partially structured species is favoured and β₂m assembles rapidly into worm-like fibrils in a kinetically driven mechanism via monomer addition (Figure 5). At lower pH values, nucleation competes with monomer addition, resulting in the thermody-namically driven⁹ formation of long straight amyloid-like fibrils as the product of assembly (Figure 5). While analysis, for example of the protein concentration-dependence of the rate of fibril assembly, can be used to infer a kinetic mechanism for fibril assembly,^6,35 the direct identification of assembly intermediates along the reaction coordinate established in this study provides unique insights into the species formed early during amyloid assembly.

Conclusion

Biology utilises protein assembly, often together with different nucleic acid components, to create large and complex macromolecular machines that perform some of the most important functions in life. Understanding how these assembling systems are created and controlled is an enormous challenge, often because of their ephemeral nature, inherent instability and/or heterogeneity. The self-assembly of polypeptide chains into amyloid fibrils is no exception, and delineating the nature of transient oligomeric species that initiate this process presents an important challenge. Here, using β₂m as a model system, we have demonstrated the power of non-covalent nanoESI-MS to reveal new insights into the early stages of amyloid fibril formation, showing that individual oligomeric species within complex mixtures of self-assembling molecules can be identified unambiguously and their behaviour monitored in real time without the need for prior separation. This approach has unique advantages over other methods of determining the molecularity of the assembly reaction and the identity of assembly intermediates, as the mass of individual species within the ensemble is measured directly, without recourse to assumptions about the hydrodynamic properties of the assembling species. Combined with complementary techniques, such as AUC and Thio-T fluorescence, the data presented illustrate the power of nanoESI-MS to monitor the change in the concentration of the monomer during fibril assembly and to discriminate between different assembly mechanisms by the direct observation of oligomer formation in real time. Species ranging from dimers to 28-mers have been shown to be the culprits of amyloid-associated toxicity and to be the most efficient initiators of TSE diseases.^43–45 Further work quantifying the concentration of individual species, for example by the construction of suitable calibration curves, may allow correlation between different oligomeric species and biological function. The high sensitivity, resolving power and reproducibility of modern methods of nanoESI-MS, combined with the exciting potentials of ion mobility, isotope labelling and MS/MS methods,^17,23,46 also offer exciting opportunities to screen for molecules that enhance or disfavour amyloid formation and to determine their mode of action, including the direct identification of oligomeric states to which they bind.

Abbreviations

ESI-MS

electrospray ionisation mass spectrometry

β₂m

β₂-microglobulin

AUC

analytical ultracentrifugation

Thio-T

thioflavin-T

AFM

atomic force microscopy

Supplementary Material

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.jmb.2006.08.081