Mass and charge distributions of amyloid fibers involved in neurodegenerative diseases: mapping heterogeneity and polymorphism

Characterization by charge detection mass spectrometry of amyloid fibers involved in neurodegenerative diseases: Aβ peptide, tau and α-synuclein.


Introduction
The most frequent age-related neurodegenerative diseases, Alzheimer's and Parkinson's diseases, are related to the accumulation of amyloid deposits due to the aggregation of specic proteins. 1,2 In the case of Alzheimer's disease, Ab peptides form extracellular plaques and tau protein accumulates as intraneuronal inclusion bodies. 3,4 This is observed in relation to synaptic dysfunction, neuron death, brain shrinkage and, ultimately, dementia. Parkinson's disease is associated to the appearance of intracellular deposits made of a-synuclein, so-called Lewy bodies, leading to dopaminergic neuron death (lack of DOPA synthesis) and to motor system disorder. 5 The proteins involved in these deposits are in the so-called amyloid state, with common structural features: 6,7 high aspect-ratio bers, with diameters of a few nanometers and lengths around a micrometer, stabilized by hydrogen-bonded b-strands perpendicular to the ber axis and forming b-sheets. Beyond these generic features, amyloid bers are characterized by a polymorphism which is observed within in vivo amyloid deposits 8,9 and within samples prepared in vitro. [10][11][12] As amyloid bers are oen an association of protobrils, their heterogeneity, i.e. their polymorphism, depends on the number of pro-tobrils, the arrangement of protobrils or the conformation of polypeptide. 10 According to the classical view, the formation of amyloid bers follows a nucleation/growth mechanism, i.e. a primary nucleation mechanism. 13 The initial step, which is also the slowest one, is the formation of oligomers which act as nuclei for the growth of the bers. In the case of Ab 1-42 peptide, these oligomers can be of various sizes from dimers up to dodecamers. 14-16 Moreover, preformed bers potentially enable an additional nucleation pathway, so-called secondary nucleation mechanism. [17][18][19] Their surface can act as template for the formation of oligomers and protobrils, and their fragmentation can generate new growth sites. At high ratio of preformed bers, the bril-dependent secondary nucleation mechanism can surpass the primary nucleation and becomes the main source of new nuclei. [19][20][21] This could provide a relationship between the accumulation of amyloid deposits and toxicity in vivo, 19 through the constant release of oligomers, thought to be the most toxic species. [22][23][24][25][26] We have shown recently that charge detection mass spectrometry (CDMS) can be used to accurately measure masses of individual amyloid brils, 27 while previous MS-based studies of brillation have been limited to the early steps in aggregation. 28,29 The mass of the biological assembles of few megadaltons to 18 MDa can be measured by native mass spectrometry [30][31][32][33][34] but it is very challenging due to the charge states resolving problems. 35,36 Single-molecule CDMS technique, where mass and charge are measured simultaneously, has previously been used for DNA, polymers and various virus capsids. [37][38][39][40][41][42][43][44][45] In our previous study, samples containing a single population of amyloid brils have been characterized. 27 This provided important information about amyloid brils, such as their mass, charge density and the number of proteins involved. However, disease-related amyloid brils are also characterized by signicant heterogeneity and polymorphism; 8-10 mass and charge distribution of large heterogeneous and polymorph amyloid brils have never been characterized by any methods. We report here the characterization by CDMS of amyloid bers made of the proteins involved in neurodegenerative diseases: Ab 1-42 peptide, tau and a-synuclein. Beside the mass distribution for the different amyloid bers, this technique allows to highlight and characterize the heterogeneity of the populations, with the possibility to distinguish several species, as illustrated with Ab 1-42 peptide and tau, and to quantify the polymorphism, as illustrated with a-synuclein. In that case, we show how the polymorphism affects the mass and charge distributions.

Results and discussion
The formation of amyloid bers by tau was triggered by the addition of heparin at a molar ratio of 2.2 (see "Methods" within ESI †). 46 Transmission electron microscopy (TEM) of this sample showed that the main species corresponded to well-dened straight bers together with few more-or-less spherical oligomers (Fig. 1A). Two populations could be seen also on the CDMS 2D-graph (Fig. 1B). They could be further distinguished by their time-of-ight ( Fig. 1C and S1 †). The main population, the "high" mass population, had a mean mass of 113.5 MDa (Fig. 1C). Because of the heparin, only an estimation of the number of proteins per bril could be given. Assuming that the ratio tau/heparin was the same within the bers and in the bulk, i.e. 2.2, we obtained around 1835 proteins per ber on average (M tau ¼ 45.85 kDa and M heparin ¼ 8 kDa). The length distribution estimated from electron microscopy image was quite broad (Fig. S1 †) and this impeded to estimate a mass per length value with some meaning.
The "low" mass population, on the bottom detection limit of the CDMS experiment (around 13 MDa), was probably due to the spherical oligomers seen in the electron microscopy image (Fig. 1A). Based on a protein density at 1.41 g cm À3 (ref. 47), a mean molar mass of 13 MDa corresponded to spherical oligomers of about 30 nm, in agreement with the size of the oligomers observed by electron microscopy. The differences of the charge vs. molar mass slopes corresponding to the two populations ( Fig. 1B) indicated that the oligomers were highly charged; much more than the bers. The binding of heparin was necessary for the formation of tau bers in order to screen the charges within the fuzzy coat of the bers. 46,48,49 A possible explanation for the difference of charge density between the spherical oligomers and the bers was that no heparin was bound to the oligomers. Then, these later contained 283 proteins. Moreover, based on number of macro-ions counted for each population; i.e. 364 vs. 6836 for oligomers and brils respectively, the former represented 5% of the detected macroions, and the later 95% of them (Fig. 1C).
Amyloid bers made of Ab 1-42 peptides were obtained upon incubation at pH 6.5 and 37 C. At least two ber populations: short and curly protobrils with a strong tendency to aggregate into clusters and elongated straight bers, could be distinguished on the electron microscopy image (Fig. 2A). From TEM images, the predominant population was attributed to clusters of protobrils ( Fig. 2A and S2 †). On a higher magnication image (Fig. S2 †), we could see that our sample showed strong similarities with observations reported earlier; 50 branching on the ber sides (Fig. S2, † arrows) showed that some secondary nucleation was occurring. Therefore, the nal state of our sample was the result of a competition between primary and secondary nucleation mechanisms. Mass measurements have been performed on 9642 single Ab ber macro-ions and the results were gathered into a 2D graph (charge vs. mass) (Fig. 2B). Although less evident than in the case of tau, two different charge vs. mass dependencies could be seen on the 2D graph. Then, based on their time-of-ight, two populations could be extracted from the 2D-graph: a "low" mass (centered on 20 MDa) and a "high" mass population (centered on 55 MDa) (Fig. 2C). We have shown with nanoparticles that CD-MS allowed to distinguish different types of clusters and provided estimations of their relative populations in agreement with TEM measurements. 51 According to the amount of macro-ions counted for each population: 759 for "low" mass vs. 9642 for "high" mass, the "low" mass population counted for 8% of the macro-ions. Based on the obvious ratio of populations within the electron microscopy images ( Fig. 2A and S2 †), the "low" mass population could be assigned to the elongated bers, and the "high" mass population to protobril clusters.
For the elongated bers, a molar mass of 20 MDa gave 4400 peptides per ber (M Ab ¼ 4.51 kDa). Their length distribution was centered on 0.9 mm (Fig. S2 †), this gave a mass-per-length (MPL) value around 22 kDa nm À1 , in agreement with the values based on electron cryomicroscopy image processing, i.e. $20 kDa nm À1 . 52 These parameters could not be determined in the case of the "high" mass population which corresponded to protobril clusters. Nevertheless, the fact that their charge density was lower (weaker slope of the charge vs. mass dependency) indicated either that they were less electrically charged or that their association induced some charge screening.
Two types of a-synuclein bers have been obtained. Although the reasons for the differences between the two samples are not fully understood (see ESI †), their characteristic in terms of mass and charge could be clearly discriminated by CD-MS. According to TEM images ( Fig. 3A and B), isolated bers, referred to as type I, were formed in the rst sample (Fig. 3A), while irregular ribbons, referred to as type II, were observed in the second sample (Fig. 3B). These ribbons resulted from the heterogeneous association of brils of variable lengths. Moreover, bers involved in ribbons were obviously shorter than those observed in the sample with isolated bers. According to their respective length distributions extracted from several TEM images (Fig. S3 †), isolated bers of the rst sample type had a mean length of 0.9 mm, while those involved within ribbons had a mean length of 0.5 mm. Still according to TEM images, the most populated ribbon species was that made of the association of two bers (i.e. $58%), and the probability decreased when the number of associated bers increased (Fig. S3 †). Moreover, the probability to have ribbons made of an even numbers of bers was much higher than that of ribbons with an odd number of brils. Both types of bers were further characterized by Atomic Force Microscopy (AFM) (Fig. S4 †). The height proles of isolated bers showed a single maximum around 8 nm. In the case of ribbons, the proles were much  respectively. The mass distribution is histogrammed using a given bin-size (10 MDa). Each bar represents the number of measured ions whose masses correspond to the mass range of the bin. broader, with several peaks corresponding to aligned bers. At the exception of regions with overlapping bers, the average height of the ribbons was around 6-8 nm. This was close to the height of isolated bers suggesting that the ribbons were mostly the results of the lateral association of bers into 2D structures.
The CDMS 2D graph recorded with the isolated bers showed a well-dened charge vs. mass dependency (Fig. 3C). According to the mass distribution (Fig. 3D), the mean mass was 85.4 MDa. Hence, these a-synuclein amyloid bers were made of 5900 molecules on average (M a-syn. ¼ 14.46 kDa). According to the length distribution estimated from electron microscopy (Fig. S3 †), the average ber length was 0.90 mm, giving an estimation of the mean MPL value around 95 kDa nm À1 , to be compared with that determined from electron microscopy image processing, i.e. 60 kDa nm À1 . 53 Our value must be taken with caution because of the poor quality of the length distribution extracted from the TEM images (Fig. S3 †). Given the disparity in length, a much larger sampling would be required to obtain a precise value.
In the case of ribbons, the charge vs. mass dependency was not so well dened (Fig. 3E), resulting in a much broader mass distribution, with a mean mass at 147.8 MDa (Fig. 3F). The broadness of mass distribution reected the heterogeneity of the sample, with ribbons of varying lengths and widths. It was obvious on the 2D-graph that the charge of the ribbons was signicantly lower than in the case of the isolated bers; a dashed line corresponding to the charge vs. mass dependency of the isolated bers was reported on both 2D-graphs for comparison. The pH being identical for both types of sample, this suggested that charges were at least partially hidden due to the organization of the bers into ribbons. In our electrospray ionization (ESI) mass spectrometry (MS) investigation with a positive polarity mode, only positively charged gas-phase electrosprayed brils were measured, resulting from a complex desolvation process of highly positively charged solvent droplets. The net charge of electrosprayed brils produced by ESI was mainly determined by the number of positively charged sites on their surface. In a previous work, 54 we demonstrated with latex nanoparticles that the magnitude of charging of ions produced in the gas phase was correlated with the surface charge in solution, however their values cannot be directly compared. Therefore the values of the charge of a-synuclein bers reported in the 2D graph (Fig. 3C) cannot be compared to those extracted from electrophoretic mobility measurements on brils in solution. 55 However, the phenomena described to explain the fact the bers in solution are drastically less charged than expected from the charge of monomers, i.e. shi of ionizable residue pK a values and/or incorporation of counter ions into oligomers, 55,56 must occur in our experiments. Thus, the net charge of electrosprayed amyloid bers was signicantly smaller (more than ten times) than expected from the net charge of monomers monitored by ESI-MS. 27 According to the electron microscopy and AFM images, the ribbons were due to the 2D association of the bers. Likely, the weaker charge density of the ribbons was due to the further burying of ionizable groups into the interface, allowed by a shi of their pK a values or to the incorporation of counter ions. This allowed to use a simple model assuming some counter ion incorporation (easier to visualize than a shi of pK a ) (Fig. 4) to estimate the effect of the brils association on the charge density (charge per mass unit). The ratio between the slopes of the two associated 'charge vs. mass' graphs, either for single or associated brils, is equivalent to the ratio: r, between the charges present within a set of either isolated brils or laterally associated into ribbons: where n is the number of considered brils, N tot is the number of charges carried by a single bril and N int is the number of charges involved in the ribbon association.
According to this equation and the value of the ratio between the slopes of the 'charge vs. mass' graphs ( Fig. 3C and E), i.e. 1.7, about 20% of the charges of bers are involved in the lateral association.

Conclusions
Charge-detection-mass-spectrometry provides a wealth of information on amyloid ber samples. Beside the mass and charge of individual bers, this technique enables to characterize the heterogeneity of the population and to detect the i.e. the number of fibers involved in the ribbon, considering a ratio between the two 'charge vs. mass' graphs equal to 1.7, which was extracted from Fig. 3. About 20% of the charges are involved in the interaction; the value of N int /N tot tends toward 0.206 for high values of n.
presence of different types of bers. This is of prime importance with amyloid ber samples, well-known to be highly heterogeneous and, as a consequence, difficult to accurately characterize. In association with time-resolved experiment, this will allow to investigate the mechanisms of formation and maturation of amyloid bers, so important to get insight into the development of the neurodegenerative diseases. The association of classical MS and CDMS with separative methods will allow the complete characterization of the species involved from monomer to amyloid bers, through oligomers.

Conflicts of interest
There are no conicts to declare.