Coexisting Order and Disorder Within a Common 40-Residue Amyloid-β Fibril Structure in Alzheimer's Disease Brain Tissue

Abstract

Fibrils formed by 40- and 42-residue amyloid-β (Aβ40 and Aβ42) peptides exhibit molecular-level structural polymorphisms. A recent screen of fibrils derived from brain tissue of Alzheimer's disease patients revealed a single predominant Aβ40 polymorph. We present solid state nuclear magnetic resonance (ssNMR) data that define its coexisting structurally ordered and disordered segments.

Graphical abstract

Only about half of the 40-residue amyloid-β sequence is structurally ordered in a common fibril structure from Alzheimer’s disease brain.

Molecular structures within amyloid fibrils formed by amyloid-β peptides in vitro are not determined uniquely by their amino acid sequences. Instead Aβ40 and Aβ42 fibril structures are sensitive to subtle variations in growth conditions in de novo preparations.^1–3 Multiple distinct structures can also develop simultaneously under a single set of conditions.^4–7 Moreover, structural differences are self-propagating in seeded fibril growth, i.e., when new fibrils grow from fragments of pre-existing fibrils.^{1, 8} This phenomenon of molecular-level structural polymorphism in vitro raises the following questions^9–12: Are the Aβ40 and Aβ42 fibrils that develop in brain tissue of Alzheimer's disease (AD) patients similarly polymorphic? Are molecular structural variations in Aβ40 and Aβ42 fibrils correlated with variations in clinical and pathological characteristics of AD?

To address these questions, our lab developed methods for seeded growth of Aβ40 and Aβ42 fibrils that can be studied by solid state nuclear magnetic resonance (ssNMR) and transmission electron microscopy (TEM), using amyloid-containing cortical tissue of deceased AD patients as the source of seeds.^13–15 Experiments by Lu et al.¹⁵ found different predominant structures in Aβ40 fibrils derived from cortical tissue of two AD patients (patients 1 and 2), indicated by different fibril morphologies in TEM images and different crosspeak signal patterns in two-dimensional (2D) ssNMR spectra. The same predominant polymorph was found throughout the cerebral cortex of each patient. Interestingly, patient 1 had an atypical clinical history, while patient 2 had typical AD symptoms and history. Qiang et al. subsequently examined 37 cortical tissue samples from 18 patients, recording TEM images and 2D ssNMR spectra of Aβ40 and Aβ42 fibrils derived from these tissue samples by seeded growth.¹⁴ For Aβ40, a single predominant fibril polymorph was found in most samples from patients from two AD subtypes, namely typical long-duration AD and the posterior cortical atrophy variant of AD. Based on its "fingerprint" in 2D ssNMR, this Aβ40 polymorph was the same as the predominant polymorph from patient 2 of Lu et al. Thus, we have evidence that most AD patients develop a specific Aβ40 fibril polymorph in their cerebral cortex. However, the detailed molecular structure of this polymorph is not yet known.

As a first step towards the development of a full molecular structural model for this predominant Aβ40 fibril polymorph in AD, we prepared Aβ40 fibrils from seven synthetic peptides with ¹⁵N,¹³C-labeled residues at the specific positions shown in Fig. 1A. "First-generation" fibrils, prepared by seeding Aβ40-sl1 with cortical tissue extracts, were used in ssNMR measurements by Lu et al. and Qiang et al.^{14, 15} We used one of these tissue-seeded Aβ40-sl1 samples (identified as PCA3f in Table 1 of Qiang et al. and shown in Fig. 1B) as the source of seeds for "second-generation" fibrils, which were prepared with the remaining six labeling patterns (see Materials and Methods in ESI for experimental conditions). Second-generation Aβ40-sl6 fibrils were also used as the source of seeds for "third-generation" fibrils. Second- and third-generation fibrils appear morphologically homogeneous in TEM images with negative staining, with approximately 90% of visible fibrils showing periodic modulations in apparent width with periods approximately 130 ± 30 nm.

Fig. 1.

(A) Amino acid sequence of Aβ40 and positions of ¹⁵N,¹³C-labeled residues in the seven selectively labeled samples. (B) Negatively-stained TEM image of first-generation brain-derived Aβ40-sl1 fibrils, used as the source of seeds for second-generation samples. Cyan arrows indicate collagen fibers. (C,D) TEM images of second-generation Aβ40-sl5 and Aβ40-sl4 fibrils. (E) TEM image of third-generation Aβ40-sl6 fibrils, grown from second-generation Aβ40-sl6 fibril seeds. Red arrows indicate the periodicity of apparent width modulation for the most common morphology.

As shown in Fig. 1A, first-generation fibril samples contained a large proportion of extraneous material from the cortical tissue. This proportion was greatly reduced in subsequent generations, becoming nearly absent from TEM images. It is conceivable that the Aβ40 fibril structure could be stabilized or otherwise affected by the presence of certain chemical compounds from the tissue. 2D ¹³C-¹³C ssNMR spectra of second-generation Aβ40-sl2 fibrils and first-generation Aβ40-sl1 fibrils show good agreement of crosspeak signal patterns (see Fig. S1 of ESI), as do 2D¹³C-¹³C ssNMR spectra of third-generation and second-generation Aβ40-sl6 fibrils (see Fig. S2 of ESI). These results indicate faithful propagation of the molecular structure despite the progressive dilution of extraneous material in successive generations.

Figs. 2 and 3 show 2D ¹³C-¹³C and ¹⁵N-¹³C ssNMR spectra of the second-generation fibrils (see Materials and Methods of ESI for measurement conditions). Crosspeak signals were readily assigned, based on the known labeling patterns and residue-specific chemical shift ranges for labeled sites.¹⁶ When the same residue was labeled in more than one sample (e.g., I31 in Aβ40-sl2 and Aβ40-sl3), good agreement of chemical shifts and ssNMR line widths was observed. (Apparent differences in line widths for the same sites in the contour plots in Fig. 2 are due primarily to differences in signal-to-noise for 2D spectra of different samples.)

Fig. 2.

Two-dimensional 13C-13C spectra of selectively-labeled, second-generation Aβ40 fibrils, with color-coded resonance assignments. Broad crosspeaks are indicated by colored circles and ellipses. Contour levels increase by successive factors of 1.5.

Fig. 3.

Two-dimensional ¹⁵N-¹³C spectra of selectively-labeled, second-generation Aβ40 fibrils, with color-coded resonance assignments. Contour levels increase by successive factors of 1.5.

Importantly, signal intensities and line widths for different residues varied significantly. While certain residues contributed relatively strong, sharp crosspeaks to the 2D spectra (residues 19, 21, 23–25, 27–35), crosspeaks from other residues were weak and broad (residues 2, 7, 9, 12, 17, 20, 22, 26, 38, 39) or undetectable (residues 8 and 10). We attribute these large variations in properties of ssNMR signals to a coexistence of structurally ordered and disordered segments of the Aβ40 sequence within a single fibril. The morphological uniformity of the fibrils in our TEM images argues against an alternative interpretation that broad ssNMR signals arise from the presence of multiple distinct polymorphs in our samples, as does the reproducibility of ssNMR signals in multiple independent sample preparations.

In addition to measurements on samples with specific ¹⁵N,¹³C-labeled residues, we performed three-dimensional (3D) ssNMR measurements on a uniformly ¹⁵N,¹³C-labeled first-generation Aβ40 fibril sample, derived from cortical tissue of patient 2 of Lu et al.¹⁵ Strong crosspeaks in 3D NCACX and 3D NCOCX spectra (see Fig. 4 and Fig. S3 of ESI) were observed for residues 18–25 and 28–37, consistent with assignments from specifically labeled samples and providing additional assignments for V18, V36, and G37. The final set of assignments, together with site-specific line widths and predictions of backbone torsion angles from the TALOS-N program¹⁷, are given in Table S1 of ESI.

Fig 4.

Crosspeak heights in 3D ssNMR spectra of uniformly ¹⁵N,¹³C-labeled first-generation Aβ40 fibrils. Values are shown for N(i+1)/CO(i)/Cα (i) crosspeaks in the 3D NCOCX spectrum and N(i)/Cα(i)/Cβ(i) crosspeaks in the 3D NCACX spectrum, where i is the residue number. Values are shown only for residues with detectable, resolved crosspeaks.

In previous studies of Aβ40 and Aβ42 fibril structures by ssNMR and cryo-electron microscopy, certain polymorphs were found to be structurally ordered across the entire peptide sequence^{15, 18–20}, while other polymorphs were found to contain coexisting ordered and disordered segments.^{6, 8, 21–24} Naïvely, one might have expected that the predominant Aβ40 fibril polymorph in AD brain tissue would be highly ordered, based on the ideas that the predominant polymorph might be an especially stable one and that a high degree of structural order goes hand-in-hand with high stability. The 2D and 3D ssNMR spectra discussed above clearly contradict this expectation, with only about half of the Aβ40 sequence being highly ordered. One might then speculate that partial disorder facilitates the breakage, transport, and propagation of fibrils within brain tissue, so that the predominant fibril structures in AD brain tissue might be partially disordered. However, Aβ40 fibrils derived from patient 1 of Lu et al. (which we now know to be an uncommon polymorph, based on the results of Qiang et al.¹⁴) were found to be highly ordered, exhibiting strong, sharp crosspeaks in 2D and 3D ssNMR spectra from all residues except H14 and E22.¹⁵ Thus, the relationship between structural order and relevance to AD is not yet understood.

Finally, we note that the predominant fibril morphology for our brain-derived Aβ40 fibrils, shown in Figs. 1C, 1D, and 1E, is remarkably similar to the morphology of Aβ40 fibrils prepared in vitro by Paravastu et al., which have a width modulation period of 120 ± 20 nm.⁸ Moreover, both types of fibrils have mass-per-length values of approximately 27 kDa/nm.^{8, 15} However, ssNMR chemical shifts are significantly different (see Fig. S4A of ESI). We therefore anticipate that the molecular structure of our brain-derived Aβ40 fibrils resembles the structure of in vitro fibrils prepared by Paravastu et al., but is not the same. In contrast, ssNMR chemical shift differences relative to brain-derived Aβ40 fibrils from patient 1 of Lu et al.¹⁵ are somewhat larger (see Fig. S4B of ESI). Additional ssNMR data to provide experimental restraints on long-range inter-residue contacts will be required to enable development of a structural model for our brain-derived Aβ40 fibrils.