Electron tomography reveals the fibril structure and lipid interactions in amyloid deposits

Significance

Although considerable previous efforts have been dedicated to studying the molecular assembly of individual amyloid fibrils, much less is known about their 3D arrangement within a pathological deposit. In this study, we use electron tomography, an extremely powerful method for studying the detailed structure of cellular assemblies or macromolecular complexes, to unravel the superstructure of fibril networks. The structural views provided by our analysis enable a better understanding of the properties and pathogenic features of amyloid fibrils. The fibril network structure is also a crucial determinant of possible applications of such fibrils in the field of biotechnology or material sciences.

Abstract

Electron tomography is an increasingly powerful method to study the detailed architecture of macromolecular complexes or cellular structures. Applied to amyloid deposits formed in a cell culture model of systemic amyloid A amyloidosis, we could determine the structural morphology of the fibrils directly in the deposit. The deposited fibrils are arranged in different networks, and depending on the relative fibril orientation, we can distinguish between fibril meshworks, fibril bundles, and amyloid stars. These networks are frequently infiltrated by vesicular lipid inclusions that may originate from the death of the amyloid-forming cells. Our data support the role of nonfibril components for constructing fibril deposits and provide structural views of different types of lipid–fibril interactions.

Amyloid fibrils are fibrillary polypeptide aggregates with a cross-β conformation (1, 2) and define a group of debilitating human diseases that includes, besides Alzheimer’s disease and type II diabetes, different forms of systemic amyloidosis (1, 3). Considerable previous research has focused on the structure of individual fibrils and determined their β-strand conformation (1, 4), their global topology, and the position of the polypeptide chains within the fibril (5), but comparatively little is known about how fibrils are organized into a 3D amyloid deposit.

Amyloid fibrils make, specifically in systemic amyloidosis, substantial contributions to the disease process and physically impair the functionality of the affected tissue (6, 7). The way in which fibrils are arranged can vary between different forms of a disease, suggesting that the deposit structure may contribute to the disease process. In systemic transthyretin (ATTR) amyloidosis, for example, the deposit structure was found to correlate with the Congo red (CR) staining properties and tissue involvement in different variants of the disease (8). In Alzheimer’s disease, there is evidence that the amyloid β (Aβ) amyloid plaque structure in affected patients is histologically different from the Aβ plaques in the brains of nondemented individuals (9).

In this research, we have used electron tomography, an increasingly powerful method to characterize macromolecular assemblies and cellular structures at very high resolution (10), to investigate the network structure within an amyloid deposit. The analyzed amyloid deposits were formed by a cell culture model of systemic amyloid A (AA) amyloidosis (11–13), an amyloid disease that affects humans and other mammals, and birds (14). AA amyloid deposits occur in multiple organs and are frequently found in spleen, kidneys, and liver. AA amyloid fibrils consist of AA protein, a proteolytic fragment of serum amyloid A1 (SAA1) protein (14). This acute-phase protein is secreted by the liver in response to chronic inflammation and released into the blood, where it associates with high-density lipoprotein (HDL) particles. The cell model reproduces crucial features of fibril formation from SAA1 protein in vivo, such as the involvement of macrophages (11); the fragmentation of SAA1 protein (15); and the association of fibrils with secondary components, such as glycosaminoglycans (GAGs) and lipids (13, 16). Due to its ready accessibility, we use it here for further analysis with electron tomography.

Our study shows that electron microscopy is a powerful technique that is able to visualize the fibril network structure and the fibril morphology directly in the deposits. The deposited filaments can adopt multiple network structures, and they are frequently infiltrated by vesicular lipid inclusions.

Results

Formation of Amyloid-Like Fibrils from SAA1 Protein in Cell Culture.

Exposure of murine J774A.1 cells, a widely used model of macrophage and monocytic function, to acute-phase levels of murine SAA1 protein and HDL was previously shown to lead to the formation of extracellular amyloid deposits (13). The formed amyloid deposits stain with CR (Fig. S1A) and produce green birefringence in the polarizing microscope (Fig. S1B), which is the gold standard of amyloid diagnosis in pathology (2). By using scanning electron microscopy (SEM), we here find that the formed amyloid deposits consist of extracellular fibrils that can be immunogold-labeled via a primary antibody raised against full-length SAA1.1 protein (Fig. S1 C–E). Transmission electron microscopy (TEM) analysis of ultrathin sections further confirmed this finding and showed the immunogold labeling of the fibrils (Fig. S1F). The formed deposits comprise typical components of pathological AA amyloid deposits, and in addition to AA protein, they contain GAGs, which we demonstrate here with Alcian blue staining (Fig. S1G) and with antibody reacting with heparan sulfate (Fig. S1H), as well as a serum amyloid P (SAP) component (Fig. S1I).

Fig. S1.

Amyloid deposits contain SAA-derived fibrils and GAGs. Bright-field (A) and dark-field (B) microscopy images of cells incubated with 1 mg/mL SAA1 and HDL after incubation for 6 d, and CR staining. (C–E) SEM images of an amyloid deposit that was immunogold-labeled with a rabbit anti–full-length SAA1.1 primary antibody and a secondary antibody coupled with 10-nm gold particles. (C) Backscattered electron signal showing distribution of gold particles. (D) Secondary electron signal displays the fibril surfaces. (E) Merged image. (F) TEM image of a 70-nm section of a freeze-substituted amyloid deposit immunogold-labeled with anti–full-length SAA1.1 primary antibody. Light microscopic images of cells that were incubated with SAA1 and HDL, as indicated in the figure, for 7 d and then stained with Alcian blue (G) or with an anti-heparan sulfate primary antibody (H) are shown. (I) Coomassie-stained denaturing protein gel (Left) and Western blot (Right; anti-mouse SAP primary antibody) of fibrils extracted from cell culture and lysates from spleen tissue of an AA amyloid-laden mouse. The black arrow indicates the expected size of glycosylated SAP.

Fibrils extracted from the deposits (Fig. S2A) consist of partially fragmented SAA1 protein (Fig. S2B), resembling the C-terminal truncation of SAA1 protein in AA amyloidosis (17). Similar C-terminal truncation was reported previously to occur if primary murine macrophages are exposed to SAA1 protein in vitro (15). The formed fibrils exhibit bona fide amyloid characteristics and give rise to X-ray reflections at 4.74 ± 0.03 Å and 10.7 ± 0.1 Å (Fig. S3A) that demonstrate the presence of a cross–β-sheet conformation. Extracted fibrils bind amyloid binding dyes, such as thioflavin T (Fig. S3B) and CR (Fig. S2C), and produce CR green birefringence in the polarizing microscope (Fig. S2C), similar to the properties of the amyloid deposits before fibril extraction (Fig. S1 A and B). This structure differs from globular SAA1 and SAA3 proteins (18, 19), which are significantly α-helical and devoid of β-sheet structure (Fig. S3C). These globular proteins exhibit an α-helical content of 70–74% as determined by the program DSSP, whereas attenuated total reflectance Fourier-transform infrared spectroscopy and deconvolution of the amide I spectral region reveal a β-sheet content of the presently analyzed fibrils of 73% (Fig. S3D). These data show that the protein has undergone a drastic conformational change as it converted into the fibril state.

Fig. S2.

Structural characteristics of fibrils extracted from cell culture. (A) TEM image of fibrils extracted from cell culture. (B) Coomassie-stained denaturing protein gel (Left) and Western blot [Right; anti-AA(1–76) primary antibody] of fibrils extracted from cell culture and freshly dissolved recombinant SAA1 protein. (C) Bright-field (Left) and dark-field (Right) polarizing microscopy images of extracted fibrils stained with CR.

Fig. S3.

Amyloid properties and secondary structural analysis. (A) X-ray diffraction pattern of fibrils extracted from cell culture shows reflections at 4.74 ± 0.03 Å and 10.7 ± 0.1 Å. (B) Thioflavin T (ThT) fluorescence spectrum with (red) or without (green) cell culture-derived fibrils. (C) Ribbon diagrams of the globular states of human SAA1.1 [green; Protein Data Bank (PDB) ID code 4IP9] and murine SAA3 protein (blue; PDB ID code 4Q5G) as taken from available crystal structures (18, 19) and superimposed at an all-atom root mean square deviation of 0.51 Å. (D) Amide I and II regions of the attenuated total reflectance Fourier-transform infrared spectroscopy (ATR-FTIR) spectrum of cell culture-derived fibrils. Black circles illustrate the measured spectrum, gray illustrates components of the fit, and red illustrates the sum of the fit, closely overlapping with the original spectrum. The secondary structure of cell culture-derived fibrils was obtained by deconvolution of the amide I region of the ATR-FTIR spectrum. Assignments of the components of the amide I region: 1,626 cm⁻¹ (β = 68%), 1,656 cm⁻¹ (α = 27%), 1,681 cm⁻¹ (β = 5%).

Cell Culture Fibrils Are Toxic to Neighboring Cells.

Addition of fibrils extracted from the cell culture to a fresh culture of J774A.1 cells induces cellular toxicity and metabolic changes, as demonstrated by the 2,5-diphenyltetrazolium bromide (MTT) assay (Fig. S4A), as well as apoptosis, as shown by increased caspase 3/7 activity (Fig. S4B). These effects were measured with cells kept without HDL-SAA1; that is, the cells do not themselves form amyloid, and the strength of the effect depends on the fibril concentration applied to the cells. Analysis of the mechanism underlying this effect with confocal laser-scanning microscopy shows that SAA1 fibrils from the cell culture that were fluorescently labeled with Alexa Fluor 488 dye are avidly taken up by the cells, where they accumulate at substantial levels (Fig. S4 C and D). Uptake occurs mainly via phagocytosis (Fig. S4E) and leads to the accumulation of fibrils in the endocytic pathway, where they cause lysosomal leakage and the disruption of intracellular membranes, as shown by a flow cytometric analysis of cells that had additionally internalized Acridine Orange (Fig. S4F). This dye shows a pH-dependent reduction of its fluorescence intensity as soon as it exits the lysosomes into the cytoplasm.

Fig. S4.

Cell culture-derived fibrils are toxic to cells that do not form amyloid themselves. (A) MTT-based cell viability measurement of cells incubated with or without cell culture-derived SAA1 fibrils for 24 h. The sample without fibrils was set to 100% (n = 4; **P < 0.01). (B) Apoptosis induction measured by caspase-3/7 up-regulation in cells exposed to cell culture-derived fibrils for 8 h. The untreated sample was set to 100% (n = 3; **P < 0.01). (C) Laser-scanning microscopy (LSM) image of trypsinized cells after incubation with 50 μg/mL AF488-labeled cell culture-derived fibrils for 24 h. (D) Flow cytometric analysis of trypsinized cells exposed to different concentrations of AF488-labeled cell culture-derived fibrils (n = 4). (E) LSM analysis of trypsinized cells incubated with 50 μg/mL AF488-labeled cell culture fibrils and an uptake marker for phagocytosis (pHrodo Red Zymosan A BioParticles Conjugate) for 3 h. (F) Flow cytometric analysis of AO-stained cells incubated with or without cell culture-derived fibrils for 24 h. Gating was used to assess the percentage of cells with low AO fluorescence, exhibiting lysosomal leakage (n = 4; **P < 0.01). a.u., arbitrary units.

Electron Tomography Reveals the Deposit to Consist of Multiple Fibril Network Structures.

We then prepared freeze-substituted specimens of the fibril deposits to cut out sections that we analyzed with scanning TEM at 300 kV. Based on image series with incrementally varied tilt angles, we computed the 3D tomograms of the analyzed amyloid deposits. The tomograms show well-resolved fibrils that could be tracked through multiple virtual sections (Fig. 1). The fibrils are arranged into networks that exhibit significant local order. We can discern three major types of networks that we term here the fibril meshwork, fibril bundle, and amyloid star. Fibril meshworks present no preferential overall orientation of the constituting filaments, whereas fibrils in a bundle are significantly aligned in parallel. An amyloid star consists of fibrils that radiate out in different x/y directions. However, analysis of different horizontal planes of the tomogram cannot reveal well-defined star core and the star represents a stack of fibril bundles with different orientations relative to each other (Fig. S5). The three types of network structures usually co-occur within the same amyloid deposit (Fig. S6).

Fig. 1.

Electron tomograms showing different fibril network structures. Fibril meshwork (Top), fibril bundle (Middle), and amyloid star (Bottom). (Left) Two-dimensional projections of a 500-nm-thick slice of a deposit. (Center) Virtual sections through the tomograms; thickness of the virtual sections: fibril meshwork = 1.29 nm, fibril bundle = 1.29 nm, and amyloid star = 1.84 nm. (Right) Three-dimensional model. Fibrils are shown in blue, and gold particles used for image alignment are shown as yellow arrows.

Fig. S5.

Bundle substructure of an amyloid star. Rendered fibrils of a tomogram of an amyloid star (Fig. 1 G–I). (A) Different bundles constituting the star are indicated in different colors. Total size: 1884 × 1884 × 460 nm. (B) Sections of the tomogram at different z-axis positions as indicated. (C) Direction of fibril segments was computationally analyzed in two dimensions and shown as a circular plot such that density values are specified by the length of the corresponding blue lines. (D) The 3D distribution of the directions of fibril segments is visualized using the Lambert azimuthal equal-area projection of a sphere onto a disk (52) with color-coded density values.

Fig. S6.

Deposit can comprise multiple network structures. TEM image of a 500-nm-thick section from a freeze-substituted deposit. Three network substructures can be discerned: (1) fibril meshwork, (2) fibril bundle, and (3) fibril star. The section was plasma-cleaned with an Edwards plasma cleaning system immediately before TEM.

Morphological Analysis of the Fibrils Within the Deposit.

We then measured, within our tomograms, the distribution of the fibril width W and the persistence length P, which reflect the bending propensity of the filament (Fig. 2A). The obtained values of W present a roughly bell-shaped distribution, centered at 11–12 nm (Fig. 2A). We obtained relatively similar distributions of W for the fibril meshwork, the fibril bundle, and the amyloid star (Fig. 2A). Also, the parameter P shows a very similar distribution for the fibrils in the three deposit structures (Fig. 2A), which, taken together with the similarity of W, suggests that the amyloid star, fibril bundle, and meshwork are constructed from morphologically similar fibrils.

Fig. 2.

Fibril morphology in the amyloid deposit. (A) Distribution of the values of fibril width (W) and persistence length (P) obtained from the tomograms of fibril deposits and by conventional TEM of fibrils extracted from the cell culture or murine AA amyloidotic spleen or of fibrils grown from pure SAA1 protein in vitro. (A, Insets) Close-up view of the region of P < 1 μm. (B–D) TEM images of fibrils extracted from cell culture (B) and from mouse spleen (C) and of fibrils that were formed from pure SAA1 protein in vitro (D) are shown.

The two properties W and P could also be measured with fibrils that were extracted from the cell culture; immobilized onto a formvar-carbon–coated grid; negatively stained; and viewed by conventional TEM techniques, that is, without using tomography (Fig. 2B). These measurements show a slightly higher average value of W than in the tomography-based measurements (Fig. 2A), presumably arising from the lateral association of counterstaining agent, whereas the distribution of the values of P corresponds well to the measurements performed on the fibrils in the deposit (Fig. 2A). By negative-stain TEM, we could also analyze that fibril morphology of filaments extracted from murine amyloidotic spleen (Fig. 2C) and the distribution of values for P and W resemble the distribution of values of cell culture fibrils (Fig. 2A). By contrast, we found much more pronounced differences when we compared cell culture fibrils with fibrils grown from pure SAA1 protein in simple phosphate buffer in vitro (Fig. 2D and Fig. S7B). These in vitro fibrils were formed according to a previously published protocol (20). They exhibit a relatively curvilinear structure in the original TEM images (Fig. 2D) and give rise to much smaller P values than cell culture-derived fibrils and AA amyloid fibrils (Fig. 2A).

Fig. S7.

Fibril formation in vitro at physiological conditions or in phosphate buffer. (Left) Negative-stain TEM images of SAA1 protein that has been incubated in vitro for 4, 8, and 12 d in PBS buffer at pH 7.4 (A) or phosphate buffer at pH 3.0 (B). (Right) SAA1 protein incubated after addition of 5% (vol/vol) extracted fibrils from mouse spleen at pH 7.4 (A) or pH 3.0 (B).

We also tested other conditions of fibril formation in vitro, such as incubation of SAA1 protein at pH 7.4. However, the fibril yield obtained under this condition was much smaller than at pH 3.0, as judged from the amount of material visible on the TEM grid (Fig. S7). Seeding with 5% (vol/vol) murine ex vivo AA amyloid fibrils slightly increased the number of visible fibrils, but the amount of fibrous material attached to the grid remained low. The observed fibrils were relatively long and straight and resembled in their morphologies the AA fibrils that we used as seeds.

Amyloid Deposits Comprise Vesicular Lipid Inclusions.

The formed deposits were frequently infiltrated by vesicular lipid inclusions that possess a hollow architecture and the typical double-layer structure of lipid membranes (Fig. S8 A–D). The thickness of the bilayers (7.3 ± 0.7 nm; Fig. S8E) is also consistent with literature values for biological membranes (21–23). Lipids represent, besides SAP and GAGs, an additional class of nonfibril components of disease-associated, extracellular amyloid deposits (16, 24) and were previously shown to interact with many amyloid-forming polypeptide chains (25), to promote fibril formation in vitro (25), and to mediate the amyloid-dependent enhanced infectivity of HIV-1 to cultured cells (26).

Fig. S8.

Lipid bilayer of vesicular inclusions in amyloid deposits. (A–D) Virtual sections of tomograms with different vesicular inclusions revealing lipid bilayers. Lipid bilayers are shown for the lipid inclusions on varying sections of the tomograms. The z axis of the virtual sections shown in A–D is increasing from left to right in the images. The values of the distance (∆Z) between the shown consecutive sections are 9 nm (A), 4 nm (B), 7 nm (C), and 9 nm (D). (E) Distribution of lipid bilayer width of vesicular lipid inclusions obtained from tomograms with vesicular inclusions. The width of 40 lipid bilayers was measured using the GNU Image Manipulation Program 2 software (version 2.8.14).

Our tomograms now reveal that the lipid inclusions exhibit a vesicular organization with diameters ranging from 20 nm to sometimes over 500 nm. The overall guise of these inclusions encompassed elongated tubular networks to spherical structures and multivesicular assemblies. Some exhibited a particularly electron dense structure, resembling apoptotic bodies (Fig. 3 A–D), which suggests that the inclusions originate from the death of the amyloid-producing cells. Indeed, caspase activity and MTT assay constituently demonstrate fibril formation to be toxic to the cells in culture. We also observe a disintegration of the plasma membrane of amyloid-producing cells with SEM (Fig. S9). The vesicular lipid inclusions are also seen if amyloid deposits were formed by cells that were kept under serum-free conditions and in the absence of HDL, demonstrating that the cells are a cause of the encountered lipid structures (Fig. S10).

Fig. 3.

Vesicular lipid inclusions and fibril–lipid interactions in amyloid deposits. (A) Virtual section through a tomogram of an amyloid deposit with extensive lipid inclusions. The green arrow points to a densely filled apoptotic body, the yellow arrow points to a multivesicular structure, and the magenta arrow points to a vesicle with low-electron-dense content. A 2D projection (B), virtual section (C), and 3D model (D) of an amyloid bundle show extensive lipid inclusions. (E–J) Tomogram of a deposit with lipid inclusions. Virtual sections (E and H) and regions of the 3D model (F, G, I, and J) are shown. (E–G) Lipid interaction via the fibril tip, distorting the lipid bilayer and producing a notable hump at the interaction site. (H–J) Lipid interaction via the lateral surface of the fibril. (K) Series of virtual sections showing a vesicular inclusion where the inclusion displaces a fibril bundle without making direct lateral contacts to the filaments. The figures in G and J are rotated by 90° relative to F and I.

Fig. S9.

Cellular toxicity of amyloid deposit formation. (A) Apoptosis induction measured by caspase-3/7 up-regulation in amyloid-forming cells exposed to SAA1 and HDL for 2 d (orange) and in non–amyloid-forming cells that were incubated for 2 d without HDL/SAA1 (black). (B) MTT assay of amyloid-forming cells exposed to SAA1 and HDL for 6 d (orange) and in non–amyloid-forming cells that were incubated for 6 d without HDL/SAA1 (black). Untreated samples were set to 100% in both panels (n = 4; **P < 0.01). (C) SEM image of an amyloid-forming cell exposed to SAA1 and HDL for 6 d showing no membrane ruffles; the surface is disintegrated and exhibits holes and a porous appearance. (D) SEM image of a nonamyloidogenic cell that was cultured for 6 d without HDL/SAA1.

Fig. S10.

Lipid vesicles in fibril deposits formed in cell culture under serum-free conditions. A TEM image of a 70-nm ultrathin section shows amyloid deposits derived from cells cultured with SAA1 in Dulbecco’s modified Eagle’s medium without FBS for 3 d. Vesicular lipid inclusions can be identified within the amyloid deposits.

We can discern, in the recorded tomograms, two modes by which fibrils interact with the lipids. These modes depend on whether fibrils contact the lipid bilayers through their tips (Fig. 3 E–G) or through their lateral surfaces (Fig. 3 H–J). Interactions with the fibril tip can lead to a significant distortion of the membrane structure at the contact point (Fig. 3 E–G). These interactions are also more abundant and occur with 6% of the fibril tips in lipid-rich tomograms, whereas only 3% of the cumulative length of the same fibrils in the same tomogram is in lateral contact with the lipid membranes. In addition, we frequently saw fibrils form a cage around the lipid inclusion (Fig. 3K), apparently avoiding direct interactions and suggesting that lateral fibril–lipid interactions are more incidental in our system or restricted to certain membrane regions only. In our tomograms, we almost never saw fibrils that penetrated through a lipid bilayer.

Discussion

In this study, we have used electron tomography to investigate the deposit structure of the fibrils formed by a cell culture model of AA amyloidosis. Our analysis reveals three types of network structures that we term fibril meshwork, fibril bundle, and amyloid star. These data are consistent with previous analyses of histological tissue sections that used topo-optical reactions or conventional TEM methods that solely reported 2D projections of a tissue slice but did not generate 3D structural data. Such 2D data have been reported, for example, for Aβ fibrils in Alzheimer’s disease and for transthyretin-derived fibrils in ATTR amyloidosis, as well as for fibril deposits formed in other types of systemic amyloidosis, notably including AA. The observed features are consistent with star-like assemblies (27–29) and fibril bundles (8, 30–32), as well as with fibril meshwork structures (8). Based on the currently used tomographic techniques, 3D structural views of the respective network structures can now be provided.

The investigated fibrils are toxic when added to other cells (Fig. S4), and all the above examples of similarly structured amyloid deposits refer to pathogenic amyloids within the extracellular space. It is thus possible that intracellular or cytoplasmic cross-β filaments may have a different structural organization, specifically if they are of functional relevance to their host cells. Examples are the fibrils formed from sup35 or pml17 protein. Sup35-derived fibrils play a role in phenotypic inheritance in yeast cells (33) and were found to adopt ring-like assemblies underneath the plasma membrane or dot-like cytoplasmic inclusions with a regular mosaic structure (33). Pml17 fibrils are important for melanosome formation, and tomographic studies revealed intracellular sheet like-assemblies that consist of parallel fibrils of pml17 protein (34). The different network structures of these intracellular filaments compared with the presently investigated extracellular amyloids raise the possibility that intracellular networks are influenced by one or several intracellular factors modifying their assembly.

Such modifying factors may also exist for the extracellular fibril networks because the fibrils structuring the presently analyzed deposits do not show a random orientation but structural order in terms of a fibril bundle or an amyloid star. The factors responsible for these effects could arise from the environment in which fibril formation takes place, for example, if fibrils form within a cavity or under shear flow conditions that induce a preferential orientation of the formed filaments (24, 35). Alternatively, it is possible that fibrils interact with molecules like GAGs or lipids that exhibit regular surfaces and were shown to affect the nucleation and growth of amyloid-like fibrils in vitro (36, 37). However, lipid vesicles may also form obstacles that sterically restrict the room in which fibril growth can take place, and several of the presently analyzed fibril bundles have apparently grown around a preexisting vesicle (Fig. 3K).

A second possible source of ordering factors is the fibrillation mechanism. For example, if fibrils become nucleated on the surface of a preexisting fibril (38), it is possible that the seed partially orients the daughter filament. Furthermore, total internal reflection fluorescence microscopy revealed the outgrowth of amyloid-like fibrils in vitro and the formation of fibril bundles due to fibril branching reactions (39) or the formation of star-like spherulites due to a radial extension of fibrils from a seed (40, 41).

A particularly interesting feature of the recorded tomograms is the association of the fibril deposits with lipids. Lipids have long been known to be present in pathological amyloid deposits (16, 24). Thin layer chromatography previously established their molecular composition and enrichment in cholesterol, cholesterol esters, and sphingolipids (16), whereas conventional TEM analyses reported the presence of spherical lipid inclusions in histological sections of amyloid-laden tissues (24), reminiscent of the vesicular structures seen here by tomography.

Lipid inclusions may interact with the fibrils, and respective interactions occur either via the fibril tip or through the lateral fibril surface. We find here that the interactions via the fibril tip are more abundant and strong enough to distort the lipid bilayer structure locally (Fig. 3 E–G). They possibly involve hydrophobic interactions between the fatty acid tails and the relatively hydrophobic fibril core that is exposed at the fibril tip, as suggested by a recent structural model of Aβ(1–42) fibrils (42). Interactions of the tips of in vitro-formed β2-microglobulin fibrils were previously observed with artificial liposomes and found to perturb the structural integrity of their lipid bilayer structures (43), one possible toxic mechanism of amyloid fibrils (25, 43). Interactions of the lateral fibril surface with the lipid bilayer are likely to involve greater contributions from polar or electrostatic groups and may underlie the binding of cross-β fibrils to lipid-enveloped viral particles in the course of viral infection enhancements (26). The present study, as well as the previous β2-microglobulin study, can detect such lateral interactions, but they are particularly rare in the present case.

Having established electron tomography as a tool to investigate the structure of disease-associated amyloid deposits, it is now possible to apply this method to other biological samples, such as histological sections, and to explore the 3D structure and pathogenic effects of in vivo-formed amyloid deposits on their neighboring tissues. Moreover, cross-β fibrils have recently given rise to biotechnological utilities, such as hydrogels, cell growth support structures, filtration units for purification, and enhancers of viral infections (26, 44–46). These applications may also depend on the fibril network structure, and through the presently described techniques, it is now possible to characterize the 3D structure of these assemblies as a prerequisite to fine-tuning their functions.

Materials and Methods

Growth of Amyloid Deposits in Cell Culture.

J774A.1 cells (Sigma–Aldrich) were seeded in a 96-well plate (Greiner Bio-One) at a concentration of 350,000 cells per milliliter. The wells contained either sapphire disks (3 mm; M. Wohlwend GmbH) if samples were dedicated for SEM or TEM or glass coverslips if samples were produced for light microscopy (Thermo Fisher Scientific). Sapphire disks were pretreated by coating with a 20-nm-thick carbon layer using a Balzers BAF 300 (Bal-Tec) instrument. The coated disks were dried overnight in an oven (120 °C) and sterilized by UV irradiation at 320 nm for 10 min immediately before use. The cultured cells were always cultivated in an atmosphere containing 5% (vol/vol) CO₂ at 37 °C and in Dulbecco’s modified Eagle’s medium (Life Technologies) that was supplemented with 10% (vol/vol) of heat-inactivated FBS (Life Technologies) and 1% (vol/vol) of Antibiotic-Antimycotic (Life Technologies). To induce the formation of amyloid deposits, we added aliquots from stock solutions of SAA1 and HDL to the cell culture medium. We exchanged the cell culture medium, together with HDL and SAA1 if applicable, every 24 h or 48 h until the end of incubation.

The SAA1 stock was prepared by dissolving lyophilized murine SAA1.1 in distilled water at a concentration of 10 mg/mL. Residual trifluoroacetate was removed by washing the protein twice with pure water using a 3-kDa membrane filter (Amicon Ultra-0.5 mL 3K; Merck Millipore) that was centrifuged for 10 min at 16,900 × g. The sample remaining in the membrane filter was filled up with pure water to adjust the protein concentration to 10 mg/mL Elution of the protein from the membrane was done by inversion of the filter and centrifugation for 1 min at 100 × g. Stock SAA1 solution was added to the culture to reach a final SAA1 concentration of 1 mg/mL. The stock solution of human HDL was obtained commercially (AppliChem). The amount of HDL stock that we added to the culture depended on the HDL concentration that was provided by the vendor based on the triglyceride content of the sample. Hence, the amount of HDL stock that we added was adjusted such that the final concentration of triglycerides in milligrams per milliliter was 9% (wt/vol) of the SAA1 protein concentration in milligrams per milliliter that we added to the culture supernatant.

Preparation of TEM Specimens Embedded in Epoxy Resin.

Sapphire disks were removed from 96-well plates and plunged into 95% (vol/vol) 1-hexadecene (Sigma–Aldrich). Excess solution was removed via touching onto filter paper. Two sapphire disks were stacked up face-to-face, separated by a gold ring (diameter = 3.05 mm, central bore = 2 mm; Plano), mounted in a holder (Engineering Office, M. Wohlwend GmbH), and inserted into a Wohlwend HPF Compact 01 high-pressure freezer (Engineering Office, M. Wohlwend GmbH). The stack was disassembled with a set of pliers, and each sapphire disk was transferred into a separate precooled (−87 °C) 1.5-mL sample tube (Eppendorf). To each tube we added 1 mL of freeze substitution solution, which consisted of 0.2% (wt/vol) osmium tetroxide, 0.1% (wt/vol) uranyl acetate, and 5% (vol/vol) distilled water in acetone. The tubes were warmed up to 0 °C over a period of 24 h. Finally, they were left to equilibrate at room temperature for 1 h. The solution was removed, and the samples were washed with 1 mL of 100% (vol/vol) acetone. Each sample was incubated in an ascending epoxy resin (Fluka) series [33%, 50%, or 66% (vol/vol) epoxy resin in acetone]. Each incubation step (1 mL) lasted for 1 h at room temperature. The solution was replaced with 1 mL of pure epoxy resin and incubated for 24 h at room temperature. Afterwards, the sample was transferred into a new sample tube (Eppendorf) containing 0.25 mL of pure epoxy resin and heated to 60 °C for 24 h to polymerize the resin. The resin blocks were stored at room temperature.

TEM.

The analysis of thin sections without acquisition of tilt angle series and of extracted fibrils on formvar-carbon–coated copper grids (Plano) and counterstained with 2% (wt/vol) uranyl acetate grids was done at 120 kV using a JEM-1400 electron microscope (Jeol) equipped with a VELETA 2,000 × 2,000 side-mounted TEM camera (Olympus). Sections were mounted onto a 200 mesh, quadratic formvar-carbon–coated grid or on 300 mesh parallel copper grids (Plano) counterstained with 0.3% (wt/vol) lead citrate in water.

Scanning Transmission Electron Tomography.

From the resin block, we cut 500-nm-thick sections with a microtome (Leica Ultracut UCT ultramicrotome) using a diamond knife (Diatome). The resin block was cut in the direction parallel to the plane of the sapphire disk, and the slice was placed onto a copper grid (300 mesh parallel) that was plasma-cleaned with an Edwards plasma cleaning system and dried for 10 min at room temperature. A droplet of 10% (wt/vol) poly-l-lysine (Sigma–Aldrich) in water was added onto the grid holding the slice and dried for 5 min on a heating block (37 °C). Afterward, we put 15 μL of a solution of 15-nm colloidal gold particles (Aurion) that was diluted 1:1 with water onto each side of the grid. The gold particles serve as markers for image alignment in the processing of the electron tomograms. Finally, the grid holding the slice was coated on each side with a 5-nm carbon layer using a Balzers BAF 300 electron beam evaporation apparatus. The grid was plasma-cleaned with a Solarus Model 950 Advanced Plasma Cleaning System (Gatan) for 10 s immediately before electron microscopy. Image series at different tilt angles were recorded with a FEI-300 kV Titan scanning transmission electron microscope (FEI) operated at an acceleration voltage of 300 kV. Electron micrographs were recorded with a bright-field detector (FEI) at a size of 1,024 × 1,024 pixels. A single-axis tilt-series was recorded from −72° to +72° using tilt increments of 1°. Hence, the final image series consisted of 145 original images. Each image was acquired using an exposure time of 18 s and a convergence angle of 0.58 mrad. From the acquired tilt angle series, we reconstructed the tomograms and produced a 3D model using the IMOD software package (version 4.7.6); that is, the individual images were first aligned to form an image stack. In the second step, the tomogram was computationally reconstructed using a weighted back-projection algorithm. The 3D model was generated by tracing the fibrils and lipid vesicles manually through the different virtual sections of the tomogram.

Measurement of the Fibril Width, Contour Length, and Persistence Length.

The fibril width W, the contour length L, and the end-to-end distance R were determined from negative-stain TEM images of 500 cell culture-derived fibrils, 500 AA amyloid fibrils that were extracted from murine spleen, and 500 amyloid-like fibrils formed from murine SAA1 in vitro. Measurements were carried out using iTEM software (Olympus). The persistence length P was calculated from L and R using Eq. 1, assuming that the fibrils were deposited in a 2D manner on the grid surface in an energetically equilibrated conformation:

$${\left(R\right)}^2=4PL*\left[1-\frac{2P}{L}*\left(1-e^{\left(\frac{-L}{2P}\right)}\right)\right].$$ [1]

The values of W in the tomograms were measured for 250 fibrils per deposit type by analysis of the virtual sections using GNU Image Manipulation Program 2 software (version 2.8.14). In addition, P has been calculated for all fibrils contained in the 3D models using Eq. 2. Because Eqs. 1 and 2 cannot be solved for P analytically, the solution has been approximated numerically using Newton’s method (47), with a constant initial value of 1 and a target accuracy of 10⁻⁷:

$${\left(R\right)}^2=2PL*\left[1-\frac{P}{L}*\left(1-e^{\left(\frac{-L}{P}\right)}\right)\right].$$ [2]

Induction of AA-Amyloid in Mice.

AA amyloidosis was induced in a female 6-wk-old NMRI mouse as described (48). Animal experiments were conducted based on permission from the Thüringer Landesamt für Lebensmittelsicherheit und Verbraucherschutz Abteilung Gesundheitlicher Verbraucherschutz, Veterinärwesen, Pharmazie (registration no. 03-010/12). The animal was anesthetized with 480 mg/kg Ketamin (Inresa) and 64 mg/kg Rompun (Bayer) or 600 mg/kg isoflurane (Forene; Abbott) and killed by cervical dislocation, and the spleen was removed and stored at −80 °C or −20 °C.

SI Materials and Methods

Recombinant Expression and Purification of SAA1 Protein.

Murine SAA1.1 protein was recombinantly expressed in Escherichia coli RV308 as described previously (16). In brief, the coding region of murine SAA1.1 was cloned to the C terminus of a His-tagged maltose-binding protein in a pMAL-c2X vector (New England Biolabs) separated by a cleavage site for tobacco etch virus protease. Protein purification was done in five steps: (i) amylose resin high flow (New England Biolabs), (ii) nickel-Sepharose fast flow (GE Healthcare) chromatography, (iii) fusion protein cleavage by overnight incubation with tobacco etch virus protease at 34 °C, (iv) nickel chelate chromatography to separate SAA from the fusion protein and maltose-binding protein, and (v) Source 15 RPC (GE Healthcare) reversed-phase chromatography. The purified protein was lyophilized using an alpha 2-4 LD plus freeze dryer (Christ).

In Vitro Fibrillation of SAA1 Protein.

SAA1 was incubated at 1 mg/mL in 50 mM sodium phosphate buffer, pH 3.0, for 12 d at 37 °C.

SAA1 Binding Antibodies and Sera.

We used rabbit serum against mouse AA(1–76), batch 138 25/5-93 (49) or polyclonal rabbit antibody that we obtained commercially (Pineda) by injection of full-length, recombinant murine SAA1.1 protein. After 160 d, the antiserum was obtained, and the anti-SAA antibody was purified from the serum with SAA1-immobilized Sepharose 4B.

Purification of Fibrils from J774A.1 Cell Culture.

To obtain large quantities of amyloid, we cultured J774A.1 cells in a six-well plate in the presence of 1 mg/mL SAA1 and HDL over a period of 7 d. After incubation, the medium in each well was replaced with 2 mL of water and the amyloid deposits were scraped off with a Techno Plastic Products cell scraper. The 2 mL of water was transferred into a 50-mL tube (Greiner Bio-One), and the extraction cycle was repeated once such that we obtained 4 mL of amyloid suspension per well in the 50-mL tube. Next, we added 4 mL of homogenization buffer [20 mM Tris(hydroxymethyl)aminomethane (Tris) pH 7.4, 150 mM sodium chloride, 3 mM EDTA] to the 4-mL amyloid suspension and centrifuged the sample at 16,000 × g for 30 min at 4 °C with an Avanti J-26 XP centrifuge (Beckman Coulter) using a JLA-16250 rotor (Beckman Coulter). The supernatant was removed, and the pellet was suspended in 8 mL of homogenization buffer and centrifuged again at 16,000 × g for 30 min at 4 °C. The supernatant was discarded, and the pellet was suspended once more in 8 mL of homogenization buffer and centrifuged again using the same conditions as described before. The supernatant was discarded, and the pellet was finally resuspended in 0.5 mL of water to yield the fibril extract, which was stored at 4 °C until use.

Purification of Fibrils from Mouse Spleen Tissue.

Eighty milligrams of frozen spleen tissue was kept on ice and diced into pieces using a scalpel (Braun). The diced tissue was transferred to a 1.5-mL sample tube (Eppendorf) to which 0.5 mL of Tris calcium buffer [TCB; 20 mM Tris, 138 mM NaCl, 2 mM CaCl₂, 0.1% (wt/vol) NaN₃ (pH 8.0)] was added. The sample was mixed by gently flipping the tube and centrifuged at 800 × g for 5 min at 4 °C with a 6418R centrifuge (Eppendorf). The supernatant was discarded. The pellet was resuspended in 0.5 mL of TCB, mixed, and centrifuged as before. This cycle was repeated four more times. After the last centrifugation step, the pellet was resuspended in 0.5 mL of freshly prepared collagenase/protease inhibitor solution [one protease inhibitor EDTA-free tablet (Roche) in 7 mL of TCB, 2 mg/mL crude collagenase from Clostridium histolyticum (Sigma)] and incubated for 3 h at 37 °C under constant agitation (using an IKA MTS 2/4 digital table shaker at 250 rpm, horizontal position of the tube). Afterward, the sample was centrifuged at 800 × g for 30 min at 4 °C. The supernatant was discarded, and the pellet was washed twice by addition of 0.5 mL of ice-cold Tris EDTA buffer [20 mM Tris, 140 mM NaCl, 10 mM EDTA, 0.1% (wt/vol) NaN₃ (pH 8.0)] and centrifugation at 800 × g for 15 min at 4 °C. The samples were then homogenized in 250 μL of water with a pellet pestle (Kontes) that we used in a pulsative manner (1 s on, 1 s off) for a total of 10 s. The homogenate was centrifuged at 800 × g for 10 min at 4 °C. The supernatant was removed, and the pellet was resuspended again in 250 μL of water. The homogenization cycle was repeated five more times. The supernatants from the six homogenization steps were kept at 4 °C.

Preparation of Lysates from Mouse Spleen Tissue.

One milligram of spleen tissue was kept on ice and washed by dropping the tissue several times into 1 mL of PBS [137 mM sodium chloride, 2.7 mM potassium chloride, 8 mM di-sodium hydrogen phosphate, 2 mM potassium dihydrogen phosphate (pH 7.4); Thermo Fisher Scientific]. The tissue was then homogenized with a pellet pestle that we used in a pulsative manner (1 s on, 1 s off) for a total of 20 s. We added 50 μL of radioimmunoprecipitation assay buffer [1% (vol/vol) Nonidet P-40 (Roche), 0.5% (wt/vol) sodium desoxycholate, 0.1% (wt/vol) SDS in PBS] supplemented with 0.02% (vol/vol) phosphatase inhibitor [6.25 mM sodium fluoride (Sigma–Aldrich), 12.5 mM β-glycerophosphate (Calbiochem), 1.25 mM sodium orthovanadate, 12.5 mM p-nitropheny-phosphate disodium hexahydrate (Sigma–Aldrich)] and 0.01% (wt/vol) protease inhibitor (Calbiochem). The sample was incubated for a total of 30 min, during which time we kept the sample on ice and vortexed it twice for 15 s before and after we kept the sample in an ultrasonic bath at 4 °C for 15 min. The sample was finally centrifuged at 16,000 × g for 5 min at 4 °C, and the supernatant was immediately analyzed with reduced denaturing protein gel electrophoresis.

Unspecific N-Fluorescent Labeling of Cell Culture-Derived Fibrils.

Seventy microliters of cell culture-derived fibrils (15 mg/mL) was diluted in 180 μL of sodium carbonate buffer (140 mM, pH 8). We added 25 μL of Alexa Fluor 488 dye (AF488) and succinimidyl ester (4 mg/mL; Life Technologies) in dimethyl sulfoxide and incubated the solution under continuous shaking (using an IKA MTS 2/4 digital table shaker at 300 rpm) at room temperature for 1 h. We stopped the reaction by adding 100 μL of 1.5 M hydroxyl amine (pH 8.5) and dialyzed the sample three times for 2 h against 1 L of water with a Slide-A-Lyzer Dialysis Cassette (3 kDa; Thermo Scientific) to remove the hydroxyl amine and dimethyl sulfoxide. The sample was then transferred into a tube and centrifuged (16,000 × g, 4 °C, 30 min). We removed the supernatant and dissolved the pellet in 70 μL of water.

Amyloid Detection with CR Green Birefringence.

For analysis of the amyloid deposits grown on glass coverslips in cell culture, the medium was removed from the wells and the glass coverslips were washed once with 100 μL of PBS. The PBS was immediately removed, and 100 μL of ice-cold methanol (Sigma–Aldrich) was added to the specimens and incubated for 10 min at 4 °C. The ice-cold methanol was then replaced by 100 μL of CR solution containing 80% (vol/vol) ethanol, 3% (wt/vol) sodium chloride, and 0.6% (wt/vol) CR (Carl Roth). Specimens were incubated under shaking at 75 rpm for 45 min at room temperature using an orbital platform shaker (Heidolph Rotamax 120). Samples were washed three times by the addition of 100 μL of distilled water, which was removed immediately, after which 100 μL of Mayer’s Hemalaun solution (Carl Roth) was added and incubated for 2 min at room temperature. The Hemalaun solution was then replaced with 100 μL of 70% (vol/vol) ethanol, and the samples were washed three times with 100 μL of distilled water. The CR-stained samples were finally dehydrated using an ascending alcohol series. For this purpose, we removed the coverslips from a 96-well plate using a pair of tweezers and plunged them sequentially three times into a flask containing 90% (vol/vol) ethanol, three times into a flask with 100% (vol/vol) ethanol, and once each into two flasks of 100% (vol/vol) xylol. Between each plunging step, the residual ethanol or xylol was removed by dipping the coverslips onto filter paper. Coverslips were mounted on microscope slides with Roti-Histokitt (Carl Roth), and CR green birefringence was assessed in an Eclipse 80i polarizing microscope (Nikon).

For analysis of fibril solutions, 150 μL of a 1-mg/mL fibril solution was pipetted onto a glass slide and air-dried on a heating block (37 °C). The slide was washed for 20 min with 1 mL of 80% (vol/vol) ethanol containing 1% (wt/vol) NaCl and 0.01% (vol/vol) NaOH, and then incubated for 20 min with 1 mL of 80% (vol/vol) ethanol containing 1% (wt/vol) NaCl, 0.01% (vol/vol) NaOH, and 0.2% (wt/vol) CR (Carl Roth) at room temperature. Samples were washed twice with 100% (vol/vol) ethanol for 10 s each, mounted with 50 μL of Roti-Histokitt, and examined with a polarizing microscope (Eclipse 80i).

GAG Staining.

For Alcian blue staining, the medium in a well of a 96-well plate was exchanged with 100 μL of ice-cold methanol and incubated for 10 min at 4 °C. Afterward, the wells were washed three times with 100 μL of water and incubated with 100 μL of 1% (wt/vol) Alcian blue (Carl Roth) in 3% (vol/vol) acetic acid for 5 min at room temperature. The samples were then washed three times with water for 5 min each time. The samples were additionally stained with 100 μL of nuclear fast red solution (Carl Roth) for 10 min at room temperature, and then dehydrated and fixed with an ascending alcohol series [90% (vol/vol) ethanol, 100% (vol/vol) ethanol, 100% (vol/vol) xylol]. For that purpose, 100 μL of each solution was added onto the slide and removed immediately after addition. The samples were mounted on microscopic slides with 50 μL of Roti-Histokitt and examined with a light microscope (Eclipse 80i).

For staining of GAGs with an antibody, we removed the medium from the wells of a 96-well plate and added 100 μL of PBS that we immediately replaced with 100 μL of 0.1 M sodium phosphate buffer, pH 7.3, containing 4% (wt/vol) paraformaldehyde, 0.05% (vol/vol) glutaraldehyde, and 1% (wt/vol) sucrose. After a 30-min incubation at room temperature, we exchanged the buffer with 1% (wt/vol) BSA in PBS and incubated the samples for 1 h at room temperature. This BSA solution was then exchanged with 50 μL of 1% (wt/vol) BSA in PBS supplemented with 0.7% (vol/vol) mouse anti-heparan sulfate primary antibody (10E4; Amsbio). After a 1-h incubation period, the primary antibody was removed and the samples were washed three times with 200 μL of PBS for 5 min each time. We then added 50 μL of a 1% (wt/vol) BSA solution in PBS supplemented with 0.7% (vol/vol) goat anti-mouse secondary antibody that was conjugated to horseradish peroxidase (Dako) to the well and incubated the sample for 1 h. Afterward, we replaced the secondary antibody solution with 15 μL of 0.1% (wt/vol) 3,3′-diaminobenzidine tetrahydrochloride (Applichem) and 0.01% (vol/vol) hydrogen peroxide in PBS, and incubated the samples for 15 min at room temperature. Finally, samples were washed three times with 100 μL of water and dehydrated with 100 μL of 90% (vol/vol) ethanol in water, 100 μL of 100% (vol/vol) ethanol, and 100 μL of 100% (vol/vol) xylol, and mounted with 50 μL of Roti-Histokitt. We used an Eclipse 80i light microscope to assess the samples.

Apoptosis and Cell Viability Assays.

J774A.1 cells were seeded at a density of 3.5 * 10⁵ cells per milliliter in black 96-well plates (Greiner Bio-One). After a 24-h incubation period without SAA1, the medium was replaced with fresh medium supplemented with 1 mg/mL SAA1 and HDL. The apoptosis assay was carried out after incubation for 8 h or 2 d by determining the caspase-3/7 activity using a SensoLyte Homogeneous Rh110 Caspase-3/7 Assay Kit (AnaSpec) that we utilized according to the manufacturer’s protocol (n = 3). Fluorescence measurements were carried out with a FLUOstar OMEGA plate reader (BMG Labtech) using an excitation wavelength of 485 nm and an emission wavelength of 520 nm. The cell viability with the MTT assay was carried out after incubation for 24 h or 6 d, replenishing the medium and SAA1 and HDL at days 2 and 4, using a Cell Proliferation Kit I (Roche) according to the manufacturer’s protocol. The absorbance measurement was done in a FLUOstar OMEGA plate reader at 550 nm and 690 nm, blanked against buffer. The absorbance at 690 nm was subtracted from the measured absorbance at 550 nm. The value obtained with cells without HDL-SAA1 was set to 100% (n = 4).

Lysosomal Leakage Assay.

J774A.1 cells were seeded at a density of 3.5 * 10⁵ cells per milliliter in 96-well plates. After a 24-h incubation period, medium was exchanged with 100 μL of medium supplemented with cell culture-derived fibrils for 24 h. We removed the fibril-containing medium and added 40 μL of medium (without FBS) with 5 μg/mL Acridine Orange (AO; Sigma–Aldrich) and incubated the samples at room temperature for 15 min. The staining solution was discarded, and samples were washed twice with 100 μL of PBS for 5 s; the staining solution was then replaced with 100 μL of trypsin-EDTA for 10 s. We removed the trypsin-EDTA and incubated cells at 37 °C for 5 min. We then added 100 μL of medium with FBS to each well to inactivate remaining trypsin and immediately assessed the AO fluorescence of cells with excitation at 488 nm and emission at 700 ± 27 nm for 10,000 events using the BD VACSVerse flow cytometer (BD Biosciences). Gating of cells was done to determine the percentage of cells exposing low AO fluorescence, which represent cells with leaking lysosomes.

Measurement of AF488-Labeled Cell Culture-Derived Fibril Uptake by Flow Cytometry.

J774A.1 cells were seeded at a density of 3.5 * 10⁵ cells per milliliter in 24-well plates (Thermo Fisher Scientific). After a 24-h incubation period, we replaced the medium with 350 μL of culture medium containing AF488-labeled, cell culture-derived fibrils for 24 h. Afterward, we removed the medium and added 100 μL of trypsin-EDTA (Invitrogen) for 10 s. The solution was discarded, and samples were then incubated at 37 °C for 5 min. The remaining trypsin was neutralized by the addition of 350 μL of fresh culture medium. We immediately scraped off the cells with a Techno Plastic Products cell scraper, transferred the cell suspension into 2-mL tubes, and centrifuged the specimens (200 × g, 5 min, 4 °C). The supernatant was discarded, and the pellet was resuspended in 1 mL of flow cytometry buffer [2 mM EDTA, 0.5% (wt/vol) BSA, 0.1% (wt/vol) sodium azide in PBS (pH 7.3)]. We centrifuged the cells again and resuspended the cell pellet in 200 μL of flow cytometry buffer with 2% (wt/vol) paraformaldehyde (Carl Roth). After an incubation period of 15 min at room temperature, we centrifuged the samples a third time, discarded the supernatant, and resuspended the pellet in 200 μL of flow cytometry buffer. Samples were transferred into flow cytometry tubes (Sarstedt) and analyzed by flow cytometry (BD VACSVerse flow cytometer) with excitation at 488 nm and emission at 543 ± 16 nm. We measured 10,000 events.

Intracellular Colocalization Analyses.

We seeded J774A.1 cells on eight-well chambered cover glasses at a density of 350,000 cells per milliliter. After a 24-h incubation period, we removed the medium and added a 200-μL mixture of fresh medium with 50 μg/mL AF488-labeled, cell culture-derived fibrils and an uptake marker for phagocytosis (0.5 mg/mL pHrodo Red Zymosan A BioParticles Conjugate; Life Technologies). We incubated the cells at 37 °C for 3 h. Afterward, we removed the medium, washed the cells once with 200 μL of PBS, and added 100 μL of trypsin-EDTA for 10 s. The trypsin-EDTA was discarded, and cells were incubated at 37 °C for 5 min. We then washed the samples once with 200 μL of medium, added another 200 μL of fresh culture medium, and imaged the cells using an Eclipse Ti-E confocal laser-scanning microscope (Nikon) with a bright field, excitation at 488 nm and 561 nm, and an emission range of 500–550 nm and 570–1,000 nm, respectively.

Denaturing Protein Gel Electrophoresis.

Electrophoretic samples were prepared by mixing 15 μL of the sample solution to be analyzed with 5 μL of 4× NuPAGE LDS Sample Buffer (Life Technologies) and boiling for 10 min at 95 °C. Fifteen microliters of this sample was loaded onto a 4–12% NuPAGE Bis-Tris Gel (17 wells; Life Technologies). The gel was run for 38 min at 180 V in NuPAGE MES-SDS Running Buffer (Life Technologies) according to the manufacturer’s protocol using 3 μL of SeeBlue Plus2 Pre-Stained Standard (Life Technologies) as a marker. For reduced denaturing protein gel electrophoresis, we loaded 5 μL of 4× NuPAGE LDS Sample Buffer, 2 μL of NuPAGE Sample Reducing Agent, and 13 μL of the sample onto the gel after boiling for 10 min at 95 °C. The gel was run for 38 min at 180 V in NuPAGE MES-SDS Running Buffer supplemented with 0.25% (vol/vol) NuPAGE Antioxidant (Life Technologies).

Western Blot.

The proteins separated out by denaturing gel electrophoresis were transferred onto a 0.45-μm nitrocellulose membrane (Protran BA85; Whatman) using a Semi-Dry Blotting System (Biorad). Filter paper (Whatman) and the membrane were incubated in transfer buffer [1× NuPAGE Transfer Buffer (Life Technologies) in 20% (vol/vol) methanol] for 5 min before assembly. After transferring the proteins for 35 min at 20 V, the membrane was blocked in 50 mL of 5% (wt/vol) milk powder in PBST [PBS + 0.1% (vol/vol) Tween 20 (Carl Roth)] overnight at 4 °C. On the next day, 1 mL of 5% (wt/vol) milk powder in PBST containing 0.5% (vol/vol) primary antibody [rabbit serum against mouse AA(1–76) or rabbit anti-mouse SAP (Amsbio)] was added to the membrane. After an incubation period of 1 h at room temperature, the membrane was washed three times for 5 min in PBST. We then added 0.1% (vol/vol) secondary antibody (anti-rabbit from goat, horseradish peroxidase-conjugated; Dako) in PBST containing 5% (wt/vol) milk powder. The membrane was washed three times for 5 min in PBST, and the proteins were visualized by incubating the membrane with SuperSignal West Femto Chemiluminescent Substrate (Pierce) for 5 min and analyzing the chemiluminescence with a G:Box (Syngene).

SEM.

After a 6-d incubation period of the cells with or without SAA1 and HDL as indicated in the respective experiments, the medium was replaced with 100 μL of 0.1 mM sodium phosphate buffer (pH 7.3) supplemented with 2.5% (vol/vol) of glutaraldehyde and 1% (wt/vol) sucrose and incubated for 3 h at room temperature. Afterward, we removed the solution and washed the samples twice with 100 μL of PBS for 2 min at room temperature. We then dehydrated the samples using an ascending series of 30%, 50%, 70%, and 90% (vol/vol) propanol. To that end, samples were plunged into flasks containing 100 mL of the different propanol concentrations for 1 min at room temperature. After dehydration, we plunged the specimens twice into a flask containing 100% (vol/vol) propanol for 10 min at room temperature. The dehydrated specimens were then critical point-dried with a CPD Bal-Tec 030 Critical Point Dryer such that the 100% (vol/vol) propanol was exchanged within 2 min with pure carbon dioxide over a total of 14 steps at 8 °C. Samples were heated to 41 °C to raise the pressure above the critical point of CO₂ (73 bar, 31.1 °C) and coated with a 5-nm carbon layer by electron beam evaporation using a Balzers BAF 300 (Bal-Tec). Samples were analyzed with a Hitachi S-5200 scanning electron microscope (Hitachi) operated at an acceleration voltage of 10 kV. We used a secondary electron detector (Hitachi) and an yttrium-aluminum-garnet–backscattered electron detector (Hitachi). Images of 1,280 × 960 pixels were taken.

Immunogold Labeling.

For SEM, the immunogold labeling was done before the sample preparation for SEM. In this case, we incubated the sample with 100 μL of a 1% (wt/vol) BSA solution in PBS (blocking solution) for 20 min. This solution was removed, and the sample was exposed to 100 μL of fresh blocking solution supplemented with 5% (vol/vol) of polyclonal, anti–full-length SAA1.1 primary antibody for 30 min. After primary antibody staining, samples were washed three times with 100 μL of PBS for 5 min at room temperature. Next, the samples were incubated with 100 μL of blocking solution supplemented with 5% (vol/vol) of 10-nm gold-conjugated, polyclonal, anti-rabbit secondary antibody purified from goat (Aurion) for 30 min at room temperature. We removed the secondary antibody solution and washed the samples three times with 100 μL of PBS for 5 min at room temperature.

For TEM, samples were high-pressure-frozen as described and the frozen samples were exposed to 1 mL of freeze substitution solution consisting of acetone supplemented with 0.1% (wt/vol) uranyl acetate and 5.2% (vol/vol) water. The temperature of the samples was raised from −90 °C to −20 °C over a period of 16 h, and the acetone solution was removed. The samples were briefly washed three times with 1 mL of 100% (vol/vol) propanol each time. The propanol was removed, and 1 mL of 33% (vol/vol) LR Gold resin (Plano-Agar) in propanol was added to the sample and incubated for 1 h at room temperature. The 33% (vol/vol) LR Gold resin solution was then exchanged with 50% (vol/vol) LR Gold resin in propanol, 66% (vol/vol) LR Gold resin, and 100% (vol/vol) LR Gold resin in propanol, and incubated for 1 h during each step. Finally, the 100% (vol/vol) LR Gold resin was removed, and 0.25 mL of fresh 100% (vol/vol) LR Gold resin was added to the samples. We then exposed samples to UV light for 2 d to polymerize the resin. Resin blocks were stored at room temperature. From these resin blocks, we cut out 70-nm sections and collected them on formvar-carbon–coated grids (Plano). The grids were then incubated in 100 μL of 1% (wt/vol) BSA in PBS for 10 min at room temperature. Afterward, the solution was exchanged with 100 μL of fresh blocking solution supplemented with 0.5% (vol/vol) rabbit anti–full-length SAA1.1 antibody, and samples were incubated for 30 min at room temperature. Next, we washed samples with 100 μL of 1% (wt/vol) BSA in PBS six times with a 2-min incubation per step, followed by incubation with 100 μL of 1% (wt/vol) BSA in PBS supplemented with 2% (vol/vol) secondary antibody (goat anti-rabbit antibody conjugated with 10-nm gold particles; Aurion) for 30 min. Samples were then washed three times with 100 μL of blocking solution and three times with 100 μL of PBS (2 min per washing step). Finally, we stained samples with 1% (wt/vol) uranyl acetate in PBS for 5 min and briefly washed them three times with water.

Three-Dimensional and 2D Analysis of Fibril Orientations.

The 3D distributions of fibril orientations have been calculated using kernel density estimation on the unit sphere with a Gaussian kernel and a bandwidth of $h=0.2$ (50, 51). The fibrils extracted from the 3D tomograms are represented as polygonal lines (i.e., they consist of multiple straight line segments). The (normalized) direction vectors $x_1,\dots ,x_n$ of these line segments can be interpreted as points on the unit sphere and are used as data points in kernel density estimation, where each direction, $x_i,$ is weighted with the length, $M_i$, of the corresponding line segment. Then, the estimated probability density of a direction $x$ on the unit sphere is given by Eq. S1, where $k$ is the Gaussian kernel, $M={\displaystyle \sum _{i=1}^n}{M}_i$ is the length of the line segment, and $c$ is a normalizing constant. Distances between pairs of directions have been computed using the orthodromic distance [i.e., $d(x_1,x_2)$ is the length of the shortest path on the surface of the unit sphere, which connects $x_1$ and $x_2$]. The constant $c$ is needed to adjust for the fact that the Gaussian kernel does not integrate to 1 on a bounded domain, and is computed via numerical integration. When calculating direction distributions in a cutout of the domain, polygonal lines have been cut off at the edges of the region of interest and only the parts within have been considered:

$$\widehat{f}\left(x\right)=c\frac{1}{Mh}{\displaystyle \sum }_{i=1}^n{M}_i\text{k}\left(\frac{d(x,\hspace{0.17em}{x}_i)}{h}\right).$$ [S1]

The 2D direction distributions have been computed analogously using kernel density estimation on the unit circle instead of the unit sphere. Here, the polygonal lines in the region of interest have been projected onto the plane $z=0$, and the directions and lengths of the projected segments have been used as data points.

X-Ray Diffraction.

Fibrils extracted from the cell culture were centrifuged at 100,000 rpm for 30 min at 25 °C using a S100AT3-204-100 KRPM rotor and a SORVALL RC-M120GX centrifuge. The pellet was scraped out, and a portion of the wet pellet was placed on top of a VariMax HR cryoloop holder (Rigaku). X-ray diffraction images of the fully hydrated fibrils were recorded using an R-AXIS IV++ detector (Rigaku) at room temperature. The sample-to-detector distance was set to 20 cm. The exposure time was 60 s. Images were processed with Crystal Clear 1.3.6 SP3 software (Rigaku). The diffraction spacings were taken at 10 points along the reflection arc and subsequently averaged.

Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy.

Spectra were recorded at room temperature on a Tensor 27 FTIR spectrometer (Bruker) equipped with BIO-ATR II cell and a mercury cadmium telluride detector cooled with liquid nitrogen. Twenty-five microliters of the sample (1 mg/mL protein in water) was placed onto the crystal of the BIO-ATR II cell and incubated for 2 min. Spectra represent averages of 64 scans, using an aperture of 4 mm and an instrument resolution of 4 cm⁻¹ with four times zero filling. The amide I and II regions were fit with Gaussian and Lorentzian curves. The peak position of these components was determined from the second derivative of the spectrum.

Thioflavin T Fluorescence Measurement.

Spectra were measured at room temperature using an LS 55 fluorescence spectrometer (PerkinElmer) and a SUPRASIL 10-mm 105.253-QS quartz fluorescence cuvette (Hellma). Samples (160-μL volume) contained 15 μM thioflavin T, 4 mM sodium phosphate buffer (pH 7.4), and 20 μM fibrils where indicated. Fluorescence emission spectra represent the averages of five scans that were recorded from 460 to 700 nm, using an excitation wavelength of 450 nm, a scan speed of 100 nm/min, and excitation and emission slit settings of 7 nm each.

Statistical Analysis.

Results were analyzed by the Student’s t test (unpaired, unequal variances). Error bars show the SD.