Amyloid and the Cross-Beta Architecture

Amyloid fibrils were first identified within a group of diseases known as Amyloidoses, in which proteinaceous fibrils accumulate in the extracellular spaces of tissues¹. Each disease is characterized by a particular precursor protein which undergoes a conformational change to form highly ordered, stable protein fibrils^2,3. The more recent definition of amyloid fibrils now incorporates fibrils that accumulate within the cellular environment. More recently, a number of functional systems have been identified that utilize the amyloid structure to provide stability, adhesion, or even information storage. Despite the wide range of protein structures that can self-assemble to form amyloid fibrils, they all share a cross-β structure⁴. This cross-β arrangement was first described by Astbury for hen egg white⁵ and later explained further by Geddes for cross-β silk⁶. In this protein architecture, the fibrils are made up of β-strands that run perpendicular to the fibril axis, which are then arranged into β-sheets which associate via the side chains to form what was later referred to as steric zippers⁷. This creates a highly stable organization held together via hydrogen bonds that run parallel to the fiber, alongside hydrophobic, electrostatic, and polar interactions via side chains⁸.

Early structural studies used X-ray fiber diffraction to interrogate the structure of amyloid fibrils, whereby the fibrils were aligned and oriented to form a bundle of amyloid fibrils which could be placed in the X-ray beam of a diffractometer^9,10. A generic structure of amyloid fibrils was generated to fit the diffraction patterns from several amyloid fibrils extracted from human disease tissue or grown in vitro^4, and this has provided a basis for further structural descriptions¹¹. In recent years, advances in solid state nuclear magnetic resonance (ssNMR), microcrystal X-ray crystallography, and cryogenic electron microscopy (CryoEM) have provided atomic detail for the cross-β structure¹¹. However, the X-ray fiber diffraction method remains important and useful, and provides information regarding the repetitive organization of the proteins. Furthermore, in some cases, information relating to a longer-range order may be accessible. Fundamentally, X-ray fiber diffraction is often used to confirm the presence of amyloid fibrils formed from different proteins, and the cross-β pattern is one of the principal requirements for a fibrous assembly to be classified as amyloid¹².

Here, we describe the methodology used to prepare amyloid fibrils and align them for X-ray fiber diffraction data collection. Furthermore, we provide details of how the diffraction data may be further analyzed to generate a model structure, its predicted diffraction pattern, and compared to diffraction data from ex-vivo-derived fibrils.

1. Preparation of fibrillar samples for X-ray fiber diffraction

1. Dissolve protein or peptide of interest in optimal conditions for fibril formation for the selected amyloidogenic peptide and assess fibril formation using electron microscopy or atomic force microscopy. Details of how to prepare samples and conduct imaging are available elsewhere¹³. Ensure that the fibrils are long, straight, and unbranched as shown in Figure 1. For example, the fibrils shown in Figure 1 were imaged using transmission electron microscopy (TEM). Place a droplet containing fibrils on an electron microscope grid and then negatively stain using uranyl acetate¹⁴.
   NOTE: An even covering and distribution of the fibrils on the grid is desirable.
2. Align to produce a bundle of fibrils whose fiber axes are ideally all parallel (Figure 2A).
   1. Melt candle wax until it is liquid using a hot plate. Using a glass cutter such as a diamond pen, cut 1 mm borosilicate glass capillaries to a length of about 2-3 cm and dip them into the liquid wax briefly. Draw the candle wax up by capillary action (approximately 1-2 mm in length within the capillary). Ensure the wax surface is flat so that the fibril suspension can rest evenly.
   2. Use plasticine to attach the sealed capillaries horizontally in a Petri dish with a 0.5 mm to 2 mm gap between the wax-filled ends (Figure 2A).
   3. Using a pipette, carefully suspend a 10 µL droplet of solution between the wax-filled ends of the two capillaries. Ensure that the solution makes contact with both wax ends but does not extend onto the glass.
   4. Protect the sample from dust by covering the Petri dish. Allow the sample to evaporate overnight, resulting in alignment of the fibrils (see Figure 2A).
      NOTE: It may take several days for the sample to dry fully, and this will depend on the ambient temperature and humidity, and the viscosity of the sample.
   5. Seal the plate; incubate at lower temperatures in a refrigerator or incubate using humidity in a sealed box to slow down drying, which may result in better alignment of the sample.
3. Align amyloid fibrils in a glass capillary (Figure 2B).
   1. Fix a 1 mL syringe fitted with a small piece of rubber tubing to the wide end to a 0.7 mm diameter, siliconized glass capillary and aspirate up 2-3 cm of fibril suspension.
   2. Seal one end of the capillary using melted wax and allow the other end to remain open to allow evaporation. Arrange the capillary vertically and allow it to dry over several days or weeks until it has formed a dried disk in the capillary tube (Figure 2B).
4. Align fibrils as a "mat" or thin film (​​Figure 2C).
   1. Place a high concentration (e.g.,>10 mg/mL) of fibril sample onto a glass slide, parafilm, or piece of Teflon, cover, and allow it to air dry for at least 24 h. This should dry to form a thin, flat film on the surface. Carefully lift this using a scalpel and then attach it to a glass capillary tube using glue or plasticine.
   2. The mat will have a random arrangement of fibrils (see Figure 2C). Carefully mount the sample on a goniometer head so that the X-ray beam passes through the edge of the sample and parallel to the plane of the film (Figure 2C).
      NOTE: Diffraction through the face of the film will yield diffraction rings due to random alignment (see Figure 2C).
5. Examine the sample.
   1. Investigate how well the sample is aligned by examining it through a cross-polariser under a light microscope. If birefringence is present, it may suggest that there is good alignment in the fibril bundle.
   2. Attach the capillary fiber sample onto a goniometer head using plasticine. Consider whether the sample is a fiber bundle, mat, or disk when placing it in the X-ray beam.
      1. Note that the sample rotation axis will be around the fiber axis for a fiber bundle (Figure 2A), while in disks the fibers tend to align across the diameter of the tube (Figure 2B). A mat will have two distinct directions (parallel to the plane of the film and perpendicular to it) (Figure 2C).
   3. Collect a diffraction pattern (as shown in Figure 3A) using in-house X-ray equipment (CuKα for protein crystallography) or using synchrotron radiation. Exposure times will vary depending on the intensity of the X-ray beam; adjust this depending on recommended collection times for the diffraction source.

2. Data analysis

1. Use of CLEARER for data analysis¹⁵.
   NOTE: The X-ray fiber diffraction pattern analysis program is available free to academic users from the corresponding author. Other methods for diffraction analysis are available^16,17 (www.esrf.fr/computing/expg/subgroups/data_analysis/FIT2D)¹⁸. Italics within parentheses indicate menu commands.
   1. Convert the diffraction pattern to a .tiff file and upload the file into CLEARER (Figure 4).
   2. Input the diffraction settings (Diffraction Pattern: Diffraction settings), including the pixel size in the TIFF image, the X-ray source, and the sample to detector distance.
   3. Ensure that the diffraction pattern is oriented so that the meridian is vertical (Image Processing: Rotate). Use the Centering module (Diffraction Pattern: Centre Image) to centre the image. Define the two main axes (meridian and equator):- This module compares intensity along these two axes of the pattern. Once centering is completed, save the image and ensure that it is used for subsequent steps.
   4. Specify a suitable angle (set along the two main axes [meridian and equator]) and output a graph showing diffraction signal position against intensity using the Radially Average module (Diffraction Pattern: Radially Average) (Figure 4A).
   5. Measure and automatically output the diffraction signal positions using the Peak Find module (Diffraction Pattern: Radially Average: Peak Find) (Figure 4B). Ensure that all diffraction signals are being detected accurately by adjusting the peak search width while making sure that noise is not being detected.
   6. Check the signal positions output by Peak Find by using the Zoom & Measure module (Diffraction Pattern: Zoom & Measure). Verify all signal positions are listed and are assigned to the correct axis.
   7. Explore unit cell dimensions (b and c) that will generate the observed diffraction signals using the Unit Cell Optimization module (Diffraction Pattern: Unit Cell Optimization) (Figure 5A). Examine the strong signals found on the equator to generate an initial guess to input into the unit cell input table and specify the min and max [hkl] indices over which CLEARER searches for potential cells (a dimension is meridional and defaults to 4.76 Å).
   8. Calculate potential unit cell dimensions and compare the predicted diffraction signal positions with observed reflections. Unit cells are scored depending on the degree of agreement between the predicted and experimental signals. The top twenty matches are then displayed. Consider the peptide precursor length and other information from electron microscopy, AFM, or solid state nuclear magnetic resonance (ssNMR) to help select the best possible unit cells to fit the data and the sample.

3. Structural modeling and testing the model using CLEARER

1. Construct molecular models (PDB format).
   ​NOTE: Utilize protein modelling programmes such as Pymol (DeLano)¹⁹ or Insight II (Accelerys) and others, to generate PDB coordinates of the precursor peptide in β-strand conformation and in a β-sheet arrangement. This section describes a simple process used for an orthorhombic unit cell (all 90° angles). Non-orthogonal organizations are possible where more expert knowledge of diffraction is required. These can be explored in CLEARER for more complex unit cells.
   1. Build three-dimensional arrangements of the peptides in a β-sheet arrangement to fit into the determined unit cell(s). Ensure this model contains all symmetry-related objects within the cell.
   2. Load the PDB model into the Structure Chain Generator module (Diffraction Simulation: Structures: Structure Chain Generator) to check the constructed structure coordinates. Ensure that the fiber axis and beam axis are correctly assigned. Input the determined unit cell dimensions. Check whether clashing occurs for the selected cell dimensions (Figure 5B).
   3. Upload PDB model coordinates into the Fiber Diffraction Simulation module (Diffraction Simulation: Fiber Diffraction Simulation).
      1. Input the fiber axis direction, beam orientation, unit cell dimensions selected from step 3.1.2, X-ray source and detector settings, and click simulate. For an orthorhombic unit cell simulation, the fiber axis is defined by the Cartesian axis related to the orientation of the fiber axis in the model used.
      2. Ensure that the beam axes are perpendicular to fiber axis, i.e., on one of the other two possible axes. Save the simulation settings as .html files (File: Save As). These can be uploaded and edited later to iteratively explore the optimal model structure.
      3. Save the resulting simulated pattern in a raw format, which can be uploaded later for further analysis, or as a .tiff (this format has additional contrast information discarded) (Simulation: Save Viewable Image as TIFF). Then, compare these to the experimental pattern.
2. Compare simulated diffraction patterns with experimental patterns for presentation.
   1. Use steps 2.1.5-2.1.6 to measure simulated diffraction signals and compare with those measured from experimental patterns (Figure 6). Record these in a table.
   2. Use a module called Layers (Diffraction Pattern: Layers) to overlay simulated and experimental patterns to compare signal positions ( Figure 6C).
   3. Compare signal positions and relative intensities between the experimental and simulated patterns by arranging a quadrant of the simulated pattern alongside the experimental pattern (as shown in Figure 6B)^8,20,21.
   4. Evaluate the validity of a suggested unit cell and model, and then iteratively alter the model and cell to improve the similarity between the simulated pattern and experimental diffraction pattern.
   5. Extensively compare the results of different models and cells using the procedure outlined in step 3.1 to test several alternatives. Ensure that information gathered from other techniques, such as TEM, AFM, and NMR, is considered to constrain modeling and to arrive at the best possible structural representation that fits the data^8,21,22,23.

CcbMet example24: Figure 1 shows electron micrographs showing a high density of amyloid fibrils on a grid suitable for preparation of an X-ray fiber sample. Figure 2 describes the different textures that can be generated from fibrous samples fiber bundle (Figure 2A), disk (Figure 2B), mat/film (Figure 2C), and shows the direction of the X-ray beam. Figure 3A shows the diffraction pattern obtained from stretch-frame aligned fibers (texture shown in Figure 2A) and portrays the characteristic cross-ß diffraction pattern with a meridional reflection observed at 4.76 Å and a set of equatorial reflections arising from the chain length, sheet spacing, and size of the protofilaments. The idealized cross-ß diffraction pattern shown in Figure 3B shows that meridional and equatorial reflections may be directly related to repetitive spacings on corresponding axes of the fiber, as shown in Figure 3C. Figure 4 and Figure 5 show the process of analysis of the pattern using the program CLEARER. A well-centred pattern will be symmetric, and so one expects the equatorial and meridional peaks to overlap in position, as in Figure 4A. Peaks can then be automatically found as in Figure 4B. These peaks are used to find a plausible unit cell, as in Figure 5A. Figure 5B,C depicts the preparation of a simulation using the identified unit cell and the visualization of the structure used. A comparison of the simulated diffraction data from the potential model structure (shown in Figure 6A) with the experimental data is presented in Figure 6B, demonstrating a reasonable match in reflection positions and relative intensities. This may be further compared by overlaying the simulated and experimental patterns as shown in Figure 6C.

Figure 1: Electron micrograph of amyloid fibrils. Negatively stained grid, showing the characteristic long, straight and unbranching fibrils distributed over the microscopy grid. Scale bar is 0.2 mm. A 4 mL droplet of fibrils formed from 10mg/ml islet amyloid polypeptide in filtered, milliQ water were placed on a copper 400 mesh grid coated with formvar and carbon and negatively stained using uranyl acetate. Please click here to view a larger version of this figure.

Figure 2: Fiber alignment. (A) The placement of a droplet of fiber suspension between two wax tipped capillaries and subsequent drying to form a fiber bundle. (B) The drying of a disk within a siliconized capillary tube. (C) The formation of a flat film or mat. It is necessary to consider the direction of the alignment of fibers when mounting the diffraction sample in the beam and interpreting the diffraction pattern (see 1.5.2). Please click here to view a larger version of this figure.

Figure 3: Cross-β pattern from amyloid fibrils. (A) The diffraction pattern obtained from aligned amyloid fibrils (cc b -Met). (B) A schematic diagram showing the essential elements of a cross-b pattern. (C) A schematic showing the structural element of a cross-b structure. Please click here to view a larger version of this figure.

Figure 4: Use of CLEARER. (A) Uploading the diffraction pattern. (B) Reflection measurement Please click here to view a larger version of this figure.

Figure 5: CLEARER. (A) Unit cell identification. (B) Fiber structure visualisation. (C) Simulation parameters. Please click here to view a larger version of this figure.

Figure 6: Testing model structures. (A) Contents of a potential model structure of a unit cell shown from the top and side. (B) Comparison of the experimental pattern with the calculated pattern (insert) as a quadrant. (C) Comparison using layers. Please click here to view a larger version of this figure.

X-ray fiber diffraction can be used to evaluate whether a sample is forming an amyloid structure and to gain further insight into the molecular organization of the precursor protein or peptide within the fibers. Here, we have described the methods by which fibrous samples may be prepared for X-ray fiber diffraction, data may be obtained, and then analyzed. Critically, the quality of the data obtained will depend on both the order within the individual fibers. Therefore, initial sample preparation is of major importance, and whether the fibers are synthetic or extracted from tissue will also affect data quality. Exploration of optimal buffers for amyloid fibril assembly may be important since excessive salt may obscure the patterns. Washing of fibrils made in salt buffers, such as phosphate buffer, can improve both the quality of the fibrils and the degree of alignment, resulting in higher quality data. Optimization of alignment methods can also enhance the quality by selecting from methods in section 2, or by slowing drying of the sample to allow more time for alignment. The more aligned the fibrils are relative to one another, the better oriented the diffraction pattern, and the more information can be obtained from the data.

For data analysis, intensive exploration of possible unit cells may be required to arrive at the most suitable unit cell for the sample. Taking into account the size of the peptide or protein will provide further guidance. The contribution of AFM or TEM morphology information may provide additional detail to guide the selection of unit cell(s) to be tested. Model building is then required to test whether the model placed in a unit cell is able to produce a diffraction pattern that matches experimental diffraction patterns. Problems can occur during diffraction simulation if the incorrect fiber and beam directions are selected, so it is essential to check these in the Structure chain generator as described in steps 3.1.1-3.1.2. Once the diffraction pattern is generated, comparison of patterns is important, and this will require intensive attention to detail and optimization of the model structure, as well as small changes to the unit cell, to arrive at the best model that fits the data.

Here, we have described model testing for a simple system in which the unit cell is orthorhombic, allowing straightforward testing of models. In more complicated systems, where one or more angles are not 90°, careful input of diffraction settings is necessary, which is outside the scope of this manuscript.

X-ray fiber diffraction was famously used to generate the structure of DNA, underlining its importance in structural biology. Since then, it has contributed to our understanding of numerous protein structures, including keratin, collagen, actin/myosin, and muscle, as well as amyloid fibrils. The method suffers limitations compared to X-ray crystallography since the cylindrical averaging and overlapping of diffraction signals result in difficulty in interpreting a structure directly and unambiguously. However, X-ray crystallographic methods can not be used for most fibrous structures due to heterogeneity and insolubility. Major advances have been made in cryoEM analysis of fibrous structures in recent years, and these now utilize single particle overlap to generate structures of the building blocks that generate the fibers. X-ray fiber diffraction provides a relatively quick and straightforward method to perform an initial characterization of fibers, to confirm that they are cross-β. Highly ordered samples may provide details that can be analysed further to generate model structures as described here. Therefore, despite major advances in structural biology methods, X-ray fiber diffraction remains important and relevant.

Overall, the use of fiber diffraction has played a key role in the generation of the generic model for amyloid fibrils^4,9,25. Furthermore, the cross-β pattern is included as a requirement for the definition of a protein fibril as Amyloid². Finally, numerous model structures have been generated using X-ray fiber diffraction data as a guide. This has resulted in a basis for the addition of further structural detail obtained from other methods.