Key Charged Residues Influence the Amyloidogenic Propensity of the Helix-1 Region of Serum Amyloid A

Abstract

Increased plasma levels of serum amyloid A (SAA), an acute-phase protein that is secreted in response to inflammation, may lead to the accumulation of amyloid in various organs thereby obstructing their functions. Severe cases can lead to a systemic disorder called AA amyloidosis. Previous studies suggest that the N-terminal helix is the most amyloidogenic region of SAA. Moreover, computational studies implicated a significant role for Arg-1 and the residue-specific interactions formed during the fibrillization process. With a focus on the N-terminal region of helix-1, SAA_1–13, mutational analysis was employed to interrogate the roles of the amino acid residues, Arg-1, Ser-5, Glu-9, and Asp-12. The truncated SAA_1–13 fragment was systematically modified by substituting the key residues with alanine or uncharged but structurally similar amino acids. We monitored the changes in the amyloidogenic propensities, associated conformational markers, and morphology of the amyloids resulting from the mutation of SAA_1–13. Mutating out Arg-1 resulted in much reduced aggregation propensity and a lack of detectable β-structures alluding to the importance of salt-bridge interactions involving Arg-1. Our data revealed that by systematically mutating the key amino acid residues, we can modulate the amyloidogenic propensity and alter the time-dependent conformational variation of the peptide. When the behaviors of each mutant peptide were analyzed, they provided evidence consistent with the aggregation pathway predicted by MD simulation studies. Here, we detail the important temporal molecular interactions formed by Arg-1 with Ser-5, Glu-9, and Asp-12 and discuss its mechanistic implications on the self-assembly of the helix-1 region of SAA.

Graphical Abstract

INTRODUCTION

AA amyloidosis, a systemic disorder resulting from the deposition of circulating amyloid fibrils derived from serum amyloid A (SAA), can lead to complications of chronic inflammatory conditions. SAA is an acute-phase protein produced in response to inflammatory disorders such as rheumatoid arthritis, Crohn’s disease, inflammatory bowel disease, and infection.¹ In AA amyloidosis, circulating amyloid is frequently deposited in the liver, spleen, and kidney. Renal failure is the most prominent cause of death in patients suffering from AA amyloidosis. There is accruing evidence that overexpression of SAA is strongly correlated with the severity of inflammation caused by COVID-19, making SAA a potential therapeutic target for associated long-term inflammatory diseases.^2–6

It is believed that the overexpression and subsequent increase in levels of circulating SAA protein, amyloid fibril precursors responsible for AA amyloidosis, poses a significant risk in the development of the disease.⁷ While SAA has been identified as a key contributor to the development of AA amyloidosis, the exact molecular mechanisms driving protein misfolding that leads to amyloid deposits have yet to be determined. Human SAA is encoded by three different genes: Saa1, Saa2, and Saa4.⁸ Currently, it is known that only Saa1 and Saa2 genes code for acute phase isoforms of SAA. Saa4 is a constitutive isoform that is not overexpressed during inflammation.^9–11 In a study using SAA isoforms from CE/J mice, results show that SAA1 can produce cytotoxic soluble oligomers that can persist for an extended period of time, making it much more pathogenic than SAA2.^12–14

Full-length SAA1 (henceforth referred to as SAA) is a 104-residue polypeptide that is present in plasma as an apolipoprotein on a high-density lipoprotein. However, amyloid deposits of SAA are typically composed of a truncated version of the polypeptide, containing only the first 66 or 76 amino acids.^{15, 16} These amyloid deposits also contain glycosaminoglycans and serum amyloid P.^{17, 18} Predictive analysis of the amino acid sequence suggests that human SAA, consistent with its lipid binding properties, contains two potential amphiphilic α-helical regions.¹⁹ Recent investigations have identified residues 1–27 as the lipid binding region of SAA, with residues 1–11 forming the helical sequence essential for lipid binding.²⁰

In an effort to gain insight into the misfolding and structure of SAA, Nordling and Abraham-Nordling conducted ab initio calculations and computational modeling of the polypeptide.²¹ Their findings indicated that SAA is composed of a five-membered helical bundle with a fold related to the tetratricopeptide repeat domain. Of the five helical regions, the N-terminal helix, which included the first 13 residues (SAA_1–13), was found to be the least stable. Molecular dynamics (MD) simulations (10 ns) revealed that SAA_1–13 spent less than 35% of the simulation time in a helical conformation. In contrast, helices 2 through 5 were relatively stable. Longer (100 ns) simulations indicated that the N-terminal region of SAA completely loses its helical structure and adopts a low energy β-hairpin conformation that persists for the last 30 ns of the calculation.

In silico simulations monitoring the trajectory of the residues encompassing helix-1 revealed key contacts and interactions involved in stabilizing the β-hairpin structure.^21–24 It was evident from the published simulation studies that the initial interactions between the guanidino group of Arg-1 and the side chains of Ser-5 and Glu-9 stabilize the SAA_1–13 helix. As the interactions between Arg-1 and Ser-5 dissolve, Arg-1 subsequently forms a salt bridge with the carboxylate end of Asp-12. Moreover, the initial Arg-1 and Glu-9 contact devolves as SAA_1–13 transitions to the β-hairpin conformation stabilized by the Asp-12/Arg-1 ion pair. Multiple hairpins eventually can stack upon one another to produce the characteristic amyloid fibrils.²¹

Bernhardt et al. used a different force field and an implicit solvent while extending the simulation window to 200 ns.²³ Within this time frame, they observed that SAA_1–13 changes structure from helix to β-hairpin and back to helix. Consistent with an earlier study, they contend that the loss of Arg-1-Ser-5 and Arg-1-Glu-9 contacts promotes the formation of β-hairpin. They modeled various SAA_1–13 structures where each key residue was replaced by alanine and found that replacing Ser-5 and/or Glu-9 increased β-hairpin formation while substituting Asp-12 yielded higher helical content.

Computational studies done by Wang et al. revealed that the full helix-1 sequence (SAA_1–27) became more flexible when separated from the native sequence.^{22, 24} They showed that an isolated SAA_1–27 fragment forms a kink at residues 12–13 dividing the sequence into two helical segments: residues 1–11 (helix-1a) and 14–27 (helix-1b). Helix-1a, which has a lower helicity, dramatically changes becoming elongated and disordered.²² Without the stabilizing interaction between Phe-11 and Tyr-21, helix-1a devolves and becomes amyloidogenic. In vitro studies by Skibiszewska et al. revealed that SAA_1–12 exhibited a greater amyloidogenic propensity than SAA_1–27^.25

The crystal structure of the native SAA revealed that the protein exists as a hexamer with each monomer characterized by an antiparallel four-helix bundle fold with helices 1, 2, 3, and 4 involving residues 1–27, 32–47, 50–69, and 73–88, respectively.²⁶ The C-terminal “tail”, composed of residues 89–104, forms an ordered loop structure by establishing interactions, like salt bridges and H-bonds, with helices 1, 2, and 4. The ordered loop structure wraps around one end of the four-helix bundle stabilizing the native helices. This accounts for why amyloid deposits of SAA contain C-terminal truncated forms of the polypeptide.²⁶

In agreement with published computational studies, Lu et al. found that the peptide corresponding to the helix-1 region of SAA readily formed amyloid fibrils and was the most amyloidogenic of all four helices.²⁶ These observations corroborate studies where deletion of the first 11 residues of SAA dramatically decreases the ability of polypeptides to form amyloid.^{27, 28} Lu et al. speculate that dissociation of the hexamer and proteolytic cleavage of the C-terminal region may lead to disruption of the four-helix bundle and conformational changes that leads to amyloid formation.²⁶

The amyloidogenic propensity of peptides or proteins is heavily influenced by its amino acid sequence. Numerous studies employing diverse techniques, including combinatorial and computational chemistry, kinetic aggregation assays, spectroscopy, and microscopy, have shown that even an amino acid substitution can dramatically alter the polypeptide’s ability to form amyloid fibrils, either by facilitating or inhibiting the process.^29–33 Non-polar, particularly aromatic, residues play a crucial role, as evidenced by reduced amyloid fibril formation upon their substitution in various amyloidogenic peptides^34–36 and their prevalence in aggregating sequences including animal and yeast prions.³⁷ The position and side chain characteristics of substituted residues are equally important, as demonstrated by point mutations of aromatic residues in nucleophosmin 1 fragment.³⁸ Electrostatic interactions also significantly impact amyloidogenicity, influencing native structure, fibril morphology, and amyloid growth kinetics, as observed in curli-specific gene encoding protein A mutational studies.³⁹ The putative helix-1 region of SAA, SAA_1–27, was subjected to N-terminal modifications specifically targeting the charged Arg-1. Carbamoylation of the N-terminal amino group delayed the onset of amyloid formation and changed fibril morphologies. Meanwhile, acetylation of the same amino group and truncation of Arg-1 significantly decreased the amyloidogenic propensity of the peptide.⁴⁰ These findings collectively underscore the importance of amino acid composition and placement in determining amyloidogenic behavior. They provide valuable insights in elucidating the mechanism of amyloidogenesis and developing pathways for targeted therapeutic strategies.

With a focus on the peptide fragment SAA_1–13, mutational analysis was employed to interrogate the roles of the amino acid residues, Arg-1, Ser-5, Glu-9, and Asp-12, implicated in the MD simulation-based misfolding mechanism. Using various techniques, we monitored the changes in the amyloidogenic propensities, associated conformational markers, and morphology of the amyloids resulting from mutating SAA_1–13. Collected results were corroborated using MD simulation studies. Here, we detail the mechanistic implications of perturbing Arg-1, Ser-5, Glu-9, and Asp-12 in SAA_1–13.

MATERIALS AND METHODS

Fmoc-protected amino acids, Rink amide resin, and coupling reagents for peptide synthesis were obtained from Advanced Chemtech (Louisville, KY). Triisopropylsilane (TIPS) and 1,1,1,3,3,3-hexafluoroisopropanol (HFIP) were purchased from TCI America (Portland, OR). Dimethylformamide (DMF), dichloromethane, and acetonitrile were acquired from Pharmco-AAPER (Brookfield, CT). All other chemicals were from Sigma-Aldrich (St. Louis, MO).

Peptide Synthesis

Synthesized peptides were derived from the first 13 amino acid residues of human SAA protein (SAA1, UniProtKB P0DJI8).⁴¹ Peptide synthesis was performed using Fmoc-solid phase chemistry. Low substitution Rink amide resin (substitution level 0.30 mmol/g) was used with 1-hydroxybenzotriazole hydrate, N,N,N’,N’-Tetramethyl-O-(1H-benzotriazol-1-yl) uranium hexafluorophosphate, and 4-methylmorpholine as activating agents. Deprotection of the Fmoc group was done by 20 % (v/v) piperidine in DMF. Fmoc-deprotection and coupling steps were carried out by shaking for 20 min and 1 h, respectively, at room temperature using Wrist Action™ Shaker (Burrel Scientific, Pittsburg, PA). Cleavage of peptides from the resin was achieved using a 95:2.5:2.5 volume ratio of trifluoroacetic acid (TFA), water, and TIPS. Crude peptides were purified by reverse-phase HPLC (Varian Pro-Star, Palo Alto, CA) on a Vydac C18 column (22 × 250 mm) using a linear gradient of acetonitrile/water containing 0.1% TFA. Appropriate fractions were pooled, concentrated, and lyophilized to provide purified peptides as white powders. Peptide purity was assessed by analytical reverse-phase HPLC using a Vydac C18 (4.6 × 250 mm) column. Peptide structure was confirmed using matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) LTQ XL mass spectrometry (ThermoScientific, Somerset, NJ).

Amyloidogenic Propensity by Thioflavin T (ThT) Fluorescence Assay

Samples for ThT fluorescence assay contained 50 μg/mL of peptide and 10 mM ThT dissolved in 10 mM phosphate buffer (pH 7.4) with 100 mM NaCl and 0.02% (w/v) sodium azide. Peptide samples were added from a pre-sonicated stock that was prepared by dissolving each peptide in 100% HFIP. Whenever needed, heparin stocks were added to a final concentration of 40 µg/mL.⁴² Measurements were performed using VICTOR³ V 1420 Multilabel Counter (Perkin Elmer, Shelton, CT). Solutions were excited at 450 nm and fluorescence emission was monitored at 486 nm. Fluorescence intensity was measured every 15 min for at least 20 hours at 25 °C. A control sample containing all components except the peptide was included in each set of measurements to account for background fluorescence and potential interactions between non-peptide elements.

Circular Dichroism (CD) Measurements and Conformational Analysis

Measurements were performed on a Jasco J-810 spectropolarimeter (Jasco Incorporated, Easton, MD). All CD spectra were taken every 1 h for 24 h at room temperature. An appropriate blank containing all components except the peptide sample was used to correct each CD spectrum. Samples were prepared similarly to those used for the ThT assay without the fluorescent dye. Data obtained were analyzed, based on the profile of the obtained CD spectra, using a deconvolution algorithm found in the Beta Structure Selection (BeStSel) web server. BeStSel is a freely accessible program for secondary structure prediction and fold recognition of CD spectra. It uses an algorithm that accounts for the twisting angles between β-strands to accurately analyze the structure of proteins, especially protein aggregates and amyloid fibrils.⁴³

Fourier Transform Infrared (FTIR) Characterization of Peptide Aggregates

Samples for FTIR measurements were prepared by first incubating 50 µg/mL peptide and 40 µg/mL heparin in 10 mM Tris-HCl buffer (pH 7.4) for at least 20 hours at room temperature.⁴² Aliquots of the incubated samples were applied into a single AgCl cell and allowed to dry in a desiccator to probe the characteristics of the bulk solution. These aliquots will ostensibly contain a mix of aggregated and unaggregated forms of the peptides. Isolated aggregates were prepared by centrifuging aliquots from the incubated samples, removing the supernatant, and washing the aggregates with distilled water. After three washing cycles, the samples were applied on a single AgCl cell and allowed to dry. The AgCl cells, with either the dried-up aliquot samples or isolated aggregates, were then placed in the sample compartment Nicolet 8700 FTIR spectrometer, pre-purged with CO₂-free dry air. The sample spectra were measured using a liquid nitrogen-cooled mercury-cadmium-tellurium detector (Thermo Scientific, Madison, WI). Final FTIR spectra were an average of 128 scans collected at a data spacing of 4 cm⁻¹. The reproducibility of the band position measurements was better than 2 cm⁻¹ based on sample-to-sample variations. To better identify peak positions, a second derivative of the spectra was taken using Omnic 8.3 FT-IR software (Thermo Scientific, Madison, WI).

Fibril Morphologies Probed Using Transmission Electron Microscopy (TEM)

TEM samples were preincubated as described for FTIR except with increased incubation time to maximize fibrillization. Aliquots containing aggregates were deposited onto 300 mesh copper grids with carbon film (Ted Pella, Redding, CA) and negatively stained using 2% uranyl acetate twice. The images were taken using a calibrated Tecnai Spirit 120kV TEM (ThermoFisher Scientific, Hillsboro, OR) equipped with an Eagle 4k CCD camera with a resolution of 2048 × 2048 pixels. The pixel size varies since images were taken at various magnifications.

Molecular Dynamic (MD) Simulations

The SAA_1–13 peptide fragment was obtained from the X-ray-resolved crystal structure of SAA in the Protein Data Bank (PDB ID: 4IP8).²⁶ MD simulations were performed using the GROMACS 2021 software package,⁴⁴ employing the OPLS-AA/L all-atom force field.⁴⁵ The peptide was placed in a cubic box with edges distanced 1 nm away from the peptide. The box was filled with SPC/E modeled water molecules,⁴⁶ and the system was neutralized with sodium and chloride ions. The energy of the system was minimized, and the system was equilibrated twice, first with a constant number of particles, volume, and temperature and then with constant pressure instead of volume. The simulation was run for 100 ns at 300 K with a time step of 2 fs with the data being stored every 2 ps. Particle-Mesh Ewald,⁴⁷ velocity rescaling,⁴⁸ and Parrinello–Rahman⁴⁹ methods were employed in the simulations. The results were analyzed using the GROMACS tools, and the secondary structures were deconvoluted using the Definition of Secondary Structure of Protein (DSSP) algorithm.⁵⁰ Structural trajectories during the simulations were visualized using Visual Molecular Dynamics (VMD)⁵¹ and images at different time frames were generated using PyMOL.⁵²

RESULTS AND DISCUSSION

To probe the roles of the relevant amino acid residues Arg-1, Ser-5, Glu-9, and Asp-12 in SAA_1–13, each residue was systematically replaced with alanine or a structurally conservative analog; Arg-1 was replaced with citrulline (Cit), Glu-9 with glutamine, and Asp-12 with asparagine (Table 1). Once synthesized, we probed the amyloidogenic propensities and conformational characteristics of these mutant peptides in the absence or presence of heparin. Glycosaminoglycans like heparin are known to accelerate the aggregation of SAA peptides even at low concentrations.⁵³ Published MD simulation studies agree that charged residues play an important role in the transition of SAA_1–13 from α-helix to β-hairpin. The reorganization of helix-1 begins with Arg-1 interactions shifting from Ser-5 to Glu-9 which are located on the same face of the helix. Interaction with Ser-5 is quickly lost with Arg-1 subsequently interacting with Glu-9 and Asp-12. Electrostatic interactions between Arg-1 and these two acidic residues favor the peptide adopting a transient bent conformation. The interaction of Arg-1 with Glu-9 rapidly decays, leaving Arg-1 salt-bridged to Asp-12, a characteristic of the final β-hairpin structure.^21–23

Table 1.

Peptide sequences synthesized along with their corresponding mass-to-charge ratio (m/z) from MALDI-TOF measurements.

Peptide	Sequence	Calculated MW	Obtained m/z
SAA1–13	RSFFSFLGEAFDG-NH2	1478.61	1477.67
SAA1–13R1A	ASFFSFLGEAFDG-NH2	1393.50	1394.92
SAA1–13S5A	RSFFAFLGEAFDG-NH2	1462.61	1463.67
SAA1–13E9A	RSFFSFLGAAFDG-NH2	1420.58	1420.92
SAA1–13D12A	RSFFSFLGEAFAG-NH2	1376.57	1377.08
SAA1–13R1Cit	CitSFFSFLGEAFDG-NH2	1479.59	1478.58
SAA1–13E9Q	RSFFSFLGQAFDG-NH2	1477.62	1477.58
SAA1–13D12N	RSFFSFLGEAFNG-NH2	1477.62	1478.50

Amyloidogenic Propensities of the Mutant Peptides

The ThT fluorescence assay of the synthesized peptides (Table 1) revealed that in the absence of heparin, SAA_1–13 and its mutants, except for SAA_1–13S5A, did not aggregate within the 24-hr incubation period (Fig. 1A). Samples of SAA_1–13S5A displayed a relatively sharp increase in fluorescence intensity as a function of time while similarly incubated wild type and mutant samples yielded ThT plots that remained relatively flat even after 24 hrs.

Figure 1.

Time course ThT fluorescence assay of wild type SAA_1–13 and its mutants in the absence (A) or presence (B) of heparin. Data were obtained at room temperature using 50 µg/mL peptide solutions in 10 mM Tris-HCl buffer (pH 7.4) placed in a 96-well plate and run on a plate reader set to measure ThT fluorescence. Samples were excited at 450 nm, and their ThT fluorescence emission intensities were monitored at 486 nm. Measurements were taken every 15 min for at least 20 h. Whenever needed, heparin was added to a final concentration of 41 µg/mL. The blank in each set of measurements contains all components except the peptide.

When incubated with the aggregation-inducing heparin (Fig. 1B), the wild type yielded a ThT assay indicative of fibrilization. Moreover, most of the mutants displayed varying degrees of aggregation propensities relative to the wild type, as indicated by the steepness of the rise in the ThT plots and the maximum saturation points. While samples of SAA_1–13S5A and the Asp-12 mutants were slightly more aggregating than the wild type, Glu-9 mutants aggregated the fastest and yielded relatively the most aggregates. SAA_1–13S5A aggregated faster than SAA_1–13 and both produced about the same aggregate yields after a 24-hr incubation. Substitution of Glu-9 with alanine (SAA_1–13E9A) leads to an increase in the amyloidogenic potential. The ThT data further revealed that samples of SAA_1–13E9Q (Fig. 1B) were four-fold more amyloidogenic than SAA_1–13E9A and even more so compared to the wild-type fragment. Similarly, the alanine substitution of Asp-12 (SAA_1–13D12A) yielded a ThT fluorescence profile that reflects that this mutant is relatively more amyloidogenic than SAA_1–13 but slightly less than SAA_1–13D12N. The differing degrees of mutant amyloidogenicity perhaps hint at the varied role of each residue in the self-assembly pathway.

The replacement of Arg-1 with alanine (SAA_1–13R1A) or citrulline (SAA_1–13R1Cit) significantly reduced amyloidogenic propensity. The ThT fluorescence assay (Fig. 1B) revealed that for these Arg-mutants, the aggregation was significantly lower and closely resembles that of the negative control. Clearly, aggregation of Arg-1 mutants was limited regardless of whether the samples were incubated with or sans heparin.

Time-Dependent Changes in the Secondary Structure of the Mutants

To better understand the differences in the amyloidogenicity of the different mutants relative to the wild-type fragment, CD spectra were collected every hour for 24 hours to see the time-dependent modulation of the peptide conformation. When co-incubated with heparin (Fig. 2), the initial CD spectra of SAA_1–13 (Fig. 2A) revealed minima around 205–206 and 220 nm indicating a mixture of different conformations. Several studies have shown that the transition of α-helical structures to random coils shifts the 208 nm negative band to a lower wavelength.^{54, 55} Moreover, a study on the amyloid formation of protein albebetin shows that an increase (more negative) in intensity at ca. 205 nm relative to the ellipticity at 220 nm indicates the presence of intermediate aggregates composed of various structures including β-sheets.⁵⁶ As time elapses, the CD spectra of SAA_1–13 show a decrease in the ratio of ellipticity taken at 205 and 220 nm, indicating a consequent transformation to amyloid fibrils. In contrast, the CD spectra of SAA_1–13R1A (Fig. 2B) and SAA_1–13R1Cit (Fig. 2C) revealed peptides that have random coil conformation, as evidenced by a minimum near 200 nm and low ellipticity above 215 nm,⁵⁷ which persisted with incubation time. The small negative band around 229 nm indicates the presence of β turns.^{58, 59} The CD spectra confirmed that Arg-1 mutants do not yield detectable aggregates after 24 hours of incubation regardless of the presence of heparin.

Figure 2.

CD spectra of peptides (A) SAA_1–13, (B) SAA_1–13R1A, (C) SAA_1–13R1Cit, (D) SAA_1–13S5A, (G) SAA_1–13E9A, (H) SAA_1–13E9Q, (I) SAA_1–13D12A, and (J) SAA_1–13D12N incubated in heparin and 10 mM phosphate buffer (pH 7.4). CD spectra of peptides (E) SAA_1–13 and (F) SAA_1–13S5A incubated without heparin are also shown. CD measurements from 200 nm to 250 nm were done every hour for 24 hours at room temperature.

In agreement with published MD simulation studies,^{21, 23} the absence of Arg-1 makes it impossible to form the necessary salt-bridge that stabilizes the key bent intermediate of the misfolding pathway. Implications of the ThT assays and CD data on the Arg-1 mutants also corroborate published in vitro studies on various truncated SAA peptides. SAA_2–12 and SAA_2–15 have been found to display reduced aggregation propensity compared to SAA_1–12.⁶⁰ Placing a methionine residue on the N-terminal of SAA_1–12 revealed a similar reduction in aggregation propensity.⁶⁰ Clearly, these results highlighted the importance of the free guanidino group on Arg-1 in stabilizing the purported final β-hairpin structure.^{21, 23} Jannone et al. substituted glutamine for Arg-1 in SAA_1–11 (SAA_1–11R1Q) and found that the mutant peptide formed globular amorphous aggregates with low ThT fluorescence intensity and weak FTIR amide β-sheet peaks.⁶¹ These outcomes indicated that the glutamine-replaced Arg-1 mutant was weakly aggregating and yielded aggregates of various conformations. Altogether, our findings raised the plausibility that the amyloidogenesis of SAA_1–13 peptides is facilitated by molecular interactions involving Arg-1.

The behavior of SAA_1–13S5A was dependent on whether it was co-incubated with heparin. In the presence of heparin, the amyloidogenicity and conformational variation of SAA_1–13S5A (Fig. 2D) appear to support a higher amyloid potential than a similarly incubated wild-type fragment (Fig. 2A). Their CD spectra imply structures that are initially a mixture of conformers, as indicated by negative peaks in the regions 200 to 210 cm⁻¹. As time lapses, the conformation shifts to one dominated by β-structures, evidenced by a decrease in ellipticity at ca. 205 nm and a negative peak at 227 nm.^{56, 58, 59} Though both CD spectra failed to visibly show the β-sheet bands around 218 nm, the deconvolution algorithm (BeStSel web server) still recognized the transition to an increasing β-sheet composition as a function of time (Table S1).⁴³ An analysis of the CD data taken in the presence of heparin indicates that SAA_1–13S5A formed a higher percentage of β-structures than SAA_1–13. Moreover, it was clear that within two hours, the S5A mutant reached its highest β-sheet content while wild-type samples required five hours to reach this apex. A closer inspection of the ThT assays revealed that within the first five hours, the S5A mutant aggregated faster relative to the wild-type fragment (Fig. S1). When heparin is present, it appears that replacing Ser-5 with alanine makes the SAA_1–13 fragment more amyloidogenic with structures quickly dominated by β-structures.

Interestingly, without heparin, SAA_1–13S5A appeared to be more amyloidogenic while SAA_1–13 did not appear to aggregate. Under these conditions, the amyloidogenic difference between SAA_1–13 and SAA_1–13S5A was more evident compared to when the peptides were incubated with heparin. While the corresponding CD spectra of SAA_1–13 (Fig. 2E) suggest that its conformation remains mainly random coil regardless of incubation time, the conformation of SAA_1–13S5A (Fig. 2F), even at the onset, was dominated by β-rich structures with negative bands at 217 nm (antiparallel β-sheet) and 227 nm (β-turns)^57–59 that persisted during the 24-hr incubation. Deconvolution of the initial CD data indicated that compared to the wild type, the SAA_1–13S5A mutant sans heparin exhibited a substantially higher percentage composition of β-structures that only increased with time (Table S1).⁴³ The results of measurements involving SAA_1–13S5A suggest a pivotal role for Ser-5 in SAA_1–13 fibrilization.

The findings also align with the MD simulations previously done on SAA_1–13S5A wherein the mutant exhibited a higher frequency of β-hairpin conformation compared to the SAA_1–13 wild type.²³ Based on the published MD simulation models, Arg-1 interacts with Ser-5 and Glu-9, stabilizing the helical structure of SAA_1–13. The eventual dissolution of Arg-1 and Ser-5 contacts causes the peptide helix to bend and form the β-hairpin conformation.^{21, 23} The results of MD simulations performed in the lab on SAA_1–13 (Fig. S2A) and SAA_1–13S5A (Fig. S2B), indicated that initially, Arg-1 in SAA_1–13S5A still makes contact with Glu-9 even in the absence of Ser-5. However, it is evident that the structure is less helical compared to the model SAA_1–13. Moreover, in the modeled SAA_1–13S5A dynamics, Arg-1 quickly (~10 ns) shifts to form stable contacts with Asp-12 while the same occurrence took almost ~60 ns in the modeled wild type. During the 100 ns simulation, β-sheets were observed for the modeled SAA_1–13S5A (Fig. S3B) but not for the wild type (Fig. S3A). If indeed Ser-5 helps stabilize the existing helix, it is perhaps logical to expect that alanine replacement of Ser-5 increases amyloidogenicity and would hasten the shift to more β-structures.

While the data on the SAA_1–13S5A mutant supports the MD simulation-based model of self-assembly, aggregation tendencies were clearly dependent on the presence or absence of heparin. It should be noted that the in silico models did not factor in heparin in their MD simulations.^{21, 23} Hence, it cannot be ruled out that the fibrilization process of SAA_1–13S5A with heparin follows an aggregation pathway different from when heparin is absent. It is clear, however, that the replacement of Ser-5 with alanine destabilizes the structure, increases amyloidogenicity, and hastens β-sheet formation.

Replacement of the acidic residues Glu-9 and Asp-12 also revealed clear changes in how the secondary structure of the mutant peptides evolved with time as reflected in their CD spectra. Though SAA_1–13E9A and SAA_1–13E9Q failed to aggregate in the absence of heparin (Fig 1A), the ThT assays taken with heparin (Fig. 1B) revealed that Glu-9 mutants were evidently more amyloidogenic than the wild type fragment. The corresponding CD spectra (Figs 2G and 2H) revealed that at the onset, the structures were already predominantly β-sheets as indicated by the ellipticity around 218 and 227 nm.^57–59 Moreover, the data also reflected fast transitions to β-structures as shown by the intensifying negative bands at 218 and 227 nm. Clearly, the change is more prominent for SAA_1–13E9Q than SAA_1–13E9A, thus corroborating the ThT data. The deconvolution of the CD spectra shows that the change in the structural composition of SAA_1–13E9A was closely similar to that of the non-heparin-induced SAA_1–13S5A mutant (Table S1).⁴³ Meanwhile, the CD spectra of SAA_1–13E9Q implied the highest rate of β-structure formation and final percent β-sheet content. The distinct amyloidogenic propensities and observed conformational distribution indicate that the charge at residue-9 helps mediate the stabilization of the helix. An uncharged residue in position 9 led to higher amyloidogenicity and a quicker transition to β-sheets. The absence of charge on residue-9 appears to quicken the transition to β-hairpins, characterized by interactions between Arg-1 and Asp-12. A more conservative Glu-9 to glutamine replacement appears to best facilitate the transition to β-hairpins than with alanine at position-9.

According to the in silico models, the acidic residue Glu-9 is not only one of the helix-stabilizing partners of Arg-1, but it also facilitates the transition to the final Arg-1 and Asp-12 interaction.^{21, 23} Data collected indicate that mutating out Glu-9 favor the formation of β-structures. Without the negative charge at position 9, Arg-1 cannot form the necessary helix-stabilizing salt bridge. This promotes Arg-1 interaction with Asp-12, which in turn promotes β-hairpin structures. Our simulation data showed that replacing Glu-9 resulted in similar molecular dynamics as that of the modeled SAA_1–13S5A. With Glu-9 replaced, the initial Arg-1 and Ser-5 interaction remains. However, early structures of SAA_1–13E9Q and SAA_1–13E9A were much less helical (Fig. S2C and S2D, respectively) compared to that of the wild type (Fig. S2A). Moreover, after ~10 ns in the 100-ns simulation window, Arg-1 in SAA_1–13E9Q forms stable contacts with Asp-12. This was significantly faster than what was obtained for the modeled SAA_1–13E9A, where stable Arg-1 and Asp-12 interaction appeared after ~24 ns. These were much quicker transitions than was reflected from the modeled SAA_1–13, ~60 ns.

It is therefore not surprising to find that Glu-9 substitution hastens fibril formation. The effect is magnified in the conservative Glu-9 to glutamine substitution, possibly because the uncharged glutamine somehow facilitates the formation of the β-hairpin favoring Arg-1 to Asp-12 interaction. Perhaps, when alanine replaces Glu-9, Arg-1 is not immediately channeled to Asp-12. Nevertheless, the Arg-1/Asp-12 ion pair eventually forms without the helix stabilizing salt bridge with Glu-9. Our MD simulation data shows that in SAA_1–13E9Q, Gln-9 facilitated the quick formation of the interaction between Arg-1 and Asp-12 (Fig. S2C). This was not as evident when we modeled the dynamics of SAA_1–13E9A. Published computational studies on SAA_1–13E9A²³ and SAA_1–27E9A²² show that for these mutants, more β-rich configurations formed relative to the wild-type fragments. Moreover, in vitro work by Jannone et al. showed evidence that the substitution of glutamine for Glu-9 in SAA_1–11 (SAA_1–11E9Q) still yielded detectable amyloid fibrils that had substantial structural heterogeneity.⁶¹

In the presence of heparin, the substitution of Asp-12 yielded CD data that appeared to depend on whether it was replaced by alanine or asparagine. Again, based on the ThT assay taken in the presence of heparin, the Asp-12 mutants were more amyloidogenic than the wild-type fragment. These mutant peptides did not aggregate without heparin. When incubated with heparin, the CD spectra of SAA_1–13D12A (Fig. 2I) appeared to be mostly random coils that edged towards a β-sheet conformation as time evolved, like the wild type (Fig. 2A) albeit at a much slower rate. There was a stronger resemblance between the CD spectra of SAA_1–13D12N (Fig. 2J) and SAA_1–13 (Fig. 2A), which visibly reflects the conformational transition to β-structures. Analysis using BeStSel⁴³ shows that the SAA_1–13 and its corresponding Asp-12 mutants eventually formed approximately the same percentage of β-structures when incubated with heparin. The deconvolution algorithm also revealed that SAA_1–13D12A transitioned to β-sheet at the slowest rate (Table S1). The CD spectra for SAA_1–13D12A (Fig. 2I) implied the eventual formation of β-structures, but the time evolution of this structural transition is significantly slower such that very different time-dependent profiles were observed. The prediction tool was able to establish the eventual formation of β-structures after longer incubation.

At first glance, it appeared that the data collected for the Asp-12 were at odds with the in silico model^{21, 23} especially if indeed the Arg-1/Asp-12 ion pair is critical in stabilizing the final β-hairpin structure. Without the negatively charged aspartic acid in position 12, a salt bridge with Arg-1 would be impossible and therefore should render the Asp-12 mutants less amyloidogenic with fewer β-structures. However, it is possible that alternative interactions and conformations are exploited to allow the misfolding trajectory towards fibrilization.

The similarity in the time evolution of secondary structures for SAA_1–13D12N and SAA_1–13 also raised the plausibility that Arg-1 still forms a final interaction, albeit not through a salt bridge, with residue-12. Perhaps Arg-1 and asparagine form hydrogen bonding and/or hydrophobic interactions that replace the salt bridge. Asparagine might even be preferred over aspartic acid as it alleviates the possible repulsive effect of Glu-9 within the vicinity allowing the mutant to form the final β-hairpin spontaneously. While the in silico model predicted Asp-12 and Glu-9 to be relatively distant, there is an instant in the proposed misfolding pathway that these acidic residues are simultaneously coordinated with Arg-1 as it traversed towards the bent intermediate.^{21, 23} MD simulations done in the lab showed that for SAA_1–13D12N, the eventual loss of Arg-1 and Ser-5 contact edges Arg-1 to interact with both Glu-9 and Asn-12 (Fig. S2E). This resulted in the formation of a bent intermediate at ~20 ns of a 100-ns simulation window. As the peptide folds into the final β-hairpin structure, the contact between Arg-1 and Asn-12 dissolves with Arg-1 retaining its salt bridge with Glu-9 until the simulation terminates.

In contrast, the dynamic simulation of SAA_1–13D12A (Fig. S2F) revealed that with alanine in position-12, the relevant interaction with Arg-1 was not possible. Hence, the bent intermediate was not observed until ~42 ns into the simulation thus delaying the formation of the final β-hairpin structure. The alternative interactions mentioned above would certainly be weaker if alanine occupies position-12. Alanine is too small and hydrophobic to mediate the formation of the purported Arg-1 to residue-12 interaction hence the much-delayed appearance of β-structures relative to that of the wild type and SAA_1–13D12N. Published in silico models on the low energy model of SAA_1–13 predicted a final β-hairpin reinforced by the interaction between the N-terminal amino group and Ser-5. This perhaps contributes to the stability of β-hairpin formed in SAA_1–13D12A.^{21, 23} However, this was not evident in the MD simulations we performed on SAA_1–13D12A. Nevertheless, without Asp-12 it appears that Arg-1 still mitigated β-hairpin formation involving a Glu-9 salt bridge. Interestingly, Asp-12 mutants of SAA_1–12, SAA_1–12D12A, and SAA_1–12D12L, were able to form amyloid fibrils with morphology similar to that of the wild type.⁶²

Comparing the Secondary Structures of the Peptide in Solution and When Aggregated

To delineate the conformational populations of each mutant in the bulk solution and when aggregated, FTIR measurements were done. It has long been established that the peptide Amide I (1611 – 1690 cm⁻¹),^63–65 Amide II (1520 – 1540 cm⁻¹),⁶⁶ and Amide III (1220 – 1240 cm⁻¹)^{61, 67, 68} modes help characterize secondary structures. The FTIR spectra, displayed as a double derivative to accentuate peak positions, were taken for aliquots of each peptide solution incubated with heparin and allowed to dry onto an AgCl window (Fig. S4). The peak positions of Amides I, II, and III obtained for each peptide are summarized in Table 2.

Table 2.

Summary of FTIR data for the sample bulk solutions of SAA_1–13 and its associated mutants. FTIR of the washed fibrils are also shown for samples that displayed helical bands on the FTIR spectra of their bulk solutions.

Preparation	Peptide†	FTIR (cm−1)‡
		amide III	amide II	amide I
Bulk Solution	SAA1–13	1230	1527	1626, 1697
	SAA1–13S5A	1230	1527	1628, 1697
	SAA1–13D12N	1230	1527	1626, 1697
	SAA1–13D12A	1230	1529	1626, 1697
	SAA1–13E9Q	1234, 1315	1547	1626, 1656, 1680, 1699
	SAA1–13E9A	1234, 1313	1545	1626, 1656, 1680, 1699
	SAA1–13R1Cit	1242	–	1624, 1658, 1681, 1697
	SAA1–13R1A	1242	–	1624, 1655, 1678, 1699
	SAA1–13S5A (no heparin)	1230, 1315	1527, 1547	1625, 1656, 1680, 1697
Washed Fibrils	SAA1–13E9Q	–	1527	1626, 1697
	SAA1–13E9A	–	1527	1626, 1697
	SAA1–13S5A (no heparin)	–	1527	1625, 1697

^†Samples were incubated with heparin except for the last entry.

^‡Amide I, II, and III peaks taken as the minima of the second-derivative spectra.

^§Amyloid morphology as deduced from the TEM data.

The bulk solution of SAA_1–13 depicts peaks at 1697 and 1626 cm⁻¹ in the Amide I region, 1527 cm⁻¹ for Amide II, and 1230 cm⁻¹ for III region, that indicates the presence mainly of β-sheet structures (Table 2, Fig. S4). Barth (2007)⁶⁵ also observed the two peaks in the amide I region for β-sheet structures and explained that it may be due to transition dipole coupling, through-bond coupling, and hydrogen bonding. The aliquot solutions of amyloidogenic mutants SAA_1–13S5A, SAA_1–13D12A, and SAA_1–13D12N incubated with heparin were distinctly composed of β-sheet conformers analogous to SAA_1–13 wild type (Table 2, Fig. S4). This corroborates analyses of the ThT assay and CD spectroscopy findings. In the presence of heparin, these peptide mutants, while slightly more amyloidogenic, followed the same conformational transition towards fibril formation, albeit at varying rates. The FTIR data confirmed that the bulk solutions of these mutants, like the wild-type fragment, are dominated by β-sheets.

Peptides SAA_1–13R1Cit and SAA_1–13R1A did not readily form detectable fibrils under the experimental conditions regardless of whether each was incubated with heparin or not. FTIR spectra of an aliquot of these Arg-1 mutants (Table 2, Fig. S4), which were much weaker, revealed multiple amides II and III peaks indicating a mixture of different conformers. Additionally, the amide I region displayed band markers for both α- and β-conformers as well as random coils (1660 – 1685 cm⁻¹).⁶⁷ The presence of markers for the β-structures may be attributed to some aggregated peptides. According to a previous study,⁶² SAA can still form fibrils even with the deletion of Arg-1 when incubated at pH 2 and when a 10 mg/mL peptide concentration is used. It is clear, however, that the Arg-1 mutants did not behave like the wild-type or pseudo-wild-type peptides, which formed aggregates dominated by β-sheets.

Conversely, aliquots of SAA_1–13E9Q and SAA_1–13E9A mutants incubated with heparin as well as the non-heparin-induced samples of SAA_1–13S5A yielded FTIR spectra that indicated the presence of both α- and β-conformers (Table 2, Fig. 3A). In addition to the marker bands indicating the presence of β-sheets, the FTIR spectra of these mutants revealed peaks at 1650 – 1660 cm⁻¹ in the Amide I region,⁶⁹ 1540 – 1550 cm⁻¹ in the Amide II region,⁶⁶ and ~1300 cm⁻¹ in the Amide III region, which are indicative of α-helical structures. Plotted alongside the FTIR data of the aggregated peptides (Fig. 3B), the fibrils formed by the mutants evidently have β-sheet conformations only, with peaks at 1626 and 1697 cm⁻¹. This proves that the helical structures detected in the bulk solution are transitory and are not part of the fibrils.

Figure 3.

FTIR spectra, displayed as its second derivate to accentuate the peaks, of the amide I region of (A) aliquots taken from SAA_1–13 mutants peptide solutions which exhibited formation of helical structures and (B) washed aggregates collected from the same peptide samples.

These Ser- and Glu-mutants, i.e. SAA_1–13E9Q and SAA_1–13E9A mutants incubated with heparin as well as the non-heparin induced samples of SAA_1–13S5A, were also highly amyloidogenic and readily form β-structures based on CD data. Without the helix stabilizing residues Ser-5 or Glu-9, the structure of SAA_1–13 unravels with Arg-1 being channeled to form interactions with Asp-12 and the bent helix.^{21, 23} The helical structures detected in the FTIR data of the bulk solution could be (1) the parent stable helix that is in equilibrium with the bent structures or (2) an intermediary labile helix transitioning from the stable helix to β-hairpins. Since these mutants are highly amyloidogenic, it is less likely that there will be a detectible stable helix. If an equilibrium exists between the stable helix and β-hairpins, it will be well shifted towards the β-hairpin side. It is more plausible that the helix structures detected in the FTIR spectra of the bulk solution are intermediary labile helix. It is possible that the weaker Arg-1 interaction led to the formation of labile helical intermediates which promoted amyloid formation. A logical consequence if we consider three moieties, labile α-helices, β-structures, and fibrils, that are in equilibrium. Accordingly, an accumulation of labile α-helical conformers shifts the equilibrium toward increased fibril formation. It has previously been identified that extensive local changes in conformation prior to self-aggregation are not essential in the fibril formation of some proteins. In fact, in most amyloidogenic proteins, the rate of fibril growth is directly proportional to the number of α-helical intermediates that can self-assemble into oligomers.^70–73 Vestergaard et al. investigated the insulin amyloid formation mechanism and found that the growth rate of fibrils and fibril elongation were dependent on the concentration of helical oligomeric species present in the solution. They also proposed that these helical oligomers could be involved in secondary nucleation mechanisms such as branching, fragmentation, and nucleation on the fibril surface.⁷⁴

Morphology of the Aggregated Peptides

The morphology of the aggregates appears to vary with each mutant and whether heparin is present. SAA_1–13 form long thin fibrils that are closely packed (Fig. 4A). TEM data of heparin-incubated SAA_1–13S5A (Fig. 4B), SAA_1–13D12A (Fig. 4C), and SAA_1–13D12N (Fig. 4D) were like that of the wild type. The fibrils formed by each mutant were less compact and more elongated. At higher magnification (Fig. S6), the right-handed twist is quite visible for all heparin-incubated samples above, with SAA_1–13D12A forming thinner and longer fibrils compared to those of SAA_1–13, SAA_1–13S5A, and SAA_1–13D12N. These observations are consistent with previously published TEM analyses showing that various SAA fragments also exhibit fibril chirality and amyloid polymorphism like other amyloidogenic peptides.^{61, 62, 75} Perhaps the relatively slower rate of aggregation, evident from the spectroscopic data, allowed for long-thin fibrils to develop.

Figure 4.

TEM images of the peptides solutions of (A) SAA_1–13, (B) SAA_1–13S5A, (C) SAA_1–13D12A, (D) SAA_1–13D12N, (E) SAA_1–13E9A, and (F) SAA_1–13E9Q, (H) SAA_1–13R1A, and (I) SAA_1–13R1Cit after incubation with heparin. TEM of (G) SAA_1–13S5A fibrils from the solution incubated without heparin is also shown. All samples were incubated for 24 hours in Tris-HCl buffer (pH 7.4) at room temperature. Images were taken from aged samples deposited on 300 mesh copper grids with carbon film negatively stained with 2% uranyl acetate. Scale bars = 1 µm and 2µm in H and I, respectively, and 500 nm in other panels.

In contrast, aggregates of SAA_1–13S5A incubated without heparin appear to assume morphologies different from the heparin-incubated wild type, SAA_1–13S5A, and Asp-12 mutants discussed above. Without heparin, SAA_1–13S5A peptides formed multiple short amyloid fibrils that were mostly clumped together (Fig. 4G). Interestingly, the TEM data of Glu-9 mutants also exhibited numerous very short fibrils that were lumped in “small colonies” like heaps (Figs. 5E & 5F). These heaps of aggregates were scattered more through the visible frame and at times more overlapping compared to the Ser-5 mutant incubated without heparin. Nonetheless, there appeared to be more nucleation sites for fibril formation, most likely because of the abundance of labile helical intermediates, as evidenced by the spectroscopic data. It is possible that the presence of abundant labile helical structures, which are thought to initiate fibrillogenesis,^{70, 72, 74} provided numerous nucleation sites that compete for limited monomeric structures needed for fibril elongation.

Figure 5.

Rendering of a plausible misfolding mechanism for SAA_1–13-based peptides. (A) SAA_1–13 helices start to unravel to form (B) a labile helical intermediate that can self-assemble into (C) oligomeric species. Labile helices then transition to (D) bent intermediates. When the critical concentration is reached, the peptide spontaneously folds to a (E) β-hairpin (U-shaped) structure which pile up and grow to (F) mature amyloid fibrils with the characteristic cross-β motif (F).

These differences in morphology may explain why their respective time-course CD spectra were not identical. Perhaps the changes in the misfolding pathway due to mutation could account for the difference in fibril morphology based on TEM images. It appears that the morphology of the aggregates generated depends on how the mutation affects the path of misfolding. Polymorphism, which is naturally occurring among amyloid fibrils formed by peptides and proteins, complicates the analyses and understanding of amyloid formation. Polymorphism was thought to be attributed to changes in the environment, but studies have shown that it can be observed under the same conditions. The complexity and occurrence of polymorphism, in fact, increases with the level of protein structure.⁷⁶ A study on KFFEAAAKKFFE, a sequence derived from amyloid-forming fragments of different amyloidogenic proteins, showed that mutations yielded fibrils of varying morphologies. For instance, alanine replacement of lysine increased the amyloidogenic potential of the peptide but made the fibrils narrower. They surmised that alterations in charge and hydrophobicity could affect the resulting ultrastructure of amyloid fibrils.⁷⁷

CONCLUSIONS

We employed single amino acid mutations targeting the important residues Arg-1, Ser-5, Glu-9, and Asp-12 to investigate changes in molecular interactions that define their roles in the aggregation of SAA helix-1. The truncated SAA_1–13 fragment, which displayed similar aggregation characteristics as the full-length helix-1 domain SAA_1–27, was systematically modified by substituting the key residues with alanine or with uncharged but structurally similar amino acids. Replacing Arg-1 with alanine or citrulline resulted in much reduced aggregation propensity and a lack of detectable β-structures alluding to the importance of salt-bridge interactions involving Arg-1. Regardless of whether the aggregate-inducing heparin is present, replacing Ser-5 with alanine enhanced the aggregation propensity and β-sheets preference of SAA_1–13. This hints at a pivotal role for residue-5 in modulating the aggregation of SAA_1–13.

The wild-type fragment along with the rest of the mutants studied required incubation with heparin to induce aggregation. When co-incubated with heparin, mutating out Glu-9 increased the aggregation potential. This suggests that without a negative moiety on position 9, the kinetics of the misfolding pathway were accelerated. Interestingly, the enhancements were magnified when Glu-9 was replaced by glutamine instead of alanine. MD simulation based on the Ser- and Glu-mutants hinted that these Ser-5 and Glu-9 residues stabilize the helical form of SAA_1–13. It is not surprising that mutating out either Ser-5 or Glu-9 hastens aggregation, with the helix-stabilizing interactions absent, the conformation readily transitions to β-hairpins. MD simulation also revealed that glutamine at position-9 forms temporary interaction with Arg-1 helping with the transition to β-sheet. This was not possible when alanine replaced Glu-9.

In contrast, Asp-12 mutants displayed levels of aggregation closely at par with that of the wild type. Furthermore, detection of the β-conformers in the bulk solution was delayed when alanine, instead of asparagine, replaced Asp-12. MD simulations revealed marginal differences in β-composition of the wild type and Asp-12 mutants. Interestingly, the transition to β-hairpin is slower for the alanine-substituted mutant relative to when asparagine replaced Asp-12. Nevertheless, Glu-9 makes the final salt bridge Arg-1 in both mutants.

When the behaviors of each mutant peptide were analyzed, they provided evidence consistent with the aggregation pathway proposed by previously published models, which highlighted the important role of temporal molecular interactions formed by Arg-1 with Ser-5, Glu-9, and Asp-12, respectively.^{21, 23} Bernhardt et al.²³ used MD simulations on alanine mutants of SAA_1–13 where they systematically replaced the highlighted residues. Their findings were mostly consistent with our solution studies as well as the in silico model²¹ except that for SAA_1–13S5A and both Glu-9 mutants, they did not observe the transition through helical conformers. FTIR analyses presented here clearly revealed α-conformers in the bulk solutions of SAA_1–13S5A (sans heparin), SAA_1–13E9A, and SAA_1–13E9Q. These findings highlight the subtleties that experimental studies may provide as our work provided further evidence revealing a pathway involving helical intermediates. Our findings presented additional context over previously reported computational works as we were able to probe the actual chemical nature of the transition to amyloid fibrils as they are modulated by each systematic mutation.

It is now possible to envision a mechanism that corroborates the in silico models and clarifies the roles of the interactions formed by the key amino acids. Building upon the misfolding pathway first proposed by Karamanos et al.,⁷⁸ we believe that the SAA_1–13 helix is stabilized by the initial interaction of Arg-1 with Ser-5 and Glu-9 (Fig. 5A). The helix then begins to weaken as Arg-1 leaves Ser-5 (Fig. 5B). This results in the formation of labile helical intermediates which consequently self-assemble into oligomers (Fig. 5C). Arg-1 then forms salt-bridges with both Glu-9 and Asp-12 as the peptide folds into a bent intermediate (Fig. 5D). When the critical concentration of the intermediates is reached, the structure devolves into a β-hairpin dominated conformation wherein only the Arg-1 and Asp-12 salt-bridge are conserved (Fig. 5E). Resulting β-hairpins then pile up to generate the critical nucleus which commences the rapid growth of amyloid fibrils (Fig. 5F). During the nucleation phase, the dynamic equilibrium between two dominant states, labile helical intermediates and β-rich structures, appears to determine the rate of fibril formation. Mutations such as converting Ser-5 to alanine or Glu-9 to glutamine seem to weaken the helix and increase the population of labile helical intermediates, thereby increasing the overall rate of aggregate formation. Moreover, replacing Asp-12 with asparagine appears to somewhat promote aggregation signifying perhaps that the final Agr-1/Asp-12 salt bridge is not critical. We have identified a peptide system that potentially can be modulated to help unravel some of the critical conformational changes associated with the fibrillization pathway. Knowing the role of these residues may facilitate the development of inhibitors of SAA aggregation.

Finally, it must be recognized that the in silico models do not consider the presence of heparin. With heparin present, an entirely different self-assembly mechanism may be at work with disparately different intermediates involved in the aggregation process. Whether heparin merely serves as a platform to increase the relative local concentration of peptide to promote aggregation or provides an alternative self-assembly pathway is not clear. In either case, our findings reveal that Arg-1, Ser-5, Glu-9, and Asp-12 play a significant role in the aggregation of SAA_1–13. Studies are underway in our to probe the longer SAA fragments, SAA_1–27 and SAA_1–76. The former is the full helix-1 chain while the latter truncated polypeptide found in amyloid deposits of SAA.

Highlights.

• SAA_1–13 fragment displays aggregation characteristics similar to the known behavior of the full-length helix-1 SAA_1–27 sequence

• Amyloidogenic propensity of SAA_1–13 is significantly reduced by Arg-1 mutation, suggesting the importance of salt-bridge interactions involving Arg-1 to SAA_1–13 amyloid formation

• Asp-12 substitutions reveal that alternative interactions and conformations may be exploited to allow the misfolding trajectory toward fibrilization

• Mutation of either Ser-5 or Glu-9 enhances the amyloidogenicity of SAA_1–13 and facilitates the formation of labile helical intermediates

• Favoring the formation of labile helical intermediates appears to promote the amyloid fibril formation by SAA_1–13

ABBREVIATIONS

SAA

serum amyloid A

TFA

trifluoroacetic acid

DMF

dimethylformamide

TIPS

triisopropylsilane

HFIP

1,1,1,3,3,3-hexafluoroisopropanol

Cit

citrulline

MALDI-TOF

matrix-assisted laser desorption/ionization-time of flight

ThT

thioflavin T

CD

circular dichroism

FTIR

Fourier transform infrared

TEM

transmission electron microscopy

MD

molecular dynamics

DSSP

Definition of Secondary Structure of Protein

BeStSel

beta structure selection

VMD

Visual Molecular Dynamics