Analysis of the Amyloidogenic Potential of Pufferfish (Takifugu rubripes) Islet Amyloid Polypeptide Highlights the Limitations of Thioflavin-T Assays and the Difficulties in Defining Amyloidogenicity

Abstract

Islet amyloid polypeptide (IAPP, amylin) forms pancreatic amyloid in type-2 diabetes, a process that contributes to the loss of β-cell mass in the disease. IAPP has been found in all higher organisms examined, but not all species form amyloid and the ability to do so correlates with the primary sequence. The amyloidogenic potential of fish IAPPs have not been examined, although fish have been proposed as a source for xenobiotic transplantation. The sequence of pufferfish IAPP (Takifugu rubripes) is known and is the most divergent from human IAPP of any reported IAPP sequence, differing at eleven positions including seven located within residues 20 to 29, a segment of the molecule that is important for controlling amyloidogenicity. Several of the substitutions found in pufferfish IAPP are non-conservative including Ser to Pro, Asn to Thr, Ala to Tyr and Leu to Tyr replacements and several of these have not been reported in mammalian IAPP sequences. Amyloid prediction programs give conflicting results for pufferfish IAPP. CD spectroscopy, FTIR, and transmission electron microscopy reveal that pufferfish IAPP forms amyloid and does so more rapidly than human IAPP in tris buffer at pH 7.4, but does so more slowly in phosphate buffered saline (PBS) at pH 7.4. Molecular dynamics simulations indicate that the pufferfish sequence is compatible with models of IAPP amyloid. The fish polypeptide does not significantly bind to thioflavin-T in tris and does so only weakly in PBS. The results highlight difficulties with thioflavin-T assays and the ambiguity in defining amyloidogenicity.

Amyloid formation and aberrant protein aggregation play important roles in a diverse range of disorders including Alzheimer’s disease, Parkinson’s disease, and type-2 diabetes. The neuropancreatic polypeptide hormone human islet amyloid polypeptide (h-IAPP, also known as amylin) forms islet amyloid in the pancreas in type-2 diabetes.^1–3 The process of islet amyloid formation contributes to β-cell dysfunction in the disease and to the failure of islet transplants.^1–9 Mature IAPP is a 37 residue polypeptide which has been found in all higher organisms examined and is highly conserved, but not all species form amyloid and the ability to do so correlates with the primary sequence of IAPP. Teleostean fish have been proposed as a source for xenobiotic transplantation and for studies of diabetes mellitus, in part because in teleostean fish the pancreatic exocrine tissue is separated from the pancreatic endocrine cells, facilitating the isolation of the endocrine cells.^10–15 However, the amyloidogenicity of fish IAPP amyloidogenicity is not well understood and this is an issue for islet transplantation.^{8, 9} The amyloidogenic potential of any full-length fish IAPP has not been examined, although a ten residue fragment of salmon IAPP has been studied experimentally and larger fragments derived from different fish have been analyzed computationally.^{11, 12} The complete sequence of pufferfish IAPP (Takifugu rubripes) is known and diverges the most from human IAPP of any reported IAPP sequence, differing at eleven positions, including multiple substitutions in regions believed to be important for amyloid formation (Figures-1, 2).

Figure-1.

(A) Comparison of the primary sequence of human IAPP and pufferfish IAPP. Residues which differ from the human peptide are colored red. Both sequences have amidated C-termini and contain a disulfide bond between residues 2 and 7. (B) Cross-section of a model of the human IAPP amyloid fibril based on crystal structures of fragments of the molecule. Residues in pufferfish IAPP which differ from the human sequence are built into the model and are shown in space filling format. (C) Ribbon diagram of the fibril.

Figure-2.

Primary sequences of IAPP from different species. Residues that differ from the human sequence are colored. Substitutions which are in pufferfish IAPP but not in mammalian IAPP sequences are color coded blue. All of the variants contain a disulfide bridge between Cys-2 and Cys-7 and all have an amidated C-terminus. Primates and cats have been reported to develop diabetes and form islet amyloid, but rodents do not. Dog IAPP does form islet amyloid in canine pancreatic endocrine tumors.⁵⁹ Ferret and porcine IAPP are reported to be significantly less amyloidogenic than human IAPP. The ability of baboon, bear, bovine and spectacled bear IAPP to form amyloid in vitro has not been investigated. Islet amyloid is found in the degu, a rodent, but degu islet amyloid is derived from insulin. Only partial sequences are available for rabbit, hare, wolffish, zebrafish and Atlantic salmon IAPP.

Here we examine the ability of pufferfish IAPP (p-IAPP) to form amyloid in vitro. The analysis reveals that the relative amyloidogenicity, as defined by the kinetics of amyloid formation, of human and pufferfish IAPP depend on solution conditions and thus amyloidogencity is context dependent. The analysis also highlights complications with the widely applied thioflavin-T assay of amyloid formation and provides more evidence that they can lead to false negatives.

Amyloid formation in vitro is commonly followed using fluorescence detected thioflavin-T binding assays.^{16, 17} Thioflavin-T is a small dye whose quantum yield increases and whose emission maximum shifts upon binding to amyloid fibrils.¹⁸ There are no high resolution structures of thioflavin-T bound to amyloid fibrils, but the dye is believed to bind to the surface of the cross-β structure of the amyloid fibrils.^17–21 The parallel, in register β-sheet structures of typical amyloid fibrils create a series of grooves that run parallel to the long axis of the fibril and these are believed to form the thioflavin-T binding sites. Binding of the dye fixes the position of the dimethylaminobenzyl and benzothiazole rings of thioflavin-T and reduces self-quenching, thereby leading to the enhancement in quantum yield. Thioflavin-T assays are the most widely applied biophysical technique used to follow amyloid formation. They are simple, easy to apply and have proven to be very informative, but the dye provides an extrinsic probe of amyloid formation and the signal depends on the amount of dye bound and the quantum yield of the bound dye.^{17, 18} Thus, the relationship between the intensity of thioflavin-T fluorescence and the amount of amyloid formed is not always clear and it is formally possible that the assay could give false positives and false negatives.²²

Materials and Methods

Peptide Synthesis and Purification

Peptides were synthesized with a CEM microwave peptide synthesizer on a 0.10 mmol scale utilizing 9-fluorenylmethoxycarbonyl (Fmoc) chemistry. 5-(4’-Fmoc-aminomethyl-3’,5-dimethoxyphenol) valeric acid (PAL-PEG) resin was used to provide an amidated C-terminus. Fmoc-protected pseudoproline (oxazolidine) dipeptide derivatives were utilized as previously described.^{23, 24} Solvents used were ACS-grade. β-branched residues, the first residue attached to the resin, pseudoproline dipeptide derivatives and the residues following the pseudoproline dipeptide derivatives were double-coupled. Peptides were cleaved from the resin via standard trifluoroacetic acid (TFA) methods. The cleaved crude peptides were dissolved into 15% (v/v) acetic acid and lyophilized. The disulfide bond was formed in 100% dimethyl sulfoxide at room temperature. Peptides were purified via reverse-phase high-performance liquid chromatography (RP-HPLC) using a Higgins Analytical Proto 300 C18 preparative column (10 mm × 250 mm). The purity of the peptides was tested using analytical HPLC. The masses of the pure peptides were confirmed with MALDI time-of-flight mass spectrometry. h-IAPP, expected 3903.6, observed 3902.9; p-IAPP, expected 3948.9, observed 3949.4.

Sample Preparation and Fluorescence Assays

Stock solutions were prepared by dissolving peptide into 100% hexafluoroisopropanol (HFIP) at 1.6 mM. Solutions were filtered with 0.45 µM Acrodisc syringe filters and the required amount was lyophilized overnight to remove HFIP. Dry peptide was then dissolved into tris buffer or PBS for the fluorescence assays. The kinetics of amyloid formation were monitored using thioflavin-T binding assays conducted at 25 °C. Fluorescence measurements were performed using a Beckman Coulter DTX 880 plate reader with a multimode detector using an excitation wavelength of 430 nm and an emission wavelength of 485 nm.

Transmission Electron Microscopy (TEM)

TEM images were collected at the Life Science Microscopy Center at the State University of New York at Stony Brook. At the end of each experiment, 15 µL aliquots of the samples used for the kinetic studies were removed, blotted on a carbon-coated 300-mesh copper grid for 1 min and then negatively stained with saturated uranyl acetate for 1 min.

Circular Dichroism

Far UV CD experiments were performed on an Applied Photophysics Chirascan CD spectrophotometer. The sample was incubated on the bench for two weeks and the spectrum was recorded. The spectrum was an average of three repeats recorded over a range of 190 – 260 nm, at 1 nm intervals. A 10 mm quartz cuvette was used and a background spectrum was subtracted from the collected data. Experiments were performed at 25 °C in 20 mM tris buffer at pH 7.4. The sample concentration was 40 µM.

FTIR Spectroscopy

FTIR spectra were recorded on a Bruker Vertex 80 FTIR spectrometer equipped with a globar source, KBr beamsplitter and a liquid nitrogen-cooled mercury cadmium telluride (MCT) detector. Spectra were recorded as the result of 2048 scans at a resolution of 1 cm⁻¹ using a Bruker BioATR II cell. The effective path length of the cell was approximately 6–8 µm. Experiments were performed at 20 °C and a protein concentration of 600 µM.

Molecular Dynamics Simulations and Modeling of Pufferfish and Human Amyloid Fibrils

The two initial fibril structures of human IAPP (two layers of 5 peptides) were taken from two sources: one derived from solid state NMR studies of full length human IAPP and the other from crystal structures of small “steric zipper” peptides derived from the human IAPP sequence. The two initial fibril structures of pufferfish IAPP were derived from the two experimental fibril structures of human IAPP by mutating the corresponding residues (Figure-1A). Each of the four initial fibril structures was immersed in a truncated octahedral box (a = b = c = ~69 Å, α = β = γ = 109.47°) filled with water molecules. The all-atom point-charge force field (AMBER ff03)²⁵ was used to represent the peptides. This force field has been successfully used to model the binding of Aβ(39–42) to Aβ40/ Aβ42 peptides²⁶, the binding among Aβ protofibrils²⁷ and the binding of fluorescent dyes to Aβ protofibrils.²⁸ The water solvent was explicitly represented by the TIP3P model.²⁹ The AMBER 14 simulation suite³⁰ was used in molecular dynamics simulations and data analysis. After an initial energy minimization, a total of 4 simulations (one run for each fibril system) were performed with different initial random velocities. The random velocities of atoms were generated according to the Maxwell-Boltzmann distribution at 300 K. The production run (500 ns) was performed at 300 K and was comprised of 1 ns of molecular dynamics in the NPT ensemble mode (constant pressure and temperature) to equilibrate the solvent and 499 ns dynamics in the NVT ensemble mode (constant volume and temperature). Periodic boundary conditions were imposed on the system. The particle-mesh Ewald method³¹ was used to treat the long-range electrostatic interactions. SHAKE³² was applied to constrain all bonds connecting hydrogen atoms, enabling a 2 fs time step used in the dynamics. To reduce computation time, non-bonded forces were calculated using a two-stage RESPA approach³³ where the short-range forces within a 10 Å radius were updated every step and the long range forces beyond 10 Å are updated every two steps. Langevin dynamics were used to control the temperature (300 K) using a collision frequency of 1 ps⁻¹. The center of mass translation and rotation was removed every 500 steps to remove the “block of ice” problem.^{34, 35} The trajectories were saved at 50 ps intervals for analysis.

Results

The primary sequence of h-IAPP and p-IAPP are compared in Figure-1A, the location of the residues in p-IAPP that differ from those in h-IAPP is also displayed within the context of a structural model of the h-IAPP amyloid fibril (Figure-1B).³⁶ The structured core of the h-IAPP amyloid fibril is thought to be comprised of residues 8 to 37.^36–38 All eleven of the changes in p-IAPP relative to h-IAPP are found within this 30 residue segment. Of the eleven differences in the sequence, seven are found between residues 22 and 29, a segment which has been proposed to be important for modulating amyloidogenicity.^39–42 p-IAPP also contains a His-18 to Arg substitution. This replacement is known to reduce the amyloidogenicity of h-IAPP and modulates IAPP toxicity in cell culture.^{43, 44} The pufferfish polypeptide also includes an Asn-14 to Asp substitution relative to h-IAPP which has been reported in other fish IAPPs, but not in mammals. Other substitutions which have not been reported in mammals include a Leu-27 to Tyr replacement and the replacement of Phe-23 by Ile (Figure-2).

Amyloid Prediction Algorithms Give Conflicting Results for the Relative Amyloidogenicity of Pufferfish IAPP

A range of amyloid prediction programs have been developed and most rely on physicochemical analysis of the properties of the primary sequence, although the ZipperDB algorithm uses a template approach based on steric zippers.^45–50 The different methods lead to different predictions for the relative amyloidogenicity of h-IAPP and p-IAPP (Supporting Table-1). AGGRESCAN and AmylPred predict h-IAPP is more amyloidogenic than p-IAPP, while TANGO and ZipperDB predict that it is less amyloidogenic and Zyggregator and PASTA predict that there is no change. Thus, existing methods to predict amyloidogenicity give ambiguous results when applied to p-IAPP. Amyloid prediction algorithms have also been shown to improperly predict the amyloidogenicity of scrambled peptide sequences derived from the N-terminal segment of the Huntingtin peptide.⁵¹ The ability of any full-length fish IAPP to form amyloid has not yet been experimentally evaluated.

Pufferfish IAPP Forms Amyloid, But Does not Bind to Thioflavin-T in Tris Buffer

We compared the time course of amyloid formation by human and pufferfish IAPP using thioflavin-T fluorescence assays and by transmission electron microscopy (TEM). We conducted initial experiments in 20 mM tris at pH 7.4, chosen because this buffer has been used extensively in studies of amyloid formation by IAPP. The concentration of IAPP used was 16 µM, again chosen because it is typical of values used for biophysical studies with the polypeptide. A 2-fold excess of thioflavin-T was used. Most experiments with h-IAPP avoid a large excess of thioflavin-T to avoid any perturbation of the kinetics of amyloid formation. Thioflavin-T does not impact the kinetics of IAPP aggregation under these conditions.⁵² A sigmoidal thioflavin-T fluorescence time course, consisting of a lag period, a growth phase and final plateau is observed in the presence of h-IAPP which is characteristic of amyloid formation (Figure-3A). No change in thioflavin-T intensity is observed for the sample of p-IAPP, even for a time which is 6-fold longer than the lag time of human IAPP amyloid formation (Figure-3A). The standard interpretation of these results would be that pufferfish IAPP is not amyloidogenic under these conditions. However, time-dependent TEM studies show that this is not the case (Figure-3B). Aliquots were collected at five different time points, including two time points within the lag phase of h-IAPP amyloid formation. Samples were removed at t= 0, 0.25t₅₀, 2t₅₀, 3t₅₀ and 5t₅₀, where t₅₀ refers to the time required for human IAPP to reach half maximum fluorescence intesnsity in a thioflavin-T assay. As expected, no amyloid fibrils are detected in the samples of h-IAPP removed at t = 0, and t = 0.25t₅₀, In contrast, amyloid fibrils are clearly present in the p-IAPP sample at the 0.25t₅₀ time point (Figure-3B). Mats of amyloid fibrils are observed for both h-IAPP and p-IAPP in samples removed at 24 hours and at subsequent time points even though no change in thioflavin-T fluorescence intensity is detected for the p-IAPP sample. These experiments clearly indicate that thioflavin-T based assays give a misleading view of p-IAPP amyloid formation. The thioflavin-T assays displayed in Figure-3A involved monitoring the fluorescence intensity at 485 nm. We confirmed that the lack of a change in the signal of the p-IAPP sample is not a consequence of the emission wavelength chosen by collecting complete fluorescence emission spectra of thioflavin-T in the absence of peptide and in the presence of either human or pufferfish IAPP amyloid fibrils. The spectra of thioflavin-T alone or in the presence of p-IAPP amyloid are essentially identical with a weak emission maximum near 520 nm (Figure-4). In contrast, the emission maximum is shifted to near 480 nm in the presence of h-IAPP amyloid fibrils and there is a significant enhancement in quantum yield. The lack of thioflavin-T signal could be due to changes in the surface structure of the amyloid fibrils that lead to weaker dye binding, or be caused by changes in the quantum yield of the bound dye. A lower yield of fibrils will also lead to a weaker thioflavin-T signal. Another potential factor that can lead to lower thioflavin-T is changes in the association of fibrils. If the fibrils tightly associate, the dye binding surfacing could potentially be occluded, leading to fewer thioflavin-T binding sites.

Figure-3.

Analysis of amyloid formation by pufferfish IAPP (blue) and human IAPP (black) in 20 mM tris-HCl at pH 7.4. (A) Fluorescence monitored thioflavin-T assays of amyloid formation. No change in thioflavin-T fluorescence is detected for the pufferfish peptide over the entire time course of the experiment. Arrows indicate times, at which aliquots where collected for TEM analysis. (B) TEM images of samples of human IAPP (top) and pufferfish IAPP (bottom) collected at the different time points. Experiments were conducted with 16 µM IAPP, 32 µM thioflavin-T at 25 °C in 20 mM tris-HCl at pH 7.4. Scale bars in TEM images represent 100 nm.

Figure-4.

Thioflavin-T emission spectra of samples of 32 µM thioflavin-T, with human IAPP (black), pufferfish IAPP amyloid fibrils (blue), and thioflavin-T alone (grey). IAPP when present is at 16 µM in monomer concentration. Spectra were recorded at the end of the kinetic experiments depicted in Figure 2. Samples contained 20 mM tris-HCl at pH 7.4 and measurements were made at 25 °C. Excitation at 450 nm.

The TEM images of h-IAPP and p-IAPP amyloid fibrils are very similar (Figure-3B), but it is difficult to detect molecular level changes in structure by TEM, given the resolution of the method and the negative stains employed. Thus, we turned to circular dichroism (CD) to probe the secondary structure of the p-IAPP fibrils. A sample of p-IAPP was prepared and incubated until amyloid fibrils were formed, as judged by TEM, and the CD spectrum recorded (Figure-5A). The CD spectrum of the pufferfish peptide exhibits a broad minimum near 218 nm, consistent with β-sheet structure, and the shape of the spectrum is very similar to that reported for the h-IAPP amyloid fibrils.

Figure-5.

Pufferfish IAPP forms fibrils with significant β-sheet secondary structure. (A) CD spectrum of a sample of the fibril material formed by a 40 µM sample of pufferfish IAPP. The spectrum is the smoothed average of three scans. (B) FTIR spectra of p-IAPP (blue) and h-IAPP (black). (C) TEM image of the pufferfish IAPP sample collected at time of CD measurement. Samples were incubated at 25 °C. Scale bars in TEM images represent 100 nm.

We also recorded the Fourier-transform infrared (FTIR) spectra of h-IAPP and p-IAPP (Figure-5B). The amide-I mode is sensitive to secondary structure, exhibiting characteristic shifts to lower energy, lower wavenumber, observed in β-sheets. There is considerable variability in the FTIR spectra of amyloid, but most amyloids exhibit a broad peak near the region of 1615 to 1630 cm⁻¹ while random coil features are found at higher wavenumbers.⁵³ Monomeric h-IAPP has been reported to give an FTIR spectrum with a broad peak centered near 1645 to 1650 cm⁻¹. ⁵⁴ Our experimental spectrum of h-IAPP exhibits a peak at 1627 cm⁻¹ while p-IAPP displays a peak at 1635 cm⁻¹. In both cases, the peak is shifted away from the random coil position. The relatively modest difference in band position for the two polypeptides suggests that there could be differences in the structure of the aggregates. MD simulations, described in a subsequent section, provide a possible explanation for the difference.

The Relative Rates of Amyloid Formation by Human and Pufferfish IAPP are Dependent on the Choice of Buffer

We next examined the behavior of h-IAPP and p-IAPP in a more physiological buffer consisting of 20 mM sodium phosphate, 140 mM KCl at pH 7.4 (PBS). A typical thioflavin-T curve is observed for h-IAPP in PBS with a shorter lag phase than observed in tris (Figure-6). The shorter lag phase is consistent with the known effects of salts on IAPP amyloid formation.⁵⁵ A weak thioflavin-T signal is observed from the p-IAPP sample under these conditions, but the final thioflavin-T intensity is 15 to 20-fold weaker than observed for the h-IAPP sample. Somewhat surprisingly, the time required to reach the plateau of the thioflavin-T response is longer for the p-IAPP sample in PBS relative to the time required for h-IAPP to form amyloid even though p-IAPP forms amyloid more rapidly in tris buffer. A TEM time course confirms this observation. Amyloid fibrils are detected from the sample of h-IAPP before they are observed in the p-IAPP sample. Thus, the relative rate of amyloid formation by p-IAPP and h-IAPP depends upon the choice of buffer, an observation which indicates discussions of relative amyloidogenicity should be treated with caution.

Figure-6.

Analysis of the ability of pufferfish IAPP to form amyloid in phosphate buffered saline solution. (A) Fluorescence monitored thioflavin-T assays of human IAPP (black) and pufferfish IAPP (blue) amyloid formation in 20 mM sodium phosphate, 140 mM potassium chloride at pH 7.4. Arrows indicate times at which aliquots where collected for TEM analysis. Aliquots were collected at t = 0 (black), 3 hrs (red), 7 hrs (green), and 24 hrs (blue) (B) TEM images of samples human IAPP (top) and pufferfish IAPP (bottom) collected at the different time points. Experiments were conducted with 16 µM IAPP, 32 µM thioflavin-T at 25 °C in 20 mM sodium phosphate, 140 mM potassium chloride at pH 7.4. Scale bars in TEM images represent 100 nm.

Pufferfish IAPP is Compatible with Existing Models of Human IAPP Amyloid Fibrils

We next performed molecular dynamics simulations to test if the pufferfish sequence is compatible with existing models of the h-IAPP amyloid fibrils. There are two high resolution models of the h-IAPP fibrils: one derived from solid state NMR studies of full length h-IAPP and the other from crystal structures of small “steric zipper” peptides derived from the h-IAPP sequence.^{36, 37} While the two models differ in their details, they have many common features and their similarities outweigh their differences. In both structures, the fibril core is made up of two C2 symmetric stacks of monomers (Figure 1B). Each monomer adopts a U-shaped structure which includes two β-strands connected by a less structured linker. The N-terminal strand is on the exterior of the fibril and the C-terminal strand forms the interface between two peptide molecules in one layer. We threaded the pufferfish sequence onto each model structure, here using two stacks of five monomers as the model, and conducted all-atom molecular dynamics simulations of both with explicit solvent model, as detailed in the Methods section. We also ran MD simulations of each model using the human peptide as a control. The 500 ns simulations indicate that the p-IAPP sequence is compatible with the proposed fibril models for h-IAPP, retaining the overall strand-loop-strand fold. The Cα-RMSD at the end of the simulations are shown in Supporting Figure-S1 and the last snapshots of the four simulations are shown in Supporting Figure-S2. The NMR-derived fibril model fits the p-IAPP sequence particularly well, with less than a 1 Å deviation in Cα-RMSD between the h-IAPP and the p-IAPP forms, suggesting that the backbone fold is very similar between human and pufferfish. However, as can be seen from the two representative structures shown in Figure-7, while the overall U-shape fold (β-strand-loop-β-strand) is maintained in p-IAPP, the region adjacent to the loop has more coil and less β-sheet structure than in the case of h-IAPP. This is an interesting observation in light of the FTIR results. The loop which connects the two β-strands also appears to be more flexible in p-IAPP. A notable feature is that the sidechains on the surface of the pufferfish fibrils are more disordered relative to the models of h-IAPP. Our early computational studies of the binding of thioflavin-T to a number of amyloid fibril models showed that the primary binding sites for thioflavin-T were regular grooves formed by ordered side-chains on the surface of amyloid fibrils, with minor sites located the ends of the fibril.^19–21 The lack of ordered grooves on the surface of the p-IAPP fibrils may explain the weak thioflavin-T fluorescence signal, despite clear evidence of fibril formation from TEM.

Figure-7.

Last snapshots of two representative simulations. (A) Human IAPP fibril model from the solid state NMR study. (B) Pufferfish IAPP fibril model is derived from the NMR model by threading. 3₁₀, α-helical, β-strand, turn and coiled conformations are colored in blue, yellow, cyan and white respectively. Sidechains are color-coded as; blue, positively charged; red, negatively charged; black, hydrophobic; and green, hydrophilic.

Thioflavin-T Assays Do Not Accurately Report on Pufferfish IAPP Amyloid Formation Even If the Dye is in Excess

We also examined the effect of increasing the concentration of thioflavin-T to a 20-fold excess relative to IAPP in monomer units. Normally, these concentrations are avoided for fear that high dye concentrations might perturb the kinetics of amyloid formation. The dye has a small impact on the kinetics of h-IAPP amyloid formation, leading to a slightly more rapid time course (Figure-8). A weak thioflavin-T signal is detected when 20-fold excess of dye is added to pufferfish IAPP in tris buffer, but the signal is still weaker than observed for the sample of h-IAPP with a 10-fold lower concentration of dye. The small thioflavin-T signal for the pufferfish IAPP sample exhibits a time course with a longer apparent lag phase (Figure-8A) than h-IAPP. The normal interpretation would be that pufferfish IAPP forms amyloid more slowly than h-IAPP, but generates less amyloid fibers than h-IAPP. However, TEM reveals that this is not the case. Images recorded of samples removed during the lag phase of h-IAPP amyloid formation revealed no detectable fibrils in the h-IAPP samples, but mats of amyloid fibrils were evident in the pufferfish IAPP samples (Figure-8B). This clearly indicates that pufferfish IAPP forms amyloid more rapidly than does h-IAPP under these conditions and provides further evidence that thioflavin-T assays can give misleading results.

Figure-8.

Analysis of amyloid formation by pufferfish IAPP and human IAPP in the presence of excess thioflavin-T in 20 mM tris-HCl at pH 7.4. (A) Fluorescence monitored thioflavin-T assays of human IAPP with 2-fold and 20-fold excess of thioflavin-T (black and orange respectively) and of pufferfish IAPP with 2-fold and 20-fold excess of the dye (blue and purple respectively). Arrows indicate times at which aliquots were collected for TEM analysis. Aliquots were collected at t = 0 (black), 0.25t₅₀ (red), 2t₅₀ (green), 3t₅₀ (blue), and 5t₅₀ (yellow), where t₅₀ refers to the time required for human IAPP to reach half maximum fluorescence intensity in a thioflavin-T assay. (B) TEM images of samples of human IAPP (top) and pufferfish IAPP (bottom) in the presence of 20-fold excess of thioflavin-T collected at different time points. Experiments were conducted with 16 µM IAPP, 32 µM or 320 µM thioflavin-T at 25 °C in 20 mM tris-HCl at pH 7.4. Scale bars in TEM images represent 100 nm.

Discussion

The data presented here demonstrates that p-IAPP is capable of forming amyloid, but shows that the relative rate of amyloid formation as compared to human IAPP depends upon the choice of buffer. This observation has clear implications for studies which attempt to define and compare the relative amyloidogenicity of different polypeptides. In the present case, the differential effects may reflect the different charge distributions of the two polypeptides. Analysis of the effects of anions on h-IAPP amyloid formation have shown that their effects follow the ion selectivity series at low to moderate concentrations, arguing for a contribution from ion binding.⁵⁵ p-IAPP has a lower net charge than h-IAPP due to the Asn to Asp substitution and this may reduce the affinity for phosphate. In this scenario, phosphate accelerates h-IAPP amyloid formation to a greater extent than it does amyloid formation by p-IAPP and this leads to a reversal in the relative rates of amyloid formation. Irrespective of the mechanistic details, the results indicate that solution conditions are important in evaluating relative amyloidogenicity. Our observation that different amyloid prediction programs give differing results for p-IAPP and h-IAPP, together with the context dependent effects described above, and earlier studies on the reliability of amyloidogenicity algorithms indicates that caution should be applied when interpreting the results of amyloidogenicity algorithms.⁵¹

We believe that the observation of weak or non-existent thioflavin-T binding by p-IAPP is also important. Thioflavin-T is the most widely used probe of amyloid formation and the present study provides a clear cautionary example of the assay yielding a false negative. Thioflavin-T is believed to bind to surface grooves found on amyloid fibrils that are a consequence of their cross-β structure. The large number of substitutions in p-IAPP, 11 out of the 30 residues which are believed to form the core of the fibril, could change the structure and physicochemical nature of the surface grooves. Our all atom MD simulations show that the pufferfish sequence is compatible with the solid state NMR model of the h-IAPP amyloid fibrils as well as with the model derived from crystal structures of small peptide fragments of h-IAPP. However, two significant differences are observed. First, the part of the region located adjacent to the loop in the U-shape motif adopts coil-like conformations rather than the extended β-sheet conformations seen in h-IAPP. Importantly, the sidechains of the pufferfish IAPP model are more disordered than in the case of the human fibrils and, as a result, do not form the well-ordered surface grooves on the face of the fibril that serve as binding sites for thioflavin-T. The differences in the FTIR spectra of human and pufferfish IAPP are consistent with some structural changes. These differences may contribute to the reduced binding of thioflavin-T to pufferfish IAPP. A reduced yield of amyloid fibrils by the pufferfish peptide could potentially contribute to the lower thioflavin-T signal. In this scenario, p-IAPP forms fewer fibrils and there would be a significant fraction of the polypeptide that remain unaggregated. This seems unlikely, especially given that the CD spectrum and FTIR spectrum indicate that the conformational ensemble is dominated by β-structure. Gravimetric data could formally rule out this possibility but is not available. Again, the mechanistic details of the reduced thioflavin-T signal are not completely clear, but the key result is that thioflavin-T assays can give rise to false negatives and the final thioflavin-T intensity should not be interpreted as a quantitative measure of the amount of amyloid formed when comparing different peptides. Work on Aβ polymorphs and poly-Q peptides as well as a comparison of Aβ(1–40) and Aβ(1–42) also highlights the dangers of quantitative interpretation of thioflavin-T intensities.^56–58

Finally, we note that teleostean fish have been proposed to be a source of xenobiotic islet transplants. Encouraging success has been obtained using porcine islets for xenobiotic islet transplants, but porcine IAPP is much less amyloidogenic than h-IAPP under all conditions tested and its reduced amyloidogencity correlates with enhanced islet survival.⁹ The comparable amyloidogenicity of pufferfish and human IAPP indicates that caution should be used if fish are employed for transplant studies.^{12, 14, 15} However, it is also worth noting that the partial sequence of Atlantic wolffish (Anarhichas lupus) includes additional proline substitutions as well as additional charge substitutions which likely render the sequence less amyloidogenic and it may be a more attractive candidate for xenobiotic studies.¹² Of course, other factors also need to be addressed when considering candidates for xenobiotic studies. These include the activity relative to human IAPP and immunogenicity. Nonetheless, it will be interesting to examine the behavior of the full-length wolfish sequence once it becomes available (Figure-2).

Abbreviations

CD

circular dichroism

Cα-RMSD

root mean square deviations of the Cα atomic coordinates

Fmoc

9-fluorenylmethoxycarbonyl

FTIR

Fourier-transform infrared spectroscopy

HFIP

hexafluoroisopropanol

h-IAPP

human IAPP

IAPP

islet amyloid polypeptide or amylin

MALDI

matrix assisted laser desorption/ionization

MD

molecular dynamics

PAL-PEG

5-(4’-Fmoc-aminomethyl-3’,5-dimethoxyphenol) valeric acid

PBS

sodium phosphate saline buffer

p-IAPP

pufferfish IAPP

t₅₀

time required for a human IAPP to reach half maximum fluorescence

TEM

transmission electron microscopy

TFA

trifluoroacetic acid;