Introduction

Increasing evidence correlates the process of amyloid formation by human islet amyloid polypeptide (hIAPP) with pathogenic β-cell apoptosis in diabetes mellitus type II (T2DM) [1–8]. hIAPP is a 37-residue peptide co-secreted with insulin from pancreatic β-cells, which normally plays an adaptive role in glucose metabolism. In T2DM this peptide is overexpressed and misfolds from its natively disordered state [8–11], forming pervasive amyloid plaques in the pancreas. Amyloid formation by hIAPP also contributes to the failure of islet transplants [12, 13]. As with other amyloid diseases, hIAPP plaques are composed of macroscopic, structured fibrils containing stacked parallel protein β-sheets [1, 14]. Recent findings suggest that it is not the plaques but early soluble oligomers that are the most pathogenic agent [2, 3, 15, 16]. It is therefore critical to understand early oligomer assembly and structure, a process hampered by the transient nature of the soluble oligomers. Ion mobility spectrometry-mass spectrometry (IMS-MS) has proven to be a powerful tool for investigating the distribution of these low-order amyloid oligomers [17–21]. Previous IMS-MS studies have revealed key differences between the oligomers formed by hIAPP and the oligomers formed by nontoxic rat IAPP [22, 23]. IMS-MS has also been used to study the inhibition of hIAPP amyloid formation in vitro by insulin and small molecules [20, 21, 24].

The aggregation dynamics of hIAPP peptide segments have been studied for their potential as aggregation nucleation sites while modified fragments serve as promising inhibitors of hIAPP aggregation [25–30]. Much attention has focused on hIAPP20-29 (consisting of the 20–29 residue region, SNNFGAILSS) and hIAPP22-29 as early work linked the primary sequence in this region to hIAPP amyloidogenicity. Subsequent work showed the situation is more complex but the 22-29 region is still considered to be a key contributor to hIAPP aggregation [31–33]. Several other IAPP segments have been found capable of self-association, including the N-terminal region [27, 34, 35]. This region, which contains an intramolecular disulfide bond between cysteine residues 2 and 7, is thought to be important to hIAPP’s in vivo activity [36–38]. Additionally, while hIAPP has been found to endogenously decrease insulin secretion from β-cells, the introduction of hIAPP8-37 had an antagonistic effect on hIAPP-mediated insulin inhibition [39]. There are conflicting reports on the role of the N-terminus in hIAPP aggregation; early studies showed that amyloid formation by an 8-37 fragment of hIAPP occurs more rapidly in buffer but more slowly in the presence of 1.5% hexafluoroisopropanol (HFIP) [40, 41]. In either case, the effect was modest but scanning transmission electron microscopy (STEM) appears to indicate that truncation changes the morphology of the fibrils [40]. A mutant form of hIAPP, with alanine substitutions at both cysteine residues, positions 2 and 7, showed a modest increase in the nucleation lag-phase compared with wildtype hIAPP [42]. A study by Cope et al. has shown that the hIAPP1-8 peptide fragment is capable of non-amyloid aggregation into large macroscopic structures and that cleavage of the disulfide bond apparently inhibits this aggregation process [43]. Given both the physiological and biophysical significance of the disulfide bond in hIAPP, understanding the role of the N-terminal region is important to the overall picture of IAPP aggregation and pathology.

In this paper, we investigate the role of the N-terminus fragment in aggregation using a combination of ion-mobility mass spectrometry (IMS-MS), atomic force microscopy (AFM), and replica-exchange molecular dynamics (REMD) (Figure 1). IMS-MS allows for the separation of conformational isomers based on their cross section and, when coupled with a soft ionization source, clear characterization of isomeric and oligomeric species is possible [44, 45]. While the combination of ion mobility and molecular simulations allows for the determination of early small oligomer structures, AFM adds the ability to observe bulk characteristics of the peptide in solution as well as larger oligomeric aggregates, which correlate with longer aggregation times [46–49]. By combining these methods, significant insight into the structural details of peptide oligomerization in solution can be obtained, particularly over a broad range of oligomer sizes.

Figure 1
figure 1

Sequences of the peptides studied in this work; the oxidized WT 1-8 residue fragment (WT1-8 ox) is indicated in blue, reduced 1-8 residue fragment (WT1-8 red) is indicated in green, and the cysteine to serine double mutant fragment (C2S/C7S1-8) is indicated in orange. The presence of a disulfide bond is indicated by a bracket. All of the peptides contain an amidated C-terminus. The sequence of full length hIAPP is shown in black for comparison

Full size image

Materials and Methods

Peptide Synthesis and Preparation

All peptides were synthesized using Fmoc-based microwave-assisted methods and purified by reverse-phase HPLC as previously described [50, 51]. All peptides contain an amidated C-terminus. Samples were dissolved in 100% hexafluoroisopropanol (HFIP) (Sigma-Aldrich, St. Louis, MO, USA) to a final stock concentration of 1 mM. Aliquots of the stock were lyophilized and redissolved in 1:1 methanol:water for final peptide concentrations of 50–100 μM. In order to reduce the disulfide bond between cysteine residues 2 and 7, tris(2-carboxyethyl)phosphine (TCEP) (Alfa Aesar, Ward Hill, MA, USA) was added at a 10:1 excess to peptide solutions and briefly incubated before measurements were taken. Reduction of the disulfide bond was verified via quadrupole time-of-flight mass spectrometry (Supplementary Figure S4). All solvents had their pH adjusted to physiological pH (7.4) using concentrated ammonium hydroxide and acetic acid unless otherwise noted.

Ion Mobility Experiments

An in-depth description of the IMS-MS instrumentation has been given previously [52]. In summary, samples were loaded into gold-coated borosilicate glass capillaries that had been pulled to a fine point using an in-house capillary puller (Sutter Instrument Co., Novato, CA, USA). Using a nanoelectrospray ionization source (nano-ESI), ions are pulled from the capillary under the influence of a voltage differential and channeled through a small (~70 μm) orifice into an ion funnel. The funnel then collects and focuses the ions; these can either be continuously funneled (for mass spectra) or pulsed (for ion mobility) into a 4.5-cm drift cell filled with ~3.5 Torr of helium gas. Under the influence of a weak electric field, ions are drifted through the cell, mass analyzed by a quadrupole, and detected by an electron multiplier detector. Total time for analysis is between 400 and 2000 μs.

Ion drift is affected by two factors; a uniform electric field, E, applied across the cell as well as collisions with the cell buffer gas. These two opposing forces result in a constant drift velocity proportional to the electric field with an ion mobility proportionality constant K;

$$ {v}_d=K\cdot E $$
(1)

The ion mobility is dependent upon the buffer gas temperature T and pressure P; ion mobility is typically expressed without this dependence as a reduced mobility, K 0 . The electric field is generated by a fixed voltage V across a drift cell of length L;

$$ {K}_0=K\frac{P}{760\;\mathrm{Torr}}\frac{273K}{T}\kern1em E=\frac{V}{L} $$
(2)

The reduced mobility can therefore be determined by plotting drift time as a function of pressure to voltage (P/V) ratios;

$$ {v}_d=\frac{L}{t_d}=K\cdot E\kern1em {t}_d=\frac{L}{K\cdot E}=\frac{L^2\left(273\;\mathrm{K}\right)}{K_0\left(760\;\mathrm{Torr}\right)T}\cdot \frac{P}{V} $$
(3)

Ion mobility is then directly related to the ion’s collision cross section via the collision theory of gases [53];

$$ {K}_0=\left(\frac{3e}{16{N}_0}\right){\left(\frac{2\pi }{\mu {k}_BT}\right)}^{\frac{1}{2}}\frac{1}{\sigma } $$
(4)

where e is the ionic charge, N 0 is the buffer gas number density, μ is the reduced mass of the ion and buffer gas, and σ is the collision cross section.

Molecular Dynamics

Explicit solvent temperature-REMD (T-REMD) simulations were performed for the wildtype 1-8 residue fragment (WT1-8 ox, containing a disulfide bond between the second and seventh cysteine residues), the reduced 1-8 residue fragment (WT1-8 red), and the cysteine → serine C2S/C7S mutant fragment (C2S/C7S1-8) using the GROMACS 4.5.3 package and the all-atom Optimized Potentials for Liquid Simulations (OPLS-AA) force field in TIP3P water [54–56]. The peptide monomers have charged termini throughout the simulation, assuming neutral pH. The initial structure was minimized using the steepest decent method and solvated in a TIP3P water box. The size of the solvated cubic system was 4.50 nm for peptide fragment dimer and tetramer simulations. In each system, negative charge chlorine ions were added to neutralize the system. Solvent and volume equilibration simulations in NPT ensemble (T = 300 K and P = 1 bar) were performed to optimize the box size, followed by 6-ns NVT equilibration at 300 K. The LINCS algorithm was employed to constrain bonds between heavy atoms and hydrogen [57], and the SETTLE algorithm was used for water molecules [58]. These constraints allowed an integration time step of 2.0 fs. The electrostatic and dispersion forces were computed with a real space cutoff of 1.2 nm, and particle mesh Ewald method was used to treat long-range electrostatics [59]. The temperature was maintained by the Nose-Hoover thermostat [60]. The temperature and pressure coupling constants were 0.1 and 1.0 ps, respectively. The equations of motion were integrated according to the leap-frog algorithm.

The initial guess for temperatures in T-REMD simulations (24 replicas for dimer and 32 replicas for tetramers) was taken from Patriksson and Spoel’s temperature predictor (http://folding.bmc.uu.se/remd/index.php) and then adjusted to obtain the exchange rate of approximately 20%–25% [61]. Temperature lists can be found in the supporting information.

Each replica was equilibrated at the desired temperature for 12 ns before the production run for T-REMD began. Exchanges between replicas were attempted every 2 ps. The production run was 300-ns long per replica for dimers and 400-ns per replica for tetramers. Only the last 150-ns data of the trajectory at 299 K were subjected to analysis (Supplementary Figure S5). Clustering was performed using the Daura algorithm [62]. In order to obtain theoretical cross sections, 50 structures per cluster were subjected to a short minimization in the gas phase to mimic the desolvation process after the electrospray ionization process. After that, the theoretical cross sections were computed using the projected superposition approximation (PSA) method available at (http://luschka.bic.ucsb.edu:8080/WebPSA/index.jsp) and the average values were reported [63, 64].

Atomic Force Microscopy

AFM experiments were carried out in tapping mode in air using an MFP-3D Atomic Force Microscope (Asylum Research, Goleta, CA, USA). A silicon probe was used with a cantilever spring constant of 7 N/m and a resonant frequency of 155 kHz (MikroMasch USA, Lady’s Island, SC, USA). Protein incubation conditions matched those used in the IMS-MS experiments. Oxidized and mutant peptide fragment samples were incubated in 1:1 methanol:water; chemically reduced peptide samples were incubated with TCEP reducing agent at a 10:1 ratio of reducing agent to peptide. Five μL aliquots were removed at defined times over the course of incubation and deposited on freshly cleaved V1-grade mica (TedPella, Redding, CA, USA). The samples were then dried in a vacuum desiccator. Height analysis of globular aggregates was performed using a mask at 1 nm to the image and adjusting to include all aggregates in the image. A particle analysis routine within the MFP-3D software was then applied.

Results and Discussion

Early Oligomer Formation Probed by IMS-MS Reveals a Distinct Pathway for the Aggregation of WT1-8 ox Compared to Those of WT1-8 red and C2S/C7S1-8

Positive-ion mass-spectra showed similar predominant features for all three peptide fragments with peaks corresponding to a range of oligomers up to heptamer (Figure 2b). WT1-8 red has an additional peak at 1070 m/z not found in the other samples; this is likely a complex between a singly charged peptide and an adduct of the reducing agent, tris(2-carboxyethyl)phosphine (TCEP) (810 amu + 250 amu). All three samples show extensive oligomerization at both vesicular (pH = 5.5) and physiological pH (pH = 7.4). Arrival time distributions (ATDs) of all major features were collected and representative ATDs are shown in Figure 3. Several aspects of the ATDs are of note. In the z/n = +1/1 ATD more doubly charged dimers are seen for the reduced WT1-8 red versus the oxidized WT1-8 ox or mutant C2S/C7S1-8 (Figure 3a). Secondly, more higher-order oligomers are seen in the reduced and mutant z/n = +1/2 ATDs than in WT1-8 ox (Figure 3b). This suggests that the presence of the disulfide bond inhibits oligomer formation. For all fragments the +2/2 oligomers show larger cross sections than the corresponding singly charged dimers. This is likely the result of charge repulsion between lysines at position 1, indicating that these amino acids are in close proximity within the dimer. Comparing relative collision cross sections, the higher-order oligomers of the WT1-8 red and C2S/C7S1-8 peptides are more compact than those of the WT1-8 ox peptide (Table 1), whereas the lower-order WT1-8 ox oligomers are smaller than those of the reduced or mutant peptides. The REMD simulations of the dimers confirm this trend, with WT1-8 red and C2S/C7S1-8 favoring extended conformations with extensive intermolecular bonding, whereas WT1-8 ox takes on compact conformations with low intermolecular bonding (Figure 4). The REMD results also show WT1-8 ox has fewer hydrogen bonds between monomers than WT1-8 red and C2S/C7S1-8 in the tetramers (Figure 5e, f). Taken together, the IMS-MS and REMD results suggest WT1-8 ox aggregates via a different pathway than the peptides without a disulfide bond.

Figure 2
figure 2

Mass spectra of IAPP peptide fragments at 50 μM and 7.4 pH. WT1-8 ox is indicated in blue, WT1-8 red is indicated in green, and C2S/C7S1-8 is indicated in orange. Spectra are annotated by z/n where z is the charge and n the oligomer number. WT1-8 red shows an additional peak not present in the other spectra corresponding to an adduction of the peptide and the TCEP reducing agent

Full size image
Figure 3
figure 3

Representative ATD of the human IAPP fragment peptides at 50 μM and 7.4 pH; WT1-8 ox is indicated in blue, WT1-8 red is indicated in green, and C2S/C7S1-8 is indicated in orange. All peaks are annotated by z/n, where z is the charge and n the oligomer number. The cross sections for all peaks are also noted. The injection energy is 40 V. Each peak is fit using the procedure described in the supporting information

Full size image
Table 1 Collision Cross Section in Å2 of IAPP Fragments in the z/n = +1/1 and +1/2 ATDs*
Full size table
Figure 4
figure 4

Conformational spaces sampled by the peptide dimers and distributions of the number of hydrogen bonds between peptide monomers for WT1-8 ox (a), (d), WT1-8 red (b), (e), and C2S/C7S1-8 (c), (f). A sample structure containing the mean number of hydrogen bonds is shown by each distribution. The potential mean force (PMF) is generated from the distributions of the end-to-end distances (Ree, nm) of the dimer; Ree = (Ree1 2 + Ree2 2)1/2, and the distributions of the radius of gyration (Rg, nm). The disulfide bonds keep the WT1-8 ox dimers in a narrow conformational space, whereas both C2S/C7S1-8 and WT1-8 red dimers adopt a wider range of structures including significant populations of extended structures (Ree ≥ 2 nm). The extended structures are stabilized through inter-chain hydrogen bonding, with an average of five hydrogen bonds per structure

Full size image
Figure 5
figure 5

Conformational spaces sampled by the peptide tetramers and distributions of the number of hydrogen bonds between peptide monomers for WT1-8 ox (a), (d), WT1-8 red (b), (e), and C2S/C7S1-8 (c), (f). A sample structure containing the mean number of hydrogen bonds is shown by each distribution. The potential mean force (PMF) is generated from the distributions of the end-to-end distances (Ree, nm) of the tetramer; Ree = (Ree1 2 + Ree2 2 + Ree3 2 + Ree4 2)1/2, and the distributions of the radius of gyration (Rg, nm). The WT1-8 ox tetramers sample only compact structures with small amounts of hydrogen bond formation, similar to its dimers. The C2S/C7S1-8 prefers to adopt elongated structure (the majority have Ree > 3 nm), however, its conformational space is disordered with multiple distinct local minima. The conformation space of WT1-8 red is less disordered than that of the mutant with significant hydrogen bond formation

Full size image

AFM Reveals WT1-8 red and C2S/C7S1-8 Show Opposite Morphological Transformations

AFM results allow the extrapolation of aggregation trends and morphologies for larger oligomers in solution (Figure 6). WT1-8 ox exhibited relatively slow, gradual fibril growth over the course of the 1-hour incubation period consistent with oligomer trends observed by IMS-MS (Figure 6a). After 5 min of incubation, few fibrils were seen on the mica substrate with the dominant structure being small, unstructured globular aggregates with an average height of 1.00 ± (0.25) nm. Fibril-like structures that are present are approximately 0.5 nm in height. After 30 min, there were indications of growth via a seeding core mechanism; fibrils with heights of 0.5 nm were observed growing from central cores of ~1 nm in height. This growth motif has been previously observed for the full-length hIAPP peptide [65, 66]. From this time onward, growth of several fibrils from a central seed becomes a pervasive growth motif. After 60 min, a large degree of lateral growth has occurred and the fibril diameters are much larger, ranging from one to several tens of nanometers. The three representative images shown in Figure 6a reveal that fibril nucleating cores progressively increase in size with time. These data, along with the apparent bundling of fibrils when incubation time is extended to 24 h (Supplementary Figure S2), are consistent with observations of WT1-8 ox made by Cope et al. but the fibrils we observe show more curvature [43]. Given the relatively low amount of intermolecular hydrogen bonding predicted by the REMD simulations for smaller oligomers, the observed fibril bundling could be a result of lateral stabilization that allows for the increase in linear growth and the observed curvature.

Figure 6
figure 6

Time-lapse incubation of WT1-8 ox (a), WT1-8 red (b), and C2S/C7S1-8 (c) fragments at 50 μM and 7.4 pH with incubation times of 5, 30, and 60 min. WT 1-8 hIAPP shows gradual fibril growth over the course of incubation. WT1-8 red shows growth from initially short, closely associated structures into more traditional fibril-like aggregates. C2S/C7S1-8 shows rapid ordering into fibril-like aggregates after only 5 min followed by self-aggregation of these fibrils and further conversion into needle-like structures

Full size image

As seen in Figure 6b, c, both WT1-8 red and C2S/C7S1-8 rapidly form highly ordered aggregate structures (Figure 6, 5 min), consistent with both the higher order oligomerization seen in the IMS-MS experiments and the extensive intermolecular hydrogen bonding predicted by REMD. C2S/C7S1-8 fibrils of sizes ranging in width from 1 nm at the smallest to 10 nm at the largest were observed after just 5 min of incubation (Figure 6c). At longer incubation times, these fibrils appear to self-associate resulting in an interwoven morphology (Figure 6c, 30 min) that then rearranges into short, needle-like aggregates (Figure 6c, 60 min). WT1-8 red also formed ordered aggregates rapidly; however, its time evolution differs from C2S/C7S1-8. Short, closely associated fibril structures are favored at short incubation times and more extended fibrils favored at longer incubation times (Figure 6b). After 1 h, both WT1-8 red and C2S/C7S1-8 exhibited feature lengths significantly shorter than WT1-8 ox. This could explain the apparent lack of aggregation observed in prior studies for WT1-8 red, given that structures of this size would not be visible under a standard light microscope [43]. Again it is intriguing that the peptides for which REMD predicted higher intermolecular bonding appear to follow a discrete aggregation pathway disparate from WT1-8 ox, a result consistent with IMS-MS data.

Conclusions

The presence of the disulfide bond in the N-terminal fragment does not completely inhibit self-aggregation of the fragment, but reduction of the disulfide bond or its removal by mutation increases the rate of aggregation. Overall, IMS-MS, REMD simulations, and AFM indicate WT1-8 ox is capable of forming stable aggregates along an entirely different pathway than both WT1-8 red and C2S/C7S1-8. WT1-8 red and C2S/C7S1-8 form extended oligomer conformations with high intermolecular hydrogen bonding (Figures 4 and 5). AFM images recorded at short incubation times corroborate these conclusions, with WT1-8 red and C2S/C7S1-8 showing extensive aggregation and rearrangement. Long-term macroscopic fibril formation varies drastically between the peptides, WT1-8 ox prefers extended fibril bundling seeded by globular aggregates; WT1-8 red forms short fibrils of varying widths, and C2S/C7S1-8 favors densely packed needle-like structures. These observations, coupled with the REMD results, suggest that the cleavage of the disulfide bond has an agonistic effect toward oligomerization of the full hIAPP peptide by increasing the probability of intermolecular hydrogen bonding and β-sheet formation. Perhaps the most important conclusion we make is that the 2-7 disulfide bond at the N-terminus limits the aggregation of the peptide relative to both the reduced and mutated peptides. This occurs primarily due to a significant reduction in intermolecular hydrogen bonds relative to the noncyclic peptides and suggests the disulfide bond may actually protect full-length hIAPP from amyloid formation. This central conclusion is at odds with the conclusion of Cope et al. [43], who concluded that “loop–loop” interaction might enhance hIAPP oligomerization and could be a target for therapeutic intervention.