Keywords

1 Introduction

Islet amyloid polypeptide (IAPP) is a peptide hormone secreted by the endocrine β-cells of the pancreas together with insulin [1]. It has 37 amino acids with a disulfide bond between residue 2 and 7 in the N-terminus. In solution, IAPP is characterized as a natively disordered protein [2–6]. In patients with type 2 diabetes, IAPP changes its conformation to form amyloid fibers [7]. The process of IAPP amyloid formation and the prefibrillar oligomers are supposed to be one of the pathogenic agencies causing pancreatic β-cell dysfunction [8–11]. Thus the structural characterization of IAPP in the form of monomer or small oligomer states would be beneficial towards the full understanding of the toxicity mechanism of IAPP oligomers.

The human IAPP (hIAPP) aggregates easily and therefore, it is difficult to characterize its structural features by standard biophysical tools. Whereas the rat version of IAPP (rIAPP), that differ by six amino acids compared with hIAPP, is not prone to aggregation and does not form amyloid fibrils. But similar to hIAPP it also demonstrates random-coiled nature in solution. The structural propensity of rIAPP has been studied as a hIAPP mimic in several of the recent works [2–5, 12, 13]. However, the overall shape of it in solution still remains elusive. Here, we focus on resolving the low resolution structure of this peptide in solution by small angle X-ray scattering (SAXS) measurements and with complementary nuclear magnetic resonance (NMR) data. Further, these structural information were utilized to assess the ability of the three commonly used classic energy functions (force fields) for simulating the intrinsic disordered peptides/proteins.

2 Materials and Methods

2.1 Systems and NMR Spectroscopy

The 15N uniformly labeled rIAPP sample preparation and NMR experiments have been described in details in our last publication [3]. Briefly, 15N-HSQC, 15N-TOCSY-HSQC (τmix = 120 ms) and 15N-NOESY-HSQC (τmix = 200 ms) were performed using a Bruker Advance II 700 MHz spectrometer at 25 °C. The data were collected on a sample containing 50 μM 15N uniformly labeled rIAPP (in 5 mM potassium phosphate buffer, 10 mM KCl, 3 % D2O, pH 6). The spectra were analyzed and the chemical shifts were assigned with CARA software (www.nmr.ch). Peaks were picked manually from the 3D 15N-NOESY-HSQC spectrum. The peak list, together with the chemical shift assignments were used as the input for structure calculations by CYANA 2.0 [14].

2.2 Small Angle X-ray Scattering Experiments

The synchrotron radiation X-ray scattering data for rIAPP were collected following standard procedures on the X33 SAXS camera of the EMBL Hamburg located on a bending magnet (sector D) on the storage ring DORIS III of the Deutsches Elektronen Synchrotron (DESY).

2.3 Molecular Dynamics Simulations

Classic molecular dynamics simulations were performed on rIAPP using three different force fields, AMBER03 [15] with recent modifications [16], OPLSAA [17] and CHARMM with CMAP [18, 19]. All simulations lasted 50 ns, which began with the NMR model 1 that has a Rg value of 1.78 nm. A dodecahedron box of size 7 nm was used with 7,869 water molecules (SPC model) and four chloride ions. The simulation was performed using Gromacs simulation package [20], during which all bonds involving hydrogen atoms were constrained in length according to LINCS protocol [21] with the integration step 2 fs. Non-bonded pair lists were updated every five integration steps. The protein and the water were separately coupled to the external heat bath with the relaxation time of 0.1 ps. The structure snapshots were output every 1 ps. Electrostatic interactions were treated with the particle mesh Ewald method [22] with a cutoff of 0.9 nm, while for the van der Waals interactions a cutoff of 1.4 nm was used. The simulations were repeated three times for each force field with different initial velocities. The helical structures of peptides were assessed by DSSP algorithm [23].

3 Results

3.1 Solution Structure Obtained from SAXS Measurements

Solution X-ray scattering experiments have been performed with the aim to determine the low resolution structure of rIAPP in solution. SAXS patterns from solutions of the peptide were recorded as described in “Materials and Methods” to yield the final composite scattering curve in Fig. 7.1a, showing that the peptide is monodispersed in the solution. Inspection of the Guinier plots at low angles indicated good data quality and no protein aggregation. The radius of gyration R g of rIAPP is 1.83 ± 0.1 nm and the maximum dimension D max of the peptide is 6.4 ± 0.4 nm (Fig. 7.1b). The solution shape of rIAPP was restored ab initio from the scattering pattern in Fig. 7.1a using the dummy residues modeling program DAMMIN [24], which fitted well to the experimental data in the entire scattering range (a typical fit displayed in Fig. 7.1a, red curve, has the discrepancy of χ2 = 1.061). Ten independent reconstructions yielded reproducible models and the average model is displayed in Fig. 7.2. rIAPP appears as an elongated molecule with a length of 6.4 nm and an overall spiral-like shape.

Fig. 7.1
figure 1

Small-angle X-ray scattering data of rIAPP. a Experimental scattering data (circle) and the fitting curves (line; green experimental, red calculated from ab initio model) for rIAPP. b The distance distribution function of the same peptide

Full size image
Fig. 7.2
figure 2

Superposition of the DAMMIN model of rIAPP (mashed shape) with the NMR solution structure (blue) of the same peptide. The two models are rotated clockwise by around 90° along the Y-axis. The two helical regions are residue 9–12 and residue 15–17

Full size image

The solution structure of rIAPP has also been studied by us through NMR spectroscopy in our previous work [3]. Although the residual helical structure was relatively well-defined, the global structures generated by applying NMR constraints were quite heterogeneous, which was due to the lack of long-range constraints. Twenty structural models generated from NMR constrains are shown in Fig. 7.3. The radius of gyration (R g) for this structure ensemble ranges from 1.1 to 1.78 nm. Out of the 20 structural models we do found one model (NMR1) with an R g of 1.78 nm. This NMR1 model and the solution shape, determined by SAXS (R g = 1.83 nm), were superimposed with the program SUBCOMP [25] which showed good fitting with an r.m.s. deviation of 1.47 Å (Fig. 7.2). When we repeated the structural refinement processes using NMR constrains, only 5 % of the generated structural models have R g larger than 1.73 nm. Clearly combing local structural information from NMR measurement with the global structural profile from SAXS can greatly narrow down the configuration space of the model structures. Thus the SAXS refined structural models of rIAPP are quite extended. Previously the dynamics of contact formation between the N- and C-termini in monomeric IAPP from human and rat were probed by triplet quenching technique [5]. This showed that the relaxation rates are approximately 2-fold faster for hIAPP than for rIAPP, which indicated that rIAPP is always more expanded than hIAPP.

Fig. 7.3
figure 3

The superimposing of 20 structure models from NMR constraints. The average helical residue number is 5.75

Full size image

3.2 Evaluation of Three Force Fields

Three all-atom force fields, AMBER03d [16], CHARMM [18] and OPLSAA [17], were employed to simulate this peptide in the presence of explicit water. The initial structure was taken from the elongated NMR1 model (Fig. 7.2 and 7.4a). The back-calculated scattering curve from the NMR1 model with CRYSOL program [26] gave a χ2 value of 2.93 after superimposition onto the experimental data (Fig. 7.4c, solid line), indicating the consistency with both the experimental NMR- and SAXS data. Unfortunately, the configuration of the solution model cannot be maintained within the three force fields. In the AMBER03d and OPLSAA force field simulations the peptide collapses quickly in the first 10 ns from the initial R g value of 1.78–1.1 nm (Fig. 7.5a). These compact conformations (Fig. 7.5b) were nearly unchanged during the following 40 ns simulations. The back-calculated scattering curve from such compact model (Fig. 7.5b) with CRYSOL resulted in a χ2 value of 7.49 after superimposition with the experimental data (Fig. 7.4c, blue dashed line).

Fig. 7.4
figure 4

Comparison of NMR1 model (a) and simulation model of AMBER03d (b) scattering intensities between experimental data and calculated from structural models (c)

Full size image
Fig. 7.5
figure 5

Evolution of radius of gyration, Rg, a and the number of backbone hydrogen bonds b of rat IAPP from three different force fields simulations. Each data point is an averaged value during 1 ns simulation

Full size image

This compact feature of rIAPP has also been proposed from theoretical simulations and infrared spectroscopy data [4]. However, a compact model is not consistent with the experimental SAXS- and NMR data presented. The simulated conformations using CHARMM force field have larger R g values, than the other two force fields (Fig. 7.5a), however, they are highly helical (Fig. 7.5b). The average number of helical residues is above 19 (more than half of the residues of the peptide) during the 50 ns trajectory. Such highly helical propensity is in contradiction with the overall random-coil nature of the peptide resolved by NMR measurement [2, 3]. The average helical residue number of 20 NMR structural models (Fig. 7.3) is only 5.75. Two more simulations in each type of force fields have been performed with different initial velocities in which similar results were obtained.

4 Conclusions

In summary we combined NMR measurements, which mainly provided local secondary structure information in this case, and SAXS data, which delivered a global structural profile, to resolve an extended, random-coiled structural model for the rat IAPP peptide. The presented structure will provide an invaluable reference to further study the conformational propensity of more disease related hIAPP.