β-Hairpin of Islet Amyloid Polypeptide Bound to an Aggregation Inhibitor

In type 2 diabetes, the formation of islet amyloid consisting of islet amyloid polypeptide (IAPP) is associated with reduction in β-cell mass and contributes to the failure of islet cell transplantation. Rational design of inhibitors of IAPP amyloid formation has therapeutic potential, but is hampered by the lack of structural information on inhibitor complexes of the conformationally flexible, aggregation-prone IAPP. Here we characterize a β-hairpin conformation of IAPP in complex with the engineered binding protein β-wrapin HI18. The β-strands correspond to two amyloidogenic motifs, 12-LANFLVH-18 and 22-NFGAILS-28, which are connected by a turn established around Ser-20. Besides backbone hydrogen bonding, the IAPP:HI18 interaction surface is dominated by non-polar contacts involving hydrophobic side chains of the IAPP β-strands. Apart from monomers, HI18 binds oligomers and fibrils and inhibits IAPP aggregation and toxicity at low substoichiometric concentrations. The IAPP β-hairpin can serve as a molecular recognition motif enabling control of IAPP aggregation.


Results and Discussion
Structure of the IAPP:HI18 complex. In ( 1 H-15 N) HSQC NMR spectroscopy, free [U-13 C, 15 N]-IAPP exhibits limited resonance dispersion as a consequence of its intrinsically disordered nature (Fig. 1b). Upon addition of [NA]-HI18 resonance dispersion increases, with several resonances experiencing a downfield shift into the chemical shift region characteristic of β -sheet conformation, indicating that binding to HI18 is coupled to folding of IAPP. The structure of the IAPP:HI18 complex was determined by high-resolution liquid-state NMR spectroscopy ( Fig. 1c and Supplementary Table 1, PDB: 5K5G). The folded core of the complex comprises residues 10-30 of IAPP and residues 14-56 of the HI18 subunits (Fig. 1a). IAPP adopts a β -hairpin conformation upon binding to HI18. The β -hairpin is wrapped by the two HI18 subunits, each contributing two α -helices and a short β -strand that extends the IAPP β -hairpin, giving rise to a four-stranded intermolecular antiparallel β -sheet (Figs 1c and 2a). The IAPP:HI18 complex exhibits C2 pseudosymmetry, with a C2 axis passing through the center of the IAPP β -hairpin and through the Cys-28 disulfide bond that connects the HI18 subunits.
Molecular dynamics simulations have observed significant population of β -hairpin structures in the conformational pool of the intrinsically disordered IAPP monomer, in particular β -hairpins with a turn in the region around Ser-20 and highest β -sheet probability in those sequence regions that also form β -strands in complex with The two subunits of HI18 are linked by a disulfide bond involving the Cys-28 residues (yellow). Residues that are exchanged in HI18 compared to ZAβ 3 and AS10 are underlined in blue and magenta, respectively. The segments that constitute the folded core of the IAPP:HI18 complex are underlined in black. The positions of α -helical and β -sheet secondary structure in the complex are indicated by cylinders and arrows, respectively.  HI18 [22][23][24] . These β -hairpins have been described as the most stable conformations of monomeric, free IAPP 23 . The good agreement of β -hairpin conformations observed in MD simulations of free IAPP with the structure of HI18-bound IAPP suggests that conformational selection may be involved in the binding of HI18 to IAPP. However, the present experimental data does not allow to determine if and to what extent binding is governed by conformational selection.
Interestingly, a high sensitivity of IAPP amyloidogenicity and toxicity to mutations at residues 20, located at the center of the β -hairpin turn, has been reported [25][26][27] . The Ser-20 to Gly substitution is the only known polymorphism of human IAPP and is associated with earlier onset of T2D. It leads to increased amyloid formation and toxicity 26,27 and has been proposed to support IAPP aggregation by initiating a turn-like structure 28 .

Aromatic-aromatic and aromatic-aliphatic interactions in the IAPP:HI18 complex.
Aromatic-aromatic interactions have been detected in monomeric IAPP 29 , influence the rate of fibril formation 30 , and can play a major role for the inhibition of IAPP aggregation 31 . In the IAPP:HI18 complex, Phe-23 of IAPP is in contact with Phe-15 of IAPP and Phe-30 of one HI18 subunit (Fig. 2e). Phe-15 and Phe-23 are located in the central core of the IAPP:HI18 complex, where they pack against a cluster of hydrophobic side chains of HI18, including one of the Ile-34 residues that were exchanged in HI18 in comparison to the ZAβ 3 scaffold sequence.
For the overall amino acid sequences a comparatively high similarity between IAPP and Aβ has been noted (24% identity and 47% similarity) 32,33 . Interestingly, the sequence alignment based on the positions of the β -strands in the HI18-and ZAβ 3 -complexes does not match the alignment based on sequence similarity (Fig. 2d). Obviously, the sequence features determining the common adoption of the β -hairpin conformation are not directly reflected in the sequence similarity alignment.
There is increasing evidence for molecular links between different protein aggregation diseases. For example, Aβ plaques frequently occur in Lewy body disease while α -synuclein Lewy bodies are a feature of most of the AD cases 34,35 , and α -synuclein cross-seeds tau aggregation 36 . Epidemiological studies show an increased risk of AD in people with diabetes 37 . Intravenous injection with preformed Aβ fibrils into human IAPP transgenic mice triggered IAPP amyloid formation in the pancreas, and IAPP was detected in cerebral and vascular Aβ deposits in brains of AD patients 38,39 . IAPP and Aβ moreover interact at the pre-fibrillar stage by forming hetero-oligomers, a reaction affecting the formation of homo-oligomers and homo-fibrils of both polypeptides 40,41 . Interestingly, the regions of IAPP and Aβ involved in hetero-oligomer formation are in good agreement with those forming β -hairpins in complex with HI18 and ZAβ 3, respectively 32 . Thus, it will be interesting to test the hypothesis that hairpin structures such as those displayed in Fig. 2c, which present largely hydrophobic surfaces, are involved in hetero-oligomer formation, in analogy to the previously reported formation of Aβ homo-oligomers containing β -hairpins 42,43 . HI18 inhibits IAPP aggregation at substoichiometric concentrations. The inhibition of IAPP aggregation by HI18 was studied by a Thioflavin T fluorescence assay and atomic force microscopy (AFM). In the absence of HI18, the aggregation of a solution of 10 μ M IAPP occurred with a lag time of only a few minutes and a half-time (t 1/2 ) of 1.2 h (Fig. 3a), yielding IAPP amyloid fibrils (Fig. 3b). Addition of HI18 led to a concentration-dependent inhibition of aggregation, with a prolongation of the half-time of aggregation by 3-fold (t 1/2 = 3.6 h), 12-fold (t 1/2 = 14 h), and 28-fold (t 1/2 = 35 h) in the presence of a 1:100, 1:10, and 1:1 HI18:IAPP molar ratio, respectively (Fig. 3a). At a 10:1 HI18:IAPP molar ratio, no increase in Thioflavin T fluorescence was detected during a 5-day experiment, and no amyloid fibrils or other IAPP aggregates were discernible by AFM (Fig. 3a,b).
To assess the impact of HI18 on IAPP toxicity in a pancreatic islet cell culture system, we evaluated the viability of human 1.1B4 cells 44 in the absence and presence of HI18. When IAPP was aged at a concentration of 10 μM for 1 hour and diluted into the cell culture medium to a final concentration of 1 μ M, a significant decrease in cell viability was observed in an MTT assay (Fig. 3c), in agreement with cytotoxicity of IAPP oligomers and/or fibrils 45 . However, when IAPP was aged in the presence of HI18, the cell viability was largely rescued, demonstrating that HI18 inhibits toxic aggregate formation (Fig. 3c).
The potent aggregation inhibition at substoichiometric HI18 concentrations reveals that HI18 interferes with IAPP aggregation not only by removal of IAPP monomers from the aggregation reaction. In addition, HI18 must exert an activity which effectively impedes fibril nucleation or elongation. Substoichiometric concentrations of HI18 inhibited IAPP aggregation also in seeded aggregation assays, in which the primary nucleation step of the aggregation reaction is bypassed by the addition of preformed fibril seeds (Supplementary Fig. 1). This demonstrates that HI18 affects at least one of the reaction steps subsequent to primary nucleation, i.e. secondary nucleation or elongation 46 . Interference with fibril nucleation or elongation at substoichiometric concentrations requires a high-affinity interaction with oligomeric or fibrillar species. We could not observe such interactions during the lag-phase of aggregation due to the absence of detectable amounts of higher-order assemblies (Fig. 3b). Therefore, we enriched IAPP oligomers and fibrils by permitting aggregation to proceed in the absence of HI18, added HI18 when approximately the middle of the growth phase was achieved, and analyzed the binding of HI18 to IAPP aggregates by HI18-targeted immunogold-labeling and transmission electron microscopy (TEM) (Fig. 3d). HI18 was indeed localized on fibril surfaces, where it could potentially interfere with secondary nucleation (Fig. 3d, upper panels). In addition, HI18 was associated with smaller assemblies, which might be involved in primary nucleation (Fig. 3d, arrows). Due to limited resolution and contrast, however, we cannot exclude that these smaller assemblies represent aggregates other than IAPP oligomers. Enrichment of HI18 at fibril ends, in line with a potential effect on fibril elongation, was not evident from the TEM images. When HI18 addition was omitted nanogold particles were rarely observed and did not co-localize with IAPP aggregates, demonstrating the absence of unspecific immunogold labeling of IAPP aggregates (Fig. 3d, bottom panels).
A possible explanation for the potent interference of HI18 with the nucleation of aggregation would be a high-affinity interaction of HI18 with β -sheets present in intermediates on the IAPP aggregation pathway. The formation of intermediate β -sheet structures has in fact been implicated in the early steps of the amyloid formation reaction: the population of β -hairpin conformations in the monomeric state correlates with the amyloidogenicity of IAPP variants, and dimer formation by β -hairpin stacking has been suggested to occur on-pathway based on MD simulations 22,24,47 . Moreover, two-dimensional IR spectroscopy detected an initial formation of β -sheet secondary structure exactly in those regions of the IAPP sequence that also form β -strands in the HI18-complex 48 .

Conclusions
This study identifies a β -hairpin conformation of IAPP that is accessible for molecular recognition. It demonstrates how the conformational preferences of IAPP can be targeted to achieve inhibition of amyloid formation. Comparison with previously observed structures of bound Aβ and α -synuclein highlights common β -hairpin molecular recognition motifs which exhibit only modest sequence similarity. The β -hairpin might represent a favorable compact motif for molecular recognition also by other classes of molecules aiming at IAPP inhibition. The β -wrapin HI18 acts as an exceptionally potent aggregation inhibitor, suggesting that these principles should be further exploited in future development of inhibitors of amyloid formation.

Preparation of IAPP.
To prepare monomeric starting material, synthetic human IAPP (EMD Millipore) amidated at the C-terminus was dissolved in 6 M guanidine hydrochloride (GdnHCl), 20 mM sodium phosphate, pH 6.0 and loaded onto a Superdex 75 10/300 GL column (GE Healthcare) equilibrated in 20 mM sodium phosphate, pH 6.0. Monomeric IAPP eluted at a volume of ~17 mL and was stored at − 80 °C. For NMR experiments, recombinant human IAPP was produced using a cleavable fusion construct as described 17 . Preparation of β-wrapin HI18. HI18 was selected from a β -wrapin phage display library as described 17 .
Protein expression was carried out using a pET302/NT-His vector encoding HI18 with an N-terminal His 6 -tag in E. coli BL21(DE3) cells in LB medium. Expression was induced with 1 mM IPTG at OD 0.6-0.8 followed by overnight incubation at 27 °C. The cell pellet was resuspended in 20 mM imidazole, 500 mM NaCl, 20 mM sodium phosphate, pH 8.0, containing EDTA-free protease inhibitors (Roche Applied Sciences) and lysed by a cell disrupter (Constant Systems). Insoluble material was removed by centrifugation at 28,000 × g and the supernatant was loaded on a HisTrap FF 5 ml column (GE Healthcare). The dimeric form of HI18 was collected from a HiLoad 16/60 Superdex 75 size-exclusion chromatography column (GE Healthcare) in 20 mM sodium phosphate, pH 6.0. For NMR experiments, HI18 was expressed in M9 minimal medium supplemented with 15 N-NH 4 Cl (1 g/l) and 13 C 6 -glucose (2 g/l) and purified as above.
NMR and structure determination. The NMR spectra were acquired at 25 °C using VNMRS instruments (Varian) at proton frequencies of 800 and 900 MHz, each equipped with a cryogenically cooled Z-axis pulse-field-gradient (PFG) triple resonance probe. NMR samples contained [U-13 C, 15  Structure calculations based on NOE distance restraints and dihedral angle restraints were accomplished with a modified version of CNS v. 1.2.1 53,54 using an optimized version of the PARALLHDG force field. The MD protocol contained 30 ps high temperature torsion angle dynamics (10,000 K) and 20,000 steps during two cooling phases (2,000 K and 50 K). Ten lowest energy structures (overall CNS energy) out of 100 calculated were selected and deposited (PDB:5K5G). 90.0% of the residues are found in the core regions, 9.9% in the allowed regions and 0.1% in the generous regions of the Ramachandran diagram. All structural figures were generated using PyMOL Molecular Graphics System (Schrödinger).
Aggregation assay. Fluorescence was monitored in samples containing HI18 and IAPP at indicated concentrations in 20 mM sodium phosphate, pH 6.0 containing 40 μ M Thioflavin T (ThT) in a final volume of 100 μl in a round-bottom 96-well black plate (Nunc). Aggregation was performed at 28 °C on an Infinite M1000 plate reader (Tecan). Fluorescence was measured at 490 nm with an excitation Wavelength of 450 nm. To obtain half-times of aggregation the data was fitted to a model of secondary nucleation dominated aggregation kinetics using the program AmyloFit 55,56 . Atomic force microscopy (AFM). 50 μ l of 10 μ M solutions of IAPP aggregated in presence and absence of HI18 were adsorbed for 1 h to a freshly cleaved mica surface followed by washing with milliQ water and drying with a gentle stream of N 2 gas. Imaging was performed under air dried conditions in intermittent contact mode in a JPK Nano Wizard II atomic force microscope using a silicon cantilever with silicon tip (OMCL-AC160TS, Olympus) with a typical tip radius of 9 ± 2 nm, a force constant of 42 N/m and a line rate of 0.5 Hz. The images were processed using JPK Data Processing software.
MTT cell viability assay. The viability of a human pancreatic beta cell line in the presence of IAPP and HI18 was tested with an MTT assay (Cell Proliferation Kit I, Roche Diagnostic). 1.1B4 cells (purchased from Public Health England) were seeded in 96-well tissue culture plate at a density of 10,000 cells/well in 100 μ l of media (RPMI-1640, 10% fetal calf serum) and incubated for 24 h. Protein samples were aged under amyloid formation conditions as described above for 1 h at an IAPP concentration of 10 μ M and diluted into the cell culture medium to a final concentration of 1 μ M (the concentration of HI18 in the samples is given in Fig. 3c). Following application of the protein samples, cells were further incubated for 24 h. To assess the effects of test samples on the cells, MTT was added to the cells at a final concentration of 0.5 mg/ml in PBS followed by incubation for an additional 4 h at 37 °C. Next, 100 μ l of the solubilization solution (DMSO) was added and the absorbance at 590 nm was measured with 620 nm as a reference wavelength in an Infinite M1000 plate reader (Tecan). The data was normalized to the value of untreated control cells. All cell cultures were maintained in a 5% CO 2 humidified atmosphere at 37 °C.
Immunogold labeling and transmission electron microscopy. A sample of IAPP at a corresponding monomer concentration of 25 μ M was withdrawn during the growth phase of aggregation and incubated for 1 h after addition of HI18 at a final concentration of 2.5 μ M. For immunogold-labeling of β -wrapin HI18, protein sample was adsorbed on formvar-carbon coated copper grid (Plano) for 30 min. After blocking using 5% (w/v) BSA in PBS for 60 min, grids were incubated with goat anti-affibody antibody (Abcam) which recognizes HI18 as proved in preceding experiments, diluted 1:200 in 1% (w/v) BSA in PBS for 1 h. After being washed six times in PBS, grids were incubated with gold-labeled rabbit anti-goat IgG (Abcam, 10-nm gold) diluted 1:20 in 1% (w/v) BSA in PBS for 1 h, followed by six washes in PBS. Postfixation was done with 1% (w/v) glutaraldehyde in PBS for 5 min. After excessive washing in water, the sample was negatively stained using 2% (w/v) uranyl-acetate in water for 5 min. Analysis was performed on a Zeiss EM 902A transmission electron microscope operated at 80 kV.