Elucidating the Role of Disulfide Bond on Amyloid Formation and Fibril Reversibility of Somatostatin-14

Background: Peptide/protein hormones are stored as amyloids within endocrine secretory granules. Results: Disulfide bond cleavage enhances conformational dynamics and aggregation kinetics in somatostatin-14, resulting in amyloid fibrils with increased resistance to denaturing conditions and decreased reversibility. Conclusion: Disulfide bond could be a key modulating factor in somatostatin-14 amyloid formation associated with secretory granule biogenesis. Significance: Defective disulfide bonding might cause dysregulation of hormone storage/secretion. The storage of protein/peptide hormones within subcellular compartments and subsequent release are crucial for their native function, and hence these processes are intricately regulated in mammalian systems. Several peptide hormones were recently suggested to be stored as amyloids within endocrine secretory granules. This leads to an apparent paradox where storage requires formation of aggregates, and their function requires a supply of non-aggregated peptides on demand. The precise mechanism behind amyloid formation by these hormones and their subsequent release remain an open question. To address this, we examined aggregation and fibril reversibility of a cyclic peptide hormone somatostatin (SST)-14 using various techniques. After proving that SST gets stored as amyloid in vivo, we investigated the role of native structure in modulating its conformational dynamics and self-association by disrupting the disulfide bridge (Cys3–Cys14) in SST. Using two-dimensional NMR, we resolved the initial structure of somatostatin-14 leading to aggregation and further probed its conformational dynamics in silico. The perturbation in native structure (S-S cleavage) led to a significant increase in conformational flexibility and resulted in rapid amyloid formation. The fibrils formed by disulfide-reduced noncyclic SST possess greater resistance to denaturing conditions with decreased monomer releasing potency. MD simulations reveal marked differences in the intermolecular interactions in SST and noncyclic SST providing plausible explanation for differential aggregation and fibril reversibility observed experimentally in these structural variants. Our findings thus emphasize that subtle changes in the native structure of peptide hormone(s) could alter its conformational dynamics and amyloid formation, which might have significant implications on their reversible storage and secretion.

The storage of protein/peptide hormones within subcellular compartments and subsequent release are crucial for their native function, and hence these processes are intricately regulated in mammalian systems. Several peptide hormones were recently suggested to be stored as amyloids within endocrine secretory granules. This leads to an apparent paradox where storage requires formation of aggregates, and their function requires a supply of non-aggregated peptides on demand. The precise mechanism behind amyloid formation by these hormones and their subsequent release remain an open question. To address this, we examined aggregation and fibril reversibility of a cyclic peptide hormone somatostatin (SST)-14 using various techniques. After proving that SST gets stored as amyloid in vivo, we investigated the role of native structure in modulating its conformational dynamics and self-association by disrupting the disulfide bridge (Cys 3 -Cys 14 ) in SST. Using two-dimensional NMR, we resolved the initial structure of somatostatin-14 leading to aggregation and further probed its conformational dynamics in silico. The perturbation in native structure (S-S cleavage) led to a significant increase in conformational flexibility and resulted in rapid amyloid formation. The fibrils formed by disulfide-reduced noncyclic SST possess greater resistance to denaturing conditions with decreased monomer releasing potency. MD simulations reveal marked differences in the intermolecular interactions in SST and noncyclic SST providing plausible explanation for differential aggregation and fibril reversibility observed experimentally in these structural variants. Our findings thus emphasize that subtle changes in the native structure of peptide hormone(s) could alter its conformational dynamics and amyloid formation, which might have significant implications on their reversible storage and secretion.
Protein/peptide hormones secreted through the regulated secretory pathway are known to be stored in a highly concentrated form within specialized membrane-enclosed structures, known as "secretory granules" (1)(2)(3). Because of the intricate regulation of storage and secretion, the stored form of these hormones must possess properties of stability as well as reversibility. The exact structural form of these aggregates within the granule is thus of great interest. Recently, it was suggested that peptide/protein hormones under the regulated secretory pathway are stored in the form of amyloid-like structures within the secretory granules (4).
Amyloids are highly ordered protein/peptide aggregates that have conventionally been associated with diseases like Alzheimer, Parkinson, and type 2 diabetes mellitus (5). However, increasing examples of amyloid-like structures in the nondisease context suggest that amyloid formation might be a generic property of proteins/peptides (6). Studies focusing on "functional amyloids," i.e. amyloids involved in native biological functions, have thus grown in recent times. Examples include the curli protein fibrils (7) of Escherichia coli that aid in surface attachment and colonization of the bacteria and the amyloidlike form of mammalian protein Pmel17 that is used as a template for synthesis of pigment melanin (8). Augmenting our understanding of the potential functional role of amyloids in mammalian systems was the finding that various peptide/protein hormones showed the ability to form amyloid under granule relevant conditions in vitro (4). Furthermore, the reversible nature of these hormone aggregates has also been reported (4,9). However, the "factors" modulating hormone aggregation and reversibility of these aggregates must be understood to elucidate the mechanism(s) by which amyloids are utilized as a storage depot for the hormones. Additionally, the correct posttranslation modification of proteins within the Golgi apparatus is important in maintaining their native structure. This could be crucial for the regulated formation of amyloid and subsequent release of the hormones. One of the essential factors governing protein/peptide structure and stability is the disulfide bond (10 -12). The role of disulfide bond on structure, oligomerization, and amyloidogenicity of several proteins relevant to diseases/biological functions has previously been postulated by various research groups (10,(13)(14)(15).
In this study, we chose a representative peptide hormone somatostatin-14 ( Fig. 1), which is targeted to the regulated secretory pathway, as a model system to understand the role of disulfide bond (Cys 3 -Cys 14 ) in controlling its conformational flexibility, self-association, and fibril disassembly (monomer release) profiles. Somatostatin-14 (SST-14) 4 is a small cyclic peptide hormone secreted by hypothalamus, one of the well known functions of which is to inhibit the release of growth hormone (16). Other important functions of somatostatin-14 include inhibition of gastrin and gastric acid secretion, inhibition of insulin and glucagon secretion in pancreas, and regulation of amyloid-␤ peptide, A␤42, in brain (17)(18)(19) Somatostatin-14 has previously been reported to form amyloid fibrils in vitro (4,20). In this study, we first show that somatostatin-14 is stored as amyloid-like aggregates in rat hypothalamus tissues (Fig. 2). Next, to understand the role played by the disulfide bond in modulating the structure of somatostatin, we performed structural analysis of two forms of the peptide as follows: the disulfide-bonded form (SST-14) and the disulfidereduced "noncyclic" form . Although ncSST has previously been shown to exist endogenously within cells (21), its biological implication is not yet demonstrated. Our NMR and MD simulation results reveal significant differences in conformational flexibilities of the native (cyclic) and noncyclic form of the peptide. Interestingly, our in vitro aggregation studies display accelerated aggregation kinetics of ncSST-14 compared with SST-14 in the presence of a secretory granule relevant glycosaminoglycan, heparin (22)(23)(24). Furthermore, our in silico aggregation studies reveal marked variation in the interpeptide hydrogen bonding network upon cleavage of the disulfide bridge. The simulations also reveal the involvement of almost the entire peptide in self-association, which is further supported by our H/D exchange and proteinase-K digestion results. Finally, we observed considerable differences in fibril  Photomicrographs of the hypothalamic periventricular nucleus neurons showing somatostatin immunofluorescence visualized using AlexaFluor647-conjugated secondary antibody (red). Note the co-localization of somatostatin (red) with the amyloid-specific dye thioflavin S (green, A) or the amyloid-specific antibody OC (green, B) indicating that SST-14 is stored as amyloid within the secretory granules. Scale bar, 10 m. reversibility and resistance to thermal or chemical denaturation of the SST and ncSST fibrils, with the latter showing slower monomer releasing potency and increased resistance to denaturing conditions. This study not only underlines the pivotal role of disulfide bridge(s) in regulation of protein/peptide amyloid formation and the reversibility of the aggregates, it also provides the detailed mechanism of somatostatin amyloid formation, which has substantial value for understanding its storage within secretory granules.

EXPERIMENTAL PROCEDURES
Peptides and Reagents-Somatostatin-14 used for the aggregation studies was purchased from Bachem (Bubendorf, Switzerland). Other chemicals and reagents used were purchased from Sigma and Calbiochem. Water was double-distilled and deionized using a TKA Lab Tower AFT (Niederelbert, Germany).
Amyloid Formation-One milligram of somatostatin-14 was dissolved in 500 l of 5% D-mannitol, 0.01% sodium azide and divided equally into two microcentrifuge tubes such that the final peptide concentration was ϳ1200 M. To one of these fractions, the glycosaminoglycan heparin (from a 10 mM stock in 5% D-mannitol, 0.01% sodium azide) was added such that the resulting heparin concentration was 400 M. Thus, the peptide/ heparin ratio in solution was finally 3:1. For the preparation of noncyclic somatostatin, dithiothreitol (DTT) was first added to the peptide with a final concentration of 5 mM and incubated overnight. Heparin was further added to the noncyclic peptide in the ratio mentioned above. The solutions were incubated at 37°C, and at regular intervals, thioflavin T (ThT) fluorescence and circular dichroism were performed to monitor the amyloid fibril formation, and EM was done to confirm the morphology of fibrils. All the biophysical characterization experiments, including circular dichroism, Fourier transform infrared spectroscopy, ThT fluorescence, and nuclear magnetic resonance were performed at 25°C.
Thioflavin T Fluorescence-The amyloid formation was monitored by binding and fluorescence intensity changes of the ␤-sheet-sensitive dye, ThT. 2 l of ThT solution (stock 1 mM) was added in 200 l of diluted sample (final peptide concentration 15 M). The ThT fluorescence was measured immediately in 10-mm path length quartz cuvette cells (Hellma, Forest Hills, NY) on a spectrofluorometer (Shimadzu-RF-530) with an excitation wavelength 450 nm, and emission spectra were recorded between 460 and 500 nm. The excitation and emission slit width was 5 nm for all the studies. ThT control in the experiment contained 2 l of ThT dye solution (1 mM) in 200 l of D-mannitol. ThT fluorescence at 482 nm was plotted against incubation time.
Circular Dichroism (CD)-Circular dichroism was performed to monitor the secondary structural transition during the course of SST aggregation. For CD, 5 l of incubated peptide sample was taken up and diluted to 200 l with 5% D-mannitol (final peptide concentration 30 M). The sample was transferred into a 0.1-cm path length quartz cell (Hellma), and spectra were obtained using Jasco spectropolarimeter (model J-180). All measurements were done at 25°C. Spectra were recorded over the wavelength range of 198 -260 nm. Three independent readings were taken with each sample. Raw data were processed by spectra smoothing and subtraction of D-mannitol spectra. Only heparin (10 M) did not show any significant CD signal compared with peptide during the entire incubation period (data not shown). CD results were represented in molar ellipticity (kilodegrees ϫ cm 2 ϫ dmol Ϫ1 ).
Fourier Transform Infrared Spectroscopy (FTIR)-FTIR is used for the determination of protein/peptide secondary structure. For FTIR analysis, the thin translucent pellet of KBr was made by compressing the ground potassium bromide (KBr) powder at the pressure of 7 tons by using a hydraulic pressure pump. The pellet was then kept under an IR lamp, and 5 l of the SST-14 solution (1200 M) was spotted on it and dried immediately. For background spectrum, 5 l of the 5% D-mannitol was spotted on another KBr pellet and dried. The pellet was then kept in a transmission holder, and the IR spectra in the range of 1800 -1500 cm Ϫ1 were acquired by using Bruker-Vertex-80 instrument equipped with deuterated glycine sulfate detector. For each spectrum, 32 scans at the resolution of 4 cm Ϫ1 were recorded. Raw data corresponding to amide-I region (1700 -1600 cm Ϫ1 ) were deconvoluted by the Fourier self-deconvolution method. The deconvoluted spectra were then subjected to the Lorentzian curve fitting procedure by using OPUS-65 software.
Tryptophan Fluorescence-Tryptophan (Trp) fluorescence is used to probe changes in the microenvironment of the Trp residue in the proteins/peptides (25) and thus was utilized for monitoring the protein folding and aggregation (26 -28). For tryptophan fluorescence, 1200 M stock of peptide was diluted with 5% D-mannitol such that final protein concentration was 15 M. Sample was placed in a 10-mm path length quartz cuvette cell (Hellma), and spectra were acquired using spectrofluorometer (Shimadzu-RF-530). The spectra were recorded with an excitation wavelength of 280 nm and emission wavelength range of 290 -500 nm. The excitation and emission slit width was 5 nm for all the studies.
The time-resolved Trp fluorescence intensity decay kinetic experiments were carried out using a rhodamine 6G dye laser (Spectra Physics, Mountain View, CA), pumped by an Nd:YAG laser (Millenia X, Spectra Physics) and a time-correlated singlephoton counting setup coupled to micro-channel plate photomultiplier (model R2809u; Hamamatsu Corp). Pulses (1-ps duration) of 885 nm radiation from the laser were frequency tripled to 295 nm by using a frequency tripler (GWU, Spectra physics). Each sample (100 M) was excited at 295 nm at a pulse repetition rate of 4 MHz. For each sample, the emission was measured at their emission maxima ( max ) as determined from Trp fluorescence spectra. The instrument-response function was obtained at the excitation wavelength using a diluted colloidal suspension of dried non-dairy coffee whitener. The width (full width at half-maximum) of the instrument-response function was ϳ40 ps. Peak counts of 10,000 were collected with the emission polarizer oriented at the magic angle (54.7°) with respect to the excitation polarizer. A three-exponential function was used to fit the time-resolved fluorescence intensity decays. The mean fluorescence lifetime ( m ) was calculated using m ϭ Α␣ i i, where ␣ i is the amplitude associated with each fluorescence lifetime i and Α␣ i ϭ 1.
Acrylamide Quenching-Acrylamide quenching was carried out to determine the relative solvent exposure of tryptophan in somatostatin before and after amyloid formation. Somatostatin monomer/fibril (100 M) was prepared in 5% D-mannitol containing 0.01% (w/v) sodium azide. Peptide samples (150 l) were mixed with increasing concentrations of acrylamide (0 -0.25 M) from a 5 M stock solution of acrylamide. For each quencher concentration, fluorescence lifetime measurements were performed as described above. The ratio 0 / was calculated and plotted against concentrations of acrylamide [Q]. 0 and are the mean fluorescence lifetimes in the absence of acrylamide and in the presence of different concentrations of acrylamide, respectively. Stern-Volmer equation (Equation 1) was utilized to calculate the k q , bimolecular rate constant of dynamic quenching (M Ϫ1 s Ϫ1 ).
N-Acetyltryptophan amide (NATA) was used as standard reference control for the fluorescence quenching studies. In case of ncSST samples, as DTT was used to reduce SST to ncSST, DTT was added in the standard reference sample (NATA ϩ DTT).
Transmission Electron Microscopy-To characterize the morphology of the somatostatin aggregates, transmission electron microscopy was performed at various time points during aggregation until amyloid formation was completed. A 2.5-l aliquot of protein of a 1200 M sample was diluted to 75 l with 5% D-mannitol, so that the final protein concentration was ϳ40 M, spotted on a glow-discharged, carbon-coated Formvar grid (Electron Microscopy Sciences, Fort Washington, PA), incubated for 5 min, washed with distilled water, and then stained with 1% (w/v) aqueous uranyl formate solution. Uranyl formate solution was freshly prepared and filtered through 0.2-micron sterile syringe filter (Millex TM , Millipore, Bedford, MA) before use. The imaging of samples was performed using a transmission electron microscope (TECNAI12 D312 FEI, Netherlands) at 120 kV with nominal magnifications between ϫ26,000 and ϫ43,000. Images were recorded digitally using the Megaview imaging system.
Congo Red Birefringence-Somatostatin solutions (1200 M) were incubated at 37°C with slight agitation until fibrils were formed. Somatostatin fibrils (SST and ncSST) were isolated by ultracentrifugation (90,000 rpm for 45 min) and stained with alkaline Congo Red (CR) dye (100 M CR in PBS containing 10% ethanol) for 15 min and spread evenly on a glass slide. After drying under vacuum, the slides were analyzed using a light microscope (Olympus SZ61 stereo zoom microscope, Japan) equipped with two polarizers. The images were first viewed under bright field and followed by observation under crosspolarized light. Those regions that display CR binding (red) were analyzed under cross-polarizer, and images were captured using an attached CCD camera.
Congo Red Absorbance-The peptide fibril/monomer (5 l) was diluted with 80 l of PBS buffer (containing 10% ethanol). Then 15 l of a 0.1 mM Congo red solution prepared in PBS containing 10% ethanol was added. The solution was used for UV absorbance measurements in the range 400 -600 nm. The Jasco V-650 spectrophotometer was used for this study. 15 l of CR solution added to 85 l of PBS containing 10% ethanol was used as a control.
Monomer Release Assay-Amyloid fibrils of the SST and ncSST were tested for their ability to release their monomer counterparts. For this, 100 l of 400 M of the fibrils in 5% D-mannitol, 0.01% sodium azide were subjected to dialysis through a 3.5-kDa cutoff membrane against 500 l of Tris-HCl buffer (pH 7.4), 0.01% sodium azide. The fibril samples were placed into a Slide-A-Lyzer mini dialysis unit system (Pierce), capped, and positioned into a 1-ml cryo-tube (Nunc, Denmark) containing 500 l of Tris-HCl buffer (pH 7.4), 0.01% sodium azide. The Slide-A-Lyzer dialysis unit was placed such that its membrane was in contact with the Tris buffer solution. Magnetic bars were kept into individual tubes and the assembled units were placed on a magnetic stirrer (Spinot, Jaibro Scientific Works, New Delhi, India). The rotation of magnetic bars provided the constant movement in the outer solution further allowing monomer release. After suitable time intervals (0, 3, 6, 12, 24, 36, and 48 h), the solution from outside the dialysis membrane was taken for analysis. To monitor the monomer release, an aliquot of 150-l solutions from the cryo-tube (outside the membrane) was taken, and tryptophan fluorescence was measured. The outside solution was returned to the cryotube, and the dialysis system was reassembled after each reading.
Denaturation Study of Fibrils-The fibrils of cyclic or noncyclic somatostatin-14 were subjected to chemical and thermal denaturation to compare their relative tolerance to denaturing conditions. Chemical denaturation was carried out by incubating the 30 M fibrils from each sample in the presence of increasing concentrations of GdnHCl (0, 0.5, 1, 1.25, 1.5, 1.75, 2-4, and 6 M) prepared in 5% D-mannitol containing 0.01% sodium azide. The solutions were incubated overnight for equilibrium unfolding at room temperature. Trp fluorescence of these samples was recorded for each concentration of GdnHCl. The shift in wavelength corresponding to maximum fluorescence intensity ( max ) was plotted against each GdnHCl concentration. For thermal denaturation study, the 30 M SST or ncSST fibrils prepared in 5% D-mannitol containing 0.01% sodium azide were placed in spectrofluorometer attached with a temperature control system. Samples were then subjected to an increasing temperature gradient (25-95°C) with a 10°C increment, and Trp fluorescence of these samples was recorded at each temperature. The shift in Trp max was plotted against each temperature. All spectra were acquired using a spectrofluorometer (Fluoromax-4, Horiba Scientific).
Proteinase K Digestion-To determine the "core" of the somatostatin amyloid, the preformed fibrils were subjected to proteinase K digestion, because of the fact that the "amyloid core" is resistant to enzyme digestion. Both fibrils and monomers (100 M) were separately treated with 20 g/ml proteinase K and incubated at 37°C. At different time points (10,20,30,45,60,80, and 480 min), the digestion profiles of the treated fibril samples relative to monomers were analyzed by mass spectrometry (Autoflex Speed, Bruker MALDI-TOF-TOF mass spectrometer; mass range 1100 -1800; mass tolerance 3 Da). The spectra obtained at different time points were plotted after normalizing peak intensities with respect to the highest intensity peak (base peak) in each spectrum.
Hydrogen/Deuterium Exchange of Somatostatin Fibrils-To determine the involvement of amide hydrogens in both forms of SST-14 fibrils, the H/D exchange experiment was performed. For this experiment, SST was dissolved in 5% D-mannitol (containing 0.01% NaN 3 ), such that the final concentration of the peptide was 1200 M. Heparin was added to the peptide solution to allow fibril formation where the peptide/heparin molar ratio was 3:1. Solutions were incubated at 37°C with slight agitation until fibril formation. Amyloid fibrils were confirmed by high ThT binding and appearance of fibrillar morphology using transmission electron microscopy. Similarly, the ncSST fibrils were prepared in the presence of heparin. The SST or ncSST fibrils were then pelleted down by ultracentrifugation at 90,000 rpm for 45 min at room temperature. The fibril pellet was resuspended in 5% D-mannitol, 0.01% sodium azide and again pelleted down by ultracentrifugation (wash). The washing step was repeated twice, and the fibril pellet was retrieved. This pellet was then thoroughly resuspended in 5% D-mannitol, and equal volumes of the resuspension were aliquoted into 2 microcentrifuge tubes. The tubes were quickly frozen in liquid nitrogen and were subjected to lyophilization. The lyophilized fibrils were used for the H/D exchange. One of the samples was added with autoclaved distilled water, and the other sample was treated with D 2 O. The D 2 O-treated sample was kept under slight rotation (50 rpm) at room temperature, allowing H/D exchange, and was quick-frozen in liquid nitrogen to stop the H/D reaction after 8 days (HD8). The water-treated sample was quickfrozen on day 0 (HD0). Both the frozen samples were then lyophilized. To determine the H/D exchange, the lyophilized samples (fibrils) were dissolved in DMSO-d 6 , 2.5% TFA, and NMR experiments were performed.
NMR Spectroscopy-All 1 H NMR experiments were performed on a Bruker Avance 500 MHz NMR spectrometer equipped with triple resonance gradient probe at 298 K. Data were processed using the Topspin 2.1 version and analyzed with Sparky 3.114. 4,4-Dimethyl-4-silapentane-1-sulfonic acid was used as an internal reference for calibration of proton chemical shifts. Spectra of peptides (SST and ncSST) were collected at a concentration of ϳ600 M in a H 2 O/D 2 O (90:10) containing 5% D-mannitol. Series of one-dimensional spectra of peptides were recorded with 1024 scans and 16 K data points in presence and absence of heparin. For sequence-specific resonance assignments, two-dimensional TOCSY with an MLEV17 sequence with a mixing time of 60 -80 ms and NOESY spectra with mixing times of 200 ms were recorded. For H/D exchange experiments, the solvent used was DMSO-d 6 , and the mixing time was 80 and 200 ms for TOCSY and NOESY, respectively. All two-dimensional experiments were acquired with 60 scans having (2048 ϫ 320) data matrix with an acquisition time of 17.04 and 2.66 ms during direct and indirect dimension. The data were zero-filled to give a (4096 ϫ 1024) matrix and processed prior to Fourier transformation with sine square bell. In all experiments, suppression of the water signal was achieved with the Excitation Sculpting scheme prior to acquisition.
Structure Calculation-Structure of SST-14 in the presence of heparin in 5% D-mannitol was calculated, whereas in the absence of heparin, NOESY signals could not be obtained, which disallowed calculation of its structure. The structural restraints for the somatostatin-14 peptide were obtained based upon NMR-derived NOEs, although the dihedral constraints were obtained using the GRIDSEARCH and FOUND (HABAS) macro of CYANA 2.1 (29), which converts the upper distance limits into torsion angle restraints. We utilized 79 NOE constraints and 47 angle constraints to calculate the conformers (supplemental Table S1). Initially, 100 conformers were generated using simulated annealing protocol in the program CYANA, and 10 of those conformers with the lowest target function values were utilized for energy minimization using molecular dynamics software package NAMD (30) and the CHARMM 2.7 force field (31,32). Energy minimization was performed using a stepwise steepest descent algorithm followed by the conjugate gradient algorithm for 15,000 steps. The energy-minimized structures were then validated using Ramachandran Plot assessment program RAMPAGE (33) (supplemental Table S1). The resulting conformers were then equilibrated in a TIP3P explicit water box for 200 ns and were finally used as a basis of the solution structure of somatostatin-14. The starting structure for studying conformational dynamics of ncSST in silico was modeled using the NMR structure of SST-14 with a cleaved disulfide bond (Cys 3 -Cys 14 ). The conformational dynamics of the two forms of the peptide was then assessed from the 200-ns trajectories. The structures and the molecular dynamics trajectories were analyzed using PyMOL (34) and VMD (35). The coordinates and NMR restraints of the resolved structure have been deposited in the Protein Data Bank under code 2MI1 and Biological Magnetic Resonance Bank (BMRB accession number 19663).
MD Simulations-All atom MD simulations were performed to characterize the conformational dynamics of the SST and ncSST, using structures resolved by NMR spectroscopy. The simulations were performed using molecular dynamics software package NAMD 2.9 (30) and the CHARMM 2.7 (31) protein force field. The trajectories were analyzed using CARMA (36) trajectory analysis package. The various conformations accessed by the peptide were assessed by clustering them into conformations with root mean square deviation within a tolerance range in software package CHIMERA (37). The most frequently accessed states and the frequency with which they are accessed were reported.
To study the aggregation propensity of somatostatin and the role of heparin in promoting their aggregation, 24 molecules of the peptide were simulated in the presence and absence of heparin for 100 ns each. The initial conformations of the 24 peptides in the simulation box were chosen according to the frequencies with which the monomeric peptide accessed various states. These trajectories were analyzed for their interpeptide hydrogen bonding and peptide-heparin hydrogen bonds using VMD (35). The snapshots were rendered using PyMOL (34). The simulations were performed with an electrostatic cutoff of 12 Å units and a van der Waals cutoff of 10 Å units along with periodic boundary conditions. The systems were first energyminimized and then heated to 310 K with a gradual scaled increase in temperature. This was followed by a constant pressure equilibration for 3 ns and then a 100-ns production run in NPT (isothermal-isobaric ensemble). All simulations were performed until the simulations reached convergence.
Immunofluorescence-To characterize the nature of somatostatin aggregates within secretory granules, rat hypothalamus tissues were immunohistochemically analyzed. Adult male Sprague-Dawley rats (200 -250 g) were used. Animals were maintained under a light/dark 12:12-h cycle maintaining the standard temperature and humidity of the animal facility. Food and water were provided ad libitum. All experimental protocols were reviewed and approved by the Institutional Animal Ethical Committee at NISER, Bhubaneswar, India. Rats were deeply anesthetized with intraperitoneal administration of a mixture of ketamine (100 mg/kg) and xylazine (10 mg/kg) and perfused transcardially with ice-cold phosphate-buffered saline (PBS) (pH 7.4) followed by 50 ml of 4% paraformaldehyde in phosphate buffer (PB) (pH 7.4). The brains were removed, postfixed in 4% paraformaldehyde followed by cryoprotection in 25% sucrose in PBS at 4°C overnight, rapidly frozen on dry ice, and mounted using Tissue-Tek in cryostat, and a block containing the hypothalamus was isolated. Serial 20-m thick coronal sections through the rostro-caudal extent of the hypothalamus were cut on a cryostat (Leica CM3050S, Leica Microsystems, Nussloch GmbH, Germany), collected on poly-L-lysine (Sigma)coated glass slides, and stored at Ϫ20°C until processed further. Before proceeding with immunostaining, slides were kept at room temperature for 15 min. The sections were first washed twice with PBS (pH 7.4). Antigen retrieval was performed on the sections by treating them with 0.5% Triton X-100 (in PBS) for 20 min at room temperature. The sections were next washed with PBST (PBS with Tween 20 (pH 7.4)). Nonspecific antigenic sites were blocked using PBST containing 2% bovine serum albumin (BSA). The sections were sequentially immunostained, first with rabbit polyclonal somatostatin antiserum (Chemicon) at 1:1000 dilution and then with anti-amyloid OC antibody (rabbit polyclonal, Millipore, 1:1000), with respective secondary antibody staining after each primary antibody. Although the sections were incubated in primary antibodies overnight at 4°C, secondary antibody incubation was carried out at room temperature for 1 h. For SST and amyloid staining (using OC antibody), AlexaFluor647-conjugated (red) and AlexaFluor488-conjugated (green) secondary antibodies (Invitrogen) were used, respectively. The sections were rinsed with PBST, mounted with mounting media containing 1% 1,4diazabicyclo[2.2.3]octane (Sigma) in 90% glycerol and 10% PBS, and visualized under Olympus IX 81 confocal microscope. For thioflavin S staining, anti-SST immunostained sections were first incubated with AlexaFluor647-conjugated secondary antibody. The sections were rinsed with PBST and treated with 1% thioflavin S (Sigma) for 10 min in dark. The sections were then washed twice with 50% ethanol/PBST followed by PBST. Sections were mounted and observed under a confocal microscope.

RESULTS
Somatostatin-14 is a 14-residue peptide hormone, where Cys 3 is disulfide-bonded with Cys 14 to form a cyclic structure. Somatostatin has been shown to form amyloid fibrils in vitro, suggesting that amyloid could be the storage state of SST in secretory granules (4,38). To substantiate this finding within mammalian tissue, we performed immunohistochemical studies using rat hypothalamus tissue sections. We utilized the rabbit polyclonal somatostatin antiserum and OC antibody (or thioflavin S dye) for somatostatin and amyloid detection within tissues, respectively. Application of somatostatin antisera (raised against somatostatin-14 peptide as well as somatostatin) from two different sources (Chemicon and Abcam) resulted in a similar immunostaining pattern in the hypothalamic periventricular neurons. Furthermore, the somatostatin antibody used in this study has recently been used for the localization of somatostatin neurons in the periventricular nucleus in the hypothalamus (39). Results of this study showing co-localization of SST and thioflavin S (or OC antibody) staining in hypothalamic tissues suggest that SST is stored within the secretory granules as amyloids (Fig. 2).
To investigate the role of native structure in the regulation of amyloid formation and its subsequent monomer release (from fibrils), we studied both native cyclic SST and ncSST using various biophysical and computational approaches. The glycosaminoglycan heparin was used as an "aggregation inducer" in our studies, as it has been suggested to accelerate protein aggregation and regulate protein/peptide hormone aggregation associated with their secretory granule biogenesis (9,22,23,40,41).
In Vitro Aggregation of Somatostatin-14-Far UV-CD spectroscopy (198 -260 nm) was used to monitor the possible conformational changes in SST during aggregation. As evident from the CD spectra, SST initially displayed unstructured conformation (random coil) because it showed negative minima at ϳ198 nm (data not shown). However, significant secondary structural change was observed during the course of aggregation (Fig. 3A) in the presence of heparin. After 6.5 h of incubation, a conformational change was observed from random coil to a '"helix-rich" structure displaying CD spectra with two distinct minima, one at ϳ208 nm and another at ϳ222 nm (Fig.  3A). The helix-rich state, however, did not further convert to classical ␤-sheet structure (single minima at ϳ218 nm) after 15 days of incubation. Instead, it transitioned to a secondary structure, where CD spectra showed two minima one at ϳ215 nm and another at ϳ240 nm after fibril formation. The observation of two negative minima in SST amyloids has been reported earlier (4). Moreover, it was previously shown that the disulfide bond chromophore can give rise to a negative minima at 240 nm in L-cystine crystals (42). Therefore, the peak at ϳ240 nm for SST fibrils could be due to the alignment/ordering of a disulfide bond after fibrillation. In the absence of heparin, SST mostly remained in unstructured conformation (random coil) throughout the period of incubation (data not shown).
The amyloid-specific dye ThT was utilized to monitor the kinetics of SST amyloid formation. ThT has been shown to bind specifically to cross ␤-sheet structures of amyloids (43,44), thus routinely used to monitor amyloid formation kinetics (45)(46)(47)(48). At the beginning of the experiment (0 h), low ThT binding was observed in both samples (SST or SST-Hep) (Fig. 3B). However, SST in the presence of heparin showed an increase in ThT binding after 3 h of incubation, which reached the maximum at ϳ6.5 h (Fig. 3B). After 13 h of incubation and thereafter, ThT fluorescence intensity steadily decreased and then remained constant after 72 h of incubation. The enhanced ThT fluores-cence during the course of incubation indicates the formation of amyloid-like structures by somatostatin. The decrease in ThT fluorescence at later time points perhaps was due to inaccessibility of the ThT-binding sites in the higher ordered "bundled packs" of SST fibrils (Fig. 3F, 360 h). There was no appreciable change in ThT fluorescence intensity observed in somatostatin control (without heparin) with prolonged incubation time (Fig. 3B).
We also utilized Trp fluorescence to evaluate structure-specific changes during SST fibrillation. A time-dependent blue shift in wavelength corresponding to Trp fluorescence maxima ( max ) was observed during aggregation (ϳ348 to ϳ337 nm) when SST was incubated in the presence of heparin (Fig. 3C). Additionally, a gradual decrease in Trp fluorescence intensity was observed during SST fibrillation (Fig. 3C). The low quantum yield of Trp in the fibril state might be due to the effect of potential quencher residues in the vicinity of Trp at this state and/or precipitation of SST fibrils. In the absence of heparin, the Trp fluorescence of SST remained unaltered even after 360 h of incubation (Fig. 3C).
We further studied the microenvironment of Trp in monomer and fibrils using time-resolved fluorescence intensity decay kinetics (Fig. 3D). All the decay kinetics were fitted to a three-exponential function, and the mean lifetime ( m ϭ ⌺␣ i i ) values derived showed lesser fluorescence lifetime for SST fibrils ( m ϭ 0.64 ns) compared with SST monomers ( m ϭ 1.71 ns). The data indicate an alteration of Trp microenvironment upon amyloid formation in SST. In addition, we performed dynamic quenching of Trp fluorescence using lifetime measurements to delineate solvent exposure of Trp before and after amyloid formation (Fig. 3E). It was interesting to observe that the k q values of SST (k q ϭ 6.27 ϫ 10 9 M Ϫ1 s Ϫ1 ) and SST fibrils (k q ϭ 9.07 ϫ 10 9 M Ϫ1 s Ϫ1 ) were of the same order of magnitude as that of NATA (k q ϭ 9.49 ϫ 10 9 M Ϫ1 s Ϫ1 ). The standard devi- Overall, these results suggest that irrespective of a different microenvironment experienced by Trp in the amyloid state compared with monomer, the solvent exposure of Trp in monomeric versus fibrillar state is not significantly altered. Although there is a blue shift of max observed during aggregation, a similar extent of solvent exposure in monomeric and fibrillar state indicates that the Trp gets only partially buried during amyloid formation.
To examine the morphological development of SST during aggregation, peptide solution was analyzed during the incubation using transmission electron microscopy. Immediately after dissolution and addition of heparin, SST showed mostly amorphous oligomers with sparsely populated thin filaments. However, a major population of filamentous fibrils (fibril width ϳ10 nm) was observed in SST incubated in the presence of heparin after ϳ6.5 h. These structures transformed into slightly thicker and straight fibrils (Fig. 3F, 13-120 h), which eventually formed thick fibrillar bundles at the end (Fig. 3F, 360 h) of the study. These mature fibrils were composed of several laterally associated filaments, which range between 10 and 12 in numbers. We observed no fibrillar structures in SST in the absence of heparin after incubation up to 360 h (data not shown). Although SST formed amyloid fibrils displaying increased ThT binding, these did not show a traditional ␤-sheet conformation in CD (Fig.  3A). To confirm the ␤-sheet conformation, FTIR study was performed. The FTIR data with fibrils showed a peak ϳ1638 cm Ϫ1 suggesting the presence of ␤-sheet content in the fibrils (Fig. 3G). Furthermore, the SST fibrils exhibited apple-green birefringence upon Congo red staining under cross-polarizer (Fig. 3H) and increased CR absorbance (Fig. 3I) indicating their amyloidogenic nature.
Noncyclic Somatostatin-14 Forms Amyloid Fibrils Instantaneously-To probe the role of disulfide bond of somatostatin-14 in its aggregation, ncSST was incubated in the presence and absence of heparin. The ncSST was random coil initially as shown by single minima at ϳ198 nm in CD (data not shown); however, after 12 min of incubation in the presence of heparin, the peptide showed transition to a helix-rich state (two minima in CD spectra at ϳ208 and ϳ222 nm). After 20 min of incubation, it converted to a ␤-sheet rich structure (negative minima at ϳ216) (Fig. 4A), which remains mostly unaltered up to 3 days of measurement. The data indicate the rapid amyloid formation by ncSST. Consistent with CD data, ThT binding study also showed a significant increase in fluorescence from the beginning of the aggregation reaction itself in the presence of heparin, which did not change during the first 20 min of incubation and fibril formation. However, after 20 min, ThT fluorescence gradually increased during fibrillation (Fig. 4B), which becomes stationary after 3 h of incubation. In the absence of heparin, however, ncSST neither showed any conformational transition (data not shown) nor any substantial change in ThT binding during the entire incubation period (Fig. 4B). Additionally, Trp fluorescence was monitored during the course of aggregation, and the data suggest that immediately after addition of heparin, a blue shift (ϳ352 to ϳ342 nm) in wavelength maxima ( max ) was observed (Fig. 4C), which remains unaltered during rest of the incubation period. We also observed an increase in fluorescence intensity immediately after addition of heparin (at 0 min). However, during the course of aggregation, the Trp fluorescence intensity gradually decreased (Fig. 4C), which could be due to the partial precipitation and/or Trp fluorescence quenching by nearby residues.
The Trp microenvironment of ncSST in monomer and fibrils (72 h old) were further probed by fluorescence intensity decay kinetics (Fig. 4D). The three-exponential fit of the data and the derived mean fluorescence lifetime value suggest that Trp in fibrils possess a shorter lifetime ( m ϭ 0.79 ns) and thus a different microenvironment compared with monomers ( m ϭ 1.89 ns). The results from dynamic fluorescence quenching using lifetime measurements (Fig. 4E), however, reveal that the extent of solvent exposure of Trp in ncSST fibrils (k q ϭ 7.72 ϫ 10 9 M Ϫ1 s Ϫ1 ) was of the same order of magnitude as that of ncSST monomer (k q ϭ 5.83 ϫ 10 9 M Ϫ1 s Ϫ1 ) and NATA (in the presence of DTT) (k q ϭ 9.81 ϫ 10 9 M Ϫ1 s Ϫ1 ). The standard deviations in m and k q were about 5-10% of the values. Collectively, the fluorescence data suggest that only a partial burial of tryptophan occurs during fibrillation of ncSST similar to SST fibrillation.
Electron microscopic studies for the morphological characterization of ncSST fibrillation displayed thin fibrillar structures in samples at early time points (Fig. 4F, 6 min). Electron micrographs at different time points during aggregation indicated the presence of long, flexible filamentous fibrils (ϳ6 -12 nm width) of ncSST (with Hep) (Fig. 4F), whereas only small oligomeric forms with no defined morphology were seen in absence of heparin (data not shown). Consistent with CD data showing ␤-sheet secondary structure (negative minima at 216 nm), FTIR of the ncSST fibrils showed a major peak at ϳ1633 cm Ϫ1 (Fig. 4G), indicative of parallel ␤-sheet content in the amyloid. The amyloid nature of the ncSST fibrils was also confirmed by the display of apple-green birefringence under crosspolarizer upon Congo red staining (Fig. 4H) as well as an increase in CR absorbance (Fig. 4I) in UV spectroscopy. Overall, our in vitro aggregation experiments suggest that the disulfide bond between Cys 3 and Cys 14 in SST-14 can significantly influence amyloid formation of the peptide.
Initial Structure of Somatostatin Leading to Fibrillation-Characterizing the underlying structural features of the cyclic and noncyclic somatostatin contributing to their differential aggregation behavior is very crucial to understand the "conformation-aggregation" relationship. To structurally characterize the monomeric somatostatin peptide (cyclic and noncyclic), we used NMR spectroscopy both in the presence and absence of heparin. We performed 1 H NMR experiments of SST to investigate its monomeric structure in water (containing 5% D-mannitol). The one-dimensional proton spectrum of SST monomer showed distinct and sharp peaks, which shifted to lower parts/ million values upon addition of heparin (Fig. 5A), indicating subtle structural changes in the peptide. The peaks in the latter case also appeared relatively broader. Our TOCSY (two-dimensional) spectrum also showed the same result, wherein the peak shifts of individual amino acids upon heparin addition were clearly observed (Fig. 5B). Because of the absence of long range NOESY cross-peaks, the structure of SST monomer could not be calculated. However, heparin addition allowed the peptide to assume a definite structure, the TOCSY and NOESY peaks of which were used to determine its structure. The NOESY spectrum in Fig. 5C shows the long range and medium range NOEs used for structure calculation. The structure calculation was performed, as discussed under "Experimental Procedures." An ensemble of 10 structures of SST (Fig. 5D) was obtained using the CYANA program, and the average structure from the ensemble is represented in Fig. 5E. We find the structure of somatostatin-14 determined in this work to be comparable with already reported structures of SST analogs (49).
We also performed 1 H NMR experiments with ncSST (in 5% D-mannitol) to resolve its structural characteristics. In the absence of heparin, ncSST showed distinct peaks in the one-dimensional proton spectra that disappeared instantaneously upon addition of heparin. Furthermore, peak broadening was seen in this case suggesting the formation of higher order structures immediately upon heparin addition (Fig. 5F). Only few peaks indicative of side chains from aromatic amino acids were visible in the "ncSST ϩ Hep" sample (Fig. 5F, bottom). The chemical shift assignments of proton resonances in ncSST were carried out using the two-dimensional NMR experiments (Fig. 5G). We could not, however, assign the resonances corresponding to cysteine residues (Cys 3 and Cys 14 ) in the peptide. Although TOCSY data aided assignment of most of the amino acids in ncSST, the experimental NOEs were obtained only for sequential amino acids (Fig. 5H) and therefore did not yield meaningful distant restraints to determine the structure of ncSST in solution (water). This was further sup- ported by the observation of near-zero secondary chemical shift indices displayed by most of the amino acid residues in ncSST (Fig. 5I).
Furthermore, it was interesting to note that the TOCSY spectra of SST and ncSST showed significant differences (Fig. 5, B and G). The reason for the observed differences

Disulfide Bond Regulates Somatostatin-14 Amyloid Formation
could be attributed to a change in conformational ensemble properties of somatostatin resulting from an altered local environment of the amino acid residues in peptide upon release of the disulfide bridge.
Somatostatin-14 structure determined from this study was further analyzed using computational approaches. We utilized the NMR-derived structure for aggregation studies of somatostatin in silico.
Structural Dynamics of Somatostatin-14 in Silico-To understand the time-resolved properties of NMR-derived SST structure, classical MD simulations were performed in explicit solvent. All-atom MD simulations are powerful tools to study protein structural dynamics (50,51) and aggregation (52)(53)(54). Our MD simulation results reveal the tendency of SST to assume predominantly two structures with varying frequencies as evident from the distribution of the radius of gyration of the peptide main chain during the simulation (Fig. 6A, SST). The peptide also accesses other states, however, with relatively low frequencies (Fig. 6B, SST, 3, 4, and 5). The noncyclic form of somatostatin, however, displayed largely multistate dynamics (Fig. 6A, ncSST) suggesting that the cleavage of the disulfide bond had a major effect on the conformational flexibility of the peptide. It was interesting to note that ncSST assumes various states in solution, where the most frequently accessed state showed the higher helicity. However, the peptide accesses this state (state 1 in Fig. 6B, lower panel) with low probability (ϳ0.08). Interestingly, the other states assumed by ncSST (i.e. states 2-5 in Fig. 6B, lower panel) also exhibit lower frequencies. The divergence observed in ncSST structures during the simulation was due to loss of native contacts with time. Additionally, both SST and ncSST tend to deviate away from the starting structure during the simulation reflecting on the loss of native contacts with time (supplemental Fig. S1A, top). Because we used the NMR structure of SST determined in the presence of heparin for the simulation of SST, the observed R g value changes (ϳ7 to ϳ5.5 Å) during simulation could be due to the different conformation (more compact) the peptide assumes in the absence of heparin. The root mean square deviation, radii of gyration, surface area, and end-end distance profiles were smoother in the case of SST, while displaying a dynamic behavior in ncSST (supplemental Fig. S1, A and B). Somatostatin thus demonstrates a significantly high conformational flexibility with vastly fluctuating profiles of reaction coordinates in the absence of the disulfide bond.
Somatostatin Aggregation in Silico-To delineate the similarities and differences in aggregation of SST and ncSST, 100-ns MD simulations of the peptides in explicit solvent were performed. The starting positions of the peptides in the four systems (SST, SST ϩ heparin, ncSST, and ncSST ϩ heparin) were initially unbiased and randomly positioned to avoid any intermolecular contacts at the beginning of the simulation. The initial configurations of the systems can be visualized from Figs. 7A (left) and 8A (left). It is interesting to note that both the cyclic and noncyclic forms of the peptide showed very negligible selfassociation tendencies in the absence of the glycosaminoglycan heparin in the solvent box consistent with our in vitro data. However, addition of heparin was observed to significantly promote self-association of peptides in both SST and ncSST systems, as evident from the snapshot of the final state of the system in Figs. 7A (right) and 8A (right). A more detailed timeline of progression of the four systems and their higher self-association tendencies in the presence of heparin is represented in supplemental Fig. S1C. Because the interpeptide contacts are one of the major driving forces for peptide/protein aggregation, we analyzed all four simulation systems for interpeptide contacts during the simulation time scale. Interestingly, both the cyclic and noncyclic forms of somatostatin-14 in the absence of heparin showed negligible interpeptide hydrogen bond development during the course of the 100-ns simulation (Figs. 7B and 8B). However, the addition of heparin to the solvent box seemed to promote the self-association of both the peptides with an increase in interpeptide hydrogen bonding (Figs. 7B and 8B and supplemental Fig. S2A). Furthermore, although both SST and ncSST systems demonstrate increased heparin interactions during the simulation, the noncyclic form of the peptide shows significantly higher interacting tendency with heparin (supplemental Fig. S2B). One of the measures of characterizing the self-association of the peptides is to monitor the cumulative surface area of the peptides over time. A gradual decline in the total surface area of SST and ncSST peptide sys- tems over time in the presence of heparin indicates self-association of peptides during the simulation (Figs. 7D and 8D). Such a decline was, however, not observed in the absence of heparin in the solvent box, thereby enforcing the role of heparin in somatostatin aggregation. The self-assembling tendencies of the two peptides in the presence of heparin were further confirmed by a decrease in the R g value of the system during the simulation (Figs. 7E and 8E). We further analyzed the specific differences in the interpeptide hydrogen bonding network in the SST and ncSST systems to identify whether disulfide bond cleavage leads to differences in hydrogen bonding network in the aggregated state. A clear difference in hydrogen bonding patterns was observed in SST and ncSST. Although the most frequently observed hydrogen bonds in both the cases remained same (Lys 4 -Cys 14 and Lys 9 -Cys 14 ), there was a major change in the hydrogen bonding patterns of other residues involved in interpeptide contacts among the two somatostatin variants (Figs. 7C and 8C). The most prominent difference was the prevalence of hydrogen bonding between hydrophobic residues Phe 6 , Phe 11 , and Trp 8 in ncSST (Fig. 8C). This was, however, not evident from the simulations of the SST system in the presence of heparin (Fig. 7C). Also, although the residues Phe 6 -Thr 12 are all involved as hydrogen bond donors in the case of SST, the H-bonding pattern in case of ncSST shows the region Lys 4 -Lys 9 acting as major H-bonding donors, whereas the C terminus of the peptide (residues Phe 11 -Cys 14 ) plays a key role of H-bonding acceptor (Fig. 8C).
Amyloid Core of Somatostatin Fibrils-With our MD simulation data showing that most of the amino acid residues are involved in interpeptide H-bonding in a "self-assembled" state of somatostatin, we performed H/D exchange (coupled with NMR) and proteinase K digestion (coupled with mass spectrometry) experiments with the fibrils to determine the involvement of peptide segment in SST and ncSST amyloid core formation. The fibrils were treated with proteinase K, and the samples were analyzed by mass spectrometry at different time points as mentioned previously (see under "Experimental Procedures"). In parallel, only the SST and ncSST monomer treated with the enzyme was used as controls.
The SST fibrils upon proteinase K treatment did not show any digested product(s) up to 80 min. At this time point, however, the SST monomer control showed two prominent fragments as follows: peaks at 1277 Da (peptide fragment excluding Lys 9 -Thr 10 -Phe 11 ) and 1404 Da (peptide fragment excluding Phe 11 -Thr 12 ). We found that the peak corresponding to fulllength peptide 1655 Da (1637 ϩ 1 H 2 O) totally diminished by this time (Fig. 9A, Monomer). The same peak, however, remained unaltered in the SST fibrils, indicating that the entire stretch of the peptide is perhaps proteinase-resistant. After 480 min of enzyme reaction, the SST monomer displayed complete degradation (Fig.  9A, Monomer), whereas the fibrils showed only partial digestion at this time point (Fig. 9A, Fibril). Apart from the major peak at 1637 Da corresponding to the intact full-length SST peptide, the smaller peaks that were observed in the fibril sample were 1277 and 1404 Da (Fig. 9A, Fibril, peaks a and b, respectively). The digestion profiles of SST fibrils therefore suggest that the entire peptide in the fibrillar state is resistant to enzyme-mediated degradation, indicating that almost all amino acid residues in somatostatin are protected and form the amyloid core. This is also consistent with the results of aggregation prediction algorithm Zyggregator (Fig. 9B) (55). To further delineate the amyloid core at residue-specific resolution, H/D exchange experiments coupled with NMR spectroscopy was carried out. The data showed that the intensity of SST amide protons of each amino acid was mostly unchanged even after subjecting the peptide fibrils to deuterium exchange for 8 days (Fig. 9C). The peak intensities of day 0 versus day 8 are shown in Fig. 9D, wherein we observe only a few peaks (Gly 2 , Lys 9 , and Phe 11 ) showing slight reduction in intensities relative to other peaks after 8 days of H/D exchange.
Similarly, the proteinase K digestion of ncSST fibrils showed that ncSST in the fibrils was not digested after 480 min of proteinase K treatment (Fig. 9E, left), as indicated by the intact 1638-Da peak corresponding to full-length ncSST in the fibrils throughout the reaction. In contrast, the ncSST monomer was digested within 10 min of the proteinase K reaction (Fig. 9E,  right). The resultant peak (1350 Da) at 10 min corresponds to the ncSST fragment excluding the sequence Thr 12 -Ser 13 -Cys 14 . At later time points, the monomer was completely digested. Our data therefore suggest that almost the entire peptide is involved in amyloid formation. The H/D exchange also shows that most of the amide protons in the fibril state were protected from H/D exchange (Fig. 9F) except for three peaks (Gly 2 , Asn 5 , and Trp 8 ), which showed a decrease in peak intensity (thus increased exchange) upon 8 day of incubation with deuterium in the solution (Fig. 9G). The partial H/D exchange of these residues after a prolonged period might be due to the partial solvent exposure of these residues in the fibrillar state.
Monomer Release and Denaturation of SST Versus ncSST Fibrils-As both the cyclic and noncyclic form of the peptide exhibited differences in conformation, aggregation, and fibril characteristics, we were interested to study whether these differences arising from structural change (S-S cleavage) in somatostatin can result in any variation in the "reversible" property of the fibrils. Previous reports show that the hormone amyloids are reversible and therefore can successfully release monomers under suitable experimental conditions (4,9). On this basis, we designed monomer release experiments by exposing the two different fibrils to physiological pH 7.4. SST or ncSST fibrils were allowed to release monomers by subjecting them to dialysis using a 3.5-kDa cutoff membrane, and the peptide release was monitored by measuring Trp fluorescence of the solution outside the membrane at different time points. We find that SST fibrils displayed spontaneous release of monomers up to 12 h, after which there was a gradual increasing pattern observed until the end of the study (48 h) (Fig. 10A, SST fib). In contrast, the ncSST fibrils showed only a slow release of monomers, with a slight increment in Trp fluorescence after 24 h (Fig. 10A, ncSST fib). Our results thus suggest that the ncSST fibrils release monomers at a much slower rate compared with the SST fibrils, indicating that the fibrils formed from the noncyclic SST might be structurally more intact as compared with SST fibrils.
To decipher whether the slower monomer-releasing tendency possessed by ncSST fibrils compared with SST fibrils is an inherent property of the fibrils, we evaluated the thermal and chemical denaturation profiles of SST and ncSST fibrils. Two methods were employed, viz. equilibrium unfolding using GdnHCl and temperature-induced denaturation. Guanidine hydrochloride-mediated denaturation as monitored by tryptophan fluorescence showed that SST fibrils were dissociated at low concentrations (0.5 or 1 M) of GdnHCl, and above these concentrations saturation was reached (Fig. 10B, SST fib). Conversely, ncSST fibrils showed gradual denaturation with increasing concentrations of GdnHCl (Fig. 10B, ncSST fib) indicating that resistance to chemical denaturation of these fibrils is relatively higher than that of SST fibrils. Similar studies using GdnHCl have previously been performed to analyze in detail the conformational strength of amyloid fibrils in vitro (56 -58). Temperature-induced denaturation is another method to study the structural stability of proteins and amyloid fibrils (59 -62). To evaluate thermal denaturation of SST versus ncSST fibrils, both fibrils were subjected to increasing temperatures (25-95°C), and Trp fluorescence of the samples was monitored. For SST fibrils, Trp fluorescence experiments showed considerable red shift of max (ϳ340 to ϳ380 nm) at 45°C itself and attained saturation, whereas for ncSST fibrils, only a relatively small max red shift of ϳ340 to ϳ350 nm was observed at this temperature, which almost remained constant thereafter (Fig. 10C). Our thermal denaturation results thus suggest that SST fibrils are thermally more labile compared with ncSST fibrils. Overall, we find the results from chemical/thermal denaturation studies to be consistent with that of the monomer release assay, all of which clearly demonstrate that ncSST fibrils exhibit a relatively higher thermal and GdnHCl resistance, which might influence its monomer releasing potency.

DISCUSSION
Although it was previously known that the protein/peptide hormones are stored as aggregates within SGs, the molecular organization was not clearly understood until recent evidence suggested that these hormones are stored as amyloids within SGs of the endocrine cells (4). The effective utilization of amyloids as a depot for hormone storage/secretion would depend primarily on the reversible nature of aggregates. In other words, the aggregates must encompass the property to store the peptide/proteins in a stable fashion and release monomers when necessary. The aggregation of the stored proteins in SGs has been suggested to be influenced by changes in environment conditions and various modulating factors (38,(63)(64)(65). Additionally, the native conformation of the proteins/peptides and its dynamicity could also alter its aggregation behavior.
This study is aimed at understanding the effect of subtle changes in the native structure of a peptide hormone and its potential manifestation on the aggregation and release profiles of its aggregate. To probe the relationship between the native structure and aggregation of peptide, we chose somatostatin-14, an important peptide hormone that is secreted via the regulated secretory pathway (66). The presence of a single disulfide bridge between residue Cys 3 and Cys 14 in somatostatin-14 makes it a simple yet attractive model system to probe altered conformational dynamics and aggregation in response to slight modifications in native structure (S-S bond cleavage). Disulfide bonds have been previously reported to play a key role in modulating the intrinsic dynamics of proteins that are crucial to their native function (10,67,68). In somatostatin and its peptidic analogs, the importance of the disulfide bond on its structural and aggregation features has been demonstrated previously (69,70).
In this study, we compared the cyclic and noncyclic somatostatin for their conformational flexibility, amyloid formation kinetics, and fibril reversibility to get insights on the role of native "cyclic" structure (-S-S-integrity) relevant to storage and secretion of the hormone.
Conformational Dynamics Controlling Aggregation and Fibril Formation by SST and ncSST-Somatostatin during the course of aggregation was observed to transit multiple stages before forming fibrils (Fig. 3A). However, it was interesting to observe that ncSST displays faster aggregation kinetics compared with SST (Figs. 3A and 4A). One possible explanation for this event could be that the linearization of the peptide possibly exposed the "aggregation-prone" regions for interpeptide interaction allowing instant amyloid aggregation in presence of heparin. Comparing the secondary structure of the SST and ncSST in fibrillar state (from CD and FTIR studies), we found SST to display a "mixed" secondary structure, whereas ncSST shows a classical ␤-sheet (Figs. 3, A and G, and 4, A and G). We infer from these results that the disulfide bond could affect the secondary structure in SST fibrils.
Next, we investigated whether any differences in conformational dynamics of SST and ncSST could be the underlying factor for observed variation in aggregation kinetics and fibril properties. To answer this question, we first determined the structure of the peptide using solution-state NMR, and we probed the effect of the disulfide bond on its conformational flexibility using all-atom molecular dynamics simulations. The lack of any distinct structure for SST or ncSST is clear from our FIGURE 9. Elucidation of somatostatin fibril core. A, proteinase K digestion profiles of fibril (left) and monomer (right) as analyzed by mass spectrometry at different time points of digestion. The result shows a peak at ϳ1637 Da corresponding to full-length peptide in the fibril sample suggesting the monomer is resistant to enzyme digestion in the fibrillar state (left); however, it gets degraded in its monomeric state (right). The masses of the digestion products are labeled adjacent to each peak as follows: peak a and peak b in the FIBRIL sample representing a fragment mass of 1277 Da (SST fragment excluding Lys 9 -Thr 10 -Phe 11 ) and 1404 Da (SST fragment excluding Phe 11 -Thr 12 ), respectively. B, aggregation-prediction algorithm Zyggregator suggests almost the entire peptide participates in amyloid formation. C, H/D exchange coupled with two-dimensional NMR of SST in fibrillar state shown at time 0 (d0) and after day 8 (d8). The overlay (rightmost) of day 0 and day 8 shows slight alteration in peak intensities of all the residues. D, histogram showing the relative variation in peak intensity in individual residues on day 0 and 8 of SST fibrils. Most of the peaks showed only a slight difference in intensity after 8 days of exchange. E, proteinase K digestion profiles of ncSST fibril (left) and monomer (right) as analyzed by mass spectrometry at the mentioned time points. The fibrils upon proteinase K digestion did not show any fragments up to 480 min, although the MONOMER was susceptible to proteinase digestion showing fragmentation by 10 min. The FIBRIL sample shows a peak at ϳ1638 (full-length SST) throughout the incubation period with proteinase K, suggesting that almost the entire peptide performs as the fibril core.  NMR results. However, upon addition of heparin, SST assumed a structure that was resolvable (Fig. 5E) by NMR, and we used this as the starting structure for our MD simulations. The MD simulations reveal that the restriction in backbone flexibility of SST due to the S-S bond causes it to display a two-state dynamics (Fig. 6A). However, the disulfide-reduced structure accesses a wider range of conformations and is highly dynamic (Fig. 6). Disulfide bond has previously been suggested to reduce conformational dynamics, increase mechanical stability, and reduce entropy in proteins/peptides (10,67,71). Additionally, this linkage has been shown to play a role in minimizing protein aggregation (14,15,72) thereby making it an effective modulator of aggregation. In this context, disulfide bonds have been suggested to limit the aggregation tendencies of highly aggregation-prone stretches in human islet amyloid polypeptide and insulin (73). The findings of this study imply that the presence of the disulfide bond in SST not only plays a crucial role in limiting the conformations accessed by the peptide but also alters its aggregation profile.
We further studied the differences in aggregation of SST and ncSST at an atomistic level using MD simulation. Simulations in the absence of heparin in the solvent box reveal an inherent tendency of somatostatin to form small aggregated clusters, which are short lived and relatively smaller in size (supplemental Fig. S1C). Interestingly, the addition of heparin shifted the dynamics toward formation of larger aggregated clumps composed of more numbers of monomers. The immediate clustering might be due to the formation of stable interactions between heparin backbone and peptides and reduction of the intrinsic dynamicity of the peptide structure. Addition of heparin may also increase the local concentration of the peptide around heparin that leads the stable interpeptide interactions resulting in higher order aggregates. The 100-ns simulations of both SST and ncSST in the presence of heparin showed a strong tendency of the peptide to self-associate with formation of intermolecular hydrogen bonds during the course of aggregation ( Figs. 7 and 8). However, a detailed analysis of the selfassembled clusters in the presence of heparin revealed that the network of stable H-bonds in ncSST was markedly different in comparison with its cyclic counterpart (Figs. 7C and 8C). It is interesting to note that an organized interpeptide H-bond network was observed in ncSST, whereas in the case of SST, the H-bond receptors and donors were scattered along the length of the peptide. These differences in H-bonding pattern suggest that, upon release of disulfide bond, the hydrogen bonding network shows greater organization in terms of the location of H-bond donors and acceptors along the primary structure.
When we specifically probed for the amino acid residue stretch that acts as the amyloid core using H/D exchange (coupled with NMR) and proteinase K digestion experiments, we found that almost the entire peptide participated in the amyloid formation ( Fig. 9), further corroborating our in silico findings. Consistent with our observation, the aggregation-prediction algorithm Zyggregator also displayed that residues 3-14 in the peptide are aggregation-prone (Fig. 9B). Similar findings, where almost the entire peptide/protein was observed to be structured or protected within the fibrils, have been reported for other proteins previously (74,75).
Reversibility and Denaturation of SST and ncSST Fibrils-For secretory granule biogenesis, the controlled formation of protein aggregates during maturation of SG at trans-Golgi is of utmost importance (2, 76 -78). However, it also required that the granules dock onto the plasma membrane and release their contents in a controlled manner upon receiving the external signal (2,38,79). Therefore, the peptide/protein hormone amyloids inside the granules require stability for it to be suitably stored in addition to being reversible, thus allowing it to release monomers to the extracellular space. Dysregulation of release could thus lead to hormone-related disorder(s). For instance, it was suggested that the R183H mutation in growth hormone FIGURE 11. Native structure controlling aggregation of somatostatin. The native somatostatin (cyclic) is restricted to display a two-state conformational dynamics due to the constraint imposed by the disulfide linkage between Cys 3 and Cys 14 residues. The peptide aggregates form amyloid fibrils upon prolonged incubation, and the resulting amyloids give away the monomers instantly. In contrast, the cleavage of the disulfide bond in somatostatin leads to increased conformational flexibility and rapid aggregation kinetics resulting in fibrils with increased structural integrity (resistance to denaturation), which release monomers at a slower rate as compared with its native form. results in prolonged retention and thus impaired the release of this hormone (80), leading to autosomal dominant growth hormone deficiency. However, the role played by the native structure of the protein/peptide in controlling its aggregation and amyloid formation associated with secretory granule biogenesis is not well understood. When we probed for differences in "fibril reversibility" resulting from disrupted disulfide linkage in somatostatin, we found that SST amyloids release the monomers at a relatively faster rate than the ncSST fibrils when exposed to physiological pH (Fig. 10A). This could most likely be due to the difference in structural arrangement of monomers within the respective fibrils. However, we cannot negate the possibility of heparin's role in modulating fibril intactness.
Furthermore, thermal and guanidine hydrochloride-mediated denaturation of fibrils reveal that ncSST amyloids possess relatively higher resistance to denaturing conditions compared with SST amyloids (Fig. 10, B and C). This could be due to the greater accessibility to the aggregation-prone state upon release of disulfide bond constraints, different interpeptide H-bonding pattern with increased involvement of hydrophobic residues, as well as increased peptide-heparin interactions. For hormones to be released to the extracellular space, the fibrils should readily release the monomers, which might not happen if the naturally occurring fibrils were composed of ncSST. Therefore, SST fibrils must have been selected by nature for regulated storage and secretion over the ncSST fibrils.
Conclusion-Peptide/protein hormones have recently been demonstrated to be stored as amyloids within the secretory granules. Storage and secretion of these hormones might be influenced by a multitude of factors, including the post-translation modification and native structure of the protein. Here, we attempted to understand whether a change in the native structure (disulfide reduction) of the representative peptide hormone somatostatin-14 leads to any difference in its conformational dynamics, aggregation profile, and fibril disassembly in the context of hormone aggregation, storage, and release. We propose that the increased conformational flexibility due to linearization (cyclic3noncyclic) plays a vital role in modulating not only the kinetics of amyloid formation but also the nature of resultant aggregates in terms of fibril intactness and reversibility (Fig. 11). Defective disulfide bond formation during posttranslational modifications could thus have significant implications related to storage, secretion, and function of the peptide hormones.