Fibrillation of human insulin B-chain by pulsed hydrogen-deuterium exchange mass spectrometry

Abstract

Insulin forms amyloid fibrils under slightly destabilizing conditions, and B-chain residues are thought to play an important role in insulin fibrillation. Here, pulsed hydrogen-deuterium exchange mass spectrometry (HDX-MS), far-UV circular dichroism spectroscopy, thioflavin T (ThioT) fluorescence, turbidity, and soluble fraction measurements were used to monitor the kinetics and mechanisms of fibrillation of human insulin B-chain (INSB) in acidic solution (1 mg/mL, pH 4.5) under stressed conditions (40°C, continuous shaking). Initially, INSB rapidly formed β-sheet-rich oligomers that were protected from HD exchange and showed weak ThioT binding. Subsequent fibril growth and maturation was accompanied by even greater protection from HD exchange and stronger ThioT binding. With peptic digestion of deuterated INSB, HDX-MS suggested early involvement of the N-terminal (1–11, 1–15) and central (12–15, 16–25) fragments in fibril-forming interactions, whereas the C-terminal fragment (25–30) showed limited involvement. The results provide mechanistic understanding of the intermolecular interactions and structural changes during INSB fibrillation under stressed conditions and demonstrate the application of pulsed HDX-MS to probe peptide fibrillation.

Significance

Amide hydrogen-deuterium exchange mass spectrometry coupled with proteolytic digestion provided mechanistic understanding of intermolecular interactions in the fibrillation of human insulin B-chain (INSB). Differences in the rate and extent of deuteration among proteolytic fragments of INSB enabled identification of the residues involved in the early interactions leading to fibrillation. Since INSB is known to be important in the fibrillation of native insulin, the results improve our understanding of the mechanisms of insulin fibrillation and can help inform the development of fibrillation-resistant insulin analogs.

Introduction

Insulin is a peptide hormone that regulates blood glucose levels and has been used in the treatment of diabetes for more than a century (1). Today, an estimated 150–200 million patients worldwide require insulin therapy (2). The safety and efficacy of insulin are compromised by its tendency to form amyloid fibrils, an instability that can lead to decreased potency and an increased risk of immunogenic side effects (3). Fibrillation also poses challenges in the production, storage, and delivery of insulin products (3). Insulin consists of two polypeptide chains, A and B, with 21 and 30 amino acids respectively (4). The chains are linked by two interchain (A7–B7 and A20–B19) disulfide bonds and one intra-A-chain (A6–A11) disulfide bond that play critical roles in its stability and biological activity (5). Despite its relatively small size, insulin has well-defined secondary and tertiary structural elements typical of larger proteins (6). The A-chain forms two antiparallel α-helices (A2–A8, A13–A19), whereas the B-chain contains a type II′ β-turn (B7–B10), a central α-helix (B9–B19), a type I β-turn (B20–B23), and a C-terminal β-strand (B24–B28) (6,7,8). The three α-helices form a globular domain against which the N- and C-terminal segments of the B-chain are packed (9). These structural features of crystalline (T-state) insulin are also observed in solution (10). Insulin can reversibly self-assemble into dimers, tetramers, and hexamers depending on solution conditions and insulin concentration (3). In the presence of Zn⁺², insulin self-assembles into hexamers containing three identical dimers, stabilized by intermolecular interactions and by a coordination complex between Zn⁺² ions and B-chain histidine (H₁₀) residues (3,6). For pharmaceutical applications, insulin is often formulated as zinc-coordinated hexamers, in part because of their increased resistance to physical and chemical degradation (11).

Under stressed conditions (e.g., low pH, high temperature, agitation), insulin hexamers dissociate into dimers and monomers that are susceptible to fibrillation (3,12). Attempts have been made to elucidate the mechanism of fibrillation of native insulin (9,13,14,15,16). NMR studies under amyloidogenic conditions (pH 2.4, 60°C) revealed that partially folded monomers are important intermediates in insulin fibrillation (9). In the partially folded state, the N-terminal segments of the A and B chains detach from the helical core and unfold, whereas the C-terminal segment of the B-chain is disordered but remains attached to the core (9). Unfolding of the N-terminal α-helix in the A-chain and disordering of the N- and C-terminal regions of the B-chain expose hydrophobic residues that may serve as early interaction sites in the formation of nuclei and fibrils (9).

The B-chain central sequence, L₁₁VEALYL₁₇, is thought to form the core of insulin fibrils and is the smallest segment able to nucleate the fibrillation of full-length insulin (13,17). X-ray crystallographic studies of L₁₁VEALYL₁₇ microcrystals have shown that the extended β-strands of fragment L₁₁VEALYL₁₇ pack in register into parallel β-sheets, such that the side chains of alternating residues L₁₁, E₁₃, L₁₅, and L₁₇ of one β-sheet intermesh with the same set of side chains of the second β-sheet to form the highly complementary “steric zipper” interface typical of amyloid-like fibrils (13). Based on this crystal structure, Ivanova et al. proposed a model of insulin fibrillation in which the B-chain central segment L₁₁VEALYL₁₇ undergoes an α-helix to β-strand conversion, which forces the disulfide-linked helical A-chain segment L₁₃YQLENY₁₉ to unfold into an extended β-strand and interface with the B-chain central segment (13). Thus, the cross-β spine of insulin fibrils consists of four β-sheets: an inner pair of B-chain L₁₁VEALYL₁₇ β-sheets situated between an outer auxiliary pair of A-chain L₁₃YQLENY₁₉ β-sheets (13). Structural transitions from α-helix to β-sheet conformations have been observed during the fibrillation of native insulin using circular dichroism (CD) spectroscopy and Fourier-transform infrared spectroscopy (18,19). Moreover, exposing the hydrophobic residues in the B-chain central region (B9–B19) by truncating the C-terminal segment (B26–B30) has been shown to increase the fibrillation propensity of native insulin (15,20), whereas point mutations in this region to more polar residues (H10D, L17Q, Y16E) delayed insulin fibrillation (16), findings that underscore the importance of the B-chain central segment. Chemical modifications to the B-chain affect the oligomeric stability of insulin and have been used to achieve desired pharmacokinetic profiles and therapeutic outcomes (21,22,23,24,25). For example, in rapid-acting insulin lispro, the positions of proline (P_B28) and lysine (K_B29) residues in native insulin have been reversed, weakening dimer-forming interactions and facilitating rapid dissociation and in vivo absorption (26).

Previous studies have used hydrogen-deuterium exchange with mass spectrometric analysis (HDX-MS) to assess the conformational stability of native insulin under stressed conditions (19,27,28) and to compare the oligomeric stability of therapeutic insulin analogs (29,30). In the studies reported here, pulsed HDX-MS and biophysical characterization methods were used to monitor the fibrillation of the human insulin B-chain (INSB) peptide with the goal of identifying early interactions involved in fibril formation. In pulsed HDX-MS, fibrillating peptides are exposed to D₂O for short time periods (e.g., in minutes) over the longer fibrillation time course (e.g., in hours) (31,32,33,34,35). The backbone amide protons in monomeric peptides are solvent exposed and exchange deuterons to a greater extent than those involved in hydrogen-bonded interactions or buried in solvent-inaccessible regions of fibrils (36). Mass spectrometry (MS) is used to quantify deuterium incorporation based on the greater isotopic mass of deuterons (2 amu) compared with protons (1 amu) (37,38). Proteolytic digestion of deuterated samples can localize deuteration sites and identify the regions of the peptide that participate in the early stages of fibrillation based on their differential deuterium exchange rates (31,32,33,34,35). The results presented here demonstrate the involvement of the central and N-terminal regions in the early stages of INSB fibrillation at acidic pH, whereas the C-terminal region showed limited involvement even in mature INSB fibrils.

Materials and methods

Materials

Research-grade human INSB was synthesized by GenScript Biotech (Piscataway, NJ), with greater than 95% purity as determined by the manufacturer and with the amino acid sequence F₁VNQHLCGSH₁₀LVEALYLVCG₂₀ERGFFYTPKT₃₀. Fibrillation kinetics studies were performed using 96-well nonbinding, black-wall, clear-bottom microtiter plates from Greiner Bio-One (#655906; Frickenhausen, Germany) covered with MicroAmp Optical Adhesive Film (Thermo Fisher Scientific, Waltham, MA) to minimize evaporation. Thioflavin T (ThioT) was purchased from Abcam (Cambridge, MA). Deuterated water (D₂O) was purchased from Cambridge Isotope Laboratories (Andover, MA). Poroszyme Immobilized Pepsin Bulk Media was purchased from Applied Biosystems (Foster City, CA) and packed into a high-performance liquid chromatography column (50 × 2.1 mm, Grace Davison Discovery Sciences, Deerfield, IL). Mass spectrometry-grade water, acetonitrile (ACN), and formic acid (FA) were from Thermo Fisher Scientific (Waltham, MA). All other chemicals used were at least reagent grade and used as received.

Sample preparation

INSB, as received, was dissolved in deionized (DI) water to obtain a clear solution at room temperature, followed by the addition of 100 mM sodium citrate buffer to produce a final buffer concentration of 10 mM (pH 4.5) and a peptide concentration of ∼1 mg/mL. Any undissolved material was removed by centrifugation at 14,000 rpm for 15 min at room temperature. The supernatant was collected and INSB concentration determined by measuring UV absorbance at 276 nm, using an extinction coefficient of 2560 M⁻¹cm⁻¹ calculated from the amino acid sequence (39).

ThioT fluorescence and turbidity measurements

The kinetics of INSB fibrillation was monitored by measuring the increases in ThioT fluorescence intensity and solution turbidity with time (31). 1.5 mL of freshly prepared INSB solution (1 mg/mL, pH 4.5) was transferred to 2 mL low protein binding Eppendorf tubes and stressed at 40°C with continuous orbital shaking at 800 rpm using a benchtop Thermomixer R incubator-shaker (Eppendorf, Hamburg, Germany). At intervals over the fibrillation time course (0–1440 min), 100-μL samples were withdrawn and transferred to 96-well microtiter plates. The extent of fibrillation was monitored by measuring ThioT fluorescence and turbidity using a Synergy Neo2 Multi-Detection Microplate Reader (BioTek Instruments, Winooski, VT). For ThioT fluorescence measurements, the withdrawn samples were incubated with 10 μM ThioT dye at 40°C for 5 min before the readings. ThioT fluorescence was bottom-read at excitation and emission wavelengths of 445 nm and 485 nm, respectively. Solution turbidity was measured at 405 nm without adding the ThioT dye. Triplicate measurements were recorded at each time point and peptide-free buffer samples served as controls.

Fibrillation kinetics was also monitored by incubating 100 μL of freshly prepared INSB solution (1 mg/mL, pH 4.5) in a 96-well microtiter plate placed inside a Synergy Neo2 plate reader and stressed at 40°C with continuous orbital shaking at 807 cpm for 13 h. For ThioT fluorescence measurements, 5 μL of ThioT stock solution (200 μM) was added to 95 μL of INSB solution before the start of the kinetic study to give a final ThioT concentration of 10 μM per well. Five replicate measurements were recorded at 15-min intervals over 13 h, and peptide-free buffer samples served as controls.

Soluble peptide fraction

To determine the loss in soluble peptide due to the formation of insoluble fibrillar aggregates, 100 μL of fibrillating INSB solution (1 mg/mL, pH 4.5, 40°C, continuous shaking at 800 rpm, Thermomixer R) was withdrawn at intervals over the fibrillation time course (0–1440 min) and centrifuged at 14,000 rpm for 15 min at room temperature. The INSB concentration in the supernatant was determined by reverse-phase high-performance liquid chromatography (rpHPLC) using an Agilent 1200 HPLC system (Agilent Technologies, Santa Clara, CA), with a C-18 analytical column (4.6 mm × 150 mm, 3 μm particle size, Agilent Technologies) and a gradient of ACN (20%–40%) and DI water with 0.1% trifluoroacetic acid over 40 min. Peptide concentration was determined using the peak areas at UV 214 nm and a calibration curve prepared using standards of known concentration (1–0.1 mg/mL INSB). INSB concentrations below the limit of quantitation (0.1 mg/mL) were considered to be zero.

Circular dichroism spectroscopy

Changes in the secondary structure of fibrillating INSB were monitored by circular dichroism (CD) spectroscopy using a Jasco J-815 spectrometer (JASCO Analytical Instruments, Easton, MD). INSB solution (1 mg/mL, pH 4.5) was allowed to fibrillate under stressed conditions (40°C, continuous shaking at 800 rpm) in low protein binding Eppendorf tubes placed inside a Thermomixer R incubator-shaker. At intervals over the fibrillation time course (0–1440 min), samples were withdrawn and diluted four-fold with DI water. CD spectra were recorded in a quartz cuvette with a 1-mm pathlength and scanned in the far-UV region from 190 to 260 nm at a scan speed of 50 nm/min at 20°C. Spectra were reported as an average of three scans, and the CD signals were plotted as mean residue ellipticity $\left[\theta \right]$. Conformational changes in fibrillating INSB were monitored by plotting the CD minima at 218 nm $\left[\theta \right]$ over the fibrillation time course (to 1440 min).

Amide HDX-MS

Pulsed HDX-MS of INSB during fibrillation

INSB fibrillation was monitored by measuring the decrease in deuterium incorporation during fibril formation (31,32). 1.5 mL of freshly prepared INSB solution (1 mg/mL, pH 4.5) was transferred to a 2-mL low protein binding Eppendorf tube and allowed to fibrillate under stressed conditions (40°C, continuous orbital shaking at 800 rpm) using a Thermomixer R incubator-shaker. The onset and progression of INSB fibrillation differed considerably among solutions incubated in different Eppendorf tubes, perhaps due to the stochastic nature of the fibrillation process. Thus, fibrillation was followed in a single Eppendorf tube by withdrawing triplicate samples for HDX-MS analysis at intervals over the fibrillation time course (0–1440 min). The time t = 0 min sample was collected from INSB solution at room temperature (25°C) before transferring to the incubator-shaker, from which subsequent samples (2–1440 min) were collected under stressed conditions (40°C, continuous shaking). INSB solution took <5 min to equilibrate to 40°C on incubation.

A 5-min pulse deuterium-labeling period was initiated by diluting a 10-μL sample of fibrillating INSB in 90 μL of D₂O (99.9%) at room temperature, resulting in a final concentration of ∼90% D₂O (v/v) and ∼0.1 mg/mL INSB (pH_read 4.7). Preliminary studies indicated that the 5-min pulse deuterium-labeling period was sufficient to ensure maximal deuteration of INSB, while taking up a small fraction of the longer fibrillation time course (to 24 h) (31,32). A 10-fold dilution with D₂O enhances deuterium uptake by the backbone amide groups and is assumed to quench fibrillation during the pulse-labeling period. The rate of amide HD exchange is strongly dependent on pH and temperature (40,41,42). Generally, the intrinsic exchange rate increases ∼ten-fold for every pH unit above pH 4.0 and ∼ three-fold for every 10°C increase in temperature (40,41,42). Here, fibrillating INSB was pulse labeled using unbuffered D₂O at pH_read 4.7 so that labeling was conducted at the pH of fibrillation. Though exchange rates are ∼1000-fold slower at pH 4.7 than at neutral pH, the time course of labeling (∼5 min) was still shorter than the time course of fibrillation (∼24 h), allowing use of the pulse-labeling approach. At the end of the labeling period, the exchange reaction itself was quenched by adding 200 μL of quench buffer (3 M urea, 5% v/v methanol, 0.2% v/v FA, pH 2.7) followed by immersion in liquid nitrogen. HD exchange rates are at a minimum at quench conditions of low pH (2.0–3.0) and temperature (0°C–4°C), which helps to minimize the reverse reaction (“back exchange”) so that the deuterium label is retained for analysis (40,41,42). The deuterated samples were stored at $-$80°C until analysis.

HDX-MS instrumentation

For pulsed HDX-MS studies, deuterium uptake was measured using a liquid chromatography mass spectrometry (LC-MS) system (Agilent 6520 QTOF, Agilent Technologies, Santa Clara, CA), equipped with a custom-built column-switching and refrigeration unit to maintain the low temperatures (∼2°C) necessary to minimize back exchange and to desalt the injected sample before MS analysis. Deuterated samples were quickly thawed, and ∼200 pmol INSB was injected into the LC-MS system. The sample was retained on a peptide microtrap (Michrom Bioresources, Auburn, CA) and desalted with 0.1% FA in water for 1.7 min at a flow rate of 0.2 mL/min. The peptide was then eluted on a C18 analytical column (Zorbax 300SB-C18; Agilent Technologies, Santa Clara, CA) using a gradient of ACN (25–59%), water, and 0.1% FA at a flow rate of 50 μL/min over 4.5 min, for a total run time of 11.5 min. Mass spectra were collected over the m/z range of 100–1700. For analysis of deuterium uptake at the fragment level, proteolytic digestion was carried out by injecting ∼200 pmole of peptide into an inline immobilized pepsin column housed in a heating oven placed inside the refrigerated unit and maintained at 25°C. Inline digestion was carried out for 2 min in MS-grade water containing 0.1% FA, at a flow rate of 0.2 mL/min. The peptic fragments were retained on the microtrap, desalted, and eluted on a C18 analytical column as described above.

HDX-MS data analysis

All samples were analyzed using MassHunter Workstation software equipped with the BioConfirm package (Agilent Technologies, version B.04.00). The number of deuterons incorporated and percent deuteration in peptides showing unimodal or bimodal distributions were calculated by the HDExaminer software (version 3.3.0, Sierra Analytics, Modesto, CA). For unimodal distributions, deuterium uptake was calculated from the difference between the centroids of the deuterated peptide and the undeuterated control. For bimodal distributions, deuterium uptake in each population was calculated using their respective centroids. In cases where bimodal fit was not satisfactory, unimodal analysis was applied, and the number of deuterons incorporated was calculated as a weighted average of the deuterium uptake in the protected and more accessible populations. Population fractions were determined by the HDExaminer software (v 3.3.0) from the peak intensities, expressed as a fraction of the total (1 or 100%). For HDX-MS analysis at the fragment level, a list of INSB fragments was generated from the undeuterated control using the BioConfirm software package and mapped onto subsequent deuteration experiments using the HDExaminer software. Deuterium uptake values were not subjected to back exchange correction as the time between thawing and analysis was similar for all samples. A 5-min pulse deuterium-labeling period resulted in ∼57% deuterium uptake in monomeric INSB at acidic pH (see Figs. 3 A and S2), suggesting an upper limit for back exchange of ∼43% in this system, in agreement with previous HDX-MS studies in our lab of similar length peptides at acidic pH (32). All values are reported as the mean of three independent HDX experiments. Additional details regarding the HDX-MS experiments are provided in supporting material using the format suggested by Masson et al. (43).

Figure 3.

Pulsed HDX-MS of fibrillating human insulin B-chain (INSB) at intact level. Fibrillating INSB was pulse labeled with deuterium for 5 min at intervals over the fibrillation time course (to 1440 min) and analyzed for deuterium uptake by LC-MS. (A) Mass envelope of deuterated and undeuterated INSB (m/z = 687.022; z = +5) at the intact level showing a bimodal peak distribution (peaks I and II) during fibrillation. %D is the average percent deuterium uptake by INSB relative to the theoretical maximum (n = 3). (B) Kinetics of changing relative intensities of peaks I and II (“population fraction”), determined using bimodal analysis of the HDX-MS data of fibrillating INSB at intact level, fitted to a biexponential decay model (Eq. 1). Insert shows the kinetics for the initial 0–60 min of fibrillation time, fitted to a monoexponential decay model (Eq. 3), n = 3, mean ± SD. (C) Deuterium uptake kinetics of fibrillating INSB (to 1440 min), determined by unimodal analysis of the HDX-MS data at intact level, fitted to a biexponential decay model (Eq. 1). Insert shows the kinetics for the initial 0–60 min of fibrillation time, fitted to a monoexponential decay model (Eq. 3), n = 3, mean ± SD. Error bars not shown when smaller than the symbol.

Fibrillation kinetic data by HDX-MS at the intact and fragment levels were fitted to a biexponential decay model (Eq. 1) representing two pools of amide hydrogens:

$$D\left(t\right)=D_{\infty }+D_{fast}\left(e^{-k_{fast}t}\right)+D_{slow}\left(e^{-k_{slow}t}\right)\text{,}$$ (1)

where $D\left(t\right)$ is the number of deuterons taken up at time $t$ of fibrillation using a 5-min pulse deuterium-labeling period. The preexponential terms, $D_{fast}$ and $D_{slow}$, represent the number of amide groups that show “fast” and “slow” decreases in deuterium incorporation, with corresponding rate constants, $k_{fast}$ and $k_{slow}$. $D_{\infty }$ is the measured deuterium uptake at long fibrillation times $\left(t=\infty \right)$, and the initial deuterium uptake is given by $D\left(t=0\right)=D_{\infty }+D_{fast}+D_{slow}$. Half-life (t₅₀) values for the “fast” and “slow” phases were calculated from the regression parameters as, $(t_{50}=\ln 2/k)$.

Kinetic data for the initial 0–60 min of INSB fibrillation measured by HDX-MS at the intact level, ThioT fluorescence, turbidity, soluble peptide fraction, and CD spectroscopy were fitted to monoexponential association (Eq. 2) or decay (Eq. 3) models:

$$\text{Y}\left(t\right)=Y_0+\left(Y_{\infty }-Y_0\right)\left(1-e^{-kx}\right)\text{and}$$ (2)

$$\text{Y}\left(t\right)=\left(Y_0-Y_{\infty }\right)e^{-kx}+Y_{\infty }\text{,}$$ (3)

where $k$ is the first-order rate constant, and $Y_0$ and $Y_{\infty }$ are signal intensities at initial $(x=0)$ and long $(x=\infty )$ fibrillation times, respectively. Half-life (t₅₀) values for the monoexponential fits were calculated from the regression parameters as above. One-way ANOVA followed by Tukey’s test was used for determination of statistical significance between multiple comparisons of regression parameters. Graphing, regression analysis to calculate kinetic parameters, and statistical analysis were performed using GraphPad Prism version 9.1 (GraphPad Software, La Jolla, CA).

Results

Fibrillation kinetics using ThioT fluorescence, turbidity, and soluble fraction measurements

ThioT is a dye commonly used to detect amyloid fibril formation due to its increased fluorescence on binding to the hydrophobic pockets of β-sheet fibrils (44,45). ThioT binds poorly to early prefibrillar oligomers with low fluorescence emission but binds strongly to mature β-sheet fibrils with high fluorescence emission (44). Fibrillation often also results in an increase in solution turbidity due to the formation of higher order oligomers or aggregates capable of scattering incident light (46). For INSB solution stressed inside the Thermomixer R incubator-shaker (40°C, continuous orbital shaking at 800 rpm), there was a slight increase in both ThioT fluorescence and turbidity signal intensities for the initial 0–60 min of fibrillation time, followed by larger increases from 60–1440 min (Fig. 1 A and B). The overall change in solution turbidity was small for the entire fibrillation time course (0–1440 min) (Fig. 1 B). In contrast to the slight increases in ThioT fluorescence and turbidity signals (Fig. 1 A and B), there was a rapid decrease in soluble INSB concentration from ∼1 mg/mL to ∼0.2 mg/mL in the initial 0–60 min of fibrillation time (Fig. 1 C), suggesting the formation of higher order oligomers that showed poor ThioT binding and little light scattering. Beyond 60 min, the soluble INSB concentration was below the 0.1 mg/mL threshold of the assay (Fig. 1 C), consistent with depletion of INSB monomers and the formation of insoluble species. Moreover, the corresponding gradual increase in ThioT intensity from 60–1440 min suggests the formation of fibrils that bind strongly to ThioT (Fig. 1 A).

Figure 1.

Fibrillation of human insulin B-chain (INSB) (1 mg/mL, pH 4.5, 40°C, orbital shaking at 800 rpm using a Thermomixer R incubator-shaker) by ThioT fluorescence (n = 3) (A), turbidity (n = 3) (B), and soluble fraction (n = 2) (C) measurements over the fibrillation time course (to 1440 min). Lines show trends in data and do not represent regression. Insert shows the initial 0–60 min of fibrillation time, fitted to monoexponential association (Eq. 2) or decay (Eq. 3) models. Mean ± SD. Error bars not shown when smaller than the symbol.

For INSB solutions stressed inside the microplate reader (40°C, continuous orbital shaking at 807 cpm), ThioT fluorescence measurements showed sigmoidal behavior with lag, growth, and plateau phases characteristic of nucleation-dependent aggregation (Fig. S1 A) (47). INSB solution (1 mg/mL, pH 4.5) was subjected to similar stress conditions in the Thermomixer R incubator-shaker and in the microplate reader. The increase in ThioT fluorescence intensity during the late lag phase (∼120–240 min) in the microplate reader corresponded to the increase in ThioT fluorescence intensity in the Thermomixer R for the same duration (Figs. 1 A and S1 A). Like the Thermomixer R studies (Fig. 1 B), the microplate reader studies did not show significant increases in solution turbidity in INSB samples compared with buffer controls (Fig. S1 B). On visual inspection, translucent gel-like deposits were observed at the bottom of the 96-well plates and inside the Eppendorf tubes, which may be responsible for the low turbidity (Figs. 1 B and S1 B), perhaps due to limited light scattering by these translucent deposits that are below the light path (46). Although fibrillation conditions were similar for Thermomixer R and microplate reader studies, fibrillation kinetics can be influenced by factors such as mixing dynamics (1.5 mL solution in Eppendorf tubes versus 100 μL solution in microplate wells) (48), type of polymer surface (polypropylene Eppendorf tubes versus polystyrene microplates) (48,49), and perturbation of the fibrillation medium on sampling from Eppendorf tubes.

Conformational changes by CD spectroscopy

The changes in the secondary structure of INSB during fibrillation, as detected by CD spectroscopy (Fig. 2), coincided with the changes observed by ThioT and soluble fraction measurements (Fig. 1 A and C). At the start of the fibrillation experiment (t = 0 min), the CD spectra of INSB showed a negative minimum at 204 nm and a weak positive signal at 190 nm (Fig. 2 A), indicative of a predominantly unstructured peptide with some secondary structure of unknown type (50). Under stressed conditions (40°C, continuous orbital shaking at 800 rpm, Thermomixer R), the spectra showed rapid conversion to β-sheet features, as evidenced by a broad negative minimum around 218 nm and a strong positive maximum below 195 nm (Fig. 2 A). There was a rapid increase in the negative CD signal intensity at 218 nm during the initial 0–60 min of fibrillation (Fig. 2 B), corresponding to the rapid decrease in soluble INSB concentration and slight increase in ThioT fluorescence intensity over the same period (Fig. 1 A and C), suggesting the formation of oligomers with β-sheet-like secondary structure that do not yet bind strongly to ThioT. From 60 to 1440 min of fibrillation time, the negative CD signal intensity at 218 nm decreased gradually with near-complete loss in signal intensity at 1440 min (Fig. 2), consistent with the loss in soluble INSB concentration and formation of insoluble fibrils (Fig. 1 C).

Figure 2.

Circular dichroism (CD) spectra of fibrillating human insulin B-chain (INSB). (A) Overlay of CD spectra at different times during fibrillation (0, 2, 5, 10, 15, 30, 45, 60, 90, 120, 180, 240, 300, 360, 480, 720, 1440 min). (B) Changes in CD signal at 218 nm over the fibrillation time course. Lines show trends in data and do not represent regression. Insert shows CD signal at 218 nm for the initial 0–60 min of fibrillation time, fitted to a monoexponential association model (Eq. 2) (n = 1). To see this figure in color, go online.

Amide HDX-MS

HDX-MS for fibrillating INSB at intact level

At the start of the fibrillation experiment (t = 0 min), deuterated INSB (1 mg/mL, 25°C) showed a bimodal deuterated mass envelope with two distinct peaks: peak I had higher intensity and appeared at higher m/z, whereas peak II had lower intensity and appeared at lower m/z (Fig. 3 A). The two INSB populations (peaks I and II) showed differences in their deuteration states (Fig. 3 A). Peak I corresponds to INSB monomers that are more solvent exposed and therefore more deuterated (31,36), an assignment supported by the spectra for deuterated INSB in DI water (Fig. S2). Peak II is assigned to INSB oligomers or protofibrils that are protected from deuterium incorporation by intermolecular association (31,36). The presence of the protected population (peak II) at time t = 0 min is consistent with low levels of oligomeric species in INSB solution at the start of fibrillation study (Fig. 3 A). After 2 min under stressed conditions (40°C, continuous orbital shaking at 800 rpm), there was a marked increase in the intensity of the protected population (peak II) (Fig. 3 A). During the initial 0–30 min of fibrillation time, the intensity of peak II increased rapidly with a corresponding decrease in the intensity of peak I and a rapid decrease in the percent deuterium incorporation from 56% to 35% (Fig. 3 A). At 240 min, peak I had diminished to a shoulder suggesting a near-complete shift from monomeric to oligomeric INSB (Fig. 3 A). At times greater than 240 min, the isotopic distribution was mostly unimodal with only peak II appearing (Fig. 3 A). The percent deuterium incorporation decreased gradually from 35% to 23% between 30 and 1440 min, in contrast to its earlier more rapid decrease (Fig. 3 A). After 1440 min, the percent deuterium uptake was 23%, suggesting that some solvent-accessible regions remain as the INSB fibrils mature (Fig. 3 A).

Deuterium incorporation was analyzed quantitatively in two ways. In the first approach, bimodal analysis was used in which the two peaks in the bimodal mass envelopes were treated separately, and deuterium incorporation for each population was determined by deconvolution (Figs. S3 and 3 B). With this approach, although the peak intensities changed over time (Fig. 3 A and B), the percent deuterium incorporation in the solvent-accessible (peak I) and protected (peak II) populations remained relatively constant at 64% (#D ∼ 18) and 29% (#D ∼ 8), respectively (Fig. S3). At the intact level, the kinetics of INSB fibrillation by HDX-MS showed biphasic behavior and was fitted to a biexponential decay model (Eq. 1) consisting of “fast” and “slow” phases (Fig. 3 B and C). The half-lives for the growth of peak II and for the decrease in peak I were identical (t_{50, fast} = 2.4 ± 0.2 min, t_{50, slow} = 121 ± 13 min) (Table S1), consistent with conversion of the solvent-exposed population to the protected population. In the second approach, unimodal analysis was used in which the overall loss in deuteration was determined as the weighted average of the number of deuterons in both the solvent-accessible and protected populations (Fig. 3 C). The resulting %D values decrease over time, reflecting the shift to the protected population, but they do not correspond to the extent of deuteration in either population individually (Fig. 3 C). With unimodal analysis, the half-life of the “fast” phase (t_{50, fast} = 3.1 ± 0.3 min) was ∼50$\times$ shorter than that of the “slow” phase (t_{50, slow} = 161 ± 19 min) and comparable to the half-lives determined using bimodal analysis (Table S1). The decrease in deuterium uptake during the “fast” phase (D_fast ≈ 18%) was only slightly greater than that during the “slow” phase (D_slow ≈ 14%) (Table S2), indicating the involvement of comparable numbers of amide groups during the two phases of fibrillation. In the text below and in Fig. 3 A, the %D values are those determined by unimodal analysis.

INSB showed differences in fibrillation behavior during early and late fibrillation times. At early fibrillation times (0–60 min), the decrease in deuterium uptake (56%–35%, Fig. 3 A and C), corresponded to a decrease in soluble INSB concentration (∼1–0.2 mg/mL) and an increase in CD signal intensity at 218 nm (Figs. 1, 2, and S4). In contrast, there was only a small increase in ThioT fluorescence intensity and a negligible increase in solution turbidity from 0–60 min (Figs. 1 and S4). These findings suggest that the emergence of HDX-protected species corresponds in time to the formation of INSB oligomers with β-sheet-like secondary structure that show weak ThioT binding and little increase in light scattering. The fibrillation half-life (t₅₀) of INSB for the early period (0–60 min) was similar for HDX-MS and CD measurements, slightly longer for the ThioT and soluble fraction measurements, and longest for turbidity measurements (Fig. 4; Tables S3 and S4). Except for turbidity and HDX-MS t₅₀ values, there were no significant differences (p > 0.05) in the t₅₀ values determined by other methods (Fig. 4, Table S4). Rate constant (k) values obtained from different analytical methods showed a similar trend (Fig. S5, Table S4). These findings suggest that the structural changes detected by HDX-MS coincide with the changes detected by CD spectroscopy and occur somewhat sooner than changes detected by the other methods, although relatively weak signals for ThioT and turbidity measurements in the first 60 min suggest that t₅₀ and k values for these measurements should be interpreted cautiously (Figs. 1, 4, S4, and S5). INSB solution in Eppendorf tubes took <5 min to equilibrate to 40°C on incubation. Although the time to temperature equilibration may have some effect on early fibrillation kinetics (e.g., at 2 min), the overall effect is probably low considering that the “fast” phase of fibrillation extends up to 30 min (Fig. S4). At later fibrillation times (60–1440 min), INSB showed a gradual decrease in deuterium uptake (∼35%–23%) (Fig. 3 C), which corresponded to near-complete loss in soluble monomer concentration (<0.1 mg/mL) (Fig. 1 C) and reduction in the CD signal intensity (Fig. 2). Interestingly, additional protection from HD exchange during this time interval (60–1440 min) corresponded to a larger increase in ThioT fluorescence intensity (Figs. 1 A and 3 C), suggesting that structural changes in INSB fibrils at longer fibrillation times were accompanied by greater binding to ThioT.

Figure 4.

Half-life (t₅₀) of INSB fibrillation for the initial 0–60 min measured by HDX-MS at intact level (n = 3), ThioT fluorescence (n = 3), turbidity (n = 3), soluble fraction (SF, n = 2), and CD spectroscopy (n = 1). Values were obtained from monoexponential fits (Eqs. 2 and 3) to kinetic data for each method. Mean ± SE of the regression. One-way ANOVA with Tukey’s test was used for multiple comparisons; see Table S4 for p-values.

HDX-MS for fibrillating INSB at fragment level

Fibrillating INSB was subjected to HDX-MS with pepsin digestion to identify the residues involved in fibrillation throughout the time course. Deuterium uptake was monitored in five INSB fragments (1–11, 1–15, 12–15, 16–25, and 25–30) that showed strong and reproducible MS signals and provided 100% sequence coverage with some overlap (Fig. 5). Overlapping fragments 1–11 and 1–15, from the N-terminus (Fig. 5 A and B), and central fragments 12–15 and 16–25 (Fig. 5 C and D) all showed rapid decreases in deuterium uptake from 0–30 min, followed by relatively slow decreases in deuterium uptake from 30–1440 min. In contrast, the C-terminal fragment 25–30 showed little change in deuterium uptake over the fibrillation time course (to 1440 min) (Fig. 5 E). Similar to observations at the intact level (Fig. 3 A), the deuterated mass envelopes of INSB fragments 1–15, 12–15, and 16–25 showed bimodal isotopic distributions at time t = 0 min, consistent with the presence of some exchange protected oligomeric species at the start of the fibrillation study (Fig. S6). Under stressed conditions (40°C, continuous orbital shaking at 800 rpm), the deuterated mass envelopes of N-terminal (1–11, 1–15) and central (12–15, 16–25) fragments showed rapid shifts toward a more protected population from 0–30 min, followed by a more gradual shift from 30–1440 min (Fig. S6). In contrast, the C-terminal fragment (25–30) only showed minor peak broadening at 1440 min (Fig. S6). Fragment 1–15, which contains overlapping residues from fragments 1–11 and 12–15, showed intermediate behavior (Figs. 5 B and S6). At 1440 min, the mass envelope of central fragment 12–15 showed a near-complete shift to a more protected population with low deuterium uptake (∼6%) (Figs. 5 C and S6), whereas the mass envelopes of fragments 1–11 and 16–25 showed somewhat greater deuterium uptake at this time (∼14% and ∼20%, respectively) (Figs. 5 A, D, and S6). These findings suggest that the central region (12–15) is protected from exchange by the fibril structure, whereas neighboring regions (1–11 and 16–25) are somewhat more solvent exposed. In contrast, the C-terminal fragment 25–30 showed little decrease in deuterium uptake and little change in the deuterated mass envelope even after 1440 min (Figs. 5 E and S6), suggesting that this region does not participate in fibril formation under these conditions.

Figure 5.

Pulsed HDX-MS of fibrillating INSB at fragment level. Fibrillating INSB was pulse labeled with deuterium for 5 min at intervals over the fibrillation time course (to 1440 min) and analyzed for deuterium uptake by MS after pepsin digestion. Percent deuterium uptake for INSB fragments 1–11 (A), 1–15 (B), 12–15 (C), 16–25 (D), and 25–30 (E) is plotted as a function of time of fibrillation and determined by unimodal analysis of the HDX-MS data. Inserts show deuterium uptake for the initial 0–60 min of fibrillation time. Data were fitted to a biexponential decay model (Eq. 1) in cases where good fit was achieved. n = 3, mean ± SD. Error bars not shown when smaller than the symbol.

Although a bimodal peak distribution was observed for some INSB fragments (1–15, 12–15, 16–25) at early times (Fig. S6), a good bimodal fit could not be achieved probably due to poor resolution of the two populations. Thus, unimodal analysis was applied, and deuteration kinetic data for fibrillating INSB fragments was fitted to a biexponential decay model (Eq. 1), consisting of “fast” and “slow” phases (Fig. 5). This analysis was not applied to the C-terminal fragment 25–30, which showed little overall change in deuterium uptake (Fig. 5 E). For the N-terminal (1–11, 1–15) and central (12–15, 16–25) fragments, the deuteration half-life for the “fast” phase (t_{50, fast}) was significantly shorter than for the “slow” phase (Fig. 6 A and B; Table S1), consistent with the rapid decrease in deuterium uptake for these fragments in <15 min of fibrillation time (Fig. 5 A–D). Also, these fragments showed a greater decrease in deuterium uptake and greater protection from exchange during the “fast” phase than the “slow” phase (D_fast > D_slow) (Fig. 6 C and D; Table S2). The “slow” phase was characterized by a uniform decrease in deuterium uptake, with similar rates among the fragments (Figs. 6 D and S7 B). These findings suggest that the N-terminal (1–11, 1–15) and central (12–15, 16–25) fragments participate in early interactions leading to fibrillation and that these interactions gradually increase at later times. Central fragment 12–15 had the shortest t₅₀ values and highest rate constant (k) values for both the “fast” and “slow” phases and showed the greatest loss in deuteration during the “fast” phase (D_fast), as compared with fragments 1–11 and 16–25 (Figs. 6 and S7). Although the observed differences in the rates and extents of deuteration in INSB fragments in the “fast” phase are statistically significant (Table S5), the comparison may be confounded by the inherent limitations associated with capturing rapid fibrillation kinetics and measuring deuterium uptake in fragments of short length (e.g., fragment 12–15).

Figure 6.

Pulsed HDX-MS kinetic parameters, t_{50, fast} (A), t_{50, slow} (B), D_fast (C), and D_slow (D), for INSB fragments during fibrillation, obtained from biexponential fits (Eq. 1) to HDX-MS data at the fragment level. n = 3, mean ± SE of the regression. One-way ANOVA with Tukey’s test was used for multiple comparisons; see Table S5 for p-values.

Discussion

Pulsed HDX-MS, far-UV CD spectroscopy, ThioT fluorescence, turbidity, and soluble fraction measurements were used to monitor the kinetics of human insulin B-chain (INSB) fibrillation under stressed conditions (40°C, continuous shaking). At early fibrillation times (0–60 min), INSB rapidly formed oligomeric species (Fig. 1 C) with β-sheet-like secondary structure (Fig. 2) that was protected from HD exchange (Fig. 3) and showed weak ThioT binding (Fig. 1 A). At later fibrillation times (60–1440 min), insoluble species were formed (Fig. 1 C) that showed even greater protection from HD exchange (Fig. 3) and stronger ThioT binding (Fig. 1 A). Pulsed HDX-MS of fibrillating INSB at the intact level showed biphasic kinetics, with rapid growth of a single protected population at early fibrillation times (0–30 min) followed by a more gradual increase in protection at later times (30–1440 min) (Fig. 3 A). The “fast” and “slow” phases involved similar numbers of amide groups (D_fast ≈ 18%, D_slow ≈ 14%) (Fig. 3, Table S2). At the end of the fibrillation time course, deuterium uptake of ∼23% remained, suggesting that some amide groups remain solvent exposed as the INSB fibrils mature (Fig. 3). Previous studies by Hong et al. showed that INSB at high concentrations in acidic solution (>0.2 mg/mL, pH 1.6, 37°C, continuous shaking) was present as β-sheet oligomers that preferentially formed stable protofilaments (51). The protofilaments showed low fluorescence intensity on ThioT binding and did not progress to mature fibrils when incubated for up to a week (51). The studies reported here were conducted at higher concentration and pH (1 mg/mL, pH 4.5) and suggest that progression of protofilaments to mature INSB fibrils does occur under these conditions (40°C, continuous shaking) (Figs. 1 and S1).

Proteolytic digestion of deuterated samples showed comparable rates of protection for N-terminal (1–11, 1–15) and central (12–15, 16–25) INSB fragments throughout the fibrillation time course. However, loss of deuterium incorporation was somewhat faster and more complete for the central (12–15) fragment than for other regions (Figs. 5, 6 A, and B; Fig. S7). In contrast, the C-terminal fragment (25–30) showed little to no change in deuterium incorporation during fibrillation (Fig. 5 E), indicating that the C-terminal region is not incorporated in the hydrogen-bonded β-sheet structure of INSB fibrils. Together, these findings are consistent with the rapid initial formation of INSB oligomers through interaction of central and N-terminal regions, followed by more gradual involvement of additional amide protons in these regions during fibril growth and maturation, perhaps through structural rearrangement. In a previous report, substitution of positively charged arginine residues at the C-terminus of the B-chain produced variants that resisted fibrillation in both native insulin and INSB alone, suggesting the involvement of INSB C-terminal residues in fibrillation (14). Our findings are not in agreement with this result. The previous study was conducted at different solution composition (2 mg/mL peptide in 20% acetic acid, 150 mM NaCl, pH 2.0) and stress conditions (60°C, quiescent) (14) than the studies conducted here, which may account for the apparent differences in fibrillation kinetics and the involvement of the C-terminus.

Several previous reports have identified the central B-chain sequence L₁₁VEALYL₁₇ as the primary contributor to cross-β spine formation in insulin fibrils (13,17,52). The central fragment 12–15 studied here is completely contained within this amyloidogenic sequence. As noted above, the rate of loss of deuteration in pulsed HDX-MS studies was somewhat greater for fragment 12–15 than for other fragments (Fig. 6 A and B; Fig. S7). In addition, at the end of the fibrillation time course (1440 min), fragment 12–15 showed less deuteration (6%) than neighboring fragments (1–11, 14%; 16–25, 20%) (Figs. 5 and S6), consistent with greater incorporation into the fibril core. Our results thus support the importance of the central region in INSB fibrillation, though similar rates and extents of deuterium loss in neighboring regions (Figs. 5, 6, and S7) suggest that they also contribute, and differences among the N-terminal and central fragments are not large.

In pulsed HDX-MS studies, fibrillating INSB showed a bimodal distribution of deuterated mass envelopes, with two distinct and well-resolved populations (peaks I and II) at the intact level (Fig. 3 A) and a poorly resolved peak broadening at the fragment level (Fig. S6). At the intact level, fibrillation was monitored from the growth of the exchange protected population (peak II) or the decay of the solvent-accessible population (peak I) (Fig. 3 B). Bimodal fit was used to determine the deuterium incorporation in the solvent-accessible (peak I) and protected (peak II) populations individually from their respective centroids using HDExaminer software (v 3.3.0); however, deuterium incorporation in these populations remained relatively constant over the fibrillation time course (Fig. S3). Although deuteration levels in peaks I and II provide information on the extent of protection in the amide protons involved (Fig. S3), they do not provide information on the rate at which protection occurs as fibrils grow. Several approaches have been developed to fit bi- or multimodal isotopic distributions commonly observed with multiple coexisting protein conformations or EX1 exchange mechanisms (34,53,54). In the present study, INSB fibrillation was monitored by measuring the overall loss in deuteration in both the protected and solvent-accessible populations weighted against their relative abundances (Fig. 3 C). Interestingly, the growth kinetics of the protected population (peak II) (Fig. 3 B) tracked closely with the weighted deuteration kinetics of fibrillating INSB at the intact level (Fig. 3 C), as seen from their comparable half-lives (t₅₀) (Table S1), highlighting the contributions of relative population abundances to these measurements. This approach may be beneficial particularly in cases where good bimodal fits cannot be achieved due to poor resolution of isotopic populations, as observed for INSB fragments in the present study (Fig. S6), or when more than two populations coexist, as observed for fibrillating Aβ and disulfide-reduced human calcitonin peptides in previous studies (31,53).

Here and in previous work, we have studied the fibrillation of several peptides including human insulin B-chain (INSB), glucagon, human calcitonin (hCT), and their structural analogs (31,32). Although these peptides were of similar lengths (29–32 residues) and, in some cases, had similar sequences, the kinetics and mechanisms of fibrillation differed and were influenced by differences in primary sequence, formulation, and stress conditions employed, as monitored by pulsed HDX-MS. For example, under identical conditions (100 μM peptide, pH 7.4, 25°C, continuous shaking), fibrillating native hCT showed sigmoidal deuteration kinetics corresponding to lag, growth, and plateau phases, whereas reduced hCT (i.e., reduction of Cys1-Cys7) showed faster fibrillation and monoexponential loss of deuteration (31). Similarly, fibrillating glucagon (0.6 mg/mL, pH 2.5, 23°C, quiescent) showed monoexponential deuteration kinetics, whereas fibrillating INSB (1 mg/mL, pH 4.5, 40°C, continuous shaking) showed biexponential loss of deuterium uptake, as noted above. Although sigmoidal fibrillation kinetics is typical of a nucleation-dependent aggregation process (47), the mono- and biexponential kinetics suggest different underlying mechanisms or rate determining steps.

In some instances, fibrillation proceeded sequentially with early involvement of specific regions. For example, the C-terminal fragment 22–29 of glucagon showed a rapid initial decrease in deuterium incorporation, during which time the N-terminal fragment 1–6 showed no changes in deuteration, implicating C-terminal interactions in the early stages of glucagon fibrillation (32). For fibrillating hCT, the N-terminal (1–11) and central (12–19) fragments were involved in early interactions, whereas the C-terminal fragments (20–32, 26–32) showed limited involvement (31). Interestingly, the hCT C-terminal residues (20–32) showed greater protection from HD exchange and greater involvement in mature fibrils in disulfide-reduced hCT than in native hCT (31), suggesting structural differences between the two types of fibrils. In the studies reported here, the central (12–15, 16–25) and N-terminal (1–11, 1–15) regions of INSB were involved in the early stages of INSB fibrillation, though rapid fibrillation kinetics and comparable rates of decrease in deuteration in these fragments made it difficult to better resolve differences among them (Figs. 5 A–D, 6, S7). Interestingly, proteolytic fragments overlapping with known amyloidogenic sequences in INSB (fragment 12–15; sequence L₁₁VEALYL₁₇) and hCT (fragment 12–19; sequence D₁₅FNKF₁₉) showed early loss of deuteration and greater protection from exchange in mature fibrils (31), underscoring the importance of these core segments in the fibrillation of these peptides. Residue-level information in pulsed HDX-MS studies of hCT and glucagon enabled us to develop fibrillation-resistant analogs using site-specific phosphorylation of serine or threonine residues, which interfered with fibril-forming interactions through charge repulsion and/or steric effects (31,32,55). Thus, pulsed HDX-MS studies can provide mechanistic understanding of the intermolecular interactions during fibrillation, enable identification of the residues involved, and guide the rational development of fibrillation-resistant therapeutic peptides.

Conclusions

Under stressed conditions (40°C, continuous shaking), INSB (1 mg/mL, pH 4.5) rapidly formed β-sheet rich oligomers that were protected from HD exchange but showed weak ThioT binding. Subsequent fibril growth and maturation were accompanied by greater protection from HD exchange and stronger ThioT binding. Pulsed HDX-MS with proteolytic digestion showed that the N-terminal (1–11) and central (12–15, 16–25) regions were involved in early interactions and that the C-terminus (25–30) does not participate in the formation of INSB fibrils under these conditions. Since INSB is known to be important in the fibrillation of native insulin, the results suggest that modifications to the N-terminal or central fragments of INSB may be effective in inhibiting insulin fibrillation.

Author contributions

H.K.R performed the studies, analyzed data, and drafted the manuscript. E.M.T. supervised the project, designing the studies and analyzing data in collaboration with H.K.R., and oversaw data analysis and manuscript preparation.

Editor: Tuomas Knowles.

Supporting citations

Reference (56) appears in the supporting material.

Supporting material

Document S3. Article plus supporting

material