Sulfate Anion Delays the Self-Assembly of Human Insulin by Modifying the Aggregation Pathway

Abstract

The understanding of the molecular mechanisms underlying protein self-assembly and of their dependence on solvent composition has implications in a large number of biological and biotechnological systems. In this work, we characterize the aggregation process of human insulin at acidic pH in the presence of sulfate ions using a combination of Thioflavin T fluorescence, dynamic light scattering, size exclusion chromatography, Fourier transform infrared spectroscopy, and transmission electron microscopy. It is found that the increase of sulfate concentration inhibits the conversion of insulin molecules into aggregates by modifying the aggregation pathway. At low sulfate concentrations (0–5 mM) insulin forms amyloid fibrils following the nucleated polymerization mechanism commonly observed under acidic conditions in the presence of monovalent anions. When the sulfate concentration is increased above 5 mM, the sulfate anion induces the salting-out of ∼18–20% of insulin molecules into reversible amorphous aggregates, which retain a large content of α-helix structures. During time these aggregates undergo structure rearrangements into β-sheet structures, which are able to recruit monomers and bind to the Thioflavin T dye. The alternative aggregation mechanism observed at large sulfate concentrations is characterized by a larger activation energy and leads to more polymorphic structures with respect to the self-assembly in the presence of chloride ions. The system shown in this work represents a case where amorphous aggregates on pathway to the formation of structures with amyloid features could be detected and analyzed.

Introduction

In the last decades, the interest in understanding the molecular mechanisms responsible for protein self-assembling has increased significantly due to the large number of implications in life sciences and technology. Examples range from the aberrant aggregation of peptides and proteins involved in several neurodegenerative diseases (1), to the stability of protein solutions in food and pharmaceutical industry (2–4) as well as the engineering of mechanically and chemically stable biomaterials (5).

A large number of aggregation pathways and aggregate structures has been observed depending on the specific protein and the environmental conditions under consideration: many proteins under suitable conditions are able to self-assemble into regular fibrillar structures known as amyloids (6–10), although in other systems proteins aggregate into amorphous precipitates (11–13). In addition, different aggregation pathways may occur simultaneously, thereby leading to a heterogeneous composition of the final products.

One of the most relevant external factors affecting protein aggregation pathways and aggregate morphologies is the buffer composition. Ions and other cosolutes, such as sugars, polyols, and amino acids, mediate protein intra- and intermolecular interactions via a combination of several effects such as charge screening, ion binding, preferential exclusion, and dipole interactions, therefore affecting the protein secondary and quaternary structure (14–19) as well as the individual microscopic events underlying the aggregation process (20). Peculiar effects such as ion specificity and restabilization behavior at large salt concentrations have been observed in the aggregation of many different proteins, ranging from short peptides to globular multidomain proteins (21–25). In the amyloid field, specific ion effects have been observed for instance in the aggregation of the Aβ peptide (26), α-synuclein (27), islet amyloid polypeptide (28), prion protein (29), and HypF-N (30). The propensity of the different ions to induce protein aggregation is commonly related to either the electroselectivity or the Hofmeister series, a ranking that has been observed in a broad range of phenomena in physics and chemistry (31). However, the exact order of the effectiveness of different ions in promoting aggregation is strongly dependent on the pH value, the salt concentration, and the specific protein under consideration.

In this study, we concentrate our attention on the effect of sulfate anion on the fibrillization process of human insulin under acidic conditions. Sulfate is a kosmotropic anion that stabilizes protein structure and reduces protein solubility in solution (29). In the form of ammonium salt, sulfate is one of the most common anions used for inducing salting out of proteins in crystallography and purification processes (32,33). Although the effect on the precipitation of native proteins is well characterized, the effect of sulfate on nonnative protein aggregation behavior is more complex. On one hand, the salting-out property promotes aggregation due to the decrease of the total amount of exposed surface that accompanies the formation of the aggregates; on the other hand, the kosmotropic nature of the anion stabilizes the protein structure under nonnative conditions, therefore reducing the formation of nonnative aggregation-prone conformations. In addition, the divalency of the anion confers peculiar properties in terms of charge screening, anion binding, and salt bridge formation. The overimposition of the different electrostatic effects can either promote (34–36) or inhibit (37) the protein aggregation propensity. In the case of an IgG2 monoclonal antibody, we have recently reported a maximum of the aggregation propensity as a function of sulfate concentration (22).

The delicate balance between the different sulfate properties modulates not only the protein aggregation propensity, but also the morphology of the final aggregates. When monovalent anions are replaced by sulfate, either longer fibrils or, on the contrary, amorphous aggregates and shorter fibrils are observed (20,36,38). Sulfate can also promote fibril lateral aggregation and increase fibril stability (30). The different effect depends strongly on the anion concentration and the protein primary sequence. For instance, sulfate has been found to promote the fibril formation of the EAK16-II peptide (39), and to disrupt the regular fibrillar structure of the EMK16-II peptide (40).

The understanding of the complex effect of sulfate anion on protein aggregation may represent, to our knowledge, a first step toward the fundamental rationalization of the even more complex interactions between proteins and sulfonate biomacromolecules, such as glycosaminoglycans. These physiologically relevant components are polyelectrolyte macromolecules that are involved in the in vivo aggregation propensity of several amyloidogenic proteins (41–46).

In this work, we analyze the effect of sulfate on the aggregation behavior of human insulin, a hormone produced in pancreatic β cells and responsible for the metabolism of glucose (47). Insulin is one of the therapeutic proteins with the largest production volume due to the use as regulator of the level of sugar in the blood of people affected by diabetes (47). Insulin is well known to form amyloid fibrils characterized by large β-sheet content under destabilizing conditions such as low pH and high temperature (38,47–50). Several biophysical studies investigated the molecular pathway underlying the fibril formation process using monovalent anions, typically chloride (9,20,51–55). When chloride is replaced by sulfate, destabilizing and stabilizing effects have been reported at low and high ionic strength, respectively (56), and a shortening of the fibril length has been observed (38).

Here, we investigate whether the changes in the kinetics and in the aggregate morphology observed when chloride is replaced by sulfate are connected to a change in the aggregation mechanism. In particular, we characterize in detail the aggregation of insulin in the presence of sodium sulfate under conditions where formation of amyloid fibrils is commonly observed, i.e., 25 mM HCl at pH 1.6 and 60°C under quiescent conditions (57). At this pH value positive charges are largely dominating on the protein surface, and therefore the specific effect of monovalent cations can be considered negligible with respect to the anion effect. Indeed, the negligible effect of monovalent cations at low pH has been observed previously with other proteins (22,30). In addition, Hofmeister effects are normally dominated by anions with respect to cations.

The study is performed in parallel with the investigation of the aggregation behavior in the presence of sodium chloride, to allow the comparison with the well-known behavior reported in the literature (9,56,58–61).

We investigated the aggregation kinetics and the aggregate morphology using a combination of Thioflavin T (ThT) fluorescence, dynamic light scattering (DLS), size exclusion chromatography (SEC), Fourier transform infrared spectroscopy (FTIR), and transmission electron microscopy (TEM). We show that at concentrations larger than 5 mM sulfate ions induce an alternative aggregation pathway with respect to chloride ions. This alternative process involves an initial precipitation of amorphous aggregates, which during time undergo structural rearrangements into β-sheet structures, which exhibit features of amyloid fibrils. This study represents a relevant example where amorphous intermediates on-pathway to β-sheet-rich structures could be isolated and characterized.

Materials and Methods

Materials and aggregation conditions

Human insulin was kindly donated by Novo Nordisk (Bagsvaerd, Denmark). Insulin solutions in the concentration range 0.5–5 g/L (86–860 μM) were freshly prepared before each experiment by dissolving insulin powder in 25 mM HCl solution (Fluka Chemika, Buchs, Switzerland) at pH 1.6 with 100 mM sodium chloride (VWR International BVBA, Leuven, Belgium) or 100 mM sodium sulfate (Merck kGaA, Darmstadt, Germany). To remove potential seeds, before the aggregation studies the solutions were filtered by low protein binding hydrophilic LCR, Millex-LG syringe filters with 200 nm cut-off membrane (Merck Millipore, Merck KGaA, Darmstadt, Germany). The protein concentration after filtration was measured by ultraviolet absorbance at 280 nm. Aggregation was induced by incubating the protein solutions in the temperature range from 57 to 63°C.

It is known that the insulin fibrillation process can be affected by the presence of surfaces, in particular by hydrophobic surfaces (63). To probe for the presence of possible artifacts induced by this effect, we monitored the aggregation of insulin solutions incubated in cuvettes or tubes made of different materials (glass, polystyrene, polypropylene, and polymetylmetacrylate) using different solution volumes. No significant difference was observed under the different conditions, indicating that, at least under the investigated conditions, both the material of the container and the total sample volume do not affect the conclusions of this work.

ThT fluorescence assay

ThT (Sigma-Aldrich GmbH, Steinheim, Germany) fluorescence was measured in a 96-well plate (96 Isoplate, Perkin Elmer, Waltham, MA) using an EnSpire 2300 Multilabel Reader fluorometer (Perkin Elmer). 10 μM ThT was added to threefold diluted samples taken at different time points and emission fluorescence values were measured at 485 nm after excitation at 450 nm.

On-line kinetic experiments were performed by incubating the protein solution in the presence of 10 μM ThT and monitoring the fluorescence signal during aggregation. No significant difference was observed between the off-line and the on-line kinetics (data reported in the Supporting Material, Fig. S1).

DLS

DLS measurements were performed on-line using a Zetasizer Nano (Malvern, Worcestershire, United Kingdom), operating in the backscattering mode at a fixed angle of 173° with a laser beam with wavelength of 633 nm.

SEC coupled with multiangle static light scattering (MALS)

SEC analysis was performed with a Superdex Peptide 10/300 GL, 10 mm × 300 mm size-exclusion column (GE Healthcare, Uppsala, Sweden) assembled on a Agilent 1100 series HPLC unit (Santa Clara, CA). Each sample was eluted for 70 min at a constant flow rate of 0.4 mL/min using as mobile phase 10 mM HCl solution at pH 2.0. The pH of the mobile phase was slightly greater than the pH of the analyzed samples (pH = 1.6) to avoid damage of the equipment. Ultraviolet absorbance was detected at 280 nm. To evaluate the amount of residual soluble insulin during the aggregation process, protein samples were centrifuged for 15 min at 10,000 rpm to precipitate the aggregates, and 40 μl aliquots of the supernatant were injected into the column after filtration by low protein binding hydrophilic LCR, Millex-LG syringe filters with 200 nm cut-off membrane (Merck Millipore).

MALS of fractionated samples eluting from the SEC column was measured on-line by a Wyatt light scattering detector (Wyatt, Dernbach, Germany) with a laser beam with a wavelength of 658 nm and scattering angles from 14° to 163°.

FTIR

Hydrated thin film attenuated total reflectance Fourier transform infrared spectroscopy (ATR-FTIR) measurements were performed on a Nicolet Nexus 870 FTIR ESP instrument equipped with a ATR Nicolet Omni-Sampler device (Nicolet, Madison, WI). Aliquots of 10 μL were spotted on the crystal surface and dried under nitrogen flux. The spectra were collected in the wavenumber range from 1700 to 1600 cm⁻¹ at 1 cm⁻¹ resolution and smoothed using the Savitsky-Golay function after buffer and atmospheric subtraction.

TEM

The TEM pictures were taken using a FEI Morgagni 268 microscope. 5 μL of 75-fold diluted samples were spotted on a carbon support films—300 mesh Cu grids (Quantifoil, Jena, Germany) for 30 s, washed with distilled water, and negative stained with a 2% uranyl acetate solution.

Results

Aggregation kinetics

Before inducing aggregation by incubating the protein samples at high temperature, we analyzed the initial quaternary structure of insulin under the investigated conditions by SEC coupled with MALS measurements. A single peak corresponding to a species with a molecular mass of 5800 ± 100 Da was observed in the SEC chromatograms (data reported in the Supporting Material, Fig. S2). This value is in excellent agreement with the molecular mass of insulin monomer (5808 Da), thereby indicating that under the investigated conditions insulin is initially present as monomer.

The formation of the aggregates was monitored by ThT fluorescence assay and by DLS, whereas the monomer conversion was measured by SEC. The time evolution of the SEC chromatograms is reported in the Supporting Material (Fig. S3).

The ThT fluorescence values and the monomer conversion during the aggregation kinetics of a 5 g/L (860 μM) insulin solution with 100 mM NaCl or Na₂SO₄ are shown in Fig. 1, a and b, respectively. ThT fluorescence assay is a common technique applied to monitor the formation of amyloid fibrils (64), which relies on the increase of the fluorescence yield of the dye upon binding to β-sheet structures.

Figure 1.

Insulin aggregation kinetics in the presence of chloride and sulfate anion. Time evolution of the ThT fluorescence values (▪) and of the residual monomer amount evaluated by SEC technique (▼) for a 5 g/L (860 μM) insulin solution at pH 1.6 and 60°C with (a) 100 mM NaCl or (b) 100 mM Na₂SO₄; (c) average hydrodynamic diameter measured by DLS for a 5 g/L (860 μM) insulin solution at pH 1.6 and 60°C with 100 mM NaCl (▲) or Na₂SO₄ (•). To see this figure in color, go online.

In the presence of sodium chloride the time evolution of the ThT signal shown in Fig. 1 a exhibits the sigmoidal profile commonly observed during the in vitro aggregation of several amyloidogenic proteins: a lag-phase is followed by rapid growth until a plateau corresponding to monomer depletion is reached (56). The amount of soluble monomer as a function of time decreases according to a specular sigmoidal profile, being negligible during the lag-phase and decreasing rapidly during fibril growth until complete conversion is reached. This result indicates that essentially all the converted monomers are present in the form of fibrils characterized by β-sheet structures that are able to bind ThT.

In analogy with the behavior in the presence of sodium chloride, also in the presence of sodium sulfate the time evolution of the ThT fluorescence follows a sigmoidal profile, as shown in Fig. 1 b. However, in contrast with the situation with sodium chloride, after 6 h incubation ∼20% of monomer conversion is observed despite the ThT fluorescence signal is still in the background noise level (Fig. 1 b). In addition, the increase of ThT fluorescence during time in the presence of sodium sulfate is slower than in the presence of sodium chloride. The results suggest that sulfate induces a first aggregation step leading to non-ThT-binding aggregates, which depletes ∼20% of the monomer. This aggregation step is then followed by the formation of aggregates with amyloid-like content (see also Fig. S4).

To confirm the initial aggregation step in the presence of sulfate, we followed the aggregation kinetics by DLS (Fig. 1 c). The ThT assay is unable to detect the formation of aggregates lacking β-sheet structures. In contrast, light scattering techniques are sensitive to the formation of aggregates independently of the morphology. The results reported in Fig. 1 c show that in the presence of NaCl the DLS signal increases after ∼5 h, in correspondence with the increase of ThT fluorescence (Fig. 1 a), indicating that all the aggregates formed in the presence of sodium chloride contain amyloid structures Conversely, in the presence of sulfate the DLS signal increases already after 1 h of incubation, indicating the formation of aggregates in the micron size range and confirming the monomer conversion measured by SEC. The formation of these aggregates is not accompanied by an increase of the ThT signal, which increases only after 5 h of incubation. This result confirms that in the case of sulfate the formation of ThT-binding aggregates is preceded by the appearance of aggregates that are not able to bind ThT.

The effect of temperature on the aggregation kinetics in the presence of both salts has been analyzed in the temperature range 57–63°C. This small range of temperatures has been selected to analyze the aggregation kinetics without changing the aggregation mechanism or the aggregate morphology, which are strongly affected by the initial temperature value. The fibril formation measured by ThT assay and the time evolution of the monomer conversion at different temperatures are shown in Fig. 2. In the investigated temperature range in the presence of chloride the kinetics are only slightly affected by the initial temperature value and, as expected, they increase as temperature increases (Fig. 2, a and c). In contrast, in the presence of sulfate the kinetics are significantly affected by the initial temperature value: a difference of 3° induces a difference in the lag-phase of several hours (Fig. 2, b and d). The monomer depletion corresponding to the initial aggregation step is independent of the temperature and equal to ∼18–20%.

Figure 2.

Temperature dependence of the aggregation kinetics. Time evolution of ThT fluorescence values with (a) 100 mM NaCl or (b) 100 mM Na₂SO₄ and of the residual monomer amount with (c) 100 mM NaCl or (d) 100 mM Na₂SO₄ for a 5 g/L (860 μM) insulin solution at pH 1.6 and different temperatures: 57°C (▲), 60°C (•), and 63°C (▪). When nonvisible error bars are smaller than the symbols. To see this figure in color, go online.

On the basis of these observations, we conclude that in the presence of sulfate two different aggregation processes occur: i), an initial aggregation step that consumes ∼18–20% of the initial monomer and leads to the formation of amorphous aggregates; and ii), a second aggregation phase that forms ThT-binding structures. This second aggregation process leading to amyloid-like structure is strongly affected by temperature, and is therefore characterized by larger activation energy with respect to the fibrillation process in the presence of sodium chloride.

To support the hypothesis of the formation of two populations of aggregates characterized by different structures, in addition to ThT binding assay, which provides indirect information about the β-sheet content of the aggregates, we investigated the aggregate morphology at different times by FTIR and TEM.

Aggregate morphology

The FTIR spectra of samples taken at different time points during the aggregation reaction are shown in Fig. 3, a and b. In the presence of both chloride and sulfate at time 0 the FTIR spectrum exhibits a peak with maximum intensity at the wavenumber 1656 cm⁻¹, characteristic of the α-helix content of the protein. In the presence of chloride after 6 h of incubation the position of the peak shifts toward 1632 cm⁻¹, corresponding to the rich amount of intermolecular β-sheet structures of amyloid fibrils. This observation is consistent with the formation of fibrils monitored by ThT fluorescence (Fig. 1 a) and the almost complete monomer conversion measured by SEC at 6 h incubation (Fig. 1 a). Further analysis by TEM imaging confirms the presence of amyloid fibrils as unique final product (Fig. 3 c).

Figure 3.

Change of the aggregate morphology during the aggregation reaction. (a and b) FTIR spectra of 5 g/L (860 μM) insulin samples collected at different time points during aggregation at 60°C in 25 mM HCl at pH 1.6 with (a) 100 mM NaCl or (b) 100 mM Na₂SO₄. (c–e) TEM pictures of the final aggregates formed in the presence of (c) 100 mM NaCl and (d and e) 100 mM Na₂SO₄. In the presence of sulfate both fibrils (d) and amorphous aggregates (e) are observed. (f) Position of the maximum intensity in the FTIR spectra and (g) ThT fluorescence values at different temperatures (57°C (▲), 60°C (•), and 63°C (▪)) as a function of the monomer conversion during the aggregation of 5 g/L (860 μM) insulin in 25 mM HCl with 100 mM Na₂SO₄ at pH 1.6. To see this figure in color, go online.

In the presence of sulfate, the FTIR spectrum of the sample collected after 12 h of incubation is similar to the spectrum of the initial monomeric solution (Fig. 3 b). The absence of β-sheet structures is in agreement with the corresponding low ThT fluorescence value (Fig. 1 b). Considering that after 12 h of incubation ∼25% of the initial monomers is already converted into aggregates (Fig. 1 b), the FTIR analysis supports the hypothesis that the sulfate anion induces an initial precipitation of amorphous aggregates characterized by a secondary structure content that is similar to the initial monomeric insulin. The FTIR spectrum corresponding to 74% conversion shows a larger amount of β-sheet structures (Fig. 3 b), in agreement with the increase of ThT fluorescence (Fig. 1 b), confirming that the initial precipitation of amorphous aggregates is followed by the formation of amyloid-like structures. However, in the presence of sulfate, the samples corresponding to high conversion values retain a large amount of α-helix content together with intermolecular β-sheet structures, suggesting a different morphology of the final aggregates with respect to the fibrils produced in the presence of sodium chloride. The polymorphism of the aggregates obtained in the presence of sulfate is confirmed by the TEM pictures of samples taken after 30 h of incubation (Fig. 3, d and e): the images show the presence of both fibrils and amorphous aggregates. Moreover, the fibrils obtained in the presence of sulfate anion are shorter and more branched than the fibrils produced in the presence of chloride anion (Fig. 3, c and d).

The behavior described previously is confirmed by the analysis of the time evolution of the aggregate morphology obtained at different temperatures. In Fig. 3 f, we report the wavenumber corresponding to the maximum intensity in the FTIR spectrum as a function of monomer conversion. The data corresponding to the different temperature values overlap on a single curve. This result indicates that despite the aggregation kinetics differ significantly at different temperatures (Fig. 2, b and d), the aggregation mechanism is unaffected: at low conversion values insulin forms amorphous aggregates with α-helix content and a maximum peak in the FTIR spectrum in the range from 1650 to1656 cm⁻¹. At conversion values larger than 50% these aggregates evolve to β-sheet structures characterized by a maximum peak in the FTIR spectrum at ∼1633 cm⁻¹. The FTIR results are in excellent agreement with the ThT measurements, as shown in Fig. 3 g: the increase of ThT fluorescence with increasing conversion occurs simultaneously to the shift of the maximum peak in the FTIR spectra from 1655 cm⁻¹ to 1633 cm⁻¹.

To investigate further the structure transition, we isolated by centrifugation the amorphous aggregates produced after 18% of monomer conversion (sample I in Fig. 4 a) and we analyzed both the precipitated aggregates and the supernatant.

Figure 4.

Structure rearrangements of the aggregates formed in the presence of sulfate. (a) Time evolution of monomer conversion of a 5 g/L (860 μM) insulin solution in 25 mM HCl with 100 mM Na₂SO₄ at 60°C. The insets show a TEM image and the FTIR spectrum of the amorphous aggregates isolated by centrifugation from the sample corresponding to 18% monomer conversion (sample I). (b) Time evolution of the ThT fluorescence of the amorphous aggregates collected from sample I in (a); the insets show a TEM picture and the FTIR spectrum of the sample at the end of the incubation (sample II). The star symbol represents the ThT fluorescence value corresponding to a sample with the same amount of mature amyloid fibrils produced under the same conditions but with 100 mM NaCl instead of sulfate. (c) Aggregation kinetics of solutions that were seeded with 1% amorphous aggregates isolated during the early stages (▪) (sample I), and with 1% ordered aggregates originating from the structure reorganization of the amorphous aggregates during incubation at 60°C in the absence of monomer (▲) (sample II). Blue circles (•) represent the control unseeded experiment. Aggregation conditions are the same as in (a). To see this figure in color, go online.

The supernatant contains monomeric proteins that after a few minutes reform the amorphous aggregates at the same amount of 18–20%, indicating that these aggregates are in equilibrium with the monomers. TEM images of the isolated aggregates (Fig. 4 a) confirm the amorphous morphology of the aggregates detected by ThT and FTIR analysis.

We then investigated the stability of these amorphous aggregates in the absence of monomers by incubating the isolated aggregates at 60°C in the presence of 10 μM ThT. Notably, the ThT fluorescence increases during incubation time, as shown in Fig. 4 b, indicating structure rearrangements that lead to the formation of β-sheet structures able to bind ThT. It is interesting to note that the final ThT value is comparable but slightly smaller than the ThT fluorescence value corresponding to a sample with the same amount of mature amyloid fibrils produced under the same conditions but with 100 mM NaCl instead of sulfate (star symbol in Fig. 4 b). The structure rearrangement of the amorphous aggregates is confirmed by FTIR analysis, which shows the presence of β-sheet structures after incubation at 60°C (insets in Fig. 4, a and b). In addition, we tested the reactivity of the aggregates before and after incubation by performing seeded kinetic experiments (Fig. 4 c). The amorphous aggregates have essentially no effect on the aggregation kinetics with respect to unseeded conditions. Conversely, the ordered aggregates obtained after the structure rearrangements accelerate significantly the fibril formation.

From the robust analysis obtained with the different applied techniques (ThT binding, SEC, DLS, FTIR, TEM, and seeded kinetic experiments), we can conclude that in the presence of sulfate the amorphous aggregates formed during the early stages convert during time into amyloid-like structures. This structure change is likely driven by the minimization of the system free energy and could occur either internally or via continuous release and reincorporation of the monomers from and into the aggregates. SEC measurements of samples taken during the incubation of the isolated aggregates at high temperature show no monomer in the system (data not shown). However, rapid exchange of monomers, which is out of the detection limit of the SEC analysis, cannot be excluded.

Effect of sulfate and insulin concentration

To investigate further the effect of sulfate on the insulin aggregation pathway, we performed the analysis described in the previous sections at different protein and sulfate concentrations. In Fig. 5, we show the aggregation kinetics followed by ThT assay and SEC measurements at 60°C. We started our analysis by considering solutions at 100 mM Na₂SO₄ and varying the insulin concentration from 0.5 to 5 g/L (86–860 μM) (Fig. 5, a and b). As expected, the aggregation rate increases with increasing protein concentration. At concentration lower than 1 g/L (172 μM) the ThT signal after 30 h incubation is still in the background noise level, indicating the absence of significant amount of fibrils. The corresponding SEC measurements indicate that after 12 h incubation ∼18–20% of the monomer is consumed, and no further aggregation occurs during the incubation time. The relative amount of monomer converted into the amorphous aggregates during the initial step is independent of the initial protein concentration, and we can therefore exclude that the formation of these aggregates is due to a solubility effect.

Figure 5.

Effect of protein and Na₂SO₄ concentration on insulin aggregation in 25 mM HCl at pH 1.6 and 60°C. Time evolution of ThT fluorescence (a and c) and monomer conversion (b and d) for: (a and b) 0.5 g/L (86 μM) (♦), 1 g/L (172 μM) (◄), 3 g/L (516.6 μM) (★), and 5 g/L (860 μM) (•) insulin solutions with 100 mM Na₂SO₄; (c and d) 5 g/L (860 μM) insulin solutions with 2.5 mM (▪), 5 mM (▼), 50 mM (▲), or 100 mM (•) Na₂SO₄. To see this figure in color, go online.

After analyzing the effect of insulin concentration, we measured the aggregation kinetics at the reference insulin concentration of 5 g/L (860 μM) changing the sulfate concentration from 2.5 to 100 mM (Fig. 5, c and d). The corresponding values of the lag-phase and of the half-time measured by means of ThT fluorescence assay are reported in Table 1.

Table 1.

Lag-phases and half-times (t₅₀) of the aggregation of a 5 g/L (860 μM) insulin solution in 25 mM HCl at pH 1.6 and 60°C at different sodium sulfate concentrations as measured by ThT fluorescence assays

SO42- Conc. [mM]	Lag time [h]	t50 [h]
100	7.2	12.0
50	5.8	10.6
5	4.3	6.0
2.5	4.2	5.0

The rate of the aggregation process decreases with increasing the sulfate concentration. At 2.5 mM sodium sulfate, no precipitation of amorphous aggregates is observed, and the time evolution of monomer conversion and ThT signal is similar to the one observed in the presence of chloride (Fig. 5, c and d, and Fig. 1 a). The increase of the sulfate concentration from 2.5 to 5 mM inhibits slightly the aggregation kinetics, whereas at sulfate concentrations of 50 and 100 mM the comparison between ThT and SEC data indicate clearly the formation of amorphous aggregates in the early stages. In this high sulfate concentration regime, the overall aggregation rate decreases significantly with increasing the sulfate concentration. Based on this analysis, we can consider the concentration of 5 mM sulfate, which, at the working insulin concentration of 860 μM, corresponds to a sulfate/insulin molar ratio equal to 6, as an approximate threshold concentration above which the alternative aggregation pathway overimposes the nucleation fibrillation pathway. The alternative nucleation pathway dominates at larger sulfate concentrations, and explains the decrease of the overall aggregation rate with increasing sulfate concentration.

Discussion

In the physiological range of salt concentrations typically considered in protein aggregation studies (0–150 mM), the increase in the concentration of monovalent salts induces commonly an increase of the aggregation rate, as verified in our system by kinetic experiments in the presence of sodium chloride (see Fig. S5). By contrast, it is found that the increase in sulfate concentration has an inhibitory effect on insulin aggregation kinetics at low pH. The results described in the previous sections show that this effect is associated with a change in the insulin aggregation pathway with increasing the sulfate concentration.

At low sulfate concentrations (0–5 mM) insulin aggregation follows a nucleation polymerization mechanism, which forms amyloid fibrils. This mechanism has been widely described in the literature in the presence of monovalent anions (47,51,52,55–58,60–63,65) (Pathway 1 in Fig. 6).

Figure 6.

Schematic diagram showing the alternative aggregation mechanisms of insulin in the presence of sulfate. (a) At low sulfate concentrations (0–5 mM) insulin forms amyloid fibrils following the nucleated polymerization mechanism commonly observed under acidic conditions in the presence of monovalent anions; (b) When the sulfate concentration is increased above 5 mM, the sulfate anion induces the salting-out of ∼18–20% of insulin molecules into reversible amorphous aggregates that retain a large content of α-helix structures. During time, these aggregates undergo structure rearrangements into polymorphic species that exhibit β-sheet structures. To see this figure in color, go online.

At sulfate concentrations larger than ∼5 mM, ∼18–20% of the insulin monomers are initially converted into reversible amorphous aggregates (Pathway 2 in Fig. 6). When these amorphous aggregates are removed from the solution, they are reformed in a few minutes at the same amount of 18–20% of the total protein concentration, indicating that the aggregates are in equilibrium with the monomers. During incubation, these aggregates undergo structure reorganization from α-helix to β-sheet structures, which are able to bind ThT dye and act as seeds that recruit soluble monomers. The aggregation process in the presence of sulfate is characterized by larger activation energy with respect to the aggregation in the presence of chloride. This larger activation energy may be related to either the structure rearrangements of the amorphous aggregates or the elongation reaction between monomeric proteins and the seed-competent aggregates.

Not only the aggregation pathway but also the morphology of the final aggregates obtained in the presence of the two anions is different. Although the final value of the ThT signal is comparable for the two anions (Fig. 1, a and b), the TEM pictures show different macroscopic structures of the final aggregates: in particular, in the case of chloride anion only fibrils are detected (Fig. 3 c), whereas in the presence of sulfate anion both fibrillar and amorphous aggregates are observed (Fig. 3, d and e). These results are consistent with the FTIR analysis of the final aggregates, which shows a mixture of α-helix and β-sheet structures in the case of sulfate and reveals the presence of β-sheet structures in the case of chloride (Fig. 3).

The alternative aggregation mechanism in the presence of sulfate, which has been characterized in detail in this kinetic study, is in agreement with previous findings reported in the literature, which show a different morphology of insulin aggregates obtained in the presence of sulfate (38,56,66).

The microscopic mechanism responsible for the formation of the amorphous aggregates is likely due to a combination of the peculiar binding and salting-out properties of the sulfate anion. Sulfate anion has a large propensity to bind to insulin molecules (38) and to induce salt bridges (40,67). It has been recently demonstrated that, even at low salt concentrations (few to 10 mM), ion-binding and additional ion-specific effects, on top of the Debye-Hückel charge screening effect, modify protein electrostatic interactions and aggregation kinetics (20). It is expected that at larger salt concentrations (10–100 mM) the ion binding, together with the salting-out effect, can affect even more significantly both the protein structure and the intermolecular interactions.

According to circular dichroism analysis, the presence of sulfate does not modify either the initial protein structure or the thermal stability with respect to sodium chloride (data reported in the Supporting Material, Fig. S6), although small changes not detectable by circular dichroism cannot be excluded.

The reduction of insulin net charge upon binding, together with the increase of hydrophobic interactions due to preferential exclusion, increases the net attractive intermolecular interactions between insulin molecules, thus promoting aggregation, and in particular the nucleation of aggregates characterized by native-like content. These aggregates are likely formed too rapidly to reach the most stable configuration. The kinetically trapped species reorganize during time into more thermodynamically favored β-sheet structures, which are able to bind ThT dyes and recruit additional monomers.

The system described in this work shows how the kinetic and the thermodynamic features of the protein aggregation process are highly sensitive to the environmental parameters, which affect the intra- and intermolecular interactions driving the microscopic aggregation events and the overall reaction mechanism.

These aspects are relevant for instance in the biotechnology context for the development of strategies to improve the formulation of protein-based drugs. Salt-induced aggregation of proteins in aqueous solutions is a major problem that is often encountered during the production and the storage of therapeutic proteins. The results shown in this work, although obtained under strong denaturing conditions, may suggest some caution in the use of sulfate and sulfonated compounds in the processing and formulation of insulin, because these compounds can rapidly trigger the formation of amorphous precipitates, which cannot be detected by ThT fluorescence assay and can convert slowly into ordered aggregates with seeding properties.

In the context of protein aggregation involved with human diseases, it remains challenging to understand at a fundamental level the interactions between proteins and all the possible cellular components. On a long term, the results obtained on the effect of sulfate on in vitro protein aggregation may represent a first step toward the understanding of the complex in vivo interactions between proteins and sulfonated macromolecules such as glycosaminoglycans.

Conclusions

In this study, we investigated the effect of sulfate anion on the kinetics and the mechanism of insulin aggregation at low pH. It is found that the increase of sulfate concentration inhibits insulin aggregation kinetics.

This effect is associated with a change in the aggregation mechanism with increasing sulfate concentration. At low sulfate concentrations (0–5 mM) insulin monomers form amyloid fibrils following the nucleated-polymerization mechanism commonly observed with monovalent anions.

At sulfate concentrations larger than 5 mM, an alternative aggregation mechanism is observed: initially, 18–20% of monomers are converted into reversible aggregates that retain a large amount of native α-helix structure. During time, according to FTIR, ThT binding, and seeded kinetic assays, these intermediate amorphous aggregates undergo reorganization into β-sheet structures, which are able to bind to the ThT dye and seed fibril formation. TEM analysis shows that the final aggregates consist of a mixture of amorphous and fibrillar aggregates, and differ from the regular amyloid fibrils obtained in the presence of chloride.

The system characterized in this work represents a case where amorphous aggregates on-pathway to the formation of structures with amyloid-like content could be detected and characterized.