Exploring the occurrence of thioflavin-T-positive insulin amyloid aggregation intermediates

Insulin aggregation

Human recombinant insulin powder (Sigma-Aldrich cat. No. 91077C) was dissolved in a 100 mM sodium phosphate buffer (pH 2.4) containing 100 mM NaCl (reaction buffer). ThT (Sigma-Aldrich cat. No. T3516) was dissolved in H₂O to a final concentration of ∼12 mM and mixed for 10 min using vigorous agitation, after which the dye solution was filtered through a 0.22 µm pore syringe filter. An aliquot of the ThT stock solution was diluted 200 times and the exact dye concentration was determined by measuring the solution’s absorbance at 412 nm (ε₄₁₂ = 23250 M⁻¹cm⁻¹). The ThT stock solution was then diluted to a final concentration of 10 mM. The protein solution was then combined with the reaction buffer and a 10 mM ThTsolution to a final protein and ThT concentration of 100 µM (insulin ε₂₈₀ = 6,335 M⁻¹cm⁻¹, MW=5808 Da) and distributed into 200 µL test tubes (20 µL final volume). These conditions result in both types of aggregation, with a random appearance of double-sigmoidal kinetic curves.

The aggregation kinetics were tracked as previously described (Milto, Michailova & Smirnovas, 2014). In short, sample ThT fluorescence intensity was monitored using a Qiagen Rotorgene Q real-time analyzer at a constant 60 °C temperature with measurements taken every minute. A total of one thousand samples were measured in batches of 36. After the aggregation reaction, the samples were stored at 4 °C.

For seeded aggregation, fibril samples were sonicated for 10 min using a Bandelin Sonopuls ultrasonic homogenizer with a MS73 tip (40% power, with 30 s sonication/30 s rest intervals). Then insulin, ThT and fibril solutions were combined to a final protein concentration of 100 µM, ThT concentration of 100 µM and 1% or 10⁻⁵% fibrils (% of total protein mass in solution). The reaction was monitored as in the non-seeded aggregation experiment.

Kinetic data analysis

A first-order derivative was calculated for each sample’s kinetic data, using a 40-point averaging range. The maximum value of the derivative curve corresponds to the rate of aggregation, while its position—to the time at which the rate is highest (t_r). Samples which had one clear peak in the first-order derivative were regarded as normal, while ones which were composed of two peaks (regular, high-rate peak and a small, low-rate peak preceding it)—as double-sigmoidal. Data processing was done using Origin 2018 software.

Fluorescence measurements

Each sample was diluted 5 times to 100 µL with the reaction buffer containing 100 µM of ThT and their fluorescence intensity was measured using a Varian Cary Eclipse Fluorescence Spectrophotometer with 440 nm excitation (slit width—5 nm) and 480 nm emission (slit width—5 nm) wavelengths. For each case, three measurements were taken and averaged.

Atomic force microscopy

The samples were separated into two groups based on their aggregation kinetic profiles and mixed to result in a homogenous solution. 30 µL aliquots were deposited on freshly cleaved mica, incubated for 1 min, washed with 1 mL of MilliQ water and dried under airflow. In the case of intermediate aggregates during the double-sigmoidal aggregation, the real-time analyzer was stopped when the aggregation curve reached the first minor plateau. Then the samples were quickly removed and placed on freshly cleaved mica as mentioned earlier. For each condition, three 10 × 10 µm AFM images were recorded as previously described (Sneideris et al., 2019) using a Dimension Icon (Bruker) atomic force microscope, operating in tapping mode with a silicon cantilever Tap300AI-G (40 N m⁻¹, Budget Sensors). High resolution (1,024 × 1,024 pixels) images were flattened and analyzed using Gwyddion 2.5.5 and SPIP 6.7.8. Each fibril’s height was determined by tracing lines perpendicular to the fibril’s axis. Height statistical analysis was conducted by taking into consideration all three repeats for each condition, i.e., a similar number of aggregates were examined in every image.

Fourier-transform infrared spectroscopy

The two sample groups were centrifuged at 10,000 g for 30 min and resuspended in 1 mL of D₂O. The centrifugation and resuspension step was repeated 3 times and the final resuspension volume was 0.25 mL. Before measurements, both samples were sonicated using a Bandelin Sonopuls ultrasonic homogenizer with a MS72 tip (20% power and constant sonication for 30 s). Sonication helps to break fibril clumps, which leads to less scattering effects and better-quality FTIR spectra. During sample preparation and measurement, H-D exchange is insignificant, as in the case of insulin fibrils it is very slow (Dzwolak, Loksztejn & Smirnovas, 2006). The spectra were recorded as previously described (Sneideris et al., 2019). In short, the concentrated fibril samples were scanned in near-vacuum conditions (∼2 mbar) at room temperature using a Vertex 80v (Bruker) IR spectrometer. 256 interferograms were averaged for each spectrum. A D₂O spectrum was subtracted and the resulting spectra were normalized to the same area of amide I/I’ band (1,700–1,595 cm⁻¹). Data processing was performed using GRAMS software.

ThT fluorescence excitation-emission matrices

The aggregation solution was prepared as described in the insulin aggregation section to a final volume of 3 mL. The solution was then placed in a 10 mm pathlength cuvette, sealed with a plug cap to prevent evaporation and incubated at 60 °C without agitation. EEMs were scanned every 5 min using a Varian Cary Eclipse fluorescence spectrophotometer using an excitation range from 440 to 465 nm and emission range from 475 to 500 nm (excitation and emission slit widths—5 nm, wavelength step—1 nm, scan rate—600 points/min). Due to the analysis being conducted on a transitioning system, the EEM size was optimized to be as minimal as possible to lower the impact of an intensity drift, which results from different concentrations of aggregates present at the start and finish of each scan cycle. This was done by first acquiring a larger EEM, which encompassed the maximum ThT fluorescence zone, then it was narrowed down as much as possible to reduce scan time.

Each EEM was corrected for the inner filter effect caused by 100 µM of ThT as described previously (Ziaunys & Smirnovas, 2019b). In short, the correction was made by using the absorbance spectra of 100 µM non-bound ThT, as it is extremely difficult to account for absorbance changes throughout the entire reaction resulting from ThT becoming bound to fibrils and because the majority of ThT remains non-bound even when all insulin is aggregated (Ziaunys & Smirnovas, 2019b). The EEM “center of mass” was then calculated for the entire EEM after the inner filter correction. This was done to prevent signal noise, caused by light scattering, from affecting the maximum intensity position.

Aggregation kinetics

A large number (n = 1,000) of low volume insulin samples were aggregated under the exact same conditions and their kinetics were tracked by monitoring changes in ThT fluorescence intensity. Analysis of all the data revealed that a majority of samples experienced normal, sigmoidal spontaneous aggregation kinetics, with one rate maximum seen in the first order derivative (Figs. 1A and 1C). A fraction of samples displayed double-sigmoidal kinetics, with two peaks in the first order derivative curve (Figs. 1B and 1D). The first increase in fluorescence intensity of the double-sigmoidal aggregation kinetics occurred roughly 100 min before the second increase and its rate was, on average, nearly 10-fold lower.

A total of 55 samples possessed such unusual aggregation kinetics, constituting a probability of such an occurrence being at least 5.5% under the tested conditions. 77 samples had a mixed kinetic profile (very small first peak or a large overlap between both peaks), which could not be accurately attributed to either type of aggregation. The normal and double-sigmoidal samples were then separated for further analysis.

In order to determine if there are any links between the rate of aggregation, the time at which this rate is highest and the resulting fluorescence intensity of fibrils, all three factor dependencies were examined. We can see that all three parameters are mostly independent from one another (Figs. 2A–2C). The time at which the aggregation rate is highest, does not influence the rate at which fibril elongation occurs, neither does it change the final fluorescence intensity of the formed fibrils. When we compare these three factors between the normal and double-sigmoidal samples (Figs. 2D–2I), it appears that the double-sigmoidal fibrillization has a slightly lower time at which the maximum aggregation rate is reached (Figs. 2D and 2G) and it has no effect on the rate itself (Figs. 2E and 2H). There is, however, a considerable difference in the final fluorescence intensity distribution (Figs. 2F and 2I). The distribution maximum is more than 10% lower when the aggregation kinetics are double-sigmoidal, suggesting that there are either off-pathway aggregates or a fraction of fibrils possess a different ThT binding mode. Despite this distinction in average fluorescence intensity, a large portion of all three data sets overlap with one another due to a large spread caused by the stochastic nature of non-seeded insulin aggregation (Foderà et al., 2008).

Fibril structure and seeding properties

The fibril samples were examined using atomic force microscopy (AFM), Fourier-transform infrared spectroscopy (FTIR) and used as seeds to examine their rate of self-replication. In the AFM images acquired during the first increase in signal intensity during double-sigmoidal kinetics, we observe small, round oligomeric aggregate species (with most having a height of 1–2 nm) and short protofibrils (0.1–0.5 µm in length) (Fig. 3A, Fig. S1). When compared to a sample obtained before an increase in dye fluorescence is observed (Fig. S1), there are far more aggregates present in the case of ThT-positive intermediates. The ThT-negative samples also contain a higher number of 0.5–1.5 nm height assemblies and very few elongated structures (Figs. S2A–S2C). When the aggregation reactions are concluded, the fibrils are considerably longer and have a greater height, however, there do not seem to be any major differences between both cases, neither visually nor by their height distribution (Figs. 3B–3D, Fig. S1). Due to formation of large aggregate clumps, the fibrils were sonicated to better examine any possible differences in their height. While the height distribution average values are similar, there is a wider spread in the case of the sonicated double-sigmoidal sample (Fig. S3), likely caused by the existence of several smaller fibrils or amorphous aggregates. The FTIR second derivative spectra (Fig. 3E) and seeding kinetics (Fig. 3F) are also nearly identical for both cases, indicating that the double-sigmoidal aggregation does not have a significant influence on the final fibril secondary structure or self-replication properties.

ThT-positive intermediates

The normal and double-sigmoidal aggregation reactions were examined by scanning excitation-emission matrices of ThT fluorescence during aggregation. The kinetic curves were plotted as the maximum EEM signal intensity over time. When examining the kinetics of double-sigmoidal aggregation, we see that the lag phase is followed by a slow increase in ThT fluorescence intensity, then a sudden jump in intensity (marked as *), which then quickly returns to a low value and is continued by the second growth phase (Fig. 4A). Such a jump is not visible in the normal aggregation data (Fig. 4A), nor in any of the previous experiments, where samples were only scanned once a minute. This indicates that it may only be visible for a very short time during the EEM scan. When the excitation and emission wavelengths of the EEM “center of mass” are calculated (Figs. 4B and 4C), we see that there are significant changes in both of these parameters during the first, anomalous increase in ThT intensity, as compared to relatively minor variations in the normal aggregation. The excitation wavelength shifts from 457 nm to 453 nm and then rises back to 454 nm, while the emission intensity shifts between 489 nm and 487 nm and gradually reaches 488 nm. Both wavelengths reach a constant value at roughly the same time after the sudden increase in ThT fluorescence intensity, marking the end of the anomalous phase.

If we examine the top fluorescence intensity value distributions in the EEMs before and after the sudden signal jump (Fig. 4D), there are both significant shifts in the top value positions during the anomalous phase, as well as single, high intensity lines. Such lines can be caused by either a large particle floating past the optical path during a scan or by sudden association and dissociation of a ThT-positive aggregate. After the high intensity jump, once the signal returns to normal, these lines are no longer present in any of the EEMs (as seen after 290 min (Fig. 4D)) and they become nearly identical. Such an occurrence has been observed multiple times throughout the study and it ranged from being mild (Fig. S4) to very extreme (Fig. S5).