Label-Free Measurement of Amyloid Elongation by Suspended

Dec 24, 2014 - amyloid elongation rates have been obtained in the past,20,21 .... in the bypass and resonator channels were passivated by injecting a ...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Label-Free Measurement of Amyloid Elongation by Suspended Microchannel Resonators Yu Wang,† Mario Matteo Modena,† Mitja Platen,‡ Iwan Alexander Taco Schaap,‡,§ and Thomas Peter Burg*,† †

Max Planck Institute for Biophysical Chemistry, Am Fassberg 11, 37077 Göttingen, Germany Third Institute of Physics, University of Goettingen, Friedrich-Hund-Platz 1, 37077 Göttingen, Germany § Center for Nanoscale Microscopy and Molecular Physiology of the Brain (CNMPB), 37073 Göttingen, Germany ‡

S Supporting Information *

ABSTRACT: Protein aggregation is a widely studied phenomenon that is associated with many human diseases and with the degradation of biotechnological products. Here, we establish a new label-free method for characterizing the aggregation kinetics of proteins into amyloid fibrils by suspended microchannel resonators (SMR). SMR devices are unique in their ability to provide mass-based measurements under reaction-limited conditions in a 10 pL volume. To demonstrate the method, insulin seed fibrils of defined length, characterized by atomic force microscopy (AFM) and transmission electron microscopy (TEM), were covalently immobilized inside microchannels embedded within a micromechanical resonator, and the elongation of these fibrils under a continuous flow of monomer solution (rate ∼1 nL/s) was measured by monitoring the resonance frequency shift. The kinetics for concentrations below ∼0.6 mg/mL fits well with an irreversible bimolecular binding model with the rate constant kon = (1.2 ± 0.1) × 103 M−1 s−1. Rate saturation occurred at higher concentrations. The nonlinear on-rate for monomer concentrations from 0 to 6 mg/mL and for temperatures from 20 to 42 °C fit well globally with an energy landscape model characterized by a single activation barrier. Finally, elongation rates were studied under different solution conditions and in the presence of a small molecule inhibitor of amyloid growth. Due to the low volume requirements, high precision, and speed of SMR measurements, the method may become a valuable new tool in the screening for inhibitors and the study of fundamental biophysical mechanisms of protein aggregation processes. binding to amyloid fibrils, can be detected with high sensitivity using optical spectroscopic methods and thereby serve as reporters for the aggregation. For example, the fluorescencebased ThT assay has become the de facto standard for measuring the kinetics of amyloid formation. However, ThT and many other dyes are insensitive to early aggregation species. Furthermore, these dyes are very likely to interfere with the aggregation process. Congo Red, for example, has been suggested to inhibit amyloid formation, and similar concerns are often raised over ThT.14−16 Another disadvantage is that biochemical reporter systems do not allow quantitative measurements of binding rate constants in solution due to the conformation-dependent response and the changing solution concentrations of both reactants and products. Label-free methods have increasingly been used in recent years for the precision measurement of amyloid formation kinetics. Several such methods allow one to monitor the growth

M

any proteins of different sequence, size, and function tend to polymerize into amyloid fibrils sharing a common structure of pleated β-sheets.1,2 Recent studies show that in vitro this process can be promoted by exposing proteins to extreme conditions, such as low pH, chemical denaturants, or heat.3−5 However, protein aggregation is sometimes difficult to avoid even under nondenaturing conditions. This creates great challenges in the production, handling, and storage of biotechnological products and in the delivery of protein based drugs.6−8 In vivo, amyloid deposition in tissues is associated with many severe human diseases, including Alzheimer’s disease, Parkinson’s disease, and Huntington’s disease.9 While it is not yet established whether these deposits themselves play a causative role or whether they are merely a consequence of the disease, some of the intermediates, which arise on the aggregation pathway, do exhibit pronounced cytotoxicity.4,10−13 Methods that help to study the pathways of amyloid formation by measuring aggregation kinetics under different conditions and for different mutants are therefore of great interest. By far, the most widely used class of techniques for measuring amyloid formation today is the dye-binding assays. Specific dyes, which change their spectral properties upon © XXXX American Chemical Society

Received: October 14, 2014 Accepted: December 24, 2014

A

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

we show that the method can be applied to enable quantitative and highly repeatable measurements of amyloid elongation rates. An important technical advance that made this possible was the development of a covalent coupling procedure to immobilize a defined ensemble of seed fibrils on the inner silicon walls of the resonator microchannel. This is challenging due to the propensity of these high aspect-ratio channels to clog and the need to regenerate and refunctionalize devices frequently and reproducibly for repeated experiments.24 Furthermore, optimizing procedures to generate seed fibrils with a consistent size distribution proved to be critical for the quantitative interpretation of the kinetic data. Human insulin was chosen as a model system for the development of the new method. Insulin is particularly interesting as it represents a rapidly growing class of protein and peptide drugs. Fibrillation is an important concern in the production and storage of such drugs, as aggregation may dramatically increase the immunogenicity and impair the bioavailability of the molecules.6−8,25 Moreover, insulin is among the most widely used models for in vitro studies of amyloid formation, making it well suited for comparison of the SMR method with established techniques.26 We first describe the immobilization of insulin seed fibrils inside the SMR and characterize the number and surface coverage of immobilized fibrils by a combination of mass measurements, AFM, and transmission electron microscopy (TEM). The rate of monomer addition to the immobilized seed fibrils is then measured while monomer solutions continuously flow through the SMR sensor. These measurements combined with theoretical considerations show that the label-free SMR approach enables kinetic measurements of amyloid elongation in a reaction limited regime at extremely low volumetric flow rates compared to conventional methods. Another advantage is that the system allows high-throughput detection by rapidly changing the solutions in the microchannels. Measurements are conducted over a wide range of monomer concentrations with and without incorporated metal ions, natural products, or small organic molecules. Finally, the temperature of the sensor is precisely controlled, and by varying the temperature, we show how insight into the thermodynamics of the elongation process can be gained using the SMR.

of aggregates on a surface by physical means. First, high-end atomic force microscopy (AFM) can visualize the growth of individual aggregates directly and in real time with excellent spatial resolution.17 While the AFM is perhaps the most information-rich among the available techniques, it requires very high-quality sample preparation, has limited throughput, and is technically complex. In contrast, it is possible to follow the emergence and growth of aggregates in the mean, i.e., without information on size distribution, by using surface plasmon resonance (SPR) or the quartz crystal microbalance (QCM). In particular, the latter has been used extensively for monitoring the elongation kinetics of amyloid fibrils. The signals in SPR and QCM both depend on the total mass adsorbed to the sensor surface. While the SPR signal reflects the change in the effective refractive index and thickness of the adsorbed layer, QCM measures a combination of the change in mass and damping of the adlayer. Furthermore, microcantilever array sensors, similar to AFM-cantilevers and developed using micro electro-mechanical systems (MEMS) techniques, have been used to monitor growth of amyloid fibrils by monitoring changes in surface stress.18 One limitation of surface sensitive methods is that the measured kinetics can be affected by mass transport. Although detailed models can account for this effect, they introduce additional free parameters in the data analysis, which inevitably increases the uncertainty of extracted rate constants.19 Another way to minimize the effect of transport limitations is through convection. Fast flow, however, comes at the expense of large sample consumption. Due to the size of flow cells in commercially available SPR and QCM instruments, rates up to 1 mL/min are common. This is sometimes prohibitive for experiments with rare and expensive reagents. Compared to the height of the flow cell, which typically exceeds 50 μm, the sensitive region of SPR and QCM is relatively small, extending only a few hundred nanometers from the surface. When monitoring the growth kinetics of large protein aggregates, this nonlinear sensitivity profile must be taken into account. Despite these limitations, very high quality label-free measurements of amyloid elongation rates have been obtained in the past,20,21 and improvements in sample consumption, linearity, and resolution would be desirable to allow more extensive studies and facilitate screening of large sets of conditions in a short time. Here, a new label-free method is introduced for precision measurements of amyloid elongation kinetics. The method uses a recently developed class of microfluidic mass sensors known as suspended microchannel resonators (SMR), which have an internal swept volume of only ∼10 pL and can resolve mass changes on the order of 1 fg. The combination of small volume and high sensitivity enable measurements in a short time and at high flow velocity while consuming only minute amounts of sample (less than 1 μL/min). Although the connections required for sample delivery to the resonator have a volume of a few microliters, unused fluid contained therein is not lost and can be recovered. Measurements are conducted inside 3 μm × 8 μm microchannels embedded inside vacuum-packed micromechanical resonators. Small shifts in the mechanical resonance frequency provide a direct measure of the change in mass of the resonator and the density of the fluid contained within it.22,23 It has been shown previously that SMR devices in principle enable the measurement of biomolecular interaction kinetics by monitoring the increase in mass as molecules bind to and accumulate on the inner surface of the resonator. In this work,

■

EXPERIMENTAL SECTION Chemicals. Human insulin, 3-aminopropyl trimethoxysilane (APTMS), glutaraldehyde solution (10% in water), sodium cyanoborohydride (NaCNBH3), and thioflavin T (ThT) were purchased from Sigma-Aldrich, Steinheim, Germany. 4-(2Hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), sodium chloride (NaCl), calcium chloride (CaCl 2), and magnesium chloride were purchased from Carl Roth, Karlsruhe, Germany. Glycine, sulfuric acid (97%), ethanol (99.9%), and sodium hydroxide (NaOH) solution (1 M) were obtained from Merck, Darmstadt, Germany. Hydrogen peroxide (30% w/v) was obtained from Fisher Scientific GmbH (Schwerte, Germany). Pthalocyanine tetrasulfonate (PcTS) was purchased from MP Biomedicals, Heidelberg, Germany. Water was purified in a TKA MicroPure Ultra Pure Water System produced by Thermo Fisher Scientific, Niederelbert, Germany. Silicon wafers (100 mm × 0.5 mm) were obtained from ABC GmbH, Brunnthal, Germany. Preparation and Characterization of Insulin Seed Fibrils. To prepare long insulin amyloid fibrils, fresh monomer B

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 1. Schematics of protein immobilization procedure in closed SMR microchannels (a). Real-time monitoring of immobilization of insulin seed fibrils (b). AFM image of Si-substrate with immobilized insulin fibrils. The scale bars are 200 nm (c).

detected using a custom-built optical lever system. The resonance frequency shift was calibrated as a function of fluid density by monitoring the signals of sodium chloride solutions at different concentrations. Continuous monitoring of the resonance frequency was done by incorporating the micromechanical resonator as the frequency-determining element into an oscillator circuit. A pressure-driven fluid delivery system was applied to inject sample solutions. Additionally, the temperature of the SMR chips was regulated using a Peltier element. Surface Functionalization. Functionalization of the inner surfaces of the embedded silicon microchannel was done in several steps, as shown in Figure 1a. First, reactive amine groups were introduced by gas phase silanization using a process similar to the one described in ref 24. After plasma treatment of the SMR chips for 10 min to clean the external surfaces, the chips were sealed in 10 mL glass vials together with 3 μL APTMS under argon and incubated at 40 °C for 6 h (Step I, Figure 1a). In Step II, the devices were further functionalized with aldehyde groups. Glutaraldehyde solution was diluted to 10% in water and centrifuged at 30 000g for 15 min at 4 °C to reduce the content of large glutaraldehyde polymer aggregates in solution. The supernatant solution was introduced into the resonator through both bypass channels for 5 min. Afterward, the channels were rinsed with water. The formed Schiff base bonds were stabilized using NaCNBH3 (10 mg/mL in water) for 10 min. In Step III, seed fibrils were immobilized. Clear fibril solutions were prepared in HEPES buffer as described above and flown through the small

solutions (6 mg/mL in 10 mM HEPES buffer at pH 2.0) were incubated under continuous stirring at 37 °C in 1.5 mL borosilicate glass vials with rubber-lined closures. Aliquots (3 μL) were withdrawn from the solutions at 2 h intervals, and the fluorescence was measured by ThT assay. The fibril formation was complete when the fluorescence intensity did not increase. Alternatively, long fibrils were also prepared by seeding 6 mg/mL monomer solution with short fibril fragments and incubating the solution for 2−4 h. The solutions were stored at room temperature. To prepare short insulin seed fibrils of uniform size, the long fibril solutions were first diluted to 1 mg/mL and then ultrasonicated for 150 min with 15 min intervals at 4 °C. The samples were quickly transferred into 0.2 μm centrifugal filter tubes (Millipore) and filtered at 12 000g at room temperature. Electron microscopy was conducted for insulin monomers and fibrils. The sample solutions (1 mg/mL) were diluted 50− 100 times, deposited on Formvar-coated 200 mesh copper grids, washed with water, and stained with freshly prepared 1% (w/v) uranyl acetate. The samples were evaluated with a CM 120 transmission electron microscope (FEI, Eindhoven, and The Netherlands). The pictures were taken with a TemCam 224A slow scan CCD camera (TVIPS, Gauting, Germany). SMR Measurements and Calibration. Torsional suspended microchannel resonators, consisting of two paddles with an embedded microchannel of 3 μm × 8 μm (height × width) and cross section as described previously, were used for all measurements.27 Electrostatic excitation was used to drive the SMR into mechanical resonance, and the vibration was C

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

into (8.56 ± 1.03) × 10−16 g/μm2. Although the size distribution of seed fibrils is normally very broad, it can be narrowed significantly by a self-limiting fragmentation under ultrasound treatment.28 After ultrasonication, transmission electron microscopy revealed an average seed length of 97.6 ± 20.1 nm and a width of 12.3 ± 0.8 nm (Figure S-2d, Supporting Information). The fibril width value implies that each produced seed fibril consists of two protofibrils that have been reported to be 5−7 nm wide.17,29 As the mass-per-length value of the fibrils with two protofibrils is 2.85 ± 0.35 kDa/ Å,8,30 the averaged buoyant mass of a single short seed fibril was calculated to be 1.2 ± 0.4 ag, yielding a surface fibril density of 734 ± 89 molecules/μm2 after the immobilization. In this calibration, it is assumed that the layer of seed fibrils forms a homogeneous coating across high-sensitivity and low-sensitivity regions of the SMR. Support for this assumption comes from the fact that the transient mass signal during immobilization always settled, indicating full saturation of the sensitive surface. In addition, fluorescence-based experiments using transparent microchannels of similar size had confirmed earlier that the functionalization protocol resulted in a uniform surface coverage of binding sites.24 Independent surface density measurements of immobilized seed fibrils were performed via AFM on open silicon surfaces by following a similar protein coupling procedure. Compared to the immobilization inside the SMR, there were two differences on the open surface: first, the temperature was kept at 25 °C instead of at 37 °C; second, open surfaces were incubated in fibril solutions with fixed fluid volume under stirring. As shown in Figure 1c, densely packed insulin seed fibrils were clearly observed in contrast to the aminosilane-modified silicon surface without fibril immobilization. By counting the fibrils in multiple 1 × 1 μm scans acquired on different parts of the surface, we obtain a surface fibril density of 235 ± 25 molecules/μm2 (mean ± SD), which is on the same order as the value obtained by the SMR measurements. Due to unavoidable differences in immobilization protocols, the surface densities are not expected to match exactly, and this is not required for the elongation rate measurements. Although the solution concentrations applied using both methods were sufficiently high to fully saturate the surface area, the lower density as observed by AFM may, for example, be due to the temperature or due to experimental variability in the coverage of the open surface sample. Nevertheless, this supports that the SMR signal gives a realistic measure of the surface density of the deposited fibrils. Only the SMR derived values of seed fibril coverage are later used to calculate the absolute rate constant of the insulin fibril elongation kinetics. Figure 2a shows the SMR resonance frequency shift during the elongation of immobilized insulin seed fibrils exposed to a continuous flow of fresh 1 mg/mL monomer solution at pH 2.0 and 37 °C. The frequency drop induced by the density change of the buffer was subtracted from the elongation signal for clarity. The linear signal decrease over time indicates that the fibrils grow at a constant rate. Calibrating the mass sensitivity of the resonator and normalizing the rate of total mass change by the buoyant mass of monomeric insulin (2.9 zg) and by the number of immobilized fibrils, the measurements revealed an elongation rate of dN/dt =12.2 ± 1.2 monomer molecules per fibril per minute. The linear model of dN/dt = k × NFibril × cMonomer then yields a rate constant of k = (1.2 ± 0.1) × 103 M−1 s−1. This rate measured with the SMR for human insulin is comparable to but slower than the rate of (9.2 ± 3) × 103

resonator channel for 15 min via one of the two bypasses. The other bypass channel was filled with HEPES buffer (10 mM at pH 2.0). The seed fibril solutions were then replaced by the HEPES buffer, followed by water and 10 mg/mL NaCNBH3 solution for 20 min. In Step IV, any remaining aldehyde groups in the bypass and resonator channels were passivated by injecting a glycine buffer solution (50 mM at pH 2.0), followed by rinsing with water and NaCNBH3 for 20 min. Step V represents the actual measurement of insulin fibril elongation. This was done using monomer solutions at different concentrations, different temperatures, and with and without several additives. After each series of experiments, the device was cleaned (Step VI) using piranha solution (2:1 sulfuric acid/ hydrogen peroxide) and thoroughly rinsed with water. (CAUTION: “Piranha” solution reacts violently with organic materials; it must be handled with extreme care.) Atomic Force Microscopy (AFM). Short insulin seed fibrils were immobilized on 3 mm × 6 mm silicon substrates at 25 °C following the previous immobilization procedure in 1.5 mL Eppendorf tubes on a shaker with 2000 rotations per minute. The samples were then dried with nitrogen gas and imaged with a Cervantes FullMode AFM system (Nanotec, Spain) using silicon cantilevers (Olympus OMCL-AC240TS, tip radius of 7 nm, spring constant of 1.8 N/m, resonance frequency at ca. 70 kHz). Force curves were performed on the substrate to estimate the sensitivity (68 nm/V). The images were acquired at room temperature using amplitude modulation by oscillating the cantilever at 65 kHz with the amplitude of 3.5−7 nm.

■

RESULTS AND DISCUSSION Immobilization of Seed Fibrils and Quantitative Measurements of Elongation Rates. For the measurement of insulin fibril elongation, seed fibrils were covalently immobilized on the aminated surfaces of the 3 μm high and 8 μm wide microchannel as shown in Figure 1b. Glutaraldehyde was first injected (Figure 1a,II), followed by water and HEPES immobilization buffer. Figure 1b shows the measured mass change over time. The buffer injection was accompanied by a sudden frequency drop due to the difference in fluid density. The seed fibril solution was then introduced to the reactive aldehyde groups, showing a saturating signal within 15 min (Figure 1a,III,b). After rinsing with HEPES buffer and water, a total baseline shift of Δf Fibrils = 71.3 ± 8.6 Hz due to the mass of immobilized seed fibrils was recorded. The bound seed fibril ensemble was subsequently stabilized, and remaining aldehyde groups on the surface were passivated (Figure 1a,IV). The uncertainty in the value of Δf Fibrils represents the standard deviation over five complete regeneration/functionalization cycles. Each cycle started with a fresh silanization after cleaning with H2O2/H2SO4 (piranha solution), thorough rinsing with water, and drying. The final frequency shift of Δf Fibrils was always within 10% of the mean, indicating that the developed immobilization technique is highly reproducible. The frequency shift induced by the immobilized ensemble of seed fibrils was used to estimate the number of seeds on the surface. By conducting the calibration with NaCl solutions at different concentrations (Figure S-1, Supporting Information), the slope of the linear fit shows that a density change of 1 mg/cm3 results in a frequency shift of 76.11 ± 1.19 Hz, corresponding to the density sensitivity of Sp = 53979 ± 844 ppm/(g·cm−3) within the 3 μm × 8 μm (height × width) channel. As such, the detected frequency shift can be translated D

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

growth resulted in a total length increase of about 88 nm. The length increase was negligible compared to the 3 μm by 8 μm microchannel dimension, implying that the fibril growth neither clogged the channel nor affected the fluid flow in the channel. Furthermore, the constancy of the slope in Figure 2a clearly verified that the growth rate was independent of the fibril length. This is important for further quantitative measurements under a series of conditions, as rates measured sequentially on the same fibril ensemble can be directly compared. The stability and the reliability of the SMR measurements were corroborated by repeating the elongation rate measurements on the same ensemble of surface fibrils with the same monomer concentration (0.4 mg/mL) on different days. As shown in Figure 2b, the measured elongation rates were consistent for at least 1 week, with small variations possibly due to small differences in concentrations of the freshly prepared monomer solutions on different days. To ensure that quantitative measurements of amyloid elongation rates were fully reaction limited and not influenced by mass transport, we analyzed the relevant transport coefficients for our system. Since the flow was controlled by applying a constant pressure, the actual flow velocity was determined experimentally. This was done by running solutions containing polystyrene tracer particles and measuring the transit time through the resonator from the mass signal as described previously.27 At the slowest setting, a transit time of about 10 ms through the 144 μm long channel was measured. This should be considered a lower limit, since the actual elongation rate measurements were typically conducted at higher differential pressure with correspondingly faster flow. The Peclet number, Pe = vc × H/2D, for insulin monomers in the SMR microchannel can thus be estimated to be larger than 200, indicating that any concentration gradients will be predominantly across a boundary layer adjacent to the surface. Here, vc is the velocity at the center of the channel, H is the height of the channel, and D is the diffusion coefficient that is 1.0 × 10−10 m2/s for insulin. Importantly, diffusion of monomers across the boundary layer is much faster than the reaction time. While every monomer addition took place at the scale of several seconds, the monomer diffusion time for a full channel height is only on the order of 20 ms. This corresponds to a Damkohler number Da ≪ 1 (Da = time(diffusion)/ time(reaction)), indicating that mass transport does not limit the amyloid elongation process considered here. Dependence of the Elongation Rate on Temperature and Monomer Concentration. Having established the repeatability and precision of amyloid elongation measurements by the SMR, we applied the method to study the kinetics and thermodynamics of insulin fibril elongation by measuring over a range of temperatures and monomer concentrations. The obtained elongation rates, summarized in Figure 3, show a rapid increase with temperature at all concentrations. At any fixed temperature, the elongation rate first increased in proportion to the monomer concentration up to approximately 0.6 mg/mL and approached a constant value thereafter. No measurable dissociation of fibrils was observed after rinsing at any of the tested temperatures. A nonlinear dependence of elongation rate on monomer concentration has been reported previously for insulin and other amyloid systems using surface sensitive detection techniques.31,37,38 However, there are significant differences in the quantitative models used to explain this observation in the literature. For example, an induced fit mechanism has been suggested for Aβ fibril elongation, which

Figure 2. Kinetic measurements of insulin seed fibrils on the SMR channel surfaces in the presence of 1 mg/mL insulin monomer solution for 30 min at 37 °C (a). Stability monitoring of surface fibrils by measuring elongation rates with 0.4 mg/mL insulin monomers for at least 1 week (b).

M−1 s−1 reported for bovine insulin using the QCM (measured at the same monomer concentration of 1 mg/mL and the same pH 2.0 but at room temperature).31 The on-rate for bovine insulin measured by QCM at room temperature can be extrapolated to 37 °C using the Arrhenius law. This yields an on-rate of 4.8 × 104 M−1 s−1 (Table 1) based on the reported Table 1. Thermodynamic Parameters for Different AmyloidRelated Proteins Using Various Approaches protein human insulin human insulin bovine insulin human Aβ human Aβ

SMR QLS48 QCM31 QCM20 QLS49

ΔH‡ or EA [kcal/mol]

TΔS‡ (25 °C) [kcal/mol]

ΔG‡ [kcal/mol]

± ± ± ± ±

20.2 ± 0.1 14.9 ± 0.6 16.8 ± 2.0 14.4 ± 3.2 ca. 16

6.1 ± 2 8±3 6±2 1.4 ± 1.0 ca. 7

26.3 25 24.4 15.8 23.0

8.5 2 1.0 1.9 0.6

activation energy. The faster elongation process observed by QCM could be due to the fact that bovine insulin is more prone to form fibrils than human insulin due to different amino acid residues in three positions.32 In addition, differences due to the surface chemistry and the preparation of seed ensembles can also not be ruled out with certainty based on the information available. In the future, a side-by-side comparison of methods using a wider range of model systems should make it possible to clearly differentiate systematic differences due to the detection technology from the influence of different protocols and reagents. A control experiment (Figure 2a, black line), in which step III in Figure 1a (fibril immobilization) had been omitted, shows no significant change in the resonance frequency over 30 min, verifying that nonspecific adsorption of insulin monomers was below the detection limit. As each seed fibril was assumed to consist of two twisted protofibrils detected by EM (Figure S-2, Supporting Information) and each monomer contributes to a 0.48 nm length increase along the fibril axis,33−36 the half-hour fibril E

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Influence of Additives on Elongation Kinetics. As another application of amyloid elongation rate measurements by SMR, we measured the influence of 10 compounds selected from a large variety of candidates, including metal ions, natural products, and small aromatic organic molecules that have been reported to interfere with amyloid formation. The additives were dissolved in 0.4 mg/mL monomer solutions at 37 °C and with concentrations spanning several orders of magnitude. The pH was kept at pH 2.0 due to the reduced solubility of insulin monomers at higher pH values (Figure S-3, Supporting Information). The elongation rates were measured while successively introducing solutions with increasing additive concentrations. Two reference measurements with the same monomer concentration but without additives were performed at the start and at the end of each sequence. The rate comparison provides more insight into the interaction mechanisms between the additives and the monomers/fibrils. Except for the natural products that did not show any significant effect on the elongation rate, the observed rate changes obtained with several metal ion additives and organic molecules are discussed below. We first selected four metal ion salts: NaCl, ZnCl2, MgCl2, and CaCl2. Surprisingly, no significant inhibition was observed with any of the four compounds. Interestingly, all compounds exerted a similar nonmonotonic acceleration effect on the insulin elongation with maxima around 60 mM, as summarized in Figure 4. In the case of NaCl, a common component in

Figure 3. Plots showing summarized on-rates at different concentrations and temperature, and the fitting model for insulin elongation with a single energy barrier.

consists of two steps, a highly reversible dock-step followed by a slow irreversible lock-step.39 A three-step polymerization model has been concluded on the basis of kinetic data measured by SPR, which suggested that a maturation step takes place in addition to the initial attachment (docking) and the conformational rearrangement (locking). It was suggested that conformational rearrangement of monomers that are loosely attached was the rate-limiting step at higher concentrations. An interesting alternative description has been given by Buell et al., who showed that a single activation barrier is sufficient to account for the nonlinear relationship between monomer concentration and fibril elongation rate.40 In this model, rate saturation arises from the fact that only one monomer at a time can reside inside the reaction volume, and the rate r of monomer addition to each seed fibril on the surface is described by 1 ck r= (1) 1 + c /c ̃ Equation 1 matches the functional form of the Michaelis− Menten model known from enzyme kinetics. Here, c̃ represents a critical concentration above which the elongation rate saturates. For concentrations c ≪ c̃, r is directly proportional ‡ to c with the proportionality constant k = Γe−G /RT. The prefactor Γ can be predicted on the basis of polymer theory as described by Buell et al.,40 yielding Γ = 9.03 × 106 M−1 s−1 for human insulin.20 A global fit of eq 1 (shown in Figure 3) over the measured range of temperatures and concentrations was carried out using c̃, the activation enthalpy H‡, and the activation entropy S‡ as parameters and substituting G‡ = H‡ − TS‡. The resulting critical concentration of c̃ = 0.35 ± 0.1 mg/mL is smaller than previously reported for bovine insulin.40 This difference could be explained by several factors, including differences between the human protein used in this work and bovine insulin, the different coupling of seed fibrils to the surface of the sensor, and differences in the preparation of the seed fibril ensembles. The obtained enthalpy from the global fitting is consistent with the activation energy determined by the Arrhenius plots at all tested concentrations. Moreover, extracted values for the enthalpy and entropy of activation are close to values measured by other methods, as shown in Table 1. Importantly, the similarity between the values obtained by QLS (quasielastic light scattering) in solution and by SMR suggests that the SMR measurement of the thermodynamic parameters is not strongly influenced by the fact that it is surface based.

Figure 4. Insulin elongation rates with metal ions (NaCl, ZnCl2, MgCl2, and CaCl2). The test procedure started with the initial injection of the 0.4 mg/mL monomer concentration as the first reference test, followed by the injection of the monomer solutions with increased additive concentrations and the final injection of the monomer solution for a second reference test.

biological buffers, our results agree with the reported observation of a slightly accelerated amyloid formation rate with increasing concentrations up to 50 mM, as assayed by ThT.41 Zn(II) ions did not decelerate the rates, which is not unexpected as all experiments were conducted at a low pH.42 The greatest acceleratory effect was observed for MgCl2 and CaCl2. Metal ion addition in our experiments never showed any remnant effect on the elongation rates. This was confirmed by injecting pure monomer solutions before and after each series of additives and comparing the resulting elongation rates. Finally, the nonmonotonic dependence of elongation rate on the ionic strength suggests the involvement of at least two competing mechanisms. Charge-screening and changes in the F

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

profile of the SMR varies strongly along the channel, with fibrils located at vibrational nodes contributing zero and fibrils at antinodes contributing maximally to the signal, these differences are averaged in the ensemble measurement. A natural limitation of the method is given by the channel height. Only samples that are free of aggregates larger than ∼1 μm pass easily through the devices used here. Stringent cleanliness and filtration are therefore paramount with this method. An important basis for the characterization of amyloid elongation rates was the development of a stable and reproducible method for immobilizing a defined ensemble of seed fibrils. Gas phase silanization was used for the first time to enable covalent coupling of biomolecules in the high-aspect ratio embedded silicon microchannels of the SMR. Furthermore, the possibility to regenerate the devices and obtain quantitatively highly reproducible results after repeated functionalization was shown. Using insulin as a model system, the growth rates of the elementary elongation step have been systematically monitored at different temperatures and for a wide range of monomer concentrations. Mechanistic models and thermodynamic parameters can be readily extracted from the kinetic data. Furthermore, the effects of different additives on amyloid elongation kinetics were studied as a demonstration of the ability to screen for potential inhibitors of amyloid growth. In the future, we envision that the SMR method will prove a valuable new tool for the label-free characterization of amyloid formation kinetics and other processes that involve the assembly of large macromolecular complexes in vitro. Significant progress toward enhancing the sensitivity by miniaturization has already been made, and it is likely that this development will enable the detection of far smaller and fewer aggregates than presented here. This would provide the exciting opportunity to follow such processes with great precision even during the early prefibrillar stages, which are difficult to characterize by other means.

interaction with water are likely significant, and precision SMR measurements over a more extensive range of ionic and nonionic cosolvents should make it possible to differentiate these effects.43,44 We further measured the change in insulin fibril elongation rates in the presence of pthalocyanine tetrasulfonate (PcTS), which has been recently described as an inhibitor of amyloid formation.45,46 PcTS was added to the monomer solutions at concentrations ranging from 1 to 100 μM. As shown in Figure 5, a relevant inhibitory effect is clearly observed at the highest

Figure 5. Inhibitory effect of small organic PcTS molecules for insulin elongation: (a) shows the frequency monitoring of monomer solutions before, during, and after the PcTS addition; (b) shows the measured elongation rates.

■

concentration of ca. 100 μM, which corresponded to an approximately equimolar mixture of PcTS and insulin monomers. Furthermore, the reference rate measurements showed no significant change before and after the PcTS addition. Both observations are consistent with a process by which the inhibitory effect of PcTS results from interactions between the compound and the monomers in solution rather than interactions with the seed fibrils.

ASSOCIATED CONTENT

S Supporting Information *

Sensitivity calibration of the SMR device, electron micrographs of prepared seed fibrils, and solubility test of insulin monomers. This material is available free of charge via the Internet at http://pubs.acs.org.

■

■

CONCLUSIONS We have introduced a new method for the label-free measurement of amyloid elongation kinetics by nanomechanical mass sensing. Compared with other surface-based label-free methods, including QCM and SPR, there are several differences. First, as the channel in which the measurement is conducted is embedded inside the micromechanical resonator, sample consumption is on the order of 100 nL/min even at a high linear flow velocity. At this rate, the system is shown to be well within the reaction limited regime. Therefore, data analysis is relatively straightforward. Quantitative interpretation of results in SMR measurements is further facilitated by the fact that the signal is directly proportional to the total buoyant mass of fibrils in the channel and does not depend on dissipation or thickness of the surface-attached layer. Due to the large stiffness and symmetry of the device, changes in surface stress do not measurably affect the SMR signal.47 Although the sensitivity

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Fax: +49-551-2011577. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS Suspended Microchannel Resonator devices were generously provided by the laboratory of Prof. Scott Manalis (MIT, Cambridge, MA). The authors would like to thank Ms. Gudrun Heim and Dr. Dietmar Riedel in Max Planck Institute for Biophysical Chemistry for carrying out the EM measurements. This work was supported by the Max Planck Society and the Max Planck Institute for Biophysical Chemistry. I.A.T.S. was funded through the Cluster of Excellence and DFG Research Center Nanoscale Microscopy and Molecular Physiology of the Brain. M.P. was supported by the DFG (SFB860) and a stipend G

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

(32) Nielsen, L.; Frokjaer, S.; Brange, J.; Uversky, V. N.; Fink, A. L. Biochemistry 2001, 40, 8397−8409. (33) Burke, M. J.; Rougvie, M. A. Biochemistry 1972, 11, 2435−2439. (34) Kumar, S.; Walter, J. Aging 2011, 3, 803−812. (35) Cohen, S. I. A.; Vendruscolo, M.; Dobson, C. M.; Knowles, T. P. J. J. Mol. Biol. 2012, 421, 160−171. (36) Westermark, P.; Benson, M. D.; Buxbaum, J. N.; Cohen, A. S.; Frangione, B.; Ikeda, S. I.; Masters, C. L.; Merlini, G.; Saraiva, M. J.; Sipe, J. D. Amyloid 2005, 12, 1−4. (37) Collins, S. R.; Douglass, A.; Vale, R. D.; Weissman, J. S. PLoS Biol. 2004, 2, 1582−1590. (38) Rogers, S. S.; Krebs, M. R.; Bromley, E. H.; van der Linden, E.; Donald, A. M. Biophys. J. 2006, 90, 1043−1054. (39) Esler, W. P.; Stimson, E. R.; Jennings, J. M.; Vinters, H. V.; Ghilardi, J. R.; Lee, J. P.; Mantyh, P. W.; Maggio, J. E. Biochemistry 2000, 39, 6288−6295. (40) Buell, A. K.; Blundell, J. R.; Dobson, C. M.; Welland, M. E.; Terentjev, E. M.; Knowles, T. P. J. Phys. Rev. Lett. 2010, 104, 2281011−2281014. (41) Nielsen, L.; Khurana, R.; Coats, A.; Frokjaer, S.; Brange, J.; Vyas, S.; Uversky, V. N.; Fink, A. L. Biochemistry 2001, 40, 6036−6046. (42) Noormagi, A.; Gavrilova, J.; Smirnova, J.; Tougu, V.; Palumaa, P. Biochem. J. 2010, 430, 511−518. (43) Ruckenstein, E.; Shulgin, I. L. Adv. Colloid Interface Sci. 2006, 123, 97−103. (44) Shimizu, S.; McLaren, W. M.; Matubayasi, N. J. Chem. Phys. 2006, 124, 2349051−2349054. (45) Akoury, E.; Gajda, M.; Pickhard, M.; Biernat, J.; Soraya, P.; Griesinger, C.; Mandelkow, E.; Zweckstetter, M. J. Am. Chem. Soc. 2013, 135, 2853−2862. (46) Levy-Sakin, M.; Shreberk, M.; Daniel, Y.; Gazit, E. Islets 2009, 1, 210−215. (47) Burg, T. P.; Mirza, A. R.; Milovic, N.; Tsau, C. H.; Popescu, G. A.; Foster, J. S.; Manalis, S. R. J. Microelectromech. Syst. 2006, 15, 1466−1476. (48) Mauro, M.; Craparo, E. F.; Podesta, A.; Bulone, D.; Carrotta, R.; Martorana, V.; Tiana, G.; San Biagio, P. L. J. Mol. Biol. 2007, 366, 258−274. (49) Kusumoto, Y.; Lomakin, A.; Teplow, D. B.; Benedek, G. B. Proc. Natl. Acad. Sci. U.S.A. 1998, 95, 12277−12282.

from the Göttingen Graduate School for Neurosciences, Biophysics, and Molecular Biosciences.

■

REFERENCES

(1) Dobson, C. M. Trends Biochem. Sci. 1999, 24, 329−332. (2) Carulla, N.; Caddy, G. L.; Hall, D. R.; Zurdo, J.; Gairi, M.; Feliz, M.; Giralt, E.; Robinson, C. V.; Dobson, C. M. Nature 2005, 436, 554−558. (3) Booth, D. R.; Sunde, M.; Bellotti, V.; Robinson, C. V.; Hutchinson, W. L.; Fraser, P. E.; Hawkins, P. N.; Dobson, C. M.; Radford, S. E.; Blake, C. C. F.; Pepys, M. B. Nature 1997, 385, 787− 793. (4) Dobson, C. M. Nature 2003, 426, 884−890. (5) Bradbury, E. M.; Brown, L.; Downie, A. R.; Elliott, A.; Fraser, R. D. B.; Hanby, W. E.; Mcdonald, T. R. R. J. Mol. Biol. 1960, 2, 276− 286. (6) Brange, J.; Andersen, L.; Laursen, E. D.; Meyn, G.; Rasmussen, E. J. Pharm. Sci. 1997, 86, 517−525. (7) Dische, F. E.; Wernstedt, C.; Westermark, G. T.; Westermark, P.; Pepys, M. B.; Rennie, J. A.; Gilbey, S. G.; Watkins, P. J. Diabetologia 1988, 31, 158−161. (8) Ivanova, M. I.; Sievers, S. A.; Sawaya, M. R.; Wall, J. S.; Eisenberg, D. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 18990−18995. (9) Hartl, F. U.; Hayer-Hartl, M. Nat. Struct. Mol. Biol. 2009, 16, 574−581. (10) Kisilevsky, R. J. Struct. Biol. 2000, 130, 99−108. (11) Soto, C. Nat. Rev. Neurosci. 2003, 4, 49−60. (12) Kelly, J. W. Nat. Struct. Biol. 2000, 7, 824−826. (13) Goldsbury, C.; Frey, P.; Olivieri, V.; Aebi, U.; Muller, S. A. J. Mol. Biol. 2005, 352, 282−298. (14) Buell, A. K.; Dobson, C. M.; Knowles, T. P. J.; Welland, M. E. Biophys. J. 2010, 99, 3492−3497. (15) Hudson, S. A.; Ecroyd, H.; Kee, T. W.; Carver, J. A. FEBS J. 2009, 276, 5960−5972. (16) Noormai, A.; Primar, K.; Tougu, V.; Palumaa, P. J. Pept. Sci. 2012, 18, 59−64. (17) Adamcik, J.; Mezzenga, R. Curr. Opin. Colloid Interface Sci. 2012, 17, 369−376. (18) Knowles, T. P. J.; Shu, W.; Huber, F.; Lang, H. P.; Gerber, C.; Dobson, C. M.; Welland, M. E. Nanotechnology 2008, 19, 3840071− 3840075. (19) Gervais, T.; Jensen, K. F. Chem. Eng. Sci. 2006, 61, 1102−1121. (20) Buell, A. K.; Dhulesia, A.; White, D. A.; Knowles, T. P. J.; Dobson, C. M.; Welland, M. E. Angew. Chem., Int. Ed. 2012, 51, 5247− 5251. (21) Cannon, M. J.; Williams, A. D.; Wetzel, R.; Myszka, D. G. Anal. Biochem. 2004, 328, 67−75. (22) Burg, T. P.; Godin, M.; Knudsen, S. M.; Shen, W.; Carlson, G.; Foster, J. S.; Babcock, K.; Manalis, S. R. Nature 2007, 446, 1066−1069. (23) Godin, M.; Bryan, A. K.; Burg, T. P.; Babcock, K.; Manalis, S. R. Appl. Phys. Lett. 2007, 91, 1231211−1231213. (24) Wang, Y.; Lowe, R. D.; Mejia, Y. X.; Feindt, H.; Steltenkamp, S.; Burg, T. P. Biomicrofluidics 2013, 7, 0265031−02650311. (25) Frokjaer, S.; Otzen, D. E. Nat. Rev. Drug Discovery 2005, 4, 298− 306. (26) Mohanty, J.; Choudhury, S. D.; Pal, H.; Bhasikuttan, A. C. Chem. Commun. 2012, 48, 2403−2405. (27) Modena, M. M.; Wang, Y.; Riedel, D.; Burg, T. P. Lab Chip 2014, 14, 342−350. (28) Chatani, E.; Lee, Y. H.; Yagi, H.; Yoshimura, Y.; Naiki, H.; Goto, Y. Proc. Natl. Acad. Sci. U.S.A. 2009, 106, 11119−11124. (29) Jansen, R.; Dzwolak, W.; Winter, R. Biophys. J. 2005, 88, 1344− 1353. (30) Sipe, J. D.; Benson, M. D.; Buxbaum, J. N.; Ikeda, S.; Merlini, G.; Saraiva, M. J. M.; Westermark, P. Amyloid 2010, 17, 101−104. (31) Knowles, T. P. J.; Shu, W. M.; Devlin, G. L.; Meehan, S.; Auer, S.; Dobson, C. M.; Welland, M. E. Proc. Natl. Acad. Sci. U.S.A. 2007, 104, 10016−10021. H

DOI: 10.1021/ac503845f Anal. Chem. XXXX, XXX, XXX−XXX