Article pubs.acs.org/ac
Automated Circular Dichroism Spectroscopy for MediumThroughput Analysis of Protein Conformation Sebastian Fiedler,† Lindsay Cole,‡ and Sandro Keller*,† †
Molecular Biophysics, University of Kaiserslautern, Erwin-Schrödinger-Str. 13, 67663 Kaiserslautern, Germany Applied Photophysics Ltd., 21 Mole Business Park, Leatherhead, Surrey KT22 7BA, Leatherhead, United Kingdom
‡
ABSTRACT: Circular dichroism (CD) spectroscopy is a powerful method for monitoring conformational changes of biomolecules. For peptides and proteins, it is highly sensitive to changes in secondary structure, which may be caused by alterations in amino acid composition or solution conditions (e.g., temperature, pH, salts, detergents, denaturants, and excipients), post-translational modifications, self-association, or ligand binding. The assets of CD spectroscopy are that the signal is directly linked to structure, the analyte is measured without labels and in solution, the technique requires low sample amounts, and data analysis is straightforward. However, CD spectroscopy has remained a low-throughput method because it imposes high requirements on the optical quality of sample cells and thus cannot be performed in microplate-reader format. Here, we introduce an automated CD spectrometer equipped with a low-birefringence flow-through cell that is coupled to a three-axis robotic liquid-handling system. This enables unattended CD measurements on up to 384 samples, including sample transfer from 96-well plates into the flow-through cell, data acquisition, and cell cleaning. We show that the accuracy, precision, and reproducibility afforded by the new instrument are excellent and exemplify how the advantages offered by automated CD spectroscopy can be exploited to quantify protein stability by titration with chemical denaturants.
C
as the absence of birefringence resulting from mechanical strain, do not allow measurements in a microplate-reader format. In addition, sample cells have to be cleaned and dried thoroughly, rendering CD a rather time-consuming method that requires human attendance. Multicell holders accommodating up to eight samples represent a first step toward speeding up and automating the technique to some extent,5 and automated titrators that mix two solutions inside the cell are available for commercial CD spectrometers.8 To this date, however, no CD system has been reported that could substantially increase the throughput by automated measurement of more than a few independent samples. Herein, we describe an automated CD (ACD) setup consisting of a liquid-handling robot interfaced with a CD spectrometer equipped with a flow-through cell. This allows unattended, serial data acquisition on up to 384 samples by analyte transfer from 96-well plates into the flow-through cell and automated cell cleaning. We determine the accuracy, precision, and reproducibility of ACD as well as the optimal protein concentration range and demonstrate how it can be applied to the quantification of protein stability using, as an
ircular dichroism (CD) spectroscopy provides lowresolution information on the secondary and tertiary structure of biological macromolecules.1,2 The technique measures the difference in absorbance between left- and right-handed circularly polarized light by a chiral chromophore and has been extensively applied in peptide and protein research,3 taking advantage of the n−π* and π−π* transitions of the amides in the polypeptide backbone, which are locally planar but coupled in a chiral manner. CD measurements are performed on label-free samples in solution, require low sample amounts, and data analysis is straightforward. Although it is often used to estimate the absolute secondary-structure content of peptides and proteins, the primary strength of CD spectroscopy lies in its high sensitivity toward relative conformational changes in response to alterations in solution conditions such as temperature, pH, and solvent composition. In a series of experimental protocols, Greenfield provides an excellent introduction to some common applications of protein CD spectroscopy, including estimation of secondary-structure content,4 analysis of thermal stability and binding interactions,5 study of folding kinetics,6 and titration with chemical denaturants.7 In principle, CD spectroscopy would be perfectly suited for screening applications because it can provide quick yes-or-no answers, for instance, whether a protein is folded under certain conditions.4 Unfortunately, the demanding requirements imposed on the optical properties of CD sample cells, such © 2012 American Chemical Society
Received: November 7, 2012 Accepted: December 19, 2012 Published: December 19, 2012 1868
dx.doi.org/10.1021/ac303244g | Anal. Chem. 2013, 85, 1868−1872
Analytical Chemistry
Article
H2O (1×), and 500 of μL acetone (1×). Data were analyzed according to the linear-extrapolation (LEM)10 to extract the best-fit values of the thermodynamic parameters and the preand post-transition baselines. Error surface projections were constructed to yield the SDs of the adjustable parameters as the half-widths of their 68.3% confidence intervals.11
example, chemical unfolding titrations of the model protein lysozyme.
■
EXPERIMENTAL SECTION Materials. For sample preparation and storage during ACD experiments, we used 96-deep-well plates with a volume of 2 mL/well from Abimed (Langenfeld, Germany). Acetone was from Sigma-Aldrich (Steinheim, Germany) and Decon 90 from Decon Laboratories (Hove, UK). Glycine, guanidinium chloride (GdmCl), sodium dihydrogen phosphate, and sodium hydrogen phosphate were obtained from Carl Roth (Karlsruhe, Germany). Hen egg-white lysozyme (HEWL) was from Carl Roth and Roche Diagnostics (Burgess Hill, UK). Hydrochloric acid and sodium fluoride were purchased from Merck (Darmstadt, Germany). All chemicals were purchased at the highest purity grades available. Methods. CD and Absorbance Spectra of HEWL. The ACD setup combines a robotic liquid-handling system (Tecan Systems, San Jose, CA, USA) with a Chirascan-plus spectropolarimeter (Applied Photophysics, Leatherhead, UK) equipped with a large-area avalanche photodiode detector (Advanced Photonix, Ann Arbor, MI, USA) and a 0.2-mm fused silica flow-through cell (Optiglass, Hainault, UK) having a prime volume of 40 μL. HEWL solutions and buffer blanks (10 mM Na2HPO4/NaH2PO4, 154 mM NaF, pH 7.4) were dispensed into a 96-well plate in an alternating pattern, that is, with buffer in odd-numbered wells and protein solutions in even-numbered wells. Each well contained a total volume of 100 μL. The plate was positioned on a temperature-controlled plate holder and thermostatted at 20 °C. Before each data acquisition step, 40 μL of buffer or protein solution was transferred from a well into the flow-through cell in an automated manner. CD and absorbance spectra were recorded simultaneously at 20 °C from 320 to 180 nm with a wavelength increment of 1 nm, a spectral bandwidth of 1 nm, and a digital integration time (DIT) of 0.25 s. Each sample was scanned 10 times, thus yielding the standard deviation (SD) of the spectroscopic signal for a DIT of 0.25 s, σθ,0.25s. From the latter, the SD for a total DIT of 2.5 s corresponding to a set of 10 scans was calculated as σθ,2.5s = σθ,0.25s/10−0.5. Each protein spectrum was corrected by subtracting the spectrum of the preceding buffer blank and zeroing the resulting intensity averaged over 310−320 nm, which was treated as offset. After data acquisition, the cell and the sample probe were rinsed sequentially with 500 μL of H2O (1×), 500 μL of 10% (v/v) Decon 90 (2×), 500 μL of H2O (1×), and 500 μL of acetone (1×) before drying by vacuum pumping for 1 min. After each rinsing step, the cleaning solution was pipetted up and down twice inside the sample cell before being discarded. Chemical Unfolding of HEWL. A stock solution containing 10 mg/mL of HEWL in 50 mM glycine buffer, pH 2.9, was diluted into buffer A (50 mM glycine, pH 2.9) and buffer B (7 M GdmCl, 50 mM glycine, pH 2.9) to a final protein concentration of 0.5 mg/mL each. GdmCl concentrations were checked by refractometry with an Abbemat 500 digital refractometer (Anton Paar, Ostfildern, Germany).9 Protein solutions A and B were mixed to yield 48 different GdmCl concentrations equally distributed over 0−6 M, and samples were allowed to equilibrate for 16 h at 20 °C. ACD data were obtained and processed as described above, but cleaning of the flow-through cell was more extensive to avoid GdmCl crystallization, comprising sequential flushes with 700 μL of H2O (4×), 500 μL of 10% (v/v) Decon 90 (3×), 500 μL of
■
INSTRUMENT DESIGN The ACD autosampler system consists of a Chirascan-plus spectropolarimeter coupled to a sample chamber with a flowthrough cell and a three-axis robotic liquid-handling system that can access up to four 96-well plates (Figure 1). The plate
Figure 1. Schematic diagram of the autosampler including the fluidics of the sampling, washing, and sample-cell systems. The pipetting robot aspirates samples from 96-well plates and dispenses them through the injection port into the sample cell. This is controlled by a 1-mL syringe pump using water as a reference liquid in the control circuit (green). The unidirectional cleaning flow lines containing water (blue), detergent (yellow), and a water-miscible solvent (gray) for flushing the sample cell are controlled by separate diaphragm pumps. The line common to the control and cleaning circuits is shown in red. During dispensing of sample or cleaning solution by the sample probe, the three-way waste valve is set to the gravity drain waste system. For removal of solution and cell drying, the sample probe is removed from the injection port, and the three-way waste valve is connected to the vacuum waste system.
holders consist of Peltier-controlled aluminum blocks designed to provide reliable temperature control for 2-mL deep-well plates. This allows for both far- and near-ultraviolet (UV) CD measurements, which require sample volumes of ∼100 and ∼900 μL, respectively. The plate holders and the cell cartridge accommodating the low-birefringence fused silica flow-through cell with an optical path length of 0.2, 0.5, or 10 mm can be temperature-controlled independently. The fluidics consists of a sample probe connected through a polytetrafluoroethylene tubing loop and a six-port flow-selection valve to a syringe 1869
dx.doi.org/10.1021/ac303244g | Anal. Chem. 2013, 85, 1868−1872
Analytical Chemistry
Article
which potentially could result from residual protein attached to the walls of the sample cell. The absorbance spectra reveal two peaks at 190 and 220 nm characteristic of peptide backbone absorption (Figure 2A, inset). Signal-to-Noise Ratio and Optimal Protein Concentration. To assess the precision of the data acquisition process, we performed 10 scans on each sample using a DIT of 0.25 s, which is the duration over which the spectroscopic signal is averaged at a given wavelength during one scan. From each set of 10 scans, we determined the average ellipticity, θ, and the associated SD of the spectroscopic signal for a total DIT of 2.5 s, σθ,2.5s (Figure 2B, inset). The latter reflects the precision of the spectroscopic signal itself, that is, in the absence of sample-to-sample variability. At 208 and 222 nm, σθ,2.5s is ≤0.05 m° over the entire protein concentration range tested; at 190 nm, it assumes similar values at low protein concentrations but increases substantially at higher ones. To obtain a measure of the signal-to-noise ratio (S/N), we also calculated the relative SD as the ratio σθ,2.5s/θ (Figure 2B). With decreasing protein concentration, a reduction in CD signal intensity at nearly constant σθ,2.5s leads to poorer S/N values at all three wavelengths, as indicated by a rapid increase in σθ,2.5s/θ below ∼0.4 mg/mL. Conversely, the rise in σθ,2.5s at 190 nm with increasing protein concentration leads to a noticeable deterioration in S/N above ∼1.0 mg/mL Thus, the optimal protein concentration range yielding an excellent S/N over the entire wavelength range used is ∼0.4−1.0 mg/mL, where σθ,2.5s/θ is always ∼0.5%. In general, the acquisition of multiple scans in a stepped-scan mode allows for the determination of the spectroscopic SD at each wavelength, which provides a direct and rigorous approach to spectral quality assessment and statistical weighting in subsequent data analysis (see Denaturant-Induced Protein Unfolding below). Reproducibility. Since cell-cleaning and sample-transfer procedures are potential sources of sample dilution and crosscontamination by residual analyte or cleaning solutions, the sample-to-sample reproducibility is another key parameter in validating an automated method. At each of the above protein concentrations, we measured four independent samples, that is, four pairs of protein solutions and corresponding buffer blanks, and determined the sample-to-sample SD. Figure 3 exemplarily depicts the data obtained at HEWL concentrations of 0.1, 0.5, and 1.2 mg/mL. Above 200 nm, reproducibility is always excellent, as the spectra recorded at any given protein concentration superimpose virtually perfectly with one another. Below 200 nm, this still holds at protein concentrations of 0.1 and 0.5 mg/mL, but reproducibility slightly deteriorates in this wavelength range at the highest protein concentration of 1.2 mg/mL, which correlates with the above-mentioned decrease in S/N under these conditions (Figure 2B, inset). Most importantly, however, it becomes obvious that the sample-to-sample reproducibility (Figure 3B) is virtually as good as the precision of the spectroscopic signal proper (Figure 3A), thus attesting to the robustness of the liquid-handling system and the cell-cleaning and sample-transfer procedures employed. Denaturant-Induced Protein Unfolding. CD spectroscopy is used routinely in chemical unfolding titrations for quantitative analysis of protein stability.13 Such experiments are based on the stepwise addition of a denaturant such as GdmCl to an initially folded protein, leading to a gradual increase in the fraction of unfolded protein. According to the LEM,10 the Gibbs free energies of unfolding measured around the midpoint
pump and three microdiaphragm wash pumps. The syringe pump is for accurate aspiration and dispensing of liquids. Each of the diaphragm pumps draws from a separate reservoir of water, an aqueous detergent solution, or a water-miscible volatile solvent, allowing the sample probe loop to be rapidly primed with any of these liquids. Automated, simultaneous CD and absorbance measurements follow a three-step procedure: First, the robotic system aspirates a sample from a well, moves to the injection port of the cell cartridge, and injects the sample into the flow-through cell. Second, after a defined equilibration period, data acquisition is performed automatically as specified by the user. Third, a programmable cleaning procedure flushes a sequence of cleaning solutionstypically water, detergent, and a water-miscible solvent such as acetone or methanolthrough the flow-through cell. For drying and reconditioning, a vacuum waste system comprising a diaphragm pump draws air across the cell. This sequence can be repeated for up to 384 samples.
■
RESULTS Simultaneous CD and Absorbance Measurements. We simultaneously recorded far-UV CD and absorbance spectra for a series of HEWL samples at a protein concentration of 0.1− 1.2 mg/mL (Figure 2A). All CD spectra display a profile characteristic of HEWL, featuring a negative shoulder at 222 nm, a minimum at 208 nm, and a maximum at 190 nm.12 At 200 nm, the spectra exhibit a well-defined isosbestic point on the x-axis, indicating the absence of cross-contamination,
Figure 2. Fully automated acquisition of CD and absorbance spectra. (A) CD in ellipticity, θ, and absorbance, A (inset). The protein concentration was increased from 0.1 to 1.2 mg/mL in increments of 0.1 mg/mL. Each spectrum is the average of four independent samples, each of which was measured in 10 scans with a DIT of 0.25 s per scan. (B) Relative SD, σθ,2.5s/θ, and absolute SD, σθ,2.5s (inset), as functions of HEWL concentration, cHEWL. 1870
dx.doi.org/10.1021/ac303244g | Anal. Chem. 2013, 85, 1868−1872
Analytical Chemistry
Article
Figure 3. Spectroscopic precision and sample-to-sample reproducibility. (A) Four individual CD spectra at each HEWL concentration (solid lines) with SDs of the spectroscopic signal, σθ,2.5s (error bars; cf. Figure 2B, inset). (B) Average CD spectra obtained from four independent samples at each HEWL concentration (solid lines) with sample-to-sample SDs (error bars).
Figure 4. GdmCl-induced unfolding of HEWL monitored by ACD. (A) CD spectra of 1.0 mg/mL HEWL in the presence of 0−6 M GdmCl and absolute SD, σθ,2.5s (inset), at 0 and 6 M GdmCl. (B) Unfolding isotherms depicting θ versus GdmCl concentration, cGdmCl. For clarity, only isotherms at 215, 218, 221, 224, 227, and 230 nm are shown. Experimental data (open circles; spectroscopic SDs are much smaller than symbols) and fits (solid lines) according to the LEM10 with ΔG°(H2O) and m as global fitting parameters.
■
of the unfolding transition, ΔG°(denat), can be extrapolated linearly to yield the Gibbs free energy in the absence of denaturant, ΔG°(H2O), and the so-called m value as the slope of this linear relation. Gradual unfolding of HEWL induced by increasing the GdmCl concentration from 0 to 6 M is reflected in a strong decrease in the absolute ellipticity below 240 nm (Figure 4A). The most dramatic changes occur at 215−230 nm, where the pronounced CD signatures of native HEWL give way to weak signals characteristic of an unordered polypeptide. With σθ,2.5s < 0.1 m° at wavelengths ≥213 nm, the spectroscopic SD is very low both in the absence and in the presence of 6 M GdmCl (Figure 4A, inset), which is attributed to the weak background absorbance resulting from the short optical path length of the 0.2-mm flow-through cell. Plotting the ellipticity versus the GdmCl concentration yielded a series of unfolding isotherms (Figure 4B). We globally fitted 16 isotherms at 215−230 nm, thereby using σθ,2.5s for statistical weighting and allowing each isotherm its own baselines while restricting the thermodynamic parameters to be uniform across the entire data set. This yielded ΔG°(H2O) = 32.8 kJ/mol and m = −10.8 kJ/(mol M), which are in good agreement with published values.10 With the aid of error surface projections,11 we found the SDs to amount to 2.5 kJ/mol for ΔG°(H2O) and 0.8 kJ/(mol M) for m, that is, 75% of this time is spent on cell cleaning. Thus, while ACD is still slower than dedicated and less demanding high-throughput methods, automation brings about a drastic productivity enhancement without sacrificing data quality. ACD holds great promise for many applications involving medium-throughput assessment of protein conformation, as exemplified above for the determination of protein stability by denaturant titration. Here, a major difficulty is the pronounced far-UV absorbance of denaturants.13 The traditional solution consists in the use of demountable cells with optical path lengths
