Urszula
Łapińska
a,
Kadi L.
Saar
a,
Emma V.
Yates
a,
Therese W.
Herling
a,
Thomas
Müller
ab,
Pavan K.
Challa
a,
Christopher M.
Dobson
a and
Tuomas P. J.
Knowles
*ac
aDepartment of Chemistry, University of Cambridge, Lensfield Road, Cambridge CB2 1EW, UK. E-mail: tpjk2@cam.ac.uk; Fax: +44 (0)1223 336362; Tel: +44 (0)1223 336344
bFluidic Analytics, Unit 5 Chesterton Mill, French's Road, Cambridge CB4 3NP, UK
cDepartment of Physics, Cavendish Laboratory, 19 J J Thomson Avenue, Cambridge CB3 0HE, UK
First published on 17th August 2017
The isoelectric point (pI) of a protein is a key characteristic that influences its overall electrostatic behaviour. The majority of conventional methods for the determination of the isoelectric point of a molecule rely on the use of spatial gradients in pH, although significant practical challenges are associated with such techniques, notably the difficulty in generating a stable and well controlled pH gradient. Here, we introduce a gradient-free approach, exploiting a microfluidic platform which allows us to perform rapid pH change on chip and probe the electrophoretic mobility of species in a controlled field. In particular, in this approach, the pH of the electrolyte solution is modulated in time rather than in space, as in the case for conventional determinations of the isoelectric point. To demonstrate the general approachability of this platform, we have measured the isoelectric points of representative set of seven proteins, bovine serum albumin, β-lactoglobulin, ribonuclease A, ovalbumin, human transferrin, ubiquitin and myoglobin in microlitre sample volumes. The ability to conduct measurements in free solution thus provides the basis for the rapid determination of isoelectric points of proteins under a wide variety of solution conditions and in small volumes.
A range of conventional methods can be used to determine the isoelectric point of proteins, including isoelectric precipitation,5 techniques based on ion-exchange adsorption,6–8 zeta potential measurements,9–11 capillary electrophoresis12 or a recently developed nanoparticle-based approach.13 However, the predominant way to separate proteins and investigate their pI is isoelectric focusing (IEF), in which a pH gradient is generated across a chamber when an electric field is applied.14 Proteins that are then introduced into the system migrate in the gradient until they reach their pI and start to precipitate.15 A significant challenge underlying this approach is the requirement to generate and maintain a spatially stable pH gradient of a known and well controlled magnitude. One possible way to characterise such pH gradients is the use of a set of molecular markers, such as fluorescently labelled peptides or proteins, of known pI1 to which the pI investigated sample can be added. This general principle of isoelectric focusing (IEF) can be applied using capillaries IEF,1,16 gel slabs IEF17 or macro scale Free Flow Isoelectric Focusing (FFIEF).18
Microfluidic platforms are powerful technologies offering many advantages over bulk measurements, including high resolution, well controlled experimental conditions, low analyte volume requirement, short analysis time and low cost.19–22 One currently used micro-scale approach for determining pI is microfluidic Free-Flow Isoelectric Focusing (μFFIEF),23–25 which is based on the same principle as IEF with a pH gradient in the direction perpendicular to the advective flow in a microchannel.26 The advantages offered by μFFIEF are rapid focusing of protein mixtures and protein complexes, accurate control of laminar flow and negligible Joule heating, although the quantitative control of maintenance and characteristic of spatial pH gradients remains a limitation.27 IEF devices with spatial pH gradients have been created simply by mixing acidic and basic buffers, but this approach can be challenging to implement, because of difficulties in ensuring the stability of the gradient.28 A significant improvement in this context is the invention of a “natural” pH gradient,29 generated by the simultaneous use of several carrier ampholytes (CAs), amphoteric compounds with pI values close to each other. These compounds undergo a differential drift in an applied electric field and form a gradient,30 that is nonlinear.31 Moreover, CAs and often markers can also interact with the protein under investigation and thus affect its pI.16,31 Another potential drawback of CAs-gradients is the cathodic drift that can be due to the electromigration of CAs, electrolyte diffusion or electroosmosis,30,32 and, moreover, the CAs approach is most reliable typically for samples under low salt concentration conditions.32 Despite the limitations of CAs, including their relatively high cost,33 this μFFIEF gradient-model is constantly being improved14,27 and is now commonly used for protein separation as well as the determination of pI values.
An alternative to the CAs approach is the use of an immobilized spatial pH gradient (IPG), in which monomeric buffering species are covalently linked to a polyacrylamide gel, this overcomes many of the drawbacks of the CAs approach, and many successful examples have been demonstrated.32,34 This approach however, requires the casting of polyacrylamide gel with a spatial pH gradient, a process which can be challenging to automate and standarise to achieve a highly controlled linear gradient.30 In addition to the gradient generated by carrier ampholytes or IPGs, there are less common CAs-free methods, such as thermally generated gradients,33,35 electrolysis-induced pH gradients36–38 and the promising technique of gradients created using diffusion potentials.39
Although micron scale approaches for pI determination have great potential, and present many advantages over bulk techniques, the control of spatial pH gradients remains a challenge. To overcome this limitation, we introduce here a gradient-free method based on microfluidic Free-Flow Electrophoresis (μFFE) to determine the pI of a protein. We developed an approach that exploits temporal rather then spatial pH gradients in combination with μFFE as a tool to determine the isoelectric point of proteins. In this approach, the separation of a mixture of molecules is achieved by exploiting the difference in their charge-to-size ratios, which allows us to control the movement of molecules along the main advective direction of a separation chamber while applying an electric field perpendicularly to this direction. The positively charged molecules deflect towards the negative cathode, whereas negatively charged species migrate towards the positive anode.40
In order to measure the deflection in the electric fields, the proteins studied here were initially labelled with ortho-pthalaldehyde (OPA) and imaged using an inverted fluorescence microscope. The analyte solution was introduced into the chip under native conditions (2 mM phosphate buffer pH 7.7) and the pH was only changed once the sample was within the microfluidic device (Fig. 1a); this process avoided the need to modulate the pH of the labelled protein off the chip prior its introduction to the device. To quantify accurately the electrophoretic mobility of the proteins under an applied electric field, we also monitored the DC current through the device, that together with the knowledge of the conductance of the buffer solution, allowed the magnitude of the electric field to be quantified.41 We then tested a series of buffers with different pH values to determine the pI from the dependence of the protein mobility on the pH, this approach has been extensively used in the macro scale.12,42,43 We also investigated the influence of Tween-20, a biocompatible surfactant that aids in maintaining solubility of the protein samples close to their pI values, and also showed the effect of OPA labelling, on the isoelectric point value of proteins.
Fig. 1 (a) Schematic diagram of the experimental setup for μFFE and gradient-free determination of protein isoelectric points. The microfluidic chip was designed such that a protein stream is flanked on either side by streams of buffer with variable pH. In the electrophoresis chamber (length = 5000 μm, height = 25 μm and width = 550 μm), the protein stream is deflected by a transverse electric field generated by lateral electrodes added in a single step during the PDMS chip fabrication.41 (b) Deflection of the BSA stream in a phosphate buffer at pH 7.7 > pI with no voltage applied (top), in a phosphate buffer at pH 7.7 > pI with 4 V applied (middle) and in a phosphate buffer at pH 3.0 < pI with 4 V applied (bottom). |
A dye stock solution was prepared by dissolving OPA and BME in a ratio of 1:1.5 in 50 mM phosphate buffer at pH 8, to obtain final concentrations of 60 mM OPA and 90 mM BME. To prepare the analyte samples 2 mg of protein were dissolved in 990 μL of 2 mM phosphate buffer pH 7.7 (containing 0 v/v%, 0.01 v/v%, 0.1 v/v%, 1 v/v%, 2 v/v% of Tween-20) and then 10 μL of the dye stock solution were added, to give a final protein concentration of 2 mg mL−1. In each case, the protein samples were freshly labelled prior to each experiment. For the rapid 2-point method no Tween-20 was added to the analyte samples, and no OPA was used for the label-free BSA* test. The 2 mM phosphate buffers solutions (pH 3, 3.3, 3.5, 3.8, 5.6, 6.6, 7.7, 8.0, 11.2) were prepared just prior to use by dilution of the 20 mM stocks, prepared in advance. The ionic strength for the 2 mM buffer solutions used during the experiment for pH range (3.0–7.7) was varied between 0.0017–0.0054 M. The pH was measured with the pH meter (Mettler Toledo MP 225, US).
Devices for the measurements of pIs were replicated from the master using a mixture of 1:10 curing agent:polydimethylsiloxane elastomer (PDMS) (Sylgard 184, Dow Corning), degassed in a vacuum desiccator for 30 min and baked at 65 °C for 90 min. Each PDMS device was provided with fluidic access by punching inlets/outlet with a 0.75 mm biopsy puncher (Harris UniCore), washed with isopropanol (IPA) (Sigma Aldrich) and dried with nitrogen gas. The device was then chemically bonded to a microscope glass slide (Thermo Scientific) by activating both surfaces with an oxygen plasma (10 s, 40 mW) (Electronic Diener Femto Plasma Bonder) and baked at 65 °C for 10 min to improve adhesion. To prevent protein molecules sticking to the PDMS surface, a second plasma treatment (500 s, 80 mW) was applied to increase its hydrophylicity after bonding.44
Electrodes were fabricated as reported previously.41 The microfluidic device was heated at 78 °C and alloy wires (51% In, 32.5% Bi, 16.5% Sn) (Indalloy, Conro Electronics) were inserted into the appropriate electrode inlets, and prevented from entering the main channel by carefully designed pillars. The diameter of the pillars was 25 μm and the distance between them was also 25 μm. As a next step the main channel of each device was quickly filled with water to maintain the surface hydrophilicity. Finally, copper wires were soldered to the alloy electrodes and connected to an external power supply.
The total fluid flow (250 μL h−1) through the devices was controlled by applying a reduced pressure via a neMESYS syringe pump (Cetoni) at the outlets, and the electrodes were connected to a power supply via copper wires. After 8 min of equilibration at the standard flow-rate the intensity of the protein beam was constant and deflection measurements were acquired. A range of electric potentials was applied across the device in 0.5 V steps up to a maximum of 4 V. For each voltage interval, three snapshots of the deflected protein beam were taken (at the end of the main channel) and three measurements of the electrical current were made using a digital multimeter. An inverted microscope (Zeiss AxioObserver, Germany) equipped with a UV LED, a filter Chromo 49000 DAPI and a 10× objective was used for all the optical measurements. For label-free measurements, which relied on the intrinsic fluorescence of aromatic amino acids, a custom-built deep UV-fluorescence microscope was used. Pictures of the deflected protein beam were acquired with a camera (Evolve 512) at an exposure time of 500 ms, and the same procedure was repeated for phosphate buffers with different pH values. For each pH value used the deflection measurement was repeated four times in 4 min intervals. To change the buffer the main channel was washed through the outlet with 200 μL of Milli-Q water using the tubing and a 1 mL Norm-Ject (Fisher Scientific) plastic syringe to avoid a significal increase in pressure which can damage the electrodes.
To quantify the deflection of the beam of the protein solution, the optical images acquired were analysed with ImageJ and custom software written in Python. After a series of the deflection measurements for all buffers were recorded with the different pH values, the device cell constant was calibrated for a KCl conductivity standard (Sigma-Aldrich) of 500 μS cm−1 and for all tested buffers.41 Using a conductivity standard and the calculated conductance G, the cell constant K for each device was obtained as K = σ/G. The average cell constant of all the data presented in this paper was (49.6 ± 3.2 cm−1). Recording the current, I, during the deflection measurements and from the conductance value obtained for the specific buffers, the effective voltage was calculated from Ve = I/G. Knowing Ve and the distance between the electrodes (w = 600 μm) allowed the electric field, E = Ve/w, to be calculated. The electrophoretic velocity v was determined by dividing the deflection δ by the residence time t, v = δ/t. In low aspect ratio channels the residence time at the centre of the channel can be approximated as the ratio of the device volume to the flow rate. Finally, to obtain the electrophoretic mobility μ, the migration velocity was divided by the electric field μ = v/E.
Serial dilutions of BSA, BLG, UB, and lysine varying in concentration between 30 μM and 1.8 nM were prepared in 2 mM phosphate buffer solution at pH 8. Then, 2 mg mL−1 of OPA-pre-labelled BSA, BLG and UB samples were prepared and diluted to 0.3 μM, and all the samples, and a background solution consisting of buffer and OPA, were placed in the wells of a half-area non-protein binding microplate (#3881, Corning) in triplicate. Using a CLARIOstar microplate reader (BMG LabTech), 50 μL of labelling cocktail was injected (430 μL s−1) into each well containing 50 μL of unlabelled BSA, BLG, UB and lysine solutions to quantitatively label the samples, and the fluorescence intensity of each sample measured 3 s after dye injection. The fluorescence intensities of the pre-labelled BSA, BLG and UB solutions, corresponding to the “partially labelled” conditions used in this paper, were also measured, and the measurement for BSA repeated after 2 h (the maximum amount of time which passed during sample preparation and measurement for the experiments reported in this paper). The results were analysed using a custom Matlab program.
The deflection was measured four times in each device for each pH value, and was found to increase monotonously for enhancing voltages (Fig. 2a). When the maximum voltage (4 V) used in this study was reached, the highest average deflection (−17.8 ± 1.7 μm) was observed for solutions pH 7.7, while a lower average value (12.4 ± 2.3 μm) was observed for a solution with pH 3.0, which is closer to the pI for BSA. Moreover, deflection data for pH 7.7 were more reproducible than for pH 3. This fact can be explained by observing that BSA was always labelled with OPA off-chip in a buffer at pH 7.7. Hence during on-chip measurements at low pH (3, 3.3, 3.8) the buffer transitioned from pH 7.7 to lower pH at the nozzle, promoting BSA precipitation at pH = pI and the formation of a deposit, which has the propensity to affect the uniformity of the protein beam in the main channel. We explored two different paths (in this section, and in Section 3.2) to address this limiting factor both by modulating the flow rates and through the presence of a bio-compatible surfactant.
The flow rate during the deflection measurements was typically maintained at 250 μL h−1; however, to avoid precipitation at the nozzle, for the low pH solutions, a higher flow rate (500 μL h−1) was applied for the first 4 min of the equilibration step and afterwards decreased to 250 μL h−1 and this procedure led to accurate measurements of the deflection. The beam deflection was found to correlate linearly with the simultaneously measured current41 (Fig. 2b); similarly, the electrophoretic velocity correlates linearly with the electric field (Fig. 2c) for pH 3.0 and 7.7 solutions and for all the other investigated pH values.
In order to estimate the protein isoelectric point several methods have been previously reported.12,42,43,47–49 Here we used linear interpolation. By using this approach, only the two points closest to the pI play a crucial role in analysing its value, although for a protein with unknown pI it is necessary to test a range of pH values. In our case values for the isoelectric point of BSA have been reported, so appropriate buffers were chosen accordingly. The pI obtained, in this way, which was based on the average for three devices, was 4.6 ± 0.1, and was the intersection of the curve and the μ = 0 line (Fig. 2d). This result is consistent with the values previously reported in the literature (4.5–5.1).50–53 This confirms that our technique provides an accurate measurement of the protein's pI.
It is commonly assumed that the presence of Tween-20 does not affect the pIs of proteins.57 Here, with the microfluidic measurement device, we set out to test this hypothesis. Addition of 0.01 v/v% Tween-20 to the BSA solution (Fig. 3e) resulted in a pI value (4.5 ± 0.1), that was essentially identical to that determined for the measurement without Tween-20 (Fig. 2d). Nevertheless, this amount of Tween-20 was not sufficient to avoid the formation of any deposits in the solution. The concentration of the surfactant was increased to 0.1 v/v%, a value commonly used in the literature,23,26,27 and with this amount of Tween-20, we did not observe the formation of any deposits of protein at any pH value (Fig. 3d). We measured an isoelectric point of 4.7 ± 0.1 (Fig. 3f) in accordance with the value obtained without the surfactant. This demonstrates that 0.1 v/v% Tween-20 concentration does not affect the protein pI measurement while avoiding the formation of protein deposit. At the concentration of Tween-20 of 1 v/v% (Fig. 3g) and 2 v/v% (Fig. 3h, inset) the pI changed significantly (5.0 ± 0.1) and (5.1 ± 0.2) respectively in comparison to the one obtained without the surfactant (Fig. 2d). The results, as shown in Fig. 3h, indicate that by increasing the amount of the surfactant Tween-20 in the analyte solution, a slow increase in pI was observed. These data indicate, therefore that the higher concentration of Tween-20 may affect the pI of protein. A possible reason behind this observation could be related to the critical micelle concentration of Tween-20 (CMC = 0.007–0.05 v/v%)55,56 above which surfactant micelles formed,59 and thus non-ionic micelles interact hydrophobically with protein ions in the applied electric field can induce changes in the electrophoretic mobilities57,60 or modulate the deprotonation free energies of acidic residues.
We first determined the number of lysine residues that have reacted with the OPA, under the conditions used in this study. To obtain this information, we compared the fluorescence intensity measured for BSA which had been pre-labelled with OPA using our method, versus using another method46 where lysine residues are quantitatively labelled with OPA as described in the Methods section. From the point (Fig. 4, intersection of the black/grey dashed lines) for quantitatively labelled BSA at 0.3 μM, we know the fluorescence intensity when all lysines react with the dye. The fluorescence intensity of pre-labelled BSA (Fig. 4, dark blue diamond) allowed us to calculate the percentage of lysines which reacted with OPA during our experiment. We showed that only 9 Lys (15%) reacted with OPA. We also showed that the number of labelled Lys did not increase with the time of experiment (Fig. 4, light blue diamond), but even slightly decreased, due to the instability of the isoindole fluorophore formed in this reaction.46 Additionally to confirm those data we repeated the experiment for BLG and UB, which had similar labelling efficiencies, at 10.5% and 10.6% primary amines respectively. It can thus be suggested that in our electrophoretic experiments, OPA labelling does not affect significantly the values of isoelectric points for proteins.
We further explored the generality of this approach by measuring the pIs of β-lactoglobulin, ovalbumin, transferrin, ubiquitin and myoglobin (Fig. 5b). We also examined ribonuclease A, as an example of the protein which is outside the pH range (4–7) (Fig. 5c); by increasing the higher value of pH to 11.2. The obtained result (8.6 ± 0.2) was in accordance with previously reported values (8.6–9.6).53,63,64 Similarly, the values measured for other proteins were in accordance with literature values: BLG 5.1 ± 0.2 (5.1–5.2),7,65,66 OVA 4.9 ± 0.3 (4.6–4.9),67 TF 5.4 ± 0.1 (5.2–5.5),68,69 UB 6.2 ± 0.1 (6.5–6.8)70–72 and MYO 6.5 ± 0.1 (6.8–7.0).73 By plotting the dependence of the literature and the experimental values of pIs (Fig. 5d), we observed a linear correlation with the high coefficient of R2 = 0.98.
This journal is © the Owner Societies 2017 |