Introduction

Adsorption of proteins at solid-liquid interfaces represents an important process that has significant applications in both nature and industry. For example, many biochemical processes within living systems rely on the adsorption of proteins to lipid layers,¹ several bioseparation procedures utilize the specific adsorption of proteins to a stationary phase,² and the fabrication and operation of biodevices such as biosensors requires bioactive protein molecules on a surface.³ Of particular interest is the controlled, selective, rapid, durable, and nondenaturing attachment of proteins, particularly enzymes, to a surface. Specifically, if the properties of the fully functional protein can be maintained, then the proteins may be used to perform analogous tasks to those they do in nature-catalyze difficult transformations, transfer electrons, sense specific molecules, harvest energy, and a host of other functions.

The quartz crystal microbalance (QCM) has recently been used to examine cell and protein adsorption onto a variety of modified and unmodified solid surfaces.^4-15 These studies have determined the amount and kinetics of adsorption as well as provided information relating to protein surface charge, specific and nonspecific interactions, and structural changes associated with the adsorption process. Changes in the fundamental resonant or overtone frequency, f, of an oscillating surface (typically a gold coated quartz crystal) can be converted into a mass change, m, having the units of ng·cm^-2, via the Sauerbrey equation¹⁶

Recently, a quartz crystal microbalance-dissipation or QCM-D instrument was designed which can measure the change in the energy dissipation factor, D, while simultaneously measuring f.¹⁸ The energy dissipation factor, D, can be simply defined as shown in eq 3

Azurin (14.6 kDa), the protein of interest in the present study, is responsible for electron transfer in the respiratory system of several bacteria.²⁰ It contains a single copper atom, bound by two histidine (N), a cysteine (S), methionine (S), and glycine (O) residues, that undergoes a one-electron redox process, converting it between the Cu^II and the Cu^I redox states.²¹ Azurin is believed to interact with its redox partners via an area of hydrophobic residues located close to the copper site.²² There are numerous reports on the exploitation of azurin adsorption onto solid surfaces, particularly on gold electrodes modified with alkanethiol self-assembled monolayers (SAMs).^23-27 Figure 1 provides a simplified model of an idealized azurin monolayer formed after adsorption onto an alkanethiol SAM on gold. The majority of these investigations have been for the purpose of establishing its electron-transfer properties. The adsorbed azurin system is now regarded as a model of an essentially ideal surface-confined electron-transfer system. However, little information is available on the fundamental adsorption process. In order to address this issue we now provide the first electrochemical QCM-D study of azurin adsorption onto a SAM of octanethiol-modifed gold.

Figure 1 Simplified model (not to scale) of azurin adsorbed onto an alkanethiol-modified gold electrode. The molecules are shown in their idealized orientation based on the expected interactions between the hydrophobic surface residues of azurin (red) and the alkane SAM.

Experimental Section

Chemicals. All chemicals were commercially available and used as supplied, including 1-octanethiol (98%, Lancaster Synthesis), ethanol (Merck), sodium chloride (Fluka), and isopropanol and sodium acetate (Sigma). A stock solution of purified Pseudomonas aeruginosa azurin (0.93 mM) was diluted into acetate buffer (0.02 M, pH 4.0, 0.1 M NaCl) to the desired concentration. Ultrapure water (18.2 M) was used to prepare all solutions (Sartorius Arium 611 system).

QCM. QCM measurements were performed with a Q-sense E4 instrument (Q-Sense, Sweden) using gold 'chips' that consisted of a thin quartz crystal coated with a gold layer (Q-Sense, f = 5 MHz, C = 17.7 ng·cm^-2·Hz^-1 geometric area of gold working surface = 0.78 cm²). The gold chips also served as the working electrode in the electrochemical measurements. The chips were housed in a specially designed temperature-controlled cell that served as a chip holder and as an electrochemical cell (volume 0.7 mL; Q-sense; Echem module, QEC401, Sweden). All experiments were performed at a temperature of 22.00 ± 0.05 C. Prior to each experiment, chips were cleaned in a NH₄OH (28%):H₂O₂ (30%):H₂O (1:1:3 v/v/v) mixture at 70 C for 15 min, washed several times with water and ethanol, and dried under a stream of nitrogen. The cleaned chip was immediately immersed into an octanethiol solution (1 mM in isopropanol), and the SAM developed on the gold surface over a period of at least 12 h. The chip was rinsed with isopropanol to remove any physisorbed thiol molecules, dried under a nitrogen stream, and then immediately positioned and sealed in the QCM cell. A peristaltic pump was used to introduce solutions into the cell. Following inclusion of the sample into the measurement chamber, the f and D changes associated with four overtone frequencies (15, 25, 35, and 45 MHz) were recorded. These frequencies correlate with the third, fifth, seventh, and ninth harmonics. The f and D changes associated with the fundamental frequency (5 MHz) were sensitive to changes in bulk solution. The Q-sense E4 instrument provides the overtone-normalized f data, that is f/n, and this is the value that is generally reported throughout this paper. This allows us to compare directly the frequency change for each harmonic. To help distinguish this, we refer to the normalized f data as f_N. The stability of the QCM response was first tested in pure water, and if stable after 10 min, the cell contents were exchanged with buffered electrolyte solution. Again, a period of time (10-15 min) was allowed to check for stability in the f_N and D responses. Following this, approximately 1 mL of azurin solution was slowly pumped (50 L/min) into the cell and f_N and D were monitored with time. After a maximum f_N value had been reached (5-40 min) the flow was stopped and the response monitored for at least 15 min. Buffered electrolyte solution containing no azurin was then pumped (300 L/min) into the cell for 5 min to remove unadsorbed or weakly bound azurin. The flow was stopped and f_N and D were monitored with time. The frequency change observed at this stage was designated to be f_max, the maximum change in frequency.

Electrochemistry. The QCM cell was coupled to a -Autolab (EcoChemie, Utrecht, Netherlands) or an Epsilon (BAS, West Lafayette, USA) electrochemical workstation. Cyclic voltammetry experiments were performed using a conventional three-electrode configuration. The gold chip was used as the working electrode, a platinum disc served as the counter electrode, and a "no leak" Ag/AgCl (3 M KCl) electrode (Cypress Systems, Chelmsford, MA) was used as the reference electrode. All potential values are quoted with respect to the standard hydrogen electrode (SHE), where SHE is 0.207 V versus Ag/AgCl (3 M KCl) at 22 C. The working electrode area was calculated electrochemically to be 0.87 cm² (geometric area = 0.78 cm²) based on the peak current obtained from a 0.75 mM potassium ferricyanide solution and applying the Randles-Sevick equation.²⁸ Voltammetric scans were performed in 0.1 M NaCl solution containing 0.02 M acetate buffered at pH 4.0. The potential was cycled between 0.4 and -0.1 V vs Ag/AgCl (3 M KCl). The scan rate ranged from 0.01 to 100 V·s^-1. The uncompensated cell resistance was <100 . All experiments were performed at a temperature of 22.00 ± 0.05 C.

Results

QCM. Figure 2 shows the typical change in f_N and D observed during a QCM-D experiment for the third, fifth, seventh, and ninth overtones. Within a few minutes of commencing the flow (50 L/min) of buffered azurin solution into the cell, a decrease in f_N of the order of 20 Hz was observed. After stopping the flow of buffered azurin solution, f_N did not change significantly over a period of 15-30 min. Buffered electrolyte solution containing no azurin was then pumped (300 L/min) into the cell for 5 min to remove unadsorbed or weakly bound azurin, which resulted in a f_N increase of ~4-5 Hz. This frequency increase was related to the increased flow rate used to purge the cell of solution-phase azurin. On stopping the flow of buffered solution into the cell, f_N slowly decreased over a period of around 5 min to its previous level prior to the high flow rinse. The value of f_N observed at this stage (t 50 min) was designated to be f_max, the maximum change in frequency. The average f_max value for all concentrations studied was -20.5 ± 1.0 Hz. The surface coverage calculated via use of eqs 1 and 2 was 25.1 ± 1.1 pmol·cm^-2 (see discussion below).

Figure 2 Change in (a) fN and (b) D observed for the third (), fifth (·), seventh (), and ninth () overtones during a QCM experiment when azurin is adsorbed onto an octanethiol SAM on gold. The timing of flow and solution changes is indicated with dotted lines. The plots shown are from an experiment using an azurin concentration of 0.35 M.

During the initially rapid change in frequency (5-20 min) or after f_max had been reached, very little change in dissipation (<0.1 × 10^-6) was observed. Small changes in D were observed for each overtone frequency during the high-flow rinse stage (35-40 min), the magnitude and direction of which were dependent on the overtone.

The solution concentration of azurin employed during the QCM measurement was varied between 0.035 and 5.5 M. Figure 3 shows the change in f_N prior to the high-flow rinse step for three different azurin concentrations under the same flow conditions (50 L/min). Only data obtained from the ninth overtone are shown in Figure 3 as the other overtones showed similar profiles. Each QCM profile could be characterized by an initial linear change in frequency, f_init, with values of 1.6, 3.5, and 10.7 Hz/min for the 0.035, 0.35, and 2.75 M solutions, respectively. Following this, the decrease in f_N was approximately exponential with an apparent time constant related to the azurin concentration. Interestingly, and importantly, f_N decayed to a similar value for each concentration of azurin studied.

Figure 3 Change in fN observed prior to the high flow rinse step for azurin concentrations of () 0.035, (·) 0.35, and () 2.75 M. Data shown are from the ninth overtone frequency with a flow rate of 50 L/min.

Voltammetry. dc cyclic voltammetric studies were undertaken after the f_max value had been obtained in solutions containing 0.1 M NaCl and 0.02 M sodium acetate buffered at pH 4.0. The voltammetric data, as was the case with QCM-D data, confirmed the presence of azurin adsorbed onto the 1-octanethiol-modified gold electrodes. Figure 4 shows a dc cyclic voltammogram for azurin bound to a 1-octanethiol-modified gold electrode. The reversible potential value, E^', calculated as (E_p^ox + E_p^red)/2, where E_p^ox and E_p^red are the oxidation and reduction peak potentials, respectively, was 0.360 ± 0.005 V vs SHE. This E^' value was in reasonable agreement with values of 0.342 ± 0.005 and 0.368 ± 0.003 V previously reported for azurin adsorbed at similar SAM electrodes.^24,25 The average peak-to-peak separation, E_p, calculated as E_p^ox - E_p^red was 15 ± 2 mV, and the peak width at half-height, _1/2, was 95 ± 5 mV at a scan rate of 0.02 V·s^-1. These values were slightly larger than the Nernstian values of 0 and 89.6 mV, respectively, predicted for a one-electron reversible process at 22 C.²⁹ The oxidation and reduction peak current values, I_p^ox and I_p^red, varied linearly with scan rate, commensurate with an adsorbed redox process. The interfacial electron-transfer rate constant, k^0', was estimated to be 550 s^-1 using the scan rate dependence (1-100 V·s^-1) of E_p and use of Laviron's theory.^29,30 This was consistent with k^0' values obtained on similarly modified surfaces.^26,27,31

Figure 4 Cyclic voltammogram of azurin adsorbed onto an octanethiol-modified gold electrode. The voltammogram was measured at a scan rate of 0.02 V·s-1 in a solution containing 0.1 M NaCl and 0.02 M sodium acetate buffered at pH 4.0.

Importantly, the voltammetric data can also be used to obtain a comparative measure of the azurin surface coverage. The surface coverage of electrochemically active azurin, _echem, was determined by integration of the area under the reduction peak (charge consumed) from a series of slow-scan (0.02 V s^-l) cyclic voltammograms using the relationship

Discussion

It is clear from both the QCM and the voltammetric data that azurin adsorbs onto the octanethiol-modified gold surface and in such a manner that fast electron transfer between the modified electrode and at least some of the protein is possible. According to the E-QCM data, the kinetics of the adsorption process is strongly dependent on the concentration of azurin in solution. For example, the adsorption maximum for a solution of 2.75 M azurin was reached within 2 min compared with around 20 min for 0.35 M. Notably, and importantly, the magnitude of the adsorption maximum was always the same with little variation observed over 3-4 h of constant QCM measurement. This suggests that once a monolayer of azurin is formed, no further significant adsorption takes place. Chi et al.²⁶ measured a sigmoidal-type adsorption isotherm for azurin over the concentration range from ~10 to 300 mM that showed that saturation coverage occurs from a concentration of about 150 M. This is a significantly higher concentration needed for monolayer formation than on the basis of our QCM results. These authors also indicated that at least 4 h was required to achieve saturation adsorption.²⁴ One possible factor contributing to their higher values needed for monolayer coverage is that the adsorption isotherm was based on voltammetric data (see below), which reflects only azurin adsorbed in an orientation suitable for rapid electron transfer.

The very low change in dissipation observed in the QCM experiments indicates that the adsorbed azurin layer is rigid and not easily deformed by the shear stress induced by the oscillating crystal or flow of solution.⁵ We hypothesize that azurin is strongly coupled to the electrode, primarily via interactions between a patch of hydrophobic residues at the surface of the protein and the alkyl groups at the modified electrode surface.²³ The rigidity of the adsorbed azurin layer provides confidence that the Sauerbrey equation will be valid in this case. If we therefore assume that the adsorption of azurin is solely responsible for the change in frequency, then by using eq 1 this equates to an average adsorbed mass of 366 ± 16 ng·cm^-2. Furthermore, using eq 2 we can then further calculate the surface concentration of azurin, _QCM, which in this case was 25.1 ± 1.1 pmol·cm^-2. This is significantly larger than the _echem (7 pmol·cm^-2) value calculated electrochemically. An analogous outcome was reported for cytochrome c self-assembly onto 11-mercaptoundecanoic acid monolayer electrodes.¹⁵ Clearly, reconciliation of the differences between these values needs to be made.

Only molecules that are electrochemically addressable within the time scale of the cyclic voltammetric measurement contribute to the _echem value. Therefore, azurin molecules adsorbed on the SAM in orientations that do not result in a significant rate of electron transfer between electrode and protein will be excluded from this value. Any denatured or unfolded protein that adsorbs on the SAM would also not contribute to the _echem value. On the basis of its crystal structure, one molecule of azurin has the approximate dimensions of 4.5 nm high and 2.5 × 3.0 nm wide.²⁰ If all molecules adsorbed onto the SAM in the same orientation, as indicated in Figure 1, then the monolayer surface coverage would be ~25 pmol·cm^-2 (assuming each molecule is a sphere of radius 1.5 nm). Binding to the electrode via the hydrophobic patch of residues near the copper atom would certainly be one favored orientation environment. Dense packing on the SAM could be expected as the high electrolyte concentration (0.1 M NaCl) would help to reduce the repulsive interactions between azurin molecules. However, saturation monolayer coverage of molecules in similar orientations would be unlikely given the surface heterogeneity and roughness typically observed with SAM-modified electrode surfaces and other less favored hydrophobic binding regions on the surface of the protein.

The frequency changes observed in QCM solution measurements can also reflect the added mass of trapped water and other molecules (counterions from the electrolyte and/or buffer) associated with the adsorbed protein layer.^4,5,8 This means that using the absolute value of f_N measured by QCM would produce an overestimate of the adsorbed mass of azurin. Given the strongly hydrophobic nature of the octanethiol-modified surface, wetting and water association at the SAM surface would be minimal. However, azurin adsorption would help reduce the interfacial tension, and as such make the surface more amenable to water being present, be that entrapped water between proteins or associated with the protein water shell. An accurate measure of the water contribution to the f_N value is difficult to obtain. A conservative estimate of the water associated with the protein shell can be deduced from examination of the crystal structure. For azurin from P. aeruginosa, approximately 80 water molecules were incorporated in the crystal. This equates to at least a 10% increase in mass on top of the total mass of amino acid groups in the azurin sequence. Höök et al. reported that ~94% of the mass associated with mussel adhesive protein adsorbed at a methyl-terminated gold surface was due to water.⁸ For the mass measured by QCM for hemoglobin adsorption, 30% was estimated to be due to protein-bound water.⁶ In their study of ferritin adsorption on gold, Caruso et al.⁴ compared the frequency changes between dried surfaces-before and after protein adsorption. This approach indicated that water could account for about 50% of the frequency change. In our case, the current experimental design requires the SAM-modified gold chip to be removed from the cell in order for it to be carefully and completely dried, a process which itself can impart significant errors in the final measurement of the frequency difference.

Conclusions

Coupling data from QCM-D and electrochemical measurements provides a powerful approach to the analysis of adsorption processes involving redox proteins. We applied such an approach, for the first time, to azurin adsorption onto an alkanethiol SAM. As expected, both techniques confirmed that azurin adsorbed irreversibly onto a gold surface modified with an octanethiol SAM with a significant fraction, but not all, of the adsorbed azurin remaining electrochemically active. The QCM data revealed that the kinetics of azurin adsorption onto such electrodes could be controlled by the solution concentration of azurin. With only modest azurin concentrations (~micromolar) the saturation coverage was attained quite rapidly (<3 min). This has implications for experiments where an adsorbed layer of protein is required. Typically, azurin concentrations ranging from 1 to 300 M and adsorption times from 1 to 16 h have been reported.^23-26 Our results suggest that such long time frames need not be necessarily required.

Acknowledgment

Financial support from the Australian Research Council is gratefully acknowledged.