The water contact angle at the SH surface, determined using 1-2 μL droplets of distilled water, was 65-70°, in agreement with literature data [28]. The activity of the surface SH groups was determined by exposing the SAM to DTNB (5,5'-dithiobis-(2-nitrobenzoic acid), Elman's reagent). This reaction is rapid and stoichiometric. The same procedure was used to determine the activity of the SH groups in the protein. In both cases the appearance of a yellow color indicated that the SH groups were active and capable of forming S-S bonds. It was found that as-prepared SAMs were quickly oxidized (probably by ambient oxygen) to form surface S-S bonds. This oxidation is reversible and the disulfides could be reduced back to surface SH groups with the use of TCEP.
Curve fitting in XPS measurements is a well-established and reliable procedure that uses a Gaussian-Lorentzian function, and provides an accuracy of two decimal digits [29]. The fitting was performed using 5600 Multi-technique system software (PHI, USA). The accuracy of fitting depends on the signal-to noise ratio for the measured curve, and in the present studies the S2p spectrum was rather noisy. Therefore the calculated areas are provided with accuracy up to an integer number: AS-Au = 43%, AC-SH = 39%, AS-O = 18%, (Figure 3). The S-Au bonds ( AS-Au = 43%) are at Au/dithiol interface, and their S2p signal is measured after attenuation because of the SAM layer. On the other hand, the ω-SH group is at the SAM surface; hence, there is no attenuation of the S2p electrons.
To address the concentration of loops, the real quantity of the S-Au bonding should first be obtained, taking into consideration electron attenuation: L = exp(-d/λ) = 0.719, where L is the attenuation factor for the electrons coming through the monolayer, d is the layer thickness, assumed to be about 11 Ǻ (considering bond lengths and molecular tilt angle), and λ is the inelastic mean-free-path of electrons (33 Ǻ for this kind of a molecule) [30]. The real quantity of the S-Au bonding can be obtained by the equation A(S-Au) real = A(S-Au) measured : L = 43: 0.719 ≈ 60%. This 60% are composed of dithiol molecules, whose other end is SH (AC-SH = 39%) and bridged (B) thiol molecules (or "loops"), with both ends connected to Au. For convenience they will be marked as 2B, accounting for two molecular end-groups connected to Au. We did not use attenuation for the electrons coming from the loops. therefore if A(S-Au) real = S + 2B, then 60 = 39 + 2B, and B = 10.5 ≈ 11%. Thus, about 20% of adsorbed 1,6-hexanedithiol form loops.
X-ray photoelectron spectroscopy (XPS) showed clearly the two types of sulfur atom, one attached to the gold (thiolate) and the other (SH/S-S) at the ω-position for the hexane-1,6-dithiol SAMs (Figure 3) [31, 32]. Notice that alkanethiolates in self-assembled monolayers on gold oxidize in air, in the dark, to form sulfinates and sulfonates and the thiolate gold interface (The S- is more prone to oxidation than the SH), and gives rise to the S-O peak in the XPS spectra (~ 168.5 eV) [33–37].
XPS spectra of the immobilized protein samples revealed significant increase in the carbon content, with the different protein carbons observed (Figure 4).
After the protein was chemically attached to the surface, XPS provided bonding percentage of S-Au and C-SH bonds, as A(S-Au) measured ≈ 35%, and A(C-S-protein) measured ≈ 28%, accordingly. If the attenuation in the dithiol layer for S2p electrons involved in S-Au bonding is taken into consideration, one obtains A(S-Au) real = A(S-Au) measured : L = 35: 0.719 ≈ 49%
There is also attenuation of electrons passing through the protein for both S-Au and C-SH electrons, and assumed to be the same for both of them, because of the large size of the protein, and hence is not considered here. The quantity of the forming loops is unchanged, 11%. The ratio of the A(S-Au) real to A(C-SH) measured for dithiol SAM and for dithiol SAM with protein (including bridged molecules) is 60:39 ≈ 1.54, and 49:28 ≈ 1.75 respectively. The ratio between the two numbers (1.54: 1.75 = 0.88) suggests that protein binding is accompanies by desorption of ~12% of 1,6-hexanedithiol molecules from the SAM in SAM. This loss might be understood if one assumes that the immobilized proteins are in a brush-like structure. Loading of more protein molecules at the surface results in less space available for each protein and thus the stretching of its chain to decrease intermolecular repulsion. such stretching can eventually result in breaking of the Au-S bond, which is relatively weak (~44 Kcal/mol) [38, 39]. In fact, the same desorption phenomenon was observed for polystyrene brushes formed by surface-initiated living anionic polymerization of styrene using rigid lithiated biphenyl SAM surfaces as initiator. There, surface morphology studies by AFM showed holes in the brush and the desorbed chains accumulated on their edges [40].
Enzymatic activity tests for the immobilized Adenylate kinase were performed as described by Valero and coworkers [23]. Figure 5 shows a schematic diagram with a step-by-step graphical representation of the grafting of Adenylate kinase on the SAM surface and the study of its activity. Briefly, it consists of preparing the reaction medium from NADH, potassium acetate, MgCl2, AMP (previously treated with apyrase) [2], PEP(phospho-enol-pyruvate), pyruvate kinase, and L-lactate dehydrogenase imidazole/acetic acid buffer (pH 7.5). Immediately after the immobilized protein sample was added, the reaction was initiated by the addition of ATP and the kinetics was followed by measuring the disappearance of NADH at 37°C, following the 340 nm band in the absorption spectrum. This band results from a number of conjugated processes described in detail in the literature [23]. Figure 6 shows a typical kinetics curve.
The distribution of slopes of the linear portions of the kinetic curves of substrate disappearance, which provides information on Adenylate-kinase activity, ranged from -0.0732 to - 0.1198 (arbitrary units), with an average of -0.097. One of the samples was rechecked in a different solution, to ensure that the observed activity was from immobilized Adenylate kinase only and not from proteins that desorbed from the SAM surface. The fitted results of the samples in the first (-0.1198), and second (-0.1130) tests are well within experimental error. These numbers are averages, obtained from the kinetics slopes of at least five independent experiments. The enzymatic activity of 52 nanomoles of protein dissolved in TRIS buffer solution was measured, and the slope obtained was -0.1370. The method described by Valero [26] was used many times by the Haas group with excellent reproducibility. When experiments were carried out, in which one of the components was missing, no catalytic reaction was observed. This confirms that no loss of specificity results from immobilization on the surface.
The theoretical maximum number of protein molecules immobilized is about 2.3 × 10-8 mole/sample if a hexagonal arrangement of molecules is assumed, and 1.62 × 10-10 for cubic arrangement. It was suggested by Patolsky [41], Granot [42] and Willner [43], that globular proteins tend to pack in homogeneous cubic-like form. However, the actual form of packing is not critical, since the controlling factors are the number of immobilized molecules, and the significant free space around the immobilized protein molecules, as is discussed below.
To prove that the observed activity came exclusively from immobilized protein, and to obtain an experimental value for the amount of protein on the surface, a series of QCM experiments were performed.
The average frequency of the crystal that was used, at ambient temperatures, was 8.9869614 MHz, and the addition of hexane-1,6-dithiol to the crystal resulted in a measured frequency of 8.9868661 MHz. The change in frequency (Δf) was 95 Hz. Protein attachment resulted in an average frequency of 8.9864262 MHz. Δf was approximately 43.99 Hz. According to the Sauerbrey equation, at the abovementioned basic sensitivity of the crystal, a change in frequency of 1 Hz indicates the addition of a mass of 1 ng/cm2. Since the effective area of crystal is lower than 1 cm2 (0.392 cm2), the addition of the protein mass could be represented as
. If a crystal area of 1 cm2 is assumed, and with a knowledge of the molecular weight of Adenylate kinase - 24000 g/mol - the calculated number of moles for the protein is
moles. Multiplying the calculated number of moles by Avogadro's number gives the number of molecules per cm2, which is approximately 2.82 × 1012. Since the effective area of the crystal is 0.392 cm2, the actual number of protein molecules on the crystal surface was ~ 1.1 × 1012. The total area of the protein molecules was estimated by multiplying previously estimated number of molecules by the reported cross-sectional area of a single protein molecule [19].
The area covered with protein molecules, estimated by dividing the substrate surface area by the reported cross-sectional area of the protein, was approximately 0.2 cm2, which is ≤60% of the crystal surface area and suggests a homogeneous protein monolayer [41–43].
If the molecules are dispersed homogeneously over the surface and cover only maximum 60% of it, 40% of the surface is empty. Since this ratio is 40/60 = 0.67, for every nm2 covered by protein molecules, there is 0.67 nm2 of free surface. Since the protein is 5 nm in diameter [17], it occupies 19.63 nm2 (Figure 7). Therefore, the free area associated with each protein molecule should be, on the average, 19.63 × 0.67 = 13.15 nm2. If one assumes that the protein occupies a circle of diameter 5 nm, concentric with a larger circle, whose area is that occupied by protein, combined with the area of free space around it -- 32.78 nm2 -- then the diameter of the outer circle should be 6.5 nm. This means that the distance between the protein and the boundary of the larger circle is 1.5 nm and the average distance between two protein molecules is 3 nm.
This results in a significant free volume that should allow easy opening of the LID and hence efficient catalysis and high reactivity. In the case presented, the effective coverage is even less than 60%, which increases the average distance between protein molecules. The fact that only about 60% coverage is obtained, might suggest that in addition to repulsion forces between the surface-attached proteins, molecular motion could be an important factor in determining the coverage.
According to the results of QCM experiments, the amount of adsorbed surface protein, which produced activity equal to that of a nanomole of non-immobilized protein, was 10-11 mole. This surprising result is encouraging, and supports our initial hypothesis that, when the system is properly designed, the activity of surface-adsorbed proteins can be higher than that in solution. Indeed, QCM measurements suggest that the activity of an Adenylate kinase protein attached through an S-S bond to a gold surface with the use of a hexane-1,6-dithiol SAM is about 100 times higher than that of the protein molecule in solution. We note that enzymatic activity depends on sample size (the amount of immobilized protein), as shown in Figure 8. Since it is difficult to achieve high precision with such small samples, one must consider up to 10% error in the sample dimension. This means that the error in the calculated amount of the immobilized protein could be up to 10%, and the average observed activity is 90 times or more higher than that of the protein molecule in solution.
While immobilization could increase protein activity[44], the high activity observed in this case proves that the activity of the immobilized protein results from the molecular design of the protein. With the cysteine at the 75 residue forming the S-S bond with the SAM, the protein at the surface is positioned with its active site facing outwards, and hence fully available, in analogy to the Cyclodextrin-SAM system [26]. In addition, the free volume resulting from 60% coverage is advantageous, because it provides the space needed for active-site operation, i.e. for the LID and AMP-binding domain movements. This means that any substrate molecule passing close enough to the active site has a high probability of being converted to product.
Finally, and probably most interesting is that, while attempts to immobilize the mutated protein directly to the gold surface resulted in no activity, connecting it to a hexane-1,6-dithiol SAM, via an S-S linkage, was enough not only to maintain functionality, but also to exhibit very high reactivity. The lack of activity for the direct immobilization on gold was unexpected, since cysteine-containing engineered IgG-binding protein on a gold surface retained the same IgG-binding activity as the native protein [44]. However, in that case, the antigen-binding activity of immobilized antibody molecules on a gold surface was about 4.3 times higher than that of physically-adsorbed antibody molecules. Our conclusion is that if the formation of S-S bonds cannot take place, the mutated Adenylate kinase might be physically adsorbed on the bare gold surface with its active site hindered from contact with the solution.
It is also possible that the immobilized Adenylate kinase forms H-bonds with remaining surface SH groups, which, weak as they may be, affect protein conformation. For example In crystalline 2-mercaptobenzoic acid, the S-H groups were found to form an infinite S-H ⋯ S-H ⋯ S-H hydrogen-bond chain [45].
Finally, it is possible that interactions between the adsorbed protein and the underlying thiolate/Au system affect its structure and reactivity. To understand the mechanism that might be behind such interaction we first point to a study by Miller and Abbot which showed that the hexadecane contact angles on alkanethiolate SAMs on gold, taken in air, are measurably influenced by van der Waals forces that act between the liquid and the metallic substrate (through the SAMs) [46]. This effect decrease with increasing thiolate-alkyl chain length. We used surface-potential measurements to study thiolate SAMs on gold and showed that image charges are formed in the gold because of thiolate adsorption [47]. Thus, it is possible that the direction is space of bond dipoles in the protein is affected by interactions with the underlying image dipole structure.
We recognize that these final arguments require systematic studies, part of which are being conducted right now, but we have decided to discuss those issues with the hope that they will incite more studies, especially because of the importance of immobilizing proteins on magnetic and non-magnetic nanoparticles [7].