A. Surface and solution chemistry

In this article, a phosphate-buffer solution was used with different concentrations of calf serum. The amount of calf serum was varied over one order of magnitude in the electrolyte, but after about half an hour in contact with the titanium surfaces, all concentrations of calf serum gave a similar response in terms of adsorbed weight. Assuming water replacement at the surface and a protein density of 136 ng nm−1 cm−2,1616. I. Frateur, J. Lecoeur, S. Zanna, C.-O. A. Olsson, D. Landolt, and P. Marcus, Electrochim. Acta 52, 7660 (2007). https://doi.org/10.1016/j.electacta.2006.12.060 the adsorption on the rough surface represented more than 50 monolayers of adsorbate. In this study, a comparatively large variation in protein concentration had no marked effect on the amount of adsorbed protein. It does, however, appear to influence the adsorption kinetics with higher concentrations giving faster adsorption. Changes in pH and other critical parameters could alter this behavior.

To somewhat account for differences in surface chemistry between the samples studied, Auger sputter depth profiles were recorded before and after adsorption experiments. The smooth surface had a thinner deposit to allow for acquiring a wider range of harmonics. The deposit was also cleaner in the sense that the rough surface had higher oxygen contents dissolved in the titanium metal, which can take as much as 30% in its metallic state. For most practical applications, the titanium is in its passive state, covered by an anodic oxide film. For the passive state, the oxygen content in the underlying titanium metal should have only a minor influence on the interaction with proteins. The present set of exposures did not leave any trace of change in oxide chemistry following exposure at open circuit potential for an hour.

In the present context, the surface chemistry of the Ti/TiO2 system corresponds to variations in stoichiometry, crystal structure, and oxide film thickness. Titanium is a valve metal. During anodic potential sweeps, it forms an anodic film that grows with the applied potential. Titanium belongs to a subclass of valve metals that allows for growing anodic films with thicknesses up to several tens of nanometers (e.g., Al, Ti, Nb, Zr, and Hf), whereas other valve metals (e.g., Fe, Cr, and Co) only allow for growing films in the ranges of 1–3 nm. One condition for film formation is that the oxide has low or no solubility in the electrolyte. As there is a linear relation between potential and film thickness, the potential at the oxide/electrolyte interface remains quasiconstant during the forward sweep—the major part of the applied potential is accommodated as an electric field within the film. The potential distribution across a titanium oxide/electrolyte interface was studied by Vergé et al. using a rotating ring disk EQCM, which allowed for investigating potential distributions and side reactions.1717. M.-G. Vergé, P. Mettraux, C.-O. A. Olsson, and D. Landolt, J. Electroanal. Chem. 566, 361 (2004). https://doi.org/10.1016/j.jelechem.2003.11.047 For this study, mass and dissipation changes were monitored during a potential sweep, see Fig. 3. The curves were recorded from −0.5 to +0.5 VAg/AgCl. Wider scans occasionally led to deposit decohesion. Within this potential region, no correlation between the applied potential and protein adsorption was observed.Literature data on protein adsorption on metal surfaces with anodic films show some divergence. It is possible to obtain differences in surface chemistry by preoxidizing, using either electrochemical methods or thermal oxidation. Kusakawa et al. compared the response of ZrO2 and TiO2 surfaces to exposure of albumin and fibronectin.1818. Y. Kusakawa, E. Yoshida, and T. Hayakawa, BioMed Res. Int. 2017, e1521593 (2017). https://doi.org/10.1155/2017/1521593 They used 27 MHz as base frequency and observed a frequency decrease of about −600 Hz for the fibronectin on TiO2 and −400 Hz for the same protein on ZrO2. The time scale of about half an hour matches the rates observed in this article. The frequency shifts would correspond to about 14 Hz on a 5 MHz crystal, which is in the same magnitude as observed for smooth quartz crystals, keeping in mind that the experiments in this study were performed at 37 °C, as opposed to 25 °C for the Kusakawa study. The difference in frequency shift between ZrO2 and TiO2 was about a factor of two, i.e., one order of magnitude smaller than the shift between the two types of surfaces in the present study.The limited dependence on surface chemistry is also in accordance with what has already been reported by van de Keere et al., who studied adsorption of lysozymes on thermally evaporated titanium which they tested in a PBS solution. They did not find any influence of the applied potential on adsorption.1919. I. Van De Keere, S. Svedhem, H. Högberg, J. Vereecken, B. Kasemo, and A. Hubin, ACS Appl. Mater. Interfaces 1, 301 (2009). https://doi.org/10.1021/am800029y Their protein adsorption amplitudes were in the same magnitude as those observed for the smooth quartz surface in this study, although their experiments were performed at 22 °C. When studying albumin in lower-conducting electrolytes, Brusatori et al. did find a potential dependence on adsorption using optical waveguide lightmode spectroscopy with a two-electrode system at ambient temperature.2020. M. A. Brusatori, Y. Tie, and P. R. Van Tassel, Langmuir 19, 5089 (2003). https://doi.org/10.1021/la0269558 For the other protein studied—horse heart cytochrome c—there was no apparent correlation between the applied potential and adsorption during an initial phase. The adsorption of albumin was less pronounced in electrolytes with higher conductivity, as used in the present study. A higher ionic strength in the solution is one important factor that reduces migration by screening of electrostatic interactions.

B. Roughness

The surfaces used in this study were titanium deposited on quartz crystals, which were either atomically smooth or with roughness features in the micrometer range. This is rougher than most surfaces used in EQCM studies. The difference between the two surfaces was striking, with some 80 μg cm−2 of adsorbed matter on the rough surface compared to 4 μg cm−2 for the smooth surface, see Fig. 6. It appears that the final amount of organic matter adsorbed is independent of calf serum concentration in the solution. From the graph in Fig. 7, the most rapid adsorption is found for the highest concentration. The base line in PBS with no calf serum added is a flat line, indicating a stable experiment. The adsorption kinetics appear to be rather complex, as already outlined in the review by Rabe et al.11. M. Rabe, D. Verdes, and S. Seeger, Adv. Colloid Interfaces 162, 87 (2011). https://doi.org/10.1016/j.cis.2010.12.007Many types of models for protein adsorption are based on the Langmuir isotherm. In this study, the amount of adsorbed matter is more than one order of magnitude thicker than a monolayer before adsorption abates. This observation rules out most adsorption isotherms. Instead, a possible interpretation of the present experiment could be that proteins start adsorbing at nucleation sites, which are not present on the smooth surface. In Fig. 1, there are local regions with considerably smoother appearance on rough samples. If this region is found in a groove, this could be a site for nucleation. As it is a surface process, it will grow by order r2 until the disks start to touch. Once all nucleation sites are filled, the adsorption process follows a slower growth law. A model that accounts for these basic assumptions is the model first introduced by Kolmogorov in 1937.2121. A. N. Kolmogorov, B Acad. Sci. USSR (CI Sci. Math. Nat.) 1, 355 (1937), see http://mi.mathnet.ru/izv3359. It is also known as Johnson–Mehl–Avrami–Kolmogorov (JMAK) following independent development in North America some years later. It has been reviewed by Fanfoni and Tomellini, who also discussed its application to thin films.2222. M. Fanfoni and M. Tomellini, Riv. Nuovo Cim. 20D, 1171 (1998). https://doi.org/10.1007/BF03185527

For precipitation kinetics on a 2D surface, the projection of the overlayer on the substrate can be considered 2D transformation. In this case, the growth of the adhered proteins is considered three-dimensional, whereas the limiting surface is two-dimensional. As parallel to the three-dimensional case, it corresponds to assembling all nucleation sites on one wall of the volume to be transformed. When they are large enough to interact, the overall growth rate abates.

When precipitates are growing independently of each other in three dimensions, the extended fraction of adsorbed substance (index β), VβeV=π3∂N∂t[∂r∂t]3t4,(2)where N is the number of nucleation sites and size of the precipitate and V is a volume near the interface with an arbitrary thickness. A full list of parameters with explanations can be found in Tables I. For short times, the number of nuclei grows as Nt, and the radius of each precipitate as (∂r/∂t)⋅t; hence the volume grows as t3, and the total volume fraction will grow as t4. As the phase cannot nucleate in areas where adsorption has already taken place, it is useful to adjust the extended volume estimate to get the real volume,Rewriting and introducing Y = Vβ/V gives a total differential expression,which, applying the boundary condition that the radii are 0 at t = 0, can be integrated into −ln(1−Y)=VβeV=π3∂N∂t[∂r∂t]3tn.(5)It is frequently of interest to define the n-value from a linear fit according to ln(−ln[1−Y])=lnK+nln⁡t,(6)where Yβ is the volume fraction of phase β and n is the order of the reaction, t4, for free growing particles in three dimensions.In the present case, Vβ is replaced by mass fraction m/mmax, where the corresponding plot for the highest serum concentration can be found in Fig. 8. In this analysis, it was chosen to ignore the initial phase during the first 100 s of the experiment. Different growth modes can be distinguished; two of which are marked by dashed lines. The n-values for the adsorption process were calculated for all concentrations, and their respective values can be found in Tables II. For an ideally flat surface with a two-dimensional adsorption process, the n-value is 3. For a three-dimensional process, n = 4. There is a difference between the lower concentrations (0.5 and 2 g/l) and the higher levels (8 and 32 g/l). The steepest slopes, i.e., the highest adsorption rates, were concentration dependent, as also seen in the differential adsorption plot in Fig. 7.TABLE II. n-values from Eq. (3) calculated as indicated in Fig. 8 for the rough sample. Two different growth phases were distinguished, a faster around 200 s (ln t ≈ 5–5.5) abating to a slower regime (ln t > 6).Addition to PBS
g/lnln t ≈ 5–5.5ln t ≈ 6.50.53.10.722.31.183.61.1324.50.7

TABLE I. List of parameters.

LatinAWEArea of the working electrode-to-groundcm2AOscOscillating (mass sensing) areacm2Cs38.6 (rough) and 29.2 (smooth) values for nominal surfacesng Hz−1ΔDEnergy dissipation differenceppmΔfmeasMeasured frequency shift, Δf(t = 0) = 0HzΔfviscFrequency shift corrected for viscous loadHzkvisc2.51Hz ppm−1KLinear fit parameter, see Eq. (6)m3 s−4ΔmMass change, calculated from the Sauerbrey equationNgNNumber of nucleation sites—nLinear fit parameter, see Eq. (6)—SaSurface roughness parametermRRadius of the nucleated particlesmTTimesVeβExtended volume of the β matter—VTotal volume where the β is growingm3YVβ/V, volume fraction of β—GreekβPrecipitating (growing/adsorbing) phase—ηlViscosity of the liquidN s cm−2ρlDensity of the liquidg cm−3

The highest n-value measured was 4.5, which could be understood as a fast filling of all available nucleation sites where growth would occur as half-spheres placed on a rough surface with an effective surface area considerably larger than its nominal value. For lower concentrations, the nucleation sites would fill slower. After the initial growth phase, when the nucleation sites are occupied, the growth abates to a slower mode.

The adsorption data appear consistent with a model based on growth limitation by the number of nucleation sites. In this context, the lower adsorption on the smooth surface would be explained by the absence of nucleation sites. The proteins also appear to have weaker bonds to the smooth surface, as indicated by Figs. 4(a) and 4(b), where desorption became dominating after some 1200 s.Figure 7 indicates different stages of adsorption, disregarding the initial incubation phase. As the calf serum contains different proteins, this is consistent with the Vroman effect.2323. L. Vroman, A. Adams, G. Fischer, and P. Munoz, Blood 55, 156 (1980). https://doi.org/10.1182/blood.V55.1.156.156 The calf serum contains both albumin and gamma globulin. This makes a replacement reaction possible where the albumin would be expected to adsorb first. The different concentrations of the proteins in calf serum would contribute to variations in adsorption rates.

The present study has shown the importance of the surface condition for protein adsorption. This has practical implications for the interaction of biomaterials with corporal liquids. The roughness scale on the rough quartz crystal is of the same scale as the surface finish of an implant in the human body. Surface defects in the micrometer range would have a strong impact on the adsorption of organic matter. It also illustrates the difficulty in comparing results from different studies, as it is cumbersome to maintain equal surface conditions between different laboratories. Nevertheless, the EQCM remains a powerful tool for studying adsorption phenomena related to proteins and metal surfaces.