Surface Organization and Cooperativity during Nonspecific Protein

J. Phys. Chem. B 2008, 112, 13971–13980

13971

Surface Organization and Cooperativity during Nonspecific Protein Adsorption Events Michael Rabe, Dorinel Verdes, Jan Zimmermann, and Stefan Seeger* Institute of Physical Chemistry, UniVersity of Zurich, Winterthurerstrasse 190, 8057 Zurich, Switzerland ReceiVed: May 22, 2008; ReVised Manuscript ReceiVed: August 6, 2008

Despite many experimental studies on cooperative effects during protein adsorption events, this phenomenon is still poorly characterized and subject of much controversy. In this study, we address the topic of cooperativity using two distinct experimental approaches, namely, kinetic analysis and surface imaging, both based on supercritical angle fluorescence (SAF) microscopy. Several model systems comprising the two proteins BSA and fibrinogen, two different ionic strength conditions and varying pH environments were investigated. The combination of the experimental information obtained from kinetic analysis and from real-time in situ scan images unravel a clear correlation between cooperative adsorption and a heterogeneous protein layer buildup. We propose a mechanistic model of protein adsorption based on an overlap of classical Langmuir-type adsorption on unoccupied surface areas and an additional cooperative adsorption pathway near preadsorbed proteins which is consistent with the experimental observations. Moreover, the growth of two-dimensional surface clusters as an often assumed element of cooperativity could be excluded for the studied systems. The model includes the often observed phenomenon that the adsorption rate decelerates abruptly above a certain coverage limit. Furthermore, the observed evolution of the heterogeneous protein distribution on the surface is in good agreement with the proposed model. Introduction Protein adsorption on solid interfaces is a common event in nature and has promoted research activities in many scientific fields such as microbiology, biomaterials, pharmacology, or nanotechnology.1-4 For more than two decades, comprehensive experimental efforts on interfacial binding phenomena have given rise to evidence that cooperative effects play a central role in many protein adsorption mechanisms.5-9 Specifically, in the field of cell biology, the demand for further investigations on this topic results from the awareness that protein adsorption affects and regulates the activity of biological interfaces and cellular function.10-12 With respect to the surface organization of proteins, cooperativity is typically considered as molecular self-assembly and aggregation. The consequences of protein aggregation can be dramatic for biological systems. It is for instance considered to be the first step in blood coagulation finally triggering thrombosis.13,14 This issue has initiated the search for efficient protein-resistant surfaces as biocompatible or biomimetic materials.15-17 Furthermore, many important human diseases such as Alzheimer’s or Parkinson’s disease, type II diabetes, and many more have been found to be associated with precursor protein aggregates.18,19 Cooperativity and selfassembly of adsorbing proteins has also been suggested to be vital in some cellular systems as it regulates biological activity.6,20 The characteristics of cooperative protein adsorption events are typically increasing adsorption rates in the initial binding kinetics, sigmoidal adsorption isotherms, and molecular selfassembly at the interface. Particularly, the advances in highly accurate and sensitive technologies and the developments of detailed theoretical models have promoted new insights into the molecular events behind cooperative phenomena in protein adsorption processes.5,7,21-24 In one of the most comprehensive * Corresponding author. Fax: +41-(0)44-6356813. Phone: +41-(0)446354451. E-mail: [email protected].

theoretical studies in this field, Minton associates positive cooperativity with the growth of two-dimensional surface clusters.21-23 Indeed, self-association of surface bound proteins has been confirmed mainly through scanning force microscopy images on adsorbed proteins.6,20,25,26 However, this technique does not allow for real time and noninvasive analyses impeding a straightforward and general interpretation of the experimental data. To this end, time and spatially resolved data that allow a tight integration into theoretical models are needed to obtain deeper insights into the relation between molecular adsorption mechanisms and the organization of surface bound proteins. In this work, we present experimental data showing that cooperative adsorption is not necessarily connected to a tight self-assembly of proteins leading to the formation of twodimensional surface clusters. Nevertheless, a clear connection between the cooperative adsorption and the surface organization of the protein layer is shown. Combining spatial information from in situ fluorescence scan images with real-time kinetic analysis, we succeeded to quantitatively describe the adsorption mechanism with a general model. The experimental data are obtained using the noninvasive supercritical angle fluorescence (SAF) microscopy technique27-29 which has proven to represent a highly reliable and accurate tool for in situ protein adsorption studies.30-32 To evaluate the applicability of the proposed mechanism, the model was tested with two proteins (bovine serum albumin, BSA and Fibrinogen, Fib) on a hydrophilic glass surface at different pH and ionic strength conditions. Experimental Methods Preparation of Buffers and Protein Solutions. All adsorption experiments were conducted in citrate buffer which can be readily adjusted to any required pH value in the range of pH 3 to pH 7 since the three pKa values (3.13, 4.76, 6.40) differ by less than two units. A 10.5 or 1.05 g sample of citric acid (Riedel de Haen, Switzerland) was dissolved in 1.0 L of water to obtain a 50 mM and 5 mM buffer solution, respectively. Subsequent

10.1021/jp804532v CCC: $40.75  2008 American Chemical Society Published on Web 10/09/2008

13972 J. Phys. Chem. B, Vol. 112, No. 44, 2008

Figure 1. Surface scan images recorded at different stages during the adsorption of BSA (8 nM) in citrate buffer (pH 3) at 5 mM ionic strength (left column) and 50 mM ionic strength (right column). Red color indicates patches of higher protein densities; blue color indicates regions of lower protein density. The scale of each scan image is adapted to its maximum intensity values. An intensity of 100% refers to the saturation coverage. The profile plots underneath the images show the evolution of the fluorescence intensity along the cross sections marked by the dotted lines.

titration with NaOH (5 N; Fluka, Switzerland) led to the desired pH value. All citrate buffer solutions were filtrated using a 0.22 µm pore size membrane filter (Millipore, Switzerland) before use. Carbonate buffer (50 mM) was prepared by dissolving 4.2 g of NaHCO3 (Fluka, Switzerland) in 1.0 L of water and adjusting the pH to 9.3 with HCl (1 N; Acros, Switzerland). Double distilled water was used for preparing all solutions. The proteins fibrinogen (Fib, F-4753, >61% protein) and bovine serum albumin (BSA, A-4378, >97% protein) were obtained from Sigma (Switzerland) and used as received. DY647-monofunctional N-hydroxysuccinimide ester dye (Dyomics, Germany) was used as fluorescent label for both proteins. The coupling of the dye to the proteins was conducted by adding 1.0 mL of protein dissolved in carbonate buffer, pH 9.3 (20 mg/mL, BSA; 10 mg/mL, Fib) to 0.2 mg of lyophilized DY647-NHS. The mixture was kept at room temperature for about 1 h under slight agitation leading to dye-to-protein ratios of 5.1 (fibrinogen) and 1.05 (BSA) respectively, as determined by UV/ vis spectroscopy. Unreacted dye was separated from labeled proteins via size exclusion chromatography using a Superdex 200 10/300 GL column (Amersham, UK). To allow for quantitative comparisons of the labeled proteins, the final ratios of labeled to unlabeled protein (p*/p ratios) of the stock solutions were adjusted to 1.0 (surface imaging) and to 0.25 or 0.125

Rabe et al.

Figure 2. (A) Adsorption kinetics of BSA at bulk concentrations of 16 nM (O), 8 nM (0), and 4 nM ()) in citrate buffer at pH 3 at low ionic strength (5 mM). The three lower curves refer to identical measurements in which only half as much proteins were labeled as in the upper curves (p*/p ratio ) 12.5%). The solid lines refer to the best fit of the model proposed in this work. (B) Adsorption kinetics of BSA at bulk concentrations of 16 nM (O), 8 nM (0), and 4 nM ()) in citrate buffer at pH 3 at high ionic strength (50 mM). The solid lines refer to the best fit of the model proposed in this work; the dashed lines refer to the best fit of a model based on the work of Minton.21

(kinetic measurements) by dilution with unlabeled proteins. All solutions were stored at 4 °C in the dark. Adsorption Experiments. All adsorption experiments were performed in a flowing cell setup (V ) 0.2 mL) at a pump rate of 0.25 mL/min which is high enough to ensure a quasi-constant bulk concentration near the interface. A further increase of the pump rate to 0.5 mL/min showed no alteration of the measured initial adsorption rates of BSA at any bulk concentration proving that the adsorption rate is not transport limited. Common glass coverslips (Menzel, Germany) were attached to the measuring cell and served as target surface. Before use, they were immersed in 0.2% aqueous deconex solution for several days (Borer Chemie, Switzerland) rendering them very hydrophilic and ensuring high reproducibility. Images of fluorescently labeled proteins were recorded with a custom-made microscope based on super critical angle fluorescence (SAF) detection. The technique is based on the selective and highly sensitive detection of fluorophores near the dielectric interface whereas the fluorescence coming from the bulk solution is efficiently rejected.27-29,33 A scanning microscope table moves the measuring cell over the objective of the microscope and, hence, up to 75 µm × 75 µm areas can be imaged with diffraction limited resolution (see refs 29 and 32 for a detailed description of the optical setup).

Nonspecific Protein Adsorption Events

J. Phys. Chem. B, Vol. 112, No. 44, 2008 13973

Figure 4. Schematic representation of the proposed model. Proteins can either adsorb at any available binding site in an unoccupied region (pathway 1: Langmuir-type adsorption) or in close proximity to preadsorbed proteins (pathway 2: cooperative adsorption). In the cooperative case, proteins are not rejected due to an occupied binding site but are guided to a free area nearby the preadsorbed proteins.

Figure 3. (A) Rate-coverage plots of the adsorption kinetics of BSA at bulk concentrations of 16 nM (O), 8 nM (0), and 4 nM ()) in citrate buffer at pH 3 at low ionic strength (5 mM). The solid lines refer to the best fit of the model proposed in this work. (B) Rate-coverage plots of the adsorption kinetics of BSA at bulk concentrations of 16 nM (O), 8 nM (0), and 4 nM ()) in citrate buffer at pH 3 at high ionic strength (50 mM). The solid lines refer to the best fit of the model proposed in this work; the dashed lines refer to the best fit of a model based on the work of Minton.21

The kinetic measurements were conducted using the same SAF principle but via a simplified optical setup which we refer to as SAF surface sensor.28,30,31 To yield optimal accuracy and reproducibility, the laser focus was enlarged to 60 µm in diameter at a total laser intensity of approximately 5 µW. In this manner, the total fluorescence of a large area was summed up, and inhomogeneities in the protein layer structure were averaged out. A shutter across the laser beam allowed for the collection of the total fluorescence every 60 s over a long period of up to several hours without facing noticeable signal loss due to photodestruction processes. Kinetics were measured with reduced p*/p ratios of 0.25 and 0.125 to avoid self-quenching effects at very high surface coverages. All presented kinetics of BSA adsorption at pH 3 are averaged curves of duplicate measurements. Note that the plots in this paper typically show only every second, third, or fourth data point for the sake of clarity. Results and Discussion Protein Adsorption at pH 3. We investigate the formation of the protein layer structure during the adsorption of BSA on a bare, hydrophilic glass surface in citrate buffer at pH 3. BSA is the most abundant plasma protein, exhibits a heart-like shape,

and has a molecular weight of approximately 67 kDa. It has become one of the most important blocking and stabilization agents in assay technologies which renders it a perfect model protein for the present study. Under the chosen condition, BSA is net positively charged whereas the target glass surface is negatively charged.34 To investigate the effect of salt ions shielding electrostatic interactions, all adsorption experiments at pH 3 were conducted under low (5 mM) and high (50 mM) ionic strength conditions. In situ surface images of fluorescently labeled BSA adsorption were recorded by scanning 4.5 µm × 4.5 µm surface areas using the SAF microscope. The adsorption process was monitored using a constant bulk protein concentration of 8 nM. Figure 1 presents fluorescence intensity images that illustrate the density distribution of a protein covered surface area with increasing adsorption times. The average coverage levels range between 10% and 60% of the saturation limit. Images of BSA adsorbed to a glass surface at low (5 mM, left column) and high (50 mM, right column) ionic strength conditions are shown. The images clearly show that proteins do not adsorb in a uniform layer under these conditions. Instead, an inhomogeneous distribution on the surface comprising regions of high and low protein density is observed. Furthermore, initial patches of high density retain a relatively high protein density throughout the adsorption process; that is, density differences are not balanced out with time. The profile plots underneath the scan images displaying the evolution of the protein density along the marked lines in Figure 1 emphasize this observation. Density maxima at an early stage of the adsorption can still be noticed as maxima at a later stage. As a result of the optical resolution limit of our experimental setup (0.5 µm) the dimensions of the patches of high or low density depicted in Figure 1 are in the range of micrometers and hence accommodate thousands of adsorbed proteins within these areas. Hence, only large-scale density fluctuations can be resolved and observed over time. The appearance of these fluctuations in the density of adsorbed proteins may be due to random processes or surface inhomogeneities inherent to the glass surface35,36 and are not treated explicitly in the framework of this study. However, according to the classical Langmuir theory, such density fluctuations would be balanced out as the probability of adsorption decreases with the increasing surface density of adsorbed proteins. Since the density differences between regions of lower and higher protein density are maintained during the course of adsorption, we conclude that the simple Langmuir model is an incomplete description of the observed adsorption behavior for BSA at pH 3. Accurate quantitative information about the adsorption of BSA were obtained by analyzing adsorption kinetics using the


Rabe et al.

TABLE 1: Estimated Parameters for BSA Adsorption in Citrate Buffer at pH 3 protein conc. (M)

ionic strength (M)

θmax (counts ms-1)

6 kon 1 (10 counts ms-1 M-1 s-1)


(kon 1 + on kon 2 )/k1

R

R2

plot

16 × 10-9 8 × 10-9 4 × 10-9 16 × 10-9 8 × 10-9 4 × 10-9

5 × 10-3 5 × 10-3 5 × 10-3 50 × 10-3 50 × 10-3 50 × 10-3

238 ( 14 238 ( 14 238 ( 14 340 ( 13 340 ( 13 340 ( 13

6.35 ( 0.26 6.35 ( 0.26 6.35 ( 0.26 4.69 ( 0.50 4.69 ( 0.50 4.69 ( 0.50

8.63 ( 0.70 8.63 ( 0.70 8.63 ( 0.70 5.73 ( 0.69 5.73 ( 0.69 5.73 ( 0.69

2.36 2.36 2.36 2.22 2.22 2.22

0.82 ( 0.09 0.82 ( 0.09 0.82 ( 0.09 0.80 ( 0.07 0.80 ( 0.07 0.80 ( 0.07

0.9503 0.9987 0.9985 0.9985 0.9995 0.9987

Figure 2A, 3A Figure 2A, 3A Figure 2A, 3A Figure 2B, 3B Figure 2B, 3B Figure 2B, 3B

SAF surface sensor. Under the applied conditions of pH 3, BSA exhibits an elongated conformation as compared with its natural form.37 The adsorption rate is dominated by electrostatic forces whereas the protein-surface interactions most probably are of electrostatic and hydrophobic nature. Figure 2 presents the adsorption kinetics of BSA in citrate buffer of pH 3 at low (A) and high (B) ionic strength conditions, respectively. Both plots contain the kinetic data recorded at three different bulk protein concentrations of 4 nM, 8 nM, and 16 nM, respectively. As expected, the initial adsorption rates are proportional to the applied concentrations. Also, the saturation levels at a given ionic strength are equal for the three applied bulk protein concentrations. This is a clear indication that a monolayer coverage is formed in the saturation stage which is in agreement with other experimental studies on BSA.38-40 Rinsing of adsorbed proteins with protein-free buffer yielded no measurable

Figure 5. (A) Adsorption kinetics of BSA in citrate buffer (50 mM) at varying pH: pH 4.85 (64 nM, O), pH 6 (200nM, 0), and pH 7 (200 nM, )). Arrows indicate the points in time when rinsing with pure buffer was started. (B) Rate-coverage plots of the kinetics presented in Figure 5A. Solid lines refer to the best fit obtained with the proposed model (see text).

desorption at any point indicating that the adsorption process is essentially irreversible under the given conditions. The most important observation is that the adsorption kinetics clearly differ from an exponential-like growth that is characteristic for conventional concepts like the Langmuir adsorption model. Here, the adsorption rate is apparently not proportional to the available surface area but stays almost constant until about 80% of the maximum coverage is reached. In other words, it seems like adsorbing proteins do not sense the shrinkage of free adsorption sites until the surface is almost filled. Even more precisely, the adsorption rate tends to slightly increase within the first stage of the adsorption as can be seen when the adsorption rate is plotted as a function of the normalized surface coverage which we refer to as the rate-coverage plot. (Figure 3 A and B). In fact, this representation highlights a key feature of the adsorption kinetics of BSA at both ionic strength conditions (Figure 3A, 5 mM; Figure 3B, 50 mM), namely a slightly linearly increasing adsorption rate in the beginning followed by a sudden sharp decline at a certain coverage level. The fact that the adsorption rate does not decrease during the major part of the adsorption process is an unambiguous indication for the presence of cooperative effects.5,7,8 Moreover, increasing adsorption rates imply that adsorbed proteins do not hinder but even promote further protein adsorption. By taking into consideration the observation that protein adsorption at pH 3 is generally not decelerated in regions of high protein density (see Figure 1), it can be concluded that proteins preferentially adsorb in the vicinity of other preadsorbed proteins. Consequently, the decelerating effect of Langmuir-type adsorption, in which adsorbed proteins reduce the free surface area and hinder further adsorption, is compensated or even overcompensated by cooperative effects. By comparing the adsorption kinetics of BSA at low ionic strength (Figure 2A and Figure 3A) with those measured at high ionic strength (Figure 2B and Figure 3B), it is noticed that there are two significant differences although they look qualitatively very similar. First, the saturation level is about 30% lower at 5 mM ionic strength than at 50 mM ionic strength, and second of all, the initial adsorption rates are much higher at low ionic strength than at high ionic strength. The reduction of the buffer ionic strength from 50 mM to 5 mM increases the Debye length from 1.4 to 4.3 nm; that is, the protein’s electrostatic potential is less efficiently shielded.41 As a consequence, lateral electrostatic repulsions between the charged molecules are more pronounced at low ionic strength forcing the proteins to adsorb on the surface in greater distances to each other. In contrast, at higher ionic strength, proteins are allowed to adsorb in a more closely packed layer because of less lateral repulsion. The difference in the adsorption rates, on the other hand, is explained by the attractive electrostatic interactions between the negatively charged surface and the net positively charged proteins. A stronger shielding of surface and protein charges at high ionic strength naturally reduces the adsorption rate.42



TABLE 2: Estimated Parameters for BSA Adsorption in Citrate Buffer (50 mM) at pH 4.85, 6, and 7 protein conc. (M) 64 × 10-9 200 × 10-9 200 × 10-9 a

pH

θmax(counts ms-1)


4.85 6.0 7.0

888 ( 59a 198 ( 18 30.5 ( 3.2

2.44 ( 0.69 0.076 ( 0.002 0.013 ( 0.002


0.223 ( 0.011 0.038 ( 0.005

(kon 1 + on kon 2 )/k1 3.93 3.92

R

koff (10-7 s-1)

R2

plot

0.31 ( 0.06 0.32 ( 0.08

7.5 ( 1.4 250 ( 50 500 ( 80

0.9985 0.9979 0.9756

Figure 5 Figure 5 Figure 5

The surface function Φ1′ is included.

Before extracting quantitative information from the presented kinetic data, it is essential to verify that the fluorescent tag attached to the protein does not significantly alter its diffusion or adsorption behavior as has been observed in some experimental works.43-45 Therefore, all adsorption kinetics of BSA depicted in Figure 2A were measured with two different ratios of labeled to unlabeled proteins, namely, p*/p ) 25.0% and p*/p ) 12.5%. Within the experimental error, the adsorption kinetics acquired at p*/p ) 12.5% are by factor two lower than the kinetics acquired at p*/p ) 25.0%. A preferential adsorption of labeled over unlabeled BSA can therefore be excluded as this would have yielded a factor of less than two. Conclusively, the observed adsorption behavior of fluorescently labeled BSA can be considered to reflect the true behavior of native BSA. Modeling. The time evolution of regions of high and low occupancy on the one side and the nonexponential-like adsorption kinetics on the other side is explainable by a preferential adsorption of proteins in the vicinity of other preadsorbed proteins. In this respect, the influence that adsorbed proteins have on the adsorption behavior of the approaching proteins is referred to as cooperative effects. If these effects just compensate for the decelerating behavior of surface occupancy, the resulting adsorption kinetics appear to be linear. In the case of even stronger cooperative effects, a slightly upward concave shape of the adsorption kinetics is observed (see Figure 2). Similar protein adsorption kinetics have been observed in previous studies by fluorescence methods, radiolabeling, and label-free techniques.46-48 Clearly, a consistent mathematical description of the observations presented in Figures 1-3 must include the issue of cooperativity. To this end, the kinetic model proposed by Minton21,23 is a promising approach because it describes the protein layer build-up through a cooperative growth of twodimensional surface clusters. In this framework, proteins are either allowed to adsorb as individual species on the surface which can then diffuse and aggregate to a pre-existing cluster or to directly deposit at the edge of a two-dimensional surface cluster via a “piggyback” pathway. The model was successfully applied to simulate the cooperative adsorption of the protein ezrin to model membranes.6 Herein, we tested a version of this model by neglecting surface diffusion and size exclusion effects because they have a small influence on positive cooperativity.21,23 Least-squares fits of Minton’s model to the experimentally obtained adsorption kinetics are presented as dashed lines in Figure 2B and Figure 3B. Despite an overall qualitative agreement, the experimental data measured in this study are not adequately described by this model. In particular, the model does not reflect the linearly increasing rate in the beginning and the abrupt decrease at a certain coverage level. This disagreement is predominantly obvious when the rate-coverage plots of the experimental and calculated data are compared (Figure 3B). As a first conclusion from this discrepancy, we neglected the appearance of growing two-dimensional surface clusters for the studied systems. This concept would imply that the observable adsorption rate depends on the average cluster size such that it becomes a nonlinear function of the surface coverage.21,23 However, the experimentally obtained adsorption

Figure 6. Surface scan images recorded at different stages during the adsorption of BSA in citrate buffer at pH 4.85 (64 nM, left column) and at pH 6 (200 nM, right column). The profile plots underneath the images show the evolution of the fluorescence intensity along the cross sections marked by the dotted lines.

rate is a linear function of the surface coverage until about 80% of the saturation limit is reached (see Figure 3). Additionally, it has been reported that the saturation coverage of BSA on a hydrophilic surface is by a factor of 5 lower than that at pH ) pI because charged proteins generally tend to adsorb apart from each other.39,40 Given the layer thickness of a BSA monolayer at pH 3 of 30 Å to 40 Å, it was calculated that at least half of the total surface area remains uncovered at saturation. Hence, in the studied system adsorbed proteins form a layer of rather loosely distributed molecules instead of surface clusters in which proteins would be densely ordered.38-40 The surface distribution of adsorbed proteins is connected with the electrostatic interactions between bulk proteins and surface during the adsorption process. It is accepted that the electrostatic field at a protein covered interface arises from the coupled fields of the surface and the inhomogeneously distributed charged residues within adsorbed proteins.42,49 On negatively charged surfaces, proteins adapt an orientation in which the majority of their positively charged regions are directed toward the surface and the majority


Rabe et al.

dθ2 θ - koff ) kon 2 · c · Φ2 · 2 · θ2 dt θmax

(3)

θ eR θmax θ >R θmax

(4)

Φ2 )

{

1

(

1 θ · 11-R θmax

)

In these equations, θ1 and θ2 denote the fraction of surface proteins adsorbed via the Langmuir-type or the cooperative adsorption pathway, θ denotes the total surface coverage, and θmax is the experimentally observed maximum coverage. kon 1 , off off kon 2 and k1 , k2 are the adsorption and desorption rate constants for the corresponding pathways, c is the bulk protein concentration, and Φ1,Φ2 refers to the available surface function. As will be discussed in more detail below, the parameter R defines the coverage limit at which the cooperative adsorption pathway changes from the regime in which the overall coverage has no influence on the adsorption rate into the regime where the increasing coverage slows down the adsorption rate. In eq 4, the factor (1 - R)-1 has to be included in the second part of the surface function to avoid a mathematical point of discontinuity at θ ) R · θmax. Summing up the time-integrated adsorption rates of the two pathways gives the total surface coverage.

θ ) θ1 + θ2

Figure 7. (A) Adsorption kinetics of Fib in citrate buffer (50 mM) at varying pH: pH 3 (8 nM, O), pH 4 (8nM, 0), and pH 5 (8 nM, )). (B) Rate-coverage plots of the kinetics presented in Figure 7A. Solid lines refer to the best fit obtained with the model proposed in this work.

of their negatively charged regions are directed toward the bulk solution and thereby enhance the attraction of further bulk proteins.40,49 Because of the fixed orientation of surface bound proteins, equally charged regions within the proteins oppose each other in lateral direction with respect to the surface causing a repulsive force. Hence, adsorbed proteins induce an attractive force in vertical direction toward diffusing bulk proteins and simultaneously cause a lateral repulsive force toward other proteins on the surface. Consequently, we suggest that bulk proteins coming close to surface bound proteins are guided to a free adsorption area in their vicinity but not as close as in a surface cluster. The protein density in a certain region hence determines the probability with which this cooperative adsorption pathway takes place. Bulk proteins that reach the surface in a larger distance from other adsorbed proteins are considered to adsorb without further influence through the classical Langmuir-type adsorption pathway. Thus, the overlapping of both, the Langmuir type pathway (pathway 1) and the cooperative adsorption pathway (pathway 2), leads to the observed behavior (Figure 4). The description of pathway 1 is as follows:

dθ1 off ) kon 1 · c · Φ1 - k1 · θ1 dt θ Φ1 ) 1 θmax

(1) (2)

The description of the additional cooperative adsorption pathway (indicated by the subscript 2) is as follows:

(5)

To account for the cooperative effects in the second pathway the cooperative adsorption rate is set proportional to the normalized surface coverage (θ/θmax). Mechanistically, this ensures that the probability of a bulk protein to adsorb close to another preadsorbed protein increases with the increasing surface coverage. However, the cooperative adsorption rate constant kon 2 is independent of the number of neighboring proteins. As discussed before, the intermolecular distances between adsorbed proteins are larger than in a surface cluster, and hence the accumulation of attractive cooperative forces is weak when proteins assemble in a certain region. An important effect accompanying cooperative adsorption is that approaching bulk proteins do not adsorb at random positions but are directed to favorable adsorption areas in the vicinity of preadsorbed proteins. This is explained by a kind of electrostatic selforganization mediated by lateral interactions from neighboring surface bound proteins.5,24,50 As a consequence, the adsorption rate of cooperatively adsorbing proteins does not decrease with increasing occupancy at low to moderate surface densities which causes the observed linear adsorption kinetics (see Figure 2). In order to model the sudden decrease of the adsorption rate at a fixed adsorption stage (see Figure 3), we define a coverage limit (R · θmax) up to which the available surface function Φ2 is set constant (Φ2 ) 1). Only above this limit the surface function is set proportional to the number of free binding sites. Thus, the complete model defined by eqs 1-5 comprises the adsorption of proteins by the Langmuir-type adsorption pathway on any available binding site, and an additional cooperative adsorption pathway which is only encountered in the vicinity of preadsorbed proteins. In this respect, kon 1 is a measure for the general probability of proteins to adsorb on the surface whereas kon 2 corresponds to the additional cooperative adsorption probability that is encountered in the vicinity of other surface bound proteins. Because of the overlap of Langmuir-type adsorption on and cooperative adsorption, the sum of kon 1 and k2 can be considered a measure for the total probability of proteins to adsorb in the vicinity of other surface bound proteins. In on mathematical terms, the two rate constants kon 1 and k2 are most



TABLE 3: Estimated Parameters for Fib Adsorption in Citrate Buffer (50 mM) at pH 3, 4, and 5 protein conc. (M)

pH

θmax(counts ms-1)

8 × 10-9 8 × 10-9 8 × 10-9

3.0 4.0 5.0

61 ( 1 64 ( 5 97 ( 6


1.68 ( 0.10 0.81 ( 0.07 0.99 ( 0.02


(6)

Analogously, the probability by which a protein adsorbs to the position in the vicinity of preadsorbed proteins is given by on p** ) (kon 1 + k2 ) · c

R

koff (10-7 s-1)

R2

plot

2.60 2.34 2.04

0.66 ( 0.06 0.83 ( 0.02 0.73 ( 0.02

0 0 0

0.9980 0.9991 0.9991

Figure 7 Figure 7 Figure 7

2.68 ( 0.24 1.09 ( 0.09 1.03 ( 0.10

easily understood by regarding an empty position in an unoccupied region and an empty position in the vicinity of preadsorbed proteins. The probability by which a protein adsorbs to the position within the empty region is given by:

p* ) kon 1 · c

on on (kon 1 + k2 )/k1

(7)

From eqs 6 and 7, it follows directly that the relation of the on on rate constants (kon 1 + k2 )/k1 describes by which factor the probability of a protein to adsorb in the vicinity of preadsorbed proteins is increased over the probability to adsorb at any random position. The parameter R determines the surface coverage level when the adsorption rate abruptly starts to decline toward zero. In the first stage of the adsorption, the decreasing rate of the Langmuir-type adsorption pathway is overlapped by the increasing rate of the cooperative adsorption pathway. As discussed before, cooperatively adsorbing proteins are guided to preferential adsorption areas close to preadsorbed proteins and are hence not affected by the shrinkage of unoccupied surface. Naturally, this mechanism breaks down when the surface becomes strongly crowded such that not every bulk protein diffusing close to other adsorbed proteins is guided to a free adsorption area. In this respect, R points to the coverage limit (R · θmax) at which the cooperative adsorption rate starts to decrease. Phenomenologically, a large value of R can be explained by attractive forces ranging over a rather long distance caused by surface bound proteins. Then, the probability that a bulk protein is guided to a free adsorption area in the vicinity of adsorbed proteins is still high at increased coverage levels. Small values of R, on the other hand, indicate attractive forces that range only over short distances. In that case, the cooperative adsorption pathway is already influenced by the quantity of available adsorption sites at low coverage levels. That means, when an incoming bulk protein reaches an occupied region, there is a high probability that it is not rejected but guided to the nearest adsorption area provided that R is high enough. From the experimental data, it is evident that the adsorption rate undergoes a sharp transition between the first stage in which the available surface is irrelevant (θ e R · θmax) and the second stage in which the available surface is relevant (θ > R · θmax; see Figure 3). Hence, from an experimental point of view, it is justified to approximate this behavior using a step-like transition between these two stages in the proposed model as realized by the definition of Φ2 (eq 4). Naturally, the step function can only represent a simplified description of the real transition between the two stages. An alternative approach could be to perform Monte Carlo simulations of the proposed mechanism from which the macroscopic adsorption kinetics could be extracted. However, as the experimental error at the transition stage is quite significant, a great improvement of the data fits is not expected. Also, a mathematically exact treatment of the protein adsorption

behavior under consideration of the microenvironment of each adsorbing molecule would extend the number of rate equations to infinity.21 To quantitatively evaluate the parameters describing the Langmuir-type adsorption pathway (k1on) and the cooperative adsorption pathway (kon 2 , R), the proposed model was fitted to the experimentally acquired kinetics of BSA. Since all adsorption kinetics of BSA measured at pH 3 were found to be off irreversible, the off-rate constants koff 1 and k2 were set to zero leading to the following expression for the total rate of the irreversible adsorption: dθirr ) dt

{

θ +n θmax θ 1 · (m + n) · 1-R θmax

m·

θ eR θmax 2 1 θ θ R + +n >R · m+ ·n · 1-R 1-R θmax θmax

( ) (

)

(8) on m ) (kon 2 - k1 ) · c

n ) kon 1

·c

(9) (10)

The function given by eqs 8-10 is composed of a linear part characterized by the slope m and the intercept n and a quadratic function whose parameters are fixed by m, n, and R. To find appropriate parameters for the proposed model, the maximum coverage (θmax) of each data set was determined as the final saturation limit of the respective adsorption curve. The count rates of the plateaus of the kinetics measured at three different bulk protein concentrations were therefore averaged and used as maximum surfaces coverage. Subsequently, the derivations of the adsorption kinetics were calculated and plotted as functions of the normalized surface coverage θ/θmax, that is, as rate-coverage plots. The function defined by eq 8 was then fitted to all rate-coverage plots yielding estimated values for the three on parameters R, kon 1 , and k2 (Table 1). For the three curves obtained in one system but at different bulk protein concentrations, we performed a global fit ensuring that all parameters except the given concentration are indifferent. Finally, the estimated parameters were used to numerically integrate the set of rate equations given by eq 15 leading to the calculated adsorption kinetics presented as solid lines in Figures 2 and 3. To evaluate the correlation between experimental data and calculated kinetics, the coefficient of determination R2 for each data set is added to Table 1. The obvious agreement between experiment and model (most R2 values given in Table 1 are above 0.990) highlights the validity of the overlap of Langmuirtype adsorption and cooperative adsorption. It confirms in particular that an accumulation of cooperative forces plays no important role which is a clear indication that two-dimensional surface clusters do not exist in the studied systems. Instead, the model suggests a layer of more loosely distributed proteins with a certain degree of self-organization. This means that one fraction of all adsorbed proteins has been guided to preferential adsorption areas due to cooperative effects whereas the other fraction has adsorbed on random surface positions.


Rabe et al.

In relation to the proposed model, the evolution of the protein density distribution observed in the scan images (Figure 1A,B) is easily explained. The probability that proteins adsorb in close proximity to preadsorbed proteins is by a factor of more than two higher than the probability to adsorb in an unoccupied on on region (see (kon 1 + k2 )/k1 values in Table 1). Thus, the reduced number of available binding sites in regions of high protein occupancy is overcompensated by an increased adsorption probability in the circumference of adsorbed proteins. Consequently, variations in the protein density are not balanced out. Cooperative Adsorption of BSA at Higher pH Values. To evaluate if the proposed adsorption mechanism is also valid in higher pH environments, adsorption kinetics were recorded for BSA in citrate buffer (50 mM) at pH 4.85, 6, and 7, respectively (Figure 5A,B). Not surprisingly, the adsorption kinetics at different pH differ strikingly in terms of curve shape, initial adsorption rate, and saturation limit from those acquired at pH 3. Additionally, rinsing experiments revealed that the adsorption becomes reversible at higher pH environments. To perform a curve fit in the case of reversible protein adsorption, eq 8 is extended by the term -koff · θ:

dθreV dθirr ) - koff · θ dt dt

(11)

off Herein, the two off-rate constants koff 1 , k2 were set equal for simplification. Due to this approximation, the fitting procedure explained above is still valid once the off-rate constant koff has been determined separately by fitting the desorption curve to a monoexponential decay.51

θ(t) ) (θ0 - θ∞) · e-koff · (t-t0) + θ∞

(12)

θ0 and θ∞ denote the surface coverages at the onset of rinsing and after long desorption time; t0 is the point in time when rinsing is started. Of course, θmax is always higher than the observable equilibrium level in the case of reversible adsorption and is therefore only indirectly accessible from the adsorption kinetics. An efficient way to determine all parameters in these on cases is to choose initial values for the parameters R, kon 1 , k2 and to numerically optimize θmax by time-integrating the on extended model (eq 11). Subsequently, R, kon 1 , k2 are estimated through fitting the extended model to the rate-coverage plot and used for a new optimization of θmax. This cycle is repeated until self-consistency. All parameters and the coefficients of determination for the fits of BSA at pH 4.85, 6, and 7 are summarized in Table 2. The fit of the model to the adsorption kinetics of BSA at pH 6 and pH 7 displays an excellent agreement of the proposed mechanism with the observed behavior (solid lines in Figure 5A,B). From the rate-coverage plots (Figure 5B, 0, ]), it is evident that a contribution of the cooperative adsorption pathway still plays an important role in the initial stage. In comparison to the adsorption behavior at pH < pI, much slower rates and lower saturation levels are noticed in these conditions (see Table 2). This is because at rising pH values above the pI the number of negative charges within the protein and on the surface increases leading to stronger repulsions between protein molecules and between proteins and the surface. Explanations why proteins can bind to a negatively charged surface despite their own net negative charge, that is, on the wrong side of the isoelectricpoint,arecomprehensivelydiscussedintheliterature.52-54 Clearly, as R is rather small in these conditions (R ) 0.31 at pH 6, R ) 0.32 at pH 7), the cooperative effects which are

most obvious at low surface coverages can be easily disregarded especially at high bulk concentrations. When BSA adsorbs in conditions of pH ) pI (4.85), it exhibits the highest observable saturation level in its equilibrium state which is in agreement with the commonly reported behavior of this protein38-40 and of many other proteins.49,55,56 At pH ) pI, the curve shape of the kinetics is exponential-like during the whole adsorption process and hence differs clearly from the kinetics at pH 3. Plotting the adsorption rate as a function of surface coverage (Figure 5B, inset) reveals no characteristic of positive cooperativity. On the contrary, the upward concave shape of the curve is a strong indication that size exclusion effects which are comprehensively investigated in the literature have to be considered in this system.57-60 The best fit to the adsorption kinetics at pH 4.85 was obtained by eliminating the cooperative adsorption pathway (kon 2 ) 0) and using a new surface function Φ1′ that approximately accounts for size exclusion effects according to the random sequential adsorption (RSA) theory61 (parameters summarized in Table 2). Φ1 )

( ) ( ) ( ) 1-

1 - 0.812 · 1 -

θ θj

3

θ θ + 0.2336 · 1 θj θj

2

( )

+ 0.0845 · 1 -

θ θj

3

(13) In this equation, θj is defined as the jamming limit which refers to a surface density of 55% of a close-packed monolayer. At its isoelectric point, BSA has a net charge of zero such that lateral repulsions of adsorbed proteins are minimized. This explains the observed maximum surface coverage of all pH conditions evaluated. The surface function Φ1′ allows a more accurate fit of the observed kinetics at pI than using the simple Langmuir model expressed by Φ1 (R2 ) 0.9985 when Φ1′ is used and R2 ) 0.9977 when Φ1 is used). In the absence of cooperativity, proteins apparently bind to random surface sites, and this is better accounted for by the RSA theory than by the simple Langmuir theory. However, the difference between these two theories is rather subtle and is only strong enough pronounced in the case of pure noncooperative adsorption at pH ) pI. In the case of an overlap of cooperative and Langmuirtype adsorption pathways (as found for all other pH conditions), our experimental data do not allow the direct observation of size exclusion effects. For this reason, we generally implemented the simple surface function Φ1 into the Langmuir-type adsorption pathway instead of the more complicated surface function Φ1′ which refers to the RSA model. We also compared the evolution of the protein density distribution at pH 4.85 and pH 6 by recording scan images of a chosen area at different adsorption stages. Only very slight intensity variations were visible during the adsorption of BSA at pH 4.85 (Figure 6, left column). Accordingly, the profile plots along the marked lines show that density fluctuations are rapidly balanced out during the adsorption resulting in a uniform protein distribution. In agreement with the exponential-like adsorption kinetics (Figure 5A), this is explained by the absence of cooperative effects. In contrast, the images recorded during the adsorption of BSA at pH 6 (Figure 6, right column) reveal a clear distinction between regions of high protein density and regions of low density. The profile plots also confirm the preferential adsorption of bulk proteins in the proximity of other adsorbed proteins. Cooperative Adsorption of Fib. In the following, we present analogies between the adsorption behavior of BSA and that of

Nonspecific Protein Adsorption Events fibrinogen (Fib), to demonstrate that the discussed mechanism is not restricted to one single protein species. Like BSA, Fib is a highly abundant plasma protein but with a significantly different structure (rod-like), a five times higher molecular weight (340 kDa), and a slightly higher isoelectric point (pI ) 5.5). Nevertheless, the adsorption kinetics of Fib at pH 3, 4, 5 were found to be comparable to the kinetics of BSA (Figure 7). In fact, the curve shapes of these two proteins are almost identical at pH 3 and only differ from each other at higher pH values. Conclusively, a contribution of cooperativity during the adsorption of Fib is evident. The excellent fit of the proposed model defined by eq 15 to the recorded adsorption kinetics (Figure 7A) and to the rate-coverage plot (Figure 7B) suggests that the same mechanistic description as used for BSA adsorption can be applied to the adsorption of Fib. The parameters for these fits are summarized in Table 3. These results suggest a rather general validity of the discussed adsorption model. Conclusions In the present study, we experimentally investigated the phenomenon of cooperative protein adsorption on solid interfaces by the combination of in situ surface imaging and kinetic analyses. Scan images during the course of adsorption enabled the observation of a heterogeneous protein density distribution as expressed by regions of high and low occupancy. We showed that the adsorption of bulk proteins into regions of high protein density is related to positive cooperative adsorption whereas adsorption to any available binding site further away from preadsorbed proteins correlates to Langmuir-type adsorption. In the absence of positive cooperativity, proteins adsorb on random surface sites leading to a more homogeneous protein distribution which has been found for the adsorption of BSA at its isoelectric point (see Figure 6 left column). On the other hand, cooperative adsorption is connected to a kind of selforganization forcing incoming proteins to adsorb to binding areas close to other proteins which leads to a more heterogeneous protein distribution (see Figure 1 and Figure 6, right column). As a consequence, the growing surface occupancy does not impede the adsorption rate of cooperatively adsorbing molecules causing the observed linear or even upward concave shaped adsorption kinetics.5,24,50 However, we emphasize that the cooperative effects observed in this study are not related to the formation of surface aggregates or two-dimensional clusters in which proteins bind to the surface and to each other. The experimentally acquired adsorption kinetics prove that a model based on growing surface clusters yields unsatisfying fitting results. Instead, the cooperative adsorption pathway results in a rather loose distribution of surface proteins with a certain degree of self-organization. We show experimentally that an abrupt change in the cooperative adsorption behavior takes place at a certain coverage limit above which the surface density starts to decelerate the adsorption rate. Furthermore, the model provides a simple and very efficient mechanistic explanation for linearly increasing adsorption kinetics based on the overlap of Langmuir-type and cooperative adsorption pathways. Given the considerable number of equivalent experimental observations on a variety of different proteins in the literature,7,46-48,62-64 the proposed model apparently describes a rather general adsorption behavior. Dependent on the importance of the respective Langmuir-type and the cooperative adsorption pathon way (as expressed by kon 1 and k2 ), the model perfectly fits to on on on upward concave (k1 < k2 ), linear (kon 1 ) k2 ), or exponentialon on like (k1 > k2 ) adsorption kinetics. Hence, it allows for extracting quantitative information about the positive cooperative

J. Phys. Chem. B, Vol. 112, No. 44, 2008 13979 effects from the curve shapes of protein adsorption kinetics. The wide applicability of the model is also highlighted by its ability to describe the adsorption kinetics of BSA and Fib in different pH environments. Additionally, in the case of a more complex protein adsorption behavior, the model can be readily extended to account for additional phenomena such as structural reorientation, lateral repulsion, or overshootings. Acknowledgment. The authors wish to thank the Swiss National Science Foundation (SNSF) and the Forschungskredit of the University of Zurich for financial support. References and Notes (1) Gray, J. J. Curr. Opin. Struct. Biol. 2004, 14, 110. (2) Hlady, V.; Buijs, J. Curr. Opin. Biotechnol. 1996, 7, 72. (3) Nakanishi, K.; Sakiyama, T.; Imamura, K. J. Biosci. Bioeng. 2001, 91, 233. (4) Norde, W. Colloids Surf., B 2008, 61, 1. (5) Ball, V.; Ramsden, J. J. Phys. Chem. B 1997, 101, 5465. (6) Herrig, A.; Janke, M.; Austermann, J.; Gerke, V.; Janshoff, A.; Steinem, C. Biochemistry 2006, 45, 13025. (7) Nygren, H.; Alaeddin, S.; Lundstro¨m, I.; Magnusson, K. E. Biophys. Chem. 1994, 49, 263. (8) Van der Veen, M.; Stuart, M. C.; Norde, W. Colloids Surf., B 2007, 54, 136. (9) Cuypers, P. A.; Willems, G. M.; Hemker, H. C.; Hermens, W. T. Ann. N.Y. Acad. Sci. 1987, 516, 244. (10) Hinderliter, A.; Almeida, P. F. F.; Creutz, C. E.; Biltonen, R. L. Biochemistry 2001, 40, 4181. (11) Hinderliter, A.; Biltonen, R. L.; Almeida, P. F. F. Biochemistry 2004, 43, 7102. (12) Hinderliter, A.; May, S. J. Phys.: Condens. Matter 2006, 18, 1257. (13) Malmsten, M.; Muller, D.; Lassen, B. J. Colloid Interface Sci. 1997, 193, 88. (14) Walker, R. K.; Krishnawamy, S. J. Biol. Chem. 1994, 269, 27441. (15) Heuberger, R.; Sukhorukov, G.; Voros, J.; Textor, M.; Mohwald, H. AdV. Funct. Mater. 2005, 15, 357. (16) Huang, N. P.; Michel, R.; Vo¨ro¨s, J.; Textor, M.; Hofer, R.; Rossi, A.; Elbert, D. L.; Hubbell, J. A.; Spencer, N. D. Langmuir 2001, 17, 489. (17) Zurcher, S.; Wackerlin, D.; Bethuel, Y.; Malisova, B.; Textor, M.; Tosatti, S.; Gademann, K. J. Am. Chem. Soc. 2006, 128, 1064. (18) Dobson, C. M. Phil. Trans. R. Soc. B 2001, 356, 133. (19) Sunde, M.; Blake, C. AdV. Protein Chem. 1997, 50, 123. (20) Richter, R. P.; Him, J. L. K.; Tessier, B.; Tessier, C.; Brisson, A. R. Biophys. J. 2005, 89, 3372. (21) Minton, A. P. Biophys. J. 2001, 80, 1641. (22) Minton, A. P. Biophys. J. 1999, 76, 176. (23) Minton, A. Biophys. Chem. 2000, 86, 239. (24) Ball, V.; Ramsden, J. Colloids Surf., B 2000, 17, 81. (25) Ortega-Vinuesa, J. L.; Tengvall, P.; Lundstrom, I. J. Colloid Interface Sci. 1998, 207, 228. (26) Mulheran, P. A.; Pellenc, D.; Bennett, R. A.; Green, R. J.; Sperrin, M. Phys. ReV. Lett. 2008, 100, 068102. (27) Ruckstuhl, T.; Enderlein, J.; Jung, S.; Seeger, S. Anal. Chem. 2000, 72, 2117. (28) Ruckstuhl, T.; Rankl, M.; Seeger, S. Biosens. Bioelectron. 2003, 18, 1193. (29) Ruckstuhl, T.; Verdes, D. Opt. Express 2004, 12, 4246. (30) Rabe, M.; Verdes, D.; Rankl, M.; Artus, G. R. J.; Seeger, S. ChemPhysChem 2007, 8, 862. (31) Rankl, M.; Ruckstuhl, T.; Rabe, M.; Artus, G. R. J.; Walser, A.; Seeger, S. ChemPhysChem 2006, 7, 837. (32) Verdes, D.; Ruckstuhl, T.; Seeger, S. J. Biomed. Opt. 2007, 12, 034012. (33) Enderlein, J.; Ruckstuhl, T.; Seeger, S. Appl. Opt. 1999, 38, 724. (34) Rezwan, K.; Meier, L. P.; Gauckler, L. J. Biomaterials 2005, 26, 4351. (35) Calonder, C.; Tie, Y.; Van Tassel, P. R. Proc. Natl. Acad. Sci. U. S. A. 2001, 98, 10664. (36) Luthi, Y.; Ricka, J.; Borkovec, M. J. Colloid Interface Sci. 1998, 206, 314. (37) El Kadi, N.; Taulier, N.; Le Huerou, J. Y.; Gindre, M.; Urbach, W.; Nwigwe, I.; Kahn, P. C.; Waks, M. Biophys. J. 2006, 91, 3397. (38) Lu, J. R.; Zhao, X. B.; Yaseen, M. Curr. Opin. Colloid Interface Sci. 2007, 12, 9. (39) Su, T. J.; Lu, J. R.; Thomas, R. K.; Cui, Z. F. J. Phys. Chem. B 1999, 103, 3727. (40) Su, T.; Lu, J.; Thomas, R.; Cui, Z.; Penfold, J. J. Phys. Chem. B 1998, 102, 8100.

13980 J. Phys. Chem. B, Vol. 112, No. 44, 2008 (41) Israelachvili, J. Intermolecular and Surface Forces, 2nd ed.; Academic Press: London, 1991; Vol. 201. (42) Fang, F.; Szleifer, I. J. Chem. Phys. 2003, 119, 1053. (43) Gajraj, A.; Ofoli, R. Y. Langmuir 2000, 16, 8085. (44) Mollmann, S. H.; Elofsson, U.; Bukrinsky, J. T.; Frokjaer, S. Pharm. Res. 2005, 22, 1931. (45) Sapsford, K. E.; Ligler, F. S. Biosens. Bioelectron. 2004, 19, 1045. (46) Ball, V.; Lustig, A.; Ramsden, J. J. Phys. Chem. Chem. Phys. 1999, 1, 3667. (47) Daly, S. M.; Przybycien, T. M.; Tilton, R. D. Langmuir 2003, 19, 3848. (48) Vasina, E. N.; Dejardin, P. Langmuir 2004, 20, 8699. (49) Bremer, M. G. E. G.; Duval, J.; Norde, W.; Lyklema, J. Colloids Surf., A 2004, 250, 29. (50) Vasina, E.; Dejardin, P. Langmuir 2004, 20, 8699. (51) Ball, V. Colloids Surf., B 2004, 33, 129. (52) Hollmann, O.; Gutberlet, T.; Czeslik, C. Langmuir 2007, 23, 1347. (53) Hollmann, O.; Reichhart, C.; Czeslik, C. Z. Phys. Chemie-Int. J. Res. Phys. Chem. Chem. Phys 2008, 222, 205.

Rabe et al. (54) Hollmann, O.; Steitz, R.; Czeslik, C. Phys. Chem. Chem. Phys. 2008, 10, 1448. (55) Ho¨o¨k, F.; Rodahl, M.; Kasemo, B.; Brzezinski, P. Proc. Natl. Acad. Sci. U. S. A. 1998, 95, 12271. (56) Haynes, C.; Norde, W. Colloids Surf., B 1994, 2, 517. (57) Schaaf, P.; Talbot, J. Phys. ReV. Lett. 1989, 62, 175. (58) Schaaf, P.; Voegel, J. C.; Senger, B. J. Phys. Chem. B 2000, 104, 2204. (59) Talbot, J.; Schaaf, P. Phys. ReV. A 1989, 40, 422. (60) Talbot, J.; Tarjus, G.; Van Tassel, P. R.; Viot, P. Colloids Surf., A 2000, 165, 287. (61) Schaaf, P.; Talbot, J. J. Chem. Phys. 1989, 91, 4401. (62) Asanov, A. N.; DeLucas, L. J.; Oldham, P. B.; Wilson, W. W. J. Colloid Interface Sci. 1997, 191, 222. (63) Fair, B. D.; Jamieson, A. M. J. Colloid Interface Sci. 1980, 77, 525. (64) Wahlgren, M.; Elofsson, U. J. Colloid Interface Sci. 1997, 188, 121.

JP804532V

Surface Organization and Cooperativity during Nonspecific Protein

Recommend Documents