Dynamic Alterations of Fibronectin Layers on Copolymer Substrates

2928

Langmuir 2004, 20, 2928-2933

Dynamic Alterations of Fibronectin Layers on Copolymer Substrates with Graded Physicochemical Characteristics Lars Renner, Tilo Pompe,* Katrin Salchert, and Carsten Werner Institute of Polymer Research Dresden & The Max Bergmann Center of Biomaterials Dresden, Hohe Strasse 6, 01069 Dresden, Germany Received December 2, 2003. In Final Form: January 15, 2004 Desorption and exchange of preadsorbed fibronectin layers in pure buffer solution and solutions of human serum albumin or fibronectin, respectively, were studied in dependence on the physicochemical characteristics of maleic acid copolymer films used as substrates. Although the preadsorbed amount of fibronectin differed only slightly, the protein was found to exhibit a significantly enhanced anchorage at the more hydrophobic polymer surface as compared to the more hydrophilic and more negatively charged polymer surface. The preadsorbed fibronectin layer was most efficiently exchanged by fibronectin (i.e., in the homodisplacement process) while pure buffer solution and human serum albumin solutions induced desorption or exchange of fibronectin to lower and similar degrees. An increase of the total adsorbed amount of protein due to additional adsorption of fibronectin or human serum albumin accompanied the partial exchange of the preadsorbed fibronectin in the displacement experiments. Evaluation of the kinetics of desorption and exchange of fibronectin at any of the substrates revealed two kinds of surface-attached protein populationssa fast desorbing species and a species with a slow desorption and exchange rate. By a multivariate regression analysis the surface characteristics of the polymer substrate were confirmed to determine the degree of protein desorption and exchange while the dynamics of the layer alteration was found to solely depend on the diffusion behavior of the proteins.

Introduction Protein adsorption processes, occurring at solid/liquid interfaces, are known to play a key role in the interaction between biological systems and artificial surfaces. There is a great technological significance in protein adsorption for the control of such phenomena like blood compatibility of implants, cell in-growth in tissue engineering scaffolds, or the inhibition of biofouling. Due to the importance of protein adsorption and exchange, numerous studies already addressed the understanding of the interaction of proteins with solid surfaces, their competitive adsorption behavior, and the kinetics of these phenomena.1-8 While characteristic features could be elaborated for model systems and basic principles of protein interaction and adsorption have been established, control over the composition of protein layers, the lateral distribution and pattern formation of proteins at interfaces, and in particular the dynamic changes within protein adsorption layers are still most demanding tasks for many applications. In particular, the current knowledge about exchange and desorption phenomena of proteins cannot yet provide a clear picture concerning the relevance of interaction parameters. In many studies protein adsorption was concluded to be an irreversible process.9,10 Additionally, several studies indicate that the reversibility of adsorption * To whom correspondence should be addressed: E-mail: [email protected]. Phone: +49-351-4658274. Fax: +49-3514658533. (1) Haynes, C. A.; Norde, W. Colloids Surf. 1994, B2, 517. (2) van Tassel P. R.; Brusatori, M. A.. J. Colloid Interface Sci. 1999, 219, 333. (3) Ramsden, J. J. In Biopolymers at Interfaces; Malmsten, M., Ed.; Marcel Dekker: New York, 1998; p 321. (4) Ball, V.; Bentaleb, A.; Hemmerle, J.; Voegel, J.-C.; Schaaf, P. Langmuir 1996, 12, 1614. (5) Evans, J. W. Rev. Mod. Phys. 1993, 65, 1281. (6) Calonder, C.; Van Tassel, P. R. Langmuir 2001, 17, 4392. (7) Minton, A. P. Biophys. J. 2001, 80, 1641. (8) Wertz, C. F.; Santore, M. M. Langmuir 1999, 15, 8884.

depends on solution characteristics including the presence of competing proteins and the variation of pH and temperature as well as on the substrate characteristic. Protein exchange processes were analyzed for homomolecular systems of serum albumin and immunoglobulins on silica and polystyrene sulfonate surfaces.4,11-13 In more application oriented experiments, heteromolecular exchange processes were investigated.14,15 The exchange of preadsorbed proteins by other proteins in a size-depending manner (exchange of smaller proteins by larger ones) has been passed into the scientific canon as the “Vroman effect”.16 Theoretical approaches were reported for the analytical description of protein desorption dynamics based on the ideas of nearly irreversible adsorption or for the simple exponential decay of the adsorbed amount.3,14,17,18 The basic idea of the current study was 2-fold: At first, we intended to improve the understanding of protein exchange for in vitro cell culture systems as frequently applied to trigger growth and differentiation of adherent cells on biomaterial surfaces. Therefore, the exchange of fibronectin (FN) and human serum albumin (HSA) was studied on different maleic acid copolymer thin films (9) Hibbert, D. B.; Gooding, J. J.; Erokhin, P. Langmuir 2002, 18, 1770. (10) Norde, W.; Haynes, C. A. In Proteins at Interfaces II; Horbett, T. A., Brash, J. L., Eds.; American Chemical Society: Washington, DC, 1994; p 26. (11) Giacomelli, C. E.; Norde, W. J. Colloid Interface Sci. 2001, 233, 234. (12) Sukhishvili, S. A.; Granick, S. J. Chem. Sci. 1999, 20, 10153. (13) Brash, J. L.; Samak, Q. M. J. Colloid Interface Sci. 1978, 65, 495. (14) Huetz, Ph.; Ball, V.; Voegel, J.-C.; Schaaf, P. Langmuir 1995, 11, 3145. (15) Green, R. J.; Davies, M. C.; Roberts, C. J.; Tendler, S. J. B. Biomaterials 1999, 20, 385. (16) The Vroman Effect; Bamford, C. H., Cooper, S. L., Tsurutta, T., Eds.; VSP: Utrecht, 1992. (17) Dijt, J. C.; Cohen Stuart, M. A.; Fleer, G. J. Macromolecules 1992, 25, 5416. (18) Wertz, C. F.; Santore, M. M. Langmuir 2002, 18, 706.

10.1021/la0362627 CCC: $27.50 © 2004 American Chemical Society Published on Web 02/21/2004

Alterations of Fibronectin Layers

Langmuir, Vol. 20, No. 7, 2004 2929

providing model surfaces with graded physicochemical characteristics.19 The variation of the size of the comonomer results in a change of the relative concentration of unpolar alkyl components and polar acidic groups in the copolymer films. An increased surface concentration of polar acidic groups leads to higher charge density and surface energy of the polymer films and was also found to induce structural transitions of the surface-confined polymer depending on the degree of ionization of the maleic acid.20 Because of the chemical similarity of the utilized substratessonly differing in the ratio of unpolar alkyl groups and negatively charged carboxyl groupssthe surface hydrophobicity correlates with the surface concentration of the maleic acid groups. These features may vary the protein-substrate interaction strength; however, according to well-known principles of proteins at interfaces1 the graduation of the hydrophobicity can be reasonably assumed to play a dominating role. The choice of the two proteins results from their relevance in the context of cell-matrix adhesion. FN is a major component of the extracellular matrix providing adhesion ligands for cellular integrin receptors. Its binding strength and exchange properties are crucial for the cellular behavior with respect to adhesion and differentiation.21-26 HSA can be considered as the major competing protein since it is present at high concentration in the blood and in serum containing cell culture media. Beyond the clarification of the fate of preadsorbed FN at cell culture carriers, our study similarly aimed at the introduction of a general approach to treat the dependence of the desorption and exchange dynamics on relevant parameters, such as the substrate hydrophobicity or the displacing protein, in a quantitative way. For the latter purpose, relevant parameters of substrate and protein properties were implemented in a linear regression model. Theoretical Model and Multivariate Regression Analysis In a common approach Huetz and co-workers14 suggested a simple kinetics for the description of protein desorption depending only on the amount of immobilized protein, which is characterized by the differential equation

∂Γ/∂t ) -kΓ

(1)

where Γ and k are surface coverage and desorption time constant, respectively. A straightforward solution assumes two species of immobilized protein, fast (A) and slow desorbing (B), leading to eq 2

Γ(t) ) ΓA exp(-kAt) + ΓB exp(-kBt)

(2)

To achieve a satisfactory interpretation of our experimental data, a multivariate regression analysis with a simple linear model was performed to reveal the influence of various experimental parameters of the analyzed amounts (ΓA, ΓB) and time constants (kA, kB) of desorption (19) Pompe, T.; Zschoche, S.; Herold, N.; Salchert, K.; Gouzy, M.-F.; Sperling, C.; Werner, C. Biomacromolecules 2003, 4, 1072. (20) Osaki, T.; Werner, C. Langmuir 2003, 19, 5787. (21) Katz, B. Z.; Zamir, E.; Bershadsky, A.; Kam, Z.; Yamada, K. M.; Geiger, B. Mol. Biol. Cell 2000, 11, 1047. (22) Groth, T.; Altankov, G.; Kostadinova, A.; Krasteva, N.; Albrecht, W.; Paul, D. J Biomed. Mater. Res. 1999, 44, 341. (23) Pompe T.; Markowski, M.; Werner, C. Tissue Eng., in press. (24) Capadona, J. C.; Collard, D. M.; Garcia, A. S. Langmuir 2003, 19, 1847. (25) Pettit, D. K.; Horbett, T. A.; Hoffman, A. S. J. Biomed. Mater. Res. 1992, 26, 1259-1275. (26) McClary, K. B.; Ugarova, T.; Grainger, D. W. J. Biomed. Mater. Res. 2000, 50, 428-439.

and exchange. In a first approach, the following simple linear model was used

y ) a0 +

aixi ∑ i)1

(3)

where substrate hydrophobicity, size of the adsorbed protein, and size of the exchange protein are the experimental parameters xi. Consecutively, their coefficients will be attributed by ai (i ) 1, 2, 3), respectively. For the multivariate regression analysis the proteinsubstrate interaction was assumed to depend on the substrate surface energy, which was characterized by contact angle measurements. The water contact angle for the used copolymer surfaces was determined to 100° and 38° ((3°) for poly(octadecene-alt-maleic acid) and poly(propene-alt-maleic acid), respectively. The substrate hydrophobicity is described by the water contact angle cos Θ resulting in an argument of the characteristic of surface tension. The parameters for the preadsorbed and exchange proteins were described in two ways. For the analysis of the protein amounts ΓA and ΓB the protein size was introduced as the area, which the protein covers at the substrate surface. Hence, the parameter relates to MW2/3 (MW, molecular weight) with the assumption that the radius r of a globular protein scales as MW1/3 and the area as r2. In the case of the time constants kA and kB, the parameters of the protein size were scaled by MW-1/3 due to the fact, that the diffusion constant D for a globular protein scales approximately with r-1, which can be derived via the Stokes equation and the frictional coefficient. In a first-order approach of (1). Fick’s law for the desorption process, the time constant k could be assumed to be proportional to D when comparing with the simple rate and coverage-dependent kinetics used in eq 1.27 MW for FN and HSA are 440 000 and 66 000, respectively. All parameters have been normalized before calculating the postulated coefficients. The overdetermined linear equation system was solved in the least-squares sense. The significance of the calculated parameters was tested with STATMOD software (Technische Universita¨t Dresden, Dresden, Germany) by means of a Student t test. Furthermore, the determination coefficient was calculated to estimate the quality of the applied model. Experimental Section Substrate Preparation. Thin films of poly(octadecene-altmaleic anhydride) (POMA) (Polysciences Inc., Warrington, PA) and poly(propene-alt-maleic anhydride) (PPMA) (Leuna-Werke AG, Germany) were produced by spin coating (RC5, Suess Microtec, Garching, Germany) of 0.08% and 0.1% copolymer solutions in tetrahydrofuran (Fluka, Deisenhofen, Germany), respectively, on top of glass coverslips. The coverslips had been freshly oxidized before in a mixture of aqueous solutions of ammonia (Acros Organics, Geel, Belgium) and hydrogen peroxide (Merck, Darmstadt, Germany) and were subsequently surface modified with (3-aminopropyl)dimethylethoxysilane (ABCR, Karlsruhe, Germany) prior spin coating of the copolymer solutions to allow a covalent fixation of the thin copolymer films. Stable covalent binding of the polymer films to the glass carriers was achieved by annealing at 120 °C for 2 h. The polymer films were thoroughly characterized with respect to water contact angle, film thickness, surface roughness, and chemical composition as published recently.19 In particular, the static advancing water contact angle was determined by the sessile drop method (droplet diameter approximately 2 mm) (G40, Kruess, Hamburg, Germany). Exposure of protein solution was performed with polymer surfaces after autoclaving. Autoclaving induced hydrolysis of (27) Iordanskii, A. L.; Dmitriev, E. V.; Kamaev, P. P.; Zaikov, G. E. J. Appl. Polym. Sci. 1999, 74, 595.

2930

Langmuir, Vol. 20, No. 7, 2004

the anhydride moieties to provide a surface exclusively bearing carboxylic acid groups. In control experiments covalent protein binding to anhydride moieties via free amine groups on lysine side chains was achieved on freshly annealed (120 °C, 2 h) polymer surfaces. Protein Labeling and Adsorption. For fluorescence experiments the proteins had to be labeled by the fluorescent dye tetramethylrhodamine isothiocyanate (TRITC). HSA-TRITC was used as commercially available (Rockland, Gilbertsville, PA). Unlabeled HSA was purchased from Sigma (Mu¨nchen, Germany). FN (Roche, Basel, Switzerland) was conjugated by the FluoReporter Tetramethylrhodamine Protein Labeling Kit (Molecular Probes, Eugene, OR). The average degree of FN labeling was determined according to the kit protocol with a spectrophotometer (Specord S10, Zeiss, Jena, Germany) to be 3.8 ((1.1) mol of dye per mol of protein. The effect of fluorescent self-quenching in the adsorbed protein layers could be excluded for the performed experiments. Labeling degrees of 10 mol of dye per mol of protein or higher were calculated to become critical with respect to selfquenching in the adsorbed protein layers from the known spacing of self-quenching octadecylrhodamine fluorochromes in lipid bilayers.28 Unlabeled proteins were used for the HPLC experiments (see below). For protein adsorption polymer surfaces were immersed in solutions of protein (FN and HSA at a concentration of 50 µg mL-1 in phosphate buffered saline (PBS) (Sigma, Steinheim, Germany) at pH 7.4) for 1 h. Desorption and Exchange. After 1 h of incubation, the surfaces were rinsed three times by PBS followed by addition of buffer (PBS) or exchange protein solution (FN or HSA concentrated 50 µg/mL) to start desorption and exchange processes, respectively. Fluorescence confocal laser scanning microscopy (TCS SP, Leica, Bensheim, Germany) was performed to reveal information from the plane of adsorption directly on the substrate surface. All experiments have been repeated at least three times. Laser intensities were calibrated prior each measurement by using the InSpeck Orange calibration kit (Molecular Probes). With the calibration beads of different intensity, a simple correlation between the photomultiplier gain of the confocal laser scanning microscopy (cLSM) and the measured intensities could be established to convert the measured relative intensities to an absolute scale. The conversion allows a more valuable comparison between the adsorbed protein amounts on both polymer surfaces. Furthermore, a direct comparison with high-performance liquid chromatography (HPLC) data could be performed to convert the qualitative cLSM data to a quantitative scale of the surface coverage. Additionally, comparison with HPLC-based protein quantification provided evidence that the fluorochrome labeling of the proteins did not affect their adsorption, desorption, and exchange characteristics. HPLC Quantification. The amount of immobilized FN and HSA on the copolymer surfaces after adsorption and after 24 h of exposure toward desorption or exchange solution was quantified by HPLC-based amino acid analysis of the protein mixtures as described elsewhere.29 Briefly, after acidic vapor phase hydrolysis, amino acids were fluorescence labeled, separated with a HPLC system, and quantified by a fluorescence detector (series 1100, Agilent Technologies, Bo¨blingen, Germany). The amounts of different proteins in adsorbed mixtures were determined by numerical analysis (by means of least-squares sense) utilizing the known amino acid sequences of HSA and FN.

Results and Discussion Adsorption. Upon the basis of preceding experimental studies,29 adsorption of HSA and FN onto the polymeric surfaces was quantified by HPLC-based amino acid analysis. Time-dependent and concentration-dependent adsorption experiments revealed an almost steady-state surface coverage at the chosen solution concentration of 50 µg mL-1 after 1 h of adsorption. Figure 1 indicates the (28) Hoekstra, D.; de Boer, T.; Klappe, K.; Wilschut, J. Biochemistry 1984, 23, 5675. (29) Salchert, K., Pompe, T., Sperling, C., Werner, C. J. Chromatogr. A 2003, 1005, 113.

Renner et al.

Figure 1. Initial adsorbed surface coverage of FN and HSA on POMA and PPMA (HPLC data).

Figure 2. Desorption and exchange of FN by PBS, FN, and HSA on POMA, relative amounts of preadsorbed protein measured by cLSM.

quantity of adsorbed protein on the hydrophobic POMA and hydrophilic PPMA surface. The differences in the protein surface concentration were found to be more pronounced for HSA. Furthermore, the amount of FN implies no simple side-on adsorption for the protein. An assumptive monolayer, supposing a sideon adsorption of a molecule with two arms of 60 nm length and 2.5 nm in diameter,30 is calculated to 0.25 µg cm-2. However, in a realistic scenario adsorbed protein layers cannot be considered to create a homogeneous structure. The gained amount of FN may be attributed to the heterogeneity in the orientation of the adsorbed protein molecules, because a heterogeneous protein layer could combine both, side-on and end-on orientation in a random distribution. A hypothetical HSA monolayer can be calculated to 0.24 µg cm-2.12 Hence, the measured surface concentration of HSA is compatible with, but does not necessarily represent, a complete monolayer of molecules adsorbed in a side-on orientation. Desorption and Exchange. On the basis of the adsorption experiments desorption and exchange processes were measured on POMA and PPMA copolymer surfaces. They revealed distinct differences for the compared polymer films and exposed proteins. The microscopy experiments showed a general dynamic trend as displayed in Figure 2. Figure 2 represents the desorption and the exchange process at POMA surfaces with preadsorbed FN, applying various exchange solutions. FN exerted a bigger impact (30) Human protein data; Haeberli, A., Ed.; Wiley-VCH Verlag: Weinheim, 1998.

Alterations of Fibronectin Layers

Langmuir, Vol. 20, No. 7, 2004 2931

Figure 3. Exchange of FN by FN on both polymer surfaces measured by cLSM. Relative amount of preadsorbed FN versus time (A) and negative rate versus relative amount of preadsorbed FN (B).

toward the displacement of the preadsorbed FN than HSA and PBS. FN was desorbed in the presence of PBS, as well as it was exchanged by added FN solutions, while HSA added to PBS solutions was hardly capable of displacing preadsorbed FN more efficiently than the pure PBS. Differences between the exchange process of preadsorbed FN by FN on POMA and PPMA (Figure 3A) demonstrate the impact of the physicochemical characteristics of the polymer substrates. A higher amount of FN was exchanged on the hydrophilic PPMA surface, correlating with a stronger anchorage of FN at hydrophobic surfaces. Noticeable in Figures 2 and 3A is a fast decline during the first few hours of observation and the attenuation of the curves observed toward the end. We detected this as a general dynamic trend in all displacement experiments, which is in agreement with earlier published findings.14,18 Huetz et al.14 reported data indicating a similar behavior for a model system of immunoglobulin molecules displaced by fibrinogen on a silica surface. The exchange mechanism was classified into a fast exchange step, followed by a slower step independent from solution containing molecules. The latter is apparently in contrast to our observation. Nevertheless, we have been able to describe the process of desorption and exchange with two exponential functions on the basis of this theoretical approach. Two examples are given in Figures 2 and 3A. Plotting the negative desorption rate versus the relative amount, a graph is gained, which resembles exponential growth (Figure 3B). In the early and late regime (at high or low relative amounts) the influence of the two different species can be clearly separated giving a two straight lines with different gradients corresponding to the coefficients -ki. With regard to this, the exponential approach is a reasonable explanation for the observed process. Furthermore, it is interesting to note that protein desorption and exchange were observed for FN on the hydrophobic POMA, which is in contrast to several earlier findings24-26 on various different hydrophobic surfaces. However, desorption of small fractions of adsorbed protein layers was also reported in the literature.18 This may indicate that protein adsorption phenomena do not simply depend on any single integral surface property, like hydrophobicity, but depend on a balance of different interactions including electrical charges, van der Waals force, hydrogen bonding, and entropic and structural forces, as convincingly summarized in ref 1. We have been able to reliably convert the relative amount of adsorbed protein into absolute intensities by

Table 1. Comparison of HPLC Data and Converted Absolute cLSM Data by Intensity Ratios APPMA/APOMA cLSM HPLC

FN

HSA

0.89 ( 0.2 0.92 ( 0.15

0.46 ( 0.17 0.62 ( 0.1

the use of calibration beads, as described in the materials and methods section. The conversion allows a more valuable comparison between the exchange dynamics of the protein amounts on both polymer surfaces and a rescaling to quantitative surface coverage by the HPLC data. The calibrated absolute fluorescence intensities can be shown to be comparable to the complementary HPLC data by introducing the ratio between the amounts adsorbed onto PPMA and POMA (APP/APO). We gained values, which differ in the range of the mean error (Table 1). We conclude that the HPLC data correspond with the converted absolute amount and the exchange dynamics measured by cLSM can be rescaled to a real surface coverage. In this context the homo- and heteromolecular exchange of FN on the different copolymers could be analyzed in more detail. The results shown in Figure 4 for the homomolecular exchange indicate an exchange and further adsorption process on both polymer surfaces. The HPLC data indicate that after the initial adsorption time of 1 h the amount of adsorbed FN increases further on, however very slowly. The process can be attributed to conformational changes, remodeling within the adsorbed protein layer or exchange of protein with adsorption in different orientations. Such processes are already known in the literature to occur over long time scales.1,8,31,32 The dynamic exchange process is illustrated by the cLSM data indicating the amount of remaining FN from the initial adsorption step with a more intense exchange on the hydrophilic PPMA surface. A similar behavior was measured for the heteromolecular exchange of FN on the two copolymer surfaces (Figure 5). On the hydrophilic PPMA substrate more FN is displaced by HSA. Again, an increase in the total amount of adsorbed protein could be observed which can be attributed to similar conformational changes and exchange processes as observed for the homomolecular system. The exchange dynamics indicates clearly a higher affinity of FN toward hydrophobic surfaces. Furthermore, Figure 5 provides further evidence for the good agreement of HPLC (31) Calonder, C.; Tie, Y.; Van Tassel, P. R. Proc. Natl. Acad. Sci. U.S.A. 2001, 98, 10664. (32) Nygren, H. Biophys. J. 1993, 65, 1508.

2932

Langmuir, Vol. 20, No. 7, 2004

Renner et al.

Figure 4. Homomolecular exchange of FN on POMA (A) and on PPMA (B), measured by HPLC and cLSM.

Figure 5. Heteromolecular exchange of FN by HSA solutions on POMA (A) and PPMA (B) (cLSM and HPLC data).

and cLSM analysis. Additionally, it proves the reliability of the cLSM measurements. While the influence of the fluorochrome on the adsorption and the exchange properties of the proteins might be critical in the cLSM experiments, the agreement with the HPLC data of the unlabeled proteins proves that the label has no impact on the adsorption and the exchange processes. While HPLC provides more quantitative data, the method is limited to a smaller number of time points for the measurement because of its complexity and its susceptibility to possible contaminations of the samples. To verify the stability of the dye in the cLSM experiments, freshly annealed polymer surfaces were used to bind the labeled proteins covalently onto the polymer films. Exposed toward buffer, the experiment revealed that the fluorescent dye was totally bound to the protein and no release occurred. An apparent loss of intensity (estimated to a maximum of 10%) can be explained by a small amount of noncovalently bound protein, which easily emerged into the surrounding medium. After performing 200 measurements on one sample, the dye lost around 5% of its intensity. Hence, photobleaching is not considered to distort the measurement, because standard experiments were done with only up to 60 measurements per sample. As the aim of the current study was directed toward the understanding of FN interaction with different copolymer surfaces at conditions of desorption and displacement, we can qualitatively resume at first that FN exhibits a higher affinity to the hydrophobic POMA substrates as compared to the more hydrophilic and more negatively charged PPMA substrates. A dependence of desorption and displacement on the kind of exchange solution was observed indicating that FN is most efficient with respect to displacing preadsorbed FN (homodisplacement) while

Table 2. Parameters of Evaluated Model Containing Only Significant Parameters after Reduction by t Test Analysis (R2, Determination Coefficient) parameter

kA

kB

ΓA

ΓB

a0 a1 a2 a3 R2

0.64

0.007 -0.0005 0.41

0.015 -0.06 0.14

0.03 0.57

0.015 -0.04 0.195 0.05 0.93

0.86

pure PBS and HSA solutions in PBS could exchange the preadsorbed FN layers to lower and similar degrees. Linear Model Analysis. To evaluate the influence of the different copolymer surfaces on the different types of exchange processes in a more general and quantitative way, a multivariate regression analysis was performed using a simple linear model (eq 3). The set values of this model were elaborated from the kinetic parameters of the exponential fits (e.g., Figures 2 and 3A). With regard to the intensity calibration of cLSM data and the good agreement of the converted absolute amounts with HPLC data (Table 1), we were able to calculate the model with the absolute surface concentrations. The use of absolute amounts pinpoints a further quality of the model, because the differences of the distinct polymer surfaces for the adsorbed, desorbed, and exchanged proteins can be considered in terms of the total surface concentration and the ratio of both proteins. Results of the multivariate regression analysis are shown in Table 2. The model parameters were found to clarify highly relevant properties of proteins and surfaces. The determination coefficient R2 indicates the quality of the applied model. The regression for the fit of the time constants kA and kB is not fully satisfying but nevertheless provides an explanation for the tendencies characterizing

Alterations of Fibronectin Layers

exchange processes. Because of the simplicity of the model, it certainly cannot describe the protein exchange mechanism in a fully satisfying way. However, the utilized approach may demonstrate the possibility to generalize information gained from rather complex experimental settings, which is currently not achieved in the vast majority of studies on proteins at interfaces. Further refinement of the model and implementation of more experimental data from different surfaces and other proteins will be certainly necessary to improve this approach. Modeling the interaction parameters ai in the reported way was carefully scrutinized by analyzing the statistical significance and the determination coefficient of multivariate regression analysis. Diffusion- and adsorption-limited regimes were tested as well as different concepts for the incorporation of a surface energy parameter via a Boltzmann activation term. On the basis of these attempts, the best-fit model parameters were applied for the final analysis as described in the introductory section. A surface energy parameter derived from contact angle data and the molecular weight of the protein introduced as a second parameter determining the protein area on the surface and the protein diffusion coefficient were considered to influence the adsorbed amount and the time constants of the exchange process. kA, representing the initial displacement phase, depends significantly on the applied exchange solution. The same is true for kB although it is obvious by the low determination coefficient that the model cannot provide the real picture of the process. According to this fact homo- and heteromolecular exchanges are dependent on the provided exchange solution. Interestingly, the amounts of the two species of preadsorbed protein ΓA and ΓB depend strongly on the surface hydrophobicity and the size of the preadsorbed protein. Additionally, the fast desorbing species ΓA is influenced by the exchange solution. This correlates very well with the results reported by Huetz14 claiming that the slow desorption step is independent of the molecules in solution. In contrast, the analysis of the time constants kA and kB indicated a dependency on the exchange proteins as stated above. Further predictions of the protein exchange behavior could be drawn utilizing the estimated coefficients ai and their algebraic signs, which affects the direction of impact. Up to now there is no direct physical law at hand describing protein-exchange processes in general. However, with our approach of using physically meaningful parameters xi as outlined in the introduction section, certain correlations can be found and described in a quantitative way. For the amount of adsorbed protein ΓA and ΓB, it can be stated that surface energy is a crucial parameter: The amounts ΓA and ΓB increase with lowering the surface tension (increasing hydrophobicity) and with the area covered by the proteins (increased interaction area). Both parameters point toward the well-known strong interaction of proteins with hydrophobic low-energy surfaces1 due to entropically favored hydrophobic dehydration. (Note that the structural features of the substrates were not considered in this evaluation but may additionally contribute to the observed graduation of the desorption and displacement effects.) Conclusions Significant dynamic alterations of preadsorbed FN layers on polymer substrates were demonstrated to occur

Langmuir, Vol. 20, No. 7, 2004 2933

in dependence on both the characteristics of the polymer substrate and the solution applied for desorption or displacement. A stronger anchorage of FN was in general observed on the more hydrophobic POMA. While for all investigated cases the most efficient exchange of the preadsorbed FN layer was achieved in the homodisplacement (i.e., by FN from solution), about similar degrees of desorption were obtained in pure PBS and in HSA solution. An increase of the total adsorbed amount of protein due to additional adsorption of FN or HSA accompanied the partial exchange of the preadsorbed FN in the displacement experiments. In this study, emphasis was put on the experimental determination of the kinetics of FN desorption and displacement. For that purpose, a novel combination of analytical methods was established to obtain quantitative in situ information on the composition of binary layers containing preadsorbed FN and a second type of protein (FN or HSA): HPLC-based amino acid analysis was utilized to verify the quantification of protein surface concentrations achieved from fluorescence intensities determined by cLSM and to translate these relative fluorescence intensities into absolute values. With this approach, desorbed and displaced FN from both POMA and PPMA could be classified into two distinct populations according to the dynamics of removal from the substrate. Parameters, fitted by the assumption of an exponential decay, were used to reveal the influence of surface hydrophobicity and size of the preadsorbed and exchanged protein in a linear multivariate analysis. The results show that the degree of desorption/displacement is determined by the surface characteristics of the polymer substrate pointing at a stronger anchorage of FN at the more hydrophobic POMA. The time constants for the protein exchange indicated a dependency only on the kind of exchange protein and could be described in a diffusion dependent manner. The experiments were designed to analyze the fate of preadsorbed FN layers on cell culture carriers at application-related settings. From the results obtained we may conclude that a fractional desorption/displacement of the preadsorbed protein will be hardly avoidable while the characteristics of both the substrate and the solution determine the degree and dynamics of the layer alteration as recently reported.24 The retained amount of predeposited FN can be considered sufficient to trigger cell adhesion in all of the analyzed cases. For the evaluation of the protein displacement kinetics at rather complex settings, a linear multivariate analysis was introduced. The approach advantageously provides quantitative parameters to describe a variety of factors affecting the protein-substrate interaction. Despite of several limitations, the present study may demonstrate a valid strategy to analyze complex interfacial phenomena in a generalized manner. Acknowledgment. Assistance and provision of STATMOD software by Dr. Boehlmann (Technische Universita¨t Dresden, Department of Process Engineering) is gratefully acknowledged. This work was granted in part by the Bundesministerium fu¨r Bildung, Forschung und Technologie, Berlin, Germany, as BMBF-Kompetenzzentrum fu¨r Materialien im Blut- und Gewebekontakt (Grant No. 03N4022). LA0362627