Fibrinogen Adsorption on Hydrophilic and Hydrophobic Surfaces

---

706

Langmuir 2002, 18, 706-715

Fibrinogen Adsorption on Hydrophilic and Hydrophobic Surfaces: Geometrical and Energetic Aspects of Interfacial Relaxations Christian F. Wertz and Maria M. Santore*,† Department of Chemical Engineering, Lehigh University, 111 Research Drive, Bethlehem, Pennsylvania 18015 Received July 13, 2001. In Final Form: October 24, 2001 This work examined the relationship between footprint growth and adsorption energetics for fibrinogen on model hydrophobic and hydrophilic surfaces. For adsorption runs at different free solution concentrations and flow rates in a slit-shear flow cell, single surface-dependent relaxation times were found on each of two model surfaces. By use of each relaxation time in an exponential growth model, predictions of the footprint size distributions were made for different adsorption histories. Then, from desorption rate measurements and a reversible binding model, the sizes of loosely and tightly bound populations were determined, in addition to the binding energy of the loosely bound population. These binding energies were then compared with the footprint size distributions to reveal markedly different behavior on the hydrophobic and hydrophilic surfaces. On the hydrophobic surface, a single footprint size correlated with a binding energy of 6kT, a feature that was independent of adsorption history and the footprint size distribution. On the hydrophilic surface, the footprint size associated with a similar binding energy for the loosely bound population depended on the adsorption history. These observations are discussed in the context of potentially different relaxation mechanisms on the two surfaces.

Introduction One of the first processes to occur upon contact between solid surfaces and blood is the adsorption of plasma proteins at the interface.1 Of the numerous proteins available for investigation, fibrinogen has been studied most intensely due to its abundance in plasma,2 its role in coagulation,3-5 and its ability to promote platelet adhesion.6-8 Also, the conformation and/or orientation of adsorbed fibrinogen on the surface is believed to play a major role in determining the biocompatibility of a particular material. Therefore, a detailed understanding of fibrinogensurface interactions and binding energies is needed in order to facilitate better control over biological response to certain materials. It is widely accepted that conditions of high protein flux to the interface result in different fibrinogen adsorption behavior than conditions of low protein flux, indicating a history dependence on the ultimate surface coverage.9-13 This history dependence, along with the essential irreversibility of fibrinogen adsorption to most surfaces, has precluded a rigorous analysis of binding data according to the conventional Langmuir adsorption isotherm.14,15 On the basis of nonequilibrium data, it has been suggested †

Present address: Department of Polymer Science and Engineering, University of Massachusetts, Conte Building, 120 Governors Dr., Amherst, MA 01003. (1) Baier, R. E.; Dutton, R. C. J. Biomed. Mater. Res. 1969, 3, 191. (2) Slack, S. M.; Horbett, T. A. J. Biomed. Mater. Res. 1992, 26, 1633. (3) Brash, J. L. In Protein Interactions with Artificial Surfaces; Salzman, E. W., Ed.; Marcel Dekker: New York, 1981; p 37. (4) Horbett, T. A.; Weathersby, P. K.; Hoffman, A. S. J. Bioeng. 1976, 1, 61. (5) Barbucci, R.; Magnami, A. Biomaterials 1994, 15, 955. (6) Packham, M. A.; Evans, G.; Glynn, M. F.; Mustard, J. F. J. Lab. Clin. Med. 1969, 73, 686. (7) Phillips, D. R.; Charo, I. F.; Parise, L. V.; Fitzgerald, L. A. Blood 1988, 71, 831. (8) Plow, E. F.; Ginsberg, M. H. In Progress in Hemostasis and Thrombosis; Coller, B. S., Ed.; W. B. Saunders: Philadelphia, PA, 1989; p 117.

that fibrinogen can exist in two distinct conformations on surfaces,15,16 resulting in two experimentally observable populations: a larger, irreversibly adsorbed layer and a smaller, reversibly adsorbed layer.17-19 However, recent work by our lab has suggested that fibrinogen can exist in many possible orientations and/or conformations, depending on the adsorption history and the surface chemistry of the substrate.12,13 Here, fibrinogen adsorbs nonspecifically to both hydrophobic and hydrophilic surfaces, resulting in a random mixture of end-on and side-on orientated molecules initially. Following attachment to the surface, fibrinogen begins to increase its footprint (i.e., number of segment-surface contacts) in a manner that is consistent with denaturation (unfolding) on hydrophobic surfaces and reorientation (rolling over) on hydrophilic surfaces. In this context, reversibly and irreversibly bound populations may consist of many different protein footprint sizes, not only as two distinct conformations. Based on the experimentally observable existence of reversible and irreversible adsorbed populations, most of the recent theoretical work on protein adsorption only considers the possibility of two adsorbed states.20-28 Generally, proteins modeled in this manner can only (9) Nygren, H.; Stenberg, M. J. Biomed. Mater. Res. 1988, 22, 1. (10) Brynda, E.; Houska, M.; Lednicky, F. J. Colloid Interface Sci. 1986, 113, 164. (11) Nygren, H.; Stenberg, M.; Karlsson, C. J. Biomed. Mater. Res. 1992, 26, 77. (12) Wertz, C. F.; Santore, M. M. Langmuir 1999, 15, 8884. (13) Wertz, C. F.; Santore, M. M. Langmuir 2001, 17, 3006. (14) Horbett, T. A.; Brash, J. L. In Proteins at Interfaces; Brash, J. L., Horbett, T. A., Eds.; American Chemical Society: Washington, DC, 1987; p 1. (15) Schmitt, A.; Varoqui, R.; Uniyal, S.; Brash, J. L.; Pusineri, C. J. Colloid Interface Sci. 1983, 92, 25. (16) Retzinger, G. S.; Cook, B. C.; DeAnglis, A. P. J. Colloid Interface Sci. 1994, 168, 514. (17) Bosco, M.; Chan, C.; Brash, J. L. J. Colloid Interface Sci. 1981, 82, 217. (18) MacRitchie, F. J. Colloid Interface Sci. 1972, 38, 484. (19) Brash, J. L.; Uniyal, S. J. Polym. Sci. 1979, C66, 377.

10.1021/la011075z CCC: $22.00 © 2002 American Chemical Society Published on Web 12/29/2001

Fibrinogen Adsorption on Surfaces

adsorb in a single, reversible state. However, once attached, these molecules may undergo a transition to a second adsorbed state, which is usually considered to be irreversibly bound to the surface. This transition from a reversible to an irreversible adsorbed state is often associated with a conformational change of the adsorbed molecule. Although these models may be appropriate for certain systems,28 recent work conducted in our lab,12,13,29 as well as in other labs,30,31 suggests that multiple adsorbed conformations cannot be neglected. To this end, kinetic models allowing for the existence of multiple surface conformations have been developed.32,33 Although these models may provide better predictive capabilities for protein-surface systems associated with large extents of conformational change, very few comparisons between experimental data and theoretical models can be found in the literature. Motivated by the importance of fibrinogen in protein adsorption applications and the absence, in the current literature, of studies addressing the distribution of energetic and areal states of adsorbed proteins, the current work examines the relationship between footprint growth and adsorption energetics. Taking a first-order approach to interpret fibrinogen behavior on model hydrophobic and hydrophilic surfaces, the current study obtains simplistic estimates of footprint size distributions and adsorption energies: on the basis of relaxation times inferred from measurements by total internal reflectance fluorescence (TIRF), kinetic data were interpreted in the context of an exponential growth model in order to quantify footprint size distributions for different adsorption histories. These distributions were then compared with adsorption energies distinguishing loosely and tightly bound populations, determined from desorption kinetics and a reversible binding model. This made possible the isolation of potentially different relaxation mechanisms for fibrinogen on the hydrophobic and hydrophilic surface. Exponential Growth Model The exponential growth model is based on the “growing disk” model originally proposed by Pefferkorn and Elaissari34 and further developed by Van Eijk and Cohen Stuart32 In this model, individual protein molecules adsorb on the surface with an initial area, a0. Immediately following the attachment process, the protein molecules begin to unfold and/or reorient themselves on the surface, (20) Beissinger, R. L.; Leonard, E. F. J. Colloid Interface Sci. 1981, 85, 521. (21) Lundstrom, I.; Elwing, H. J. Colloid Interface Sci. 1989, 136, 68. (22) Norde, W.; Haynes, C. A. In Proteins at Interfaces II: Fundamentals and Applications; ACS Symposium 602; Brash, J. L., Horbett, T. A., Eds.; American Chemical Society: Washington, DC, 1995; p 26. (23) Kurrat, R.; Prenosil, J. E.; Ramsden, J. J. J. Colloid Interface Sci. 1997, 185, 1. (24) Van Tassel, P. R.; Viot, P.; Tarjus, G. J. Chem. Phys. 1997, 106, 761. (25) Van Tassel, P. R.; Guemouri, L.; Ramsden, J. J.; Tarjus, G.; Viot, P.; Talbot, J. J. Colloid Interface Sci. 1998, 207, 317. (26) Brusatori, M. A.; Van Tassel, P. R. J. Colloid Interface Sci. 1999, 219, 333. (27) Talbot, J. J. Chem. Phys. 1997, 106, 4696. (28) Wertz, C. F.; Santore, M. M. Langmuir, in press. (29) Wertz, C. F.; Santore, M. M. Manuscript in preparation. (30) Brynda, E.; Houska, M.; Lednicky, F. J. Colloid Interface Sci. 1986, 113, 164. (31) Wahlgren, M.; Arnebrant, T.; Lundstrom, I. J. Colloid Interface Sci. 1995, 175, 506. (32) Van Eijk, M. C. P.; Cohen Stuart, M. A. Langmuir 1997, 13, 5447. (33) Minton, A. P. Biophys. J. 1999, 76, 176. (34) Pefferkorn, E.; Elaissari, A. J. Colloid Interface Sci. 1990, 138, 187.

Langmuir, Vol. 18, No. 3, 2002 707

increasing their footprint. The area occupied by each molecule increases with a growth rate characterized by a spreading time, τs, and a limiting footprint size, a0 + R. For a protein molecule adsorbing at time t′, its occupied area at some later time, t, is given by

a(t,t′) ) a0 + R(1 - e-(t-t′)/τs)

(1)

On the basis of eq 1, early-arriving molecules have more time to relax on the surface than those arriving later, resulting in a distribution of protein footprints at any time during the adsorption process. The adsorption history begins at t ) 0, and the adsorption rate, dΓ/dt, is proportional to the free surface area fraction, β:

dΓ/dt ) Jβ

(2)

where Γ is the number of protein molecules per unit area and J is the flux of protein that would occur on a bare surface under the same adsorption conditions, when β ) 1. During a time interval dt′, the number of adsorbed protein molecules will increase by an amount dΓ ) Jβ(t′) dt′. At a later time t, this population will contribute to the fraction of occupied area, 1 - β, by an amount Jβ(t′)a(t,t′) dt′. Therefore, the time dependence of the free surface area fraction is found by integrating over all values of t′ (when each population initially adsorbed):

β(t) ) 1 - J

∫0t dt′ β(t′)[a0 + R(1 - e-(t-t′)/τs)]

(3)

Substituting the dimensionless variables

J* ) Ja0τs

t* ) t/τs

R* ) R/a0

(4)

into eq 3, one obtains

β(t*) ) 1 - J*

∫0t* dt′* β(t*){1 + R*[1 - e-(t*-t′*)]}

(5)

After eq 5 is differentiated twice with respect to t*, the following linear, homogeneous, second-order differential equation results:32

dβ d2β + J*(1 + R*)β(t*) ) 0 + (1 + J*) dt* dt*2

(6)

Equation 6 is solved with the following boundary conditions:

β(0) ) 1

dβ (dt* )

) -J*

t*)0

(7)

Equation 6 is the analogue of the differential equation describing the movement of a damped pendulum and has been solved previously.32 After solving for the time dependence of the free surface area fraction, β(t*), the adsorbed amount is calculated by integrating β(t*) in eq 2:

Γ* ) J*

∫0t* dt′* β(t′*)

(8)

where Γ* is also equal to Γa0. It is important to note that eq 8 is physically meaningful only for increasing Γ*, even though a full solution will give Γ*(t*), which oscillates in time. The model is not intended to describe reversible adsorption, and only the solutions of the integral in eq 8 until the time where β(t*) ) 0 for the first time (Γ ) Γmax) are retained. The oscillations beyond the time of surface saturation are a

708

Langmuir, Vol. 18, No. 3, 2002

result of the general form of eq 6 and are not meaningful in the context of eq 5. A few points about the exponential growth model are worth noting. First, this treatment was chosen because of its ability to quickly and easily predict the distribution of interfacial protein footprints, though we admit it is a much simpler approach than other treatments in the literature that account for the influence of finite surface loading on the adsorption probability.24-27 The strength of the current treatment is, however, that it makes no assumptions (e.g., disks or spheres) about the shape of the initially adsorbed proteins. Statistical mechanical treatments that rely heavily on such knowledge, though they are fundamentally correct in their treatment of the influence of surface excluded area on the adsorption rate, suffer from the constraints of the geometrical assumption. The second point about the exponential growth treatment is that it works remarkably well, as we show below. This alone is worth noting, despite the flaws in the model. With these caveats in mind, we proceed to employ the growing disk model to estimate a characteristic relaxation time for the adsorbed proteins and to provide an estimate for the distribution of interfacial footprints. Experimental Section Bovine plasma fibrinogen, type IV, was purchased from Sigma (catalog no. F-4753) and was 95% clottable.35 Gel electrophoresis has shown no detectable impurities,36 and consistent results in the current study have been obtained for the same fibrinogen product from different lots. We have previously demonstrated how contaminants would affect our data and its interpretation and why our protein-handling procedures yield materials of adequate purity.12 Fluorescein isothiocyanate was covalently attached to fibrinogen by reaction at room temperature in carbonate buffer for several hours, according to established procedures.37 Free fluorescein and other potential contaminants were removed from the protein solutions by size-exclusion chromatography with a Bio-Gel P-6 polyacrylamide gel column (Bio-Rad). The column eluent was phosphate buffer, such that the purified, labeled product was at pH 7.4 rather than the pH 9 corresponding to the reaction solution. The extent of fluorescein labeling was measured with absorbance at 494 nm and, for different labeling batches, was between 0.6 and 1.2 labels/fibrinogen molecule. The buffer solutions employed salts from Fisher Scientific. The phosphate buffer was composed of 0.008 M Na2HPO4 and 0.002 M KH2PO4. The carbonate buffer was made of 0.004 M Na2CO3 and 0.046 M NaHCO3. C16 (hexadecyltrichlorosilane, Huls) self-assembled monolayers served as model hydrophobic surfaces. They were formed on microscope slides, as described previously,38 resulting in water contact angles ranging from 109° to 111°. OH [N-[3-(triethoxysilyl)propyl]-4-hydroxybutyramide, Gelest] self-assembled monolayers were used as model uncharged, hydrophilic surfaces. OH monolayers were also formed on microscope slides,39 resulting in water contact angles ranging from 54° to 56°. TIRF (total internal reflectance fluorescence) was used to measure protein adsorption kinetics from gentle shearing flow. Our TIRF instrument, which has been described in detail previously,40,41 employs an Ar+ ion laser at 488 nm for excitation and a photon counting detector. The microscope slide substrates, which comprised one wall of the adsorption flow cell, were (35) Protein characterization data provided by Sigma Chemical Co. (36) Malmstem, M.; Johansson, J.-A.; Burns, N. L.; Yasuda, H. K. Colloids Surf. B: Biointerfaces 1996, 6, 191. (37) Robeson, J. L.; Tilton, R. D. Biophys. J. 1995, 68, 2145. (38) Chaudhury, M. K.; Whitesides, G. M. Langmuir 1991, 7, 1013. (39) P. E. Laibinis, Massachusetts Institute of Technology, private communication. (40) Kelly, M. S.; Santore, M. M. Colloids Surf. A: Physiochem. Eng. Aspects 1995, 96, 199. (41) Rebar, V. A.; Santore, M. M. Macromolecules 1996, 29, 6263.

Wertz and Santore optically coupled to a waveguide prism with index-matching oil. The cell was made of a black Teflon block into which a 0.5 mm deep channel was machined, per modification of Shibata’s design.42 Adsorption experiments were conducted in a slit shear flow cell, which continually replenishes the bulk solution and maintains a constant bulk protein concentration. TIRF can be used in single-species studies to track adsorption kinetics directly, or labeled populations can be monitored in more complex mixtures to probe competitive behavior.43 The calibration to convert fluorescence to surface coverage identifies regions of transport-limited kinetics and employs known values for the diffusion coefficients, as discussed previously.12 Also, in fluorescence tracer experiments, two issues must be addressed: the potential invasiveness of the fluorescent label, and the proper interpretation of the fluorescent signal. A thorough discussion,12 based on previous work,44-49 has shown the fluorescein to be noninvasive in protein studies at the concentrations currently employed and the fluorescence signal to be proportional to the interfacial mass of the tagged species.

Results and Discussion Adsorption Dynamics and Footprint Distributions. Before we compared the exponential growth model to experimental kinetic traces, values for the model parameters were determined. From the definition of the dimensionless variables in eq 4, the free parameters were the initial fibrinogen footprint a0, the maximum unfolded footprint a0 + R, the flux of protein to the interface J, and the characteristic spreading time τs. a0 and R were assigned on the basis of our previous studies of fibrinogen adsorption, while J depends on the flow geometry and rate, leaving only τs to fit multiple adsorption runs for varied concentration and wall shear rate. In our previous studies of fibrinogen adsorption onto hydrophobic and hydrophilic SAMs, fibrinogen initially adsorbed in either an end-on or a side-on configuration, with the distribution of end-on and side-on adsorbed molecules the same on both SAMs.13 Therefore, the initial fibrinogen footprint was identical on both surfaces and equal to 100 nm2/molecule (a0 ) 0.177 m2/mg). To determine the footprint of fully relaxed fibrinogen molecules on both the hydrophilic and hydrophobic SAMs, additional studies of unsaturated fibrinogen layers were conducted.29 On the basis of these measurements, the maximum footprints have been determined: 175 nm2/ molecule (R ) 0.133 m2/mg, R* ) 0.75) on the OH SAM and 475 nm2/molecule (R ) 0.664 m2/mg, R* ) 3.75) on the C16 SAM. The flux term J, for use in eq 4 is fixed, depending only on the limiting zero-coverage flux (transport-limited flux) and the protein diffusivity. When steady-state conditions are realized, the flux is described by

J ) M(C - C*)

(9)

where M is the mass transfer coefficient, C is the bulk solution concentration, and C* is the concentration of protein near the interface. For high-affinity adsorption and at low surface coverage levels, C* can be neglected. (42) Shibata, C. T.; Lenhoff, A. M. J. Colloid Interface Sci. 1992, 148, 485. (43) Fu, Z.; Santore, M. M. Macromolecules 1998, 31, 7014. (44) Fu, Z.; Santore, M. M. Colloids Surf. A: Physiochem. Eng. Aspects 1998, 135, 63. (45) Rodelez, F.; Ausserre, D.; Hervet, H. Annu. Rev. Phys. Chem. 1987, 38, 317. (46) Rebar, V. A.; Santore, M. M. J. Colloid Interface Sci. 1996, 178, 29. (47) Fu, Z.; Santore, M. M. Langmuir 1998, 14, 4300. (48) Lui, X.; Kim, D. J.; Santore, M. M. Langmuir, submitted for publication. (49) Mubarekyan, E.; Santore, M. M. J. Colloid Interface Sci. 2000, 227, 334.

Fibrinogen Adsorption on Surfaces

Langmuir, Vol. 18, No. 3, 2002 709

Figure 1. (A, top panel) Fibrinogen adsorption kinetics on the OH SAM at a wall shear rate of 5 s-1 and varying bulk solution concentration of 25, 50, and 100 mg/L. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 0.75, and τs ) 700 s. (B, bottom panel) Fibrinogen adsorption kinetics on the OH SAM at a bulk solution concentration of 25 mg/L and varying wall shear rate of 2, 5, and 30 s-1. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 0.75, and τs ) 700 s.

Further, for slit shear flow cells such as ours, M is given by

M ) 0.538(γ/L)1/3D2/3

(10)

where γ is the wall shear rate, L is the distance from the entrance of the flow cell to the point of observation, and D is the diffusion coefficient. For the concentrations and wall shear rates used here, the initial adsorption kinetics have been shown to be transport-limited on both surfaces;13 therefore, the initial adsorption rate obeys

dΓ/dt ) J

(11)

By use of eqs 9-11, the flux of fibrinogen was determined for each experiment in which the bulk solution concentration and the wall shear rate are varied:

J ) 0.538(γ/L)1/3D2/3C

(12)

At this point, the only parameter remaining in eq 8 is the characteristic spreading time, τs, which was determined from comparison between the model and experimental data. A single value of τs was able to adequately predict the initial adsorption rate and ultimate surface coverage for all runs in which the bulk solution concentration and wall shear rate were varied. Figure 1 shows the fit of the model to fibrinogen kinetic traces on the hydrophilic SAM. In Figure 1A, the bulk solution concentration is varied at a constant wall shear rate of 5 s-1, while in Figure 1B, the wall shear rate is varied at a constant bulk solution concentration of 25 mg/ L. The fit gives a value for τs of 700 s. Similarly, Figure 2 shows the fit of the model to fibrinogen kinetic traces on the hydrophobic SAM. In Figure 2A, the bulk solution concentration is varied at a constant wall shear of 5 s-1,

Figure 2. (A, top panel) Fibrinogen adsorption kinetics on the C16 SAM at a wall shear rate of 5 s-1 and varying bulk solution concentration of 25, 50, and 100 mg/L. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 3.75, and τs ) 1500 s. (B, bottom panel) Fibrinogen adsorption kinetics on the C16 SAM at bulk solution concentration of 25 mg/L and varying wall shear rate of 1.5, 5, and 22 s-1. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 3.75, and τs ) 1500 s.

while in Figure 2B, the wall shear rate is varied at a constant bulk solution concentration of 50 mg/L. Here, a fit of the model to experimental data gives a value for τs of 1500 s. As Figures 1 and 2 indicate, the model reasonably predicts both the initial adsorption rate, dΓ*/dt*, as well as the ultimate surface coverage, Γ*max, on both SAMs. It is important to note that, at short times, the initial adsorption rate is slower than that predicted by the model; however, this is a result of axial diffusion of fibrinogen during the initial cell-filling process (before a constant concentration gradient is achieved) and not a physical breakdown of the model. Also, several of the runs in Figures 1 and 2 show a slight increase in the adsorbed amount with time after the surface saturates. Here, it is important to keep in mind that the model only describes the adsorption process up until the moment the free surface area fraction goes to zero and the surface saturates. It does not take into account the possibility that at the highest surface coverages, where small amounts of open area are still available for adsorption, there might be a slow adsorption mechanism involving specific regions of the adsorbing protein or rearrangements of those already adsorbed. To determine the error in our determination of τs, Figure 3 examines how variations in τs affect the fit of single experimental runs. Figure 3 presents the fit of the model to kinetic runs on both the hydrophilic (Figure 3A) and hydrophobic (Figure 3B) SAMs at a bulk solution concentration of 100 mg/L and a wall shear rate of 5 s-1. In Figure 3A, τs has been varied from 400 to 1000 s, while in Figure 3B, τs has been varied from 1200 to 1800 s. In each figure, the same kinetic run has been plotted three times; however, because of the different τs values being examined for the fit, different dimensionless times, t*,

710

Langmuir, Vol. 18, No. 3, 2002

Figure 3. (A, top panel) Fibrinogen adsorption kinetics on the OH SAM at bulk solution concentration of 100 mg/L and a wall shear rate of 5 s-1. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 0.75, and τs ) 400, 700, and 1000 s. (B, bottom panel) Fibrinogen adsorption kinetics on the C16 SAM at bulk solution concentration of 100 mg/L and a wall shear rate of 5 s-1. The smooth lines represent the fit of the exponential growth model for a0 ) 0.177 m2/mg, R ) 3.75, and τs ) 1200, 1500, and 1800 s.

give the appearance of separate experimental runs. As Figure 3 shows, reducing τs below the previously determined values (700 s on OH SAM, 1500 s on C16 SAM) leads to faster predicted relaxation and lower overall coverage levels than what is observed experimentally. Conversely, increasing τs above the predetermined values leads to a slower predicted relaxation rate and higher ultimate coverage levels than those observed experimentally. Considering that a single value of τs was able to reasonably predict the initial adsorption rate and ultimate surface coverage for all variations in concentration and wall shear rate on both SAMs, Figure 3 reveals that the model is quite sensitive to the value of τs determined by comparison to experimental data and that differences in τs are meaningful. It is important to consider what Figures 1 and 2 tell us about the unfolding process and relaxation rates for fibrinogen on the two different SAMs. Fibrinogen is a 340 kDa dimeric molecule consisting of two sets of three intertwined polypeptide chains (AR, Bβ, and γ) held together by 29 disulfide bonds.50-52 Electron microscopy studies show that the molecule can be described as a trinodular structure in which the N-terminal regions of all six chains are folded into a globular, central E domain and each set of C-terminal regions is folded into two globular, outer D domains.53 The amino acid residues between the D and E domains fold into an R-helical structure with all three chains intertwined to form a linear coiled-coil region.54 (50) Hoeprich, P. D.; Doolittle, R. F. Biochemistry 1983, 22, 2049. (51) Garlund, B.; Hessel, B.; Marguerie, G. Eur. J. Biochem. 1977, 77, 595. (52) Hantgan, R. R.; Francis, C. W.; Marder, V. J. In Hemostasis and Thrombosis: Basic Principles and Clinical Practice, 3rd ed.; Colman, R. W., Hirsh, J., Marder, V. J., Salzman, E. W., Eds.; J. B. Lippincott: Philadelphia, PA, 1994; p 277.

Wertz and Santore

From the perspective that fibrinogen possesses a complex, multidomain structure, it is not surprising that the exponential growth model (with its distribution of possible adsorbed states) can predict the experimental data. A number of different processes potentially contribute to fibrinogen’s interfacial footprint growth on the C16 SAM. For instance, molecules adsorbed in an end-on configuration may reorient to the side-on configuration in order to form additional segment-surface contacts. Furthermore, the D and E domains do not have the same affinity for a particular surface and can achieve different extents of conformational change as the adsorption proceeds.55 These domains can unfold independently of each other, resulting in multiple unfolded states existing on the surface at any given time. Another possibility is that the unfolding or reorienting of one domain may trigger the cooperative unfolding of a neighboring domain, leading to a pathway of different intermediate states. Regardless of the true spreading mechanisms, a distribution of footprints exists, ranging in size from native, end-on adsorbed molecules to side-on adsorbed molecules that are substantially denatured.12,13,29 Many intermediate footprints are also possible, with the orientation of each molecule and the extent of unfolding depending on the surface chemistry of the substrate, the orientation of the molecule upon adsorption, and lateral interactions that may stunt unfolding of a particular domain. To gain a better perspective on the extent of spreading and the distribution of footprints on both the hydrophilic and hydrophobic SAMs, it is necessary to more closely examine the extent and rate of relaxation on each surface. Fibrinogen is able to achieve much larger footprints on the hydrophobic SAM than on a hydrophilic one. Such large increases in footprint on the C16 SAM are driven by hydrophobic interactions and must be accompanied by protein denaturation, although reorientations may also contribute to footprint growth.13 This is not necessarily the case on the OH SAM, where disruption of internal, hydrophobic interactions are not energetically favored. Here, the small increases in adsorbed area are consistent with a reconfiguration of the protein on the surface. Assuming that there is minimal unfolding of fibrinogen on the hydrophilic SAM, the different characteristic spreading times determined on each surface suggest that surface reorientations happen on a faster time scale than unfolding. In either case, the τs values are quite large, suggesting that both relaxation processes are significantly slower than those for flexible polymers, which have been shown to unfold at time scales of about 1 s or less.56 Furthermore, Van Eijk and Cohen Stuart32 have recently reported a τs value of 100 s for savinase adsorption on SiO2 using the same exponential growth model. Although a direct comparison between the τs values is confounded by the different proteins used and the presence of surface charge on the SiO2 surface, these spreading times suggest that the relaxation of individual protein molecules involves large energy barriers as a result of dissociation and reattachment processes and the strong internal structures of proteins. Finally, it is worth noting the fact that the fit of the fibrinogen adsorption traces to an exponential form is remarkable, given the complex sets of mechanisms likely to be involved in interfacial fibrinogen relaxation. Though (53) Hall, C. E.; Slater, H. S. J. Biophys. Biochem. Cytol. 1959, 5, 11. (54) Doolittle, R. F.; Everse, S. J.; Spraggon, G. FASEB J. 1996, 10, 1464. (55) Tang, L. J. Biomater. Sci. Polym. Ed. 1998, 12, 1257. (56) Adachi, Y.; Cohen Stuart, M. A.; Fokkink, R. J. Colloid Interface Sci. 1994, 167, 346.

Fibrinogen Adsorption on Surfaces

Figure 4. (A, top panel) Fibrinogen footprint distributions on the OH SAM: (i) C ) 100 mg/L, γ ) 5 s-1; (ii) C ) 25 mg/L, γ ) 30 s-1; (iii) C ) 25 mg/L, γ ) 5 s-1; (iv) C ) 25 mg/L, γ ) 2 s-1. The curves were generated from the exponential growth model for a0 ) 0.177 m2/mg, R ) 0.75, and τs ) 700 s. The shaded boxes and straight lines represent the footprints corresponding to the fraction of loosely bound fibrinogen for each desorption trace shown in Figure 7A. (B, bottom panel) Fibrinogen footprint distributions on the C16 SAM: (i) C ) 100 mg/L, γ ) 5 s-1; (ii) C ) 50 mg/L, γ ) 30 s-1; (iii) C ) 25 mg/L, γ ) 22 s-1; (iv) C ) 25 mg/L, γ ) 5 s-1; (v) C ) 25 mg/L, γ ) 1.5 s-1. The curves were generated from the exponential growth model for a0 ) 0.177 m2/mg, R ) 3.75, and τs ) 1500 s. The shaded boxes and straight lines represent the footprints corresponding to the fraction of loosely bound fibrinogen for each desorption trace shown in Figure 7B.

it may not be the case for the current system, the simplest mechanism yielding the exponential form is a single unfolding event, characterized by one dominant kinetic energy barrier. Although we are unable to measure the unfolding dynamics of individual protein molecules to confirm this, relaxation studies of unsaturated fibrinogen layers on the same C16 SAM have also revealed an exponential growth rate with a characteristic spreading time of 1425 s.29 The excellent agreement between the two characteristic times is quite remarkable considering the vastly different nature of the two experiments. In the case of the unsaturated layers, different proteins were used as probe molecules in order to measure the consumption of free surface space as partial fibrinogen layers unfolded. By varying the relaxation time of the unsaturated layer, the entire unfolding process could be observed. In the present study, the surface has been allowed to saturate under different adsorption conditions, resulting in protein footprint distributions that are dependent on the flux of protein to the interface. Here, we are only measuring the initial unfolding kinetics because lateral interactions stunt the fibrinogen molecules from realizing their maximum footprint. The consistency of both measurements is a strong indication that a distribution of different fibrinogen states does exist on the surface. One of the strengths of the exponential growth model is that it predicts distributions in the footprints of adsorbed molecules, on the basis of their arrival time and the current time in the course of an experimental run. Figure 4 shows

Langmuir, Vol. 18, No. 3, 2002 711

examples of these distributions, in panel A for the OH SAM and in panel B for the C16 SAM. Since a footprint distribution will evolve over the course of an adsorption run, a distribution graph provides a snapshot of the interfacial state at some instant. As the interface approaches the final jammed state where the surface fraction nears unity, the distribution ceases to evolve and its shape is constant for longer times. The distributions in Figure 4 represent these final interfacial states for different runs having various free solution concentrations and/or wall shear rates. The footprint distributions in Figure 4 have on their x-axes the dimensionless extent of spreading, from eq 1, where each value of a* corresponds to some population of molecules that have arrived to the interface at some instant t′*, prior to the present and final time t*. For the specific graphs in Figure 4, the x-axis represents footprints ranging from the initial protein footprint up to the maximum spreading that is known to occur for fibrinogen on the two surfaces. The y-axis describes the relative size of each population having footprint a*. The population size derives directly from the adsorption rate in eq 2. The molecules that currently have footprint a* had arrived at the interface at a prior time t′*. At the instant of their arrival, the adsorption rate was given by eq 2. Therefore the size of this population is dΓ*/dt* multiplied by some arbitrary differential time element, dt*, which applies to all interfacial populations. The relative size of each interfacial population is therefore proportional to the instantaneous flux at the time that population arrived at the interface, on the y-axis, and the multiplicative constant dt* is eliminated. The different areas under the various cusps are proportional to the different adsorbed amounts that are observed for the variety of adsorption histories examined. In Figure 4, the cusp-shaped graphs represent the footprint distributions, while the vertical bars are additional features that will be discussed in the final section of this paper. Within each graph, the largest and leftmost distributions correspond to conditions with the greatest concentrations and wall shear rates; in other words, the fastest rates of protein arrival at the interface or the shortest characteristic adsorption times. The graphs indicate for these fast runs, more protein adsorbs, i.e., the area under the cusp is greater than for runs having more dilute free solution concentrations or gentler flow rates. Also evident, the runs at the faster adsorption conditions yield smaller extents of spreading and narrower footprint distributions, while broader footprint distributions are realized for adsorption runs with the smallest flux of protein to the interface (low concentration and wall shear rate). The broadest distributions are found in Figure 4B for the C16 SAM for all adsorption conditions, as a result of the significant protein unfolding associated with that surface. Desorption Kinetics and Binding Energies. Application of the exponential growth model to the adsorption data has provided insight into the characteristic spreading rates of fibrinogen on hydrophobic and hydrophilic surfaces and the sizes of interfacial populations occupying different interfacial footprints. This latter information is geometrical in nature. We ultimately would like to address the energetic distribution of the adsorbed proteins and relate this to the protein-surface contact area. To this end, this section examines desorption data to obtain information about the tightness of binding between adsorbed proteins and the two surfaces. For each of the adsorption runs in Figures 1 and 2, buffer was introduced into the flow cell for 30 min following

712

Langmuir, Vol. 18, No. 3, 2002

Wertz and Santore

Figure 6. Schematic representing multiple binding strengths for different protein populations. Proteins with the largest footprints have the largest binding energies and binding constants (represented by arrow proportion). Figure 5. Fibrinogen adsorption kinetics on the OH SAM at a wall shear rate of 5 s-1 and varying bulk solution concentration of 25, 50, and 100 mg/L. Following 1 h of protein exposure, buffer is introduced into the flow cell (indicated by the arrow).

time scale, the simple Langmuir isotherm was chosen:

the initial 1-h adsorption period in which protein solution flowed past the surface. The desorption traces were discovered to be highly correlated with the prior adsorption history (choice of concentration and wall shear rate), as indicated by the example in Figure 5 for three runs involving different initial free solution fibrinogen concentrations on OH surfaces. It was always observed that runs with the highest coverages at the end of the adsorption period (with high concentrations and wall shear rates during adsorption) suffered the greatest loss of protein on washing with solvent. The population analysis in the previous section suggests that these runs with large adsorbed amounts contain substantial populations of protein that have not yet had the opportunity to spread or reorient on the surface. Such populations, which are more prevalent in runs with high concentrations and wall shear rates, may have a lower binding energy with the surface, since their contact area is smaller. We would ultimately like to be able to correlate adsorption energy with footprint size, a goal which is complicated by the distribution of interfacial states that exist in an adsorbed layer. We therefore commenced with the simplification that an adsorbed layer contains two conceptual populations: tightly and loosely bound proteins. Tightly bound proteins are those that tend not to desorb on an experimentally accessible time scale while loosely bound proteins are those that could ultimately be removed by washing in buffer. To quantify the binding tightness of the two populations, or at least that of the loosely bound population within each layer, we explored the following concept for local constrained equilibrium in adsorbed protein layers. We understand bound layers to contain proteins whose footprints are frozen as the surface saturates and the interface becomes jammed. Because jamming prevents further adsorption and produces a history-dependent coverage, we know that adsorbed layers are not truly equilibrated. However, in determining the binding tightness of each interfacial population, we considered the possibility that each population, with its particular footprint, has a binding energy associated with that footprint and could, in principle, maintain local equilibrium with free proteins in solution. Figure 6 illustrates the concept that multiple populations could have different binding strengths that could be represented by different equilibrium constants. Populations with large footprints would have greater binding energies and therefore greater binding constants. The present work made the simplification of having two main interfacial populations and therefore developed two separate binding expressions. For the weakly bound proteins, which can desorb on an experimentally accessible

Here K is the binding constant, C, is the free solution concentration, and L is a proportionality constant allowing the coverage to be represented in milligrams per square meter. It is noted that layers of adsorbed proteins, including the loosely bound population, possess more complex physics (e.g., interactions among proteins at the surface) than are included in eq 13. The Langmuir model nonetheless provides a first-order estimate of the energy difference between free and loosely bound proteins, where this “binding energy” includes interactions beyond those between individual proteins and the substrate. For the tightly bound population, we were inspired by the mean-field treatments for adsorbed polymers developed by Scheutjens and Fleer.57 Here, adsorbed molecules maintain many segment surface contacts but experience a configurational entropy loss on adsorption. Numerical solutions to this model generally produced isotherms that appeared linear on a semilog plot:

Γ(C) ) L

(1 +KCKC)

Γ(C) ) Γ0 + p[log (C) - log (C0)]

(13)

(14)

In eq 14, C is the bulk solution concentration, p is the slope of the isotherm in semilog coordinates, and Γ0 is the coverage measured at an arbitrary reference concentration C0, needed to fix an intercept on the logarithmic x-scale. While this treatment clearly cannot be applied quantitatively for protein adsorption, we invoke it here simply as a means to quantify the amount of tightly bound protein and do not overinterpret the parameters. Since our protein layers comprised the loosely and tightly bound populations, the two isotherm forms above were combined to yield an overall expression:

Γ(C) ) L

(1 +KCKC) + Γ

0

+ p[log (C) - log (C0)]

(15)

Now K describes the tightness of binding for the loosely bound population, and L and Γ0 represent the relative sizes of the loosely and tightly bound populations, respectively. Next, we considered the possibility that, despite the complexities of the jammed protein layer, each population of molecules is capable of desorption at its transportlimited rate. While kinetically limited desorption may seem, at first, to be the more appropriate interfacial behavior, Dijt et al.58 have shown that slow transportlimited desorption does indeed occur for tightly bound and entangled layers of homopolymers. It turns out to be (57) Scheutjens, J. M. H. M.; Fleer, G. J. J. Phys. Chem. 1979, 83, 1619. (58) Dijt, J. C.; Fleer, G. J.; Cohen Stuart, M. A. Macromolecules 1992, 25, 5416.

Fibrinogen Adsorption on Surfaces

Langmuir, Vol. 18, No. 3, 2002 713

Figure 8. Fibrinogen desorption kinetics on the C16 SAM. Time zero represents the introduction of buffer at two different flow rates [(i) γ ) 5 s-1; (ii) γ ) 30 s-1], following 1 h of fibrinogen exposure at C ) 50 mg/L and γ ) 5 s-1. The lines are fits to the runs for K ) 0.0025 L/mg, p ) 0.001 mg/m2, Γ0 ) 2.4 mg/m2, and C0 ) 5 mg/L.

Figure 7. (A, top panel) Fibrinogen desorption kinetics on the OH SAM. Time zero represents the introduction of buffer following 1 h of fibrinogen exposure at the following condtions: (i) C ) 100 mg/L, γ ) 5 s-1; (ii) C ) 25 mg/L, ) 30 s-1; (iii) C ) 25 mg/L, γ ) 5 s-1; (iv) C ) 25 mg/L, γ ) 2 s-1. The lines are fits to the runs for K ) 0.0025 L/mg, p ) 0.001 mg/m2, Γ0 ) 3.0 mg/m2, and C0 ) 5 mg/L. (B, bottom panel) Fibrinogen desorption kinetics on the C16 SAM. Time zero represents the introduction of buffer following 1 h of fibrinogen exposure at the following condtions: (i) C ) 100 mg/L, γ ) 5 s-1; (ii) C ) 25 mg/L, γ ) 22 s-1; (iii) C ) 25 mg/L, γ ) 5 s-1; (iv) C ) 25 mg/L, γ ) 1.5 s-1. The lines are fits to the runs for K ) 0.0025 L/mg, p ) 0.001 mg/m2, Γ0 ) 2.4 mg/m2, and C0 ) 5 mg/L.

wrong to assume, without proof, that slow desorption processes are kinetically limited. Tight binding energies can give very slow transport-limited behavior. Assuming that desorption was indeed transport-limited, its kinetics were described by eq 9 with C set to 0, since buffer was continuously flowed through the cell during desorption. Then, local equilibrium between the solution nearest the interface and the adsorbed layer maintained C* in accordance with eq 15. This led to

t)-

1 M

dΓ ∫ΓΓ(t) C*(Γ) init

(16)

which was solved numerically. Here, C*(Γ) is the inversion of the isotherm function in eq 15; hence, local concentration is used as a function of coverage, instead of coverage as a function of local concentration. Predictions from eq 16 are compared with desorption runs in Figure 7 panels A (OH SAM) and B (C16 SAM). The different data sets (O) represent variations in both bulk solution concentration and wall shear rate, while the solid lines represent the best fit of eq 16 to the experimental data for p ) 0.001 L/m2. Although it may appear that there are many parameters which could be chosen to accomplish the fits in Figure 7, most of these are eliminated independently. Since we have observed that changes in the bulk solution concentration have very little effect on the coverage level of the tightly bound layer, p was set at an arbitrarily small value, in this case 0.001 L/m2. Variations in p at this order of magnitude did not significantly change the shape of the

adsorption isotherm and had little effect on the desorption kinetics predicted by eq 16. Thus, an intermediate value for the reference concentration (C0 ) 5 mg/L) was chosen and used for each run. In the case of the relative amounts of loosely and tightly bound fibrinogen, Γ0 was estimated from runs that had minimal desorption. Once a reasonable value of Γ0 was set, this value was chosen to be identical for all runs (3.0 and 2.4 mg/m2 on the OH and C16 SAMs, respectively) and allowed for L to be determined from the experimental data. With p, Γ0, and L set to reasonable values, the primary fitting parameter for all the runs was K. Much to our surprise, K ) 0.0025 L/mg was obtained for all runs on all surfaces, a point to which we shall return. As Figure 7 indicates, eq 16 does an excellent job of predicting the desorption kinetics for all variations of free solution concentration and wall shear rate on both surfaces. Furthermore, these fits were obtained by keeping the coverage level corresponding to the tightly bound layer constant for each SAM, suggesting that the number of irreversibly adsorbed fibrinogen molecules is independent of the adsorption history. However, increasing the flux does result in higher coverage levels, leading to larger extents of loosely adsorbed molecules. On the OH SAM, the coverage level contribution from the loosely bound population ranges from 0.6 mg/m2 for the run with the smallest flux to 2.2 mg/m2 for the run with the largest flux. These values correspond to fractional coverages of the loosely bound layer ranging from 0.17 to 0.42. Similarly on the C16 SAM, the coverage levels of the loosely bound layer range from 0.3 to 3.6 mg/m2, corresponding to fractional coverages ranging from 0.11 to 0.6. The fit of eq 16 to the desorption data suggests that the observed dynamics are, indeed, transport-limited. As further proof, Figure 8 shows two runs with identical adsorption histories (C ) 50 mg/L, γ ) 5.0 s-1) that were subjected to different wall shear rates during desorption. In curve i, the wall shear rate during desorption is 5.0 s-1, while in curve ii, the wall shear rate during desorption is 30 s-1. Again, the fits to the desoption data have identical values of K, L, Γ0, and p. Only the mass transfer coefficient varied, per eq 10, as the wall shear rate was changed for the desorption portion of the kinetic traces. Again, there is excellent agreement between eq 16 and the data. The spread in the two data sets follows exactly from the mass transfer coefficients, indicating that the desorption of the loosely bound layer is transport-limited. At this point, it is important to point out that the experimental adsorption traces have been kept relatively short (30 min). Although these short times do not allow for complete desorption of the loosely bound fraction, it

714

Langmuir, Vol. 18, No. 3, 2002

is essential to keep the desorption times small in order to avoid protein spreading during the desorption process itself. At saturation, the surface is crowded and latearriving proteins have no free surface space to unfold and increase their number of segment-surface contacts. However, as protein molecules desorb back to the bulk solution, free surface space becomes available, allowing for further protein spreading during desorption. Longer desorption times would only lead to larger extents of unfolding during desorption, making the interpretation of the data exceedingly difficult. The different coverage levels associated with the irreversibly adsorbed layer on each SAM suggest that different protein-surface interactions are involved on the hydrophilic and hydrophobic surfaces. On the other hand, identical K values for the loosely bound layers on each SAM suggest that the average binding energy of the loosely bound proteins, and therefore the minimum energy needed to tightly bind a fibrinogen molecule, is the same regardless of surface chemistry, a surprising conclusion and one that warrants further investigation. These K values correspond to the adsorption energy of the entire loosely bound population and indicate a similar threshold energy that must be achieved in order for a molecule to become irreversibly adsorbed. The K value of 0.0025 L/mg corresponds to an adsorption energy, including proteinprotein interactions at the interface, of 6kT per protein molecule.57,58 Significant in interpreting fibrinogen adsorption on the C16 SAM, this value is within the range of 5-13kT reported by several laboratories for hydrophobically modified polymer adsorption on several hydrophobic surfaces,57-60 suggesting that the minimal energy needed for irreversible fibrinogen adsorption is comparable to the energy associated with a linear chain consisting of 16-18 carbon molecules adsorbing onto a hydrophobic surface. The significance of the 6kT value on the OH SAM is less clear but may result from the formation of a few hydrogen bonds. While we report adsorption energies similar to those of other hydrophically driven adsorption processes, it is important to keep in mind that our value corresponds to the association energy of the loosely bound proteins only. Our K value does not correspond to the irreversibly adsorbed layers present on the hydrophilic and hydrophobic SAMs; thus, it does not describe protein surface interactions that can be as high as 30-200kT per molecule for fibrinogen adsorption.61,62 These higher adsorption energies are captured by p and Γ0 in eq 14. Relationship between Footprint Size and Binding Energy. In the first section of this work, we demonstrated that the exponential spreading process yielded a substratedependent distribution of protein footprints. In the previous section, we interpreted desorption data in the context of loosely and tightly bound populations. In this section we relate the threshold energy of binding and the size of the loosely bound populations to the predicted footprint distributions from the first section. One of the key observations in the prior section was that adsorption runs giving high coverages and smaller

Wertz and Santore

footprints had larger populations that could be readily washed from the surface. This suggests that the last proteins to arrive at the interface, which have had the least opportunity to establish contact with the surface, are the most readily displaced upon washing. We now quantify this effect in Figure 4, by placing bars on the footprint distributions calculated for each of the adsorption runs. Each bar divides the particular footprint distribution into two populations: the loosely bound proteins on the left and the tightly bound proteins on the right, based on the fraction of loosely bound, L/(L + Γ0), and tightly bound, Γ0/(L + Γ0), populations. The width of the each bar simply represents the error in making this determination. The heights of the bars decrease from left to right to aid in pairing each bar with the appropriate footprint distribution. In Figure 4A (OH SAM), the bars, representing the threshold binding energies for the tightly bound populations, are spread across the range of possible footprints, while in Figure 4B (C16 SAM), the bars tend to be clustered near a footprint twice the size of that for the native protein. This is an interesting distinction between the nature of spreading on hydrophobic and hydrophilic surfaces. For the hydrophobic surface, the results suggest that once a protein reaches a particular size on the surface, it establishes sufficient hydrophobic attractions (on the order of 6kT) to the surface that it will not be easily removed. Fibrinogen will continue to unfold with time, but any additional unfolding does not influence the ability of the protein to be removed. It is worth noting in Figure 4B that for the run giving the leftmost footprint distribution (C ) 100 mg/L, γ ) 5 s-1), the bar does not overlap with the other bars. In fact, for this particular run, the bar indicating the threshold desorption energy cannot overlap with the other bars because none of the proteins in that particular distribution have sufficiently large footprints. Instead, the bar for this distribution lies on the far right side, within the possible range of footprint sizes for this run. In contrast to the behavior with the hydrophobic surface, on the hydrophilic surface in Figure 4A, the bars move to the right as the distribution evolves. This signature is more difficult to interpret; however, one possible interpretation could be that fibrinogen relaxations on the hydrophilic surface have to do with reorienting rather than unfolding. As fibrinogen molecules roll over on the surface to increase their footprint, some contacts are lost while new ones are made. In layers with more fibrinogens rolled over to occupy greater area, there is more interaction with the surface, so that the threshold energy is associated with different-sized footprints or different protein orientations, as the adsorption history is varied. This feature may be a signature of rolling over, while on the C16 surface in Figure 4B, there may be little rolling or loss of the original surface contacts as new ones are made. Instead, when fibrinogen adsorbs to a hydrophobic surface, the original contacts are retained while new ones develop as a result of denaturing. Summary and Conclusions

(59) Jenkins, R. D.; Silebi, C. A.; El-Asser, M. S. In Polymers as Rheology Modifiers; Schulz, D. N., Glass J. E., Eds.; ACS Symposium Series 462; American Chemical Society: Washington, DC, 1991. (60) Huang, Y.; Santore, M. M. Langmuir, submitted for publication, 2001. (61) Hemmerle, J.; Altmann, S. M.; Maaloum, M.; Horber, J. K. H.; Heinrich, L.; Voegel, J.-C.; Schaaf, P. Proc. Natl. Acad. Sci. U.S.A. 1999, 96, 6705. (62) Gergely, C.; Voegel, J.-C.; Schaaf, P.; Senger, B.; Maaloum, M.; Horber, J. K. H.; Hemmerle, J. Proc. Natl. Acad. Sci. U.S.A. 2000, 97, 10802.

This work demonstrated that, during adsorption of fibrinogen on hydrophobic and hydrophilic surfaces, interfacial relaxations occur. These relaxations are manifest by an increased molecular footprint as a function of time, they limit the ultimate coverage, and they can lead to the appearance of a history-dependent coverage through the competitive time scales of (mass transfer-limited) adsorption and relaxation. Experimental runs having

Fibrinogen Adsorption on Surfaces

variations in concentration and wall shear rate could be described by an exponential footprint growth model, with a single characteristic growth-relaxation time for each surface. These relaxation times agreed quantitatively with our previous assessments of interfacial spreading times by very different experimental procedures. By use of the relaxation times, it was possible to reconstruct the distributions of interfacial footprints resulting from adsorption at different conditions (concentration and wall shear rate). A broader range of footprint distributions with greater ultimate extents of footprint growth were found for the hydrophobic surface compared with the hydrophilic surface, suggesting substantial interfacial denaturing on the former. The more modest range of interfacial molecular areas on the hydrophilic surface suggested a substantial influence of reorientation. It was not necessary to invoke a denaturing process to explain the range of footprints observed on the hydrophilic surface. For each of the adsorption studies analyzed with the exponential growth model, the final adsorbed layers were subject to controlled rinsing in flowing buffer. It was observed that adsorbed layers with high coverages (which had been deposited from high concentrations or fast wall shear rates) were most susceptible to loss of interfacial mass on rinsing. This suggested that these particular runs had significant protein populations that had not experienced sufficient opportunity to make intimate contact with the surface by reorienting or denaturing. The observations therefore motivated a comparison of the desorption traces with a kinetic model that allowed the loosely and tightly bound masses to be distinguished and the energy of the loosely bound molecules to be quantified. It was found that for a particular substrate, the mass of tightly bound material was somewhat invariant, while the amount of loosely bound protein was strongly dependent on experimental history. More surprising, however, was that on both surfaces the loosely bound proteins had a binding

Langmuir, Vol. 18, No. 3, 2002 715

energy of 6kT, roughly equivalent to hydrophobic associations of C16-C18 chains or to a few hydrogen bonds for adsorption onto hydrophobic or hydrophilic surfaces, respectively. This implied that the tightly bound fibrinogen fractions, which could withstand rinsing in buffer, had average adsorption energies greater than 6kT. It was also shown that desorption of the loosely bound fraction was transport-limited, following a predictable dependence on flow rate. The relative sizes of the loosely and tightly bound populations were then compared with the footprint distributions originally calculated from the exponential growth model for each adsorption history. On the hydrophobic surface, the 6kT binding energy corresponded to an interfacial footprint twice the size of initial (natively configured) value. This finding was independent of adsorption history, suggesting that when fibrinogen adsorbs on hydrophobic surfaces, its footprint grows in a way that the initial protein-surface contacts are retained as new ones are generated from interfacial denaturing. In contrast, on the hydrophilic surface, the 6kT threshold binding energy corresponded to different footprints, depending on the adsorption history. This suggested that old protein-surface contacts were sacrificed as the protein footprint grew and new contacts were made. These characteristics of retaining contacts on hydrophobic surfaces and sacrificing contacts on hydrophilic surfaces as the overall footprint increases appeared to be characteristic of unfolding and reorienting processes, respectively, on the two surfaces. Acknowledgment. This work was supported through grants from The Whitaker Foundation and NSF (CTS9817048). LA011075Z