pubs.acs.org/Langmuir © 2010 American Chemical Society
Fibrinogen Adsorption on Mica Studied by AFM and in Situ Streaming Potential Measurements Monika Wasilewska* and Zbigniew Adamczyk Institute of Catalysis and Surface Chemistry, Polish Academy of Science, Niezapominajek 8, 30-239 Cracow, Poland Received July 23, 2010. Revised Manuscript Received September 15, 2010 Adsorption of fibrinogen from aqueous solutions on mica was studied using AFM and in situ streaming potential measurements. In the first stage, bulk physicochemical properties of fibrinogen and the mica substrate were characterized for various ionic strength and pH. The zeta potential and number of uncompensated (electrokinetic) charges on the protein surfaces were determined from microelectrophoretic measurements. Analogously, using streaming potential measurements, the electrokinetic charge density of mica was determined for pH range 3-10 and the NaCl background electrolyte concentration of 10-3 and 10-2 M. Next, the kinetics of fibrinogen adsorption at pH 3.5 and 7.4 in the diffusion cell was studied using a direct AFM determination of the number of molecules per unit area of the mica substrate. Then, streaming potential measurements were performed to determine the apparent zeta potential of fibrinogen-covered mica for different pH and ionic strength in terms of its surface concentration. A quantitative interpretation of these streaming potential measurements was achieved in terms of the theoretical model postulating a side-on adsorption of fibrinogen molecules as discrete particles. On the basis of these results, the maximum coverage of fibrinogen Θ close to 0.29 was predicted, in accordance with previous theoretical predictions. It was also suggested that anomalous adsorption for pH 7.4, where fibrinogen and the mica substrate were both negatively charged, can be explained in terms of a heterogeneous charge distribution on fibrinogen molecules. It was estimated that the positive charge was 12 e (for NaCl concentration of 10-2 M and pH 7.4) compared with the net charge of fibrinogen at this pH, equal to -21 e. Results obtained in this work proved that the coverage of fibrinogen can be quantitatively determined using the streaming potential method, especially for Θ5.8. As discussed before, knowing the electrophoretic mobility, one can calculate the average number of charges per molecule from the Lorenz-Stokes relationship28 Nc ¼
6πη 3 108 RH μ e 1:602
ð4Þ
where Nc is expressed as the number of elementary charges (e) per molecule. It is assumed hereafter that e = 1.60 10-19 Coulomb. It should be noted, however, that eq 4 becomes less accurate -1 if κa >1, that is, if the thickness P of the double-layer κ = (εkT/2e2I)1/2, where I = 1/2( icizi2) is the ionic strength and ci is the ion concentrations, zi is the ion valency and becomes smaller than the characteristic protein dimension a. Even with this limitation, the microelectrophoretic method of (32) Galisteo, F.; Norde, W. Colloids Surf., B 1995, 4, 389.
690 DOI: 10.1021/la102931a
ζp ¼
3η μ 2εFðκaÞ e
ð5Þ
where F(κa) is the function of the dimensionless parameter κa. For thin double layers (κa>10), F(κa) approaches unity, and for thick double layers κa>1, F(κa) approaches 3/2. It was calculated from eq 5 that ζp = 28 mV for I = 1.3 10-3 M and pH 3.5, and ζp = 22.2 mV for I = 10-2 M. For pH 7.4, ζp = -20.7 mV for I = 10-3 M, and ζp = -19.4 mV for I = 10-2 M. (See Table 1.) b. Mica Substrate Surface Characteristics. Zeta potential of the mica substrate ζi was determined via the streaming potential measurements according to the procedure described above. The dependence of ζi on the pH for different NaCl electrolyte concentration of 10-3 and 10-2 M is shown in Figure 3. As can be observed, for NaCl concentration for I = 1.3 10-3 and I = 10-2 M, the zeta potential of mica was negative and decreased moderately with the increase in pH, varying between -63 mV for pH 3.5 and -100 mV at pH 7.4. For NaCl concentration for I = 1.3 10-2 M, the zeta potential of mica varied between -52 mV for pH 3.5 and -74 mV at pH 7.4. It is interesting to note that in the case of the TRIS buffer, the zeta potential of mica for pH 7.4 was slightly lower, equal to -112 and -80 mV, for NaCl concentration for I = 1.3 10-3 and 10-2 M, respectively. (See Table 2.) Knowing the zeta potential, one can calculate the electrokinetic (uncompensated charge) of mica for given pH and ionic strength using the well-known relationship valid for a symmetric 1:1 electrolyte34 σ0 ¼ -
ð8εkTnb Þ1=2 eζ sinh 2kT 0:160
ð6Þ
where σ0 is the electrokinetic charge density expressed in e nm-2 and nb is the number concentration of ions expressed in m-3 . Using the above zeta potential values, one can calculate from eq 6 that σ0 = -0.042 e nm-2 for pH 3.5, I = 1.3 10-3 M NaCl, σ0 = -0.090 e nm-2 for pH 3.5, 10-2 NaCl, σ0 =-0.105 e nm-2 for pH 7.4, I = 1 10-3 M NaCl, and σ0 = -0.17 e nm-2 for pH 3.5, 10-2 NaCl. (For the sake of convenience, these values are collected in Table 2.) As can be noticed, the electrokinetic charge density on mica decreases significantly with the ionic strength and pH. It should be observed, however, that even the lowest value of -0.17 e nm-2 remains much higher than the lattice charge of the basal plane of mica, equal to -2.1 e nm-2.35 This effect is caused (33) Galisteo, F.; Norde, W. J. Colloid Interface Sci. 1995, 172, 502. (34) Adamczyk, Z. Particles at Interfaces: Interactions, Depositions, Structure; Elsevier/Academic Press: Amsterdam, 2006; pp 214-219. (35) Rojas, O. Adsorption of Polyelectrolytes on Mica. In Encyclopedia of Surface and Colloid Science; Marcel Dekker: New York, 2002; p 517.
Langmuir 2011, 27(2), 686–696
Wasilewska and Adamczyk
Article Table 2. Zeta Potential ζi and the Number of Uncompensated Charge Density Σ0 of Mica under Various Physicochemical Conditions conditions
zeta potential, ζi [mV]a charge density, σ0[e nm-2]b
pH 3.5 I = 1.3 10-3 M pH 3.5 I = 10-2 M pH 7.4 I = 10-3 M pH 7.4 I = 10-2 M a Determined from lated using eq 6.
-63 ( 4
-0.042 ( 0.005
-52 ( 5
-0.090 ( 0.01
-112 ( 4 (TRIS)
-0.105 ( 0.008
-80 ( 3 (TRIS) -0.17 ( 0.01 the streaming potential measurements. b Calcu-
Figure 3. Dependence of the zeta potential of bare mica ζi on pH, T = 293 K, b, I = 10-2 M, pH regulated by the addition of HCl/ NaOH; O, I=10-2 M, pH regulated by the addition of the TRIS buffer (the solid line 1 denotes the interpolation of experimental results). [, I = 10-3 M, pH regulated by the addition of HCl/ NaOH; ], I=10-3 M, pH regulated by the addition of the TRIS buffer (the solid line 2 denotes the interpolation of experimental results).
Figure 5. AFM images (semicontact mode, air) of fibrinogen monolayers on mica (deposition conditions: pH 7.4, I = 10-2, cb = 0.35 ppm). (a) Surface concentration of fibrinogen N = 27 μm-2. (b) Surface concentration of fibrinogen N = 123 μm-2.
Figure 4. AFM images (semicontact mode, air) of fibrinogen monolayers on mica (deposition conditions: pH 3.5, I = 1.3 10-3, cb = 0.3 ppm). (a) Surface concentration of fibrinogen N = 20 μm-2. (b) Surface concentration of fibrinogen N = 125 μm-2.
by preferential adsorption of cations, including Hþ, which was extensively studied by Scales et al.36 and others.26 The experimental results shown in Figures 1-3 suggest, therefore, that an electrostatically driven adsorption of fibrinogen on mica is expected for pH < 5.8, where mica remains strongly negatively charged and fibrinogen charge becomes increasingly positive. This hypothesis was checked experimentally, as described in the next section. c. Fibrinogen Adsorption Kinetics. Fibrinogen adsorption experiments were conducted under the diffusion-controlled transport according to the procedure described above. The surface concentration of fibrinogen molecules N adsorbed after a prescribed time t (expressed as the number of molecules per square micrometer) was determined by a direct AFM counting. Although this procedure is tedious and inaccurate for higher surface concentration, it is the only one that allows for direct determination of fibrinogen adsorption kinetics. In consequence, streaming potential (36) Scales, P. J.; Grieser, F.; Healy, T. W. Langmuir 1990, 6, 582.
Langmuir 2011, 27(2), 686–696
measurements, discussed next, can be unequivocally interpreted in terms of the protein surface concentration rather than adsorption time or protein bulk concentration, as usually done. Typical fibrinogen monolayers of different concentration, observed by AFM, are shown in Figures 4 and 5 for pH 3.5 and 7.4, respectively. As can be seen, the fibrinogen molecules appear as isolated particles, which facilitates their enumeration. It should be mentioned, however, that no attempt was undertaken in this work to determine fibrinogen molecule dimensions on surfaces because of the possibility of tip artifacts. In Figure 6, the kinetics of fibrinogen adsorption on mica, determined according to this direct counting procedure, is plotted in the reduced form, that is, as the dependence of N/cb on the square root of the time t1/2, which is relevant for an irreversible, diffusion-controlled adsorption.17 As can be seen in Figure 6, the experimental results obtained for various bulk concentration of fibrinogen (varied between 0.35 and 2 ppm) and pH 3.5 are well reflected in this coordinate system by one universal straight-line dependence (dashed line in Figure 6) having the slope sf of 61 μm-2 ppm-1 min-1/2. This behavior can be quite naturally interpreted in terms of the strong electrostatic attraction between positively charged fibrinogen molecules, having zeta potential of 28 mV, and the negatively charged mica surface, characterized by zeta potential of -63 mV (pH 3.5, I=1.3 10-3 M and 10-2 M NaCl). Quite unexpectedly, however, fibrinogen molecules adsorbed on mica with exactly the same rate for pH 7.4 (Figure 6), where the zeta potential of fibrinogen was negative, equal to -19.0 mV, and mica surface was also strongly negative, having zeta potential of -80 mV (pH 7.4, 10-2 M TRIS buffer). Such anomalous fibrinogen deposition under physiological conditions (pH 7.4) was previously observed by Zembala and Dejardin,24 Malmsten,13 DOI: 10.1021/la102931a
691
Article
Wasilewska and Adamczyk
experiments in comparison with direct adsorption measurements were found within the experimental error, estimated to be of ∼5%. Negligible desorption was qualitatively observed for higher fibrinogen concentration range as well, although quantitative measurements by AFM were not feasible. d. Streaming Potential of Fibrinogen Covered Mica. Streaming potential measurements for fibrinogen-covered mica were carried out according to the procedure described above to elucidate mechanism of fibrinogen adsorption. The primary result of this series of experiments was the dependence of the apparent zeta potential of mica ζ on the nominal fibrinogen surface concentration N. This was calculated using the previous experimental slope (Figure 6) using the constitutive dependence N ¼ sf t1=2 cb Figure 6. Dependence of the reduced surface concentration of fibrinogen N/cb (μm-2 ppm-1) on the square of adsorption time t1/2, I = 1.3 10-3 M. The full points denote experimental results obtained by a direct AFM enumeration for pH 3.5 (various bulk concentration of fibrinogen), and the empty points denote results obtained for pH 7.4, I = 10-3. The dashed line represents a linear regression on experimental points. The solid line shows exact theoretical results obtained using the RSA model.
Ortega-Vinuesa et al.,5,6 Toscano and Santore,11 Kalasin and Santore,15 and others.16 One can speculate that this effect is caused by additional interactions, most likely of the van der Waals type. Another contribution to attractive interactions at this pH may originate from the heterogeneous charge distribution over the fibrinogen molecule, which was previously suggested.9 Therefore, even if the overall charge of the fibrinogen molecule remains negative, there are patches (most likely the side arms) bearing positive charge. However, little quantitative information on this issue can be gained without additional experiments in which electrostatic properties of fibrinogen monolayers are analyzed under in situ conditions. It is interesting to mention that experimentally determined fibrinogen adsorption kinetics shown in Figure 6 can be well accounted for by the theoretical model of an irreversible, side-on adsorption, analyzed in detail theoretically in ref 17 Results obtained using this model are plotted in Figure 6 in the form of the solid line. These theoretical results were obtained by a numerical solution of the governing mass transport equation, in which the surface blocking effects were considered in an exact form. As discussed in ref 17, for a not too high range of N/cb, the results obtained from this exact solution can be the straight-line dependence because the overall adsorption rate is governed by the bulk transport. It is interesting to observe that analogous kinetic results were obtained by Toscano and Santore11 in the case of the convectioncontrolled adsorption of fibrinogen on modified glass substrate in the parallel-plate channel. In this case, however, the initial adsorption kinetics was a linear function of the adsorption time rather than the square root of time. The good agreement of experimental and theoretical results shown in Figure 6 suggests that fibrinogen adsorption on mica for the low surface coverage range proceeds according to the irreversible, diffusion-controlled mechanism. Additional desorption experiments were carried out to clarify the reversibility issue. In this series of experiments, the fibrinogen monolayer was kept for a desired time period (up to 100 min) in a pure electrolyte without fibrinogen. Then, the monolayer was examined under AFM, and the surface concentration of fibrinogen was quantitatively determined. The differences in N in these 692 DOI: 10.1021/la102931a
ð7Þ
It is interesting to mention that for the fibrinogen surface concentration calculated from eq 7 remains a well-established, experimental value for N< 200 μm-2. In Figure 7a,b, experimental results are presented, obtained for pH 3.5, I = 1.3 10-3 M and 10-2 M, respectively. Some characteristic features of experimental dependencies of ζ on N can be distinguished: (i) a steep, quasi-linear increase in ζ for N < 1000 μm-2, (ii) inversion of the initially negative zeta potential to positive for N ca. 1200 μm-2, (iii) attaining the saturation value for N > 2200 μm-2, which was markedly lower than the bulk value of the zeta potential of fibrinogen (equal to 28 and 22.2 mV for I = 1.3 10-3 M and 10-2 M, respectively), and (iv) little sensitivity of the ζ versus N dependencies to the change in the ionic strength of the solution. It should be mentioned that the limiting value of N for which the saturation values of ζ were observed is close to the exact value determined theoretically,17 equal to 2260 μm-2, pertinent to the irreversible, side-on adsorption mechanism. Analogous results were previously reported for the PEI/mica system, where the polyelectrolyte was positively charged37 and the polystyrene latex particles/mica system.30 In the latter case, the dependence of the apparent zeta potential of mica was plotted against the coverage of latex rather than its surface concentration. Results of streaming potential measurements for pH 7.4, I = 10-2 M, TRIS are shown in Figure 8. In this case, however, because the bare mica zeta potential and the bulk zeta potential of fibrinogen were both negative, no inversion of the mica zeta potential was observed, and the saturation value of the micacovered zeta potential coincided with the bulk zeta potential of fibrinogen, equal to -21 mV. It is interesting that our data shown in Figure 8 resemble closely previous experimental results obtained by Zembala and Dejardin24 who used the same streaming-potential method for fibrinogencovered silica capillaries. In their work, the apparent zeta potential was plotted as a function of the fibrinogen concentration in the bulk. The fibrinogen monolayers were formed under convectioncontrolled transport for the time of 3600 s. It was found that for fibrinogen bulk concentration between 5 and 100 ppm, the apparent zeta potential of silica capillaries covered by fibrinogen varied slightly between -22 and -20 mV, that is, close to our results shown in Figure 8. Similar experiments were reported recently by Kalasin and Santore,15 who determined the zeta potential of negatively charged silica spheres (1 μm of diameter) covered by a controlled amount of (37) Adamczyk, Z.; Michna, A.; Szaraniec, M.; Bratek, A.; Barbasz, J. J. Colloid Interface Sci. 2007, 313, 86.
Langmuir 2011, 27(2), 686–696
Wasilewska and Adamczyk
Article
Figure 8. Dependence of the zeta potential of mica ζ on the surface concentration of fibrinogen N (μm-2). The points denote experimental results obtained from the streaming potential measurements. I = 10-2 M, pH 7.4. The dashed vertical line shows the maximum surface concentration predicted theoretically.
irreversibly under arbitrary orientations (including the end-on configuration). Accordingly, in case (i), because charge additivity is assumed, the net charge density of the mica after fibrinogen adsorption, characterized by the surface concentration, N, is calculated from the charge balance equation37 σ ¼ σ 0 þ NNc
ð8Þ
Because all parameters appearing in eq 8 are accessible experimentally, as discussed above, one can unequivocally determine the zeta potential of the interface covered by fibrinogen using the GC relationship34 Figure 7. Dependence of the zeta potential of mica ζ on the surface concentration of fibrinogen N (μm-2). The points denote experimental results obtained from the streaming potential measurements: (a) I = 1.3 10-3 M, pH 3.5. (b) I = 10-2 M, pH 3.5. The dashed vertical line shows the maximum surface concentration predicted theoretically.
fibrinogen (whose coverage was expressed in terms of mg m-2) for pH 7.4, I = 5 to 176 10-3 M. For the ionic strength similar to our experiments, that is, 5 to 2610-3 M, they observed the saturation value of zeta potential equal to -20 mV for fibrinogen coverage >1 mg m-2, which corresponds to surface concentration N = 1500 μm-2. Both the saturation zeta potential value and the critical surface concentration determined by Kalasin and Santore are in a good agreement with our results shown in Figure 8. The agreement of our results with previous data obtained for colloid particles26 and polyelectrolytes,37 which were successfully interpreted theoretically, suggests that such analysis is feasible for fibrinogen adsorption as well. This is so because the necessary physicochemical properties of fibrinogen in the bulk and the mica substrate were thoroughly determined in our work. Accordingly, we attempt to analyze our experimental data in terms of three basic models: (i) the patchy adsorption model, based on the Gouy-Chapmann (GC) concept, where the adsorbate (fibrinogen) is treated as a flat (2D) object having a uniform charge density distribution, (ii) the irreversible, side-on adsorption model, considering a 3D charge distribution over the adsorbate molecules applied before for colloid particles and polyelectrolytes,27 and (iii) 3D adsorption model where the molecules are assumed to adsorb Langmuir 2011, 27(2), 686–696
ζi ¼ (
2kT jσj þ ðσ 2 þ 4Þ1=2 ln e 2
ð9Þ
where the plus and minus sign correspond to positive and negative signs of σ, σh = σ/(2εkTnb)1/2 is the dimensionless electrokinetic charge density, and nb = 6.023 1020 I (where I is expressed in M) is the number concentration of cations in the bulk. It should be noted that according to the above GC model, the entire substrate surface covered by adsorbed species is treated as uniformy charged, characterized by one macroscopic value of the zeta potential ζi. Hence, charge distribution is assumed to be strictly 2D, and no charge heterogeneity or charge fluctuation effects are considered. This disadvantage is eliminated in the 3D models (ii) and (iii), in which the electrokinetic flow of charge in the vicinity of adsorbed species, treated as isolated entities, is considered in an exact way.38 As discussed in ref 27, the general expression for the streaming potential of an interface covered uniformly by particles (molecules) is given by ζp ζ Es ðΘÞ ¼ ¼ 1 - Ai ðθÞθ þ Ap ðθÞ θ ζi Es 0 ζi
ð10Þ
where Es(θ) is the streaming potential for particle-covered substrate, Es0 = (εΔP/ηKe)ζi is the streaming potential for the bare substrate, and Ai (Θ) and Ap (Θ) are the dimensionless functions of (38) Zembala, M.; Adamczyk, Z.; Warszynski, P. Colloids Surf., A 2001, 195, 3.
DOI: 10.1021/la102931a
693
Article
Wasilewska and Adamczyk
the particle coverage Θ, particle shape, and the a/Le parameter (Le = κ-1 is the thickness of the double-layer). In the limit of low coverage and spherical particles, the functions Ai (Θ) and Ap (Θ), approach the constant values of C0i = 10.2 and C0p = 6.51, respectively.38 These constants are practically independent of a/Le for the range of a/Le > 2 (thin double-layers) as shown in ref 38. For a/Le < 2 (thicker double-layers), the Ci0 constant decreases slightly, and the C0p constant increases. Exact values of these constants for 0.1 < a/Le < 100 are reported in ref 27. Exact theoretical values of C0i and C0p in the limit of thin double layers are also available for anisotropic (elongated) particles composed of strings of touching beads, adsorbing side-on or end-on (arbitrary angle) on planar interfaces.27 In these cases, the Ci0 and the Cp0 constants increased significantly with the aspect ratio (widthto-length ratio of particles). For the aspect ratio of 10, in the case of unoriented, end-on adsorption, ci = 17.1 and C0p = 10.7,27 whereas for aspect ratio of 20, Ci0 = 25.8 and C0p = 15.6. Recently, exact theoretical results were reported,25 which allowed one to determine the Ai (Θ) and Ap (Θ) functions for spherical particles in the limit of thin double layers. These results were obtained by evaluating numerically the flow in the vicinity of adsorbed particles using the multipole expansion method for the coverage range up to 0.5. The exact numerical results were interpolated by the following analytical functions assuring precision better than 1%. Ai ðΘÞ ¼
10:2 - 5:75Θ 1 þ 5:46Θ
Ap ðΘÞ ¼
6:51 - 2:38Θ 1 þ 5:46Θ
ð11Þ
These functions can be also well interpolated by more concise exponential forms, convenient for practical applications 1 - expð- Ci0 ΘÞ Θ 0 1 pffiffiffi 1 @1 - expð- 2Cp0 ΘÞA Ap ðΘÞ ¼ pffiffiffi Θ 2 Ai ðΘÞ ¼
ð12Þ
It is interesting to mention that this 3D model was applied successfully for a quantitative interpretation of streaming potential of colloid particles26,30 and polyelectrolyte covered surfaces.37 However, application of this model requires the knowledge of the solute coverage Θ rather than the surface concentration N. Both quantities are connected by the constitutive dependence Θ ¼ Sg N
ð13Þ
where Sg is the characteristic cross-section of the solute particle or molecule. In the case of spherical particles, the choice of Sg = πa2 is obvious and unique.27 However, this is not unique for anisotropic molecules of irregular shape such as fibrinogen because various characteristic cross sections can be distinguished. Therefore, to simplify the theoretical analysis, we adopted in our work the bead model of fibrinogen (Table 1), which was applied successfully before17 for determining adsorption kinetics and the jamming limit. Consequently, the characteristic cross-section Sg for the side-on adsorption equals 128 nm2, and that for the end-on adsorption equals 35.3 nm2. Using the characteristic cross-section (39) Triantaphyllopoulos, E.; Triantaphyllopoulos, D. C. Biochem. J. 1967, 105, 393.
694 DOI: 10.1021/la102931a
for the side-on adsorption, the jamming coverage for fibrinogen becomes 0.29.17 Using this coverage definition, the experimental streaming potential results for fibrinogen can be quantitatively interpreted in terms of the above three models. In Figure 9a,b, results are shown for pH 3.5 (positively charged fibrinogen molecules) and NaCl concentration of 10-3 and 10-2 M, respectively. The ζp/ζi parameter was -0.44, for I = 1.3 10-3 M, NaCl and -0.46 for I=10-2 M, NaCl. As can be seen in Figure 9, the GC model does not properly reflect the experimental results. For low ionic strength, it predicts results too high for Θf < 0.2, where Θf = SgN is the fibrinogen coverage, and much too low for Θf>0.2. It should be observed, however, that values of Θf > 0.29 are apparent ones, calculated from eq 7. A significant role of the ionic strength is also predicted by the GC model, which was not confirmed by experimental results shown in Figure 9. On the contrary, the experimental ζ/ζi dependence was insensitive to the ionic strength, which was properly reflected by the 3D model postulating a side-on adsorption of fibrinogen (based on eqs 11 and 13). The agreement was especially satisfactory for higher ionic strength, where the a/Le parameter based on the fibrinogen main nodule (having the radius of 3.35 nm) was close to unity. It is interesting to mention that in this case the experimental results for fibrinogen were in good agreement with previous results obtained for positively charged latex particles26 shown also in Figure 10b. It is interesting to mention that for such a system of oppositely charged interface and the protein, the sensitivity of the streaming potential measurements is high because the initial slope of the ζ/ζi versus Θf dependence is much larger than unity (in absolute terms). Therefore, the results shown in Figure 9 suggest that the coverage of fibrinogen, and consequently the concentration of this protein in the bulk solution, can be precisely determined by the streaming potential method, especially for Θf