Viscoelastic Modeling of Highly Hydrated Laminin Layers at

Aug 11, 2007 - The adsorption of proteins at material surfaces is important in applications .... Interactions between Exogenous FGF-2 and Sulfonic Gro...
0 downloads 0 Views 235KB Size
Download PDF
9760

Langmuir 2007, 23, 9760-9768

Viscoelastic Modeling of Highly Hydrated Laminin Layers at Homogeneous and Nanostructured Surfaces: Quantification of Protein Layer Properties Using QCM-D and SPR Jenny Malmstro¨m, Hossein Agheli, Peter Kingshott, and Duncan S. Sutherland* Interdisciplinary Nanoscience Center, iNANO, UniVersity of Aarhus, Aarhus 8000, Denmark ReceiVed April 27, 2007. In Final Form: June 27, 2007 The adsorption of proteins at material surfaces is important in applications such as biomaterials, drug delivery, and diagnostics. The interaction of cells with artificial surfaces is mediated through adsorbed proteins, where the type of protein, amount, orientation, and conformation are of consequence for the cell response. Laminin, an important cell adhesive protein that is central in developmental biology, is studied by a combination of quartz crystal microbalance with dissipation (QCM-D) and surface plasmon resonance (SPR) to characterize the adsorption of laminin on surfaces of different surface chemistries. The combination of these two techniques allows for the determination of the thickness and effective density of the protein layer as well as the adsorbed mass and viscoelastic properties. We also evaluate the capacity of QCM-D to be used as a quantitative technique on a nanostructured surface, where protein is adsorbed specifically in a nanopattern exploiting PLL-g-PEG as a protein-resistant background. We show that laminin forms a highly hydrated protein layer with different characteristics depending on the underlying substrate. Using a combination of QCM-D and atomic force microscopy (AFM) data from nanostructured surfaces, we model laminin and antibody binding to nanometer-scale patches. A higher amount of laminin was found to adsorb in a thicker layer of a lower effective density in nanopatches compared to equivalent homogeneous surfaces. These results suggest that modeling of QCM-D data of soft viscoelastic layers arranged in nanopatterns may be applied where an independent measure of the “dry” mass is known.

Introduction The adsorption of proteins at material surfaces has been the focus of considerable research effort for several decades, and it continues to receive increasing attention1-3 because of its importance in applications such as biomaterials, drug delivery, and diagnostics.4 The interaction of cells with biomaterial surfaces is mediated through adsorbed proteins, where the type of protein, amount, orientation, and conformation are important to the cellular response.5 The adsorption of cell adhesive extracellular matrix (ECM) proteins such as fibronectin and their influence on cell behavior have been widely studied.5,6 Laminin, another major cell adhesive protein of importance in developmental biology, is often used as a coating on surfaces for the culture of neuronal cells.7,8 Laminin is a large, flexible glycoprotein with active domains for collagen binding, cell adhesion, and heparin binding.9 A recent active research area has been the interaction of biological systems with nanostructured materials. Both nanoscale topographic features10,11 and nanopatterns with, for example, the * Corresponding author. E-mail: [email protected]. Tel: +4589425547. Fax +4589423690. (1) Andrade, J. D.; Hlady, V. AdV. Polym. Sci. 1986, 79, 1-63. (2) Mrksich, M.; Whitesides, G. M. Annu. ReV. Biophys. Biomol. Struct. 1996, 25, 55-78. (3) Rechendorff, K.; Hovgaard, M. B.; Foss, M.; Zhdanov, V. P.; Besenbacher, F. Langmuir 2006, 22, 10885-10888. (4) Kasemo, B. Surf. Sci. 2002, 500, 656-677. (5) Keselowsky, B. G.; Collard, D. M.; Garcia, A. J. J. Biomed. Mater. Res. Part A 2003, 66A, 247-259. (6) Klueh, U.; Seery, T.; Castner, D. G.; Bryers, J. D.; Kreutzer, D. L. Biomaterials 2003, 24, 3877-3884. (7) He, W.; Bellamkonda, R. V. Biomaterials 2005, 26, 2983-2990. (8) Kreis, T., Vale, R., Eds. Guidebook to the Extracellular Matrix, Anchor, and Adhesion Proteins, 2nd ed.; Oxford University Press: Oxford, U.K., 1999. (9) Kleinman, H. K.; Luckenbill-Edds, L.; Cannon, F. W.; Sephel, G. C. Anal. Biochem. 1987, 166, 1-13. (10) Dalby, M. J.; Riehle, M. O.; Sutherland, D. S.; Agheli, H.; Curtis, A. S. G. Biomaterials 2004, 25, 5415-5422. (11) Andersson, A. S.; Brink, J.; Lidberg, U.; Sutherland, D. S. IEEE Trans. Nanobiosci. 2003, 2, 49-57.

nanoscale distribution of proteins12,13 or peptides14 have been shown to be capable of influencing cell behavior. A critical parameter in such studies is the surface density of adsorbed protein. A number of surface-sensitive techniques have been used for the quantification of protein adsorption.4 Optical techniques such as surface plasmon resonance (SPR),15 optical waveguide lightmode spectroscopy (OWLS),16 ellipsometry,16 and total internal reflection fluorescence spectroscopy (TIRF)17 are well suited for the study of homogeneous substrates but are less likely to be useful for nanostructured substrates. In this study, we utilize a combination of the quartz crystal microbalance with dissipation (QCM-D)18 and SPR to characterize the adsorption of laminin on substrates of different surface chemistries. The combination of these two techniques allows for the determination of not only the adsorbed mass but also the effective density of the protein layer and the viscoelastic properties of that layer.19 We also evaluate the capacity of QCM-D to be used as a quantitative technique on a nanostructured surface, where protein is adsorbed specifically in a nanopattern exploiting PLL-g-PEG as a protein-resistant background. We show that laminin forms a highly hydrated protein layer with different characteristics depending on the underlying substrate. Using a (12) Christman, K. L.; Enriquez-Rios, V. D.; Maynard, H. D. Soft Matter 2006, 2, 928-939. (13) Berry, C. C.; Curtis, A. C. G.; Oreffo, R. O. C.; Agheli, H.; Sutherland, D. S. IEEE Trans. Nanobiosci., accepted for publication, 2007. (14) Cavalcanti-Adam, E. A.; Micoulet, A.; Blummel, J.; Auernheimer, J.; Kessler, H.; Spatz, J. P. Eur. J. Cell Biol. 2006, 85, 219-224. (15) Stenberg, E.; Persson, B.; Roos, H.; Urbaniczky, C. J. Colloid Interface Sci. 1991, 143, 513-526. (16) Hook, F.; Voros, J.; Rodahl, M.; Kurrat, R.; Boni, P.; Ramsden, J. J.; Textor, M.; Spencer, N. D.; Tengvall, P.; Gold, J.; Kasemo, B. Colloids Surf., B 2002, 24, 155-170. (17) Lassen, B.; Malmsten, M. J. Colloid Interface Sci. 1997, 186, 9-16. (18) Rodahl, M.; Hook, F.; Fredriksson, C.; Keller, C. A.; Krozer, A.; Brzezinski, P.; Voinova, M.; Kasemo, B. Faraday Discuss. 1997, 229-246. (19) Reimhult, E.; Larsson, C.; Kasemo, B.; Hook, F. Anal. Chem. 2004, 76, 7211-7220.

10.1021/la701233y CCC: $37.00 © 2007 American Chemical Society Published on Web 08/11/2007

Viscoelastic Modeling of Laminin

combination of QCM-D and atomic force microscopy (AFM) data20 from nanostructured surfaces, we model laminin and subsequent antilaminin binding to nanometer-scale patches. The protein that is bound to these nanopatches has a higher surface density of adsorbed protein, which couples more water per molecule than protein layers at equivalent flat homogeneous surfaces.

Langmuir, Vol. 23, No. 19, 2007 9761

Materials. Buffers were prepared with MQ water (MilliQ A10, Millipore, Switzerland), filtered through a 0.2 µm pore filter, and degassed by the use of sonication prior to use. The buffers used were 10 mM HEPES (2-[4-(2-hydroxyethyl)-1-piperazinyl ]ethanesulfonic acid, pH 7.4) and 10 mM TRIS (tris(hydroxymethyl)aminomethan), 2.7 mM KCl, and 137 mM NaCl at pH 7.4. Laminin (L2020), polyclonal anti-laminin (L9393), and bovine serum albumin (BSA) were purchased from Sigma-Aldrich. Laminin and anti-laminin were thawed in a refrigerator, divided into aliquots, and stored at -20 °C until use, when they were diluted to the desired concentration. PolyL-ornithine (Sigma P3655) powder was dissolved in MQ water to a concentration of 10 mg/mL, divided into aliquots, and frozen. After defrosting, it was stored in a refrigerator for a maximum of 2 weeks. PLL(20)-g[3.5]-PEG(2) was provided from the group of Professor Marcus Textor, ETH Zurich, dissolved in HEPES buffer to a concentration of 0.25 mg/mL, and sterile filtered before use. The solution was divided into aliquots and stored at -20 °C until use. Substrates. QCM-D crystals, AT- cut quartz crystals with a fundamental resonance frequency of ∼5 MHz, were purchased from Q-Sense AB (Gothenburg, Sweden) with bare gold electrodes (QSX 301) or with silicon oxide coatings (QSX 303) and were used as purchased, and some gold-coated crystals were coated with 1 nm of Ti and 20 nm of SiO2 (rf magnetron sputtering (home made), 2 × 10-3 mbar argon pressure, Ti deposition rate 1 nm/s (6.45 W/cm2), SiO2 deposition rate 0.6 nm/s (2.54 W/cm2). The nanostructured surfaces were prepared on Q-Sense QSX 303 crystals by colloidal lithography21 and characterized with AFM.20 In brief, SiO2-coated QCM-D crystals were further coated with 19.5 nm of gold (0.5 nm Ti adhesion layer, AVAC HVC-600 electron beam evaporator 99.5 vol %) to a concentration of at least 10 mg/mL. Surfaces were immersed in the thiol bath after cleaning for a minimum of 6 h, transferred into clean ethanol, ultrasonicated for 5 min, transferred into MQ water, and left there for a minimum of 4-6 h or sonicated for 5 min in the cases in which surfaces survived this treatment. QCM-D. The quartz crystal microbalance is an ultrasensitive weighing device capable of sensing mass changes in the nanogram range. In the QCM-D technique,18 the damping of the crystal oscillation (the dissipation factor) is measured in addition to the frequency shift, allowing us also to determine the adsorbed mass of thick and viscous layers. The so-called Voigt mass derived from QCM-D data includes any water coupled to and in the layer.26 Measurements were performed with a Q-sense D 300 system from Q-Sense AB (Gothenburg, Sweden) where the third, fifth, and seventh overtones were recorded simultaneously with the damping of the crystal (i.e., the dissipation factor). Each experiment was run in batch mode with a gravity feed of liquid through the sensor when rinsing or exchanging buffers. Before the introduction of liquid into the sensor, it was preheated in the temperature loop for at least 2 min to ensure a stable temperature in the sensor. The liquid (1 mL) was passed to the temperature loop, followed by 0.5 mL to the sensor. Measurements were made on hydrophobic gold surfaces and on the nanostructured thiol-treated surfaces for 20 µg/mL of laminin. PLL-g-PEG was first adsorbed for 30 min in HEPES buffer (10 mM, pH 7.4), and the chamber was rinsed twice with HEPES 7.4. The buffer was then changed to TRIS (10 mM, 2.7 mM KCl, 137 mM NaCl) by rinsing twice and waiting for 10 min, after which the laminin solution of 20 µg/mL was introduced and allowed to adsorb onto the surface for 2 h. The chamber was then rinsed three times (15 min), after which 1 mg/mL BSA was introduced for 30 min (in order to block the surface against unspecific binding of the antibody) with three subsequent rinses with buffer. The polyclonal antibody (1:100 dilution, i.e., 6 µg/mL) was allowed to attach for 2 h, followed by three rinses (15 min). Polyornithine was adsorbed onto clean SiO2 QCM-D crystals. When a stable baseline was attained in MQ water, a polyornithine solution of 50 µg/mL was introduced for 1 h. After adsorption, the chamber was rinsed three times with MQ water (15 min), the crystal was taken out of the chamber, and both the crystal and chamber were dried with N2 before remounting the crystal. Laminin, BSA, and polyclonal anti-laminin were adsorbed onto the polyornithine-coated crystal as described above.

(20) Agheli, H.; Malmstrom, J.; Larsson, E. M.; Textor, M.; Sutherland, D. S. Nano Lett. 2006, 6, 1165-1171. (21) Hanarp, P.; Sutherland, D. S.; Gold, J.; Kasemo, B. Colloids Surf., A 2003, 214, 23-36. (22) Agheli, H.; Malmstrom, J.; Hanarp, P.; Sutherland, D. S. Mater. Sci. Eng. C 2006, 26, 911-917.

(23) Krozer, A.; Rodahl, M. J. Vac. Sci. Technol., A 1997, 15, 1704-1709. (24) Rodahl, M.; Hook, F.; Krozer, A.; Brzezinski, P.; Kasemo, B. ReV. Sci. Instrum. 1995, 66, 3924-3930. (25) Nomura, T.; Okuhara, M. Anal. Chim. Acta 1982, 142, 281-284. (26) Hook, F.; Kasemo, B.; Nylander, T.; Fant, C.; Sott, K.; Elwing, H. Anal. Chem. 2001, 73, 5796-5804.

Materials and Methods

9762 Langmuir, Vol. 23, No. 19, 2007 The QCM-D data was modeled using Q-tools software 2.0.6. The data from the 15 and 24 MHz signals was fitted to the viscoelastic model using one layer. Parameters used as fixed were the fluid density (1000 kg/m3) and fluid viscosity (0.001 kg/ms), and the density of the layer was kept fixed at each fit but was iterated to find the correct value. The layer parameters to be fitted were kept within boundaries: viscosity (0.0001-0.1 kg/ms), shear ((100-1) × 105 Pa), and thickness (1 × 10-11-1 × 10-6 m). Descending incremental fitting was used. A successful fit was identified by the absence of jumps between solutions in the resulting thickness, viscosity, and shear. The offset for the modeling should be set in the correct buffer and for the experiments where this buffer change was not done. A representative buffer change, induced by the pure buffer shift on a clean crystal, was added to the data. SPR. SPR utilizes surface plasmon polariton exitations, which consists of charge-density waves propagating along the interface between a metal and a dielectric material, to sense the local refractive index in the liquid close to the gold surface. The surface plasmon can be optically excited in some metals, creating an evanescent field extending into the bulk and leading to a dip in the intensity of the reflected light at the resonance wavelength. The plasmon resonance is, in our case, monitored as a shift in the angle of resonance for single-wavelength light. SPR is sensitive to changes in refractive index at the metal/liquid interface, which makes it suitable for biosensing applications. If proteins are adsorbed on a surface, then they replace the buffer next to the surface, and the change in refractive index can be converted to adsorbed (dry or optical) mass.15 Measurements were performed with a Biacore 2000 system from Biacore AB (Uppsala, Sweden). A flow rate of 5 µL/min was used, and a stable baseline was attained before adsorption. The procedure followed that of a QCM-D experiment with the same buffers and chemicals for both SiO2 surfaces and hydrophobic gold surfaces. Structured surfaces cannot be run in the SPR because of the localized surface plasmon created by the structures that would greatly influence the result. Injections were made for a maximum of 65 min because of instrument configurations, but in general the adsorption finished faster for the SPR measurements than for the QCM-D experiments because of the flow versus static conditions. X-ray Photoelectron Spectroscopy (XPS). Substrates coated in different molecular layers were dried in nitrogen after rinsing in MQ water for 5 min. XPS analysis was performed using a Kratos Axis Ultra-DLD instrument operating with a monochromatic Al KR X-ray source at a power of 150 W and an analysis area of 300-700 µm. The compositional data obtained from survey scans (0-1100 eV) was acquired at a pass energy of 80 eV. High-resolution scans were acquired at a pass energy of 20 eV. The generated data was converted to Vamas format and processed using CasaXPS software. Core-level spectra were peak fitted using a 70% Gaussian line shape and a Shirley model as the background. Accurate binding energies of the components in the spectra were determined by referencing to the C-H/C-C peak in the C 1s spectra at 285.0 eV. Statistics. The Student’s t test for equal variance was used for the statistical analysis of the data. Probability values of 97% water and a water factor of ∼40. After antibody binding, the density was found to increase slightly to 1010.9 ( 0.64 kg/m3, and the thickness was found to increase to 147 ( 8.0 nm. The shear viscosity of the layer was found to be 3.0 ( 0.5 g/ms before antibody binding and 4.1 ( 0.6 g/ms after. The elastic shear modulus was seen to have relatively large standard deviations, giving 53 ( 18 kPa for the laminin layer and 66 ( 20kPa after antibody binding. (Such limited precision for the elastic shear modulus has been reported for fibrinogen adsorption on polymer surfaces, for example.43) The Voigt masses for the two layers were 11 600 ( 420 and 14 800 ( 810 ng/cm2. The thickness of the laminin layer on the nanopatterned surface determined by the combination of QCM-D and AFM is significantly larger than that on the homogeneous surfaces as is the dry mass of protein determined by AFM but remains below the characteristic length of the long axis of the protein. The values determined by the model are reasonable, which suggests that modeling of QCM-D data of soft viscoelastic layers arranged in nanopatterns maybe applied where an independent measure of the “dry” mass is known. It may be of importance to the applicability of the modeling that the nanopattern used here had structures that were homogeneously distributed across the entire sensitive region of the crystal sensor and that these patterns had only short-range order21 with no defined orientation of the pattern relative to the quartz crystal oscillation direction. The binding of polyclonal antibodies to laminin adsorbed on the nanopatterns, calculated from AFM data, resulted in ∼4.2 antibodies per laminin. The binding is large and similar to that observed at homogeneous gold surfaces and reflects the high accessibility of the laminin protein on all of the surfaces. It is important to note here that a simple comparison of the frequency shifts observed via QCM, using the Sauerbrey equation, would (43) Weber, N.; Pesnell, A.; Bolikal, D.; Zeltinger, J.; Kohn, J. Langmuir 2007, 23, 3298-3304.

Langmuir, Vol. 23, No. 19, 2007 9767

imply that significantly more antibodies are binding to laminin on the nanostructured surface. However, the Sauerbrey equation significantly underestimates the mass of the layers, and the density of the layers changes before and after antibody binding. These two effects lead to an apparent difference that is likely an artifact. The modeling also gives values for the elastic shear modulus and shear viscosity that indicate a more rigid film for the proteins adsorbed on the nanopatterns compared to that on homogeneous surfaces (elastic shear moduli of 53 ( 18 and 8 ( 2 kPa and shear viscosities of 3.0 ( 0.5 and 1.8 ( 0.1 g/ms for the nanopatterned and homogeneous layers, respectively), which is somewhat surprising given the higher level of hydration. However, the values for the two homogeneous surface chemistries are relatively similar despite changes in the density of the films, so the situation is clearly complex. Both the homogeneous and nanopatterned layers are extremely soft, and the observed differences could easily reflect changes in the film structure both laterally (nanopatterning) and vertically (changes in film density). This is to our knowledge the first attempt to model QCM-D data of nanopatterned macromolecular films. The larger amount of protein per active surface area binding to nanoscale patches compared to the amount binding to homogeneous surfaces is seen directly from the unmodeled data comparing the AFM data on the nanopatterns to SPR data on homogeneous hydrophobic gold surfaces (371 ( 27 ng/cm2 for structured surface and 280 ( 15 ng/cm2 for homogeneous). The increased binding is not surprising if one considers the protein patch to be a 3D structure. In a hydrated state, the protein is unlikely to form a cylindrical structure sticking up from the surface given that the patch size is comparable to the long axis of the protein. The protein may well spread out the further it is from the surface, leading to a more hemispherical arrangement. Such an arrangement has been predicted for nanopatterned polymers44 and may allow a large binding capacity per active surface area on the patch compared to that on a homogeneous surface. The modeling approach used treats the frequency and dissipation shifts coming from different parts of the surface as being independent such that the contribution from the nanopatches can be modeled separately from the PLLg-PEG-coated parts of the surface. This modeling indicated a lower density layer with a larger layer thickness for protein adsorbed on nanoscale patches compared to that on homogeneous surfaces with a larger overall Voigt mass. There is a source of uncertainty in the absolute surface density of protein obtained by AFM, which originates from both the assumed density of dried protein and tip convolution effects. We cannot rule out the fact that this contributes to the larger absolute value of protein on the nanopatterned versus homogeneous interfaces. The numbers obtained from the modeling fit well with the large frequency and dissipation shifts observed for the binding of protein to the nanopatches with the frequency and dissipation shifts being similar to those at the homogeneous surface, despite the protein binding to only approximately 1/3 of the surface area. The model assumes a single layer with a constant density; real protein films are likely to have changes in density within the film, and the modeled data gives effective “average” values for films. However, a comparison of these effective values from different surfaces and conditions can give valuable input. We postulate that the protein extending out from the nanoscale patches is able to spread out because of the comparative size of the nanopattern and the size of the protein and with a significantly reduced density far from the surface, leading to a reduced effective density. (44) Patra, M.; Linse, P. Nano Lett. 2006, 6, 133-137.

9768 Langmuir, Vol. 23, No. 19, 2007

Malmstro¨m et al.

Conclusions

viscoelastic layers arranged in nanopatterns may be applied where an independent measure of the “dry” mass is known.

The use of QCM-D and SPR has allowed us to quantify the properties of laminin layers formed at homogeneous surfaces with two surface chemistries. Laminin forms very hydrated layers on the hydrophilically optimized cell culture substrate and even more so on the hydrophobic gold surface. The level of polyclonal antibody binding per surface-bound laminin is large at both surfaces, indicating accessible layers. We show that there is a significant difference in the density of the protein layers for the two surface chemistries and that despite having a lower surface density of laminin, the protein layer on the hydrophobic gold surface is significantly thicker than the layer on the hydrophilic SiO2 surface. Furthermore, we use the Voigt model in combination with data from AFM to obtain information about laminin adsorption in a nanopattern. Despite the limitations of the model, we achieve a realistic result with a larger amount of laminin adsorbing in a thicker layer of a lower density on the nanostructures. The result also fits well with indications from the raw data that further suggests that modeling of QCM-D data of soft

Acknowledgment. Some of the experimental work was carried out at the Department of Applied Physics at Chalmers, Gothenburg, Sweden. This work was funded through EC FP6 STREP NANOCUES and the Danish Research Agency (internationalisation grant 645-05-0016). We acknowledge the help of Lara Gamble and Dave Castner for the XPS experiments carried out at the National ESCA and Surface Analysis Center for Biomedical Problems (University of Washington, Seattle), which is funded by NIH grant EB-002027 from the National Institute of Biomedical Imaging and Bioengineering. Supporting Information Available: All results from both homogeneous and nanopatterned interfaces and XPS data from the homogeneous interfaces. This material is available free of charge via the Internet at http://pubs.acs.org. LA701233Y