Simultaneous Nanoplasmonic and Quartz Crystal Microbalance

Oct 4, 2008 - This paper presents a study of supported lipid bilayer (SLB) formation and subsequent protein binding using a sensor that combines local...
0 downloads 0 Views 649KB Size
Download PDF
Anal. Chem. 2008, 80, 7988–7995

Simultaneous Nanoplasmonic and Quartz Crystal Microbalance Sensing: Analysis of Biomolecular Conformational Changes and Quantification of the Bound Molecular Mass Magnus P. Jonsson,*,† Peter Jo¨nsson, and Fredrik Ho ¨ o¨k*,† Division of Solid State Physics, Lund University, SE-22100 Lund, Sweden This paper presents a study of supported lipid bilayer (SLB) formation and subsequent protein binding using a sensor that combines localized surface plasmon resonance (LSPR) and quartz crystal microbalance with dissipation (QCM-D) monitoring. The LSPR activity arises from silicon oxide (SiOx) coated nanometric apertures in a thin gold film, which also serves as the active electrode of a QCM-D crystal. Both transducer principles provide signatures for the formation of a SLB upon adsorption and subsequent rupture of adsorbed lipid vesicles. However, the two techniques are sensitive over different regions of the sample: LSPR primarily inside and on the rim of the holes and QCM-D primarily on the planar areas between the holes. Although the dimension of the lipid vesicles is on the same order as the dimension of the nanoholes, it is concluded from the response of the combined system that vesicle rupture in the nanoholes and on the planar region between the holes is synchronized. Furthermore, by determining the thickness of the SLB from the QCM-D response, the characteristic decay length of the LSPR field intensity could be determined. This made it possible not only to determine the mass and refractive index of the homogeneous SLB but also to postulate a generic means to quantify the LSPR response in terms of mass-uptake also for nonhomogeneous films. This is exemplified by measuring the adsorbed lipid mass during vesicle adsorption, yielding the critical lipid vesicle coverage at which spontaneous rupture into a planar bilayer occurs. The generic applicability and versatility of the method is demonstrated from specific protein binding to a functionalized SLB. From the absolute refractive index of the protein, provided from the LSPR data alone, it was possible to determine both the effective thickness of the protein film and the molecular mass (or number) of bound protein. The investigation of processes associated with the cell membrane is important both from a fundamental perspective and from the fact that more than half of all drugs are directed toward * Corresponding authors. E-mail: [email protected] (M.P.J.); [email protected] (F.H.). † Current address: Department of Applied Physics, Chalmers University of Technology, SE-41296, Go ¨teborg, Sweden.

7988

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

membrane-residing proteins. This, in turn, is due to the large fraction (approximately 30%) of proteins that are associated with the cell membrane and the importance of the life-sustaining processes they regulate.1 To be able to study these processes, as well as biophysical properties of cell membranes, such as their fluidity, rigidity, and the ability to form lipid domains, mimics of the cell membrane are often used as substitutes of whole cells. One popular such mimic is the planar supported lipid bilayer (SLB). SLBs can be prepared on various materials, including unmodified or chemically functionalized metals2,3 and metal oxides.4 However, on silica-based materials and mica, SLBs can be formed spontaneously upon adsorption and subsequent rupture of small unilamellar vesicles (SUVs).5,6 Because of the simplicity of this preparation procedure, these materials are by far the most commonly used supports. Examples where SLBs have been proven successful include studies of membrane proteins,7 crystallization of water-soluble proteins,8 and the interaction between membranes and living cells,9 including multivalent binding reactions.10 Recently, we and others have demonstrated successful SLB formation from vesicles also on nanostructured surfaces, of high relevance for the design of new miniaturized biosensors for studies of cell membrane related biorecognition reactions.11-13 The SLB formation process from vesicle adsorption on planar surfaces have been extensively studied using various fluorescence(1) Cooper, M. A. J. Mol. Recognit. 2004, 17, 286–315. (2) Atanasov, V.; Atanasova, P. P.; Vockenroth, I. K.; Knorr, N.; Koper, I. Bioconjugate Chem. 2006, 17, 631–637. (3) Lahiri, J.; Kalal, P.; Frutos, A. G.; Jonas, S. T.; Schaeffler, R. Langmuir 2000, 16, 7805–7810. (4) Rossetti, F. F.; Bally, M.; Michel, R.; Textor, M.; Reviakine, I. Langmuir 2005, 21, 6443–6450. (5) Brian, A. A.; McConnell, H. M. Proc. Natl. Acad. Sci. U.S.A. 1984, 81, 6159–6163. (6) Richter, R. P.; Berat, R.; Brisson, A. R. Langmuir 2006, 22, 3497–3505. (7) Salafsky, J.; Groves, J. T.; Boxer, S. G. Biochemistry 1996, 35, 14773–14781. (8) Reviakine, I.; Brisson, A. Langmuir 2001, 17, 8293–8299. (9) Mossman, K. D.; Campi, G.; Groves, J. T.; Dustin, M. L. Science 2005, 310, 1191–1193. (10) Vogel, J.; Bendas, G.; Bakowsky, U.; Hummel, G.; Schmidt, R. R.; Kettmann, U.; Rothe, U. Biochim. Biophys. Acta 1998, 1372, 205–215. (11) Huang, S. C. J.; Artyukhin, A. B.; Martinez, J. A.; Sirbuly, D. J.; Wang, Y.; Ju, J. W.; Stroeve, P.; Noy, A. Nano Lett. 2007, 7, 3355–3359. (12) Jonsson, M. P.; Jonsson, P.; Dahlin, A. B.; Hook, F. Nano Lett. 2007, 7, 3462–3468. (13) Pfeiffer, I.; Seantier, B.; Petronis, S.; Sutherland, D.; Kasemo, B.; Zach, M. J. Phys. Chem. B 2008, 112, 5175–5181. 10.1021/ac8008753 CCC: $40.75  2008 American Chemical Society Published on Web 10/04/2008

based methods14 as well as by a large number of techniques that do not require molecular labeling, including scanning probe microscopy,15 surface plasmon resonance,16 ellipsometry,17 and quartz crystal microbalance with dissipation (QCM-D) monitoring.18 The latter method is based on measurements of adsorptioninduced changes in frequency, f (cf. adsorbed mass), and energy dissipation, D (cf. rigidity), of piezoelectric AT-cut quartz crystal resonators.19 The technique is particularly valuable owing to its unique signature for the successful formation of a SLB.18 This stems from the fact that, for adsorbed vesicles, the mass contribution to the change in f includes solvent that is coupled both within and between vesicles. Upon vesicle rupture into a planar SLB, the solvent no longer remains coupled to the crystal oscillation, resulting in a decrease of the coupled mass. Vesicle rupture also results in an increased film rigidity, being reflected in a characteristic decrease in D.18 A signature for bilayer formation was also reported by us in a recent work12 using a sensor based on localized surface plasmon resonance (LSPR), which is a transducer principle that has been increasingly utilized in bioanalytical sensors the last 10 years.20,21 The LSPR concept is based on metal nanostructures, whose conduction electrons couple with the field oscillations of electromagnetic waves. The coupling is strongest at certain resonance wavelengths, which results in characteristic extinction peaks. The fact that the LSPR resonance condition, which determines the wavelength at maximum extinction (peak position), is highly sensitive to changes in refractive index within the rapidly decaying (tens of nanometers) nanoplasmonic field forms the base for the use of the concept in sensor applications. To mention a few examples, LSPR has been utilized for detection of biomarkers for Alzheimer’s disease22 as well as for both imaging and treatment of cancer cells.23-26 Our extension of the LSPR concept to studies of continuous SLBs was realized using silicon oxide (SiOx) coated thin gold or silver films perforated with short-range ordered nanometer-sized apertures.12 The signature for SLB formation, manifested as an acceleration in the temporal sensor response, was attributed to the fact that upon vesicle rupture, the average lipid distribution is moved closer to the sensor surface and thus into a region of the sensor where the LSPR field is stronger. In the present work we have utilized the fact that a SiOx-coated gold film perforated with nanoholes remains electrically conductive, which enables the LSPR active surface to be used as one of the electrodes of a QCM-D crystal, as depicted schematically in (14) Johnson, J. M.; Ha, T.; Chu, S.; Boxer, S. G. Biophys. J. 2002, 83, 3371– 3379. (15) Leonenko, Z. V.; Carnini, A.; Cramb, D. T. Biochim. Biophys. Acta 2000, 1509, 131–147. (16) Keller, C. A.; Glasmastar, K.; Zhdanov, V. P.; Kasemo, B. Phys. Rev. Lett. 2000, 84, 5443–5446. (17) Puu, G.; Gustafson, I. Biochim. Biophys. Acta 1997, 1327, 149–161. (18) Keller, C. A.; Kasemo, B. Biophys. J. 1998, 75, 1397–1402. (19) Rodahl, M.; Hook, F.; Krozer, A.; Brzezinski, P.; Kasemo, B. Rev. Sci. Instrum. 1995, 66, 3924–3930. (20) Englebienne, P. Analyst 1998, 123, 1599–1603. (21) Willets, K. A.; Van Duyne, R. P. Annu. Rev. Phys. Chem. 2007, 58, 267– 297. (22) Haes, A. J.; Chang, L.; Klein, W. L.; Van Duyne, R. P. J. Am. Chem. Soc. 2005, 127, 2264–2271. (23) El-Sayed, I. H.; Huang, X.; El-Sayed, M. A. Nano Lett. 2005, 5, 829–834. (24) El-Sayed, I. H.; Huang, X.; El-Sayed, M. A. Cancer Lett. 2006, 239, 129– 135. (25) Jain, P. K.; El-Sayed, I. H.; El-Sayed, M. A. Nano Today 2007, 2, 18–29. (26) Kumar, S.; Harrison, N.; Richards-Kortum, R.; Sokolov, K. Nano Lett. 2007, 7, 1338–1343.

Figure 1. Schematic illustration of the sensor concept. (A) The flashlight and the blue box represent the light source and the spectrometer, respectively. The electric circuit represents the QCM-D read-out system, which together with the spectrometer is connected to a computer. (B) A schematic illustration depicting a small area of the sensor surface. (C) Illustration of the cross section through a nanohole after bilayer formation (not drawn to scale).

Figure 1. Hence, simultaneous QCM-D and LSPR measurements of reactions that occur on the same surface can be obtained, as recently demonstrated in a previous study of antigen-antibody interactions.27 The localized nature of the LSPR field makes the LSPR signal most sensitive to processes that take place in the void of the holes.28 In contrast, because only around 12% of the total projected surface area of the sample is constituted by holes, the QCM-D response will be dominated by binding processes that occur on the planar area surrounding the holes. This means that the LSPR and QCM-D responses can, to a good approximation, be used to deduce information of processes that occur in and between the nanoholes, respectively. In the present work, this was utilized to investigate adsorption and subsequent rupture of lipid vesicles upon SLB formation on the nanostructured surface, a process that was recently shown to depend on the nanoscale topography.13 It is also emphasized that information about the thickness of the SLB obtained from the QCM-D response is, together with LSPR bulk-sensitivity measurements, shown to provide a new way of quantifying the characteristic decay length of the electromagnetic field outside the nanostructured surface. Apart from providing an improved understanding of the plasmonic properties of the nanostructure, the decay length was used to quantify the LSPR response in terms of adsorbed molecular mass. The generic applicability of the quantification was demonstrated by determining the mass of both adsorbed vesicles at the critical surface coverage, at which the vesicles start to rupture into a SLB, and the protein NeutrAvidin bound to a SLB functionalized with a fraction (1%) of biotinylated lipids. This new and generic way of quantifying the LSPR response relies on the information provided from the QCM-D data and a determination of the absolute refractive index of the bound entities, provided from bulk sensitivity measurements of the LSPR sensor prior to and after saturated film formation. (27) Dahlin, A. B.; Jonsson, P.; Jonsson, M. P.; Schmid, E.; Zhou, Y.; Hook, F. ACS Nano 2008, accepted. (28) Dahlin, A. B.; Jonsson, M.; Ho ¨o ¨k, F. Adv. Mater. 2008, 20, 1–7.

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

7989

MATERIALS AND METHODS Fabrication. The sensor substrates were fabricated from 1 in. QCM crystals with a fundamental resonance frequency of 5 MHz (MaxTec Inc.). An approximately 4 mm diameter hole in the backside electrode was defined by conventional UV-photolithography (MJB3 mask aligner, Karl Su¨ss) followed by wet etching of the gold (4 g of KI, 1 g of I2, and 40 mL of deionized water) and the chrome (20 g of (NH4)2Ce(NO2)6 (CAN), 35 mL of CH2COOH) in the electrode. Similarly, the frontside electrode was removed completely, although a thin stripe connecting the frontside electrode with the backside electrode was kept. The substrates were then immersed in a mixture of Milli-Q water, 25% ammonia, and 30% hydrogen peroxide (1:1:5 by volume) at approximately 70 °C for 15 min followed by thoroughly rinsing with Milli-Q water (RCA 1 cleaning). The nanostructured surface was fabricated by colloidal lithography to obtain a long-range randomly ordered structure, yet with a narrow nearest neighbor spacing distribution (short-range ordered).29 In brief, 140 nm in diameter polystyrene colloids (Interfacial Dynamics Corporation) were self-assembled on the sensor surface after functionalization of the surface with a triple layer of electrolytes. The triple layer was subsequently removed by a short oxygen plasma etch (PlasmaPreen) to improve the quality of the sample. At the same time, the colloids were etched approximately 10 nm. Chrome (1.5 nm) and 30 nm gold were thermally evaporated followed by removal of the colloids using clean room tape leaving a thin LSPR active QCM frontside electrode on the crystal. The substrate was finalized by sputtering an approximately 20 nm thick SiOx layer (Orion 5, AJA International) to fully coat the metal. A thin layer (around 2 nm) of chrome on top of the metal was used for improved adhesion also in this step. SEM inspections (Nanolab 600, FEI Company) before and after sputtering confirmed that SiOx was deposited also on the sides of the nanoholes (data not shown). Combined QCM-D and LSPR Measurements. The combined setup was built in-house, where the QCM-D setup was based on the work by Rodahl and Kasemo.30 The crystal oscillations were driven by a frequency generator (33250A, Agilent) that was controlled via GPIB, while a digital oscilloscope (model 490, Nicolet) monitored the decay of the oscillations. A thermoelectric module, together with a temperature controller (MPT-5000, Wavelength Electronics), was used to achieve a stabilized temperature during the measurements. LSPR extinction spectra were recorded continuously using a conventional spectrometer (BTC611E, B&WTek). Light from a tungstenhalogen lamp (HL-2000, Ocean Optics) was guided by an optical fiber and collimated by a lens through the hole in the backside electrode. Light transmitted through the nanostructure was then coupled, via a second lens, into an optical fiber that was connected to the spectrometer. To obtain more accurate absolute values of the extinction spectra, a reference sample was used. This sample was identical to the active substrates but without an electrode facing the liquid. QCM-D frequency and dissipation at several overtones was calculated and plotted continuously together with the LSPR peak position in a custom designed LabView program (LabVIEW 8.0, National Instruments). In all experiments, a syringe was used to ensure that the liquid in the flow cell was completely exchanged. (29) Hanarp, P.; Sutherland, D. S.; Gold, J.; Kasemo, B. Colloids Surf., A 2003, 214, 23–36. (30) Rodahl, M.; Kasemo, B. Rev. Sci. Instrum. 1996, 67, 3238–3241.

7990

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

Table 1. Experimental and Theoretical Values of Frequency Shifts Induced by the Modification of the QCM-D Crystal with a Nanostructured Electrodea ν

medium

3 3 5 5

air water air water

∆fexp/kHz

∆ftheory/kHz

57.8 57.75 96.5 97.0

60 60 100 100

a ν is the overtone number and ∆fexp and ∆ftheory are the measured and theoretically predicted shifts, respectively.

Median averaging was used in the analysis of the data to remove spikes from the liquid exchanges. Preparation of Lipid Vesicles. SUVs were prepared by the vesicle extrusion principle.31 Egg PC (99 wt %) (L-a-phosphatidylcholine) and 1 wt % biotin-PE (1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(biotinyl)(sodium salt) (5 mg in total and both from Avanti Polar Lipids) was dissolved in 1 mL of methanol in a round flask and dried under a nitrogen environment for at least 1 h. The lipids were rehydrated in 1 mL of NaTrisEDTA buffer (100 mM NaCl, 10 mM TRIS, 1 mM EDTA [ethylenediaminetetraacetic acid], pH 8.0) to yield a lipid concentration of 5 mg/mL. The solution was extruded 11 times through a 100 nm polycarbonate membrane followed by extrusion 11 times through a 30 nm membrane (Whatman) to yield approximately 35 nm in diameter unilamellar lipid vesicles.31 The vesicles were stored at 4 °C and used within 2 weeks. A diluted suspension of vesicles in buffer (100 µg/mL) was used in all experiments. RESULTS AND DISCUSSION Characterization of the Piezoelectric Properties of the Sensor. The QCM resonance frequency of the sensor crystal was measured at various steps in the fabrication process. The resonance frequency increased significantly after etching the hole in the backside electrode. The modification of the frontside electrode with a LSPR active electrode resulted in an additional increase in peak position, which is attributed to the lower mass of the nanostructured surface compared with the initial electrode. The expected peak shift, ∆ftheory, due to a reduction in the mass of the electrode can be estimated from32 ∆ftheory ) -

f0ν ∆Γ Fqtq

(1)

where f0 is the fundamental frequency, ν is the overtone number, Fq is the density of the quartz (2.65 g/cm3),33 tq is the thickness of the crystal (334 µm), and ∆Γ is the change in mass per unit area. With an initial gold electrode thickness of 210 nm, the frequency shift expected from modifying the frontside electrode with a 30 nm thick gold film perforated with nanoholes is around 20 kHz multiplied with the overtone number. The experimental and theoretical shifts are summarized in Table 1, where differences in the adhesion layers and the SiOx coating are not considered, since these do not contribute significantly to the change in mass, while the holes, which cover around 12% of the total projected surface area, were taken into account. For both (31) Patty, P. J.; Frisken, B. J. Biophys. J. 2003, 85, 996–1004. (32) Rodahl, M.; Kasemo, B. Sens. Actuators, A 1996, 54, 448–456.

Figure 2. (A) Frequency (blue) and dissipation shifts (red) at the third overtone for various concentrations of glycerol. Squares represent experimental values and the solid lines are the theoretical predictions. (B) The squares are experimental peak position shifts versus refractive index. The full line is a linear fit to the values around the refractive index of water giving the bulk sensitivity 47.4 nm/RIU and the dashed line is a second-order polynomial fit including also the peak shift from air to water. The inset shows an extinction spectrum acquired in water for the sample used in the analysis.

the third and fifth overtone, there is an excellent match (within 4%) between the experimental and theoretical values. When immersed in a viscous liquid, both f and D are expected to change according to32

∆fv ) -

1 2tqFq

∆Dν )

1 tqFq





f0vFη π

(2)

Fη νπf0

(3)

where η is the viscosity and F is the density of the liquid. As shown in Figure 2A, there is qualitative agreement between the experimental and theoretical values when the sensor is immersed in increasing concentrations of glycerol (0-35 wt % in steps of 5 wt %). However, the experimental values for ∆f3 and ∆D3 are approximately 19% and 17% lower than expected from theory, respectively. One explanation to this deviation could be inefficient liquid exchange. However, the same measurement using a traditional QCM crystal, instead of a LSPR active crystal, showed significantly smaller deviations (data not shown), which excludes this source of error. The observation is rather attributed to the small thickness (30 nm) of the conductive part of the active electrode, which increases its electrical resistance. Hence, already minor changes in conductivity of the liquid bulk influences the

resistance of the effective electrode and therefore also the frequency and dissipation.34 This interpretation is supported by the fact that the deviation was significantly reduced for an approximately 2 times thicker gold film perforated with similar nanoholes.27 It is also worthwhile to point out that additional peaks close to the resonance position become more pronounced for the modified crystals, which may also influence the QCM-D response upon large external perturbations. However, as illustrated below, the minor shifts induced upon adsorption of, e.g., biomolecules to the nanostructured thin electrode are identical to those observed for unmodified crystals. Characterization of the Optical Properties of the Sensor. The LSPR sensitivity to changes in bulk refractive index was measured in increasing concentrations of glycerol (0-35 wt % in steps of 5 wt %) (see Figure 2B). Tabulated values of the refractive index at different concentrations were obtained from Hoyt et al.35 As shown in Figure 2B, a second-order polynomial fit rather than a linear fit to the peak position versus refractive index represents the data best. This is in agreement with previous observations, which showed that the sensitivity for gold nanohole structures obeys a slight increase with increasing refractive index.12 Although less pronounced, this observation is also in agreement with the behavior of conventional surface plasmon resonance sensors.36 However, the difference in bulk sensitivity, S, between typical aqueous solutions, as well as for biological materials, can in most practical cases be neglected. At a refractive index of water, the sensitivity was determined to 47.4 nm/RIU. This value is somewhat lower than previously reported by us for a similar LSPR sensor,12 a difference that is attributed to a slightly thicker SiOx layer originating from a refined fabrication procedure. This, in turn, significantly improved the reusability of samples between experiments. However, the difference in substrate material (quartz instead of borosilicate glass) may also influence the plasmonic properties. Structural Information Contained in the Combined Sensor Response. There exists extensive literature that addresses spontaneous formation of a SLB from adsorption and subsequent rupture of lipid vesicles using the QCM-D technique.4,6,16,18,37 This makes the process ideal as a model system to characterize the potential of the combined QCM-D/LSPR system. Figure 3A shows the temporal variation in ∆f3/3 and ∆D3 during SLB formation. SUVs with a nominal diameter of approximately 35 nm were injected at t ) 0 min and subsequent adsorption to the surface resulted in significant frequency and dissipation shifts. In agreement with results for SLB formation on planar silica-coated QCM crystals, there was, at around t ) 5.5 min, an increase and a decrease in ∆f and ∆D, respectively. These features in the responses are attributed to the release of coupled solvent and an increase in rigidity upon vesicle rupture. Finally, ∆f3/3 saturated at approximately -24 Hz at complete SLB formation with a total increase in D3 of less than 1 × 10-6. These values are consistent with those expected from SLB formation on both unstructured and nanostructured silica-coated QCM crystals,13,18 which proves Kanazawa, K. K.; Gordon, J. G. Anal. Chim. Acta 1985, 175, 99–105. Rodahl, M.; Hook, F.; Kasemo, B. Anal. Chem. 1996, 68, 2219–2227. Hoyt, L. F. Ind. Eng. Chem. 1934, 26, 329–332. Jung, L. S.; Campbell, C. T.; Chinowsky, T. M.; Mar, M. N.; Yee, S. S. Langmuir 1998, 14, 5636–5648. (37) Reimhult, E.; Hook, F.; Kasemo, B. Langmuir 2003, 19, 1681–1691. (33) (34) (35) (36)

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

7991

Figure 3. (A) Temporal variation of the dissipation, ∆D, (red) and the frequency, ∆f/3, (blue) at the 3rd overtone during bilayer formation on the nanostructured surface. (B) Same as in part A but instead showing the dissipation (red) and the LSPR peak shift (green). An initial (at around 0 min) drop in peak shift due to a wavelength dependence on the scattering of vesicles has been removed. The vertical dashed lines in both parts A and B are there as visual aids and to demonstrate the temporal correlation between the turnover in the dissipation and the LSPR kink, respectively. The short black line and the arrow demonstrate the possibility to investigate the initiation of vesicle rupture from the LSPR response.

the successful formation of a SLB formation. This is also in agreement with our previous report, in which fluorescence imaging and fluorescence recovery after photobleaching38 were used to confirm the formation of a continuous and laterally mobile SLB on the same type of support.12 It is also interesting to note that despite a hole coverage of 12%, the saturated frequency shift is essentially identical to that observed on a planar silica substrate. This is consistent with the fact that the QCM technique senses coupled solvent. As a consequence, the contribution from the fraction of the SLB that replaces the coupled solvent within the hole will, due to the buoyancy match between the lipids and the solvent, not contribute to changes in the frequency, while the additional solvent coupled due to the thickness increase at the top of the SLB-containing nanohole will be sensed as an additional rigid mass. With the QCM-D response as an independent verification of successful SLB formation on the nanohole-containing surface, the synchronized LSPR response (see Figure 3B) provides information that allows the SLB formation process to be analyzed in further detail. Initially, there is a monotonic increase in the peak shift, ∆λ, due to adsorption of vesicles that induces local changes in (38) Jonsson, P.; Jonsson, M. P.; Tegenfeldt, J. O.; Hook, F. Biophys. J. DOI: 10.1529/biophysj.108.134874.

7992

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

the refractive index in the proximity of the LSPR active surface. At around 5.5 min, there is a significant acceleration of the response, resulting in a kink in the curve. Thereafter, the signal increases monotonically until saturation of ∆λ at a value of 1.37 nm. As argued in our previous work,12 the kink is attributed to the rupture process of adsorbed vesicles, which leads to a change in the average lipid distribution toward the surface of the nanostructure. Because of a rapidly decaying LSPR field (see below),39 the lipids thus enters a region with a stronger field, which, in turn, results in an increase in the peak shift. Because the LSPR and the QCM-D responses are dominated by binding reactions in the void of the holes and on the planar regions between the holes, respectively, the excellent temporal correlation between the kink and the turning point in ∆D suggests that the onset of bilayer formation occurs at the same time over the entire surface. Hence, in contrast to previous reports,13 vesicle binding to regions of positive or negative curvature in the holes seems neither to accelerate nor hamper the cooperative SLB formation process under these conditions. It is also worthwhile to emphasize that the turning point in ∆D corresponds to the point in time when the positive contribution to D from vesicle adsorption equals the decrease in D due to the rupture of vesicles. Hence, the actual onset of bilayer formation cannot be directly obtained from QCM-D data. In contrast, the LSPR response provides a unique signature in this respect, because the onset of the vesicle rupture process corresponds to the point in time at which the acceleration starts. As illustrated by the black line and the arrow in Figure 3B, the rupture is initiated at a coverage that corresponds to roughly 90% of the coverage at which ∆D reaches it maximum value. Quantification of the Sensor Response in Terms of Bound Molecular Mass. Besides new structural information about the SLB formation process, the combined LSPR and QCM-D responses provide the base for an improved way to quantify the LSPR response in terms of adsorbed molecular mass as well as a determination of the absolute refractive index of the adsorbed entities. This is exemplified for both the SLB formation process and a nonhomogeneous protein film specifically bound on top of the SLB. Determination of the Effective Decay Length of the LSPR Field. Because of the similarity in dimension of the biomolecular systems under investigation and the decay length, L, of the evanescent field associated with the LSPR active substrate, a first requirement to be able to quantify the LSPR response is a determination of L. Following the reasoning for conventional SPR sensors, introduced by Jung et al.,36 a film refractive index n(z), where z is the distance from the surface, can be translated into an effective refractive index, neff, of the probed volume by weighting n(z) with the field intensity, I(z), according to

∫ I(z)n(z) dz ) ∫ I(z) dz ∞

neff

z)0

∞

(4)

z)0

Although the exact geometry of the LSPR field for the particular nanostructure used in this work is not known, it is (39) Rindzevicius, T.; Alaverdyan, Y.; Dahlin, A.; Hook, F.; Sutherland, D. S.; Kall, M. Nano Lett. 2005, 5, 2335–2339.

represented with an effective exponential decay, which has proven adequate as a sufficient approximation for various LSPR structures.40,41 For an exponentially decaying field, eq 4 is reduced to

neff )

1 L

∫

∞

z)0

n(z) e-z⁄L dz

(5)

where L is the decay length of the LSPR field intensity. The peak shift, ∆λ, induced by a change in the refractive index, (neff - nbuffer), is given by ∆λ ) S(neff - nbuffer)

(6)

where S is the bulk sensitivity of the sensor. For the situation with a SLB with a thickness t and a refractive index nSLB, the effective refractive index becomes

neff )

nSLB L

∫

t

z)0

e-z⁄L dz +

n L

∫

∞

z)t

e-z⁄L dz ) nSLB(1 - e-t⁄L) + ne-t⁄L(7)

where n is the refractive index of the surrounding medium. By inserting eq 7 into eq 6, the following expression is obtained ∆λ ) S(1 - e-t⁄L)(nSLB - nbuffer) + Se-t⁄L(n - nbuffer)

(8)

With n ) nbuffer, this expression yields the peak position shift induced upon SLB formation (∼1.37 nm). Equation 8 also describes the second glycerol cycle in Figure 4B (after bilayer formation), where n is the refractive index of the glycerol suspensions. Thus, the slope of the curve, S′, is S e-t/L, and the following relation describes the ratio of the slopes of the sensitivity curves prior to and after SLB formation: S ) e-t⁄L S

(9)

where S and S′ were determined to 47.4 nm/RIU and 37.9 nm/ RIU, respectively (see Figure 4). The reduction in sensitivity is thus represented by the exponential factor in eq 9, which equals the fraction of the total LSPR field intensity that is not occupied by the SLB. Since the SLB is a homogeneous and acoustically rigid (low ∆D),42 the sensed mass per unit area, ∆ΓQCM, and thickness, t, can be obtained from the QCM-D data using the Sauerbrey relation, which is a rearrangement of eq 1:42 C ∆ΓQCM ) tF ) - ∆fν ν

(10)

where F is the effective density of the adsorbed film, t is the thickness, and C ) 17.7 ng Hz-1 cm-2 is the mass sensitivity constant.42 The mass of the SLB was determined to 423 ng/cm2. (40) Haes, A. J.; Zou, S. L.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2004, 108, 109–116. (41) Rindzevicius, T.; Alaverdyan, Y.; Kall, M.; Murray, W. A.; Barnes, W. L. J. Phys. Chem. C 2007, 111, 11806–11810. (42) Reimhult, E.; Larsson, C.; Kasemo, B.; Hook, F. Anal. Chem. 2004, 76, 7211–7220.

Figure 4. (A) Temporal variations of the peak position, ∆λ, due to the following sequential steps: (1) Exchange of the bulk solution (and hence the bulk refractive index) using various concentrations of glycerol (0-35 wt % in steps of 5 wt %). (2) Bilayer formation from SUVs. (3) Exchange of the bulk solution using various glycerol concentrations again. (4) Adsorption of NeutrAvidin to the functionalized bilayer. (5) A third and final cycle of addition of various glycerol concentrations to the bulk. The refractive index of the buffer, nbuffer, was determined to 1.335 from the bulk sensitivity and the LSPR shift induced upon exchanging water with buffer. (B) The peak shift versus the refractive index for the experimental values (squares) extracted from part A for the glycerol cycle before (blue), after bilayer formation (red), and after adsorption of NeutrAvidin (green). The full lines are linear fits to the experimental data. The inset is a close-up of the same graph, and the black dashed lines mark the intersections that represent the refractive index of the SLB and NeutrAvidin, respectively.

With the use of F ) 1.02 g/cm3 for egg PC lipids at 20 °C,43 this gives a corresponding thickness of 4.16 nm, which is close to the values obtained using other techniques.44,45 Using eq 9, this yields a decay length of ∼19 nm, which is in good agreement with previous reports on studies of gold nanohole structures.41 Determination of the Absolute Refractive Index of Adsorbed Entities. A second requirement in order to estimate the bound molecular mass using sensors based on measurements of changes in interfacial refractive index is an estimate of either the specific refractivity, r, of the adsorbed entities or knowledge about their refractive index increment versus concentration, dn/dC.42 To(43) Tirosh, O.; Barenholz, Y.; Katzhendler, J.; Priev, A. Biophys. J. 1998, 74, 1371–1379. (44) Schonherr, H.; Johnson, J. M.; Lenz, P.; Frank, C. W.; Boxer, S. G. Langmuir 2004, 20, 11600–11606. (45) Tawa, K.; Morigaki, K. Biophys. J. 2005, 89, 2750–2758.

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

7993

gether with information of the film thickness, t, and the effective refractive index of the film, nfilm, the adsorbed mass can then be obtained from42,46,47

∆ΓLSPR )

≈

3t(nfilm - nbuffer)(nfilm + nbuffer) 2 (nfilm + 2)[r(nbuffer2 + 2) - v(nbuffer2 - 1)] t(nfilm - nbuffer) dn ⁄ dC

(11)

where ν is the specific volume (inverse of the density) of the adsorbed entities. While dn/dC can be determined experimentally by estimating the refractive index of suspensions at increasing concentrations of the probed entities,47 the small sensing volume associated with LSPR active nanoparticles enables a determination of the absolute refractive index of the probed entities in the adsorbed state, as previously shown for an alkanethiol monolayer on silver nanoparticles.48 With information of the absolute refractive index of the entities, nmolecule, the specific refractivity can be obtained from46

r)v

nmolecule2 - 1 nmolecule2 + 2

(12)

The estimate of nmolecule in the present work was obtained from the LSPR bulk sensitivities (see above) measured before (47.4 nm/RIU, blue curve) and after bilayer formation (37.9 nm/RIU, red curve) (Figure 4). In Figure 4B, the sensitivity curves obtained before and after SLB formation are separated by the shift induced by the SLB (1.37 nm at nbuffer). As discussed above, the observed decrease in sensitivity after bilayer formation is attributed to a significant fraction of the total LSPR field being occupied by the SLB. Furthermore, the decreasing separation between the curves as the bulk refractive index increases (Figure 4B) illustrates the expected sensor response induced upon SLB formation at different bulk refractive indices. Hence, the absolute refractive index of the SLB corresponds to the bulk refractive index at the intercept of the two sensitivity curves (see Figure 4B). Under the reasonable assumption that the linear extrapolation is a sufficiently good approximation of the peak shifts at increasing refractive index values, this yields an absolute refractive index of the SLB, nSLB, of 1.48 ± 0.01 (obtained from three independent measurements) (see Figure 4B). It is in this context relevant to note that SLBs are known to be optically anisotropic.49 However, considering the presumed geometry of the SLB on the nanostructure (see Figure 1) the measured value of the refractive index should be considered an averaged value. Indeed, the obtained value from the extrapolation method is in excellent agreement with reported values for the average refractive index of SLBs composed of similar lipids.49 The specific refractivity, obtained using eq 12 and a specific volume of 0.979 cm3/g,43 becomes 0.279 cm3/g, which is in perfect (46) Cuypers, P. A.; Corsel, J. W.; Janssen, M. P.; Kop, J. M.; Hermens, W. T.; Hemker, H. C. J. Biol. Chem. 1983, 258, 2426–2431. (47) Defeijter, J. A.; Benjamins, J.; Veer, F. A. Biopolymers 1978, 17, 1759– 1772. (48) Haes, A. J.; Zou, S. L.; Schatz, G. C.; Van Duyne, R. P. J. Phys. Chem. B 2004, 108, 6961–6968. (49) Salamon, Z.; Tollin, G. Biophys. J. 2001, 80, 1557–1567.

7994

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

agreement with literature values.50 Similarly, the specific refractivity of adsorbed proteins was determined using a SLB containing 1% biotinylated lipids, to which NeutrAvidin was specifically bound (Figure 4A). With the performance of an additional bulk-sensitivity measurement after NeutrAvidin adsorption (green curve in Figure 4B), the refractive index of the protein was determined to be 1.55, which, using 0.73 cm3/g for the specific volume of NeutrAvidin,51 yields a specific refractivity of 0.233 cm3/g. This is close to literature values for the similar avidin analogue streptavidin,42 demonstrating the generic applicability of the method to characterize the refractive index of molecular entities in the adsorbed state. Determination of the Bound Molecular Mass for Inhomogeneous Lipid Vesicle and Protein Films. While the mass of the rigid and homogeneous SLB, which does not couple a substantial amount of solvent, could be estimated directly from the QCM-D response (see above), the situation is more complicated for nonrigid and/or inhomogeneous films, such as films composed of adsorbed vesicles or proteins.52 In this section, it is demonstrated how the estimates of the decay length of the LSPR field intensity and the specific refractivity of the SLB and bound proteins, respectively, can be utilized to quantify the bound molecular mass of inhomogeneous films. This is first exemplified with the adsorption process of vesicles prior to rupture, with the specific aim to quantify the molecular mass at the critical coverage (the turnover in ∆D). The effective refractive index of a nonhomogeneous film, nfilm, (not to be confused with the effective refractive index of the probed volume, neff) can be obtained from eq 7 by replacing nSLB with nfilm and setting n ) nbuffer. This yields

nfilm )

neff - nbuffer (1 - e-t⁄L)

+ nbuffer

(13)

where the thickness, t, is defined as the height of the adsorbed entities. With ∆λ equal to the peak shift induced at the critical vesicle coverage, neff is given directly from eq 6. The mass of the adsorbed molecules, ΓLSPR, can then be directly determined from eq 11,42,46 given that the film thickness is known. At the critical coverage ∆λ ) 0.75 nm (see Figure 3B), which using S ) 47.4 nm/RIU inserted into eq 6 yields neff ) 1.351. A complicating factor in the case of adsorbed vesicles is the fact that the film thickness cannot be directly obtained from the dimension of the vesicles in suspension, because they tend to deform upon adsorption,53 or from the QCM-D data, because an estimate of the film thickness is a good approximation only at high coverage (see below). The magnitude of the contraction depends on the rigidity of the vesicles as well as the adhesion strength of the binding, and on silica, a width-to-height ratio of 2:1 has been suggested.53 If the surface area of a 35 nm vesicle is assumed to be retained after deformation, the film thickness becomes approximately 21 nm, which from eq 13 gives nfilm ≈ 1.358. Inserted into eq 11, this yields, at the critical coverage, a mass of the adsorbed vesicles of 357 ng/cm2. Interestingly, this corresponds to around 84% of the total mass of the final SLB, which was determined from the QCM-D data (see above) and this is close to previously published values of the vesicle mass at the critical coverage.53 The corresponding mass at the critical coverage obtained by QCM-D, from

∆f3/3 in Figure 3A and eq 4, gives an adsorbed mass of 1158 ng/ cm2, which illustrates the high amount of coupled solvent measured by the QCM-D in the situation of adsorbed vesicles. Finally, it is demonstrated how the combined QCM-D/LSPR setup can be used to quantify the molecular mass in the case of bound proteins, which represents the class of biological entities most frequently probed using LSPR sensing. This is exemplified using NeutrAvidin bound to a biotin-modified SLB (Figure 4), and it is emphasized that the quantification requires no other information than the density (or specific volume) of the proteins, which is in most cases a well-known parameter. At high surface coverage and if the adsorbed entities form a rigid film, essentially all solvent in between the adsorbed molecules can be assumed to be coupled to the QCM oscillation.50 The thickness of the adsorbed entities can therefore be determined from eq 10 using an effective film density described by42 ΓQCM ΓLSPR ΓQCM - ΓLSPR + Fmolecule Fbuffer

Ffilm )

(14)

Inserting eq 14 into eq 10 gives

(

t ) ΓLSPR

)

ΓQCM 1 1 + Fmolecule Fbuffer Fbuffer

(15)

By replacement of ΓLSPR in eq 15 with the expression in eq 11 and nfilm with the expression in eq 13, the following equation is obtained

t

(

3(R2 - nbuffer2)

)

1 1 + (R2 + 2)(r(nbuffer2 + 2) - v(nbuffer2 - 1)) Fmolecule Fbuffer ΓQCM - t ) 0 (16) Fbuffer

where R)

(

neff - e-t⁄Lnbuffer 1 - e-t⁄L

)

With a LSPR bulk sensitivity of 38.9 nm/RIU, as determined after bilayer formation, and with ∆λ)0.48 nm upon NeutrAvidin binding, neff becomes 1.347 (see eq 6). From the simultaneously measured QCM-D response, the QCM mass was determined to 478 ng/cm2 (data not shown). Hence, using the measured value of 0.233 cm3/g for the specific refractivity and with densities of NeutrAvidin51 and the buffer of 1.370 and 0.998 g/cm3, respectively, eq 16 yields a film thickness of the NeutrAvidin layer of 4.3 nm. With the use of eq 11, this finally gives a mass uptake of 177 ng/cm2, which is in excellent agreement with the expected value using an SLB containing 1% biotin.42,54 CONCLUDING REMARKS In this work, a LSPR active nanohole surface was used as one of the operating electrodes of a QCM-D crystal to probe

SLB formation on SiOx and subsequent adsorption of NeutrAvidin. The synchronized responses were used to show that vesicle rupture is, within experimental uncertainty, initiated at the same time in the nanoholes and on the planar regions between the holes. This reveals the potential of the combined sensor design to provide valuable information in studies of biomacromolecular interactions that involve conformational changes. Furthermore, the combined sensor enables a new approach to characterize the LSPR response and was shown to provide a means for quantitative mass determinations, exemplified using both adsorbed lipid vesicles and proteins. The proposed mass determination requires information about the specific volume (density) and the specific refractivity (see eq 11) of the molecular system. While the density of molecular entities is generally known, it was shown how the latter value (see eq 5) could be determined by simply extrapolating the sensitivity curves obtained prior to and after SLB formation or protein binding. Because this approach of determining the absolute refractive index of adsorbed molecules applies also for inhomogeneous films, such as those formed upon protein adsorption, the method is considered applicable on any biomolecular system and thus adds a generic way of quantifying the response of LSPR sensors. It is also worthwhile to point out that due to the fact that the decay length of the evanescent LSPR field is on the same order as the dimension of the probed entities, the mass determination is sensitive to the choice of the effective film thickness, in this work estimated from the QCM-D response in the case of adsorbed proteins or from the molecular dimension in the case of vesicles. However, a change in effective film thickness results in an opposite change in the magnitude of the film refractive index (see eq 13). Considering that the mass depends nearly linearly on both these parameters (see eq 11), an uncertainty in the assumed thickness of 20% in the case of vesicles (t ∼ 21 nm), translates into an uncertainty of around 8% in the estimated mass uptake for the adsorbed vesicles. For thinner films, such as the film composed of proteins (t ∼ 4.3 nm), this uncertainty is significantly reduced. ACKNOWLEDGMENT We thank Dr. Michael Rodahl for valuable discussions. The work was financially supported by the Swedish Research Council for Engineering Sciences, Contract Number 2005-3140, and the Ingvar grant from the Strategic Research Foundation.

Received for review April 30, 2008. Accepted August 15, 2008. AC8008753 (50) Wang, G.; Rodahl, M.; Edvardsson, M.; Svedhem, S.; Ohlsson, G.; Hook, F.; Kasemo, B. Rev. Sci. Instrum. 2008, 79, 075107. (51) Burrows, S. M.; Pappas, D. Analyst 2008, 133, 870–873. (52) Hook, F.; Kasemo, B.; Nylander, T.; Fant, C.; Sott, K.; Elwing, H. Anal. Chem. 2001, 73, 5796–5804. (53) Reimhult, E.; Zach, M.; Hook, F.; Kasemo, B. Langmuir 2006, 22, 3313– 3319. (54) Patel, A. R.; Frank, C. W. Langmuir 2006, 22, 7587–7599.

Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

7995