Immunosensor for Detection of Inhibitory Neurotransmitter Î³

Anal. Chem. 2008, 80, 8576–8582

Immunosensor for Detection of Inhibitory Neurotransmitter γ-Aminobutyric Acid Using Quartz Crystal Microbalance Tingting Wang and Jit Muthuswamy* Harrington Department of Bioengineering, Arizona State University, Tempe, Arizona 85287 A piezoelectric immunosensor for sensing the low molecular weight neurotransmitter γ-aminobutyric acid (GABA), one of two major inhibitory neurotransmitters in the central nervous system, is described. The sensing interface consists of a dextran layer covalently attached to a self-assembled monolayer of thiolamine compound on the surface of gold electrodes of the crystals. The dextran layer is further modified with GABA molecules to act as the biosensing layer. The affinity binding of monoclonal antiGABA antibody on the modified piezoelectric crystals is studied in real time without any additional labels. The equilibrium association constant, Keq for binding between anti-GABA antibody and GABA molecules is 14.5 µg · mL-1. The detection limit for anti-GABA is ∼10 nM. The sensitivity of the sensor at a concentration corresponding to half-maximal response is 13.6 ng/mL · Hz. The functionalized sensor substrate is subsequently used for competitive determination of different concentrations of free GABA (range of 5 µM-50 mM) in PBS-BSA buffer. The detection limit of the immunosensor for sensing GABA with maximum sensitivity is ∼42 µM. γ-Aminobutyric acid (GABA) is the principal inhibitory neurotransmitter in the central nervous system.1 Since 30-50% of all central synapses are GABAergic, GABA plays a critical role in the overall balance between neuronal excitation and inhibition. Alterations in GABAergic activity and receptor function impacts a wide range of physiological functions involved in eating, sleeping, learning, and memory.2 Thus, direct measurement of GABA concentrations in the neuronal extracellular space is clinically significant. Current techniques to detect or measure the extracellular concentrations of GABA are liquid chromatography with electrochemical detection by prepost column derivation,3,4 fluorescence,5,6 electrochromatography, and laser-induced fluorescence detection.7 Since GABA * To whom correspondence should be addressed. Phone: (480) 965 1599. Fax: (480) 727 7624. E-mail: [email protected]. (1) McCormick, D. A. J. Neurophysiol. 1989, 62, 1018–1027. (2) Paredes, R. G.; Agmo, A. Neurosci. Biobehavior R. 1992, 16, 145–170. (3) Murai, S.; Saito, H.; Nagahama, H.; Miyate, H.; Masuda, Y.; Itoh, T. J. Chromatogr.: Biomed. 1989, 497, 363–366. (4) Smith, S.; Sharp, T. J. Chromatogr., B 1994, 652, 228–233. (5) Frye, G. D.; Breese, G. R. J. Pharmacol. Exp. Ther. 1982, 223, 750–756. (6) Allison, L. A.; Mayer, G. S.; Shoup, R. E. Anal. Chem. 1984, 56, 1089– 1096. (7) Rada, P.; Tucci, S.; Teneud, L.; Paez, X.; Alba, G.; Garcia, Y.; Sacchettoni, S.; del Corral, J.; Hernandez, L. J. Chromatogr., B 1999, 735, 1–10.

8576

Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

cannot be detected directly by fluorescence or electrochemical methods, it has to be first modified to electroactive derivatives. Some clever electrochemical approaches to detect small concentrations of GABA have also been reported recently.8-10 Thus, a sensor technique for simple, direct, real-time measurement of GABA concentration is needed. Quartz crystal microbalance (QCM) measurements involving mass-sensitive quartz transducers in thickness shear mode vibration have been used extensively as chemical-sensing devices for the studies of antibody and antigen affinity interactions in solid-fluid interfaces.11 QCM measurements are simple and easy to use, without potentially hazardous labeling materials.12 QCM provides a direct measurement of the antigenantibody affinity reactions at the sensor surface. A change in the crystal mass due to the formation of an immune complex on its surface is assumed to produce a proportional shift in resonant frequency according to Sauerbrey’s equation.13 When operated in a liquid environment, the response of the quartz resonator also depends on the density and viscosity of the liquid.14 Martin et al. derived a modified Butterworth van Dyke equivalent circuit model combining the influences of both mass loading and liquid damping.15 Mass effect induced by viscoelastic changes in boundary interface and the film itself due to antibody binding reactions, also contribute significantly to sensor response.16 A strategy for quantitative measurement of mass effect due to solvent and immunoreaction based on rheological changes of shear acoustic impedance is outlined by Schmitt et al.17 Based on their work, we further derived a quantitative relationship between changes in the real and imaginary components of shear acoustic impedance for a viscoelastic film. (8) Cook, C. J. J. Neurosci. Methods 1998, 82, 145–150. (9) Niwa, O.; Kurita, R.; Horiuchi, T.; Torimitsu, K. Anal. Chem. 1998, 70, 89–93. (10) Sekioka, N.; Kato, D.; Kurita, R.; Hirono, S.; Niwa, O. Sens. Actuators, B: Chem. 2008, 129, 442–449. (11) Janshoff, A.; Galla, H. J.; Steinem, C. Angew. Chem., Int. Ed. 2000, 39, 4004–4032. (12) Bunde, R. L.; Jarvi, E. J.; Rosentreter, J. J. Talanta 1998, 46, 1223–1236. (13) Sauerbrey, G. Z. Phys. 1959, 155, 206–222. (14) Kanazawa, K. K.; Gordon, J. G. Anal. Chem. 1985, 57, 1770–1771. (15) Martin, S. J.; Granstaff, V. E.; Frye, G. C. Anal. Chem. 1991, 63, 2272– 2281. (16) Martin, S. J.; Frye, G. C.; Senturia, S. D. Anal. Chem. 1994, 66, 2201– 2219. (17) Schmitt, N.; Tessier, L.; Watier, H.; Patat, F. Sens. Actuators, B: Chem. 1997, 43, 217–223. 10.1021/ac801463a CCC: $40.75  2008 American Chemical Society Published on Web 10/24/2008

The aim of this study was to develop a direct immunosensor for detecting small-molecule GABA without the need of enzyme or fluorescence labeled tracers. For sensing antigens with low molecular weight such as GABA (mw ) 103.7) using a masssensitive approach, direct binding may not produce sufficient measurable frequency shift.18 A competitive binding assay design is more appropriate for overcoming the limitations of low sensitivity in detecting small molecules. In our design, GABA is immobilized on a gold surface of QCM through a layer of polysaccharide dextran. Hydrogels such as dextran have been widely used on gold19 after the success of Biacore surface plasmon resonance technology.20 A predetermined amount of anti-GABA was mixed with free and immobilized GABA on the sensor surface. Bound and free GABA molecules compete for the binding sites on the antibodies. Molecular interaction of surface-immobilized GABA molecules with anti-GABA antibody on the surface can be detected by the change in resonant frequency that corresponds to the amount of antibody bound to the GABA immobilized sensor surface. The decrease in resonant frequency of the sensor is then inversely related to the free GABA concentration in the sample. EXPERIMENTAL METHODS Chemicals and Materials. Dextran with molecular mass of 40 kDa, 11-amino-1-undecanethiol, alkaline phosphatase, GABA, monoclonal mouse anti-GABA, rabbit antimouse IgG, substrate 2,2- azinobis(3-ethylbenzthiazoline-6-sulfonic acid), and concentrated hydrogen peroxide were all purchased from Sigma (St. Louis, MO). Analytical grade sodium periodate was purchased from BDH Chemicals and sodium borohydride cyanoborohydride from Aldrich without purification. Surface Modification Procedure. GABA molecules are covalently attached to the gold electrode of the QCM sensor in three steps.21 The procedure published by us earlier is summarized here briefly. As the first step, the gold electrode of the quartz crystal is aminated by incubating in a 2 mM solution of NH2(CH2)10SH (11-amino-1-undecanethiol) ethanol solution to form a submonolayer of amines. It is then incubated with oxidized dextran (0.15 g/mL) in PBS buffer for 16 h. Hemiacetal-containing units of oxidized dextran react with amine groups presented on the substrate surface to form a Schiff base linkage. Such linkages are unstable and therefore are reduced by 0.1 M solution of sodium borohydride for 2 h to quench any free aldehyde groups present on the oxidized dextran chain. The final step is covalent binding of GABA with Dextran. GABA has a 4-hydrocarbon chain, an alcohol group (COH) at one end and an amine group (CNH2) at the other end. The amine group of GABA reacts with hemiacetal-containing units of oxidized dextran to form a Schiff base linkage. The 0.1 M sodium metaperiodate solution is then added to the dextran-bound sample to produce aldehyde groups on the surface. Then 1 M GABA solution is added and left at room Zhou, A.; Muthuswamy, J. Sens. Actuators, B: Chem. 2004, 101, 8–19. Rusmini, F.; Zhong, Z.; Feijen, J. Biomacromolecules 2007, 8, 1775–1789. Johnsson, B.; Lofas, S.; Lindquist, G. Anal. Biochem. 1991, 198, 268–277. Wang, T. T.; Ehteshami, G.; Massia, S.; Muthuswamy, J. J. Biomed. Mater. Res. A 2006, 79A, 201–209. (22) Tessier, L.; Patat, F.; Schmitt, N.; Lethiecq, M.; Frangin, Y.; Guilloteau, D. Sens. Actuators, B: Chem. 1994, 19, 698–703. (23) Granstaff, V. E.; Martin, S. J. J. Appl. Phys. 1994, 75, 1319–1329. (24) Mason, W. P. Electromechanical Transducers and Wave Filters; Van Nostrand-Reinhold: New York, 1948. (18) (19) (20) (21)

temperature for at least 48 h, and the reaction is stopped by incubating with 0.1 M solution of sodium borohydride as described above. After the three steps, GABA molecules are covalently attached to the gold substrate. General Experimental Setup. Polished quartz crystals (International Crystal Manufacturing Co. Inc., Oklahoma City, OK), AT-cut (denoting the orientation of plane of the crystal with respect to the crystalline axes) having a center frequency of 10 MHz, a 14-mm diameter, and a 0.168 mm thickness with an active area of 0.196 cm2 are used. The electrodes with diameter of 5 mm are made of evaporated gold (1000 Å) and chromium layers (100 Å). The crystals are placed in the liquid cells by O-ring seals with one face in contact with the liquid solution and the other side facing air. Impedance data are acquired using a HP network analyzer E5100A connected via a 50 Ω coaxial cable to a π-fixture. Measurements are made with frequency centered about the fundamental resonant frequencies of the dry QCM (quartz crystal in air). Impedance measurements are made after network analyzer scanned 201 points centered on the resonant frequency, 200-kHz bandwidth, and 0.5-dBm with an incident power of 1 mW. The parameters of the equivalent electrical circuit elements are estimated internally from the impedance measurements by the analyzer. The resonant frequency f0 and the quality factor of the resonator Q are measured directly by the network analyzer. A visual basic program running in a PC is used to control HP network analyzer and acquire the parameters every second. The shift in resonant frequency, ∆f, and ∆Q-1 (inverse of change in Q-factor) are calculated offline to determine the complex shear acoustic impedance of the loading layers. The QCM sensor after immobilization with GABA is stabilized in buffer solution for 30 min prior to the start of the experiment. The functionalized sensor surface is first blocked with serum proteins (3% BSA) for ∼10 min to eliminate nonspecific binding before the start of the experiment. After blocking, different dilutions of the samples are injected into the flow cell. Each of the steps is monitored via changes in QCM oscillator resonant frequency and Q-factor. The sample volume was 100 µL for all experiments, and the time between two consecutive measurements was ∼10 min. After a strong positive sensor response, the sensor surface is regenerated with a series of wash steps in order to obtain free binding sites again before subsequent measurements. In our studies, the efficiency of HCl buffer (100 mM, pH 2.0) and 10 mM NaOH is confirmed, and the sensor can be regenerated more than 10 times without loss of sensitivity by simple alternate wash of the buffer. Kinetic Calculations. The binding between surface-immobilized GABA and anti-GABA antibody in solution can be interpreted in terms of pseudo-first-order kinetics with three characteristic parameters ka (association rate constant), kd (dissociation rate constant), and equilibrium association constant Keq. The change in resonant frequency, ∆f of the QCM sensor in response to such binding event can be written as a function of time as ∆f ) ∆feq[1 - exp(-kobst)];

kobs ) kaC + kd

(1)

Equation 1 can also be written as

(

ln 1 -

)

∆f(t) ) -kabst ∆feq


(2) 8577

Here, ∆f(t) is related with the surface concentration of the bound antibody at time t, while ∆feq is the change in the resonant frequency of the sensor at equilibrium. Equilibrium or steady state for each concentration of the analyte is defined as the time beyond which resonant frequency does not change more than 2% of its value or stays within the band of frequencies given by, feq ± 0.02feq, where feq is the resonant frequency at steady state or equilibrium. ∆feq (equilibrium or steady-state change in the resonant frequency for each concentration of the analyte) is related to the concentration of free antibody in solution (C) via the langmuir isotherm expression

∆feq )

∆fmax Keq 1+ C

(3)

where ∆fmax represents the maximum response that would be obtained if all available ligand sites are saturated. Nonlinear regression analysis of experimental traces is performed for each concentration of antibody or antibody GABA mixture to reveal estimates of binding parameters. Also, nonlinear regression analysis of ∆feq values obtained at each antibody concentrations is fitted as the simple langmuir isotherm to reveal estimates of Keq and fmax. Curve fitting was done using the Origin software (Originlab, Northampton, MA). Theoretical Basis. The energy stored and dissipated in the loading material is related to the complex shear acoustic impedance of the load ZL = R + jX in Rayleigh, Ra (1 Rayleigh ) 2ng cm-2). The real part of the shear impedance R, representing mechanical power dissipation in the load, and the imaginary part X, corresponding to the kinetic energy stored in the load, are deduced from the responses of the resonator (change in resonant frequency f0 and quality factor Q) as shown in the following equations:22 ∆f X )f0 πZQ

(4)

2R πZQ

(5)

∆Q-1 )

where f0 is the fundamental resonant frequency of the quartz resonance and ZQ is the shear acoustic impedance of the quartz (determined by the material properties of quartz). The complex acoustic impedance of a viscoelastic layer is23 ZL ) √FG tanh(jωd √F ⁄ G)

(6)

G is the complex shear modulus of the coating material; d and F are the thickness and density of the coating, respectively. In the case of a Newtonian liquid, coupling of the liquid to the surface of the resonator causes energy dissipation and thus a decrease in the quality factor. A Newtonian liquid is characterized by very low viscosity and its mechanical impedance is24

ZL ) √jωF1η1 ) (1 + j) 8578

ωF1η1 2

(7)


Here, F1 and η1 are the viscosity and density of the liquid, respectively. The real and imaginary parts of its shear acoustic impedance are equivalent (∆X ) ∆R). The equilibrium resonant frequency shift due to contact to a Newtonian liquid is given by14 ∆f ) f03⁄2(F1η1 ⁄ πFq µq)1⁄2

(8)

where Fq and µq are the density and shear modulus of quartz, respectively. However, viscoelastic materials are characterized by a complex shear modulus G ) G′ + iG″. The chemically sensitive film of the biosensor is usually thin, and the phase shift the acoustic wave undergoes is usually small.25 Under this circumstance, using the third-order series expansion of tanh(x) ≈ x + x3/3 in eq 6 gives26

∆R )

[

∆X ) ωFd +

m3ω3 F2ω3d3G″ ) J 2 3F 3|G|

(9)

]

(10)

m3ω3 F2ω3d3G' ) mω + J″ 2 3F 3|G|

with compliance J ) 1/G )J′ + iJ″, and m = Fd. Clearly, shift in resonant frequency depends on not only the inertia of the deposited mass (first term mω in eq 10) but also the elastic contribution (second term J″m3ω3/3F in eq 10), which would vanish for very thin films. Equations 9 and 10 can be further rewritten as

∆X )

() () J 3F ∆R + J J

1/3

∆R1/3

(11)

Equation 11 describes the relationship between the changes of real and imaginary components of shear acoustic impedance of the coating, which depends only on the mechanical properties of the coating itself. From eq 11, it can be seen that when ∆R is small, as is the case for very thin films

∆X ≈

( ) 3F J

1/3

∆R1/3 ) mω

(12)

When increasing film thickness and subsequently increasing R,

∆X ≈

()

J ∆R J

(13)

X changes linearly with ∆R, and the slope is determined only by the shear modulus of the film itself (∆X/∆R ) J″/J′ ) -G″/ G′). Therefore, by finding out the slope of ∆X versus ∆R, changes in the resonant frequency of the sensor can be compensated for the viscoelastic contribution of the film. (25) Lucklum, R.; Hauptmann, P. Anal. Bioanal. Chem. 2006, 384, 667–682. (26) Johannsmann, D.; Mathauer, K.; Wegner, G.; Knoll, W. Phys. Rev. B 1992, 46, 7808–7815.

Figure 2. ELISA measurements of biofunctionality of immobilized GABA. The labels GABA is for GABA-immobilized surface, control 1 is for positive control with dextran-coated surface to study the nonspecific adsorption on the surface, and control 2 is for negative control with only a bare plastic surface. ANOVA results indicate significant specific adsorption on the GABA-coated surface (p < 0.001).

Figure 1. Response of QCM sensor during surface modification procedure. Changes in resonant frequency of the sensor are plotted as a function of time during (A) amination of the gold electrode and (B) covalent binding of dextran. In (C), the imaginary component of the shear acoustic impedance of the thin film on the surface of the gold electrode is plotted against the real component during amination (curve 1) and coating of dextran (curve 2). The evolution of the shear acoustic impedance through the surface reaction reveals accumulation of mass and also different boundary liquid properties. Red line indicates theoretical change in shear acoustic impedance due to a Newtonian liquid with ∆X ) ∆R. The inset is a magnified version of the main plots within the rectangular box close to the origin indicating linearity in the initial stages of the immobilization process.

RESULTS AND DISCUSSION Sensor Surface Modification. The frequency responses of the sensor during surface modification are shown in Figure 1. During surface amination, a monolayer of 11-amino-1-undecanethiol is self-assembled on the Au electrodes (Figure 1A). The reaction takes place at room temperature and requires 3 h to reach equilibrium. The net decrease in resonant frequency at equilibrium is ∼253 Hz; thus, the mass of the material (per unit area) bound on the surface as determined by Sauerbrey equation is ∼101 ng/ cm2. In the next step, dextran is covalently linked to the amine group of the monolayer present on the sensor surface (Figure 1B), which requires at least 16 h to reach equilibrium. The net decrease in resonant frequency at equilibrium is ∼600 Hz, corresponding to a surface mass (per unit area) of bound dextran 264 ng/cm2. Estimation of sensor sensitivity toward added mass on the surface by the Sauerbrey equation is only good for thin, rigid layers. However, when the phase change across the film is significant, the mass effect induced by the viscoelasticity of biological films has to be taken into consideration. The contributions due to viscoelasticity of the surface coatings are illustrated by the plot of changes in real (∆R) and imaginary (∆X) components of its shear acoustic impedance (Figure 1C). On this graph, a Newtonian liquid effect (pure viscous effect) results in a

straight red line with a slope equal to 1 since ∆X ) ∆R. The two distinct traces represent the formation of a self-assembled monolayer (SAM) layer (trace 1) and a dextran layer (trace 2), respectively. The shapes of the curves serve as a qualitative indicator of the viscoelastic properties of the deposited films. Clearly, the two traces indicate a significant deviation from the rigid mass effect. Trace 1, indicating a SAM layer forming on the surface, has a slope greater than 1 (∼2.7). Trace 2 has a smaller slope (∼1.6) than trace 1, and its ∆X initially increases sharply with increasing ∆R, but subsequently begins to increase at a slower rate. Presumably, for rubbery materials like dextran, the increasing distance of the liquid film interface from the quartz surface leads to increased interfacial displacement thereby producing greater energy dissipation and greater ∆R.The viscosity contribution ∆Xvisc of the SAM layer (trace 1) represents ∼37% of the total ∆X response (21 Ra). The mass contribution ∆Xmass equals 13.2 Ra and corresponds to actual surface mass per unit area of 69.3 ng/cm2. For the covalently bounded dextran, the viscosity contribution of ∆Xvisc (trace 2) equals to 34.4 Ra (∼62% of total ∆X response). Therefore, the actual mass per unit area of dextran is ∼100 ng/cm2, which is considerably smaller than the 264 ng/cm2 predicted by direct application of the Sauerbrey equation. ELISA Test. A preliminary study of the activity of the immobilized GABA is analyzed by ELISA (Figure 2). GABA molecules are immobilized on top of the dextran layer according to the same procedure described earlier. The biofunctionality of the GABA molecules after immobilization procedures is determined by comparing the absorbance values obtained with bare plastic surface (control 2), nonfunctionalized dextran surface (control 1), and GABA-functionalized surface. The absorbance of GABA-immobilized surfaces is significantly larger than those of both control surfaces (p < 0.001) as tested by analysis of variance (ANOVA) statistical test. The result indicates significant specific antibody adsorption on the GABA-functionalized surface, while nonspecific protein adsorption is not statistically significant. This is due to the outstanding biocompatibility of polysaccharide dextran as the stable anchoring layer.27 Dextran in aqueous media would substantially hydrate and form a highly hydrophilic surface, (27) Massia, S. P.; Stark, J.; Letbetter, D. S. Biomaterials 2000, 21, 2253–2261.


8579

Figure 4. Time evolution of (A) imaginary and (B) real components of the shear acoustic impedance during specific surface interaction between immobilized GABA and various dilutions of anti-GABA antibody. Antibody dilutions corresponding to the different traces are (1) 1:1640, (2) 1:820, (3) 1:410, and (4) 1:205.

Figure 3. Changes in the resonant frequency of a functionalized quartz crystal in response to various dilutions of anti-GABA antibody (2 mg/mL) plotted against time. (A). Traces 1-6 correspond to increasing dilutions of antibody (1) 1: 40, (2) 1:100, (3) 1:205, (4) 1:410, (5) 1:820, and (6) 1:1640. (B). Calibration curve for detection of anti-GABA antibody in PBS buffer. The curve is fitted to the equation ∆f ) ∆f0 + ∆fmax/(1 + Keq/C). The estimated constants are, ∆f0 ) 7.6 ( 3.1 Hz, ∆fmax ) 1065.6 ( 180.7 Hz, Keq) 14.5 ( 6.0 µg/mL, and R2 ) 0.98. Maximum sensitivity of sensor to anti-GABA antibody is determined as the slope (red line) at the midpoint Keq and is ∼13.6 ng/mL · Hz. The sensing interface consists of SAM/ dextran/GABA on a gold electrode.

with very little affinity for nonspecific adsorption of proteins at the interface.28 In contact with a watery liquid, a hydrophilic surface results in a better coupling between the sensor surface and the liquid, because the adhesion between the surface and the water is stronger. In a recent study comparing the nonspecific adsorption on surfaces coated with 11-mercaptoundecanoic acid (MUA), Nhydroxysuccinimide (NHS) ester of MUA with undecanethiol, and NHS ester of 16-mercaptohexadecanoicacid (MHA) with a colayer of hexadecanethiol, it was demonstrated that the surface NHS ester of MHA had lower nonspecific adsorption than even CM-dextran surfaces that are commonly used.29 It is possible therefore that the immobilization chemistry reported in this article can be further optimized to improve its nonspecific adsorption properties. In addition, the mass sensing approach proposed in (28) Crepon, B.; Jozefonvicz, J.; Chytry, V.; Rihova, B.; Kopecek, J. Biomaterials 1991, 12, 550–554. (29) Masson, J.-F.; Battaglia, T. M.; Cramer, J.; Beaudoin, S.; Sierks, M.; Booksh, K. S. Anal. Bioanal. Chem. 2006, 386, 1951–1959.

8580


this study provides an advantage over electrochemical approaches in that longer chain SAM layers can be used to further reduce nonspecific adsorption. Longer chain SAM layers have been proven to significantly reduce nonspecific adsorption30 but combined with electrochemical approaches lead to impaired electron transfer across the layers covering the sensing surface. Sensor Response to Different Antibody Concentrations. Piezoelectric immunosensing has been attempted with the sensor surface modified with GABA. The sensor is first washed with PBS and then with 3% BSA solution until the response stabilizes, indicating full blockage of any nonspecific binding sites for protein on the sensor surface. The interaction of the sensor with 3% BSA results in a frequency drop of ∼5 Hz due to nonspecific adsorption of the protein. Addition of specific antibodies however results in a significant frequency drop due to specific binding of antibody to the antigen on the surface. Thus, pretreatment with BSA does block nonspecific adsorption sites on the sensor surface, and the BSA-blocked surface still retains its affinity for sensing the antiGABA antibody. The frequency response or change in the resonant frequency of the resonator, ∆f, during the specific binding of anti-GABA antibody and GABA on the sensor surface is shown in Figure 3A. Increasing antibody concentrations were tested in the measuring chamber, and real-time traces (labeled 1-6, corresponding to dilution range 1:40-1:3200) of the change in the resonant frequencies were recorded. A concentrated antibody solution resulted in more abrupt decrease in frequency than a dilute solution. For example, in the presence of anti-GABA at a dilution of 1:410 (curve 4) and 1:100 (curve 2), the resonant frequency decreased by approximately 150 and 300 Hz, respectively. The calibration curve for different antibody concentrations is determined by fitting the measurement values of the dilution series to a sigmoid curve (eq 3), as shown in Figure 3B. The slope at the midpoint value corresponding to Keq (14.5 ± 6.0 µg/mL), the concentration for half-maximal response, indicates the sensitivity of the sensor, which is ∼13.6 ng/mL · Hz. The lowest detectable anti-GABA concentration is determined as 10% of Keq, which is 1.45 µg/mL and corresponds to a change in resonant frequency (30) Baldrich, E.; Laczka, O.; Campo, F. J. D.; Munoz, F. X. J. Immunol. Methods, in press.

Figure 5. Response of the sensor to five different concentrations of free GABA premixed with 1:200 dilution of anti-GABA antibody in the presence of 3% BSA serum. GABA concentration corresponds to (1) 50 mM, (2) 5 mM, (3) 500 µM, (4) 50 µM, and (5) 5 µM in both (A) and (B). (a) Changes in resonant frequency plotted against time. (b) Changes in imaginary and real parts of the shear acoustic impedance of the QCM sensor in response to five different GABA concentrations. The straight line indicates the response due to a Newtonian fluid where changes in real and imaginary components of shear acoustic impedance are the same (∆X ) ∆R). (c) Standard curve for GABA. A sigmoid curve is fit using the equation, ∆f ) A/[1 + (c/c0)], where the estimated parameters are A ) 240 Hz, c0 ) 80 µM. Curve fitting was done using the Origin software (OriginLab, Northampton, MA).

of ∼100 Hz. Assuming a molecular mass of 150 kDa for the antiGABA antibody, the detection limit of the sensor is estimated as 10 nM anti-GABA. Acoustic Impedance Measurements. The variation in the real, ∆R(t), and imaginary, ∆X(t), components of the acoustic impedance with time in the presence of four different dilutions of GABA antibody is shown in Figure 4. A positive differential response proportional to the antibody concentrations was observed in both ∆X(t) and ∆R(t) as shown in Figure 4A. The results again demonstrate specificity of the immunosensor. The immunocomplex formed on the surface of crystal may lead to a small but significant variation in the boundary layer properties. In order to know whether the variations of ∆X(t) are due to rigid mass accumulation on the sensor surface or to viscosity variations, it is useful to look at the corresponding ∆R(t) signal. As shown in Figure 4B, only small variations in the real component ∆R(t) were observed even at high antibody concentrations (curves 3 and 4 where ∆R ≈ 10) indicating changes in ∆X(t) predominantly due to changes in mass from antibody binding reactions. Detection of GABA in PBS-BSA Solution. The functionalized sensing interface is aimed at detecting free GABA in buffer solution (PBS with 3% BSA). In the experiments using a competitive binding assay technique, a preincubated mixture of various concentrations of GABA and constant antibody dilution 1:200 is added to the measurement chamber with the QCM sensor surface modified with SAM/dextran/GABA. Antibody dilution 1:200 (concentration 10 µg/mL) is chosen since it is close to the Keq

value determined previously to achieve maximum sensitivity and dynamic range. A preincubation time of 10 min is chosen for providing sharp decreases in resonant frequency. Longer incubation time results in a decreased binding of antibody to the sensor surface since more binding sites of antibodies were occupied by free GABA molecules in solution. The changes in resonant frequency and the surface acoustic impedance of the sensor in response to various GABA concentrations are shown in Figure 5. The five numbered traces correspond to GABA concentrations 5 µM, 50 µM, 500 µM, 5 mM, and 50 mM, respectively. In Figure 5A, the individual curves are overlaid on top of each other, demonstrating a decrease in ∆f (change in resonant frequency) with increasing concentration of free GABA as expected for a competitive assay. However, changes in shear acoustic impedance (∆X versus ∆R), as shown in Figure 5B appear to be independent of GABA concentrations. The slight differences in traces are probably due to the variance in surface roughness (liquid trapping), sensor sealing (stress generation, spurious mode), and the properties of the sensor electronics (phase stability, capacitance compensation). Standard curve showing the dependence of resonant frequency changes at equilibrium (∆feq) on the concentration of free GABA is illustrated in Figure 5C. In the same figure, 50% of ∆fmax (parameter c0 of the logistic sigmoidal fit) is observed at 380 µM. The expected limit of detection for GABA should be at 10% below the maximum binding, which is ∼42 µM when the sensor is operated at maximum sensitivity (as indicated by the Analytical Chemistry, Vol. 80, No. 22, November 15, 2008

8581

vertical line in Figure 5C). Lower detection limits can be achieved by increasing the number and concentration of immobilized GABA molecules on the surface of the sensor, which then increases the dynamic range of operation. The concentrations of GABA tested in Figure 5, particularly 5, 50, and 500 µM, are in the range of typical clinical measurements of extracellular GABA concentrations from normal and pathological brain tissues.31,32 The detection limits reported in this work, while clinically useful, are higher than what has been reported earlier using electrochemical approaches.8-10 However, the sensitivity and detection limits of this approach can be enhanced, if necessary, by optimizing parameters such as the thickness of the resonator (and hence the resonant frequency of oscillation), electrode diameters, and surface chemistry as mentioned earlier. For in vivo experiments, samples will have to be obtained using microdialysis with cannulas implanted in appropriate GABAergic regions of the brain. As mentioned earlier, the sample volume on top of the sensor at any given time is 100 µL with continuous inflow and outflow of sample solutions. However, the sensor begins to respond immediately to any addition of the antibody as shown in Figures 3 and 5. The slope and the time constant of the response can be used to obtain an immediate estimate of the unknown concentration of the analyte. Therefore, the sample volume required for sensing can be a few microliters. In addition, the diameter of the current electrodes can be easily scaled down to a fourth (or less) of its current diameter of 5 mm without the use of advanced microfabrication processes. Such a reduction will enable the sensor to be operated at sample volumes in the range of 5 µL or less, in addition to improving the sensitivity (due to reduction in surface area).

CONCLUSIONS The present work describes development of a novel biosensor based on quartz crystal microbalance for the detection of neurotransmitter GABA. A specific sensing interface for the antiGABA antibody was designed, and an appropriate analysis protocol was evaluated. The configuration of the sensing surface involves the attachment of a SAM layer on the gold electrode, followed by covalent attachment of a dextran matrix. Subsequent binding of GABA to the dextran layer yields the specific interface for the anti-GABA antibody. The combination of dextran and GABA in the affinity matrix results in a stable and biocompatible interface with minimal nonspecific protein adsorption. The analysis protocol includes the injection of either antibody or preincubated antibody and GABA mixture to the sensor, followed by rinsing with PBS buffer. Changes in resonant frequency in response to different dilutions of antibody enable us to determine a calibration curve for the antibody. A standard curve for measurement of GABA is also determined, and the expected detection limit for GABA is ∼42 µM when the sensor is operated at maximum sensitivity. The results presented here demonstrate that direct piezoelectric immunosensors are a novel and viable approach for sensing low molecular weight neurotransmitter GABA. Also, the technology could yield new information about the molecular interactions at the interface (such as neurotransmitter-receptor binding interactions), which cannot be observed using more conventional bulk measurement techniques.

(31) Kuzniecky, R.; Ho, S.; Pan, J.; Martin, R.; Gilliam, F.; Faught, E.; Hetherington, H. Neurology 2002, 58, 368–372. (32) Kuzniecky, R.; Pan, J.; Burns, A.; Devinsky, O.; Hetherington, H. Epilepsy Behavior 2008, 12, 242–244.

Received for review July 14, 2008. Accepted August 28, 2008.

8582


ACKNOWLEDGMENT We would like to thank Jonathan Rogul for his assistance in reviewing the manuscript.

AC801463A

Immunosensor for Detection of Inhibitory Neurotransmitter Î³

Recommend Documents