Anal. Chem. 1997, 69, 3651 -3657
Accelerated Articles
Capacitance Measurements of Antibody-Antigen Interactions in a Flow System chlirtine Berggml and Gills Joh#rUon* Analytical Chemistry. Chemical Center, Univetsily of Lund, P.O. Box 124, S-22100Lund, Sweden
Capacitive immunosenwmwere made by coupling monoclonal antibodies to thioctic acid, which had self-assembled on a gold electrode. Surface areas that were not covered were plugged with l-dodecanethiol to make the layer dense and insulating. Cyclic voltammetry showed that the hexacyanoferrate redox reactions were biwked by this procedure. The capacitance of the electrode was evaluated from the current transients obtained when a potentioatatic step was applied. The immunosensor was placed in a flow system, and a capacitance decrease could be observed after injection of an unlabeled antigen. Itwas linear over almost three decades when plotted vs the logarithm of the antigen concentmtion. Human chorionic gonadotropin hormone could be determined in the range 1 pg/mL- 1 ng/mL, with a detection limit of 0.5 pg/mL (15 10-l5M). A similar response was obtained with immobilized F(ab):! fragments. No cross-reactivitywas ob.served with the thyrotropic hormone, which has one chain in common with gonadotropin. Monoclonal antibodies toward interleukin-2 immobilized on the immunosensor gave also a response Over 1 pg/mL-1 ng/mL, with a detection limit of 1 W m L An immunosensorwith monoclonal antibodies toward human albumin gave a calibration curve with lower slope than the other proteins but still with a detection limit of 1pg/mL.
The development of immunoassays during the last three decades has revolutionized determination of drugs and hormones in clinical and pharmaceutical chemism as well as contaminants in the environmental area. Almost all immunomethods require labels attached to either the antibody or the antigen, but some attempts have been made to detect the binding itself by potentiometric.’ piezoelectric.” or optical (1) Taylor. K. I:.: Marenchic, 1. G.: Rmcer. R H. Anal. Ckim. Ada 1991, 249, 67.- 70. (2) Kurderer. J. E.; Ihatiaans. 6. J. Anal. Ckrm. 19I13.55. B33-23:%.
S0003-2700(97)00203-5 CCC: $10.00 10 1997 American Chemical Society
Lately, capacitance measurements have also been investigated as a highly sensitive approach.+IR When an immunosensor based on a capacitive transducer is constructed, the immobilization of the recognition element is of vital importance for the ability to detect the antibody-antigen interaction. The grafted recognition layer should be electrically insulating to prevent interferences kom redox couples in the electrolyte solution and high faradaic background currents. On the other hand, it should be as thin as possible in order to achieve high sensitivity. Different immobilization procedures onto different substrates have been reported in the literature. Antibodies have been grafted covalently via silanization to a Si02 semiconductor substrate,”--14 transformed to a modified Sios semiconductor substrate via a Langmuir-Blodgett film15or coupled via silanization to a TaOz e l e ~ t r o d e . ’ Antigen ~ , ~ ~ as a long-chain, dialkyl disulfide with a (3) LafAs. S. Pun Appl. Chem. 1995, 67.829-834. (4) L4fAs. S.: Johnsson. B.J. Chcn. Soe, Chen. Conrmvn. 1990.1526-1528. (5) Johnsson. B.: LafPts. S.: Lindquist. G. Anal. Biwhem. 1991. 198, 268277. (6) Chaiken. I.; Rose, S.; Karlsson, R. Anal. Biochem. 1992, 201. 197-210. (7) H u h . W.:Bamer. R.: Fattinger. C.; Hiibscher, J.; Koller. H.: Muller. F.: Schlatter, D.: lukosz. W. Seas. Actuators B 1992.6. 122-126. (8) Nellen, P. M.: Lukosz. b‘.Biosens. Bioelectron. 1991, 6, 517-525. 63(9) Souteyrand, E.;Martin, J. R Martelet, C. &as. Actvaton B 1994,B. 69. (10) Saby. C.; Jaffrezic-Renault.N.; Martelet, C.;Colin, B.;Charles, M.H.;Delair. T.: Mandrand. B. Sew. Actuators, B 1993,lo”.458-462. (11) Billard. V.; Martelet. C.: Binder. P.; Therasse. J. Anal. Chim. Acta 1991. 249. 367-372. (12) Gardies. F.; Martelet. C.; Colin, B.; Mandrand. B. Sew. Achrotots 1989. 17,461-464. (13) Bataillard, P.;Gardies. F.: JahzicRenault. N.; Martelet, C.: Colin, B.; Mandrand. B. Anal. Chem. 1988.60.2374-2379. (14) Maupas. 11.: %by. C.; Martelet. C.: Jaffresic-Kenault, N.:Soldatkin. A. P.; Charles. M.-H.; De1air.T.; Mandrand. B. J. Elednxanal. Chem. 1996,406, 5348. (15) Barraud. A: Perrot. H.; Billard. V.; Martelet. C.: Themme. J. Bioscnr. Bioelectron. 1993, 8. 39-48. (16) Gebbert. A.: Alvarez-lcaza. M.; Peters. H.;JMer, V.; Bilitrwski. [I.; Sehmid, K. D. J. Biotechnol. 1994. 32.213-220. (17) Gebbert. A; Nvareokm, M.; StWklcin. W.;*hmid. R. D. A n d . clkn. 1992.64, !397-10(13. (18) Taira. H.; Nakano. K: Maeda. M.:Takagi, M. Anal. sri. 1993. 9. 199-
a%.
Analyticai Chemistry, Vd. 6Q,No. 18. SepMmber 15, 1997 3661
I H
Gold aubstnte
Flgulv 1. Antibodies immobilized on the gold sensor surface via an amide bond to the self-assembled thiocti acid. After the immobilization of the antibody a long hydrocarbon thiol, l-dodecanthiol, was introduced into the system to block any uncovered spots of the
sensor surface. terminal dinitrophenyl group has been synthesized and immobilized via chemisorptionto form a self-assembled monolayer (SAM) on a gold electrode.I8 A peptide bound to an alkanethiol was also immobilized as a self-assembled layer on gold.Ig Self-assembled monolayers of thiols, sumdes, and disulfides on gold electrodes have been widely studied and longchain alkanethiols are known to form insulating well-organkd structures on gold sdmates?0 The binding formed between the sulfur atom and gold is strong, and the formed self-assembled monolayers are stable in air, water, and organic solvents at room temperature.21 It has been suggested that microcontact printingz and ph~tolithography~~ can be used to pattern surfaces with functionalized self-assembled monolayers for biosensor production with low cost for a diversity of applications. It would therefore appear appealing to develop an immunosensor with antibodies immobilized onto a self-assembled functionalized monolayer. In the present study we immobilized first monoclonal antibodies raised against the hormone human chorionic gonadotropin (HCG) on gold substrates. Anti-HCG antibodies have previously been coupled by thiol chemistry onto gold?' and this was used by us as a model system. Antibodies toward interleukin-2 (IL2) and human serum albumin (HSA) were also used as model substances. The antibody layers and their interactions with antigens in solution were investigated with capacitance measurements and with cyclic voltammetry. EXPERIMENTAL S E C " Antibody immobilization to the working electrode.The anthdycovered electrode is schematically shown in Figure 1. me electrode was a gold rod (99.99%Aldrich. 3 mm in diameter) cut up into thin sections threaded to stainless steel holders. Prior to immobilization, the gold rod was polished with alumina slurries down to 0.04-0.05 pm (Struers, Copenhagen). M e r being mounted into the Teflon holder, the electrode was plasma cleaned for 15 min (Hanick Sci. Co., New York, PDC-3XG) and immediately placed in a solution of 2% (w/w) ~/t-thiocticacid in (19)
Rickert. J.: Wolfgang. G.; Beck, W.: Jung. C.: Heiduschka. P. Bioseas.
Bioelectmr. 1996,II. 757-768. (20) Porter. M. D.; Bright, T. B.; Allara. D. L; Chidsey. C. E. D. J. Am. Ckcn. Sot. 1987,109.3559-3568. (21) Bain. C. D.:Troughton. E. B.:Tao.Y.-T.: Evall. J.; Whitesides. G. M.; Nuzzo. R. G.J. Am. Chem. Soc. 1989,111.321-335. (22) Mrksich. M.: Whitesides, G.M. Tmndc Bioteckrol. 1995, 13. 228-235. (23) 13hatia. S. IC: Hiekman. J. J.: tigler. F. S. J. Am. Chem. Soc. 1992, 114. 4432-440. (24) 1)usii. C.; Meyerhoff. M. E. AIIU~.Ckcm. 1994.66, 1369-1377.
3652 Analytical Chemistry, Vd. 69, No. 18, seprember 15, 1997
absolute ethanol. The electrode was taken from the solution after 24 h, thoroughly rinsed in absolute ethanol, and allowed to dry. Thereafter the electrode was put into a solution of 1%(w/w) 1-13. ( d i m e t h y l a m i n o ) p r o W l J - ~ t h y ~hydrochloride ~de in dried acetonitrile for 5 h. A 5 p L sample (-1 mg/mI.) of antibody solution was placed on the electrode surface, and the coupling procedure was performed at 4 "C for 24 h. The coupling procedure followed essentially that described by Duan and Meyerhoff." A long thiol, ldodecanethiol,was used to block any unblocked spots on the electrode surface. At least half of the electrodes did not produce any changes in the capacity when antigens were injected. The reasons for this inactivity are unknown, and the yield did not seem to depend on variations in the pretreatment. Chemicals. HSA, monoclonal HSA antibodies, human thyrotropic hormone, and HCG and its monoclonal antibodies were all obtained from Sigma. Monoclonal IL2 antibodies were taken from the kit, D 2050, from R&D Systems Inc., Minneapolis, MN. The cytokine IL2 was also from R&D Systems. All other reagents were of analytical grade. Capacitance Measumments. The capacitance changes were evaluated from the transient current response obtained when a potentiostatic step was applied to the electrode. An alternative measurement principle relies on the evaluation of the currents at a number of sinusoidal wave frequencies, usually called impedance spectroscopy. The two methods were compared by using the same potentiostat and electrodeand found to give almost the same results in terms of equivalent capacitances and resistances. A detailed description of the instrumentation will be published separately.'" The potentiostaticpulse method is faster and more convenient and is therefore used throughout in this paper. The measuring setup consisted of a three-electrode system, with an extra reference electrode, connected to a fast potentiostat. The potentiostat was connected to a computer (486,33 MHz) vis. a Keithley 575 measurement and control system, containing 16 bit A/D and D/A converters. The Keithley system was powered from the computer through a gavanically isolated power line in the box. The potentiostat was powered from the Keithley in order to isolate the analog parts from the noisy digital circuits. The sampling frequency of 50 k€kwas determined by an internal clock in the Keithley box. The current values were taken as the mean of 10 measurements. The rest potential was 0 mV vs an Ag/AgCI reference electrode throughout. A potential step of 50 mV was applied, and the current transient that followed was sampled. An identical current transient but of opposite direction was obtained when the potential was stepped back to the rest value. Taking the logarithm of the current gives an almost linear cutve from which R, and C1 can be calculated (see F ~ i r e3a). The first 10 current values were used for the calculation, and a correlation coefficient better than 0.99 was obtained. The measuring cell, a flow cell with a volume of 2 mL, is shown schematically in Figure 2. The working electrode (a) was a gold rod (diameter 3 mm) mounted in a Teflon holder, the auxiliary electrode (b) a disk-shaped platinum foil with a hole in the center, and the reference electrode (c) a platinum wire (diameter 1mm). The platinum wire was used as a reference electrode because it can be placed closer to the working electrode than a Luggin capillary of glass without causing any shielding. The platinum reference electrode, though, has no defined potential 90 its (25) Herggren. C.; Hjarnimwn. B.: Johanswn. G.. nianuscript in Vrt'parPtinn.
c2
d a.
Inlet -+ C
-
-+ outlet
' b
--.I I_-
Flgur, a Measuring flow cell: (a) working electrode, (b) auxiliary electrode, (c) platinum reference ektrode, and (d) AgAgCl reference electrode.
.-
-,--c
Md
odddar
4 potential was compared to a commercial A g W / refmnce electrode (d) just before the potentiostatic pulse was applied. The carrier solution. 10 mM citrate buffer, pH 7.4 was pumped with a flow rate of -0.5 Wm n i t h m g h the flow cell. An injector with a loop of 250 p L was connected to the flow system. Cyclicvoltamm*. cyclic VOltammO~Swere recorded in a three-elechde system in a batch cell. The working electrode was the unmodiied or modified gold rod (diameter 3 mm) in a Teflon holder, the auxiliary electrode was a platinum foil, and the reference electrode was a saturated calomel electrode (%E). A 5 mM I&@e(CN)s) solution was used for the measurements. The instrumentation used for cyclic voltammetry was a Princeton Applied 273 A potentiostat, controlled by a computer.
RESULT SAND^ PrOpertiesoftbeG~AntibodyLayer.Thecapacitance is determined from the current response obtained when a potentiostatic step of 50 mV is applied on theelectrode. Evaluation is made by assuming that the response follows a simpk RC modeL With this model, the current transient after a potential step will approximately follow eq 1. i(t) = w/R, exp(-f/R,C,)
(1)
where i(f) is the current in the circuit as a function of time, Y is the applied pulse potential, R, is the dynamic resistance of the self-assembled layer, CIis the capacitance measured between the gold electrode and the solution, and f is the time elapsed after the potentiostaticpulse was applied. The correspondinge q u h lent cirmit is shown in F i i 3a, and the current vs time response is shown in figure 4. figure 3c shows how Ct is built up from a series of capacitors representingessentiallythe thioctic, antibody and doublelayer capacitances, respectively. The response is initially fairly linear in a log i vs time plot. A more accurate description can usually be obtained Over a larger range by using a pseudo Randles equivalent circuit, Fi3b.z R, is the dynamic resistance which dominates immediately the pulse is applied. The charge transfer resistance, R.1,dominates when the &tor, Cz, is fully charged. A hgh value of Rd indicates that the electrode blocks almost all h d a i c reactions, giving a steady-state residual current orders of magnitude lower than for a bare metalekch.ode. 'Ihe Capscitive sensor described in this paper relies on the measurement of
Fl#uro 3, Models for evaluation of the current vs time data: (a) a simple RC circuit, where f?, is the dynamic resistance of the organic layer, Ct is the capacitance measured between the gold etactrode and the electrolyte solution, and (b) a pseudo Randlescircuit, where & is the dynamic resistanceof the organic layer, Cr is the cepadtence measured between the gold electrode and the electrolyte solution, C,is a pseudocapacitance.and & is the charge transfer resistance of the immobilized recognitionlayer; (c) Describeshow the capacitors for the thioctic, antibody, and double layers together constitute the capacitance G.
0.0
0.2
0.4
0.8
0.8
1.0
1.2
1.4
1.6
1.8
nmim Flmm4. Current response obtained after applyinga potentiostatic step of 50 mV to the thioctic acid-modified electrode.
changes in CI.It is best evaluated h m the immediate m n t response after applying the pulse. A gold surface covered with a longchain akanethiol layer blocks almost all faradaic currents and is highly insulating with an equivalent transfer and dynamic resistance of -2OOO and 69 S2 cm2,respectively, for a surface covered with butanethim1.n A layer of thioctic acid was much less insulating with an equivalent transfer and dynamic resistanceof 470 and 40 R an2,respectively. The permeability of ions through the layer is so hwh that a redox couple can penetrate it, giving almost the samecurrents in a cyclic voltammogram as on a bare gold electrode; see Figure 5,curve a. This is consistent with the results obtained by Chemg and Brajter-TothZ8a Immobilization of an antibody reduces the penetration of the redox couple ( F i i 5,c u m b). Insultion is further improved when the electrode is treated with ldodecmethiol, as can be seen from the absence of redox peaks for such (n)Swktkm. A; Skw,M.: Johmwn. C. Ekchwad. 1902,4,921-928. ('28) Chew, Q.;Braliter-Toth. A A d . (Mu. 1905,67.2767-2775. (29)C h g , Q.:Bnjl~rlbth.A A d . clkm. 1992,64,1998-2000.
Am&thI CMstty, Vd. 6Q, No. 18,
15, lm7 3863
"[
-20'
'
-20
-100
0
100
200
300
400
500
600
10'
rd
105
104
106
Id
ElmV Figun 5. Cyclic voltamogramms recorded in a 5 mM K3[Fe(CN)e] solution. The Scan rate was 10 mV/s. All potentialsare given vs SCE. The voltamogramms shows (a) gold modified with dL-thioctic acid, (b) as in (a) but with anti-HCG antibody added, and (c) as in (b) but with 1-dodecanethiol added.
Hormone I 1Ol5 M Fig6. Capacitance change vs the logarithm of the hormone concentration, measured in a flow system with a carrier solution of 10 mM cirrate buffer, pH 7.4. The capacitance change was taken after 15 min, when the capacitance had reached a stable value. The upper curve shows the specific response for HCG and the lower curve the nonspecific response for thyrotropic hormone.
an electrode Figure 5. curve c). The shape of the current response with immobilized antibodies on top of thioctic acid became much more complex and could not be fitted to a fourparameter model. The most disturbins discrepancy occurred immediately after the pulse was applied, and the simple RC model is therefore much better in thiscase than the Randles circuit. The Rsvalue obtained with the simple RC model was 58 !J cmz, Le., not much different from that of thioctic acid alone. The capacitance C1 decreases from 24 600 to 7100 nF for a typical electrode when thioctic acid is bound to the surface. It will decrease further to -1400 nF when an antibody is bound to the thioctic acid and treated with alkanethiol. This can be compared with a C1 value of 2480 nF for butanethiol! at an ionic strength of 10 mM. In spite of the huge size of the antibody in comparison with the butane chain, the capacitance is not that much smaller. Each Iayer can be considered to contribute to the total capacitance. Since the inverse Cl is the sum of the inverse of each layer's capacitance, a small capacitance in one layer will result in a low total CI.This will reduce the sensitivity of the capacitive biosensor because the small capacitor is in series with the layer of the antibody antigen reaction. The size of the circuit elements is also important from an instrumental point of view because the product ClR,is the time constant. A large time constant makes the current response decrease slower and there will therefore be time to collect many more current readings in the computer. The chemical reason for the high capacitance of the electrode described here is that the thioctic acid layer is thin and has a very hgh polarity compared to the well-ordered, almost crystalline polyethylene layer for alkanethiols. The antibody is also much more polar than a polyethyIene layer. Antigen Detection. When an antigen binds to the antibody immobilized on the electrode, there will be an additional layer decreasing the total CIfurther. The b i n d w between the antigen and antibody is therefore detected directly. No label is necessary for the antigen. The physical basis for the response is thought to arise from displacement of the polar water further out from the electrodesurface, replacing it with a much less polar molecule. The hormone HCG was used as one model substance. HCG is a glycoprotein with a molecular weight of 30 OOO. The hormone consists of an a and a /3 chain. The a chain is the same as in the
thyrotropichormone, but the /3 chains differ in the two hormones. The monoclonal antibody immobilized on the electrode was directed toward the @ chain specific for HCG. Thyrotropic hormone and HCG are known to have a cross-reactivity uf less than 0.05%."' The thyrotropichormone was used as a control for testing the selectivity of the immunosensor. Samples with HCG concentrations as low as 30 x lo-]$M (1 pg/mL) were injected into the flow system. The capacitancewas continuously measured and found to decrease after an injection until it reached a stable value, which took -15 min with a flow rate of 0.5mWmin. The change in capacitancevs the logarithm of the concentration was found to give a linear relationship up to M (0.3ng/mL) and to reach a saturating a concentration of value at lo-'" M; see figure 6, upper curve. The detection limit M (0.5pg/mL) hormone. It was calculated from was -15 x a comparison between the signal and the irreproducibility of measurements on the antibody surface alone. The irreproducibility corresponds to 15 nF cnr2. As usual in flow injection analysis, the sensitivity and detection limit can be changed by changing the injection volume. A larger sample size will thus decrease the detection limit in proportion. No cross-reactivity whatsoever was observed on the capacitive biosensor when the control antigen, thyrotropic hormone, was injected into the flow system: see F i r e 6, lower curve, This suggests that the observed capacitance change is specific and not caused by an unspecific adsorption of protein to the sensor surface. Injection of a semm sample without added HCG produced a 13% increase in the capacitance when the sample entered the cell. The signal returned to the previous value when buffer filled the cell again. The increase in capacity is due to the increased ionic strength of the solution. Ihe experiment thus shows that semm as such does not give rise to any permanent change. c a p a & m e ~ f o r A n ~ ~ F m g m e n tToincrew s. the sensitivityof the signal,antibodies against HCG were digested with ficin to F(ab'), fragments. The idea is to remove an inactive part of the antibody and to move the binding sites closer to the electrode surface. The fragments wen! immobilized to the electrode surface in the same way as described above. The analytical properties were similar to those obtained with electrodes
3654 Analytic81Chemisw, Vd, SQ,No. 18, Sept8mbOr 15, 1987
(30)Sigma Chemical Co.. uwluc? specitication. C-7659.
-210
i"
t
0 15n\lkm
-190
1
-250
-170
t
9
-150
-50
-100
-30 -101
Id
IO'
103
10'
l(r
HCG I 1015 M I.raun7. Capacitance change vs the logarithm of antlgen concentration for a surface with immobilized F(ab')2 fragments from antiHCG antibodies. Conditions as in Figure 6.
HCG
"
10'
ld
10'
l(r
105
ld
Antigen conmntntionI 10"' M
Fig-
8. Influence of molecular weight on capacitance change studied for electrodesspecific for HCG, 11-2, and HSA. The capacitance was measured in a flow system with a carrier solution of 10 mM citrate buffer. pH 7.4. Conditions as in Figure 6.
covered with the native antibody, as shown in Fire 7. The capacitance, CI,of the F(ab'h electrode was 4500 nFcm+ compared to 1400nFcm-2for an electrode with a native antibody. The resistivities were -63 S2 c m 2 in both casts. The slopes of the calibration curves were about the same in both cases, and an increased sensitivity was not obtained. The increased capacitance will improve the signal-tcmoiseratio somewhat. Capacitance Changes forhtigens with DifferentMolecular Size. To study the influence of the size of the antigen, electrodes specific to antigens with different molecular weights were prepared. The three different antigensused were IL2 (M, 15 700). HCC (Mw30 OOO) and HSA (Mw69 OOO). The results (see Figure 8) indicate that the capacitance change was about twice as large for HCG as for IL2. This can be explained by the larger molecular weight of HCG. The IL2 a n t i i y was taken from a commercial sandwich ELISA kit for determination of IL2 after incubation in microliter plates with a stated sensitivity of 6 pg/mL in medium and 10 pg/ mL in serum. The detection limit for the immunosensor is -1 pg/mL. Serum samples from apparently healthy donors were all below 31 pg/mbR1i.e., the commercial kit could not reliably measure IL2 levels in healthy individuals. The response for HSA was lower than for IL2, which suggests that more factors than molecular size has to be taken into account. ~~
(31)
R & D Systems, Inc.. Quantikine. IL2 manual.
lo'
,
,
,
10'
104
10s
HSAI 1015 M
Fbum 0. Influence of flow rate on capacitance change for HSAspecific electrodes. Conditionsas in Figure 6.
Such factors can be the structure of the antigen, that is, if it has a compact or a loose configuration, charges of the antigen, and the affinity constant for the antibody-antigen complex. One possibility is that albumin is penetrated by the aqueous phase, resulting in an increased polarity of the antigen layer. Another possibility is that the antigen binds in such a way that aqueous solution can penetrate between molecules to some extent. There might be steric hindrance for two large HSA molecules to bind to an antibody with about the same molecular weight. Intluence of Flour Rate on Capacitaoce Change. The capacitance changes obtained for different flow rates were studied for the HSA system, and the results are shown in Figure 9. The capacitance change was found to increase from a flow rate of 0.6 mWmin to 0.15 mL/min. A longer residence time in the cell will allow more HSA molecules to be transported to the biosensor surface by diffusion and hydrodynamicmovements in the solution. An increased sensitivity with decreasing flow rate is therefore generally expected. A closer look at the curve shapes for HSA at 0.3 and 0.6 mW min shows that they differ from those of the other antigens and from that of HSA at 0.15 mL/min. The lower flow rate gives HSA more time to interact with the antibody and to rearrange itself on the biosensor surface. The sensitivity per molecule seems also to increase when the concentration decreases. The original reason for the large volume of the flow cell was that the current response immediately after the pulse rise edge became distorted when the auxiliary electrode was moved closer to the biosensor. It was also evident that the Luggin capillary to the reference electrode reduced the sharpness of the initial current response. The cell was therefore provided with a fourth electrode. a thin partially insulated platinum wire, inserted through a hole in the middle of the a m i electrode. It wrves as the reference electrode during the potentiostatic pulses, but it will of course drift and it has not the properties expected from a reference. The computer will therefore make potentiometric comparisons be tween the platinum potential and a commercial reference in a side arm just before a pulse is applied. Ihe difference will be calculated and used to select a potential for the platinum wire, which makes the potentiostat behave as it had an Ag/@I reference. The equivalent circuit for the platinum wire will be simpler at high frequencies, and it will therefore increase the sharpness of the current response curve. The four-electrde arrangement made it possible to make a new flow cell with a dead volume of only 10pL The sensitivity Analytical Chemistry. Vd. 69, No. 18, -r
15, 1997 3656
A
2
3
4
5
6
7
PH Figwm 10. Capacitance response as a function of pH for Pmernormalcaptoethand, thioctic acid, and 16-1nercapto-l-hexadeca~, ized to pH 7.0. The capacitance was measuredin a buffer Now of 10 mM phosphate buffer, pH 7, before and after exposure to other pHs. of the biosensor was essentially the same in the two flow cells. The diffusional distances are much shorter in the new cell but the mean residence time will, at the same flow rate, be much shorter, resulting in binding of approximately the same proportion of the sample antigen molecules. pH stability studies a€ Ditkrat SAMs on Gdd. The behavior of self-assembled monolayem from Zmercaptoethanol, thio& acid, and 16mer~apt~l-hexadecanol on gold was studied at different pHs. The capacitance was measured in buffer, pH 7, before and after exposure to other pHs. As can be seen in F i r e 10, the self-assembled monolayers are all influenced by the treatment in acidic solutions. Once exposed to a lower pH, the capacitance will permanently increase to a higher value. The increase presumably reflects formation of pinholes in the layers, giving the aqueous solution access to the metal surface, thereby short-circuiting the lowcapacitance layers. It is furthermore evident that the behavior of the dfierent self-assembled compounds is similar, indicating a common cause. Research in other field^^^*^ points to pit corrosion of the gold surface itself with a start at defect sites. It has been observed in this laboratory that use of high-purity gold rods is very important for the quality of the capacitive biosensors. Thiols will not bnd well to heteroatoms l i e copper, and the surface will therefore be less blocking. The surface roughness, as seen in atomic force microscope pictures, is probably of little importance. The quality of the self-assembled monolayers is dependent on the underlying gold surface. Steinberg et al. have found that if the gold is thermally annealed at 400 "C the surface appears to improve in cleanliness and smoothness.33 Furthennore, they always observed pinholes for nonthemdy treated gold, but for annealed gold, pinholes disappeared after repeated treatments in a blocking thiol solution. It is suggested that monolayer surfaces without pinholes can resist acid treatment, as the uncoating is thought to start at these sites. This would lead to more stable and resistant sensors suitable as regenerable immunosurfaces. Annealing the gold rods in a flamP has not given any improvements so far. Acid wash with 100 mM glycine hydrochloride, pH 2.8, normally used to break antibody-antigen interactions, always (32) Heller, A, persona! communic8tion. 1M. (33) Steinberg, S.; Tor. Y.; shuwr, 11,Rubinskin. I. 7kin Films 1995, W, 183-205.
3656 Analytical Chemisrty, Vd. 69, No. 18, Sqfmb@r 15, 1997
desb-oyed the electricali n w a propertiecs of the ensor. W this treatment, the capacitance inmaeed with timc md cy& voltammognmsshowedoxidationreductionpealuintheprecmur of the redox couple &(Fe(CN)& Testswith 3.5 M MgClz, !N% ethylene glycol, and 8 M urea were also u n - 4 and destroyed the electrodes. A new electrode is therefore used for each measurement. A new cell for disposable plates of sputtered gold on silicon seems to be promising. comparieon WW M e r Reported crprrcitive Immunommom. The detection limits repotted here are at least 2 ordm of magnitude lower than those reported previously for capacitive immunosensors, and a comparison is therefore necessary in order to understand the factors that are of importance. Capacitive immunosensors with covalent bonds to an electrode have been reported using either surfaces with oxygen functionalities~Q-*~ or self-assembly by coupling sulfur to gold.'* Other immobilization methods to oxide surfaces are adsorption11 or through transfer of Langmuir-Blodgett films.15The space between interd@tated electrodes has also been filled with polymers holding antibodies.' Oxide layers are not well ordered, and less perfect coverage can therefore be expected in the immobilized layers. Any part of the surface that allows the aqueous solution to penetrate below the plane where the immunoreaction takes place will act like a short-circuiting element. The capacitance will therefore increase due to the higher dielectric constant of the penetrating aqueous solution. In an equivalent circuit model for the layer, a single capacitor will be replaced by a resistor in parallel with a capacitor that is larger than for a dense layer. Binding still decreases the capacitance but the relative decrease is less, due to the increased size of the capacitor. Tantalum oxide surfaces should have some advantages over silicon oxides, but they have also a disadvantage as discussed by Gebbert et aI.,"Q7namely, that a layer thick enough to be stable has a very low capacitance. This capacitance appears as a series capacitor with a very low value, and it will dominate in the expression leading to the total capacitance. Changes in the capacitance of the antibody layer will therefore have less influence on the measured value. A comparison with an interdigitated array' is not possible because no capacitance changes have been obsemed when the antigen binds; the same holds true for the Langmuir-Blodgett film e~periments.~~ Polymeric layers containing antibodies on electrodes10have only recently produced a signal with a detection limit of 10 ng/mL-l.l4 Self-assembled thiols with antigenic terminating groups were reported to have coverages of only 14, 19, or 31%for different electrodes.18 The lowest measured value was at an antibody concentration of 10 ng/mL, which can be compared to 1pg/mL of antigen reported in this paper (See Fwre 6). The saturation seems to occur at similar concentrations in the two cases if the larger bulk of the antibody compared to the antigen is taken into account This comparison thus supports the arguments given above that a dense layer is of great importance for a hgh sensitivity. A closer comparison of the curves reported with antibodies on oxide surfaces and our own presented in this paper similarly shows that the binding saturates at similar concentrations. The measuring ranges of the oxide immunosensom were fairly short in these cases,indicating that the high senddvity of our electmdes is obtained because their ranges have been pudrcd downward. In F v 6, saturation occurs at 2OOO timesthe stated detection
limit, corresponding to an antigen binding to only 0.05%of the antibodies at the detection limit. Self-assembled thioctic acid has been shown to have much higher dielectric constant than that expected from its hydrocarbon layer thickness.?Ra It has also been shown that redox pairs like Fe(CN)& and Ru(NH3)s3+penetrate the layers, a result that is reproduced in F i r e 5. Immobilization of antibodies on top of the thioctic acid decreases the penetration substantially and decreases the capacitance. Treatments with dodecanethiol after the antibody immobiition demases the penebation further. The capacitance decreases also during the dodecanethiol treatment,
although less than during the antibody layer i m m o b i i . The alkanethiol seems to plug some h o h in the thiactic acid layer and will thus contribute to make the surface compact urd nonconducting. AcKwormEDQMare
This research was supported by NUlEK, Sweden. Received for review February 19, 1997. Accepted July 8, 1997.e AC970203E Abstract published in Adwin ACS AMmcts, August 1.1997.
