A

etaag©

K

1

su[bhabuipaTt

psnana

0d

Electrochimica Acta 53 (2007) 803–810

Development of a membrane-based electrochemical immunosensor

Guiwan Koh, Shuchi Agarwal, Pui-Sze Cheow, Chee-Seng Toh ∗Department of Chemistry, Faculty of Science, National University of Singapore, Singapore 117543, Singapore

Received 31 May 2007; received in revised form 23 July 2007; accepted 25 July 2007Available online 31 July 2007

bstract

A membrane-based electrochemical immunosensor sensitive towards proteins is described, based on nanoporous alumina-modified platinum wirelectrodes. The sensing mechanism depends on the blocking of pore channels when the protein antigen molecules bind to antibody molecules attached

o the channel walls, impeding the diffusion of redox probe, ferrocenemethanol, towards the sensing platinum wire overlaid by the nanoporouslumina film. The antibody and antigen used in this work were anti-glucose oxidase and glucose oxidase, respectively. The immunosensor showedlow limit of detection of 100 ng L−1 antigen concentration and was selective towards glucose oxidase protein in the presence of another protein,lucothione S-transferase.

2007 Elsevier Ltd. All rights reserved.

globu

mattwacfcolhuauosTqr

eywords: Alumina; Immunosensor; Glucose oxidase; Nanostructure; Immuno

. Introduction

Development of immunosensors with capability for rapid,ensitive and selective detection of infectious diseases, contin-es to be an important subject for research and development1–4]. It is equally desirable to have fast response sensing capa-ility towards other analytes, including proteins, DNAs andaptens in environmental studies, pharmaceutical applicationsnd biomedical diagnostics, as long as their complementaryinding immunoglobulins can be produced. Current methodssed by in vitro immunoassay such as ELISA incorporatesmmunoglobulins or antigens tagged with markers into appro-riate biorecognition materials and coupled to a transducer suchs optical [5], fluorescence [6] or electrochemical sensor [7–9].ypically, these techniques give a linear response of 10 �g L−1

o 150 mg L−1 and detection limit of 10 �g L−1 [10].Herein, we describe a method for measuring the amount of

rotein antigen based on monitoring the magnitude of diffu-ion limited faradiac current of a redox probe diffusing withinarrow channels of a nanoporous alumina matrix. Nanoporous

lumina is a highly regular, rigid and dense porous material withominal pore sizes ranging from 10 to 200 nm pore density ofbout 1 × 1010 pores cm−2 [11,12] and is chemically and ther-

∗ Corresponding author. Tel.: +65 6516 3887; fax: +65 6779 1691.E-mail address: chmtohcs@nus.edu.sg (C.-S. Toh).

ictDirt

013-4686/$ – see front matter © 2007 Elsevier Ltd. All rights reserved.oi:10.1016/j.electacta.2007.07.055

lin

ally stable [13]. These features are relatively easy to achievend inexpensive by comparison to conventional lithographicechniques. We reason that these same features could be usedo trap specific-binding antibody such as immunoglobulin Githin the confined spaces of the vertical channels within an

lumina matrix. In addition, these immunoglobulin G coatedhannels could function as diffusion paths for a redox probe,errocenemethanol, chosen for its neutral charge and electro-hemically reversible behaviour. Fig. 1 shows the basic designf the biosensor which explains its scheme of operation. Aayer of aluminum (ca. 400 nm thick) was sputtered onto aome-made platinum wire electrode tip and anodized to aluminasing a pipette anodization method, which yields barrier-freelumina [14]. A sub-monolayer or monolayer of immunoglob-lin G was then immobilized along the nano-channel wallsf the porous alumina, followed by immobilization of bovineerum albumin (BSA) to block the unspecific adsorption sites.he alumina-modified platinum wire electrode was subse-uently used for antigen detection, in the presence of theedox probe. Upon binding the complementary antigen to themmunoglobulin G, formation of the antigen–antibody (Ag–Ab)omplexes blocked the approach of ferrocenemethanol towardshe exposed platinum surface beneath the porous alumina layer.

ifferential pulse voltammetry (DPV) was employed to mon-

tor the faradiac current limited by the diffusion rate of theedox probe diffusing towards the underlying platinum elec-rode.

804 G. Koh et al. / Electrochimica

Fig. 1. Scheme of antigen detection. A monolayer or submonolayer of antibodywas physically adsorbed onto the channel walls of the nanosized alumina pores,followed by blocking of non-specific adsorption sites with BSA and finally thealumina-modified electrode was used for antigen detection. Formation of theAau

o(mcmi

2

2

mtsp(bav(tw10aGti

2

me

mfired1AuIpamtstasmteds1psr

2e

eedoitef6tt6cBrratsrpl

b–Ag complex resulted in the narrowing and blocking of the nanosized poresnd the subsequent decrease in signal response towards a redox probe measuredsing differential pulse voltammetry.

We demonstrated the feasibility of using this scheme ofperation for antigen detection, by choosing glucose oxidaseGOx) and anti-glucose oxidase (immunoglobulin G) as ourodel antigen-antibody system. The antigen GOx system was

hosen because GOx is a well-studied protein with knownolecular mass, size and shows high binding specificity for its

mmunoglobulin G antibody.

. Experimental

.1. Reagents and materials

Monoclonal anti-glucose oxidase (anti-GOx) provided asouse ascites fluid with 0.1% sodium azide as preserva-

ive, glucose oxidase (GOx) from Aspergillus niger, bovineerum albumin (BSA, >98%), ferrocenemethanol (>99%) andhosphoric acid (85%) were purchased from Sigma–AldrichSingapore). All proteins solutions were prepared in phosphateuffered saline (PBS; containing 100 mM sodium phosphatend 1.7 M sodium chloride; pH 6.8). All chemicals and sol-ents of analytical grade were used as received. Ultrapure waterNanopure Ultrapure Water System) was used for all prepara-ions, unless otherwise stated. Glutathione S-transferase (GST)as supplied by BST Scientific Pte Ltd. in the form of amg mL−1 solution in phosphate buffered saline containing.02% Tween 20 and 0.02% sodium azide. Mouse monoclonalntibody (IgG2b-kappa, clone GST 3-4C) raised against 26 kDaST protein from S. japonica was supplied by ZYMED labora-

ories Inc. as a 200 mL aliquot at a concentration of 0.5 mg mL−1

n PBS, pH 7.4, containing 0.1% sodium azide (NaN3).

.2. Electrode fabrication

Home-made electrodes were fabricated using epoxy glue,icropipette tips and platinum wire (0.076 mm diameter). The

lectrode tip was polished with 1.0 and 0.3 �m diameter alu-

rmtt

Acta 53 (2007) 803–810

ina powder before being sputter-coated with aluminum metallm. Sub-micrometer thick aluminum films in the thicknessange of 300–500 nm were sputter-coated onto the platinumlectrodes using 99.999% purity aluminum target, Dentoniscovery® 18 Sputtering System and sputtering power of00W in an atmosphere of research-grade Ar at 5 × 10−3 Torr.nodization of aluminum-coated electrodes were conductedsing Apelex electrophoresis power supply model P304 minipacI. All aluminum-coated electrodes were rinsed with ultra-ure water before anodized potentiostatically in 0.1 M oxaliccid at 40 V. Anodization using the surface contact anodizationethod described elsewhere [14], was carried out by posi-

ioning a platinum coated glass pipette in contact with theurface of the aluminum-coated electrode. A two-step anodiza-ion process was used in which the aluminum was first anodizedt 40 V in 0.1 M oxalic acid solution, followed by immer-ion in 3% H3PO4 solution containing 0.2 M CrO3 for oneinute and a second anodization step until the current decreased

o zero. Electrochemical behaviours of the alumina-modifiedlectrodes were characterized using cyclic voltammetry andifferential pulsed voltammetry techniques (CHI440 potentio-tat/galvanostat, data acquisition software) in the presence of.0 mM ferrocenemethanol in 0.1 M phosphate buffer solution,H 6.8 (refer Ref. [14] for details). All potentials were mea-ured with respect to the silver–silver chloride (saturated KCl)eference electrode.

.3. Preparation of immunosensor derived from aluminalectrode

Fig. 2A shows the signal responses of the membrane-basedlectrochemical immunosensor electrode obtained by differ-ntial pulse voltammetry technique towards the redox probe,uring each step of the sensor preparation procedure carriedut consecutively as follows. A sub-monolayer to monolayer ofmmunoglobulin G was first immobilized via physical adsorp-ion onto the alumina by immersing the alumina-modifiedlectrode in 5 �L of 0.4 mg L−1 anti-GOx IgG solution for 1 h,ollowed by rinsing in 0.1 M phosphate buffer solution (pH.8). The DPV signal response of this alumina-IgG electrodeowards ferrocenemethanol was measured in a solution con-aining 1 mM ferrocenemethanol, 0.1 M phosphate buffer (pH.8). The electrode was subsequently transferred to and physi-ally adsorbed with BSA in a solution containing 200 mg L−1

SA for 30 min, followed by rinsing, transferred to the fer-ocenemethanol solution and measurement of immunosensoresponse towards the redox probe using DPV. Adsorption of IgGnd BSA onto the alumina electrode reduced the signal responseowards ferrocenemethanol (Fig. 2A). The immunosensor wastable, with no significant change (±2%) in the DPV signalesponse towards redox probe in the absence of antigen over aeriod of 2 h. A precautionary measure we had taken to estab-ish this stability was to immerse the immunosensor in the

edox probe solution for 30 min before commencing measure-ent of antigen concentrations in order for the redox probe

o infiltrate the nano-channels of the alumina layer. Withouthis preparative step, the electrode showed erratic behaviour

G. Koh et al. / Electrochimica Acta 53 (2007) 803–810 805

Fig. 2. (A) DPV signal response obtained at a 60 min etched alumina electrode immersed in 1 mM FeMeOH, after each step of the sensor preparation procedure (a)before adsorption of IgG, (b) after adsorption of anti-GOx IgG, (c) after adsorption of BSA and (d) after formation of IgG-antigen complex. Experimental conditionsw he reds kgrouo (A) an

dt

2

sBsco1abaecsbsowtsdG

cta

Ds

3

3

tppcdmubemwtaisGit

ere described in text. (B) DPV signal response of an immunosensor towards tignal response for an immunosensor prepared similarly as in (A). Average bacf all the experiments were ±0.5 and ±0.4 nA for both electrodes presented at

uring measurement of the first one or two antigen concentra-ions.

.4. Glucose oxidase detection

Detection of the GOx antigen was carried out using a batchystem in which the prepared immunosensor (alumina-IgG-SA electrode) was kept immersed in 15 mL of a buffered

olution containing 1 mM ferrocenemethanol, pH 6.8. Con-entration of GOx in the solution was varied by additionf aliquots of a 10 mg L−1 GOx stock solution containingmM ferrocenemethanol. Further quartz crystal microbal-nce (QCM) studies revealed that significant antigen–antibodyinding occurred within the first 2 min of immersing thelumina-modified electrode in the antigen solution while near-quilibrium was reached after ca. 30 min, when the observedurrent in cyclic voltammetry coupled to QCM stabilized. In aeparate study, during this immersion time in the antigen solutionetween 2–30 min, continuous differential pulse voltammetricignal response obtained at intervals of 2 min similarly decreasedver time and stabilized after 25–30 min. Detection of antigenas thus, carried out after immersion times of immunosensor in

he antigen solution for 30 min. Fig. 2B shows the decreasingignal responses of the immunosensor towards the redox probe,uring successive additions of increasing concentrations of theOx, using differential pulse voltammetry.

All differential pulsed voltammetry (DPV) experiments were

onducted using the prepared electrode as the working elec-rode, a platinum disk electrode as the auxillary electrode andsilver/silver chloride reference electrode. Parameters used for

do(o

ox probe in increasing concentration of GOx, with corresponding decrease innd noise determined from standard deviations of three consecutive DPV scansd (B), respectively.

PV were as follows: amplitude of 50 mV, pulse width of 0.05 s,ample width of 0.0167 s and scan range from −0.2 to 0.8 V.

. Results and discussion

.1. Control experiments and selectivity studies

Several control studies were carried out using a high concen-ration of 0.1 mg L−1 GOx antigen concentration. First, a barelatinum electrode was prepared under identical immunosensorreparation procedure, but without the alumina overlayer, indi-ated high constant signal response towards the redox probe, atifferent antigen concentrations. In the second control experi-ent, a 60 min etched alumina-modified electrode was prepared

nder the same conditions as the working immunosensor, immo-ilized with BSA, but without immunoglobulin G (alumina-BSAlectrode). The third control used a 60 min etched alumina-odified electrode immobilized with immunoglobulin G, butithout BSA (alumina-IgG electrode). Fig. 3 shows the differen-

ial pulse voltammograms obtained for the three controls beforend after the addition of glucose oxidase antigen into the sens-ng solution. Both the alumina-BSA and alumina-IgG electrodeshowed decrease in signal response in the presence of antigenOx (Fig. 3B and C). It was expected the presence of BSA

n the alumina-BSA electrode prevented non-specific adsorp-ion along the BSA coated alumina channel walls. The observed

ecrease in signal response was likely the result of displacementf the BSA molecules (65 kDa) by the bulkier GOx molecules180 kDa). It was unclear whether the decrease in signal responsef the alumina-IgG electrode in the presence of GOx was due

806 G. Koh et al. / Electrochimica Acta 53 (2007) 803–810

F electf DPV

tFfnteatbptaotlnets

at

uilFIGwcodAs

ig. 3. DPV signal responses of (A) platinum wire electrode, (B) alumina-BSAor the various control studies. Average background noise of three consecutive

o specific or non-specific binding. However, it is obvious fromig. 3D, the decreases in signal responses in the presence of GOxor both alumina-BSA and alumina-IgG control electrodes wereot cumulative, but were significantly less than that obtained athe alumina-IgG-BSA electrode. The signal responses of theselectrodes towards the redox probe obtained in the presence ofntigen GOx were compared to the signal responses towardshe redox probe by their respective bare platinum electrodesefore coating with alumina overlayers. It was found that in theresence of 0.1 mg L−1 GOx, the alumina-BSA control elec-rode showed 74 ± 7% reduction and the alumina-IgG electrodesimilar 65 ± 5% reduction with respect to the signal responsesbtained at their respective uncoated platinum electrode. In con-rast, the alumina-IgG-BSA electrode showed a significantlyarger signal reduction of 82 ± 1% which indicated the chan-

els within the alumina-IgG-BSA electrode were blocked moreffectively compared to the alumina-BSA and alumina-IgG elec-rodes. This suggests specific binding which yields a largerize immunocomplex probably occurs to greater extent in the

0rag

rode, (C) alumina-IgG electrode and (D) alumina-IgG-BSA electrode obtainedfor the control experiments was in the range of ±0.1 to ±2 nA.

lumina-IgG-BSA electrode, compared to the alumina-IgG elec-rode.

We carried out further study on alumina-IgG-BSA electrodesing another protein, glutathione S-transferase (GST) whichndicated non-specific binding of antigen molecules was neg-igible when both IgG and BSA were present in the electrode.ig. 4A shows the responses of a GOx immunosensor (alumina-gG-BSA electrode) towards GOx and non-specific antigenST. A decrease in signal response of the GOx immunosensoras observed in the presence of 0.03 mg L−1of GOx as expected,

ompared to a buffer solution without GOx. Subsequent additionf 0.1 mg L−1 non-specific antigen GST resulted in no furtherecrease in signal response of the same GOx immunosensor.dditional decrease in signal response was observed when a

olution containing both GOx and GST at concentrations of

.10 mg L−1 was subsequently added. The experiment was car-ied out in reverse order using a second GOx immunosensor, bydding the non-specific antigen GST first, followed by the anti-en GOx. The GOx immunosensor showed no change in signal

G. Koh et al. / Electrochimica Acta 53 (2007) 803–810 807

Fig. 4. Signal response obtained by DPV technique towards the redox probe for (A) GOx immunosensor and (B) GST immunosensor. Each bar represents the signalr specifib ise oft

rnSIpi0Gtfaer(Isaaδ

rolaTaiidtscutan

3

t

m6itistttWreebwtuantigen concentrations for the unetched electrode was only ca.0.8 times of the initial signal response in the absence of antigen.

It was expected alumina-modified electrodes with poresizes significantly larger than the immunoglobulin and GOx

Fig. 5. DPV signal response towards the redox probe for an immunosensor

esponse obtained towards the redox probe after adding aliquots of antigen, non-uffered solution of 1 mM ferrocenemethanol (pH 7.5). Average background noo ±1.6 nA.

esponse in the presence of GST, but significant change in sig-al responses in the presence of GOx or GOx–GST mixture.imilarly, using an immunosensor immobilized with anti-GSTgG, negligible signal response change was observed in theresence of non-specific antigen GOx (Fig. 4B). The GSTmmunosensor was immersed sequentially in 0.03 mg L−1 GST,.03 mg L−1 GOx and finally, a mixture containing 0.10 mg L−1

ST and GOx. Thus, non-specific binding of GST or GOx ifhese occurred, resulted in negligible channel blocking effector the alumina-IgG-BSA electrodes. Signal responses of anlumina-modified electrode towards the redox probe in the pres-nce of antigen (δiGOx) were normalized against the signalesponse of the same alumina electrode immobilized with IgGalumina-IgG) (δianti-GOx), instead of IgG and BSA (alumina-gG-BSA) (δianti-GOx/BSA). δianti-GOx and δianti-GOx/BSA refer toignal responses obtained at the same alumina electrode after thentibody and BSA immobilization step, respectively, and in thebsence of antigen. Normalization of signal responses againstianti-GOx using the ratio of δiGOx/δianti-GOx gave reproducibleesults for different electrodes and good linear logarithmic plotsf normalized signal response against concentrations. This isikely due to irreproducible non-specific binding between BSAnd IgG close to or at the antigen-binding site of IgG molecules.he BSA molecules bound at these parts of the IgG moleculesre readily displaced by the GOx antigen which binds specif-cally to IgG. This method of data presentation significantlymproved the reproducibility of the immunosensor for threeifferent electrodes. For example, the relative standard devia-ion for three different 60 min etched electrodes prepared underimilar conditions ranged from ±4.1% at 200 ng L−1 antigenoncentration to ±13.8% at 10 �g L−1 antigen concentrationsing the normalized signal responses (δiGOx/δianti-GOx). In con-rast, the relative standard deviation were ±26.4% (200 ng L−1)nd ±15.6% (10 �g L−1) respectively when the same data wereormalised using δiGOx/δianti-GOx/BSA.

.2. Study of channel diameters

In order to understand the effect of channel diameters onhe sensing response of the immunosensor, we etched the alu-

pceo2

c antigen and mixture of antigen/non-specific antigen consecutively to a 15 mL3 consecutive DPV after each step of the experiment was in the range of ±0.1

ina overlayer in 3% phosphoric acids for durations of 0, 30 or0 min. Each alumina-modified electrode was thoroughly rinsedn nanopure water after the etching process in order to removehe dissolved alumina which caused significant electrode foul-ng problem. Fig. 5 shows the decreasing signal response of theensor towards increasing concentrations of GOx antigen forhree different alumina-modified electrodes etched for differentime durations in phosphoric acid, where the 60 min etched elec-rode exhibited the steepest slope of all the calibration curves.

e observed two general trends. First, the decrease in signalesponse per unit antigen concentration (sensitivity) was high-st for electrodes etched at longest time of 60 min. Second, thelectrodes etched at 60 min also exhibited the most effectivelocking of pore channels at high antigen concentrations, inhich its signal response decreased to less than 0.5 times of

he initial measured signal response when the channels werenblocked. In contrast, the signal response in antigen at high

repared using 0.4 mg L−1 anti-GOx and 200 mg L−1 BSA over varying con-entrations of GOx. Immunosensors were derived from alumina electrodestched for 0, 30 and 60 min in 3% phosphoric acid solutions. Concentrationf ferrocenemethanol used was 1 mM and all experiments were conducted at97 K.

8 mica

mrgm1mdZafm(cratwIetamCrBott

psiataan

3

rtmaotpfioat

Ft(

08 G. Koh et al. / Electrochi

olecules would display poor sensitive response and lowesistance towards diffusion of the redox probe at high anti-en concentrations. The dimension of an immunoglobulin Golecule obtained from X-ray diffraction studies was known,

4.5 nm × 8.5 nm × 4 nm [15] and average size of a GOxolecule determined in phosphate buffered saline (PBS), using

ynamic light scattering technique (Malvern Zetasizer NanoS) was 14.0 ± 0.3 nm. SEM studies of the side profiles oflumina overlayers after different duration of etching timesurther revealed narrow channels extended throughout the alu-ina overlayer for an unetched alumina-modified electrode

Fig. 6). For a 15 min etched alumina-modified electrode, thehannels were widened at the side exposed to solution and nar-ower at the alumina-electrode interface. For the 60 min etchedlumina-modified electrode, the alumina overlayer showed twoypes of layers, a highly porous reticulate layer at the surfaceith pore sizes at least 10 times larger than either antigen or

gG molecule. The inner layer of the 60 min etched aluminalectrode resembled unetched alumina with channel diame-ers ca. 30–50 nm (Fig. 6), the result of slower etching ratest the ‘inner’ part of channels due to depletion and slowass transfer of acid reagent within the alumina overlayer.hannel diameters within the inner alumina overlayer thus

emained relatively unchanged over the varying etching time.

esides the change in pore sizes and channel diameters, thether effect of chemical etching on alumina-modified elec-rodes was a reduction in the ratio of outer surface area to theotal internal channel volume of the alumina overlayer. It was

ripp

ig. 6. Scanning electron micrographs of side profiles of alumina overlayers etchedhe outer films show extensive merging with poorly defined shapes. Pore sizes of theC) 30–75 nm and (D) 30–100 nm. Pore densities of inner films range from 5.9 × 109

Acta 53 (2007) 803–810

ossible that binding of antigen molecules to IgG moleculesited along the alumina outer surface occurred which could notnfluence the sensor response. Whereas, an extensively etchedlumina-modified electrode with reduced ratio of outer surface-o-channel volume likely exhibited sensitive response towardsntigen due to majority of antigen-antibody binding occurringlong the diffusion path of the redox probe within the chan-els.

.3. Electrochemical studies

During a potential step experiment of an electrochemicaleversible redox couple, the faradaic current flow at a planar elec-rode can be described by a Cottrell-type equation in which the

agnitude of the transient current is inversely related to time1/2

s well as the equilibrium surface concentration ratio of thexidized and reduced forms of the redox couple after the poten-ial step [16]. The signal response obtained during a differentialulse voltammetric experiment derived from the difference in thearadaic current flow before and after the pulse, is thus expressedn the form of the same Cottrell-type equation with considerationf the equilibrium surface concentration ratios of the oxidizednd reduced forms of the redox couple before and after the poten-ial step. For a nernstian system, these concentration ratios are

elated exponentially to the applied potentials and it can be read-ly shown at E = Emax (where Emax = E0′ − �E/2 and �E is theulse height applied during the DPV experiment), the maximumeak height is linearly proportional to concentration of the redox

in 3% phosphoric acid for (A) 15, (B) 30, (C) 45 and (D) 60 min. The pores ofinner film estimated from SEM micrographs were (A) 20–50 nm (B) 25–70 nmto 7 × 109 pores cm−2 for (B), (C) and (D).

mica

s

(

(nc((

1reIfwmt(cofordid

Fatopes

aisetoocdrtwus[uwTuoatm

G. Koh et al. / Electrochi

pecies for an electrochemically reversible system [16]:

δi)max = nFAD1/2Cbulk

π1/2(τ − τ′)1/2

[1 − σ

1 + σ

](1)

δi)max is the maximum peak height or DPV signal response,= 1, F = Faraday constant, D = diffusion coefficient of ferro-enemethanol, Cbulk = bulk concentration of ferrocenemethanol,τ − τ′) = pulse duration, �E = pulse amplitude and σ = exp−nF�E/2RT) for the oxidation process.

Fig. 7 shows the plot of maximum peak height (δi)max against− σ/1 + σ obtained during DPV measurements towards the

edox probe over varying pulse heights for unetched and 60 mintched alumina electrodes immobilized with IgG, IgG/BSA orgG/BSA/antigen. Plots are shown for pulse amplitude valuesrom 0 to 130 mV, which demonstrated good linearity, consistentith an electrochemically reversible system under the experi-ental conditions described in Fig. 7. The inset at Fig. 7 shows

he magnitude of AD1/2 derived from the slopes of the plots ofδi)max against 1 − σ/1 + σ. In our immunosensor system, theoncentration of ferrocenemethanol was kept constant through-ut the experiments. Thus changes in the maximum peak heightor the different experiments could only arise due to variationf the electrode active area and/or diffusion coefficient of the

edox probe, according to Eq. (1). Fig. 7 inset revealed gradualecrease in the magnitude of AD1/2 for the IgG and IgG/BSAmmobilized 60 min etched alumina electrode, but significantecrease for the IgG/BSA/antigen immobilized 60 min etched

ig. 7. Plots of (δi)max against (1 − σ)/(1 + �) obtained by varying DPV pulsemplitude from 10 to 130 mV with a pulse width of 50 ms in an unstirred solu-ion containing 1 mM ferrocenemethanol at 297 K. Relative standard deviationf (δi)max ranges from ±4.2% for 10 mV pulse amplitude to ±22.7% for 130 mVulse amplitude. Inset: variation of AD1/2 with respect to types of aluminalectrodes immobilized with IgG, IgG/BSA or IgG/BSA/antigen, derived fromlopes and slope errors of best-fit lines through the plots presented in this figure.

i2ootdtvacGwectwteott[

4

rGvaaai

Acta 53 (2007) 803–810 809

lumina electrode. This trend agrees with the result obtainedn Fig. 2A in which the DPV waveforms similarly revealedignificant decrease in the maximum peak height in the pres-nce of antigen. The magnitude of AD1/2 values are slightly lesshan the known diffusion coefficient value of ferrocenemethanolf 1 × 10−5 cm2 s−1 and that calculated using the surface areaf the 76 �m diameter electrode overlaid by an alumina layeromprising ca. 30 nm size pores of ca. 7 × 109 pores cm−2 poreensity. It is likely due to reduced diffusion coefficient of theedox probe when the channel diameters were narrowed inhe presence of antigen due to formation of immunocomplexithin the channels, as determined by molecular dynamic stim-lation study carried out for system comprising channel sizemaller than five times the diameter of the diffusing species17]. Interestingly, the magnitude of AD1/2 obtained at thenetched alumina-modified electrode remained fairly constanthen immobilized using IgG, IgG/BSA and IgG/BSA/antigen.his was due to ineffective blocking of the channels for thenetched alumina electrode, likely the results of a large numberf available binding sites along the outer surface of the unetchedlumina overlayer and a smaller channel size in comparison tohe etched electrode, reduce the ease of access by the biological

olecules.Measurement of the signal-to-noise ratio revealed a decreas-

ng signal-to-noise (S/N) ratio, for example, from S/N = 28 at00 ng L−1 to S/N = 16 at 50 �g L−1 antigen concentration forne of the immunosensor electrodes. This has the advantagef increased sensitivity towards the antigen at low concentra-ions, unlike most sensors in which the signal-to-noise ratioecreases as concentration of analyte decreases. Limit of detec-ion was derived from the minimum antigen concentrationalue which produced a reduction of sensor signal equiv-lent to three times the background noise at zero antigenoncentration. A low-detection limit (LOD) of 100 ng L−1 ofOx antigen was achieved using this simple sensing approachith a linear logarithmic concentration dependence range,

xtending from 200 ng L−1 to 11.2 �g L−1 (correlation coeffi-ient = 0.993). In comparison, commercially available ELISAechnique has a higher LOD of ca. 10 �g L−1. Unlike ELISA,here large quantity of antibody (50 �L of 20 mg L−1) is

ypically used for protein detection, this membrane-basedlectrochemical immunosensor system requires only ca. 5 �Lf 0.4 mg L−1 antibody for sensor preparation. The detec-ion limit of this technique is significantly lower comparedo other similar immunosensors using static batch system3,18,19].

. Conclusion

Overall, a membrane-based immunosensor with sensitiveesponse has been developed for the detection of the protein,Ox. An extensively etched alumina membrane overlayer pro-ides the greatest decrease in signal response towards the protein

ntigen molecules over the concentration range tested. The samepproach may be applicable to detection of other proteins, usingppropriate complementary binding immunoglobulin antibod-es.

8 mica

A

fn

R

[

[[[[

[

[Applications, John Wiley and Sons, New York, 2000.

10 G. Koh et al. / Electrochi

cknowledgements

We thank NUS (Grant No. R143-000-219-101) for projectunding, NUSNNI for use of sputtering system and DSKH tech-ology (Material Science Division) for DLS equipment.

eferences

[1] A.J. Baeumner, N.A. Schlesinger, N.S. Slutzki, J. Romano, E.M. Lee, R.A.Montagna, Anal. Chem. 74 (2002) 1442.

[2] S. Booth, C. Baleriola, W.D. Rawlinson, J. Med. Virol. 78 (2006) 619.[3] M. Okochi, H. Ohta, T. Taguchi, H. Ohta, T. Matsunaga, Electrochim. Acta

51 (2005) 952.[4] S. Vetcha, I. Abdel-Hamid, P. Artanasov, D. Ivnitski, E. Wikkins, B. Hjelle,

Electroanalysis 12 (2000) 1034.[5] M. Tudorache, M. Co, H. Lifgren, J. Emneus, Anal. Chem. 77 (2005) 7156.[6] H. Volland, L.M. Neuburger, E. Schultz, J. Grassi, F. Perraut, C. Creminon,

Anal. Chem. 77 (2005) 1896.[7] L. Wu, J. Chen, D. Du, H. Ju, Electrochim. Acta 51 (2006) 1208.

[[

[

Acta 53 (2007) 803–810

[8] E. Palecek, R. Kizek, L. Havran, S. Billova, M. Fojta, Anal. Chim. Acta469 (2002) 73.

[9] M. O’Connor, S.N. Kim, A. Killard, R.J. Forster, M.R. Smyth, F. Papadim-itrakopoulos, J.F. Rusling, Analyst 129 (2004) 1176.

10] Invitrogen, 2006, http://probes.invitrogen.com/lit/catalog/1/sections/7830.html. (Accessed October 5, 2006).

11] J.W. Diggle, T.C. Downie, C.W. Goulding, Chem. Rev. 69 (1969) 365.12] J.P. O’Sullivan, G.C. Wood, Proc. Roy. Soc. A 317 (1970) 511.13] H.-S. Kang, Y.-H. Chang, C.-W. Lee, Korean J. Chem. Eng. 17 (2000) 266.14] G. Koh, S. Agarwal, P.S. Cheow, C.-S. Toh, Electrochim. Acta 52 (2007)

2815.15] C.J. Roberts, P.M. Williams, J. Davies, A.C. Dawkes, J. Sefton, J.C.

Edwards, A.G. Haymes, C. Bestwick, M.C. Davies, S.J.B. Tendler, Lang-muir 11 (1995) 1822.

16] A.J. Bard, L.R. Faulkner, Electrochemical Methods: Fundamentals and

17] Y.W. Tang, I. Szalai, K.-Y. Chan, J. Phys. Chem. A 105 (2001) 9616.18] D. Tang, R. Yuan, Y. Chai, Y. Fu, J. Dai, Y. Liu, X. Zhong, Biosens.

Bioelectron. 21 (2005) 539.19] M.S. Wilson, R.D. Rauh, Biosens. Bioelectron. 19 (2004) 693.