AC electrokinetics-enhanced capacitive immunosensorfor point-of-care serodiagnosis of infectious diseases

Shanshan Li a,b, Haochen Cui a, Quan Yuan a, Jie Wu a,n, Ashutosh Wadhwa c,Shigetoshi Eda c, Hongyuan Jiang b

a Department of Electrical Engineering and Computer Science, The University of Tennessee, Knoxville, TN 37996, USAb School of Mechatronics Engineering, Harbin Institute of Technology, Harbin, 150001, PR Chinac Department of Forestry, Wildlife and Fisheries, The University of Tennessee, Knoxville, TN 37996, USA

a r t i c l e i n f o

Article history:Received 30 May 2013Received in revised form8 August 2013Accepted 12 August 2013Available online 23 August 2013

Keywords:AC electrokineticsDielectrophoresisImmunosensorCapacitive sensingPoint of care diagnosis

a b s t r a c t

The current serological diagnostic method can be time consuming and labor intensive, which is notpractical for on-site diagnosis and screening of infectious diseases. Capacitive bioaffinity detection usingmicroelectrodes is considered as a promising label-free method for point-of-care diagnosis, though withchallenges in sensitivity and the time “from sample to result.” With recent development in ACelectrokinetics (ACEK), especially in dielectrophoresis (DEP), we are able to develop an ACEK enhancedcapacitive bioaffinity sensing method to realize simple, fast and sensitive diagnosis from serum samples.The capacitive immunosensor presented here employs elevated AC potentials at a fixed frequency forimpedimetric interrogation of the microelectrodes. According to prior work, such an AC signal is capableof inducing dielectrophoresis and other ACEK effects, so as to realize in-situ enrichment of macro-molecules at microelectrodes and hence accelerated detection. Experimental study of the ACEK-enhanced capacitive sensing method was conducted, and the results corroborate our hypothesis. Thecapacitive sensing responses showed clear frequency dependence and voltage-level dependency, whichsupports the hypothesis that ACEK aids the antigen–antibody binding, and these dependencies wereused to optimize our detection protocol. Our capacitive sensing method was shown to work with bovinesera to differentiate disease-positive samples from negative samples within 2 min, while conventionalimmunoassay would require multiple processing steps and take hours to complete. The results showedhigh accuracy and sensitivity. The detection limit is found to reach 10 ng/ml in 2 min. The ACEK-enhanced capacitive immunosensor is a platform technology, and can be employed to detect anycombination of probe (e.g. antigen) and analyte (e.g. serum antibody) in a small volume of bodily fluids.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

The capability of rapidly diagnosing the disease and identifyingits causative agent is critical to combat diseases and halt epi-demics. As of now, enzyme-linked immunosorbent assay (ELISA) isone of the most common test formats used for laboratorydiagnosis of infections (Wright et al., 1993; Lequin, 2005). Diag-nosis of infectious disease with ELISA is mostly based on thespecific detection of disease-specific antibodies in body fluids byusing antigen of the causative agent as a specific probe. However,ELISA requires testing in a laboratory facility and, therefore, can betime consuming and labor intensive (Lequin, 2005; Nakamuraet al., 1992; Xiang et al., 2006).

Recent development in the areas of microfluidics and biosensortechnology is leading to improved diagnostic systems and proce-dures. Among them, capacitive sensing by microelectrodes is alabel free detection method for characterizing material depositionat electrode surfaces, well suited for sensing the formation ofantigen–antibody complexes or assembly of protein arrays atelectrode surfaces (Berggren et al., 2001; Hedström et al., 2005;Daniels and Pourmand, 2007). The underlying mechanism is thatmaterial changes near the electrode surface will cause changes inthe capacitance at the electrolyte/microelectrode interfaces, thusproviding a direct means of detecting target–probe binding reac-tions on the sensor (Berggren and Johansson, 1997; Guan et al.,2004). The capacitive response is generally proportional to theamount of captured biomolecules.

Capacitive sensing does not require molecular labeling of a targetanalyte, which simplifies the sample preparation and reduces thenumber of steps in the assay workflow. Other inherent advantagesinclude easy miniaturization, small analyte volumes, and compatibility

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/bios

Biosensors and Bioelectronics

0956-5663/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.bios.2013.08.016

n Corresponding author. Tel.: þ1 865 974 5494; fax: þ1 865 974 5483.E-mail address: jaynewu@utk.edu (J. Wu).

Biosensors and Bioelectronics 51 (2014) 437–443

with turbid biofluids. Therefore, capacitive biosensing by microelec-trodes holds great promise as a direct, single-step assay procedure forpoint-of-care (POC) applications.

While there has been extensive research on capacitive affinitysensors for on-site immunological diagnosis (Klein et al., 1995;Betty et al., 2004; Diem et al., 2004; Cosnier, 2005; Daniels andPourmand, 2007; Labib et al., 2010), capacitive affinity sensors stillfaces several challenges in order to expand their application scope.One obstacle is long testing time for diluted samples or lack ofsensitivity. Of the reported capacitive immunosensors, the time toreach a reliable detection response relies strongly on the diffusionof target macromolecules towards the electrode surfaces (Sheehanand Whitman, 2005). As a result, each binding step typically takesfrom 30 min to several hours to complete, even longer time “fromsample to result” due to the need for sample pretreatment andmultiple washing/incubation steps, not practical for on-sitescreening.

This paper presents an AC electrokinetic (ACEK)-enhancedcapacitive sensing method for low cost, rapid and accurate POCdiagnosis. Commercially available microelectrode chips were mod-ified and adopted as single use sensors, which cost less than $1 pertest. The diagnostic process only involves loading the serumsample and applying an AC signal for impedimetric interrogation.Different from conventional impedimetric sensing, which uses alow AC voltage of 5–10 mVrms before and after the binding, ourmethod reads the electrode impedance continuously with avoltage of 100 mVrms at 100 kHz. With this AC signal, the targetmacromolecules will experience sufficiently strong attraction forcetowards the electrode surface. As a result, binding between theantigen–antibody is accelerated. Our system can achieve “fromsample to results” within several minutes. This work used rawserum samples from cows with Johne's disease (JD) (Stabel, 1998).ELISA test for JD is currently conducted in diagnostic laboratories,causing a long turn-around time and consequently hampering itscontrol and prevention. Once commercialized, the capacitiveimmunosensor reported in this work would reduce thediagnosis-related costs and thereby facilitate JD control efforts inglobal dairy industries. As this is a platform technology, its furtherdevelopment could offer exciting opportunities for numerousdecentralized clinical applications for not only JD but also forother infectious diseases.

2. Mechanisms

2.1. Detection of surface binding by interfacial capacitance

Capacitive sensing of macromolecule binding is usually achievedby extracting any change in the interfacial capacitance at theelectrode surface before and after the binding reaction. The inter-facial capacitance can be attributed to the formation of electricdouble layer and the deposition of macromolecules on the electrodesurface. When electrodes are immersed into a liquid electrolyte, theelectrode surface will acquire net charges. To maintain chargeneutrality, a thin layer of counter ions will accumulate in the fluidnear the electrode surface, which is known as electric double layer(EDL) (Labib et al., 2009). The EDL can be regarded as the dielectriclayer of the interfacial capacitor. When macromolecules areadsorbed onto the electrode surface, it will increase the thicknessof the dielectric layer for the interfacial capacitor, consequently theinterfacial capacitor will change.

As schematically shown in Fig. 1, before the antigen–antibodybinding, the dielectric layer of the interfacial capacitor is made up bythe EDL λd and deposited antigens dag, and the initial interfacialcapacitance can be expressed as Cint ;o ¼ Aint=ð ð1=εpÞdagþð1=εsÞλdÞ,where εs and εp are the permittivities of the sample fluid and

proteins. The relative permittivity of protein is around 2–3 (Sculleyet al., 1980), while that for water is �80. Aint is the surface area of theinterfacial capacitor of the functionalized electrode, λd is the EDLthickness, and dag is the antigen thickness. After the binding reaction,the dielectric layer becomes the sum of the EDL, the probe dag, andthe bounded antibodies dab, and the interfacial capacitance changesto Cb ¼ Ab=½ 1=εp

� �ðdagþdabÞþð1=εsÞλd�, where Ab is the surfacearea of the interfacial capacitor after binding, dab is the antibodythickness. The relative change in capacitance is ΔC=Cint ;o ¼ðCb�Cint ;oÞ=Cint ;o ¼�dab=½ ðdagþdabÞþðεp=εsÞλd� assuming that Ain-tEAb. Consequently, the surface binding, i.e. the addition of dab, canbe detected through ΔC/Cint,o.

In this work, normalized capacitance change rate (i.e. ΔC/Cint,oover a fixed test period) is employed to indicate the surfacebinding events. Because ΔC/Cint,o can be directly correlated withthe amount of bounded macromolecules on the electrode surface,this is a quantitative detection method to a certain extend. Also,using capacitance change rate, i.e. ΔC/Cint,o, helps to improve thetest repeatability. In reality, when measuring the impedance of anelectrode/electrolyte cell, there always exist inconsistencies fromsensor to sensor, mostly due to the variations in electric propertiesof real life samples and the bottom-up nature of surface functio-nalization. As a result, the baseline and initial capacitances willhave a range instead of a well-defined value. Using the capacitancechange rate, the uncertainty in test results can be reduced, sincethe aforementioned test variations will be factored out.

2.2. Acceleration of binding by AC electric field

The obstacle to rapid detection is the diffusion of targetanalytes to the sensor, which may take hours even days for theanalytes to accumulate to a detectable level at the sensor (Sculleyet al., 1980; Qureshi et al., 2010). To achieve a useful sensitivitywithin practical time scales beyond what is dictated by thelimitation of molecular diffusion, active guidance of biomoleculesis needed to direct them to sensors. A number of strategies havebeen investigated to provide sample preconcentration in liquids,such as magnetic particles (Smith et al., 2011; Tennico andRemcho, 2010) and optical tweezers (Fan and White, 2011;Velasco et al., 2012). However, those methods are still toocomplicated for POC testing (Urdea et al., 2006). In contrast,

Fig. 1. Changes at the electrode surface due to the binding of specific antibody toimmobilized antigen.

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443438

electrokinetic in-situ concentration of analytes can be seamlesslyincorporated into capacitive detection to accelerate the incubationprocess (Wu 2010; Wang et al., 2012).

In our tests, the capacitive sensing was conducted at a pre-determined AC frequency with a continuously applied voltage of100 mVrms, which is much higher than typically used 5 mVrms.The reason for higher voltage is to induce ACEK forces, especiallyDEP forces, on antibodies to direct them towards the electrodesurface. In our prior work, we have demonstrated by both numer-ical study (Yang and Wu, 2010) and experiments (Liu et al., 2011)that the binding of antigen–antibody pairs at the bottom ofmicrochannel can be significantly accelerated by applying anappropriate AC signal over the interdigitated microelectrodes atthe channel bottom. In the immunosensor work by Liu et al. (2011),a fluorescently-labeled secondary antibody was used to detect thepresence of serum antibody bound to the antigen on the micro-electrode surface. By powering the microelectrodes at the channelbottom, the time needed for incubation decreased from 30 min bytraditional ELISA to less than 5 min by our method. Our prior studyon ACEK enhanced incubation using fluorescent labeling for ima-ging has enabled us to narrow down an AC frequency range foroptimal attraction of antibodies to the electrodes.

Based on our prior work (Liu et al., 2011; Yang et al., 2011),there are two possible mechanisms that are responsible for anti-body concentration. One is dielectrophoresis (DEP), and the otheris AC electrothermal effect (ACET). In the presence of an inhomo-geneous electric field, DEP can attract or repel particles based onthe difference between the particle polarizability and that of themedium solution (Morgan et al., 1999). The DEP force on aspherical particle can be expressed as

FDEP ¼ πεma3Reεnp�εnmεnpþ2εnm

" #∇ Ej2�� ð1Þ

where εm is the medium permitivity, a is the particle radiater andE is electric field strength. εnp, and εnm are the complex permittiv-ities of particle and medium, respectively. Complex permittivity isdefined as εn ¼ ε�jðs=ωεÞ, where ω is angular frequency and j isthe imaginary unit, defined as

ffiffiffiffiffiffiffi�1

p. Because complex permittiv-

ities are frequency dependent, the DEP force is also a function ofAC frequency.

As it can be seen from Eq. (1), the magnitude of DEP forcedepends on the particle volume. In case of enriching macromole-cules, DEP is effective mostly for macromolecules that are locatedwithin a very short distance to the electrodes (e.g. o1 mm). Formacromolecules that are farther away from the electrode, thebinding can be aided by fluidic vortices induced above themicroelectrodes by AC electric fields to convect macromoleculesto the electrodes. At ionic strength of interest to this work,microflows are generated by AC electrothermal (ACET) effect.Previous study showed that ACET effect has played an importantrole in increasing detection sensitivity (Feldman et al., 2007;Lian et al., 2007).

Time averaged ACET fluidic force can be given as (Green et al.,2001)

Fe ¼12Re

sεmðα�βÞsþ iωεm

ð∇T EÞEn�12εmαjEj2∇T

� �ð2Þ

in which α¼ ð1=εmÞð∂εm=∂TÞ � �0:004 K�1 andβ¼ ð1=sÞð∂s=∂TÞ � 0:02 K�1 for aqueous solution (Lian et al.,2007; Green et al., 2001; Birdi, 2008). Because fluidic forces haveno dependence on particle size (Eq. (2)), ACET microflows will bewell suited for transporting macromolecules to the electrodes.DEP together with ACET effect may have a much larger effectiverange for nanoparticle concentration.

3. Experimental methods

3.1. Reagent preparation

3.1.1. Buffer solutionsOur capacitive assay is based on dilute phosphate-buffered

saline (PBS) solutions, denoted as 0.1� PBS in this work. 0.1� PBSis also used as control sample in the testing. 0.1x PBS is preparedby 1:10 volume dilution of physiological strength PBS withdeionized water to obtain 1 mM phosphate buffer (pH 7.0) con-taining 15 mM NaCl. 0.1� PBS-T, washing solution, is 0.1� PBScontaining 0.05 v/v% Tween 20 (Fisher Scientific, Pittsburgh, PA).0.1� Buffer B is 0.1� PBS-T containing 10 v/v% SuperBlock(PIERCE Biotechnology, Rockford, IL), and it is used for dilutingserum samples as well as blocking the electrode chips.

3.1.2. JD-specific antigenA bacterial pathogen Mycobacterium avium subsp. paratubercu-

losis (MAP), which is the causative agent of JD, was used to extractantigen as described previously elsewhere (Liu et al., 2011; Edaet al., 2006). MAP was obtained from Dr. John P. Bannantine in theUSDA/NADC (Ames, IA) and was cultured in Middlebrook 7H9medium (Becton Dickinson Microbiology Systems, Franklin Lakes,NJ) with 10% OADC (oleic acid–albumin–dextrose–NaCl) (BectonDickinson Microbiology Systems, Franklin Lakes, NJ). The mediumwas supplemented with 2 mg/ml of Mycobactin J (Allied Monitor,Fayette, MO). The cultures were maintained at 37 1C withoutshaking until they reached an optical density of approximately0.7 at 600 nm. MAP was harvested from the liquid culture atstationary phase and centrifuged at 2600g for 10 min; the pelletwas resuspended in 80% ethanol, agitated by vortex at roomtemperature for 2 min, and centrifuged at 10,621g for 10 min.The supernatant was then collected, diluted in 100% ethanol (1:80v/v) and used as antigen for immunoassay.

3.1.3. JD serum samplesTen JD-positive serum samples were obtained from 10 different

dairy cows (breed, Holstein; average age, 2.8 years old; gender,female) tested positive for JD by bacterial culture and ELISA. TenJD-negative serum samples were obtained from 10 different cows(breed, Holstein; average age, 2.4 years old; gender, female) testednegative for JD by the procedures mentioned above. The bovineserum samples used in this work were kindly provided by NationalAnimal Disease Center, USDA, Ames, IA. For testing, the serumsamples were 1:20 v/v diluted with 0.1� PBS before use.

3.1.4. Antigen and antibody for limit of detection testsThe limit of detection (LOD) was determined by measuring

binding of goat anti-bovine IgG (HþL) antibody (unlabeled,Jackson ImmunoResearch Laboratories Inc., West Grove, PA) tobovine IgG whole molecules (Jackson ImmunoResearch Labora-tories Inc., West Grove, PA). The concentration of the bovine IgGwhole molecules was 10 μg/mL (diluted in 0.1� Buffer B). Differ-ent concentrations of the anti-bovine IgG antibody were used atconcentrations ranging 1–10,000 ng/mL (diluted in 0.1� Buffer B).

3.2. Preparation of microelectrode sensors

The interdigitated microelectrode arrays for antigen–antibodybinding assay were obtained by modifying commercially availablesurface acoustic wave (SAW) resonator PARS 433.92 (AVX, Corp.).Similar types of SAW chips have been used by other groups(Stanke et al., 2011) to study ACEK effects by the interdigitatedmicroelectrodes. They observed fluid streaming above the electro-des along with DEP experiments from 10 Hz to 1 GHz, proving that

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443 439

ACET flows could be induced concomitantly with DEP for this typeof microelectrodes.

To modify the SAW resonator for immunosensing, the metalcover of the SAW resonator was removed mechanically, and theworking electrode array was exposed for use (The images aregiven as Fig. S1a in the supplementary material). The electrodesare made of aluminum deposited on quartz substrate; the inter-digitated electrodes are 1.4 μm wide with 1.1 μm spacing betweenthe electrodes.

Before use, the SAW chips are thoroughly cleaned by acetone,isopropyl alcohol and de-ionized water to remove possible con-tamination during chip packaging step. Then the chips are plasma-treated for 30 s at 50 W by PE-50 Plasma Cleaner (Plasma Etch Inc.,NV, USA) to make the electrode surface hydrophilic. The the metalhousing around the electrode chip is about 4 mm (L)�2.5 mm(W)�1 mm (H)¼10 mL, serving as a microchamber for samplesolutions.

The subsequent step is surface functionalization of microelec-trodes for intended detection. For JD experiments, 5 mL MAPantigen sample was loaded into the SAW chip chamber andincubated at room temperature (20–25 1C) for 20 minutes. Forlimit of detection experiments, SAW chips were functionalized byloading the chips with bovine IgG whole molecules at a concen-tration of 10 mg/mL (diluted in Buffer B) and keeping in a humidorfor 10 h. Then the SAW chips were blocked by incubating thefunctionalized chips in 0.1� Buffer B for 30 min at room tem-perature. Afterwards, excessive antigens and blocking moleculesthat were not bound to the electrode surface were flushed away by0.1� PBST (Phosphate Buffered Saline with Tween).

3.3. ACEK impedance measurement procedure

In this work, impedance data were acquired using a high pre-cision impedance analyzer (Agilents 4294A) and the data wererecorded through its LAN port onto a computer using softwareData Transfer V3.0 (SEKONICs). ACEK impedance assay is per-formed by continuously measuring the impedance of the electrodecell with a fixed AC signal for up to 5 min. For the macromoleculesand buffer solutions used here, a 100 kHz AC signal of no higherthan 500 mVrms would be applied onto the electrode cell byAgilent 4294A as the measuring signal. 100 kHz was chosenbecause strong DEP enrichment effect was found at 100 kHz inour past experiments with similar types of macromolecules in0.1� PBS (Li et al., 2013). Measurement for every data point wasrepeated 3–5 times with a new electrode chip each time.

To minimize the effect of impedance inconsistence from oneelectrode chip to another, ACEK impedance sensing employs thenormalized change in the measured capacitance as the detectionmetric to correlate with antibody deposition on the electrodesurface, which is the slope of normalized capacitance as a functionof time and found by least square linear fitting method. Normal-ized capacitance is calculated as Norm(Ct)¼Ct/C0, where Ct and C0are the capacitance values at time t and time zero, respectively.

4. Results and discussion

4.1. Impedimetric characterization of the electrode cell

In order to ensure good electrical connections and properimpedimetric behaviors of the electrodes, prior to and after theACEK binding assay, the impedance spectra of the sensor cell weremeasured from 40 Hz to 6 MHz with an excitation voltage of5 mVrms. Such a voltage level is too low to induce appreciableACEK effect, so that the fluid/electrode system was not disturbedby the spectral scan. The impedance spectra will also allow us to

establish an appropriate equivalent circuit of the electrode cell fordata processing.

Representative impedance spectra of an electrode chip beforeand after binding reaction were presented in Fig. 2. The specificbinding reactions were conducted with JD-positive serum byapplying a 100 kHz AC voltage at 100 mVrms for 60 s. The Bodeplots were curve-fitted from 10 k to 1 MHz by the equivalentcircuit shown as the inset of Fig. 2. The equivalent circuit simplyconsists of a serial connection of Rs�Cint. At low to mediumfrequencies (low MHz), the impedance between the interdigitatedmicroelectrodes, Zsens, can be approximated as a serial connectionof the resistance through electrolyte Rs, and the interfacial reac-tance between microelectrodes and electrolyte, 1=jωCint, i.e.Zsens ¼ Rsþ1=jωCint. Such an equivalent circuit is commonlyadopted (Yang et al., 2004; Luong et al., 2001) and has beenpreviously used in our work (Li et al., 2013) to characterize thesame type of SAW electrode chips in PBS. The curve-fitting agreeswell with the experimental data around the frequency of interest,i.e. 100 kHz. (The Kendall correlation coefficients between themeasured data and curve fitting are better than 0.96 for theimpedance magnitude and 0.83 for the phase angle.) Thus, around100 kHz, the interfacial capacitance of the electrode cell can becalculated straight forward from the impedance magnitude and itsphase angle as Xc¼ 1=ωCint ¼�jZj sin ðθÞ, which is indicative ofmacromolecule deposition at electrode surface.

Also to note, the impedance of the electrode cell at 100 kHz ishighly resistive with a phase angle of��51. As such, most of theAC potential would be applied over the resistive fluid bulk, andabout 8.7% of the potential (8.7 mVrms) would be dropped betweenthe electrode and the electrolyte, which poses very little risk ofcausing electrochemical reactions.

4.2. JD detection by ACEK impedance sensing

First, we conducted proof of concept experiments to demon-strate that our impedimetric assay is capable of distinguishingJD-positive/negative serum samples. A 100 kHz AC signal at100 mVrms was employed as the measuring signal.

The normalized capacitance change rates were plotted as afunction of time in Fig. 3 for the three types of samples. As shown

10 100 1k 10k 100k 1M 10M 100M1k

10k

|Z|()

before bindingafter binding

10 100 1k 10k 100k 1M 10M 100M

-60

-40

-20

0

(deg)

before bindingafter binding

C : 23. 0 nF 21. 3nFRs :1728 1834int

Rs Cint

before binding by curve fittingafter binding by curve fitting

before binding by curve fittingafter binding by curve fitting

Frequency (Hz)

Frequency (Hz)

Fig. 2. Bode plots of the electrode chip withpositive serum before and afterperforming ACEK binding by a 100 kHz AC signal of 100 mVrms for 60 s. Anequivalent circuit as shown in the inset was used to perform curve-fitting of theimpedance from 10 k to 1 MHz, which is plotted alongside the measured data.Top: impedance magnitude; bottom: phase angle.

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443440

in Fig. 3, the capacitances of the JD-negative serum and the controlsample stay rather constant during the test, while that of JD-positive serum capacitance decreases monotonically with timedue to binding reaction. For the positive sample, the capacitancechange rate reaches almost �54.6‰/min, while for disease-negative serum and control samples the capacitance change ratesare �1.64‰/min and �2.62‰/min, respectively. The low capaci-tance change from the negative sample indicates that this testprotocol can minimize the interference arising from nonspecificbinding of antibodies to the electrode surface, and the significantcapacitance change from JD-positive sample is not a result ofelectrochemical reactions and other impedimetric artifacts in themeasurements.

The next set of experiments tests whether ACEK impedancesensing is capable of quantitative measurement, in which JD-positive serum was diluted with 0.1� PBS with a volume ratiofrom 1:20 to 1:200 dilution. The capacitance change rates forvarious dilutions of JD-positive sera are also given in Fig. 3. Asexpected, when the sample became more diluted, the capacitancechange rate decreased from (�54.672.15)‰/min at 1:20 dilutionto (�1.0171.88)‰/min at 1:200 dilution. Highly diluted sera (e.g.1:160 dilution and 1:200 dilution) cannot be differentiated fromthe control sample. For the JD-positive serum samples withdilution from 1:120 to 1:40, the normalized capacitance changerate shows a logarithmic dependence on serum dilution. The

dependence could be expressed as

yð‰=minÞ ¼ �73:717 log ðserum dilutionÞ�157:47 ð3Þ

with Pearson correlation coefficient R2¼0.912. At 1:20 dilution,the capacitance change of �54.6‰/min is lower than thatcalculated from Eq. (3) (�64.2‰/min), which could be attributedto saturation of binding sites at high concentration of antibodies.

The results in Fig. 3 demonstrate that ACEK impedance assay is aquantitative method within a range of antibody concentrations.Because the serum sample tested here is from a JD-positive dairycow, the concentration of antibodies within is unknown. Experimentsto obtain the analytical sensitivity of ACEK impedance assay wereconducted by measuring the binding of goat anti-bovine IgG (HþL)antibody (unlabeled, Jackson ImmunoResearch Laboratories, Inc., WestGrove, PA) to bovine IgG whole molecules (Jackson ImmunoResearchLaboratories, Inc., West Grove, PA) that were immobilized on themicroelectrodes. The experiments again yielded a logarithmic depen-dence on the antibody concentration for the capacitance change rate,which can be expressed as y(‰ per minute)¼�10.058 log(x (ng/mL))þ1.562. The sensitivity is calculated as �10:058‰=ðng=mLÞ, andthe limit of detection is determined to be around 10 ng/mL. Details ofthe experimental procedures and results can be found in the supple-mentary material section.

4.3. Effects of AC electric field

According to Eqs. (1) and (2), the forces on particles by ACEKeffects increase with electric field strength. If the short “sample toresults” time is a result of electrokinetic enrichment, then thecapacitance change rate should exhibit a monotonically increasingdependence on the applied voltage amplitude. Therefore, ourexperiments tested different amplitudes of 100 kHz AC signalsfor this impedance assay. The experiments were repeated threetimes at each voltage amplitude.

The capacitance change rates were obtained and plotted in Fig. 4(a) as a function of voltage amplitude from 5 mVrms to 500 mVrms.As shown in Fig. 4(a), with JD-positive sample, the capacitancechange rate did increase monotonically with the magnitude of ACsignals. The change rates from the positive sample are significantlyhigher than those from the control samples except for the results at5 mVrms, while the change rate remained consistently low(o73:73‰ =min) for the control experiments (0.1� PBS) duringthe same voltage range. The voltage of 5 mVrms is too low to induceany ACEK effects to enhance antigen–antibody binding. With5 mVrms applied, the capacitance change rate was merely(�5.1770.784)‰/min, only slightly higher than that of the controlsamples, (�3.7372.08)‰/min. Such a small difference between

0 50 100 150 200

0.80

0.85

0.90

0.95

1.00

100mV@100 kHz

Normalizedcapacitance

Time (s)

P-1:20P-1:40P-1:80P-1:120P-1:160P-1:200N-1:20control buffer

/min

/min

/min-54.6

-34.2

-22.9

Fig. 3. Transient behavior of normalized capacitances from microelectrode arraysfunctionalized with JD specific antigen and blocking agents. The samples werevarious dilutions of disease positive and negative sera and 0.1x PBS as the controlbuffer. An AC potential of 100 mVrms at100 kHz was applied.

1k 10k 100k 1M 10M0

-10

-20

-30

-40

-50

-60

positive serum control buffer

dC/C

per

min

ute

()

frequency (Hz)

100 mV

0 100 200 300 400 500

0.0

-20.0

-40.0

-60.0

-80.0 positive serum control buffer

dC/C

per

min

ute

()

voltage (mV)

100 kHz

Fig. 4. Capacitance change rates of JD-positive serum sample and control buffer (0.1� PBS) when they are tested at (a) 100 kHz AC signals of different AC voltage levels,and (b) 100 mVrms AC signals at different frequencies.

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443 441

JD-positive sample and the control samples would not warrant areliable detection of antibody–antigen binding.

The capacitance change rate increased rapidly when theapplied voltage increased from 5 mVrms to 100 mVrms. The changerate of capacitance at 50 mVrms is (�29.170.41)‰/min, while100 Vrms yielded a change rate of (�54.672.15 )‰/min. Theincrease in capacitance change rate slowed down when thevoltage rose above 100 mVrms, which could be caused by possibleadverse effect that electric field may have on the activities oradhesion of antibodies on the electrodes. Nevertheless, theincrease of capacitance change rate with voltage amplitude servesto corroborate the contribution of ACEK effect in this capacitiveassay. The results in Fig. 4(a) also let us to set 100 mVrms as theoptimal operating voltage of the ACEK impedance assay. Addition-ally, the control solution consistently exhibited low capacitancechange for the voltage range of 100�500 mVrms at 100 kHz,indicating that the capacitive response from the JD-positivesample was not caused by electrochemical reactions or otherimpedimetric artifacts in the measurements.

Further, the ACEK impedance assay is expected to exhibitfrequency dependence. According to Eq. (1), DEP force is frequencydependent. Consequently, there would exist a specific frequency atwhich the target macromolecules will experience maximal attrac-tion force to the electrodes (i.e. positive DEP). The accompanyingACET flows (by Eq. (2)) may add to the particle movement causedby DEP, but will not affect its frequency properties, since ACETeffect is independent of AC frequency (Green et al., 2001; Honget al., 2011). As a result, the DEP property of the antibody willdetermine the frequency dependence of antigen–antibody bindingand further affect the magnitude of capacitance change.

Therefore, to optimize the DEP enhanced capacitive assay, it isnecessary to investigate the frequency dependence of antibodycapture. Capacitance change rates for the JD-positive serumsample were measured at various AC frequencies from 5 kHz to5 MHz at 100 mVrms. The capacitance change rates at differentfrequencies are given in Fig. 4(b). The capacitance change rateshowed obvious frequency dependence. The change rate was thelargest between 100 kHz and 300 kHz, and quickly dropped whenAC frequency deviated from this frequency range. The bell-shapedfrequency dependence clearly indicates DEP characteristics. WhenAC frequency was outside the optimal range, the capacitancechange rates were still much higher than those of control sample(in contrast to the results of 5 mVrms assay), which could beattributed to ACET microflows.

4.4. Reliability

Repeatability is an essential figure of merit in a disposablesensing system. Therefore, we also investigated the responses that

a same serum sample would produce on different electrode chips.A JD-positive sample was tested on ten different chips and thecoefficient of variance (CV) was calculated to be 11.66%, showinggood repeatability. It should be noted that the mycobacteriaantigen used in our study was immobilized on the electrodes bypassive absorption; the repeatability is expected to improve withthe use of chemical cross-linkers.

In order to examine the accuracy of the ACEK capacitiveimmunosensor developed in this work, blind tests were performedusing 10 JD-negative and 10 JD-positive samples. The examinerwas not informed of identities of the 20 JD samples before testing.Fig. 5 shows the result of the blind tests, in which JD-negative andJD-positive samples can be clearly differentiated based on theircapacitance changes. From Fig. 5, it can be derived that a cut-offvale of �10‰/min would safely distinguish JD-negative andpositive samples. JD-negative serum samples consistently exhibita capacitance change no more than 76‰/min.

5. Conclusions

This work presents a rapid immunosensing method that holdshigh promise for POC disease diagnosis. The detection is realizedthrough a modified capacitive sensing technique with a micro-electrode array. The capacitive immunoassay procedure presentedhere can achieve ACEK enrichment of macromolecules and label-free detection of deposition on the microelectrode surface simul-taneously. The experiments proved that DEP and other ACEK effectcould speed up the binding process to achieve rapid and sensitivedetection. Using bovine serum samples, our experiments showedthat 2 min is sufficient to differentiate disease positive andnegative samples, which is much faster than conventional meth-ods using a static fluid chamber. The ACEK enhanced capacitiveimmunosensor also demonstrated a high sensitivity with a limit ofdetection reaching 10 ng/mL. In summary, the ACEK enhancedcapacitive sensing method has a shorter assay time, reducedreagent and device cost, simpler device design and operation thanconventional diagnostic system. With some minor adjustment,this technology is expected to be able to detect antibody–antigenbinding for diagnosis of other infectious diseases.

Acknowledgment

This study was supported by the U.S. National Science Founda-tion under Grant no. ECS-0448896, and The University of Tennes-see Research Foundation Maturation Fund. S. Li gratefullyacknowledge financial support from China Scholarship Council(No. 2010612182) and China′s Natural Science Foundation underGrant no. 51075087.

Appendix A. Supporting information

Supplementary data associated with this article can be found inthe online version at http://dx.doi.org/10.1016/j.bios.2013.08.016

References

Berggren, C., Johansson, G., 1997. Analytical Chemistry 69, 3651–3657.Berggren, C., Bjarnason, B., Johansson, G., 2001. Electroanalysis 13, 173–180.Betty, C., Lal, R., Sharma, D., Yakhmi, J., Mittal, J., 2004. Sensors and Actuators B:

Chemical 97, 334–343.Birdi, K., 2008. Handbook of Surface and Colloid Chemistry. CRC.Cosnier, S., 2005. Electroanalysis 17, 1701–1715.Daniels, J.S., Pourmand, N., 2007. Electroanalysis 19, 1239–1257.Diem, P., Kalt, L., Haueter, U., Krinelke, L., Fajfr, R., Reihl, B., Beyer, U., 2004. Diabetes

Technology & Therapeutics 6, 790–799.

Fig. 5. Capacitance change of blind Johne′s disease serum (1:20). MAP antigen(1:80) incubation overnight, 0.1� buffer B block for 30 min and 100 mVrms at100 kHz AC field is applied for 1 min.

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443442

Eda, S., Bannantine, J.P., Waters, W., Mori, Y., Whitlock, R.H., Scott, M.C., Speer, C.,2006. Clinical and Vaccine Immunology 13, 837–844.

Fan, X., White, I.M., 2011. Nature Photonics 5, 591–597.Feldman, H.C., Sigurdson, M., Meinhart, C.D., 2007. Lab on a Chip 7, 1553–1559.Green, N.G., Ramos, A., González, A., Castellanos, A., Morgan, H., 2001. Journal of

Electrostatics 53, 71–87.Guan, J.G., Miao, Y.Q., Zhang, Q.J., 2004. Journal of Bioscience and Bioengineering 97,

219–226.Hedström, M., Galaev, I.Y., Mattiasson, B., 2005. Biosensors and Bioelectronics 21,

41–48.Hong, F., Cao, J., Cheng, P., 2011. International Communications in Heat and Mass

Transfer 38, 275–279.Klein, M., Kates, R., Chucholowski, N., Scmitt, M., Lyden, C., 1995. Sensors and

Actuators B: Chemical 27, 474–476.Labib, M., Hedström, M., Amin, M., Mattiasson, B., 2009. Biotechnology and

Bioengineering 104, 312–320.Labib, M., Hedström, M., Amin, M., Mattiasson, B., 2010. Analytica Chimica Acta 659,

194–200.Lequin, R.M., 2005. Clinical Chemistry 51, 2415–2418.Li, S., Yuan, Q., Morshed, B., Ke, C., Wu, J., Jiang, H., 2013. Biosensors and

Bioelectronics 41, 649–655.Lian, M., Islam, N., Wu, J., 2007. Nanobiotechnology, IET 1, 36–43.Liu, X., Yang, K., Wadhwa, A., Eda, S., Li, S., Wu, J., 2011. Sensors and Actuators A:

Physical.Luong, J.H.T., Habibi-Rezaei, M., Meghrous, J., Xiao, C., Keith, B., Kamen, A., 2001.

Analytical Chemistry 73, 1844–1848.Morgan, H., Hughes, M.P., Green, N.G., 1999. Biophysical Journal 77, 516–525.

Nakamura, R.M., Kasahara, Y., Rechnitz, G.A., 1992. American Society for Micro-biology, 1992. Immunochemical Assays and Biosensor Technology for the1990s. American Society for Microbiology, Washington, D.C.

Qureshi, A., Gurbuz, Y., Kallempudi, S., Niazi, J.H., 2010. Physical Chemistry ChemicalPhysics 12, 9176–9182.

Sculley, M.J., et al., 1980. Archives of Biochemistry and Biophysics 201 (1), 339–346.Sheehan, P.E., Whitman, L.J., 2005. Nano Letters 5, 803–807.Smith, J.E., Sapsford, K.E., Tan, W., Ligler, F.S., 2011. Analytical Biochemistry 410,

124–132.Stabel, J., 1998. Journal of Dairy Science 81, 283–288.Stanke, S., Bier, F.F., Hölzel, R., 2011. Electrophoresis.Tennico, Y.H., Remcho, V.T., 2010. Electrophoresis 31, 2548–2557.Urdea, M., Penny, L.A., Olmsted, S.S., Giovanni, M.Y., Kaspar, P., Shepherd, A., Wilson,

P., Dahl, C.A., Buchsbaum, S., Moeller, G., 2006. Nature 444, 73–79.Velasco, V., Work, A.H., Williams, S.J., 2012. Electrophoresis 33, 1931–1937.Wang, Y., Ye, Z., Ying, Y., 2012. Sensors 12, 3449–3471.Wright, P., Nilsson, E., Van Rooij, E., Lelenta, M., Jeggo, M., 1993. Revue scientifique

et technique (International Office of Epizootics) 12, 435.Wu, J., 2010. Microfluidic Devices in Nanotechnology, 295–316.Xiang, Q., Hu, G., Gao, Y., Li, D., 2006. Biosensors and Bioelectronics 21, 2006–2009.Yang, K., Wu, J., 2010. Biomicrofluidics 4, 034106.Yang, K., Liu, X., Yuan, Q., Wu, J., Wadhwa, A, Eda, S., 2011. 6th International

Conference on Microtechnologies in Medicine and Biology (MMB 2011),Lucerne, Switzerland.

Yang, L., Li, Y., Griffis, C.L., Johnson, M.G., 2004. Biosensors and Bioelectronics 19,1139–1147.

S. Li et al. / Biosensors and Bioelectronics 51 (2014) 437–443 443