Cs

Ca

b

c

a

ARRAA

KEUcCaPCm

1

tcdmr[ftfc

h0

Sensors and Actuators B 212 (2015) 544–550

Contents lists available at ScienceDirect

Sensors and Actuators B: Chemical

jo u r nal homep age: www.elsev ier .com/ locate /snb

onductometric analysis in bio-applications: A universal impedancepectroscopy-based approach using modified electrodes

hiara Canali a, Layla Bashir Larsena, Ørjan Grøttem Martinsenb,c, Arto Heiskanena,∗

Department of Micro- and Nanotechnology, Technical University of Denmark, Produktionstorvet 423, 2800 Kgs Lyngby, DenmarkDepartment of Physics, University of Oslo, Sem Sælands vei 24, Fysikkbygningen, 0371 Oslo, NorwayDepartment of Biomedical and Clinical Engineering, Oslo University Hospital, 0372 Oslo, Norway

r t i c l e i n f o

rticle history:eceived 19 December 2014eceived in revised form 5 February 2015ccepted 7 February 2015vailable online 16 February 2015

eywords:lectrochemical impedance spectroscopyniversal protocol for impedance-basedonductivity determinationustomised Matlab-based algorithm forutomated spectral analysisrotein-repellent electrode modificationonductivity measurements in cell cultureedium

a b s t r a c t

We present a universal protocol for quick and reproducible conductivity determinations in bio-applications using electrochemical impedance spectroscopy (EIS), electrode modification and automatedspectral analysis. Two-terminal EIS measurements may be acquired using any standard impedance anal-yser adjusting the applied sinusoidal potential and frequency range for spectral analysis. An implementedMatlab algorithm displays the acquired spectra, automatically identifies the frequency at which the phaseangle (ϕ) is closest to 0◦ and determines the impedance magnitude, i.e. the solution resistance (RS). Thecorresponding conductivity value is immediately calculated as the ratio of the conductivity cell constant(�), determined based on calibration, and RS. This protocol eliminates the need for evaluating a spe-cific equivalent circuit followed by non-linear regression based curve fitting that is generally required inEIS-based conductivity determinations. The protocol is applicable to conductivity determinations usingdifferent conductivity cell configurations in any electrolyte solution regardless of its composition, i.e. insolutions with or without electroactive species that give rise to faradaic interface impedance. Conductedmeasurements showed high reproducibility in good agreement with a commercial conductometer in awide range of ionic strengths up to five times that of physiological PBS. Since measurements in cell cul-

ture medium with bare gold electrodes indicated the need for recalibration to counteract the effect ofbiomolecule physisorption, the validity of the protocol was further extended using a protein-repellentcoating of poly(ethylene glycol) methyl ether thiol self-assembled monolayer. This effectively elimi-nated electrode fouling, facilitating high reproducibility in repeated conductivity determinations in thepresence of proteins.

© 2015 Elsevier B.V. All rights reserved.

. Introduction

Measurement of solution conductivity is a classical analyticalechnique that finds application in a wide variety of chemi-al and biochemical studies, e.g., evaluation of solvent purity,etermination of relative ionic strength in solutions, assess-ent of critical micelle concentration, monitoring of enzymatic

eactions and calculation of basic thermodynamic quantities1], as well as biosensing [2,3]. Although bio-applications mostrequently involve measurements in the physiological conduc-

ivity range [4,5], low conductivity solutions have been usedor, e.g., impedance-based affinity biosensing [6,7], while highonductivity solutions are used as supporting electrolytes in

∗ Corresponding author. Tel.: +45 45 25 68 39; fax: +45 45 88 77 62.E-mail address: arto.heiskanen@nanotech.dtu.dk (A. Heiskanen).

ttp://dx.doi.org/10.1016/j.snb.2015.02.029925-4005/© 2015 Elsevier B.V. All rights reserved.

voltammetric analysis, where concentrations may range up to 3 M[8,9].

Solution conductivity can be explained in terms of ionicmobility, which is directly proportional to temperature with animmediate effect of typically 1–3% per ◦C with respect to tabu-lated values at 25 ◦C. It is also influenced by shifts in pH; acidic orbasic solutions increase conductivity since hydrated protons andhydroxyl ions are the most mobile cations and anions, respec-tively. Variation of conductivity over time may also convey ahigh degree of information about the chemical dynamics of bio-logical processes. A culture medium containing proliferating ordying cells is an extremely heterogeneous chemical environment,the composition of which is continuously changing depending

on cellular metabolism [10]. Conductivity of culture media wascorrelated to microbial concentration and fermentation activityalready in 1898 [11], paving the way to impedance microbi-ology [12], which gained increased significance in the 1970s

Actua

wavmaSmpcgepmctwcgvfp

asrnoiwcictoeaembfa

citkcGttsor[tco

uoifvflap

C. Canali et al. / Sensors and

ith the emergence of frequency response analysers. However,lthough electrical impedance spectroscopy (EIS) potentially pro-ides extensive information about electron transfer mechanisms,ass transfer phenomena and biochemical activity, the design of

n optimised analytical strategy may be highly demanding [13].implified approaches have been devised based on conductivityeasurements. Parsons and Sturges developed and tested a sim-

le and relatively quick approach to correlate culture mediumonductivity to the total amount of ammonia and amino nitro-en produced by putrefactive anaerobic bacteria [14,15]. Allisont al. extended this technique to metabolic studies on aerobesroducing ammonia and lactic acid [10]. Similarly, cell cultureedium conductivity may be a key parameter for analysing the

omplex dynamics that modulate mammalian cell cultures overime [16]. This is particularly relevant for biomedical researchhich is currently undergoing a shift of paradigm toward 3D cell

ultures, needing new non-invasive on-line monitoring technolo-ies [17,18]. Furthermore, EIS has already been shown to be aaluable tool for defining the porosity of 3D cell culture scaf-olds by yielding the conductivity of the bulk electrolyte filling theores [19,20].

The simplest approach for conductivity determination is topply an alternating electric field between two electrodes and mea-ure the impedance magnitude as an estimation of the solutionesistance (RS), which can be used to calculate conductivity. Alter-ating current at frequencies above 1 kHz should be used insteadf direct current that may lead to electrolysis and electrode polar-zation [21]. Commercial conductometers are standalone devices

hich typically operate maximally at few predetermined frequen-ies. During calibration using a conductivity standard solution, annternal algorithm selects the frequency value depending on theonductivity range to give an accurate estimation of RS. However,he algorithm may vary between manufacturers and dependingn the instrument quality, increasing the possibility of systematicrrors in the analysis. Moreover, although commercially avail-ble conductivity sensors are manufactured from robust metals,.g., platinum, stainless steel and titanium, that provide highechanical stability, they are not designed for measurements in

iological solutions. Biomolecule adsorption on the electrode sur-ace increases the measured impedance, hence, yielding a lowerpparent conductivity value.

EIS-based conductivity measurements, relying on analysis ofomplete spectra using equivalent circuits, have been proposedn order to eliminate the need of choosing a specific frequencyhat may result in measurement inaccuracy [22]. To our bestnowledge, such approach has been used especially for lab-on-a-hip devices in impedance microbiological applications [23–25].enerally, such analysis depends on the electrolyte composi-

ion and hence, whether the measured impedance is contributedo by faradaic and/or non-faradaic processes. Additionally, thetructure/geometry of a conductivity cell imposes requirementsn the used equivalent circuit [22], especially when microfab-icated devices having interdigitated electrodes (IDEs) are used26]. Hence, this approach complicates conductivity determina-ions, requiring a different equivalent circuit depending on theomposition of the used electrolyte solution [13,27] or structuref the conductivity cell [22,26].

To avoid the limitations described above, we hereby present aniversal protocol that relies on acquisition and automated analysisf complete impedance spectra as the basis for quick conductiv-ty determinations, as well as electrode modification applicableor measurements in biomolecule containing electrolytes. Con-

entional two-terminal EIS measurements can be used with theexibility of choosing both the sinusoidal potential of excitationnd frequency range for analysis. A Matlab-based algorithm dis-lays Bode plots for impedance magnitude (|Z|) and phase angle

tors B 212 (2015) 544–550 545

(ϕ) while it automatically identifies the most suitable frequencyfor determining RS, i.e. |Z| at the frequency where ϕ is closest to 0◦.This approach provides two major advantages: the calculated solu-tion conductivity value is (1) based on an accurately determined RSunlike in the case of an algorithm relying on a few predeterminedfrequencies, and (2) independent of validation and analysis of aspecific equivalent circuit. A commercial conductometer was usedfor validation and a good agreement between the two methodswas found for ionic strength (I) values up to five times that ofphysiological PBS. Our universal protocol facilitated stable, preciseand accurate measurements using inexpensive electrodes and astandard impedance analyser. Moreover, the validity was furtherextended to conductivity measurements in biomolecule contain-ing electrolytes (i.e. cell culture medium). The influence of proteinphysisorption on conductivity determination was eliminated usinga protein-repellent self-assembled monolayer (SAM) modificationof poly(ethylene glycol)-terminated alkanethiol. High precisionand reproducibility was achieved in repeated measurements in cellculture medium without the requirement for recalibration duringmeasurements or labour-intensive electrode cleaning after mea-surements.

2. Materials and methods

2.1. Chemicals and solutions

Potassium hydroxide (semiconductor grade), hydrogen perox-ide (30% solution in water), cell culture tested PBS (physiologicaland 10× concentrate), Roswell Park Memorial Institute medium1640 (RPMI), fetal bovine serum (FBS), penicillin/streptomycin(P/S), and poly(ethylene glycol) methyl ether thiol (mPEG, aver-age Mn 800) were purchased from Sigma–Aldrich Corporation(St. Louis, MO, USA). Electrochemical measurements were per-formed on serial dilutions of PBS in the range of 0–1.5 M NaCl(chemical composition of the 10× concentrate given by the sup-plier: 0.03 M Na(PO4)3, 1.5 M NaCl, 0.0105 M KH2PO4). Ultrapuredeionized water (18.2 M� cm) from Milli-Q system (Millipore Cor-poration, Billerica, MA, USA) was used for diluting PBS samplesand rinsing electrodes. Standard solutions of known conductivity(8.4 × 10−3, 1.4 × 10−2, 1.3, 11.2 S/m) were purchased from HannaInstruments (Kungsbacka, Sweden).

2.2. Measurement protocol

Prior to the first EIS measurement, gold plate electrodes,fabricated by e-beam evaporation on oxidized silicon wafers as pre-viously described [18,20], were cleaned by a 10-min treatment witha mixture of 25% (v/v) H2O2 and 50 mM KOH followed by a potentialsweep from −200 mV to −1200 mV in 50 mM KOH [27]. Electrodeswere mounted on a 3D printed conductivity cell as described inSupplementary Information S1 and shown in Fig. S-1. Impedancespectra were acquired using a Reference 600 potentiostat (GamryInstruments, Warminster, PA, USA). An alternating potential (rms)of 10 mV was applied in the frequency range between 10 Hz and1 MHz. The geometrically determined cell constant (�) was verifiedby acquiring and analysing three EIS spectra in triplicate samplesof each conductivity standard solution. This allowed calculation ofthe experimental value for �, which is defined as the product ofthe solution conductivity and solution resistance (RS). The resultsare presented as average ± standard error of mean (s.e.m.), n = 9. EISdata processing for automated generation of Bode plots and con-

ductivity determination were implemented as a Matlab (versionR2012b) script (detailed description in Supplementary InformationS2). For-loops allowed presentation of spectra acquired in differ-ent solutions in the same Bode plot (|Z| or ϕ). To calculate the

5 Actuators B 212 (2015) 544–550

cqtsldcTmt

2

cmupTmpwcCmw

3

3

aN((acR�iitcb

iddibTaeiiotts(tggi

Fig. 1. Averaged EIS spectra (±s.e.m., n = 9) for PBS dilutions ranging from very low

46 C. Canali et al. / Sensors and

onductivity, RS was determined based on the |Z| value in a fre-uency range where ϕ is closest to 0◦. This was computed usinghe min(X,Y) function. Three EIS spectra were acquired in triplicateamples of different PBS dilutions. The relationship between calcu-ated conductivity and I (approximated by NaCl concentration) wasetermined and validated against the performance of a commer-ial conductometer (CDM 92, Radiometer Analytical, Copenhagen).he results are presented as average ± s.e.m., n = 9. All the measure-ents were conducted at room temperature in solutions exposed

o atmospheric CO2.

.3. Electrode modification for bio-applications

Three EIS spectra were acquired in triplicate samples of cellulture medium. Before and after measurements in cell cultureedium, three samples of physiological PBS was analysed to eval-

ate the influence of medium serum proteins on the electrodeerformance. The results are presented as average ± s.e.m., n = 9.hen, three sets of electrodes were used for measurements afterodification with 10 mM mPEG in water for 16 h. The results are

resented as average ± s.e.m., n = 27. Impedance measurementsere performed as described in Section 2.2. CO2 level in cell

ulture medium samples was equilibrated in an incubator (5%O2/95% air) for 30 min prior to measurements. Between measure-ents in different samples, electrodes were rinsed using Milli-Qater.

. Results and discussion

.1. Analysis of PBS dilutions

As first validation of our universal protocol, EIS spectra werecquired in samples of different PBS dilutions (range of 0–1.5 MaCl) and analysed using the custom-made Matlab algorithm

Supplementary Information S2) that presents data as Bode plotsFig. 1a, b) and determines RS in a frequency range where ϕ ispproximately 0◦ (Fig. 1b). The corresponding conductivity was cal-ulated as the ratio between the conductivity cell constant (�) andS. The calculations were based on the experimentally determined, 1.77 ± 0.02 cm−1 (average ± s.e.m., n = 9), obtained by calibration

n standard solutions with known conductivity. The geometric def-nition of � is the ratio between the distance of the electrodes andhe area exposed to the solution. Based on these factors, � wasalculated to be 1.85 cm−1 (Fig. S1b), indicating a good agreementetween the experimental and geometric value.

The relationship between the determined conductivity and theonic strength (I) of the PBS dilutions showed very high repro-ucibility, the overall relative standard error (RSE%) for all theilutions being only 0.5%. The value also comprises measurements

n solutions with very low conductivity, which are more prone toe affected by artifacts due to parasitic conduction paths [28,29].his means that in solutions with higher conductivity, such as thoseround physiological I, the obtained variation in measurements wasven lower than the above mentioned overall RSE%. Correspond-ng measurements using the commercial conductometer resultedn an overall RSE% of 3.0%, which is clearly higher than the valuebtained with our universal protocol. Further comparison betweenhe methods is shown in Fig. 2. The linear correlation betweenhe determined conductivity and calculated I of the PBS dilutionshown in Fig. 2a indicates that the coefficient of determinationR2) for our universal protocol was slightly higher (0.998) than

he one obtained for the commercial instrument (R2 = 0.991). Theeneral observation based on Fig. 2a is that both approaches showood agreement in conductivity determination for solutions reach-ng five times the I value of physiological PBS (I = 8.12 × 10−1 M).

to high conductivity: Bode plots for (a) impedance magnitude (|Z|) and (b) phaseangle (ϕ).

However, for highly concentrated electrolytes (10 times the I ofphysiological PBS), the commercial conductometer deviated signif-icantly from the value obtained using our universal protocol, whichretained linearity throughout the used I range. Performed regres-sion analysis (Fig. 2b) indicated a very good agreement (R2 = 0.999)until I five times that of physiological PBS. These results showedthat the established universal protocol provides stable, accurateand reproducible measurements in a wide conductivity range.

The impedance spectra acquired in the PBS dilutions werealso used to define four frequency ranges applicable forquicker determination of conductivity with single frequencyanalysis: (1) 1–4 kHz for very low conductivity solutions(I ≤ 1.62 × 10−3 M); (2) 20–50 kHz for low conductivity solutions(1.62 × 10−3 < I ≤ 8.12 × 10−2 M); (3) 200–250 kHz for solutionsclose to the physiological I (8.12 × 10−2 < I ≤ 1.62 × 10−1 M);(4) 300–400 kHz for solutions above the physiological I(I > 1.62 × 10−1 M).

In the above described conductivity determinations based onEIS measurements, the applied sinusoidal potential was 10 mVrms

and the used electrodes had large area with a deposited thin goldfilm. When using an impedance analyser, the amplitude of the

sinusoidal potential may be freely chosen as long as it ensureslinearity of the current–voltage response. However, despite thechosen potential amplitude, the measured impedance, and hence

C. Canali et al. / Sensors and Actua

Fig. 2. Comparison of the universal conductivity determination protocol with acNa

tdmva1ShtqoSccmeo

3

temoe

3.2.1. Effect of biological solutions on bare gold electrodesThe impedance spectra presented in Fig. 3a indicate that the

spectral behaviour is in accordance with the one previously shown

ommercial conductometer: (a) trend of data point distribution relating increasingaCl concentration (M) in PBS dilutions to the corresponding conductivity (S/m)nd (b) regression analysis. The error bars represent s.e.m. (n = 9).

he determined conductivity, is the same provided that the setupoes not have a high intrinsic impedance, which could decreaseeasurement accuracy at low applied potential amplitudes. The

alidity of our proposed universal protocol was additionally evalu-ted using other potential amplitudes (25 and 50 mVrms) aside from0 mVrms. The results are shown in Supplementary Information3. For EIS measurements using IDEs, different equivalent circuitsave been proposed in literature; however, a common feature ishe presence of a parallel geometric capacitance [26]. The conse-uence is that in a frequency range below 1 MHz the absolute valuef ϕ may significantly deviate from 0◦. Supplementary Information4 shows conductivity determinations using IDEs applied in twoonfigurations. The results demonstrate that our universal proto-ol provides possibility for reliable conductivity determinations ininiaturised devices where the conductivity cell significantly influ-

nces the characteristic equivalent circuit, making it different fromne that is valid for parallel plate electrodes.

.2. Electrode modification for bio-applications

To further evaluate the validity of the presented universal pro-ocol for conductivity measurements in biomolecule-containing

lectrolytes, we determined the conductivity of cell cultureedium containing serum proteins. Physisorption of biomolecules

n bare gold surfaces is well-known. To easily see the effect oflectrode fouling, we determined the conductivity of physiological

tors B 212 (2015) 544–550 547

PBS before and after measurements in cell culture medium. Theperformed conductivity measurements clearly indicated that theused electrode surfaces were fouled by physisorbed biomolecules,requiring intensive electrode cleaning and recalibration aftereach measurement. Therefore, we optimised a protein-repellentelectrode modification using self-assembled monolayer (SAM)of a poly(ethylene glycol)-terminated alkanethiol. This approacheliminated the need for labour-intensive electrode cleaning andrecalibration that are crucial when using a commercial instrument.

Fig. 3. Averaged EIS spectra (±s.e.m.) for physiological PBS and cell culture mediumusing (a) bare gold electrodes (n = 9) and (b) mPEG-modified gold electrodes (n = 27).Measurements in PBS were performed before and after analysis of cell culturemedium.

548 C. Canali et al. / Sensors and Actua

Table 1Influence of biomolecule-containing electrolytes on conductivity measurements:conductivity [S/m] values for physiological PBS and cell culture medium determinedusing bare and mPEG-modified gold electrodes.

Type of electrolyte Conductivity [S/m]

Bare gold electrodesa

PBS before measurements in cell culture medium 1.66 ± 0.01PBS after measurements in cell culture medium 1.12 ± 0.01PBS after measurements in cell culture medium

and recalibration1.49 ± 0.01

Cell culture medium (RPMI) 1.55 ± 0.01

mPEG-modified gold electrodesb

PBS before measurements in cell culture medium 1.596 ± 0.007PBS after measurements in cell culture medium 1.603 ± 0.007Cell culture medium (RPMI) 1.44 ± 0.01

t

ftadplfcrivwelcucsrcwscoaodiEqptso

3e

aat3sbctep

a Conductivity values are reported as average ± s.e.m., n = 9.b Conductivity values are reported as average ± s.e.m., n = 27. Three sets of elec-

rodes were evaluated.

or PBS dilutions (Fig. 1). The decrease in conductivity correspondso a slight upward shift in the Bode plots for |Z| at frequenciesbove 200 Hz and a clear separation above 5 kHz. Values of con-uctivity were calculated in the frequency range proposed forerforming measurements on solutions with I close to the physio-

ogical range, further confirming that the protocol is fully suitableor single frequency analysis. Table 1 summarises the determinedonductivity values for PBS and cell culture medium. The obtainedesults clearly show that measurements in cell culture mediumnfluenced the behaviour of bare gold electrodes. The conductivityalue determined for PBS before measurements in medium agreedell with previously shown results [30], whereas after electrode

xposure to biomolecules the determined conductivity was clearlyower. Moreover, despite the small distribution of the determinedonductivity values (s.e.m.) for cell culture medium, the individ-al measurements showed a clear trend in terms of decreasingonductivity, which indicates that even during a series of mea-urements in a biological solution the individual measurementesults may be influenced by electrode fouling. When performingonductivity measurements using a commercial conductometer,hich has fixed electrodes, calibration using conductivity standard

olutions may alleviate the effect of exposure to biomolecule-ontaining electrolytes. To evaluate this possibility, we recalibratedur conductivity cell after measurements in cell culture mediumnd repeated conductivity determination of physiological PBS. Thebtained result (Table 1) indicates that, despite recalibration, theetermined conductivity was still lower than obtained in the orig-

nal measurements prior to any exposure to cell culture medium.xtensive cleaning might improve electrode behaviour in subse-uent measurements; however, it would easily deteriorate theerformance of microfabricated thin-film electrodes, shorteningheir lifetime. This encourages the use of modified electrodes with atable protein-repellent SAM so that they can be applied to a seriesf measurements without the influence of electrode fouling.

.2.2. Effect of biological solutions on mPEG-modified goldlectrodes

Analogously to chemical electrode modifications, which behaves capacitors in series with the double layer capacitance [27],dsorbed biomolecules also change the capacitive behaviour ofhe electrode–electrolyte interface. Results described in Section.2.1 clearly indicate that |Z| increases as a consequence of mea-urements in biomolecule-containing electrolytes when usingare gold electrodes. To eliminate this effect and be able to

onduct EIS measurements without the need for extensive elec-rode cleaning and/or recalibration after each measurement, wevaluated the possibility of using electrodes modified with arotein-repellent coating. Polyethylene glycol (PEG) is a nontoxic

tors B 212 (2015) 544–550

and non-immunogenic polymer which is able to decrease attractiveforces between solid surfaces and proteins. Harder et al. showedhow this ability is related to PEG molecular conformation andbecomes most effective when PEG chain length allows the forma-tion of dense and predominantly helical films of PEG [31]. Theyspeculated that the helical conformation may correlate with thedegree of solvation and, consequently, with the stability of theinterfacial layer of water, which may repel proteins reaching theelectrode surface by diffusion. Hence, PEG-terminated alkanethiolSAMs on gold surfaces have been successfully used to facilitateselective cell patterning and reduce surface fouling [32].

To readily modify gold surfaces and increase the biocompati-bility, we used water soluble mPEG-terminated alkanethiol. Initialtests followed by EIS characterisation indicated that a modificationtime of at least 16 h yielded a robust well-behaved functionaliza-tion (data not shown). Three sets of electrodes were calibrated aftermPEG modification and � was calculated to be (1.52 ± 0.01) cm−1

(average ± s.e.m., n = 9). Based on the geometric factors, � resulted1.85 cm−1. The larger difference between the calculated and thegeometrically determined � is due to the fact that although� is generally referred to as the ratio between electrode dis-tance and area, calibration takes into account the condition ofthe electrode–electrolyte interface, which is different for chemi-cally modified electrodes in comparison with non-modified ones.Fig. 3b shows a set of averaged impedance spectra for PBS andcell culture medium acquired using three sets of modified elec-trodes (three repetitive spectra acquired for each electrode setin three electrolyte samples). Each spectrum is reported as aver-age ± s.e.m. (n = 27). In the frequency range 200–250 kHz, where ϕis approximately 0◦, the spectra acquired in PBS before and aftermeasurements in cell culture medium completely overlap, whichmeans that the mPEG modification was able to protect the electrodesurface from the influence of physisorbed biomolecules. The spec-tra acquired in PBS indicate that the measurements in cell culturemedium only had a slight influence on the capacitive behaviour ofthe electrodes as seen in the shift of ϕ at frequencies below 100 Hz.The conductivity values for PBS are shown in Table 1 and agree wellwith values reported in literature [30]. Moreover, the values shownin Table 1 also indicate that, on virtue of the electrode modification,the conducted measurements are highly reproducible between setsof electrodes.

4. Conclusions

We established and verified a universal protocol for quickconductivity determinations using electrochemical impedancespectroscopy (EIS) and electrode modification for bio-applicationswith high reproducibility. The approach relies on simple spectralanalysis to determine the frequency at which the phase angle(ϕ) is closest to 0◦. The impedance magnitude (|Z|) at that fre-quency corresponds to the solution resistance (RS), which can beused to calculate the conductivity. Unlike commercial instruments,typically operating at a few predetermined frequencies, which,depending on the solution composition, may not yield an accu-rate estimation of RS, our universal protocol, based on analysisof complete spectra, has the advantage that RS, and consequentlythe corresponding conductivity, can be determined with certaintyindependent of the solution composition. A Matlab-based algo-rithm was implemented to display impedance spectra as Bode plotsand automatically identify the suitable frequency for determiningRS and calculating solution conductivity. The universal protocolgave results in good agreement with a commercial conductome-

ter in a wide range of electrolyte solutions for ionic strength upto five times that of physiological PBS. The validity of the pro-tocol was extended to bio-applications by applying an optimisedelectrode modification method using a protein-repellent coating

Actua

oaonpwiaoad

A

EptF

A

t

R

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

C. Canali et al. / Sensors and

f poly(ethylene glycol) methyl ether thiol. This allowed precisend reproducible conductivity determinations without influencef electrode fouling, which clearly affected measurements withon-modified electrodes. Our universal protocol combined withrotein-repellent electrode modification provides a fast and simpleay of performing programmed conductivity monitoring using any

mpedance analyser, opening possibilities for diverse applicationreas, ranging from biosensing to cell-based assays and microbi-logical studies. Moreover, the entire protocol for data treatmentnd electrode modification may be particularly beneficial for theevelopment of stable, precise and accurate miniaturised devices.

cknowledgements

This work and the Ph.D. fellowship of C.C. were supported by theU-funded project NanoBio4Trans (“A new nanotechnology-basedaradigm for engineering vascularised liver tissue for transplanta-ion”, grant no: 304842). Additionally, A.H. acknowledges Lundbeckoundation (grant no. R69-A6408) for financial support.

ppendix A. Supplementary data

Supplementary material related to this article can be found, inhe online version, at http://dx.doi.org/10.1016/j.snb.2015.02.029.

eferences

[1] L. Coury, Conductance measurements part 1: theory, Curr. Sep. 18 (1999) 91–96.[2] N. Jaffrezic-Renault, S.V. Dzyadevych, Conductometric microbiosensors for

environmental monitoring, Sensors 8 (2008) 2569–2588, http://dx.doi.org/10.3390/s8042569.

[3] M.M. Barsan, M.E. Ghica, C.M.A. Brett, Electrochemical biosensors, in: D.P.Nikolelis, T. Varzakas, A. Erdem, G.-P. Nikoleli (Eds.), Portable Biosensing FoodToxicants Environ. Pollut., Taylor & Francis Group, Boca Raton, 2014, pp. 33–70.

[4] G. Pucihar, T. Kotnik, M. Kanduser, D. Miklavcic, The influence ofmedium conductivity on electropermeabilization and survival of cells invitro, Bioelectrochemistry 54 (2001) 107–115, http://dx.doi.org/10.1016/S1567-5394(01)00117-7.

[5] M. Pavlin, M. Kanduser, M. Rebersek, G. Pucihar, F.X. Hart, R. Magjarevic, et al.,Effect of cell electroporation on the conductivity of a cell suspension, Biophys.J. 88 (2005) 4378–4390, http://dx.doi.org/10.1529/biophysj.104.048975.

[6] J. Ramón-Azcón, E. Valera, A. Rodríguez, A. Barranco, B. Alfaro, F. Sanchez-Baeza,et al., An impedimetric immunosensor based on interdigitated microelectrodes(ID�E) for the determination of atrazine residues in food samples, Biosens. Bio-electron. 23 (2008) 1367–1373, http://dx.doi.org/10.1016/j.bios.2007.12.010.

[7] G.L. Damhorst, C.E. Smith, E.M. Salm, M.M. Sobieraj, H. Ni, H. Kong, et al., Aliposome-based ion release impedance sensor for biological detection, Biomed.Microdev. 15 (2013) 895–905, http://dx.doi.org/10.1007/s10544-013-9778-4.

[8] E.J.F. Dickinson, J.G. Limon-petersen, N.V. Rees, R.G. Compton, Quantitativelydiffusional? A theoretical and experimental investigation, J. Phys. Chem. C. 113(2009) 11157–11171.

[9] Q. Li, C. Batchelor-McAuley, N.S. Lawrence, R.S. Hartshorne, C.J.V. Jones, R.G.Compton, A flow system for hydrogen peroxide production at reticulated vitre-ous carbon via electroreduction of oxygen, J. Solid State Electrochem. 18 (2013)1215–1221, http://dx.doi.org/10.1007/s10008-013-2250-9.

10] J.B. Allison, J.A. Anderson, W.H. Cole, The method of electrical conductivity instudies on bacterial metabolism, J. Bacteriol. 36 (1938) 571–586.

11] G.N. Stewart, The changes produced by the growth of bacteria in the molecularconcentration and electrical conductivity of culture media, J. Exp. Med. 4 (1898)235–243.

12] P. Silley, S. Forsythe, Impedance microbiology-a rapid change for microbiolo-gist, J. Appl. Bacteriol. 80 (1996) 233–243.

13] X. Dominguez-Benetton, S. Sevda, K. Vanbroekhoven, D. Pant, The accurateuse of impedance analysis for the study of microbial electrochemical systems,Chem. Soc. Rev. 41 (2012) 7228–7246, http://dx.doi.org/10.1039/c2cs35026b.

14] L.B. Parsons, W.S. Sturges, The possibilities of the conductivity methodas applied to studies of bacterial metabolism, J. Bacteriol. 11 (1926)177–188.

15] L.B. Parsons, W.S. Sturges, Conductivity as applied to studies of bac-

terial metabolism II. Parallelism between ammonia and condictivity innutrient gelatin cultures of putrefactive anaerobes, J. Bacteriol. 12 (1926)267–272.

16] E. Petiot, A. El-Wajgali, G. Esteban, C. Gény, H. Pinton, A. Marc, Real-time monitoring of adherent Vero cell density and apoptosis in bioreactorprocesses, Cytotechnology 64 (2012) 429–441, http://dx.doi.org/10.1007/s10616-011-9421-2.

tors B 212 (2015) 544–550 549

17] K.F. Lei, M.-H. Wu, C.-W. Hsu, Y.-D. Chen, Real-time and non-invasive impedi-metric monitoring of cell proliferation and chemosensitivity in a perfusion3D cell culture microfluidic chip, Biosens. Bioelectron. 51 (2014) 16–21,http://dx.doi.org/10.1016/j.bios.2013.07.031.

18] C. Canali, A. Heiskanen, H.B. Muhammad, P. Høyum, F.-J. Pettersen, M. Hem-mingsen, et al., Bioimpedance monitoring of 3D cell culturing-complementaryelectrode configurations for enhanced spatial sensitivity, Biosens. Bioelectron.63 (2015) 72–79, http://dx.doi.org/10.1016/j.bios.2014.07.020.

19] S. Tully-Dartez, H.E. Cardenas, P.-F.S. Sit, Pore characteristics of chitosan scaf-folds studied by electrochemical impedance spectroscopy, Tissue Eng. Part C.Methods 16 (2010) 339–345, http://dx.doi.org/10.1089/ten.TEC.2009.0132.

20] C. Canali, S. Mohanty, A. Heiskanen, H.B. Muhammad, Ø.G. Martinsen, M. Dufva,et al., Impedance spectroscopic characterisation of porosity in 3D cell culturescaffolds with different channel networks, Electroanalysis 27 (2015) 193–199,http://dx.doi.org/10.1002/elan.201400413.

21] P.W. Atkins, Physical Chemistry, 8th ed., Oxford University Press, Oxford, 2006,http://dx.doi.org/10.1007/SpringerReference 77053.

22] J.P. Diard, P. Landaud, B. Le Gorrec, C. Montella, Automatic measurement ofthe conductivity of an electrolyte solution by FFT electrochemical impedancespectroscopy, J. Appl. Electrochem. 22 (1992) 1180–1184, http://dx.doi.org/10.1007/BF01297421.

23] B. Timmer, W. Sparreboom, W. Olthuis, P. Bergveld, A. van den Berg,Optimization of an electrolyte conductivity detector for measuring low ion con-centrations, Lab Chip 2 (2002) 121–124, http://dx.doi.org/10.1039/b201225a.

24] R. Gómez-Sjöberg, D.T. Morisette, R. Bashir, S. Member, Impedancemicrobiology-on-a-chip: microfluidic bioprocessor for rapid detection of bac-terial metabolism, J. Microelectromech. Syst. 14 (2005) 829–838.

25] K. Lei, Review on impedance detection of cellular responses inmicro/nano environment, Micromachines 5 (2014) 1–12, http://dx.doi.org/10.3390/mi5010001.

26] N.F. Sheppard, R.C. Tucker, C. Wu, Electrical conductivity measurements usingmicrofabricated interdigitated electrodes, Anal. Chem. 65 (1993) 1199–1202.

27] A.R. Heiskanen, C.F. Spegel, N. Kostesha, T. Ruzgas, J. Emnéus, Monitor-ing of Saccharomyces cerevisiae cell proliferation on thiol-modified planargold microelectrodes using impedance spectroscopy, Langmuir 24 (2008)9066–9073.

28] S. Chechirlian, M. Keddam, H. Takenouti, Specific aspects of impedance mea-surements in low conductivity media, in: J. Scully, D. Silverman, M. Kendig(Eds.), Electrochem. Impedance Anal. Interpret., ASTM International, Philadel-phia, 1993, pp. 23–36.

29] K.C. Stewart, D.G. Kolman, S.R. Taylor, The effect of parasitic conduction path-ways on EIS measurements in low conductivity media, in: J. Scully, D. Silverman,M. Kendig (Eds.), Electrochem. Impedance Anal. Interpret, ASTM International,Philadelphia, 1993, pp. 73–93.

30] A.M. Johnson, D.R. Sadoway, M.J. Cima, R. Langer, Design and testing of animpedance-based sensor for monitoring drug delivery, J. Electrochem. Soc. 152(2005) H6–H11, http://dx.doi.org/10.1149/1.1824045.

31] P. Harder, M. Grunze, R. Dahint, G.M. Whitesides, P.E. Laibinis, Molecular con-formation in oligo (ethylene glycol)-terminated self-assembled monolayers ongold and silver surfaces determines their ability to resist protein adsorption, J.Phys. Chem. B 5647 (1998) 426–436.

32] S. Lan, M. Veiseh, M. Zhang, Surface modification of silicon and gold-patternedsilicon surfaces for improved biocompatibility and cell patterning selectiv-ity, Biosens. Bioelectron. 20 (2005) 1697–1708, http://dx.doi.org/10.1016/j.bios.2004.06.025.

Biographies

Chiara Canali received the M.Sc. degree in pharmaceutical biotechnology (summacum laude) from University of Bologna in Italy in 2011. From 2009 to 2012 herresearch projects were focusing on bioanalytical chemistry and bioluminescenceimaging working at University of Bologna, Leiden University Medical Center andInstitut Polytechnique de Grenoble. She is currently a Ph.D. student at the TechnicalUniversity of Denmark (Department of Micro- and Nanotechnology). Her projectis focused on the design of bioimpedance-based sensing for applications in tissueengineering.

Layla Bashir Larsen is postdoc at the Technical University of Denmark (Departmentof Micro- and Nanotechnology). She is a biomedical engineer with a Ph.D. degreein Micro and Nano-engineering (University of Birmingham, UK) and prior clinicalexperience (within the National Health Service, UK) in the field of rehabilitationengineering. Her research focuses on the applications of engineering in medicineand includes development of sensors, microfluidic devices and tissue engineeringsupport systems.

Ørjan Grøttem Martinsen is professor at the University of Oslo (Department ofPhysics) since 2002 and head of the Bioimpedance group. He also has a part timeposition as researcher at the Department of Clinical and Biomedical Engineering atOslo University Hospital. He became electrical engineer in 1983 and Dr. Scient. in1995 defending a thesis on the electrical properties of human skin. His research

mainly focuses on the passive electrical properties of biological materials andapplications of bioimpedance technology in medicine, biology and physiology. Hisparticular expertise is within bioimpedance basic theory, instrumentation, electrodesystems and modeling. He is co-editor in chief of Journal of Electrical Bioimpedance(JEB) and President of the International Society for Electrical Bioimpedance (ISEBI).

5 Actua

A(mHF

50 C. Canali et al. / Sensors and

rto Heiskanen is associate professor at the Technical University of DenmarkDepartment of Micro- and Nanotechnology) focusing on the development of

icrofluidic systems with electrochemical detection for cell-based applications.e received the B.Sc. degree in biochemistry from Åbo Akademi University, Turku,inland, in 1987, the B.Sc. degree in chemistry from the University of the Philippines,

The author has requested enhancement of the downloaded file. All in-textThe author has requested enhancement of the downloaded file. All in-text

tors B 212 (2015) 544–550

Diliman, Quezon City, in 2001, and both the M.Sc. degree in 2004 and the Ph.D.degree in 2009 from Lund University, Lund, Sweden. He has coauthored over 50peer-reviewed publications on electrode fabrication and modification as well asdevelopment of microfluidic platforms for applications in electrochemical monitor-ing of cellular functions.

references underlined in blue are linked to publications on ResearchGate. references underlined in blue are linked to publications on ResearchGate.