Affinity Sensor based on Immobilized Molecular Imprinted Synthetic Recognition Elements

Authors

Pieterjan Lenain[1]*, Sarah De Saeger[1], Bo Mattiasson[2], Martin Hedström[2].

[1] Laboratory of Food Analysis, Ghent University, Ottergemsesteenweg 460, B-9000, Ghent, Belgium. * Corresponding authorE-mail: pieterjan.lenain@ugent.beTel : +32 9 264 81 17Fax : +32 9 264 81 99

[2] Capsenze HB, Medicon Village, SE-22381 Lund, Sweden.

Abstract

An affinity sensor based on capacitive transduction was developed to detect a model compound, metergoline, in a continuous flow system. This system simulates the monitoring of low-molecular weight organic compounds in natural flowing waters, i.e. rivers and streams. During operation in such scenarios, control of the experimental parameters is not possible, which poses a true analytical challenge. A two-step approach was used to produce a sensor for metergoline. Submicron spherical molecularly imprinted polymers, used as recognition elements, were obtained through emulsion polymerization and subsequently coupled to the sensor surface by electropolymerization. This way, a robust and reusable sensor was obtained that regenerated spontaneously under the natural conditions in a river. Small organic compounds could be analyzed in water without manipulating the binding or regeneration conditions, thereby offering a viable tool for on-site application.

Keywords

Molecularly Imprinted Polymer • Capacitive Affinity Sensor • Monitoring • Sensors • Analytical Methods • Metergoline

Introduction

Continuous sampling and monitoring of target compounds in rivers and streams is an extremely challenging analytical task where the assay response time is of crucial importance. Depending on public health risk, quick action might be required. Pollutants in rivers have a great impact on communities and eco-systems both close to the polluted area and downstream. To address this issue, a robust, stand-alone, and preferably cheap solution is needed. Therefore, a sensing unit, which can be submerged in a river to continuously monitor specific target substances, could be of

1

1

2

3

456789101112131415

1617181920212223242526

27

2829

30

313233343536

1

interest. The most important requirement is that the device can operate independently over a longer period of time, e.g. several weeks or months. We have in this work developed an affinity sensor based on capacitive transduction, able to perform as the sensing device (detector) in such a unit. Molecularly imprinted polymers (MIPs) were used as recognition elements, where binding of the analyte to the sensor surface caused a capacitance change, which was further assessed by a constant current pulse to the sensor transducer, as described by Erlandsson et al. (Erlandsson et al., 2014).

The choice of using MIPs as recognition elements was based on the specific conditions in which the system needs to perform (Ruigrok et al., 2011). Contrary to natural antibodies, MIPs have no biological origin, are robust, and are chemically and thermally stable (Sellergren, 2001). This robustness allows for application in environments where biological recognition elements are unsuitable or will denature. MIPs can also be reused multiple times, which is essential to meet the stand-alone requirement. Metergoline was chosen as a model compound for small organic molecules. Small molecules, such as pesticides, herbicides, antibiotics, are widely discarded and encountered in naturally flowing waters. Imprinted polymers where metergoline was used as the template molecule have been described (Lenain et al., 2012). Metergoline, a drug used to treat disorders associated with hyperprolactinaemia and prophylaxis of migraine, was used as a model compound in this study, representative of small organic structures such as pharmaceutical residues (Reynolds, 1993).

Early attempts comprised the incorporation of MIPs in carbon paste through powder processing. An imprinted bulk-polymerized monolith was crushed, sieved, and mixed with carbon paste to produce a sensor (Blanco-López et al., 2004). A variation on this two-step method was the inclusion of particles in a sol-gel, where the swelling in an aqueous environment intended to facilitate better mass transfer of the analyte to the recognition elements, thus enhancing the sensitivity. However, these attempts suffered from reduced sensitivities, slow kinetics and long response times (Mazzotta et al., 2008). Proper regeneration of the surface was also inhibited, often limiting a sensor to a single use.

Most techniques described in the literature for combining MIPs and electro-chemical sensors utilize in situ approaches where the imprinting process and the coupling to the transducer occur in one step (Alexander et al., 2006; Blanco-López et al., 2004), and where the most frequently employed technique for producing MIPs to be immobilized onto the transducer surface is direct bulk polymerization (Morita et al., 1997). Similarly, monomers can be physically adsorbed onto a conducting surface in the presence of template, and electro-polymerized, leaving a functionalized layer behind (Suryanarayanan et al., 2010). Other film immobilization techniques include the use of agar gels, silanes or self-assembled monolayer (SAM) formation on the surface (Suryanarayanan et al., 2010). More refined sensor configurations could be obtained by applying grafting procedures, thereby producing functionalized imprinted monolayers (Panasyuk-Delaney et al., 2001). The performance of these sensors relies on the density and proper orientation of the SAM. One disadvantage of these films is the limited surface available for interaction with

2

37383940414243

444546474849505152535455

5657585960616263

646566676869707172737475

2

analytes. Another general drawback is the removal of template after polymerization. In fact, since these are label-free techniques, template leakage can cause false measurements. However, in situ prepared sensors cannot be subjected to harsh conditions for removal of template since this affects adsorption-based functionalized surfaces. The authors propose a two-step approach to overcome these drawbacks.

It was opted to produce small, uniformly sized, spherical MIPs, and their non-imprinted counterpart, NIPs, in a separate step by use of emulsion polymerization (Dvorakova et al., 2012; Perez-Moral and Mayes, 2004). NIPs lack specific binding sites and the comparison of their performance with MIPs therefore enables discrimination between specific and non-specific binding. The obtained beads were attached to a gold electrode surface via a thin but well-insulating polytyramine layer. Finally, residual pinholes in the sensing layer were covered with 1-dodecanethiol. This approach offered important advantages. The issue of template leakage was addressed by the off-line bead production since this allowed thorough and complete template removal with organic solvents, bases and increased temperature, which would otherwise damage in situ prepared electrodes. Also, employing spherical particles effectively enlarged the surface available for interaction with analyte molecules.

Experimental

Materials

Chloroform (98%) was purchased from Novolab (Geraardsbergen, Belgium). Methanol (MeOH, LC-MS grade) was obtained from Biosolve BV (Valkenswaard, Netherlands), acetone (>99.5%) from Fiers (Kuurne, Belgium), and ethanol (EtOH absolute, Analar Normapure) from VWR International (Leuven, Belgium). Sulfuric acid (95-97%), potassium dihydrogen phosphate (KH2PO4, p.a.), and potassium chloride (KCl, p.a.) were bought from Merck (Darmstadt, Germany). Azobisisobutyronitrile (98%), sodium dodecyl sulphate (≥98.5), metergoline (p.a.), methacrylic acid (MAA, 99%), ethylene glycol dimethacrylate (EGDMA, 98%), hexadecane (98%), hydrogen peroxide (30 wt%), tyramine (99%), dipotassium hydrogen phosphate (K2HPO4, ≥98%), 1-dodecanethiol (>98%), potassium ferricyanide (K3[Fe(CN)6], ≥99.0%), and triethylamine (≥99%) were purchased from Sigma Aldrich (Bornem, Belgium). Ultrapure water was obtained with a MilliQ system from Millipore (Brussels, Belgium).

Emulsion polymerization of metergoline-imprinted polymers

First, stock solutions of the disperse and continuous phases, respectively organic and aqueous phases, were prepared in order to minimize weighing errors. The disperse phase consisted of 5 ml chloroform and 113 mg azobisisobutyronitrile, dissolved in a 10 ml round-bottomed flask. The continuous phase comprised/(consisted of?) 25 ml ultrapure water and 720 mg sodium dodecyl sulphate mixed in a 50 ml flask, thus obtaining a 0.1M surfactant solution. Template (metergoline), functional monomer (MAA) and crosslinker (EGDMA) were added in a 1:6:24 molar ratio. In a long, small 20 ml glass flask, 50 mg metergoline (0.124 mmol), 63 µl MAA

3

7677787980

8182838485868788899091

92

93

949596979899

100101102103104

105

106107108109110111112

3

(0.743 mmol), 19.33 µl hexadecane, 0.561 ml EGDMA (2.97 mmol) and 0.95 ml disperse phase were added and stirred at 500 rpm with a magnetic stirrer. Hexadecane served as a co-stabilizer for the surfactant micelles. Next, 5 ml of the continuous phase was added and the content homogenized at 24000 rpm for 60 seconds using an IKA-mixer (IKA-Werke, Yellow line DI 25 Basic, 12 mm diameter mixer bar). The obtained mixture was transferred to a 25 ml round-bottomed flask together with a small magnetic stirrer bar and a tap was placed on the flask. This was submerged in an ice bath (0°C) and allowed to cool down for a few minutes. Cooling is necessary to reduce the evaporation of chloroform in the next step. Vacuum was applied to the flask by use of a vacuum pump after which the tap was sealed. Nitrogen was introduced into the flask which then was placed in a thermostatic cooling chamber (Thermotron, set at 10°C) above a magnetic stirrer plate and stirred at 250 rpm during the entire reaction. The flask was located 20 – 22 cm away from the UV light (75 mWcm-²; λ = 365 nm) and the polymerization reaction was initiated and allowed to react for 1 h.

After polymerization, the mixture was collected in aliquots and centrifuged at 12000 g (Awel MF48-R, NuAire, Chateau Gontier, France). The supernatant was discarded and MeOH was added to remove the template and unreacted monomers. The aliquots were placed in a shaker for 30 minutes, then centrifuged (12000 g) and the supernatant was discarded. After the centrifugation step, a sample of the supernatant was injected in a time-of-flight-mass spectrometer to verify the absence of template. This process was repeated several times until all template was removed. Subsequently, the aliquots were placed in an oven, overnight, at 40°C. Non-imprinted polymers were prepared in exactly the same manner, however, without addition of metergoline. The size of the beads was measured with a Zetasizer Nano ZS (Malvern Instruments Ltd, Worcestershire, UK) and the size distribution of the MIP obtained by dynamic light scattering (DLS) is displayed in Supplementary Figure 1. Ninety percent of the MIP beads were distributed between 200 – 800 nm, with the median at 350 nm.

Coupling of MIP beads to an electrode

Disposable electrodes were prepared by use of thermal evaporation to coat a silicon wafer with chromium (50 nm) and gold (200 nm) as described by Teeparuksapun et al. (Teeparuksapun et al., 2009; Teeparuksapun et al., 2012). Before coupling, the electrode surface was cleaned to remove protective coatings on the surface. The electrode was submerged and sonicated for 10 minutes in acetone, ethanol, and piranha solution (H2SO4:H2O2; 3:1), successively, and subsequently dried under a stream of nitrogen.

MIP or NIP particles were suspended by sonication in a 10 mM conductive tyramine solution. This solution was prepared by dissolving 0.0137 g of tyramine (0.100 mmol) in 2.5 ml ethanol. Once all tyramine was dissolved, 7.5 ml of phosphate buffer (10 mM KH2PO4/K2HPO4 in ultrapure water, pH = 7.2) was added. An electrode was fixed in a reaction cell and the suspended solution was transferred to this cell. The beads were allowed to sediment onto the electrode’s surface for 30 minutes. The electrode acted as a working electrode, and the reaction cell was

4

113114115116117118119120121122123124125

126127128129130131132133134135136137

138

139140141142143144

145146147148149150

4

further equipped with a platinum reference and auxiliary electrode that allowed the electro-oxidation of tyramine by variation of the potential. After 15 potential sweeps between 0 V and 1.5 V with a 0.05 V.s-1 potential step, a polytyramine layer with a thickness around 80 nm was formed (Tenreiro et al., 2007). Finally, the electrodes were placed in a 10 mM 1-dodecanethiol in ethanol solution for 30 minutes to insulate the remaining pinholes on the gold surface. The proper formation of each layer onto the surface was verified with cyclic voltammetry, recorded in 10 mM K3[Fe(CN)6] in 0.1 M KCl.

Automated flow injection system

The capacitance measurements were performed with an automated flow injection system developed by Capsenze HB (Lund, Sweden). This system is based on current pulse capacitive measurements, and was described by Erlandsson et al. (Erlandsson et al., 2014). The feed buffer used was a 10 mM KH2PO4/K2HPO4 buffer at pH = 6.32. A multi-port valve that can be used to inject controlled volumes of up to six different samples was utilized. Regeneration of the working electrode after binding was performed by injecting a regeneration buffer consisting of MeOH and 10 mM KH2PO4/K2HPO4 buffer at pH = 6.32 with 5% triethylamine (47.5/47.5/5). The electrode with immobilized MIP or NIP was inserted in the electrochemical flow-cell and connected to the gold working electrode and the platinum auxiliary- and reference electrodes. The capacitance measurement was performed via the current step method where a constant current of +10 µA and -10 µA were alternately supplied to the surface of the working electrode. The capacitance was calculated from the resulting registered potential profile and plotted as function of time. The tubing system was connected to the inlet of the flow-cell where a constant flow-rate of running buffer was aspirated by the syringe pump. Prior to the analysis, the regeneration buffer was injected to the system via port 1 to clean the surface and remove weakly bound compounds. The binding event between the target analyte and the immobilized MIPs (flow rate 10 µL/min, sample volume 250 µL) on the electrode surface resulted in a decrease in the registered capacitance. The capacitive responses for both the MIP and NIP functionalized electrodes were sampled with 1 minute intervals.

Preparation of solutions for determination of working range and cross reactivity

The electrodes were tested towards increasing concentrations of metergoline dissolved in 10 mM phosphate buffer (pH 6.32). A stock solution was made and dilutions of 0.5, 1, 2.5, 5, 10, 20, 30, 40 and 50 µM were prepared. The first injection was pure feed buffer solution, which served as a control measurement.

A first cross-reactivity test was performed with pergolide methanesulfonate and 2-bromo-α-ergocryptine methanesulfonate. Solutions with different concentrations were prepared for both compounds by use of 10 mM phosphate buffer (pH 6.32). Pergolide solutions with 1, 5, 25 and 50 µM concentration were prepared, as well as 2-bromo-α-ergocryptine solutions with 1, 5, 10 and 25 µM concentration.

5

151152153154155156157

158

159160161162163164165166167168169170171172173174175176177

178

179180181182

183184185186187

5

A second cross-reactivity test was performed in which each sample contained a fixed concentration of 10 µM pergolide and 10 µM 2-bromo-α-ergocryptine, and increasing amounts of metergoline. The concentrations of metergoline were 1, 3, 10 and 30 µM, and a blank sample of phosphate buffer without metergoline was included as well.

Results and discussion

The synthesized MIP particles were suspended in a tyramine solution and attached to the electrode by electropolymerization. The formation of polytyramine occurred at the gold surface and the beads were integrated in this layer through matrix entrapment (Tenreiro et al., 2007). The electrode surface was examined with SEM after coupling of the imprinted beads (Fig. 1). These pictures clearly show the distribution of the particles on the gold surface. Figure 1d. shows the surface of the electrode at the boundary of the silicon wafer and the deposited gold layer. The spherical MIP particles are attached to the gold layer and stretch about 25 µm further onto the silicon wafer. This latter could be explained by assuming that tyramine polymerized from the gold layer onto the silicon wafer for a limited distance. This indicated that the coupling could be attributed to matrix entrapment in the polytyramine layer and not to other types of interaction, e.g. static, hydrophobic, or non-specific. Electropolymerization is commonly used to oxidize pyrrole and deposit layers of overoxidized polypyrrole (oPPy) (Tokonami et al., 2012). However, the deposition of an oPPy layer on a metallic electrode surface is prone to exfoliation. To the best of our knowledge, this problem is addressed for the first time by use of polytyramine matrix entrapment.

Evaluation of the sensor signal

Since the binding between recognition element (i.e. MIPs) and target analyte relies on relatively weak interactions (e.g. ionic interactions, hydrogen bonds), it was assumed that the regeneration could occur spontaneously with uncontaminated water. The case of multiple, successive, time-separated contaminant discharges was considered. After some time, the contamination would pass the sensor node and interact with the surface. Since the signal at any moment will reflect the level of contaminant in the medium, the sensor has the property to allow on-line measurements.

The capacitive sensor from Capsenze HB (Lund, Sweden) was used to simulate these circumstances in a laboratory environment. Initially, 50 µM of metergoline was injected and feed buffer was allowed to flow for 3 h without any regeneration, illustrated in Fig. 2. Considering the flow rate of 100 µl/min and the small dimensions of the flow cell, it was assumed that changes for the MIP and NIP electrodes at any given time point were caused by the same factors. The capacitive response profiles of the MIP and NIP functionalized electrodes were markedly different from each other. The response of the MIP electrode was characterized by a steep inclination after introduction of metergoline, whereas the NIP electrode showed a less pronounced deviation. After 3 h, the signal of both electrodes returned to baseline level, thereby offering validity to the hypothesis of regeneration through spontaneous dissociation. The curve obtained after subtracting the NIP signal from the MIP signal (red curve in Fig. 2) showed a steep

6

188189190191

192

193194195196197198199200201202203204205206207

208

209210211212213214

215216217218219220221222223224225

6

inclination, followed by a descent towards baseline level within one hour. The signal of a NIP electrode was considered to be the accumulation of all non-specific interactions, and the MIP electrode registered all specific and non-specific binding. The curve after subtraction of the NIP signal from that of the MIP represented specific binding due to the target compound. The concentration of target analyte was proportional to the maximum amplitude of this curve, defined as the difference in capacitance between the start of the capacitance drop and the maximum decline (indicated by ΔC in Fig. 2). This approach should be plausible since the measurement occurred in non-equilibrium conditions. It should also be considered that evaluating the sensor signal after a fixed time interval could result in improper evaluation of the capacitance change.

A possible explanation for this behavior could be provided by considering the electrical double layer theorem (Bard and Faulkner, 2001). A 10 mM phosphate buffer solution was flowing over the electrodes at a constant rate before injection of the target compound. A stable electrical double layer was formed and the measured capacitance values remained constant. Once the compound was injected, it bound to the synthetic receptors on the surface, thereby displacing the double layer further away from the surface. This displacement caused a decrease in capacitance, in accordance with the electrical double layer theorem. This behavior has been reported by other investigators (Yang et al., 2004; Zhou et al., 2007). This decrease occurred very rapid, which was attributed to the sudden entry of target compound and subsequent fast adsorption to the receptors. Since the difference in molarity of the buffer solution and the compound in solution is three orders of magnitude, differences in conductivity could not account for such a large decrease. Additionally, the influence of compound concentration on the conductivity, and thereby capacitance, was corrected for by subtracting MIP and NIP capacitive responses. Conversely, desorption of the target compound led to an increase of the capacitance. This occurred in a more gradual fashion compared to the adsorption process because of higher affinity of metergoline and the receptor, and thus a slower release rate.

The evaluation of the sensor signal was performed according the difference in capacitance, expressed as ΔC (in nF). This was obtained by subtracting the capacitive response of the NIP electrode from the MIP electrode for each pulse, and was plotted in function of the concentration.

Spontaneous regeneration and robustness

The proportional relation between the metergoline concentration and the capacitance changes as observed and illustrated in Fig. 3, indicated that metergoline could be detected and quantified. It was also apparent that metergoline dissociated over time until the original baseline conditions were reached, as indicated by the dotted line. This is important as it demonstrates full regeneration of the electrode surface without the use of a regeneration solvent, thereby demonstrating the sensor’s reusability. The electrodes used to obtain the data in Figure 3 were mounted into the capacitive device immediately after production, without overnight equilibration. The drift during the injection of control solution with no active substance and 10 µM metergoline solution was typical behavior for electrodes that were not sufficiently equilibrated prior to first

7

226227228229230231232233234

235236237238239240241242243244245246247248249250

251252253

254

255256257258259260261262263

7

use. This behavior could be related to the integration of (non-) imprinted polymers. Depending on the solvent in which they are submerged, these polymer will possess a different degree of swelling. This directly influenced the distance between the electrical double layer and the gold surface of the electrode, and thereby affecting the capacitance. However, once the electrodes were fully equilibrated, the base level showed a very stable course.

One interesting feature is that the MIP can be regenerated almost immediately just by exposing it to water. When MIPs with higher binding strength are used, regeneration will take longer. Lower affinity results in convenient dissociation/regeneration, but at the expense of the sensitivity of the assay. Many analyses using capacitive biosensors based on e.g. antibody recognition elements are orders of magnitude more sensitive (Loyprasert et al., 2010).

Next, regeneration buffer (phosphate buffer/MeOH/triethylamine; 47.5/47.5/5) was injected between every sample to facilitate and expedite the analysis. Increasing concentrations of metergoline were injected and the measurements were plotted in Fig. 4. As before, a proportional relation between measured signal and concentration was established in the range 10 – 50 µM. Except for the lowest concentration, the standard deviation values were below 4.2%. The response profile was similar whether regeneration solvent was used or not. The retained performance of the sensor after the harsh conditions imposed by the regeneration buffer, illustrates the robustness offered by synthetic recognition elements.

Working range, limit of detection, and cross-reactivity

The limit of detection was determined to be 1 µM and a working range from 1 to 50 µM, the maximum solubility in aqueous solvent, was established (Fig. 5a). Cross-reactivity experiments were performed with the structural analogues pergolide methanesulfonate and 2-bromo-α-ergocryptine methanesulfonate. Their concentrations were in the same order as for metergoline, to obtain a relevant comparison. 2-Bromo-α-ergocryptine methanesulfonate was not soluble at the highest concentration of 50 µM. All the measurements were performed in triplicate and reported with error bars. However, these latter are not always visible due to a very high reproducibility and the associated low rate and magnitude of experimental errors. The differences in capacitance changes generated by separate injection of these compounds are plotted in Fig. 5a, and the signal generated by metergoline was three times higher than that of the nearest analogue.

The cross-reactivity test in which the samples contained a fixed amount of the structural analogues and increasing amounts of metergoline is displayed in Fig. 5b. Here, the compounds competed for the selective binding sites and simulate the presence of cross-reacting compounds in river water. The curve was corrected for the contribution of the cross-reacting compounds. Standard deviations are not visible due to very high reproducibility, and are between 0.062 – 0.097 nF. (see Supplementary Table 1). The curve clearly indicates the proportional relation between the metergoline concentration and the capacitive change.

8

264265266267268

269270271272273

274275276277278279280281

282

283284285286287288289290291292

293294295296297298299

8

Although there was binding of the structural analogues, this amounted to a maximum contribution of 1.3 nF. This illustrated that analogous structures provoked a less intense signal in the same concentration range compared to the target compound. This was, however, true when measurements are performed in a controlled environment, when the content of the samples is known. This observation is in accordance with the results from the first cross-reactivity test.

Influence of ionic strength on the behavior of the sensor

Capacitive measurements are based on the electrical double layer theorem (Helmholtz, 1879). As such, the ionic strength of the solution influences the interactions at the electrode/solution interface during measurement. Therefore, the influence of ionic strength on the behavior of the sensor was investigated as well. Buffer solution and increasing concentrations of metergoline were prepared in the same 100 mM phosphate buffer and the capacitance values were compared to the same dilutions in 10 mM phosphate buffer with the same set of electrodes (Fig. 6). It should be remarked that the peaks belonging to the injection of metergoline samples of the two curves did not overlap due to the fact that a different run-time was used for the different series. However, this graph illustrates the difference in general behavior of the sensor at different ionic strengths.

The sensor response curve was more stable at higher ionic strength. However, it was noticed that the extent of capacitance change for different concentrations of analyte was less pronounced with higher ionic strength of the buffer. These findings are coherent with each other and could be explained by the Gouy-Chapman theory (Bard and Faulkner, 2001). If the electrolyte concentration was increased, the electrical double layer was compressed at the electrode surface, resulting in a capacitance rise. Conversely, in lower electrolyte concentrations, this layer dispersed further into the buffer solution and was therefore more susceptible to changes, which resulted in more fluctuations of the baseline and the profile of the capacitance values curve.

Conclusions

The MIP electrode was able to bind more metergoline compared to the NIP, resulting in an enhanced difference in capacitance. A proportional relation between the analyte concentration and capacitance change was established in the range 1 – 50 µM, characterized by highly reproducible measurements. The increased capacitance difference of structurally related compounds could not be attributed solely to non-specific binding, however, it constituted one third or less of the differences caused by metergoline binding. The buffer ionic strength played an important role in the sensors’ behavior. This parameter can easily be controlled in a laboratory environment, however, this is less straightforward in naturally flowing waters and should therefore be considered. The sensor was able to withstand harsh environments. Most important, a complete and spontaneous regeneration of the sensor was established in conditions imposed in an environmental setting where external manipulation is restricted. All these findings make the developed sensor a viable tool for on-site applications.

9

300301302303304

305

306307308309310311312313314315

316317318319320321322323

324

325326327328329330331332333334335336

9

Cross-reactivity is a crucial performance indicator for sensors. The currently presented set-up with two electrodes, would probably experience difficulties in terms of cross-reactivity in river waters. To address this issue, a multi-channel approach with several sensing electrodes which allows simultaneous monitoring could be considered. The channels would be modified with imprinted polymers with different chemical constitution, selective towards the target compound, and their correspondent non-imprinted polymers. Pattern recognition would be used to extract characteristic signals, or fingerprints, for the target compound from the obtained data matrix.

A sensor of this type could be a useful tool to monitor e.g. pollution from production of pharmaceuticals. It has been demonstrated that release of bioactive compounds from pharmaceutical industries in India have caused severe problems. The release of e.g. antibiotics has contributed to the creation of multiresistant bacteria (Johnning et al., 2013; Kristiansson et al., 2011; Marathe et al., 2013). Therefore, surveillance of these threats to the quality of surface water is becoming increasingly important and urgent.

The issue of biofouling will become exceedingly important when operating such system in open waters for extended periods. One solution to this problem can be to utilize shielding membranes over the sensor electrode with a suitable cut-off. In this direction, experiments need to be performed where dissociation of the target molecule size is optimized.

Acknowledgements

The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 269985. Biotage AB (Lund, Sweden) is highly acknowledged for their help and support in this work.

References

Alexander, C., Andersson, H.S., Andersson, L.I., Ansell, R.J., Kirsch, N., Nicholls, I.A., O'Mahony, J., Whitcombe, M.J., 2006. J. Mol. Recognit. 19, 106-180.

Bard, A.J., Faulkner, L.R., 2001. John Wiley & Sons, Inc., New York (USA).

Blanco-López, M.C., Lobo-Castañón, M.J., Miranda-Ordieres, A.J., Tuñón-Blanco, P., 2004. TrAC Trend. Anal. Chem. 23, 36-48.

Dvorakova, G., Haschick, R., Klapper, M., Müllen, K., Biffis, A., 2012. J. Polym. Sci. Part A. 51, 267-274.

Erlandsson, D., Teeparuksapun, K., Mattiasson, B., Hedström, M., 2014. Sensors and Actuat. B: Chem. 190, 295-304.

Helmholtz, H., 1879. Annalen der Physik 243, 337-382.

10

337338339340341342343

344345346347348349

350351352353

354

355356357

358

359360361362363364365366367368369370371372373

10

Johnning, A., Moore, E.R.B., Svensson-Stadler, L., Shouche, Y.S., Larsson, D.G.J., Kristiansson, E., 2013. Appl. Environ. Microbiol. 79, 7256-7263.

Kristiansson, E., Fick, J., Janzon, A., Grabic, R., Rutgersson, C., Weijdegard, B., Soderstrom, H., Larsson, D.G.J., 2011. PloS one 6, 7.

Lenain, P., Di Mavungu, J.D., Dubruel, P., Robbens, J., De Saeger, S., 2012. Anal. Chem. 84, 10411-10418.

Loyprasert, S., Hedstrom, M., Thavarungkul, P., Kanatharana, P., Mattiasson, B., 2010. Biosens. Bioelectron. 25, 1977-1983.

Marathe, N.P., Regina, V.R., Walujkar, S.A., Charan, S.S., Moore, E.R., Larsson, D.G., Shouche, Y.S., 2013. PloS one 8.

Mazzotta, E., Picca, R.A., Malitesta, C., Piletsky, S.A., Piletska, E.V., 2008. Biosens. Bioelectron. 23, 1152-1156.

Morita, M., Niwa, O., Horiuchi, T., 1997. Electrochim. Acta 42, 3177-3183.

Panasyuk-Delaney, T., Mirsky, V.M., Ulbricht, M., Wolfbeis, O.S., 2001. Analytica Chimica Acta 435, 157-162.

Perez-Moral, N., Mayes, A.G., 2004. Analytica Chimica Acta 504, 15-21.

Reynolds, J.E.F.e., 1993. The Pharmaceutical Press, London (England).

Ruigrok, V.J.B., Levisson, M., Eppink, M.H.M., Smidt, H., van der Oost, J., 2011. Biochem. J. 436, 1-13.

Sellergren, B., 2001. Elsevier Science, Amsterdam.

Suryanarayanan, V., Wu, C.T., Ho, K.C., 2010. Electroanal. 22, 1795-1811.

Teeparuksapun, K., Kanatharana, P., Limbut, W., Thammakhet, C., Asawatreratanakul, P., Mattiasson, B., Wongkittisuksa, B., Limsakul, C., Thavarungkul, P., 2009. Electroanal. 21, 1066-1074.

Teeparuksapun, K., Hedstrom, M., Kanatharana, P., Thavarungkul, P., Mattiasson, B., 2012. J. Biotechnol. 157, 207-213.

11

374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414

11

Tenreiro, A.M., Nabais, C., Correia, J.P., Fernandes, F.M.S.S., Romero, J.R., Abrantes, L.M., 2007. J. Solid State Electr. 11, 1059-1069.

Tokonami, S., Shiigi, H., Nagaoka, T., 2012. In: Ge, S.L., Lunec, S.A.P. (Eds.), Molecularly Imprinted Sensors, pp. 73-89. Elsevier, Amsterdam.

Yang, L., Wei, W.Z., Xia, J.J., Tao, H., 2004. Anal. Lett. 37, 2303-2319.

Zhou, L., Ye, G.R., Yuan, R., Chai, Y.Q., Chen, S.M., 2007. Sci. China Ser. B-Chem. 50, 547-553.

12

415416417418419420421422423424

425

426

12

Figure captions

Figure 1. Overview of SEM pictures of the electrode surface after functionalization with imprinted polymers. From left to right, top to bottom: (a) Overview of electrode surface with SEM; (b, c) at the center of the electrode; (d) at the border between the gold layer and wafer.

Figure 2. Capacitive response (nF) plotted towards time (minutes), each measurement was obtained with a one-minute interval. The capacitance values of a (○) MIP and (●) NIP functionalized electrode, and (□, red curve) the difference between both, were normalized for the baseline values (set equal to 0 nF) of each curve. The maximum difference in capacitance (ΔC) between the MIP and NIP electrode is indicated by the arrow.

Figure 3. Differences between capacitance changes (nF) of the MIP and NIP functionalized electrode plotted as a function of time (minutes). Regeneration occurred spontaneously in uncontaminated feed buffer. The respective metergoline concentrations are indicated at their time of injection. The dotted line indicates the baseline level when no analyte is bound to the functionalized sensor surface.

Figure 4. Difference between capacitance changes (nF) of the MIP and NIP functionalized electrode in function of metergoline concentration (µM). The measurements with the use of regeneration solvent between each metergoline injection were repeated in four-fold.

Figure 5. Differences between capacitance changes (nF) of the MIP and NIP functionalized electrode in function of concentration (µM) for (a) separate injections of metergoline, pergolide methanesulfonate and 2-bromo-α-ergocryptine methanesulfonate (not soluble at 50 µM) and for (b) metergoline in the presence of 10 µM of both structurally related compounds.

Figure 6. The influence of the ionic strength on the behavior of the capacitive change differences. The curve with the red squared markings represents the metergoline solution in 10 mM phosphate buffer and the curve with the black circular markings shows the solutions with 100 mM phosphate buffer, both with pH 6.32.

13

427

428429430

431432433434435

436437438439440

441442443

444445446447

448449450451

452

453

13