A novel sensitive and selective electrochemical sensor based on molecularly imprinted polymer on a nanoporous gold leaf modi ﬁ ed electrode for warfarin sodium determination †

Warfarin sodium (WFS) is a widely used oral anticoagulant drug but has a narrow therapeutic window. Since traditional detection methods applied for therapeutic drug monitoring su ﬀ er some shortcomings, including di ﬃ culty in obtaining timely reports, high costs and tedious operation, it is necessary to develop a detection system for the rapid monitoring of WFS in biological samples. Here we report a novel electrochemical sensor, which was facilely fabricated by coupling nanoporous gold leaf (NPGL) with molecularly imprinted polymer (MIP) to a ﬀ ord ultrasensitive and selective determination of WFS. The morphological characterization via scanning electron microscopy proved the successful modi ﬁ cation of the sensor by NPGL followed with MIP layer modi ﬁ cation. The in ﬂ uencing parameters including the type of monomer, pH and molar ratio of template to monomer were optimized during electro-polymerization. Using Fe(CN) 63 (cid:1) /4 (cid:1) as a probe acting as an electrical indicator, a linear relationship of the current response versus the concentration of WFS was obtained in the range from 1.0 (cid:3) 10 (cid:1) 10 to 8.0 (cid:3) 10 (cid:1) 8 M under the optimal experimental conditions, with a detection limit of 4.1 (cid:3) 10 (cid:1) 11 M (S/N ¼ 3). In addition, the as-prepared sensor exhibited speci ﬁ c detection of WFS over its structural analogues and interferents, and the established electrochemical approach was validated using the standard method – high performance liquid chromatography. Eventually, rapid and accurate determination of WFS in human blood was carried out after easy sample pretreatment.


Introduction
Warfarin sodium (WFS), a coumarin derivative, is an oral anticoagulant widely used in multifarious cerebrovascular and cardiovascular disorders, for instance, venous pulmonary embolism, atrial brillation, coronary heart diseases, etc. Despite its effectiveness, treatment with warfarin has several shortcomings, one of which is a narrow therapeutic window. Dosage out of the window may cause either treatment failure or unwanted bleeding and even threat to life. Therefore, the activity of WFS has to be monitored by blood testing to ensure an adequate yet safe dose. [1][2][3][4] The methods of determining WFS in biological samples include high performance liquid chromatography (HPLC) using uorescence 5 or ultraviolet detectors, 6 liquid chromatographytandem mass spectrometry (LC-MS/MS), 7 capillary zone electrophoresis (CZE), 8 and so on. However, particular concerns for these large instrument-based strategies are high energy and money consumption, a long working time, tedious pretreatment, etc., which is especially unsuitable for fast therapeutic drug monitoring. An alternative to the aforementioned techniques is electrochemical sensing, which gains much attention owing to easy preparation, high sensitivity, low detection limit, etc. 9,10 Specicity, however, is a common problem for sensorbased determination since normally no separation system is involved in the detection procedure. Therefore, physical, chemical or biological modication is oen necessary to endow a sensor with a specic recognition ability toward target molecules. [11][12][13][14] A very promising modier candidate is molecularly imprinted polymer (MIP) since it is able to provide antibodylike specicity and long-term stability. 15 Another barrier encountered in electrochemical sensing is that analytes must have electrochemical activity, [16][17][18][19][20][21] which means that the redox reaction of the target analyte should happen at the sensor surface under certain experimental conditions. As redox reaction conditions for different analytes could be very different, tedious optimization work is oen needed in order to nd suitable conditions and achieve sensitive electrical signals. In our previous work, we have developed a versatile way of measuring various substances by introducing active probes as electrical signal indicators (e.g. ferri/ferro-cyanide) along with an MIP-modied electrochemical sensor. [22][23][24] Such a strategy is independent of the electrochemical activity of the analyte itself, and thus it is possible to test any species using an electrochemical sensor.
Apart from specicity and feasibility, another important issue in sensor development is sensitivity. As materials with nanostructures are expected to enlarge the effective surface area of a planar sensor surface, various nano-agents have been employed to modify sensors for the purpose of improving sensitivity. [25][26][27][28][29][30] Among them, a very suitable material is nanoporous gold leaf (NPGL). The unique features of NPGL include high conductivity, interconnected three-dimensional (3D) architecture, uniformly distributed nanopores and nanoligaments, good biocompatibility, a relatively low cost, and possibility for mass production. [31][32][33] Compared with gold nanoparticles, NPGL is a free-standing mesoporous thin lm with a rigid 3D framework structure which avoids particle stacking, thus assuring controllability and stability of the NPGL modied sensor. In the present work, we fabricated an electrochemical sensor by combining NPGL and MIP and furthermore the composite-lm decorated sensor was applied for the detection of WFS. Several parameters inuencing the sensing performance have been carefully optimized and the resulting sensor has been successfully adopted to analyze WFS in human blood.

Materials and apparatus
Warfarin sodium (WFS), aspirin (ASP), hydrochlorothiazide (HCT), and vitamin K 4 (VK 4  Electrochemical measurements were performed on a CHI 760E Electrochemical Workstation (CHI Instruments Co., Shanghai, China) connected to a PC at room temperature. A conventional three-electrode system was employed, consisting of a bare or a modied planar gold electrode (GE, 4 mm in diameter) as the working electrode, a saturated calomel electrode (SCE) as the reference electrode and a platinum wire (0.5 mm in diameter, 34 mm in length) as the counter electrode. All potentials given in this paper were referenced to the SCE. The surface morphology of the NPGL was characterized using a Zeiss Supra55VP scanning electron microscope (SEM) operating at 20 kV. HPLC was performed with a Shimadzu (Japan) system comprising of LC-10A pumps and an SPD-10A UVdetector. The LC conditions were as follows: chromatographic separation was performed on a WondaSil C 18 column (150 mm Â 4.6 mm i.d., 5 mm) which was purchased from Dalian Elite Analytical Instruments Co. (Dalian, China). The mobile phase for WFS was acetonitrile-methanol-water (70 : 30 : 1, v/v/v) with a ow rate of 1.0 mL min À1 and the detection wavelength was 308 nm. All experiments involving the use of human subjects were performed in compliance with the relevant laws and institutional guidelines, and the institutional committees have approved the experiments. The informed consent was obtained for any experimentation with human subjects.

Preparation of sensor
The Ag/Au alloy leaves were pruned into a 7 Â 7 mm sheet and oated onto the surface of 65 wt% nitric acid. Erosion was stopped aer 60 min, and then the sample was washed with distilled water thoroughly and the NPGL was obtained. GE was rst polished repeatedly to a mirror nish with 0.3 and 0.05 mm Al 2 O 3 and thoroughly cleaned with distilled water before use. NPGL was carefully affixed to the GE surface and the NPGL-modied GE (NPGL/GE) was dried under an infrared lamp for 15 min. The MIP was electro-polymerized on the electrode surface using cyclic voltammetry (CV), which was performed between 0 and 1.0 V (vs. SCE) for 15 cycles at a scan rate of 50 mV s À1 in a solution containing functional monomers and WFS. Aerwards, the polymer lm-covered NPGL/GE was immersed in 0.1 M NaOH to remove embedded WFS by scanning from À0.5 V to +0.5 V for several cycles until obvious and stable redox peaks could be found in the probe solution GE) was prepared using the same method but in the absence of the template-WFS during electro-polymerization. 34

Electrochemical measurement
The electrochemical behavior of different electrodes was studied in the probe solution using CV, which was operated in the scanning range of À0.2 to +0.6 V at a scan rate of 100 mV s À1 . A sensor was rst incubated in a sample solution containing the analyte for 10 min, aer which the electrode was washed with water and applied for voltammetric measurement. The peak current shi (DI) was calculated from the reduction peak currents of Fe(CN) 6 3À/4À obtained before and aer binding with WFS. The shi was used to explore the inuence of different modications on the sensor performance and to establish a calibration curve for sample determination. Aer each analysis, the sensor was recovered by immersion in 0.1 M NaOH with CV scanning between À0.5 to +0.5 V for several cycles to remove WFS at the electrode surface. The electrode was reusable aer this cleaning procedure. Electrochemical impedance spectroscopy (EIS) experiments were carried out in a solution containing 5 mM Fe(CN) 6 3À/4À and 0.1 M KCl within a frequency range from 0.01 Hz to 100 kHz. All of the electrochemical experiments were performed at room temperature.

Detection of WFS in biological samples
A certain amount of WFS was added into human blood samples for a spiked recovery experiment. The sample was centrifuged at 4000 rpm for 10 min and methanol was added into the supernatant at a volumetric ratio of 1 : 1, followed by further centrifugation at 4000 rpm for 10 min to get rid of protein. The nal supernatant was used for WFS detection.

Preparation and characterization of WFS-MIP/NPGL/GE electrode
The process of preparing a hybrid sensor modied with MIP and NPGL composite lm is illustrated in Fig. 1. The morphology of the electrode surface was characterized using SEM, and it exhibits a sponge-like conformation with metal ligaments and nanopore channels ( Fig. 2A). The modication of NPGL with an electro-synthesized MIP layer is obvious as the width of the nanopores decreases from $30 nm to $20 nm and the image gets darker compared with bare NPGL (Fig. 2B). Ag/ Au alloy leaves, NPGL and molecularly imprinted polymer decorated NPGL (MIP/NPGL) aer the removal of WFS were probed via energy dispersive spectroscopy (EDS) and element mapping. All of the related information is given in Fig. S1-S4 in the ESI le. † Compared with the EDS spectrum of Ag/Au alloy leaf ( Fig. S1a †), there are only Au signals in NPGL (Fig. S1b †), indicating that Ag was removed aer dealloying. When MIP covered the surface of NPGL, the elements carbon and oxygen can be observed from the EDS spectra (Fig. S1c †), and their relative contents correspond to the calculated values of the functional monomer (resorcin, C 6 H 6 O 2 ), which conrms the formation of an MIP network on NPGL. From the element mapping images (Fig. S2 to S4 †), one can see that the elements Au, C and O are uniformly distributed on the electrode surface, implying that a homogeneous MIP lm was obtained on NPGL by the in situ electropolymerization of the functional monomers. The electrochemical behavior of the stepwise production  process was investigated via CV in a probe solution. As shown in Fig. 3A, compared with bare GE (curve a), the peak currents of NPGL/GE (curve b) increase obviously due to its enlarged surface area, which can enhance the detection sensitivity. Aer MIP modication, the redox peaks of Fe(CN) 6 3À/4À disappear (curve c). This phenomenon can be explained by MIP not being conductive and aer the whole electrode surface was densely covered with the polymeric lm, there was virtually no channel for the active probe ions to access the electrode surface. This also proves the successful preparation of WFS-MIP lm on the entire 3D surface of the NPGL electrode. Aer that, the removal of the template leads to an increase of the peak currents (curve d), as the cavities generated in the rigid polymer matrix open doors for the probe ions to transfer to the electrolyte/electrode interface. When the electrode was immersed in the solution of WFS, the rebinding of WFS by MIP impeded the electron transportation of Fe(CN) 6 3À/4À , resulting in current reduction (curve e). Electrochemical impedance spectroscopy (EIS) is also a helpful way of characterizing the stepwise sensor construction

View Article Online
process and provides useful information on impedance change at the electrode/electrolyte interface. 35 Fig. 3B depicts the EIS diagrams of electrodes at each preparation step. Bare GE (curve a) yielded a small semicircle, implying that the electrode has quite good conductivity. In comparison, the semicircle at NPGL/ GE (curve b) got much smaller, suggesting that the nanoporous structure of NPGL benets electron migration effectively. A signicant increase in the diameter of the semicircle was observed in curve c, which is ascribed to MIP modication and the polymer acting as a compact barrier for electron transfer. The conductivity of the electrode was recovered by getting rid of the template (curve d) due to the cavities created in the polymer matrix. Finally, the rebinding of WFS increased the impedance of the electrode (curve e), as a result of WFS embedding in binding sites and consequently inhibiting the redox reactions of probe ions. In order to investigate the role of NPGL on sensor performance, MIP/GE was prepared as a comparison sensor. Fig. 4A shows the cyclic voltammograms of MIP/NPGL/GE and MIP/GE before and aer binding with WFS, showing that the current shi of MIP/NPGL/GE is about 5.3 times larger than that of MIP/ GE. The enhanced sensing response of the hybrid sensor can be ascribed to the nanoporous architecture of NPGL which enhanced the active surface area and the conductivity of the electrode. Fig. 4B depicts the change in the cyclic voltammograms of MIP/NPGL/GE with varied scan rates. A linear relationship was observed between the redox peak currents and the square root of the scan rate, revealing a diffusion-controlled mechanism at the sensor surface. 36 3.2 Optimization of MIP/NPGL/GE preparation 3.2.1 Inuence of pH value on electro-polymerization of MIP. Four monomers (phenylenediamine, dopamine, o-aminophenol and resorcinol) were tested as functional monomer candidates for achieving the optimal imprinting effect toward WFS. First, the pH values during polymerization were optimized for each monomer with a molar ratio of template to monomer (T : M) of 1 : 3. Since the precipitation of WFS occurs when the pH value is less than 6.8, the pH range in the optimization work started at 6.8. Aer polymerization and the extraction of WFS, the electrodes modied with different MIP lms were employed for detection of WFS at the same concentration of 1.0 Â 10 À9 M and their reduction peak current shis (DI) before and aer the rebinding of WFS were used to evaluate the individual response level. As shown in Fig. 5A, the optimum pH values for phenylenediamine, dopamine, o-aminophenol and resorcinol are 7.5, 7.0, 7.5 and 7.0, respectively.
3.2.2 Optimization of the molar ratio of template to monomer (T : M). Aer choosing the most ideal pH value for each monomer, the molar ratio between the template molecule and monomer (T : M) was studied in the ratio range from 1 : 1 to 1 : 5. The effect of T : M is related to the number of binding sites generated during MIP preparation. Inadequate functional monomer results in insufficient binding sites available for WFS binding, while an excessive amount of monomer during  polymerization causes a thick MIP lm on the electrode surface, and consequently many binding sites are buried in the polymeric matrix and become non-effective. As shown in Fig. 5B, the MIP/GE with resorcinol as the monomer displayed the largest current response to WFS at the pH value of 7.5 and a T : M of 1 : 3.

Calibration curve and detection limit
Under the optimized experimental conditions, the CV responses and calibration curve for detecting WFS using the proposed MIP/NPGL/GE sensor were investigated. Fig. 6 shows the correlation between the current shi (DI) and the logarithm of the concentration of WFS (ln C), and the corresponding linear regression equation is DI (mA) ¼ 5.2513 ln C + 129.6810 (R 2 ¼ 0.9948) in the range of 1 Â 10 À10 to 8 Â 10 À8 M. The limit of detection (LOD) is down to 4.1 Â 10 À11 M (S/N ¼ 3), which is lower than the data from all of the reported WFS sensors we could nd. This comparison is summarized in Table 1.

Selectivity of the WFS-MIP/NPGL/GE
In order to assess the specicity of the MIP-modied sensor, HCT, VK 4 and ASP, the structural analogues of WFS, were detected. As shown in Fig. 7, the MIP/NPGL/GE exhibited a much higher response to WFS than to the other analogues, whereas NIP/NPGL/GE shows no obvious difference in sensing all of the analytes. The t-test shown in Fig. 7A also proved the remarkable difference in the results of the sensor response of MIP/NPGL/GE towards the analogues and WFS with a signicant level of 0.01. The high specicity of MIP can be ascribed to the classic molecular imprinting effect. 41,42 In particular, during MIP preparation, WFS was incorporated into polymeric networks via several kinds of non-covalent interactions and the following removal of WFS le behind the imprinting cavities that are complementary to the template in size, shape and functionality. NIP, due to the lack of imprinting effect, can only yield non-specic binding toward different substances. The calibration curves of WFS and its structural analogues in the range of 1 Â 10 À10 to 8 Â 10 À8 M are shown in Fig. S5, † in order to further investigate the sensitivity of MIP/NPGL/GE towards different compounds. It is obvious that MIP/NPGL/GE has a higher current shi for WFS than the analogues with the slopes of WFS (0.0531) being 3 times larger than that of HCT (0.0139), VK 4 (0.0150) and ASP (0.0173). In addition, it is known that some co-existing ions may cause interference in WFS detection in real samples. Hence, a mixture of WFS and several ions (NH 4 + , Ca 2+ , Na + , K + , NO 3 À , SO 4 2À and Cl À ) was assayed with the concentration of these ions being ten times higher than that of WFS. It was found that the sensing response from WFS alone is 97.8% of the response from the mixture, implying an excellent anti-interfering ability of MIP/NPGL/GE even in a complicated matrix where the amount of interferents is remarkably higher than the analyte.

Repeatability and stability
WFS at three different concentrations (8 Â 10 À8 , 8 Â 10 À9 and 8 Â 10 À10 M) was measured in triplicate using the same MIP/ NPGL/GE sensor. A relative standard deviation (RSD) of less than 2.5% was obtained, which reects the good repeatability of the sensor. When not in use, all of the electrodes were stored in open air at room temperature. Compared with the initial electro-signal, the current response decreased by about 3% over the rst week, then little change was observed in a one-month period of time, exhibiting the decent stability of the hybrid electrode, which could be ascribed to the inherent stability of NPGL and MIP.

Real sample analysis
In order to evaluate the capability of the developed sensor in detecting real clinical samples, WFS in human blood was tested. The results shown in Table 2 exhibited satisfying accuracy with recoveries ranging from 102.9% to 104.9%. In addition, HPLC was employed as the standard method to analyze the same blood samples and the results were in good accordance with those from our sensor, which proved the validity of the newly developed approach.

Conclusions
In this study, a novel hybrid sensor system was established using a two-step coating tactic of integrating NPGL modication and in situ MIP electro-synthesis for the specic detection of WFS. This is to our knowledge the rst report on the combination of WFS-imprinted polymeric lm with NPGL for establishing a hybrid membrane-modied electrochemical sensor. First of all, a 3D open and continuous nanoporous skeleton of NPGL helps to enhance electron transmission on one hand and allow for an enlarged platform for MIP loading on the other. Secondly, the MIP layer provides a number of imprinting cavities that match the template molecules based on geometrical size and the number and the steric arrangement of functional groups, guaranteeing specic recognition toward WFS. Application of the modied electrode in real sample analysis is proven to be more sensitive and convenient than the reference method, HPLC, with comparable accuracy and repeatability. Compared with the reported work of other systems, the sensor fabrication technique is facile and highly controllable, ensuring very good reproducibility and ease for developing admirable sensors in mass production. In summary, the sensor preparation procedure is simple and cost-effective, and the detection of WFS is far quicker than the commonly employed method, HPLC. Therefore, it can be expected that the developed sensor has great potential in therapeutic drug monitoring.