Functionalized TiO 2 Nanotube Platform for Gliadin Electroanalysis

: The present paper presents a gliadin detection method. This method is based on a modiﬁed Ti electrode. Modiﬁcation was performed by a simple and cheap anodization. Then, a layer of graphene oxide was added, and gliadin antibody was ﬁxed on the electrode surface. Using this complex system, electrochemical impedance spectroscopy was used for gliadin detection. Solutions with known gliadin (a fraction from gluten) content were used for analysis. Impedance measured at a certain frequency and coating resistance were analyzed. Better results (good linearity and lower detection limit) were obtained by plotting impedance at a certain frequency versus gliadin concentration. Coating resistance was proved to be in linear dependency with gliadin concentration only at lower concentrations. This system based on titanium nanostructured electrode has the potential to be used for gluten contamination detection from foods.


Introduction
Celiac disease is a well-known chronic immune-mediated disorder initiated by dietary gluten, having a prevalence of approximately 1% worldwide [1][2][3]. Although research is being done, a life-long gluten-free diet, first treatment from more than a half-century, remains the only treatment currently recognized [3,4]. For celiac disease patients, adherence to a gluten-free diet prevents further complications. There also some other related gluten disorders in which adherence to a gluten-free diet is required [5,6].
Food products can be labeled as "gluten-free" if their gluten content is <20 ppm (20 mg/kg product) and as "very low gluten content" if the gluten content is in [1,9] the range 20-100 ppm [1,9,10]. A common method is using the sandwich ELISA test kit based on R5 and Skerrit antibodies. However, this method has some disadvantages: it cannot identify the source of gluten-whether it is from wheat, rye, or barley; it takes a long time (between 30 min and 2.5 h depending on the manufacturer); it has difficulties in testing the heat-processed food samples (especially glutenin) [11,12]. R5 sandwich ELISA cannot be applied for hydrolyzed protein products, so R5 competitive ELISA, ELISA G12, and mass spectrometry was developed for hydrolyzed food [7]. Given these problems, other different methods were developed. One method is a genomic method-PCR (polymerase chain reaction) using specific DNA sequences [12,13]. A double PCR system was developed with DNA isolated from different cereals. First, genomic eukaryote DNA sequences were used to prove isolated nucleic acid substrate accessibility to PCR amplification. Then a wheat genomic DNA segment was amplified by PCR for wheat detection. DNA samples only are screened for wheat genomic DNA when the tested sample has proved positive for the first test [13]. It is much more sensitive compared to ELISA and can distinguish the source of gluten or cereal proteins but cannot be applied to the hydrolyzed proteins either [11,12]. Moreover, non-immunological matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) was applied for gluten detection with high sensitivity but with high costs and specialized equipment. Chromatographic techniques like: HPLC coupled to mass spectrometry (LC-MS/MS), gel permeation (GP) chromatography, reversephase (RP) chromatography, have also been used to detect gluten [12]. Another group of methods applied for this allergen detection are optical methods: near-infrared (NIR) [12] or recently luminescent sensing based on phosphorescent switch-on probe [14]. Many of these mentioned methods are using expensive equipment and considerable sample/reagent consumption, require complicated sample pre-treatment and are time-consuming [14].
Recently, electrochemical methods were also used for gluten detection from food or human samples-amperometry [15][16][17], surface plasmon resonance (SPR) [18], impedance [19], voltammetry [20,21]. Amperometric methods have used aptamers in different approaches. In two studies, aptamers were immobilized onto the surface of streptavidin-modified magnetic particles on carbon screen printed electrodes to allow chronoamperometric detection of gluten from different foods, including treated ones (heat-treated and hydrolyzed) [15,17]. Amperometric flow system was used with an immunodominant peptide from gliadin coupled to a competitive ELISA. A reduction of the enzymatic reaction time was obtained, and the results were well corelated with optical measurements [16]. Antibody was linked on gold SPR electrode via self-assembled monolayer (SAM) and it was used for gluten detection from patients' urine samples [18]. A sensor based on electrochemical impedance spectrometry also used aptamers immobilized on gold working electrode using poly (amidoamine) dendrimers [19]. Voltametric methods were based on differential pulsed voltammetry (DPV). One study used reduced graphene oxide modified electrodes with antibodies [21] and the other just a pencil graphite unmodified electrode [20].
Graphene oxide (GO) was used for sensor applications because of very good charge transfer capabilities, and because it can serve as supporting layers for biomolecules and presents high surface-to-volume ratio and low electrical noise [22]. Graphene oxide was already used for sensing gliadin (component of gluten) [21]. In another study, our group also used Ti electrode with graphene oxide deposited using cyclic voltammetry to test gluten presence [9].
Titanium dioxide (TiO 2 ) nanoparticles and nanotubes were used for example in constructing a glucose, cholesterol, or cancer cells sensor [22][23][24]. They were used in sensors applications because they can be easily prepared with high reproducibility and low cost, they have a large surface area and they are non-toxic, hydrophilic, and biocompatible [22,24]. These nanostructures have also other important characteristics: they have antibacterial activities, confer UV protection [25], and have the ability to incorporate compounds for controlled release [26].
Several studies combined these two nanomaterials-TiO 2 nanotubes and graphene oxide for sensors applications, to overcome the disadvantages of each of them and benefit from their advantages [22].
In this study, combined TiO 2 nanotubes-graphene oxide antigliadin antibody was tested for gluten presence evaluation. An electrochemical method was used-electrochemical impedance spectroscopy at open circuit potential. Two result interpretations were analyzed: impedance at a certain frequency versus gliadin concentration and coating resistance obtained by fitting the results versus gliadin concentrations.

Materials and Methods
Titanium round bars with 8 mm diameter (99.99% purity, Alpha Aesar) were cut into small pieces (3 mm thick). Pieces were polished using a Beta 2 grinder-polisher with CarbiMet (Buehler) abrasive papers, having porosities from 800 to 1200. Then, to remove impurities and obtain degreasing/pickling, samples were immersed in the ultrasound bath (model Ti-H-5 from Transsonic) for 15 minutes in distilled water, ethanol, and then acetone.
After samples were dried at room temperature, they were mounted in Origalys system to have a controlled surface.
TiNT were prepared by 2 hours' anodization at room temperature in ethylene glycol (Chimreactiv, anhydrous, 99.8% purity) with 0.5% w NH4F (Sigma Aldrich) and 2% distilled water. A spiral platinum wire (model 50SP17, Redox Research & Analytic, Otopeni, Romania) was used as counter electrode. Electrolysis was performed with an MPS-7163 (MATRIX) power source, at 50 V. After anodization, the sample was rinsed with distilled water to remove any electrolyte remains and was subjected to thermal treatment, 4 h at 400 • C in air, using an LEF-1035 (Daihan Labtech, Gyeonggi-do, Republic of Korea) oven.
TiNT GO antigliadin antibody was prepared in several steps, all at room temperature and light, each in sterile containers. In the first step, TiNT GO was immersed in 1 mM aqueous solution of 1-pyrenecarboxylic acid (Sigma Aldrich). After 1 hour, samples were rinsed with distilled water and then immersed another half an hour in a mixture of 15 mM N-(3-dimethylaminopropyl)-N -ethylcarbodiimide hydrochloride (EDC, Sigma Aldrich)/15 mM Nhydroxysuccinimide (NHS, Sigma Aldrich) in 0.1 M phophate buffer pH 7.4 (made from 0.2 M stock aqueous solutions of Na2HPO4 and NaH2PO4 from Sigma Aldrich). Two drops of 7 microliters each from a buffered aqueous solution of anti-gliadin (wheat) antibody produced in rabbit (Sigma Aldrich) were added using a Gilson micropipette on electrode surface. After 40 minutes, the electrode was rinsed with phosphate buffer. The final step was immersion in 2 w/v % bovine serum albumin (BSA, Sigma Aldrich) in phosphate buffer pH 7.0. After 1 hour, TiNT GO antibody was rinsed with phosphate buffer and allowed to dry.
Contact angle was measured with distilled water at room temperature and light using sessile drop method, with a CAM 100 contact angle meter (KSV Instruments LTD). A small drop was deposited from a Pyrex glass syringe with a blunt tip needle. Five measurements were performed on each sample.
Atomic force microscopy images were recorded with a machine from APE Research, Italy and processed with Gwyddion 2.9 software (A.P.E. Research srl, Trieste -Italy).
Electrochemical impedance spectroscopy measurements were performed with three electrode system using Autolab 302N (Metrohm, Eco Chemie, Utrecht, The Netherlands) with Nova 1.11 software. Measurements were performed in buffer solutions pH 7. Impedance spectra were recorded between 0.1-10 5 Hz, with 0.01 amplitude at open circuit potential. Only obtained experimental points between 0.1 and 5963 Hz were fitted using Nova 1.11 software (Metrohm, Nova 1.11, Eco Chemie, Utrecht, The Netherlands).
To avoid contaminations, sterile containers were used as electrochemical cell. Between measurements, electrodes were rinsed. Before recording a spectrum in a solution with gliadin, the working electrode was let in contact 30 min in this solution, to allow antibodyantigen coupling. For working electrode regeneration between measurements, a solution of NaOH/H 3 PO 4 , pH = 12, was used. Gliadin standard solutions were prepared using alcohol 60% and phosphate buffer.

Results
The scheme of titanium electrode modification with nanotubes, graphene oxide, and gliadin antibody are presented in Figure 1. Nanotubes were created because they allowed a higher surface. They were subjected to thermal treatment to increase crystallinity.
GO was deposited to increase the conductivity. It also increases the specific surface, allowing a high amount of antibody to be immobilized. It is also mentioned in literature that GO promotes diffusion of the analyte through interconnected pores [21].
Pyrene carboxylic acid was added to allow covalent bond with antigliadin antibody thru NHS and EDC chemistry.
Gliadin antibody was used to render the electrode specificity for gliadin (a gluten fraction).
In Figure 2a, Origalys system is presented. The Origalys system allowed a controlled surface, which is very important in electrochemical experiments. System design allowed round-shaped samples. It was used for all experiments. During thermal treatment, the samples were removed easily. For GO deposition and antibody deposition, the sample remained in the Origalys system. After each modification step, the sample morphology was changed from thin-walled nanotubes (Figure 2b), to nanotubes having thick walls (Figure 2c) after GO addition. After all steps for antibody immobilization (presented in Figure 1), partially covered nanotubes were obtained (Figure 2d).  The software for contact angle measuring device, gave left, right, and mean value for each measurement. The mean value of five different consecutive measurements for each of four samples was written in the Excel program. The average (presented in Figure 3) and the standard deviation for water contact angle were calculated for each sample. The standard deviation values were: 1.90 for Ti, 0.69 for Ti NT, 1.14 for TiNT GO, and 0.67 for TiNT GO antibody. The relatively small standard deviation values show that the sample surface is uniform. Ti substrate was freshly polished before contact angle recording.
Comparative FT-IR spectra are presented in Figure 4, after each modification step. They were recorded with 4 cm −1 resolution. Four consecutive scans were applied for each sample. Each sample has its own color, the same in all graphs. Using the equipment-corresponding software, the spectra were processed using background correction and automatic smoothing. For the TiNT sample (Figure 4a, red line), C-O-C bond and CH 2 groups at 1360 cm −1 (in inset) and respective at 2924 cm −1 are present, peak at 1638 cm −1 correspond to coordinated H 2 O as well as Ti-OH, and C-C stretching appears at 1074 cm −1 [27]. Traces of organic electrolyte remained.
The TiNT GO sample, blue line Figure 4b, has characteristic peaks. The peak at 1711 cm −1 it is assigned for C=O stretching, the one at 1404 cm −1 for O-H deformation, and C-O stretching appears at 1066 cm −1 [28].
When pyrene carboxylic acid was added (Figure 4c, green line), peaks for -COOH group are visible: at -OH in carboxylic acid which form intramolecular bonds at 3524 cm −1 and -OH stretching vibration at 3430 cm −1 . C=O stretching appears at 1684 cm −1 . The C-H stretching vibrations of aromatic and heteroaromatic stretching vibrations appears at 3040 cm −1 and 2994 cm −1 [29].
The modified electrode was tested in buffer solutions and buffer solutions containing known gliadin concentrations. The concentrations were selected between 10 and 80 ppm gliadin.
In Figure 5, a Bode modulus diagram is shown for TiNT GO antibody in electrolyte without and with different gliadin concentrations. The obtained signal (impedance) is not proportional to the concentration increase for all tested gliadin concentrations (between 10 and 80 ppm), at all frequencies in 0.1-10 5 domain. In this diagram, frequency and impedance (Z) are on a logarithmic scale. In the inset can be seen an increased magnification corresponding to the frequency of 0.543 Hz. Nyquist and Bode phase diagrams are not presented.
To observe signal stability, we recorded three consecutive scans in each concentration. The electrode was carefuly rinsed between measurements. For 0.543 Hz, the obtained impedance, at different concentrations, was added in Excel to allow the calculation of the standard deviation. The following standard deviation values were obtained for TiNT GO antibodyelectrode in phosfate buffer with gliadin: 0.531 in 0 ppm, 3.764 in 10 ppm, 0.808 in 15 ppm, 1.320 in 20 ppm, 2.089 in 40 ppm, and 2.279 in 80 ppm. The standard deviation is highest for 10 ppm gliadin concentration. The difference between the values obtained during three consecutive scans for one concentration is not big, the signal being stable.
In Figure 6, a graph of impedance vs. concentration is presented. It has a linear tendency, the Pearson's R being 0.990. The graph and linear fitting of the data were obtained with Origin software. Only six points were selected. Impedance data were fitted using equivalent circuit from Figure 7a. In this circuit, Rs is the solution resistance. Other resistance was added as Rcoating for coating layer (barrier oxide + nanotubes + graphene oxide + linker -pyrene carboxylic acid and NHS EDC system + antigliadin antibody) in parallel with CPE. CPE is the constant phase element. CPE has two components: Y and N. The first one is the numerical value of admittance. The second one ranges between 0 and 1. It is an empirical constant. Value 1 indicates a pure capacitor, and value 0, a pure resistor. Rcoating was plotted against concentration. It has also a linear tendency. Results from fitting are presented in Table 1 for all seven concentrations, including the values for the two components of constant phase element (CPE).

Discussion
The entire electrode modification process presented schematically in Figure 1 has many steps and takes many hours (2 h anodization, 4 h calcination, several minutes GO deposition, 3 h antibody deposition).
In Figure 2b, a very ordered nanotubular layer can be seen, having internal diameters of around 60 nanometers.
It is known that after anodization nanotubes are amorphous. Literature studies show that an increased voltage used for nanotube formation is conducive to an increase of the crystallinity of anodic TiO 2 films [31]. In our case, anodization was performed at a higher voltage (50 V), so nanotubes are expected to have a certain degree of crystallinity. However, thermal treatment is still needed to reach a crystalline phase.
In Figure 2c, after graphene oxide deposition, the nanotubes walls are much thicker. After the final step, antibody deposition, Figure 2d, the nanotubes' surface is covered. The structure is still porous.
As presented in Figure 3, for untreated Ti, an average contact angle value of 66 • was obtained. The substrate is hydrophilic, with a roughness of 39 nm. For TiNT sample, the contact angle value was lower (average 29 • ), and the roughness was increased from 39 to 192.7 nm. In our case, a hydrophilic substrate, by nano-structuration, becomes more hydrophilic with a higher surface roughness. According to literature [32], this corresponds to Wenzel state.
After GO deposition on titanium dioxide nanotubes, the contact angle is not influenced as much. It remains also close to 20 • (Figure 3e), but the roughness is very much increased (Figure 3g). Antigliadin antibody deposition does not influence much hydrophilicity (Figure 3f, about 20 • ) but this sample is the roughest (Figure 3h).
In Figure 4a, a comparation can be seen between Ti and TiNT spectra. Between 3000 and 3500 cm −1 , the broad peak is assigned for -OH groups [27]. Other characteristic peaks are visible for the anodized sample. In Figure 4b-d, the appearance or disappearance of some peaks is visible after each step. These peaks are marked with a dotted line.
The used method was impedance spectroscopy because it is a simple method and it disturbs the interface at minimum, measurements being performed at open circuit potential.
To make a calibration curve, the impedance at 0.543 Hz was selected. At this frequency, the impedance is proportional with concentration.
The limit of detection (LOD) and the limit of quantification (LOQ) were determined using linear regression according with method presented in literature [33]. LOD was 14 ppm and LOQ 45 ppm. According to presented literature, a food can be labeled as "gluten-free" if it has gluten content below 20 ppm. So, further work will be necessary to lower LOQ to 20 ppm. From Figure 7, it can be seen that R coating values are linear only in the 0-20 ppm domain, with Pearson's R of 0.991. After this concentration, it does not have good linearity. In this case, LOD and LOQ were not calculated. It seems to work only for lower concentrations. For many patients with disorders related to gluten, contaminations can be very harmful [5,6].
In this case, a better result was obtained by monitoring the impedance at one frequency than to fit the data and use obtained resistance for a calibration curve.
From Table 1, it can be seen that the solution resistance has similar values for all the samples, as it was expected. N have values around 0.9 in all the cases, indicating a pseudocapacitive behavior. R coating increased when the gliadin concentration was increased, except one case, at 40 ppm.

Conclusions
A complex electrode modification was obtained on titanium substrate. This modification is easy to achieve. A cheap method, electrolysis, was used to obtain the nanostructured electrode.
To determine the ability of this system to be used for gluten detection from foods, antigliadin antibody was immobilized on electrode surface using a chemical linker NHS:EDC mixture. It was tested against known concentrations of gliadin using a simple electrochemical method: impedance spectroscopy.
This complex system has the potential to be used for gluten presence evaluation from foods. Future work is necessary to lower LOQ and verify additional parameters required for a sensor. The system works better at lower gliadin concentrations, and could be suitable to verify food contamination, which is a problem for some very sensitive patients.