A Third Generation Glucose Biosensor Based on Cellobiose Dehydrogenase Immobilized on a Glassy Carbon Electrode Decorated with Electrodeposited Gold Nanoparticles: Characterization and Application in Human Saliva

Efficient direct electron transfer (DET) between a cellobiose dehydrogenase mutant from Corynascus thermophilus (CtCDH C291Y) and a novel glassy carbon (GC)-modified electrode, obtained by direct electrodeposition of gold nanoparticles (AuNPs) was realized. The electrode was further modified with a mixed self-assembled monolayer of 4-aminothiophenol (4-APh) and 4-mercaptobenzoic acid (4-MBA), by using glutaraldehyde (GA) as cross-linking agent. The CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC platform showed an apparent heterogeneous electron transfer rate constant (ks) of 19.4 ± 0.6 s−1, with an enhanced theoretical and real enzyme surface coverage (Γtheor and Γreal) of 5287 ± 152 pmol cm−2 and 27 ± 2 pmol cm−2, respectively. The modified electrode was successively used as glucose biosensor exhibiting a detection limit of 6.2 μM, an extended linear range from 0.02 to 30 mM, a sensitivity of 3.1 ± 0.1 μA mM−1 cm−2 (R2 = 0.995), excellent stability and good selectivity. These performances compared favourably with other glucose biosensors reported in the literature. Finally, the biosensor was tested to quantify the glucose content in human saliva samples with successful results in terms of both recovery and correlation with glucose blood levels, allowing further considerations on the development of non-invasive glucose monitoring devices.


Introduction
Glucose monitoring has attracted great attention in several fields, ranging from biomedical applications to ecological fields [1]. In particular, for clinical trials, glucose monitoring has been considered one of the key factor in early diagnosis of diabetes mellitus, which is a main cause of death or other diseases around the world. Diabetes is a metabolic disease generally related to non-/under-production of insulin in the pancreas and hyperglycemia, reflected by blood glucose concentrations higher or lower than the normal range of 80-120 mg dL −1 [2]. It is possible to distinguish between three types of diabetes: (i) type 1, which most affects young people, with non-production of insulin in the pancreas and involves about 10% of diabetic people [3]; (ii) type 2, which occurs in (CYT CDH ) [30]. DH CDH domain is structurally similar to the FAD domain of most GMC-oxidoreductase enzymes and is fully reduced by di-/mono-saccharides, transferring the electrons through internal electron transfer (IET) to the CYT CDH , which finally shuttles the electrons to properly modified electrodes [31]. Among II class CDHs, Corynascus thermophilus CDH (CtCDH) was genetically mutated in its active site (CtCDH C291Y mutant) to enhance its sensitivity toward glucose and reduce the maltose cross-reactivity [32].
In this work, we report an improved DET efficiency between CtCDH C291Y and a novel GC modified electrode, obtained through direct electrodeposition of gold nanoparticles (AuNPs) on the GC electrode, further modified with a mixed self-assembled monolayer of 4-aminothiophenol (4-APh) and 4-mercaptobenzoic acid (4-MBA) using glutaraldehyde as cross-linking agent, as shown in Scheme 1. The proposed electrodeposition method allowed to monitor the nanoparticles surface coverage as well as the surface area available for the biomodification, which is directly related to the biosensor sensitivity. The so modified AuNPs/GC electrode was used to develop a third generation biosensor for glucose detection. The performances of the proposed biosensor were investigated in human saliva samples, demonstrating that the constructed AuNPs/GC biosensor has great potentials to realize electrochemical devices for non-invasive diabetes mellitus monitoring.
Sensors 2017, 17,1912 3 of 14 cytochrome domain (CYTCDH) [30]. DHCDH domain is structurally similar to the FAD domain of most GMC-oxidoreductase enzymes and is fully reduced by di-/mono-saccharides, transferring the electrons through internal electron transfer (IET) to the CYTCDH, which finally shuttles the electrons to properly modified electrodes [31]. Among II class CDHs, Corynascus thermophilus CDH (CtCDH) was genetically mutated in its active site (CtCDH C291Y mutant) to enhance its sensitivity toward glucose and reduce the maltose cross-reactivity [32].
In this work, we report an improved DET efficiency between CtCDH C291Y and a novel GC modified electrode, obtained through direct electrodeposition of gold nanoparticles (AuNPs) on the GC electrode, further modified with a mixed self-assembled monolayer of 4-aminothiophenol (4-APh) and 4-mercaptobenzoic acid (4-MBA) using glutaraldehyde as cross-linking agent, as shown in Scheme 1. The proposed electrodeposition method allowed to monitor the nanoparticles surface coverage as well as the surface area available for the biomodification, which is directly related to the biosensor sensitivity. The so modified AuNPs/GC electrode was used to develop a third generation biosensor for glucose detection. The performances of the proposed biosensor were investigated in human saliva samples, demonstrating that the constructed AuNPs/GC biosensor has great potentials to realize electrochemical devices for non-invasive diabetes mellitus monitoring. Scheme 1. The electrode modification pathway for CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC platform has been reported. Initially, AuNPs were directly electrodeposited onto cleaned GC electrode by sweeping the potential. Afterward, the electrode was incubated in a thiol mixture (1:1 v/v 4-APh and 4-MBA) overnight, followed by cross-linking reaction (glutaraldehyde GA: crosslinking agent) to covalently link the enzyme (CtCDH C291Y) to the so modified electrode surface.

Electrode Preparation and Modification
GC electrodes (Bioanalytical Systems Inc., West Lafayette, IN, USA, d = 3 mm) were polished with alumina slurries (Al 2 O 3 , particle size of 1 and 0.1 µm) on cloth pads wet with Milli-Q water (Struers ApS, Ballerup, Denmark), thoroughly rinsed with Milli-Q water and further sonicated for 5 min between each polishing step. GC electrodes were successively modified by electrodeposition of gold nanoparticles (AuNPs) by sweeping the potential between 1.1 and −0.1 V vs. Ag|AgCl sat for a given number of scans (5, 10, 15, 20, 25, 30, 35 scans) in 10 mM HAuCl 4 [33]. Then, the modified electrodes were activated in 0.5 M H 2 SO 4 by running 25 scans between 0 and +1.7 vs. Ag|AgCl sat at a scan rate of 0.1 V s −1 until a well-defined cyclic voltammogram (CV) was obtained. The best modified electrode was selected on the basis of the electroactive and real surface area, heterogeneous electron transfer rate constant (k 0 , cm s −1 ) and roughness factor (ρ), calculated from CV measurements carried out in 10 mM Fe(CN) 6 3−/4− (50 mM TRIS buffer pH 7.4). It was further dipped into a volumetric 1:1 mixture of 1 mM 4-APh/4-MBA ethanol solution. Then, the electrode was thoroughly rinsed with ethanol and dried under N 2 stream. For the biomodification, 1 µL of GA solution (2.5% v/v in distilled water) and 3 µL of CtCDH C291Y solution (16 mg mL −1 ) were drop-cast, gently mixed on the top of thiol-modified AuNPs/GC electrode and allowed to react in a moisturised atmosphere for 2 h to avoid evaporation of the reactants. Finally, the so modified electrode was gently rinsed with 50 mM TRIS buffer (pH 7.4) in order to remove any possible unbounded enzyme molecule [34].

Whole Saliva and Blood Samples Collection and Analysis
Saliva samples collection was performed at 8.30 a.m. from three healthy male and female patients refrained from eating, drinking and oral hygiene procedures (at least for 1 h before). The patients were given drinking bottled water and asked to rinse well their mouths. After 5 min, the patients were asked to spit whole saliva (WS) into a 50 mL sterile Falcon ® tube, once a minute for up to 10 min until sampling 5 mL of WS [35]. At the same time the patients were punched on their fingers to collect a drop of blood sufficient to measure glucose with the commercial GlucoContour XT (Bayer, Leverkusen, Germany) used by diabetic patients for self-monitoring and with the glucose oxidase-peroxidase method [36,37] by using the Glucose (GO) Assay Kit, which is the standard reference method for WS samples [38].

SEM Experiments
Scanned electron microscopy (SEM) measurements were performed with a JSM-7600F Schottky Field Emission Scanning Electron Microscope (JEOL Nordic AB, Sollentuna, Sweden). All samples were prepared according to the electrodeposition protocol, reported in Section 2.2, using glassy carbon plates (25 × 25 × 1 mm, ALS Co. Ltd., Tokyo, Japan) instead of GC electrodes. The samples were paced on a clip SEM sample holder (JEOL Nordic AB).

Electrochemical Measurements and Electrochemical Apparatus
Cyclic voltammograms (CVs) were recorded by using a PGSTAT 30 potentiostat (equipped with GPES 4.9, Autolab, Utrecht, The Netherlands). CVs were performed in a three-electrode electrochemical cell containing a standard silver chloride electrode (Ag|AgCl, sat. KCl), a platinum wire counter electrode and a modified glassy carbon (GC) electrode as working electrode. The temperature controlled experiments were carried out by using a cryostatic bath (T ± 0.01 • C, LAUDA RM6, Delran, NJ, USA). Flow injection analysis (FIA) data have been collected by using an analogic potentiostat (Zäta Elektronik, Höör, Sweden) connected with a strip chart recorder (Kipp & Zonen, Utrecht, The Netherlands). The modified GC electrode, an Ag|AgCl (0.1 M KCl) reference electrode and a platinum wire counter electrode were fitted into a wall-jet cell. The electrochemical system was equipped with a flow system consisting of a peristaltic pump (Gilson, Villier-le-Bel, France) and a six-port valve electrical injector (Rheodyne, Cotati, CA, USA).

SEM and Electrochemical Characterization of AuNPs Modified GC Electrodes
SEMs were used to evaluate the physical appearance and surface characteristics of the AuNPs on the electrode surfaces for a given number of scans. Figure 1a-g show the SEM images relative to increasing number of scans. It is clearly visible that the surface coverage of the AuNPs increases with increasing number of scans until 25 scans, when the electrode surface is completely covered by a single layer of AuNPs. For electrodes prepared with 30 and 35 scans (Figure 1f,g), it is possible to observe multiple layers of AuNPs with possible formation of AuNPs agglomerates.

SEM and Electrochemical Characterization of AuNPs Modified GC Electrodes
SEMs were used to evaluate the physical appearance and surface characteristics of the AuNPs on the electrode surfaces for a given number of scans. Figure 1a-g show the SEM images relative to increasing number of scans. It is clearly visible that the surface coverage of the AuNPs increases with increasing number of scans until 25 scans, when the electrode surface is completely covered by a single layer of AuNPs. For electrodes prepared with 30 and 35 scans (Figure 1f,g), it is possible to observe multiple layers of AuNPs with possible formation of AuNPs agglomerates. All AuNP-modified electrodes were successively characterized by cyclic voltammetry (CV) experiments in a solution of Fe(CN)6 3−/4− (data not shown) in order to calculate the electroactive area (AEA, cm 2 ), the heterogeneous electron transfer rate constant (k 0 , cm s −1 ) and the roughness factor (electroactive/geometrical area ratio, ρ) and in 0.5 M H2SO4 (Figure 2) in order to calculate the real surface area (Areal). All data are shown in Table 1. The AEA has been evaluated using the Randles- All AuNP-modified electrodes were successively characterized by cyclic voltammetry (CV) experiments in a solution of Fe(CN) 6 3−/4− (data not shown) in order to calculate the electroactive area (A EA , cm 2 ), the heterogeneous electron transfer rate constant (k 0 , cm s −1 ) and the roughness factor (electroactive/geometrical area ratio, ρ) and in 0.5 M H 2 SO 4 ( Figure 2) in order to calculate the real surface area (A real ). All data are shown in Table 1. The A EA has been evaluated using the Randles-Sevcik equation by the slope of the peak current vs. square root of scan rate (υ 1/2 ) [39], whereas the real surface area (A real ) was calculated by integration of the peak current related to the gold oxide reduction process occurring by running CVs in 0.5 M H 2 SO 4 [40,41]. The theoretical charge density considered for gold oxide reduction is 390 ± 10 µC cm −2 [42]. k 0 was calculated using the extended method which merges the Klingler-Kochi and Nicholson-Shain methods for totally irreversible and reversible systems, respectively [43,44]. It is possible to observe in Table 1 that all the electrochemical parameters are highly influenced by the number of scans, showing the best results after 25 scans with an A EA of 12.96 ± 0.18 cm 2 and a roughness factor of 183.6 ± 1.2, probably related to the increase in AuNPs surface coverage with the scan number. With electrodes prepared with 30 and 35 scans the decrease in the electrochemical parameters reported in Table 1 might be due to the presence of multiple layers and possible AuNPs agglomeration (Figure 1f,g).
reduction process occurring by running CVs in 0.5 M H2SO4 [40,41]. The theoretical charge density considered for gold oxide reduction is 390 ± 10 μC cm −2 [42]. k 0 was calculated using the extended method which merges the Klingler-Kochi and Nicholson-Shain methods for totally irreversible and reversible systems, respectively [43,44]. It is possible to observe in Table 1 that all the electrochemical parameters are highly influenced by the number of scans, showing the best results after 25 scans with an AEA of 12.96 ± 0.18 cm 2 and a roughness factor of 183.6 ± 1.2, probably related to the increase in AuNPs surface coverage with the scan number. With electrodes prepared with 30 and 35 scans the decrease in the electrochemical parameters reported in Table 1 might be due to the presence of multiple layers and possible AuNPs agglomeration (Figure 1f,g).

Electrochemistry of CtCDH C291Y on Modified GA/4-APh,4-MBA/AuNPs/GC Electrode
After preliminary characterization, the modified AuNPs/GC electrode obtained after 25 scans of electrodeposition was further modified with CtCDH C291Y covalently linked through GA with a mixed SAM consisting of 4-APh and 4-MBA. Figure 3a depicts the typical CVs of the enzyme electrode at different scan rates, showing an increased peak-to-peak separation (ΔEp) between the anodic and cathodic peak potentials. The modified electrode exhibited a clear linear dependence of both anodic and cathodic peak current densities versus the scan rate over the range 2-500 mV s −1 , as

Electrochemistry of CtCDH C291Y on Modified GA/4-APh,4-MBA/AuNPs/GC Electrode
After preliminary characterization, the modified AuNPs/GC electrode obtained after 25 scans of electrodeposition was further modified with CtCDH C291Y covalently linked through GA with a mixed SAM consisting of 4-APh and 4-MBA. Figure 3a depicts the typical CVs of the enzyme electrode at different scan rates, showing an increased peak-to-peak separation (∆E p ) between the anodic and cathodic peak potentials. The modified electrode exhibited a clear linear dependence of both anodic and cathodic peak current densities versus the scan rate over the range 2-500 mV s −1 , as shown in the inset of Figure 3a. The presented results fitted with thin-layer electrochemical behaviour, as generally reported for immobilized systems.  It is possible to observe in Figure 3b (black curve) a couple of peaks related to DET of CDH through the CYTCDH subunit containing the heme b, which displayed a midpoint potential (E 0 ') of −98 mV vs. Ag|AgClsat, close to the values reported in the literature for Ascomycota CDHs immobilized on gold electrodes [19]. The apparent heterogeneous electron transfer rate constant (ks) was calculated by considering an electron transfer coefficient of 0.53, obtained by fitting the linear part of the trumpet plot, as shown in the inset of Figure 3a. Therefore, the ks value was estimated to be 19.4 ± 0.6 s −1 , according to Laviron's equation [45] reported below: where α is the electron transfer coefficient, n the number of electrons, ΔEp the separation of the redox peak potentials and ν the scan rate (F = 96.495 C mol −1 , T = 298 K, R = 8.31 J mol −1 K −1 ). By integration of the redox peaks relative to the DET of CYTCDH it was possible to evaluate the enzyme surface coverage using the Faraday's law below reported in Equation (2): where ΓT is the total surface concentration of electroactive protein (mol cm −2 ), A the electrode area (cm 2 ), F the Faraday's constant (96 495 C mol −1 of electrons), Q the charge underlying the redox wave and n the number of electrons [46]. The theoretical surface coverage (Γtheor) was estimated to be 5287 ± 152 pmol cm −2 (Ageom = 0.073 cm 2 ), while the real surface coverage (Γreal) resulted to be 27 ± 2 pmol cm −2 (Areal = 13.83 ± 0.04 cm 2 , as shown in Table 1). Afterwards, the electrocatalytic behaviour of the It is possible to observe in Figure 3b (black curve) a couple of peaks related to DET of CDH through the CYT CDH subunit containing the heme b, which displayed a midpoint potential (E 0 ') of −98 mV vs. Ag|AgCl sat , close to the values reported in the literature for Ascomycota CDHs immobilized on gold electrodes [19]. The apparent heterogeneous electron transfer rate constant (k s ) was calculated by considering an electron transfer coefficient of 0.53, obtained by fitting the linear part of the trumpet plot, as shown in the inset of Figure 3a. Therefore, the k s value was estimated to be 19.4 ± 0.6 s −1 , according to Laviron's equation [45] reported below: where α is the electron transfer coefficient, n the number of electrons, ∆E p the separation of the redox peak potentials and ν the scan rate (F = 96.495 C mol −1 , T = 298 K, R = 8.31 J mol −1 K −1 ). By integration of the redox peaks relative to the DET of CYT CDH it was possible to evaluate the enzyme surface coverage using the Faraday's law below reported in Equation (2): where Γ T is the total surface concentration of electroactive protein (mol cm −2 ), A the electrode area (cm 2 ), F the Faraday's constant (96 495 C mol −1 of electrons), Q the charge underlying the redox wave and n the number of electrons [46]. The theoretical surface coverage (Γ theor ) was estimated to be 5287 ± 152 pmol cm −2 (A geom = 0.073 cm 2 ), while the real surface coverage (Γ real ) resulted to be 27 ± 2 pmol cm −2 (A real = 13.83 ± 0.04 cm 2 , as shown in Table 1). Afterwards, the electrocatalytic behaviour of the CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC electrode was studied by performing CVs in the presence of 5 mM glucose as substrate (Figure 3b, red curve), showing excellent performances with a current density of about 30 µA cm −2 , probably due to the high nanostructuration of the electrode surface and the covalent immobilization of the enzyme.

Glucose Biosensor Development
The amperometric response to glucose was studied by injecting glucose solutions at different concentrations by using the flow injection analysis (FIA) system, in order to investigate the electroanalytical and kinetic parameters of the modified CtCDH C291Y/GA/4-APh, 4-MBA/AuNPs/GC electrode. The biosensor showed a fast peak response (5 s), probably due to the enlarged surface area related to the electrodeposition of the AuNPs and the cross-linking of the enzyme, which ensure high number of immobilized enzyme molecules and stable enzyme layer.
The calibration curve displayed a linear response range between 0.02 and 30 mM (R 2 = 0.995, n = 5) with a sensitivity of 3.1 ± 0.1 µA mM −1 cm −2 , as shown in the inset of Figure 4a. At higher concentrations the amperometric response is no longer linear due to the saturation of the enzyme active site. The detection limit for CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC biosensor was found to be 6.2 µM, calculated using the relation 3σ/S, where σ is the absolute standard deviation of the intercept and S is the slope of the calibration curve [47]. The analytical performances of the glucose biosensor and the kinetic parameters are listed in Table 2. The apparent kinetic parameters (I max , K M app ) are in good agreement with the values reported in the literature for nanostructured electrodes [48]. It is interesting to underline that today very few third generation glucose biosensors based on Ascomycota CDHs have been reported in the literature while most other glucose biosensors are based on first and second generation electron transfer mechanism of other GMC oxidoreductase enzymes (e.g., GOx). The proposed CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC biosensor shows a clear increase in terms of sensitivity, selectivity, stability, extended linear range and lower detection limit compared to other second and third generation glucose biosensors reported in the literature, as shown in Table 3.

Effect of pH and Temperature, Interferences and Stability Studies
The effects of pH and temperature on the proposed glucose biosensor were evaluated and the results are reported in Figure 4b. The optimum pH resulted to be pH 7 in TRIS buffer at a temperature of 35 • C. A significant decrease in the current densities occurs below pH 5.5 and above 8, in perfect agreement with the data reported in the literature about the optimum pH of the free CDH [56]. The dependence on the temperature is shown in Figure 4c where it is possible to see that the amperometric response increased from 20 to 30-35 • C and drastically decreased above 37 • C, due to a possible inactivation of the enzyme caused by the temperature.
The stability and lifetime of the CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC biosensor was evaluated using the FIA system by monitoring the signal decrease within 20 days when the biosensor is used for one measurement per day, as reported in Figure 4d. The modified biosensor seems to retain about 90% of its initial activity after 20 days, probably due to the stability of the enzyme layer directly related to the nanostructuration of the electrode surface.
Finally, the selectivity of the proposed biosensor was studied in order to see the influence of possible interfering compounds generally present in human saliva such as maltose, cortisol, ascorbic acid, urea and calcium ions (Ca 2+ ). The signal obtained for a fixed concentration of glucose (750 µM) was compared to that obtained with a sample containing the same glucose concentration plus equal amounts of the possible interfering compounds. The amperometric signal is lower than 10% for all compounds tested with the exception of Ca 2+ ions, which potentially may interfere in real measurements (30% of glucose signal), probably because of its interaction with some amino acid residues present between the DH CDH and CYT CDH domains [56].

Glucose Detection in Human Saliva
In order to demonstrate the feasibility of the modified electrode for the non-invasive detection of glucose, the proposed biosensor was used to detect the concentration of glucose in human saliva samples. The samples were collected according to the procedure reported in Section 2.3, referred to literature on saliva analysis. The reliability of the amperometric biosensor platform CtCDH Sensors 2017, 17, 1912 11 of 14 C291Y/GA/4-APh,4-MBA/AuNPs/GC was evaluated by comparing the results with those obtained with the glucose oxidase-peroxidase method. The proposed biosensor showed satisfactory results in all samples tested with a recovery between 95.0 and 97.4% (RSD values lower than 4%), as reported in Table 4. The glucose content was measured also in blood samples collected from the same healthy patients with a commercial self-monitoring system (GlucoContour XT) in order to evaluate the correlation between glucose saliva and blood levels, for future potential development of devices for non-invasive glucose monitoring [57]. Figure 5 shows a good correlation between salivary and blood glucose concentration, opening the doors to the development of possible self-non invasive glucose monitoring devices.

Conclusions
We have demonstrated the possibility to carefully monitor the surface coverage of AuNPs on the electrode surface through a direct electrochemical deposition method of AuNPs onto a glassy carbon electrode which allows to achieve an efficient DET thanks to the effective nanostructure and the cross-linking of CDH molecules. AuNPs resulted to be very efficient for retaining the enzyme activity and promoting the electron transfer. The CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC biosensor showed great performances in terms of extended linear range and higher sensitivity, selectivity and stability compared to other glucose biosensors. The promising platform allowed the detection of glucose in human saliva with results in very good agreement with those obtained with the standard spectrophotometric method showing also a good correlation with glucose blood levels. For these reasons, the proposed biosensor may represent the basis for the development of a portable non-invasive device for glucose monitoring in diabetes mellitus patients.

Conclusions
We have demonstrated the possibility to carefully monitor the surface coverage of AuNPs on the electrode surface through a direct electrochemical deposition method of AuNPs onto a glassy carbon electrode which allows to achieve an efficient DET thanks to the effective nanostructure and the cross-linking of CDH molecules. AuNPs resulted to be very efficient for retaining the enzyme activity and promoting the electron transfer. The CtCDH C291Y/GA/4-APh,4-MBA/AuNPs/GC biosensor showed great performances in terms of extended linear range and higher sensitivity, selectivity and stability compared to other glucose biosensors. The promising platform allowed the detection of glucose in human saliva with results in very good agreement with those obtained with the standard spectrophotometric method showing also a good correlation with glucose blood levels. For these reasons, the proposed biosensor may represent the basis for the development of a portable non-invasive device for glucose monitoring in diabetes mellitus patients.