Diabetes mellitus is considered to be the most rapidly growing chronic disease of this century. It is a disease in which a person suffers from high blood sugar, either because the pancreas does not produce enough insulin (Type I), or because cells do not respond to the insulin that is produced (Type II). In 2014, 29.1 million Americans, about 9.3% of the total US population, were diagnosed with diabetes [1
]. Besides inducing negative impacts on the life quality of patients, the epidemic of diabetes is resulting in more costs to the government health care budget each year. For the self-management of diabetes, most patients need to test their blood glucose (BG) levels periodically, actively dosing with insulin, or taking an oral drug in combination with controlling their diet. Typically, a blood sample for analysis is obtained through a finger prick, which often causes physical and mental stress to patients. Therefore, there is a great need for a point-of-care glucose monitoring system that allows for noninvasive, rapid and painless measurements. Saliva has been studied as a better indicator of disorders and diseases than blood [2
]. Specifically, the salivary glucose (SG) level is considered an indicator of diabetes. According to preliminary studies, a good correlation between the BG level and the SG level has been verified [3
]. Thus, the tracking of BG through the continuous monitoring of SG levels is very promising. Saliva offers great advantages as a diagnostic fluid over other body fluids such as blood, tears, sweat, urine and so forth. First, it is easily tested by individuals with modest training; second, it is noninvasive, so the risk of infection or cross-contamination caused by frequent finger pricks is eliminated; third, it is convenient for people who face difficulty extracting blood samples such as infants, the elderly and haemophiliacs; and, last but not the least, saliva contains numerous disease-related biomarkers, including those typically found in blood [7
]. Taking these into account, this study developed a noninvasive SG sensing system which can be transformed into a saliva-based multiplex biomarkers detection system for the management of diabetes.
Various technologies, including infrared spectroscopy, fluorescence spectroscopy, raman spectroscopy, liquid chromatography–mass spectrometry (LC-MS) and gas chromatography–mass spectrometry (GC-MS), were proved capable of detecting glucose in clinical specimens [8
]. However, the requirements of expensive equipment and complicated operation prohibited them from being applied to home-based care. Hence, SG sensors introduced in this study were developed based on electrochemical sensing technology, which has attracted more and more attention due to its high selectivity, repeatability, accuracy, strong biocompatibility, and low cost [13
]. A typical sensor consists of a bioreceptor and transducers. The bioreceptor recognizes a target analyte and specifically reacts with it. Transducers are then used to convert the recognition event into a measurable electrical signal. For the purpose of glucose detection, the most commonly used bioreceptor is glucose oxidase (GOx) enzyme. In 1978 the first BG biosensor was designed relying on catalytic enzyme reactions [14
]. It detected the hydrogen peroxide (H2
) product that was generated from the glucose and GOx reaction when oxygen (O2
) was present. However, the measurement of H2
was easily altered by the product (H2
) accumulation and also limited by the concentration of O2
]. Due to these concerns, the second-generation glucose sensor was rendered by replacing O2
with a synthetic electron acceptor (mediator), which was capable of shuttling electrons from the redox center of the enzyme to the surface of the electrode [13
]. Measuring the electron transfer carried by mediators instead of products has successfully eliminated the drawbacks of the first generation. Nevertheless, there were still unsolved problems such as the potential leaching, low stability, and toxicity of mediators. That was when the third-generation glucose sensor came along, with the promise of a mediator-free glucose detection system based on direct electron transfer (DET). It is crucial to have a short distance or efficient electron-transfer pathway between the electrode and the reaction site which allows DET to happen. The third generation eliminated the interferences caused by O2
accumulation, and mediators leaching during the reaction, so more stable and precise electron transfer was captured [16
]. Because of these advantages, we developed a third-generation glucose sensor using screen-printed sensor chips as bases.
Screen printing technology was widely used to mass produce disposable sensor chips. This technology was favored since it had flexible choices of cheap bulk materials, a mature process development, a fast production line, and high manufacturing capability and capacity. Table 1
displays a list of screen-printed glucose biosensors: here, it is evident that the third-generation glucose sensors achieved a lower limit of detection (LOD) with a faster response time compared to the first- and second-generation. Some sensors in this table display a broader range of detection or a faster response time when compared to our sensors. However, none of them have ever successfully detected glucose from real saliva samples. Unlike the other similar electrochemical biosensors reported in literature, developed sensors in this study demonstrate the capability of a real time SG measurement within 2 min (including saliva collection and testing). In addition, sensors presented in this paper provide a sufficient glucose detection range of 1.1–45 mg/dL, pinpointing the biological range of human SG (0.5–20 mg/dL), permitting them to be applicable for diabetic control and health surveillance [17
]. Extremely low SG levels (<1.1 mg/dL) and high SG levels (>10.1 mg/dL) have not been examined yet since saliva samples of diabetic patients are excluded from this study. Here, an accurate, rapid, and portable SG measurement method is introduced to improve its clinical practicability and reduce pain for real-time diabetes management. Moving forward, our future goal is to measure extremely low/high SG levels in patients with diabetes as part of SG clinical validation study.
2. Experimental Section
In this study, the developed glucose-sensing system can accurately measure the SG levels of healthy subjects. Sensors were fabricated through a layer-by-layer (LBL) self-assembly process on screen-printed electrodes. They were designed to help diabetics manage their health conditions and treatment results by monitoring SG levels. However, this study is not intended to replace BG tests. We believe that SG itself could potentially be considered as means for diabetes monitoring. The clinical applicability of SG sensors will be clearer especially when the next steps of clinical studies are finished.
2.1. Sensor Fabrication
DS550 platinum (Pt) screen-printed sensor chips were purchased from DropSens (Metrohm USA Incorporated). The three metallic electrodes were a Pt working electrode (WE), a silver (Ag) reference electrode (RE), and a Pt counter electrode (CE) (Figure 1
). Phosphate buffered saline (PBS, pH 7.4, Sigma-Aldrich Co. LLC.) was dissolved in deionized (DI) water to yield 0.1 M PBS aqueous buffer solution. Carboxyl groups functionalized single-walled carbon nanotubes (SWNT–COOH, diameter: 1~2 nm; length: 2~5 μm, 4000 mg/L in DI water with 5~7 wt% COOH groups at the ends) were obtained from Brewer Science Company. Chitosan (CS, low molecular weight), spherical gold nanoparticles (GNp, colloid gold, 20 nm diameter stabilized suspension), GOx (type II lyophilized powder with at least 17,300 units/g solid, enzyme commission (EC) 18.104.22.168 enzyme from Aspergillus niger), and acetate buffer solution (pH 4.65) were purchased from Sigma-Aldrich Co. LLC. Before use, 400 μL of the SWNT suspension was dispersed in 10 mL DI water with the aid of ultrasonication for 2 h to obtain stable black stock suspension; 10 mg CS was dissolved thoroughly into 5 mL acetate buffer solution to yield 2 mg/mL CS suspension; 5 mg GOx was dissolved into 5 mL 0.1 M PBS to produce 1 mg/mL GOx suspension. Experiments were performed at room temperature at approximately 23 °C.
For sensor fabrication, all sensors were first rinsed with DI water and left to air dry. Only working electrodes were exposed while the other two electrodes were covered with dielectric tape (MSC industrial supply Co.). Then, 10 μL of SWNT suspension was cast onto each sensor and allowed to dry in desiccator (Terra Universal) under 10% relative humidity (RH). After 25 min, sensors were washed with 0.1 M PBS and dried up. The washing step was applied after the deposition of each layer (unless otherwise stated). Then, 10 μL of 2 mg/mL CS, 10 μL of GNp, and 10 μL of 1mg/mL GOx were cast onto the exposed electrode sequentially to form the first (CS/GNp/GOx) multi-layer film: each layer took 20 min to complete. Subsequently, two more multi-layer films were cast, and sensors were dried in the desiccator for 1 h without a washing step. After removing the dielectric tape, all resulting sensors were packed in gel-boxes (Gel-Pak) and then sealed in vacuum bags using a vacuum packaging machine (VACmaster pro110). Sensors were stored at 4 °C when not in use.
2.2. Micro-Fabrication Imaging
All micro-fabrication images were produced using a Supra 25 scanning electron microscope (SEM) from Gorge J. Kostas Nanoscale Technology and Manufacturing Research Center located at Northeastern University.
2.3. Sensor Measurement
Electrochemical measurements were carried out using a potentiostat (DY2100 mini, Digi-ivy), which was connected to a laptop with pre-installed post data processing software. A boxed adaptor for solid connection between single-use three-electrode sensors and potentiostat was purchased from DropSens. First, cyclic voltammetry (CV) electro-analytical tests were conducted and a suitable voltage to be applied between the WE and the RE was determined, under which the output current between the WE and the CE corresponded to the change of glucose concentrations. During the test, the applied voltage was sweeping from −0.2 V to 0.4 V at a scan rate of 50 mV/s. Current and voltage profiles of the third cycle were analyzed to determine a proper working potential. Then under a fixed working potential, amperometric tests were conducted to measure the glucose level, which is proportional to the output current signal. A 100 μL sample was dropped to cover all three electrodes before the selected voltage was applied between the WE and the RE constantly for 30 s. As glucose-GOx redox reactions happened on the WE, the output current between the WE and the CE illustrated how much charge passed along, quantitatively indicating the amount of glucose. In short, glucose concentrations were determined as a function of output current densities. As the reactions reached a steady state within 30 s, the output current in a time window of 27~30 s was integrated and used as the analytical signal. At the end of this 30 s test, a matlab script ran automatically to analyze the data and display results on a laptop.
2.4. Glucose Analysis with Reference Method
Spectrophotometric analysis based on enzymatic reactions by using ultraviolet-visible (UV-vis) spectroscopy was considered as a standard reference method for quantitative detection of glucose. Glucose assay kits (K606-100, Biovision incorporated) were used along with an ultraviolet (UV) spectrophotometer (mini1240, Shimadzu) and ultra-micro UV cuvettes (Brandtech scientific incorporated). This method detects 0.018–180 mg/dL glucose samples with a resolution of 7.2 × 10−5
]. For each batch of measurements, a set of six standard glucose solutions were prepared following the manufacturer’s protocol to provide a calibration curve. Samples were incubated in a water bath at 37 °C for 30–40 min and tested at a 570 nm wavelength. Absorbance readings were then converted into glucose concentrations using the calibration curve. All glucose concentrations in saliva/buffer samples were validated with standard UV tests.
2.5. Blood Glucose and Salivary Glucose Monitoring Test
Ten healthy volunteers in an age group of 20–60 years participated in an anonymous human subjects study. Diabetic and pre-diabetic subjects were excluded from this study. This project protocol was approved by Institutional Review Board of Northeastern University Human Subject Research Protection in February 2013 and was identified as #12-11-31. Each subject signed consent forms after fully understanding the purpose, procedure, and risks of the study, and were offered $12 compensation to purchase a lunch box at the completion of each session. Subjects were required to fast overnight without drinking/eating anything (except water) after 10 p.m. prior to the test date. In the morning, both blood and saliva samples of each subject in their fasted state were taken for analysis. Blood samples were measured by finger prick method using FreeStyle Lite BG monitoring systems (Abbott). Saliva samples for sensor tests were collected within 1–2 min following the saliva collection protocol as described in Section 2.6
. The pH and viscosity of saliva samples were also recorded using pH test papers (6.0–8.0 range, FisherBrand) and a portable viscometer (Core-Parmer). The remaining saliva samples were prepared for spectrophotometric analysis. The preparation procedures included boiling samples at 100 °C for 30 to 60 min, and centrifuging them at 12,000× g for 6 min. The supernatant was then collected and analyzed for glucose concentrations as a reference method.
2.6. Saliva Collection Protocol
Wait for 5 min after rinsing mouth with water;
Minimize swallowing and hold saliva in mouth;
Place sterilized dental cotton sponge in mouth and chew until it is soaked with saliva (typically < 1 min);
Deposit sponge into syringe directly from the mouth without touching it to avoid contamination;
Insert plunger into syringe;
Squeeze saliva through pre-installed Westran S 0.2 μm polyvinylidene fluoride (PVDF) membrane (Sigma-Aldrich Co. LLC.) at the bottom of syringe and into sterilized tubes. Usually 1 ml of saliva samples is obtained through this process;
Use pipette to drop 100 μL saliva onto a sensor to cover all three electrodes;
Dispose of sensor after washing out residual salivary specimen.
The saliva collection method provides a reliable filtering performance to remove large biomolecules such as mucins from saliva samples. The collected saliva’s viscosity usually ranges from 1.05 to 1.15 mPa·s, which is close to the viscosity of buffer solutions [30
]. This is because the 0.2 μm pore size filter that comes pre-installed in collection devices is comprised of a fine PVDF material, which is widely used for protein blotting due to its high protein binding capability (over 200 µg/cm2
]. Its high efficiency in mucins removal eliminates the matrix effects for saliva tests.