INTRODUCTION
Wearable health monitoring technologies and devices are of great and continuous interest in clinical healthcare due to their ability to monitor physiological signals and to help maintain an optimal health status as well as assess the physical fitness of outpatients (
1,
2). In particular, wearable biosensors aim to replace centralized hospital-based care systems with home-based personal diagnostics to reduce healthcare costs and time to diagnosis by providing noninvasive, real-time analysis (
3,
4) .Therefore, a wide variety of approaches have been proposed to bring such analysis methodologies closer to patients in both time and space (
5,
6). Early research activities on continuous health monitoring using wearable sensors focused on physical sensing (
7–
9). These efforts have resulted predominantly in temperature, pressure, and electric field sensors for monitoring biophysical signals including heart rate (
6), respiration rate (
10), skin temperature (
11), and brain activity (
12). Recent interest, however, focuses on chemical and biochemical sensing to monitor clinically relevant biomarkers using wearable devices to broaden the range of measurable quantities (
13,
14). Among many bodily fluids, sweat provides a significant amount of information about a patient’s health status and is readily accessible, making it suitable for wearable, noninvasive biosensing (
15). Sweat contains important electrolytes, metabolites, amino acids, proteins, and hormones, which allows monitoring of metabolic diseases, physiological conditions, or a person’s intoxication level (
16,
17).
Stress plays an important role in the overall health of a patient; when under stress, the adrenal gland releases cortisol and adrenaline into the bloodstream. The cortisol levels in various bodily fluids can range from 4 pM to 70 μM depending on the fluid. In sweat, the optimum level of cortisol ranges from 0.02 to 0.5 μM (
18,
19). Increased levels of cortisol have a detrimental effect on the regulation of various physiological processes such as blood pressure, glucose levels, and carbohydrate metabolism, and sustained stress can disrupt homeostasis in the cardiovascular, immune, renal, skeletal, and endocrine systems, leading to development of chronic diseases (
19). Therefore, continuous monitoring of cortisol levels in bodily fluids has great relevance in maintaining healthy physiological conditions. As a result, there is much interest in devising wearable devices able to monitor stress levels. Most stress sensors described in the literature are based on physical sensing and mainly focus on monitoring skin perspiration or conductivity, heart rate, and temperature (
6,
20). These approaches are promising in terms of fabrication using novel functional materials having desirable mechanical properties such as stretchability, flexibility, and high durability. However, the alteration of bodily physical parameters can also be induced by nonstress-related causes such as weather conditions and fever, making these sensors generally vulnerable to false positives. Furthermore, recent devices often show poor performance in terms of invasiveness, stability of recognition, selectivity, and sample acquisition (
19). However, in one recent study, Jang
et al. demonstrated a field-effect transistor-based cortisol sensor by embedding a cortisol antibody into the synthetic polymer matrix to generate a cortisol-selective/sensitive membrane. The designed sensor shows high sensitivity and a low limit of detection (down to 1 pg/ml) (
18). Here, we describe the development of a wearable biosensor using an organic electrochemical device for the detection of stress by selectively sensing cortisol in sweat.
Recently, electrochemical transducing elements have been developed to directly detect biomarkers from patients (
3,
21,
22). Among electrochemical transducing elements, organic electrochemical transistors (OECTs) are preferred in the field of bioelectronics due to their exceptional ability to interface electronics with biology (
23,
24). An OECT consists of a semiconductor polymer channel, typically poly(ethylenedioxythiophene):poly(styrenesulfonate) (PEDOT:PSS), that can be gated through an electrolyte solution (
25). The ions in solution are pushed by the gate potential to dope/de-dope the entire volume in the organic semiconductor channel, thereby strongly modulating its conductivity (
26). Hence, OECTs are able to transduce biological ion-based signals into electrical signals with high gain at relatively low voltages (<0.5 V) (
27). They have been widely proposed for many applications including water and food safety and health monitoring through detection of biologically relevant ions (
28), metabolites (
22), and pathogens (
29), which are commonly used for diagnostics (
30). Because cortisol is not charged at physiological pH, OECTs are, in principle, not suitable for its detection. To obviate this drawback and target specific molecules, common biocomponents such as antibodies or enzymes have been used as the specific recognition elements in the fabrication of OECT-based biosensors (
31). However, these sensors have inherent limitations such as vulnerability to interference from other ions present in the sample solutions and lack of selectivity to a particular analyte even if high-affinity biomolecules are used. Given the often poor chemical and physical stability of these biomolecules, which is especially problematic in wearable devices, artificial receptors have been gaining in importance as a possible alternative to natural systems (
32). Artificial receptors, such as molecularly imprinted polymers (MIPs), are increasingly becoming recognized as a versatile tool, particularly for the preparation of synthetic polymers containing tailor-made recognition sites (
33). Hence, to overcome the existing limitations of unstable biorecognition, we developed an MIP-based artificial recognition membrane that is interposed between the PEDOT:PSS channel layer and the analyte (sweat) reservoir to control and regulate the selective molecular transport of cortisol directly from the skin to the OECT sensing channel (
Fig. 1A). Here, we report the development of new skin-mounted patch-type molecularly selective OECT (MS-OECT) as a cortisol biosensor by implementing a multifunctional layered device, which can be adapted to other types of sweat-based wearable sensors.
DISCUSSION
In summary, we have demonstrated the integration of an artificial receptor as a biomimetic polymeric membrane for stable and selective molecular recognition using OECTs to produce a wearable sweat diagnostics platform for real-time analysis of the human stress hormone cortisol. This wearable sensing device for cortisol detection was realized using a conductive polymer channel functionalized with a cortisol-selective membrane produced on a flexible and stretchable elastomeric substrate. The molecularly selective polymer-based membrane shows high chemical and physical stability at body temperature, as well as resistance to physical deformation. The presented sensor tolerates mechanical testing such as bending and stretching in conditions similar to those found in the normal range of motion of the human epidermis. Moreover, we used a simple strategy to generate a passive fluid control system consisting of a laser-patterned microcapillary channel array that provides fast and precise delivery of sweat directly to the sensor interface. The resulting wearable sensor was used for measuring cortisol concentration in a real human sweat sample collected during exercise. Considering that traditional blood analysis is often used for cortisol sensing, the wearable device provides many advantages including noninvasiveness, ease of operation, and user comfort. The initial results suggest that the design principles presented here can be adapted for the selective detection of various other molecules, especially noncharged biomolecules and hormones. Moreover, future efforts will be aimed at miniaturizing the device, combining different sensing interfaces for multiplexing, data evaluation, processing and transmission of the results to a user interface, noninvasive sweat induction, and harvesting energy directly from bodily fluids to power the device.
MATERIALS AND METHODS
Preparation of cortisol-selective MIPs
All MIPs were synthesized by bulk polymerization. For all polymerization conditions, 0.2 mmol of the template (cortisol) was used. The initiator 2,2′-azobis (2-methylpropionitrile), azobisisobutyronitrile (AIBN) [0.2 mole percent (mol %) of template], methacrylic acid (MAA) functional monomer (1.2 mmol), and ethylene glycol methacrylate (EDMA) cross-linker (from 3.6 to 12 mmol depending on optimization parameters) were dissolved in corresponding solvent the reaction chamber was sealed and then the mixture was degassed with N2(g) for 10 min. The polymerization was initiated by UV irradiation at either room temperature or 4°C for 30 hours. The control NIPs were prepared by following the same conditions, except neglecting to add the template molecules to the reaction chamber. After mechanically grinding the as-synthesized polymer, they were processed by washing in acetic acid/methanol mixture (8:2, v/v). The washing solution was applied to polymers by first dispersing the particles in the centrifuge tube, shaking for 30 min, and then centrifuging at 1000 rpm for 15 min. This process was repeated five times successively for each polymer, and all particles were dried at vacuum overnight.
Device fabrication
The SEBS substrate was prepared successively by dissolving the SEBS polymer in toluene (200 mg/ml) and curing in a petri dish overnight. After obtaining a thin elastomer film (100 μm thick) as a flexible and stretchable substrate, it was cleaned by handwashing with soap followed by sonication in baths of methanol and then isopropanol for 10 min each. The substrates were treated with UV-ozone for 30 min. Ag/AgCl paste was brush-printed for gate, source, and drain contacts, and the whole substrate was baked at 110°C for 20 min to cure the conducting contacts. After stabilizing the conducting paste, drain source and gate contacts were masked to spin coat the first layer of PEDOT:PSS. The PEDOT:PSS blend was prepared by adding ethylene glycol (6 volume %) (Sigma), (3-glycidyloxypropyl)trimethoxysilane cross-linking agent (1 volume %), and dodecylbenzenesulfonic acid (one drop per 10 ml) to a stock Clevios PH 1000 PEDOT:PSS solution (Heraeus). The PEDOT:PSS solution is then spin-coated onto clean substrates at 2000 rpm to achieve a film thickness of approximately 100 nm. The molecularly selective and/or control membranes were coated on top of PEDOT:PSS by spin coating a tetrahydrofuran (2 ml) mixture containing a high–molecular weight PVC (50.0 mg), the bis(2-ethylhexyl) plasticizing solvent mediator (120.0 μl), and the anion excluder potassium tetrakis(4-chloropheny1) borate (15.0 mg). We varied the composition of MIP in the PVC matrix (10, 25, and 40% by weight) and observed the highest selectivity and sensitivity with 40 weight % (wt %) MIP content. Further increase of the concentration of MIP in PVC resulted in a highly dense, slurry-like mixture that could not be processed for further applications. After coating the membrane, contacts were exposed to the electrode connection for further characterization.
Device characterization
All output, transfer, and drain current measurements were performed with the Keithley 2612 SourceMeter using custom LabVIEW software. All voltammetric measurements were carried out using a BioLogic SP-200 potentiostat. Cyclic voltammetry measurements were performed in conventional three-electrode setup, where Ag/AgCl was used as the reference electrode, the platinum wire as the auxiliary counter electrode, and the whole device as the working electrode.
BET measurement
The specific surface area of the different templates was evaluated by nitrogen physiosorption using a BET Quantachrome Autosorb iQ3 instrument. The samples were degassed under vacuum at 80°C for 6 hours before analysis. The pore size was determined using the desorption isotherms in a relative pressure window of 0.35 to 0.99 and following the BJH approach.
Collection of real samples for ex situ and on-body measurements
The sweat samples used in all experiments were collected with a Macroduct sweat collector placed on the skin and firmly strapped in place during outdoor running exercises. Sweat secreted by the sweat glands was forced from the ducts under hydraulic pressure and flowed between the skin and the concave under the surface of the collector and into the micropore tubing spiral. The sweat samples were collected from healthy human subjects. The on-body measurements were performed in compliance with institutional review board guidelines. All subjects gave written, informed consent before participation in the study.
Fabrication of skin-like microfluidic device
The laser-patterned microfluidic device was fabricated using a multilayered fabrication approach. The base layer of the device consisted of a 125-μm-thick PMMA foil, while the skin-like laser pattern was defined on a 50-μm-thick PMMA foil. Laser ablation of 60 μm through holes on PMMA was performed using a high-power density focusing optics (VLS6.60, Universal Laser Systems) to achieve high resolution during laser ablation. A single layer of a double-sided medical-grade pressure-sensitive adhesive (PSA) was laser cut to define the branched microfluidic structure. The three layers were then manually aligned and laminated together to obtain stable and irreversible bonding of the PSA with the top and bottom PMMA layers.
Fabrication of laser-patterned microcapillary channels
Laser-patterned microfluidic channels were introduced into the medical-grade tape using a femtosecond laser (500 mW, 808 nm, 1 kHz; Spectra-Physics) to locally etch through the thickness of the tape without melting using a spot size of approximately 20 μm. A shutter was used to block the beam as the sample was translated by 100 μm in the x and y axes to fabricate an array of microfluidic pores.
Setup for on-body testing
Double- and single-sided medical-grade tapes were integrated with the MSM-based sensor surface after laser patterning to provide efficient sweat collection via capillary action. These laser-patterned channels and reservoir layers absorb and collect sweat to provide stable and reliable sensor readings while also preventing direct mechanical contact between the sensor and skin. During on-body tests, the newly generated sweat would fill the reservoir and was delivered to the sensor interface. The on-body measurement results were also consistent with ex situ tests using freshly collected sweat samples.
As a safety note, it is highly advised that the electrochemical measurement (output measurement) apparatus should be voltage- and current-limited to reduce the likelihood of unintentional electric shock.
Standard measurements using cortisol ELISA
The Human Cortisol ELISA Kit (Biotang Inc.) was used for the quantitative determination of the concentrations of human cortisol in sweat. The kit assays human cortisol in the samples by double-antibody sandwich ELISA. The human cortisol antibody-coated microplate was used. For the binding, biotinylated cortisol antibody, the horseradish peroxidase (HRP)– streptavidin enzyme conjugate was added to the wells. After washing to remove the unbound antibody enzyme reagent, the tetramethylbenzidine (TMB) substrate was added. The TMB substrate produced a blue color when the HRP enzyme was catalyzed. After reaching the desired color intensity, the reaction was terminated by the addition of an acidic stop solution, which changed the solution color from blue to yellow. The absorbance at a wavelength of 450 nm was measured by the microwell reader and the concentration of cortisol was calculated according to the standard curve. Standard kits and solutions were kept at 4°C unless otherwise used.