Simulated wound fluid exudate (BZ292) was obtained from Bio Chemazone (Alberta T6B 3P3 Canada) while the other reagents such as Prussian blue (PB), potassium hexacyanoferrate (III) {K₃ [Fe(CN)₆]}, potassium chloride (KCl), hydrochloric acid (HCl), iron (III) chloride (FeCl₃, thiol cross-linker acid solution, N′-ethylcarbodiimide hydrochloride (EDC), N-Hydroxysuccinimide (NHS), bovine serum albumin (BSA), graphene ink, silver/silver chloride (Ag/AgCl) paste, hydrogen tetrachloroaurate (III) hydrate (HAuCl₄), and trisodium citrate were purchased from Sigma Aldrich (St. Louis, MO). The multi-walled carbon nanotube (MWCNT) was obtained from ACS Material (Pasadena, CA). The Interleukin-6 and Interlukin-10 protein and antibody were purchased from R&D Systems (Minneapolis, MN). IL-6 and IL-10 proteins were spiked in simulated wound fluid (SWF) at varying concentrations ranging from 0.1 pg/ml to 1,000 pg/ml.

Cyclic voltammetry (CV) and chronoamperometry analyses were performed through CHI 660E electrochemical workstation (CH Instruments, Inc.). The following parameters were used for CV: range, −0.2–+2.0 V; scan rate, 50 mV/s; incremental potential, 0.01 V. The following parameters were used for CA: pulse amplitude, 0.5 V; time interval, 0.1 s, and run time, 20 s. It is to be noted that, similar CV results were obtained with our custom-made potentiostat (Figure 1B), as explained under the Results and Discussion section. A PrismCut Vinyl Cutter (USCutter) was used for cutting and screen printing the electrodes. For characterizing mechanical deformations of the sensor, a compact, motorized translation stage (Thorlabs) was used. The motorized base was programmed to bend the sensor from 0° to 90° angle and return to the initial 0° position. This entire cycle of bending was conducted 100 times.

Zeiss Supra 55VP scanning electron microscope (SEM) and Nicolet Avatar 360 E.S.P. ATR-FTIR spectrometer were utilized to characterize the AuNP decorated MWCNT coatings.

The electrochemical sensor design contained four layers: 1) Screen-printed electrodes (SPEs) having two working electrodes (WEs) for detecting two cytokines, one counter electrode (CE), and one reference electrode (RE), 2) a hydrophobic polyethylene terephthalate (PET) sheet patterned with the SPEs, 3) a medical-grade sterile wound dressing as the underlying substrate, and 4) a second medical-grade porous gauge for protecting the electrode surface.

A low-cost and roll-to-roll technique was used to manufacture the sensors on a large scale (Figure 2A) (Pereira et al., 2021). Figure 2B demonstrates the stepwise fabrication of the SPEs. The electrode patterns were designed in AutoCAD Fusion 360 software and then imported to a benchtop craft cutter for cutting the electrode patterns on an 85 μm thick polyvinyl chloride (PVC) sheet (HTVRont). The PVC sheet was covered with a transfer tape that worked as the masking layer. Firstly, the taped PVC sheet was loaded into the craft cutter, and the SPE patterns were cut through the sheet (Figures 2Bi–iii). The optimized parameters for the cutter blade were force 4.5 N and speed 30 mm/s. The blade was run twice over the PVC sheet to generate a precise pattern. Next, the transfer tape was peeled off from the WEs and the CE areas, and graphene ink was uniformly screen-printed to the exposed regions with a squeegee (Figure 2Biv). Likewise, the exposed regions of RE were screen printed with Ag/AgCl paste (Figure 2Biv). The electrodes were cured at 100°C temperature for 60 min in a convection oven. Afterward, the SPE pieces were removed from the base and transferred to a pre-cut 125 μm thick PET sheet (Grainger) (Figures 2Bv, vi). Finally, the PET sheet was placed on a sterile cotton dressing. A second medical-grade porous gauge was attached on the electrode surface such that wound exudates can get absorbed by the gauge and flow to the electrode surface, while simultaneously protecting the delicate wound area from foreign body response (due to the sensor electrodes). The cross-section of the sensor is shown in Figure 3A. The overall dimension of the sensing area was 2 cm × 1.4 cm. The screen-printing method with a benchtop craft cutter enables low-cost and rapid bulk manufacturing of the electrodes.

To selectively detect IL-6 and IL-10, the WEs were immobilized with a number of chemical layers. This work features gold nanoparticles decorated multiwalled carbon nanotube (AuNP-MWCNT) composite coating as the protein-sensitive layer. Preparation of the AuNP-MWCNT dispersion started with the formation of AuNPs using the Turkevich method (Turkevich et al., 1951; Shakila and Pandian, 2006). Briefly, 30 ml of 0.01% (w/v) tetra chloroauric (III) acid (HAuCl₄) was set to boil on a hotplate with constant stirring. After the boiling started, 1 ml of 1% sodium citrate was added slowly. The solution was stirred for 15 min to complete the reaction. The obtained dispersion was centrifuged at 14,000 rpm for 20 min to remove the unreacted HAuCl₄ and sodium citrate. The resulting precipitate was collected, which was redispersed in 200 ml of Milli-Q water. The above procedure resulted in mean AuNPs of 25 nm (Turkevich et al.,1951), which was also confirmed through scanning electron microscope images (Figure 3). We selected this specific size of AuNPs because it provided highest current response, as described in the Supplementary Table S1 and Supplementary Figure S2.

Synthesis of -OH functionalized MWCNT was carried out by treating the MWCNT with 3M nitric acid solution under constant stirring for 24 h. The functionalized MWCNT was washed thoroughly with Milli-Q water and subsequently dried at 80°C for 12 h in a convection oven (Ghica and Brett, 2013; Silva et al., 2018).

The AuNP-MWCNT composite coating was synthesized by dispersing equal volumes of AuNP and MWCNT stock solutions in 1% (w/v) chitosan dissolved in 1% (v/v) acetic acid solution and sonicating for 8 h. Four different AuNP-MWCNT dispersions (0.1, 0.5, 1 and 2% (w/v)) were prepared by following this procedure.

The WEs were first electrodeposited with Prussian Blue (PB) (worked as a redox mediator) and AuNP-MWCNT by running 200 voltammetric cycles for potential ranging from −0.3 to +1.2 V at a scan rate of 50 mV/s in a plating solution containing 2.5 mM K₃ [Fe (CN)₆], 2.5 mM FeCl₃, 0.1 M KCl, 0.1 M HCl and 1 mg ml⁻¹ AuNP-MWCNT. Four different dispersions of AuNP-MWCNT, 0.1, 0.5, 1 and 2% (w/v), were used to do electrodeposition on four electrodes (Shakila and Pandian, 2006; Noushin and Tabassum, 2022).

Secondly, the AuNP-MWCNT-PB immobilized WEs were coated with a thiol cross-linker [HS (CH₂)₄–COOH]. The thiol functional groups of the linker formed Au-S bonds with the AuNPs, while the carboxyl groups facilitated covalent binding to EDC molecules of the next layer to form an intermediate O-acylisourea. A 1 mM of thiolated linker solution was prepared in PBS (pH = 7) and 4 μL of this solution was drop cast on the WEs.

Thirdly, the WEs were functionalized with anti-IL6/anti-IL10 molecules via EDC-NHS coupling chemistry (Ali et al., 2018). In this regard, 0.2M EDC and 0.05M NHS solutions were mixed with a 1 mg/ml antibody solution at a 1:1 volume ratio. The NHS molecules reacted with primary amines of antibody to form amine-reactive stable NHS ester. A 4 μL of antibody solution was drop cast on the WEs that were stored inside a humid chamber at 4°C for 12 h. This step allows the conjugation of antibodies to the WE surface.

Finally, 2 mg/ml solution of bovine serum albumin (BSA) was drop-cast on the WEs to block the non-specific binding sites.

The AuNP-MWCNT coated and uncoated electrodes were characterized by scanning electron microscopy (SEM), Fourier transform infrared (FTIR) spectroscopy, cyclic voltammetry (CV), and chronoamperometry (CA).

The effects of different parameters including changes in the coating composition, wound pH, and bending motions, on the electrochemical behaviour of the sensors are explained in this section.

We developed an IoT-enabled potentiostat capable of measuring cyclic voltammetry responses with a predefined potential sweeping from −0.2 to 2.0 V and a scan rate of 50 mV/s. The detailed block diagram is shown in Figure 1C and explained in Supplementary Figure S4. The entire electronics was enclosed in a 3D printed box that was wearable on the arm (Supplementary Figure S1). The wearable potentiostat was used to measure the CV responses of the sensor for eight different concentrations of IL-10 and IL-6. The Blynk IoT interface was used to transmit the data wirelessly to a smartphone application (Figure 1C). Figures 8A,B show the cyclic voltammograms in response to IL-6 and IL-10 concentrations, respectively, while the resulting calibration curves are shown in Figures 8C,D. The CV calibration curves obtained by the custom-developed potentiostat were nearly the same as the calibration plots generated by the commercial CHI 660E potentiostat. Error bars represent the three consecutive measurements taken for each protein concentration. The coefficient of variance between the redox currents measured by the two systems (CHI 660E vs. the developed potentiostat) was < ±1%.

The developed potentiostat was also used to measure the dynamic response of the sensor, as shown in Figure 9A. The sensor was spiked with increasing IL-10 concentrations ranging from 0.1 pg/ml to 1,000 pg/ml. Four consecutive readings were taken for each concentration in a 6-min timeframe where the sensor response was stable. We determined the sensor’s accuracy for analyzing unknown protein concentrations in SWF. The sensor could accurately estimate the concentrations, with a high Pearson correlation coefficient of 0.997, suggesting the excellent reliability of the sensor (Figure 9B).

The reproducibility, repeatability, and reversibility characteristics were analyzed using the CV technique. In order to test for the reproducibility of our device, four identical sensors were tested with the same set of IL-10 levels, and the calibration curves were plotted accordingly (shown in Figure 10A). Our sensor indicated an acceptable reproducible behavior with a less than 8% coefficient of variance among the four calibration curves, which is vital for on-body measurements.

Sensor repeatability was tested by repeating the calibration of the same sensor four times in an interval of 1 h. Figure 10B demonstrates the four calibration plots obtained with the same sensor. Although the coefficient of variance of the intercept was calculated to be <0.1%, the slope values showed a nearly 10% deviation. This might be due to the continuous exposure of the sensor surface to an uncontrolled humid environment for more than 4 h. Such exposures can be avoided during on-body applications because a medical gauge covers the sensor surface (also explained in the Materials and Methods section). The similar calibration curves demonstrated the repeatable nature of the sensor.

The sensor also demonstrated excellent reversibility, as is illustrated in Figure 10C. The sensor was exposed to increasing, followed by decreasing concentrations of IL-10 protein biomarker, and the cycle was repeated four times. The peak oxidation current was almost the same for different cycles with less than 0.1% deviation.

The drift characteristic of the sensor was analyzed using three IL-10 concentrations, including 0.1 pg/ml, 50 pg/ml, and 1,000 pg/ml. The voltammograms were measured every hour over 12 h (Figure 11A). The sensor was stored at 4°C between two consecutive test sessions. The overall coefficient of variance was <0.5% indicating the minimal drift displayed by the sensor. The drift behavior was further analyzed in 5 min interval for the same set of IL-10 concentrations over an hour (Figure 11B). Again, a minute drift was observed with a coefficient of variance of only 0.001%.

To successfully apply the sensor in an actual wound, it is essential to evaluate the sensor response to interfering species present at the wound site. Therefore, the sensor was tested against some common interfering species found at the wound site, such as glucose (180 g/L), cortisol (100 ng/ml), and C-reactive protein (10 pg/ml) (Figures 12A,B). The sensor was spiked with three different SWF solutions: 1) a solution containing only interfering species, 2) 10 pg/ml of the target protein (IL-6 or IL-10) mixed with the interfering species, and 3) 500 pg/ml of the target protein (IL-6 or IL-10) mixed with the interfering species. In the absence of target proteins, no current variations were observed with respect to the baseline, while the introduction of target protein (IL-6 or IL-10) generated noticeable current peaks. Further, the effect of interfering species on the sensitivity and response time of the sensors were analyzed (as demonstrated in Supplementary Figures S5, S6). It was evident that in the presence of interferents, the sensitivity reduced by <8% and the response was delayed by only 135 s.

To evaluate the long-term stability of the sensor, three sensors were tested for three different IL-10 concentrations over a week. The test results are demonstrated in Figure 12C. The sensor was stored at 4°C after each test. The coefficient of variance in the sensor response was measured to be <0.1% over the first 3 days and approximately 5% for the remaining 4 days, indicating an acceptable stable response for on-body measurements under an appropriate storage condition. The stability of the AuNPs-MWCNT coating was further analyzed. The results are reported in the Supplementary Figure S7.