High sensitivity chemiresistive biosensor prepared via enzyme-catalyzed redox and nanoparticle conduction network

Abstract

The demand is increasing for compact, low-cost biosensors suitable for point-of-care diagnostics. In this study, we developed a novel chemiresistive biosensor based on a platinum (Pt) nanoparticle–polymer composite matrix functionalized with a creatinine enzyme cascade. The sensor detects creatinine through resistance changes triggered by redox reactions of enzymatically generated hydrogen peroxide at the Pt nanoparticle interface. Operating near the percolation threshold of metallic nanoparticles enhances sensor sensitivity, as it promotes the formation of efficient electron conduction paths through hopping and tunneling mechanisms. The simplified two-electrode structure of the device eliminates the need for a reference electrode, enabling miniaturization and facilitating fabrication. Both direct and alternating current measurements confirm that the electrical response arises from interfacial charge redistribution combined with bulk conduction network formation. The biosensor exhibits a wide detection range (1–300 mg/dL), fast response time (~35 s), and strong correlation between analyte concentration and electrical signal. This platform offers a promising approach for high sensitivity, high selectivity, real-time biosensing of creatinine and other biomarkers.

Introduction

Biosensors have garnered significant attention in recent years because of their potential for cost-effective detection across various fields¹. Biosensors integrate bio-recognition elements—such as receptors, enzymes, and antibodies—with sensing components, facilitating real-time, highly specific analyte detection² while ensuring high sensitivity and selectivity. Biosensors can be classified according to the type of transduction mechanism employed, including optical^3,4, thermal⁵, and electrochemical^6,7 biosensors. Of these, electrochemical biosensors, which measure parameters such as potential, current, or conductivity, are particularly promising for their ability to operate based on potentiometric, amperometric, or conductometric principles. However, their miniaturization is hindered by the need for a reference electrode, which stabilizes detection but also increases device size and cost. Although enzyme-based calorimetric biosensors offer an alternative approach to miniaturization⁸, they tend to have lower sensitivity and require thermal insulation to reduce noise⁹. Similarly, optical-detection biosensors are limited by their sensitivity to environmental interference³.

Chemiresistive sensors detect analytes through resistance changes in the sensing material, offering a promising platform for portable sensor design. These sensors have been developed and applied in various fields, including environmental monitoring¹⁰, food safety^11,12, and healthcare testing^13,14. Chemiresistive sensors typically consist of two electrodes bridged by a sensing material, eliminating the need for a reference electrode and thus enabling a compact and cost-effective sensor design. Studies on chemiresistive sensors primarily investigated the detection of gas-phase analytes¹⁵. However, utilizing chemiresistive sensors in liquid media is viable but mostly underexplored. The sensing mechanism of chemiresistive liquid sensors depends on the sensing material used. The conductivity of a carbon nanotube-based chemiresistor is influenced by charge transfer and electrostatic gating by molecules adsorbed onto its surface¹⁶. These interactions alter the hole concentration within the p-doped single-walled carbon nanotube network, causing observable changes in resistance. Functionalized few-layer graphene-based chemiresistors exhibit conductivity changes due to the formation of an analyte complex on a p-type graphene surface, increasing the hole concentration and, consequently, the current¹⁷. Another study synthesized a chemiresistor by combining graphene with a conductive polymer (poly(3,4-ethylenedioxythiophene) polystyrene sulfonate)¹⁸. The chemiresistor alters the resistance caused by analyte oxidation, thereby disrupting the polymer’s conductive pathways, causing resistance changes. The change in resistivity is contingent upon the chemiresistor’s response to analyte recognition, thereby significantly hindering the synthesis of chemiresistive sensors with high sensitivity, selectivity, and stability^19,20. Consequently, the recognition element of the analyte will be a crucial component of the chemiresistor. The integration of bio-recognition in biosensors can enhance measurement accuracy by achieving high sensitivity and specificity²¹.

In this study, an enzyme-based chemiresistive sensor was developed to evaluate analyte concentration by monitoring changes in chemiresistor resistance. Creatinine is a well-established biomarker in clinical diagnostics^22–24, making it an attractive detection target for chemiresistive biosensors. We specifically focus on creatinine in urine due to its potential for non-invasive collection, making it ideal for preventive screening by non-experts. Therefore, this study evaluates the 0-300 mg/dL range to demonstrate the biosensor’s potential as a monitoring instrument for random urine testing. Creatinine concentration can be analyzed through several methods, the most common of which is by spectrophotometric methodology using Jaffé’s reaction²⁵, offering high accuracy and reproducibility. However, this method is hindered by interferences from certain drugs and species, such as blood and urine²⁶. Furthermore, the detection device is too large for portable use and expensive to operate. Similar issues have been encountered in various healthcare detection systems. Hence, a detection method with a portable design, high sensitivity, fast response, and low operating costs needs to be established.

To address this challenge, we proposed a novel method for detecting analyte concentration by applying chemiresistive sensing. The sensing material consisted of metal nanoparticles dispersed in a non-conductive polymer at a concentration near the percolation threshold, a state maximizing the sensitivity of the resistive element to chemical disturbances. Enzymatic recognition offers high selectivity for the analyte and also enables reusability. Furthermore, the design can be adapted for the bio-recognition of other analytes. The proposed device was designed specifically for personal, repetitive use.

Chemiresistive biosensor working principle

Figure 1a illustrates the working principle of the chemiresistive biosensor, which functions by sensing the resistance changes in a solid-state resistive element (chemiresistor) caused by catalytic chemical reactions between enzymes and biomarkers. The chemiresistor consists of platinum (Pt) nanoparticles with a diameter of 4 nm, three enzymes, and a UV-patternable hydrogel composed of a copolymer of vinyl alcohol, which functions as a scaffold (see the “Preparation of solution for chemiresistor” section). The chemiresistor is bridged across a 10 μm gap between two electrodes on a glass substrate, with the area above the electrode gap, which is coated with a chemically resistant material, serving as the sensing area. When the analyte solution permeates the chemiresistor, it triggers enzymatic catalysis that generates hydrogen peroxide (H₂O₂). Subsequent redox reactions occurring on the surface of Pt nanoparticles induce charge transfer phenomena, modulating chemiresistor conductivity, thereby altering its resistance. Enzyme-based creatinine sensing employs three enzymes: creatinine amidohydrolase (CA), creatine amidinohydrolase (CI), and sarcosine oxidase (SO). The reaction between creatinine and the three enzymes induces the conversion of creatinine to creatine, creatine to sarcosine, and sarcosine to glycine, where hydrogen peroxide (H₂O₂) is liberated in the latter²⁷. The reaction sequence is as follows:

$$\mathrm{Creatinine}+{\mathrm{H}}_2\mathrm{O}\stackrel{\mathrm{CA}}{\to }\mathrm{Creatine}$$ 1

$$\mathrm{Creatine}+{\mathrm{H}}_2\mathrm{O}\stackrel{\mathrm{CI}}{\to }\mathrm{Urea}+\mathrm{Sarcosine}$$ 2

$$\mathrm{Sarcosine}+{\mathrm{H}}_2\mathrm{O}+{\mathrm{O}}_2\stackrel{\mathrm{SO}}{\to }\mathrm{Formaldehyde}+\mathrm{Glycine}+{\mathrm{H}}_2{\mathrm{O}}_2$$ 3

Fig. 1. Working principle of a chemiresistive biosensor.

a Structure of the chemiresistive biosensor and schematic of the resistive elements (chemiresistor). b Schematic of enzyme catalysis showing the production of hydrogen peroxide (H₂O₂) and reaction products by the analyte (creatinine). c Schematic of the redox reaction of H₂O₂ catalyzed by a platinum (Pt) nanoparticle

In the final step of the reaction, sarcosine oxidase utilizes flavin adenine dinucleotide (FAD) as a covalently bound cofactor. The reduced form (FADH₂) is reoxidized by molecular oxygen dissolved in the solution, enabling continuous enzyme activity. The released H₂O₂ contributes to the sensing mechanism because it can undergo redox reactions with Pt particles in the chemiresistors. Furthermore, Pt serves as a catalyst that facilitates both the oxidation and reduction of H₂O₂²⁸. Redox processes alter the electrical conductivity of the chemiresistor through two possible mechanisms. One mechanism (Fig. 1b) involves the predominant reduction reaction of H₂O₂, which generates charge carriers via the electrochemical reaction. The resulting conductivity changes are likely governed by charge transfer processes rather than by direct electron or hole transport. Charge transfer processes include localized electron hopping between adjacent Pt nanoparticles and/or ion-mediated conduction pathways within the polymer matrix. When an electric field is applied, these processes enhance charge mobility through the material, thereby effectively modulating the sensor’s electrical response.

Furthermore, polyvinylpyrrolidone (PVP)-modified Pt nanoparticles are predominantly negatively charged²⁹, making electrons the primary charge carriers. Since free electrons are not bound to specific atoms, they facilitate electrical current flow, resulting in higher conductivity. However, when the oxidation reaction of H₂O₂ predominates, electron-charged Pt particles and free electrons are released. Free electrons directly enhance chemiresistor conductivity, whereas electron-carrying Pt particles achieve this indirectly by facilitating electron hopping. Both redox reactions can occur in chemiresistive biosensors, thereby enhancing chemiresistor conductivity. A chemiresistor with higher creatinine concentrations produces more H₂O₂, thereby causing more redox reactions and increasing chemiresistor conductivity. This concentration-dependent conductivity change of the chemiresistor thus forms the basis for creatinine detection.

Experimental setup and schematic diagram of chemiresistive biosensor

Biosensor performance and response to the analyte were assessed by implementing a series of experiments, including direct current (DC) resistance evaluation and alternating current (AC) impedance evaluation (Fig. 2a and b; see “Materials and Methods” section). The volume of creatinine solution used for DC voltage/Voltage scan is 30 mL, whereas for EIS measurement, it is 100 μL. Figure 2c and d present optical micrographs of the electrodes before and after chemiresistor synthesis.

Fig. 2. Fundamental measurement setup and biosensor images.

Measurement setup used to evaluate the biosensor by (a) a direct current (DC) voltage/Voltage scan and (b) by electrochemical impedance spectroscopy. The electrodes pattern of the sensor c before and d after coating the chemiresistive biosensor

Results

Current–voltage characteristics of the chemiresistive biosensor

The current–voltage (I–V) characteristics of the chemiresistor were evaluated in creatinine solutions with concentrations ranging from 0 to 300 mg/dL (Fig. 3a). The I–V curves exhibited clear nonlinearity and evident hysteresis, indicating significant dependence on analyte concentration (Supplementary Fig. S3). In addition, the scanning results demonstrate an exponential shape, which is the feature of a tunneling-based conduction mechanism. Therefore, the resistance of the chemiresistor was further evaluated using a tunneling electron model (Fig. 3b; see the section on I–V characteristics measurement in “Materials and Methods”). The model-fitted resistance decreased with increasing creatinine concentration, aligning well with the expected working principle of the biosensor. Specifically, higher creatinine concentrations increased the H₂O₂ generated via enzymatic reactions. The generated H₂O₂ molecules promoted electron transfer at the Pt nanoparticle interfaces, thereby increasing the overall chemiresistive sensor conductivity. This enhanced charge transfer significantly reduced resistance with increasing analyte concentration. At higher creatinine concentrations, the resistance variation became saturated, which may be caused by certain reasons. One is Michaelis–Menten saturation³⁰, where the enzyme reaction rate is limited at high analyte concentrations. The second reason may involve the saturation of Pt surface reactions and the nonlinear characteristics of the carrier conduction pathways. Diffusion, interfacial capacitance, and percolation effects may also contribute to this complex phenomenon.

Fig. 3. Current–voltage characteristics of the chemiresistive biosensor.

a Current–voltage (I–V) characteristics of the chemiresistive biosensor tested in creatinine solutions with different concentrations. b The resistances of the chemiresistive biosensor are calculated from the I–V curves using the tunnel injection equation

Evaluation of the chemiresistive biosensor by AC impedance

Figure 4a presents the Nyquist plots of the chemiresistive biosensor at varying creatinine concentrations. Specifically, the inset presents the wide-frequency-range plots, which exhibit a clear two-semicircle pattern. The high-frequency semicircle clearly decreased in diameter with increasing creatinine concentration. To analyze this behavior, we adopted an equivalent circuit model composed of two R-CPE (R: resistor, CPE: constant phase element) in series, i.e., (R₁/CPE₁) + (R₂/CPE₂) (Fig. 4b). The R-CPE model consists of a resistor (R) in parallel with a constant phase element (CPE), which accounts for the non-ideal capacitive behavior often observed in systems with structural heterogeneity or surface roughness^31,32. The impedance of CPE_s is expressed as follows:

$${\mathrm{CPE}}_{\mathrm{s}}=1/\left[{Q_s\left(j\omega \right)}^{n_{\mathrm{s}}}\right],\left(s=1\mathrm{or}2\right)$$ 4

where Q_s (s = 1 and 2 for high- and low-frequency semicircles, respectively) is the pseudo-capacitance; ω is the angular frequency; and n_s(0 < n ≤ 1) is an empirical factor indicating the deviation from ideal capacitive behavior. As demonstrated by the fitting result (Supplementary Fig. S4), R₁ systematically decreased with increasing creatinine concentration (Fig. 4c), while Q₁ and n₁ remained relatively stable (Q₁ ≈ 5.2 × 10⁻⁹ $\text{F}\cdot {\text{s}}^{\left(n_1-1\right)}$; n₁≈ 0.82); thus, R₁ indicates the charge transport resistance through the polymer matrix among the Pt nanoparticle network. As shown in Fig. 4f, R₂ exhibited minimal variation across the tested concentration range, indicating that the low-frequency impedance component is largely insensitive to analyte concentration. Furthermore, Q₂ and n₂ remained nearly constant (Q₂ ≈ 2.2 × 10⁻⁷ $\text{F}\cdot {\text{s}}^{\left(n_2-1\right)}$; n₂≈ 0.86), signifying that the associated interfacial polarization processes remained stable and were not significantly influenced by the presence of creatinine. This result implies that the concentration-dependent impedance response is governed by the high-frequency semicircle (R₁/CPE₁), whereas the R₂/CPE₂ branch primarily indicates a background or parasitic contribution related to intrinsic matrix or interface properties.

Fig. 4. Chemiresistive biosensor evaluation by electrochemical impedance spectroscopy (EIS).

a Nyquist plot of the biosensor measured at different creatinine concentrations, showing a decreasing semicircle diameter with increasing creatinine concentration. b Equivalent circuit model used to fit the EIS data, consisting of two resistor–constant phase elements (R-CPEs) representing bulk and interfacial processes. Concentration dependence of c the high-frequency resistance R₁, d the constant phase element Q₁, and e the empirical factor n₁ associated with fast charge transfer. Concentration dependence of f the resistance R₂, g the constant phase element Q₂, and h the empirical factor n₂, corresponding to the low-frequency semicircle and slower interfacial or diffusion-limited processes

The charge transport resistance R₁ exhibited the most significant change among all circuit elements, decreasing with increasing creatinine concentration. This behavior is attributed to enhanced redox activity at the Pt nanoparticle interface, which was driven by the enzymatic generation of H₂O₂ at higher creatinine levels. This accelerated charge transfer reduces resistance and increases conductivity. Notably, the concentration-dependent trend of R₁ aligned well with the resistance changes observed in the I–V characteristics (Fig. 3b), further validating that R₁ reflects the sensor’s primary response to creatinine.

The first constant phase element, Q₁, arises from both the bulk dielectric properties of the water/polymer matrix and the interfacial electric double-layer capacitance at the Pt nanoparticle–matrix interface. This capacitance component is essentially independent of the analyte concentration.

The second constant phase element, Q₂, indicates slower polarization processes and is primarily attributed to the deep interfacial or space–charge polarization effects within the composite, which may arise from three reasons. One is proton charge accumulation at extended interfaces or grain boundaries within the percolated Pt network. Second is ionic diffusion and polarization within the water-rich domains absorbed into the polymer matrix. Third is macroscopic heterogeneity introduced by water channels, enzyme clusters, or incomplete Pt dispersion, resulting in frequency-dispersed capacitive behavior at lower frequencies. Thus, Q₂ represents the contribution of macroscopic interfacial polarization or heterogeneous diffusion-related dielectric relaxation in the hydrated nanocomposite structure, which manifests at lower frequencies and does not interfere with the high-frequency response indicated by Q₁.

The resistance R₂ showed minimal variation with increasing creatinine concentration, indicating that the underlying impedance mechanism associated with this element is relatively unaffected by analyte levels. Therefore, R₂ primarily reflects a stable background process, such as long-range ionic diffusion or deep interfacial polarization within the polymer matrix, which is minimally affected by changes in the local redox environment. The constancy of R₂ implies that the low-frequency region of the impedance spectrum is governed by intrinsic material properties rather than by analyte-induced modulation of charge transport.

Although H₂O₂ generation could, in principle, induce the partial deactivation or surface coverage of Pt catalytic sites, potentially altering interfacial charge transfer dynamics, we found that the R₂ component remained largely unchanged across creatinine concentrations. This stability suggests that such surface effects marginally influence low-frequency polarization resistance under the test conditions. Furthermore, any accumulation of reaction byproducts may have a limited impact on the electrostatic landscape at the Pt–polymer interface. Consequently, the dominant impedance change was primarily attributed to the R₁ component, reaffirming that the chemiresistive response of the device is governed by fast interfacial charge transfer processes rather than by deep interfacial polarization.

Previous DC resistance measurements were determined from the I–V curve. The DC resistance reflects steady-state current flow, including contributions from bulk conduction, charge hopping or tunneling mechanisms, and contact resistance. In contrast, charge transport resistance R₁ was primarily determined by the kinetic barrier to charge transfer at the electrode–electrolyte interface and the reaction resistance of redox reactions. Therefore, AC measurement results are more significantly affected by enzyme activity and H₂O₂ concentration.

Although a simplified two-electrode EIS setup was employed, we acknowledge its limitations for quantitative mechanistic interpretation. The equivalent circuit model (two R–CPE elements) serves primarily as a phenomenological fit to the composite system. R₁ reflects high-frequency interfacial responses influenced by redox activity, while R₂ captures slower dielectric relaxation. Warburg-type diffusion or electrode asymmetry effects, which may contribute to low-frequency features, were not explicitly separated. Thus, the EIS analysis is interpreted qualitatively, without implying precise electrochemical assignments.

DC potential experiment

Compared with more complex techniques, the simple DC measurement method is advantageous because of its ease of use and rapid response. DC measurements were performed under a constant potential using creatinine solutions with varying concentrations. As shown in the inset of Fig. 5, the typical response to creatinine solutions was ~35 s and was characterized by a rapid change and subsequent stabilization of the output current. The biosensor output current showed an upward trend with increasing creatinine concentration, confirming the working principle that higher creatinine concentrations increase the biosensor’s conductivity, producing a higher output current. A significant difference in output current was also observed at low concentrations, indicating the biosensor’s capability of detecting small variations in creatinine levels. Additionally, the error bars in the results represent the variability between different biosensor devices. Despite the considerable variation in the biosensor’s response at high concentrations, a uniform trend is exhibited across all biosensors with respect to creatinine concentration alterations. In order to understand the enzyme’s capacity to validate the proposed mechanism, the same tests were conducted using a non-enzyme biosensor. The results demonstrated a clear distinction between the enzyme- and non-enzyme-containing biosensors, suggesting that the variations in output current responses were attributable to the enzymatic reaction.

Fig. 5. Chemiresistive biosensor evaluation with the application under a constant DC voltage.

Average output currents were measured at 150 s after immersion in creatinine solutions with different concentrations. The figure demonstrates the typical response of the chemiresistive biosensor’s time-dependent output current when tested with creatinine solutions under a constant voltage. A notable distinction was observed between the enzyme- and non-enzyme-containing biosensors in their responses to variations in creatinine concentrations

Real-time response to creatinine concentration changes

A simple DC measurement approach is more beneficial for practical use than the impedance spectroscopy. Here, the current response was observed under a constant DC voltage of 0.5 V applied to the chemiresistor, for testing creatinine concentrations of 0–300 mg/dL, with each concentration applied for 100 s. Before testing, the biosensor surface was cleaned with deionized (DI) water to prevent cross-contamination between samples. The output current increased significantly with increasing creatinine concentration (Fig. 6). However, during testing, the solution was absorbed into the chemiresistor, which contains a polymer. This resulted in a minor hysteresis, likely attributable to trace retention of the analyte. Nevertheless, this issue did not compromise the biosensor’s capacity to measure distinct, absolute concentrations in real-time.

Fig. 6. Real-time response of the chemiresistive biosensor to creatinine concentration changes (0–300 mg/dL).

The observed output current increased stepwise with some fluctuations, indicating that the biosensor’s conductivity changed in response to creatinine concentration

Discussion

In this study, we developed a chemiresistive biosensor based on a Pt nanoparticle-embedded PVA-based matrix functionalized with creatinine-reactive enzymes. The sensing mechanism operates via changes in electrical resistance induced by enzymatically generated H₂O₂, which undergoes redox reactions at the Pt surface. These reactions modulate the charge transport pathways within the material. The device meets essential criteria for point-of-care diagnostics³³, offering a rapid, highly sensitive, and user-friendly biosensor platform.

To elucidate the sensing mechanism, we investigated both the DC resistance and AC impedance characteristics of the biosensor. DC measurements reflect steady-state electronic transport across the sensor matrix, including contributions from electron hopping, tunneling, and percolation pathways. In contrast, AC impedance analysis, particularly the charge transport resistance, captures dynamic processes at the Pt nanoparticle–electrolyte interface, including redox kinetics and interfacial polarization.

Two key physical phenomena dominate the electrical behavior of the composite: Maxwell–Wagner–Sillars (MWS) interfacial polarization and percolation-driven DC conductivity^34,35. MWS polarization occurs in heterogeneous dielectric systems containing an insulating matrix with embedded metallic nanoparticles. Upon exposure to an alternating electric field, interfacial charges accumulate at boundaries with varying permittivity and conductivity. This results in frequency-dependent impedance relaxation, which is especially pronounced at low frequencies. MWS polarization can be modeled by the equivalent circuit (R₁/CPE₁ + R₂/CPE₂), as described previously^36,37.

Although the device operates without a reference electrode, the I–V curves do not display the nonlinear or diffusion-limited behavior typical of Faradaic processes. Additionally, measurements were conducted in low-ionic-strength DI water, reducing the likelihood of significant H₂O₂ redox activity. These findings suggest that the main conduction mechanism arises from the polymer–Pt network rather than electrode-driven electrochemistry.

In our biosensor, the absorption of aqueous creatinine solutions increased ionic mobility, enhancing interfacial polarization³⁸. These mechanisms can be modeled with a resistor–capacitor network in which nanoparticle clusters act as distributed resistive and capacitive elements. The two R/CPE elements in our system were defined as follows: R₁/CPE₁ is the bulk impedance through the hydrated PVA-Pt network, while R₂/CPE₂ indicates the redox reaction-dominated interfacial impedance at Pt nanoparticle–electrolyte interfaces.

Importantly, the redox activity of H₂O₂ synergizes with the percolated conductive network. The generation of charge carriers through enzymatic reactions and their subsequent transport via percolated pathways enhances current flow. This dual-mode conduction–interfacial capacitive polarization and bulk electron percolation is responsible for the biosensor’s high sensitivity, rapid response, and nonlinear resistance–concentration characteristics.

These findings collectively demonstrate that rational material design near the percolation threshold, combined with controlled interfacial electrochemistry, offers a robust strategy for optimizing chemiresistive biosensor performance.

The performance of this sensor was compared with other electrochemical methods, as shown in Table 1. The chemiresistive biosensor exhibited a broad detection range and fast response. Although the biosensor was not tested on real urine or blood samples in this study, its enzymatic cascade design suggests high selectivity. Future work will validate its performance in complex biological matrices. The detection volume used in this study was limited to 100 μL, resulting in extremely small quantities of reaction byproducts such as formaldehyde and hydrogen peroxide. This microscale configuration minimizes safety concerns and supports the biosensor’s suitability for point-of-care applications. This working principle may be applied to develop inexpensive and disposable biosensors, which would be suitable for various biomarker detection purposes.

Table 1.

Performance comparison between this research and other detection methods

Detection method	Detection range	Response time	Ref.
Amperometric	0.05–1.4 mM	3 min	42
Amperometric	1–10 mM	20 s	43
Amperometric	0.01–10 mM	N/A	44
Potentiometric	1–50 mM	15–60 min	45
Potentiometric	0.177–44.2 mM	3 min	46
Chemiresistive (This work)	0.009–26.5 mM	35 s	-

Materials and methods

Preparation of the solution for chemiresistor

All chemicals were prepared in deionized distilled water. The electrical conductivity of deionized distilled water was measured at 2.73 μS/cm, which corresponds to an estimated ionic strength of approximately 4.4 × 10⁻⁵ mol/L. Creatinine powder, sarcosine oxidase (SO) (S7897-1KU) from Bacillus sp., and creatininase (C3172-1KU) from microorganisms were purchased from Sigma Chemical Company. Creatinase (C2-AT) from recombinant Escherichia coli was purchased from Kikkoman Biochemifa company. PVP-coated Pt nanoparticle solution was purchased from Prexnet, Tanaka Kikinzoku company. PVA-based polymer (Biosurfine AWP-MRH) was purchased from Toyo Gosei Company. Sodium hydroxide was purchased from FUJIFILM Wako Pure Chemical Corporation, Japan.

The chemiresistive solution consisted of three enzymes (creatinine amidohydrolase [CA], creatine amidinohydrolase [CI], and sarcosine oxidase [SO]), the Pt particle solution, and the PVA-based polymer. This PVA-based polymer is a water-soluble photopolymer, which forms the cross-link with enzymes and nanoparticles under UV light. To prepare the chemiresistor, 0.5 mg CA, 5.7 mg CI, and 2.3 mg SO were dissolved in 2 mL distilled water, according to the ratio described previously³⁹. The Pt particle solution (3 mL) and PVA-based polymer (1 mL) were then added to the enzyme solution. The volume of the Pt particle solution was measured and found to be within the range required to reach the percolation threshold. The mixture was then homogenized in an ultrasonic homogenizer at 50 Hz for 2 min.

Fabrication of chemiresistive biosensor

The fabrication process of the chemiresistive biosensor is based on a photolithography process. Chromium and gold (Cr-Au) films with thicknesses of 20 nm and 200 nm, respectively, were deposited on a 300-µm-thick glass substrate via sputtering for use as sensing electrodes. A positive photoresist (Tokyo Ohka Kogyo Co., Ltd., Kanagawa, Japan, OFPR 200 cp) was then deposited onto the Cr-Au films using the spinner. After photolithography, the electrodes were etched by wet etching onto the Cr and Au films. Copper wires were connected to the electrode pads using a conductive paste. After fixing the wafer on the board, 2.5 µL of a solution containing a polymer, Pt nanoparticles, and enzymes was coated onto the gap at the sensing area, forming a thin film. The sensing area was defined as the region of the biosensor coated with the chemiresistive solution (4 × 2 mm²). The PVA-based polymer was polymerized and immobilized by UV irradiation for 200 s, employing a wavelength of 365 nm and an intensity of 11.8 mW/cm². All electrodes and metal lines outside the sensing area were insulated with Kapton tape to prevent unexpected surface reactions, thereby enabling the chemiresistive biosensor to be immersed in solutions for testing.

Impedance measurement

The chemiresistive sensor was then immersed in creatinine solutions with varying concentrations, and its impedance was evaluated by I–V characteristic measurement, DC resistance measurement, and AC impedance measurement. A magnetic stir bar was used to maintain the homogeneity of the solution, while a computer was programmed to control the voltage and record the results.

I–V characteristics measurement

The I–V characteristics of the biosensor were measured by applying a voltage to the chemiresistor using a voltage source (Keithley 2400). The applied voltage ranged from −1 to 1 V in creatinine solutions across the device, and the resulting flowing current was measured. The I–V curves exhibited clear nonlinearity and significant hysteresis. This non-ohmic behavior was analyzed quantitatively using the exponential tunneling model proposed by Hampl et al.⁴⁰, in which the current and voltage interact in a tunneling junction as follows (See supplementary information, Fig. S3):

$$I=a_{\mathrm{t}}{V}^2\exp \left(-\frac{b_{\mathrm{t}}}{V}\right),$$ 5

$$R=\left(\frac{1}{a_{\mathrm{t}}V}\right)\exp \left(\frac{b_{\mathrm{t}}}{V}\right),$$ 6

where I is the current, V is the applied voltage, R is the resistance, and a_t and b_t are constants, independent of the applied voltage.

Resistance values were calculated using the forward scan data from Fig. 3a and by applying Eq. (5). The results are shown in Fig. 3b.

DC current measurement

To ensure that the biosensor responds to the different creatinine concentrations and suppresses unwanted electrolysis, a DC constant voltage of +0.5 V was applied to the biosensor. Creatinine concentrations ranging from 0 to 300 mg/dL were prepared, and the fabricated biosensor was then immersed in each creatinine solution. The biosensor’s output current was measured for 150 s and then compared to the corresponding creatinine concentrations (inset of Fig. 5). During the measurement, the creatinine solution was homogenized using a magnetic stirrer. The biosensor’s output current was affected by a series of enzyme-catalyzed reactions.

AC impedance measurement

Electrode impedance spectroscopy (EIS), an effective method for evaluating electrode impedance, was performed using an electrochemical measurement system (Hokuto Denko Co., HZ-7000) to evaluate the impedance of the chemiresistor against the creatinine solution concentration⁴¹. For the measurements, the frequency of the applied AC voltage, with an amplitude of 100 mV, ranged from 500 kHz to 1 Hz. A potential bias of +0.2 V was applied to the chemiresistor to facilitate the electrochemical reaction on the Pt nanoparticles. The creatinine solutions were initially adjusted using 0.01 M NaOH to a final pH of 7.0 ± 0.1. To minimize potential pH drift in the unbuffered environment, freshly prepared solutions were used for each measurement within 10 minutes of adjustment.

Author contributions

Y.K. performed most experiments. Y.K. and T.O. analyzed data, T.A. I.V. proposed experimental design, N.T. and I.V. analyze data, Y.K. and T.O. wrote the manuscript with comments from all authors.

Data availability

All data that support the results of this study are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41378-025-01155-3.