Introduction

Advances in bioelectronic technologies have enabled the integration of electrodes with living tissues to advance human healthcare1,2,3. More recently, similar strategies have been applied to plants, where bioelectronic devices are used to monitor physiological signals for use in agriculture and environmental monitoring, building platforms for smart farming practices4,5,6,7. On-leaf wearable sensors have emerged as a promising approach for directly monitoring these physiological signals8,9. Prior work has investigated the use of wearable sensors to monitor plant growth10,11, leaf microclimate7,12,13, organic and inorganic chemical markers14,15, and phytohormones16,17,18. This method of on-leaf sensing also enables monitoring of plant electrical signaling which provides information about plant health and responses to environmental stimuli such as photosynthesis, drought stress, and wounding19,20,21. Plants utilize electrical signaling to enable long-distance communication between cells22,23,24, which plays a crucial role in their ability to sense and respond to stimuli such as touch25,26,27, herbivore-induced wounding28,29, light exposure30,31, and temperature fluctuations25,32,33. These electrophysiological signals are inherently self-generated and provide real-time feedback on complex environmental stimuli. When harnessed with stable and conformal electrodes, these electrical signals can be transduced to create a nature-based sensing network, where each plant serves as a responsive node for real-time distributed sensing. A primary issue in monitoring plant electrophysiology, however, has been the challenge of creating biocompatible and robust electrodes capable of recording and stimulating plant electrical activity over extended periods of time34,35. Certain touch-sensitive plants, such as the Venus flytrap (Dionaea muscipula, VFT) and the Mimosa pudica, are capable of producing electrical signals that trigger thigmonastic movements, allowing them to serve as models for studying plant electrophysiology and developing a platform for biological integrated sensor networks36. The VFT serves as an excellent model for plant electrophysiology research and for developing plant-electronic interfaces that can be adapted for broader applications in smart agriculture and distributed environmental sensing due to its well-characterized and reproducible action potentials (APs). These APs are generated in response to the stimulation of mechanosensitive trigger hairs inside the lobes of the trap36,37, where if two APs are elicited within 20 s, the stimulated lobes of the VFT will close38.

Though widely used, conventional Ag/AgCl electrodes are bulky, invasive, and poorly suited for long-term interfacing on plant tissues, such as the inside of VFT lobes39,40,41. In order to address this issue, studies have explored conformable electrodes, with a focus on monitoring and applying APs from the outside surface of the VFT lobes. Action potentials have been successfully recorded for short durations, but sustained adhesion to plant tissues and operation over extended time periods remains an important challenge40,42. In this work, we present an adhesive bioelectrode that integrates inkjet-printed PEDOT:PSS traces with a poly(vinyl alcohol) (PVA) hydrogel substrate specifically designed to adhere to plant tissues such as the inner surface of a VFT lobe using a thin layer of methyl cellulose (MC). While PEDOT:PSS-based bioelectronics have been applied to plant tissue in prior studies40,42,43,44, our approach differs in combining a printed conductive trace with a mechanically adhesive hydrogel that forms a reversible, conformable, non-invasive interface suitable for both recording and stimulation. When combined with cost-effective electronics, this forms the foundation for a biologically integrated environmental sensing platform. Compared to rigid Ag/AgCl electrodes, the gel electrodes exhibit excellent biocompatibility as they do not induce necrotic tissue formation after continuous contact over many days, and show enhanced long-term performance with minimal impedance change when measured over 120 hrs. The adhesive gel electrodes are also capable of monitoring environmental variables, such as diurnal cycling, insect activity, and temperature fluctuations, and can deliver electrical stimulation to close the lobes of other VFT plants remotely. These results demonstrate the compatibility of this materials set with plant tissues, and the ability of these electrodes to monitor plant conditions continuously. Remote actuation additionally enables communication between plants which could serve as the foundation for distributed environmental sensing systems, where multiple plants act as interconnected sensor nodes that collectively report changes in conditions22,23,24,45. The integration of conformal gel electrodes with wireless electronics that transduce electrophysiological signals from plants paves the way for developing intelligent, self-sustaining nature-based sensor networks.

Results and discussion

Printed electrode optimization

PEDOT:PSS traces were first inkjet printed and then transferred to a flexible, uncrosslinked PVA substrate (Fig. 1) to act as gel electrodes for plant electrophysiological monitoring. The three inkjet-printed layers that made up each trace were deposited using a Dimatix DMP-2800 inkjet printer using wet-on-wet printing onto a glass slide, then dried on a hot plate at 90 ∘C under ambient conditions for 10 min. The final thickness of the printed trace was 1.3 μm  ± 0.02 μm (n = 10). Once dried, a PVA layer was drop-cast on top of the traces and left to dry for 24 h under ambient conditions. The electrodes were then peeled off the glass slide, and a thin wire was attached using silver epoxy. To evaluate the performance of the gel electrodes, which were adhered to the plant tissue using methyl cellulose, we compared their biocompatibility, adhesion, and long-term stability, which are all crucial for continuous electrophysiological monitoring, against Ag/AgCl wire electrodes. The adhesive gel electrodes showed improved biocompatibility compared to the commercial electrodes, as seen in Fig. 2a, over nine days. The bioelectrodes did not cause any visible necrosis or tissue damage on the lobe, while the commercial electrodes caused significant amounts of necrotic tissue. To quantitatively assess the impact of the electrodes on the plant tissue, Pulse Amplitude Modulated (PAM) chlorophyll fluorometry was performed at the same intervals on different plants for both the gel and rigid Ag/AgCl electrodes (n = 3 for each). PAM measures the fluorescence of chlorophyll to evaluate photosynthetic performance, and Fv/Fm represents the maximum quantum efficiency of Photosystem II, providing a sensitive indicator of plant stress and overall photosynthetic efficiency46. The Fv/Fm values taken as controls showed no statistically significant difference between the plants selected for commercial or gel electrodes before placing them (0.79 ± 0.3 vs. 0.78 ± 0.3, p = 0.69) confirming that there was no baseline bias (Fig. S1). At day 0 there were no significant differences between plants in contact with the commercial electrode vs. the gel electrode (0.79 ± 0.02 vs. 0.81 ± 0.02, p = 0.46). At day 3 the Fv/Fm values for the commercial electrode were lower than those for the gel electrode, although this difference was not significant (0.70 ± 0.03 vs. 0.79 ± 0.03, p = 0.11). Statistically significant reductions in Fv/Fm for plants with the commercial electrode compared to those with the gel electrode were observed on day 5 (0.64 ± 0.02 vs. 0.79 ± 0.02, p = 0.005) and day 9 (0.41 ± 0.03 vs. 0.78 ± 0.03, p = 0.001) confirming the results from the visual analysis. The combination of the Fv/Fm values reducing to  ~ 0.4 and the visual necrotic tissue suggests that the damage was severe and likely irreversible47. In contrast, this decline in Fv/Fm was not observed when the gel electrode was used. Overall, Fv/Fm values closely tracked visible damage: leaves showing necrosis exhibited low Fv/Fm, whereas leaves without visible damage maintained higher values. This tissue damage suggests that the commercial electrode was cytotoxic to the plant when in contact with the tissue over extended periods. Additionally, we assessed the impact of each electrode on the plant’s photosynthetic efficiency over time by performing a Dunnett’s test to compare the Fv/Fm means to the controls recorded before day 0 for each electrode (Fig. S2). For the 9-day experiment, the gel electrode had no significant impact on the photosynthetic efficiency of the plant compared to the control (day 0: p = 0.58; day 3: p = 0.89; day 5: p = 0.94; day 9: p = 0.997). In contrast, there was a significant decrease in the photosynthetic efficiency by day 9 of the plants with the commercial electrode compared to the control (day 0: p = 0.9999; day 3: p = 0.18; day 5: p = 0.026; day 9: p < 0.0001). These results indicate that the small gel electrodes had no observable impact on the plants’ photosynthetic abilities compared to the conventional Ag/AgCl electrodes.

Fig. 1: Fabrication steps for manufacturing and adhering the gel electrode platform onto the inner surface of a Venus flytrap lobe.
Fig. 1: Fabrication steps for manufacturing and adhering the gel electrode platform onto the inner surface of a Venus flytrap lobe.The alternative text for this image may have been generated using AI.
Full size image

Fabrication steps for manufacturing and adhering the gel electrode platform onto the inner surface of a Venus flytrap lobe.

Fig. 2: Bioelectrode adhesion, mechanics, and electrical stability on Venus Flytrap lobe.
Fig. 2: Bioelectrode adhesion, mechanics, and electrical stability on Venus Flytrap lobe.The alternative text for this image may have been generated using AI.
Full size image

a Photographs of a commercial electrode and a bioelectrode attached to the outside of Venus flytrap lobes for 9 days (scale bars are 5 mm). b A stress-strain plot of a bioelectrode adhered to a glass slide with methyl cellulose under uniaxial tension and (c) a photograph of the electrode adhered to the glass slide before stretching (scale bar is 5 mm). d A photograph of the electrode configuration for the long-term electrochemical impedance spectroscopy (EIS) experiment (scale bar is 5 mm) and (e) the impedance modulus at 1 kHz for the commercial and adhesive gel electrodes as a function of time (insert is an expanded view of the same data).

Due to their small size, the adhesive electrodes were able to be adhered to the inside of the VFT lobe, using methyl cellulose applied to the PVA gel and then placed onto the plant tissue near the bases of the mechanosensitive trigger hairs. In contrast, the Ag/AgCl wire electrodes needed to be placed in contact with the outside of the flytrap’s lobe as they could not be placed on the inside of the lobes due to their size, which frequently stimulated the trigger hairs, causing the closure of the lobes. The commercial electrodes relied on a gel to maintain constant contact with the plant tissue. However, due to their inability to adhere to the plant tissue, these electrodes easily lost connection due to mechanical perturbation. This loss of connection occurs as the Ag/AgCl wire electrodes rest against the plant tissue, unadhered, while the gel electrodes are physically adhered to the plant tissue using methyl cellulose. A tensile test was conducted to analyze the properties of the adhesive electrode under uniaxial tension. The electrode was adhered to a glass slide using a thin layer of methyl cellulose, was left to dry for an hour, and then tested using a Test Resources Dynamic and Fatigue Test Machine (Figs. 2b-c). The interfacial shear modulus of the adhesive gel/glass interface was  ~ 19 kPa, showing that the adhesive can withstand deformations without failing. The bioelectrode began to delaminate from the glass slide at a strain of 10% and a shear stress of 145 kPa. During delamination, the average shear stress was 153  ± 7 kPa. The electrode fully delaminated at a strain of 47%, demonstrating that the electrode is mechanically robust enough to maintain contact with the plant tissue under moderate stress. To address the electrode’s adhesive performance under cyclic fatigue, an electrode was adhered to a glass slide then stretched to 10% strain for 1000 cycles under uniaxial tension at a rate of 30 mm min−1 The average recorded interfacial shear stress was 15.3 kPa ± 1.90 kPa (n = 4, Fig. S3) demonstrating the electrode’s ability to remain adhered under repeated mechanical loading. To further address the electrodes mechanical stability on the VFT lobe, a tensile test under uniaxial tension with an electrode adhered to an VFT lobe was performed. The plant tissue tore and failed at a maximum shear stress of 115 kPa, which occurred without any delamination of the electrode, indicating electrode adhesion was stronger than the VFT lobe tissue (Supplementary Video 1 and Fig. S4). The long-term stability of the commercial and the printed adhesive gel electrodes was also evaluated using electrochemical impedance spectroscopy (EIS) in a climate-controlled grow chamber in a laboratory setting. A commercial Ag/AgCl wire electrode was placed in contact with one lobe of a plant, while the adhesive electrode was attached to the inside of a lobe of another leaf on the same plant. Although the electrodes were placed on separate lobes, they were exposed to the same environmental conditions as the experiments were run simultaneously. The commercial and adhesive electrodes were attached to a common counter electrode, a needle in the hydrated sphagnum moss growing media (Fig. 2d). The commercial electrode exhibited a  ~30x increase in impedance modulus and a relative standard deviation (RSD) of 74.6% at 1 kHz over the course of the experiment (120 hrs). In contrast, the adhesive gel electrode maintained a relatively stable impedance profile with a 2.94% RSD over the same period, remaining at ~1x its initial impedance (Fig. 2e). The EIS data at 0 hrs and 120 h was also fit using a double-distributed circuit element (DCE) model: R0 − (R1∥CPE1) − (R2∥CPE2). R0 represents the bulk resistance of the ionic pathway of the system, R1 and the first constant phase element (CPE1) represent faster charge transfer and double-layer capacitance occurring at the interface between the electrode and the plant tissue, and R2 and CPE2 represent slower ion transport and dielectric polarization within the plant tissue (Figs. S5–S8)48,49,50. For the adhesive bioelectrode there was a minimal change in the bulk resistance (R0: 26 kΩ to 23 kΩ), the electrode-plant interface (R1: 1.4 MΩ to 0.37 MΩ; CPE1: Q = 1.4 × 10−9, α = 0.98 to Q = 9.8 × 10−10, α = 0.70), and resistive and capacitive behavior within the plant (R2: 1.5 MΩ to 2.1 MΩ; CPE2: Q = 1.1 × 10−8, α = 0.51 to Q = 1.4 × 10−9, α = 0.83) over the course of the experiment (NRMSEt = 0 hrs  =  2.5% and NRMSEt = 120 hrs  =  2.8%). The minor changes in the capacitive behavior of the bioelectrode can be attributed to changes in electrode hydration; however, the stability of the bulk resistance fit parameter indicates that the adhesive gel electrodes maintained a consistent connection with healthy plant tissue. In contrast, however, the EIS data from the rigid Ag/AgCl electrode showed a close fit to the equivalent circuit (NRMSEt = 0 hrs  =  2.0%) but had an increased NRMSE (18%) by the end of the experiment, although the fitted equivalent circuit model remained in close qualitative agreement with the data, due to somewhat noisy data from a poor connection between the electrode and the plant tissue. These large changes can also be seen in the bulk resistance (R0: 36 kΩ to 0 kΩ), the electrode-plant interface (R1: 0.46 MΩ to 39 MΩ; CPE1: Q = 4.5 × 10−10, α = 0.77 to Q = 8.5 × 10−10, α = 0.67), and resistive and capacitive behavior within the plant (R2: 0.69 MΩ to 74 MΩ; CPE2: Q = 2.0 × 10−9, α = 0.74 to Q = 8.1 × 10−12, α = 0.98). The increases in R1 and R2 show that there were significant impacts on charge transfer between the electrode and the plant tissue over the course of the experiment. This increase in resistance for the Ag/AgCl electrodes at 120 hrs (5 days) is also consistent with the observed statistically significant development of necrotic tissue from both a visual inspection and the PAM fluorescence measurements, which would prevent ionic conduction between the plant tissue and the electrodes.

In addition to fitting the EIS data, the impedance modulus at 1 kHz was also compared to the environmental data (plotted with respect to time in S9-S15) by calculating the Pearson correlation coefficients (r, Table 1), using Equation (1).

$$r\,=\,\frac{\sum \left(x-\bar{x}\right)\left(y-\bar{y}\right)}{\sqrt{\sum {\left(x-\bar{x}\right)}^{2}}\sqrt{\sum {\left(y-\bar{y}\right)}^{2}}}$$
(1)
Table 1 Pearson correlation coefficients between the impedance modulus at 1 kHz of adhesive gel electrodes/commercial electrodes and environmental variables
Full size table

Although both electrodes demonstrated susceptibility to drying (r = −0.80 and r = −0.88 for the adhesive and commercial electrodes, with respect to dew point), impedance correlations revealed that the adhesive electrode demonstrated enhanced stability against drying under environmental conditions with a moderate negative correlation to relative humidity (RH) (r = −0.43) and negligible correlation to vapor pressure deficit (VPD) (r = −0.12). The commercial Ag/AgCl electrode, however, was sensitive to drying and had a strong negative correlation to RH (r = −0.79) and a moderate positive correlation to VPD (r = 0.43), leading to large increases in impedance over time. Additionally, the sensitivity coefficients \(\left(\Delta M\Omega /\Delta {{{\rm{Environmental\; Variable}}}}\right)\) were calculated for the change in electrode impedance as a function of RH, VPD, temperature, and light intensity. The adhesive gel electrode demonstrated significantly lower sensitivity coefficients than the conventional Ag/AgCl electrode for RH (0.014 MΩ/% vs. 10.8 MΩ/%), temperature (0.073 MΩ/°C vs. 48.2 MΩ/°C), and VPD (0.615 MΩ/kPa vs. 394 MΩ/kPa). The lower sensitivity coefficients for the adhesive electrode with respect to RH, temperature, and VPD demonstrates that the gel electrodes are less susceptible to fluctuations in impedance due to environmental conditions, showing that they are more suitable for long-term electrophysiological monitoring. Additionally, the gel electrode showed a lower change in impedance as a function of the ambient light intensity (lux) compared to the Ag/AgCl electrode (−0.013 MΩ/lux vs. −8.89 MΩ/lux). The Pearson correlation coefficient between the light intensity and the impedance at 1 kHz also showed a moderate negative correlation (r = −0.47), demonstrating a capacity to track diurnal cycles, thereby indicating a stable connection to the plant tissue. The diurnal cycling can be seen in the impedance fluctuations during the day/night cycles in Fig. 2e as the plant undergoes changes in photosynthetic activity51,52,53. The Ag/AgCl electrode was unable to track the plant’s diurnal cycling and had a negligible correlation to light intensity (r = −0.03).

Action potential recording and electrical stimulation

To evaluate the ability of the adhesive gel electrodes to record APs in VFTs reliably, we used the electrodes to record APs generated from manual stimulation of the trigger hairs using a wire (Fig. 3a, 3b) and performed an analysis of the resulting signals. The recorded APs were processed as follows: first, the potentials were resampled to ensure a constant time interval of 1 ms, then, a Savitzky-Golay filter with a window length of 65 fitted by a first-order polynomial was applied. The adhesive gel electrodes’ performance was compared to the commercial Ag/AgCl electrodes, showing that the bioelectrodes had a similar signal-to-noise ratio (SNR) as the commercial electrodes, which was determined from the data before applying the Savitzky-Golay filter. A Oneway Welch ANOVA testing for equal means with potentially unequal variances was performed between the SNR for the bioelectrode (n = 12) and the Ag/AgCl electrode (n = 13), resulting in a p-value of 0.0032, suggesting a trend toward the bioelectrodes having a higher SNR than the commercial electrodes (Fig. 3c). The Welch’s Test also demonstrated that the bioelectrodes recorded significantly higher average amplitudes of the APs compared to the commercial Ag/AgCl wire electrodes (64.3 mV  ± S.E. 2.7 mV vs. 37.5 mV  ± S.E. 2.6 mV, p  < 0.0001, Fig. 3e). The difference indicates that the adhesive electrodes have enhanced sensitivity and improved performance when recording APs. The analysis also showed that APs recorded with the adhesive electrodes have a significantly shorter average full-width at half maximum (FWHM) compared to the commercial Ag/AgCl wire electrodes (891 ms  ± S.E. 63 ms vs. 620 ms  ± S.E. 61 ms, Welch’s Test, p = 0.0050, Fig. 3g). The decreased recorded amplitude and increased FWHM while using the commercial electrode are likely due to signal distortion caused by poor connection to the plant tissue, a limitation effectively mitigated with the bioelectrode due to its ability to adhere effectively to the lobe40,42.

Fig. 3: Recording and stimulating Venus Flytrap action potentials using bioelectrodes.
Fig. 3: Recording and stimulating Venus Flytrap action potentials using bioelectrodes.The alternative text for this image may have been generated using AI.
Full size image

a Photograph of VFT with adhesive gel electrode adhered to the inside of a lobe with a wire in position to deliver mechanical stimulation to generate an AP with all trigger hairs identified (scale bars are 5 mm). b The average and standard deviation of the action potentials for the bioelectrode and the commercial electrode used in the analysis. c The signal-to-noise ratio of the AP signals for the bioelectrode and the commercial electrode. d The signal-to-noise ratio for an adhesive gel electrode that was removed and attached to a VFT over 14 days. e The amplitude of APs recorded using the bioelectrode and the commercial electrode. f Plot of applied electrical stimulation and recorded response from gel electrodes adhered to the inside of a VFT lobe and grounded through a needle in the sphagnum moss growing media, and images of the plant receiving electrical stimulation and closing (scale bars are 5 mm). g The FWHM of APs recorded using the bioelectrode and the commercial electrode.

The APs were recorded at 60 Hz which provides a Nyquist frequency of 30 Hz which is sufficient for capturing the ~2 s APs54. However, to ensure that faster transients (e.g., pre-depolarization spikes) were not missed, APs were also recorded at ~740 Hz for comparison. A fast Fourier transform (FFT) and the power spectral density (PSD) analysis were conducted on the high-frequency APs which showed that VFT APs fall primarily below 20––25 Hz (Fig. S16), confirming that 60 Hz sampling is sufficient to capture the AP. Additionally, the PSD analysis of the 740 Hz data showed that >99% of AP power resides below 200 Hz and >87% below 100 Hz (n = 15), with only noise at higher frequencies (n = 15 for each control). When down sampled to 60 Hz with a 25 Hz low-pass, >99.9% of AP power was retained which shows that the 60 Hz sampling rate is sufficient to capture the AP. Additionally, the higher frequency AP data was processed identically to the data recorded at 60 Hz with a Savitzky-Golay filter for comparison. A Oneway Welch ANOVA and post-hoc t-tests showed that signals recorded at 60 Hz had higher amplitudes (64.3 mV  ± S.E. 0.9 mV vs. 34.7 mV  ± S.E. 0.9 mV, p  < 0.0001) and SNR (32.1 ± S.E. 1.5 vs. 17.8  ± S.E. 1.4, p < 0.0001), while the 740 Hz sampling rate produced broader AP waveforms (1148 ms ± S.E. 61 ms vs. 891 ms ± S.E. 68 ms, p = 0.0095) (Fig. S17). There were, however, no signals that were unique to the APs recorded at 740 Hz (Fig. S18) compared to those recorded at 60 Hz. For both sampling rates, the gel electrodes were able to reliably record the AP events.

The robustness of the adhesive gel electrodes was also demonstrated through an experiment where an electrode was attached for recordings and then removed each day over 14 days. Over the course of the long-term experiment where the electrodes were removed and reattached, the recorded APs had an average amplitude of 48.0 mV  ± S.E. 1.5 mV, an average FWHM of 851 ms  ± S.E. 26 ms, and an average signal-to-noise ratio of 24.2 dB  ± S.E. 0.8 dB (Fig. 3d) with a statistically significant difference between signal and noise levels (Wilcoxon Signed Rank test, p < 0.0001). The averaged amplitude and FWHM collected during the long-term experiment align with the baseline values collected while comparing the bioelectrodes to the commercial Ag/AgCl electrodes. The stable SNR and minimal variability in the amplitude and FWHM of the APs recorded during the 14-day experiment support the bioelectrodes’ mechanical robustness and ability to adhere to the plant tissue. An additional benefit of the ability to remove and reattach the bioelectrodes is the potential to replace electrodes that may be exhibiting drift or other signal degradation.

A mechanically stimulated AP was extracted from a VFT and amplified to have a voltage of -20 V applied via a Keithley 2600 source-measure unit (SMU) to electrically stimulate VFTs to induce them to close. The electrical stimulation was delivered via an adhesive electrode attached to the inside of a lobe near a trigger hair, and a needle was placed in the hydrated sphagnum moss growing media. Lower stimulation voltages (≤15 V) failed to induce lobe closure reliably, likely due to the impedance of the bioelectrode-plant tissue interface. Impedance matching of the electrodes or the incorporation of amplification circuits could also aid in reducing the magnitude of the applied potential. However, a stimulation voltage of ±20 V was sufficient to cause the VFT lobes to close reliably for this study. The applied signal was also monitored using another adhesive electrode that was adhered to the inside of the VFT lobe. APs delivered with a 15 s gap between them were sufficient to cause the lobes of the VFT to close. As shown in Fig. 4, the VFT remained open after the first applied AP (t2) but began to close after the second applied AP (t3), which caused it to close fully (t4) and is consistent with the expected physiological response (Fig. 3f and Supplementary Video 2). The current from the applied AP was recorded between the stimulating electrode and the needle in the hydrated sphagnum moss which was then integrated to find the total charge magnitude. The average absolute charge transferred per pulse was 1.1  ×  10−10 C  ±  1.8  ×  10−11 C (n = 33) which is below previously reported charge densities for VFT electrostimulation54.

Fig. 4: Wireless plant-to-plant communication using printed adhesive gel bioelectrodes and low-cost electronics.
Fig. 4: Wireless plant-to-plant communication using printed adhesive gel bioelectrodes and low-cost electronics.The alternative text for this image may have been generated using AI.
Full size image

a The labeled components of the transmitter and receiver electronics (scale bar 5 mm). b Illustration of plant-to-plant communication at a distance of 15 m with pictures of the electronics for the transmitter and receiver [AP y scale bar is 10 mV, AP x scale bar is 0.25 s; transmitter and receiver scale bars are 10 mm. Created in BioRender. Crichton, C. (2026) https://BioRender.com/3j1fyfy]. c The transmitter microcontroller is attached to a printed adhesive gel electrode grounded through a needle in the sphagnum moss growing media and transmits a packet when an event is detected from the transmitter plant, stimulated by hand (scale bars are 10 mm). d An identical microcontroller, the receiver microcontroller, is connected to a printed electrode attached to the plant and grounded through a needle in the sphagnum moss growing media (scale bars are 10 mm). Upon receiving a packet, a 20 V voltage regulator is enabled for a pulse length determined by the packet content, simulating an AP for the receiver plant.

Plant-to-plant communication

Building on the electrodes’ ability to record and electrically stimulate APs reliably, we integrated the adhesive gel electrodes with low-cost wireless electronics to enable plant-to-plant communication. This system reliably records APs and electrically stimulates VFTs to induce trap closure at a distance tested up to 45 m (Fig. S19). While this has previously been achieved in controlled settings using laboratory-grade electronics, we implemented a low-cost, printed electronics-based approach to monitor and stimulate plant electrical signals. Two additional contributions are described here: the use of low-cost microcontrollers to detect unamplified AP signals and deliver electrical stimulation, and a demonstration of mirrored plant activation at a distance. In prior literature, instruments have been used to detect and apply APs to individual plants54,55,56. Here, two identical ESP32 microcontrollers (Fig. 4a), which are low-cost and ubiquitous, were used to detect, transmit, and apply APs from a transmitter plant to a receiver plant located 15 m apart (Fig. 4b); other tests were also conducted with a maximum of 45 m separation (Fig. S19). The microcontrollers communicated over the ESP-NOW protocol (Espressif Systems) using on-board PCB antennas with firmware that was written using the Arduino IDE. This protocol supports up to 10 encrypted peers, packet sizes up to 250 bytes which are sent rapidly reducing potential interference from denser sensor networks, a range of over 200 m, and low transmission latency. In this application, the small packet size is appropriate due to only sending an event-detected signal to the receiver. A potential issue with the telemetry system, however, is that it has a relatively low range (220 m) without the inclusion of an external antenna. A more mature system might combine ESP-NOW with a longer-distance protocol such as LoRa, or utilize a satellite or GSM-connected gateway with an on-board ESP32 radio. Both boards were powered with independent 3.7 V Lithium-polymer batteries, though the ESP32 platform described here includes provisions for external power (visible in the lower right-hand corner of the PCB, in Fig. 4b). Additionally, a small solar panel could be integrated to fully charge the on-board 2500 mAh battery daily, providing long-term autonomous power. In environments where a solar panel is not available, a standard 5 V wall charger can be used to keep the battery charged. To ensure long-term robustness in outdoor environments, a suitable housing to protect the electronics system could be used.

For the transmitter microcontroller, an analog input GPIO pin was connected to the plant’s printed electrode, a GPIO ground pin was connected to a pin grounded in the sphagnum moss growing media, and analog voltage readings were taken continuously at a frequency of 48 Hz. A FFT and PSD analysis on AP data recorded at 740 Hz showed that VFT APs fall primarily below 20–25 Hz (Fig. S16), confirming that 48 Hz sampling is sufficient for AP event detection. Two data signal processing algorithms were used in the transmitter firmware to allow for sensitive event detection with low-cost electronics: first, noise was reduced from analog readings by rapidly taking 10 analog input readings, calculating the average deviation in the signal, and filtering out any readings that fell outside of that deviation. This filtered out noise associated with the ADC. Second, new readings were written to adjustable arrays sized to detect sustained changes in voltage below an adjustable threshold, compared to a baseline array, which is updated in real time. This approach prevented single or sporadic readings from triggering an event, requiring several sequential readings below the threshold to determine that an event had occurred. Observed APs were used to determine the magnitude and duration of signal change associated with events; in the firmware used, the voltage threshold was set to 20 mV below baseline, and to be considered an actual event, four sequential readings had to be detected below that threshold. The receiver was similarly connected to a plant with a voltage regulator output pin connected to the electrode, and the GPIO ground connected to a needle in the sphagnum moss growing media. On detection of an event, the transmitter sends a packet containing a pulse duration command to the receiver. Based on observations of logged APs, a pulse length of 3 s was selected for this experiment. On receipt of a packet, the receiver enabled a 20 V voltage regulator (Pololu Robotics) for the duration sent in the transmitted packet. The regulator applied a square-wave pulse to the electrode on the receiver plant, simulating an AP. For ease of use, the receiver and transmitter provided visual and audio feedback via an onboard LED and a piezo buzzer when events were detected on the transmitter side or received on the receiver side. The ability of this low-cost system to wirelessly mirror APs between two instrumented plants placed 15 m apart and independently powered was demonstrated. Manual stimulation of the sensory hairs on the transmitting plant were detected as events by the attached microcontroller (Fig. 4c), which transmitted a command to the receiving microcontroller, applying a 3-second, 20 V pulse to the receiving plant (Fig. 4d). Two events were recorded and transmitted, resulting in both plants closing (Supplementary Video 3).

Environmental monitoring

The ability of the VFT monitored with adhesive gel electrodes to function as an environmental sensor was tested using two primary stimuli: mechanical stimulation via insect interactions, and APs induced due to temperature changes. VFTs with gel electrodes adhered to the inside of their lobes were placed inside an enclosure where large crickets (Acheta domesticus) were introduced and allowed to move freely (Fig. 5a). Occasionally, the crickets stimulated the VFT’s sensory trigger hairs, generating APs. These APs were recorded in real-time using an electrode that was adhered to the inside of the VFT’s lobe and a needle located in the hydrated sphagnum moss growing media and connected to a Keithley DMM6500 multimeter. The crickets were able to mechanically stimulate the trigger hairs to generate APs without closing the VFT, APs in close enough proximity to close the VFT but without capturing the cricket, and APs that would close the VFT and capture the cricket. If a cricket was captured, it would generate a substantial number of APs from inside the lobes after the plant had closed (Fig. 5b and Supplementary Video 4). In the VFT experiments involving crickets, the recorded APs were preprocessed similarly as previously described. The preprocessed APs were binned into different events of interest (Baseline, Closed, Closed but No Capture, Post Close, and Random) and analyzed using a one-way analysis of variance (ANOVA) for each response variable: amplitude and FWHM. The ANOVA results (Supplementary Fig. 10) indicate that the events have a significant effect on the two response variables (amplitude p < 0.0001 and FWHM p < 0.0001). Further, a post-hoc non-parametric all-pair analysis was performed using the Steel-Dwass method, which effectively accounts for uneven samples per event. The post-hoc analysis results show that the amplitude for the post close events (46.0 mV ± S.E. 1.2 mV, n = 46) was significantly different than the baseline events (35.4 mV  ± S.E. 2.3 mV, n = 12, p = 0.017), while the remaining comparisons were insignificant. The FWHM for the post close events (1312 ms ± S.E. 39 ms, n = 46), were significantly different from the baseline events (678 ms  ± S.E. 76 ms, n = 12, p < 0.001), closed events (910 ms ± S.E. 96 ms, n = 11, p = 0.0002), and random events (641 ms  ± S.E. 86 ms, n = 10, p < 0.0001). Overall, the APs for the post close events had higher amplitudes and FWHM than the baseline, manually triggered, events.

Fig. 5: Recording of Venus Flytrap action potential generation due to environmental stressors using printed adhesive gel bioelectrodes.
Fig. 5: Recording of Venus Flytrap action potential generation due to environmental stressors using printed adhesive gel bioelectrodes.The alternative text for this image may have been generated using AI.
Full size image

a Illustration of VFT being introduced to an enclosure of crickets and (b) the recorded potential of a VFT capturing a cricket with additional action potentials generated from the trapped cricket (scale bars are 5 mm; insert is an expanded view of the same data). c Illustration of VFT being put inside a freezer and (d) the recorded potential of a VFT and temperature in a cooling freezer (scale bars are 5 mm). Figures (a) and (c) were created in BioRender. Crichton, C. (2026) https://BioRender.com/g94w927.

Another environmental factor that causes AP generation is cold shock57. Cold shocking depolarizes membrane potentials similar to the response observed during mechanical stimulation of the trigger hairs57,58. To assess the adhesive gel electrode’s ability to monitor the VFT’s response to temperature, plants were placed in a freezer initially at room temperature. The freezer was then cooled gradually (Fig. 5c). Once the temperature reached 9.0 ± 0.7 °C (n = 3), spontaneous APs occurred within the VFTs, causing the lobes to close (Fig. 5d and Supplementary Video 5). The APs were preprocessed similarly as previously described. However, apart from binning the different events of interest (baseline AP, freezer AP, freezer AP - closed) as done for the cricket experiment, we also include the temperature data for different APs. We chose a least squares fit regression analysis (Figs. S20–S26) to study the effect of the different events (categorical variable) and temperature (continuous variable). The results reveal that the effects of temperature and events on FWHM are significant (p < 0.0001 and p = 0.0095) while insignificant for amplitude (p = 0.28 and p = 0.48). Additionally, an ad-hoc Tukey-Kramer Honestly Significant Difference (HSD) multiple comparison test was performed for the different events. The results indicate that FWHM for baseline significantly differs from freezer AP - closed (p = 0.016), while other comparisons were insignificant (p  > 0.1). Also, the all-pairs comparison of events on the amplitude measure was insignificant for all combinations (p > 0.47). The results from both the insect interaction and temperature sensitivity experiments highlight the adhesive gel electrode’s ability to monitor bioelectronic signals of the VFT’s response to environmental changes.

Conclusion

In this work, we demonstrated the ability of an electrode made from inkjet-printed PEDOT:PSS on a soft PVA-gel substrate to adhere to the inside of VFT lobes using a methyl cellulose adhesive layer. These electrodes exhibit mechanical robustness, a stable impedance over extended periods, and an ability to adhere to the plant tissue without cytotoxic effects which enables longer-term electrophysiological monitoring and stimulation in VFTs when compared with rigid Ag/AgCl electrodes. The adhesive gel electrodes also demonstrated a comparable SNR and higher average amplitude and FWHM when compared to the rigid commercial electrodes. The bioelectrodes could also be removed from the plant and reattached over the course of 14 days. These values, averaged over the long-term experiment (14 days), are in line with baseline values. The soft adhesive electrodes can be used to repeatedly deliver electrical stimulation on the order of  ± 20 V to close the lobes of the VFT, and can be paired with low-cost electronics to enable wireless plant-to-plant communication. These devices provided real-time feedback on the VFT’s responses to environmental stimuli such as insect activity and temperature. While the VFT plants provide a reproducible, well-studied platform for developing biohybrid systems, translating this system to other plant species may require adapting the gel electrodes for the appropriate geometry and mechanical properties. This work demonstrates the potential of these biocompatible printed adhesive gel electrodes to be used for long-term interfacing with plant tissues. Opening up the possibility of their use to enable instrumented plants to act as living, biohybrid environmental sensors that take advantage of electrophysiological responses for real-time monitoring for natural systems.

Methods

Materials and ink formulation

PEDOT:PSS Clevios™ FHC. ink was purchased from Heraeus. Following previous reports, PEDOT FHC was mixed with 1 wt% of glycidoxypropyltrimethoxysilane (GOPS, Sigma Aldrich), which is a concentration that produces a film with high electrical conductivity while maintaining mechanical stability and softness, to crosslink the PEDOT:PSS conductive trace59,60,61. The PEDOT:PSS ink concentration was 1–1.5%, and its viscosity was 8–70 mPa s before the GOPS was added. The PEDOT:PSS inks were sonicated in an ultrasonic bath for 30 min and filtered using a 1.2 μm PTFE filter. For the substrates and the adhesive, PVA (Mw 89,000-98,000 g mol−1; Sigma Aldrich), phosphoric acid (Sigma Aldrich) for better transfer of the PEDOT:PSS from the glass slide, methyl cellulose (Sigma Aldrich), and deionized water were used. The thickness of the PEDOT:PSS traces was evaluated using a DektakXTA profilometer from Bruker.

Device fabrication and application to plant tissue

Each layer was deposited using a Dimatix DMP-2800 inkjet printer with a 10 μm drop spacing, using wet-on-wet printing on a glass substrate for a total of three layers (Fig. S27)4,62. The printer plate temperature was kept at 50 °C and the ink at 30 °C. After printing, the conducting traces were dried on a hot plate at 90 °C under ambient conditions for 10 min. PVA gels (100:10 water:PVA ratio, Mw = 89,000–98,000 g mol−1) were prepared by dissolving PVA in deionized water at 90 °C under magnetic stirring. Methyl cellulose was prepared by dissolving methyl cellulose (deionized water : methyl cellulose weight ratio 30:1) in deionized water on a hot plate at 90 °C using magnetic stirring. For the electronic adhesive gels, after curing the conducting electrode, a layer of PVA gel was deposited and dried for 24 h at room temperature. The device was then peeled, thin wires were attached using silver epoxy (MG Chemicals 8331D), and left to dry for 12 hrs at room temperature. Once dried, T-Rex tape (purchased from Home Depot) was applied over the silver epoxy/PVA junction. When attaching the electrodes to the VFTs, a thin layer of methyl cellulose was added to the PVA to act as an adhesive to attach the electrode pad (1 × 1.5 mm) to the plant (Fig. S28). When attaching the electrodes to the plant tissue, the methyl cellulose/PVA were attached directly to the plant tissue.

A similar method was employed for the 14-day removal and reattachment experiment for the gel bioelectrodes. After collecting APs, tweezers were used to remove the gel electrode from the plant tissue. To reattach the same electrode the subsequent day, a thin layer of methyl cellulose was reapplied to the PVA substrate, then the electrode was attached to the same location of the inside of the VFT lobe using tweezers. All APs were manually generated by touching a trigger hair and were monitored using a Keithley DMM6500, which was connected to the bioelectrode and a needle in the hydrated sphagnum moss growing media.

Commercial electrodes

The commercial electrodes were purchased from Backyard Brains as part of their Plant SpikerBox and consisted of an Ag/AgCl wire wrapped around a plastic rod. SignaGel electrode ionic gel (purchased from Backyard Brains) was used as a conductive electrolyte between the commercial electrode and the plant tissues.

Characterization of adhesive gel electrodes

Electrical characterization

Electrochemical impedance spectroscopy (EIS) was performed using an EmStat Pico (Palmsens) impedance analyzer over a frequency range of 22 Hz to 200 kHz with a 100 mV applied potential. Action potentials were monitored using a Keithley DMM6500 multimeter.

Mechanical and adhesion testing

A Test Resources Dynamic and Fatigue Test Machine was used to perform the mechanical characterization of the electronic gel-based devices and their adhesion to glass slides under uniaxial tension at a rate of 30 mm min−1. A 445 N load cell was used for the experiments. For the experiments analyzing the shear stress as a function of strain for the gel electrodes adhered to plant tissue, the electrodes were adhered using methyl cellulose then stretched at a rate of 30 mm min-1 until failure. The electrodes robustness to mechanical stresses over time were analyzed by adhering an electrode to a glass slide then stretching it to 10% strain then relaxing it both at a rate of 30 mm min−1 for 1000 cycles.

Chlorophyll fluorescence measurements

Chlorophyll fluorescence was measured with a PAM-101 chlorophyll fluorometer (Walz, Effeltrich, Germany) to quantify maximal photochemical efficiency of photosystem II (PSII) as a measure of photosynthetic competence of Dionaea muscipula leaf tissue in contact with the electrodes. Plants of D. muscipula were kept in darkness for 30 min prior to fluorescence measurements. After 30 min, Fo (with all PSII centers in the open state ready to perform photochemistry) was determined, followed by Fm (with all PSII centers reduced and unavailable to perform photochemistry). Variable fluorescence Fv was calculated as Fm-Fo. For both treatments (with the different electrode types), control values of Fv/Fm were determined before electrodes were placed into the respective leaves. Once electrodes were in contact with the leaves, measurements were conducted in triplicate on days 0, 3, 5, and 9.

Electrophysiological experiments

Action potential monitoring

All action potentials (APs) were monitored using a Keithley DMM6500 multimeter, with voltage recorded between a bioelectrode adhered to the inside of the VFT lobe and a needle that was inserted into the hydrated sphagnum moss growing media. The Keithley DMM6500 collected data at a sampling rate of 60 Hz. No additional signal processing, filtering, or noise rejection methods were applied. A wire was used to manually touch a sensory trigger hair to stimulate an AP. To prevent premature closing of the lobes while collecting APs, a 2 min waiting period between mechanical stimulation attempts was implemented.

Electrical stimulation and plant-to-plant communication

Electrical stimulation was delivered through the gel electrodes using a Keithley 2600 source-measure unit (SMU) and a 20 V voltage regulator (Pololu Robotics). For experiments conducted using the Keithley SMU, a recorded AP was amplified to have a minimum of -20 V, which was then applied to the VFT twice with a 15 s interval. Electrical stimulation using the 20 V voltage regulator was initiated using a set of two ESP32 microcontrollers (Adafruit, Inc.), which were used to detect, transmit, and apply APs between a transmitter and receiver plant, tested at a maximum of 45 m, apart using the ESP-NOW protocol to enable plant-to-plant communication. The transmitter microcontroller, connected to a plant electrode, sampled voltage at 48 Hz and employed a two-step signal processing algorithm to filter noise and detect sustained voltage drops of at least 20 mV below baseline. When an event was detected, the transmitter microcontroller would trigger the receiver microcontroller to activate a 20 V voltage regulator (Pololu Robotics) to apply a 3-second square-wave pulse.

Plant growth conditions

Venus flytrap plants (Dionaea muscipula) were grown in a semi-hydroponic system using sphagnum moss and reverse osmosis water in a temperature-controlled greenhouse environment. The experiments were conducted in a controlled environment where the temperature and humidity were kept constant and monitored every minute using a PulsePro environmental monitor. An 86-watt LED grow light maintained a 12 hr:12 hr light-dark cycle.

Environmental monitoring experiments

Insect experiments

Healthy large crickets (Acheta domesticus) were purchased from Petco in batches of 15 crickets and kept in an enclosure with damp cotton balls for water and Fluker’s Orange Cube Cricket Diet for food. The potential of the VFT during the experiments involving crickets was monitored using a Keithley DMM6500 multimeter, measuring the voltage between a bioelectrode and a needle in the hydrated sphagnum moss growing media.

Temperature variation experiments

APs were recorded using a Keithley DMM6500 multimeter to monitor the voltage of a bioelectrode on the inside of a VFT lobe and a needle in the hydrated sphagnum moss growing media. Environmental conditions, including temperature, humidity, and light intensity, were monitored using a PulsePro environmental logger, with data collected every minute.

Imaging

Optical photographs were taken with an iPhone 12 Pro, a Canon EOS Rebel T8i, and a Nikon D7000. Video monitoring of the VFTs during long-term experiments was conducted using a WYZE Cam v3. Images pulled from videos of experiments in Fig. 4 were upscaled using Gigapixel AI, and the images from videos of experiments in Fig. 5 were upscaled using Upscayl AI Image Upscaler.