Multifunctional Nanomesh Enables Cellular-Resolution, Elastic Neuroelectronics

Abstract

Silicone-based devices have the potential to achieve an ideal interface with nervous tissue but suffer from scalability, primarily due to the mechanical mismatch between established electronic materials and soft elastomer substrates. This study presents a novel approach using conventional electrode materials through multifunctional nanomesh to achieve reliable elastic microelectrodes directly on polydimethylsiloxane (PDMS) silicone with an unprecedented cellular resolution. This engineered nanomesh features an in-plane nanoscale mesh pattern, physically embodied by a stack of three thin-film materials by design, namely Parylene-C for mechanical buffering, gold (Au) for electrical conduction, and Poly(3,4-ethylenedioxythiophene)-poly(styrenesulfonate) (PEDOT:PSS) for improved electrochemical interfacing. We have validated nanomesh elastic neuroelectronics using single-unit recording from the small and curvilinear epidural surface of mouse dorsal root ganglia (DRG) with device self-conformed and demonstrated superior recording quality compared to plastic control devices requiring manual pressing. Electrode scaling studies from in vivo epidural recording further revealed the need for cellular resolution for high-fidelity recording of single-unit activities and compound action potentials. In addition to creating a minimally invasive device to effectively interface with DRG sensory afferents at a single-cell resolution, our study establishes nanomeshing as a practical pathway to leverage traditional electrode materials for a new class of elastic neuroelectronics.

Graphical Abstract

The results presented here provide multifunctional materials, fabrication, and microsystem strategies that establish the foundations for elastic neuroelectronics, where conventional thin-film-based electrode materials make stretchable microelectrodes through in-plane nanomesh structuring, achieving unparalleled scalability down to cellular resolution and reliable elasticity. Detailed in vivo studies suggest the ability of these devices to provide tissue-conforming and high-quality interfaces with individual neurons in intact nervous tissues.

1. Introduction

Soft, elastic electronic devices are ideal platforms for interfacing with biological tissue due to their ability to conform to the irregular morphology of organs.^[1] Compelling recent examples include hydrogel-based electrodes for localized low-voltage neuromodulation and nanofiber-scaffold-based macroporous electrodes for dynamically monitoring pulsing cardiomyocytes.^[2–4] Among various elastic polymers, silicone stands out due to its unique combination of biocompatibility, great dielectric performance, high water resistivity, and easy processability and has seen widespread uses in breast and facial implants and bioelectronic devices such as cochlear implants and pacemaker leads. However, current silicone-based leads are not truly elastic nor scalable, due to the rigidity of the metal wires/foils being used and the largely manual manufacturing processes.^[5] On the other hand, thin-film based micro- and nano-fabrication has enabled remarkable electronic scaling in the past five decades, but is not readily compatible with silicone substrates. Indeed, recent bioelectronic demonstrations on silicones have to leverage microcracks, which has established a promising approach towards addressing this challenge and demonstrated efficient nerve recording and stimulation, but only with a feature resolution generally over 200 μm, in part due to the unpredictable nature of the micro-scale cracks.^[6–7] Previous approaches have also leveraged microscale structuring or serpentine designs to achieve elastic devices, but also at limited electrode size or resolution.^[8–9] This scalability challenge is especially prominent for cellular activity mapping in neuroelectronics, where electrodes/sensors require a sub-20-μm resolution to achieve neuronal-level specificity and high-density interfacing.^[10–11]

Here, we demonstrate a multifunctional nanomesh approach to fabricate highly scalable elastic neuroelectronics featuring crack-free elasticity directly from conventional thin film materials. The nanomesh here features a lithographically defined nanoscale mesh pattern in plane, and consists of a vertical stack of three different thin-film materials purposefully by design, namely Parylene-C for mechanical buffering on PDMS, Au for electrical conduction, and PEDOT:PSS for improved electrochemical interfacing.^[12–16] As such, the resulting elastic electrodes are scalable down to less than 20 μm with electrode performance comparable to conventional rigid ones. Due to its spring-like curvilinear pattern in-plane in the mesh structure, the multifunctional nanomesh has a modulus comparable to PDMS. It is deformable and free of cracks by up to ~ 50%, further improvable from in-plane structure optimization. This elasticity is also highly reliable, as revealed by rigorous stretching and soak tests. We validated our nanomesh elastic bioelectronics through extracellular recordings from the mouse DRG epidural surface. Specifically, in vivo electrode scaling studies vindicate the need for cellular resolution, and high-density nanomesh elastic electrode arrays demonstrate single-unit recording from the DRG surface through self-conformed with unprecedented quality and spatiotemporal dynamics. Together, our study not only establishes nanomesh elastic neuroelectronics as a minimally-invasive device to effectively interface with sensory afferents at the DRG, it also creates a new material platform, multifunctional nanomesh, for elastic neuroelectronics with unparalleled scalability.

2. Results and Discussion

2.1. Thin-film Fabrication of Nanomesh Elastic Microelectrodes

The fabrication here starts with a thin substrate of PDMS (thickness, 15 μm) and forming a stack of three different thin films with the same deformable nanomesh pattern in-plane (Figure 1a,b and Figure S1). These three thin films include, first, Parylene-C (thickness, 100–500 nm), establishing a mechanical buffer layer on PDMS to prevent cracks of thin-film electronic materials during subsequent fabrication processes, enabling a high-yield fabrication.^[17] Due to its production from chemical vapor deposition, Parylene-C is also known to form partial caulking into PDMS and achieve reliable interface.^[18–19] Second, a Au layer (thickness, 20–100 nm) with Titanium (Ti) adhesion achieves electrical conduction on Parylene-C. We utilized Indium (In) grain boundary lithography for Au nanomesh patterning.^[20] Grain boundary lithography primarily employs a bilayer lift-off metallization methodology. In this study, the bilayer comprises In grains as the mask layer and poly(methyl methacrylate) (PMMA) as the sacrificial layer for the lift-off process. The gaps between the In grains are enlarged via nitric acid (HNO₃) etching within the initially deposited In film, which is constituted of single-layered In grains. Subsequently, reactive ion etching (RIE) is conducted to etch the mask stack down to the Parylene-C. Following the Ti/Au metallization and lift-off process, the resulting nanomesh forms a structure comprising a network of fully interconnected Au wires. Fully interconnected Au nanomesh can have excellent electrical conductivity since it does not have wire-to-wire junction resistance. Third, a PEDOT:PSS (thickness, 25–107 nm) coating results in a low-impedance interface on Au. Leveraging its porous structure and electrochemically active nature, PEDOT:PSS increases the electrode’s effective surface area while facilitating efficient ion exchange at the recording site and effectively converting ionic current to electronic current.^[21] Fabrication details are elaborated in the Methods section. Consequently, we have fabricated a multi-layer nanomesh on PDMS that exhibits mechanical stiffness comparable to pure PDMS substrates of the same thickness, making it ideal for interfacing with biological tissues (Figure 1c). The PDMS with the multi-layer nanomesh appeared to be stiffer than the pristine PDMS, as a result of including Au and Parylene-C in the laminate system. It should be noted that the tensile stiffness of such laminate was a sum of each layer, as detailed in the Supplementary Text section (Table S1). Even though the thickness was dominated by PDMS, the contributions from other layers are not negligible. And a good agreement has been made between the experimental values and the theoretical calculations.

Figure 1.

Nanomesh elastic microelectrodes and arrays. a) Schematic and an SEM image of the multifunctional nanomesh, showing mechanical buffering, electrical conduction, and electrochemical interface. b) SEM image of a nanomesh elastic microelectrode with 40% strain. c) Tensile testing result of 20-μm-thick PDMS with and without Parylene-C/Au (300 nm/60 nm) nanomesh. The PDMS is composed of a 15-μm-thick 4:1 PDMS substrate and a 5-μm-thick 10:1 PDMS encapsulation. d) Impedance spectra of a representative nanomesh elastic microelectrode as a function of strain. e) FEA simulation result of Parylene-C/Au nanomeh on PDMS when 5% strain was applied. f) Impedance histogram of a representative 256-ch nanomesh elastic microelectrode array. Inset: image of a 256-ch array. g) Upper: comparison of the throughput and electrode site area in nanomesh elastic electrode arrays with other reported elastomer-based bioelectronics in literature. Lower: Comparison of the electrode impedance in nanomesh elastic electrode arrays with other reported elastomer-based bioelectronics in literature.

As a result of the multifunctional nanomesh, we successfully achieved elastic electrodes scalable down to cellular resolution. In brief, the elastic microelectrodes and interconnects were lithographically patterned as if the nanomesh were a continuous film and finally encapsulated with PDMS (thickness, 5 μm). When measured using electrochemical impedance spectroscopy, the impedance and phase spectrum exhibited negligible changes as a function of applied strain up to 50% using strain rate of 5%/s (Figure 1d) and exposure area increased consistent with the applied tensile strain (Figure S2). We also measured impedance spectra of representative nanomesh elastic microelectrodes under 0%, 50% uniaxial strain at varying strain rates ranging from 0.2%/s to 5%/s (Figure S3). Considering the strain rate used in the characterization of the latest bioelectronics devices as 10%/min or 3%/sec or 120%/s or, it is quite encouraging to observe a negligible change in impedance when 50% strain is applied at a rate of 5%/s.^{[14, 22]} This strain immunity is ideal for bio interfacing and can be illustrated by a Finite Element Analysis (FEA) model of multi film nanomesh comprising 5 μm PDMS, 300 nm Parylene-C, and 60 nm Au with a 5% strain applied to the PDMS (Figure 1e and Figure S4 and S5). The strain distribution in the top Au layer reveals that the local deformation of the nanomesh-patterned Au is significantly lower than the global deformation of the substrate. As a result, the nanomesh electrode would undergo a global elastic deformation upon applying strain, which does not significantly change its electrode area, hence the impedance. The FEA simulations for the multi-layer nanomesh stretched up to 30% are presented in Figure S6. It should be noted that predictions for large strains can be less accurate without incorporating a fracture model, which is why we demonstrated only a small applied strain. However, the effects of the random, spring-like networks remain valid under stretching. Specifically, these effects include the disparity between the globally applied strain on the soft substrate and the local strain on the mesh layer, as well as the randomized stress distribution throughout the network. The combination of these effects helps maintain the stretchability of the multi-layer nanomesh. Furthermore, on a larger scale with hundreds of traces, the breakage or local fracture of some traces will not compromise the overall integrity of the structure. These local fractures, which release strain energy, can actually protect the surrounding network, preventing catastrophic failure and thereby maintaining the conductivity of the mesh (Figure S7). We note that although limited by the random nature of the In grain boundary lithography,^[20] this stretchability is sufficient to enable excellent PDMS compatibility and a wide range of bioelectronic applications.

With the ability to achieve cellular-resolution electrodes and scaled interconnects, we created high-density arrays with up to 256 nanomesh elastic microelectrodes (Figure 1f). Previously reported microelectrodes on PDMS or hydrogel substrates are all limited by large electrode site areas and low density (Table S2). In contrast, the multifunctional nanomesh approach has enabled elastic neuroelectronics with record throughput and density, reaching levels on par with the commercial silicon-based Michigan arrays while maintaining a cellular-level electrode site area (down to 20-μm diameter, or 314 μm²) and low impedance (Figure 1g). This low impedance is also comparable with state-of-the-art neuroelectrodes.^{[6–7, 23–30]} The impedance histogram of all 256 neuroelectrodes on a nanomesh elastic array (electrode site area: 20×20 μm²) exhibited an average impedance of 200 kΩ at 1 kHz, achieving a 90.6% yield (Figure S8).

2.2. Multifunctional Nanomesh Structure-property Relationship

To study the structure-property relationship of the multifunctional nanomesh, we investigated the electrical, mechanical, and electrochemical performance of the nanomesh and its electrode as a function of each constituent film’s thickness (Figure 2a) while controlling the in-plane nanomesh pattern to be nearly identical from the same process. We first fabricated Parylene-C/Au nanomesh interconnects with various lengths and a fixed width of 20 μm to extract the sheet resistance values as a function of Au thicknesses (20–100 nm). We extracted sheet resistance of Au nanomesh and Au-film control with various thicknesses by using the Transfer Length Method (TLM) (Figure 2b), as extracted by the following equation, where R_tot is the total resistance measured, R_c is the contact resistance, R_s represents the sheet resistance of the interconnect, w is the interconnect width which we fixed as 20 μm, and d is the length of interconnect.

Figure 2.

Multifunctional nanomesh structure-property relationship. a) Exploded-view illustration of the constituent films of the multifunctional nanomesh on a PDMS substrate: Parylene-C, Au, and PEDOT:PSS. b) TLM study of Au nanomesh as a function of its thickness. c) The sheet resistance of the Au nanomesh as a function of its thickness with FEA simulation result. d) FEA simulation of the surface current density (magnitude) in Au nanomesh. e) Stretchability of Au nanomesh as a function of Parylene-C thickness. f) 1kHz-impedance of nanomesh elastic electrode as a function of PEDOT:PSS thickness. g) Impedance and phase spectra of a representative nanomesh elastic electrode with equivalent circuit modeling fitting. Inset: equivalent circuit model for the electrode-electrolyte interface of the nanomesh elastic electrode. h) Extracted parameters from equivalent circuit modeling as a function of PEDOT:PSS thickness. Here, C_dl is double-layer capacitance, C_d is diffusion capacitance, R_d is diffusion resistance, and R_ct is charge-transfer resistance.

$$R_{tot}=\frac{R_s}{w}d+2R_c$$ (1)

The sheet resistances of the Au nanomesh ranged from 2.74 to 18.79 Ω/□ (Figure 2c and Figure S9), representing values ten times higher than those of the Au film at the same thicknesses. Nevertheless, the Au nanomesh exhibits competitive sheet resistance within stretchable bioelectronics (Table S3), with excellent scalability.^{[3, 6–7, 23, 25, 29, 31–33]} The sheet resistance and surface current density (magnitude) are predicted in FEA (Figure 2c,d), where a good agreement of the FEA and experimental results for sheet resistance as a function of Au thickness is observed. Guided by these experimental outcomes, we established the thickness of Au as 60 nm for the subsequent experiment, as this value yields satisfactory conductivity for the interconnects. For example, using 60-nm-thick Au nanomesh, the 4-cm-long interconnection intended for use in the in vivo application exhibited a resistance of approximately 10.6 kΩ, which is substantially smaller than the desired impedance level for typical neuroelectrodes.

Next, Parylene-C/Au nanomesh interconnects, 2 cm in length and 20 μm in width, were fabricated on a PDMS substrate to study its stretchability and effective modulus as a function of the Parylene-C thickness (100–500 nm). We observed higher interconnect stretchability as the Parylene-C thickness increased (Figure 2e). The resistance of the interconnects remained nearly unchanged as the applied strain before rupture (Figure S10). This phenomenon also contributed to the performance stability of the nanomesh elastic electrodes during strain. It is beneficial to use thicker Parylene-C to enhance the overall stretchability, presumably by reducing the local strain in Au. As a tradeoff, the effective modulus of the nanomesh will increase (Figure S11), and the fabrication of Parylene-C with larger thickness is also more challenging due to the undercut during the dry etching process to achieve the nanomesh pattern in Parylene-C. Guided by the experimental outcomes and tradeoffs, we established the optimal thickness of Parylene-C as 300 nm, which could lead to around 50% stretchability, significant for bioelectronics. This attribute, combined with the ultrathin nature of the entire system at a 20-μm thickness, makes it highly suitable for interfacing with biological tissues, with the ability to conform to highly irregular shapes of organs and anatomy.

For the electrochemical properties, we studied 20-μm-diameter electrodes from nanomesh comprising all three films of Parylene-C, Au, and PEDOT:PSS with varying PEDOT:PSS thicknesses (26–107 nm). The thickness of PEDOT:PSS was measured by atomic force microscopy (AFM) (Figure S12). The coating of PEDOT:PSS resulted in a significant reduction in the electrode impedance (Figure 2f), consequently lowering thermal noise and signal loss before amplification. Through AFM scanning, we identified a linear correlation between the thickness of the coated PEDOT:PSS layer and the deposition charge density^[34]. Notably, there was no significant lateral growth of PEDOT:PSS for all deposition conditions, maintaining the nanomesh pattern after coating (Figure S13). The PEDOT:PSS layer, being both thin and soft, has a negligible contribution to the mechanical properties of the multi-layer nanomesh compared to the Au and Parylene-C layers. To investigate the effect of Au thickness on stretchability, we conducted FEA simulations by varying the Au thickness from 30 nm to 150 nm in intervals of 30 nm. The results are summarized in Figure S14. Our preliminary observations indicate that there is no significant difference in the overall strain distributions among these simulations at the same global stretching. This suggests that stretchability is not highly dependent on the Au thickness, particularly at scales around 100 nm. This may be due to the fact that the tensile stiffness and deformation behavior are predominantly governed by the Parylene-C layer. Consequently, this coating approach achieves highly scalable nanomesh elastic electrodes.

To shed light on the electrochemical interface of the multifunctional nanomesh with PEDOT:PSS, the extracted equivalent circuit components from the impedance and phase spectra from all electrodes as a function of PEDOT:PSS thickness were fit into an electrode-electrolyte equivalent circuit model (Figure 2g,h, Figure S15, and Table S4). The circuit model includes a double-layer capacitance denoted as C_dl, in parallel with a faradic impedance Z_F, comprised of a charge-transfer resistance R_ct in series with the parallel conduction of a linear diffusion component involving the capacitance from electrolyte ion adsorption, C_d, and the diffusion resistance R_d. The resistance R_ct embodies electron transfer resistance from the electrolyte to the electrode during Faradic reactions. Both C_dl and C_d correspond to the non-Faradic double-layer capacitance, depicted by the equation $Z=1/{(j2\pi f)}^n·Q$, where 2πƒ represents angular frequency, and Q determines the magnitude of impedance.^[35] For both C_dl and C_d, we could set n=1, a value for ideal capacitors, while achieving good fitting with experimental measurements. A parasitic resistance R_P from solution and routing resistances further connects the above circuit, in addition to a parasitic capacitance C_p, combining stray capacitance between the electrolyte and the PDMS encapsulation above the routing trace, along with the measurement device’s parasitic capacitance. Modeled impedance magnitudes and phase values were able to fit well with measured data despite the simplicity of this model (Figure 2g). Our model also suggests that the impedance reduction from increasing PEDOT:PSS thickness is primarily driven by the increase of the interfacial capacitances C_dl and C_d and the reduction of interfacial diffusion resistance R_d, while not the charge transfer resistance R_ct (Figure 2h). It should be emphasized that the current circuit modeling has been previously implemented by Wang et al. (2021), focusing specifically on the impedance characterization and modeling of microelectrodes with different materials and sizes coated with PEDOT:PSS, including Au/PEDOT electrodes.^[35] The comparable double-layer capacitance and diffusion capacitance observed may stem from the fact that PEDOT:PSS has a significantly large electroactive surface area due to its porous structure. This porosity allows the electrolyte to penetrate and access the internal surface effectively, leading to a faster diffusion process for a thin coating.^[36] In 3D porous media, the physical time scales of double-layer formation and the diffusion process can be similar, which differs from the solid electrode/electrolyte interface.^[37] By plotting the capacitance vs. thickness in log-log scale with linear fitting, the scaling law has been revealed, namely, C_d~t^1.32(R²=0.9) and C_dl~t^1.93 (R²=0.98), which is similar to the reported scaling for different thicknesses (Figure S16).^[38] It was indicated that the double-layer capacitance, determined by the electroactive surface area, will grow faster than the diffusion capacitance, as the ratio between them C_d/C_dl~t^-0.61. This may be because of the nanomesh structure, which played a role in further enhancing the increment of electroactive surface area by introducing large-scale pores. However, it is difficult to conclude at the current stage since only limited experimental data were used for the fitting. And future investigations will be necessary to further undisclose the mechanism here.

2.3. Performance Scaling and Array Connection of Nanomesh Elastic Microelectrodes

To study the performance scaling of nanomesh elastic electrodes, we fabricated electrodes with varying diameters from nanomesh comprising 300 nm Parylene-C, 60 nm Au, and 43 nm PEDOT:PSS. As shown in Figure S17, the electrode impedance decreased as its site area increased. Eventually, with the expansion of the site area, the parasitic resistance from the interconnects and solution assumed a more substantial proportion of the overall impedance (Table S5). The PDMS compatibility and device elasticity are highly reliable, as validated by subjecting the arrays to 30% strain for up to 1,000,000 cycles (Figure 3a). This mechanical testing reveals no substantial yield or impedance changes. We further conducted accelerated soak tests of the nanomesh elastic electrode (N=5) in a 0.01M phosphate-buffered saline (PBS) solution (with a pH of 7.4) at 57 °C for up to 4 weeks, which is equivalent to 16 weeks at 37 °C, assuming an accelerated aging factor Q10 of 2 (Figure 3a). This aging factor is typically used in the ASTM F1980 standard for accelerated aging of medical devices.^[39] Throughout these tests, we monitored impedance and observed negligible changes, with no instances of film delamination detected upon microscopic inspection. We also validated the reliability of the Parylene-C-PDMS and Parylene-C-metal interfaces. We fabricated single electrodes and measured impedance while immersing the samples under 30% strain in a 57°C PBS solution over time. The impedance remained stable throughout the 7-day soaking period at 57°C, equivalent to 28 days of soaking at 37°C based on an accelerated aging factor (Q10) of 2. During these tests, we observed negligible changes in impedance, with no instances of film delamination detected upon microscopic inspection (Figure S18). These results confirm that our multifunctional nanomesh possesses reliable Parylene-C-PDMS and Parylene-C-metal interfaces. We further investigated the relationship between charge injection limit (CIL) and site area. The CIL measurement was achieved by delivering a series of 256 biphasic pulse trains, initiating a cathodic-first current pulse with 266.5 μs pulse duration. These pulses varied in current amplitude, enabling us to assess the electrode’s voltage response to the stimulation current. This evaluation was conducted during reversible charge transfer at the electrode interface, ensuring that the potential remained within the water window limits. The outcomes, as depicted in Figure 3b, reveal a linear correlation between the charge injection limit and the site area.^[40] This investigation into electrode scaling underscored the predictable performance of nanomesh elastic electrodes, enabling easy bioelectronic design for different neural interfacing applications.

Figure 3.

Performance scaling and array connection of nanomesh elastic microelectrodes. a) Upper: cyclic stretching test result of a representative 32-ch nanomesh elastic electrode array at a 30% strain. Lower: averaged and normalized impedance of representative elastic electrodes (N=5) with 314 μm² electrode site area as a function of weeks after soaking immersed in PBS (pH 7.4) under 57 ºC. b) CIL as a function of the site area of nanomesh elastic electrodes. Inset: Examples of corresponding voltage transients to injected current limits (±480 μA) for a nanomesh elastic electrode with 38,400 μm² electrode site area, limited at water electrolysis window (−0.9V to 0.6V). c) Set-up of the PDMS encapsulation scalability test. d) Leakage current as a function of PDMS encapsulation thickness. e) Illustration of the 256-ch elastic electrode array connection. f) Strain distribution on the elastic array pad area during connection.

We also conducted a scaling study on the thickness of the PDMS encapsulation, leveraging accelerated soak tests. In this study, we applied PDMS coatings of varying thicknesses (1–100 μm) onto the 20-μm-diameter electrodes (Figure 3c). Subsequently, we covered the interconnect area with a 2-mm-thick PDMS layer. This configuration was designed to eliminate the impact on the interconnect to evaluate encapsulation on a well-defined electrode area. DC leakages and impedances were measured as we emersed the samples in a 77 °C PBS solution over time (Figure 3d and Figure S19). For all conditions except the 1-μm-thick PDMS encapsulation, the encapsulation performance remained stable throughout the 8-day soaking period at 77 °C, equivalent to 128 days of soaking at 37 °C assuming the same aging factor (Q10) of 2. We hypothesize that the pore structure of the 1-μm-thick PDMS might have facilitated the infiltration of ions and water, leading to its less stable performance.^[41] Based on this observation, we determined that a 5-μm-thick PDMS encapsulation was optimal.

A significant challenge in elastic bioelectronics lies in connecting it to rigid electronics-based data acquisition systems. We demonstrated a compact matrix connector to achieve an easy, robust, and reusable connection (Figure 3e). Employing a sandwiching structure, this connector establishes physical and electrical connections by uniformly applying strain to the pad matrix array of the device and the solder ball array of the connector with the help of cushion layer between top-board and pad matrix array. The assembly procedures are outlined in detail in the Methods section. This matrix connection can be elucidated through an FEA model. The model reveals that the localized strain of the pad array in contact with the solder ball array is significantly higher than the global deformation of the pad array, enabling secured connections (Figure 3f and Figure S20). Essentially, the pad array undergoes a global deformation with small strain, which does not lead to catastrophic failure. As such, we have achieved a high-yield and low-resistance connection for nanomesh elastic bioelectronics at the 256-channel level.

2.4. High-density Single-unit Mapping from the DRG Surface

To validate the in vivo functionality of nanomesh neuroelectronics, we utilized 64 ch high-density elastic arrays at an unprecedented 25.9 electrodes/mm² and performed neural electrophysiology mapping on the L4 DRG epidural surface in mice. Our nanomesh elastic array conformed seamlessly to cover the entire L4 DRG surface (Figure 4a,b). Additionally, the transparency of nanomesh elastic bioelectronics facilitated the precise placement onto the targeted L4 DRG region (Figure 4a and Figure S21 and S22). As a control, we also implanted Kapton-based electrode arrays with an identical layout and similar thickness (approximately 20 μm) onto the same L4 DRG in each experiment. The nanomesh elastic array naturally “self-conformed” to the contoured DRG surface to establish stable and close contact, whereas an external perpendicular force was required via a micromanipulator to establish neural recordings from Kapton-based electrode arrays. In the absence of the perpendicular force, the Kapton-based electrode array tended to detach from the DRG surface, thereby hindering any neural recording (Figure 4b).

Figure 4.

In vivo validation and high-density single-unit mapping from the DRG surface. a) Illustration of in vivo set up (left). In vivo images of the same DRG after exposure (middle), and with a 64-ch nanomesh elastic electrode array (right upper) and 64-ch Kapton array of the same design (right lower). b) Illustration of a nanomesh elastic electrode array self-conformed to DRG surface with high conformality compared to a Kapton-based array attaching to DRG surface, with the necessary need of an external perpendicular force (left). Representative wideband signals recorded from the DRG surface using the nanomesh elastic electrode and Kapton electrode (right). c) Averaged single unit waveforms mapped from a DRG surface using a 64-ch nanomesh elastic electrode array, demonstrating high yield and high fidelity single-unit recording. d) Spike firing rate (upper) and LFP (lower) mapping from a 64-ch nanomesh elastic electrode array. e) Upper: PSTH analysis of the recording from a nanomesh elastic electrode, Lower: LFP band power (1–500 Hz) of the data recorded by the same electrode, showing a strong positive correlation between spike firing and LFP. f) Spike raster plot of all recorded single units from the data shown in (c).

Overall, the nanomesh elastic array captured high-fidelity neural signals across a broad area of L4 DRG surface (Figure 4b). In contrast, though we pressed the Kapton array on the DRG surface at a certain angle, the Kapton array can still experience poor local contact with the DRG surface due to its sophisticated anatomy and motion artifact relevant to respiration and involuntary limb activities. The difference in performance is likely due to the superior surface conformability of the nanomesh array, which enables them to cover the DRG surface more completely and intimately. The interfacing challenge with Kapton array also led to increased channels contaminated with motion artifacts and attenuated action potentials in Kapton array recordings, compromising the fidelity of bio-activity mapping (Figure S23).

Single units were sorted using principal component analysis (PCA)-based algorithms and curated with rigorous metrics, as detailed in the Methods section. The nanomesh elastic array was able to demonstrate mapping of single units across the entire L4 DRG surface with an extraordinary yield (Figure 4c). Taking advantage of the conformal coverage and high-density capabilities, we recorded both the neuronal firings and local field potentials (LFPs) from the epidural surface of L4 DRG evoked by mechanical stimuli delivered to the hind paw, i.e., forceps pinching. We observed spatiotemporally-resolved evoked response throughout the entire L4 DRG (Figure 4d), which was repeatable in multiple trials. LFP spectrogram revealed an evident increase in power within 1–500 Hz during the application of pinching (Figure S24). Moreover, LFP recordings from the DRG surface exhibited high spatial resolution, with clearly distinguishable signals from electrodes separated by 195 μm (Figure S25).^[38] This spatial heterogeneity is interesting compared to the homogeneity in signal distribution typically observed within similar distances in micro-electrocorticography (μECoG) recordings from mouse brain.^[42–43] This distinction, revealed by our nanomesh elastic bioelectronics, can be attributed to the predominant presence of densely packed somata in the DRG and relatively low levels of axons, in stark contrast to the abundant dendrites and axons in the brain. Importantly, the increase in LFP power correlated positively with spike firing across the recording channels, indicating that LFP data from the DRG surface are indicative of action potentials, particularly for neurons with high firing rates (Figure 4e). These observations are supported by the spatiotemporal dynamics of action potentials and LFPs provided in the Movie S1 and S2.

Previous studies have employed optical imaging techniques to record neural activities from individual afferent somata in rodent DRG surfaces as an alternative minimally invasive approach. Although recordings of Ca²⁺ transients (using GCaMP6f) have resolved individual neuron firing across a large volume of DRG tissue, experimental evidence suggests that the recorded calcium transients do not strongly correlate with the afferent spiking patterns, and this correlation shows significant cell-to-cell variations.^[44–45] In contrast, by leveraging the electrophysiological capabilities of our nanomesh elastic bioelectronics (30 ksps sampling rate), we can easily capture sub-millisecond-level dynamics of action potentials, as depicted in the spike raster data in Figure 4f. The spike firing rates observed across individual recording channels consistently demonstrated a strong positive correlation with mechanical stimuli, also displaying diverse types of temporal responses. Specifically, noxious mechanical pinching at the L4 DRG dermatome in the hind paw area reliably evoked afferent responses in a majority of somata within the L4 DRG, which presumably originated from the mechanosensitive afferents, including nociceptors that encode tissue-injurious mechanical stimuli. The intense mechanical stimulation leads to peripheral sensitization in some neurons following the hind paw pinching, as evidenced by the carryover effects of continuous spiking even after terminating the stimulation. Meanwhile, non-nociceptive afferent neurons demonstrating rapid or slow adaptation to mechanical stimuli were also recorded, exhibiting reduced firing rates following stepped mechanical stimuli (3^rd and 4^th in the raster plot). Interestingly, a small portion of afferent neurons initially showed no response or very low firing frequency (<0.5 Hz) to noxious mechanical pinching. They became sensitized after several repetitive pinchings and started to respond with robust spiking patterns (>3 Hz), which are hallmark features of nociceptor sensitization. These results provide the proof-of-concept for achieving effective recordings of afferent spikes minimally invasively from the epidural DRG surface, which can be potentially harnessed by the next-generation prostheses and closed-loop neural stimulators that incorporate robust sensory feedback for improved performance.

2.5. Electrode Scaling Study of Epidural DRG Recording

To highlight the scalability of the nanomesh elastic microelectrodes, we performed the epidural recording from the small and curvilinear surface of mouse L4 DRG. Neuroelectronics requires cellular scale electrodes to resolve single-cell activities. In the context of extracellular recording in the brain, a common assumption is that electrodes with a size below 20 μm are suitable for single-unit recordings.^[11] However, to our knowledge, no studies have investigated the effect of electrode scaling on surface recording epidurally from small and curvilinear DRG. To shed light on this scaling, we implanted a 16-channel nanomesh elastic bioelectronics with four distinct electrode sizes (20–400 μm) and recorded electrophysiology signals on the surface of mouse L4 DRG (Figure 5a). Details on the surgical exposure of the DRG and the extracellular recording set-up are elaborated in the Methods section. Electrodes from all four sizes successfully captured evoked multi-unit activities and compound action potentials (Figure 5b-d). Utilizing the sorted single units recorded from electrodes of various sizes, we generated plots depicting the spike signal-to-noise ratio (SSNR), signal amplitude, and noise amplitude as a function of the microelectrode diameter (Figure 5e-g). Based on the experimental findings, it can be deduced that sub-40 μm-diameter electrode exhibits the most favorable performance for single-unit recording, with the highest spike SNR and most robust spike sorting result (Figure S26). This observation is consistent with neurobiology since there exists a diversity of soma sizes, ranging from large (>35 μm), medium (20–35 μm), to small (15–20 μm). Electrodes possessing dimensions similar to those of DRG surface cells can ensure close proximity of the electrodes to the targeted neurons and avoid the signal-averaging effect if using larger electrodes.^[11] Since the cellular density of the mouse L4 DRG region ranges from approximately 170 to 350 cells/mm², we believe arrays with electrode densities commensurate with this value will further enhance the exploration of rich spatiotemporal dynamics of electrophysiology from the DRG surface.

Figure 5.

Electrode scaling study of epidural DRG recording. a) In vivo image (upper) showing a 16-ch nanomesh elastic electrode array implanted on the epidural surface of a mouse L4 DRG with four different electrode sizes and the layout (lower) of the 16-ch array. b) Illustration showing the importance of electrode size. Electrodes of comparable size to DRG surface cells are ideal for single-unit recording. c) Representative multi-unit band signals and sorted waveforms from each size of elastic electrodes. d) Overlays of Aβ and Aδ components of compound action potentials evoked by electrical stimulation, recorded from four different sizes of elastic electrodes. e) SNR of spikes recorded from four different sizes of elastic electrodes as a function of the electrode diameter. f) Amplitude of spikes recorded from four different sizes of elastic electrodes as a function of the electrode diameter. g) Noise amplitude in recordings from four different sizes of elastic electrodes as a function of the electrode diameter. h) Histogram showing the number of Aδ fibers distributed across their respective ranges of conduction velocities. i) Histogram showing the number of Aβ fibers distributed across their respective ranges of conduction velocities. j) Recorded Aδ fiber action potential amplitude from four different sizes of elastic electrodes as a function of the electrode diameter.

To determine whether recorded spikes are from fast-conducting A-fiber afferents or slow-conducting C-fiber ones, we also measured the conduction velocity by L4 surface DRG recordings of action potentials evoked by electrical stimulation at the mouse hind paw.^[46] We successfully recorded components of compound action potentials evoked by electrical stimulation from the same mouse and observed clear conduction velocity cutoffs between Aβ and Aδ fibers (Figure 5d). We then isolated distinct components of action potentials by discerning their conduction velocities from the gathered data. Our analysis revealed the detection of Aβ and Aδ fibers (Figure 5h and i), although not C-fibers, possibly because the intensity of our stimulus was insufficient to trigger their response.^[47] Consistent with the outcomes of the single unit analysis, we found a similar trend regarding the amplitude of Aδ components as a function of the electrode site area (Figure 5j).

3. Conclusion

In summary, this research presents an innovative and scalable approach towards thin-film-fabricated elastic neuroelectronics through multifunctional nanomesh. The multifunctional nanomesh approach integrates and synergizes the nanomesh property as an elastic platform and key functionalities from individual constituting materials, where Parylene-C serves as a mechanical buffer on PDMS, Au provides electrical conduction, and PEDOT:PSS enhances electrochemical interface, all in the form of thin films. Enabled by this nanomesh strategy, the cellular resolutions and high densities of elastic microelectrode array are unprecedented, and their excellent performance demonstrates the scalability of the approach. Leveraging conventional neuroelectrode materials and thin-film fabrication, the multifunctional nanomesh approach is also highly reliable, as vindicated by cyclic stretching tests and rigorous accelerated soak tests. The properties of the multifunctional nanomesh are highly tunable by varying each constituent thin film’s thickness, enabling versatile materials design for different applications.

Moreover, the ability to record single-unit action potentials from the epidural DRG surface highlights the potential of nanomesh elastic neuroelectronics for minimally invasive neural interfacing applications. Specifically, our study establishes the robust relationship between the electrode size and the single-unit recording performance from the DRG surface, highlighting the significance of cellular-scale electrodes. We also showed that epidural DRG recordings were able to record concurrently from a variety of afferent neurons that are anatomically spread through the entire DRG with functionally distinct features, i.e., nociceptors that can sensitize to noxious mechanical stimuli and non-nociceptive mechanoreceptors that rapidly or slowly adapt to the stepped mechanical forces. Further, high-density nanomesh elastic neuroelectronics demonstrated superior conformability and recording capabilities with self-conformed and revealed unprecedented spatiotemporal dynamics of the neuronal activities from the epidural DRG surface. Overall, the multifunctional nanomesh approach establishes a promising avenue in bioelectronics, elastic neuroelectronics, with outstanding scalability and demonstrates the potential of effectively interfacing with soft and curvilinear electrically active cells and tissues. By addressing challenges at the fundamental neural interface, nanomesh elastic neuroelectronics hold promise for advancing our understanding of population neural dynamics in sensory neural encoding. This novel minimally invasive capability to record high-resolution neuronal spiking from a population of afferents will likely open a new avenue of research for developing next-generation prostheses and closed-loop neural stimulators that incorporate spatiotemporally rich sensory information. Beyond neuroelectronics, the concepts and technical approaches here broadly apply to elastic-electronics scaling with widespread applications.

4. Experimental Section/Methods

4.1. Thin-film Fabrication of Multifunctional Nanomesh on PDMS

First, we coated a 4:1 PDMS film (thickness ~15 μm) on a glass slide as a handling substrate that was pre-spin-coated with a dextran layer (thickness ~30 μm) which serves as a sacrificial layer during the final neuroelectrode arrays release. Then, a 300 nm-thick Parylene-C film was deposited via chemical vapor deposition using a Parylene deposition system (PDS 2010). This Parylene-C film prevents micro-cracking in the subsequent film throughout the subsequent micro-fabrication process and acts as a stress-balancing layer between the PDMS substrate and the rest of the bilayer nanomesh. An electron-beam (e-beam) evaporator then deposited Ti (3 nm)/SiO₂ (20 nm) as an etch stopper for the latter reactive-ion etching (RIE) step. To achieve the Au nanomesh, we leveraged the In grain boundary lithography method^[20]. Grain boundary lithography employs a bilayer lift-off metallization procedure, presenting benefits in resolution, removal, process simplicity, undercut control, and yield compared to traditional lift-off process. In this work, the bilayer consists of Ti (3 nm)/SiO₂ (20 nm)/In (210 nm) mask layer and a Polymethyl methacrylate (PMMA) sacrificial layer for undercut formation. We used PMMA as the sacrificial layer under In grains. The first SiO₂ layer is an etch stop, which prevents the Parylene-C stress-balancing layer during PMMA dry etching. The second SiO₂ layer is known to be important for the wetting of In grains to form suitable grain shapes on PMMA. The In film forms micro-sized grains with irregular structures. These grains can function as a lift-off mask, enabling the creation of stretchable nanomesh based on the boundaries of the In grains. After the In deposition, the utilization of diluted nitric acid (HNO₃) facilitated the removal of nanoscale In particles and concurrently enabled the controlled widening of the gap between neighboring In grains. RIE process then etched SiO₂/PMMA/SiO₂ layer with In as the hard mask, leaving also lateral undercuts facilitating the lift-off. After e-beam evaporation deposited Ti (5 nm)/Au (60 nm)/Ni (30 nm), lift-off was performed in acetone with a gentle swabbing. Ti acts as an adhesion layer between the Parylene-C and Au films, and Ni works as an etch stop layer during Parylene-C etching. Consecutive steps involved RIE for SiO₂ and Parylene-C removal, while wet etching of Ni was performed using a 20% solution of iron chloride (FeCl₃). The Ni was eliminated by immersing the sample for 30 seconds, followed by a rinse with deionized water to eliminate both Ni and any remaining residues, resulting in the creation of Parylene-C/Au bilayer nanomesh on a PDMS film. PEDOT:PSS (Sigma-Aldrich) was electrochemically deposited with a current density of 0.2 mA/cm², and deposition times varied from 0 to 60 s on the Au nanomesh layer to form Parylene-C/Au/PEDOT:PSS nanomesh on PDMS film. Initially, O₂ plasma treatment was conducted to eliminate any residue on the mesh surface. Subsequently, PEDOT:PSS was electrodeposited onto the Au nanomesh template via electropolymerization of a blend comprising EDOT (Sigma-Aldrich) monomer and NaPSS (Sigma-Aldrich) powder. For the preparation, NaPSS powder (0.1 mol/liter) was introduced into 150ml of DI water and agitated for 30 minutes. Then, EDOT (0.001 mol/liter) monomer was added and stirred for an additional hour before utilization. The resulting solution was employed for electroplating using a standard three-electrode setup. An Ag/AgCl electrode served as the reference electrode, while a platinum wire was employed as the counter electrode. Ultimately, PEDOT:PSS was electrochemically deposited at a current density of 0.2 mA/cm², with deposition times ranging from 0 to 60 seconds on the Au nanomesh layer, yielding the formation of Parylene-C/Au/PEDOT:PSS nanomesh on PDMS film.

4.2. Fabrication of Nanomesh Elastic Electrodes and Arrays

The fabrication process began with e-beam evaporation of a 30-nm-thick layer of Ni onto the multifunctional nanomesh, followed by photolithography steps to define electrode and interconnect patterns using positive Photoresist (PR), S1805. In this process, Ni acts as a sacrificial layer, facilitating the patterning process and exhibiting superior adhesion to the PR compared to PEDOT:PSS. The unpatterned region was subjected to ion milling to generate complete patterns, subsequently accompanied by a 30-second oxygen plasma treatment utilizing RIE. This treatment was employed to facilitate the easy removal of the thermally cross-linked PR. To achieve patterned PDMS encapsulation layer, the process began by defining 20-μm-thick negative NR9–8000 PR posts on arrays of electrodes and contact pads. Following this, the entire pattern underwent exposure to Ni etchant to etch the exposed Ni layer. Furthermore, a 5-μm-thick PDMS film mixture with a 10:1 ratio was applied utilizing the spin-coating technique onto a sample that had been previously subjected to a 10-second oxygen plasma treatment via RIE. This treatment significantly enhanced the adhesion of the PDMS encapsulation. As a result of the subsequent lift-off of these posts in acetone after curing PDMS at 70 °C hotplate for 30 minutes, openings corresponding to the electrode and pad arrays were formed within the PDMS encapsulation layer. Finally, Ni etchant removed the Ni mask on the electrode and pad areas by soaking the sample for 10 min in a vacuum chamber. In order to detach the device from the handling glass substrate, we spin-coated a supporting layerk of 50 μm-thick SU-8 first. Following this step, the sacrificial dextran layer was dissolved in water. After the device was separated from the glass, half-mil Kapton film with Ti/ (3 nm)/SiO₂ (20 nm) layer was partially laminated on the interconnects and contact pad array using Si/SiO₂ bonding, and then the SU-8 was removed using PG remover. Kapton film was applied to reinforce the rigidity of the contact pad area and to establish alignment and stable connection with the matrix connector.

4.3. Finite Element Analysis

The 3D finite element model of multifunctional nanomesh was established in ABAQUS. The representative three layers, as in Figure 1e, were composed of PDMS (5 μm), Parylene-C (100–500 nm), and Au (60 nm) in which the grain boundary nanomesh patterns extracted from SEM were applied for only the Parylene-C and Au films (Figure S4). The 8-node linear brick, reduced integration element C3D8R was used for all layers. The perfect adhesion was assumed between layers with tie constraints in the model. The deformation of Au nanomesh was predicted when a 5% applied strain loaded on one side of the PDMS substrate as the boundary condition for a uniaxial stretching test when the other side of the PDMS was fixed. The effect of Parylene-C as a stress relief layer was investigated by tuning its thickness. The according strain distribution and effective modulus were calculated in Figure S2 and S6. The materials properties (Elastic modulus and Poisson’s ratio) used in the simulations are E_PDMS=1.6 MPa, ν=0.45; E_Parylene-C=2.7 GPa, ν=0.4; and E_Au=79 GPa, ν=0.42. To perform the FEA simulation under large strain, the Neo Hooke hyperelastic constitutive model was used for the PDMS layer, with C₁₀=250 MPa, D₁=0.24 GPa^-1. The thickness of the Parylene-C and PDMS was 300 nm and 5 μm respectively. The boundary conditions are the same as used for Figure 1(e).

The electrical properties of nanomesh were predicted by simulations in COMSOL (AC/DC module). The Au nanomesh was simulated as a conductive shell, with one boundary applied an electrical potential (1.5V) and the other grounded. The surface current density and sheet resistance as a function of Au thickness were predicated as in Figure 2c and d. Importantly, the Au nanomesh offers multiple pathways for electric current, as predicted in Figure S4c. This feature is advantageous in maintaining conductivity even when local fractures or cracks occur. The electric conductivity and relative permittivity used in the simulation for Au are 4.5×10⁷ S⁄m and 6.9.

The strain distribution of the 256-channel pad matrix was predicted through ABAQUS simulations. The four representative layers consisted of an acrylic top plate with four holes at the corner (deformable, C3D8R), device pad array matrix with cushion layer (deformable, C3D8R), and solder ball matrix integrated PCB layer (rigid), as shown in Figure S20. The simulation sequence to mimic real cases consisted of two steps. First, we established an initial aligned contact between the top board with the device layer and the PCB board. In the second step, the top board was pressed downwards with 50 μm as if tightening the screws. The encastre boundary condition was applied for the solder ball part in both steps, and the motion of the top plate was limited in plane by applying the displacement constrains (${\mathrm{U}}_{\mathrm{X}}={\mathrm{U}}_{\mathrm{Y}}=0$) on the four holes to mimic the fixation effect of screws. The corresponding strain distribution on the device is shown in Figure 3f and Figure S20. The material properties (Elastic modulus and Poisson’s ratio) used in the simulations are E_Acrylic=3.3 Gpa, v=0.37; E_device=2.5 Gpa, v=0.34.

4.4. Assemble Nanomesh Elastic Array Utilizing Matrix Connector

In the initial stage, the nanomesh elastic bioelectronics were positioned in alignment on a polycarbonate top plate with an intermediate layer of silicone rubber cushion. The application of pressure, facilitated by four force screws, was employed to establish a low-resistance connection between the meshed array pads and the solder balls located on the PCB connector (Figure 3e and f). Afterward, the assembled device was connected to an RHS stimulation/recording system (Intan Technologies), which was employed for measuring the electrode impedance at 1 kHz, recording bench signals to evaluate device performance, and conducting in vivo recordings.

4.5. Surgical Procedure for Anesthetized Recording

The University of Connecticut Animal Care and Use Committee approved all animal procedures, and all experiments were performed in accordance with relevant guidelines and regulations. Experiments were carried out in a surgery room. C57BL/6 mice of both sexes (8–16 weeks of age, 25–35 g) were laid down in a prone position and deeply anesthetized with 2 % Isoflurane inhalation throughout the surgery procedure. The surgical plane of anesthesia was verified by a lack of toe pinch reflex. After shaving the hair at the surgical site and sterilizing the skin with 70 % ethanol, an incision was made in the dorsal skin at the L3-L5 level of the spine. Under a stereomicroscope, muscles, and ligaments covering the dorsal side of the L4 vertebrae were dissected away. Using a high-speed micro-drill, mouse L4 pedicles were removed by grinding the bone structure until a very thin layer of pedicle bone was left. A fine-tipped tweezer was used to remove the last layer of pedicle bone to expose the L4 DRGs. Afterward, the lamina and the spinous process were removed for easier implantation of the recording electrode array. Sterile surgical swabs were used to control bleeding, and the exposed DRGs and spinal cord were irrigated with sterile artificial cerebrospinal fluid (aCSF) to keep the neural tissues from drying. After the surgery procedure, the isoflurane concentration was reduced to 0.5–1 %. Then, urethane was administrated through intraperitoneal injection (0.6 g/kg, Sigma Aldrich) to preserve the integrity of neural transmission and autonomic reflexes for recording. Right before the implantation of the electrode array, the aCSF, blood, and any connective tissue on top of the DRG was carefully removed, leaving an intact DRG enclosed by dura. This will not only ensure easier implantation with improved adhesion between the electrode array and DRG but also reduce the possibility of short-circuiting caused by the conductive liquid within the space between the electrode array and DRG. Throughout the surgical procedure and the following recording process, the animal’s body temperature was maintained at ~37 °C through an integrated temperature control heating pad.

4.6. Electrophysiology

The nanomesh elastic array was connected to the 128-ch stim/recording system (Intan Technology) for electrophysiology recording. The sampling rate of the recording system was 30 ksps. To increase the signal-to-noise ratio, we inserted a platinum wire into the muscle adjacent to the recording site as a reference for the electrode. We first performed mechanical stimulation to evoke neural response from somatosensory neurons through pinching each toe and glabrous paw area, which corresponds to the dermatomes of L4 DRG in mice. To synchronize mechanical stimulation and electrophysiological recordings, we generated a pulse train sequence of 5-second 5 V as a mechanical stimulation cue and 15-second 0 V as a non-stimulation cue and applied it as an analog input to the Intan system. Following the above analog signal, mechanical pinching was periodically delivered to the animal’s toe and paw for 2 minutes in the aforementioned sequence of 5 seconds of pinching and 15 seconds of resting using a grounded tweezer to minimize noise. After that, we delivered electrical stimulation to the glabrous paw area through platinum-iridium wire to electrically evoke neural response. Electrical stimulation was delivered at 0.5 Hz using charge-balanced biphasic stimuli (1–5 mA, cathodic first) with a pulse width of 100 μs and a ramp-up and ramp-down time of 100 μs for each phase. We estimated the conduction path from the paw to the L4 DRG to be 3 cm to extract conduction velocity. To allow the desensitized sensory endings recover, 15 minutes were given between consecutive stimulation.

4.7. Data Analysis

The electrophysiology data was processed with custom Python codes. Data were filtered to 1–500 Hz band for LFP analysis and 300–6000 Hz band for spike analysis. The spike sorting was performed using MountainSort.^[48] We curated the sorted spike clusters that satisfy the following criteria simultaneously as well-isolated single unit clusters: Firing rate > 0.1 Hz; Inter-spike-interval (ISI) violation (< 2ms) < 1%; Full-width half maximum (FWHM) of the mean waveform between 0.15–0.75 ms; Peak-to-valley time (PVT) of the mean waveform between 0.15–0.85 ms; L-ratio < 0.1; Median spike SNR > 4. All single-unit analyses used the curated spike clusters as described above.

Data Availability Statement

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