A soft multimodal optoelectronic array interface for multiparametric mapping of heart function in vivo

Abstract

Multiparametric investigation of cardiac physiology is crucial for the diagnosis and therapy of heart disease. However, no method exists to simultaneously map multiple parameters that govern cardiac (patho)physiology from beating hearts in vivo. Here, we present a cardiac sensing platform that addresses this challenge, functioning with a wireless interface. Advanced fabrication and assembling strategies enable the heterogeneous integration of transparent microelectrodes, light-emitting diodes, photodiodes, and optical filters into a multilayer array structure on soft substrates. The microelectrodes exhibit superior electrochemical performance for measuring electrical potentials and excellent transparency for co-localized fluorescence measurement. The device shows excellent biocompatibility and records the fluorescence of calcium reporter with performance comparable to imaging cameras. Multiparametric in vivo mapping of electrical excitation, calcium dynamics, and their combined effects on cardiac excitation-contraction coupling is demonstrated during normal rhythm, arrhythmia, and treatment. This technology offers potential widespread use in cardiac research to support scientific discoveries and advance clinical life-saving diagnostics and therapies.

Soft multimodal cardiac device allows for electrical and optical mapping of cardiac physiology in vivo.

INTRODUCTION

Heart disease is among the most challenging human diseases to diagnose and treat, and it remains the leading cause of mortality worldwide (1, 2). Its complex underlying pathophysiology has driven the development of advanced techniques to study heart function during sinus rhythm, dysfunction, and treatment (3–9). Soft microelectrode arrays (MEAs) have revolutionized cardiac research by enabling the monitoring of physiological properties and abnormal states with a high temporal resolution from electrically and mechanically active cardiomyocytes (10–12). Nevertheless, MEAs alone do not provide information on other key biophysical parameters involved in cardiac function, such as calcium, a key messenger of the excitation-contraction coupling, which triggers muscle contraction. Abnormal calcium cycling is also responsible for various cardiac pathologies: ventricular arrhythmias, atrial fibrillation, heart failure, etc. (13, 14). During heart function, membrane depolarization triggers intracellular calcium releases from the sarcoplasmic reticulum that drives heart contraction. In turn, calcium modulates local electrical potential and its propagation via multiple calcium-dependent ionic currents (e.g., L-type calcium and sodium-calcium exchanger currents). Those parameters work in concert and cannot be fully understood in isolation (Fig. 1A) (15–18). Optical mapping complements MEA approaches and records these parameters using appropriate fluorescent reporters (19, 20). Recent advances in genetically encoded fluorescent reporters enable optical mapping from specific cardiac cell types, proteins, or parameters (21, 22). Therefore, optical mapping has greatly contributed to our understanding of heart disease mechanisms [e.g., role of rotors in atrial fibrillation (23) and regulation of cardiac arrhythmias (24)]. However, in vivo cardiac optical mapping remains a grand challenge and lags behind current motionless organ systems due to problems associated with persistent motion artifacts from heartbeat. The toxicity of small-molecule fluorescent dyes and the spectral overlap of existing genetically encoded voltage and calcium reporters further prevent in vivo multiparametric cardiac measurements using optical mapping alone. In addition, the temporal resolution of cardiac optical mapping is limited by the inherent kinetics of the fluorescent reporters (20). As a result, there is growing interest in combining electrical and optical approaches for multiparametric studying cardiac physiology and disease conditions from the same myocardial sites, particularly in vivo, to verify experimental hypotheses, generate important insights into arrhythmogenic drivers (e.g., sequential events in excitation-contraction coupling and the distinct process that links cardiac excitation and contraction), and evaluate heart disease treatments while overcoming the limitations of each modality.

Fig. 1. Soft multimodal optoelectronic array device for multiparametric electrical and optical mapping of cardiac physiology.

(A) Schematic illustration of pathological changes in calcium handling, electrical excitation, and excitation-contraction coupling during heart disease as compared to normal sinus rhythm. (B) Schematic exploded-view illustration of a device integrating a 2 × 2 array of Au-Ag NW transparent microelectrodes on top of a 2 × 2 array of μ-LEDs, μ-PDs, and optical filters. The microelectrodes record cardiac electrical excitation while the μ-LEDs and μ-PDs work together to excite and record calcium fluorescence from GCaMP6f (excitation/emission maxima at 462/520 nm). (C) Optical image of a fully integrated device with four blue μ-LEDs on. (D) SEM image of the percolative Au-Ag NW networks in the microelectrodes. (E) EDS elemental mapping images of Au and Ag in the Au-Ag NW, and a merged image confirming the core-shell structure.

To date, state-of-the-art soft MEAs for cardiac electrophysiology predominantly rely on opaque microelectrodes that impede co-localized optical mapping experiments because they generate optical shadows that obstruct the delivery of excitation light to and fluorescence collection from the microelectrode sites (25–28). These opaque microelectrodes also produce severe light-induced electrical artifacts that scale with optical irradiance (29, 30). In recent years, our group and others have developed various types of transparent microelectrode materials to surmount these challenges (31, 32), such as carbon nanotube (CNT) (33), graphene (34–36), metal nanostructures (37–41), and poly(3,4-ethylenedioxythiophene):poly(styrene sulfonate) (PEDOT:PSS) (42). Nonetheless, all now available transparent MEAs are coupled with fluorescence microscopy for optical mapping studies. Fluorescence microscopy setups use externally positioned light sources coupled with high-speed cameras and objective lenses. While they are suitable for certain applications in motion-arrested ex vivo heart preparations, they are incompatible with in vivo experiments on beating hearts with intact circulation and nervous systems. Recently, optoelectronic devices that integrate microscale light-emitting diodes (μ-LEDs), microscale photodetectors (μ-PDs), and optical filters enable single-channel calcium fluorescence measurement of neural dynamics in the brain, eliminating the need for external imaging setups (43, 44). Therefore, we hypothesize that an emerging multimodal optoelectronic cardiac mapping platform could (i) contain and align electrical (such as transparent microelectrodes) and fluorescence (such as μ-LEDs, μ-PDs, and optical filters) recording channels for co-localized monitoring of electrophysiological parameters, calcium transients, and their interactions to study excitation-contraction coupling in hearts; (ii) have a programmable multiplexed array configuration to probe the spatiotemporal patterns of cardiac activity; (iii) demonstrate sufficient mechanical flexibility to enable an intimate contact with the curvilinear heart surfaces. Despite the increasing demand, developing such a platform has not been feasible due to substantial technical challenges in microfabrication and device integration required to combine multiple functionalities into a soft array form factor.

To close this technological gap, here, we present a multimodal, multichannel, soft, optoelectronic cardiac mapping device that contains a 2 × 2 array of (i) blue μ-LED–based optical emitters to excite one of the most popular genetically encoded calcium indicators, GCaMP6f, which shows reduced toxicity than conventional small-molecule fluorescent dyes used in ex vivo cardiac preparations and is more suited for in vivo experiments; (ii) μ-PDs with optical filters to spectrally reject the excitation light background and selectively record green calcium fluorescence signals from GCaMP6f labeled cardiomyocytes; (iii) gold-coated silver (Au-Ag) nanowire (NW) transparent microelectrodes for electrophysiological mapping. The microelectrodes demonstrate a low 1-kHz electrochemical impedance of 7.0 ± 0.14 kilohms and an excellent optical transmittance exceeding 80% in the 400- to 800-nm range for co-localized cross-talk–free electrophysiological and optical measurements. The μ-LED– and μ-PD–based optical sensing subsystem features a high wavelength selectivity >1400 for mapping GCaMP6f fluorescence signals. The device exhibits a mechanical compliance with stable sensing performance under cyclic mechanical bending against a 5-mm radius for 4000 cycles. Benchtop characterizations show that the array platform is fully compatible with Bluetooth-based wireless communication systems for wireless control and acquisition of electrical and calcium fluorescence signals. Ion leaching measurement and histological analysis confirm the biocompatibility of the device. The device successfully achieves co-localized multiparametric mapping of the electrophysiological properties, calcium dynamics, and the excitation-contraction coupling from the beating hearts during sinus rhythm, arrhythmia, and electrotherapy in vivo. These capabilities could provide valuable physiological insights to decode heart function and arrhythmogenesis in a way not possible using existing technologies. Together, the soft multimodal optoelectronic array platform will create exciting opportunities for a broad spectrum of cardiac research, such as heart disease modeling, cell-type–specific electrophysiology, and drug screening.

RESULTS

Design and fabrication of the soft multimodal optoelectronic array device

Figure 1B presents a schematic illustration of the soft multimodal optoelectronic array device. The device consists of a transparent polyethylene terephthalate (PET) substrate, transparent SU-8 planarization/insulation/encapsulation layers, blue μ-LEDs, μ-PDs with optical filters to excite and record GCaMP6f fluorescence, and transparent Au-Ag NW microelectrodes for electrophysiological recording. Our previous Monte Carlo simulation results suggest that the major calcium fluorescence measurement volume is at the interface between the μ-LED and μ-PD (43). As a result, one μ-LED stands next to a μ-PD, forming an efficient optical recording pair, while each microelectrode sits above this optical recording pair with their centers well aligned. This design enables the recording and correlation of electrical and optical signals from the same site. Previous work shows that the irradiance from blue μ-LEDs attenuates exponentially to ~10% of the initial value after traveling 0.5 mm on the epicardial surfaces (45). Thus, we use 0.5 mm as the separation between two nearest μ-LEDs to avoid optical cross-talk between different channels. The resulting pitch values (center to center distance between the two nearest recording channels) also match well with the space constant of cardiac tissue to map the propagation of electrical signals. The overall active device dimension (~1.5 mm by 1.5 mm) is smaller than the ventricle sizes of small animals (46). Device fabrication begins with laminating an ultrathin PET substrate on a glass handling carrier. A transparent SU-8 coating serves as the planarization layer on PET. Electron beam evaporation with a shadow mask deposits the bilayer chromium/copper interconnects for the μ-LEDs and μ-PDs. Micro-transfer printing and a reflow soldering process place the 2 × 2 μ-LED and μ-PD array onto the desired locations on the interconnects. A photolithographically defined thin narrowband absorber serves as an on-chip optical filter on top of each μ-PD in the array. A SU-8 layer fully insulates the optical mapping subsystem. Afterward, photolithographic patterning creates the 2 × 2 Ag NW transparent MEA and interconnects on SU-8. Another SU-8 layer defines the MEA openings and encapsulates the device. Last, electroplating a thin Au outer layer on the unencapsulated Ag NW microelectrode recording areas completes the fabrication. Details of the device fabrication process appear in Materials and Methods. It is noted that, although this work presents a 2 × 2 array, the channel count exceeds that of most state-of-the-art single-modality optical sensing implants containing both μ-LEDs and μ-PDs, which typically include only one to two channels (43, 44, 47–49).

Figure 1C presents an optical image of a fully integrated device with four blue μ-LEDs turned on. The circuit connection design of the sensing components is in fig. S1. The blue μ-LEDs exhibit emission maxima at 462 nm (fig. S2), which is suitable to excite GCaMP6f (50). It is apparent that the MEA and SU-8 layers on top of the optical mapping subsystem allow blue photons from the underneath μ-LEDs to pass through and illuminate the areas above the device. The scanning electron microscopy (SEM) image in Fig. 1D reveals a percolative network of Au-Ag NWs in the microelectrodes, showing a fine distribution of the NWs. X-ray energy-dispersive spectroscopy (EDS) elemental mapping results in Fig. 1E and fig. S3, and transmission electron microscopy (TEM) image in fig. S4 illustrate the core-shell structure of the Au-Ag NW with a uniform Au coating (thickness, 10 ± 0.3 nm). The NWs have a diameter of ~110 nm and a length of ~10 to 20 μm.

Characterizations of the soft multimodal optoelectronic array device

The optical and electrochemical properties of the microelectrodes are tunable via adjusting the thickness of the Au shell. Figure 2A displays the optical transmission spectra of the percolative NW networks, where the average transmittance values in the visible spectrum (between 400 and 800 nm) decrease from 85 ± 2.0% to 81 ± 2.0%, and 76 ± 2.2% with 0-nm (Ag NW), 10-nm (Au-Ag NW-10), and 18-nm (Au-Ag NW-18) Au shell layers on Ag NW, respectively. The average deposition density of Ag NW, Au-Ag NW-10, and Au-Ag NW-18 is 51.3 ± 3.7, 92.9 ± 7.5, and 128 ± 11 μg/cm², respectively. Electrochemical impedance spectroscopy (EIS) evaluates the electrochemical characteristics of microelectrodes. Figure 2B presents the electrochemical impedance curves of the microelectrodes (single microelectrode dimension, 250 μm by 250 μm) with various Au shell thicknesses in phosphate-buffered saline (PBS). As expected, the 1-kHz electrochemical impedance values decrease from 28 to 6.9 and 3.7 kilohms for Ag NW, Au-Ag NW-10, and Au-Ag NW-18 microelectrodes, respectively. The improved electrochemical performance with increased Au shell thickness is attributed to the increased surface area, enhanced conductivity, and superior electrochemical performance of Au compared to Ag (40, 51, 52). The average sheet resistance of Ag NW, Au-Ag NW-10, and Au-Ag NW-18 networks decreases from 14.7 ± 1.7, to 4.54 ± 0.16, and 1.13 ± 0.04 ohms per square, respectively. The Au coating shields the Ag NW from leaching, oxidation, and other adverse interactions with the biofluids. To test the leaching of Ag ions, inductively coupled plasma mass spectrometry (ICP-MS) analyzes the dissolved Ag ions after soaking Ag NW and Au-Ag NW-10 samples in PBS at 37°C for 3 days. Compared to pristine Ag NW, the Au shell reduces Ag ion leaching by 95%, from 314 ± 16 to 14.7 ± 6.0 parts per billion (fig. S5). Oxygen plasma treatment further evaluates the stability of the Au-Ag NW-10. The resistance of Au-Ag NW-10 remains unchanged, while the resistance of the pristine Ag NW increases substantially due to oxidation (fig. S6). Figure 2C compares the normalized 1-kHz electrochemical impedance of the Au-Ag NW-10 microelectrodes to state-of-the-art transparent microelectrodes with comparable 550-nm optical transparency (80 to 85%) for electrophysiological studies, including indium tin oxide (53), CNT (33), Au nanogrid (37), PEDOT:PSS (42), and graphene (34). The electrochemical performance of the Au-Ag NW-10 microelectrodes ranks among the most competitive ones for transparent microelectrodes reported to date. Because of their balanced electrochemical and optical properties, Au-Ag NW-10 microelectrodes are used in subsequent experiments. The EIS results of Au-Ag NW-10 microelectrodes are fit to an equivalent circuit model, as shown in Fig. 2D (inset), to gain deeper insights into the interfacial properties between the microelectrodes and the electrolyte. The circuit model comprises solution resistance (R_S), parallel constant phase element (C_PE), charge transfer resistance (R_CT), and Warburg element for diffusion (Z_W) (40). C_PE is defined by $\frac{1}{Y_0{(j\mathrm{\omega })}^n}$, where Y₀ is the C_PE constant, j is the imaginary unit, ω is the angular frequency, and n is the exponent with values between 0 and 1. When n = 1, the C_PE represents an ideal capacitor, while, when n = 0, it represents a pure resistor. The fitting results align well with the experimental EIS results. The n value (0.9) for the Au-Ag NW-10 microelectrodes (table S1) indicates a highly desired double-layer capacitive interface for recording electrical field potentials. The phase responses confirm the Au-Ag NW-10 microelectrodes are more capacitive at physiologically relevant low frequencies (1 Hz to 1 kHz) and become more resistive at higher frequencies (fig. S7). Figure 2E displays the impedance characteristics of the four microelectrodes in a 2 × 2 Au-Ag NW-10 MEA, showing an average 1-kHz impedance value of 7.0 ± 0.14 kilohms. The inset impedance color map illustrates the spatial distribution of the 1-kHz impedance, providing a visualization of the uniform performance. The superior electrochemical performance of the Au-Ag NW microelectrodes, along with the robust fabrication process, enables both the upscaling and downscaling of microelectrode sizes for high-quality electrophysiological recordings as needed. For example, the 1-kHz electrochemical impedance of Au-Ag NW-10 microelectrodes with dimensions at 50 μm by 50 μm (comparable to the size of a single cardiomyocyte) and 750 μm by 750 μm is 167 and 0.689 kilohms, respectively (fig. S8).

Fig. 2. Optical, electrochemical, and optoelectrical characteristics of the soft multimodal optoelectronic array device.

Transmission (A) and impedance (B) spectra of the Au-Ag NW microelectrodes with different Au shell thicknesses at 0, 10, and 18 nm, respectively. (C) Normalized electrochemical impedance versus transmittance comparison of Au-Ag NW-10 microelectrodes and literature transparent microelectrodes for electrophysiological studies. (D) Equivalent circuit model fitting of the Au-Ag NW-10 microelectrodes. Inset: The equivalent circuit model used. (E) Impedance spectra of the four Au-Ag NW-10 microelectrodes in a MEA. Inset: 1-kHz impedance color map of the microelectrodes with respect to their actual positions in the MEA. (F) Current-voltage characteristics of the blue μ-LED array. (G) Transmission spectra of the optical filters with and without Au-Ag NW-10 microelectrodes on top. (H) Average EQE spectra from the four μ-PDs in the device with and without the optical filters and Au-Ag NW-10 microelectrodes on top. (I) Average short circuit current (I_sc) from the four μ-PDs in the device under varying green irradiances.

Figure 2F and fig. S9 show the current-voltage characteristics of the four μ-LEDs and a single μ-LED in the device. The irradiance from the excitation μ-LEDs typically exceeds the fluorescence signals by several orders of magnitude. We previously designed effective optical filters for detecting GCaMP6f fluorescence using epoxy-mixed absorber dyes (43, 44, 54). We adopt this approach here, and Fig. 2G presents the transmission spectra of the optimized optical filter with and without Au-Ag NW-10 microelectrodes on top. The optical filter incorporating Au-Ag NW-10 microelectrodes exhibits a moderate 520 nm (emission peak of GCaMP6f) transmittance value of 30% to allow green fluorescence to pass and blocks the blue μ-LED excitation light from reaching to μ-PDs. Figure 2H shows the external quantum efficiency (EQE) results from the μ-PDs. Benefiting from the optical filter and unique spectral responses of the μ-PDs, the average EQE ratio at 520 to 462 nm from all four μ-PDs in the device is 1.44 × 10³. The high wavelength selectivity is crucial for high-fidelity green fluorescence measurement. Furthermore, in our device design, the μ-LEDs emit the excitation light from the same plane as the μ-PDs. This is beneficial for the background rejection because the μ-LED light needs to be backscattered 180° to make it back into the μ-PD. Figure 2I presents the near-precise linear response of the μ-PD photocurrent to typical fluorescence irradiances of green fluorescent reporters, demonstrating the μ-PDs are suitable for green fluorescence measurements.

The Au-Ag NW-10 microelectrodes record programmed 10- and 6-Hz sine wave inputs with peak-to-peak amplitudes from 1 to 10 mV in PBS without any noticeable decrease in signal amplitudes (Fig. 3A and fig. S10). This indicates that the microelectrodes could facilitate high-fidelity measurement of cardiac field potentials, which typically exhibit amplitudes within a few millivolts and frequencies of several hertz. The power spectral density (PSD) results in fig. S11 offer detailed frequency-domain insights into the recorded signals. The large peak in the PSD curve is from the input signal. The average signal-to-noise ratio (SNR) from the recording results in Fig. 3A is 33.4 ± 0.35 dB. In parallel, fluorescence signals reflecting intracellular calcium levels in cardiomyocytes at rest (~0.1 μM) and during stimulation (10 to 15 μM) (55) could be reliably captured by the device with a linear response using Oregon Green 488 BAPTA-1 (emission maximum at 522 nm when excited at 493 nm) as the calcium indicator (Fig. 3B). These results illustrate that the overall changes in calcium fluorescence associated with heart function are within the dynamic range of the device. Mechanical flexibility is another crucial feature for cardiac devices, as it allows for good contact with curvilinear heart surfaces, minimizes tissue damage, and reduces stress at the device-tissue interface. The device maintains stable 1-kHz electrochemical impedance (Fig. 3C) and calcium fluorescence sensing performance of 1 μM calcium in PBS (Fig. 3D) over 4000 bending cycles. The bending radius (5 mm) is comparable to the anatomical features of interest in mouse hearts. Those results indicate the mechanical flexibility of the device. Tissue damage from overheating by μ-LEDs is a concern for biological applications. Figure 3 (E and F) summarizes the maximum temperature increases on the surface of the device in air at different duty cycles, frequencies, and irradiances of μ-LED light pulses relevant to cardiac optical mapping. The temperature increase is <0.6°C for duty cycles ≤25% when the four blue μ-LEDs in the device operate at 0.5, 5, and 50 Hz with output irradiances of 5, 15, and 25 mW/mm², respectively. This falls well below the acceptable temperature increase (<2°C) for cardiac experiments and ensures the thermal safety of the device (56). It is noteworthy that biofluids show higher thermal conductivities than air. Therefore, the risk of thermally induced tissue damage during device operation would be further mitigated during ex vivo and in vivo experiments.

Fig. 3. Benchtop measurements of the soft multimodal optoelectronic array device.

(A) Electrical recording output of a programmed 10 Hz, 1 mV peak-to-peak (1 mV_P-P) amplitude sine wave input by the Au-Ag NW-10 microelectrode in the device. (B) Fluorescence measurement results from Oregon Green 488 BAPTA-1 calcium dye solutions using the device. Calcium concentrations range from 0.1 to 15 μM. a.u., arbitrary units. One-kilohertz electrochemical impedance from the Au-Ag NW-10 microelectrodes (C) and calcium fluorescence signals recorded by the μ-PD and μ-LED pairs (D) as a function of bending cycles at a radius of 5 mm. Z and I_ph are the impedance and fluorescence (calcium concentration at 1 μM) values at a specific bending cycle, and Z₀ and I_ph0 represent the initial impedance and calcium fluorescence values, respectively. Changes in temperature from the device versus μ-LED duty cycles in air at (E) 0.5, 5, and 50 Hz and (F) output irradiance at 5, 15, and 25 mW/mm², respectively.

Wireless compatibility is a key feature for the chronic in vivo operation of cardiac implants. We developed a battery-powered dual-sided wireless data acquisition system to record electrical and calcium fluorescence signals from the device. The block diagram of circuit function appears in fig. S12. The wireless flexible printed circuit board (fPCB) incorporates several subsystems (fig. S13A), including electrical connection to the device, front-end analog circuits with different designs for μ-PDs and microelectrodes, analog-to-digital conversion (ADC) circuit, microcontroller unit (MCU) with Bluetooth module for wireless control, and streaming the data to computers. The MCU programs the μ-LED driver circuit to adjust the irradiance and operating frequency of the μ-LED array through pulse-width modulation. Voltage regulator circuits generate specific voltage values from an on-board battery to power different subsystems. The microelectrodes wirelessly record programmed 10-Hz, 1-mV (fig. S13B), 2-mV (fig. S13C), and 10-mV (fig. S13D) peak-to-peak amplitude sine wave inputs with no changes in signal amplitude and morphology. The average SNR from the 1-mV peak-to-peak amplitude sine wave recording signals is 32.6 ± 0.80 dB. Meanwhile, fig. S13E demonstrates the wireless and linear fluorescence capture from Oregon Green 488 BAPTA-1 calcium indicator solutions with calcium levels ranging from 0.1 to 15 μM. The multimodal wireless sensing performance is comparable to those from the measurements in Fig. 3 (A and B). A flexible, miniaturized version of the wireless electronics system is now under development, with physiological experiments on small animals to be demonstrated in future work.

Ex vivo validation of multiparametric cardiac physiological mapping

Experiments on ex vivo Langendorff-perfused GCaMP6f expressing mouse hearts validate the multiparametric mapping capability of the device. The hearts are also stained with potentiometric dye Di-4-ANBDQBS (excitation/emission maxima at 620/722 nm), for ex vivo recording of optical action potential (OAP). The device’s mechanical flexibility allows it to conform and attach to the epicardial surface of the left ventricle (LV). An external optical mapping system serves as the reference, as illustrated in Fig. 4A. Figure 4 (B and C) demonstrates the representative time-aligned electrogram, OAP, and GCaMP6f fluorescence signals recorded by the device and a complementary metal-oxide-semiconductor (CMOS) imaging camera during pacing (voltage, 5 V; pulse width, 3 ms; and frequency, 6 Hz). In both cases, the external optical mapping data are from the closest locations next to the device channels. The electrogram and calcium fluorescence recorded by the device show strong correlations and similar shapes compared to those from CMOS mapping. Slight discrepancies are due to the recording location mismatch and different sensing mechanisms. The microelectrodes collect cardiac conduction properties from the epicardium, while the fluorescence signals reveal the average activity of cardiomyocytes to a depth of about 0.5 mm within the cardiac tissue (4). The action potential duration at 80% repolarization (APD₈₀) and the calcium transient duration at 80% reuptake (CaTD₈₀) are extracted from the electrograms, OAPs, and calcium reporter GCaMP6f fluorescence signals. APD₈₀ is defined as the time interval between activation and repolarization. Similarly, CaTD₈₀ is defined as the time interval between calcium release and its reuptake into the sarcoplasmic reticulum. The APD₈₀ and CaTD₈₀ results measured using the device and CMOS imaging camera show no statistically significant difference (Fig. 4D). The similarity in calcium rise time and calcium decay constant (time constant τ of an exponential fit) between the waveforms measured by the device and those captured by the CMOS imaging camera further confirms the high-fidelity cardiac calcium fluorescence recording performance of the device (Fig. 4E and fig. S14). Rise time is defined as the duration required for depolarization to progress from 20 to 80% of the signal upstroke and τ measures the decay rate of the calcium transients. Both parameters serve as time constants to characterize the calcium responses.

Fig. 4. Ex vivo demonstration of multiparametric electrical and optical mapping of cardiac physiology.

(A) Illustration of the device on the LV of a GCaMP6f expressing mouse heart with an external optical mapping system for side-by-side validation. (B) Comparison of the device electrical signals (red) and OAPs (orange) from CMOS imaging camera in one cycle. (C) Comparison of the calcium fluorescence signals recorded by the device (blue) and CMOS imaging camera (olive) in one cycle. Statistical analysis of APD₈₀ and CaTD₈₀ (D) and calcium rise time (E) calculated from the electrograms, OAPs, calcium fluorescence signals measured by the device and CMOS imaging camera. Data represent means ± SD. A two-sample t test comparison between device (n = 5) and CMOS (n = 5) demonstrates no statistically significant difference [not significant (n.s.), P > 0.5]. (F) Representative traces of simultaneously recorded ECG signals (black) from needle electrodes, and electrograms (red) and calcium fluorescence signals (blue) from all channels in the device on the surface of LV. The time points of pacing are indicated by the dashed lines on each plot. Electrogram (left) and calcium (right) activation maps from the device (G) and CMOS (H) recording results during pacing. (I) Voltage-calcium delay statistics from the device mapping results at varying PCLs.

After validating the device’s efficacy to accurately detect electrophysiological and calcium parameters, its capability for multichannel co-localized cardiac mapping is demonstrated. Figure 4F illustrates the time-aligned far-field electrocardiogram (ECG) signals (black) from needle electrodes, electrograms (red), and calcium fluorescence signals (blue) mapped by the four device channels in each modality during pacing. No observable light-induced electrical artifacts occur in the electrograms during optical mapping. The ECG signals, electrograms, and calcium transients are highly correlated. During membrane depolarization, inward calcium channels open, allowing calcium ions to flow into the cell. These ions subsequently stimulate the opening of intracellular calcium-release channels (ryanodine receptors) in the sarcoplasmic reticulum membrane, leading to a substantial release of calcium into the cytosol and the activation of muscle excitation-contraction coupling. This calcium-induced calcium release process introduces an activation delay between membrane depolarization and the calcium transient (voltage-calcium delay), which is essential for synchronizing cardiac contraction and determining the tight excitation-contraction coupling. Figure 4G displays the synchronized electrical (left) and optical calcium (right) activation maps, derived from the depolarization (activation) times of the recorded electrograms and calcium signals by the device. The results match well with the activation maps derived from CMOS imaging camera recording results (Fig. 4H). The comparison between the electrical and calcium activation maps clearly visualizes the spatiotemporal distribution of the voltage-calcium delays. The calculated average voltage-calcium delay values from device and CMOS mapping results are −9.7 ± 1.5 ms and −8.8 ± 2.2 ms, respectively. Figure 4I presents the average voltage-calcium delay values extracted from the device mapping results at different pacing cycle lengths (PCLs). The prolonged delay from 75- to 100-ms PCL indicates the uncoupling between electrical excitation and contraction. Further increasing the PCL results in shortened delays, possibly due to a slower depolarization (57). Overall, these results highlight the accuracy and capability of the soft multimodal optoelectronic array device for co-localized electrical and optical mapping of electrical excitation, calcium transients, and the excitation-contraction coupling to monitor cardiac physiology.

In vivo mapping of cardiac electrophysiology, calcium dynamics, and excitation-contraction coupling

In vivo studies, compared to ex vivo experiments, provide a better understanding of the interactions between blood flow, autonomic nervous system, electrical conduction, calcium transients, and mechanical contraction during normal and pathological function of the heart. Following ex vivo validations, systematic in vivo open-chest studies are conducted to demonstrate the device’s ability and advantages for detailed multiparametric mapping of the dynamic cardiac physiology from the blood-perfused freely beating hearts, a challenge for state-of-the-art cardiac optical mapping systems. Figure 5A presents a schematic illustration of the in vivo open-chest test setup, where a device attaches to the LV of a mouse heart as a temporary implant. Figure 5B shows the simultaneously recorded electrical (red) and calcium fluorescence (blue) signals during electrical pacing (voltage, 5 V; pulse width, 3 ms; frequency, 6 Hz) by the four device channels in each modality. Figure 5C displays electrical and calcium activation maps from the device recording results during pacing. It is evident that the device captures the spatiotemporal propagation of cardiac activation, moving from the bottom left of the device area to the top and lastly to the bottom right. The electrical and calcium wave propagation patterns during pacing differ from those during atrioventricular (AV) block, a common arrhythmia (fig. S15). This is because the ventricular activation originates from the pacing site during electrical pacing, while, during AV block, ventricular activation initiates from the His-Purkinje system. Figure 5D demonstrates the spatiotemporal voltage (top) and calcium (bottom) maps at five sequential time points of pacing, highlighting the sensing of anisotropic cardiac excitation and calcium wave propagations on the LV surface. Figure 5 (E to G) compares the APD₈₀, CaTD₈₀, and voltage-calcium delay values from the device mapping results during sinus rhythm, pacing, and second-degree AV block, to elucidate changes in electrical excitation, calcium dynamics, and excitation-contraction coupling under different cardiac conditions in vivo. Compared to second-degree AV block, pacing greatly increases APD₈₀, from 14.2 ± 2.9 ms to 29.5 ± 5.2 ms, reduces CaTD₈₀, from 158 ± 20.1 ms to 72.0 ± 1.4 ms, and decreases the voltage-calcium delay, from −36.7 ± 5.0 ms to −8.9 ± 2.3 ms, respectively. Meanwhile, the APD₈₀, CaTD₈₀, and voltage-calcium delay values show no statistically significant difference between sinus rhythm and pacing. Those results suggest that AV block shortens the duration of the cardiac cycle, prolongs the time interval between calcium release and reuptake, and induces a much-increased degree of uncoupling of muscle excitation-contraction compared to sinus rhythm and pacing. In addition, the device captures increased CaTD₈₀ values as the heart transitions from second-degree AV block to more severe third-degree AV block (fig. S16), suggesting that calcium flux dynamics vary with disease stage and progression. These findings demonstrate the device’s capability to capture calcium transients, electrophysiological properties, and excitation-contraction coupling from the in vivo dynamically beating hearts at different conditions.

Fig. 5. In vivo demonstration of multiparametric electrical and optical mapping of cardiac physiology.

(A) Illustration of the in vivo open-chest test setup. (B) Simultaneously recorded electrograms and calcium fluorescence signals from the device during external pacing with a platinum electrode. The time points of pacing are indicated by the dashed lines on each plot. (C) Activation maps from the electrograms and calcium fluorescence signals in (B). (D) Spatiotemporal voltage and calcium maps at five sequential time points from the simultaneously recorded electrograms and calcium fluorescence signals. APD₈₀ (E), CaTD₈₀ (F), and voltage-calcium delay (G) values from the co-localized electrical and optical mapping results during external pacing, sinus rhythm (SR), and second-degree AV block. Data represent means ± SD, n = 5. One-way analysis of variance (ANOVA) in conjunction with a two-sample t test is performed. n.s., not significant, **P < 0.01, ***P < 0.001, and ****P < 0.0001.

To evaluate the in vivo biocompatibility, the devices are implanted subcutaneously in the dorsal paramedial region of GCaMP6f mice. Ag ion leaching tests with blood serum and histological evaluations of the tissues investigate the biocompatibility at different time points (1 and 3 weeks) after device implantation. No statistically significant changes in the Ag ion concentrations exist in the blood serum samples among the control, device, and sham groups, respectively (fig. S17A), suggesting the robustness of device encapsulation and Au coating. Masson’s trichrome (fig. S17B) and hematoxylin and eosin (H&E) (fig. S17C) staining after 1 and 3 weeks of implantation show no obvious adverse reactions compared to control and sham groups, such as myocyte volume (fig. S17D). Together, these results suggest that the device is fully biocompatible.

Multiparametric mapping of cardiac physiology during pharmacological interventions

To demonstrate the ability of the device to perform a detailed assessment of pharmacological interventions, 1 μM verapamil is administered to the heart to observe changes in cardiac electrophysiological properties, calcium dynamics, and excitation-contraction coupling. Verapamil is a well-studied L-type calcium channel blocker affecting calcium handling (57, 58). Figure 6 (A and B) shows the representative traces of electrograms and calcium fluorescence signals during pacing (6 Hz) with and without administering 1 μM verapamil. Specifically, verapamil treatment does not notably alter APD₈₀, CaTD₈₀, and τ (Fig. 6, C and D). This indicates that verapamil does not induce obvious changes in the repolarization and calcium reuptake phase of the electrograms and calcium signals. In addition, no statistically significant changes in the upstroke rise time of electrograms occur with verapamil treatment (Fig. 6E). Meanwhile, a prolonged upstroke rise time of calcium transients appears in verapamil-treated mouse hearts (Fig. 6E), indicating slowed calcium entry from the extracellular space via calcium channels and slowed release from the sarcoplasmic reticulum, as expected with the L-type calcium channel blocker. Minimal changes in the voltage-calcium delay occur after verapamil treatment (Fig. 6F), suggesting that verapamil does not alter excitation-contraction coupling. Similar verapamil effects are evident at a higher dosage (5 μM, fig. S18) and at other physiologically relevant pacing frequencies (fig. S19). Verapamil induces a dose-dependent increase in the rise time of calcium transients, suggesting further decreased calcium release at higher doses. The effects of pinacidil, a nonselective adenosine triphosphate–sensitive potassium channel opener, on cardiac physiology are also assessed at dosages of 20 and 100 μM (fig. S20). APD₈₀ and CaTD₈₀ shorten with pinacidil treatment at both doses, although the shortening of CaTD₈₀ is less pronounced than APD₈₀. In addition, a higher dose of pinacidil leads to greater reductions in APD₈₀ and CaTD₈₀, along with an increase in voltage-calcium delay. The observed changes in calcium handling and electrical excitation properties are consistent with known cardiac responses to pinacidil (59). Together, those multiparametric electrical and optical mapping measurements highlight the use of the device for drug safety and efficacy screening in vivo.

Fig. 6. Multiparametric mapping of cardiac physiology in drug testing using L-type calcium channel blocker verapamil.

Representative traces of electrograms (A) along with calcium fluorescence signals (B) simultaneously recorded by the device before and after treatment with 1 μM verapamil. Statistical analysis of APD₈₀ and CaTD₈₀ (C), τ (D), action potential (AP) and calcium rise time (E), and voltage-calcium delay (F) values from the simultaneously recorded electrograms and calcium fluorescence signals by the device with and without 1 μM verapamil during pacing at 6 Hz, 3-ms pulse width, and 5 V. Data represent means ± SD, n = 5. A two-sample t test is used for comparison. n.s., not significant, ****P < 0.0001.

DISCUSSION

In summary, here, we demonstrate a soft multimodal optoelectronic array platform that combines Au-Ag NW transparent microelectrodes, μ-LEDs, μ-PDs, and optical filters on a flexible substrate for cross-talk–free, multiparametric, co-localized mapping of cardiac electrophysiological properties, calcium dynamics, and excitation-contraction coupling. The microelectrodes offer low electrochemical impedance and high optical transparency for high-fidelity cardiac electrophysiological recording. The optical mapping subsystem exhibits a superior selectivity for sensing calcium fluorescence. The integrated device displays excellent mechanical flexibility to achieve good contact with the heart. Future developments could lead to fully stretchable devices for seamless in vivo integration with the beating hearts over time. Benchtop characterizations reveal that the soft multimodal cardiac mapping platform is compatible with Bluetooth wireless operation electronics, highlighting a promising avenue for future development as an untethered chronic cardiac implant for analysis of the heart in freely behaving animals over extended periods of time. Ex vivo validations confirm that the device demonstrates a comparable sensing performance to state-of-the-art external optical mapping systems. Systematic in vivo demonstrations highlight the device’s unique strengths in simultaneously investigating cardiac electrical excitation, calcium signaling, and their combined effects on excitation-contraction coupling in blood-perfused beating hearts during normal rhythm, heart disease, and pacing treatment. The demonstrated parameters and metrics measured from the device mapping results include calcium and voltage activation maps, sequential maps, APD₈₀, CaTD₈₀, calcium and voltage rise times, calcium decay constant, and voltage-calcium delay. The voltage-calcium delay provides crucial insights into the excitation-contraction coupling during heart function and is extremely challenging to investigate in vivo using existing tools. We show that the device can capture important information on cardiac physiological modulation upon administration of therapeutic drugs. Similar to arrhythmias, simultaneous but distinct changes in electrical potential and calcium handling occur during heart failure. The extent of these changes varies depending on the specific type of arrhythmia (e.g., focal versus reentrant) and heart failure (e.g., ischemic versus nonischemic) (18). Future studies will use this multiparametric sensing platform, along with its design and integration strategies, to conduct basic studies of heart function and disease mechanisms, as well as translational research aimed at developing advanced diagnostics and targeted treatments for various heart diseases. The current device features four electrical mapping and four calcium fluorescence mapping channels as a proof-of-concept demonstration. An important direction for future improvements is to increase the channel counts (e.g., to a few hundred) for high-resolution spatial mapping of cardiac signal propagations in broader areas. This will allow for a more comprehensive capture of both regional and global heart activity in large animal models and, eventually, in patients. The optical profiles of the μ-LEDs and μ-PDs could be tuned to target other fluorescent reporters by adjusting their emission and detection wavelengths. For example, a future version of the device may include additional optical excitation and sensing elements in the ultraviolet and blue wavelength regions to record cardiac metabolism. This would enable a comprehensive investigation of the metabolism-excitation-contraction coupling in the heart. In addition, future versions of this technology might incorporate strain sensors (60, 61) to directly monitor mechanical signals together with electrical excitation and calcium dynamics, enabling a more comprehensive investigation of cardiac physiology, or include electrical pacing capability for closed-loop feedback control of cardiac function and delivering real-time clinically relevant cardiac interventions.

MATERIALS AND METHODS

Fabrication of the soft multimodal optoelectronic array

Device fabrication began with the lamination of a flexible PET film (25 μm thick, CS Hyde Company) on a glass handling carrier with polydimethylsiloxane (Sylgard 184, Dow Corning) adhesive. Spin coating a 7-μm photodefinable epoxy (SU-8 2007, MicroChem) planarized the PET film. Electron beam evaporation with a custom stainless steel shadow mask (Micron Laser Technology Inc.) defined the double-layer chromium/copper interconnects and contact pads (thickness, 1.1 μm) for the μ-LED and μ-PD arrays. We used copper as the interconnect material due to its low cost, which reduces fabrication expenses, and its excellent thermal conductivity, which effectively dissipates heat generated by the μ-LEDs during device operation. A flip-chip soldering process using Indalloy 4 (Indium Corporation) and SAC305 (Chip Quik Inc.) solder pastes joined μ-PDs (L × W × H: 300 μm by 300 μm by 100 μm, TCE12-589, Three Five Materials Inc.) and μ-LEDs (L × W × H: 270 μm by 220 μm by 50 μm, C460TR2227-0216, Cree Inc.) on the interconnects. A narrowband molecular absorber (ABS 473, Exciton) was mixed into SU-8 (7 μm, 2.5% weight percentage) and patterned on the μ-PDs as optical filters via photolithography. A 100-μm-thick SU-8 encapsulated the optoelectronic components and created a smooth surface for patterning NW microelectrodes on top. Another 2-μm SU-8 adhesive layer was spin coated and exposed with a dose of 55 mJ/cm² to define MEA and its interconnect patterns, followed by spin coating a 100 μl of Ag NW solution (ACS Material) on SU-8, post-exposure bake at 95°C for 4 min, development in the SU-8 developer solution for 2 min, and a hard bake at 110°C for 5 min. The sheet resistance of the interconnects remained relatively stable with the line width varied between 100 and 250 μm (fig. S21). Consequently, the line width of the interconnects was set at 100 μm to achieve miniaturization. A 7-μm-thick SU-8 layer defined the MEA windows (single microelectrode dimension, 250 μm by 250 μm) and bonding pads via photolithography. Layers of Au with varying thicknesses were electroplated galvanostatically onto the surfaces of Ag NW microelectrodes using a sulfite-based solution (pH 6.0 to 7.0, TSG-250, Transene) diluted with water at a 1:8 volume ratio (40). The microelectrodes remained in the solution under stirring for 1 hour at 60°C before electroplating. A potentiostat (Reference 600+, Gamry Instruments Inc.) applied a current density of 0.1 mA/cm² for varying lengths of time to control the thickness of Au. After deposition, the devices were rinsed with water and dried to remove any excess solution. Laser cutting (LPKF ProtoLaser U4) defined the device geometry, followed by delamination from the glass handling carrier to complete the fabrication.

Wireless circuit design and assembly

The fPCB was designed using EAGLE (Autodesk Inc.) and manufactured off-site by a PCB manufacturing company (PCBWay). A framed stencil and screen printer (FP2636, Neoden USA) deposited solder paste (EP256, Kester) before the placement of circuit components. The assembled boards were soldered in a reflow oven. Through-hole components were soldered manually.

Optical, electrical, and optoelectrical measurements

A spectrophotometer (V-770 UV-visible/NIR, Jasco Inc.) characterized the transmittance spectra of the NW microelectrodes and optical filters. A Keithley 2614B source meter measured the current-voltage characteristics of the μ-LEDs and μ-PDs. A quantum efficiency system (QE-RS, Enlitech) measured the EQE spectra of the μ-PDs. An externally mounted green LED source (M530L3, ThorLabs Inc.) determined the responsivity of the μ-PDs. A spectrometer (ILT560, International Light) measured the irradiance values of the light source and μ-LEDs. A four-point probe (SRM-232, Guardian Manufacturing Inc.) determined the sheet resistance of the NWs. An oxygen plasma cleaner (PDC-001-HP, Harrick Plasma) performed the oxidation treatment of the NWs.

EIS measurements

A potentiostat (Reference 600+, Gamry Instruments Inc.) conducted the EIS measurements of the microelectrodes at a frequency range from 1 Hz to 100 kHz via a three-electrode configuration in PBS (Sigma-Aldrich) within a Faraday Shield (fig. S22). Here, the microelectrodes served as the working electrode, an Ag/AgCl electrode served as the reference electrode, and a platinum electrode served as the counter electrode, respectively.

Benchtop and morphological measurements

For benchtop electrical signal recording, a platinum electrode inputs 1- to 10-mV peak-to-peak amplitude sine waves at 6 and 10 Hz in PBS from a PowerLab data acquisition system (ADInstruments Inc.). For benchtop fluorescence recording measurements, the μ-PD responses were recorded with excitation by the μ-LEDs at different calcium levels in Oregon Green 488 BAPTA-1 calcium dye solutions. A motorized test stand (ESM 1500, Mark-10) tested the mechanical flexibility of the device at a 5-mm bending radius. Thermal imaging with an infrared camera (E8-XT, FLIR) monitored temperature changes of the device in air at room temperature. SEM (PIONEER EBL, Raith Inc.), high-resolution scanning TEM (Talos F200X G2 TEM, Thermo Fisher Scientific Inc.), and EDS examined the Au-Ag NW morphology, Au thickness, and interface properties. A Thermo Fisher Scientific iCAP Q quadrupole ICP-MS measured the Ag ion concentrations in the ion leaching tests.

Animal experiments

All animal procedures (62) followed protocols approved by Northwestern University’s Institutional Animal Care and Use Committee and coincided with National Institutes of Health guidelines.

Mouse model

The JAX stock strains #031968 and #011038 were cross bred for cardiomyocyte specific expression of GCaMP6f, a genetically encoded calcium indicator. Adult mice were randomly sampled and between 10 and 16 weeks of age.

Ex vivo electrical and optical mapping experiments

Figure S23 illustrates the setup for ex vivo investigations. Mice were anesthetized by inhalation of 5% isoflurane at an oxygen flow of 2 ml/min in an induction chamber (E-Z Systems Inc.). Deep anesthetic induction was verified by the absence of response to a toe pinch, confirming the cessation of pain. Cervical dislocation was performed to euthanize mice before the excision of the heart. We performed a transdiaphragmatic thoracotomy and separated the rib cage to create a large opening with a clear view of the heart and vasculature. A supra-cardiac transection of the great vessels was performed to remove and transfer the heart to be cannulated (via 20-gauge blunt tip) through the aorta for perfusion and exsanguination. The heart was then attached to a Langendorff-perfusion of a modified Tyrode’s solution [140 mM NaCl, 4.7 mM KCl, 1.05 mM MgCl₂, 1.3 mM CaCl₂, 10 mM Hepes, and 11.1 mM glucose (pH 7.4) at 37°C] oxygenated with 100% O₂. Hydrostatic pressure was maintained at 60 to 80 mm Hg through modifying the perfusion flow rate. Far-field needle electrodes (MLA1203, ADInstruments Inc.) monitored ECG in a three-electrode configuration. Blebbistatin (Cayman Chemical) was added at 15 μM concentration followed by letting the heart stabilize with electromechanical uncoupling for a minimum of 10 min or until motion arrested. In tandem with loading blebbistatin, the heart was stained with dye Di-4-ANBDQBS [30 μl of dye stock solution (1.66 mg/ml) and 970 μl of Tyrode’s solution] to measure transmembrane voltage. The dye was perfused for at least 10 min prior imaging to enable complete staining of tissue and removal of any excess dye. Multispectral excitation was performed on the Multi-LED Light System LEX9 (SciMedia). LED units (460 and 620 nm) stimulated GCaMP6f and Di-4-ANBDQBS. A 695-nm-long-pass filter and a 520 ± 15–nm band-pass filter isolated the OAP and GCaMP6f calcium fluorescence signals, respectively. A three-dimensionally printed assembly placed the soft device on the LV during the experiment (fig. S23A). A customized platinum bipolar pacing electrode paced the heart from an inferior direction, effectively stimulating the apex. Total length of the ex vivo experimental procedure was executed within 3 hours.

In vivo electrical and optical mapping experiments

Figure S24 presents the workflow for in vivo studies. GCaMP6f mice were anesthetized by a 5% isoflurane inhalation in an induction chamber for 3 min or until unconscious and respiration slows. We lowered the isoflurane to 2 to 3% and transferred the mice to the mouse intubation platform (Kent Scientific) in supine position. After sedation, buprenorphine was administered via subcutaneous injection at a dosage of 0.5 to 1.0 mg/kg for pain relief. The cessation of pain was verified as previously described (63). Mice were intubated [20-gauge 1-inch (2.54 cm) or 22-gauge 1-inch (2.54 cm) dulled safelet catheter] and placed on volumetric controlled ventilation, using the VentElite small animal ventilator (Harvard Apparatus). The ventilator was set to 180 breaths per min using volume-controlled ventilation with a tidal volume of 10 ml/kg, or peak inspiratory pressure of 14 to 20 cmH₂O, maximum pressure of 30 cmH₂O, and positive end-expiratory pressure of 2 to 5 cmH₂O. ECG leads were connected for cardiac monitoring in Lead I configuration. Body temperature was maintained at 37°C. The left lateral chest region was shaved, and Nair (hair removal cream) was applied for 1 min to remove hair cleanly. To view the base of the LV in mice, an incision was made between the fourth and fifth intercostal spaces laterally, left of the midline. This approach provided direct access and a clear view of the base of the LV. To verify the point of entry for the incision within the intercostal space, palpitations were used to feel the area with maximum pulsation. Scissors were used to make one clean cut through the skin and then retracted via Braintree Scientific’s magnetic animal surgery board. The muscles were torn across the long axis of the muscle between the ribs, and the magnetic retractors were adjusted to hold open the muscle, ribs, and skin. A cotton swab was used to gently retract the lungs. The device was slowly inserted into the thoracic cavity, placed gently on the left lateral side of the heart for acute measurement. For future chronic in vivo operations, sutures or adhesives (64) could be used for device attachment. Mice were euthanized after experiments via cervical dislocation and heart excision. The total length of the in vivo procedure was executed within 2 hours.

Biocompatibility tests

n = 3 GCaMP6f mice for each group. Devices were implanted subcutaneously in the dorsal paramedial region. For implantation, an incision was made, and blunt dissection created a subcutaneous pocket for the device. Absorbable sutures secured the devices and closed the incision. Regular checking ensured the health conditions of the mice. Independent collection of blood and tissue samples happened at 0 (control), 1, and 3 weeks after implantation. In addition, a sham group was evaluated at 1 week as an incision-only control. Dermal samples (1 cm by 1 cm) were collected, fixed in 4% paraformaldehyde for 24 hours, and stored in PBS. The tissues were then paraffin-embedded, sectioned, and stained with H&E and Masson’s trichrome. Blood samples were collected through exsanguination and centrifuged in BD Microtainer tubes at 10,000 times gravity.

Data acquisition and processing

Electrical and optical signals were recorded from the μ-PDs and MEA with the Intan RHS stimulation/recording system using two RHS headstages: one for the μ-PD array channels and another for the MEA channels. Both signals were recorded at a 20-kHz sampling rate with a low-pass filter at 2.5 kHz and high-pass filter at 0.78 Hz. All data were post-processed using custom MATLAB programs to apply a 100-Hz low-pass filter, a 60-Hz notch filter, or a 1-Hz high-pass filter for analysis. Optical traces by CMOS imaging cameras were analyzed using custom MATLAB software, Rhythm 3.0, which is available as open source on Github.

Statistical analysis

Where applicable, data were represented as means ± SD with three or five samples per group. Comparisons between two groups were performed using a two-sample t test. For comparisons between more than two groups, a one-way analysis of variance (ANOVA) was used, followed by Tukey’s post hoc test to assess significant changes. Differences were considered statistically significant when P < 0.05 (*P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001).

Supplementary Materials

This PDF file includes:

Figs. S1 to S24

Table S1